public inbox for [email protected]
help / color / mirror / Atom feed[PATCH 1/4] demote: setter functions for LocalXLogInsert local variable
22+ messages / 10 participants
[nested] [flat]
* [PATCH 1/4] demote: setter functions for LocalXLogInsert local variable
@ 2020-07-31 08:58 Jehan-Guillaume de Rorthais <[email protected]>
0 siblings, 0 replies; 22+ messages in thread
From: Jehan-Guillaume de Rorthais @ 2020-07-31 08:58 UTC (permalink / raw)
Adds functions extern LocalSetXLogInsertNotAllowed() and
LocalSetXLogInsertCheckRecovery() to set the local variable
LocalXLogInsert respectively to 0 and -1.
These functions are declared as extern for future need in
the demote patch.
Function LocalSetXLogInsertAllowed() already exists and
declared as static as it is not needed outside of xlog.h.
---
src/backend/access/transam/xlog.c | 27 +++++++++++++++++++++++----
src/include/access/xlog.h | 2 ++
2 files changed, 25 insertions(+), 4 deletions(-)
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 09c01ed4ae..c0d79f192c 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -7711,7 +7711,7 @@ StartupXLOG(void)
Insert->fullPageWrites = lastFullPageWrites;
LocalSetXLogInsertAllowed();
UpdateFullPageWrites();
- LocalXLogInsertAllowed = -1;
+ LocalSetXLogInsertCheckRecovery();
if (InRecovery)
{
@@ -8219,6 +8219,25 @@ LocalSetXLogInsertAllowed(void)
InitXLOGAccess();
}
+/*
+ * Make XLogInsertAllowed() return false in the current process only.
+ */
+void
+LocalSetXLogInsertNotAllowed(void)
+{
+ LocalXLogInsertAllowed = 0;
+}
+
+/*
+ * Make XLogInsertCheckRecovery() return false in the current process only.
+ */
+void
+LocalSetXLogInsertCheckRecovery(void)
+{
+ LocalXLogInsertAllowed = -1;
+}
+
+
/*
* Subroutine to try to fetch and validate a prior checkpoint record.
*
@@ -9004,9 +9023,9 @@ CreateCheckPoint(int flags)
if (shutdown)
{
if (flags & CHECKPOINT_END_OF_RECOVERY)
- LocalXLogInsertAllowed = -1; /* return to "check" state */
+ LocalSetXLogInsertCheckRecovery(); /* return to "check" state */
else
- LocalXLogInsertAllowed = 0; /* never again write WAL */
+ LocalSetXLogInsertNotAllowed(); /* never again write WAL */
}
/*
@@ -9159,7 +9178,7 @@ CreateEndOfRecoveryRecord(void)
END_CRIT_SECTION();
- LocalXLogInsertAllowed = -1; /* return to "check" state */
+ LocalSetXLogInsertCheckRecovery(); /* return to "check" state */
}
/*
diff --git a/src/include/access/xlog.h b/src/include/access/xlog.h
index 221af87e71..8c9cadc6da 100644
--- a/src/include/access/xlog.h
+++ b/src/include/access/xlog.h
@@ -306,6 +306,8 @@ extern RecoveryState GetRecoveryState(void);
extern bool HotStandbyActive(void);
extern bool HotStandbyActiveInReplay(void);
extern bool XLogInsertAllowed(void);
+extern void LocalSetXLogInsertNotAllowed(void);
+extern void LocalSetXLogInsertCheckRecovery(void);
extern void GetXLogReceiptTime(TimestampTz *rtime, bool *fromStream);
extern XLogRecPtr GetXLogReplayRecPtr(TimeLineID *replayTLI);
extern XLogRecPtr GetXLogInsertRecPtr(void);
--
2.20.1
--MP_/6xG3U2Dzv0UR=.BCNK7fiL6
Content-Type: text/x-patch
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment;
filename=v5-0002-demote-support-demoting-instance-from-production-to-.patch
^ permalink raw reply [nested|flat] 22+ messages in thread
* SLRUs in the main buffer pool, redux
@ 2022-01-13 13:59 Thomas Munro <[email protected]>
0 siblings, 2 replies; 22+ messages in thread
From: Thomas Munro @ 2022-01-13 13:59 UTC (permalink / raw)
To: pgsql-hackers
Hi,
I was re-reviewing the proposed batch of GUCs for controlling the SLRU
cache sizes[1], and I couldn't resist sketching out $SUBJECT as an
obvious alternative. This patch is highly experimental and full of
unresolved bits and pieces (see below for some), but it passes basic
tests and is enough to start trying the idea out and figuring out
where the real problems lie. The hypothesis here is that CLOG,
multixact, etc data should compete for space with relation data in one
unified buffer pool so you don't have to tune them, and they can
benefit from the better common implementation (mapping, locking,
replacement, bgwriter, checksums, etc and eventually new things like
AIO, TDE, ...).
I know that many people have talked about doing this and maybe they
already have patches along these lines too; I'd love to know what
others imagined differently/better.
In the attached sketch, the SLRU caches are psuedo-relations in
pseudo-database 9. Yeah. That's a straw-man idea stolen from the
Zheap/undo project[2] (I also stole DiscardBuffer() from there);
better ideas for identifying these buffers without making BufferTag
bigger are very welcome. You can list SLRU buffers with:
WITH slru(relfilenode, path) AS (VALUES (0, 'pg_xact'),
(1, 'pg_multixact/offsets'),
(2, 'pg_multixact/members'),
(3, 'pg_subtrans'),
(4, 'pg_serial'),
(5, 'pg_commit_ts'),
(6, 'pg_notify'))
SELECT bufferid, path, relblocknumber, isdirty, usagecount, pinning_backends
FROM pg_buffercache NATURAL JOIN slru
WHERE reldatabase = 9
ORDER BY path, relblocknumber;
Here are some per-cache starter hypotheses about locking that might be
completely wrong and obviously need real analysis and testing.
pg_xact:
I couldn't easily get rid of XactSLRULock, because it didn't just
protect buffers, it's also used to negotiate "group CLOG updates". (I
think it'd be nice to replace that system with an atomic page update
scheme so that concurrent committers stay on CPU, something like [3],
but that's another topic.) I decided to try a model where readers
only have to pin the page (the reads are sub-byte values that we can
read atomically, and you'll see a value as least as fresh as the time
you took the pin, right?), but writers have to take an exclusive
content lock because otherwise they'd clobber each other at byte
level, and because they need to maintain the page LSN consistently.
Writing back is done with a share lock as usual and log flushing can
be done consistently. I also wanted to try avoiding the extra cost of
locking and accessing the buffer mapping table in common cases, so I
use ReadRecentBuffer() for repeat access to the same page (this
applies to the other SLRUs too).
pg_subtrans:
I got rid of SubtransSLRULock because it only protected page contents.
Can be read with only a pin. Exclusive page content lock to write.
pg_multixact:
I got rid of the MultiXact{Offset,Members}SLRULock locks. Can be read
with only a pin. Writers take exclusive page content lock. The
multixact.c module still has its own MultiXactGenLock.
pg_commit_ts:
I got rid of CommitTsSLRULock since it only protected buffers, but
here I had to take shared content locks to read pages, since the
values can't be read atomically. Exclusive content lock to write.
pg_serial:
I could not easily get rid of SerialSLRULock, because it protects the
SLRU + also some variables in serialControl. Shared and exclusive
page content locks.
pg_notify:
I got rid of NotifySLRULock. Shared and exclusive page content locks
are used for reading and writing. The module still has a separate
lock NotifyQueueLock to coordinate queue positions.
Some problems tackled incompletely:
* I needed to disable checksums and in-page LSNs, since SLRU pages
hold raw data with no header. We'd probably eventually want regular
(standard? formatted?) pages (the real work here may be implementing
FPI for SLRUs so that checksums don't break your database on torn
writes). In the meantime, suppressing those things is done by the
kludge of recognising database 9 as raw data, but there should be
something better than this. A separate array of size NBuffer holds
"external" page LSNs, to drive WAL flushing.
* The CLOG SLRU also tracks groups of async commit LSNs in a fixed
sized array. The obvious translation would be very wasteful (an array
big enough for NBuffers * groups per page), but I hope that there is a
better way to do this... in the sketch patch I changed it to use the
single per-page LSN for simplicity (basically group size is 32k
instead of 32...), which is certainly not good enough.
Some stupid problems not tackled yet:
* It holds onto the virtual file descriptor for the last segment
accessed, but there is no invalidation for when segment files are
recycled; that could be fixed with a cycle counter or something like
that.
* It needs to pin buffers during the critical section in commit
processing, but that crashes into the ban on allocating memory while
dealing with resowner.c book-keeping. It's also hard to know how many
buffers you'll need to pin in advance. For now, I just commented out
the assertions...
* While hacking on the pg_stat_slru view I realised that there is
support for "other" SLRUs, presumably for extensions to define their
own. Does anyone actually do that? I, erm, didn't support that in
this sketch (not too hard though, I guess).
* For some reason this is failing on Windows CI, but I haven't looked
into that yet.
Thoughts on the general concept, technical details? Existing patches
for this that are further ahead/better?
[1] https://commitfest.postgresql.org/36/2627/
[2] https://commitfest.postgresql.org/36/3228/
[3] http://www.vldb.org/pvldb/vol13/p3195-kodandaramaih.pdf
Attachments:
[application/gzip] 0001-Move-SLRU-data-into-the-regular-buffer-pool.patch.gz (42.2K, ../../CA+hUKGKAYze99B-jk9NoMp-2BDqAgiRC4oJv+bFxghNgdieq8Q@mail.gmail.com/2-0001-Move-SLRU-data-into-the-regular-buffer-pool.patch.gz)
download
^ permalink raw reply [nested|flat] 22+ messages in thread
* Re: SLRUs in the main buffer pool, redux
@ 2022-01-13 17:46 Robert Haas <[email protected]>
parent: Thomas Munro <[email protected]>
1 sibling, 0 replies; 22+ messages in thread
From: Robert Haas @ 2022-01-13 17:46 UTC (permalink / raw)
To: Thomas Munro <[email protected]>; +Cc: pgsql-hackers
On Thu, Jan 13, 2022 at 9:00 AM Thomas Munro <[email protected]> wrote:
> I was re-reviewing the proposed batch of GUCs for controlling the SLRU
> cache sizes[1], and I couldn't resist sketching out $SUBJECT as an
> obvious alternative. This patch is highly experimental and full of
> unresolved bits and pieces (see below for some), but it passes basic
> tests and is enough to start trying the idea out and figuring out
> where the real problems lie. The hypothesis here is that CLOG,
> multixact, etc data should compete for space with relation data in one
> unified buffer pool so you don't have to tune them, and they can
> benefit from the better common implementation (mapping, locking,
> replacement, bgwriter, checksums, etc and eventually new things like
> AIO, TDE, ...).
I endorse this hypothesis. The performance cliff when the XID range
you're regularly querying exceeds the hardcoded constant is quite
steep, and yet we can't just keep pushing that constant up. Linear
search does not scale well to infinitely large arrays.[citation
needed]
> [ long list of dumpster-fire level problems with the patch ]
Honestly, none of this sounds that bad. I mean, it sounds bad in the
sense that you're going to have to fix all of this somehow and I'm
going to unhelpfully give you no advice whatsoever about how to do
that, but my guess is that a moderate amount of persistence will be
sufficient for you to get the job done. None of it sounds hopeless.
Before fixing all of that, one thing you might want to consider is
whether it uh, works. And by "work" I don't mean "get the right
answer" even though I agree with my esteemed fellow hacker that this
is an important thing to do.[1] What I mean is that it would be good
to see some evidence that the number of buffers that end up being used
to cache any particular SLRU is somewhat sensible, and varies by
workload. For example, consider a pgbench workload. As you increase
the scale factor, the age of the oldest XIDs that you regularly
encounter will also increase, because on the average, the row you're
now updating will not have been updated for a larger number of
transactions. So it would be interesting to know whether all of the
CLOG buffers that are regularly being accessed do in fact remain in
cache - and maybe even whether buffers that stop being regularly
accessed get evicted in the face of cache pressure.
Also, the existing clog code contains a guard that absolutely prevents
the latest CLOG buffer from being evicted. Because - at least in a
pgbench test like the one postulated above, and probably in general -
the frequency of access to older CLOG buffers decays exponentially,
evicting the newest or even the second or third newest CLOG buffer is
really bad. At present, there's a hard-coded guard to prevent the
newest buffer from being evicted, which is a band-aid, but an
effective one. Even with that band-aid, evicting any of the most
recent few can produce a system-wide stall, where every backend ends
up waiting for the evicted buffer to be retrieved. It would be
interesting to know whether the same problem can be recreated with
your patch, because the buffer eviction algorithm for shared buffers
is only a little bit less dumb than the one for SLRUs, and can pretty
commonly devolve into little more than evict-at-random.
Evict-at-random is very bad here, because evicting a hot CLOG page is
probably even worse than evicting, say, a btree root page.
Another interesting test might be one that puts pressure on some other
SLRU, like pg_multixact or pg_subtrans. In general SLRU pages that are
actually being used are hot enough that we should keep them in cache
almost no matter what else is competing for cache space ... but the
number of such pages can be different from one SLRU to another, and
can change over time.
--
Robert Haas
EDB: http://www.enterprisedb.com
[1] http://postgr.es/m/[email protected]
^ permalink raw reply [nested|flat] 22+ messages in thread
* Re: SLRUs in the main buffer pool, redux
@ 2022-01-17 10:23 Heikki Linnakangas <[email protected]>
parent: Thomas Munro <[email protected]>
1 sibling, 1 reply; 22+ messages in thread
From: Heikki Linnakangas @ 2022-01-17 10:23 UTC (permalink / raw)
To: Thomas Munro <[email protected]>; pgsql-hackers
On 13/01/2022 15:59, Thomas Munro wrote:
> Hi,
>
> I was re-reviewing the proposed batch of GUCs for controlling the SLRU
> cache sizes[1], and I couldn't resist sketching out $SUBJECT as an
> obvious alternative. This patch is highly experimental and full of
> unresolved bits and pieces (see below for some), but it passes basic
> tests and is enough to start trying the idea out and figuring out
> where the real problems lie. The hypothesis here is that CLOG,
> multixact, etc data should compete for space with relation data in one
> unified buffer pool so you don't have to tune them, and they can
> benefit from the better common implementation (mapping, locking,
> replacement, bgwriter, checksums, etc and eventually new things like
> AIO, TDE, ...).
+1
> I know that many people have talked about doing this and maybe they
> already have patches along these lines too; I'd love to know what
> others imagined differently/better.
IIRC one issue with this has been performance. When an SLRU is working
well, a cache hit in the SLRU is very cheap. Linearly scanning the SLRU
array is cheap, compared to computing the hash and looking up a buffer
in the buffer cache. Would be good to do some benchmarking of that.
I wanted to do this with the CSN (Commit Sequence Number) work a long
time ago. That would benefit from something like an SLRU with very fast
access, to look up CSNs. Even without CSNs, if the CLOG was very fast to
access we might not need hint bits anymore. I guess trying to make SLRUs
faster than they already are is moving the goalposts, but at a minimum
let's make sure they don't get any slower.
- Heikki
^ permalink raw reply [nested|flat] 22+ messages in thread
* Re: SLRUs in the main buffer pool, redux
@ 2022-01-17 20:05 Thomas Munro <[email protected]>
parent: Heikki Linnakangas <[email protected]>
0 siblings, 1 reply; 22+ messages in thread
From: Thomas Munro @ 2022-01-17 20:05 UTC (permalink / raw)
To: Heikki Linnakangas <[email protected]>; +Cc: pgsql-hackers
On Mon, Jan 17, 2022 at 11:23 PM Heikki Linnakangas <[email protected]> wrote:
> IIRC one issue with this has been performance. When an SLRU is working
> well, a cache hit in the SLRU is very cheap. Linearly scanning the SLRU
> array is cheap, compared to computing the hash and looking up a buffer
> in the buffer cache. Would be good to do some benchmarking of that.
One trick I want to experiment with is trying to avoid the mapping
table lookup by using ReadRecentBuffer(). In the prototype patch I do
that for one buffer per SLRU cache (so that'll often point to the head
CLOG page), but it could be extended to a small array of recently
accessed buffers to scan linearly, much like the current SLRU happy
case except that it's not shared and doesn't need a lock so it's even
happier. I'm half joking here, but that would let us keep calling
this subsystem SLRU :-)
> I wanted to do this with the CSN (Commit Sequence Number) work a long
> time ago. That would benefit from something like an SLRU with very fast
> access, to look up CSNs. Even without CSNs, if the CLOG was very fast to
> access we might not need hint bits anymore. I guess trying to make SLRUs
> faster than they already are is moving the goalposts, but at a minimum
> let's make sure they don't get any slower.
One idea I've toyed with is putting a bitmap into shared memory where
each bit corresponds to a range of (say) 256 xids, accessed purely
with atomics. If a bit is set we know they're all committed, and
otherwise you have to do more pushups. I've attached a quick and
dirty experimental patch along those lines, but I don't think it's
quite right, especially with people waving 64 bit xid patches around.
Perhaps it'd need to be a smaller sliding window, somehow, and perhaps
people will balk at the wild and unsubstantiated assumption that
transactions generally commit.
From ee0db89b7c054ca0297d42ef94058b37bb8b9436 Mon Sep 17 00:00:00 2001
From: Thomas Munro <[email protected]>
Date: Fri, 14 Jan 2022 11:15:48 +1300
Subject: [PATCH] WIP: Try to cache CLOG results.
To make CLOG lookups faster, maintain a bitmap of xid groups that are
known to be all committed, using atomic ops.
XXX Just an experiment...
---
src/backend/access/transam/clog.c | 144 +++++++++++++++++++++++++++++-
1 file changed, 143 insertions(+), 1 deletion(-)
diff --git a/src/backend/access/transam/clog.c b/src/backend/access/transam/clog.c
index de787c3d37..0678fa828c 100644
--- a/src/backend/access/transam/clog.c
+++ b/src/backend/access/transam/clog.c
@@ -41,6 +41,7 @@
#include "miscadmin.h"
#include "pg_trace.h"
#include "pgstat.h"
+#include "port/atomics.h"
#include "storage/proc.h"
#include "storage/sync.h"
@@ -81,6 +82,127 @@
*/
#define THRESHOLD_SUBTRANS_CLOG_OPT 5
+/*
+ * The number of transactions to summarize into a single bit of shared memory
+ * that says 'everything in this group has committed'. Must be a power of two
+ * and <= CLOG_XACTS_PER_PAGE. Currently set so that it corresponds to one
+ * cache line of CLOG page, so that it's cheap to scan a group on every
+ * commit. That results in a 1MB summary array. XXX Many other policies
+ * including lazy computation and course granularity are possible.
+ */
+#define CLOG_XACTS_PER_SUMMARY_GROUP (64 * CLOG_XACTS_PER_BYTE)
+
+/* Fixed size of the array of all-committed summary bits, in bytes. */
+#define CLOG_SUMMARY_SIZE (0x80000000 / CLOG_XACTS_PER_SUMMARY_GROUP / 8)
+
+static pg_atomic_uint32 *CLOGSummary;
+
+/*
+ * Write the all-committed bit for the summary group containing 'xid'.
+ */
+static void
+TransactionXidSetSummaryBit(TransactionId xid, bool value)
+{
+ size_t bit_index = (xid & 0x7fffffff) / CLOG_XACTS_PER_SUMMARY_GROUP;
+ size_t bits_per_word = sizeof(uint32) * 8;
+ size_t word_index = bit_index / bits_per_word;
+ uint32 mask = 1 << (bit_index % bits_per_word);
+
+ if (value)
+ pg_atomic_fetch_or_u32(&CLOGSummary[word_index], mask);
+ else
+ pg_atomic_fetch_and_u32(&CLOGSummary[word_index], ~mask);
+}
+
+/*
+ * Read the all-committed bit for the summary group containing 'xid'.
+ */
+static bool
+TransactionXidGetSummaryBit(TransactionId xid)
+{
+ size_t bit_index = (xid & 0x7fffffff) / CLOG_XACTS_PER_SUMMARY_GROUP;
+ size_t bits_per_word = sizeof(uint32) * 8;
+ size_t word_index = bit_index / bits_per_word;
+ uint32 mask = 1 << (bit_index % bits_per_word);
+
+ return (pg_atomic_read_u32(&CLOGSummary[word_index]) & mask) != 0;
+}
+
+/*
+ * Summarize a cache line's worth of CLOG statuses into an all-committed bit,
+ * if possible. Called for each CLOG commit, so had better be fast.
+ */
+static void
+TransactionXidSummarizeGroup(TransactionId xid, char *page)
+{
+ int first_chunk;
+ int end_chunk;
+ uint64 all_committed_chunk;
+ uint64 *page_chunks = (uint64 *) page;
+
+ /* Already set? */
+ if (unlikely(TransactionXidGetSummaryBit(xid)))
+ return;
+
+ /*
+ * XXX Async commit LSNs should prevent this from happening.
+ */
+
+ /*
+ What does a uint64 full of commits look like? This is computed
+ symbolically here, but should be a constant-folded at compile time.
+ */
+ all_committed_chunk = 0;
+ for (int i = 0; i < CLOG_XACTS_PER_BYTE * sizeof(all_committed_chunk); ++i)
+ {
+ all_committed_chunk <<= CLOG_BITS_PER_XACT;
+ all_committed_chunk |= TRANSACTION_STATUS_COMMITTED;
+ }
+
+ /* Rewind to start of group. */
+ xid &= ~(CLOG_XACTS_PER_SUMMARY_GROUP - 1);
+
+ /* Scan summary group looking for non-committed xacts, chunk at a time. */
+ first_chunk = TransactionIdToByte(xid) / sizeof(all_committed_chunk);
+ end_chunk =
+ first_chunk + (CLOG_XACTS_PER_SUMMARY_GROUP /
+ (CLOG_XACTS_PER_BYTE * sizeof(all_committed_chunk)));
+ for (int i = first_chunk; i < end_chunk; ++i)
+ if (page_chunks[i] != all_committed_chunk)
+ return;
+
+ TransactionXidSetSummaryBit(xid, true);
+}
+
+#if 0
+/*
+ * Summarize a whole page of CLOG statuses into all-committed bits, if
+ * possible. XXX Could be called after loading due to cache miss (?), but
+ * currently this fact is not exposed by slru.c.
+ */
+static void
+TransactionXidSummarizePage(int pageno, char *page)
+{
+ for (TransactionId xid = pageno * CLOG_XACTS_PER_PAGE;
+ xid < (pageno + 1) * CLOG_XACTS_PER_PAGE;
+ xid += CLOG_XACTS_PER_SUMMARY_GROUP)
+ TransactionXidSummarizeGroup(xid, page);
+}
+#endif
+
+/*
+ * Summarize a zero'd page of CLOG statuses into zero all-committed bits.
+ * Called when extending a the CLOG.
+ */
+static void
+TransactionXidSummarizeNewPage(int pageno)
+{
+ for (TransactionId xid = pageno * CLOG_XACTS_PER_PAGE;
+ xid < (pageno + 1) * CLOG_XACTS_PER_PAGE;
+ xid += CLOG_XACTS_PER_SUMMARY_GROUP)
+ TransactionXidSetSummaryBit(xid, false);
+}
+
/*
* Link to shared-memory data structures for CLOG control
*/
@@ -618,6 +740,11 @@ TransactionIdSetStatusBit(TransactionId xid, XidStatus status, XLogRecPtr lsn, i
if (XactCtl->shared->group_lsn[lsnindex] < lsn)
XactCtl->shared->group_lsn[lsnindex] = lsn;
}
+
+ /* Scan a cache line of statuses to see if we can set a summary bit. */
+ /* XXX: Move this up so we do it just once for page update? */
+ if (status == TRANSACTION_STATUS_COMMITTED)
+ TransactionXidSummarizeGroup(xid, XactCtl->shared->page_buffer[slotno]);
}
/*
@@ -646,6 +773,10 @@ TransactionIdGetStatus(TransactionId xid, XLogRecPtr *lsn)
char *byteptr;
XidStatus status;
+ /* See if we can take the lock-free path using a summary bit. */
+ if (TransactionXidGetSummaryBit(xid))
+ return TRANSACTION_STATUS_COMMITTED;
+
/* lock is acquired by SimpleLruReadPage_ReadOnly */
slotno = SimpleLruReadPage_ReadOnly(XactCtl, pageno, xid);
@@ -689,17 +820,26 @@ CLOGShmemBuffers(void)
Size
CLOGShmemSize(void)
{
- return SimpleLruShmemSize(CLOGShmemBuffers(), CLOG_LSNS_PER_PAGE);
+ return SimpleLruShmemSize(CLOGShmemBuffers(), CLOG_LSNS_PER_PAGE) +
+ CLOG_SUMMARY_SIZE;
}
void
CLOGShmemInit(void)
{
+ bool found;
+
XactCtl->PagePrecedes = CLOGPagePrecedes;
SimpleLruInit(XactCtl, "Xact", CLOGShmemBuffers(), CLOG_LSNS_PER_PAGE,
XactSLRULock, "pg_xact", LWTRANCHE_XACT_BUFFER,
SYNC_HANDLER_CLOG);
SlruPagePrecedesUnitTests(XactCtl, CLOG_XACTS_PER_PAGE);
+ CLOGSummary = ShmemInitStruct("pg_xact summary", CLOG_SUMMARY_SIZE, &found);
+ if (!found)
+ {
+ for (int i = 0; i < CLOG_SUMMARY_SIZE / sizeof(pg_atomic_uint32); ++i)
+ pg_atomic_init_u32(&CLOGSummary[i], 0);
+ }
}
/*
@@ -739,6 +879,8 @@ ZeroCLOGPage(int pageno, bool writeXlog)
{
int slotno;
+ TransactionXidSummarizeNewPage(pageno);
+
slotno = SimpleLruZeroPage(XactCtl, pageno);
if (writeXlog)
--
2.30.2
Attachments:
[text/plain] 0001-WIP-Try-to-cache-CLOG-results.patch.txt (6.6K, ../../CA+hUKG+qzHB-mY3kym7JCQ-SbiO88hhiWKkUCfLMQmm7CbaCBw@mail.gmail.com/2-0001-WIP-Try-to-cache-CLOG-results.patch.txt)
download | inline diff:
From ee0db89b7c054ca0297d42ef94058b37bb8b9436 Mon Sep 17 00:00:00 2001
From: Thomas Munro <[email protected]>
Date: Fri, 14 Jan 2022 11:15:48 +1300
Subject: [PATCH] WIP: Try to cache CLOG results.
To make CLOG lookups faster, maintain a bitmap of xid groups that are
known to be all committed, using atomic ops.
XXX Just an experiment...
---
src/backend/access/transam/clog.c | 144 +++++++++++++++++++++++++++++-
1 file changed, 143 insertions(+), 1 deletion(-)
diff --git a/src/backend/access/transam/clog.c b/src/backend/access/transam/clog.c
index de787c3d37..0678fa828c 100644
--- a/src/backend/access/transam/clog.c
+++ b/src/backend/access/transam/clog.c
@@ -41,6 +41,7 @@
#include "miscadmin.h"
#include "pg_trace.h"
#include "pgstat.h"
+#include "port/atomics.h"
#include "storage/proc.h"
#include "storage/sync.h"
@@ -81,6 +82,127 @@
*/
#define THRESHOLD_SUBTRANS_CLOG_OPT 5
+/*
+ * The number of transactions to summarize into a single bit of shared memory
+ * that says 'everything in this group has committed'. Must be a power of two
+ * and <= CLOG_XACTS_PER_PAGE. Currently set so that it corresponds to one
+ * cache line of CLOG page, so that it's cheap to scan a group on every
+ * commit. That results in a 1MB summary array. XXX Many other policies
+ * including lazy computation and course granularity are possible.
+ */
+#define CLOG_XACTS_PER_SUMMARY_GROUP (64 * CLOG_XACTS_PER_BYTE)
+
+/* Fixed size of the array of all-committed summary bits, in bytes. */
+#define CLOG_SUMMARY_SIZE (0x80000000 / CLOG_XACTS_PER_SUMMARY_GROUP / 8)
+
+static pg_atomic_uint32 *CLOGSummary;
+
+/*
+ * Write the all-committed bit for the summary group containing 'xid'.
+ */
+static void
+TransactionXidSetSummaryBit(TransactionId xid, bool value)
+{
+ size_t bit_index = (xid & 0x7fffffff) / CLOG_XACTS_PER_SUMMARY_GROUP;
+ size_t bits_per_word = sizeof(uint32) * 8;
+ size_t word_index = bit_index / bits_per_word;
+ uint32 mask = 1 << (bit_index % bits_per_word);
+
+ if (value)
+ pg_atomic_fetch_or_u32(&CLOGSummary[word_index], mask);
+ else
+ pg_atomic_fetch_and_u32(&CLOGSummary[word_index], ~mask);
+}
+
+/*
+ * Read the all-committed bit for the summary group containing 'xid'.
+ */
+static bool
+TransactionXidGetSummaryBit(TransactionId xid)
+{
+ size_t bit_index = (xid & 0x7fffffff) / CLOG_XACTS_PER_SUMMARY_GROUP;
+ size_t bits_per_word = sizeof(uint32) * 8;
+ size_t word_index = bit_index / bits_per_word;
+ uint32 mask = 1 << (bit_index % bits_per_word);
+
+ return (pg_atomic_read_u32(&CLOGSummary[word_index]) & mask) != 0;
+}
+
+/*
+ * Summarize a cache line's worth of CLOG statuses into an all-committed bit,
+ * if possible. Called for each CLOG commit, so had better be fast.
+ */
+static void
+TransactionXidSummarizeGroup(TransactionId xid, char *page)
+{
+ int first_chunk;
+ int end_chunk;
+ uint64 all_committed_chunk;
+ uint64 *page_chunks = (uint64 *) page;
+
+ /* Already set? */
+ if (unlikely(TransactionXidGetSummaryBit(xid)))
+ return;
+
+ /*
+ * XXX Async commit LSNs should prevent this from happening.
+ */
+
+ /*
+ What does a uint64 full of commits look like? This is computed
+ symbolically here, but should be a constant-folded at compile time.
+ */
+ all_committed_chunk = 0;
+ for (int i = 0; i < CLOG_XACTS_PER_BYTE * sizeof(all_committed_chunk); ++i)
+ {
+ all_committed_chunk <<= CLOG_BITS_PER_XACT;
+ all_committed_chunk |= TRANSACTION_STATUS_COMMITTED;
+ }
+
+ /* Rewind to start of group. */
+ xid &= ~(CLOG_XACTS_PER_SUMMARY_GROUP - 1);
+
+ /* Scan summary group looking for non-committed xacts, chunk at a time. */
+ first_chunk = TransactionIdToByte(xid) / sizeof(all_committed_chunk);
+ end_chunk =
+ first_chunk + (CLOG_XACTS_PER_SUMMARY_GROUP /
+ (CLOG_XACTS_PER_BYTE * sizeof(all_committed_chunk)));
+ for (int i = first_chunk; i < end_chunk; ++i)
+ if (page_chunks[i] != all_committed_chunk)
+ return;
+
+ TransactionXidSetSummaryBit(xid, true);
+}
+
+#if 0
+/*
+ * Summarize a whole page of CLOG statuses into all-committed bits, if
+ * possible. XXX Could be called after loading due to cache miss (?), but
+ * currently this fact is not exposed by slru.c.
+ */
+static void
+TransactionXidSummarizePage(int pageno, char *page)
+{
+ for (TransactionId xid = pageno * CLOG_XACTS_PER_PAGE;
+ xid < (pageno + 1) * CLOG_XACTS_PER_PAGE;
+ xid += CLOG_XACTS_PER_SUMMARY_GROUP)
+ TransactionXidSummarizeGroup(xid, page);
+}
+#endif
+
+/*
+ * Summarize a zero'd page of CLOG statuses into zero all-committed bits.
+ * Called when extending a the CLOG.
+ */
+static void
+TransactionXidSummarizeNewPage(int pageno)
+{
+ for (TransactionId xid = pageno * CLOG_XACTS_PER_PAGE;
+ xid < (pageno + 1) * CLOG_XACTS_PER_PAGE;
+ xid += CLOG_XACTS_PER_SUMMARY_GROUP)
+ TransactionXidSetSummaryBit(xid, false);
+}
+
/*
* Link to shared-memory data structures for CLOG control
*/
@@ -618,6 +740,11 @@ TransactionIdSetStatusBit(TransactionId xid, XidStatus status, XLogRecPtr lsn, i
if (XactCtl->shared->group_lsn[lsnindex] < lsn)
XactCtl->shared->group_lsn[lsnindex] = lsn;
}
+
+ /* Scan a cache line of statuses to see if we can set a summary bit. */
+ /* XXX: Move this up so we do it just once for page update? */
+ if (status == TRANSACTION_STATUS_COMMITTED)
+ TransactionXidSummarizeGroup(xid, XactCtl->shared->page_buffer[slotno]);
}
/*
@@ -646,6 +773,10 @@ TransactionIdGetStatus(TransactionId xid, XLogRecPtr *lsn)
char *byteptr;
XidStatus status;
+ /* See if we can take the lock-free path using a summary bit. */
+ if (TransactionXidGetSummaryBit(xid))
+ return TRANSACTION_STATUS_COMMITTED;
+
/* lock is acquired by SimpleLruReadPage_ReadOnly */
slotno = SimpleLruReadPage_ReadOnly(XactCtl, pageno, xid);
@@ -689,17 +820,26 @@ CLOGShmemBuffers(void)
Size
CLOGShmemSize(void)
{
- return SimpleLruShmemSize(CLOGShmemBuffers(), CLOG_LSNS_PER_PAGE);
+ return SimpleLruShmemSize(CLOGShmemBuffers(), CLOG_LSNS_PER_PAGE) +
+ CLOG_SUMMARY_SIZE;
}
void
CLOGShmemInit(void)
{
+ bool found;
+
XactCtl->PagePrecedes = CLOGPagePrecedes;
SimpleLruInit(XactCtl, "Xact", CLOGShmemBuffers(), CLOG_LSNS_PER_PAGE,
XactSLRULock, "pg_xact", LWTRANCHE_XACT_BUFFER,
SYNC_HANDLER_CLOG);
SlruPagePrecedesUnitTests(XactCtl, CLOG_XACTS_PER_PAGE);
+ CLOGSummary = ShmemInitStruct("pg_xact summary", CLOG_SUMMARY_SIZE, &found);
+ if (!found)
+ {
+ for (int i = 0; i < CLOG_SUMMARY_SIZE / sizeof(pg_atomic_uint32); ++i)
+ pg_atomic_init_u32(&CLOGSummary[i], 0);
+ }
}
/*
@@ -739,6 +879,8 @@ ZeroCLOGPage(int pageno, bool writeXlog)
{
int slotno;
+ TransactionXidSummarizeNewPage(pageno);
+
slotno = SimpleLruZeroPage(XactCtl, pageno);
if (writeXlog)
--
2.30.2
^ permalink raw reply [nested|flat] 22+ messages in thread
* Re: SLRUs in the main buffer pool, redux
@ 2022-05-27 11:24 Thomas Munro <[email protected]>
parent: Thomas Munro <[email protected]>
0 siblings, 3 replies; 22+ messages in thread
From: Thomas Munro @ 2022-05-27 11:24 UTC (permalink / raw)
To: Heikki Linnakangas <[email protected]>; +Cc: pgsql-hackers
Rebased, debugged and fleshed out a tiny bit more, but still with
plenty of TODO notes and questions. I will talk about this idea at
PGCon, so I figured it'd help to have a patch that actually applies,
even if it doesn't work quite right yet. It's quite a large patch but
that's partly because it removes a lot of lines...
Attachments:
[text/x-patch] v2-0001-WIP-Move-SLRU-data-into-the-regular-buffer-pool.patch (170.5K, ../../CA+hUKGKCkbtOutcz5M8Z=0pgAkwdiR57Lxk7803rGsgiBNST6w@mail.gmail.com/2-v2-0001-WIP-Move-SLRU-data-into-the-regular-buffer-pool.patch)
download | inline diff:
From d17b57e8036d51dce5495b3cb68bc42207ed8bb0 Mon Sep 17 00:00:00 2001
From: Thomas Munro <[email protected]>
Date: Sun, 2 Jan 2022 11:39:23 +1300
Subject: [PATCH v2 1/2] WIP: Move SLRU data into the regular buffer pool.
Previously, slru.c managed 7 small buffer pools for special queue-like
data. It had very simple sizing, locking, mapping, replacement and
write back. Switch to the regular buffer pool, to benefit from the
standard infrastructure.
The layout of SLRU files on disk remains unchanged. File access is done
through new callback functions exposed through smgr.c.
Switching SLRU page format to "standard" format (and thus enabling
in-page LSNs and checksums) is left for a later project. For now, add
support for raw (headerless) pages in the buffer pool.
XXX Among the problems not yet addressed:
* SMgrRelation objects cache slru_file descriptors, but there is no
invalidation for when SLRUs wrap around so they might wrap around
* CLOG lookups are a bit slower if you have to do a buffer mapping table
lookup first; to do better than that you need a little cache, which
for now is single item, and could be generalised; maybe CLOG is
actually so hot that it needs summary data that can be access
atomically?
* CLOG LSNs are now one-per-page, which probably isn't good ("group_lsn"
dropped by this patch)
* "external LSNs" are in a new separate array, but could perhaps be
stuffed into BufferDescriptor
* The locking strategies for all SLRUs need to be reviewed
* Support for "extension" SLRUs (ie SLRUs owned by code in core) is
removed -- did anyone ever acutally use that? Code searches seem to
indicate no
* While committing, in a critical section, the buffer pinning code
allocates memory which isn't allowed; I commented that stuff out for
now; perhaps we can make the resowner stuff non-allocating?
* Something is occasionally corrupting memory (locking problem, or
bad pointer arithmetic?)
XXX Work in progress!
XXX Unfinished, experimental!
XXX Contains known bugs!
Discussion: https://postgr.es/m/CA%2BhUKGKAYze99B-jk9NoMp-2BDqAgiRC4oJv%2BbFxghNgdieq8Q%40mail.gmail.com
---
doc/src/sgml/monitoring.sgml | 56 +-
src/backend/access/transam/clog.c | 199 +--
src/backend/access/transam/commit_ts.c | 156 +-
src/backend/access/transam/multixact.c | 358 ++---
src/backend/access/transam/slru.c | 1734 ++++++----------------
src/backend/access/transam/subtrans.c | 108 +-
src/backend/access/transam/xlog.c | 19 +-
src/backend/commands/async.c | 67 +-
src/backend/postmaster/checkpointer.c | 18 +
src/backend/storage/buffer/buf_init.c | 17 +-
src/backend/storage/buffer/bufmgr.c | 151 +-
src/backend/storage/ipc/ipci.c | 4 -
src/backend/storage/lmgr/lwlock.c | 14 -
src/backend/storage/lmgr/lwlocknames.txt | 10 +-
src/backend/storage/lmgr/predicate.c | 62 +-
src/backend/storage/smgr/smgr.c | 20 +-
src/backend/storage/sync/sync.c | 22 +-
src/backend/utils/activity/pgstat_slru.c | 35 +-
src/backend/utils/activity/wait_event.c | 4 +-
src/backend/utils/adt/pgstatfuncs.c | 8 +-
src/backend/utils/mmgr/mcxt.c | 12 +-
src/include/access/clog.h | 6 -
src/include/access/commit_ts.h | 3 -
src/include/access/multixact.h | 3 -
src/include/access/slru.h | 193 +--
src/include/access/subtrans.h | 3 -
src/include/pgstat.h | 1 -
src/include/storage/buf_internals.h | 6 +
src/include/storage/bufmgr.h | 9 +
src/include/storage/lwlock.h | 9 +-
src/include/storage/smgr.h | 8 +
src/include/storage/sync.h | 6 +-
src/include/utils/pgstat_internal.h | 27 +-
src/include/utils/wait_event.h | 2 +-
34 files changed, 1022 insertions(+), 2328 deletions(-)
diff --git a/doc/src/sgml/monitoring.sgml b/doc/src/sgml/monitoring.sgml
index 56d9b375ec..5366058bd5 100644
--- a/doc/src/sgml/monitoring.sgml
+++ b/doc/src/sgml/monitoring.sgml
@@ -1443,9 +1443,8 @@ postgres 27093 0.0 0.0 30096 2752 ? Ss 11:34 0:00 postgres: ser
<entry>Waiting for a write to a replication slot control file.</entry>
</row>
<row>
- <entry><literal>SLRUFlushSync</literal></entry>
- <entry>Waiting for SLRU data to reach durable storage during a checkpoint
- or database shutdown.</entry>
+ <entry><literal>SLRUFlush</literal></entry>
+ <entry>Initiating writeback of SLRU data.</entry>
</row>
<row>
<entry><literal>SLRURead</literal></entry>
@@ -1453,8 +1452,7 @@ postgres 27093 0.0 0.0 30096 2752 ? Ss 11:34 0:00 postgres: ser
</row>
<row>
<entry><literal>SLRUSync</literal></entry>
- <entry>Waiting for SLRU data to reach durable storage following a page
- write.</entry>
+ <entry>Waiting for SLRU data to reach durable storage.</entry>
</row>
<row>
<entry><literal>SLRUWrite</literal></entry>
@@ -1949,14 +1947,6 @@ postgres 27093 0.0 0.0 30096 2752 ? Ss 11:34 0:00 postgres: ser
<entry>Waiting to read or update the last value set for a
transaction commit timestamp.</entry>
</row>
- <row>
- <entry><literal>CommitTsBuffer</literal></entry>
- <entry>Waiting for I/O on a commit timestamp SLRU buffer.</entry>
- </row>
- <row>
- <entry><literal>CommitTsSLRU</literal></entry>
- <entry>Waiting to access the commit timestamp SLRU cache.</entry>
- </row>
<row>
<entry><literal>ControlFile</literal></entry>
<entry>Waiting to read or update the <filename>pg_control</filename>
@@ -1986,31 +1976,10 @@ postgres 27093 0.0 0.0 30096 2752 ? Ss 11:34 0:00 postgres: ser
<entry><literal>MultiXactGen</literal></entry>
<entry>Waiting to read or update shared multixact state.</entry>
</row>
- <row>
- <entry><literal>MultiXactMemberBuffer</literal></entry>
- <entry>Waiting for I/O on a multixact member SLRU buffer.</entry>
- </row>
- <row>
- <entry><literal>MultiXactMemberSLRU</literal></entry>
- <entry>Waiting to access the multixact member SLRU cache.</entry>
- </row>
- <row>
- <entry><literal>MultiXactOffsetBuffer</literal></entry>
- <entry>Waiting for I/O on a multixact offset SLRU buffer.</entry>
- </row>
- <row>
- <entry><literal>MultiXactOffsetSLRU</literal></entry>
- <entry>Waiting to access the multixact offset SLRU cache.</entry>
- </row>
<row>
<entry><literal>MultiXactTruncation</literal></entry>
<entry>Waiting to read or truncate multixact information.</entry>
</row>
- <row>
- <entry><literal>NotifyBuffer</literal></entry>
- <entry>Waiting for I/O on a <command>NOTIFY</command> message SLRU
- buffer.</entry>
- </row>
<row>
<entry><literal>NotifyQueue</literal></entry>
<entry>Waiting to read or update <command>NOTIFY</command> messages.</entry>
@@ -2020,11 +1989,6 @@ postgres 27093 0.0 0.0 30096 2752 ? Ss 11:34 0:00 postgres: ser
<entry>Waiting to update limit on <command>NOTIFY</command> message
storage.</entry>
</row>
- <row>
- <entry><literal>NotifySLRU</literal></entry>
- <entry>Waiting to access the <command>NOTIFY</command> message SLRU
- cache.</entry>
- </row>
<row>
<entry><literal>OidGen</literal></entry>
<entry>Waiting to allocate a new OID.</entry>
@@ -2109,11 +2073,6 @@ postgres 27093 0.0 0.0 30096 2752 ? Ss 11:34 0:00 postgres: ser
<entry><literal>ReplicationSlotIO</literal></entry>
<entry>Waiting for I/O on a replication slot.</entry>
</row>
- <row>
- <entry><literal>SerialBuffer</literal></entry>
- <entry>Waiting for I/O on a serializable transaction conflict SLRU
- buffer.</entry>
- </row>
<row>
<entry><literal>SerializableFinishedList</literal></entry>
<entry>Waiting to access the list of finished serializable
@@ -2141,11 +2100,6 @@ postgres 27093 0.0 0.0 30096 2752 ? Ss 11:34 0:00 postgres: ser
<entry>Waiting to read or update information about serializable
transactions.</entry>
</row>
- <row>
- <entry><literal>SerialSLRU</literal></entry>
- <entry>Waiting to access the serializable transaction conflict SLRU
- cache.</entry>
- </row>
<row>
<entry><literal>SharedTidBitmap</literal></entry>
<entry>Waiting to access a shared TID bitmap during a parallel bitmap
@@ -2213,10 +2167,6 @@ postgres 27093 0.0 0.0 30096 2752 ? Ss 11:34 0:00 postgres: ser
<entry>Waiting to update limits on transaction id and multixact
consumption.</entry>
</row>
- <row>
- <entry><literal>XactBuffer</literal></entry>
- <entry>Waiting for I/O on a transaction status SLRU buffer.</entry>
- </row>
<row>
<entry><literal>XactSLRU</literal></entry>
<entry>Waiting to access the transaction status SLRU cache.</entry>
diff --git a/src/backend/access/transam/clog.c b/src/backend/access/transam/clog.c
index 3d9088a704..cbe537ae72 100644
--- a/src/backend/access/transam/clog.c
+++ b/src/backend/access/transam/clog.c
@@ -41,6 +41,8 @@
#include "miscadmin.h"
#include "pg_trace.h"
#include "pgstat.h"
+#include "storage/bufmgr.h"
+#include "storage/buf_internals.h"
#include "storage/proc.h"
#include "storage/sync.h"
@@ -81,15 +83,8 @@
*/
#define THRESHOLD_SUBTRANS_CLOG_OPT 5
-/*
- * Link to shared-memory data structures for CLOG control
- */
-static SlruCtlData XactCtlData;
-
-#define XactCtl (&XactCtlData)
-
-static int ZeroCLOGPage(int pageno, bool writeXlog);
+static Buffer ZeroCLOGPage(int pageno, bool writeXlog);
static bool CLOGPagePrecedes(int page1, int page2);
static void WriteZeroPageXlogRec(int pageno);
static void WriteTruncateXlogRec(int pageno, TransactionId oldestXact,
@@ -99,7 +94,7 @@ static void TransactionIdSetPageStatus(TransactionId xid, int nsubxids,
XLogRecPtr lsn, int pageno,
bool all_xact_same_page);
static void TransactionIdSetStatusBit(TransactionId xid, XidStatus status,
- XLogRecPtr lsn, int slotno);
+ XLogRecPtr lsn, Buffer buffer);
static void set_status_by_pages(int nsubxids, TransactionId *subxids,
XidStatus status, XLogRecPtr lsn);
static bool TransactionGroupUpdateXidStatus(TransactionId xid,
@@ -341,13 +336,12 @@ TransactionIdSetPageStatusInternal(TransactionId xid, int nsubxids,
TransactionId *subxids, XidStatus status,
XLogRecPtr lsn, int pageno)
{
- int slotno;
+ Buffer buffer;
int i;
Assert(status == TRANSACTION_STATUS_COMMITTED ||
status == TRANSACTION_STATUS_ABORTED ||
(status == TRANSACTION_STATUS_SUB_COMMITTED && !TransactionIdIsValid(xid)));
- Assert(LWLockHeldByMeInMode(XactSLRULock, LW_EXCLUSIVE));
/*
* If we're doing an async commit (ie, lsn is valid), then we must wait
@@ -358,7 +352,8 @@ TransactionIdSetPageStatusInternal(TransactionId xid, int nsubxids,
* write-busy, since we don't care if the update reaches disk sooner than
* we think.
*/
- slotno = SimpleLruReadPage(XactCtl, pageno, XLogRecPtrIsInvalid(lsn), xid);
+ buffer = ReadSlruBuffer(SLRU_CLOG_REL_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
/*
* Set the main transaction id, if any.
@@ -376,25 +371,26 @@ TransactionIdSetPageStatusInternal(TransactionId xid, int nsubxids,
{
for (i = 0; i < nsubxids; i++)
{
- Assert(XactCtl->shared->page_number[slotno] == TransactionIdToPage(subxids[i]));
+ Assert(pageno == TransactionIdToPage(subxids[i]));
TransactionIdSetStatusBit(subxids[i],
TRANSACTION_STATUS_SUB_COMMITTED,
- lsn, slotno);
+ lsn, buffer);
}
}
/* ... then the main transaction */
- TransactionIdSetStatusBit(xid, status, lsn, slotno);
+ TransactionIdSetStatusBit(xid, status, lsn, buffer);
}
/* Set the subtransactions */
for (i = 0; i < nsubxids; i++)
{
- Assert(XactCtl->shared->page_number[slotno] == TransactionIdToPage(subxids[i]));
- TransactionIdSetStatusBit(subxids[i], status, lsn, slotno);
+ Assert(pageno == TransactionIdToPage(subxids[i]));
+ TransactionIdSetStatusBit(subxids[i], status, lsn, buffer);
}
- XactCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -568,7 +564,7 @@ TransactionGroupUpdateXidStatus(TransactionId xid, XidStatus status,
* Must be called with XactSLRULock held
*/
static void
-TransactionIdSetStatusBit(TransactionId xid, XidStatus status, XLogRecPtr lsn, int slotno)
+TransactionIdSetStatusBit(TransactionId xid, XidStatus status, XLogRecPtr lsn, Buffer buffer)
{
int byteno = TransactionIdToByte(xid);
int bshift = TransactionIdToBIndex(xid) * CLOG_BITS_PER_XACT;
@@ -576,7 +572,10 @@ TransactionIdSetStatusBit(TransactionId xid, XidStatus status, XLogRecPtr lsn, i
char byteval;
char curval;
- byteptr = XactCtl->shared->page_buffer[slotno] + byteno;
+ Assert(LWLockHeldByMeInMode(BufferDescriptorGetContentLock(GetBufferDescriptor(buffer - 1)),
+ LW_EXCLUSIVE));
+
+ byteptr = BufferGetPage(buffer) + byteno;
curval = (*byteptr >> bshift) & CLOG_XACT_BITMASK;
/*
@@ -605,7 +604,7 @@ TransactionIdSetStatusBit(TransactionId xid, XidStatus status, XLogRecPtr lsn, i
*byteptr = byteval;
/*
- * Update the group LSN if the transaction completion LSN is higher.
+ * Update the buffer LSN if the transaction completion LSN is higher.
*
* Note: lsn will be invalid when supplied during InRecovery processing,
* so we don't need to do anything special to avoid LSN updates during
@@ -614,10 +613,8 @@ TransactionIdSetStatusBit(TransactionId xid, XidStatus status, XLogRecPtr lsn, i
*/
if (!XLogRecPtrIsInvalid(lsn))
{
- int lsnindex = GetLSNIndex(slotno, xid);
-
- if (XactCtl->shared->group_lsn[lsnindex] < lsn)
- XactCtl->shared->group_lsn[lsnindex] = lsn;
+ if (BufferGetExternalLSN(GetBufferDescriptor(buffer)) < lsn)
+ BufferSetExternalLSN(GetBufferDescriptor(buffer), lsn);
}
}
@@ -642,67 +639,22 @@ TransactionIdGetStatus(TransactionId xid, XLogRecPtr *lsn)
int pageno = TransactionIdToPage(xid);
int byteno = TransactionIdToByte(xid);
int bshift = TransactionIdToBIndex(xid) * CLOG_BITS_PER_XACT;
- int slotno;
- int lsnindex;
char *byteptr;
XidStatus status;
+ Buffer buffer;
- /* lock is acquired by SimpleLruReadPage_ReadOnly */
-
- slotno = SimpleLruReadPage_ReadOnly(XactCtl, pageno, xid);
- byteptr = XactCtl->shared->page_buffer[slotno] + byteno;
+ buffer = ReadSlruBuffer(SLRU_CLOG_REL_ID, pageno);
+ byteptr = BufferGetPage(buffer) + byteno;
status = (*byteptr >> bshift) & CLOG_XACT_BITMASK;
- lsnindex = GetLSNIndex(slotno, xid);
- *lsn = XactCtl->shared->group_lsn[lsnindex];
+ *lsn = BufferGetExternalLSN(GetBufferDescriptor(buffer));
- LWLockRelease(XactSLRULock);
+ ReleaseBuffer(buffer);
return status;
}
-/*
- * Number of shared CLOG buffers.
- *
- * On larger multi-processor systems, it is possible to have many CLOG page
- * requests in flight at one time which could lead to disk access for CLOG
- * page if the required page is not found in memory. Testing revealed that we
- * can get the best performance by having 128 CLOG buffers, more than that it
- * doesn't improve performance.
- *
- * Unconditionally keeping the number of CLOG buffers to 128 did not seem like
- * a good idea, because it would increase the minimum amount of shared memory
- * required to start, which could be a problem for people running very small
- * configurations. The following formula seems to represent a reasonable
- * compromise: people with very low values for shared_buffers will get fewer
- * CLOG buffers as well, and everyone else will get 128.
- */
-Size
-CLOGShmemBuffers(void)
-{
- return Min(128, Max(4, NBuffers / 512));
-}
-
-/*
- * Initialization of shared memory for CLOG
- */
-Size
-CLOGShmemSize(void)
-{
- return SimpleLruShmemSize(CLOGShmemBuffers(), CLOG_LSNS_PER_PAGE);
-}
-
-void
-CLOGShmemInit(void)
-{
- XactCtl->PagePrecedes = CLOGPagePrecedes;
- SimpleLruInit(XactCtl, "Xact", CLOGShmemBuffers(), CLOG_LSNS_PER_PAGE,
- XactSLRULock, "pg_xact", LWTRANCHE_XACT_BUFFER,
- SYNC_HANDLER_CLOG);
- SlruPagePrecedesUnitTests(XactCtl, CLOG_XACTS_PER_PAGE);
-}
-
/*
* This func must be called ONCE on system install. It creates
* the initial CLOG segment. (The CLOG directory is assumed to
@@ -712,18 +664,15 @@ CLOGShmemInit(void)
void
BootStrapCLOG(void)
{
- int slotno;
-
- LWLockAcquire(XactSLRULock, LW_EXCLUSIVE);
+ Buffer buffer;
/* Create and zero the first page of the commit log */
- slotno = ZeroCLOGPage(0, false);
+ buffer = ZeroCLOGPage(0, false);
/* Make sure it's written out */
- SimpleLruWritePage(XactCtl, slotno);
- Assert(!XactCtl->shared->page_dirty[slotno]);
+ FlushOneBuffer(buffer);
- LWLockRelease(XactSLRULock);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -735,17 +684,18 @@ BootStrapCLOG(void)
*
* Control lock must be held at entry, and will be held at exit.
*/
-static int
+static Buffer
ZeroCLOGPage(int pageno, bool writeXlog)
{
- int slotno;
+ Buffer buffer;
- slotno = SimpleLruZeroPage(XactCtl, pageno);
+ buffer = ZeroSlruBuffer(SLRU_CLOG_REL_ID, pageno);
+ MarkBufferDirty(buffer);
if (writeXlog)
WriteZeroPageXlogRec(pageno);
- return slotno;
+ return buffer;
}
/*
@@ -755,17 +705,6 @@ ZeroCLOGPage(int pageno, bool writeXlog)
void
StartupCLOG(void)
{
- TransactionId xid = XidFromFullTransactionId(ShmemVariableCache->nextXid);
- int pageno = TransactionIdToPage(xid);
-
- LWLockAcquire(XactSLRULock, LW_EXCLUSIVE);
-
- /*
- * Initialize our idea of the latest page number.
- */
- XactCtl->shared->latest_page_number = pageno;
-
- LWLockRelease(XactSLRULock);
}
/*
@@ -777,8 +716,6 @@ TrimCLOG(void)
TransactionId xid = XidFromFullTransactionId(ShmemVariableCache->nextXid);
int pageno = TransactionIdToPage(xid);
- LWLockAcquire(XactSLRULock, LW_EXCLUSIVE);
-
/*
* Zero out the remainder of the current clog page. Under normal
* circumstances it should be zeroes already, but it seems at least
@@ -795,40 +732,24 @@ TrimCLOG(void)
{
int byteno = TransactionIdToByte(xid);
int bshift = TransactionIdToBIndex(xid) * CLOG_BITS_PER_XACT;
- int slotno;
char *byteptr;
+ Buffer buffer;
- slotno = SimpleLruReadPage(XactCtl, pageno, false, xid);
- byteptr = XactCtl->shared->page_buffer[slotno] + byteno;
+ buffer = ReadSlruBuffer(SLRU_CLOG_REL_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ byteptr = BufferGetPage(buffer) + byteno;
/* Zero so-far-unused positions in the current byte */
*byteptr &= (1 << bshift) - 1;
/* Zero the rest of the page */
MemSet(byteptr + 1, 0, BLCKSZ - byteno - 1);
- XactCtl->shared->page_dirty[slotno] = true;
- }
+ MarkBufferDirty(buffer);
- LWLockRelease(XactSLRULock);
-}
-
-/*
- * Perform a checkpoint --- either during shutdown, or on-the-fly
- */
-void
-CheckPointCLOG(void)
-{
- /*
- * Write dirty CLOG pages to disk. This may result in sync requests
- * queued for later handling by ProcessSyncRequests(), as part of the
- * checkpoint.
- */
- TRACE_POSTGRESQL_CLOG_CHECKPOINT_START(true);
- SimpleLruWriteAll(XactCtl, true);
- TRACE_POSTGRESQL_CLOG_CHECKPOINT_DONE(true);
+ UnlockReleaseBuffer(buffer);
+ }
}
-
/*
* Make sure that CLOG has room for a newly-allocated XID.
*
@@ -852,12 +773,8 @@ ExtendCLOG(TransactionId newestXact)
pageno = TransactionIdToPage(newestXact);
- LWLockAcquire(XactSLRULock, LW_EXCLUSIVE);
-
/* Zero the page and make an XLOG entry about it */
- ZeroCLOGPage(pageno, true);
-
- LWLockRelease(XactSLRULock);
+ UnlockReleaseBuffer(ZeroCLOGPage(pageno, true));
}
@@ -888,7 +805,8 @@ TruncateCLOG(TransactionId oldestXact, Oid oldestxid_datoid)
cutoffPage = TransactionIdToPage(oldestXact);
/* Check to see if there's any files that could be removed */
- if (!SlruScanDirectory(XactCtl, SlruScanDirCbReportPresence, &cutoffPage))
+ if (!SlruScanDirectory(SLRU_CLOG_REL_ID, CLOGPagePrecedes,
+ SlruScanDirCbReportPresence, &cutoffPage))
return; /* nothing to remove */
/*
@@ -909,7 +827,7 @@ TruncateCLOG(TransactionId oldestXact, Oid oldestxid_datoid)
WriteTruncateXlogRec(cutoffPage, oldestXact, oldestxid_datoid);
/* Now we can remove the old CLOG segment(s) */
- SimpleLruTruncate(XactCtl, cutoffPage);
+ SimpleLruTruncate(SLRU_CLOG_REL_ID, CLOGPagePrecedes, cutoffPage);
}
@@ -994,17 +912,13 @@ clog_redo(XLogReaderState *record)
if (info == CLOG_ZEROPAGE)
{
int pageno;
- int slotno;
+ Buffer buffer;
memcpy(&pageno, XLogRecGetData(record), sizeof(int));
- LWLockAcquire(XactSLRULock, LW_EXCLUSIVE);
-
- slotno = ZeroCLOGPage(pageno, false);
- SimpleLruWritePage(XactCtl, slotno);
- Assert(!XactCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(XactSLRULock);
+ buffer = ZeroCLOGPage(pageno, false);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
else if (info == CLOG_TRUNCATE)
{
@@ -1014,17 +928,8 @@ clog_redo(XLogReaderState *record)
AdvanceOldestClogXid(xlrec.oldestXact);
- SimpleLruTruncate(XactCtl, xlrec.pageno);
+ SimpleLruTruncate(SLRU_CLOG_REL_ID, CLOGPagePrecedes, xlrec.pageno);
}
else
elog(PANIC, "clog_redo: unknown op code %u", info);
}
-
-/*
- * Entrypoint for sync.c to sync clog files.
- */
-int
-clogsyncfiletag(const FileTag *ftag, char *path)
-{
- return SlruSyncFileTag(XactCtl, ftag, path);
-}
diff --git a/src/backend/access/transam/commit_ts.c b/src/backend/access/transam/commit_ts.c
index 4dc8d402bd..1eac761b39 100644
--- a/src/backend/access/transam/commit_ts.c
+++ b/src/backend/access/transam/commit_ts.c
@@ -73,13 +73,6 @@ typedef struct CommitTimestampEntry
#define TransactionIdToCTsEntry(xid) \
((xid) % (TransactionId) COMMIT_TS_XACTS_PER_PAGE)
-/*
- * Link to shared-memory data structures for CommitTs control
- */
-static SlruCtlData CommitTsCtlData;
-
-#define CommitTsCtl (&CommitTsCtlData)
-
/*
* We keep a cache of the last value set in shared memory.
*
@@ -110,7 +103,7 @@ static void SetXidCommitTsInPage(TransactionId xid, int nsubxids,
static void TransactionIdSetCommitTs(TransactionId xid, TimestampTz ts,
RepOriginId nodeid, int slotno);
static void error_commit_ts_disabled(void);
-static int ZeroCommitTsPage(int pageno, bool writeXlog);
+static Buffer ZeroCommitTsPage(int pageno, bool writeXlog);
static bool CommitTsPagePrecedes(int page1, int page2);
static void ActivateCommitTs(void);
static void DeactivateCommitTs(void);
@@ -219,30 +212,27 @@ SetXidCommitTsInPage(TransactionId xid, int nsubxids,
TransactionId *subxids, TimestampTz ts,
RepOriginId nodeid, int pageno)
{
- int slotno;
int i;
+ Buffer buffer;
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
-
- slotno = SimpleLruReadPage(CommitTsCtl, pageno, true, xid);
+ buffer = ReadSlruBuffer(SLRU_COMMITTS_REL_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
- TransactionIdSetCommitTs(xid, ts, nodeid, slotno);
+ TransactionIdSetCommitTs(xid, ts, nodeid, buffer);
for (i = 0; i < nsubxids; i++)
- TransactionIdSetCommitTs(subxids[i], ts, nodeid, slotno);
+ TransactionIdSetCommitTs(subxids[i], ts, nodeid, buffer);
- CommitTsCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
- LWLockRelease(CommitTsSLRULock);
+ UnlockReleaseBuffer(buffer);
}
/*
* Sets the commit timestamp of a single transaction.
- *
- * Must be called with CommitTsSLRULock held
*/
static void
TransactionIdSetCommitTs(TransactionId xid, TimestampTz ts,
- RepOriginId nodeid, int slotno)
+ RepOriginId nodeid, Buffer buffer)
{
int entryno = TransactionIdToCTsEntry(xid);
CommitTimestampEntry entry;
@@ -252,8 +242,7 @@ TransactionIdSetCommitTs(TransactionId xid, TimestampTz ts,
entry.time = ts;
entry.nodeid = nodeid;
- memcpy(CommitTsCtl->shared->page_buffer[slotno] +
- SizeOfCommitTimestampEntry * entryno,
+ memcpy(BufferGetPage(buffer) + SizeOfCommitTimestampEntry * entryno,
&entry, SizeOfCommitTimestampEntry);
}
@@ -271,10 +260,10 @@ TransactionIdGetCommitTsData(TransactionId xid, TimestampTz *ts,
{
int pageno = TransactionIdToCTsPage(xid);
int entryno = TransactionIdToCTsEntry(xid);
- int slotno;
CommitTimestampEntry entry;
TransactionId oldestCommitTsXid;
TransactionId newestCommitTsXid;
+ Buffer buffer;
if (!TransactionIdIsValid(xid))
ereport(ERROR,
@@ -328,10 +317,11 @@ TransactionIdGetCommitTsData(TransactionId xid, TimestampTz *ts,
return false;
}
- /* lock is acquired by SimpleLruReadPage_ReadOnly */
- slotno = SimpleLruReadPage_ReadOnly(CommitTsCtl, pageno, xid);
+ buffer = ReadSlruBuffer(SLRU_COMMITTS_REL_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_SHARE);
+
memcpy(&entry,
- CommitTsCtl->shared->page_buffer[slotno] +
+ BufferGetPage(buffer) +
SizeOfCommitTimestampEntry * entryno,
SizeOfCommitTimestampEntry);
@@ -339,7 +329,7 @@ TransactionIdGetCommitTsData(TransactionId xid, TimestampTz *ts,
if (nodeid)
*nodeid = entry.nodeid;
- LWLockRelease(CommitTsSLRULock);
+ UnlockReleaseBuffer(buffer);
return *ts != 0;
}
@@ -508,27 +498,13 @@ pg_xact_commit_timestamp_origin(PG_FUNCTION_ARGS)
PG_RETURN_DATUM(HeapTupleGetDatum(htup));
}
-/*
- * Number of shared CommitTS buffers.
- *
- * We use a very similar logic as for the number of CLOG buffers (except we
- * scale up twice as fast with shared buffers, and the maximum is twice as
- * high); see comments in CLOGShmemBuffers.
- */
-Size
-CommitTsShmemBuffers(void)
-{
- return Min(256, Max(4, NBuffers / 256));
-}
-
/*
* Shared memory sizing for CommitTs
*/
Size
CommitTsShmemSize(void)
{
- return SimpleLruShmemSize(CommitTsShmemBuffers(), 0) +
- sizeof(CommitTimestampShared);
+ return sizeof(CommitTimestampShared);
}
/*
@@ -540,12 +516,7 @@ CommitTsShmemInit(void)
{
bool found;
- CommitTsCtl->PagePrecedes = CommitTsPagePrecedes;
- SimpleLruInit(CommitTsCtl, "CommitTs", CommitTsShmemBuffers(), 0,
- CommitTsSLRULock, "pg_commit_ts",
- LWTRANCHE_COMMITTS_BUFFER,
- SYNC_HANDLER_COMMIT_TS);
- SlruPagePrecedesUnitTests(CommitTsCtl, COMMIT_TS_XACTS_PER_PAGE);
+ SlruPagePrecedesUnitTests(CommitTsPagePrecedes, COMMIT_TS_XACTS_PER_PAGE);
commitTsShared = ShmemInitStruct("CommitTs shared",
sizeof(CommitTimestampShared),
@@ -589,17 +560,18 @@ BootStrapCommitTs(void)
*
* Control lock must be held at entry, and will be held at exit.
*/
-static int
+static Buffer
ZeroCommitTsPage(int pageno, bool writeXlog)
{
- int slotno;
+ Buffer buffer;
- slotno = SimpleLruZeroPage(CommitTsCtl, pageno);
+ buffer = ZeroSlruBuffer(SLRU_COMMITTS_REL_ID, pageno);
+ MarkBufferDirty(buffer);
if (writeXlog)
WriteZeroPageXlogRec(pageno);
- return slotno;
+ return buffer;
}
/*
@@ -697,13 +669,6 @@ ActivateCommitTs(void)
xid = XidFromFullTransactionId(ShmemVariableCache->nextXid);
pageno = TransactionIdToCTsPage(xid);
- /*
- * Re-Initialize our idea of the latest page number.
- */
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
- CommitTsCtl->shared->latest_page_number = pageno;
- LWLockRelease(CommitTsSLRULock);
-
/*
* If CommitTs is enabled, but it wasn't in the previous server run, we
* need to set the oldest and newest values to the next Xid; that way, we
@@ -726,15 +691,14 @@ ActivateCommitTs(void)
LWLockRelease(CommitTsLock);
/* Create the current segment file, if necessary */
- if (!SimpleLruDoesPhysicalPageExist(CommitTsCtl, pageno))
+ if (!SimpleLruDoesPhysicalPageExist(SLRU_COMMITTS_REL_ID, pageno))
{
- int slotno;
+ Buffer buffer;
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
- slotno = ZeroCommitTsPage(pageno, false);
- SimpleLruWritePage(CommitTsCtl, slotno);
- Assert(!CommitTsCtl->shared->page_dirty[slotno]);
- LWLockRelease(CommitTsSLRULock);
+ buffer = ZeroSlruBuffer(SLRU_COMMITTS_REL_ID, pageno);
+ MarkBufferDirty(buffer);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
/* Change the activation status in shared memory. */
@@ -783,23 +747,9 @@ DeactivateCommitTs(void)
* be overwritten anyway when we wrap around, but it seems better to be
* tidy.)
*/
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
- (void) SlruScanDirectory(CommitTsCtl, SlruScanDirCbDeleteAll, NULL);
- LWLockRelease(CommitTsSLRULock);
-}
-
-/*
- * Perform a checkpoint --- either during shutdown, or on-the-fly
- */
-void
-CheckPointCommitTs(void)
-{
- /*
- * Write dirty CommitTs pages to disk. This may result in sync requests
- * queued for later handling by ProcessSyncRequests(), as part of the
- * checkpoint.
- */
- SimpleLruWriteAll(CommitTsCtl, true);
+ (void) SlruScanDirectory(SLRU_COMMITTS_REL_ID,
+ CommitTsPagePrecedes,
+ SlruScanDirCbDeleteAll, NULL);
}
/*
@@ -837,12 +787,8 @@ ExtendCommitTs(TransactionId newestXact)
pageno = TransactionIdToCTsPage(newestXact);
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
-
/* Zero the page and make an XLOG entry about it */
- ZeroCommitTsPage(pageno, !InRecovery);
-
- LWLockRelease(CommitTsSLRULock);
+ UnlockReleaseBuffer(ZeroCommitTsPage(pageno, !InRecovery));
}
/*
@@ -863,7 +809,9 @@ TruncateCommitTs(TransactionId oldestXact)
cutoffPage = TransactionIdToCTsPage(oldestXact);
/* Check to see if there's any files that could be removed */
- if (!SlruScanDirectory(CommitTsCtl, SlruScanDirCbReportPresence,
+ if (!SlruScanDirectory(SLRU_COMMITTS_REL_ID,
+ CommitTsPagePrecedes,
+ SlruScanDirCbReportPresence,
&cutoffPage))
return; /* nothing to remove */
@@ -871,7 +819,7 @@ TruncateCommitTs(TransactionId oldestXact)
WriteTruncateXlogRec(cutoffPage, oldestXact);
/* Now we can remove the old CommitTs segment(s) */
- SimpleLruTruncate(CommitTsCtl, cutoffPage);
+ SimpleLruTruncate(SLRU_COMMITTS_REL_ID, CommitTsPagePrecedes, cutoffPage);
}
/*
@@ -995,17 +943,14 @@ commit_ts_redo(XLogReaderState *record)
if (info == COMMIT_TS_ZEROPAGE)
{
int pageno;
- int slotno;
+ Buffer buffer;
memcpy(&pageno, XLogRecGetData(record), sizeof(int));
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
-
- slotno = ZeroCommitTsPage(pageno, false);
- SimpleLruWritePage(CommitTsCtl, slotno);
- Assert(!CommitTsCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(CommitTsSLRULock);
+ buffer = ZeroSlruBuffer(SLRU_COMMITTS_REL_ID, pageno);
+ MarkBufferDirty(buffer);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
else if (info == COMMIT_TS_TRUNCATE)
{
@@ -1013,23 +958,8 @@ commit_ts_redo(XLogReaderState *record)
AdvanceOldestCommitTsXid(trunc->oldestXid);
- /*
- * During XLOG replay, latest_page_number isn't set up yet; insert a
- * suitable value to bypass the sanity test in SimpleLruTruncate.
- */
- CommitTsCtl->shared->latest_page_number = trunc->pageno;
-
- SimpleLruTruncate(CommitTsCtl, trunc->pageno);
+ SimpleLruTruncate(SLRU_COMMITTS_REL_ID, CommitTsPagePrecedes, trunc->pageno);
}
else
elog(PANIC, "commit_ts_redo: unknown op code %u", info);
}
-
-/*
- * Entrypoint for sync.c to sync commit_ts files.
- */
-int
-committssyncfiletag(const FileTag *ftag, char *path)
-{
- return SlruSyncFileTag(CommitTsCtl, ftag, path);
-}
diff --git a/src/backend/access/transam/multixact.c b/src/backend/access/transam/multixact.c
index 8f7d12950e..25a348d1b1 100644
--- a/src/backend/access/transam/multixact.c
+++ b/src/backend/access/transam/multixact.c
@@ -181,15 +181,6 @@
#define PreviousMultiXactId(xid) \
((xid) == FirstMultiXactId ? MaxMultiXactId : (xid) - 1)
-/*
- * Links to shared-memory data structures for MultiXact control
- */
-static SlruCtlData MultiXactOffsetCtlData;
-static SlruCtlData MultiXactMemberCtlData;
-
-#define MultiXactOffsetCtl (&MultiXactOffsetCtlData)
-#define MultiXactMemberCtl (&MultiXactMemberCtlData)
-
/*
* MultiXact state shared across all backends. All this state is protected
* by MultiXactGenLock. (We also use MultiXactOffsetSLRULock and
@@ -353,10 +344,9 @@ static void mXactCachePut(MultiXactId multi, int nmembers,
static char *mxstatus_to_string(MultiXactStatus status);
/* management of SLRU infrastructure */
-static int ZeroMultiXactOffsetPage(int pageno, bool writeXlog);
-static int ZeroMultiXactMemberPage(int pageno, bool writeXlog);
+static Buffer ZeroMultiXactOffsetPage(int pageno, bool writeXlog);
+static Buffer ZeroMultiXactMemberPage(int pageno, bool writeXlog);
static bool MultiXactOffsetPagePrecedes(int page1, int page2);
-static bool MultiXactMemberPagePrecedes(int page1, int page2);
static bool MultiXactOffsetPrecedes(MultiXactOffset offset1,
MultiXactOffset offset2);
static void ExtendMultiXactOffset(MultiXactId multi);
@@ -865,34 +855,25 @@ RecordNewMultiXact(MultiXactId multi, MultiXactOffset offset,
int pageno;
int prev_pageno;
int entryno;
- int slotno;
MultiXactOffset *offptr;
int i;
-
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
+ Buffer buffer;
pageno = MultiXactIdToOffsetPage(multi);
entryno = MultiXactIdToOffsetEntry(multi);
- /*
- * Note: we pass the MultiXactId to SimpleLruReadPage as the "transaction"
- * to complain about if there's any I/O error. This is kinda bogus, but
- * since the errors will always give the full pathname, it should be clear
- * enough that a MultiXactId is really involved. Perhaps someday we'll
- * take the trouble to generalize the slru.c error reporting code.
- */
- slotno = SimpleLruReadPage(MultiXactOffsetCtl, pageno, true, multi);
- offptr = (MultiXactOffset *) MultiXactOffsetCtl->shared->page_buffer[slotno];
+ /* XXX set up error context? */
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_OFFSET_REL_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ offptr = (MultiXactOffset *) BufferGetPage(buffer);
offptr += entryno;
*offptr = offset;
- MultiXactOffsetCtl->shared->page_dirty[slotno] = true;
-
- /* Exchange our lock */
- LWLockRelease(MultiXactOffsetSLRULock);
+ MarkBufferDirty(buffer);
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
+ UnlockReleaseBuffer(buffer);
+ buffer = InvalidBuffer;
prev_pageno = -1;
@@ -914,27 +895,28 @@ RecordNewMultiXact(MultiXactId multi, MultiXactOffset offset,
if (pageno != prev_pageno)
{
- slotno = SimpleLruReadPage(MultiXactMemberCtl, pageno, true, multi);
+ if (BufferIsValid(buffer))
+ UnlockReleaseBuffer(buffer);
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_MEMBER_REL_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
prev_pageno = pageno;
}
- memberptr = (TransactionId *)
- (MultiXactMemberCtl->shared->page_buffer[slotno] + memberoff);
+ memberptr = (TransactionId *) (BufferGetPage(buffer) + memberoff);
*memberptr = members[i].xid;
- flagsptr = (uint32 *)
- (MultiXactMemberCtl->shared->page_buffer[slotno] + flagsoff);
+ flagsptr = (uint32 *) (BufferGetPage(buffer) + flagsoff);
flagsval = *flagsptr;
flagsval &= ~(((1 << MXACT_MEMBER_BITS_PER_XACT) - 1) << bshift);
flagsval |= (members[i].status << bshift);
*flagsptr = flagsval;
- MultiXactMemberCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
}
- LWLockRelease(MultiXactMemberSLRULock);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -1226,7 +1208,6 @@ GetMultiXactIdMembers(MultiXactId multi, MultiXactMember **members,
int pageno;
int prev_pageno;
int entryno;
- int slotno;
MultiXactOffset *offptr;
MultiXactOffset offset;
int length;
@@ -1237,6 +1218,7 @@ GetMultiXactIdMembers(MultiXactId multi, MultiXactMember **members,
MultiXactId tmpMXact;
MultiXactOffset nextOffset;
MultiXactMember *ptr;
+ Buffer buffer;
debug_elog3(DEBUG2, "GetMembers: asked for %u", multi);
@@ -1340,13 +1322,12 @@ GetMultiXactIdMembers(MultiXactId multi, MultiXactMember **members,
* time on every multixact creation.
*/
retry:
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
-
pageno = MultiXactIdToOffsetPage(multi);
entryno = MultiXactIdToOffsetEntry(multi);
- slotno = SimpleLruReadPage(MultiXactOffsetCtl, pageno, true, multi);
- offptr = (MultiXactOffset *) MultiXactOffsetCtl->shared->page_buffer[slotno];
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_OFFSET_REL_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_SHARE);
+ offptr = (MultiXactOffset *) BufferGetPage(buffer);
offptr += entryno;
offset = *offptr;
@@ -1377,16 +1358,20 @@ retry:
entryno = MultiXactIdToOffsetEntry(tmpMXact);
if (pageno != prev_pageno)
- slotno = SimpleLruReadPage(MultiXactOffsetCtl, pageno, true, tmpMXact);
+ {
+ UnlockReleaseBuffer(buffer);
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_OFFSET_REL_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_SHARE);
+ }
- offptr = (MultiXactOffset *) MultiXactOffsetCtl->shared->page_buffer[slotno];
+ offptr = (MultiXactOffset *) BufferGetPage(buffer);
offptr += entryno;
nextMXOffset = *offptr;
if (nextMXOffset == 0)
{
/* Corner case 2: next multixact is still being filled in */
- LWLockRelease(MultiXactOffsetSLRULock);
+ UnlockReleaseBuffer(buffer);
CHECK_FOR_INTERRUPTS();
pg_usleep(1000L);
goto retry;
@@ -1394,14 +1379,11 @@ retry:
length = nextMXOffset - offset;
}
-
- LWLockRelease(MultiXactOffsetSLRULock);
+ UnlockReleaseBuffer(buffer);
+ buffer = InvalidBuffer;
ptr = (MultiXactMember *) palloc(length * sizeof(MultiXactMember));
- /* Now get the members themselves. */
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
-
truelength = 0;
prev_pageno = -1;
for (i = 0; i < length; i++, offset++)
@@ -1417,12 +1399,14 @@ retry:
if (pageno != prev_pageno)
{
- slotno = SimpleLruReadPage(MultiXactMemberCtl, pageno, true, multi);
+ if (BufferIsValid(buffer))
+ UnlockReleaseBuffer(buffer);
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_MEMBER_REL_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_SHARE);
prev_pageno = pageno;
}
- xactptr = (TransactionId *)
- (MultiXactMemberCtl->shared->page_buffer[slotno] + memberoff);
+ xactptr = (TransactionId *) (BufferGetPage(buffer) + memberoff);
if (!TransactionIdIsValid(*xactptr))
{
@@ -1433,14 +1417,13 @@ retry:
flagsoff = MXOffsetToFlagsOffset(offset);
bshift = MXOffsetToFlagsBitShift(offset);
- flagsptr = (uint32 *) (MultiXactMemberCtl->shared->page_buffer[slotno] + flagsoff);
+ flagsptr = (uint32 *) (BufferGetPage(buffer) + flagsoff);
ptr[truelength].xid = *xactptr;
ptr[truelength].status = (*flagsptr >> bshift) & MXACT_MEMBER_XACT_BITMASK;
truelength++;
}
-
- LWLockRelease(MultiXactMemberSLRULock);
+ UnlockReleaseBuffer(buffer);
/* A multixid with zero members should not happen */
Assert(truelength > 0);
@@ -1834,8 +1817,6 @@ MultiXactShmemSize(void)
mul_size(sizeof(MultiXactId) * 2, MaxOldestSlot))
size = SHARED_MULTIXACT_STATE_SIZE;
- size = add_size(size, SimpleLruShmemSize(NUM_MULTIXACTOFFSET_BUFFERS, 0));
- size = add_size(size, SimpleLruShmemSize(NUM_MULTIXACTMEMBER_BUFFERS, 0));
return size;
}
@@ -1847,22 +1828,6 @@ MultiXactShmemInit(void)
debug_elog2(DEBUG2, "Shared Memory Init for MultiXact");
- MultiXactOffsetCtl->PagePrecedes = MultiXactOffsetPagePrecedes;
- MultiXactMemberCtl->PagePrecedes = MultiXactMemberPagePrecedes;
-
- SimpleLruInit(MultiXactOffsetCtl,
- "MultiXactOffset", NUM_MULTIXACTOFFSET_BUFFERS, 0,
- MultiXactOffsetSLRULock, "pg_multixact/offsets",
- LWTRANCHE_MULTIXACTOFFSET_BUFFER,
- SYNC_HANDLER_MULTIXACT_OFFSET);
- SlruPagePrecedesUnitTests(MultiXactOffsetCtl, MULTIXACT_OFFSETS_PER_PAGE);
- SimpleLruInit(MultiXactMemberCtl,
- "MultiXactMember", NUM_MULTIXACTMEMBER_BUFFERS, 0,
- MultiXactMemberSLRULock, "pg_multixact/members",
- LWTRANCHE_MULTIXACTMEMBER_BUFFER,
- SYNC_HANDLER_MULTIXACT_MEMBER);
- /* doesn't call SimpleLruTruncate() or meet criteria for unit tests */
-
/* Initialize our shared state struct */
MultiXactState = ShmemInitStruct("Shared MultiXact State",
SHARED_MULTIXACT_STATE_SIZE,
@@ -1893,29 +1858,17 @@ MultiXactShmemInit(void)
void
BootStrapMultiXact(void)
{
- int slotno;
-
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
+ Buffer buffer;
/* Create and zero the first page of the offsets log */
- slotno = ZeroMultiXactOffsetPage(0, false);
-
- /* Make sure it's written out */
- SimpleLruWritePage(MultiXactOffsetCtl, slotno);
- Assert(!MultiXactOffsetCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(MultiXactOffsetSLRULock);
-
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
+ buffer = ZeroMultiXactOffsetPage(0, false);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
/* Create and zero the first page of the members log */
- slotno = ZeroMultiXactMemberPage(0, false);
-
- /* Make sure it's written out */
- SimpleLruWritePage(MultiXactMemberCtl, slotno);
- Assert(!MultiXactMemberCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(MultiXactMemberSLRULock);
+ buffer = ZeroMultiXactMemberPage(0, false);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -1927,33 +1880,35 @@ BootStrapMultiXact(void)
*
* Control lock must be held at entry, and will be held at exit.
*/
-static int
+static Buffer
ZeroMultiXactOffsetPage(int pageno, bool writeXlog)
{
- int slotno;
+ Buffer buffer;
- slotno = SimpleLruZeroPage(MultiXactOffsetCtl, pageno);
+ buffer = ZeroSlruBuffer(SLRU_MULTIXACT_OFFSET_REL_ID, pageno);
+ MarkBufferDirty(buffer);
if (writeXlog)
WriteMZeroPageXlogRec(pageno, XLOG_MULTIXACT_ZERO_OFF_PAGE);
- return slotno;
+ return buffer;
}
/*
* Ditto, for MultiXactMember
*/
-static int
+static Buffer
ZeroMultiXactMemberPage(int pageno, bool writeXlog)
{
- int slotno;
+ Buffer buffer;
- slotno = SimpleLruZeroPage(MultiXactMemberCtl, pageno);
+ buffer = ZeroSlruBuffer(SLRU_MULTIXACT_MEMBER_REL_ID, pageno);
+ MarkBufferDirty(buffer);
if (writeXlog)
WriteMZeroPageXlogRec(pageno, XLOG_MULTIXACT_ZERO_MEM_PAGE);
- return slotno;
+ return buffer;
}
/*
@@ -1978,22 +1933,14 @@ MaybeExtendOffsetSlru(void)
pageno = MultiXactIdToOffsetPage(MultiXactState->nextMXact);
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
-
- if (!SimpleLruDoesPhysicalPageExist(MultiXactOffsetCtl, pageno))
+ if (!SimpleLruDoesPhysicalPageExist(SLRU_MULTIXACT_OFFSET_REL_ID, pageno))
{
- int slotno;
+ Buffer buffer;
- /*
- * Fortunately for us, SimpleLruWritePage is already prepared to deal
- * with creating a new segment file even if the page we're writing is
- * not the first in it, so this is enough.
- */
- slotno = ZeroMultiXactOffsetPage(pageno, false);
- SimpleLruWritePage(MultiXactOffsetCtl, slotno);
+ buffer = ZeroMultiXactOffsetPage(pageno, false);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
-
- LWLockRelease(MultiXactOffsetSLRULock);
}
/*
@@ -2007,21 +1954,6 @@ MaybeExtendOffsetSlru(void)
void
StartupMultiXact(void)
{
- MultiXactId multi = MultiXactState->nextMXact;
- MultiXactOffset offset = MultiXactState->nextOffset;
- int pageno;
-
- /*
- * Initialize offset's idea of the latest page number.
- */
- pageno = MultiXactIdToOffsetPage(multi);
- MultiXactOffsetCtl->shared->latest_page_number = pageno;
-
- /*
- * Initialize member's idea of the latest page number.
- */
- pageno = MXOffsetToMemberPage(offset);
- MultiXactMemberCtl->shared->latest_page_number = pageno;
}
/*
@@ -2045,14 +1977,7 @@ TrimMultiXact(void)
oldestMXactDB = MultiXactState->oldestMultiXactDB;
LWLockRelease(MultiXactGenLock);
- /* Clean up offsets state */
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
-
- /*
- * (Re-)Initialize our idea of the latest page number for offsets.
- */
- pageno = MultiXactIdToOffsetPage(nextMXact);
- MultiXactOffsetCtl->shared->latest_page_number = pageno;
+ pageno = MXOffsetToMemberPage(offset);
/*
* Zero out the remainder of the current offsets page. See notes in
@@ -2065,29 +1990,20 @@ TrimMultiXact(void)
entryno = MultiXactIdToOffsetEntry(nextMXact);
if (entryno != 0)
{
- int slotno;
MultiXactOffset *offptr;
+ Buffer buffer;
- slotno = SimpleLruReadPage(MultiXactOffsetCtl, pageno, true, nextMXact);
- offptr = (MultiXactOffset *) MultiXactOffsetCtl->shared->page_buffer[slotno];
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_OFFSET_REL_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ offptr = (MultiXactOffset *) BufferGetPage(buffer);
offptr += entryno;
MemSet(offptr, 0, BLCKSZ - (entryno * sizeof(MultiXactOffset)));
- MultiXactOffsetCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
+ UnlockReleaseBuffer(buffer);
}
- LWLockRelease(MultiXactOffsetSLRULock);
-
- /* And the same for members */
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
-
- /*
- * (Re-)Initialize our idea of the latest page number for members.
- */
- pageno = MXOffsetToMemberPage(offset);
- MultiXactMemberCtl->shared->latest_page_number = pageno;
-
/*
* Zero out the remainder of the current members page. See notes in
* TrimCLOG() for motivation.
@@ -2095,14 +2011,14 @@ TrimMultiXact(void)
flagsoff = MXOffsetToFlagsOffset(offset);
if (flagsoff != 0)
{
- int slotno;
TransactionId *xidptr;
int memberoff;
+ Buffer buffer;
memberoff = MXOffsetToMemberOffset(offset);
- slotno = SimpleLruReadPage(MultiXactMemberCtl, pageno, true, offset);
- xidptr = (TransactionId *)
- (MultiXactMemberCtl->shared->page_buffer[slotno] + memberoff);
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_MEMBER_REL_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ xidptr = (TransactionId *) (BufferGetPage(buffer) + memberoff);
MemSet(xidptr, 0, BLCKSZ - memberoff);
@@ -2112,11 +2028,10 @@ TrimMultiXact(void)
* writing.
*/
- MultiXactMemberCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
+ UnlockReleaseBuffer(buffer);
}
- LWLockRelease(MultiXactMemberSLRULock);
-
/* signal that we're officially up */
LWLockAcquire(MultiXactGenLock, LW_EXCLUSIVE);
MultiXactState->finishedStartup = true;
@@ -2148,25 +2063,6 @@ MultiXactGetCheckptMulti(bool is_shutdown,
*nextMulti, *nextMultiOffset, *oldestMulti, *oldestMultiDB);
}
-/*
- * Perform a checkpoint --- either during shutdown, or on-the-fly
- */
-void
-CheckPointMultiXact(void)
-{
- TRACE_POSTGRESQL_MULTIXACT_CHECKPOINT_START(true);
-
- /*
- * Write dirty MultiXact pages to disk. This may result in sync requests
- * queued for later handling by ProcessSyncRequests(), as part of the
- * checkpoint.
- */
- SimpleLruWriteAll(MultiXactOffsetCtl, true);
- SimpleLruWriteAll(MultiXactMemberCtl, true);
-
- TRACE_POSTGRESQL_MULTIXACT_CHECKPOINT_DONE(true);
-}
-
/*
* Set the next-to-be-assigned MultiXactId and offset
*
@@ -2415,12 +2311,8 @@ ExtendMultiXactOffset(MultiXactId multi)
pageno = MultiXactIdToOffsetPage(multi);
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
-
/* Zero the page and make an XLOG entry about it */
- ZeroMultiXactOffsetPage(pageno, true);
-
- LWLockRelease(MultiXactOffsetSLRULock);
+ UnlockReleaseBuffer(ZeroMultiXactOffsetPage(pageno, true));
}
/*
@@ -2456,12 +2348,8 @@ ExtendMultiXactMember(MultiXactOffset offset, int nmembers)
pageno = MXOffsetToMemberPage(offset);
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
-
/* Zero the page and make an XLOG entry about it */
- ZeroMultiXactMemberPage(pageno, true);
-
- LWLockRelease(MultiXactMemberSLRULock);
+ UnlockReleaseBuffer(ZeroMultiXactMemberPage(pageno, true));
}
/*
@@ -2736,8 +2624,9 @@ find_multixact_start(MultiXactId multi, MultiXactOffset *result)
MultiXactOffset offset;
int pageno;
int entryno;
- int slotno;
MultiXactOffset *offptr;
+ Buffer buffer;
+ RelFileNode rnode = SlruRelFileNode(SLRU_MULTIXACT_OFFSET_REL_ID);
Assert(MultiXactState->finishedStartup);
@@ -2745,20 +2634,19 @@ find_multixact_start(MultiXactId multi, MultiXactOffset *result)
entryno = MultiXactIdToOffsetEntry(multi);
/*
- * Write out dirty data, so PhysicalPageExists can work correctly.
+ * Cope with missing/bogus oldest MultiXact in inconsistent states (see
+ * commit 068cfadf9).
*/
- SimpleLruWriteAll(MultiXactOffsetCtl, true);
- SimpleLruWriteAll(MultiXactMemberCtl, true);
-
- if (!SimpleLruDoesPhysicalPageExist(MultiXactOffsetCtl, pageno))
+ if (!BufferProbe(rnode, MAIN_FORKNUM, pageno) &&
+ !SimpleLruDoesPhysicalPageExist(SLRU_MULTIXACT_OFFSET_REL_ID, pageno))
return false;
- /* lock is acquired by SimpleLruReadPage_ReadOnly */
- slotno = SimpleLruReadPage_ReadOnly(MultiXactOffsetCtl, pageno, multi);
- offptr = (MultiXactOffset *) MultiXactOffsetCtl->shared->page_buffer[slotno];
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_OFFSET_REL_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_SHARE);
+ offptr = (MultiXactOffset *) BufferGetPage(buffer);
offptr += entryno;
offset = *offptr;
- LWLockRelease(MultiXactOffsetSLRULock);
+ UnlockReleaseBuffer(buffer);
*result = offset;
return true;
@@ -2865,12 +2753,13 @@ typedef struct mxtruncinfo
* This callback determines the earliest existing page number.
*/
static bool
-SlruScanDirCbFindEarliest(SlruCtl ctl, char *filename, int segpage, void *data)
+SlruScanDirCbFindEarliest(Oid rel_id, SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage, void *data)
{
mxtruncinfo *trunc = (mxtruncinfo *) data;
if (trunc->earliestExistingPage == -1 ||
- ctl->PagePrecedes(segpage, trunc->earliestExistingPage))
+ PagePrecedes(segpage, trunc->earliestExistingPage))
{
trunc->earliestExistingPage = segpage;
}
@@ -2902,7 +2791,7 @@ PerformMembersTruncation(MultiXactOffset oldestOffset, MultiXactOffset newOldest
while (segment != endsegment)
{
elog(DEBUG2, "truncating multixact members segment %x", segment);
- SlruDeleteSegment(MultiXactMemberCtl, segment);
+ SlruDeleteSegment(SLRU_MULTIXACT_MEMBER_REL_ID, segment);
/* move to next segment, handling wraparound correctly */
if (segment == maxsegment)
@@ -2925,7 +2814,8 @@ PerformOffsetsTruncation(MultiXactId oldestMulti, MultiXactId newOldestMulti)
* didn't subtract one, we'd trigger SimpleLruTruncate's wraparound
* detection.
*/
- SimpleLruTruncate(MultiXactOffsetCtl,
+ SimpleLruTruncate(SLRU_MULTIXACT_OFFSET_REL_ID,
+ MultiXactOffsetPagePrecedes,
MultiXactIdToOffsetPage(PreviousMultiXactId(newOldestMulti)));
}
@@ -2999,7 +2889,9 @@ TruncateMultiXact(MultiXactId newOldestMulti, Oid newOldestMultiDB)
* been truncated away, and we crashed before updating oldestMulti.
*/
trunc.earliestExistingPage = -1;
- SlruScanDirectory(MultiXactOffsetCtl, SlruScanDirCbFindEarliest, &trunc);
+ SlruScanDirectory(SLRU_MULTIXACT_OFFSET_REL_ID,
+ MultiXactOffsetPagePrecedes,
+ SlruScanDirCbFindEarliest, &trunc);
earliest = trunc.earliestExistingPage * MULTIXACT_OFFSETS_PER_PAGE;
if (earliest < FirstMultiXactId)
earliest = FirstMultiXactId;
@@ -3131,24 +3023,6 @@ MultiXactOffsetPagePrecedes(int page1, int page2)
multi2 + MULTIXACT_OFFSETS_PER_PAGE - 1));
}
-/*
- * Decide whether a MultiXactMember page number is "older" for truncation
- * purposes. There is no "invalid offset number" so use the numbers verbatim.
- */
-static bool
-MultiXactMemberPagePrecedes(int page1, int page2)
-{
- MultiXactOffset offset1;
- MultiXactOffset offset2;
-
- offset1 = ((MultiXactOffset) page1) * MULTIXACT_MEMBERS_PER_PAGE;
- offset2 = ((MultiXactOffset) page2) * MULTIXACT_MEMBERS_PER_PAGE;
-
- return (MultiXactOffsetPrecedes(offset1, offset2) &&
- MultiXactOffsetPrecedes(offset1,
- offset2 + MULTIXACT_MEMBERS_PER_PAGE - 1));
-}
-
/*
* Decide which of two MultiXactIds is earlier.
*
@@ -3243,32 +3117,18 @@ multixact_redo(XLogReaderState *record)
if (info == XLOG_MULTIXACT_ZERO_OFF_PAGE)
{
int pageno;
- int slotno;
memcpy(&pageno, XLogRecGetData(record), sizeof(int));
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
-
- slotno = ZeroMultiXactOffsetPage(pageno, false);
- SimpleLruWritePage(MultiXactOffsetCtl, slotno);
- Assert(!MultiXactOffsetCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(MultiXactOffsetSLRULock);
+ UnlockReleaseBuffer(ZeroMultiXactOffsetPage(pageno, false));
}
else if (info == XLOG_MULTIXACT_ZERO_MEM_PAGE)
{
int pageno;
- int slotno;
memcpy(&pageno, XLogRecGetData(record), sizeof(int));
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
-
- slotno = ZeroMultiXactMemberPage(pageno, false);
- SimpleLruWritePage(MultiXactMemberCtl, slotno);
- Assert(!MultiXactMemberCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(MultiXactMemberSLRULock);
+ UnlockReleaseBuffer(ZeroMultiXactMemberPage(pageno, false));
}
else if (info == XLOG_MULTIXACT_CREATE_ID)
{
@@ -3302,7 +3162,6 @@ multixact_redo(XLogReaderState *record)
else if (info == XLOG_MULTIXACT_TRUNCATE_ID)
{
xl_multixact_truncate xlrec;
- int pageno;
memcpy(&xlrec, XLogRecGetData(record),
SizeOfMultiXactTruncate);
@@ -3328,13 +3187,6 @@ multixact_redo(XLogReaderState *record)
PerformMembersTruncation(xlrec.startTruncMemb, xlrec.endTruncMemb);
- /*
- * During XLOG replay, latest_page_number isn't necessarily set up
- * yet; insert a suitable value to bypass the sanity test in
- * SimpleLruTruncate.
- */
- pageno = MultiXactIdToOffsetPage(xlrec.endTruncOff);
- MultiXactOffsetCtl->shared->latest_page_number = pageno;
PerformOffsetsTruncation(xlrec.startTruncOff, xlrec.endTruncOff);
LWLockRelease(MultiXactTruncationLock);
@@ -3407,21 +3259,3 @@ pg_get_multixact_members(PG_FUNCTION_ARGS)
SRF_RETURN_DONE(funccxt);
}
-
-/*
- * Entrypoint for sync.c to sync offsets files.
- */
-int
-multixactoffsetssyncfiletag(const FileTag *ftag, char *path)
-{
- return SlruSyncFileTag(MultiXactOffsetCtl, ftag, path);
-}
-
-/*
- * Entrypoint for sync.c to sync members files.
- */
-int
-multixactmemberssyncfiletag(const FileTag *ftag, char *path)
-{
- return SlruSyncFileTag(MultiXactMemberCtl, ftag, path);
-}
diff --git a/src/backend/access/transam/slru.c b/src/backend/access/transam/slru.c
index b65cb49d7f..241f34b4bd 100644
--- a/src/backend/access/transam/slru.c
+++ b/src/backend/access/transam/slru.c
@@ -1,41 +1,9 @@
/*-------------------------------------------------------------------------
*
* slru.c
- * Simple LRU buffering for transaction status logfiles
+ * Simple buffering for transaction status logfiles
*
- * We use a simple least-recently-used scheme to manage a pool of page
- * buffers. Under ordinary circumstances we expect that write
- * traffic will occur mostly to the latest page (and to the just-prior
- * page, soon after a page transition). Read traffic will probably touch
- * a larger span of pages, but in any case a fairly small number of page
- * buffers should be sufficient. So, we just search the buffers using plain
- * linear search; there's no need for a hashtable or anything fancy.
- * The management algorithm is straight LRU except that we will never swap
- * out the latest page (since we know it's going to be hit again eventually).
- *
- * We use a control LWLock to protect the shared data structures, plus
- * per-buffer LWLocks that synchronize I/O for each buffer. The control lock
- * must be held to examine or modify any shared state. A process that is
- * reading in or writing out a page buffer does not hold the control lock,
- * only the per-buffer lock for the buffer it is working on.
- *
- * "Holding the control lock" means exclusive lock in all cases except for
- * SimpleLruReadPage_ReadOnly(); see comments for SlruRecentlyUsed() for
- * the implications of that.
- *
- * When initiating I/O on a buffer, we acquire the per-buffer lock exclusively
- * before releasing the control lock. The per-buffer lock is released after
- * completing the I/O, re-acquiring the control lock, and updating the shared
- * state. (Deadlock is not possible here, because we never try to initiate
- * I/O when someone else is already doing I/O on the same buffer.)
- * To wait for I/O to complete, release the control lock, acquire the
- * per-buffer lock in shared mode, immediately release the per-buffer lock,
- * reacquire the control lock, and then recheck state (since arbitrary things
- * could have happened while we didn't have the lock).
- *
- * As with the regular buffer manager, it is possible for another process
- * to re-dirty a page that is currently being written out. This is handled
- * by re-setting the page's page_dirty flag.
+ * XXX write me
*
*
* Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group
@@ -60,561 +28,81 @@
#include "storage/fd.h"
#include "storage/shmem.h"
-#define SlruFileName(ctl, path, seg) \
- snprintf(path, MAXPGPATH, "%s/%04X", (ctl)->Dir, seg)
+#define SlruFileName(rel_id, path, seg) \
+ snprintf(path, MAXPGPATH, "%s/%04X", defs[(rel_id)].path, seg)
+
+struct SlruDef {
+ const char *name;
+ const char *path;
+ bool synchronize;
+};
+
+static const struct SlruDef defs[] = {
+ [SLRU_CLOG_REL_ID] = {
+ .name = "Xact",
+ .path = "pg_xact",
+ .synchronize = true,
+ },
+ [SLRU_SUBTRANS_REL_ID] = {
+ .name = "Subtrans",
+ .path = "pg_subtrans",
+ },
+ [SLRU_MULTIXACT_OFFSET_REL_ID] = {
+ .name = "MultiXactOffset",
+ .path = "pg_multixact/offsets",
+ .synchronize = true,
+ },
+ [SLRU_MULTIXACT_MEMBER_REL_ID] = {
+ .name = "MultiXactMember",
+ .path = "pg_multixact/members",
+ .synchronize = true,
+ },
+ [SLRU_COMMITTS_REL_ID] = {
+ .name = "CommitTs",
+ .path = "pg_commit_ts",
+ .synchronize = true,
+ },
+ [SLRU_SERIAL_REL_ID] = {
+ .name = "Serial",
+ .path = "pg_serial",
+ },
+ [SLRU_NOTIFY_REL_ID] = {
+ .name = "Notify",
+ .path = "pg_notify",
+ },
+};
/*
- * During SimpleLruWriteAll(), we will usually not need to write more than one
- * or two physical files, but we may need to write several pages per file. We
- * can consolidate the I/O requests by leaving files open until control returns
- * to SimpleLruWriteAll(). This data structure remembers which files are open.
+ * We'll maintain a little cache of recently seen buffers, to try to avoid the
+ * buffer mapping table on repeat access (ie the busy end of the CLOG). One
+ * entry per SLRU relation.
*/
-#define MAX_WRITEALL_BUFFERS 16
+struct SlruRecentBuffer {
+ int pageno;
+ Buffer recent_buffer;
+};
-typedef struct SlruWriteAllData
-{
- int num_files; /* # files actually open */
- int fd[MAX_WRITEALL_BUFFERS]; /* their FD's */
- int segno[MAX_WRITEALL_BUFFERS]; /* their log seg#s */
-} SlruWriteAllData;
-
-typedef struct SlruWriteAllData *SlruWriteAll;
+static struct SlruRecentBuffer slru_recent_buffers[lengthof(defs)];
/*
- * Populate a file tag describing a segment file. We only use the segment
- * number, since we can derive everything else we need by having separate
- * sync handler functions for clog, multixact etc.
+ * Populate a file tag identifying an SLRU segment file.
*/
-#define INIT_SLRUFILETAG(a,xx_handler,xx_segno) \
+#define INIT_SLRUFILETAG(a,xx_rel_id,xx_segno) \
( \
memset(&(a), 0, sizeof(FileTag)), \
- (a).handler = (xx_handler), \
+ (a).handler = SYNC_HANDLER_SLRU, \
+ (a).rnode = SlruRelFileNode(xx_rel_id), \
(a).segno = (xx_segno) \
)
-/*
- * Macro to mark a buffer slot "most recently used". Note multiple evaluation
- * of arguments!
- *
- * The reason for the if-test is that there are often many consecutive
- * accesses to the same page (particularly the latest page). By suppressing
- * useless increments of cur_lru_count, we reduce the probability that old
- * pages' counts will "wrap around" and make them appear recently used.
- *
- * We allow this code to be executed concurrently by multiple processes within
- * SimpleLruReadPage_ReadOnly(). As long as int reads and writes are atomic,
- * this should not cause any completely-bogus values to enter the computation.
- * However, it is possible for either cur_lru_count or individual
- * page_lru_count entries to be "reset" to lower values than they should have,
- * in case a process is delayed while it executes this macro. With care in
- * SlruSelectLRUPage(), this does little harm, and in any case the absolute
- * worst possible consequence is a nonoptimal choice of page to evict. The
- * gain from allowing concurrent reads of SLRU pages seems worth it.
- */
-#define SlruRecentlyUsed(shared, slotno) \
- do { \
- int new_lru_count = (shared)->cur_lru_count; \
- if (new_lru_count != (shared)->page_lru_count[slotno]) { \
- (shared)->cur_lru_count = ++new_lru_count; \
- (shared)->page_lru_count[slotno] = new_lru_count; \
- } \
- } while (0)
-
-/* Saved info for SlruReportIOError */
-typedef enum
-{
- SLRU_OPEN_FAILED,
- SLRU_SEEK_FAILED,
- SLRU_READ_FAILED,
- SLRU_WRITE_FAILED,
- SLRU_FSYNC_FAILED,
- SLRU_CLOSE_FAILED
-} SlruErrorCause;
-
-static SlruErrorCause slru_errcause;
-static int slru_errno;
-
-
-static void SimpleLruZeroLSNs(SlruCtl ctl, int slotno);
-static void SimpleLruWaitIO(SlruCtl ctl, int slotno);
-static void SlruInternalWritePage(SlruCtl ctl, int slotno, SlruWriteAll fdata);
-static bool SlruPhysicalReadPage(SlruCtl ctl, int pageno, int slotno);
-static bool SlruPhysicalWritePage(SlruCtl ctl, int pageno, int slotno,
- SlruWriteAll fdata);
-static void SlruReportIOError(SlruCtl ctl, int pageno, TransactionId xid);
-static int SlruSelectLRUPage(SlruCtl ctl, int pageno);
-
-static bool SlruScanDirCbDeleteCutoff(SlruCtl ctl, char *filename,
+static bool SlruScanDirCbDeleteCutoff(Oid rel_id,
+ SlruPagePrecedesFunction PagePrecedes,
+ char *filename,
int segpage, void *data);
-static void SlruInternalDeleteSegment(SlruCtl ctl, int segno);
-
-/*
- * Initialization of shared memory
- */
-
-Size
-SimpleLruShmemSize(int nslots, int nlsns)
-{
- Size sz;
-
- /* we assume nslots isn't so large as to risk overflow */
- sz = MAXALIGN(sizeof(SlruSharedData));
- sz += MAXALIGN(nslots * sizeof(char *)); /* page_buffer[] */
- sz += MAXALIGN(nslots * sizeof(SlruPageStatus)); /* page_status[] */
- sz += MAXALIGN(nslots * sizeof(bool)); /* page_dirty[] */
- sz += MAXALIGN(nslots * sizeof(int)); /* page_number[] */
- sz += MAXALIGN(nslots * sizeof(int)); /* page_lru_count[] */
- sz += MAXALIGN(nslots * sizeof(LWLockPadded)); /* buffer_locks[] */
-
- if (nlsns > 0)
- sz += MAXALIGN(nslots * nlsns * sizeof(XLogRecPtr)); /* group_lsn[] */
-
- return BUFFERALIGN(sz) + BLCKSZ * nslots;
-}
-
-/*
- * Initialize, or attach to, a simple LRU cache in shared memory.
- *
- * ctl: address of local (unshared) control structure.
- * name: name of SLRU. (This is user-visible, pick with care!)
- * nslots: number of page slots to use.
- * nlsns: number of LSN groups per page (set to zero if not relevant).
- * ctllock: LWLock to use to control access to the shared control structure.
- * subdir: PGDATA-relative subdirectory that will contain the files.
- * tranche_id: LWLock tranche ID to use for the SLRU's per-buffer LWLocks.
- */
-void
-SimpleLruInit(SlruCtl ctl, const char *name, int nslots, int nlsns,
- LWLock *ctllock, const char *subdir, int tranche_id,
- SyncRequestHandler sync_handler)
-{
- SlruShared shared;
- bool found;
-
- shared = (SlruShared) ShmemInitStruct(name,
- SimpleLruShmemSize(nslots, nlsns),
- &found);
-
- if (!IsUnderPostmaster)
- {
- /* Initialize locks and shared memory area */
- char *ptr;
- Size offset;
- int slotno;
-
- Assert(!found);
-
- memset(shared, 0, sizeof(SlruSharedData));
-
- shared->ControlLock = ctllock;
-
- shared->num_slots = nslots;
- shared->lsn_groups_per_page = nlsns;
-
- shared->cur_lru_count = 0;
-
- /* shared->latest_page_number will be set later */
-
- shared->slru_stats_idx = pgstat_get_slru_index(name);
-
- ptr = (char *) shared;
- offset = MAXALIGN(sizeof(SlruSharedData));
- shared->page_buffer = (char **) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(char *));
- shared->page_status = (SlruPageStatus *) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(SlruPageStatus));
- shared->page_dirty = (bool *) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(bool));
- shared->page_number = (int *) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(int));
- shared->page_lru_count = (int *) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(int));
-
- /* Initialize LWLocks */
- shared->buffer_locks = (LWLockPadded *) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(LWLockPadded));
-
- if (nlsns > 0)
- {
- shared->group_lsn = (XLogRecPtr *) (ptr + offset);
- offset += MAXALIGN(nslots * nlsns * sizeof(XLogRecPtr));
- }
-
- ptr += BUFFERALIGN(offset);
- for (slotno = 0; slotno < nslots; slotno++)
- {
- LWLockInitialize(&shared->buffer_locks[slotno].lock,
- tranche_id);
-
- shared->page_buffer[slotno] = ptr;
- shared->page_status[slotno] = SLRU_PAGE_EMPTY;
- shared->page_dirty[slotno] = false;
- shared->page_lru_count[slotno] = 0;
- ptr += BLCKSZ;
- }
-
- /* Should fit to estimated shmem size */
- Assert(ptr - (char *) shared <= SimpleLruShmemSize(nslots, nlsns));
- }
- else
- Assert(found);
-
- /*
- * Initialize the unshared control struct, including directory path. We
- * assume caller set PagePrecedes.
- */
- ctl->shared = shared;
- ctl->sync_handler = sync_handler;
- strlcpy(ctl->Dir, subdir, sizeof(ctl->Dir));
-}
-
-/*
- * Initialize (or reinitialize) a page to zeroes.
- *
- * The page is not actually written, just set up in shared memory.
- * The slot number of the new page is returned.
- *
- * Control lock must be held at entry, and will be held at exit.
- */
-int
-SimpleLruZeroPage(SlruCtl ctl, int pageno)
-{
- SlruShared shared = ctl->shared;
- int slotno;
-
- /* Find a suitable buffer slot for the page */
- slotno = SlruSelectLRUPage(ctl, pageno);
- Assert(shared->page_status[slotno] == SLRU_PAGE_EMPTY ||
- (shared->page_status[slotno] == SLRU_PAGE_VALID &&
- !shared->page_dirty[slotno]) ||
- shared->page_number[slotno] == pageno);
-
- /* Mark the slot as containing this page */
- shared->page_number[slotno] = pageno;
- shared->page_status[slotno] = SLRU_PAGE_VALID;
- shared->page_dirty[slotno] = true;
- SlruRecentlyUsed(shared, slotno);
-
- /* Set the buffer to zeroes */
- MemSet(shared->page_buffer[slotno], 0, BLCKSZ);
-
- /* Set the LSNs for this new page to zero */
- SimpleLruZeroLSNs(ctl, slotno);
-
- /* Assume this page is now the latest active page */
- shared->latest_page_number = pageno;
-
- /* update the stats counter of zeroed pages */
- pgstat_count_slru_page_zeroed(shared->slru_stats_idx);
-
- return slotno;
-}
-
-/*
- * Zero all the LSNs we store for this slru page.
- *
- * This should be called each time we create a new page, and each time we read
- * in a page from disk into an existing buffer. (Such an old page cannot
- * have any interesting LSNs, since we'd have flushed them before writing
- * the page in the first place.)
- *
- * This assumes that InvalidXLogRecPtr is bitwise-all-0.
- */
-static void
-SimpleLruZeroLSNs(SlruCtl ctl, int slotno)
-{
- SlruShared shared = ctl->shared;
-
- if (shared->lsn_groups_per_page > 0)
- MemSet(&shared->group_lsn[slotno * shared->lsn_groups_per_page], 0,
- shared->lsn_groups_per_page * sizeof(XLogRecPtr));
-}
-
-/*
- * Wait for any active I/O on a page slot to finish. (This does not
- * guarantee that new I/O hasn't been started before we return, though.
- * In fact the slot might not even contain the same page anymore.)
- *
- * Control lock must be held at entry, and will be held at exit.
- */
-static void
-SimpleLruWaitIO(SlruCtl ctl, int slotno)
-{
- SlruShared shared = ctl->shared;
-
- /* See notes at top of file */
- LWLockRelease(shared->ControlLock);
- LWLockAcquire(&shared->buffer_locks[slotno].lock, LW_SHARED);
- LWLockRelease(&shared->buffer_locks[slotno].lock);
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
- /*
- * If the slot is still in an io-in-progress state, then either someone
- * already started a new I/O on the slot, or a previous I/O failed and
- * neglected to reset the page state. That shouldn't happen, really, but
- * it seems worth a few extra cycles to check and recover from it. We can
- * cheaply test for failure by seeing if the buffer lock is still held (we
- * assume that transaction abort would release the lock).
- */
- if (shared->page_status[slotno] == SLRU_PAGE_READ_IN_PROGRESS ||
- shared->page_status[slotno] == SLRU_PAGE_WRITE_IN_PROGRESS)
- {
- if (LWLockConditionalAcquire(&shared->buffer_locks[slotno].lock, LW_SHARED))
- {
- /* indeed, the I/O must have failed */
- if (shared->page_status[slotno] == SLRU_PAGE_READ_IN_PROGRESS)
- shared->page_status[slotno] = SLRU_PAGE_EMPTY;
- else /* write_in_progress */
- {
- shared->page_status[slotno] = SLRU_PAGE_VALID;
- shared->page_dirty[slotno] = true;
- }
- LWLockRelease(&shared->buffer_locks[slotno].lock);
- }
- }
-}
-
-/*
- * Find a page in a shared buffer, reading it in if necessary.
- * The page number must correspond to an already-initialized page.
- *
- * If write_ok is true then it is OK to return a page that is in
- * WRITE_IN_PROGRESS state; it is the caller's responsibility to be sure
- * that modification of the page is safe. If write_ok is false then we
- * will not return the page until it is not undergoing active I/O.
- *
- * The passed-in xid is used only for error reporting, and may be
- * InvalidTransactionId if no specific xid is associated with the action.
- *
- * Return value is the shared-buffer slot number now holding the page.
- * The buffer's LRU access info is updated.
- *
- * Control lock must be held at entry, and will be held at exit.
- */
-int
-SimpleLruReadPage(SlruCtl ctl, int pageno, bool write_ok,
- TransactionId xid)
-{
- SlruShared shared = ctl->shared;
-
- /* Outer loop handles restart if we must wait for someone else's I/O */
- for (;;)
- {
- int slotno;
- bool ok;
-
- /* See if page already is in memory; if not, pick victim slot */
- slotno = SlruSelectLRUPage(ctl, pageno);
-
- /* Did we find the page in memory? */
- if (shared->page_number[slotno] == pageno &&
- shared->page_status[slotno] != SLRU_PAGE_EMPTY)
- {
- /*
- * If page is still being read in, we must wait for I/O. Likewise
- * if the page is being written and the caller said that's not OK.
- */
- if (shared->page_status[slotno] == SLRU_PAGE_READ_IN_PROGRESS ||
- (shared->page_status[slotno] == SLRU_PAGE_WRITE_IN_PROGRESS &&
- !write_ok))
- {
- SimpleLruWaitIO(ctl, slotno);
- /* Now we must recheck state from the top */
- continue;
- }
- /* Otherwise, it's ready to use */
- SlruRecentlyUsed(shared, slotno);
-
- /* update the stats counter of pages found in the SLRU */
- pgstat_count_slru_page_hit(shared->slru_stats_idx);
-
- return slotno;
- }
-
- /* We found no match; assert we selected a freeable slot */
- Assert(shared->page_status[slotno] == SLRU_PAGE_EMPTY ||
- (shared->page_status[slotno] == SLRU_PAGE_VALID &&
- !shared->page_dirty[slotno]));
-
- /* Mark the slot read-busy */
- shared->page_number[slotno] = pageno;
- shared->page_status[slotno] = SLRU_PAGE_READ_IN_PROGRESS;
- shared->page_dirty[slotno] = false;
-
- /* Acquire per-buffer lock (cannot deadlock, see notes at top) */
- LWLockAcquire(&shared->buffer_locks[slotno].lock, LW_EXCLUSIVE);
-
- /* Release control lock while doing I/O */
- LWLockRelease(shared->ControlLock);
-
- /* Do the read */
- ok = SlruPhysicalReadPage(ctl, pageno, slotno);
-
- /* Set the LSNs for this newly read-in page to zero */
- SimpleLruZeroLSNs(ctl, slotno);
-
- /* Re-acquire control lock and update page state */
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
- Assert(shared->page_number[slotno] == pageno &&
- shared->page_status[slotno] == SLRU_PAGE_READ_IN_PROGRESS &&
- !shared->page_dirty[slotno]);
-
- shared->page_status[slotno] = ok ? SLRU_PAGE_VALID : SLRU_PAGE_EMPTY;
+static void SlruInternalDeleteSegment(Oid rel_id, int segno);
- LWLockRelease(&shared->buffer_locks[slotno].lock);
-
- /* Now it's okay to ereport if we failed */
- if (!ok)
- SlruReportIOError(ctl, pageno, xid);
-
- SlruRecentlyUsed(shared, slotno);
-
- /* update the stats counter of pages not found in SLRU */
- pgstat_count_slru_page_read(shared->slru_stats_idx);
-
- return slotno;
- }
-}
-
-/*
- * Find a page in a shared buffer, reading it in if necessary.
- * The page number must correspond to an already-initialized page.
- * The caller must intend only read-only access to the page.
- *
- * The passed-in xid is used only for error reporting, and may be
- * InvalidTransactionId if no specific xid is associated with the action.
- *
- * Return value is the shared-buffer slot number now holding the page.
- * The buffer's LRU access info is updated.
- *
- * Control lock must NOT be held at entry, but will be held at exit.
- * It is unspecified whether the lock will be shared or exclusive.
- */
-int
-SimpleLruReadPage_ReadOnly(SlruCtl ctl, int pageno, TransactionId xid)
-{
- SlruShared shared = ctl->shared;
- int slotno;
-
- /* Try to find the page while holding only shared lock */
- LWLockAcquire(shared->ControlLock, LW_SHARED);
-
- /* See if page is already in a buffer */
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- if (shared->page_number[slotno] == pageno &&
- shared->page_status[slotno] != SLRU_PAGE_EMPTY &&
- shared->page_status[slotno] != SLRU_PAGE_READ_IN_PROGRESS)
- {
- /* See comments for SlruRecentlyUsed macro */
- SlruRecentlyUsed(shared, slotno);
-
- /* update the stats counter of pages found in the SLRU */
- pgstat_count_slru_page_hit(shared->slru_stats_idx);
-
- return slotno;
- }
- }
-
- /* No luck, so switch to normal exclusive lock and do regular read */
- LWLockRelease(shared->ControlLock);
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
- return SimpleLruReadPage(ctl, pageno, true, xid);
-}
-
-/*
- * Write a page from a shared buffer, if necessary.
- * Does nothing if the specified slot is not dirty.
- *
- * NOTE: only one write attempt is made here. Hence, it is possible that
- * the page is still dirty at exit (if someone else re-dirtied it during
- * the write). However, we *do* attempt a fresh write even if the page
- * is already being written; this is for checkpoints.
- *
- * Control lock must be held at entry, and will be held at exit.
- */
-static void
-SlruInternalWritePage(SlruCtl ctl, int slotno, SlruWriteAll fdata)
-{
- SlruShared shared = ctl->shared;
- int pageno = shared->page_number[slotno];
- bool ok;
-
- /* If a write is in progress, wait for it to finish */
- while (shared->page_status[slotno] == SLRU_PAGE_WRITE_IN_PROGRESS &&
- shared->page_number[slotno] == pageno)
- {
- SimpleLruWaitIO(ctl, slotno);
- }
-
- /*
- * Do nothing if page is not dirty, or if buffer no longer contains the
- * same page we were called for.
- */
- if (!shared->page_dirty[slotno] ||
- shared->page_status[slotno] != SLRU_PAGE_VALID ||
- shared->page_number[slotno] != pageno)
- return;
-
- /*
- * Mark the slot write-busy, and clear the dirtybit. After this point, a
- * transaction status update on this page will mark it dirty again.
- */
- shared->page_status[slotno] = SLRU_PAGE_WRITE_IN_PROGRESS;
- shared->page_dirty[slotno] = false;
-
- /* Acquire per-buffer lock (cannot deadlock, see notes at top) */
- LWLockAcquire(&shared->buffer_locks[slotno].lock, LW_EXCLUSIVE);
-
- /* Release control lock while doing I/O */
- LWLockRelease(shared->ControlLock);
-
- /* Do the write */
- ok = SlruPhysicalWritePage(ctl, pageno, slotno, fdata);
-
- /* If we failed, and we're in a flush, better close the files */
- if (!ok && fdata)
- {
- int i;
-
- for (i = 0; i < fdata->num_files; i++)
- CloseTransientFile(fdata->fd[i]);
- }
-
- /* Re-acquire control lock and update page state */
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
- Assert(shared->page_number[slotno] == pageno &&
- shared->page_status[slotno] == SLRU_PAGE_WRITE_IN_PROGRESS);
-
- /* If we failed to write, mark the page dirty again */
- if (!ok)
- shared->page_dirty[slotno] = true;
-
- shared->page_status[slotno] = SLRU_PAGE_VALID;
-
- LWLockRelease(&shared->buffer_locks[slotno].lock);
-
- /* Now it's okay to ereport if we failed */
- if (!ok)
- SlruReportIOError(ctl, pageno, InvalidTransactionId);
-
- /* If part of a checkpoint, count this as a buffer written. */
- if (fdata)
- CheckpointStats.ckpt_bufs_written++;
-}
-
-/*
- * Wrapper of SlruInternalWritePage, for external callers.
- * fdata is always passed a NULL here.
- */
-void
-SimpleLruWritePage(SlruCtl ctl, int slotno)
-{
- SlruInternalWritePage(ctl, slotno, NULL);
-}
+static File slrufile(SMgrRelation reln, BlockNumber blocknum, int mode,
+ bool missing_ok);
/*
* Return whether the given page exists on disk.
@@ -623,592 +111,30 @@ SimpleLruWritePage(SlruCtl ctl, int slotno)
* large enough to contain the given page.
*/
bool
-SimpleLruDoesPhysicalPageExist(SlruCtl ctl, int pageno)
-{
- int segno = pageno / SLRU_PAGES_PER_SEGMENT;
- int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
- int offset = rpageno * BLCKSZ;
- char path[MAXPGPATH];
- int fd;
- bool result;
- off_t endpos;
-
- /* update the stats counter of checked pages */
- pgstat_count_slru_page_exists(ctl->shared->slru_stats_idx);
-
- SlruFileName(ctl, path, segno);
-
- fd = OpenTransientFile(path, O_RDONLY | PG_BINARY);
- if (fd < 0)
- {
- /* expected: file doesn't exist */
- if (errno == ENOENT)
- return false;
-
- /* report error normally */
- slru_errcause = SLRU_OPEN_FAILED;
- slru_errno = errno;
- SlruReportIOError(ctl, pageno, 0);
- }
-
- if ((endpos = lseek(fd, 0, SEEK_END)) < 0)
- {
- slru_errcause = SLRU_SEEK_FAILED;
- slru_errno = errno;
- SlruReportIOError(ctl, pageno, 0);
- }
-
- result = endpos >= (off_t) (offset + BLCKSZ);
-
- if (CloseTransientFile(fd) != 0)
- {
- slru_errcause = SLRU_CLOSE_FAILED;
- slru_errno = errno;
- return false;
- }
-
- return result;
-}
-
-/*
- * Physical read of a (previously existing) page into a buffer slot
- *
- * On failure, we cannot just ereport(ERROR) since caller has put state in
- * shared memory that must be undone. So, we return false and save enough
- * info in static variables to let SlruReportIOError make the report.
- *
- * For now, assume it's not worth keeping a file pointer open across
- * read/write operations. We could cache one virtual file pointer ...
- */
-static bool
-SlruPhysicalReadPage(SlruCtl ctl, int pageno, int slotno)
-{
- SlruShared shared = ctl->shared;
- int segno = pageno / SLRU_PAGES_PER_SEGMENT;
- int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
- off_t offset = rpageno * BLCKSZ;
- char path[MAXPGPATH];
- int fd;
-
- SlruFileName(ctl, path, segno);
-
- /*
- * In a crash-and-restart situation, it's possible for us to receive
- * commands to set the commit status of transactions whose bits are in
- * already-truncated segments of the commit log (see notes in
- * SlruPhysicalWritePage). Hence, if we are InRecovery, allow the case
- * where the file doesn't exist, and return zeroes instead.
- */
- fd = OpenTransientFile(path, O_RDONLY | PG_BINARY);
- if (fd < 0)
- {
- if (errno != ENOENT || !InRecovery)
- {
- slru_errcause = SLRU_OPEN_FAILED;
- slru_errno = errno;
- return false;
- }
-
- ereport(LOG,
- (errmsg("file \"%s\" doesn't exist, reading as zeroes",
- path)));
- MemSet(shared->page_buffer[slotno], 0, BLCKSZ);
- return true;
- }
-
- errno = 0;
- pgstat_report_wait_start(WAIT_EVENT_SLRU_READ);
- if (pg_pread(fd, shared->page_buffer[slotno], BLCKSZ, offset) != BLCKSZ)
- {
- pgstat_report_wait_end();
- slru_errcause = SLRU_READ_FAILED;
- slru_errno = errno;
- CloseTransientFile(fd);
- return false;
- }
- pgstat_report_wait_end();
-
- if (CloseTransientFile(fd) != 0)
- {
- slru_errcause = SLRU_CLOSE_FAILED;
- slru_errno = errno;
- return false;
- }
-
- return true;
-}
-
-/*
- * Physical write of a page from a buffer slot
- *
- * On failure, we cannot just ereport(ERROR) since caller has put state in
- * shared memory that must be undone. So, we return false and save enough
- * info in static variables to let SlruReportIOError make the report.
- *
- * For now, assume it's not worth keeping a file pointer open across
- * independent read/write operations. We do batch operations during
- * SimpleLruWriteAll, though.
- *
- * fdata is NULL for a standalone write, pointer to open-file info during
- * SimpleLruWriteAll.
- */
-static bool
-SlruPhysicalWritePage(SlruCtl ctl, int pageno, int slotno, SlruWriteAll fdata)
+SimpleLruDoesPhysicalPageExist(Oid rel_id, int pageno)
{
- SlruShared shared = ctl->shared;
- int segno = pageno / SLRU_PAGES_PER_SEGMENT;
int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
off_t offset = rpageno * BLCKSZ;
- char path[MAXPGPATH];
- int fd = -1;
-
- /* update the stats counter of written pages */
- pgstat_count_slru_page_written(shared->slru_stats_idx);
-
- /*
- * Honor the write-WAL-before-data rule, if appropriate, so that we do not
- * write out data before associated WAL records. This is the same action
- * performed during FlushBuffer() in the main buffer manager.
- */
- if (shared->group_lsn != NULL)
- {
- /*
- * We must determine the largest async-commit LSN for the page. This
- * is a bit tedious, but since this entire function is a slow path
- * anyway, it seems better to do this here than to maintain a per-page
- * LSN variable (which'd need an extra comparison in the
- * transaction-commit path).
- */
- XLogRecPtr max_lsn;
- int lsnindex,
- lsnoff;
-
- lsnindex = slotno * shared->lsn_groups_per_page;
- max_lsn = shared->group_lsn[lsnindex++];
- for (lsnoff = 1; lsnoff < shared->lsn_groups_per_page; lsnoff++)
- {
- XLogRecPtr this_lsn = shared->group_lsn[lsnindex++];
-
- if (max_lsn < this_lsn)
- max_lsn = this_lsn;
- }
+ off_t size;
+ File file;
+ RelFileNode rnode = SlruRelFileNode(rel_id);
+ SMgrRelation reln = smgropen(rnode, InvalidBackendId);
- if (!XLogRecPtrIsInvalid(max_lsn))
- {
- /*
- * As noted above, elog(ERROR) is not acceptable here, so if
- * XLogFlush were to fail, we must PANIC. This isn't much of a
- * restriction because XLogFlush is just about all critical
- * section anyway, but let's make sure.
- */
- START_CRIT_SECTION();
- XLogFlush(max_lsn);
- END_CRIT_SECTION();
- }
- }
-
- /*
- * During a WriteAll, we may already have the desired file open.
- */
- if (fdata)
- {
- int i;
-
- for (i = 0; i < fdata->num_files; i++)
- {
- if (fdata->segno[i] == segno)
- {
- fd = fdata->fd[i];
- break;
- }
- }
- }
-
- if (fd < 0)
- {
- /*
- * If the file doesn't already exist, we should create it. It is
- * possible for this to need to happen when writing a page that's not
- * first in its segment; we assume the OS can cope with that. (Note:
- * it might seem that it'd be okay to create files only when
- * SimpleLruZeroPage is called for the first page of a segment.
- * However, if after a crash and restart the REDO logic elects to
- * replay the log from a checkpoint before the latest one, then it's
- * possible that we will get commands to set transaction status of
- * transactions that have already been truncated from the commit log.
- * Easiest way to deal with that is to accept references to
- * nonexistent files here and in SlruPhysicalReadPage.)
- *
- * Note: it is possible for more than one backend to be executing this
- * code simultaneously for different pages of the same file. Hence,
- * don't use O_EXCL or O_TRUNC or anything like that.
- */
- SlruFileName(ctl, path, segno);
- fd = OpenTransientFile(path, O_RDWR | O_CREAT | PG_BINARY);
- if (fd < 0)
- {
- slru_errcause = SLRU_OPEN_FAILED;
- slru_errno = errno;
- return false;
- }
-
- if (fdata)
- {
- if (fdata->num_files < MAX_WRITEALL_BUFFERS)
- {
- fdata->fd[fdata->num_files] = fd;
- fdata->segno[fdata->num_files] = segno;
- fdata->num_files++;
- }
- else
- {
- /*
- * In the unlikely event that we exceed MAX_FLUSH_BUFFERS,
- * fall back to treating it as a standalone write.
- */
- fdata = NULL;
- }
- }
- }
+ /* update the stats counter of checked pages */
+ pgstat_count_slru_page_exists(rel_id);
- errno = 0;
- pgstat_report_wait_start(WAIT_EVENT_SLRU_WRITE);
- if (pg_pwrite(fd, shared->page_buffer[slotno], BLCKSZ, offset) != BLCKSZ)
+ file = slrufile(reln, pageno, O_RDWR, true);
+ if (file < 0)
{
- pgstat_report_wait_end();
- /* if write didn't set errno, assume problem is no disk space */
- if (errno == 0)
- errno = ENOSPC;
- slru_errcause = SLRU_WRITE_FAILED;
- slru_errno = errno;
- if (!fdata)
- CloseTransientFile(fd);
+ Assert(errno == ENOENT);
return false;
}
- pgstat_report_wait_end();
-
- /* Queue up a sync request for the checkpointer. */
- if (ctl->sync_handler != SYNC_HANDLER_NONE)
- {
- FileTag tag;
-
- INIT_SLRUFILETAG(tag, ctl->sync_handler, segno);
- if (!RegisterSyncRequest(&tag, SYNC_REQUEST, false))
- {
- /* No space to enqueue sync request. Do it synchronously. */
- pgstat_report_wait_start(WAIT_EVENT_SLRU_SYNC);
- if (pg_fsync(fd) != 0)
- {
- pgstat_report_wait_end();
- slru_errcause = SLRU_FSYNC_FAILED;
- slru_errno = errno;
- CloseTransientFile(fd);
- return false;
- }
- pgstat_report_wait_end();
- }
- }
+ size = FileSize(file);
+ if (size < 0)
+ elog(ERROR, "could not get size of file \"%s\": %m",
+ FilePathName(file));
- /* Close file, unless part of flush request. */
- if (!fdata)
- {
- if (CloseTransientFile(fd) != 0)
- {
- slru_errcause = SLRU_CLOSE_FAILED;
- slru_errno = errno;
- return false;
- }
- }
-
- return true;
-}
-
-/*
- * Issue the error message after failure of SlruPhysicalReadPage or
- * SlruPhysicalWritePage. Call this after cleaning up shared-memory state.
- */
-static void
-SlruReportIOError(SlruCtl ctl, int pageno, TransactionId xid)
-{
- int segno = pageno / SLRU_PAGES_PER_SEGMENT;
- int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
- int offset = rpageno * BLCKSZ;
- char path[MAXPGPATH];
-
- SlruFileName(ctl, path, segno);
- errno = slru_errno;
- switch (slru_errcause)
- {
- case SLRU_OPEN_FAILED:
- ereport(ERROR,
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not open file \"%s\": %m.", path)));
- break;
- case SLRU_SEEK_FAILED:
- ereport(ERROR,
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not seek in file \"%s\" to offset %d: %m.",
- path, offset)));
- break;
- case SLRU_READ_FAILED:
- if (errno)
- ereport(ERROR,
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not read from file \"%s\" at offset %d: %m.",
- path, offset)));
- else
- ereport(ERROR,
- (errmsg("could not access status of transaction %u", xid),
- errdetail("Could not read from file \"%s\" at offset %d: read too few bytes.", path, offset)));
- break;
- case SLRU_WRITE_FAILED:
- if (errno)
- ereport(ERROR,
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not write to file \"%s\" at offset %d: %m.",
- path, offset)));
- else
- ereport(ERROR,
- (errmsg("could not access status of transaction %u", xid),
- errdetail("Could not write to file \"%s\" at offset %d: wrote too few bytes.",
- path, offset)));
- break;
- case SLRU_FSYNC_FAILED:
- ereport(data_sync_elevel(ERROR),
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not fsync file \"%s\": %m.",
- path)));
- break;
- case SLRU_CLOSE_FAILED:
- ereport(ERROR,
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not close file \"%s\": %m.",
- path)));
- break;
- default:
- /* can't get here, we trust */
- elog(ERROR, "unrecognized SimpleLru error cause: %d",
- (int) slru_errcause);
- break;
- }
-}
-
-/*
- * Select the slot to re-use when we need a free slot.
- *
- * The target page number is passed because we need to consider the
- * possibility that some other process reads in the target page while
- * we are doing I/O to free a slot. Hence, check or recheck to see if
- * any slot already holds the target page, and return that slot if so.
- * Thus, the returned slot is *either* a slot already holding the pageno
- * (could be any state except EMPTY), *or* a freeable slot (state EMPTY
- * or CLEAN).
- *
- * Control lock must be held at entry, and will be held at exit.
- */
-static int
-SlruSelectLRUPage(SlruCtl ctl, int pageno)
-{
- SlruShared shared = ctl->shared;
-
- /* Outer loop handles restart after I/O */
- for (;;)
- {
- int slotno;
- int cur_count;
- int bestvalidslot = 0; /* keep compiler quiet */
- int best_valid_delta = -1;
- int best_valid_page_number = 0; /* keep compiler quiet */
- int bestinvalidslot = 0; /* keep compiler quiet */
- int best_invalid_delta = -1;
- int best_invalid_page_number = 0; /* keep compiler quiet */
-
- /* See if page already has a buffer assigned */
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- if (shared->page_number[slotno] == pageno &&
- shared->page_status[slotno] != SLRU_PAGE_EMPTY)
- return slotno;
- }
-
- /*
- * If we find any EMPTY slot, just select that one. Else choose a
- * victim page to replace. We normally take the least recently used
- * valid page, but we will never take the slot containing
- * latest_page_number, even if it appears least recently used. We
- * will select a slot that is already I/O busy only if there is no
- * other choice: a read-busy slot will not be least recently used once
- * the read finishes, and waiting for an I/O on a write-busy slot is
- * inferior to just picking some other slot. Testing shows the slot
- * we pick instead will often be clean, allowing us to begin a read at
- * once.
- *
- * Normally the page_lru_count values will all be different and so
- * there will be a well-defined LRU page. But since we allow
- * concurrent execution of SlruRecentlyUsed() within
- * SimpleLruReadPage_ReadOnly(), it is possible that multiple pages
- * acquire the same lru_count values. In that case we break ties by
- * choosing the furthest-back page.
- *
- * Notice that this next line forcibly advances cur_lru_count to a
- * value that is certainly beyond any value that will be in the
- * page_lru_count array after the loop finishes. This ensures that
- * the next execution of SlruRecentlyUsed will mark the page newly
- * used, even if it's for a page that has the current counter value.
- * That gets us back on the path to having good data when there are
- * multiple pages with the same lru_count.
- */
- cur_count = (shared->cur_lru_count)++;
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- int this_delta;
- int this_page_number;
-
- if (shared->page_status[slotno] == SLRU_PAGE_EMPTY)
- return slotno;
- this_delta = cur_count - shared->page_lru_count[slotno];
- if (this_delta < 0)
- {
- /*
- * Clean up in case shared updates have caused cur_count
- * increments to get "lost". We back off the page counts,
- * rather than trying to increase cur_count, to avoid any
- * question of infinite loops or failure in the presence of
- * wrapped-around counts.
- */
- shared->page_lru_count[slotno] = cur_count;
- this_delta = 0;
- }
- this_page_number = shared->page_number[slotno];
- if (this_page_number == shared->latest_page_number)
- continue;
- if (shared->page_status[slotno] == SLRU_PAGE_VALID)
- {
- if (this_delta > best_valid_delta ||
- (this_delta == best_valid_delta &&
- ctl->PagePrecedes(this_page_number,
- best_valid_page_number)))
- {
- bestvalidslot = slotno;
- best_valid_delta = this_delta;
- best_valid_page_number = this_page_number;
- }
- }
- else
- {
- if (this_delta > best_invalid_delta ||
- (this_delta == best_invalid_delta &&
- ctl->PagePrecedes(this_page_number,
- best_invalid_page_number)))
- {
- bestinvalidslot = slotno;
- best_invalid_delta = this_delta;
- best_invalid_page_number = this_page_number;
- }
- }
- }
-
- /*
- * If all pages (except possibly the latest one) are I/O busy, we'll
- * have to wait for an I/O to complete and then retry. In that
- * unhappy case, we choose to wait for the I/O on the least recently
- * used slot, on the assumption that it was likely initiated first of
- * all the I/Os in progress and may therefore finish first.
- */
- if (best_valid_delta < 0)
- {
- SimpleLruWaitIO(ctl, bestinvalidslot);
- continue;
- }
-
- /*
- * If the selected page is clean, we're set.
- */
- if (!shared->page_dirty[bestvalidslot])
- return bestvalidslot;
-
- /*
- * Write the page.
- */
- SlruInternalWritePage(ctl, bestvalidslot, NULL);
-
- /*
- * Now loop back and try again. This is the easiest way of dealing
- * with corner cases such as the victim page being re-dirtied while we
- * wrote it.
- */
- }
-}
-
-/*
- * Write dirty pages to disk during checkpoint or database shutdown. Flushing
- * is deferred until the next call to ProcessSyncRequests(), though we do fsync
- * the containing directory here to make sure that newly created directory
- * entries are on disk.
- */
-void
-SimpleLruWriteAll(SlruCtl ctl, bool allow_redirtied)
-{
- SlruShared shared = ctl->shared;
- SlruWriteAllData fdata;
- int slotno;
- int pageno = 0;
- int i;
- bool ok;
-
- /* update the stats counter of flushes */
- pgstat_count_slru_flush(shared->slru_stats_idx);
-
- /*
- * Find and write dirty pages
- */
- fdata.num_files = 0;
-
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- SlruInternalWritePage(ctl, slotno, &fdata);
-
- /*
- * In some places (e.g. checkpoints), we cannot assert that the slot
- * is clean now, since another process might have re-dirtied it
- * already. That's okay.
- */
- Assert(allow_redirtied ||
- shared->page_status[slotno] == SLRU_PAGE_EMPTY ||
- (shared->page_status[slotno] == SLRU_PAGE_VALID &&
- !shared->page_dirty[slotno]));
- }
-
- LWLockRelease(shared->ControlLock);
-
- /*
- * Now close any files that were open
- */
- ok = true;
- for (i = 0; i < fdata.num_files; i++)
- {
- if (CloseTransientFile(fdata.fd[i]) != 0)
- {
- slru_errcause = SLRU_CLOSE_FAILED;
- slru_errno = errno;
- pageno = fdata.segno[i] * SLRU_PAGES_PER_SEGMENT;
- ok = false;
- }
- }
- if (!ok)
- SlruReportIOError(ctl, pageno, InvalidTransactionId);
-
- /* Ensure that directory entries for new files are on disk. */
- if (ctl->sync_handler != SYNC_HANDLER_NONE)
- fsync_fname(ctl->Dir, true);
+ return size >= offset + BLCKSZ;
}
/*
@@ -1223,75 +149,14 @@ SimpleLruWriteAll(SlruCtl ctl, bool allow_redirtied)
* after it has accrued freshly-written data.
*/
void
-SimpleLruTruncate(SlruCtl ctl, int cutoffPage)
+SimpleLruTruncate(Oid rel_id, SlruPagePrecedesFunction PagePrecedes, int cutoffPage)
{
- SlruShared shared = ctl->shared;
- int slotno;
-
/* update the stats counter of truncates */
- pgstat_count_slru_truncate(shared->slru_stats_idx);
-
- /*
- * Scan shared memory and remove any pages preceding the cutoff page, to
- * ensure we won't rewrite them later. (Since this is normally called in
- * or just after a checkpoint, any dirty pages should have been flushed
- * already ... we're just being extra careful here.)
- */
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
-restart:;
-
- /*
- * While we are holding the lock, make an important safety check: the
- * current endpoint page must not be eligible for removal.
- */
- if (ctl->PagePrecedes(shared->latest_page_number, cutoffPage))
- {
- LWLockRelease(shared->ControlLock);
- ereport(LOG,
- (errmsg("could not truncate directory \"%s\": apparent wraparound",
- ctl->Dir)));
- return;
- }
-
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- if (shared->page_status[slotno] == SLRU_PAGE_EMPTY)
- continue;
- if (!ctl->PagePrecedes(shared->page_number[slotno], cutoffPage))
- continue;
-
- /*
- * If page is clean, just change state to EMPTY (expected case).
- */
- if (shared->page_status[slotno] == SLRU_PAGE_VALID &&
- !shared->page_dirty[slotno])
- {
- shared->page_status[slotno] = SLRU_PAGE_EMPTY;
- continue;
- }
-
- /*
- * Hmm, we have (or may have) I/O operations acting on the page, so
- * we've got to wait for them to finish and then start again. This is
- * the same logic as in SlruSelectLRUPage. (XXX if page is dirty,
- * wouldn't it be OK to just discard it without writing it?
- * SlruMayDeleteSegment() uses a stricter qualification, so we might
- * not delete this page in the end; even if we don't delete it, we
- * won't have cause to read its data again. For now, keep the logic
- * the same as it was.)
- */
- if (shared->page_status[slotno] == SLRU_PAGE_VALID)
- SlruInternalWritePage(ctl, slotno, NULL);
- else
- SimpleLruWaitIO(ctl, slotno);
- goto restart;
- }
-
- LWLockRelease(shared->ControlLock);
+ pgstat_count_slru_truncate(rel_id);
/* Now we can remove the old segment(s) */
- (void) SlruScanDirectory(ctl, SlruScanDirCbDeleteCutoff, &cutoffPage);
+ (void) SlruScanDirectory(rel_id, PagePrecedes, SlruScanDirCbDeleteCutoff,
+ &cutoffPage);
}
/*
@@ -1301,21 +166,21 @@ restart:;
* they either can't yet contain anything, or have already been cleaned out.
*/
static void
-SlruInternalDeleteSegment(SlruCtl ctl, int segno)
+SlruInternalDeleteSegment(Oid rel_id, int segno)
{
char path[MAXPGPATH];
/* Forget any fsync requests queued for this segment. */
- if (ctl->sync_handler != SYNC_HANDLER_NONE)
+ if (defs[rel_id].synchronize)
{
FileTag tag;
- INIT_SLRUFILETAG(tag, ctl->sync_handler, segno);
+ INIT_SLRUFILETAG(tag, rel_id, segno);
RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true);
}
/* Unlink the file. */
- SlruFileName(ctl, path, segno);
+ SlruFileName(rel_id, path, segno);
ereport(DEBUG2, (errmsg_internal("removing file \"%s\"", path)));
unlink(path);
}
@@ -1324,54 +189,15 @@ SlruInternalDeleteSegment(SlruCtl ctl, int segno)
* Delete an individual SLRU segment, identified by the segment number.
*/
void
-SlruDeleteSegment(SlruCtl ctl, int segno)
+SlruDeleteSegment(Oid rel_id, int segno)
{
- SlruShared shared = ctl->shared;
- int slotno;
- bool did_write;
+ RelFileNode rnode = SlruRelFileNode(rel_id);
/* Clean out any possibly existing references to the segment. */
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-restart:
- did_write = false;
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- int pagesegno = shared->page_number[slotno] / SLRU_PAGES_PER_SEGMENT;
-
- if (shared->page_status[slotno] == SLRU_PAGE_EMPTY)
- continue;
+ for (int i = 0; i < SLRU_PAGES_PER_SEGMENT; ++i)
+ DiscardBuffer(rnode, MAIN_FORKNUM, segno * SLRU_PAGES_PER_SEGMENT + i);
- /* not the segment we're looking for */
- if (pagesegno != segno)
- continue;
-
- /* If page is clean, just change state to EMPTY (expected case). */
- if (shared->page_status[slotno] == SLRU_PAGE_VALID &&
- !shared->page_dirty[slotno])
- {
- shared->page_status[slotno] = SLRU_PAGE_EMPTY;
- continue;
- }
-
- /* Same logic as SimpleLruTruncate() */
- if (shared->page_status[slotno] == SLRU_PAGE_VALID)
- SlruInternalWritePage(ctl, slotno, NULL);
- else
- SimpleLruWaitIO(ctl, slotno);
-
- did_write = true;
- }
-
- /*
- * Be extra careful and re-check. The IO functions release the control
- * lock, so new pages could have been read in.
- */
- if (did_write)
- goto restart;
-
- SlruInternalDeleteSegment(ctl, segno);
-
- LWLockRelease(shared->ControlLock);
+ SlruInternalDeleteSegment(rel_id, segno);
}
/*
@@ -1388,19 +214,21 @@ restart:
* first>=cutoff && last>=cutoff: no; every page of this segment is too young
*/
static bool
-SlruMayDeleteSegment(SlruCtl ctl, int segpage, int cutoffPage)
+SlruMayDeleteSegment(SlruPagePrecedesFunction PagePrecedes,
+ int segpage, int cutoffPage)
{
int seg_last_page = segpage + SLRU_PAGES_PER_SEGMENT - 1;
Assert(segpage % SLRU_PAGES_PER_SEGMENT == 0);
- return (ctl->PagePrecedes(segpage, cutoffPage) &&
- ctl->PagePrecedes(seg_last_page, cutoffPage));
+ return (PagePrecedes(segpage, cutoffPage) &&
+ PagePrecedes(seg_last_page, cutoffPage));
}
#ifdef USE_ASSERT_CHECKING
static void
-SlruPagePrecedesTestOffset(SlruCtl ctl, int per_page, uint32 offset)
+SlruPagePrecedesTestOffset(SlruPagePrecedesFunction PagePrecedes,
+ int per_page, uint32 offset)
{
TransactionId lhs,
rhs;
@@ -1425,19 +253,19 @@ SlruPagePrecedesTestOffset(SlruCtl ctl, int per_page, uint32 offset)
Assert(!TransactionIdPrecedes(rhs, lhs + 1));
Assert(!TransactionIdFollowsOrEquals(lhs, rhs));
Assert(!TransactionIdFollowsOrEquals(rhs, lhs));
- Assert(!ctl->PagePrecedes(lhs / per_page, lhs / per_page));
- Assert(!ctl->PagePrecedes(lhs / per_page, rhs / per_page));
- Assert(!ctl->PagePrecedes(rhs / per_page, lhs / per_page));
- Assert(!ctl->PagePrecedes((lhs - per_page) / per_page, rhs / per_page));
- Assert(ctl->PagePrecedes(rhs / per_page, (lhs - 3 * per_page) / per_page));
- Assert(ctl->PagePrecedes(rhs / per_page, (lhs - 2 * per_page) / per_page));
- Assert(ctl->PagePrecedes(rhs / per_page, (lhs - 1 * per_page) / per_page)
+ Assert(!PagePrecedes(lhs / per_page, lhs / per_page));
+ Assert(!PagePrecedes(lhs / per_page, rhs / per_page));
+ Assert(!PagePrecedes(rhs / per_page, lhs / per_page));
+ Assert(!PagePrecedes((lhs - per_page) / per_page, rhs / per_page));
+ Assert(PagePrecedes(rhs / per_page, (lhs - 3 * per_page) / per_page));
+ Assert(PagePrecedes(rhs / per_page, (lhs - 2 * per_page) / per_page));
+ Assert(PagePrecedes(rhs / per_page, (lhs - 1 * per_page) / per_page)
|| (1U << 31) % per_page != 0); /* See CommitTsPagePrecedes() */
- Assert(ctl->PagePrecedes((lhs + 1 * per_page) / per_page, rhs / per_page)
+ Assert(PagePrecedes((lhs + 1 * per_page) / per_page, rhs / per_page)
|| (1U << 31) % per_page != 0);
- Assert(ctl->PagePrecedes((lhs + 2 * per_page) / per_page, rhs / per_page));
- Assert(ctl->PagePrecedes((lhs + 3 * per_page) / per_page, rhs / per_page));
- Assert(!ctl->PagePrecedes(rhs / per_page, (lhs + per_page) / per_page));
+ Assert(PagePrecedes((lhs + 2 * per_page) / per_page, rhs / per_page));
+ Assert(PagePrecedes((lhs + 3 * per_page) / per_page, rhs / per_page));
+ Assert(!PagePrecedes(rhs / per_page, (lhs + per_page) / per_page));
/*
* GetNewTransactionId() has assigned the last XID it can safely use, and
@@ -1450,7 +278,7 @@ SlruPagePrecedesTestOffset(SlruCtl ctl, int per_page, uint32 offset)
oldestXact = newestXact + 1;
oldestXact -= 1U << 31;
oldestPage = oldestXact / per_page;
- Assert(!SlruMayDeleteSegment(ctl,
+ Assert(!SlruMayDeleteSegment(PagePrecedes,
(newestPage -
newestPage % SLRU_PAGES_PER_SEGMENT),
oldestPage));
@@ -1466,7 +294,7 @@ SlruPagePrecedesTestOffset(SlruCtl ctl, int per_page, uint32 offset)
oldestXact = newestXact + 1;
oldestXact -= 1U << 31;
oldestPage = oldestXact / per_page;
- Assert(!SlruMayDeleteSegment(ctl,
+ Assert(!SlruMayDeleteSegment(PagePrecedes,
(newestPage -
newestPage % SLRU_PAGES_PER_SEGMENT),
oldestPage));
@@ -1482,12 +310,12 @@ SlruPagePrecedesTestOffset(SlruCtl ctl, int per_page, uint32 offset)
* do not apply to them.)
*/
void
-SlruPagePrecedesUnitTests(SlruCtl ctl, int per_page)
+SlruPagePrecedesUnitTests(SlruPagePrecedesFunction PagePrecedes, int per_page)
{
/* Test first, middle and last entries of a page. */
- SlruPagePrecedesTestOffset(ctl, per_page, 0);
- SlruPagePrecedesTestOffset(ctl, per_page, per_page / 2);
- SlruPagePrecedesTestOffset(ctl, per_page, per_page - 1);
+ SlruPagePrecedesTestOffset(PagePrecedes, per_page, 0);
+ SlruPagePrecedesTestOffset(PagePrecedes, per_page, per_page / 2);
+ SlruPagePrecedesTestOffset(PagePrecedes, per_page, per_page - 1);
}
#endif
@@ -1497,11 +325,12 @@ SlruPagePrecedesUnitTests(SlruCtl ctl, int per_page)
* one containing the page passed as "data".
*/
bool
-SlruScanDirCbReportPresence(SlruCtl ctl, char *filename, int segpage, void *data)
+SlruScanDirCbReportPresence(Oid rel_id, SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage, void *data)
{
int cutoffPage = *(int *) data;
- if (SlruMayDeleteSegment(ctl, segpage, cutoffPage))
+ if (SlruMayDeleteSegment(PagePrecedes, segpage, cutoffPage))
return true; /* found one; don't iterate any more */
return false; /* keep going */
@@ -1512,12 +341,18 @@ SlruScanDirCbReportPresence(SlruCtl ctl, char *filename, int segpage, void *data
* This callback deletes segments prior to the one passed in as "data".
*/
static bool
-SlruScanDirCbDeleteCutoff(SlruCtl ctl, char *filename, int segpage, void *data)
+SlruScanDirCbDeleteCutoff(Oid rel_id, SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage, void *data)
{
+ RelFileNode rnode = SlruRelFileNode(rel_id);
int cutoffPage = *(int *) data;
- if (SlruMayDeleteSegment(ctl, segpage, cutoffPage))
- SlruInternalDeleteSegment(ctl, segpage / SLRU_PAGES_PER_SEGMENT);
+ if (SlruMayDeleteSegment(PagePrecedes, segpage, cutoffPage))
+ {
+ for (int i = 0; i < SLRU_PAGES_PER_SEGMENT; ++i)
+ DiscardBuffer(rnode, MAIN_FORKNUM, segpage + i);
+ SlruInternalDeleteSegment(rel_id, segpage / SLRU_PAGES_PER_SEGMENT);
+ }
return false; /* keep going */
}
@@ -1527,9 +362,10 @@ SlruScanDirCbDeleteCutoff(SlruCtl ctl, char *filename, int segpage, void *data)
* This callback deletes all segments.
*/
bool
-SlruScanDirCbDeleteAll(SlruCtl ctl, char *filename, int segpage, void *data)
+SlruScanDirCbDeleteAll(Oid rel_id, SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage, void *data)
{
- SlruInternalDeleteSegment(ctl, segpage / SLRU_PAGES_PER_SEGMENT);
+ SlruInternalDeleteSegment(rel_id, segpage / SLRU_PAGES_PER_SEGMENT);
return false; /* keep going */
}
@@ -1550,16 +386,20 @@ SlruScanDirCbDeleteAll(SlruCtl ctl, char *filename, int segpage, void *data)
* Note that no locking is applied.
*/
bool
-SlruScanDirectory(SlruCtl ctl, SlruScanCallback callback, void *data)
+SlruScanDirectory(Oid rel_id, SlruPagePrecedesFunction PagePrecedes,
+ SlruScanCallback callback, void *data)
{
bool retval = false;
DIR *cldir;
struct dirent *clde;
int segno;
int segpage;
+ const char *path;
+
+ path = defs[rel_id].path;
- cldir = AllocateDir(ctl->Dir);
- while ((clde = ReadDir(cldir, ctl->Dir)) != NULL)
+ cldir = AllocateDir(path);
+ while ((clde = ReadDir(cldir, path)) != NULL)
{
size_t len;
@@ -1572,8 +412,8 @@ SlruScanDirectory(SlruCtl ctl, SlruScanCallback callback, void *data)
segpage = segno * SLRU_PAGES_PER_SEGMENT;
elog(DEBUG2, "SlruScanDirectory invoking callback on %s/%s",
- ctl->Dir, clde->d_name);
- retval = callback(ctl, clde->d_name, segpage, data);
+ path, clde->d_name);
+ retval = callback(rel_id, PagePrecedes, clde->d_name, segpage, data);
if (retval)
break;
}
@@ -1583,30 +423,294 @@ SlruScanDirectory(SlruCtl ctl, SlruScanCallback callback, void *data)
return retval;
}
+void
+CheckPointSLRU(void)
+{
+ /* Ensure that directory entries for new files are on disk. */
+ for (int i = 0; i < lengthof(defs); ++i)
+ {
+ if (defs[i].synchronize)
+ fsync_fname(defs[i].path, true);
+ }
+}
+
/*
- * Individual SLRUs (clog, ...) have to provide a sync.c handler function so
- * that they can provide the correct "SlruCtl" (otherwise we don't know how to
- * build the path), but they just forward to this common implementation that
- * performs the fsync.
+ * Read a buffer. Buffer is pinned on return.
*/
-int
-SlruSyncFileTag(SlruCtl ctl, const FileTag *ftag, char *path)
+Buffer
+ReadSlruBuffer(Oid rel_id, int pageno)
+{
+ RelFileNode rnode = SlruRelFileNode(rel_id);
+ Buffer buffer;
+ bool hit;
+
+ /* Try to avoid doing a buffer mapping table lookup for repeated access. */
+ buffer = slru_recent_buffers[rel_id].recent_buffer;
+ if (slru_recent_buffers[rel_id].pageno == pageno &&
+ BufferIsValid(buffer) &&
+ ReadRecentBuffer(rnode, MAIN_FORKNUM, pageno, buffer))
+ {
+ pgstat_count_slru_page_hit(rel_id);
+ return buffer;
+ }
+
+ /* Regular lookup. */
+ buffer = ReadBufferWithoutRelcacheWithHit(rnode, MAIN_FORKNUM, pageno,
+ RBM_NORMAL, &hit);
+
+ /* Remember where this page is for next time. */
+ slru_recent_buffers[rel_id].pageno = pageno;
+ slru_recent_buffers[rel_id].recent_buffer = buffer;
+
+ if (hit)
+ pgstat_count_slru_page_hit(rel_id);
+
+ return buffer;
+}
+
+/*
+ * Zero-initialize a buffer. Buffer is pinned and exclusively locked on return.
+ */
+Buffer
+ZeroSlruBuffer(Oid rel_id, int pageno)
+{
+ RelFileNode rnode = SlruRelFileNode(rel_id);
+ Buffer buffer;
+
+ buffer = ReadBufferWithoutRelcache(rnode, MAIN_FORKNUM, pageno,
+ RBM_ZERO_AND_LOCK, NULL, true);
+
+ /* Remember where this page is for next time. */
+ slru_recent_buffers[rel_id].pageno = pageno;
+ slru_recent_buffers[rel_id].recent_buffer = buffer;
+
+ pgstat_count_slru_page_zeroed(rel_id);
+
+ return buffer;
+}
+
+Oid
+SlruRelIdByName(const char *name)
{
- int fd;
- int save_errno;
- int result;
+ for (int i = 0; i < lengthof(defs); ++i)
+ if (strcmp(defs[i].name, name) == 0)
+ return i;
+
+ elog(ERROR, "unknown SLRU \"%s\"", name);
+}
- SlruFileName(ctl, path, ftag->segno);
+const char *
+SlruName(Oid rel_id)
+{
+ if (rel_id >= SLRU_NUM_RELS)
+ elog(ERROR, "invalid SLRU rel ID %u", rel_id);
+
+ return defs[rel_id].name;
+}
- fd = OpenTransientFile(path, O_RDWR | PG_BINARY);
- if (fd < 0)
+int
+slrusyncfiletag(const FileTag *ftag, char *path)
+{
+ SMgrRelation reln;
+ File file;
+
+ reln = smgropen(ftag->rnode, InvalidBackendId);
+ file = slrufile(reln, ftag->segno * SLRU_PAGES_PER_SEGMENT, O_RDWR, true);
+ if (file < 0)
+ {
+ /* Path is reported here so the caller can make an error message */
+ SlruFileName(ftag->rnode.relNode, path, ftag->segno);
return -1;
+ }
+
+ pgstat_count_slru_flush(reln->smgr_rnode.node.relNode);
+
+ return FileSync(file, WAIT_EVENT_SLRU_SYNC);
+}
+
+static File
+slrufile(SMgrRelation reln, BlockNumber blocknum, int mode, bool missing_ok)
+{
+ int segment = blocknum / SLRU_PAGES_PER_SEGMENT;
+ char path[MAXPGPATH];
+
+ Assert(reln->smgr_rnode.node.dbNode == SLRU_DB_ID);
+ Assert(reln->smgr_rnode.node.relNode < lengthof(defs));
+ Assert(defs[reln->smgr_rnode.node.relNode].path != NULL);
+
+ /* Do we have the right file open already? */
+ if (reln->slru_file_segment == segment)
+ {
+ /* XXX How can we invalidate this if the SLRU wraps around?! */
+ Assert(reln->slru_file != -1);
+ return reln->slru_file;
+ }
+
+ /* Close the current file, if we have one open. */
+ if (reln->slru_file_segment != -1)
+ {
+ Assert(reln->slru_file != -1);
+ FileClose(reln->slru_file);
+ reln->slru_file = -1;
+ reln->slru_file_segment = -1;
+ }
+
+ /* Open the file we want. */
+ SlruFileName(reln->smgr_rnode.node.relNode, path, segment);
+ reln->slru_file = PathNameOpenFile(path, mode);
+ if (reln->slru_file >= 0)
+ reln->slru_file_segment = segment;
+ else if (!(missing_ok && errno == ENOENT))
+ ereport(ERROR,
+ (errcode_for_file_access(),
+ errmsg("could not open file \"%s\": %m",
+ path)));
+
+ /*
+ * XXX That error message doesn't contain the xid; then again, the
+ * previous SLRU error codes were all weird about xids anyway; maybe
+ * clog.c, notify.c et al should set a context that has SLRU-specific
+ * context in a more natural format
+ */
+
+ return reln->slru_file;
+}
+
+void
+slruopen(SMgrRelation reln)
+{
+ reln->slru_file = -1;
+ reln->slru_file_segment = -1;
+
+ /*
+ * We don't want this to be closed at end of transaction, which would
+ * otherwise happen, because it isn't owned by a Relation.
+ */
+ dlist_delete(&reln->node);
+}
+
+void
+slruclose(SMgrRelation reln, ForkNumber forknum)
+{
+ if (reln->slru_file != -1)
+ FileClose(reln->slru_file);
+ reln->slru_file = -1;
+ reln->slru_file_segment = -1;
+}
+
+void
+slruwrite(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
+ char *buffer, bool skipFsync)
+{
+ File file;
+ Oid rel_id;
+ off_t offset;
+ int rc;
+
+ file = slrufile(reln, blocknum, O_RDWR | O_CREAT, false);
+ offset = (blocknum % SLRU_PAGES_PER_SEGMENT) * BLCKSZ;
+
+ rc = FileWrite(file, buffer, BLCKSZ, offset, WAIT_EVENT_SLRU_WRITE);
+ if (rc < 0)
+ ereport(ERROR,
+ (errcode_for_file_access(),
+ errmsg("could not write %d bytes to file \"%s\" at offset %d: %m",
+ BLCKSZ,
+ FilePathName(file),
+ (int) offset)));
+ if (rc < BLCKSZ)
+ ereport(ERROR,
+ (errcode_for_file_access(),
+ errmsg("could not write %d bytes to file \"%s\" at offset %d, only %d bytes written",
+ BLCKSZ,
+ FilePathName(file),
+ (int) offset,
+ rc)));
+
+ rel_id = reln->smgr_rnode.node.relNode;
+ if (defs[rel_id].synchronize)
+ {
+ FileTag tag;
+
+ /* Tell checkpointer to synchronize this file. */
+ INIT_SLRUFILETAG(tag, rel_id, blocknum / SLRU_PAGES_PER_SEGMENT);
+ if (!RegisterSyncRequest(&tag, SYNC_REQUEST, false))
+ {
+ /* Queue full. Do it synchronously. */
+ if (FileSync(file, WAIT_EVENT_SLRU_SYNC) < 0)
+ ereport(data_sync_elevel(ERROR),
+ (errcode_for_file_access(),
+ errmsg("could not synchronize file \"%s\": %m",
+ FilePathName(file))));
+ }
+ }
+
+ pgstat_count_slru_page_written(reln->smgr_rnode.node.relNode);
+}
+
+void
+slruread(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
+ char *buffer)
+{
+ File file;
+ off_t offset;
+ int rc;
+
+ file = slrufile(reln, blocknum, O_RDWR, false);
+ offset = (blocknum % SLRU_PAGES_PER_SEGMENT) * BLCKSZ;
+
+ rc = FileRead(file, buffer, BLCKSZ, offset, WAIT_EVENT_SLRU_READ);
+ if (rc < 0)
+ ereport(ERROR,
+ (errcode_for_file_access(),
+ errmsg("could not read %d bytes from file \"%s\" at offset %d: %m",
+ BLCKSZ,
+ FilePathName(file),
+ (int) offset)));
+ if (rc < BLCKSZ)
+ ereport(ERROR,
+ (errcode_for_file_access(),
+ errmsg("could not read %d bytes from file \"%s\" at offset %d, only %d bytes read",
+ BLCKSZ,
+ FilePathName(file),
+ (int) offset,
+ rc)));
+
+ pgstat_count_slru_page_read(reln->smgr_rnode.node.relNode);
+}
- result = pg_fsync(fd);
- save_errno = errno;
+void
+slruwriteback(SMgrRelation reln, ForkNumber forknum,
+ BlockNumber blocknum, BlockNumber nblocks)
+{
+ Oid rel_id;
- CloseTransientFile(fd);
+ /* No point in flushing data we won't be fsyncing. */
+ rel_id = reln->smgr_rnode.node.relNode;
+ if (!defs[rel_id].synchronize)
+ return;
- errno = save_errno;
- return result;
+ while (nblocks > 0)
+ {
+ File file;
+ BlockNumber blocknum_in_this_file;
+ BlockNumber nflush;
+
+ /* File range of blocks to flush in one file. */
+ blocknum_in_this_file = blocknum % SLRU_PAGES_PER_SEGMENT;
+ nflush = Min(nblocks, SLRU_PAGES_PER_SEGMENT - blocknum_in_this_file);
+
+ /* The file might have been unlinked already, so tolerate missing. */
+ file = slrufile(reln, blocknum, O_RDWR, true);
+ if (file < 0)
+ return;
+
+ FileWriteback(file,
+ BLCKSZ * blocknum_in_this_file,
+ BLCKSZ * nflush,
+ WAIT_EVENT_SLRU_FLUSH);
+
+ nblocks -= nflush;
+ blocknum += nflush;
+ }
}
diff --git a/src/backend/access/transam/subtrans.c b/src/backend/access/transam/subtrans.c
index 7240454ca4..bdd547c2bd 100644
--- a/src/backend/access/transam/subtrans.c
+++ b/src/backend/access/transam/subtrans.c
@@ -32,6 +32,7 @@
#include "access/subtrans.h"
#include "access/transam.h"
#include "pg_trace.h"
+#include "storage/bufmgr.h"
#include "utils/snapmgr.h"
@@ -63,15 +64,7 @@
static TransactionId cachedFetchSubXid = InvalidTransactionId;
static TransactionId cachedFetchTopmostXid = InvalidTransactionId;
-/*
- * Link to shared-memory data structures for SUBTRANS control
- */
-static SlruCtlData SubTransCtlData;
-
-#define SubTransCtl (&SubTransCtlData)
-
-
-static int ZeroSUBTRANSPage(int pageno);
+static Buffer ZeroSUBTRANSPage(int pageno);
static bool SubTransPagePrecedes(int page1, int page2);
@@ -83,16 +76,15 @@ SubTransSetParent(TransactionId xid, TransactionId parent)
{
int pageno = TransactionIdToPage(xid);
int entryno = TransactionIdToEntry(xid);
- int slotno;
TransactionId *ptr;
+ Buffer buffer;
Assert(TransactionIdIsValid(parent));
Assert(TransactionIdFollows(xid, parent));
- LWLockAcquire(SubtransSLRULock, LW_EXCLUSIVE);
-
- slotno = SimpleLruReadPage(SubTransCtl, pageno, true, xid);
- ptr = (TransactionId *) SubTransCtl->shared->page_buffer[slotno];
+ buffer = ReadSlruBuffer(SLRU_SUBTRANS_REL_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ ptr = (TransactionId *) BufferGetPage(buffer);
ptr += entryno;
/*
@@ -104,10 +96,10 @@ SubTransSetParent(TransactionId xid, TransactionId parent)
{
Assert(*ptr == InvalidTransactionId);
*ptr = parent;
- SubTransCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
}
- LWLockRelease(SubtransSLRULock);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -118,9 +110,9 @@ SubTransGetParent(TransactionId xid)
{
int pageno = TransactionIdToPage(xid);
int entryno = TransactionIdToEntry(xid);
- int slotno;
TransactionId *ptr;
TransactionId parent;
+ Buffer buffer;
/* Can't ask about stuff that might not be around anymore */
Assert(TransactionIdFollowsOrEquals(xid, TransactionXmin));
@@ -129,15 +121,14 @@ SubTransGetParent(TransactionId xid)
if (!TransactionIdIsNormal(xid))
return InvalidTransactionId;
- /* lock is acquired by SimpleLruReadPage_ReadOnly */
+ buffer = ReadSlruBuffer(SLRU_SUBTRANS_REL_ID, pageno);
- slotno = SimpleLruReadPage_ReadOnly(SubTransCtl, pageno, xid);
- ptr = (TransactionId *) SubTransCtl->shared->page_buffer[slotno];
+ ptr = (TransactionId *) BufferGetPage(buffer);
ptr += entryno;
parent = *ptr;
- LWLockRelease(SubtransSLRULock);
+ ReleaseBuffer(buffer);
return parent;
}
@@ -195,26 +186,6 @@ SubTransGetTopmostTransaction(TransactionId xid)
return previousXid;
}
-
-/*
- * Initialization of shared memory for SUBTRANS
- */
-Size
-SUBTRANSShmemSize(void)
-{
- return SimpleLruShmemSize(NUM_SUBTRANS_BUFFERS, 0);
-}
-
-void
-SUBTRANSShmemInit(void)
-{
- SubTransCtl->PagePrecedes = SubTransPagePrecedes;
- SimpleLruInit(SubTransCtl, "Subtrans", NUM_SUBTRANS_BUFFERS, 0,
- SubtransSLRULock, "pg_subtrans",
- LWTRANCHE_SUBTRANS_BUFFER, SYNC_HANDLER_NONE);
- SlruPagePrecedesUnitTests(SubTransCtl, SUBTRANS_XACTS_PER_PAGE);
-}
-
/*
* This func must be called ONCE on system install. It creates
* the initial SUBTRANS segment. (The SUBTRANS directory is assumed to
@@ -228,18 +199,16 @@ SUBTRANSShmemInit(void)
void
BootStrapSUBTRANS(void)
{
- int slotno;
+ Buffer buffer;
- LWLockAcquire(SubtransSLRULock, LW_EXCLUSIVE);
+ SlruPagePrecedesUnitTests(SubTransPagePrecedes, SUBTRANS_XACTS_PER_PAGE);
/* Create and zero the first page of the subtrans log */
- slotno = ZeroSUBTRANSPage(0);
+ buffer = ZeroSUBTRANSPage(0);
/* Make sure it's written out */
- SimpleLruWritePage(SubTransCtl, slotno);
- Assert(!SubTransCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(SubtransSLRULock);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -250,10 +219,15 @@ BootStrapSUBTRANS(void)
*
* Control lock must be held at entry, and will be held at exit.
*/
-static int
+static Buffer
ZeroSUBTRANSPage(int pageno)
{
- return SimpleLruZeroPage(SubTransCtl, pageno);
+ Buffer buffer;
+
+ buffer = ZeroSlruBuffer(SLRU_SUBTRANS_REL_ID, pageno);
+ MarkBufferDirty(buffer);
+
+ return buffer;
}
/*
@@ -276,7 +250,6 @@ StartupSUBTRANS(TransactionId oldestActiveXID)
* Whenever we advance into a new page, ExtendSUBTRANS will likewise zero
* the new page without regard to whatever was previously on disk.
*/
- LWLockAcquire(SubtransSLRULock, LW_EXCLUSIVE);
startPage = TransactionIdToPage(oldestActiveXID);
nextXid = ShmemVariableCache->nextXid;
@@ -284,36 +257,15 @@ StartupSUBTRANS(TransactionId oldestActiveXID)
while (startPage != endPage)
{
- (void) ZeroSUBTRANSPage(startPage);
+ UnlockReleaseBuffer(ZeroSUBTRANSPage(startPage));
startPage++;
/* must account for wraparound */
if (startPage > TransactionIdToPage(MaxTransactionId))
startPage = 0;
}
- (void) ZeroSUBTRANSPage(startPage);
-
- LWLockRelease(SubtransSLRULock);
+ UnlockReleaseBuffer(ZeroSUBTRANSPage(startPage));
}
-/*
- * Perform a checkpoint --- either during shutdown, or on-the-fly
- */
-void
-CheckPointSUBTRANS(void)
-{
- /*
- * Write dirty SUBTRANS pages to disk
- *
- * This is not actually necessary from a correctness point of view. We do
- * it merely to improve the odds that writing of dirty pages is done by
- * the checkpoint process and not by backends.
- */
- TRACE_POSTGRESQL_SUBTRANS_CHECKPOINT_START(true);
- SimpleLruWriteAll(SubTransCtl, true);
- TRACE_POSTGRESQL_SUBTRANS_CHECKPOINT_DONE(true);
-}
-
-
/*
* Make sure that SUBTRANS has room for a newly-allocated XID.
*
@@ -337,12 +289,8 @@ ExtendSUBTRANS(TransactionId newestXact)
pageno = TransactionIdToPage(newestXact);
- LWLockAcquire(SubtransSLRULock, LW_EXCLUSIVE);
-
/* Zero the page */
- ZeroSUBTRANSPage(pageno);
-
- LWLockRelease(SubtransSLRULock);
+ UnlockReleaseBuffer(ZeroSUBTRANSPage(pageno));
}
@@ -368,7 +316,7 @@ TruncateSUBTRANS(TransactionId oldestXact)
TransactionIdRetreat(oldestXact);
cutoffPage = TransactionIdToPage(oldestXact);
- SimpleLruTruncate(SubTransCtl, cutoffPage);
+ SimpleLruTruncate(SLRU_SUBTRANS_REL_ID, SubTransPagePrecedes, cutoffPage);
}
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 71136b11a2..6df2b148c5 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -51,6 +51,7 @@
#include "access/heaptoast.h"
#include "access/multixact.h"
#include "access/rewriteheap.h"
+#include "access/slru.h"
#include "access/subtrans.h"
#include "access/timeline.h"
#include "access/transam.h"
@@ -1459,7 +1460,7 @@ WaitXLogInsertionsToFinish(XLogRecPtr upto)
*/
if (upto > reservedUpto)
{
- ereport(LOG,
+ ereport(PANIC,
(errmsg("request to flush past end of generated WAL; request %X/%X, current position %X/%X",
LSN_FORMAT_ARGS(upto), LSN_FORMAT_ARGS(reservedUpto))));
upto = reservedUpto;
@@ -4495,6 +4496,7 @@ BootStrapXLOG(void)
uint64 sysidentifier;
struct timeval tv;
pg_crc32c crc;
+ ResourceOwner resowner;
/* allow ordinary WAL segment creation, like StartupXLOG() would */
LWLockAcquire(ControlFileLock, LW_EXCLUSIVE);
@@ -4635,11 +4637,15 @@ BootStrapXLOG(void)
/* some additional ControlFile fields are set in WriteControlFile() */
WriteControlFile();
- /* Bootstrap the commit log, too */
+ /* Bootstrap the commit log too */
+ resowner = ResourceOwnerCreate(NULL, "bootstrap resowner");
+ CurrentResourceOwner = resowner;
BootStrapCLOG();
BootStrapCommitTs();
BootStrapSUBTRANS();
BootStrapMultiXact();
+ CurrentResourceOwner = NULL;
+ ResourceOwnerDelete(resowner);
pfree(buffer);
@@ -6834,15 +6840,12 @@ CheckPointGuts(XLogRecPtr checkPointRedo, int flags)
CheckPointSnapBuild();
CheckPointLogicalRewriteHeap();
CheckPointReplicationOrigin();
+ CheckPointPredicate();
+ CheckPointSLRU();
- /* Write out all dirty data in SLRUs and the main buffer pool */
+ /* Write out all dirty data in the buffer pool */
TRACE_POSTGRESQL_BUFFER_CHECKPOINT_START(flags);
CheckpointStats.ckpt_write_t = GetCurrentTimestamp();
- CheckPointCLOG();
- CheckPointCommitTs();
- CheckPointSUBTRANS();
- CheckPointMultiXact();
- CheckPointPredicate();
CheckPointBuffers(flags);
/* Perform all queued up fsyncs */
diff --git a/src/backend/commands/async.c b/src/backend/commands/async.c
index 3e1b92df03..c4249bfa86 100644
--- a/src/backend/commands/async.c
+++ b/src/backend/commands/async.c
@@ -141,6 +141,7 @@
#include "libpq/libpq.h"
#include "libpq/pqformat.h"
#include "miscadmin.h"
+#include "storage/bufmgr.h"
#include "storage/ipc.h"
#include "storage/lmgr.h"
#include "storage/proc.h"
@@ -305,12 +306,6 @@ static AsyncQueueControl *asyncQueueControl;
#define QUEUE_NEXT_LISTENER(i) (asyncQueueControl->backend[i].nextListener)
#define QUEUE_BACKEND_POS(i) (asyncQueueControl->backend[i].pos)
-/*
- * The SLRU buffer area through which we access the notification queue
- */
-static SlruCtlData NotifyCtlData;
-
-#define NotifyCtl (&NotifyCtlData)
#define QUEUE_PAGESIZE BLCKSZ
#define QUEUE_FULL_WARN_INTERVAL 5000 /* warn at most once every 5s */
@@ -521,8 +516,6 @@ AsyncShmemSize(void)
size = mul_size(MaxBackends + 1, sizeof(QueueBackendStatus));
size = add_size(size, offsetof(AsyncQueueControl, backend));
- size = add_size(size, SimpleLruShmemSize(NUM_NOTIFY_BUFFERS, 0));
-
return size;
}
@@ -565,20 +558,13 @@ AsyncShmemInit(void)
}
}
- /*
- * Set up SLRU management of the pg_notify data.
- */
- NotifyCtl->PagePrecedes = asyncQueuePagePrecedes;
- SimpleLruInit(NotifyCtl, "Notify", NUM_NOTIFY_BUFFERS, 0,
- NotifySLRULock, "pg_notify", LWTRANCHE_NOTIFY_BUFFER,
- SYNC_HANDLER_NONE);
-
if (!found)
{
/*
* During start or reboot, clean out the pg_notify directory.
*/
- (void) SlruScanDirectory(NotifyCtl, SlruScanDirCbDeleteAll, NULL);
+ (void) SlruScanDirectory(SLRU_NOTIFY_REL_ID, asyncQueuePagePrecedes,
+ SlruScanDirCbDeleteAll, NULL);
}
}
@@ -1411,10 +1397,7 @@ asyncQueueAddEntries(ListCell *nextNotify)
QueuePosition queue_head;
int pageno;
int offset;
- int slotno;
-
- /* We hold both NotifyQueueLock and NotifySLRULock during this operation */
- LWLockAcquire(NotifySLRULock, LW_EXCLUSIVE);
+ Buffer buffer;
/*
* We work with a local copy of QUEUE_HEAD, which we write back to shared
@@ -1439,13 +1422,17 @@ asyncQueueAddEntries(ListCell *nextNotify)
*/
pageno = QUEUE_POS_PAGE(queue_head);
if (QUEUE_POS_IS_ZERO(queue_head))
- slotno = SimpleLruZeroPage(NotifyCtl, pageno);
+ {
+ buffer = ZeroSlruBuffer(SLRU_NOTIFY_REL_ID, pageno);
+ }
else
- slotno = SimpleLruReadPage(NotifyCtl, pageno, true,
- InvalidTransactionId);
+ {
+ buffer = ReadSlruBuffer(SLRU_NOTIFY_REL_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ }
/* Note we mark the page dirty before writing in it */
- NotifyCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
while (nextNotify != NULL)
{
@@ -1476,7 +1463,7 @@ asyncQueueAddEntries(ListCell *nextNotify)
}
/* Now copy qe into the shared buffer page */
- memcpy(NotifyCtl->shared->page_buffer[slotno] + offset,
+ memcpy(BufferGetPage(buffer) + offset,
&qe,
qe.length);
@@ -1491,7 +1478,10 @@ asyncQueueAddEntries(ListCell *nextNotify)
* asyncQueueIsFull() ensured that there is room to create this
* page without overrunning the queue.
*/
- slotno = SimpleLruZeroPage(NotifyCtl, QUEUE_POS_PAGE(queue_head));
+ UnlockReleaseBuffer(buffer);
+ buffer = ZeroSlruBuffer(SLRU_NOTIFY_REL_ID,
+ QUEUE_POS_PAGE(queue_head));
+ MarkBufferDirty(buffer);
/*
* If the new page address is a multiple of QUEUE_CLEANUP_DELAY,
@@ -1505,12 +1495,11 @@ asyncQueueAddEntries(ListCell *nextNotify)
break;
}
}
+ UnlockReleaseBuffer(buffer);
/* Success, so update the global QUEUE_HEAD */
QUEUE_HEAD = queue_head;
- LWLockRelease(NotifySLRULock);
-
return nextNotify;
}
@@ -1983,17 +1972,16 @@ asyncQueueReadAllNotifications(void)
{
int curpage = QUEUE_POS_PAGE(pos);
int curoffset = QUEUE_POS_OFFSET(pos);
- int slotno;
int copysize;
+ Buffer buffer;
/*
- * We copy the data from SLRU into a local buffer, so as to avoid
- * holding the NotifySLRULock while we are examining the entries
- * and possibly transmitting them to our frontend. Copy only the
- * part of the page we will actually inspect.
+ * We copy the data into a local buffer, so as to avoid holding a
+ * buffer pin while we are examining the entries and possibly
+ * transmitting them to our frontend. Copy only the part of the
+ * page we will actually inspect.
*/
- slotno = SimpleLruReadPage_ReadOnly(NotifyCtl, curpage,
- InvalidTransactionId);
+ buffer = ReadSlruBuffer(SLRU_NOTIFY_REL_ID, curpage);
if (curpage == QUEUE_POS_PAGE(head))
{
/* we only want to read as far as head */
@@ -2007,10 +1995,9 @@ asyncQueueReadAllNotifications(void)
copysize = QUEUE_PAGESIZE - curoffset;
}
memcpy(page_buffer.buf + curoffset,
- NotifyCtl->shared->page_buffer[slotno] + curoffset,
+ BufferGetPage(buffer) + curoffset,
copysize);
- /* Release lock that we got from SimpleLruReadPage_ReadOnly() */
- LWLockRelease(NotifySLRULock);
+ ReleaseBuffer(buffer);
/*
* Process messages up to the stop position, end of page, or an
@@ -2207,7 +2194,7 @@ asyncQueueAdvanceTail(void)
* SimpleLruTruncate() will ask for NotifySLRULock but will also
* release the lock again.
*/
- SimpleLruTruncate(NotifyCtl, newtailpage);
+ SimpleLruTruncate(SLRU_NOTIFY_REL_ID, asyncQueuePagePrecedes, newtailpage);
/*
* Update QUEUE_STOP_PAGE. This changes asyncQueueIsFull()'s verdict
diff --git a/src/backend/postmaster/checkpointer.c b/src/backend/postmaster/checkpointer.c
index c937c39f50..529bad33fe 100644
--- a/src/backend/postmaster/checkpointer.c
+++ b/src/backend/postmaster/checkpointer.c
@@ -1172,6 +1172,24 @@ CompactCheckpointerRequestQueue(void)
/* must hold CheckpointerCommLock in exclusive mode */
Assert(LWLockHeldByMe(CheckpointerCommLock));
+ /*
+ * XXX FIXME
+ * If we were unlucky enough to be reached from
+ * TransactionIdSetTreeStatus() while trying to read in a CLOG page and
+ * having to write out some other dirty page and then finding the sync
+ * request queue full, we're now running in a critical section and we're
+ * not allowed to allocate any memory below.
+ *
+ * XXX Without this, 027_stream_regress.pl aborts below occasionally, due
+ * to pressure on its very small shared_buffers and thus also sync request
+ * queue.
+ */
+ if (CritSectionCount > 0)
+ {
+ elog(LOG, "CompactCheckpointerRequestQueue() critical section, returning");
+ return false;
+ }
+
/* Initialize skip_slot array */
skip_slot = palloc0(sizeof(bool) * CheckpointerShmem->num_requests);
diff --git a/src/backend/storage/buffer/buf_init.c b/src/backend/storage/buffer/buf_init.c
index 2862e9e412..719f27a86a 100644
--- a/src/backend/storage/buffer/buf_init.c
+++ b/src/backend/storage/buffer/buf_init.c
@@ -20,6 +20,7 @@
BufferDescPadded *BufferDescriptors;
char *BufferBlocks;
+XLogRecPtr *BufferExternalLSNs;
ConditionVariableMinimallyPadded *BufferIOCVArray;
WritebackContext BackendWritebackContext;
CkptSortItem *CkptBufferIds;
@@ -69,9 +70,11 @@ InitBufferPool(void)
{
bool foundBufs,
foundDescs,
+ foundLSNs,
foundIOCV,
foundBufCkpt;
+
/* Align descriptors to a cacheline boundary. */
BufferDescriptors = (BufferDescPadded *)
ShmemInitStruct("Buffer Descriptors",
@@ -88,6 +91,11 @@ InitBufferPool(void)
NBuffers * sizeof(ConditionVariableMinimallyPadded),
&foundIOCV);
+ BufferExternalLSNs = (XLogRecPtr *)
+ ShmemInitStruct("Buffer External LSNs",
+ NBuffers * sizeof(XLogRecPtr),
+ &foundLSNs);
+
/*
* The array used to sort to-be-checkpointed buffer ids is located in
* shared memory, to avoid having to allocate significant amounts of
@@ -99,10 +107,10 @@ InitBufferPool(void)
ShmemInitStruct("Checkpoint BufferIds",
NBuffers * sizeof(CkptSortItem), &foundBufCkpt);
- if (foundDescs || foundBufs || foundIOCV || foundBufCkpt)
+ if (foundDescs || foundBufs || foundIOCV || foundBufCkpt || foundLSNs)
{
/* should find all of these, or none of them */
- Assert(foundDescs && foundBufs && foundIOCV && foundBufCkpt);
+ Assert(foundDescs && foundBufs && foundIOCV && foundBufCkpt && foundLSNs);
/* note: this path is only taken in EXEC_BACKEND case */
}
else
@@ -133,6 +141,8 @@ InitBufferPool(void)
LWTRANCHE_BUFFER_CONTENT);
ConditionVariableInit(BufferDescriptorGetIOCV(buf));
+
+ BufferExternalLSNs[i] = InvalidXLogRecPtr;
}
/* Correct last entry of linked list */
@@ -166,6 +176,9 @@ BufferShmemSize(void)
/* size of data pages */
size = add_size(size, mul_size(NBuffers, BLCKSZ));
+ /* size of external LSNs */
+ size = add_size(size, mul_size(NBuffers, sizeof(XLogRecPtr)));
+
/* size of stuff controlled by freelist.c */
size = add_size(size, StrategyShmemSize());
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index ae13011d27..7e6ffe4e18 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -57,10 +57,21 @@
#include "utils/resowner_private.h"
#include "utils/timestamp.h"
+/*
+ * XXX Ideally we'd switch to standard pages for SLRU data, but in the
+ * meantime we need some way to identify buffers that hold raw data (no
+ * invasive LSN, no checksums).
+ */
+#define BufferHasStandardPage(bufHdr) \
+ ((bufHdr)->tag.rnode.dbNode != 9)
+
+#define BufferHasExternalLSN(bufHdr) \
+ !BufferHasStandardPage(bufHdr)
/* Note: these two macros only work on shared buffers, not local ones! */
#define BufHdrGetBlock(bufHdr) ((Block) (BufferBlocks + ((Size) (bufHdr)->buf_id) * BLCKSZ))
-#define BufferGetLSN(bufHdr) (PageGetLSN(BufHdrGetBlock(bufHdr)))
+#define BufferGetLSN(bufHdr) \
+ (BufferHasExternalLSN(bufHdr) ? BufferGetExternalLSN(bufHdr) : PageGetLSN(BufHdrGetBlock(bufHdr)))
/* Note: this macro only works on local buffers, not shared ones! */
#define LocalBufHdrGetBlock(bufHdr) \
@@ -799,6 +810,21 @@ ReadBufferWithoutRelcache(RelFileNode rnode, ForkNumber forkNum,
mode, strategy, &hit);
}
+/*
+ * Like ReadBufferWithoutRelcache, but returns the hit flag.
+ * XXX Merge
+ */
+Buffer
+ReadBufferWithoutRelcacheWithHit(RelFileNode rnode, ForkNumber forkNum,
+ BlockNumber blockNum, ReadBufferMode mode,
+ bool *hit)
+{
+ SMgrRelation smgr = smgropen(rnode, InvalidBackendId);
+
+ return ReadBuffer_common(smgr, RELPERSISTENCE_PERMANENT, forkNum, blockNum,
+ mode, NULL, hit);
+}
+
/*
* ReadBuffer_common -- common logic for all ReadBuffer variants
@@ -1019,7 +1045,8 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
}
/* check for garbage data */
- if (!PageIsVerifiedExtended((Page) bufBlock, blockNum,
+ if (BufferHasStandardPage(bufHdr) &&
+ !PageIsVerifiedExtended((Page) bufBlock, blockNum,
PIV_LOG_WARNING | PIV_REPORT_STAT))
{
if (mode == RBM_ZERO_ON_ERROR || zero_damaged_pages)
@@ -1440,6 +1467,9 @@ BufferAlloc(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
LWLockRelease(newPartitionLock);
+ if (BufferHasExternalLSN(buf))
+ BufferSetExternalLSN(buf, InvalidXLogRecPtr);
+
/*
* Buffer contents are currently invalid. Try to obtain the right to
* start I/O. If StartBufferIO returns false, then someone else managed
@@ -1560,6 +1590,84 @@ retry:
StrategyFreeBuffer(buf);
}
+/*
+ * DiscardBuffer -- drop a buffer from pool.
+ *
+ * If the buffer isn't present in shared buffers, nothing happens. If it is
+ * present and not pinned, it is discarded without making any attempt to write
+ * it back out to the operating system. If I/O is in progress, we wait for it
+ * to to complete. If it is pinned, an error is raised (some other backend
+ * must still be interested in it, so it's an error to discard it).
+ */
+void
+DiscardBuffer(RelFileNode rnode, ForkNumber forkNum, BlockNumber blockNum)
+{
+ SMgrRelation smgr = smgropen(rnode, InvalidBackendId);
+ BufferTag tag; /* identity of target block */
+ uint32 hash; /* hash value for tag */
+ LWLock *partitionLock; /* buffer partition lock for it */
+ int buf_id;
+ BufferDesc *bufHdr;
+ uint32 buf_state;
+
+ /* create a tag so we can lookup the buffer */
+ INIT_BUFFERTAG(tag, smgr->smgr_rnode.node, forkNum, blockNum);
+
+ /* determine its hash code and partition lock ID */
+ hash = BufTableHashCode(&tag);
+ partitionLock = BufMappingPartitionLock(hash);
+
+ retry:
+ /* see if the block is in the buffer pool */
+ LWLockAcquire(partitionLock, LW_SHARED);
+ buf_id = BufTableLookup(&tag, hash);
+ LWLockRelease(partitionLock);
+
+ /* didn't find it, so nothing to do */
+ if (buf_id < 0)
+ return;
+
+ /* take the buffer header lock */
+ bufHdr = GetBufferDescriptor(buf_id);
+ buf_state = LockBufHdr(bufHdr);
+
+ /*
+ * The buffer might been evicted after we released the partition lock and
+ * before we acquired the buffer header lock. If so, the buffer we've
+ * locked might contain some other data which we shouldn't touch. If the
+ * buffer hasn't been recycled, we proceed to invalidate it.
+ */
+ if (RelFileNodeEquals(bufHdr->tag.rnode, rnode) &&
+ bufHdr->tag.blockNum == blockNum &&
+ bufHdr->tag.forkNum == forkNum)
+ {
+ if (buf_state & BM_IO_IN_PROGRESS)
+ {
+ UnlockBufHdr(bufHdr, buf_state);
+ WaitIO(bufHdr);
+ goto retry;
+ }
+ else if (BUF_STATE_GET_REFCOUNT(buf_state) == 0)
+ {
+ /* Nobody has it pinned, so we can immediately invalidate it. */
+ InvalidateBuffer(bufHdr); /* releases spinlock */
+ }
+ else
+ {
+ /*
+ * XXX: Is it OK to say that the contract for DiscardBuffer() is
+ * that the caller is asserting that no one else could be
+ * interested in this buffer, and therefore it's a programming
+ * error or corruption if you reach this case?
+ */
+ UnlockBufHdr(bufHdr, buf_state);
+ elog(ERROR, "cannot discard buffer that is pinned");
+ }
+ }
+ else
+ UnlockBufHdr(bufHdr, buf_state);
+}
+
/*
* MarkBufferDirty
*
@@ -2890,7 +2998,10 @@ FlushBuffer(BufferDesc *buf, SMgrRelation reln)
* buffer, other processes might be updating hint bits in it, so we must
* copy the page to private storage if we do checksumming.
*/
- bufToWrite = PageSetChecksumCopy((Page) bufBlock, buf->tag.blockNum);
+ if (BufferHasStandardPage(buf))
+ bufToWrite = PageSetChecksumCopy((Page) bufBlock, buf->tag.blockNum);
+ else
+ bufToWrite = bufBlock;
if (track_io_timing)
INSTR_TIME_SET_CURRENT(io_start);
@@ -3019,7 +3130,10 @@ BufferGetLSNAtomic(Buffer buffer)
Assert(BufferIsPinned(buffer));
buf_state = LockBufHdr(bufHdr);
- lsn = PageGetLSN(page);
+ if (BufferHasStandardPage(bufHdr))
+ lsn = PageGetLSN(page);
+ else
+ lsn = BufferGetExternalLSN(bufHdr);
UnlockBufHdr(bufHdr, buf_state);
return lsn;
@@ -3532,7 +3646,8 @@ FlushRelationBuffers(Relation rel)
errcallback.previous = error_context_stack;
error_context_stack = &errcallback;
- PageSetChecksumInplace(localpage, bufHdr->tag.blockNum);
+ if (BufferHasStandardPage(bufHdr))
+ PageSetChecksumInplace(localpage, bufHdr->tag.blockNum);
smgrwrite(RelationGetSmgr(rel),
bufHdr->tag.forkNum,
@@ -5002,3 +5117,29 @@ TestForOldSnapshot_impl(Snapshot snapshot, Relation relation)
(errcode(ERRCODE_SNAPSHOT_TOO_OLD),
errmsg("snapshot too old")));
}
+
+/*
+ * Check if a buffer tag is currently mapped.
+ *
+ * XXX Dubious semantics; needed only for multixact's handling for
+ * inconsistent states.
+ */
+bool
+BufferProbe(RelFileNode rnode, ForkNumber forkNum, BlockNumber blockNum)
+{
+ BufferTag tag;
+ uint32 hash;
+ LWLock *partitionLock;
+ int buf_id;
+
+ INIT_BUFFERTAG(tag, rnode, forkNum, blockNum);
+
+ hash = BufTableHashCode(&tag);
+ partitionLock = BufMappingPartitionLock(hash);
+
+ LWLockAcquire(partitionLock, LW_SHARED);
+ buf_id = BufTableLookup(&tag, hash);
+ LWLockRelease(partitionLock);
+
+ return buf_id >= 0;
+}
diff --git a/src/backend/storage/ipc/ipci.c b/src/backend/storage/ipc/ipci.c
index 26372d95b3..5a9767aff4 100644
--- a/src/backend/storage/ipc/ipci.c
+++ b/src/backend/storage/ipc/ipci.c
@@ -116,9 +116,7 @@ CalculateShmemSize(int *num_semaphores)
size = add_size(size, XLogPrefetchShmemSize());
size = add_size(size, XLOGShmemSize());
size = add_size(size, XLogRecoveryShmemSize());
- size = add_size(size, CLOGShmemSize());
size = add_size(size, CommitTsShmemSize());
- size = add_size(size, SUBTRANSShmemSize());
size = add_size(size, TwoPhaseShmemSize());
size = add_size(size, BackgroundWorkerShmemSize());
size = add_size(size, MultiXactShmemSize());
@@ -240,9 +238,7 @@ CreateSharedMemoryAndSemaphores(void)
XLOGShmemInit();
XLogPrefetchShmemInit();
XLogRecoveryShmemInit();
- CLOGShmemInit();
CommitTsShmemInit();
- SUBTRANSShmemInit();
MultiXactShmemInit();
InitBufferPool();
diff --git a/src/backend/storage/lmgr/lwlock.c b/src/backend/storage/lmgr/lwlock.c
index 8aef909037..8e2dcf1519 100644
--- a/src/backend/storage/lmgr/lwlock.c
+++ b/src/backend/storage/lmgr/lwlock.c
@@ -129,20 +129,6 @@ extern slock_t *ShmemLock;
extern const char *const IndividualLWLockNames[]; /* in lwlocknames.c */
static const char *const BuiltinTrancheNames[] = {
- /* LWTRANCHE_XACT_BUFFER: */
- "XactBuffer",
- /* LWTRANCHE_COMMITTS_BUFFER: */
- "CommitTSBuffer",
- /* LWTRANCHE_SUBTRANS_BUFFER: */
- "SubtransBuffer",
- /* LWTRANCHE_MULTIXACTOFFSET_BUFFER: */
- "MultiXactOffsetBuffer",
- /* LWTRANCHE_MULTIXACTMEMBER_BUFFER: */
- "MultiXactMemberBuffer",
- /* LWTRANCHE_NOTIFY_BUFFER: */
- "NotifyBuffer",
- /* LWTRANCHE_SERIAL_BUFFER: */
- "SerialBuffer",
/* LWTRANCHE_WAL_INSERT: */
"WALInsert",
/* LWTRANCHE_BUFFER_CONTENT: */
diff --git a/src/backend/storage/lmgr/lwlocknames.txt b/src/backend/storage/lmgr/lwlocknames.txt
index 6c7cf6c295..fbd2976a60 100644
--- a/src/backend/storage/lmgr/lwlocknames.txt
+++ b/src/backend/storage/lmgr/lwlocknames.txt
@@ -17,10 +17,10 @@ WALWriteLock 8
ControlFileLock 9
# 10 was CheckpointLock
XactSLRULock 11
-SubtransSLRULock 12
+# 12 was SubtransSLRULock
MultiXactGenLock 13
-MultiXactOffsetSLRULock 14
-MultiXactMemberSLRULock 15
+# 14 was MultiXactOffsetSLRULock
+# 15 was MultiXactMemberSLRULock
RelCacheInitLock 16
CheckpointerCommLock 17
TwoPhaseStateLock 18
@@ -31,7 +31,7 @@ AutovacuumLock 22
AutovacuumScheduleLock 23
SyncScanLock 24
RelationMappingLock 25
-NotifySLRULock 26
+# 26 was NotifySLRULock
NotifyQueueLock 27
SerializableXactHashLock 28
SerializableFinishedListLock 29
@@ -43,7 +43,7 @@ DynamicSharedMemoryControlLock 34
AutoFileLock 35
ReplicationSlotAllocationLock 36
ReplicationSlotControlLock 37
-CommitTsSLRULock 38
+# 38 was CommitTsSLRULock
CommitTsLock 39
ReplicationOriginLock 40
MultiXactTruncationLock 41
diff --git a/src/backend/storage/lmgr/predicate.c b/src/backend/storage/lmgr/predicate.c
index 25e7e4e37b..40888e2c33 100644
--- a/src/backend/storage/lmgr/predicate.c
+++ b/src/backend/storage/lmgr/predicate.c
@@ -312,14 +312,6 @@
((targethash) ^ ((uint32) PointerGetDatum((predicatelocktag)->myXact)) \
<< LOG2_NUM_PREDICATELOCK_PARTITIONS)
-
-/*
- * The SLRU buffer area through which we access the old xids.
- */
-static SlruCtlData SerialSlruCtlData;
-
-#define SerialSlruCtl (&SerialSlruCtlData)
-
#define SERIAL_PAGESIZE BLCKSZ
#define SERIAL_ENTRYSIZE sizeof(SerCommitSeqNo)
#define SERIAL_ENTRIESPERPAGE (SERIAL_PAGESIZE / SERIAL_ENTRYSIZE)
@@ -331,8 +323,8 @@ static SlruCtlData SerialSlruCtlData;
#define SerialNextPage(page) (((page) >= SERIAL_MAX_PAGE) ? 0 : (page) + 1)
-#define SerialValue(slotno, xid) (*((SerCommitSeqNo *) \
- (SerialSlruCtl->shared->page_buffer[slotno] + \
+#define SerialValue(buffer, xid) (*((SerCommitSeqNo *) \
+ (BufferGetPage(buffer) + \
((((uint32) (xid)) % SERIAL_ENTRIESPERPAGE) * SERIAL_ENTRYSIZE))))
#define SerialPage(xid) (((uint32) (xid)) / SERIAL_ENTRIESPERPAGE)
@@ -867,17 +859,10 @@ SerialInit(void)
{
bool found;
- /*
- * Set up SLRU management of the pg_serial data.
- */
- SerialSlruCtl->PagePrecedes = SerialPagePrecedesLogically;
- SimpleLruInit(SerialSlruCtl, "Serial",
- NUM_SERIAL_BUFFERS, 0, SerialSLRULock, "pg_serial",
- LWTRANCHE_SERIAL_BUFFER, SYNC_HANDLER_NONE);
#ifdef USE_ASSERT_CHECKING
SerialPagePrecedesLogicallyUnitTests();
#endif
- SlruPagePrecedesUnitTests(SerialSlruCtl, SERIAL_ENTRIESPERPAGE);
+ SlruPagePrecedesUnitTests(SerialPagePrecedesLogically, SERIAL_ENTRIESPERPAGE);
/*
* Create or attach to the SerialControl structure.
@@ -907,9 +892,9 @@ SerialAdd(TransactionId xid, SerCommitSeqNo minConflictCommitSeqNo)
{
TransactionId tailXid;
int targetPage;
- int slotno;
int firstZeroPage;
bool isNewPage;
+ Buffer buffer;
Assert(TransactionIdIsValid(xid));
@@ -954,16 +939,22 @@ SerialAdd(TransactionId xid, SerCommitSeqNo minConflictCommitSeqNo)
/* Initialize intervening pages. */
while (firstZeroPage != targetPage)
{
- (void) SimpleLruZeroPage(SerialSlruCtl, firstZeroPage);
+ buffer = ZeroSlruBuffer(SLRU_SERIAL_REL_ID, firstZeroPage);
+ MarkBufferDirty(buffer);
+ UnlockReleaseBuffer(buffer);
firstZeroPage = SerialNextPage(firstZeroPage);
}
- slotno = SimpleLruZeroPage(SerialSlruCtl, targetPage);
+ buffer = ZeroSlruBuffer(SLRU_SERIAL_REL_ID, targetPage);
}
else
- slotno = SimpleLruReadPage(SerialSlruCtl, targetPage, true, xid);
+ {
+ buffer = ReadSlruBuffer(SLRU_SERIAL_REL_ID, targetPage);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ }
- SerialValue(slotno, xid) = minConflictCommitSeqNo;
- SerialSlruCtl->shared->page_dirty[slotno] = true;
+ SerialValue(buffer, xid) = minConflictCommitSeqNo;
+ MarkBufferDirty(buffer);
+ UnlockReleaseBuffer(buffer);
LWLockRelease(SerialSLRULock);
}
@@ -979,7 +970,7 @@ SerialGetMinConflictCommitSeqNo(TransactionId xid)
TransactionId headXid;
TransactionId tailXid;
SerCommitSeqNo val;
- int slotno;
+ Buffer buffer;
Assert(TransactionIdIsValid(xid));
@@ -1001,9 +992,9 @@ SerialGetMinConflictCommitSeqNo(TransactionId xid)
* The following function must be called without holding SerialSLRULock,
* but will return with that lock held, which must then be released.
*/
- slotno = SimpleLruReadPage_ReadOnly(SerialSlruCtl,
- SerialPage(xid), xid);
- val = SerialValue(slotno, xid);
+ buffer = ReadSlruBuffer(SLRU_SERIAL_REL_ID, SerialPage(xid));
+ val = SerialValue(buffer, xid);
+ ReleaseBuffer(buffer);
LWLockRelease(SerialSLRULock);
return val;
}
@@ -1122,19 +1113,7 @@ CheckPointPredicate(void)
LWLockRelease(SerialSLRULock);
/* Truncate away pages that are no longer required */
- SimpleLruTruncate(SerialSlruCtl, tailPage);
-
- /*
- * Write dirty SLRU pages to disk
- *
- * This is not actually necessary from a correctness point of view. We do
- * it merely as a debugging aid.
- *
- * We're doing this after the truncation to avoid writing pages right
- * before deleting the file in which they sit, which would be completely
- * pointless.
- */
- SimpleLruWriteAll(SerialSlruCtl, true);
+ SimpleLruTruncate(SLRU_SERIAL_REL_ID, SerialPagePrecedesLogically, tailPage);
}
/*------------------------------------------------------------------------*/
@@ -1396,7 +1375,6 @@ PredicateLockShmemSize(void)
/* Shared memory structures for SLRU tracking of old committed xids. */
size = add_size(size, sizeof(SerialControlData));
- size = add_size(size, SimpleLruShmemSize(NUM_SERIAL_BUFFERS, 0));
return size;
}
diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c
index a477f70f0e..107c1e3be4 100644
--- a/src/backend/storage/smgr/smgr.c
+++ b/src/backend/storage/smgr/smgr.c
@@ -17,6 +17,7 @@
*/
#include "postgres.h"
+#include "access/slru.h"
#include "access/xlogutils.h"
#include "lib/ilist.h"
#include "storage/bufmgr.h"
@@ -82,6 +83,14 @@ static const f_smgr smgrsw[] = {
.smgr_nblocks = mdnblocks,
.smgr_truncate = mdtruncate,
.smgr_immedsync = mdimmedsync,
+ },
+ /* "SLRU" storage */
+ {
+ .smgr_open = slruopen,
+ .smgr_close = slruclose,
+ .smgr_read = slruread,
+ .smgr_write = slruwrite,
+ .smgr_writeback = slruwriteback,
}
};
@@ -176,13 +185,18 @@ smgropen(RelFileNode rnode, BackendId backend)
reln->smgr_targblock = InvalidBlockNumber;
for (int i = 0; i <= MAX_FORKNUM; ++i)
reln->smgr_cached_nblocks[i] = InvalidBlockNumber;
- reln->smgr_which = 0; /* we only have md.c at present */
- /* implementation-specific initialization */
- smgrsw[reln->smgr_which].smgr_open(reln);
+ /* XXX find some elegant way to do this, or something better */
+ if (rnode.dbNode == SLRU_DB_ID)
+ reln->smgr_which = 1; /* slru.c */
+ else
+ reln->smgr_which = 0; /* md.c */
/* it has no owner yet */
dlist_push_tail(&unowned_relns, &reln->node);
+
+ /* implementation-specific initialization */
+ smgrsw[reln->smgr_which].smgr_open(reln);
}
return reln;
diff --git a/src/backend/storage/sync/sync.c b/src/backend/storage/sync/sync.c
index e1fb631003..cc3acde56f 100644
--- a/src/backend/storage/sync/sync.c
+++ b/src/backend/storage/sync/sync.c
@@ -18,9 +18,7 @@
#include <fcntl.h>
#include <sys/file.h>
-#include "access/commit_ts.h"
-#include "access/clog.h"
-#include "access/multixact.h"
+#include "access/slru.h"
#include "access/xlog.h"
#include "access/xlogutils.h"
#include "commands/tablespace.h"
@@ -106,21 +104,9 @@ static const SyncOps syncsw[] = {
.sync_unlinkfiletag = mdunlinkfiletag,
.sync_filetagmatches = mdfiletagmatches
},
- /* pg_xact */
- [SYNC_HANDLER_CLOG] = {
- .sync_syncfiletag = clogsyncfiletag
- },
- /* pg_commit_ts */
- [SYNC_HANDLER_COMMIT_TS] = {
- .sync_syncfiletag = committssyncfiletag
- },
- /* pg_multixact/offsets */
- [SYNC_HANDLER_MULTIXACT_OFFSET] = {
- .sync_syncfiletag = multixactoffsetssyncfiletag
- },
- /* pg_multixact/members */
- [SYNC_HANDLER_MULTIXACT_MEMBER] = {
- .sync_syncfiletag = multixactmemberssyncfiletag
+ /* SLRU */
+ [SYNC_HANDLER_SLRU] = {
+ .sync_syncfiletag = slrusyncfiletag
}
};
diff --git a/src/backend/utils/activity/pgstat_slru.c b/src/backend/utils/activity/pgstat_slru.c
index d0b85b62a5..4d981f90da 100644
--- a/src/backend/utils/activity/pgstat_slru.c
+++ b/src/backend/utils/activity/pgstat_slru.c
@@ -31,7 +31,7 @@ static void pgstat_reset_slru_counter_internal(int index, TimestampTz ts);
* SLRU counters are reported within critical sections so we use static memory
* in order to avoid memory allocation.
*/
-static PgStat_SLRUStats pending_SLRUStats[SLRU_NUM_ELEMENTS];
+static PgStat_SLRUStats pending_SLRUStats[SLRU_NUM_RELS];
bool have_slrustats = false;
@@ -110,37 +110,21 @@ pgstat_fetch_slru(void)
}
/*
- * Returns SLRU name for an index. The index may be above SLRU_NUM_ELEMENTS,
- * in which case this returns NULL. This allows writing code that does not
- * know the number of entries in advance.
+ * Returns SLRU name for an index.
*/
const char *
pgstat_get_slru_name(int slru_idx)
{
- if (slru_idx < 0 || slru_idx >= SLRU_NUM_ELEMENTS)
- return NULL;
-
- return slru_names[slru_idx];
+ return SlruName(slru_idx);
}
/*
- * Determine index of entry for a SLRU with a given name. If there's no exact
- * match, returns index of the last "other" entry used for SLRUs defined in
- * external projects.
+ * Determine index of entry for a SLRU with a given name.
*/
int
pgstat_get_slru_index(const char *name)
{
- int i;
-
- for (i = 0; i < SLRU_NUM_ELEMENTS; i++)
- {
- if (strcmp(slru_names[i], name) == 0)
- return i;
- }
-
- /* return index of the last entry (which is the "other" one) */
- return (SLRU_NUM_ELEMENTS - 1);
+ return SlruRelIdByName(name);
}
/*
@@ -168,7 +152,7 @@ pgstat_slru_flush(bool nowait)
else if (!LWLockConditionalAcquire(&stats_shmem->lock, LW_EXCLUSIVE))
return true;
- for (i = 0; i < SLRU_NUM_ELEMENTS; i++)
+ for (i = 0; i < SLRU_NUM_RELS; i++)
{
PgStat_SLRUStats *sharedent = &stats_shmem->stats[i];
PgStat_SLRUStats *pendingent = &pending_SLRUStats[i];
@@ -197,7 +181,7 @@ pgstat_slru_flush(bool nowait)
void
pgstat_slru_reset_all_cb(TimestampTz ts)
{
- for (int i = 0; i < SLRU_NUM_ELEMENTS; i++)
+ for (int i = 0; i < SLRU_NUM_RELS; i++)
pgstat_reset_slru_counter_internal(i, ts);
}
@@ -215,8 +199,7 @@ pgstat_slru_snapshot_cb(void)
}
/*
- * Returns pointer to entry with counters for given SLRU (based on the name
- * stored in SlruCtl as lwlock tranche name).
+ * Returns pointer to entry with counters for given SLRU.
*/
static inline PgStat_SLRUStats *
get_slru_entry(int slru_idx)
@@ -229,7 +212,7 @@ get_slru_entry(int slru_idx)
*/
Assert(IsUnderPostmaster || !IsPostmasterEnvironment);
- Assert((slru_idx >= 0) && (slru_idx < SLRU_NUM_ELEMENTS));
+ Assert((slru_idx >= 0) && (slru_idx < SLRU_NUM_RELS));
have_slrustats = true;
diff --git a/src/backend/utils/activity/wait_event.c b/src/backend/utils/activity/wait_event.c
index 87c15b9c6f..5d464cf734 100644
--- a/src/backend/utils/activity/wait_event.c
+++ b/src/backend/utils/activity/wait_event.c
@@ -657,8 +657,8 @@ pgstat_get_wait_io(WaitEventIO w)
case WAIT_EVENT_REPLICATION_SLOT_WRITE:
event_name = "ReplicationSlotWrite";
break;
- case WAIT_EVENT_SLRU_FLUSH_SYNC:
- event_name = "SLRUFlushSync";
+ case WAIT_EVENT_SLRU_FLUSH:
+ event_name = "SLRUFlush";
break;
case WAIT_EVENT_SLRU_READ:
event_name = "SLRURead";
diff --git a/src/backend/utils/adt/pgstatfuncs.c b/src/backend/utils/adt/pgstatfuncs.c
index 893690dad5..12cf36ec27 100644
--- a/src/backend/utils/adt/pgstatfuncs.c
+++ b/src/backend/utils/adt/pgstatfuncs.c
@@ -15,6 +15,7 @@
#include "postgres.h"
#include "access/htup_details.h"
+#include "access/slru.h"
#include "access/xlog.h"
#include "access/xlogprefetcher.h"
#include "catalog/pg_authid.h"
@@ -1823,7 +1824,7 @@ pg_stat_get_slru(PG_FUNCTION_ARGS)
/* request SLRU stats from the cumulative stats system */
stats = pgstat_fetch_slru();
- for (i = 0;; i++)
+ for (i = 0; i < SLRU_NUM_RELS; i++)
{
/* for each row */
Datum values[PG_STAT_GET_SLRU_COLS];
@@ -1831,10 +1832,7 @@ pg_stat_get_slru(PG_FUNCTION_ARGS)
PgStat_SLRUStats stat;
const char *name;
- name = pgstat_get_slru_name(i);
-
- if (!name)
- break;
+ name = SlruName(i);
stat = stats[i];
MemSet(values, 0, sizeof(values));
diff --git a/src/backend/utils/mmgr/mcxt.c b/src/backend/utils/mmgr/mcxt.c
index e12be1b9bd..7373538e2f 100644
--- a/src/backend/utils/mmgr/mcxt.c
+++ b/src/backend/utils/mmgr/mcxt.c
@@ -865,7 +865,9 @@ MemoryContextAlloc(MemoryContext context, Size size)
void *ret;
AssertArg(MemoryContextIsValid(context));
- AssertNotInCriticalSection(context);
+ // XXX horrible hack: while pinning clog buffers during commit, resowner
+ // stuff allocates!
+ //AssertNotInCriticalSection(context);
if (!AllocSizeIsValid(size))
elog(ERROR, "invalid memory alloc request size %zu", size);
@@ -1072,7 +1074,13 @@ palloc(Size size)
MemoryContext context = CurrentMemoryContext;
AssertArg(MemoryContextIsValid(context));
- AssertNotInCriticalSection(context);
+ /*
+ * XXX:TM I commented this out for now, because otherwise it can fail
+ * while writing back md.c buffers due to cache pressure while reading in
+ * CLOG buffers during commit, in a critical section (md.c allocates
+ * memory to build paths...); FIXME!
+ */
+ //AssertNotInCriticalSection(context);
if (!AllocSizeIsValid(size))
elog(ERROR, "invalid memory alloc request size %zu", size);
diff --git a/src/include/access/clog.h b/src/include/access/clog.h
index 543f2e2643..af16a2afcb 100644
--- a/src/include/access/clog.h
+++ b/src/include/access/clog.h
@@ -40,18 +40,12 @@ extern void TransactionIdSetTreeStatus(TransactionId xid, int nsubxids,
TransactionId *subxids, XidStatus status, XLogRecPtr lsn);
extern XidStatus TransactionIdGetStatus(TransactionId xid, XLogRecPtr *lsn);
-extern Size CLOGShmemBuffers(void);
-extern Size CLOGShmemSize(void);
-extern void CLOGShmemInit(void);
extern void BootStrapCLOG(void);
extern void StartupCLOG(void);
extern void TrimCLOG(void);
-extern void CheckPointCLOG(void);
extern void ExtendCLOG(TransactionId newestXact);
extern void TruncateCLOG(TransactionId oldestXact, Oid oldestxid_datoid);
-extern int clogsyncfiletag(const FileTag *ftag, char *path);
-
/* XLOG stuff */
#define CLOG_ZEROPAGE 0x00
#define CLOG_TRUNCATE 0x10
diff --git a/src/include/access/commit_ts.h b/src/include/access/commit_ts.h
index 7662f8e1a9..4e986fc023 100644
--- a/src/include/access/commit_ts.h
+++ b/src/include/access/commit_ts.h
@@ -27,7 +27,6 @@ extern bool TransactionIdGetCommitTsData(TransactionId xid,
extern TransactionId GetLatestCommitTsData(TimestampTz *ts,
RepOriginId *nodeid);
-extern Size CommitTsShmemBuffers(void);
extern Size CommitTsShmemSize(void);
extern void CommitTsShmemInit(void);
extern void BootStrapCommitTs(void);
@@ -41,8 +40,6 @@ extern void SetCommitTsLimit(TransactionId oldestXact,
TransactionId newestXact);
extern void AdvanceOldestCommitTsXid(TransactionId oldestXact);
-extern int committssyncfiletag(const FileTag *ftag, char *path);
-
/* XLOG stuff */
#define COMMIT_TS_ZEROPAGE 0x00
#define COMMIT_TS_TRUNCATE 0x10
diff --git a/src/include/access/multixact.h b/src/include/access/multixact.h
index a5600a320a..a6bbe53e1b 100644
--- a/src/include/access/multixact.h
+++ b/src/include/access/multixact.h
@@ -118,9 +118,6 @@ extern bool MultiXactIdPrecedes(MultiXactId multi1, MultiXactId multi2);
extern bool MultiXactIdPrecedesOrEquals(MultiXactId multi1,
MultiXactId multi2);
-extern int multixactoffsetssyncfiletag(const FileTag *ftag, char *path);
-extern int multixactmemberssyncfiletag(const FileTag *ftag, char *path);
-
extern void AtEOXact_MultiXact(void);
extern void AtPrepare_MultiXact(void);
extern void PostPrepare_MultiXact(TransactionId xid);
diff --git a/src/include/access/slru.h b/src/include/access/slru.h
index 130c41c863..4b967bb35d 100644
--- a/src/include/access/slru.h
+++ b/src/include/access/slru.h
@@ -1,7 +1,7 @@
/*-------------------------------------------------------------------------
*
* slru.h
- * Simple LRU buffering for transaction status logfiles
+ * Buffering for transaction status logfiles
*
* Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group
* Portions Copyright (c) 1994, Regents of the University of California
@@ -14,9 +14,34 @@
#define SLRU_H
#include "access/xlogdefs.h"
+#include "catalog/pg_tablespace_d.h"
+#include "storage/buf.h"
#include "storage/lwlock.h"
+#include "storage/smgr.h"
#include "storage/sync.h"
+/* Pseudo database ID used for SLRU data. */
+#define SLRU_DB_ID 9
+
+/* Pseudo relation IDs used by each cache. */
+#define SLRU_CLOG_REL_ID 0
+#define SLRU_MULTIXACT_OFFSET_REL_ID 1
+#define SLRU_MULTIXACT_MEMBER_REL_ID 2
+#define SLRU_SUBTRANS_REL_ID 3
+#define SLRU_SERIAL_REL_ID 4
+#define SLRU_COMMITTS_REL_ID 5
+#define SLRU_NOTIFY_REL_ID 6
+#define SLRU_NUM_RELS 7
+
+typedef bool (*SlruPagePrecedesFunction) (int, int);
+
+static inline RelFileNode
+SlruRelFileNode(Oid relNode)
+{
+ RelFileNode rnode = {DEFAULTTABLESPACE_OID, SLRU_DB_ID, relNode};
+ return rnode;
+}
+
/*
* Define SLRU segment size. A page is the same BLCKSZ as is used everywhere
@@ -33,142 +58,54 @@
*/
#define SLRU_PAGES_PER_SEGMENT 32
-/*
- * Page status codes. Note that these do not include the "dirty" bit.
- * page_dirty can be true only in the VALID or WRITE_IN_PROGRESS states;
- * in the latter case it implies that the page has been re-dirtied since
- * the write started.
- */
-typedef enum
-{
- SLRU_PAGE_EMPTY, /* buffer is not in use */
- SLRU_PAGE_READ_IN_PROGRESS, /* page is being read in */
- SLRU_PAGE_VALID, /* page is valid and not being written */
- SLRU_PAGE_WRITE_IN_PROGRESS /* page is being written out */
-} SlruPageStatus;
-
-/*
- * Shared-memory state
- */
-typedef struct SlruSharedData
-{
- LWLock *ControlLock;
-
- /* Number of buffers managed by this SLRU structure */
- int num_slots;
-
- /*
- * Arrays holding info for each buffer slot. Page number is undefined
- * when status is EMPTY, as is page_lru_count.
- */
- char **page_buffer;
- SlruPageStatus *page_status;
- bool *page_dirty;
- int *page_number;
- int *page_lru_count;
- LWLockPadded *buffer_locks;
-
- /*
- * Optional array of WAL flush LSNs associated with entries in the SLRU
- * pages. If not zero/NULL, we must flush WAL before writing pages (true
- * for pg_xact, false for multixact, pg_subtrans, pg_notify). group_lsn[]
- * has lsn_groups_per_page entries per buffer slot, each containing the
- * highest LSN known for a contiguous group of SLRU entries on that slot's
- * page.
- */
- XLogRecPtr *group_lsn;
- int lsn_groups_per_page;
-
- /*----------
- * We mark a page "most recently used" by setting
- * page_lru_count[slotno] = ++cur_lru_count;
- * The oldest page is therefore the one with the highest value of
- * cur_lru_count - page_lru_count[slotno]
- * The counts will eventually wrap around, but this calculation still
- * works as long as no page's age exceeds INT_MAX counts.
- *----------
- */
- int cur_lru_count;
-
- /*
- * latest_page_number is the page number of the current end of the log;
- * this is not critical data, since we use it only to avoid swapping out
- * the latest page.
- */
- int latest_page_number;
-
- /* SLRU's index for statistics purposes (might not be unique) */
- int slru_stats_idx;
-} SlruSharedData;
-
-typedef SlruSharedData *SlruShared;
-
-/*
- * SlruCtlData is an unshared structure that points to the active information
- * in shared memory.
- */
-typedef struct SlruCtlData
-{
- SlruShared shared;
-
- /*
- * Which sync handler function to use when handing sync requests over to
- * the checkpointer. SYNC_HANDLER_NONE to disable fsync (eg pg_notify).
- */
- SyncRequestHandler sync_handler;
-
- /*
- * Decide whether a page is "older" for truncation and as a hint for
- * evicting pages in LRU order. Return true if every entry of the first
- * argument is older than every entry of the second argument. Note that
- * !PagePrecedes(a,b) && !PagePrecedes(b,a) need not imply a==b; it also
- * arises when some entries are older and some are not. For SLRUs using
- * SimpleLruTruncate(), this must use modular arithmetic. (For others,
- * the behavior of this callback has no functional implications.) Use
- * SlruPagePrecedesUnitTests() in SLRUs meeting its criteria.
- */
- bool (*PagePrecedes) (int, int);
-
- /*
- * Dir is set during SimpleLruInit and does not change thereafter. Since
- * it's always the same, it doesn't need to be in shared memory.
- */
- char Dir[64];
-} SlruCtlData;
-
-typedef SlruCtlData *SlruCtl;
-
-
-extern Size SimpleLruShmemSize(int nslots, int nlsns);
-extern void SimpleLruInit(SlruCtl ctl, const char *name, int nslots, int nlsns,
- LWLock *ctllock, const char *subdir, int tranche_id,
- SyncRequestHandler sync_handler);
-extern int SimpleLruZeroPage(SlruCtl ctl, int pageno);
-extern int SimpleLruReadPage(SlruCtl ctl, int pageno, bool write_ok,
- TransactionId xid);
-extern int SimpleLruReadPage_ReadOnly(SlruCtl ctl, int pageno,
- TransactionId xid);
-extern void SimpleLruWritePage(SlruCtl ctl, int slotno);
-extern void SimpleLruWriteAll(SlruCtl ctl, bool allow_redirtied);
#ifdef USE_ASSERT_CHECKING
-extern void SlruPagePrecedesUnitTests(SlruCtl ctl, int per_page);
+extern void SlruPagePrecedesUnitTests(SlruPagePrecedesFunction PagePrecedes,
+ int per_page);
#else
#define SlruPagePrecedesUnitTests(ctl, per_page) do {} while (0)
#endif
-extern void SimpleLruTruncate(SlruCtl ctl, int cutoffPage);
-extern bool SimpleLruDoesPhysicalPageExist(SlruCtl ctl, int pageno);
+extern void SimpleLruTruncate(Oid rel_id, SlruPagePrecedesFunction PagePrecedes,
+ int cutoffPage);
+extern bool SimpleLruDoesPhysicalPageExist(Oid rel_id, int pageno);
-typedef bool (*SlruScanCallback) (SlruCtl ctl, char *filename, int segpage,
+typedef bool (*SlruScanCallback) (Oid rel_id,
+ SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage,
void *data);
-extern bool SlruScanDirectory(SlruCtl ctl, SlruScanCallback callback, void *data);
-extern void SlruDeleteSegment(SlruCtl ctl, int segno);
-
-extern int SlruSyncFileTag(SlruCtl ctl, const FileTag *ftag, char *path);
+extern bool SlruScanDirectory(Oid rel_id, SlruPagePrecedesFunction PagePrecedes,
+ SlruScanCallback callback, void *data);
+extern void SlruDeleteSegment(Oid rel_id, int segno);
/* SlruScanDirectory public callbacks */
-extern bool SlruScanDirCbReportPresence(SlruCtl ctl, char *filename,
+extern bool SlruScanDirCbReportPresence(Oid rel_id,
+ SlruPagePrecedesFunction PagePrecedes,
+ char *filename,
int segpage, void *data);
-extern bool SlruScanDirCbDeleteAll(SlruCtl ctl, char *filename, int segpage,
+extern bool SlruScanDirCbDeleteAll(Oid rel_id, SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage,
void *data);
+extern void CheckPointSLRU(void);
+
+/* Buffer access */
+extern Buffer ReadSlruBuffer(Oid rel_id, int pageno);
+extern Buffer ZeroSlruBuffer(Oid rel_id, int pageno);
+
+/* Interfaces use by stats view */
+extern Oid SlruRelIdByName(const char *name);
+extern const char *SlruName(Oid rel_id);
+
+/* Sync callback */
+extern int slrusyncfiletag(const FileTag *ftag, char *path);
+
+/* SMGR callbacks */
+extern void slruopen(SMgrRelation reln);
+extern void slruclose(SMgrRelation reln, ForkNumber forknum);
+extern void slruread(SMgrRelation reln, ForkNumber forknum,
+ BlockNumber blocknum, char *buffer);
+extern void slruwrite(SMgrRelation reln, ForkNumber forknum,
+ BlockNumber blocknum, char *buffer, bool skipFsync);
+extern void slruwriteback(SMgrRelation reln, ForkNumber forknum,
+ BlockNumber blocknum, BlockNumber nblocks);
+
#endif /* SLRU_H */
diff --git a/src/include/access/subtrans.h b/src/include/access/subtrans.h
index f94e116640..4685a05bc9 100644
--- a/src/include/access/subtrans.h
+++ b/src/include/access/subtrans.h
@@ -18,11 +18,8 @@ extern void SubTransSetParent(TransactionId xid, TransactionId parent);
extern TransactionId SubTransGetParent(TransactionId xid);
extern TransactionId SubTransGetTopmostTransaction(TransactionId xid);
-extern Size SUBTRANSShmemSize(void);
-extern void SUBTRANSShmemInit(void);
extern void BootStrapSUBTRANS(void);
extern void StartupSUBTRANS(TransactionId oldestActiveXID);
-extern void CheckPointSUBTRANS(void);
extern void ExtendSUBTRANS(TransactionId newestXact);
extern void TruncateSUBTRANS(TransactionId oldestXact);
diff --git a/src/include/pgstat.h b/src/include/pgstat.h
index ac28f813b4..bb7099fc80 100644
--- a/src/include/pgstat.h
+++ b/src/include/pgstat.h
@@ -691,5 +691,4 @@ extern PGDLLIMPORT SessionEndType pgStatSessionEndCause;
/* updated directly by backends and background processes */
extern PGDLLIMPORT PgStat_WalStats PendingWalStats;
-
#endif /* PGSTAT_H */
diff --git a/src/include/storage/buf_internals.h b/src/include/storage/buf_internals.h
index a17e7b28a5..3feee46fa1 100644
--- a/src/include/storage/buf_internals.h
+++ b/src/include/storage/buf_internals.h
@@ -230,6 +230,11 @@ typedef union BufferDescPadded
#define BufferDescriptorGetContentLock(bdesc) \
((LWLock*) (&(bdesc)->content_lock))
+#define BufferGetExternalLSN(bufHdr) \
+ BufferExternalLSNs[(bufHdr)->buf_id]
+#define BufferSetExternalLSN(bufHdr, lsn) \
+ BufferExternalLSNs[(bufHdr)->buf_id] = (lsn)
+
extern PGDLLIMPORT ConditionVariableMinimallyPadded *BufferIOCVArray;
/*
@@ -276,6 +281,7 @@ typedef struct WritebackContext
/* in buf_init.c */
extern PGDLLIMPORT BufferDescPadded *BufferDescriptors;
+extern PGDLLIMPORT XLogRecPtr *BufferExternalLSNs;
extern PGDLLIMPORT WritebackContext BackendWritebackContext;
/* in localbuf.c */
diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h
index 58391406f6..ce8d59e2b1 100644
--- a/src/include/storage/bufmgr.h
+++ b/src/include/storage/bufmgr.h
@@ -186,12 +186,21 @@ extern Buffer ReadBufferWithoutRelcache(RelFileNode rnode,
ForkNumber forkNum, BlockNumber blockNum,
ReadBufferMode mode, BufferAccessStrategy strategy,
bool permanent);
+extern Buffer ReadBufferWithoutRelcacheWithHit(RelFileNode rnode,
+ ForkNumber forkNum, BlockNumber blockNum,
+ ReadBufferMode mode, bool *hit);
extern void ReleaseBuffer(Buffer buffer);
extern void UnlockReleaseBuffer(Buffer buffer);
extern void MarkBufferDirty(Buffer buffer);
extern void IncrBufferRefCount(Buffer buffer);
extern Buffer ReleaseAndReadBuffer(Buffer buffer, Relation relation,
BlockNumber blockNum);
+extern void DiscardBuffer(RelFileNode rnode,
+ ForkNumber forkNum,
+ BlockNumber blockNum);
+extern bool BufferProbe(RelFileNode rnode,
+ ForkNumber forkNum,
+ BlockNumber blockNum);
extern void InitBufferPool(void);
extern void InitBufferPoolAccess(void);
diff --git a/src/include/storage/lwlock.h b/src/include/storage/lwlock.h
index e8c91139f8..c6136f4660 100644
--- a/src/include/storage/lwlock.h
+++ b/src/include/storage/lwlock.h
@@ -166,14 +166,7 @@ extern void LWLockInitialize(LWLock *lock, int tranche_id);
*/
typedef enum BuiltinTrancheIds
{
- LWTRANCHE_XACT_BUFFER = NUM_INDIVIDUAL_LWLOCKS,
- LWTRANCHE_COMMITTS_BUFFER,
- LWTRANCHE_SUBTRANS_BUFFER,
- LWTRANCHE_MULTIXACTOFFSET_BUFFER,
- LWTRANCHE_MULTIXACTMEMBER_BUFFER,
- LWTRANCHE_NOTIFY_BUFFER,
- LWTRANCHE_SERIAL_BUFFER,
- LWTRANCHE_WAL_INSERT,
+ LWTRANCHE_WAL_INSERT = NUM_INDIVIDUAL_LWLOCKS,
LWTRANCHE_BUFFER_CONTENT,
LWTRANCHE_REPLICATION_ORIGIN_STATE,
LWTRANCHE_REPLICATION_SLOT_IO,
diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h
index 6b63c60fbd..d2ae6b5058 100644
--- a/src/include/storage/smgr.h
+++ b/src/include/storage/smgr.h
@@ -16,6 +16,7 @@
#include "lib/ilist.h"
#include "storage/block.h"
+#include "storage/fd.h"
#include "storage/relfilenode.h"
/*
@@ -68,6 +69,13 @@ typedef struct SMgrRelationData
int md_num_open_segs[MAX_FORKNUM + 1];
struct _MdfdVec *md_seg_fds[MAX_FORKNUM + 1];
+ /*
+ * for slru.c; we'll just cache one File for now.
+ * XXX Unionize with md stuff somehow so we don't waste space
+ */
+ int slru_file_segment;
+ File slru_file;
+
/* if unowned, list link in list of all unowned SMgrRelations */
dlist_node node;
} SMgrRelationData;
diff --git a/src/include/storage/sync.h b/src/include/storage/sync.h
index 9737e1eb67..76c3ebc8bd 100644
--- a/src/include/storage/sync.h
+++ b/src/include/storage/sync.h
@@ -35,11 +35,7 @@ typedef enum SyncRequestType
typedef enum SyncRequestHandler
{
SYNC_HANDLER_MD = 0,
- SYNC_HANDLER_CLOG,
- SYNC_HANDLER_COMMIT_TS,
- SYNC_HANDLER_MULTIXACT_OFFSET,
- SYNC_HANDLER_MULTIXACT_MEMBER,
- SYNC_HANDLER_NONE
+ SYNC_HANDLER_SLRU
} SyncRequestHandler;
/*
diff --git a/src/include/utils/pgstat_internal.h b/src/include/utils/pgstat_internal.h
index 9303d05427..eec150af73 100644
--- a/src/include/utils/pgstat_internal.h
+++ b/src/include/utils/pgstat_internal.h
@@ -14,6 +14,7 @@
#define PGSTAT_INTERNAL_H
+#include "access/slru.h"
#include "common/hashfn.h"
#include "lib/dshash.h"
#include "lib/ilist.h"
@@ -260,28 +261,6 @@ typedef struct PgStat_KindInfo
} PgStat_KindInfo;
-/*
- * List of SLRU names that we keep stats for. There is no central registry of
- * SLRUs, so we use this fixed list instead. The "other" entry is used for
- * all SLRUs without an explicit entry (e.g. SLRUs in extensions).
- *
- * This is only defined here so that SLRU_NUM_ELEMENTS is known for later type
- * definitions.
- */
-static const char *const slru_names[] = {
- "CommitTs",
- "MultiXactMember",
- "MultiXactOffset",
- "Notify",
- "Serial",
- "Subtrans",
- "Xact",
- "other" /* has to be last */
-};
-
-#define SLRU_NUM_ELEMENTS lengthof(slru_names)
-
-
/* ----------
* Types and definitions for different kinds of fixed-amount stats.
*
@@ -333,7 +312,7 @@ typedef struct PgStatShared_SLRU
{
/* lock protects ->stats */
LWLock lock;
- PgStat_SLRUStats stats[SLRU_NUM_ELEMENTS];
+ PgStat_SLRUStats stats[SLRU_NUM_RELS];
} PgStatShared_SLRU;
typedef struct PgStatShared_Wal
@@ -442,7 +421,7 @@ typedef struct PgStat_Snapshot
PgStat_CheckpointerStats checkpointer;
- PgStat_SLRUStats slru[SLRU_NUM_ELEMENTS];
+ PgStat_SLRUStats slru[SLRU_NUM_RELS];
PgStat_WalStats wal;
diff --git a/src/include/utils/wait_event.h b/src/include/utils/wait_event.h
index b578e2ec75..8c463b7d88 100644
--- a/src/include/utils/wait_event.h
+++ b/src/include/utils/wait_event.h
@@ -202,7 +202,7 @@ typedef enum
WAIT_EVENT_REPLICATION_SLOT_RESTORE_SYNC,
WAIT_EVENT_REPLICATION_SLOT_SYNC,
WAIT_EVENT_REPLICATION_SLOT_WRITE,
- WAIT_EVENT_SLRU_FLUSH_SYNC,
+ WAIT_EVENT_SLRU_FLUSH,
WAIT_EVENT_SLRU_READ,
WAIT_EVENT_SLRU_SYNC,
WAIT_EVENT_SLRU_WRITE,
--
2.30.2
^ permalink raw reply [nested|flat] 22+ messages in thread
* Re: SLRUs in the main buffer pool, redux
@ 2022-05-28 01:13 Thomas Munro <[email protected]>
parent: Thomas Munro <[email protected]>
2 siblings, 2 replies; 22+ messages in thread
From: Thomas Munro @ 2022-05-28 01:13 UTC (permalink / raw)
To: Heikki Linnakangas <[email protected]>; +Cc: pgsql-hackers
On Fri, May 27, 2022 at 11:24 PM Thomas Munro <[email protected]> wrote:
> Rebased, debugged and fleshed out a tiny bit more, but still with
> plenty of TODO notes and questions. I will talk about this idea at
> PGCon, so I figured it'd help to have a patch that actually applies,
> even if it doesn't work quite right yet. It's quite a large patch but
> that's partly because it removes a lot of lines...
FWIW, here are my PGCon slides about this:
https://speakerdeck.com/macdice/improving-the-slru-subsystem
There was a little bit of discussion on #pgcon-stream2 which I could
summarise as: can we figure out a way to keep parts of the CLOG pinned
so that backends don't have to do that for each lookup? Then CLOG
checks become simple reads. There may be some relation to the idea of
'nailing' btree root pages that I've heard of from a couple of people
now (with ProcSignalBarrier or something more fine grained along those
lines if you need to unnail anything). Something to think about.
I'm also wondering if it would be possible to do "optimistic" pinning
instead for reads that normally need only a pin, using some kind of
counter scheme with read barriers to tell you if the page might have
been evicted after you read the data...
^ permalink raw reply [nested|flat] 22+ messages in thread
* Re: SLRUs in the main buffer pool, redux
@ 2022-05-29 20:57 Andres Freund <[email protected]>
parent: Thomas Munro <[email protected]>
1 sibling, 0 replies; 22+ messages in thread
From: Andres Freund @ 2022-05-29 20:57 UTC (permalink / raw)
To: Thomas Munro <[email protected]>; +Cc: Heikki Linnakangas <[email protected]>; pgsql-hackers
Hi,
On 2022-05-28 13:13:20 +1200, Thomas Munro wrote:
> There was a little bit of discussion on #pgcon-stream2 which I could
> summarise as: can we figure out a way to keep parts of the CLOG pinned
> so that backends don't have to do that for each lookup? Then CLOG
> checks become simple reads.
Included in that is not needing to re-check that the identity of the buffer
changed since the last use and to not need a PrivateRefCountEntry. Neither is
cheap...
I'd structure it so that there's a small list of slru buffers that's pinned in
a "shared" mode. Entering the buffer into that increases the BufferDesc's
refcount, but is *not* memorialized in the backend's refcount structures,
because it's "owned by the SLRU".
> There may be some relation to the idea of
> 'nailing' btree root pages that I've heard of from a couple of people
> now (with ProcSignalBarrier or something more fine grained along those
> lines if you need to unnail anything). Something to think about.
I'm very doubtful it's a good idea to combine those things - I think it's
quite different to come up with a design for SLRUs, of which there's a
constant number and shared memory ownership datastructures, and btree root
pages etc, of which there are arbitrary many.
For the nbtree (and similar) cases, I think it'd make sense to give backends a
size-limited number of pages they can keep pinned, but in a backend local
way. With, as you suggest, a procsignal barrier or such to force release.
> I'm also wondering if it would be possible to do "optimistic" pinning
> instead for reads that normally need only a pin, using some kind of
> counter scheme with read barriers to tell you if the page might have
> been evicted after you read the data...
-many
That seems fragile and complicated, without, at least to me, a clear need.
Greetings,
Andres Freund
^ permalink raw reply [nested|flat] 22+ messages in thread
* Re: SLRUs in the main buffer pool, redux
@ 2022-06-16 17:13 Konstantin Knizhnik <[email protected]>
parent: Thomas Munro <[email protected]>
1 sibling, 1 reply; 22+ messages in thread
From: Konstantin Knizhnik @ 2022-06-16 17:13 UTC (permalink / raw)
To: Thomas Munro <[email protected]>; Heikki Linnakangas <[email protected]>; +Cc: pgsql-hackers
On 28.05.2022 04:13, Thomas Munro wrote:
> On Fri, May 27, 2022 at 11:24 PM Thomas Munro<[email protected]> wrote:
>> Rebased, debugged and fleshed out a tiny bit more, but still with
>> plenty of TODO notes and questions. I will talk about this idea at
>> PGCon, so I figured it'd help to have a patch that actually applies,
>> even if it doesn't work quite right yet. It's quite a large patch but
>> that's partly because it removes a lot of lines...
> FWIW, here are my PGCon slides about this:
> https://speakerdeck.com/macdice/improving-the-slru-subsystem
>
> There was a little bit of discussion on #pgcon-stream2 which I could
> summarise as: can we figure out a way to keep parts of the CLOG pinned
> so that backends don't have to do that for each lookup? Then CLOG
> checks become simple reads. There may be some relation to the idea of
> 'nailing' btree root pages that I've heard of from a couple of people
> now (with ProcSignalBarrier or something more fine grained along those
> lines if you need to unnail anything). Something to think about.
>
> I'm also wondering if it would be possible to do "optimistic" pinning
> instead for reads that normally need only a pin, using some kind of
> counter scheme with read barriers to tell you if the page might have
> been evicted after you read the data...
>
>
I wonder if there are some tests which can illustrate advantages of
storing SLRU pages in shared buffers?
In PgPro we had a customer which run PL-PgSql code with recursively
called function containing exception handling code. Each exception block
creates subtransaction
and subxids SLRU becomes bottleneck.
I have simulated this workload with large number subxids using the
following function:
create or replace function do_update(id integer, level integer) returns
void as $$
begin
  begin
    if level > 0 then
      perform do_update(id, level-1);
    else
      update pgbench_accounts SET abalance = abalance + 1 WHERE
aid = id;
    end if;
  exception WHEN OTHERS THEN
    raise notice '% %', SQLERRM, SQLSTATE;
  end;
end; $$ language plpgsql;
With the following test script:
   \set aid random(1, 1000)
    select do_update(:aid,100)
I got the following results:
knizhnik@xps:~/db$ pgbench postgres -f update.sql -c 10 -T 100 -P 1 -M
prepared
pgbench (15beta1)
starting vacuum...end.
progress: 1.0 s, 3030.8 tps, lat 3.238 ms stddev 1.110, 0 failed
progress: 2.0 s, 3018.0 tps, lat 3.303 ms stddev 1.088, 0 failed
progress: 3.0 s, 3000.4 tps, lat 3.329 ms stddev 1.063, 0 failed
progress: 4.0 s, 2855.6 tps, lat 3.494 ms stddev 1.152, 0 failed
progress: 5.0 s, 2747.0 tps, lat 3.631 ms stddev 1.306, 0 failed
progress: 6.0 s, 2664.0 tps, lat 3.743 ms stddev 1.410, 0 failed
progress: 7.0 s, 2498.0 tps, lat 3.992 ms stddev 1.659, 0 failed
...
progress: 93.0 s, 670.0 tps, lat 14.964 ms stddev 10.555, 0 failed
progress: 94.0 s, 615.0 tps, lat 16.222 ms stddev 11.419, 0 failed
progress: 95.0 s, 580.0 tps, lat 17.251 ms stddev 11.622, 0 failed
progress: 96.0 s, 568.0 tps, lat 17.582 ms stddev 11.679, 0 failed
progress: 97.0 s, 573.0 tps, lat 17.389 ms stddev 11.771, 0 failed
progress: 98.0 s, 611.0 tps, lat 16.428 ms stddev 11.768, 0 failed
progress: 99.0 s, 568.0 tps, lat 17.622 ms stddev 11.912, 0 failed
progress: 100.0 s, 568.0 tps, lat 17.631 ms stddev 11.672, 0 failed
tps = 1035.566054 (without initial connection time)
With Thomas patch results are the following:
progress: 1.0 s, 2949.8 tps, lat 3.332 ms stddev 1.285, 0 failed
progress: 2.0 s, 3009.1 tps, lat 3.317 ms stddev 1.077, 0 failed
progress: 3.0 s, 2993.6 tps, lat 3.338 ms stddev 1.099, 0 failed
progress: 4.0 s, 3034.4 tps, lat 3.291 ms stddev 1.056, 0 failed
...
progress: 97.0 s, 1113.0 tps, lat 8.972 ms stddev 3.885, 0 failed
progress: 98.0 s, 1138.0 tps, lat 8.803 ms stddev 3.496, 0 failed
progress: 99.0 s, 1174.8 tps, lat 8.471 ms stddev 3.875, 0 failed
progress: 100.0 s, 1094.1 tps, lat 9.123 ms stddev 3.842, 0 failed
tps = 2133.240094 (without initial connection time)
So there is still degrade of performance but smaller than in case of
vanilla and total TPS are almost two times higher.
And this is another example demonstrating degrade of performance from
presentation by Alexander Korotkov:
pgbench script:
\setaid random(1, 100000 * :scale)
\setbid random(1, 1 * :scale)
\settid random(1, 10 * :scale)
\setdelta random(-5000, 5000)
BEGIN;
INSERT INTOpgbench_history (tid, bid, aid, delta, mtime)
VALUES(:tid, :bid, :aid, :delta,CURRENT_TIMESTAMP);
SAVEPOINT s1;
INSERT INTOpgbench_history (tid, bid, aid, delta, mtime)
VALUES(:tid, :bid, :aid, :delta,CURRENT_TIMESTAMP);
....
SAVEPOINT sN;
INSERT INTOpgbench_history (tid, bid, aid, delta, mtime)
VALUES(:tid, :bid, :aid, :delta,CURRENT_TIMESTAMP);
SELECTpg_sleep(1.0);
END;
I wonder which workload can cause CLOG to become a bottleneck?
Usually Postgres uses hint bits to avoid clog access. So standard
pgbench doesn't demonstrate any degrade of performance even in case of
presence of long living transactions,
which keeps XMIN horizon.
Attachments:
[image/png] k0jVyQplzHekz3Kn.png (72.8K, ../../[email protected]/3-k0jVyQplzHekz3Kn.png)
download | view image
^ permalink raw reply [nested|flat] 22+ messages in thread
* Re: SLRUs in the main buffer pool, redux
@ 2022-06-20 16:21 Robert Haas <[email protected]>
parent: Konstantin Knizhnik <[email protected]>
0 siblings, 0 replies; 22+ messages in thread
From: Robert Haas @ 2022-06-20 16:21 UTC (permalink / raw)
To: Konstantin Knizhnik <[email protected]>; +Cc: Thomas Munro <[email protected]>; Heikki Linnakangas <[email protected]>; pgsql-hackers
On Thu, Jun 16, 2022 at 1:13 PM Konstantin Knizhnik <[email protected]> wrote:
> I wonder which workload can cause CLOG to become a bottleneck?
> Usually Postgres uses hint bits to avoid clog access. So standard pgbench doesn't demonstrate any degrade of performance even in case of presence of long living transactions,
> which keeps XMIN horizon.
I haven't done research on this in a number of years and a bunch of
other improvements have been made since then, but I remember
discovering that CLOG could become a bottleneck on standard pgbench
tests especially when using unlogged tables. The higher you raised the
scale factor, the more of a bottleneck CLOG became. That makes sense:
no matter what the scale factor is, you're constantly updating rows
that have not previously been hinted, but as the scale factor gets
bigger, those rows are likely to be older (in terms of XID age) on
average, and so you need to cache more CLOG buffers to maintain
performance. But the SLRU system provides no such flexibility: it
doesn't scale to large numbers of buffers the way the code is written,
and it certainly can't vary the number of buffers devoted to this
purpose at runtime. So I think that the approach we're talking about
here has potential in that sense.
However, another problem with the SLRU code is that it's old and
crufty and hard to work with. It's hard to imagine anyone being able
to improve things very much as long as that's the basis. I don't know
that whatever code Thomas has written or will write is better, but if
it is, that would be good, because I don't see a lot of improvement in
this area being possible otherwise.
--
Robert Haas
EDB: http://www.enterprisedb.com
^ permalink raw reply [nested|flat] 22+ messages in thread
* Re: SLRUs in the main buffer pool, redux
@ 2022-07-21 13:23 Yura Sokolov <[email protected]>
parent: Thomas Munro <[email protected]>
2 siblings, 1 reply; 22+ messages in thread
From: Yura Sokolov @ 2022-07-21 13:23 UTC (permalink / raw)
To: Thomas Munro <[email protected]>; Heikki Linnakangas <[email protected]>; +Cc: pgsql-hackers
Good day, Thomas
В Пт, 27/05/2022 в 23:24 +1200, Thomas Munro пишет:
> Rebased, debugged and fleshed out a tiny bit more, but still with
> plenty of TODO notes and questions. I will talk about this idea at
> PGCon, so I figured it'd help to have a patch that actually applies,
> even if it doesn't work quite right yet. It's quite a large patch but
> that's partly because it removes a lot of lines...
Looks like it have to be rebased again.
^ permalink raw reply [nested|flat] 22+ messages in thread
* Re: SLRUs in the main buffer pool, redux
@ 2022-07-25 06:54 Heikki Linnakangas <[email protected]>
parent: Yura Sokolov <[email protected]>
0 siblings, 2 replies; 22+ messages in thread
From: Heikki Linnakangas @ 2022-07-25 06:54 UTC (permalink / raw)
To: Yura Sokolov <[email protected]>; Thomas Munro <[email protected]>; +Cc: pgsql-hackers
On 21/07/2022 16:23, Yura Sokolov wrote:
> Good day, Thomas
>
> В Пт, 27/05/2022 в 23:24 +1200, Thomas Munro пишет:
>> Rebased, debugged and fleshed out a tiny bit more, but still with
>> plenty of TODO notes and questions. I will talk about this idea at
>> PGCon, so I figured it'd help to have a patch that actually applies,
>> even if it doesn't work quite right yet. It's quite a large patch but
>> that's partly because it removes a lot of lines...
>
> Looks like it have to be rebased again.
Here's a rebase.
I'll write a separate post with my thoughts on the high-level design of
this, but first a couple of more detailed issues:
In RecordTransactionCommit(), we enter a critical section, and then call
TransactionIdCommitTree() to update the CLOG pages. That now involves a
call to ReadBuffer_common(), which in turn calls
ResourceOwnerEnlargeBuffers(). That can fail, because it might require
allocating memory, which is forbidden in a critical section. I ran into
an assertion about that with "make check" when I was playing around with
a heavily modified version of this patch. Haven't seen it with your
original one, but I believe that's just luck.
Calling ResourceOwnerEnlargeBuffers() before entering the critical
section would probably fix that, although I'm a bit worried about having
the Enlarge call so far away from the point where it's needed.
> +void
> +CheckPointSLRU(void)
> +{
> + /* Ensure that directory entries for new files are on disk. */
> + for (int i = 0; i < lengthof(defs); ++i)
> + {
> + if (defs[i].synchronize)
> + fsync_fname(defs[i].path, true);
> + }
> +}
> +
Is it really necessary to fsync() the directories? We don't do that
today for the SLRUs, and we don't do it for the tablespace/db
directories holding relations.
- Heikki
Attachments:
[text/x-patch] v3-0001-WIP-Move-SLRU-data-into-the-regular-buffer-pool.patch (171.0K, ../../[email protected]/2-v3-0001-WIP-Move-SLRU-data-into-the-regular-buffer-pool.patch)
download | inline diff:
From 241a8e7610e699ddaeeb38066b347bdac5d5e535 Mon Sep 17 00:00:00 2001
From: Thomas Munro <[email protected]>
Date: Sun, 2 Jan 2022 11:39:23 +1300
Subject: [PATCH v3 1/1] WIP: Move SLRU data into the regular buffer pool.
Previously, slru.c managed 7 small buffer pools for special queue-like
data. It had very simple sizing, locking, mapping, replacement and
write back. Switch to the regular buffer pool, to benefit from the
standard infrastructure.
The layout of SLRU files on disk remains unchanged. File access is done
through new callback functions exposed through smgr.c.
Switching SLRU page format to "standard" format (and thus enabling
in-page LSNs and checksums) is left for a later project. For now, add
support for raw (headerless) pages in the buffer pool.
XXX Among the problems not yet addressed:
* SMgrRelation objects cache slru_file descriptors, but there is no
invalidation for when SLRUs wrap around so they might wrap around
* CLOG lookups are a bit slower if you have to do a buffer mapping table
lookup first; to do better than that you need a little cache, which
for now is single item, and could be generalised; maybe CLOG is
actually so hot that it needs summary data that can be access
atomically?
* CLOG LSNs are now one-per-page, which probably isn't good ("group_lsn"
dropped by this patch)
* "external LSNs" are in a new separate array, but could perhaps be
stuffed into BufferDescriptor
* The locking strategies for all SLRUs need to be reviewed
* Support for "extension" SLRUs (ie SLRUs owned by code in core) is
removed -- did anyone ever acutally use that? Code searches seem to
indicate no
* While committing, in a critical section, the buffer pinning code
allocates memory which isn't allowed; I commented that stuff out for
now; perhaps we can make the resowner stuff non-allocating?
* Something is occasionally corrupting memory (locking problem, or
bad pointer arithmetic?)
XXX Work in progress!
XXX Unfinished, experimental!
XXX Contains known bugs!
Discussion: https://postgr.es/m/CA%2BhUKGKAYze99B-jk9NoMp-2BDqAgiRC4oJv%2BbFxghNgdieq8Q%40mail.gmail.com
---
doc/src/sgml/monitoring.sgml | 56 +-
src/backend/access/transam/clog.c | 199 +--
src/backend/access/transam/commit_ts.c | 156 +-
src/backend/access/transam/multixact.c | 358 ++---
src/backend/access/transam/slru.c | 1734 ++++++----------------
src/backend/access/transam/subtrans.c | 108 +-
src/backend/access/transam/xlog.c | 19 +-
src/backend/commands/async.c | 67 +-
src/backend/postmaster/checkpointer.c | 18 +
src/backend/storage/buffer/buf_init.c | 17 +-
src/backend/storage/buffer/bufmgr.c | 151 +-
src/backend/storage/ipc/ipci.c | 4 -
src/backend/storage/lmgr/lwlock.c | 14 -
src/backend/storage/lmgr/lwlocknames.txt | 10 +-
src/backend/storage/lmgr/predicate.c | 62 +-
src/backend/storage/smgr/smgr.c | 20 +-
src/backend/storage/sync/sync.c | 22 +-
src/backend/utils/activity/pgstat_slru.c | 35 +-
src/backend/utils/activity/wait_event.c | 4 +-
src/backend/utils/adt/pgstatfuncs.c | 8 +-
src/backend/utils/mmgr/mcxt.c | 12 +-
src/include/access/clog.h | 6 -
src/include/access/commit_ts.h | 3 -
src/include/access/multixact.h | 3 -
src/include/access/slru.h | 193 +--
src/include/access/subtrans.h | 3 -
src/include/pgstat.h | 1 -
src/include/storage/buf_internals.h | 6 +
src/include/storage/bufmgr.h | 9 +
src/include/storage/lwlock.h | 9 +-
src/include/storage/smgr.h | 8 +
src/include/storage/sync.h | 6 +-
src/include/utils/pgstat_internal.h | 27 +-
src/include/utils/wait_event.h | 2 +-
34 files changed, 1022 insertions(+), 2328 deletions(-)
diff --git a/doc/src/sgml/monitoring.sgml b/doc/src/sgml/monitoring.sgml
index 7dbbab6f5c9..83ab5459339 100644
--- a/doc/src/sgml/monitoring.sgml
+++ b/doc/src/sgml/monitoring.sgml
@@ -1448,9 +1448,8 @@ postgres 27093 0.0 0.0 30096 2752 ? Ss 11:34 0:00 postgres: ser
<entry>Waiting for a write to a replication slot control file.</entry>
</row>
<row>
- <entry><literal>SLRUFlushSync</literal></entry>
- <entry>Waiting for SLRU data to reach durable storage during a checkpoint
- or database shutdown.</entry>
+ <entry><literal>SLRUFlush</literal></entry>
+ <entry>Initiating writeback of SLRU data.</entry>
</row>
<row>
<entry><literal>SLRURead</literal></entry>
@@ -1458,8 +1457,7 @@ postgres 27093 0.0 0.0 30096 2752 ? Ss 11:34 0:00 postgres: ser
</row>
<row>
<entry><literal>SLRUSync</literal></entry>
- <entry>Waiting for SLRU data to reach durable storage following a page
- write.</entry>
+ <entry>Waiting for SLRU data to reach durable storage.</entry>
</row>
<row>
<entry><literal>SLRUWrite</literal></entry>
@@ -1954,14 +1952,6 @@ postgres 27093 0.0 0.0 30096 2752 ? Ss 11:34 0:00 postgres: ser
<entry>Waiting to read or update the last value set for a
transaction commit timestamp.</entry>
</row>
- <row>
- <entry><literal>CommitTsBuffer</literal></entry>
- <entry>Waiting for I/O on a commit timestamp SLRU buffer.</entry>
- </row>
- <row>
- <entry><literal>CommitTsSLRU</literal></entry>
- <entry>Waiting to access the commit timestamp SLRU cache.</entry>
- </row>
<row>
<entry><literal>ControlFile</literal></entry>
<entry>Waiting to read or update the <filename>pg_control</filename>
@@ -1991,31 +1981,10 @@ postgres 27093 0.0 0.0 30096 2752 ? Ss 11:34 0:00 postgres: ser
<entry><literal>MultiXactGen</literal></entry>
<entry>Waiting to read or update shared multixact state.</entry>
</row>
- <row>
- <entry><literal>MultiXactMemberBuffer</literal></entry>
- <entry>Waiting for I/O on a multixact member SLRU buffer.</entry>
- </row>
- <row>
- <entry><literal>MultiXactMemberSLRU</literal></entry>
- <entry>Waiting to access the multixact member SLRU cache.</entry>
- </row>
- <row>
- <entry><literal>MultiXactOffsetBuffer</literal></entry>
- <entry>Waiting for I/O on a multixact offset SLRU buffer.</entry>
- </row>
- <row>
- <entry><literal>MultiXactOffsetSLRU</literal></entry>
- <entry>Waiting to access the multixact offset SLRU cache.</entry>
- </row>
<row>
<entry><literal>MultiXactTruncation</literal></entry>
<entry>Waiting to read or truncate multixact information.</entry>
</row>
- <row>
- <entry><literal>NotifyBuffer</literal></entry>
- <entry>Waiting for I/O on a <command>NOTIFY</command> message SLRU
- buffer.</entry>
- </row>
<row>
<entry><literal>NotifyQueue</literal></entry>
<entry>Waiting to read or update <command>NOTIFY</command> messages.</entry>
@@ -2025,11 +1994,6 @@ postgres 27093 0.0 0.0 30096 2752 ? Ss 11:34 0:00 postgres: ser
<entry>Waiting to update limit on <command>NOTIFY</command> message
storage.</entry>
</row>
- <row>
- <entry><literal>NotifySLRU</literal></entry>
- <entry>Waiting to access the <command>NOTIFY</command> message SLRU
- cache.</entry>
- </row>
<row>
<entry><literal>OidGen</literal></entry>
<entry>Waiting to allocate a new OID.</entry>
@@ -2114,11 +2078,6 @@ postgres 27093 0.0 0.0 30096 2752 ? Ss 11:34 0:00 postgres: ser
<entry><literal>ReplicationSlotIO</literal></entry>
<entry>Waiting for I/O on a replication slot.</entry>
</row>
- <row>
- <entry><literal>SerialBuffer</literal></entry>
- <entry>Waiting for I/O on a serializable transaction conflict SLRU
- buffer.</entry>
- </row>
<row>
<entry><literal>SerializableFinishedList</literal></entry>
<entry>Waiting to access the list of finished serializable
@@ -2146,11 +2105,6 @@ postgres 27093 0.0 0.0 30096 2752 ? Ss 11:34 0:00 postgres: ser
<entry>Waiting to read or update information about serializable
transactions.</entry>
</row>
- <row>
- <entry><literal>SerialSLRU</literal></entry>
- <entry>Waiting to access the serializable transaction conflict SLRU
- cache.</entry>
- </row>
<row>
<entry><literal>SharedTidBitmap</literal></entry>
<entry>Waiting to access a shared TID bitmap during a parallel bitmap
@@ -2218,10 +2172,6 @@ postgres 27093 0.0 0.0 30096 2752 ? Ss 11:34 0:00 postgres: ser
<entry>Waiting to update limits on transaction id and multixact
consumption.</entry>
</row>
- <row>
- <entry><literal>XactBuffer</literal></entry>
- <entry>Waiting for I/O on a transaction status SLRU buffer.</entry>
- </row>
<row>
<entry><literal>XactSLRU</literal></entry>
<entry>Waiting to access the transaction status SLRU cache.</entry>
diff --git a/src/backend/access/transam/clog.c b/src/backend/access/transam/clog.c
index 3d9088a7048..cbe537ae72a 100644
--- a/src/backend/access/transam/clog.c
+++ b/src/backend/access/transam/clog.c
@@ -41,6 +41,8 @@
#include "miscadmin.h"
#include "pg_trace.h"
#include "pgstat.h"
+#include "storage/bufmgr.h"
+#include "storage/buf_internals.h"
#include "storage/proc.h"
#include "storage/sync.h"
@@ -81,15 +83,8 @@
*/
#define THRESHOLD_SUBTRANS_CLOG_OPT 5
-/*
- * Link to shared-memory data structures for CLOG control
- */
-static SlruCtlData XactCtlData;
-
-#define XactCtl (&XactCtlData)
-
-static int ZeroCLOGPage(int pageno, bool writeXlog);
+static Buffer ZeroCLOGPage(int pageno, bool writeXlog);
static bool CLOGPagePrecedes(int page1, int page2);
static void WriteZeroPageXlogRec(int pageno);
static void WriteTruncateXlogRec(int pageno, TransactionId oldestXact,
@@ -99,7 +94,7 @@ static void TransactionIdSetPageStatus(TransactionId xid, int nsubxids,
XLogRecPtr lsn, int pageno,
bool all_xact_same_page);
static void TransactionIdSetStatusBit(TransactionId xid, XidStatus status,
- XLogRecPtr lsn, int slotno);
+ XLogRecPtr lsn, Buffer buffer);
static void set_status_by_pages(int nsubxids, TransactionId *subxids,
XidStatus status, XLogRecPtr lsn);
static bool TransactionGroupUpdateXidStatus(TransactionId xid,
@@ -341,13 +336,12 @@ TransactionIdSetPageStatusInternal(TransactionId xid, int nsubxids,
TransactionId *subxids, XidStatus status,
XLogRecPtr lsn, int pageno)
{
- int slotno;
+ Buffer buffer;
int i;
Assert(status == TRANSACTION_STATUS_COMMITTED ||
status == TRANSACTION_STATUS_ABORTED ||
(status == TRANSACTION_STATUS_SUB_COMMITTED && !TransactionIdIsValid(xid)));
- Assert(LWLockHeldByMeInMode(XactSLRULock, LW_EXCLUSIVE));
/*
* If we're doing an async commit (ie, lsn is valid), then we must wait
@@ -358,7 +352,8 @@ TransactionIdSetPageStatusInternal(TransactionId xid, int nsubxids,
* write-busy, since we don't care if the update reaches disk sooner than
* we think.
*/
- slotno = SimpleLruReadPage(XactCtl, pageno, XLogRecPtrIsInvalid(lsn), xid);
+ buffer = ReadSlruBuffer(SLRU_CLOG_REL_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
/*
* Set the main transaction id, if any.
@@ -376,25 +371,26 @@ TransactionIdSetPageStatusInternal(TransactionId xid, int nsubxids,
{
for (i = 0; i < nsubxids; i++)
{
- Assert(XactCtl->shared->page_number[slotno] == TransactionIdToPage(subxids[i]));
+ Assert(pageno == TransactionIdToPage(subxids[i]));
TransactionIdSetStatusBit(subxids[i],
TRANSACTION_STATUS_SUB_COMMITTED,
- lsn, slotno);
+ lsn, buffer);
}
}
/* ... then the main transaction */
- TransactionIdSetStatusBit(xid, status, lsn, slotno);
+ TransactionIdSetStatusBit(xid, status, lsn, buffer);
}
/* Set the subtransactions */
for (i = 0; i < nsubxids; i++)
{
- Assert(XactCtl->shared->page_number[slotno] == TransactionIdToPage(subxids[i]));
- TransactionIdSetStatusBit(subxids[i], status, lsn, slotno);
+ Assert(pageno == TransactionIdToPage(subxids[i]));
+ TransactionIdSetStatusBit(subxids[i], status, lsn, buffer);
}
- XactCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -568,7 +564,7 @@ TransactionGroupUpdateXidStatus(TransactionId xid, XidStatus status,
* Must be called with XactSLRULock held
*/
static void
-TransactionIdSetStatusBit(TransactionId xid, XidStatus status, XLogRecPtr lsn, int slotno)
+TransactionIdSetStatusBit(TransactionId xid, XidStatus status, XLogRecPtr lsn, Buffer buffer)
{
int byteno = TransactionIdToByte(xid);
int bshift = TransactionIdToBIndex(xid) * CLOG_BITS_PER_XACT;
@@ -576,7 +572,10 @@ TransactionIdSetStatusBit(TransactionId xid, XidStatus status, XLogRecPtr lsn, i
char byteval;
char curval;
- byteptr = XactCtl->shared->page_buffer[slotno] + byteno;
+ Assert(LWLockHeldByMeInMode(BufferDescriptorGetContentLock(GetBufferDescriptor(buffer - 1)),
+ LW_EXCLUSIVE));
+
+ byteptr = BufferGetPage(buffer) + byteno;
curval = (*byteptr >> bshift) & CLOG_XACT_BITMASK;
/*
@@ -605,7 +604,7 @@ TransactionIdSetStatusBit(TransactionId xid, XidStatus status, XLogRecPtr lsn, i
*byteptr = byteval;
/*
- * Update the group LSN if the transaction completion LSN is higher.
+ * Update the buffer LSN if the transaction completion LSN is higher.
*
* Note: lsn will be invalid when supplied during InRecovery processing,
* so we don't need to do anything special to avoid LSN updates during
@@ -614,10 +613,8 @@ TransactionIdSetStatusBit(TransactionId xid, XidStatus status, XLogRecPtr lsn, i
*/
if (!XLogRecPtrIsInvalid(lsn))
{
- int lsnindex = GetLSNIndex(slotno, xid);
-
- if (XactCtl->shared->group_lsn[lsnindex] < lsn)
- XactCtl->shared->group_lsn[lsnindex] = lsn;
+ if (BufferGetExternalLSN(GetBufferDescriptor(buffer)) < lsn)
+ BufferSetExternalLSN(GetBufferDescriptor(buffer), lsn);
}
}
@@ -642,67 +639,22 @@ TransactionIdGetStatus(TransactionId xid, XLogRecPtr *lsn)
int pageno = TransactionIdToPage(xid);
int byteno = TransactionIdToByte(xid);
int bshift = TransactionIdToBIndex(xid) * CLOG_BITS_PER_XACT;
- int slotno;
- int lsnindex;
char *byteptr;
XidStatus status;
+ Buffer buffer;
- /* lock is acquired by SimpleLruReadPage_ReadOnly */
-
- slotno = SimpleLruReadPage_ReadOnly(XactCtl, pageno, xid);
- byteptr = XactCtl->shared->page_buffer[slotno] + byteno;
+ buffer = ReadSlruBuffer(SLRU_CLOG_REL_ID, pageno);
+ byteptr = BufferGetPage(buffer) + byteno;
status = (*byteptr >> bshift) & CLOG_XACT_BITMASK;
- lsnindex = GetLSNIndex(slotno, xid);
- *lsn = XactCtl->shared->group_lsn[lsnindex];
+ *lsn = BufferGetExternalLSN(GetBufferDescriptor(buffer));
- LWLockRelease(XactSLRULock);
+ ReleaseBuffer(buffer);
return status;
}
-/*
- * Number of shared CLOG buffers.
- *
- * On larger multi-processor systems, it is possible to have many CLOG page
- * requests in flight at one time which could lead to disk access for CLOG
- * page if the required page is not found in memory. Testing revealed that we
- * can get the best performance by having 128 CLOG buffers, more than that it
- * doesn't improve performance.
- *
- * Unconditionally keeping the number of CLOG buffers to 128 did not seem like
- * a good idea, because it would increase the minimum amount of shared memory
- * required to start, which could be a problem for people running very small
- * configurations. The following formula seems to represent a reasonable
- * compromise: people with very low values for shared_buffers will get fewer
- * CLOG buffers as well, and everyone else will get 128.
- */
-Size
-CLOGShmemBuffers(void)
-{
- return Min(128, Max(4, NBuffers / 512));
-}
-
-/*
- * Initialization of shared memory for CLOG
- */
-Size
-CLOGShmemSize(void)
-{
- return SimpleLruShmemSize(CLOGShmemBuffers(), CLOG_LSNS_PER_PAGE);
-}
-
-void
-CLOGShmemInit(void)
-{
- XactCtl->PagePrecedes = CLOGPagePrecedes;
- SimpleLruInit(XactCtl, "Xact", CLOGShmemBuffers(), CLOG_LSNS_PER_PAGE,
- XactSLRULock, "pg_xact", LWTRANCHE_XACT_BUFFER,
- SYNC_HANDLER_CLOG);
- SlruPagePrecedesUnitTests(XactCtl, CLOG_XACTS_PER_PAGE);
-}
-
/*
* This func must be called ONCE on system install. It creates
* the initial CLOG segment. (The CLOG directory is assumed to
@@ -712,18 +664,15 @@ CLOGShmemInit(void)
void
BootStrapCLOG(void)
{
- int slotno;
-
- LWLockAcquire(XactSLRULock, LW_EXCLUSIVE);
+ Buffer buffer;
/* Create and zero the first page of the commit log */
- slotno = ZeroCLOGPage(0, false);
+ buffer = ZeroCLOGPage(0, false);
/* Make sure it's written out */
- SimpleLruWritePage(XactCtl, slotno);
- Assert(!XactCtl->shared->page_dirty[slotno]);
+ FlushOneBuffer(buffer);
- LWLockRelease(XactSLRULock);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -735,17 +684,18 @@ BootStrapCLOG(void)
*
* Control lock must be held at entry, and will be held at exit.
*/
-static int
+static Buffer
ZeroCLOGPage(int pageno, bool writeXlog)
{
- int slotno;
+ Buffer buffer;
- slotno = SimpleLruZeroPage(XactCtl, pageno);
+ buffer = ZeroSlruBuffer(SLRU_CLOG_REL_ID, pageno);
+ MarkBufferDirty(buffer);
if (writeXlog)
WriteZeroPageXlogRec(pageno);
- return slotno;
+ return buffer;
}
/*
@@ -755,17 +705,6 @@ ZeroCLOGPage(int pageno, bool writeXlog)
void
StartupCLOG(void)
{
- TransactionId xid = XidFromFullTransactionId(ShmemVariableCache->nextXid);
- int pageno = TransactionIdToPage(xid);
-
- LWLockAcquire(XactSLRULock, LW_EXCLUSIVE);
-
- /*
- * Initialize our idea of the latest page number.
- */
- XactCtl->shared->latest_page_number = pageno;
-
- LWLockRelease(XactSLRULock);
}
/*
@@ -777,8 +716,6 @@ TrimCLOG(void)
TransactionId xid = XidFromFullTransactionId(ShmemVariableCache->nextXid);
int pageno = TransactionIdToPage(xid);
- LWLockAcquire(XactSLRULock, LW_EXCLUSIVE);
-
/*
* Zero out the remainder of the current clog page. Under normal
* circumstances it should be zeroes already, but it seems at least
@@ -795,40 +732,24 @@ TrimCLOG(void)
{
int byteno = TransactionIdToByte(xid);
int bshift = TransactionIdToBIndex(xid) * CLOG_BITS_PER_XACT;
- int slotno;
char *byteptr;
+ Buffer buffer;
- slotno = SimpleLruReadPage(XactCtl, pageno, false, xid);
- byteptr = XactCtl->shared->page_buffer[slotno] + byteno;
+ buffer = ReadSlruBuffer(SLRU_CLOG_REL_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ byteptr = BufferGetPage(buffer) + byteno;
/* Zero so-far-unused positions in the current byte */
*byteptr &= (1 << bshift) - 1;
/* Zero the rest of the page */
MemSet(byteptr + 1, 0, BLCKSZ - byteno - 1);
- XactCtl->shared->page_dirty[slotno] = true;
- }
+ MarkBufferDirty(buffer);
- LWLockRelease(XactSLRULock);
-}
-
-/*
- * Perform a checkpoint --- either during shutdown, or on-the-fly
- */
-void
-CheckPointCLOG(void)
-{
- /*
- * Write dirty CLOG pages to disk. This may result in sync requests
- * queued for later handling by ProcessSyncRequests(), as part of the
- * checkpoint.
- */
- TRACE_POSTGRESQL_CLOG_CHECKPOINT_START(true);
- SimpleLruWriteAll(XactCtl, true);
- TRACE_POSTGRESQL_CLOG_CHECKPOINT_DONE(true);
+ UnlockReleaseBuffer(buffer);
+ }
}
-
/*
* Make sure that CLOG has room for a newly-allocated XID.
*
@@ -852,12 +773,8 @@ ExtendCLOG(TransactionId newestXact)
pageno = TransactionIdToPage(newestXact);
- LWLockAcquire(XactSLRULock, LW_EXCLUSIVE);
-
/* Zero the page and make an XLOG entry about it */
- ZeroCLOGPage(pageno, true);
-
- LWLockRelease(XactSLRULock);
+ UnlockReleaseBuffer(ZeroCLOGPage(pageno, true));
}
@@ -888,7 +805,8 @@ TruncateCLOG(TransactionId oldestXact, Oid oldestxid_datoid)
cutoffPage = TransactionIdToPage(oldestXact);
/* Check to see if there's any files that could be removed */
- if (!SlruScanDirectory(XactCtl, SlruScanDirCbReportPresence, &cutoffPage))
+ if (!SlruScanDirectory(SLRU_CLOG_REL_ID, CLOGPagePrecedes,
+ SlruScanDirCbReportPresence, &cutoffPage))
return; /* nothing to remove */
/*
@@ -909,7 +827,7 @@ TruncateCLOG(TransactionId oldestXact, Oid oldestxid_datoid)
WriteTruncateXlogRec(cutoffPage, oldestXact, oldestxid_datoid);
/* Now we can remove the old CLOG segment(s) */
- SimpleLruTruncate(XactCtl, cutoffPage);
+ SimpleLruTruncate(SLRU_CLOG_REL_ID, CLOGPagePrecedes, cutoffPage);
}
@@ -994,17 +912,13 @@ clog_redo(XLogReaderState *record)
if (info == CLOG_ZEROPAGE)
{
int pageno;
- int slotno;
+ Buffer buffer;
memcpy(&pageno, XLogRecGetData(record), sizeof(int));
- LWLockAcquire(XactSLRULock, LW_EXCLUSIVE);
-
- slotno = ZeroCLOGPage(pageno, false);
- SimpleLruWritePage(XactCtl, slotno);
- Assert(!XactCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(XactSLRULock);
+ buffer = ZeroCLOGPage(pageno, false);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
else if (info == CLOG_TRUNCATE)
{
@@ -1014,17 +928,8 @@ clog_redo(XLogReaderState *record)
AdvanceOldestClogXid(xlrec.oldestXact);
- SimpleLruTruncate(XactCtl, xlrec.pageno);
+ SimpleLruTruncate(SLRU_CLOG_REL_ID, CLOGPagePrecedes, xlrec.pageno);
}
else
elog(PANIC, "clog_redo: unknown op code %u", info);
}
-
-/*
- * Entrypoint for sync.c to sync clog files.
- */
-int
-clogsyncfiletag(const FileTag *ftag, char *path)
-{
- return SlruSyncFileTag(XactCtl, ftag, path);
-}
diff --git a/src/backend/access/transam/commit_ts.c b/src/backend/access/transam/commit_ts.c
index 4dc8d402bd3..1eac761b393 100644
--- a/src/backend/access/transam/commit_ts.c
+++ b/src/backend/access/transam/commit_ts.c
@@ -73,13 +73,6 @@ typedef struct CommitTimestampEntry
#define TransactionIdToCTsEntry(xid) \
((xid) % (TransactionId) COMMIT_TS_XACTS_PER_PAGE)
-/*
- * Link to shared-memory data structures for CommitTs control
- */
-static SlruCtlData CommitTsCtlData;
-
-#define CommitTsCtl (&CommitTsCtlData)
-
/*
* We keep a cache of the last value set in shared memory.
*
@@ -110,7 +103,7 @@ static void SetXidCommitTsInPage(TransactionId xid, int nsubxids,
static void TransactionIdSetCommitTs(TransactionId xid, TimestampTz ts,
RepOriginId nodeid, int slotno);
static void error_commit_ts_disabled(void);
-static int ZeroCommitTsPage(int pageno, bool writeXlog);
+static Buffer ZeroCommitTsPage(int pageno, bool writeXlog);
static bool CommitTsPagePrecedes(int page1, int page2);
static void ActivateCommitTs(void);
static void DeactivateCommitTs(void);
@@ -219,30 +212,27 @@ SetXidCommitTsInPage(TransactionId xid, int nsubxids,
TransactionId *subxids, TimestampTz ts,
RepOriginId nodeid, int pageno)
{
- int slotno;
int i;
+ Buffer buffer;
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
-
- slotno = SimpleLruReadPage(CommitTsCtl, pageno, true, xid);
+ buffer = ReadSlruBuffer(SLRU_COMMITTS_REL_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
- TransactionIdSetCommitTs(xid, ts, nodeid, slotno);
+ TransactionIdSetCommitTs(xid, ts, nodeid, buffer);
for (i = 0; i < nsubxids; i++)
- TransactionIdSetCommitTs(subxids[i], ts, nodeid, slotno);
+ TransactionIdSetCommitTs(subxids[i], ts, nodeid, buffer);
- CommitTsCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
- LWLockRelease(CommitTsSLRULock);
+ UnlockReleaseBuffer(buffer);
}
/*
* Sets the commit timestamp of a single transaction.
- *
- * Must be called with CommitTsSLRULock held
*/
static void
TransactionIdSetCommitTs(TransactionId xid, TimestampTz ts,
- RepOriginId nodeid, int slotno)
+ RepOriginId nodeid, Buffer buffer)
{
int entryno = TransactionIdToCTsEntry(xid);
CommitTimestampEntry entry;
@@ -252,8 +242,7 @@ TransactionIdSetCommitTs(TransactionId xid, TimestampTz ts,
entry.time = ts;
entry.nodeid = nodeid;
- memcpy(CommitTsCtl->shared->page_buffer[slotno] +
- SizeOfCommitTimestampEntry * entryno,
+ memcpy(BufferGetPage(buffer) + SizeOfCommitTimestampEntry * entryno,
&entry, SizeOfCommitTimestampEntry);
}
@@ -271,10 +260,10 @@ TransactionIdGetCommitTsData(TransactionId xid, TimestampTz *ts,
{
int pageno = TransactionIdToCTsPage(xid);
int entryno = TransactionIdToCTsEntry(xid);
- int slotno;
CommitTimestampEntry entry;
TransactionId oldestCommitTsXid;
TransactionId newestCommitTsXid;
+ Buffer buffer;
if (!TransactionIdIsValid(xid))
ereport(ERROR,
@@ -328,10 +317,11 @@ TransactionIdGetCommitTsData(TransactionId xid, TimestampTz *ts,
return false;
}
- /* lock is acquired by SimpleLruReadPage_ReadOnly */
- slotno = SimpleLruReadPage_ReadOnly(CommitTsCtl, pageno, xid);
+ buffer = ReadSlruBuffer(SLRU_COMMITTS_REL_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_SHARE);
+
memcpy(&entry,
- CommitTsCtl->shared->page_buffer[slotno] +
+ BufferGetPage(buffer) +
SizeOfCommitTimestampEntry * entryno,
SizeOfCommitTimestampEntry);
@@ -339,7 +329,7 @@ TransactionIdGetCommitTsData(TransactionId xid, TimestampTz *ts,
if (nodeid)
*nodeid = entry.nodeid;
- LWLockRelease(CommitTsSLRULock);
+ UnlockReleaseBuffer(buffer);
return *ts != 0;
}
@@ -508,27 +498,13 @@ pg_xact_commit_timestamp_origin(PG_FUNCTION_ARGS)
PG_RETURN_DATUM(HeapTupleGetDatum(htup));
}
-/*
- * Number of shared CommitTS buffers.
- *
- * We use a very similar logic as for the number of CLOG buffers (except we
- * scale up twice as fast with shared buffers, and the maximum is twice as
- * high); see comments in CLOGShmemBuffers.
- */
-Size
-CommitTsShmemBuffers(void)
-{
- return Min(256, Max(4, NBuffers / 256));
-}
-
/*
* Shared memory sizing for CommitTs
*/
Size
CommitTsShmemSize(void)
{
- return SimpleLruShmemSize(CommitTsShmemBuffers(), 0) +
- sizeof(CommitTimestampShared);
+ return sizeof(CommitTimestampShared);
}
/*
@@ -540,12 +516,7 @@ CommitTsShmemInit(void)
{
bool found;
- CommitTsCtl->PagePrecedes = CommitTsPagePrecedes;
- SimpleLruInit(CommitTsCtl, "CommitTs", CommitTsShmemBuffers(), 0,
- CommitTsSLRULock, "pg_commit_ts",
- LWTRANCHE_COMMITTS_BUFFER,
- SYNC_HANDLER_COMMIT_TS);
- SlruPagePrecedesUnitTests(CommitTsCtl, COMMIT_TS_XACTS_PER_PAGE);
+ SlruPagePrecedesUnitTests(CommitTsPagePrecedes, COMMIT_TS_XACTS_PER_PAGE);
commitTsShared = ShmemInitStruct("CommitTs shared",
sizeof(CommitTimestampShared),
@@ -589,17 +560,18 @@ BootStrapCommitTs(void)
*
* Control lock must be held at entry, and will be held at exit.
*/
-static int
+static Buffer
ZeroCommitTsPage(int pageno, bool writeXlog)
{
- int slotno;
+ Buffer buffer;
- slotno = SimpleLruZeroPage(CommitTsCtl, pageno);
+ buffer = ZeroSlruBuffer(SLRU_COMMITTS_REL_ID, pageno);
+ MarkBufferDirty(buffer);
if (writeXlog)
WriteZeroPageXlogRec(pageno);
- return slotno;
+ return buffer;
}
/*
@@ -697,13 +669,6 @@ ActivateCommitTs(void)
xid = XidFromFullTransactionId(ShmemVariableCache->nextXid);
pageno = TransactionIdToCTsPage(xid);
- /*
- * Re-Initialize our idea of the latest page number.
- */
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
- CommitTsCtl->shared->latest_page_number = pageno;
- LWLockRelease(CommitTsSLRULock);
-
/*
* If CommitTs is enabled, but it wasn't in the previous server run, we
* need to set the oldest and newest values to the next Xid; that way, we
@@ -726,15 +691,14 @@ ActivateCommitTs(void)
LWLockRelease(CommitTsLock);
/* Create the current segment file, if necessary */
- if (!SimpleLruDoesPhysicalPageExist(CommitTsCtl, pageno))
+ if (!SimpleLruDoesPhysicalPageExist(SLRU_COMMITTS_REL_ID, pageno))
{
- int slotno;
+ Buffer buffer;
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
- slotno = ZeroCommitTsPage(pageno, false);
- SimpleLruWritePage(CommitTsCtl, slotno);
- Assert(!CommitTsCtl->shared->page_dirty[slotno]);
- LWLockRelease(CommitTsSLRULock);
+ buffer = ZeroSlruBuffer(SLRU_COMMITTS_REL_ID, pageno);
+ MarkBufferDirty(buffer);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
/* Change the activation status in shared memory. */
@@ -783,23 +747,9 @@ DeactivateCommitTs(void)
* be overwritten anyway when we wrap around, but it seems better to be
* tidy.)
*/
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
- (void) SlruScanDirectory(CommitTsCtl, SlruScanDirCbDeleteAll, NULL);
- LWLockRelease(CommitTsSLRULock);
-}
-
-/*
- * Perform a checkpoint --- either during shutdown, or on-the-fly
- */
-void
-CheckPointCommitTs(void)
-{
- /*
- * Write dirty CommitTs pages to disk. This may result in sync requests
- * queued for later handling by ProcessSyncRequests(), as part of the
- * checkpoint.
- */
- SimpleLruWriteAll(CommitTsCtl, true);
+ (void) SlruScanDirectory(SLRU_COMMITTS_REL_ID,
+ CommitTsPagePrecedes,
+ SlruScanDirCbDeleteAll, NULL);
}
/*
@@ -837,12 +787,8 @@ ExtendCommitTs(TransactionId newestXact)
pageno = TransactionIdToCTsPage(newestXact);
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
-
/* Zero the page and make an XLOG entry about it */
- ZeroCommitTsPage(pageno, !InRecovery);
-
- LWLockRelease(CommitTsSLRULock);
+ UnlockReleaseBuffer(ZeroCommitTsPage(pageno, !InRecovery));
}
/*
@@ -863,7 +809,9 @@ TruncateCommitTs(TransactionId oldestXact)
cutoffPage = TransactionIdToCTsPage(oldestXact);
/* Check to see if there's any files that could be removed */
- if (!SlruScanDirectory(CommitTsCtl, SlruScanDirCbReportPresence,
+ if (!SlruScanDirectory(SLRU_COMMITTS_REL_ID,
+ CommitTsPagePrecedes,
+ SlruScanDirCbReportPresence,
&cutoffPage))
return; /* nothing to remove */
@@ -871,7 +819,7 @@ TruncateCommitTs(TransactionId oldestXact)
WriteTruncateXlogRec(cutoffPage, oldestXact);
/* Now we can remove the old CommitTs segment(s) */
- SimpleLruTruncate(CommitTsCtl, cutoffPage);
+ SimpleLruTruncate(SLRU_COMMITTS_REL_ID, CommitTsPagePrecedes, cutoffPage);
}
/*
@@ -995,17 +943,14 @@ commit_ts_redo(XLogReaderState *record)
if (info == COMMIT_TS_ZEROPAGE)
{
int pageno;
- int slotno;
+ Buffer buffer;
memcpy(&pageno, XLogRecGetData(record), sizeof(int));
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
-
- slotno = ZeroCommitTsPage(pageno, false);
- SimpleLruWritePage(CommitTsCtl, slotno);
- Assert(!CommitTsCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(CommitTsSLRULock);
+ buffer = ZeroSlruBuffer(SLRU_COMMITTS_REL_ID, pageno);
+ MarkBufferDirty(buffer);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
else if (info == COMMIT_TS_TRUNCATE)
{
@@ -1013,23 +958,8 @@ commit_ts_redo(XLogReaderState *record)
AdvanceOldestCommitTsXid(trunc->oldestXid);
- /*
- * During XLOG replay, latest_page_number isn't set up yet; insert a
- * suitable value to bypass the sanity test in SimpleLruTruncate.
- */
- CommitTsCtl->shared->latest_page_number = trunc->pageno;
-
- SimpleLruTruncate(CommitTsCtl, trunc->pageno);
+ SimpleLruTruncate(SLRU_COMMITTS_REL_ID, CommitTsPagePrecedes, trunc->pageno);
}
else
elog(PANIC, "commit_ts_redo: unknown op code %u", info);
}
-
-/*
- * Entrypoint for sync.c to sync commit_ts files.
- */
-int
-committssyncfiletag(const FileTag *ftag, char *path)
-{
- return SlruSyncFileTag(CommitTsCtl, ftag, path);
-}
diff --git a/src/backend/access/transam/multixact.c b/src/backend/access/transam/multixact.c
index 8f7d12950e5..b35e73fc1f6 100644
--- a/src/backend/access/transam/multixact.c
+++ b/src/backend/access/transam/multixact.c
@@ -181,15 +181,6 @@
#define PreviousMultiXactId(xid) \
((xid) == FirstMultiXactId ? MaxMultiXactId : (xid) - 1)
-/*
- * Links to shared-memory data structures for MultiXact control
- */
-static SlruCtlData MultiXactOffsetCtlData;
-static SlruCtlData MultiXactMemberCtlData;
-
-#define MultiXactOffsetCtl (&MultiXactOffsetCtlData)
-#define MultiXactMemberCtl (&MultiXactMemberCtlData)
-
/*
* MultiXact state shared across all backends. All this state is protected
* by MultiXactGenLock. (We also use MultiXactOffsetSLRULock and
@@ -353,10 +344,9 @@ static void mXactCachePut(MultiXactId multi, int nmembers,
static char *mxstatus_to_string(MultiXactStatus status);
/* management of SLRU infrastructure */
-static int ZeroMultiXactOffsetPage(int pageno, bool writeXlog);
-static int ZeroMultiXactMemberPage(int pageno, bool writeXlog);
+static Buffer ZeroMultiXactOffsetPage(int pageno, bool writeXlog);
+static Buffer ZeroMultiXactMemberPage(int pageno, bool writeXlog);
static bool MultiXactOffsetPagePrecedes(int page1, int page2);
-static bool MultiXactMemberPagePrecedes(int page1, int page2);
static bool MultiXactOffsetPrecedes(MultiXactOffset offset1,
MultiXactOffset offset2);
static void ExtendMultiXactOffset(MultiXactId multi);
@@ -865,34 +855,25 @@ RecordNewMultiXact(MultiXactId multi, MultiXactOffset offset,
int pageno;
int prev_pageno;
int entryno;
- int slotno;
MultiXactOffset *offptr;
int i;
-
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
+ Buffer buffer;
pageno = MultiXactIdToOffsetPage(multi);
entryno = MultiXactIdToOffsetEntry(multi);
- /*
- * Note: we pass the MultiXactId to SimpleLruReadPage as the "transaction"
- * to complain about if there's any I/O error. This is kinda bogus, but
- * since the errors will always give the full pathname, it should be clear
- * enough that a MultiXactId is really involved. Perhaps someday we'll
- * take the trouble to generalize the slru.c error reporting code.
- */
- slotno = SimpleLruReadPage(MultiXactOffsetCtl, pageno, true, multi);
- offptr = (MultiXactOffset *) MultiXactOffsetCtl->shared->page_buffer[slotno];
+ /* XXX set up error context? */
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_OFFSET_REL_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ offptr = (MultiXactOffset *) BufferGetPage(buffer);
offptr += entryno;
*offptr = offset;
- MultiXactOffsetCtl->shared->page_dirty[slotno] = true;
-
- /* Exchange our lock */
- LWLockRelease(MultiXactOffsetSLRULock);
+ MarkBufferDirty(buffer);
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
+ UnlockReleaseBuffer(buffer);
+ buffer = InvalidBuffer;
prev_pageno = -1;
@@ -914,27 +895,28 @@ RecordNewMultiXact(MultiXactId multi, MultiXactOffset offset,
if (pageno != prev_pageno)
{
- slotno = SimpleLruReadPage(MultiXactMemberCtl, pageno, true, multi);
+ if (BufferIsValid(buffer))
+ UnlockReleaseBuffer(buffer);
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_MEMBER_REL_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
prev_pageno = pageno;
}
- memberptr = (TransactionId *)
- (MultiXactMemberCtl->shared->page_buffer[slotno] + memberoff);
+ memberptr = (TransactionId *) (BufferGetPage(buffer) + memberoff);
*memberptr = members[i].xid;
- flagsptr = (uint32 *)
- (MultiXactMemberCtl->shared->page_buffer[slotno] + flagsoff);
+ flagsptr = (uint32 *) (BufferGetPage(buffer) + flagsoff);
flagsval = *flagsptr;
flagsval &= ~(((1 << MXACT_MEMBER_BITS_PER_XACT) - 1) << bshift);
flagsval |= (members[i].status << bshift);
*flagsptr = flagsval;
- MultiXactMemberCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
}
- LWLockRelease(MultiXactMemberSLRULock);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -1226,7 +1208,6 @@ GetMultiXactIdMembers(MultiXactId multi, MultiXactMember **members,
int pageno;
int prev_pageno;
int entryno;
- int slotno;
MultiXactOffset *offptr;
MultiXactOffset offset;
int length;
@@ -1237,6 +1218,7 @@ GetMultiXactIdMembers(MultiXactId multi, MultiXactMember **members,
MultiXactId tmpMXact;
MultiXactOffset nextOffset;
MultiXactMember *ptr;
+ Buffer buffer;
debug_elog3(DEBUG2, "GetMembers: asked for %u", multi);
@@ -1340,13 +1322,12 @@ GetMultiXactIdMembers(MultiXactId multi, MultiXactMember **members,
* time on every multixact creation.
*/
retry:
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
-
pageno = MultiXactIdToOffsetPage(multi);
entryno = MultiXactIdToOffsetEntry(multi);
- slotno = SimpleLruReadPage(MultiXactOffsetCtl, pageno, true, multi);
- offptr = (MultiXactOffset *) MultiXactOffsetCtl->shared->page_buffer[slotno];
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_OFFSET_REL_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_SHARE);
+ offptr = (MultiXactOffset *) BufferGetPage(buffer);
offptr += entryno;
offset = *offptr;
@@ -1377,16 +1358,20 @@ retry:
entryno = MultiXactIdToOffsetEntry(tmpMXact);
if (pageno != prev_pageno)
- slotno = SimpleLruReadPage(MultiXactOffsetCtl, pageno, true, tmpMXact);
+ {
+ UnlockReleaseBuffer(buffer);
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_OFFSET_REL_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_SHARE);
+ }
- offptr = (MultiXactOffset *) MultiXactOffsetCtl->shared->page_buffer[slotno];
+ offptr = (MultiXactOffset *) BufferGetPage(buffer);
offptr += entryno;
nextMXOffset = *offptr;
if (nextMXOffset == 0)
{
/* Corner case 2: next multixact is still being filled in */
- LWLockRelease(MultiXactOffsetSLRULock);
+ UnlockReleaseBuffer(buffer);
CHECK_FOR_INTERRUPTS();
pg_usleep(1000L);
goto retry;
@@ -1394,14 +1379,11 @@ retry:
length = nextMXOffset - offset;
}
-
- LWLockRelease(MultiXactOffsetSLRULock);
+ UnlockReleaseBuffer(buffer);
+ buffer = InvalidBuffer;
ptr = (MultiXactMember *) palloc(length * sizeof(MultiXactMember));
- /* Now get the members themselves. */
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
-
truelength = 0;
prev_pageno = -1;
for (i = 0; i < length; i++, offset++)
@@ -1417,12 +1399,14 @@ retry:
if (pageno != prev_pageno)
{
- slotno = SimpleLruReadPage(MultiXactMemberCtl, pageno, true, multi);
+ if (BufferIsValid(buffer))
+ UnlockReleaseBuffer(buffer);
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_MEMBER_REL_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_SHARE);
prev_pageno = pageno;
}
- xactptr = (TransactionId *)
- (MultiXactMemberCtl->shared->page_buffer[slotno] + memberoff);
+ xactptr = (TransactionId *) (BufferGetPage(buffer) + memberoff);
if (!TransactionIdIsValid(*xactptr))
{
@@ -1433,14 +1417,13 @@ retry:
flagsoff = MXOffsetToFlagsOffset(offset);
bshift = MXOffsetToFlagsBitShift(offset);
- flagsptr = (uint32 *) (MultiXactMemberCtl->shared->page_buffer[slotno] + flagsoff);
+ flagsptr = (uint32 *) (BufferGetPage(buffer) + flagsoff);
ptr[truelength].xid = *xactptr;
ptr[truelength].status = (*flagsptr >> bshift) & MXACT_MEMBER_XACT_BITMASK;
truelength++;
}
-
- LWLockRelease(MultiXactMemberSLRULock);
+ UnlockReleaseBuffer(buffer);
/* A multixid with zero members should not happen */
Assert(truelength > 0);
@@ -1834,8 +1817,6 @@ MultiXactShmemSize(void)
mul_size(sizeof(MultiXactId) * 2, MaxOldestSlot))
size = SHARED_MULTIXACT_STATE_SIZE;
- size = add_size(size, SimpleLruShmemSize(NUM_MULTIXACTOFFSET_BUFFERS, 0));
- size = add_size(size, SimpleLruShmemSize(NUM_MULTIXACTMEMBER_BUFFERS, 0));
return size;
}
@@ -1847,22 +1828,6 @@ MultiXactShmemInit(void)
debug_elog2(DEBUG2, "Shared Memory Init for MultiXact");
- MultiXactOffsetCtl->PagePrecedes = MultiXactOffsetPagePrecedes;
- MultiXactMemberCtl->PagePrecedes = MultiXactMemberPagePrecedes;
-
- SimpleLruInit(MultiXactOffsetCtl,
- "MultiXactOffset", NUM_MULTIXACTOFFSET_BUFFERS, 0,
- MultiXactOffsetSLRULock, "pg_multixact/offsets",
- LWTRANCHE_MULTIXACTOFFSET_BUFFER,
- SYNC_HANDLER_MULTIXACT_OFFSET);
- SlruPagePrecedesUnitTests(MultiXactOffsetCtl, MULTIXACT_OFFSETS_PER_PAGE);
- SimpleLruInit(MultiXactMemberCtl,
- "MultiXactMember", NUM_MULTIXACTMEMBER_BUFFERS, 0,
- MultiXactMemberSLRULock, "pg_multixact/members",
- LWTRANCHE_MULTIXACTMEMBER_BUFFER,
- SYNC_HANDLER_MULTIXACT_MEMBER);
- /* doesn't call SimpleLruTruncate() or meet criteria for unit tests */
-
/* Initialize our shared state struct */
MultiXactState = ShmemInitStruct("Shared MultiXact State",
SHARED_MULTIXACT_STATE_SIZE,
@@ -1893,29 +1858,17 @@ MultiXactShmemInit(void)
void
BootStrapMultiXact(void)
{
- int slotno;
-
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
+ Buffer buffer;
/* Create and zero the first page of the offsets log */
- slotno = ZeroMultiXactOffsetPage(0, false);
-
- /* Make sure it's written out */
- SimpleLruWritePage(MultiXactOffsetCtl, slotno);
- Assert(!MultiXactOffsetCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(MultiXactOffsetSLRULock);
-
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
+ buffer = ZeroMultiXactOffsetPage(0, false);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
/* Create and zero the first page of the members log */
- slotno = ZeroMultiXactMemberPage(0, false);
-
- /* Make sure it's written out */
- SimpleLruWritePage(MultiXactMemberCtl, slotno);
- Assert(!MultiXactMemberCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(MultiXactMemberSLRULock);
+ buffer = ZeroMultiXactMemberPage(0, false);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -1927,33 +1880,35 @@ BootStrapMultiXact(void)
*
* Control lock must be held at entry, and will be held at exit.
*/
-static int
+static Buffer
ZeroMultiXactOffsetPage(int pageno, bool writeXlog)
{
- int slotno;
+ Buffer buffer;
- slotno = SimpleLruZeroPage(MultiXactOffsetCtl, pageno);
+ buffer = ZeroSlruBuffer(SLRU_MULTIXACT_OFFSET_REL_ID, pageno);
+ MarkBufferDirty(buffer);
if (writeXlog)
WriteMZeroPageXlogRec(pageno, XLOG_MULTIXACT_ZERO_OFF_PAGE);
- return slotno;
+ return buffer;
}
/*
* Ditto, for MultiXactMember
*/
-static int
+static Buffer
ZeroMultiXactMemberPage(int pageno, bool writeXlog)
{
- int slotno;
+ Buffer buffer;
- slotno = SimpleLruZeroPage(MultiXactMemberCtl, pageno);
+ buffer = ZeroSlruBuffer(SLRU_MULTIXACT_MEMBER_REL_ID, pageno);
+ MarkBufferDirty(buffer);
if (writeXlog)
WriteMZeroPageXlogRec(pageno, XLOG_MULTIXACT_ZERO_MEM_PAGE);
- return slotno;
+ return buffer;
}
/*
@@ -1978,22 +1933,14 @@ MaybeExtendOffsetSlru(void)
pageno = MultiXactIdToOffsetPage(MultiXactState->nextMXact);
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
-
- if (!SimpleLruDoesPhysicalPageExist(MultiXactOffsetCtl, pageno))
+ if (!SimpleLruDoesPhysicalPageExist(SLRU_MULTIXACT_OFFSET_REL_ID, pageno))
{
- int slotno;
+ Buffer buffer;
- /*
- * Fortunately for us, SimpleLruWritePage is already prepared to deal
- * with creating a new segment file even if the page we're writing is
- * not the first in it, so this is enough.
- */
- slotno = ZeroMultiXactOffsetPage(pageno, false);
- SimpleLruWritePage(MultiXactOffsetCtl, slotno);
+ buffer = ZeroMultiXactOffsetPage(pageno, false);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
-
- LWLockRelease(MultiXactOffsetSLRULock);
}
/*
@@ -2007,21 +1954,6 @@ MaybeExtendOffsetSlru(void)
void
StartupMultiXact(void)
{
- MultiXactId multi = MultiXactState->nextMXact;
- MultiXactOffset offset = MultiXactState->nextOffset;
- int pageno;
-
- /*
- * Initialize offset's idea of the latest page number.
- */
- pageno = MultiXactIdToOffsetPage(multi);
- MultiXactOffsetCtl->shared->latest_page_number = pageno;
-
- /*
- * Initialize member's idea of the latest page number.
- */
- pageno = MXOffsetToMemberPage(offset);
- MultiXactMemberCtl->shared->latest_page_number = pageno;
}
/*
@@ -2045,14 +1977,7 @@ TrimMultiXact(void)
oldestMXactDB = MultiXactState->oldestMultiXactDB;
LWLockRelease(MultiXactGenLock);
- /* Clean up offsets state */
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
-
- /*
- * (Re-)Initialize our idea of the latest page number for offsets.
- */
- pageno = MultiXactIdToOffsetPage(nextMXact);
- MultiXactOffsetCtl->shared->latest_page_number = pageno;
+ pageno = MXOffsetToMemberPage(offset);
/*
* Zero out the remainder of the current offsets page. See notes in
@@ -2065,29 +1990,20 @@ TrimMultiXact(void)
entryno = MultiXactIdToOffsetEntry(nextMXact);
if (entryno != 0)
{
- int slotno;
MultiXactOffset *offptr;
+ Buffer buffer;
- slotno = SimpleLruReadPage(MultiXactOffsetCtl, pageno, true, nextMXact);
- offptr = (MultiXactOffset *) MultiXactOffsetCtl->shared->page_buffer[slotno];
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_OFFSET_REL_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ offptr = (MultiXactOffset *) BufferGetPage(buffer);
offptr += entryno;
MemSet(offptr, 0, BLCKSZ - (entryno * sizeof(MultiXactOffset)));
- MultiXactOffsetCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
+ UnlockReleaseBuffer(buffer);
}
- LWLockRelease(MultiXactOffsetSLRULock);
-
- /* And the same for members */
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
-
- /*
- * (Re-)Initialize our idea of the latest page number for members.
- */
- pageno = MXOffsetToMemberPage(offset);
- MultiXactMemberCtl->shared->latest_page_number = pageno;
-
/*
* Zero out the remainder of the current members page. See notes in
* TrimCLOG() for motivation.
@@ -2095,14 +2011,14 @@ TrimMultiXact(void)
flagsoff = MXOffsetToFlagsOffset(offset);
if (flagsoff != 0)
{
- int slotno;
TransactionId *xidptr;
int memberoff;
+ Buffer buffer;
memberoff = MXOffsetToMemberOffset(offset);
- slotno = SimpleLruReadPage(MultiXactMemberCtl, pageno, true, offset);
- xidptr = (TransactionId *)
- (MultiXactMemberCtl->shared->page_buffer[slotno] + memberoff);
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_MEMBER_REL_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ xidptr = (TransactionId *) (BufferGetPage(buffer) + memberoff);
MemSet(xidptr, 0, BLCKSZ - memberoff);
@@ -2112,11 +2028,10 @@ TrimMultiXact(void)
* writing.
*/
- MultiXactMemberCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
+ UnlockReleaseBuffer(buffer);
}
- LWLockRelease(MultiXactMemberSLRULock);
-
/* signal that we're officially up */
LWLockAcquire(MultiXactGenLock, LW_EXCLUSIVE);
MultiXactState->finishedStartup = true;
@@ -2148,25 +2063,6 @@ MultiXactGetCheckptMulti(bool is_shutdown,
*nextMulti, *nextMultiOffset, *oldestMulti, *oldestMultiDB);
}
-/*
- * Perform a checkpoint --- either during shutdown, or on-the-fly
- */
-void
-CheckPointMultiXact(void)
-{
- TRACE_POSTGRESQL_MULTIXACT_CHECKPOINT_START(true);
-
- /*
- * Write dirty MultiXact pages to disk. This may result in sync requests
- * queued for later handling by ProcessSyncRequests(), as part of the
- * checkpoint.
- */
- SimpleLruWriteAll(MultiXactOffsetCtl, true);
- SimpleLruWriteAll(MultiXactMemberCtl, true);
-
- TRACE_POSTGRESQL_MULTIXACT_CHECKPOINT_DONE(true);
-}
-
/*
* Set the next-to-be-assigned MultiXactId and offset
*
@@ -2415,12 +2311,8 @@ ExtendMultiXactOffset(MultiXactId multi)
pageno = MultiXactIdToOffsetPage(multi);
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
-
/* Zero the page and make an XLOG entry about it */
- ZeroMultiXactOffsetPage(pageno, true);
-
- LWLockRelease(MultiXactOffsetSLRULock);
+ UnlockReleaseBuffer(ZeroMultiXactOffsetPage(pageno, true));
}
/*
@@ -2456,12 +2348,8 @@ ExtendMultiXactMember(MultiXactOffset offset, int nmembers)
pageno = MXOffsetToMemberPage(offset);
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
-
/* Zero the page and make an XLOG entry about it */
- ZeroMultiXactMemberPage(pageno, true);
-
- LWLockRelease(MultiXactMemberSLRULock);
+ UnlockReleaseBuffer(ZeroMultiXactMemberPage(pageno, true));
}
/*
@@ -2736,8 +2624,9 @@ find_multixact_start(MultiXactId multi, MultiXactOffset *result)
MultiXactOffset offset;
int pageno;
int entryno;
- int slotno;
MultiXactOffset *offptr;
+ Buffer buffer;
+ RelFileLocator rlocator = SlruRelFileLocator(SLRU_MULTIXACT_OFFSET_REL_ID);
Assert(MultiXactState->finishedStartup);
@@ -2745,20 +2634,19 @@ find_multixact_start(MultiXactId multi, MultiXactOffset *result)
entryno = MultiXactIdToOffsetEntry(multi);
/*
- * Write out dirty data, so PhysicalPageExists can work correctly.
+ * Cope with missing/bogus oldest MultiXact in inconsistent states (see
+ * commit 068cfadf9).
*/
- SimpleLruWriteAll(MultiXactOffsetCtl, true);
- SimpleLruWriteAll(MultiXactMemberCtl, true);
-
- if (!SimpleLruDoesPhysicalPageExist(MultiXactOffsetCtl, pageno))
+ if (!BufferProbe(rlocator, MAIN_FORKNUM, pageno) &&
+ !SimpleLruDoesPhysicalPageExist(SLRU_MULTIXACT_OFFSET_REL_ID, pageno))
return false;
- /* lock is acquired by SimpleLruReadPage_ReadOnly */
- slotno = SimpleLruReadPage_ReadOnly(MultiXactOffsetCtl, pageno, multi);
- offptr = (MultiXactOffset *) MultiXactOffsetCtl->shared->page_buffer[slotno];
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_OFFSET_REL_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_SHARE);
+ offptr = (MultiXactOffset *) BufferGetPage(buffer);
offptr += entryno;
offset = *offptr;
- LWLockRelease(MultiXactOffsetSLRULock);
+ UnlockReleaseBuffer(buffer);
*result = offset;
return true;
@@ -2865,12 +2753,13 @@ typedef struct mxtruncinfo
* This callback determines the earliest existing page number.
*/
static bool
-SlruScanDirCbFindEarliest(SlruCtl ctl, char *filename, int segpage, void *data)
+SlruScanDirCbFindEarliest(Oid rel_id, SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage, void *data)
{
mxtruncinfo *trunc = (mxtruncinfo *) data;
if (trunc->earliestExistingPage == -1 ||
- ctl->PagePrecedes(segpage, trunc->earliestExistingPage))
+ PagePrecedes(segpage, trunc->earliestExistingPage))
{
trunc->earliestExistingPage = segpage;
}
@@ -2902,7 +2791,7 @@ PerformMembersTruncation(MultiXactOffset oldestOffset, MultiXactOffset newOldest
while (segment != endsegment)
{
elog(DEBUG2, "truncating multixact members segment %x", segment);
- SlruDeleteSegment(MultiXactMemberCtl, segment);
+ SlruDeleteSegment(SLRU_MULTIXACT_MEMBER_REL_ID, segment);
/* move to next segment, handling wraparound correctly */
if (segment == maxsegment)
@@ -2925,7 +2814,8 @@ PerformOffsetsTruncation(MultiXactId oldestMulti, MultiXactId newOldestMulti)
* didn't subtract one, we'd trigger SimpleLruTruncate's wraparound
* detection.
*/
- SimpleLruTruncate(MultiXactOffsetCtl,
+ SimpleLruTruncate(SLRU_MULTIXACT_OFFSET_REL_ID,
+ MultiXactOffsetPagePrecedes,
MultiXactIdToOffsetPage(PreviousMultiXactId(newOldestMulti)));
}
@@ -2999,7 +2889,9 @@ TruncateMultiXact(MultiXactId newOldestMulti, Oid newOldestMultiDB)
* been truncated away, and we crashed before updating oldestMulti.
*/
trunc.earliestExistingPage = -1;
- SlruScanDirectory(MultiXactOffsetCtl, SlruScanDirCbFindEarliest, &trunc);
+ SlruScanDirectory(SLRU_MULTIXACT_OFFSET_REL_ID,
+ MultiXactOffsetPagePrecedes,
+ SlruScanDirCbFindEarliest, &trunc);
earliest = trunc.earliestExistingPage * MULTIXACT_OFFSETS_PER_PAGE;
if (earliest < FirstMultiXactId)
earliest = FirstMultiXactId;
@@ -3131,24 +3023,6 @@ MultiXactOffsetPagePrecedes(int page1, int page2)
multi2 + MULTIXACT_OFFSETS_PER_PAGE - 1));
}
-/*
- * Decide whether a MultiXactMember page number is "older" for truncation
- * purposes. There is no "invalid offset number" so use the numbers verbatim.
- */
-static bool
-MultiXactMemberPagePrecedes(int page1, int page2)
-{
- MultiXactOffset offset1;
- MultiXactOffset offset2;
-
- offset1 = ((MultiXactOffset) page1) * MULTIXACT_MEMBERS_PER_PAGE;
- offset2 = ((MultiXactOffset) page2) * MULTIXACT_MEMBERS_PER_PAGE;
-
- return (MultiXactOffsetPrecedes(offset1, offset2) &&
- MultiXactOffsetPrecedes(offset1,
- offset2 + MULTIXACT_MEMBERS_PER_PAGE - 1));
-}
-
/*
* Decide which of two MultiXactIds is earlier.
*
@@ -3243,32 +3117,18 @@ multixact_redo(XLogReaderState *record)
if (info == XLOG_MULTIXACT_ZERO_OFF_PAGE)
{
int pageno;
- int slotno;
memcpy(&pageno, XLogRecGetData(record), sizeof(int));
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
-
- slotno = ZeroMultiXactOffsetPage(pageno, false);
- SimpleLruWritePage(MultiXactOffsetCtl, slotno);
- Assert(!MultiXactOffsetCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(MultiXactOffsetSLRULock);
+ UnlockReleaseBuffer(ZeroMultiXactOffsetPage(pageno, false));
}
else if (info == XLOG_MULTIXACT_ZERO_MEM_PAGE)
{
int pageno;
- int slotno;
memcpy(&pageno, XLogRecGetData(record), sizeof(int));
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
-
- slotno = ZeroMultiXactMemberPage(pageno, false);
- SimpleLruWritePage(MultiXactMemberCtl, slotno);
- Assert(!MultiXactMemberCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(MultiXactMemberSLRULock);
+ UnlockReleaseBuffer(ZeroMultiXactMemberPage(pageno, false));
}
else if (info == XLOG_MULTIXACT_CREATE_ID)
{
@@ -3302,7 +3162,6 @@ multixact_redo(XLogReaderState *record)
else if (info == XLOG_MULTIXACT_TRUNCATE_ID)
{
xl_multixact_truncate xlrec;
- int pageno;
memcpy(&xlrec, XLogRecGetData(record),
SizeOfMultiXactTruncate);
@@ -3328,13 +3187,6 @@ multixact_redo(XLogReaderState *record)
PerformMembersTruncation(xlrec.startTruncMemb, xlrec.endTruncMemb);
- /*
- * During XLOG replay, latest_page_number isn't necessarily set up
- * yet; insert a suitable value to bypass the sanity test in
- * SimpleLruTruncate.
- */
- pageno = MultiXactIdToOffsetPage(xlrec.endTruncOff);
- MultiXactOffsetCtl->shared->latest_page_number = pageno;
PerformOffsetsTruncation(xlrec.startTruncOff, xlrec.endTruncOff);
LWLockRelease(MultiXactTruncationLock);
@@ -3407,21 +3259,3 @@ pg_get_multixact_members(PG_FUNCTION_ARGS)
SRF_RETURN_DONE(funccxt);
}
-
-/*
- * Entrypoint for sync.c to sync offsets files.
- */
-int
-multixactoffsetssyncfiletag(const FileTag *ftag, char *path)
-{
- return SlruSyncFileTag(MultiXactOffsetCtl, ftag, path);
-}
-
-/*
- * Entrypoint for sync.c to sync members files.
- */
-int
-multixactmemberssyncfiletag(const FileTag *ftag, char *path)
-{
- return SlruSyncFileTag(MultiXactMemberCtl, ftag, path);
-}
diff --git a/src/backend/access/transam/slru.c b/src/backend/access/transam/slru.c
index b65cb49d7ff..12eeca67e27 100644
--- a/src/backend/access/transam/slru.c
+++ b/src/backend/access/transam/slru.c
@@ -1,41 +1,9 @@
/*-------------------------------------------------------------------------
*
* slru.c
- * Simple LRU buffering for transaction status logfiles
+ * Simple buffering for transaction status logfiles
*
- * We use a simple least-recently-used scheme to manage a pool of page
- * buffers. Under ordinary circumstances we expect that write
- * traffic will occur mostly to the latest page (and to the just-prior
- * page, soon after a page transition). Read traffic will probably touch
- * a larger span of pages, but in any case a fairly small number of page
- * buffers should be sufficient. So, we just search the buffers using plain
- * linear search; there's no need for a hashtable or anything fancy.
- * The management algorithm is straight LRU except that we will never swap
- * out the latest page (since we know it's going to be hit again eventually).
- *
- * We use a control LWLock to protect the shared data structures, plus
- * per-buffer LWLocks that synchronize I/O for each buffer. The control lock
- * must be held to examine or modify any shared state. A process that is
- * reading in or writing out a page buffer does not hold the control lock,
- * only the per-buffer lock for the buffer it is working on.
- *
- * "Holding the control lock" means exclusive lock in all cases except for
- * SimpleLruReadPage_ReadOnly(); see comments for SlruRecentlyUsed() for
- * the implications of that.
- *
- * When initiating I/O on a buffer, we acquire the per-buffer lock exclusively
- * before releasing the control lock. The per-buffer lock is released after
- * completing the I/O, re-acquiring the control lock, and updating the shared
- * state. (Deadlock is not possible here, because we never try to initiate
- * I/O when someone else is already doing I/O on the same buffer.)
- * To wait for I/O to complete, release the control lock, acquire the
- * per-buffer lock in shared mode, immediately release the per-buffer lock,
- * reacquire the control lock, and then recheck state (since arbitrary things
- * could have happened while we didn't have the lock).
- *
- * As with the regular buffer manager, it is possible for another process
- * to re-dirty a page that is currently being written out. This is handled
- * by re-setting the page's page_dirty flag.
+ * XXX write me
*
*
* Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group
@@ -60,561 +28,81 @@
#include "storage/fd.h"
#include "storage/shmem.h"
-#define SlruFileName(ctl, path, seg) \
- snprintf(path, MAXPGPATH, "%s/%04X", (ctl)->Dir, seg)
+#define SlruFileName(rel_id, path, seg) \
+ snprintf(path, MAXPGPATH, "%s/%04X", defs[(rel_id)].path, seg)
+
+struct SlruDef {
+ const char *name;
+ const char *path;
+ bool synchronize;
+};
+
+static const struct SlruDef defs[] = {
+ [SLRU_CLOG_REL_ID] = {
+ .name = "Xact",
+ .path = "pg_xact",
+ .synchronize = true,
+ },
+ [SLRU_SUBTRANS_REL_ID] = {
+ .name = "Subtrans",
+ .path = "pg_subtrans",
+ },
+ [SLRU_MULTIXACT_OFFSET_REL_ID] = {
+ .name = "MultiXactOffset",
+ .path = "pg_multixact/offsets",
+ .synchronize = true,
+ },
+ [SLRU_MULTIXACT_MEMBER_REL_ID] = {
+ .name = "MultiXactMember",
+ .path = "pg_multixact/members",
+ .synchronize = true,
+ },
+ [SLRU_COMMITTS_REL_ID] = {
+ .name = "CommitTs",
+ .path = "pg_commit_ts",
+ .synchronize = true,
+ },
+ [SLRU_SERIAL_REL_ID] = {
+ .name = "Serial",
+ .path = "pg_serial",
+ },
+ [SLRU_NOTIFY_REL_ID] = {
+ .name = "Notify",
+ .path = "pg_notify",
+ },
+};
/*
- * During SimpleLruWriteAll(), we will usually not need to write more than one
- * or two physical files, but we may need to write several pages per file. We
- * can consolidate the I/O requests by leaving files open until control returns
- * to SimpleLruWriteAll(). This data structure remembers which files are open.
+ * We'll maintain a little cache of recently seen buffers, to try to avoid the
+ * buffer mapping table on repeat access (ie the busy end of the CLOG). One
+ * entry per SLRU relation.
*/
-#define MAX_WRITEALL_BUFFERS 16
+struct SlruRecentBuffer {
+ int pageno;
+ Buffer recent_buffer;
+};
-typedef struct SlruWriteAllData
-{
- int num_files; /* # files actually open */
- int fd[MAX_WRITEALL_BUFFERS]; /* their FD's */
- int segno[MAX_WRITEALL_BUFFERS]; /* their log seg#s */
-} SlruWriteAllData;
-
-typedef struct SlruWriteAllData *SlruWriteAll;
+static struct SlruRecentBuffer slru_recent_buffers[lengthof(defs)];
/*
- * Populate a file tag describing a segment file. We only use the segment
- * number, since we can derive everything else we need by having separate
- * sync handler functions for clog, multixact etc.
+ * Populate a file tag identifying an SLRU segment file.
*/
-#define INIT_SLRUFILETAG(a,xx_handler,xx_segno) \
+#define INIT_SLRUFILETAG(a,xx_rel_number,xx_segno) \
( \
memset(&(a), 0, sizeof(FileTag)), \
- (a).handler = (xx_handler), \
+ (a).handler = SYNC_HANDLER_SLRU, \
+ (a).rlocator = SlruRelFileLocator(xx_rel_number), \
(a).segno = (xx_segno) \
)
-/*
- * Macro to mark a buffer slot "most recently used". Note multiple evaluation
- * of arguments!
- *
- * The reason for the if-test is that there are often many consecutive
- * accesses to the same page (particularly the latest page). By suppressing
- * useless increments of cur_lru_count, we reduce the probability that old
- * pages' counts will "wrap around" and make them appear recently used.
- *
- * We allow this code to be executed concurrently by multiple processes within
- * SimpleLruReadPage_ReadOnly(). As long as int reads and writes are atomic,
- * this should not cause any completely-bogus values to enter the computation.
- * However, it is possible for either cur_lru_count or individual
- * page_lru_count entries to be "reset" to lower values than they should have,
- * in case a process is delayed while it executes this macro. With care in
- * SlruSelectLRUPage(), this does little harm, and in any case the absolute
- * worst possible consequence is a nonoptimal choice of page to evict. The
- * gain from allowing concurrent reads of SLRU pages seems worth it.
- */
-#define SlruRecentlyUsed(shared, slotno) \
- do { \
- int new_lru_count = (shared)->cur_lru_count; \
- if (new_lru_count != (shared)->page_lru_count[slotno]) { \
- (shared)->cur_lru_count = ++new_lru_count; \
- (shared)->page_lru_count[slotno] = new_lru_count; \
- } \
- } while (0)
-
-/* Saved info for SlruReportIOError */
-typedef enum
-{
- SLRU_OPEN_FAILED,
- SLRU_SEEK_FAILED,
- SLRU_READ_FAILED,
- SLRU_WRITE_FAILED,
- SLRU_FSYNC_FAILED,
- SLRU_CLOSE_FAILED
-} SlruErrorCause;
-
-static SlruErrorCause slru_errcause;
-static int slru_errno;
-
-
-static void SimpleLruZeroLSNs(SlruCtl ctl, int slotno);
-static void SimpleLruWaitIO(SlruCtl ctl, int slotno);
-static void SlruInternalWritePage(SlruCtl ctl, int slotno, SlruWriteAll fdata);
-static bool SlruPhysicalReadPage(SlruCtl ctl, int pageno, int slotno);
-static bool SlruPhysicalWritePage(SlruCtl ctl, int pageno, int slotno,
- SlruWriteAll fdata);
-static void SlruReportIOError(SlruCtl ctl, int pageno, TransactionId xid);
-static int SlruSelectLRUPage(SlruCtl ctl, int pageno);
-
-static bool SlruScanDirCbDeleteCutoff(SlruCtl ctl, char *filename,
+static bool SlruScanDirCbDeleteCutoff(Oid rel_id,
+ SlruPagePrecedesFunction PagePrecedes,
+ char *filename,
int segpage, void *data);
-static void SlruInternalDeleteSegment(SlruCtl ctl, int segno);
-
-/*
- * Initialization of shared memory
- */
-
-Size
-SimpleLruShmemSize(int nslots, int nlsns)
-{
- Size sz;
-
- /* we assume nslots isn't so large as to risk overflow */
- sz = MAXALIGN(sizeof(SlruSharedData));
- sz += MAXALIGN(nslots * sizeof(char *)); /* page_buffer[] */
- sz += MAXALIGN(nslots * sizeof(SlruPageStatus)); /* page_status[] */
- sz += MAXALIGN(nslots * sizeof(bool)); /* page_dirty[] */
- sz += MAXALIGN(nslots * sizeof(int)); /* page_number[] */
- sz += MAXALIGN(nslots * sizeof(int)); /* page_lru_count[] */
- sz += MAXALIGN(nslots * sizeof(LWLockPadded)); /* buffer_locks[] */
-
- if (nlsns > 0)
- sz += MAXALIGN(nslots * nlsns * sizeof(XLogRecPtr)); /* group_lsn[] */
-
- return BUFFERALIGN(sz) + BLCKSZ * nslots;
-}
-
-/*
- * Initialize, or attach to, a simple LRU cache in shared memory.
- *
- * ctl: address of local (unshared) control structure.
- * name: name of SLRU. (This is user-visible, pick with care!)
- * nslots: number of page slots to use.
- * nlsns: number of LSN groups per page (set to zero if not relevant).
- * ctllock: LWLock to use to control access to the shared control structure.
- * subdir: PGDATA-relative subdirectory that will contain the files.
- * tranche_id: LWLock tranche ID to use for the SLRU's per-buffer LWLocks.
- */
-void
-SimpleLruInit(SlruCtl ctl, const char *name, int nslots, int nlsns,
- LWLock *ctllock, const char *subdir, int tranche_id,
- SyncRequestHandler sync_handler)
-{
- SlruShared shared;
- bool found;
-
- shared = (SlruShared) ShmemInitStruct(name,
- SimpleLruShmemSize(nslots, nlsns),
- &found);
-
- if (!IsUnderPostmaster)
- {
- /* Initialize locks and shared memory area */
- char *ptr;
- Size offset;
- int slotno;
-
- Assert(!found);
-
- memset(shared, 0, sizeof(SlruSharedData));
-
- shared->ControlLock = ctllock;
-
- shared->num_slots = nslots;
- shared->lsn_groups_per_page = nlsns;
-
- shared->cur_lru_count = 0;
-
- /* shared->latest_page_number will be set later */
-
- shared->slru_stats_idx = pgstat_get_slru_index(name);
-
- ptr = (char *) shared;
- offset = MAXALIGN(sizeof(SlruSharedData));
- shared->page_buffer = (char **) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(char *));
- shared->page_status = (SlruPageStatus *) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(SlruPageStatus));
- shared->page_dirty = (bool *) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(bool));
- shared->page_number = (int *) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(int));
- shared->page_lru_count = (int *) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(int));
-
- /* Initialize LWLocks */
- shared->buffer_locks = (LWLockPadded *) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(LWLockPadded));
-
- if (nlsns > 0)
- {
- shared->group_lsn = (XLogRecPtr *) (ptr + offset);
- offset += MAXALIGN(nslots * nlsns * sizeof(XLogRecPtr));
- }
-
- ptr += BUFFERALIGN(offset);
- for (slotno = 0; slotno < nslots; slotno++)
- {
- LWLockInitialize(&shared->buffer_locks[slotno].lock,
- tranche_id);
-
- shared->page_buffer[slotno] = ptr;
- shared->page_status[slotno] = SLRU_PAGE_EMPTY;
- shared->page_dirty[slotno] = false;
- shared->page_lru_count[slotno] = 0;
- ptr += BLCKSZ;
- }
-
- /* Should fit to estimated shmem size */
- Assert(ptr - (char *) shared <= SimpleLruShmemSize(nslots, nlsns));
- }
- else
- Assert(found);
-
- /*
- * Initialize the unshared control struct, including directory path. We
- * assume caller set PagePrecedes.
- */
- ctl->shared = shared;
- ctl->sync_handler = sync_handler;
- strlcpy(ctl->Dir, subdir, sizeof(ctl->Dir));
-}
-
-/*
- * Initialize (or reinitialize) a page to zeroes.
- *
- * The page is not actually written, just set up in shared memory.
- * The slot number of the new page is returned.
- *
- * Control lock must be held at entry, and will be held at exit.
- */
-int
-SimpleLruZeroPage(SlruCtl ctl, int pageno)
-{
- SlruShared shared = ctl->shared;
- int slotno;
+static void SlruInternalDeleteSegment(Oid rel_id, int segno);
- /* Find a suitable buffer slot for the page */
- slotno = SlruSelectLRUPage(ctl, pageno);
- Assert(shared->page_status[slotno] == SLRU_PAGE_EMPTY ||
- (shared->page_status[slotno] == SLRU_PAGE_VALID &&
- !shared->page_dirty[slotno]) ||
- shared->page_number[slotno] == pageno);
-
- /* Mark the slot as containing this page */
- shared->page_number[slotno] = pageno;
- shared->page_status[slotno] = SLRU_PAGE_VALID;
- shared->page_dirty[slotno] = true;
- SlruRecentlyUsed(shared, slotno);
-
- /* Set the buffer to zeroes */
- MemSet(shared->page_buffer[slotno], 0, BLCKSZ);
-
- /* Set the LSNs for this new page to zero */
- SimpleLruZeroLSNs(ctl, slotno);
-
- /* Assume this page is now the latest active page */
- shared->latest_page_number = pageno;
-
- /* update the stats counter of zeroed pages */
- pgstat_count_slru_page_zeroed(shared->slru_stats_idx);
-
- return slotno;
-}
-
-/*
- * Zero all the LSNs we store for this slru page.
- *
- * This should be called each time we create a new page, and each time we read
- * in a page from disk into an existing buffer. (Such an old page cannot
- * have any interesting LSNs, since we'd have flushed them before writing
- * the page in the first place.)
- *
- * This assumes that InvalidXLogRecPtr is bitwise-all-0.
- */
-static void
-SimpleLruZeroLSNs(SlruCtl ctl, int slotno)
-{
- SlruShared shared = ctl->shared;
-
- if (shared->lsn_groups_per_page > 0)
- MemSet(&shared->group_lsn[slotno * shared->lsn_groups_per_page], 0,
- shared->lsn_groups_per_page * sizeof(XLogRecPtr));
-}
-
-/*
- * Wait for any active I/O on a page slot to finish. (This does not
- * guarantee that new I/O hasn't been started before we return, though.
- * In fact the slot might not even contain the same page anymore.)
- *
- * Control lock must be held at entry, and will be held at exit.
- */
-static void
-SimpleLruWaitIO(SlruCtl ctl, int slotno)
-{
- SlruShared shared = ctl->shared;
-
- /* See notes at top of file */
- LWLockRelease(shared->ControlLock);
- LWLockAcquire(&shared->buffer_locks[slotno].lock, LW_SHARED);
- LWLockRelease(&shared->buffer_locks[slotno].lock);
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
- /*
- * If the slot is still in an io-in-progress state, then either someone
- * already started a new I/O on the slot, or a previous I/O failed and
- * neglected to reset the page state. That shouldn't happen, really, but
- * it seems worth a few extra cycles to check and recover from it. We can
- * cheaply test for failure by seeing if the buffer lock is still held (we
- * assume that transaction abort would release the lock).
- */
- if (shared->page_status[slotno] == SLRU_PAGE_READ_IN_PROGRESS ||
- shared->page_status[slotno] == SLRU_PAGE_WRITE_IN_PROGRESS)
- {
- if (LWLockConditionalAcquire(&shared->buffer_locks[slotno].lock, LW_SHARED))
- {
- /* indeed, the I/O must have failed */
- if (shared->page_status[slotno] == SLRU_PAGE_READ_IN_PROGRESS)
- shared->page_status[slotno] = SLRU_PAGE_EMPTY;
- else /* write_in_progress */
- {
- shared->page_status[slotno] = SLRU_PAGE_VALID;
- shared->page_dirty[slotno] = true;
- }
- LWLockRelease(&shared->buffer_locks[slotno].lock);
- }
- }
-}
-
-/*
- * Find a page in a shared buffer, reading it in if necessary.
- * The page number must correspond to an already-initialized page.
- *
- * If write_ok is true then it is OK to return a page that is in
- * WRITE_IN_PROGRESS state; it is the caller's responsibility to be sure
- * that modification of the page is safe. If write_ok is false then we
- * will not return the page until it is not undergoing active I/O.
- *
- * The passed-in xid is used only for error reporting, and may be
- * InvalidTransactionId if no specific xid is associated with the action.
- *
- * Return value is the shared-buffer slot number now holding the page.
- * The buffer's LRU access info is updated.
- *
- * Control lock must be held at entry, and will be held at exit.
- */
-int
-SimpleLruReadPage(SlruCtl ctl, int pageno, bool write_ok,
- TransactionId xid)
-{
- SlruShared shared = ctl->shared;
-
- /* Outer loop handles restart if we must wait for someone else's I/O */
- for (;;)
- {
- int slotno;
- bool ok;
-
- /* See if page already is in memory; if not, pick victim slot */
- slotno = SlruSelectLRUPage(ctl, pageno);
-
- /* Did we find the page in memory? */
- if (shared->page_number[slotno] == pageno &&
- shared->page_status[slotno] != SLRU_PAGE_EMPTY)
- {
- /*
- * If page is still being read in, we must wait for I/O. Likewise
- * if the page is being written and the caller said that's not OK.
- */
- if (shared->page_status[slotno] == SLRU_PAGE_READ_IN_PROGRESS ||
- (shared->page_status[slotno] == SLRU_PAGE_WRITE_IN_PROGRESS &&
- !write_ok))
- {
- SimpleLruWaitIO(ctl, slotno);
- /* Now we must recheck state from the top */
- continue;
- }
- /* Otherwise, it's ready to use */
- SlruRecentlyUsed(shared, slotno);
-
- /* update the stats counter of pages found in the SLRU */
- pgstat_count_slru_page_hit(shared->slru_stats_idx);
-
- return slotno;
- }
-
- /* We found no match; assert we selected a freeable slot */
- Assert(shared->page_status[slotno] == SLRU_PAGE_EMPTY ||
- (shared->page_status[slotno] == SLRU_PAGE_VALID &&
- !shared->page_dirty[slotno]));
-
- /* Mark the slot read-busy */
- shared->page_number[slotno] = pageno;
- shared->page_status[slotno] = SLRU_PAGE_READ_IN_PROGRESS;
- shared->page_dirty[slotno] = false;
-
- /* Acquire per-buffer lock (cannot deadlock, see notes at top) */
- LWLockAcquire(&shared->buffer_locks[slotno].lock, LW_EXCLUSIVE);
-
- /* Release control lock while doing I/O */
- LWLockRelease(shared->ControlLock);
-
- /* Do the read */
- ok = SlruPhysicalReadPage(ctl, pageno, slotno);
-
- /* Set the LSNs for this newly read-in page to zero */
- SimpleLruZeroLSNs(ctl, slotno);
-
- /* Re-acquire control lock and update page state */
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
- Assert(shared->page_number[slotno] == pageno &&
- shared->page_status[slotno] == SLRU_PAGE_READ_IN_PROGRESS &&
- !shared->page_dirty[slotno]);
-
- shared->page_status[slotno] = ok ? SLRU_PAGE_VALID : SLRU_PAGE_EMPTY;
-
- LWLockRelease(&shared->buffer_locks[slotno].lock);
-
- /* Now it's okay to ereport if we failed */
- if (!ok)
- SlruReportIOError(ctl, pageno, xid);
-
- SlruRecentlyUsed(shared, slotno);
-
- /* update the stats counter of pages not found in SLRU */
- pgstat_count_slru_page_read(shared->slru_stats_idx);
-
- return slotno;
- }
-}
-
-/*
- * Find a page in a shared buffer, reading it in if necessary.
- * The page number must correspond to an already-initialized page.
- * The caller must intend only read-only access to the page.
- *
- * The passed-in xid is used only for error reporting, and may be
- * InvalidTransactionId if no specific xid is associated with the action.
- *
- * Return value is the shared-buffer slot number now holding the page.
- * The buffer's LRU access info is updated.
- *
- * Control lock must NOT be held at entry, but will be held at exit.
- * It is unspecified whether the lock will be shared or exclusive.
- */
-int
-SimpleLruReadPage_ReadOnly(SlruCtl ctl, int pageno, TransactionId xid)
-{
- SlruShared shared = ctl->shared;
- int slotno;
-
- /* Try to find the page while holding only shared lock */
- LWLockAcquire(shared->ControlLock, LW_SHARED);
-
- /* See if page is already in a buffer */
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- if (shared->page_number[slotno] == pageno &&
- shared->page_status[slotno] != SLRU_PAGE_EMPTY &&
- shared->page_status[slotno] != SLRU_PAGE_READ_IN_PROGRESS)
- {
- /* See comments for SlruRecentlyUsed macro */
- SlruRecentlyUsed(shared, slotno);
-
- /* update the stats counter of pages found in the SLRU */
- pgstat_count_slru_page_hit(shared->slru_stats_idx);
-
- return slotno;
- }
- }
-
- /* No luck, so switch to normal exclusive lock and do regular read */
- LWLockRelease(shared->ControlLock);
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
- return SimpleLruReadPage(ctl, pageno, true, xid);
-}
-
-/*
- * Write a page from a shared buffer, if necessary.
- * Does nothing if the specified slot is not dirty.
- *
- * NOTE: only one write attempt is made here. Hence, it is possible that
- * the page is still dirty at exit (if someone else re-dirtied it during
- * the write). However, we *do* attempt a fresh write even if the page
- * is already being written; this is for checkpoints.
- *
- * Control lock must be held at entry, and will be held at exit.
- */
-static void
-SlruInternalWritePage(SlruCtl ctl, int slotno, SlruWriteAll fdata)
-{
- SlruShared shared = ctl->shared;
- int pageno = shared->page_number[slotno];
- bool ok;
-
- /* If a write is in progress, wait for it to finish */
- while (shared->page_status[slotno] == SLRU_PAGE_WRITE_IN_PROGRESS &&
- shared->page_number[slotno] == pageno)
- {
- SimpleLruWaitIO(ctl, slotno);
- }
-
- /*
- * Do nothing if page is not dirty, or if buffer no longer contains the
- * same page we were called for.
- */
- if (!shared->page_dirty[slotno] ||
- shared->page_status[slotno] != SLRU_PAGE_VALID ||
- shared->page_number[slotno] != pageno)
- return;
-
- /*
- * Mark the slot write-busy, and clear the dirtybit. After this point, a
- * transaction status update on this page will mark it dirty again.
- */
- shared->page_status[slotno] = SLRU_PAGE_WRITE_IN_PROGRESS;
- shared->page_dirty[slotno] = false;
-
- /* Acquire per-buffer lock (cannot deadlock, see notes at top) */
- LWLockAcquire(&shared->buffer_locks[slotno].lock, LW_EXCLUSIVE);
-
- /* Release control lock while doing I/O */
- LWLockRelease(shared->ControlLock);
-
- /* Do the write */
- ok = SlruPhysicalWritePage(ctl, pageno, slotno, fdata);
-
- /* If we failed, and we're in a flush, better close the files */
- if (!ok && fdata)
- {
- int i;
-
- for (i = 0; i < fdata->num_files; i++)
- CloseTransientFile(fdata->fd[i]);
- }
-
- /* Re-acquire control lock and update page state */
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
- Assert(shared->page_number[slotno] == pageno &&
- shared->page_status[slotno] == SLRU_PAGE_WRITE_IN_PROGRESS);
-
- /* If we failed to write, mark the page dirty again */
- if (!ok)
- shared->page_dirty[slotno] = true;
-
- shared->page_status[slotno] = SLRU_PAGE_VALID;
-
- LWLockRelease(&shared->buffer_locks[slotno].lock);
-
- /* Now it's okay to ereport if we failed */
- if (!ok)
- SlruReportIOError(ctl, pageno, InvalidTransactionId);
-
- /* If part of a checkpoint, count this as a buffer written. */
- if (fdata)
- CheckpointStats.ckpt_bufs_written++;
-}
-
-/*
- * Wrapper of SlruInternalWritePage, for external callers.
- * fdata is always passed a NULL here.
- */
-void
-SimpleLruWritePage(SlruCtl ctl, int slotno)
-{
- SlruInternalWritePage(ctl, slotno, NULL);
-}
+static File slrufile(SMgrRelation reln, BlockNumber blocknum, int mode,
+ bool missing_ok);
/*
* Return whether the given page exists on disk.
@@ -623,592 +111,30 @@ SimpleLruWritePage(SlruCtl ctl, int slotno)
* large enough to contain the given page.
*/
bool
-SimpleLruDoesPhysicalPageExist(SlruCtl ctl, int pageno)
-{
- int segno = pageno / SLRU_PAGES_PER_SEGMENT;
- int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
- int offset = rpageno * BLCKSZ;
- char path[MAXPGPATH];
- int fd;
- bool result;
- off_t endpos;
-
- /* update the stats counter of checked pages */
- pgstat_count_slru_page_exists(ctl->shared->slru_stats_idx);
-
- SlruFileName(ctl, path, segno);
-
- fd = OpenTransientFile(path, O_RDONLY | PG_BINARY);
- if (fd < 0)
- {
- /* expected: file doesn't exist */
- if (errno == ENOENT)
- return false;
-
- /* report error normally */
- slru_errcause = SLRU_OPEN_FAILED;
- slru_errno = errno;
- SlruReportIOError(ctl, pageno, 0);
- }
-
- if ((endpos = lseek(fd, 0, SEEK_END)) < 0)
- {
- slru_errcause = SLRU_SEEK_FAILED;
- slru_errno = errno;
- SlruReportIOError(ctl, pageno, 0);
- }
-
- result = endpos >= (off_t) (offset + BLCKSZ);
-
- if (CloseTransientFile(fd) != 0)
- {
- slru_errcause = SLRU_CLOSE_FAILED;
- slru_errno = errno;
- return false;
- }
-
- return result;
-}
-
-/*
- * Physical read of a (previously existing) page into a buffer slot
- *
- * On failure, we cannot just ereport(ERROR) since caller has put state in
- * shared memory that must be undone. So, we return false and save enough
- * info in static variables to let SlruReportIOError make the report.
- *
- * For now, assume it's not worth keeping a file pointer open across
- * read/write operations. We could cache one virtual file pointer ...
- */
-static bool
-SlruPhysicalReadPage(SlruCtl ctl, int pageno, int slotno)
+SimpleLruDoesPhysicalPageExist(RelFileNumber rel_id, int pageno)
{
- SlruShared shared = ctl->shared;
- int segno = pageno / SLRU_PAGES_PER_SEGMENT;
int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
off_t offset = rpageno * BLCKSZ;
- char path[MAXPGPATH];
- int fd;
-
- SlruFileName(ctl, path, segno);
-
- /*
- * In a crash-and-restart situation, it's possible for us to receive
- * commands to set the commit status of transactions whose bits are in
- * already-truncated segments of the commit log (see notes in
- * SlruPhysicalWritePage). Hence, if we are InRecovery, allow the case
- * where the file doesn't exist, and return zeroes instead.
- */
- fd = OpenTransientFile(path, O_RDONLY | PG_BINARY);
- if (fd < 0)
- {
- if (errno != ENOENT || !InRecovery)
- {
- slru_errcause = SLRU_OPEN_FAILED;
- slru_errno = errno;
- return false;
- }
-
- ereport(LOG,
- (errmsg("file \"%s\" doesn't exist, reading as zeroes",
- path)));
- MemSet(shared->page_buffer[slotno], 0, BLCKSZ);
- return true;
- }
-
- errno = 0;
- pgstat_report_wait_start(WAIT_EVENT_SLRU_READ);
- if (pg_pread(fd, shared->page_buffer[slotno], BLCKSZ, offset) != BLCKSZ)
- {
- pgstat_report_wait_end();
- slru_errcause = SLRU_READ_FAILED;
- slru_errno = errno;
- CloseTransientFile(fd);
- return false;
- }
- pgstat_report_wait_end();
-
- if (CloseTransientFile(fd) != 0)
- {
- slru_errcause = SLRU_CLOSE_FAILED;
- slru_errno = errno;
- return false;
- }
+ off_t size;
+ File file;
+ RelFileLocator rlocator = SlruRelFileLocator(rel_id);
+ SMgrRelation reln = smgropen(rlocator, InvalidBackendId);
- return true;
-}
-
-/*
- * Physical write of a page from a buffer slot
- *
- * On failure, we cannot just ereport(ERROR) since caller has put state in
- * shared memory that must be undone. So, we return false and save enough
- * info in static variables to let SlruReportIOError make the report.
- *
- * For now, assume it's not worth keeping a file pointer open across
- * independent read/write operations. We do batch operations during
- * SimpleLruWriteAll, though.
- *
- * fdata is NULL for a standalone write, pointer to open-file info during
- * SimpleLruWriteAll.
- */
-static bool
-SlruPhysicalWritePage(SlruCtl ctl, int pageno, int slotno, SlruWriteAll fdata)
-{
- SlruShared shared = ctl->shared;
- int segno = pageno / SLRU_PAGES_PER_SEGMENT;
- int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
- off_t offset = rpageno * BLCKSZ;
- char path[MAXPGPATH];
- int fd = -1;
-
- /* update the stats counter of written pages */
- pgstat_count_slru_page_written(shared->slru_stats_idx);
-
- /*
- * Honor the write-WAL-before-data rule, if appropriate, so that we do not
- * write out data before associated WAL records. This is the same action
- * performed during FlushBuffer() in the main buffer manager.
- */
- if (shared->group_lsn != NULL)
- {
- /*
- * We must determine the largest async-commit LSN for the page. This
- * is a bit tedious, but since this entire function is a slow path
- * anyway, it seems better to do this here than to maintain a per-page
- * LSN variable (which'd need an extra comparison in the
- * transaction-commit path).
- */
- XLogRecPtr max_lsn;
- int lsnindex,
- lsnoff;
-
- lsnindex = slotno * shared->lsn_groups_per_page;
- max_lsn = shared->group_lsn[lsnindex++];
- for (lsnoff = 1; lsnoff < shared->lsn_groups_per_page; lsnoff++)
- {
- XLogRecPtr this_lsn = shared->group_lsn[lsnindex++];
-
- if (max_lsn < this_lsn)
- max_lsn = this_lsn;
- }
-
- if (!XLogRecPtrIsInvalid(max_lsn))
- {
- /*
- * As noted above, elog(ERROR) is not acceptable here, so if
- * XLogFlush were to fail, we must PANIC. This isn't much of a
- * restriction because XLogFlush is just about all critical
- * section anyway, but let's make sure.
- */
- START_CRIT_SECTION();
- XLogFlush(max_lsn);
- END_CRIT_SECTION();
- }
- }
-
- /*
- * During a WriteAll, we may already have the desired file open.
- */
- if (fdata)
- {
- int i;
-
- for (i = 0; i < fdata->num_files; i++)
- {
- if (fdata->segno[i] == segno)
- {
- fd = fdata->fd[i];
- break;
- }
- }
- }
-
- if (fd < 0)
- {
- /*
- * If the file doesn't already exist, we should create it. It is
- * possible for this to need to happen when writing a page that's not
- * first in its segment; we assume the OS can cope with that. (Note:
- * it might seem that it'd be okay to create files only when
- * SimpleLruZeroPage is called for the first page of a segment.
- * However, if after a crash and restart the REDO logic elects to
- * replay the log from a checkpoint before the latest one, then it's
- * possible that we will get commands to set transaction status of
- * transactions that have already been truncated from the commit log.
- * Easiest way to deal with that is to accept references to
- * nonexistent files here and in SlruPhysicalReadPage.)
- *
- * Note: it is possible for more than one backend to be executing this
- * code simultaneously for different pages of the same file. Hence,
- * don't use O_EXCL or O_TRUNC or anything like that.
- */
- SlruFileName(ctl, path, segno);
- fd = OpenTransientFile(path, O_RDWR | O_CREAT | PG_BINARY);
- if (fd < 0)
- {
- slru_errcause = SLRU_OPEN_FAILED;
- slru_errno = errno;
- return false;
- }
-
- if (fdata)
- {
- if (fdata->num_files < MAX_WRITEALL_BUFFERS)
- {
- fdata->fd[fdata->num_files] = fd;
- fdata->segno[fdata->num_files] = segno;
- fdata->num_files++;
- }
- else
- {
- /*
- * In the unlikely event that we exceed MAX_FLUSH_BUFFERS,
- * fall back to treating it as a standalone write.
- */
- fdata = NULL;
- }
- }
- }
+ /* update the stats counter of checked pages */
+ pgstat_count_slru_page_exists(rel_id);
- errno = 0;
- pgstat_report_wait_start(WAIT_EVENT_SLRU_WRITE);
- if (pg_pwrite(fd, shared->page_buffer[slotno], BLCKSZ, offset) != BLCKSZ)
+ file = slrufile(reln, pageno, O_RDWR, true);
+ if (file < 0)
{
- pgstat_report_wait_end();
- /* if write didn't set errno, assume problem is no disk space */
- if (errno == 0)
- errno = ENOSPC;
- slru_errcause = SLRU_WRITE_FAILED;
- slru_errno = errno;
- if (!fdata)
- CloseTransientFile(fd);
+ Assert(errno == ENOENT);
return false;
}
- pgstat_report_wait_end();
-
- /* Queue up a sync request for the checkpointer. */
- if (ctl->sync_handler != SYNC_HANDLER_NONE)
- {
- FileTag tag;
-
- INIT_SLRUFILETAG(tag, ctl->sync_handler, segno);
- if (!RegisterSyncRequest(&tag, SYNC_REQUEST, false))
- {
- /* No space to enqueue sync request. Do it synchronously. */
- pgstat_report_wait_start(WAIT_EVENT_SLRU_SYNC);
- if (pg_fsync(fd) != 0)
- {
- pgstat_report_wait_end();
- slru_errcause = SLRU_FSYNC_FAILED;
- slru_errno = errno;
- CloseTransientFile(fd);
- return false;
- }
- pgstat_report_wait_end();
- }
- }
+ size = FileSize(file);
+ if (size < 0)
+ elog(ERROR, "could not get size of file \"%s\": %m",
+ FilePathName(file));
- /* Close file, unless part of flush request. */
- if (!fdata)
- {
- if (CloseTransientFile(fd) != 0)
- {
- slru_errcause = SLRU_CLOSE_FAILED;
- slru_errno = errno;
- return false;
- }
- }
-
- return true;
-}
-
-/*
- * Issue the error message after failure of SlruPhysicalReadPage or
- * SlruPhysicalWritePage. Call this after cleaning up shared-memory state.
- */
-static void
-SlruReportIOError(SlruCtl ctl, int pageno, TransactionId xid)
-{
- int segno = pageno / SLRU_PAGES_PER_SEGMENT;
- int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
- int offset = rpageno * BLCKSZ;
- char path[MAXPGPATH];
-
- SlruFileName(ctl, path, segno);
- errno = slru_errno;
- switch (slru_errcause)
- {
- case SLRU_OPEN_FAILED:
- ereport(ERROR,
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not open file \"%s\": %m.", path)));
- break;
- case SLRU_SEEK_FAILED:
- ereport(ERROR,
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not seek in file \"%s\" to offset %d: %m.",
- path, offset)));
- break;
- case SLRU_READ_FAILED:
- if (errno)
- ereport(ERROR,
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not read from file \"%s\" at offset %d: %m.",
- path, offset)));
- else
- ereport(ERROR,
- (errmsg("could not access status of transaction %u", xid),
- errdetail("Could not read from file \"%s\" at offset %d: read too few bytes.", path, offset)));
- break;
- case SLRU_WRITE_FAILED:
- if (errno)
- ereport(ERROR,
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not write to file \"%s\" at offset %d: %m.",
- path, offset)));
- else
- ereport(ERROR,
- (errmsg("could not access status of transaction %u", xid),
- errdetail("Could not write to file \"%s\" at offset %d: wrote too few bytes.",
- path, offset)));
- break;
- case SLRU_FSYNC_FAILED:
- ereport(data_sync_elevel(ERROR),
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not fsync file \"%s\": %m.",
- path)));
- break;
- case SLRU_CLOSE_FAILED:
- ereport(ERROR,
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not close file \"%s\": %m.",
- path)));
- break;
- default:
- /* can't get here, we trust */
- elog(ERROR, "unrecognized SimpleLru error cause: %d",
- (int) slru_errcause);
- break;
- }
-}
-
-/*
- * Select the slot to re-use when we need a free slot.
- *
- * The target page number is passed because we need to consider the
- * possibility that some other process reads in the target page while
- * we are doing I/O to free a slot. Hence, check or recheck to see if
- * any slot already holds the target page, and return that slot if so.
- * Thus, the returned slot is *either* a slot already holding the pageno
- * (could be any state except EMPTY), *or* a freeable slot (state EMPTY
- * or CLEAN).
- *
- * Control lock must be held at entry, and will be held at exit.
- */
-static int
-SlruSelectLRUPage(SlruCtl ctl, int pageno)
-{
- SlruShared shared = ctl->shared;
-
- /* Outer loop handles restart after I/O */
- for (;;)
- {
- int slotno;
- int cur_count;
- int bestvalidslot = 0; /* keep compiler quiet */
- int best_valid_delta = -1;
- int best_valid_page_number = 0; /* keep compiler quiet */
- int bestinvalidslot = 0; /* keep compiler quiet */
- int best_invalid_delta = -1;
- int best_invalid_page_number = 0; /* keep compiler quiet */
-
- /* See if page already has a buffer assigned */
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- if (shared->page_number[slotno] == pageno &&
- shared->page_status[slotno] != SLRU_PAGE_EMPTY)
- return slotno;
- }
-
- /*
- * If we find any EMPTY slot, just select that one. Else choose a
- * victim page to replace. We normally take the least recently used
- * valid page, but we will never take the slot containing
- * latest_page_number, even if it appears least recently used. We
- * will select a slot that is already I/O busy only if there is no
- * other choice: a read-busy slot will not be least recently used once
- * the read finishes, and waiting for an I/O on a write-busy slot is
- * inferior to just picking some other slot. Testing shows the slot
- * we pick instead will often be clean, allowing us to begin a read at
- * once.
- *
- * Normally the page_lru_count values will all be different and so
- * there will be a well-defined LRU page. But since we allow
- * concurrent execution of SlruRecentlyUsed() within
- * SimpleLruReadPage_ReadOnly(), it is possible that multiple pages
- * acquire the same lru_count values. In that case we break ties by
- * choosing the furthest-back page.
- *
- * Notice that this next line forcibly advances cur_lru_count to a
- * value that is certainly beyond any value that will be in the
- * page_lru_count array after the loop finishes. This ensures that
- * the next execution of SlruRecentlyUsed will mark the page newly
- * used, even if it's for a page that has the current counter value.
- * That gets us back on the path to having good data when there are
- * multiple pages with the same lru_count.
- */
- cur_count = (shared->cur_lru_count)++;
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- int this_delta;
- int this_page_number;
-
- if (shared->page_status[slotno] == SLRU_PAGE_EMPTY)
- return slotno;
- this_delta = cur_count - shared->page_lru_count[slotno];
- if (this_delta < 0)
- {
- /*
- * Clean up in case shared updates have caused cur_count
- * increments to get "lost". We back off the page counts,
- * rather than trying to increase cur_count, to avoid any
- * question of infinite loops or failure in the presence of
- * wrapped-around counts.
- */
- shared->page_lru_count[slotno] = cur_count;
- this_delta = 0;
- }
- this_page_number = shared->page_number[slotno];
- if (this_page_number == shared->latest_page_number)
- continue;
- if (shared->page_status[slotno] == SLRU_PAGE_VALID)
- {
- if (this_delta > best_valid_delta ||
- (this_delta == best_valid_delta &&
- ctl->PagePrecedes(this_page_number,
- best_valid_page_number)))
- {
- bestvalidslot = slotno;
- best_valid_delta = this_delta;
- best_valid_page_number = this_page_number;
- }
- }
- else
- {
- if (this_delta > best_invalid_delta ||
- (this_delta == best_invalid_delta &&
- ctl->PagePrecedes(this_page_number,
- best_invalid_page_number)))
- {
- bestinvalidslot = slotno;
- best_invalid_delta = this_delta;
- best_invalid_page_number = this_page_number;
- }
- }
- }
-
- /*
- * If all pages (except possibly the latest one) are I/O busy, we'll
- * have to wait for an I/O to complete and then retry. In that
- * unhappy case, we choose to wait for the I/O on the least recently
- * used slot, on the assumption that it was likely initiated first of
- * all the I/Os in progress and may therefore finish first.
- */
- if (best_valid_delta < 0)
- {
- SimpleLruWaitIO(ctl, bestinvalidslot);
- continue;
- }
-
- /*
- * If the selected page is clean, we're set.
- */
- if (!shared->page_dirty[bestvalidslot])
- return bestvalidslot;
-
- /*
- * Write the page.
- */
- SlruInternalWritePage(ctl, bestvalidslot, NULL);
-
- /*
- * Now loop back and try again. This is the easiest way of dealing
- * with corner cases such as the victim page being re-dirtied while we
- * wrote it.
- */
- }
-}
-
-/*
- * Write dirty pages to disk during checkpoint or database shutdown. Flushing
- * is deferred until the next call to ProcessSyncRequests(), though we do fsync
- * the containing directory here to make sure that newly created directory
- * entries are on disk.
- */
-void
-SimpleLruWriteAll(SlruCtl ctl, bool allow_redirtied)
-{
- SlruShared shared = ctl->shared;
- SlruWriteAllData fdata;
- int slotno;
- int pageno = 0;
- int i;
- bool ok;
-
- /* update the stats counter of flushes */
- pgstat_count_slru_flush(shared->slru_stats_idx);
-
- /*
- * Find and write dirty pages
- */
- fdata.num_files = 0;
-
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- SlruInternalWritePage(ctl, slotno, &fdata);
-
- /*
- * In some places (e.g. checkpoints), we cannot assert that the slot
- * is clean now, since another process might have re-dirtied it
- * already. That's okay.
- */
- Assert(allow_redirtied ||
- shared->page_status[slotno] == SLRU_PAGE_EMPTY ||
- (shared->page_status[slotno] == SLRU_PAGE_VALID &&
- !shared->page_dirty[slotno]));
- }
-
- LWLockRelease(shared->ControlLock);
-
- /*
- * Now close any files that were open
- */
- ok = true;
- for (i = 0; i < fdata.num_files; i++)
- {
- if (CloseTransientFile(fdata.fd[i]) != 0)
- {
- slru_errcause = SLRU_CLOSE_FAILED;
- slru_errno = errno;
- pageno = fdata.segno[i] * SLRU_PAGES_PER_SEGMENT;
- ok = false;
- }
- }
- if (!ok)
- SlruReportIOError(ctl, pageno, InvalidTransactionId);
-
- /* Ensure that directory entries for new files are on disk. */
- if (ctl->sync_handler != SYNC_HANDLER_NONE)
- fsync_fname(ctl->Dir, true);
+ return size >= offset + BLCKSZ;
}
/*
@@ -1223,75 +149,14 @@ SimpleLruWriteAll(SlruCtl ctl, bool allow_redirtied)
* after it has accrued freshly-written data.
*/
void
-SimpleLruTruncate(SlruCtl ctl, int cutoffPage)
+SimpleLruTruncate(Oid rel_id, SlruPagePrecedesFunction PagePrecedes, int cutoffPage)
{
- SlruShared shared = ctl->shared;
- int slotno;
-
/* update the stats counter of truncates */
- pgstat_count_slru_truncate(shared->slru_stats_idx);
-
- /*
- * Scan shared memory and remove any pages preceding the cutoff page, to
- * ensure we won't rewrite them later. (Since this is normally called in
- * or just after a checkpoint, any dirty pages should have been flushed
- * already ... we're just being extra careful here.)
- */
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
-restart:;
-
- /*
- * While we are holding the lock, make an important safety check: the
- * current endpoint page must not be eligible for removal.
- */
- if (ctl->PagePrecedes(shared->latest_page_number, cutoffPage))
- {
- LWLockRelease(shared->ControlLock);
- ereport(LOG,
- (errmsg("could not truncate directory \"%s\": apparent wraparound",
- ctl->Dir)));
- return;
- }
-
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- if (shared->page_status[slotno] == SLRU_PAGE_EMPTY)
- continue;
- if (!ctl->PagePrecedes(shared->page_number[slotno], cutoffPage))
- continue;
-
- /*
- * If page is clean, just change state to EMPTY (expected case).
- */
- if (shared->page_status[slotno] == SLRU_PAGE_VALID &&
- !shared->page_dirty[slotno])
- {
- shared->page_status[slotno] = SLRU_PAGE_EMPTY;
- continue;
- }
-
- /*
- * Hmm, we have (or may have) I/O operations acting on the page, so
- * we've got to wait for them to finish and then start again. This is
- * the same logic as in SlruSelectLRUPage. (XXX if page is dirty,
- * wouldn't it be OK to just discard it without writing it?
- * SlruMayDeleteSegment() uses a stricter qualification, so we might
- * not delete this page in the end; even if we don't delete it, we
- * won't have cause to read its data again. For now, keep the logic
- * the same as it was.)
- */
- if (shared->page_status[slotno] == SLRU_PAGE_VALID)
- SlruInternalWritePage(ctl, slotno, NULL);
- else
- SimpleLruWaitIO(ctl, slotno);
- goto restart;
- }
-
- LWLockRelease(shared->ControlLock);
+ pgstat_count_slru_truncate(rel_id);
/* Now we can remove the old segment(s) */
- (void) SlruScanDirectory(ctl, SlruScanDirCbDeleteCutoff, &cutoffPage);
+ (void) SlruScanDirectory(rel_id, PagePrecedes, SlruScanDirCbDeleteCutoff,
+ &cutoffPage);
}
/*
@@ -1301,21 +166,21 @@ restart:;
* they either can't yet contain anything, or have already been cleaned out.
*/
static void
-SlruInternalDeleteSegment(SlruCtl ctl, int segno)
+SlruInternalDeleteSegment(Oid rel_id, int segno)
{
char path[MAXPGPATH];
/* Forget any fsync requests queued for this segment. */
- if (ctl->sync_handler != SYNC_HANDLER_NONE)
+ if (defs[rel_id].synchronize)
{
FileTag tag;
- INIT_SLRUFILETAG(tag, ctl->sync_handler, segno);
+ INIT_SLRUFILETAG(tag, rel_id, segno);
RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true);
}
/* Unlink the file. */
- SlruFileName(ctl, path, segno);
+ SlruFileName(rel_id, path, segno);
ereport(DEBUG2, (errmsg_internal("removing file \"%s\"", path)));
unlink(path);
}
@@ -1324,54 +189,15 @@ SlruInternalDeleteSegment(SlruCtl ctl, int segno)
* Delete an individual SLRU segment, identified by the segment number.
*/
void
-SlruDeleteSegment(SlruCtl ctl, int segno)
+SlruDeleteSegment(RelFileNumber rel_number, int segno)
{
- SlruShared shared = ctl->shared;
- int slotno;
- bool did_write;
+ RelFileLocator rlocator = SlruRelFileLocator(rel_number);
/* Clean out any possibly existing references to the segment. */
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-restart:
- did_write = false;
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- int pagesegno = shared->page_number[slotno] / SLRU_PAGES_PER_SEGMENT;
-
- if (shared->page_status[slotno] == SLRU_PAGE_EMPTY)
- continue;
-
- /* not the segment we're looking for */
- if (pagesegno != segno)
- continue;
-
- /* If page is clean, just change state to EMPTY (expected case). */
- if (shared->page_status[slotno] == SLRU_PAGE_VALID &&
- !shared->page_dirty[slotno])
- {
- shared->page_status[slotno] = SLRU_PAGE_EMPTY;
- continue;
- }
+ for (int i = 0; i < SLRU_PAGES_PER_SEGMENT; ++i)
+ DiscardBuffer(rlocator, MAIN_FORKNUM, segno * SLRU_PAGES_PER_SEGMENT + i);
- /* Same logic as SimpleLruTruncate() */
- if (shared->page_status[slotno] == SLRU_PAGE_VALID)
- SlruInternalWritePage(ctl, slotno, NULL);
- else
- SimpleLruWaitIO(ctl, slotno);
-
- did_write = true;
- }
-
- /*
- * Be extra careful and re-check. The IO functions release the control
- * lock, so new pages could have been read in.
- */
- if (did_write)
- goto restart;
-
- SlruInternalDeleteSegment(ctl, segno);
-
- LWLockRelease(shared->ControlLock);
+ SlruInternalDeleteSegment(rel_number, segno);
}
/*
@@ -1388,19 +214,21 @@ restart:
* first>=cutoff && last>=cutoff: no; every page of this segment is too young
*/
static bool
-SlruMayDeleteSegment(SlruCtl ctl, int segpage, int cutoffPage)
+SlruMayDeleteSegment(SlruPagePrecedesFunction PagePrecedes,
+ int segpage, int cutoffPage)
{
int seg_last_page = segpage + SLRU_PAGES_PER_SEGMENT - 1;
Assert(segpage % SLRU_PAGES_PER_SEGMENT == 0);
- return (ctl->PagePrecedes(segpage, cutoffPage) &&
- ctl->PagePrecedes(seg_last_page, cutoffPage));
+ return (PagePrecedes(segpage, cutoffPage) &&
+ PagePrecedes(seg_last_page, cutoffPage));
}
#ifdef USE_ASSERT_CHECKING
static void
-SlruPagePrecedesTestOffset(SlruCtl ctl, int per_page, uint32 offset)
+SlruPagePrecedesTestOffset(SlruPagePrecedesFunction PagePrecedes,
+ int per_page, uint32 offset)
{
TransactionId lhs,
rhs;
@@ -1425,19 +253,19 @@ SlruPagePrecedesTestOffset(SlruCtl ctl, int per_page, uint32 offset)
Assert(!TransactionIdPrecedes(rhs, lhs + 1));
Assert(!TransactionIdFollowsOrEquals(lhs, rhs));
Assert(!TransactionIdFollowsOrEquals(rhs, lhs));
- Assert(!ctl->PagePrecedes(lhs / per_page, lhs / per_page));
- Assert(!ctl->PagePrecedes(lhs / per_page, rhs / per_page));
- Assert(!ctl->PagePrecedes(rhs / per_page, lhs / per_page));
- Assert(!ctl->PagePrecedes((lhs - per_page) / per_page, rhs / per_page));
- Assert(ctl->PagePrecedes(rhs / per_page, (lhs - 3 * per_page) / per_page));
- Assert(ctl->PagePrecedes(rhs / per_page, (lhs - 2 * per_page) / per_page));
- Assert(ctl->PagePrecedes(rhs / per_page, (lhs - 1 * per_page) / per_page)
+ Assert(!PagePrecedes(lhs / per_page, lhs / per_page));
+ Assert(!PagePrecedes(lhs / per_page, rhs / per_page));
+ Assert(!PagePrecedes(rhs / per_page, lhs / per_page));
+ Assert(!PagePrecedes((lhs - per_page) / per_page, rhs / per_page));
+ Assert(PagePrecedes(rhs / per_page, (lhs - 3 * per_page) / per_page));
+ Assert(PagePrecedes(rhs / per_page, (lhs - 2 * per_page) / per_page));
+ Assert(PagePrecedes(rhs / per_page, (lhs - 1 * per_page) / per_page)
|| (1U << 31) % per_page != 0); /* See CommitTsPagePrecedes() */
- Assert(ctl->PagePrecedes((lhs + 1 * per_page) / per_page, rhs / per_page)
+ Assert(PagePrecedes((lhs + 1 * per_page) / per_page, rhs / per_page)
|| (1U << 31) % per_page != 0);
- Assert(ctl->PagePrecedes((lhs + 2 * per_page) / per_page, rhs / per_page));
- Assert(ctl->PagePrecedes((lhs + 3 * per_page) / per_page, rhs / per_page));
- Assert(!ctl->PagePrecedes(rhs / per_page, (lhs + per_page) / per_page));
+ Assert(PagePrecedes((lhs + 2 * per_page) / per_page, rhs / per_page));
+ Assert(PagePrecedes((lhs + 3 * per_page) / per_page, rhs / per_page));
+ Assert(!PagePrecedes(rhs / per_page, (lhs + per_page) / per_page));
/*
* GetNewTransactionId() has assigned the last XID it can safely use, and
@@ -1450,7 +278,7 @@ SlruPagePrecedesTestOffset(SlruCtl ctl, int per_page, uint32 offset)
oldestXact = newestXact + 1;
oldestXact -= 1U << 31;
oldestPage = oldestXact / per_page;
- Assert(!SlruMayDeleteSegment(ctl,
+ Assert(!SlruMayDeleteSegment(PagePrecedes,
(newestPage -
newestPage % SLRU_PAGES_PER_SEGMENT),
oldestPage));
@@ -1466,7 +294,7 @@ SlruPagePrecedesTestOffset(SlruCtl ctl, int per_page, uint32 offset)
oldestXact = newestXact + 1;
oldestXact -= 1U << 31;
oldestPage = oldestXact / per_page;
- Assert(!SlruMayDeleteSegment(ctl,
+ Assert(!SlruMayDeleteSegment(PagePrecedes,
(newestPage -
newestPage % SLRU_PAGES_PER_SEGMENT),
oldestPage));
@@ -1482,12 +310,12 @@ SlruPagePrecedesTestOffset(SlruCtl ctl, int per_page, uint32 offset)
* do not apply to them.)
*/
void
-SlruPagePrecedesUnitTests(SlruCtl ctl, int per_page)
+SlruPagePrecedesUnitTests(SlruPagePrecedesFunction PagePrecedes, int per_page)
{
/* Test first, middle and last entries of a page. */
- SlruPagePrecedesTestOffset(ctl, per_page, 0);
- SlruPagePrecedesTestOffset(ctl, per_page, per_page / 2);
- SlruPagePrecedesTestOffset(ctl, per_page, per_page - 1);
+ SlruPagePrecedesTestOffset(PagePrecedes, per_page, 0);
+ SlruPagePrecedesTestOffset(PagePrecedes, per_page, per_page / 2);
+ SlruPagePrecedesTestOffset(PagePrecedes, per_page, per_page - 1);
}
#endif
@@ -1497,11 +325,12 @@ SlruPagePrecedesUnitTests(SlruCtl ctl, int per_page)
* one containing the page passed as "data".
*/
bool
-SlruScanDirCbReportPresence(SlruCtl ctl, char *filename, int segpage, void *data)
+SlruScanDirCbReportPresence(Oid rel_id, SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage, void *data)
{
int cutoffPage = *(int *) data;
- if (SlruMayDeleteSegment(ctl, segpage, cutoffPage))
+ if (SlruMayDeleteSegment(PagePrecedes, segpage, cutoffPage))
return true; /* found one; don't iterate any more */
return false; /* keep going */
@@ -1512,12 +341,18 @@ SlruScanDirCbReportPresence(SlruCtl ctl, char *filename, int segpage, void *data
* This callback deletes segments prior to the one passed in as "data".
*/
static bool
-SlruScanDirCbDeleteCutoff(SlruCtl ctl, char *filename, int segpage, void *data)
+SlruScanDirCbDeleteCutoff(RelFileNumber rel_number, SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage, void *data)
{
+ RelFileLocator rlocator = SlruRelFileLocator(rel_number);
int cutoffPage = *(int *) data;
- if (SlruMayDeleteSegment(ctl, segpage, cutoffPage))
- SlruInternalDeleteSegment(ctl, segpage / SLRU_PAGES_PER_SEGMENT);
+ if (SlruMayDeleteSegment(PagePrecedes, segpage, cutoffPage))
+ {
+ for (int i = 0; i < SLRU_PAGES_PER_SEGMENT; ++i)
+ DiscardBuffer(rlocator, MAIN_FORKNUM, segpage + i);
+ SlruInternalDeleteSegment(rel_number, segpage / SLRU_PAGES_PER_SEGMENT);
+ }
return false; /* keep going */
}
@@ -1527,9 +362,10 @@ SlruScanDirCbDeleteCutoff(SlruCtl ctl, char *filename, int segpage, void *data)
* This callback deletes all segments.
*/
bool
-SlruScanDirCbDeleteAll(SlruCtl ctl, char *filename, int segpage, void *data)
+SlruScanDirCbDeleteAll(Oid rel_id, SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage, void *data)
{
- SlruInternalDeleteSegment(ctl, segpage / SLRU_PAGES_PER_SEGMENT);
+ SlruInternalDeleteSegment(rel_id, segpage / SLRU_PAGES_PER_SEGMENT);
return false; /* keep going */
}
@@ -1550,16 +386,20 @@ SlruScanDirCbDeleteAll(SlruCtl ctl, char *filename, int segpage, void *data)
* Note that no locking is applied.
*/
bool
-SlruScanDirectory(SlruCtl ctl, SlruScanCallback callback, void *data)
+SlruScanDirectory(Oid rel_id, SlruPagePrecedesFunction PagePrecedes,
+ SlruScanCallback callback, void *data)
{
bool retval = false;
DIR *cldir;
struct dirent *clde;
int segno;
int segpage;
+ const char *path;
- cldir = AllocateDir(ctl->Dir);
- while ((clde = ReadDir(cldir, ctl->Dir)) != NULL)
+ path = defs[rel_id].path;
+
+ cldir = AllocateDir(path);
+ while ((clde = ReadDir(cldir, path)) != NULL)
{
size_t len;
@@ -1572,8 +412,8 @@ SlruScanDirectory(SlruCtl ctl, SlruScanCallback callback, void *data)
segpage = segno * SLRU_PAGES_PER_SEGMENT;
elog(DEBUG2, "SlruScanDirectory invoking callback on %s/%s",
- ctl->Dir, clde->d_name);
- retval = callback(ctl, clde->d_name, segpage, data);
+ path, clde->d_name);
+ retval = callback(rel_id, PagePrecedes, clde->d_name, segpage, data);
if (retval)
break;
}
@@ -1583,30 +423,294 @@ SlruScanDirectory(SlruCtl ctl, SlruScanCallback callback, void *data)
return retval;
}
+void
+CheckPointSLRU(void)
+{
+ /* Ensure that directory entries for new files are on disk. */
+ for (int i = 0; i < lengthof(defs); ++i)
+ {
+ if (defs[i].synchronize)
+ fsync_fname(defs[i].path, true);
+ }
+}
+
/*
- * Individual SLRUs (clog, ...) have to provide a sync.c handler function so
- * that they can provide the correct "SlruCtl" (otherwise we don't know how to
- * build the path), but they just forward to this common implementation that
- * performs the fsync.
+ * Read a buffer. Buffer is pinned on return.
*/
-int
-SlruSyncFileTag(SlruCtl ctl, const FileTag *ftag, char *path)
+Buffer
+ReadSlruBuffer(RelFileNumber rel_number, int pageno)
+{
+ RelFileLocator rlocator = SlruRelFileLocator(rel_number);
+ Buffer buffer;
+ bool hit;
+
+ /* Try to avoid doing a buffer mapping table lookup for repeated access. */
+ buffer = slru_recent_buffers[rel_number].recent_buffer;
+ if (slru_recent_buffers[rel_number].pageno == pageno &&
+ BufferIsValid(buffer) &&
+ ReadRecentBuffer(rlocator, MAIN_FORKNUM, pageno, buffer))
+ {
+ pgstat_count_slru_page_hit(rel_number);
+ return buffer;
+ }
+
+ /* Regular lookup. */
+ buffer = ReadBufferWithoutRelcacheWithHit(rlocator, MAIN_FORKNUM, pageno,
+ RBM_NORMAL, &hit);
+
+ /* Remember where this page is for next time. */
+ slru_recent_buffers[rel_number].pageno = pageno;
+ slru_recent_buffers[rel_number].recent_buffer = buffer;
+
+ if (hit)
+ pgstat_count_slru_page_hit(rel_number);
+
+ return buffer;
+}
+
+/*
+ * Zero-initialize a buffer. Buffer is pinned and exclusively locked on return.
+ */
+Buffer
+ZeroSlruBuffer(RelFileNumber rel_number, int pageno)
+{
+ RelFileLocator rlocator = SlruRelFileLocator(rel_number);
+ Buffer buffer;
+
+ buffer = ReadBufferWithoutRelcache(rlocator, MAIN_FORKNUM, pageno,
+ RBM_ZERO_AND_LOCK, NULL, true);
+
+ /* Remember where this page is for next time. */
+ slru_recent_buffers[rel_number].pageno = pageno;
+ slru_recent_buffers[rel_number].recent_buffer = buffer;
+
+ pgstat_count_slru_page_zeroed(rel_number);
+
+ return buffer;
+}
+
+Oid
+SlruRelIdByName(const char *name)
{
- int fd;
- int save_errno;
- int result;
+ for (int i = 0; i < lengthof(defs); ++i)
+ if (strcmp(defs[i].name, name) == 0)
+ return i;
+
+ elog(ERROR, "unknown SLRU \"%s\"", name);
+}
- SlruFileName(ctl, path, ftag->segno);
+const char *
+SlruName(Oid rel_id)
+{
+ if (rel_id >= SLRU_NUM_RELS)
+ elog(ERROR, "invalid SLRU rel ID %u", rel_id);
- fd = OpenTransientFile(path, O_RDWR | PG_BINARY);
- if (fd < 0)
+ return defs[rel_id].name;
+}
+
+int
+slrusyncfiletag(const FileTag *ftag, char *path)
+{
+ SMgrRelation reln;
+ File file;
+
+ reln = smgropen(ftag->rlocator, InvalidBackendId);
+ file = slrufile(reln, ftag->segno * SLRU_PAGES_PER_SEGMENT, O_RDWR, true);
+ if (file < 0)
+ {
+ /* Path is reported here so the caller can make an error message */
+ SlruFileName(ftag->rlocator.relNumber, path, ftag->segno);
return -1;
+ }
+
+ pgstat_count_slru_flush(reln->smgr_rlocator.locator.relNumber);
+
+ return FileSync(file, WAIT_EVENT_SLRU_SYNC);
+}
+
+static File
+slrufile(SMgrRelation reln, BlockNumber blocknum, int mode, bool missing_ok)
+{
+ int segment = blocknum / SLRU_PAGES_PER_SEGMENT;
+ char path[MAXPGPATH];
+
+ Assert(reln->smgr_rlocator.locator.dbOid == SLRU_DB_ID);
+ Assert(reln->smgr_rlocator.locator.relNumber < lengthof(defs));
+ Assert(defs[reln->smgr_rlocator.locator.relNumber].path != NULL);
+
+ /* Do we have the right file open already? */
+ if (reln->slru_file_segment == segment)
+ {
+ /* XXX How can we invalidate this if the SLRU wraps around?! */
+ Assert(reln->slru_file != -1);
+ return reln->slru_file;
+ }
+
+ /* Close the current file, if we have one open. */
+ if (reln->slru_file_segment != -1)
+ {
+ Assert(reln->slru_file != -1);
+ FileClose(reln->slru_file);
+ reln->slru_file = -1;
+ reln->slru_file_segment = -1;
+ }
+
+ /* Open the file we want. */
+ SlruFileName(reln->smgr_rlocator.locator.relNumber, path, segment);
+ reln->slru_file = PathNameOpenFile(path, mode);
+ if (reln->slru_file >= 0)
+ reln->slru_file_segment = segment;
+ else if (!(missing_ok && errno == ENOENT))
+ ereport(ERROR,
+ (errcode_for_file_access(),
+ errmsg("could not open file \"%s\": %m",
+ path)));
+
+ /*
+ * XXX That error message doesn't contain the xid; then again, the
+ * previous SLRU error codes were all weird about xids anyway; maybe
+ * clog.c, notify.c et al should set a context that has SLRU-specific
+ * context in a more natural format
+ */
+
+ return reln->slru_file;
+}
+
+void
+slruopen(SMgrRelation reln)
+{
+ reln->slru_file = -1;
+ reln->slru_file_segment = -1;
+
+ /*
+ * We don't want this to be closed at end of transaction, which would
+ * otherwise happen, because it isn't owned by a Relation.
+ */
+ dlist_delete(&reln->node);
+}
+
+void
+slruclose(SMgrRelation reln, ForkNumber forknum)
+{
+ if (reln->slru_file != -1)
+ FileClose(reln->slru_file);
+ reln->slru_file = -1;
+ reln->slru_file_segment = -1;
+}
+
+void
+slruwrite(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
+ char *buffer, bool skipFsync)
+{
+ File file;
+ RelFileNumber rel_number;
+ off_t offset;
+ int rc;
+
+ file = slrufile(reln, blocknum, O_RDWR | O_CREAT, false);
+ offset = (blocknum % SLRU_PAGES_PER_SEGMENT) * BLCKSZ;
+
+ rc = FileWrite(file, buffer, BLCKSZ, offset, WAIT_EVENT_SLRU_WRITE);
+ if (rc < 0)
+ ereport(ERROR,
+ (errcode_for_file_access(),
+ errmsg("could not write %d bytes to file \"%s\" at offset %d: %m",
+ BLCKSZ,
+ FilePathName(file),
+ (int) offset)));
+ if (rc < BLCKSZ)
+ ereport(ERROR,
+ (errcode_for_file_access(),
+ errmsg("could not write %d bytes to file \"%s\" at offset %d, only %d bytes written",
+ BLCKSZ,
+ FilePathName(file),
+ (int) offset,
+ rc)));
+
+ rel_number = reln->smgr_rlocator.locator.relNumber;
+ if (defs[rel_number].synchronize)
+ {
+ FileTag tag;
+
+ /* Tell checkpointer to synchronize this file. */
+ INIT_SLRUFILETAG(tag, rel_number, blocknum / SLRU_PAGES_PER_SEGMENT);
+ if (!RegisterSyncRequest(&tag, SYNC_REQUEST, false))
+ {
+ /* Queue full. Do it synchronously. */
+ if (FileSync(file, WAIT_EVENT_SLRU_SYNC) < 0)
+ ereport(data_sync_elevel(ERROR),
+ (errcode_for_file_access(),
+ errmsg("could not synchronize file \"%s\": %m",
+ FilePathName(file))));
+ }
+ }
+
+ pgstat_count_slru_page_written(reln->smgr_rlocator.locator.relNumber);
+}
+
+void
+slruread(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
+ char *buffer)
+{
+ File file;
+ off_t offset;
+ int rc;
+
+ file = slrufile(reln, blocknum, O_RDWR, false);
+ offset = (blocknum % SLRU_PAGES_PER_SEGMENT) * BLCKSZ;
+
+ rc = FileRead(file, buffer, BLCKSZ, offset, WAIT_EVENT_SLRU_READ);
+ if (rc < 0)
+ ereport(ERROR,
+ (errcode_for_file_access(),
+ errmsg("could not read %d bytes from file \"%s\" at offset %d: %m",
+ BLCKSZ,
+ FilePathName(file),
+ (int) offset)));
+ if (rc < BLCKSZ)
+ ereport(ERROR,
+ (errcode_for_file_access(),
+ errmsg("could not read %d bytes from file \"%s\" at offset %d, only %d bytes read",
+ BLCKSZ,
+ FilePathName(file),
+ (int) offset,
+ rc)));
+
+ pgstat_count_slru_page_read(reln->smgr_rlocator.locator.relNumber);
+}
- result = pg_fsync(fd);
- save_errno = errno;
+void
+slruwriteback(SMgrRelation reln, ForkNumber forknum,
+ BlockNumber blocknum, BlockNumber nblocks)
+{
+ RelFileNumber rel_number;
- CloseTransientFile(fd);
+ /* No point in flushing data we won't be fsyncing. */
+ rel_number = reln->smgr_rlocator.locator.relNumber;
+ if (!defs[rel_number].synchronize)
+ return;
- errno = save_errno;
- return result;
+ while (nblocks > 0)
+ {
+ File file;
+ BlockNumber blocknum_in_this_file;
+ BlockNumber nflush;
+
+ /* File range of blocks to flush in one file. */
+ blocknum_in_this_file = blocknum % SLRU_PAGES_PER_SEGMENT;
+ nflush = Min(nblocks, SLRU_PAGES_PER_SEGMENT - blocknum_in_this_file);
+
+ /* The file might have been unlinked already, so tolerate missing. */
+ file = slrufile(reln, blocknum, O_RDWR, true);
+ if (file < 0)
+ return;
+
+ FileWriteback(file,
+ BLCKSZ * blocknum_in_this_file,
+ BLCKSZ * nflush,
+ WAIT_EVENT_SLRU_FLUSH);
+
+ nblocks -= nflush;
+ blocknum += nflush;
+ }
}
diff --git a/src/backend/access/transam/subtrans.c b/src/backend/access/transam/subtrans.c
index 66d35481552..98c4caea8d8 100644
--- a/src/backend/access/transam/subtrans.c
+++ b/src/backend/access/transam/subtrans.c
@@ -32,6 +32,7 @@
#include "access/subtrans.h"
#include "access/transam.h"
#include "pg_trace.h"
+#include "storage/bufmgr.h"
#include "utils/snapmgr.h"
@@ -55,15 +56,7 @@
#define TransactionIdToEntry(xid) ((xid) % (TransactionId) SUBTRANS_XACTS_PER_PAGE)
-/*
- * Link to shared-memory data structures for SUBTRANS control
- */
-static SlruCtlData SubTransCtlData;
-
-#define SubTransCtl (&SubTransCtlData)
-
-
-static int ZeroSUBTRANSPage(int pageno);
+static Buffer ZeroSUBTRANSPage(int pageno);
static bool SubTransPagePrecedes(int page1, int page2);
@@ -75,16 +68,15 @@ SubTransSetParent(TransactionId xid, TransactionId parent)
{
int pageno = TransactionIdToPage(xid);
int entryno = TransactionIdToEntry(xid);
- int slotno;
TransactionId *ptr;
+ Buffer buffer;
Assert(TransactionIdIsValid(parent));
Assert(TransactionIdFollows(xid, parent));
- LWLockAcquire(SubtransSLRULock, LW_EXCLUSIVE);
-
- slotno = SimpleLruReadPage(SubTransCtl, pageno, true, xid);
- ptr = (TransactionId *) SubTransCtl->shared->page_buffer[slotno];
+ buffer = ReadSlruBuffer(SLRU_SUBTRANS_REL_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ ptr = (TransactionId *) BufferGetPage(buffer);
ptr += entryno;
/*
@@ -96,10 +88,10 @@ SubTransSetParent(TransactionId xid, TransactionId parent)
{
Assert(*ptr == InvalidTransactionId);
*ptr = parent;
- SubTransCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
}
- LWLockRelease(SubtransSLRULock);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -110,9 +102,9 @@ SubTransGetParent(TransactionId xid)
{
int pageno = TransactionIdToPage(xid);
int entryno = TransactionIdToEntry(xid);
- int slotno;
TransactionId *ptr;
TransactionId parent;
+ Buffer buffer;
/* Can't ask about stuff that might not be around anymore */
Assert(TransactionIdFollowsOrEquals(xid, TransactionXmin));
@@ -121,15 +113,14 @@ SubTransGetParent(TransactionId xid)
if (!TransactionIdIsNormal(xid))
return InvalidTransactionId;
- /* lock is acquired by SimpleLruReadPage_ReadOnly */
+ buffer = ReadSlruBuffer(SLRU_SUBTRANS_REL_ID, pageno);
- slotno = SimpleLruReadPage_ReadOnly(SubTransCtl, pageno, xid);
- ptr = (TransactionId *) SubTransCtl->shared->page_buffer[slotno];
+ ptr = (TransactionId *) BufferGetPage(buffer);
ptr += entryno;
parent = *ptr;
- LWLockRelease(SubtransSLRULock);
+ ReleaseBuffer(buffer);
return parent;
}
@@ -177,26 +168,6 @@ SubTransGetTopmostTransaction(TransactionId xid)
return previousXid;
}
-
-/*
- * Initialization of shared memory for SUBTRANS
- */
-Size
-SUBTRANSShmemSize(void)
-{
- return SimpleLruShmemSize(NUM_SUBTRANS_BUFFERS, 0);
-}
-
-void
-SUBTRANSShmemInit(void)
-{
- SubTransCtl->PagePrecedes = SubTransPagePrecedes;
- SimpleLruInit(SubTransCtl, "Subtrans", NUM_SUBTRANS_BUFFERS, 0,
- SubtransSLRULock, "pg_subtrans",
- LWTRANCHE_SUBTRANS_BUFFER, SYNC_HANDLER_NONE);
- SlruPagePrecedesUnitTests(SubTransCtl, SUBTRANS_XACTS_PER_PAGE);
-}
-
/*
* This func must be called ONCE on system install. It creates
* the initial SUBTRANS segment. (The SUBTRANS directory is assumed to
@@ -210,18 +181,16 @@ SUBTRANSShmemInit(void)
void
BootStrapSUBTRANS(void)
{
- int slotno;
+ Buffer buffer;
- LWLockAcquire(SubtransSLRULock, LW_EXCLUSIVE);
+ SlruPagePrecedesUnitTests(SubTransPagePrecedes, SUBTRANS_XACTS_PER_PAGE);
/* Create and zero the first page of the subtrans log */
- slotno = ZeroSUBTRANSPage(0);
+ buffer = ZeroSUBTRANSPage(0);
/* Make sure it's written out */
- SimpleLruWritePage(SubTransCtl, slotno);
- Assert(!SubTransCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(SubtransSLRULock);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -232,10 +201,15 @@ BootStrapSUBTRANS(void)
*
* Control lock must be held at entry, and will be held at exit.
*/
-static int
+static Buffer
ZeroSUBTRANSPage(int pageno)
{
- return SimpleLruZeroPage(SubTransCtl, pageno);
+ Buffer buffer;
+
+ buffer = ZeroSlruBuffer(SLRU_SUBTRANS_REL_ID, pageno);
+ MarkBufferDirty(buffer);
+
+ return buffer;
}
/*
@@ -258,7 +232,6 @@ StartupSUBTRANS(TransactionId oldestActiveXID)
* Whenever we advance into a new page, ExtendSUBTRANS will likewise zero
* the new page without regard to whatever was previously on disk.
*/
- LWLockAcquire(SubtransSLRULock, LW_EXCLUSIVE);
startPage = TransactionIdToPage(oldestActiveXID);
nextXid = ShmemVariableCache->nextXid;
@@ -266,36 +239,15 @@ StartupSUBTRANS(TransactionId oldestActiveXID)
while (startPage != endPage)
{
- (void) ZeroSUBTRANSPage(startPage);
+ UnlockReleaseBuffer(ZeroSUBTRANSPage(startPage));
startPage++;
/* must account for wraparound */
if (startPage > TransactionIdToPage(MaxTransactionId))
startPage = 0;
}
- (void) ZeroSUBTRANSPage(startPage);
-
- LWLockRelease(SubtransSLRULock);
+ UnlockReleaseBuffer(ZeroSUBTRANSPage(startPage));
}
-/*
- * Perform a checkpoint --- either during shutdown, or on-the-fly
- */
-void
-CheckPointSUBTRANS(void)
-{
- /*
- * Write dirty SUBTRANS pages to disk
- *
- * This is not actually necessary from a correctness point of view. We do
- * it merely to improve the odds that writing of dirty pages is done by
- * the checkpoint process and not by backends.
- */
- TRACE_POSTGRESQL_SUBTRANS_CHECKPOINT_START(true);
- SimpleLruWriteAll(SubTransCtl, true);
- TRACE_POSTGRESQL_SUBTRANS_CHECKPOINT_DONE(true);
-}
-
-
/*
* Make sure that SUBTRANS has room for a newly-allocated XID.
*
@@ -319,12 +271,8 @@ ExtendSUBTRANS(TransactionId newestXact)
pageno = TransactionIdToPage(newestXact);
- LWLockAcquire(SubtransSLRULock, LW_EXCLUSIVE);
-
/* Zero the page */
- ZeroSUBTRANSPage(pageno);
-
- LWLockRelease(SubtransSLRULock);
+ UnlockReleaseBuffer(ZeroSUBTRANSPage(pageno));
}
@@ -350,7 +298,7 @@ TruncateSUBTRANS(TransactionId oldestXact)
TransactionIdRetreat(oldestXact);
cutoffPage = TransactionIdToPage(oldestXact);
- SimpleLruTruncate(SubTransCtl, cutoffPage);
+ SimpleLruTruncate(SLRU_SUBTRANS_REL_ID, SubTransPagePrecedes, cutoffPage);
}
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15ab8d90d4b..9b484382f80 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -51,6 +51,7 @@
#include "access/heaptoast.h"
#include "access/multixact.h"
#include "access/rewriteheap.h"
+#include "access/slru.h"
#include "access/subtrans.h"
#include "access/timeline.h"
#include "access/transam.h"
@@ -1459,7 +1460,7 @@ WaitXLogInsertionsToFinish(XLogRecPtr upto)
*/
if (upto > reservedUpto)
{
- ereport(LOG,
+ ereport(PANIC,
(errmsg("request to flush past end of generated WAL; request %X/%X, current position %X/%X",
LSN_FORMAT_ARGS(upto), LSN_FORMAT_ARGS(reservedUpto))));
upto = reservedUpto;
@@ -4501,6 +4502,7 @@ BootStrapXLOG(void)
uint64 sysidentifier;
struct timeval tv;
pg_crc32c crc;
+ ResourceOwner resowner;
/* allow ordinary WAL segment creation, like StartupXLOG() would */
LWLockAcquire(ControlFileLock, LW_EXCLUSIVE);
@@ -4641,11 +4643,15 @@ BootStrapXLOG(void)
/* some additional ControlFile fields are set in WriteControlFile() */
WriteControlFile();
- /* Bootstrap the commit log, too */
+ /* Bootstrap the commit log too */
+ resowner = ResourceOwnerCreate(NULL, "bootstrap resowner");
+ CurrentResourceOwner = resowner;
BootStrapCLOG();
BootStrapCommitTs();
BootStrapSUBTRANS();
BootStrapMultiXact();
+ CurrentResourceOwner = NULL;
+ ResourceOwnerDelete(resowner);
pfree(buffer);
@@ -6851,15 +6857,12 @@ CheckPointGuts(XLogRecPtr checkPointRedo, int flags)
CheckPointSnapBuild();
CheckPointLogicalRewriteHeap();
CheckPointReplicationOrigin();
+ CheckPointPredicate();
+ CheckPointSLRU();
- /* Write out all dirty data in SLRUs and the main buffer pool */
+ /* Write out all dirty data in the buffer pool */
TRACE_POSTGRESQL_BUFFER_CHECKPOINT_START(flags);
CheckpointStats.ckpt_write_t = GetCurrentTimestamp();
- CheckPointCLOG();
- CheckPointCommitTs();
- CheckPointSUBTRANS();
- CheckPointMultiXact();
- CheckPointPredicate();
CheckPointBuffers(flags);
/* Perform all queued up fsyncs */
diff --git a/src/backend/commands/async.c b/src/backend/commands/async.c
index 3e1b92df030..c4249bfa861 100644
--- a/src/backend/commands/async.c
+++ b/src/backend/commands/async.c
@@ -141,6 +141,7 @@
#include "libpq/libpq.h"
#include "libpq/pqformat.h"
#include "miscadmin.h"
+#include "storage/bufmgr.h"
#include "storage/ipc.h"
#include "storage/lmgr.h"
#include "storage/proc.h"
@@ -305,12 +306,6 @@ static AsyncQueueControl *asyncQueueControl;
#define QUEUE_NEXT_LISTENER(i) (asyncQueueControl->backend[i].nextListener)
#define QUEUE_BACKEND_POS(i) (asyncQueueControl->backend[i].pos)
-/*
- * The SLRU buffer area through which we access the notification queue
- */
-static SlruCtlData NotifyCtlData;
-
-#define NotifyCtl (&NotifyCtlData)
#define QUEUE_PAGESIZE BLCKSZ
#define QUEUE_FULL_WARN_INTERVAL 5000 /* warn at most once every 5s */
@@ -521,8 +516,6 @@ AsyncShmemSize(void)
size = mul_size(MaxBackends + 1, sizeof(QueueBackendStatus));
size = add_size(size, offsetof(AsyncQueueControl, backend));
- size = add_size(size, SimpleLruShmemSize(NUM_NOTIFY_BUFFERS, 0));
-
return size;
}
@@ -565,20 +558,13 @@ AsyncShmemInit(void)
}
}
- /*
- * Set up SLRU management of the pg_notify data.
- */
- NotifyCtl->PagePrecedes = asyncQueuePagePrecedes;
- SimpleLruInit(NotifyCtl, "Notify", NUM_NOTIFY_BUFFERS, 0,
- NotifySLRULock, "pg_notify", LWTRANCHE_NOTIFY_BUFFER,
- SYNC_HANDLER_NONE);
-
if (!found)
{
/*
* During start or reboot, clean out the pg_notify directory.
*/
- (void) SlruScanDirectory(NotifyCtl, SlruScanDirCbDeleteAll, NULL);
+ (void) SlruScanDirectory(SLRU_NOTIFY_REL_ID, asyncQueuePagePrecedes,
+ SlruScanDirCbDeleteAll, NULL);
}
}
@@ -1411,10 +1397,7 @@ asyncQueueAddEntries(ListCell *nextNotify)
QueuePosition queue_head;
int pageno;
int offset;
- int slotno;
-
- /* We hold both NotifyQueueLock and NotifySLRULock during this operation */
- LWLockAcquire(NotifySLRULock, LW_EXCLUSIVE);
+ Buffer buffer;
/*
* We work with a local copy of QUEUE_HEAD, which we write back to shared
@@ -1439,13 +1422,17 @@ asyncQueueAddEntries(ListCell *nextNotify)
*/
pageno = QUEUE_POS_PAGE(queue_head);
if (QUEUE_POS_IS_ZERO(queue_head))
- slotno = SimpleLruZeroPage(NotifyCtl, pageno);
+ {
+ buffer = ZeroSlruBuffer(SLRU_NOTIFY_REL_ID, pageno);
+ }
else
- slotno = SimpleLruReadPage(NotifyCtl, pageno, true,
- InvalidTransactionId);
+ {
+ buffer = ReadSlruBuffer(SLRU_NOTIFY_REL_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ }
/* Note we mark the page dirty before writing in it */
- NotifyCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
while (nextNotify != NULL)
{
@@ -1476,7 +1463,7 @@ asyncQueueAddEntries(ListCell *nextNotify)
}
/* Now copy qe into the shared buffer page */
- memcpy(NotifyCtl->shared->page_buffer[slotno] + offset,
+ memcpy(BufferGetPage(buffer) + offset,
&qe,
qe.length);
@@ -1491,7 +1478,10 @@ asyncQueueAddEntries(ListCell *nextNotify)
* asyncQueueIsFull() ensured that there is room to create this
* page without overrunning the queue.
*/
- slotno = SimpleLruZeroPage(NotifyCtl, QUEUE_POS_PAGE(queue_head));
+ UnlockReleaseBuffer(buffer);
+ buffer = ZeroSlruBuffer(SLRU_NOTIFY_REL_ID,
+ QUEUE_POS_PAGE(queue_head));
+ MarkBufferDirty(buffer);
/*
* If the new page address is a multiple of QUEUE_CLEANUP_DELAY,
@@ -1505,12 +1495,11 @@ asyncQueueAddEntries(ListCell *nextNotify)
break;
}
}
+ UnlockReleaseBuffer(buffer);
/* Success, so update the global QUEUE_HEAD */
QUEUE_HEAD = queue_head;
- LWLockRelease(NotifySLRULock);
-
return nextNotify;
}
@@ -1983,17 +1972,16 @@ asyncQueueReadAllNotifications(void)
{
int curpage = QUEUE_POS_PAGE(pos);
int curoffset = QUEUE_POS_OFFSET(pos);
- int slotno;
int copysize;
+ Buffer buffer;
/*
- * We copy the data from SLRU into a local buffer, so as to avoid
- * holding the NotifySLRULock while we are examining the entries
- * and possibly transmitting them to our frontend. Copy only the
- * part of the page we will actually inspect.
+ * We copy the data into a local buffer, so as to avoid holding a
+ * buffer pin while we are examining the entries and possibly
+ * transmitting them to our frontend. Copy only the part of the
+ * page we will actually inspect.
*/
- slotno = SimpleLruReadPage_ReadOnly(NotifyCtl, curpage,
- InvalidTransactionId);
+ buffer = ReadSlruBuffer(SLRU_NOTIFY_REL_ID, curpage);
if (curpage == QUEUE_POS_PAGE(head))
{
/* we only want to read as far as head */
@@ -2007,10 +1995,9 @@ asyncQueueReadAllNotifications(void)
copysize = QUEUE_PAGESIZE - curoffset;
}
memcpy(page_buffer.buf + curoffset,
- NotifyCtl->shared->page_buffer[slotno] + curoffset,
+ BufferGetPage(buffer) + curoffset,
copysize);
- /* Release lock that we got from SimpleLruReadPage_ReadOnly() */
- LWLockRelease(NotifySLRULock);
+ ReleaseBuffer(buffer);
/*
* Process messages up to the stop position, end of page, or an
@@ -2207,7 +2194,7 @@ asyncQueueAdvanceTail(void)
* SimpleLruTruncate() will ask for NotifySLRULock but will also
* release the lock again.
*/
- SimpleLruTruncate(NotifyCtl, newtailpage);
+ SimpleLruTruncate(SLRU_NOTIFY_REL_ID, asyncQueuePagePrecedes, newtailpage);
/*
* Update QUEUE_STOP_PAGE. This changes asyncQueueIsFull()'s verdict
diff --git a/src/backend/postmaster/checkpointer.c b/src/backend/postmaster/checkpointer.c
index 5fc076fc149..f9c31143ae0 100644
--- a/src/backend/postmaster/checkpointer.c
+++ b/src/backend/postmaster/checkpointer.c
@@ -1172,6 +1172,24 @@ CompactCheckpointerRequestQueue(void)
/* must hold CheckpointerCommLock in exclusive mode */
Assert(LWLockHeldByMe(CheckpointerCommLock));
+ /*
+ * XXX FIXME
+ * If we were unlucky enough to be reached from
+ * TransactionIdSetTreeStatus() while trying to read in a CLOG page and
+ * having to write out some other dirty page and then finding the sync
+ * request queue full, we're now running in a critical section and we're
+ * not allowed to allocate any memory below.
+ *
+ * XXX Without this, 027_stream_regress.pl aborts below occasionally, due
+ * to pressure on its very small shared_buffers and thus also sync request
+ * queue.
+ */
+ if (CritSectionCount > 0)
+ {
+ elog(LOG, "CompactCheckpointerRequestQueue() critical section, returning");
+ return false;
+ }
+
/* Initialize skip_slot array */
skip_slot = palloc0(sizeof(bool) * CheckpointerShmem->num_requests);
diff --git a/src/backend/storage/buffer/buf_init.c b/src/backend/storage/buffer/buf_init.c
index 2862e9e412c..719f27a86a4 100644
--- a/src/backend/storage/buffer/buf_init.c
+++ b/src/backend/storage/buffer/buf_init.c
@@ -20,6 +20,7 @@
BufferDescPadded *BufferDescriptors;
char *BufferBlocks;
+XLogRecPtr *BufferExternalLSNs;
ConditionVariableMinimallyPadded *BufferIOCVArray;
WritebackContext BackendWritebackContext;
CkptSortItem *CkptBufferIds;
@@ -69,9 +70,11 @@ InitBufferPool(void)
{
bool foundBufs,
foundDescs,
+ foundLSNs,
foundIOCV,
foundBufCkpt;
+
/* Align descriptors to a cacheline boundary. */
BufferDescriptors = (BufferDescPadded *)
ShmemInitStruct("Buffer Descriptors",
@@ -88,6 +91,11 @@ InitBufferPool(void)
NBuffers * sizeof(ConditionVariableMinimallyPadded),
&foundIOCV);
+ BufferExternalLSNs = (XLogRecPtr *)
+ ShmemInitStruct("Buffer External LSNs",
+ NBuffers * sizeof(XLogRecPtr),
+ &foundLSNs);
+
/*
* The array used to sort to-be-checkpointed buffer ids is located in
* shared memory, to avoid having to allocate significant amounts of
@@ -99,10 +107,10 @@ InitBufferPool(void)
ShmemInitStruct("Checkpoint BufferIds",
NBuffers * sizeof(CkptSortItem), &foundBufCkpt);
- if (foundDescs || foundBufs || foundIOCV || foundBufCkpt)
+ if (foundDescs || foundBufs || foundIOCV || foundBufCkpt || foundLSNs)
{
/* should find all of these, or none of them */
- Assert(foundDescs && foundBufs && foundIOCV && foundBufCkpt);
+ Assert(foundDescs && foundBufs && foundIOCV && foundBufCkpt && foundLSNs);
/* note: this path is only taken in EXEC_BACKEND case */
}
else
@@ -133,6 +141,8 @@ InitBufferPool(void)
LWTRANCHE_BUFFER_CONTENT);
ConditionVariableInit(BufferDescriptorGetIOCV(buf));
+
+ BufferExternalLSNs[i] = InvalidXLogRecPtr;
}
/* Correct last entry of linked list */
@@ -166,6 +176,9 @@ BufferShmemSize(void)
/* size of data pages */
size = add_size(size, mul_size(NBuffers, BLCKSZ));
+ /* size of external LSNs */
+ size = add_size(size, mul_size(NBuffers, sizeof(XLogRecPtr)));
+
/* size of stuff controlled by freelist.c */
size = add_size(size, StrategyShmemSize());
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index b7488b5d89e..709949fdb87 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -57,10 +57,21 @@
#include "utils/resowner_private.h"
#include "utils/timestamp.h"
+/*
+ * XXX Ideally we'd switch to standard pages for SLRU data, but in the
+ * meantime we need some way to identify buffers that hold raw data (no
+ * invasive LSN, no checksums).
+ */
+#define BufferHasStandardPage(bufHdr) \
+ ((bufHdr)->tag.rlocator.dbOid != 9)
+
+#define BufferHasExternalLSN(bufHdr) \
+ !BufferHasStandardPage(bufHdr)
/* Note: these two macros only work on shared buffers, not local ones! */
#define BufHdrGetBlock(bufHdr) ((Block) (BufferBlocks + ((Size) (bufHdr)->buf_id) * BLCKSZ))
-#define BufferGetLSN(bufHdr) (PageGetLSN(BufHdrGetBlock(bufHdr)))
+#define BufferGetLSN(bufHdr) \
+ (BufferHasExternalLSN(bufHdr) ? BufferGetExternalLSN(bufHdr) : PageGetLSN(BufHdrGetBlock(bufHdr)))
/* Note: this macro only works on local buffers, not shared ones! */
#define LocalBufHdrGetBlock(bufHdr) \
@@ -809,6 +820,21 @@ ReadBufferWithoutRelcache(RelFileLocator rlocator, ForkNumber forkNum,
mode, strategy, &hit);
}
+/*
+ * Like ReadBufferWithoutRelcache, but returns the hit flag.
+ * XXX Merge
+ */
+Buffer
+ReadBufferWithoutRelcacheWithHit(RelFileLocator rlocator, ForkNumber forkNum,
+ BlockNumber blockNum, ReadBufferMode mode,
+ bool *hit)
+{
+ SMgrRelation smgr = smgropen(rlocator, InvalidBackendId);
+
+ return ReadBuffer_common(smgr, RELPERSISTENCE_PERMANENT, forkNum, blockNum,
+ mode, NULL, hit);
+}
+
/*
* ReadBuffer_common -- common logic for all ReadBuffer variants
@@ -1029,7 +1055,8 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
}
/* check for garbage data */
- if (!PageIsVerifiedExtended((Page) bufBlock, blockNum,
+ if (BufferHasStandardPage(bufHdr) &&
+ !PageIsVerifiedExtended((Page) bufBlock, blockNum,
PIV_LOG_WARNING | PIV_REPORT_STAT))
{
if (mode == RBM_ZERO_ON_ERROR || zero_damaged_pages)
@@ -1450,6 +1477,9 @@ BufferAlloc(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
LWLockRelease(newPartitionLock);
+ if (BufferHasExternalLSN(buf))
+ BufferSetExternalLSN(buf, InvalidXLogRecPtr);
+
/*
* Buffer contents are currently invalid. Try to obtain the right to
* start I/O. If StartBufferIO returns false, then someone else managed
@@ -1570,6 +1600,84 @@ retry:
StrategyFreeBuffer(buf);
}
+/*
+ * DiscardBuffer -- drop a buffer from pool.
+ *
+ * If the buffer isn't present in shared buffers, nothing happens. If it is
+ * present and not pinned, it is discarded without making any attempt to write
+ * it back out to the operating system. If I/O is in progress, we wait for it
+ * to to complete. If it is pinned, an error is raised (some other backend
+ * must still be interested in it, so it's an error to discard it).
+ */
+void
+DiscardBuffer(RelFileLocator rlocator, ForkNumber forkNum, BlockNumber blockNum)
+{
+ SMgrRelation smgr = smgropen(rlocator, InvalidBackendId);
+ BufferTag tag; /* identity of target block */
+ uint32 hash; /* hash value for tag */
+ LWLock *partitionLock; /* buffer partition lock for it */
+ int buf_id;
+ BufferDesc *bufHdr;
+ uint32 buf_state;
+
+ /* create a tag so we can lookup the buffer */
+ INIT_BUFFERTAG(tag, smgr->smgr_rlocator.locator, forkNum, blockNum);
+
+ /* determine its hash code and partition lock ID */
+ hash = BufTableHashCode(&tag);
+ partitionLock = BufMappingPartitionLock(hash);
+
+ retry:
+ /* see if the block is in the buffer pool */
+ LWLockAcquire(partitionLock, LW_SHARED);
+ buf_id = BufTableLookup(&tag, hash);
+ LWLockRelease(partitionLock);
+
+ /* didn't find it, so nothing to do */
+ if (buf_id < 0)
+ return;
+
+ /* take the buffer header lock */
+ bufHdr = GetBufferDescriptor(buf_id);
+ buf_state = LockBufHdr(bufHdr);
+
+ /*
+ * The buffer might been evicted after we released the partition lock and
+ * before we acquired the buffer header lock. If so, the buffer we've
+ * locked might contain some other data which we shouldn't touch. If the
+ * buffer hasn't been recycled, we proceed to invalidate it.
+ */
+ if (RelFileLocatorEquals(bufHdr->tag.rlocator, rlocator) &&
+ bufHdr->tag.blockNum == blockNum &&
+ bufHdr->tag.forkNum == forkNum)
+ {
+ if (buf_state & BM_IO_IN_PROGRESS)
+ {
+ UnlockBufHdr(bufHdr, buf_state);
+ WaitIO(bufHdr);
+ goto retry;
+ }
+ else if (BUF_STATE_GET_REFCOUNT(buf_state) == 0)
+ {
+ /* Nobody has it pinned, so we can immediately invalidate it. */
+ InvalidateBuffer(bufHdr); /* releases spinlock */
+ }
+ else
+ {
+ /*
+ * XXX: Is it OK to say that the contract for DiscardBuffer() is
+ * that the caller is asserting that no one else could be
+ * interested in this buffer, and therefore it's a programming
+ * error or corruption if you reach this case?
+ */
+ UnlockBufHdr(bufHdr, buf_state);
+ elog(ERROR, "cannot discard buffer that is pinned");
+ }
+ }
+ else
+ UnlockBufHdr(bufHdr, buf_state);
+}
+
/*
* MarkBufferDirty
*
@@ -2900,7 +3008,10 @@ FlushBuffer(BufferDesc *buf, SMgrRelation reln)
* buffer, other processes might be updating hint bits in it, so we must
* copy the page to private storage if we do checksumming.
*/
- bufToWrite = PageSetChecksumCopy((Page) bufBlock, buf->tag.blockNum);
+ if (BufferHasStandardPage(buf))
+ bufToWrite = PageSetChecksumCopy((Page) bufBlock, buf->tag.blockNum);
+ else
+ bufToWrite = bufBlock;
if (track_io_timing)
INSTR_TIME_SET_CURRENT(io_start);
@@ -3029,7 +3140,10 @@ BufferGetLSNAtomic(Buffer buffer)
Assert(BufferIsPinned(buffer));
buf_state = LockBufHdr(bufHdr);
- lsn = PageGetLSN(page);
+ if (BufferHasStandardPage(bufHdr))
+ lsn = PageGetLSN(page);
+ else
+ lsn = BufferGetExternalLSN(bufHdr);
UnlockBufHdr(bufHdr, buf_state);
return lsn;
@@ -3541,7 +3655,8 @@ FlushRelationBuffers(Relation rel)
errcallback.previous = error_context_stack;
error_context_stack = &errcallback;
- PageSetChecksumInplace(localpage, bufHdr->tag.blockNum);
+ if (BufferHasStandardPage(bufHdr))
+ PageSetChecksumInplace(localpage, bufHdr->tag.blockNum);
smgrwrite(RelationGetSmgr(rel),
bufHdr->tag.forkNum,
@@ -5011,3 +5126,29 @@ TestForOldSnapshot_impl(Snapshot snapshot, Relation relation)
(errcode(ERRCODE_SNAPSHOT_TOO_OLD),
errmsg("snapshot too old")));
}
+
+/*
+ * Check if a buffer tag is currently mapped.
+ *
+ * XXX Dubious semantics; needed only for multixact's handling for
+ * inconsistent states.
+ */
+bool
+BufferProbe(RelFileLocator rlocator, ForkNumber forkNum, BlockNumber blockNum)
+{
+ BufferTag tag;
+ uint32 hash;
+ LWLock *partitionLock;
+ int buf_id;
+
+ INIT_BUFFERTAG(tag, rlocator, forkNum, blockNum);
+
+ hash = BufTableHashCode(&tag);
+ partitionLock = BufMappingPartitionLock(hash);
+
+ LWLockAcquire(partitionLock, LW_SHARED);
+ buf_id = BufTableLookup(&tag, hash);
+ LWLockRelease(partitionLock);
+
+ return buf_id >= 0;
+}
diff --git a/src/backend/storage/ipc/ipci.c b/src/backend/storage/ipc/ipci.c
index 1a6f5270518..c063a201a4f 100644
--- a/src/backend/storage/ipc/ipci.c
+++ b/src/backend/storage/ipc/ipci.c
@@ -116,9 +116,7 @@ CalculateShmemSize(int *num_semaphores)
size = add_size(size, XLogPrefetchShmemSize());
size = add_size(size, XLOGShmemSize());
size = add_size(size, XLogRecoveryShmemSize());
- size = add_size(size, CLOGShmemSize());
size = add_size(size, CommitTsShmemSize());
- size = add_size(size, SUBTRANSShmemSize());
size = add_size(size, TwoPhaseShmemSize());
size = add_size(size, BackgroundWorkerShmemSize());
size = add_size(size, MultiXactShmemSize());
@@ -240,9 +238,7 @@ CreateSharedMemoryAndSemaphores(void)
XLOGShmemInit();
XLogPrefetchShmemInit();
XLogRecoveryShmemInit();
- CLOGShmemInit();
CommitTsShmemInit();
- SUBTRANSShmemInit();
MultiXactShmemInit();
InitBufferPool();
diff --git a/src/backend/storage/lmgr/lwlock.c b/src/backend/storage/lmgr/lwlock.c
index 38317edaf96..d1e1add7484 100644
--- a/src/backend/storage/lmgr/lwlock.c
+++ b/src/backend/storage/lmgr/lwlock.c
@@ -129,20 +129,6 @@ extern slock_t *ShmemLock;
extern const char *const IndividualLWLockNames[]; /* in lwlocknames.c */
static const char *const BuiltinTrancheNames[] = {
- /* LWTRANCHE_XACT_BUFFER: */
- "XactBuffer",
- /* LWTRANCHE_COMMITTS_BUFFER: */
- "CommitTSBuffer",
- /* LWTRANCHE_SUBTRANS_BUFFER: */
- "SubtransBuffer",
- /* LWTRANCHE_MULTIXACTOFFSET_BUFFER: */
- "MultiXactOffsetBuffer",
- /* LWTRANCHE_MULTIXACTMEMBER_BUFFER: */
- "MultiXactMemberBuffer",
- /* LWTRANCHE_NOTIFY_BUFFER: */
- "NotifyBuffer",
- /* LWTRANCHE_SERIAL_BUFFER: */
- "SerialBuffer",
/* LWTRANCHE_WAL_INSERT: */
"WALInsert",
/* LWTRANCHE_BUFFER_CONTENT: */
diff --git a/src/backend/storage/lmgr/lwlocknames.txt b/src/backend/storage/lmgr/lwlocknames.txt
index 6c7cf6c2956..fbd2976a600 100644
--- a/src/backend/storage/lmgr/lwlocknames.txt
+++ b/src/backend/storage/lmgr/lwlocknames.txt
@@ -17,10 +17,10 @@ WALWriteLock 8
ControlFileLock 9
# 10 was CheckpointLock
XactSLRULock 11
-SubtransSLRULock 12
+# 12 was SubtransSLRULock
MultiXactGenLock 13
-MultiXactOffsetSLRULock 14
-MultiXactMemberSLRULock 15
+# 14 was MultiXactOffsetSLRULock
+# 15 was MultiXactMemberSLRULock
RelCacheInitLock 16
CheckpointerCommLock 17
TwoPhaseStateLock 18
@@ -31,7 +31,7 @@ AutovacuumLock 22
AutovacuumScheduleLock 23
SyncScanLock 24
RelationMappingLock 25
-NotifySLRULock 26
+# 26 was NotifySLRULock
NotifyQueueLock 27
SerializableXactHashLock 28
SerializableFinishedListLock 29
@@ -43,7 +43,7 @@ DynamicSharedMemoryControlLock 34
AutoFileLock 35
ReplicationSlotAllocationLock 36
ReplicationSlotControlLock 37
-CommitTsSLRULock 38
+# 38 was CommitTsSLRULock
CommitTsLock 39
ReplicationOriginLock 40
MultiXactTruncationLock 41
diff --git a/src/backend/storage/lmgr/predicate.c b/src/backend/storage/lmgr/predicate.c
index 5136da6ea36..7c200bc8745 100644
--- a/src/backend/storage/lmgr/predicate.c
+++ b/src/backend/storage/lmgr/predicate.c
@@ -312,14 +312,6 @@
((targethash) ^ ((uint32) PointerGetDatum((predicatelocktag)->myXact)) \
<< LOG2_NUM_PREDICATELOCK_PARTITIONS)
-
-/*
- * The SLRU buffer area through which we access the old xids.
- */
-static SlruCtlData SerialSlruCtlData;
-
-#define SerialSlruCtl (&SerialSlruCtlData)
-
#define SERIAL_PAGESIZE BLCKSZ
#define SERIAL_ENTRYSIZE sizeof(SerCommitSeqNo)
#define SERIAL_ENTRIESPERPAGE (SERIAL_PAGESIZE / SERIAL_ENTRYSIZE)
@@ -331,8 +323,8 @@ static SlruCtlData SerialSlruCtlData;
#define SerialNextPage(page) (((page) >= SERIAL_MAX_PAGE) ? 0 : (page) + 1)
-#define SerialValue(slotno, xid) (*((SerCommitSeqNo *) \
- (SerialSlruCtl->shared->page_buffer[slotno] + \
+#define SerialValue(buffer, xid) (*((SerCommitSeqNo *) \
+ (BufferGetPage(buffer) + \
((((uint32) (xid)) % SERIAL_ENTRIESPERPAGE) * SERIAL_ENTRYSIZE))))
#define SerialPage(xid) (((uint32) (xid)) / SERIAL_ENTRIESPERPAGE)
@@ -867,17 +859,10 @@ SerialInit(void)
{
bool found;
- /*
- * Set up SLRU management of the pg_serial data.
- */
- SerialSlruCtl->PagePrecedes = SerialPagePrecedesLogically;
- SimpleLruInit(SerialSlruCtl, "Serial",
- NUM_SERIAL_BUFFERS, 0, SerialSLRULock, "pg_serial",
- LWTRANCHE_SERIAL_BUFFER, SYNC_HANDLER_NONE);
#ifdef USE_ASSERT_CHECKING
SerialPagePrecedesLogicallyUnitTests();
#endif
- SlruPagePrecedesUnitTests(SerialSlruCtl, SERIAL_ENTRIESPERPAGE);
+ SlruPagePrecedesUnitTests(SerialPagePrecedesLogically, SERIAL_ENTRIESPERPAGE);
/*
* Create or attach to the SerialControl structure.
@@ -907,9 +892,9 @@ SerialAdd(TransactionId xid, SerCommitSeqNo minConflictCommitSeqNo)
{
TransactionId tailXid;
int targetPage;
- int slotno;
int firstZeroPage;
bool isNewPage;
+ Buffer buffer;
Assert(TransactionIdIsValid(xid));
@@ -954,16 +939,22 @@ SerialAdd(TransactionId xid, SerCommitSeqNo minConflictCommitSeqNo)
/* Initialize intervening pages. */
while (firstZeroPage != targetPage)
{
- (void) SimpleLruZeroPage(SerialSlruCtl, firstZeroPage);
+ buffer = ZeroSlruBuffer(SLRU_SERIAL_REL_ID, firstZeroPage);
+ MarkBufferDirty(buffer);
+ UnlockReleaseBuffer(buffer);
firstZeroPage = SerialNextPage(firstZeroPage);
}
- slotno = SimpleLruZeroPage(SerialSlruCtl, targetPage);
+ buffer = ZeroSlruBuffer(SLRU_SERIAL_REL_ID, targetPage);
}
else
- slotno = SimpleLruReadPage(SerialSlruCtl, targetPage, true, xid);
+ {
+ buffer = ReadSlruBuffer(SLRU_SERIAL_REL_ID, targetPage);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ }
- SerialValue(slotno, xid) = minConflictCommitSeqNo;
- SerialSlruCtl->shared->page_dirty[slotno] = true;
+ SerialValue(buffer, xid) = minConflictCommitSeqNo;
+ MarkBufferDirty(buffer);
+ UnlockReleaseBuffer(buffer);
LWLockRelease(SerialSLRULock);
}
@@ -979,7 +970,7 @@ SerialGetMinConflictCommitSeqNo(TransactionId xid)
TransactionId headXid;
TransactionId tailXid;
SerCommitSeqNo val;
- int slotno;
+ Buffer buffer;
Assert(TransactionIdIsValid(xid));
@@ -1001,9 +992,9 @@ SerialGetMinConflictCommitSeqNo(TransactionId xid)
* The following function must be called without holding SerialSLRULock,
* but will return with that lock held, which must then be released.
*/
- slotno = SimpleLruReadPage_ReadOnly(SerialSlruCtl,
- SerialPage(xid), xid);
- val = SerialValue(slotno, xid);
+ buffer = ReadSlruBuffer(SLRU_SERIAL_REL_ID, SerialPage(xid));
+ val = SerialValue(buffer, xid);
+ ReleaseBuffer(buffer);
LWLockRelease(SerialSLRULock);
return val;
}
@@ -1122,19 +1113,7 @@ CheckPointPredicate(void)
LWLockRelease(SerialSLRULock);
/* Truncate away pages that are no longer required */
- SimpleLruTruncate(SerialSlruCtl, tailPage);
-
- /*
- * Write dirty SLRU pages to disk
- *
- * This is not actually necessary from a correctness point of view. We do
- * it merely as a debugging aid.
- *
- * We're doing this after the truncation to avoid writing pages right
- * before deleting the file in which they sit, which would be completely
- * pointless.
- */
- SimpleLruWriteAll(SerialSlruCtl, true);
+ SimpleLruTruncate(SLRU_SERIAL_REL_ID, SerialPagePrecedesLogically, tailPage);
}
/*------------------------------------------------------------------------*/
@@ -1396,7 +1375,6 @@ PredicateLockShmemSize(void)
/* Shared memory structures for SLRU tracking of old committed xids. */
size = add_size(size, sizeof(SerialControlData));
- size = add_size(size, SimpleLruShmemSize(NUM_SERIAL_BUFFERS, 0));
return size;
}
diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c
index c1a5febcbfd..bda3f02fccd 100644
--- a/src/backend/storage/smgr/smgr.c
+++ b/src/backend/storage/smgr/smgr.c
@@ -17,6 +17,7 @@
*/
#include "postgres.h"
+#include "access/slru.h"
#include "access/xlogutils.h"
#include "lib/ilist.h"
#include "storage/bufmgr.h"
@@ -82,6 +83,14 @@ static const f_smgr smgrsw[] = {
.smgr_nblocks = mdnblocks,
.smgr_truncate = mdtruncate,
.smgr_immedsync = mdimmedsync,
+ },
+ /* "SLRU" storage */
+ {
+ .smgr_open = slruopen,
+ .smgr_close = slruclose,
+ .smgr_read = slruread,
+ .smgr_write = slruwrite,
+ .smgr_writeback = slruwriteback,
}
};
@@ -176,13 +185,18 @@ smgropen(RelFileLocator rlocator, BackendId backend)
reln->smgr_targblock = InvalidBlockNumber;
for (int i = 0; i <= MAX_FORKNUM; ++i)
reln->smgr_cached_nblocks[i] = InvalidBlockNumber;
- reln->smgr_which = 0; /* we only have md.c at present */
- /* implementation-specific initialization */
- smgrsw[reln->smgr_which].smgr_open(reln);
+ /* XXX find some elegant way to do this, or something better */
+ if (rlocator.dbOid == SLRU_DB_ID)
+ reln->smgr_which = 1; /* slru.c */
+ else
+ reln->smgr_which = 0; /* md.c */
/* it has no owner yet */
dlist_push_tail(&unowned_relns, &reln->node);
+
+ /* implementation-specific initialization */
+ smgrsw[reln->smgr_which].smgr_open(reln);
}
return reln;
diff --git a/src/backend/storage/sync/sync.c b/src/backend/storage/sync/sync.c
index 9d6a9e91090..6015e8e79b4 100644
--- a/src/backend/storage/sync/sync.c
+++ b/src/backend/storage/sync/sync.c
@@ -18,9 +18,7 @@
#include <fcntl.h>
#include <sys/file.h>
-#include "access/commit_ts.h"
-#include "access/clog.h"
-#include "access/multixact.h"
+#include "access/slru.h"
#include "access/xlog.h"
#include "access/xlogutils.h"
#include "commands/tablespace.h"
@@ -106,21 +104,9 @@ static const SyncOps syncsw[] = {
.sync_unlinkfiletag = mdunlinkfiletag,
.sync_filetagmatches = mdfiletagmatches
},
- /* pg_xact */
- [SYNC_HANDLER_CLOG] = {
- .sync_syncfiletag = clogsyncfiletag
- },
- /* pg_commit_ts */
- [SYNC_HANDLER_COMMIT_TS] = {
- .sync_syncfiletag = committssyncfiletag
- },
- /* pg_multixact/offsets */
- [SYNC_HANDLER_MULTIXACT_OFFSET] = {
- .sync_syncfiletag = multixactoffsetssyncfiletag
- },
- /* pg_multixact/members */
- [SYNC_HANDLER_MULTIXACT_MEMBER] = {
- .sync_syncfiletag = multixactmemberssyncfiletag
+ /* SLRU */
+ [SYNC_HANDLER_SLRU] = {
+ .sync_syncfiletag = slrusyncfiletag
}
};
diff --git a/src/backend/utils/activity/pgstat_slru.c b/src/backend/utils/activity/pgstat_slru.c
index 28ef736735f..d3f5d558e71 100644
--- a/src/backend/utils/activity/pgstat_slru.c
+++ b/src/backend/utils/activity/pgstat_slru.c
@@ -31,7 +31,7 @@ static void pgstat_reset_slru_counter_internal(int index, TimestampTz ts);
* SLRU counters are reported within critical sections so we use static memory
* in order to avoid memory allocation.
*/
-static PgStat_SLRUStats pending_SLRUStats[SLRU_NUM_ELEMENTS];
+static PgStat_SLRUStats pending_SLRUStats[SLRU_NUM_RELS];
bool have_slrustats = false;
@@ -110,37 +110,21 @@ pgstat_fetch_slru(void)
}
/*
- * Returns SLRU name for an index. The index may be above SLRU_NUM_ELEMENTS,
- * in which case this returns NULL. This allows writing code that does not
- * know the number of entries in advance.
+ * Returns SLRU name for an index.
*/
const char *
pgstat_get_slru_name(int slru_idx)
{
- if (slru_idx < 0 || slru_idx >= SLRU_NUM_ELEMENTS)
- return NULL;
-
- return slru_names[slru_idx];
+ return SlruName(slru_idx);
}
/*
- * Determine index of entry for a SLRU with a given name. If there's no exact
- * match, returns index of the last "other" entry used for SLRUs defined in
- * external projects.
+ * Determine index of entry for a SLRU with a given name.
*/
int
pgstat_get_slru_index(const char *name)
{
- int i;
-
- for (i = 0; i < SLRU_NUM_ELEMENTS; i++)
- {
- if (strcmp(slru_names[i], name) == 0)
- return i;
- }
-
- /* return index of the last entry (which is the "other" one) */
- return (SLRU_NUM_ELEMENTS - 1);
+ return SlruRelIdByName(name);
}
/*
@@ -166,7 +150,7 @@ pgstat_slru_flush(bool nowait)
else if (!LWLockConditionalAcquire(&stats_shmem->lock, LW_EXCLUSIVE))
return true;
- for (i = 0; i < SLRU_NUM_ELEMENTS; i++)
+ for (i = 0; i < SLRU_NUM_RELS; i++)
{
PgStat_SLRUStats *sharedent = &stats_shmem->stats[i];
PgStat_SLRUStats *pendingent = &pending_SLRUStats[i];
@@ -195,7 +179,7 @@ pgstat_slru_flush(bool nowait)
void
pgstat_slru_reset_all_cb(TimestampTz ts)
{
- for (int i = 0; i < SLRU_NUM_ELEMENTS; i++)
+ for (int i = 0; i < SLRU_NUM_RELS; i++)
pgstat_reset_slru_counter_internal(i, ts);
}
@@ -213,8 +197,7 @@ pgstat_slru_snapshot_cb(void)
}
/*
- * Returns pointer to entry with counters for given SLRU (based on the name
- * stored in SlruCtl as lwlock tranche name).
+ * Returns pointer to entry with counters for given SLRU.
*/
static inline PgStat_SLRUStats *
get_slru_entry(int slru_idx)
@@ -227,7 +210,7 @@ get_slru_entry(int slru_idx)
*/
Assert(IsUnderPostmaster || !IsPostmasterEnvironment);
- Assert((slru_idx >= 0) && (slru_idx < SLRU_NUM_ELEMENTS));
+ Assert((slru_idx >= 0) && (slru_idx < SLRU_NUM_RELS));
have_slrustats = true;
diff --git a/src/backend/utils/activity/wait_event.c b/src/backend/utils/activity/wait_event.c
index da57a93034e..836e7a9b482 100644
--- a/src/backend/utils/activity/wait_event.c
+++ b/src/backend/utils/activity/wait_event.c
@@ -660,8 +660,8 @@ pgstat_get_wait_io(WaitEventIO w)
case WAIT_EVENT_REPLICATION_SLOT_WRITE:
event_name = "ReplicationSlotWrite";
break;
- case WAIT_EVENT_SLRU_FLUSH_SYNC:
- event_name = "SLRUFlushSync";
+ case WAIT_EVENT_SLRU_FLUSH:
+ event_name = "SLRUFlush";
break;
case WAIT_EVENT_SLRU_READ:
event_name = "SLRURead";
diff --git a/src/backend/utils/adt/pgstatfuncs.c b/src/backend/utils/adt/pgstatfuncs.c
index d9e2a793829..f376e5b0a68 100644
--- a/src/backend/utils/adt/pgstatfuncs.c
+++ b/src/backend/utils/adt/pgstatfuncs.c
@@ -15,6 +15,7 @@
#include "postgres.h"
#include "access/htup_details.h"
+#include "access/slru.h"
#include "access/xlog.h"
#include "access/xlogprefetcher.h"
#include "catalog/pg_authid.h"
@@ -1813,7 +1814,7 @@ pg_stat_get_slru(PG_FUNCTION_ARGS)
/* request SLRU stats from the cumulative stats system */
stats = pgstat_fetch_slru();
- for (i = 0;; i++)
+ for (i = 0; i < SLRU_NUM_RELS; i++)
{
/* for each row */
Datum values[PG_STAT_GET_SLRU_COLS] = {0};
@@ -1821,10 +1822,7 @@ pg_stat_get_slru(PG_FUNCTION_ARGS)
PgStat_SLRUStats stat;
const char *name;
- name = pgstat_get_slru_name(i);
-
- if (!name)
- break;
+ name = SlruName(i);
stat = stats[i];
diff --git a/src/backend/utils/mmgr/mcxt.c b/src/backend/utils/mmgr/mcxt.c
index e12be1b9bd8..7373538e2f7 100644
--- a/src/backend/utils/mmgr/mcxt.c
+++ b/src/backend/utils/mmgr/mcxt.c
@@ -865,7 +865,9 @@ MemoryContextAlloc(MemoryContext context, Size size)
void *ret;
AssertArg(MemoryContextIsValid(context));
- AssertNotInCriticalSection(context);
+ // XXX horrible hack: while pinning clog buffers during commit, resowner
+ // stuff allocates!
+ //AssertNotInCriticalSection(context);
if (!AllocSizeIsValid(size))
elog(ERROR, "invalid memory alloc request size %zu", size);
@@ -1072,7 +1074,13 @@ palloc(Size size)
MemoryContext context = CurrentMemoryContext;
AssertArg(MemoryContextIsValid(context));
- AssertNotInCriticalSection(context);
+ /*
+ * XXX:TM I commented this out for now, because otherwise it can fail
+ * while writing back md.c buffers due to cache pressure while reading in
+ * CLOG buffers during commit, in a critical section (md.c allocates
+ * memory to build paths...); FIXME!
+ */
+ //AssertNotInCriticalSection(context);
if (!AllocSizeIsValid(size))
elog(ERROR, "invalid memory alloc request size %zu", size);
diff --git a/src/include/access/clog.h b/src/include/access/clog.h
index 543f2e2643a..af16a2afcb9 100644
--- a/src/include/access/clog.h
+++ b/src/include/access/clog.h
@@ -40,18 +40,12 @@ extern void TransactionIdSetTreeStatus(TransactionId xid, int nsubxids,
TransactionId *subxids, XidStatus status, XLogRecPtr lsn);
extern XidStatus TransactionIdGetStatus(TransactionId xid, XLogRecPtr *lsn);
-extern Size CLOGShmemBuffers(void);
-extern Size CLOGShmemSize(void);
-extern void CLOGShmemInit(void);
extern void BootStrapCLOG(void);
extern void StartupCLOG(void);
extern void TrimCLOG(void);
-extern void CheckPointCLOG(void);
extern void ExtendCLOG(TransactionId newestXact);
extern void TruncateCLOG(TransactionId oldestXact, Oid oldestxid_datoid);
-extern int clogsyncfiletag(const FileTag *ftag, char *path);
-
/* XLOG stuff */
#define CLOG_ZEROPAGE 0x00
#define CLOG_TRUNCATE 0x10
diff --git a/src/include/access/commit_ts.h b/src/include/access/commit_ts.h
index 7662f8e1a9c..4e986fc023d 100644
--- a/src/include/access/commit_ts.h
+++ b/src/include/access/commit_ts.h
@@ -27,7 +27,6 @@ extern bool TransactionIdGetCommitTsData(TransactionId xid,
extern TransactionId GetLatestCommitTsData(TimestampTz *ts,
RepOriginId *nodeid);
-extern Size CommitTsShmemBuffers(void);
extern Size CommitTsShmemSize(void);
extern void CommitTsShmemInit(void);
extern void BootStrapCommitTs(void);
@@ -41,8 +40,6 @@ extern void SetCommitTsLimit(TransactionId oldestXact,
TransactionId newestXact);
extern void AdvanceOldestCommitTsXid(TransactionId oldestXact);
-extern int committssyncfiletag(const FileTag *ftag, char *path);
-
/* XLOG stuff */
#define COMMIT_TS_ZEROPAGE 0x00
#define COMMIT_TS_TRUNCATE 0x10
diff --git a/src/include/access/multixact.h b/src/include/access/multixact.h
index a5600a320ae..a6bbe53e1bb 100644
--- a/src/include/access/multixact.h
+++ b/src/include/access/multixact.h
@@ -118,9 +118,6 @@ extern bool MultiXactIdPrecedes(MultiXactId multi1, MultiXactId multi2);
extern bool MultiXactIdPrecedesOrEquals(MultiXactId multi1,
MultiXactId multi2);
-extern int multixactoffsetssyncfiletag(const FileTag *ftag, char *path);
-extern int multixactmemberssyncfiletag(const FileTag *ftag, char *path);
-
extern void AtEOXact_MultiXact(void);
extern void AtPrepare_MultiXact(void);
extern void PostPrepare_MultiXact(TransactionId xid);
diff --git a/src/include/access/slru.h b/src/include/access/slru.h
index 130c41c8632..ed85dd019d9 100644
--- a/src/include/access/slru.h
+++ b/src/include/access/slru.h
@@ -1,7 +1,7 @@
/*-------------------------------------------------------------------------
*
* slru.h
- * Simple LRU buffering for transaction status logfiles
+ * Buffering for transaction status logfiles
*
* Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group
* Portions Copyright (c) 1994, Regents of the University of California
@@ -14,9 +14,34 @@
#define SLRU_H
#include "access/xlogdefs.h"
+#include "catalog/pg_tablespace_d.h"
+#include "storage/buf.h"
#include "storage/lwlock.h"
+#include "storage/smgr.h"
#include "storage/sync.h"
+/* Pseudo database ID used for SLRU data. */
+#define SLRU_DB_ID 9
+
+/* Pseudo relation IDs used by each cache. */
+#define SLRU_CLOG_REL_ID 0
+#define SLRU_MULTIXACT_OFFSET_REL_ID 1
+#define SLRU_MULTIXACT_MEMBER_REL_ID 2
+#define SLRU_SUBTRANS_REL_ID 3
+#define SLRU_SERIAL_REL_ID 4
+#define SLRU_COMMITTS_REL_ID 5
+#define SLRU_NOTIFY_REL_ID 6
+#define SLRU_NUM_RELS 7
+
+typedef bool (*SlruPagePrecedesFunction) (int, int);
+
+static inline RelFileLocator
+SlruRelFileLocator(RelFileNumber relNumber)
+{
+ RelFileLocator rlocator = {DEFAULTTABLESPACE_OID, SLRU_DB_ID, relNumber};
+ return rlocator;
+}
+
/*
* Define SLRU segment size. A page is the same BLCKSZ as is used everywhere
@@ -33,142 +58,54 @@
*/
#define SLRU_PAGES_PER_SEGMENT 32
-/*
- * Page status codes. Note that these do not include the "dirty" bit.
- * page_dirty can be true only in the VALID or WRITE_IN_PROGRESS states;
- * in the latter case it implies that the page has been re-dirtied since
- * the write started.
- */
-typedef enum
-{
- SLRU_PAGE_EMPTY, /* buffer is not in use */
- SLRU_PAGE_READ_IN_PROGRESS, /* page is being read in */
- SLRU_PAGE_VALID, /* page is valid and not being written */
- SLRU_PAGE_WRITE_IN_PROGRESS /* page is being written out */
-} SlruPageStatus;
-
-/*
- * Shared-memory state
- */
-typedef struct SlruSharedData
-{
- LWLock *ControlLock;
-
- /* Number of buffers managed by this SLRU structure */
- int num_slots;
-
- /*
- * Arrays holding info for each buffer slot. Page number is undefined
- * when status is EMPTY, as is page_lru_count.
- */
- char **page_buffer;
- SlruPageStatus *page_status;
- bool *page_dirty;
- int *page_number;
- int *page_lru_count;
- LWLockPadded *buffer_locks;
-
- /*
- * Optional array of WAL flush LSNs associated with entries in the SLRU
- * pages. If not zero/NULL, we must flush WAL before writing pages (true
- * for pg_xact, false for multixact, pg_subtrans, pg_notify). group_lsn[]
- * has lsn_groups_per_page entries per buffer slot, each containing the
- * highest LSN known for a contiguous group of SLRU entries on that slot's
- * page.
- */
- XLogRecPtr *group_lsn;
- int lsn_groups_per_page;
-
- /*----------
- * We mark a page "most recently used" by setting
- * page_lru_count[slotno] = ++cur_lru_count;
- * The oldest page is therefore the one with the highest value of
- * cur_lru_count - page_lru_count[slotno]
- * The counts will eventually wrap around, but this calculation still
- * works as long as no page's age exceeds INT_MAX counts.
- *----------
- */
- int cur_lru_count;
-
- /*
- * latest_page_number is the page number of the current end of the log;
- * this is not critical data, since we use it only to avoid swapping out
- * the latest page.
- */
- int latest_page_number;
-
- /* SLRU's index for statistics purposes (might not be unique) */
- int slru_stats_idx;
-} SlruSharedData;
-
-typedef SlruSharedData *SlruShared;
-
-/*
- * SlruCtlData is an unshared structure that points to the active information
- * in shared memory.
- */
-typedef struct SlruCtlData
-{
- SlruShared shared;
-
- /*
- * Which sync handler function to use when handing sync requests over to
- * the checkpointer. SYNC_HANDLER_NONE to disable fsync (eg pg_notify).
- */
- SyncRequestHandler sync_handler;
-
- /*
- * Decide whether a page is "older" for truncation and as a hint for
- * evicting pages in LRU order. Return true if every entry of the first
- * argument is older than every entry of the second argument. Note that
- * !PagePrecedes(a,b) && !PagePrecedes(b,a) need not imply a==b; it also
- * arises when some entries are older and some are not. For SLRUs using
- * SimpleLruTruncate(), this must use modular arithmetic. (For others,
- * the behavior of this callback has no functional implications.) Use
- * SlruPagePrecedesUnitTests() in SLRUs meeting its criteria.
- */
- bool (*PagePrecedes) (int, int);
-
- /*
- * Dir is set during SimpleLruInit and does not change thereafter. Since
- * it's always the same, it doesn't need to be in shared memory.
- */
- char Dir[64];
-} SlruCtlData;
-
-typedef SlruCtlData *SlruCtl;
-
-
-extern Size SimpleLruShmemSize(int nslots, int nlsns);
-extern void SimpleLruInit(SlruCtl ctl, const char *name, int nslots, int nlsns,
- LWLock *ctllock, const char *subdir, int tranche_id,
- SyncRequestHandler sync_handler);
-extern int SimpleLruZeroPage(SlruCtl ctl, int pageno);
-extern int SimpleLruReadPage(SlruCtl ctl, int pageno, bool write_ok,
- TransactionId xid);
-extern int SimpleLruReadPage_ReadOnly(SlruCtl ctl, int pageno,
- TransactionId xid);
-extern void SimpleLruWritePage(SlruCtl ctl, int slotno);
-extern void SimpleLruWriteAll(SlruCtl ctl, bool allow_redirtied);
#ifdef USE_ASSERT_CHECKING
-extern void SlruPagePrecedesUnitTests(SlruCtl ctl, int per_page);
+extern void SlruPagePrecedesUnitTests(SlruPagePrecedesFunction PagePrecedes,
+ int per_page);
#else
#define SlruPagePrecedesUnitTests(ctl, per_page) do {} while (0)
#endif
-extern void SimpleLruTruncate(SlruCtl ctl, int cutoffPage);
-extern bool SimpleLruDoesPhysicalPageExist(SlruCtl ctl, int pageno);
+extern void SimpleLruTruncate(Oid rel_id, SlruPagePrecedesFunction PagePrecedes,
+ int cutoffPage);
+extern bool SimpleLruDoesPhysicalPageExist(Oid rel_id, int pageno);
-typedef bool (*SlruScanCallback) (SlruCtl ctl, char *filename, int segpage,
+typedef bool (*SlruScanCallback) (Oid rel_id,
+ SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage,
void *data);
-extern bool SlruScanDirectory(SlruCtl ctl, SlruScanCallback callback, void *data);
-extern void SlruDeleteSegment(SlruCtl ctl, int segno);
-
-extern int SlruSyncFileTag(SlruCtl ctl, const FileTag *ftag, char *path);
+extern bool SlruScanDirectory(Oid rel_id, SlruPagePrecedesFunction PagePrecedes,
+ SlruScanCallback callback, void *data);
+extern void SlruDeleteSegment(Oid rel_id, int segno);
/* SlruScanDirectory public callbacks */
-extern bool SlruScanDirCbReportPresence(SlruCtl ctl, char *filename,
+extern bool SlruScanDirCbReportPresence(Oid rel_id,
+ SlruPagePrecedesFunction PagePrecedes,
+ char *filename,
int segpage, void *data);
-extern bool SlruScanDirCbDeleteAll(SlruCtl ctl, char *filename, int segpage,
+extern bool SlruScanDirCbDeleteAll(Oid rel_id, SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage,
void *data);
+extern void CheckPointSLRU(void);
+
+/* Buffer access */
+extern Buffer ReadSlruBuffer(Oid rel_id, int pageno);
+extern Buffer ZeroSlruBuffer(Oid rel_id, int pageno);
+
+/* Interfaces use by stats view */
+extern Oid SlruRelIdByName(const char *name);
+extern const char *SlruName(Oid rel_id);
+
+/* Sync callback */
+extern int slrusyncfiletag(const FileTag *ftag, char *path);
+
+/* SMGR callbacks */
+extern void slruopen(SMgrRelation reln);
+extern void slruclose(SMgrRelation reln, ForkNumber forknum);
+extern void slruread(SMgrRelation reln, ForkNumber forknum,
+ BlockNumber blocknum, char *buffer);
+extern void slruwrite(SMgrRelation reln, ForkNumber forknum,
+ BlockNumber blocknum, char *buffer, bool skipFsync);
+extern void slruwriteback(SMgrRelation reln, ForkNumber forknum,
+ BlockNumber blocknum, BlockNumber nblocks);
+
#endif /* SLRU_H */
diff --git a/src/include/access/subtrans.h b/src/include/access/subtrans.h
index f94e116640b..4685a05bc92 100644
--- a/src/include/access/subtrans.h
+++ b/src/include/access/subtrans.h
@@ -18,11 +18,8 @@ extern void SubTransSetParent(TransactionId xid, TransactionId parent);
extern TransactionId SubTransGetParent(TransactionId xid);
extern TransactionId SubTransGetTopmostTransaction(TransactionId xid);
-extern Size SUBTRANSShmemSize(void);
-extern void SUBTRANSShmemInit(void);
extern void BootStrapSUBTRANS(void);
extern void StartupSUBTRANS(TransactionId oldestActiveXID);
-extern void CheckPointSUBTRANS(void);
extern void ExtendSUBTRANS(TransactionId newestXact);
extern void TruncateSUBTRANS(TransactionId oldestXact);
diff --git a/src/include/pgstat.h b/src/include/pgstat.h
index ac28f813b4e..bb7099fc809 100644
--- a/src/include/pgstat.h
+++ b/src/include/pgstat.h
@@ -691,5 +691,4 @@ extern PGDLLIMPORT SessionEndType pgStatSessionEndCause;
/* updated directly by backends and background processes */
extern PGDLLIMPORT PgStat_WalStats PendingWalStats;
-
#endif /* PGSTAT_H */
diff --git a/src/include/storage/buf_internals.h b/src/include/storage/buf_internals.h
index 69e45900bae..f071539281a 100644
--- a/src/include/storage/buf_internals.h
+++ b/src/include/storage/buf_internals.h
@@ -230,6 +230,11 @@ typedef union BufferDescPadded
#define BufferDescriptorGetContentLock(bdesc) \
((LWLock*) (&(bdesc)->content_lock))
+#define BufferGetExternalLSN(bufHdr) \
+ BufferExternalLSNs[(bufHdr)->buf_id]
+#define BufferSetExternalLSN(bufHdr, lsn) \
+ BufferExternalLSNs[(bufHdr)->buf_id] = (lsn)
+
extern PGDLLIMPORT ConditionVariableMinimallyPadded *BufferIOCVArray;
/*
@@ -276,6 +281,7 @@ typedef struct WritebackContext
/* in buf_init.c */
extern PGDLLIMPORT BufferDescPadded *BufferDescriptors;
+extern PGDLLIMPORT XLogRecPtr *BufferExternalLSNs;
extern PGDLLIMPORT WritebackContext BackendWritebackContext;
/* in localbuf.c */
diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h
index bf8cce7ccf6..b7c7a1eae9d 100644
--- a/src/include/storage/bufmgr.h
+++ b/src/include/storage/bufmgr.h
@@ -116,12 +116,21 @@ extern Buffer ReadBufferWithoutRelcache(RelFileLocator rlocator,
ForkNumber forkNum, BlockNumber blockNum,
ReadBufferMode mode, BufferAccessStrategy strategy,
bool permanent);
+extern Buffer ReadBufferWithoutRelcacheWithHit(RelFileLocator rlocator,
+ ForkNumber forkNum, BlockNumber blockNum,
+ ReadBufferMode mode, bool *hit);
extern void ReleaseBuffer(Buffer buffer);
extern void UnlockReleaseBuffer(Buffer buffer);
extern void MarkBufferDirty(Buffer buffer);
extern void IncrBufferRefCount(Buffer buffer);
extern Buffer ReleaseAndReadBuffer(Buffer buffer, Relation relation,
BlockNumber blockNum);
+extern void DiscardBuffer(RelFileLocator rlocator,
+ ForkNumber forkNum,
+ BlockNumber blockNum);
+extern bool BufferProbe(RelFileLocator rlocator,
+ ForkNumber forkNum,
+ BlockNumber blockNum);
extern void InitBufferPool(void);
extern void InitBufferPoolAccess(void);
diff --git a/src/include/storage/lwlock.h b/src/include/storage/lwlock.h
index e03d317eeac..d71bf58de22 100644
--- a/src/include/storage/lwlock.h
+++ b/src/include/storage/lwlock.h
@@ -166,14 +166,7 @@ extern void LWLockInitialize(LWLock *lock, int tranche_id);
*/
typedef enum BuiltinTrancheIds
{
- LWTRANCHE_XACT_BUFFER = NUM_INDIVIDUAL_LWLOCKS,
- LWTRANCHE_COMMITTS_BUFFER,
- LWTRANCHE_SUBTRANS_BUFFER,
- LWTRANCHE_MULTIXACTOFFSET_BUFFER,
- LWTRANCHE_MULTIXACTMEMBER_BUFFER,
- LWTRANCHE_NOTIFY_BUFFER,
- LWTRANCHE_SERIAL_BUFFER,
- LWTRANCHE_WAL_INSERT,
+ LWTRANCHE_WAL_INSERT = NUM_INDIVIDUAL_LWLOCKS,
LWTRANCHE_BUFFER_CONTENT,
LWTRANCHE_REPLICATION_ORIGIN_STATE,
LWTRANCHE_REPLICATION_SLOT_IO,
diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h
index a07715356ba..7398c0d2c9f 100644
--- a/src/include/storage/smgr.h
+++ b/src/include/storage/smgr.h
@@ -16,6 +16,7 @@
#include "lib/ilist.h"
#include "storage/block.h"
+#include "storage/fd.h"
#include "storage/relfilelocator.h"
/*
@@ -68,6 +69,13 @@ typedef struct SMgrRelationData
int md_num_open_segs[MAX_FORKNUM + 1];
struct _MdfdVec *md_seg_fds[MAX_FORKNUM + 1];
+ /*
+ * for slru.c; we'll just cache one File for now.
+ * XXX Unionize with md stuff somehow so we don't waste space
+ */
+ int slru_file_segment;
+ File slru_file;
+
/* if unowned, list link in list of all unowned SMgrRelations */
dlist_node node;
} SMgrRelationData;
diff --git a/src/include/storage/sync.h b/src/include/storage/sync.h
index 049af878dec..45778b185bc 100644
--- a/src/include/storage/sync.h
+++ b/src/include/storage/sync.h
@@ -35,11 +35,7 @@ typedef enum SyncRequestType
typedef enum SyncRequestHandler
{
SYNC_HANDLER_MD = 0,
- SYNC_HANDLER_CLOG,
- SYNC_HANDLER_COMMIT_TS,
- SYNC_HANDLER_MULTIXACT_OFFSET,
- SYNC_HANDLER_MULTIXACT_MEMBER,
- SYNC_HANDLER_NONE
+ SYNC_HANDLER_SLRU
} SyncRequestHandler;
/*
diff --git a/src/include/utils/pgstat_internal.h b/src/include/utils/pgstat_internal.h
index 9303d05427f..eec150af73f 100644
--- a/src/include/utils/pgstat_internal.h
+++ b/src/include/utils/pgstat_internal.h
@@ -14,6 +14,7 @@
#define PGSTAT_INTERNAL_H
+#include "access/slru.h"
#include "common/hashfn.h"
#include "lib/dshash.h"
#include "lib/ilist.h"
@@ -260,28 +261,6 @@ typedef struct PgStat_KindInfo
} PgStat_KindInfo;
-/*
- * List of SLRU names that we keep stats for. There is no central registry of
- * SLRUs, so we use this fixed list instead. The "other" entry is used for
- * all SLRUs without an explicit entry (e.g. SLRUs in extensions).
- *
- * This is only defined here so that SLRU_NUM_ELEMENTS is known for later type
- * definitions.
- */
-static const char *const slru_names[] = {
- "CommitTs",
- "MultiXactMember",
- "MultiXactOffset",
- "Notify",
- "Serial",
- "Subtrans",
- "Xact",
- "other" /* has to be last */
-};
-
-#define SLRU_NUM_ELEMENTS lengthof(slru_names)
-
-
/* ----------
* Types and definitions for different kinds of fixed-amount stats.
*
@@ -333,7 +312,7 @@ typedef struct PgStatShared_SLRU
{
/* lock protects ->stats */
LWLock lock;
- PgStat_SLRUStats stats[SLRU_NUM_ELEMENTS];
+ PgStat_SLRUStats stats[SLRU_NUM_RELS];
} PgStatShared_SLRU;
typedef struct PgStatShared_Wal
@@ -442,7 +421,7 @@ typedef struct PgStat_Snapshot
PgStat_CheckpointerStats checkpointer;
- PgStat_SLRUStats slru[SLRU_NUM_ELEMENTS];
+ PgStat_SLRUStats slru[SLRU_NUM_RELS];
PgStat_WalStats wal;
diff --git a/src/include/utils/wait_event.h b/src/include/utils/wait_event.h
index c3ade011206..60c221a5729 100644
--- a/src/include/utils/wait_event.h
+++ b/src/include/utils/wait_event.h
@@ -203,7 +203,7 @@ typedef enum
WAIT_EVENT_REPLICATION_SLOT_RESTORE_SYNC,
WAIT_EVENT_REPLICATION_SLOT_SYNC,
WAIT_EVENT_REPLICATION_SLOT_WRITE,
- WAIT_EVENT_SLRU_FLUSH_SYNC,
+ WAIT_EVENT_SLRU_FLUSH,
WAIT_EVENT_SLRU_READ,
WAIT_EVENT_SLRU_SYNC,
WAIT_EVENT_SLRU_WRITE,
--
2.30.2
^ permalink raw reply [nested|flat] 22+ messages in thread
* Re: SLRUs in the main buffer pool, redux
@ 2022-07-25 08:54 Heikki Linnakangas <[email protected]>
parent: Heikki Linnakangas <[email protected]>
1 sibling, 1 reply; 22+ messages in thread
From: Heikki Linnakangas @ 2022-07-25 08:54 UTC (permalink / raw)
To: Thomas Munro <[email protected]>; +Cc: pgsql-hackers; Yura Sokolov <[email protected]>
On 25/07/2022 09:54, Heikki Linnakangas wrote:
> In RecordTransactionCommit(), we enter a critical section, and then call
> TransactionIdCommitTree() to update the CLOG pages. That now involves a
> call to ReadBuffer_common(), which in turn calls
> ResourceOwnerEnlargeBuffers(). That can fail, because it might require
> allocating memory, which is forbidden in a critical section. I ran into
> an assertion about that with "make check" when I was playing around with
> a heavily modified version of this patch. Haven't seen it with your
> original one, but I believe that's just luck.
>
> Calling ResourceOwnerEnlargeBuffers() before entering the critical
> section would probably fix that, although I'm a bit worried about having
> the Enlarge call so far away from the point where it's needed.
Oh I just saw that you had a comment about that in the patch and had
hacked around it. Anyway, calling ResourceOwnerEnlargeBuffers() might be
a solution. Or switch to a separate "CriticalResourceOwner" that's
guaranteed to have enough pre-allocated space, before entering the
critical section.
- Heikki
^ permalink raw reply [nested|flat] 22+ messages in thread
* Re: SLRUs in the main buffer pool, redux
@ 2022-07-25 09:59 Heikki Linnakangas <[email protected]>
parent: Heikki Linnakangas <[email protected]>
1 sibling, 0 replies; 22+ messages in thread
From: Heikki Linnakangas @ 2022-07-25 09:59 UTC (permalink / raw)
To: Thomas Munro <[email protected]>; +Cc: pgsql-hackers; Yura Sokolov <[email protected]>
On 25/07/2022 09:54, Heikki Linnakangas wrote:
> I'll write a separate post with my thoughts on the high-level design of
> this, ...
This patch represents each SLRU as a relation. The CLOG is one relation,
pg_subtrans is another relations, and so forth. The SLRU relations use a
different SMGR implementation, which is implemented in slru.c.
As you know, I'd like to make the SMGR implementation replaceable by
extensions. We need that for Neon, and I'd imagine it to be useful for
many other things, too, like compression, encryption, or restoring data
from a backup on-demand. I'd like all file operations to go through the
smgr API as much as possible, so that an extension can intercept SLRU
file operations too. If we introduce another internal SMGR
implementation, then an extension would need to replace both
implementations separately. I'd prefer to use the current md.c
implementation for SLRUs too, instead.
Thus I propose:
Let's represent each SLRU *segment* as a separate relation, giving each
SLRU segment a separate relNumber. Then we can use md.c for SLRUs, too.
Dropping an SLRU segment can be done by calling smgrunlink(). You won't
need to deal with missing segments in md.c, because each individual SLRU
file is a complete file, with no holes. Dropping buffers for one SLRU
segment can be done with DropRelationBuffers(), instead of introducing
the new DiscardBuffer() function. You can let md.c handle the caching of
the file descriptors, you won't need to reimplement that with
'slru_file_segment'.
SLRUs won't need the segmentation into 1 GB segments that md.c does,
because each SLRU file is just 256 kB in size. That's OK. (BTW, I
propose that we bump the SLRU segment size up to a whopping 1 MB or even
more, while we're at it. But one step at a time.)
SLRUs also won't need the concept of relation forks. That's fine, we can
just use MAIN_FORKNUM. elated to that, I'm somewhat bothered by the way
that SMgrRelation currently bundles all the relation forks together. A
comment in smgr.h says:
> smgr.c maintains a table of SMgrRelation objects, which are essentially
> cached file handles.
But when we introduced relation forks, that got a bit muddled. Each
SMgrRelation object is now a file handle for a bunch of related relation
forks, and each fork is a separate file that can be created and
truncated separately.
That means that an SMGR implementation, like md.c, needs to track the
file handles for each fork. I think things would be more clear if we
unbundled the forks at the SMGR level, so that we would have a separate
SMgrRelation struct for each fork. And let's rename it to SMgrFile to
make the role more clear. I think that would reduce the confusion when
we start using it for SLRUs; an SLRU is not a relation, after all. md.c
would still segment each logical file into 1 GB segments, but it would
not need to deal with forks.
Attached is a draft patch to refactor it that way, and a refactored
version of your SLRU patch over that.
The relation cache now needs to hold a separate reference to the
SMgrFile of each fork of a relation. And smgr cache invalidation still
works at relation granularity. Doing it per SmgrFile would be more clean
in smgr.c, but in practice all the forks of a relation are unlinked and
truncated together, so sending a separate invalidation event for each
SMgrFile would increase the cache invalidation traffic.
In the passing, I moved the DropRelationBuffers() calls from smgr.c to
the callers. smgr.c doesn't otherwise make any effort to keep the buffer
manager in sync with the state on-disk, that responsibility is normally
with the code that *uses* the smgr functions, so I think that's more
logical.
The first patch currently causes the '018_wal_optimize.pl' test to fail.
I guess I messed up something in the relation truncation code, but I
haven't investigated it yet. I wanted to post this to get comments on
the design, before spending more time on that.
What do you think?
- Heikki
Attachments:
[text/x-patch] 0001-WIP-Have-separate-SMmgrRelation-per-fork-rename-it-t.patch (161.6K, ../../[email protected]/2-0001-WIP-Have-separate-SMmgrRelation-per-fork-rename-it-t.patch)
download | inline diff:
From 9302b6498453ea4dcb5c906a07633baa2cc1f8ad Mon Sep 17 00:00:00 2001
From: Heikki Linnakangas <[email protected]>
Date: Mon, 25 Jul 2022 12:50:44 +0300
Subject: [PATCH 1/2] WIP: Have separate SMmgrRelation per fork, rename it to
SMgrFile
---
contrib/amcheck/verify_nbtree.c | 2 +-
contrib/bloom/blinsert.c | 6 +-
contrib/pg_prewarm/autoprewarm.c | 2 +-
contrib/pg_prewarm/pg_prewarm.c | 4 +-
contrib/pg_visibility/pg_visibility.c | 5 +-
src/backend/access/gist/gistbuild.c | 11 +-
src/backend/access/hash/hashpage.c | 2 +-
src/backend/access/heap/heapam_handler.c | 32 +-
src/backend/access/heap/rewriteheap.c | 6 +-
src/backend/access/heap/visibilitymap.c | 41 ++-
src/backend/access/nbtree/nbtree.c | 6 +-
src/backend/access/nbtree/nbtsort.c | 8 +-
src/backend/access/spgist/spginsert.c | 17 +-
src/backend/access/table/tableam.c | 4 +-
src/backend/access/transam/xlogprefetcher.c | 28 +-
src/backend/access/transam/xlogutils.c | 17 +-
src/backend/catalog/catalog.c | 53 +--
src/backend/catalog/index.c | 4 +-
src/backend/catalog/storage.c | 205 +++++++----
src/backend/commands/dbcommands.c | 2 +-
src/backend/commands/sequence.c | 8 +-
src/backend/commands/tablecmds.c | 19 +-
src/backend/storage/buffer/bufmgr.c | 264 +++++++-------
src/backend/storage/buffer/localbuf.c | 25 +-
src/backend/storage/freespace/freespace.c | 41 ++-
src/backend/storage/smgr/md.c | 359 ++++++++-----------
src/backend/storage/smgr/smgr.c | 376 +++++++++-----------
src/backend/utils/cache/inval.c | 17 +-
src/backend/utils/cache/relcache.c | 15 +-
src/common/relpath.c | 4 +-
src/include/catalog/storage.h | 11 +-
src/include/common/relpath.h | 12 +-
src/include/storage/buf_internals.h | 5 +-
src/include/storage/bufmgr.h | 25 +-
src/include/storage/md.h | 28 +-
src/include/storage/relfilelocator.h | 29 +-
src/include/storage/smgr.h | 98 +++--
src/include/utils/inval.h | 2 +-
src/include/utils/rel.h | 38 +-
39 files changed, 885 insertions(+), 946 deletions(-)
diff --git a/contrib/amcheck/verify_nbtree.c b/contrib/amcheck/verify_nbtree.c
index 2beeebb1635..3aa20bb5ccc 100644
--- a/contrib/amcheck/verify_nbtree.c
+++ b/contrib/amcheck/verify_nbtree.c
@@ -323,7 +323,7 @@ bt_index_check_internal(Oid indrelid, bool parentcheck, bool heapallindexed,
bool heapkeyspace,
allequalimage;
- if (!smgrexists(RelationGetSmgr(indrel), MAIN_FORKNUM))
+ if (!smgrexists(RelationGetSmgr(indrel, MAIN_FORKNUM)))
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("index \"%s\" lacks a main relation fork",
diff --git a/contrib/bloom/blinsert.c b/contrib/bloom/blinsert.c
index e64291e049a..49f3a46c125 100644
--- a/contrib/bloom/blinsert.c
+++ b/contrib/bloom/blinsert.c
@@ -177,9 +177,9 @@ blbuildempty(Relation index)
* this even when wal_level=minimal.
*/
PageSetChecksumInplace(metapage, BLOOM_METAPAGE_BLKNO);
- smgrwrite(RelationGetSmgr(index), INIT_FORKNUM, BLOOM_METAPAGE_BLKNO,
+ smgrwrite(RelationGetSmgr(index, INIT_FORKNUM), BLOOM_METAPAGE_BLKNO,
(char *) metapage, true);
- log_newpage(&(RelationGetSmgr(index))->smgr_rlocator.locator, INIT_FORKNUM,
+ log_newpage(&index->rd_locator, INIT_FORKNUM,
BLOOM_METAPAGE_BLKNO, metapage, true);
/*
@@ -187,7 +187,7 @@ blbuildempty(Relation index)
* write did not go through shared_buffers and therefore a concurrent
* checkpoint may have moved the redo pointer past our xlog record.
*/
- smgrimmedsync(RelationGetSmgr(index), INIT_FORKNUM);
+ smgrimmedsync(RelationGetSmgr(index, INIT_FORKNUM));
}
/*
diff --git a/contrib/pg_prewarm/autoprewarm.c b/contrib/pg_prewarm/autoprewarm.c
index b2d60260934..92d23f45b60 100644
--- a/contrib/pg_prewarm/autoprewarm.c
+++ b/contrib/pg_prewarm/autoprewarm.c
@@ -535,7 +535,7 @@ autoprewarm_database_main(Datum main_arg)
*/
if (blk->forknum > InvalidForkNumber &&
blk->forknum <= MAX_FORKNUM &&
- smgrexists(RelationGetSmgr(rel), blk->forknum))
+ smgrexists(RelationGetSmgr(rel, blk->forknum)))
nblocks = RelationGetNumberOfBlocksInFork(rel, blk->forknum);
else
nblocks = 0;
diff --git a/contrib/pg_prewarm/pg_prewarm.c b/contrib/pg_prewarm/pg_prewarm.c
index caff5c4a80f..1e57481ef92 100644
--- a/contrib/pg_prewarm/pg_prewarm.c
+++ b/contrib/pg_prewarm/pg_prewarm.c
@@ -109,7 +109,7 @@ pg_prewarm(PG_FUNCTION_ARGS)
aclcheck_error(aclresult, get_relkind_objtype(rel->rd_rel->relkind), get_rel_name(relOid));
/* Check that the fork exists. */
- if (!smgrexists(RelationGetSmgr(rel), forkNumber))
+ if (!smgrexists(RelationGetSmgr(rel, forkNumber)))
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("fork \"%s\" does not exist for this relation",
@@ -177,7 +177,7 @@ pg_prewarm(PG_FUNCTION_ARGS)
for (block = first_block; block <= last_block; ++block)
{
CHECK_FOR_INTERRUPTS();
- smgrread(RelationGetSmgr(rel), forkNumber, block, blockbuffer.data);
+ smgrread(RelationGetSmgr(rel, forkNumber), block, blockbuffer.data);
++blocks_done;
}
}
diff --git a/contrib/pg_visibility/pg_visibility.c b/contrib/pg_visibility/pg_visibility.c
index a95f73ec796..38cb56c9bce 100644
--- a/contrib/pg_visibility/pg_visibility.c
+++ b/contrib/pg_visibility/pg_visibility.c
@@ -388,13 +388,14 @@ pg_truncate_visibility_map(PG_FUNCTION_ARGS)
check_relation_relkind(rel);
/* Forcibly reset cached file size */
- RelationGetSmgr(rel)->smgr_cached_nblocks[VISIBILITYMAP_FORKNUM] = InvalidBlockNumber;
+ RelationGetSmgr(rel, VISIBILITYMAP_FORKNUM)->smgr_cached_nblocks = InvalidBlockNumber;
block = visibilitymap_prepare_truncate(rel, 0);
if (BlockNumberIsValid(block))
{
fork = VISIBILITYMAP_FORKNUM;
- smgrtruncate(RelationGetSmgr(rel), &fork, 1, &block);
+ DropRelationBuffers(rel->rd_locator, rel->rd_backend, &fork, 1, &block);
+ smgrtruncate_multi(rel->rd_locator, rel->rd_backend, &fork, 1, &block);
}
if (RelationNeedsWAL(rel))
diff --git a/src/backend/access/gist/gistbuild.c b/src/backend/access/gist/gistbuild.c
index 374e64e8086..0a581cd8cac 100644
--- a/src/backend/access/gist/gistbuild.c
+++ b/src/backend/access/gist/gistbuild.c
@@ -415,7 +415,7 @@ gist_indexsortbuild(GISTBuildState *state)
* replaced with the real root page at the end.
*/
page = palloc0(BLCKSZ);
- smgrextend(RelationGetSmgr(state->indexrel), MAIN_FORKNUM, GIST_ROOT_BLKNO,
+ smgrextend(RelationGetSmgr(state->indexrel, MAIN_FORKNUM), GIST_ROOT_BLKNO,
page, true);
state->pages_allocated++;
state->pages_written++;
@@ -459,7 +459,7 @@ gist_indexsortbuild(GISTBuildState *state)
/* Write out the root */
PageSetLSN(levelstate->pages[0], GistBuildLSN);
PageSetChecksumInplace(levelstate->pages[0], GIST_ROOT_BLKNO);
- smgrwrite(RelationGetSmgr(state->indexrel), MAIN_FORKNUM, GIST_ROOT_BLKNO,
+ smgrwrite(RelationGetSmgr(state->indexrel, MAIN_FORKNUM), GIST_ROOT_BLKNO,
levelstate->pages[0], true);
if (RelationNeedsWAL(state->indexrel))
log_newpage(&state->indexrel->rd_locator, MAIN_FORKNUM, GIST_ROOT_BLKNO,
@@ -478,7 +478,7 @@ gist_indexsortbuild(GISTBuildState *state)
* still not be on disk when the crash occurs.
*/
if (RelationNeedsWAL(state->indexrel))
- smgrimmedsync(RelationGetSmgr(state->indexrel), MAIN_FORKNUM);
+ smgrimmedsync(RelationGetSmgr(state->indexrel, MAIN_FORKNUM));
}
/*
@@ -656,7 +656,7 @@ gist_indexsortbuild_flush_ready_pages(GISTBuildState *state)
PageSetLSN(page, GistBuildLSN);
PageSetChecksumInplace(page, blkno);
- smgrextend(RelationGetSmgr(state->indexrel), MAIN_FORKNUM, blkno, page,
+ smgrextend(RelationGetSmgr(state->indexrel, MAIN_FORKNUM), blkno, page,
true);
state->pages_written++;
@@ -942,8 +942,7 @@ gistBuildCallback(Relation index,
*/
if ((buildstate->buildMode == GIST_BUFFERING_AUTO &&
buildstate->indtuples % BUFFERING_MODE_SWITCH_CHECK_STEP == 0 &&
- effective_cache_size < smgrnblocks(RelationGetSmgr(index),
- MAIN_FORKNUM)) ||
+ effective_cache_size < smgrnblocks(RelationGetSmgr(index, MAIN_FORKNUM))) ||
(buildstate->buildMode == GIST_BUFFERING_STATS &&
buildstate->indtuples >= BUFFERING_MODE_TUPLE_SIZE_STATS_TARGET))
{
diff --git a/src/backend/access/hash/hashpage.c b/src/backend/access/hash/hashpage.c
index d2edcd46172..96133b0fd49 100644
--- a/src/backend/access/hash/hashpage.c
+++ b/src/backend/access/hash/hashpage.c
@@ -1026,7 +1026,7 @@ _hash_alloc_buckets(Relation rel, BlockNumber firstblock, uint32 nblocks)
true);
PageSetChecksumInplace(page, lastblock);
- smgrextend(RelationGetSmgr(rel), MAIN_FORKNUM, lastblock, zerobuf.data,
+ smgrextend(RelationGetSmgr(rel, MAIN_FORKNUM), lastblock, zerobuf.data,
false);
return true;
diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c
index a3414a76e8d..0f18f51a56b 100644
--- a/src/backend/access/heap/heapam_handler.c
+++ b/src/backend/access/heap/heapam_handler.c
@@ -572,7 +572,7 @@ heapam_relation_set_new_filelocator(Relation rel,
TransactionId *freezeXid,
MultiXactId *minmulti)
{
- SMgrRelation srel;
+ SMgrFileHandle sfile;
/*
* Initialize to the minimum XID that could put tuples in the table. We
@@ -591,7 +591,7 @@ heapam_relation_set_new_filelocator(Relation rel,
*/
*minmulti = GetOldestMultiXactId();
- srel = RelationCreateStorage(*newrlocator, persistence, true);
+ sfile = RelationCreateStorage(*newrlocator, persistence, true);
/*
* If required, set up an init fork for an unlogged table so that it can
@@ -604,15 +604,18 @@ heapam_relation_set_new_filelocator(Relation rel,
*/
if (persistence == RELPERSISTENCE_UNLOGGED)
{
+ SMgrFileHandle sfile_init;
+
Assert(rel->rd_rel->relkind == RELKIND_RELATION ||
rel->rd_rel->relkind == RELKIND_MATVIEW ||
rel->rd_rel->relkind == RELKIND_TOASTVALUE);
- smgrcreate(srel, INIT_FORKNUM, false);
+ sfile_init = smgropen(*newrlocator, InvalidBackendId, INIT_FORKNUM);
+ smgrcreate(sfile_init, false);
log_smgrcreate(newrlocator, INIT_FORKNUM);
- smgrimmedsync(srel, INIT_FORKNUM);
+ smgrimmedsync(sfile);
}
- smgrclose(srel);
+ smgrclose(sfile);
}
static void
@@ -624,9 +627,7 @@ heapam_relation_nontransactional_truncate(Relation rel)
static void
heapam_relation_copy_data(Relation rel, const RelFileLocator *newrlocator)
{
- SMgrRelation dstrel;
-
- dstrel = smgropen(*newrlocator, rel->rd_backend);
+ SMgrFileHandle dstmain;
/*
* Since we copy the file directly without looking at the shared buffers,
@@ -646,16 +647,21 @@ heapam_relation_copy_data(Relation rel, const RelFileLocator *newrlocator)
RelationCreateStorage(*newrlocator, rel->rd_rel->relpersistence, true);
/* copy main fork */
- RelationCopyStorage(RelationGetSmgr(rel), dstrel, MAIN_FORKNUM,
+ dstmain = smgropen(*newrlocator, rel->rd_backend, MAIN_FORKNUM);
+ RelationCopyStorage(RelationGetSmgr(rel, MAIN_FORKNUM), dstmain,
rel->rd_rel->relpersistence);
/* copy those extra forks that exist */
for (ForkNumber forkNum = MAIN_FORKNUM + 1;
forkNum <= MAX_FORKNUM; forkNum++)
{
- if (smgrexists(RelationGetSmgr(rel), forkNum))
+ SMgrFileHandle src_fork = RelationGetSmgr(rel, forkNum);
+
+ if (smgrexists(src_fork))
{
- smgrcreate(dstrel, forkNum, false);
+ SMgrFileHandle dst_fork = smgropen(*newrlocator, rel->rd_backend, forkNum);
+
+ smgrcreate(dst_fork, false);
/*
* WAL log creation if the relation is persistent, or this is the
@@ -665,7 +671,7 @@ heapam_relation_copy_data(Relation rel, const RelFileLocator *newrlocator)
(rel->rd_rel->relpersistence == RELPERSISTENCE_UNLOGGED &&
forkNum == INIT_FORKNUM))
log_smgrcreate(newrlocator, forkNum);
- RelationCopyStorage(RelationGetSmgr(rel), dstrel, forkNum,
+ RelationCopyStorage(RelationGetSmgr(rel, forkNum), dst_fork,
rel->rd_rel->relpersistence);
}
}
@@ -673,7 +679,7 @@ heapam_relation_copy_data(Relation rel, const RelFileLocator *newrlocator)
/* drop old relation, and close new one */
RelationDropStorage(rel);
- smgrclose(dstrel);
+ smgrclose(dstmain);
}
static void
diff --git a/src/backend/access/heap/rewriteheap.c b/src/backend/access/heap/rewriteheap.c
index 197f06b5eca..8a86dc8aaea 100644
--- a/src/backend/access/heap/rewriteheap.c
+++ b/src/backend/access/heap/rewriteheap.c
@@ -326,7 +326,7 @@ end_heap_rewrite(RewriteState state)
PageSetChecksumInplace(state->rs_buffer, state->rs_blockno);
- smgrextend(RelationGetSmgr(state->rs_new_rel), MAIN_FORKNUM,
+ smgrextend(RelationGetSmgr(state->rs_new_rel, MAIN_FORKNUM),
state->rs_blockno, (char *) state->rs_buffer, true);
}
@@ -338,7 +338,7 @@ end_heap_rewrite(RewriteState state)
* wrote before the checkpoint.
*/
if (RelationNeedsWAL(state->rs_new_rel))
- smgrimmedsync(RelationGetSmgr(state->rs_new_rel), MAIN_FORKNUM);
+ smgrimmedsync(RelationGetSmgr(state->rs_new_rel, MAIN_FORKNUM));
logical_end_heap_rewrite(state);
@@ -692,7 +692,7 @@ raw_heap_insert(RewriteState state, HeapTuple tup)
*/
PageSetChecksumInplace(page, state->rs_blockno);
- smgrextend(RelationGetSmgr(state->rs_new_rel), MAIN_FORKNUM,
+ smgrextend(RelationGetSmgr(state->rs_new_rel, MAIN_FORKNUM),
state->rs_blockno, (char *) page, true);
state->rs_blockno++;
diff --git a/src/backend/access/heap/visibilitymap.c b/src/backend/access/heap/visibilitymap.c
index ed72eb7b631..796f57ee046 100644
--- a/src/backend/access/heap/visibilitymap.c
+++ b/src/backend/access/heap/visibilitymap.c
@@ -460,7 +460,7 @@ visibilitymap_prepare_truncate(Relation rel, BlockNumber nheapblocks)
* If no visibility map has been created yet for this relation, there's
* nothing to truncate.
*/
- if (!smgrexists(RelationGetSmgr(rel), VISIBILITYMAP_FORKNUM))
+ if (!smgrexists(RelationGetSmgr(rel, VISIBILITYMAP_FORKNUM)))
return InvalidBlockNumber;
/*
@@ -527,7 +527,7 @@ visibilitymap_prepare_truncate(Relation rel, BlockNumber nheapblocks)
else
newnblocks = truncBlock;
- if (smgrnblocks(RelationGetSmgr(rel), VISIBILITYMAP_FORKNUM) <= newnblocks)
+ if (smgrnblocks(RelationGetSmgr(rel, VISIBILITYMAP_FORKNUM)) <= newnblocks)
{
/* nothing to do, the file was already smaller than requested size */
return InvalidBlockNumber;
@@ -546,29 +546,29 @@ static Buffer
vm_readbuf(Relation rel, BlockNumber blkno, bool extend)
{
Buffer buf;
- SMgrRelation reln;
+ SMgrFileHandle vm_sfile;
/*
* Caution: re-using this smgr pointer could fail if the relcache entry
* gets closed. It's safe as long as we only do smgr-level operations
* between here and the last use of the pointer.
*/
- reln = RelationGetSmgr(rel);
+ vm_sfile = RelationGetSmgr(rel, VISIBILITYMAP_FORKNUM);
/*
* If we haven't cached the size of the visibility map fork yet, check it
* first.
*/
- if (reln->smgr_cached_nblocks[VISIBILITYMAP_FORKNUM] == InvalidBlockNumber)
+ if (vm_sfile->smgr_cached_nblocks == InvalidBlockNumber)
{
- if (smgrexists(reln, VISIBILITYMAP_FORKNUM))
- smgrnblocks(reln, VISIBILITYMAP_FORKNUM);
+ if (smgrexists(vm_sfile))
+ smgrnblocks(vm_sfile);
else
- reln->smgr_cached_nblocks[VISIBILITYMAP_FORKNUM] = 0;
+ vm_sfile->smgr_cached_nblocks = 0;
}
/* Handle requests beyond EOF */
- if (blkno >= reln->smgr_cached_nblocks[VISIBILITYMAP_FORKNUM])
+ if (blkno >= vm_sfile->smgr_cached_nblocks)
{
if (extend)
vm_extend(rel, blkno + 1);
@@ -595,8 +595,7 @@ vm_readbuf(Relation rel, BlockNumber blkno, bool extend)
* long as it doesn't depend on the page header having correct contents.
* Current usage is safe because PageGetContents() does not require that.
*/
- buf = ReadBufferExtended(rel, VISIBILITYMAP_FORKNUM, blkno,
- RBM_ZERO_ON_ERROR, NULL);
+ buf = ReadBufferExtended(rel, VISIBILITYMAP_FORKNUM, blkno, RBM_ZERO_ON_ERROR, NULL);
if (PageIsNew(BufferGetPage(buf)))
{
LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
@@ -616,7 +615,7 @@ vm_extend(Relation rel, BlockNumber vm_nblocks)
{
BlockNumber vm_nblocks_now;
PGAlignedBlock pg;
- SMgrRelation reln;
+ SMgrFileHandle vm_sfile;
PageInit((Page) pg.data, BLCKSZ, 0);
@@ -637,27 +636,27 @@ vm_extend(Relation rel, BlockNumber vm_nblocks)
* gets closed. It's safe as long as we only do smgr-level operations
* between here and the last use of the pointer.
*/
- reln = RelationGetSmgr(rel);
+ vm_sfile = RelationGetSmgr(rel, VISIBILITYMAP_FORKNUM);
/*
* Create the file first if it doesn't exist. If smgr_vm_nblocks is
* positive then it must exist, no need for an smgrexists call.
*/
- if ((reln->smgr_cached_nblocks[VISIBILITYMAP_FORKNUM] == 0 ||
- reln->smgr_cached_nblocks[VISIBILITYMAP_FORKNUM] == InvalidBlockNumber) &&
- !smgrexists(reln, VISIBILITYMAP_FORKNUM))
- smgrcreate(reln, VISIBILITYMAP_FORKNUM, false);
+ if ((vm_sfile->smgr_cached_nblocks == 0 ||
+ vm_sfile->smgr_cached_nblocks == InvalidBlockNumber) &&
+ !smgrexists(vm_sfile))
+ smgrcreate(vm_sfile, false);
/* Invalidate cache so that smgrnblocks() asks the kernel. */
- reln->smgr_cached_nblocks[VISIBILITYMAP_FORKNUM] = InvalidBlockNumber;
- vm_nblocks_now = smgrnblocks(reln, VISIBILITYMAP_FORKNUM);
+ vm_sfile->smgr_cached_nblocks = InvalidBlockNumber;
+ vm_nblocks_now = smgrnblocks(vm_sfile);
/* Now extend the file */
while (vm_nblocks_now < vm_nblocks)
{
PageSetChecksumInplace((Page) pg.data, vm_nblocks_now);
- smgrextend(reln, VISIBILITYMAP_FORKNUM, vm_nblocks_now, pg.data, false);
+ smgrextend(vm_sfile, vm_nblocks_now, pg.data, false);
vm_nblocks_now++;
}
@@ -668,7 +667,7 @@ vm_extend(Relation rel, BlockNumber vm_nblocks)
* to keep checking for creation or extension of the file, which happens
* infrequently.
*/
- CacheInvalidateSmgr(reln->smgr_rlocator);
+ CacheInvalidateSmgr(rel->rd_locator, rel->rd_backend);
UnlockRelationForExtension(rel, ExclusiveLock);
}
diff --git a/src/backend/access/nbtree/nbtree.c b/src/backend/access/nbtree/nbtree.c
index b52eca8f38b..644431a7e3c 100644
--- a/src/backend/access/nbtree/nbtree.c
+++ b/src/backend/access/nbtree/nbtree.c
@@ -164,9 +164,9 @@ btbuildempty(Relation index)
* this even when wal_level=minimal.
*/
PageSetChecksumInplace(metapage, BTREE_METAPAGE);
- smgrwrite(RelationGetSmgr(index), INIT_FORKNUM, BTREE_METAPAGE,
+ smgrwrite(RelationGetSmgr(index, INIT_FORKNUM), BTREE_METAPAGE,
(char *) metapage, true);
- log_newpage(&RelationGetSmgr(index)->smgr_rlocator.locator, INIT_FORKNUM,
+ log_newpage(&index->rd_locator, INIT_FORKNUM,
BTREE_METAPAGE, metapage, true);
/*
@@ -174,7 +174,7 @@ btbuildempty(Relation index)
* write did not go through shared_buffers and therefore a concurrent
* checkpoint may have moved the redo pointer past our xlog record.
*/
- smgrimmedsync(RelationGetSmgr(index), INIT_FORKNUM);
+ smgrimmedsync(RelationGetSmgr(index, INIT_FORKNUM));
}
/*
diff --git a/src/backend/access/nbtree/nbtsort.c b/src/backend/access/nbtree/nbtsort.c
index bd1685c441b..53ddddfd69f 100644
--- a/src/backend/access/nbtree/nbtsort.c
+++ b/src/backend/access/nbtree/nbtsort.c
@@ -662,7 +662,7 @@ _bt_blwritepage(BTWriteState *wstate, Page page, BlockNumber blkno)
if (!wstate->btws_zeropage)
wstate->btws_zeropage = (Page) palloc0(BLCKSZ);
/* don't set checksum for all-zero page */
- smgrextend(RelationGetSmgr(wstate->index), MAIN_FORKNUM,
+ smgrextend(RelationGetSmgr(wstate->index, MAIN_FORKNUM),
wstate->btws_pages_written++,
(char *) wstate->btws_zeropage,
true);
@@ -677,14 +677,14 @@ _bt_blwritepage(BTWriteState *wstate, Page page, BlockNumber blkno)
if (blkno == wstate->btws_pages_written)
{
/* extending the file... */
- smgrextend(RelationGetSmgr(wstate->index), MAIN_FORKNUM, blkno,
+ smgrextend(RelationGetSmgr(wstate->index, MAIN_FORKNUM), blkno,
(char *) page, true);
wstate->btws_pages_written++;
}
else
{
/* overwriting a block we zero-filled before */
- smgrwrite(RelationGetSmgr(wstate->index), MAIN_FORKNUM, blkno,
+ smgrwrite(RelationGetSmgr(wstate->index, MAIN_FORKNUM), blkno,
(char *) page, true);
}
@@ -1431,7 +1431,7 @@ _bt_load(BTWriteState *wstate, BTSpool *btspool, BTSpool *btspool2)
* still not be on disk when the crash occurs.
*/
if (wstate->btws_use_wal)
- smgrimmedsync(RelationGetSmgr(wstate->index), MAIN_FORKNUM);
+ smgrimmedsync(RelationGetSmgr(wstate->index, MAIN_FORKNUM));
}
/*
diff --git a/src/backend/access/spgist/spginsert.c b/src/backend/access/spgist/spginsert.c
index c6821b59524..75eb09543d9 100644
--- a/src/backend/access/spgist/spginsert.c
+++ b/src/backend/access/spgist/spginsert.c
@@ -155,8 +155,11 @@ spgbuild(Relation heap, Relation index, IndexInfo *indexInfo)
void
spgbuildempty(Relation index)
{
+ SMgrFileHandle sfile;
Page page;
+ sfile = RelationGetSmgr(index, INIT_FORKNUM);
+
/* Construct metapage. */
page = (Page) palloc(BLCKSZ);
SpGistInitMetapage(page);
@@ -169,27 +172,27 @@ spgbuildempty(Relation index)
* replayed.
*/
PageSetChecksumInplace(page, SPGIST_METAPAGE_BLKNO);
- smgrwrite(RelationGetSmgr(index), INIT_FORKNUM, SPGIST_METAPAGE_BLKNO,
+ smgrwrite(sfile, SPGIST_METAPAGE_BLKNO,
(char *) page, true);
- log_newpage(&(RelationGetSmgr(index))->smgr_rlocator.locator, INIT_FORKNUM,
+ log_newpage(&index->rd_locator, INIT_FORKNUM,
SPGIST_METAPAGE_BLKNO, page, true);
/* Likewise for the root page. */
SpGistInitPage(page, SPGIST_LEAF);
PageSetChecksumInplace(page, SPGIST_ROOT_BLKNO);
- smgrwrite(RelationGetSmgr(index), INIT_FORKNUM, SPGIST_ROOT_BLKNO,
+ smgrwrite(sfile, SPGIST_ROOT_BLKNO,
(char *) page, true);
- log_newpage(&(RelationGetSmgr(index))->smgr_rlocator.locator, INIT_FORKNUM,
+ log_newpage(&index->rd_locator, INIT_FORKNUM,
SPGIST_ROOT_BLKNO, page, true);
/* Likewise for the null-tuples root page. */
SpGistInitPage(page, SPGIST_LEAF | SPGIST_NULLS);
PageSetChecksumInplace(page, SPGIST_NULL_BLKNO);
- smgrwrite(RelationGetSmgr(index), INIT_FORKNUM, SPGIST_NULL_BLKNO,
+ smgrwrite(sfile, SPGIST_NULL_BLKNO,
(char *) page, true);
- log_newpage(&(RelationGetSmgr(index))->smgr_rlocator.locator, INIT_FORKNUM,
+ log_newpage(&index->rd_locator, INIT_FORKNUM,
SPGIST_NULL_BLKNO, page, true);
/*
@@ -197,7 +200,7 @@ spgbuildempty(Relation index)
* writes did not go through shared buffers and therefore a concurrent
* checkpoint may have moved the redo pointer past our xlog record.
*/
- smgrimmedsync(RelationGetSmgr(index), INIT_FORKNUM);
+ smgrimmedsync(sfile);
}
/*
diff --git a/src/backend/access/table/tableam.c b/src/backend/access/table/tableam.c
index b3d1a6c3f8f..18c4e87ff52 100644
--- a/src/backend/access/table/tableam.c
+++ b/src/backend/access/table/tableam.c
@@ -632,10 +632,10 @@ table_block_relation_size(Relation rel, ForkNumber forkNumber)
if (forkNumber == InvalidForkNumber)
{
for (int i = 0; i < MAX_FORKNUM; i++)
- nblocks += smgrnblocks(RelationGetSmgr(rel), i);
+ nblocks += smgrnblocks(RelationGetSmgr(rel, i));
}
else
- nblocks = smgrnblocks(RelationGetSmgr(rel), forkNumber);
+ nblocks = smgrnblocks(RelationGetSmgr(rel, forkNumber));
return nblocks * BLCKSZ;
}
diff --git a/src/backend/access/transam/xlogprefetcher.c b/src/backend/access/transam/xlogprefetcher.c
index 87d14210be7..bc2ecec2205 100644
--- a/src/backend/access/transam/xlogprefetcher.c
+++ b/src/backend/access/transam/xlogprefetcher.c
@@ -650,7 +650,7 @@ XLogPrefetcherNextBlock(uintptr_t pgsr_private, XLogRecPtr *lsn)
{
int block_id = prefetcher->next_block_id++;
DecodedBkpBlock *block = &record->blocks[block_id];
- SMgrRelation reln;
+ SMgrFileHandle sfile;
PrefetchBufferResult result;
if (!block->in_use)
@@ -720,7 +720,7 @@ XLogPrefetcherNextBlock(uintptr_t pgsr_private, XLogRecPtr *lsn)
* same relation (with some scheme to handle invalidations
* safely), but for now we'll call smgropen() every time.
*/
- reln = smgropen(block->rlocator, InvalidBackendId);
+ sfile = smgropen(block->rlocator, InvalidBackendId, block->forknum);
/*
* If the relation file doesn't exist on disk, for example because
@@ -729,14 +729,14 @@ XLogPrefetcherNextBlock(uintptr_t pgsr_private, XLogRecPtr *lsn)
* further prefetching in the relation until this record is
* replayed.
*/
- if (!smgrexists(reln, MAIN_FORKNUM))
+ if (!smgrexists(sfile))
{
#ifdef XLOGPREFETCHER_DEBUG_LEVEL
elog(XLOGPREFETCHER_DEBUG_LEVEL,
"suppressing all prefetch in relation %u/%u/%u until %X/%X is replayed, because the relation does not exist on disk",
- reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
- reln->smgr_rlocator.locator.relNumber,
+ sfile->smgr_rlocator.locator.spcOid,
+ sfile->smgr_rlocator.locator.dbOid,
+ sfile->smgr_rlocator.locator.relNumber,
LSN_FORMAT_ARGS(record->lsn));
#endif
XLogPrefetcherAddFilter(prefetcher, block->rlocator, 0,
@@ -750,14 +750,14 @@ XLogPrefetcherNextBlock(uintptr_t pgsr_private, XLogRecPtr *lsn)
* block yet, suppress prefetching of this block and higher until
* this record is replayed.
*/
- if (block->blkno >= smgrnblocks(reln, block->forknum))
+ if (block->blkno >= smgrnblocks(sfile))
{
#ifdef XLOGPREFETCHER_DEBUG_LEVEL
elog(XLOGPREFETCHER_DEBUG_LEVEL,
"suppressing prefetch in relation %u/%u/%u from block %u until %X/%X is replayed, because the relation is too small",
- reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
- reln->smgr_rlocator.locator.relNumber,
+ sfile->smgr_rlocator.locator.spcOid,
+ sfile->smgr_rlocator.locator.dbOid,
+ sfile->smgr_rlocator.locator.relNumber,
block->blkno,
LSN_FORMAT_ARGS(record->lsn));
#endif
@@ -768,7 +768,7 @@ XLogPrefetcherNextBlock(uintptr_t pgsr_private, XLogRecPtr *lsn)
}
/* Try to initiate prefetching. */
- result = PrefetchSharedBuffer(reln, block->forknum, block->blkno);
+ result = PrefetchSharedBuffer(sfile, block->blkno);
if (BufferIsValid(result.recent_buffer))
{
/* Cache hit, nothing to do. */
@@ -794,9 +794,9 @@ XLogPrefetcherNextBlock(uintptr_t pgsr_private, XLogRecPtr *lsn)
*/
elog(ERROR,
"could not prefetch relation %u/%u/%u block %u",
- reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
- reln->smgr_rlocator.locator.relNumber,
+ sfile->smgr_locator.locator.spcOid,
+ sfile->smgr_locator.locator.dbOid,
+ sfile->smgr_locator.locator.relNumber,
block->blkno);
}
}
diff --git a/src/backend/access/transam/xlogutils.c b/src/backend/access/transam/xlogutils.c
index 0cda22597fe..7901fc5ffc8 100644
--- a/src/backend/access/transam/xlogutils.c
+++ b/src/backend/access/transam/xlogutils.c
@@ -475,7 +475,7 @@ XLogReadBufferExtended(RelFileLocator rlocator, ForkNumber forknum,
{
BlockNumber lastblock;
Buffer buffer;
- SMgrRelation smgr;
+ SMgrFileHandle sfile;
Assert(blkno != P_NEW);
@@ -489,7 +489,7 @@ XLogReadBufferExtended(RelFileLocator rlocator, ForkNumber forknum,
}
/* Open the relation at smgr level */
- smgr = smgropen(rlocator, InvalidBackendId);
+ sfile = smgropen(rlocator, InvalidBackendId, forknum);
/*
* Create the target file if it doesn't already exist. This lets us cope
@@ -499,9 +499,9 @@ XLogReadBufferExtended(RelFileLocator rlocator, ForkNumber forknum,
* filesystem loses an inode during a crash. Better to write the data
* until we are actually told to delete the file.)
*/
- smgrcreate(smgr, forknum, true);
+ smgrcreate(sfile, true);
- lastblock = smgrnblocks(smgr, forknum);
+ lastblock = smgrnblocks(sfile);
if (blkno < lastblock)
{
@@ -629,7 +629,7 @@ CreateFakeRelcacheEntry(RelFileLocator rlocator)
rel->rd_lockInfo.lockRelId.dbId = rlocator.dbOid;
rel->rd_lockInfo.lockRelId.relId = rlocator.relNumber;
- rel->rd_smgr = NULL;
+ MemSet(rel->rd_smgr, 0, sizeof(rel->rd_smgr));
return rel;
}
@@ -641,8 +641,11 @@ void
FreeFakeRelcacheEntry(Relation fakerel)
{
/* make sure the fakerel is not referenced by the SmgrRelation anymore */
- if (fakerel->rd_smgr != NULL)
- smgrclearowner(&fakerel->rd_smgr, fakerel->rd_smgr);
+ for (int i = 0; i <= MAX_FORKNUM; i++)
+ {
+ if (fakerel->rd_smgr[i] != NULL)
+ smgrclearowner(&fakerel->rd_smgr[i], fakerel->rd_smgr[i]);
+ }
pfree(fakerel);
}
diff --git a/src/backend/catalog/catalog.c b/src/backend/catalog/catalog.c
index 6f43870779f..03d163f7882 100644
--- a/src/backend/catalog/catalog.c
+++ b/src/backend/catalog/catalog.c
@@ -499,10 +499,10 @@ GetNewOidWithIndex(Relation relation, Oid indexId, AttrNumber oidcolumn)
RelFileNumber
GetNewRelFileNumber(Oid reltablespace, Relation pg_class, char relpersistence)
{
- RelFileLocatorBackend rlocator;
- char *rpath;
- bool collides;
+ RelFileLocator rlocator;
BackendId backend;
+ SMgrFileHandle sfile;
+ bool collides;
/*
* If we ever get here during pg_upgrade, there's something wrong; all
@@ -511,6 +511,11 @@ GetNewRelFileNumber(Oid reltablespace, Relation pg_class, char relpersistence)
*/
Assert(!IsBinaryUpgrade);
+ /*
+ * The relpath will vary based on the backend ID, so we must initialize
+ * that properly here to make sure that any collisions based on filename
+ * are properly detected.
+ */
switch (relpersistence)
{
case RELPERSISTENCE_TEMP:
@@ -526,53 +531,29 @@ GetNewRelFileNumber(Oid reltablespace, Relation pg_class, char relpersistence)
}
/* This logic should match RelationInitPhysicalAddr */
- rlocator.locator.spcOid = reltablespace ? reltablespace : MyDatabaseTableSpace;
- rlocator.locator.dbOid =
- (rlocator.locator.spcOid == GLOBALTABLESPACE_OID) ?
+ rlocator.spcOid = reltablespace ? reltablespace : MyDatabaseTableSpace;
+ rlocator.dbOid =
+ (rlocator.spcOid == GLOBALTABLESPACE_OID) ?
InvalidOid : MyDatabaseId;
- /*
- * The relpath will vary based on the backend ID, so we must initialize
- * that properly here to make sure that any collisions based on filename
- * are properly detected.
- */
- rlocator.backend = backend;
-
do
{
CHECK_FOR_INTERRUPTS();
/* Generate the OID */
if (pg_class)
- rlocator.locator.relNumber = GetNewOidWithIndex(pg_class, ClassOidIndexId,
+ rlocator.relNumber = GetNewOidWithIndex(pg_class, ClassOidIndexId,
Anum_pg_class_oid);
else
- rlocator.locator.relNumber = GetNewObjectId();
+ rlocator.relNumber = GetNewObjectId();
/* Check for existing file of same name */
- rpath = relpath(rlocator, MAIN_FORKNUM);
-
- if (access(rpath, F_OK) == 0)
- {
- /* definite collision */
- collides = true;
- }
- else
- {
- /*
- * Here we have a little bit of a dilemma: if errno is something
- * other than ENOENT, should we declare a collision and loop? In
- * practice it seems best to go ahead regardless of the errno. If
- * there is a colliding file we will get an smgr failure when we
- * attempt to create the new relation file.
- */
- collides = false;
- }
-
- pfree(rpath);
+ sfile = smgropen(rlocator, backend, MAIN_FORKNUM);
+ collides = smgrexists(sfile);
+ smgrclose(sfile);
} while (collides);
- return rlocator.locator.relNumber;
+ return rlocator.relNumber;
}
/*
diff --git a/src/backend/catalog/index.c b/src/backend/catalog/index.c
index d7192f35e3f..07cf7b66787 100644
--- a/src/backend/catalog/index.c
+++ b/src/backend/catalog/index.c
@@ -3026,9 +3026,9 @@ index_build(Relation heapRelation,
* relfilenumber won't change, and nothing needs to be done here.
*/
if (indexRelation->rd_rel->relpersistence == RELPERSISTENCE_UNLOGGED &&
- !smgrexists(RelationGetSmgr(indexRelation), INIT_FORKNUM))
+ !smgrexists(RelationGetSmgr(indexRelation, INIT_FORKNUM)))
{
- smgrcreate(RelationGetSmgr(indexRelation), INIT_FORKNUM, false);
+ smgrcreate(RelationGetSmgr(indexRelation, INIT_FORKNUM), false);
indexRelation->rd_indam->ambuildempty(indexRelation);
}
diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c
index d708af19ed2..5d47864e3a9 100644
--- a/src/backend/catalog/storage.c
+++ b/src/backend/catalog/storage.c
@@ -116,11 +116,11 @@ AddPendingSync(const RelFileLocator *rlocator)
* that does not want the storage to be destroyed in case of an abort may
* pass register_delete = false.
*/
-SMgrRelation
+SMgrFileHandle
RelationCreateStorage(RelFileLocator rlocator, char relpersistence,
bool register_delete)
{
- SMgrRelation srel;
+ SMgrFileHandle sfile;
BackendId backend;
bool needs_wal;
@@ -145,11 +145,11 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence,
return NULL; /* placate compiler */
}
- srel = smgropen(rlocator, backend);
- smgrcreate(srel, MAIN_FORKNUM, false);
+ sfile = smgropen(rlocator, backend, MAIN_FORKNUM);
+ smgrcreate(sfile, false);
if (needs_wal)
- log_smgrcreate(&srel->smgr_rlocator.locator, MAIN_FORKNUM);
+ log_smgrcreate(&rlocator, MAIN_FORKNUM);
/*
* Add the relation to the list of stuff to delete at abort, if we are
@@ -175,7 +175,7 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence,
AddPendingSync(&rlocator);
}
- return srel;
+ return sfile;
}
/*
@@ -292,16 +292,18 @@ RelationTruncate(Relation rel, BlockNumber nblocks)
ForkNumber forks[MAX_FORKNUM];
BlockNumber blocks[MAX_FORKNUM];
int nforks = 0;
- SMgrRelation reln;
/*
* Make sure smgr_targblock etc aren't pointing somewhere past new end.
* (Note: don't rely on this reln pointer below this loop.)
*/
- reln = RelationGetSmgr(rel);
- reln->smgr_targblock = InvalidBlockNumber;
- for (int i = 0; i <= MAX_FORKNUM; ++i)
- reln->smgr_cached_nblocks[i] = InvalidBlockNumber;
+ for (int i = 0; i <= MAX_FORKNUM; i++)
+ {
+ SMgrFileHandle sfile = RelationGetSmgr(rel, i);
+
+ sfile->smgr_targblock = InvalidBlockNumber;
+ sfile->smgr_cached_nblocks = InvalidBlockNumber;
+ }
/* Prepare for truncation of MAIN fork of the relation */
forks[nforks] = MAIN_FORKNUM;
@@ -309,7 +311,7 @@ RelationTruncate(Relation rel, BlockNumber nblocks)
nforks++;
/* Prepare for truncation of the FSM if it exists */
- fsm = smgrexists(RelationGetSmgr(rel), FSM_FORKNUM);
+ fsm = smgrexists(RelationGetSmgr(rel, FSM_FORKNUM));
if (fsm)
{
blocks[nforks] = FreeSpaceMapPrepareTruncateRel(rel, nblocks);
@@ -322,7 +324,7 @@ RelationTruncate(Relation rel, BlockNumber nblocks)
}
/* Prepare for truncation of the visibility map too if it exists */
- vm = smgrexists(RelationGetSmgr(rel), VISIBILITYMAP_FORKNUM);
+ vm = smgrexists(RelationGetSmgr(rel, VISIBILITYMAP_FORKNUM));
if (vm)
{
blocks[nforks] = visibilitymap_prepare_truncate(rel, nblocks);
@@ -390,11 +392,12 @@ RelationTruncate(Relation rel, BlockNumber nblocks)
}
/*
- * This will first remove any buffers from the buffer pool that should no
+ * First remove any buffers from the buffer pool that should no
* longer exist after truncation is complete, and then truncate the
* corresponding files on disk.
*/
- smgrtruncate(RelationGetSmgr(rel), forks, nforks, blocks);
+ DropRelationBuffers(rel->rd_locator, rel->rd_backend, forks, nforks, blocks);
+ smgrtruncate_multi(rel->rd_locator, rel->rd_backend, forks, nforks, blocks);
/* We've done all the critical work, so checkpoints are OK now. */
MyProc->delayChkptFlags &= ~DELAY_CHKPT_COMPLETE;
@@ -428,7 +431,7 @@ RelationPreTruncate(Relation rel)
return;
pending = hash_search(pendingSyncHash,
- &(RelationGetSmgr(rel)->smgr_rlocator.locator),
+ &rel->rd_locator,
HASH_FIND, NULL);
if (pending)
pending->is_truncated = true;
@@ -444,12 +447,12 @@ RelationPreTruncate(Relation rel)
* Also note that this is frequently called via locutions such as
* RelationCopyStorage(RelationGetSmgr(rel), ...);
* That's safe only because we perform only smgr and WAL operations here.
- * If we invoked anything else, a relcache flush could cause our SMgrRelation
+ * If we invoked anything else, a relcache flush could cause our SMgrFileHandle
* argument to become a dangling pointer.
*/
void
-RelationCopyStorage(SMgrRelation src, SMgrRelation dst,
- ForkNumber forkNum, char relpersistence)
+RelationCopyStorage(SMgrFileHandle src, SMgrFileHandle dst,
+ char relpersistence)
{
PGAlignedBlock buf;
Page page;
@@ -466,7 +469,7 @@ RelationCopyStorage(SMgrRelation src, SMgrRelation dst,
* it needs to be synced to disk.
*/
copying_initfork = relpersistence == RELPERSISTENCE_UNLOGGED &&
- forkNum == INIT_FORKNUM;
+ src->smgr_locator.forknum == INIT_FORKNUM;
/*
* We need to log the copied data in WAL iff WAL archiving/streaming is
@@ -477,14 +480,14 @@ RelationCopyStorage(SMgrRelation src, SMgrRelation dst,
use_wal = XLogIsNeeded() &&
(relpersistence == RELPERSISTENCE_PERMANENT || copying_initfork);
- nblocks = smgrnblocks(src, forkNum);
+ nblocks = smgrnblocks(src);
for (blkno = 0; blkno < nblocks; blkno++)
{
/* If we got a cancel signal during the copy of the data, quit */
CHECK_FOR_INTERRUPTS();
- smgrread(src, forkNum, blkno, buf.data);
+ smgrread(src, blkno, buf.data);
if (!PageIsVerifiedExtended(page, blkno,
PIV_LOG_WARNING | PIV_REPORT_STAT))
@@ -496,9 +499,7 @@ RelationCopyStorage(SMgrRelation src, SMgrRelation dst,
* (errcontext callbacks shouldn't be risking any such thing, but
* people have been known to forget that rule.)
*/
- char *relpath = relpathbackend(src->smgr_rlocator.locator,
- src->smgr_rlocator.backend,
- forkNum);
+ char *relpath = smgrfilepath(src->smgr_locator);
ereport(ERROR,
(errcode(ERRCODE_DATA_CORRUPTED),
@@ -512,7 +513,7 @@ RelationCopyStorage(SMgrRelation src, SMgrRelation dst,
* space.
*/
if (use_wal)
- log_newpage(&dst->smgr_rlocator.locator, forkNum, blkno, page, false);
+ log_newpage(&dst->smgr_locator.locator, dst->smgr_locator.forknum, blkno, page, false);
PageSetChecksumInplace(page, blkno);
@@ -521,7 +522,7 @@ RelationCopyStorage(SMgrRelation src, SMgrRelation dst,
* need for smgr to schedule an fsync for this write; we'll do it
* ourselves below.
*/
- smgrextend(dst, forkNum, blkno, buf.data, true);
+ smgrextend(dst, blkno, buf.data, true);
}
/*
@@ -534,7 +535,7 @@ RelationCopyStorage(SMgrRelation src, SMgrRelation dst,
* they might still not be on disk when the crash occurs.
*/
if (use_wal || copying_initfork)
- smgrimmedsync(dst, forkNum);
+ smgrimmedsync(dst);
}
/*
@@ -653,9 +654,9 @@ smgrDoPendingDeletes(bool isCommit)
PendingRelDelete *pending;
PendingRelDelete *prev;
PendingRelDelete *next;
- int nrels = 0,
- maxrels = 0;
- SMgrRelation *srels = NULL;
+ int nlocators = 0,
+ maxlocators = 0;
+ RelFileLocatorBackend *locators = NULL;
prev = NULL;
for (pending = pendingDeletes; pending != NULL; pending = next)
@@ -676,23 +677,21 @@ smgrDoPendingDeletes(bool isCommit)
/* do deletion if called for */
if (pending->atCommit == isCommit)
{
- SMgrRelation srel;
-
- srel = smgropen(pending->rlocator, pending->backend);
+ RelFileLocatorBackend rlocator = { pending->rlocator, pending->backend };
/* allocate the initial array, or extend it, if needed */
- if (maxrels == 0)
+ if (maxlocators == 0)
{
- maxrels = 8;
- srels = palloc(sizeof(SMgrRelation) * maxrels);
+ maxlocators = 8;
+ locators = palloc(sizeof(RelFileLocatorBackend) * maxlocators);
}
- else if (maxrels <= nrels)
+ else if (maxlocators <= nlocators)
{
- maxrels *= 2;
- srels = repalloc(srels, sizeof(SMgrRelation) * maxrels);
+ maxlocators *= 2;
+ locators = repalloc(locators, sizeof(RelFileLocatorBackend) * maxlocators);
}
- srels[nrels++] = srel;
+ locators[nlocators++] = rlocator;
}
/* must explicitly free the list entry */
pfree(pending);
@@ -700,15 +699,58 @@ smgrDoPendingDeletes(bool isCommit)
}
}
- if (nrels > 0)
+ if (nlocators > 0)
{
- smgrdounlinkall(srels, nrels, false);
+ ForkNumber forks[MAX_FORKNUM + 1];
+
+ for (int i = 0; i <= MAX_FORKNUM; i++)
+ forks[i] = i;
+
+ /*
+ * Get rid of any remaining buffers for the relations. bufmgr will just
+ * drop them without bothering to write the contents.
+ */
+ DropRelationsAllBuffers(locators, nlocators);
- for (int i = 0; i < nrels; i++)
- smgrclose(srels[i]);
+ for (int i = 0; i < nlocators; i++)
+ smgrunlink_multi(locators[i].locator, locators[i].backend, forks, MAX_FORKNUM + 1, false);
+ pfree(locators);
+ }
+}
+
+/*
+ * DropRelationFiles -- drop files of all given relations
+ */
+void
+DropRelationFiles(RelFileLocator *delrels, int ndelrels, bool isRedo)
+{
+ RelFileLocatorBackend *locators;
+ int i;
+ ForkNumber all_forks[MAX_FORKNUM + 1];
- pfree(srels);
+ locators = palloc(sizeof(RelFileLocatorBackend) * ndelrels);
+ for (i = 0; i < ndelrels; i++)
+ {
+ if (isRedo)
+ {
+ for (int fork = 0; fork <= MAX_FORKNUM; fork++)
+ XLogDropRelation(delrels[i], fork);
+ }
+ locators[i].locator = delrels[i];
+ locators[i].backend = InvalidBackendId;
}
+
+ /*
+ * Get rid of any remaining buffers for the relations. bufmgr will just
+ * drop them without bothering to write the contents.
+ */
+ DropRelationsAllBuffers(locators, ndelrels);
+
+ for (int fork = 0; fork <= MAX_FORKNUM; fork++)
+ all_forks[fork] = fork;
+
+ for (i = 0; i < ndelrels; i++)
+ smgrunlink_multi(locators[i].locator, locators[i].backend, all_forks, MAX_FORKNUM + 1, true);
}
/*
@@ -718,9 +760,9 @@ void
smgrDoPendingSyncs(bool isCommit, bool isParallelWorker)
{
PendingRelDelete *pending;
- int nrels = 0,
- maxrels = 0;
- SMgrRelation *srels = NULL;
+ int nlocators = 0,
+ maxlocators = 0;
+ RelFileLocator *locators = NULL;
HASH_SEQ_STATUS scan;
PendingRelSync *pendingsync;
@@ -757,9 +799,6 @@ smgrDoPendingSyncs(bool isCommit, bool isParallelWorker)
ForkNumber fork;
BlockNumber nblocks[MAX_FORKNUM + 1];
BlockNumber total_blocks = 0;
- SMgrRelation srel;
-
- srel = smgropen(pendingsync->rlocator, InvalidBackendId);
/*
* We emit newpage WAL records for smaller relations.
@@ -773,9 +812,12 @@ smgrDoPendingSyncs(bool isCommit, bool isParallelWorker)
{
for (fork = 0; fork <= MAX_FORKNUM; fork++)
{
- if (smgrexists(srel, fork))
+ SMgrFileHandle sfile;
+
+ sfile = smgropen(pendingsync->rlocator, InvalidBackendId, fork);
+ if (smgrexists(sfile))
{
- BlockNumber n = smgrnblocks(srel, fork);
+ BlockNumber n = smgrnblocks(sfile);
/* we shouldn't come here for unlogged relations */
Assert(fork != INIT_FORKNUM);
@@ -803,18 +845,19 @@ smgrDoPendingSyncs(bool isCommit, bool isParallelWorker)
total_blocks * BLCKSZ / 1024 >= wal_skip_threshold)
{
/* allocate the initial array, or extend it, if needed */
- if (maxrels == 0)
+ if (maxlocators == 0)
{
- maxrels = 8;
- srels = palloc(sizeof(SMgrRelation) * maxrels);
+ maxlocators = 8;
+ locators = palloc(sizeof(RelFileLocatorBackend) * maxlocators);
}
- else if (maxrels <= nrels)
+ else if (maxlocators <= nlocators)
{
- maxrels *= 2;
- srels = repalloc(srels, sizeof(SMgrRelation) * maxrels);
+ maxlocators *= 2;
+ locators = repalloc(locators, sizeof(RelFileLocatorBackend) * maxlocators);
}
- srels[nrels++] = srel;
+ locators[nlocators] = pendingsync->rlocator;
+ nlocators++;
}
else
{
@@ -833,7 +876,7 @@ smgrDoPendingSyncs(bool isCommit, bool isParallelWorker)
* page including any unused space. ReadBufferExtended()
* counts some pgstat events; unfortunately, we discard them.
*/
- rel = CreateFakeRelcacheEntry(srel->smgr_rlocator.locator);
+ rel = CreateFakeRelcacheEntry(pendingsync->rlocator);
log_newpage_range(rel, fork, 0, n, false);
FreeFakeRelcacheEntry(rel);
}
@@ -842,11 +885,20 @@ smgrDoPendingSyncs(bool isCommit, bool isParallelWorker)
pendingSyncHash = NULL;
- if (nrels > 0)
+ FlushRelationsAllBuffers(locators, nlocators);
+
+ for (int i = 0; i < nlocators; i++)
{
- smgrdosyncall(srels, nrels);
- pfree(srels);
+ for (int fork = 0; fork <= MAX_FORKNUM; fork++)
+ {
+ SMgrFileHandle sfile = smgropen(locators[i], InvalidBackendId, fork);
+
+ if (smgrexists(sfile))
+ smgrimmedsync(sfile);
+ }
}
+ if (locators != NULL)
+ pfree(locators);
}
/*
@@ -966,22 +1018,22 @@ smgr_redo(XLogReaderState *record)
if (info == XLOG_SMGR_CREATE)
{
xl_smgr_create *xlrec = (xl_smgr_create *) XLogRecGetData(record);
- SMgrRelation reln;
+ SMgrFileHandle sfile;
- reln = smgropen(xlrec->rlocator, InvalidBackendId);
- smgrcreate(reln, xlrec->forkNum, true);
+ sfile = smgropen(xlrec->rlocator, InvalidBackendId, xlrec->forkNum);
+ smgrcreate(sfile, true);
}
else if (info == XLOG_SMGR_TRUNCATE)
{
xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record);
- SMgrRelation reln;
+ SMgrFileHandle sfile;
Relation rel;
ForkNumber forks[MAX_FORKNUM];
BlockNumber blocks[MAX_FORKNUM];
int nforks = 0;
bool need_fsm_vacuum = false;
- reln = smgropen(xlrec->rlocator, InvalidBackendId);
+ sfile = smgropen(xlrec->rlocator, InvalidBackendId, MAIN_FORKNUM);
/*
* Forcibly create relation if it doesn't exist (which suggests that
@@ -989,7 +1041,7 @@ smgr_redo(XLogReaderState *record)
* XLogReadBufferForRedo, we prefer to recreate the rel and replay the
* log as best we can until the drop is seen.
*/
- smgrcreate(reln, MAIN_FORKNUM, true);
+ smgrcreate(sfile, true);
/*
* Before we perform the truncation, update minimum recovery point to
@@ -1022,8 +1074,10 @@ smgr_redo(XLogReaderState *record)
/* Prepare for truncation of FSM and VM too */
rel = CreateFakeRelcacheEntry(xlrec->rlocator);
+ DropRelationBuffers(xlrec->rlocator, InvalidBackendId, forks, nforks, blocks);
+
if ((xlrec->flags & SMGR_TRUNCATE_FSM) != 0 &&
- smgrexists(reln, FSM_FORKNUM))
+ smgrexists(smgropen(xlrec->rlocator, InvalidBackendId, FSM_FORKNUM)))
{
blocks[nforks] = FreeSpaceMapPrepareTruncateRel(rel, xlrec->blkno);
if (BlockNumberIsValid(blocks[nforks]))
@@ -1034,7 +1088,7 @@ smgr_redo(XLogReaderState *record)
}
}
if ((xlrec->flags & SMGR_TRUNCATE_VM) != 0 &&
- smgrexists(reln, VISIBILITYMAP_FORKNUM))
+ smgrexists(smgropen(xlrec->rlocator, InvalidBackendId, VISIBILITYMAP_FORKNUM)))
{
blocks[nforks] = visibilitymap_prepare_truncate(rel, xlrec->blkno);
if (BlockNumberIsValid(blocks[nforks]))
@@ -1046,7 +1100,10 @@ smgr_redo(XLogReaderState *record)
/* Do the real work to truncate relation forks */
if (nforks > 0)
- smgrtruncate(reln, forks, nforks, blocks);
+ {
+ DropRelationBuffers(xlrec->rlocator, InvalidBackendId, forks, nforks, blocks);
+ smgrtruncate_multi(xlrec->rlocator, InvalidBackendId, forks, nforks, blocks);
+ }
/*
* Update upper-level FSM pages to account for the truncation. This is
diff --git a/src/backend/commands/dbcommands.c b/src/backend/commands/dbcommands.c
index 099d369b2f4..80d55c1375f 100644
--- a/src/backend/commands/dbcommands.c
+++ b/src/backend/commands/dbcommands.c
@@ -280,7 +280,7 @@ ScanSourceDatabasePgClass(Oid tbid, Oid dbid, char *srcpath)
* invalidations.
*/
rel = CreateFakeRelcacheEntry(rlocator);
- nblocks = smgrnblocks(RelationGetSmgr(rel), MAIN_FORKNUM);
+ nblocks = smgrnblocks(RelationGetSmgr(rel, MAIN_FORKNUM));
FreeFakeRelcacheEntry(rel);
/* Use a buffer access strategy since this is a bulk read operation. */
diff --git a/src/backend/commands/sequence.c b/src/backend/commands/sequence.c
index b0b211891c3..0888b5107b3 100644
--- a/src/backend/commands/sequence.c
+++ b/src/backend/commands/sequence.c
@@ -345,14 +345,14 @@ fill_seq_with_data(Relation rel, HeapTuple tuple)
if (rel->rd_rel->relpersistence == RELPERSISTENCE_UNLOGGED)
{
- SMgrRelation srel;
+ SMgrFileHandle sfile;
- srel = smgropen(rel->rd_locator, InvalidBackendId);
- smgrcreate(srel, INIT_FORKNUM, false);
+ sfile = smgropen(rel->rd_locator, InvalidBackendId, INIT_FORKNUM);
+ smgrcreate(sfile, false);
log_smgrcreate(&rel->rd_locator, INIT_FORKNUM);
fill_seq_fork_with_data(rel, tuple, INIT_FORKNUM);
FlushRelationBuffers(rel);
- smgrclose(srel);
+ smgrclose(sfile);
}
}
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 7fbee0c1f71..512876aa953 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -14593,9 +14593,7 @@ AlterTableMoveAll(AlterTableMoveAllStmt *stmt)
static void
index_copy_data(Relation rel, RelFileLocator newrlocator)
{
- SMgrRelation dstrel;
-
- dstrel = smgropen(newrlocator, rel->rd_backend);
+ SMgrFileHandle dstmain;
/*
* Since we copy the file directly without looking at the shared buffers,
@@ -14615,16 +14613,20 @@ index_copy_data(Relation rel, RelFileLocator newrlocator)
RelationCreateStorage(newrlocator, rel->rd_rel->relpersistence, true);
/* copy main fork */
- RelationCopyStorage(RelationGetSmgr(rel), dstrel, MAIN_FORKNUM,
+ dstmain = smgropen(newrlocator, rel->rd_backend, MAIN_FORKNUM);
+ RelationCopyStorage(RelationGetSmgr(rel, MAIN_FORKNUM), dstmain,
rel->rd_rel->relpersistence);
/* copy those extra forks that exist */
for (ForkNumber forkNum = MAIN_FORKNUM + 1;
forkNum <= MAX_FORKNUM; forkNum++)
{
- if (smgrexists(RelationGetSmgr(rel), forkNum))
+ if (smgrexists(RelationGetSmgr(rel, forkNum)))
{
- smgrcreate(dstrel, forkNum, false);
+ SMgrFileHandle src_fork = RelationGetSmgr(rel, forkNum);
+ SMgrFileHandle dst_fork = smgropen(newrlocator, rel->rd_backend, forkNum);
+
+ smgrcreate(dst_fork, false);
/*
* WAL log creation if the relation is persistent, or this is the
@@ -14634,14 +14636,15 @@ index_copy_data(Relation rel, RelFileLocator newrlocator)
(rel->rd_rel->relpersistence == RELPERSISTENCE_UNLOGGED &&
forkNum == INIT_FORKNUM))
log_smgrcreate(&newrlocator, forkNum);
- RelationCopyStorage(RelationGetSmgr(rel), dstrel, forkNum,
+ RelationCopyStorage(src_fork, dst_fork,
rel->rd_rel->relpersistence);
+ smgrclose(dst_fork);
}
}
/* drop old relation, and close new one */
RelationDropStorage(rel);
- smgrclose(dstrel);
+ smgrclose(dstmain);
}
/*
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index b7488b5d89e..2ba75f676ad 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -117,19 +117,6 @@ typedef struct CkptTsStatus
int index;
} CkptTsStatus;
-/*
- * Type for array used to sort SMgrRelations
- *
- * FlushRelationsAllBuffers shares the same comparator function with
- * DropRelationsAllBuffers. Pointer to this struct and RelFileLocator must be
- * compatible.
- */
-typedef struct SMgrSortArray
-{
- RelFileLocator rlocator; /* This must be the first member */
- SMgrRelation srel;
-} SMgrSortArray;
-
/* GUC variables */
bool zero_damaged_pages = false;
int bgwriter_lru_maxpages = 100;
@@ -459,8 +446,8 @@ ForgetPrivateRefCountEntry(PrivateRefCountEntry *ref)
)
-static Buffer ReadBuffer_common(SMgrRelation reln, char relpersistence,
- ForkNumber forkNum, BlockNumber blockNum,
+static Buffer ReadBuffer_common(SMgrFileHandle sfile, char relpersistence,
+ BlockNumber blockNum,
ReadBufferMode mode, BufferAccessStrategy strategy,
bool *hit);
static bool PinBuffer(BufferDesc *buf, BufferAccessStrategy strategy);
@@ -476,13 +463,12 @@ static void TerminateBufferIO(BufferDesc *buf, bool clear_dirty,
uint32 set_flag_bits);
static void shared_buffer_write_error_callback(void *arg);
static void local_buffer_write_error_callback(void *arg);
-static BufferDesc *BufferAlloc(SMgrRelation smgr,
+static BufferDesc *BufferAlloc(SMgrFileHandle smgr,
char relpersistence,
- ForkNumber forkNum,
BlockNumber blockNum,
BufferAccessStrategy strategy,
bool *foundPtr);
-static void FlushBuffer(BufferDesc *buf, SMgrRelation reln);
+static void FlushBuffer(BufferDesc *buf, SMgrFileHandle sfile);
static void FindAndDropRelationBuffers(RelFileLocator rlocator,
ForkNumber forkNum,
BlockNumber nForkBlock,
@@ -502,9 +488,7 @@ static int ts_ckpt_progress_comparator(Datum a, Datum b, void *arg);
* Implementation of PrefetchBuffer() for shared buffers.
*/
PrefetchBufferResult
-PrefetchSharedBuffer(SMgrRelation smgr_reln,
- ForkNumber forkNum,
- BlockNumber blockNum)
+PrefetchSharedBuffer(SMgrFileHandle sfile, BlockNumber blockNum)
{
PrefetchBufferResult result = {InvalidBuffer, false};
BufferTag newTag; /* identity of requested block */
@@ -515,8 +499,8 @@ PrefetchSharedBuffer(SMgrRelation smgr_reln,
Assert(BlockNumberIsValid(blockNum));
/* create a tag so we can lookup the buffer */
- INIT_BUFFERTAG(newTag, smgr_reln->smgr_rlocator.locator,
- forkNum, blockNum);
+ INIT_BUFFERTAG(newTag, sfile->smgr_locator.locator,
+ sfile->smgr_locator.forknum, blockNum);
/* determine its hash code and partition lock ID */
newHash = BufTableHashCode(&newTag);
@@ -535,7 +519,7 @@ PrefetchSharedBuffer(SMgrRelation smgr_reln,
* Try to initiate an asynchronous read. This returns false in
* recovery if the relation file doesn't exist.
*/
- if (smgrprefetch(smgr_reln, forkNum, blockNum))
+ if (smgrprefetch(sfile, blockNum))
result.initiated_io = true;
#endif /* USE_PREFETCH */
}
@@ -603,12 +587,12 @@ PrefetchBuffer(Relation reln, ForkNumber forkNum, BlockNumber blockNum)
errmsg("cannot access temporary tables of other sessions")));
/* pass it off to localbuf.c */
- return PrefetchLocalBuffer(RelationGetSmgr(reln), forkNum, blockNum);
+ return PrefetchLocalBuffer(RelationGetSmgr(reln, forkNum), blockNum);
}
else
{
/* pass it to the shared buffer version */
- return PrefetchSharedBuffer(RelationGetSmgr(reln), forkNum, blockNum);
+ return PrefetchSharedBuffer(RelationGetSmgr(reln, forkNum), blockNum);
}
}
@@ -777,8 +761,9 @@ ReadBufferExtended(Relation reln, ForkNumber forkNum, BlockNumber blockNum,
* miss.
*/
pgstat_count_buffer_read(reln);
- buf = ReadBuffer_common(RelationGetSmgr(reln), reln->rd_rel->relpersistence,
- forkNum, blockNum, mode, strategy, &hit);
+ buf = ReadBuffer_common(RelationGetSmgr(reln, forkNum),
+ reln->rd_rel->relpersistence,
+ blockNum, mode, strategy, &hit);
if (hit)
pgstat_count_buffer_hit(reln);
return buf;
@@ -802,10 +787,10 @@ ReadBufferWithoutRelcache(RelFileLocator rlocator, ForkNumber forkNum,
{
bool hit;
- SMgrRelation smgr = smgropen(rlocator, InvalidBackendId);
+ SMgrFileHandle sfile = smgropen(rlocator, InvalidBackendId, forkNum);
- return ReadBuffer_common(smgr, permanent ? RELPERSISTENCE_PERMANENT :
- RELPERSISTENCE_UNLOGGED, forkNum, blockNum,
+ return ReadBuffer_common(sfile, permanent ? RELPERSISTENCE_PERMANENT :
+ RELPERSISTENCE_UNLOGGED, blockNum,
mode, strategy, &hit);
}
@@ -816,7 +801,7 @@ ReadBufferWithoutRelcache(RelFileLocator rlocator, ForkNumber forkNum,
* *hit is set to true if the request was satisfied from shared buffer cache.
*/
static Buffer
-ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
+ReadBuffer_common(SMgrFileHandle sfile, char relpersistence,
BlockNumber blockNum, ReadBufferMode mode,
BufferAccessStrategy strategy, bool *hit)
{
@@ -824,7 +809,7 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
Block bufBlock;
bool found;
bool isExtend;
- bool isLocalBuf = SmgrIsTemp(smgr);
+ bool isLocalBuf = SmgrIsTemp(sfile);
*hit = false;
@@ -833,29 +818,30 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
isExtend = (blockNum == P_NEW);
- TRACE_POSTGRESQL_BUFFER_READ_START(forkNum, blockNum,
- smgr->smgr_rlocator.locator.spcOid,
- smgr->smgr_rlocator.locator.dbOid,
- smgr->smgr_rlocator.locator.relNumber,
- smgr->smgr_rlocator.backend,
+ TRACE_POSTGRESQL_BUFFER_READ_START(smgr->smgr_locator.forkNum,
+ blockNum,
+ smgr->smgr_locator.locator.spcOid,
+ smgr->smgr_locator.locator.dbOid,
+ smgr->smgr_locator.locator.relNumber,
+ smgr->smgr_locator.backend,
isExtend);
/* Substitute proper block number if caller asked for P_NEW */
if (isExtend)
{
- blockNum = smgrnblocks(smgr, forkNum);
+ blockNum = smgrnblocks(sfile);
/* Fail if relation is already at maximum possible length */
if (blockNum == P_NEW)
ereport(ERROR,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
errmsg("cannot extend relation %s beyond %u blocks",
- relpath(smgr->smgr_rlocator, forkNum),
+ smgrfilepath(sfile->smgr_locator),
P_NEW)));
}
if (isLocalBuf)
{
- bufHdr = LocalBufferAlloc(smgr, forkNum, blockNum, &found);
+ bufHdr = LocalBufferAlloc(sfile, blockNum, &found);
if (found)
pgBufferUsage.local_blks_hit++;
else if (isExtend)
@@ -870,7 +856,7 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
* lookup the buffer. IO_IN_PROGRESS is set if the requested block is
* not currently in memory.
*/
- bufHdr = BufferAlloc(smgr, relpersistence, forkNum, blockNum,
+ bufHdr = BufferAlloc(sfile, relpersistence, blockNum,
strategy, &found);
if (found)
pgBufferUsage.shared_blks_hit++;
@@ -895,11 +881,12 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
if (VacuumCostActive)
VacuumCostBalance += VacuumCostPageHit;
- TRACE_POSTGRESQL_BUFFER_READ_DONE(forkNum, blockNum,
- smgr->smgr_rlocator.locator.spcOid,
- smgr->smgr_rlocator.locator.dbOid,
- smgr->smgr_rlocator.locator.relNumber,
- smgr->smgr_rlocator.backend,
+ TRACE_POSTGRESQL_BUFFER_READ_DONE(sfile->smgr_forknum,
+ blockNum,
+ sfile->smgr_rlocator.locator.spcOid,
+ sfile->smgr_rlocator.locator.dbOid,
+ sfile->smgr_rlocator.locator.relNumber,
+ sfile->smgr_rlocator.backend,
isExtend,
found);
@@ -936,7 +923,7 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
if (!PageIsNew((Page) bufBlock))
ereport(ERROR,
(errmsg("unexpected data beyond EOF in block %u of relation %s",
- blockNum, relpath(smgr->smgr_rlocator, forkNum)),
+ blockNum, smgrfilepath(sfile->smgr_locator)),
errhint("This has been seen to occur with buggy kernels; consider updating your system.")));
/*
@@ -993,7 +980,7 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
/* new buffers are zero-filled */
MemSet((char *) bufBlock, 0, BLCKSZ);
/* don't set checksum for all-zero page */
- smgrextend(smgr, forkNum, blockNum, (char *) bufBlock, false);
+ smgrextend(sfile, blockNum, (char *) bufBlock, false);
/*
* NB: we're *not* doing a ScheduleBufferTagForWriteback here;
@@ -1018,7 +1005,7 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
if (track_io_timing)
INSTR_TIME_SET_CURRENT(io_start);
- smgrread(smgr, forkNum, blockNum, (char *) bufBlock);
+ smgrread(sfile, blockNum, (char *) bufBlock);
if (track_io_timing)
{
@@ -1038,7 +1025,7 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
(errcode(ERRCODE_DATA_CORRUPTED),
errmsg("invalid page in block %u of relation %s; zeroing out page",
blockNum,
- relpath(smgr->smgr_rlocator, forkNum))));
+ smgrfilepath(sfile->smgr_locator))));
MemSet((char *) bufBlock, 0, BLCKSZ);
}
else
@@ -1046,7 +1033,7 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
(errcode(ERRCODE_DATA_CORRUPTED),
errmsg("invalid page in block %u of relation %s",
blockNum,
- relpath(smgr->smgr_rlocator, forkNum))));
+ smgrfilepath(sfile->smgr_locator))));
}
}
}
@@ -1085,11 +1072,12 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
if (VacuumCostActive)
VacuumCostBalance += VacuumCostPageMiss;
- TRACE_POSTGRESQL_BUFFER_READ_DONE(forkNum, blockNum,
- smgr->smgr_rlocator.locator.spcOid,
- smgr->smgr_rlocator.locator.dbOid,
- smgr->smgr_rlocator.locator.relNumber,
- smgr->smgr_rlocator.backend,
+ TRACE_POSTGRESQL_BUFFER_READ_DONE(sfile->smgr_locator.forknum,
+ blockNum,
+ sfile->smgr_locator.locator.spcOid,
+ sfile->smgr_locator.locator.dbOid,
+ sfile->smgr_locator.locator.relNumber,
+ sfile->smgr_locator.backend,
isExtend,
found);
@@ -1116,7 +1104,7 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
* No locks are held either at entry or exit.
*/
static BufferDesc *
-BufferAlloc(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
+BufferAlloc(SMgrFileHandle sfile, char relpersistence,
BlockNumber blockNum,
BufferAccessStrategy strategy,
bool *foundPtr)
@@ -1134,7 +1122,7 @@ BufferAlloc(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
uint32 buf_state;
/* create a tag so we can lookup the buffer */
- INIT_BUFFERTAG(newTag, smgr->smgr_rlocator.locator, forkNum, blockNum);
+ INIT_BUFFERTAG(newTag, sfile->smgr_locator.locator, sfile->smgr_locator.forknum, blockNum);
/* determine its hash code and partition lock ID */
newHash = BufTableHashCode(&newTag);
@@ -1264,10 +1252,11 @@ BufferAlloc(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
}
/* OK, do the I/O */
- TRACE_POSTGRESQL_BUFFER_WRITE_DIRTY_START(forkNum, blockNum,
- smgr->smgr_rlocator.locator.spcOid,
- smgr->smgr_rlocator.locator.dbOid,
- smgr->smgr_rlocator.locator.relNumber);
+ TRACE_POSTGRESQL_BUFFER_WRITE_DIRTY_START(sfile->smgr_locator.forknum,
+ blockNum,
+ sfile->smgr_locator.locator.spcOid,
+ sfile->smgr_locator.locator.dbOid,
+ sfile->smgr_locator.locator.relNumber);
FlushBuffer(buf, NULL);
LWLockRelease(BufferDescriptorGetContentLock(buf));
@@ -1275,10 +1264,11 @@ BufferAlloc(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
ScheduleBufferTagForWriteback(&BackendWritebackContext,
&buf->tag);
- TRACE_POSTGRESQL_BUFFER_WRITE_DIRTY_DONE(forkNum, blockNum,
- smgr->smgr_rlocator.locator.spcOid,
- smgr->smgr_rlocator.locator.dbOid,
- smgr->smgr_rlocator.locator.relNumber);
+ TRACE_POSTGRESQL_BUFFER_WRITE_DIRTY_DONE(sfile->smgr_locator.forknum,
+ blockNum,
+ sfile->smgr_locator.locator.spcOid,
+ sfile->smgr_locator.locator.dbOid,
+ sfile->smgr_locator.locator.relNumber);
}
else
{
@@ -1434,7 +1424,7 @@ BufferAlloc(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
buf_state &= ~(BM_VALID | BM_DIRTY | BM_JUST_DIRTIED |
BM_CHECKPOINT_NEEDED | BM_IO_ERROR | BM_PERMANENT |
BUF_USAGECOUNT_MASK);
- if (relpersistence == RELPERSISTENCE_PERMANENT || forkNum == INIT_FORKNUM)
+ if (relpersistence == RELPERSISTENCE_PERMANENT || sfile->smgr_locator.forknum == INIT_FORKNUM)
buf_state |= BM_TAG_VALID | BM_PERMANENT | BUF_USAGECOUNT_ONE;
else
buf_state |= BM_TAG_VALID | BUF_USAGECOUNT_ONE;
@@ -2822,7 +2812,7 @@ BufferGetTag(Buffer buffer, RelFileLocator *rlocator, ForkNumber *forknum,
* as the second parameter. If not, pass NULL.
*/
static void
-FlushBuffer(BufferDesc *buf, SMgrRelation reln)
+FlushBuffer(BufferDesc *buf, SMgrFileHandle sfile)
{
XLogRecPtr recptr;
ErrorContextCallback errcallback;
@@ -2847,14 +2837,14 @@ FlushBuffer(BufferDesc *buf, SMgrRelation reln)
error_context_stack = &errcallback;
/* Find smgr relation for buffer */
- if (reln == NULL)
- reln = smgropen(buf->tag.rlocator, InvalidBackendId);
+ if (sfile == NULL)
+ sfile = smgropen(buf->tag.rlocator, InvalidBackendId, buf->tag.forkNum);
- TRACE_POSTGRESQL_BUFFER_FLUSH_START(buf->tag.forkNum,
+ TRACE_POSTGRESQL_BUFFER_FLUSH_START(sfile->smgr_locator.forknum,
buf->tag.blockNum,
- reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
- reln->smgr_rlocator.locator.relNumber);
+ sfile->smgr_locator.locator.spcOid,
+ sfile->smgr_locator.locator.dbOid,
+ sfile->smgr_locator.locator.relNumber);
buf_state = LockBufHdr(buf);
@@ -2908,8 +2898,7 @@ FlushBuffer(BufferDesc *buf, SMgrRelation reln)
/*
* bufToWrite is either the shared buffer or a copy, as appropriate.
*/
- smgrwrite(reln,
- buf->tag.forkNum,
+ smgrwrite(sfile,
buf->tag.blockNum,
bufToWrite,
false);
@@ -2930,11 +2919,11 @@ FlushBuffer(BufferDesc *buf, SMgrRelation reln)
*/
TerminateBufferIO(buf, true, 0);
- TRACE_POSTGRESQL_BUFFER_FLUSH_DONE(buf->tag.forkNum,
+ TRACE_POSTGRESQL_BUFFER_FLUSH_DONE(sfile->smgr_locator.forknum,
buf->tag.blockNum,
- reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
- reln->smgr_rlocator.locator.relNumber);
+ sfile->smgr_locator.locator.spcOid,
+ sfile->smgr_locator.locator.dbOid,
+ sfile->smgr_locator.locator.relNumber);
/* Pop the error context stack */
error_context_stack = errcallback.previous;
@@ -2967,7 +2956,7 @@ RelationGetNumberOfBlocksInFork(Relation relation, ForkNumber forkNum)
}
else if (RELKIND_HAS_STORAGE(relation->rd_rel->relkind))
{
- return smgrnblocks(RelationGetSmgr(relation), forkNum);
+ return smgrnblocks(RelationGetSmgr(relation, forkNum));
}
else
Assert(false);
@@ -3057,26 +3046,20 @@ BufferGetLSNAtomic(Buffer buffer)
* --------------------------------------------------------------------
*/
void
-DropRelationBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum,
+DropRelationBuffers(RelFileLocator rlocator, BackendId backend, ForkNumber *forkNum,
int nforks, BlockNumber *firstDelBlock)
{
int i;
int j;
- RelFileLocatorBackend rlocator;
BlockNumber nForkBlock[MAX_FORKNUM];
uint64 nBlocksToInvalidate = 0;
- rlocator = smgr_reln->smgr_rlocator;
-
/* If it's a local relation, it's localbuf.c's problem. */
- if (RelFileLocatorBackendIsTemp(rlocator))
+ if (backend == MyBackendId)
{
- if (rlocator.backend == MyBackendId)
- {
- for (j = 0; j < nforks; j++)
- DropRelationLocalBuffers(rlocator.locator, forkNum[j],
- firstDelBlock[j]);
- }
+ for (j = 0; j < nforks; j++)
+ DropRelationLocalBuffers(rlocator, forkNum[j],
+ firstDelBlock[j]);
return;
}
@@ -3105,7 +3088,10 @@ DropRelationBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum,
for (i = 0; i < nforks; i++)
{
/* Get the number of blocks for a relation's fork */
- nForkBlock[i] = smgrnblocks_cached(smgr_reln, forkNum[i]);
+ SMgrFileHandle sfile;
+
+ sfile = smgropen(rlocator, backend, forkNum[i]);
+ nForkBlock[i] = smgrnblocks_cached(sfile);
if (nForkBlock[i] == InvalidBlockNumber)
{
@@ -3125,7 +3111,7 @@ DropRelationBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum,
nBlocksToInvalidate < BUF_DROP_FULL_SCAN_THRESHOLD)
{
for (j = 0; j < nforks; j++)
- FindAndDropRelationBuffers(rlocator.locator, forkNum[j],
+ FindAndDropRelationBuffers(rlocator, forkNum[j],
nForkBlock[j], firstDelBlock[j]);
return;
}
@@ -3151,14 +3137,14 @@ DropRelationBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum,
* We could check forkNum and blockNum as well as the rlocator, but
* the incremental win from doing so seems small.
*/
- if (!RelFileLocatorEquals(bufHdr->tag.rlocator, rlocator.locator))
+ if (!RelFileLocatorEquals(bufHdr->tag.rlocator, rlocator))
continue;
buf_state = LockBufHdr(bufHdr);
for (j = 0; j < nforks; j++)
{
- if (RelFileLocatorEquals(bufHdr->tag.rlocator, rlocator.locator) &&
+ if (RelFileLocatorEquals(bufHdr->tag.rlocator, rlocator) &&
bufHdr->tag.forkNum == forkNum[j] &&
bufHdr->tag.blockNum >= firstDelBlock[j])
{
@@ -3180,12 +3166,11 @@ DropRelationBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum,
* --------------------------------------------------------------------
*/
void
-DropRelationsAllBuffers(SMgrRelation *smgr_reln, int nlocators)
+DropRelationsAllBuffers(RelFileLocatorBackend *rlocators, int nlocators)
{
int i;
int j;
int n = 0;
- SMgrRelation *rels;
BlockNumber (*block)[MAX_FORKNUM + 1];
uint64 nBlocksToInvalidate = 0;
RelFileLocator *locators;
@@ -3195,18 +3180,18 @@ DropRelationsAllBuffers(SMgrRelation *smgr_reln, int nlocators)
if (nlocators == 0)
return;
- rels = palloc(sizeof(SMgrRelation) * nlocators); /* non-local relations */
+ locators = palloc(sizeof(RelFileLocator) * nlocators); /* non-local relations */
/* If it's a local relation, it's localbuf.c's problem. */
for (i = 0; i < nlocators; i++)
{
- if (RelFileLocatorBackendIsTemp(smgr_reln[i]->smgr_rlocator))
+ if (rlocators[i].backend != InvalidBackendId)
{
- if (smgr_reln[i]->smgr_rlocator.backend == MyBackendId)
- DropRelationAllLocalBuffers(smgr_reln[i]->smgr_rlocator.locator);
+ if (rlocators[i].backend == MyBackendId)
+ DropRelationAllLocalBuffers(rlocators[i].locator);
}
else
- rels[n++] = smgr_reln[i];
+ locators[n++] = rlocators[i].locator;
}
/*
@@ -3215,7 +3200,7 @@ DropRelationsAllBuffers(SMgrRelation *smgr_reln, int nlocators)
*/
if (n == 0)
{
- pfree(rels);
+ pfree(locators);
return;
}
@@ -3235,12 +3220,13 @@ DropRelationsAllBuffers(SMgrRelation *smgr_reln, int nlocators)
for (j = 0; j <= MAX_FORKNUM; j++)
{
/* Get the number of blocks for a relation's fork. */
- block[i][j] = smgrnblocks_cached(rels[i], j);
+ SMgrFileHandle sfile = smgropen(locators[i], InvalidBackendId, j);
+ block[i][j] = smgrnblocks_cached(sfile);
/* We need to only consider the relation forks that exists. */
if (block[i][j] == InvalidBlockNumber)
{
- if (!smgrexists(rels[i], j))
+ if (!smgrexists(sfile))
continue;
cached = false;
break;
@@ -3266,20 +3252,17 @@ DropRelationsAllBuffers(SMgrRelation *smgr_reln, int nlocators)
continue;
/* drop all the buffers for a particular relation fork */
- FindAndDropRelationBuffers(rels[i]->smgr_rlocator.locator,
+ FindAndDropRelationBuffers(locators[i],
j, block[i][j], 0);
}
}
pfree(block);
- pfree(rels);
+ pfree(locators);
return;
}
pfree(block);
- locators = palloc(sizeof(RelFileLocator) * n); /* non-local relations */
- for (i = 0; i < n; i++)
- locators[i] = rels[i]->smgr_rlocator.locator;
/*
* For low number of relations to drop just use a simple walk through, to
@@ -3336,7 +3319,6 @@ DropRelationsAllBuffers(SMgrRelation *smgr_reln, int nlocators)
}
pfree(locators);
- pfree(rels);
}
/* ---------------------------------------------------------------------
@@ -3543,8 +3525,7 @@ FlushRelationBuffers(Relation rel)
PageSetChecksumInplace(localpage, bufHdr->tag.blockNum);
- smgrwrite(RelationGetSmgr(rel),
- bufHdr->tag.forkNum,
+ smgrwrite(RelationGetSmgr(rel, bufHdr->tag.forkNum),
bufHdr->tag.blockNum,
localpage,
false);
@@ -3584,7 +3565,7 @@ FlushRelationBuffers(Relation rel)
{
PinBuffer_Locked(bufHdr);
LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), LW_SHARED);
- FlushBuffer(bufHdr, RelationGetSmgr(rel));
+ FlushBuffer(bufHdr, RelationGetSmgr(rel, bufHdr->tag.forkNum));
LWLockRelease(BufferDescriptorGetContentLock(bufHdr));
UnpinBuffer(bufHdr, true);
}
@@ -3603,25 +3584,20 @@ FlushRelationBuffers(Relation rel)
* --------------------------------------------------------------------
*/
void
-FlushRelationsAllBuffers(SMgrRelation *smgrs, int nrels)
+FlushRelationsAllBuffers(RelFileLocator *rels, int nrels)
{
int i;
- SMgrSortArray *srels;
+ RelFileLocator *locators = NULL;
bool use_bsearch;
if (nrels == 0)
return;
/* fill-in array for qsort */
- srels = palloc(sizeof(SMgrSortArray) * nrels);
+ locators = palloc(sizeof(RelFileLocator) * nrels);
for (i = 0; i < nrels; i++)
- {
- Assert(!RelFileLocatorBackendIsTemp(smgrs[i]->smgr_rlocator));
-
- srels[i].rlocator = smgrs[i]->smgr_rlocator.locator;
- srels[i].srel = smgrs[i];
- }
+ locators[i] = rels[i];
/*
* Save the bsearch overhead for low number of relations to sync. See
@@ -3629,16 +3605,16 @@ FlushRelationsAllBuffers(SMgrRelation *smgrs, int nrels)
*/
use_bsearch = nrels > RELS_BSEARCH_THRESHOLD;
- /* sort the list of SMgrRelations if necessary */
+ /* sort the list of locators if necessary */
if (use_bsearch)
- pg_qsort(srels, nrels, sizeof(SMgrSortArray), rlocator_comparator);
+ pg_qsort(locators, nrels, sizeof(RelFileLocator), rlocator_comparator);
/* Make sure we can handle the pin inside the loop */
ResourceOwnerEnlargeBuffers(CurrentResourceOwner);
for (i = 0; i < NBuffers; i++)
{
- SMgrSortArray *srelent = NULL;
+ RelFileLocator *found = NULL;
BufferDesc *bufHdr = GetBufferDescriptor(i);
uint32 buf_state;
@@ -3653,33 +3629,35 @@ FlushRelationsAllBuffers(SMgrRelation *smgrs, int nrels)
for (j = 0; j < nrels; j++)
{
- if (RelFileLocatorEquals(bufHdr->tag.rlocator, srels[j].rlocator))
+ if (RelFileLocatorEquals(bufHdr->tag.rlocator, locators[j]))
{
- srelent = &srels[j];
+ found = &locators[j];
break;
}
}
}
else
{
- srelent = bsearch((const void *) &(bufHdr->tag.rlocator),
- srels, nrels, sizeof(SMgrSortArray),
- rlocator_comparator);
+ found = bsearch((const void *) &(bufHdr->tag.rlocator),
+ locators, nrels, sizeof(RelFileLocator),
+ rlocator_comparator);
}
/* buffer doesn't belong to any of the given relfilelocators; skip it */
- if (srelent == NULL)
+ if (found == NULL)
continue;
+ /* FIXME: cache SMgrFileHandles for the rels, and pass to FlushBuffer */
+
ReservePrivateRefCountEntry();
buf_state = LockBufHdr(bufHdr);
- if (RelFileLocatorEquals(bufHdr->tag.rlocator, srelent->rlocator) &&
+ if (RelFileLocatorEquals(bufHdr->tag.rlocator, *found) &&
(buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY))
{
PinBuffer_Locked(bufHdr);
LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), LW_SHARED);
- FlushBuffer(bufHdr, srelent->srel);
+ FlushBuffer(bufHdr, NULL);
LWLockRelease(BufferDescriptorGetContentLock(bufHdr));
UnpinBuffer(bufHdr, true);
}
@@ -3687,7 +3665,7 @@ FlushRelationsAllBuffers(SMgrRelation *smgrs, int nrels)
UnlockBufHdr(bufHdr, buf_state);
}
- pfree(srels);
+ pfree(locators);
}
/* ---------------------------------------------------------------------
@@ -3722,7 +3700,7 @@ RelationCopyStorageUsingBuffer(Relation src, Relation dst, ForkNumber forkNum,
use_wal = XLogIsNeeded() && (permanent || forkNum == INIT_FORKNUM);
/* Get number of blocks in the source relation. */
- nblocks = smgrnblocks(RelationGetSmgr(src), forkNum);
+ nblocks = smgrnblocks(RelationGetSmgr(src, forkNum));
/* Nothing to copy; just return. */
if (nblocks == 0)
@@ -3820,9 +3798,9 @@ CreateAndCopyRelationData(RelFileLocator src_rlocator,
for (ForkNumber forkNum = MAIN_FORKNUM + 1;
forkNum <= MAX_FORKNUM; forkNum++)
{
- if (smgrexists(RelationGetSmgr(src_rel), forkNum))
+ if (smgrexists(RelationGetSmgr(src_rel, forkNum)))
{
- smgrcreate(RelationGetSmgr(dst_rel), forkNum, false);
+ smgrcreate(RelationGetSmgr(dst_rel, forkNum), false);
/*
* WAL log creation if the relation is persistent, or this is the
@@ -4952,7 +4930,7 @@ IssuePendingWritebacks(WritebackContext *context)
{
PendingWriteback *cur;
PendingWriteback *next;
- SMgrRelation reln;
+ SMgrFileHandle sfile;
int ahead;
BufferTag tag;
Size nblocks = 1;
@@ -4988,8 +4966,8 @@ IssuePendingWritebacks(WritebackContext *context)
i += ahead;
/* and finally tell the kernel to write the data to storage */
- reln = smgropen(tag.rlocator, InvalidBackendId);
- smgrwriteback(reln, tag.forkNum, tag.blockNum, nblocks);
+ sfile = smgropen(tag.rlocator, InvalidBackendId, tag.forkNum);
+ smgrwriteback(sfile, tag.blockNum, nblocks);
}
context->nr_pending = 0;
diff --git a/src/backend/storage/buffer/localbuf.c b/src/backend/storage/buffer/localbuf.c
index 9c038851d75..92767122797 100644
--- a/src/backend/storage/buffer/localbuf.c
+++ b/src/backend/storage/buffer/localbuf.c
@@ -61,14 +61,13 @@ static Block GetLocalBufferStorage(void);
* No-op if prefetching isn't compiled in.
*/
PrefetchBufferResult
-PrefetchLocalBuffer(SMgrRelation smgr, ForkNumber forkNum,
- BlockNumber blockNum)
+PrefetchLocalBuffer(SMgrFileHandle sfile, BlockNumber blockNum)
{
PrefetchBufferResult result = {InvalidBuffer, false};
BufferTag newTag; /* identity of requested block */
LocalBufferLookupEnt *hresult;
- INIT_BUFFERTAG(newTag, smgr->smgr_rlocator.locator, forkNum, blockNum);
+ INIT_BUFFERTAG(newTag, sfile->smgr_locator.locator, sfile->smgr_locator.forknum, blockNum);
/* Initialize local buffers if first request in this session */
if (LocalBufHash == NULL)
@@ -87,7 +86,7 @@ PrefetchLocalBuffer(SMgrRelation smgr, ForkNumber forkNum,
{
#ifdef USE_PREFETCH
/* Not in buffers, so initiate prefetch */
- smgrprefetch(smgr, forkNum, blockNum);
+ smgrprefetch(sfile, blockNum);
result.initiated_io = true;
#endif /* USE_PREFETCH */
}
@@ -106,8 +105,7 @@ PrefetchLocalBuffer(SMgrRelation smgr, ForkNumber forkNum,
* (hence, usage_count is always advanced).
*/
BufferDesc *
-LocalBufferAlloc(SMgrRelation smgr, ForkNumber forkNum, BlockNumber blockNum,
- bool *foundPtr)
+LocalBufferAlloc(SMgrFileHandle sfile, BlockNumber blockNum, bool *foundPtr)
{
BufferTag newTag; /* identity of requested block */
LocalBufferLookupEnt *hresult;
@@ -117,7 +115,7 @@ LocalBufferAlloc(SMgrRelation smgr, ForkNumber forkNum, BlockNumber blockNum,
bool found;
uint32 buf_state;
- INIT_BUFFERTAG(newTag, smgr->smgr_rlocator.locator, forkNum, blockNum);
+ INIT_BUFFERTAG(newTag, sfile->smgr_locator.locator, sfile->smgr_locator.forknum, blockNum);
/* Initialize local buffers if first request in this session */
if (LocalBufHash == NULL)
@@ -134,7 +132,7 @@ LocalBufferAlloc(SMgrRelation smgr, ForkNumber forkNum, BlockNumber blockNum,
Assert(BUFFERTAGS_EQUAL(bufHdr->tag, newTag));
#ifdef LBDEBUG
fprintf(stderr, "LB ALLOC (%u,%d,%d) %d\n",
- smgr->smgr_rlocator.locator.relNumber, forkNum, blockNum, -b - 1);
+ sfile->smgr_locator.locator.relNumber, sfile->smgr_locator.forknum, blockNum, -b - 1);
#endif
buf_state = pg_atomic_read_u32(&bufHdr->state);
@@ -162,7 +160,7 @@ LocalBufferAlloc(SMgrRelation smgr, ForkNumber forkNum, BlockNumber blockNum,
#ifdef LBDEBUG
fprintf(stderr, "LB ALLOC (%u,%d,%d) %d\n",
- smgr->smgr_rlocator.locator.relNumber, forkNum, blockNum,
+ sfile->smgr_locator.locator.relNumber, sfile->smgr_locator.forknum, blockNum,
-nextFreeLocalBuf - 1);
#endif
@@ -211,17 +209,16 @@ LocalBufferAlloc(SMgrRelation smgr, ForkNumber forkNum, BlockNumber blockNum,
*/
if (buf_state & BM_DIRTY)
{
- SMgrRelation oreln;
+ SMgrFileHandle ofile;
Page localpage = (char *) LocalBufHdrGetBlock(bufHdr);
- /* Find smgr relation for buffer */
- oreln = smgropen(bufHdr->tag.rlocator, MyBackendId);
+ /* Find smgr file handle for buffer */
+ ofile = smgropen(bufHdr->tag.rlocator, MyBackendId, bufHdr->tag.forkNum);
PageSetChecksumInplace(localpage, bufHdr->tag.blockNum);
/* And write... */
- smgrwrite(oreln,
- bufHdr->tag.forkNum,
+ smgrwrite(ofile,
bufHdr->tag.blockNum,
localpage,
false);
diff --git a/src/backend/storage/freespace/freespace.c b/src/backend/storage/freespace/freespace.c
index 005def56dcb..73173849bfd 100644
--- a/src/backend/storage/freespace/freespace.c
+++ b/src/backend/storage/freespace/freespace.c
@@ -271,7 +271,7 @@ FreeSpaceMapPrepareTruncateRel(Relation rel, BlockNumber nblocks)
* If no FSM has been created yet for this relation, there's nothing to
* truncate.
*/
- if (!smgrexists(RelationGetSmgr(rel), FSM_FORKNUM))
+ if (!smgrexists(RelationGetSmgr(rel, FSM_FORKNUM)))
return InvalidBlockNumber;
/* Get the location in the FSM of the first removed heap block */
@@ -317,7 +317,7 @@ FreeSpaceMapPrepareTruncateRel(Relation rel, BlockNumber nblocks)
else
{
new_nfsmblocks = fsm_logical_to_physical(first_removed_address);
- if (smgrnblocks(RelationGetSmgr(rel), FSM_FORKNUM) <= new_nfsmblocks)
+ if (smgrnblocks(RelationGetSmgr(rel, FSM_FORKNUM)) <= new_nfsmblocks)
return InvalidBlockNumber; /* nothing to do; the FSM was already
* smaller */
}
@@ -532,14 +532,14 @@ fsm_readbuf(Relation rel, FSMAddress addr, bool extend)
{
BlockNumber blkno = fsm_logical_to_physical(addr);
Buffer buf;
- SMgrRelation reln;
+ SMgrFileHandle fsm_file;
/*
* Caution: re-using this smgr pointer could fail if the relcache entry
* gets closed. It's safe as long as we only do smgr-level operations
* between here and the last use of the pointer.
*/
- reln = RelationGetSmgr(rel);
+ fsm_file = RelationGetSmgr(rel, FSM_FORKNUM);
/*
* If we haven't cached the size of the FSM yet, check it first. Also
@@ -547,19 +547,19 @@ fsm_readbuf(Relation rel, FSMAddress addr, bool extend)
* value might be stale. (We send smgr inval messages on truncation, but
* not on extension.)
*/
- if (reln->smgr_cached_nblocks[FSM_FORKNUM] == InvalidBlockNumber ||
- blkno >= reln->smgr_cached_nblocks[FSM_FORKNUM])
+ if (fsm_file->smgr_cached_nblocks == InvalidBlockNumber ||
+ blkno >= fsm_file->smgr_cached_nblocks)
{
/* Invalidate the cache so smgrnblocks asks the kernel. */
- reln->smgr_cached_nblocks[FSM_FORKNUM] = InvalidBlockNumber;
- if (smgrexists(reln, FSM_FORKNUM))
- smgrnblocks(reln, FSM_FORKNUM);
+ fsm_file->smgr_cached_nblocks = InvalidBlockNumber;
+ if (smgrexists(fsm_file))
+ smgrnblocks(fsm_file);
else
- reln->smgr_cached_nblocks[FSM_FORKNUM] = 0;
+ fsm_file->smgr_cached_nblocks = 0;
}
/* Handle requests beyond EOF */
- if (blkno >= reln->smgr_cached_nblocks[FSM_FORKNUM])
+ if (blkno >= fsm_file->smgr_cached_nblocks)
{
if (extend)
fsm_extend(rel, blkno + 1);
@@ -609,7 +609,7 @@ fsm_extend(Relation rel, BlockNumber fsm_nblocks)
{
BlockNumber fsm_nblocks_now;
PGAlignedBlock pg;
- SMgrRelation reln;
+ SMgrFileHandle fsm_file;
PageInit((Page) pg.data, BLCKSZ, 0);
@@ -630,29 +630,28 @@ fsm_extend(Relation rel, BlockNumber fsm_nblocks)
* gets closed. It's safe as long as we only do smgr-level operations
* between here and the last use of the pointer.
*/
- reln = RelationGetSmgr(rel);
+ fsm_file = RelationGetSmgr(rel, FSM_FORKNUM);
/*
* Create the FSM file first if it doesn't exist. If
* smgr_cached_nblocks[FSM_FORKNUM] is positive then it must exist, no
* need for an smgrexists call.
*/
- if ((reln->smgr_cached_nblocks[FSM_FORKNUM] == 0 ||
- reln->smgr_cached_nblocks[FSM_FORKNUM] == InvalidBlockNumber) &&
- !smgrexists(reln, FSM_FORKNUM))
- smgrcreate(reln, FSM_FORKNUM, false);
+ if ((fsm_file->smgr_cached_nblocks == 0 ||
+ fsm_file->smgr_cached_nblocks == InvalidBlockNumber) &&
+ !smgrexists(fsm_file))
+ smgrcreate(fsm_file, false);
/* Invalidate cache so that smgrnblocks() asks the kernel. */
- reln->smgr_cached_nblocks[FSM_FORKNUM] = InvalidBlockNumber;
- fsm_nblocks_now = smgrnblocks(reln, FSM_FORKNUM);
+ fsm_file->smgr_cached_nblocks = InvalidBlockNumber;
+ fsm_nblocks_now = smgrnblocks(fsm_file);
/* Extend as needed. */
while (fsm_nblocks_now < fsm_nblocks)
{
PageSetChecksumInplace((Page) pg.data, fsm_nblocks_now);
- smgrextend(reln, FSM_FORKNUM, fsm_nblocks_now,
- pg.data, false);
+ smgrextend(fsm_file, fsm_nblocks_now, pg.data, false);
fsm_nblocks_now++;
}
diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c
index 3998296a62f..4528e45a5ca 100644
--- a/src/backend/storage/smgr/md.c
+++ b/src/backend/storage/smgr/md.c
@@ -66,10 +66,10 @@
* out to an unlinked old copy of a segment file that will eventually
* disappear.
*
- * File descriptors are stored in the per-fork md_seg_fds arrays inside
- * SMgrRelation. The length of these arrays is stored in md_num_open_segs.
- * Note that a fork's md_num_open_segs having a specific value does not
- * necessarily mean the relation doesn't have additional segments; we may
+ * File descriptors are stored in the md_seg_fds array inside
+ * SMgrFileData. The length of the array is stored in md_num_open_segs.
+ * Note that md_num_open_segs having a specific value does not
+ * necessarily mean the file doesn't have additional segments; we may
* just not have opened the next segment yet. (We could not have "all
* segments are in the array" as an invariant anyway, since another backend
* could extend the relation while we aren't looking.) We do not have
@@ -121,26 +121,18 @@ static MemoryContext MdCxt; /* context for all MdfdVec objects */
/* local routines */
-static void mdunlinkfork(RelFileLocatorBackend rlocator, ForkNumber forkNum,
- bool isRedo);
-static MdfdVec *mdopenfork(SMgrRelation reln, ForkNumber forknum, int behavior);
-static void register_dirty_segment(SMgrRelation reln, ForkNumber forknum,
- MdfdVec *seg);
-static void register_unlink_segment(RelFileLocatorBackend rlocator, ForkNumber forknum,
- BlockNumber segno);
-static void register_forget_request(RelFileLocatorBackend rlocator, ForkNumber forknum,
- BlockNumber segno);
-static void _fdvec_resize(SMgrRelation reln,
- ForkNumber forknum,
- int nseg);
-static char *_mdfd_segpath(SMgrRelation reln, ForkNumber forknum,
- BlockNumber segno);
-static MdfdVec *_mdfd_openseg(SMgrRelation reln, ForkNumber forkno,
+static void mdunlinkfork(SMgrFileLocator slocator, bool isRedo);
+static MdfdVec *mdopenfork(SMgrFileHandle sfile, int behavior);
+static void register_dirty_segment(SMgrFileHandle sfile, MdfdVec *seg);
+static void register_unlink_segment(SMgrFileLocator slocator, BlockNumber segno);
+static void register_forget_request(SMgrFileLocator slocator, BlockNumber segno);
+static void _fdvec_resize(SMgrFileHandle sfile, int nseg);
+static char *_mdfd_segpath(SMgrFileHandle sfile, BlockNumber segno);
+static MdfdVec *_mdfd_openseg(SMgrFileHandle sfile,
BlockNumber segno, int oflags);
-static MdfdVec *_mdfd_getseg(SMgrRelation reln, ForkNumber forkno,
+static MdfdVec *_mdfd_getseg(SMgrFileHandle sfile,
BlockNumber blkno, bool skipFsync, int behavior);
-static BlockNumber _mdnblocks(SMgrRelation reln, ForkNumber forknum,
- MdfdVec *seg);
+static BlockNumber _mdnblocks(SMgrFileHandle sfile, MdfdVec *seg);
/*
@@ -160,7 +152,7 @@ mdinit(void)
* Note: this will return true for lingering files, with pending deletions
*/
bool
-mdexists(SMgrRelation reln, ForkNumber forkNum)
+mdexists(SMgrFileHandle sfile)
{
/*
* Close it first, to ensure that we notice if the fork has been unlinked
@@ -168,9 +160,9 @@ mdexists(SMgrRelation reln, ForkNumber forkNum)
* which already closes relations when dropping them.
*/
if (!InRecovery)
- mdclose(reln, forkNum);
+ mdclose(sfile);
- return (mdopenfork(reln, forkNum, EXTENSION_RETURN_NULL) != NULL);
+ return (mdopenfork(sfile, EXTENSION_RETURN_NULL) != NULL);
}
/*
@@ -179,16 +171,16 @@ mdexists(SMgrRelation reln, ForkNumber forkNum)
* If isRedo is true, it's okay for the relation to exist already.
*/
void
-mdcreate(SMgrRelation reln, ForkNumber forkNum, bool isRedo)
+mdcreate(SMgrFileHandle sfile, bool isRedo)
{
MdfdVec *mdfd;
char *path;
File fd;
- if (isRedo && reln->md_num_open_segs[forkNum] > 0)
+ if (isRedo && sfile->md_num_open_segs > 0)
return; /* created and opened already... */
- Assert(reln->md_num_open_segs[forkNum] == 0);
+ Assert(sfile->md_num_open_segs == 0);
/*
* We may be using the target table space for the first time in this
@@ -199,11 +191,11 @@ mdcreate(SMgrRelation reln, ForkNumber forkNum, bool isRedo)
* should be here and not in commands/tablespace.c? But that would imply
* importing a lot of stuff that smgr.c oughtn't know, either.
*/
- TablespaceCreateDbspace(reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
+ TablespaceCreateDbspace(sfile->smgr_locator.locator.spcOid,
+ sfile->smgr_locator.locator.dbOid,
isRedo);
- path = relpath(reln->smgr_rlocator, forkNum);
+ path = smgrfilepath(sfile->smgr_locator);
fd = PathNameOpenFile(path, O_RDWR | O_CREAT | O_EXCL | PG_BINARY);
@@ -225,8 +217,8 @@ mdcreate(SMgrRelation reln, ForkNumber forkNum, bool isRedo)
pfree(path);
- _fdvec_resize(reln, forkNum, 1);
- mdfd = &reln->md_seg_fds[forkNum][0];
+ _fdvec_resize(sfile, 1);
+ mdfd = &sfile->md_seg_fds[0];
mdfd->mdfd_vfd = fd;
mdfd->mdfd_segno = 0;
}
@@ -278,16 +270,9 @@ mdcreate(SMgrRelation reln, ForkNumber forkNum, bool isRedo)
* we are usually not in a transaction anymore when this is called.
*/
void
-mdunlink(RelFileLocatorBackend rlocator, ForkNumber forkNum, bool isRedo)
+mdunlink(SMgrFileLocator slocator, bool isRedo)
{
- /* Now do the per-fork work */
- if (forkNum == InvalidForkNumber)
- {
- for (forkNum = 0; forkNum <= MAX_FORKNUM; forkNum++)
- mdunlinkfork(rlocator, forkNum, isRedo);
- }
- else
- mdunlinkfork(rlocator, forkNum, isRedo);
+ mdunlinkfork(slocator, isRedo);
}
/*
@@ -315,25 +300,25 @@ do_truncate(const char *path)
}
static void
-mdunlinkfork(RelFileLocatorBackend rlocator, ForkNumber forkNum, bool isRedo)
+mdunlinkfork(SMgrFileLocator slocator, bool isRedo)
{
char *path;
int ret;
- path = relpath(rlocator, forkNum);
+ path = smgrfilepath(slocator);
/*
* Delete or truncate the first segment.
*/
- if (isRedo || forkNum != MAIN_FORKNUM || RelFileLocatorBackendIsTemp(rlocator))
+ if (isRedo || slocator.forknum != MAIN_FORKNUM || SMgrFileLocatorIsTemp(slocator))
{
- if (!RelFileLocatorBackendIsTemp(rlocator))
+ if (!SMgrFileLocatorIsTemp(slocator))
{
/* Prevent other backends' fds from holding on to the disk space */
ret = do_truncate(path);
/* Forget any pending sync requests for the first segment */
- register_forget_request(rlocator, forkNum, 0 /* first seg */ );
+ register_forget_request(slocator, 0 /* first seg */ );
}
else
ret = 0;
@@ -354,7 +339,7 @@ mdunlinkfork(RelFileLocatorBackend rlocator, ForkNumber forkNum, bool isRedo)
ret = do_truncate(path);
/* Register request to unlink first segment later */
- register_unlink_segment(rlocator, forkNum, 0 /* first seg */ );
+ register_unlink_segment(slocator, 0 /* first seg */ );
}
/*
@@ -373,7 +358,7 @@ mdunlinkfork(RelFileLocatorBackend rlocator, ForkNumber forkNum, bool isRedo)
{
sprintf(segpath, "%s.%u", path, segno);
- if (!RelFileLocatorBackendIsTemp(rlocator))
+ if (!SMgrFileLocatorIsTemp(slocator))
{
/*
* Prevent other backends' fds from holding on to the disk
@@ -386,7 +371,7 @@ mdunlinkfork(RelFileLocatorBackend rlocator, ForkNumber forkNum, bool isRedo)
* Forget any pending sync requests for this segment before we
* try to unlink.
*/
- register_forget_request(rlocator, forkNum, segno);
+ register_forget_request(slocator, segno);
}
if (unlink(segpath) < 0)
@@ -415,7 +400,7 @@ mdunlinkfork(RelFileLocatorBackend rlocator, ForkNumber forkNum, bool isRedo)
* causes intervening file space to become filled with zeroes.
*/
void
-mdextend(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
+mdextend(SMgrFileHandle sfile, BlockNumber blocknum,
char *buffer, bool skipFsync)
{
off_t seekpos;
@@ -424,7 +409,7 @@ mdextend(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
/* This assert is too expensive to have on normally ... */
#ifdef CHECK_WRITE_VS_EXTEND
- Assert(blocknum >= mdnblocks(reln, forknum));
+ Assert(blocknum >= mdnblocks(sfile));
#endif
/*
@@ -437,10 +422,10 @@ mdextend(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
ereport(ERROR,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
errmsg("cannot extend file \"%s\" beyond %u blocks",
- relpath(reln->smgr_rlocator, forknum),
+ smgrfilepath(sfile->smgr_locator),
InvalidBlockNumber)));
- v = _mdfd_getseg(reln, forknum, blocknum, skipFsync, EXTENSION_CREATE);
+ v = _mdfd_getseg(sfile, blocknum, skipFsync, EXTENSION_CREATE);
seekpos = (off_t) BLCKSZ * (blocknum % ((BlockNumber) RELSEG_SIZE));
@@ -463,10 +448,10 @@ mdextend(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
errhint("Check free disk space.")));
}
- if (!skipFsync && !SmgrIsTemp(reln))
- register_dirty_segment(reln, forknum, v);
+ if (!skipFsync && !SmgrIsTemp(sfile))
+ register_dirty_segment(sfile, v);
- Assert(_mdnblocks(reln, forknum, v) <= ((BlockNumber) RELSEG_SIZE));
+ Assert(_mdnblocks(sfile, v) <= ((BlockNumber) RELSEG_SIZE));
}
/*
@@ -480,17 +465,17 @@ mdextend(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
* invent one out of whole cloth.
*/
static MdfdVec *
-mdopenfork(SMgrRelation reln, ForkNumber forknum, int behavior)
+mdopenfork(SMgrFileHandle sfile, int behavior)
{
MdfdVec *mdfd;
char *path;
File fd;
/* No work if already open */
- if (reln->md_num_open_segs[forknum] > 0)
- return &reln->md_seg_fds[forknum][0];
+ if (sfile->md_num_open_segs > 0)
+ return &sfile->md_seg_fds[0];
- path = relpath(reln->smgr_rlocator, forknum);
+ path = smgrfilepath(sfile->smgr_locator);
fd = PathNameOpenFile(path, O_RDWR | PG_BINARY);
@@ -509,12 +494,12 @@ mdopenfork(SMgrRelation reln, ForkNumber forknum, int behavior)
pfree(path);
- _fdvec_resize(reln, forknum, 1);
- mdfd = &reln->md_seg_fds[forknum][0];
+ _fdvec_resize(sfile, 1);
+ mdfd = &sfile->md_seg_fds[0];
mdfd->mdfd_vfd = fd;
mdfd->mdfd_segno = 0;
- Assert(_mdnblocks(reln, forknum, mdfd) <= ((BlockNumber) RELSEG_SIZE));
+ Assert(_mdnblocks(sfile, mdfd) <= ((BlockNumber) RELSEG_SIZE));
return mdfd;
}
@@ -523,20 +508,19 @@ mdopenfork(SMgrRelation reln, ForkNumber forknum, int behavior)
* mdopen() -- Initialize newly-opened relation.
*/
void
-mdopen(SMgrRelation reln)
+mdopen(SMgrFileHandle sfile)
{
/* mark it not open */
- for (int forknum = 0; forknum <= MAX_FORKNUM; forknum++)
- reln->md_num_open_segs[forknum] = 0;
+ sfile->md_num_open_segs = 0;
}
/*
* mdclose() -- Close the specified relation, if it isn't closed already.
*/
void
-mdclose(SMgrRelation reln, ForkNumber forknum)
+mdclose(SMgrFileHandle sfile)
{
- int nopensegs = reln->md_num_open_segs[forknum];
+ int nopensegs = sfile->md_num_open_segs;
/* No work if already closed */
if (nopensegs == 0)
@@ -545,10 +529,10 @@ mdclose(SMgrRelation reln, ForkNumber forknum)
/* close segments starting from the end */
while (nopensegs > 0)
{
- MdfdVec *v = &reln->md_seg_fds[forknum][nopensegs - 1];
+ MdfdVec *v = &sfile->md_seg_fds[nopensegs - 1];
FileClose(v->mdfd_vfd);
- _fdvec_resize(reln, forknum, nopensegs - 1);
+ _fdvec_resize(sfile, nopensegs - 1);
nopensegs--;
}
}
@@ -557,13 +541,13 @@ mdclose(SMgrRelation reln, ForkNumber forknum)
* mdprefetch() -- Initiate asynchronous read of the specified block of a relation
*/
bool
-mdprefetch(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum)
+mdprefetch(SMgrFileHandle sfile, BlockNumber blocknum)
{
#ifdef USE_PREFETCH
off_t seekpos;
MdfdVec *v;
- v = _mdfd_getseg(reln, forknum, blocknum, false,
+ v = _mdfd_getseg(sfile, blocknum, false,
InRecovery ? EXTENSION_RETURN_NULL : EXTENSION_FAIL);
if (v == NULL)
return false;
@@ -585,8 +569,7 @@ mdprefetch(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum)
* considerably more efficient than doing so individually.
*/
void
-mdwriteback(SMgrRelation reln, ForkNumber forknum,
- BlockNumber blocknum, BlockNumber nblocks)
+mdwriteback(SMgrFileHandle sfile, BlockNumber blocknum, BlockNumber nblocks)
{
/*
* Issue flush requests in as few requests as possible; have to split at
@@ -600,7 +583,7 @@ mdwriteback(SMgrRelation reln, ForkNumber forknum,
int segnum_start,
segnum_end;
- v = _mdfd_getseg(reln, forknum, blocknum, true /* not used */ ,
+ v = _mdfd_getseg(sfile, blocknum, true /* not used */ ,
EXTENSION_DONT_OPEN);
/*
@@ -637,20 +620,21 @@ mdwriteback(SMgrRelation reln, ForkNumber forknum,
* mdread() -- Read the specified block from a relation.
*/
void
-mdread(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
+mdread(SMgrFileHandle sfile, BlockNumber blocknum,
char *buffer)
{
off_t seekpos;
int nbytes;
MdfdVec *v;
- TRACE_POSTGRESQL_SMGR_MD_READ_START(forknum, blocknum,
- reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
- reln->smgr_rlocator.locator.relNumber,
- reln->smgr_rlocator.backend);
+ TRACE_POSTGRESQL_SMGR_MD_READ_START(sfile->smgr_forknum,
+ blocknum,
+ sfile->smgr_rlocator.locator.spcOid,
+ sfile->smgr_rlocator.locator.dbOid,
+ sfile->smgr_rlocator.locator.relNumber,
+ sfile->smgr_rlocator.backend);
- v = _mdfd_getseg(reln, forknum, blocknum, false,
+ v = _mdfd_getseg(sfile, blocknum, false,
EXTENSION_FAIL | EXTENSION_CREATE_RECOVERY);
seekpos = (off_t) BLCKSZ * (blocknum % ((BlockNumber) RELSEG_SIZE));
@@ -659,11 +643,12 @@ mdread(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
nbytes = FileRead(v->mdfd_vfd, buffer, BLCKSZ, seekpos, WAIT_EVENT_DATA_FILE_READ);
- TRACE_POSTGRESQL_SMGR_MD_READ_DONE(forknum, blocknum,
- reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
- reln->smgr_rlocator.locator.relNumber,
- reln->smgr_rlocator.backend,
+ TRACE_POSTGRESQL_SMGR_MD_READ_DONE(sfile->smgr_forknum,
+ blocknum,
+ sfile->smgr_rlocator.locator.spcOid,
+ sfile->smgr_rlocator.locator.dbOid,
+ sfile->smgr_rlocator.locator.relNumber,
+ sfile->smgr_rlocator.backend,
nbytes,
BLCKSZ);
@@ -702,7 +687,7 @@ mdread(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
* use mdextend().
*/
void
-mdwrite(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
+mdwrite(SMgrFileHandle sfile, BlockNumber blocknum,
char *buffer, bool skipFsync)
{
off_t seekpos;
@@ -711,16 +696,17 @@ mdwrite(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
/* This assert is too expensive to have on normally ... */
#ifdef CHECK_WRITE_VS_EXTEND
- Assert(blocknum < mdnblocks(reln, forknum));
+ Assert(blocknum < mdnblocks(sfile));
#endif
- TRACE_POSTGRESQL_SMGR_MD_WRITE_START(forknum, blocknum,
- reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
- reln->smgr_rlocator.locator.relNumber,
- reln->smgr_rlocator.backend);
+ TRACE_POSTGRESQL_SMGR_MD_WRITE_START(sfile->smgr_forknum,
+ blocknum,
+ sfile->smgr_rlocator.locator.spcOid,
+ sfile->smgr_rlocator.locator.dbOid,
+ sfile->smgr_rlocator.locator.relNumber,
+ sfile->smgr_rlocator.backend);
- v = _mdfd_getseg(reln, forknum, blocknum, skipFsync,
+ v = _mdfd_getseg(sfile, blocknum, skipFsync,
EXTENSION_FAIL | EXTENSION_CREATE_RECOVERY);
seekpos = (off_t) BLCKSZ * (blocknum % ((BlockNumber) RELSEG_SIZE));
@@ -729,11 +715,12 @@ mdwrite(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
nbytes = FileWrite(v->mdfd_vfd, buffer, BLCKSZ, seekpos, WAIT_EVENT_DATA_FILE_WRITE);
- TRACE_POSTGRESQL_SMGR_MD_WRITE_DONE(forknum, blocknum,
- reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
- reln->smgr_rlocator.locator.relNumber,
- reln->smgr_rlocator.backend,
+ TRACE_POSTGRESQL_SMGR_MD_WRITE_DONE(sfile->smgr_forknum,
+ blocknum,
+ sfile->smgr_rlocator.locator.spcOid,
+ sfile->smgr_rlocator.locator.dbOid,
+ sfile->smgr_rlocator.locator.relNumber,
+ sfile->smgr_rlocator.backend,
nbytes,
BLCKSZ);
@@ -754,8 +741,8 @@ mdwrite(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
errhint("Check free disk space.")));
}
- if (!skipFsync && !SmgrIsTemp(reln))
- register_dirty_segment(reln, forknum, v);
+ if (!skipFsync && !SmgrIsTemp(sfile))
+ register_dirty_segment(sfile, v);
}
/*
@@ -767,16 +754,16 @@ mdwrite(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
* are present in the array.
*/
BlockNumber
-mdnblocks(SMgrRelation reln, ForkNumber forknum)
+mdnblocks(SMgrFileHandle sfile)
{
MdfdVec *v;
BlockNumber nblocks;
BlockNumber segno;
- mdopenfork(reln, forknum, EXTENSION_FAIL);
+ mdopenfork(sfile, EXTENSION_FAIL);
/* mdopen has opened the first segment */
- Assert(reln->md_num_open_segs[forknum] > 0);
+ Assert(sfile->md_num_open_segs > 0);
/*
* Start from the last open segments, to avoid redundant seeks. We have
@@ -791,12 +778,12 @@ mdnblocks(SMgrRelation reln, ForkNumber forknum)
* that's OK because the checkpointer never needs to compute relation
* size.)
*/
- segno = reln->md_num_open_segs[forknum] - 1;
- v = &reln->md_seg_fds[forknum][segno];
+ segno = sfile->md_num_open_segs - 1;
+ v = &sfile->md_seg_fds[segno];
for (;;)
{
- nblocks = _mdnblocks(reln, forknum, v);
+ nblocks = _mdnblocks(sfile, v);
if (nblocks > ((BlockNumber) RELSEG_SIZE))
elog(FATAL, "segment too big");
if (nblocks < ((BlockNumber) RELSEG_SIZE))
@@ -814,7 +801,7 @@ mdnblocks(SMgrRelation reln, ForkNumber forknum)
* undermines _mdfd_getseg's attempts to notice and report an error
* upon access to a missing segment.
*/
- v = _mdfd_openseg(reln, forknum, segno, 0);
+ v = _mdfd_openseg(sfile, segno, 0);
if (v == NULL)
return segno * ((BlockNumber) RELSEG_SIZE);
}
@@ -824,7 +811,7 @@ mdnblocks(SMgrRelation reln, ForkNumber forknum)
* mdtruncate() -- Truncate relation to specified number of blocks.
*/
void
-mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks)
+mdtruncate(SMgrFileHandle sfile, BlockNumber nblocks)
{
BlockNumber curnblk;
BlockNumber priorblocks;
@@ -834,7 +821,7 @@ mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks)
* NOTE: mdnblocks makes sure we have opened all active segments, so that
* truncation loop will get them all!
*/
- curnblk = mdnblocks(reln, forknum);
+ curnblk = mdnblocks(sfile);
if (nblocks > curnblk)
{
/* Bogus request ... but no complaint if InRecovery */
@@ -842,7 +829,7 @@ mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks)
return;
ereport(ERROR,
(errmsg("could not truncate file \"%s\" to %u blocks: it's only %u blocks now",
- relpath(reln->smgr_rlocator, forknum),
+ smgrfilepath(sfile->smgr_locator),
nblocks, curnblk)));
}
if (nblocks == curnblk)
@@ -852,14 +839,14 @@ mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks)
* Truncate segments, starting at the last one. Starting at the end makes
* managing the memory for the fd array easier, should there be errors.
*/
- curopensegs = reln->md_num_open_segs[forknum];
+ curopensegs = sfile->md_num_open_segs;
while (curopensegs > 0)
{
MdfdVec *v;
priorblocks = (curopensegs - 1) * RELSEG_SIZE;
- v = &reln->md_seg_fds[forknum][curopensegs - 1];
+ v = &sfile->md_seg_fds[curopensegs - 1];
if (priorblocks > nblocks)
{
@@ -873,14 +860,14 @@ mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks)
errmsg("could not truncate file \"%s\": %m",
FilePathName(v->mdfd_vfd))));
- if (!SmgrIsTemp(reln))
- register_dirty_segment(reln, forknum, v);
+ if (!SmgrIsTemp(sfile))
+ register_dirty_segment(sfile, v);
/* we never drop the 1st segment */
- Assert(v != &reln->md_seg_fds[forknum][0]);
+ Assert(v != &sfile->md_seg_fds[0]);
FileClose(v->mdfd_vfd);
- _fdvec_resize(reln, forknum, curopensegs - 1);
+ _fdvec_resize(sfile, curopensegs - 1);
}
else if (priorblocks + ((BlockNumber) RELSEG_SIZE) > nblocks)
{
@@ -899,8 +886,8 @@ mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks)
errmsg("could not truncate file \"%s\" to %u blocks: %m",
FilePathName(v->mdfd_vfd),
nblocks)));
- if (!SmgrIsTemp(reln))
- register_dirty_segment(reln, forknum, v);
+ if (!SmgrIsTemp(sfile))
+ register_dirty_segment(sfile, v);
}
else
{
@@ -926,7 +913,7 @@ mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks)
* segment may survive recovery, reintroducing unwanted data into the table.
*/
void
-mdimmedsync(SMgrRelation reln, ForkNumber forknum)
+mdimmedsync(SMgrFileHandle sfile)
{
int segno;
int min_inactive_seg;
@@ -935,9 +922,9 @@ mdimmedsync(SMgrRelation reln, ForkNumber forknum)
* NOTE: mdnblocks makes sure we have opened all active segments, so that
* fsync loop will get them all!
*/
- mdnblocks(reln, forknum);
+ mdnblocks(sfile);
- min_inactive_seg = segno = reln->md_num_open_segs[forknum];
+ min_inactive_seg = segno = sfile->md_num_open_segs;
/*
* Temporarily open inactive segments, then close them after sync. There
@@ -945,12 +932,12 @@ mdimmedsync(SMgrRelation reln, ForkNumber forknum)
* is harmless. We don't bother to clean them up and take a risk of
* further trouble. The next mdclose() will soon close them.
*/
- while (_mdfd_openseg(reln, forknum, segno, 0) != NULL)
+ while (_mdfd_openseg(sfile, segno, 0) != NULL)
segno++;
while (segno > 0)
{
- MdfdVec *v = &reln->md_seg_fds[forknum][segno - 1];
+ MdfdVec *v = &sfile->md_seg_fds[segno - 1];
if (FileSync(v->mdfd_vfd, WAIT_EVENT_DATA_FILE_IMMEDIATE_SYNC) < 0)
ereport(data_sync_elevel(ERROR),
@@ -962,7 +949,7 @@ mdimmedsync(SMgrRelation reln, ForkNumber forknum)
if (segno > min_inactive_seg)
{
FileClose(v->mdfd_vfd);
- _fdvec_resize(reln, forknum, segno - 1);
+ _fdvec_resize(sfile, segno - 1);
}
segno--;
@@ -979,14 +966,14 @@ mdimmedsync(SMgrRelation reln, ForkNumber forknum)
* enough to be a performance problem).
*/
static void
-register_dirty_segment(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg)
+register_dirty_segment(SMgrFileHandle sfile, MdfdVec *seg)
{
FileTag tag;
- INIT_MD_FILETAG(tag, reln->smgr_rlocator.locator, forknum, seg->mdfd_segno);
+ INIT_MD_FILETAG(tag, sfile->smgr_locator.locator, sfile->smgr_locator.forknum, seg->mdfd_segno);
/* Temp relations should never be fsync'd */
- Assert(!SmgrIsTemp(reln));
+ Assert(!SmgrIsTemp(sfile));
if (!RegisterSyncRequest(&tag, SYNC_REQUEST, false /* retryOnError */ ))
{
@@ -1005,15 +992,14 @@ register_dirty_segment(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg)
* register_unlink_segment() -- Schedule a file to be deleted after next checkpoint
*/
static void
-register_unlink_segment(RelFileLocatorBackend rlocator, ForkNumber forknum,
- BlockNumber segno)
+register_unlink_segment(SMgrFileLocator slocator, BlockNumber segno)
{
FileTag tag;
- INIT_MD_FILETAG(tag, rlocator.locator, forknum, segno);
+ INIT_MD_FILETAG(tag, slocator.locator, slocator.forknum, segno);
/* Should never be used with temp relations */
- Assert(!RelFileLocatorBackendIsTemp(rlocator));
+ Assert(!SMgrFileLocatorIsTemp(slocator));
RegisterSyncRequest(&tag, SYNC_UNLINK_REQUEST, true /* retryOnError */ );
}
@@ -1022,12 +1008,11 @@ register_unlink_segment(RelFileLocatorBackend rlocator, ForkNumber forknum,
* register_forget_request() -- forget any fsyncs for a relation fork's segment
*/
static void
-register_forget_request(RelFileLocatorBackend rlocator, ForkNumber forknum,
- BlockNumber segno)
+register_forget_request(SMgrFileLocator slocator, BlockNumber segno)
{
FileTag tag;
- INIT_MD_FILETAG(tag, rlocator.locator, forknum, segno);
+ INIT_MD_FILETAG(tag, slocator.locator, slocator.forknum, segno);
RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ );
}
@@ -1050,57 +1035,25 @@ ForgetDatabaseSyncRequests(Oid dbid)
RegisterSyncRequest(&tag, SYNC_FILTER_REQUEST, true /* retryOnError */ );
}
-/*
- * DropRelationFiles -- drop files of all given relations
- */
-void
-DropRelationFiles(RelFileLocator *delrels, int ndelrels, bool isRedo)
-{
- SMgrRelation *srels;
- int i;
-
- srels = palloc(sizeof(SMgrRelation) * ndelrels);
- for (i = 0; i < ndelrels; i++)
- {
- SMgrRelation srel = smgropen(delrels[i], InvalidBackendId);
-
- if (isRedo)
- {
- ForkNumber fork;
-
- for (fork = 0; fork <= MAX_FORKNUM; fork++)
- XLogDropRelation(delrels[i], fork);
- }
- srels[i] = srel;
- }
-
- smgrdounlinkall(srels, ndelrels, isRedo);
-
- for (i = 0; i < ndelrels; i++)
- smgrclose(srels[i]);
- pfree(srels);
-}
-
/*
* _fdvec_resize() -- Resize the fork's open segments array
*/
static void
-_fdvec_resize(SMgrRelation reln,
- ForkNumber forknum,
+_fdvec_resize(SMgrFileHandle sfile,
int nseg)
{
if (nseg == 0)
{
- if (reln->md_num_open_segs[forknum] > 0)
+ if (sfile->md_num_open_segs > 0)
{
- pfree(reln->md_seg_fds[forknum]);
- reln->md_seg_fds[forknum] = NULL;
+ pfree(sfile->md_seg_fds);
+ sfile->md_seg_fds = NULL;
}
}
- else if (reln->md_num_open_segs[forknum] == 0)
+ else if (sfile->md_num_open_segs == 0)
{
- reln->md_seg_fds[forknum] =
+ sfile->md_seg_fds =
MemoryContextAlloc(MdCxt, sizeof(MdfdVec) * nseg);
}
else
@@ -1111,12 +1064,12 @@ _fdvec_resize(SMgrRelation reln,
* FileClose(), and the memory context internally will sometimes avoid
* doing an actual reallocation.
*/
- reln->md_seg_fds[forknum] =
- repalloc(reln->md_seg_fds[forknum],
+ sfile->md_seg_fds =
+ repalloc(sfile->md_seg_fds,
sizeof(MdfdVec) * nseg);
}
- reln->md_num_open_segs[forknum] = nseg;
+ sfile->md_num_open_segs = nseg;
}
/*
@@ -1124,12 +1077,12 @@ _fdvec_resize(SMgrRelation reln,
* returned string is palloc'd.
*/
static char *
-_mdfd_segpath(SMgrRelation reln, ForkNumber forknum, BlockNumber segno)
+_mdfd_segpath(SMgrFileHandle sfile, BlockNumber segno)
{
char *path,
*fullpath;
- path = relpath(reln->smgr_rlocator, forknum);
+ path = smgrfilepath(sfile->smgr_locator);
if (segno > 0)
{
@@ -1147,14 +1100,14 @@ _mdfd_segpath(SMgrRelation reln, ForkNumber forknum, BlockNumber segno)
* and make a MdfdVec object for it. Returns NULL on failure.
*/
static MdfdVec *
-_mdfd_openseg(SMgrRelation reln, ForkNumber forknum, BlockNumber segno,
+_mdfd_openseg(SMgrFileHandle sfile, BlockNumber segno,
int oflags)
{
MdfdVec *v;
File fd;
char *fullpath;
- fullpath = _mdfd_segpath(reln, forknum, segno);
+ fullpath = _mdfd_segpath(sfile, segno);
/* open the file */
fd = PathNameOpenFile(fullpath, O_RDWR | PG_BINARY | oflags);
@@ -1168,16 +1121,16 @@ _mdfd_openseg(SMgrRelation reln, ForkNumber forknum, BlockNumber segno,
* Segments are always opened in order from lowest to highest, so we must
* be adding a new one at the end.
*/
- Assert(segno == reln->md_num_open_segs[forknum]);
+ Assert(segno == sfile->md_num_open_segs);
- _fdvec_resize(reln, forknum, segno + 1);
+ _fdvec_resize(sfile, segno + 1);
/* fill the entry */
- v = &reln->md_seg_fds[forknum][segno];
+ v = &sfile->md_seg_fds[segno];
v->mdfd_vfd = fd;
v->mdfd_segno = segno;
- Assert(_mdnblocks(reln, forknum, v) <= ((BlockNumber) RELSEG_SIZE));
+ Assert(_mdnblocks(sfile, v) <= ((BlockNumber) RELSEG_SIZE));
/* all done */
return v;
@@ -1192,7 +1145,7 @@ _mdfd_openseg(SMgrRelation reln, ForkNumber forknum, BlockNumber segno,
* EXTENSION_CREATE case.
*/
static MdfdVec *
-_mdfd_getseg(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
+_mdfd_getseg(SMgrFileHandle sfile, BlockNumber blkno,
bool skipFsync, int behavior)
{
MdfdVec *v;
@@ -1207,9 +1160,9 @@ _mdfd_getseg(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
targetseg = blkno / ((BlockNumber) RELSEG_SIZE);
/* if an existing and opened segment, we're done */
- if (targetseg < reln->md_num_open_segs[forknum])
+ if (targetseg < sfile->md_num_open_segs)
{
- v = &reln->md_seg_fds[forknum][targetseg];
+ v = &sfile->md_seg_fds[targetseg];
return v;
}
@@ -1224,19 +1177,19 @@ _mdfd_getseg(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
* 'behavior'). Start with either the last opened, or the first segment if
* none was opened before.
*/
- if (reln->md_num_open_segs[forknum] > 0)
- v = &reln->md_seg_fds[forknum][reln->md_num_open_segs[forknum] - 1];
+ if (sfile->md_num_open_segs > 0)
+ v = &sfile->md_seg_fds[sfile->md_num_open_segs - 1];
else
{
- v = mdopenfork(reln, forknum, behavior);
+ v = mdopenfork(sfile, behavior);
if (!v)
return NULL; /* if behavior & EXTENSION_RETURN_NULL */
}
- for (nextsegno = reln->md_num_open_segs[forknum];
+ for (nextsegno = sfile->md_num_open_segs;
nextsegno <= targetseg; nextsegno++)
{
- BlockNumber nblocks = _mdnblocks(reln, forknum, v);
+ BlockNumber nblocks = _mdnblocks(sfile, v);
int flags = 0;
Assert(nextsegno == v->mdfd_segno + 1);
@@ -1265,7 +1218,7 @@ _mdfd_getseg(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
{
char *zerobuf = palloc0(BLCKSZ);
- mdextend(reln, forknum,
+ mdextend(sfile,
nextsegno * ((BlockNumber) RELSEG_SIZE) - 1,
zerobuf, skipFsync);
pfree(zerobuf);
@@ -1296,11 +1249,11 @@ _mdfd_getseg(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not open file \"%s\" (target block %u): previous segment is only %u blocks",
- _mdfd_segpath(reln, forknum, nextsegno),
+ _mdfd_segpath(sfile, nextsegno),
blkno, nblocks)));
}
- v = _mdfd_openseg(reln, forknum, nextsegno, flags);
+ v = _mdfd_openseg(sfile, nextsegno, flags);
if (v == NULL)
{
@@ -1310,7 +1263,7 @@ _mdfd_getseg(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not open file \"%s\" (target block %u): %m",
- _mdfd_segpath(reln, forknum, nextsegno),
+ _mdfd_segpath(sfile, nextsegno),
blkno)));
}
}
@@ -1322,7 +1275,7 @@ _mdfd_getseg(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
* Get number of blocks present in a single disk file
*/
static BlockNumber
-_mdnblocks(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg)
+_mdnblocks(SMgrFileHandle sfile, MdfdVec *seg)
{
off_t len;
@@ -1345,16 +1298,16 @@ _mdnblocks(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg)
int
mdsyncfiletag(const FileTag *ftag, char *path)
{
- SMgrRelation reln = smgropen(ftag->rlocator, InvalidBackendId);
+ SMgrFileHandle sfile = smgropen(ftag->rlocator, InvalidBackendId, ftag->forknum);
File file;
bool need_to_close;
int result,
save_errno;
/* See if we already have the file open, or need to open it. */
- if (ftag->segno < reln->md_num_open_segs[ftag->forknum])
+ if (ftag->segno < sfile->md_num_open_segs)
{
- file = reln->md_seg_fds[ftag->forknum][ftag->segno].mdfd_vfd;
+ file = sfile->md_seg_fds[ftag->segno].mdfd_vfd;
strlcpy(path, FilePathName(file), MAXPGPATH);
need_to_close = false;
}
@@ -1362,7 +1315,7 @@ mdsyncfiletag(const FileTag *ftag, char *path)
{
char *p;
- p = _mdfd_segpath(reln, ftag->forknum, ftag->segno);
+ p = _mdfd_segpath(sfile, ftag->segno);
strlcpy(path, p, MAXPGPATH);
pfree(p);
diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c
index c1a5febcbfd..6e1ad31654e 100644
--- a/src/backend/storage/smgr/smgr.c
+++ b/src/backend/storage/smgr/smgr.c
@@ -41,27 +41,24 @@ typedef struct f_smgr
{
void (*smgr_init) (void); /* may be NULL */
void (*smgr_shutdown) (void); /* may be NULL */
- void (*smgr_open) (SMgrRelation reln);
- void (*smgr_close) (SMgrRelation reln, ForkNumber forknum);
- void (*smgr_create) (SMgrRelation reln, ForkNumber forknum,
- bool isRedo);
- bool (*smgr_exists) (SMgrRelation reln, ForkNumber forknum);
- void (*smgr_unlink) (RelFileLocatorBackend rlocator, ForkNumber forknum,
- bool isRedo);
- void (*smgr_extend) (SMgrRelation reln, ForkNumber forknum,
+ void (*smgr_open) (SMgrFileHandle sfile);
+ void (*smgr_close) (SMgrFileHandle sfile);
+ void (*smgr_create) (SMgrFileHandle sfile, bool isRedo);
+ bool (*smgr_exists) (SMgrFileHandle sfile);
+ void (*smgr_unlink) (SMgrFileLocator slocator, bool isRedo);
+ void (*smgr_extend) (SMgrFileHandle sfile,
BlockNumber blocknum, char *buffer, bool skipFsync);
- bool (*smgr_prefetch) (SMgrRelation reln, ForkNumber forknum,
+ bool (*smgr_prefetch) (SMgrFileHandle sfile,
BlockNumber blocknum);
- void (*smgr_read) (SMgrRelation reln, ForkNumber forknum,
+ void (*smgr_read) (SMgrFileHandle sfile,
BlockNumber blocknum, char *buffer);
- void (*smgr_write) (SMgrRelation reln, ForkNumber forknum,
+ void (*smgr_write) (SMgrFileHandle sfile,
BlockNumber blocknum, char *buffer, bool skipFsync);
- void (*smgr_writeback) (SMgrRelation reln, ForkNumber forknum,
+ void (*smgr_writeback) (SMgrFileHandle sfile,
BlockNumber blocknum, BlockNumber nblocks);
- BlockNumber (*smgr_nblocks) (SMgrRelation reln, ForkNumber forknum);
- void (*smgr_truncate) (SMgrRelation reln, ForkNumber forknum,
- BlockNumber nblocks);
- void (*smgr_immedsync) (SMgrRelation reln, ForkNumber forknum);
+ BlockNumber (*smgr_nblocks) (SMgrFileHandle sfile);
+ void (*smgr_truncate) (SMgrFileHandle sfile, BlockNumber nblocks);
+ void (*smgr_immedsync) (SMgrFileHandle sfile);
} f_smgr;
static const f_smgr smgrsw[] = {
@@ -88,12 +85,12 @@ static const f_smgr smgrsw[] = {
static const int NSmgr = lengthof(smgrsw);
/*
- * Each backend has a hashtable that stores all extant SMgrRelation objects.
- * In addition, "unowned" SMgrRelation objects are chained together in a list.
+ * Each backend has a hashtable that stores all extant SMgrFileData objects.
+ * In addition, "unowned" SMgrFile objects are chained together in a list.
*/
-static HTAB *SMgrRelationHash = NULL;
+static HTAB *SMgrFileHash = NULL;
-static dlist_head unowned_relns;
+static dlist_head unowned_sfiles;
/* local function prototypes */
static void smgrshutdown(int code, Datum arg);
@@ -142,50 +139,50 @@ smgrshutdown(int code, Datum arg)
*
* This does not attempt to actually open the underlying file.
*/
-SMgrRelation
-smgropen(RelFileLocator rlocator, BackendId backend)
+SMgrFileHandle
+smgropen(RelFileLocator rlocator, BackendId backend, ForkNumber forkNum)
{
- RelFileLocatorBackend brlocator;
- SMgrRelation reln;
+ SMgrFileLocator slocator;
+ SMgrFileHandle sfile;
bool found;
- if (SMgrRelationHash == NULL)
+ if (SMgrFileHash == NULL)
{
/* First time through: initialize the hash table */
HASHCTL ctl;
- ctl.keysize = sizeof(RelFileLocatorBackend);
- ctl.entrysize = sizeof(SMgrRelationData);
- SMgrRelationHash = hash_create("smgr relation table", 400,
+ ctl.keysize = sizeof(SMgrFileLocator);
+ ctl.entrysize = sizeof(SMgrFileData);
+ SMgrFileHash = hash_create("smgr relation table", 400,
&ctl, HASH_ELEM | HASH_BLOBS);
- dlist_init(&unowned_relns);
+ dlist_init(&unowned_sfiles);
}
/* Look up or create an entry */
- brlocator.locator = rlocator;
- brlocator.backend = backend;
- reln = (SMgrRelation) hash_search(SMgrRelationHash,
- (void *) &brlocator,
- HASH_ENTER, &found);
+ slocator.locator = rlocator;
+ slocator.backend = backend;
+ slocator.forknum = forkNum;
+ sfile = (SMgrFileHandle) hash_search(SMgrFileHash,
+ (void *) &slocator,
+ HASH_ENTER, &found);
/* Initialize it if not present before */
if (!found)
{
/* hash_search already filled in the lookup key */
- reln->smgr_owner = NULL;
- reln->smgr_targblock = InvalidBlockNumber;
- for (int i = 0; i <= MAX_FORKNUM; ++i)
- reln->smgr_cached_nblocks[i] = InvalidBlockNumber;
- reln->smgr_which = 0; /* we only have md.c at present */
+ sfile->smgr_owner = NULL;
+ sfile->smgr_targblock = InvalidBlockNumber;
+ sfile->smgr_cached_nblocks = InvalidBlockNumber;
+ sfile->smgr_which = 0; /* we only have md.c at present */
/* implementation-specific initialization */
- smgrsw[reln->smgr_which].smgr_open(reln);
+ smgrsw[sfile->smgr_which].smgr_open(sfile);
/* it has no owner yet */
- dlist_push_tail(&unowned_relns, &reln->node);
+ dlist_push_tail(&unowned_sfiles, &sfile->node);
}
- return reln;
+ return sfile;
}
/*
@@ -195,7 +192,7 @@ smgropen(RelFileLocator rlocator, BackendId backend)
* the only such owners exist in the relcache.
*/
void
-smgrsetowner(SMgrRelation *owner, SMgrRelation reln)
+smgrsetowner(SMgrFileHandle *owner, SMgrFileHandle sfile)
{
/* We don't support "disowning" an SMgrRelation here, use smgrclearowner */
Assert(owner != NULL);
@@ -206,68 +203,66 @@ smgrsetowner(SMgrRelation *owner, SMgrRelation reln)
* depending on the order of processing. It's ok to close the old
* relcache entry early in that case.)
*
- * If there isn't an old owner, then the reln should be in the unowned
+ * If there isn't an old owner, then the sfile should be in the unowned
* list, and we need to remove it.
*/
- if (reln->smgr_owner)
- *(reln->smgr_owner) = NULL;
+ if (sfile->smgr_owner)
+ *(sfile->smgr_owner) = NULL;
else
- dlist_delete(&reln->node);
+ dlist_delete(&sfile->node);
/* Now establish the ownership relationship. */
- reln->smgr_owner = owner;
- *owner = reln;
+ sfile->smgr_owner = owner;
+ *owner = sfile;
}
/*
- * smgrclearowner() -- Remove long-lived reference to an SMgrRelation object
+ * smgrclearowner() -- Remove long-lived reference to an SMgrFileHandle object
* if one exists
*/
void
-smgrclearowner(SMgrRelation *owner, SMgrRelation reln)
+smgrclearowner(SMgrFileHandle *owner, SMgrFileHandle sfile)
{
/* Do nothing if the SMgrRelation object is not owned by the owner */
- if (reln->smgr_owner != owner)
+ if (sfile->smgr_owner != owner)
return;
/* unset the owner's reference */
*owner = NULL;
/* unset our reference to the owner */
- reln->smgr_owner = NULL;
+ sfile->smgr_owner = NULL;
/* add to list of unowned relations */
- dlist_push_tail(&unowned_relns, &reln->node);
+ dlist_push_tail(&unowned_sfiles, &sfile->node);
}
/*
- * smgrexists() -- Does the underlying file for a fork exist?
+ * smgrexists() -- Does the underlying file exist?
*/
bool
-smgrexists(SMgrRelation reln, ForkNumber forknum)
+smgrexists(SMgrFileHandle sfile)
{
- return smgrsw[reln->smgr_which].smgr_exists(reln, forknum);
+ return smgrsw[sfile->smgr_which].smgr_exists(sfile);
}
/*
- * smgrclose() -- Close and delete an SMgrRelation object.
+ * smgrclose() -- Close and delete an SMgrFile object.
*/
void
-smgrclose(SMgrRelation reln)
+smgrclose(SMgrFileHandle sfile)
{
- SMgrRelation *owner;
- ForkNumber forknum;
+ SMgrFileHandle *owner;
- for (forknum = 0; forknum <= MAX_FORKNUM; forknum++)
- smgrsw[reln->smgr_which].smgr_close(reln, forknum);
+ smgrsw[sfile->smgr_which].smgr_close(sfile);
- owner = reln->smgr_owner;
+ owner = sfile->smgr_owner;
if (!owner)
- dlist_delete(&reln->node);
+ dlist_delete(&sfile->node);
- if (hash_search(SMgrRelationHash,
- (void *) &(reln->smgr_rlocator),
+ if (hash_search(SMgrFileHash,
+ (void *) &(sfile->smgr_locator),
HASH_REMOVE, NULL) == NULL)
elog(ERROR, "SMgrRelation hashtable corrupted");
@@ -284,14 +279,11 @@ smgrclose(SMgrRelation reln)
*
* The object remains valid.
*/
-void
-smgrrelease(SMgrRelation reln)
+static void
+smgrrelease(SMgrFileHandle sfile)
{
- for (ForkNumber forknum = 0; forknum <= MAX_FORKNUM; forknum++)
- {
- smgrsw[reln->smgr_which].smgr_close(reln, forknum);
- reln->smgr_cached_nblocks[forknum] = InvalidBlockNumber;
- }
+ smgrsw[sfile->smgr_which].smgr_close(sfile);
+ sfile->smgr_cached_nblocks = InvalidBlockNumber;
}
/*
@@ -303,16 +295,16 @@ void
smgrreleaseall(void)
{
HASH_SEQ_STATUS status;
- SMgrRelation reln;
+ SMgrFileHandle sfile;
/* Nothing to do if hashtable not set up */
- if (SMgrRelationHash == NULL)
+ if (SMgrFileHash == NULL)
return;
- hash_seq_init(&status, SMgrRelationHash);
+ hash_seq_init(&status, SMgrFileHash);
- while ((reln = (SMgrRelation) hash_seq_search(&status)) != NULL)
- smgrrelease(reln);
+ while ((sfile = (SMgrFileHandle) hash_seq_search(&status)) != NULL)
+ smgrrelease(sfile);
}
/*
@@ -322,16 +314,16 @@ void
smgrcloseall(void)
{
HASH_SEQ_STATUS status;
- SMgrRelation reln;
+ SMgrFileHandle sfile;
/* Nothing to do if hashtable not set up */
- if (SMgrRelationHash == NULL)
+ if (SMgrFileHash == NULL)
return;
- hash_seq_init(&status, SMgrRelationHash);
+ hash_seq_init(&status, SMgrFileHash);
- while ((reln = (SMgrRelation) hash_seq_search(&status)) != NULL)
- smgrclose(reln);
+ while ((sfile = (SMgrFileHandle) hash_seq_search(&status)) != NULL)
+ smgrclose(sfile);
}
/*
@@ -343,111 +335,65 @@ smgrcloseall(void)
* such entry exists already.
*/
void
-smgrcloserellocator(RelFileLocatorBackend rlocator)
+smgrcloserellocator(RelFileLocator rlocator, BackendId backend)
{
- SMgrRelation reln;
+ SMgrFileHandle sfile;
/* Nothing to do if hashtable not set up */
- if (SMgrRelationHash == NULL)
+ if (SMgrFileHash == NULL)
return;
- reln = (SMgrRelation) hash_search(SMgrRelationHash,
- (void *) &rlocator,
- HASH_FIND, NULL);
- if (reln != NULL)
- smgrclose(reln);
+ for (int i = 0; i <= MAX_FORKNUM; i++)
+ {
+ SMgrFileLocator slocator = { rlocator, backend, i };
+
+ sfile = (SMgrFileHandle) hash_search(SMgrFileHash,
+ (void *) &slocator,
+ HASH_FIND, NULL);
+ if (sfile != NULL)
+ smgrclose(sfile);
+ }
}
/*
- * smgrcreate() -- Create a new relation.
+ * smgrcreate() -- Create a new file.
*
- * Given an already-created (but presumably unused) SMgrRelation,
+ * Given an already-created (but presumably unused) SMgrFileHandle,
* cause the underlying disk file or other storage for the fork
* to be created.
*/
void
-smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo)
-{
- smgrsw[reln->smgr_which].smgr_create(reln, forknum, isRedo);
-}
-
-/*
- * smgrdosyncall() -- Immediately sync all forks of all given relations
- *
- * All forks of all given relations are synced out to the store.
- *
- * This is equivalent to FlushRelationBuffers() for each smgr relation,
- * then calling smgrimmedsync() for all forks of each relation, but it's
- * significantly quicker so should be preferred when possible.
- */
-void
-smgrdosyncall(SMgrRelation *rels, int nrels)
+smgrcreate(SMgrFileHandle sfile, bool isRedo)
{
- int i = 0;
- ForkNumber forknum;
-
- if (nrels == 0)
- return;
-
- FlushRelationsAllBuffers(rels, nrels);
-
- /*
- * Sync the physical file(s).
- */
- for (i = 0; i < nrels; i++)
- {
- int which = rels[i]->smgr_which;
-
- for (forknum = 0; forknum <= MAX_FORKNUM; forknum++)
- {
- if (smgrsw[which].smgr_exists(rels[i], forknum))
- smgrsw[which].smgr_immedsync(rels[i], forknum);
- }
- }
+ smgrsw[sfile->smgr_which].smgr_create(sfile, isRedo);
}
/*
- * smgrdounlinkall() -- Immediately unlink all forks of all given relations
+ * smgrunlink_multi() -- Immediately unlink given forks of given relation
*
- * All forks of all given relations are removed from the store. This
+ * The given forks of the relation are removed from the store. This
* should not be used during transactional operations, since it can't be
* undone.
*
+ * This handles multiple forks in one call, because the cache invalidation
+ * happens at relation granularity. If we had an smgrunlink() function
+ * to unlink just a single smgr file, and the caller wanted to delete
+ * multiple forks of a single relation, each call would send a new
+ * cache invalidation event, which would be wasteful.
+ *
* If isRedo is true, it is okay for the underlying file(s) to be gone
* already.
*/
void
-smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo)
+smgrunlink_multi(RelFileLocator rlocator, BackendId backend, ForkNumber *forks, int nforks, bool isRedo)
{
+ int which;
int i = 0;
- RelFileLocatorBackend *rlocators;
- ForkNumber forknum;
-
- if (nrels == 0)
- return;
-
- /*
- * Get rid of any remaining buffers for the relations. bufmgr will just
- * drop them without bothering to write the contents.
- */
- DropRelationsAllBuffers(rels, nrels);
-
- /*
- * create an array which contains all relations to be dropped, and close
- * each relation's forks at the smgr level while at it
- */
- rlocators = palloc(sizeof(RelFileLocatorBackend) * nrels);
- for (i = 0; i < nrels; i++)
- {
- RelFileLocatorBackend rlocator = rels[i]->smgr_rlocator;
- int which = rels[i]->smgr_which;
- rlocators[i] = rlocator;
+ which = 0; /* we only have md.c at present */
- /* Close the forks at smgr level */
- for (forknum = 0; forknum <= MAX_FORKNUM; forknum++)
- smgrsw[which].smgr_close(rels[i], forknum);
- }
+ /* Close the forks at smgr level */
+ smgrcloserellocator(rlocator, backend);
/*
* Send a shared-inval message to force other backends to close any
@@ -457,8 +403,7 @@ smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo)
* back to this backend, too, and thereby provide a backstop that we
* closed our own smgr rel.
*/
- for (i = 0; i < nrels; i++)
- CacheInvalidateSmgr(rlocators[i]);
+ CacheInvalidateSmgr(rlocator, backend);
/*
* Delete the physical file(s).
@@ -467,16 +412,11 @@ smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo)
* ERROR, because we've already decided to commit or abort the current
* xact.
*/
-
- for (i = 0; i < nrels; i++)
+ for (i = 0; i < nforks; i++)
{
- int which = rels[i]->smgr_which;
-
- for (forknum = 0; forknum <= MAX_FORKNUM; forknum++)
- smgrsw[which].smgr_unlink(rlocators[i], forknum, isRedo);
+ SMgrFileLocator slocator = { rlocator, backend, forks[i] };
+ smgrsw[which].smgr_unlink(slocator, isRedo);
}
-
- pfree(rlocators);
}
@@ -490,21 +430,21 @@ smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo)
* causes intervening file space to become filled with zeroes.
*/
void
-smgrextend(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
+smgrextend(SMgrFileHandle sfile, BlockNumber blocknum,
char *buffer, bool skipFsync)
{
- smgrsw[reln->smgr_which].smgr_extend(reln, forknum, blocknum,
- buffer, skipFsync);
+ smgrsw[sfile->smgr_which].smgr_extend(sfile, blocknum,
+ buffer, skipFsync);
/*
* Normally we expect this to increase nblocks by one, but if the cached
* value isn't as expected, just invalidate it so the next call asks the
* kernel.
*/
- if (reln->smgr_cached_nblocks[forknum] == blocknum)
- reln->smgr_cached_nblocks[forknum] = blocknum + 1;
+ if (sfile->smgr_cached_nblocks == blocknum)
+ sfile->smgr_cached_nblocks = blocknum + 1;
else
- reln->smgr_cached_nblocks[forknum] = InvalidBlockNumber;
+ sfile->smgr_cached_nblocks = InvalidBlockNumber;
}
/*
@@ -515,13 +455,13 @@ smgrextend(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
* record).
*/
bool
-smgrprefetch(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum)
+smgrprefetch(SMgrFileHandle sfile, BlockNumber blocknum)
{
- return smgrsw[reln->smgr_which].smgr_prefetch(reln, forknum, blocknum);
+ return smgrsw[sfile->smgr_which].smgr_prefetch(sfile, blocknum);
}
/*
- * smgrread() -- read a particular block from a relation into the supplied
+ * smgrread() -- read a particular block from a file into the supplied
* buffer.
*
* This routine is called from the buffer manager in order to
@@ -529,10 +469,9 @@ smgrprefetch(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum)
* return pages in the format that POSTGRES expects.
*/
void
-smgrread(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
- char *buffer)
+smgrread(SMgrFileHandle sfile, BlockNumber blocknum, char *buffer)
{
- smgrsw[reln->smgr_which].smgr_read(reln, forknum, blocknum, buffer);
+ smgrsw[sfile->smgr_which].smgr_read(sfile, blocknum, buffer);
}
/*
@@ -551,11 +490,11 @@ smgrread(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
* do not require fsync.
*/
void
-smgrwrite(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
+smgrwrite(SMgrFileHandle sfile, BlockNumber blocknum,
char *buffer, bool skipFsync)
{
- smgrsw[reln->smgr_which].smgr_write(reln, forknum, blocknum,
- buffer, skipFsync);
+ smgrsw[sfile->smgr_which].smgr_write(sfile, blocknum,
+ buffer, skipFsync);
}
@@ -564,11 +503,11 @@ smgrwrite(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
* blocks.
*/
void
-smgrwriteback(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
+smgrwriteback(SMgrFileHandle sfile, BlockNumber blocknum,
BlockNumber nblocks)
{
- smgrsw[reln->smgr_which].smgr_writeback(reln, forknum, blocknum,
- nblocks);
+ smgrsw[sfile->smgr_which].smgr_writeback(sfile, blocknum,
+ nblocks);
}
/*
@@ -576,18 +515,18 @@ smgrwriteback(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
* supplied relation.
*/
BlockNumber
-smgrnblocks(SMgrRelation reln, ForkNumber forknum)
+smgrnblocks(SMgrFileHandle sfile)
{
BlockNumber result;
/* Check and return if we get the cached value for the number of blocks. */
- result = smgrnblocks_cached(reln, forknum);
+ result = smgrnblocks_cached(sfile);
if (result != InvalidBlockNumber)
return result;
- result = smgrsw[reln->smgr_which].smgr_nblocks(reln, forknum);
+ result = smgrsw[sfile->smgr_which].smgr_nblocks(sfile);
- reln->smgr_cached_nblocks[forknum] = result;
+ sfile->smgr_cached_nblocks = result;
return result;
}
@@ -600,38 +539,41 @@ smgrnblocks(SMgrRelation reln, ForkNumber forknum)
* fork size is not cached.
*/
BlockNumber
-smgrnblocks_cached(SMgrRelation reln, ForkNumber forknum)
+smgrnblocks_cached(SMgrFileHandle sfile)
{
/*
* For now, we only use cached values in recovery due to lack of a shared
* invalidation mechanism for changes in file size.
*/
- if (InRecovery && reln->smgr_cached_nblocks[forknum] != InvalidBlockNumber)
- return reln->smgr_cached_nblocks[forknum];
+ if (InRecovery && sfile->smgr_cached_nblocks != InvalidBlockNumber)
+ return sfile->smgr_cached_nblocks;
return InvalidBlockNumber;
}
/*
- * smgrtruncate() -- Truncate the given forks of supplied relation to
- * each specified numbers of blocks
+ * smgrtruncate_multi() -- Truncate the given forks of supplied relation to
+ * each specified numbers of blocks
*
* The truncation is done immediately, so this can't be rolled back.
*
* The caller must hold AccessExclusiveLock on the relation, to ensure that
* other backends receive the smgr invalidation event that this function sends
* before they access any forks of the relation again.
+ *
+ * Like smgrunlink_multi(), this handles multiple forks in one call because
+ * the cache invalidation happens at relation granularity.
+ *
+ * NB: The caller is responsible for dropping buffers! Before v16, this
+ * function did it.
*/
void
-smgrtruncate(SMgrRelation reln, ForkNumber *forknum, int nforks, BlockNumber *nblocks)
+smgrtruncate_multi(RelFileLocator rlocator, BackendId backend, ForkNumber *forks,
+ int nforks, BlockNumber *nblocks)
{
int i;
- /*
- * Get rid of any buffers for the about-to-be-deleted blocks. bufmgr will
- * just drop them without bothering to write the contents.
- */
- DropRelationBuffers(reln, forknum, nforks, nblocks);
+ Assert(nforks < MAX_FORKNUM + 1);
/*
* Send a shared-inval message to force other backends to close any smgr
@@ -643,15 +585,19 @@ smgrtruncate(SMgrRelation reln, ForkNumber *forknum, int nforks, BlockNumber *nb
* is a performance-critical path.) As in the unlink code, we want to be
* sure the message is sent before we start changing things on-disk.
*/
- CacheInvalidateSmgr(reln->smgr_rlocator);
+ CacheInvalidateSmgr(rlocator, backend);
- /* Do the truncation */
+ /* Do the truncations */
for (i = 0; i < nforks; i++)
{
+ SMgrFileHandle sfile;
+
+ sfile = smgropen(rlocator, backend, forks[i]);
+
/* Make the cached size is invalid if we encounter an error. */
- reln->smgr_cached_nblocks[forknum[i]] = InvalidBlockNumber;
+ sfile->smgr_cached_nblocks = InvalidBlockNumber;
- smgrsw[reln->smgr_which].smgr_truncate(reln, forknum[i], nblocks[i]);
+ smgrsw[sfile->smgr_which].smgr_truncate(sfile, nblocks[i]);
/*
* We might as well update the local smgr_cached_nblocks values. The
@@ -660,7 +606,7 @@ smgrtruncate(SMgrRelation reln, ForkNumber *forknum, int nforks, BlockNumber *nb
* smgr_vm_nblocks, and these ones too at the next command boundary.
* But these ensure they aren't outright wrong until then.
*/
- reln->smgr_cached_nblocks[forknum[i]] = nblocks[i];
+ sfile->smgr_cached_nblocks = nblocks[i];
}
}
@@ -688,9 +634,9 @@ smgrtruncate(SMgrRelation reln, ForkNumber *forknum, int nforks, BlockNumber *nb
* otherwise the sync is not very meaningful.
*/
void
-smgrimmedsync(SMgrRelation reln, ForkNumber forknum)
+smgrimmedsync(SMgrFileHandle sfile)
{
- smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum);
+ smgrsw[sfile->smgr_which].smgr_immedsync(sfile);
}
/*
@@ -714,14 +660,14 @@ AtEOXact_SMgr(void)
* Zap all unowned SMgrRelations. We rely on smgrclose() to remove each
* one from the list.
*/
- dlist_foreach_modify(iter, &unowned_relns)
+ dlist_foreach_modify(iter, &unowned_sfiles)
{
- SMgrRelation rel = dlist_container(SMgrRelationData, node,
- iter.cur);
+ SMgrFileHandle sfile = dlist_container(SMgrFileData, node,
+ iter.cur);
- Assert(rel->smgr_owner == NULL);
+ Assert(sfile->smgr_owner == NULL);
- smgrclose(rel);
+ smgrclose(sfile);
}
}
diff --git a/src/backend/utils/cache/inval.c b/src/backend/utils/cache/inval.c
index eb5782f82a4..bd435215dca 100644
--- a/src/backend/utils/cache/inval.c
+++ b/src/backend/utils/cache/inval.c
@@ -661,11 +661,12 @@ LocalExecuteInvalidationMessage(SharedInvalidationMessage *msg)
* We could have smgr entries for relations of other databases, so no
* short-circuit test is possible here.
*/
- RelFileLocatorBackend rlocator;
+ RelFileLocator rlocator;
+ BackendId backend;
- rlocator.locator = msg->sm.rlocator;
- rlocator.backend = (msg->sm.backend_hi << 16) | (int) msg->sm.backend_lo;
- smgrcloserellocator(rlocator);
+ rlocator = msg->sm.rlocator;
+ backend = (msg->sm.backend_hi << 16) | (int) msg->sm.backend_lo;
+ smgrcloserellocator(rlocator, backend);
}
else if (msg->id == SHAREDINVALRELMAP_ID)
{
@@ -1459,14 +1460,14 @@ CacheInvalidateRelcacheByRelid(Oid relid)
* Thus, the maximum possible backend ID is 2^23-1.
*/
void
-CacheInvalidateSmgr(RelFileLocatorBackend rlocator)
+CacheInvalidateSmgr(RelFileLocator rlocator, BackendId backend)
{
SharedInvalidationMessage msg;
msg.sm.id = SHAREDINVALSMGR_ID;
- msg.sm.backend_hi = rlocator.backend >> 16;
- msg.sm.backend_lo = rlocator.backend & 0xffff;
- msg.sm.rlocator = rlocator.locator;
+ msg.sm.backend_hi = backend >> 16;
+ msg.sm.backend_lo = backend & 0xffff;
+ msg.sm.rlocator = rlocator;
/* check AddCatcacheInvalidationMessage() for an explanation */
VALGRIND_MAKE_MEM_DEFINED(&msg, sizeof(msg));
diff --git a/src/backend/utils/cache/relcache.c b/src/backend/utils/cache/relcache.c
index bdb771d278f..90a1eb91980 100644
--- a/src/backend/utils/cache/relcache.c
+++ b/src/backend/utils/cache/relcache.c
@@ -418,7 +418,7 @@ AllocateRelationDesc(Form_pg_class relp)
relation = (Relation) palloc0(sizeof(RelationData));
/* make sure relation is marked as having no open file yet */
- relation->rd_smgr = NULL;
+ MemSet(relation->rd_smgr, 0, sizeof(relation->rd_smgr));
/*
* Copy the relation tuple form
@@ -1247,7 +1247,7 @@ retry:
RelationInitPhysicalAddr(relation);
/* make sure relation is marked as having no open file yet */
- relation->rd_smgr = NULL;
+ MemSet(relation->rd_smgr, 0, sizeof(relation->rd_smgr));
/*
* now we can free the memory allocated for pg_class_tuple
@@ -1876,7 +1876,7 @@ formrdesc(const char *relationName, Oid relationReltype,
relation = (Relation) palloc0(sizeof(RelationData));
/* make sure relation is marked as having no open file yet */
- relation->rd_smgr = NULL;
+ MemSet(relation->rd_smgr, 0, sizeof(relation->rd_smgr));
/*
* initialize reference count: 1 because it is nailed in cache
@@ -2693,7 +2693,8 @@ RelationClearRelation(Relation relation, bool rebuild)
}
/* rd_smgr must not be swapped, due to back-links from smgr level */
- SWAPFIELD(SMgrRelation, rd_smgr);
+ for (int i = 0; i <= MAX_FORKNUM; i++)
+ SWAPFIELD(SMgrFileHandle, rd_smgr[i]);
/* rd_refcnt must be preserved */
SWAPFIELD(int, rd_refcnt);
/* isnailed shouldn't change */
@@ -3524,7 +3525,7 @@ RelationBuildLocalRelation(const char *relname,
rel = (Relation) palloc0(sizeof(RelationData));
/* make sure relation is marked as having no open file yet */
- rel->rd_smgr = NULL;
+ MemSet(rel->rd_smgr, 0, sizeof(rel->rd_smgr));
/* mark it nailed if appropriate */
rel->rd_isnailed = nailit;
@@ -3748,7 +3749,7 @@ RelationSetNewRelfilenumber(Relation relation, char persistence)
else if (RELKIND_HAS_STORAGE(relation->rd_rel->relkind))
{
/* handle these directly, at least for now */
- SMgrRelation srel;
+ SMgrFileHandle srel;
srel = RelationCreateStorage(newrlocator, persistence, true);
smgrclose(srel);
@@ -6235,7 +6236,7 @@ load_relcache_init_file(bool shared)
/*
* Reset transient-state fields in the relcache entry
*/
- rel->rd_smgr = NULL;
+ MemSet(rel->rd_smgr, 0, sizeof(rel->rd_smgr));
if (rel->rd_isnailed)
rel->rd_refcnt = 1;
else
diff --git a/src/common/relpath.c b/src/common/relpath.c
index 1b6b620ce83..ae87843ae7e 100644
--- a/src/common/relpath.c
+++ b/src/common/relpath.c
@@ -129,7 +129,7 @@ GetDatabasePath(Oid dbOid, Oid spcOid)
}
/*
- * GetRelationPath - construct path to a relation's file
+ * GetSMgrFilePath - construct path to a relation's file
*
* Result is a palloc'd string.
*
@@ -138,7 +138,7 @@ GetDatabasePath(Oid dbOid, Oid spcOid)
* the trouble considering BackendId is just int anyway.
*/
char *
-GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber,
+GetSMgrFilePath(Oid dbOid, Oid spcOid, RelFileNumber relNumber,
int backendId, ForkNumber forkNumber)
{
char *path;
diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h
index 9964c312aa2..04114305547 100644
--- a/src/include/catalog/storage.h
+++ b/src/include/catalog/storage.h
@@ -22,19 +22,20 @@
/* GUC variables */
extern PGDLLIMPORT int wal_skip_threshold;
-extern SMgrRelation RelationCreateStorage(RelFileLocator rlocator,
- char relpersistence,
- bool register_delete);
+extern SMgrFileHandle RelationCreateStorage(RelFileLocator rlocator,
+ char relpersistence,
+ bool register_delete);
extern void RelationDropStorage(Relation rel);
extern void RelationPreserveStorage(RelFileLocator rlocator, bool atCommit);
extern void RelationPreTruncate(Relation rel);
extern void RelationTruncate(Relation rel, BlockNumber nblocks);
-extern void RelationCopyStorage(SMgrRelation src, SMgrRelation dst,
- ForkNumber forkNum, char relpersistence);
+extern void RelationCopyStorage(SMgrFileHandle src, SMgrFileHandle dst,
+ char relpersistence);
extern bool RelFileLocatorSkippingWAL(RelFileLocator rlocator);
extern Size EstimatePendingSyncsSpace(void);
extern void SerializePendingSyncs(Size maxSize, char *startAddress);
extern void RestorePendingSyncs(char *startAddress);
+extern void DropRelationFiles(RelFileLocator *delrels, int ndelrels, bool isRedo);
/*
* These functions used to be in storage/smgr/smgr.c, which explains the
diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h
index 3ab713247f3..be92980fb34 100644
--- a/src/include/common/relpath.h
+++ b/src/include/common/relpath.h
@@ -66,25 +66,25 @@ extern int forkname_chars(const char *str, ForkNumber *fork);
*/
extern char *GetDatabasePath(Oid dbOid, Oid spcOid);
-extern char *GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber,
+extern char *GetSMgrFilePath(Oid dbOid, Oid spcOid, RelFileNumber relNumber,
int backendId, ForkNumber forkNumber);
/*
* Wrapper macros for GetRelationPath. Beware of multiple
- * evaluation of the RelFileLocator or RelFileLocatorBackend argument!
+ * evaluation of the RelFileLocator or SMgrFileLocator argument!
*/
/* First argument is a RelFileLocator */
#define relpathbackend(rlocator, backend, forknum) \
- GetRelationPath((rlocator).dbOid, (rlocator).spcOid, (rlocator).relNumber, \
+ GetSMgrFilePath((rlocator).dbOid, (rlocator).spcOid, (rlocator).relNumber, \
backend, forknum)
/* First argument is a RelFileLocator */
#define relpathperm(rlocator, forknum) \
relpathbackend(rlocator, InvalidBackendId, forknum)
-/* First argument is a RelFileLocatorBackend */
-#define relpath(rlocator, forknum) \
- relpathbackend((rlocator).locator, (rlocator).backend, forknum)
+/* First argument is a SMgrFileLocator */
+#define smgrfilepath(slocator) \
+ GetSMgrFilePath((slocator).locator.dbOid, (slocator).locator.spcOid, (slocator).locator.relNumber, (slocator).backend, (slocator).forknum)
#endif /* RELPATH_H */
diff --git a/src/include/storage/buf_internals.h b/src/include/storage/buf_internals.h
index 69e45900bae..4e232e5d794 100644
--- a/src/include/storage/buf_internals.h
+++ b/src/include/storage/buf_internals.h
@@ -331,10 +331,9 @@ extern int BufTableInsert(BufferTag *tagPtr, uint32 hashcode, int buf_id);
extern void BufTableDelete(BufferTag *tagPtr, uint32 hashcode);
/* localbuf.c */
-extern PrefetchBufferResult PrefetchLocalBuffer(SMgrRelation smgr,
- ForkNumber forkNum,
+extern PrefetchBufferResult PrefetchLocalBuffer(SMgrFileHandle smgr,
BlockNumber blockNum);
-extern BufferDesc *LocalBufferAlloc(SMgrRelation smgr, ForkNumber forkNum,
+extern BufferDesc *LocalBufferAlloc(SMgrFileHandle smgr,
BlockNumber blockNum, bool *foundPtr);
extern void MarkLocalBufferDirty(Buffer buffer);
extern void DropRelationLocalBuffers(RelFileLocator rlocator,
diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h
index bf8cce7ccf6..897f2893860 100644
--- a/src/include/storage/bufmgr.h
+++ b/src/include/storage/bufmgr.h
@@ -59,7 +59,7 @@ typedef struct PrefetchBufferResult
struct WritebackContext;
/* forward declared, to avoid including smgr.h here */
-struct SMgrRelationData;
+struct SMgrFileData;
/* in globals.c ... this duplicates miscadmin.h */
extern PGDLLIMPORT int NBuffers;
@@ -101,8 +101,7 @@ extern PGDLLIMPORT int32 *LocalRefCount;
/*
* prototypes for functions in bufmgr.c
*/
-extern PrefetchBufferResult PrefetchSharedBuffer(struct SMgrRelationData *smgr_reln,
- ForkNumber forkNum,
+extern PrefetchBufferResult PrefetchSharedBuffer(struct SMgrFileData *smgr_file,
BlockNumber blockNum);
extern PrefetchBufferResult PrefetchBuffer(Relation reln, ForkNumber forkNum,
BlockNumber blockNum);
@@ -131,18 +130,26 @@ extern void CheckPointBuffers(int flags);
extern BlockNumber BufferGetBlockNumber(Buffer buffer);
extern BlockNumber RelationGetNumberOfBlocksInFork(Relation relation,
ForkNumber forkNum);
-extern void FlushOneBuffer(Buffer buffer);
-extern void FlushRelationBuffers(Relation rel);
-extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels);
extern void CreateAndCopyRelationData(RelFileLocator src_rlocator,
RelFileLocator dst_rlocator,
bool permanent);
+
+extern void FlushOneBuffer(Buffer buffer);
+extern void FlushRelationBuffers(Relation rel);
+extern void FlushRelationsAllBuffers(RelFileLocator *locators, int nlocators);
extern void FlushDatabaseBuffers(Oid dbid);
-extern void DropRelationBuffers(struct SMgrRelationData *smgr_reln,
+
+extern void DropRelationBuffers(RelFileLocator rlocator, BackendId backend,
ForkNumber *forkNum,
int nforks, BlockNumber *firstDelBlock);
-extern void DropRelationsAllBuffers(struct SMgrRelationData **smgr_reln,
- int nlocators);
+
+typedef struct RelFileLocatorBackend
+{
+ RelFileLocator locator;
+ BackendId backend;
+} RelFileLocatorBackend;
+
+extern void DropRelationsAllBuffers(RelFileLocatorBackend *locators, int nlocators);
extern void DropDatabaseBuffers(Oid dbid);
#define RelationGetNumberOfBlocks(reln) \
diff --git a/src/include/storage/md.h b/src/include/storage/md.h
index 10aa1b0109b..bcb87d56295 100644
--- a/src/include/storage/md.h
+++ b/src/include/storage/md.h
@@ -21,28 +21,26 @@
/* md storage manager functionality */
extern void mdinit(void);
-extern void mdopen(SMgrRelation reln);
-extern void mdclose(SMgrRelation reln, ForkNumber forknum);
-extern void mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo);
-extern bool mdexists(SMgrRelation reln, ForkNumber forknum);
-extern void mdunlink(RelFileLocatorBackend rlocator, ForkNumber forknum, bool isRedo);
-extern void mdextend(SMgrRelation reln, ForkNumber forknum,
+extern void mdopen(SMgrFileHandle sfile);
+extern void mdclose(SMgrFileHandle sfile);
+extern void mdcreate(SMgrFileHandle sfile, bool isRedo);
+extern bool mdexists(SMgrFileHandle sfile);
+extern void mdunlink(SMgrFileLocator slocator, bool isRedo);
+extern void mdextend(SMgrFileHandle sfile,
BlockNumber blocknum, char *buffer, bool skipFsync);
-extern bool mdprefetch(SMgrRelation reln, ForkNumber forknum,
+extern bool mdprefetch(SMgrFileHandle sfile,
BlockNumber blocknum);
-extern void mdread(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
+extern void mdread(SMgrFileHandle sfile, BlockNumber blocknum,
char *buffer);
-extern void mdwrite(SMgrRelation reln, ForkNumber forknum,
+extern void mdwrite(SMgrFileHandle sfile,
BlockNumber blocknum, char *buffer, bool skipFsync);
-extern void mdwriteback(SMgrRelation reln, ForkNumber forknum,
+extern void mdwriteback(SMgrFileHandle sfile,
BlockNumber blocknum, BlockNumber nblocks);
-extern BlockNumber mdnblocks(SMgrRelation reln, ForkNumber forknum);
-extern void mdtruncate(SMgrRelation reln, ForkNumber forknum,
- BlockNumber nblocks);
-extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum);
+extern BlockNumber mdnblocks(SMgrFileHandle sfile);
+extern void mdtruncate(SMgrFileHandle sfile, BlockNumber nblocks);
+extern void mdimmedsync(SMgrFileHandle sfile);
extern void ForgetDatabaseSyncRequests(Oid dbid);
-extern void DropRelationFiles(RelFileLocator *delrels, int ndelrels, bool isRedo);
/* md sync callbacks */
extern int mdsyncfiletag(const FileTag *ftag, char *path);
diff --git a/src/include/storage/relfilelocator.h b/src/include/storage/relfilelocator.h
index 10f41f3abb3..296cb1b8e51 100644
--- a/src/include/storage/relfilelocator.h
+++ b/src/include/storage/relfilelocator.h
@@ -53,6 +53,8 @@
* Note: various places use RelFileLocator in hashtable keys. Therefore,
* there *must not* be any unused padding bytes in this struct. That
* should be safe as long as all the fields are of type Oid.
+ *
+ * See also SMgrFileLocator in smgr.h.
*/
typedef struct RelFileLocator
{
@@ -62,38 +64,15 @@ typedef struct RelFileLocator
} RelFileLocator;
/*
- * Augmenting a relfilelocator with the backend ID provides all the information
- * we need to locate the physical storage. The backend ID is InvalidBackendId
- * for regular relations (those accessible to more than one backend), or the
- * owning backend's ID for backend-local relations. Backend-local relations
- * are always transient and removed in case of a database crash; they are
- * never WAL-logged or fsync'd.
- */
-typedef struct RelFileLocatorBackend
-{
- RelFileLocator locator;
- BackendId backend;
-} RelFileLocatorBackend;
-
-#define RelFileLocatorBackendIsTemp(rlocator) \
- ((rlocator).backend != InvalidBackendId)
-
-/*
- * Note: RelFileLocatorEquals and RelFileLocatorBackendEquals compare relNumber
+ * Note: RelFileLocatorEquals compares relNumber
* first since that is most likely to be different in two unequal
* RelFileLocators. It is probably redundant to compare spcOid if the other
* fields are found equal, but do it anyway to be sure. Likewise for checking
- * the backend ID in RelFileLocatorBackendEquals.
+ * the backend ID in SMgrFileLocatorBackendEquals.
*/
#define RelFileLocatorEquals(locator1, locator2) \
((locator1).relNumber == (locator2).relNumber && \
(locator1).dbOid == (locator2).dbOid && \
(locator1).spcOid == (locator2).spcOid)
-#define RelFileLocatorBackendEquals(locator1, locator2) \
- ((locator1).locator.relNumber == (locator2).locator.relNumber && \
- (locator1).locator.dbOid == (locator2).locator.dbOid && \
- (locator1).backend == (locator2).backend && \
- (locator1).locator.spcOid == (locator2).locator.spcOid)
-
#endif /* RELFILELOCATOR_H */
diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h
index a07715356ba..130e59241e3 100644
--- a/src/include/storage/smgr.h
+++ b/src/include/storage/smgr.h
@@ -18,31 +18,56 @@
#include "storage/block.h"
#include "storage/relfilelocator.h"
+
/*
- * smgr.c maintains a table of SMgrRelation objects, which are essentially
- * cached file handles. An SMgrRelation is created (if not already present)
+ * SMgrFileLocator contains all the information needed to locate the physical
+ * storage of a relation fork, or some other file that is managed by the buffer
+ * manager.
+ *
+ * The backend ID is InvalidBackendId for regular relations (those accessible
+ * to more than one backend), or the owning backend's ID for backend-local
+ * relations. Backend-local relations are always transient and removed in
+ * case of a database crash; they are never WAL-logged or fsync'd.
+ */
+typedef struct SMgrFileLocator
+{
+ RelFileLocator locator;
+ BackendId backend;
+ ForkNumber forknum;
+} SMgrFileLocator;
+
+#define SMgrFileLocatorIsTemp(slocator) \
+ ((slocator).backend != InvalidBackendId)
+
+/*
+ * smgr.c maintains a table of SMgrFileData objects, which are essentially
+ * cached file handles. An SMgrFile is created (if not already present)
* by smgropen(), and destroyed by smgrclose(). Note that neither of these
* operations imply I/O, they just create or destroy a hashtable entry.
* (But smgrclose() may release associated resources, such as OS-level file
* descriptors.)
*
- * An SMgrRelation may have an "owner", which is just a pointer to it from
- * somewhere else; smgr.c will clear this pointer if the SMgrRelation is
+ * An SMgrFile may have an "owner", which is just a pointer to it from
+ * somewhere else; smgr.c will clear this pointer if the SMgrFile is
* closed. We use this to avoid dangling pointers from relcache to smgr
* without having to make the smgr explicitly aware of relcache. There
* can't be more than one "owner" pointer per SMgrRelation, but that's
* all we need.
*
- * SMgrRelations that do not have an "owner" are considered to be transient,
+ * SMgrFiles that do not have an "owner" are considered to be transient,
* and are deleted at end of transaction.
+ *
+ * A file that is represented by an SMgrFile can be managed by the buffer
+ * manager. Currently, it's only used for relation files, but could be used
+ * for SLRUs and other things in the future.
*/
-typedef struct SMgrRelationData
+typedef struct SMgrFileData
{
- /* rlocator is the hashtable lookup key, so it must be first! */
- RelFileLocatorBackend smgr_rlocator; /* relation physical identifier */
+ /* locator is the hashtable lookup key, so must be first! */
+ SMgrFileLocator smgr_locator; /* file physical identifier */
/* pointer to owning pointer, or NULL if none */
- struct SMgrRelationData **smgr_owner;
+ struct SMgrFileData **smgr_owner;
/*
* The following fields are reset to InvalidBlockNumber upon a cache flush
@@ -51,7 +76,7 @@ typedef struct SMgrRelationData
* invalidation for fork extension.
*/
BlockNumber smgr_targblock; /* current insertion target block */
- BlockNumber smgr_cached_nblocks[MAX_FORKNUM + 1]; /* last known size */
+ BlockNumber smgr_cached_nblocks; /* last known size */
/* additional public fields may someday exist here */
@@ -65,46 +90,45 @@ typedef struct SMgrRelationData
* for md.c; per-fork arrays of the number of open segments
* (md_num_open_segs) and the segments themselves (md_seg_fds).
*/
- int md_num_open_segs[MAX_FORKNUM + 1];
- struct _MdfdVec *md_seg_fds[MAX_FORKNUM + 1];
+ int md_num_open_segs;
+ struct _MdfdVec *md_seg_fds;
- /* if unowned, list link in list of all unowned SMgrRelations */
+ /* if unowned, list link in list of all unowned SMgrFiles */
dlist_node node;
-} SMgrRelationData;
+} SMgrFileData;
-typedef SMgrRelationData *SMgrRelation;
+typedef SMgrFileData *SMgrFileHandle;
#define SmgrIsTemp(smgr) \
- RelFileLocatorBackendIsTemp((smgr)->smgr_rlocator)
+ SMgrFileLocatorIsTemp((smgr)->smgr_locator)
extern void smgrinit(void);
-extern SMgrRelation smgropen(RelFileLocator rlocator, BackendId backend);
-extern bool smgrexists(SMgrRelation reln, ForkNumber forknum);
-extern void smgrsetowner(SMgrRelation *owner, SMgrRelation reln);
-extern void smgrclearowner(SMgrRelation *owner, SMgrRelation reln);
-extern void smgrclose(SMgrRelation reln);
+extern SMgrFileHandle smgropen(RelFileLocator rlocator, BackendId backend, ForkNumber forkNum);
+extern bool smgrexists(SMgrFileHandle sfile);
+extern void smgrsetowner(SMgrFileHandle *owner, SMgrFileHandle sfile);
+extern void smgrclearowner(SMgrFileHandle *owner, SMgrFileHandle sfile);
+extern void smgrclose(SMgrFileHandle sfile);
extern void smgrcloseall(void);
-extern void smgrcloserellocator(RelFileLocatorBackend rlocator);
-extern void smgrrelease(SMgrRelation reln);
extern void smgrreleaseall(void);
-extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo);
-extern void smgrdosyncall(SMgrRelation *rels, int nrels);
-extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo);
-extern void smgrextend(SMgrRelation reln, ForkNumber forknum,
+extern void smgrcreate(SMgrFileHandle sfile, bool isRedo);
+extern void smgrextend(SMgrFileHandle sfile,
BlockNumber blocknum, char *buffer, bool skipFsync);
-extern bool smgrprefetch(SMgrRelation reln, ForkNumber forknum,
- BlockNumber blocknum);
-extern void smgrread(SMgrRelation reln, ForkNumber forknum,
+extern bool smgrprefetch(SMgrFileHandle sfile, BlockNumber blocknum);
+extern void smgrread(SMgrFileHandle sfile,
BlockNumber blocknum, char *buffer);
-extern void smgrwrite(SMgrRelation reln, ForkNumber forknum,
+extern void smgrwrite(SMgrFileHandle sfile,
BlockNumber blocknum, char *buffer, bool skipFsync);
-extern void smgrwriteback(SMgrRelation reln, ForkNumber forknum,
+extern void smgrwriteback(SMgrFileHandle sfile,
BlockNumber blocknum, BlockNumber nblocks);
-extern BlockNumber smgrnblocks(SMgrRelation reln, ForkNumber forknum);
-extern BlockNumber smgrnblocks_cached(SMgrRelation reln, ForkNumber forknum);
-extern void smgrtruncate(SMgrRelation reln, ForkNumber *forknum,
- int nforks, BlockNumber *nblocks);
-extern void smgrimmedsync(SMgrRelation reln, ForkNumber forknum);
+extern BlockNumber smgrnblocks(SMgrFileHandle sfile);
+extern BlockNumber smgrnblocks_cached(SMgrFileHandle sfile);
+extern void smgrimmedsync(SMgrFileHandle sfile);
+
+extern void smgrtruncate_multi(RelFileLocator rlocator, BackendId backend, ForkNumber *forks, int nforks, BlockNumber *nblocks);
+extern void smgrunlink_multi(RelFileLocator rlocator, BackendId backend, ForkNumber *forks, int nforks, bool isRedo);
+
+extern void smgrcloserellocator(RelFileLocator rlocator, BackendId backend);
+
extern void AtEOXact_SMgr(void);
extern bool ProcessBarrierSmgrRelease(void);
diff --git a/src/include/utils/inval.h b/src/include/utils/inval.h
index 23748b72caf..c494c01056f 100644
--- a/src/include/utils/inval.h
+++ b/src/include/utils/inval.h
@@ -48,7 +48,7 @@ extern void CacheInvalidateRelcacheByTuple(HeapTuple classTuple);
extern void CacheInvalidateRelcacheByRelid(Oid relid);
-extern void CacheInvalidateSmgr(RelFileLocatorBackend rlocator);
+extern void CacheInvalidateSmgr(RelFileLocator rlocator, BackendId backend);
extern void CacheInvalidateRelmap(Oid databaseId);
diff --git a/src/include/utils/rel.h b/src/include/utils/rel.h
index 7dc401cf0df..6e849a901ce 100644
--- a/src/include/utils/rel.h
+++ b/src/include/utils/rel.h
@@ -54,7 +54,7 @@ typedef LockInfoData *LockInfo;
typedef struct RelationData
{
RelFileLocator rd_locator; /* relation physical identifier */
- SMgrRelation rd_smgr; /* cached file handle, or NULL */
+ SMgrFileHandle rd_smgr[MAX_FORKNUM + 1]; /* cached file handles, or NULLs */
int rd_refcnt; /* reference count */
BackendId rd_backend; /* owning backend id, if temporary relation */
bool rd_islocaltemp; /* rel is a temp rel of this session */
@@ -562,15 +562,15 @@ typedef struct ViewOptions
* Note: since a relcache flush can cause the file handle to be closed again,
* it's unwise to hold onto the pointer returned by this function for any
* long period. Recommended practice is to just re-execute RelationGetSmgr
- * each time you need to access the SMgrRelation. It's quite cheap in
+ * each time you need to access the SMgrFileHandle. It's quite cheap in
* comparison to whatever an smgr function is going to do.
*/
-static inline SMgrRelation
-RelationGetSmgr(Relation rel)
+static inline SMgrFileHandle
+RelationGetSmgr(Relation rel, ForkNumber forkNum)
{
- if (unlikely(rel->rd_smgr == NULL))
- smgrsetowner(&(rel->rd_smgr), smgropen(rel->rd_locator, rel->rd_backend));
- return rel->rd_smgr;
+ if (unlikely(rel->rd_smgr[forkNum] == NULL))
+ smgrsetowner(&(rel->rd_smgr[forkNum]), smgropen(rel->rd_locator, rel->rd_backend, forkNum));
+ return rel->rd_smgr[forkNum];
}
#endif
@@ -580,14 +580,18 @@ RelationGetSmgr(Relation rel)
*
* Note: smgrclose should unhook from owner pointer, hence the Assert.
*/
-#define RelationCloseSmgr(relation) \
- do { \
- if ((relation)->rd_smgr != NULL) \
- { \
- smgrclose((relation)->rd_smgr); \
- Assert((relation)->rd_smgr == NULL); \
- } \
- } while (0)
+static inline void
+RelationCloseSmgr(Relation relation)
+{
+ for (int i = 0; i <= MAX_FORKNUM; i++)
+ {
+ if (relation->rd_smgr[i] != NULL)
+ {
+ smgrclose(relation->rd_smgr[i]);
+ Assert(relation->rd_smgr[i] == NULL);
+ }
+ }
+}
/*
* RelationGetTargetBlock
@@ -598,7 +602,7 @@ RelationGetSmgr(Relation rel)
* so there's no need to re-open the smgr handle if it's not currently open.
*/
#define RelationGetTargetBlock(relation) \
- ( (relation)->rd_smgr != NULL ? (relation)->rd_smgr->smgr_targblock : InvalidBlockNumber )
+ ( (relation)->rd_smgr[MAIN_FORKNUM] != NULL ? (relation)->rd_smgr[MAIN_FORKNUM]->smgr_targblock : InvalidBlockNumber )
/*
* RelationSetTargetBlock
@@ -606,7 +610,7 @@ RelationGetSmgr(Relation rel)
*/
#define RelationSetTargetBlock(relation, targblock) \
do { \
- RelationGetSmgr(relation)->smgr_targblock = (targblock); \
+ RelationGetSmgr(relation, MAIN_FORKNUM)->smgr_targblock = (targblock); \
} while (0)
/*
--
2.30.2
[text/x-patch] 0002-WIP-SLRUs.patch (145.8K, ../../[email protected]/3-0002-WIP-SLRUs.patch)
download | inline diff:
From 8b2c71dad454bb81fe8ab9cb1e8ff726ce38750d Mon Sep 17 00:00:00 2001
From: Heikki Linnakangas <[email protected]>
Date: Mon, 25 Jul 2022 12:07:49 +0300
Subject: [PATCH 2/2] WIP: SLRUs
This is Thomas's patch refactored over the per-fork SMgrFile patch:
- each SLRU segment is represented by a separate SMgrFile.
- md.c implementation handles SLRUs, too
---
src/backend/access/transam/clog.c | 199 +---
src/backend/access/transam/commit_ts.c | 156 +--
src/backend/access/transam/multixact.c | 357 ++----
src/backend/access/transam/slru.c | 1487 +++---------------------
src/backend/access/transam/subtrans.c | 108 +-
src/backend/access/transam/xact.c | 2 +
src/backend/access/transam/xlog.c | 15 +-
src/backend/commands/async.c | 67 +-
src/backend/storage/buffer/buf_init.c | 17 +-
src/backend/storage/buffer/bufmgr.c | 62 +-
src/backend/storage/ipc/ipci.c | 4 -
src/backend/storage/lmgr/predicate.c | 62 +-
src/backend/storage/smgr/md.c | 9 +-
src/backend/storage/smgr/smgr.c | 43 +
src/backend/storage/sync/sync.c | 20 +-
src/backend/utils/mmgr/mcxt.c | 13 +-
src/common/relpath.c | 27 +-
src/include/access/clog.h | 6 -
src/include/access/commit_ts.h | 3 -
src/include/access/multixact.h | 3 -
src/include/access/slru.h | 180 +--
src/include/access/slrulist.h | 30 +
src/include/access/subtrans.h | 3 -
src/include/common/relpath.h | 3 +
src/include/storage/buf_internals.h | 6 +
src/include/storage/bufmgr.h | 8 +-
src/include/storage/smgr.h | 1 +
27 files changed, 689 insertions(+), 2202 deletions(-)
create mode 100644 src/include/access/slrulist.h
diff --git a/src/backend/access/transam/clog.c b/src/backend/access/transam/clog.c
index 3d9088a7048..fbda2c69ead 100644
--- a/src/backend/access/transam/clog.c
+++ b/src/backend/access/transam/clog.c
@@ -41,6 +41,8 @@
#include "miscadmin.h"
#include "pg_trace.h"
#include "pgstat.h"
+#include "storage/bufmgr.h"
+#include "storage/buf_internals.h"
#include "storage/proc.h"
#include "storage/sync.h"
@@ -81,15 +83,8 @@
*/
#define THRESHOLD_SUBTRANS_CLOG_OPT 5
-/*
- * Link to shared-memory data structures for CLOG control
- */
-static SlruCtlData XactCtlData;
-
-#define XactCtl (&XactCtlData)
-
-static int ZeroCLOGPage(int pageno, bool writeXlog);
+static Buffer ZeroCLOGPage(int pageno, bool writeXlog);
static bool CLOGPagePrecedes(int page1, int page2);
static void WriteZeroPageXlogRec(int pageno);
static void WriteTruncateXlogRec(int pageno, TransactionId oldestXact,
@@ -99,7 +94,7 @@ static void TransactionIdSetPageStatus(TransactionId xid, int nsubxids,
XLogRecPtr lsn, int pageno,
bool all_xact_same_page);
static void TransactionIdSetStatusBit(TransactionId xid, XidStatus status,
- XLogRecPtr lsn, int slotno);
+ XLogRecPtr lsn, Buffer buffer);
static void set_status_by_pages(int nsubxids, TransactionId *subxids,
XidStatus status, XLogRecPtr lsn);
static bool TransactionGroupUpdateXidStatus(TransactionId xid,
@@ -341,13 +336,12 @@ TransactionIdSetPageStatusInternal(TransactionId xid, int nsubxids,
TransactionId *subxids, XidStatus status,
XLogRecPtr lsn, int pageno)
{
- int slotno;
+ Buffer buffer;
int i;
Assert(status == TRANSACTION_STATUS_COMMITTED ||
status == TRANSACTION_STATUS_ABORTED ||
(status == TRANSACTION_STATUS_SUB_COMMITTED && !TransactionIdIsValid(xid)));
- Assert(LWLockHeldByMeInMode(XactSLRULock, LW_EXCLUSIVE));
/*
* If we're doing an async commit (ie, lsn is valid), then we must wait
@@ -358,7 +352,8 @@ TransactionIdSetPageStatusInternal(TransactionId xid, int nsubxids,
* write-busy, since we don't care if the update reaches disk sooner than
* we think.
*/
- slotno = SimpleLruReadPage(XactCtl, pageno, XLogRecPtrIsInvalid(lsn), xid);
+ buffer = ReadSlruBuffer(SLRU_CLOG_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
/*
* Set the main transaction id, if any.
@@ -376,25 +371,26 @@ TransactionIdSetPageStatusInternal(TransactionId xid, int nsubxids,
{
for (i = 0; i < nsubxids; i++)
{
- Assert(XactCtl->shared->page_number[slotno] == TransactionIdToPage(subxids[i]));
+ Assert(pageno == TransactionIdToPage(subxids[i]));
TransactionIdSetStatusBit(subxids[i],
TRANSACTION_STATUS_SUB_COMMITTED,
- lsn, slotno);
+ lsn, buffer);
}
}
/* ... then the main transaction */
- TransactionIdSetStatusBit(xid, status, lsn, slotno);
+ TransactionIdSetStatusBit(xid, status, lsn, buffer);
}
/* Set the subtransactions */
for (i = 0; i < nsubxids; i++)
{
- Assert(XactCtl->shared->page_number[slotno] == TransactionIdToPage(subxids[i]));
- TransactionIdSetStatusBit(subxids[i], status, lsn, slotno);
+ Assert(pageno == TransactionIdToPage(subxids[i]));
+ TransactionIdSetStatusBit(subxids[i], status, lsn, buffer);
}
- XactCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -568,7 +564,7 @@ TransactionGroupUpdateXidStatus(TransactionId xid, XidStatus status,
* Must be called with XactSLRULock held
*/
static void
-TransactionIdSetStatusBit(TransactionId xid, XidStatus status, XLogRecPtr lsn, int slotno)
+TransactionIdSetStatusBit(TransactionId xid, XidStatus status, XLogRecPtr lsn, Buffer buffer)
{
int byteno = TransactionIdToByte(xid);
int bshift = TransactionIdToBIndex(xid) * CLOG_BITS_PER_XACT;
@@ -576,7 +572,10 @@ TransactionIdSetStatusBit(TransactionId xid, XidStatus status, XLogRecPtr lsn, i
char byteval;
char curval;
- byteptr = XactCtl->shared->page_buffer[slotno] + byteno;
+ Assert(LWLockHeldByMeInMode(BufferDescriptorGetContentLock(GetBufferDescriptor(buffer - 1)),
+ LW_EXCLUSIVE));
+
+ byteptr = BufferGetPage(buffer) + byteno;
curval = (*byteptr >> bshift) & CLOG_XACT_BITMASK;
/*
@@ -605,7 +604,7 @@ TransactionIdSetStatusBit(TransactionId xid, XidStatus status, XLogRecPtr lsn, i
*byteptr = byteval;
/*
- * Update the group LSN if the transaction completion LSN is higher.
+ * Update the buffer LSN if the transaction completion LSN is higher.
*
* Note: lsn will be invalid when supplied during InRecovery processing,
* so we don't need to do anything special to avoid LSN updates during
@@ -614,10 +613,8 @@ TransactionIdSetStatusBit(TransactionId xid, XidStatus status, XLogRecPtr lsn, i
*/
if (!XLogRecPtrIsInvalid(lsn))
{
- int lsnindex = GetLSNIndex(slotno, xid);
-
- if (XactCtl->shared->group_lsn[lsnindex] < lsn)
- XactCtl->shared->group_lsn[lsnindex] = lsn;
+ if (BufferGetExternalLSN(GetBufferDescriptor(buffer)) < lsn)
+ BufferSetExternalLSN(GetBufferDescriptor(buffer), lsn);
}
}
@@ -642,67 +639,22 @@ TransactionIdGetStatus(TransactionId xid, XLogRecPtr *lsn)
int pageno = TransactionIdToPage(xid);
int byteno = TransactionIdToByte(xid);
int bshift = TransactionIdToBIndex(xid) * CLOG_BITS_PER_XACT;
- int slotno;
- int lsnindex;
char *byteptr;
XidStatus status;
+ Buffer buffer;
- /* lock is acquired by SimpleLruReadPage_ReadOnly */
-
- slotno = SimpleLruReadPage_ReadOnly(XactCtl, pageno, xid);
- byteptr = XactCtl->shared->page_buffer[slotno] + byteno;
+ buffer = ReadSlruBuffer(SLRU_CLOG_ID, pageno);
+ byteptr = BufferGetPage(buffer) + byteno;
status = (*byteptr >> bshift) & CLOG_XACT_BITMASK;
- lsnindex = GetLSNIndex(slotno, xid);
- *lsn = XactCtl->shared->group_lsn[lsnindex];
+ *lsn = BufferGetExternalLSN(GetBufferDescriptor(buffer));
- LWLockRelease(XactSLRULock);
+ ReleaseBuffer(buffer);
return status;
}
-/*
- * Number of shared CLOG buffers.
- *
- * On larger multi-processor systems, it is possible to have many CLOG page
- * requests in flight at one time which could lead to disk access for CLOG
- * page if the required page is not found in memory. Testing revealed that we
- * can get the best performance by having 128 CLOG buffers, more than that it
- * doesn't improve performance.
- *
- * Unconditionally keeping the number of CLOG buffers to 128 did not seem like
- * a good idea, because it would increase the minimum amount of shared memory
- * required to start, which could be a problem for people running very small
- * configurations. The following formula seems to represent a reasonable
- * compromise: people with very low values for shared_buffers will get fewer
- * CLOG buffers as well, and everyone else will get 128.
- */
-Size
-CLOGShmemBuffers(void)
-{
- return Min(128, Max(4, NBuffers / 512));
-}
-
-/*
- * Initialization of shared memory for CLOG
- */
-Size
-CLOGShmemSize(void)
-{
- return SimpleLruShmemSize(CLOGShmemBuffers(), CLOG_LSNS_PER_PAGE);
-}
-
-void
-CLOGShmemInit(void)
-{
- XactCtl->PagePrecedes = CLOGPagePrecedes;
- SimpleLruInit(XactCtl, "Xact", CLOGShmemBuffers(), CLOG_LSNS_PER_PAGE,
- XactSLRULock, "pg_xact", LWTRANCHE_XACT_BUFFER,
- SYNC_HANDLER_CLOG);
- SlruPagePrecedesUnitTests(XactCtl, CLOG_XACTS_PER_PAGE);
-}
-
/*
* This func must be called ONCE on system install. It creates
* the initial CLOG segment. (The CLOG directory is assumed to
@@ -712,18 +664,15 @@ CLOGShmemInit(void)
void
BootStrapCLOG(void)
{
- int slotno;
-
- LWLockAcquire(XactSLRULock, LW_EXCLUSIVE);
+ Buffer buffer;
/* Create and zero the first page of the commit log */
- slotno = ZeroCLOGPage(0, false);
+ buffer = ZeroCLOGPage(0, false);
/* Make sure it's written out */
- SimpleLruWritePage(XactCtl, slotno);
- Assert(!XactCtl->shared->page_dirty[slotno]);
+ FlushOneBuffer(buffer);
- LWLockRelease(XactSLRULock);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -735,17 +684,18 @@ BootStrapCLOG(void)
*
* Control lock must be held at entry, and will be held at exit.
*/
-static int
+static Buffer
ZeroCLOGPage(int pageno, bool writeXlog)
{
- int slotno;
+ Buffer buffer;
- slotno = SimpleLruZeroPage(XactCtl, pageno);
+ buffer = ZeroSlruBuffer(SLRU_CLOG_ID, pageno);
+ MarkBufferDirty(buffer);
if (writeXlog)
WriteZeroPageXlogRec(pageno);
- return slotno;
+ return buffer;
}
/*
@@ -755,17 +705,6 @@ ZeroCLOGPage(int pageno, bool writeXlog)
void
StartupCLOG(void)
{
- TransactionId xid = XidFromFullTransactionId(ShmemVariableCache->nextXid);
- int pageno = TransactionIdToPage(xid);
-
- LWLockAcquire(XactSLRULock, LW_EXCLUSIVE);
-
- /*
- * Initialize our idea of the latest page number.
- */
- XactCtl->shared->latest_page_number = pageno;
-
- LWLockRelease(XactSLRULock);
}
/*
@@ -777,8 +716,6 @@ TrimCLOG(void)
TransactionId xid = XidFromFullTransactionId(ShmemVariableCache->nextXid);
int pageno = TransactionIdToPage(xid);
- LWLockAcquire(XactSLRULock, LW_EXCLUSIVE);
-
/*
* Zero out the remainder of the current clog page. Under normal
* circumstances it should be zeroes already, but it seems at least
@@ -795,40 +732,24 @@ TrimCLOG(void)
{
int byteno = TransactionIdToByte(xid);
int bshift = TransactionIdToBIndex(xid) * CLOG_BITS_PER_XACT;
- int slotno;
char *byteptr;
+ Buffer buffer;
- slotno = SimpleLruReadPage(XactCtl, pageno, false, xid);
- byteptr = XactCtl->shared->page_buffer[slotno] + byteno;
+ buffer = ReadSlruBuffer(SLRU_CLOG_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ byteptr = BufferGetPage(buffer) + byteno;
/* Zero so-far-unused positions in the current byte */
*byteptr &= (1 << bshift) - 1;
/* Zero the rest of the page */
MemSet(byteptr + 1, 0, BLCKSZ - byteno - 1);
- XactCtl->shared->page_dirty[slotno] = true;
- }
+ MarkBufferDirty(buffer);
- LWLockRelease(XactSLRULock);
-}
-
-/*
- * Perform a checkpoint --- either during shutdown, or on-the-fly
- */
-void
-CheckPointCLOG(void)
-{
- /*
- * Write dirty CLOG pages to disk. This may result in sync requests
- * queued for later handling by ProcessSyncRequests(), as part of the
- * checkpoint.
- */
- TRACE_POSTGRESQL_CLOG_CHECKPOINT_START(true);
- SimpleLruWriteAll(XactCtl, true);
- TRACE_POSTGRESQL_CLOG_CHECKPOINT_DONE(true);
+ UnlockReleaseBuffer(buffer);
+ }
}
-
/*
* Make sure that CLOG has room for a newly-allocated XID.
*
@@ -852,12 +773,8 @@ ExtendCLOG(TransactionId newestXact)
pageno = TransactionIdToPage(newestXact);
- LWLockAcquire(XactSLRULock, LW_EXCLUSIVE);
-
/* Zero the page and make an XLOG entry about it */
- ZeroCLOGPage(pageno, true);
-
- LWLockRelease(XactSLRULock);
+ UnlockReleaseBuffer(ZeroCLOGPage(pageno, true));
}
@@ -888,7 +805,8 @@ TruncateCLOG(TransactionId oldestXact, Oid oldestxid_datoid)
cutoffPage = TransactionIdToPage(oldestXact);
/* Check to see if there's any files that could be removed */
- if (!SlruScanDirectory(XactCtl, SlruScanDirCbReportPresence, &cutoffPage))
+ if (!SlruScanDirectory(SLRU_CLOG_ID, CLOGPagePrecedes,
+ SlruScanDirCbReportPresence, &cutoffPage))
return; /* nothing to remove */
/*
@@ -909,7 +827,7 @@ TruncateCLOG(TransactionId oldestXact, Oid oldestxid_datoid)
WriteTruncateXlogRec(cutoffPage, oldestXact, oldestxid_datoid);
/* Now we can remove the old CLOG segment(s) */
- SimpleLruTruncate(XactCtl, cutoffPage);
+ SimpleLruTruncate(SLRU_CLOG_ID, CLOGPagePrecedes, cutoffPage);
}
@@ -994,17 +912,13 @@ clog_redo(XLogReaderState *record)
if (info == CLOG_ZEROPAGE)
{
int pageno;
- int slotno;
+ Buffer buffer;
memcpy(&pageno, XLogRecGetData(record), sizeof(int));
- LWLockAcquire(XactSLRULock, LW_EXCLUSIVE);
-
- slotno = ZeroCLOGPage(pageno, false);
- SimpleLruWritePage(XactCtl, slotno);
- Assert(!XactCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(XactSLRULock);
+ buffer = ZeroCLOGPage(pageno, false);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
else if (info == CLOG_TRUNCATE)
{
@@ -1014,17 +928,8 @@ clog_redo(XLogReaderState *record)
AdvanceOldestClogXid(xlrec.oldestXact);
- SimpleLruTruncate(XactCtl, xlrec.pageno);
+ SimpleLruTruncate(SLRU_CLOG_ID, CLOGPagePrecedes, xlrec.pageno);
}
else
elog(PANIC, "clog_redo: unknown op code %u", info);
}
-
-/*
- * Entrypoint for sync.c to sync clog files.
- */
-int
-clogsyncfiletag(const FileTag *ftag, char *path)
-{
- return SlruSyncFileTag(XactCtl, ftag, path);
-}
diff --git a/src/backend/access/transam/commit_ts.c b/src/backend/access/transam/commit_ts.c
index 4dc8d402bd3..7eded0d1921 100644
--- a/src/backend/access/transam/commit_ts.c
+++ b/src/backend/access/transam/commit_ts.c
@@ -73,13 +73,6 @@ typedef struct CommitTimestampEntry
#define TransactionIdToCTsEntry(xid) \
((xid) % (TransactionId) COMMIT_TS_XACTS_PER_PAGE)
-/*
- * Link to shared-memory data structures for CommitTs control
- */
-static SlruCtlData CommitTsCtlData;
-
-#define CommitTsCtl (&CommitTsCtlData)
-
/*
* We keep a cache of the last value set in shared memory.
*
@@ -110,7 +103,7 @@ static void SetXidCommitTsInPage(TransactionId xid, int nsubxids,
static void TransactionIdSetCommitTs(TransactionId xid, TimestampTz ts,
RepOriginId nodeid, int slotno);
static void error_commit_ts_disabled(void);
-static int ZeroCommitTsPage(int pageno, bool writeXlog);
+static Buffer ZeroCommitTsPage(int pageno, bool writeXlog);
static bool CommitTsPagePrecedes(int page1, int page2);
static void ActivateCommitTs(void);
static void DeactivateCommitTs(void);
@@ -219,30 +212,27 @@ SetXidCommitTsInPage(TransactionId xid, int nsubxids,
TransactionId *subxids, TimestampTz ts,
RepOriginId nodeid, int pageno)
{
- int slotno;
int i;
+ Buffer buffer;
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
-
- slotno = SimpleLruReadPage(CommitTsCtl, pageno, true, xid);
+ buffer = ReadSlruBuffer(SLRU_COMMIT_TS_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
- TransactionIdSetCommitTs(xid, ts, nodeid, slotno);
+ TransactionIdSetCommitTs(xid, ts, nodeid, buffer);
for (i = 0; i < nsubxids; i++)
- TransactionIdSetCommitTs(subxids[i], ts, nodeid, slotno);
+ TransactionIdSetCommitTs(subxids[i], ts, nodeid, buffer);
- CommitTsCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
- LWLockRelease(CommitTsSLRULock);
+ UnlockReleaseBuffer(buffer);
}
/*
* Sets the commit timestamp of a single transaction.
- *
- * Must be called with CommitTsSLRULock held
*/
static void
TransactionIdSetCommitTs(TransactionId xid, TimestampTz ts,
- RepOriginId nodeid, int slotno)
+ RepOriginId nodeid, Buffer buffer)
{
int entryno = TransactionIdToCTsEntry(xid);
CommitTimestampEntry entry;
@@ -252,8 +242,7 @@ TransactionIdSetCommitTs(TransactionId xid, TimestampTz ts,
entry.time = ts;
entry.nodeid = nodeid;
- memcpy(CommitTsCtl->shared->page_buffer[slotno] +
- SizeOfCommitTimestampEntry * entryno,
+ memcpy(BufferGetPage(buffer) + SizeOfCommitTimestampEntry * entryno,
&entry, SizeOfCommitTimestampEntry);
}
@@ -271,10 +260,10 @@ TransactionIdGetCommitTsData(TransactionId xid, TimestampTz *ts,
{
int pageno = TransactionIdToCTsPage(xid);
int entryno = TransactionIdToCTsEntry(xid);
- int slotno;
CommitTimestampEntry entry;
TransactionId oldestCommitTsXid;
TransactionId newestCommitTsXid;
+ Buffer buffer;
if (!TransactionIdIsValid(xid))
ereport(ERROR,
@@ -328,10 +317,11 @@ TransactionIdGetCommitTsData(TransactionId xid, TimestampTz *ts,
return false;
}
- /* lock is acquired by SimpleLruReadPage_ReadOnly */
- slotno = SimpleLruReadPage_ReadOnly(CommitTsCtl, pageno, xid);
+ buffer = ReadSlruBuffer(SLRU_COMMIT_TS_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_SHARE);
+
memcpy(&entry,
- CommitTsCtl->shared->page_buffer[slotno] +
+ BufferGetPage(buffer) +
SizeOfCommitTimestampEntry * entryno,
SizeOfCommitTimestampEntry);
@@ -339,7 +329,7 @@ TransactionIdGetCommitTsData(TransactionId xid, TimestampTz *ts,
if (nodeid)
*nodeid = entry.nodeid;
- LWLockRelease(CommitTsSLRULock);
+ UnlockReleaseBuffer(buffer);
return *ts != 0;
}
@@ -508,27 +498,13 @@ pg_xact_commit_timestamp_origin(PG_FUNCTION_ARGS)
PG_RETURN_DATUM(HeapTupleGetDatum(htup));
}
-/*
- * Number of shared CommitTS buffers.
- *
- * We use a very similar logic as for the number of CLOG buffers (except we
- * scale up twice as fast with shared buffers, and the maximum is twice as
- * high); see comments in CLOGShmemBuffers.
- */
-Size
-CommitTsShmemBuffers(void)
-{
- return Min(256, Max(4, NBuffers / 256));
-}
-
/*
* Shared memory sizing for CommitTs
*/
Size
CommitTsShmemSize(void)
{
- return SimpleLruShmemSize(CommitTsShmemBuffers(), 0) +
- sizeof(CommitTimestampShared);
+ return sizeof(CommitTimestampShared);
}
/*
@@ -540,12 +516,7 @@ CommitTsShmemInit(void)
{
bool found;
- CommitTsCtl->PagePrecedes = CommitTsPagePrecedes;
- SimpleLruInit(CommitTsCtl, "CommitTs", CommitTsShmemBuffers(), 0,
- CommitTsSLRULock, "pg_commit_ts",
- LWTRANCHE_COMMITTS_BUFFER,
- SYNC_HANDLER_COMMIT_TS);
- SlruPagePrecedesUnitTests(CommitTsCtl, COMMIT_TS_XACTS_PER_PAGE);
+ SlruPagePrecedesUnitTests(CommitTsPagePrecedes, COMMIT_TS_XACTS_PER_PAGE);
commitTsShared = ShmemInitStruct("CommitTs shared",
sizeof(CommitTimestampShared),
@@ -589,17 +560,18 @@ BootStrapCommitTs(void)
*
* Control lock must be held at entry, and will be held at exit.
*/
-static int
+static Buffer
ZeroCommitTsPage(int pageno, bool writeXlog)
{
- int slotno;
+ Buffer buffer;
- slotno = SimpleLruZeroPage(CommitTsCtl, pageno);
+ buffer = ZeroSlruBuffer(SLRU_COMMIT_TS_ID, pageno);
+ MarkBufferDirty(buffer);
if (writeXlog)
WriteZeroPageXlogRec(pageno);
- return slotno;
+ return buffer;
}
/*
@@ -697,13 +669,6 @@ ActivateCommitTs(void)
xid = XidFromFullTransactionId(ShmemVariableCache->nextXid);
pageno = TransactionIdToCTsPage(xid);
- /*
- * Re-Initialize our idea of the latest page number.
- */
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
- CommitTsCtl->shared->latest_page_number = pageno;
- LWLockRelease(CommitTsSLRULock);
-
/*
* If CommitTs is enabled, but it wasn't in the previous server run, we
* need to set the oldest and newest values to the next Xid; that way, we
@@ -726,15 +691,14 @@ ActivateCommitTs(void)
LWLockRelease(CommitTsLock);
/* Create the current segment file, if necessary */
- if (!SimpleLruDoesPhysicalPageExist(CommitTsCtl, pageno))
+ if (!SimpleLruDoesPhysicalPageExist(SLRU_COMMIT_TS_ID, pageno))
{
- int slotno;
+ Buffer buffer;
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
- slotno = ZeroCommitTsPage(pageno, false);
- SimpleLruWritePage(CommitTsCtl, slotno);
- Assert(!CommitTsCtl->shared->page_dirty[slotno]);
- LWLockRelease(CommitTsSLRULock);
+ buffer = ZeroSlruBuffer(SLRU_COMMIT_TS_ID, pageno);
+ MarkBufferDirty(buffer);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
/* Change the activation status in shared memory. */
@@ -783,23 +747,9 @@ DeactivateCommitTs(void)
* be overwritten anyway when we wrap around, but it seems better to be
* tidy.)
*/
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
- (void) SlruScanDirectory(CommitTsCtl, SlruScanDirCbDeleteAll, NULL);
- LWLockRelease(CommitTsSLRULock);
-}
-
-/*
- * Perform a checkpoint --- either during shutdown, or on-the-fly
- */
-void
-CheckPointCommitTs(void)
-{
- /*
- * Write dirty CommitTs pages to disk. This may result in sync requests
- * queued for later handling by ProcessSyncRequests(), as part of the
- * checkpoint.
- */
- SimpleLruWriteAll(CommitTsCtl, true);
+ (void) SlruScanDirectory(SLRU_COMMIT_TS_ID,
+ CommitTsPagePrecedes,
+ SlruScanDirCbDeleteAll, NULL);
}
/*
@@ -837,12 +787,8 @@ ExtendCommitTs(TransactionId newestXact)
pageno = TransactionIdToCTsPage(newestXact);
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
-
/* Zero the page and make an XLOG entry about it */
- ZeroCommitTsPage(pageno, !InRecovery);
-
- LWLockRelease(CommitTsSLRULock);
+ UnlockReleaseBuffer(ZeroCommitTsPage(pageno, !InRecovery));
}
/*
@@ -863,7 +809,9 @@ TruncateCommitTs(TransactionId oldestXact)
cutoffPage = TransactionIdToCTsPage(oldestXact);
/* Check to see if there's any files that could be removed */
- if (!SlruScanDirectory(CommitTsCtl, SlruScanDirCbReportPresence,
+ if (!SlruScanDirectory(SLRU_COMMIT_TS_ID,
+ CommitTsPagePrecedes,
+ SlruScanDirCbReportPresence,
&cutoffPage))
return; /* nothing to remove */
@@ -871,7 +819,7 @@ TruncateCommitTs(TransactionId oldestXact)
WriteTruncateXlogRec(cutoffPage, oldestXact);
/* Now we can remove the old CommitTs segment(s) */
- SimpleLruTruncate(CommitTsCtl, cutoffPage);
+ SimpleLruTruncate(SLRU_COMMIT_TS_ID, CommitTsPagePrecedes, cutoffPage);
}
/*
@@ -995,17 +943,14 @@ commit_ts_redo(XLogReaderState *record)
if (info == COMMIT_TS_ZEROPAGE)
{
int pageno;
- int slotno;
+ Buffer buffer;
memcpy(&pageno, XLogRecGetData(record), sizeof(int));
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
-
- slotno = ZeroCommitTsPage(pageno, false);
- SimpleLruWritePage(CommitTsCtl, slotno);
- Assert(!CommitTsCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(CommitTsSLRULock);
+ buffer = ZeroSlruBuffer(SLRU_COMMIT_TS_ID, pageno);
+ MarkBufferDirty(buffer);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
else if (info == COMMIT_TS_TRUNCATE)
{
@@ -1013,23 +958,8 @@ commit_ts_redo(XLogReaderState *record)
AdvanceOldestCommitTsXid(trunc->oldestXid);
- /*
- * During XLOG replay, latest_page_number isn't set up yet; insert a
- * suitable value to bypass the sanity test in SimpleLruTruncate.
- */
- CommitTsCtl->shared->latest_page_number = trunc->pageno;
-
- SimpleLruTruncate(CommitTsCtl, trunc->pageno);
+ SimpleLruTruncate(SLRU_COMMIT_TS_ID, CommitTsPagePrecedes, trunc->pageno);
}
else
elog(PANIC, "commit_ts_redo: unknown op code %u", info);
}
-
-/*
- * Entrypoint for sync.c to sync commit_ts files.
- */
-int
-committssyncfiletag(const FileTag *ftag, char *path)
-{
- return SlruSyncFileTag(CommitTsCtl, ftag, path);
-}
diff --git a/src/backend/access/transam/multixact.c b/src/backend/access/transam/multixact.c
index 8f7d12950e5..aabe8b53452 100644
--- a/src/backend/access/transam/multixact.c
+++ b/src/backend/access/transam/multixact.c
@@ -181,15 +181,6 @@
#define PreviousMultiXactId(xid) \
((xid) == FirstMultiXactId ? MaxMultiXactId : (xid) - 1)
-/*
- * Links to shared-memory data structures for MultiXact control
- */
-static SlruCtlData MultiXactOffsetCtlData;
-static SlruCtlData MultiXactMemberCtlData;
-
-#define MultiXactOffsetCtl (&MultiXactOffsetCtlData)
-#define MultiXactMemberCtl (&MultiXactMemberCtlData)
-
/*
* MultiXact state shared across all backends. All this state is protected
* by MultiXactGenLock. (We also use MultiXactOffsetSLRULock and
@@ -353,10 +344,9 @@ static void mXactCachePut(MultiXactId multi, int nmembers,
static char *mxstatus_to_string(MultiXactStatus status);
/* management of SLRU infrastructure */
-static int ZeroMultiXactOffsetPage(int pageno, bool writeXlog);
-static int ZeroMultiXactMemberPage(int pageno, bool writeXlog);
+static Buffer ZeroMultiXactOffsetPage(int pageno, bool writeXlog);
+static Buffer ZeroMultiXactMemberPage(int pageno, bool writeXlog);
static bool MultiXactOffsetPagePrecedes(int page1, int page2);
-static bool MultiXactMemberPagePrecedes(int page1, int page2);
static bool MultiXactOffsetPrecedes(MultiXactOffset offset1,
MultiXactOffset offset2);
static void ExtendMultiXactOffset(MultiXactId multi);
@@ -865,34 +855,25 @@ RecordNewMultiXact(MultiXactId multi, MultiXactOffset offset,
int pageno;
int prev_pageno;
int entryno;
- int slotno;
MultiXactOffset *offptr;
int i;
-
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
+ Buffer buffer;
pageno = MultiXactIdToOffsetPage(multi);
entryno = MultiXactIdToOffsetEntry(multi);
- /*
- * Note: we pass the MultiXactId to SimpleLruReadPage as the "transaction"
- * to complain about if there's any I/O error. This is kinda bogus, but
- * since the errors will always give the full pathname, it should be clear
- * enough that a MultiXactId is really involved. Perhaps someday we'll
- * take the trouble to generalize the slru.c error reporting code.
- */
- slotno = SimpleLruReadPage(MultiXactOffsetCtl, pageno, true, multi);
- offptr = (MultiXactOffset *) MultiXactOffsetCtl->shared->page_buffer[slotno];
+ /* XXX set up error context? */
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_OFFSET_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ offptr = (MultiXactOffset *) BufferGetPage(buffer);
offptr += entryno;
*offptr = offset;
- MultiXactOffsetCtl->shared->page_dirty[slotno] = true;
-
- /* Exchange our lock */
- LWLockRelease(MultiXactOffsetSLRULock);
+ MarkBufferDirty(buffer);
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
+ UnlockReleaseBuffer(buffer);
+ buffer = InvalidBuffer;
prev_pageno = -1;
@@ -914,27 +895,28 @@ RecordNewMultiXact(MultiXactId multi, MultiXactOffset offset,
if (pageno != prev_pageno)
{
- slotno = SimpleLruReadPage(MultiXactMemberCtl, pageno, true, multi);
+ if (BufferIsValid(buffer))
+ UnlockReleaseBuffer(buffer);
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_MEMBER_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
prev_pageno = pageno;
}
- memberptr = (TransactionId *)
- (MultiXactMemberCtl->shared->page_buffer[slotno] + memberoff);
+ memberptr = (TransactionId *) (BufferGetPage(buffer) + memberoff);
*memberptr = members[i].xid;
- flagsptr = (uint32 *)
- (MultiXactMemberCtl->shared->page_buffer[slotno] + flagsoff);
+ flagsptr = (uint32 *) (BufferGetPage(buffer) + flagsoff);
flagsval = *flagsptr;
flagsval &= ~(((1 << MXACT_MEMBER_BITS_PER_XACT) - 1) << bshift);
flagsval |= (members[i].status << bshift);
*flagsptr = flagsval;
- MultiXactMemberCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
}
- LWLockRelease(MultiXactMemberSLRULock);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -1226,7 +1208,6 @@ GetMultiXactIdMembers(MultiXactId multi, MultiXactMember **members,
int pageno;
int prev_pageno;
int entryno;
- int slotno;
MultiXactOffset *offptr;
MultiXactOffset offset;
int length;
@@ -1237,6 +1218,7 @@ GetMultiXactIdMembers(MultiXactId multi, MultiXactMember **members,
MultiXactId tmpMXact;
MultiXactOffset nextOffset;
MultiXactMember *ptr;
+ Buffer buffer;
debug_elog3(DEBUG2, "GetMembers: asked for %u", multi);
@@ -1340,13 +1322,12 @@ GetMultiXactIdMembers(MultiXactId multi, MultiXactMember **members,
* time on every multixact creation.
*/
retry:
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
-
pageno = MultiXactIdToOffsetPage(multi);
entryno = MultiXactIdToOffsetEntry(multi);
- slotno = SimpleLruReadPage(MultiXactOffsetCtl, pageno, true, multi);
- offptr = (MultiXactOffset *) MultiXactOffsetCtl->shared->page_buffer[slotno];
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_OFFSET_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_SHARE);
+ offptr = (MultiXactOffset *) BufferGetPage(buffer);
offptr += entryno;
offset = *offptr;
@@ -1377,16 +1358,20 @@ retry:
entryno = MultiXactIdToOffsetEntry(tmpMXact);
if (pageno != prev_pageno)
- slotno = SimpleLruReadPage(MultiXactOffsetCtl, pageno, true, tmpMXact);
+ {
+ UnlockReleaseBuffer(buffer);
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_OFFSET_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_SHARE);
+ }
- offptr = (MultiXactOffset *) MultiXactOffsetCtl->shared->page_buffer[slotno];
+ offptr = (MultiXactOffset *) BufferGetPage(buffer);
offptr += entryno;
nextMXOffset = *offptr;
if (nextMXOffset == 0)
{
/* Corner case 2: next multixact is still being filled in */
- LWLockRelease(MultiXactOffsetSLRULock);
+ UnlockReleaseBuffer(buffer);
CHECK_FOR_INTERRUPTS();
pg_usleep(1000L);
goto retry;
@@ -1394,14 +1379,11 @@ retry:
length = nextMXOffset - offset;
}
-
- LWLockRelease(MultiXactOffsetSLRULock);
+ UnlockReleaseBuffer(buffer);
+ buffer = InvalidBuffer;
ptr = (MultiXactMember *) palloc(length * sizeof(MultiXactMember));
- /* Now get the members themselves. */
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
-
truelength = 0;
prev_pageno = -1;
for (i = 0; i < length; i++, offset++)
@@ -1417,12 +1399,14 @@ retry:
if (pageno != prev_pageno)
{
- slotno = SimpleLruReadPage(MultiXactMemberCtl, pageno, true, multi);
+ if (BufferIsValid(buffer))
+ UnlockReleaseBuffer(buffer);
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_MEMBER_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_SHARE);
prev_pageno = pageno;
}
- xactptr = (TransactionId *)
- (MultiXactMemberCtl->shared->page_buffer[slotno] + memberoff);
+ xactptr = (TransactionId *) (BufferGetPage(buffer) + memberoff);
if (!TransactionIdIsValid(*xactptr))
{
@@ -1433,14 +1417,13 @@ retry:
flagsoff = MXOffsetToFlagsOffset(offset);
bshift = MXOffsetToFlagsBitShift(offset);
- flagsptr = (uint32 *) (MultiXactMemberCtl->shared->page_buffer[slotno] + flagsoff);
+ flagsptr = (uint32 *) (BufferGetPage(buffer) + flagsoff);
ptr[truelength].xid = *xactptr;
ptr[truelength].status = (*flagsptr >> bshift) & MXACT_MEMBER_XACT_BITMASK;
truelength++;
}
-
- LWLockRelease(MultiXactMemberSLRULock);
+ UnlockReleaseBuffer(buffer);
/* A multixid with zero members should not happen */
Assert(truelength > 0);
@@ -1834,8 +1817,6 @@ MultiXactShmemSize(void)
mul_size(sizeof(MultiXactId) * 2, MaxOldestSlot))
size = SHARED_MULTIXACT_STATE_SIZE;
- size = add_size(size, SimpleLruShmemSize(NUM_MULTIXACTOFFSET_BUFFERS, 0));
- size = add_size(size, SimpleLruShmemSize(NUM_MULTIXACTMEMBER_BUFFERS, 0));
return size;
}
@@ -1847,22 +1828,6 @@ MultiXactShmemInit(void)
debug_elog2(DEBUG2, "Shared Memory Init for MultiXact");
- MultiXactOffsetCtl->PagePrecedes = MultiXactOffsetPagePrecedes;
- MultiXactMemberCtl->PagePrecedes = MultiXactMemberPagePrecedes;
-
- SimpleLruInit(MultiXactOffsetCtl,
- "MultiXactOffset", NUM_MULTIXACTOFFSET_BUFFERS, 0,
- MultiXactOffsetSLRULock, "pg_multixact/offsets",
- LWTRANCHE_MULTIXACTOFFSET_BUFFER,
- SYNC_HANDLER_MULTIXACT_OFFSET);
- SlruPagePrecedesUnitTests(MultiXactOffsetCtl, MULTIXACT_OFFSETS_PER_PAGE);
- SimpleLruInit(MultiXactMemberCtl,
- "MultiXactMember", NUM_MULTIXACTMEMBER_BUFFERS, 0,
- MultiXactMemberSLRULock, "pg_multixact/members",
- LWTRANCHE_MULTIXACTMEMBER_BUFFER,
- SYNC_HANDLER_MULTIXACT_MEMBER);
- /* doesn't call SimpleLruTruncate() or meet criteria for unit tests */
-
/* Initialize our shared state struct */
MultiXactState = ShmemInitStruct("Shared MultiXact State",
SHARED_MULTIXACT_STATE_SIZE,
@@ -1893,29 +1858,17 @@ MultiXactShmemInit(void)
void
BootStrapMultiXact(void)
{
- int slotno;
-
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
+ Buffer buffer;
/* Create and zero the first page of the offsets log */
- slotno = ZeroMultiXactOffsetPage(0, false);
-
- /* Make sure it's written out */
- SimpleLruWritePage(MultiXactOffsetCtl, slotno);
- Assert(!MultiXactOffsetCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(MultiXactOffsetSLRULock);
-
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
+ buffer = ZeroMultiXactOffsetPage(0, false);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
/* Create and zero the first page of the members log */
- slotno = ZeroMultiXactMemberPage(0, false);
-
- /* Make sure it's written out */
- SimpleLruWritePage(MultiXactMemberCtl, slotno);
- Assert(!MultiXactMemberCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(MultiXactMemberSLRULock);
+ buffer = ZeroMultiXactMemberPage(0, false);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -1927,33 +1880,35 @@ BootStrapMultiXact(void)
*
* Control lock must be held at entry, and will be held at exit.
*/
-static int
+static Buffer
ZeroMultiXactOffsetPage(int pageno, bool writeXlog)
{
- int slotno;
+ Buffer buffer;
- slotno = SimpleLruZeroPage(MultiXactOffsetCtl, pageno);
+ buffer = ZeroSlruBuffer(SLRU_MULTIXACT_OFFSET_ID, pageno);
+ MarkBufferDirty(buffer);
if (writeXlog)
WriteMZeroPageXlogRec(pageno, XLOG_MULTIXACT_ZERO_OFF_PAGE);
- return slotno;
+ return buffer;
}
/*
* Ditto, for MultiXactMember
*/
-static int
+static Buffer
ZeroMultiXactMemberPage(int pageno, bool writeXlog)
{
- int slotno;
+ Buffer buffer;
- slotno = SimpleLruZeroPage(MultiXactMemberCtl, pageno);
+ buffer = ZeroSlruBuffer(SLRU_MULTIXACT_MEMBER_ID, pageno);
+ MarkBufferDirty(buffer);
if (writeXlog)
WriteMZeroPageXlogRec(pageno, XLOG_MULTIXACT_ZERO_MEM_PAGE);
- return slotno;
+ return buffer;
}
/*
@@ -1978,22 +1933,14 @@ MaybeExtendOffsetSlru(void)
pageno = MultiXactIdToOffsetPage(MultiXactState->nextMXact);
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
-
- if (!SimpleLruDoesPhysicalPageExist(MultiXactOffsetCtl, pageno))
+ if (!SimpleLruDoesPhysicalPageExist(SLRU_MULTIXACT_OFFSET_ID, pageno))
{
- int slotno;
+ Buffer buffer;
- /*
- * Fortunately for us, SimpleLruWritePage is already prepared to deal
- * with creating a new segment file even if the page we're writing is
- * not the first in it, so this is enough.
- */
- slotno = ZeroMultiXactOffsetPage(pageno, false);
- SimpleLruWritePage(MultiXactOffsetCtl, slotno);
+ buffer = ZeroMultiXactOffsetPage(pageno, false);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
-
- LWLockRelease(MultiXactOffsetSLRULock);
}
/*
@@ -2007,21 +1954,6 @@ MaybeExtendOffsetSlru(void)
void
StartupMultiXact(void)
{
- MultiXactId multi = MultiXactState->nextMXact;
- MultiXactOffset offset = MultiXactState->nextOffset;
- int pageno;
-
- /*
- * Initialize offset's idea of the latest page number.
- */
- pageno = MultiXactIdToOffsetPage(multi);
- MultiXactOffsetCtl->shared->latest_page_number = pageno;
-
- /*
- * Initialize member's idea of the latest page number.
- */
- pageno = MXOffsetToMemberPage(offset);
- MultiXactMemberCtl->shared->latest_page_number = pageno;
}
/*
@@ -2045,14 +1977,7 @@ TrimMultiXact(void)
oldestMXactDB = MultiXactState->oldestMultiXactDB;
LWLockRelease(MultiXactGenLock);
- /* Clean up offsets state */
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
-
- /*
- * (Re-)Initialize our idea of the latest page number for offsets.
- */
- pageno = MultiXactIdToOffsetPage(nextMXact);
- MultiXactOffsetCtl->shared->latest_page_number = pageno;
+ pageno = MXOffsetToMemberPage(offset);
/*
* Zero out the remainder of the current offsets page. See notes in
@@ -2065,29 +1990,20 @@ TrimMultiXact(void)
entryno = MultiXactIdToOffsetEntry(nextMXact);
if (entryno != 0)
{
- int slotno;
MultiXactOffset *offptr;
+ Buffer buffer;
- slotno = SimpleLruReadPage(MultiXactOffsetCtl, pageno, true, nextMXact);
- offptr = (MultiXactOffset *) MultiXactOffsetCtl->shared->page_buffer[slotno];
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_OFFSET_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ offptr = (MultiXactOffset *) BufferGetPage(buffer);
offptr += entryno;
MemSet(offptr, 0, BLCKSZ - (entryno * sizeof(MultiXactOffset)));
- MultiXactOffsetCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
+ UnlockReleaseBuffer(buffer);
}
- LWLockRelease(MultiXactOffsetSLRULock);
-
- /* And the same for members */
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
-
- /*
- * (Re-)Initialize our idea of the latest page number for members.
- */
- pageno = MXOffsetToMemberPage(offset);
- MultiXactMemberCtl->shared->latest_page_number = pageno;
-
/*
* Zero out the remainder of the current members page. See notes in
* TrimCLOG() for motivation.
@@ -2095,14 +2011,14 @@ TrimMultiXact(void)
flagsoff = MXOffsetToFlagsOffset(offset);
if (flagsoff != 0)
{
- int slotno;
TransactionId *xidptr;
int memberoff;
+ Buffer buffer;
memberoff = MXOffsetToMemberOffset(offset);
- slotno = SimpleLruReadPage(MultiXactMemberCtl, pageno, true, offset);
- xidptr = (TransactionId *)
- (MultiXactMemberCtl->shared->page_buffer[slotno] + memberoff);
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_MEMBER_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ xidptr = (TransactionId *) (BufferGetPage(buffer) + memberoff);
MemSet(xidptr, 0, BLCKSZ - memberoff);
@@ -2112,11 +2028,10 @@ TrimMultiXact(void)
* writing.
*/
- MultiXactMemberCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
+ UnlockReleaseBuffer(buffer);
}
- LWLockRelease(MultiXactMemberSLRULock);
-
/* signal that we're officially up */
LWLockAcquire(MultiXactGenLock, LW_EXCLUSIVE);
MultiXactState->finishedStartup = true;
@@ -2148,25 +2063,6 @@ MultiXactGetCheckptMulti(bool is_shutdown,
*nextMulti, *nextMultiOffset, *oldestMulti, *oldestMultiDB);
}
-/*
- * Perform a checkpoint --- either during shutdown, or on-the-fly
- */
-void
-CheckPointMultiXact(void)
-{
- TRACE_POSTGRESQL_MULTIXACT_CHECKPOINT_START(true);
-
- /*
- * Write dirty MultiXact pages to disk. This may result in sync requests
- * queued for later handling by ProcessSyncRequests(), as part of the
- * checkpoint.
- */
- SimpleLruWriteAll(MultiXactOffsetCtl, true);
- SimpleLruWriteAll(MultiXactMemberCtl, true);
-
- TRACE_POSTGRESQL_MULTIXACT_CHECKPOINT_DONE(true);
-}
-
/*
* Set the next-to-be-assigned MultiXactId and offset
*
@@ -2415,12 +2311,8 @@ ExtendMultiXactOffset(MultiXactId multi)
pageno = MultiXactIdToOffsetPage(multi);
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
-
/* Zero the page and make an XLOG entry about it */
- ZeroMultiXactOffsetPage(pageno, true);
-
- LWLockRelease(MultiXactOffsetSLRULock);
+ UnlockReleaseBuffer(ZeroMultiXactOffsetPage(pageno, true));
}
/*
@@ -2456,12 +2348,8 @@ ExtendMultiXactMember(MultiXactOffset offset, int nmembers)
pageno = MXOffsetToMemberPage(offset);
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
-
/* Zero the page and make an XLOG entry about it */
- ZeroMultiXactMemberPage(pageno, true);
-
- LWLockRelease(MultiXactMemberSLRULock);
+ UnlockReleaseBuffer(ZeroMultiXactMemberPage(pageno, true));
}
/*
@@ -2736,8 +2624,8 @@ find_multixact_start(MultiXactId multi, MultiXactOffset *result)
MultiXactOffset offset;
int pageno;
int entryno;
- int slotno;
MultiXactOffset *offptr;
+ Buffer buffer;
Assert(MultiXactState->finishedStartup);
@@ -2745,20 +2633,19 @@ find_multixact_start(MultiXactId multi, MultiXactOffset *result)
entryno = MultiXactIdToOffsetEntry(multi);
/*
- * Write out dirty data, so PhysicalPageExists can work correctly.
+ * Cope with missing/bogus oldest MultiXact in inconsistent states (see
+ * commit 068cfadf9).
*/
- SimpleLruWriteAll(MultiXactOffsetCtl, true);
- SimpleLruWriteAll(MultiXactMemberCtl, true);
-
- if (!SimpleLruDoesPhysicalPageExist(MultiXactOffsetCtl, pageno))
+ if (!ProbeSlruBuffer(SLRU_MULTIXACT_OFFSET_ID, pageno) &&
+ !SimpleLruDoesPhysicalPageExist(SLRU_MULTIXACT_OFFSET_ID, pageno))
return false;
- /* lock is acquired by SimpleLruReadPage_ReadOnly */
- slotno = SimpleLruReadPage_ReadOnly(MultiXactOffsetCtl, pageno, multi);
- offptr = (MultiXactOffset *) MultiXactOffsetCtl->shared->page_buffer[slotno];
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_OFFSET_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_SHARE);
+ offptr = (MultiXactOffset *) BufferGetPage(buffer);
offptr += entryno;
offset = *offptr;
- LWLockRelease(MultiXactOffsetSLRULock);
+ UnlockReleaseBuffer(buffer);
*result = offset;
return true;
@@ -2865,12 +2752,13 @@ typedef struct mxtruncinfo
* This callback determines the earliest existing page number.
*/
static bool
-SlruScanDirCbFindEarliest(SlruCtl ctl, char *filename, int segpage, void *data)
+SlruScanDirCbFindEarliest(int slru_id, SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage, void *data)
{
mxtruncinfo *trunc = (mxtruncinfo *) data;
if (trunc->earliestExistingPage == -1 ||
- ctl->PagePrecedes(segpage, trunc->earliestExistingPage))
+ PagePrecedes(segpage, trunc->earliestExistingPage))
{
trunc->earliestExistingPage = segpage;
}
@@ -2902,7 +2790,7 @@ PerformMembersTruncation(MultiXactOffset oldestOffset, MultiXactOffset newOldest
while (segment != endsegment)
{
elog(DEBUG2, "truncating multixact members segment %x", segment);
- SlruDeleteSegment(MultiXactMemberCtl, segment);
+ SlruDeleteSegment(SLRU_MULTIXACT_MEMBER_ID, segment);
/* move to next segment, handling wraparound correctly */
if (segment == maxsegment)
@@ -2925,7 +2813,8 @@ PerformOffsetsTruncation(MultiXactId oldestMulti, MultiXactId newOldestMulti)
* didn't subtract one, we'd trigger SimpleLruTruncate's wraparound
* detection.
*/
- SimpleLruTruncate(MultiXactOffsetCtl,
+ SimpleLruTruncate(SLRU_MULTIXACT_OFFSET_ID,
+ MultiXactOffsetPagePrecedes,
MultiXactIdToOffsetPage(PreviousMultiXactId(newOldestMulti)));
}
@@ -2999,7 +2888,9 @@ TruncateMultiXact(MultiXactId newOldestMulti, Oid newOldestMultiDB)
* been truncated away, and we crashed before updating oldestMulti.
*/
trunc.earliestExistingPage = -1;
- SlruScanDirectory(MultiXactOffsetCtl, SlruScanDirCbFindEarliest, &trunc);
+ SlruScanDirectory(SLRU_MULTIXACT_OFFSET_ID,
+ MultiXactOffsetPagePrecedes,
+ SlruScanDirCbFindEarliest, &trunc);
earliest = trunc.earliestExistingPage * MULTIXACT_OFFSETS_PER_PAGE;
if (earliest < FirstMultiXactId)
earliest = FirstMultiXactId;
@@ -3131,24 +3022,6 @@ MultiXactOffsetPagePrecedes(int page1, int page2)
multi2 + MULTIXACT_OFFSETS_PER_PAGE - 1));
}
-/*
- * Decide whether a MultiXactMember page number is "older" for truncation
- * purposes. There is no "invalid offset number" so use the numbers verbatim.
- */
-static bool
-MultiXactMemberPagePrecedes(int page1, int page2)
-{
- MultiXactOffset offset1;
- MultiXactOffset offset2;
-
- offset1 = ((MultiXactOffset) page1) * MULTIXACT_MEMBERS_PER_PAGE;
- offset2 = ((MultiXactOffset) page2) * MULTIXACT_MEMBERS_PER_PAGE;
-
- return (MultiXactOffsetPrecedes(offset1, offset2) &&
- MultiXactOffsetPrecedes(offset1,
- offset2 + MULTIXACT_MEMBERS_PER_PAGE - 1));
-}
-
/*
* Decide which of two MultiXactIds is earlier.
*
@@ -3243,32 +3116,18 @@ multixact_redo(XLogReaderState *record)
if (info == XLOG_MULTIXACT_ZERO_OFF_PAGE)
{
int pageno;
- int slotno;
memcpy(&pageno, XLogRecGetData(record), sizeof(int));
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
-
- slotno = ZeroMultiXactOffsetPage(pageno, false);
- SimpleLruWritePage(MultiXactOffsetCtl, slotno);
- Assert(!MultiXactOffsetCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(MultiXactOffsetSLRULock);
+ UnlockReleaseBuffer(ZeroMultiXactOffsetPage(pageno, false));
}
else if (info == XLOG_MULTIXACT_ZERO_MEM_PAGE)
{
int pageno;
- int slotno;
memcpy(&pageno, XLogRecGetData(record), sizeof(int));
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
-
- slotno = ZeroMultiXactMemberPage(pageno, false);
- SimpleLruWritePage(MultiXactMemberCtl, slotno);
- Assert(!MultiXactMemberCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(MultiXactMemberSLRULock);
+ UnlockReleaseBuffer(ZeroMultiXactMemberPage(pageno, false));
}
else if (info == XLOG_MULTIXACT_CREATE_ID)
{
@@ -3302,7 +3161,6 @@ multixact_redo(XLogReaderState *record)
else if (info == XLOG_MULTIXACT_TRUNCATE_ID)
{
xl_multixact_truncate xlrec;
- int pageno;
memcpy(&xlrec, XLogRecGetData(record),
SizeOfMultiXactTruncate);
@@ -3328,13 +3186,6 @@ multixact_redo(XLogReaderState *record)
PerformMembersTruncation(xlrec.startTruncMemb, xlrec.endTruncMemb);
- /*
- * During XLOG replay, latest_page_number isn't necessarily set up
- * yet; insert a suitable value to bypass the sanity test in
- * SimpleLruTruncate.
- */
- pageno = MultiXactIdToOffsetPage(xlrec.endTruncOff);
- MultiXactOffsetCtl->shared->latest_page_number = pageno;
PerformOffsetsTruncation(xlrec.startTruncOff, xlrec.endTruncOff);
LWLockRelease(MultiXactTruncationLock);
@@ -3407,21 +3258,3 @@ pg_get_multixact_members(PG_FUNCTION_ARGS)
SRF_RETURN_DONE(funccxt);
}
-
-/*
- * Entrypoint for sync.c to sync offsets files.
- */
-int
-multixactoffsetssyncfiletag(const FileTag *ftag, char *path)
-{
- return SlruSyncFileTag(MultiXactOffsetCtl, ftag, path);
-}
-
-/*
- * Entrypoint for sync.c to sync members files.
- */
-int
-multixactmemberssyncfiletag(const FileTag *ftag, char *path)
-{
- return SlruSyncFileTag(MultiXactMemberCtl, ftag, path);
-}
diff --git a/src/backend/access/transam/slru.c b/src/backend/access/transam/slru.c
index b65cb49d7ff..8890503b7cf 100644
--- a/src/backend/access/transam/slru.c
+++ b/src/backend/access/transam/slru.c
@@ -1,41 +1,9 @@
/*-------------------------------------------------------------------------
*
* slru.c
- * Simple LRU buffering for transaction status logfiles
+ * Simple buffering for transaction status logfiles
*
- * We use a simple least-recently-used scheme to manage a pool of page
- * buffers. Under ordinary circumstances we expect that write
- * traffic will occur mostly to the latest page (and to the just-prior
- * page, soon after a page transition). Read traffic will probably touch
- * a larger span of pages, but in any case a fairly small number of page
- * buffers should be sufficient. So, we just search the buffers using plain
- * linear search; there's no need for a hashtable or anything fancy.
- * The management algorithm is straight LRU except that we will never swap
- * out the latest page (since we know it's going to be hit again eventually).
- *
- * We use a control LWLock to protect the shared data structures, plus
- * per-buffer LWLocks that synchronize I/O for each buffer. The control lock
- * must be held to examine or modify any shared state. A process that is
- * reading in or writing out a page buffer does not hold the control lock,
- * only the per-buffer lock for the buffer it is working on.
- *
- * "Holding the control lock" means exclusive lock in all cases except for
- * SimpleLruReadPage_ReadOnly(); see comments for SlruRecentlyUsed() for
- * the implications of that.
- *
- * When initiating I/O on a buffer, we acquire the per-buffer lock exclusively
- * before releasing the control lock. The per-buffer lock is released after
- * completing the I/O, re-acquiring the control lock, and updating the shared
- * state. (Deadlock is not possible here, because we never try to initiate
- * I/O when someone else is already doing I/O on the same buffer.)
- * To wait for I/O to complete, release the control lock, acquire the
- * per-buffer lock in shared mode, immediately release the per-buffer lock,
- * reacquire the control lock, and then recheck state (since arbitrary things
- * could have happened while we didn't have the lock).
- *
- * As with the regular buffer manager, it is possible for another process
- * to re-dirty a page that is currently being written out. This is handled
- * by re-setting the page's page_dirty flag.
+ * XXX write me
*
*
* Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group
@@ -60,561 +28,34 @@
#include "storage/fd.h"
#include "storage/shmem.h"
-#define SlruFileName(ctl, path, seg) \
- snprintf(path, MAXPGPATH, "%s/%04X", (ctl)->Dir, seg)
-
/*
- * During SimpleLruWriteAll(), we will usually not need to write more than one
- * or two physical files, but we may need to write several pages per file. We
- * can consolidate the I/O requests by leaving files open until control returns
- * to SimpleLruWriteAll(). This data structure remembers which files are open.
+ * SLRU ID to path mapping
*/
-#define MAX_WRITEALL_BUFFERS 16
+#define PG_SLRU(symname,name,path,synchronize) \
+ path,
-typedef struct SlruWriteAllData
+static char *slru_dirs[] =
{
- int num_files; /* # files actually open */
- int fd[MAX_WRITEALL_BUFFERS]; /* their FD's */
- int segno[MAX_WRITEALL_BUFFERS]; /* their log seg#s */
-} SlruWriteAllData;
-
-typedef struct SlruWriteAllData *SlruWriteAll;
+#include "access/slrulist.h"
+};
/*
- * Populate a file tag describing a segment file. We only use the segment
- * number, since we can derive everything else we need by having separate
- * sync handler functions for clog, multixact etc.
+ * We'll maintain a little cache of recently seen buffers, to try to avoid the
+ * buffer mapping table on repeat access (ie the busy end of the CLOG). One
+ * entry per SLRU.
*/
-#define INIT_SLRUFILETAG(a,xx_handler,xx_segno) \
-( \
- memset(&(a), 0, sizeof(FileTag)), \
- (a).handler = (xx_handler), \
- (a).segno = (xx_segno) \
-)
-
-/*
- * Macro to mark a buffer slot "most recently used". Note multiple evaluation
- * of arguments!
- *
- * The reason for the if-test is that there are often many consecutive
- * accesses to the same page (particularly the latest page). By suppressing
- * useless increments of cur_lru_count, we reduce the probability that old
- * pages' counts will "wrap around" and make them appear recently used.
- *
- * We allow this code to be executed concurrently by multiple processes within
- * SimpleLruReadPage_ReadOnly(). As long as int reads and writes are atomic,
- * this should not cause any completely-bogus values to enter the computation.
- * However, it is possible for either cur_lru_count or individual
- * page_lru_count entries to be "reset" to lower values than they should have,
- * in case a process is delayed while it executes this macro. With care in
- * SlruSelectLRUPage(), this does little harm, and in any case the absolute
- * worst possible consequence is a nonoptimal choice of page to evict. The
- * gain from allowing concurrent reads of SLRU pages seems worth it.
- */
-#define SlruRecentlyUsed(shared, slotno) \
- do { \
- int new_lru_count = (shared)->cur_lru_count; \
- if (new_lru_count != (shared)->page_lru_count[slotno]) { \
- (shared)->cur_lru_count = ++new_lru_count; \
- (shared)->page_lru_count[slotno] = new_lru_count; \
- } \
- } while (0)
-
-/* Saved info for SlruReportIOError */
-typedef enum
-{
- SLRU_OPEN_FAILED,
- SLRU_SEEK_FAILED,
- SLRU_READ_FAILED,
- SLRU_WRITE_FAILED,
- SLRU_FSYNC_FAILED,
- SLRU_CLOSE_FAILED
-} SlruErrorCause;
+struct SlruRecentBuffer {
+ int pageno;
+ Buffer recent_buffer;
+};
-static SlruErrorCause slru_errcause;
-static int slru_errno;
+static struct SlruRecentBuffer slru_recent_buffers[SLRU_NEXT_ID];
-
-static void SimpleLruZeroLSNs(SlruCtl ctl, int slotno);
-static void SimpleLruWaitIO(SlruCtl ctl, int slotno);
-static void SlruInternalWritePage(SlruCtl ctl, int slotno, SlruWriteAll fdata);
-static bool SlruPhysicalReadPage(SlruCtl ctl, int pageno, int slotno);
-static bool SlruPhysicalWritePage(SlruCtl ctl, int pageno, int slotno,
- SlruWriteAll fdata);
-static void SlruReportIOError(SlruCtl ctl, int pageno, TransactionId xid);
-static int SlruSelectLRUPage(SlruCtl ctl, int pageno);
-
-static bool SlruScanDirCbDeleteCutoff(SlruCtl ctl, char *filename,
+static bool SlruScanDirCbDeleteCutoff(int slru_id,
+ SlruPagePrecedesFunction PagePrecedes,
+ char *filename,
int segpage, void *data);
-static void SlruInternalDeleteSegment(SlruCtl ctl, int segno);
-
-/*
- * Initialization of shared memory
- */
-
-Size
-SimpleLruShmemSize(int nslots, int nlsns)
-{
- Size sz;
-
- /* we assume nslots isn't so large as to risk overflow */
- sz = MAXALIGN(sizeof(SlruSharedData));
- sz += MAXALIGN(nslots * sizeof(char *)); /* page_buffer[] */
- sz += MAXALIGN(nslots * sizeof(SlruPageStatus)); /* page_status[] */
- sz += MAXALIGN(nslots * sizeof(bool)); /* page_dirty[] */
- sz += MAXALIGN(nslots * sizeof(int)); /* page_number[] */
- sz += MAXALIGN(nslots * sizeof(int)); /* page_lru_count[] */
- sz += MAXALIGN(nslots * sizeof(LWLockPadded)); /* buffer_locks[] */
-
- if (nlsns > 0)
- sz += MAXALIGN(nslots * nlsns * sizeof(XLogRecPtr)); /* group_lsn[] */
-
- return BUFFERALIGN(sz) + BLCKSZ * nslots;
-}
-
-/*
- * Initialize, or attach to, a simple LRU cache in shared memory.
- *
- * ctl: address of local (unshared) control structure.
- * name: name of SLRU. (This is user-visible, pick with care!)
- * nslots: number of page slots to use.
- * nlsns: number of LSN groups per page (set to zero if not relevant).
- * ctllock: LWLock to use to control access to the shared control structure.
- * subdir: PGDATA-relative subdirectory that will contain the files.
- * tranche_id: LWLock tranche ID to use for the SLRU's per-buffer LWLocks.
- */
-void
-SimpleLruInit(SlruCtl ctl, const char *name, int nslots, int nlsns,
- LWLock *ctllock, const char *subdir, int tranche_id,
- SyncRequestHandler sync_handler)
-{
- SlruShared shared;
- bool found;
-
- shared = (SlruShared) ShmemInitStruct(name,
- SimpleLruShmemSize(nslots, nlsns),
- &found);
-
- if (!IsUnderPostmaster)
- {
- /* Initialize locks and shared memory area */
- char *ptr;
- Size offset;
- int slotno;
-
- Assert(!found);
-
- memset(shared, 0, sizeof(SlruSharedData));
-
- shared->ControlLock = ctllock;
-
- shared->num_slots = nslots;
- shared->lsn_groups_per_page = nlsns;
-
- shared->cur_lru_count = 0;
-
- /* shared->latest_page_number will be set later */
-
- shared->slru_stats_idx = pgstat_get_slru_index(name);
-
- ptr = (char *) shared;
- offset = MAXALIGN(sizeof(SlruSharedData));
- shared->page_buffer = (char **) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(char *));
- shared->page_status = (SlruPageStatus *) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(SlruPageStatus));
- shared->page_dirty = (bool *) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(bool));
- shared->page_number = (int *) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(int));
- shared->page_lru_count = (int *) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(int));
-
- /* Initialize LWLocks */
- shared->buffer_locks = (LWLockPadded *) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(LWLockPadded));
-
- if (nlsns > 0)
- {
- shared->group_lsn = (XLogRecPtr *) (ptr + offset);
- offset += MAXALIGN(nslots * nlsns * sizeof(XLogRecPtr));
- }
-
- ptr += BUFFERALIGN(offset);
- for (slotno = 0; slotno < nslots; slotno++)
- {
- LWLockInitialize(&shared->buffer_locks[slotno].lock,
- tranche_id);
-
- shared->page_buffer[slotno] = ptr;
- shared->page_status[slotno] = SLRU_PAGE_EMPTY;
- shared->page_dirty[slotno] = false;
- shared->page_lru_count[slotno] = 0;
- ptr += BLCKSZ;
- }
-
- /* Should fit to estimated shmem size */
- Assert(ptr - (char *) shared <= SimpleLruShmemSize(nslots, nlsns));
- }
- else
- Assert(found);
-
- /*
- * Initialize the unshared control struct, including directory path. We
- * assume caller set PagePrecedes.
- */
- ctl->shared = shared;
- ctl->sync_handler = sync_handler;
- strlcpy(ctl->Dir, subdir, sizeof(ctl->Dir));
-}
-
-/*
- * Initialize (or reinitialize) a page to zeroes.
- *
- * The page is not actually written, just set up in shared memory.
- * The slot number of the new page is returned.
- *
- * Control lock must be held at entry, and will be held at exit.
- */
-int
-SimpleLruZeroPage(SlruCtl ctl, int pageno)
-{
- SlruShared shared = ctl->shared;
- int slotno;
-
- /* Find a suitable buffer slot for the page */
- slotno = SlruSelectLRUPage(ctl, pageno);
- Assert(shared->page_status[slotno] == SLRU_PAGE_EMPTY ||
- (shared->page_status[slotno] == SLRU_PAGE_VALID &&
- !shared->page_dirty[slotno]) ||
- shared->page_number[slotno] == pageno);
-
- /* Mark the slot as containing this page */
- shared->page_number[slotno] = pageno;
- shared->page_status[slotno] = SLRU_PAGE_VALID;
- shared->page_dirty[slotno] = true;
- SlruRecentlyUsed(shared, slotno);
-
- /* Set the buffer to zeroes */
- MemSet(shared->page_buffer[slotno], 0, BLCKSZ);
-
- /* Set the LSNs for this new page to zero */
- SimpleLruZeroLSNs(ctl, slotno);
-
- /* Assume this page is now the latest active page */
- shared->latest_page_number = pageno;
-
- /* update the stats counter of zeroed pages */
- pgstat_count_slru_page_zeroed(shared->slru_stats_idx);
-
- return slotno;
-}
-
-/*
- * Zero all the LSNs we store for this slru page.
- *
- * This should be called each time we create a new page, and each time we read
- * in a page from disk into an existing buffer. (Such an old page cannot
- * have any interesting LSNs, since we'd have flushed them before writing
- * the page in the first place.)
- *
- * This assumes that InvalidXLogRecPtr is bitwise-all-0.
- */
-static void
-SimpleLruZeroLSNs(SlruCtl ctl, int slotno)
-{
- SlruShared shared = ctl->shared;
-
- if (shared->lsn_groups_per_page > 0)
- MemSet(&shared->group_lsn[slotno * shared->lsn_groups_per_page], 0,
- shared->lsn_groups_per_page * sizeof(XLogRecPtr));
-}
-
-/*
- * Wait for any active I/O on a page slot to finish. (This does not
- * guarantee that new I/O hasn't been started before we return, though.
- * In fact the slot might not even contain the same page anymore.)
- *
- * Control lock must be held at entry, and will be held at exit.
- */
-static void
-SimpleLruWaitIO(SlruCtl ctl, int slotno)
-{
- SlruShared shared = ctl->shared;
-
- /* See notes at top of file */
- LWLockRelease(shared->ControlLock);
- LWLockAcquire(&shared->buffer_locks[slotno].lock, LW_SHARED);
- LWLockRelease(&shared->buffer_locks[slotno].lock);
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
- /*
- * If the slot is still in an io-in-progress state, then either someone
- * already started a new I/O on the slot, or a previous I/O failed and
- * neglected to reset the page state. That shouldn't happen, really, but
- * it seems worth a few extra cycles to check and recover from it. We can
- * cheaply test for failure by seeing if the buffer lock is still held (we
- * assume that transaction abort would release the lock).
- */
- if (shared->page_status[slotno] == SLRU_PAGE_READ_IN_PROGRESS ||
- shared->page_status[slotno] == SLRU_PAGE_WRITE_IN_PROGRESS)
- {
- if (LWLockConditionalAcquire(&shared->buffer_locks[slotno].lock, LW_SHARED))
- {
- /* indeed, the I/O must have failed */
- if (shared->page_status[slotno] == SLRU_PAGE_READ_IN_PROGRESS)
- shared->page_status[slotno] = SLRU_PAGE_EMPTY;
- else /* write_in_progress */
- {
- shared->page_status[slotno] = SLRU_PAGE_VALID;
- shared->page_dirty[slotno] = true;
- }
- LWLockRelease(&shared->buffer_locks[slotno].lock);
- }
- }
-}
-
-/*
- * Find a page in a shared buffer, reading it in if necessary.
- * The page number must correspond to an already-initialized page.
- *
- * If write_ok is true then it is OK to return a page that is in
- * WRITE_IN_PROGRESS state; it is the caller's responsibility to be sure
- * that modification of the page is safe. If write_ok is false then we
- * will not return the page until it is not undergoing active I/O.
- *
- * The passed-in xid is used only for error reporting, and may be
- * InvalidTransactionId if no specific xid is associated with the action.
- *
- * Return value is the shared-buffer slot number now holding the page.
- * The buffer's LRU access info is updated.
- *
- * Control lock must be held at entry, and will be held at exit.
- */
-int
-SimpleLruReadPage(SlruCtl ctl, int pageno, bool write_ok,
- TransactionId xid)
-{
- SlruShared shared = ctl->shared;
-
- /* Outer loop handles restart if we must wait for someone else's I/O */
- for (;;)
- {
- int slotno;
- bool ok;
-
- /* See if page already is in memory; if not, pick victim slot */
- slotno = SlruSelectLRUPage(ctl, pageno);
-
- /* Did we find the page in memory? */
- if (shared->page_number[slotno] == pageno &&
- shared->page_status[slotno] != SLRU_PAGE_EMPTY)
- {
- /*
- * If page is still being read in, we must wait for I/O. Likewise
- * if the page is being written and the caller said that's not OK.
- */
- if (shared->page_status[slotno] == SLRU_PAGE_READ_IN_PROGRESS ||
- (shared->page_status[slotno] == SLRU_PAGE_WRITE_IN_PROGRESS &&
- !write_ok))
- {
- SimpleLruWaitIO(ctl, slotno);
- /* Now we must recheck state from the top */
- continue;
- }
- /* Otherwise, it's ready to use */
- SlruRecentlyUsed(shared, slotno);
-
- /* update the stats counter of pages found in the SLRU */
- pgstat_count_slru_page_hit(shared->slru_stats_idx);
-
- return slotno;
- }
-
- /* We found no match; assert we selected a freeable slot */
- Assert(shared->page_status[slotno] == SLRU_PAGE_EMPTY ||
- (shared->page_status[slotno] == SLRU_PAGE_VALID &&
- !shared->page_dirty[slotno]));
-
- /* Mark the slot read-busy */
- shared->page_number[slotno] = pageno;
- shared->page_status[slotno] = SLRU_PAGE_READ_IN_PROGRESS;
- shared->page_dirty[slotno] = false;
-
- /* Acquire per-buffer lock (cannot deadlock, see notes at top) */
- LWLockAcquire(&shared->buffer_locks[slotno].lock, LW_EXCLUSIVE);
-
- /* Release control lock while doing I/O */
- LWLockRelease(shared->ControlLock);
-
- /* Do the read */
- ok = SlruPhysicalReadPage(ctl, pageno, slotno);
-
- /* Set the LSNs for this newly read-in page to zero */
- SimpleLruZeroLSNs(ctl, slotno);
-
- /* Re-acquire control lock and update page state */
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
- Assert(shared->page_number[slotno] == pageno &&
- shared->page_status[slotno] == SLRU_PAGE_READ_IN_PROGRESS &&
- !shared->page_dirty[slotno]);
-
- shared->page_status[slotno] = ok ? SLRU_PAGE_VALID : SLRU_PAGE_EMPTY;
-
- LWLockRelease(&shared->buffer_locks[slotno].lock);
-
- /* Now it's okay to ereport if we failed */
- if (!ok)
- SlruReportIOError(ctl, pageno, xid);
-
- SlruRecentlyUsed(shared, slotno);
-
- /* update the stats counter of pages not found in SLRU */
- pgstat_count_slru_page_read(shared->slru_stats_idx);
-
- return slotno;
- }
-}
-
-/*
- * Find a page in a shared buffer, reading it in if necessary.
- * The page number must correspond to an already-initialized page.
- * The caller must intend only read-only access to the page.
- *
- * The passed-in xid is used only for error reporting, and may be
- * InvalidTransactionId if no specific xid is associated with the action.
- *
- * Return value is the shared-buffer slot number now holding the page.
- * The buffer's LRU access info is updated.
- *
- * Control lock must NOT be held at entry, but will be held at exit.
- * It is unspecified whether the lock will be shared or exclusive.
- */
-int
-SimpleLruReadPage_ReadOnly(SlruCtl ctl, int pageno, TransactionId xid)
-{
- SlruShared shared = ctl->shared;
- int slotno;
-
- /* Try to find the page while holding only shared lock */
- LWLockAcquire(shared->ControlLock, LW_SHARED);
-
- /* See if page is already in a buffer */
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- if (shared->page_number[slotno] == pageno &&
- shared->page_status[slotno] != SLRU_PAGE_EMPTY &&
- shared->page_status[slotno] != SLRU_PAGE_READ_IN_PROGRESS)
- {
- /* See comments for SlruRecentlyUsed macro */
- SlruRecentlyUsed(shared, slotno);
-
- /* update the stats counter of pages found in the SLRU */
- pgstat_count_slru_page_hit(shared->slru_stats_idx);
-
- return slotno;
- }
- }
-
- /* No luck, so switch to normal exclusive lock and do regular read */
- LWLockRelease(shared->ControlLock);
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
- return SimpleLruReadPage(ctl, pageno, true, xid);
-}
-
-/*
- * Write a page from a shared buffer, if necessary.
- * Does nothing if the specified slot is not dirty.
- *
- * NOTE: only one write attempt is made here. Hence, it is possible that
- * the page is still dirty at exit (if someone else re-dirtied it during
- * the write). However, we *do* attempt a fresh write even if the page
- * is already being written; this is for checkpoints.
- *
- * Control lock must be held at entry, and will be held at exit.
- */
-static void
-SlruInternalWritePage(SlruCtl ctl, int slotno, SlruWriteAll fdata)
-{
- SlruShared shared = ctl->shared;
- int pageno = shared->page_number[slotno];
- bool ok;
-
- /* If a write is in progress, wait for it to finish */
- while (shared->page_status[slotno] == SLRU_PAGE_WRITE_IN_PROGRESS &&
- shared->page_number[slotno] == pageno)
- {
- SimpleLruWaitIO(ctl, slotno);
- }
-
- /*
- * Do nothing if page is not dirty, or if buffer no longer contains the
- * same page we were called for.
- */
- if (!shared->page_dirty[slotno] ||
- shared->page_status[slotno] != SLRU_PAGE_VALID ||
- shared->page_number[slotno] != pageno)
- return;
-
- /*
- * Mark the slot write-busy, and clear the dirtybit. After this point, a
- * transaction status update on this page will mark it dirty again.
- */
- shared->page_status[slotno] = SLRU_PAGE_WRITE_IN_PROGRESS;
- shared->page_dirty[slotno] = false;
-
- /* Acquire per-buffer lock (cannot deadlock, see notes at top) */
- LWLockAcquire(&shared->buffer_locks[slotno].lock, LW_EXCLUSIVE);
-
- /* Release control lock while doing I/O */
- LWLockRelease(shared->ControlLock);
-
- /* Do the write */
- ok = SlruPhysicalWritePage(ctl, pageno, slotno, fdata);
-
- /* If we failed, and we're in a flush, better close the files */
- if (!ok && fdata)
- {
- int i;
-
- for (i = 0; i < fdata->num_files; i++)
- CloseTransientFile(fdata->fd[i]);
- }
-
- /* Re-acquire control lock and update page state */
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
- Assert(shared->page_number[slotno] == pageno &&
- shared->page_status[slotno] == SLRU_PAGE_WRITE_IN_PROGRESS);
-
- /* If we failed to write, mark the page dirty again */
- if (!ok)
- shared->page_dirty[slotno] = true;
-
- shared->page_status[slotno] = SLRU_PAGE_VALID;
-
- LWLockRelease(&shared->buffer_locks[slotno].lock);
-
- /* Now it's okay to ereport if we failed */
- if (!ok)
- SlruReportIOError(ctl, pageno, InvalidTransactionId);
-
- /* If part of a checkpoint, count this as a buffer written. */
- if (fdata)
- CheckpointStats.ckpt_bufs_written++;
-}
-
-/*
- * Wrapper of SlruInternalWritePage, for external callers.
- * fdata is always passed a NULL here.
- */
-void
-SimpleLruWritePage(SlruCtl ctl, int slotno)
-{
- SlruInternalWritePage(ctl, slotno, NULL);
-}
+static void SlruInternalDeleteSegment(int slru_id, int segno);
/*
* Return whether the given page exists on disk.
@@ -623,592 +64,24 @@ SimpleLruWritePage(SlruCtl ctl, int slotno)
* large enough to contain the given page.
*/
bool
-SimpleLruDoesPhysicalPageExist(SlruCtl ctl, int pageno)
-{
- int segno = pageno / SLRU_PAGES_PER_SEGMENT;
- int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
- int offset = rpageno * BLCKSZ;
- char path[MAXPGPATH];
- int fd;
- bool result;
- off_t endpos;
-
- /* update the stats counter of checked pages */
- pgstat_count_slru_page_exists(ctl->shared->slru_stats_idx);
-
- SlruFileName(ctl, path, segno);
-
- fd = OpenTransientFile(path, O_RDONLY | PG_BINARY);
- if (fd < 0)
- {
- /* expected: file doesn't exist */
- if (errno == ENOENT)
- return false;
-
- /* report error normally */
- slru_errcause = SLRU_OPEN_FAILED;
- slru_errno = errno;
- SlruReportIOError(ctl, pageno, 0);
- }
-
- if ((endpos = lseek(fd, 0, SEEK_END)) < 0)
- {
- slru_errcause = SLRU_SEEK_FAILED;
- slru_errno = errno;
- SlruReportIOError(ctl, pageno, 0);
- }
-
- result = endpos >= (off_t) (offset + BLCKSZ);
-
- if (CloseTransientFile(fd) != 0)
- {
- slru_errcause = SLRU_CLOSE_FAILED;
- slru_errno = errno;
- return false;
- }
-
- return result;
-}
-
-/*
- * Physical read of a (previously existing) page into a buffer slot
- *
- * On failure, we cannot just ereport(ERROR) since caller has put state in
- * shared memory that must be undone. So, we return false and save enough
- * info in static variables to let SlruReportIOError make the report.
- *
- * For now, assume it's not worth keeping a file pointer open across
- * read/write operations. We could cache one virtual file pointer ...
- */
-static bool
-SlruPhysicalReadPage(SlruCtl ctl, int pageno, int slotno)
-{
- SlruShared shared = ctl->shared;
- int segno = pageno / SLRU_PAGES_PER_SEGMENT;
- int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
- off_t offset = rpageno * BLCKSZ;
- char path[MAXPGPATH];
- int fd;
-
- SlruFileName(ctl, path, segno);
-
- /*
- * In a crash-and-restart situation, it's possible for us to receive
- * commands to set the commit status of transactions whose bits are in
- * already-truncated segments of the commit log (see notes in
- * SlruPhysicalWritePage). Hence, if we are InRecovery, allow the case
- * where the file doesn't exist, and return zeroes instead.
- */
- fd = OpenTransientFile(path, O_RDONLY | PG_BINARY);
- if (fd < 0)
- {
- if (errno != ENOENT || !InRecovery)
- {
- slru_errcause = SLRU_OPEN_FAILED;
- slru_errno = errno;
- return false;
- }
-
- ereport(LOG,
- (errmsg("file \"%s\" doesn't exist, reading as zeroes",
- path)));
- MemSet(shared->page_buffer[slotno], 0, BLCKSZ);
- return true;
- }
-
- errno = 0;
- pgstat_report_wait_start(WAIT_EVENT_SLRU_READ);
- if (pg_pread(fd, shared->page_buffer[slotno], BLCKSZ, offset) != BLCKSZ)
- {
- pgstat_report_wait_end();
- slru_errcause = SLRU_READ_FAILED;
- slru_errno = errno;
- CloseTransientFile(fd);
- return false;
- }
- pgstat_report_wait_end();
-
- if (CloseTransientFile(fd) != 0)
- {
- slru_errcause = SLRU_CLOSE_FAILED;
- slru_errno = errno;
- return false;
- }
-
- return true;
-}
-
-/*
- * Physical write of a page from a buffer slot
- *
- * On failure, we cannot just ereport(ERROR) since caller has put state in
- * shared memory that must be undone. So, we return false and save enough
- * info in static variables to let SlruReportIOError make the report.
- *
- * For now, assume it's not worth keeping a file pointer open across
- * independent read/write operations. We do batch operations during
- * SimpleLruWriteAll, though.
- *
- * fdata is NULL for a standalone write, pointer to open-file info during
- * SimpleLruWriteAll.
- */
-static bool
-SlruPhysicalWritePage(SlruCtl ctl, int pageno, int slotno, SlruWriteAll fdata)
+SimpleLruDoesPhysicalPageExist(int slru_id, int pageno)
{
- SlruShared shared = ctl->shared;
int segno = pageno / SLRU_PAGES_PER_SEGMENT;
int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
off_t offset = rpageno * BLCKSZ;
- char path[MAXPGPATH];
- int fd = -1;
-
- /* update the stats counter of written pages */
- pgstat_count_slru_page_written(shared->slru_stats_idx);
-
- /*
- * Honor the write-WAL-before-data rule, if appropriate, so that we do not
- * write out data before associated WAL records. This is the same action
- * performed during FlushBuffer() in the main buffer manager.
- */
- if (shared->group_lsn != NULL)
- {
- /*
- * We must determine the largest async-commit LSN for the page. This
- * is a bit tedious, but since this entire function is a slow path
- * anyway, it seems better to do this here than to maintain a per-page
- * LSN variable (which'd need an extra comparison in the
- * transaction-commit path).
- */
- XLogRecPtr max_lsn;
- int lsnindex,
- lsnoff;
-
- lsnindex = slotno * shared->lsn_groups_per_page;
- max_lsn = shared->group_lsn[lsnindex++];
- for (lsnoff = 1; lsnoff < shared->lsn_groups_per_page; lsnoff++)
- {
- XLogRecPtr this_lsn = shared->group_lsn[lsnindex++];
-
- if (max_lsn < this_lsn)
- max_lsn = this_lsn;
- }
-
- if (!XLogRecPtrIsInvalid(max_lsn))
- {
- /*
- * As noted above, elog(ERROR) is not acceptable here, so if
- * XLogFlush were to fail, we must PANIC. This isn't much of a
- * restriction because XLogFlush is just about all critical
- * section anyway, but let's make sure.
- */
- START_CRIT_SECTION();
- XLogFlush(max_lsn);
- END_CRIT_SECTION();
- }
- }
-
- /*
- * During a WriteAll, we may already have the desired file open.
- */
- if (fdata)
- {
- int i;
-
- for (i = 0; i < fdata->num_files; i++)
- {
- if (fdata->segno[i] == segno)
- {
- fd = fdata->fd[i];
- break;
- }
- }
- }
-
- if (fd < 0)
- {
- /*
- * If the file doesn't already exist, we should create it. It is
- * possible for this to need to happen when writing a page that's not
- * first in its segment; we assume the OS can cope with that. (Note:
- * it might seem that it'd be okay to create files only when
- * SimpleLruZeroPage is called for the first page of a segment.
- * However, if after a crash and restart the REDO logic elects to
- * replay the log from a checkpoint before the latest one, then it's
- * possible that we will get commands to set transaction status of
- * transactions that have already been truncated from the commit log.
- * Easiest way to deal with that is to accept references to
- * nonexistent files here and in SlruPhysicalReadPage.)
- *
- * Note: it is possible for more than one backend to be executing this
- * code simultaneously for different pages of the same file. Hence,
- * don't use O_EXCL or O_TRUNC or anything like that.
- */
- SlruFileName(ctl, path, segno);
- fd = OpenTransientFile(path, O_RDWR | O_CREAT | PG_BINARY);
- if (fd < 0)
- {
- slru_errcause = SLRU_OPEN_FAILED;
- slru_errno = errno;
- return false;
- }
-
- if (fdata)
- {
- if (fdata->num_files < MAX_WRITEALL_BUFFERS)
- {
- fdata->fd[fdata->num_files] = fd;
- fdata->segno[fdata->num_files] = segno;
- fdata->num_files++;
- }
- else
- {
- /*
- * In the unlikely event that we exceed MAX_FLUSH_BUFFERS,
- * fall back to treating it as a standalone write.
- */
- fdata = NULL;
- }
- }
- }
-
- errno = 0;
- pgstat_report_wait_start(WAIT_EVENT_SLRU_WRITE);
- if (pg_pwrite(fd, shared->page_buffer[slotno], BLCKSZ, offset) != BLCKSZ)
- {
- pgstat_report_wait_end();
- /* if write didn't set errno, assume problem is no disk space */
- if (errno == 0)
- errno = ENOSPC;
- slru_errcause = SLRU_WRITE_FAILED;
- slru_errno = errno;
- if (!fdata)
- CloseTransientFile(fd);
- return false;
- }
- pgstat_report_wait_end();
-
- /* Queue up a sync request for the checkpointer. */
- if (ctl->sync_handler != SYNC_HANDLER_NONE)
- {
- FileTag tag;
-
- INIT_SLRUFILETAG(tag, ctl->sync_handler, segno);
- if (!RegisterSyncRequest(&tag, SYNC_REQUEST, false))
- {
- /* No space to enqueue sync request. Do it synchronously. */
- pgstat_report_wait_start(WAIT_EVENT_SLRU_SYNC);
- if (pg_fsync(fd) != 0)
- {
- pgstat_report_wait_end();
- slru_errcause = SLRU_FSYNC_FAILED;
- slru_errno = errno;
- CloseTransientFile(fd);
- return false;
- }
- pgstat_report_wait_end();
- }
- }
-
- /* Close file, unless part of flush request. */
- if (!fdata)
- {
- if (CloseTransientFile(fd) != 0)
- {
- slru_errcause = SLRU_CLOSE_FAILED;
- slru_errno = errno;
- return false;
- }
- }
-
- return true;
-}
-
-/*
- * Issue the error message after failure of SlruPhysicalReadPage or
- * SlruPhysicalWritePage. Call this after cleaning up shared-memory state.
- */
-static void
-SlruReportIOError(SlruCtl ctl, int pageno, TransactionId xid)
-{
- int segno = pageno / SLRU_PAGES_PER_SEGMENT;
- int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
- int offset = rpageno * BLCKSZ;
- char path[MAXPGPATH];
-
- SlruFileName(ctl, path, segno);
- errno = slru_errno;
- switch (slru_errcause)
- {
- case SLRU_OPEN_FAILED:
- ereport(ERROR,
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not open file \"%s\": %m.", path)));
- break;
- case SLRU_SEEK_FAILED:
- ereport(ERROR,
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not seek in file \"%s\" to offset %d: %m.",
- path, offset)));
- break;
- case SLRU_READ_FAILED:
- if (errno)
- ereport(ERROR,
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not read from file \"%s\" at offset %d: %m.",
- path, offset)));
- else
- ereport(ERROR,
- (errmsg("could not access status of transaction %u", xid),
- errdetail("Could not read from file \"%s\" at offset %d: read too few bytes.", path, offset)));
- break;
- case SLRU_WRITE_FAILED:
- if (errno)
- ereport(ERROR,
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not write to file \"%s\" at offset %d: %m.",
- path, offset)));
- else
- ereport(ERROR,
- (errmsg("could not access status of transaction %u", xid),
- errdetail("Could not write to file \"%s\" at offset %d: wrote too few bytes.",
- path, offset)));
- break;
- case SLRU_FSYNC_FAILED:
- ereport(data_sync_elevel(ERROR),
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not fsync file \"%s\": %m.",
- path)));
- break;
- case SLRU_CLOSE_FAILED:
- ereport(ERROR,
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not close file \"%s\": %m.",
- path)));
- break;
- default:
- /* can't get here, we trust */
- elog(ERROR, "unrecognized SimpleLru error cause: %d",
- (int) slru_errcause);
- break;
- }
-}
-
-/*
- * Select the slot to re-use when we need a free slot.
- *
- * The target page number is passed because we need to consider the
- * possibility that some other process reads in the target page while
- * we are doing I/O to free a slot. Hence, check or recheck to see if
- * any slot already holds the target page, and return that slot if so.
- * Thus, the returned slot is *either* a slot already holding the pageno
- * (could be any state except EMPTY), *or* a freeable slot (state EMPTY
- * or CLEAN).
- *
- * Control lock must be held at entry, and will be held at exit.
- */
-static int
-SlruSelectLRUPage(SlruCtl ctl, int pageno)
-{
- SlruShared shared = ctl->shared;
-
- /* Outer loop handles restart after I/O */
- for (;;)
- {
- int slotno;
- int cur_count;
- int bestvalidslot = 0; /* keep compiler quiet */
- int best_valid_delta = -1;
- int best_valid_page_number = 0; /* keep compiler quiet */
- int bestinvalidslot = 0; /* keep compiler quiet */
- int best_invalid_delta = -1;
- int best_invalid_page_number = 0; /* keep compiler quiet */
-
- /* See if page already has a buffer assigned */
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- if (shared->page_number[slotno] == pageno &&
- shared->page_status[slotno] != SLRU_PAGE_EMPTY)
- return slotno;
- }
-
- /*
- * If we find any EMPTY slot, just select that one. Else choose a
- * victim page to replace. We normally take the least recently used
- * valid page, but we will never take the slot containing
- * latest_page_number, even if it appears least recently used. We
- * will select a slot that is already I/O busy only if there is no
- * other choice: a read-busy slot will not be least recently used once
- * the read finishes, and waiting for an I/O on a write-busy slot is
- * inferior to just picking some other slot. Testing shows the slot
- * we pick instead will often be clean, allowing us to begin a read at
- * once.
- *
- * Normally the page_lru_count values will all be different and so
- * there will be a well-defined LRU page. But since we allow
- * concurrent execution of SlruRecentlyUsed() within
- * SimpleLruReadPage_ReadOnly(), it is possible that multiple pages
- * acquire the same lru_count values. In that case we break ties by
- * choosing the furthest-back page.
- *
- * Notice that this next line forcibly advances cur_lru_count to a
- * value that is certainly beyond any value that will be in the
- * page_lru_count array after the loop finishes. This ensures that
- * the next execution of SlruRecentlyUsed will mark the page newly
- * used, even if it's for a page that has the current counter value.
- * That gets us back on the path to having good data when there are
- * multiple pages with the same lru_count.
- */
- cur_count = (shared->cur_lru_count)++;
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- int this_delta;
- int this_page_number;
-
- if (shared->page_status[slotno] == SLRU_PAGE_EMPTY)
- return slotno;
- this_delta = cur_count - shared->page_lru_count[slotno];
- if (this_delta < 0)
- {
- /*
- * Clean up in case shared updates have caused cur_count
- * increments to get "lost". We back off the page counts,
- * rather than trying to increase cur_count, to avoid any
- * question of infinite loops or failure in the presence of
- * wrapped-around counts.
- */
- shared->page_lru_count[slotno] = cur_count;
- this_delta = 0;
- }
- this_page_number = shared->page_number[slotno];
- if (this_page_number == shared->latest_page_number)
- continue;
- if (shared->page_status[slotno] == SLRU_PAGE_VALID)
- {
- if (this_delta > best_valid_delta ||
- (this_delta == best_valid_delta &&
- ctl->PagePrecedes(this_page_number,
- best_valid_page_number)))
- {
- bestvalidslot = slotno;
- best_valid_delta = this_delta;
- best_valid_page_number = this_page_number;
- }
- }
- else
- {
- if (this_delta > best_invalid_delta ||
- (this_delta == best_invalid_delta &&
- ctl->PagePrecedes(this_page_number,
- best_invalid_page_number)))
- {
- bestinvalidslot = slotno;
- best_invalid_delta = this_delta;
- best_invalid_page_number = this_page_number;
- }
- }
- }
-
- /*
- * If all pages (except possibly the latest one) are I/O busy, we'll
- * have to wait for an I/O to complete and then retry. In that
- * unhappy case, we choose to wait for the I/O on the least recently
- * used slot, on the assumption that it was likely initiated first of
- * all the I/Os in progress and may therefore finish first.
- */
- if (best_valid_delta < 0)
- {
- SimpleLruWaitIO(ctl, bestinvalidslot);
- continue;
- }
-
- /*
- * If the selected page is clean, we're set.
- */
- if (!shared->page_dirty[bestvalidslot])
- return bestvalidslot;
-
- /*
- * Write the page.
- */
- SlruInternalWritePage(ctl, bestvalidslot, NULL);
-
- /*
- * Now loop back and try again. This is the easiest way of dealing
- * with corner cases such as the victim page being re-dirtied while we
- * wrote it.
- */
- }
-}
-
-/*
- * Write dirty pages to disk during checkpoint or database shutdown. Flushing
- * is deferred until the next call to ProcessSyncRequests(), though we do fsync
- * the containing directory here to make sure that newly created directory
- * entries are on disk.
- */
-void
-SimpleLruWriteAll(SlruCtl ctl, bool allow_redirtied)
-{
- SlruShared shared = ctl->shared;
- SlruWriteAllData fdata;
- int slotno;
- int pageno = 0;
- int i;
- bool ok;
-
- /* update the stats counter of flushes */
- pgstat_count_slru_flush(shared->slru_stats_idx);
-
- /*
- * Find and write dirty pages
- */
- fdata.num_files = 0;
-
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- SlruInternalWritePage(ctl, slotno, &fdata);
-
- /*
- * In some places (e.g. checkpoints), we cannot assert that the slot
- * is clean now, since another process might have re-dirtied it
- * already. That's okay.
- */
- Assert(allow_redirtied ||
- shared->page_status[slotno] == SLRU_PAGE_EMPTY ||
- (shared->page_status[slotno] == SLRU_PAGE_VALID &&
- !shared->page_dirty[slotno]));
- }
+ off_t size;
+ RelFileLocator rlocator = SlruRelFileLocator(slru_id, segno);
+ SMgrFileHandle sfile = smgropen(rlocator, InvalidBackendId, MAIN_FORKNUM);
- LWLockRelease(shared->ControlLock);
+ /* update the stats counter of checked pages */
+ pgstat_count_slru_page_exists(slru_id);
- /*
- * Now close any files that were open
- */
- ok = true;
- for (i = 0; i < fdata.num_files; i++)
- {
- if (CloseTransientFile(fdata.fd[i]) != 0)
- {
- slru_errcause = SLRU_CLOSE_FAILED;
- slru_errno = errno;
- pageno = fdata.segno[i] * SLRU_PAGES_PER_SEGMENT;
- ok = false;
- }
- }
- if (!ok)
- SlruReportIOError(ctl, pageno, InvalidTransactionId);
+ if (smgrexists(sfile))
+ size = smgrnblocks(sfile);
+ else
+ size = 0;
- /* Ensure that directory entries for new files are on disk. */
- if (ctl->sync_handler != SYNC_HANDLER_NONE)
- fsync_fname(ctl->Dir, true);
+ return size >= offset + BLCKSZ;
}
/*
@@ -1223,75 +96,14 @@ SimpleLruWriteAll(SlruCtl ctl, bool allow_redirtied)
* after it has accrued freshly-written data.
*/
void
-SimpleLruTruncate(SlruCtl ctl, int cutoffPage)
+SimpleLruTruncate(int slru_id, SlruPagePrecedesFunction PagePrecedes, int cutoffPage)
{
- SlruShared shared = ctl->shared;
- int slotno;
-
/* update the stats counter of truncates */
- pgstat_count_slru_truncate(shared->slru_stats_idx);
-
- /*
- * Scan shared memory and remove any pages preceding the cutoff page, to
- * ensure we won't rewrite them later. (Since this is normally called in
- * or just after a checkpoint, any dirty pages should have been flushed
- * already ... we're just being extra careful here.)
- */
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
-restart:;
-
- /*
- * While we are holding the lock, make an important safety check: the
- * current endpoint page must not be eligible for removal.
- */
- if (ctl->PagePrecedes(shared->latest_page_number, cutoffPage))
- {
- LWLockRelease(shared->ControlLock);
- ereport(LOG,
- (errmsg("could not truncate directory \"%s\": apparent wraparound",
- ctl->Dir)));
- return;
- }
-
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- if (shared->page_status[slotno] == SLRU_PAGE_EMPTY)
- continue;
- if (!ctl->PagePrecedes(shared->page_number[slotno], cutoffPage))
- continue;
-
- /*
- * If page is clean, just change state to EMPTY (expected case).
- */
- if (shared->page_status[slotno] == SLRU_PAGE_VALID &&
- !shared->page_dirty[slotno])
- {
- shared->page_status[slotno] = SLRU_PAGE_EMPTY;
- continue;
- }
-
- /*
- * Hmm, we have (or may have) I/O operations acting on the page, so
- * we've got to wait for them to finish and then start again. This is
- * the same logic as in SlruSelectLRUPage. (XXX if page is dirty,
- * wouldn't it be OK to just discard it without writing it?
- * SlruMayDeleteSegment() uses a stricter qualification, so we might
- * not delete this page in the end; even if we don't delete it, we
- * won't have cause to read its data again. For now, keep the logic
- * the same as it was.)
- */
- if (shared->page_status[slotno] == SLRU_PAGE_VALID)
- SlruInternalWritePage(ctl, slotno, NULL);
- else
- SimpleLruWaitIO(ctl, slotno);
- goto restart;
- }
-
- LWLockRelease(shared->ControlLock);
+ pgstat_count_slru_truncate(slru_id);
/* Now we can remove the old segment(s) */
- (void) SlruScanDirectory(ctl, SlruScanDirCbDeleteCutoff, &cutoffPage);
+ (void) SlruScanDirectory(slru_id, PagePrecedes, SlruScanDirCbDeleteCutoff,
+ &cutoffPage);
}
/*
@@ -1301,77 +113,22 @@ restart:;
* they either can't yet contain anything, or have already been cleaned out.
*/
static void
-SlruInternalDeleteSegment(SlruCtl ctl, int segno)
+SlruInternalDeleteSegment(int slru_id, int segno)
{
- char path[MAXPGPATH];
-
- /* Forget any fsync requests queued for this segment. */
- if (ctl->sync_handler != SYNC_HANDLER_NONE)
- {
- FileTag tag;
-
- INIT_SLRUFILETAG(tag, ctl->sync_handler, segno);
- RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true);
- }
+ RelFileLocator rlocator = SlruRelFileLocator(slru_id, segno);
+ SMgrFileHandle sfile = smgropen(rlocator, InvalidBackendId, MAIN_FORKNUM);
/* Unlink the file. */
- SlruFileName(ctl, path, segno);
- ereport(DEBUG2, (errmsg_internal("removing file \"%s\"", path)));
- unlink(path);
+ smgrunlink(sfile, false);
}
/*
* Delete an individual SLRU segment, identified by the segment number.
*/
void
-SlruDeleteSegment(SlruCtl ctl, int segno)
+SlruDeleteSegment(int slru_id, int segno)
{
- SlruShared shared = ctl->shared;
- int slotno;
- bool did_write;
-
- /* Clean out any possibly existing references to the segment. */
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-restart:
- did_write = false;
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- int pagesegno = shared->page_number[slotno] / SLRU_PAGES_PER_SEGMENT;
-
- if (shared->page_status[slotno] == SLRU_PAGE_EMPTY)
- continue;
-
- /* not the segment we're looking for */
- if (pagesegno != segno)
- continue;
-
- /* If page is clean, just change state to EMPTY (expected case). */
- if (shared->page_status[slotno] == SLRU_PAGE_VALID &&
- !shared->page_dirty[slotno])
- {
- shared->page_status[slotno] = SLRU_PAGE_EMPTY;
- continue;
- }
-
- /* Same logic as SimpleLruTruncate() */
- if (shared->page_status[slotno] == SLRU_PAGE_VALID)
- SlruInternalWritePage(ctl, slotno, NULL);
- else
- SimpleLruWaitIO(ctl, slotno);
-
- did_write = true;
- }
-
- /*
- * Be extra careful and re-check. The IO functions release the control
- * lock, so new pages could have been read in.
- */
- if (did_write)
- goto restart;
-
- SlruInternalDeleteSegment(ctl, segno);
-
- LWLockRelease(shared->ControlLock);
+ SlruInternalDeleteSegment(slru_id, segno);
}
/*
@@ -1388,19 +145,21 @@ restart:
* first>=cutoff && last>=cutoff: no; every page of this segment is too young
*/
static bool
-SlruMayDeleteSegment(SlruCtl ctl, int segpage, int cutoffPage)
+SlruMayDeleteSegment(SlruPagePrecedesFunction PagePrecedes,
+ int segpage, int cutoffPage)
{
int seg_last_page = segpage + SLRU_PAGES_PER_SEGMENT - 1;
Assert(segpage % SLRU_PAGES_PER_SEGMENT == 0);
- return (ctl->PagePrecedes(segpage, cutoffPage) &&
- ctl->PagePrecedes(seg_last_page, cutoffPage));
+ return (PagePrecedes(segpage, cutoffPage) &&
+ PagePrecedes(seg_last_page, cutoffPage));
}
#ifdef USE_ASSERT_CHECKING
static void
-SlruPagePrecedesTestOffset(SlruCtl ctl, int per_page, uint32 offset)
+SlruPagePrecedesTestOffset(SlruPagePrecedesFunction PagePrecedes,
+ int per_page, uint32 offset)
{
TransactionId lhs,
rhs;
@@ -1425,19 +184,19 @@ SlruPagePrecedesTestOffset(SlruCtl ctl, int per_page, uint32 offset)
Assert(!TransactionIdPrecedes(rhs, lhs + 1));
Assert(!TransactionIdFollowsOrEquals(lhs, rhs));
Assert(!TransactionIdFollowsOrEquals(rhs, lhs));
- Assert(!ctl->PagePrecedes(lhs / per_page, lhs / per_page));
- Assert(!ctl->PagePrecedes(lhs / per_page, rhs / per_page));
- Assert(!ctl->PagePrecedes(rhs / per_page, lhs / per_page));
- Assert(!ctl->PagePrecedes((lhs - per_page) / per_page, rhs / per_page));
- Assert(ctl->PagePrecedes(rhs / per_page, (lhs - 3 * per_page) / per_page));
- Assert(ctl->PagePrecedes(rhs / per_page, (lhs - 2 * per_page) / per_page));
- Assert(ctl->PagePrecedes(rhs / per_page, (lhs - 1 * per_page) / per_page)
+ Assert(!PagePrecedes(lhs / per_page, lhs / per_page));
+ Assert(!PagePrecedes(lhs / per_page, rhs / per_page));
+ Assert(!PagePrecedes(rhs / per_page, lhs / per_page));
+ Assert(!PagePrecedes((lhs - per_page) / per_page, rhs / per_page));
+ Assert(PagePrecedes(rhs / per_page, (lhs - 3 * per_page) / per_page));
+ Assert(PagePrecedes(rhs / per_page, (lhs - 2 * per_page) / per_page));
+ Assert(PagePrecedes(rhs / per_page, (lhs - 1 * per_page) / per_page)
|| (1U << 31) % per_page != 0); /* See CommitTsPagePrecedes() */
- Assert(ctl->PagePrecedes((lhs + 1 * per_page) / per_page, rhs / per_page)
+ Assert(PagePrecedes((lhs + 1 * per_page) / per_page, rhs / per_page)
|| (1U << 31) % per_page != 0);
- Assert(ctl->PagePrecedes((lhs + 2 * per_page) / per_page, rhs / per_page));
- Assert(ctl->PagePrecedes((lhs + 3 * per_page) / per_page, rhs / per_page));
- Assert(!ctl->PagePrecedes(rhs / per_page, (lhs + per_page) / per_page));
+ Assert(PagePrecedes((lhs + 2 * per_page) / per_page, rhs / per_page));
+ Assert(PagePrecedes((lhs + 3 * per_page) / per_page, rhs / per_page));
+ Assert(!PagePrecedes(rhs / per_page, (lhs + per_page) / per_page));
/*
* GetNewTransactionId() has assigned the last XID it can safely use, and
@@ -1450,7 +209,7 @@ SlruPagePrecedesTestOffset(SlruCtl ctl, int per_page, uint32 offset)
oldestXact = newestXact + 1;
oldestXact -= 1U << 31;
oldestPage = oldestXact / per_page;
- Assert(!SlruMayDeleteSegment(ctl,
+ Assert(!SlruMayDeleteSegment(PagePrecedes,
(newestPage -
newestPage % SLRU_PAGES_PER_SEGMENT),
oldestPage));
@@ -1466,7 +225,7 @@ SlruPagePrecedesTestOffset(SlruCtl ctl, int per_page, uint32 offset)
oldestXact = newestXact + 1;
oldestXact -= 1U << 31;
oldestPage = oldestXact / per_page;
- Assert(!SlruMayDeleteSegment(ctl,
+ Assert(!SlruMayDeleteSegment(PagePrecedes,
(newestPage -
newestPage % SLRU_PAGES_PER_SEGMENT),
oldestPage));
@@ -1482,12 +241,12 @@ SlruPagePrecedesTestOffset(SlruCtl ctl, int per_page, uint32 offset)
* do not apply to them.)
*/
void
-SlruPagePrecedesUnitTests(SlruCtl ctl, int per_page)
+SlruPagePrecedesUnitTests(SlruPagePrecedesFunction PagePrecedes, int per_page)
{
/* Test first, middle and last entries of a page. */
- SlruPagePrecedesTestOffset(ctl, per_page, 0);
- SlruPagePrecedesTestOffset(ctl, per_page, per_page / 2);
- SlruPagePrecedesTestOffset(ctl, per_page, per_page - 1);
+ SlruPagePrecedesTestOffset(PagePrecedes, per_page, 0);
+ SlruPagePrecedesTestOffset(PagePrecedes, per_page, per_page / 2);
+ SlruPagePrecedesTestOffset(PagePrecedes, per_page, per_page - 1);
}
#endif
@@ -1497,11 +256,12 @@ SlruPagePrecedesUnitTests(SlruCtl ctl, int per_page)
* one containing the page passed as "data".
*/
bool
-SlruScanDirCbReportPresence(SlruCtl ctl, char *filename, int segpage, void *data)
+SlruScanDirCbReportPresence(int slru_id, SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage, void *data)
{
int cutoffPage = *(int *) data;
- if (SlruMayDeleteSegment(ctl, segpage, cutoffPage))
+ if (SlruMayDeleteSegment(PagePrecedes, segpage, cutoffPage))
return true; /* found one; don't iterate any more */
return false; /* keep going */
@@ -1512,12 +272,15 @@ SlruScanDirCbReportPresence(SlruCtl ctl, char *filename, int segpage, void *data
* This callback deletes segments prior to the one passed in as "data".
*/
static bool
-SlruScanDirCbDeleteCutoff(SlruCtl ctl, char *filename, int segpage, void *data)
+SlruScanDirCbDeleteCutoff(int slru_id, SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage, void *data)
{
int cutoffPage = *(int *) data;
- if (SlruMayDeleteSegment(ctl, segpage, cutoffPage))
- SlruInternalDeleteSegment(ctl, segpage / SLRU_PAGES_PER_SEGMENT);
+ if (SlruMayDeleteSegment(PagePrecedes, segpage, cutoffPage))
+ {
+ SlruDeleteSegment(slru_id, segpage / SLRU_PAGES_PER_SEGMENT);
+ }
return false; /* keep going */
}
@@ -1527,9 +290,10 @@ SlruScanDirCbDeleteCutoff(SlruCtl ctl, char *filename, int segpage, void *data)
* This callback deletes all segments.
*/
bool
-SlruScanDirCbDeleteAll(SlruCtl ctl, char *filename, int segpage, void *data)
+SlruScanDirCbDeleteAll(int slru_id, SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage, void *data)
{
- SlruInternalDeleteSegment(ctl, segpage / SLRU_PAGES_PER_SEGMENT);
+ SlruInternalDeleteSegment(slru_id, segpage / SLRU_PAGES_PER_SEGMENT);
return false; /* keep going */
}
@@ -1550,16 +314,20 @@ SlruScanDirCbDeleteAll(SlruCtl ctl, char *filename, int segpage, void *data)
* Note that no locking is applied.
*/
bool
-SlruScanDirectory(SlruCtl ctl, SlruScanCallback callback, void *data)
+SlruScanDirectory(int slru_id, SlruPagePrecedesFunction PagePrecedes,
+ SlruScanCallback callback, void *data)
{
bool retval = false;
DIR *cldir;
struct dirent *clde;
int segno;
int segpage;
+ const char *path;
- cldir = AllocateDir(ctl->Dir);
- while ((clde = ReadDir(cldir, ctl->Dir)) != NULL)
+ path = slru_dirs[slru_id];
+
+ cldir = AllocateDir(path);
+ while ((clde = ReadDir(cldir, path)) != NULL)
{
size_t len;
@@ -1572,8 +340,8 @@ SlruScanDirectory(SlruCtl ctl, SlruScanCallback callback, void *data)
segpage = segno * SLRU_PAGES_PER_SEGMENT;
elog(DEBUG2, "SlruScanDirectory invoking callback on %s/%s",
- ctl->Dir, clde->d_name);
- retval = callback(ctl, clde->d_name, segpage, data);
+ path, clde->d_name);
+ retval = callback(slru_id, PagePrecedes, clde->d_name, segpage, data);
if (retval)
break;
}
@@ -1584,29 +352,78 @@ SlruScanDirectory(SlruCtl ctl, SlruScanCallback callback, void *data)
}
/*
- * Individual SLRUs (clog, ...) have to provide a sync.c handler function so
- * that they can provide the correct "SlruCtl" (otherwise we don't know how to
- * build the path), but they just forward to this common implementation that
- * performs the fsync.
+ * Read a buffer. Buffer is pinned on return.
*/
-int
-SlruSyncFileTag(SlruCtl ctl, const FileTag *ftag, char *path)
+Buffer
+ReadSlruBuffer(int slru_id, int pageno)
{
- int fd;
- int save_errno;
- int result;
+ int segno = pageno / SLRU_PAGES_PER_SEGMENT;
+ int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
+ RelFileLocator rlocator = SlruRelFileLocator(slru_id, segno);
+ Buffer buffer;
+ bool hit;
+
+ /* Try to avoid doing a buffer mapping table lookup for repeated access. */
+ buffer = slru_recent_buffers[slru_id].recent_buffer;
+ if (slru_recent_buffers[slru_id].pageno == pageno &&
+ BufferIsValid(buffer) &&
+ ReadRecentBuffer(rlocator, MAIN_FORKNUM, pageno, buffer))
+ {
+ pgstat_count_slru_page_hit(slru_id);
+ return buffer;
+ }
+
+ /* Regular lookup. */
+ buffer = ReadBufferWithoutRelcacheWithHit(rlocator, MAIN_FORKNUM, rpageno,
+ RBM_NORMAL, NULL, true, &hit);
- SlruFileName(ctl, path, ftag->segno);
+ /* Remember where this page is for next time. */
+ slru_recent_buffers[slru_id].pageno = pageno;
+ slru_recent_buffers[slru_id].recent_buffer = buffer;
- fd = OpenTransientFile(path, O_RDWR | PG_BINARY);
- if (fd < 0)
- return -1;
+ if (hit)
+ pgstat_count_slru_page_hit(slru_id);
- result = pg_fsync(fd);
- save_errno = errno;
+ return buffer;
+}
+
+/*
+ * Zero-initialize a buffer. Buffer is pinned and exclusively locked on return.
+ */
+Buffer
+ZeroSlruBuffer(int slru_id, int pageno)
+{
+ int segno = pageno / SLRU_PAGES_PER_SEGMENT;
+ int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
+ RelFileLocator rlocator = SlruRelFileLocator(slru_id, segno);
+ Buffer buffer;
+ SMgrFileHandle sfile;
- CloseTransientFile(fd);
+ if (rpageno == 0)
+ {
+ sfile = smgropen(rlocator, InvalidBackendId, MAIN_FORKNUM);
+ if (!smgrexists(sfile))
+ smgrcreate(sfile, false);
+ }
+
+ buffer = ReadBufferWithoutRelcache(rlocator, MAIN_FORKNUM, rpageno,
+ RBM_ZERO_AND_LOCK, NULL, true);
+
+ /* Remember where this page is for next time. */
+ slru_recent_buffers[slru_id].pageno = pageno;
+ slru_recent_buffers[slru_id].recent_buffer = buffer;
+
+ pgstat_count_slru_page_zeroed(slru_id);
+
+ return buffer;
+}
+
+bool
+ProbeSlruBuffer(int slru_id, int pageno)
+{
+ int segno = pageno / SLRU_PAGES_PER_SEGMENT;
+ int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
+ RelFileLocator rlocator = SlruRelFileLocator(slru_id, segno);
- errno = save_errno;
- return result;
+ return BufferProbe(rlocator, MAIN_FORKNUM, rpageno);
}
diff --git a/src/backend/access/transam/subtrans.c b/src/backend/access/transam/subtrans.c
index 66d35481552..cefd928021d 100644
--- a/src/backend/access/transam/subtrans.c
+++ b/src/backend/access/transam/subtrans.c
@@ -32,6 +32,7 @@
#include "access/subtrans.h"
#include "access/transam.h"
#include "pg_trace.h"
+#include "storage/bufmgr.h"
#include "utils/snapmgr.h"
@@ -55,15 +56,7 @@
#define TransactionIdToEntry(xid) ((xid) % (TransactionId) SUBTRANS_XACTS_PER_PAGE)
-/*
- * Link to shared-memory data structures for SUBTRANS control
- */
-static SlruCtlData SubTransCtlData;
-
-#define SubTransCtl (&SubTransCtlData)
-
-
-static int ZeroSUBTRANSPage(int pageno);
+static Buffer ZeroSUBTRANSPage(int pageno);
static bool SubTransPagePrecedes(int page1, int page2);
@@ -75,16 +68,15 @@ SubTransSetParent(TransactionId xid, TransactionId parent)
{
int pageno = TransactionIdToPage(xid);
int entryno = TransactionIdToEntry(xid);
- int slotno;
TransactionId *ptr;
+ Buffer buffer;
Assert(TransactionIdIsValid(parent));
Assert(TransactionIdFollows(xid, parent));
- LWLockAcquire(SubtransSLRULock, LW_EXCLUSIVE);
-
- slotno = SimpleLruReadPage(SubTransCtl, pageno, true, xid);
- ptr = (TransactionId *) SubTransCtl->shared->page_buffer[slotno];
+ buffer = ReadSlruBuffer(SLRU_SUBTRANS_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ ptr = (TransactionId *) BufferGetPage(buffer);
ptr += entryno;
/*
@@ -96,10 +88,10 @@ SubTransSetParent(TransactionId xid, TransactionId parent)
{
Assert(*ptr == InvalidTransactionId);
*ptr = parent;
- SubTransCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
}
- LWLockRelease(SubtransSLRULock);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -110,9 +102,9 @@ SubTransGetParent(TransactionId xid)
{
int pageno = TransactionIdToPage(xid);
int entryno = TransactionIdToEntry(xid);
- int slotno;
TransactionId *ptr;
TransactionId parent;
+ Buffer buffer;
/* Can't ask about stuff that might not be around anymore */
Assert(TransactionIdFollowsOrEquals(xid, TransactionXmin));
@@ -121,15 +113,14 @@ SubTransGetParent(TransactionId xid)
if (!TransactionIdIsNormal(xid))
return InvalidTransactionId;
- /* lock is acquired by SimpleLruReadPage_ReadOnly */
+ buffer = ReadSlruBuffer(SLRU_SUBTRANS_ID, pageno);
- slotno = SimpleLruReadPage_ReadOnly(SubTransCtl, pageno, xid);
- ptr = (TransactionId *) SubTransCtl->shared->page_buffer[slotno];
+ ptr = (TransactionId *) BufferGetPage(buffer);
ptr += entryno;
parent = *ptr;
- LWLockRelease(SubtransSLRULock);
+ ReleaseBuffer(buffer);
return parent;
}
@@ -177,26 +168,6 @@ SubTransGetTopmostTransaction(TransactionId xid)
return previousXid;
}
-
-/*
- * Initialization of shared memory for SUBTRANS
- */
-Size
-SUBTRANSShmemSize(void)
-{
- return SimpleLruShmemSize(NUM_SUBTRANS_BUFFERS, 0);
-}
-
-void
-SUBTRANSShmemInit(void)
-{
- SubTransCtl->PagePrecedes = SubTransPagePrecedes;
- SimpleLruInit(SubTransCtl, "Subtrans", NUM_SUBTRANS_BUFFERS, 0,
- SubtransSLRULock, "pg_subtrans",
- LWTRANCHE_SUBTRANS_BUFFER, SYNC_HANDLER_NONE);
- SlruPagePrecedesUnitTests(SubTransCtl, SUBTRANS_XACTS_PER_PAGE);
-}
-
/*
* This func must be called ONCE on system install. It creates
* the initial SUBTRANS segment. (The SUBTRANS directory is assumed to
@@ -210,18 +181,16 @@ SUBTRANSShmemInit(void)
void
BootStrapSUBTRANS(void)
{
- int slotno;
+ Buffer buffer;
- LWLockAcquire(SubtransSLRULock, LW_EXCLUSIVE);
+ SlruPagePrecedesUnitTests(SubTransPagePrecedes, SUBTRANS_XACTS_PER_PAGE);
/* Create and zero the first page of the subtrans log */
- slotno = ZeroSUBTRANSPage(0);
+ buffer = ZeroSUBTRANSPage(0);
/* Make sure it's written out */
- SimpleLruWritePage(SubTransCtl, slotno);
- Assert(!SubTransCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(SubtransSLRULock);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -232,10 +201,15 @@ BootStrapSUBTRANS(void)
*
* Control lock must be held at entry, and will be held at exit.
*/
-static int
+static Buffer
ZeroSUBTRANSPage(int pageno)
{
- return SimpleLruZeroPage(SubTransCtl, pageno);
+ Buffer buffer;
+
+ buffer = ZeroSlruBuffer(SLRU_SUBTRANS_ID, pageno);
+ MarkBufferDirty(buffer);
+
+ return buffer;
}
/*
@@ -258,7 +232,6 @@ StartupSUBTRANS(TransactionId oldestActiveXID)
* Whenever we advance into a new page, ExtendSUBTRANS will likewise zero
* the new page without regard to whatever was previously on disk.
*/
- LWLockAcquire(SubtransSLRULock, LW_EXCLUSIVE);
startPage = TransactionIdToPage(oldestActiveXID);
nextXid = ShmemVariableCache->nextXid;
@@ -266,36 +239,15 @@ StartupSUBTRANS(TransactionId oldestActiveXID)
while (startPage != endPage)
{
- (void) ZeroSUBTRANSPage(startPage);
+ UnlockReleaseBuffer(ZeroSUBTRANSPage(startPage));
startPage++;
/* must account for wraparound */
if (startPage > TransactionIdToPage(MaxTransactionId))
startPage = 0;
}
- (void) ZeroSUBTRANSPage(startPage);
-
- LWLockRelease(SubtransSLRULock);
+ UnlockReleaseBuffer(ZeroSUBTRANSPage(startPage));
}
-/*
- * Perform a checkpoint --- either during shutdown, or on-the-fly
- */
-void
-CheckPointSUBTRANS(void)
-{
- /*
- * Write dirty SUBTRANS pages to disk
- *
- * This is not actually necessary from a correctness point of view. We do
- * it merely to improve the odds that writing of dirty pages is done by
- * the checkpoint process and not by backends.
- */
- TRACE_POSTGRESQL_SUBTRANS_CHECKPOINT_START(true);
- SimpleLruWriteAll(SubTransCtl, true);
- TRACE_POSTGRESQL_SUBTRANS_CHECKPOINT_DONE(true);
-}
-
-
/*
* Make sure that SUBTRANS has room for a newly-allocated XID.
*
@@ -319,12 +271,8 @@ ExtendSUBTRANS(TransactionId newestXact)
pageno = TransactionIdToPage(newestXact);
- LWLockAcquire(SubtransSLRULock, LW_EXCLUSIVE);
-
/* Zero the page */
- ZeroSUBTRANSPage(pageno);
-
- LWLockRelease(SubtransSLRULock);
+ UnlockReleaseBuffer(ZeroSUBTRANSPage(pageno));
}
@@ -350,7 +298,7 @@ TruncateSUBTRANS(TransactionId oldestXact)
TransactionIdRetreat(oldestXact);
cutoffPage = TransactionIdToPage(oldestXact);
- SimpleLruTruncate(SubTransCtl, cutoffPage);
+ SimpleLruTruncate(SLRU_SUBTRANS_ID, SubTransPagePrecedes, cutoffPage);
}
diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 116de1175b2..61c25bc0a82 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -67,6 +67,7 @@
#include "utils/inval.h"
#include "utils/memutils.h"
#include "utils/relmapper.h"
+#include "utils/resowner_private.h"
#include "utils/snapmgr.h"
#include "utils/timeout.h"
#include "utils/timestamp.h"
@@ -1393,6 +1394,7 @@ RecordTransactionCommit(void)
* are delaying the checkpoint a bit fuzzy, but it doesn't matter.
*/
Assert((MyProc->delayChkptFlags & DELAY_CHKPT_START) == 0);
+
START_CRIT_SECTION();
MyProc->delayChkptFlags |= DELAY_CHKPT_START;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15ab8d90d4b..e9506175239 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -4501,6 +4501,7 @@ BootStrapXLOG(void)
uint64 sysidentifier;
struct timeval tv;
pg_crc32c crc;
+ ResourceOwner resowner;
/* allow ordinary WAL segment creation, like StartupXLOG() would */
LWLockAcquire(ControlFileLock, LW_EXCLUSIVE);
@@ -4642,10 +4643,14 @@ BootStrapXLOG(void)
WriteControlFile();
/* Bootstrap the commit log, too */
+ resowner = ResourceOwnerCreate(NULL, "bootstrap resowner");
+ CurrentResourceOwner = resowner;
BootStrapCLOG();
BootStrapCommitTs();
BootStrapSUBTRANS();
BootStrapMultiXact();
+ CurrentResourceOwner = NULL;
+ ResourceOwnerDelete(resowner);
pfree(buffer);
@@ -4654,6 +4659,8 @@ BootStrapXLOG(void)
* otherwise never run the checks and GUC related initializations therein.
*/
ReadControlFile();
+
+ smgrcloseall();
}
static char *
@@ -6851,15 +6858,11 @@ CheckPointGuts(XLogRecPtr checkPointRedo, int flags)
CheckPointSnapBuild();
CheckPointLogicalRewriteHeap();
CheckPointReplicationOrigin();
+ CheckPointPredicate();
- /* Write out all dirty data in SLRUs and the main buffer pool */
+ /* Write out all dirty data in the buffer pool */
TRACE_POSTGRESQL_BUFFER_CHECKPOINT_START(flags);
CheckpointStats.ckpt_write_t = GetCurrentTimestamp();
- CheckPointCLOG();
- CheckPointCommitTs();
- CheckPointSUBTRANS();
- CheckPointMultiXact();
- CheckPointPredicate();
CheckPointBuffers(flags);
/* Perform all queued up fsyncs */
diff --git a/src/backend/commands/async.c b/src/backend/commands/async.c
index 3e1b92df030..760705caf90 100644
--- a/src/backend/commands/async.c
+++ b/src/backend/commands/async.c
@@ -141,6 +141,7 @@
#include "libpq/libpq.h"
#include "libpq/pqformat.h"
#include "miscadmin.h"
+#include "storage/bufmgr.h"
#include "storage/ipc.h"
#include "storage/lmgr.h"
#include "storage/proc.h"
@@ -305,12 +306,6 @@ static AsyncQueueControl *asyncQueueControl;
#define QUEUE_NEXT_LISTENER(i) (asyncQueueControl->backend[i].nextListener)
#define QUEUE_BACKEND_POS(i) (asyncQueueControl->backend[i].pos)
-/*
- * The SLRU buffer area through which we access the notification queue
- */
-static SlruCtlData NotifyCtlData;
-
-#define NotifyCtl (&NotifyCtlData)
#define QUEUE_PAGESIZE BLCKSZ
#define QUEUE_FULL_WARN_INTERVAL 5000 /* warn at most once every 5s */
@@ -521,8 +516,6 @@ AsyncShmemSize(void)
size = mul_size(MaxBackends + 1, sizeof(QueueBackendStatus));
size = add_size(size, offsetof(AsyncQueueControl, backend));
- size = add_size(size, SimpleLruShmemSize(NUM_NOTIFY_BUFFERS, 0));
-
return size;
}
@@ -565,20 +558,13 @@ AsyncShmemInit(void)
}
}
- /*
- * Set up SLRU management of the pg_notify data.
- */
- NotifyCtl->PagePrecedes = asyncQueuePagePrecedes;
- SimpleLruInit(NotifyCtl, "Notify", NUM_NOTIFY_BUFFERS, 0,
- NotifySLRULock, "pg_notify", LWTRANCHE_NOTIFY_BUFFER,
- SYNC_HANDLER_NONE);
-
if (!found)
{
/*
* During start or reboot, clean out the pg_notify directory.
*/
- (void) SlruScanDirectory(NotifyCtl, SlruScanDirCbDeleteAll, NULL);
+ (void) SlruScanDirectory(SLRU_NOTIFY_ID, asyncQueuePagePrecedes,
+ SlruScanDirCbDeleteAll, NULL);
}
}
@@ -1411,10 +1397,7 @@ asyncQueueAddEntries(ListCell *nextNotify)
QueuePosition queue_head;
int pageno;
int offset;
- int slotno;
-
- /* We hold both NotifyQueueLock and NotifySLRULock during this operation */
- LWLockAcquire(NotifySLRULock, LW_EXCLUSIVE);
+ Buffer buffer;
/*
* We work with a local copy of QUEUE_HEAD, which we write back to shared
@@ -1439,13 +1422,17 @@ asyncQueueAddEntries(ListCell *nextNotify)
*/
pageno = QUEUE_POS_PAGE(queue_head);
if (QUEUE_POS_IS_ZERO(queue_head))
- slotno = SimpleLruZeroPage(NotifyCtl, pageno);
+ {
+ buffer = ZeroSlruBuffer(SLRU_NOTIFY_ID, pageno);
+ }
else
- slotno = SimpleLruReadPage(NotifyCtl, pageno, true,
- InvalidTransactionId);
+ {
+ buffer = ReadSlruBuffer(SLRU_NOTIFY_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ }
/* Note we mark the page dirty before writing in it */
- NotifyCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
while (nextNotify != NULL)
{
@@ -1476,7 +1463,7 @@ asyncQueueAddEntries(ListCell *nextNotify)
}
/* Now copy qe into the shared buffer page */
- memcpy(NotifyCtl->shared->page_buffer[slotno] + offset,
+ memcpy(BufferGetPage(buffer) + offset,
&qe,
qe.length);
@@ -1491,7 +1478,10 @@ asyncQueueAddEntries(ListCell *nextNotify)
* asyncQueueIsFull() ensured that there is room to create this
* page without overrunning the queue.
*/
- slotno = SimpleLruZeroPage(NotifyCtl, QUEUE_POS_PAGE(queue_head));
+ UnlockReleaseBuffer(buffer);
+ buffer = ZeroSlruBuffer(SLRU_NOTIFY_ID,
+ QUEUE_POS_PAGE(queue_head));
+ MarkBufferDirty(buffer);
/*
* If the new page address is a multiple of QUEUE_CLEANUP_DELAY,
@@ -1505,12 +1495,11 @@ asyncQueueAddEntries(ListCell *nextNotify)
break;
}
}
+ UnlockReleaseBuffer(buffer);
/* Success, so update the global QUEUE_HEAD */
QUEUE_HEAD = queue_head;
- LWLockRelease(NotifySLRULock);
-
return nextNotify;
}
@@ -1983,17 +1972,16 @@ asyncQueueReadAllNotifications(void)
{
int curpage = QUEUE_POS_PAGE(pos);
int curoffset = QUEUE_POS_OFFSET(pos);
- int slotno;
int copysize;
+ Buffer buffer;
/*
- * We copy the data from SLRU into a local buffer, so as to avoid
- * holding the NotifySLRULock while we are examining the entries
- * and possibly transmitting them to our frontend. Copy only the
- * part of the page we will actually inspect.
+ * We copy the data into a local buffer, so as to avoid holding a
+ * buffer pin while we are examining the entries and possibly
+ * transmitting them to our frontend. Copy only the part of the
+ * page we will actually inspect.
*/
- slotno = SimpleLruReadPage_ReadOnly(NotifyCtl, curpage,
- InvalidTransactionId);
+ buffer = ReadSlruBuffer(SLRU_NOTIFY_ID, curpage);
if (curpage == QUEUE_POS_PAGE(head))
{
/* we only want to read as far as head */
@@ -2007,10 +1995,9 @@ asyncQueueReadAllNotifications(void)
copysize = QUEUE_PAGESIZE - curoffset;
}
memcpy(page_buffer.buf + curoffset,
- NotifyCtl->shared->page_buffer[slotno] + curoffset,
+ BufferGetPage(buffer) + curoffset,
copysize);
- /* Release lock that we got from SimpleLruReadPage_ReadOnly() */
- LWLockRelease(NotifySLRULock);
+ ReleaseBuffer(buffer);
/*
* Process messages up to the stop position, end of page, or an
@@ -2207,7 +2194,7 @@ asyncQueueAdvanceTail(void)
* SimpleLruTruncate() will ask for NotifySLRULock but will also
* release the lock again.
*/
- SimpleLruTruncate(NotifyCtl, newtailpage);
+ SimpleLruTruncate(SLRU_NOTIFY_ID, asyncQueuePagePrecedes, newtailpage);
/*
* Update QUEUE_STOP_PAGE. This changes asyncQueueIsFull()'s verdict
diff --git a/src/backend/storage/buffer/buf_init.c b/src/backend/storage/buffer/buf_init.c
index 2862e9e412c..719f27a86a4 100644
--- a/src/backend/storage/buffer/buf_init.c
+++ b/src/backend/storage/buffer/buf_init.c
@@ -20,6 +20,7 @@
BufferDescPadded *BufferDescriptors;
char *BufferBlocks;
+XLogRecPtr *BufferExternalLSNs;
ConditionVariableMinimallyPadded *BufferIOCVArray;
WritebackContext BackendWritebackContext;
CkptSortItem *CkptBufferIds;
@@ -69,9 +70,11 @@ InitBufferPool(void)
{
bool foundBufs,
foundDescs,
+ foundLSNs,
foundIOCV,
foundBufCkpt;
+
/* Align descriptors to a cacheline boundary. */
BufferDescriptors = (BufferDescPadded *)
ShmemInitStruct("Buffer Descriptors",
@@ -88,6 +91,11 @@ InitBufferPool(void)
NBuffers * sizeof(ConditionVariableMinimallyPadded),
&foundIOCV);
+ BufferExternalLSNs = (XLogRecPtr *)
+ ShmemInitStruct("Buffer External LSNs",
+ NBuffers * sizeof(XLogRecPtr),
+ &foundLSNs);
+
/*
* The array used to sort to-be-checkpointed buffer ids is located in
* shared memory, to avoid having to allocate significant amounts of
@@ -99,10 +107,10 @@ InitBufferPool(void)
ShmemInitStruct("Checkpoint BufferIds",
NBuffers * sizeof(CkptSortItem), &foundBufCkpt);
- if (foundDescs || foundBufs || foundIOCV || foundBufCkpt)
+ if (foundDescs || foundBufs || foundIOCV || foundBufCkpt || foundLSNs)
{
/* should find all of these, or none of them */
- Assert(foundDescs && foundBufs && foundIOCV && foundBufCkpt);
+ Assert(foundDescs && foundBufs && foundIOCV && foundBufCkpt && foundLSNs);
/* note: this path is only taken in EXEC_BACKEND case */
}
else
@@ -133,6 +141,8 @@ InitBufferPool(void)
LWTRANCHE_BUFFER_CONTENT);
ConditionVariableInit(BufferDescriptorGetIOCV(buf));
+
+ BufferExternalLSNs[i] = InvalidXLogRecPtr;
}
/* Correct last entry of linked list */
@@ -166,6 +176,9 @@ BufferShmemSize(void)
/* size of data pages */
size = add_size(size, mul_size(NBuffers, BLCKSZ));
+ /* size of external LSNs */
+ size = add_size(size, mul_size(NBuffers, sizeof(XLogRecPtr)));
+
/* size of stuff controlled by freelist.c */
size = add_size(size, StrategyShmemSize());
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 2ba75f676ad..fa060864773 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -57,10 +57,21 @@
#include "utils/resowner_private.h"
#include "utils/timestamp.h"
+/*
+ * XXX Ideally we'd switch to standard pages for SLRU data, but in the
+ * meantime we need some way to identify buffers that hold raw data (no
+ * invasive LSN, no checksums).
+ */
+#define BufferHasStandardPage(bufHdr) \
+ ((bufHdr)->tag.rlocator.spcOid != 9)
+
+#define BufferHasExternalLSN(bufHdr) \
+ !BufferHasStandardPage(bufHdr)
/* Note: these two macros only work on shared buffers, not local ones! */
#define BufHdrGetBlock(bufHdr) ((Block) (BufferBlocks + ((Size) (bufHdr)->buf_id) * BLCKSZ))
-#define BufferGetLSN(bufHdr) (PageGetLSN(BufHdrGetBlock(bufHdr)))
+#define BufferGetLSN(bufHdr) \
+ (BufferHasExternalLSN(bufHdr) ? BufferGetExternalLSN(bufHdr) : PageGetLSN(BufHdrGetBlock(bufHdr)))
/* Note: this macro only works on local buffers, not shared ones! */
#define LocalBufHdrGetBlock(bufHdr) \
@@ -794,6 +805,18 @@ ReadBufferWithoutRelcache(RelFileLocator rlocator, ForkNumber forkNum,
mode, strategy, &hit);
}
+Buffer
+ReadBufferWithoutRelcacheWithHit(RelFileLocator rlocator, ForkNumber forkNum,
+ BlockNumber blockNum, ReadBufferMode mode,
+ BufferAccessStrategy strategy, bool permanent, bool *hit)
+{
+ SMgrFileHandle sfile = smgropen(rlocator, InvalidBackendId, forkNum);
+
+ return ReadBuffer_common(sfile, permanent ? RELPERSISTENCE_PERMANENT :
+ RELPERSISTENCE_UNLOGGED, blockNum,
+ mode, strategy, hit);
+}
+
/*
* ReadBuffer_common -- common logic for all ReadBuffer variants
@@ -1016,7 +1039,8 @@ ReadBuffer_common(SMgrFileHandle sfile, char relpersistence,
}
/* check for garbage data */
- if (!PageIsVerifiedExtended((Page) bufBlock, blockNum,
+ if (BufferHasStandardPage(bufHdr) &&
+ !PageIsVerifiedExtended((Page) bufBlock, blockNum,
PIV_LOG_WARNING | PIV_REPORT_STAT))
{
if (mode == RBM_ZERO_ON_ERROR || zero_damaged_pages)
@@ -1440,6 +1464,9 @@ BufferAlloc(SMgrFileHandle sfile, char relpersistence,
LWLockRelease(newPartitionLock);
+ if (BufferHasExternalLSN(buf))
+ BufferSetExternalLSN(buf, InvalidXLogRecPtr);
+
/*
* Buffer contents are currently invalid. Try to obtain the right to
* start I/O. If StartBufferIO returns false, then someone else managed
@@ -3018,7 +3045,10 @@ BufferGetLSNAtomic(Buffer buffer)
Assert(BufferIsPinned(buffer));
buf_state = LockBufHdr(bufHdr);
- lsn = PageGetLSN(page);
+ if (BufferHasStandardPage(bufHdr))
+ lsn = PageGetLSN(page);
+ else
+ lsn = BufferGetExternalLSN(bufHdr);
UnlockBufHdr(bufHdr, buf_state);
return lsn;
@@ -4989,3 +5019,29 @@ TestForOldSnapshot_impl(Snapshot snapshot, Relation relation)
(errcode(ERRCODE_SNAPSHOT_TOO_OLD),
errmsg("snapshot too old")));
}
+
+/*
+ * Check if a buffer tag is currently mapped.
+ *
+ * XXX Dubious semantics; needed only for multixact's handling for
+ * inconsistent states.
+ */
+bool
+BufferProbe(RelFileLocator rlocator, ForkNumber forkNum, BlockNumber blockNum)
+{
+ BufferTag tag;
+ uint32 hash;
+ LWLock *partitionLock;
+ int buf_id;
+
+ INIT_BUFFERTAG(tag, rlocator, forkNum, blockNum);
+
+ hash = BufTableHashCode(&tag);
+ partitionLock = BufMappingPartitionLock(hash);
+
+ LWLockAcquire(partitionLock, LW_SHARED);
+ buf_id = BufTableLookup(&tag, hash);
+ LWLockRelease(partitionLock);
+
+ return buf_id >= 0;
+}
diff --git a/src/backend/storage/ipc/ipci.c b/src/backend/storage/ipc/ipci.c
index 1a6f5270518..c063a201a4f 100644
--- a/src/backend/storage/ipc/ipci.c
+++ b/src/backend/storage/ipc/ipci.c
@@ -116,9 +116,7 @@ CalculateShmemSize(int *num_semaphores)
size = add_size(size, XLogPrefetchShmemSize());
size = add_size(size, XLOGShmemSize());
size = add_size(size, XLogRecoveryShmemSize());
- size = add_size(size, CLOGShmemSize());
size = add_size(size, CommitTsShmemSize());
- size = add_size(size, SUBTRANSShmemSize());
size = add_size(size, TwoPhaseShmemSize());
size = add_size(size, BackgroundWorkerShmemSize());
size = add_size(size, MultiXactShmemSize());
@@ -240,9 +238,7 @@ CreateSharedMemoryAndSemaphores(void)
XLOGShmemInit();
XLogPrefetchShmemInit();
XLogRecoveryShmemInit();
- CLOGShmemInit();
CommitTsShmemInit();
- SUBTRANSShmemInit();
MultiXactShmemInit();
InitBufferPool();
diff --git a/src/backend/storage/lmgr/predicate.c b/src/backend/storage/lmgr/predicate.c
index 5136da6ea36..91d2764890a 100644
--- a/src/backend/storage/lmgr/predicate.c
+++ b/src/backend/storage/lmgr/predicate.c
@@ -312,14 +312,6 @@
((targethash) ^ ((uint32) PointerGetDatum((predicatelocktag)->myXact)) \
<< LOG2_NUM_PREDICATELOCK_PARTITIONS)
-
-/*
- * The SLRU buffer area through which we access the old xids.
- */
-static SlruCtlData SerialSlruCtlData;
-
-#define SerialSlruCtl (&SerialSlruCtlData)
-
#define SERIAL_PAGESIZE BLCKSZ
#define SERIAL_ENTRYSIZE sizeof(SerCommitSeqNo)
#define SERIAL_ENTRIESPERPAGE (SERIAL_PAGESIZE / SERIAL_ENTRYSIZE)
@@ -331,8 +323,8 @@ static SlruCtlData SerialSlruCtlData;
#define SerialNextPage(page) (((page) >= SERIAL_MAX_PAGE) ? 0 : (page) + 1)
-#define SerialValue(slotno, xid) (*((SerCommitSeqNo *) \
- (SerialSlruCtl->shared->page_buffer[slotno] + \
+#define SerialValue(buffer, xid) (*((SerCommitSeqNo *) \
+ (BufferGetPage(buffer) + \
((((uint32) (xid)) % SERIAL_ENTRIESPERPAGE) * SERIAL_ENTRYSIZE))))
#define SerialPage(xid) (((uint32) (xid)) / SERIAL_ENTRIESPERPAGE)
@@ -867,17 +859,10 @@ SerialInit(void)
{
bool found;
- /*
- * Set up SLRU management of the pg_serial data.
- */
- SerialSlruCtl->PagePrecedes = SerialPagePrecedesLogically;
- SimpleLruInit(SerialSlruCtl, "Serial",
- NUM_SERIAL_BUFFERS, 0, SerialSLRULock, "pg_serial",
- LWTRANCHE_SERIAL_BUFFER, SYNC_HANDLER_NONE);
#ifdef USE_ASSERT_CHECKING
SerialPagePrecedesLogicallyUnitTests();
#endif
- SlruPagePrecedesUnitTests(SerialSlruCtl, SERIAL_ENTRIESPERPAGE);
+ SlruPagePrecedesUnitTests(SerialPagePrecedesLogically, SERIAL_ENTRIESPERPAGE);
/*
* Create or attach to the SerialControl structure.
@@ -907,9 +892,9 @@ SerialAdd(TransactionId xid, SerCommitSeqNo minConflictCommitSeqNo)
{
TransactionId tailXid;
int targetPage;
- int slotno;
int firstZeroPage;
bool isNewPage;
+ Buffer buffer;
Assert(TransactionIdIsValid(xid));
@@ -954,16 +939,22 @@ SerialAdd(TransactionId xid, SerCommitSeqNo minConflictCommitSeqNo)
/* Initialize intervening pages. */
while (firstZeroPage != targetPage)
{
- (void) SimpleLruZeroPage(SerialSlruCtl, firstZeroPage);
+ buffer = ZeroSlruBuffer(SLRU_SERIAL_ID, firstZeroPage);
+ MarkBufferDirty(buffer);
+ UnlockReleaseBuffer(buffer);
firstZeroPage = SerialNextPage(firstZeroPage);
}
- slotno = SimpleLruZeroPage(SerialSlruCtl, targetPage);
+ buffer = ZeroSlruBuffer(SLRU_SERIAL_ID, targetPage);
}
else
- slotno = SimpleLruReadPage(SerialSlruCtl, targetPage, true, xid);
+ {
+ buffer = ReadSlruBuffer(SLRU_SERIAL_ID, targetPage);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ }
- SerialValue(slotno, xid) = minConflictCommitSeqNo;
- SerialSlruCtl->shared->page_dirty[slotno] = true;
+ SerialValue(buffer, xid) = minConflictCommitSeqNo;
+ MarkBufferDirty(buffer);
+ UnlockReleaseBuffer(buffer);
LWLockRelease(SerialSLRULock);
}
@@ -979,7 +970,7 @@ SerialGetMinConflictCommitSeqNo(TransactionId xid)
TransactionId headXid;
TransactionId tailXid;
SerCommitSeqNo val;
- int slotno;
+ Buffer buffer;
Assert(TransactionIdIsValid(xid));
@@ -1001,9 +992,9 @@ SerialGetMinConflictCommitSeqNo(TransactionId xid)
* The following function must be called without holding SerialSLRULock,
* but will return with that lock held, which must then be released.
*/
- slotno = SimpleLruReadPage_ReadOnly(SerialSlruCtl,
- SerialPage(xid), xid);
- val = SerialValue(slotno, xid);
+ buffer = ReadSlruBuffer(SLRU_SERIAL_ID, SerialPage(xid));
+ val = SerialValue(buffer, xid);
+ ReleaseBuffer(buffer);
LWLockRelease(SerialSLRULock);
return val;
}
@@ -1122,19 +1113,7 @@ CheckPointPredicate(void)
LWLockRelease(SerialSLRULock);
/* Truncate away pages that are no longer required */
- SimpleLruTruncate(SerialSlruCtl, tailPage);
-
- /*
- * Write dirty SLRU pages to disk
- *
- * This is not actually necessary from a correctness point of view. We do
- * it merely as a debugging aid.
- *
- * We're doing this after the truncation to avoid writing pages right
- * before deleting the file in which they sit, which would be completely
- * pointless.
- */
- SimpleLruWriteAll(SerialSlruCtl, true);
+ SimpleLruTruncate(SLRU_SERIAL_ID, SerialPagePrecedesLogically, tailPage);
}
/*------------------------------------------------------------------------*/
@@ -1396,7 +1375,6 @@ PredicateLockShmemSize(void)
/* Shared memory structures for SLRU tracking of old committed xids. */
size = add_size(size, sizeof(SerialControlData));
- size = add_size(size, SimpleLruShmemSize(NUM_SERIAL_BUFFERS, 0));
return size;
}
diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c
index 4528e45a5ca..9fc3cb1caaf 100644
--- a/src/backend/storage/smgr/md.c
+++ b/src/backend/storage/smgr/md.c
@@ -191,9 +191,12 @@ mdcreate(SMgrFileHandle sfile, bool isRedo)
* should be here and not in commands/tablespace.c? But that would imply
* importing a lot of stuff that smgr.c oughtn't know, either.
*/
- TablespaceCreateDbspace(sfile->smgr_locator.locator.spcOid,
- sfile->smgr_locator.locator.dbOid,
- isRedo);
+ if (sfile->smgr_locator.locator.spcOid != SLRU_SPC_OID)
+ {
+ TablespaceCreateDbspace(sfile->smgr_locator.locator.spcOid,
+ sfile->smgr_locator.locator.dbOid,
+ isRedo);
+ }
path = smgrfilepath(sfile->smgr_locator);
diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c
index 6e1ad31654e..de103320574 100644
--- a/src/backend/storage/smgr/smgr.c
+++ b/src/backend/storage/smgr/smgr.c
@@ -419,6 +419,49 @@ smgrunlink_multi(RelFileLocator rlocator, BackendId backend, ForkNumber *forks,
}
}
+/*
+ * smgrdounlink() -- Immediately unlink a file
+ *
+ * If isRedo is true, it is okay for the underlying file(s) to be gone
+ * already.
+ *
+ * To remove a relation transactionally, see RelationDropStorage() instead.
+ * This will cause cache invalidation of all forks of the relation, not just
+ * this one.
+ */
+void
+smgrunlink(SMgrFileHandle sfile, bool isRedo)
+{
+ SMgrFileLocator locator;
+ int which;
+
+ /* remember before closing */
+ which = sfile->smgr_which;
+ locator = sfile->smgr_locator;
+
+ /* Close the file at smgr level */
+ smgrclose(sfile);
+
+ /*
+ * Send a shared-inval message to force other backends to close any
+ * dangling smgr references they may have for these rels. We should do
+ * this before starting the actual unlinking, in case we fail partway
+ * through that step. Note that the sinval messages will eventually come
+ * back to this backend, too, and thereby provide a backstop that we
+ * closed our own smgr rel.
+ */
+ CacheInvalidateSmgr(locator.locator, locator.backend);
+
+ /*
+ * Delete the physical file(s).
+ *
+ * Note: smgr_unlink must treat deletion failure as a WARNING, not an
+ * ERROR, because we've already decided to commit or abort the current
+ * xact.
+ */
+ smgrsw[which].smgr_unlink(locator, isRedo);
+}
+
/*
* smgrextend() -- Add a new block to a file.
diff --git a/src/backend/storage/sync/sync.c b/src/backend/storage/sync/sync.c
index 9d6a9e91090..3572078ba79 100644
--- a/src/backend/storage/sync/sync.c
+++ b/src/backend/storage/sync/sync.c
@@ -18,9 +18,7 @@
#include <fcntl.h>
#include <sys/file.h>
-#include "access/commit_ts.h"
-#include "access/clog.h"
-#include "access/multixact.h"
+#include "access/slru.h"
#include "access/xlog.h"
#include "access/xlogutils.h"
#include "commands/tablespace.h"
@@ -106,22 +104,6 @@ static const SyncOps syncsw[] = {
.sync_unlinkfiletag = mdunlinkfiletag,
.sync_filetagmatches = mdfiletagmatches
},
- /* pg_xact */
- [SYNC_HANDLER_CLOG] = {
- .sync_syncfiletag = clogsyncfiletag
- },
- /* pg_commit_ts */
- [SYNC_HANDLER_COMMIT_TS] = {
- .sync_syncfiletag = committssyncfiletag
- },
- /* pg_multixact/offsets */
- [SYNC_HANDLER_MULTIXACT_OFFSET] = {
- .sync_syncfiletag = multixactoffsetssyncfiletag
- },
- /* pg_multixact/members */
- [SYNC_HANDLER_MULTIXACT_MEMBER] = {
- .sync_syncfiletag = multixactmemberssyncfiletag
- }
};
/*
diff --git a/src/backend/utils/mmgr/mcxt.c b/src/backend/utils/mmgr/mcxt.c
index e12be1b9bd8..f0255f895c6 100644
--- a/src/backend/utils/mmgr/mcxt.c
+++ b/src/backend/utils/mmgr/mcxt.c
@@ -69,9 +69,20 @@ static void MemoryContextStatsPrint(MemoryContext context, void *passthru,
* You should not do memory allocations within a critical section, because
* an out-of-memory error will be escalated to a PANIC. To enforce that
* rule, the allocation functions Assert that.
+ *
+ * FIXME: bypass this for the critical section in RecordTransactionCommit()
+ * for now. It does a lot of things that can allocate:
+ * - calls TransactionIdCommitTree, which pins buffers, which requires
+ * space in the ResourceOwner for the pin (ResourceOwnerEnlargeBuffers())
+ * - same for TransactionTreeSetCommitTsData() call.
+ * - reading a page can require flushing other pages, which in turn
+ * can call CompactCheckpointerRequestQueue(), which allocates
+ * - reading a page calls smgropen(), which allocates the SMgrFile entry
+ * if it's not open already
*/
#define AssertNotInCriticalSection(context) \
- Assert(CritSectionCount == 0 || (context)->allowInCritSection)
+ Assert(CritSectionCount == 0 || (context)->allowInCritSection || \
+ (MyProc != NULL && (MyProc->delayChkptFlags & DELAY_CHKPT_START != 0)))
/*****************************************************************************
diff --git a/src/common/relpath.c b/src/common/relpath.c
index ae87843ae7e..007b0898049 100644
--- a/src/common/relpath.c
+++ b/src/common/relpath.c
@@ -22,6 +22,16 @@
#include "common/relpath.h"
#include "storage/backendid.h"
+/*
+ * SLRU ID to path mapping
+ */
+#define PG_SLRU(symname,name,path,synchronize) \
+ path,
+
+static char *slru_dirs[] =
+{
+#include "access/slrulist.h"
+};
/*
* Lookup table of fork name by fork number.
@@ -143,7 +153,22 @@ GetSMgrFilePath(Oid dbOid, Oid spcOid, RelFileNumber relNumber,
{
char *path;
- if (spcOid == GLOBALTABLESPACE_OID)
+ if (spcOid == SLRU_SPC_OID)
+ {
+ if (dbOid >= lengthof(slru_dirs) || forkNumber != 0 || backendId != InvalidBackendId)
+ {
+#ifndef FRONTEND
+ ereport(ERROR,
+ (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
+ errmsg("invalid SLRU file locator %u/%u/%u/%u/%u",
+ spcOid, dbOid, relNumber, backendId, forkNumber)));
+#else
+ return NULL;
+#endif
+ }
+ path = psprintf("%s/%u", slru_dirs[dbOid], relNumber);
+ }
+ else if (spcOid == GLOBALTABLESPACE_OID)
{
/* Shared system relations live in {datadir}/global */
Assert(dbOid == 0);
diff --git a/src/include/access/clog.h b/src/include/access/clog.h
index 543f2e2643a..af16a2afcb9 100644
--- a/src/include/access/clog.h
+++ b/src/include/access/clog.h
@@ -40,18 +40,12 @@ extern void TransactionIdSetTreeStatus(TransactionId xid, int nsubxids,
TransactionId *subxids, XidStatus status, XLogRecPtr lsn);
extern XidStatus TransactionIdGetStatus(TransactionId xid, XLogRecPtr *lsn);
-extern Size CLOGShmemBuffers(void);
-extern Size CLOGShmemSize(void);
-extern void CLOGShmemInit(void);
extern void BootStrapCLOG(void);
extern void StartupCLOG(void);
extern void TrimCLOG(void);
-extern void CheckPointCLOG(void);
extern void ExtendCLOG(TransactionId newestXact);
extern void TruncateCLOG(TransactionId oldestXact, Oid oldestxid_datoid);
-extern int clogsyncfiletag(const FileTag *ftag, char *path);
-
/* XLOG stuff */
#define CLOG_ZEROPAGE 0x00
#define CLOG_TRUNCATE 0x10
diff --git a/src/include/access/commit_ts.h b/src/include/access/commit_ts.h
index 7662f8e1a9c..4e986fc023d 100644
--- a/src/include/access/commit_ts.h
+++ b/src/include/access/commit_ts.h
@@ -27,7 +27,6 @@ extern bool TransactionIdGetCommitTsData(TransactionId xid,
extern TransactionId GetLatestCommitTsData(TimestampTz *ts,
RepOriginId *nodeid);
-extern Size CommitTsShmemBuffers(void);
extern Size CommitTsShmemSize(void);
extern void CommitTsShmemInit(void);
extern void BootStrapCommitTs(void);
@@ -41,8 +40,6 @@ extern void SetCommitTsLimit(TransactionId oldestXact,
TransactionId newestXact);
extern void AdvanceOldestCommitTsXid(TransactionId oldestXact);
-extern int committssyncfiletag(const FileTag *ftag, char *path);
-
/* XLOG stuff */
#define COMMIT_TS_ZEROPAGE 0x00
#define COMMIT_TS_TRUNCATE 0x10
diff --git a/src/include/access/multixact.h b/src/include/access/multixact.h
index a5600a320ae..a6bbe53e1bb 100644
--- a/src/include/access/multixact.h
+++ b/src/include/access/multixact.h
@@ -118,9 +118,6 @@ extern bool MultiXactIdPrecedes(MultiXactId multi1, MultiXactId multi2);
extern bool MultiXactIdPrecedesOrEquals(MultiXactId multi1,
MultiXactId multi2);
-extern int multixactoffsetssyncfiletag(const FileTag *ftag, char *path);
-extern int multixactmemberssyncfiletag(const FileTag *ftag, char *path);
-
extern void AtEOXact_MultiXact(void);
extern void AtPrepare_MultiXact(void);
extern void PostPrepare_MultiXact(TransactionId xid);
diff --git a/src/include/access/slru.h b/src/include/access/slru.h
index 130c41c8632..ee1b2f40c7e 100644
--- a/src/include/access/slru.h
+++ b/src/include/access/slru.h
@@ -1,7 +1,7 @@
/*-------------------------------------------------------------------------
*
* slru.h
- * Simple LRU buffering for transaction status logfiles
+ * Buffering for transaction status logfiles
*
* Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group
* Portions Copyright (c) 1994, Regents of the University of California
@@ -14,9 +14,35 @@
#define SLRU_H
#include "access/xlogdefs.h"
+#include "catalog/pg_tablespace_d.h"
+#include "storage/buf.h"
#include "storage/lwlock.h"
+#include "storage/smgr.h"
#include "storage/sync.h"
+/* Pseudo database ID used for SLRU data. */
+#define SLRU_SPC_ID 9
+
+/* Pseudo database IDs used by each cache. */
+#define PG_SLRU(symname,name,path, synchronize) \
+ symname,
+
+typedef enum SlruIds
+{
+#include "access/slrulist.h"
+ SLRU_NEXT_ID
+} SlruIds;
+#undef PG_SLRU
+
+typedef bool (*SlruPagePrecedesFunction) (int, int);
+
+static inline RelFileLocator
+SlruRelFileLocator(uint32 slru_db_id, uint32 segment_id)
+{
+ RelFileLocator rlocator = {SLRU_SPC_ID, slru_db_id, segment_id};
+ return rlocator;
+}
+
/*
* Define SLRU segment size. A page is the same BLCKSZ as is used everywhere
@@ -33,142 +59,40 @@
*/
#define SLRU_PAGES_PER_SEGMENT 32
-/*
- * Page status codes. Note that these do not include the "dirty" bit.
- * page_dirty can be true only in the VALID or WRITE_IN_PROGRESS states;
- * in the latter case it implies that the page has been re-dirtied since
- * the write started.
- */
-typedef enum
-{
- SLRU_PAGE_EMPTY, /* buffer is not in use */
- SLRU_PAGE_READ_IN_PROGRESS, /* page is being read in */
- SLRU_PAGE_VALID, /* page is valid and not being written */
- SLRU_PAGE_WRITE_IN_PROGRESS /* page is being written out */
-} SlruPageStatus;
-
-/*
- * Shared-memory state
- */
-typedef struct SlruSharedData
-{
- LWLock *ControlLock;
-
- /* Number of buffers managed by this SLRU structure */
- int num_slots;
-
- /*
- * Arrays holding info for each buffer slot. Page number is undefined
- * when status is EMPTY, as is page_lru_count.
- */
- char **page_buffer;
- SlruPageStatus *page_status;
- bool *page_dirty;
- int *page_number;
- int *page_lru_count;
- LWLockPadded *buffer_locks;
-
- /*
- * Optional array of WAL flush LSNs associated with entries in the SLRU
- * pages. If not zero/NULL, we must flush WAL before writing pages (true
- * for pg_xact, false for multixact, pg_subtrans, pg_notify). group_lsn[]
- * has lsn_groups_per_page entries per buffer slot, each containing the
- * highest LSN known for a contiguous group of SLRU entries on that slot's
- * page.
- */
- XLogRecPtr *group_lsn;
- int lsn_groups_per_page;
-
- /*----------
- * We mark a page "most recently used" by setting
- * page_lru_count[slotno] = ++cur_lru_count;
- * The oldest page is therefore the one with the highest value of
- * cur_lru_count - page_lru_count[slotno]
- * The counts will eventually wrap around, but this calculation still
- * works as long as no page's age exceeds INT_MAX counts.
- *----------
- */
- int cur_lru_count;
-
- /*
- * latest_page_number is the page number of the current end of the log;
- * this is not critical data, since we use it only to avoid swapping out
- * the latest page.
- */
- int latest_page_number;
-
- /* SLRU's index for statistics purposes (might not be unique) */
- int slru_stats_idx;
-} SlruSharedData;
-
-typedef SlruSharedData *SlruShared;
-
-/*
- * SlruCtlData is an unshared structure that points to the active information
- * in shared memory.
- */
-typedef struct SlruCtlData
-{
- SlruShared shared;
-
- /*
- * Which sync handler function to use when handing sync requests over to
- * the checkpointer. SYNC_HANDLER_NONE to disable fsync (eg pg_notify).
- */
- SyncRequestHandler sync_handler;
-
- /*
- * Decide whether a page is "older" for truncation and as a hint for
- * evicting pages in LRU order. Return true if every entry of the first
- * argument is older than every entry of the second argument. Note that
- * !PagePrecedes(a,b) && !PagePrecedes(b,a) need not imply a==b; it also
- * arises when some entries are older and some are not. For SLRUs using
- * SimpleLruTruncate(), this must use modular arithmetic. (For others,
- * the behavior of this callback has no functional implications.) Use
- * SlruPagePrecedesUnitTests() in SLRUs meeting its criteria.
- */
- bool (*PagePrecedes) (int, int);
-
- /*
- * Dir is set during SimpleLruInit and does not change thereafter. Since
- * it's always the same, it doesn't need to be in shared memory.
- */
- char Dir[64];
-} SlruCtlData;
-
-typedef SlruCtlData *SlruCtl;
-
-
-extern Size SimpleLruShmemSize(int nslots, int nlsns);
-extern void SimpleLruInit(SlruCtl ctl, const char *name, int nslots, int nlsns,
- LWLock *ctllock, const char *subdir, int tranche_id,
- SyncRequestHandler sync_handler);
-extern int SimpleLruZeroPage(SlruCtl ctl, int pageno);
-extern int SimpleLruReadPage(SlruCtl ctl, int pageno, bool write_ok,
- TransactionId xid);
-extern int SimpleLruReadPage_ReadOnly(SlruCtl ctl, int pageno,
- TransactionId xid);
-extern void SimpleLruWritePage(SlruCtl ctl, int slotno);
-extern void SimpleLruWriteAll(SlruCtl ctl, bool allow_redirtied);
#ifdef USE_ASSERT_CHECKING
-extern void SlruPagePrecedesUnitTests(SlruCtl ctl, int per_page);
+extern void SlruPagePrecedesUnitTests(SlruPagePrecedesFunction PagePrecedes,
+ int per_page);
#else
#define SlruPagePrecedesUnitTests(ctl, per_page) do {} while (0)
#endif
-extern void SimpleLruTruncate(SlruCtl ctl, int cutoffPage);
-extern bool SimpleLruDoesPhysicalPageExist(SlruCtl ctl, int pageno);
+extern void SimpleLruTruncate(int slru_id, SlruPagePrecedesFunction PagePrecedes,
+ int cutoffPage);
+extern bool SimpleLruDoesPhysicalPageExist(int slru_id, int pageno);
-typedef bool (*SlruScanCallback) (SlruCtl ctl, char *filename, int segpage,
+typedef bool (*SlruScanCallback) (int slru_id,
+ SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage,
void *data);
-extern bool SlruScanDirectory(SlruCtl ctl, SlruScanCallback callback, void *data);
-extern void SlruDeleteSegment(SlruCtl ctl, int segno);
-
-extern int SlruSyncFileTag(SlruCtl ctl, const FileTag *ftag, char *path);
+extern bool SlruScanDirectory(int slru_id, SlruPagePrecedesFunction PagePrecedes,
+ SlruScanCallback callback, void *data);
+extern void SlruDeleteSegment(int slru_id, int segno);
/* SlruScanDirectory public callbacks */
-extern bool SlruScanDirCbReportPresence(SlruCtl ctl, char *filename,
+extern bool SlruScanDirCbReportPresence(int slru_id,
+ SlruPagePrecedesFunction PagePrecedes,
+ char *filename,
int segpage, void *data);
-extern bool SlruScanDirCbDeleteAll(SlruCtl ctl, char *filename, int segpage,
+extern bool SlruScanDirCbDeleteAll(int slru_id, SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage,
void *data);
+/* Buffer access */
+extern Buffer ReadSlruBuffer(int slru_id, int pageno);
+extern Buffer ZeroSlruBuffer(int slru_id, int pageno);
+extern bool ProbeSlruBuffer(int slru_id, int pageno);
+
+/* Interfaces use by stats view */
+extern Oid SlruRelIdByName(const char *name);
+extern const char *SlruName(int slru_id);
+
#endif /* SLRU_H */
diff --git a/src/include/access/slrulist.h b/src/include/access/slrulist.h
new file mode 100644
index 00000000000..c1289a1326c
--- /dev/null
+++ b/src/include/access/slrulist.h
@@ -0,0 +1,30 @@
+/*---------------------------------------------------------------------------
+ * slrulist.h
+ *
+ * The SLRU list is kept in its own source file for possible
+ * use by automatic tools. The exact representation of a rmgr is determined
+ * by the PG_SLRU macro, which is not defined in this file; it can be
+ * defined by the caller for special purposes.
+ *
+ * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group
+ * Portions Copyright (c) 1994, Regents of the University of California
+ *
+ * src/include/access/slrulist.h
+ *---------------------------------------------------------------------------
+ */
+
+/* there is deliberately not an #ifndef SLRULIST_H here */
+
+/*
+ * List of SLRU entries. Note that order of entries defines the
+ * numerical values of each SLRU's ID, which is used in in-memory structus.
+ */
+
+/* symbol name, textual name, path, synchronize */
+PG_SLRU(SLRU_CLOG_ID, "Xact", "pg_xact", true)
+PG_SLRU(SLRU_SUBTRANS_ID, "Subtrans", "pg_subtrans", false)
+PG_SLRU(SLRU_MULTIXACT_OFFSET_ID, "MultiXactOffset", "pg_multixact/offsets", true)
+PG_SLRU(SLRU_MULTIXACT_MEMBER_ID, "MultiXactMember", "pg_multixact/members", true)
+PG_SLRU(SLRU_COMMIT_TS_ID, "CommitTs", "pg_commit_ts", true)
+PG_SLRU(SLRU_SERIAL_ID, "Serial", "pg_serial", false)
+PG_SLRU(SLRU_NOTIFY_ID, "Notify", "pg_notify", false)
diff --git a/src/include/access/subtrans.h b/src/include/access/subtrans.h
index f94e116640b..4685a05bc92 100644
--- a/src/include/access/subtrans.h
+++ b/src/include/access/subtrans.h
@@ -18,11 +18,8 @@ extern void SubTransSetParent(TransactionId xid, TransactionId parent);
extern TransactionId SubTransGetParent(TransactionId xid);
extern TransactionId SubTransGetTopmostTransaction(TransactionId xid);
-extern Size SUBTRANSShmemSize(void);
-extern void SUBTRANSShmemInit(void);
extern void BootStrapSUBTRANS(void);
extern void StartupSUBTRANS(TransactionId oldestActiveXID);
-extern void CheckPointSUBTRANS(void);
extern void ExtendSUBTRANS(TransactionId newestXact);
extern void TruncateSUBTRANS(TransactionId oldestXact);
diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h
index be92980fb34..1ed0c8ed127 100644
--- a/src/include/common/relpath.h
+++ b/src/include/common/relpath.h
@@ -56,6 +56,9 @@ typedef enum ForkNumber
#define FORKNAMECHARS 4 /* max chars for a fork name */
+/* Pseudo tablespace ID used for SLRUs. */
+#define SLRU_SPC_OID 9
+
extern PGDLLIMPORT const char *const forkNames[];
extern ForkNumber forkname_to_number(const char *forkName);
diff --git a/src/include/storage/buf_internals.h b/src/include/storage/buf_internals.h
index 4e232e5d794..a93a7771ca8 100644
--- a/src/include/storage/buf_internals.h
+++ b/src/include/storage/buf_internals.h
@@ -230,6 +230,11 @@ typedef union BufferDescPadded
#define BufferDescriptorGetContentLock(bdesc) \
((LWLock*) (&(bdesc)->content_lock))
+#define BufferGetExternalLSN(bufHdr) \
+ BufferExternalLSNs[(bufHdr)->buf_id]
+#define BufferSetExternalLSN(bufHdr, lsn) \
+ BufferExternalLSNs[(bufHdr)->buf_id] = (lsn)
+
extern PGDLLIMPORT ConditionVariableMinimallyPadded *BufferIOCVArray;
/*
@@ -276,6 +281,7 @@ typedef struct WritebackContext
/* in buf_init.c */
extern PGDLLIMPORT BufferDescPadded *BufferDescriptors;
+extern PGDLLIMPORT XLogRecPtr *BufferExternalLSNs;
extern PGDLLIMPORT WritebackContext BackendWritebackContext;
/* in localbuf.c */
diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h
index 897f2893860..e5957c5965d 100644
--- a/src/include/storage/bufmgr.h
+++ b/src/include/storage/bufmgr.h
@@ -115,13 +115,19 @@ extern Buffer ReadBufferWithoutRelcache(RelFileLocator rlocator,
ForkNumber forkNum, BlockNumber blockNum,
ReadBufferMode mode, BufferAccessStrategy strategy,
bool permanent);
+extern Buffer ReadBufferWithoutRelcacheWithHit(RelFileLocator rlocator,
+ ForkNumber forkNum, BlockNumber blockNum,
+ ReadBufferMode mode, BufferAccessStrategy strategy,
+ bool permanent, bool *hit);
extern void ReleaseBuffer(Buffer buffer);
extern void UnlockReleaseBuffer(Buffer buffer);
extern void MarkBufferDirty(Buffer buffer);
extern void IncrBufferRefCount(Buffer buffer);
extern Buffer ReleaseAndReadBuffer(Buffer buffer, Relation relation,
BlockNumber blockNum);
-
+extern bool BufferProbe(RelFileLocator rlocator, ForkNumber forkNum,
+ BlockNumber blockNum);
+
extern void InitBufferPool(void);
extern void InitBufferPoolAccess(void);
extern void AtEOXact_Buffers(bool isCommit);
diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h
index 130e59241e3..0ed569b2836 100644
--- a/src/include/storage/smgr.h
+++ b/src/include/storage/smgr.h
@@ -123,6 +123,7 @@ extern void smgrwriteback(SMgrFileHandle sfile,
extern BlockNumber smgrnblocks(SMgrFileHandle sfile);
extern BlockNumber smgrnblocks_cached(SMgrFileHandle sfile);
extern void smgrimmedsync(SMgrFileHandle sfile);
+extern void smgrunlink(SMgrFileHandle sfile, bool isRedo);
extern void smgrtruncate_multi(RelFileLocator rlocator, BackendId backend, ForkNumber *forks, int nforks, BlockNumber *nblocks);
extern void smgrunlink_multi(RelFileLocator rlocator, BackendId backend, ForkNumber *forks, int nforks, bool isRedo);
--
2.30.2
^ permalink raw reply [nested|flat] 22+ messages in thread
* Re: SLRUs in the main buffer pool, redux
@ 2022-09-17 00:41 Bagga, Rishu <[email protected]>
parent: Thomas Munro <[email protected]>
2 siblings, 1 reply; 22+ messages in thread
From: Bagga, Rishu @ 2022-09-17 00:41 UTC (permalink / raw)
To: Thomas Munro <[email protected]>; Heikki Linnakangas <[email protected]>; +Cc: pgsql-hackers
Hi Thomas,
While I was working on adding the page headers to SLRU pages on your patch, I came across this code where it seems like "MultiXactIdToMemberPage" is mistakenly being used instead of MultiXactIdToOffsetPage in the TrimMultiXact function.
Below is the area of concern in the patch:
@@ -2045,14 +1977,7 @@ TrimMultiXact(void)
oldestMXactDB = MultiXactState->oldestMultiXactDB;
LWLockRelease(MultiXactGenLock);
- /* Clean up offsets state */
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
-
- /*
- * (Re-)Initialize our idea of the latest page number for offsets.
- */
- pageno = MultiXactIdToOffsetPage(nextMXact);
- MultiXactOffsetCtl->shared->latest_page_number = pag0eno;
+ pageno = MXOffsetToMemberPage(offset);
Let us know if I am missing something here or if it is an error.
Sincerely,
Rishu Bagga (Amazon Web Services)
On 9/16/22, 5:37 PM, "Thomas Munro" <[email protected]> wrote:
CAUTION: This email originated from outside of the organization. Do not click links or open attachments unless you can confirm the sender and know the content is safe.
Rebased, debugged and fleshed out a tiny bit more, but still with
plenty of TODO notes and questions. I will talk about this idea at
PGCon, so I figured it'd help to have a patch that actually applies,
even if it doesn't work quite right yet. It's quite a large patch but
that's partly because it removes a lot of lines...
^ permalink raw reply [nested|flat] 22+ messages in thread
* Re: SLRUs in the main buffer pool, redux
@ 2022-09-21 09:32 Thomas Munro <[email protected]>
parent: Bagga, Rishu <[email protected]>
0 siblings, 0 replies; 22+ messages in thread
From: Thomas Munro @ 2022-09-21 09:32 UTC (permalink / raw)
To: Bagga, Rishu <[email protected]>; +Cc: Heikki Linnakangas <[email protected]>; pgsql-hackers
On Sat, Sep 17, 2022 at 12:41 PM Bagga, Rishu <[email protected]> wrote:
> While I was working on adding the page headers to SLRU pages on your patch, I came across this code where it seems like "MultiXactIdToMemberPage" is mistakenly being used instead of MultiXactIdToOffsetPage in the TrimMultiXact function.
Thanks Rishu. Right. Will fix soon in the next version, along with
my long overdue replies to Heikki and Konstantin.
^ permalink raw reply [nested|flat] 22+ messages in thread
* Re: SLRUs in the main buffer pool, redux
@ 2023-01-20 17:00 Shawn Debnath <[email protected]>
parent: Heikki Linnakangas <[email protected]>
0 siblings, 1 reply; 22+ messages in thread
From: Shawn Debnath @ 2023-01-20 17:00 UTC (permalink / raw)
To: Heikki Linnakangas <[email protected]>; +Cc: Thomas Munro <[email protected]>; pgsql-hackers; Yura Sokolov <[email protected]>
On Mon, Jul 25, 2022 at 11:54:36AM +0300, Heikki Linnakangas wrote:
> Oh I just saw that you had a comment about that in the patch and had hacked
> around it. Anyway, calling ResourceOwnerEnlargeBuffers() might be a
> solution. Or switch to a separate "CriticalResourceOwner" that's guaranteed
> to have enough pre-allocated space, before entering the critical section.
Wanted to bump up this thread. Rishu in my team posted a patch in the other
SLRU thread [1] with the latest updates and fixes and looks like performance
numbers do not show any regression. This change is currently in the
January commitfest [2] as well. Any feedback would be appreciated!
[1]
https://www.postgresql.org/message-id/A09EAE0D-0D3F-4A34-ADE9-8AC1DCBE7D57%40amazon.com
[2] https://commitfest.postgresql.org/41/3514/
Shawn
Amazon Web Services (AWS)
^ permalink raw reply [nested|flat] 22+ messages in thread
* Re: SLRUs in the main buffer pool, redux
@ 2023-02-27 13:31 Heikki Linnakangas <[email protected]>
parent: Shawn Debnath <[email protected]>
0 siblings, 1 reply; 22+ messages in thread
From: Heikki Linnakangas @ 2023-02-27 13:31 UTC (permalink / raw)
To: Shawn Debnath <[email protected]>; +Cc: Thomas Munro <[email protected]>; pgsql-hackers; Yura Sokolov <[email protected]>
On 20/01/2023 19:00, Shawn Debnath wrote:
> On Mon, Jul 25, 2022 at 11:54:36AM +0300, Heikki Linnakangas wrote:
>
>> Oh I just saw that you had a comment about that in the patch and had hacked
>> around it. Anyway, calling ResourceOwnerEnlargeBuffers() might be a
>> solution. Or switch to a separate "CriticalResourceOwner" that's guaranteed
>> to have enough pre-allocated space, before entering the critical section.
>
> Wanted to bump up this thread. Rishu in my team posted a patch in the other
> SLRU thread [1] with the latest updates and fixes and looks like performance
> numbers do not show any regression. This change is currently in the
> January commitfest [2] as well. Any feedback would be appreciated!
Here's a rebased set of patches.
The second patch is failing the pg_upgrade tests. Before I dig into
that, I'd love to get some feedback on this general approach.
- Heikki
Attachments:
[text/x-patch] v2-0001-Have-separate-SMmgrRelation-per-fork-rename-it-to.patch (165.4K, ../../[email protected]/2-v2-0001-Have-separate-SMmgrRelation-per-fork-rename-it-to.patch)
download | inline diff:
From 40e81bea853243cdedb46f2ef8b5c4780562cbf1 Mon Sep 17 00:00:00 2001
From: Heikki Linnakangas <[email protected]>
Date: Mon, 27 Feb 2023 14:53:19 +0200
Subject: [PATCH v2 1/2] Have separate SMmgrRelation per fork, rename it to
SMgrFile
---
contrib/amcheck/verify_nbtree.c | 2 +-
contrib/bloom/blinsert.c | 6 +-
contrib/pg_prewarm/autoprewarm.c | 2 +-
contrib/pg_prewarm/pg_prewarm.c | 4 +-
contrib/pg_visibility/pg_visibility.c | 5 +-
src/backend/access/gist/gistbuild.c | 11 +-
src/backend/access/hash/hashpage.c | 2 +-
src/backend/access/heap/heapam_handler.c | 32 +-
src/backend/access/heap/rewriteheap.c | 6 +-
src/backend/access/heap/visibilitymap.c | 41 ++-
src/backend/access/nbtree/nbtree.c | 6 +-
src/backend/access/nbtree/nbtsort.c | 8 +-
src/backend/access/spgist/spginsert.c | 17 +-
src/backend/access/table/tableam.c | 4 +-
src/backend/access/transam/xlogprefetcher.c | 28 +-
src/backend/access/transam/xlogutils.c | 17 +-
src/backend/catalog/catalog.c | 53 +--
src/backend/catalog/index.c | 4 +-
src/backend/catalog/storage.c | 205 +++++++----
src/backend/commands/dbcommands.c | 10 +-
src/backend/commands/sequence.c | 8 +-
src/backend/commands/tablecmds.c | 19 +-
src/backend/storage/buffer/bufmgr.c | 274 +++++++-------
src/backend/storage/buffer/localbuf.c | 24 +-
src/backend/storage/freespace/freespace.c | 41 ++-
src/backend/storage/smgr/md.c | 361 ++++++++-----------
src/backend/storage/smgr/smgr.c | 379 +++++++++-----------
src/backend/utils/cache/inval.c | 17 +-
src/backend/utils/cache/relcache.c | 34 +-
src/common/relpath.c | 4 +-
src/include/catalog/storage.h | 11 +-
src/include/common/relpath.h | 12 +-
src/include/storage/buf_internals.h | 7 +-
src/include/storage/bufmgr.h | 25 +-
src/include/storage/md.h | 28 +-
src/include/storage/relfilelocator.h | 29 +-
src/include/storage/smgr.h | 98 +++--
src/include/utils/inval.h | 2 +-
src/include/utils/rel.h | 32 +-
39 files changed, 907 insertions(+), 961 deletions(-)
diff --git a/contrib/amcheck/verify_nbtree.c b/contrib/amcheck/verify_nbtree.c
index 257cff671bd..2ab837c3062 100644
--- a/contrib/amcheck/verify_nbtree.c
+++ b/contrib/amcheck/verify_nbtree.c
@@ -324,7 +324,7 @@ bt_index_check_internal(Oid indrelid, bool parentcheck, bool heapallindexed,
bool heapkeyspace,
allequalimage;
- if (!smgrexists(RelationGetSmgr(indrel), MAIN_FORKNUM))
+ if (!smgrexists(RelationGetSmgr(indrel, MAIN_FORKNUM)))
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("index \"%s\" lacks a main relation fork",
diff --git a/contrib/bloom/blinsert.c b/contrib/bloom/blinsert.c
index dcd81208953..df4a821369e 100644
--- a/contrib/bloom/blinsert.c
+++ b/contrib/bloom/blinsert.c
@@ -177,9 +177,9 @@ blbuildempty(Relation index)
* this even when wal_level=minimal.
*/
PageSetChecksumInplace(metapage, BLOOM_METAPAGE_BLKNO);
- smgrwrite(RelationGetSmgr(index), INIT_FORKNUM, BLOOM_METAPAGE_BLKNO,
+ smgrwrite(RelationGetSmgr(index, INIT_FORKNUM), BLOOM_METAPAGE_BLKNO,
metapage, true);
- log_newpage(&(RelationGetSmgr(index))->smgr_rlocator.locator, INIT_FORKNUM,
+ log_newpage(&index->rd_locator, INIT_FORKNUM,
BLOOM_METAPAGE_BLKNO, metapage, true);
/*
@@ -187,7 +187,7 @@ blbuildempty(Relation index)
* write did not go through shared_buffers and therefore a concurrent
* checkpoint may have moved the redo pointer past our xlog record.
*/
- smgrimmedsync(RelationGetSmgr(index), INIT_FORKNUM);
+ smgrimmedsync(RelationGetSmgr(index, INIT_FORKNUM));
}
/*
diff --git a/contrib/pg_prewarm/autoprewarm.c b/contrib/pg_prewarm/autoprewarm.c
index 93835449c0e..74cc8df6481 100644
--- a/contrib/pg_prewarm/autoprewarm.c
+++ b/contrib/pg_prewarm/autoprewarm.c
@@ -535,7 +535,7 @@ autoprewarm_database_main(Datum main_arg)
*/
if (blk->forknum > InvalidForkNumber &&
blk->forknum <= MAX_FORKNUM &&
- smgrexists(RelationGetSmgr(rel), blk->forknum))
+ smgrexists(RelationGetSmgr(rel, blk->forknum)))
nblocks = RelationGetNumberOfBlocksInFork(rel, blk->forknum);
else
nblocks = 0;
diff --git a/contrib/pg_prewarm/pg_prewarm.c b/contrib/pg_prewarm/pg_prewarm.c
index 54209924aed..479bbafb745 100644
--- a/contrib/pg_prewarm/pg_prewarm.c
+++ b/contrib/pg_prewarm/pg_prewarm.c
@@ -109,7 +109,7 @@ pg_prewarm(PG_FUNCTION_ARGS)
aclcheck_error(aclresult, get_relkind_objtype(rel->rd_rel->relkind), get_rel_name(relOid));
/* Check that the fork exists. */
- if (!smgrexists(RelationGetSmgr(rel), forkNumber))
+ if (!smgrexists(RelationGetSmgr(rel, forkNumber)))
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("fork \"%s\" does not exist for this relation",
@@ -177,7 +177,7 @@ pg_prewarm(PG_FUNCTION_ARGS)
for (block = first_block; block <= last_block; ++block)
{
CHECK_FOR_INTERRUPTS();
- smgrread(RelationGetSmgr(rel), forkNumber, block, blockbuffer.data);
+ smgrread(RelationGetSmgr(rel, forkNumber), block, blockbuffer.data);
++blocks_done;
}
}
diff --git a/contrib/pg_visibility/pg_visibility.c b/contrib/pg_visibility/pg_visibility.c
index 2a4acfd1eee..bb7cd8ec43d 100644
--- a/contrib/pg_visibility/pg_visibility.c
+++ b/contrib/pg_visibility/pg_visibility.c
@@ -386,13 +386,14 @@ pg_truncate_visibility_map(PG_FUNCTION_ARGS)
check_relation_relkind(rel);
/* Forcibly reset cached file size */
- RelationGetSmgr(rel)->smgr_cached_nblocks[VISIBILITYMAP_FORKNUM] = InvalidBlockNumber;
+ RelationGetSmgr(rel, VISIBILITYMAP_FORKNUM)->smgr_cached_nblocks = InvalidBlockNumber;
block = visibilitymap_prepare_truncate(rel, 0);
if (BlockNumberIsValid(block))
{
fork = VISIBILITYMAP_FORKNUM;
- smgrtruncate(RelationGetSmgr(rel), &fork, 1, &block);
+ DropRelationBuffers(rel->rd_locator, rel->rd_backend, &fork, 1, &block);
+ smgrtruncate_multi(rel->rd_locator, rel->rd_backend, &fork, 1, &block);
}
if (RelationNeedsWAL(rel))
diff --git a/src/backend/access/gist/gistbuild.c b/src/backend/access/gist/gistbuild.c
index 7a6d93bb879..7ae5618377c 100644
--- a/src/backend/access/gist/gistbuild.c
+++ b/src/backend/access/gist/gistbuild.c
@@ -416,7 +416,7 @@ gist_indexsortbuild(GISTBuildState *state)
* replaced with the real root page at the end.
*/
page = palloc0(BLCKSZ);
- smgrextend(RelationGetSmgr(state->indexrel), MAIN_FORKNUM, GIST_ROOT_BLKNO,
+ smgrextend(RelationGetSmgr(state->indexrel, MAIN_FORKNUM), GIST_ROOT_BLKNO,
page, true);
state->pages_allocated++;
state->pages_written++;
@@ -460,7 +460,7 @@ gist_indexsortbuild(GISTBuildState *state)
/* Write out the root */
PageSetLSN(levelstate->pages[0], GistBuildLSN);
PageSetChecksumInplace(levelstate->pages[0], GIST_ROOT_BLKNO);
- smgrwrite(RelationGetSmgr(state->indexrel), MAIN_FORKNUM, GIST_ROOT_BLKNO,
+ smgrwrite(RelationGetSmgr(state->indexrel, MAIN_FORKNUM), GIST_ROOT_BLKNO,
levelstate->pages[0], true);
if (RelationNeedsWAL(state->indexrel))
log_newpage(&state->indexrel->rd_locator, MAIN_FORKNUM, GIST_ROOT_BLKNO,
@@ -479,7 +479,7 @@ gist_indexsortbuild(GISTBuildState *state)
* still not be on disk when the crash occurs.
*/
if (RelationNeedsWAL(state->indexrel))
- smgrimmedsync(RelationGetSmgr(state->indexrel), MAIN_FORKNUM);
+ smgrimmedsync(RelationGetSmgr(state->indexrel, MAIN_FORKNUM));
}
/*
@@ -657,7 +657,7 @@ gist_indexsortbuild_flush_ready_pages(GISTBuildState *state)
PageSetLSN(page, GistBuildLSN);
PageSetChecksumInplace(page, blkno);
- smgrextend(RelationGetSmgr(state->indexrel), MAIN_FORKNUM, blkno, page,
+ smgrextend(RelationGetSmgr(state->indexrel, MAIN_FORKNUM), blkno, page,
true);
state->pages_written++;
@@ -943,8 +943,7 @@ gistBuildCallback(Relation index,
*/
if ((buildstate->buildMode == GIST_BUFFERING_AUTO &&
buildstate->indtuples % BUFFERING_MODE_SWITCH_CHECK_STEP == 0 &&
- effective_cache_size < smgrnblocks(RelationGetSmgr(index),
- MAIN_FORKNUM)) ||
+ effective_cache_size < smgrnblocks(RelationGetSmgr(index, MAIN_FORKNUM))) ||
(buildstate->buildMode == GIST_BUFFERING_STATS &&
buildstate->indtuples >= BUFFERING_MODE_TUPLE_SIZE_STATS_TARGET))
{
diff --git a/src/backend/access/hash/hashpage.c b/src/backend/access/hash/hashpage.c
index 2d8fdec98ed..166bf81e028 100644
--- a/src/backend/access/hash/hashpage.c
+++ b/src/backend/access/hash/hashpage.c
@@ -1030,7 +1030,7 @@ _hash_alloc_buckets(Relation rel, BlockNumber firstblock, uint32 nblocks)
true);
PageSetChecksumInplace(page, lastblock);
- smgrextend(RelationGetSmgr(rel), MAIN_FORKNUM, lastblock, zerobuf.data,
+ smgrextend(RelationGetSmgr(rel, MAIN_FORKNUM), lastblock, zerobuf.data,
false);
return true;
diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c
index c4b1916d36e..8af225502cc 100644
--- a/src/backend/access/heap/heapam_handler.c
+++ b/src/backend/access/heap/heapam_handler.c
@@ -572,7 +572,7 @@ heapam_relation_set_new_filelocator(Relation rel,
TransactionId *freezeXid,
MultiXactId *minmulti)
{
- SMgrRelation srel;
+ SMgrFileHandle sfile;
/*
* Initialize to the minimum XID that could put tuples in the table. We
@@ -591,7 +591,7 @@ heapam_relation_set_new_filelocator(Relation rel,
*/
*minmulti = GetOldestMultiXactId();
- srel = RelationCreateStorage(*newrlocator, persistence, true);
+ sfile = RelationCreateStorage(*newrlocator, persistence, true);
/*
* If required, set up an init fork for an unlogged table so that it can
@@ -604,15 +604,18 @@ heapam_relation_set_new_filelocator(Relation rel,
*/
if (persistence == RELPERSISTENCE_UNLOGGED)
{
+ SMgrFileHandle sfile_init;
+
Assert(rel->rd_rel->relkind == RELKIND_RELATION ||
rel->rd_rel->relkind == RELKIND_MATVIEW ||
rel->rd_rel->relkind == RELKIND_TOASTVALUE);
- smgrcreate(srel, INIT_FORKNUM, false);
+ sfile_init = smgropen(*newrlocator, InvalidBackendId, INIT_FORKNUM);
+ smgrcreate(sfile_init, false);
log_smgrcreate(newrlocator, INIT_FORKNUM);
- smgrimmedsync(srel, INIT_FORKNUM);
+ smgrimmedsync(sfile);
}
- smgrclose(srel);
+ smgrclose(sfile);
}
static void
@@ -624,9 +627,7 @@ heapam_relation_nontransactional_truncate(Relation rel)
static void
heapam_relation_copy_data(Relation rel, const RelFileLocator *newrlocator)
{
- SMgrRelation dstrel;
-
- dstrel = smgropen(*newrlocator, rel->rd_backend);
+ SMgrFileHandle dstmain;
/*
* Since we copy the file directly without looking at the shared buffers,
@@ -646,16 +647,21 @@ heapam_relation_copy_data(Relation rel, const RelFileLocator *newrlocator)
RelationCreateStorage(*newrlocator, rel->rd_rel->relpersistence, true);
/* copy main fork */
- RelationCopyStorage(RelationGetSmgr(rel), dstrel, MAIN_FORKNUM,
+ dstmain = smgropen(*newrlocator, rel->rd_backend, MAIN_FORKNUM);
+ RelationCopyStorage(RelationGetSmgr(rel, MAIN_FORKNUM), dstmain,
rel->rd_rel->relpersistence);
/* copy those extra forks that exist */
for (ForkNumber forkNum = MAIN_FORKNUM + 1;
forkNum <= MAX_FORKNUM; forkNum++)
{
- if (smgrexists(RelationGetSmgr(rel), forkNum))
+ SMgrFileHandle src_fork = RelationGetSmgr(rel, forkNum);
+
+ if (smgrexists(src_fork))
{
- smgrcreate(dstrel, forkNum, false);
+ SMgrFileHandle dst_fork = smgropen(*newrlocator, rel->rd_backend, forkNum);
+
+ smgrcreate(dst_fork, false);
/*
* WAL log creation if the relation is persistent, or this is the
@@ -665,7 +671,7 @@ heapam_relation_copy_data(Relation rel, const RelFileLocator *newrlocator)
(rel->rd_rel->relpersistence == RELPERSISTENCE_UNLOGGED &&
forkNum == INIT_FORKNUM))
log_smgrcreate(newrlocator, forkNum);
- RelationCopyStorage(RelationGetSmgr(rel), dstrel, forkNum,
+ RelationCopyStorage(RelationGetSmgr(rel, forkNum), dst_fork,
rel->rd_rel->relpersistence);
}
}
@@ -673,7 +679,7 @@ heapam_relation_copy_data(Relation rel, const RelFileLocator *newrlocator)
/* drop old relation, and close new one */
RelationDropStorage(rel);
- smgrclose(dstrel);
+ smgrclose(dstmain);
}
static void
diff --git a/src/backend/access/heap/rewriteheap.c b/src/backend/access/heap/rewriteheap.c
index ae0282a70ee..30b700b0e8e 100644
--- a/src/backend/access/heap/rewriteheap.c
+++ b/src/backend/access/heap/rewriteheap.c
@@ -325,7 +325,7 @@ end_heap_rewrite(RewriteState state)
PageSetChecksumInplace(state->rs_buffer, state->rs_blockno);
- smgrextend(RelationGetSmgr(state->rs_new_rel), MAIN_FORKNUM,
+ smgrextend(RelationGetSmgr(state->rs_new_rel, MAIN_FORKNUM),
state->rs_blockno, state->rs_buffer, true);
}
@@ -337,7 +337,7 @@ end_heap_rewrite(RewriteState state)
* wrote before the checkpoint.
*/
if (RelationNeedsWAL(state->rs_new_rel))
- smgrimmedsync(RelationGetSmgr(state->rs_new_rel), MAIN_FORKNUM);
+ smgrimmedsync(RelationGetSmgr(state->rs_new_rel, MAIN_FORKNUM));
logical_end_heap_rewrite(state);
@@ -691,7 +691,7 @@ raw_heap_insert(RewriteState state, HeapTuple tup)
*/
PageSetChecksumInplace(page, state->rs_blockno);
- smgrextend(RelationGetSmgr(state->rs_new_rel), MAIN_FORKNUM,
+ smgrextend(RelationGetSmgr(state->rs_new_rel, MAIN_FORKNUM),
state->rs_blockno, page, true);
state->rs_blockno++;
diff --git a/src/backend/access/heap/visibilitymap.c b/src/backend/access/heap/visibilitymap.c
index 74ff01bb172..96ca3b0cc71 100644
--- a/src/backend/access/heap/visibilitymap.c
+++ b/src/backend/access/heap/visibilitymap.c
@@ -468,7 +468,7 @@ visibilitymap_prepare_truncate(Relation rel, BlockNumber nheapblocks)
* If no visibility map has been created yet for this relation, there's
* nothing to truncate.
*/
- if (!smgrexists(RelationGetSmgr(rel), VISIBILITYMAP_FORKNUM))
+ if (!smgrexists(RelationGetSmgr(rel, VISIBILITYMAP_FORKNUM)))
return InvalidBlockNumber;
/*
@@ -535,7 +535,7 @@ visibilitymap_prepare_truncate(Relation rel, BlockNumber nheapblocks)
else
newnblocks = truncBlock;
- if (smgrnblocks(RelationGetSmgr(rel), VISIBILITYMAP_FORKNUM) <= newnblocks)
+ if (smgrnblocks(RelationGetSmgr(rel, VISIBILITYMAP_FORKNUM)) <= newnblocks)
{
/* nothing to do, the file was already smaller than requested size */
return InvalidBlockNumber;
@@ -554,29 +554,29 @@ static Buffer
vm_readbuf(Relation rel, BlockNumber blkno, bool extend)
{
Buffer buf;
- SMgrRelation reln;
+ SMgrFileHandle vm_sfile;
/*
* Caution: re-using this smgr pointer could fail if the relcache entry
* gets closed. It's safe as long as we only do smgr-level operations
* between here and the last use of the pointer.
*/
- reln = RelationGetSmgr(rel);
+ vm_sfile = RelationGetSmgr(rel, VISIBILITYMAP_FORKNUM);
/*
* If we haven't cached the size of the visibility map fork yet, check it
* first.
*/
- if (reln->smgr_cached_nblocks[VISIBILITYMAP_FORKNUM] == InvalidBlockNumber)
+ if (vm_sfile->smgr_cached_nblocks == InvalidBlockNumber)
{
- if (smgrexists(reln, VISIBILITYMAP_FORKNUM))
- smgrnblocks(reln, VISIBILITYMAP_FORKNUM);
+ if (smgrexists(vm_sfile))
+ smgrnblocks(vm_sfile);
else
- reln->smgr_cached_nblocks[VISIBILITYMAP_FORKNUM] = 0;
+ vm_sfile->smgr_cached_nblocks = 0;
}
/* Handle requests beyond EOF */
- if (blkno >= reln->smgr_cached_nblocks[VISIBILITYMAP_FORKNUM])
+ if (blkno >= vm_sfile->smgr_cached_nblocks)
{
if (extend)
vm_extend(rel, blkno + 1);
@@ -603,8 +603,7 @@ vm_readbuf(Relation rel, BlockNumber blkno, bool extend)
* long as it doesn't depend on the page header having correct contents.
* Current usage is safe because PageGetContents() does not require that.
*/
- buf = ReadBufferExtended(rel, VISIBILITYMAP_FORKNUM, blkno,
- RBM_ZERO_ON_ERROR, NULL);
+ buf = ReadBufferExtended(rel, VISIBILITYMAP_FORKNUM, blkno, RBM_ZERO_ON_ERROR, NULL);
if (PageIsNew(BufferGetPage(buf)))
{
LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
@@ -624,7 +623,7 @@ vm_extend(Relation rel, BlockNumber vm_nblocks)
{
BlockNumber vm_nblocks_now;
PGAlignedBlock pg;
- SMgrRelation reln;
+ SMgrFileHandle vm_sfile;
PageInit((Page) pg.data, BLCKSZ, 0);
@@ -645,27 +644,27 @@ vm_extend(Relation rel, BlockNumber vm_nblocks)
* gets closed. It's safe as long as we only do smgr-level operations
* between here and the last use of the pointer.
*/
- reln = RelationGetSmgr(rel);
+ vm_sfile = RelationGetSmgr(rel, VISIBILITYMAP_FORKNUM);
/*
* Create the file first if it doesn't exist. If smgr_vm_nblocks is
* positive then it must exist, no need for an smgrexists call.
*/
- if ((reln->smgr_cached_nblocks[VISIBILITYMAP_FORKNUM] == 0 ||
- reln->smgr_cached_nblocks[VISIBILITYMAP_FORKNUM] == InvalidBlockNumber) &&
- !smgrexists(reln, VISIBILITYMAP_FORKNUM))
- smgrcreate(reln, VISIBILITYMAP_FORKNUM, false);
+ if ((vm_sfile->smgr_cached_nblocks == 0 ||
+ vm_sfile->smgr_cached_nblocks == InvalidBlockNumber) &&
+ !smgrexists(vm_sfile))
+ smgrcreate(vm_sfile, false);
/* Invalidate cache so that smgrnblocks() asks the kernel. */
- reln->smgr_cached_nblocks[VISIBILITYMAP_FORKNUM] = InvalidBlockNumber;
- vm_nblocks_now = smgrnblocks(reln, VISIBILITYMAP_FORKNUM);
+ vm_sfile->smgr_cached_nblocks = InvalidBlockNumber;
+ vm_nblocks_now = smgrnblocks(vm_sfile);
/* Now extend the file */
while (vm_nblocks_now < vm_nblocks)
{
PageSetChecksumInplace((Page) pg.data, vm_nblocks_now);
- smgrextend(reln, VISIBILITYMAP_FORKNUM, vm_nblocks_now, pg.data, false);
+ smgrextend(vm_sfile, vm_nblocks_now, pg.data, false);
vm_nblocks_now++;
}
@@ -676,7 +675,7 @@ vm_extend(Relation rel, BlockNumber vm_nblocks)
* to keep checking for creation or extension of the file, which happens
* infrequently.
*/
- CacheInvalidateSmgr(reln->smgr_rlocator);
+ CacheInvalidateSmgr(rel->rd_locator, rel->rd_backend);
UnlockRelationForExtension(rel, ExclusiveLock);
}
diff --git a/src/backend/access/nbtree/nbtree.c b/src/backend/access/nbtree/nbtree.c
index 3f7b541e9d2..d4db2c6d0fa 100644
--- a/src/backend/access/nbtree/nbtree.c
+++ b/src/backend/access/nbtree/nbtree.c
@@ -164,9 +164,9 @@ btbuildempty(Relation index)
* this even when wal_level=minimal.
*/
PageSetChecksumInplace(metapage, BTREE_METAPAGE);
- smgrwrite(RelationGetSmgr(index), INIT_FORKNUM, BTREE_METAPAGE,
+ smgrwrite(RelationGetSmgr(index, INIT_FORKNUM), BTREE_METAPAGE,
metapage, true);
- log_newpage(&RelationGetSmgr(index)->smgr_rlocator.locator, INIT_FORKNUM,
+ log_newpage(&index->rd_locator, INIT_FORKNUM,
BTREE_METAPAGE, metapage, true);
/*
@@ -174,7 +174,7 @@ btbuildempty(Relation index)
* write did not go through shared_buffers and therefore a concurrent
* checkpoint may have moved the redo pointer past our xlog record.
*/
- smgrimmedsync(RelationGetSmgr(index), INIT_FORKNUM);
+ smgrimmedsync(RelationGetSmgr(index, INIT_FORKNUM));
}
/*
diff --git a/src/backend/access/nbtree/nbtsort.c b/src/backend/access/nbtree/nbtsort.c
index 02b9601bece..4ae27cc57c6 100644
--- a/src/backend/access/nbtree/nbtsort.c
+++ b/src/backend/access/nbtree/nbtsort.c
@@ -662,7 +662,7 @@ _bt_blwritepage(BTWriteState *wstate, Page page, BlockNumber blkno)
if (!wstate->btws_zeropage)
wstate->btws_zeropage = (Page) palloc0(BLCKSZ);
/* don't set checksum for all-zero page */
- smgrextend(RelationGetSmgr(wstate->index), MAIN_FORKNUM,
+ smgrextend(RelationGetSmgr(wstate->index, MAIN_FORKNUM),
wstate->btws_pages_written++,
wstate->btws_zeropage,
true);
@@ -677,14 +677,14 @@ _bt_blwritepage(BTWriteState *wstate, Page page, BlockNumber blkno)
if (blkno == wstate->btws_pages_written)
{
/* extending the file... */
- smgrextend(RelationGetSmgr(wstate->index), MAIN_FORKNUM, blkno,
+ smgrextend(RelationGetSmgr(wstate->index, MAIN_FORKNUM), blkno,
page, true);
wstate->btws_pages_written++;
}
else
{
/* overwriting a block we zero-filled before */
- smgrwrite(RelationGetSmgr(wstate->index), MAIN_FORKNUM, blkno,
+ smgrwrite(RelationGetSmgr(wstate->index, MAIN_FORKNUM), blkno,
page, true);
}
@@ -1431,7 +1431,7 @@ _bt_load(BTWriteState *wstate, BTSpool *btspool, BTSpool *btspool2)
* still not be on disk when the crash occurs.
*/
if (wstate->btws_use_wal)
- smgrimmedsync(RelationGetSmgr(wstate->index), MAIN_FORKNUM);
+ smgrimmedsync(RelationGetSmgr(wstate->index, MAIN_FORKNUM));
}
/*
diff --git a/src/backend/access/spgist/spginsert.c b/src/backend/access/spgist/spginsert.c
index 718a88335d0..3de90846a10 100644
--- a/src/backend/access/spgist/spginsert.c
+++ b/src/backend/access/spgist/spginsert.c
@@ -155,8 +155,11 @@ spgbuild(Relation heap, Relation index, IndexInfo *indexInfo)
void
spgbuildempty(Relation index)
{
+ SMgrFileHandle sfile;
Page page;
+ sfile = RelationGetSmgr(index, INIT_FORKNUM);
+
/* Construct metapage. */
page = (Page) palloc(BLCKSZ);
SpGistInitMetapage(page);
@@ -169,27 +172,27 @@ spgbuildempty(Relation index)
* replayed.
*/
PageSetChecksumInplace(page, SPGIST_METAPAGE_BLKNO);
- smgrwrite(RelationGetSmgr(index), INIT_FORKNUM, SPGIST_METAPAGE_BLKNO,
+ smgrwrite(sfile, SPGIST_METAPAGE_BLKNO,
page, true);
- log_newpage(&(RelationGetSmgr(index))->smgr_rlocator.locator, INIT_FORKNUM,
+ log_newpage(&index->rd_locator, INIT_FORKNUM,
SPGIST_METAPAGE_BLKNO, page, true);
/* Likewise for the root page. */
SpGistInitPage(page, SPGIST_LEAF);
PageSetChecksumInplace(page, SPGIST_ROOT_BLKNO);
- smgrwrite(RelationGetSmgr(index), INIT_FORKNUM, SPGIST_ROOT_BLKNO,
+ smgrwrite(sfile, SPGIST_ROOT_BLKNO,
page, true);
- log_newpage(&(RelationGetSmgr(index))->smgr_rlocator.locator, INIT_FORKNUM,
+ log_newpage(&index->rd_locator, INIT_FORKNUM,
SPGIST_ROOT_BLKNO, page, true);
/* Likewise for the null-tuples root page. */
SpGistInitPage(page, SPGIST_LEAF | SPGIST_NULLS);
PageSetChecksumInplace(page, SPGIST_NULL_BLKNO);
- smgrwrite(RelationGetSmgr(index), INIT_FORKNUM, SPGIST_NULL_BLKNO,
+ smgrwrite(sfile, SPGIST_NULL_BLKNO,
page, true);
- log_newpage(&(RelationGetSmgr(index))->smgr_rlocator.locator, INIT_FORKNUM,
+ log_newpage(&index->rd_locator, INIT_FORKNUM,
SPGIST_NULL_BLKNO, page, true);
/*
@@ -197,7 +200,7 @@ spgbuildempty(Relation index)
* writes did not go through shared buffers and therefore a concurrent
* checkpoint may have moved the redo pointer past our xlog record.
*/
- smgrimmedsync(RelationGetSmgr(index), INIT_FORKNUM);
+ smgrimmedsync(sfile);
}
/*
diff --git a/src/backend/access/table/tableam.c b/src/backend/access/table/tableam.c
index ef0d34fceee..9ebf174b189 100644
--- a/src/backend/access/table/tableam.c
+++ b/src/backend/access/table/tableam.c
@@ -631,10 +631,10 @@ table_block_relation_size(Relation rel, ForkNumber forkNumber)
if (forkNumber == InvalidForkNumber)
{
for (int i = 0; i < MAX_FORKNUM; i++)
- nblocks += smgrnblocks(RelationGetSmgr(rel), i);
+ nblocks += smgrnblocks(RelationGetSmgr(rel, i));
}
else
- nblocks = smgrnblocks(RelationGetSmgr(rel), forkNumber);
+ nblocks = smgrnblocks(RelationGetSmgr(rel, forkNumber));
return nblocks * BLCKSZ;
}
diff --git a/src/backend/access/transam/xlogprefetcher.c b/src/backend/access/transam/xlogprefetcher.c
index 046e40d143a..50674cd8b53 100644
--- a/src/backend/access/transam/xlogprefetcher.c
+++ b/src/backend/access/transam/xlogprefetcher.c
@@ -652,7 +652,7 @@ XLogPrefetcherNextBlock(uintptr_t pgsr_private, XLogRecPtr *lsn)
{
int block_id = prefetcher->next_block_id++;
DecodedBkpBlock *block = &record->blocks[block_id];
- SMgrRelation reln;
+ SMgrFileHandle sfile;
PrefetchBufferResult result;
if (!block->in_use)
@@ -722,7 +722,7 @@ XLogPrefetcherNextBlock(uintptr_t pgsr_private, XLogRecPtr *lsn)
* same relation (with some scheme to handle invalidations
* safely), but for now we'll call smgropen() every time.
*/
- reln = smgropen(block->rlocator, InvalidBackendId);
+ sfile = smgropen(block->rlocator, InvalidBackendId, block->forknum);
/*
* If the relation file doesn't exist on disk, for example because
@@ -731,14 +731,14 @@ XLogPrefetcherNextBlock(uintptr_t pgsr_private, XLogRecPtr *lsn)
* further prefetching in the relation until this record is
* replayed.
*/
- if (!smgrexists(reln, MAIN_FORKNUM))
+ if (!smgrexists(sfile))
{
#ifdef XLOGPREFETCHER_DEBUG_LEVEL
elog(XLOGPREFETCHER_DEBUG_LEVEL,
"suppressing all prefetch in relation %u/%u/%u until %X/%X is replayed, because the relation does not exist on disk",
- reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
- reln->smgr_rlocator.locator.relNumber,
+ sfile->smgr_rlocator.locator.spcOid,
+ sfile->smgr_rlocator.locator.dbOid,
+ sfile->smgr_rlocator.locator.relNumber,
LSN_FORMAT_ARGS(record->lsn));
#endif
XLogPrefetcherAddFilter(prefetcher, block->rlocator, 0,
@@ -752,14 +752,14 @@ XLogPrefetcherNextBlock(uintptr_t pgsr_private, XLogRecPtr *lsn)
* block yet, suppress prefetching of this block and higher until
* this record is replayed.
*/
- if (block->blkno >= smgrnblocks(reln, block->forknum))
+ if (block->blkno >= smgrnblocks(sfile))
{
#ifdef XLOGPREFETCHER_DEBUG_LEVEL
elog(XLOGPREFETCHER_DEBUG_LEVEL,
"suppressing prefetch in relation %u/%u/%u from block %u until %X/%X is replayed, because the relation is too small",
- reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
- reln->smgr_rlocator.locator.relNumber,
+ sfile->smgr_rlocator.locator.spcOid,
+ sfile->smgr_rlocator.locator.dbOid,
+ sfile->smgr_rlocator.locator.relNumber,
block->blkno,
LSN_FORMAT_ARGS(record->lsn));
#endif
@@ -770,7 +770,7 @@ XLogPrefetcherNextBlock(uintptr_t pgsr_private, XLogRecPtr *lsn)
}
/* Try to initiate prefetching. */
- result = PrefetchSharedBuffer(reln, block->forknum, block->blkno);
+ result = PrefetchSharedBuffer(sfile, block->blkno);
if (BufferIsValid(result.recent_buffer))
{
/* Cache hit, nothing to do. */
@@ -796,9 +796,9 @@ XLogPrefetcherNextBlock(uintptr_t pgsr_private, XLogRecPtr *lsn)
*/
elog(ERROR,
"could not prefetch relation %u/%u/%u block %u",
- reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
- reln->smgr_rlocator.locator.relNumber,
+ sfile->smgr_locator.locator.spcOid,
+ sfile->smgr_locator.locator.dbOid,
+ sfile->smgr_locator.locator.relNumber,
block->blkno);
}
}
diff --git a/src/backend/access/transam/xlogutils.c b/src/backend/access/transam/xlogutils.c
index 2c28956b1aa..5e4034f7cd7 100644
--- a/src/backend/access/transam/xlogutils.c
+++ b/src/backend/access/transam/xlogutils.c
@@ -477,7 +477,7 @@ XLogReadBufferExtended(RelFileLocator rlocator, ForkNumber forknum,
{
BlockNumber lastblock;
Buffer buffer;
- SMgrRelation smgr;
+ SMgrFileHandle sfile;
Assert(blkno != P_NEW);
@@ -491,7 +491,7 @@ XLogReadBufferExtended(RelFileLocator rlocator, ForkNumber forknum,
}
/* Open the relation at smgr level */
- smgr = smgropen(rlocator, InvalidBackendId);
+ sfile = smgropen(rlocator, InvalidBackendId, forknum);
/*
* Create the target file if it doesn't already exist. This lets us cope
@@ -501,9 +501,9 @@ XLogReadBufferExtended(RelFileLocator rlocator, ForkNumber forknum,
* filesystem loses an inode during a crash. Better to write the data
* until we are actually told to delete the file.)
*/
- smgrcreate(smgr, forknum, true);
+ smgrcreate(sfile, true);
- lastblock = smgrnblocks(smgr, forknum);
+ lastblock = smgrnblocks(sfile);
if (blkno < lastblock)
{
@@ -631,7 +631,7 @@ CreateFakeRelcacheEntry(RelFileLocator rlocator)
rel->rd_lockInfo.lockRelId.dbId = rlocator.dbOid;
rel->rd_lockInfo.lockRelId.relId = rlocator.relNumber;
- rel->rd_smgr = NULL;
+ MemSet(rel->rd_smgr, 0, sizeof(rel->rd_smgr));
return rel;
}
@@ -643,8 +643,11 @@ void
FreeFakeRelcacheEntry(Relation fakerel)
{
/* make sure the fakerel is not referenced by the SmgrRelation anymore */
- if (fakerel->rd_smgr != NULL)
- smgrclearowner(&fakerel->rd_smgr, fakerel->rd_smgr);
+ for (int i = 0; i <= MAX_FORKNUM; i++)
+ {
+ if (fakerel->rd_smgr[i] != NULL)
+ smgrclearowner(&fakerel->rd_smgr[i], fakerel->rd_smgr[i]);
+ }
pfree(fakerel);
}
diff --git a/src/backend/catalog/catalog.c b/src/backend/catalog/catalog.c
index 1bf6c5633cd..6b01158615b 100644
--- a/src/backend/catalog/catalog.c
+++ b/src/backend/catalog/catalog.c
@@ -501,10 +501,10 @@ GetNewOidWithIndex(Relation relation, Oid indexId, AttrNumber oidcolumn)
RelFileNumber
GetNewRelFileNumber(Oid reltablespace, Relation pg_class, char relpersistence)
{
- RelFileLocatorBackend rlocator;
- char *rpath;
- bool collides;
+ RelFileLocator rlocator;
BackendId backend;
+ SMgrFileHandle sfile;
+ bool collides;
/*
* If we ever get here during pg_upgrade, there's something wrong; all
@@ -513,6 +513,11 @@ GetNewRelFileNumber(Oid reltablespace, Relation pg_class, char relpersistence)
*/
Assert(!IsBinaryUpgrade);
+ /*
+ * The relpath will vary based on the backend ID, so we must initialize
+ * that properly here to make sure that any collisions based on filename
+ * are properly detected.
+ */
switch (relpersistence)
{
case RELPERSISTENCE_TEMP:
@@ -528,53 +533,29 @@ GetNewRelFileNumber(Oid reltablespace, Relation pg_class, char relpersistence)
}
/* This logic should match RelationInitPhysicalAddr */
- rlocator.locator.spcOid = reltablespace ? reltablespace : MyDatabaseTableSpace;
- rlocator.locator.dbOid =
- (rlocator.locator.spcOid == GLOBALTABLESPACE_OID) ?
+ rlocator.spcOid = reltablespace ? reltablespace : MyDatabaseTableSpace;
+ rlocator.dbOid =
+ (rlocator.spcOid == GLOBALTABLESPACE_OID) ?
InvalidOid : MyDatabaseId;
- /*
- * The relpath will vary based on the backend ID, so we must initialize
- * that properly here to make sure that any collisions based on filename
- * are properly detected.
- */
- rlocator.backend = backend;
-
do
{
CHECK_FOR_INTERRUPTS();
/* Generate the OID */
if (pg_class)
- rlocator.locator.relNumber = GetNewOidWithIndex(pg_class, ClassOidIndexId,
+ rlocator.relNumber = GetNewOidWithIndex(pg_class, ClassOidIndexId,
Anum_pg_class_oid);
else
- rlocator.locator.relNumber = GetNewObjectId();
+ rlocator.relNumber = GetNewObjectId();
/* Check for existing file of same name */
- rpath = relpath(rlocator, MAIN_FORKNUM);
-
- if (access(rpath, F_OK) == 0)
- {
- /* definite collision */
- collides = true;
- }
- else
- {
- /*
- * Here we have a little bit of a dilemma: if errno is something
- * other than ENOENT, should we declare a collision and loop? In
- * practice it seems best to go ahead regardless of the errno. If
- * there is a colliding file we will get an smgr failure when we
- * attempt to create the new relation file.
- */
- collides = false;
- }
-
- pfree(rpath);
+ sfile = smgropen(rlocator, backend, MAIN_FORKNUM);
+ collides = smgrexists(sfile);
+ smgrclose(sfile);
} while (collides);
- return rlocator.locator.relNumber;
+ return rlocator.relNumber;
}
/*
diff --git a/src/backend/catalog/index.c b/src/backend/catalog/index.c
index 7777e7ec770..f5a0a16d3d5 100644
--- a/src/backend/catalog/index.c
+++ b/src/backend/catalog/index.c
@@ -3042,9 +3042,9 @@ index_build(Relation heapRelation,
* relfilenumber won't change, and nothing needs to be done here.
*/
if (indexRelation->rd_rel->relpersistence == RELPERSISTENCE_UNLOGGED &&
- !smgrexists(RelationGetSmgr(indexRelation), INIT_FORKNUM))
+ !smgrexists(RelationGetSmgr(indexRelation, INIT_FORKNUM)))
{
- smgrcreate(RelationGetSmgr(indexRelation), INIT_FORKNUM, false);
+ smgrcreate(RelationGetSmgr(indexRelation, INIT_FORKNUM), false);
indexRelation->rd_indam->ambuildempty(indexRelation);
}
diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c
index af1491aa1d1..3816610c87c 100644
--- a/src/backend/catalog/storage.c
+++ b/src/backend/catalog/storage.c
@@ -116,11 +116,11 @@ AddPendingSync(const RelFileLocator *rlocator)
* that does not want the storage to be destroyed in case of an abort may
* pass register_delete = false.
*/
-SMgrRelation
+SMgrFileHandle
RelationCreateStorage(RelFileLocator rlocator, char relpersistence,
bool register_delete)
{
- SMgrRelation srel;
+ SMgrFileHandle sfile;
BackendId backend;
bool needs_wal;
@@ -145,11 +145,11 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence,
return NULL; /* placate compiler */
}
- srel = smgropen(rlocator, backend);
- smgrcreate(srel, MAIN_FORKNUM, false);
+ sfile = smgropen(rlocator, backend, MAIN_FORKNUM);
+ smgrcreate(sfile, false);
if (needs_wal)
- log_smgrcreate(&srel->smgr_rlocator.locator, MAIN_FORKNUM);
+ log_smgrcreate(&rlocator, MAIN_FORKNUM);
/*
* Add the relation to the list of stuff to delete at abort, if we are
@@ -175,7 +175,7 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence,
AddPendingSync(&rlocator);
}
- return srel;
+ return sfile;
}
/*
@@ -292,16 +292,18 @@ RelationTruncate(Relation rel, BlockNumber nblocks)
ForkNumber forks[MAX_FORKNUM];
BlockNumber blocks[MAX_FORKNUM];
int nforks = 0;
- SMgrRelation reln;
/*
* Make sure smgr_targblock etc aren't pointing somewhere past new end.
* (Note: don't rely on this reln pointer below this loop.)
*/
- reln = RelationGetSmgr(rel);
- reln->smgr_targblock = InvalidBlockNumber;
- for (int i = 0; i <= MAX_FORKNUM; ++i)
- reln->smgr_cached_nblocks[i] = InvalidBlockNumber;
+ for (int i = 0; i <= MAX_FORKNUM; i++)
+ {
+ SMgrFileHandle sfile = RelationGetSmgr(rel, i);
+
+ sfile->smgr_targblock = InvalidBlockNumber;
+ sfile->smgr_cached_nblocks = InvalidBlockNumber;
+ }
/* Prepare for truncation of MAIN fork of the relation */
forks[nforks] = MAIN_FORKNUM;
@@ -309,7 +311,7 @@ RelationTruncate(Relation rel, BlockNumber nblocks)
nforks++;
/* Prepare for truncation of the FSM if it exists */
- fsm = smgrexists(RelationGetSmgr(rel), FSM_FORKNUM);
+ fsm = smgrexists(RelationGetSmgr(rel, FSM_FORKNUM));
if (fsm)
{
blocks[nforks] = FreeSpaceMapPrepareTruncateRel(rel, nblocks);
@@ -322,7 +324,7 @@ RelationTruncate(Relation rel, BlockNumber nblocks)
}
/* Prepare for truncation of the visibility map too if it exists */
- vm = smgrexists(RelationGetSmgr(rel), VISIBILITYMAP_FORKNUM);
+ vm = smgrexists(RelationGetSmgr(rel, VISIBILITYMAP_FORKNUM));
if (vm)
{
blocks[nforks] = visibilitymap_prepare_truncate(rel, nblocks);
@@ -390,11 +392,12 @@ RelationTruncate(Relation rel, BlockNumber nblocks)
}
/*
- * This will first remove any buffers from the buffer pool that should no
+ * First remove any buffers from the buffer pool that should no
* longer exist after truncation is complete, and then truncate the
* corresponding files on disk.
*/
- smgrtruncate(RelationGetSmgr(rel), forks, nforks, blocks);
+ DropRelationBuffers(rel->rd_locator, rel->rd_backend, forks, nforks, blocks);
+ smgrtruncate_multi(rel->rd_locator, rel->rd_backend, forks, nforks, blocks);
/* We've done all the critical work, so checkpoints are OK now. */
MyProc->delayChkptFlags &= ~DELAY_CHKPT_COMPLETE;
@@ -428,7 +431,7 @@ RelationPreTruncate(Relation rel)
return;
pending = hash_search(pendingSyncHash,
- &(RelationGetSmgr(rel)->smgr_rlocator.locator),
+ &rel->rd_locator,
HASH_FIND, NULL);
if (pending)
pending->is_truncated = true;
@@ -444,12 +447,12 @@ RelationPreTruncate(Relation rel)
* Also note that this is frequently called via locutions such as
* RelationCopyStorage(RelationGetSmgr(rel), ...);
* That's safe only because we perform only smgr and WAL operations here.
- * If we invoked anything else, a relcache flush could cause our SMgrRelation
+ * If we invoked anything else, a relcache flush could cause our SMgrFileHandle
* argument to become a dangling pointer.
*/
void
-RelationCopyStorage(SMgrRelation src, SMgrRelation dst,
- ForkNumber forkNum, char relpersistence)
+RelationCopyStorage(SMgrFileHandle src, SMgrFileHandle dst,
+ char relpersistence)
{
PGAlignedBlock buf;
Page page;
@@ -466,7 +469,7 @@ RelationCopyStorage(SMgrRelation src, SMgrRelation dst,
* it needs to be synced to disk.
*/
copying_initfork = relpersistence == RELPERSISTENCE_UNLOGGED &&
- forkNum == INIT_FORKNUM;
+ src->smgr_locator.forknum == INIT_FORKNUM;
/*
* We need to log the copied data in WAL iff WAL archiving/streaming is
@@ -477,14 +480,14 @@ RelationCopyStorage(SMgrRelation src, SMgrRelation dst,
use_wal = XLogIsNeeded() &&
(relpersistence == RELPERSISTENCE_PERMANENT || copying_initfork);
- nblocks = smgrnblocks(src, forkNum);
+ nblocks = smgrnblocks(src);
for (blkno = 0; blkno < nblocks; blkno++)
{
/* If we got a cancel signal during the copy of the data, quit */
CHECK_FOR_INTERRUPTS();
- smgrread(src, forkNum, blkno, buf.data);
+ smgrread(src, blkno, buf.data);
if (!PageIsVerifiedExtended(page, blkno,
PIV_LOG_WARNING | PIV_REPORT_STAT))
@@ -496,9 +499,7 @@ RelationCopyStorage(SMgrRelation src, SMgrRelation dst,
* (errcontext callbacks shouldn't be risking any such thing, but
* people have been known to forget that rule.)
*/
- char *relpath = relpathbackend(src->smgr_rlocator.locator,
- src->smgr_rlocator.backend,
- forkNum);
+ char *relpath = smgrfilepath(src->smgr_locator);
ereport(ERROR,
(errcode(ERRCODE_DATA_CORRUPTED),
@@ -512,7 +513,7 @@ RelationCopyStorage(SMgrRelation src, SMgrRelation dst,
* space.
*/
if (use_wal)
- log_newpage(&dst->smgr_rlocator.locator, forkNum, blkno, page, false);
+ log_newpage(&dst->smgr_locator.locator, dst->smgr_locator.forknum, blkno, page, false);
PageSetChecksumInplace(page, blkno);
@@ -521,7 +522,7 @@ RelationCopyStorage(SMgrRelation src, SMgrRelation dst,
* need for smgr to schedule an fsync for this write; we'll do it
* ourselves below.
*/
- smgrextend(dst, forkNum, blkno, buf.data, true);
+ smgrextend(dst, blkno, buf.data, true);
}
/*
@@ -534,7 +535,7 @@ RelationCopyStorage(SMgrRelation src, SMgrRelation dst,
* they might still not be on disk when the crash occurs.
*/
if (use_wal || copying_initfork)
- smgrimmedsync(dst, forkNum);
+ smgrimmedsync(dst);
}
/*
@@ -653,9 +654,9 @@ smgrDoPendingDeletes(bool isCommit)
PendingRelDelete *pending;
PendingRelDelete *prev;
PendingRelDelete *next;
- int nrels = 0,
- maxrels = 0;
- SMgrRelation *srels = NULL;
+ int nlocators = 0,
+ maxlocators = 0;
+ RelFileLocatorBackend *locators = NULL;
prev = NULL;
for (pending = pendingDeletes; pending != NULL; pending = next)
@@ -676,23 +677,21 @@ smgrDoPendingDeletes(bool isCommit)
/* do deletion if called for */
if (pending->atCommit == isCommit)
{
- SMgrRelation srel;
-
- srel = smgropen(pending->rlocator, pending->backend);
+ RelFileLocatorBackend rlocator = { pending->rlocator, pending->backend };
/* allocate the initial array, or extend it, if needed */
- if (maxrels == 0)
+ if (maxlocators == 0)
{
- maxrels = 8;
- srels = palloc(sizeof(SMgrRelation) * maxrels);
+ maxlocators = 8;
+ locators = palloc(sizeof(RelFileLocatorBackend) * maxlocators);
}
- else if (maxrels <= nrels)
+ else if (maxlocators <= nlocators)
{
- maxrels *= 2;
- srels = repalloc(srels, sizeof(SMgrRelation) * maxrels);
+ maxlocators *= 2;
+ locators = repalloc(locators, sizeof(RelFileLocatorBackend) * maxlocators);
}
- srels[nrels++] = srel;
+ locators[nlocators++] = rlocator;
}
/* must explicitly free the list entry */
pfree(pending);
@@ -700,15 +699,58 @@ smgrDoPendingDeletes(bool isCommit)
}
}
- if (nrels > 0)
+ if (nlocators > 0)
{
- smgrdounlinkall(srels, nrels, false);
+ ForkNumber forks[MAX_FORKNUM + 1];
+
+ for (int i = 0; i <= MAX_FORKNUM; i++)
+ forks[i] = i;
+
+ /*
+ * Get rid of any remaining buffers for the relations. bufmgr will just
+ * drop them without bothering to write the contents.
+ */
+ DropRelationsAllBuffers(locators, nlocators);
- for (int i = 0; i < nrels; i++)
- smgrclose(srels[i]);
+ for (int i = 0; i < nlocators; i++)
+ smgrunlink_multi(locators[i].locator, locators[i].backend, forks, MAX_FORKNUM + 1, false);
+ pfree(locators);
+ }
+}
+
+/*
+ * DropRelationFiles -- drop files of all given relations
+ */
+void
+DropRelationFiles(RelFileLocator *delrels, int ndelrels, bool isRedo)
+{
+ RelFileLocatorBackend *locators;
+ int i;
+ ForkNumber all_forks[MAX_FORKNUM + 1];
- pfree(srels);
+ locators = palloc(sizeof(RelFileLocatorBackend) * ndelrels);
+ for (i = 0; i < ndelrels; i++)
+ {
+ if (isRedo)
+ {
+ for (int fork = 0; fork <= MAX_FORKNUM; fork++)
+ XLogDropRelation(delrels[i], fork);
+ }
+ locators[i].locator = delrels[i];
+ locators[i].backend = InvalidBackendId;
}
+
+ /*
+ * Get rid of any remaining buffers for the relations. bufmgr will just
+ * drop them without bothering to write the contents.
+ */
+ DropRelationsAllBuffers(locators, ndelrels);
+
+ for (int fork = 0; fork <= MAX_FORKNUM; fork++)
+ all_forks[fork] = fork;
+
+ for (i = 0; i < ndelrels; i++)
+ smgrunlink_multi(locators[i].locator, locators[i].backend, all_forks, MAX_FORKNUM + 1, true);
}
/*
@@ -718,9 +760,9 @@ void
smgrDoPendingSyncs(bool isCommit, bool isParallelWorker)
{
PendingRelDelete *pending;
- int nrels = 0,
- maxrels = 0;
- SMgrRelation *srels = NULL;
+ int nlocators = 0,
+ maxlocators = 0;
+ RelFileLocator *locators = NULL;
HASH_SEQ_STATUS scan;
PendingRelSync *pendingsync;
@@ -757,9 +799,6 @@ smgrDoPendingSyncs(bool isCommit, bool isParallelWorker)
ForkNumber fork;
BlockNumber nblocks[MAX_FORKNUM + 1];
BlockNumber total_blocks = 0;
- SMgrRelation srel;
-
- srel = smgropen(pendingsync->rlocator, InvalidBackendId);
/*
* We emit newpage WAL records for smaller relations.
@@ -773,9 +812,12 @@ smgrDoPendingSyncs(bool isCommit, bool isParallelWorker)
{
for (fork = 0; fork <= MAX_FORKNUM; fork++)
{
- if (smgrexists(srel, fork))
+ SMgrFileHandle sfile;
+
+ sfile = smgropen(pendingsync->rlocator, InvalidBackendId, fork);
+ if (smgrexists(sfile))
{
- BlockNumber n = smgrnblocks(srel, fork);
+ BlockNumber n = smgrnblocks(sfile);
/* we shouldn't come here for unlogged relations */
Assert(fork != INIT_FORKNUM);
@@ -803,18 +845,19 @@ smgrDoPendingSyncs(bool isCommit, bool isParallelWorker)
total_blocks * BLCKSZ / 1024 >= wal_skip_threshold)
{
/* allocate the initial array, or extend it, if needed */
- if (maxrels == 0)
+ if (maxlocators == 0)
{
- maxrels = 8;
- srels = palloc(sizeof(SMgrRelation) * maxrels);
+ maxlocators = 8;
+ locators = palloc(sizeof(RelFileLocatorBackend) * maxlocators);
}
- else if (maxrels <= nrels)
+ else if (maxlocators <= nlocators)
{
- maxrels *= 2;
- srels = repalloc(srels, sizeof(SMgrRelation) * maxrels);
+ maxlocators *= 2;
+ locators = repalloc(locators, sizeof(RelFileLocatorBackend) * maxlocators);
}
- srels[nrels++] = srel;
+ locators[nlocators] = pendingsync->rlocator;
+ nlocators++;
}
else
{
@@ -833,7 +876,7 @@ smgrDoPendingSyncs(bool isCommit, bool isParallelWorker)
* page including any unused space. ReadBufferExtended()
* counts some pgstat events; unfortunately, we discard them.
*/
- rel = CreateFakeRelcacheEntry(srel->smgr_rlocator.locator);
+ rel = CreateFakeRelcacheEntry(pendingsync->rlocator);
log_newpage_range(rel, fork, 0, n, false);
FreeFakeRelcacheEntry(rel);
}
@@ -842,11 +885,20 @@ smgrDoPendingSyncs(bool isCommit, bool isParallelWorker)
pendingSyncHash = NULL;
- if (nrels > 0)
+ FlushRelationsAllBuffers(locators, nlocators);
+
+ for (int i = 0; i < nlocators; i++)
{
- smgrdosyncall(srels, nrels);
- pfree(srels);
+ for (int fork = 0; fork <= MAX_FORKNUM; fork++)
+ {
+ SMgrFileHandle sfile = smgropen(locators[i], InvalidBackendId, fork);
+
+ if (smgrexists(sfile))
+ smgrimmedsync(sfile);
+ }
}
+ if (locators != NULL)
+ pfree(locators);
}
/*
@@ -966,22 +1018,22 @@ smgr_redo(XLogReaderState *record)
if (info == XLOG_SMGR_CREATE)
{
xl_smgr_create *xlrec = (xl_smgr_create *) XLogRecGetData(record);
- SMgrRelation reln;
+ SMgrFileHandle sfile;
- reln = smgropen(xlrec->rlocator, InvalidBackendId);
- smgrcreate(reln, xlrec->forkNum, true);
+ sfile = smgropen(xlrec->rlocator, InvalidBackendId, xlrec->forkNum);
+ smgrcreate(sfile, true);
}
else if (info == XLOG_SMGR_TRUNCATE)
{
xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record);
- SMgrRelation reln;
+ SMgrFileHandle sfile;
Relation rel;
ForkNumber forks[MAX_FORKNUM];
BlockNumber blocks[MAX_FORKNUM];
int nforks = 0;
bool need_fsm_vacuum = false;
- reln = smgropen(xlrec->rlocator, InvalidBackendId);
+ sfile = smgropen(xlrec->rlocator, InvalidBackendId, MAIN_FORKNUM);
/*
* Forcibly create relation if it doesn't exist (which suggests that
@@ -989,7 +1041,7 @@ smgr_redo(XLogReaderState *record)
* XLogReadBufferForRedo, we prefer to recreate the rel and replay the
* log as best we can until the drop is seen.
*/
- smgrcreate(reln, MAIN_FORKNUM, true);
+ smgrcreate(sfile, true);
/*
* Before we perform the truncation, update minimum recovery point to
@@ -1022,8 +1074,10 @@ smgr_redo(XLogReaderState *record)
/* Prepare for truncation of FSM and VM too */
rel = CreateFakeRelcacheEntry(xlrec->rlocator);
+ DropRelationBuffers(xlrec->rlocator, InvalidBackendId, forks, nforks, blocks);
+
if ((xlrec->flags & SMGR_TRUNCATE_FSM) != 0 &&
- smgrexists(reln, FSM_FORKNUM))
+ smgrexists(smgropen(xlrec->rlocator, InvalidBackendId, FSM_FORKNUM)))
{
blocks[nforks] = FreeSpaceMapPrepareTruncateRel(rel, xlrec->blkno);
if (BlockNumberIsValid(blocks[nforks]))
@@ -1034,7 +1088,7 @@ smgr_redo(XLogReaderState *record)
}
}
if ((xlrec->flags & SMGR_TRUNCATE_VM) != 0 &&
- smgrexists(reln, VISIBILITYMAP_FORKNUM))
+ smgrexists(smgropen(xlrec->rlocator, InvalidBackendId, VISIBILITYMAP_FORKNUM)))
{
blocks[nforks] = visibilitymap_prepare_truncate(rel, xlrec->blkno);
if (BlockNumberIsValid(blocks[nforks]))
@@ -1046,7 +1100,10 @@ smgr_redo(XLogReaderState *record)
/* Do the real work to truncate relation forks */
if (nforks > 0)
- smgrtruncate(reln, forks, nforks, blocks);
+ {
+ DropRelationBuffers(xlrec->rlocator, InvalidBackendId, forks, nforks, blocks);
+ smgrtruncate_multi(xlrec->rlocator, InvalidBackendId, forks, nforks, blocks);
+ }
/*
* Update upper-level FSM pages to account for the truncation. This is
diff --git a/src/backend/commands/dbcommands.c b/src/backend/commands/dbcommands.c
index a0259cc5938..0d9ed2e0e0c 100644
--- a/src/backend/commands/dbcommands.c
+++ b/src/backend/commands/dbcommands.c
@@ -258,7 +258,7 @@ ScanSourceDatabasePgClass(Oid tbid, Oid dbid, char *srcpath)
List *rlocatorlist = NIL;
LockRelId relid;
Snapshot snapshot;
- SMgrRelation smgr;
+ SMgrFileHandle sfile;
BufferAccessStrategy bstrategy;
/* Get pg_class relfilenumber. */
@@ -275,9 +275,9 @@ ScanSourceDatabasePgClass(Oid tbid, Oid dbid, char *srcpath)
rlocator.dbOid = dbid;
rlocator.relNumber = relfilenumber;
- smgr = smgropen(rlocator, InvalidBackendId);
- nblocks = smgrnblocks(smgr, MAIN_FORKNUM);
- smgrclose(smgr);
+ sfile = smgropen(rlocator, InvalidBackendId, MAIN_FORKNUM);
+ nblocks = smgrnblocks(sfile);
+ smgrclose(sfile);
/* Use a buffer access strategy since this is a bulk read operation. */
bstrategy = GetAccessStrategy(BAS_BULKREAD);
@@ -439,7 +439,7 @@ ScanSourceDatabasePgClassTuple(HeapTupleData *tuple, Oid tbid, Oid dbid,
relinfo->rlocator.relNumber = relfilenumber;
relinfo->reloid = classForm->oid;
- /* Temporary relations were rejected above. */
+ /* Temporary relations were rejected ove. */
Assert(classForm->relpersistence != RELPERSISTENCE_TEMP);
relinfo->permanent =
(classForm->relpersistence == RELPERSISTENCE_PERMANENT) ? true : false;
diff --git a/src/backend/commands/sequence.c b/src/backend/commands/sequence.c
index bfe279cddf1..460e2bb1822 100644
--- a/src/backend/commands/sequence.c
+++ b/src/backend/commands/sequence.c
@@ -353,14 +353,14 @@ fill_seq_with_data(Relation rel, HeapTuple tuple)
if (rel->rd_rel->relpersistence == RELPERSISTENCE_UNLOGGED)
{
- SMgrRelation srel;
+ SMgrFileHandle sfile;
- srel = smgropen(rel->rd_locator, InvalidBackendId);
- smgrcreate(srel, INIT_FORKNUM, false);
+ sfile = smgropen(rel->rd_locator, InvalidBackendId, INIT_FORKNUM);
+ smgrcreate(sfile, false);
log_smgrcreate(&rel->rd_locator, INIT_FORKNUM);
fill_seq_fork_with_data(rel, tuple, INIT_FORKNUM);
FlushRelationBuffers(rel);
- smgrclose(srel);
+ smgrclose(sfile);
}
}
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 62d9917ca36..2033aa3e35e 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -14665,9 +14665,7 @@ AlterTableMoveAll(AlterTableMoveAllStmt *stmt)
static void
index_copy_data(Relation rel, RelFileLocator newrlocator)
{
- SMgrRelation dstrel;
-
- dstrel = smgropen(newrlocator, rel->rd_backend);
+ SMgrFileHandle dstmain;
/*
* Since we copy the file directly without looking at the shared buffers,
@@ -14687,16 +14685,20 @@ index_copy_data(Relation rel, RelFileLocator newrlocator)
RelationCreateStorage(newrlocator, rel->rd_rel->relpersistence, true);
/* copy main fork */
- RelationCopyStorage(RelationGetSmgr(rel), dstrel, MAIN_FORKNUM,
+ dstmain = smgropen(newrlocator, rel->rd_backend, MAIN_FORKNUM);
+ RelationCopyStorage(RelationGetSmgr(rel, MAIN_FORKNUM), dstmain,
rel->rd_rel->relpersistence);
/* copy those extra forks that exist */
for (ForkNumber forkNum = MAIN_FORKNUM + 1;
forkNum <= MAX_FORKNUM; forkNum++)
{
- if (smgrexists(RelationGetSmgr(rel), forkNum))
+ if (smgrexists(RelationGetSmgr(rel, forkNum)))
{
- smgrcreate(dstrel, forkNum, false);
+ SMgrFileHandle src_fork = RelationGetSmgr(rel, forkNum);
+ SMgrFileHandle dst_fork = smgropen(newrlocator, rel->rd_backend, forkNum);
+
+ smgrcreate(dst_fork, false);
/*
* WAL log creation if the relation is persistent, or this is the
@@ -14706,14 +14708,15 @@ index_copy_data(Relation rel, RelFileLocator newrlocator)
(rel->rd_rel->relpersistence == RELPERSISTENCE_UNLOGGED &&
forkNum == INIT_FORKNUM))
log_smgrcreate(&newrlocator, forkNum);
- RelationCopyStorage(RelationGetSmgr(rel), dstrel, forkNum,
+ RelationCopyStorage(src_fork, dst_fork,
rel->rd_rel->relpersistence);
+ smgrclose(dst_fork);
}
}
/* drop old relation, and close new one */
RelationDropStorage(rel);
- smgrclose(dstrel);
+ smgrclose(dstmain);
}
/*
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 0a05577b68d..1305eb7dee1 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -117,19 +117,6 @@ typedef struct CkptTsStatus
int index;
} CkptTsStatus;
-/*
- * Type for array used to sort SMgrRelations
- *
- * FlushRelationsAllBuffers shares the same comparator function with
- * DropRelationsAllBuffers. Pointer to this struct and RelFileLocator must be
- * compatible.
- */
-typedef struct SMgrSortArray
-{
- RelFileLocator rlocator; /* This must be the first member */
- SMgrRelation srel;
-} SMgrSortArray;
-
/* GUC variables */
bool zero_damaged_pages = false;
int bgwriter_lru_maxpages = 100;
@@ -450,8 +437,8 @@ ForgetPrivateRefCountEntry(PrivateRefCountEntry *ref)
)
-static Buffer ReadBuffer_common(SMgrRelation smgr, char relpersistence,
- ForkNumber forkNum, BlockNumber blockNum,
+static Buffer ReadBuffer_common(SMgrFileHandle sfile, char relpersistence,
+ BlockNumber blockNum,
ReadBufferMode mode, BufferAccessStrategy strategy,
bool *hit);
static bool PinBuffer(BufferDesc *buf, BufferAccessStrategy strategy);
@@ -467,13 +454,12 @@ static void TerminateBufferIO(BufferDesc *buf, bool clear_dirty,
uint32 set_flag_bits);
static void shared_buffer_write_error_callback(void *arg);
static void local_buffer_write_error_callback(void *arg);
-static BufferDesc *BufferAlloc(SMgrRelation smgr,
+static BufferDesc *BufferAlloc(SMgrFileHandle sfile,
char relpersistence,
- ForkNumber forkNum,
BlockNumber blockNum,
BufferAccessStrategy strategy,
bool *foundPtr, IOContext *io_context);
-static void FlushBuffer(BufferDesc *buf, SMgrRelation reln,
+static void FlushBuffer(BufferDesc *buf, SMgrFileHandle sfile,
IOObject io_object, IOContext io_context);
static void FindAndDropRelationBuffers(RelFileLocator rlocator,
ForkNumber forkNum,
@@ -494,9 +480,7 @@ static int ts_ckpt_progress_comparator(Datum a, Datum b, void *arg);
* Implementation of PrefetchBuffer() for shared buffers.
*/
PrefetchBufferResult
-PrefetchSharedBuffer(SMgrRelation smgr_reln,
- ForkNumber forkNum,
- BlockNumber blockNum)
+PrefetchSharedBuffer(SMgrFileHandle sfile, BlockNumber blockNum)
{
PrefetchBufferResult result = {InvalidBuffer, false};
BufferTag newTag; /* identity of requested block */
@@ -507,8 +491,8 @@ PrefetchSharedBuffer(SMgrRelation smgr_reln,
Assert(BlockNumberIsValid(blockNum));
/* create a tag so we can lookup the buffer */
- InitBufferTag(&newTag, &smgr_reln->smgr_rlocator.locator,
- forkNum, blockNum);
+ InitBufferTag(&newTag, &sfile->smgr_locator.locator,
+ sfile->smgr_locator.forknum, blockNum);
/* determine its hash code and partition lock ID */
newHash = BufTableHashCode(&newTag);
@@ -527,7 +511,7 @@ PrefetchSharedBuffer(SMgrRelation smgr_reln,
* Try to initiate an asynchronous read. This returns false in
* recovery if the relation file doesn't exist.
*/
- if (smgrprefetch(smgr_reln, forkNum, blockNum))
+ if (smgrprefetch(sfile, blockNum))
result.initiated_io = true;
#endif /* USE_PREFETCH */
}
@@ -595,12 +579,12 @@ PrefetchBuffer(Relation reln, ForkNumber forkNum, BlockNumber blockNum)
errmsg("cannot access temporary tables of other sessions")));
/* pass it off to localbuf.c */
- return PrefetchLocalBuffer(RelationGetSmgr(reln), forkNum, blockNum);
+ return PrefetchLocalBuffer(RelationGetSmgr(reln, forkNum), blockNum);
}
else
{
/* pass it to the shared buffer version */
- return PrefetchSharedBuffer(RelationGetSmgr(reln), forkNum, blockNum);
+ return PrefetchSharedBuffer(RelationGetSmgr(reln, forkNum), blockNum);
}
}
@@ -769,8 +753,9 @@ ReadBufferExtended(Relation reln, ForkNumber forkNum, BlockNumber blockNum,
* miss.
*/
pgstat_count_buffer_read(reln);
- buf = ReadBuffer_common(RelationGetSmgr(reln), reln->rd_rel->relpersistence,
- forkNum, blockNum, mode, strategy, &hit);
+ buf = ReadBuffer_common(RelationGetSmgr(reln, forkNum),
+ reln->rd_rel->relpersistence,
+ blockNum, mode, strategy, &hit);
if (hit)
pgstat_count_buffer_hit(reln);
return buf;
@@ -794,10 +779,10 @@ ReadBufferWithoutRelcache(RelFileLocator rlocator, ForkNumber forkNum,
{
bool hit;
- SMgrRelation smgr = smgropen(rlocator, InvalidBackendId);
+ SMgrFileHandle sfile = smgropen(rlocator, InvalidBackendId, forkNum);
- return ReadBuffer_common(smgr, permanent ? RELPERSISTENCE_PERMANENT :
- RELPERSISTENCE_UNLOGGED, forkNum, blockNum,
+ return ReadBuffer_common(sfile, permanent ? RELPERSISTENCE_PERMANENT :
+ RELPERSISTENCE_UNLOGGED, blockNum,
mode, strategy, &hit);
}
@@ -808,7 +793,7 @@ ReadBufferWithoutRelcache(RelFileLocator rlocator, ForkNumber forkNum,
* *hit is set to true if the request was satisfied from shared buffer cache.
*/
static Buffer
-ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
+ReadBuffer_common(SMgrFileHandle sfile, char relpersistence,
BlockNumber blockNum, ReadBufferMode mode,
BufferAccessStrategy strategy, bool *hit)
{
@@ -818,7 +803,7 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
IOContext io_context;
IOObject io_object;
bool isExtend;
- bool isLocalBuf = SmgrIsTemp(smgr);
+ bool isLocalBuf = SmgrIsTemp(sfile);
*hit = false;
@@ -827,23 +812,24 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
isExtend = (blockNum == P_NEW);
- TRACE_POSTGRESQL_BUFFER_READ_START(forkNum, blockNum,
- smgr->smgr_rlocator.locator.spcOid,
- smgr->smgr_rlocator.locator.dbOid,
- smgr->smgr_rlocator.locator.relNumber,
- smgr->smgr_rlocator.backend,
+ TRACE_POSTGRESQL_BUFFER_READ_START(smgr->smgr_locator.forkNum,
+ blockNum,
+ smgr->smgr_locator.locator.spcOid,
+ smgr->smgr_locator.locator.dbOid,
+ smgr->smgr_locator.locator.relNumber,
+ smgr->smgr_locator.backend,
isExtend);
/* Substitute proper block number if caller asked for P_NEW */
if (isExtend)
{
- blockNum = smgrnblocks(smgr, forkNum);
+ blockNum = smgrnblocks(sfile);
/* Fail if relation is already at maximum possible length */
if (blockNum == P_NEW)
ereport(ERROR,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
errmsg("cannot extend relation %s beyond %u blocks",
- relpath(smgr->smgr_rlocator, forkNum),
+ smgrfilepath(sfile->smgr_locator),
P_NEW)));
}
@@ -856,7 +842,7 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
* rely on IOContextForStrategy() to set the right IOContext for us.
* This may happen in cases like CREATE TEMPORARY TABLE AS...
*/
- bufHdr = LocalBufferAlloc(smgr, forkNum, blockNum, &found, &io_context);
+ bufHdr = LocalBufferAlloc(sfile, blockNum, &found, &io_context);
if (found)
pgBufferUsage.local_blks_hit++;
else if (isExtend)
@@ -871,7 +857,7 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
* lookup the buffer. IO_IN_PROGRESS is set if the requested block is
* not currently in memory.
*/
- bufHdr = BufferAlloc(smgr, relpersistence, forkNum, blockNum,
+ bufHdr = BufferAlloc(sfile, relpersistence, blockNum,
strategy, &found, &io_context);
if (found)
pgBufferUsage.shared_blks_hit++;
@@ -896,11 +882,12 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
if (VacuumCostActive)
VacuumCostBalance += VacuumCostPageHit;
- TRACE_POSTGRESQL_BUFFER_READ_DONE(forkNum, blockNum,
- smgr->smgr_rlocator.locator.spcOid,
- smgr->smgr_rlocator.locator.dbOid,
- smgr->smgr_rlocator.locator.relNumber,
- smgr->smgr_rlocator.backend,
+ TRACE_POSTGRESQL_BUFFER_READ_DONE(sfile->smgr_forknum,
+ blockNum,
+ sfile->smgr_rlocator.locator.spcOid,
+ sfile->smgr_rlocator.locator.dbOid,
+ sfile->smgr_rlocator.locator.relNumber,
+ sfile->smgr_rlocator.backend,
isExtend,
found);
@@ -937,7 +924,7 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
if (!PageIsNew((Page) bufBlock))
ereport(ERROR,
(errmsg("unexpected data beyond EOF in block %u of relation %s",
- blockNum, relpath(smgr->smgr_rlocator, forkNum)),
+ blockNum, smgrfilepath(sfile->smgr_locator)),
errhint("This has been seen to occur with buggy kernels; consider updating your system.")));
/*
@@ -1003,7 +990,7 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
/* new buffers are zero-filled */
MemSet((char *) bufBlock, 0, BLCKSZ);
/* don't set checksum for all-zero page */
- smgrextend(smgr, forkNum, blockNum, bufBlock, false);
+ smgrextend(sfile, blockNum, bufBlock, false);
pgstat_count_io_op(io_object, io_context, IOOP_EXTEND);
@@ -1032,7 +1019,7 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
else
INSTR_TIME_SET_ZERO(io_start);
- smgrread(smgr, forkNum, blockNum, bufBlock);
+ smgrread(sfile, blockNum, bufBlock);
pgstat_count_io_op(io_object, io_context, IOOP_READ);
@@ -1054,7 +1041,7 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
(errcode(ERRCODE_DATA_CORRUPTED),
errmsg("invalid page in block %u of relation %s; zeroing out page",
blockNum,
- relpath(smgr->smgr_rlocator, forkNum))));
+ smgrfilepath(sfile->smgr_locator))));
MemSet((char *) bufBlock, 0, BLCKSZ);
}
else
@@ -1062,7 +1049,7 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
(errcode(ERRCODE_DATA_CORRUPTED),
errmsg("invalid page in block %u of relation %s",
blockNum,
- relpath(smgr->smgr_rlocator, forkNum))));
+ smgrfilepath(sfile->smgr_locator))));
}
}
}
@@ -1101,11 +1088,12 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
if (VacuumCostActive)
VacuumCostBalance += VacuumCostPageMiss;
- TRACE_POSTGRESQL_BUFFER_READ_DONE(forkNum, blockNum,
- smgr->smgr_rlocator.locator.spcOid,
- smgr->smgr_rlocator.locator.dbOid,
- smgr->smgr_rlocator.locator.relNumber,
- smgr->smgr_rlocator.backend,
+ TRACE_POSTGRESQL_BUFFER_READ_DONE(sfile->smgr_locator.forknum,
+ blockNum,
+ sfile->smgr_locator.locator.spcOid,
+ sfile->smgr_locator.locator.dbOid,
+ sfile->smgr_locator.locator.relNumber,
+ sfile->smgr_locator.backend,
isExtend,
found);
@@ -1136,7 +1124,7 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
* No locks are held either at entry or exit.
*/
static BufferDesc *
-BufferAlloc(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
+BufferAlloc(SMgrFileHandle sfile, char relpersistence,
BlockNumber blockNum,
BufferAccessStrategy strategy,
bool *foundPtr, IOContext *io_context)
@@ -1155,7 +1143,7 @@ BufferAlloc(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
uint32 buf_state;
/* create a tag so we can lookup the buffer */
- InitBufferTag(&newTag, &smgr->smgr_rlocator.locator, forkNum, blockNum);
+ InitBufferTag(&newTag, &sfile->smgr_locator.locator, sfile->smgr_locator.forknum, blockNum);
/* determine its hash code and partition lock ID */
newHash = BufTableHashCode(&newTag);
@@ -1290,10 +1278,11 @@ BufferAlloc(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
}
/* OK, do the I/O */
- TRACE_POSTGRESQL_BUFFER_WRITE_DIRTY_START(forkNum, blockNum,
- smgr->smgr_rlocator.locator.spcOid,
- smgr->smgr_rlocator.locator.dbOid,
- smgr->smgr_rlocator.locator.relNumber);
+ TRACE_POSTGRESQL_BUFFER_WRITE_DIRTY_START(sfile->smgr_locator.forknum,
+ blockNum,
+ sfile->smgr_locator.locator.spcOid,
+ sfile->smgr_locator.locator.dbOid,
+ sfile->smgr_locator.locator.relNumber);
FlushBuffer(buf, NULL, IOOBJECT_RELATION, *io_context);
LWLockRelease(BufferDescriptorGetContentLock(buf));
@@ -1301,10 +1290,11 @@ BufferAlloc(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
ScheduleBufferTagForWriteback(&BackendWritebackContext,
&buf->tag);
- TRACE_POSTGRESQL_BUFFER_WRITE_DIRTY_DONE(forkNum, blockNum,
- smgr->smgr_rlocator.locator.spcOid,
- smgr->smgr_rlocator.locator.dbOid,
- smgr->smgr_rlocator.locator.relNumber);
+ TRACE_POSTGRESQL_BUFFER_WRITE_DIRTY_DONE(sfile->smgr_locator.forknum,
+ blockNum,
+ sfile->smgr_locator.locator.spcOid,
+ sfile->smgr_locator.locator.dbOid,
+ sfile->smgr_locator.locator.relNumber);
}
else
{
@@ -1460,7 +1450,7 @@ BufferAlloc(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
buf_state &= ~(BM_VALID | BM_DIRTY | BM_JUST_DIRTIED |
BM_CHECKPOINT_NEEDED | BM_IO_ERROR | BM_PERMANENT |
BUF_USAGECOUNT_MASK);
- if (relpersistence == RELPERSISTENCE_PERMANENT || forkNum == INIT_FORKNUM)
+ if (relpersistence == RELPERSISTENCE_PERMANENT || sfile->smgr_locator.forknum == INIT_FORKNUM)
buf_state |= BM_TAG_VALID | BM_PERMANENT | BUF_USAGECOUNT_ONE;
else
buf_state |= BM_TAG_VALID | BUF_USAGECOUNT_ONE;
@@ -2868,7 +2858,7 @@ BufferGetTag(Buffer buffer, RelFileLocator *rlocator, ForkNumber *forknum,
* as the second parameter. If not, pass NULL.
*/
static void
-FlushBuffer(BufferDesc *buf, SMgrRelation reln, IOObject io_object,
+FlushBuffer(BufferDesc *buf, SMgrFileHandle sfile, IOObject io_object,
IOContext io_context)
{
XLogRecPtr recptr;
@@ -2894,14 +2884,14 @@ FlushBuffer(BufferDesc *buf, SMgrRelation reln, IOObject io_object,
error_context_stack = &errcallback;
/* Find smgr relation for buffer */
- if (reln == NULL)
- reln = smgropen(BufTagGetRelFileLocator(&buf->tag), InvalidBackendId);
+ if (sfile == NULL)
+ sfile = smgropen(BufTagGetRelFileLocator(&buf->tag), InvalidBackendId, BufTagGetForkNum(&buf->tag));
TRACE_POSTGRESQL_BUFFER_FLUSH_START(BufTagGetForkNum(&buf->tag),
buf->tag.blockNum,
- reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
- reln->smgr_rlocator.locator.relNumber);
+ sfile->smgr_locator.locator.spcOid,
+ sfile->smgr_locator.locator.dbOid,
+ sfile->smgr_locator.locator.relNumber);
buf_state = LockBufHdr(buf);
@@ -2957,8 +2947,7 @@ FlushBuffer(BufferDesc *buf, SMgrRelation reln, IOObject io_object,
/*
* bufToWrite is either the shared buffer or a copy, as appropriate.
*/
- smgrwrite(reln,
- BufTagGetForkNum(&buf->tag),
+ smgrwrite(sfile,
buf->tag.blockNum,
bufToWrite,
false);
@@ -3001,9 +2990,9 @@ FlushBuffer(BufferDesc *buf, SMgrRelation reln, IOObject io_object,
TRACE_POSTGRESQL_BUFFER_FLUSH_DONE(BufTagGetForkNum(&buf->tag),
buf->tag.blockNum,
- reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
- reln->smgr_rlocator.locator.relNumber);
+ sfile->smgr_locator.locator.spcOid,
+ sfile->smgr_locator.locator.dbOid,
+ sfile->smgr_locator.locator.relNumber);
/* Pop the error context stack */
error_context_stack = errcallback.previous;
@@ -3036,7 +3025,7 @@ RelationGetNumberOfBlocksInFork(Relation relation, ForkNumber forkNum)
}
else if (RELKIND_HAS_STORAGE(relation->rd_rel->relkind))
{
- return smgrnblocks(RelationGetSmgr(relation), forkNum);
+ return smgrnblocks(RelationGetSmgr(relation, forkNum));
}
else
Assert(false);
@@ -3126,26 +3115,20 @@ BufferGetLSNAtomic(Buffer buffer)
* --------------------------------------------------------------------
*/
void
-DropRelationBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum,
+DropRelationBuffers(RelFileLocator rlocator, BackendId backend, ForkNumber *forkNum,
int nforks, BlockNumber *firstDelBlock)
{
int i;
int j;
- RelFileLocatorBackend rlocator;
BlockNumber nForkBlock[MAX_FORKNUM];
uint64 nBlocksToInvalidate = 0;
- rlocator = smgr_reln->smgr_rlocator;
-
/* If it's a local relation, it's localbuf.c's problem. */
- if (RelFileLocatorBackendIsTemp(rlocator))
+ if (backend == MyBackendId)
{
- if (rlocator.backend == MyBackendId)
- {
- for (j = 0; j < nforks; j++)
- DropRelationLocalBuffers(rlocator.locator, forkNum[j],
- firstDelBlock[j]);
- }
+ for (j = 0; j < nforks; j++)
+ DropRelationLocalBuffers(rlocator, forkNum[j],
+ firstDelBlock[j]);
return;
}
@@ -3174,7 +3157,10 @@ DropRelationBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum,
for (i = 0; i < nforks; i++)
{
/* Get the number of blocks for a relation's fork */
- nForkBlock[i] = smgrnblocks_cached(smgr_reln, forkNum[i]);
+ SMgrFileHandle sfile;
+
+ sfile = smgropen(rlocator, backend, forkNum[i]);
+ nForkBlock[i] = smgrnblocks_cached(sfile);
if (nForkBlock[i] == InvalidBlockNumber)
{
@@ -3194,7 +3180,7 @@ DropRelationBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum,
nBlocksToInvalidate < BUF_DROP_FULL_SCAN_THRESHOLD)
{
for (j = 0; j < nforks; j++)
- FindAndDropRelationBuffers(rlocator.locator, forkNum[j],
+ FindAndDropRelationBuffers(rlocator, forkNum[j],
nForkBlock[j], firstDelBlock[j]);
return;
}
@@ -3220,14 +3206,14 @@ DropRelationBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum,
* We could check forkNum and blockNum as well as the rlocator, but
* the incremental win from doing so seems small.
*/
- if (!BufTagMatchesRelFileLocator(&bufHdr->tag, &rlocator.locator))
+ if (!BufTagMatchesRelFileLocator(&bufHdr->tag, &rlocator))
continue;
buf_state = LockBufHdr(bufHdr);
for (j = 0; j < nforks; j++)
{
- if (BufTagMatchesRelFileLocator(&bufHdr->tag, &rlocator.locator) &&
+ if (BufTagMatchesRelFileLocator(&bufHdr->tag, &rlocator) &&
BufTagGetForkNum(&bufHdr->tag) == forkNum[j] &&
bufHdr->tag.blockNum >= firstDelBlock[j])
{
@@ -3249,11 +3235,10 @@ DropRelationBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum,
* --------------------------------------------------------------------
*/
void
-DropRelationsAllBuffers(SMgrRelation *smgr_reln, int nlocators)
+DropRelationsAllBuffers(RelFileLocatorBackend *rlocators, int nlocators)
{
int i;
int n = 0;
- SMgrRelation *rels;
BlockNumber (*block)[MAX_FORKNUM + 1];
uint64 nBlocksToInvalidate = 0;
RelFileLocator *locators;
@@ -3263,18 +3248,18 @@ DropRelationsAllBuffers(SMgrRelation *smgr_reln, int nlocators)
if (nlocators == 0)
return;
- rels = palloc(sizeof(SMgrRelation) * nlocators); /* non-local relations */
+ locators = palloc(sizeof(RelFileLocator) * nlocators); /* non-local relations */
/* If it's a local relation, it's localbuf.c's problem. */
for (i = 0; i < nlocators; i++)
{
- if (RelFileLocatorBackendIsTemp(smgr_reln[i]->smgr_rlocator))
+ if (rlocators[i].backend != InvalidBackendId)
{
- if (smgr_reln[i]->smgr_rlocator.backend == MyBackendId)
- DropRelationAllLocalBuffers(smgr_reln[i]->smgr_rlocator.locator);
+ if (rlocators[i].backend == MyBackendId)
+ DropRelationAllLocalBuffers(rlocators[i].locator);
}
else
- rels[n++] = smgr_reln[i];
+ locators[n++] = rlocators[i].locator;
}
/*
@@ -3283,7 +3268,7 @@ DropRelationsAllBuffers(SMgrRelation *smgr_reln, int nlocators)
*/
if (n == 0)
{
- pfree(rels);
+ pfree(locators);
return;
}
@@ -3303,12 +3288,13 @@ DropRelationsAllBuffers(SMgrRelation *smgr_reln, int nlocators)
for (int j = 0; j <= MAX_FORKNUM; j++)
{
/* Get the number of blocks for a relation's fork. */
- block[i][j] = smgrnblocks_cached(rels[i], j);
+ SMgrFileHandle sfile = smgropen(locators[i], InvalidBackendId, j);
+ block[i][j] = smgrnblocks_cached(sfile);
/* We need to only consider the relation forks that exists. */
if (block[i][j] == InvalidBlockNumber)
{
- if (!smgrexists(rels[i], j))
+ if (!smgrexists(sfile))
continue;
cached = false;
break;
@@ -3334,20 +3320,17 @@ DropRelationsAllBuffers(SMgrRelation *smgr_reln, int nlocators)
continue;
/* drop all the buffers for a particular relation fork */
- FindAndDropRelationBuffers(rels[i]->smgr_rlocator.locator,
+ FindAndDropRelationBuffers(locators[i],
j, block[i][j], 0);
}
}
pfree(block);
- pfree(rels);
+ pfree(locators);
return;
}
pfree(block);
- locators = palloc(sizeof(RelFileLocator) * n); /* non-local relations */
- for (i = 0; i < n; i++)
- locators[i] = rels[i]->smgr_rlocator.locator;
/*
* For low number of relations to drop just use a simple walk through, to
@@ -3407,7 +3390,6 @@ DropRelationsAllBuffers(SMgrRelation *smgr_reln, int nlocators)
}
pfree(locators);
- pfree(rels);
}
/* ---------------------------------------------------------------------
@@ -3616,8 +3598,7 @@ FlushRelationBuffers(Relation rel)
PageSetChecksumInplace(localpage, bufHdr->tag.blockNum);
- smgrwrite(RelationGetSmgr(rel),
- BufTagGetForkNum(&bufHdr->tag),
+ smgrwrite(RelationGetSmgr(rel, BufTagGetForkNum(&bufHdr->tag)),
bufHdr->tag.blockNum,
localpage,
false);
@@ -3659,7 +3640,8 @@ FlushRelationBuffers(Relation rel)
{
PinBuffer_Locked(bufHdr);
LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), LW_SHARED);
- FlushBuffer(bufHdr, RelationGetSmgr(rel), IOOBJECT_RELATION, IOCONTEXT_NORMAL);
+ FlushBuffer(bufHdr, RelationGetSmgr(rel, BufTagGetForkNum(&bufHdr->tag)),
+ IOOBJECT_RELATION, IOCONTEXT_NORMAL);
LWLockRelease(BufferDescriptorGetContentLock(bufHdr));
UnpinBuffer(bufHdr);
}
@@ -3678,25 +3660,20 @@ FlushRelationBuffers(Relation rel)
* --------------------------------------------------------------------
*/
void
-FlushRelationsAllBuffers(SMgrRelation *smgrs, int nrels)
+FlushRelationsAllBuffers(RelFileLocator *rels, int nrels)
{
int i;
- SMgrSortArray *srels;
+ RelFileLocator *locators = NULL;
bool use_bsearch;
if (nrels == 0)
return;
/* fill-in array for qsort */
- srels = palloc(sizeof(SMgrSortArray) * nrels);
+ locators = palloc(sizeof(RelFileLocator) * nrels);
for (i = 0; i < nrels; i++)
- {
- Assert(!RelFileLocatorBackendIsTemp(smgrs[i]->smgr_rlocator));
-
- srels[i].rlocator = smgrs[i]->smgr_rlocator.locator;
- srels[i].srel = smgrs[i];
- }
+ locators[i] = rels[i];
/*
* Save the bsearch overhead for low number of relations to sync. See
@@ -3704,16 +3681,16 @@ FlushRelationsAllBuffers(SMgrRelation *smgrs, int nrels)
*/
use_bsearch = nrels > RELS_BSEARCH_THRESHOLD;
- /* sort the list of SMgrRelations if necessary */
+ /* sort the list of locators if necessary */
if (use_bsearch)
- pg_qsort(srels, nrels, sizeof(SMgrSortArray), rlocator_comparator);
+ pg_qsort(locators, nrels, sizeof(RelFileLocator), rlocator_comparator);
/* Make sure we can handle the pin inside the loop */
ResourceOwnerEnlargeBuffers(CurrentResourceOwner);
for (i = 0; i < NBuffers; i++)
{
- SMgrSortArray *srelent = NULL;
+ RelFileLocator *found = NULL;
BufferDesc *bufHdr = GetBufferDescriptor(i);
uint32 buf_state;
@@ -3728,9 +3705,9 @@ FlushRelationsAllBuffers(SMgrRelation *smgrs, int nrels)
for (j = 0; j < nrels; j++)
{
- if (BufTagMatchesRelFileLocator(&bufHdr->tag, &srels[j].rlocator))
+ if (BufTagMatchesRelFileLocator(&bufHdr->tag, &locators[j]))
{
- srelent = &srels[j];
+ found = &locators[j];
break;
}
}
@@ -3740,24 +3717,26 @@ FlushRelationsAllBuffers(SMgrRelation *smgrs, int nrels)
RelFileLocator rlocator;
rlocator = BufTagGetRelFileLocator(&bufHdr->tag);
- srelent = bsearch((const void *) &(rlocator),
- srels, nrels, sizeof(SMgrSortArray),
- rlocator_comparator);
+ found = bsearch((const void *) &rlocator,
+ locators, nrels, sizeof(RelFileLocator),
+ rlocator_comparator);
}
/* buffer doesn't belong to any of the given relfilelocators; skip it */
- if (srelent == NULL)
+ if (found == NULL)
continue;
+ /* FIXME: cache SMgrFileHandles for the rels, and pass to FlushBuffer */
+
ReservePrivateRefCountEntry();
buf_state = LockBufHdr(bufHdr);
- if (BufTagMatchesRelFileLocator(&bufHdr->tag, &srelent->rlocator) &&
+ if (BufTagMatchesRelFileLocator(&bufHdr->tag, found) &&
(buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY))
{
PinBuffer_Locked(bufHdr);
LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), LW_SHARED);
- FlushBuffer(bufHdr, srelent->srel, IOOBJECT_RELATION, IOCONTEXT_NORMAL);
+ FlushBuffer(bufHdr, NULL, IOOBJECT_RELATION, IOCONTEXT_NORMAL);
LWLockRelease(BufferDescriptorGetContentLock(bufHdr));
UnpinBuffer(bufHdr);
}
@@ -3765,7 +3744,7 @@ FlushRelationsAllBuffers(SMgrRelation *smgrs, int nrels)
UnlockBufHdr(bufHdr, buf_state);
}
- pfree(srels);
+ pfree(locators);
}
/* ---------------------------------------------------------------------
@@ -3802,8 +3781,7 @@ RelationCopyStorageUsingBuffer(RelFileLocator srclocator,
use_wal = XLogIsNeeded() && (permanent || forkNum == INIT_FORKNUM);
/* Get number of blocks in the source relation. */
- nblocks = smgrnblocks(smgropen(srclocator, InvalidBackendId),
- forkNum);
+ nblocks = smgrnblocks(smgropen(srclocator, InvalidBackendId, forkNum));
/* Nothing to copy; just return. */
if (nblocks == 0)
@@ -3814,7 +3792,7 @@ RelationCopyStorageUsingBuffer(RelFileLocator srclocator,
* relation before starting to copy block by block.
*/
memset(buf.data, 0, BLCKSZ);
- smgrextend(smgropen(dstlocator, InvalidBackendId), forkNum, nblocks - 1,
+ smgrextend(smgropen(dstlocator, InvalidBackendId, forkNum), nblocks - 1,
buf.data, true);
/* This is a bulk operation, so use buffer access strategies. */
@@ -3872,7 +3850,6 @@ void
CreateAndCopyRelationData(RelFileLocator src_rlocator,
RelFileLocator dst_rlocator, bool permanent)
{
- RelFileLocatorBackend rlocator;
char relpersistence;
/* Set the relpersistence. */
@@ -3895,9 +3872,9 @@ CreateAndCopyRelationData(RelFileLocator src_rlocator,
for (ForkNumber forkNum = MAIN_FORKNUM + 1;
forkNum <= MAX_FORKNUM; forkNum++)
{
- if (smgrexists(smgropen(src_rlocator, InvalidBackendId), forkNum))
+ if (smgrexists(smgropen(src_rlocator, InvalidBackendId, forkNum)))
{
- smgrcreate(smgropen(dst_rlocator, InvalidBackendId), forkNum, false);
+ smgrcreate(smgropen(dst_rlocator, InvalidBackendId, forkNum), false);
/*
* WAL log creation if the relation is persistent, or this is the
@@ -3913,13 +3890,8 @@ CreateAndCopyRelationData(RelFileLocator src_rlocator,
}
/* close source and destination smgr if exists. */
- rlocator.backend = InvalidBackendId;
-
- rlocator.locator = src_rlocator;
- smgrcloserellocator(rlocator);
-
- rlocator.locator = dst_rlocator;
- smgrcloserellocator(rlocator);
+ smgrcloserellocator(src_rlocator, InvalidBackendId);
+ smgrcloserellocator(dst_rlocator, InvalidBackendId);
}
/* ---------------------------------------------------------------------
@@ -5033,7 +5005,7 @@ IssuePendingWritebacks(WritebackContext *context)
{
PendingWriteback *cur;
PendingWriteback *next;
- SMgrRelation reln;
+ SMgrFileHandle sfile;
int ahead;
BufferTag tag;
RelFileLocator currlocator;
@@ -5073,8 +5045,8 @@ IssuePendingWritebacks(WritebackContext *context)
i += ahead;
/* and finally tell the kernel to write the data to storage */
- reln = smgropen(currlocator, InvalidBackendId);
- smgrwriteback(reln, BufTagGetForkNum(&tag), tag.blockNum, nblocks);
+ sfile = smgropen(currlocator, InvalidBackendId, BufTagGetForkNum(&tag));
+ smgrwriteback(sfile, tag.blockNum, nblocks);
}
context->nr_pending = 0;
diff --git a/src/backend/storage/buffer/localbuf.c b/src/backend/storage/buffer/localbuf.c
index 5325ddb663d..62849fb2d11 100644
--- a/src/backend/storage/buffer/localbuf.c
+++ b/src/backend/storage/buffer/localbuf.c
@@ -62,14 +62,13 @@ static Block GetLocalBufferStorage(void);
* No-op if prefetching isn't compiled in.
*/
PrefetchBufferResult
-PrefetchLocalBuffer(SMgrRelation smgr, ForkNumber forkNum,
- BlockNumber blockNum)
+PrefetchLocalBuffer(SMgrFileHandle sfile, BlockNumber blockNum)
{
PrefetchBufferResult result = {InvalidBuffer, false};
BufferTag newTag; /* identity of requested block */
LocalBufferLookupEnt *hresult;
- InitBufferTag(&newTag, &smgr->smgr_rlocator.locator, forkNum, blockNum);
+ InitBufferTag(&newTag, &sfile->smgr_locator.locator, sfile->smgr_locator.forknum, blockNum);
/* Initialize local buffers if first request in this session */
if (LocalBufHash == NULL)
@@ -88,7 +87,7 @@ PrefetchLocalBuffer(SMgrRelation smgr, ForkNumber forkNum,
{
#ifdef USE_PREFETCH
/* Not in buffers, so initiate prefetch */
- smgrprefetch(smgr, forkNum, blockNum);
+ smgrprefetch(sfile, blockNum);
result.initiated_io = true;
#endif /* USE_PREFETCH */
}
@@ -107,7 +106,7 @@ PrefetchLocalBuffer(SMgrRelation smgr, ForkNumber forkNum,
* (hence, usage_count is always advanced).
*/
BufferDesc *
-LocalBufferAlloc(SMgrRelation smgr, ForkNumber forkNum, BlockNumber blockNum,
+LocalBufferAlloc(SMgrFileHandle sfile, BlockNumber blockNum,
bool *foundPtr, IOContext *io_context)
{
BufferTag newTag; /* identity of requested block */
@@ -118,7 +117,7 @@ LocalBufferAlloc(SMgrRelation smgr, ForkNumber forkNum, BlockNumber blockNum,
bool found;
uint32 buf_state;
- InitBufferTag(&newTag, &smgr->smgr_rlocator.locator, forkNum, blockNum);
+ InitBufferTag(&newTag, &sfile->smgr_locator.locator, sfile->smgr_locator.forknum, blockNum);
/* Initialize local buffers if first request in this session */
if (LocalBufHash == NULL)
@@ -143,7 +142,7 @@ LocalBufferAlloc(SMgrRelation smgr, ForkNumber forkNum, BlockNumber blockNum,
Assert(BufferTagsEqual(&bufHdr->tag, &newTag));
#ifdef LBDEBUG
fprintf(stderr, "LB ALLOC (%u,%d,%d) %d\n",
- smgr->smgr_rlocator.locator.relNumber, forkNum, blockNum, -b - 1);
+ sfile->smgr_locator.locator.relNumber, sfile->smgr_locator.forknum, blockNum, -b - 1);
#endif
buf_state = pg_atomic_read_u32(&bufHdr->state);
@@ -171,7 +170,7 @@ LocalBufferAlloc(SMgrRelation smgr, ForkNumber forkNum, BlockNumber blockNum,
#ifdef LBDEBUG
fprintf(stderr, "LB ALLOC (%u,%d,%d) %d\n",
- smgr->smgr_rlocator.locator.relNumber, forkNum, blockNum,
+ sfile->smgr_locator.locator.relNumber, sfile->smgr_locator.forknum, blockNum,
-nextFreeLocalBuf - 1);
#endif
@@ -220,17 +219,16 @@ LocalBufferAlloc(SMgrRelation smgr, ForkNumber forkNum, BlockNumber blockNum,
*/
if (buf_state & BM_DIRTY)
{
- SMgrRelation oreln;
+ SMgrFileHandle ofile;
Page localpage = (char *) LocalBufHdrGetBlock(bufHdr);
- /* Find smgr relation for buffer */
- oreln = smgropen(BufTagGetRelFileLocator(&bufHdr->tag), MyBackendId);
+ /* Find smgr file handle for buffer */
+ ofile = smgropen(BufTagGetRelFileLocator(&bufHdr->tag), MyBackendId, BufTagGetForkNum(&bufHdr->tag));
PageSetChecksumInplace(localpage, bufHdr->tag.blockNum);
/* And write... */
- smgrwrite(oreln,
- BufTagGetForkNum(&bufHdr->tag),
+ smgrwrite(ofile,
bufHdr->tag.blockNum,
localpage,
false);
diff --git a/src/backend/storage/freespace/freespace.c b/src/backend/storage/freespace/freespace.c
index 3e9693b293b..a825a0b5af3 100644
--- a/src/backend/storage/freespace/freespace.c
+++ b/src/backend/storage/freespace/freespace.c
@@ -271,7 +271,7 @@ FreeSpaceMapPrepareTruncateRel(Relation rel, BlockNumber nblocks)
* If no FSM has been created yet for this relation, there's nothing to
* truncate.
*/
- if (!smgrexists(RelationGetSmgr(rel), FSM_FORKNUM))
+ if (!smgrexists(RelationGetSmgr(rel, FSM_FORKNUM)))
return InvalidBlockNumber;
/* Get the location in the FSM of the first removed heap block */
@@ -317,7 +317,7 @@ FreeSpaceMapPrepareTruncateRel(Relation rel, BlockNumber nblocks)
else
{
new_nfsmblocks = fsm_logical_to_physical(first_removed_address);
- if (smgrnblocks(RelationGetSmgr(rel), FSM_FORKNUM) <= new_nfsmblocks)
+ if (smgrnblocks(RelationGetSmgr(rel, FSM_FORKNUM)) <= new_nfsmblocks)
return InvalidBlockNumber; /* nothing to do; the FSM was already
* smaller */
}
@@ -532,14 +532,14 @@ fsm_readbuf(Relation rel, FSMAddress addr, bool extend)
{
BlockNumber blkno = fsm_logical_to_physical(addr);
Buffer buf;
- SMgrRelation reln;
+ SMgrFileHandle fsm_file;
/*
* Caution: re-using this smgr pointer could fail if the relcache entry
* gets closed. It's safe as long as we only do smgr-level operations
* between here and the last use of the pointer.
*/
- reln = RelationGetSmgr(rel);
+ fsm_file = RelationGetSmgr(rel, FSM_FORKNUM);
/*
* If we haven't cached the size of the FSM yet, check it first. Also
@@ -547,19 +547,19 @@ fsm_readbuf(Relation rel, FSMAddress addr, bool extend)
* value might be stale. (We send smgr inval messages on truncation, but
* not on extension.)
*/
- if (reln->smgr_cached_nblocks[FSM_FORKNUM] == InvalidBlockNumber ||
- blkno >= reln->smgr_cached_nblocks[FSM_FORKNUM])
+ if (fsm_file->smgr_cached_nblocks == InvalidBlockNumber ||
+ blkno >= fsm_file->smgr_cached_nblocks)
{
/* Invalidate the cache so smgrnblocks asks the kernel. */
- reln->smgr_cached_nblocks[FSM_FORKNUM] = InvalidBlockNumber;
- if (smgrexists(reln, FSM_FORKNUM))
- smgrnblocks(reln, FSM_FORKNUM);
+ fsm_file->smgr_cached_nblocks = InvalidBlockNumber;
+ if (smgrexists(fsm_file))
+ smgrnblocks(fsm_file);
else
- reln->smgr_cached_nblocks[FSM_FORKNUM] = 0;
+ fsm_file->smgr_cached_nblocks = 0;
}
/* Handle requests beyond EOF */
- if (blkno >= reln->smgr_cached_nblocks[FSM_FORKNUM])
+ if (blkno >= fsm_file->smgr_cached_nblocks)
{
if (extend)
fsm_extend(rel, blkno + 1);
@@ -609,7 +609,7 @@ fsm_extend(Relation rel, BlockNumber fsm_nblocks)
{
BlockNumber fsm_nblocks_now;
PGAlignedBlock pg;
- SMgrRelation reln;
+ SMgrFileHandle fsm_file;
PageInit((Page) pg.data, BLCKSZ, 0);
@@ -630,29 +630,28 @@ fsm_extend(Relation rel, BlockNumber fsm_nblocks)
* gets closed. It's safe as long as we only do smgr-level operations
* between here and the last use of the pointer.
*/
- reln = RelationGetSmgr(rel);
+ fsm_file = RelationGetSmgr(rel, FSM_FORKNUM);
/*
* Create the FSM file first if it doesn't exist. If
* smgr_cached_nblocks[FSM_FORKNUM] is positive then it must exist, no
* need for an smgrexists call.
*/
- if ((reln->smgr_cached_nblocks[FSM_FORKNUM] == 0 ||
- reln->smgr_cached_nblocks[FSM_FORKNUM] == InvalidBlockNumber) &&
- !smgrexists(reln, FSM_FORKNUM))
- smgrcreate(reln, FSM_FORKNUM, false);
+ if ((fsm_file->smgr_cached_nblocks == 0 ||
+ fsm_file->smgr_cached_nblocks == InvalidBlockNumber) &&
+ !smgrexists(fsm_file))
+ smgrcreate(fsm_file, false);
/* Invalidate cache so that smgrnblocks() asks the kernel. */
- reln->smgr_cached_nblocks[FSM_FORKNUM] = InvalidBlockNumber;
- fsm_nblocks_now = smgrnblocks(reln, FSM_FORKNUM);
+ fsm_file->smgr_cached_nblocks = InvalidBlockNumber;
+ fsm_nblocks_now = smgrnblocks(fsm_file);
/* Extend as needed. */
while (fsm_nblocks_now < fsm_nblocks)
{
PageSetChecksumInplace((Page) pg.data, fsm_nblocks_now);
- smgrextend(reln, FSM_FORKNUM, fsm_nblocks_now,
- pg.data, false);
+ smgrextend(fsm_file, fsm_nblocks_now, pg.data, false);
fsm_nblocks_now++;
}
diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c
index 352958e1feb..eea7ce944c3 100644
--- a/src/backend/storage/smgr/md.c
+++ b/src/backend/storage/smgr/md.c
@@ -66,10 +66,10 @@
* out to an unlinked old copy of a segment file that will eventually
* disappear.
*
- * File descriptors are stored in the per-fork md_seg_fds arrays inside
- * SMgrRelation. The length of these arrays is stored in md_num_open_segs.
- * Note that a fork's md_num_open_segs having a specific value does not
- * necessarily mean the relation doesn't have additional segments; we may
+ * File descriptors are stored in the md_seg_fds array inside
+ * SMgrFileData. The length of the array is stored in md_num_open_segs.
+ * Note that md_num_open_segs having a specific value does not
+ * necessarily mean the file doesn't have additional segments; we may
* just not have opened the next segment yet. (We could not have "all
* segments are in the array" as an invariant anyway, since another backend
* could extend the relation while we aren't looking.) We do not have
@@ -121,26 +121,18 @@ static MemoryContext MdCxt; /* context for all MdfdVec objects */
/* local routines */
-static void mdunlinkfork(RelFileLocatorBackend rlocator, ForkNumber forknum,
- bool isRedo);
-static MdfdVec *mdopenfork(SMgrRelation reln, ForkNumber forknum, int behavior);
-static void register_dirty_segment(SMgrRelation reln, ForkNumber forknum,
- MdfdVec *seg);
-static void register_unlink_segment(RelFileLocatorBackend rlocator, ForkNumber forknum,
- BlockNumber segno);
-static void register_forget_request(RelFileLocatorBackend rlocator, ForkNumber forknum,
- BlockNumber segno);
-static void _fdvec_resize(SMgrRelation reln,
- ForkNumber forknum,
- int nseg);
-static char *_mdfd_segpath(SMgrRelation reln, ForkNumber forknum,
- BlockNumber segno);
-static MdfdVec *_mdfd_openseg(SMgrRelation reln, ForkNumber forknum,
+static void mdunlinkfork(SMgrFileLocator slocator, bool isRedo);
+static MdfdVec *mdopenfork(SMgrFileHandle sfile, int behavior);
+static void register_dirty_segment(SMgrFileHandle sfile, MdfdVec *seg);
+static void register_unlink_segment(SMgrFileLocator slocator, BlockNumber segno);
+static void register_forget_request(SMgrFileLocator slocator, BlockNumber segno);
+static void _fdvec_resize(SMgrFileHandle sfile, int nseg);
+static char *_mdfd_segpath(SMgrFileHandle sfile, BlockNumber segno);
+static MdfdVec *_mdfd_openseg(SMgrFileHandle sfile,
BlockNumber segno, int oflags);
-static MdfdVec *_mdfd_getseg(SMgrRelation reln, ForkNumber forknum,
+static MdfdVec *_mdfd_getseg(SMgrFileHandle sfile,
BlockNumber blkno, bool skipFsync, int behavior);
-static BlockNumber _mdnblocks(SMgrRelation reln, ForkNumber forknum,
- MdfdVec *seg);
+static BlockNumber _mdnblocks(SMgrFileHandle sfile, MdfdVec *seg);
/*
@@ -160,7 +152,7 @@ mdinit(void)
* Note: this will return true for lingering files, with pending deletions
*/
bool
-mdexists(SMgrRelation reln, ForkNumber forknum)
+mdexists(SMgrFileHandle sfile)
{
/*
* Close it first, to ensure that we notice if the fork has been unlinked
@@ -168,9 +160,9 @@ mdexists(SMgrRelation reln, ForkNumber forknum)
* which already closes relations when dropping them.
*/
if (!InRecovery)
- mdclose(reln, forknum);
+ mdclose(sfile);
- return (mdopenfork(reln, forknum, EXTENSION_RETURN_NULL) != NULL);
+ return (mdopenfork(sfile, EXTENSION_RETURN_NULL) != NULL);
}
/*
@@ -179,16 +171,16 @@ mdexists(SMgrRelation reln, ForkNumber forknum)
* If isRedo is true, it's okay for the relation to exist already.
*/
void
-mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo)
+mdcreate(SMgrFileHandle sfile, bool isRedo)
{
MdfdVec *mdfd;
char *path;
File fd;
- if (isRedo && reln->md_num_open_segs[forknum] > 0)
+ if (isRedo && sfile->md_num_open_segs > 0)
return; /* created and opened already... */
- Assert(reln->md_num_open_segs[forknum] == 0);
+ Assert(sfile->md_num_open_segs == 0);
/*
* We may be using the target table space for the first time in this
@@ -199,11 +191,11 @@ mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo)
* should be here and not in commands/tablespace.c? But that would imply
* importing a lot of stuff that smgr.c oughtn't know, either.
*/
- TablespaceCreateDbspace(reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
+ TablespaceCreateDbspace(sfile->smgr_locator.locator.spcOid,
+ sfile->smgr_locator.locator.dbOid,
isRedo);
- path = relpath(reln->smgr_rlocator, forknum);
+ path = smgrfilepath(sfile->smgr_locator);
fd = PathNameOpenFile(path, O_RDWR | O_CREAT | O_EXCL | PG_BINARY);
@@ -225,8 +217,8 @@ mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo)
pfree(path);
- _fdvec_resize(reln, forknum, 1);
- mdfd = &reln->md_seg_fds[forknum][0];
+ _fdvec_resize(sfile, 1);
+ mdfd = &sfile->md_seg_fds[0];
mdfd->mdfd_vfd = fd;
mdfd->mdfd_segno = 0;
}
@@ -293,16 +285,9 @@ mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo)
* we are usually not in a transaction anymore when this is called.
*/
void
-mdunlink(RelFileLocatorBackend rlocator, ForkNumber forknum, bool isRedo)
+mdunlink(SMgrFileLocator slocator, bool isRedo)
{
- /* Now do the per-fork work */
- if (forknum == InvalidForkNumber)
- {
- for (forknum = 0; forknum <= MAX_FORKNUM; forknum++)
- mdunlinkfork(rlocator, forknum, isRedo);
- }
- else
- mdunlinkfork(rlocator, forknum, isRedo);
+ mdunlinkfork(slocator, isRedo);
}
/*
@@ -330,29 +315,29 @@ do_truncate(const char *path)
}
static void
-mdunlinkfork(RelFileLocatorBackend rlocator, ForkNumber forknum, bool isRedo)
+mdunlinkfork(SMgrFileLocator slocator, bool isRedo)
{
char *path;
int ret;
int save_errno;
- path = relpath(rlocator, forknum);
+ path = smgrfilepath(slocator);
/*
* Truncate and then unlink the first segment, or just register a request
* to unlink it later, as described in the comments for mdunlink().
*/
- if (isRedo || IsBinaryUpgrade || forknum != MAIN_FORKNUM ||
- RelFileLocatorBackendIsTemp(rlocator))
+ if (isRedo || IsBinaryUpgrade || slocator.forknum != MAIN_FORKNUM ||
+ SMgrFileLocatorIsTemp(slocator))
{
- if (!RelFileLocatorBackendIsTemp(rlocator))
+ if (!SMgrFileLocatorIsTemp(slocator))
{
/* Prevent other backends' fds from holding on to the disk space */
ret = do_truncate(path);
/* Forget any pending sync requests for the first segment */
save_errno = errno;
- register_forget_request(rlocator, forknum, 0 /* first seg */ );
+ register_forget_request(slocator, 0 /* first seg */ );
errno = save_errno;
}
else
@@ -379,7 +364,7 @@ mdunlinkfork(RelFileLocatorBackend rlocator, ForkNumber forknum, bool isRedo)
/* Register request to unlink first segment later */
save_errno = errno;
- register_unlink_segment(rlocator, forknum, 0 /* first seg */ );
+ register_unlink_segment(slocator, 0 /* first seg */ );
errno = save_errno;
}
@@ -404,7 +389,7 @@ mdunlinkfork(RelFileLocatorBackend rlocator, ForkNumber forknum, bool isRedo)
{
sprintf(segpath, "%s.%u", path, segno);
- if (!RelFileLocatorBackendIsTemp(rlocator))
+ if (!SMgrFileLocatorIsTemp(slocator))
{
/*
* Prevent other backends' fds from holding on to the disk
@@ -417,7 +402,7 @@ mdunlinkfork(RelFileLocatorBackend rlocator, ForkNumber forknum, bool isRedo)
* Forget any pending sync requests for this segment before we
* try to unlink.
*/
- register_forget_request(rlocator, forknum, segno);
+ register_forget_request(slocator, segno);
}
if (unlink(segpath) < 0)
@@ -446,7 +431,7 @@ mdunlinkfork(RelFileLocatorBackend rlocator, ForkNumber forknum, bool isRedo)
* causes intervening file space to become filled with zeroes.
*/
void
-mdextend(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
+mdextend(SMgrFileHandle sfile, BlockNumber blocknum,
const void *buffer, bool skipFsync)
{
off_t seekpos;
@@ -455,7 +440,7 @@ mdextend(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
/* This assert is too expensive to have on normally ... */
#ifdef CHECK_WRITE_VS_EXTEND
- Assert(blocknum >= mdnblocks(reln, forknum));
+ Assert(blocknum >= mdnblocks(sfile));
#endif
/*
@@ -468,10 +453,10 @@ mdextend(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
ereport(ERROR,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
errmsg("cannot extend file \"%s\" beyond %u blocks",
- relpath(reln->smgr_rlocator, forknum),
+ smgrfilepath(sfile->smgr_locator),
InvalidBlockNumber)));
- v = _mdfd_getseg(reln, forknum, blocknum, skipFsync, EXTENSION_CREATE);
+ v = _mdfd_getseg(sfile, blocknum, skipFsync, EXTENSION_CREATE);
seekpos = (off_t) BLCKSZ * (blocknum % ((BlockNumber) RELSEG_SIZE));
@@ -494,10 +479,10 @@ mdextend(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
errhint("Check free disk space.")));
}
- if (!skipFsync && !SmgrIsTemp(reln))
- register_dirty_segment(reln, forknum, v);
+ if (!skipFsync && !SmgrIsTemp(sfile))
+ register_dirty_segment(sfile, v);
- Assert(_mdnblocks(reln, forknum, v) <= ((BlockNumber) RELSEG_SIZE));
+ Assert(_mdnblocks(sfile, v) <= ((BlockNumber) RELSEG_SIZE));
}
/*
@@ -511,17 +496,17 @@ mdextend(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
* invent one out of whole cloth.
*/
static MdfdVec *
-mdopenfork(SMgrRelation reln, ForkNumber forknum, int behavior)
+mdopenfork(SMgrFileHandle sfile, int behavior)
{
MdfdVec *mdfd;
char *path;
File fd;
/* No work if already open */
- if (reln->md_num_open_segs[forknum] > 0)
- return &reln->md_seg_fds[forknum][0];
+ if (sfile->md_num_open_segs > 0)
+ return &sfile->md_seg_fds[0];
- path = relpath(reln->smgr_rlocator, forknum);
+ path = smgrfilepath(sfile->smgr_locator);
fd = PathNameOpenFile(path, O_RDWR | PG_BINARY);
@@ -540,12 +525,12 @@ mdopenfork(SMgrRelation reln, ForkNumber forknum, int behavior)
pfree(path);
- _fdvec_resize(reln, forknum, 1);
- mdfd = &reln->md_seg_fds[forknum][0];
+ _fdvec_resize(sfile, 1);
+ mdfd = &sfile->md_seg_fds[0];
mdfd->mdfd_vfd = fd;
mdfd->mdfd_segno = 0;
- Assert(_mdnblocks(reln, forknum, mdfd) <= ((BlockNumber) RELSEG_SIZE));
+ Assert(_mdnblocks(sfile, mdfd) <= ((BlockNumber) RELSEG_SIZE));
return mdfd;
}
@@ -554,20 +539,19 @@ mdopenfork(SMgrRelation reln, ForkNumber forknum, int behavior)
* mdopen() -- Initialize newly-opened relation.
*/
void
-mdopen(SMgrRelation reln)
+mdopen(SMgrFileHandle sfile)
{
/* mark it not open */
- for (int forknum = 0; forknum <= MAX_FORKNUM; forknum++)
- reln->md_num_open_segs[forknum] = 0;
+ sfile->md_num_open_segs = 0;
}
/*
* mdclose() -- Close the specified relation, if it isn't closed already.
*/
void
-mdclose(SMgrRelation reln, ForkNumber forknum)
+mdclose(SMgrFileHandle sfile)
{
- int nopensegs = reln->md_num_open_segs[forknum];
+ int nopensegs = sfile->md_num_open_segs;
/* No work if already closed */
if (nopensegs == 0)
@@ -576,10 +560,10 @@ mdclose(SMgrRelation reln, ForkNumber forknum)
/* close segments starting from the end */
while (nopensegs > 0)
{
- MdfdVec *v = &reln->md_seg_fds[forknum][nopensegs - 1];
+ MdfdVec *v = &sfile->md_seg_fds[nopensegs - 1];
FileClose(v->mdfd_vfd);
- _fdvec_resize(reln, forknum, nopensegs - 1);
+ _fdvec_resize(sfile, nopensegs - 1);
nopensegs--;
}
}
@@ -588,13 +572,13 @@ mdclose(SMgrRelation reln, ForkNumber forknum)
* mdprefetch() -- Initiate asynchronous read of the specified block of a relation
*/
bool
-mdprefetch(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum)
+mdprefetch(SMgrFileHandle sfile, BlockNumber blocknum)
{
#ifdef USE_PREFETCH
off_t seekpos;
MdfdVec *v;
- v = _mdfd_getseg(reln, forknum, blocknum, false,
+ v = _mdfd_getseg(sfile, blocknum, false,
InRecovery ? EXTENSION_RETURN_NULL : EXTENSION_FAIL);
if (v == NULL)
return false;
@@ -616,8 +600,7 @@ mdprefetch(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum)
* considerably more efficient than doing so individually.
*/
void
-mdwriteback(SMgrRelation reln, ForkNumber forknum,
- BlockNumber blocknum, BlockNumber nblocks)
+mdwriteback(SMgrFileHandle sfile, BlockNumber blocknum, BlockNumber nblocks)
{
/*
* Issue flush requests in as few requests as possible; have to split at
@@ -631,7 +614,7 @@ mdwriteback(SMgrRelation reln, ForkNumber forknum,
int segnum_start,
segnum_end;
- v = _mdfd_getseg(reln, forknum, blocknum, true /* not used */ ,
+ v = _mdfd_getseg(sfile, blocknum, true /* not used */ ,
EXTENSION_DONT_OPEN);
/*
@@ -668,20 +651,21 @@ mdwriteback(SMgrRelation reln, ForkNumber forknum,
* mdread() -- Read the specified block from a relation.
*/
void
-mdread(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
+mdread(SMgrFileHandle sfile, BlockNumber blocknum,
void *buffer)
{
off_t seekpos;
int nbytes;
MdfdVec *v;
- TRACE_POSTGRESQL_SMGR_MD_READ_START(forknum, blocknum,
- reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
- reln->smgr_rlocator.locator.relNumber,
- reln->smgr_rlocator.backend);
+ TRACE_POSTGRESQL_SMGR_MD_READ_START(sfile->smgr_forknum,
+ blocknum,
+ sfile->smgr_rlocator.locator.spcOid,
+ sfile->smgr_rlocator.locator.dbOid,
+ sfile->smgr_rlocator.locator.relNumber,
+ sfile->smgr_rlocator.backend);
- v = _mdfd_getseg(reln, forknum, blocknum, false,
+ v = _mdfd_getseg(sfile, blocknum, false,
EXTENSION_FAIL | EXTENSION_CREATE_RECOVERY);
seekpos = (off_t) BLCKSZ * (blocknum % ((BlockNumber) RELSEG_SIZE));
@@ -690,11 +674,12 @@ mdread(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
nbytes = FileRead(v->mdfd_vfd, buffer, BLCKSZ, seekpos, WAIT_EVENT_DATA_FILE_READ);
- TRACE_POSTGRESQL_SMGR_MD_READ_DONE(forknum, blocknum,
- reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
- reln->smgr_rlocator.locator.relNumber,
- reln->smgr_rlocator.backend,
+ TRACE_POSTGRESQL_SMGR_MD_READ_DONE(sfile->smgr_forknum,
+ blocknum,
+ sfile->smgr_rlocator.locator.spcOid,
+ sfile->smgr_rlocator.locator.dbOid,
+ sfile->smgr_rlocator.locator.relNumber,
+ sfile->smgr_rlocator.backend,
nbytes,
BLCKSZ);
@@ -733,7 +718,7 @@ mdread(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
* use mdextend().
*/
void
-mdwrite(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
+mdwrite(SMgrFileHandle sfile, BlockNumber blocknum,
const void *buffer, bool skipFsync)
{
off_t seekpos;
@@ -742,16 +727,17 @@ mdwrite(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
/* This assert is too expensive to have on normally ... */
#ifdef CHECK_WRITE_VS_EXTEND
- Assert(blocknum < mdnblocks(reln, forknum));
+ Assert(blocknum < mdnblocks(sfile));
#endif
- TRACE_POSTGRESQL_SMGR_MD_WRITE_START(forknum, blocknum,
- reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
- reln->smgr_rlocator.locator.relNumber,
- reln->smgr_rlocator.backend);
+ TRACE_POSTGRESQL_SMGR_MD_WRITE_START(sfile->smgr_forknum,
+ blocknum,
+ sfile->smgr_rlocator.locator.spcOid,
+ sfile->smgr_rlocator.locator.dbOid,
+ sfile->smgr_rlocator.locator.relNumber,
+ sfile->smgr_rlocator.backend);
- v = _mdfd_getseg(reln, forknum, blocknum, skipFsync,
+ v = _mdfd_getseg(sfile, blocknum, skipFsync,
EXTENSION_FAIL | EXTENSION_CREATE_RECOVERY);
seekpos = (off_t) BLCKSZ * (blocknum % ((BlockNumber) RELSEG_SIZE));
@@ -760,11 +746,12 @@ mdwrite(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
nbytes = FileWrite(v->mdfd_vfd, buffer, BLCKSZ, seekpos, WAIT_EVENT_DATA_FILE_WRITE);
- TRACE_POSTGRESQL_SMGR_MD_WRITE_DONE(forknum, blocknum,
- reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
- reln->smgr_rlocator.locator.relNumber,
- reln->smgr_rlocator.backend,
+ TRACE_POSTGRESQL_SMGR_MD_WRITE_DONE(sfile->smgr_forknum,
+ blocknum,
+ sfile->smgr_rlocator.locator.spcOid,
+ sfile->smgr_rlocator.locator.dbOid,
+ sfile->smgr_rlocator.locator.relNumber,
+ sfile->smgr_rlocator.backend,
nbytes,
BLCKSZ);
@@ -785,8 +772,8 @@ mdwrite(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
errhint("Check free disk space.")));
}
- if (!skipFsync && !SmgrIsTemp(reln))
- register_dirty_segment(reln, forknum, v);
+ if (!skipFsync && !SmgrIsTemp(sfile))
+ register_dirty_segment(sfile, v);
}
/*
@@ -798,16 +785,16 @@ mdwrite(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
* are present in the array.
*/
BlockNumber
-mdnblocks(SMgrRelation reln, ForkNumber forknum)
+mdnblocks(SMgrFileHandle sfile)
{
MdfdVec *v;
BlockNumber nblocks;
BlockNumber segno;
- mdopenfork(reln, forknum, EXTENSION_FAIL);
+ mdopenfork(sfile, EXTENSION_FAIL);
/* mdopen has opened the first segment */
- Assert(reln->md_num_open_segs[forknum] > 0);
+ Assert(sfile->md_num_open_segs > 0);
/*
* Start from the last open segments, to avoid redundant seeks. We have
@@ -822,12 +809,12 @@ mdnblocks(SMgrRelation reln, ForkNumber forknum)
* that's OK because the checkpointer never needs to compute relation
* size.)
*/
- segno = reln->md_num_open_segs[forknum] - 1;
- v = &reln->md_seg_fds[forknum][segno];
+ segno = sfile->md_num_open_segs - 1;
+ v = &sfile->md_seg_fds[segno];
for (;;)
{
- nblocks = _mdnblocks(reln, forknum, v);
+ nblocks = _mdnblocks(sfile, v);
if (nblocks > ((BlockNumber) RELSEG_SIZE))
elog(FATAL, "segment too big");
if (nblocks < ((BlockNumber) RELSEG_SIZE))
@@ -845,7 +832,7 @@ mdnblocks(SMgrRelation reln, ForkNumber forknum)
* undermines _mdfd_getseg's attempts to notice and report an error
* upon access to a missing segment.
*/
- v = _mdfd_openseg(reln, forknum, segno, 0);
+ v = _mdfd_openseg(sfile, segno, 0);
if (v == NULL)
return segno * ((BlockNumber) RELSEG_SIZE);
}
@@ -855,7 +842,7 @@ mdnblocks(SMgrRelation reln, ForkNumber forknum)
* mdtruncate() -- Truncate relation to specified number of blocks.
*/
void
-mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks)
+mdtruncate(SMgrFileHandle sfile, BlockNumber nblocks)
{
BlockNumber curnblk;
BlockNumber priorblocks;
@@ -865,7 +852,7 @@ mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks)
* NOTE: mdnblocks makes sure we have opened all active segments, so that
* truncation loop will get them all!
*/
- curnblk = mdnblocks(reln, forknum);
+ curnblk = mdnblocks(sfile);
if (nblocks > curnblk)
{
/* Bogus request ... but no complaint if InRecovery */
@@ -873,7 +860,7 @@ mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks)
return;
ereport(ERROR,
(errmsg("could not truncate file \"%s\" to %u blocks: it's only %u blocks now",
- relpath(reln->smgr_rlocator, forknum),
+ smgrfilepath(sfile->smgr_locator),
nblocks, curnblk)));
}
if (nblocks == curnblk)
@@ -883,14 +870,14 @@ mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks)
* Truncate segments, starting at the last one. Starting at the end makes
* managing the memory for the fd array easier, should there be errors.
*/
- curopensegs = reln->md_num_open_segs[forknum];
+ curopensegs = sfile->md_num_open_segs;
while (curopensegs > 0)
{
MdfdVec *v;
priorblocks = (curopensegs - 1) * RELSEG_SIZE;
- v = &reln->md_seg_fds[forknum][curopensegs - 1];
+ v = &sfile->md_seg_fds[curopensegs - 1];
if (priorblocks > nblocks)
{
@@ -904,14 +891,14 @@ mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks)
errmsg("could not truncate file \"%s\": %m",
FilePathName(v->mdfd_vfd))));
- if (!SmgrIsTemp(reln))
- register_dirty_segment(reln, forknum, v);
+ if (!SmgrIsTemp(sfile))
+ register_dirty_segment(sfile, v);
/* we never drop the 1st segment */
- Assert(v != &reln->md_seg_fds[forknum][0]);
+ Assert(v != &sfile->md_seg_fds[0]);
FileClose(v->mdfd_vfd);
- _fdvec_resize(reln, forknum, curopensegs - 1);
+ _fdvec_resize(sfile, curopensegs - 1);
}
else if (priorblocks + ((BlockNumber) RELSEG_SIZE) > nblocks)
{
@@ -930,8 +917,8 @@ mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks)
errmsg("could not truncate file \"%s\" to %u blocks: %m",
FilePathName(v->mdfd_vfd),
nblocks)));
- if (!SmgrIsTemp(reln))
- register_dirty_segment(reln, forknum, v);
+ if (!SmgrIsTemp(sfile))
+ register_dirty_segment(sfile, v);
}
else
{
@@ -957,7 +944,7 @@ mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks)
* segment may survive recovery, reintroducing unwanted data into the table.
*/
void
-mdimmedsync(SMgrRelation reln, ForkNumber forknum)
+mdimmedsync(SMgrFileHandle sfile)
{
int segno;
int min_inactive_seg;
@@ -966,9 +953,9 @@ mdimmedsync(SMgrRelation reln, ForkNumber forknum)
* NOTE: mdnblocks makes sure we have opened all active segments, so that
* fsync loop will get them all!
*/
- mdnblocks(reln, forknum);
+ mdnblocks(sfile);
- min_inactive_seg = segno = reln->md_num_open_segs[forknum];
+ min_inactive_seg = segno = sfile->md_num_open_segs;
/*
* Temporarily open inactive segments, then close them after sync. There
@@ -976,12 +963,12 @@ mdimmedsync(SMgrRelation reln, ForkNumber forknum)
* is harmless. We don't bother to clean them up and take a risk of
* further trouble. The next mdclose() will soon close them.
*/
- while (_mdfd_openseg(reln, forknum, segno, 0) != NULL)
+ while (_mdfd_openseg(sfile, segno, 0) != NULL)
segno++;
while (segno > 0)
{
- MdfdVec *v = &reln->md_seg_fds[forknum][segno - 1];
+ MdfdVec *v = &sfile->md_seg_fds[segno - 1];
/*
* fsyncs done through mdimmedsync() should be tracked in a separate
@@ -1002,7 +989,7 @@ mdimmedsync(SMgrRelation reln, ForkNumber forknum)
if (segno > min_inactive_seg)
{
FileClose(v->mdfd_vfd);
- _fdvec_resize(reln, forknum, segno - 1);
+ _fdvec_resize(sfile, segno - 1);
}
segno--;
@@ -1019,14 +1006,14 @@ mdimmedsync(SMgrRelation reln, ForkNumber forknum)
* enough to be a performance problem).
*/
static void
-register_dirty_segment(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg)
+register_dirty_segment(SMgrFileHandle sfile, MdfdVec *seg)
{
FileTag tag;
- INIT_MD_FILETAG(tag, reln->smgr_rlocator.locator, forknum, seg->mdfd_segno);
+ INIT_MD_FILETAG(tag, sfile->smgr_locator.locator, sfile->smgr_locator.forknum, seg->mdfd_segno);
/* Temp relations should never be fsync'd */
- Assert(!SmgrIsTemp(reln));
+ Assert(!SmgrIsTemp(sfile));
if (!RegisterSyncRequest(&tag, SYNC_REQUEST, false /* retryOnError */ ))
{
@@ -1058,15 +1045,14 @@ register_dirty_segment(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg)
* register_unlink_segment() -- Schedule a file to be deleted after next checkpoint
*/
static void
-register_unlink_segment(RelFileLocatorBackend rlocator, ForkNumber forknum,
- BlockNumber segno)
+register_unlink_segment(SMgrFileLocator slocator, BlockNumber segno)
{
FileTag tag;
- INIT_MD_FILETAG(tag, rlocator.locator, forknum, segno);
+ INIT_MD_FILETAG(tag, slocator.locator, slocator.forknum, segno);
/* Should never be used with temp relations */
- Assert(!RelFileLocatorBackendIsTemp(rlocator));
+ Assert(!SMgrFileLocatorIsTemp(slocator));
RegisterSyncRequest(&tag, SYNC_UNLINK_REQUEST, true /* retryOnError */ );
}
@@ -1075,12 +1061,11 @@ register_unlink_segment(RelFileLocatorBackend rlocator, ForkNumber forknum,
* register_forget_request() -- forget any fsyncs for a relation fork's segment
*/
static void
-register_forget_request(RelFileLocatorBackend rlocator, ForkNumber forknum,
- BlockNumber segno)
+register_forget_request(SMgrFileLocator slocator, BlockNumber segno)
{
FileTag tag;
- INIT_MD_FILETAG(tag, rlocator.locator, forknum, segno);
+ INIT_MD_FILETAG(tag, slocator.locator, slocator.forknum, segno);
RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ );
}
@@ -1103,57 +1088,25 @@ ForgetDatabaseSyncRequests(Oid dbid)
RegisterSyncRequest(&tag, SYNC_FILTER_REQUEST, true /* retryOnError */ );
}
-/*
- * DropRelationFiles -- drop files of all given relations
- */
-void
-DropRelationFiles(RelFileLocator *delrels, int ndelrels, bool isRedo)
-{
- SMgrRelation *srels;
- int i;
-
- srels = palloc(sizeof(SMgrRelation) * ndelrels);
- for (i = 0; i < ndelrels; i++)
- {
- SMgrRelation srel = smgropen(delrels[i], InvalidBackendId);
-
- if (isRedo)
- {
- ForkNumber fork;
-
- for (fork = 0; fork <= MAX_FORKNUM; fork++)
- XLogDropRelation(delrels[i], fork);
- }
- srels[i] = srel;
- }
-
- smgrdounlinkall(srels, ndelrels, isRedo);
-
- for (i = 0; i < ndelrels; i++)
- smgrclose(srels[i]);
- pfree(srels);
-}
-
/*
* _fdvec_resize() -- Resize the fork's open segments array
*/
static void
-_fdvec_resize(SMgrRelation reln,
- ForkNumber forknum,
+_fdvec_resize(SMgrFileHandle sfile,
int nseg)
{
if (nseg == 0)
{
- if (reln->md_num_open_segs[forknum] > 0)
+ if (sfile->md_num_open_segs > 0)
{
- pfree(reln->md_seg_fds[forknum]);
- reln->md_seg_fds[forknum] = NULL;
+ pfree(sfile->md_seg_fds);
+ sfile->md_seg_fds = NULL;
}
}
- else if (reln->md_num_open_segs[forknum] == 0)
+ else if (sfile->md_num_open_segs == 0)
{
- reln->md_seg_fds[forknum] =
+ sfile->md_seg_fds =
MemoryContextAlloc(MdCxt, sizeof(MdfdVec) * nseg);
}
else
@@ -1164,12 +1117,12 @@ _fdvec_resize(SMgrRelation reln,
* FileClose(), and the memory context internally will sometimes avoid
* doing an actual reallocation.
*/
- reln->md_seg_fds[forknum] =
- repalloc(reln->md_seg_fds[forknum],
+ sfile->md_seg_fds =
+ repalloc(sfile->md_seg_fds,
sizeof(MdfdVec) * nseg);
}
- reln->md_num_open_segs[forknum] = nseg;
+ sfile->md_num_open_segs = nseg;
}
/*
@@ -1177,12 +1130,12 @@ _fdvec_resize(SMgrRelation reln,
* returned string is palloc'd.
*/
static char *
-_mdfd_segpath(SMgrRelation reln, ForkNumber forknum, BlockNumber segno)
+_mdfd_segpath(SMgrFileHandle sfile, BlockNumber segno)
{
char *path,
*fullpath;
- path = relpath(reln->smgr_rlocator, forknum);
+ path = smgrfilepath(sfile->smgr_locator);
if (segno > 0)
{
@@ -1200,14 +1153,14 @@ _mdfd_segpath(SMgrRelation reln, ForkNumber forknum, BlockNumber segno)
* and make a MdfdVec object for it. Returns NULL on failure.
*/
static MdfdVec *
-_mdfd_openseg(SMgrRelation reln, ForkNumber forknum, BlockNumber segno,
+_mdfd_openseg(SMgrFileHandle sfile, BlockNumber segno,
int oflags)
{
MdfdVec *v;
File fd;
char *fullpath;
- fullpath = _mdfd_segpath(reln, forknum, segno);
+ fullpath = _mdfd_segpath(sfile, segno);
/* open the file */
fd = PathNameOpenFile(fullpath, O_RDWR | PG_BINARY | oflags);
@@ -1221,16 +1174,16 @@ _mdfd_openseg(SMgrRelation reln, ForkNumber forknum, BlockNumber segno,
* Segments are always opened in order from lowest to highest, so we must
* be adding a new one at the end.
*/
- Assert(segno == reln->md_num_open_segs[forknum]);
+ Assert(segno == sfile->md_num_open_segs);
- _fdvec_resize(reln, forknum, segno + 1);
+ _fdvec_resize(sfile, segno + 1);
/* fill the entry */
- v = &reln->md_seg_fds[forknum][segno];
+ v = &sfile->md_seg_fds[segno];
v->mdfd_vfd = fd;
v->mdfd_segno = segno;
- Assert(_mdnblocks(reln, forknum, v) <= ((BlockNumber) RELSEG_SIZE));
+ Assert(_mdnblocks(sfile, v) <= ((BlockNumber) RELSEG_SIZE));
/* all done */
return v;
@@ -1245,7 +1198,7 @@ _mdfd_openseg(SMgrRelation reln, ForkNumber forknum, BlockNumber segno,
* EXTENSION_CREATE case.
*/
static MdfdVec *
-_mdfd_getseg(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
+_mdfd_getseg(SMgrFileHandle sfile, BlockNumber blkno,
bool skipFsync, int behavior)
{
MdfdVec *v;
@@ -1260,9 +1213,9 @@ _mdfd_getseg(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
targetseg = blkno / ((BlockNumber) RELSEG_SIZE);
/* if an existing and opened segment, we're done */
- if (targetseg < reln->md_num_open_segs[forknum])
+ if (targetseg < sfile->md_num_open_segs)
{
- v = &reln->md_seg_fds[forknum][targetseg];
+ v = &sfile->md_seg_fds[targetseg];
return v;
}
@@ -1277,19 +1230,19 @@ _mdfd_getseg(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
* 'behavior'). Start with either the last opened, or the first segment if
* none was opened before.
*/
- if (reln->md_num_open_segs[forknum] > 0)
- v = &reln->md_seg_fds[forknum][reln->md_num_open_segs[forknum] - 1];
+ if (sfile->md_num_open_segs > 0)
+ v = &sfile->md_seg_fds[sfile->md_num_open_segs - 1];
else
{
- v = mdopenfork(reln, forknum, behavior);
+ v = mdopenfork(sfile, behavior);
if (!v)
return NULL; /* if behavior & EXTENSION_RETURN_NULL */
}
- for (nextsegno = reln->md_num_open_segs[forknum];
+ for (nextsegno = sfile->md_num_open_segs;
nextsegno <= targetseg; nextsegno++)
{
- BlockNumber nblocks = _mdnblocks(reln, forknum, v);
+ BlockNumber nblocks = _mdnblocks(sfile, v);
int flags = 0;
Assert(nextsegno == v->mdfd_segno + 1);
@@ -1318,7 +1271,7 @@ _mdfd_getseg(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
{
char *zerobuf = palloc0(BLCKSZ);
- mdextend(reln, forknum,
+ mdextend(sfile,
nextsegno * ((BlockNumber) RELSEG_SIZE) - 1,
zerobuf, skipFsync);
pfree(zerobuf);
@@ -1349,11 +1302,11 @@ _mdfd_getseg(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not open file \"%s\" (target block %u): previous segment is only %u blocks",
- _mdfd_segpath(reln, forknum, nextsegno),
+ _mdfd_segpath(sfile, nextsegno),
blkno, nblocks)));
}
- v = _mdfd_openseg(reln, forknum, nextsegno, flags);
+ v = _mdfd_openseg(sfile, nextsegno, flags);
if (v == NULL)
{
@@ -1363,7 +1316,7 @@ _mdfd_getseg(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not open file \"%s\" (target block %u): %m",
- _mdfd_segpath(reln, forknum, nextsegno),
+ _mdfd_segpath(sfile, nextsegno),
blkno)));
}
}
@@ -1375,7 +1328,7 @@ _mdfd_getseg(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
* Get number of blocks present in a single disk file
*/
static BlockNumber
-_mdnblocks(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg)
+_mdnblocks(SMgrFileHandle sfile, MdfdVec *seg)
{
off_t len;
@@ -1398,16 +1351,16 @@ _mdnblocks(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg)
int
mdsyncfiletag(const FileTag *ftag, char *path)
{
- SMgrRelation reln = smgropen(ftag->rlocator, InvalidBackendId);
+ SMgrFileHandle sfile = smgropen(ftag->rlocator, InvalidBackendId, ftag->forknum);
File file;
bool need_to_close;
int result,
save_errno;
/* See if we already have the file open, or need to open it. */
- if (ftag->segno < reln->md_num_open_segs[ftag->forknum])
+ if (ftag->segno < sfile->md_num_open_segs)
{
- file = reln->md_seg_fds[ftag->forknum][ftag->segno].mdfd_vfd;
+ file = sfile->md_seg_fds[ftag->segno].mdfd_vfd;
strlcpy(path, FilePathName(file), MAXPGPATH);
need_to_close = false;
}
@@ -1415,7 +1368,7 @@ mdsyncfiletag(const FileTag *ftag, char *path)
{
char *p;
- p = _mdfd_segpath(reln, ftag->forknum, ftag->segno);
+ p = _mdfd_segpath(sfile, ftag->segno);
strlcpy(path, p, MAXPGPATH);
pfree(p);
diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c
index dc466e54145..58a2322d018 100644
--- a/src/backend/storage/smgr/smgr.c
+++ b/src/backend/storage/smgr/smgr.c
@@ -41,27 +41,24 @@ typedef struct f_smgr
{
void (*smgr_init) (void); /* may be NULL */
void (*smgr_shutdown) (void); /* may be NULL */
- void (*smgr_open) (SMgrRelation reln);
- void (*smgr_close) (SMgrRelation reln, ForkNumber forknum);
- void (*smgr_create) (SMgrRelation reln, ForkNumber forknum,
- bool isRedo);
- bool (*smgr_exists) (SMgrRelation reln, ForkNumber forknum);
- void (*smgr_unlink) (RelFileLocatorBackend rlocator, ForkNumber forknum,
- bool isRedo);
- void (*smgr_extend) (SMgrRelation reln, ForkNumber forknum,
+ void (*smgr_open) (SMgrFileHandle sfile);
+ void (*smgr_close) (SMgrFileHandle sfile);
+ void (*smgr_create) (SMgrFileHandle sfile, bool isRedo);
+ bool (*smgr_exists) (SMgrFileHandle sfile);
+ void (*smgr_unlink) (SMgrFileLocator slocator, bool isRedo);
+ void (*smgr_extend) (SMgrFileHandle sfile,
BlockNumber blocknum, const void *buffer, bool skipFsync);
- bool (*smgr_prefetch) (SMgrRelation reln, ForkNumber forknum,
+ bool (*smgr_prefetch) (SMgrFileHandle sfile,
BlockNumber blocknum);
- void (*smgr_read) (SMgrRelation reln, ForkNumber forknum,
+ void (*smgr_read) (SMgrFileHandle sfile,
BlockNumber blocknum, void *buffer);
- void (*smgr_write) (SMgrRelation reln, ForkNumber forknum,
+ void (*smgr_write) (SMgrFileHandle sfile,
BlockNumber blocknum, const void *buffer, bool skipFsync);
- void (*smgr_writeback) (SMgrRelation reln, ForkNumber forknum,
+ void (*smgr_writeback) (SMgrFileHandle sfile,
BlockNumber blocknum, BlockNumber nblocks);
- BlockNumber (*smgr_nblocks) (SMgrRelation reln, ForkNumber forknum);
- void (*smgr_truncate) (SMgrRelation reln, ForkNumber forknum,
- BlockNumber nblocks);
- void (*smgr_immedsync) (SMgrRelation reln, ForkNumber forknum);
+ BlockNumber (*smgr_nblocks) (SMgrFileHandle sfile);
+ void (*smgr_truncate) (SMgrFileHandle sfile, BlockNumber nblocks);
+ void (*smgr_immedsync) (SMgrFileHandle sfile);
} f_smgr;
static const f_smgr smgrsw[] = {
@@ -88,12 +85,12 @@ static const f_smgr smgrsw[] = {
static const int NSmgr = lengthof(smgrsw);
/*
- * Each backend has a hashtable that stores all extant SMgrRelation objects.
- * In addition, "unowned" SMgrRelation objects are chained together in a list.
+ * Each backend has a hashtable that stores all extant SMgrFileData objects.
+ * In addition, "unowned" SMgrFile objects are chained together in a list.
*/
-static HTAB *SMgrRelationHash = NULL;
+static HTAB *SMgrFileHash = NULL;
-static dlist_head unowned_relns;
+static dlist_head unowned_sfiles;
/* local function prototypes */
static void smgrshutdown(int code, Datum arg);
@@ -142,50 +139,50 @@ smgrshutdown(int code, Datum arg)
*
* This does not attempt to actually open the underlying file.
*/
-SMgrRelation
-smgropen(RelFileLocator rlocator, BackendId backend)
+SMgrFileHandle
+smgropen(RelFileLocator rlocator, BackendId backend, ForkNumber forkNum)
{
- RelFileLocatorBackend brlocator;
- SMgrRelation reln;
+ SMgrFileLocator slocator;
+ SMgrFileHandle sfile;
bool found;
- if (SMgrRelationHash == NULL)
+ if (SMgrFileHash == NULL)
{
/* First time through: initialize the hash table */
HASHCTL ctl;
- ctl.keysize = sizeof(RelFileLocatorBackend);
- ctl.entrysize = sizeof(SMgrRelationData);
- SMgrRelationHash = hash_create("smgr relation table", 400,
- &ctl, HASH_ELEM | HASH_BLOBS);
- dlist_init(&unowned_relns);
+ ctl.keysize = sizeof(SMgrFileLocator);
+ ctl.entrysize = sizeof(SMgrFileData);
+ SMgrFileHash = hash_create("smgr relation table", 400,
+ &ctl, HASH_ELEM | HASH_BLOBS);
+ dlist_init(&unowned_sfiles);
}
/* Look up or create an entry */
- brlocator.locator = rlocator;
- brlocator.backend = backend;
- reln = (SMgrRelation) hash_search(SMgrRelationHash,
- &brlocator,
- HASH_ENTER, &found);
+ slocator.locator = rlocator;
+ slocator.backend = backend;
+ slocator.forknum = forkNum;
+ sfile = (SMgrFileHandle) hash_search(SMgrFileHash,
+ &slocator,
+ HASH_ENTER, &found);
/* Initialize it if not present before */
if (!found)
{
/* hash_search already filled in the lookup key */
- reln->smgr_owner = NULL;
- reln->smgr_targblock = InvalidBlockNumber;
- for (int i = 0; i <= MAX_FORKNUM; ++i)
- reln->smgr_cached_nblocks[i] = InvalidBlockNumber;
- reln->smgr_which = 0; /* we only have md.c at present */
+ sfile->smgr_owner = NULL;
+ sfile->smgr_targblock = InvalidBlockNumber;
+ sfile->smgr_cached_nblocks = InvalidBlockNumber;
+ sfile->smgr_which = 0; /* we only have md.c at present */
/* implementation-specific initialization */
- smgrsw[reln->smgr_which].smgr_open(reln);
+ smgrsw[sfile->smgr_which].smgr_open(sfile);
/* it has no owner yet */
- dlist_push_tail(&unowned_relns, &reln->node);
+ dlist_push_tail(&unowned_sfiles, &sfile->node);
}
- return reln;
+ return sfile;
}
/*
@@ -195,7 +192,7 @@ smgropen(RelFileLocator rlocator, BackendId backend)
* the only such owners exist in the relcache.
*/
void
-smgrsetowner(SMgrRelation *owner, SMgrRelation reln)
+smgrsetowner(SMgrFileHandle *owner, SMgrFileHandle sfile)
{
/* We don't support "disowning" an SMgrRelation here, use smgrclearowner */
Assert(owner != NULL);
@@ -206,68 +203,66 @@ smgrsetowner(SMgrRelation *owner, SMgrRelation reln)
* depending on the order of processing. It's ok to close the old
* relcache entry early in that case.)
*
- * If there isn't an old owner, then the reln should be in the unowned
+ * If there isn't an old owner, then the sfile should be in the unowned
* list, and we need to remove it.
*/
- if (reln->smgr_owner)
- *(reln->smgr_owner) = NULL;
+ if (sfile->smgr_owner)
+ *(sfile->smgr_owner) = NULL;
else
- dlist_delete(&reln->node);
+ dlist_delete(&sfile->node);
/* Now establish the ownership relationship. */
- reln->smgr_owner = owner;
- *owner = reln;
+ sfile->smgr_owner = owner;
+ *owner = sfile;
}
/*
- * smgrclearowner() -- Remove long-lived reference to an SMgrRelation object
+ * smgrclearowner() -- Remove long-lived reference to an SMgrFileHandle object
* if one exists
*/
void
-smgrclearowner(SMgrRelation *owner, SMgrRelation reln)
+smgrclearowner(SMgrFileHandle *owner, SMgrFileHandle sfile)
{
/* Do nothing if the SMgrRelation object is not owned by the owner */
- if (reln->smgr_owner != owner)
+ if (sfile->smgr_owner != owner)
return;
/* unset the owner's reference */
*owner = NULL;
/* unset our reference to the owner */
- reln->smgr_owner = NULL;
+ sfile->smgr_owner = NULL;
/* add to list of unowned relations */
- dlist_push_tail(&unowned_relns, &reln->node);
+ dlist_push_tail(&unowned_sfiles, &sfile->node);
}
/*
- * smgrexists() -- Does the underlying file for a fork exist?
+ * smgrexists() -- Does the underlying file exist?
*/
bool
-smgrexists(SMgrRelation reln, ForkNumber forknum)
+smgrexists(SMgrFileHandle sfile)
{
- return smgrsw[reln->smgr_which].smgr_exists(reln, forknum);
+ return smgrsw[sfile->smgr_which].smgr_exists(sfile);
}
/*
- * smgrclose() -- Close and delete an SMgrRelation object.
+ * smgrclose() -- Close and delete an SMgrFile object.
*/
void
-smgrclose(SMgrRelation reln)
+smgrclose(SMgrFileHandle sfile)
{
- SMgrRelation *owner;
- ForkNumber forknum;
+ SMgrFileHandle *owner;
- for (forknum = 0; forknum <= MAX_FORKNUM; forknum++)
- smgrsw[reln->smgr_which].smgr_close(reln, forknum);
+ smgrsw[sfile->smgr_which].smgr_close(sfile);
- owner = reln->smgr_owner;
+ owner = sfile->smgr_owner;
if (!owner)
- dlist_delete(&reln->node);
+ dlist_delete(&sfile->node);
- if (hash_search(SMgrRelationHash,
- &(reln->smgr_rlocator),
+ if (hash_search(SMgrFileHash,
+ &sfile->smgr_locator,
HASH_REMOVE, NULL) == NULL)
elog(ERROR, "SMgrRelation hashtable corrupted");
@@ -284,14 +279,11 @@ smgrclose(SMgrRelation reln)
*
* The object remains valid.
*/
-void
-smgrrelease(SMgrRelation reln)
+static void
+smgrrelease(SMgrFileHandle sfile)
{
- for (ForkNumber forknum = 0; forknum <= MAX_FORKNUM; forknum++)
- {
- smgrsw[reln->smgr_which].smgr_close(reln, forknum);
- reln->smgr_cached_nblocks[forknum] = InvalidBlockNumber;
- }
+ smgrsw[sfile->smgr_which].smgr_close(sfile);
+ sfile->smgr_cached_nblocks = InvalidBlockNumber;
}
/*
@@ -303,16 +295,16 @@ void
smgrreleaseall(void)
{
HASH_SEQ_STATUS status;
- SMgrRelation reln;
+ SMgrFileHandle sfile;
/* Nothing to do if hashtable not set up */
- if (SMgrRelationHash == NULL)
+ if (SMgrFileHash == NULL)
return;
- hash_seq_init(&status, SMgrRelationHash);
+ hash_seq_init(&status, SMgrFileHash);
- while ((reln = (SMgrRelation) hash_seq_search(&status)) != NULL)
- smgrrelease(reln);
+ while ((sfile = (SMgrFileHandle) hash_seq_search(&status)) != NULL)
+ smgrrelease(sfile);
}
/*
@@ -322,16 +314,16 @@ void
smgrcloseall(void)
{
HASH_SEQ_STATUS status;
- SMgrRelation reln;
+ SMgrFileHandle sfile;
/* Nothing to do if hashtable not set up */
- if (SMgrRelationHash == NULL)
+ if (SMgrFileHash == NULL)
return;
- hash_seq_init(&status, SMgrRelationHash);
+ hash_seq_init(&status, SMgrFileHash);
- while ((reln = (SMgrRelation) hash_seq_search(&status)) != NULL)
- smgrclose(reln);
+ while ((sfile = (SMgrFileHandle) hash_seq_search(&status)) != NULL)
+ smgrclose(sfile);
}
/*
@@ -343,111 +335,64 @@ smgrcloseall(void)
* such entry exists already.
*/
void
-smgrcloserellocator(RelFileLocatorBackend rlocator)
+smgrcloserellocator(RelFileLocator rlocator, BackendId backend)
{
- SMgrRelation reln;
-
/* Nothing to do if hashtable not set up */
- if (SMgrRelationHash == NULL)
+ if (SMgrFileHash == NULL)
return;
- reln = (SMgrRelation) hash_search(SMgrRelationHash,
- &rlocator,
- HASH_FIND, NULL);
- if (reln != NULL)
- smgrclose(reln);
+ for (int forknum = 0; forknum <= MAX_FORKNUM; forknum++)
+ {
+ SMgrFileLocator slocator = {rlocator, backend, forknum};
+ SMgrFileHandle sfile;
+
+ sfile = (SMgrFileHandle) hash_search(SMgrFileHash,
+ &slocator,
+ HASH_FIND, NULL);
+ if (sfile != NULL)
+ smgrclose(sfile);
+ }
}
/*
- * smgrcreate() -- Create a new relation.
+ * smgrcreate() -- Create a new file.
*
- * Given an already-created (but presumably unused) SMgrRelation,
+ * Given an already-created (but presumably unused) SMgrFileHandle,
* cause the underlying disk file or other storage for the fork
* to be created.
*/
void
-smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo)
-{
- smgrsw[reln->smgr_which].smgr_create(reln, forknum, isRedo);
-}
-
-/*
- * smgrdosyncall() -- Immediately sync all forks of all given relations
- *
- * All forks of all given relations are synced out to the store.
- *
- * This is equivalent to FlushRelationBuffers() for each smgr relation,
- * then calling smgrimmedsync() for all forks of each relation, but it's
- * significantly quicker so should be preferred when possible.
- */
-void
-smgrdosyncall(SMgrRelation *rels, int nrels)
+smgrcreate(SMgrFileHandle sfile, bool isRedo)
{
- int i = 0;
- ForkNumber forknum;
-
- if (nrels == 0)
- return;
-
- FlushRelationsAllBuffers(rels, nrels);
-
- /*
- * Sync the physical file(s).
- */
- for (i = 0; i < nrels; i++)
- {
- int which = rels[i]->smgr_which;
-
- for (forknum = 0; forknum <= MAX_FORKNUM; forknum++)
- {
- if (smgrsw[which].smgr_exists(rels[i], forknum))
- smgrsw[which].smgr_immedsync(rels[i], forknum);
- }
- }
+ smgrsw[sfile->smgr_which].smgr_create(sfile, isRedo);
}
/*
- * smgrdounlinkall() -- Immediately unlink all forks of all given relations
+ * smgrunlink_multi() -- Immediately unlink given forks of given relation
*
- * All forks of all given relations are removed from the store. This
+ * The given forks of the relation are removed from the store. This
* should not be used during transactional operations, since it can't be
* undone.
*
+ * This handles multiple forks in one call, because the cache invalidation
+ * happens at relation granularity. If we had an smgrunlink() function
+ * to unlink just a single smgr file, and the caller wanted to delete
+ * multiple forks of a single relation, each call would send a new
+ * cache invalidation event, which would be wasteful.
+ *
* If isRedo is true, it is okay for the underlying file(s) to be gone
* already.
*/
void
-smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo)
+smgrunlink_multi(RelFileLocator rlocator, BackendId backend,
+ ForkNumber *forks, int nforks, bool isRedo)
{
- int i = 0;
- RelFileLocatorBackend *rlocators;
- ForkNumber forknum;
-
- if (nrels == 0)
- return;
-
- /*
- * Get rid of any remaining buffers for the relations. bufmgr will just
- * drop them without bothering to write the contents.
- */
- DropRelationsAllBuffers(rels, nrels);
-
- /*
- * create an array which contains all relations to be dropped, and close
- * each relation's forks at the smgr level while at it
- */
- rlocators = palloc(sizeof(RelFileLocatorBackend) * nrels);
- for (i = 0; i < nrels; i++)
- {
- RelFileLocatorBackend rlocator = rels[i]->smgr_rlocator;
- int which = rels[i]->smgr_which;
+ int which;
- rlocators[i] = rlocator;
+ which = 0; /* we only have md.c at present */
- /* Close the forks at smgr level */
- for (forknum = 0; forknum <= MAX_FORKNUM; forknum++)
- smgrsw[which].smgr_close(rels[i], forknum);
- }
+ /* Close the forks at smgr level */
+ smgrcloserellocator(rlocator, backend);
/*
* Send a shared-inval message to force other backends to close any
@@ -457,8 +402,7 @@ smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo)
* back to this backend, too, and thereby provide a backstop that we
* closed our own smgr rel.
*/
- for (i = 0; i < nrels; i++)
- CacheInvalidateSmgr(rlocators[i]);
+ CacheInvalidateSmgr(rlocator, backend);
/*
* Delete the physical file(s).
@@ -468,15 +412,12 @@ smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo)
* xact.
*/
- for (i = 0; i < nrels; i++)
+ for (int i = 0; i < nforks; i++)
{
- int which = rels[i]->smgr_which;
+ SMgrFileLocator slocator = {rlocator, backend, forks[i]};
- for (forknum = 0; forknum <= MAX_FORKNUM; forknum++)
- smgrsw[which].smgr_unlink(rlocators[i], forknum, isRedo);
+ smgrsw[which].smgr_unlink(slocator, isRedo);
}
-
- pfree(rlocators);
}
@@ -490,21 +431,21 @@ smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo)
* causes intervening file space to become filled with zeroes.
*/
void
-smgrextend(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
+smgrextend(SMgrFileHandle sfile, BlockNumber blocknum,
const void *buffer, bool skipFsync)
{
- smgrsw[reln->smgr_which].smgr_extend(reln, forknum, blocknum,
- buffer, skipFsync);
+ smgrsw[sfile->smgr_which].smgr_extend(sfile, blocknum,
+ buffer, skipFsync);
/*
* Normally we expect this to increase nblocks by one, but if the cached
* value isn't as expected, just invalidate it so the next call asks the
* kernel.
*/
- if (reln->smgr_cached_nblocks[forknum] == blocknum)
- reln->smgr_cached_nblocks[forknum] = blocknum + 1;
+ if (sfile->smgr_cached_nblocks == blocknum)
+ sfile->smgr_cached_nblocks = blocknum + 1;
else
- reln->smgr_cached_nblocks[forknum] = InvalidBlockNumber;
+ sfile->smgr_cached_nblocks = InvalidBlockNumber;
}
/*
@@ -515,13 +456,13 @@ smgrextend(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
* record).
*/
bool
-smgrprefetch(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum)
+smgrprefetch(SMgrFileHandle sfile, BlockNumber blocknum)
{
- return smgrsw[reln->smgr_which].smgr_prefetch(reln, forknum, blocknum);
+ return smgrsw[sfile->smgr_which].smgr_prefetch(sfile, blocknum);
}
/*
- * smgrread() -- read a particular block from a relation into the supplied
+ * smgrread() -- read a particular block from a file into the supplied
* buffer.
*
* This routine is called from the buffer manager in order to
@@ -529,10 +470,9 @@ smgrprefetch(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum)
* return pages in the format that POSTGRES expects.
*/
void
-smgrread(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
- void *buffer)
+smgrread(SMgrFileHandle sfile, BlockNumber blocknum, void *buffer)
{
- smgrsw[reln->smgr_which].smgr_read(reln, forknum, blocknum, buffer);
+ smgrsw[sfile->smgr_which].smgr_read(sfile, blocknum, buffer);
}
/*
@@ -551,11 +491,11 @@ smgrread(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
* do not require fsync.
*/
void
-smgrwrite(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
+smgrwrite(SMgrFileHandle sfile, BlockNumber blocknum,
const void *buffer, bool skipFsync)
{
- smgrsw[reln->smgr_which].smgr_write(reln, forknum, blocknum,
- buffer, skipFsync);
+ smgrsw[sfile->smgr_which].smgr_write(sfile, blocknum,
+ buffer, skipFsync);
}
@@ -564,11 +504,11 @@ smgrwrite(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
* blocks.
*/
void
-smgrwriteback(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
+smgrwriteback(SMgrFileHandle sfile, BlockNumber blocknum,
BlockNumber nblocks)
{
- smgrsw[reln->smgr_which].smgr_writeback(reln, forknum, blocknum,
- nblocks);
+ smgrsw[sfile->smgr_which].smgr_writeback(sfile, blocknum,
+ nblocks);
}
/*
@@ -576,18 +516,18 @@ smgrwriteback(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
* supplied relation.
*/
BlockNumber
-smgrnblocks(SMgrRelation reln, ForkNumber forknum)
+smgrnblocks(SMgrFileHandle sfile)
{
BlockNumber result;
/* Check and return if we get the cached value for the number of blocks. */
- result = smgrnblocks_cached(reln, forknum);
+ result = smgrnblocks_cached(sfile);
if (result != InvalidBlockNumber)
return result;
- result = smgrsw[reln->smgr_which].smgr_nblocks(reln, forknum);
+ result = smgrsw[sfile->smgr_which].smgr_nblocks(sfile);
- reln->smgr_cached_nblocks[forknum] = result;
+ sfile->smgr_cached_nblocks = result;
return result;
}
@@ -600,38 +540,41 @@ smgrnblocks(SMgrRelation reln, ForkNumber forknum)
* fork size is not cached.
*/
BlockNumber
-smgrnblocks_cached(SMgrRelation reln, ForkNumber forknum)
+smgrnblocks_cached(SMgrFileHandle sfile)
{
/*
* For now, we only use cached values in recovery due to lack of a shared
* invalidation mechanism for changes in file size.
*/
- if (InRecovery && reln->smgr_cached_nblocks[forknum] != InvalidBlockNumber)
- return reln->smgr_cached_nblocks[forknum];
+ if (InRecovery && sfile->smgr_cached_nblocks != InvalidBlockNumber)
+ return sfile->smgr_cached_nblocks;
return InvalidBlockNumber;
}
/*
- * smgrtruncate() -- Truncate the given forks of supplied relation to
- * each specified numbers of blocks
+ * smgrtruncate_multi() -- Truncate the given forks of supplied relation to
+ * each specified numbers of blocks
*
* The truncation is done immediately, so this can't be rolled back.
*
* The caller must hold AccessExclusiveLock on the relation, to ensure that
* other backends receive the smgr invalidation event that this function sends
* before they access any forks of the relation again.
+ *
+ * Like smgrunlink_multi(), this handles multiple forks in one call because
+ * the cache invalidation happens at relation granularity.
+ *
+ * NB: The caller is responsible for dropping buffers! Before v16, this
+ * function did it.
*/
void
-smgrtruncate(SMgrRelation reln, ForkNumber *forknum, int nforks, BlockNumber *nblocks)
+smgrtruncate_multi(RelFileLocator rlocator, BackendId backend, ForkNumber *forks,
+ int nforks, BlockNumber *nblocks)
{
int i;
- /*
- * Get rid of any buffers for the about-to-be-deleted blocks. bufmgr will
- * just drop them without bothering to write the contents.
- */
- DropRelationBuffers(reln, forknum, nforks, nblocks);
+ Assert(nforks < MAX_FORKNUM + 1);
/*
* Send a shared-inval message to force other backends to close any smgr
@@ -643,15 +586,19 @@ smgrtruncate(SMgrRelation reln, ForkNumber *forknum, int nforks, BlockNumber *nb
* is a performance-critical path.) As in the unlink code, we want to be
* sure the message is sent before we start changing things on-disk.
*/
- CacheInvalidateSmgr(reln->smgr_rlocator);
+ CacheInvalidateSmgr(rlocator, backend);
- /* Do the truncation */
+ /* Do the truncations */
for (i = 0; i < nforks; i++)
{
+ SMgrFileHandle sfile;
+
+ sfile = smgropen(rlocator, backend, forks[i]);
+
/* Make the cached size is invalid if we encounter an error. */
- reln->smgr_cached_nblocks[forknum[i]] = InvalidBlockNumber;
+ sfile->smgr_cached_nblocks = InvalidBlockNumber;
- smgrsw[reln->smgr_which].smgr_truncate(reln, forknum[i], nblocks[i]);
+ smgrsw[sfile->smgr_which].smgr_truncate(sfile, nblocks[i]);
/*
* We might as well update the local smgr_cached_nblocks values. The
@@ -660,7 +607,7 @@ smgrtruncate(SMgrRelation reln, ForkNumber *forknum, int nforks, BlockNumber *nb
* smgr_vm_nblocks, and these ones too at the next command boundary.
* But these ensure they aren't outright wrong until then.
*/
- reln->smgr_cached_nblocks[forknum[i]] = nblocks[i];
+ sfile->smgr_cached_nblocks = nblocks[i];
}
}
@@ -688,9 +635,9 @@ smgrtruncate(SMgrRelation reln, ForkNumber *forknum, int nforks, BlockNumber *nb
* otherwise the sync is not very meaningful.
*/
void
-smgrimmedsync(SMgrRelation reln, ForkNumber forknum)
+smgrimmedsync(SMgrFileHandle sfile)
{
- smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum);
+ smgrsw[sfile->smgr_which].smgr_immedsync(sfile);
}
/*
@@ -714,14 +661,14 @@ AtEOXact_SMgr(void)
* Zap all unowned SMgrRelations. We rely on smgrclose() to remove each
* one from the list.
*/
- dlist_foreach_modify(iter, &unowned_relns)
+ dlist_foreach_modify(iter, &unowned_sfiles)
{
- SMgrRelation rel = dlist_container(SMgrRelationData, node,
- iter.cur);
+ SMgrFileHandle sfile = dlist_container(SMgrFileData, node,
+ iter.cur);
- Assert(rel->smgr_owner == NULL);
+ Assert(sfile->smgr_owner == NULL);
- smgrclose(rel);
+ smgrclose(sfile);
}
}
diff --git a/src/backend/utils/cache/inval.c b/src/backend/utils/cache/inval.c
index 0008826f67c..7a1d8c012c2 100644
--- a/src/backend/utils/cache/inval.c
+++ b/src/backend/utils/cache/inval.c
@@ -661,11 +661,12 @@ LocalExecuteInvalidationMessage(SharedInvalidationMessage *msg)
* We could have smgr entries for relations of other databases, so no
* short-circuit test is possible here.
*/
- RelFileLocatorBackend rlocator;
+ RelFileLocator rlocator;
+ BackendId backend;
- rlocator.locator = msg->sm.rlocator;
- rlocator.backend = (msg->sm.backend_hi << 16) | (int) msg->sm.backend_lo;
- smgrcloserellocator(rlocator);
+ rlocator = msg->sm.rlocator;
+ backend = (msg->sm.backend_hi << 16) | (int) msg->sm.backend_lo;
+ smgrcloserellocator(rlocator, backend);
}
else if (msg->id == SHAREDINVALRELMAP_ID)
{
@@ -1459,14 +1460,14 @@ CacheInvalidateRelcacheByRelid(Oid relid)
* Thus, the maximum possible backend ID is 2^23-1.
*/
void
-CacheInvalidateSmgr(RelFileLocatorBackend rlocator)
+CacheInvalidateSmgr(RelFileLocator rlocator, BackendId backend)
{
SharedInvalidationMessage msg;
msg.sm.id = SHAREDINVALSMGR_ID;
- msg.sm.backend_hi = rlocator.backend >> 16;
- msg.sm.backend_lo = rlocator.backend & 0xffff;
- msg.sm.rlocator = rlocator.locator;
+ msg.sm.backend_hi = backend >> 16;
+ msg.sm.backend_lo = backend & 0xffff;
+ msg.sm.rlocator = rlocator;
/* check AddCatcacheInvalidationMessage() for an explanation */
VALGRIND_MAKE_MEM_DEFINED(&msg, sizeof(msg));
diff --git a/src/backend/utils/cache/relcache.c b/src/backend/utils/cache/relcache.c
index 13f79873733..45c22e0c137 100644
--- a/src/backend/utils/cache/relcache.c
+++ b/src/backend/utils/cache/relcache.c
@@ -419,7 +419,7 @@ AllocateRelationDesc(Form_pg_class relp)
relation = (Relation) palloc0(sizeof(RelationData));
/* make sure relation is marked as having no open file yet */
- relation->rd_smgr = NULL;
+ MemSet(relation->rd_smgr, 0, sizeof(relation->rd_smgr));
/*
* Copy the relation tuple form
@@ -1248,7 +1248,7 @@ retry:
RelationInitPhysicalAddr(relation);
/* make sure relation is marked as having no open file yet */
- relation->rd_smgr = NULL;
+ MemSet(relation->rd_smgr, 0, sizeof(relation->rd_smgr));
/*
* now we can free the memory allocated for pg_class_tuple
@@ -1877,7 +1877,7 @@ formrdesc(const char *relationName, Oid relationReltype,
relation = (Relation) palloc0(sizeof(RelationData));
/* make sure relation is marked as having no open file yet */
- relation->rd_smgr = NULL;
+ MemSet(relation->rd_smgr, 0, sizeof(relation->rd_smgr));
/*
* initialize reference count: 1 because it is nailed in cache
@@ -2701,7 +2701,8 @@ RelationClearRelation(Relation relation, bool rebuild)
}
/* rd_smgr must not be swapped, due to back-links from smgr level */
- SWAPFIELD(SMgrRelation, rd_smgr);
+ for (int i = 0; i <= MAX_FORKNUM; i++)
+ SWAPFIELD(SMgrFileHandle, rd_smgr[i]);
/* rd_refcnt must be preserved */
SWAPFIELD(int, rd_refcnt);
/* isnailed shouldn't change */
@@ -3532,7 +3533,7 @@ RelationBuildLocalRelation(const char *relname,
rel = (Relation) palloc0(sizeof(RelationData));
/* make sure relation is marked as having no open file yet */
- rel->rd_smgr = NULL;
+ MemSet(rel->rd_smgr, 0, sizeof(rel->rd_smgr));
/* mark it nailed if appropriate */
rel->rd_isnailed = nailit;
@@ -3764,7 +3765,11 @@ RelationSetNewRelfilenumber(Relation relation, char persistence)
*/
if (IsBinaryUpgrade)
{
- SMgrRelation srel;
+ ForkNumber forks[MAX_FORKNUM + 1];
+ RelFileLocatorBackend locator = {
+ relation->rd_locator,
+ relation->rd_backend
+ };
/*
* During a binary upgrade, we use this code path to ensure that
@@ -3781,9 +3786,16 @@ RelationSetNewRelfilenumber(Relation relation, char persistence)
* fails at this stage, the new cluster will need to be recreated
* anyway.
*/
- srel = smgropen(relation->rd_locator, relation->rd_backend);
- smgrdounlinkall(&srel, 1, false);
- smgrclose(srel);
+
+ /*
+ * Get rid of any remaining buffers for the relations. bufmgr will just
+ * drop them without bothering to write the contents.
+ */
+ DropRelationsAllBuffers(&locator, 1);
+
+ for (int i = 0; i <= MAX_FORKNUM; i++)
+ forks[i] = i;
+ smgrunlink_multi(relation->rd_locator, relation->rd_backend, forks, MAX_FORKNUM + 1, false);
}
else
{
@@ -3811,7 +3823,7 @@ RelationSetNewRelfilenumber(Relation relation, char persistence)
else if (RELKIND_HAS_STORAGE(relation->rd_rel->relkind))
{
/* handle these directly, at least for now */
- SMgrRelation srel;
+ SMgrFileHandle srel;
srel = RelationCreateStorage(newrlocator, persistence, true);
smgrclose(srel);
@@ -6298,7 +6310,7 @@ load_relcache_init_file(bool shared)
/*
* Reset transient-state fields in the relcache entry
*/
- rel->rd_smgr = NULL;
+ MemSet(rel->rd_smgr, 0, sizeof(rel->rd_smgr));
if (rel->rd_isnailed)
rel->rd_refcnt = 1;
else
diff --git a/src/common/relpath.c b/src/common/relpath.c
index 87de5f6c960..ae2d384fb34 100644
--- a/src/common/relpath.c
+++ b/src/common/relpath.c
@@ -129,7 +129,7 @@ GetDatabasePath(Oid dbOid, Oid spcOid)
}
/*
- * GetRelationPath - construct path to a relation's file
+ * GetSMgrFilePath - construct path to a relation's file
*
* Result is a palloc'd string.
*
@@ -138,7 +138,7 @@ GetDatabasePath(Oid dbOid, Oid spcOid)
* the trouble considering BackendId is just int anyway.
*/
char *
-GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber,
+GetSMgrFilePath(Oid dbOid, Oid spcOid, RelFileNumber relNumber,
int backendId, ForkNumber forkNumber)
{
char *path;
diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h
index 45a3c7835cb..c96f448225b 100644
--- a/src/include/catalog/storage.h
+++ b/src/include/catalog/storage.h
@@ -22,19 +22,20 @@
/* GUC variables */
extern PGDLLIMPORT int wal_skip_threshold;
-extern SMgrRelation RelationCreateStorage(RelFileLocator rlocator,
- char relpersistence,
- bool register_delete);
+extern SMgrFileHandle RelationCreateStorage(RelFileLocator rlocator,
+ char relpersistence,
+ bool register_delete);
extern void RelationDropStorage(Relation rel);
extern void RelationPreserveStorage(RelFileLocator rlocator, bool atCommit);
extern void RelationPreTruncate(Relation rel);
extern void RelationTruncate(Relation rel, BlockNumber nblocks);
-extern void RelationCopyStorage(SMgrRelation src, SMgrRelation dst,
- ForkNumber forkNum, char relpersistence);
+extern void RelationCopyStorage(SMgrFileHandle src, SMgrFileHandle dst,
+ char relpersistence);
extern bool RelFileLocatorSkippingWAL(RelFileLocator rlocator);
extern Size EstimatePendingSyncsSpace(void);
extern void SerializePendingSyncs(Size maxSize, char *startAddress);
extern void RestorePendingSyncs(char *startAddress);
+extern void DropRelationFiles(RelFileLocator *delrels, int ndelrels, bool isRedo);
/*
* These functions used to be in storage/smgr/smgr.c, which explains the
diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h
index 511c21682ec..12df11c7dfc 100644
--- a/src/include/common/relpath.h
+++ b/src/include/common/relpath.h
@@ -73,25 +73,25 @@ extern int forkname_chars(const char *str, ForkNumber *fork);
*/
extern char *GetDatabasePath(Oid dbOid, Oid spcOid);
-extern char *GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber,
+extern char *GetSMgrFilePath(Oid dbOid, Oid spcOid, RelFileNumber relNumber,
int backendId, ForkNumber forkNumber);
/*
* Wrapper macros for GetRelationPath. Beware of multiple
- * evaluation of the RelFileLocator or RelFileLocatorBackend argument!
+ * evaluation of the RelFileLocator or SMgrFileLocator argument!
*/
/* First argument is a RelFileLocator */
#define relpathbackend(rlocator, backend, forknum) \
- GetRelationPath((rlocator).dbOid, (rlocator).spcOid, (rlocator).relNumber, \
+ GetSMgrFilePath((rlocator).dbOid, (rlocator).spcOid, (rlocator).relNumber, \
backend, forknum)
/* First argument is a RelFileLocator */
#define relpathperm(rlocator, forknum) \
relpathbackend(rlocator, InvalidBackendId, forknum)
-/* First argument is a RelFileLocatorBackend */
-#define relpath(rlocator, forknum) \
- relpathbackend((rlocator).locator, (rlocator).backend, forknum)
+/* First argument is a SMgrFileLocator */
+#define smgrfilepath(slocator) \
+ GetSMgrFilePath((slocator).locator.dbOid, (slocator).locator.spcOid, (slocator).locator.relNumber, (slocator).backend, (slocator).forknum)
#endif /* RELPATH_H */
diff --git a/src/include/storage/buf_internals.h b/src/include/storage/buf_internals.h
index 0b448147407..6a37e0ce6b4 100644
--- a/src/include/storage/buf_internals.h
+++ b/src/include/storage/buf_internals.h
@@ -415,11 +415,10 @@ extern int BufTableInsert(BufferTag *tagPtr, uint32 hashcode, int buf_id);
extern void BufTableDelete(BufferTag *tagPtr, uint32 hashcode);
/* localbuf.c */
-extern PrefetchBufferResult PrefetchLocalBuffer(SMgrRelation smgr,
- ForkNumber forkNum,
+extern PrefetchBufferResult PrefetchLocalBuffer(SMgrFileHandle smgr,
BlockNumber blockNum);
-extern BufferDesc *LocalBufferAlloc(SMgrRelation smgr, ForkNumber forkNum,
- BlockNumber blockNum, bool *foundPtr, IOContext *io_context);
+extern BufferDesc *LocalBufferAlloc(SMgrFileHandle smgr, BlockNumber blockNum,
+ bool *foundPtr, IOContext *io_context);
extern void MarkLocalBufferDirty(Buffer buffer);
extern void DropRelationLocalBuffers(RelFileLocator rlocator,
ForkNumber forkNum,
diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h
index b8a18b8081f..7de50bf71b7 100644
--- a/src/include/storage/bufmgr.h
+++ b/src/include/storage/bufmgr.h
@@ -64,7 +64,7 @@ typedef struct PrefetchBufferResult
struct WritebackContext;
/* forward declared, to avoid including smgr.h here */
-struct SMgrRelationData;
+struct SMgrFileData;
/* in globals.c ... this duplicates miscadmin.h */
extern PGDLLIMPORT int NBuffers;
@@ -115,8 +115,7 @@ extern PGDLLIMPORT int32 *LocalRefCount;
/*
* prototypes for functions in bufmgr.c
*/
-extern PrefetchBufferResult PrefetchSharedBuffer(struct SMgrRelationData *smgr_reln,
- ForkNumber forkNum,
+extern PrefetchBufferResult PrefetchSharedBuffer(struct SMgrFileData *smgr_file,
BlockNumber blockNum);
extern PrefetchBufferResult PrefetchBuffer(Relation reln, ForkNumber forkNum,
BlockNumber blockNum);
@@ -144,18 +143,26 @@ extern void CheckPointBuffers(int flags);
extern BlockNumber BufferGetBlockNumber(Buffer buffer);
extern BlockNumber RelationGetNumberOfBlocksInFork(Relation relation,
ForkNumber forkNum);
-extern void FlushOneBuffer(Buffer buffer);
-extern void FlushRelationBuffers(Relation rel);
-extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels);
extern void CreateAndCopyRelationData(RelFileLocator src_rlocator,
RelFileLocator dst_rlocator,
bool permanent);
+
+extern void FlushOneBuffer(Buffer buffer);
+extern void FlushRelationBuffers(Relation rel);
+extern void FlushRelationsAllBuffers(RelFileLocator *locators, int nlocators);
extern void FlushDatabaseBuffers(Oid dbid);
-extern void DropRelationBuffers(struct SMgrRelationData *smgr_reln,
+
+extern void DropRelationBuffers(RelFileLocator rlocator, BackendId backend,
ForkNumber *forkNum,
int nforks, BlockNumber *firstDelBlock);
-extern void DropRelationsAllBuffers(struct SMgrRelationData **smgr_reln,
- int nlocators);
+
+typedef struct RelFileLocatorBackend
+{
+ RelFileLocator locator;
+ BackendId backend;
+} RelFileLocatorBackend;
+
+extern void DropRelationsAllBuffers(RelFileLocatorBackend *locators, int nlocators);
extern void DropDatabaseBuffers(Oid dbid);
#define RelationGetNumberOfBlocks(reln) \
diff --git a/src/include/storage/md.h b/src/include/storage/md.h
index 8f32af9ef3d..4046584a830 100644
--- a/src/include/storage/md.h
+++ b/src/include/storage/md.h
@@ -21,28 +21,26 @@
/* md storage manager functionality */
extern void mdinit(void);
-extern void mdopen(SMgrRelation reln);
-extern void mdclose(SMgrRelation reln, ForkNumber forknum);
-extern void mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo);
-extern bool mdexists(SMgrRelation reln, ForkNumber forknum);
-extern void mdunlink(RelFileLocatorBackend rlocator, ForkNumber forknum, bool isRedo);
-extern void mdextend(SMgrRelation reln, ForkNumber forknum,
+extern void mdopen(SMgrFileHandle sfile);
+extern void mdclose(SMgrFileHandle sfile);
+extern void mdcreate(SMgrFileHandle sfile, bool isRedo);
+extern bool mdexists(SMgrFileHandle sfile);
+extern void mdunlink(SMgrFileLocator slocator, bool isRedo);
+extern void mdextend(SMgrFileHandle sfile,
BlockNumber blocknum, const void *buffer, bool skipFsync);
-extern bool mdprefetch(SMgrRelation reln, ForkNumber forknum,
+extern bool mdprefetch(SMgrFileHandle sfile,
BlockNumber blocknum);
-extern void mdread(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
+extern void mdread(SMgrFileHandle sfile, BlockNumber blocknum,
void *buffer);
-extern void mdwrite(SMgrRelation reln, ForkNumber forknum,
+extern void mdwrite(SMgrFileHandle sfile,
BlockNumber blocknum, const void *buffer, bool skipFsync);
-extern void mdwriteback(SMgrRelation reln, ForkNumber forknum,
+extern void mdwriteback(SMgrFileHandle sfile,
BlockNumber blocknum, BlockNumber nblocks);
-extern BlockNumber mdnblocks(SMgrRelation reln, ForkNumber forknum);
-extern void mdtruncate(SMgrRelation reln, ForkNumber forknum,
- BlockNumber nblocks);
-extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum);
+extern BlockNumber mdnblocks(SMgrFileHandle sfile);
+extern void mdtruncate(SMgrFileHandle sfile, BlockNumber nblocks);
+extern void mdimmedsync(SMgrFileHandle sfile);
extern void ForgetDatabaseSyncRequests(Oid dbid);
-extern void DropRelationFiles(RelFileLocator *delrels, int ndelrels, bool isRedo);
/* md sync callbacks */
extern int mdsyncfiletag(const FileTag *ftag, char *path);
diff --git a/src/include/storage/relfilelocator.h b/src/include/storage/relfilelocator.h
index 61cf0169bd7..e2c22b8e6f9 100644
--- a/src/include/storage/relfilelocator.h
+++ b/src/include/storage/relfilelocator.h
@@ -53,6 +53,8 @@
* Note: various places use RelFileLocator in hashtable keys. Therefore,
* there *must not* be any unused padding bytes in this struct. That
* should be safe as long as all the fields are of type Oid.
+ *
+ * See also SMgrFileLocator in smgr.h.
*/
typedef struct RelFileLocator
{
@@ -62,38 +64,15 @@ typedef struct RelFileLocator
} RelFileLocator;
/*
- * Augmenting a relfilelocator with the backend ID provides all the information
- * we need to locate the physical storage. The backend ID is InvalidBackendId
- * for regular relations (those accessible to more than one backend), or the
- * owning backend's ID for backend-local relations. Backend-local relations
- * are always transient and removed in case of a database crash; they are
- * never WAL-logged or fsync'd.
- */
-typedef struct RelFileLocatorBackend
-{
- RelFileLocator locator;
- BackendId backend;
-} RelFileLocatorBackend;
-
-#define RelFileLocatorBackendIsTemp(rlocator) \
- ((rlocator).backend != InvalidBackendId)
-
-/*
- * Note: RelFileLocatorEquals and RelFileLocatorBackendEquals compare relNumber
+ * Note: RelFileLocatorEquals compares relNumber
* first since that is most likely to be different in two unequal
* RelFileLocators. It is probably redundant to compare spcOid if the other
* fields are found equal, but do it anyway to be sure. Likewise for checking
- * the backend ID in RelFileLocatorBackendEquals.
+ * the backend ID in SMgrFileLocatorBackendEquals.
*/
#define RelFileLocatorEquals(locator1, locator2) \
((locator1).relNumber == (locator2).relNumber && \
(locator1).dbOid == (locator2).dbOid && \
(locator1).spcOid == (locator2).spcOid)
-#define RelFileLocatorBackendEquals(locator1, locator2) \
- ((locator1).locator.relNumber == (locator2).locator.relNumber && \
- (locator1).locator.dbOid == (locator2).locator.dbOid && \
- (locator1).backend == (locator2).backend && \
- (locator1).locator.spcOid == (locator2).locator.spcOid)
-
#endif /* RELFILELOCATOR_H */
diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h
index 0935144f425..fe7282127ed 100644
--- a/src/include/storage/smgr.h
+++ b/src/include/storage/smgr.h
@@ -18,31 +18,56 @@
#include "storage/block.h"
#include "storage/relfilelocator.h"
+
/*
- * smgr.c maintains a table of SMgrRelation objects, which are essentially
- * cached file handles. An SMgrRelation is created (if not already present)
+ * SMgrFileLocator contains all the information needed to locate the physical
+ * storage of a relation fork, or some other file that is managed by the buffer
+ * manager.
+ *
+ * The backend ID is InvalidBackendId for regular relations (those accessible
+ * to more than one backend), or the owning backend's ID for backend-local
+ * relations. Backend-local relations are always transient and removed in
+ * case of a database crash; they are never WAL-logged or fsync'd.
+ */
+typedef struct SMgrFileLocator
+{
+ RelFileLocator locator;
+ BackendId backend;
+ ForkNumber forknum;
+} SMgrFileLocator;
+
+#define SMgrFileLocatorIsTemp(slocator) \
+ ((slocator).backend != InvalidBackendId)
+
+/*
+ * smgr.c maintains a table of SMgrFileData objects, which are essentially
+ * cached file handles. An SMgrFile is created (if not already present)
* by smgropen(), and destroyed by smgrclose(). Note that neither of these
* operations imply I/O, they just create or destroy a hashtable entry.
* (But smgrclose() may release associated resources, such as OS-level file
* descriptors.)
*
- * An SMgrRelation may have an "owner", which is just a pointer to it from
- * somewhere else; smgr.c will clear this pointer if the SMgrRelation is
+ * An SMgrFile may have an "owner", which is just a pointer to it from
+ * somewhere else; smgr.c will clear this pointer if the SMgrFile is
* closed. We use this to avoid dangling pointers from relcache to smgr
* without having to make the smgr explicitly aware of relcache. There
* can't be more than one "owner" pointer per SMgrRelation, but that's
* all we need.
*
- * SMgrRelations that do not have an "owner" are considered to be transient,
+ * SMgrFiles that do not have an "owner" are considered to be transient,
* and are deleted at end of transaction.
+ *
+ * A file that is represented by an SMgrFile can be managed by the buffer
+ * manager. Currently, it's only used for relation files, but could be used
+ * for SLRUs and other things in the future.
*/
-typedef struct SMgrRelationData
+typedef struct SMgrFileData
{
- /* rlocator is the hashtable lookup key, so it must be first! */
- RelFileLocatorBackend smgr_rlocator; /* relation physical identifier */
+ /* locator is the hashtable lookup key, so must be first! */
+ SMgrFileLocator smgr_locator; /* file physical identifier */
/* pointer to owning pointer, or NULL if none */
- struct SMgrRelationData **smgr_owner;
+ struct SMgrFileData **smgr_owner;
/*
* The following fields are reset to InvalidBlockNumber upon a cache flush
@@ -51,7 +76,7 @@ typedef struct SMgrRelationData
* invalidation for fork extension.
*/
BlockNumber smgr_targblock; /* current insertion target block */
- BlockNumber smgr_cached_nblocks[MAX_FORKNUM + 1]; /* last known size */
+ BlockNumber smgr_cached_nblocks; /* last known size */
/* additional public fields may someday exist here */
@@ -65,46 +90,45 @@ typedef struct SMgrRelationData
* for md.c; per-fork arrays of the number of open segments
* (md_num_open_segs) and the segments themselves (md_seg_fds).
*/
- int md_num_open_segs[MAX_FORKNUM + 1];
- struct _MdfdVec *md_seg_fds[MAX_FORKNUM + 1];
+ int md_num_open_segs;
+ struct _MdfdVec *md_seg_fds;
- /* if unowned, list link in list of all unowned SMgrRelations */
+ /* if unowned, list link in list of all unowned SMgrFiles */
dlist_node node;
-} SMgrRelationData;
+} SMgrFileData;
-typedef SMgrRelationData *SMgrRelation;
+typedef SMgrFileData *SMgrFileHandle;
#define SmgrIsTemp(smgr) \
- RelFileLocatorBackendIsTemp((smgr)->smgr_rlocator)
+ SMgrFileLocatorIsTemp((smgr)->smgr_locator)
extern void smgrinit(void);
-extern SMgrRelation smgropen(RelFileLocator rlocator, BackendId backend);
-extern bool smgrexists(SMgrRelation reln, ForkNumber forknum);
-extern void smgrsetowner(SMgrRelation *owner, SMgrRelation reln);
-extern void smgrclearowner(SMgrRelation *owner, SMgrRelation reln);
-extern void smgrclose(SMgrRelation reln);
+extern SMgrFileHandle smgropen(RelFileLocator rlocator, BackendId backend, ForkNumber forkNum);
+extern bool smgrexists(SMgrFileHandle sfile);
+extern void smgrsetowner(SMgrFileHandle *owner, SMgrFileHandle sfile);
+extern void smgrclearowner(SMgrFileHandle *owner, SMgrFileHandle sfile);
+extern void smgrclose(SMgrFileHandle sfile);
extern void smgrcloseall(void);
-extern void smgrcloserellocator(RelFileLocatorBackend rlocator);
-extern void smgrrelease(SMgrRelation reln);
extern void smgrreleaseall(void);
-extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo);
-extern void smgrdosyncall(SMgrRelation *rels, int nrels);
-extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo);
-extern void smgrextend(SMgrRelation reln, ForkNumber forknum,
+extern void smgrcreate(SMgrFileHandle sfile, bool isRedo);
+extern void smgrextend(SMgrFileHandle sfile,
BlockNumber blocknum, const void *buffer, bool skipFsync);
-extern bool smgrprefetch(SMgrRelation reln, ForkNumber forknum,
- BlockNumber blocknum);
-extern void smgrread(SMgrRelation reln, ForkNumber forknum,
+extern bool smgrprefetch(SMgrFileHandle sfile, BlockNumber blocknum);
+extern void smgrread(SMgrFileHandle sfile,
BlockNumber blocknum, void *buffer);
-extern void smgrwrite(SMgrRelation reln, ForkNumber forknum,
+extern void smgrwrite(SMgrFileHandle sfile,
BlockNumber blocknum, const void *buffer, bool skipFsync);
-extern void smgrwriteback(SMgrRelation reln, ForkNumber forknum,
+extern void smgrwriteback(SMgrFileHandle sfile,
BlockNumber blocknum, BlockNumber nblocks);
-extern BlockNumber smgrnblocks(SMgrRelation reln, ForkNumber forknum);
-extern BlockNumber smgrnblocks_cached(SMgrRelation reln, ForkNumber forknum);
-extern void smgrtruncate(SMgrRelation reln, ForkNumber *forknum,
- int nforks, BlockNumber *nblocks);
-extern void smgrimmedsync(SMgrRelation reln, ForkNumber forknum);
+extern BlockNumber smgrnblocks(SMgrFileHandle sfile);
+extern BlockNumber smgrnblocks_cached(SMgrFileHandle sfile);
+extern void smgrimmedsync(SMgrFileHandle sfile);
+
+extern void smgrtruncate_multi(RelFileLocator rlocator, BackendId backend, ForkNumber *forks, int nforks, BlockNumber *nblocks);
+extern void smgrunlink_multi(RelFileLocator rlocator, BackendId backend, ForkNumber *forks, int nforks, bool isRedo);
+
+extern void smgrcloserellocator(RelFileLocator rlocator, BackendId backend);
+
extern void AtEOXact_SMgr(void);
extern bool ProcessBarrierSmgrRelease(void);
diff --git a/src/include/utils/inval.h b/src/include/utils/inval.h
index 14b4eac0630..6c279422b2e 100644
--- a/src/include/utils/inval.h
+++ b/src/include/utils/inval.h
@@ -48,7 +48,7 @@ extern void CacheInvalidateRelcacheByTuple(HeapTuple classTuple);
extern void CacheInvalidateRelcacheByRelid(Oid relid);
-extern void CacheInvalidateSmgr(RelFileLocatorBackend rlocator);
+extern void CacheInvalidateSmgr(RelFileLocator rlocator, BackendId backend);
extern void CacheInvalidateRelmap(Oid databaseId);
diff --git a/src/include/utils/rel.h b/src/include/utils/rel.h
index 67f994cb3e2..7ced81d869b 100644
--- a/src/include/utils/rel.h
+++ b/src/include/utils/rel.h
@@ -54,7 +54,7 @@ typedef LockInfoData *LockInfo;
typedef struct RelationData
{
RelFileLocator rd_locator; /* relation physical identifier */
- SMgrRelation rd_smgr; /* cached file handle, or NULL */
+ SMgrFileHandle rd_smgr[MAX_FORKNUM + 1]; /* cached file handles, or NULLs */
int rd_refcnt; /* reference count */
BackendId rd_backend; /* owning backend id, if temporary relation */
bool rd_islocaltemp; /* rel is a temp rel of this session */
@@ -562,15 +562,15 @@ typedef struct ViewOptions
* Note: since a relcache flush can cause the file handle to be closed again,
* it's unwise to hold onto the pointer returned by this function for any
* long period. Recommended practice is to just re-execute RelationGetSmgr
- * each time you need to access the SMgrRelation. It's quite cheap in
+ * each time you need to access the SMgrFileHandle. It's quite cheap in
* comparison to whatever an smgr function is going to do.
*/
-static inline SMgrRelation
-RelationGetSmgr(Relation rel)
+static inline SMgrFileHandle
+RelationGetSmgr(Relation rel, ForkNumber forkNum)
{
- if (unlikely(rel->rd_smgr == NULL))
- smgrsetowner(&(rel->rd_smgr), smgropen(rel->rd_locator, rel->rd_backend));
- return rel->rd_smgr;
+ if (unlikely(rel->rd_smgr[forkNum] == NULL))
+ smgrsetowner(&(rel->rd_smgr[forkNum]), smgropen(rel->rd_locator, rel->rd_backend, forkNum));
+ return rel->rd_smgr[forkNum];
}
/*
@@ -580,11 +580,15 @@ RelationGetSmgr(Relation rel)
static inline void
RelationCloseSmgr(Relation relation)
{
- if (relation->rd_smgr != NULL)
- smgrclose(relation->rd_smgr);
-
- /* smgrclose should unhook from owner pointer */
- Assert(relation->rd_smgr == NULL);
+ for (int i = 0; i <= MAX_FORKNUM; i++)
+ {
+ if (relation->rd_smgr[i] != NULL)
+ {
+ smgrclose(relation->rd_smgr[i]);
+ /* smgrclose should unhook from owner pointer */
+ Assert(relation->rd_smgr[i] == NULL);
+ }
+ }
}
#endif /* !FRONTEND */
@@ -597,7 +601,7 @@ RelationCloseSmgr(Relation relation)
* so there's no need to re-open the smgr handle if it's not currently open.
*/
#define RelationGetTargetBlock(relation) \
- ( (relation)->rd_smgr != NULL ? (relation)->rd_smgr->smgr_targblock : InvalidBlockNumber )
+ ( (relation)->rd_smgr[MAIN_FORKNUM] != NULL ? (relation)->rd_smgr[MAIN_FORKNUM]->smgr_targblock : InvalidBlockNumber )
/*
* RelationSetTargetBlock
@@ -605,7 +609,7 @@ RelationCloseSmgr(Relation relation)
*/
#define RelationSetTargetBlock(relation, targblock) \
do { \
- RelationGetSmgr(relation)->smgr_targblock = (targblock); \
+ RelationGetSmgr(relation, MAIN_FORKNUM)->smgr_targblock = (targblock); \
} while (0)
/*
--
2.30.2
[text/x-patch] v2-0002-WIP-SLRUs.patch (147.0K, ../../[email protected]/3-v2-0002-WIP-SLRUs.patch)
download | inline diff:
From e1bdd6ff380d0d80f349d1f096e5581c3ef4a953 Mon Sep 17 00:00:00 2001
From: Heikki Linnakangas <[email protected]>
Date: Mon, 27 Feb 2023 15:19:01 +0200
Subject: [PATCH v2 2/2] WIP: SLRUs
This is Thomas's patch refactored over the per-fork SMgrFile patch:
- each SLRU segment is represented by a separate SMgrFile.
- md.c implementation handles SLRUs, too
---
src/backend/access/transam/clog.c | 199 +---
src/backend/access/transam/commit_ts.c | 156 +--
src/backend/access/transam/multixact.c | 357 ++----
src/backend/access/transam/slru.c | 1495 +++---------------------
src/backend/access/transam/subtrans.c | 108 +-
src/backend/access/transam/xact.c | 2 +
src/backend/access/transam/xlog.c | 15 +-
src/backend/commands/async.c | 67 +-
src/backend/storage/buffer/buf_init.c | 17 +-
src/backend/storage/buffer/bufmgr.c | 62 +-
src/backend/storage/ipc/ipci.c | 4 -
src/backend/storage/lmgr/predicate.c | 62 +-
src/backend/storage/smgr/md.c | 9 +-
src/backend/storage/smgr/smgr.c | 42 +
src/backend/storage/sync/sync.c | 20 +-
src/backend/utils/mmgr/mcxt.c | 40 +-
src/common/relpath.c | 27 +-
src/include/access/clog.h | 6 -
src/include/access/commit_ts.h | 3 -
src/include/access/multixact.h | 3 -
src/include/access/slru.h | 180 +--
src/include/access/subtrans.h | 3 -
src/include/common/relpath.h | 3 +
src/include/storage/buf_internals.h | 13 +
src/include/storage/bufmgr.h | 6 +
src/include/storage/smgr.h | 1 +
src/test/modules/Makefile | 3 +-
27 files changed, 695 insertions(+), 2208 deletions(-)
diff --git a/src/backend/access/transam/clog.c b/src/backend/access/transam/clog.c
index 4a431d58767..b6f5ae987b1 100644
--- a/src/backend/access/transam/clog.c
+++ b/src/backend/access/transam/clog.c
@@ -41,6 +41,8 @@
#include "miscadmin.h"
#include "pg_trace.h"
#include "pgstat.h"
+#include "storage/bufmgr.h"
+#include "storage/buf_internals.h"
#include "storage/proc.h"
#include "storage/sync.h"
@@ -81,15 +83,8 @@
*/
#define THRESHOLD_SUBTRANS_CLOG_OPT 5
-/*
- * Link to shared-memory data structures for CLOG control
- */
-static SlruCtlData XactCtlData;
-
-#define XactCtl (&XactCtlData)
-
-static int ZeroCLOGPage(int pageno, bool writeXlog);
+static Buffer ZeroCLOGPage(int pageno, bool writeXlog);
static bool CLOGPagePrecedes(int page1, int page2);
static void WriteZeroPageXlogRec(int pageno);
static void WriteTruncateXlogRec(int pageno, TransactionId oldestXact,
@@ -99,7 +94,7 @@ static void TransactionIdSetPageStatus(TransactionId xid, int nsubxids,
XLogRecPtr lsn, int pageno,
bool all_xact_same_page);
static void TransactionIdSetStatusBit(TransactionId xid, XidStatus status,
- XLogRecPtr lsn, int slotno);
+ XLogRecPtr lsn, Buffer buffer);
static void set_status_by_pages(int nsubxids, TransactionId *subxids,
XidStatus status, XLogRecPtr lsn);
static bool TransactionGroupUpdateXidStatus(TransactionId xid,
@@ -339,13 +334,12 @@ TransactionIdSetPageStatusInternal(TransactionId xid, int nsubxids,
TransactionId *subxids, XidStatus status,
XLogRecPtr lsn, int pageno)
{
- int slotno;
+ Buffer buffer;
int i;
Assert(status == TRANSACTION_STATUS_COMMITTED ||
status == TRANSACTION_STATUS_ABORTED ||
(status == TRANSACTION_STATUS_SUB_COMMITTED && !TransactionIdIsValid(xid)));
- Assert(LWLockHeldByMeInMode(XactSLRULock, LW_EXCLUSIVE));
/*
* If we're doing an async commit (ie, lsn is valid), then we must wait
@@ -356,7 +350,8 @@ TransactionIdSetPageStatusInternal(TransactionId xid, int nsubxids,
* write-busy, since we don't care if the update reaches disk sooner than
* we think.
*/
- slotno = SimpleLruReadPage(XactCtl, pageno, XLogRecPtrIsInvalid(lsn), xid);
+ buffer = ReadSlruBuffer(SLRU_CLOG_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
/*
* Set the main transaction id, if any.
@@ -374,25 +369,26 @@ TransactionIdSetPageStatusInternal(TransactionId xid, int nsubxids,
{
for (i = 0; i < nsubxids; i++)
{
- Assert(XactCtl->shared->page_number[slotno] == TransactionIdToPage(subxids[i]));
+ Assert(pageno == TransactionIdToPage(subxids[i]));
TransactionIdSetStatusBit(subxids[i],
TRANSACTION_STATUS_SUB_COMMITTED,
- lsn, slotno);
+ lsn, buffer);
}
}
/* ... then the main transaction */
- TransactionIdSetStatusBit(xid, status, lsn, slotno);
+ TransactionIdSetStatusBit(xid, status, lsn, buffer);
}
/* Set the subtransactions */
for (i = 0; i < nsubxids; i++)
{
- Assert(XactCtl->shared->page_number[slotno] == TransactionIdToPage(subxids[i]));
- TransactionIdSetStatusBit(subxids[i], status, lsn, slotno);
+ Assert(pageno == TransactionIdToPage(subxids[i]));
+ TransactionIdSetStatusBit(subxids[i], status, lsn, buffer);
}
- XactCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -566,7 +562,7 @@ TransactionGroupUpdateXidStatus(TransactionId xid, XidStatus status,
* Must be called with XactSLRULock held
*/
static void
-TransactionIdSetStatusBit(TransactionId xid, XidStatus status, XLogRecPtr lsn, int slotno)
+TransactionIdSetStatusBit(TransactionId xid, XidStatus status, XLogRecPtr lsn, Buffer buffer)
{
int byteno = TransactionIdToByte(xid);
int bshift = TransactionIdToBIndex(xid) * CLOG_BITS_PER_XACT;
@@ -574,7 +570,10 @@ TransactionIdSetStatusBit(TransactionId xid, XidStatus status, XLogRecPtr lsn, i
char byteval;
char curval;
- byteptr = XactCtl->shared->page_buffer[slotno] + byteno;
+ Assert(LWLockHeldByMeInMode(BufferDescriptorGetContentLock(GetBufferDescriptor(buffer - 1)),
+ LW_EXCLUSIVE));
+
+ byteptr = BufferGetPage(buffer) + byteno;
curval = (*byteptr >> bshift) & CLOG_XACT_BITMASK;
/*
@@ -603,7 +602,7 @@ TransactionIdSetStatusBit(TransactionId xid, XidStatus status, XLogRecPtr lsn, i
*byteptr = byteval;
/*
- * Update the group LSN if the transaction completion LSN is higher.
+ * Update the buffer LSN if the transaction completion LSN is higher.
*
* Note: lsn will be invalid when supplied during InRecovery processing,
* so we don't need to do anything special to avoid LSN updates during
@@ -612,10 +611,8 @@ TransactionIdSetStatusBit(TransactionId xid, XidStatus status, XLogRecPtr lsn, i
*/
if (!XLogRecPtrIsInvalid(lsn))
{
- int lsnindex = GetLSNIndex(slotno, xid);
-
- if (XactCtl->shared->group_lsn[lsnindex] < lsn)
- XactCtl->shared->group_lsn[lsnindex] = lsn;
+ if (BufferGetExternalLSN(GetBufferDescriptor(buffer)) < lsn)
+ BufferSetExternalLSN(GetBufferDescriptor(buffer), lsn);
}
}
@@ -640,67 +637,22 @@ TransactionIdGetStatus(TransactionId xid, XLogRecPtr *lsn)
int pageno = TransactionIdToPage(xid);
int byteno = TransactionIdToByte(xid);
int bshift = TransactionIdToBIndex(xid) * CLOG_BITS_PER_XACT;
- int slotno;
- int lsnindex;
char *byteptr;
XidStatus status;
+ Buffer buffer;
- /* lock is acquired by SimpleLruReadPage_ReadOnly */
-
- slotno = SimpleLruReadPage_ReadOnly(XactCtl, pageno, xid);
- byteptr = XactCtl->shared->page_buffer[slotno] + byteno;
+ buffer = ReadSlruBuffer(SLRU_CLOG_ID, pageno);
+ byteptr = BufferGetPage(buffer) + byteno;
status = (*byteptr >> bshift) & CLOG_XACT_BITMASK;
- lsnindex = GetLSNIndex(slotno, xid);
- *lsn = XactCtl->shared->group_lsn[lsnindex];
+ *lsn = BufferGetExternalLSN(GetBufferDescriptor(buffer));
- LWLockRelease(XactSLRULock);
+ ReleaseBuffer(buffer);
return status;
}
-/*
- * Number of shared CLOG buffers.
- *
- * On larger multi-processor systems, it is possible to have many CLOG page
- * requests in flight at one time which could lead to disk access for CLOG
- * page if the required page is not found in memory. Testing revealed that we
- * can get the best performance by having 128 CLOG buffers, more than that it
- * doesn't improve performance.
- *
- * Unconditionally keeping the number of CLOG buffers to 128 did not seem like
- * a good idea, because it would increase the minimum amount of shared memory
- * required to start, which could be a problem for people running very small
- * configurations. The following formula seems to represent a reasonable
- * compromise: people with very low values for shared_buffers will get fewer
- * CLOG buffers as well, and everyone else will get 128.
- */
-Size
-CLOGShmemBuffers(void)
-{
- return Min(128, Max(4, NBuffers / 512));
-}
-
-/*
- * Initialization of shared memory for CLOG
- */
-Size
-CLOGShmemSize(void)
-{
- return SimpleLruShmemSize(CLOGShmemBuffers(), CLOG_LSNS_PER_PAGE);
-}
-
-void
-CLOGShmemInit(void)
-{
- XactCtl->PagePrecedes = CLOGPagePrecedes;
- SimpleLruInit(XactCtl, "Xact", CLOGShmemBuffers(), CLOG_LSNS_PER_PAGE,
- XactSLRULock, "pg_xact", LWTRANCHE_XACT_BUFFER,
- SYNC_HANDLER_CLOG);
- SlruPagePrecedesUnitTests(XactCtl, CLOG_XACTS_PER_PAGE);
-}
-
/*
* This func must be called ONCE on system install. It creates
* the initial CLOG segment. (The CLOG directory is assumed to
@@ -710,18 +662,15 @@ CLOGShmemInit(void)
void
BootStrapCLOG(void)
{
- int slotno;
-
- LWLockAcquire(XactSLRULock, LW_EXCLUSIVE);
+ Buffer buffer;
/* Create and zero the first page of the commit log */
- slotno = ZeroCLOGPage(0, false);
+ buffer = ZeroCLOGPage(0, false);
/* Make sure it's written out */
- SimpleLruWritePage(XactCtl, slotno);
- Assert(!XactCtl->shared->page_dirty[slotno]);
+ FlushOneBuffer(buffer);
- LWLockRelease(XactSLRULock);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -733,17 +682,18 @@ BootStrapCLOG(void)
*
* Control lock must be held at entry, and will be held at exit.
*/
-static int
+static Buffer
ZeroCLOGPage(int pageno, bool writeXlog)
{
- int slotno;
+ Buffer buffer;
- slotno = SimpleLruZeroPage(XactCtl, pageno);
+ buffer = ZeroSlruBuffer(SLRU_CLOG_ID, pageno);
+ MarkBufferDirty(buffer);
if (writeXlog)
WriteZeroPageXlogRec(pageno);
- return slotno;
+ return buffer;
}
/*
@@ -753,17 +703,6 @@ ZeroCLOGPage(int pageno, bool writeXlog)
void
StartupCLOG(void)
{
- TransactionId xid = XidFromFullTransactionId(ShmemVariableCache->nextXid);
- int pageno = TransactionIdToPage(xid);
-
- LWLockAcquire(XactSLRULock, LW_EXCLUSIVE);
-
- /*
- * Initialize our idea of the latest page number.
- */
- XactCtl->shared->latest_page_number = pageno;
-
- LWLockRelease(XactSLRULock);
}
/*
@@ -775,8 +714,6 @@ TrimCLOG(void)
TransactionId xid = XidFromFullTransactionId(ShmemVariableCache->nextXid);
int pageno = TransactionIdToPage(xid);
- LWLockAcquire(XactSLRULock, LW_EXCLUSIVE);
-
/*
* Zero out the remainder of the current clog page. Under normal
* circumstances it should be zeroes already, but it seems at least
@@ -793,40 +730,24 @@ TrimCLOG(void)
{
int byteno = TransactionIdToByte(xid);
int bshift = TransactionIdToBIndex(xid) * CLOG_BITS_PER_XACT;
- int slotno;
char *byteptr;
+ Buffer buffer;
- slotno = SimpleLruReadPage(XactCtl, pageno, false, xid);
- byteptr = XactCtl->shared->page_buffer[slotno] + byteno;
+ buffer = ReadSlruBuffer(SLRU_CLOG_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ byteptr = BufferGetPage(buffer) + byteno;
/* Zero so-far-unused positions in the current byte */
*byteptr &= (1 << bshift) - 1;
/* Zero the rest of the page */
MemSet(byteptr + 1, 0, BLCKSZ - byteno - 1);
- XactCtl->shared->page_dirty[slotno] = true;
- }
+ MarkBufferDirty(buffer);
- LWLockRelease(XactSLRULock);
-}
-
-/*
- * Perform a checkpoint --- either during shutdown, or on-the-fly
- */
-void
-CheckPointCLOG(void)
-{
- /*
- * Write dirty CLOG pages to disk. This may result in sync requests
- * queued for later handling by ProcessSyncRequests(), as part of the
- * checkpoint.
- */
- TRACE_POSTGRESQL_CLOG_CHECKPOINT_START(true);
- SimpleLruWriteAll(XactCtl, true);
- TRACE_POSTGRESQL_CLOG_CHECKPOINT_DONE(true);
+ UnlockReleaseBuffer(buffer);
+ }
}
-
/*
* Make sure that CLOG has room for a newly-allocated XID.
*
@@ -850,12 +771,8 @@ ExtendCLOG(TransactionId newestXact)
pageno = TransactionIdToPage(newestXact);
- LWLockAcquire(XactSLRULock, LW_EXCLUSIVE);
-
/* Zero the page and make an XLOG entry about it */
- ZeroCLOGPage(pageno, true);
-
- LWLockRelease(XactSLRULock);
+ UnlockReleaseBuffer(ZeroCLOGPage(pageno, true));
}
@@ -886,7 +803,8 @@ TruncateCLOG(TransactionId oldestXact, Oid oldestxid_datoid)
cutoffPage = TransactionIdToPage(oldestXact);
/* Check to see if there's any files that could be removed */
- if (!SlruScanDirectory(XactCtl, SlruScanDirCbReportPresence, &cutoffPage))
+ if (!SlruScanDirectory(SLRU_CLOG_ID, CLOGPagePrecedes,
+ SlruScanDirCbReportPresence, &cutoffPage))
return; /* nothing to remove */
/*
@@ -907,7 +825,7 @@ TruncateCLOG(TransactionId oldestXact, Oid oldestxid_datoid)
WriteTruncateXlogRec(cutoffPage, oldestXact, oldestxid_datoid);
/* Now we can remove the old CLOG segment(s) */
- SimpleLruTruncate(XactCtl, cutoffPage);
+ SimpleLruTruncate(SLRU_CLOG_ID, CLOGPagePrecedes, cutoffPage);
}
@@ -992,17 +910,13 @@ clog_redo(XLogReaderState *record)
if (info == CLOG_ZEROPAGE)
{
int pageno;
- int slotno;
+ Buffer buffer;
memcpy(&pageno, XLogRecGetData(record), sizeof(int));
- LWLockAcquire(XactSLRULock, LW_EXCLUSIVE);
-
- slotno = ZeroCLOGPage(pageno, false);
- SimpleLruWritePage(XactCtl, slotno);
- Assert(!XactCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(XactSLRULock);
+ buffer = ZeroCLOGPage(pageno, false);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
else if (info == CLOG_TRUNCATE)
{
@@ -1012,17 +926,8 @@ clog_redo(XLogReaderState *record)
AdvanceOldestClogXid(xlrec.oldestXact);
- SimpleLruTruncate(XactCtl, xlrec.pageno);
+ SimpleLruTruncate(SLRU_CLOG_ID, CLOGPagePrecedes, xlrec.pageno);
}
else
elog(PANIC, "clog_redo: unknown op code %u", info);
}
-
-/*
- * Entrypoint for sync.c to sync clog files.
- */
-int
-clogsyncfiletag(const FileTag *ftag, char *path)
-{
- return SlruSyncFileTag(XactCtl, ftag, path);
-}
diff --git a/src/backend/access/transam/commit_ts.c b/src/backend/access/transam/commit_ts.c
index b897fabc702..69f34624b08 100644
--- a/src/backend/access/transam/commit_ts.c
+++ b/src/backend/access/transam/commit_ts.c
@@ -70,13 +70,6 @@ typedef struct CommitTimestampEntry
#define TransactionIdToCTsEntry(xid) \
((xid) % (TransactionId) COMMIT_TS_XACTS_PER_PAGE)
-/*
- * Link to shared-memory data structures for CommitTs control
- */
-static SlruCtlData CommitTsCtlData;
-
-#define CommitTsCtl (&CommitTsCtlData)
-
/*
* We keep a cache of the last value set in shared memory.
*
@@ -107,7 +100,7 @@ static void SetXidCommitTsInPage(TransactionId xid, int nsubxids,
static void TransactionIdSetCommitTs(TransactionId xid, TimestampTz ts,
RepOriginId nodeid, int slotno);
static void error_commit_ts_disabled(void);
-static int ZeroCommitTsPage(int pageno, bool writeXlog);
+static Buffer ZeroCommitTsPage(int pageno, bool writeXlog);
static bool CommitTsPagePrecedes(int page1, int page2);
static void ActivateCommitTs(void);
static void DeactivateCommitTs(void);
@@ -216,30 +209,27 @@ SetXidCommitTsInPage(TransactionId xid, int nsubxids,
TransactionId *subxids, TimestampTz ts,
RepOriginId nodeid, int pageno)
{
- int slotno;
int i;
+ Buffer buffer;
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
-
- slotno = SimpleLruReadPage(CommitTsCtl, pageno, true, xid);
+ buffer = ReadSlruBuffer(SLRU_COMMIT_TS_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
- TransactionIdSetCommitTs(xid, ts, nodeid, slotno);
+ TransactionIdSetCommitTs(xid, ts, nodeid, buffer);
for (i = 0; i < nsubxids; i++)
- TransactionIdSetCommitTs(subxids[i], ts, nodeid, slotno);
+ TransactionIdSetCommitTs(subxids[i], ts, nodeid, buffer);
- CommitTsCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
- LWLockRelease(CommitTsSLRULock);
+ UnlockReleaseBuffer(buffer);
}
/*
* Sets the commit timestamp of a single transaction.
- *
- * Must be called with CommitTsSLRULock held
*/
static void
TransactionIdSetCommitTs(TransactionId xid, TimestampTz ts,
- RepOriginId nodeid, int slotno)
+ RepOriginId nodeid, Buffer buffer)
{
int entryno = TransactionIdToCTsEntry(xid);
CommitTimestampEntry entry;
@@ -249,8 +239,7 @@ TransactionIdSetCommitTs(TransactionId xid, TimestampTz ts,
entry.time = ts;
entry.nodeid = nodeid;
- memcpy(CommitTsCtl->shared->page_buffer[slotno] +
- SizeOfCommitTimestampEntry * entryno,
+ memcpy(BufferGetPage(buffer) + SizeOfCommitTimestampEntry * entryno,
&entry, SizeOfCommitTimestampEntry);
}
@@ -268,10 +257,10 @@ TransactionIdGetCommitTsData(TransactionId xid, TimestampTz *ts,
{
int pageno = TransactionIdToCTsPage(xid);
int entryno = TransactionIdToCTsEntry(xid);
- int slotno;
CommitTimestampEntry entry;
TransactionId oldestCommitTsXid;
TransactionId newestCommitTsXid;
+ Buffer buffer;
if (!TransactionIdIsValid(xid))
ereport(ERROR,
@@ -325,10 +314,11 @@ TransactionIdGetCommitTsData(TransactionId xid, TimestampTz *ts,
return false;
}
- /* lock is acquired by SimpleLruReadPage_ReadOnly */
- slotno = SimpleLruReadPage_ReadOnly(CommitTsCtl, pageno, xid);
+ buffer = ReadSlruBuffer(SLRU_COMMIT_TS_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_SHARE);
+
memcpy(&entry,
- CommitTsCtl->shared->page_buffer[slotno] +
+ BufferGetPage(buffer) +
SizeOfCommitTimestampEntry * entryno,
SizeOfCommitTimestampEntry);
@@ -336,7 +326,7 @@ TransactionIdGetCommitTsData(TransactionId xid, TimestampTz *ts,
if (nodeid)
*nodeid = entry.nodeid;
- LWLockRelease(CommitTsSLRULock);
+ UnlockReleaseBuffer(buffer);
return *ts != 0;
}
@@ -487,27 +477,13 @@ pg_xact_commit_timestamp_origin(PG_FUNCTION_ARGS)
PG_RETURN_DATUM(HeapTupleGetDatum(htup));
}
-/*
- * Number of shared CommitTS buffers.
- *
- * We use a very similar logic as for the number of CLOG buffers (except we
- * scale up twice as fast with shared buffers, and the maximum is twice as
- * high); see comments in CLOGShmemBuffers.
- */
-Size
-CommitTsShmemBuffers(void)
-{
- return Min(256, Max(4, NBuffers / 256));
-}
-
/*
* Shared memory sizing for CommitTs
*/
Size
CommitTsShmemSize(void)
{
- return SimpleLruShmemSize(CommitTsShmemBuffers(), 0) +
- sizeof(CommitTimestampShared);
+ return sizeof(CommitTimestampShared);
}
/*
@@ -519,12 +495,7 @@ CommitTsShmemInit(void)
{
bool found;
- CommitTsCtl->PagePrecedes = CommitTsPagePrecedes;
- SimpleLruInit(CommitTsCtl, "CommitTs", CommitTsShmemBuffers(), 0,
- CommitTsSLRULock, "pg_commit_ts",
- LWTRANCHE_COMMITTS_BUFFER,
- SYNC_HANDLER_COMMIT_TS);
- SlruPagePrecedesUnitTests(CommitTsCtl, COMMIT_TS_XACTS_PER_PAGE);
+ SlruPagePrecedesUnitTests(CommitTsPagePrecedes, COMMIT_TS_XACTS_PER_PAGE);
commitTsShared = ShmemInitStruct("CommitTs shared",
sizeof(CommitTimestampShared),
@@ -568,17 +539,18 @@ BootStrapCommitTs(void)
*
* Control lock must be held at entry, and will be held at exit.
*/
-static int
+static Buffer
ZeroCommitTsPage(int pageno, bool writeXlog)
{
- int slotno;
+ Buffer buffer;
- slotno = SimpleLruZeroPage(CommitTsCtl, pageno);
+ buffer = ZeroSlruBuffer(SLRU_COMMIT_TS_ID, pageno);
+ MarkBufferDirty(buffer);
if (writeXlog)
WriteZeroPageXlogRec(pageno);
- return slotno;
+ return buffer;
}
/*
@@ -676,13 +648,6 @@ ActivateCommitTs(void)
xid = XidFromFullTransactionId(ShmemVariableCache->nextXid);
pageno = TransactionIdToCTsPage(xid);
- /*
- * Re-Initialize our idea of the latest page number.
- */
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
- CommitTsCtl->shared->latest_page_number = pageno;
- LWLockRelease(CommitTsSLRULock);
-
/*
* If CommitTs is enabled, but it wasn't in the previous server run, we
* need to set the oldest and newest values to the next Xid; that way, we
@@ -705,15 +670,14 @@ ActivateCommitTs(void)
LWLockRelease(CommitTsLock);
/* Create the current segment file, if necessary */
- if (!SimpleLruDoesPhysicalPageExist(CommitTsCtl, pageno))
+ if (!SimpleLruDoesPhysicalPageExist(SLRU_COMMIT_TS_ID, pageno))
{
- int slotno;
+ Buffer buffer;
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
- slotno = ZeroCommitTsPage(pageno, false);
- SimpleLruWritePage(CommitTsCtl, slotno);
- Assert(!CommitTsCtl->shared->page_dirty[slotno]);
- LWLockRelease(CommitTsSLRULock);
+ buffer = ZeroSlruBuffer(SLRU_COMMIT_TS_ID, pageno);
+ MarkBufferDirty(buffer);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
/* Change the activation status in shared memory. */
@@ -762,23 +726,9 @@ DeactivateCommitTs(void)
* be overwritten anyway when we wrap around, but it seems better to be
* tidy.)
*/
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
- (void) SlruScanDirectory(CommitTsCtl, SlruScanDirCbDeleteAll, NULL);
- LWLockRelease(CommitTsSLRULock);
-}
-
-/*
- * Perform a checkpoint --- either during shutdown, or on-the-fly
- */
-void
-CheckPointCommitTs(void)
-{
- /*
- * Write dirty CommitTs pages to disk. This may result in sync requests
- * queued for later handling by ProcessSyncRequests(), as part of the
- * checkpoint.
- */
- SimpleLruWriteAll(CommitTsCtl, true);
+ (void) SlruScanDirectory(SLRU_COMMIT_TS_ID,
+ CommitTsPagePrecedes,
+ SlruScanDirCbDeleteAll, NULL);
}
/*
@@ -816,12 +766,8 @@ ExtendCommitTs(TransactionId newestXact)
pageno = TransactionIdToCTsPage(newestXact);
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
-
/* Zero the page and make an XLOG entry about it */
- ZeroCommitTsPage(pageno, !InRecovery);
-
- LWLockRelease(CommitTsSLRULock);
+ UnlockReleaseBuffer(ZeroCommitTsPage(pageno, !InRecovery));
}
/*
@@ -842,7 +788,9 @@ TruncateCommitTs(TransactionId oldestXact)
cutoffPage = TransactionIdToCTsPage(oldestXact);
/* Check to see if there's any files that could be removed */
- if (!SlruScanDirectory(CommitTsCtl, SlruScanDirCbReportPresence,
+ if (!SlruScanDirectory(SLRU_COMMIT_TS_ID,
+ CommitTsPagePrecedes,
+ SlruScanDirCbReportPresence,
&cutoffPage))
return; /* nothing to remove */
@@ -850,7 +798,7 @@ TruncateCommitTs(TransactionId oldestXact)
WriteTruncateXlogRec(cutoffPage, oldestXact);
/* Now we can remove the old CommitTs segment(s) */
- SimpleLruTruncate(CommitTsCtl, cutoffPage);
+ SimpleLruTruncate(SLRU_COMMIT_TS_ID, CommitTsPagePrecedes, cutoffPage);
}
/*
@@ -974,17 +922,14 @@ commit_ts_redo(XLogReaderState *record)
if (info == COMMIT_TS_ZEROPAGE)
{
int pageno;
- int slotno;
+ Buffer buffer;
memcpy(&pageno, XLogRecGetData(record), sizeof(int));
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
-
- slotno = ZeroCommitTsPage(pageno, false);
- SimpleLruWritePage(CommitTsCtl, slotno);
- Assert(!CommitTsCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(CommitTsSLRULock);
+ buffer = ZeroSlruBuffer(SLRU_COMMIT_TS_ID, pageno);
+ MarkBufferDirty(buffer);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
else if (info == COMMIT_TS_TRUNCATE)
{
@@ -992,23 +937,8 @@ commit_ts_redo(XLogReaderState *record)
AdvanceOldestCommitTsXid(trunc->oldestXid);
- /*
- * During XLOG replay, latest_page_number isn't set up yet; insert a
- * suitable value to bypass the sanity test in SimpleLruTruncate.
- */
- CommitTsCtl->shared->latest_page_number = trunc->pageno;
-
- SimpleLruTruncate(CommitTsCtl, trunc->pageno);
+ SimpleLruTruncate(SLRU_COMMIT_TS_ID, CommitTsPagePrecedes, trunc->pageno);
}
else
elog(PANIC, "commit_ts_redo: unknown op code %u", info);
}
-
-/*
- * Entrypoint for sync.c to sync commit_ts files.
- */
-int
-committssyncfiletag(const FileTag *ftag, char *path)
-{
- return SlruSyncFileTag(CommitTsCtl, ftag, path);
-}
diff --git a/src/backend/access/transam/multixact.c b/src/backend/access/transam/multixact.c
index fe6698d5ffa..8e06a0e9a91 100644
--- a/src/backend/access/transam/multixact.c
+++ b/src/backend/access/transam/multixact.c
@@ -181,15 +181,6 @@
#define PreviousMultiXactId(xid) \
((xid) == FirstMultiXactId ? MaxMultiXactId : (xid) - 1)
-/*
- * Links to shared-memory data structures for MultiXact control
- */
-static SlruCtlData MultiXactOffsetCtlData;
-static SlruCtlData MultiXactMemberCtlData;
-
-#define MultiXactOffsetCtl (&MultiXactOffsetCtlData)
-#define MultiXactMemberCtl (&MultiXactMemberCtlData)
-
/*
* MultiXact state shared across all backends. All this state is protected
* by MultiXactGenLock. (We also use MultiXactOffsetSLRULock and
@@ -354,10 +345,9 @@ static void mXactCachePut(MultiXactId multi, int nmembers,
static char *mxstatus_to_string(MultiXactStatus status);
/* management of SLRU infrastructure */
-static int ZeroMultiXactOffsetPage(int pageno, bool writeXlog);
-static int ZeroMultiXactMemberPage(int pageno, bool writeXlog);
+static Buffer ZeroMultiXactOffsetPage(int pageno, bool writeXlog);
+static Buffer ZeroMultiXactMemberPage(int pageno, bool writeXlog);
static bool MultiXactOffsetPagePrecedes(int page1, int page2);
-static bool MultiXactMemberPagePrecedes(int page1, int page2);
static bool MultiXactOffsetPrecedes(MultiXactOffset offset1,
MultiXactOffset offset2);
static void ExtendMultiXactOffset(MultiXactId multi);
@@ -867,34 +857,25 @@ RecordNewMultiXact(MultiXactId multi, MultiXactOffset offset,
int pageno;
int prev_pageno;
int entryno;
- int slotno;
MultiXactOffset *offptr;
int i;
-
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
+ Buffer buffer;
pageno = MultiXactIdToOffsetPage(multi);
entryno = MultiXactIdToOffsetEntry(multi);
- /*
- * Note: we pass the MultiXactId to SimpleLruReadPage as the "transaction"
- * to complain about if there's any I/O error. This is kinda bogus, but
- * since the errors will always give the full pathname, it should be clear
- * enough that a MultiXactId is really involved. Perhaps someday we'll
- * take the trouble to generalize the slru.c error reporting code.
- */
- slotno = SimpleLruReadPage(MultiXactOffsetCtl, pageno, true, multi);
- offptr = (MultiXactOffset *) MultiXactOffsetCtl->shared->page_buffer[slotno];
+ /* XXX set up error context? */
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_OFFSET_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ offptr = (MultiXactOffset *) BufferGetPage(buffer);
offptr += entryno;
*offptr = offset;
- MultiXactOffsetCtl->shared->page_dirty[slotno] = true;
-
- /* Exchange our lock */
- LWLockRelease(MultiXactOffsetSLRULock);
+ MarkBufferDirty(buffer);
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
+ UnlockReleaseBuffer(buffer);
+ buffer = InvalidBuffer;
prev_pageno = -1;
@@ -916,27 +897,28 @@ RecordNewMultiXact(MultiXactId multi, MultiXactOffset offset,
if (pageno != prev_pageno)
{
- slotno = SimpleLruReadPage(MultiXactMemberCtl, pageno, true, multi);
+ if (BufferIsValid(buffer))
+ UnlockReleaseBuffer(buffer);
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_MEMBER_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
prev_pageno = pageno;
}
- memberptr = (TransactionId *)
- (MultiXactMemberCtl->shared->page_buffer[slotno] + memberoff);
+ memberptr = (TransactionId *) (BufferGetPage(buffer) + memberoff);
*memberptr = members[i].xid;
- flagsptr = (uint32 *)
- (MultiXactMemberCtl->shared->page_buffer[slotno] + flagsoff);
+ flagsptr = (uint32 *) (BufferGetPage(buffer) + flagsoff);
flagsval = *flagsptr;
flagsval &= ~(((1 << MXACT_MEMBER_BITS_PER_XACT) - 1) << bshift);
flagsval |= (members[i].status << bshift);
*flagsptr = flagsval;
- MultiXactMemberCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
}
- LWLockRelease(MultiXactMemberSLRULock);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -1228,7 +1210,6 @@ GetMultiXactIdMembers(MultiXactId multi, MultiXactMember **members,
int pageno;
int prev_pageno;
int entryno;
- int slotno;
MultiXactOffset *offptr;
MultiXactOffset offset;
int length;
@@ -1239,6 +1220,7 @@ GetMultiXactIdMembers(MultiXactId multi, MultiXactMember **members,
MultiXactId tmpMXact;
MultiXactOffset nextOffset;
MultiXactMember *ptr;
+ Buffer buffer;
debug_elog3(DEBUG2, "GetMembers: asked for %u", multi);
@@ -1342,13 +1324,12 @@ GetMultiXactIdMembers(MultiXactId multi, MultiXactMember **members,
* time on every multixact creation.
*/
retry:
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
-
pageno = MultiXactIdToOffsetPage(multi);
entryno = MultiXactIdToOffsetEntry(multi);
- slotno = SimpleLruReadPage(MultiXactOffsetCtl, pageno, true, multi);
- offptr = (MultiXactOffset *) MultiXactOffsetCtl->shared->page_buffer[slotno];
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_OFFSET_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_SHARE);
+ offptr = (MultiXactOffset *) BufferGetPage(buffer);
offptr += entryno;
offset = *offptr;
@@ -1379,16 +1360,20 @@ retry:
entryno = MultiXactIdToOffsetEntry(tmpMXact);
if (pageno != prev_pageno)
- slotno = SimpleLruReadPage(MultiXactOffsetCtl, pageno, true, tmpMXact);
+ {
+ UnlockReleaseBuffer(buffer);
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_OFFSET_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_SHARE);
+ }
- offptr = (MultiXactOffset *) MultiXactOffsetCtl->shared->page_buffer[slotno];
+ offptr = (MultiXactOffset *) BufferGetPage(buffer);
offptr += entryno;
nextMXOffset = *offptr;
if (nextMXOffset == 0)
{
/* Corner case 2: next multixact is still being filled in */
- LWLockRelease(MultiXactOffsetSLRULock);
+ UnlockReleaseBuffer(buffer);
CHECK_FOR_INTERRUPTS();
pg_usleep(1000L);
goto retry;
@@ -1396,14 +1381,11 @@ retry:
length = nextMXOffset - offset;
}
-
- LWLockRelease(MultiXactOffsetSLRULock);
+ UnlockReleaseBuffer(buffer);
+ buffer = InvalidBuffer;
ptr = (MultiXactMember *) palloc(length * sizeof(MultiXactMember));
- /* Now get the members themselves. */
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
-
truelength = 0;
prev_pageno = -1;
for (i = 0; i < length; i++, offset++)
@@ -1419,12 +1401,14 @@ retry:
if (pageno != prev_pageno)
{
- slotno = SimpleLruReadPage(MultiXactMemberCtl, pageno, true, multi);
+ if (BufferIsValid(buffer))
+ UnlockReleaseBuffer(buffer);
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_MEMBER_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_SHARE);
prev_pageno = pageno;
}
- xactptr = (TransactionId *)
- (MultiXactMemberCtl->shared->page_buffer[slotno] + memberoff);
+ xactptr = (TransactionId *) (BufferGetPage(buffer) + memberoff);
if (!TransactionIdIsValid(*xactptr))
{
@@ -1435,14 +1419,13 @@ retry:
flagsoff = MXOffsetToFlagsOffset(offset);
bshift = MXOffsetToFlagsBitShift(offset);
- flagsptr = (uint32 *) (MultiXactMemberCtl->shared->page_buffer[slotno] + flagsoff);
+ flagsptr = (uint32 *) (BufferGetPage(buffer) + flagsoff);
ptr[truelength].xid = *xactptr;
ptr[truelength].status = (*flagsptr >> bshift) & MXACT_MEMBER_XACT_BITMASK;
truelength++;
}
-
- LWLockRelease(MultiXactMemberSLRULock);
+ UnlockReleaseBuffer(buffer);
/* A multixid with zero members should not happen */
Assert(truelength > 0);
@@ -1834,8 +1817,6 @@ MultiXactShmemSize(void)
mul_size(sizeof(MultiXactId) * 2, MaxOldestSlot))
size = SHARED_MULTIXACT_STATE_SIZE;
- size = add_size(size, SimpleLruShmemSize(NUM_MULTIXACTOFFSET_BUFFERS, 0));
- size = add_size(size, SimpleLruShmemSize(NUM_MULTIXACTMEMBER_BUFFERS, 0));
return size;
}
@@ -1847,22 +1828,6 @@ MultiXactShmemInit(void)
debug_elog2(DEBUG2, "Shared Memory Init for MultiXact");
- MultiXactOffsetCtl->PagePrecedes = MultiXactOffsetPagePrecedes;
- MultiXactMemberCtl->PagePrecedes = MultiXactMemberPagePrecedes;
-
- SimpleLruInit(MultiXactOffsetCtl,
- "MultiXactOffset", NUM_MULTIXACTOFFSET_BUFFERS, 0,
- MultiXactOffsetSLRULock, "pg_multixact/offsets",
- LWTRANCHE_MULTIXACTOFFSET_BUFFER,
- SYNC_HANDLER_MULTIXACT_OFFSET);
- SlruPagePrecedesUnitTests(MultiXactOffsetCtl, MULTIXACT_OFFSETS_PER_PAGE);
- SimpleLruInit(MultiXactMemberCtl,
- "MultiXactMember", NUM_MULTIXACTMEMBER_BUFFERS, 0,
- MultiXactMemberSLRULock, "pg_multixact/members",
- LWTRANCHE_MULTIXACTMEMBER_BUFFER,
- SYNC_HANDLER_MULTIXACT_MEMBER);
- /* doesn't call SimpleLruTruncate() or meet criteria for unit tests */
-
/* Initialize our shared state struct */
MultiXactState = ShmemInitStruct("Shared MultiXact State",
SHARED_MULTIXACT_STATE_SIZE,
@@ -1893,29 +1858,17 @@ MultiXactShmemInit(void)
void
BootStrapMultiXact(void)
{
- int slotno;
-
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
+ Buffer buffer;
/* Create and zero the first page of the offsets log */
- slotno = ZeroMultiXactOffsetPage(0, false);
-
- /* Make sure it's written out */
- SimpleLruWritePage(MultiXactOffsetCtl, slotno);
- Assert(!MultiXactOffsetCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(MultiXactOffsetSLRULock);
-
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
+ buffer = ZeroMultiXactOffsetPage(0, false);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
/* Create and zero the first page of the members log */
- slotno = ZeroMultiXactMemberPage(0, false);
-
- /* Make sure it's written out */
- SimpleLruWritePage(MultiXactMemberCtl, slotno);
- Assert(!MultiXactMemberCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(MultiXactMemberSLRULock);
+ buffer = ZeroMultiXactMemberPage(0, false);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -1927,33 +1880,35 @@ BootStrapMultiXact(void)
*
* Control lock must be held at entry, and will be held at exit.
*/
-static int
+static Buffer
ZeroMultiXactOffsetPage(int pageno, bool writeXlog)
{
- int slotno;
+ Buffer buffer;
- slotno = SimpleLruZeroPage(MultiXactOffsetCtl, pageno);
+ buffer = ZeroSlruBuffer(SLRU_MULTIXACT_OFFSET_ID, pageno);
+ MarkBufferDirty(buffer);
if (writeXlog)
WriteMZeroPageXlogRec(pageno, XLOG_MULTIXACT_ZERO_OFF_PAGE);
- return slotno;
+ return buffer;
}
/*
* Ditto, for MultiXactMember
*/
-static int
+static Buffer
ZeroMultiXactMemberPage(int pageno, bool writeXlog)
{
- int slotno;
+ Buffer buffer;
- slotno = SimpleLruZeroPage(MultiXactMemberCtl, pageno);
+ buffer = ZeroSlruBuffer(SLRU_MULTIXACT_MEMBER_ID, pageno);
+ MarkBufferDirty(buffer);
if (writeXlog)
WriteMZeroPageXlogRec(pageno, XLOG_MULTIXACT_ZERO_MEM_PAGE);
- return slotno;
+ return buffer;
}
/*
@@ -1978,22 +1933,14 @@ MaybeExtendOffsetSlru(void)
pageno = MultiXactIdToOffsetPage(MultiXactState->nextMXact);
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
-
- if (!SimpleLruDoesPhysicalPageExist(MultiXactOffsetCtl, pageno))
+ if (!SimpleLruDoesPhysicalPageExist(SLRU_MULTIXACT_OFFSET_ID, pageno))
{
- int slotno;
+ Buffer buffer;
- /*
- * Fortunately for us, SimpleLruWritePage is already prepared to deal
- * with creating a new segment file even if the page we're writing is
- * not the first in it, so this is enough.
- */
- slotno = ZeroMultiXactOffsetPage(pageno, false);
- SimpleLruWritePage(MultiXactOffsetCtl, slotno);
+ buffer = ZeroMultiXactOffsetPage(pageno, false);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
-
- LWLockRelease(MultiXactOffsetSLRULock);
}
/*
@@ -2007,21 +1954,6 @@ MaybeExtendOffsetSlru(void)
void
StartupMultiXact(void)
{
- MultiXactId multi = MultiXactState->nextMXact;
- MultiXactOffset offset = MultiXactState->nextOffset;
- int pageno;
-
- /*
- * Initialize offset's idea of the latest page number.
- */
- pageno = MultiXactIdToOffsetPage(multi);
- MultiXactOffsetCtl->shared->latest_page_number = pageno;
-
- /*
- * Initialize member's idea of the latest page number.
- */
- pageno = MXOffsetToMemberPage(offset);
- MultiXactMemberCtl->shared->latest_page_number = pageno;
}
/*
@@ -2045,14 +1977,7 @@ TrimMultiXact(void)
oldestMXactDB = MultiXactState->oldestMultiXactDB;
LWLockRelease(MultiXactGenLock);
- /* Clean up offsets state */
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
-
- /*
- * (Re-)Initialize our idea of the latest page number for offsets.
- */
- pageno = MultiXactIdToOffsetPage(nextMXact);
- MultiXactOffsetCtl->shared->latest_page_number = pageno;
+ pageno = MXOffsetToMemberPage(offset);
/*
* Zero out the remainder of the current offsets page. See notes in
@@ -2065,29 +1990,20 @@ TrimMultiXact(void)
entryno = MultiXactIdToOffsetEntry(nextMXact);
if (entryno != 0)
{
- int slotno;
MultiXactOffset *offptr;
+ Buffer buffer;
- slotno = SimpleLruReadPage(MultiXactOffsetCtl, pageno, true, nextMXact);
- offptr = (MultiXactOffset *) MultiXactOffsetCtl->shared->page_buffer[slotno];
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_OFFSET_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ offptr = (MultiXactOffset *) BufferGetPage(buffer);
offptr += entryno;
MemSet(offptr, 0, BLCKSZ - (entryno * sizeof(MultiXactOffset)));
- MultiXactOffsetCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
+ UnlockReleaseBuffer(buffer);
}
- LWLockRelease(MultiXactOffsetSLRULock);
-
- /* And the same for members */
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
-
- /*
- * (Re-)Initialize our idea of the latest page number for members.
- */
- pageno = MXOffsetToMemberPage(offset);
- MultiXactMemberCtl->shared->latest_page_number = pageno;
-
/*
* Zero out the remainder of the current members page. See notes in
* TrimCLOG() for motivation.
@@ -2095,14 +2011,14 @@ TrimMultiXact(void)
flagsoff = MXOffsetToFlagsOffset(offset);
if (flagsoff != 0)
{
- int slotno;
TransactionId *xidptr;
int memberoff;
+ Buffer buffer;
memberoff = MXOffsetToMemberOffset(offset);
- slotno = SimpleLruReadPage(MultiXactMemberCtl, pageno, true, offset);
- xidptr = (TransactionId *)
- (MultiXactMemberCtl->shared->page_buffer[slotno] + memberoff);
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_MEMBER_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ xidptr = (TransactionId *) (BufferGetPage(buffer) + memberoff);
MemSet(xidptr, 0, BLCKSZ - memberoff);
@@ -2112,11 +2028,10 @@ TrimMultiXact(void)
* writing.
*/
- MultiXactMemberCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
+ UnlockReleaseBuffer(buffer);
}
- LWLockRelease(MultiXactMemberSLRULock);
-
/* signal that we're officially up */
LWLockAcquire(MultiXactGenLock, LW_EXCLUSIVE);
MultiXactState->finishedStartup = true;
@@ -2148,25 +2063,6 @@ MultiXactGetCheckptMulti(bool is_shutdown,
*nextMulti, *nextMultiOffset, *oldestMulti, *oldestMultiDB);
}
-/*
- * Perform a checkpoint --- either during shutdown, or on-the-fly
- */
-void
-CheckPointMultiXact(void)
-{
- TRACE_POSTGRESQL_MULTIXACT_CHECKPOINT_START(true);
-
- /*
- * Write dirty MultiXact pages to disk. This may result in sync requests
- * queued for later handling by ProcessSyncRequests(), as part of the
- * checkpoint.
- */
- SimpleLruWriteAll(MultiXactOffsetCtl, true);
- SimpleLruWriteAll(MultiXactMemberCtl, true);
-
- TRACE_POSTGRESQL_MULTIXACT_CHECKPOINT_DONE(true);
-}
-
/*
* Set the next-to-be-assigned MultiXactId and offset
*
@@ -2415,12 +2311,8 @@ ExtendMultiXactOffset(MultiXactId multi)
pageno = MultiXactIdToOffsetPage(multi);
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
-
/* Zero the page and make an XLOG entry about it */
- ZeroMultiXactOffsetPage(pageno, true);
-
- LWLockRelease(MultiXactOffsetSLRULock);
+ UnlockReleaseBuffer(ZeroMultiXactOffsetPage(pageno, true));
}
/*
@@ -2456,12 +2348,8 @@ ExtendMultiXactMember(MultiXactOffset offset, int nmembers)
pageno = MXOffsetToMemberPage(offset);
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
-
/* Zero the page and make an XLOG entry about it */
- ZeroMultiXactMemberPage(pageno, true);
-
- LWLockRelease(MultiXactMemberSLRULock);
+ UnlockReleaseBuffer(ZeroMultiXactMemberPage(pageno, true));
}
/*
@@ -2737,8 +2625,8 @@ find_multixact_start(MultiXactId multi, MultiXactOffset *result)
MultiXactOffset offset;
int pageno;
int entryno;
- int slotno;
MultiXactOffset *offptr;
+ Buffer buffer;
Assert(MultiXactState->finishedStartup);
@@ -2746,20 +2634,19 @@ find_multixact_start(MultiXactId multi, MultiXactOffset *result)
entryno = MultiXactIdToOffsetEntry(multi);
/*
- * Write out dirty data, so PhysicalPageExists can work correctly.
+ * Cope with missing/bogus oldest MultiXact in inconsistent states (see
+ * commit 068cfadf9).
*/
- SimpleLruWriteAll(MultiXactOffsetCtl, true);
- SimpleLruWriteAll(MultiXactMemberCtl, true);
-
- if (!SimpleLruDoesPhysicalPageExist(MultiXactOffsetCtl, pageno))
+ if (!ProbeSlruBuffer(SLRU_MULTIXACT_OFFSET_ID, pageno) &&
+ !SimpleLruDoesPhysicalPageExist(SLRU_MULTIXACT_OFFSET_ID, pageno))
return false;
- /* lock is acquired by SimpleLruReadPage_ReadOnly */
- slotno = SimpleLruReadPage_ReadOnly(MultiXactOffsetCtl, pageno, multi);
- offptr = (MultiXactOffset *) MultiXactOffsetCtl->shared->page_buffer[slotno];
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_OFFSET_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_SHARE);
+ offptr = (MultiXactOffset *) BufferGetPage(buffer);
offptr += entryno;
offset = *offptr;
- LWLockRelease(MultiXactOffsetSLRULock);
+ UnlockReleaseBuffer(buffer);
*result = offset;
return true;
@@ -2862,12 +2749,13 @@ typedef struct mxtruncinfo
* This callback determines the earliest existing page number.
*/
static bool
-SlruScanDirCbFindEarliest(SlruCtl ctl, char *filename, int segpage, void *data)
+SlruScanDirCbFindEarliest(int slru_id, SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage, void *data)
{
mxtruncinfo *trunc = (mxtruncinfo *) data;
if (trunc->earliestExistingPage == -1 ||
- ctl->PagePrecedes(segpage, trunc->earliestExistingPage))
+ PagePrecedes(segpage, trunc->earliestExistingPage))
{
trunc->earliestExistingPage = segpage;
}
@@ -2899,7 +2787,7 @@ PerformMembersTruncation(MultiXactOffset oldestOffset, MultiXactOffset newOldest
while (segment != endsegment)
{
elog(DEBUG2, "truncating multixact members segment %x", segment);
- SlruDeleteSegment(MultiXactMemberCtl, segment);
+ SlruDeleteSegment(SLRU_MULTIXACT_MEMBER_ID, segment);
/* move to next segment, handling wraparound correctly */
if (segment == maxsegment)
@@ -2922,7 +2810,8 @@ PerformOffsetsTruncation(MultiXactId oldestMulti, MultiXactId newOldestMulti)
* didn't subtract one, we'd trigger SimpleLruTruncate's wraparound
* detection.
*/
- SimpleLruTruncate(MultiXactOffsetCtl,
+ SimpleLruTruncate(SLRU_MULTIXACT_OFFSET_ID,
+ MultiXactOffsetPagePrecedes,
MultiXactIdToOffsetPage(PreviousMultiXactId(newOldestMulti)));
}
@@ -2996,7 +2885,9 @@ TruncateMultiXact(MultiXactId newOldestMulti, Oid newOldestMultiDB)
* been truncated away, and we crashed before updating oldestMulti.
*/
trunc.earliestExistingPage = -1;
- SlruScanDirectory(MultiXactOffsetCtl, SlruScanDirCbFindEarliest, &trunc);
+ SlruScanDirectory(SLRU_MULTIXACT_OFFSET_ID,
+ MultiXactOffsetPagePrecedes,
+ SlruScanDirCbFindEarliest, &trunc);
earliest = trunc.earliestExistingPage * MULTIXACT_OFFSETS_PER_PAGE;
if (earliest < FirstMultiXactId)
earliest = FirstMultiXactId;
@@ -3128,24 +3019,6 @@ MultiXactOffsetPagePrecedes(int page1, int page2)
multi2 + MULTIXACT_OFFSETS_PER_PAGE - 1));
}
-/*
- * Decide whether a MultiXactMember page number is "older" for truncation
- * purposes. There is no "invalid offset number" so use the numbers verbatim.
- */
-static bool
-MultiXactMemberPagePrecedes(int page1, int page2)
-{
- MultiXactOffset offset1;
- MultiXactOffset offset2;
-
- offset1 = ((MultiXactOffset) page1) * MULTIXACT_MEMBERS_PER_PAGE;
- offset2 = ((MultiXactOffset) page2) * MULTIXACT_MEMBERS_PER_PAGE;
-
- return (MultiXactOffsetPrecedes(offset1, offset2) &&
- MultiXactOffsetPrecedes(offset1,
- offset2 + MULTIXACT_MEMBERS_PER_PAGE - 1));
-}
-
/*
* Decide which of two MultiXactIds is earlier.
*
@@ -3240,32 +3113,18 @@ multixact_redo(XLogReaderState *record)
if (info == XLOG_MULTIXACT_ZERO_OFF_PAGE)
{
int pageno;
- int slotno;
memcpy(&pageno, XLogRecGetData(record), sizeof(int));
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
-
- slotno = ZeroMultiXactOffsetPage(pageno, false);
- SimpleLruWritePage(MultiXactOffsetCtl, slotno);
- Assert(!MultiXactOffsetCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(MultiXactOffsetSLRULock);
+ UnlockReleaseBuffer(ZeroMultiXactOffsetPage(pageno, false));
}
else if (info == XLOG_MULTIXACT_ZERO_MEM_PAGE)
{
int pageno;
- int slotno;
memcpy(&pageno, XLogRecGetData(record), sizeof(int));
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
-
- slotno = ZeroMultiXactMemberPage(pageno, false);
- SimpleLruWritePage(MultiXactMemberCtl, slotno);
- Assert(!MultiXactMemberCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(MultiXactMemberSLRULock);
+ UnlockReleaseBuffer(ZeroMultiXactMemberPage(pageno, false));
}
else if (info == XLOG_MULTIXACT_CREATE_ID)
{
@@ -3299,7 +3158,6 @@ multixact_redo(XLogReaderState *record)
else if (info == XLOG_MULTIXACT_TRUNCATE_ID)
{
xl_multixact_truncate xlrec;
- int pageno;
memcpy(&xlrec, XLogRecGetData(record),
SizeOfMultiXactTruncate);
@@ -3325,13 +3183,6 @@ multixact_redo(XLogReaderState *record)
PerformMembersTruncation(xlrec.startTruncMemb, xlrec.endTruncMemb);
- /*
- * During XLOG replay, latest_page_number isn't necessarily set up
- * yet; insert a suitable value to bypass the sanity test in
- * SimpleLruTruncate.
- */
- pageno = MultiXactIdToOffsetPage(xlrec.endTruncOff);
- MultiXactOffsetCtl->shared->latest_page_number = pageno;
PerformOffsetsTruncation(xlrec.startTruncOff, xlrec.endTruncOff);
LWLockRelease(MultiXactTruncationLock);
@@ -3401,21 +3252,3 @@ pg_get_multixact_members(PG_FUNCTION_ARGS)
SRF_RETURN_DONE(funccxt);
}
-
-/*
- * Entrypoint for sync.c to sync offsets files.
- */
-int
-multixactoffsetssyncfiletag(const FileTag *ftag, char *path)
-{
- return SlruSyncFileTag(MultiXactOffsetCtl, ftag, path);
-}
-
-/*
- * Entrypoint for sync.c to sync members files.
- */
-int
-multixactmemberssyncfiletag(const FileTag *ftag, char *path)
-{
- return SlruSyncFileTag(MultiXactMemberCtl, ftag, path);
-}
diff --git a/src/backend/access/transam/slru.c b/src/backend/access/transam/slru.c
index 5ab86238a92..1204468c039 100644
--- a/src/backend/access/transam/slru.c
+++ b/src/backend/access/transam/slru.c
@@ -1,41 +1,9 @@
/*-------------------------------------------------------------------------
*
* slru.c
- * Simple LRU buffering for transaction status logfiles
+ * Simple buffering for transaction status logfiles
*
- * We use a simple least-recently-used scheme to manage a pool of page
- * buffers. Under ordinary circumstances we expect that write
- * traffic will occur mostly to the latest page (and to the just-prior
- * page, soon after a page transition). Read traffic will probably touch
- * a larger span of pages, but in any case a fairly small number of page
- * buffers should be sufficient. So, we just search the buffers using plain
- * linear search; there's no need for a hashtable or anything fancy.
- * The management algorithm is straight LRU except that we will never swap
- * out the latest page (since we know it's going to be hit again eventually).
- *
- * We use a control LWLock to protect the shared data structures, plus
- * per-buffer LWLocks that synchronize I/O for each buffer. The control lock
- * must be held to examine or modify any shared state. A process that is
- * reading in or writing out a page buffer does not hold the control lock,
- * only the per-buffer lock for the buffer it is working on.
- *
- * "Holding the control lock" means exclusive lock in all cases except for
- * SimpleLruReadPage_ReadOnly(); see comments for SlruRecentlyUsed() for
- * the implications of that.
- *
- * When initiating I/O on a buffer, we acquire the per-buffer lock exclusively
- * before releasing the control lock. The per-buffer lock is released after
- * completing the I/O, re-acquiring the control lock, and updating the shared
- * state. (Deadlock is not possible here, because we never try to initiate
- * I/O when someone else is already doing I/O on the same buffer.)
- * To wait for I/O to complete, release the control lock, acquire the
- * per-buffer lock in shared mode, immediately release the per-buffer lock,
- * reacquire the control lock, and then recheck state (since arbitrary things
- * could have happened while we didn't have the lock).
- *
- * As with the regular buffer manager, it is possible for another process
- * to re-dirty a page that is currently being written out. This is handled
- * by re-setting the page's page_dirty flag.
+ * XXX write me
*
*
* Portions Copyright (c) 1996-2023, PostgreSQL Global Development Group
@@ -60,562 +28,31 @@
#include "storage/fd.h"
#include "storage/shmem.h"
-#define SlruFileName(ctl, path, seg) \
- snprintf(path, MAXPGPATH, "%s/%04X", (ctl)->Dir, seg)
-
-/*
- * During SimpleLruWriteAll(), we will usually not need to write more than one
- * or two physical files, but we may need to write several pages per file. We
- * can consolidate the I/O requests by leaving files open until control returns
- * to SimpleLruWriteAll(). This data structure remembers which files are open.
- */
-#define MAX_WRITEALL_BUFFERS 16
+#define PG_SLRU(symname,name,path,synchronize) \
+ path,
-typedef struct SlruWriteAllData
+static char *slru_dirs[] =
{
- int num_files; /* # files actually open */
- int fd[MAX_WRITEALL_BUFFERS]; /* their FD's */
- int segno[MAX_WRITEALL_BUFFERS]; /* their log seg#s */
-} SlruWriteAllData;
-
-typedef struct SlruWriteAllData *SlruWriteAll;
-
-/*
- * Populate a file tag describing a segment file. We only use the segment
- * number, since we can derive everything else we need by having separate
- * sync handler functions for clog, multixact etc.
- */
-#define INIT_SLRUFILETAG(a,xx_handler,xx_segno) \
-( \
- memset(&(a), 0, sizeof(FileTag)), \
- (a).handler = (xx_handler), \
- (a).segno = (xx_segno) \
-)
+#include "access/slrulist.h"
+};
/*
- * Macro to mark a buffer slot "most recently used". Note multiple evaluation
- * of arguments!
- *
- * The reason for the if-test is that there are often many consecutive
- * accesses to the same page (particularly the latest page). By suppressing
- * useless increments of cur_lru_count, we reduce the probability that old
- * pages' counts will "wrap around" and make them appear recently used.
- *
- * We allow this code to be executed concurrently by multiple processes within
- * SimpleLruReadPage_ReadOnly(). As long as int reads and writes are atomic,
- * this should not cause any completely-bogus values to enter the computation.
- * However, it is possible for either cur_lru_count or individual
- * page_lru_count entries to be "reset" to lower values than they should have,
- * in case a process is delayed while it executes this macro. With care in
- * SlruSelectLRUPage(), this does little harm, and in any case the absolute
- * worst possible consequence is a nonoptimal choice of page to evict. The
- * gain from allowing concurrent reads of SLRU pages seems worth it.
- */
-#define SlruRecentlyUsed(shared, slotno) \
- do { \
- int new_lru_count = (shared)->cur_lru_count; \
- if (new_lru_count != (shared)->page_lru_count[slotno]) { \
- (shared)->cur_lru_count = ++new_lru_count; \
- (shared)->page_lru_count[slotno] = new_lru_count; \
- } \
- } while (0)
-
-/* Saved info for SlruReportIOError */
-typedef enum
-{
- SLRU_OPEN_FAILED,
- SLRU_SEEK_FAILED,
- SLRU_READ_FAILED,
- SLRU_WRITE_FAILED,
- SLRU_FSYNC_FAILED,
- SLRU_CLOSE_FAILED
-} SlruErrorCause;
-
-static SlruErrorCause slru_errcause;
-static int slru_errno;
-
-
-static void SimpleLruZeroLSNs(SlruCtl ctl, int slotno);
-static void SimpleLruWaitIO(SlruCtl ctl, int slotno);
-static void SlruInternalWritePage(SlruCtl ctl, int slotno, SlruWriteAll fdata);
-static bool SlruPhysicalReadPage(SlruCtl ctl, int pageno, int slotno);
-static bool SlruPhysicalWritePage(SlruCtl ctl, int pageno, int slotno,
- SlruWriteAll fdata);
-static void SlruReportIOError(SlruCtl ctl, int pageno, TransactionId xid);
-static int SlruSelectLRUPage(SlruCtl ctl, int pageno);
-
-static bool SlruScanDirCbDeleteCutoff(SlruCtl ctl, char *filename,
+ * We'll maintain a little cache of recently seen buffers, to try to avoid the
+ * buffer mapping table on repeat access (ie the busy end of the CLOG). One
+ * entry per SLRU.
+ */
+struct SlruRecentBuffer {
+ int pageno;
+ Buffer recent_buffer;
+};
+
+static struct SlruRecentBuffer slru_recent_buffers[SLRU_NEXT_ID];
+
+static bool SlruScanDirCbDeleteCutoff(int slru_id,
+ SlruPagePrecedesFunction PagePrecedes,
+ char *filename,
int segpage, void *data);
-static void SlruInternalDeleteSegment(SlruCtl ctl, int segno);
-
-/*
- * Initialization of shared memory
- */
-
-Size
-SimpleLruShmemSize(int nslots, int nlsns)
-{
- Size sz;
-
- /* we assume nslots isn't so large as to risk overflow */
- sz = MAXALIGN(sizeof(SlruSharedData));
- sz += MAXALIGN(nslots * sizeof(char *)); /* page_buffer[] */
- sz += MAXALIGN(nslots * sizeof(SlruPageStatus)); /* page_status[] */
- sz += MAXALIGN(nslots * sizeof(bool)); /* page_dirty[] */
- sz += MAXALIGN(nslots * sizeof(int)); /* page_number[] */
- sz += MAXALIGN(nslots * sizeof(int)); /* page_lru_count[] */
- sz += MAXALIGN(nslots * sizeof(LWLockPadded)); /* buffer_locks[] */
-
- if (nlsns > 0)
- sz += MAXALIGN(nslots * nlsns * sizeof(XLogRecPtr)); /* group_lsn[] */
-
- return BUFFERALIGN(sz) + BLCKSZ * nslots;
-}
-
-/*
- * Initialize, or attach to, a simple LRU cache in shared memory.
- *
- * ctl: address of local (unshared) control structure.
- * name: name of SLRU. (This is user-visible, pick with care!)
- * nslots: number of page slots to use.
- * nlsns: number of LSN groups per page (set to zero if not relevant).
- * ctllock: LWLock to use to control access to the shared control structure.
- * subdir: PGDATA-relative subdirectory that will contain the files.
- * tranche_id: LWLock tranche ID to use for the SLRU's per-buffer LWLocks.
- * sync_handler: which set of functions to use to handle sync requests
- */
-void
-SimpleLruInit(SlruCtl ctl, const char *name, int nslots, int nlsns,
- LWLock *ctllock, const char *subdir, int tranche_id,
- SyncRequestHandler sync_handler)
-{
- SlruShared shared;
- bool found;
-
- shared = (SlruShared) ShmemInitStruct(name,
- SimpleLruShmemSize(nslots, nlsns),
- &found);
-
- if (!IsUnderPostmaster)
- {
- /* Initialize locks and shared memory area */
- char *ptr;
- Size offset;
- int slotno;
-
- Assert(!found);
-
- memset(shared, 0, sizeof(SlruSharedData));
-
- shared->ControlLock = ctllock;
-
- shared->num_slots = nslots;
- shared->lsn_groups_per_page = nlsns;
-
- shared->cur_lru_count = 0;
-
- /* shared->latest_page_number will be set later */
-
- shared->slru_stats_idx = pgstat_get_slru_index(name);
-
- ptr = (char *) shared;
- offset = MAXALIGN(sizeof(SlruSharedData));
- shared->page_buffer = (char **) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(char *));
- shared->page_status = (SlruPageStatus *) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(SlruPageStatus));
- shared->page_dirty = (bool *) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(bool));
- shared->page_number = (int *) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(int));
- shared->page_lru_count = (int *) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(int));
-
- /* Initialize LWLocks */
- shared->buffer_locks = (LWLockPadded *) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(LWLockPadded));
-
- if (nlsns > 0)
- {
- shared->group_lsn = (XLogRecPtr *) (ptr + offset);
- offset += MAXALIGN(nslots * nlsns * sizeof(XLogRecPtr));
- }
-
- ptr += BUFFERALIGN(offset);
- for (slotno = 0; slotno < nslots; slotno++)
- {
- LWLockInitialize(&shared->buffer_locks[slotno].lock,
- tranche_id);
-
- shared->page_buffer[slotno] = ptr;
- shared->page_status[slotno] = SLRU_PAGE_EMPTY;
- shared->page_dirty[slotno] = false;
- shared->page_lru_count[slotno] = 0;
- ptr += BLCKSZ;
- }
-
- /* Should fit to estimated shmem size */
- Assert(ptr - (char *) shared <= SimpleLruShmemSize(nslots, nlsns));
- }
- else
- Assert(found);
-
- /*
- * Initialize the unshared control struct, including directory path. We
- * assume caller set PagePrecedes.
- */
- ctl->shared = shared;
- ctl->sync_handler = sync_handler;
- strlcpy(ctl->Dir, subdir, sizeof(ctl->Dir));
-}
-
-/*
- * Initialize (or reinitialize) a page to zeroes.
- *
- * The page is not actually written, just set up in shared memory.
- * The slot number of the new page is returned.
- *
- * Control lock must be held at entry, and will be held at exit.
- */
-int
-SimpleLruZeroPage(SlruCtl ctl, int pageno)
-{
- SlruShared shared = ctl->shared;
- int slotno;
-
- /* Find a suitable buffer slot for the page */
- slotno = SlruSelectLRUPage(ctl, pageno);
- Assert(shared->page_status[slotno] == SLRU_PAGE_EMPTY ||
- (shared->page_status[slotno] == SLRU_PAGE_VALID &&
- !shared->page_dirty[slotno]) ||
- shared->page_number[slotno] == pageno);
-
- /* Mark the slot as containing this page */
- shared->page_number[slotno] = pageno;
- shared->page_status[slotno] = SLRU_PAGE_VALID;
- shared->page_dirty[slotno] = true;
- SlruRecentlyUsed(shared, slotno);
-
- /* Set the buffer to zeroes */
- MemSet(shared->page_buffer[slotno], 0, BLCKSZ);
-
- /* Set the LSNs for this new page to zero */
- SimpleLruZeroLSNs(ctl, slotno);
-
- /* Assume this page is now the latest active page */
- shared->latest_page_number = pageno;
-
- /* update the stats counter of zeroed pages */
- pgstat_count_slru_page_zeroed(shared->slru_stats_idx);
-
- return slotno;
-}
-
-/*
- * Zero all the LSNs we store for this slru page.
- *
- * This should be called each time we create a new page, and each time we read
- * in a page from disk into an existing buffer. (Such an old page cannot
- * have any interesting LSNs, since we'd have flushed them before writing
- * the page in the first place.)
- *
- * This assumes that InvalidXLogRecPtr is bitwise-all-0.
- */
-static void
-SimpleLruZeroLSNs(SlruCtl ctl, int slotno)
-{
- SlruShared shared = ctl->shared;
-
- if (shared->lsn_groups_per_page > 0)
- MemSet(&shared->group_lsn[slotno * shared->lsn_groups_per_page], 0,
- shared->lsn_groups_per_page * sizeof(XLogRecPtr));
-}
-
-/*
- * Wait for any active I/O on a page slot to finish. (This does not
- * guarantee that new I/O hasn't been started before we return, though.
- * In fact the slot might not even contain the same page anymore.)
- *
- * Control lock must be held at entry, and will be held at exit.
- */
-static void
-SimpleLruWaitIO(SlruCtl ctl, int slotno)
-{
- SlruShared shared = ctl->shared;
-
- /* See notes at top of file */
- LWLockRelease(shared->ControlLock);
- LWLockAcquire(&shared->buffer_locks[slotno].lock, LW_SHARED);
- LWLockRelease(&shared->buffer_locks[slotno].lock);
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
- /*
- * If the slot is still in an io-in-progress state, then either someone
- * already started a new I/O on the slot, or a previous I/O failed and
- * neglected to reset the page state. That shouldn't happen, really, but
- * it seems worth a few extra cycles to check and recover from it. We can
- * cheaply test for failure by seeing if the buffer lock is still held (we
- * assume that transaction abort would release the lock).
- */
- if (shared->page_status[slotno] == SLRU_PAGE_READ_IN_PROGRESS ||
- shared->page_status[slotno] == SLRU_PAGE_WRITE_IN_PROGRESS)
- {
- if (LWLockConditionalAcquire(&shared->buffer_locks[slotno].lock, LW_SHARED))
- {
- /* indeed, the I/O must have failed */
- if (shared->page_status[slotno] == SLRU_PAGE_READ_IN_PROGRESS)
- shared->page_status[slotno] = SLRU_PAGE_EMPTY;
- else /* write_in_progress */
- {
- shared->page_status[slotno] = SLRU_PAGE_VALID;
- shared->page_dirty[slotno] = true;
- }
- LWLockRelease(&shared->buffer_locks[slotno].lock);
- }
- }
-}
-
-/*
- * Find a page in a shared buffer, reading it in if necessary.
- * The page number must correspond to an already-initialized page.
- *
- * If write_ok is true then it is OK to return a page that is in
- * WRITE_IN_PROGRESS state; it is the caller's responsibility to be sure
- * that modification of the page is safe. If write_ok is false then we
- * will not return the page until it is not undergoing active I/O.
- *
- * The passed-in xid is used only for error reporting, and may be
- * InvalidTransactionId if no specific xid is associated with the action.
- *
- * Return value is the shared-buffer slot number now holding the page.
- * The buffer's LRU access info is updated.
- *
- * Control lock must be held at entry, and will be held at exit.
- */
-int
-SimpleLruReadPage(SlruCtl ctl, int pageno, bool write_ok,
- TransactionId xid)
-{
- SlruShared shared = ctl->shared;
-
- /* Outer loop handles restart if we must wait for someone else's I/O */
- for (;;)
- {
- int slotno;
- bool ok;
-
- /* See if page already is in memory; if not, pick victim slot */
- slotno = SlruSelectLRUPage(ctl, pageno);
-
- /* Did we find the page in memory? */
- if (shared->page_number[slotno] == pageno &&
- shared->page_status[slotno] != SLRU_PAGE_EMPTY)
- {
- /*
- * If page is still being read in, we must wait for I/O. Likewise
- * if the page is being written and the caller said that's not OK.
- */
- if (shared->page_status[slotno] == SLRU_PAGE_READ_IN_PROGRESS ||
- (shared->page_status[slotno] == SLRU_PAGE_WRITE_IN_PROGRESS &&
- !write_ok))
- {
- SimpleLruWaitIO(ctl, slotno);
- /* Now we must recheck state from the top */
- continue;
- }
- /* Otherwise, it's ready to use */
- SlruRecentlyUsed(shared, slotno);
-
- /* update the stats counter of pages found in the SLRU */
- pgstat_count_slru_page_hit(shared->slru_stats_idx);
-
- return slotno;
- }
-
- /* We found no match; assert we selected a freeable slot */
- Assert(shared->page_status[slotno] == SLRU_PAGE_EMPTY ||
- (shared->page_status[slotno] == SLRU_PAGE_VALID &&
- !shared->page_dirty[slotno]));
-
- /* Mark the slot read-busy */
- shared->page_number[slotno] = pageno;
- shared->page_status[slotno] = SLRU_PAGE_READ_IN_PROGRESS;
- shared->page_dirty[slotno] = false;
-
- /* Acquire per-buffer lock (cannot deadlock, see notes at top) */
- LWLockAcquire(&shared->buffer_locks[slotno].lock, LW_EXCLUSIVE);
-
- /* Release control lock while doing I/O */
- LWLockRelease(shared->ControlLock);
-
- /* Do the read */
- ok = SlruPhysicalReadPage(ctl, pageno, slotno);
-
- /* Set the LSNs for this newly read-in page to zero */
- SimpleLruZeroLSNs(ctl, slotno);
-
- /* Re-acquire control lock and update page state */
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
- Assert(shared->page_number[slotno] == pageno &&
- shared->page_status[slotno] == SLRU_PAGE_READ_IN_PROGRESS &&
- !shared->page_dirty[slotno]);
-
- shared->page_status[slotno] = ok ? SLRU_PAGE_VALID : SLRU_PAGE_EMPTY;
-
- LWLockRelease(&shared->buffer_locks[slotno].lock);
-
- /* Now it's okay to ereport if we failed */
- if (!ok)
- SlruReportIOError(ctl, pageno, xid);
-
- SlruRecentlyUsed(shared, slotno);
-
- /* update the stats counter of pages not found in SLRU */
- pgstat_count_slru_page_read(shared->slru_stats_idx);
-
- return slotno;
- }
-}
-
-/*
- * Find a page in a shared buffer, reading it in if necessary.
- * The page number must correspond to an already-initialized page.
- * The caller must intend only read-only access to the page.
- *
- * The passed-in xid is used only for error reporting, and may be
- * InvalidTransactionId if no specific xid is associated with the action.
- *
- * Return value is the shared-buffer slot number now holding the page.
- * The buffer's LRU access info is updated.
- *
- * Control lock must NOT be held at entry, but will be held at exit.
- * It is unspecified whether the lock will be shared or exclusive.
- */
-int
-SimpleLruReadPage_ReadOnly(SlruCtl ctl, int pageno, TransactionId xid)
-{
- SlruShared shared = ctl->shared;
- int slotno;
-
- /* Try to find the page while holding only shared lock */
- LWLockAcquire(shared->ControlLock, LW_SHARED);
-
- /* See if page is already in a buffer */
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- if (shared->page_number[slotno] == pageno &&
- shared->page_status[slotno] != SLRU_PAGE_EMPTY &&
- shared->page_status[slotno] != SLRU_PAGE_READ_IN_PROGRESS)
- {
- /* See comments for SlruRecentlyUsed macro */
- SlruRecentlyUsed(shared, slotno);
-
- /* update the stats counter of pages found in the SLRU */
- pgstat_count_slru_page_hit(shared->slru_stats_idx);
-
- return slotno;
- }
- }
-
- /* No luck, so switch to normal exclusive lock and do regular read */
- LWLockRelease(shared->ControlLock);
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
- return SimpleLruReadPage(ctl, pageno, true, xid);
-}
-
-/*
- * Write a page from a shared buffer, if necessary.
- * Does nothing if the specified slot is not dirty.
- *
- * NOTE: only one write attempt is made here. Hence, it is possible that
- * the page is still dirty at exit (if someone else re-dirtied it during
- * the write). However, we *do* attempt a fresh write even if the page
- * is already being written; this is for checkpoints.
- *
- * Control lock must be held at entry, and will be held at exit.
- */
-static void
-SlruInternalWritePage(SlruCtl ctl, int slotno, SlruWriteAll fdata)
-{
- SlruShared shared = ctl->shared;
- int pageno = shared->page_number[slotno];
- bool ok;
-
- /* If a write is in progress, wait for it to finish */
- while (shared->page_status[slotno] == SLRU_PAGE_WRITE_IN_PROGRESS &&
- shared->page_number[slotno] == pageno)
- {
- SimpleLruWaitIO(ctl, slotno);
- }
-
- /*
- * Do nothing if page is not dirty, or if buffer no longer contains the
- * same page we were called for.
- */
- if (!shared->page_dirty[slotno] ||
- shared->page_status[slotno] != SLRU_PAGE_VALID ||
- shared->page_number[slotno] != pageno)
- return;
-
- /*
- * Mark the slot write-busy, and clear the dirtybit. After this point, a
- * transaction status update on this page will mark it dirty again.
- */
- shared->page_status[slotno] = SLRU_PAGE_WRITE_IN_PROGRESS;
- shared->page_dirty[slotno] = false;
-
- /* Acquire per-buffer lock (cannot deadlock, see notes at top) */
- LWLockAcquire(&shared->buffer_locks[slotno].lock, LW_EXCLUSIVE);
-
- /* Release control lock while doing I/O */
- LWLockRelease(shared->ControlLock);
-
- /* Do the write */
- ok = SlruPhysicalWritePage(ctl, pageno, slotno, fdata);
-
- /* If we failed, and we're in a flush, better close the files */
- if (!ok && fdata)
- {
- int i;
-
- for (i = 0; i < fdata->num_files; i++)
- CloseTransientFile(fdata->fd[i]);
- }
-
- /* Re-acquire control lock and update page state */
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
- Assert(shared->page_number[slotno] == pageno &&
- shared->page_status[slotno] == SLRU_PAGE_WRITE_IN_PROGRESS);
-
- /* If we failed to write, mark the page dirty again */
- if (!ok)
- shared->page_dirty[slotno] = true;
-
- shared->page_status[slotno] = SLRU_PAGE_VALID;
-
- LWLockRelease(&shared->buffer_locks[slotno].lock);
-
- /* Now it's okay to ereport if we failed */
- if (!ok)
- SlruReportIOError(ctl, pageno, InvalidTransactionId);
-
- /* If part of a checkpoint, count this as a buffer written. */
- if (fdata)
- CheckpointStats.ckpt_bufs_written++;
-}
-
-/*
- * Wrapper of SlruInternalWritePage, for external callers.
- * fdata is always passed a NULL here.
- */
-void
-SimpleLruWritePage(SlruCtl ctl, int slotno)
-{
- SlruInternalWritePage(ctl, slotno, NULL);
-}
+static void SlruInternalDeleteSegment(int slru_id, int segno);
/*
* Return whether the given page exists on disk.
@@ -624,592 +61,24 @@ SimpleLruWritePage(SlruCtl ctl, int slotno)
* large enough to contain the given page.
*/
bool
-SimpleLruDoesPhysicalPageExist(SlruCtl ctl, int pageno)
+SimpleLruDoesPhysicalPageExist(int slru_id, int pageno)
{
- int segno = pageno / SLRU_PAGES_PER_SEGMENT;
- int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
- int offset = rpageno * BLCKSZ;
- char path[MAXPGPATH];
- int fd;
- bool result;
- off_t endpos;
-
- /* update the stats counter of checked pages */
- pgstat_count_slru_page_exists(ctl->shared->slru_stats_idx);
-
- SlruFileName(ctl, path, segno);
-
- fd = OpenTransientFile(path, O_RDONLY | PG_BINARY);
- if (fd < 0)
- {
- /* expected: file doesn't exist */
- if (errno == ENOENT)
- return false;
-
- /* report error normally */
- slru_errcause = SLRU_OPEN_FAILED;
- slru_errno = errno;
- SlruReportIOError(ctl, pageno, 0);
- }
-
- if ((endpos = lseek(fd, 0, SEEK_END)) < 0)
- {
- slru_errcause = SLRU_SEEK_FAILED;
- slru_errno = errno;
- SlruReportIOError(ctl, pageno, 0);
- }
-
- result = endpos >= (off_t) (offset + BLCKSZ);
-
- if (CloseTransientFile(fd) != 0)
- {
- slru_errcause = SLRU_CLOSE_FAILED;
- slru_errno = errno;
- return false;
- }
-
- return result;
-}
-
-/*
- * Physical read of a (previously existing) page into a buffer slot
- *
- * On failure, we cannot just ereport(ERROR) since caller has put state in
- * shared memory that must be undone. So, we return false and save enough
- * info in static variables to let SlruReportIOError make the report.
- *
- * For now, assume it's not worth keeping a file pointer open across
- * read/write operations. We could cache one virtual file pointer ...
- */
-static bool
-SlruPhysicalReadPage(SlruCtl ctl, int pageno, int slotno)
-{
- SlruShared shared = ctl->shared;
int segno = pageno / SLRU_PAGES_PER_SEGMENT;
int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
off_t offset = rpageno * BLCKSZ;
- char path[MAXPGPATH];
- int fd;
-
- SlruFileName(ctl, path, segno);
-
- /*
- * In a crash-and-restart situation, it's possible for us to receive
- * commands to set the commit status of transactions whose bits are in
- * already-truncated segments of the commit log (see notes in
- * SlruPhysicalWritePage). Hence, if we are InRecovery, allow the case
- * where the file doesn't exist, and return zeroes instead.
- */
- fd = OpenTransientFile(path, O_RDONLY | PG_BINARY);
- if (fd < 0)
- {
- if (errno != ENOENT || !InRecovery)
- {
- slru_errcause = SLRU_OPEN_FAILED;
- slru_errno = errno;
- return false;
- }
-
- ereport(LOG,
- (errmsg("file \"%s\" doesn't exist, reading as zeroes",
- path)));
- MemSet(shared->page_buffer[slotno], 0, BLCKSZ);
- return true;
- }
-
- errno = 0;
- pgstat_report_wait_start(WAIT_EVENT_SLRU_READ);
- if (pg_pread(fd, shared->page_buffer[slotno], BLCKSZ, offset) != BLCKSZ)
- {
- pgstat_report_wait_end();
- slru_errcause = SLRU_READ_FAILED;
- slru_errno = errno;
- CloseTransientFile(fd);
- return false;
- }
- pgstat_report_wait_end();
-
- if (CloseTransientFile(fd) != 0)
- {
- slru_errcause = SLRU_CLOSE_FAILED;
- slru_errno = errno;
- return false;
- }
-
- return true;
-}
-
-/*
- * Physical write of a page from a buffer slot
- *
- * On failure, we cannot just ereport(ERROR) since caller has put state in
- * shared memory that must be undone. So, we return false and save enough
- * info in static variables to let SlruReportIOError make the report.
- *
- * For now, assume it's not worth keeping a file pointer open across
- * independent read/write operations. We do batch operations during
- * SimpleLruWriteAll, though.
- *
- * fdata is NULL for a standalone write, pointer to open-file info during
- * SimpleLruWriteAll.
- */
-static bool
-SlruPhysicalWritePage(SlruCtl ctl, int pageno, int slotno, SlruWriteAll fdata)
-{
- SlruShared shared = ctl->shared;
- int segno = pageno / SLRU_PAGES_PER_SEGMENT;
- int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
- off_t offset = rpageno * BLCKSZ;
- char path[MAXPGPATH];
- int fd = -1;
-
- /* update the stats counter of written pages */
- pgstat_count_slru_page_written(shared->slru_stats_idx);
-
- /*
- * Honor the write-WAL-before-data rule, if appropriate, so that we do not
- * write out data before associated WAL records. This is the same action
- * performed during FlushBuffer() in the main buffer manager.
- */
- if (shared->group_lsn != NULL)
- {
- /*
- * We must determine the largest async-commit LSN for the page. This
- * is a bit tedious, but since this entire function is a slow path
- * anyway, it seems better to do this here than to maintain a per-page
- * LSN variable (which'd need an extra comparison in the
- * transaction-commit path).
- */
- XLogRecPtr max_lsn;
- int lsnindex,
- lsnoff;
-
- lsnindex = slotno * shared->lsn_groups_per_page;
- max_lsn = shared->group_lsn[lsnindex++];
- for (lsnoff = 1; lsnoff < shared->lsn_groups_per_page; lsnoff++)
- {
- XLogRecPtr this_lsn = shared->group_lsn[lsnindex++];
-
- if (max_lsn < this_lsn)
- max_lsn = this_lsn;
- }
-
- if (!XLogRecPtrIsInvalid(max_lsn))
- {
- /*
- * As noted above, elog(ERROR) is not acceptable here, so if
- * XLogFlush were to fail, we must PANIC. This isn't much of a
- * restriction because XLogFlush is just about all critical
- * section anyway, but let's make sure.
- */
- START_CRIT_SECTION();
- XLogFlush(max_lsn);
- END_CRIT_SECTION();
- }
- }
-
- /*
- * During a WriteAll, we may already have the desired file open.
- */
- if (fdata)
- {
- int i;
-
- for (i = 0; i < fdata->num_files; i++)
- {
- if (fdata->segno[i] == segno)
- {
- fd = fdata->fd[i];
- break;
- }
- }
- }
-
- if (fd < 0)
- {
- /*
- * If the file doesn't already exist, we should create it. It is
- * possible for this to need to happen when writing a page that's not
- * first in its segment; we assume the OS can cope with that. (Note:
- * it might seem that it'd be okay to create files only when
- * SimpleLruZeroPage is called for the first page of a segment.
- * However, if after a crash and restart the REDO logic elects to
- * replay the log from a checkpoint before the latest one, then it's
- * possible that we will get commands to set transaction status of
- * transactions that have already been truncated from the commit log.
- * Easiest way to deal with that is to accept references to
- * nonexistent files here and in SlruPhysicalReadPage.)
- *
- * Note: it is possible for more than one backend to be executing this
- * code simultaneously for different pages of the same file. Hence,
- * don't use O_EXCL or O_TRUNC or anything like that.
- */
- SlruFileName(ctl, path, segno);
- fd = OpenTransientFile(path, O_RDWR | O_CREAT | PG_BINARY);
- if (fd < 0)
- {
- slru_errcause = SLRU_OPEN_FAILED;
- slru_errno = errno;
- return false;
- }
-
- if (fdata)
- {
- if (fdata->num_files < MAX_WRITEALL_BUFFERS)
- {
- fdata->fd[fdata->num_files] = fd;
- fdata->segno[fdata->num_files] = segno;
- fdata->num_files++;
- }
- else
- {
- /*
- * In the unlikely event that we exceed MAX_FLUSH_BUFFERS,
- * fall back to treating it as a standalone write.
- */
- fdata = NULL;
- }
- }
- }
-
- errno = 0;
- pgstat_report_wait_start(WAIT_EVENT_SLRU_WRITE);
- if (pg_pwrite(fd, shared->page_buffer[slotno], BLCKSZ, offset) != BLCKSZ)
- {
- pgstat_report_wait_end();
- /* if write didn't set errno, assume problem is no disk space */
- if (errno == 0)
- errno = ENOSPC;
- slru_errcause = SLRU_WRITE_FAILED;
- slru_errno = errno;
- if (!fdata)
- CloseTransientFile(fd);
- return false;
- }
- pgstat_report_wait_end();
-
- /* Queue up a sync request for the checkpointer. */
- if (ctl->sync_handler != SYNC_HANDLER_NONE)
- {
- FileTag tag;
-
- INIT_SLRUFILETAG(tag, ctl->sync_handler, segno);
- if (!RegisterSyncRequest(&tag, SYNC_REQUEST, false))
- {
- /* No space to enqueue sync request. Do it synchronously. */
- pgstat_report_wait_start(WAIT_EVENT_SLRU_SYNC);
- if (pg_fsync(fd) != 0)
- {
- pgstat_report_wait_end();
- slru_errcause = SLRU_FSYNC_FAILED;
- slru_errno = errno;
- CloseTransientFile(fd);
- return false;
- }
- pgstat_report_wait_end();
- }
- }
-
- /* Close file, unless part of flush request. */
- if (!fdata)
- {
- if (CloseTransientFile(fd) != 0)
- {
- slru_errcause = SLRU_CLOSE_FAILED;
- slru_errno = errno;
- return false;
- }
- }
-
- return true;
-}
-
-/*
- * Issue the error message after failure of SlruPhysicalReadPage or
- * SlruPhysicalWritePage. Call this after cleaning up shared-memory state.
- */
-static void
-SlruReportIOError(SlruCtl ctl, int pageno, TransactionId xid)
-{
- int segno = pageno / SLRU_PAGES_PER_SEGMENT;
- int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
- int offset = rpageno * BLCKSZ;
- char path[MAXPGPATH];
-
- SlruFileName(ctl, path, segno);
- errno = slru_errno;
- switch (slru_errcause)
- {
- case SLRU_OPEN_FAILED:
- ereport(ERROR,
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not open file \"%s\": %m.", path)));
- break;
- case SLRU_SEEK_FAILED:
- ereport(ERROR,
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not seek in file \"%s\" to offset %d: %m.",
- path, offset)));
- break;
- case SLRU_READ_FAILED:
- if (errno)
- ereport(ERROR,
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not read from file \"%s\" at offset %d: %m.",
- path, offset)));
- else
- ereport(ERROR,
- (errmsg("could not access status of transaction %u", xid),
- errdetail("Could not read from file \"%s\" at offset %d: read too few bytes.", path, offset)));
- break;
- case SLRU_WRITE_FAILED:
- if (errno)
- ereport(ERROR,
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not write to file \"%s\" at offset %d: %m.",
- path, offset)));
- else
- ereport(ERROR,
- (errmsg("could not access status of transaction %u", xid),
- errdetail("Could not write to file \"%s\" at offset %d: wrote too few bytes.",
- path, offset)));
- break;
- case SLRU_FSYNC_FAILED:
- ereport(data_sync_elevel(ERROR),
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not fsync file \"%s\": %m.",
- path)));
- break;
- case SLRU_CLOSE_FAILED:
- ereport(ERROR,
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not close file \"%s\": %m.",
- path)));
- break;
- default:
- /* can't get here, we trust */
- elog(ERROR, "unrecognized SimpleLru error cause: %d",
- (int) slru_errcause);
- break;
- }
-}
-
-/*
- * Select the slot to re-use when we need a free slot.
- *
- * The target page number is passed because we need to consider the
- * possibility that some other process reads in the target page while
- * we are doing I/O to free a slot. Hence, check or recheck to see if
- * any slot already holds the target page, and return that slot if so.
- * Thus, the returned slot is *either* a slot already holding the pageno
- * (could be any state except EMPTY), *or* a freeable slot (state EMPTY
- * or CLEAN).
- *
- * Control lock must be held at entry, and will be held at exit.
- */
-static int
-SlruSelectLRUPage(SlruCtl ctl, int pageno)
-{
- SlruShared shared = ctl->shared;
-
- /* Outer loop handles restart after I/O */
- for (;;)
- {
- int slotno;
- int cur_count;
- int bestvalidslot = 0; /* keep compiler quiet */
- int best_valid_delta = -1;
- int best_valid_page_number = 0; /* keep compiler quiet */
- int bestinvalidslot = 0; /* keep compiler quiet */
- int best_invalid_delta = -1;
- int best_invalid_page_number = 0; /* keep compiler quiet */
+ off_t size;
+ RelFileLocator rlocator = SlruRelFileLocator(slru_id, segno);
+ SMgrFileHandle sfile = smgropen(rlocator, InvalidBackendId, MAIN_FORKNUM);
- /* See if page already has a buffer assigned */
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- if (shared->page_number[slotno] == pageno &&
- shared->page_status[slotno] != SLRU_PAGE_EMPTY)
- return slotno;
- }
-
- /*
- * If we find any EMPTY slot, just select that one. Else choose a
- * victim page to replace. We normally take the least recently used
- * valid page, but we will never take the slot containing
- * latest_page_number, even if it appears least recently used. We
- * will select a slot that is already I/O busy only if there is no
- * other choice: a read-busy slot will not be least recently used once
- * the read finishes, and waiting for an I/O on a write-busy slot is
- * inferior to just picking some other slot. Testing shows the slot
- * we pick instead will often be clean, allowing us to begin a read at
- * once.
- *
- * Normally the page_lru_count values will all be different and so
- * there will be a well-defined LRU page. But since we allow
- * concurrent execution of SlruRecentlyUsed() within
- * SimpleLruReadPage_ReadOnly(), it is possible that multiple pages
- * acquire the same lru_count values. In that case we break ties by
- * choosing the furthest-back page.
- *
- * Notice that this next line forcibly advances cur_lru_count to a
- * value that is certainly beyond any value that will be in the
- * page_lru_count array after the loop finishes. This ensures that
- * the next execution of SlruRecentlyUsed will mark the page newly
- * used, even if it's for a page that has the current counter value.
- * That gets us back on the path to having good data when there are
- * multiple pages with the same lru_count.
- */
- cur_count = (shared->cur_lru_count)++;
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- int this_delta;
- int this_page_number;
-
- if (shared->page_status[slotno] == SLRU_PAGE_EMPTY)
- return slotno;
- this_delta = cur_count - shared->page_lru_count[slotno];
- if (this_delta < 0)
- {
- /*
- * Clean up in case shared updates have caused cur_count
- * increments to get "lost". We back off the page counts,
- * rather than trying to increase cur_count, to avoid any
- * question of infinite loops or failure in the presence of
- * wrapped-around counts.
- */
- shared->page_lru_count[slotno] = cur_count;
- this_delta = 0;
- }
- this_page_number = shared->page_number[slotno];
- if (this_page_number == shared->latest_page_number)
- continue;
- if (shared->page_status[slotno] == SLRU_PAGE_VALID)
- {
- if (this_delta > best_valid_delta ||
- (this_delta == best_valid_delta &&
- ctl->PagePrecedes(this_page_number,
- best_valid_page_number)))
- {
- bestvalidslot = slotno;
- best_valid_delta = this_delta;
- best_valid_page_number = this_page_number;
- }
- }
- else
- {
- if (this_delta > best_invalid_delta ||
- (this_delta == best_invalid_delta &&
- ctl->PagePrecedes(this_page_number,
- best_invalid_page_number)))
- {
- bestinvalidslot = slotno;
- best_invalid_delta = this_delta;
- best_invalid_page_number = this_page_number;
- }
- }
- }
-
- /*
- * If all pages (except possibly the latest one) are I/O busy, we'll
- * have to wait for an I/O to complete and then retry. In that
- * unhappy case, we choose to wait for the I/O on the least recently
- * used slot, on the assumption that it was likely initiated first of
- * all the I/Os in progress and may therefore finish first.
- */
- if (best_valid_delta < 0)
- {
- SimpleLruWaitIO(ctl, bestinvalidslot);
- continue;
- }
-
- /*
- * If the selected page is clean, we're set.
- */
- if (!shared->page_dirty[bestvalidslot])
- return bestvalidslot;
-
- /*
- * Write the page.
- */
- SlruInternalWritePage(ctl, bestvalidslot, NULL);
-
- /*
- * Now loop back and try again. This is the easiest way of dealing
- * with corner cases such as the victim page being re-dirtied while we
- * wrote it.
- */
- }
-}
-
-/*
- * Write dirty pages to disk during checkpoint or database shutdown. Flushing
- * is deferred until the next call to ProcessSyncRequests(), though we do fsync
- * the containing directory here to make sure that newly created directory
- * entries are on disk.
- */
-void
-SimpleLruWriteAll(SlruCtl ctl, bool allow_redirtied)
-{
- SlruShared shared = ctl->shared;
- SlruWriteAllData fdata;
- int slotno;
- int pageno = 0;
- int i;
- bool ok;
-
- /* update the stats counter of flushes */
- pgstat_count_slru_flush(shared->slru_stats_idx);
-
- /*
- * Find and write dirty pages
- */
- fdata.num_files = 0;
-
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- SlruInternalWritePage(ctl, slotno, &fdata);
-
- /*
- * In some places (e.g. checkpoints), we cannot assert that the slot
- * is clean now, since another process might have re-dirtied it
- * already. That's okay.
- */
- Assert(allow_redirtied ||
- shared->page_status[slotno] == SLRU_PAGE_EMPTY ||
- (shared->page_status[slotno] == SLRU_PAGE_VALID &&
- !shared->page_dirty[slotno]));
- }
-
- LWLockRelease(shared->ControlLock);
+ /* update the stats counter of checked pages */
+ pgstat_count_slru_page_exists(slru_id);
- /*
- * Now close any files that were open
- */
- ok = true;
- for (i = 0; i < fdata.num_files; i++)
- {
- if (CloseTransientFile(fdata.fd[i]) != 0)
- {
- slru_errcause = SLRU_CLOSE_FAILED;
- slru_errno = errno;
- pageno = fdata.segno[i] * SLRU_PAGES_PER_SEGMENT;
- ok = false;
- }
- }
- if (!ok)
- SlruReportIOError(ctl, pageno, InvalidTransactionId);
+ if (smgrexists(sfile))
+ size = smgrnblocks(sfile);
+ else
+ size = 0;
- /* Ensure that directory entries for new files are on disk. */
- if (ctl->sync_handler != SYNC_HANDLER_NONE)
- fsync_fname(ctl->Dir, true);
+ return size >= offset + BLCKSZ;
}
/*
@@ -1224,75 +93,14 @@ SimpleLruWriteAll(SlruCtl ctl, bool allow_redirtied)
* after it has accrued freshly-written data.
*/
void
-SimpleLruTruncate(SlruCtl ctl, int cutoffPage)
+SimpleLruTruncate(int slru_id, SlruPagePrecedesFunction PagePrecedes, int cutoffPage)
{
- SlruShared shared = ctl->shared;
- int slotno;
-
/* update the stats counter of truncates */
- pgstat_count_slru_truncate(shared->slru_stats_idx);
-
- /*
- * Scan shared memory and remove any pages preceding the cutoff page, to
- * ensure we won't rewrite them later. (Since this is normally called in
- * or just after a checkpoint, any dirty pages should have been flushed
- * already ... we're just being extra careful here.)
- */
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
-restart:
-
- /*
- * While we are holding the lock, make an important safety check: the
- * current endpoint page must not be eligible for removal.
- */
- if (ctl->PagePrecedes(shared->latest_page_number, cutoffPage))
- {
- LWLockRelease(shared->ControlLock);
- ereport(LOG,
- (errmsg("could not truncate directory \"%s\": apparent wraparound",
- ctl->Dir)));
- return;
- }
-
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- if (shared->page_status[slotno] == SLRU_PAGE_EMPTY)
- continue;
- if (!ctl->PagePrecedes(shared->page_number[slotno], cutoffPage))
- continue;
-
- /*
- * If page is clean, just change state to EMPTY (expected case).
- */
- if (shared->page_status[slotno] == SLRU_PAGE_VALID &&
- !shared->page_dirty[slotno])
- {
- shared->page_status[slotno] = SLRU_PAGE_EMPTY;
- continue;
- }
-
- /*
- * Hmm, we have (or may have) I/O operations acting on the page, so
- * we've got to wait for them to finish and then start again. This is
- * the same logic as in SlruSelectLRUPage. (XXX if page is dirty,
- * wouldn't it be OK to just discard it without writing it?
- * SlruMayDeleteSegment() uses a stricter qualification, so we might
- * not delete this page in the end; even if we don't delete it, we
- * won't have cause to read its data again. For now, keep the logic
- * the same as it was.)
- */
- if (shared->page_status[slotno] == SLRU_PAGE_VALID)
- SlruInternalWritePage(ctl, slotno, NULL);
- else
- SimpleLruWaitIO(ctl, slotno);
- goto restart;
- }
-
- LWLockRelease(shared->ControlLock);
+ pgstat_count_slru_truncate(slru_id);
/* Now we can remove the old segment(s) */
- (void) SlruScanDirectory(ctl, SlruScanDirCbDeleteCutoff, &cutoffPage);
+ (void) SlruScanDirectory(slru_id, PagePrecedes, SlruScanDirCbDeleteCutoff,
+ &cutoffPage);
}
/*
@@ -1302,77 +110,22 @@ restart:
* they either can't yet contain anything, or have already been cleaned out.
*/
static void
-SlruInternalDeleteSegment(SlruCtl ctl, int segno)
+SlruInternalDeleteSegment(int slru_id, int segno)
{
- char path[MAXPGPATH];
-
- /* Forget any fsync requests queued for this segment. */
- if (ctl->sync_handler != SYNC_HANDLER_NONE)
- {
- FileTag tag;
-
- INIT_SLRUFILETAG(tag, ctl->sync_handler, segno);
- RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true);
- }
+ RelFileLocator rlocator = SlruRelFileLocator(slru_id, segno);
+ SMgrFileHandle sfile = smgropen(rlocator, InvalidBackendId, MAIN_FORKNUM);
/* Unlink the file. */
- SlruFileName(ctl, path, segno);
- ereport(DEBUG2, (errmsg_internal("removing file \"%s\"", path)));
- unlink(path);
+ smgrunlink(sfile, false);
}
/*
* Delete an individual SLRU segment, identified by the segment number.
*/
void
-SlruDeleteSegment(SlruCtl ctl, int segno)
+SlruDeleteSegment(int slru_id, int segno)
{
- SlruShared shared = ctl->shared;
- int slotno;
- bool did_write;
-
- /* Clean out any possibly existing references to the segment. */
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-restart:
- did_write = false;
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- int pagesegno = shared->page_number[slotno] / SLRU_PAGES_PER_SEGMENT;
-
- if (shared->page_status[slotno] == SLRU_PAGE_EMPTY)
- continue;
-
- /* not the segment we're looking for */
- if (pagesegno != segno)
- continue;
-
- /* If page is clean, just change state to EMPTY (expected case). */
- if (shared->page_status[slotno] == SLRU_PAGE_VALID &&
- !shared->page_dirty[slotno])
- {
- shared->page_status[slotno] = SLRU_PAGE_EMPTY;
- continue;
- }
-
- /* Same logic as SimpleLruTruncate() */
- if (shared->page_status[slotno] == SLRU_PAGE_VALID)
- SlruInternalWritePage(ctl, slotno, NULL);
- else
- SimpleLruWaitIO(ctl, slotno);
-
- did_write = true;
- }
-
- /*
- * Be extra careful and re-check. The IO functions release the control
- * lock, so new pages could have been read in.
- */
- if (did_write)
- goto restart;
-
- SlruInternalDeleteSegment(ctl, segno);
-
- LWLockRelease(shared->ControlLock);
+ SlruInternalDeleteSegment(slru_id, segno);
}
/*
@@ -1389,19 +142,21 @@ restart:
* first>=cutoff && last>=cutoff: no; every page of this segment is too young
*/
static bool
-SlruMayDeleteSegment(SlruCtl ctl, int segpage, int cutoffPage)
+SlruMayDeleteSegment(SlruPagePrecedesFunction PagePrecedes,
+ int segpage, int cutoffPage)
{
int seg_last_page = segpage + SLRU_PAGES_PER_SEGMENT - 1;
Assert(segpage % SLRU_PAGES_PER_SEGMENT == 0);
- return (ctl->PagePrecedes(segpage, cutoffPage) &&
- ctl->PagePrecedes(seg_last_page, cutoffPage));
+ return (PagePrecedes(segpage, cutoffPage) &&
+ PagePrecedes(seg_last_page, cutoffPage));
}
#ifdef USE_ASSERT_CHECKING
static void
-SlruPagePrecedesTestOffset(SlruCtl ctl, int per_page, uint32 offset)
+SlruPagePrecedesTestOffset(SlruPagePrecedesFunction PagePrecedes,
+ int per_page, uint32 offset)
{
TransactionId lhs,
rhs;
@@ -1426,19 +181,19 @@ SlruPagePrecedesTestOffset(SlruCtl ctl, int per_page, uint32 offset)
Assert(!TransactionIdPrecedes(rhs, lhs + 1));
Assert(!TransactionIdFollowsOrEquals(lhs, rhs));
Assert(!TransactionIdFollowsOrEquals(rhs, lhs));
- Assert(!ctl->PagePrecedes(lhs / per_page, lhs / per_page));
- Assert(!ctl->PagePrecedes(lhs / per_page, rhs / per_page));
- Assert(!ctl->PagePrecedes(rhs / per_page, lhs / per_page));
- Assert(!ctl->PagePrecedes((lhs - per_page) / per_page, rhs / per_page));
- Assert(ctl->PagePrecedes(rhs / per_page, (lhs - 3 * per_page) / per_page));
- Assert(ctl->PagePrecedes(rhs / per_page, (lhs - 2 * per_page) / per_page));
- Assert(ctl->PagePrecedes(rhs / per_page, (lhs - 1 * per_page) / per_page)
+ Assert(!PagePrecedes(lhs / per_page, lhs / per_page));
+ Assert(!PagePrecedes(lhs / per_page, rhs / per_page));
+ Assert(!PagePrecedes(rhs / per_page, lhs / per_page));
+ Assert(!PagePrecedes((lhs - per_page) / per_page, rhs / per_page));
+ Assert(PagePrecedes(rhs / per_page, (lhs - 3 * per_page) / per_page));
+ Assert(PagePrecedes(rhs / per_page, (lhs - 2 * per_page) / per_page));
+ Assert(PagePrecedes(rhs / per_page, (lhs - 1 * per_page) / per_page)
|| (1U << 31) % per_page != 0); /* See CommitTsPagePrecedes() */
- Assert(ctl->PagePrecedes((lhs + 1 * per_page) / per_page, rhs / per_page)
+ Assert(PagePrecedes((lhs + 1 * per_page) / per_page, rhs / per_page)
|| (1U << 31) % per_page != 0);
- Assert(ctl->PagePrecedes((lhs + 2 * per_page) / per_page, rhs / per_page));
- Assert(ctl->PagePrecedes((lhs + 3 * per_page) / per_page, rhs / per_page));
- Assert(!ctl->PagePrecedes(rhs / per_page, (lhs + per_page) / per_page));
+ Assert(PagePrecedes((lhs + 2 * per_page) / per_page, rhs / per_page));
+ Assert(PagePrecedes((lhs + 3 * per_page) / per_page, rhs / per_page));
+ Assert(!PagePrecedes(rhs / per_page, (lhs + per_page) / per_page));
/*
* GetNewTransactionId() has assigned the last XID it can safely use, and
@@ -1451,7 +206,7 @@ SlruPagePrecedesTestOffset(SlruCtl ctl, int per_page, uint32 offset)
oldestXact = newestXact + 1;
oldestXact -= 1U << 31;
oldestPage = oldestXact / per_page;
- Assert(!SlruMayDeleteSegment(ctl,
+ Assert(!SlruMayDeleteSegment(PagePrecedes,
(newestPage -
newestPage % SLRU_PAGES_PER_SEGMENT),
oldestPage));
@@ -1467,7 +222,7 @@ SlruPagePrecedesTestOffset(SlruCtl ctl, int per_page, uint32 offset)
oldestXact = newestXact + 1;
oldestXact -= 1U << 31;
oldestPage = oldestXact / per_page;
- Assert(!SlruMayDeleteSegment(ctl,
+ Assert(!SlruMayDeleteSegment(PagePrecedes,
(newestPage -
newestPage % SLRU_PAGES_PER_SEGMENT),
oldestPage));
@@ -1483,12 +238,12 @@ SlruPagePrecedesTestOffset(SlruCtl ctl, int per_page, uint32 offset)
* do not apply to them.)
*/
void
-SlruPagePrecedesUnitTests(SlruCtl ctl, int per_page)
+SlruPagePrecedesUnitTests(SlruPagePrecedesFunction PagePrecedes, int per_page)
{
/* Test first, middle and last entries of a page. */
- SlruPagePrecedesTestOffset(ctl, per_page, 0);
- SlruPagePrecedesTestOffset(ctl, per_page, per_page / 2);
- SlruPagePrecedesTestOffset(ctl, per_page, per_page - 1);
+ SlruPagePrecedesTestOffset(PagePrecedes, per_page, 0);
+ SlruPagePrecedesTestOffset(PagePrecedes, per_page, per_page / 2);
+ SlruPagePrecedesTestOffset(PagePrecedes, per_page, per_page - 1);
}
#endif
@@ -1498,11 +253,12 @@ SlruPagePrecedesUnitTests(SlruCtl ctl, int per_page)
* one containing the page passed as "data".
*/
bool
-SlruScanDirCbReportPresence(SlruCtl ctl, char *filename, int segpage, void *data)
+SlruScanDirCbReportPresence(int slru_id, SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage, void *data)
{
int cutoffPage = *(int *) data;
- if (SlruMayDeleteSegment(ctl, segpage, cutoffPage))
+ if (SlruMayDeleteSegment(PagePrecedes, segpage, cutoffPage))
return true; /* found one; don't iterate any more */
return false; /* keep going */
@@ -1513,12 +269,15 @@ SlruScanDirCbReportPresence(SlruCtl ctl, char *filename, int segpage, void *data
* This callback deletes segments prior to the one passed in as "data".
*/
static bool
-SlruScanDirCbDeleteCutoff(SlruCtl ctl, char *filename, int segpage, void *data)
+SlruScanDirCbDeleteCutoff(int slru_id, SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage, void *data)
{
int cutoffPage = *(int *) data;
- if (SlruMayDeleteSegment(ctl, segpage, cutoffPage))
- SlruInternalDeleteSegment(ctl, segpage / SLRU_PAGES_PER_SEGMENT);
+ if (SlruMayDeleteSegment(PagePrecedes, segpage, cutoffPage))
+ {
+ SlruDeleteSegment(slru_id, segpage / SLRU_PAGES_PER_SEGMENT);
+ }
return false; /* keep going */
}
@@ -1528,9 +287,10 @@ SlruScanDirCbDeleteCutoff(SlruCtl ctl, char *filename, int segpage, void *data)
* This callback deletes all segments.
*/
bool
-SlruScanDirCbDeleteAll(SlruCtl ctl, char *filename, int segpage, void *data)
+SlruScanDirCbDeleteAll(int slru_id, SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage, void *data)
{
- SlruInternalDeleteSegment(ctl, segpage / SLRU_PAGES_PER_SEGMENT);
+ SlruInternalDeleteSegment(slru_id, segpage / SLRU_PAGES_PER_SEGMENT);
return false; /* keep going */
}
@@ -1551,16 +311,20 @@ SlruScanDirCbDeleteAll(SlruCtl ctl, char *filename, int segpage, void *data)
* Note that no locking is applied.
*/
bool
-SlruScanDirectory(SlruCtl ctl, SlruScanCallback callback, void *data)
+SlruScanDirectory(int slru_id, SlruPagePrecedesFunction PagePrecedes,
+ SlruScanCallback callback, void *data)
{
bool retval = false;
DIR *cldir;
struct dirent *clde;
int segno;
int segpage;
+ const char *path;
- cldir = AllocateDir(ctl->Dir);
- while ((clde = ReadDir(cldir, ctl->Dir)) != NULL)
+ path = slru_dirs[slru_id];
+
+ cldir = AllocateDir(path);
+ while ((clde = ReadDir(cldir, path)) != NULL)
{
size_t len;
@@ -1573,8 +337,8 @@ SlruScanDirectory(SlruCtl ctl, SlruScanCallback callback, void *data)
segpage = segno * SLRU_PAGES_PER_SEGMENT;
elog(DEBUG2, "SlruScanDirectory invoking callback on %s/%s",
- ctl->Dir, clde->d_name);
- retval = callback(ctl, clde->d_name, segpage, data);
+ path, clde->d_name);
+ retval = callback(slru_id, PagePrecedes, clde->d_name, segpage, data);
if (retval)
break;
}
@@ -1585,29 +349,78 @@ SlruScanDirectory(SlruCtl ctl, SlruScanCallback callback, void *data)
}
/*
- * Individual SLRUs (clog, ...) have to provide a sync.c handler function so
- * that they can provide the correct "SlruCtl" (otherwise we don't know how to
- * build the path), but they just forward to this common implementation that
- * performs the fsync.
+ * Read a buffer. Buffer is pinned on return.
*/
-int
-SlruSyncFileTag(SlruCtl ctl, const FileTag *ftag, char *path)
+Buffer
+ReadSlruBuffer(int slru_id, int pageno)
{
- int fd;
- int save_errno;
- int result;
+ int segno = pageno / SLRU_PAGES_PER_SEGMENT;
+ int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
+ RelFileLocator rlocator = SlruRelFileLocator(slru_id, segno);
+ Buffer buffer;
+ bool hit;
+
+ /* Try to avoid doing a buffer mapping table lookup for repeated access. */
+ buffer = slru_recent_buffers[slru_id].recent_buffer;
+ if (slru_recent_buffers[slru_id].pageno == pageno &&
+ BufferIsValid(buffer) &&
+ ReadRecentBuffer(rlocator, MAIN_FORKNUM, pageno, buffer))
+ {
+ pgstat_count_slru_page_hit(slru_id);
+ return buffer;
+ }
+
+ /* Regular lookup. */
+ buffer = ReadBufferWithoutRelcacheWithHit(rlocator, MAIN_FORKNUM, rpageno,
+ RBM_NORMAL, NULL, true, &hit);
- SlruFileName(ctl, path, ftag->segno);
+ /* Remember where this page is for next time. */
+ slru_recent_buffers[slru_id].pageno = pageno;
+ slru_recent_buffers[slru_id].recent_buffer = buffer;
- fd = OpenTransientFile(path, O_RDWR | PG_BINARY);
- if (fd < 0)
- return -1;
+ if (hit)
+ pgstat_count_slru_page_hit(slru_id);
- result = pg_fsync(fd);
- save_errno = errno;
+ return buffer;
+}
+
+/*
+ * Zero-initialize a buffer. Buffer is pinned and exclusively locked on return.
+ */
+Buffer
+ZeroSlruBuffer(int slru_id, int pageno)
+{
+ int segno = pageno / SLRU_PAGES_PER_SEGMENT;
+ int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
+ RelFileLocator rlocator = SlruRelFileLocator(slru_id, segno);
+ Buffer buffer;
+ SMgrFileHandle sfile;
- CloseTransientFile(fd);
+ if (rpageno == 0)
+ {
+ sfile = smgropen(rlocator, InvalidBackendId, MAIN_FORKNUM);
+ if (!smgrexists(sfile))
+ smgrcreate(sfile, false);
+ }
+
+ buffer = ReadBufferWithoutRelcache(rlocator, MAIN_FORKNUM, rpageno,
+ RBM_ZERO_AND_LOCK, NULL, true);
+
+ /* Remember where this page is for next time. */
+ slru_recent_buffers[slru_id].pageno = pageno;
+ slru_recent_buffers[slru_id].recent_buffer = buffer;
+
+ pgstat_count_slru_page_zeroed(slru_id);
+
+ return buffer;
+}
+
+bool
+ProbeSlruBuffer(int slru_id, int pageno)
+{
+ int segno = pageno / SLRU_PAGES_PER_SEGMENT;
+ int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
+ RelFileLocator rlocator = SlruRelFileLocator(slru_id, segno);
- errno = save_errno;
- return result;
+ return BufferProbe(rlocator, MAIN_FORKNUM, rpageno);
}
diff --git a/src/backend/access/transam/subtrans.c b/src/backend/access/transam/subtrans.c
index 62bb610167c..1ab4e5ae557 100644
--- a/src/backend/access/transam/subtrans.c
+++ b/src/backend/access/transam/subtrans.c
@@ -32,6 +32,7 @@
#include "access/subtrans.h"
#include "access/transam.h"
#include "pg_trace.h"
+#include "storage/bufmgr.h"
#include "utils/snapmgr.h"
@@ -55,15 +56,7 @@
#define TransactionIdToEntry(xid) ((xid) % (TransactionId) SUBTRANS_XACTS_PER_PAGE)
-/*
- * Link to shared-memory data structures for SUBTRANS control
- */
-static SlruCtlData SubTransCtlData;
-
-#define SubTransCtl (&SubTransCtlData)
-
-
-static int ZeroSUBTRANSPage(int pageno);
+static Buffer ZeroSUBTRANSPage(int pageno);
static bool SubTransPagePrecedes(int page1, int page2);
@@ -75,16 +68,15 @@ SubTransSetParent(TransactionId xid, TransactionId parent)
{
int pageno = TransactionIdToPage(xid);
int entryno = TransactionIdToEntry(xid);
- int slotno;
TransactionId *ptr;
+ Buffer buffer;
Assert(TransactionIdIsValid(parent));
Assert(TransactionIdFollows(xid, parent));
- LWLockAcquire(SubtransSLRULock, LW_EXCLUSIVE);
-
- slotno = SimpleLruReadPage(SubTransCtl, pageno, true, xid);
- ptr = (TransactionId *) SubTransCtl->shared->page_buffer[slotno];
+ buffer = ReadSlruBuffer(SLRU_SUBTRANS_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ ptr = (TransactionId *) BufferGetPage(buffer);
ptr += entryno;
/*
@@ -96,10 +88,10 @@ SubTransSetParent(TransactionId xid, TransactionId parent)
{
Assert(*ptr == InvalidTransactionId);
*ptr = parent;
- SubTransCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
}
- LWLockRelease(SubtransSLRULock);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -110,9 +102,9 @@ SubTransGetParent(TransactionId xid)
{
int pageno = TransactionIdToPage(xid);
int entryno = TransactionIdToEntry(xid);
- int slotno;
TransactionId *ptr;
TransactionId parent;
+ Buffer buffer;
/* Can't ask about stuff that might not be around anymore */
Assert(TransactionIdFollowsOrEquals(xid, TransactionXmin));
@@ -121,15 +113,14 @@ SubTransGetParent(TransactionId xid)
if (!TransactionIdIsNormal(xid))
return InvalidTransactionId;
- /* lock is acquired by SimpleLruReadPage_ReadOnly */
+ buffer = ReadSlruBuffer(SLRU_SUBTRANS_ID, pageno);
- slotno = SimpleLruReadPage_ReadOnly(SubTransCtl, pageno, xid);
- ptr = (TransactionId *) SubTransCtl->shared->page_buffer[slotno];
+ ptr = (TransactionId *) BufferGetPage(buffer);
ptr += entryno;
parent = *ptr;
- LWLockRelease(SubtransSLRULock);
+ ReleaseBuffer(buffer);
return parent;
}
@@ -177,26 +168,6 @@ SubTransGetTopmostTransaction(TransactionId xid)
return previousXid;
}
-
-/*
- * Initialization of shared memory for SUBTRANS
- */
-Size
-SUBTRANSShmemSize(void)
-{
- return SimpleLruShmemSize(NUM_SUBTRANS_BUFFERS, 0);
-}
-
-void
-SUBTRANSShmemInit(void)
-{
- SubTransCtl->PagePrecedes = SubTransPagePrecedes;
- SimpleLruInit(SubTransCtl, "Subtrans", NUM_SUBTRANS_BUFFERS, 0,
- SubtransSLRULock, "pg_subtrans",
- LWTRANCHE_SUBTRANS_BUFFER, SYNC_HANDLER_NONE);
- SlruPagePrecedesUnitTests(SubTransCtl, SUBTRANS_XACTS_PER_PAGE);
-}
-
/*
* This func must be called ONCE on system install. It creates
* the initial SUBTRANS segment. (The SUBTRANS directory is assumed to
@@ -210,18 +181,16 @@ SUBTRANSShmemInit(void)
void
BootStrapSUBTRANS(void)
{
- int slotno;
+ Buffer buffer;
- LWLockAcquire(SubtransSLRULock, LW_EXCLUSIVE);
+ SlruPagePrecedesUnitTests(SubTransPagePrecedes, SUBTRANS_XACTS_PER_PAGE);
/* Create and zero the first page of the subtrans log */
- slotno = ZeroSUBTRANSPage(0);
+ buffer = ZeroSUBTRANSPage(0);
/* Make sure it's written out */
- SimpleLruWritePage(SubTransCtl, slotno);
- Assert(!SubTransCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(SubtransSLRULock);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -232,10 +201,15 @@ BootStrapSUBTRANS(void)
*
* Control lock must be held at entry, and will be held at exit.
*/
-static int
+static Buffer
ZeroSUBTRANSPage(int pageno)
{
- return SimpleLruZeroPage(SubTransCtl, pageno);
+ Buffer buffer;
+
+ buffer = ZeroSlruBuffer(SLRU_SUBTRANS_ID, pageno);
+ MarkBufferDirty(buffer);
+
+ return buffer;
}
/*
@@ -258,7 +232,6 @@ StartupSUBTRANS(TransactionId oldestActiveXID)
* Whenever we advance into a new page, ExtendSUBTRANS will likewise zero
* the new page without regard to whatever was previously on disk.
*/
- LWLockAcquire(SubtransSLRULock, LW_EXCLUSIVE);
startPage = TransactionIdToPage(oldestActiveXID);
nextXid = ShmemVariableCache->nextXid;
@@ -266,36 +239,15 @@ StartupSUBTRANS(TransactionId oldestActiveXID)
while (startPage != endPage)
{
- (void) ZeroSUBTRANSPage(startPage);
+ UnlockReleaseBuffer(ZeroSUBTRANSPage(startPage));
startPage++;
/* must account for wraparound */
if (startPage > TransactionIdToPage(MaxTransactionId))
startPage = 0;
}
- (void) ZeroSUBTRANSPage(startPage);
-
- LWLockRelease(SubtransSLRULock);
+ UnlockReleaseBuffer(ZeroSUBTRANSPage(startPage));
}
-/*
- * Perform a checkpoint --- either during shutdown, or on-the-fly
- */
-void
-CheckPointSUBTRANS(void)
-{
- /*
- * Write dirty SUBTRANS pages to disk
- *
- * This is not actually necessary from a correctness point of view. We do
- * it merely to improve the odds that writing of dirty pages is done by
- * the checkpoint process and not by backends.
- */
- TRACE_POSTGRESQL_SUBTRANS_CHECKPOINT_START(true);
- SimpleLruWriteAll(SubTransCtl, true);
- TRACE_POSTGRESQL_SUBTRANS_CHECKPOINT_DONE(true);
-}
-
-
/*
* Make sure that SUBTRANS has room for a newly-allocated XID.
*
@@ -319,12 +271,8 @@ ExtendSUBTRANS(TransactionId newestXact)
pageno = TransactionIdToPage(newestXact);
- LWLockAcquire(SubtransSLRULock, LW_EXCLUSIVE);
-
/* Zero the page */
- ZeroSUBTRANSPage(pageno);
-
- LWLockRelease(SubtransSLRULock);
+ UnlockReleaseBuffer(ZeroSUBTRANSPage(pageno));
}
@@ -350,7 +298,7 @@ TruncateSUBTRANS(TransactionId oldestXact)
TransactionIdRetreat(oldestXact);
cutoffPage = TransactionIdToPage(oldestXact);
- SimpleLruTruncate(SubTransCtl, cutoffPage);
+ SimpleLruTruncate(SLRU_SUBTRANS_ID, SubTransPagePrecedes, cutoffPage);
}
diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index b8764012607..7d4800a5f24 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -68,6 +68,7 @@
#include "utils/inval.h"
#include "utils/memutils.h"
#include "utils/relmapper.h"
+#include "utils/resowner_private.h"
#include "utils/snapmgr.h"
#include "utils/timeout.h"
#include "utils/timestamp.h"
@@ -1397,6 +1398,7 @@ RecordTransactionCommit(void)
* are delaying the checkpoint a bit fuzzy, but it doesn't matter.
*/
Assert((MyProc->delayChkptFlags & DELAY_CHKPT_START) == 0);
+
START_CRIT_SECTION();
MyProc->delayChkptFlags |= DELAY_CHKPT_START;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f9f0f6db8d1..68917d17299 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -4638,6 +4638,7 @@ BootStrapXLOG(void)
uint64 sysidentifier;
struct timeval tv;
pg_crc32c crc;
+ ResourceOwner resowner;
/* allow ordinary WAL segment creation, like StartupXLOG() would */
SetInstallXLogFileSegmentActive();
@@ -4777,10 +4778,14 @@ BootStrapXLOG(void)
WriteControlFile();
/* Bootstrap the commit log, too */
+ resowner = ResourceOwnerCreate(NULL, "bootstrap resowner");
+ CurrentResourceOwner = resowner;
BootStrapCLOG();
BootStrapCommitTs();
BootStrapSUBTRANS();
BootStrapMultiXact();
+ CurrentResourceOwner = NULL;
+ ResourceOwnerDelete(resowner);
pfree(buffer);
@@ -4789,6 +4794,8 @@ BootStrapXLOG(void)
* otherwise never run the checks and GUC related initializations therein.
*/
ReadControlFile();
+
+ smgrcloseall();
}
static char *
@@ -6997,15 +7004,11 @@ CheckPointGuts(XLogRecPtr checkPointRedo, int flags)
CheckPointSnapBuild();
CheckPointLogicalRewriteHeap();
CheckPointReplicationOrigin();
+ CheckPointPredicate();
- /* Write out all dirty data in SLRUs and the main buffer pool */
+ /* Write out all dirty data in the buffer pool */
TRACE_POSTGRESQL_BUFFER_CHECKPOINT_START(flags);
CheckpointStats.ckpt_write_t = GetCurrentTimestamp();
- CheckPointCLOG();
- CheckPointCommitTs();
- CheckPointSUBTRANS();
- CheckPointMultiXact();
- CheckPointPredicate();
CheckPointBuffers(flags);
/* Perform all queued up fsyncs */
diff --git a/src/backend/commands/async.c b/src/backend/commands/async.c
index ef909cf4e08..f944766ec2b 100644
--- a/src/backend/commands/async.c
+++ b/src/backend/commands/async.c
@@ -141,6 +141,7 @@
#include "libpq/libpq.h"
#include "libpq/pqformat.h"
#include "miscadmin.h"
+#include "storage/bufmgr.h"
#include "storage/ipc.h"
#include "storage/lmgr.h"
#include "storage/proc.h"
@@ -305,12 +306,6 @@ static AsyncQueueControl *asyncQueueControl;
#define QUEUE_NEXT_LISTENER(i) (asyncQueueControl->backend[i].nextListener)
#define QUEUE_BACKEND_POS(i) (asyncQueueControl->backend[i].pos)
-/*
- * The SLRU buffer area through which we access the notification queue
- */
-static SlruCtlData NotifyCtlData;
-
-#define NotifyCtl (&NotifyCtlData)
#define QUEUE_PAGESIZE BLCKSZ
#define QUEUE_FULL_WARN_INTERVAL 5000 /* warn at most once every 5s */
@@ -521,8 +516,6 @@ AsyncShmemSize(void)
size = mul_size(MaxBackends + 1, sizeof(QueueBackendStatus));
size = add_size(size, offsetof(AsyncQueueControl, backend));
- size = add_size(size, SimpleLruShmemSize(NUM_NOTIFY_BUFFERS, 0));
-
return size;
}
@@ -565,20 +558,13 @@ AsyncShmemInit(void)
}
}
- /*
- * Set up SLRU management of the pg_notify data.
- */
- NotifyCtl->PagePrecedes = asyncQueuePagePrecedes;
- SimpleLruInit(NotifyCtl, "Notify", NUM_NOTIFY_BUFFERS, 0,
- NotifySLRULock, "pg_notify", LWTRANCHE_NOTIFY_BUFFER,
- SYNC_HANDLER_NONE);
-
if (!found)
{
/*
* During start or reboot, clean out the pg_notify directory.
*/
- (void) SlruScanDirectory(NotifyCtl, SlruScanDirCbDeleteAll, NULL);
+ (void) SlruScanDirectory(SLRU_NOTIFY_ID, asyncQueuePagePrecedes,
+ SlruScanDirCbDeleteAll, NULL);
}
}
@@ -1411,10 +1397,7 @@ asyncQueueAddEntries(ListCell *nextNotify)
QueuePosition queue_head;
int pageno;
int offset;
- int slotno;
-
- /* We hold both NotifyQueueLock and NotifySLRULock during this operation */
- LWLockAcquire(NotifySLRULock, LW_EXCLUSIVE);
+ Buffer buffer;
/*
* We work with a local copy of QUEUE_HEAD, which we write back to shared
@@ -1439,13 +1422,17 @@ asyncQueueAddEntries(ListCell *nextNotify)
*/
pageno = QUEUE_POS_PAGE(queue_head);
if (QUEUE_POS_IS_ZERO(queue_head))
- slotno = SimpleLruZeroPage(NotifyCtl, pageno);
+ {
+ buffer = ZeroSlruBuffer(SLRU_NOTIFY_ID, pageno);
+ }
else
- slotno = SimpleLruReadPage(NotifyCtl, pageno, true,
- InvalidTransactionId);
+ {
+ buffer = ReadSlruBuffer(SLRU_NOTIFY_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ }
/* Note we mark the page dirty before writing in it */
- NotifyCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
while (nextNotify != NULL)
{
@@ -1476,7 +1463,7 @@ asyncQueueAddEntries(ListCell *nextNotify)
}
/* Now copy qe into the shared buffer page */
- memcpy(NotifyCtl->shared->page_buffer[slotno] + offset,
+ memcpy(BufferGetPage(buffer) + offset,
&qe,
qe.length);
@@ -1491,7 +1478,10 @@ asyncQueueAddEntries(ListCell *nextNotify)
* asyncQueueIsFull() ensured that there is room to create this
* page without overrunning the queue.
*/
- slotno = SimpleLruZeroPage(NotifyCtl, QUEUE_POS_PAGE(queue_head));
+ UnlockReleaseBuffer(buffer);
+ buffer = ZeroSlruBuffer(SLRU_NOTIFY_ID,
+ QUEUE_POS_PAGE(queue_head));
+ MarkBufferDirty(buffer);
/*
* If the new page address is a multiple of QUEUE_CLEANUP_DELAY,
@@ -1505,12 +1495,11 @@ asyncQueueAddEntries(ListCell *nextNotify)
break;
}
}
+ UnlockReleaseBuffer(buffer);
/* Success, so update the global QUEUE_HEAD */
QUEUE_HEAD = queue_head;
- LWLockRelease(NotifySLRULock);
-
return nextNotify;
}
@@ -1983,17 +1972,16 @@ asyncQueueReadAllNotifications(void)
{
int curpage = QUEUE_POS_PAGE(pos);
int curoffset = QUEUE_POS_OFFSET(pos);
- int slotno;
int copysize;
+ Buffer buffer;
/*
- * We copy the data from SLRU into a local buffer, so as to avoid
- * holding the NotifySLRULock while we are examining the entries
- * and possibly transmitting them to our frontend. Copy only the
- * part of the page we will actually inspect.
+ * We copy the data into a local buffer, so as to avoid holding a
+ * buffer pin while we are examining the entries and possibly
+ * transmitting them to our frontend. Copy only the part of the
+ * page we will actually inspect.
*/
- slotno = SimpleLruReadPage_ReadOnly(NotifyCtl, curpage,
- InvalidTransactionId);
+ buffer = ReadSlruBuffer(SLRU_NOTIFY_ID, curpage);
if (curpage == QUEUE_POS_PAGE(head))
{
/* we only want to read as far as head */
@@ -2007,10 +1995,9 @@ asyncQueueReadAllNotifications(void)
copysize = QUEUE_PAGESIZE - curoffset;
}
memcpy(page_buffer.buf + curoffset,
- NotifyCtl->shared->page_buffer[slotno] + curoffset,
+ BufferGetPage(buffer) + curoffset,
copysize);
- /* Release lock that we got from SimpleLruReadPage_ReadOnly() */
- LWLockRelease(NotifySLRULock);
+ ReleaseBuffer(buffer);
/*
* Process messages up to the stop position, end of page, or an
@@ -2207,7 +2194,7 @@ asyncQueueAdvanceTail(void)
* SimpleLruTruncate() will ask for NotifySLRULock but will also
* release the lock again.
*/
- SimpleLruTruncate(NotifyCtl, newtailpage);
+ SimpleLruTruncate(SLRU_NOTIFY_ID, asyncQueuePagePrecedes, newtailpage);
/*
* Update QUEUE_STOP_PAGE. This changes asyncQueueIsFull()'s verdict
diff --git a/src/backend/storage/buffer/buf_init.c b/src/backend/storage/buffer/buf_init.c
index 20946c47cb4..9746838bf9f 100644
--- a/src/backend/storage/buffer/buf_init.c
+++ b/src/backend/storage/buffer/buf_init.c
@@ -20,6 +20,7 @@
BufferDescPadded *BufferDescriptors;
char *BufferBlocks;
+XLogRecPtr *BufferExternalLSNs;
ConditionVariableMinimallyPadded *BufferIOCVArray;
WritebackContext BackendWritebackContext;
CkptSortItem *CkptBufferIds;
@@ -69,9 +70,11 @@ InitBufferPool(void)
{
bool foundBufs,
foundDescs,
+ foundLSNs,
foundIOCV,
foundBufCkpt;
+
/* Align descriptors to a cacheline boundary. */
BufferDescriptors = (BufferDescPadded *)
ShmemInitStruct("Buffer Descriptors",
@@ -88,6 +91,11 @@ InitBufferPool(void)
NBuffers * sizeof(ConditionVariableMinimallyPadded),
&foundIOCV);
+ BufferExternalLSNs = (XLogRecPtr *)
+ ShmemInitStruct("Buffer External LSNs",
+ NBuffers * sizeof(XLogRecPtr),
+ &foundLSNs);
+
/*
* The array used to sort to-be-checkpointed buffer ids is located in
* shared memory, to avoid having to allocate significant amounts of
@@ -99,10 +107,10 @@ InitBufferPool(void)
ShmemInitStruct("Checkpoint BufferIds",
NBuffers * sizeof(CkptSortItem), &foundBufCkpt);
- if (foundDescs || foundBufs || foundIOCV || foundBufCkpt)
+ if (foundDescs || foundBufs || foundIOCV || foundBufCkpt || foundLSNs)
{
/* should find all of these, or none of them */
- Assert(foundDescs && foundBufs && foundIOCV && foundBufCkpt);
+ Assert(foundDescs && foundBufs && foundIOCV && foundBufCkpt && foundLSNs);
/* note: this path is only taken in EXEC_BACKEND case */
}
else
@@ -133,6 +141,8 @@ InitBufferPool(void)
LWTRANCHE_BUFFER_CONTENT);
ConditionVariableInit(BufferDescriptorGetIOCV(buf));
+
+ BufferExternalLSNs[i] = InvalidXLogRecPtr;
}
/* Correct last entry of linked list */
@@ -166,6 +176,9 @@ BufferShmemSize(void)
/* size of data pages */
size = add_size(size, mul_size(NBuffers, BLCKSZ));
+ /* size of external LSNs */
+ size = add_size(size, mul_size(NBuffers, sizeof(XLogRecPtr)));
+
/* size of stuff controlled by freelist.c */
size = add_size(size, StrategyShmemSize());
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 1305eb7dee1..b3f7be2e05f 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -57,10 +57,21 @@
#include "utils/resowner_private.h"
#include "utils/timestamp.h"
+/*
+ * XXX Ideally we'd switch to standard pages for SLRU data, but in the
+ * meantime we need some way to identify buffers that hold raw data (no
+ * invasive LSN, no checksums).
+ */
+#define BufferHasStandardPage(bufHdr) \
+ ((bufHdr)->tag.spcOid != 9)
+
+#define BufferHasExternalLSN(bufHdr) \
+ !BufferHasStandardPage(bufHdr)
/* Note: these two macros only work on shared buffers, not local ones! */
#define BufHdrGetBlock(bufHdr) ((Block) (BufferBlocks + ((Size) (bufHdr)->buf_id) * BLCKSZ))
-#define BufferGetLSN(bufHdr) (PageGetLSN(BufHdrGetBlock(bufHdr)))
+#define BufferGetLSN(bufHdr) \
+ (BufferHasExternalLSN(bufHdr) ? BufferGetExternalLSN(bufHdr) : PageGetLSN(BufHdrGetBlock(bufHdr)))
/* Note: this macro only works on local buffers, not shared ones! */
#define LocalBufHdrGetBlock(bufHdr) \
@@ -786,6 +797,18 @@ ReadBufferWithoutRelcache(RelFileLocator rlocator, ForkNumber forkNum,
mode, strategy, &hit);
}
+Buffer
+ReadBufferWithoutRelcacheWithHit(RelFileLocator rlocator, ForkNumber forkNum,
+ BlockNumber blockNum, ReadBufferMode mode,
+ BufferAccessStrategy strategy, bool permanent, bool *hit)
+{
+ SMgrFileHandle sfile = smgropen(rlocator, InvalidBackendId, forkNum);
+
+ return ReadBuffer_common(sfile, permanent ? RELPERSISTENCE_PERMANENT :
+ RELPERSISTENCE_UNLOGGED, blockNum,
+ mode, strategy, hit);
+}
+
/*
* ReadBuffer_common -- common logic for all ReadBuffer variants
@@ -1032,7 +1055,8 @@ ReadBuffer_common(SMgrFileHandle sfile, char relpersistence,
}
/* check for garbage data */
- if (!PageIsVerifiedExtended((Page) bufBlock, blockNum,
+ if (BufferHasStandardPage(bufHdr) &&
+ !PageIsVerifiedExtended((Page) bufBlock, blockNum,
PIV_LOG_WARNING | PIV_REPORT_STAT))
{
if (mode == RBM_ZERO_ON_ERROR || zero_damaged_pages)
@@ -1433,6 +1457,9 @@ BufferAlloc(SMgrFileHandle sfile, char relpersistence,
UnpinBuffer(buf);
}
+ if (BufferHasExternalLSN(buf))
+ BufferSetExternalLSN(buf, InvalidXLogRecPtr);
+
/*
* Okay, it's finally safe to rename the buffer.
*
@@ -3087,7 +3114,10 @@ BufferGetLSNAtomic(Buffer buffer)
Assert(BufferIsPinned(buffer));
buf_state = LockBufHdr(bufHdr);
- lsn = PageGetLSN(page);
+ if (BufferHasStandardPage(bufHdr))
+ lsn = PageGetLSN(page);
+ else
+ lsn = BufferGetExternalLSN(bufHdr);
UnlockBufHdr(bufHdr, buf_state);
return lsn;
@@ -5068,3 +5098,29 @@ TestForOldSnapshot_impl(Snapshot snapshot, Relation relation)
(errcode(ERRCODE_SNAPSHOT_TOO_OLD),
errmsg("snapshot too old")));
}
+
+/*
+ * Check if a buffer tag is currently mapped.
+ *
+ * XXX Dubious semantics; needed only for multixact's handling for
+ * inconsistent states.
+ */
+bool
+BufferProbe(RelFileLocator rlocator, ForkNumber forkNum, BlockNumber blockNum)
+{
+ BufferTag tag;
+ uint32 hash;
+ LWLock *partitionLock;
+ int buf_id;
+
+ InitBufferTag(&tag, &rlocator, forkNum, blockNum);
+
+ hash = BufTableHashCode(&tag);
+ partitionLock = BufMappingPartitionLock(hash);
+
+ LWLockAcquire(partitionLock, LW_SHARED);
+ buf_id = BufTableLookup(&tag, hash);
+ LWLockRelease(partitionLock);
+
+ return buf_id >= 0;
+}
diff --git a/src/backend/storage/ipc/ipci.c b/src/backend/storage/ipc/ipci.c
index 8f1ded7338f..8601e1c0dfb 100644
--- a/src/backend/storage/ipc/ipci.c
+++ b/src/backend/storage/ipc/ipci.c
@@ -117,9 +117,7 @@ CalculateShmemSize(int *num_semaphores)
size = add_size(size, XLogPrefetchShmemSize());
size = add_size(size, XLOGShmemSize());
size = add_size(size, XLogRecoveryShmemSize());
- size = add_size(size, CLOGShmemSize());
size = add_size(size, CommitTsShmemSize());
- size = add_size(size, SUBTRANSShmemSize());
size = add_size(size, TwoPhaseShmemSize());
size = add_size(size, BackgroundWorkerShmemSize());
size = add_size(size, MultiXactShmemSize());
@@ -241,9 +239,7 @@ CreateSharedMemoryAndSemaphores(void)
XLOGShmemInit();
XLogPrefetchShmemInit();
XLogRecoveryShmemInit();
- CLOGShmemInit();
CommitTsShmemInit();
- SUBTRANSShmemInit();
MultiXactShmemInit();
InitBufferPool();
diff --git a/src/backend/storage/lmgr/predicate.c b/src/backend/storage/lmgr/predicate.c
index bfc352aed86..f72fc99762c 100644
--- a/src/backend/storage/lmgr/predicate.c
+++ b/src/backend/storage/lmgr/predicate.c
@@ -313,14 +313,6 @@
((targethash) ^ ((uint32) PointerGetDatum((predicatelocktag)->myXact)) \
<< LOG2_NUM_PREDICATELOCK_PARTITIONS)
-
-/*
- * The SLRU buffer area through which we access the old xids.
- */
-static SlruCtlData SerialSlruCtlData;
-
-#define SerialSlruCtl (&SerialSlruCtlData)
-
#define SERIAL_PAGESIZE BLCKSZ
#define SERIAL_ENTRYSIZE sizeof(SerCommitSeqNo)
#define SERIAL_ENTRIESPERPAGE (SERIAL_PAGESIZE / SERIAL_ENTRYSIZE)
@@ -332,8 +324,8 @@ static SlruCtlData SerialSlruCtlData;
#define SerialNextPage(page) (((page) >= SERIAL_MAX_PAGE) ? 0 : (page) + 1)
-#define SerialValue(slotno, xid) (*((SerCommitSeqNo *) \
- (SerialSlruCtl->shared->page_buffer[slotno] + \
+#define SerialValue(buffer, xid) (*((SerCommitSeqNo *) \
+ (BufferGetPage(buffer) + \
((((uint32) (xid)) % SERIAL_ENTRIESPERPAGE) * SERIAL_ENTRYSIZE))))
#define SerialPage(xid) (((uint32) (xid)) / SERIAL_ENTRIESPERPAGE)
@@ -803,17 +795,10 @@ SerialInit(void)
{
bool found;
- /*
- * Set up SLRU management of the pg_serial data.
- */
- SerialSlruCtl->PagePrecedes = SerialPagePrecedesLogically;
- SimpleLruInit(SerialSlruCtl, "Serial",
- NUM_SERIAL_BUFFERS, 0, SerialSLRULock, "pg_serial",
- LWTRANCHE_SERIAL_BUFFER, SYNC_HANDLER_NONE);
#ifdef USE_ASSERT_CHECKING
SerialPagePrecedesLogicallyUnitTests();
#endif
- SlruPagePrecedesUnitTests(SerialSlruCtl, SERIAL_ENTRIESPERPAGE);
+ SlruPagePrecedesUnitTests(SerialPagePrecedesLogically, SERIAL_ENTRIESPERPAGE);
/*
* Create or attach to the SerialControl structure.
@@ -843,9 +828,9 @@ SerialAdd(TransactionId xid, SerCommitSeqNo minConflictCommitSeqNo)
{
TransactionId tailXid;
int targetPage;
- int slotno;
int firstZeroPage;
bool isNewPage;
+ Buffer buffer;
Assert(TransactionIdIsValid(xid));
@@ -890,16 +875,22 @@ SerialAdd(TransactionId xid, SerCommitSeqNo minConflictCommitSeqNo)
/* Initialize intervening pages. */
while (firstZeroPage != targetPage)
{
- (void) SimpleLruZeroPage(SerialSlruCtl, firstZeroPage);
+ buffer = ZeroSlruBuffer(SLRU_SERIAL_ID, firstZeroPage);
+ MarkBufferDirty(buffer);
+ UnlockReleaseBuffer(buffer);
firstZeroPage = SerialNextPage(firstZeroPage);
}
- slotno = SimpleLruZeroPage(SerialSlruCtl, targetPage);
+ buffer = ZeroSlruBuffer(SLRU_SERIAL_ID, targetPage);
}
else
- slotno = SimpleLruReadPage(SerialSlruCtl, targetPage, true, xid);
+ {
+ buffer = ReadSlruBuffer(SLRU_SERIAL_ID, targetPage);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ }
- SerialValue(slotno, xid) = minConflictCommitSeqNo;
- SerialSlruCtl->shared->page_dirty[slotno] = true;
+ SerialValue(buffer, xid) = minConflictCommitSeqNo;
+ MarkBufferDirty(buffer);
+ UnlockReleaseBuffer(buffer);
LWLockRelease(SerialSLRULock);
}
@@ -915,7 +906,7 @@ SerialGetMinConflictCommitSeqNo(TransactionId xid)
TransactionId headXid;
TransactionId tailXid;
SerCommitSeqNo val;
- int slotno;
+ Buffer buffer;
Assert(TransactionIdIsValid(xid));
@@ -937,9 +928,9 @@ SerialGetMinConflictCommitSeqNo(TransactionId xid)
* The following function must be called without holding SerialSLRULock,
* but will return with that lock held, which must then be released.
*/
- slotno = SimpleLruReadPage_ReadOnly(SerialSlruCtl,
- SerialPage(xid), xid);
- val = SerialValue(slotno, xid);
+ buffer = ReadSlruBuffer(SLRU_SERIAL_ID, SerialPage(xid));
+ val = SerialValue(buffer, xid);
+ ReleaseBuffer(buffer);
LWLockRelease(SerialSLRULock);
return val;
}
@@ -1058,19 +1049,7 @@ CheckPointPredicate(void)
LWLockRelease(SerialSLRULock);
/* Truncate away pages that are no longer required */
- SimpleLruTruncate(SerialSlruCtl, tailPage);
-
- /*
- * Write dirty SLRU pages to disk
- *
- * This is not actually necessary from a correctness point of view. We do
- * it merely as a debugging aid.
- *
- * We're doing this after the truncation to avoid writing pages right
- * before deleting the file in which they sit, which would be completely
- * pointless.
- */
- SimpleLruWriteAll(SerialSlruCtl, true);
+ SimpleLruTruncate(SLRU_SERIAL_ID, SerialPagePrecedesLogically, tailPage);
}
/*------------------------------------------------------------------------*/
@@ -1331,7 +1310,6 @@ PredicateLockShmemSize(void)
/* Shared memory structures for SLRU tracking of old committed xids. */
size = add_size(size, sizeof(SerialControlData));
- size = add_size(size, SimpleLruShmemSize(NUM_SERIAL_BUFFERS, 0));
return size;
}
diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c
index eea7ce944c3..2cfee0deaad 100644
--- a/src/backend/storage/smgr/md.c
+++ b/src/backend/storage/smgr/md.c
@@ -191,9 +191,12 @@ mdcreate(SMgrFileHandle sfile, bool isRedo)
* should be here and not in commands/tablespace.c? But that would imply
* importing a lot of stuff that smgr.c oughtn't know, either.
*/
- TablespaceCreateDbspace(sfile->smgr_locator.locator.spcOid,
- sfile->smgr_locator.locator.dbOid,
- isRedo);
+ if (sfile->smgr_locator.locator.spcOid != SLRU_SPC_OID)
+ {
+ TablespaceCreateDbspace(sfile->smgr_locator.locator.spcOid,
+ sfile->smgr_locator.locator.dbOid,
+ isRedo);
+ }
path = smgrfilepath(sfile->smgr_locator);
diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c
index 58a2322d018..d47695b808b 100644
--- a/src/backend/storage/smgr/smgr.c
+++ b/src/backend/storage/smgr/smgr.c
@@ -420,6 +420,48 @@ smgrunlink_multi(RelFileLocator rlocator, BackendId backend,
}
}
+/*
+ * smgrdounlink() -- Immediately unlink a file
+ *
+ * If isRedo is true, it is okay for the underlying file(s) to be gone
+ * already.
+ *
+ * To remove a relation transactionally, see RelationDropStorage() instead.
+ * This will cause cache invalidation of all forks of the relation, not just
+ * this one.
+ */
+void
+smgrunlink(SMgrFileHandle sfile, bool isRedo)
+{
+ SMgrFileLocator locator;
+ int which;
+
+ /* remember before closing */
+ which = sfile->smgr_which;
+ locator = sfile->smgr_locator;
+
+ /* Close the file at smgr level */
+ smgrclose(sfile);
+
+ /*
+ * Send a shared-inval message to force other backends to close any
+ * dangling smgr references they may have for these rels. We should do
+ * this before starting the actual unlinking, in case we fail partway
+ * through that step. Note that the sinval messages will eventually come
+ * back to this backend, too, and thereby provide a backstop that we
+ * closed our own smgr rel.
+ */
+ CacheInvalidateSmgr(locator.locator, locator.backend);
+
+ /*
+ * Delete the physical file(s).
+ *
+ * Note: smgr_unlink must treat deletion failure as a WARNING, not an
+ * ERROR, because we've already decided to commit or abort the current
+ * xact.
+ */
+ smgrsw[which].smgr_unlink(locator, isRedo);
+}
/*
* smgrextend() -- Add a new block to a file.
diff --git a/src/backend/storage/sync/sync.c b/src/backend/storage/sync/sync.c
index 768d1dbfc41..41886288644 100644
--- a/src/backend/storage/sync/sync.c
+++ b/src/backend/storage/sync/sync.c
@@ -18,9 +18,7 @@
#include <fcntl.h>
#include <sys/file.h>
-#include "access/commit_ts.h"
-#include "access/clog.h"
-#include "access/multixact.h"
+#include "access/slru.h"
#include "access/xlog.h"
#include "access/xlogutils.h"
#include "commands/tablespace.h"
@@ -106,22 +104,6 @@ static const SyncOps syncsw[] = {
.sync_unlinkfiletag = mdunlinkfiletag,
.sync_filetagmatches = mdfiletagmatches
},
- /* pg_xact */
- [SYNC_HANDLER_CLOG] = {
- .sync_syncfiletag = clogsyncfiletag
- },
- /* pg_commit_ts */
- [SYNC_HANDLER_COMMIT_TS] = {
- .sync_syncfiletag = committssyncfiletag
- },
- /* pg_multixact/offsets */
- [SYNC_HANDLER_MULTIXACT_OFFSET] = {
- .sync_syncfiletag = multixactoffsetssyncfiletag
- },
- /* pg_multixact/members */
- [SYNC_HANDLER_MULTIXACT_MEMBER] = {
- .sync_syncfiletag = multixactmemberssyncfiletag
- }
};
/*
diff --git a/src/backend/utils/mmgr/mcxt.c b/src/backend/utils/mmgr/mcxt.c
index 0b00802df70..4f9fa85d51a 100644
--- a/src/backend/utils/mmgr/mcxt.c
+++ b/src/backend/utils/mmgr/mcxt.c
@@ -162,9 +162,47 @@ static void MemoryContextStatsPrint(MemoryContext context, void *passthru,
* You should not do memory allocations within a critical section, because
* an out-of-memory error will be escalated to a PANIC. To enforce that
* rule, the allocation functions Assert that.
+ *
+ * FIXME: bypass this for the critical section in RecordTransactionCommit()
+ * for now. It does a lot of things that can allocate:
+ * - calls TransactionIdCommitTree, which pins buffers, which requires
+ * space in the ResourceOwner for the pin (ResourceOwnerEnlargeBuffers())
+ * - same for TransactionTreeSetCommitTsData() call.
+ * - reading a page can require flushing other pages, which in turn
+ * can call CompactCheckpointerRequestQueue(), which allocates
+ * - reading a page calls smgropen(), which allocates the SMgrFile entry
+ * if it's not open already
+ *
+ * FIXME: Here's another codepath that reaches this, reproducable with
+ * the 'lock-committed-update' isolation test:
+ *
+ * #5 0x000056230e91788d in MemoryContextAllocExtended (context=0x562310709c40, size=4048, flags=2) at mcxt.c:1137
+ * #6 0x000056230e8e9655 in DynaHashAlloc (size=4048) at dynahash.c:292
+ * #7 0x000056230e8ebadf in element_alloc (hashp=0x562310709d58, nelem=46, freelist_idx=0) at dynahash.c:1715
+ * #8 0x000056230e8eaef8 in get_hash_entry (hashp=0x562310709d58, freelist_idx=0) at dynahash.c:1324
+ * #9 0x000056230e8ea993 in hash_search_with_hash_value (hashp=0x562310709d58, keyPtr=0x7ffc30cdd4f0, hashvalue=1219519527, action=HASH_ENTER, foundPtr=0x7ffc30cdd4ef) at dynahash.c:1097
+ * #10 0x000056230e8ea578 in hash_search (hashp=0x562310709d58, keyPtr=0x7ffc30cdd4f0, action=HASH_ENTER, foundPtr=0x7ffc30cdd4ef) at dynahash.c:958
+ * #11 0x000056230e70f8fa in smgropen (rlocator=..., backend=-1, forkNum=MAIN_FORKNUM) at smgr.c:165
+ * #12 0x000056230e6c7f58 in ReadBufferWithoutRelcacheWithHit (rlocator=..., forkNum=MAIN_FORKNUM, blockNum=0, mode=RBM_NORMAL, strategy=0x0, permanent=true, hit=0x7ffc30cdd597)
+ * at bufmgr.c:805
+ * #13 0x000056230e2b45ce in ReadSlruBuffer (slru_id=3, pageno=0) at slru.c:377
+ * #14 0x000056230e2ad192 in RecordNewMultiXact (multi=5, offset=9, nmembers=2, members=0x7ffc30cdd690) at multixact.c:902
+ * #15 0x000056230e2acfbb in MultiXactIdCreateFromMembers (nmembers=2, members=0x7ffc30cdd690) at multixact.c:833
+ * #16 0x000056230e2ac8d3 in MultiXactIdCreate (xid1=753, status1=MultiXactStatusForKeyShare, xid2=754, status2=MultiXactStatusNoKeyUpdate) at multixact.c:402
+ * #17 0x000056230e248ff2 in compute_new_xmax_infomask (xmax=753, old_infomask=402, old_infomask2=2, add_to_xmax=754, mode=LockTupleNoKeyExclusive, is_update=true, result_xmax=0x7ffc30cdd79c,
+ * result_infomask=0x7ffc30cdd79a, result_infomask2=0x7ffc30cdd798) at heapam.c:5017
+ * #18 0x000056230e24632c in heap_update (relation=0x7f99454cb168, otid=0x7ffc30cddaba, newtup=0x56231073e840, cid=0, crosscheck=0x0, wait=true, tmfd=0x7ffc30cddaf0, lockmode=0x7ffc30cdda34)
+ * at heapam.c:3345
+ *
+ * Disabled this completely because of that.
*/
+#if 0
#define AssertNotInCriticalSection(context) \
- Assert(CritSectionCount == 0 || (context)->allowInCritSection)
+ Assert(CritSectionCount == 0 || (context)->allowInCritSection || \
+ (MyProc != NULL && (MyProc->delayChkptFlags & DELAY_CHKPT_START != 0)))
+#else
+#define AssertNotInCriticalSection(context) ((void)true)
+#endif
/*
* Call the given function in the MemoryContextMethods for the memory context
diff --git a/src/common/relpath.c b/src/common/relpath.c
index ae2d384fb34..4715d06d287 100644
--- a/src/common/relpath.c
+++ b/src/common/relpath.c
@@ -22,6 +22,16 @@
#include "common/relpath.h"
#include "storage/backendid.h"
+/*
+ * SLRU ID to path mapping
+ */
+#define PG_SLRU(symname,name,path,synchronize) \
+ path,
+
+static char *slru_dirs[] =
+{
+#include "access/slrulist.h"
+};
/*
* Lookup table of fork name by fork number.
@@ -143,7 +153,22 @@ GetSMgrFilePath(Oid dbOid, Oid spcOid, RelFileNumber relNumber,
{
char *path;
- if (spcOid == GLOBALTABLESPACE_OID)
+ if (spcOid == SLRU_SPC_OID)
+ {
+ if (dbOid >= lengthof(slru_dirs) || forkNumber != 0 || backendId != InvalidBackendId)
+ {
+#ifndef FRONTEND
+ ereport(ERROR,
+ (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
+ errmsg("invalid SLRU file locator %u/%u/%u/%u/%u",
+ spcOid, dbOid, relNumber, backendId, forkNumber)));
+#else
+ return NULL;
+#endif
+ }
+ path = psprintf("%s/%04X", slru_dirs[dbOid], relNumber);
+ }
+ else if (spcOid == GLOBALTABLESPACE_OID)
{
/* Shared system relations live in {datadir}/global */
Assert(dbOid == 0);
diff --git a/src/include/access/clog.h b/src/include/access/clog.h
index d99444f073f..aacf10ca522 100644
--- a/src/include/access/clog.h
+++ b/src/include/access/clog.h
@@ -40,18 +40,12 @@ extern void TransactionIdSetTreeStatus(TransactionId xid, int nsubxids,
TransactionId *subxids, XidStatus status, XLogRecPtr lsn);
extern XidStatus TransactionIdGetStatus(TransactionId xid, XLogRecPtr *lsn);
-extern Size CLOGShmemBuffers(void);
-extern Size CLOGShmemSize(void);
-extern void CLOGShmemInit(void);
extern void BootStrapCLOG(void);
extern void StartupCLOG(void);
extern void TrimCLOG(void);
-extern void CheckPointCLOG(void);
extern void ExtendCLOG(TransactionId newestXact);
extern void TruncateCLOG(TransactionId oldestXact, Oid oldestxid_datoid);
-extern int clogsyncfiletag(const FileTag *ftag, char *path);
-
/* XLOG stuff */
#define CLOG_ZEROPAGE 0x00
#define CLOG_TRUNCATE 0x10
diff --git a/src/include/access/commit_ts.h b/src/include/access/commit_ts.h
index 5087cdce51e..605dc75b3b0 100644
--- a/src/include/access/commit_ts.h
+++ b/src/include/access/commit_ts.h
@@ -27,7 +27,6 @@ extern bool TransactionIdGetCommitTsData(TransactionId xid,
extern TransactionId GetLatestCommitTsData(TimestampTz *ts,
RepOriginId *nodeid);
-extern Size CommitTsShmemBuffers(void);
extern Size CommitTsShmemSize(void);
extern void CommitTsShmemInit(void);
extern void BootStrapCommitTs(void);
@@ -41,8 +40,6 @@ extern void SetCommitTsLimit(TransactionId oldestXact,
TransactionId newestXact);
extern void AdvanceOldestCommitTsXid(TransactionId oldestXact);
-extern int committssyncfiletag(const FileTag *ftag, char *path);
-
/* XLOG stuff */
#define COMMIT_TS_ZEROPAGE 0x00
#define COMMIT_TS_TRUNCATE 0x10
diff --git a/src/include/access/multixact.h b/src/include/access/multixact.h
index 246f757f6ab..5848e4072ba 100644
--- a/src/include/access/multixact.h
+++ b/src/include/access/multixact.h
@@ -118,9 +118,6 @@ extern bool MultiXactIdPrecedes(MultiXactId multi1, MultiXactId multi2);
extern bool MultiXactIdPrecedesOrEquals(MultiXactId multi1,
MultiXactId multi2);
-extern int multixactoffsetssyncfiletag(const FileTag *ftag, char *path);
-extern int multixactmemberssyncfiletag(const FileTag *ftag, char *path);
-
extern void AtEOXact_MultiXact(void);
extern void AtPrepare_MultiXact(void);
extern void PostPrepare_MultiXact(TransactionId xid);
diff --git a/src/include/access/slru.h b/src/include/access/slru.h
index a8a424d92da..fcae11ce599 100644
--- a/src/include/access/slru.h
+++ b/src/include/access/slru.h
@@ -1,7 +1,7 @@
/*-------------------------------------------------------------------------
*
* slru.h
- * Simple LRU buffering for transaction status logfiles
+ * Buffering for transaction status logfiles
*
* Portions Copyright (c) 1996-2023, PostgreSQL Global Development Group
* Portions Copyright (c) 1994, Regents of the University of California
@@ -14,9 +14,35 @@
#define SLRU_H
#include "access/xlogdefs.h"
+#include "catalog/pg_tablespace_d.h"
+#include "storage/buf.h"
#include "storage/lwlock.h"
+#include "storage/smgr.h"
#include "storage/sync.h"
+/* Pseudo database ID used for SLRU data. */
+#define SLRU_SPC_ID 9
+
+/* Pseudo database IDs used by each cache. */
+#define PG_SLRU(symname,name,path, synchronize) \
+ symname,
+
+typedef enum SlruIds
+{
+#include "access/slrulist.h"
+ SLRU_NEXT_ID
+} SlruIds;
+#undef PG_SLRU
+
+typedef bool (*SlruPagePrecedesFunction) (int, int);
+
+static inline RelFileLocator
+SlruRelFileLocator(uint32 slru_db_id, uint32 segment_id)
+{
+ RelFileLocator rlocator = {SLRU_SPC_ID, slru_db_id, segment_id};
+ return rlocator;
+}
+
/*
* Define SLRU segment size. A page is the same BLCKSZ as is used everywhere
@@ -33,142 +59,40 @@
*/
#define SLRU_PAGES_PER_SEGMENT 32
-/*
- * Page status codes. Note that these do not include the "dirty" bit.
- * page_dirty can be true only in the VALID or WRITE_IN_PROGRESS states;
- * in the latter case it implies that the page has been re-dirtied since
- * the write started.
- */
-typedef enum
-{
- SLRU_PAGE_EMPTY, /* buffer is not in use */
- SLRU_PAGE_READ_IN_PROGRESS, /* page is being read in */
- SLRU_PAGE_VALID, /* page is valid and not being written */
- SLRU_PAGE_WRITE_IN_PROGRESS /* page is being written out */
-} SlruPageStatus;
-
-/*
- * Shared-memory state
- */
-typedef struct SlruSharedData
-{
- LWLock *ControlLock;
-
- /* Number of buffers managed by this SLRU structure */
- int num_slots;
-
- /*
- * Arrays holding info for each buffer slot. Page number is undefined
- * when status is EMPTY, as is page_lru_count.
- */
- char **page_buffer;
- SlruPageStatus *page_status;
- bool *page_dirty;
- int *page_number;
- int *page_lru_count;
- LWLockPadded *buffer_locks;
-
- /*
- * Optional array of WAL flush LSNs associated with entries in the SLRU
- * pages. If not zero/NULL, we must flush WAL before writing pages (true
- * for pg_xact, false for multixact, pg_subtrans, pg_notify). group_lsn[]
- * has lsn_groups_per_page entries per buffer slot, each containing the
- * highest LSN known for a contiguous group of SLRU entries on that slot's
- * page.
- */
- XLogRecPtr *group_lsn;
- int lsn_groups_per_page;
-
- /*----------
- * We mark a page "most recently used" by setting
- * page_lru_count[slotno] = ++cur_lru_count;
- * The oldest page is therefore the one with the highest value of
- * cur_lru_count - page_lru_count[slotno]
- * The counts will eventually wrap around, but this calculation still
- * works as long as no page's age exceeds INT_MAX counts.
- *----------
- */
- int cur_lru_count;
-
- /*
- * latest_page_number is the page number of the current end of the log;
- * this is not critical data, since we use it only to avoid swapping out
- * the latest page.
- */
- int latest_page_number;
-
- /* SLRU's index for statistics purposes (might not be unique) */
- int slru_stats_idx;
-} SlruSharedData;
-
-typedef SlruSharedData *SlruShared;
-
-/*
- * SlruCtlData is an unshared structure that points to the active information
- * in shared memory.
- */
-typedef struct SlruCtlData
-{
- SlruShared shared;
-
- /*
- * Which sync handler function to use when handing sync requests over to
- * the checkpointer. SYNC_HANDLER_NONE to disable fsync (eg pg_notify).
- */
- SyncRequestHandler sync_handler;
-
- /*
- * Decide whether a page is "older" for truncation and as a hint for
- * evicting pages in LRU order. Return true if every entry of the first
- * argument is older than every entry of the second argument. Note that
- * !PagePrecedes(a,b) && !PagePrecedes(b,a) need not imply a==b; it also
- * arises when some entries are older and some are not. For SLRUs using
- * SimpleLruTruncate(), this must use modular arithmetic. (For others,
- * the behavior of this callback has no functional implications.) Use
- * SlruPagePrecedesUnitTests() in SLRUs meeting its criteria.
- */
- bool (*PagePrecedes) (int, int);
-
- /*
- * Dir is set during SimpleLruInit and does not change thereafter. Since
- * it's always the same, it doesn't need to be in shared memory.
- */
- char Dir[64];
-} SlruCtlData;
-
-typedef SlruCtlData *SlruCtl;
-
-
-extern Size SimpleLruShmemSize(int nslots, int nlsns);
-extern void SimpleLruInit(SlruCtl ctl, const char *name, int nslots, int nlsns,
- LWLock *ctllock, const char *subdir, int tranche_id,
- SyncRequestHandler sync_handler);
-extern int SimpleLruZeroPage(SlruCtl ctl, int pageno);
-extern int SimpleLruReadPage(SlruCtl ctl, int pageno, bool write_ok,
- TransactionId xid);
-extern int SimpleLruReadPage_ReadOnly(SlruCtl ctl, int pageno,
- TransactionId xid);
-extern void SimpleLruWritePage(SlruCtl ctl, int slotno);
-extern void SimpleLruWriteAll(SlruCtl ctl, bool allow_redirtied);
#ifdef USE_ASSERT_CHECKING
-extern void SlruPagePrecedesUnitTests(SlruCtl ctl, int per_page);
+extern void SlruPagePrecedesUnitTests(SlruPagePrecedesFunction PagePrecedes,
+ int per_page);
#else
#define SlruPagePrecedesUnitTests(ctl, per_page) do {} while (0)
#endif
-extern void SimpleLruTruncate(SlruCtl ctl, int cutoffPage);
-extern bool SimpleLruDoesPhysicalPageExist(SlruCtl ctl, int pageno);
+extern void SimpleLruTruncate(int slru_id, SlruPagePrecedesFunction PagePrecedes,
+ int cutoffPage);
+extern bool SimpleLruDoesPhysicalPageExist(int slru_id, int pageno);
-typedef bool (*SlruScanCallback) (SlruCtl ctl, char *filename, int segpage,
+typedef bool (*SlruScanCallback) (int slru_id,
+ SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage,
void *data);
-extern bool SlruScanDirectory(SlruCtl ctl, SlruScanCallback callback, void *data);
-extern void SlruDeleteSegment(SlruCtl ctl, int segno);
-
-extern int SlruSyncFileTag(SlruCtl ctl, const FileTag *ftag, char *path);
+extern bool SlruScanDirectory(int slru_id, SlruPagePrecedesFunction PagePrecedes,
+ SlruScanCallback callback, void *data);
+extern void SlruDeleteSegment(int slru_id, int segno);
/* SlruScanDirectory public callbacks */
-extern bool SlruScanDirCbReportPresence(SlruCtl ctl, char *filename,
+extern bool SlruScanDirCbReportPresence(int slru_id,
+ SlruPagePrecedesFunction PagePrecedes,
+ char *filename,
int segpage, void *data);
-extern bool SlruScanDirCbDeleteAll(SlruCtl ctl, char *filename, int segpage,
+extern bool SlruScanDirCbDeleteAll(int slru_id, SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage,
void *data);
+/* Buffer access */
+extern Buffer ReadSlruBuffer(int slru_id, int pageno);
+extern Buffer ZeroSlruBuffer(int slru_id, int pageno);
+extern bool ProbeSlruBuffer(int slru_id, int pageno);
+
+/* Interfaces use by stats view */
+extern Oid SlruRelIdByName(const char *name);
+extern const char *SlruName(int slru_id);
+
#endif /* SLRU_H */
diff --git a/src/include/access/subtrans.h b/src/include/access/subtrans.h
index 46a473c77f5..14e3bf720fe 100644
--- a/src/include/access/subtrans.h
+++ b/src/include/access/subtrans.h
@@ -18,11 +18,8 @@ extern void SubTransSetParent(TransactionId xid, TransactionId parent);
extern TransactionId SubTransGetParent(TransactionId xid);
extern TransactionId SubTransGetTopmostTransaction(TransactionId xid);
-extern Size SUBTRANSShmemSize(void);
-extern void SUBTRANSShmemInit(void);
extern void BootStrapSUBTRANS(void);
extern void StartupSUBTRANS(TransactionId oldestActiveXID);
-extern void CheckPointSUBTRANS(void);
extern void ExtendSUBTRANS(TransactionId newestXact);
extern void TruncateSUBTRANS(TransactionId oldestXact);
diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h
index 12df11c7dfc..7f132864592 100644
--- a/src/include/common/relpath.h
+++ b/src/include/common/relpath.h
@@ -63,6 +63,9 @@ typedef enum ForkNumber
#define FORKNAMECHARS 4 /* max chars for a fork name */
+/* Pseudo tablespace ID used for SLRUs. */
+#define SLRU_SPC_OID 9
+
extern PGDLLIMPORT const char *const forkNames[];
extern ForkNumber forkname_to_number(const char *forkName);
diff --git a/src/include/storage/buf_internals.h b/src/include/storage/buf_internals.h
index 6a37e0ce6b4..276e3c55c3b 100644
--- a/src/include/storage/buf_internals.h
+++ b/src/include/storage/buf_internals.h
@@ -307,6 +307,7 @@ typedef struct WritebackContext
/* in buf_init.c */
extern PGDLLIMPORT BufferDescPadded *BufferDescriptors;
+extern PGDLLIMPORT XLogRecPtr *BufferExternalLSNs;
extern PGDLLIMPORT ConditionVariableMinimallyPadded *BufferIOCVArray;
extern PGDLLIMPORT WritebackContext BackendWritebackContext;
@@ -344,6 +345,18 @@ BufferDescriptorGetContentLock(const BufferDesc *bdesc)
return (LWLock *) (&bdesc->content_lock);
}
+static inline XLogRecPtr
+BufferGetExternalLSN(const BufferDesc *bdesc)
+{
+ return BufferExternalLSNs[bdesc->buf_id];
+}
+
+static inline void
+BufferSetExternalLSN(const BufferDesc *bdesc, XLogRecPtr lsn)
+{
+ BufferExternalLSNs[bdesc->buf_id] = lsn;
+}
+
/*
* The freeNext field is either the index of the next freelist entry,
* or one of these special values:
diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h
index 7de50bf71b7..4338752826c 100644
--- a/src/include/storage/bufmgr.h
+++ b/src/include/storage/bufmgr.h
@@ -129,12 +129,18 @@ extern Buffer ReadBufferWithoutRelcache(RelFileLocator rlocator,
ForkNumber forkNum, BlockNumber blockNum,
ReadBufferMode mode, BufferAccessStrategy strategy,
bool permanent);
+extern Buffer ReadBufferWithoutRelcacheWithHit(RelFileLocator rlocator,
+ ForkNumber forkNum, BlockNumber blockNum,
+ ReadBufferMode mode, BufferAccessStrategy strategy,
+ bool permanent, bool *hit);
extern void ReleaseBuffer(Buffer buffer);
extern void UnlockReleaseBuffer(Buffer buffer);
extern void MarkBufferDirty(Buffer buffer);
extern void IncrBufferRefCount(Buffer buffer);
extern Buffer ReleaseAndReadBuffer(Buffer buffer, Relation relation,
BlockNumber blockNum);
+extern bool BufferProbe(RelFileLocator rlocator, ForkNumber forkNum,
+ BlockNumber blockNum);
extern void InitBufferPoolAccess(void);
extern void AtEOXact_Buffers(bool isCommit);
diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h
index fe7282127ed..bc11bc70f56 100644
--- a/src/include/storage/smgr.h
+++ b/src/include/storage/smgr.h
@@ -123,6 +123,7 @@ extern void smgrwriteback(SMgrFileHandle sfile,
extern BlockNumber smgrnblocks(SMgrFileHandle sfile);
extern BlockNumber smgrnblocks_cached(SMgrFileHandle sfile);
extern void smgrimmedsync(SMgrFileHandle sfile);
+extern void smgrunlink(SMgrFileHandle sfile, bool isRedo);
extern void smgrtruncate_multi(RelFileLocator rlocator, BackendId backend, ForkNumber *forks, int nforks, BlockNumber *nblocks);
extern void smgrunlink_multi(RelFileLocator rlocator, BackendId backend, ForkNumber *forks, int nforks, bool isRedo);
diff --git a/src/test/modules/Makefile b/src/test/modules/Makefile
index c629cbe3830..3dbdb1d769e 100644
--- a/src/test/modules/Makefile
+++ b/src/test/modules/Makefile
@@ -32,10 +32,11 @@ SUBDIRS = \
test_regex \
test_rls_hooks \
test_shm_mq \
- test_slru \
unsafe_tests \
worker_spi
+# test_slru \ # FIXME: Broken
+
ifeq ($(with_ssl),openssl)
SUBDIRS += ssl_passphrase_callback
else
--
2.30.2
^ permalink raw reply [nested|flat] 22+ messages in thread
* Re: SLRUs in the main buffer pool, redux
@ 2023-02-27 13:36 Heikki Linnakangas <[email protected]>
parent: Heikki Linnakangas <[email protected]>
0 siblings, 1 reply; 22+ messages in thread
From: Heikki Linnakangas @ 2023-02-27 13:36 UTC (permalink / raw)
To: Shawn Debnath <[email protected]>; +Cc: Thomas Munro <[email protected]>; pgsql-hackers; Yura Sokolov <[email protected]>
On 27/02/2023 15:31, Heikki Linnakangas wrote:
> On 20/01/2023 19:00, Shawn Debnath wrote:
>> On Mon, Jul 25, 2022 at 11:54:36AM +0300, Heikki Linnakangas wrote:
>>
>>> Oh I just saw that you had a comment about that in the patch and had hacked
>>> around it. Anyway, calling ResourceOwnerEnlargeBuffers() might be a
>>> solution. Or switch to a separate "CriticalResourceOwner" that's guaranteed
>>> to have enough pre-allocated space, before entering the critical section.
>>
>> Wanted to bump up this thread. Rishu in my team posted a patch in the other
>> SLRU thread [1] with the latest updates and fixes and looks like performance
>> numbers do not show any regression. This change is currently in the
>> January commitfest [2] as well. Any feedback would be appreciated!
>
> Here's a rebased set of patches.
>
> The second patch is failing the pg_upgrade tests. Before I dig into
> that, I'd love to get some feedback on this general approach.
Forgot to include the new "slrulist.h" file in the previous patch, fixed
here.
- Heikki
Attachments:
[text/x-patch] v3-0001-Have-separate-SMmgrRelation-per-fork-rename-it-to.patch (165.4K, ../../[email protected]/2-v3-0001-Have-separate-SMmgrRelation-per-fork-rename-it-to.patch)
download | inline diff:
From b5b1efd94eb0eec68c18b8eac52f0988e334edf8 Mon Sep 17 00:00:00 2001
From: Heikki Linnakangas <[email protected]>
Date: Mon, 27 Feb 2023 15:34:16 +0200
Subject: [PATCH v3 1/2] Have separate SMmgrRelation per fork, rename it to
SMgrFile
---
contrib/amcheck/verify_nbtree.c | 2 +-
contrib/bloom/blinsert.c | 6 +-
contrib/pg_prewarm/autoprewarm.c | 2 +-
contrib/pg_prewarm/pg_prewarm.c | 4 +-
contrib/pg_visibility/pg_visibility.c | 5 +-
src/backend/access/gist/gistbuild.c | 11 +-
src/backend/access/hash/hashpage.c | 2 +-
src/backend/access/heap/heapam_handler.c | 32 +-
src/backend/access/heap/rewriteheap.c | 6 +-
src/backend/access/heap/visibilitymap.c | 41 ++-
src/backend/access/nbtree/nbtree.c | 6 +-
src/backend/access/nbtree/nbtsort.c | 8 +-
src/backend/access/spgist/spginsert.c | 17 +-
src/backend/access/table/tableam.c | 4 +-
src/backend/access/transam/xlogprefetcher.c | 28 +-
src/backend/access/transam/xlogutils.c | 17 +-
src/backend/catalog/catalog.c | 53 +--
src/backend/catalog/index.c | 4 +-
src/backend/catalog/storage.c | 205 +++++++----
src/backend/commands/dbcommands.c | 10 +-
src/backend/commands/sequence.c | 8 +-
src/backend/commands/tablecmds.c | 19 +-
src/backend/storage/buffer/bufmgr.c | 274 +++++++-------
src/backend/storage/buffer/localbuf.c | 24 +-
src/backend/storage/freespace/freespace.c | 41 ++-
src/backend/storage/smgr/md.c | 361 ++++++++-----------
src/backend/storage/smgr/smgr.c | 379 +++++++++-----------
src/backend/utils/cache/inval.c | 17 +-
src/backend/utils/cache/relcache.c | 34 +-
src/common/relpath.c | 4 +-
src/include/catalog/storage.h | 11 +-
src/include/common/relpath.h | 12 +-
src/include/storage/buf_internals.h | 7 +-
src/include/storage/bufmgr.h | 25 +-
src/include/storage/md.h | 28 +-
src/include/storage/relfilelocator.h | 29 +-
src/include/storage/smgr.h | 98 +++--
src/include/utils/inval.h | 2 +-
src/include/utils/rel.h | 32 +-
39 files changed, 907 insertions(+), 961 deletions(-)
diff --git a/contrib/amcheck/verify_nbtree.c b/contrib/amcheck/verify_nbtree.c
index 257cff671bd..2ab837c3062 100644
--- a/contrib/amcheck/verify_nbtree.c
+++ b/contrib/amcheck/verify_nbtree.c
@@ -324,7 +324,7 @@ bt_index_check_internal(Oid indrelid, bool parentcheck, bool heapallindexed,
bool heapkeyspace,
allequalimage;
- if (!smgrexists(RelationGetSmgr(indrel), MAIN_FORKNUM))
+ if (!smgrexists(RelationGetSmgr(indrel, MAIN_FORKNUM)))
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("index \"%s\" lacks a main relation fork",
diff --git a/contrib/bloom/blinsert.c b/contrib/bloom/blinsert.c
index dcd81208953..df4a821369e 100644
--- a/contrib/bloom/blinsert.c
+++ b/contrib/bloom/blinsert.c
@@ -177,9 +177,9 @@ blbuildempty(Relation index)
* this even when wal_level=minimal.
*/
PageSetChecksumInplace(metapage, BLOOM_METAPAGE_BLKNO);
- smgrwrite(RelationGetSmgr(index), INIT_FORKNUM, BLOOM_METAPAGE_BLKNO,
+ smgrwrite(RelationGetSmgr(index, INIT_FORKNUM), BLOOM_METAPAGE_BLKNO,
metapage, true);
- log_newpage(&(RelationGetSmgr(index))->smgr_rlocator.locator, INIT_FORKNUM,
+ log_newpage(&index->rd_locator, INIT_FORKNUM,
BLOOM_METAPAGE_BLKNO, metapage, true);
/*
@@ -187,7 +187,7 @@ blbuildempty(Relation index)
* write did not go through shared_buffers and therefore a concurrent
* checkpoint may have moved the redo pointer past our xlog record.
*/
- smgrimmedsync(RelationGetSmgr(index), INIT_FORKNUM);
+ smgrimmedsync(RelationGetSmgr(index, INIT_FORKNUM));
}
/*
diff --git a/contrib/pg_prewarm/autoprewarm.c b/contrib/pg_prewarm/autoprewarm.c
index 93835449c0e..74cc8df6481 100644
--- a/contrib/pg_prewarm/autoprewarm.c
+++ b/contrib/pg_prewarm/autoprewarm.c
@@ -535,7 +535,7 @@ autoprewarm_database_main(Datum main_arg)
*/
if (blk->forknum > InvalidForkNumber &&
blk->forknum <= MAX_FORKNUM &&
- smgrexists(RelationGetSmgr(rel), blk->forknum))
+ smgrexists(RelationGetSmgr(rel, blk->forknum)))
nblocks = RelationGetNumberOfBlocksInFork(rel, blk->forknum);
else
nblocks = 0;
diff --git a/contrib/pg_prewarm/pg_prewarm.c b/contrib/pg_prewarm/pg_prewarm.c
index 54209924aed..479bbafb745 100644
--- a/contrib/pg_prewarm/pg_prewarm.c
+++ b/contrib/pg_prewarm/pg_prewarm.c
@@ -109,7 +109,7 @@ pg_prewarm(PG_FUNCTION_ARGS)
aclcheck_error(aclresult, get_relkind_objtype(rel->rd_rel->relkind), get_rel_name(relOid));
/* Check that the fork exists. */
- if (!smgrexists(RelationGetSmgr(rel), forkNumber))
+ if (!smgrexists(RelationGetSmgr(rel, forkNumber)))
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
errmsg("fork \"%s\" does not exist for this relation",
@@ -177,7 +177,7 @@ pg_prewarm(PG_FUNCTION_ARGS)
for (block = first_block; block <= last_block; ++block)
{
CHECK_FOR_INTERRUPTS();
- smgrread(RelationGetSmgr(rel), forkNumber, block, blockbuffer.data);
+ smgrread(RelationGetSmgr(rel, forkNumber), block, blockbuffer.data);
++blocks_done;
}
}
diff --git a/contrib/pg_visibility/pg_visibility.c b/contrib/pg_visibility/pg_visibility.c
index 2a4acfd1eee..bb7cd8ec43d 100644
--- a/contrib/pg_visibility/pg_visibility.c
+++ b/contrib/pg_visibility/pg_visibility.c
@@ -386,13 +386,14 @@ pg_truncate_visibility_map(PG_FUNCTION_ARGS)
check_relation_relkind(rel);
/* Forcibly reset cached file size */
- RelationGetSmgr(rel)->smgr_cached_nblocks[VISIBILITYMAP_FORKNUM] = InvalidBlockNumber;
+ RelationGetSmgr(rel, VISIBILITYMAP_FORKNUM)->smgr_cached_nblocks = InvalidBlockNumber;
block = visibilitymap_prepare_truncate(rel, 0);
if (BlockNumberIsValid(block))
{
fork = VISIBILITYMAP_FORKNUM;
- smgrtruncate(RelationGetSmgr(rel), &fork, 1, &block);
+ DropRelationBuffers(rel->rd_locator, rel->rd_backend, &fork, 1, &block);
+ smgrtruncate_multi(rel->rd_locator, rel->rd_backend, &fork, 1, &block);
}
if (RelationNeedsWAL(rel))
diff --git a/src/backend/access/gist/gistbuild.c b/src/backend/access/gist/gistbuild.c
index 7a6d93bb879..7ae5618377c 100644
--- a/src/backend/access/gist/gistbuild.c
+++ b/src/backend/access/gist/gistbuild.c
@@ -416,7 +416,7 @@ gist_indexsortbuild(GISTBuildState *state)
* replaced with the real root page at the end.
*/
page = palloc0(BLCKSZ);
- smgrextend(RelationGetSmgr(state->indexrel), MAIN_FORKNUM, GIST_ROOT_BLKNO,
+ smgrextend(RelationGetSmgr(state->indexrel, MAIN_FORKNUM), GIST_ROOT_BLKNO,
page, true);
state->pages_allocated++;
state->pages_written++;
@@ -460,7 +460,7 @@ gist_indexsortbuild(GISTBuildState *state)
/* Write out the root */
PageSetLSN(levelstate->pages[0], GistBuildLSN);
PageSetChecksumInplace(levelstate->pages[0], GIST_ROOT_BLKNO);
- smgrwrite(RelationGetSmgr(state->indexrel), MAIN_FORKNUM, GIST_ROOT_BLKNO,
+ smgrwrite(RelationGetSmgr(state->indexrel, MAIN_FORKNUM), GIST_ROOT_BLKNO,
levelstate->pages[0], true);
if (RelationNeedsWAL(state->indexrel))
log_newpage(&state->indexrel->rd_locator, MAIN_FORKNUM, GIST_ROOT_BLKNO,
@@ -479,7 +479,7 @@ gist_indexsortbuild(GISTBuildState *state)
* still not be on disk when the crash occurs.
*/
if (RelationNeedsWAL(state->indexrel))
- smgrimmedsync(RelationGetSmgr(state->indexrel), MAIN_FORKNUM);
+ smgrimmedsync(RelationGetSmgr(state->indexrel, MAIN_FORKNUM));
}
/*
@@ -657,7 +657,7 @@ gist_indexsortbuild_flush_ready_pages(GISTBuildState *state)
PageSetLSN(page, GistBuildLSN);
PageSetChecksumInplace(page, blkno);
- smgrextend(RelationGetSmgr(state->indexrel), MAIN_FORKNUM, blkno, page,
+ smgrextend(RelationGetSmgr(state->indexrel, MAIN_FORKNUM), blkno, page,
true);
state->pages_written++;
@@ -943,8 +943,7 @@ gistBuildCallback(Relation index,
*/
if ((buildstate->buildMode == GIST_BUFFERING_AUTO &&
buildstate->indtuples % BUFFERING_MODE_SWITCH_CHECK_STEP == 0 &&
- effective_cache_size < smgrnblocks(RelationGetSmgr(index),
- MAIN_FORKNUM)) ||
+ effective_cache_size < smgrnblocks(RelationGetSmgr(index, MAIN_FORKNUM))) ||
(buildstate->buildMode == GIST_BUFFERING_STATS &&
buildstate->indtuples >= BUFFERING_MODE_TUPLE_SIZE_STATS_TARGET))
{
diff --git a/src/backend/access/hash/hashpage.c b/src/backend/access/hash/hashpage.c
index 2d8fdec98ed..166bf81e028 100644
--- a/src/backend/access/hash/hashpage.c
+++ b/src/backend/access/hash/hashpage.c
@@ -1030,7 +1030,7 @@ _hash_alloc_buckets(Relation rel, BlockNumber firstblock, uint32 nblocks)
true);
PageSetChecksumInplace(page, lastblock);
- smgrextend(RelationGetSmgr(rel), MAIN_FORKNUM, lastblock, zerobuf.data,
+ smgrextend(RelationGetSmgr(rel, MAIN_FORKNUM), lastblock, zerobuf.data,
false);
return true;
diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c
index c4b1916d36e..8af225502cc 100644
--- a/src/backend/access/heap/heapam_handler.c
+++ b/src/backend/access/heap/heapam_handler.c
@@ -572,7 +572,7 @@ heapam_relation_set_new_filelocator(Relation rel,
TransactionId *freezeXid,
MultiXactId *minmulti)
{
- SMgrRelation srel;
+ SMgrFileHandle sfile;
/*
* Initialize to the minimum XID that could put tuples in the table. We
@@ -591,7 +591,7 @@ heapam_relation_set_new_filelocator(Relation rel,
*/
*minmulti = GetOldestMultiXactId();
- srel = RelationCreateStorage(*newrlocator, persistence, true);
+ sfile = RelationCreateStorage(*newrlocator, persistence, true);
/*
* If required, set up an init fork for an unlogged table so that it can
@@ -604,15 +604,18 @@ heapam_relation_set_new_filelocator(Relation rel,
*/
if (persistence == RELPERSISTENCE_UNLOGGED)
{
+ SMgrFileHandle sfile_init;
+
Assert(rel->rd_rel->relkind == RELKIND_RELATION ||
rel->rd_rel->relkind == RELKIND_MATVIEW ||
rel->rd_rel->relkind == RELKIND_TOASTVALUE);
- smgrcreate(srel, INIT_FORKNUM, false);
+ sfile_init = smgropen(*newrlocator, InvalidBackendId, INIT_FORKNUM);
+ smgrcreate(sfile_init, false);
log_smgrcreate(newrlocator, INIT_FORKNUM);
- smgrimmedsync(srel, INIT_FORKNUM);
+ smgrimmedsync(sfile);
}
- smgrclose(srel);
+ smgrclose(sfile);
}
static void
@@ -624,9 +627,7 @@ heapam_relation_nontransactional_truncate(Relation rel)
static void
heapam_relation_copy_data(Relation rel, const RelFileLocator *newrlocator)
{
- SMgrRelation dstrel;
-
- dstrel = smgropen(*newrlocator, rel->rd_backend);
+ SMgrFileHandle dstmain;
/*
* Since we copy the file directly without looking at the shared buffers,
@@ -646,16 +647,21 @@ heapam_relation_copy_data(Relation rel, const RelFileLocator *newrlocator)
RelationCreateStorage(*newrlocator, rel->rd_rel->relpersistence, true);
/* copy main fork */
- RelationCopyStorage(RelationGetSmgr(rel), dstrel, MAIN_FORKNUM,
+ dstmain = smgropen(*newrlocator, rel->rd_backend, MAIN_FORKNUM);
+ RelationCopyStorage(RelationGetSmgr(rel, MAIN_FORKNUM), dstmain,
rel->rd_rel->relpersistence);
/* copy those extra forks that exist */
for (ForkNumber forkNum = MAIN_FORKNUM + 1;
forkNum <= MAX_FORKNUM; forkNum++)
{
- if (smgrexists(RelationGetSmgr(rel), forkNum))
+ SMgrFileHandle src_fork = RelationGetSmgr(rel, forkNum);
+
+ if (smgrexists(src_fork))
{
- smgrcreate(dstrel, forkNum, false);
+ SMgrFileHandle dst_fork = smgropen(*newrlocator, rel->rd_backend, forkNum);
+
+ smgrcreate(dst_fork, false);
/*
* WAL log creation if the relation is persistent, or this is the
@@ -665,7 +671,7 @@ heapam_relation_copy_data(Relation rel, const RelFileLocator *newrlocator)
(rel->rd_rel->relpersistence == RELPERSISTENCE_UNLOGGED &&
forkNum == INIT_FORKNUM))
log_smgrcreate(newrlocator, forkNum);
- RelationCopyStorage(RelationGetSmgr(rel), dstrel, forkNum,
+ RelationCopyStorage(RelationGetSmgr(rel, forkNum), dst_fork,
rel->rd_rel->relpersistence);
}
}
@@ -673,7 +679,7 @@ heapam_relation_copy_data(Relation rel, const RelFileLocator *newrlocator)
/* drop old relation, and close new one */
RelationDropStorage(rel);
- smgrclose(dstrel);
+ smgrclose(dstmain);
}
static void
diff --git a/src/backend/access/heap/rewriteheap.c b/src/backend/access/heap/rewriteheap.c
index ae0282a70ee..30b700b0e8e 100644
--- a/src/backend/access/heap/rewriteheap.c
+++ b/src/backend/access/heap/rewriteheap.c
@@ -325,7 +325,7 @@ end_heap_rewrite(RewriteState state)
PageSetChecksumInplace(state->rs_buffer, state->rs_blockno);
- smgrextend(RelationGetSmgr(state->rs_new_rel), MAIN_FORKNUM,
+ smgrextend(RelationGetSmgr(state->rs_new_rel, MAIN_FORKNUM),
state->rs_blockno, state->rs_buffer, true);
}
@@ -337,7 +337,7 @@ end_heap_rewrite(RewriteState state)
* wrote before the checkpoint.
*/
if (RelationNeedsWAL(state->rs_new_rel))
- smgrimmedsync(RelationGetSmgr(state->rs_new_rel), MAIN_FORKNUM);
+ smgrimmedsync(RelationGetSmgr(state->rs_new_rel, MAIN_FORKNUM));
logical_end_heap_rewrite(state);
@@ -691,7 +691,7 @@ raw_heap_insert(RewriteState state, HeapTuple tup)
*/
PageSetChecksumInplace(page, state->rs_blockno);
- smgrextend(RelationGetSmgr(state->rs_new_rel), MAIN_FORKNUM,
+ smgrextend(RelationGetSmgr(state->rs_new_rel, MAIN_FORKNUM),
state->rs_blockno, page, true);
state->rs_blockno++;
diff --git a/src/backend/access/heap/visibilitymap.c b/src/backend/access/heap/visibilitymap.c
index 74ff01bb172..96ca3b0cc71 100644
--- a/src/backend/access/heap/visibilitymap.c
+++ b/src/backend/access/heap/visibilitymap.c
@@ -468,7 +468,7 @@ visibilitymap_prepare_truncate(Relation rel, BlockNumber nheapblocks)
* If no visibility map has been created yet for this relation, there's
* nothing to truncate.
*/
- if (!smgrexists(RelationGetSmgr(rel), VISIBILITYMAP_FORKNUM))
+ if (!smgrexists(RelationGetSmgr(rel, VISIBILITYMAP_FORKNUM)))
return InvalidBlockNumber;
/*
@@ -535,7 +535,7 @@ visibilitymap_prepare_truncate(Relation rel, BlockNumber nheapblocks)
else
newnblocks = truncBlock;
- if (smgrnblocks(RelationGetSmgr(rel), VISIBILITYMAP_FORKNUM) <= newnblocks)
+ if (smgrnblocks(RelationGetSmgr(rel, VISIBILITYMAP_FORKNUM)) <= newnblocks)
{
/* nothing to do, the file was already smaller than requested size */
return InvalidBlockNumber;
@@ -554,29 +554,29 @@ static Buffer
vm_readbuf(Relation rel, BlockNumber blkno, bool extend)
{
Buffer buf;
- SMgrRelation reln;
+ SMgrFileHandle vm_sfile;
/*
* Caution: re-using this smgr pointer could fail if the relcache entry
* gets closed. It's safe as long as we only do smgr-level operations
* between here and the last use of the pointer.
*/
- reln = RelationGetSmgr(rel);
+ vm_sfile = RelationGetSmgr(rel, VISIBILITYMAP_FORKNUM);
/*
* If we haven't cached the size of the visibility map fork yet, check it
* first.
*/
- if (reln->smgr_cached_nblocks[VISIBILITYMAP_FORKNUM] == InvalidBlockNumber)
+ if (vm_sfile->smgr_cached_nblocks == InvalidBlockNumber)
{
- if (smgrexists(reln, VISIBILITYMAP_FORKNUM))
- smgrnblocks(reln, VISIBILITYMAP_FORKNUM);
+ if (smgrexists(vm_sfile))
+ smgrnblocks(vm_sfile);
else
- reln->smgr_cached_nblocks[VISIBILITYMAP_FORKNUM] = 0;
+ vm_sfile->smgr_cached_nblocks = 0;
}
/* Handle requests beyond EOF */
- if (blkno >= reln->smgr_cached_nblocks[VISIBILITYMAP_FORKNUM])
+ if (blkno >= vm_sfile->smgr_cached_nblocks)
{
if (extend)
vm_extend(rel, blkno + 1);
@@ -603,8 +603,7 @@ vm_readbuf(Relation rel, BlockNumber blkno, bool extend)
* long as it doesn't depend on the page header having correct contents.
* Current usage is safe because PageGetContents() does not require that.
*/
- buf = ReadBufferExtended(rel, VISIBILITYMAP_FORKNUM, blkno,
- RBM_ZERO_ON_ERROR, NULL);
+ buf = ReadBufferExtended(rel, VISIBILITYMAP_FORKNUM, blkno, RBM_ZERO_ON_ERROR, NULL);
if (PageIsNew(BufferGetPage(buf)))
{
LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
@@ -624,7 +623,7 @@ vm_extend(Relation rel, BlockNumber vm_nblocks)
{
BlockNumber vm_nblocks_now;
PGAlignedBlock pg;
- SMgrRelation reln;
+ SMgrFileHandle vm_sfile;
PageInit((Page) pg.data, BLCKSZ, 0);
@@ -645,27 +644,27 @@ vm_extend(Relation rel, BlockNumber vm_nblocks)
* gets closed. It's safe as long as we only do smgr-level operations
* between here and the last use of the pointer.
*/
- reln = RelationGetSmgr(rel);
+ vm_sfile = RelationGetSmgr(rel, VISIBILITYMAP_FORKNUM);
/*
* Create the file first if it doesn't exist. If smgr_vm_nblocks is
* positive then it must exist, no need for an smgrexists call.
*/
- if ((reln->smgr_cached_nblocks[VISIBILITYMAP_FORKNUM] == 0 ||
- reln->smgr_cached_nblocks[VISIBILITYMAP_FORKNUM] == InvalidBlockNumber) &&
- !smgrexists(reln, VISIBILITYMAP_FORKNUM))
- smgrcreate(reln, VISIBILITYMAP_FORKNUM, false);
+ if ((vm_sfile->smgr_cached_nblocks == 0 ||
+ vm_sfile->smgr_cached_nblocks == InvalidBlockNumber) &&
+ !smgrexists(vm_sfile))
+ smgrcreate(vm_sfile, false);
/* Invalidate cache so that smgrnblocks() asks the kernel. */
- reln->smgr_cached_nblocks[VISIBILITYMAP_FORKNUM] = InvalidBlockNumber;
- vm_nblocks_now = smgrnblocks(reln, VISIBILITYMAP_FORKNUM);
+ vm_sfile->smgr_cached_nblocks = InvalidBlockNumber;
+ vm_nblocks_now = smgrnblocks(vm_sfile);
/* Now extend the file */
while (vm_nblocks_now < vm_nblocks)
{
PageSetChecksumInplace((Page) pg.data, vm_nblocks_now);
- smgrextend(reln, VISIBILITYMAP_FORKNUM, vm_nblocks_now, pg.data, false);
+ smgrextend(vm_sfile, vm_nblocks_now, pg.data, false);
vm_nblocks_now++;
}
@@ -676,7 +675,7 @@ vm_extend(Relation rel, BlockNumber vm_nblocks)
* to keep checking for creation or extension of the file, which happens
* infrequently.
*/
- CacheInvalidateSmgr(reln->smgr_rlocator);
+ CacheInvalidateSmgr(rel->rd_locator, rel->rd_backend);
UnlockRelationForExtension(rel, ExclusiveLock);
}
diff --git a/src/backend/access/nbtree/nbtree.c b/src/backend/access/nbtree/nbtree.c
index 3f7b541e9d2..d4db2c6d0fa 100644
--- a/src/backend/access/nbtree/nbtree.c
+++ b/src/backend/access/nbtree/nbtree.c
@@ -164,9 +164,9 @@ btbuildempty(Relation index)
* this even when wal_level=minimal.
*/
PageSetChecksumInplace(metapage, BTREE_METAPAGE);
- smgrwrite(RelationGetSmgr(index), INIT_FORKNUM, BTREE_METAPAGE,
+ smgrwrite(RelationGetSmgr(index, INIT_FORKNUM), BTREE_METAPAGE,
metapage, true);
- log_newpage(&RelationGetSmgr(index)->smgr_rlocator.locator, INIT_FORKNUM,
+ log_newpage(&index->rd_locator, INIT_FORKNUM,
BTREE_METAPAGE, metapage, true);
/*
@@ -174,7 +174,7 @@ btbuildempty(Relation index)
* write did not go through shared_buffers and therefore a concurrent
* checkpoint may have moved the redo pointer past our xlog record.
*/
- smgrimmedsync(RelationGetSmgr(index), INIT_FORKNUM);
+ smgrimmedsync(RelationGetSmgr(index, INIT_FORKNUM));
}
/*
diff --git a/src/backend/access/nbtree/nbtsort.c b/src/backend/access/nbtree/nbtsort.c
index 02b9601bece..4ae27cc57c6 100644
--- a/src/backend/access/nbtree/nbtsort.c
+++ b/src/backend/access/nbtree/nbtsort.c
@@ -662,7 +662,7 @@ _bt_blwritepage(BTWriteState *wstate, Page page, BlockNumber blkno)
if (!wstate->btws_zeropage)
wstate->btws_zeropage = (Page) palloc0(BLCKSZ);
/* don't set checksum for all-zero page */
- smgrextend(RelationGetSmgr(wstate->index), MAIN_FORKNUM,
+ smgrextend(RelationGetSmgr(wstate->index, MAIN_FORKNUM),
wstate->btws_pages_written++,
wstate->btws_zeropage,
true);
@@ -677,14 +677,14 @@ _bt_blwritepage(BTWriteState *wstate, Page page, BlockNumber blkno)
if (blkno == wstate->btws_pages_written)
{
/* extending the file... */
- smgrextend(RelationGetSmgr(wstate->index), MAIN_FORKNUM, blkno,
+ smgrextend(RelationGetSmgr(wstate->index, MAIN_FORKNUM), blkno,
page, true);
wstate->btws_pages_written++;
}
else
{
/* overwriting a block we zero-filled before */
- smgrwrite(RelationGetSmgr(wstate->index), MAIN_FORKNUM, blkno,
+ smgrwrite(RelationGetSmgr(wstate->index, MAIN_FORKNUM), blkno,
page, true);
}
@@ -1431,7 +1431,7 @@ _bt_load(BTWriteState *wstate, BTSpool *btspool, BTSpool *btspool2)
* still not be on disk when the crash occurs.
*/
if (wstate->btws_use_wal)
- smgrimmedsync(RelationGetSmgr(wstate->index), MAIN_FORKNUM);
+ smgrimmedsync(RelationGetSmgr(wstate->index, MAIN_FORKNUM));
}
/*
diff --git a/src/backend/access/spgist/spginsert.c b/src/backend/access/spgist/spginsert.c
index 718a88335d0..3de90846a10 100644
--- a/src/backend/access/spgist/spginsert.c
+++ b/src/backend/access/spgist/spginsert.c
@@ -155,8 +155,11 @@ spgbuild(Relation heap, Relation index, IndexInfo *indexInfo)
void
spgbuildempty(Relation index)
{
+ SMgrFileHandle sfile;
Page page;
+ sfile = RelationGetSmgr(index, INIT_FORKNUM);
+
/* Construct metapage. */
page = (Page) palloc(BLCKSZ);
SpGistInitMetapage(page);
@@ -169,27 +172,27 @@ spgbuildempty(Relation index)
* replayed.
*/
PageSetChecksumInplace(page, SPGIST_METAPAGE_BLKNO);
- smgrwrite(RelationGetSmgr(index), INIT_FORKNUM, SPGIST_METAPAGE_BLKNO,
+ smgrwrite(sfile, SPGIST_METAPAGE_BLKNO,
page, true);
- log_newpage(&(RelationGetSmgr(index))->smgr_rlocator.locator, INIT_FORKNUM,
+ log_newpage(&index->rd_locator, INIT_FORKNUM,
SPGIST_METAPAGE_BLKNO, page, true);
/* Likewise for the root page. */
SpGistInitPage(page, SPGIST_LEAF);
PageSetChecksumInplace(page, SPGIST_ROOT_BLKNO);
- smgrwrite(RelationGetSmgr(index), INIT_FORKNUM, SPGIST_ROOT_BLKNO,
+ smgrwrite(sfile, SPGIST_ROOT_BLKNO,
page, true);
- log_newpage(&(RelationGetSmgr(index))->smgr_rlocator.locator, INIT_FORKNUM,
+ log_newpage(&index->rd_locator, INIT_FORKNUM,
SPGIST_ROOT_BLKNO, page, true);
/* Likewise for the null-tuples root page. */
SpGistInitPage(page, SPGIST_LEAF | SPGIST_NULLS);
PageSetChecksumInplace(page, SPGIST_NULL_BLKNO);
- smgrwrite(RelationGetSmgr(index), INIT_FORKNUM, SPGIST_NULL_BLKNO,
+ smgrwrite(sfile, SPGIST_NULL_BLKNO,
page, true);
- log_newpage(&(RelationGetSmgr(index))->smgr_rlocator.locator, INIT_FORKNUM,
+ log_newpage(&index->rd_locator, INIT_FORKNUM,
SPGIST_NULL_BLKNO, page, true);
/*
@@ -197,7 +200,7 @@ spgbuildempty(Relation index)
* writes did not go through shared buffers and therefore a concurrent
* checkpoint may have moved the redo pointer past our xlog record.
*/
- smgrimmedsync(RelationGetSmgr(index), INIT_FORKNUM);
+ smgrimmedsync(sfile);
}
/*
diff --git a/src/backend/access/table/tableam.c b/src/backend/access/table/tableam.c
index ef0d34fceee..9ebf174b189 100644
--- a/src/backend/access/table/tableam.c
+++ b/src/backend/access/table/tableam.c
@@ -631,10 +631,10 @@ table_block_relation_size(Relation rel, ForkNumber forkNumber)
if (forkNumber == InvalidForkNumber)
{
for (int i = 0; i < MAX_FORKNUM; i++)
- nblocks += smgrnblocks(RelationGetSmgr(rel), i);
+ nblocks += smgrnblocks(RelationGetSmgr(rel, i));
}
else
- nblocks = smgrnblocks(RelationGetSmgr(rel), forkNumber);
+ nblocks = smgrnblocks(RelationGetSmgr(rel, forkNumber));
return nblocks * BLCKSZ;
}
diff --git a/src/backend/access/transam/xlogprefetcher.c b/src/backend/access/transam/xlogprefetcher.c
index 046e40d143a..50674cd8b53 100644
--- a/src/backend/access/transam/xlogprefetcher.c
+++ b/src/backend/access/transam/xlogprefetcher.c
@@ -652,7 +652,7 @@ XLogPrefetcherNextBlock(uintptr_t pgsr_private, XLogRecPtr *lsn)
{
int block_id = prefetcher->next_block_id++;
DecodedBkpBlock *block = &record->blocks[block_id];
- SMgrRelation reln;
+ SMgrFileHandle sfile;
PrefetchBufferResult result;
if (!block->in_use)
@@ -722,7 +722,7 @@ XLogPrefetcherNextBlock(uintptr_t pgsr_private, XLogRecPtr *lsn)
* same relation (with some scheme to handle invalidations
* safely), but for now we'll call smgropen() every time.
*/
- reln = smgropen(block->rlocator, InvalidBackendId);
+ sfile = smgropen(block->rlocator, InvalidBackendId, block->forknum);
/*
* If the relation file doesn't exist on disk, for example because
@@ -731,14 +731,14 @@ XLogPrefetcherNextBlock(uintptr_t pgsr_private, XLogRecPtr *lsn)
* further prefetching in the relation until this record is
* replayed.
*/
- if (!smgrexists(reln, MAIN_FORKNUM))
+ if (!smgrexists(sfile))
{
#ifdef XLOGPREFETCHER_DEBUG_LEVEL
elog(XLOGPREFETCHER_DEBUG_LEVEL,
"suppressing all prefetch in relation %u/%u/%u until %X/%X is replayed, because the relation does not exist on disk",
- reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
- reln->smgr_rlocator.locator.relNumber,
+ sfile->smgr_rlocator.locator.spcOid,
+ sfile->smgr_rlocator.locator.dbOid,
+ sfile->smgr_rlocator.locator.relNumber,
LSN_FORMAT_ARGS(record->lsn));
#endif
XLogPrefetcherAddFilter(prefetcher, block->rlocator, 0,
@@ -752,14 +752,14 @@ XLogPrefetcherNextBlock(uintptr_t pgsr_private, XLogRecPtr *lsn)
* block yet, suppress prefetching of this block and higher until
* this record is replayed.
*/
- if (block->blkno >= smgrnblocks(reln, block->forknum))
+ if (block->blkno >= smgrnblocks(sfile))
{
#ifdef XLOGPREFETCHER_DEBUG_LEVEL
elog(XLOGPREFETCHER_DEBUG_LEVEL,
"suppressing prefetch in relation %u/%u/%u from block %u until %X/%X is replayed, because the relation is too small",
- reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
- reln->smgr_rlocator.locator.relNumber,
+ sfile->smgr_rlocator.locator.spcOid,
+ sfile->smgr_rlocator.locator.dbOid,
+ sfile->smgr_rlocator.locator.relNumber,
block->blkno,
LSN_FORMAT_ARGS(record->lsn));
#endif
@@ -770,7 +770,7 @@ XLogPrefetcherNextBlock(uintptr_t pgsr_private, XLogRecPtr *lsn)
}
/* Try to initiate prefetching. */
- result = PrefetchSharedBuffer(reln, block->forknum, block->blkno);
+ result = PrefetchSharedBuffer(sfile, block->blkno);
if (BufferIsValid(result.recent_buffer))
{
/* Cache hit, nothing to do. */
@@ -796,9 +796,9 @@ XLogPrefetcherNextBlock(uintptr_t pgsr_private, XLogRecPtr *lsn)
*/
elog(ERROR,
"could not prefetch relation %u/%u/%u block %u",
- reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
- reln->smgr_rlocator.locator.relNumber,
+ sfile->smgr_locator.locator.spcOid,
+ sfile->smgr_locator.locator.dbOid,
+ sfile->smgr_locator.locator.relNumber,
block->blkno);
}
}
diff --git a/src/backend/access/transam/xlogutils.c b/src/backend/access/transam/xlogutils.c
index 2c28956b1aa..5e4034f7cd7 100644
--- a/src/backend/access/transam/xlogutils.c
+++ b/src/backend/access/transam/xlogutils.c
@@ -477,7 +477,7 @@ XLogReadBufferExtended(RelFileLocator rlocator, ForkNumber forknum,
{
BlockNumber lastblock;
Buffer buffer;
- SMgrRelation smgr;
+ SMgrFileHandle sfile;
Assert(blkno != P_NEW);
@@ -491,7 +491,7 @@ XLogReadBufferExtended(RelFileLocator rlocator, ForkNumber forknum,
}
/* Open the relation at smgr level */
- smgr = smgropen(rlocator, InvalidBackendId);
+ sfile = smgropen(rlocator, InvalidBackendId, forknum);
/*
* Create the target file if it doesn't already exist. This lets us cope
@@ -501,9 +501,9 @@ XLogReadBufferExtended(RelFileLocator rlocator, ForkNumber forknum,
* filesystem loses an inode during a crash. Better to write the data
* until we are actually told to delete the file.)
*/
- smgrcreate(smgr, forknum, true);
+ smgrcreate(sfile, true);
- lastblock = smgrnblocks(smgr, forknum);
+ lastblock = smgrnblocks(sfile);
if (blkno < lastblock)
{
@@ -631,7 +631,7 @@ CreateFakeRelcacheEntry(RelFileLocator rlocator)
rel->rd_lockInfo.lockRelId.dbId = rlocator.dbOid;
rel->rd_lockInfo.lockRelId.relId = rlocator.relNumber;
- rel->rd_smgr = NULL;
+ MemSet(rel->rd_smgr, 0, sizeof(rel->rd_smgr));
return rel;
}
@@ -643,8 +643,11 @@ void
FreeFakeRelcacheEntry(Relation fakerel)
{
/* make sure the fakerel is not referenced by the SmgrRelation anymore */
- if (fakerel->rd_smgr != NULL)
- smgrclearowner(&fakerel->rd_smgr, fakerel->rd_smgr);
+ for (int i = 0; i <= MAX_FORKNUM; i++)
+ {
+ if (fakerel->rd_smgr[i] != NULL)
+ smgrclearowner(&fakerel->rd_smgr[i], fakerel->rd_smgr[i]);
+ }
pfree(fakerel);
}
diff --git a/src/backend/catalog/catalog.c b/src/backend/catalog/catalog.c
index 1bf6c5633cd..6b01158615b 100644
--- a/src/backend/catalog/catalog.c
+++ b/src/backend/catalog/catalog.c
@@ -501,10 +501,10 @@ GetNewOidWithIndex(Relation relation, Oid indexId, AttrNumber oidcolumn)
RelFileNumber
GetNewRelFileNumber(Oid reltablespace, Relation pg_class, char relpersistence)
{
- RelFileLocatorBackend rlocator;
- char *rpath;
- bool collides;
+ RelFileLocator rlocator;
BackendId backend;
+ SMgrFileHandle sfile;
+ bool collides;
/*
* If we ever get here during pg_upgrade, there's something wrong; all
@@ -513,6 +513,11 @@ GetNewRelFileNumber(Oid reltablespace, Relation pg_class, char relpersistence)
*/
Assert(!IsBinaryUpgrade);
+ /*
+ * The relpath will vary based on the backend ID, so we must initialize
+ * that properly here to make sure that any collisions based on filename
+ * are properly detected.
+ */
switch (relpersistence)
{
case RELPERSISTENCE_TEMP:
@@ -528,53 +533,29 @@ GetNewRelFileNumber(Oid reltablespace, Relation pg_class, char relpersistence)
}
/* This logic should match RelationInitPhysicalAddr */
- rlocator.locator.spcOid = reltablespace ? reltablespace : MyDatabaseTableSpace;
- rlocator.locator.dbOid =
- (rlocator.locator.spcOid == GLOBALTABLESPACE_OID) ?
+ rlocator.spcOid = reltablespace ? reltablespace : MyDatabaseTableSpace;
+ rlocator.dbOid =
+ (rlocator.spcOid == GLOBALTABLESPACE_OID) ?
InvalidOid : MyDatabaseId;
- /*
- * The relpath will vary based on the backend ID, so we must initialize
- * that properly here to make sure that any collisions based on filename
- * are properly detected.
- */
- rlocator.backend = backend;
-
do
{
CHECK_FOR_INTERRUPTS();
/* Generate the OID */
if (pg_class)
- rlocator.locator.relNumber = GetNewOidWithIndex(pg_class, ClassOidIndexId,
+ rlocator.relNumber = GetNewOidWithIndex(pg_class, ClassOidIndexId,
Anum_pg_class_oid);
else
- rlocator.locator.relNumber = GetNewObjectId();
+ rlocator.relNumber = GetNewObjectId();
/* Check for existing file of same name */
- rpath = relpath(rlocator, MAIN_FORKNUM);
-
- if (access(rpath, F_OK) == 0)
- {
- /* definite collision */
- collides = true;
- }
- else
- {
- /*
- * Here we have a little bit of a dilemma: if errno is something
- * other than ENOENT, should we declare a collision and loop? In
- * practice it seems best to go ahead regardless of the errno. If
- * there is a colliding file we will get an smgr failure when we
- * attempt to create the new relation file.
- */
- collides = false;
- }
-
- pfree(rpath);
+ sfile = smgropen(rlocator, backend, MAIN_FORKNUM);
+ collides = smgrexists(sfile);
+ smgrclose(sfile);
} while (collides);
- return rlocator.locator.relNumber;
+ return rlocator.relNumber;
}
/*
diff --git a/src/backend/catalog/index.c b/src/backend/catalog/index.c
index 7777e7ec770..f5a0a16d3d5 100644
--- a/src/backend/catalog/index.c
+++ b/src/backend/catalog/index.c
@@ -3042,9 +3042,9 @@ index_build(Relation heapRelation,
* relfilenumber won't change, and nothing needs to be done here.
*/
if (indexRelation->rd_rel->relpersistence == RELPERSISTENCE_UNLOGGED &&
- !smgrexists(RelationGetSmgr(indexRelation), INIT_FORKNUM))
+ !smgrexists(RelationGetSmgr(indexRelation, INIT_FORKNUM)))
{
- smgrcreate(RelationGetSmgr(indexRelation), INIT_FORKNUM, false);
+ smgrcreate(RelationGetSmgr(indexRelation, INIT_FORKNUM), false);
indexRelation->rd_indam->ambuildempty(indexRelation);
}
diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c
index af1491aa1d1..3816610c87c 100644
--- a/src/backend/catalog/storage.c
+++ b/src/backend/catalog/storage.c
@@ -116,11 +116,11 @@ AddPendingSync(const RelFileLocator *rlocator)
* that does not want the storage to be destroyed in case of an abort may
* pass register_delete = false.
*/
-SMgrRelation
+SMgrFileHandle
RelationCreateStorage(RelFileLocator rlocator, char relpersistence,
bool register_delete)
{
- SMgrRelation srel;
+ SMgrFileHandle sfile;
BackendId backend;
bool needs_wal;
@@ -145,11 +145,11 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence,
return NULL; /* placate compiler */
}
- srel = smgropen(rlocator, backend);
- smgrcreate(srel, MAIN_FORKNUM, false);
+ sfile = smgropen(rlocator, backend, MAIN_FORKNUM);
+ smgrcreate(sfile, false);
if (needs_wal)
- log_smgrcreate(&srel->smgr_rlocator.locator, MAIN_FORKNUM);
+ log_smgrcreate(&rlocator, MAIN_FORKNUM);
/*
* Add the relation to the list of stuff to delete at abort, if we are
@@ -175,7 +175,7 @@ RelationCreateStorage(RelFileLocator rlocator, char relpersistence,
AddPendingSync(&rlocator);
}
- return srel;
+ return sfile;
}
/*
@@ -292,16 +292,18 @@ RelationTruncate(Relation rel, BlockNumber nblocks)
ForkNumber forks[MAX_FORKNUM];
BlockNumber blocks[MAX_FORKNUM];
int nforks = 0;
- SMgrRelation reln;
/*
* Make sure smgr_targblock etc aren't pointing somewhere past new end.
* (Note: don't rely on this reln pointer below this loop.)
*/
- reln = RelationGetSmgr(rel);
- reln->smgr_targblock = InvalidBlockNumber;
- for (int i = 0; i <= MAX_FORKNUM; ++i)
- reln->smgr_cached_nblocks[i] = InvalidBlockNumber;
+ for (int i = 0; i <= MAX_FORKNUM; i++)
+ {
+ SMgrFileHandle sfile = RelationGetSmgr(rel, i);
+
+ sfile->smgr_targblock = InvalidBlockNumber;
+ sfile->smgr_cached_nblocks = InvalidBlockNumber;
+ }
/* Prepare for truncation of MAIN fork of the relation */
forks[nforks] = MAIN_FORKNUM;
@@ -309,7 +311,7 @@ RelationTruncate(Relation rel, BlockNumber nblocks)
nforks++;
/* Prepare for truncation of the FSM if it exists */
- fsm = smgrexists(RelationGetSmgr(rel), FSM_FORKNUM);
+ fsm = smgrexists(RelationGetSmgr(rel, FSM_FORKNUM));
if (fsm)
{
blocks[nforks] = FreeSpaceMapPrepareTruncateRel(rel, nblocks);
@@ -322,7 +324,7 @@ RelationTruncate(Relation rel, BlockNumber nblocks)
}
/* Prepare for truncation of the visibility map too if it exists */
- vm = smgrexists(RelationGetSmgr(rel), VISIBILITYMAP_FORKNUM);
+ vm = smgrexists(RelationGetSmgr(rel, VISIBILITYMAP_FORKNUM));
if (vm)
{
blocks[nforks] = visibilitymap_prepare_truncate(rel, nblocks);
@@ -390,11 +392,12 @@ RelationTruncate(Relation rel, BlockNumber nblocks)
}
/*
- * This will first remove any buffers from the buffer pool that should no
+ * First remove any buffers from the buffer pool that should no
* longer exist after truncation is complete, and then truncate the
* corresponding files on disk.
*/
- smgrtruncate(RelationGetSmgr(rel), forks, nforks, blocks);
+ DropRelationBuffers(rel->rd_locator, rel->rd_backend, forks, nforks, blocks);
+ smgrtruncate_multi(rel->rd_locator, rel->rd_backend, forks, nforks, blocks);
/* We've done all the critical work, so checkpoints are OK now. */
MyProc->delayChkptFlags &= ~DELAY_CHKPT_COMPLETE;
@@ -428,7 +431,7 @@ RelationPreTruncate(Relation rel)
return;
pending = hash_search(pendingSyncHash,
- &(RelationGetSmgr(rel)->smgr_rlocator.locator),
+ &rel->rd_locator,
HASH_FIND, NULL);
if (pending)
pending->is_truncated = true;
@@ -444,12 +447,12 @@ RelationPreTruncate(Relation rel)
* Also note that this is frequently called via locutions such as
* RelationCopyStorage(RelationGetSmgr(rel), ...);
* That's safe only because we perform only smgr and WAL operations here.
- * If we invoked anything else, a relcache flush could cause our SMgrRelation
+ * If we invoked anything else, a relcache flush could cause our SMgrFileHandle
* argument to become a dangling pointer.
*/
void
-RelationCopyStorage(SMgrRelation src, SMgrRelation dst,
- ForkNumber forkNum, char relpersistence)
+RelationCopyStorage(SMgrFileHandle src, SMgrFileHandle dst,
+ char relpersistence)
{
PGAlignedBlock buf;
Page page;
@@ -466,7 +469,7 @@ RelationCopyStorage(SMgrRelation src, SMgrRelation dst,
* it needs to be synced to disk.
*/
copying_initfork = relpersistence == RELPERSISTENCE_UNLOGGED &&
- forkNum == INIT_FORKNUM;
+ src->smgr_locator.forknum == INIT_FORKNUM;
/*
* We need to log the copied data in WAL iff WAL archiving/streaming is
@@ -477,14 +480,14 @@ RelationCopyStorage(SMgrRelation src, SMgrRelation dst,
use_wal = XLogIsNeeded() &&
(relpersistence == RELPERSISTENCE_PERMANENT || copying_initfork);
- nblocks = smgrnblocks(src, forkNum);
+ nblocks = smgrnblocks(src);
for (blkno = 0; blkno < nblocks; blkno++)
{
/* If we got a cancel signal during the copy of the data, quit */
CHECK_FOR_INTERRUPTS();
- smgrread(src, forkNum, blkno, buf.data);
+ smgrread(src, blkno, buf.data);
if (!PageIsVerifiedExtended(page, blkno,
PIV_LOG_WARNING | PIV_REPORT_STAT))
@@ -496,9 +499,7 @@ RelationCopyStorage(SMgrRelation src, SMgrRelation dst,
* (errcontext callbacks shouldn't be risking any such thing, but
* people have been known to forget that rule.)
*/
- char *relpath = relpathbackend(src->smgr_rlocator.locator,
- src->smgr_rlocator.backend,
- forkNum);
+ char *relpath = smgrfilepath(src->smgr_locator);
ereport(ERROR,
(errcode(ERRCODE_DATA_CORRUPTED),
@@ -512,7 +513,7 @@ RelationCopyStorage(SMgrRelation src, SMgrRelation dst,
* space.
*/
if (use_wal)
- log_newpage(&dst->smgr_rlocator.locator, forkNum, blkno, page, false);
+ log_newpage(&dst->smgr_locator.locator, dst->smgr_locator.forknum, blkno, page, false);
PageSetChecksumInplace(page, blkno);
@@ -521,7 +522,7 @@ RelationCopyStorage(SMgrRelation src, SMgrRelation dst,
* need for smgr to schedule an fsync for this write; we'll do it
* ourselves below.
*/
- smgrextend(dst, forkNum, blkno, buf.data, true);
+ smgrextend(dst, blkno, buf.data, true);
}
/*
@@ -534,7 +535,7 @@ RelationCopyStorage(SMgrRelation src, SMgrRelation dst,
* they might still not be on disk when the crash occurs.
*/
if (use_wal || copying_initfork)
- smgrimmedsync(dst, forkNum);
+ smgrimmedsync(dst);
}
/*
@@ -653,9 +654,9 @@ smgrDoPendingDeletes(bool isCommit)
PendingRelDelete *pending;
PendingRelDelete *prev;
PendingRelDelete *next;
- int nrels = 0,
- maxrels = 0;
- SMgrRelation *srels = NULL;
+ int nlocators = 0,
+ maxlocators = 0;
+ RelFileLocatorBackend *locators = NULL;
prev = NULL;
for (pending = pendingDeletes; pending != NULL; pending = next)
@@ -676,23 +677,21 @@ smgrDoPendingDeletes(bool isCommit)
/* do deletion if called for */
if (pending->atCommit == isCommit)
{
- SMgrRelation srel;
-
- srel = smgropen(pending->rlocator, pending->backend);
+ RelFileLocatorBackend rlocator = { pending->rlocator, pending->backend };
/* allocate the initial array, or extend it, if needed */
- if (maxrels == 0)
+ if (maxlocators == 0)
{
- maxrels = 8;
- srels = palloc(sizeof(SMgrRelation) * maxrels);
+ maxlocators = 8;
+ locators = palloc(sizeof(RelFileLocatorBackend) * maxlocators);
}
- else if (maxrels <= nrels)
+ else if (maxlocators <= nlocators)
{
- maxrels *= 2;
- srels = repalloc(srels, sizeof(SMgrRelation) * maxrels);
+ maxlocators *= 2;
+ locators = repalloc(locators, sizeof(RelFileLocatorBackend) * maxlocators);
}
- srels[nrels++] = srel;
+ locators[nlocators++] = rlocator;
}
/* must explicitly free the list entry */
pfree(pending);
@@ -700,15 +699,58 @@ smgrDoPendingDeletes(bool isCommit)
}
}
- if (nrels > 0)
+ if (nlocators > 0)
{
- smgrdounlinkall(srels, nrels, false);
+ ForkNumber forks[MAX_FORKNUM + 1];
+
+ for (int i = 0; i <= MAX_FORKNUM; i++)
+ forks[i] = i;
+
+ /*
+ * Get rid of any remaining buffers for the relations. bufmgr will just
+ * drop them without bothering to write the contents.
+ */
+ DropRelationsAllBuffers(locators, nlocators);
- for (int i = 0; i < nrels; i++)
- smgrclose(srels[i]);
+ for (int i = 0; i < nlocators; i++)
+ smgrunlink_multi(locators[i].locator, locators[i].backend, forks, MAX_FORKNUM + 1, false);
+ pfree(locators);
+ }
+}
+
+/*
+ * DropRelationFiles -- drop files of all given relations
+ */
+void
+DropRelationFiles(RelFileLocator *delrels, int ndelrels, bool isRedo)
+{
+ RelFileLocatorBackend *locators;
+ int i;
+ ForkNumber all_forks[MAX_FORKNUM + 1];
- pfree(srels);
+ locators = palloc(sizeof(RelFileLocatorBackend) * ndelrels);
+ for (i = 0; i < ndelrels; i++)
+ {
+ if (isRedo)
+ {
+ for (int fork = 0; fork <= MAX_FORKNUM; fork++)
+ XLogDropRelation(delrels[i], fork);
+ }
+ locators[i].locator = delrels[i];
+ locators[i].backend = InvalidBackendId;
}
+
+ /*
+ * Get rid of any remaining buffers for the relations. bufmgr will just
+ * drop them without bothering to write the contents.
+ */
+ DropRelationsAllBuffers(locators, ndelrels);
+
+ for (int fork = 0; fork <= MAX_FORKNUM; fork++)
+ all_forks[fork] = fork;
+
+ for (i = 0; i < ndelrels; i++)
+ smgrunlink_multi(locators[i].locator, locators[i].backend, all_forks, MAX_FORKNUM + 1, true);
}
/*
@@ -718,9 +760,9 @@ void
smgrDoPendingSyncs(bool isCommit, bool isParallelWorker)
{
PendingRelDelete *pending;
- int nrels = 0,
- maxrels = 0;
- SMgrRelation *srels = NULL;
+ int nlocators = 0,
+ maxlocators = 0;
+ RelFileLocator *locators = NULL;
HASH_SEQ_STATUS scan;
PendingRelSync *pendingsync;
@@ -757,9 +799,6 @@ smgrDoPendingSyncs(bool isCommit, bool isParallelWorker)
ForkNumber fork;
BlockNumber nblocks[MAX_FORKNUM + 1];
BlockNumber total_blocks = 0;
- SMgrRelation srel;
-
- srel = smgropen(pendingsync->rlocator, InvalidBackendId);
/*
* We emit newpage WAL records for smaller relations.
@@ -773,9 +812,12 @@ smgrDoPendingSyncs(bool isCommit, bool isParallelWorker)
{
for (fork = 0; fork <= MAX_FORKNUM; fork++)
{
- if (smgrexists(srel, fork))
+ SMgrFileHandle sfile;
+
+ sfile = smgropen(pendingsync->rlocator, InvalidBackendId, fork);
+ if (smgrexists(sfile))
{
- BlockNumber n = smgrnblocks(srel, fork);
+ BlockNumber n = smgrnblocks(sfile);
/* we shouldn't come here for unlogged relations */
Assert(fork != INIT_FORKNUM);
@@ -803,18 +845,19 @@ smgrDoPendingSyncs(bool isCommit, bool isParallelWorker)
total_blocks * BLCKSZ / 1024 >= wal_skip_threshold)
{
/* allocate the initial array, or extend it, if needed */
- if (maxrels == 0)
+ if (maxlocators == 0)
{
- maxrels = 8;
- srels = palloc(sizeof(SMgrRelation) * maxrels);
+ maxlocators = 8;
+ locators = palloc(sizeof(RelFileLocatorBackend) * maxlocators);
}
- else if (maxrels <= nrels)
+ else if (maxlocators <= nlocators)
{
- maxrels *= 2;
- srels = repalloc(srels, sizeof(SMgrRelation) * maxrels);
+ maxlocators *= 2;
+ locators = repalloc(locators, sizeof(RelFileLocatorBackend) * maxlocators);
}
- srels[nrels++] = srel;
+ locators[nlocators] = pendingsync->rlocator;
+ nlocators++;
}
else
{
@@ -833,7 +876,7 @@ smgrDoPendingSyncs(bool isCommit, bool isParallelWorker)
* page including any unused space. ReadBufferExtended()
* counts some pgstat events; unfortunately, we discard them.
*/
- rel = CreateFakeRelcacheEntry(srel->smgr_rlocator.locator);
+ rel = CreateFakeRelcacheEntry(pendingsync->rlocator);
log_newpage_range(rel, fork, 0, n, false);
FreeFakeRelcacheEntry(rel);
}
@@ -842,11 +885,20 @@ smgrDoPendingSyncs(bool isCommit, bool isParallelWorker)
pendingSyncHash = NULL;
- if (nrels > 0)
+ FlushRelationsAllBuffers(locators, nlocators);
+
+ for (int i = 0; i < nlocators; i++)
{
- smgrdosyncall(srels, nrels);
- pfree(srels);
+ for (int fork = 0; fork <= MAX_FORKNUM; fork++)
+ {
+ SMgrFileHandle sfile = smgropen(locators[i], InvalidBackendId, fork);
+
+ if (smgrexists(sfile))
+ smgrimmedsync(sfile);
+ }
}
+ if (locators != NULL)
+ pfree(locators);
}
/*
@@ -966,22 +1018,22 @@ smgr_redo(XLogReaderState *record)
if (info == XLOG_SMGR_CREATE)
{
xl_smgr_create *xlrec = (xl_smgr_create *) XLogRecGetData(record);
- SMgrRelation reln;
+ SMgrFileHandle sfile;
- reln = smgropen(xlrec->rlocator, InvalidBackendId);
- smgrcreate(reln, xlrec->forkNum, true);
+ sfile = smgropen(xlrec->rlocator, InvalidBackendId, xlrec->forkNum);
+ smgrcreate(sfile, true);
}
else if (info == XLOG_SMGR_TRUNCATE)
{
xl_smgr_truncate *xlrec = (xl_smgr_truncate *) XLogRecGetData(record);
- SMgrRelation reln;
+ SMgrFileHandle sfile;
Relation rel;
ForkNumber forks[MAX_FORKNUM];
BlockNumber blocks[MAX_FORKNUM];
int nforks = 0;
bool need_fsm_vacuum = false;
- reln = smgropen(xlrec->rlocator, InvalidBackendId);
+ sfile = smgropen(xlrec->rlocator, InvalidBackendId, MAIN_FORKNUM);
/*
* Forcibly create relation if it doesn't exist (which suggests that
@@ -989,7 +1041,7 @@ smgr_redo(XLogReaderState *record)
* XLogReadBufferForRedo, we prefer to recreate the rel and replay the
* log as best we can until the drop is seen.
*/
- smgrcreate(reln, MAIN_FORKNUM, true);
+ smgrcreate(sfile, true);
/*
* Before we perform the truncation, update minimum recovery point to
@@ -1022,8 +1074,10 @@ smgr_redo(XLogReaderState *record)
/* Prepare for truncation of FSM and VM too */
rel = CreateFakeRelcacheEntry(xlrec->rlocator);
+ DropRelationBuffers(xlrec->rlocator, InvalidBackendId, forks, nforks, blocks);
+
if ((xlrec->flags & SMGR_TRUNCATE_FSM) != 0 &&
- smgrexists(reln, FSM_FORKNUM))
+ smgrexists(smgropen(xlrec->rlocator, InvalidBackendId, FSM_FORKNUM)))
{
blocks[nforks] = FreeSpaceMapPrepareTruncateRel(rel, xlrec->blkno);
if (BlockNumberIsValid(blocks[nforks]))
@@ -1034,7 +1088,7 @@ smgr_redo(XLogReaderState *record)
}
}
if ((xlrec->flags & SMGR_TRUNCATE_VM) != 0 &&
- smgrexists(reln, VISIBILITYMAP_FORKNUM))
+ smgrexists(smgropen(xlrec->rlocator, InvalidBackendId, VISIBILITYMAP_FORKNUM)))
{
blocks[nforks] = visibilitymap_prepare_truncate(rel, xlrec->blkno);
if (BlockNumberIsValid(blocks[nforks]))
@@ -1046,7 +1100,10 @@ smgr_redo(XLogReaderState *record)
/* Do the real work to truncate relation forks */
if (nforks > 0)
- smgrtruncate(reln, forks, nforks, blocks);
+ {
+ DropRelationBuffers(xlrec->rlocator, InvalidBackendId, forks, nforks, blocks);
+ smgrtruncate_multi(xlrec->rlocator, InvalidBackendId, forks, nforks, blocks);
+ }
/*
* Update upper-level FSM pages to account for the truncation. This is
diff --git a/src/backend/commands/dbcommands.c b/src/backend/commands/dbcommands.c
index a0259cc5938..0d9ed2e0e0c 100644
--- a/src/backend/commands/dbcommands.c
+++ b/src/backend/commands/dbcommands.c
@@ -258,7 +258,7 @@ ScanSourceDatabasePgClass(Oid tbid, Oid dbid, char *srcpath)
List *rlocatorlist = NIL;
LockRelId relid;
Snapshot snapshot;
- SMgrRelation smgr;
+ SMgrFileHandle sfile;
BufferAccessStrategy bstrategy;
/* Get pg_class relfilenumber. */
@@ -275,9 +275,9 @@ ScanSourceDatabasePgClass(Oid tbid, Oid dbid, char *srcpath)
rlocator.dbOid = dbid;
rlocator.relNumber = relfilenumber;
- smgr = smgropen(rlocator, InvalidBackendId);
- nblocks = smgrnblocks(smgr, MAIN_FORKNUM);
- smgrclose(smgr);
+ sfile = smgropen(rlocator, InvalidBackendId, MAIN_FORKNUM);
+ nblocks = smgrnblocks(sfile);
+ smgrclose(sfile);
/* Use a buffer access strategy since this is a bulk read operation. */
bstrategy = GetAccessStrategy(BAS_BULKREAD);
@@ -439,7 +439,7 @@ ScanSourceDatabasePgClassTuple(HeapTupleData *tuple, Oid tbid, Oid dbid,
relinfo->rlocator.relNumber = relfilenumber;
relinfo->reloid = classForm->oid;
- /* Temporary relations were rejected above. */
+ /* Temporary relations were rejected ove. */
Assert(classForm->relpersistence != RELPERSISTENCE_TEMP);
relinfo->permanent =
(classForm->relpersistence == RELPERSISTENCE_PERMANENT) ? true : false;
diff --git a/src/backend/commands/sequence.c b/src/backend/commands/sequence.c
index bfe279cddf1..460e2bb1822 100644
--- a/src/backend/commands/sequence.c
+++ b/src/backend/commands/sequence.c
@@ -353,14 +353,14 @@ fill_seq_with_data(Relation rel, HeapTuple tuple)
if (rel->rd_rel->relpersistence == RELPERSISTENCE_UNLOGGED)
{
- SMgrRelation srel;
+ SMgrFileHandle sfile;
- srel = smgropen(rel->rd_locator, InvalidBackendId);
- smgrcreate(srel, INIT_FORKNUM, false);
+ sfile = smgropen(rel->rd_locator, InvalidBackendId, INIT_FORKNUM);
+ smgrcreate(sfile, false);
log_smgrcreate(&rel->rd_locator, INIT_FORKNUM);
fill_seq_fork_with_data(rel, tuple, INIT_FORKNUM);
FlushRelationBuffers(rel);
- smgrclose(srel);
+ smgrclose(sfile);
}
}
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 62d9917ca36..2033aa3e35e 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -14665,9 +14665,7 @@ AlterTableMoveAll(AlterTableMoveAllStmt *stmt)
static void
index_copy_data(Relation rel, RelFileLocator newrlocator)
{
- SMgrRelation dstrel;
-
- dstrel = smgropen(newrlocator, rel->rd_backend);
+ SMgrFileHandle dstmain;
/*
* Since we copy the file directly without looking at the shared buffers,
@@ -14687,16 +14685,20 @@ index_copy_data(Relation rel, RelFileLocator newrlocator)
RelationCreateStorage(newrlocator, rel->rd_rel->relpersistence, true);
/* copy main fork */
- RelationCopyStorage(RelationGetSmgr(rel), dstrel, MAIN_FORKNUM,
+ dstmain = smgropen(newrlocator, rel->rd_backend, MAIN_FORKNUM);
+ RelationCopyStorage(RelationGetSmgr(rel, MAIN_FORKNUM), dstmain,
rel->rd_rel->relpersistence);
/* copy those extra forks that exist */
for (ForkNumber forkNum = MAIN_FORKNUM + 1;
forkNum <= MAX_FORKNUM; forkNum++)
{
- if (smgrexists(RelationGetSmgr(rel), forkNum))
+ if (smgrexists(RelationGetSmgr(rel, forkNum)))
{
- smgrcreate(dstrel, forkNum, false);
+ SMgrFileHandle src_fork = RelationGetSmgr(rel, forkNum);
+ SMgrFileHandle dst_fork = smgropen(newrlocator, rel->rd_backend, forkNum);
+
+ smgrcreate(dst_fork, false);
/*
* WAL log creation if the relation is persistent, or this is the
@@ -14706,14 +14708,15 @@ index_copy_data(Relation rel, RelFileLocator newrlocator)
(rel->rd_rel->relpersistence == RELPERSISTENCE_UNLOGGED &&
forkNum == INIT_FORKNUM))
log_smgrcreate(&newrlocator, forkNum);
- RelationCopyStorage(RelationGetSmgr(rel), dstrel, forkNum,
+ RelationCopyStorage(src_fork, dst_fork,
rel->rd_rel->relpersistence);
+ smgrclose(dst_fork);
}
}
/* drop old relation, and close new one */
RelationDropStorage(rel);
- smgrclose(dstrel);
+ smgrclose(dstmain);
}
/*
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 0a05577b68d..1305eb7dee1 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -117,19 +117,6 @@ typedef struct CkptTsStatus
int index;
} CkptTsStatus;
-/*
- * Type for array used to sort SMgrRelations
- *
- * FlushRelationsAllBuffers shares the same comparator function with
- * DropRelationsAllBuffers. Pointer to this struct and RelFileLocator must be
- * compatible.
- */
-typedef struct SMgrSortArray
-{
- RelFileLocator rlocator; /* This must be the first member */
- SMgrRelation srel;
-} SMgrSortArray;
-
/* GUC variables */
bool zero_damaged_pages = false;
int bgwriter_lru_maxpages = 100;
@@ -450,8 +437,8 @@ ForgetPrivateRefCountEntry(PrivateRefCountEntry *ref)
)
-static Buffer ReadBuffer_common(SMgrRelation smgr, char relpersistence,
- ForkNumber forkNum, BlockNumber blockNum,
+static Buffer ReadBuffer_common(SMgrFileHandle sfile, char relpersistence,
+ BlockNumber blockNum,
ReadBufferMode mode, BufferAccessStrategy strategy,
bool *hit);
static bool PinBuffer(BufferDesc *buf, BufferAccessStrategy strategy);
@@ -467,13 +454,12 @@ static void TerminateBufferIO(BufferDesc *buf, bool clear_dirty,
uint32 set_flag_bits);
static void shared_buffer_write_error_callback(void *arg);
static void local_buffer_write_error_callback(void *arg);
-static BufferDesc *BufferAlloc(SMgrRelation smgr,
+static BufferDesc *BufferAlloc(SMgrFileHandle sfile,
char relpersistence,
- ForkNumber forkNum,
BlockNumber blockNum,
BufferAccessStrategy strategy,
bool *foundPtr, IOContext *io_context);
-static void FlushBuffer(BufferDesc *buf, SMgrRelation reln,
+static void FlushBuffer(BufferDesc *buf, SMgrFileHandle sfile,
IOObject io_object, IOContext io_context);
static void FindAndDropRelationBuffers(RelFileLocator rlocator,
ForkNumber forkNum,
@@ -494,9 +480,7 @@ static int ts_ckpt_progress_comparator(Datum a, Datum b, void *arg);
* Implementation of PrefetchBuffer() for shared buffers.
*/
PrefetchBufferResult
-PrefetchSharedBuffer(SMgrRelation smgr_reln,
- ForkNumber forkNum,
- BlockNumber blockNum)
+PrefetchSharedBuffer(SMgrFileHandle sfile, BlockNumber blockNum)
{
PrefetchBufferResult result = {InvalidBuffer, false};
BufferTag newTag; /* identity of requested block */
@@ -507,8 +491,8 @@ PrefetchSharedBuffer(SMgrRelation smgr_reln,
Assert(BlockNumberIsValid(blockNum));
/* create a tag so we can lookup the buffer */
- InitBufferTag(&newTag, &smgr_reln->smgr_rlocator.locator,
- forkNum, blockNum);
+ InitBufferTag(&newTag, &sfile->smgr_locator.locator,
+ sfile->smgr_locator.forknum, blockNum);
/* determine its hash code and partition lock ID */
newHash = BufTableHashCode(&newTag);
@@ -527,7 +511,7 @@ PrefetchSharedBuffer(SMgrRelation smgr_reln,
* Try to initiate an asynchronous read. This returns false in
* recovery if the relation file doesn't exist.
*/
- if (smgrprefetch(smgr_reln, forkNum, blockNum))
+ if (smgrprefetch(sfile, blockNum))
result.initiated_io = true;
#endif /* USE_PREFETCH */
}
@@ -595,12 +579,12 @@ PrefetchBuffer(Relation reln, ForkNumber forkNum, BlockNumber blockNum)
errmsg("cannot access temporary tables of other sessions")));
/* pass it off to localbuf.c */
- return PrefetchLocalBuffer(RelationGetSmgr(reln), forkNum, blockNum);
+ return PrefetchLocalBuffer(RelationGetSmgr(reln, forkNum), blockNum);
}
else
{
/* pass it to the shared buffer version */
- return PrefetchSharedBuffer(RelationGetSmgr(reln), forkNum, blockNum);
+ return PrefetchSharedBuffer(RelationGetSmgr(reln, forkNum), blockNum);
}
}
@@ -769,8 +753,9 @@ ReadBufferExtended(Relation reln, ForkNumber forkNum, BlockNumber blockNum,
* miss.
*/
pgstat_count_buffer_read(reln);
- buf = ReadBuffer_common(RelationGetSmgr(reln), reln->rd_rel->relpersistence,
- forkNum, blockNum, mode, strategy, &hit);
+ buf = ReadBuffer_common(RelationGetSmgr(reln, forkNum),
+ reln->rd_rel->relpersistence,
+ blockNum, mode, strategy, &hit);
if (hit)
pgstat_count_buffer_hit(reln);
return buf;
@@ -794,10 +779,10 @@ ReadBufferWithoutRelcache(RelFileLocator rlocator, ForkNumber forkNum,
{
bool hit;
- SMgrRelation smgr = smgropen(rlocator, InvalidBackendId);
+ SMgrFileHandle sfile = smgropen(rlocator, InvalidBackendId, forkNum);
- return ReadBuffer_common(smgr, permanent ? RELPERSISTENCE_PERMANENT :
- RELPERSISTENCE_UNLOGGED, forkNum, blockNum,
+ return ReadBuffer_common(sfile, permanent ? RELPERSISTENCE_PERMANENT :
+ RELPERSISTENCE_UNLOGGED, blockNum,
mode, strategy, &hit);
}
@@ -808,7 +793,7 @@ ReadBufferWithoutRelcache(RelFileLocator rlocator, ForkNumber forkNum,
* *hit is set to true if the request was satisfied from shared buffer cache.
*/
static Buffer
-ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
+ReadBuffer_common(SMgrFileHandle sfile, char relpersistence,
BlockNumber blockNum, ReadBufferMode mode,
BufferAccessStrategy strategy, bool *hit)
{
@@ -818,7 +803,7 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
IOContext io_context;
IOObject io_object;
bool isExtend;
- bool isLocalBuf = SmgrIsTemp(smgr);
+ bool isLocalBuf = SmgrIsTemp(sfile);
*hit = false;
@@ -827,23 +812,24 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
isExtend = (blockNum == P_NEW);
- TRACE_POSTGRESQL_BUFFER_READ_START(forkNum, blockNum,
- smgr->smgr_rlocator.locator.spcOid,
- smgr->smgr_rlocator.locator.dbOid,
- smgr->smgr_rlocator.locator.relNumber,
- smgr->smgr_rlocator.backend,
+ TRACE_POSTGRESQL_BUFFER_READ_START(smgr->smgr_locator.forkNum,
+ blockNum,
+ smgr->smgr_locator.locator.spcOid,
+ smgr->smgr_locator.locator.dbOid,
+ smgr->smgr_locator.locator.relNumber,
+ smgr->smgr_locator.backend,
isExtend);
/* Substitute proper block number if caller asked for P_NEW */
if (isExtend)
{
- blockNum = smgrnblocks(smgr, forkNum);
+ blockNum = smgrnblocks(sfile);
/* Fail if relation is already at maximum possible length */
if (blockNum == P_NEW)
ereport(ERROR,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
errmsg("cannot extend relation %s beyond %u blocks",
- relpath(smgr->smgr_rlocator, forkNum),
+ smgrfilepath(sfile->smgr_locator),
P_NEW)));
}
@@ -856,7 +842,7 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
* rely on IOContextForStrategy() to set the right IOContext for us.
* This may happen in cases like CREATE TEMPORARY TABLE AS...
*/
- bufHdr = LocalBufferAlloc(smgr, forkNum, blockNum, &found, &io_context);
+ bufHdr = LocalBufferAlloc(sfile, blockNum, &found, &io_context);
if (found)
pgBufferUsage.local_blks_hit++;
else if (isExtend)
@@ -871,7 +857,7 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
* lookup the buffer. IO_IN_PROGRESS is set if the requested block is
* not currently in memory.
*/
- bufHdr = BufferAlloc(smgr, relpersistence, forkNum, blockNum,
+ bufHdr = BufferAlloc(sfile, relpersistence, blockNum,
strategy, &found, &io_context);
if (found)
pgBufferUsage.shared_blks_hit++;
@@ -896,11 +882,12 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
if (VacuumCostActive)
VacuumCostBalance += VacuumCostPageHit;
- TRACE_POSTGRESQL_BUFFER_READ_DONE(forkNum, blockNum,
- smgr->smgr_rlocator.locator.spcOid,
- smgr->smgr_rlocator.locator.dbOid,
- smgr->smgr_rlocator.locator.relNumber,
- smgr->smgr_rlocator.backend,
+ TRACE_POSTGRESQL_BUFFER_READ_DONE(sfile->smgr_forknum,
+ blockNum,
+ sfile->smgr_rlocator.locator.spcOid,
+ sfile->smgr_rlocator.locator.dbOid,
+ sfile->smgr_rlocator.locator.relNumber,
+ sfile->smgr_rlocator.backend,
isExtend,
found);
@@ -937,7 +924,7 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
if (!PageIsNew((Page) bufBlock))
ereport(ERROR,
(errmsg("unexpected data beyond EOF in block %u of relation %s",
- blockNum, relpath(smgr->smgr_rlocator, forkNum)),
+ blockNum, smgrfilepath(sfile->smgr_locator)),
errhint("This has been seen to occur with buggy kernels; consider updating your system.")));
/*
@@ -1003,7 +990,7 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
/* new buffers are zero-filled */
MemSet((char *) bufBlock, 0, BLCKSZ);
/* don't set checksum for all-zero page */
- smgrextend(smgr, forkNum, blockNum, bufBlock, false);
+ smgrextend(sfile, blockNum, bufBlock, false);
pgstat_count_io_op(io_object, io_context, IOOP_EXTEND);
@@ -1032,7 +1019,7 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
else
INSTR_TIME_SET_ZERO(io_start);
- smgrread(smgr, forkNum, blockNum, bufBlock);
+ smgrread(sfile, blockNum, bufBlock);
pgstat_count_io_op(io_object, io_context, IOOP_READ);
@@ -1054,7 +1041,7 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
(errcode(ERRCODE_DATA_CORRUPTED),
errmsg("invalid page in block %u of relation %s; zeroing out page",
blockNum,
- relpath(smgr->smgr_rlocator, forkNum))));
+ smgrfilepath(sfile->smgr_locator))));
MemSet((char *) bufBlock, 0, BLCKSZ);
}
else
@@ -1062,7 +1049,7 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
(errcode(ERRCODE_DATA_CORRUPTED),
errmsg("invalid page in block %u of relation %s",
blockNum,
- relpath(smgr->smgr_rlocator, forkNum))));
+ smgrfilepath(sfile->smgr_locator))));
}
}
}
@@ -1101,11 +1088,12 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
if (VacuumCostActive)
VacuumCostBalance += VacuumCostPageMiss;
- TRACE_POSTGRESQL_BUFFER_READ_DONE(forkNum, blockNum,
- smgr->smgr_rlocator.locator.spcOid,
- smgr->smgr_rlocator.locator.dbOid,
- smgr->smgr_rlocator.locator.relNumber,
- smgr->smgr_rlocator.backend,
+ TRACE_POSTGRESQL_BUFFER_READ_DONE(sfile->smgr_locator.forknum,
+ blockNum,
+ sfile->smgr_locator.locator.spcOid,
+ sfile->smgr_locator.locator.dbOid,
+ sfile->smgr_locator.locator.relNumber,
+ sfile->smgr_locator.backend,
isExtend,
found);
@@ -1136,7 +1124,7 @@ ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
* No locks are held either at entry or exit.
*/
static BufferDesc *
-BufferAlloc(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
+BufferAlloc(SMgrFileHandle sfile, char relpersistence,
BlockNumber blockNum,
BufferAccessStrategy strategy,
bool *foundPtr, IOContext *io_context)
@@ -1155,7 +1143,7 @@ BufferAlloc(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
uint32 buf_state;
/* create a tag so we can lookup the buffer */
- InitBufferTag(&newTag, &smgr->smgr_rlocator.locator, forkNum, blockNum);
+ InitBufferTag(&newTag, &sfile->smgr_locator.locator, sfile->smgr_locator.forknum, blockNum);
/* determine its hash code and partition lock ID */
newHash = BufTableHashCode(&newTag);
@@ -1290,10 +1278,11 @@ BufferAlloc(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
}
/* OK, do the I/O */
- TRACE_POSTGRESQL_BUFFER_WRITE_DIRTY_START(forkNum, blockNum,
- smgr->smgr_rlocator.locator.spcOid,
- smgr->smgr_rlocator.locator.dbOid,
- smgr->smgr_rlocator.locator.relNumber);
+ TRACE_POSTGRESQL_BUFFER_WRITE_DIRTY_START(sfile->smgr_locator.forknum,
+ blockNum,
+ sfile->smgr_locator.locator.spcOid,
+ sfile->smgr_locator.locator.dbOid,
+ sfile->smgr_locator.locator.relNumber);
FlushBuffer(buf, NULL, IOOBJECT_RELATION, *io_context);
LWLockRelease(BufferDescriptorGetContentLock(buf));
@@ -1301,10 +1290,11 @@ BufferAlloc(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
ScheduleBufferTagForWriteback(&BackendWritebackContext,
&buf->tag);
- TRACE_POSTGRESQL_BUFFER_WRITE_DIRTY_DONE(forkNum, blockNum,
- smgr->smgr_rlocator.locator.spcOid,
- smgr->smgr_rlocator.locator.dbOid,
- smgr->smgr_rlocator.locator.relNumber);
+ TRACE_POSTGRESQL_BUFFER_WRITE_DIRTY_DONE(sfile->smgr_locator.forknum,
+ blockNum,
+ sfile->smgr_locator.locator.spcOid,
+ sfile->smgr_locator.locator.dbOid,
+ sfile->smgr_locator.locator.relNumber);
}
else
{
@@ -1460,7 +1450,7 @@ BufferAlloc(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
buf_state &= ~(BM_VALID | BM_DIRTY | BM_JUST_DIRTIED |
BM_CHECKPOINT_NEEDED | BM_IO_ERROR | BM_PERMANENT |
BUF_USAGECOUNT_MASK);
- if (relpersistence == RELPERSISTENCE_PERMANENT || forkNum == INIT_FORKNUM)
+ if (relpersistence == RELPERSISTENCE_PERMANENT || sfile->smgr_locator.forknum == INIT_FORKNUM)
buf_state |= BM_TAG_VALID | BM_PERMANENT | BUF_USAGECOUNT_ONE;
else
buf_state |= BM_TAG_VALID | BUF_USAGECOUNT_ONE;
@@ -2868,7 +2858,7 @@ BufferGetTag(Buffer buffer, RelFileLocator *rlocator, ForkNumber *forknum,
* as the second parameter. If not, pass NULL.
*/
static void
-FlushBuffer(BufferDesc *buf, SMgrRelation reln, IOObject io_object,
+FlushBuffer(BufferDesc *buf, SMgrFileHandle sfile, IOObject io_object,
IOContext io_context)
{
XLogRecPtr recptr;
@@ -2894,14 +2884,14 @@ FlushBuffer(BufferDesc *buf, SMgrRelation reln, IOObject io_object,
error_context_stack = &errcallback;
/* Find smgr relation for buffer */
- if (reln == NULL)
- reln = smgropen(BufTagGetRelFileLocator(&buf->tag), InvalidBackendId);
+ if (sfile == NULL)
+ sfile = smgropen(BufTagGetRelFileLocator(&buf->tag), InvalidBackendId, BufTagGetForkNum(&buf->tag));
TRACE_POSTGRESQL_BUFFER_FLUSH_START(BufTagGetForkNum(&buf->tag),
buf->tag.blockNum,
- reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
- reln->smgr_rlocator.locator.relNumber);
+ sfile->smgr_locator.locator.spcOid,
+ sfile->smgr_locator.locator.dbOid,
+ sfile->smgr_locator.locator.relNumber);
buf_state = LockBufHdr(buf);
@@ -2957,8 +2947,7 @@ FlushBuffer(BufferDesc *buf, SMgrRelation reln, IOObject io_object,
/*
* bufToWrite is either the shared buffer or a copy, as appropriate.
*/
- smgrwrite(reln,
- BufTagGetForkNum(&buf->tag),
+ smgrwrite(sfile,
buf->tag.blockNum,
bufToWrite,
false);
@@ -3001,9 +2990,9 @@ FlushBuffer(BufferDesc *buf, SMgrRelation reln, IOObject io_object,
TRACE_POSTGRESQL_BUFFER_FLUSH_DONE(BufTagGetForkNum(&buf->tag),
buf->tag.blockNum,
- reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
- reln->smgr_rlocator.locator.relNumber);
+ sfile->smgr_locator.locator.spcOid,
+ sfile->smgr_locator.locator.dbOid,
+ sfile->smgr_locator.locator.relNumber);
/* Pop the error context stack */
error_context_stack = errcallback.previous;
@@ -3036,7 +3025,7 @@ RelationGetNumberOfBlocksInFork(Relation relation, ForkNumber forkNum)
}
else if (RELKIND_HAS_STORAGE(relation->rd_rel->relkind))
{
- return smgrnblocks(RelationGetSmgr(relation), forkNum);
+ return smgrnblocks(RelationGetSmgr(relation, forkNum));
}
else
Assert(false);
@@ -3126,26 +3115,20 @@ BufferGetLSNAtomic(Buffer buffer)
* --------------------------------------------------------------------
*/
void
-DropRelationBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum,
+DropRelationBuffers(RelFileLocator rlocator, BackendId backend, ForkNumber *forkNum,
int nforks, BlockNumber *firstDelBlock)
{
int i;
int j;
- RelFileLocatorBackend rlocator;
BlockNumber nForkBlock[MAX_FORKNUM];
uint64 nBlocksToInvalidate = 0;
- rlocator = smgr_reln->smgr_rlocator;
-
/* If it's a local relation, it's localbuf.c's problem. */
- if (RelFileLocatorBackendIsTemp(rlocator))
+ if (backend == MyBackendId)
{
- if (rlocator.backend == MyBackendId)
- {
- for (j = 0; j < nforks; j++)
- DropRelationLocalBuffers(rlocator.locator, forkNum[j],
- firstDelBlock[j]);
- }
+ for (j = 0; j < nforks; j++)
+ DropRelationLocalBuffers(rlocator, forkNum[j],
+ firstDelBlock[j]);
return;
}
@@ -3174,7 +3157,10 @@ DropRelationBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum,
for (i = 0; i < nforks; i++)
{
/* Get the number of blocks for a relation's fork */
- nForkBlock[i] = smgrnblocks_cached(smgr_reln, forkNum[i]);
+ SMgrFileHandle sfile;
+
+ sfile = smgropen(rlocator, backend, forkNum[i]);
+ nForkBlock[i] = smgrnblocks_cached(sfile);
if (nForkBlock[i] == InvalidBlockNumber)
{
@@ -3194,7 +3180,7 @@ DropRelationBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum,
nBlocksToInvalidate < BUF_DROP_FULL_SCAN_THRESHOLD)
{
for (j = 0; j < nforks; j++)
- FindAndDropRelationBuffers(rlocator.locator, forkNum[j],
+ FindAndDropRelationBuffers(rlocator, forkNum[j],
nForkBlock[j], firstDelBlock[j]);
return;
}
@@ -3220,14 +3206,14 @@ DropRelationBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum,
* We could check forkNum and blockNum as well as the rlocator, but
* the incremental win from doing so seems small.
*/
- if (!BufTagMatchesRelFileLocator(&bufHdr->tag, &rlocator.locator))
+ if (!BufTagMatchesRelFileLocator(&bufHdr->tag, &rlocator))
continue;
buf_state = LockBufHdr(bufHdr);
for (j = 0; j < nforks; j++)
{
- if (BufTagMatchesRelFileLocator(&bufHdr->tag, &rlocator.locator) &&
+ if (BufTagMatchesRelFileLocator(&bufHdr->tag, &rlocator) &&
BufTagGetForkNum(&bufHdr->tag) == forkNum[j] &&
bufHdr->tag.blockNum >= firstDelBlock[j])
{
@@ -3249,11 +3235,10 @@ DropRelationBuffers(SMgrRelation smgr_reln, ForkNumber *forkNum,
* --------------------------------------------------------------------
*/
void
-DropRelationsAllBuffers(SMgrRelation *smgr_reln, int nlocators)
+DropRelationsAllBuffers(RelFileLocatorBackend *rlocators, int nlocators)
{
int i;
int n = 0;
- SMgrRelation *rels;
BlockNumber (*block)[MAX_FORKNUM + 1];
uint64 nBlocksToInvalidate = 0;
RelFileLocator *locators;
@@ -3263,18 +3248,18 @@ DropRelationsAllBuffers(SMgrRelation *smgr_reln, int nlocators)
if (nlocators == 0)
return;
- rels = palloc(sizeof(SMgrRelation) * nlocators); /* non-local relations */
+ locators = palloc(sizeof(RelFileLocator) * nlocators); /* non-local relations */
/* If it's a local relation, it's localbuf.c's problem. */
for (i = 0; i < nlocators; i++)
{
- if (RelFileLocatorBackendIsTemp(smgr_reln[i]->smgr_rlocator))
+ if (rlocators[i].backend != InvalidBackendId)
{
- if (smgr_reln[i]->smgr_rlocator.backend == MyBackendId)
- DropRelationAllLocalBuffers(smgr_reln[i]->smgr_rlocator.locator);
+ if (rlocators[i].backend == MyBackendId)
+ DropRelationAllLocalBuffers(rlocators[i].locator);
}
else
- rels[n++] = smgr_reln[i];
+ locators[n++] = rlocators[i].locator;
}
/*
@@ -3283,7 +3268,7 @@ DropRelationsAllBuffers(SMgrRelation *smgr_reln, int nlocators)
*/
if (n == 0)
{
- pfree(rels);
+ pfree(locators);
return;
}
@@ -3303,12 +3288,13 @@ DropRelationsAllBuffers(SMgrRelation *smgr_reln, int nlocators)
for (int j = 0; j <= MAX_FORKNUM; j++)
{
/* Get the number of blocks for a relation's fork. */
- block[i][j] = smgrnblocks_cached(rels[i], j);
+ SMgrFileHandle sfile = smgropen(locators[i], InvalidBackendId, j);
+ block[i][j] = smgrnblocks_cached(sfile);
/* We need to only consider the relation forks that exists. */
if (block[i][j] == InvalidBlockNumber)
{
- if (!smgrexists(rels[i], j))
+ if (!smgrexists(sfile))
continue;
cached = false;
break;
@@ -3334,20 +3320,17 @@ DropRelationsAllBuffers(SMgrRelation *smgr_reln, int nlocators)
continue;
/* drop all the buffers for a particular relation fork */
- FindAndDropRelationBuffers(rels[i]->smgr_rlocator.locator,
+ FindAndDropRelationBuffers(locators[i],
j, block[i][j], 0);
}
}
pfree(block);
- pfree(rels);
+ pfree(locators);
return;
}
pfree(block);
- locators = palloc(sizeof(RelFileLocator) * n); /* non-local relations */
- for (i = 0; i < n; i++)
- locators[i] = rels[i]->smgr_rlocator.locator;
/*
* For low number of relations to drop just use a simple walk through, to
@@ -3407,7 +3390,6 @@ DropRelationsAllBuffers(SMgrRelation *smgr_reln, int nlocators)
}
pfree(locators);
- pfree(rels);
}
/* ---------------------------------------------------------------------
@@ -3616,8 +3598,7 @@ FlushRelationBuffers(Relation rel)
PageSetChecksumInplace(localpage, bufHdr->tag.blockNum);
- smgrwrite(RelationGetSmgr(rel),
- BufTagGetForkNum(&bufHdr->tag),
+ smgrwrite(RelationGetSmgr(rel, BufTagGetForkNum(&bufHdr->tag)),
bufHdr->tag.blockNum,
localpage,
false);
@@ -3659,7 +3640,8 @@ FlushRelationBuffers(Relation rel)
{
PinBuffer_Locked(bufHdr);
LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), LW_SHARED);
- FlushBuffer(bufHdr, RelationGetSmgr(rel), IOOBJECT_RELATION, IOCONTEXT_NORMAL);
+ FlushBuffer(bufHdr, RelationGetSmgr(rel, BufTagGetForkNum(&bufHdr->tag)),
+ IOOBJECT_RELATION, IOCONTEXT_NORMAL);
LWLockRelease(BufferDescriptorGetContentLock(bufHdr));
UnpinBuffer(bufHdr);
}
@@ -3678,25 +3660,20 @@ FlushRelationBuffers(Relation rel)
* --------------------------------------------------------------------
*/
void
-FlushRelationsAllBuffers(SMgrRelation *smgrs, int nrels)
+FlushRelationsAllBuffers(RelFileLocator *rels, int nrels)
{
int i;
- SMgrSortArray *srels;
+ RelFileLocator *locators = NULL;
bool use_bsearch;
if (nrels == 0)
return;
/* fill-in array for qsort */
- srels = palloc(sizeof(SMgrSortArray) * nrels);
+ locators = palloc(sizeof(RelFileLocator) * nrels);
for (i = 0; i < nrels; i++)
- {
- Assert(!RelFileLocatorBackendIsTemp(smgrs[i]->smgr_rlocator));
-
- srels[i].rlocator = smgrs[i]->smgr_rlocator.locator;
- srels[i].srel = smgrs[i];
- }
+ locators[i] = rels[i];
/*
* Save the bsearch overhead for low number of relations to sync. See
@@ -3704,16 +3681,16 @@ FlushRelationsAllBuffers(SMgrRelation *smgrs, int nrels)
*/
use_bsearch = nrels > RELS_BSEARCH_THRESHOLD;
- /* sort the list of SMgrRelations if necessary */
+ /* sort the list of locators if necessary */
if (use_bsearch)
- pg_qsort(srels, nrels, sizeof(SMgrSortArray), rlocator_comparator);
+ pg_qsort(locators, nrels, sizeof(RelFileLocator), rlocator_comparator);
/* Make sure we can handle the pin inside the loop */
ResourceOwnerEnlargeBuffers(CurrentResourceOwner);
for (i = 0; i < NBuffers; i++)
{
- SMgrSortArray *srelent = NULL;
+ RelFileLocator *found = NULL;
BufferDesc *bufHdr = GetBufferDescriptor(i);
uint32 buf_state;
@@ -3728,9 +3705,9 @@ FlushRelationsAllBuffers(SMgrRelation *smgrs, int nrels)
for (j = 0; j < nrels; j++)
{
- if (BufTagMatchesRelFileLocator(&bufHdr->tag, &srels[j].rlocator))
+ if (BufTagMatchesRelFileLocator(&bufHdr->tag, &locators[j]))
{
- srelent = &srels[j];
+ found = &locators[j];
break;
}
}
@@ -3740,24 +3717,26 @@ FlushRelationsAllBuffers(SMgrRelation *smgrs, int nrels)
RelFileLocator rlocator;
rlocator = BufTagGetRelFileLocator(&bufHdr->tag);
- srelent = bsearch((const void *) &(rlocator),
- srels, nrels, sizeof(SMgrSortArray),
- rlocator_comparator);
+ found = bsearch((const void *) &rlocator,
+ locators, nrels, sizeof(RelFileLocator),
+ rlocator_comparator);
}
/* buffer doesn't belong to any of the given relfilelocators; skip it */
- if (srelent == NULL)
+ if (found == NULL)
continue;
+ /* FIXME: cache SMgrFileHandles for the rels, and pass to FlushBuffer */
+
ReservePrivateRefCountEntry();
buf_state = LockBufHdr(bufHdr);
- if (BufTagMatchesRelFileLocator(&bufHdr->tag, &srelent->rlocator) &&
+ if (BufTagMatchesRelFileLocator(&bufHdr->tag, found) &&
(buf_state & (BM_VALID | BM_DIRTY)) == (BM_VALID | BM_DIRTY))
{
PinBuffer_Locked(bufHdr);
LWLockAcquire(BufferDescriptorGetContentLock(bufHdr), LW_SHARED);
- FlushBuffer(bufHdr, srelent->srel, IOOBJECT_RELATION, IOCONTEXT_NORMAL);
+ FlushBuffer(bufHdr, NULL, IOOBJECT_RELATION, IOCONTEXT_NORMAL);
LWLockRelease(BufferDescriptorGetContentLock(bufHdr));
UnpinBuffer(bufHdr);
}
@@ -3765,7 +3744,7 @@ FlushRelationsAllBuffers(SMgrRelation *smgrs, int nrels)
UnlockBufHdr(bufHdr, buf_state);
}
- pfree(srels);
+ pfree(locators);
}
/* ---------------------------------------------------------------------
@@ -3802,8 +3781,7 @@ RelationCopyStorageUsingBuffer(RelFileLocator srclocator,
use_wal = XLogIsNeeded() && (permanent || forkNum == INIT_FORKNUM);
/* Get number of blocks in the source relation. */
- nblocks = smgrnblocks(smgropen(srclocator, InvalidBackendId),
- forkNum);
+ nblocks = smgrnblocks(smgropen(srclocator, InvalidBackendId, forkNum));
/* Nothing to copy; just return. */
if (nblocks == 0)
@@ -3814,7 +3792,7 @@ RelationCopyStorageUsingBuffer(RelFileLocator srclocator,
* relation before starting to copy block by block.
*/
memset(buf.data, 0, BLCKSZ);
- smgrextend(smgropen(dstlocator, InvalidBackendId), forkNum, nblocks - 1,
+ smgrextend(smgropen(dstlocator, InvalidBackendId, forkNum), nblocks - 1,
buf.data, true);
/* This is a bulk operation, so use buffer access strategies. */
@@ -3872,7 +3850,6 @@ void
CreateAndCopyRelationData(RelFileLocator src_rlocator,
RelFileLocator dst_rlocator, bool permanent)
{
- RelFileLocatorBackend rlocator;
char relpersistence;
/* Set the relpersistence. */
@@ -3895,9 +3872,9 @@ CreateAndCopyRelationData(RelFileLocator src_rlocator,
for (ForkNumber forkNum = MAIN_FORKNUM + 1;
forkNum <= MAX_FORKNUM; forkNum++)
{
- if (smgrexists(smgropen(src_rlocator, InvalidBackendId), forkNum))
+ if (smgrexists(smgropen(src_rlocator, InvalidBackendId, forkNum)))
{
- smgrcreate(smgropen(dst_rlocator, InvalidBackendId), forkNum, false);
+ smgrcreate(smgropen(dst_rlocator, InvalidBackendId, forkNum), false);
/*
* WAL log creation if the relation is persistent, or this is the
@@ -3913,13 +3890,8 @@ CreateAndCopyRelationData(RelFileLocator src_rlocator,
}
/* close source and destination smgr if exists. */
- rlocator.backend = InvalidBackendId;
-
- rlocator.locator = src_rlocator;
- smgrcloserellocator(rlocator);
-
- rlocator.locator = dst_rlocator;
- smgrcloserellocator(rlocator);
+ smgrcloserellocator(src_rlocator, InvalidBackendId);
+ smgrcloserellocator(dst_rlocator, InvalidBackendId);
}
/* ---------------------------------------------------------------------
@@ -5033,7 +5005,7 @@ IssuePendingWritebacks(WritebackContext *context)
{
PendingWriteback *cur;
PendingWriteback *next;
- SMgrRelation reln;
+ SMgrFileHandle sfile;
int ahead;
BufferTag tag;
RelFileLocator currlocator;
@@ -5073,8 +5045,8 @@ IssuePendingWritebacks(WritebackContext *context)
i += ahead;
/* and finally tell the kernel to write the data to storage */
- reln = smgropen(currlocator, InvalidBackendId);
- smgrwriteback(reln, BufTagGetForkNum(&tag), tag.blockNum, nblocks);
+ sfile = smgropen(currlocator, InvalidBackendId, BufTagGetForkNum(&tag));
+ smgrwriteback(sfile, tag.blockNum, nblocks);
}
context->nr_pending = 0;
diff --git a/src/backend/storage/buffer/localbuf.c b/src/backend/storage/buffer/localbuf.c
index 5325ddb663d..62849fb2d11 100644
--- a/src/backend/storage/buffer/localbuf.c
+++ b/src/backend/storage/buffer/localbuf.c
@@ -62,14 +62,13 @@ static Block GetLocalBufferStorage(void);
* No-op if prefetching isn't compiled in.
*/
PrefetchBufferResult
-PrefetchLocalBuffer(SMgrRelation smgr, ForkNumber forkNum,
- BlockNumber blockNum)
+PrefetchLocalBuffer(SMgrFileHandle sfile, BlockNumber blockNum)
{
PrefetchBufferResult result = {InvalidBuffer, false};
BufferTag newTag; /* identity of requested block */
LocalBufferLookupEnt *hresult;
- InitBufferTag(&newTag, &smgr->smgr_rlocator.locator, forkNum, blockNum);
+ InitBufferTag(&newTag, &sfile->smgr_locator.locator, sfile->smgr_locator.forknum, blockNum);
/* Initialize local buffers if first request in this session */
if (LocalBufHash == NULL)
@@ -88,7 +87,7 @@ PrefetchLocalBuffer(SMgrRelation smgr, ForkNumber forkNum,
{
#ifdef USE_PREFETCH
/* Not in buffers, so initiate prefetch */
- smgrprefetch(smgr, forkNum, blockNum);
+ smgrprefetch(sfile, blockNum);
result.initiated_io = true;
#endif /* USE_PREFETCH */
}
@@ -107,7 +106,7 @@ PrefetchLocalBuffer(SMgrRelation smgr, ForkNumber forkNum,
* (hence, usage_count is always advanced).
*/
BufferDesc *
-LocalBufferAlloc(SMgrRelation smgr, ForkNumber forkNum, BlockNumber blockNum,
+LocalBufferAlloc(SMgrFileHandle sfile, BlockNumber blockNum,
bool *foundPtr, IOContext *io_context)
{
BufferTag newTag; /* identity of requested block */
@@ -118,7 +117,7 @@ LocalBufferAlloc(SMgrRelation smgr, ForkNumber forkNum, BlockNumber blockNum,
bool found;
uint32 buf_state;
- InitBufferTag(&newTag, &smgr->smgr_rlocator.locator, forkNum, blockNum);
+ InitBufferTag(&newTag, &sfile->smgr_locator.locator, sfile->smgr_locator.forknum, blockNum);
/* Initialize local buffers if first request in this session */
if (LocalBufHash == NULL)
@@ -143,7 +142,7 @@ LocalBufferAlloc(SMgrRelation smgr, ForkNumber forkNum, BlockNumber blockNum,
Assert(BufferTagsEqual(&bufHdr->tag, &newTag));
#ifdef LBDEBUG
fprintf(stderr, "LB ALLOC (%u,%d,%d) %d\n",
- smgr->smgr_rlocator.locator.relNumber, forkNum, blockNum, -b - 1);
+ sfile->smgr_locator.locator.relNumber, sfile->smgr_locator.forknum, blockNum, -b - 1);
#endif
buf_state = pg_atomic_read_u32(&bufHdr->state);
@@ -171,7 +170,7 @@ LocalBufferAlloc(SMgrRelation smgr, ForkNumber forkNum, BlockNumber blockNum,
#ifdef LBDEBUG
fprintf(stderr, "LB ALLOC (%u,%d,%d) %d\n",
- smgr->smgr_rlocator.locator.relNumber, forkNum, blockNum,
+ sfile->smgr_locator.locator.relNumber, sfile->smgr_locator.forknum, blockNum,
-nextFreeLocalBuf - 1);
#endif
@@ -220,17 +219,16 @@ LocalBufferAlloc(SMgrRelation smgr, ForkNumber forkNum, BlockNumber blockNum,
*/
if (buf_state & BM_DIRTY)
{
- SMgrRelation oreln;
+ SMgrFileHandle ofile;
Page localpage = (char *) LocalBufHdrGetBlock(bufHdr);
- /* Find smgr relation for buffer */
- oreln = smgropen(BufTagGetRelFileLocator(&bufHdr->tag), MyBackendId);
+ /* Find smgr file handle for buffer */
+ ofile = smgropen(BufTagGetRelFileLocator(&bufHdr->tag), MyBackendId, BufTagGetForkNum(&bufHdr->tag));
PageSetChecksumInplace(localpage, bufHdr->tag.blockNum);
/* And write... */
- smgrwrite(oreln,
- BufTagGetForkNum(&bufHdr->tag),
+ smgrwrite(ofile,
bufHdr->tag.blockNum,
localpage,
false);
diff --git a/src/backend/storage/freespace/freespace.c b/src/backend/storage/freespace/freespace.c
index 3e9693b293b..a825a0b5af3 100644
--- a/src/backend/storage/freespace/freespace.c
+++ b/src/backend/storage/freespace/freespace.c
@@ -271,7 +271,7 @@ FreeSpaceMapPrepareTruncateRel(Relation rel, BlockNumber nblocks)
* If no FSM has been created yet for this relation, there's nothing to
* truncate.
*/
- if (!smgrexists(RelationGetSmgr(rel), FSM_FORKNUM))
+ if (!smgrexists(RelationGetSmgr(rel, FSM_FORKNUM)))
return InvalidBlockNumber;
/* Get the location in the FSM of the first removed heap block */
@@ -317,7 +317,7 @@ FreeSpaceMapPrepareTruncateRel(Relation rel, BlockNumber nblocks)
else
{
new_nfsmblocks = fsm_logical_to_physical(first_removed_address);
- if (smgrnblocks(RelationGetSmgr(rel), FSM_FORKNUM) <= new_nfsmblocks)
+ if (smgrnblocks(RelationGetSmgr(rel, FSM_FORKNUM)) <= new_nfsmblocks)
return InvalidBlockNumber; /* nothing to do; the FSM was already
* smaller */
}
@@ -532,14 +532,14 @@ fsm_readbuf(Relation rel, FSMAddress addr, bool extend)
{
BlockNumber blkno = fsm_logical_to_physical(addr);
Buffer buf;
- SMgrRelation reln;
+ SMgrFileHandle fsm_file;
/*
* Caution: re-using this smgr pointer could fail if the relcache entry
* gets closed. It's safe as long as we only do smgr-level operations
* between here and the last use of the pointer.
*/
- reln = RelationGetSmgr(rel);
+ fsm_file = RelationGetSmgr(rel, FSM_FORKNUM);
/*
* If we haven't cached the size of the FSM yet, check it first. Also
@@ -547,19 +547,19 @@ fsm_readbuf(Relation rel, FSMAddress addr, bool extend)
* value might be stale. (We send smgr inval messages on truncation, but
* not on extension.)
*/
- if (reln->smgr_cached_nblocks[FSM_FORKNUM] == InvalidBlockNumber ||
- blkno >= reln->smgr_cached_nblocks[FSM_FORKNUM])
+ if (fsm_file->smgr_cached_nblocks == InvalidBlockNumber ||
+ blkno >= fsm_file->smgr_cached_nblocks)
{
/* Invalidate the cache so smgrnblocks asks the kernel. */
- reln->smgr_cached_nblocks[FSM_FORKNUM] = InvalidBlockNumber;
- if (smgrexists(reln, FSM_FORKNUM))
- smgrnblocks(reln, FSM_FORKNUM);
+ fsm_file->smgr_cached_nblocks = InvalidBlockNumber;
+ if (smgrexists(fsm_file))
+ smgrnblocks(fsm_file);
else
- reln->smgr_cached_nblocks[FSM_FORKNUM] = 0;
+ fsm_file->smgr_cached_nblocks = 0;
}
/* Handle requests beyond EOF */
- if (blkno >= reln->smgr_cached_nblocks[FSM_FORKNUM])
+ if (blkno >= fsm_file->smgr_cached_nblocks)
{
if (extend)
fsm_extend(rel, blkno + 1);
@@ -609,7 +609,7 @@ fsm_extend(Relation rel, BlockNumber fsm_nblocks)
{
BlockNumber fsm_nblocks_now;
PGAlignedBlock pg;
- SMgrRelation reln;
+ SMgrFileHandle fsm_file;
PageInit((Page) pg.data, BLCKSZ, 0);
@@ -630,29 +630,28 @@ fsm_extend(Relation rel, BlockNumber fsm_nblocks)
* gets closed. It's safe as long as we only do smgr-level operations
* between here and the last use of the pointer.
*/
- reln = RelationGetSmgr(rel);
+ fsm_file = RelationGetSmgr(rel, FSM_FORKNUM);
/*
* Create the FSM file first if it doesn't exist. If
* smgr_cached_nblocks[FSM_FORKNUM] is positive then it must exist, no
* need for an smgrexists call.
*/
- if ((reln->smgr_cached_nblocks[FSM_FORKNUM] == 0 ||
- reln->smgr_cached_nblocks[FSM_FORKNUM] == InvalidBlockNumber) &&
- !smgrexists(reln, FSM_FORKNUM))
- smgrcreate(reln, FSM_FORKNUM, false);
+ if ((fsm_file->smgr_cached_nblocks == 0 ||
+ fsm_file->smgr_cached_nblocks == InvalidBlockNumber) &&
+ !smgrexists(fsm_file))
+ smgrcreate(fsm_file, false);
/* Invalidate cache so that smgrnblocks() asks the kernel. */
- reln->smgr_cached_nblocks[FSM_FORKNUM] = InvalidBlockNumber;
- fsm_nblocks_now = smgrnblocks(reln, FSM_FORKNUM);
+ fsm_file->smgr_cached_nblocks = InvalidBlockNumber;
+ fsm_nblocks_now = smgrnblocks(fsm_file);
/* Extend as needed. */
while (fsm_nblocks_now < fsm_nblocks)
{
PageSetChecksumInplace((Page) pg.data, fsm_nblocks_now);
- smgrextend(reln, FSM_FORKNUM, fsm_nblocks_now,
- pg.data, false);
+ smgrextend(fsm_file, fsm_nblocks_now, pg.data, false);
fsm_nblocks_now++;
}
diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c
index 352958e1feb..eea7ce944c3 100644
--- a/src/backend/storage/smgr/md.c
+++ b/src/backend/storage/smgr/md.c
@@ -66,10 +66,10 @@
* out to an unlinked old copy of a segment file that will eventually
* disappear.
*
- * File descriptors are stored in the per-fork md_seg_fds arrays inside
- * SMgrRelation. The length of these arrays is stored in md_num_open_segs.
- * Note that a fork's md_num_open_segs having a specific value does not
- * necessarily mean the relation doesn't have additional segments; we may
+ * File descriptors are stored in the md_seg_fds array inside
+ * SMgrFileData. The length of the array is stored in md_num_open_segs.
+ * Note that md_num_open_segs having a specific value does not
+ * necessarily mean the file doesn't have additional segments; we may
* just not have opened the next segment yet. (We could not have "all
* segments are in the array" as an invariant anyway, since another backend
* could extend the relation while we aren't looking.) We do not have
@@ -121,26 +121,18 @@ static MemoryContext MdCxt; /* context for all MdfdVec objects */
/* local routines */
-static void mdunlinkfork(RelFileLocatorBackend rlocator, ForkNumber forknum,
- bool isRedo);
-static MdfdVec *mdopenfork(SMgrRelation reln, ForkNumber forknum, int behavior);
-static void register_dirty_segment(SMgrRelation reln, ForkNumber forknum,
- MdfdVec *seg);
-static void register_unlink_segment(RelFileLocatorBackend rlocator, ForkNumber forknum,
- BlockNumber segno);
-static void register_forget_request(RelFileLocatorBackend rlocator, ForkNumber forknum,
- BlockNumber segno);
-static void _fdvec_resize(SMgrRelation reln,
- ForkNumber forknum,
- int nseg);
-static char *_mdfd_segpath(SMgrRelation reln, ForkNumber forknum,
- BlockNumber segno);
-static MdfdVec *_mdfd_openseg(SMgrRelation reln, ForkNumber forknum,
+static void mdunlinkfork(SMgrFileLocator slocator, bool isRedo);
+static MdfdVec *mdopenfork(SMgrFileHandle sfile, int behavior);
+static void register_dirty_segment(SMgrFileHandle sfile, MdfdVec *seg);
+static void register_unlink_segment(SMgrFileLocator slocator, BlockNumber segno);
+static void register_forget_request(SMgrFileLocator slocator, BlockNumber segno);
+static void _fdvec_resize(SMgrFileHandle sfile, int nseg);
+static char *_mdfd_segpath(SMgrFileHandle sfile, BlockNumber segno);
+static MdfdVec *_mdfd_openseg(SMgrFileHandle sfile,
BlockNumber segno, int oflags);
-static MdfdVec *_mdfd_getseg(SMgrRelation reln, ForkNumber forknum,
+static MdfdVec *_mdfd_getseg(SMgrFileHandle sfile,
BlockNumber blkno, bool skipFsync, int behavior);
-static BlockNumber _mdnblocks(SMgrRelation reln, ForkNumber forknum,
- MdfdVec *seg);
+static BlockNumber _mdnblocks(SMgrFileHandle sfile, MdfdVec *seg);
/*
@@ -160,7 +152,7 @@ mdinit(void)
* Note: this will return true for lingering files, with pending deletions
*/
bool
-mdexists(SMgrRelation reln, ForkNumber forknum)
+mdexists(SMgrFileHandle sfile)
{
/*
* Close it first, to ensure that we notice if the fork has been unlinked
@@ -168,9 +160,9 @@ mdexists(SMgrRelation reln, ForkNumber forknum)
* which already closes relations when dropping them.
*/
if (!InRecovery)
- mdclose(reln, forknum);
+ mdclose(sfile);
- return (mdopenfork(reln, forknum, EXTENSION_RETURN_NULL) != NULL);
+ return (mdopenfork(sfile, EXTENSION_RETURN_NULL) != NULL);
}
/*
@@ -179,16 +171,16 @@ mdexists(SMgrRelation reln, ForkNumber forknum)
* If isRedo is true, it's okay for the relation to exist already.
*/
void
-mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo)
+mdcreate(SMgrFileHandle sfile, bool isRedo)
{
MdfdVec *mdfd;
char *path;
File fd;
- if (isRedo && reln->md_num_open_segs[forknum] > 0)
+ if (isRedo && sfile->md_num_open_segs > 0)
return; /* created and opened already... */
- Assert(reln->md_num_open_segs[forknum] == 0);
+ Assert(sfile->md_num_open_segs == 0);
/*
* We may be using the target table space for the first time in this
@@ -199,11 +191,11 @@ mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo)
* should be here and not in commands/tablespace.c? But that would imply
* importing a lot of stuff that smgr.c oughtn't know, either.
*/
- TablespaceCreateDbspace(reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
+ TablespaceCreateDbspace(sfile->smgr_locator.locator.spcOid,
+ sfile->smgr_locator.locator.dbOid,
isRedo);
- path = relpath(reln->smgr_rlocator, forknum);
+ path = smgrfilepath(sfile->smgr_locator);
fd = PathNameOpenFile(path, O_RDWR | O_CREAT | O_EXCL | PG_BINARY);
@@ -225,8 +217,8 @@ mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo)
pfree(path);
- _fdvec_resize(reln, forknum, 1);
- mdfd = &reln->md_seg_fds[forknum][0];
+ _fdvec_resize(sfile, 1);
+ mdfd = &sfile->md_seg_fds[0];
mdfd->mdfd_vfd = fd;
mdfd->mdfd_segno = 0;
}
@@ -293,16 +285,9 @@ mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo)
* we are usually not in a transaction anymore when this is called.
*/
void
-mdunlink(RelFileLocatorBackend rlocator, ForkNumber forknum, bool isRedo)
+mdunlink(SMgrFileLocator slocator, bool isRedo)
{
- /* Now do the per-fork work */
- if (forknum == InvalidForkNumber)
- {
- for (forknum = 0; forknum <= MAX_FORKNUM; forknum++)
- mdunlinkfork(rlocator, forknum, isRedo);
- }
- else
- mdunlinkfork(rlocator, forknum, isRedo);
+ mdunlinkfork(slocator, isRedo);
}
/*
@@ -330,29 +315,29 @@ do_truncate(const char *path)
}
static void
-mdunlinkfork(RelFileLocatorBackend rlocator, ForkNumber forknum, bool isRedo)
+mdunlinkfork(SMgrFileLocator slocator, bool isRedo)
{
char *path;
int ret;
int save_errno;
- path = relpath(rlocator, forknum);
+ path = smgrfilepath(slocator);
/*
* Truncate and then unlink the first segment, or just register a request
* to unlink it later, as described in the comments for mdunlink().
*/
- if (isRedo || IsBinaryUpgrade || forknum != MAIN_FORKNUM ||
- RelFileLocatorBackendIsTemp(rlocator))
+ if (isRedo || IsBinaryUpgrade || slocator.forknum != MAIN_FORKNUM ||
+ SMgrFileLocatorIsTemp(slocator))
{
- if (!RelFileLocatorBackendIsTemp(rlocator))
+ if (!SMgrFileLocatorIsTemp(slocator))
{
/* Prevent other backends' fds from holding on to the disk space */
ret = do_truncate(path);
/* Forget any pending sync requests for the first segment */
save_errno = errno;
- register_forget_request(rlocator, forknum, 0 /* first seg */ );
+ register_forget_request(slocator, 0 /* first seg */ );
errno = save_errno;
}
else
@@ -379,7 +364,7 @@ mdunlinkfork(RelFileLocatorBackend rlocator, ForkNumber forknum, bool isRedo)
/* Register request to unlink first segment later */
save_errno = errno;
- register_unlink_segment(rlocator, forknum, 0 /* first seg */ );
+ register_unlink_segment(slocator, 0 /* first seg */ );
errno = save_errno;
}
@@ -404,7 +389,7 @@ mdunlinkfork(RelFileLocatorBackend rlocator, ForkNumber forknum, bool isRedo)
{
sprintf(segpath, "%s.%u", path, segno);
- if (!RelFileLocatorBackendIsTemp(rlocator))
+ if (!SMgrFileLocatorIsTemp(slocator))
{
/*
* Prevent other backends' fds from holding on to the disk
@@ -417,7 +402,7 @@ mdunlinkfork(RelFileLocatorBackend rlocator, ForkNumber forknum, bool isRedo)
* Forget any pending sync requests for this segment before we
* try to unlink.
*/
- register_forget_request(rlocator, forknum, segno);
+ register_forget_request(slocator, segno);
}
if (unlink(segpath) < 0)
@@ -446,7 +431,7 @@ mdunlinkfork(RelFileLocatorBackend rlocator, ForkNumber forknum, bool isRedo)
* causes intervening file space to become filled with zeroes.
*/
void
-mdextend(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
+mdextend(SMgrFileHandle sfile, BlockNumber blocknum,
const void *buffer, bool skipFsync)
{
off_t seekpos;
@@ -455,7 +440,7 @@ mdextend(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
/* This assert is too expensive to have on normally ... */
#ifdef CHECK_WRITE_VS_EXTEND
- Assert(blocknum >= mdnblocks(reln, forknum));
+ Assert(blocknum >= mdnblocks(sfile));
#endif
/*
@@ -468,10 +453,10 @@ mdextend(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
ereport(ERROR,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
errmsg("cannot extend file \"%s\" beyond %u blocks",
- relpath(reln->smgr_rlocator, forknum),
+ smgrfilepath(sfile->smgr_locator),
InvalidBlockNumber)));
- v = _mdfd_getseg(reln, forknum, blocknum, skipFsync, EXTENSION_CREATE);
+ v = _mdfd_getseg(sfile, blocknum, skipFsync, EXTENSION_CREATE);
seekpos = (off_t) BLCKSZ * (blocknum % ((BlockNumber) RELSEG_SIZE));
@@ -494,10 +479,10 @@ mdextend(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
errhint("Check free disk space.")));
}
- if (!skipFsync && !SmgrIsTemp(reln))
- register_dirty_segment(reln, forknum, v);
+ if (!skipFsync && !SmgrIsTemp(sfile))
+ register_dirty_segment(sfile, v);
- Assert(_mdnblocks(reln, forknum, v) <= ((BlockNumber) RELSEG_SIZE));
+ Assert(_mdnblocks(sfile, v) <= ((BlockNumber) RELSEG_SIZE));
}
/*
@@ -511,17 +496,17 @@ mdextend(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
* invent one out of whole cloth.
*/
static MdfdVec *
-mdopenfork(SMgrRelation reln, ForkNumber forknum, int behavior)
+mdopenfork(SMgrFileHandle sfile, int behavior)
{
MdfdVec *mdfd;
char *path;
File fd;
/* No work if already open */
- if (reln->md_num_open_segs[forknum] > 0)
- return &reln->md_seg_fds[forknum][0];
+ if (sfile->md_num_open_segs > 0)
+ return &sfile->md_seg_fds[0];
- path = relpath(reln->smgr_rlocator, forknum);
+ path = smgrfilepath(sfile->smgr_locator);
fd = PathNameOpenFile(path, O_RDWR | PG_BINARY);
@@ -540,12 +525,12 @@ mdopenfork(SMgrRelation reln, ForkNumber forknum, int behavior)
pfree(path);
- _fdvec_resize(reln, forknum, 1);
- mdfd = &reln->md_seg_fds[forknum][0];
+ _fdvec_resize(sfile, 1);
+ mdfd = &sfile->md_seg_fds[0];
mdfd->mdfd_vfd = fd;
mdfd->mdfd_segno = 0;
- Assert(_mdnblocks(reln, forknum, mdfd) <= ((BlockNumber) RELSEG_SIZE));
+ Assert(_mdnblocks(sfile, mdfd) <= ((BlockNumber) RELSEG_SIZE));
return mdfd;
}
@@ -554,20 +539,19 @@ mdopenfork(SMgrRelation reln, ForkNumber forknum, int behavior)
* mdopen() -- Initialize newly-opened relation.
*/
void
-mdopen(SMgrRelation reln)
+mdopen(SMgrFileHandle sfile)
{
/* mark it not open */
- for (int forknum = 0; forknum <= MAX_FORKNUM; forknum++)
- reln->md_num_open_segs[forknum] = 0;
+ sfile->md_num_open_segs = 0;
}
/*
* mdclose() -- Close the specified relation, if it isn't closed already.
*/
void
-mdclose(SMgrRelation reln, ForkNumber forknum)
+mdclose(SMgrFileHandle sfile)
{
- int nopensegs = reln->md_num_open_segs[forknum];
+ int nopensegs = sfile->md_num_open_segs;
/* No work if already closed */
if (nopensegs == 0)
@@ -576,10 +560,10 @@ mdclose(SMgrRelation reln, ForkNumber forknum)
/* close segments starting from the end */
while (nopensegs > 0)
{
- MdfdVec *v = &reln->md_seg_fds[forknum][nopensegs - 1];
+ MdfdVec *v = &sfile->md_seg_fds[nopensegs - 1];
FileClose(v->mdfd_vfd);
- _fdvec_resize(reln, forknum, nopensegs - 1);
+ _fdvec_resize(sfile, nopensegs - 1);
nopensegs--;
}
}
@@ -588,13 +572,13 @@ mdclose(SMgrRelation reln, ForkNumber forknum)
* mdprefetch() -- Initiate asynchronous read of the specified block of a relation
*/
bool
-mdprefetch(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum)
+mdprefetch(SMgrFileHandle sfile, BlockNumber blocknum)
{
#ifdef USE_PREFETCH
off_t seekpos;
MdfdVec *v;
- v = _mdfd_getseg(reln, forknum, blocknum, false,
+ v = _mdfd_getseg(sfile, blocknum, false,
InRecovery ? EXTENSION_RETURN_NULL : EXTENSION_FAIL);
if (v == NULL)
return false;
@@ -616,8 +600,7 @@ mdprefetch(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum)
* considerably more efficient than doing so individually.
*/
void
-mdwriteback(SMgrRelation reln, ForkNumber forknum,
- BlockNumber blocknum, BlockNumber nblocks)
+mdwriteback(SMgrFileHandle sfile, BlockNumber blocknum, BlockNumber nblocks)
{
/*
* Issue flush requests in as few requests as possible; have to split at
@@ -631,7 +614,7 @@ mdwriteback(SMgrRelation reln, ForkNumber forknum,
int segnum_start,
segnum_end;
- v = _mdfd_getseg(reln, forknum, blocknum, true /* not used */ ,
+ v = _mdfd_getseg(sfile, blocknum, true /* not used */ ,
EXTENSION_DONT_OPEN);
/*
@@ -668,20 +651,21 @@ mdwriteback(SMgrRelation reln, ForkNumber forknum,
* mdread() -- Read the specified block from a relation.
*/
void
-mdread(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
+mdread(SMgrFileHandle sfile, BlockNumber blocknum,
void *buffer)
{
off_t seekpos;
int nbytes;
MdfdVec *v;
- TRACE_POSTGRESQL_SMGR_MD_READ_START(forknum, blocknum,
- reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
- reln->smgr_rlocator.locator.relNumber,
- reln->smgr_rlocator.backend);
+ TRACE_POSTGRESQL_SMGR_MD_READ_START(sfile->smgr_forknum,
+ blocknum,
+ sfile->smgr_rlocator.locator.spcOid,
+ sfile->smgr_rlocator.locator.dbOid,
+ sfile->smgr_rlocator.locator.relNumber,
+ sfile->smgr_rlocator.backend);
- v = _mdfd_getseg(reln, forknum, blocknum, false,
+ v = _mdfd_getseg(sfile, blocknum, false,
EXTENSION_FAIL | EXTENSION_CREATE_RECOVERY);
seekpos = (off_t) BLCKSZ * (blocknum % ((BlockNumber) RELSEG_SIZE));
@@ -690,11 +674,12 @@ mdread(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
nbytes = FileRead(v->mdfd_vfd, buffer, BLCKSZ, seekpos, WAIT_EVENT_DATA_FILE_READ);
- TRACE_POSTGRESQL_SMGR_MD_READ_DONE(forknum, blocknum,
- reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
- reln->smgr_rlocator.locator.relNumber,
- reln->smgr_rlocator.backend,
+ TRACE_POSTGRESQL_SMGR_MD_READ_DONE(sfile->smgr_forknum,
+ blocknum,
+ sfile->smgr_rlocator.locator.spcOid,
+ sfile->smgr_rlocator.locator.dbOid,
+ sfile->smgr_rlocator.locator.relNumber,
+ sfile->smgr_rlocator.backend,
nbytes,
BLCKSZ);
@@ -733,7 +718,7 @@ mdread(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
* use mdextend().
*/
void
-mdwrite(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
+mdwrite(SMgrFileHandle sfile, BlockNumber blocknum,
const void *buffer, bool skipFsync)
{
off_t seekpos;
@@ -742,16 +727,17 @@ mdwrite(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
/* This assert is too expensive to have on normally ... */
#ifdef CHECK_WRITE_VS_EXTEND
- Assert(blocknum < mdnblocks(reln, forknum));
+ Assert(blocknum < mdnblocks(sfile));
#endif
- TRACE_POSTGRESQL_SMGR_MD_WRITE_START(forknum, blocknum,
- reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
- reln->smgr_rlocator.locator.relNumber,
- reln->smgr_rlocator.backend);
+ TRACE_POSTGRESQL_SMGR_MD_WRITE_START(sfile->smgr_forknum,
+ blocknum,
+ sfile->smgr_rlocator.locator.spcOid,
+ sfile->smgr_rlocator.locator.dbOid,
+ sfile->smgr_rlocator.locator.relNumber,
+ sfile->smgr_rlocator.backend);
- v = _mdfd_getseg(reln, forknum, blocknum, skipFsync,
+ v = _mdfd_getseg(sfile, blocknum, skipFsync,
EXTENSION_FAIL | EXTENSION_CREATE_RECOVERY);
seekpos = (off_t) BLCKSZ * (blocknum % ((BlockNumber) RELSEG_SIZE));
@@ -760,11 +746,12 @@ mdwrite(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
nbytes = FileWrite(v->mdfd_vfd, buffer, BLCKSZ, seekpos, WAIT_EVENT_DATA_FILE_WRITE);
- TRACE_POSTGRESQL_SMGR_MD_WRITE_DONE(forknum, blocknum,
- reln->smgr_rlocator.locator.spcOid,
- reln->smgr_rlocator.locator.dbOid,
- reln->smgr_rlocator.locator.relNumber,
- reln->smgr_rlocator.backend,
+ TRACE_POSTGRESQL_SMGR_MD_WRITE_DONE(sfile->smgr_forknum,
+ blocknum,
+ sfile->smgr_rlocator.locator.spcOid,
+ sfile->smgr_rlocator.locator.dbOid,
+ sfile->smgr_rlocator.locator.relNumber,
+ sfile->smgr_rlocator.backend,
nbytes,
BLCKSZ);
@@ -785,8 +772,8 @@ mdwrite(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
errhint("Check free disk space.")));
}
- if (!skipFsync && !SmgrIsTemp(reln))
- register_dirty_segment(reln, forknum, v);
+ if (!skipFsync && !SmgrIsTemp(sfile))
+ register_dirty_segment(sfile, v);
}
/*
@@ -798,16 +785,16 @@ mdwrite(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
* are present in the array.
*/
BlockNumber
-mdnblocks(SMgrRelation reln, ForkNumber forknum)
+mdnblocks(SMgrFileHandle sfile)
{
MdfdVec *v;
BlockNumber nblocks;
BlockNumber segno;
- mdopenfork(reln, forknum, EXTENSION_FAIL);
+ mdopenfork(sfile, EXTENSION_FAIL);
/* mdopen has opened the first segment */
- Assert(reln->md_num_open_segs[forknum] > 0);
+ Assert(sfile->md_num_open_segs > 0);
/*
* Start from the last open segments, to avoid redundant seeks. We have
@@ -822,12 +809,12 @@ mdnblocks(SMgrRelation reln, ForkNumber forknum)
* that's OK because the checkpointer never needs to compute relation
* size.)
*/
- segno = reln->md_num_open_segs[forknum] - 1;
- v = &reln->md_seg_fds[forknum][segno];
+ segno = sfile->md_num_open_segs - 1;
+ v = &sfile->md_seg_fds[segno];
for (;;)
{
- nblocks = _mdnblocks(reln, forknum, v);
+ nblocks = _mdnblocks(sfile, v);
if (nblocks > ((BlockNumber) RELSEG_SIZE))
elog(FATAL, "segment too big");
if (nblocks < ((BlockNumber) RELSEG_SIZE))
@@ -845,7 +832,7 @@ mdnblocks(SMgrRelation reln, ForkNumber forknum)
* undermines _mdfd_getseg's attempts to notice and report an error
* upon access to a missing segment.
*/
- v = _mdfd_openseg(reln, forknum, segno, 0);
+ v = _mdfd_openseg(sfile, segno, 0);
if (v == NULL)
return segno * ((BlockNumber) RELSEG_SIZE);
}
@@ -855,7 +842,7 @@ mdnblocks(SMgrRelation reln, ForkNumber forknum)
* mdtruncate() -- Truncate relation to specified number of blocks.
*/
void
-mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks)
+mdtruncate(SMgrFileHandle sfile, BlockNumber nblocks)
{
BlockNumber curnblk;
BlockNumber priorblocks;
@@ -865,7 +852,7 @@ mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks)
* NOTE: mdnblocks makes sure we have opened all active segments, so that
* truncation loop will get them all!
*/
- curnblk = mdnblocks(reln, forknum);
+ curnblk = mdnblocks(sfile);
if (nblocks > curnblk)
{
/* Bogus request ... but no complaint if InRecovery */
@@ -873,7 +860,7 @@ mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks)
return;
ereport(ERROR,
(errmsg("could not truncate file \"%s\" to %u blocks: it's only %u blocks now",
- relpath(reln->smgr_rlocator, forknum),
+ smgrfilepath(sfile->smgr_locator),
nblocks, curnblk)));
}
if (nblocks == curnblk)
@@ -883,14 +870,14 @@ mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks)
* Truncate segments, starting at the last one. Starting at the end makes
* managing the memory for the fd array easier, should there be errors.
*/
- curopensegs = reln->md_num_open_segs[forknum];
+ curopensegs = sfile->md_num_open_segs;
while (curopensegs > 0)
{
MdfdVec *v;
priorblocks = (curopensegs - 1) * RELSEG_SIZE;
- v = &reln->md_seg_fds[forknum][curopensegs - 1];
+ v = &sfile->md_seg_fds[curopensegs - 1];
if (priorblocks > nblocks)
{
@@ -904,14 +891,14 @@ mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks)
errmsg("could not truncate file \"%s\": %m",
FilePathName(v->mdfd_vfd))));
- if (!SmgrIsTemp(reln))
- register_dirty_segment(reln, forknum, v);
+ if (!SmgrIsTemp(sfile))
+ register_dirty_segment(sfile, v);
/* we never drop the 1st segment */
- Assert(v != &reln->md_seg_fds[forknum][0]);
+ Assert(v != &sfile->md_seg_fds[0]);
FileClose(v->mdfd_vfd);
- _fdvec_resize(reln, forknum, curopensegs - 1);
+ _fdvec_resize(sfile, curopensegs - 1);
}
else if (priorblocks + ((BlockNumber) RELSEG_SIZE) > nblocks)
{
@@ -930,8 +917,8 @@ mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks)
errmsg("could not truncate file \"%s\" to %u blocks: %m",
FilePathName(v->mdfd_vfd),
nblocks)));
- if (!SmgrIsTemp(reln))
- register_dirty_segment(reln, forknum, v);
+ if (!SmgrIsTemp(sfile))
+ register_dirty_segment(sfile, v);
}
else
{
@@ -957,7 +944,7 @@ mdtruncate(SMgrRelation reln, ForkNumber forknum, BlockNumber nblocks)
* segment may survive recovery, reintroducing unwanted data into the table.
*/
void
-mdimmedsync(SMgrRelation reln, ForkNumber forknum)
+mdimmedsync(SMgrFileHandle sfile)
{
int segno;
int min_inactive_seg;
@@ -966,9 +953,9 @@ mdimmedsync(SMgrRelation reln, ForkNumber forknum)
* NOTE: mdnblocks makes sure we have opened all active segments, so that
* fsync loop will get them all!
*/
- mdnblocks(reln, forknum);
+ mdnblocks(sfile);
- min_inactive_seg = segno = reln->md_num_open_segs[forknum];
+ min_inactive_seg = segno = sfile->md_num_open_segs;
/*
* Temporarily open inactive segments, then close them after sync. There
@@ -976,12 +963,12 @@ mdimmedsync(SMgrRelation reln, ForkNumber forknum)
* is harmless. We don't bother to clean them up and take a risk of
* further trouble. The next mdclose() will soon close them.
*/
- while (_mdfd_openseg(reln, forknum, segno, 0) != NULL)
+ while (_mdfd_openseg(sfile, segno, 0) != NULL)
segno++;
while (segno > 0)
{
- MdfdVec *v = &reln->md_seg_fds[forknum][segno - 1];
+ MdfdVec *v = &sfile->md_seg_fds[segno - 1];
/*
* fsyncs done through mdimmedsync() should be tracked in a separate
@@ -1002,7 +989,7 @@ mdimmedsync(SMgrRelation reln, ForkNumber forknum)
if (segno > min_inactive_seg)
{
FileClose(v->mdfd_vfd);
- _fdvec_resize(reln, forknum, segno - 1);
+ _fdvec_resize(sfile, segno - 1);
}
segno--;
@@ -1019,14 +1006,14 @@ mdimmedsync(SMgrRelation reln, ForkNumber forknum)
* enough to be a performance problem).
*/
static void
-register_dirty_segment(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg)
+register_dirty_segment(SMgrFileHandle sfile, MdfdVec *seg)
{
FileTag tag;
- INIT_MD_FILETAG(tag, reln->smgr_rlocator.locator, forknum, seg->mdfd_segno);
+ INIT_MD_FILETAG(tag, sfile->smgr_locator.locator, sfile->smgr_locator.forknum, seg->mdfd_segno);
/* Temp relations should never be fsync'd */
- Assert(!SmgrIsTemp(reln));
+ Assert(!SmgrIsTemp(sfile));
if (!RegisterSyncRequest(&tag, SYNC_REQUEST, false /* retryOnError */ ))
{
@@ -1058,15 +1045,14 @@ register_dirty_segment(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg)
* register_unlink_segment() -- Schedule a file to be deleted after next checkpoint
*/
static void
-register_unlink_segment(RelFileLocatorBackend rlocator, ForkNumber forknum,
- BlockNumber segno)
+register_unlink_segment(SMgrFileLocator slocator, BlockNumber segno)
{
FileTag tag;
- INIT_MD_FILETAG(tag, rlocator.locator, forknum, segno);
+ INIT_MD_FILETAG(tag, slocator.locator, slocator.forknum, segno);
/* Should never be used with temp relations */
- Assert(!RelFileLocatorBackendIsTemp(rlocator));
+ Assert(!SMgrFileLocatorIsTemp(slocator));
RegisterSyncRequest(&tag, SYNC_UNLINK_REQUEST, true /* retryOnError */ );
}
@@ -1075,12 +1061,11 @@ register_unlink_segment(RelFileLocatorBackend rlocator, ForkNumber forknum,
* register_forget_request() -- forget any fsyncs for a relation fork's segment
*/
static void
-register_forget_request(RelFileLocatorBackend rlocator, ForkNumber forknum,
- BlockNumber segno)
+register_forget_request(SMgrFileLocator slocator, BlockNumber segno)
{
FileTag tag;
- INIT_MD_FILETAG(tag, rlocator.locator, forknum, segno);
+ INIT_MD_FILETAG(tag, slocator.locator, slocator.forknum, segno);
RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true /* retryOnError */ );
}
@@ -1103,57 +1088,25 @@ ForgetDatabaseSyncRequests(Oid dbid)
RegisterSyncRequest(&tag, SYNC_FILTER_REQUEST, true /* retryOnError */ );
}
-/*
- * DropRelationFiles -- drop files of all given relations
- */
-void
-DropRelationFiles(RelFileLocator *delrels, int ndelrels, bool isRedo)
-{
- SMgrRelation *srels;
- int i;
-
- srels = palloc(sizeof(SMgrRelation) * ndelrels);
- for (i = 0; i < ndelrels; i++)
- {
- SMgrRelation srel = smgropen(delrels[i], InvalidBackendId);
-
- if (isRedo)
- {
- ForkNumber fork;
-
- for (fork = 0; fork <= MAX_FORKNUM; fork++)
- XLogDropRelation(delrels[i], fork);
- }
- srels[i] = srel;
- }
-
- smgrdounlinkall(srels, ndelrels, isRedo);
-
- for (i = 0; i < ndelrels; i++)
- smgrclose(srels[i]);
- pfree(srels);
-}
-
/*
* _fdvec_resize() -- Resize the fork's open segments array
*/
static void
-_fdvec_resize(SMgrRelation reln,
- ForkNumber forknum,
+_fdvec_resize(SMgrFileHandle sfile,
int nseg)
{
if (nseg == 0)
{
- if (reln->md_num_open_segs[forknum] > 0)
+ if (sfile->md_num_open_segs > 0)
{
- pfree(reln->md_seg_fds[forknum]);
- reln->md_seg_fds[forknum] = NULL;
+ pfree(sfile->md_seg_fds);
+ sfile->md_seg_fds = NULL;
}
}
- else if (reln->md_num_open_segs[forknum] == 0)
+ else if (sfile->md_num_open_segs == 0)
{
- reln->md_seg_fds[forknum] =
+ sfile->md_seg_fds =
MemoryContextAlloc(MdCxt, sizeof(MdfdVec) * nseg);
}
else
@@ -1164,12 +1117,12 @@ _fdvec_resize(SMgrRelation reln,
* FileClose(), and the memory context internally will sometimes avoid
* doing an actual reallocation.
*/
- reln->md_seg_fds[forknum] =
- repalloc(reln->md_seg_fds[forknum],
+ sfile->md_seg_fds =
+ repalloc(sfile->md_seg_fds,
sizeof(MdfdVec) * nseg);
}
- reln->md_num_open_segs[forknum] = nseg;
+ sfile->md_num_open_segs = nseg;
}
/*
@@ -1177,12 +1130,12 @@ _fdvec_resize(SMgrRelation reln,
* returned string is palloc'd.
*/
static char *
-_mdfd_segpath(SMgrRelation reln, ForkNumber forknum, BlockNumber segno)
+_mdfd_segpath(SMgrFileHandle sfile, BlockNumber segno)
{
char *path,
*fullpath;
- path = relpath(reln->smgr_rlocator, forknum);
+ path = smgrfilepath(sfile->smgr_locator);
if (segno > 0)
{
@@ -1200,14 +1153,14 @@ _mdfd_segpath(SMgrRelation reln, ForkNumber forknum, BlockNumber segno)
* and make a MdfdVec object for it. Returns NULL on failure.
*/
static MdfdVec *
-_mdfd_openseg(SMgrRelation reln, ForkNumber forknum, BlockNumber segno,
+_mdfd_openseg(SMgrFileHandle sfile, BlockNumber segno,
int oflags)
{
MdfdVec *v;
File fd;
char *fullpath;
- fullpath = _mdfd_segpath(reln, forknum, segno);
+ fullpath = _mdfd_segpath(sfile, segno);
/* open the file */
fd = PathNameOpenFile(fullpath, O_RDWR | PG_BINARY | oflags);
@@ -1221,16 +1174,16 @@ _mdfd_openseg(SMgrRelation reln, ForkNumber forknum, BlockNumber segno,
* Segments are always opened in order from lowest to highest, so we must
* be adding a new one at the end.
*/
- Assert(segno == reln->md_num_open_segs[forknum]);
+ Assert(segno == sfile->md_num_open_segs);
- _fdvec_resize(reln, forknum, segno + 1);
+ _fdvec_resize(sfile, segno + 1);
/* fill the entry */
- v = &reln->md_seg_fds[forknum][segno];
+ v = &sfile->md_seg_fds[segno];
v->mdfd_vfd = fd;
v->mdfd_segno = segno;
- Assert(_mdnblocks(reln, forknum, v) <= ((BlockNumber) RELSEG_SIZE));
+ Assert(_mdnblocks(sfile, v) <= ((BlockNumber) RELSEG_SIZE));
/* all done */
return v;
@@ -1245,7 +1198,7 @@ _mdfd_openseg(SMgrRelation reln, ForkNumber forknum, BlockNumber segno,
* EXTENSION_CREATE case.
*/
static MdfdVec *
-_mdfd_getseg(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
+_mdfd_getseg(SMgrFileHandle sfile, BlockNumber blkno,
bool skipFsync, int behavior)
{
MdfdVec *v;
@@ -1260,9 +1213,9 @@ _mdfd_getseg(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
targetseg = blkno / ((BlockNumber) RELSEG_SIZE);
/* if an existing and opened segment, we're done */
- if (targetseg < reln->md_num_open_segs[forknum])
+ if (targetseg < sfile->md_num_open_segs)
{
- v = &reln->md_seg_fds[forknum][targetseg];
+ v = &sfile->md_seg_fds[targetseg];
return v;
}
@@ -1277,19 +1230,19 @@ _mdfd_getseg(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
* 'behavior'). Start with either the last opened, or the first segment if
* none was opened before.
*/
- if (reln->md_num_open_segs[forknum] > 0)
- v = &reln->md_seg_fds[forknum][reln->md_num_open_segs[forknum] - 1];
+ if (sfile->md_num_open_segs > 0)
+ v = &sfile->md_seg_fds[sfile->md_num_open_segs - 1];
else
{
- v = mdopenfork(reln, forknum, behavior);
+ v = mdopenfork(sfile, behavior);
if (!v)
return NULL; /* if behavior & EXTENSION_RETURN_NULL */
}
- for (nextsegno = reln->md_num_open_segs[forknum];
+ for (nextsegno = sfile->md_num_open_segs;
nextsegno <= targetseg; nextsegno++)
{
- BlockNumber nblocks = _mdnblocks(reln, forknum, v);
+ BlockNumber nblocks = _mdnblocks(sfile, v);
int flags = 0;
Assert(nextsegno == v->mdfd_segno + 1);
@@ -1318,7 +1271,7 @@ _mdfd_getseg(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
{
char *zerobuf = palloc0(BLCKSZ);
- mdextend(reln, forknum,
+ mdextend(sfile,
nextsegno * ((BlockNumber) RELSEG_SIZE) - 1,
zerobuf, skipFsync);
pfree(zerobuf);
@@ -1349,11 +1302,11 @@ _mdfd_getseg(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not open file \"%s\" (target block %u): previous segment is only %u blocks",
- _mdfd_segpath(reln, forknum, nextsegno),
+ _mdfd_segpath(sfile, nextsegno),
blkno, nblocks)));
}
- v = _mdfd_openseg(reln, forknum, nextsegno, flags);
+ v = _mdfd_openseg(sfile, nextsegno, flags);
if (v == NULL)
{
@@ -1363,7 +1316,7 @@ _mdfd_getseg(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not open file \"%s\" (target block %u): %m",
- _mdfd_segpath(reln, forknum, nextsegno),
+ _mdfd_segpath(sfile, nextsegno),
blkno)));
}
}
@@ -1375,7 +1328,7 @@ _mdfd_getseg(SMgrRelation reln, ForkNumber forknum, BlockNumber blkno,
* Get number of blocks present in a single disk file
*/
static BlockNumber
-_mdnblocks(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg)
+_mdnblocks(SMgrFileHandle sfile, MdfdVec *seg)
{
off_t len;
@@ -1398,16 +1351,16 @@ _mdnblocks(SMgrRelation reln, ForkNumber forknum, MdfdVec *seg)
int
mdsyncfiletag(const FileTag *ftag, char *path)
{
- SMgrRelation reln = smgropen(ftag->rlocator, InvalidBackendId);
+ SMgrFileHandle sfile = smgropen(ftag->rlocator, InvalidBackendId, ftag->forknum);
File file;
bool need_to_close;
int result,
save_errno;
/* See if we already have the file open, or need to open it. */
- if (ftag->segno < reln->md_num_open_segs[ftag->forknum])
+ if (ftag->segno < sfile->md_num_open_segs)
{
- file = reln->md_seg_fds[ftag->forknum][ftag->segno].mdfd_vfd;
+ file = sfile->md_seg_fds[ftag->segno].mdfd_vfd;
strlcpy(path, FilePathName(file), MAXPGPATH);
need_to_close = false;
}
@@ -1415,7 +1368,7 @@ mdsyncfiletag(const FileTag *ftag, char *path)
{
char *p;
- p = _mdfd_segpath(reln, ftag->forknum, ftag->segno);
+ p = _mdfd_segpath(sfile, ftag->segno);
strlcpy(path, p, MAXPGPATH);
pfree(p);
diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c
index dc466e54145..58a2322d018 100644
--- a/src/backend/storage/smgr/smgr.c
+++ b/src/backend/storage/smgr/smgr.c
@@ -41,27 +41,24 @@ typedef struct f_smgr
{
void (*smgr_init) (void); /* may be NULL */
void (*smgr_shutdown) (void); /* may be NULL */
- void (*smgr_open) (SMgrRelation reln);
- void (*smgr_close) (SMgrRelation reln, ForkNumber forknum);
- void (*smgr_create) (SMgrRelation reln, ForkNumber forknum,
- bool isRedo);
- bool (*smgr_exists) (SMgrRelation reln, ForkNumber forknum);
- void (*smgr_unlink) (RelFileLocatorBackend rlocator, ForkNumber forknum,
- bool isRedo);
- void (*smgr_extend) (SMgrRelation reln, ForkNumber forknum,
+ void (*smgr_open) (SMgrFileHandle sfile);
+ void (*smgr_close) (SMgrFileHandle sfile);
+ void (*smgr_create) (SMgrFileHandle sfile, bool isRedo);
+ bool (*smgr_exists) (SMgrFileHandle sfile);
+ void (*smgr_unlink) (SMgrFileLocator slocator, bool isRedo);
+ void (*smgr_extend) (SMgrFileHandle sfile,
BlockNumber blocknum, const void *buffer, bool skipFsync);
- bool (*smgr_prefetch) (SMgrRelation reln, ForkNumber forknum,
+ bool (*smgr_prefetch) (SMgrFileHandle sfile,
BlockNumber blocknum);
- void (*smgr_read) (SMgrRelation reln, ForkNumber forknum,
+ void (*smgr_read) (SMgrFileHandle sfile,
BlockNumber blocknum, void *buffer);
- void (*smgr_write) (SMgrRelation reln, ForkNumber forknum,
+ void (*smgr_write) (SMgrFileHandle sfile,
BlockNumber blocknum, const void *buffer, bool skipFsync);
- void (*smgr_writeback) (SMgrRelation reln, ForkNumber forknum,
+ void (*smgr_writeback) (SMgrFileHandle sfile,
BlockNumber blocknum, BlockNumber nblocks);
- BlockNumber (*smgr_nblocks) (SMgrRelation reln, ForkNumber forknum);
- void (*smgr_truncate) (SMgrRelation reln, ForkNumber forknum,
- BlockNumber nblocks);
- void (*smgr_immedsync) (SMgrRelation reln, ForkNumber forknum);
+ BlockNumber (*smgr_nblocks) (SMgrFileHandle sfile);
+ void (*smgr_truncate) (SMgrFileHandle sfile, BlockNumber nblocks);
+ void (*smgr_immedsync) (SMgrFileHandle sfile);
} f_smgr;
static const f_smgr smgrsw[] = {
@@ -88,12 +85,12 @@ static const f_smgr smgrsw[] = {
static const int NSmgr = lengthof(smgrsw);
/*
- * Each backend has a hashtable that stores all extant SMgrRelation objects.
- * In addition, "unowned" SMgrRelation objects are chained together in a list.
+ * Each backend has a hashtable that stores all extant SMgrFileData objects.
+ * In addition, "unowned" SMgrFile objects are chained together in a list.
*/
-static HTAB *SMgrRelationHash = NULL;
+static HTAB *SMgrFileHash = NULL;
-static dlist_head unowned_relns;
+static dlist_head unowned_sfiles;
/* local function prototypes */
static void smgrshutdown(int code, Datum arg);
@@ -142,50 +139,50 @@ smgrshutdown(int code, Datum arg)
*
* This does not attempt to actually open the underlying file.
*/
-SMgrRelation
-smgropen(RelFileLocator rlocator, BackendId backend)
+SMgrFileHandle
+smgropen(RelFileLocator rlocator, BackendId backend, ForkNumber forkNum)
{
- RelFileLocatorBackend brlocator;
- SMgrRelation reln;
+ SMgrFileLocator slocator;
+ SMgrFileHandle sfile;
bool found;
- if (SMgrRelationHash == NULL)
+ if (SMgrFileHash == NULL)
{
/* First time through: initialize the hash table */
HASHCTL ctl;
- ctl.keysize = sizeof(RelFileLocatorBackend);
- ctl.entrysize = sizeof(SMgrRelationData);
- SMgrRelationHash = hash_create("smgr relation table", 400,
- &ctl, HASH_ELEM | HASH_BLOBS);
- dlist_init(&unowned_relns);
+ ctl.keysize = sizeof(SMgrFileLocator);
+ ctl.entrysize = sizeof(SMgrFileData);
+ SMgrFileHash = hash_create("smgr relation table", 400,
+ &ctl, HASH_ELEM | HASH_BLOBS);
+ dlist_init(&unowned_sfiles);
}
/* Look up or create an entry */
- brlocator.locator = rlocator;
- brlocator.backend = backend;
- reln = (SMgrRelation) hash_search(SMgrRelationHash,
- &brlocator,
- HASH_ENTER, &found);
+ slocator.locator = rlocator;
+ slocator.backend = backend;
+ slocator.forknum = forkNum;
+ sfile = (SMgrFileHandle) hash_search(SMgrFileHash,
+ &slocator,
+ HASH_ENTER, &found);
/* Initialize it if not present before */
if (!found)
{
/* hash_search already filled in the lookup key */
- reln->smgr_owner = NULL;
- reln->smgr_targblock = InvalidBlockNumber;
- for (int i = 0; i <= MAX_FORKNUM; ++i)
- reln->smgr_cached_nblocks[i] = InvalidBlockNumber;
- reln->smgr_which = 0; /* we only have md.c at present */
+ sfile->smgr_owner = NULL;
+ sfile->smgr_targblock = InvalidBlockNumber;
+ sfile->smgr_cached_nblocks = InvalidBlockNumber;
+ sfile->smgr_which = 0; /* we only have md.c at present */
/* implementation-specific initialization */
- smgrsw[reln->smgr_which].smgr_open(reln);
+ smgrsw[sfile->smgr_which].smgr_open(sfile);
/* it has no owner yet */
- dlist_push_tail(&unowned_relns, &reln->node);
+ dlist_push_tail(&unowned_sfiles, &sfile->node);
}
- return reln;
+ return sfile;
}
/*
@@ -195,7 +192,7 @@ smgropen(RelFileLocator rlocator, BackendId backend)
* the only such owners exist in the relcache.
*/
void
-smgrsetowner(SMgrRelation *owner, SMgrRelation reln)
+smgrsetowner(SMgrFileHandle *owner, SMgrFileHandle sfile)
{
/* We don't support "disowning" an SMgrRelation here, use smgrclearowner */
Assert(owner != NULL);
@@ -206,68 +203,66 @@ smgrsetowner(SMgrRelation *owner, SMgrRelation reln)
* depending on the order of processing. It's ok to close the old
* relcache entry early in that case.)
*
- * If there isn't an old owner, then the reln should be in the unowned
+ * If there isn't an old owner, then the sfile should be in the unowned
* list, and we need to remove it.
*/
- if (reln->smgr_owner)
- *(reln->smgr_owner) = NULL;
+ if (sfile->smgr_owner)
+ *(sfile->smgr_owner) = NULL;
else
- dlist_delete(&reln->node);
+ dlist_delete(&sfile->node);
/* Now establish the ownership relationship. */
- reln->smgr_owner = owner;
- *owner = reln;
+ sfile->smgr_owner = owner;
+ *owner = sfile;
}
/*
- * smgrclearowner() -- Remove long-lived reference to an SMgrRelation object
+ * smgrclearowner() -- Remove long-lived reference to an SMgrFileHandle object
* if one exists
*/
void
-smgrclearowner(SMgrRelation *owner, SMgrRelation reln)
+smgrclearowner(SMgrFileHandle *owner, SMgrFileHandle sfile)
{
/* Do nothing if the SMgrRelation object is not owned by the owner */
- if (reln->smgr_owner != owner)
+ if (sfile->smgr_owner != owner)
return;
/* unset the owner's reference */
*owner = NULL;
/* unset our reference to the owner */
- reln->smgr_owner = NULL;
+ sfile->smgr_owner = NULL;
/* add to list of unowned relations */
- dlist_push_tail(&unowned_relns, &reln->node);
+ dlist_push_tail(&unowned_sfiles, &sfile->node);
}
/*
- * smgrexists() -- Does the underlying file for a fork exist?
+ * smgrexists() -- Does the underlying file exist?
*/
bool
-smgrexists(SMgrRelation reln, ForkNumber forknum)
+smgrexists(SMgrFileHandle sfile)
{
- return smgrsw[reln->smgr_which].smgr_exists(reln, forknum);
+ return smgrsw[sfile->smgr_which].smgr_exists(sfile);
}
/*
- * smgrclose() -- Close and delete an SMgrRelation object.
+ * smgrclose() -- Close and delete an SMgrFile object.
*/
void
-smgrclose(SMgrRelation reln)
+smgrclose(SMgrFileHandle sfile)
{
- SMgrRelation *owner;
- ForkNumber forknum;
+ SMgrFileHandle *owner;
- for (forknum = 0; forknum <= MAX_FORKNUM; forknum++)
- smgrsw[reln->smgr_which].smgr_close(reln, forknum);
+ smgrsw[sfile->smgr_which].smgr_close(sfile);
- owner = reln->smgr_owner;
+ owner = sfile->smgr_owner;
if (!owner)
- dlist_delete(&reln->node);
+ dlist_delete(&sfile->node);
- if (hash_search(SMgrRelationHash,
- &(reln->smgr_rlocator),
+ if (hash_search(SMgrFileHash,
+ &sfile->smgr_locator,
HASH_REMOVE, NULL) == NULL)
elog(ERROR, "SMgrRelation hashtable corrupted");
@@ -284,14 +279,11 @@ smgrclose(SMgrRelation reln)
*
* The object remains valid.
*/
-void
-smgrrelease(SMgrRelation reln)
+static void
+smgrrelease(SMgrFileHandle sfile)
{
- for (ForkNumber forknum = 0; forknum <= MAX_FORKNUM; forknum++)
- {
- smgrsw[reln->smgr_which].smgr_close(reln, forknum);
- reln->smgr_cached_nblocks[forknum] = InvalidBlockNumber;
- }
+ smgrsw[sfile->smgr_which].smgr_close(sfile);
+ sfile->smgr_cached_nblocks = InvalidBlockNumber;
}
/*
@@ -303,16 +295,16 @@ void
smgrreleaseall(void)
{
HASH_SEQ_STATUS status;
- SMgrRelation reln;
+ SMgrFileHandle sfile;
/* Nothing to do if hashtable not set up */
- if (SMgrRelationHash == NULL)
+ if (SMgrFileHash == NULL)
return;
- hash_seq_init(&status, SMgrRelationHash);
+ hash_seq_init(&status, SMgrFileHash);
- while ((reln = (SMgrRelation) hash_seq_search(&status)) != NULL)
- smgrrelease(reln);
+ while ((sfile = (SMgrFileHandle) hash_seq_search(&status)) != NULL)
+ smgrrelease(sfile);
}
/*
@@ -322,16 +314,16 @@ void
smgrcloseall(void)
{
HASH_SEQ_STATUS status;
- SMgrRelation reln;
+ SMgrFileHandle sfile;
/* Nothing to do if hashtable not set up */
- if (SMgrRelationHash == NULL)
+ if (SMgrFileHash == NULL)
return;
- hash_seq_init(&status, SMgrRelationHash);
+ hash_seq_init(&status, SMgrFileHash);
- while ((reln = (SMgrRelation) hash_seq_search(&status)) != NULL)
- smgrclose(reln);
+ while ((sfile = (SMgrFileHandle) hash_seq_search(&status)) != NULL)
+ smgrclose(sfile);
}
/*
@@ -343,111 +335,64 @@ smgrcloseall(void)
* such entry exists already.
*/
void
-smgrcloserellocator(RelFileLocatorBackend rlocator)
+smgrcloserellocator(RelFileLocator rlocator, BackendId backend)
{
- SMgrRelation reln;
-
/* Nothing to do if hashtable not set up */
- if (SMgrRelationHash == NULL)
+ if (SMgrFileHash == NULL)
return;
- reln = (SMgrRelation) hash_search(SMgrRelationHash,
- &rlocator,
- HASH_FIND, NULL);
- if (reln != NULL)
- smgrclose(reln);
+ for (int forknum = 0; forknum <= MAX_FORKNUM; forknum++)
+ {
+ SMgrFileLocator slocator = {rlocator, backend, forknum};
+ SMgrFileHandle sfile;
+
+ sfile = (SMgrFileHandle) hash_search(SMgrFileHash,
+ &slocator,
+ HASH_FIND, NULL);
+ if (sfile != NULL)
+ smgrclose(sfile);
+ }
}
/*
- * smgrcreate() -- Create a new relation.
+ * smgrcreate() -- Create a new file.
*
- * Given an already-created (but presumably unused) SMgrRelation,
+ * Given an already-created (but presumably unused) SMgrFileHandle,
* cause the underlying disk file or other storage for the fork
* to be created.
*/
void
-smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo)
-{
- smgrsw[reln->smgr_which].smgr_create(reln, forknum, isRedo);
-}
-
-/*
- * smgrdosyncall() -- Immediately sync all forks of all given relations
- *
- * All forks of all given relations are synced out to the store.
- *
- * This is equivalent to FlushRelationBuffers() for each smgr relation,
- * then calling smgrimmedsync() for all forks of each relation, but it's
- * significantly quicker so should be preferred when possible.
- */
-void
-smgrdosyncall(SMgrRelation *rels, int nrels)
+smgrcreate(SMgrFileHandle sfile, bool isRedo)
{
- int i = 0;
- ForkNumber forknum;
-
- if (nrels == 0)
- return;
-
- FlushRelationsAllBuffers(rels, nrels);
-
- /*
- * Sync the physical file(s).
- */
- for (i = 0; i < nrels; i++)
- {
- int which = rels[i]->smgr_which;
-
- for (forknum = 0; forknum <= MAX_FORKNUM; forknum++)
- {
- if (smgrsw[which].smgr_exists(rels[i], forknum))
- smgrsw[which].smgr_immedsync(rels[i], forknum);
- }
- }
+ smgrsw[sfile->smgr_which].smgr_create(sfile, isRedo);
}
/*
- * smgrdounlinkall() -- Immediately unlink all forks of all given relations
+ * smgrunlink_multi() -- Immediately unlink given forks of given relation
*
- * All forks of all given relations are removed from the store. This
+ * The given forks of the relation are removed from the store. This
* should not be used during transactional operations, since it can't be
* undone.
*
+ * This handles multiple forks in one call, because the cache invalidation
+ * happens at relation granularity. If we had an smgrunlink() function
+ * to unlink just a single smgr file, and the caller wanted to delete
+ * multiple forks of a single relation, each call would send a new
+ * cache invalidation event, which would be wasteful.
+ *
* If isRedo is true, it is okay for the underlying file(s) to be gone
* already.
*/
void
-smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo)
+smgrunlink_multi(RelFileLocator rlocator, BackendId backend,
+ ForkNumber *forks, int nforks, bool isRedo)
{
- int i = 0;
- RelFileLocatorBackend *rlocators;
- ForkNumber forknum;
-
- if (nrels == 0)
- return;
-
- /*
- * Get rid of any remaining buffers for the relations. bufmgr will just
- * drop them without bothering to write the contents.
- */
- DropRelationsAllBuffers(rels, nrels);
-
- /*
- * create an array which contains all relations to be dropped, and close
- * each relation's forks at the smgr level while at it
- */
- rlocators = palloc(sizeof(RelFileLocatorBackend) * nrels);
- for (i = 0; i < nrels; i++)
- {
- RelFileLocatorBackend rlocator = rels[i]->smgr_rlocator;
- int which = rels[i]->smgr_which;
+ int which;
- rlocators[i] = rlocator;
+ which = 0; /* we only have md.c at present */
- /* Close the forks at smgr level */
- for (forknum = 0; forknum <= MAX_FORKNUM; forknum++)
- smgrsw[which].smgr_close(rels[i], forknum);
- }
+ /* Close the forks at smgr level */
+ smgrcloserellocator(rlocator, backend);
/*
* Send a shared-inval message to force other backends to close any
@@ -457,8 +402,7 @@ smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo)
* back to this backend, too, and thereby provide a backstop that we
* closed our own smgr rel.
*/
- for (i = 0; i < nrels; i++)
- CacheInvalidateSmgr(rlocators[i]);
+ CacheInvalidateSmgr(rlocator, backend);
/*
* Delete the physical file(s).
@@ -468,15 +412,12 @@ smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo)
* xact.
*/
- for (i = 0; i < nrels; i++)
+ for (int i = 0; i < nforks; i++)
{
- int which = rels[i]->smgr_which;
+ SMgrFileLocator slocator = {rlocator, backend, forks[i]};
- for (forknum = 0; forknum <= MAX_FORKNUM; forknum++)
- smgrsw[which].smgr_unlink(rlocators[i], forknum, isRedo);
+ smgrsw[which].smgr_unlink(slocator, isRedo);
}
-
- pfree(rlocators);
}
@@ -490,21 +431,21 @@ smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo)
* causes intervening file space to become filled with zeroes.
*/
void
-smgrextend(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
+smgrextend(SMgrFileHandle sfile, BlockNumber blocknum,
const void *buffer, bool skipFsync)
{
- smgrsw[reln->smgr_which].smgr_extend(reln, forknum, blocknum,
- buffer, skipFsync);
+ smgrsw[sfile->smgr_which].smgr_extend(sfile, blocknum,
+ buffer, skipFsync);
/*
* Normally we expect this to increase nblocks by one, but if the cached
* value isn't as expected, just invalidate it so the next call asks the
* kernel.
*/
- if (reln->smgr_cached_nblocks[forknum] == blocknum)
- reln->smgr_cached_nblocks[forknum] = blocknum + 1;
+ if (sfile->smgr_cached_nblocks == blocknum)
+ sfile->smgr_cached_nblocks = blocknum + 1;
else
- reln->smgr_cached_nblocks[forknum] = InvalidBlockNumber;
+ sfile->smgr_cached_nblocks = InvalidBlockNumber;
}
/*
@@ -515,13 +456,13 @@ smgrextend(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
* record).
*/
bool
-smgrprefetch(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum)
+smgrprefetch(SMgrFileHandle sfile, BlockNumber blocknum)
{
- return smgrsw[reln->smgr_which].smgr_prefetch(reln, forknum, blocknum);
+ return smgrsw[sfile->smgr_which].smgr_prefetch(sfile, blocknum);
}
/*
- * smgrread() -- read a particular block from a relation into the supplied
+ * smgrread() -- read a particular block from a file into the supplied
* buffer.
*
* This routine is called from the buffer manager in order to
@@ -529,10 +470,9 @@ smgrprefetch(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum)
* return pages in the format that POSTGRES expects.
*/
void
-smgrread(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
- void *buffer)
+smgrread(SMgrFileHandle sfile, BlockNumber blocknum, void *buffer)
{
- smgrsw[reln->smgr_which].smgr_read(reln, forknum, blocknum, buffer);
+ smgrsw[sfile->smgr_which].smgr_read(sfile, blocknum, buffer);
}
/*
@@ -551,11 +491,11 @@ smgrread(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
* do not require fsync.
*/
void
-smgrwrite(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
+smgrwrite(SMgrFileHandle sfile, BlockNumber blocknum,
const void *buffer, bool skipFsync)
{
- smgrsw[reln->smgr_which].smgr_write(reln, forknum, blocknum,
- buffer, skipFsync);
+ smgrsw[sfile->smgr_which].smgr_write(sfile, blocknum,
+ buffer, skipFsync);
}
@@ -564,11 +504,11 @@ smgrwrite(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
* blocks.
*/
void
-smgrwriteback(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
+smgrwriteback(SMgrFileHandle sfile, BlockNumber blocknum,
BlockNumber nblocks)
{
- smgrsw[reln->smgr_which].smgr_writeback(reln, forknum, blocknum,
- nblocks);
+ smgrsw[sfile->smgr_which].smgr_writeback(sfile, blocknum,
+ nblocks);
}
/*
@@ -576,18 +516,18 @@ smgrwriteback(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
* supplied relation.
*/
BlockNumber
-smgrnblocks(SMgrRelation reln, ForkNumber forknum)
+smgrnblocks(SMgrFileHandle sfile)
{
BlockNumber result;
/* Check and return if we get the cached value for the number of blocks. */
- result = smgrnblocks_cached(reln, forknum);
+ result = smgrnblocks_cached(sfile);
if (result != InvalidBlockNumber)
return result;
- result = smgrsw[reln->smgr_which].smgr_nblocks(reln, forknum);
+ result = smgrsw[sfile->smgr_which].smgr_nblocks(sfile);
- reln->smgr_cached_nblocks[forknum] = result;
+ sfile->smgr_cached_nblocks = result;
return result;
}
@@ -600,38 +540,41 @@ smgrnblocks(SMgrRelation reln, ForkNumber forknum)
* fork size is not cached.
*/
BlockNumber
-smgrnblocks_cached(SMgrRelation reln, ForkNumber forknum)
+smgrnblocks_cached(SMgrFileHandle sfile)
{
/*
* For now, we only use cached values in recovery due to lack of a shared
* invalidation mechanism for changes in file size.
*/
- if (InRecovery && reln->smgr_cached_nblocks[forknum] != InvalidBlockNumber)
- return reln->smgr_cached_nblocks[forknum];
+ if (InRecovery && sfile->smgr_cached_nblocks != InvalidBlockNumber)
+ return sfile->smgr_cached_nblocks;
return InvalidBlockNumber;
}
/*
- * smgrtruncate() -- Truncate the given forks of supplied relation to
- * each specified numbers of blocks
+ * smgrtruncate_multi() -- Truncate the given forks of supplied relation to
+ * each specified numbers of blocks
*
* The truncation is done immediately, so this can't be rolled back.
*
* The caller must hold AccessExclusiveLock on the relation, to ensure that
* other backends receive the smgr invalidation event that this function sends
* before they access any forks of the relation again.
+ *
+ * Like smgrunlink_multi(), this handles multiple forks in one call because
+ * the cache invalidation happens at relation granularity.
+ *
+ * NB: The caller is responsible for dropping buffers! Before v16, this
+ * function did it.
*/
void
-smgrtruncate(SMgrRelation reln, ForkNumber *forknum, int nforks, BlockNumber *nblocks)
+smgrtruncate_multi(RelFileLocator rlocator, BackendId backend, ForkNumber *forks,
+ int nforks, BlockNumber *nblocks)
{
int i;
- /*
- * Get rid of any buffers for the about-to-be-deleted blocks. bufmgr will
- * just drop them without bothering to write the contents.
- */
- DropRelationBuffers(reln, forknum, nforks, nblocks);
+ Assert(nforks < MAX_FORKNUM + 1);
/*
* Send a shared-inval message to force other backends to close any smgr
@@ -643,15 +586,19 @@ smgrtruncate(SMgrRelation reln, ForkNumber *forknum, int nforks, BlockNumber *nb
* is a performance-critical path.) As in the unlink code, we want to be
* sure the message is sent before we start changing things on-disk.
*/
- CacheInvalidateSmgr(reln->smgr_rlocator);
+ CacheInvalidateSmgr(rlocator, backend);
- /* Do the truncation */
+ /* Do the truncations */
for (i = 0; i < nforks; i++)
{
+ SMgrFileHandle sfile;
+
+ sfile = smgropen(rlocator, backend, forks[i]);
+
/* Make the cached size is invalid if we encounter an error. */
- reln->smgr_cached_nblocks[forknum[i]] = InvalidBlockNumber;
+ sfile->smgr_cached_nblocks = InvalidBlockNumber;
- smgrsw[reln->smgr_which].smgr_truncate(reln, forknum[i], nblocks[i]);
+ smgrsw[sfile->smgr_which].smgr_truncate(sfile, nblocks[i]);
/*
* We might as well update the local smgr_cached_nblocks values. The
@@ -660,7 +607,7 @@ smgrtruncate(SMgrRelation reln, ForkNumber *forknum, int nforks, BlockNumber *nb
* smgr_vm_nblocks, and these ones too at the next command boundary.
* But these ensure they aren't outright wrong until then.
*/
- reln->smgr_cached_nblocks[forknum[i]] = nblocks[i];
+ sfile->smgr_cached_nblocks = nblocks[i];
}
}
@@ -688,9 +635,9 @@ smgrtruncate(SMgrRelation reln, ForkNumber *forknum, int nforks, BlockNumber *nb
* otherwise the sync is not very meaningful.
*/
void
-smgrimmedsync(SMgrRelation reln, ForkNumber forknum)
+smgrimmedsync(SMgrFileHandle sfile)
{
- smgrsw[reln->smgr_which].smgr_immedsync(reln, forknum);
+ smgrsw[sfile->smgr_which].smgr_immedsync(sfile);
}
/*
@@ -714,14 +661,14 @@ AtEOXact_SMgr(void)
* Zap all unowned SMgrRelations. We rely on smgrclose() to remove each
* one from the list.
*/
- dlist_foreach_modify(iter, &unowned_relns)
+ dlist_foreach_modify(iter, &unowned_sfiles)
{
- SMgrRelation rel = dlist_container(SMgrRelationData, node,
- iter.cur);
+ SMgrFileHandle sfile = dlist_container(SMgrFileData, node,
+ iter.cur);
- Assert(rel->smgr_owner == NULL);
+ Assert(sfile->smgr_owner == NULL);
- smgrclose(rel);
+ smgrclose(sfile);
}
}
diff --git a/src/backend/utils/cache/inval.c b/src/backend/utils/cache/inval.c
index 0008826f67c..7a1d8c012c2 100644
--- a/src/backend/utils/cache/inval.c
+++ b/src/backend/utils/cache/inval.c
@@ -661,11 +661,12 @@ LocalExecuteInvalidationMessage(SharedInvalidationMessage *msg)
* We could have smgr entries for relations of other databases, so no
* short-circuit test is possible here.
*/
- RelFileLocatorBackend rlocator;
+ RelFileLocator rlocator;
+ BackendId backend;
- rlocator.locator = msg->sm.rlocator;
- rlocator.backend = (msg->sm.backend_hi << 16) | (int) msg->sm.backend_lo;
- smgrcloserellocator(rlocator);
+ rlocator = msg->sm.rlocator;
+ backend = (msg->sm.backend_hi << 16) | (int) msg->sm.backend_lo;
+ smgrcloserellocator(rlocator, backend);
}
else if (msg->id == SHAREDINVALRELMAP_ID)
{
@@ -1459,14 +1460,14 @@ CacheInvalidateRelcacheByRelid(Oid relid)
* Thus, the maximum possible backend ID is 2^23-1.
*/
void
-CacheInvalidateSmgr(RelFileLocatorBackend rlocator)
+CacheInvalidateSmgr(RelFileLocator rlocator, BackendId backend)
{
SharedInvalidationMessage msg;
msg.sm.id = SHAREDINVALSMGR_ID;
- msg.sm.backend_hi = rlocator.backend >> 16;
- msg.sm.backend_lo = rlocator.backend & 0xffff;
- msg.sm.rlocator = rlocator.locator;
+ msg.sm.backend_hi = backend >> 16;
+ msg.sm.backend_lo = backend & 0xffff;
+ msg.sm.rlocator = rlocator;
/* check AddCatcacheInvalidationMessage() for an explanation */
VALGRIND_MAKE_MEM_DEFINED(&msg, sizeof(msg));
diff --git a/src/backend/utils/cache/relcache.c b/src/backend/utils/cache/relcache.c
index 13f79873733..45c22e0c137 100644
--- a/src/backend/utils/cache/relcache.c
+++ b/src/backend/utils/cache/relcache.c
@@ -419,7 +419,7 @@ AllocateRelationDesc(Form_pg_class relp)
relation = (Relation) palloc0(sizeof(RelationData));
/* make sure relation is marked as having no open file yet */
- relation->rd_smgr = NULL;
+ MemSet(relation->rd_smgr, 0, sizeof(relation->rd_smgr));
/*
* Copy the relation tuple form
@@ -1248,7 +1248,7 @@ retry:
RelationInitPhysicalAddr(relation);
/* make sure relation is marked as having no open file yet */
- relation->rd_smgr = NULL;
+ MemSet(relation->rd_smgr, 0, sizeof(relation->rd_smgr));
/*
* now we can free the memory allocated for pg_class_tuple
@@ -1877,7 +1877,7 @@ formrdesc(const char *relationName, Oid relationReltype,
relation = (Relation) palloc0(sizeof(RelationData));
/* make sure relation is marked as having no open file yet */
- relation->rd_smgr = NULL;
+ MemSet(relation->rd_smgr, 0, sizeof(relation->rd_smgr));
/*
* initialize reference count: 1 because it is nailed in cache
@@ -2701,7 +2701,8 @@ RelationClearRelation(Relation relation, bool rebuild)
}
/* rd_smgr must not be swapped, due to back-links from smgr level */
- SWAPFIELD(SMgrRelation, rd_smgr);
+ for (int i = 0; i <= MAX_FORKNUM; i++)
+ SWAPFIELD(SMgrFileHandle, rd_smgr[i]);
/* rd_refcnt must be preserved */
SWAPFIELD(int, rd_refcnt);
/* isnailed shouldn't change */
@@ -3532,7 +3533,7 @@ RelationBuildLocalRelation(const char *relname,
rel = (Relation) palloc0(sizeof(RelationData));
/* make sure relation is marked as having no open file yet */
- rel->rd_smgr = NULL;
+ MemSet(rel->rd_smgr, 0, sizeof(rel->rd_smgr));
/* mark it nailed if appropriate */
rel->rd_isnailed = nailit;
@@ -3764,7 +3765,11 @@ RelationSetNewRelfilenumber(Relation relation, char persistence)
*/
if (IsBinaryUpgrade)
{
- SMgrRelation srel;
+ ForkNumber forks[MAX_FORKNUM + 1];
+ RelFileLocatorBackend locator = {
+ relation->rd_locator,
+ relation->rd_backend
+ };
/*
* During a binary upgrade, we use this code path to ensure that
@@ -3781,9 +3786,16 @@ RelationSetNewRelfilenumber(Relation relation, char persistence)
* fails at this stage, the new cluster will need to be recreated
* anyway.
*/
- srel = smgropen(relation->rd_locator, relation->rd_backend);
- smgrdounlinkall(&srel, 1, false);
- smgrclose(srel);
+
+ /*
+ * Get rid of any remaining buffers for the relations. bufmgr will just
+ * drop them without bothering to write the contents.
+ */
+ DropRelationsAllBuffers(&locator, 1);
+
+ for (int i = 0; i <= MAX_FORKNUM; i++)
+ forks[i] = i;
+ smgrunlink_multi(relation->rd_locator, relation->rd_backend, forks, MAX_FORKNUM + 1, false);
}
else
{
@@ -3811,7 +3823,7 @@ RelationSetNewRelfilenumber(Relation relation, char persistence)
else if (RELKIND_HAS_STORAGE(relation->rd_rel->relkind))
{
/* handle these directly, at least for now */
- SMgrRelation srel;
+ SMgrFileHandle srel;
srel = RelationCreateStorage(newrlocator, persistence, true);
smgrclose(srel);
@@ -6298,7 +6310,7 @@ load_relcache_init_file(bool shared)
/*
* Reset transient-state fields in the relcache entry
*/
- rel->rd_smgr = NULL;
+ MemSet(rel->rd_smgr, 0, sizeof(rel->rd_smgr));
if (rel->rd_isnailed)
rel->rd_refcnt = 1;
else
diff --git a/src/common/relpath.c b/src/common/relpath.c
index 87de5f6c960..ae2d384fb34 100644
--- a/src/common/relpath.c
+++ b/src/common/relpath.c
@@ -129,7 +129,7 @@ GetDatabasePath(Oid dbOid, Oid spcOid)
}
/*
- * GetRelationPath - construct path to a relation's file
+ * GetSMgrFilePath - construct path to a relation's file
*
* Result is a palloc'd string.
*
@@ -138,7 +138,7 @@ GetDatabasePath(Oid dbOid, Oid spcOid)
* the trouble considering BackendId is just int anyway.
*/
char *
-GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber,
+GetSMgrFilePath(Oid dbOid, Oid spcOid, RelFileNumber relNumber,
int backendId, ForkNumber forkNumber)
{
char *path;
diff --git a/src/include/catalog/storage.h b/src/include/catalog/storage.h
index 45a3c7835cb..c96f448225b 100644
--- a/src/include/catalog/storage.h
+++ b/src/include/catalog/storage.h
@@ -22,19 +22,20 @@
/* GUC variables */
extern PGDLLIMPORT int wal_skip_threshold;
-extern SMgrRelation RelationCreateStorage(RelFileLocator rlocator,
- char relpersistence,
- bool register_delete);
+extern SMgrFileHandle RelationCreateStorage(RelFileLocator rlocator,
+ char relpersistence,
+ bool register_delete);
extern void RelationDropStorage(Relation rel);
extern void RelationPreserveStorage(RelFileLocator rlocator, bool atCommit);
extern void RelationPreTruncate(Relation rel);
extern void RelationTruncate(Relation rel, BlockNumber nblocks);
-extern void RelationCopyStorage(SMgrRelation src, SMgrRelation dst,
- ForkNumber forkNum, char relpersistence);
+extern void RelationCopyStorage(SMgrFileHandle src, SMgrFileHandle dst,
+ char relpersistence);
extern bool RelFileLocatorSkippingWAL(RelFileLocator rlocator);
extern Size EstimatePendingSyncsSpace(void);
extern void SerializePendingSyncs(Size maxSize, char *startAddress);
extern void RestorePendingSyncs(char *startAddress);
+extern void DropRelationFiles(RelFileLocator *delrels, int ndelrels, bool isRedo);
/*
* These functions used to be in storage/smgr/smgr.c, which explains the
diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h
index 511c21682ec..12df11c7dfc 100644
--- a/src/include/common/relpath.h
+++ b/src/include/common/relpath.h
@@ -73,25 +73,25 @@ extern int forkname_chars(const char *str, ForkNumber *fork);
*/
extern char *GetDatabasePath(Oid dbOid, Oid spcOid);
-extern char *GetRelationPath(Oid dbOid, Oid spcOid, RelFileNumber relNumber,
+extern char *GetSMgrFilePath(Oid dbOid, Oid spcOid, RelFileNumber relNumber,
int backendId, ForkNumber forkNumber);
/*
* Wrapper macros for GetRelationPath. Beware of multiple
- * evaluation of the RelFileLocator or RelFileLocatorBackend argument!
+ * evaluation of the RelFileLocator or SMgrFileLocator argument!
*/
/* First argument is a RelFileLocator */
#define relpathbackend(rlocator, backend, forknum) \
- GetRelationPath((rlocator).dbOid, (rlocator).spcOid, (rlocator).relNumber, \
+ GetSMgrFilePath((rlocator).dbOid, (rlocator).spcOid, (rlocator).relNumber, \
backend, forknum)
/* First argument is a RelFileLocator */
#define relpathperm(rlocator, forknum) \
relpathbackend(rlocator, InvalidBackendId, forknum)
-/* First argument is a RelFileLocatorBackend */
-#define relpath(rlocator, forknum) \
- relpathbackend((rlocator).locator, (rlocator).backend, forknum)
+/* First argument is a SMgrFileLocator */
+#define smgrfilepath(slocator) \
+ GetSMgrFilePath((slocator).locator.dbOid, (slocator).locator.spcOid, (slocator).locator.relNumber, (slocator).backend, (slocator).forknum)
#endif /* RELPATH_H */
diff --git a/src/include/storage/buf_internals.h b/src/include/storage/buf_internals.h
index 0b448147407..6a37e0ce6b4 100644
--- a/src/include/storage/buf_internals.h
+++ b/src/include/storage/buf_internals.h
@@ -415,11 +415,10 @@ extern int BufTableInsert(BufferTag *tagPtr, uint32 hashcode, int buf_id);
extern void BufTableDelete(BufferTag *tagPtr, uint32 hashcode);
/* localbuf.c */
-extern PrefetchBufferResult PrefetchLocalBuffer(SMgrRelation smgr,
- ForkNumber forkNum,
+extern PrefetchBufferResult PrefetchLocalBuffer(SMgrFileHandle smgr,
BlockNumber blockNum);
-extern BufferDesc *LocalBufferAlloc(SMgrRelation smgr, ForkNumber forkNum,
- BlockNumber blockNum, bool *foundPtr, IOContext *io_context);
+extern BufferDesc *LocalBufferAlloc(SMgrFileHandle smgr, BlockNumber blockNum,
+ bool *foundPtr, IOContext *io_context);
extern void MarkLocalBufferDirty(Buffer buffer);
extern void DropRelationLocalBuffers(RelFileLocator rlocator,
ForkNumber forkNum,
diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h
index b8a18b8081f..7de50bf71b7 100644
--- a/src/include/storage/bufmgr.h
+++ b/src/include/storage/bufmgr.h
@@ -64,7 +64,7 @@ typedef struct PrefetchBufferResult
struct WritebackContext;
/* forward declared, to avoid including smgr.h here */
-struct SMgrRelationData;
+struct SMgrFileData;
/* in globals.c ... this duplicates miscadmin.h */
extern PGDLLIMPORT int NBuffers;
@@ -115,8 +115,7 @@ extern PGDLLIMPORT int32 *LocalRefCount;
/*
* prototypes for functions in bufmgr.c
*/
-extern PrefetchBufferResult PrefetchSharedBuffer(struct SMgrRelationData *smgr_reln,
- ForkNumber forkNum,
+extern PrefetchBufferResult PrefetchSharedBuffer(struct SMgrFileData *smgr_file,
BlockNumber blockNum);
extern PrefetchBufferResult PrefetchBuffer(Relation reln, ForkNumber forkNum,
BlockNumber blockNum);
@@ -144,18 +143,26 @@ extern void CheckPointBuffers(int flags);
extern BlockNumber BufferGetBlockNumber(Buffer buffer);
extern BlockNumber RelationGetNumberOfBlocksInFork(Relation relation,
ForkNumber forkNum);
-extern void FlushOneBuffer(Buffer buffer);
-extern void FlushRelationBuffers(Relation rel);
-extern void FlushRelationsAllBuffers(struct SMgrRelationData **smgrs, int nrels);
extern void CreateAndCopyRelationData(RelFileLocator src_rlocator,
RelFileLocator dst_rlocator,
bool permanent);
+
+extern void FlushOneBuffer(Buffer buffer);
+extern void FlushRelationBuffers(Relation rel);
+extern void FlushRelationsAllBuffers(RelFileLocator *locators, int nlocators);
extern void FlushDatabaseBuffers(Oid dbid);
-extern void DropRelationBuffers(struct SMgrRelationData *smgr_reln,
+
+extern void DropRelationBuffers(RelFileLocator rlocator, BackendId backend,
ForkNumber *forkNum,
int nforks, BlockNumber *firstDelBlock);
-extern void DropRelationsAllBuffers(struct SMgrRelationData **smgr_reln,
- int nlocators);
+
+typedef struct RelFileLocatorBackend
+{
+ RelFileLocator locator;
+ BackendId backend;
+} RelFileLocatorBackend;
+
+extern void DropRelationsAllBuffers(RelFileLocatorBackend *locators, int nlocators);
extern void DropDatabaseBuffers(Oid dbid);
#define RelationGetNumberOfBlocks(reln) \
diff --git a/src/include/storage/md.h b/src/include/storage/md.h
index 8f32af9ef3d..4046584a830 100644
--- a/src/include/storage/md.h
+++ b/src/include/storage/md.h
@@ -21,28 +21,26 @@
/* md storage manager functionality */
extern void mdinit(void);
-extern void mdopen(SMgrRelation reln);
-extern void mdclose(SMgrRelation reln, ForkNumber forknum);
-extern void mdcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo);
-extern bool mdexists(SMgrRelation reln, ForkNumber forknum);
-extern void mdunlink(RelFileLocatorBackend rlocator, ForkNumber forknum, bool isRedo);
-extern void mdextend(SMgrRelation reln, ForkNumber forknum,
+extern void mdopen(SMgrFileHandle sfile);
+extern void mdclose(SMgrFileHandle sfile);
+extern void mdcreate(SMgrFileHandle sfile, bool isRedo);
+extern bool mdexists(SMgrFileHandle sfile);
+extern void mdunlink(SMgrFileLocator slocator, bool isRedo);
+extern void mdextend(SMgrFileHandle sfile,
BlockNumber blocknum, const void *buffer, bool skipFsync);
-extern bool mdprefetch(SMgrRelation reln, ForkNumber forknum,
+extern bool mdprefetch(SMgrFileHandle sfile,
BlockNumber blocknum);
-extern void mdread(SMgrRelation reln, ForkNumber forknum, BlockNumber blocknum,
+extern void mdread(SMgrFileHandle sfile, BlockNumber blocknum,
void *buffer);
-extern void mdwrite(SMgrRelation reln, ForkNumber forknum,
+extern void mdwrite(SMgrFileHandle sfile,
BlockNumber blocknum, const void *buffer, bool skipFsync);
-extern void mdwriteback(SMgrRelation reln, ForkNumber forknum,
+extern void mdwriteback(SMgrFileHandle sfile,
BlockNumber blocknum, BlockNumber nblocks);
-extern BlockNumber mdnblocks(SMgrRelation reln, ForkNumber forknum);
-extern void mdtruncate(SMgrRelation reln, ForkNumber forknum,
- BlockNumber nblocks);
-extern void mdimmedsync(SMgrRelation reln, ForkNumber forknum);
+extern BlockNumber mdnblocks(SMgrFileHandle sfile);
+extern void mdtruncate(SMgrFileHandle sfile, BlockNumber nblocks);
+extern void mdimmedsync(SMgrFileHandle sfile);
extern void ForgetDatabaseSyncRequests(Oid dbid);
-extern void DropRelationFiles(RelFileLocator *delrels, int ndelrels, bool isRedo);
/* md sync callbacks */
extern int mdsyncfiletag(const FileTag *ftag, char *path);
diff --git a/src/include/storage/relfilelocator.h b/src/include/storage/relfilelocator.h
index 61cf0169bd7..e2c22b8e6f9 100644
--- a/src/include/storage/relfilelocator.h
+++ b/src/include/storage/relfilelocator.h
@@ -53,6 +53,8 @@
* Note: various places use RelFileLocator in hashtable keys. Therefore,
* there *must not* be any unused padding bytes in this struct. That
* should be safe as long as all the fields are of type Oid.
+ *
+ * See also SMgrFileLocator in smgr.h.
*/
typedef struct RelFileLocator
{
@@ -62,38 +64,15 @@ typedef struct RelFileLocator
} RelFileLocator;
/*
- * Augmenting a relfilelocator with the backend ID provides all the information
- * we need to locate the physical storage. The backend ID is InvalidBackendId
- * for regular relations (those accessible to more than one backend), or the
- * owning backend's ID for backend-local relations. Backend-local relations
- * are always transient and removed in case of a database crash; they are
- * never WAL-logged or fsync'd.
- */
-typedef struct RelFileLocatorBackend
-{
- RelFileLocator locator;
- BackendId backend;
-} RelFileLocatorBackend;
-
-#define RelFileLocatorBackendIsTemp(rlocator) \
- ((rlocator).backend != InvalidBackendId)
-
-/*
- * Note: RelFileLocatorEquals and RelFileLocatorBackendEquals compare relNumber
+ * Note: RelFileLocatorEquals compares relNumber
* first since that is most likely to be different in two unequal
* RelFileLocators. It is probably redundant to compare spcOid if the other
* fields are found equal, but do it anyway to be sure. Likewise for checking
- * the backend ID in RelFileLocatorBackendEquals.
+ * the backend ID in SMgrFileLocatorBackendEquals.
*/
#define RelFileLocatorEquals(locator1, locator2) \
((locator1).relNumber == (locator2).relNumber && \
(locator1).dbOid == (locator2).dbOid && \
(locator1).spcOid == (locator2).spcOid)
-#define RelFileLocatorBackendEquals(locator1, locator2) \
- ((locator1).locator.relNumber == (locator2).locator.relNumber && \
- (locator1).locator.dbOid == (locator2).locator.dbOid && \
- (locator1).backend == (locator2).backend && \
- (locator1).locator.spcOid == (locator2).locator.spcOid)
-
#endif /* RELFILELOCATOR_H */
diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h
index 0935144f425..fe7282127ed 100644
--- a/src/include/storage/smgr.h
+++ b/src/include/storage/smgr.h
@@ -18,31 +18,56 @@
#include "storage/block.h"
#include "storage/relfilelocator.h"
+
/*
- * smgr.c maintains a table of SMgrRelation objects, which are essentially
- * cached file handles. An SMgrRelation is created (if not already present)
+ * SMgrFileLocator contains all the information needed to locate the physical
+ * storage of a relation fork, or some other file that is managed by the buffer
+ * manager.
+ *
+ * The backend ID is InvalidBackendId for regular relations (those accessible
+ * to more than one backend), or the owning backend's ID for backend-local
+ * relations. Backend-local relations are always transient and removed in
+ * case of a database crash; they are never WAL-logged or fsync'd.
+ */
+typedef struct SMgrFileLocator
+{
+ RelFileLocator locator;
+ BackendId backend;
+ ForkNumber forknum;
+} SMgrFileLocator;
+
+#define SMgrFileLocatorIsTemp(slocator) \
+ ((slocator).backend != InvalidBackendId)
+
+/*
+ * smgr.c maintains a table of SMgrFileData objects, which are essentially
+ * cached file handles. An SMgrFile is created (if not already present)
* by smgropen(), and destroyed by smgrclose(). Note that neither of these
* operations imply I/O, they just create or destroy a hashtable entry.
* (But smgrclose() may release associated resources, such as OS-level file
* descriptors.)
*
- * An SMgrRelation may have an "owner", which is just a pointer to it from
- * somewhere else; smgr.c will clear this pointer if the SMgrRelation is
+ * An SMgrFile may have an "owner", which is just a pointer to it from
+ * somewhere else; smgr.c will clear this pointer if the SMgrFile is
* closed. We use this to avoid dangling pointers from relcache to smgr
* without having to make the smgr explicitly aware of relcache. There
* can't be more than one "owner" pointer per SMgrRelation, but that's
* all we need.
*
- * SMgrRelations that do not have an "owner" are considered to be transient,
+ * SMgrFiles that do not have an "owner" are considered to be transient,
* and are deleted at end of transaction.
+ *
+ * A file that is represented by an SMgrFile can be managed by the buffer
+ * manager. Currently, it's only used for relation files, but could be used
+ * for SLRUs and other things in the future.
*/
-typedef struct SMgrRelationData
+typedef struct SMgrFileData
{
- /* rlocator is the hashtable lookup key, so it must be first! */
- RelFileLocatorBackend smgr_rlocator; /* relation physical identifier */
+ /* locator is the hashtable lookup key, so must be first! */
+ SMgrFileLocator smgr_locator; /* file physical identifier */
/* pointer to owning pointer, or NULL if none */
- struct SMgrRelationData **smgr_owner;
+ struct SMgrFileData **smgr_owner;
/*
* The following fields are reset to InvalidBlockNumber upon a cache flush
@@ -51,7 +76,7 @@ typedef struct SMgrRelationData
* invalidation for fork extension.
*/
BlockNumber smgr_targblock; /* current insertion target block */
- BlockNumber smgr_cached_nblocks[MAX_FORKNUM + 1]; /* last known size */
+ BlockNumber smgr_cached_nblocks; /* last known size */
/* additional public fields may someday exist here */
@@ -65,46 +90,45 @@ typedef struct SMgrRelationData
* for md.c; per-fork arrays of the number of open segments
* (md_num_open_segs) and the segments themselves (md_seg_fds).
*/
- int md_num_open_segs[MAX_FORKNUM + 1];
- struct _MdfdVec *md_seg_fds[MAX_FORKNUM + 1];
+ int md_num_open_segs;
+ struct _MdfdVec *md_seg_fds;
- /* if unowned, list link in list of all unowned SMgrRelations */
+ /* if unowned, list link in list of all unowned SMgrFiles */
dlist_node node;
-} SMgrRelationData;
+} SMgrFileData;
-typedef SMgrRelationData *SMgrRelation;
+typedef SMgrFileData *SMgrFileHandle;
#define SmgrIsTemp(smgr) \
- RelFileLocatorBackendIsTemp((smgr)->smgr_rlocator)
+ SMgrFileLocatorIsTemp((smgr)->smgr_locator)
extern void smgrinit(void);
-extern SMgrRelation smgropen(RelFileLocator rlocator, BackendId backend);
-extern bool smgrexists(SMgrRelation reln, ForkNumber forknum);
-extern void smgrsetowner(SMgrRelation *owner, SMgrRelation reln);
-extern void smgrclearowner(SMgrRelation *owner, SMgrRelation reln);
-extern void smgrclose(SMgrRelation reln);
+extern SMgrFileHandle smgropen(RelFileLocator rlocator, BackendId backend, ForkNumber forkNum);
+extern bool smgrexists(SMgrFileHandle sfile);
+extern void smgrsetowner(SMgrFileHandle *owner, SMgrFileHandle sfile);
+extern void smgrclearowner(SMgrFileHandle *owner, SMgrFileHandle sfile);
+extern void smgrclose(SMgrFileHandle sfile);
extern void smgrcloseall(void);
-extern void smgrcloserellocator(RelFileLocatorBackend rlocator);
-extern void smgrrelease(SMgrRelation reln);
extern void smgrreleaseall(void);
-extern void smgrcreate(SMgrRelation reln, ForkNumber forknum, bool isRedo);
-extern void smgrdosyncall(SMgrRelation *rels, int nrels);
-extern void smgrdounlinkall(SMgrRelation *rels, int nrels, bool isRedo);
-extern void smgrextend(SMgrRelation reln, ForkNumber forknum,
+extern void smgrcreate(SMgrFileHandle sfile, bool isRedo);
+extern void smgrextend(SMgrFileHandle sfile,
BlockNumber blocknum, const void *buffer, bool skipFsync);
-extern bool smgrprefetch(SMgrRelation reln, ForkNumber forknum,
- BlockNumber blocknum);
-extern void smgrread(SMgrRelation reln, ForkNumber forknum,
+extern bool smgrprefetch(SMgrFileHandle sfile, BlockNumber blocknum);
+extern void smgrread(SMgrFileHandle sfile,
BlockNumber blocknum, void *buffer);
-extern void smgrwrite(SMgrRelation reln, ForkNumber forknum,
+extern void smgrwrite(SMgrFileHandle sfile,
BlockNumber blocknum, const void *buffer, bool skipFsync);
-extern void smgrwriteback(SMgrRelation reln, ForkNumber forknum,
+extern void smgrwriteback(SMgrFileHandle sfile,
BlockNumber blocknum, BlockNumber nblocks);
-extern BlockNumber smgrnblocks(SMgrRelation reln, ForkNumber forknum);
-extern BlockNumber smgrnblocks_cached(SMgrRelation reln, ForkNumber forknum);
-extern void smgrtruncate(SMgrRelation reln, ForkNumber *forknum,
- int nforks, BlockNumber *nblocks);
-extern void smgrimmedsync(SMgrRelation reln, ForkNumber forknum);
+extern BlockNumber smgrnblocks(SMgrFileHandle sfile);
+extern BlockNumber smgrnblocks_cached(SMgrFileHandle sfile);
+extern void smgrimmedsync(SMgrFileHandle sfile);
+
+extern void smgrtruncate_multi(RelFileLocator rlocator, BackendId backend, ForkNumber *forks, int nforks, BlockNumber *nblocks);
+extern void smgrunlink_multi(RelFileLocator rlocator, BackendId backend, ForkNumber *forks, int nforks, bool isRedo);
+
+extern void smgrcloserellocator(RelFileLocator rlocator, BackendId backend);
+
extern void AtEOXact_SMgr(void);
extern bool ProcessBarrierSmgrRelease(void);
diff --git a/src/include/utils/inval.h b/src/include/utils/inval.h
index 14b4eac0630..6c279422b2e 100644
--- a/src/include/utils/inval.h
+++ b/src/include/utils/inval.h
@@ -48,7 +48,7 @@ extern void CacheInvalidateRelcacheByTuple(HeapTuple classTuple);
extern void CacheInvalidateRelcacheByRelid(Oid relid);
-extern void CacheInvalidateSmgr(RelFileLocatorBackend rlocator);
+extern void CacheInvalidateSmgr(RelFileLocator rlocator, BackendId backend);
extern void CacheInvalidateRelmap(Oid databaseId);
diff --git a/src/include/utils/rel.h b/src/include/utils/rel.h
index 67f994cb3e2..7ced81d869b 100644
--- a/src/include/utils/rel.h
+++ b/src/include/utils/rel.h
@@ -54,7 +54,7 @@ typedef LockInfoData *LockInfo;
typedef struct RelationData
{
RelFileLocator rd_locator; /* relation physical identifier */
- SMgrRelation rd_smgr; /* cached file handle, or NULL */
+ SMgrFileHandle rd_smgr[MAX_FORKNUM + 1]; /* cached file handles, or NULLs */
int rd_refcnt; /* reference count */
BackendId rd_backend; /* owning backend id, if temporary relation */
bool rd_islocaltemp; /* rel is a temp rel of this session */
@@ -562,15 +562,15 @@ typedef struct ViewOptions
* Note: since a relcache flush can cause the file handle to be closed again,
* it's unwise to hold onto the pointer returned by this function for any
* long period. Recommended practice is to just re-execute RelationGetSmgr
- * each time you need to access the SMgrRelation. It's quite cheap in
+ * each time you need to access the SMgrFileHandle. It's quite cheap in
* comparison to whatever an smgr function is going to do.
*/
-static inline SMgrRelation
-RelationGetSmgr(Relation rel)
+static inline SMgrFileHandle
+RelationGetSmgr(Relation rel, ForkNumber forkNum)
{
- if (unlikely(rel->rd_smgr == NULL))
- smgrsetowner(&(rel->rd_smgr), smgropen(rel->rd_locator, rel->rd_backend));
- return rel->rd_smgr;
+ if (unlikely(rel->rd_smgr[forkNum] == NULL))
+ smgrsetowner(&(rel->rd_smgr[forkNum]), smgropen(rel->rd_locator, rel->rd_backend, forkNum));
+ return rel->rd_smgr[forkNum];
}
/*
@@ -580,11 +580,15 @@ RelationGetSmgr(Relation rel)
static inline void
RelationCloseSmgr(Relation relation)
{
- if (relation->rd_smgr != NULL)
- smgrclose(relation->rd_smgr);
-
- /* smgrclose should unhook from owner pointer */
- Assert(relation->rd_smgr == NULL);
+ for (int i = 0; i <= MAX_FORKNUM; i++)
+ {
+ if (relation->rd_smgr[i] != NULL)
+ {
+ smgrclose(relation->rd_smgr[i]);
+ /* smgrclose should unhook from owner pointer */
+ Assert(relation->rd_smgr[i] == NULL);
+ }
+ }
}
#endif /* !FRONTEND */
@@ -597,7 +601,7 @@ RelationCloseSmgr(Relation relation)
* so there's no need to re-open the smgr handle if it's not currently open.
*/
#define RelationGetTargetBlock(relation) \
- ( (relation)->rd_smgr != NULL ? (relation)->rd_smgr->smgr_targblock : InvalidBlockNumber )
+ ( (relation)->rd_smgr[MAIN_FORKNUM] != NULL ? (relation)->rd_smgr[MAIN_FORKNUM]->smgr_targblock : InvalidBlockNumber )
/*
* RelationSetTargetBlock
@@ -605,7 +609,7 @@ RelationCloseSmgr(Relation relation)
*/
#define RelationSetTargetBlock(relation, targblock) \
do { \
- RelationGetSmgr(relation)->smgr_targblock = (targblock); \
+ RelationGetSmgr(relation, MAIN_FORKNUM)->smgr_targblock = (targblock); \
} while (0)
/*
--
2.30.2
[text/x-patch] v3-0002-WIP-SLRUs.patch (148.6K, ../../[email protected]/3-v3-0002-WIP-SLRUs.patch)
download | inline diff:
From 02eebb8c442850dcd9b437137eec06f6a16334e2 Mon Sep 17 00:00:00 2001
From: Heikki Linnakangas <[email protected]>
Date: Mon, 27 Feb 2023 15:34:59 +0200
Subject: [PATCH v3 2/2] WIP: SLRUs
This is Thomas's patch refactored over the per-fork SMgrFile patch:
- each SLRU segment is represented by a separate SMgrFile.
- md.c implementation handles SLRUs, too
---
src/backend/access/transam/clog.c | 199 +---
src/backend/access/transam/commit_ts.c | 156 +--
src/backend/access/transam/multixact.c | 357 ++----
src/backend/access/transam/slru.c | 1495 +++---------------------
src/backend/access/transam/subtrans.c | 108 +-
src/backend/access/transam/xact.c | 2 +
src/backend/access/transam/xlog.c | 15 +-
src/backend/commands/async.c | 67 +-
src/backend/storage/buffer/buf_init.c | 17 +-
src/backend/storage/buffer/bufmgr.c | 62 +-
src/backend/storage/ipc/ipci.c | 4 -
src/backend/storage/lmgr/predicate.c | 62 +-
src/backend/storage/smgr/md.c | 9 +-
src/backend/storage/smgr/smgr.c | 42 +
src/backend/storage/sync/sync.c | 20 +-
src/backend/utils/mmgr/mcxt.c | 40 +-
src/common/relpath.c | 27 +-
src/include/access/clog.h | 6 -
src/include/access/commit_ts.h | 3 -
src/include/access/multixact.h | 3 -
src/include/access/slru.h | 180 +--
src/include/access/slrulist.h | 30 +
src/include/access/subtrans.h | 3 -
src/include/common/relpath.h | 3 +
src/include/storage/buf_internals.h | 13 +
src/include/storage/bufmgr.h | 6 +
src/include/storage/smgr.h | 1 +
src/test/modules/Makefile | 3 +-
28 files changed, 725 insertions(+), 2208 deletions(-)
create mode 100644 src/include/access/slrulist.h
diff --git a/src/backend/access/transam/clog.c b/src/backend/access/transam/clog.c
index 4a431d58767..b6f5ae987b1 100644
--- a/src/backend/access/transam/clog.c
+++ b/src/backend/access/transam/clog.c
@@ -41,6 +41,8 @@
#include "miscadmin.h"
#include "pg_trace.h"
#include "pgstat.h"
+#include "storage/bufmgr.h"
+#include "storage/buf_internals.h"
#include "storage/proc.h"
#include "storage/sync.h"
@@ -81,15 +83,8 @@
*/
#define THRESHOLD_SUBTRANS_CLOG_OPT 5
-/*
- * Link to shared-memory data structures for CLOG control
- */
-static SlruCtlData XactCtlData;
-
-#define XactCtl (&XactCtlData)
-
-static int ZeroCLOGPage(int pageno, bool writeXlog);
+static Buffer ZeroCLOGPage(int pageno, bool writeXlog);
static bool CLOGPagePrecedes(int page1, int page2);
static void WriteZeroPageXlogRec(int pageno);
static void WriteTruncateXlogRec(int pageno, TransactionId oldestXact,
@@ -99,7 +94,7 @@ static void TransactionIdSetPageStatus(TransactionId xid, int nsubxids,
XLogRecPtr lsn, int pageno,
bool all_xact_same_page);
static void TransactionIdSetStatusBit(TransactionId xid, XidStatus status,
- XLogRecPtr lsn, int slotno);
+ XLogRecPtr lsn, Buffer buffer);
static void set_status_by_pages(int nsubxids, TransactionId *subxids,
XidStatus status, XLogRecPtr lsn);
static bool TransactionGroupUpdateXidStatus(TransactionId xid,
@@ -339,13 +334,12 @@ TransactionIdSetPageStatusInternal(TransactionId xid, int nsubxids,
TransactionId *subxids, XidStatus status,
XLogRecPtr lsn, int pageno)
{
- int slotno;
+ Buffer buffer;
int i;
Assert(status == TRANSACTION_STATUS_COMMITTED ||
status == TRANSACTION_STATUS_ABORTED ||
(status == TRANSACTION_STATUS_SUB_COMMITTED && !TransactionIdIsValid(xid)));
- Assert(LWLockHeldByMeInMode(XactSLRULock, LW_EXCLUSIVE));
/*
* If we're doing an async commit (ie, lsn is valid), then we must wait
@@ -356,7 +350,8 @@ TransactionIdSetPageStatusInternal(TransactionId xid, int nsubxids,
* write-busy, since we don't care if the update reaches disk sooner than
* we think.
*/
- slotno = SimpleLruReadPage(XactCtl, pageno, XLogRecPtrIsInvalid(lsn), xid);
+ buffer = ReadSlruBuffer(SLRU_CLOG_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
/*
* Set the main transaction id, if any.
@@ -374,25 +369,26 @@ TransactionIdSetPageStatusInternal(TransactionId xid, int nsubxids,
{
for (i = 0; i < nsubxids; i++)
{
- Assert(XactCtl->shared->page_number[slotno] == TransactionIdToPage(subxids[i]));
+ Assert(pageno == TransactionIdToPage(subxids[i]));
TransactionIdSetStatusBit(subxids[i],
TRANSACTION_STATUS_SUB_COMMITTED,
- lsn, slotno);
+ lsn, buffer);
}
}
/* ... then the main transaction */
- TransactionIdSetStatusBit(xid, status, lsn, slotno);
+ TransactionIdSetStatusBit(xid, status, lsn, buffer);
}
/* Set the subtransactions */
for (i = 0; i < nsubxids; i++)
{
- Assert(XactCtl->shared->page_number[slotno] == TransactionIdToPage(subxids[i]));
- TransactionIdSetStatusBit(subxids[i], status, lsn, slotno);
+ Assert(pageno == TransactionIdToPage(subxids[i]));
+ TransactionIdSetStatusBit(subxids[i], status, lsn, buffer);
}
- XactCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -566,7 +562,7 @@ TransactionGroupUpdateXidStatus(TransactionId xid, XidStatus status,
* Must be called with XactSLRULock held
*/
static void
-TransactionIdSetStatusBit(TransactionId xid, XidStatus status, XLogRecPtr lsn, int slotno)
+TransactionIdSetStatusBit(TransactionId xid, XidStatus status, XLogRecPtr lsn, Buffer buffer)
{
int byteno = TransactionIdToByte(xid);
int bshift = TransactionIdToBIndex(xid) * CLOG_BITS_PER_XACT;
@@ -574,7 +570,10 @@ TransactionIdSetStatusBit(TransactionId xid, XidStatus status, XLogRecPtr lsn, i
char byteval;
char curval;
- byteptr = XactCtl->shared->page_buffer[slotno] + byteno;
+ Assert(LWLockHeldByMeInMode(BufferDescriptorGetContentLock(GetBufferDescriptor(buffer - 1)),
+ LW_EXCLUSIVE));
+
+ byteptr = BufferGetPage(buffer) + byteno;
curval = (*byteptr >> bshift) & CLOG_XACT_BITMASK;
/*
@@ -603,7 +602,7 @@ TransactionIdSetStatusBit(TransactionId xid, XidStatus status, XLogRecPtr lsn, i
*byteptr = byteval;
/*
- * Update the group LSN if the transaction completion LSN is higher.
+ * Update the buffer LSN if the transaction completion LSN is higher.
*
* Note: lsn will be invalid when supplied during InRecovery processing,
* so we don't need to do anything special to avoid LSN updates during
@@ -612,10 +611,8 @@ TransactionIdSetStatusBit(TransactionId xid, XidStatus status, XLogRecPtr lsn, i
*/
if (!XLogRecPtrIsInvalid(lsn))
{
- int lsnindex = GetLSNIndex(slotno, xid);
-
- if (XactCtl->shared->group_lsn[lsnindex] < lsn)
- XactCtl->shared->group_lsn[lsnindex] = lsn;
+ if (BufferGetExternalLSN(GetBufferDescriptor(buffer)) < lsn)
+ BufferSetExternalLSN(GetBufferDescriptor(buffer), lsn);
}
}
@@ -640,67 +637,22 @@ TransactionIdGetStatus(TransactionId xid, XLogRecPtr *lsn)
int pageno = TransactionIdToPage(xid);
int byteno = TransactionIdToByte(xid);
int bshift = TransactionIdToBIndex(xid) * CLOG_BITS_PER_XACT;
- int slotno;
- int lsnindex;
char *byteptr;
XidStatus status;
+ Buffer buffer;
- /* lock is acquired by SimpleLruReadPage_ReadOnly */
-
- slotno = SimpleLruReadPage_ReadOnly(XactCtl, pageno, xid);
- byteptr = XactCtl->shared->page_buffer[slotno] + byteno;
+ buffer = ReadSlruBuffer(SLRU_CLOG_ID, pageno);
+ byteptr = BufferGetPage(buffer) + byteno;
status = (*byteptr >> bshift) & CLOG_XACT_BITMASK;
- lsnindex = GetLSNIndex(slotno, xid);
- *lsn = XactCtl->shared->group_lsn[lsnindex];
+ *lsn = BufferGetExternalLSN(GetBufferDescriptor(buffer));
- LWLockRelease(XactSLRULock);
+ ReleaseBuffer(buffer);
return status;
}
-/*
- * Number of shared CLOG buffers.
- *
- * On larger multi-processor systems, it is possible to have many CLOG page
- * requests in flight at one time which could lead to disk access for CLOG
- * page if the required page is not found in memory. Testing revealed that we
- * can get the best performance by having 128 CLOG buffers, more than that it
- * doesn't improve performance.
- *
- * Unconditionally keeping the number of CLOG buffers to 128 did not seem like
- * a good idea, because it would increase the minimum amount of shared memory
- * required to start, which could be a problem for people running very small
- * configurations. The following formula seems to represent a reasonable
- * compromise: people with very low values for shared_buffers will get fewer
- * CLOG buffers as well, and everyone else will get 128.
- */
-Size
-CLOGShmemBuffers(void)
-{
- return Min(128, Max(4, NBuffers / 512));
-}
-
-/*
- * Initialization of shared memory for CLOG
- */
-Size
-CLOGShmemSize(void)
-{
- return SimpleLruShmemSize(CLOGShmemBuffers(), CLOG_LSNS_PER_PAGE);
-}
-
-void
-CLOGShmemInit(void)
-{
- XactCtl->PagePrecedes = CLOGPagePrecedes;
- SimpleLruInit(XactCtl, "Xact", CLOGShmemBuffers(), CLOG_LSNS_PER_PAGE,
- XactSLRULock, "pg_xact", LWTRANCHE_XACT_BUFFER,
- SYNC_HANDLER_CLOG);
- SlruPagePrecedesUnitTests(XactCtl, CLOG_XACTS_PER_PAGE);
-}
-
/*
* This func must be called ONCE on system install. It creates
* the initial CLOG segment. (The CLOG directory is assumed to
@@ -710,18 +662,15 @@ CLOGShmemInit(void)
void
BootStrapCLOG(void)
{
- int slotno;
-
- LWLockAcquire(XactSLRULock, LW_EXCLUSIVE);
+ Buffer buffer;
/* Create and zero the first page of the commit log */
- slotno = ZeroCLOGPage(0, false);
+ buffer = ZeroCLOGPage(0, false);
/* Make sure it's written out */
- SimpleLruWritePage(XactCtl, slotno);
- Assert(!XactCtl->shared->page_dirty[slotno]);
+ FlushOneBuffer(buffer);
- LWLockRelease(XactSLRULock);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -733,17 +682,18 @@ BootStrapCLOG(void)
*
* Control lock must be held at entry, and will be held at exit.
*/
-static int
+static Buffer
ZeroCLOGPage(int pageno, bool writeXlog)
{
- int slotno;
+ Buffer buffer;
- slotno = SimpleLruZeroPage(XactCtl, pageno);
+ buffer = ZeroSlruBuffer(SLRU_CLOG_ID, pageno);
+ MarkBufferDirty(buffer);
if (writeXlog)
WriteZeroPageXlogRec(pageno);
- return slotno;
+ return buffer;
}
/*
@@ -753,17 +703,6 @@ ZeroCLOGPage(int pageno, bool writeXlog)
void
StartupCLOG(void)
{
- TransactionId xid = XidFromFullTransactionId(ShmemVariableCache->nextXid);
- int pageno = TransactionIdToPage(xid);
-
- LWLockAcquire(XactSLRULock, LW_EXCLUSIVE);
-
- /*
- * Initialize our idea of the latest page number.
- */
- XactCtl->shared->latest_page_number = pageno;
-
- LWLockRelease(XactSLRULock);
}
/*
@@ -775,8 +714,6 @@ TrimCLOG(void)
TransactionId xid = XidFromFullTransactionId(ShmemVariableCache->nextXid);
int pageno = TransactionIdToPage(xid);
- LWLockAcquire(XactSLRULock, LW_EXCLUSIVE);
-
/*
* Zero out the remainder of the current clog page. Under normal
* circumstances it should be zeroes already, but it seems at least
@@ -793,40 +730,24 @@ TrimCLOG(void)
{
int byteno = TransactionIdToByte(xid);
int bshift = TransactionIdToBIndex(xid) * CLOG_BITS_PER_XACT;
- int slotno;
char *byteptr;
+ Buffer buffer;
- slotno = SimpleLruReadPage(XactCtl, pageno, false, xid);
- byteptr = XactCtl->shared->page_buffer[slotno] + byteno;
+ buffer = ReadSlruBuffer(SLRU_CLOG_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ byteptr = BufferGetPage(buffer) + byteno;
/* Zero so-far-unused positions in the current byte */
*byteptr &= (1 << bshift) - 1;
/* Zero the rest of the page */
MemSet(byteptr + 1, 0, BLCKSZ - byteno - 1);
- XactCtl->shared->page_dirty[slotno] = true;
- }
+ MarkBufferDirty(buffer);
- LWLockRelease(XactSLRULock);
-}
-
-/*
- * Perform a checkpoint --- either during shutdown, or on-the-fly
- */
-void
-CheckPointCLOG(void)
-{
- /*
- * Write dirty CLOG pages to disk. This may result in sync requests
- * queued for later handling by ProcessSyncRequests(), as part of the
- * checkpoint.
- */
- TRACE_POSTGRESQL_CLOG_CHECKPOINT_START(true);
- SimpleLruWriteAll(XactCtl, true);
- TRACE_POSTGRESQL_CLOG_CHECKPOINT_DONE(true);
+ UnlockReleaseBuffer(buffer);
+ }
}
-
/*
* Make sure that CLOG has room for a newly-allocated XID.
*
@@ -850,12 +771,8 @@ ExtendCLOG(TransactionId newestXact)
pageno = TransactionIdToPage(newestXact);
- LWLockAcquire(XactSLRULock, LW_EXCLUSIVE);
-
/* Zero the page and make an XLOG entry about it */
- ZeroCLOGPage(pageno, true);
-
- LWLockRelease(XactSLRULock);
+ UnlockReleaseBuffer(ZeroCLOGPage(pageno, true));
}
@@ -886,7 +803,8 @@ TruncateCLOG(TransactionId oldestXact, Oid oldestxid_datoid)
cutoffPage = TransactionIdToPage(oldestXact);
/* Check to see if there's any files that could be removed */
- if (!SlruScanDirectory(XactCtl, SlruScanDirCbReportPresence, &cutoffPage))
+ if (!SlruScanDirectory(SLRU_CLOG_ID, CLOGPagePrecedes,
+ SlruScanDirCbReportPresence, &cutoffPage))
return; /* nothing to remove */
/*
@@ -907,7 +825,7 @@ TruncateCLOG(TransactionId oldestXact, Oid oldestxid_datoid)
WriteTruncateXlogRec(cutoffPage, oldestXact, oldestxid_datoid);
/* Now we can remove the old CLOG segment(s) */
- SimpleLruTruncate(XactCtl, cutoffPage);
+ SimpleLruTruncate(SLRU_CLOG_ID, CLOGPagePrecedes, cutoffPage);
}
@@ -992,17 +910,13 @@ clog_redo(XLogReaderState *record)
if (info == CLOG_ZEROPAGE)
{
int pageno;
- int slotno;
+ Buffer buffer;
memcpy(&pageno, XLogRecGetData(record), sizeof(int));
- LWLockAcquire(XactSLRULock, LW_EXCLUSIVE);
-
- slotno = ZeroCLOGPage(pageno, false);
- SimpleLruWritePage(XactCtl, slotno);
- Assert(!XactCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(XactSLRULock);
+ buffer = ZeroCLOGPage(pageno, false);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
else if (info == CLOG_TRUNCATE)
{
@@ -1012,17 +926,8 @@ clog_redo(XLogReaderState *record)
AdvanceOldestClogXid(xlrec.oldestXact);
- SimpleLruTruncate(XactCtl, xlrec.pageno);
+ SimpleLruTruncate(SLRU_CLOG_ID, CLOGPagePrecedes, xlrec.pageno);
}
else
elog(PANIC, "clog_redo: unknown op code %u", info);
}
-
-/*
- * Entrypoint for sync.c to sync clog files.
- */
-int
-clogsyncfiletag(const FileTag *ftag, char *path)
-{
- return SlruSyncFileTag(XactCtl, ftag, path);
-}
diff --git a/src/backend/access/transam/commit_ts.c b/src/backend/access/transam/commit_ts.c
index b897fabc702..69f34624b08 100644
--- a/src/backend/access/transam/commit_ts.c
+++ b/src/backend/access/transam/commit_ts.c
@@ -70,13 +70,6 @@ typedef struct CommitTimestampEntry
#define TransactionIdToCTsEntry(xid) \
((xid) % (TransactionId) COMMIT_TS_XACTS_PER_PAGE)
-/*
- * Link to shared-memory data structures for CommitTs control
- */
-static SlruCtlData CommitTsCtlData;
-
-#define CommitTsCtl (&CommitTsCtlData)
-
/*
* We keep a cache of the last value set in shared memory.
*
@@ -107,7 +100,7 @@ static void SetXidCommitTsInPage(TransactionId xid, int nsubxids,
static void TransactionIdSetCommitTs(TransactionId xid, TimestampTz ts,
RepOriginId nodeid, int slotno);
static void error_commit_ts_disabled(void);
-static int ZeroCommitTsPage(int pageno, bool writeXlog);
+static Buffer ZeroCommitTsPage(int pageno, bool writeXlog);
static bool CommitTsPagePrecedes(int page1, int page2);
static void ActivateCommitTs(void);
static void DeactivateCommitTs(void);
@@ -216,30 +209,27 @@ SetXidCommitTsInPage(TransactionId xid, int nsubxids,
TransactionId *subxids, TimestampTz ts,
RepOriginId nodeid, int pageno)
{
- int slotno;
int i;
+ Buffer buffer;
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
-
- slotno = SimpleLruReadPage(CommitTsCtl, pageno, true, xid);
+ buffer = ReadSlruBuffer(SLRU_COMMIT_TS_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
- TransactionIdSetCommitTs(xid, ts, nodeid, slotno);
+ TransactionIdSetCommitTs(xid, ts, nodeid, buffer);
for (i = 0; i < nsubxids; i++)
- TransactionIdSetCommitTs(subxids[i], ts, nodeid, slotno);
+ TransactionIdSetCommitTs(subxids[i], ts, nodeid, buffer);
- CommitTsCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
- LWLockRelease(CommitTsSLRULock);
+ UnlockReleaseBuffer(buffer);
}
/*
* Sets the commit timestamp of a single transaction.
- *
- * Must be called with CommitTsSLRULock held
*/
static void
TransactionIdSetCommitTs(TransactionId xid, TimestampTz ts,
- RepOriginId nodeid, int slotno)
+ RepOriginId nodeid, Buffer buffer)
{
int entryno = TransactionIdToCTsEntry(xid);
CommitTimestampEntry entry;
@@ -249,8 +239,7 @@ TransactionIdSetCommitTs(TransactionId xid, TimestampTz ts,
entry.time = ts;
entry.nodeid = nodeid;
- memcpy(CommitTsCtl->shared->page_buffer[slotno] +
- SizeOfCommitTimestampEntry * entryno,
+ memcpy(BufferGetPage(buffer) + SizeOfCommitTimestampEntry * entryno,
&entry, SizeOfCommitTimestampEntry);
}
@@ -268,10 +257,10 @@ TransactionIdGetCommitTsData(TransactionId xid, TimestampTz *ts,
{
int pageno = TransactionIdToCTsPage(xid);
int entryno = TransactionIdToCTsEntry(xid);
- int slotno;
CommitTimestampEntry entry;
TransactionId oldestCommitTsXid;
TransactionId newestCommitTsXid;
+ Buffer buffer;
if (!TransactionIdIsValid(xid))
ereport(ERROR,
@@ -325,10 +314,11 @@ TransactionIdGetCommitTsData(TransactionId xid, TimestampTz *ts,
return false;
}
- /* lock is acquired by SimpleLruReadPage_ReadOnly */
- slotno = SimpleLruReadPage_ReadOnly(CommitTsCtl, pageno, xid);
+ buffer = ReadSlruBuffer(SLRU_COMMIT_TS_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_SHARE);
+
memcpy(&entry,
- CommitTsCtl->shared->page_buffer[slotno] +
+ BufferGetPage(buffer) +
SizeOfCommitTimestampEntry * entryno,
SizeOfCommitTimestampEntry);
@@ -336,7 +326,7 @@ TransactionIdGetCommitTsData(TransactionId xid, TimestampTz *ts,
if (nodeid)
*nodeid = entry.nodeid;
- LWLockRelease(CommitTsSLRULock);
+ UnlockReleaseBuffer(buffer);
return *ts != 0;
}
@@ -487,27 +477,13 @@ pg_xact_commit_timestamp_origin(PG_FUNCTION_ARGS)
PG_RETURN_DATUM(HeapTupleGetDatum(htup));
}
-/*
- * Number of shared CommitTS buffers.
- *
- * We use a very similar logic as for the number of CLOG buffers (except we
- * scale up twice as fast with shared buffers, and the maximum is twice as
- * high); see comments in CLOGShmemBuffers.
- */
-Size
-CommitTsShmemBuffers(void)
-{
- return Min(256, Max(4, NBuffers / 256));
-}
-
/*
* Shared memory sizing for CommitTs
*/
Size
CommitTsShmemSize(void)
{
- return SimpleLruShmemSize(CommitTsShmemBuffers(), 0) +
- sizeof(CommitTimestampShared);
+ return sizeof(CommitTimestampShared);
}
/*
@@ -519,12 +495,7 @@ CommitTsShmemInit(void)
{
bool found;
- CommitTsCtl->PagePrecedes = CommitTsPagePrecedes;
- SimpleLruInit(CommitTsCtl, "CommitTs", CommitTsShmemBuffers(), 0,
- CommitTsSLRULock, "pg_commit_ts",
- LWTRANCHE_COMMITTS_BUFFER,
- SYNC_HANDLER_COMMIT_TS);
- SlruPagePrecedesUnitTests(CommitTsCtl, COMMIT_TS_XACTS_PER_PAGE);
+ SlruPagePrecedesUnitTests(CommitTsPagePrecedes, COMMIT_TS_XACTS_PER_PAGE);
commitTsShared = ShmemInitStruct("CommitTs shared",
sizeof(CommitTimestampShared),
@@ -568,17 +539,18 @@ BootStrapCommitTs(void)
*
* Control lock must be held at entry, and will be held at exit.
*/
-static int
+static Buffer
ZeroCommitTsPage(int pageno, bool writeXlog)
{
- int slotno;
+ Buffer buffer;
- slotno = SimpleLruZeroPage(CommitTsCtl, pageno);
+ buffer = ZeroSlruBuffer(SLRU_COMMIT_TS_ID, pageno);
+ MarkBufferDirty(buffer);
if (writeXlog)
WriteZeroPageXlogRec(pageno);
- return slotno;
+ return buffer;
}
/*
@@ -676,13 +648,6 @@ ActivateCommitTs(void)
xid = XidFromFullTransactionId(ShmemVariableCache->nextXid);
pageno = TransactionIdToCTsPage(xid);
- /*
- * Re-Initialize our idea of the latest page number.
- */
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
- CommitTsCtl->shared->latest_page_number = pageno;
- LWLockRelease(CommitTsSLRULock);
-
/*
* If CommitTs is enabled, but it wasn't in the previous server run, we
* need to set the oldest and newest values to the next Xid; that way, we
@@ -705,15 +670,14 @@ ActivateCommitTs(void)
LWLockRelease(CommitTsLock);
/* Create the current segment file, if necessary */
- if (!SimpleLruDoesPhysicalPageExist(CommitTsCtl, pageno))
+ if (!SimpleLruDoesPhysicalPageExist(SLRU_COMMIT_TS_ID, pageno))
{
- int slotno;
+ Buffer buffer;
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
- slotno = ZeroCommitTsPage(pageno, false);
- SimpleLruWritePage(CommitTsCtl, slotno);
- Assert(!CommitTsCtl->shared->page_dirty[slotno]);
- LWLockRelease(CommitTsSLRULock);
+ buffer = ZeroSlruBuffer(SLRU_COMMIT_TS_ID, pageno);
+ MarkBufferDirty(buffer);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
/* Change the activation status in shared memory. */
@@ -762,23 +726,9 @@ DeactivateCommitTs(void)
* be overwritten anyway when we wrap around, but it seems better to be
* tidy.)
*/
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
- (void) SlruScanDirectory(CommitTsCtl, SlruScanDirCbDeleteAll, NULL);
- LWLockRelease(CommitTsSLRULock);
-}
-
-/*
- * Perform a checkpoint --- either during shutdown, or on-the-fly
- */
-void
-CheckPointCommitTs(void)
-{
- /*
- * Write dirty CommitTs pages to disk. This may result in sync requests
- * queued for later handling by ProcessSyncRequests(), as part of the
- * checkpoint.
- */
- SimpleLruWriteAll(CommitTsCtl, true);
+ (void) SlruScanDirectory(SLRU_COMMIT_TS_ID,
+ CommitTsPagePrecedes,
+ SlruScanDirCbDeleteAll, NULL);
}
/*
@@ -816,12 +766,8 @@ ExtendCommitTs(TransactionId newestXact)
pageno = TransactionIdToCTsPage(newestXact);
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
-
/* Zero the page and make an XLOG entry about it */
- ZeroCommitTsPage(pageno, !InRecovery);
-
- LWLockRelease(CommitTsSLRULock);
+ UnlockReleaseBuffer(ZeroCommitTsPage(pageno, !InRecovery));
}
/*
@@ -842,7 +788,9 @@ TruncateCommitTs(TransactionId oldestXact)
cutoffPage = TransactionIdToCTsPage(oldestXact);
/* Check to see if there's any files that could be removed */
- if (!SlruScanDirectory(CommitTsCtl, SlruScanDirCbReportPresence,
+ if (!SlruScanDirectory(SLRU_COMMIT_TS_ID,
+ CommitTsPagePrecedes,
+ SlruScanDirCbReportPresence,
&cutoffPage))
return; /* nothing to remove */
@@ -850,7 +798,7 @@ TruncateCommitTs(TransactionId oldestXact)
WriteTruncateXlogRec(cutoffPage, oldestXact);
/* Now we can remove the old CommitTs segment(s) */
- SimpleLruTruncate(CommitTsCtl, cutoffPage);
+ SimpleLruTruncate(SLRU_COMMIT_TS_ID, CommitTsPagePrecedes, cutoffPage);
}
/*
@@ -974,17 +922,14 @@ commit_ts_redo(XLogReaderState *record)
if (info == COMMIT_TS_ZEROPAGE)
{
int pageno;
- int slotno;
+ Buffer buffer;
memcpy(&pageno, XLogRecGetData(record), sizeof(int));
- LWLockAcquire(CommitTsSLRULock, LW_EXCLUSIVE);
-
- slotno = ZeroCommitTsPage(pageno, false);
- SimpleLruWritePage(CommitTsCtl, slotno);
- Assert(!CommitTsCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(CommitTsSLRULock);
+ buffer = ZeroSlruBuffer(SLRU_COMMIT_TS_ID, pageno);
+ MarkBufferDirty(buffer);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
else if (info == COMMIT_TS_TRUNCATE)
{
@@ -992,23 +937,8 @@ commit_ts_redo(XLogReaderState *record)
AdvanceOldestCommitTsXid(trunc->oldestXid);
- /*
- * During XLOG replay, latest_page_number isn't set up yet; insert a
- * suitable value to bypass the sanity test in SimpleLruTruncate.
- */
- CommitTsCtl->shared->latest_page_number = trunc->pageno;
-
- SimpleLruTruncate(CommitTsCtl, trunc->pageno);
+ SimpleLruTruncate(SLRU_COMMIT_TS_ID, CommitTsPagePrecedes, trunc->pageno);
}
else
elog(PANIC, "commit_ts_redo: unknown op code %u", info);
}
-
-/*
- * Entrypoint for sync.c to sync commit_ts files.
- */
-int
-committssyncfiletag(const FileTag *ftag, char *path)
-{
- return SlruSyncFileTag(CommitTsCtl, ftag, path);
-}
diff --git a/src/backend/access/transam/multixact.c b/src/backend/access/transam/multixact.c
index fe6698d5ffa..8e06a0e9a91 100644
--- a/src/backend/access/transam/multixact.c
+++ b/src/backend/access/transam/multixact.c
@@ -181,15 +181,6 @@
#define PreviousMultiXactId(xid) \
((xid) == FirstMultiXactId ? MaxMultiXactId : (xid) - 1)
-/*
- * Links to shared-memory data structures for MultiXact control
- */
-static SlruCtlData MultiXactOffsetCtlData;
-static SlruCtlData MultiXactMemberCtlData;
-
-#define MultiXactOffsetCtl (&MultiXactOffsetCtlData)
-#define MultiXactMemberCtl (&MultiXactMemberCtlData)
-
/*
* MultiXact state shared across all backends. All this state is protected
* by MultiXactGenLock. (We also use MultiXactOffsetSLRULock and
@@ -354,10 +345,9 @@ static void mXactCachePut(MultiXactId multi, int nmembers,
static char *mxstatus_to_string(MultiXactStatus status);
/* management of SLRU infrastructure */
-static int ZeroMultiXactOffsetPage(int pageno, bool writeXlog);
-static int ZeroMultiXactMemberPage(int pageno, bool writeXlog);
+static Buffer ZeroMultiXactOffsetPage(int pageno, bool writeXlog);
+static Buffer ZeroMultiXactMemberPage(int pageno, bool writeXlog);
static bool MultiXactOffsetPagePrecedes(int page1, int page2);
-static bool MultiXactMemberPagePrecedes(int page1, int page2);
static bool MultiXactOffsetPrecedes(MultiXactOffset offset1,
MultiXactOffset offset2);
static void ExtendMultiXactOffset(MultiXactId multi);
@@ -867,34 +857,25 @@ RecordNewMultiXact(MultiXactId multi, MultiXactOffset offset,
int pageno;
int prev_pageno;
int entryno;
- int slotno;
MultiXactOffset *offptr;
int i;
-
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
+ Buffer buffer;
pageno = MultiXactIdToOffsetPage(multi);
entryno = MultiXactIdToOffsetEntry(multi);
- /*
- * Note: we pass the MultiXactId to SimpleLruReadPage as the "transaction"
- * to complain about if there's any I/O error. This is kinda bogus, but
- * since the errors will always give the full pathname, it should be clear
- * enough that a MultiXactId is really involved. Perhaps someday we'll
- * take the trouble to generalize the slru.c error reporting code.
- */
- slotno = SimpleLruReadPage(MultiXactOffsetCtl, pageno, true, multi);
- offptr = (MultiXactOffset *) MultiXactOffsetCtl->shared->page_buffer[slotno];
+ /* XXX set up error context? */
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_OFFSET_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ offptr = (MultiXactOffset *) BufferGetPage(buffer);
offptr += entryno;
*offptr = offset;
- MultiXactOffsetCtl->shared->page_dirty[slotno] = true;
-
- /* Exchange our lock */
- LWLockRelease(MultiXactOffsetSLRULock);
+ MarkBufferDirty(buffer);
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
+ UnlockReleaseBuffer(buffer);
+ buffer = InvalidBuffer;
prev_pageno = -1;
@@ -916,27 +897,28 @@ RecordNewMultiXact(MultiXactId multi, MultiXactOffset offset,
if (pageno != prev_pageno)
{
- slotno = SimpleLruReadPage(MultiXactMemberCtl, pageno, true, multi);
+ if (BufferIsValid(buffer))
+ UnlockReleaseBuffer(buffer);
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_MEMBER_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
prev_pageno = pageno;
}
- memberptr = (TransactionId *)
- (MultiXactMemberCtl->shared->page_buffer[slotno] + memberoff);
+ memberptr = (TransactionId *) (BufferGetPage(buffer) + memberoff);
*memberptr = members[i].xid;
- flagsptr = (uint32 *)
- (MultiXactMemberCtl->shared->page_buffer[slotno] + flagsoff);
+ flagsptr = (uint32 *) (BufferGetPage(buffer) + flagsoff);
flagsval = *flagsptr;
flagsval &= ~(((1 << MXACT_MEMBER_BITS_PER_XACT) - 1) << bshift);
flagsval |= (members[i].status << bshift);
*flagsptr = flagsval;
- MultiXactMemberCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
}
- LWLockRelease(MultiXactMemberSLRULock);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -1228,7 +1210,6 @@ GetMultiXactIdMembers(MultiXactId multi, MultiXactMember **members,
int pageno;
int prev_pageno;
int entryno;
- int slotno;
MultiXactOffset *offptr;
MultiXactOffset offset;
int length;
@@ -1239,6 +1220,7 @@ GetMultiXactIdMembers(MultiXactId multi, MultiXactMember **members,
MultiXactId tmpMXact;
MultiXactOffset nextOffset;
MultiXactMember *ptr;
+ Buffer buffer;
debug_elog3(DEBUG2, "GetMembers: asked for %u", multi);
@@ -1342,13 +1324,12 @@ GetMultiXactIdMembers(MultiXactId multi, MultiXactMember **members,
* time on every multixact creation.
*/
retry:
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
-
pageno = MultiXactIdToOffsetPage(multi);
entryno = MultiXactIdToOffsetEntry(multi);
- slotno = SimpleLruReadPage(MultiXactOffsetCtl, pageno, true, multi);
- offptr = (MultiXactOffset *) MultiXactOffsetCtl->shared->page_buffer[slotno];
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_OFFSET_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_SHARE);
+ offptr = (MultiXactOffset *) BufferGetPage(buffer);
offptr += entryno;
offset = *offptr;
@@ -1379,16 +1360,20 @@ retry:
entryno = MultiXactIdToOffsetEntry(tmpMXact);
if (pageno != prev_pageno)
- slotno = SimpleLruReadPage(MultiXactOffsetCtl, pageno, true, tmpMXact);
+ {
+ UnlockReleaseBuffer(buffer);
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_OFFSET_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_SHARE);
+ }
- offptr = (MultiXactOffset *) MultiXactOffsetCtl->shared->page_buffer[slotno];
+ offptr = (MultiXactOffset *) BufferGetPage(buffer);
offptr += entryno;
nextMXOffset = *offptr;
if (nextMXOffset == 0)
{
/* Corner case 2: next multixact is still being filled in */
- LWLockRelease(MultiXactOffsetSLRULock);
+ UnlockReleaseBuffer(buffer);
CHECK_FOR_INTERRUPTS();
pg_usleep(1000L);
goto retry;
@@ -1396,14 +1381,11 @@ retry:
length = nextMXOffset - offset;
}
-
- LWLockRelease(MultiXactOffsetSLRULock);
+ UnlockReleaseBuffer(buffer);
+ buffer = InvalidBuffer;
ptr = (MultiXactMember *) palloc(length * sizeof(MultiXactMember));
- /* Now get the members themselves. */
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
-
truelength = 0;
prev_pageno = -1;
for (i = 0; i < length; i++, offset++)
@@ -1419,12 +1401,14 @@ retry:
if (pageno != prev_pageno)
{
- slotno = SimpleLruReadPage(MultiXactMemberCtl, pageno, true, multi);
+ if (BufferIsValid(buffer))
+ UnlockReleaseBuffer(buffer);
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_MEMBER_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_SHARE);
prev_pageno = pageno;
}
- xactptr = (TransactionId *)
- (MultiXactMemberCtl->shared->page_buffer[slotno] + memberoff);
+ xactptr = (TransactionId *) (BufferGetPage(buffer) + memberoff);
if (!TransactionIdIsValid(*xactptr))
{
@@ -1435,14 +1419,13 @@ retry:
flagsoff = MXOffsetToFlagsOffset(offset);
bshift = MXOffsetToFlagsBitShift(offset);
- flagsptr = (uint32 *) (MultiXactMemberCtl->shared->page_buffer[slotno] + flagsoff);
+ flagsptr = (uint32 *) (BufferGetPage(buffer) + flagsoff);
ptr[truelength].xid = *xactptr;
ptr[truelength].status = (*flagsptr >> bshift) & MXACT_MEMBER_XACT_BITMASK;
truelength++;
}
-
- LWLockRelease(MultiXactMemberSLRULock);
+ UnlockReleaseBuffer(buffer);
/* A multixid with zero members should not happen */
Assert(truelength > 0);
@@ -1834,8 +1817,6 @@ MultiXactShmemSize(void)
mul_size(sizeof(MultiXactId) * 2, MaxOldestSlot))
size = SHARED_MULTIXACT_STATE_SIZE;
- size = add_size(size, SimpleLruShmemSize(NUM_MULTIXACTOFFSET_BUFFERS, 0));
- size = add_size(size, SimpleLruShmemSize(NUM_MULTIXACTMEMBER_BUFFERS, 0));
return size;
}
@@ -1847,22 +1828,6 @@ MultiXactShmemInit(void)
debug_elog2(DEBUG2, "Shared Memory Init for MultiXact");
- MultiXactOffsetCtl->PagePrecedes = MultiXactOffsetPagePrecedes;
- MultiXactMemberCtl->PagePrecedes = MultiXactMemberPagePrecedes;
-
- SimpleLruInit(MultiXactOffsetCtl,
- "MultiXactOffset", NUM_MULTIXACTOFFSET_BUFFERS, 0,
- MultiXactOffsetSLRULock, "pg_multixact/offsets",
- LWTRANCHE_MULTIXACTOFFSET_BUFFER,
- SYNC_HANDLER_MULTIXACT_OFFSET);
- SlruPagePrecedesUnitTests(MultiXactOffsetCtl, MULTIXACT_OFFSETS_PER_PAGE);
- SimpleLruInit(MultiXactMemberCtl,
- "MultiXactMember", NUM_MULTIXACTMEMBER_BUFFERS, 0,
- MultiXactMemberSLRULock, "pg_multixact/members",
- LWTRANCHE_MULTIXACTMEMBER_BUFFER,
- SYNC_HANDLER_MULTIXACT_MEMBER);
- /* doesn't call SimpleLruTruncate() or meet criteria for unit tests */
-
/* Initialize our shared state struct */
MultiXactState = ShmemInitStruct("Shared MultiXact State",
SHARED_MULTIXACT_STATE_SIZE,
@@ -1893,29 +1858,17 @@ MultiXactShmemInit(void)
void
BootStrapMultiXact(void)
{
- int slotno;
-
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
+ Buffer buffer;
/* Create and zero the first page of the offsets log */
- slotno = ZeroMultiXactOffsetPage(0, false);
-
- /* Make sure it's written out */
- SimpleLruWritePage(MultiXactOffsetCtl, slotno);
- Assert(!MultiXactOffsetCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(MultiXactOffsetSLRULock);
-
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
+ buffer = ZeroMultiXactOffsetPage(0, false);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
/* Create and zero the first page of the members log */
- slotno = ZeroMultiXactMemberPage(0, false);
-
- /* Make sure it's written out */
- SimpleLruWritePage(MultiXactMemberCtl, slotno);
- Assert(!MultiXactMemberCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(MultiXactMemberSLRULock);
+ buffer = ZeroMultiXactMemberPage(0, false);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -1927,33 +1880,35 @@ BootStrapMultiXact(void)
*
* Control lock must be held at entry, and will be held at exit.
*/
-static int
+static Buffer
ZeroMultiXactOffsetPage(int pageno, bool writeXlog)
{
- int slotno;
+ Buffer buffer;
- slotno = SimpleLruZeroPage(MultiXactOffsetCtl, pageno);
+ buffer = ZeroSlruBuffer(SLRU_MULTIXACT_OFFSET_ID, pageno);
+ MarkBufferDirty(buffer);
if (writeXlog)
WriteMZeroPageXlogRec(pageno, XLOG_MULTIXACT_ZERO_OFF_PAGE);
- return slotno;
+ return buffer;
}
/*
* Ditto, for MultiXactMember
*/
-static int
+static Buffer
ZeroMultiXactMemberPage(int pageno, bool writeXlog)
{
- int slotno;
+ Buffer buffer;
- slotno = SimpleLruZeroPage(MultiXactMemberCtl, pageno);
+ buffer = ZeroSlruBuffer(SLRU_MULTIXACT_MEMBER_ID, pageno);
+ MarkBufferDirty(buffer);
if (writeXlog)
WriteMZeroPageXlogRec(pageno, XLOG_MULTIXACT_ZERO_MEM_PAGE);
- return slotno;
+ return buffer;
}
/*
@@ -1978,22 +1933,14 @@ MaybeExtendOffsetSlru(void)
pageno = MultiXactIdToOffsetPage(MultiXactState->nextMXact);
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
-
- if (!SimpleLruDoesPhysicalPageExist(MultiXactOffsetCtl, pageno))
+ if (!SimpleLruDoesPhysicalPageExist(SLRU_MULTIXACT_OFFSET_ID, pageno))
{
- int slotno;
+ Buffer buffer;
- /*
- * Fortunately for us, SimpleLruWritePage is already prepared to deal
- * with creating a new segment file even if the page we're writing is
- * not the first in it, so this is enough.
- */
- slotno = ZeroMultiXactOffsetPage(pageno, false);
- SimpleLruWritePage(MultiXactOffsetCtl, slotno);
+ buffer = ZeroMultiXactOffsetPage(pageno, false);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
-
- LWLockRelease(MultiXactOffsetSLRULock);
}
/*
@@ -2007,21 +1954,6 @@ MaybeExtendOffsetSlru(void)
void
StartupMultiXact(void)
{
- MultiXactId multi = MultiXactState->nextMXact;
- MultiXactOffset offset = MultiXactState->nextOffset;
- int pageno;
-
- /*
- * Initialize offset's idea of the latest page number.
- */
- pageno = MultiXactIdToOffsetPage(multi);
- MultiXactOffsetCtl->shared->latest_page_number = pageno;
-
- /*
- * Initialize member's idea of the latest page number.
- */
- pageno = MXOffsetToMemberPage(offset);
- MultiXactMemberCtl->shared->latest_page_number = pageno;
}
/*
@@ -2045,14 +1977,7 @@ TrimMultiXact(void)
oldestMXactDB = MultiXactState->oldestMultiXactDB;
LWLockRelease(MultiXactGenLock);
- /* Clean up offsets state */
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
-
- /*
- * (Re-)Initialize our idea of the latest page number for offsets.
- */
- pageno = MultiXactIdToOffsetPage(nextMXact);
- MultiXactOffsetCtl->shared->latest_page_number = pageno;
+ pageno = MXOffsetToMemberPage(offset);
/*
* Zero out the remainder of the current offsets page. See notes in
@@ -2065,29 +1990,20 @@ TrimMultiXact(void)
entryno = MultiXactIdToOffsetEntry(nextMXact);
if (entryno != 0)
{
- int slotno;
MultiXactOffset *offptr;
+ Buffer buffer;
- slotno = SimpleLruReadPage(MultiXactOffsetCtl, pageno, true, nextMXact);
- offptr = (MultiXactOffset *) MultiXactOffsetCtl->shared->page_buffer[slotno];
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_OFFSET_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ offptr = (MultiXactOffset *) BufferGetPage(buffer);
offptr += entryno;
MemSet(offptr, 0, BLCKSZ - (entryno * sizeof(MultiXactOffset)));
- MultiXactOffsetCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
+ UnlockReleaseBuffer(buffer);
}
- LWLockRelease(MultiXactOffsetSLRULock);
-
- /* And the same for members */
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
-
- /*
- * (Re-)Initialize our idea of the latest page number for members.
- */
- pageno = MXOffsetToMemberPage(offset);
- MultiXactMemberCtl->shared->latest_page_number = pageno;
-
/*
* Zero out the remainder of the current members page. See notes in
* TrimCLOG() for motivation.
@@ -2095,14 +2011,14 @@ TrimMultiXact(void)
flagsoff = MXOffsetToFlagsOffset(offset);
if (flagsoff != 0)
{
- int slotno;
TransactionId *xidptr;
int memberoff;
+ Buffer buffer;
memberoff = MXOffsetToMemberOffset(offset);
- slotno = SimpleLruReadPage(MultiXactMemberCtl, pageno, true, offset);
- xidptr = (TransactionId *)
- (MultiXactMemberCtl->shared->page_buffer[slotno] + memberoff);
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_MEMBER_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ xidptr = (TransactionId *) (BufferGetPage(buffer) + memberoff);
MemSet(xidptr, 0, BLCKSZ - memberoff);
@@ -2112,11 +2028,10 @@ TrimMultiXact(void)
* writing.
*/
- MultiXactMemberCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
+ UnlockReleaseBuffer(buffer);
}
- LWLockRelease(MultiXactMemberSLRULock);
-
/* signal that we're officially up */
LWLockAcquire(MultiXactGenLock, LW_EXCLUSIVE);
MultiXactState->finishedStartup = true;
@@ -2148,25 +2063,6 @@ MultiXactGetCheckptMulti(bool is_shutdown,
*nextMulti, *nextMultiOffset, *oldestMulti, *oldestMultiDB);
}
-/*
- * Perform a checkpoint --- either during shutdown, or on-the-fly
- */
-void
-CheckPointMultiXact(void)
-{
- TRACE_POSTGRESQL_MULTIXACT_CHECKPOINT_START(true);
-
- /*
- * Write dirty MultiXact pages to disk. This may result in sync requests
- * queued for later handling by ProcessSyncRequests(), as part of the
- * checkpoint.
- */
- SimpleLruWriteAll(MultiXactOffsetCtl, true);
- SimpleLruWriteAll(MultiXactMemberCtl, true);
-
- TRACE_POSTGRESQL_MULTIXACT_CHECKPOINT_DONE(true);
-}
-
/*
* Set the next-to-be-assigned MultiXactId and offset
*
@@ -2415,12 +2311,8 @@ ExtendMultiXactOffset(MultiXactId multi)
pageno = MultiXactIdToOffsetPage(multi);
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
-
/* Zero the page and make an XLOG entry about it */
- ZeroMultiXactOffsetPage(pageno, true);
-
- LWLockRelease(MultiXactOffsetSLRULock);
+ UnlockReleaseBuffer(ZeroMultiXactOffsetPage(pageno, true));
}
/*
@@ -2456,12 +2348,8 @@ ExtendMultiXactMember(MultiXactOffset offset, int nmembers)
pageno = MXOffsetToMemberPage(offset);
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
-
/* Zero the page and make an XLOG entry about it */
- ZeroMultiXactMemberPage(pageno, true);
-
- LWLockRelease(MultiXactMemberSLRULock);
+ UnlockReleaseBuffer(ZeroMultiXactMemberPage(pageno, true));
}
/*
@@ -2737,8 +2625,8 @@ find_multixact_start(MultiXactId multi, MultiXactOffset *result)
MultiXactOffset offset;
int pageno;
int entryno;
- int slotno;
MultiXactOffset *offptr;
+ Buffer buffer;
Assert(MultiXactState->finishedStartup);
@@ -2746,20 +2634,19 @@ find_multixact_start(MultiXactId multi, MultiXactOffset *result)
entryno = MultiXactIdToOffsetEntry(multi);
/*
- * Write out dirty data, so PhysicalPageExists can work correctly.
+ * Cope with missing/bogus oldest MultiXact in inconsistent states (see
+ * commit 068cfadf9).
*/
- SimpleLruWriteAll(MultiXactOffsetCtl, true);
- SimpleLruWriteAll(MultiXactMemberCtl, true);
-
- if (!SimpleLruDoesPhysicalPageExist(MultiXactOffsetCtl, pageno))
+ if (!ProbeSlruBuffer(SLRU_MULTIXACT_OFFSET_ID, pageno) &&
+ !SimpleLruDoesPhysicalPageExist(SLRU_MULTIXACT_OFFSET_ID, pageno))
return false;
- /* lock is acquired by SimpleLruReadPage_ReadOnly */
- slotno = SimpleLruReadPage_ReadOnly(MultiXactOffsetCtl, pageno, multi);
- offptr = (MultiXactOffset *) MultiXactOffsetCtl->shared->page_buffer[slotno];
+ buffer = ReadSlruBuffer(SLRU_MULTIXACT_OFFSET_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_SHARE);
+ offptr = (MultiXactOffset *) BufferGetPage(buffer);
offptr += entryno;
offset = *offptr;
- LWLockRelease(MultiXactOffsetSLRULock);
+ UnlockReleaseBuffer(buffer);
*result = offset;
return true;
@@ -2862,12 +2749,13 @@ typedef struct mxtruncinfo
* This callback determines the earliest existing page number.
*/
static bool
-SlruScanDirCbFindEarliest(SlruCtl ctl, char *filename, int segpage, void *data)
+SlruScanDirCbFindEarliest(int slru_id, SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage, void *data)
{
mxtruncinfo *trunc = (mxtruncinfo *) data;
if (trunc->earliestExistingPage == -1 ||
- ctl->PagePrecedes(segpage, trunc->earliestExistingPage))
+ PagePrecedes(segpage, trunc->earliestExistingPage))
{
trunc->earliestExistingPage = segpage;
}
@@ -2899,7 +2787,7 @@ PerformMembersTruncation(MultiXactOffset oldestOffset, MultiXactOffset newOldest
while (segment != endsegment)
{
elog(DEBUG2, "truncating multixact members segment %x", segment);
- SlruDeleteSegment(MultiXactMemberCtl, segment);
+ SlruDeleteSegment(SLRU_MULTIXACT_MEMBER_ID, segment);
/* move to next segment, handling wraparound correctly */
if (segment == maxsegment)
@@ -2922,7 +2810,8 @@ PerformOffsetsTruncation(MultiXactId oldestMulti, MultiXactId newOldestMulti)
* didn't subtract one, we'd trigger SimpleLruTruncate's wraparound
* detection.
*/
- SimpleLruTruncate(MultiXactOffsetCtl,
+ SimpleLruTruncate(SLRU_MULTIXACT_OFFSET_ID,
+ MultiXactOffsetPagePrecedes,
MultiXactIdToOffsetPage(PreviousMultiXactId(newOldestMulti)));
}
@@ -2996,7 +2885,9 @@ TruncateMultiXact(MultiXactId newOldestMulti, Oid newOldestMultiDB)
* been truncated away, and we crashed before updating oldestMulti.
*/
trunc.earliestExistingPage = -1;
- SlruScanDirectory(MultiXactOffsetCtl, SlruScanDirCbFindEarliest, &trunc);
+ SlruScanDirectory(SLRU_MULTIXACT_OFFSET_ID,
+ MultiXactOffsetPagePrecedes,
+ SlruScanDirCbFindEarliest, &trunc);
earliest = trunc.earliestExistingPage * MULTIXACT_OFFSETS_PER_PAGE;
if (earliest < FirstMultiXactId)
earliest = FirstMultiXactId;
@@ -3128,24 +3019,6 @@ MultiXactOffsetPagePrecedes(int page1, int page2)
multi2 + MULTIXACT_OFFSETS_PER_PAGE - 1));
}
-/*
- * Decide whether a MultiXactMember page number is "older" for truncation
- * purposes. There is no "invalid offset number" so use the numbers verbatim.
- */
-static bool
-MultiXactMemberPagePrecedes(int page1, int page2)
-{
- MultiXactOffset offset1;
- MultiXactOffset offset2;
-
- offset1 = ((MultiXactOffset) page1) * MULTIXACT_MEMBERS_PER_PAGE;
- offset2 = ((MultiXactOffset) page2) * MULTIXACT_MEMBERS_PER_PAGE;
-
- return (MultiXactOffsetPrecedes(offset1, offset2) &&
- MultiXactOffsetPrecedes(offset1,
- offset2 + MULTIXACT_MEMBERS_PER_PAGE - 1));
-}
-
/*
* Decide which of two MultiXactIds is earlier.
*
@@ -3240,32 +3113,18 @@ multixact_redo(XLogReaderState *record)
if (info == XLOG_MULTIXACT_ZERO_OFF_PAGE)
{
int pageno;
- int slotno;
memcpy(&pageno, XLogRecGetData(record), sizeof(int));
- LWLockAcquire(MultiXactOffsetSLRULock, LW_EXCLUSIVE);
-
- slotno = ZeroMultiXactOffsetPage(pageno, false);
- SimpleLruWritePage(MultiXactOffsetCtl, slotno);
- Assert(!MultiXactOffsetCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(MultiXactOffsetSLRULock);
+ UnlockReleaseBuffer(ZeroMultiXactOffsetPage(pageno, false));
}
else if (info == XLOG_MULTIXACT_ZERO_MEM_PAGE)
{
int pageno;
- int slotno;
memcpy(&pageno, XLogRecGetData(record), sizeof(int));
- LWLockAcquire(MultiXactMemberSLRULock, LW_EXCLUSIVE);
-
- slotno = ZeroMultiXactMemberPage(pageno, false);
- SimpleLruWritePage(MultiXactMemberCtl, slotno);
- Assert(!MultiXactMemberCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(MultiXactMemberSLRULock);
+ UnlockReleaseBuffer(ZeroMultiXactMemberPage(pageno, false));
}
else if (info == XLOG_MULTIXACT_CREATE_ID)
{
@@ -3299,7 +3158,6 @@ multixact_redo(XLogReaderState *record)
else if (info == XLOG_MULTIXACT_TRUNCATE_ID)
{
xl_multixact_truncate xlrec;
- int pageno;
memcpy(&xlrec, XLogRecGetData(record),
SizeOfMultiXactTruncate);
@@ -3325,13 +3183,6 @@ multixact_redo(XLogReaderState *record)
PerformMembersTruncation(xlrec.startTruncMemb, xlrec.endTruncMemb);
- /*
- * During XLOG replay, latest_page_number isn't necessarily set up
- * yet; insert a suitable value to bypass the sanity test in
- * SimpleLruTruncate.
- */
- pageno = MultiXactIdToOffsetPage(xlrec.endTruncOff);
- MultiXactOffsetCtl->shared->latest_page_number = pageno;
PerformOffsetsTruncation(xlrec.startTruncOff, xlrec.endTruncOff);
LWLockRelease(MultiXactTruncationLock);
@@ -3401,21 +3252,3 @@ pg_get_multixact_members(PG_FUNCTION_ARGS)
SRF_RETURN_DONE(funccxt);
}
-
-/*
- * Entrypoint for sync.c to sync offsets files.
- */
-int
-multixactoffsetssyncfiletag(const FileTag *ftag, char *path)
-{
- return SlruSyncFileTag(MultiXactOffsetCtl, ftag, path);
-}
-
-/*
- * Entrypoint for sync.c to sync members files.
- */
-int
-multixactmemberssyncfiletag(const FileTag *ftag, char *path)
-{
- return SlruSyncFileTag(MultiXactMemberCtl, ftag, path);
-}
diff --git a/src/backend/access/transam/slru.c b/src/backend/access/transam/slru.c
index 5ab86238a92..1204468c039 100644
--- a/src/backend/access/transam/slru.c
+++ b/src/backend/access/transam/slru.c
@@ -1,41 +1,9 @@
/*-------------------------------------------------------------------------
*
* slru.c
- * Simple LRU buffering for transaction status logfiles
+ * Simple buffering for transaction status logfiles
*
- * We use a simple least-recently-used scheme to manage a pool of page
- * buffers. Under ordinary circumstances we expect that write
- * traffic will occur mostly to the latest page (and to the just-prior
- * page, soon after a page transition). Read traffic will probably touch
- * a larger span of pages, but in any case a fairly small number of page
- * buffers should be sufficient. So, we just search the buffers using plain
- * linear search; there's no need for a hashtable or anything fancy.
- * The management algorithm is straight LRU except that we will never swap
- * out the latest page (since we know it's going to be hit again eventually).
- *
- * We use a control LWLock to protect the shared data structures, plus
- * per-buffer LWLocks that synchronize I/O for each buffer. The control lock
- * must be held to examine or modify any shared state. A process that is
- * reading in or writing out a page buffer does not hold the control lock,
- * only the per-buffer lock for the buffer it is working on.
- *
- * "Holding the control lock" means exclusive lock in all cases except for
- * SimpleLruReadPage_ReadOnly(); see comments for SlruRecentlyUsed() for
- * the implications of that.
- *
- * When initiating I/O on a buffer, we acquire the per-buffer lock exclusively
- * before releasing the control lock. The per-buffer lock is released after
- * completing the I/O, re-acquiring the control lock, and updating the shared
- * state. (Deadlock is not possible here, because we never try to initiate
- * I/O when someone else is already doing I/O on the same buffer.)
- * To wait for I/O to complete, release the control lock, acquire the
- * per-buffer lock in shared mode, immediately release the per-buffer lock,
- * reacquire the control lock, and then recheck state (since arbitrary things
- * could have happened while we didn't have the lock).
- *
- * As with the regular buffer manager, it is possible for another process
- * to re-dirty a page that is currently being written out. This is handled
- * by re-setting the page's page_dirty flag.
+ * XXX write me
*
*
* Portions Copyright (c) 1996-2023, PostgreSQL Global Development Group
@@ -60,562 +28,31 @@
#include "storage/fd.h"
#include "storage/shmem.h"
-#define SlruFileName(ctl, path, seg) \
- snprintf(path, MAXPGPATH, "%s/%04X", (ctl)->Dir, seg)
-
-/*
- * During SimpleLruWriteAll(), we will usually not need to write more than one
- * or two physical files, but we may need to write several pages per file. We
- * can consolidate the I/O requests by leaving files open until control returns
- * to SimpleLruWriteAll(). This data structure remembers which files are open.
- */
-#define MAX_WRITEALL_BUFFERS 16
+#define PG_SLRU(symname,name,path,synchronize) \
+ path,
-typedef struct SlruWriteAllData
+static char *slru_dirs[] =
{
- int num_files; /* # files actually open */
- int fd[MAX_WRITEALL_BUFFERS]; /* their FD's */
- int segno[MAX_WRITEALL_BUFFERS]; /* their log seg#s */
-} SlruWriteAllData;
-
-typedef struct SlruWriteAllData *SlruWriteAll;
-
-/*
- * Populate a file tag describing a segment file. We only use the segment
- * number, since we can derive everything else we need by having separate
- * sync handler functions for clog, multixact etc.
- */
-#define INIT_SLRUFILETAG(a,xx_handler,xx_segno) \
-( \
- memset(&(a), 0, sizeof(FileTag)), \
- (a).handler = (xx_handler), \
- (a).segno = (xx_segno) \
-)
+#include "access/slrulist.h"
+};
/*
- * Macro to mark a buffer slot "most recently used". Note multiple evaluation
- * of arguments!
- *
- * The reason for the if-test is that there are often many consecutive
- * accesses to the same page (particularly the latest page). By suppressing
- * useless increments of cur_lru_count, we reduce the probability that old
- * pages' counts will "wrap around" and make them appear recently used.
- *
- * We allow this code to be executed concurrently by multiple processes within
- * SimpleLruReadPage_ReadOnly(). As long as int reads and writes are atomic,
- * this should not cause any completely-bogus values to enter the computation.
- * However, it is possible for either cur_lru_count or individual
- * page_lru_count entries to be "reset" to lower values than they should have,
- * in case a process is delayed while it executes this macro. With care in
- * SlruSelectLRUPage(), this does little harm, and in any case the absolute
- * worst possible consequence is a nonoptimal choice of page to evict. The
- * gain from allowing concurrent reads of SLRU pages seems worth it.
- */
-#define SlruRecentlyUsed(shared, slotno) \
- do { \
- int new_lru_count = (shared)->cur_lru_count; \
- if (new_lru_count != (shared)->page_lru_count[slotno]) { \
- (shared)->cur_lru_count = ++new_lru_count; \
- (shared)->page_lru_count[slotno] = new_lru_count; \
- } \
- } while (0)
-
-/* Saved info for SlruReportIOError */
-typedef enum
-{
- SLRU_OPEN_FAILED,
- SLRU_SEEK_FAILED,
- SLRU_READ_FAILED,
- SLRU_WRITE_FAILED,
- SLRU_FSYNC_FAILED,
- SLRU_CLOSE_FAILED
-} SlruErrorCause;
-
-static SlruErrorCause slru_errcause;
-static int slru_errno;
-
-
-static void SimpleLruZeroLSNs(SlruCtl ctl, int slotno);
-static void SimpleLruWaitIO(SlruCtl ctl, int slotno);
-static void SlruInternalWritePage(SlruCtl ctl, int slotno, SlruWriteAll fdata);
-static bool SlruPhysicalReadPage(SlruCtl ctl, int pageno, int slotno);
-static bool SlruPhysicalWritePage(SlruCtl ctl, int pageno, int slotno,
- SlruWriteAll fdata);
-static void SlruReportIOError(SlruCtl ctl, int pageno, TransactionId xid);
-static int SlruSelectLRUPage(SlruCtl ctl, int pageno);
-
-static bool SlruScanDirCbDeleteCutoff(SlruCtl ctl, char *filename,
+ * We'll maintain a little cache of recently seen buffers, to try to avoid the
+ * buffer mapping table on repeat access (ie the busy end of the CLOG). One
+ * entry per SLRU.
+ */
+struct SlruRecentBuffer {
+ int pageno;
+ Buffer recent_buffer;
+};
+
+static struct SlruRecentBuffer slru_recent_buffers[SLRU_NEXT_ID];
+
+static bool SlruScanDirCbDeleteCutoff(int slru_id,
+ SlruPagePrecedesFunction PagePrecedes,
+ char *filename,
int segpage, void *data);
-static void SlruInternalDeleteSegment(SlruCtl ctl, int segno);
-
-/*
- * Initialization of shared memory
- */
-
-Size
-SimpleLruShmemSize(int nslots, int nlsns)
-{
- Size sz;
-
- /* we assume nslots isn't so large as to risk overflow */
- sz = MAXALIGN(sizeof(SlruSharedData));
- sz += MAXALIGN(nslots * sizeof(char *)); /* page_buffer[] */
- sz += MAXALIGN(nslots * sizeof(SlruPageStatus)); /* page_status[] */
- sz += MAXALIGN(nslots * sizeof(bool)); /* page_dirty[] */
- sz += MAXALIGN(nslots * sizeof(int)); /* page_number[] */
- sz += MAXALIGN(nslots * sizeof(int)); /* page_lru_count[] */
- sz += MAXALIGN(nslots * sizeof(LWLockPadded)); /* buffer_locks[] */
-
- if (nlsns > 0)
- sz += MAXALIGN(nslots * nlsns * sizeof(XLogRecPtr)); /* group_lsn[] */
-
- return BUFFERALIGN(sz) + BLCKSZ * nslots;
-}
-
-/*
- * Initialize, or attach to, a simple LRU cache in shared memory.
- *
- * ctl: address of local (unshared) control structure.
- * name: name of SLRU. (This is user-visible, pick with care!)
- * nslots: number of page slots to use.
- * nlsns: number of LSN groups per page (set to zero if not relevant).
- * ctllock: LWLock to use to control access to the shared control structure.
- * subdir: PGDATA-relative subdirectory that will contain the files.
- * tranche_id: LWLock tranche ID to use for the SLRU's per-buffer LWLocks.
- * sync_handler: which set of functions to use to handle sync requests
- */
-void
-SimpleLruInit(SlruCtl ctl, const char *name, int nslots, int nlsns,
- LWLock *ctllock, const char *subdir, int tranche_id,
- SyncRequestHandler sync_handler)
-{
- SlruShared shared;
- bool found;
-
- shared = (SlruShared) ShmemInitStruct(name,
- SimpleLruShmemSize(nslots, nlsns),
- &found);
-
- if (!IsUnderPostmaster)
- {
- /* Initialize locks and shared memory area */
- char *ptr;
- Size offset;
- int slotno;
-
- Assert(!found);
-
- memset(shared, 0, sizeof(SlruSharedData));
-
- shared->ControlLock = ctllock;
-
- shared->num_slots = nslots;
- shared->lsn_groups_per_page = nlsns;
-
- shared->cur_lru_count = 0;
-
- /* shared->latest_page_number will be set later */
-
- shared->slru_stats_idx = pgstat_get_slru_index(name);
-
- ptr = (char *) shared;
- offset = MAXALIGN(sizeof(SlruSharedData));
- shared->page_buffer = (char **) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(char *));
- shared->page_status = (SlruPageStatus *) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(SlruPageStatus));
- shared->page_dirty = (bool *) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(bool));
- shared->page_number = (int *) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(int));
- shared->page_lru_count = (int *) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(int));
-
- /* Initialize LWLocks */
- shared->buffer_locks = (LWLockPadded *) (ptr + offset);
- offset += MAXALIGN(nslots * sizeof(LWLockPadded));
-
- if (nlsns > 0)
- {
- shared->group_lsn = (XLogRecPtr *) (ptr + offset);
- offset += MAXALIGN(nslots * nlsns * sizeof(XLogRecPtr));
- }
-
- ptr += BUFFERALIGN(offset);
- for (slotno = 0; slotno < nslots; slotno++)
- {
- LWLockInitialize(&shared->buffer_locks[slotno].lock,
- tranche_id);
-
- shared->page_buffer[slotno] = ptr;
- shared->page_status[slotno] = SLRU_PAGE_EMPTY;
- shared->page_dirty[slotno] = false;
- shared->page_lru_count[slotno] = 0;
- ptr += BLCKSZ;
- }
-
- /* Should fit to estimated shmem size */
- Assert(ptr - (char *) shared <= SimpleLruShmemSize(nslots, nlsns));
- }
- else
- Assert(found);
-
- /*
- * Initialize the unshared control struct, including directory path. We
- * assume caller set PagePrecedes.
- */
- ctl->shared = shared;
- ctl->sync_handler = sync_handler;
- strlcpy(ctl->Dir, subdir, sizeof(ctl->Dir));
-}
-
-/*
- * Initialize (or reinitialize) a page to zeroes.
- *
- * The page is not actually written, just set up in shared memory.
- * The slot number of the new page is returned.
- *
- * Control lock must be held at entry, and will be held at exit.
- */
-int
-SimpleLruZeroPage(SlruCtl ctl, int pageno)
-{
- SlruShared shared = ctl->shared;
- int slotno;
-
- /* Find a suitable buffer slot for the page */
- slotno = SlruSelectLRUPage(ctl, pageno);
- Assert(shared->page_status[slotno] == SLRU_PAGE_EMPTY ||
- (shared->page_status[slotno] == SLRU_PAGE_VALID &&
- !shared->page_dirty[slotno]) ||
- shared->page_number[slotno] == pageno);
-
- /* Mark the slot as containing this page */
- shared->page_number[slotno] = pageno;
- shared->page_status[slotno] = SLRU_PAGE_VALID;
- shared->page_dirty[slotno] = true;
- SlruRecentlyUsed(shared, slotno);
-
- /* Set the buffer to zeroes */
- MemSet(shared->page_buffer[slotno], 0, BLCKSZ);
-
- /* Set the LSNs for this new page to zero */
- SimpleLruZeroLSNs(ctl, slotno);
-
- /* Assume this page is now the latest active page */
- shared->latest_page_number = pageno;
-
- /* update the stats counter of zeroed pages */
- pgstat_count_slru_page_zeroed(shared->slru_stats_idx);
-
- return slotno;
-}
-
-/*
- * Zero all the LSNs we store for this slru page.
- *
- * This should be called each time we create a new page, and each time we read
- * in a page from disk into an existing buffer. (Such an old page cannot
- * have any interesting LSNs, since we'd have flushed them before writing
- * the page in the first place.)
- *
- * This assumes that InvalidXLogRecPtr is bitwise-all-0.
- */
-static void
-SimpleLruZeroLSNs(SlruCtl ctl, int slotno)
-{
- SlruShared shared = ctl->shared;
-
- if (shared->lsn_groups_per_page > 0)
- MemSet(&shared->group_lsn[slotno * shared->lsn_groups_per_page], 0,
- shared->lsn_groups_per_page * sizeof(XLogRecPtr));
-}
-
-/*
- * Wait for any active I/O on a page slot to finish. (This does not
- * guarantee that new I/O hasn't been started before we return, though.
- * In fact the slot might not even contain the same page anymore.)
- *
- * Control lock must be held at entry, and will be held at exit.
- */
-static void
-SimpleLruWaitIO(SlruCtl ctl, int slotno)
-{
- SlruShared shared = ctl->shared;
-
- /* See notes at top of file */
- LWLockRelease(shared->ControlLock);
- LWLockAcquire(&shared->buffer_locks[slotno].lock, LW_SHARED);
- LWLockRelease(&shared->buffer_locks[slotno].lock);
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
- /*
- * If the slot is still in an io-in-progress state, then either someone
- * already started a new I/O on the slot, or a previous I/O failed and
- * neglected to reset the page state. That shouldn't happen, really, but
- * it seems worth a few extra cycles to check and recover from it. We can
- * cheaply test for failure by seeing if the buffer lock is still held (we
- * assume that transaction abort would release the lock).
- */
- if (shared->page_status[slotno] == SLRU_PAGE_READ_IN_PROGRESS ||
- shared->page_status[slotno] == SLRU_PAGE_WRITE_IN_PROGRESS)
- {
- if (LWLockConditionalAcquire(&shared->buffer_locks[slotno].lock, LW_SHARED))
- {
- /* indeed, the I/O must have failed */
- if (shared->page_status[slotno] == SLRU_PAGE_READ_IN_PROGRESS)
- shared->page_status[slotno] = SLRU_PAGE_EMPTY;
- else /* write_in_progress */
- {
- shared->page_status[slotno] = SLRU_PAGE_VALID;
- shared->page_dirty[slotno] = true;
- }
- LWLockRelease(&shared->buffer_locks[slotno].lock);
- }
- }
-}
-
-/*
- * Find a page in a shared buffer, reading it in if necessary.
- * The page number must correspond to an already-initialized page.
- *
- * If write_ok is true then it is OK to return a page that is in
- * WRITE_IN_PROGRESS state; it is the caller's responsibility to be sure
- * that modification of the page is safe. If write_ok is false then we
- * will not return the page until it is not undergoing active I/O.
- *
- * The passed-in xid is used only for error reporting, and may be
- * InvalidTransactionId if no specific xid is associated with the action.
- *
- * Return value is the shared-buffer slot number now holding the page.
- * The buffer's LRU access info is updated.
- *
- * Control lock must be held at entry, and will be held at exit.
- */
-int
-SimpleLruReadPage(SlruCtl ctl, int pageno, bool write_ok,
- TransactionId xid)
-{
- SlruShared shared = ctl->shared;
-
- /* Outer loop handles restart if we must wait for someone else's I/O */
- for (;;)
- {
- int slotno;
- bool ok;
-
- /* See if page already is in memory; if not, pick victim slot */
- slotno = SlruSelectLRUPage(ctl, pageno);
-
- /* Did we find the page in memory? */
- if (shared->page_number[slotno] == pageno &&
- shared->page_status[slotno] != SLRU_PAGE_EMPTY)
- {
- /*
- * If page is still being read in, we must wait for I/O. Likewise
- * if the page is being written and the caller said that's not OK.
- */
- if (shared->page_status[slotno] == SLRU_PAGE_READ_IN_PROGRESS ||
- (shared->page_status[slotno] == SLRU_PAGE_WRITE_IN_PROGRESS &&
- !write_ok))
- {
- SimpleLruWaitIO(ctl, slotno);
- /* Now we must recheck state from the top */
- continue;
- }
- /* Otherwise, it's ready to use */
- SlruRecentlyUsed(shared, slotno);
-
- /* update the stats counter of pages found in the SLRU */
- pgstat_count_slru_page_hit(shared->slru_stats_idx);
-
- return slotno;
- }
-
- /* We found no match; assert we selected a freeable slot */
- Assert(shared->page_status[slotno] == SLRU_PAGE_EMPTY ||
- (shared->page_status[slotno] == SLRU_PAGE_VALID &&
- !shared->page_dirty[slotno]));
-
- /* Mark the slot read-busy */
- shared->page_number[slotno] = pageno;
- shared->page_status[slotno] = SLRU_PAGE_READ_IN_PROGRESS;
- shared->page_dirty[slotno] = false;
-
- /* Acquire per-buffer lock (cannot deadlock, see notes at top) */
- LWLockAcquire(&shared->buffer_locks[slotno].lock, LW_EXCLUSIVE);
-
- /* Release control lock while doing I/O */
- LWLockRelease(shared->ControlLock);
-
- /* Do the read */
- ok = SlruPhysicalReadPage(ctl, pageno, slotno);
-
- /* Set the LSNs for this newly read-in page to zero */
- SimpleLruZeroLSNs(ctl, slotno);
-
- /* Re-acquire control lock and update page state */
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
- Assert(shared->page_number[slotno] == pageno &&
- shared->page_status[slotno] == SLRU_PAGE_READ_IN_PROGRESS &&
- !shared->page_dirty[slotno]);
-
- shared->page_status[slotno] = ok ? SLRU_PAGE_VALID : SLRU_PAGE_EMPTY;
-
- LWLockRelease(&shared->buffer_locks[slotno].lock);
-
- /* Now it's okay to ereport if we failed */
- if (!ok)
- SlruReportIOError(ctl, pageno, xid);
-
- SlruRecentlyUsed(shared, slotno);
-
- /* update the stats counter of pages not found in SLRU */
- pgstat_count_slru_page_read(shared->slru_stats_idx);
-
- return slotno;
- }
-}
-
-/*
- * Find a page in a shared buffer, reading it in if necessary.
- * The page number must correspond to an already-initialized page.
- * The caller must intend only read-only access to the page.
- *
- * The passed-in xid is used only for error reporting, and may be
- * InvalidTransactionId if no specific xid is associated with the action.
- *
- * Return value is the shared-buffer slot number now holding the page.
- * The buffer's LRU access info is updated.
- *
- * Control lock must NOT be held at entry, but will be held at exit.
- * It is unspecified whether the lock will be shared or exclusive.
- */
-int
-SimpleLruReadPage_ReadOnly(SlruCtl ctl, int pageno, TransactionId xid)
-{
- SlruShared shared = ctl->shared;
- int slotno;
-
- /* Try to find the page while holding only shared lock */
- LWLockAcquire(shared->ControlLock, LW_SHARED);
-
- /* See if page is already in a buffer */
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- if (shared->page_number[slotno] == pageno &&
- shared->page_status[slotno] != SLRU_PAGE_EMPTY &&
- shared->page_status[slotno] != SLRU_PAGE_READ_IN_PROGRESS)
- {
- /* See comments for SlruRecentlyUsed macro */
- SlruRecentlyUsed(shared, slotno);
-
- /* update the stats counter of pages found in the SLRU */
- pgstat_count_slru_page_hit(shared->slru_stats_idx);
-
- return slotno;
- }
- }
-
- /* No luck, so switch to normal exclusive lock and do regular read */
- LWLockRelease(shared->ControlLock);
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
- return SimpleLruReadPage(ctl, pageno, true, xid);
-}
-
-/*
- * Write a page from a shared buffer, if necessary.
- * Does nothing if the specified slot is not dirty.
- *
- * NOTE: only one write attempt is made here. Hence, it is possible that
- * the page is still dirty at exit (if someone else re-dirtied it during
- * the write). However, we *do* attempt a fresh write even if the page
- * is already being written; this is for checkpoints.
- *
- * Control lock must be held at entry, and will be held at exit.
- */
-static void
-SlruInternalWritePage(SlruCtl ctl, int slotno, SlruWriteAll fdata)
-{
- SlruShared shared = ctl->shared;
- int pageno = shared->page_number[slotno];
- bool ok;
-
- /* If a write is in progress, wait for it to finish */
- while (shared->page_status[slotno] == SLRU_PAGE_WRITE_IN_PROGRESS &&
- shared->page_number[slotno] == pageno)
- {
- SimpleLruWaitIO(ctl, slotno);
- }
-
- /*
- * Do nothing if page is not dirty, or if buffer no longer contains the
- * same page we were called for.
- */
- if (!shared->page_dirty[slotno] ||
- shared->page_status[slotno] != SLRU_PAGE_VALID ||
- shared->page_number[slotno] != pageno)
- return;
-
- /*
- * Mark the slot write-busy, and clear the dirtybit. After this point, a
- * transaction status update on this page will mark it dirty again.
- */
- shared->page_status[slotno] = SLRU_PAGE_WRITE_IN_PROGRESS;
- shared->page_dirty[slotno] = false;
-
- /* Acquire per-buffer lock (cannot deadlock, see notes at top) */
- LWLockAcquire(&shared->buffer_locks[slotno].lock, LW_EXCLUSIVE);
-
- /* Release control lock while doing I/O */
- LWLockRelease(shared->ControlLock);
-
- /* Do the write */
- ok = SlruPhysicalWritePage(ctl, pageno, slotno, fdata);
-
- /* If we failed, and we're in a flush, better close the files */
- if (!ok && fdata)
- {
- int i;
-
- for (i = 0; i < fdata->num_files; i++)
- CloseTransientFile(fdata->fd[i]);
- }
-
- /* Re-acquire control lock and update page state */
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
- Assert(shared->page_number[slotno] == pageno &&
- shared->page_status[slotno] == SLRU_PAGE_WRITE_IN_PROGRESS);
-
- /* If we failed to write, mark the page dirty again */
- if (!ok)
- shared->page_dirty[slotno] = true;
-
- shared->page_status[slotno] = SLRU_PAGE_VALID;
-
- LWLockRelease(&shared->buffer_locks[slotno].lock);
-
- /* Now it's okay to ereport if we failed */
- if (!ok)
- SlruReportIOError(ctl, pageno, InvalidTransactionId);
-
- /* If part of a checkpoint, count this as a buffer written. */
- if (fdata)
- CheckpointStats.ckpt_bufs_written++;
-}
-
-/*
- * Wrapper of SlruInternalWritePage, for external callers.
- * fdata is always passed a NULL here.
- */
-void
-SimpleLruWritePage(SlruCtl ctl, int slotno)
-{
- SlruInternalWritePage(ctl, slotno, NULL);
-}
+static void SlruInternalDeleteSegment(int slru_id, int segno);
/*
* Return whether the given page exists on disk.
@@ -624,592 +61,24 @@ SimpleLruWritePage(SlruCtl ctl, int slotno)
* large enough to contain the given page.
*/
bool
-SimpleLruDoesPhysicalPageExist(SlruCtl ctl, int pageno)
+SimpleLruDoesPhysicalPageExist(int slru_id, int pageno)
{
- int segno = pageno / SLRU_PAGES_PER_SEGMENT;
- int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
- int offset = rpageno * BLCKSZ;
- char path[MAXPGPATH];
- int fd;
- bool result;
- off_t endpos;
-
- /* update the stats counter of checked pages */
- pgstat_count_slru_page_exists(ctl->shared->slru_stats_idx);
-
- SlruFileName(ctl, path, segno);
-
- fd = OpenTransientFile(path, O_RDONLY | PG_BINARY);
- if (fd < 0)
- {
- /* expected: file doesn't exist */
- if (errno == ENOENT)
- return false;
-
- /* report error normally */
- slru_errcause = SLRU_OPEN_FAILED;
- slru_errno = errno;
- SlruReportIOError(ctl, pageno, 0);
- }
-
- if ((endpos = lseek(fd, 0, SEEK_END)) < 0)
- {
- slru_errcause = SLRU_SEEK_FAILED;
- slru_errno = errno;
- SlruReportIOError(ctl, pageno, 0);
- }
-
- result = endpos >= (off_t) (offset + BLCKSZ);
-
- if (CloseTransientFile(fd) != 0)
- {
- slru_errcause = SLRU_CLOSE_FAILED;
- slru_errno = errno;
- return false;
- }
-
- return result;
-}
-
-/*
- * Physical read of a (previously existing) page into a buffer slot
- *
- * On failure, we cannot just ereport(ERROR) since caller has put state in
- * shared memory that must be undone. So, we return false and save enough
- * info in static variables to let SlruReportIOError make the report.
- *
- * For now, assume it's not worth keeping a file pointer open across
- * read/write operations. We could cache one virtual file pointer ...
- */
-static bool
-SlruPhysicalReadPage(SlruCtl ctl, int pageno, int slotno)
-{
- SlruShared shared = ctl->shared;
int segno = pageno / SLRU_PAGES_PER_SEGMENT;
int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
off_t offset = rpageno * BLCKSZ;
- char path[MAXPGPATH];
- int fd;
-
- SlruFileName(ctl, path, segno);
-
- /*
- * In a crash-and-restart situation, it's possible for us to receive
- * commands to set the commit status of transactions whose bits are in
- * already-truncated segments of the commit log (see notes in
- * SlruPhysicalWritePage). Hence, if we are InRecovery, allow the case
- * where the file doesn't exist, and return zeroes instead.
- */
- fd = OpenTransientFile(path, O_RDONLY | PG_BINARY);
- if (fd < 0)
- {
- if (errno != ENOENT || !InRecovery)
- {
- slru_errcause = SLRU_OPEN_FAILED;
- slru_errno = errno;
- return false;
- }
-
- ereport(LOG,
- (errmsg("file \"%s\" doesn't exist, reading as zeroes",
- path)));
- MemSet(shared->page_buffer[slotno], 0, BLCKSZ);
- return true;
- }
-
- errno = 0;
- pgstat_report_wait_start(WAIT_EVENT_SLRU_READ);
- if (pg_pread(fd, shared->page_buffer[slotno], BLCKSZ, offset) != BLCKSZ)
- {
- pgstat_report_wait_end();
- slru_errcause = SLRU_READ_FAILED;
- slru_errno = errno;
- CloseTransientFile(fd);
- return false;
- }
- pgstat_report_wait_end();
-
- if (CloseTransientFile(fd) != 0)
- {
- slru_errcause = SLRU_CLOSE_FAILED;
- slru_errno = errno;
- return false;
- }
-
- return true;
-}
-
-/*
- * Physical write of a page from a buffer slot
- *
- * On failure, we cannot just ereport(ERROR) since caller has put state in
- * shared memory that must be undone. So, we return false and save enough
- * info in static variables to let SlruReportIOError make the report.
- *
- * For now, assume it's not worth keeping a file pointer open across
- * independent read/write operations. We do batch operations during
- * SimpleLruWriteAll, though.
- *
- * fdata is NULL for a standalone write, pointer to open-file info during
- * SimpleLruWriteAll.
- */
-static bool
-SlruPhysicalWritePage(SlruCtl ctl, int pageno, int slotno, SlruWriteAll fdata)
-{
- SlruShared shared = ctl->shared;
- int segno = pageno / SLRU_PAGES_PER_SEGMENT;
- int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
- off_t offset = rpageno * BLCKSZ;
- char path[MAXPGPATH];
- int fd = -1;
-
- /* update the stats counter of written pages */
- pgstat_count_slru_page_written(shared->slru_stats_idx);
-
- /*
- * Honor the write-WAL-before-data rule, if appropriate, so that we do not
- * write out data before associated WAL records. This is the same action
- * performed during FlushBuffer() in the main buffer manager.
- */
- if (shared->group_lsn != NULL)
- {
- /*
- * We must determine the largest async-commit LSN for the page. This
- * is a bit tedious, but since this entire function is a slow path
- * anyway, it seems better to do this here than to maintain a per-page
- * LSN variable (which'd need an extra comparison in the
- * transaction-commit path).
- */
- XLogRecPtr max_lsn;
- int lsnindex,
- lsnoff;
-
- lsnindex = slotno * shared->lsn_groups_per_page;
- max_lsn = shared->group_lsn[lsnindex++];
- for (lsnoff = 1; lsnoff < shared->lsn_groups_per_page; lsnoff++)
- {
- XLogRecPtr this_lsn = shared->group_lsn[lsnindex++];
-
- if (max_lsn < this_lsn)
- max_lsn = this_lsn;
- }
-
- if (!XLogRecPtrIsInvalid(max_lsn))
- {
- /*
- * As noted above, elog(ERROR) is not acceptable here, so if
- * XLogFlush were to fail, we must PANIC. This isn't much of a
- * restriction because XLogFlush is just about all critical
- * section anyway, but let's make sure.
- */
- START_CRIT_SECTION();
- XLogFlush(max_lsn);
- END_CRIT_SECTION();
- }
- }
-
- /*
- * During a WriteAll, we may already have the desired file open.
- */
- if (fdata)
- {
- int i;
-
- for (i = 0; i < fdata->num_files; i++)
- {
- if (fdata->segno[i] == segno)
- {
- fd = fdata->fd[i];
- break;
- }
- }
- }
-
- if (fd < 0)
- {
- /*
- * If the file doesn't already exist, we should create it. It is
- * possible for this to need to happen when writing a page that's not
- * first in its segment; we assume the OS can cope with that. (Note:
- * it might seem that it'd be okay to create files only when
- * SimpleLruZeroPage is called for the first page of a segment.
- * However, if after a crash and restart the REDO logic elects to
- * replay the log from a checkpoint before the latest one, then it's
- * possible that we will get commands to set transaction status of
- * transactions that have already been truncated from the commit log.
- * Easiest way to deal with that is to accept references to
- * nonexistent files here and in SlruPhysicalReadPage.)
- *
- * Note: it is possible for more than one backend to be executing this
- * code simultaneously for different pages of the same file. Hence,
- * don't use O_EXCL or O_TRUNC or anything like that.
- */
- SlruFileName(ctl, path, segno);
- fd = OpenTransientFile(path, O_RDWR | O_CREAT | PG_BINARY);
- if (fd < 0)
- {
- slru_errcause = SLRU_OPEN_FAILED;
- slru_errno = errno;
- return false;
- }
-
- if (fdata)
- {
- if (fdata->num_files < MAX_WRITEALL_BUFFERS)
- {
- fdata->fd[fdata->num_files] = fd;
- fdata->segno[fdata->num_files] = segno;
- fdata->num_files++;
- }
- else
- {
- /*
- * In the unlikely event that we exceed MAX_FLUSH_BUFFERS,
- * fall back to treating it as a standalone write.
- */
- fdata = NULL;
- }
- }
- }
-
- errno = 0;
- pgstat_report_wait_start(WAIT_EVENT_SLRU_WRITE);
- if (pg_pwrite(fd, shared->page_buffer[slotno], BLCKSZ, offset) != BLCKSZ)
- {
- pgstat_report_wait_end();
- /* if write didn't set errno, assume problem is no disk space */
- if (errno == 0)
- errno = ENOSPC;
- slru_errcause = SLRU_WRITE_FAILED;
- slru_errno = errno;
- if (!fdata)
- CloseTransientFile(fd);
- return false;
- }
- pgstat_report_wait_end();
-
- /* Queue up a sync request for the checkpointer. */
- if (ctl->sync_handler != SYNC_HANDLER_NONE)
- {
- FileTag tag;
-
- INIT_SLRUFILETAG(tag, ctl->sync_handler, segno);
- if (!RegisterSyncRequest(&tag, SYNC_REQUEST, false))
- {
- /* No space to enqueue sync request. Do it synchronously. */
- pgstat_report_wait_start(WAIT_EVENT_SLRU_SYNC);
- if (pg_fsync(fd) != 0)
- {
- pgstat_report_wait_end();
- slru_errcause = SLRU_FSYNC_FAILED;
- slru_errno = errno;
- CloseTransientFile(fd);
- return false;
- }
- pgstat_report_wait_end();
- }
- }
-
- /* Close file, unless part of flush request. */
- if (!fdata)
- {
- if (CloseTransientFile(fd) != 0)
- {
- slru_errcause = SLRU_CLOSE_FAILED;
- slru_errno = errno;
- return false;
- }
- }
-
- return true;
-}
-
-/*
- * Issue the error message after failure of SlruPhysicalReadPage or
- * SlruPhysicalWritePage. Call this after cleaning up shared-memory state.
- */
-static void
-SlruReportIOError(SlruCtl ctl, int pageno, TransactionId xid)
-{
- int segno = pageno / SLRU_PAGES_PER_SEGMENT;
- int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
- int offset = rpageno * BLCKSZ;
- char path[MAXPGPATH];
-
- SlruFileName(ctl, path, segno);
- errno = slru_errno;
- switch (slru_errcause)
- {
- case SLRU_OPEN_FAILED:
- ereport(ERROR,
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not open file \"%s\": %m.", path)));
- break;
- case SLRU_SEEK_FAILED:
- ereport(ERROR,
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not seek in file \"%s\" to offset %d: %m.",
- path, offset)));
- break;
- case SLRU_READ_FAILED:
- if (errno)
- ereport(ERROR,
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not read from file \"%s\" at offset %d: %m.",
- path, offset)));
- else
- ereport(ERROR,
- (errmsg("could not access status of transaction %u", xid),
- errdetail("Could not read from file \"%s\" at offset %d: read too few bytes.", path, offset)));
- break;
- case SLRU_WRITE_FAILED:
- if (errno)
- ereport(ERROR,
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not write to file \"%s\" at offset %d: %m.",
- path, offset)));
- else
- ereport(ERROR,
- (errmsg("could not access status of transaction %u", xid),
- errdetail("Could not write to file \"%s\" at offset %d: wrote too few bytes.",
- path, offset)));
- break;
- case SLRU_FSYNC_FAILED:
- ereport(data_sync_elevel(ERROR),
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not fsync file \"%s\": %m.",
- path)));
- break;
- case SLRU_CLOSE_FAILED:
- ereport(ERROR,
- (errcode_for_file_access(),
- errmsg("could not access status of transaction %u", xid),
- errdetail("Could not close file \"%s\": %m.",
- path)));
- break;
- default:
- /* can't get here, we trust */
- elog(ERROR, "unrecognized SimpleLru error cause: %d",
- (int) slru_errcause);
- break;
- }
-}
-
-/*
- * Select the slot to re-use when we need a free slot.
- *
- * The target page number is passed because we need to consider the
- * possibility that some other process reads in the target page while
- * we are doing I/O to free a slot. Hence, check or recheck to see if
- * any slot already holds the target page, and return that slot if so.
- * Thus, the returned slot is *either* a slot already holding the pageno
- * (could be any state except EMPTY), *or* a freeable slot (state EMPTY
- * or CLEAN).
- *
- * Control lock must be held at entry, and will be held at exit.
- */
-static int
-SlruSelectLRUPage(SlruCtl ctl, int pageno)
-{
- SlruShared shared = ctl->shared;
-
- /* Outer loop handles restart after I/O */
- for (;;)
- {
- int slotno;
- int cur_count;
- int bestvalidslot = 0; /* keep compiler quiet */
- int best_valid_delta = -1;
- int best_valid_page_number = 0; /* keep compiler quiet */
- int bestinvalidslot = 0; /* keep compiler quiet */
- int best_invalid_delta = -1;
- int best_invalid_page_number = 0; /* keep compiler quiet */
+ off_t size;
+ RelFileLocator rlocator = SlruRelFileLocator(slru_id, segno);
+ SMgrFileHandle sfile = smgropen(rlocator, InvalidBackendId, MAIN_FORKNUM);
- /* See if page already has a buffer assigned */
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- if (shared->page_number[slotno] == pageno &&
- shared->page_status[slotno] != SLRU_PAGE_EMPTY)
- return slotno;
- }
-
- /*
- * If we find any EMPTY slot, just select that one. Else choose a
- * victim page to replace. We normally take the least recently used
- * valid page, but we will never take the slot containing
- * latest_page_number, even if it appears least recently used. We
- * will select a slot that is already I/O busy only if there is no
- * other choice: a read-busy slot will not be least recently used once
- * the read finishes, and waiting for an I/O on a write-busy slot is
- * inferior to just picking some other slot. Testing shows the slot
- * we pick instead will often be clean, allowing us to begin a read at
- * once.
- *
- * Normally the page_lru_count values will all be different and so
- * there will be a well-defined LRU page. But since we allow
- * concurrent execution of SlruRecentlyUsed() within
- * SimpleLruReadPage_ReadOnly(), it is possible that multiple pages
- * acquire the same lru_count values. In that case we break ties by
- * choosing the furthest-back page.
- *
- * Notice that this next line forcibly advances cur_lru_count to a
- * value that is certainly beyond any value that will be in the
- * page_lru_count array after the loop finishes. This ensures that
- * the next execution of SlruRecentlyUsed will mark the page newly
- * used, even if it's for a page that has the current counter value.
- * That gets us back on the path to having good data when there are
- * multiple pages with the same lru_count.
- */
- cur_count = (shared->cur_lru_count)++;
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- int this_delta;
- int this_page_number;
-
- if (shared->page_status[slotno] == SLRU_PAGE_EMPTY)
- return slotno;
- this_delta = cur_count - shared->page_lru_count[slotno];
- if (this_delta < 0)
- {
- /*
- * Clean up in case shared updates have caused cur_count
- * increments to get "lost". We back off the page counts,
- * rather than trying to increase cur_count, to avoid any
- * question of infinite loops or failure in the presence of
- * wrapped-around counts.
- */
- shared->page_lru_count[slotno] = cur_count;
- this_delta = 0;
- }
- this_page_number = shared->page_number[slotno];
- if (this_page_number == shared->latest_page_number)
- continue;
- if (shared->page_status[slotno] == SLRU_PAGE_VALID)
- {
- if (this_delta > best_valid_delta ||
- (this_delta == best_valid_delta &&
- ctl->PagePrecedes(this_page_number,
- best_valid_page_number)))
- {
- bestvalidslot = slotno;
- best_valid_delta = this_delta;
- best_valid_page_number = this_page_number;
- }
- }
- else
- {
- if (this_delta > best_invalid_delta ||
- (this_delta == best_invalid_delta &&
- ctl->PagePrecedes(this_page_number,
- best_invalid_page_number)))
- {
- bestinvalidslot = slotno;
- best_invalid_delta = this_delta;
- best_invalid_page_number = this_page_number;
- }
- }
- }
-
- /*
- * If all pages (except possibly the latest one) are I/O busy, we'll
- * have to wait for an I/O to complete and then retry. In that
- * unhappy case, we choose to wait for the I/O on the least recently
- * used slot, on the assumption that it was likely initiated first of
- * all the I/Os in progress and may therefore finish first.
- */
- if (best_valid_delta < 0)
- {
- SimpleLruWaitIO(ctl, bestinvalidslot);
- continue;
- }
-
- /*
- * If the selected page is clean, we're set.
- */
- if (!shared->page_dirty[bestvalidslot])
- return bestvalidslot;
-
- /*
- * Write the page.
- */
- SlruInternalWritePage(ctl, bestvalidslot, NULL);
-
- /*
- * Now loop back and try again. This is the easiest way of dealing
- * with corner cases such as the victim page being re-dirtied while we
- * wrote it.
- */
- }
-}
-
-/*
- * Write dirty pages to disk during checkpoint or database shutdown. Flushing
- * is deferred until the next call to ProcessSyncRequests(), though we do fsync
- * the containing directory here to make sure that newly created directory
- * entries are on disk.
- */
-void
-SimpleLruWriteAll(SlruCtl ctl, bool allow_redirtied)
-{
- SlruShared shared = ctl->shared;
- SlruWriteAllData fdata;
- int slotno;
- int pageno = 0;
- int i;
- bool ok;
-
- /* update the stats counter of flushes */
- pgstat_count_slru_flush(shared->slru_stats_idx);
-
- /*
- * Find and write dirty pages
- */
- fdata.num_files = 0;
-
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- SlruInternalWritePage(ctl, slotno, &fdata);
-
- /*
- * In some places (e.g. checkpoints), we cannot assert that the slot
- * is clean now, since another process might have re-dirtied it
- * already. That's okay.
- */
- Assert(allow_redirtied ||
- shared->page_status[slotno] == SLRU_PAGE_EMPTY ||
- (shared->page_status[slotno] == SLRU_PAGE_VALID &&
- !shared->page_dirty[slotno]));
- }
-
- LWLockRelease(shared->ControlLock);
+ /* update the stats counter of checked pages */
+ pgstat_count_slru_page_exists(slru_id);
- /*
- * Now close any files that were open
- */
- ok = true;
- for (i = 0; i < fdata.num_files; i++)
- {
- if (CloseTransientFile(fdata.fd[i]) != 0)
- {
- slru_errcause = SLRU_CLOSE_FAILED;
- slru_errno = errno;
- pageno = fdata.segno[i] * SLRU_PAGES_PER_SEGMENT;
- ok = false;
- }
- }
- if (!ok)
- SlruReportIOError(ctl, pageno, InvalidTransactionId);
+ if (smgrexists(sfile))
+ size = smgrnblocks(sfile);
+ else
+ size = 0;
- /* Ensure that directory entries for new files are on disk. */
- if (ctl->sync_handler != SYNC_HANDLER_NONE)
- fsync_fname(ctl->Dir, true);
+ return size >= offset + BLCKSZ;
}
/*
@@ -1224,75 +93,14 @@ SimpleLruWriteAll(SlruCtl ctl, bool allow_redirtied)
* after it has accrued freshly-written data.
*/
void
-SimpleLruTruncate(SlruCtl ctl, int cutoffPage)
+SimpleLruTruncate(int slru_id, SlruPagePrecedesFunction PagePrecedes, int cutoffPage)
{
- SlruShared shared = ctl->shared;
- int slotno;
-
/* update the stats counter of truncates */
- pgstat_count_slru_truncate(shared->slru_stats_idx);
-
- /*
- * Scan shared memory and remove any pages preceding the cutoff page, to
- * ensure we won't rewrite them later. (Since this is normally called in
- * or just after a checkpoint, any dirty pages should have been flushed
- * already ... we're just being extra careful here.)
- */
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-
-restart:
-
- /*
- * While we are holding the lock, make an important safety check: the
- * current endpoint page must not be eligible for removal.
- */
- if (ctl->PagePrecedes(shared->latest_page_number, cutoffPage))
- {
- LWLockRelease(shared->ControlLock);
- ereport(LOG,
- (errmsg("could not truncate directory \"%s\": apparent wraparound",
- ctl->Dir)));
- return;
- }
-
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- if (shared->page_status[slotno] == SLRU_PAGE_EMPTY)
- continue;
- if (!ctl->PagePrecedes(shared->page_number[slotno], cutoffPage))
- continue;
-
- /*
- * If page is clean, just change state to EMPTY (expected case).
- */
- if (shared->page_status[slotno] == SLRU_PAGE_VALID &&
- !shared->page_dirty[slotno])
- {
- shared->page_status[slotno] = SLRU_PAGE_EMPTY;
- continue;
- }
-
- /*
- * Hmm, we have (or may have) I/O operations acting on the page, so
- * we've got to wait for them to finish and then start again. This is
- * the same logic as in SlruSelectLRUPage. (XXX if page is dirty,
- * wouldn't it be OK to just discard it without writing it?
- * SlruMayDeleteSegment() uses a stricter qualification, so we might
- * not delete this page in the end; even if we don't delete it, we
- * won't have cause to read its data again. For now, keep the logic
- * the same as it was.)
- */
- if (shared->page_status[slotno] == SLRU_PAGE_VALID)
- SlruInternalWritePage(ctl, slotno, NULL);
- else
- SimpleLruWaitIO(ctl, slotno);
- goto restart;
- }
-
- LWLockRelease(shared->ControlLock);
+ pgstat_count_slru_truncate(slru_id);
/* Now we can remove the old segment(s) */
- (void) SlruScanDirectory(ctl, SlruScanDirCbDeleteCutoff, &cutoffPage);
+ (void) SlruScanDirectory(slru_id, PagePrecedes, SlruScanDirCbDeleteCutoff,
+ &cutoffPage);
}
/*
@@ -1302,77 +110,22 @@ restart:
* they either can't yet contain anything, or have already been cleaned out.
*/
static void
-SlruInternalDeleteSegment(SlruCtl ctl, int segno)
+SlruInternalDeleteSegment(int slru_id, int segno)
{
- char path[MAXPGPATH];
-
- /* Forget any fsync requests queued for this segment. */
- if (ctl->sync_handler != SYNC_HANDLER_NONE)
- {
- FileTag tag;
-
- INIT_SLRUFILETAG(tag, ctl->sync_handler, segno);
- RegisterSyncRequest(&tag, SYNC_FORGET_REQUEST, true);
- }
+ RelFileLocator rlocator = SlruRelFileLocator(slru_id, segno);
+ SMgrFileHandle sfile = smgropen(rlocator, InvalidBackendId, MAIN_FORKNUM);
/* Unlink the file. */
- SlruFileName(ctl, path, segno);
- ereport(DEBUG2, (errmsg_internal("removing file \"%s\"", path)));
- unlink(path);
+ smgrunlink(sfile, false);
}
/*
* Delete an individual SLRU segment, identified by the segment number.
*/
void
-SlruDeleteSegment(SlruCtl ctl, int segno)
+SlruDeleteSegment(int slru_id, int segno)
{
- SlruShared shared = ctl->shared;
- int slotno;
- bool did_write;
-
- /* Clean out any possibly existing references to the segment. */
- LWLockAcquire(shared->ControlLock, LW_EXCLUSIVE);
-restart:
- did_write = false;
- for (slotno = 0; slotno < shared->num_slots; slotno++)
- {
- int pagesegno = shared->page_number[slotno] / SLRU_PAGES_PER_SEGMENT;
-
- if (shared->page_status[slotno] == SLRU_PAGE_EMPTY)
- continue;
-
- /* not the segment we're looking for */
- if (pagesegno != segno)
- continue;
-
- /* If page is clean, just change state to EMPTY (expected case). */
- if (shared->page_status[slotno] == SLRU_PAGE_VALID &&
- !shared->page_dirty[slotno])
- {
- shared->page_status[slotno] = SLRU_PAGE_EMPTY;
- continue;
- }
-
- /* Same logic as SimpleLruTruncate() */
- if (shared->page_status[slotno] == SLRU_PAGE_VALID)
- SlruInternalWritePage(ctl, slotno, NULL);
- else
- SimpleLruWaitIO(ctl, slotno);
-
- did_write = true;
- }
-
- /*
- * Be extra careful and re-check. The IO functions release the control
- * lock, so new pages could have been read in.
- */
- if (did_write)
- goto restart;
-
- SlruInternalDeleteSegment(ctl, segno);
-
- LWLockRelease(shared->ControlLock);
+ SlruInternalDeleteSegment(slru_id, segno);
}
/*
@@ -1389,19 +142,21 @@ restart:
* first>=cutoff && last>=cutoff: no; every page of this segment is too young
*/
static bool
-SlruMayDeleteSegment(SlruCtl ctl, int segpage, int cutoffPage)
+SlruMayDeleteSegment(SlruPagePrecedesFunction PagePrecedes,
+ int segpage, int cutoffPage)
{
int seg_last_page = segpage + SLRU_PAGES_PER_SEGMENT - 1;
Assert(segpage % SLRU_PAGES_PER_SEGMENT == 0);
- return (ctl->PagePrecedes(segpage, cutoffPage) &&
- ctl->PagePrecedes(seg_last_page, cutoffPage));
+ return (PagePrecedes(segpage, cutoffPage) &&
+ PagePrecedes(seg_last_page, cutoffPage));
}
#ifdef USE_ASSERT_CHECKING
static void
-SlruPagePrecedesTestOffset(SlruCtl ctl, int per_page, uint32 offset)
+SlruPagePrecedesTestOffset(SlruPagePrecedesFunction PagePrecedes,
+ int per_page, uint32 offset)
{
TransactionId lhs,
rhs;
@@ -1426,19 +181,19 @@ SlruPagePrecedesTestOffset(SlruCtl ctl, int per_page, uint32 offset)
Assert(!TransactionIdPrecedes(rhs, lhs + 1));
Assert(!TransactionIdFollowsOrEquals(lhs, rhs));
Assert(!TransactionIdFollowsOrEquals(rhs, lhs));
- Assert(!ctl->PagePrecedes(lhs / per_page, lhs / per_page));
- Assert(!ctl->PagePrecedes(lhs / per_page, rhs / per_page));
- Assert(!ctl->PagePrecedes(rhs / per_page, lhs / per_page));
- Assert(!ctl->PagePrecedes((lhs - per_page) / per_page, rhs / per_page));
- Assert(ctl->PagePrecedes(rhs / per_page, (lhs - 3 * per_page) / per_page));
- Assert(ctl->PagePrecedes(rhs / per_page, (lhs - 2 * per_page) / per_page));
- Assert(ctl->PagePrecedes(rhs / per_page, (lhs - 1 * per_page) / per_page)
+ Assert(!PagePrecedes(lhs / per_page, lhs / per_page));
+ Assert(!PagePrecedes(lhs / per_page, rhs / per_page));
+ Assert(!PagePrecedes(rhs / per_page, lhs / per_page));
+ Assert(!PagePrecedes((lhs - per_page) / per_page, rhs / per_page));
+ Assert(PagePrecedes(rhs / per_page, (lhs - 3 * per_page) / per_page));
+ Assert(PagePrecedes(rhs / per_page, (lhs - 2 * per_page) / per_page));
+ Assert(PagePrecedes(rhs / per_page, (lhs - 1 * per_page) / per_page)
|| (1U << 31) % per_page != 0); /* See CommitTsPagePrecedes() */
- Assert(ctl->PagePrecedes((lhs + 1 * per_page) / per_page, rhs / per_page)
+ Assert(PagePrecedes((lhs + 1 * per_page) / per_page, rhs / per_page)
|| (1U << 31) % per_page != 0);
- Assert(ctl->PagePrecedes((lhs + 2 * per_page) / per_page, rhs / per_page));
- Assert(ctl->PagePrecedes((lhs + 3 * per_page) / per_page, rhs / per_page));
- Assert(!ctl->PagePrecedes(rhs / per_page, (lhs + per_page) / per_page));
+ Assert(PagePrecedes((lhs + 2 * per_page) / per_page, rhs / per_page));
+ Assert(PagePrecedes((lhs + 3 * per_page) / per_page, rhs / per_page));
+ Assert(!PagePrecedes(rhs / per_page, (lhs + per_page) / per_page));
/*
* GetNewTransactionId() has assigned the last XID it can safely use, and
@@ -1451,7 +206,7 @@ SlruPagePrecedesTestOffset(SlruCtl ctl, int per_page, uint32 offset)
oldestXact = newestXact + 1;
oldestXact -= 1U << 31;
oldestPage = oldestXact / per_page;
- Assert(!SlruMayDeleteSegment(ctl,
+ Assert(!SlruMayDeleteSegment(PagePrecedes,
(newestPage -
newestPage % SLRU_PAGES_PER_SEGMENT),
oldestPage));
@@ -1467,7 +222,7 @@ SlruPagePrecedesTestOffset(SlruCtl ctl, int per_page, uint32 offset)
oldestXact = newestXact + 1;
oldestXact -= 1U << 31;
oldestPage = oldestXact / per_page;
- Assert(!SlruMayDeleteSegment(ctl,
+ Assert(!SlruMayDeleteSegment(PagePrecedes,
(newestPage -
newestPage % SLRU_PAGES_PER_SEGMENT),
oldestPage));
@@ -1483,12 +238,12 @@ SlruPagePrecedesTestOffset(SlruCtl ctl, int per_page, uint32 offset)
* do not apply to them.)
*/
void
-SlruPagePrecedesUnitTests(SlruCtl ctl, int per_page)
+SlruPagePrecedesUnitTests(SlruPagePrecedesFunction PagePrecedes, int per_page)
{
/* Test first, middle and last entries of a page. */
- SlruPagePrecedesTestOffset(ctl, per_page, 0);
- SlruPagePrecedesTestOffset(ctl, per_page, per_page / 2);
- SlruPagePrecedesTestOffset(ctl, per_page, per_page - 1);
+ SlruPagePrecedesTestOffset(PagePrecedes, per_page, 0);
+ SlruPagePrecedesTestOffset(PagePrecedes, per_page, per_page / 2);
+ SlruPagePrecedesTestOffset(PagePrecedes, per_page, per_page - 1);
}
#endif
@@ -1498,11 +253,12 @@ SlruPagePrecedesUnitTests(SlruCtl ctl, int per_page)
* one containing the page passed as "data".
*/
bool
-SlruScanDirCbReportPresence(SlruCtl ctl, char *filename, int segpage, void *data)
+SlruScanDirCbReportPresence(int slru_id, SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage, void *data)
{
int cutoffPage = *(int *) data;
- if (SlruMayDeleteSegment(ctl, segpage, cutoffPage))
+ if (SlruMayDeleteSegment(PagePrecedes, segpage, cutoffPage))
return true; /* found one; don't iterate any more */
return false; /* keep going */
@@ -1513,12 +269,15 @@ SlruScanDirCbReportPresence(SlruCtl ctl, char *filename, int segpage, void *data
* This callback deletes segments prior to the one passed in as "data".
*/
static bool
-SlruScanDirCbDeleteCutoff(SlruCtl ctl, char *filename, int segpage, void *data)
+SlruScanDirCbDeleteCutoff(int slru_id, SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage, void *data)
{
int cutoffPage = *(int *) data;
- if (SlruMayDeleteSegment(ctl, segpage, cutoffPage))
- SlruInternalDeleteSegment(ctl, segpage / SLRU_PAGES_PER_SEGMENT);
+ if (SlruMayDeleteSegment(PagePrecedes, segpage, cutoffPage))
+ {
+ SlruDeleteSegment(slru_id, segpage / SLRU_PAGES_PER_SEGMENT);
+ }
return false; /* keep going */
}
@@ -1528,9 +287,10 @@ SlruScanDirCbDeleteCutoff(SlruCtl ctl, char *filename, int segpage, void *data)
* This callback deletes all segments.
*/
bool
-SlruScanDirCbDeleteAll(SlruCtl ctl, char *filename, int segpage, void *data)
+SlruScanDirCbDeleteAll(int slru_id, SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage, void *data)
{
- SlruInternalDeleteSegment(ctl, segpage / SLRU_PAGES_PER_SEGMENT);
+ SlruInternalDeleteSegment(slru_id, segpage / SLRU_PAGES_PER_SEGMENT);
return false; /* keep going */
}
@@ -1551,16 +311,20 @@ SlruScanDirCbDeleteAll(SlruCtl ctl, char *filename, int segpage, void *data)
* Note that no locking is applied.
*/
bool
-SlruScanDirectory(SlruCtl ctl, SlruScanCallback callback, void *data)
+SlruScanDirectory(int slru_id, SlruPagePrecedesFunction PagePrecedes,
+ SlruScanCallback callback, void *data)
{
bool retval = false;
DIR *cldir;
struct dirent *clde;
int segno;
int segpage;
+ const char *path;
- cldir = AllocateDir(ctl->Dir);
- while ((clde = ReadDir(cldir, ctl->Dir)) != NULL)
+ path = slru_dirs[slru_id];
+
+ cldir = AllocateDir(path);
+ while ((clde = ReadDir(cldir, path)) != NULL)
{
size_t len;
@@ -1573,8 +337,8 @@ SlruScanDirectory(SlruCtl ctl, SlruScanCallback callback, void *data)
segpage = segno * SLRU_PAGES_PER_SEGMENT;
elog(DEBUG2, "SlruScanDirectory invoking callback on %s/%s",
- ctl->Dir, clde->d_name);
- retval = callback(ctl, clde->d_name, segpage, data);
+ path, clde->d_name);
+ retval = callback(slru_id, PagePrecedes, clde->d_name, segpage, data);
if (retval)
break;
}
@@ -1585,29 +349,78 @@ SlruScanDirectory(SlruCtl ctl, SlruScanCallback callback, void *data)
}
/*
- * Individual SLRUs (clog, ...) have to provide a sync.c handler function so
- * that they can provide the correct "SlruCtl" (otherwise we don't know how to
- * build the path), but they just forward to this common implementation that
- * performs the fsync.
+ * Read a buffer. Buffer is pinned on return.
*/
-int
-SlruSyncFileTag(SlruCtl ctl, const FileTag *ftag, char *path)
+Buffer
+ReadSlruBuffer(int slru_id, int pageno)
{
- int fd;
- int save_errno;
- int result;
+ int segno = pageno / SLRU_PAGES_PER_SEGMENT;
+ int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
+ RelFileLocator rlocator = SlruRelFileLocator(slru_id, segno);
+ Buffer buffer;
+ bool hit;
+
+ /* Try to avoid doing a buffer mapping table lookup for repeated access. */
+ buffer = slru_recent_buffers[slru_id].recent_buffer;
+ if (slru_recent_buffers[slru_id].pageno == pageno &&
+ BufferIsValid(buffer) &&
+ ReadRecentBuffer(rlocator, MAIN_FORKNUM, pageno, buffer))
+ {
+ pgstat_count_slru_page_hit(slru_id);
+ return buffer;
+ }
+
+ /* Regular lookup. */
+ buffer = ReadBufferWithoutRelcacheWithHit(rlocator, MAIN_FORKNUM, rpageno,
+ RBM_NORMAL, NULL, true, &hit);
- SlruFileName(ctl, path, ftag->segno);
+ /* Remember where this page is for next time. */
+ slru_recent_buffers[slru_id].pageno = pageno;
+ slru_recent_buffers[slru_id].recent_buffer = buffer;
- fd = OpenTransientFile(path, O_RDWR | PG_BINARY);
- if (fd < 0)
- return -1;
+ if (hit)
+ pgstat_count_slru_page_hit(slru_id);
- result = pg_fsync(fd);
- save_errno = errno;
+ return buffer;
+}
+
+/*
+ * Zero-initialize a buffer. Buffer is pinned and exclusively locked on return.
+ */
+Buffer
+ZeroSlruBuffer(int slru_id, int pageno)
+{
+ int segno = pageno / SLRU_PAGES_PER_SEGMENT;
+ int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
+ RelFileLocator rlocator = SlruRelFileLocator(slru_id, segno);
+ Buffer buffer;
+ SMgrFileHandle sfile;
- CloseTransientFile(fd);
+ if (rpageno == 0)
+ {
+ sfile = smgropen(rlocator, InvalidBackendId, MAIN_FORKNUM);
+ if (!smgrexists(sfile))
+ smgrcreate(sfile, false);
+ }
+
+ buffer = ReadBufferWithoutRelcache(rlocator, MAIN_FORKNUM, rpageno,
+ RBM_ZERO_AND_LOCK, NULL, true);
+
+ /* Remember where this page is for next time. */
+ slru_recent_buffers[slru_id].pageno = pageno;
+ slru_recent_buffers[slru_id].recent_buffer = buffer;
+
+ pgstat_count_slru_page_zeroed(slru_id);
+
+ return buffer;
+}
+
+bool
+ProbeSlruBuffer(int slru_id, int pageno)
+{
+ int segno = pageno / SLRU_PAGES_PER_SEGMENT;
+ int rpageno = pageno % SLRU_PAGES_PER_SEGMENT;
+ RelFileLocator rlocator = SlruRelFileLocator(slru_id, segno);
- errno = save_errno;
- return result;
+ return BufferProbe(rlocator, MAIN_FORKNUM, rpageno);
}
diff --git a/src/backend/access/transam/subtrans.c b/src/backend/access/transam/subtrans.c
index 62bb610167c..1ab4e5ae557 100644
--- a/src/backend/access/transam/subtrans.c
+++ b/src/backend/access/transam/subtrans.c
@@ -32,6 +32,7 @@
#include "access/subtrans.h"
#include "access/transam.h"
#include "pg_trace.h"
+#include "storage/bufmgr.h"
#include "utils/snapmgr.h"
@@ -55,15 +56,7 @@
#define TransactionIdToEntry(xid) ((xid) % (TransactionId) SUBTRANS_XACTS_PER_PAGE)
-/*
- * Link to shared-memory data structures for SUBTRANS control
- */
-static SlruCtlData SubTransCtlData;
-
-#define SubTransCtl (&SubTransCtlData)
-
-
-static int ZeroSUBTRANSPage(int pageno);
+static Buffer ZeroSUBTRANSPage(int pageno);
static bool SubTransPagePrecedes(int page1, int page2);
@@ -75,16 +68,15 @@ SubTransSetParent(TransactionId xid, TransactionId parent)
{
int pageno = TransactionIdToPage(xid);
int entryno = TransactionIdToEntry(xid);
- int slotno;
TransactionId *ptr;
+ Buffer buffer;
Assert(TransactionIdIsValid(parent));
Assert(TransactionIdFollows(xid, parent));
- LWLockAcquire(SubtransSLRULock, LW_EXCLUSIVE);
-
- slotno = SimpleLruReadPage(SubTransCtl, pageno, true, xid);
- ptr = (TransactionId *) SubTransCtl->shared->page_buffer[slotno];
+ buffer = ReadSlruBuffer(SLRU_SUBTRANS_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ ptr = (TransactionId *) BufferGetPage(buffer);
ptr += entryno;
/*
@@ -96,10 +88,10 @@ SubTransSetParent(TransactionId xid, TransactionId parent)
{
Assert(*ptr == InvalidTransactionId);
*ptr = parent;
- SubTransCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
}
- LWLockRelease(SubtransSLRULock);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -110,9 +102,9 @@ SubTransGetParent(TransactionId xid)
{
int pageno = TransactionIdToPage(xid);
int entryno = TransactionIdToEntry(xid);
- int slotno;
TransactionId *ptr;
TransactionId parent;
+ Buffer buffer;
/* Can't ask about stuff that might not be around anymore */
Assert(TransactionIdFollowsOrEquals(xid, TransactionXmin));
@@ -121,15 +113,14 @@ SubTransGetParent(TransactionId xid)
if (!TransactionIdIsNormal(xid))
return InvalidTransactionId;
- /* lock is acquired by SimpleLruReadPage_ReadOnly */
+ buffer = ReadSlruBuffer(SLRU_SUBTRANS_ID, pageno);
- slotno = SimpleLruReadPage_ReadOnly(SubTransCtl, pageno, xid);
- ptr = (TransactionId *) SubTransCtl->shared->page_buffer[slotno];
+ ptr = (TransactionId *) BufferGetPage(buffer);
ptr += entryno;
parent = *ptr;
- LWLockRelease(SubtransSLRULock);
+ ReleaseBuffer(buffer);
return parent;
}
@@ -177,26 +168,6 @@ SubTransGetTopmostTransaction(TransactionId xid)
return previousXid;
}
-
-/*
- * Initialization of shared memory for SUBTRANS
- */
-Size
-SUBTRANSShmemSize(void)
-{
- return SimpleLruShmemSize(NUM_SUBTRANS_BUFFERS, 0);
-}
-
-void
-SUBTRANSShmemInit(void)
-{
- SubTransCtl->PagePrecedes = SubTransPagePrecedes;
- SimpleLruInit(SubTransCtl, "Subtrans", NUM_SUBTRANS_BUFFERS, 0,
- SubtransSLRULock, "pg_subtrans",
- LWTRANCHE_SUBTRANS_BUFFER, SYNC_HANDLER_NONE);
- SlruPagePrecedesUnitTests(SubTransCtl, SUBTRANS_XACTS_PER_PAGE);
-}
-
/*
* This func must be called ONCE on system install. It creates
* the initial SUBTRANS segment. (The SUBTRANS directory is assumed to
@@ -210,18 +181,16 @@ SUBTRANSShmemInit(void)
void
BootStrapSUBTRANS(void)
{
- int slotno;
+ Buffer buffer;
- LWLockAcquire(SubtransSLRULock, LW_EXCLUSIVE);
+ SlruPagePrecedesUnitTests(SubTransPagePrecedes, SUBTRANS_XACTS_PER_PAGE);
/* Create and zero the first page of the subtrans log */
- slotno = ZeroSUBTRANSPage(0);
+ buffer = ZeroSUBTRANSPage(0);
/* Make sure it's written out */
- SimpleLruWritePage(SubTransCtl, slotno);
- Assert(!SubTransCtl->shared->page_dirty[slotno]);
-
- LWLockRelease(SubtransSLRULock);
+ FlushOneBuffer(buffer);
+ UnlockReleaseBuffer(buffer);
}
/*
@@ -232,10 +201,15 @@ BootStrapSUBTRANS(void)
*
* Control lock must be held at entry, and will be held at exit.
*/
-static int
+static Buffer
ZeroSUBTRANSPage(int pageno)
{
- return SimpleLruZeroPage(SubTransCtl, pageno);
+ Buffer buffer;
+
+ buffer = ZeroSlruBuffer(SLRU_SUBTRANS_ID, pageno);
+ MarkBufferDirty(buffer);
+
+ return buffer;
}
/*
@@ -258,7 +232,6 @@ StartupSUBTRANS(TransactionId oldestActiveXID)
* Whenever we advance into a new page, ExtendSUBTRANS will likewise zero
* the new page without regard to whatever was previously on disk.
*/
- LWLockAcquire(SubtransSLRULock, LW_EXCLUSIVE);
startPage = TransactionIdToPage(oldestActiveXID);
nextXid = ShmemVariableCache->nextXid;
@@ -266,36 +239,15 @@ StartupSUBTRANS(TransactionId oldestActiveXID)
while (startPage != endPage)
{
- (void) ZeroSUBTRANSPage(startPage);
+ UnlockReleaseBuffer(ZeroSUBTRANSPage(startPage));
startPage++;
/* must account for wraparound */
if (startPage > TransactionIdToPage(MaxTransactionId))
startPage = 0;
}
- (void) ZeroSUBTRANSPage(startPage);
-
- LWLockRelease(SubtransSLRULock);
+ UnlockReleaseBuffer(ZeroSUBTRANSPage(startPage));
}
-/*
- * Perform a checkpoint --- either during shutdown, or on-the-fly
- */
-void
-CheckPointSUBTRANS(void)
-{
- /*
- * Write dirty SUBTRANS pages to disk
- *
- * This is not actually necessary from a correctness point of view. We do
- * it merely to improve the odds that writing of dirty pages is done by
- * the checkpoint process and not by backends.
- */
- TRACE_POSTGRESQL_SUBTRANS_CHECKPOINT_START(true);
- SimpleLruWriteAll(SubTransCtl, true);
- TRACE_POSTGRESQL_SUBTRANS_CHECKPOINT_DONE(true);
-}
-
-
/*
* Make sure that SUBTRANS has room for a newly-allocated XID.
*
@@ -319,12 +271,8 @@ ExtendSUBTRANS(TransactionId newestXact)
pageno = TransactionIdToPage(newestXact);
- LWLockAcquire(SubtransSLRULock, LW_EXCLUSIVE);
-
/* Zero the page */
- ZeroSUBTRANSPage(pageno);
-
- LWLockRelease(SubtransSLRULock);
+ UnlockReleaseBuffer(ZeroSUBTRANSPage(pageno));
}
@@ -350,7 +298,7 @@ TruncateSUBTRANS(TransactionId oldestXact)
TransactionIdRetreat(oldestXact);
cutoffPage = TransactionIdToPage(oldestXact);
- SimpleLruTruncate(SubTransCtl, cutoffPage);
+ SimpleLruTruncate(SLRU_SUBTRANS_ID, SubTransPagePrecedes, cutoffPage);
}
diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index b8764012607..7d4800a5f24 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -68,6 +68,7 @@
#include "utils/inval.h"
#include "utils/memutils.h"
#include "utils/relmapper.h"
+#include "utils/resowner_private.h"
#include "utils/snapmgr.h"
#include "utils/timeout.h"
#include "utils/timestamp.h"
@@ -1397,6 +1398,7 @@ RecordTransactionCommit(void)
* are delaying the checkpoint a bit fuzzy, but it doesn't matter.
*/
Assert((MyProc->delayChkptFlags & DELAY_CHKPT_START) == 0);
+
START_CRIT_SECTION();
MyProc->delayChkptFlags |= DELAY_CHKPT_START;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f9f0f6db8d1..68917d17299 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -4638,6 +4638,7 @@ BootStrapXLOG(void)
uint64 sysidentifier;
struct timeval tv;
pg_crc32c crc;
+ ResourceOwner resowner;
/* allow ordinary WAL segment creation, like StartupXLOG() would */
SetInstallXLogFileSegmentActive();
@@ -4777,10 +4778,14 @@ BootStrapXLOG(void)
WriteControlFile();
/* Bootstrap the commit log, too */
+ resowner = ResourceOwnerCreate(NULL, "bootstrap resowner");
+ CurrentResourceOwner = resowner;
BootStrapCLOG();
BootStrapCommitTs();
BootStrapSUBTRANS();
BootStrapMultiXact();
+ CurrentResourceOwner = NULL;
+ ResourceOwnerDelete(resowner);
pfree(buffer);
@@ -4789,6 +4794,8 @@ BootStrapXLOG(void)
* otherwise never run the checks and GUC related initializations therein.
*/
ReadControlFile();
+
+ smgrcloseall();
}
static char *
@@ -6997,15 +7004,11 @@ CheckPointGuts(XLogRecPtr checkPointRedo, int flags)
CheckPointSnapBuild();
CheckPointLogicalRewriteHeap();
CheckPointReplicationOrigin();
+ CheckPointPredicate();
- /* Write out all dirty data in SLRUs and the main buffer pool */
+ /* Write out all dirty data in the buffer pool */
TRACE_POSTGRESQL_BUFFER_CHECKPOINT_START(flags);
CheckpointStats.ckpt_write_t = GetCurrentTimestamp();
- CheckPointCLOG();
- CheckPointCommitTs();
- CheckPointSUBTRANS();
- CheckPointMultiXact();
- CheckPointPredicate();
CheckPointBuffers(flags);
/* Perform all queued up fsyncs */
diff --git a/src/backend/commands/async.c b/src/backend/commands/async.c
index ef909cf4e08..f944766ec2b 100644
--- a/src/backend/commands/async.c
+++ b/src/backend/commands/async.c
@@ -141,6 +141,7 @@
#include "libpq/libpq.h"
#include "libpq/pqformat.h"
#include "miscadmin.h"
+#include "storage/bufmgr.h"
#include "storage/ipc.h"
#include "storage/lmgr.h"
#include "storage/proc.h"
@@ -305,12 +306,6 @@ static AsyncQueueControl *asyncQueueControl;
#define QUEUE_NEXT_LISTENER(i) (asyncQueueControl->backend[i].nextListener)
#define QUEUE_BACKEND_POS(i) (asyncQueueControl->backend[i].pos)
-/*
- * The SLRU buffer area through which we access the notification queue
- */
-static SlruCtlData NotifyCtlData;
-
-#define NotifyCtl (&NotifyCtlData)
#define QUEUE_PAGESIZE BLCKSZ
#define QUEUE_FULL_WARN_INTERVAL 5000 /* warn at most once every 5s */
@@ -521,8 +516,6 @@ AsyncShmemSize(void)
size = mul_size(MaxBackends + 1, sizeof(QueueBackendStatus));
size = add_size(size, offsetof(AsyncQueueControl, backend));
- size = add_size(size, SimpleLruShmemSize(NUM_NOTIFY_BUFFERS, 0));
-
return size;
}
@@ -565,20 +558,13 @@ AsyncShmemInit(void)
}
}
- /*
- * Set up SLRU management of the pg_notify data.
- */
- NotifyCtl->PagePrecedes = asyncQueuePagePrecedes;
- SimpleLruInit(NotifyCtl, "Notify", NUM_NOTIFY_BUFFERS, 0,
- NotifySLRULock, "pg_notify", LWTRANCHE_NOTIFY_BUFFER,
- SYNC_HANDLER_NONE);
-
if (!found)
{
/*
* During start or reboot, clean out the pg_notify directory.
*/
- (void) SlruScanDirectory(NotifyCtl, SlruScanDirCbDeleteAll, NULL);
+ (void) SlruScanDirectory(SLRU_NOTIFY_ID, asyncQueuePagePrecedes,
+ SlruScanDirCbDeleteAll, NULL);
}
}
@@ -1411,10 +1397,7 @@ asyncQueueAddEntries(ListCell *nextNotify)
QueuePosition queue_head;
int pageno;
int offset;
- int slotno;
-
- /* We hold both NotifyQueueLock and NotifySLRULock during this operation */
- LWLockAcquire(NotifySLRULock, LW_EXCLUSIVE);
+ Buffer buffer;
/*
* We work with a local copy of QUEUE_HEAD, which we write back to shared
@@ -1439,13 +1422,17 @@ asyncQueueAddEntries(ListCell *nextNotify)
*/
pageno = QUEUE_POS_PAGE(queue_head);
if (QUEUE_POS_IS_ZERO(queue_head))
- slotno = SimpleLruZeroPage(NotifyCtl, pageno);
+ {
+ buffer = ZeroSlruBuffer(SLRU_NOTIFY_ID, pageno);
+ }
else
- slotno = SimpleLruReadPage(NotifyCtl, pageno, true,
- InvalidTransactionId);
+ {
+ buffer = ReadSlruBuffer(SLRU_NOTIFY_ID, pageno);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ }
/* Note we mark the page dirty before writing in it */
- NotifyCtl->shared->page_dirty[slotno] = true;
+ MarkBufferDirty(buffer);
while (nextNotify != NULL)
{
@@ -1476,7 +1463,7 @@ asyncQueueAddEntries(ListCell *nextNotify)
}
/* Now copy qe into the shared buffer page */
- memcpy(NotifyCtl->shared->page_buffer[slotno] + offset,
+ memcpy(BufferGetPage(buffer) + offset,
&qe,
qe.length);
@@ -1491,7 +1478,10 @@ asyncQueueAddEntries(ListCell *nextNotify)
* asyncQueueIsFull() ensured that there is room to create this
* page without overrunning the queue.
*/
- slotno = SimpleLruZeroPage(NotifyCtl, QUEUE_POS_PAGE(queue_head));
+ UnlockReleaseBuffer(buffer);
+ buffer = ZeroSlruBuffer(SLRU_NOTIFY_ID,
+ QUEUE_POS_PAGE(queue_head));
+ MarkBufferDirty(buffer);
/*
* If the new page address is a multiple of QUEUE_CLEANUP_DELAY,
@@ -1505,12 +1495,11 @@ asyncQueueAddEntries(ListCell *nextNotify)
break;
}
}
+ UnlockReleaseBuffer(buffer);
/* Success, so update the global QUEUE_HEAD */
QUEUE_HEAD = queue_head;
- LWLockRelease(NotifySLRULock);
-
return nextNotify;
}
@@ -1983,17 +1972,16 @@ asyncQueueReadAllNotifications(void)
{
int curpage = QUEUE_POS_PAGE(pos);
int curoffset = QUEUE_POS_OFFSET(pos);
- int slotno;
int copysize;
+ Buffer buffer;
/*
- * We copy the data from SLRU into a local buffer, so as to avoid
- * holding the NotifySLRULock while we are examining the entries
- * and possibly transmitting them to our frontend. Copy only the
- * part of the page we will actually inspect.
+ * We copy the data into a local buffer, so as to avoid holding a
+ * buffer pin while we are examining the entries and possibly
+ * transmitting them to our frontend. Copy only the part of the
+ * page we will actually inspect.
*/
- slotno = SimpleLruReadPage_ReadOnly(NotifyCtl, curpage,
- InvalidTransactionId);
+ buffer = ReadSlruBuffer(SLRU_NOTIFY_ID, curpage);
if (curpage == QUEUE_POS_PAGE(head))
{
/* we only want to read as far as head */
@@ -2007,10 +1995,9 @@ asyncQueueReadAllNotifications(void)
copysize = QUEUE_PAGESIZE - curoffset;
}
memcpy(page_buffer.buf + curoffset,
- NotifyCtl->shared->page_buffer[slotno] + curoffset,
+ BufferGetPage(buffer) + curoffset,
copysize);
- /* Release lock that we got from SimpleLruReadPage_ReadOnly() */
- LWLockRelease(NotifySLRULock);
+ ReleaseBuffer(buffer);
/*
* Process messages up to the stop position, end of page, or an
@@ -2207,7 +2194,7 @@ asyncQueueAdvanceTail(void)
* SimpleLruTruncate() will ask for NotifySLRULock but will also
* release the lock again.
*/
- SimpleLruTruncate(NotifyCtl, newtailpage);
+ SimpleLruTruncate(SLRU_NOTIFY_ID, asyncQueuePagePrecedes, newtailpage);
/*
* Update QUEUE_STOP_PAGE. This changes asyncQueueIsFull()'s verdict
diff --git a/src/backend/storage/buffer/buf_init.c b/src/backend/storage/buffer/buf_init.c
index 20946c47cb4..9746838bf9f 100644
--- a/src/backend/storage/buffer/buf_init.c
+++ b/src/backend/storage/buffer/buf_init.c
@@ -20,6 +20,7 @@
BufferDescPadded *BufferDescriptors;
char *BufferBlocks;
+XLogRecPtr *BufferExternalLSNs;
ConditionVariableMinimallyPadded *BufferIOCVArray;
WritebackContext BackendWritebackContext;
CkptSortItem *CkptBufferIds;
@@ -69,9 +70,11 @@ InitBufferPool(void)
{
bool foundBufs,
foundDescs,
+ foundLSNs,
foundIOCV,
foundBufCkpt;
+
/* Align descriptors to a cacheline boundary. */
BufferDescriptors = (BufferDescPadded *)
ShmemInitStruct("Buffer Descriptors",
@@ -88,6 +91,11 @@ InitBufferPool(void)
NBuffers * sizeof(ConditionVariableMinimallyPadded),
&foundIOCV);
+ BufferExternalLSNs = (XLogRecPtr *)
+ ShmemInitStruct("Buffer External LSNs",
+ NBuffers * sizeof(XLogRecPtr),
+ &foundLSNs);
+
/*
* The array used to sort to-be-checkpointed buffer ids is located in
* shared memory, to avoid having to allocate significant amounts of
@@ -99,10 +107,10 @@ InitBufferPool(void)
ShmemInitStruct("Checkpoint BufferIds",
NBuffers * sizeof(CkptSortItem), &foundBufCkpt);
- if (foundDescs || foundBufs || foundIOCV || foundBufCkpt)
+ if (foundDescs || foundBufs || foundIOCV || foundBufCkpt || foundLSNs)
{
/* should find all of these, or none of them */
- Assert(foundDescs && foundBufs && foundIOCV && foundBufCkpt);
+ Assert(foundDescs && foundBufs && foundIOCV && foundBufCkpt && foundLSNs);
/* note: this path is only taken in EXEC_BACKEND case */
}
else
@@ -133,6 +141,8 @@ InitBufferPool(void)
LWTRANCHE_BUFFER_CONTENT);
ConditionVariableInit(BufferDescriptorGetIOCV(buf));
+
+ BufferExternalLSNs[i] = InvalidXLogRecPtr;
}
/* Correct last entry of linked list */
@@ -166,6 +176,9 @@ BufferShmemSize(void)
/* size of data pages */
size = add_size(size, mul_size(NBuffers, BLCKSZ));
+ /* size of external LSNs */
+ size = add_size(size, mul_size(NBuffers, sizeof(XLogRecPtr)));
+
/* size of stuff controlled by freelist.c */
size = add_size(size, StrategyShmemSize());
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 1305eb7dee1..b3f7be2e05f 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -57,10 +57,21 @@
#include "utils/resowner_private.h"
#include "utils/timestamp.h"
+/*
+ * XXX Ideally we'd switch to standard pages for SLRU data, but in the
+ * meantime we need some way to identify buffers that hold raw data (no
+ * invasive LSN, no checksums).
+ */
+#define BufferHasStandardPage(bufHdr) \
+ ((bufHdr)->tag.spcOid != 9)
+
+#define BufferHasExternalLSN(bufHdr) \
+ !BufferHasStandardPage(bufHdr)
/* Note: these two macros only work on shared buffers, not local ones! */
#define BufHdrGetBlock(bufHdr) ((Block) (BufferBlocks + ((Size) (bufHdr)->buf_id) * BLCKSZ))
-#define BufferGetLSN(bufHdr) (PageGetLSN(BufHdrGetBlock(bufHdr)))
+#define BufferGetLSN(bufHdr) \
+ (BufferHasExternalLSN(bufHdr) ? BufferGetExternalLSN(bufHdr) : PageGetLSN(BufHdrGetBlock(bufHdr)))
/* Note: this macro only works on local buffers, not shared ones! */
#define LocalBufHdrGetBlock(bufHdr) \
@@ -786,6 +797,18 @@ ReadBufferWithoutRelcache(RelFileLocator rlocator, ForkNumber forkNum,
mode, strategy, &hit);
}
+Buffer
+ReadBufferWithoutRelcacheWithHit(RelFileLocator rlocator, ForkNumber forkNum,
+ BlockNumber blockNum, ReadBufferMode mode,
+ BufferAccessStrategy strategy, bool permanent, bool *hit)
+{
+ SMgrFileHandle sfile = smgropen(rlocator, InvalidBackendId, forkNum);
+
+ return ReadBuffer_common(sfile, permanent ? RELPERSISTENCE_PERMANENT :
+ RELPERSISTENCE_UNLOGGED, blockNum,
+ mode, strategy, hit);
+}
+
/*
* ReadBuffer_common -- common logic for all ReadBuffer variants
@@ -1032,7 +1055,8 @@ ReadBuffer_common(SMgrFileHandle sfile, char relpersistence,
}
/* check for garbage data */
- if (!PageIsVerifiedExtended((Page) bufBlock, blockNum,
+ if (BufferHasStandardPage(bufHdr) &&
+ !PageIsVerifiedExtended((Page) bufBlock, blockNum,
PIV_LOG_WARNING | PIV_REPORT_STAT))
{
if (mode == RBM_ZERO_ON_ERROR || zero_damaged_pages)
@@ -1433,6 +1457,9 @@ BufferAlloc(SMgrFileHandle sfile, char relpersistence,
UnpinBuffer(buf);
}
+ if (BufferHasExternalLSN(buf))
+ BufferSetExternalLSN(buf, InvalidXLogRecPtr);
+
/*
* Okay, it's finally safe to rename the buffer.
*
@@ -3087,7 +3114,10 @@ BufferGetLSNAtomic(Buffer buffer)
Assert(BufferIsPinned(buffer));
buf_state = LockBufHdr(bufHdr);
- lsn = PageGetLSN(page);
+ if (BufferHasStandardPage(bufHdr))
+ lsn = PageGetLSN(page);
+ else
+ lsn = BufferGetExternalLSN(bufHdr);
UnlockBufHdr(bufHdr, buf_state);
return lsn;
@@ -5068,3 +5098,29 @@ TestForOldSnapshot_impl(Snapshot snapshot, Relation relation)
(errcode(ERRCODE_SNAPSHOT_TOO_OLD),
errmsg("snapshot too old")));
}
+
+/*
+ * Check if a buffer tag is currently mapped.
+ *
+ * XXX Dubious semantics; needed only for multixact's handling for
+ * inconsistent states.
+ */
+bool
+BufferProbe(RelFileLocator rlocator, ForkNumber forkNum, BlockNumber blockNum)
+{
+ BufferTag tag;
+ uint32 hash;
+ LWLock *partitionLock;
+ int buf_id;
+
+ InitBufferTag(&tag, &rlocator, forkNum, blockNum);
+
+ hash = BufTableHashCode(&tag);
+ partitionLock = BufMappingPartitionLock(hash);
+
+ LWLockAcquire(partitionLock, LW_SHARED);
+ buf_id = BufTableLookup(&tag, hash);
+ LWLockRelease(partitionLock);
+
+ return buf_id >= 0;
+}
diff --git a/src/backend/storage/ipc/ipci.c b/src/backend/storage/ipc/ipci.c
index 8f1ded7338f..8601e1c0dfb 100644
--- a/src/backend/storage/ipc/ipci.c
+++ b/src/backend/storage/ipc/ipci.c
@@ -117,9 +117,7 @@ CalculateShmemSize(int *num_semaphores)
size = add_size(size, XLogPrefetchShmemSize());
size = add_size(size, XLOGShmemSize());
size = add_size(size, XLogRecoveryShmemSize());
- size = add_size(size, CLOGShmemSize());
size = add_size(size, CommitTsShmemSize());
- size = add_size(size, SUBTRANSShmemSize());
size = add_size(size, TwoPhaseShmemSize());
size = add_size(size, BackgroundWorkerShmemSize());
size = add_size(size, MultiXactShmemSize());
@@ -241,9 +239,7 @@ CreateSharedMemoryAndSemaphores(void)
XLOGShmemInit();
XLogPrefetchShmemInit();
XLogRecoveryShmemInit();
- CLOGShmemInit();
CommitTsShmemInit();
- SUBTRANSShmemInit();
MultiXactShmemInit();
InitBufferPool();
diff --git a/src/backend/storage/lmgr/predicate.c b/src/backend/storage/lmgr/predicate.c
index bfc352aed86..f72fc99762c 100644
--- a/src/backend/storage/lmgr/predicate.c
+++ b/src/backend/storage/lmgr/predicate.c
@@ -313,14 +313,6 @@
((targethash) ^ ((uint32) PointerGetDatum((predicatelocktag)->myXact)) \
<< LOG2_NUM_PREDICATELOCK_PARTITIONS)
-
-/*
- * The SLRU buffer area through which we access the old xids.
- */
-static SlruCtlData SerialSlruCtlData;
-
-#define SerialSlruCtl (&SerialSlruCtlData)
-
#define SERIAL_PAGESIZE BLCKSZ
#define SERIAL_ENTRYSIZE sizeof(SerCommitSeqNo)
#define SERIAL_ENTRIESPERPAGE (SERIAL_PAGESIZE / SERIAL_ENTRYSIZE)
@@ -332,8 +324,8 @@ static SlruCtlData SerialSlruCtlData;
#define SerialNextPage(page) (((page) >= SERIAL_MAX_PAGE) ? 0 : (page) + 1)
-#define SerialValue(slotno, xid) (*((SerCommitSeqNo *) \
- (SerialSlruCtl->shared->page_buffer[slotno] + \
+#define SerialValue(buffer, xid) (*((SerCommitSeqNo *) \
+ (BufferGetPage(buffer) + \
((((uint32) (xid)) % SERIAL_ENTRIESPERPAGE) * SERIAL_ENTRYSIZE))))
#define SerialPage(xid) (((uint32) (xid)) / SERIAL_ENTRIESPERPAGE)
@@ -803,17 +795,10 @@ SerialInit(void)
{
bool found;
- /*
- * Set up SLRU management of the pg_serial data.
- */
- SerialSlruCtl->PagePrecedes = SerialPagePrecedesLogically;
- SimpleLruInit(SerialSlruCtl, "Serial",
- NUM_SERIAL_BUFFERS, 0, SerialSLRULock, "pg_serial",
- LWTRANCHE_SERIAL_BUFFER, SYNC_HANDLER_NONE);
#ifdef USE_ASSERT_CHECKING
SerialPagePrecedesLogicallyUnitTests();
#endif
- SlruPagePrecedesUnitTests(SerialSlruCtl, SERIAL_ENTRIESPERPAGE);
+ SlruPagePrecedesUnitTests(SerialPagePrecedesLogically, SERIAL_ENTRIESPERPAGE);
/*
* Create or attach to the SerialControl structure.
@@ -843,9 +828,9 @@ SerialAdd(TransactionId xid, SerCommitSeqNo minConflictCommitSeqNo)
{
TransactionId tailXid;
int targetPage;
- int slotno;
int firstZeroPage;
bool isNewPage;
+ Buffer buffer;
Assert(TransactionIdIsValid(xid));
@@ -890,16 +875,22 @@ SerialAdd(TransactionId xid, SerCommitSeqNo minConflictCommitSeqNo)
/* Initialize intervening pages. */
while (firstZeroPage != targetPage)
{
- (void) SimpleLruZeroPage(SerialSlruCtl, firstZeroPage);
+ buffer = ZeroSlruBuffer(SLRU_SERIAL_ID, firstZeroPage);
+ MarkBufferDirty(buffer);
+ UnlockReleaseBuffer(buffer);
firstZeroPage = SerialNextPage(firstZeroPage);
}
- slotno = SimpleLruZeroPage(SerialSlruCtl, targetPage);
+ buffer = ZeroSlruBuffer(SLRU_SERIAL_ID, targetPage);
}
else
- slotno = SimpleLruReadPage(SerialSlruCtl, targetPage, true, xid);
+ {
+ buffer = ReadSlruBuffer(SLRU_SERIAL_ID, targetPage);
+ LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+ }
- SerialValue(slotno, xid) = minConflictCommitSeqNo;
- SerialSlruCtl->shared->page_dirty[slotno] = true;
+ SerialValue(buffer, xid) = minConflictCommitSeqNo;
+ MarkBufferDirty(buffer);
+ UnlockReleaseBuffer(buffer);
LWLockRelease(SerialSLRULock);
}
@@ -915,7 +906,7 @@ SerialGetMinConflictCommitSeqNo(TransactionId xid)
TransactionId headXid;
TransactionId tailXid;
SerCommitSeqNo val;
- int slotno;
+ Buffer buffer;
Assert(TransactionIdIsValid(xid));
@@ -937,9 +928,9 @@ SerialGetMinConflictCommitSeqNo(TransactionId xid)
* The following function must be called without holding SerialSLRULock,
* but will return with that lock held, which must then be released.
*/
- slotno = SimpleLruReadPage_ReadOnly(SerialSlruCtl,
- SerialPage(xid), xid);
- val = SerialValue(slotno, xid);
+ buffer = ReadSlruBuffer(SLRU_SERIAL_ID, SerialPage(xid));
+ val = SerialValue(buffer, xid);
+ ReleaseBuffer(buffer);
LWLockRelease(SerialSLRULock);
return val;
}
@@ -1058,19 +1049,7 @@ CheckPointPredicate(void)
LWLockRelease(SerialSLRULock);
/* Truncate away pages that are no longer required */
- SimpleLruTruncate(SerialSlruCtl, tailPage);
-
- /*
- * Write dirty SLRU pages to disk
- *
- * This is not actually necessary from a correctness point of view. We do
- * it merely as a debugging aid.
- *
- * We're doing this after the truncation to avoid writing pages right
- * before deleting the file in which they sit, which would be completely
- * pointless.
- */
- SimpleLruWriteAll(SerialSlruCtl, true);
+ SimpleLruTruncate(SLRU_SERIAL_ID, SerialPagePrecedesLogically, tailPage);
}
/*------------------------------------------------------------------------*/
@@ -1331,7 +1310,6 @@ PredicateLockShmemSize(void)
/* Shared memory structures for SLRU tracking of old committed xids. */
size = add_size(size, sizeof(SerialControlData));
- size = add_size(size, SimpleLruShmemSize(NUM_SERIAL_BUFFERS, 0));
return size;
}
diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c
index eea7ce944c3..2cfee0deaad 100644
--- a/src/backend/storage/smgr/md.c
+++ b/src/backend/storage/smgr/md.c
@@ -191,9 +191,12 @@ mdcreate(SMgrFileHandle sfile, bool isRedo)
* should be here and not in commands/tablespace.c? But that would imply
* importing a lot of stuff that smgr.c oughtn't know, either.
*/
- TablespaceCreateDbspace(sfile->smgr_locator.locator.spcOid,
- sfile->smgr_locator.locator.dbOid,
- isRedo);
+ if (sfile->smgr_locator.locator.spcOid != SLRU_SPC_OID)
+ {
+ TablespaceCreateDbspace(sfile->smgr_locator.locator.spcOid,
+ sfile->smgr_locator.locator.dbOid,
+ isRedo);
+ }
path = smgrfilepath(sfile->smgr_locator);
diff --git a/src/backend/storage/smgr/smgr.c b/src/backend/storage/smgr/smgr.c
index 58a2322d018..d47695b808b 100644
--- a/src/backend/storage/smgr/smgr.c
+++ b/src/backend/storage/smgr/smgr.c
@@ -420,6 +420,48 @@ smgrunlink_multi(RelFileLocator rlocator, BackendId backend,
}
}
+/*
+ * smgrdounlink() -- Immediately unlink a file
+ *
+ * If isRedo is true, it is okay for the underlying file(s) to be gone
+ * already.
+ *
+ * To remove a relation transactionally, see RelationDropStorage() instead.
+ * This will cause cache invalidation of all forks of the relation, not just
+ * this one.
+ */
+void
+smgrunlink(SMgrFileHandle sfile, bool isRedo)
+{
+ SMgrFileLocator locator;
+ int which;
+
+ /* remember before closing */
+ which = sfile->smgr_which;
+ locator = sfile->smgr_locator;
+
+ /* Close the file at smgr level */
+ smgrclose(sfile);
+
+ /*
+ * Send a shared-inval message to force other backends to close any
+ * dangling smgr references they may have for these rels. We should do
+ * this before starting the actual unlinking, in case we fail partway
+ * through that step. Note that the sinval messages will eventually come
+ * back to this backend, too, and thereby provide a backstop that we
+ * closed our own smgr rel.
+ */
+ CacheInvalidateSmgr(locator.locator, locator.backend);
+
+ /*
+ * Delete the physical file(s).
+ *
+ * Note: smgr_unlink must treat deletion failure as a WARNING, not an
+ * ERROR, because we've already decided to commit or abort the current
+ * xact.
+ */
+ smgrsw[which].smgr_unlink(locator, isRedo);
+}
/*
* smgrextend() -- Add a new block to a file.
diff --git a/src/backend/storage/sync/sync.c b/src/backend/storage/sync/sync.c
index 768d1dbfc41..41886288644 100644
--- a/src/backend/storage/sync/sync.c
+++ b/src/backend/storage/sync/sync.c
@@ -18,9 +18,7 @@
#include <fcntl.h>
#include <sys/file.h>
-#include "access/commit_ts.h"
-#include "access/clog.h"
-#include "access/multixact.h"
+#include "access/slru.h"
#include "access/xlog.h"
#include "access/xlogutils.h"
#include "commands/tablespace.h"
@@ -106,22 +104,6 @@ static const SyncOps syncsw[] = {
.sync_unlinkfiletag = mdunlinkfiletag,
.sync_filetagmatches = mdfiletagmatches
},
- /* pg_xact */
- [SYNC_HANDLER_CLOG] = {
- .sync_syncfiletag = clogsyncfiletag
- },
- /* pg_commit_ts */
- [SYNC_HANDLER_COMMIT_TS] = {
- .sync_syncfiletag = committssyncfiletag
- },
- /* pg_multixact/offsets */
- [SYNC_HANDLER_MULTIXACT_OFFSET] = {
- .sync_syncfiletag = multixactoffsetssyncfiletag
- },
- /* pg_multixact/members */
- [SYNC_HANDLER_MULTIXACT_MEMBER] = {
- .sync_syncfiletag = multixactmemberssyncfiletag
- }
};
/*
diff --git a/src/backend/utils/mmgr/mcxt.c b/src/backend/utils/mmgr/mcxt.c
index 0b00802df70..4f9fa85d51a 100644
--- a/src/backend/utils/mmgr/mcxt.c
+++ b/src/backend/utils/mmgr/mcxt.c
@@ -162,9 +162,47 @@ static void MemoryContextStatsPrint(MemoryContext context, void *passthru,
* You should not do memory allocations within a critical section, because
* an out-of-memory error will be escalated to a PANIC. To enforce that
* rule, the allocation functions Assert that.
+ *
+ * FIXME: bypass this for the critical section in RecordTransactionCommit()
+ * for now. It does a lot of things that can allocate:
+ * - calls TransactionIdCommitTree, which pins buffers, which requires
+ * space in the ResourceOwner for the pin (ResourceOwnerEnlargeBuffers())
+ * - same for TransactionTreeSetCommitTsData() call.
+ * - reading a page can require flushing other pages, which in turn
+ * can call CompactCheckpointerRequestQueue(), which allocates
+ * - reading a page calls smgropen(), which allocates the SMgrFile entry
+ * if it's not open already
+ *
+ * FIXME: Here's another codepath that reaches this, reproducable with
+ * the 'lock-committed-update' isolation test:
+ *
+ * #5 0x000056230e91788d in MemoryContextAllocExtended (context=0x562310709c40, size=4048, flags=2) at mcxt.c:1137
+ * #6 0x000056230e8e9655 in DynaHashAlloc (size=4048) at dynahash.c:292
+ * #7 0x000056230e8ebadf in element_alloc (hashp=0x562310709d58, nelem=46, freelist_idx=0) at dynahash.c:1715
+ * #8 0x000056230e8eaef8 in get_hash_entry (hashp=0x562310709d58, freelist_idx=0) at dynahash.c:1324
+ * #9 0x000056230e8ea993 in hash_search_with_hash_value (hashp=0x562310709d58, keyPtr=0x7ffc30cdd4f0, hashvalue=1219519527, action=HASH_ENTER, foundPtr=0x7ffc30cdd4ef) at dynahash.c:1097
+ * #10 0x000056230e8ea578 in hash_search (hashp=0x562310709d58, keyPtr=0x7ffc30cdd4f0, action=HASH_ENTER, foundPtr=0x7ffc30cdd4ef) at dynahash.c:958
+ * #11 0x000056230e70f8fa in smgropen (rlocator=..., backend=-1, forkNum=MAIN_FORKNUM) at smgr.c:165
+ * #12 0x000056230e6c7f58 in ReadBufferWithoutRelcacheWithHit (rlocator=..., forkNum=MAIN_FORKNUM, blockNum=0, mode=RBM_NORMAL, strategy=0x0, permanent=true, hit=0x7ffc30cdd597)
+ * at bufmgr.c:805
+ * #13 0x000056230e2b45ce in ReadSlruBuffer (slru_id=3, pageno=0) at slru.c:377
+ * #14 0x000056230e2ad192 in RecordNewMultiXact (multi=5, offset=9, nmembers=2, members=0x7ffc30cdd690) at multixact.c:902
+ * #15 0x000056230e2acfbb in MultiXactIdCreateFromMembers (nmembers=2, members=0x7ffc30cdd690) at multixact.c:833
+ * #16 0x000056230e2ac8d3 in MultiXactIdCreate (xid1=753, status1=MultiXactStatusForKeyShare, xid2=754, status2=MultiXactStatusNoKeyUpdate) at multixact.c:402
+ * #17 0x000056230e248ff2 in compute_new_xmax_infomask (xmax=753, old_infomask=402, old_infomask2=2, add_to_xmax=754, mode=LockTupleNoKeyExclusive, is_update=true, result_xmax=0x7ffc30cdd79c,
+ * result_infomask=0x7ffc30cdd79a, result_infomask2=0x7ffc30cdd798) at heapam.c:5017
+ * #18 0x000056230e24632c in heap_update (relation=0x7f99454cb168, otid=0x7ffc30cddaba, newtup=0x56231073e840, cid=0, crosscheck=0x0, wait=true, tmfd=0x7ffc30cddaf0, lockmode=0x7ffc30cdda34)
+ * at heapam.c:3345
+ *
+ * Disabled this completely because of that.
*/
+#if 0
#define AssertNotInCriticalSection(context) \
- Assert(CritSectionCount == 0 || (context)->allowInCritSection)
+ Assert(CritSectionCount == 0 || (context)->allowInCritSection || \
+ (MyProc != NULL && (MyProc->delayChkptFlags & DELAY_CHKPT_START != 0)))
+#else
+#define AssertNotInCriticalSection(context) ((void)true)
+#endif
/*
* Call the given function in the MemoryContextMethods for the memory context
diff --git a/src/common/relpath.c b/src/common/relpath.c
index ae2d384fb34..4715d06d287 100644
--- a/src/common/relpath.c
+++ b/src/common/relpath.c
@@ -22,6 +22,16 @@
#include "common/relpath.h"
#include "storage/backendid.h"
+/*
+ * SLRU ID to path mapping
+ */
+#define PG_SLRU(symname,name,path,synchronize) \
+ path,
+
+static char *slru_dirs[] =
+{
+#include "access/slrulist.h"
+};
/*
* Lookup table of fork name by fork number.
@@ -143,7 +153,22 @@ GetSMgrFilePath(Oid dbOid, Oid spcOid, RelFileNumber relNumber,
{
char *path;
- if (spcOid == GLOBALTABLESPACE_OID)
+ if (spcOid == SLRU_SPC_OID)
+ {
+ if (dbOid >= lengthof(slru_dirs) || forkNumber != 0 || backendId != InvalidBackendId)
+ {
+#ifndef FRONTEND
+ ereport(ERROR,
+ (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
+ errmsg("invalid SLRU file locator %u/%u/%u/%u/%u",
+ spcOid, dbOid, relNumber, backendId, forkNumber)));
+#else
+ return NULL;
+#endif
+ }
+ path = psprintf("%s/%04X", slru_dirs[dbOid], relNumber);
+ }
+ else if (spcOid == GLOBALTABLESPACE_OID)
{
/* Shared system relations live in {datadir}/global */
Assert(dbOid == 0);
diff --git a/src/include/access/clog.h b/src/include/access/clog.h
index d99444f073f..aacf10ca522 100644
--- a/src/include/access/clog.h
+++ b/src/include/access/clog.h
@@ -40,18 +40,12 @@ extern void TransactionIdSetTreeStatus(TransactionId xid, int nsubxids,
TransactionId *subxids, XidStatus status, XLogRecPtr lsn);
extern XidStatus TransactionIdGetStatus(TransactionId xid, XLogRecPtr *lsn);
-extern Size CLOGShmemBuffers(void);
-extern Size CLOGShmemSize(void);
-extern void CLOGShmemInit(void);
extern void BootStrapCLOG(void);
extern void StartupCLOG(void);
extern void TrimCLOG(void);
-extern void CheckPointCLOG(void);
extern void ExtendCLOG(TransactionId newestXact);
extern void TruncateCLOG(TransactionId oldestXact, Oid oldestxid_datoid);
-extern int clogsyncfiletag(const FileTag *ftag, char *path);
-
/* XLOG stuff */
#define CLOG_ZEROPAGE 0x00
#define CLOG_TRUNCATE 0x10
diff --git a/src/include/access/commit_ts.h b/src/include/access/commit_ts.h
index 5087cdce51e..605dc75b3b0 100644
--- a/src/include/access/commit_ts.h
+++ b/src/include/access/commit_ts.h
@@ -27,7 +27,6 @@ extern bool TransactionIdGetCommitTsData(TransactionId xid,
extern TransactionId GetLatestCommitTsData(TimestampTz *ts,
RepOriginId *nodeid);
-extern Size CommitTsShmemBuffers(void);
extern Size CommitTsShmemSize(void);
extern void CommitTsShmemInit(void);
extern void BootStrapCommitTs(void);
@@ -41,8 +40,6 @@ extern void SetCommitTsLimit(TransactionId oldestXact,
TransactionId newestXact);
extern void AdvanceOldestCommitTsXid(TransactionId oldestXact);
-extern int committssyncfiletag(const FileTag *ftag, char *path);
-
/* XLOG stuff */
#define COMMIT_TS_ZEROPAGE 0x00
#define COMMIT_TS_TRUNCATE 0x10
diff --git a/src/include/access/multixact.h b/src/include/access/multixact.h
index 246f757f6ab..5848e4072ba 100644
--- a/src/include/access/multixact.h
+++ b/src/include/access/multixact.h
@@ -118,9 +118,6 @@ extern bool MultiXactIdPrecedes(MultiXactId multi1, MultiXactId multi2);
extern bool MultiXactIdPrecedesOrEquals(MultiXactId multi1,
MultiXactId multi2);
-extern int multixactoffsetssyncfiletag(const FileTag *ftag, char *path);
-extern int multixactmemberssyncfiletag(const FileTag *ftag, char *path);
-
extern void AtEOXact_MultiXact(void);
extern void AtPrepare_MultiXact(void);
extern void PostPrepare_MultiXact(TransactionId xid);
diff --git a/src/include/access/slru.h b/src/include/access/slru.h
index a8a424d92da..fcae11ce599 100644
--- a/src/include/access/slru.h
+++ b/src/include/access/slru.h
@@ -1,7 +1,7 @@
/*-------------------------------------------------------------------------
*
* slru.h
- * Simple LRU buffering for transaction status logfiles
+ * Buffering for transaction status logfiles
*
* Portions Copyright (c) 1996-2023, PostgreSQL Global Development Group
* Portions Copyright (c) 1994, Regents of the University of California
@@ -14,9 +14,35 @@
#define SLRU_H
#include "access/xlogdefs.h"
+#include "catalog/pg_tablespace_d.h"
+#include "storage/buf.h"
#include "storage/lwlock.h"
+#include "storage/smgr.h"
#include "storage/sync.h"
+/* Pseudo database ID used for SLRU data. */
+#define SLRU_SPC_ID 9
+
+/* Pseudo database IDs used by each cache. */
+#define PG_SLRU(symname,name,path, synchronize) \
+ symname,
+
+typedef enum SlruIds
+{
+#include "access/slrulist.h"
+ SLRU_NEXT_ID
+} SlruIds;
+#undef PG_SLRU
+
+typedef bool (*SlruPagePrecedesFunction) (int, int);
+
+static inline RelFileLocator
+SlruRelFileLocator(uint32 slru_db_id, uint32 segment_id)
+{
+ RelFileLocator rlocator = {SLRU_SPC_ID, slru_db_id, segment_id};
+ return rlocator;
+}
+
/*
* Define SLRU segment size. A page is the same BLCKSZ as is used everywhere
@@ -33,142 +59,40 @@
*/
#define SLRU_PAGES_PER_SEGMENT 32
-/*
- * Page status codes. Note that these do not include the "dirty" bit.
- * page_dirty can be true only in the VALID or WRITE_IN_PROGRESS states;
- * in the latter case it implies that the page has been re-dirtied since
- * the write started.
- */
-typedef enum
-{
- SLRU_PAGE_EMPTY, /* buffer is not in use */
- SLRU_PAGE_READ_IN_PROGRESS, /* page is being read in */
- SLRU_PAGE_VALID, /* page is valid and not being written */
- SLRU_PAGE_WRITE_IN_PROGRESS /* page is being written out */
-} SlruPageStatus;
-
-/*
- * Shared-memory state
- */
-typedef struct SlruSharedData
-{
- LWLock *ControlLock;
-
- /* Number of buffers managed by this SLRU structure */
- int num_slots;
-
- /*
- * Arrays holding info for each buffer slot. Page number is undefined
- * when status is EMPTY, as is page_lru_count.
- */
- char **page_buffer;
- SlruPageStatus *page_status;
- bool *page_dirty;
- int *page_number;
- int *page_lru_count;
- LWLockPadded *buffer_locks;
-
- /*
- * Optional array of WAL flush LSNs associated with entries in the SLRU
- * pages. If not zero/NULL, we must flush WAL before writing pages (true
- * for pg_xact, false for multixact, pg_subtrans, pg_notify). group_lsn[]
- * has lsn_groups_per_page entries per buffer slot, each containing the
- * highest LSN known for a contiguous group of SLRU entries on that slot's
- * page.
- */
- XLogRecPtr *group_lsn;
- int lsn_groups_per_page;
-
- /*----------
- * We mark a page "most recently used" by setting
- * page_lru_count[slotno] = ++cur_lru_count;
- * The oldest page is therefore the one with the highest value of
- * cur_lru_count - page_lru_count[slotno]
- * The counts will eventually wrap around, but this calculation still
- * works as long as no page's age exceeds INT_MAX counts.
- *----------
- */
- int cur_lru_count;
-
- /*
- * latest_page_number is the page number of the current end of the log;
- * this is not critical data, since we use it only to avoid swapping out
- * the latest page.
- */
- int latest_page_number;
-
- /* SLRU's index for statistics purposes (might not be unique) */
- int slru_stats_idx;
-} SlruSharedData;
-
-typedef SlruSharedData *SlruShared;
-
-/*
- * SlruCtlData is an unshared structure that points to the active information
- * in shared memory.
- */
-typedef struct SlruCtlData
-{
- SlruShared shared;
-
- /*
- * Which sync handler function to use when handing sync requests over to
- * the checkpointer. SYNC_HANDLER_NONE to disable fsync (eg pg_notify).
- */
- SyncRequestHandler sync_handler;
-
- /*
- * Decide whether a page is "older" for truncation and as a hint for
- * evicting pages in LRU order. Return true if every entry of the first
- * argument is older than every entry of the second argument. Note that
- * !PagePrecedes(a,b) && !PagePrecedes(b,a) need not imply a==b; it also
- * arises when some entries are older and some are not. For SLRUs using
- * SimpleLruTruncate(), this must use modular arithmetic. (For others,
- * the behavior of this callback has no functional implications.) Use
- * SlruPagePrecedesUnitTests() in SLRUs meeting its criteria.
- */
- bool (*PagePrecedes) (int, int);
-
- /*
- * Dir is set during SimpleLruInit and does not change thereafter. Since
- * it's always the same, it doesn't need to be in shared memory.
- */
- char Dir[64];
-} SlruCtlData;
-
-typedef SlruCtlData *SlruCtl;
-
-
-extern Size SimpleLruShmemSize(int nslots, int nlsns);
-extern void SimpleLruInit(SlruCtl ctl, const char *name, int nslots, int nlsns,
- LWLock *ctllock, const char *subdir, int tranche_id,
- SyncRequestHandler sync_handler);
-extern int SimpleLruZeroPage(SlruCtl ctl, int pageno);
-extern int SimpleLruReadPage(SlruCtl ctl, int pageno, bool write_ok,
- TransactionId xid);
-extern int SimpleLruReadPage_ReadOnly(SlruCtl ctl, int pageno,
- TransactionId xid);
-extern void SimpleLruWritePage(SlruCtl ctl, int slotno);
-extern void SimpleLruWriteAll(SlruCtl ctl, bool allow_redirtied);
#ifdef USE_ASSERT_CHECKING
-extern void SlruPagePrecedesUnitTests(SlruCtl ctl, int per_page);
+extern void SlruPagePrecedesUnitTests(SlruPagePrecedesFunction PagePrecedes,
+ int per_page);
#else
#define SlruPagePrecedesUnitTests(ctl, per_page) do {} while (0)
#endif
-extern void SimpleLruTruncate(SlruCtl ctl, int cutoffPage);
-extern bool SimpleLruDoesPhysicalPageExist(SlruCtl ctl, int pageno);
+extern void SimpleLruTruncate(int slru_id, SlruPagePrecedesFunction PagePrecedes,
+ int cutoffPage);
+extern bool SimpleLruDoesPhysicalPageExist(int slru_id, int pageno);
-typedef bool (*SlruScanCallback) (SlruCtl ctl, char *filename, int segpage,
+typedef bool (*SlruScanCallback) (int slru_id,
+ SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage,
void *data);
-extern bool SlruScanDirectory(SlruCtl ctl, SlruScanCallback callback, void *data);
-extern void SlruDeleteSegment(SlruCtl ctl, int segno);
-
-extern int SlruSyncFileTag(SlruCtl ctl, const FileTag *ftag, char *path);
+extern bool SlruScanDirectory(int slru_id, SlruPagePrecedesFunction PagePrecedes,
+ SlruScanCallback callback, void *data);
+extern void SlruDeleteSegment(int slru_id, int segno);
/* SlruScanDirectory public callbacks */
-extern bool SlruScanDirCbReportPresence(SlruCtl ctl, char *filename,
+extern bool SlruScanDirCbReportPresence(int slru_id,
+ SlruPagePrecedesFunction PagePrecedes,
+ char *filename,
int segpage, void *data);
-extern bool SlruScanDirCbDeleteAll(SlruCtl ctl, char *filename, int segpage,
+extern bool SlruScanDirCbDeleteAll(int slru_id, SlruPagePrecedesFunction PagePrecedes,
+ char *filename, int segpage,
void *data);
+/* Buffer access */
+extern Buffer ReadSlruBuffer(int slru_id, int pageno);
+extern Buffer ZeroSlruBuffer(int slru_id, int pageno);
+extern bool ProbeSlruBuffer(int slru_id, int pageno);
+
+/* Interfaces use by stats view */
+extern Oid SlruRelIdByName(const char *name);
+extern const char *SlruName(int slru_id);
+
#endif /* SLRU_H */
diff --git a/src/include/access/slrulist.h b/src/include/access/slrulist.h
new file mode 100644
index 00000000000..c1289a1326c
--- /dev/null
+++ b/src/include/access/slrulist.h
@@ -0,0 +1,30 @@
+/*---------------------------------------------------------------------------
+ * slrulist.h
+ *
+ * The SLRU list is kept in its own source file for possible
+ * use by automatic tools. The exact representation of a rmgr is determined
+ * by the PG_SLRU macro, which is not defined in this file; it can be
+ * defined by the caller for special purposes.
+ *
+ * Portions Copyright (c) 1996-2022, PostgreSQL Global Development Group
+ * Portions Copyright (c) 1994, Regents of the University of California
+ *
+ * src/include/access/slrulist.h
+ *---------------------------------------------------------------------------
+ */
+
+/* there is deliberately not an #ifndef SLRULIST_H here */
+
+/*
+ * List of SLRU entries. Note that order of entries defines the
+ * numerical values of each SLRU's ID, which is used in in-memory structus.
+ */
+
+/* symbol name, textual name, path, synchronize */
+PG_SLRU(SLRU_CLOG_ID, "Xact", "pg_xact", true)
+PG_SLRU(SLRU_SUBTRANS_ID, "Subtrans", "pg_subtrans", false)
+PG_SLRU(SLRU_MULTIXACT_OFFSET_ID, "MultiXactOffset", "pg_multixact/offsets", true)
+PG_SLRU(SLRU_MULTIXACT_MEMBER_ID, "MultiXactMember", "pg_multixact/members", true)
+PG_SLRU(SLRU_COMMIT_TS_ID, "CommitTs", "pg_commit_ts", true)
+PG_SLRU(SLRU_SERIAL_ID, "Serial", "pg_serial", false)
+PG_SLRU(SLRU_NOTIFY_ID, "Notify", "pg_notify", false)
diff --git a/src/include/access/subtrans.h b/src/include/access/subtrans.h
index 46a473c77f5..14e3bf720fe 100644
--- a/src/include/access/subtrans.h
+++ b/src/include/access/subtrans.h
@@ -18,11 +18,8 @@ extern void SubTransSetParent(TransactionId xid, TransactionId parent);
extern TransactionId SubTransGetParent(TransactionId xid);
extern TransactionId SubTransGetTopmostTransaction(TransactionId xid);
-extern Size SUBTRANSShmemSize(void);
-extern void SUBTRANSShmemInit(void);
extern void BootStrapSUBTRANS(void);
extern void StartupSUBTRANS(TransactionId oldestActiveXID);
-extern void CheckPointSUBTRANS(void);
extern void ExtendSUBTRANS(TransactionId newestXact);
extern void TruncateSUBTRANS(TransactionId oldestXact);
diff --git a/src/include/common/relpath.h b/src/include/common/relpath.h
index 12df11c7dfc..7f132864592 100644
--- a/src/include/common/relpath.h
+++ b/src/include/common/relpath.h
@@ -63,6 +63,9 @@ typedef enum ForkNumber
#define FORKNAMECHARS 4 /* max chars for a fork name */
+/* Pseudo tablespace ID used for SLRUs. */
+#define SLRU_SPC_OID 9
+
extern PGDLLIMPORT const char *const forkNames[];
extern ForkNumber forkname_to_number(const char *forkName);
diff --git a/src/include/storage/buf_internals.h b/src/include/storage/buf_internals.h
index 6a37e0ce6b4..276e3c55c3b 100644
--- a/src/include/storage/buf_internals.h
+++ b/src/include/storage/buf_internals.h
@@ -307,6 +307,7 @@ typedef struct WritebackContext
/* in buf_init.c */
extern PGDLLIMPORT BufferDescPadded *BufferDescriptors;
+extern PGDLLIMPORT XLogRecPtr *BufferExternalLSNs;
extern PGDLLIMPORT ConditionVariableMinimallyPadded *BufferIOCVArray;
extern PGDLLIMPORT WritebackContext BackendWritebackContext;
@@ -344,6 +345,18 @@ BufferDescriptorGetContentLock(const BufferDesc *bdesc)
return (LWLock *) (&bdesc->content_lock);
}
+static inline XLogRecPtr
+BufferGetExternalLSN(const BufferDesc *bdesc)
+{
+ return BufferExternalLSNs[bdesc->buf_id];
+}
+
+static inline void
+BufferSetExternalLSN(const BufferDesc *bdesc, XLogRecPtr lsn)
+{
+ BufferExternalLSNs[bdesc->buf_id] = lsn;
+}
+
/*
* The freeNext field is either the index of the next freelist entry,
* or one of these special values:
diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h
index 7de50bf71b7..4338752826c 100644
--- a/src/include/storage/bufmgr.h
+++ b/src/include/storage/bufmgr.h
@@ -129,12 +129,18 @@ extern Buffer ReadBufferWithoutRelcache(RelFileLocator rlocator,
ForkNumber forkNum, BlockNumber blockNum,
ReadBufferMode mode, BufferAccessStrategy strategy,
bool permanent);
+extern Buffer ReadBufferWithoutRelcacheWithHit(RelFileLocator rlocator,
+ ForkNumber forkNum, BlockNumber blockNum,
+ ReadBufferMode mode, BufferAccessStrategy strategy,
+ bool permanent, bool *hit);
extern void ReleaseBuffer(Buffer buffer);
extern void UnlockReleaseBuffer(Buffer buffer);
extern void MarkBufferDirty(Buffer buffer);
extern void IncrBufferRefCount(Buffer buffer);
extern Buffer ReleaseAndReadBuffer(Buffer buffer, Relation relation,
BlockNumber blockNum);
+extern bool BufferProbe(RelFileLocator rlocator, ForkNumber forkNum,
+ BlockNumber blockNum);
extern void InitBufferPoolAccess(void);
extern void AtEOXact_Buffers(bool isCommit);
diff --git a/src/include/storage/smgr.h b/src/include/storage/smgr.h
index fe7282127ed..bc11bc70f56 100644
--- a/src/include/storage/smgr.h
+++ b/src/include/storage/smgr.h
@@ -123,6 +123,7 @@ extern void smgrwriteback(SMgrFileHandle sfile,
extern BlockNumber smgrnblocks(SMgrFileHandle sfile);
extern BlockNumber smgrnblocks_cached(SMgrFileHandle sfile);
extern void smgrimmedsync(SMgrFileHandle sfile);
+extern void smgrunlink(SMgrFileHandle sfile, bool isRedo);
extern void smgrtruncate_multi(RelFileLocator rlocator, BackendId backend, ForkNumber *forks, int nforks, BlockNumber *nblocks);
extern void smgrunlink_multi(RelFileLocator rlocator, BackendId backend, ForkNumber *forks, int nforks, bool isRedo);
diff --git a/src/test/modules/Makefile b/src/test/modules/Makefile
index c629cbe3830..3dbdb1d769e 100644
--- a/src/test/modules/Makefile
+++ b/src/test/modules/Makefile
@@ -32,10 +32,11 @@ SUBDIRS = \
test_regex \
test_rls_hooks \
test_shm_mq \
- test_slru \
unsafe_tests \
worker_spi
+# test_slru \ # FIXME: Broken
+
ifeq ($(with_ssl),openssl)
SUBDIRS += ssl_passphrase_callback
else
--
2.30.2
^ permalink raw reply [nested|flat] 22+ messages in thread
* Re: SLRUs in the main buffer pool, redux
@ 2023-07-11 11:52 Aleksander Alekseev <[email protected]>
parent: Heikki Linnakangas <[email protected]>
0 siblings, 1 reply; 22+ messages in thread
From: Aleksander Alekseev @ 2023-07-11 11:52 UTC (permalink / raw)
To: pgsql-hackers; +Cc: Heikki Linnakangas <[email protected]>; Thomas Munro <[email protected]>; Shawn Debnath <[email protected]>; Yura Sokolov <[email protected]>
Hi,
> > Here's a rebased set of patches.
> >
> > The second patch is failing the pg_upgrade tests. Before I dig into
> > that, I'd love to get some feedback on this general approach.
>
> Forgot to include the new "slrulist.h" file in the previous patch, fixed
> here.
Unfortunately the patchset rotted quite a bit since February and needs a rebase.
--
Best regards,
Aleksander Alekseev
^ permalink raw reply [nested|flat] 22+ messages in thread
* Re: SLRUs in the main buffer pool, redux
@ 2023-09-04 12:31 Aleksander Alekseev <[email protected]>
parent: Aleksander Alekseev <[email protected]>
0 siblings, 1 reply; 22+ messages in thread
From: Aleksander Alekseev @ 2023-09-04 12:31 UTC (permalink / raw)
To: pgsql-hackers; +Cc: Heikki Linnakangas <[email protected]>; Thomas Munro <[email protected]>; Shawn Debnath <[email protected]>; Yura Sokolov <[email protected]>
Hi,
> Unfortunately the patchset rotted quite a bit since February and needs a rebase.
A consensus was reached [1] to mark this patch as RwF for now. There
are many patches to be reviewed and this one doesn't seem to be in the
best shape, so we have to prioritise. Please feel free re-submitting
the patch for the next commitfest.
[1]: https://postgr.es/m/0737f444-59bb-ac1d-2753-873c40da0840%40eisentraut.org
--
Best regards,
Aleksander Alekseev
^ permalink raw reply [nested|flat] 22+ messages in thread
* Re: SLRUs in the main buffer pool, redux
@ 2023-09-04 16:02 Aleksander Alekseev <[email protected]>
parent: Aleksander Alekseev <[email protected]>
0 siblings, 0 replies; 22+ messages in thread
From: Aleksander Alekseev @ 2023-09-04 16:02 UTC (permalink / raw)
To: pgsql-hackers; +Cc: Heikki Linnakangas <[email protected]>; Thomas Munro <[email protected]>; Shawn Debnath <[email protected]>; Yura Sokolov <[email protected]>
Hi,
> > Unfortunately the patchset rotted quite a bit since February and needs a rebase.
>
> A consensus was reached to mark this patch as RwF for now. There
> are many patches to be reviewed and this one doesn't seem to be in the
> best shape, so we have to prioritise. Please feel free re-submitting
> the patch for the next commitfest.
See also [1]
"""
[...]
Also, please consider joining the efforts and having one thread
with a single patchset rather than different threads with different
competing patches. This will simplify the work of the reviewers a lot.
Personally I would suggest taking one step back and agree on a
particular RFC first and then continue working on a single patchset
according to this RFC. We did it in the past in similar cases and this
approach proved to be productive.
[...]
"""
[1]: https://postgr.es/m/CAJ7c6TME5Z8k4undYUmKavD_dQFL0ujA%2BzFCK1eTH0_pzM%3DXrA%40mail.gmail.com
--
Best regards,
Aleksander Alekseev
^ permalink raw reply [nested|flat] 22+ messages in thread
end of thread, other threads:[~2023-09-04 16:02 UTC | newest]
Thread overview: 22+ messages (download: mbox mbox.gz follow: Atom feed)
-- links below jump to the message on this page --
2020-07-31 08:58 [PATCH 1/4] demote: setter functions for LocalXLogInsert local variable Jehan-Guillaume de Rorthais <[email protected]>
2022-01-13 13:59 SLRUs in the main buffer pool, redux Thomas Munro <[email protected]>
2022-01-13 17:46 ` Re: SLRUs in the main buffer pool, redux Robert Haas <[email protected]>
2022-01-17 10:23 ` Re: SLRUs in the main buffer pool, redux Heikki Linnakangas <[email protected]>
2022-01-17 20:05 ` Re: SLRUs in the main buffer pool, redux Thomas Munro <[email protected]>
2022-05-27 11:24 ` Re: SLRUs in the main buffer pool, redux Thomas Munro <[email protected]>
2022-05-28 01:13 ` Re: SLRUs in the main buffer pool, redux Thomas Munro <[email protected]>
2022-05-29 20:57 ` Re: SLRUs in the main buffer pool, redux Andres Freund <[email protected]>
2022-06-16 17:13 ` Re: SLRUs in the main buffer pool, redux Konstantin Knizhnik <[email protected]>
2022-06-20 16:21 ` Re: SLRUs in the main buffer pool, redux Robert Haas <[email protected]>
2022-07-21 13:23 ` Re: SLRUs in the main buffer pool, redux Yura Sokolov <[email protected]>
2022-07-25 06:54 ` Re: SLRUs in the main buffer pool, redux Heikki Linnakangas <[email protected]>
2022-07-25 08:54 ` Re: SLRUs in the main buffer pool, redux Heikki Linnakangas <[email protected]>
2023-01-20 17:00 ` Re: SLRUs in the main buffer pool, redux Shawn Debnath <[email protected]>
2023-02-27 13:31 ` Re: SLRUs in the main buffer pool, redux Heikki Linnakangas <[email protected]>
2023-02-27 13:36 ` Re: SLRUs in the main buffer pool, redux Heikki Linnakangas <[email protected]>
2023-07-11 11:52 ` Re: SLRUs in the main buffer pool, redux Aleksander Alekseev <[email protected]>
2023-09-04 12:31 ` Re: SLRUs in the main buffer pool, redux Aleksander Alekseev <[email protected]>
2023-09-04 16:02 ` Re: SLRUs in the main buffer pool, redux Aleksander Alekseev <[email protected]>
2022-07-25 09:59 ` Re: SLRUs in the main buffer pool, redux Heikki Linnakangas <[email protected]>
2022-09-17 00:41 ` Re: SLRUs in the main buffer pool, redux Bagga, Rishu <[email protected]>
2022-09-21 09:32 ` Re: SLRUs in the main buffer pool, redux Thomas Munro <[email protected]>
This inbox is served by agora; see mirroring instructions
for how to clone and mirror all data and code used for this inbox