public inbox for [email protected]
help / color / mirror / Atom feedFrom: Melanie Plageman <[email protected]>
To: Chao Li <[email protected]>
Cc: Andres Freund <[email protected]>
Cc: Andrey Borodin <[email protected]>
Cc: Kirill Reshke <[email protected]>
Cc: Xuneng Zhou <[email protected]>
Cc: Robert Haas <[email protected]>
Cc: PostgreSQL Hackers <[email protected]>
Cc: Heikki Linnakangas <[email protected]>
Subject: Re: eliminate xl_heap_visible to reduce WAL (and eventually set VM on-access)
Date: Sun, 15 Mar 2026 15:10:54 -0400
Message-ID: <CAAKRu_ZbOp52rnkjf63h5mf94raEKBH7AAbz6QTx-xdH9yLfmQ@mail.gmail.com> (raw)
In-Reply-To: <CAAKRu_Z8Ry_ynNBPAzs_Ry3MQi9NaBgt1ccLgwRsDbxWpocaBg@mail.gmail.com>
References: <[email protected]>
<CAAKRu_ZCjHoRPfQ8AbMrFY8TOMCPAvZ0_m9SX7yg0edfTk45-g@mail.gmail.com>
<[email protected]>
<CAAKRu_a04jbDACwzRYwzDND31aPyf7Yvz9TAZrTr=+F5bK1aVA@mail.gmail.com>
<CALdSSPjcv25jmXm29X-MRWZBae6+HwcWfVH1PE8NfD=EMTnkAg@mail.gmail.com>
<CAAKRu_bwtBEzDwemyim1r6yYonw7FTyFr1HXG8vywCe-MdbPBQ@mail.gmail.com>
<[email protected]>
<CAAKRu_YQd=2KvomM+RHcpeDKj0bq+peJ=3W-fip+pkvzA-Jq9w@mail.gmail.com>
<7ib3sa55sapwjlaz4sijbiq7iezna27kjvvvar4dpgkmadml6t@gfpkkwmdnepx>
<CAAKRu_bs+gZ83QDacmBxunPvCGnXJ05hxP2BDPJ3BGwdbGRXzg@mail.gmail.com>
<bqc4kh5midfn44gnjiqez3bjqv4zogydguvdn446riw45jcf3y@4ez66il7ebvk>
<CAAKRu_Y1MuANdm1p47Ev13Y9EQz8z+pw-vHOh=3DVdahUTjgXg@mail.gmail.com>
<[email protected]>
<CAAKRu_a7ByG2LipFADR7UYnLP4BWQKOV1sajqJC4=R37iO05+A@mail.gmail.com>
<[email protected]>
<[email protected]>
<[email protected]>
<CAAKRu_a+hO4PCptyaPR7AMZd7FjcHfOFKKJT8ouU3KedMud0tQ@mail.gmail.com>
<CAAKRu_Z8Ry_ynNBPAzs_Ry3MQi9NaBgt1ccLgwRsDbxWpocaBg@mail.gmail.com>
On Wed, Mar 11, 2026 at 1:01 PM Melanie Plageman
<[email protected]> wrote:
>
> On Fri, Mar 6, 2026 at 6:33 PM Melanie Plageman
> <[email protected]> wrote:
> >
> > Thanks for the review! Attached is v36. I've pushed some of the early
> > patches in the set and this is what is left.
>
> I've gone ahead and pushed another of the introductory commits.
> Attached v37 has the remaining patches.
I've pushed a few more of the trivial commits in the set. Attached v38
has the remaining patches.
- Melanie
Attachments:
[text/x-patch] v38-0001-Fix-visibility-map-corruption-in-more-cases.patch (18.6K, 2-v38-0001-Fix-visibility-map-corruption-in-more-cases.patch)
download | inline diff:
From 0ca92d2ccee0e589a35a79f9046c3a7900ecacf4 Mon Sep 17 00:00:00 2001
From: Melanie Plageman <[email protected]>
Date: Wed, 25 Feb 2026 16:23:09 -0500
Subject: [PATCH v38 01/12] Fix visibility map corruption in more cases
Move VM corruption detection and repair into pruning. This allows VM
repair during on-access pruning, not only during vacuum.
Also, expand corruption detection to cover pages marked all-visible that
contain dead tuples and tuples inserted or updated by in-progress
transactions, rather than only all-visible pages with LP_DEAD items.
Pinning the correct VM page before on-access pruning is cheap when
compared to the cost of actually pruning. The vmbuffer is saved in the
scan descriptor, so a query should only need to pin each VM page once
and a single VM page covers a large number of heap pages.
Author: Melanie Plageman <[email protected]>
Reviewed-by: Chao Li <[email protected]>
---
src/backend/access/heap/pruneheap.c | 176 ++++++++++++++++++++++++---
src/backend/access/heap/vacuumlazy.c | 89 +-------------
src/include/access/heapam.h | 12 ++
3 files changed, 175 insertions(+), 102 deletions(-)
diff --git a/src/backend/access/heap/pruneheap.c b/src/backend/access/heap/pruneheap.c
index 8d9f0694206..52cafb23c6b 100644
--- a/src/backend/access/heap/pruneheap.c
+++ b/src/backend/access/heap/pruneheap.c
@@ -19,7 +19,7 @@
#include "access/htup_details.h"
#include "access/multixact.h"
#include "access/transam.h"
-#include "access/visibilitymapdefs.h"
+#include "access/visibilitymap.h"
#include "access/xlog.h"
#include "access/xloginsert.h"
#include "commands/vacuum.h"
@@ -114,6 +114,21 @@ typedef struct
*/
HeapPageFreeze pagefrz;
+ /*-------------------------------------------------------
+ * Working state for visibility map processing
+ *-------------------------------------------------------
+ */
+
+ /*
+ * Caller must provide a pinned vmbuffer corresponding to the heap block
+ * passed to heap_page_prune_and_freeze(). We will fix any corruption
+ * found in the VM.
+ */
+ Buffer vmbuffer;
+
+ /* Bits in the vmbuffer for this heap page */
+ uint8 vmbits;
+
/*-------------------------------------------------------
* Information about what was done
*
@@ -168,6 +183,7 @@ static void prune_freeze_setup(PruneFreezeParams *params,
MultiXactId *new_relmin_mxid,
PruneFreezeResult *presult,
PruneState *prstate);
+static void heap_fix_vm_corruption(PruneState *prstate, OffsetNumber offnum);
static void prune_freeze_plan(PruneState *prstate,
OffsetNumber *off_loc);
static HTSV_Result heap_prune_satisfies_vacuum(PruneState *prstate,
@@ -175,7 +191,8 @@ static HTSV_Result heap_prune_satisfies_vacuum(PruneState *prstate,
static inline HTSV_Result htsv_get_valid_status(int status);
static void heap_prune_chain(OffsetNumber maxoff,
OffsetNumber rootoffnum, PruneState *prstate);
-static void heap_prune_record_prunable(PruneState *prstate, TransactionId xid);
+static void heap_prune_record_prunable(PruneState *prstate, TransactionId xid,
+ OffsetNumber offnum);
static void heap_prune_record_redirect(PruneState *prstate,
OffsetNumber offnum, OffsetNumber rdoffnum,
bool was_normal);
@@ -209,8 +226,9 @@ static bool heap_page_will_freeze(bool did_tuple_hint_fpi, bool do_prune, bool d
* Caller must have pin on the buffer, and must *not* have a lock on it.
*
* This function may pin *vmbuffer. It's passed by reference so the caller can
- * reuse the pin across calls, avoiding repeated pin/unpin cycles. Caller is
- * responsible for unpinning it.
+ * reuse the pin across calls, avoiding repeated pin/unpin cycles. If we find
+ * VM corruption during pruning, we will fix it. Caller is responsible for
+ * unpinning *vmbuffer.
*/
void
heap_page_prune_opt(Relation relation, Buffer buffer, Buffer *vmbuffer)
@@ -277,6 +295,16 @@ heap_page_prune_opt(Relation relation, Buffer buffer, Buffer *vmbuffer)
{
OffsetNumber dummy_off_loc;
PruneFreezeResult presult;
+ PruneFreezeParams params;
+
+ visibilitymap_pin(relation, BufferGetBlockNumber(buffer), vmbuffer);
+
+ params.relation = relation;
+ params.buffer = buffer;
+ params.vmbuffer = *vmbuffer;
+ params.reason = PRUNE_ON_ACCESS;
+ params.vistest = vistest;
+ params.cutoffs = NULL;
/*
* We don't pass the HEAP_PAGE_PRUNE_MARK_UNUSED_NOW option
@@ -284,14 +312,7 @@ heap_page_prune_opt(Relation relation, Buffer buffer, Buffer *vmbuffer)
* cannot safely determine that during on-access pruning with the
* current implementation.
*/
- PruneFreezeParams params = {
- .relation = relation,
- .buffer = buffer,
- .reason = PRUNE_ON_ACCESS,
- .options = 0,
- .vistest = vistest,
- .cutoffs = NULL,
- };
+ params.options = 0;
heap_page_prune_and_freeze(¶ms, &presult, &dummy_off_loc,
NULL, NULL);
@@ -354,6 +375,12 @@ prune_freeze_setup(PruneFreezeParams *params,
prstate->buffer = params->buffer;
prstate->page = BufferGetPage(params->buffer);
+ Assert(BufferIsValid(params->vmbuffer));
+ prstate->vmbuffer = params->vmbuffer;
+ prstate->vmbits = visibilitymap_get_status(prstate->relation,
+ prstate->block,
+ &prstate->vmbuffer);
+
/*
* Our strategy is to scan the page and make lists of items to change,
* then apply the changes within a critical section. This keeps as much
@@ -770,6 +797,90 @@ heap_page_will_freeze(bool did_tuple_hint_fpi,
return do_freeze;
}
+/*
+ * Helper to fix visibility-related corruption on a heap page and its
+ * corresponding VM page. An all-visible page cannot have dead items nor can
+ * it have tuples that are not visible to all running transactions. It clears
+ * the VM corruption as well as resetting the vmbits used during pruning.
+ *
+ * This function must be called while holding an exclusive lock on the heap
+ * buffer, and any dead items must have been discovered under that same lock.
+ * Although we do not hold a lock on the VM buffer, it is pinned, and the heap
+ * buffer is exclusively locked, ensuring that no other backend can update the
+ * VM bits corresponding to this heap page.
+ *
+ * This function makes changes to the VM and, potentially, the heap page, but
+ * it does not need to be done in a critical section.
+ */
+static void
+heap_fix_vm_corruption(PruneState *prstate, OffsetNumber offnum)
+{
+ const char *relname = RelationGetRelationName(prstate->relation);
+
+ Assert(BufferIsLockedByMeInMode(prstate->buffer, BUFFER_LOCK_EXCLUSIVE));
+
+ if (PageIsAllVisible(prstate->page))
+ {
+ /*
+ * It's possible for the value returned by
+ * GetOldestNonRemovableTransactionId() to move backwards, so it's not
+ * wrong for us to see tuples that appear to not be visible to
+ * everyone yet, while PD_ALL_VISIBLE is already set. The real safe
+ * xmin value never moves backwards, but
+ * GetOldestNonRemovableTransactionId() is conservative and sometimes
+ * returns a value that's unnecessarily small, so if we see that
+ * contradiction it just means that the tuples that we think are not
+ * visible to everyone yet actually are, and the PD_ALL_VISIBLE flag
+ * is correct.
+ *
+ * However, there should never be LP_DEAD items, dead tuple versions,
+ * or tuples inserted by an in-progress transaction on a page with
+ * PD_ALL_VISIBLE set.
+ */
+ if (prstate->lpdead_items > 0)
+ {
+ ereport(WARNING,
+ (errcode(ERRCODE_DATA_CORRUPTED),
+ errmsg("dead line pointer found on page marked all-visible"),
+ errcontext("relation \"%s\", page %u, tuple %u",
+ relname, prstate->block, offnum)));
+ }
+ else
+ {
+ ereport(WARNING,
+ (errcode(ERRCODE_DATA_CORRUPTED),
+ errmsg("tuple not visible to all transactions found on page marked all-visible"),
+ errcontext("relation \"%s\", page %u, tuple %u",
+ relname, prstate->block, offnum)));
+ }
+
+ /*
+ * Mark the buffer dirty now in case we make no further changes and
+ * therefore would not mark it dirty later.
+ */
+ PageClearAllVisible(prstate->page);
+ MarkBufferDirtyHint(prstate->buffer, true);
+ }
+ else if (prstate->vmbits & VISIBILITYMAP_VALID_BITS)
+ {
+ /*
+ * As of PostgreSQL 9.2, the visibility map bit should never be set if
+ * the page-level bit is clear. However, for vacuum, it's possible
+ * that the bit got cleared after heap_vac_scan_next_block() was
+ * called, so we must recheck now that we have the buffer lock before
+ * concluding that the VM is corrupt.
+ */
+ ereport(WARNING,
+ (errcode(ERRCODE_DATA_CORRUPTED),
+ errmsg("page is not marked all-visible but visibility map bit is set"),
+ errcontext("relation \"%s\", page %u",
+ relname, prstate->block)));
+ }
+
+ visibilitymap_clear(prstate->relation, prstate->block, prstate->vmbuffer,
+ VISIBILITYMAP_VALID_BITS);
+ prstate->vmbits = 0;
+}
/*
* Prune and repair fragmentation and potentially freeze tuples on the
@@ -830,6 +941,10 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
new_relfrozen_xid, new_relmin_mxid,
presult, &prstate);
+ if ((prstate.vmbits & VISIBILITYMAP_VALID_BITS) &&
+ !PageIsAllVisible(prstate.page))
+ heap_fix_vm_corruption(&prstate, InvalidOffsetNumber);
+
/*
* Examine all line pointers and tuple visibility information to determine
* which line pointers should change state and which tuples may be frozen.
@@ -973,6 +1088,7 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
presult->set_all_visible = prstate.set_all_visible;
presult->set_all_frozen = prstate.set_all_frozen;
presult->hastup = prstate.hastup;
+ presult->vmbits = prstate.vmbits;
/*
* For callers planning to update the visibility map, the conflict horizon
@@ -1295,7 +1411,8 @@ process_chain:
/* Record lowest soon-prunable XID */
static void
-heap_prune_record_prunable(PruneState *prstate, TransactionId xid)
+heap_prune_record_prunable(PruneState *prstate, TransactionId xid,
+ OffsetNumber offnum)
{
/*
* This should exactly match the PageSetPrunable macro. We can't store
@@ -1305,6 +1422,13 @@ heap_prune_record_prunable(PruneState *prstate, TransactionId xid)
if (!TransactionIdIsValid(prstate->new_prune_xid) ||
TransactionIdPrecedes(xid, prstate->new_prune_xid))
prstate->new_prune_xid = xid;
+
+ /*
+ * It's incorrect for a page to be marked all-visible if it contains
+ * prunable items.
+ */
+ if (PageIsAllVisible(prstate->page))
+ heap_fix_vm_corruption(prstate, offnum);
}
/* Record line pointer to be redirected */
@@ -1388,6 +1512,15 @@ heap_prune_record_dead_or_unused(PruneState *prstate, OffsetNumber offnum,
heap_prune_record_unused(prstate, offnum, was_normal);
else
heap_prune_record_dead(prstate, offnum, was_normal);
+
+ /*
+ * It's incorrect for the page to be set all-visible if it contains dead
+ * items. Fix that on the heap page and check the VM for corruption as
+ * well. Do that here rather than in heap_prune_record_dead() so we also
+ * cover tuples that are directly marked LP_UNUSED via mark_unused_now.
+ */
+ if (PageIsAllVisible(prstate->page))
+ heap_fix_vm_corruption(prstate, offnum);
}
/* Record line pointer to be marked unused */
@@ -1527,7 +1660,8 @@ heap_prune_record_unchanged_lp_normal(PruneState *prstate, OffsetNumber offnum)
* that the page is reconsidered for pruning in future.
*/
heap_prune_record_prunable(prstate,
- HeapTupleHeaderGetUpdateXid(htup));
+ HeapTupleHeaderGetUpdateXid(htup),
+ offnum);
break;
case HEAPTUPLE_INSERT_IN_PROGRESS:
@@ -1542,6 +1676,10 @@ heap_prune_record_unchanged_lp_normal(PruneState *prstate, OffsetNumber offnum)
prstate->set_all_visible = false;
prstate->set_all_frozen = false;
+ /* The page should not be marked all-visible */
+ if (PageIsAllVisible(page))
+ heap_fix_vm_corruption(prstate, offnum);
+
/*
* If we wanted to optimize for aborts, we might consider marking
* the page prunable when we see INSERT_IN_PROGRESS. But we
@@ -1566,7 +1704,8 @@ heap_prune_record_unchanged_lp_normal(PruneState *prstate, OffsetNumber offnum)
* the page is reconsidered for pruning in future.
*/
heap_prune_record_prunable(prstate,
- HeapTupleHeaderGetUpdateXid(htup));
+ HeapTupleHeaderGetUpdateXid(htup),
+ offnum);
break;
default:
@@ -1632,6 +1771,13 @@ heap_prune_record_unchanged_lp_dead(PruneState *prstate, OffsetNumber offnum)
/* Record the dead offset for vacuum */
prstate->deadoffsets[prstate->lpdead_items++] = offnum;
+
+ /*
+ * It's incorrect for a page to be marked all-visible if it contains dead
+ * items.
+ */
+ if (PageIsAllVisible(prstate->page))
+ heap_fix_vm_corruption(prstate, offnum);
}
/*
diff --git a/src/backend/access/heap/vacuumlazy.c b/src/backend/access/heap/vacuumlazy.c
index 82c5b28e0ad..957322648ca 100644
--- a/src/backend/access/heap/vacuumlazy.c
+++ b/src/backend/access/heap/vacuumlazy.c
@@ -425,11 +425,6 @@ static void find_next_unskippable_block(LVRelState *vacrel, bool *skipsallvis);
static bool lazy_scan_new_or_empty(LVRelState *vacrel, Buffer buf,
BlockNumber blkno, Page page,
bool sharelock, Buffer vmbuffer);
-static void identify_and_fix_vm_corruption(Relation rel, Buffer heap_buffer,
- BlockNumber heap_blk, Page heap_page,
- int nlpdead_items,
- Buffer vmbuffer,
- uint8 *vmbits);
static int lazy_scan_prune(LVRelState *vacrel, Buffer buf,
BlockNumber blkno, Page page,
Buffer vmbuffer,
@@ -1964,81 +1959,6 @@ cmpOffsetNumbers(const void *a, const void *b)
return pg_cmp_u16(*(const OffsetNumber *) a, *(const OffsetNumber *) b);
}
-/*
- * Helper to correct any corruption detected on a heap page and its
- * corresponding visibility map page after pruning but before setting the
- * visibility map. It examines the heap page, the associated VM page, and the
- * number of dead items previously identified.
- *
- * This function must be called while holding an exclusive lock on the heap
- * buffer, and the dead items must have been discovered under that same lock.
-
- * The provided vmbits must reflect the current state of the VM block
- * referenced by vmbuffer. Although we do not hold a lock on the VM buffer, it
- * is pinned, and the heap buffer is exclusively locked, ensuring that no
- * other backend can update the VM bits corresponding to this heap page.
- *
- * If it clears corruption, it will zero out vmbits.
- */
-static void
-identify_and_fix_vm_corruption(Relation rel, Buffer heap_buffer,
- BlockNumber heap_blk, Page heap_page,
- int nlpdead_items,
- Buffer vmbuffer,
- uint8 *vmbits)
-{
- Assert(visibilitymap_get_status(rel, heap_blk, &vmbuffer) == *vmbits);
-
- Assert(BufferIsLockedByMeInMode(heap_buffer, BUFFER_LOCK_EXCLUSIVE));
-
- /*
- * As of PostgreSQL 9.2, the visibility map bit should never be set if the
- * page-level bit is clear. However, it's possible that the bit got
- * cleared after heap_vac_scan_next_block() was called, so we must recheck
- * with buffer lock before concluding that the VM is corrupt.
- */
- if (!PageIsAllVisible(heap_page) &&
- ((*vmbits & VISIBILITYMAP_VALID_BITS) != 0))
- {
- ereport(WARNING,
- (errcode(ERRCODE_DATA_CORRUPTED),
- errmsg("page is not marked all-visible but visibility map bit is set in relation \"%s\" page %u",
- RelationGetRelationName(rel), heap_blk)));
-
- visibilitymap_clear(rel, heap_blk, vmbuffer,
- VISIBILITYMAP_VALID_BITS);
- *vmbits = 0;
- }
-
- /*
- * It's possible for the value returned by
- * GetOldestNonRemovableTransactionId() to move backwards, so it's not
- * wrong for us to see tuples that appear to not be visible to everyone
- * yet, while PD_ALL_VISIBLE is already set. The real safe xmin value
- * never moves backwards, but GetOldestNonRemovableTransactionId() is
- * conservative and sometimes returns a value that's unnecessarily small,
- * so if we see that contradiction it just means that the tuples that we
- * think are not visible to everyone yet actually are, and the
- * PD_ALL_VISIBLE flag is correct.
- *
- * There should never be LP_DEAD items on a page with PD_ALL_VISIBLE set,
- * however.
- */
- else if (PageIsAllVisible(heap_page) && nlpdead_items > 0)
- {
- ereport(WARNING,
- (errcode(ERRCODE_DATA_CORRUPTED),
- errmsg("page containing LP_DEAD items is marked as all-visible in relation \"%s\" page %u",
- RelationGetRelationName(rel), heap_blk)));
-
- PageClearAllVisible(heap_page);
- MarkBufferDirty(heap_buffer);
- visibilitymap_clear(rel, heap_blk, vmbuffer,
- VISIBILITYMAP_VALID_BITS);
- *vmbits = 0;
- }
-}
-
/*
* lazy_scan_prune() -- lazy_scan_heap() pruning and freezing.
*
@@ -2070,6 +1990,7 @@ lazy_scan_prune(LVRelState *vacrel,
PruneFreezeParams params = {
.relation = rel,
.buffer = buf,
+ .vmbuffer = vmbuffer,
.reason = PRUNE_VACUUM_SCAN,
.options = HEAP_PAGE_PRUNE_FREEZE,
.vistest = vacrel->vistest,
@@ -2179,18 +2100,12 @@ lazy_scan_prune(LVRelState *vacrel,
Assert(!presult.set_all_visible || !(*has_lpdead_items));
Assert(!presult.set_all_frozen || presult.set_all_visible);
- old_vmbits = visibilitymap_get_status(vacrel->rel, blkno, &vmbuffer);
-
- identify_and_fix_vm_corruption(vacrel->rel, buf, blkno, page,
- presult.lpdead_items, vmbuffer,
- &old_vmbits);
-
if (!presult.set_all_visible)
return presult.ndeleted;
/* Set the visibility map and page visibility hint */
+ old_vmbits = presult.vmbits;
new_vmbits = VISIBILITYMAP_ALL_VISIBLE;
-
if (presult.set_all_frozen)
new_vmbits |= VISIBILITYMAP_ALL_FROZEN;
diff --git a/src/include/access/heapam.h b/src/include/access/heapam.h
index 2fdc50b865b..c649e5f1980 100644
--- a/src/include/access/heapam.h
+++ b/src/include/access/heapam.h
@@ -262,6 +262,12 @@ typedef struct PruneFreezeParams
Relation relation; /* relation containing buffer to be pruned */
Buffer buffer; /* buffer to be pruned */
+ /*
+ * Callers should provide a pinned vmbuffer corresponding to the heap
+ * block in buffer. We will check for and repair any corruption in the VM.
+ */
+ Buffer vmbuffer;
+
/*
* The reason pruning was performed. It is used to set the WAL record
* opcode which is used for debugging and analysis purposes.
@@ -324,6 +330,12 @@ typedef struct PruneFreezeResult
bool set_all_frozen;
TransactionId vm_conflict_horizon;
+ /*
+ * vmbits is the value of the vmbuffer's vmbits at the beginning of
+ * pruning. It is cleared if VM corruption is found and corrected.
+ */
+ uint8 vmbits;
+
/*
* Whether or not the page makes rel truncation unsafe. This is set to
* 'true', even if the page contains LP_DEAD items. VACUUM will remove
--
2.43.0
[text/x-patch] v38-0002-Add-pruning-fast-path-for-all-visible-and-all-fr.patch (4.5K, 3-v38-0002-Add-pruning-fast-path-for-all-visible-and-all-fr.patch)
download | inline diff:
From 4ebe52f1b060db395d8abe5255ea1a86ed4fdc4a Mon Sep 17 00:00:00 2001
From: Melanie Plageman <[email protected]>
Date: Wed, 25 Feb 2026 16:48:19 -0500
Subject: [PATCH v38 02/12] Add pruning fast path for all-visible and
all-frozen pages
Because of the SKIP_PAGES_THRESHOLD optimization or a stale prune XID,
heap_page_prune_and_freeze() can be invoked for pages with no pruning or
freezing work. To avoid this, if a page is already all-frozen or it is
all-visible and no freezing will be attempted, we exit early.
Author: Melanie Plageman <[email protected]>
Reviewed-by: Chao Li <[email protected]>
---
src/backend/access/heap/pruneheap.c | 75 +++++++++++++++++++++++++++++
1 file changed, 75 insertions(+)
diff --git a/src/backend/access/heap/pruneheap.c b/src/backend/access/heap/pruneheap.c
index 52cafb23c6b..a4a0a916f61 100644
--- a/src/backend/access/heap/pruneheap.c
+++ b/src/backend/access/heap/pruneheap.c
@@ -184,6 +184,7 @@ static void prune_freeze_setup(PruneFreezeParams *params,
PruneFreezeResult *presult,
PruneState *prstate);
static void heap_fix_vm_corruption(PruneState *prstate, OffsetNumber offnum);
+static void heap_page_bypass_prune_freeze(PruneState *prstate, PruneFreezeResult *presult);
static void prune_freeze_plan(PruneState *prstate,
OffsetNumber *off_loc);
static HTSV_Result heap_prune_satisfies_vacuum(PruneState *prstate,
@@ -882,6 +883,68 @@ heap_fix_vm_corruption(PruneState *prstate, OffsetNumber offnum)
prstate->vmbits = 0;
}
+/*
+ * If the page is already all-frozen, or already all-visible and freezing
+ * is not being attempted, there is no remaining work and we can bypass the
+ * expensive overhead of heap_page_prune_and_freeze().
+ *
+ * This can happen when the page has a stale prune hint, or if VACUUM is
+ * scanning an already all-frozen page due to SKIP_PAGES_THRESHOLD.
+ *
+ * The caller must already have examined the visibility map and saved the
+ * status for the page's VM bits in prstate->vmbits. Caller must hold a
+ * content lock on the heap page since it will examine line pointers.
+ *
+ * Before calling heap_page_bypass_prune_freeze(), the caller should first
+ * check for and fix any discrepancy between the page-level visibility hint
+ * and the visibility map. Otherwise, the fast path will always prevent us
+ * from getting them in sync. Note that if there are tuples on the page that
+ * are not visible to all but the VM is incorrectly marked
+ * all-visible/all-frozen, we will not get the chance to fix that corruption
+ * when using the fast path.
+ */
+static void
+heap_page_bypass_prune_freeze(PruneState *prstate, PruneFreezeResult *presult)
+{
+ OffsetNumber maxoff = PageGetMaxOffsetNumber(prstate->page);
+ Page page = prstate->page;
+
+ Assert(prstate->vmbits & VISIBILITYMAP_ALL_FROZEN ||
+ (prstate->vmbits & VISIBILITYMAP_ALL_VISIBLE &&
+ !prstate->attempt_freeze));
+
+ /* We'll fill in presult for the caller */
+ memset(presult, 0, sizeof(PruneFreezeResult));
+
+ presult->vmbits = prstate->vmbits;
+
+ /* Clear any stale prune hint */
+ if (TransactionIdIsValid(PageGetPruneXid(page)))
+ {
+ PageClearPrunable(page);
+ MarkBufferDirtyHint(prstate->buffer, true);
+ }
+
+ if (PageIsEmpty(page))
+ return;
+
+ presult->hastup = true;
+
+ /*
+ * Since the page is all-visible, a count of the normal ItemIds on the
+ * page should be sufficient for vacuum's live tuple count.
+ */
+ for (OffsetNumber off = FirstOffsetNumber;
+ off <= maxoff;
+ off = OffsetNumberNext(off))
+ {
+ if (ItemIdIsNormal(PageGetItemId(page, off)))
+ prstate->live_tuples++;
+ }
+
+ presult->live_tuples = prstate->live_tuples;
+}
+
/*
* Prune and repair fragmentation and potentially freeze tuples on the
* specified page.
@@ -945,6 +1008,18 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
!PageIsAllVisible(prstate.page))
heap_fix_vm_corruption(&prstate, InvalidOffsetNumber);
+ /*
+ * If the page is already all-frozen, or already all-visible when freezing
+ * is not being attempted, we can exit early. Do this after fixing any
+ * discrepancy between the page-level visibility hint and the VM.
+ */
+ if (prstate.vmbits & VISIBILITYMAP_ALL_FROZEN ||
+ (prstate.vmbits & VISIBILITYMAP_ALL_VISIBLE && !prstate.attempt_freeze))
+ {
+ heap_page_bypass_prune_freeze(&prstate, presult);
+ return;
+ }
+
/*
* Examine all line pointers and tuple visibility information to determine
* which line pointers should change state and which tuples may be frozen.
--
2.43.0
[text/x-patch] v38-0003-Use-GlobalVisState-in-vacuum-to-determine-page-l.patch (11.4K, 4-v38-0003-Use-GlobalVisState-in-vacuum-to-determine-page-l.patch)
download | inline diff:
From 07396958d6588cd82ac420555b4d4b25194ced2d Mon Sep 17 00:00:00 2001
From: Melanie Plageman <[email protected]>
Date: Wed, 17 Dec 2025 16:51:05 -0500
Subject: [PATCH v38 03/12] Use GlobalVisState in vacuum to determine page
level visibility
During vacuum's first and third phases, we examine tuples' visibility
to determine if we can set the page all-visible in the visibility map.
Previously, this check compared tuple xmins against a single XID chosen at
the start of vacuum (OldestXmin). We now use GlobalVisState, which also
enables future work to set the VM during on-access pruning, since ordinary
queries have access to GlobalVisState but not OldestXmin.
This also benefits vacuum: in some cases, GlobalVisState may advance
during a vacuum, allowing more pages to become considered all-visible.
And, in the future, we could easily add a heuristic to update
GlobalVisState more frequently during vacuums of large tables.
OldestXmin is still used for freezing and as a backstop to ensure we
don't freeze a dead tuple that wasn't yet prunable according to
GlobalVisState in the rare occurrences where GlobalVisState moves
backwards.
Because comparing a transaction ID against GlobalVisState is more
expensive than comparing against a single XID, we defer this check until
after scanning all tuples on the page. Therefore, we perform the
GlobalVisState check only once per page. This is safe because
visibility_cutoff_xid records the newest live xmin on the page;
if it is globally visible, then the entire page is all-visible.
Using GlobalVisState means on-access pruning can also maintain
visibility_cutoff_xid. This approach will result in examining more tuple
xmins than before; however, the additional cost should not be
significant. And doing so will enable us to set the visibility map on
access in the future.
Author: Melanie Plageman <[email protected]>
Reviewed-by: Andres Freund <[email protected]>
Reviewed-by: Chao Li <[email protected]>
Discussion: https://postgr.es/m/flat/bqc4kh5midfn44gnjiqez3bjqv4zogydguvdn446riw45jcf3y%404ez66il7ebvk#c755ef151507aba58471ffaca607e493
---
src/backend/access/heap/heapam_visibility.c | 22 +++++++++
src/backend/access/heap/pruneheap.c | 37 +++++++--------
src/backend/access/heap/vacuumlazy.c | 51 +++++++++++++--------
src/include/access/heapam.h | 2 +
4 files changed, 72 insertions(+), 40 deletions(-)
diff --git a/src/backend/access/heap/heapam_visibility.c b/src/backend/access/heap/heapam_visibility.c
index fc64f4343ce..d70fab3a763 100644
--- a/src/backend/access/heap/heapam_visibility.c
+++ b/src/backend/access/heap/heapam_visibility.c
@@ -1131,6 +1131,28 @@ HeapTupleSatisfiesVacuum(HeapTuple htup, TransactionId OldestXmin,
return res;
}
+/*
+ * Wrapper around GlobalVisTestIsRemovableXid() for use when examining live
+ * tuples. Returns true if the given XID may be considered running by at least
+ * one snapshot.
+ *
+ * This function alone is insufficient to determine tuple visibility; callers
+ * must also consider the XID's commit status. Its purpose is purely semantic:
+ * when applied to live tuples, GlobalVisTestIsRemovableXid() is checking
+ * whether the inserting transaction is still considered running, not whether
+ * the tuple is removable. Live tuples are, by definition, not removable, but
+ * the snapshot criteria for “transaction still running” are identical to
+ * those used for removal XIDs.
+ *
+ * See the comment above GlobalVisTestIsRemovable[Full]Xid() for details on the
+ * required preconditions for calling this function.
+ */
+bool
+GlobalVisTestXidMaybeRunning(GlobalVisState *state, TransactionId xid)
+{
+ return !GlobalVisTestIsRemovableXid(state, xid);
+}
+
/*
* Work horse for HeapTupleSatisfiesVacuum and similar routines.
*
diff --git a/src/backend/access/heap/pruneheap.c b/src/backend/access/heap/pruneheap.c
index a4a0a916f61..05fe3deeb95 100644
--- a/src/backend/access/heap/pruneheap.c
+++ b/src/backend/access/heap/pruneheap.c
@@ -1028,6 +1028,17 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
*/
prune_freeze_plan(&prstate, off_loc);
+ /*
+ * After processing all the live tuples on the page, if the newest xmin
+ * amongst them may be considered running by any snapshot, the page cannot
+ * be all-visible.
+ */
+ if (prstate.set_all_visible &&
+ TransactionIdIsNormal(prstate.visibility_cutoff_xid) &&
+ GlobalVisTestXidMaybeRunning(prstate.vistest,
+ prstate.visibility_cutoff_xid))
+ prstate.set_all_visible = prstate.set_all_frozen = false;
+
/*
* If checksums are enabled, calling heap_prune_satisfies_vacuum() while
* checking tuple visibility information in prune_freeze_plan() may have
@@ -1695,29 +1706,15 @@ heap_prune_record_unchanged_lp_normal(PruneState *prstate, OffsetNumber offnum)
}
/*
- * The inserter definitely committed. But is it old enough
- * that everyone sees it as committed? A FrozenTransactionId
- * is seen as committed to everyone. Otherwise, we check if
- * there is a snapshot that considers this xid to still be
- * running, and if so, we don't consider the page all-visible.
+ * The inserter definitely committed. But we don't know if it
+ * is old enough that everyone sees it as committed. Later,
+ * after processing all the tuples on the page, we'll check if
+ * there is any snapshot that still considers the newest xid
+ * on the page to be running. If so, we don't consider the
+ * page all-visible.
*/
xmin = HeapTupleHeaderGetXmin(htup);
- /*
- * For now always use prstate->cutoffs for this test, because
- * we only update 'set_all_visible' and 'set_all_frozen' when
- * freezing is requested. We could use
- * GlobalVisTestIsRemovableXid instead, if a non-freezing
- * caller wanted to set the VM bit.
- */
- Assert(prstate->cutoffs);
- if (!TransactionIdPrecedes(xmin, prstate->cutoffs->OldestXmin))
- {
- prstate->set_all_visible = false;
- prstate->set_all_frozen = false;
- break;
- }
-
/* Track newest xmin on page. */
if (TransactionIdFollows(xmin, prstate->visibility_cutoff_xid) &&
TransactionIdIsNormal(xmin))
diff --git a/src/backend/access/heap/vacuumlazy.c b/src/backend/access/heap/vacuumlazy.c
index 957322648ca..ad85e1e1738 100644
--- a/src/backend/access/heap/vacuumlazy.c
+++ b/src/backend/access/heap/vacuumlazy.c
@@ -461,13 +461,13 @@ static void dead_items_cleanup(LVRelState *vacrel);
#ifdef USE_ASSERT_CHECKING
static bool heap_page_is_all_visible(Relation rel, Buffer buf,
- TransactionId OldestXmin,
+ GlobalVisState *vistest,
bool *all_frozen,
TransactionId *visibility_cutoff_xid,
OffsetNumber *logging_offnum);
#endif
static bool heap_page_would_be_all_visible(Relation rel, Buffer buf,
- TransactionId OldestXmin,
+ GlobalVisState *vistest,
OffsetNumber *deadoffsets,
int ndeadoffsets,
bool *all_frozen,
@@ -2054,13 +2054,10 @@ lazy_scan_prune(LVRelState *vacrel,
Assert(presult.lpdead_items == 0);
Assert(heap_page_is_all_visible(vacrel->rel, buf,
- vacrel->cutoffs.OldestXmin, &debug_all_frozen,
+ vacrel->vistest, &debug_all_frozen,
&debug_cutoff, &vacrel->offnum));
Assert(presult.set_all_frozen == debug_all_frozen);
-
- Assert(!TransactionIdIsValid(debug_cutoff) ||
- debug_cutoff == presult.vm_conflict_horizon);
}
#endif
@@ -2816,7 +2813,7 @@ lazy_vacuum_heap_page(LVRelState *vacrel, BlockNumber blkno, Buffer buffer,
* done outside the critical section.
*/
if (heap_page_would_be_all_visible(vacrel->rel, buffer,
- vacrel->cutoffs.OldestXmin,
+ vacrel->vistest,
deadoffsets, num_offsets,
&all_frozen, &visibility_cutoff_xid,
&vacrel->offnum))
@@ -3577,14 +3574,14 @@ dead_items_cleanup(LVRelState *vacrel)
*/
static bool
heap_page_is_all_visible(Relation rel, Buffer buf,
- TransactionId OldestXmin,
+ GlobalVisState *vistest,
bool *all_frozen,
TransactionId *visibility_cutoff_xid,
OffsetNumber *logging_offnum)
{
return heap_page_would_be_all_visible(rel, buf,
- OldestXmin,
+ vistest,
NULL, 0,
all_frozen,
visibility_cutoff_xid,
@@ -3605,7 +3602,7 @@ heap_page_is_all_visible(Relation rel, Buffer buf,
* Returns true if the page is all-visible other than the provided
* deadoffsets and false otherwise.
*
- * OldestXmin is used to determine visibility.
+ * vistest is used to determine visibility.
*
* Output parameters:
*
@@ -3624,7 +3621,7 @@ heap_page_is_all_visible(Relation rel, Buffer buf,
*/
static bool
heap_page_would_be_all_visible(Relation rel, Buffer buf,
- TransactionId OldestXmin,
+ GlobalVisState *vistest,
OffsetNumber *deadoffsets,
int ndeadoffsets,
bool *all_frozen,
@@ -3705,7 +3702,7 @@ heap_page_would_be_all_visible(Relation rel, Buffer buf,
{
TransactionId xmin;
- /* Check comments in lazy_scan_prune. */
+ /* Check heap_prune_record_unchanged_lp_normal comments */
if (!HeapTupleHeaderXminCommitted(tuple.t_data))
{
all_visible = false;
@@ -3714,16 +3711,17 @@ heap_page_would_be_all_visible(Relation rel, Buffer buf,
}
/*
- * The inserter definitely committed. But is it old enough
- * that everyone sees it as committed?
+ * The inserter definitely committed. But we don't know if
+ * it is old enough that everyone sees it as committed.
+ * Don't check that now.
+ *
+ * If we scan all tuples without finding one that prevents
+ * the page from being all-visible, we then check whether
+ * any snapshot still considers the newest XID on the page
+ * to be running. In that case, the page is not considered
+ * all-visible.
*/
xmin = HeapTupleHeaderGetXmin(tuple.t_data);
- if (!TransactionIdPrecedes(xmin, OldestXmin))
- {
- all_visible = false;
- *all_frozen = false;
- break;
- }
/* Track newest xmin on page. */
if (TransactionIdFollows(xmin, *visibility_cutoff_xid) &&
@@ -3752,6 +3750,19 @@ heap_page_would_be_all_visible(Relation rel, Buffer buf,
}
} /* scan along page */
+ /*
+ * After processing all the live tuples on the page, if the newest xmin
+ * among them may still be considered running by any snapshot, the page
+ * cannot be all-visible.
+ */
+ if (all_visible &&
+ TransactionIdIsNormal(*visibility_cutoff_xid) &&
+ GlobalVisTestXidMaybeRunning(vistest, *visibility_cutoff_xid))
+ {
+ all_visible = false;
+ *all_frozen = false;
+ }
+
/* Clear the offset information once we have processed the given page. */
*logging_offnum = InvalidOffsetNumber;
diff --git a/src/include/access/heapam.h b/src/include/access/heapam.h
index c649e5f1980..bbb223dd0d2 100644
--- a/src/include/access/heapam.h
+++ b/src/include/access/heapam.h
@@ -479,6 +479,8 @@ extern TM_Result HeapTupleSatisfiesUpdate(HeapTuple htup, CommandId curcid,
Buffer buffer);
extern HTSV_Result HeapTupleSatisfiesVacuum(HeapTuple htup, TransactionId OldestXmin,
Buffer buffer);
+
+extern bool GlobalVisTestXidMaybeRunning(GlobalVisState *state, TransactionId xid);
extern HTSV_Result HeapTupleSatisfiesVacuumHorizon(HeapTuple htup, Buffer buffer,
TransactionId *dead_after);
extern void HeapTupleSetHintBits(HeapTupleHeader tuple, Buffer buffer,
--
2.43.0
[text/x-patch] v38-0004-Keep-newest-live-XID-up-to-date-even-if-page-not.patch (14.8K, 5-v38-0004-Keep-newest-live-XID-up-to-date-even-if-page-not.patch)
download | inline diff:
From a3211750778f6a8bec42edd25f5763e2ae31d21c Mon Sep 17 00:00:00 2001
From: Melanie Plageman <[email protected]>
Date: Sat, 28 Feb 2026 16:06:51 -0500
Subject: [PATCH v38 04/12] Keep newest live XID up-to-date even if page not
all-visible
During pruning, we keep track of the newest xmin of live tuples on the
page visible to all running and future transactions so that we can use
it later as the snapshot conflict horizon when setting the VM if the
page turns out to be all-visible.
Previously, we stopped updating this value once we determined the page
was not all-visible. However, maintaining it even when the page is not
all-visible is inexpensive and makes the snapshot conflict horizon
calculation clearer. This guarantees it won't contain a stale value.
Since we'll keep it up to date all the time now anyway, there's no
reason not to maintain all_visible for on-access pruning. This will
allow us to set the VM on-access in the future.
Author: Melanie Plageman <[email protected]>
Reviewed-by: Andres Freund <[email protected]>
Reviewed-by: Chao Li <[email protected]>
Discussion: https://postgr.es/m/bqc4kh5midfn44gnjiqez3bjqv4zogydguvdn446riw45jcf3y%404ez66il7ebvk
---
src/backend/access/heap/pruneheap.c | 127 +++++++++++----------------
src/backend/access/heap/vacuumlazy.c | 30 +++----
2 files changed, 65 insertions(+), 92 deletions(-)
diff --git a/src/backend/access/heap/pruneheap.c b/src/backend/access/heap/pruneheap.c
index 05fe3deeb95..01c19ca8796 100644
--- a/src/backend/access/heap/pruneheap.c
+++ b/src/backend/access/heap/pruneheap.c
@@ -129,6 +129,9 @@ typedef struct
/* Bits in the vmbuffer for this heap page */
uint8 vmbits;
+ /* The newest xmin of live tuples on the page */
+ TransactionId newest_live_xid;
+
/*-------------------------------------------------------
* Information about what was done
*
@@ -160,11 +163,6 @@ typedef struct
* all-frozen bits in the visibility map can be set for this page after
* pruning.
*
- * visibility_cutoff_xid is the newest xmin of live tuples on the page.
- * The caller can use it as the conflict horizon, when setting the VM
- * bits. It is only valid if we froze some tuples, and set_all_frozen is
- * true.
- *
* NOTE: set_all_visible and set_all_frozen initially don't include
* LP_DEAD items. That's convenient for heap_page_prune_and_freeze() to
* use them to decide whether to freeze the page or not. The
@@ -174,7 +172,6 @@ typedef struct
*/
bool set_all_visible;
bool set_all_frozen;
- TransactionId visibility_cutoff_xid;
} PruneState;
/* Local functions */
@@ -435,53 +432,35 @@ prune_freeze_setup(PruneFreezeParams *params,
prstate->deadoffsets = presult->deadoffsets;
/*
- * Vacuum may update the VM after we're done. We can keep track of
- * whether the page will be all-visible and all-frozen after pruning and
- * freezing to help the caller to do that.
- *
- * Currently, only VACUUM sets the VM bits. To save the effort, only do
- * the bookkeeping if the caller needs it. Currently, that's tied to
- * HEAP_PAGE_PRUNE_FREEZE, but it could be a separate flag if you wanted
- * to update the VM bits without also freezing or freeze without also
- * setting the VM bits.
+ * We track whether the page will be all-visible/all-frozen at the end of
+ * pruning and freezing. While examining tuple visibility, we'll set
+ * set_all_visible to false if there are tuples on the page not visible to
+ * all running and future transactions. set_all_visible is always
+ * maintained but only VACUUM will set the VM if the page ends up being
+ * all-visible.
*
- * In addition to telling the caller whether it can set the VM bit, we
- * also use 'set_all_visible' and 'set_all_frozen' for our own
- * decision-making. If the whole page would become frozen, we consider
- * opportunistically freezing tuples. We will not be able to freeze the
- * whole page if there are tuples present that are not visible to everyone
- * or if there are dead tuples which are not yet removable. However, dead
- * tuples which will be removed by the end of vacuuming should not
- * preclude us from opportunistically freezing. Because of that, we do
- * not immediately clear set_all_visible and set_all_frozen when we see
- * LP_DEAD items. We fix that after scanning the line pointers. We must
- * correct set_all_visible and set_all_frozen before we return them to the
- * caller, so that the caller doesn't set the VM bits incorrectly.
+ * We also keep track of the newest live XID, which is used to calculate
+ * the snapshot conflict horizon for a WAL record setting the VM.
*/
- if (prstate->attempt_freeze)
- {
- prstate->set_all_visible = true;
- prstate->set_all_frozen = true;
- }
- else
- {
- /*
- * Initializing to false allows skipping the work to update them in
- * heap_prune_record_unchanged_lp_normal().
- */
- prstate->set_all_visible = false;
- prstate->set_all_frozen = false;
- }
+ prstate->set_all_visible = true;
+ prstate->newest_live_xid = InvalidTransactionId;
/*
- * The visibility cutoff xid is the newest xmin of live tuples on the
- * page. In the common case, this will be set as the conflict horizon the
- * caller can use for updating the VM. If, at the end of freezing and
- * pruning, the page is all-frozen, there is no possibility that any
- * running transaction on the standby does not see tuples on the page as
- * all-visible, so the conflict horizon remains InvalidTransactionId.
+ * Currently, only VACUUM performs freezing, but other callers may in the
+ * future. Other callers must initialize prstate.set_all_frozen to false,
+ * since we will not call heap_prepare_freeze_tuple() for each tuple.
+ *
+ * We only consider opportunistic freezing if the page would become
+ * all-frozen, or if it would be all-frozen except for dead tuples that
+ * VACUUM will remove.
+ *
+ * Dead tuples that will be removed by the end of vacuum should not
+ * prevent opportunistic freezing. Therefore, we do not clear
+ * set_all_visible and set_all_frozen when we encounter LP_DEAD items.
+ * Instead, we correct them after deciding whether to freeze, but before
+ * updating the VM, to avoid setting the VM bits incorrectly.
*/
- prstate->visibility_cutoff_xid = InvalidTransactionId;
+ prstate->set_all_frozen = prstate->attempt_freeze;
}
/*
@@ -711,7 +690,6 @@ heap_page_will_freeze(bool did_tuple_hint_fpi,
if (!prstate->attempt_freeze)
{
Assert(!prstate->set_all_frozen && prstate->nfrozen == 0);
- Assert(prstate->lpdead_items == 0 || !prstate->set_all_visible);
return false;
}
@@ -966,9 +944,8 @@ heap_page_bypass_prune_freeze(PruneState *prstate, PruneFreezeResult *presult)
* HEAP_PAGE_PRUNE_FREEZE option is passed, we also set
* presult->set_all_visible and presult->set_all_frozen after determining
* whether or not to opportunistically freeze, to indicate if the VM bits can
- * be set. They are always set to false when the HEAP_PAGE_PRUNE_FREEZE
- * option is not passed, because at the moment only callers that also freeze
- * need that information.
+ * be set. 'all-frozen' is always set to false when the HEAP_PAGE_PRUNE_FREEZE
+ * option is not passed.
*
* presult contains output parameters needed by callers, such as the number of
* tuples removed and the offsets of dead items on the page after pruning.
@@ -1034,9 +1011,9 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
* be all-visible.
*/
if (prstate.set_all_visible &&
- TransactionIdIsNormal(prstate.visibility_cutoff_xid) &&
+ TransactionIdIsNormal(prstate.newest_live_xid) &&
GlobalVisTestXidMaybeRunning(prstate.vistest,
- prstate.visibility_cutoff_xid))
+ prstate.newest_live_xid))
prstate.set_all_visible = prstate.set_all_frozen = false;
/*
@@ -1187,7 +1164,7 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
if (presult->set_all_frozen)
presult->vm_conflict_horizon = InvalidTransactionId;
else
- presult->vm_conflict_horizon = prstate.visibility_cutoff_xid;
+ presult->vm_conflict_horizon = prstate.newest_live_xid;
presult->lpdead_items = prstate.lpdead_items;
/* the presult->deadoffsets array was already filled in */
@@ -1647,6 +1624,7 @@ static void
heap_prune_record_unchanged_lp_normal(PruneState *prstate, OffsetNumber offnum)
{
HeapTupleHeader htup;
+ TransactionId xmin;
Page page = prstate->page;
Assert(!prstate->processed[offnum]);
@@ -1694,32 +1672,27 @@ heap_prune_record_unchanged_lp_normal(PruneState *prstate, OffsetNumber offnum)
* See SetHintBits for more info. Check that the tuple is hinted
* xmin-committed because of that.
*/
- if (prstate->set_all_visible)
+ if (!HeapTupleHeaderXminCommitted(htup))
{
- TransactionId xmin;
+ prstate->set_all_visible = false;
+ prstate->set_all_frozen = false;
+ break;
+ }
- if (!HeapTupleHeaderXminCommitted(htup))
- {
- prstate->set_all_visible = false;
- prstate->set_all_frozen = false;
- break;
- }
+ /*
+ * The inserter definitely committed. But we don't know if it is
+ * old enough that everyone sees it as committed. Later, after
+ * processing all the tuples on the page, we'll check if there is
+ * any snapshot that still considers the newest xid on the page to
+ * be running. If so, we don't consider the page all-visible.
+ */
+ xmin = HeapTupleHeaderGetXmin(htup);
- /*
- * The inserter definitely committed. But we don't know if it
- * is old enough that everyone sees it as committed. Later,
- * after processing all the tuples on the page, we'll check if
- * there is any snapshot that still considers the newest xid
- * on the page to be running. If so, we don't consider the
- * page all-visible.
- */
- xmin = HeapTupleHeaderGetXmin(htup);
+ /* Track newest xmin on page. */
+ if (TransactionIdFollows(xmin, prstate->newest_live_xid) &&
+ TransactionIdIsNormal(xmin))
+ prstate->newest_live_xid = xmin;
- /* Track newest xmin on page. */
- if (TransactionIdFollows(xmin, prstate->visibility_cutoff_xid) &&
- TransactionIdIsNormal(xmin))
- prstate->visibility_cutoff_xid = xmin;
- }
break;
case HEAPTUPLE_RECENTLY_DEAD:
diff --git a/src/backend/access/heap/vacuumlazy.c b/src/backend/access/heap/vacuumlazy.c
index ad85e1e1738..23402e7e26c 100644
--- a/src/backend/access/heap/vacuumlazy.c
+++ b/src/backend/access/heap/vacuumlazy.c
@@ -463,7 +463,7 @@ static void dead_items_cleanup(LVRelState *vacrel);
static bool heap_page_is_all_visible(Relation rel, Buffer buf,
GlobalVisState *vistest,
bool *all_frozen,
- TransactionId *visibility_cutoff_xid,
+ TransactionId *newest_live_xid,
OffsetNumber *logging_offnum);
#endif
static bool heap_page_would_be_all_visible(Relation rel, Buffer buf,
@@ -471,7 +471,7 @@ static bool heap_page_would_be_all_visible(Relation rel, Buffer buf,
OffsetNumber *deadoffsets,
int ndeadoffsets,
bool *all_frozen,
- TransactionId *visibility_cutoff_xid,
+ TransactionId *newest_live_xid,
OffsetNumber *logging_offnum);
static void update_relstats_all_indexes(LVRelState *vacrel);
static void vacuum_error_callback(void *arg);
@@ -2789,7 +2789,7 @@ lazy_vacuum_heap_page(LVRelState *vacrel, BlockNumber blkno, Buffer buffer,
Page page = BufferGetPage(buffer);
OffsetNumber unused[MaxHeapTuplesPerPage];
int nunused = 0;
- TransactionId visibility_cutoff_xid;
+ TransactionId newest_live_xid;
TransactionId conflict_xid = InvalidTransactionId;
bool all_frozen;
LVSavedErrInfo saved_err_info;
@@ -2815,14 +2815,14 @@ lazy_vacuum_heap_page(LVRelState *vacrel, BlockNumber blkno, Buffer buffer,
if (heap_page_would_be_all_visible(vacrel->rel, buffer,
vacrel->vistest,
deadoffsets, num_offsets,
- &all_frozen, &visibility_cutoff_xid,
+ &all_frozen, &newest_live_xid,
&vacrel->offnum))
{
vmflags |= VISIBILITYMAP_ALL_VISIBLE;
if (all_frozen)
{
vmflags |= VISIBILITYMAP_ALL_FROZEN;
- Assert(!TransactionIdIsValid(visibility_cutoff_xid));
+ Assert(!TransactionIdIsValid(newest_live_xid));
}
/*
@@ -2863,7 +2863,7 @@ lazy_vacuum_heap_page(LVRelState *vacrel, BlockNumber blkno, Buffer buffer,
visibilitymap_set_vmbits(blkno,
vmbuffer, vmflags,
vacrel->rel->rd_locator);
- conflict_xid = visibility_cutoff_xid;
+ conflict_xid = newest_live_xid;
}
/*
@@ -3576,7 +3576,7 @@ static bool
heap_page_is_all_visible(Relation rel, Buffer buf,
GlobalVisState *vistest,
bool *all_frozen,
- TransactionId *visibility_cutoff_xid,
+ TransactionId *newest_live_xid,
OffsetNumber *logging_offnum)
{
@@ -3584,7 +3584,7 @@ heap_page_is_all_visible(Relation rel, Buffer buf,
vistest,
NULL, 0,
all_frozen,
- visibility_cutoff_xid,
+ newest_live_xid,
logging_offnum);
}
#endif
@@ -3607,7 +3607,7 @@ heap_page_is_all_visible(Relation rel, Buffer buf,
* Output parameters:
*
* - *all_frozen: true if every tuple on the page is frozen
- * - *visibility_cutoff_xid: newest xmin; valid only if page is all-visible
+ * - *newest_live_xid: newest xmin of live tuples on the page
* - *logging_offnum: OffsetNumber of current tuple being processed;
* used by vacuum's error callback system.
*
@@ -3625,7 +3625,7 @@ heap_page_would_be_all_visible(Relation rel, Buffer buf,
OffsetNumber *deadoffsets,
int ndeadoffsets,
bool *all_frozen,
- TransactionId *visibility_cutoff_xid,
+ TransactionId *newest_live_xid,
OffsetNumber *logging_offnum)
{
Page page = BufferGetPage(buf);
@@ -3635,7 +3635,7 @@ heap_page_would_be_all_visible(Relation rel, Buffer buf,
bool all_visible = true;
int matched_dead_count = 0;
- *visibility_cutoff_xid = InvalidTransactionId;
+ *newest_live_xid = InvalidTransactionId;
*all_frozen = true;
Assert(ndeadoffsets == 0 || deadoffsets);
@@ -3724,9 +3724,9 @@ heap_page_would_be_all_visible(Relation rel, Buffer buf,
xmin = HeapTupleHeaderGetXmin(tuple.t_data);
/* Track newest xmin on page. */
- if (TransactionIdFollows(xmin, *visibility_cutoff_xid) &&
+ if (TransactionIdFollows(xmin, *newest_live_xid) &&
TransactionIdIsNormal(xmin))
- *visibility_cutoff_xid = xmin;
+ *newest_live_xid = xmin;
/* Check whether this tuple is already frozen or not */
if (all_visible && *all_frozen &&
@@ -3756,8 +3756,8 @@ heap_page_would_be_all_visible(Relation rel, Buffer buf,
* cannot be all-visible.
*/
if (all_visible &&
- TransactionIdIsNormal(*visibility_cutoff_xid) &&
- GlobalVisTestXidMaybeRunning(vistest, *visibility_cutoff_xid))
+ TransactionIdIsNormal(*newest_live_xid) &&
+ GlobalVisTestXidMaybeRunning(vistest, *newest_live_xid))
{
all_visible = false;
*all_frozen = false;
--
2.43.0
[text/x-patch] v38-0005-Eliminate-XLOG_HEAP2_VISIBLE-from-vacuum-phase-I.patch (24.7K, 6-v38-0005-Eliminate-XLOG_HEAP2_VISIBLE-from-vacuum-phase-I.patch)
download | inline diff:
From 532fa6da3a3b691f0cafcc18a57ae2251a8a7725 Mon Sep 17 00:00:00 2001
From: Melanie Plageman <[email protected]>
Date: Tue, 2 Dec 2025 15:07:42 -0500
Subject: [PATCH v38 05/12] Eliminate XLOG_HEAP2_VISIBLE from vacuum phase I
prune/freeze
Vacuum no longer emits a separate WAL record for each page set
all-visible or all-frozen during phase I. Instead, visibility map
updates are now included in the XLOG_HEAP2_PRUNE_VACUUM_SCAN record that
is already emitted for pruning and freezing.
Previously, heap_page_prune_and_freeze() determined whether a page was
all-visible, but the corresponding VM bits were only set later in
lazy_scan_prune(). Now the VM is updated immediately in
heap_page_prune_and_freeze(), at the same time as the heap
modifications.
This change applies only to vacuum phase I, not to pruning performed
during normal page access.
Author: Melanie Plageman <[email protected]>
Reviewed-by: Andres Freund <[email protected]>
Reviewed-by: Robert Haas <[email protected]>
Reviewed-by: Kirill Reshke <[email protected]>
Reviewed-by: Chao Li <[email protected]>
Discussion: https://postgr.es/m/flat/CAAKRu_ZMw6Npd_qm2KM%2BFwQ3cMOMx1Dh3VMhp8-V7SOLxdK9-g%40mail.gmail.com
---
src/backend/access/heap/pruneheap.c | 263 +++++++++++++++++++--------
src/backend/access/heap/vacuumlazy.c | 107 +----------
src/include/access/heapam.h | 38 ++--
3 files changed, 208 insertions(+), 200 deletions(-)
diff --git a/src/backend/access/heap/pruneheap.c b/src/backend/access/heap/pruneheap.c
index 01c19ca8796..a127e29144e 100644
--- a/src/backend/access/heap/pruneheap.c
+++ b/src/backend/access/heap/pruneheap.c
@@ -72,6 +72,21 @@ typedef struct
OffsetNumber nowunused[MaxHeapTuplesPerPage];
HeapTupleFreeze frozen[MaxHeapTuplesPerPage];
+ /*
+ * set_all_visible and set_all_frozen indicate if the all-visible and
+ * all-frozen bits in the visibility map can be set for this page after
+ * pruning.
+ *
+ * NOTE: set_all_visible and set_all_frozen initially don't include
+ * LP_DEAD items. That's convenient for heap_page_prune_and_freeze() to
+ * use them to decide whether to opportunistically freeze the page or not.
+ * The set_all_visible and set_all_frozen values ultimately used to set
+ * the VM are adjusted to include LP_DEAD items after we determine whether
+ * or not to opportunistically freeze.
+ */
+ bool set_all_visible;
+ bool set_all_frozen;
+
/*-------------------------------------------------------
* Working state for HOT chain processing
*-------------------------------------------------------
@@ -122,12 +137,16 @@ typedef struct
/*
* Caller must provide a pinned vmbuffer corresponding to the heap block
* passed to heap_page_prune_and_freeze(). We will fix any corruption
- * found in the VM.
+ * found in the VM and set the VM if the page is all-visible/all-frozen.
*/
Buffer vmbuffer;
- /* Bits in the vmbuffer for this heap page */
- uint8 vmbits;
+ /*
+ * The state of the VM bits at the beginning of pruning and the state they
+ * will be in at the end.
+ */
+ uint8 old_vmbits;
+ uint8 new_vmbits;
/* The newest xmin of live tuples on the page */
TransactionId newest_live_xid;
@@ -157,21 +176,6 @@ typedef struct
*/
int lpdead_items; /* number of items in the array */
OffsetNumber *deadoffsets; /* points directly to presult->deadoffsets */
-
- /*
- * set_all_visible and set_all_frozen indicate if the all-visible and
- * all-frozen bits in the visibility map can be set for this page after
- * pruning.
- *
- * NOTE: set_all_visible and set_all_frozen initially don't include
- * LP_DEAD items. That's convenient for heap_page_prune_and_freeze() to
- * use them to decide whether to freeze the page or not. The
- * set_all_visible and set_all_frozen values returned to the caller are
- * adjusted to include LP_DEAD items after we determine whether to
- * opportunistically freeze.
- */
- bool set_all_visible;
- bool set_all_frozen;
} PruneState;
/* Local functions */
@@ -209,7 +213,7 @@ static void page_verify_redirects(Page page);
static bool heap_page_will_freeze(bool did_tuple_hint_fpi, bool do_prune, bool do_hint_prune,
PruneState *prstate);
-
+static bool heap_page_will_set_vm(PruneState *prstate, PruneReason reason);
/*
* Optionally prune and repair fragmentation in the specified page.
@@ -375,9 +379,10 @@ prune_freeze_setup(PruneFreezeParams *params,
Assert(BufferIsValid(params->vmbuffer));
prstate->vmbuffer = params->vmbuffer;
- prstate->vmbits = visibilitymap_get_status(prstate->relation,
- prstate->block,
- &prstate->vmbuffer);
+ prstate->new_vmbits = 0;
+ prstate->old_vmbits = visibilitymap_get_status(prstate->relation,
+ prstate->block,
+ &prstate->vmbuffer);
/*
* Our strategy is to scan the page and make lists of items to change,
@@ -840,7 +845,7 @@ heap_fix_vm_corruption(PruneState *prstate, OffsetNumber offnum)
PageClearAllVisible(prstate->page);
MarkBufferDirtyHint(prstate->buffer, true);
}
- else if (prstate->vmbits & VISIBILITYMAP_VALID_BITS)
+ else if (prstate->old_vmbits & VISIBILITYMAP_VALID_BITS)
{
/*
* As of PostgreSQL 9.2, the visibility map bit should never be set if
@@ -858,7 +863,43 @@ heap_fix_vm_corruption(PruneState *prstate, OffsetNumber offnum)
visibilitymap_clear(prstate->relation, prstate->block, prstate->vmbuffer,
VISIBILITYMAP_VALID_BITS);
- prstate->vmbits = 0;
+ prstate->old_vmbits = 0;
+}
+
+/*
+ * Decide whether to set the visibility map bits (all-visible and all-frozen)
+ * for heap_blk using information from the PruneState and VM.
+ *
+ * This function does not actually set the VM bits or page-level visibility
+ * hint, PD_ALL_VISIBLE.
+ *
+ * Returns true if one or both VM bits should be set and false otherwise.
+ */
+static bool
+heap_page_will_set_vm(PruneState *prstate, PruneReason reason)
+{
+ /*
+ * Though on-access pruning maintains prstate->set_all_visible, we don't
+ * consider setting the VM.
+ */
+ if (reason == PRUNE_ON_ACCESS)
+ return false;
+
+ if (!prstate->set_all_visible)
+ return false;
+
+ prstate->new_vmbits = VISIBILITYMAP_ALL_VISIBLE;
+
+ if (prstate->set_all_frozen)
+ prstate->new_vmbits |= VISIBILITYMAP_ALL_FROZEN;
+
+ if (prstate->new_vmbits == prstate->old_vmbits)
+ {
+ prstate->new_vmbits = 0;
+ return false;
+ }
+
+ return true;
}
/*
@@ -887,15 +928,13 @@ heap_page_bypass_prune_freeze(PruneState *prstate, PruneFreezeResult *presult)
OffsetNumber maxoff = PageGetMaxOffsetNumber(prstate->page);
Page page = prstate->page;
- Assert(prstate->vmbits & VISIBILITYMAP_ALL_FROZEN ||
- (prstate->vmbits & VISIBILITYMAP_ALL_VISIBLE &&
+ Assert(prstate->old_vmbits & VISIBILITYMAP_ALL_FROZEN ||
+ (prstate->old_vmbits & VISIBILITYMAP_ALL_VISIBLE &&
!prstate->attempt_freeze));
/* We'll fill in presult for the caller */
memset(presult, 0, sizeof(PruneFreezeResult));
- presult->vmbits = prstate->vmbits;
-
/* Clear any stale prune hint */
if (TransactionIdIsValid(PageGetPruneXid(page)))
{
@@ -925,7 +964,8 @@ heap_page_bypass_prune_freeze(PruneState *prstate, PruneFreezeResult *presult)
/*
* Prune and repair fragmentation and potentially freeze tuples on the
- * specified page.
+ * specified page. If the page's visibility status has changed, update it in
+ * the VM.
*
* Caller must have pin and buffer cleanup lock on the page. Note that we
* don't update the FSM information for page on caller's behalf. Caller might
@@ -940,12 +980,10 @@ heap_page_bypass_prune_freeze(PruneState *prstate, PruneFreezeResult *presult)
* tuples if it's required in order to advance relfrozenxid / relminmxid, or
* if it's considered advantageous for overall system performance to do so
* now. The 'params.cutoffs', 'presult', 'new_relfrozen_xid' and
- * 'new_relmin_mxid' arguments are required when freezing. When
- * HEAP_PAGE_PRUNE_FREEZE option is passed, we also set
- * presult->set_all_visible and presult->set_all_frozen after determining
- * whether or not to opportunistically freeze, to indicate if the VM bits can
- * be set. 'all-frozen' is always set to false when the HEAP_PAGE_PRUNE_FREEZE
- * option is not passed.
+ * 'new_relmin_mxid' arguments are required when freezing.
+ *
+ * A vmbuffer corresponding to the heap page is also passed and if the page is
+ * found to be all-visible/all-frozen, we will set it in the VM.
*
* presult contains output parameters needed by callers, such as the number of
* tuples removed and the offsets of dead items on the page after pruning.
@@ -973,15 +1011,17 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
bool do_freeze;
bool do_prune;
bool do_hint_prune;
+ bool do_set_vm;
bool did_tuple_hint_fpi;
int64 fpi_before = pgWalUsage.wal_fpi;
+ TransactionId conflict_xid = InvalidTransactionId;
/* Initialize prstate */
prune_freeze_setup(params,
new_relfrozen_xid, new_relmin_mxid,
presult, &prstate);
- if ((prstate.vmbits & VISIBILITYMAP_VALID_BITS) &&
+ if ((prstate.old_vmbits & VISIBILITYMAP_VALID_BITS) &&
!PageIsAllVisible(prstate.page))
heap_fix_vm_corruption(&prstate, InvalidOffsetNumber);
@@ -990,8 +1030,8 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
* is not being attempted, we can exit early. Do this after fixing any
* discrepancy between the page-level visibility hint and the VM.
*/
- if (prstate.vmbits & VISIBILITYMAP_ALL_FROZEN ||
- (prstate.vmbits & VISIBILITYMAP_ALL_VISIBLE && !prstate.attempt_freeze))
+ if (prstate.old_vmbits & VISIBILITYMAP_ALL_FROZEN ||
+ (prstate.old_vmbits & VISIBILITYMAP_ALL_VISIBLE && !prstate.attempt_freeze))
{
heap_page_bypass_prune_freeze(&prstate, presult);
return;
@@ -1061,6 +1101,30 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
prstate.set_all_visible = prstate.set_all_frozen = false;
Assert(!prstate.set_all_frozen || prstate.set_all_visible);
+ Assert(!prstate.set_all_visible || (prstate.lpdead_items == 0));
+
+ do_set_vm = heap_page_will_set_vm(&prstate, params->reason);
+
+ /*
+ * new_vmbits should be 0 regardless of whether or not the page is
+ * all-visible if we do not intend to set the VM.
+ */
+ Assert(do_set_vm || prstate.new_vmbits == 0);
+
+ /*
+ * The snapshot conflict horizon for the whole record is the most
+ * conservative (newest) horizon required by any change in the record.
+ */
+ if (do_set_vm)
+ conflict_xid = prstate.newest_live_xid;
+ if (do_freeze && TransactionIdFollows(prstate.pagefrz.FreezePageConflictXid, conflict_xid))
+ conflict_xid = prstate.pagefrz.FreezePageConflictXid;
+ if (do_prune && TransactionIdFollows(prstate.latest_xid_removed, conflict_xid))
+ conflict_xid = prstate.latest_xid_removed;
+
+ /* Lock vmbuffer before entering a critical section */
+ if (do_set_vm)
+ LockBuffer(prstate.vmbuffer, BUFFER_LOCK_EXCLUSIVE);
/* Any error while applying the changes is critical */
START_CRIT_SECTION();
@@ -1082,14 +1146,17 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
/*
* If that's all we had to do to the page, this is a non-WAL-logged
- * hint. If we are going to freeze or prune the page, we will mark
- * the buffer dirty below.
+ * hint. If we are going to freeze or prune the page or set
+ * PD_ALL_VISIBLE, we will mark the buffer dirty below.
+ *
+ * Setting PD_ALL_VISIBLE is fully WAL-logged because it is forbidden
+ * for the VM to be set and PD_ALL_VISIBLE to be clear.
*/
- if (!do_freeze && !do_prune)
+ if (!do_freeze && !do_prune && !do_set_vm)
MarkBufferDirtyHint(prstate.buffer, true);
}
- if (do_prune || do_freeze)
+ if (do_prune || do_freeze || do_set_vm)
{
/* Apply the planned item changes and repair page fragmentation. */
if (do_prune)
@@ -1103,6 +1170,27 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
if (do_freeze)
heap_freeze_prepared_tuples(prstate.buffer, prstate.frozen, prstate.nfrozen);
+ /* Set the visibility map and page visibility hint */
+ if (do_set_vm)
+ {
+ /*
+ * While it is valid for PD_ALL_VISIBLE to be set when the
+ * corresponding VM bit is clear, we strongly prefer to keep them
+ * in sync.
+ *
+ * The heap buffer must be marked dirty before adding it to the
+ * WAL chain when setting the VM. We don't worry about
+ * unnecessarily dirtying the heap buffer if PD_ALL_VISIBLE is
+ * already set, though. It is extremely rare to have a clean heap
+ * buffer with PD_ALL_VISIBLE already set and the VM bits clear,
+ * so there is no point in optimizing it.
+ */
+ PageSetAllVisible(prstate.page);
+ PageClearPrunable(prstate.page);
+ visibilitymap_set_vmbits(prstate.block, prstate.vmbuffer, prstate.new_vmbits,
+ prstate.relation->rd_locator);
+ }
+
MarkBufferDirty(prstate.buffer);
/*
@@ -1110,29 +1198,12 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
*/
if (RelationNeedsWAL(prstate.relation))
{
- /*
- * The snapshotConflictHorizon for the whole record should be the
- * most conservative of all the horizons calculated for any of the
- * possible modifications. If this record will prune tuples, any
- * queries on the standby older than the newest xid of the most
- * recently removed tuple this record will prune will conflict. If
- * this record will freeze tuples, any queries on the standby with
- * xids older than the newest tuple this record will freeze will
- * conflict.
- */
- TransactionId conflict_xid;
-
- if (TransactionIdFollows(prstate.pagefrz.FreezePageConflictXid,
- prstate.latest_xid_removed))
- conflict_xid = prstate.pagefrz.FreezePageConflictXid;
- else
- conflict_xid = prstate.latest_xid_removed;
-
log_heap_prune_and_freeze(prstate.relation, prstate.buffer,
- InvalidBuffer, /* vmbuffer */
- 0, /* vmflags */
+ do_set_vm ? prstate.vmbuffer : InvalidBuffer,
+ do_set_vm ? prstate.new_vmbits : 0,
conflict_xid,
- true, params->reason,
+ true, /* cleanup lock */
+ params->reason,
prstate.frozen, prstate.nfrozen,
prstate.redirected, prstate.nredirected,
prstate.nowdead, prstate.ndead,
@@ -1142,33 +1213,67 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
END_CRIT_SECTION();
+ if (do_set_vm)
+ LockBuffer(prstate.vmbuffer, BUFFER_LOCK_UNLOCK);
+
+ /*
+ * During its second pass over the heap, VACUUM calls
+ * heap_page_would_be_all_visible() to determine whether a page is
+ * all-visible and all-frozen. The logic here is similar. After completing
+ * pruning and freezing, use an assertion to verify that our results
+ * remain consistent with heap_page_would_be_all_visible().
+ */
+#ifdef USE_ASSERT_CHECKING
+ if (prstate.set_all_visible)
+ {
+ TransactionId debug_cutoff;
+ bool debug_all_frozen;
+
+ Assert(prstate.lpdead_items == 0);
+
+ Assert(heap_page_is_all_visible(prstate.relation, prstate.buffer,
+ prstate.vistest,
+ &debug_all_frozen,
+ &debug_cutoff, off_loc));
+
+ /*
+ * It's possible the page is composed entirely of frozen tuples but is
+ * not set all-frozen in the VM and did not pass
+ * HEAP_PAGE_PRUNE_FREEZE. In this case, it's possible
+ * heap_page_is_all_visible() finds the page completely frozen, even
+ * though prstate.set_all_frozen is false.
+ */
+ Assert(!prstate.set_all_frozen || debug_all_frozen);
+ }
+#endif
+
/* Copy information back for caller */
presult->ndeleted = prstate.ndeleted;
presult->nnewlpdead = prstate.ndead;
presult->nfrozen = prstate.nfrozen;
presult->live_tuples = prstate.live_tuples;
presult->recently_dead_tuples = prstate.recently_dead_tuples;
- presult->set_all_visible = prstate.set_all_visible;
- presult->set_all_frozen = prstate.set_all_frozen;
presult->hastup = prstate.hastup;
- presult->vmbits = prstate.vmbits;
-
- /*
- * For callers planning to update the visibility map, the conflict horizon
- * for that record must be the newest xmin on the page. However, if the
- * page is completely frozen, there can be no conflict and the
- * vm_conflict_horizon should remain InvalidTransactionId. This includes
- * the case that we just froze all the tuples; the prune-freeze record
- * included the conflict XID already so the caller doesn't need it.
- */
- if (presult->set_all_frozen)
- presult->vm_conflict_horizon = InvalidTransactionId;
- else
- presult->vm_conflict_horizon = prstate.newest_live_xid;
presult->lpdead_items = prstate.lpdead_items;
/* the presult->deadoffsets array was already filled in */
+ presult->new_all_visible_pages = 0;
+ presult->new_all_frozen_pages = 0;
+ presult->new_all_visible_frozen_pages = 0;
+ if (do_set_vm)
+ {
+ if ((prstate.old_vmbits & VISIBILITYMAP_ALL_VISIBLE) == 0)
+ {
+ presult->new_all_visible_pages = 1;
+ if (prstate.set_all_frozen)
+ presult->new_all_visible_frozen_pages = 1;
+ }
+ else if ((prstate.old_vmbits & VISIBILITYMAP_ALL_FROZEN) == 0 &&
+ prstate.set_all_frozen)
+ presult->new_all_frozen_pages = 1;
+ }
+
if (prstate.attempt_freeze)
{
if (presult->nfrozen > 0)
diff --git a/src/backend/access/heap/vacuumlazy.c b/src/backend/access/heap/vacuumlazy.c
index 23402e7e26c..6b5210d6393 100644
--- a/src/backend/access/heap/vacuumlazy.c
+++ b/src/backend/access/heap/vacuumlazy.c
@@ -459,13 +459,6 @@ static void dead_items_add(LVRelState *vacrel, BlockNumber blkno, OffsetNumber *
static void dead_items_reset(LVRelState *vacrel);
static void dead_items_cleanup(LVRelState *vacrel);
-#ifdef USE_ASSERT_CHECKING
-static bool heap_page_is_all_visible(Relation rel, Buffer buf,
- GlobalVisState *vistest,
- bool *all_frozen,
- TransactionId *newest_live_xid,
- OffsetNumber *logging_offnum);
-#endif
static bool heap_page_would_be_all_visible(Relation rel, Buffer buf,
GlobalVisState *vistest,
OffsetNumber *deadoffsets,
@@ -1996,8 +1989,6 @@ lazy_scan_prune(LVRelState *vacrel,
.vistest = vacrel->vistest,
.cutoffs = &vacrel->cutoffs,
};
- uint8 old_vmbits = 0;
- uint8 new_vmbits = 0;
Assert(BufferGetBlockNumber(buf) == blkno);
@@ -2038,29 +2029,6 @@ lazy_scan_prune(LVRelState *vacrel,
vacrel->new_frozen_tuple_pages++;
}
- /*
- * VACUUM will call heap_page_is_all_visible() during the second pass over
- * the heap to determine all_visible and all_frozen for the page -- this
- * is a specialized version of the logic from this function. Now that
- * we've finished pruning and freezing, make sure that we're in total
- * agreement with heap_page_is_all_visible() using an assertion.
- */
-#ifdef USE_ASSERT_CHECKING
- if (presult.set_all_visible)
- {
- TransactionId debug_cutoff;
- bool debug_all_frozen;
-
- Assert(presult.lpdead_items == 0);
-
- Assert(heap_page_is_all_visible(vacrel->rel, buf,
- vacrel->vistest, &debug_all_frozen,
- &debug_cutoff, &vacrel->offnum));
-
- Assert(presult.set_all_frozen == debug_all_frozen);
- }
-#endif
-
/*
* Now save details of the LP_DEAD items from the page in vacrel
*/
@@ -2081,6 +2049,14 @@ lazy_scan_prune(LVRelState *vacrel,
}
/* Finally, add page-local counts to whole-VACUUM counts */
+ vacrel->new_all_visible_pages += presult.new_all_visible_pages;
+ vacrel->new_all_visible_all_frozen_pages += presult.new_all_visible_frozen_pages;
+ vacrel->new_all_frozen_pages += presult.new_all_frozen_pages;
+
+ /* Capture if the page was newly set frozen */
+ *vm_page_frozen = presult.new_all_visible_frozen_pages > 0 ||
+ presult.new_all_frozen_pages > 0;
+
vacrel->tuples_deleted += presult.ndeleted;
vacrel->tuples_frozen += presult.nfrozen;
vacrel->lpdead_items += presult.lpdead_items;
@@ -2094,71 +2070,6 @@ lazy_scan_prune(LVRelState *vacrel,
/* Did we find LP_DEAD items? */
*has_lpdead_items = (presult.lpdead_items > 0);
- Assert(!presult.set_all_visible || !(*has_lpdead_items));
- Assert(!presult.set_all_frozen || presult.set_all_visible);
-
- if (!presult.set_all_visible)
- return presult.ndeleted;
-
- /* Set the visibility map and page visibility hint */
- old_vmbits = presult.vmbits;
- new_vmbits = VISIBILITYMAP_ALL_VISIBLE;
- if (presult.set_all_frozen)
- new_vmbits |= VISIBILITYMAP_ALL_FROZEN;
-
- /* Nothing to do */
- if (old_vmbits == new_vmbits)
- return presult.ndeleted;
-
- /*
- * It should never be the case that the visibility map page is set while
- * the page-level bit is clear (and if so, we cleared it above), but the
- * reverse is allowed (if checksums are not enabled). Regardless, set both
- * bits so that we get back in sync.
- *
- * The heap buffer must be marked dirty before adding it to the WAL chain
- * when setting the VM. We don't worry about unnecessarily dirtying the
- * heap buffer if PD_ALL_VISIBLE is already set, though. It is extremely
- * rare to have a clean heap buffer with PD_ALL_VISIBLE already set and
- * the VM bits clear, so there is no point in optimizing it.
- */
- PageSetAllVisible(page);
- PageClearPrunable(page);
- MarkBufferDirty(buf);
-
- /*
- * If the page is being set all-frozen, we pass InvalidTransactionId as
- * the cutoff_xid, since a snapshot conflict horizon sufficient to make
- * everything safe for REDO was logged when the page's tuples were frozen.
- */
- Assert(!presult.set_all_frozen ||
- !TransactionIdIsValid(presult.vm_conflict_horizon));
-
- visibilitymap_set(vacrel->rel, blkno, buf,
- InvalidXLogRecPtr,
- vmbuffer, presult.vm_conflict_horizon,
- new_vmbits);
-
- /*
- * If the page wasn't already set all-visible and/or all-frozen in the VM,
- * count it as newly set for logging.
- */
- if ((old_vmbits & VISIBILITYMAP_ALL_VISIBLE) == 0)
- {
- vacrel->new_all_visible_pages++;
- if (presult.set_all_frozen)
- {
- vacrel->new_all_visible_all_frozen_pages++;
- *vm_page_frozen = true;
- }
- }
- else if ((old_vmbits & VISIBILITYMAP_ALL_FROZEN) == 0 &&
- presult.set_all_frozen)
- {
- vacrel->new_all_frozen_pages++;
- *vm_page_frozen = true;
- }
-
return presult.ndeleted;
}
@@ -3572,7 +3483,7 @@ dead_items_cleanup(LVRelState *vacrel)
* that expect no LP_DEAD on the page. Currently assert-only, but there is no
* reason not to use it outside of asserts.
*/
-static bool
+bool
heap_page_is_all_visible(Relation rel, Buffer buf,
GlobalVisState *vistest,
bool *all_frozen,
diff --git a/src/include/access/heapam.h b/src/include/access/heapam.h
index bbb223dd0d2..f77a00291bb 100644
--- a/src/include/access/heapam.h
+++ b/src/include/access/heapam.h
@@ -264,7 +264,8 @@ typedef struct PruneFreezeParams
/*
* Callers should provide a pinned vmbuffer corresponding to the heap
- * block in buffer. We will check for and repair any corruption in the VM.
+ * block in buffer. We will check for and repair any corruption in the VM
+ * and set the VM after pruning if the page is all-visible/all-frozen.
*/
Buffer vmbuffer;
@@ -280,8 +281,7 @@ typedef struct PruneFreezeParams
* HEAP_PAGE_PRUNE_MARK_UNUSED_NOW indicates that dead items can be set
* LP_UNUSED during pruning.
*
- * HEAP_PAGE_PRUNE_FREEZE indicates that we will also freeze tuples, and
- * will return 'all_visible', 'all_frozen' flags to the caller.
+ * HEAP_PAGE_PRUNE_FREEZE indicates that we will also freeze tuples.
*/
int options;
@@ -315,26 +315,12 @@ typedef struct PruneFreezeResult
int recently_dead_tuples;
/*
- * set_all_visible and set_all_frozen indicate if the all-visible and
- * all-frozen bits in the visibility map should be set for this page after
- * pruning.
- *
- * vm_conflict_horizon is the newest xmin of live tuples on the page. The
- * caller can use it as the conflict horizon when setting the VM bits. It
- * is only valid if we froze some tuples (nfrozen > 0), and set_all_frozen
- * is true.
- *
- * These are only set if the HEAP_PAGE_PRUNE_FREEZE option is set.
- */
- bool set_all_visible;
- bool set_all_frozen;
- TransactionId vm_conflict_horizon;
-
- /*
- * vmbits is the value of the vmbuffer's vmbits at the beginning of
- * pruning. It is cleared if VM corruption is found and corrected.
+ * Whether or not the page was newly set all-visible and all-frozen during
+ * phase I of vacuuming.
*/
- uint8 vmbits;
+ BlockNumber new_all_visible_pages;
+ BlockNumber new_all_visible_frozen_pages;
+ BlockNumber new_all_frozen_pages;
/*
* Whether or not the page makes rel truncation unsafe. This is set to
@@ -471,7 +457,13 @@ extern void log_heap_prune_and_freeze(Relation relation, Buffer buffer,
/* in heap/vacuumlazy.c */
extern void heap_vacuum_rel(Relation rel,
const VacuumParams params, BufferAccessStrategy bstrategy);
-
+#ifdef USE_ASSERT_CHECKING
+extern bool heap_page_is_all_visible(Relation rel, Buffer buf,
+ GlobalVisState *vistest,
+ bool *all_frozen,
+ TransactionId *visibility_cutoff_xid,
+ OffsetNumber *logging_offnum);
+#endif
/* in heap/heapam_visibility.c */
extern bool HeapTupleSatisfiesVisibility(HeapTuple htup, Snapshot snapshot,
Buffer buffer);
--
2.43.0
[text/x-patch] v38-0006-Eliminate-XLOG_HEAP2_VISIBLE-from-empty-page-vac.patch (2.6K, 7-v38-0006-Eliminate-XLOG_HEAP2_VISIBLE-from-empty-page-vac.patch)
download | inline diff:
From db4cc2361ccc446b54df3d2d5afde70f6869dde1 Mon Sep 17 00:00:00 2001
From: Melanie Plageman <[email protected]>
Date: Sat, 27 Sep 2025 11:55:21 -0400
Subject: [PATCH v38 06/12] Eliminate XLOG_HEAP2_VISIBLE from empty-page vacuum
As part of removing XLOG_HEAP2_VISIBLE records, phase I of VACUUM now
marks empty pages all-visible in a XLOG_HEAP2_PRUNE_VACUUM_SCAN record.
Author: Melanie Plageman <[email protected]>
Reviewed-by: Robert Haas <[email protected]>
Reviewed-by: Andres Freund <[email protected]>
Reviewed-by: Chao Li <[email protected]>
---
src/backend/access/heap/vacuumlazy.c | 35 +++++++++++++++++++++++-----
1 file changed, 29 insertions(+), 6 deletions(-)
diff --git a/src/backend/access/heap/vacuumlazy.c b/src/backend/access/heap/vacuumlazy.c
index 6b5210d6393..1451c943644 100644
--- a/src/backend/access/heap/vacuumlazy.c
+++ b/src/backend/access/heap/vacuumlazy.c
@@ -1903,9 +1903,12 @@ lazy_scan_new_or_empty(LVRelState *vacrel, Buffer buf, BlockNumber blkno,
*/
if (!PageIsAllVisible(page))
{
+ /* Lock vmbuffer before entering critical section */
+ LockBuffer(vmbuffer, BUFFER_LOCK_EXCLUSIVE);
+
START_CRIT_SECTION();
- /* mark buffer dirty before writing a WAL record */
+ /* Mark buffer dirty before writing any WAL records */
MarkBufferDirty(buf);
/*
@@ -1923,13 +1926,33 @@ lazy_scan_new_or_empty(LVRelState *vacrel, Buffer buf, BlockNumber blkno,
PageSetAllVisible(page);
PageClearPrunable(page);
- visibilitymap_set(vacrel->rel, blkno, buf,
- InvalidXLogRecPtr,
- vmbuffer, InvalidTransactionId,
- VISIBILITYMAP_ALL_VISIBLE |
- VISIBILITYMAP_ALL_FROZEN);
+ visibilitymap_set_vmbits(blkno,
+ vmbuffer,
+ VISIBILITYMAP_ALL_VISIBLE |
+ VISIBILITYMAP_ALL_FROZEN,
+ vacrel->rel->rd_locator);
+
+ /*
+ * Emit WAL for setting PD_ALL_VISIBLE on the heap page and
+ * setting the VM.
+ */
+ if (RelationNeedsWAL(vacrel->rel))
+ log_heap_prune_and_freeze(vacrel->rel, buf,
+ vmbuffer,
+ VISIBILITYMAP_ALL_VISIBLE |
+ VISIBILITYMAP_ALL_FROZEN,
+ InvalidTransactionId, /* conflict xid */
+ false, /* cleanup lock */
+ PRUNE_VACUUM_SCAN, /* reason */
+ NULL, 0,
+ NULL, 0,
+ NULL, 0,
+ NULL, 0);
+
END_CRIT_SECTION();
+ LockBuffer(vmbuffer, BUFFER_LOCK_UNLOCK);
+
/* Count the newly all-frozen pages for logging */
vacrel->new_all_visible_pages++;
vacrel->new_all_visible_all_frozen_pages++;
--
2.43.0
[text/x-patch] v38-0007-Remove-XLOG_HEAP2_VISIBLE-entirely.patch (25.0K, 8-v38-0007-Remove-XLOG_HEAP2_VISIBLE-entirely.patch)
download | inline diff:
From 97a248e7711eaed31954dd7089790e3369b0c58a Mon Sep 17 00:00:00 2001
From: Melanie Plageman <[email protected]>
Date: Sat, 27 Sep 2025 11:55:36 -0400
Subject: [PATCH v38 07/12] Remove XLOG_HEAP2_VISIBLE entirely
There are no remaining users that emit XLOG_HEAP2_VISIBLE records, so it
can be removed. This includes deleting the xl_heap_visible struct and
all functions responsible for emitting or replaying XLOG_HEAP2_VISIBLE
records.
This changes the visibility map API, so any external users/consumers of
the VM-only WAL record will need to change.
Author: Melanie Plageman <[email protected]>
Reviewed-by: Andrey Borodin <[email protected]>
Reviewed-by: Andres Freund <[email protected]>
Reviewed-by: Chao Li <[email protected]>
---
src/backend/access/common/bufmask.c | 4 +-
src/backend/access/heap/heapam.c | 54 +-------
src/backend/access/heap/heapam_xlog.c | 156 ++---------------------
src/backend/access/heap/pruneheap.c | 4 +-
src/backend/access/heap/vacuumlazy.c | 16 +--
src/backend/access/heap/visibilitymap.c | 110 +---------------
src/backend/access/rmgrdesc/heapdesc.c | 10 --
src/backend/replication/logical/decode.c | 1 -
src/backend/storage/ipc/standby.c | 12 +-
src/include/access/heapam_xlog.h | 20 ---
src/include/access/visibilitymap.h | 13 +-
src/include/access/visibilitymapdefs.h | 9 --
src/tools/pgindent/typedefs.list | 1 -
13 files changed, 38 insertions(+), 372 deletions(-)
diff --git a/src/backend/access/common/bufmask.c b/src/backend/access/common/bufmask.c
index 8a67bfa1aff..d9042e1f91d 100644
--- a/src/backend/access/common/bufmask.c
+++ b/src/backend/access/common/bufmask.c
@@ -56,8 +56,8 @@ mask_page_hint_bits(Page page)
/*
* During replay, if the page LSN has advanced past our XLOG record's LSN,
- * we don't mark the page all-visible. See heap_xlog_visible() for
- * details.
+ * we don't mark the page all-visible. See heap_xlog_prune_freeze() for
+ * more details.
*/
PageClearAllVisible(page);
}
diff --git a/src/backend/access/heap/heapam.c b/src/backend/access/heap/heapam.c
index e5bd062de77..044f385e477 100644
--- a/src/backend/access/heap/heapam.c
+++ b/src/backend/access/heap/heapam.c
@@ -2589,11 +2589,11 @@ heap_multi_insert(Relation relation, TupleTableSlot **slots, int ntuples,
{
PageSetAllVisible(page);
PageClearPrunable(page);
- visibilitymap_set_vmbits(BufferGetBlockNumber(buffer),
- vmbuffer,
- VISIBILITYMAP_ALL_VISIBLE |
- VISIBILITYMAP_ALL_FROZEN,
- relation->rd_locator);
+ visibilitymap_set(BufferGetBlockNumber(buffer),
+ vmbuffer,
+ VISIBILITYMAP_ALL_VISIBLE |
+ VISIBILITYMAP_ALL_FROZEN,
+ relation->rd_locator);
}
/*
@@ -8886,50 +8886,6 @@ bottomup_sort_and_shrink(TM_IndexDeleteOp *delstate)
return nblocksfavorable;
}
-/*
- * Perform XLogInsert for a heap-visible operation. 'block' is the block
- * being marked all-visible, and vm_buffer is the buffer containing the
- * corresponding visibility map block. Both should have already been modified
- * and dirtied.
- *
- * snapshotConflictHorizon comes from the largest xmin on the page being
- * marked all-visible. REDO routine uses it to generate recovery conflicts.
- *
- * If checksums or wal_log_hints are enabled, we may also generate a full-page
- * image of heap_buffer. Otherwise, we optimize away the FPI (by specifying
- * REGBUF_NO_IMAGE for the heap buffer), in which case the caller should *not*
- * update the heap page's LSN.
- */
-XLogRecPtr
-log_heap_visible(Relation rel, Buffer heap_buffer, Buffer vm_buffer,
- TransactionId snapshotConflictHorizon, uint8 vmflags)
-{
- xl_heap_visible xlrec;
- XLogRecPtr recptr;
- uint8 flags;
-
- Assert(BufferIsValid(heap_buffer));
- Assert(BufferIsValid(vm_buffer));
-
- xlrec.snapshotConflictHorizon = snapshotConflictHorizon;
- xlrec.flags = vmflags;
- if (RelationIsAccessibleInLogicalDecoding(rel))
- xlrec.flags |= VISIBILITYMAP_XLOG_CATALOG_REL;
- XLogBeginInsert();
- XLogRegisterData(&xlrec, SizeOfHeapVisible);
-
- XLogRegisterBuffer(0, vm_buffer, 0);
-
- flags = REGBUF_STANDARD;
- if (!XLogHintBitIsNeeded())
- flags |= REGBUF_NO_IMAGE;
- XLogRegisterBuffer(1, heap_buffer, flags);
-
- recptr = XLogInsert(RM_HEAP2_ID, XLOG_HEAP2_VISIBLE);
-
- return recptr;
-}
-
/*
* Perform XLogInsert for a heap-update operation. Caller must already
* have modified the buffer(s) and marked them dirty.
diff --git a/src/backend/access/heap/heapam_xlog.c b/src/backend/access/heap/heapam_xlog.c
index 1da774c1536..1302bb13e18 100644
--- a/src/backend/access/heap/heapam_xlog.c
+++ b/src/backend/access/heap/heapam_xlog.c
@@ -239,7 +239,7 @@ heap_xlog_prune_freeze(XLogReaderState *record)
if (PageIsNew(vmpage))
PageInit(vmpage, BLCKSZ, 0);
- visibilitymap_set_vmbits(blkno, vmbuffer, vmflags, rlocator);
+ visibilitymap_set(blkno, vmbuffer, vmflags, rlocator);
Assert(BufferIsDirty(vmbuffer));
PageSetLSN(vmpage, lsn);
@@ -252,143 +252,6 @@ heap_xlog_prune_freeze(XLogReaderState *record)
XLogRecordPageWithFreeSpace(rlocator, blkno, freespace);
}
-/*
- * Replay XLOG_HEAP2_VISIBLE records.
- *
- * The critical integrity requirement here is that we must never end up with
- * a situation where the visibility map bit is set, and the page-level
- * PD_ALL_VISIBLE bit is clear. If that were to occur, then a subsequent
- * page modification would fail to clear the visibility map bit.
- */
-static void
-heap_xlog_visible(XLogReaderState *record)
-{
- XLogRecPtr lsn = record->EndRecPtr;
- xl_heap_visible *xlrec = (xl_heap_visible *) XLogRecGetData(record);
- Buffer vmbuffer = InvalidBuffer;
- Buffer buffer;
- Page page;
- RelFileLocator rlocator;
- BlockNumber blkno;
- XLogRedoAction action;
-
- Assert((xlrec->flags & VISIBILITYMAP_XLOG_VALID_BITS) == xlrec->flags);
-
- XLogRecGetBlockTag(record, 1, &rlocator, NULL, &blkno);
-
- /*
- * If there are any Hot Standby transactions running that have an xmin
- * horizon old enough that this page isn't all-visible for them, they
- * might incorrectly decide that an index-only scan can skip a heap fetch.
- *
- * NB: It might be better to throw some kind of "soft" conflict here that
- * forces any index-only scan that is in flight to perform heap fetches,
- * rather than killing the transaction outright.
- */
- if (InHotStandby)
- ResolveRecoveryConflictWithSnapshot(xlrec->snapshotConflictHorizon,
- xlrec->flags & VISIBILITYMAP_XLOG_CATALOG_REL,
- rlocator);
-
- /*
- * Read the heap page, if it still exists. If the heap file has dropped or
- * truncated later in recovery, we don't need to update the page, but we'd
- * better still update the visibility map.
- */
- action = XLogReadBufferForRedo(record, 1, &buffer);
- if (action == BLK_NEEDS_REDO)
- {
- /*
- * We don't bump the LSN of the heap page when setting the visibility
- * map bit (unless checksums or wal_hint_bits is enabled, in which
- * case we must). This exposes us to torn page hazards, but since
- * we're not inspecting the existing page contents in any way, we
- * don't care.
- */
- page = BufferGetPage(buffer);
-
- PageSetAllVisible(page);
- PageClearPrunable(page);
-
- if (XLogHintBitIsNeeded())
- PageSetLSN(page, lsn);
-
- MarkBufferDirty(buffer);
- }
- else if (action == BLK_RESTORED)
- {
- /*
- * If heap block was backed up, we already restored it and there's
- * nothing more to do. (This can only happen with checksums or
- * wal_log_hints enabled.)
- */
- }
-
- if (BufferIsValid(buffer))
- {
- Size space = PageGetFreeSpace(BufferGetPage(buffer));
-
- UnlockReleaseBuffer(buffer);
-
- /*
- * Since FSM is not WAL-logged and only updated heuristically, it
- * easily becomes stale in standbys. If the standby is later promoted
- * and runs VACUUM, it will skip updating individual free space
- * figures for pages that became all-visible (or all-frozen, depending
- * on the vacuum mode,) which is troublesome when FreeSpaceMapVacuum
- * propagates too optimistic free space values to upper FSM layers;
- * later inserters try to use such pages only to find out that they
- * are unusable. This can cause long stalls when there are many such
- * pages.
- *
- * Forestall those problems by updating FSM's idea about a page that
- * is becoming all-visible or all-frozen.
- *
- * Do this regardless of a full-page image being applied, since the
- * FSM data is not in the page anyway.
- */
- if (xlrec->flags & VISIBILITYMAP_VALID_BITS)
- XLogRecordPageWithFreeSpace(rlocator, blkno, space);
- }
-
- /*
- * Even if we skipped the heap page update due to the LSN interlock, it's
- * still safe to update the visibility map. Any WAL record that clears
- * the visibility map bit does so before checking the page LSN, so any
- * bits that need to be cleared will still be cleared.
- */
- if (XLogReadBufferForRedoExtended(record, 0, RBM_ZERO_ON_ERROR, false,
- &vmbuffer) == BLK_NEEDS_REDO)
- {
- Page vmpage = BufferGetPage(vmbuffer);
- Relation reln;
- uint8 vmbits;
-
- /* initialize the page if it was read as zeros */
- if (PageIsNew(vmpage))
- PageInit(vmpage, BLCKSZ, 0);
-
- /* remove VISIBILITYMAP_XLOG_* */
- vmbits = xlrec->flags & VISIBILITYMAP_VALID_BITS;
-
- /*
- * XLogReadBufferForRedoExtended locked the buffer. But
- * visibilitymap_set will handle locking itself.
- */
- LockBuffer(vmbuffer, BUFFER_LOCK_UNLOCK);
-
- reln = CreateFakeRelcacheEntry(rlocator);
-
- visibilitymap_set(reln, blkno, InvalidBuffer, lsn, vmbuffer,
- xlrec->snapshotConflictHorizon, vmbits);
-
- ReleaseBuffer(vmbuffer);
- FreeFakeRelcacheEntry(reln);
- }
- else if (BufferIsValid(vmbuffer))
- UnlockReleaseBuffer(vmbuffer);
-}
-
/*
* Given an "infobits" field from an XLog record, set the correct bits in the
* given infomask and infomask2 for the tuple touched by the record.
@@ -769,8 +632,8 @@ heap_xlog_multi_insert(XLogReaderState *record)
*
* During recovery, however, no concurrent writers exist. Therefore,
* updating the VM without holding the heap page lock is safe enough. This
- * same approach is taken when replaying xl_heap_visible records (see
- * heap_xlog_visible()).
+ * same approach is taken when replaying XLOG_HEAP2_PRUNE* records (see
+ * heap_xlog_prune_freeze()).
*/
if ((xlrec->flags & XLH_INSERT_ALL_FROZEN_SET) &&
XLogReadBufferForRedoExtended(record, 1, RBM_ZERO_ON_ERROR, false,
@@ -782,11 +645,11 @@ heap_xlog_multi_insert(XLogReaderState *record)
if (PageIsNew(vmpage))
PageInit(vmpage, BLCKSZ, 0);
- visibilitymap_set_vmbits(blkno,
- vmbuffer,
- VISIBILITYMAP_ALL_VISIBLE |
- VISIBILITYMAP_ALL_FROZEN,
- rlocator);
+ visibilitymap_set(blkno,
+ vmbuffer,
+ VISIBILITYMAP_ALL_VISIBLE |
+ VISIBILITYMAP_ALL_FROZEN,
+ rlocator);
Assert(BufferIsDirty(vmbuffer));
PageSetLSN(vmpage, lsn);
@@ -1369,9 +1232,6 @@ heap2_redo(XLogReaderState *record)
case XLOG_HEAP2_PRUNE_VACUUM_CLEANUP:
heap_xlog_prune_freeze(record);
break;
- case XLOG_HEAP2_VISIBLE:
- heap_xlog_visible(record);
- break;
case XLOG_HEAP2_MULTI_INSERT:
heap_xlog_multi_insert(record);
break;
diff --git a/src/backend/access/heap/pruneheap.c b/src/backend/access/heap/pruneheap.c
index a127e29144e..9b5a0726f2b 100644
--- a/src/backend/access/heap/pruneheap.c
+++ b/src/backend/access/heap/pruneheap.c
@@ -1187,8 +1187,8 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
*/
PageSetAllVisible(prstate.page);
PageClearPrunable(prstate.page);
- visibilitymap_set_vmbits(prstate.block, prstate.vmbuffer, prstate.new_vmbits,
- prstate.relation->rd_locator);
+ visibilitymap_set(prstate.block, prstate.vmbuffer, prstate.new_vmbits,
+ prstate.relation->rd_locator);
}
MarkBufferDirty(prstate.buffer);
diff --git a/src/backend/access/heap/vacuumlazy.c b/src/backend/access/heap/vacuumlazy.c
index 1451c943644..8bd178ae7e6 100644
--- a/src/backend/access/heap/vacuumlazy.c
+++ b/src/backend/access/heap/vacuumlazy.c
@@ -1926,11 +1926,11 @@ lazy_scan_new_or_empty(LVRelState *vacrel, Buffer buf, BlockNumber blkno,
PageSetAllVisible(page);
PageClearPrunable(page);
- visibilitymap_set_vmbits(blkno,
- vmbuffer,
- VISIBILITYMAP_ALL_VISIBLE |
- VISIBILITYMAP_ALL_FROZEN,
- vacrel->rel->rd_locator);
+ visibilitymap_set(blkno,
+ vmbuffer,
+ VISIBILITYMAP_ALL_VISIBLE |
+ VISIBILITYMAP_ALL_FROZEN,
+ vacrel->rel->rd_locator);
/*
* Emit WAL for setting PD_ALL_VISIBLE on the heap page and
@@ -2794,9 +2794,9 @@ lazy_vacuum_heap_page(LVRelState *vacrel, BlockNumber blkno, Buffer buffer,
*/
PageSetAllVisible(page);
PageClearPrunable(page);
- visibilitymap_set_vmbits(blkno,
- vmbuffer, vmflags,
- vacrel->rel->rd_locator);
+ visibilitymap_set(blkno,
+ vmbuffer, vmflags,
+ vacrel->rel->rd_locator);
conflict_xid = newest_live_xid;
}
diff --git a/src/backend/access/heap/visibilitymap.c b/src/backend/access/heap/visibilitymap.c
index e21b96281a6..f1da52b2069 100644
--- a/src/backend/access/heap/visibilitymap.c
+++ b/src/backend/access/heap/visibilitymap.c
@@ -14,8 +14,7 @@
* visibilitymap_clear - clear bits for one page in the visibility map
* visibilitymap_pin - pin a map page for setting a bit
* visibilitymap_pin_ok - check whether correct map page is already pinned
- * visibilitymap_set - set bit(s) in a previously pinned page and log
- * visibilitymap_set_vmbits - set bit(s) in a pinned page
+ * visibilitymap_set - set bit(s) in a previously pinned page
* visibilitymap_get_status - get status of bits
* visibilitymap_count - count number of bits set in visibility map
* visibilitymap_prepare_truncate -
@@ -222,112 +221,11 @@ visibilitymap_pin_ok(BlockNumber heapBlk, Buffer vmbuf)
return BufferIsValid(vmbuf) && BufferGetBlockNumber(vmbuf) == mapBlock;
}
-/*
- * visibilitymap_set - set bit(s) on a previously pinned page
- *
- * recptr is the LSN of the XLOG record we're replaying, if we're in recovery,
- * or InvalidXLogRecPtr in normal running. The VM page LSN is advanced to the
- * one provided; in normal running, we generate a new XLOG record and set the
- * page LSN to that value (though the heap page's LSN may *not* be updated;
- * see below). cutoff_xid is the largest xmin on the page being marked
- * all-visible; it is needed for Hot Standby, and can be InvalidTransactionId
- * if the page contains no tuples. It can also be set to InvalidTransactionId
- * when a page that is already all-visible is being marked all-frozen.
- *
- * Caller is expected to set the heap page's PD_ALL_VISIBLE bit before calling
- * this function. Except in recovery, caller should also pass the heap
- * buffer. When checksums are enabled and we're not in recovery, we must add
- * the heap buffer to the WAL chain to protect it from being torn.
- *
- * You must pass a buffer containing the correct map page to this function.
- * Call visibilitymap_pin first to pin the right one. This function doesn't do
- * any I/O.
- */
-void
-visibilitymap_set(Relation rel, BlockNumber heapBlk, Buffer heapBuf,
- XLogRecPtr recptr, Buffer vmBuf, TransactionId cutoff_xid,
- uint8 flags)
-{
- BlockNumber mapBlock = HEAPBLK_TO_MAPBLOCK(heapBlk);
- uint32 mapByte = HEAPBLK_TO_MAPBYTE(heapBlk);
- uint8 mapOffset = HEAPBLK_TO_OFFSET(heapBlk);
- Page page;
- uint8 *map;
- uint8 status;
-
-#ifdef TRACE_VISIBILITYMAP
- elog(DEBUG1, "vm_set flags 0x%02X for %s %d",
- flags, RelationGetRelationName(rel), heapBlk);
-#endif
-
- Assert(InRecovery || !XLogRecPtrIsValid(recptr));
- Assert(InRecovery || PageIsAllVisible(BufferGetPage(heapBuf)));
- Assert((flags & VISIBILITYMAP_VALID_BITS) == flags);
-
- /* Must never set all_frozen bit without also setting all_visible bit */
- Assert(flags != VISIBILITYMAP_ALL_FROZEN);
-
- /* Check that we have the right heap page pinned, if present */
- if (BufferIsValid(heapBuf) && BufferGetBlockNumber(heapBuf) != heapBlk)
- elog(ERROR, "wrong heap buffer passed to visibilitymap_set");
-
- Assert(!BufferIsValid(heapBuf) ||
- BufferIsLockedByMeInMode(heapBuf, BUFFER_LOCK_EXCLUSIVE));
-
- /* Check that we have the right VM page pinned */
- if (!BufferIsValid(vmBuf) || BufferGetBlockNumber(vmBuf) != mapBlock)
- elog(ERROR, "wrong VM buffer passed to visibilitymap_set");
-
- page = BufferGetPage(vmBuf);
- map = (uint8 *) PageGetContents(page);
- LockBuffer(vmBuf, BUFFER_LOCK_EXCLUSIVE);
-
- status = (map[mapByte] >> mapOffset) & VISIBILITYMAP_VALID_BITS;
- if (flags != status)
- {
- START_CRIT_SECTION();
-
- map[mapByte] |= (flags << mapOffset);
- MarkBufferDirty(vmBuf);
-
- if (RelationNeedsWAL(rel))
- {
- if (!XLogRecPtrIsValid(recptr))
- {
- Assert(!InRecovery);
- recptr = log_heap_visible(rel, heapBuf, vmBuf, cutoff_xid, flags);
-
- /*
- * If data checksums are enabled (or wal_log_hints=on), we
- * need to protect the heap page from being torn.
- *
- * If not, then we must *not* update the heap page's LSN. In
- * this case, the FPI for the heap page was omitted from the
- * WAL record inserted above, so it would be incorrect to
- * update the heap page's LSN.
- */
- if (XLogHintBitIsNeeded())
- {
- Page heapPage = BufferGetPage(heapBuf);
-
- PageSetLSN(heapPage, recptr);
- }
- }
- PageSetLSN(page, recptr);
- }
-
- END_CRIT_SECTION();
- }
-
- LockBuffer(vmBuf, BUFFER_LOCK_UNLOCK);
-}
-
/*
* Set VM (visibility map) flags in the VM block in vmBuf.
*
* This function is intended for callers that log VM changes together
* with the heap page modifications that rendered the page all-visible.
- * Callers that log VM changes separately should use visibilitymap_set().
*
* vmBuf must be pinned and exclusively locked, and it must cover the VM bits
* corresponding to heapBlk.
@@ -343,9 +241,9 @@ visibilitymap_set(Relation rel, BlockNumber heapBlk, Buffer heapBuf,
* rlocator is used only for debugging messages.
*/
void
-visibilitymap_set_vmbits(BlockNumber heapBlk,
- Buffer vmBuf, uint8 flags,
- const RelFileLocator rlocator)
+visibilitymap_set(BlockNumber heapBlk,
+ Buffer vmBuf, uint8 flags,
+ const RelFileLocator rlocator)
{
BlockNumber mapBlock = HEAPBLK_TO_MAPBLOCK(heapBlk);
uint32 mapByte = HEAPBLK_TO_MAPBYTE(heapBlk);
diff --git a/src/backend/access/rmgrdesc/heapdesc.c b/src/backend/access/rmgrdesc/heapdesc.c
index 02ae91653c1..75ae6f9d375 100644
--- a/src/backend/access/rmgrdesc/heapdesc.c
+++ b/src/backend/access/rmgrdesc/heapdesc.c
@@ -349,13 +349,6 @@ heap2_desc(StringInfo buf, XLogReaderState *record)
}
}
}
- else if (info == XLOG_HEAP2_VISIBLE)
- {
- xl_heap_visible *xlrec = (xl_heap_visible *) rec;
-
- appendStringInfo(buf, "snapshotConflictHorizon: %u, flags: 0x%02X",
- xlrec->snapshotConflictHorizon, xlrec->flags);
- }
else if (info == XLOG_HEAP2_MULTI_INSERT)
{
xl_heap_multi_insert *xlrec = (xl_heap_multi_insert *) rec;
@@ -461,9 +454,6 @@ heap2_identify(uint8 info)
case XLOG_HEAP2_PRUNE_VACUUM_CLEANUP:
id = "PRUNE_VACUUM_CLEANUP";
break;
- case XLOG_HEAP2_VISIBLE:
- id = "VISIBLE";
- break;
case XLOG_HEAP2_MULTI_INSERT:
id = "MULTI_INSERT";
break;
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index 21f03864a66..3c027bcb2f7 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -448,7 +448,6 @@ heap2_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
case XLOG_HEAP2_PRUNE_ON_ACCESS:
case XLOG_HEAP2_PRUNE_VACUUM_SCAN:
case XLOG_HEAP2_PRUNE_VACUUM_CLEANUP:
- case XLOG_HEAP2_VISIBLE:
case XLOG_HEAP2_LOCK_UPDATED:
break;
default:
diff --git a/src/backend/storage/ipc/standby.c b/src/backend/storage/ipc/standby.c
index f3ad90c7c7a..66ed51a8aa1 100644
--- a/src/backend/storage/ipc/standby.c
+++ b/src/backend/storage/ipc/standby.c
@@ -477,12 +477,12 @@ ResolveRecoveryConflictWithSnapshot(TransactionId snapshotConflictHorizon,
* If we get passed InvalidTransactionId then we do nothing (no conflict).
*
* This can happen when replaying already-applied WAL records after a
- * standby crash or restart, or when replaying an XLOG_HEAP2_VISIBLE
- * record that marks as frozen a page which was already all-visible. It's
- * also quite common with records generated during index deletion
- * (original execution of the deletion can reason that a recovery conflict
- * which is sufficient for the deletion operation must take place before
- * replay of the deletion record itself).
+ * standby crash or restart, or when replaying a record that marks as
+ * frozen a page which was already marked all-visible in the visibility
+ * map. It's also quite common with records generated during index
+ * deletion (original execution of the deletion can reason that a recovery
+ * conflict which is sufficient for the deletion operation must take place
+ * before replay of the deletion record itself).
*/
if (!TransactionIdIsValid(snapshotConflictHorizon))
return;
diff --git a/src/include/access/heapam_xlog.h b/src/include/access/heapam_xlog.h
index ce3566ba949..5eed567a8e5 100644
--- a/src/include/access/heapam_xlog.h
+++ b/src/include/access/heapam_xlog.h
@@ -60,7 +60,6 @@
#define XLOG_HEAP2_PRUNE_ON_ACCESS 0x10
#define XLOG_HEAP2_PRUNE_VACUUM_SCAN 0x20
#define XLOG_HEAP2_PRUNE_VACUUM_CLEANUP 0x30
-#define XLOG_HEAP2_VISIBLE 0x40
#define XLOG_HEAP2_MULTI_INSERT 0x50
#define XLOG_HEAP2_LOCK_UPDATED 0x60
#define XLOG_HEAP2_NEW_CID 0x70
@@ -443,20 +442,6 @@ typedef struct xl_heap_inplace
#define MinSizeOfHeapInplace (offsetof(xl_heap_inplace, nmsgs) + sizeof(int))
-/*
- * This is what we need to know about setting a visibility map bit
- *
- * Backup blk 0: visibility map buffer
- * Backup blk 1: heap buffer
- */
-typedef struct xl_heap_visible
-{
- TransactionId snapshotConflictHorizon;
- uint8 flags;
-} xl_heap_visible;
-
-#define SizeOfHeapVisible (offsetof(xl_heap_visible, flags) + sizeof(uint8))
-
typedef struct xl_heap_new_cid
{
/*
@@ -500,11 +485,6 @@ extern void heap2_desc(StringInfo buf, XLogReaderState *record);
extern const char *heap2_identify(uint8 info);
extern void heap_xlog_logical_rewrite(XLogReaderState *r);
-extern XLogRecPtr log_heap_visible(Relation rel, Buffer heap_buffer,
- Buffer vm_buffer,
- TransactionId snapshotConflictHorizon,
- uint8 vmflags);
-
/* in heapdesc.c, so it can be shared between frontend/backend code */
extern void heap_xlog_deserialize_prune_and_freeze(char *cursor, uint16 flags,
int *nplans, xlhp_freeze_plan **plans,
diff --git a/src/include/access/visibilitymap.h b/src/include/access/visibilitymap.h
index 52cde56be86..e4e0cfa989e 100644
--- a/src/include/access/visibilitymap.h
+++ b/src/include/access/visibilitymap.h
@@ -15,7 +15,6 @@
#define VISIBILITYMAP_H
#include "access/visibilitymapdefs.h"
-#include "access/xlogdefs.h"
#include "storage/block.h"
#include "storage/buf.h"
#include "storage/relfilelocator.h"
@@ -32,15 +31,9 @@ extern bool visibilitymap_clear(Relation rel, BlockNumber heapBlk,
extern void visibilitymap_pin(Relation rel, BlockNumber heapBlk,
Buffer *vmbuf);
extern bool visibilitymap_pin_ok(BlockNumber heapBlk, Buffer vmbuf);
-extern void visibilitymap_set(Relation rel,
- BlockNumber heapBlk, Buffer heapBuf,
- XLogRecPtr recptr,
- Buffer vmBuf,
- TransactionId cutoff_xid,
- uint8 flags);
-extern void visibilitymap_set_vmbits(BlockNumber heapBlk,
- Buffer vmBuf, uint8 flags,
- const RelFileLocator rlocator);
+extern void visibilitymap_set(BlockNumber heapBlk,
+ Buffer vmBuf, uint8 flags,
+ const RelFileLocator rlocator);
extern uint8 visibilitymap_get_status(Relation rel, BlockNumber heapBlk, Buffer *vmbuf);
extern void visibilitymap_count(Relation rel, BlockNumber *all_visible, BlockNumber *all_frozen);
extern BlockNumber visibilitymap_prepare_truncate(Relation rel,
diff --git a/src/include/access/visibilitymapdefs.h b/src/include/access/visibilitymapdefs.h
index 89153b3cd9a..e5794c8559e 100644
--- a/src/include/access/visibilitymapdefs.h
+++ b/src/include/access/visibilitymapdefs.h
@@ -21,14 +21,5 @@
#define VISIBILITYMAP_ALL_FROZEN 0x02
#define VISIBILITYMAP_VALID_BITS 0x03 /* OR of all valid visibilitymap
* flags bits */
-/*
- * To detect recovery conflicts during logical decoding on a standby, we need
- * to know if a table is a user catalog table. For that we add an additional
- * bit into xl_heap_visible.flags, in addition to the above.
- *
- * NB: VISIBILITYMAP_XLOG_* may not be passed to visibilitymap_set().
- */
-#define VISIBILITYMAP_XLOG_CATALOG_REL 0x04
-#define VISIBILITYMAP_XLOG_VALID_BITS (VISIBILITYMAP_VALID_BITS | VISIBILITYMAP_XLOG_CATALOG_REL)
#endif /* VISIBILITYMAPDEFS_H */
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index ec8513d90b5..4c7ce9bd4b5 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -4389,7 +4389,6 @@ xl_heap_prune
xl_heap_rewrite_mapping
xl_heap_truncate
xl_heap_update
-xl_heap_visible
xl_invalid_page
xl_invalid_page_key
xl_invalidations
--
2.43.0
[text/x-patch] v38-0008-Track-which-relations-are-modified-by-a-query.patch (5.8K, 9-v38-0008-Track-which-relations-are-modified-by-a-query.patch)
download | inline diff:
From f707d345f3ee43a9b5e914e4d496c83485ea380b Mon Sep 17 00:00:00 2001
From: Melanie Plageman <[email protected]>
Date: Wed, 3 Dec 2025 15:07:24 -0500
Subject: [PATCH v38 08/12] Track which relations are modified by a query
Save the relids of modified relations in a bitmap in the executor state.
A later commit will pass this information down to scan nodes to control
whether or not on-access pruning is allowed to set the visibility map.
Setting the visibility map during a scan is counterproductive if the
query is going to modify the page immediately after.
Relations are considered modified if they are the target of INSERT,
UPDATE, DELETE, or MERGE, or if they have any row mark (including SELECT
FOR UPDATE/SHARE). All row mark types are included, even those which
don't actually modify tuples, because this bitmap is only used as a hint
to avoid unnecessary work.
Author: Melanie Plageman <[email protected]>
Reviewed-by: Andres Freund <[email protected]>
Reviewed-by: Chao Li <[email protected]>
Discussion: https://postgr.es/m/F5CDD1B5-628C-44A1-9F85-3958C626F6A9%40gmail.com
---
src/backend/executor/execMain.c | 18 ++++++++++++++++++
src/backend/executor/execUtils.c | 31 +++++++++++++++++++++++++++++++
src/include/executor/executor.h | 3 +++
src/include/nodes/execnodes.h | 6 ++++++
4 files changed, 58 insertions(+)
diff --git a/src/backend/executor/execMain.c b/src/backend/executor/execMain.c
index bfd3ebc601e..57dcdeda056 100644
--- a/src/backend/executor/execMain.c
+++ b/src/backend/executor/execMain.c
@@ -920,6 +920,10 @@ InitPlan(QueryDesc *queryDesc, int eflags)
break;
}
+ /* If it has a rowmark, the relation may be modified */
+ estate->es_modified_relids = bms_add_member(estate->es_modified_relids,
+ rc->rti);
+
/* Check that relation is a legal target for marking */
if (relation)
CheckValidRowMarkRel(relation, rc->markType);
@@ -990,6 +994,10 @@ InitPlan(QueryDesc *queryDesc, int eflags)
*/
planstate = ExecInitNode(plan, estate, eflags);
+#ifdef USE_ASSERT_CHECKING
+ CrossCheckModifiedRelids(estate);
+#endif
+
/*
* Get the tuple descriptor describing the type of tuples to return.
*/
@@ -3033,6 +3041,12 @@ EvalPlanQualStart(EPQState *epqstate, Plan *planTree)
rcestate->es_output_cid = parentestate->es_output_cid;
rcestate->es_queryEnv = parentestate->es_queryEnv;
+ /*
+ * Use a deep copy to avoid stale pointers since bms_add_member() may
+ * reallocate the bitmap.
+ */
+ rcestate->es_modified_relids = bms_copy(parentestate->es_modified_relids);
+
/*
* ResultRelInfos needed by subplans are initialized from scratch when the
* subplans themselves are initialized.
@@ -3165,6 +3179,10 @@ EvalPlanQualStart(EPQState *epqstate, Plan *planTree)
*/
epqstate->recheckplanstate = ExecInitNode(planTree, rcestate, 0);
+#ifdef USE_ASSERT_CHECKING
+ CrossCheckModifiedRelids(rcestate);
+#endif
+
MemoryContextSwitchTo(oldcontext);
}
diff --git a/src/backend/executor/execUtils.c b/src/backend/executor/execUtils.c
index cd4d5452cfb..0f8364b8720 100644
--- a/src/backend/executor/execUtils.c
+++ b/src/backend/executor/execUtils.c
@@ -123,6 +123,8 @@ CreateExecutorState(void)
estate->es_part_prune_results = NIL;
estate->es_unpruned_relids = NULL;
+ estate->es_modified_relids = NULL;
+
estate->es_junkFilter = NULL;
estate->es_output_cid = (CommandId) 0;
@@ -871,6 +873,33 @@ ExecGetRangeTableRelation(EState *estate, Index rti, bool isResultRel)
return rel;
}
+#ifdef USE_ASSERT_CHECKING
+/*
+ * Assert that es_modified_relids includes all potentially modified RT
+ * indexes.
+ */
+void
+CrossCheckModifiedRelids(EState *estate)
+{
+ Bitmapset *expected = NULL;
+ ListCell *lc;
+
+ foreach(lc, estate->es_opened_result_relations)
+ {
+ ResultRelInfo *rri = lfirst_node(ResultRelInfo, lc);
+
+ expected = bms_add_member(expected, rri->ri_RangeTableIndex);
+ }
+ if (estate->es_rowmarks)
+ {
+ for (Index rti = 1; rti <= estate->es_range_table_size; rti++)
+ if (estate->es_rowmarks[rti - 1] != NULL)
+ expected = bms_add_member(expected, rti);
+ }
+ Assert(bms_is_subset(expected, estate->es_modified_relids));
+}
+#endif
+
/*
* ExecInitResultRelation
* Open relation given by the passed-in RT index and fill its
@@ -896,6 +925,8 @@ ExecInitResultRelation(EState *estate, ResultRelInfo *resultRelInfo,
estate->es_result_relations = (ResultRelInfo **)
palloc0(estate->es_range_table_size * sizeof(ResultRelInfo *));
estate->es_result_relations[rti - 1] = resultRelInfo;
+ estate->es_modified_relids = bms_add_member(estate->es_modified_relids,
+ rti);
/*
* Saving in the list allows to avoid needlessly traversing the whole
diff --git a/src/include/executor/executor.h b/src/include/executor/executor.h
index 82c442d23f8..1411d5276ca 100644
--- a/src/include/executor/executor.h
+++ b/src/include/executor/executor.h
@@ -705,6 +705,9 @@ extern Relation ExecGetRangeTableRelation(EState *estate, Index rti,
bool isResultRel);
extern void ExecInitResultRelation(EState *estate, ResultRelInfo *resultRelInfo,
Index rti);
+#ifdef USE_ASSERT_CHECKING
+extern void CrossCheckModifiedRelids(EState *estate);
+#endif
extern int executor_errposition(EState *estate, int location);
diff --git a/src/include/nodes/execnodes.h b/src/include/nodes/execnodes.h
index 63c067d5aae..610385df12b 100644
--- a/src/include/nodes/execnodes.h
+++ b/src/include/nodes/execnodes.h
@@ -679,6 +679,12 @@ typedef struct EState
* ExecDoInitialPruning() */
const char *es_sourceText; /* Source text from QueryDesc */
+ /*
+ * RT indexes of relations modified by the query through a
+ * UPDATE/DELETE/INSERT/MERGE or targeted by a SELECT FOR UPDATE.
+ */
+ Bitmapset *es_modified_relids;
+
JunkFilter *es_junkFilter; /* top-level junk filter, if any */
/* If query can insert/delete tuples, the command ID to mark them with */
--
2.43.0
[text/x-patch] v38-0009-Thread-flags-through-begin-scan-APIs.patch (28.1K, 10-v38-0009-Thread-flags-through-begin-scan-APIs.patch)
download | inline diff:
From e501ec27844ae056c9d5b0439e327ded450c9ce2 Mon Sep 17 00:00:00 2001
From: Melanie Plageman <[email protected]>
Date: Mon, 2 Mar 2026 16:31:17 -0500
Subject: [PATCH v38 09/12] Thread flags through begin-scan APIs
Add a flags parameter to the index_fetch_begin() table AM callback and
the begin-scan helpers so the executor can pass context for building
scan descriptors. This introduces an extension point for follow-up work
to mark relations as read-only for the current query, without changing
behavior in this patch.
Author: Melanie Plageman <[email protected]>
Reviewed-by: Chao Li <[email protected]>
Discussion: https://postgr.es/m/F5CDD1B5-628C-44A1-9F85-3958C626F6A9%40gmail.com
---
contrib/pgrowlocks/pgrowlocks.c | 2 +-
src/backend/access/brin/brin.c | 3 +-
src/backend/access/gin/gininsert.c | 3 +-
src/backend/access/heap/heapam_handler.c | 6 ++--
src/backend/access/index/genam.c | 4 +--
src/backend/access/index/indexam.c | 8 +++---
src/backend/access/nbtree/nbtsort.c | 2 +-
src/backend/access/table/tableam.c | 13 +++++----
src/backend/commands/constraint.c | 2 +-
src/backend/commands/copyto.c | 2 +-
src/backend/commands/tablecmds.c | 8 +++---
src/backend/commands/typecmds.c | 4 +--
src/backend/executor/execIndexing.c | 2 +-
src/backend/executor/execReplication.c | 8 +++---
src/backend/executor/nodeBitmapHeapscan.c | 2 +-
src/backend/executor/nodeIndexonlyscan.c | 6 ++--
src/backend/executor/nodeIndexscan.c | 8 +++---
src/backend/executor/nodeSamplescan.c | 2 +-
src/backend/executor/nodeSeqscan.c | 6 ++--
src/backend/executor/nodeTidrangescan.c | 6 ++--
src/backend/partitioning/partbounds.c | 2 +-
src/backend/utils/adt/selfuncs.c | 2 +-
src/include/access/genam.h | 5 ++--
src/include/access/heapam.h | 5 ++--
src/include/access/tableam.h | 35 ++++++++++++++---------
25 files changed, 81 insertions(+), 65 deletions(-)
diff --git a/contrib/pgrowlocks/pgrowlocks.c b/contrib/pgrowlocks/pgrowlocks.c
index f88269332b6..27f01d8055f 100644
--- a/contrib/pgrowlocks/pgrowlocks.c
+++ b/contrib/pgrowlocks/pgrowlocks.c
@@ -114,7 +114,7 @@ pgrowlocks(PG_FUNCTION_ARGS)
RelationGetRelationName(rel));
/* Scan the relation */
- scan = table_beginscan(rel, GetActiveSnapshot(), 0, NULL);
+ scan = table_beginscan(rel, GetActiveSnapshot(), 0, NULL, 0);
hscan = (HeapScanDesc) scan;
attinmeta = TupleDescGetAttInMetadata(rsinfo->setDesc);
diff --git a/src/backend/access/brin/brin.c b/src/backend/access/brin/brin.c
index 1909c3254b5..a221e032f5d 100644
--- a/src/backend/access/brin/brin.c
+++ b/src/backend/access/brin/brin.c
@@ -2842,7 +2842,8 @@ _brin_parallel_scan_and_build(BrinBuildState *state,
indexInfo->ii_Concurrent = brinshared->isconcurrent;
scan = table_beginscan_parallel(heap,
- ParallelTableScanFromBrinShared(brinshared));
+ ParallelTableScanFromBrinShared(brinshared),
+ 0);
reltuples = table_index_build_scan(heap, index, indexInfo, true, true,
brinbuildCallbackParallel, state, scan);
diff --git a/src/backend/access/gin/gininsert.c b/src/backend/access/gin/gininsert.c
index 97cea5f7d4e..74243efa74f 100644
--- a/src/backend/access/gin/gininsert.c
+++ b/src/backend/access/gin/gininsert.c
@@ -2065,7 +2065,8 @@ _gin_parallel_scan_and_build(GinBuildState *state,
indexInfo->ii_Concurrent = ginshared->isconcurrent;
scan = table_beginscan_parallel(heap,
- ParallelTableScanFromGinBuildShared(ginshared));
+ ParallelTableScanFromGinBuildShared(ginshared),
+ 0);
reltuples = table_index_build_scan(heap, index, indexInfo, true, progress,
ginBuildCallbackParallel, state, scan);
diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c
index 42bf73d3138..6122603d11e 100644
--- a/src/backend/access/heap/heapam_handler.c
+++ b/src/backend/access/heap/heapam_handler.c
@@ -79,7 +79,7 @@ heapam_slot_callbacks(Relation relation)
*/
static IndexFetchTableData *
-heapam_index_fetch_begin(Relation rel)
+heapam_index_fetch_begin(Relation rel, uint32 flags)
{
IndexFetchHeapData *hscan = palloc0_object(IndexFetchHeapData);
@@ -761,7 +761,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
tableScan = NULL;
heapScan = NULL;
- indexScan = index_beginscan(OldHeap, OldIndex, SnapshotAny, NULL, 0, 0);
+ indexScan = index_beginscan(OldHeap, OldIndex, SnapshotAny, NULL, 0, 0, 0);
index_rescan(indexScan, NULL, 0, NULL, 0);
}
else
@@ -770,7 +770,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
pgstat_progress_update_param(PROGRESS_REPACK_PHASE,
PROGRESS_REPACK_PHASE_SEQ_SCAN_HEAP);
- tableScan = table_beginscan(OldHeap, SnapshotAny, 0, (ScanKey) NULL);
+ tableScan = table_beginscan(OldHeap, SnapshotAny, 0, (ScanKey) NULL, 0);
heapScan = (HeapScanDesc) tableScan;
indexScan = NULL;
diff --git a/src/backend/access/index/genam.c b/src/backend/access/index/genam.c
index 5e89b86a62c..1fe7ffb2487 100644
--- a/src/backend/access/index/genam.c
+++ b/src/backend/access/index/genam.c
@@ -455,7 +455,7 @@ systable_beginscan(Relation heapRelation,
}
sysscan->iscan = index_beginscan(heapRelation, irel,
- snapshot, NULL, nkeys, 0);
+ snapshot, NULL, nkeys, 0, 0);
index_rescan(sysscan->iscan, idxkey, nkeys, NULL, 0);
sysscan->scan = NULL;
@@ -716,7 +716,7 @@ systable_beginscan_ordered(Relation heapRelation,
bsysscan = true;
sysscan->iscan = index_beginscan(heapRelation, indexRelation,
- snapshot, NULL, nkeys, 0);
+ snapshot, NULL, nkeys, 0, 0);
index_rescan(sysscan->iscan, idxkey, nkeys, NULL, 0);
sysscan->scan = NULL;
diff --git a/src/backend/access/index/indexam.c b/src/backend/access/index/indexam.c
index 5eb7e99ad3e..87219613f0b 100644
--- a/src/backend/access/index/indexam.c
+++ b/src/backend/access/index/indexam.c
@@ -257,7 +257,7 @@ index_beginscan(Relation heapRelation,
Relation indexRelation,
Snapshot snapshot,
IndexScanInstrumentation *instrument,
- int nkeys, int norderbys)
+ int nkeys, int norderbys, uint32 flags)
{
IndexScanDesc scan;
@@ -284,7 +284,7 @@ index_beginscan(Relation heapRelation,
scan->instrument = instrument;
/* prepare to fetch index matches from table */
- scan->xs_heapfetch = table_index_fetch_begin(heapRelation);
+ scan->xs_heapfetch = table_index_fetch_begin(heapRelation, flags);
return scan;
}
@@ -593,7 +593,7 @@ IndexScanDesc
index_beginscan_parallel(Relation heaprel, Relation indexrel,
IndexScanInstrumentation *instrument,
int nkeys, int norderbys,
- ParallelIndexScanDesc pscan)
+ ParallelIndexScanDesc pscan, uint32 flags)
{
Snapshot snapshot;
IndexScanDesc scan;
@@ -615,7 +615,7 @@ index_beginscan_parallel(Relation heaprel, Relation indexrel,
scan->instrument = instrument;
/* prepare to fetch index matches from table */
- scan->xs_heapfetch = table_index_fetch_begin(heaprel);
+ scan->xs_heapfetch = table_index_fetch_begin(heaprel, flags);
return scan;
}
diff --git a/src/backend/access/nbtree/nbtsort.c b/src/backend/access/nbtree/nbtsort.c
index 69ef1527e06..bc4eedba4ac 100644
--- a/src/backend/access/nbtree/nbtsort.c
+++ b/src/backend/access/nbtree/nbtsort.c
@@ -1927,7 +1927,7 @@ _bt_parallel_scan_and_sort(BTSpool *btspool, BTSpool *btspool2,
indexInfo = BuildIndexInfo(btspool->index);
indexInfo->ii_Concurrent = btshared->isconcurrent;
scan = table_beginscan_parallel(btspool->heap,
- ParallelTableScanFromBTShared(btshared));
+ ParallelTableScanFromBTShared(btshared), 0);
reltuples = table_index_build_scan(btspool->heap, btspool->index, indexInfo,
true, progress, _bt_build_callback,
&buildstate, scan);
diff --git a/src/backend/access/table/tableam.c b/src/backend/access/table/tableam.c
index dfda1af412e..e946cfb393a 100644
--- a/src/backend/access/table/tableam.c
+++ b/src/backend/access/table/tableam.c
@@ -163,10 +163,11 @@ table_parallelscan_initialize(Relation rel, ParallelTableScanDesc pscan,
}
TableScanDesc
-table_beginscan_parallel(Relation relation, ParallelTableScanDesc pscan)
+table_beginscan_parallel(Relation relation, ParallelTableScanDesc pscan, uint32 flags)
{
Snapshot snapshot;
- uint32 flags = SO_TYPE_SEQSCAN |
+
+ flags |= SO_TYPE_SEQSCAN |
SO_ALLOW_STRAT | SO_ALLOW_SYNC | SO_ALLOW_PAGEMODE;
Assert(RelFileLocatorEquals(relation->rd_locator, pscan->phs_locator));
@@ -190,12 +191,14 @@ table_beginscan_parallel(Relation relation, ParallelTableScanDesc pscan)
TableScanDesc
table_beginscan_parallel_tidrange(Relation relation,
- ParallelTableScanDesc pscan)
+ ParallelTableScanDesc pscan,
+ uint32 flags)
{
Snapshot snapshot;
- uint32 flags = SO_TYPE_TIDRANGESCAN | SO_ALLOW_PAGEMODE;
TableScanDesc sscan;
+ flags |= SO_TYPE_TIDRANGESCAN | SO_ALLOW_PAGEMODE;
+
Assert(RelFileLocatorEquals(relation->rd_locator, pscan->phs_locator));
/* disable syncscan in parallel tid range scan. */
@@ -248,7 +251,7 @@ table_index_fetch_tuple_check(Relation rel,
bool found;
slot = table_slot_create(rel, NULL);
- scan = table_index_fetch_begin(rel);
+ scan = table_index_fetch_begin(rel, 0);
found = table_index_fetch_tuple(scan, tid, snapshot, slot, &call_again,
all_dead);
table_index_fetch_end(scan);
diff --git a/src/backend/commands/constraint.c b/src/backend/commands/constraint.c
index cc11c47b6f2..37cfbd63938 100644
--- a/src/backend/commands/constraint.c
+++ b/src/backend/commands/constraint.c
@@ -106,7 +106,7 @@ unique_key_recheck(PG_FUNCTION_ARGS)
*/
tmptid = checktid;
{
- IndexFetchTableData *scan = table_index_fetch_begin(trigdata->tg_relation);
+ IndexFetchTableData *scan = table_index_fetch_begin(trigdata->tg_relation, 0);
bool call_again = false;
if (!table_index_fetch_tuple(scan, &tmptid, SnapshotSelf, slot,
diff --git a/src/backend/commands/copyto.c b/src/backend/commands/copyto.c
index d6ef7275a64..900199dbe29 100644
--- a/src/backend/commands/copyto.c
+++ b/src/backend/commands/copyto.c
@@ -1159,7 +1159,7 @@ CopyRelationTo(CopyToState cstate, Relation rel, Relation root_rel, uint64 *proc
AttrMap *map = NULL;
TupleTableSlot *root_slot = NULL;
- scandesc = table_beginscan(rel, GetActiveSnapshot(), 0, NULL);
+ scandesc = table_beginscan(rel, GetActiveSnapshot(), 0, NULL, 0);
slot = table_slot_create(rel, NULL);
/*
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index cd6d720386f..0455b36c41e 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -6396,7 +6396,7 @@ ATRewriteTable(AlteredTableInfo *tab, Oid OIDNewHeap)
* checking all the constraints.
*/
snapshot = RegisterSnapshot(GetLatestSnapshot());
- scan = table_beginscan(oldrel, snapshot, 0, NULL);
+ scan = table_beginscan(oldrel, snapshot, 0, NULL, 0);
/*
* Switch to per-tuple memory context and reset it for each tuple
@@ -13965,7 +13965,7 @@ validateForeignKeyConstraint(char *conname,
*/
snapshot = RegisterSnapshot(GetLatestSnapshot());
slot = table_slot_create(rel, NULL);
- scan = table_beginscan(rel, snapshot, 0, NULL);
+ scan = table_beginscan(rel, snapshot, 0, NULL, 0);
perTupCxt = AllocSetContextCreate(CurrentMemoryContext,
"validateForeignKeyConstraint",
@@ -22867,7 +22867,7 @@ MergePartitionsMoveRows(List **wqueue, List *mergingPartitions, Relation newPart
/* Scan through the rows. */
snapshot = RegisterSnapshot(GetLatestSnapshot());
- scan = table_beginscan(mergingPartition, snapshot, 0, NULL);
+ scan = table_beginscan(mergingPartition, snapshot, 0, NULL, 0);
/*
* Switch to per-tuple memory context and reset it for each tuple
@@ -23331,7 +23331,7 @@ SplitPartitionMoveRows(List **wqueue, Relation rel, Relation splitRel,
/* Scan through the rows. */
snapshot = RegisterSnapshot(GetLatestSnapshot());
- scan = table_beginscan(splitRel, snapshot, 0, NULL);
+ scan = table_beginscan(splitRel, snapshot, 0, NULL, 0);
/*
* Switch to per-tuple memory context and reset it for each tuple
diff --git a/src/backend/commands/typecmds.c b/src/backend/commands/typecmds.c
index 3dab6bb5a79..5316cea7cec 100644
--- a/src/backend/commands/typecmds.c
+++ b/src/backend/commands/typecmds.c
@@ -3185,7 +3185,7 @@ validateDomainNotNullConstraint(Oid domainoid)
/* Scan all tuples in this relation */
snapshot = RegisterSnapshot(GetLatestSnapshot());
- scan = table_beginscan(testrel, snapshot, 0, NULL);
+ scan = table_beginscan(testrel, snapshot, 0, NULL, 0);
slot = table_slot_create(testrel, NULL);
while (table_scan_getnextslot(scan, ForwardScanDirection, slot))
{
@@ -3266,7 +3266,7 @@ validateDomainCheckConstraint(Oid domainoid, const char *ccbin, LOCKMODE lockmod
/* Scan all tuples in this relation */
snapshot = RegisterSnapshot(GetLatestSnapshot());
- scan = table_beginscan(testrel, snapshot, 0, NULL);
+ scan = table_beginscan(testrel, snapshot, 0, NULL, 0);
slot = table_slot_create(testrel, NULL);
while (table_scan_getnextslot(scan, ForwardScanDirection, slot))
{
diff --git a/src/backend/executor/execIndexing.c b/src/backend/executor/execIndexing.c
index 9d071e495c6..cb3e4f67ea1 100644
--- a/src/backend/executor/execIndexing.c
+++ b/src/backend/executor/execIndexing.c
@@ -815,7 +815,7 @@ check_exclusion_or_unique_constraint(Relation heap, Relation index,
retry:
conflict = false;
found_self = false;
- index_scan = index_beginscan(heap, index, &DirtySnapshot, NULL, indnkeyatts, 0);
+ index_scan = index_beginscan(heap, index, &DirtySnapshot, NULL, indnkeyatts, 0, 0);
index_rescan(index_scan, scankeys, indnkeyatts, NULL, 0);
while (index_getnext_slot(index_scan, ForwardScanDirection, existing_slot))
diff --git a/src/backend/executor/execReplication.c b/src/backend/executor/execReplication.c
index 2497ee7edc5..5b8ca1abf62 100644
--- a/src/backend/executor/execReplication.c
+++ b/src/backend/executor/execReplication.c
@@ -205,7 +205,7 @@ RelationFindReplTupleByIndex(Relation rel, Oid idxoid,
skey_attoff = build_replindex_scan_key(skey, rel, idxrel, searchslot);
/* Start an index scan. */
- scan = index_beginscan(rel, idxrel, &snap, NULL, skey_attoff, 0);
+ scan = index_beginscan(rel, idxrel, &snap, NULL, skey_attoff, 0, 0);
retry:
found = false;
@@ -383,7 +383,7 @@ RelationFindReplTupleSeq(Relation rel, LockTupleMode lockmode,
/* Start a heap scan. */
InitDirtySnapshot(snap);
- scan = table_beginscan(rel, &snap, 0, NULL);
+ scan = table_beginscan(rel, &snap, 0, NULL, 0);
scanslot = table_slot_create(rel, NULL);
retry:
@@ -602,7 +602,7 @@ RelationFindDeletedTupleInfoSeq(Relation rel, TupleTableSlot *searchslot,
* not yet committed or those just committed prior to the scan are
* excluded in update_most_recent_deletion_info().
*/
- scan = table_beginscan(rel, SnapshotAny, 0, NULL);
+ scan = table_beginscan(rel, SnapshotAny, 0, NULL, 0);
scanslot = table_slot_create(rel, NULL);
table_rescan(scan, NULL);
@@ -666,7 +666,7 @@ RelationFindDeletedTupleInfoByIndex(Relation rel, Oid idxoid,
* not yet committed or those just committed prior to the scan are
* excluded in update_most_recent_deletion_info().
*/
- scan = index_beginscan(rel, idxrel, SnapshotAny, NULL, skey_attoff, 0);
+ scan = index_beginscan(rel, idxrel, SnapshotAny, NULL, skey_attoff, 0, 0);
index_rescan(scan, skey, skey_attoff, NULL, 0);
diff --git a/src/backend/executor/nodeBitmapHeapscan.c b/src/backend/executor/nodeBitmapHeapscan.c
index 74eac93284e..620fc7e259a 100644
--- a/src/backend/executor/nodeBitmapHeapscan.c
+++ b/src/backend/executor/nodeBitmapHeapscan.c
@@ -108,7 +108,7 @@ BitmapTableScanSetup(BitmapHeapScanState *node)
table_beginscan_bm(node->ss.ss_currentRelation,
node->ss.ps.state->es_snapshot,
0,
- NULL);
+ NULL, 0);
}
node->ss.ss_currentScanDesc->st.rs_tbmiterator = tbmiterator;
diff --git a/src/backend/executor/nodeIndexonlyscan.c b/src/backend/executor/nodeIndexonlyscan.c
index 9e8ea8ddf22..aefb792ee6e 100644
--- a/src/backend/executor/nodeIndexonlyscan.c
+++ b/src/backend/executor/nodeIndexonlyscan.c
@@ -94,7 +94,7 @@ IndexOnlyNext(IndexOnlyScanState *node)
estate->es_snapshot,
&node->ioss_Instrument,
node->ioss_NumScanKeys,
- node->ioss_NumOrderByKeys);
+ node->ioss_NumOrderByKeys, 0);
node->ioss_ScanDesc = scandesc;
@@ -788,7 +788,7 @@ ExecIndexOnlyScanInitializeDSM(IndexOnlyScanState *node,
&node->ioss_Instrument,
node->ioss_NumScanKeys,
node->ioss_NumOrderByKeys,
- piscan);
+ piscan, 0);
node->ioss_ScanDesc->xs_want_itup = true;
node->ioss_VMBuffer = InvalidBuffer;
@@ -854,7 +854,7 @@ ExecIndexOnlyScanInitializeWorker(IndexOnlyScanState *node,
&node->ioss_Instrument,
node->ioss_NumScanKeys,
node->ioss_NumOrderByKeys,
- piscan);
+ piscan, 0);
node->ioss_ScanDesc->xs_want_itup = true;
/*
diff --git a/src/backend/executor/nodeIndexscan.c b/src/backend/executor/nodeIndexscan.c
index 4513b1f7a90..477cd4fcf99 100644
--- a/src/backend/executor/nodeIndexscan.c
+++ b/src/backend/executor/nodeIndexscan.c
@@ -111,7 +111,7 @@ IndexNext(IndexScanState *node)
estate->es_snapshot,
&node->iss_Instrument,
node->iss_NumScanKeys,
- node->iss_NumOrderByKeys);
+ node->iss_NumOrderByKeys, 0);
node->iss_ScanDesc = scandesc;
@@ -207,7 +207,7 @@ IndexNextWithReorder(IndexScanState *node)
estate->es_snapshot,
&node->iss_Instrument,
node->iss_NumScanKeys,
- node->iss_NumOrderByKeys);
+ node->iss_NumOrderByKeys, 0);
node->iss_ScanDesc = scandesc;
@@ -1723,7 +1723,7 @@ ExecIndexScanInitializeDSM(IndexScanState *node,
&node->iss_Instrument,
node->iss_NumScanKeys,
node->iss_NumOrderByKeys,
- piscan);
+ piscan, 0);
/*
* If no run-time keys to calculate or they are ready, go ahead and pass
@@ -1787,7 +1787,7 @@ ExecIndexScanInitializeWorker(IndexScanState *node,
&node->iss_Instrument,
node->iss_NumScanKeys,
node->iss_NumOrderByKeys,
- piscan);
+ piscan, 0);
/*
* If no run-time keys to calculate or they are ready, go ahead and pass
diff --git a/src/backend/executor/nodeSamplescan.c b/src/backend/executor/nodeSamplescan.c
index 1b0af70fd7a..47660baf2fa 100644
--- a/src/backend/executor/nodeSamplescan.c
+++ b/src/backend/executor/nodeSamplescan.c
@@ -297,7 +297,7 @@ tablesample_init(SampleScanState *scanstate)
0, NULL,
scanstate->use_bulkread,
allow_sync,
- scanstate->use_pagemode);
+ scanstate->use_pagemode, 0);
}
else
{
diff --git a/src/backend/executor/nodeSeqscan.c b/src/backend/executor/nodeSeqscan.c
index af3c788ce8b..d9d7ec0516a 100644
--- a/src/backend/executor/nodeSeqscan.c
+++ b/src/backend/executor/nodeSeqscan.c
@@ -71,7 +71,7 @@ SeqNext(SeqScanState *node)
*/
scandesc = table_beginscan(node->ss.ss_currentRelation,
estate->es_snapshot,
- 0, NULL);
+ 0, NULL, 0);
node->ss.ss_currentScanDesc = scandesc;
}
@@ -374,7 +374,7 @@ ExecSeqScanInitializeDSM(SeqScanState *node,
estate->es_snapshot);
shm_toc_insert(pcxt->toc, node->ss.ps.plan->plan_node_id, pscan);
node->ss.ss_currentScanDesc =
- table_beginscan_parallel(node->ss.ss_currentRelation, pscan);
+ table_beginscan_parallel(node->ss.ss_currentRelation, pscan, 0);
}
/* ----------------------------------------------------------------
@@ -407,5 +407,5 @@ ExecSeqScanInitializeWorker(SeqScanState *node,
pscan = shm_toc_lookup(pwcxt->toc, node->ss.ps.plan->plan_node_id, false);
node->ss.ss_currentScanDesc =
- table_beginscan_parallel(node->ss.ss_currentRelation, pscan);
+ table_beginscan_parallel(node->ss.ss_currentRelation, pscan, 0);
}
diff --git a/src/backend/executor/nodeTidrangescan.c b/src/backend/executor/nodeTidrangescan.c
index 503817da65b..461edb8893b 100644
--- a/src/backend/executor/nodeTidrangescan.c
+++ b/src/backend/executor/nodeTidrangescan.c
@@ -245,7 +245,7 @@ TidRangeNext(TidRangeScanState *node)
scandesc = table_beginscan_tidrange(node->ss.ss_currentRelation,
estate->es_snapshot,
&node->trss_mintid,
- &node->trss_maxtid);
+ &node->trss_maxtid, 0);
node->ss.ss_currentScanDesc = scandesc;
}
else
@@ -459,7 +459,7 @@ ExecTidRangeScanInitializeDSM(TidRangeScanState *node, ParallelContext *pcxt)
shm_toc_insert(pcxt->toc, node->ss.ps.plan->plan_node_id, pscan);
node->ss.ss_currentScanDesc =
table_beginscan_parallel_tidrange(node->ss.ss_currentRelation,
- pscan);
+ pscan, 0);
}
/* ----------------------------------------------------------------
@@ -493,5 +493,5 @@ ExecTidRangeScanInitializeWorker(TidRangeScanState *node,
pscan = shm_toc_lookup(pwcxt->toc, node->ss.ps.plan->plan_node_id, false);
node->ss.ss_currentScanDesc =
table_beginscan_parallel_tidrange(node->ss.ss_currentRelation,
- pscan);
+ pscan, 0);
}
diff --git a/src/backend/partitioning/partbounds.c b/src/backend/partitioning/partbounds.c
index 0ca312ac27d..b7c4e6d1071 100644
--- a/src/backend/partitioning/partbounds.c
+++ b/src/backend/partitioning/partbounds.c
@@ -3362,7 +3362,7 @@ check_default_partition_contents(Relation parent, Relation default_rel,
econtext = GetPerTupleExprContext(estate);
snapshot = RegisterSnapshot(GetLatestSnapshot());
tupslot = table_slot_create(part_rel, &estate->es_tupleTable);
- scan = table_beginscan(part_rel, snapshot, 0, NULL);
+ scan = table_beginscan(part_rel, snapshot, 0, NULL, 0);
/*
* Switch to per-tuple memory context and reset it for each tuple
diff --git a/src/backend/utils/adt/selfuncs.c b/src/backend/utils/adt/selfuncs.c
index d4da0e8dea9..5b2165c267d 100644
--- a/src/backend/utils/adt/selfuncs.c
+++ b/src/backend/utils/adt/selfuncs.c
@@ -7161,7 +7161,7 @@ get_actual_variable_endpoint(Relation heapRel,
index_scan = index_beginscan(heapRel, indexRel,
&SnapshotNonVacuumable, NULL,
- 1, 0);
+ 1, 0, 0);
/* Set it up for index-only scan */
index_scan->xs_want_itup = true;
index_rescan(index_scan, scankeys, 1, NULL, 0);
diff --git a/src/include/access/genam.h b/src/include/access/genam.h
index 4c0429cc613..9abcc99d6c8 100644
--- a/src/include/access/genam.h
+++ b/src/include/access/genam.h
@@ -156,7 +156,7 @@ extern IndexScanDesc index_beginscan(Relation heapRelation,
Relation indexRelation,
Snapshot snapshot,
IndexScanInstrumentation *instrument,
- int nkeys, int norderbys);
+ int nkeys, int norderbys, uint32 flags);
extern IndexScanDesc index_beginscan_bitmap(Relation indexRelation,
Snapshot snapshot,
IndexScanInstrumentation *instrument,
@@ -182,7 +182,8 @@ extern IndexScanDesc index_beginscan_parallel(Relation heaprel,
Relation indexrel,
IndexScanInstrumentation *instrument,
int nkeys, int norderbys,
- ParallelIndexScanDesc pscan);
+ ParallelIndexScanDesc pscan,
+ uint32 flags);
extern ItemPointer index_getnext_tid(IndexScanDesc scan,
ScanDirection direction);
extern bool index_fetch_heap(IndexScanDesc scan, TupleTableSlot *slot);
diff --git a/src/include/access/heapam.h b/src/include/access/heapam.h
index f77a00291bb..c2621dc2fac 100644
--- a/src/include/access/heapam.h
+++ b/src/include/access/heapam.h
@@ -95,8 +95,9 @@ typedef struct HeapScanDescData
ParallelBlockTableScanWorkerData *rs_parallelworkerdata;
/*
- * For sequential scans and bitmap heap scans. The current heap block's
- * corresponding page in the visibility map.
+ * For sequential scans, bitmap heap scans, TID range scans, and sample
+ * scans. The current heap block's corresponding page in the visibility
+ * map.
*/
Buffer rs_vmbuffer;
diff --git a/src/include/access/tableam.h b/src/include/access/tableam.h
index 06084752245..f1065e30638 100644
--- a/src/include/access/tableam.h
+++ b/src/include/access/tableam.h
@@ -418,9 +418,12 @@ typedef struct TableAmRoutine
* IndexFetchTableData, which the AM will typically embed in a larger
* structure with additional information.
*
+ * 'flags' is a bitmask of SO_* flags providing hints from the executor
+ * about the scan context.
+ *
* Tuples for an index scan can then be fetched via index_fetch_tuple.
*/
- struct IndexFetchTableData *(*index_fetch_begin) (Relation rel);
+ struct IndexFetchTableData *(*index_fetch_begin) (Relation rel, uint32 flags);
/*
* Reset index fetch. Typically this will release cross index fetch
@@ -894,9 +897,9 @@ table_beginscan_common(Relation rel, Snapshot snapshot, int nkeys,
*/
static inline TableScanDesc
table_beginscan(Relation rel, Snapshot snapshot,
- int nkeys, ScanKeyData *key)
+ int nkeys, ScanKeyData *key, uint32 flags)
{
- uint32 flags = SO_TYPE_SEQSCAN |
+ flags |= SO_TYPE_SEQSCAN |
SO_ALLOW_STRAT | SO_ALLOW_SYNC | SO_ALLOW_PAGEMODE;
return table_beginscan_common(rel, snapshot, nkeys, key, NULL, flags);
@@ -939,9 +942,9 @@ table_beginscan_strat(Relation rel, Snapshot snapshot,
*/
static inline TableScanDesc
table_beginscan_bm(Relation rel, Snapshot snapshot,
- int nkeys, ScanKeyData *key)
+ int nkeys, ScanKeyData *key, uint32 flags)
{
- uint32 flags = SO_TYPE_BITMAPSCAN | SO_ALLOW_PAGEMODE;
+ flags |= SO_TYPE_BITMAPSCAN | SO_ALLOW_PAGEMODE;
return table_beginscan_common(rel, snapshot, nkeys, key, NULL, flags);
}
@@ -957,9 +960,9 @@ static inline TableScanDesc
table_beginscan_sampling(Relation rel, Snapshot snapshot,
int nkeys, ScanKeyData *key,
bool allow_strat, bool allow_sync,
- bool allow_pagemode)
+ bool allow_pagemode, uint32 flags)
{
- uint32 flags = SO_TYPE_SAMPLESCAN;
+ flags |= SO_TYPE_SAMPLESCAN;
if (allow_strat)
flags |= SO_ALLOW_STRAT;
@@ -1059,10 +1062,11 @@ table_scan_getnextslot(TableScanDesc sscan, ScanDirection direction, TupleTableS
static inline TableScanDesc
table_beginscan_tidrange(Relation rel, Snapshot snapshot,
ItemPointer mintid,
- ItemPointer maxtid)
+ ItemPointer maxtid, uint32 flags)
{
TableScanDesc sscan;
- uint32 flags = SO_TYPE_TIDRANGESCAN | SO_ALLOW_PAGEMODE;
+
+ flags |= SO_TYPE_TIDRANGESCAN | SO_ALLOW_PAGEMODE;
sscan = table_beginscan_common(rel, snapshot, 0, NULL, NULL, flags);
@@ -1139,7 +1143,8 @@ extern void table_parallelscan_initialize(Relation rel,
* Caller must hold a suitable lock on the relation.
*/
extern TableScanDesc table_beginscan_parallel(Relation relation,
- ParallelTableScanDesc pscan);
+ ParallelTableScanDesc pscan,
+ uint32 flags);
/*
* Begin a parallel tid range scan. `pscan` needs to have been initialized
@@ -1149,7 +1154,8 @@ extern TableScanDesc table_beginscan_parallel(Relation relation,
* Caller must hold a suitable lock on the relation.
*/
extern TableScanDesc table_beginscan_parallel_tidrange(Relation relation,
- ParallelTableScanDesc pscan);
+ ParallelTableScanDesc pscan,
+ uint32 flags);
/*
* Restart a parallel scan. Call this in the leader process. Caller is
@@ -1172,10 +1178,13 @@ table_parallelscan_reinitialize(Relation rel, ParallelTableScanDesc pscan)
* Prepare to fetch tuples from the relation, as needed when fetching tuples
* for an index scan.
*
+ * 'flags' is a bitmask of SO_* flags providing hints from the executor about
+ * the scan context.
+ *
* Tuples for an index scan can then be fetched via table_index_fetch_tuple().
*/
static inline IndexFetchTableData *
-table_index_fetch_begin(Relation rel)
+table_index_fetch_begin(Relation rel, uint32 flags)
{
/*
* We don't allow scans to be started while CheckXidAlive is set, except
@@ -1185,7 +1194,7 @@ table_index_fetch_begin(Relation rel)
if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
elog(ERROR, "scan started during logical decoding");
- return rel->rd_tableam->index_fetch_begin(rel);
+ return rel->rd_tableam->index_fetch_begin(rel, flags);
}
/*
--
2.43.0
[text/x-patch] v38-0010-Pass-down-information-on-table-modification-to-s.patch (14.5K, 11-v38-0010-Pass-down-information-on-table-modification-to-s.patch)
download | inline diff:
From 3a6b08fc3219afd79dc81a5219e6a543d67036f6 Mon Sep 17 00:00:00 2001
From: Melanie Plageman <[email protected]>
Date: Mon, 2 Mar 2026 16:31:33 -0500
Subject: [PATCH v38 10/12] Pass down information on table modification to scan
node
Pass down information to sequential scan, index [only] scan, bitmap
table scan, sample scan, and TID range scan nodes on whether or not the
query modifies the relation being scanned. A later commit will use this
information to update the VM during on-access pruning only if the
relation is not modified by the query.
Author: Melanie Plageman <[email protected]>
Reviewed-by: Andres Freund <[email protected]>
Reviewed-by: Andrey Borodin <[email protected]>
Reviewed-by: Chao Li <[email protected]>
Discussion: https://postgr.es/m/4379FDA3-9446-4E2C-9C15-32EFE8D4F31B%40yandex-team.ru
---
src/backend/access/heap/heapam_handler.c | 1 +
src/backend/executor/nodeBitmapHeapscan.c | 9 ++++++-
src/backend/executor/nodeIndexonlyscan.c | 25 +++++++++++++++---
src/backend/executor/nodeIndexscan.c | 32 ++++++++++++++++++++---
src/backend/executor/nodeSamplescan.c | 8 +++++-
src/backend/executor/nodeSeqscan.c | 26 +++++++++++++++---
src/backend/executor/nodeTidrangescan.c | 24 ++++++++++++++---
src/include/access/heapam.h | 6 +++++
src/include/access/tableam.h | 3 +++
9 files changed, 119 insertions(+), 15 deletions(-)
diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c
index 6122603d11e..d35b688d751 100644
--- a/src/backend/access/heap/heapam_handler.c
+++ b/src/backend/access/heap/heapam_handler.c
@@ -86,6 +86,7 @@ heapam_index_fetch_begin(Relation rel, uint32 flags)
hscan->xs_base.rel = rel;
hscan->xs_cbuf = InvalidBuffer;
hscan->xs_vmbuffer = InvalidBuffer;
+ hscan->modifies_base_rel = !(flags & SO_HINT_REL_READ_ONLY);
return &hscan->xs_base;
}
diff --git a/src/backend/executor/nodeBitmapHeapscan.c b/src/backend/executor/nodeBitmapHeapscan.c
index 620fc7e259a..a5ab5e2b37f 100644
--- a/src/backend/executor/nodeBitmapHeapscan.c
+++ b/src/backend/executor/nodeBitmapHeapscan.c
@@ -104,11 +104,18 @@ BitmapTableScanSetup(BitmapHeapScanState *node)
*/
if (!node->ss.ss_currentScanDesc)
{
+ uint32 flags = 0;
+
+ if (!bms_is_member(((Scan *) node->ss.ps.plan)->scanrelid,
+ node->ss.ps.state->es_modified_relids))
+ flags |= SO_HINT_REL_READ_ONLY;
+
node->ss.ss_currentScanDesc =
table_beginscan_bm(node->ss.ss_currentRelation,
node->ss.ps.state->es_snapshot,
0,
- NULL, 0);
+ NULL,
+ flags);
}
node->ss.ss_currentScanDesc->st.rs_tbmiterator = tbmiterator;
diff --git a/src/backend/executor/nodeIndexonlyscan.c b/src/backend/executor/nodeIndexonlyscan.c
index aefb792ee6e..6d7a32c1cb8 100644
--- a/src/backend/executor/nodeIndexonlyscan.c
+++ b/src/backend/executor/nodeIndexonlyscan.c
@@ -84,6 +84,12 @@ IndexOnlyNext(IndexOnlyScanState *node)
if (scandesc == NULL)
{
+ uint32 flags = 0;
+
+ if (!bms_is_member(((Scan *) node->ss.ps.plan)->scanrelid,
+ estate->es_modified_relids))
+ flags |= SO_HINT_REL_READ_ONLY;
+
/*
* We reach here if the index only scan is not parallel, or if we're
* serially executing an index only scan that was planned to be
@@ -94,7 +100,8 @@ IndexOnlyNext(IndexOnlyScanState *node)
estate->es_snapshot,
&node->ioss_Instrument,
node->ioss_NumScanKeys,
- node->ioss_NumOrderByKeys, 0);
+ node->ioss_NumOrderByKeys,
+ flags);
node->ioss_ScanDesc = scandesc;
@@ -761,6 +768,7 @@ ExecIndexOnlyScanInitializeDSM(IndexOnlyScanState *node,
ParallelIndexScanDesc piscan;
bool instrument = node->ss.ps.instrument != NULL;
bool parallel_aware = node->ss.ps.plan->parallel_aware;
+ uint32 flags = 0;
if (!instrument && !parallel_aware)
{
@@ -782,13 +790,18 @@ ExecIndexOnlyScanInitializeDSM(IndexOnlyScanState *node,
return;
}
+ if (!bms_is_member(((Scan *) node->ss.ps.plan)->scanrelid,
+ estate->es_modified_relids))
+ flags |= SO_HINT_REL_READ_ONLY;
+
node->ioss_ScanDesc =
index_beginscan_parallel(node->ss.ss_currentRelation,
node->ioss_RelationDesc,
&node->ioss_Instrument,
node->ioss_NumScanKeys,
node->ioss_NumOrderByKeys,
- piscan, 0);
+ piscan, flags);
+
node->ioss_ScanDesc->xs_want_itup = true;
node->ioss_VMBuffer = InvalidBuffer;
@@ -829,6 +842,7 @@ ExecIndexOnlyScanInitializeWorker(IndexOnlyScanState *node,
ParallelIndexScanDesc piscan;
bool instrument = node->ss.ps.instrument != NULL;
bool parallel_aware = node->ss.ps.plan->parallel_aware;
+ uint32 flags = 0;
if (!instrument && !parallel_aware)
{
@@ -848,13 +862,18 @@ ExecIndexOnlyScanInitializeWorker(IndexOnlyScanState *node,
return;
}
+ if (!bms_is_member(((Scan *) node->ss.ps.plan)->scanrelid,
+ node->ss.ps.state->es_modified_relids))
+ flags |= SO_HINT_REL_READ_ONLY;
+
node->ioss_ScanDesc =
index_beginscan_parallel(node->ss.ss_currentRelation,
node->ioss_RelationDesc,
&node->ioss_Instrument,
node->ioss_NumScanKeys,
node->ioss_NumOrderByKeys,
- piscan, 0);
+ piscan, flags);
+
node->ioss_ScanDesc->xs_want_itup = true;
/*
diff --git a/src/backend/executor/nodeIndexscan.c b/src/backend/executor/nodeIndexscan.c
index 477cd4fcf99..52b7fc46593 100644
--- a/src/backend/executor/nodeIndexscan.c
+++ b/src/backend/executor/nodeIndexscan.c
@@ -102,6 +102,12 @@ IndexNext(IndexScanState *node)
if (scandesc == NULL)
{
+ uint32 flags = 0;
+
+ if (!bms_is_member(((Scan *) node->ss.ps.plan)->scanrelid,
+ estate->es_modified_relids))
+ flags |= SO_HINT_REL_READ_ONLY;
+
/*
* We reach here if the index scan is not parallel, or if we're
* serially executing an index scan that was planned to be parallel.
@@ -111,7 +117,8 @@ IndexNext(IndexScanState *node)
estate->es_snapshot,
&node->iss_Instrument,
node->iss_NumScanKeys,
- node->iss_NumOrderByKeys, 0);
+ node->iss_NumOrderByKeys,
+ flags);
node->iss_ScanDesc = scandesc;
@@ -198,6 +205,12 @@ IndexNextWithReorder(IndexScanState *node)
if (scandesc == NULL)
{
+ uint32 flags = 0;
+
+ if (!bms_is_member(((Scan *) node->ss.ps.plan)->scanrelid,
+ estate->es_modified_relids))
+ flags |= SO_HINT_REL_READ_ONLY;
+
/*
* We reach here if the index scan is not parallel, or if we're
* serially executing an index scan that was planned to be parallel.
@@ -207,7 +220,8 @@ IndexNextWithReorder(IndexScanState *node)
estate->es_snapshot,
&node->iss_Instrument,
node->iss_NumScanKeys,
- node->iss_NumOrderByKeys, 0);
+ node->iss_NumOrderByKeys,
+ flags);
node->iss_ScanDesc = scandesc;
@@ -1696,6 +1710,7 @@ ExecIndexScanInitializeDSM(IndexScanState *node,
ParallelIndexScanDesc piscan;
bool instrument = node->ss.ps.instrument != NULL;
bool parallel_aware = node->ss.ps.plan->parallel_aware;
+ uint32 flags = 0;
if (!instrument && !parallel_aware)
{
@@ -1717,13 +1732,17 @@ ExecIndexScanInitializeDSM(IndexScanState *node,
return;
}
+ if (!bms_is_member(((Scan *) node->ss.ps.plan)->scanrelid,
+ estate->es_modified_relids))
+ flags |= SO_HINT_REL_READ_ONLY;
+
node->iss_ScanDesc =
index_beginscan_parallel(node->ss.ss_currentRelation,
node->iss_RelationDesc,
&node->iss_Instrument,
node->iss_NumScanKeys,
node->iss_NumOrderByKeys,
- piscan, 0);
+ piscan, flags);
/*
* If no run-time keys to calculate or they are ready, go ahead and pass
@@ -1762,6 +1781,7 @@ ExecIndexScanInitializeWorker(IndexScanState *node,
ParallelIndexScanDesc piscan;
bool instrument = node->ss.ps.instrument != NULL;
bool parallel_aware = node->ss.ps.plan->parallel_aware;
+ uint32 flags = 0;
if (!instrument && !parallel_aware)
{
@@ -1781,13 +1801,17 @@ ExecIndexScanInitializeWorker(IndexScanState *node,
return;
}
+ if (!bms_is_member(((Scan *) node->ss.ps.plan)->scanrelid,
+ node->ss.ps.state->es_modified_relids))
+ flags |= SO_HINT_REL_READ_ONLY;
+
node->iss_ScanDesc =
index_beginscan_parallel(node->ss.ss_currentRelation,
node->iss_RelationDesc,
&node->iss_Instrument,
node->iss_NumScanKeys,
node->iss_NumOrderByKeys,
- piscan, 0);
+ piscan, flags);
/*
* If no run-time keys to calculate or they are ready, go ahead and pass
diff --git a/src/backend/executor/nodeSamplescan.c b/src/backend/executor/nodeSamplescan.c
index 47660baf2fa..62eff19bc4f 100644
--- a/src/backend/executor/nodeSamplescan.c
+++ b/src/backend/executor/nodeSamplescan.c
@@ -291,13 +291,19 @@ tablesample_init(SampleScanState *scanstate)
/* Now we can create or reset the HeapScanDesc */
if (scanstate->ss.ss_currentScanDesc == NULL)
{
+ uint32 flags = 0;
+
+ if (!bms_is_member(((Scan *) scanstate->ss.ps.plan)->scanrelid,
+ scanstate->ss.ps.state->es_modified_relids))
+ flags |= SO_HINT_REL_READ_ONLY;
+
scanstate->ss.ss_currentScanDesc =
table_beginscan_sampling(scanstate->ss.ss_currentRelation,
scanstate->ss.ps.state->es_snapshot,
0, NULL,
scanstate->use_bulkread,
allow_sync,
- scanstate->use_pagemode, 0);
+ scanstate->use_pagemode, flags);
}
else
{
diff --git a/src/backend/executor/nodeSeqscan.c b/src/backend/executor/nodeSeqscan.c
index d9d7ec0516a..65349ea9c54 100644
--- a/src/backend/executor/nodeSeqscan.c
+++ b/src/backend/executor/nodeSeqscan.c
@@ -65,13 +65,20 @@ SeqNext(SeqScanState *node)
if (scandesc == NULL)
{
+ uint32 flags = 0;
+
+ if (!bms_is_member(((Scan *) node->ss.ps.plan)->scanrelid,
+ estate->es_modified_relids))
+ flags |= SO_HINT_REL_READ_ONLY;
+
/*
* We reach here if the scan is not parallel, or if we're serially
* executing a scan that was planned to be parallel.
*/
scandesc = table_beginscan(node->ss.ss_currentRelation,
estate->es_snapshot,
- 0, NULL, 0);
+ 0, NULL, flags);
+
node->ss.ss_currentScanDesc = scandesc;
}
@@ -367,14 +374,20 @@ ExecSeqScanInitializeDSM(SeqScanState *node,
{
EState *estate = node->ss.ps.state;
ParallelTableScanDesc pscan;
+ uint32 flags = 0;
pscan = shm_toc_allocate(pcxt->toc, node->pscan_len);
table_parallelscan_initialize(node->ss.ss_currentRelation,
pscan,
estate->es_snapshot);
shm_toc_insert(pcxt->toc, node->ss.ps.plan->plan_node_id, pscan);
+ if (!bms_is_member(((Scan *) node->ss.ps.plan)->scanrelid,
+ estate->es_modified_relids))
+ flags |= SO_HINT_REL_READ_ONLY;
+
node->ss.ss_currentScanDesc =
- table_beginscan_parallel(node->ss.ss_currentRelation, pscan, 0);
+ table_beginscan_parallel(node->ss.ss_currentRelation, pscan,
+ flags);
}
/* ----------------------------------------------------------------
@@ -404,8 +417,15 @@ ExecSeqScanInitializeWorker(SeqScanState *node,
ParallelWorkerContext *pwcxt)
{
ParallelTableScanDesc pscan;
+ uint32 flags = 0;
+
+ if (!bms_is_member(((Scan *) node->ss.ps.plan)->scanrelid,
+ node->ss.ps.state->es_modified_relids))
+ flags |= SO_HINT_REL_READ_ONLY;
pscan = shm_toc_lookup(pwcxt->toc, node->ss.ps.plan->plan_node_id, false);
node->ss.ss_currentScanDesc =
- table_beginscan_parallel(node->ss.ss_currentRelation, pscan, 0);
+ table_beginscan_parallel(node->ss.ss_currentRelation,
+ pscan,
+ flags);
}
diff --git a/src/backend/executor/nodeTidrangescan.c b/src/backend/executor/nodeTidrangescan.c
index 461edb8893b..7fbdf401734 100644
--- a/src/backend/executor/nodeTidrangescan.c
+++ b/src/backend/executor/nodeTidrangescan.c
@@ -242,10 +242,16 @@ TidRangeNext(TidRangeScanState *node)
if (scandesc == NULL)
{
+ uint32 flags = 0;
+
+ if (!bms_is_member(((Scan *) node->ss.ps.plan)->scanrelid,
+ estate->es_modified_relids))
+ flags |= SO_HINT_REL_READ_ONLY;
+
scandesc = table_beginscan_tidrange(node->ss.ss_currentRelation,
estate->es_snapshot,
&node->trss_mintid,
- &node->trss_maxtid, 0);
+ &node->trss_maxtid, flags);
node->ss.ss_currentScanDesc = scandesc;
}
else
@@ -451,15 +457,21 @@ ExecTidRangeScanInitializeDSM(TidRangeScanState *node, ParallelContext *pcxt)
{
EState *estate = node->ss.ps.state;
ParallelTableScanDesc pscan;
+ uint32 flags = 0;
pscan = shm_toc_allocate(pcxt->toc, node->trss_pscanlen);
table_parallelscan_initialize(node->ss.ss_currentRelation,
pscan,
estate->es_snapshot);
shm_toc_insert(pcxt->toc, node->ss.ps.plan->plan_node_id, pscan);
+
+ if (!bms_is_member(((Scan *) node->ss.ps.plan)->scanrelid,
+ estate->es_modified_relids))
+ flags |= SO_HINT_REL_READ_ONLY;
+
node->ss.ss_currentScanDesc =
table_beginscan_parallel_tidrange(node->ss.ss_currentRelation,
- pscan, 0);
+ pscan, flags);
}
/* ----------------------------------------------------------------
@@ -489,9 +501,15 @@ ExecTidRangeScanInitializeWorker(TidRangeScanState *node,
ParallelWorkerContext *pwcxt)
{
ParallelTableScanDesc pscan;
+ uint32 flags = 0;
pscan = shm_toc_lookup(pwcxt->toc, node->ss.ps.plan->plan_node_id, false);
+
+ if (!bms_is_member(((Scan *) node->ss.ps.plan)->scanrelid,
+ node->ss.ps.state->es_modified_relids))
+ flags |= SO_HINT_REL_READ_ONLY;
+
node->ss.ss_currentScanDesc =
table_beginscan_parallel_tidrange(node->ss.ss_currentRelation,
- pscan, 0);
+ pscan, flags);
}
diff --git a/src/include/access/heapam.h b/src/include/access/heapam.h
index c2621dc2fac..978ea90ffa2 100644
--- a/src/include/access/heapam.h
+++ b/src/include/access/heapam.h
@@ -131,6 +131,12 @@ typedef struct IndexFetchHeapData
/* Current heap block's corresponding page in the visibility map */
Buffer xs_vmbuffer;
+
+ /*
+ * Some optimizations can only be performed if the query does not modify
+ * the underlying relation. Track that here.
+ */
+ bool modifies_base_rel;
} IndexFetchHeapData;
/* Result codes for HeapTupleSatisfiesVacuum */
diff --git a/src/include/access/tableam.h b/src/include/access/tableam.h
index f1065e30638..57ce94a386f 100644
--- a/src/include/access/tableam.h
+++ b/src/include/access/tableam.h
@@ -63,6 +63,9 @@ typedef enum ScanOptions
/* unregister snapshot at scan end? */
SO_TEMP_SNAPSHOT = 1 << 9,
+
+ /* set if the query doesn't modify the relation */
+ SO_HINT_REL_READ_ONLY = 1 << 10,
} ScanOptions;
/*
--
2.43.0
[text/x-patch] v38-0011-Allow-on-access-pruning-to-set-pages-all-visible.patch (10.5K, 12-v38-0011-Allow-on-access-pruning-to-set-pages-all-visible.patch)
download | inline diff:
From aa05e68336207dbb64c0468ab3f017f8f66f9e05 Mon Sep 17 00:00:00 2001
From: Melanie Plageman <[email protected]>
Date: Fri, 27 Feb 2026 16:33:40 -0500
Subject: [PATCH v38 11/12] Allow on-access pruning to set pages all-visible
Many queries do not modify the underlying relation. For such queries, if
on-access pruning occurs during the scan, we can check whether the page
has become all-visible and update the visibility map accordingly.
Previously, only vacuum and COPY FREEZE marked pages as all-visible or
all-frozen.
This commit implements on-access VM setting for sequential scans as well
as for the underlying heap relation in index scans and bitmap heap
scans.
Author: Melanie Plageman <[email protected]>
Reviewed-by: Andres Freund <[email protected]>
Reviewed-by: Kirill Reshke <[email protected]>
Reviewed-by: Chao Li <[email protected]>
Discussion: https://postgr.es/m/flat/CAAKRu_ZMw6Npd_qm2KM%2BFwQ3cMOMx1Dh3VMhp8-V7SOLxdK9-g%40mail.gmail.com
---
src/backend/access/heap/heapam.c | 3 +-
src/backend/access/heap/heapam_handler.c | 6 ++-
src/backend/access/heap/pruneheap.c | 46 ++++++++++++++-----
src/backend/access/heap/vacuumlazy.c | 2 +-
src/include/access/heapam.h | 12 +++--
.../t/035_standby_logical_decoding.pl | 3 +-
6 files changed, 52 insertions(+), 20 deletions(-)
diff --git a/src/backend/access/heap/heapam.c b/src/backend/access/heap/heapam.c
index 044f385e477..dbdf6521c42 100644
--- a/src/backend/access/heap/heapam.c
+++ b/src/backend/access/heap/heapam.c
@@ -633,7 +633,8 @@ heap_prepare_pagescan(TableScanDesc sscan)
/*
* Prune and repair fragmentation for the whole page, if possible.
*/
- heap_page_prune_opt(scan->rs_base.rs_rd, buffer, &scan->rs_vmbuffer);
+ heap_page_prune_opt(scan->rs_base.rs_rd, buffer, &scan->rs_vmbuffer,
+ (sscan->rs_flags & SO_HINT_REL_READ_ONLY));
/*
* We must hold share lock on the buffer content while examining tuple
diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c
index d35b688d751..a083b69ffcd 100644
--- a/src/backend/access/heap/heapam_handler.c
+++ b/src/backend/access/heap/heapam_handler.c
@@ -147,7 +147,8 @@ heapam_index_fetch_tuple(struct IndexFetchTableData *scan,
*/
if (prev_buf != hscan->xs_cbuf)
heap_page_prune_opt(hscan->xs_base.rel, hscan->xs_cbuf,
- &hscan->xs_vmbuffer);
+ &hscan->xs_vmbuffer,
+ !hscan->modifies_base_rel);
}
/* Obtain share-lock on the buffer so we can examine visibility */
@@ -2542,7 +2543,8 @@ BitmapHeapScanNextBlock(TableScanDesc scan,
/*
* Prune and repair fragmentation for the whole page, if possible.
*/
- heap_page_prune_opt(scan->rs_rd, buffer, &hscan->rs_vmbuffer);
+ heap_page_prune_opt(scan->rs_rd, buffer, &hscan->rs_vmbuffer,
+ scan->rs_flags & SO_HINT_REL_READ_ONLY);
/*
* We must hold share lock on the buffer content while examining tuple
diff --git a/src/backend/access/heap/pruneheap.c b/src/backend/access/heap/pruneheap.c
index 9b5a0726f2b..3cdc1a36441 100644
--- a/src/backend/access/heap/pruneheap.c
+++ b/src/backend/access/heap/pruneheap.c
@@ -44,6 +44,8 @@ typedef struct
bool mark_unused_now;
/* whether to attempt freezing tuples */
bool attempt_freeze;
+ /* whether to attempt setting the VM */
+ bool attempt_set_vm;
struct VacuumCutoffs *cutoffs;
Relation relation;
@@ -213,7 +215,8 @@ static void page_verify_redirects(Page page);
static bool heap_page_will_freeze(bool did_tuple_hint_fpi, bool do_prune, bool do_hint_prune,
PruneState *prstate);
-static bool heap_page_will_set_vm(PruneState *prstate, PruneReason reason);
+static bool heap_page_will_set_vm(PruneState *prstate, PruneReason reason,
+ bool do_prune, bool do_freeze);
/*
* Optionally prune and repair fragmentation in the specified page.
@@ -233,7 +236,8 @@ static bool heap_page_will_set_vm(PruneState *prstate, PruneReason reason);
* unpinning *vmbuffer.
*/
void
-heap_page_prune_opt(Relation relation, Buffer buffer, Buffer *vmbuffer)
+heap_page_prune_opt(Relation relation, Buffer buffer, Buffer *vmbuffer,
+ bool rel_read_only)
{
Page page = BufferGetPage(buffer);
TransactionId prune_xid;
@@ -315,6 +319,8 @@ heap_page_prune_opt(Relation relation, Buffer buffer, Buffer *vmbuffer)
* current implementation.
*/
params.options = 0;
+ if (rel_read_only)
+ params.options = HEAP_PAGE_PRUNE_SET_VM;
heap_page_prune_and_freeze(¶ms, &presult, &dummy_off_loc,
NULL, NULL);
@@ -371,6 +377,7 @@ prune_freeze_setup(PruneFreezeParams *params,
/* cutoffs must be provided if we will attempt freezing */
Assert(!(params->options & HEAP_PAGE_PRUNE_FREEZE) || params->cutoffs);
prstate->attempt_freeze = (params->options & HEAP_PAGE_PRUNE_FREEZE) != 0;
+ prstate->attempt_set_vm = (params->options & HEAP_PAGE_PRUNE_SET_VM) != 0;
prstate->cutoffs = params->cutoffs;
prstate->relation = params->relation;
prstate->block = BufferGetBlockNumber(params->buffer);
@@ -440,9 +447,8 @@ prune_freeze_setup(PruneFreezeParams *params,
* We track whether the page will be all-visible/all-frozen at the end of
* pruning and freezing. While examining tuple visibility, we'll set
* set_all_visible to false if there are tuples on the page not visible to
- * all running and future transactions. set_all_visible is always
- * maintained but only VACUUM will set the VM if the page ends up being
- * all-visible.
+ * all running and future transactions. If enabled for this scan, we will
+ * set the VM if the page ends up being all-visible.
*
* We also keep track of the newest live XID, which is used to calculate
* the snapshot conflict horizon for a WAL record setting the VM.
@@ -873,21 +879,37 @@ heap_fix_vm_corruption(PruneState *prstate, OffsetNumber offnum)
* This function does not actually set the VM bits or page-level visibility
* hint, PD_ALL_VISIBLE.
*
+ * This should be called only after do_freeze has been decided (and do_prune
+ * has been set), as these factor into our heuristic-based decision.
+ *
* Returns true if one or both VM bits should be set and false otherwise.
*/
static bool
-heap_page_will_set_vm(PruneState *prstate, PruneReason reason)
+heap_page_will_set_vm(PruneState *prstate, PruneReason reason,
+ bool do_prune, bool do_freeze)
{
- /*
- * Though on-access pruning maintains prstate->set_all_visible, we don't
- * consider setting the VM.
- */
- if (reason == PRUNE_ON_ACCESS)
+ if (!prstate->attempt_set_vm)
return false;
if (!prstate->set_all_visible)
return false;
+ /*
+ * If this is an on-access call and we're not actually pruning, avoid
+ * setting the visibility map if it would newly dirty the heap page or, if
+ * the page is already dirty, if doing so would require including a
+ * full-page image (FPI) of the heap page in the WAL. This situation
+ * should be rare, as on-access pruning is only attempted when
+ * pd_prune_xid is valid.
+ */
+ if (reason == PRUNE_ON_ACCESS && !do_prune && !do_freeze &&
+ (!BufferIsDirty(prstate->buffer) || XLogCheckBufferNeedsBackup(prstate->buffer)))
+ {
+ prstate->set_all_visible = false;
+ prstate->set_all_frozen = false;
+ return false;
+ }
+
prstate->new_vmbits = VISIBILITYMAP_ALL_VISIBLE;
if (prstate->set_all_frozen)
@@ -1103,7 +1125,7 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
Assert(!prstate.set_all_frozen || prstate.set_all_visible);
Assert(!prstate.set_all_visible || (prstate.lpdead_items == 0));
- do_set_vm = heap_page_will_set_vm(&prstate, params->reason);
+ do_set_vm = heap_page_will_set_vm(&prstate, params->reason, do_prune, do_freeze);
/*
* new_vmbits should be 0 regardless of whether or not the page is
diff --git a/src/backend/access/heap/vacuumlazy.c b/src/backend/access/heap/vacuumlazy.c
index 8bd178ae7e6..d2cae77b52a 100644
--- a/src/backend/access/heap/vacuumlazy.c
+++ b/src/backend/access/heap/vacuumlazy.c
@@ -2008,7 +2008,7 @@ lazy_scan_prune(LVRelState *vacrel,
.buffer = buf,
.vmbuffer = vmbuffer,
.reason = PRUNE_VACUUM_SCAN,
- .options = HEAP_PAGE_PRUNE_FREEZE,
+ .options = HEAP_PAGE_PRUNE_FREEZE | HEAP_PAGE_PRUNE_SET_VM,
.vistest = vacrel->vistest,
.cutoffs = &vacrel->cutoffs,
};
diff --git a/src/include/access/heapam.h b/src/include/access/heapam.h
index 978ea90ffa2..768d442c39c 100644
--- a/src/include/access/heapam.h
+++ b/src/include/access/heapam.h
@@ -42,6 +42,7 @@
/* "options" flag bits for heap_page_prune_and_freeze */
#define HEAP_PAGE_PRUNE_MARK_UNUSED_NOW (1 << 0)
#define HEAP_PAGE_PRUNE_FREEZE (1 << 1)
+#define HEAP_PAGE_PRUNE_SET_VM (1 << 2)
typedef struct BulkInsertStateData *BulkInsertState;
typedef struct GlobalVisState GlobalVisState;
@@ -97,7 +98,8 @@ typedef struct HeapScanDescData
/*
* For sequential scans, bitmap heap scans, TID range scans, and sample
* scans. The current heap block's corresponding page in the visibility
- * map.
+ * map. If the relation is not modified by the query, on-access pruning
+ * may set the VM.
*/
Buffer rs_vmbuffer;
@@ -129,7 +131,11 @@ typedef struct IndexFetchHeapData
*/
Buffer xs_cbuf;
- /* Current heap block's corresponding page in the visibility map */
+ /*
+ * Current heap block's corresponding page in the visibility map. For
+ * index scans that do not modify the underlying heap table, on-access
+ * pruning may set the VM on-access.
+ */
Buffer xs_vmbuffer;
/*
@@ -440,7 +446,7 @@ extern TransactionId heap_index_delete_tuples(Relation rel,
/* in heap/pruneheap.c */
extern void heap_page_prune_opt(Relation relation, Buffer buffer,
- Buffer *vmbuffer);
+ Buffer *vmbuffer, bool rel_read_only);
extern void heap_page_prune_and_freeze(PruneFreezeParams *params,
PruneFreezeResult *presult,
OffsetNumber *off_loc,
diff --git a/src/test/recovery/t/035_standby_logical_decoding.pl b/src/test/recovery/t/035_standby_logical_decoding.pl
index d264a698ff6..a5536ba4ff6 100644
--- a/src/test/recovery/t/035_standby_logical_decoding.pl
+++ b/src/test/recovery/t/035_standby_logical_decoding.pl
@@ -296,6 +296,7 @@ wal_level = 'logical'
max_replication_slots = 4
max_wal_senders = 4
autovacuum = off
+hot_standby_feedback = on
});
$node_primary->dump_info;
$node_primary->start;
@@ -748,7 +749,7 @@ check_pg_recvlogical_stderr($handle,
$logstart = -s $node_standby->logfile;
reactive_slots_change_hfs_and_wait_for_xmins('shared_row_removal_',
- 'no_conflict_', 0, 1);
+ 'no_conflict_', 1, 0);
# This should not trigger a conflict
wait_until_vacuum_can_remove(
--
2.43.0
[text/x-patch] v38-0012-Set-pd_prune_xid-on-insert.patch (10.9K, 13-v38-0012-Set-pd_prune_xid-on-insert.patch)
download | inline diff:
From fc597950684dad6328114ac0d10f791bc52b53c4 Mon Sep 17 00:00:00 2001
From: Melanie Plageman <[email protected]>
Date: Tue, 29 Jul 2025 16:12:56 -0400
Subject: [PATCH v38 12/12] Set pd_prune_xid on insert
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Now that visibility map (VM) updates can occur during read-only queries,
it makes sense to also set the page’s pd_prune_xid hint during inserts
and on the new page during updates.
This enables heap_page_prune_and_freeze() to run and set the VM
all-visible after a page is filled with newly inserted tuples the first
time it is read.
This change also addresses a long-standing note in heap_insert() and
heap_multi_insert(), which observed that setting pd_prune_xid would
help clean up aborted insertions sooner. Without it, such tuples might
linger until VACUUM, whereas now they can be pruned earlier.
The index killtuples test had to be updated to reflect a larger number
of hits by some accesses. Since the prune_xid is set by the fill/insert
step, on-access pruning can happen during the first access step (before
the DELETE). This is when the VM is extended. After the DELETE, the next
access hits the VM block instead of extending it. Thus, an additional
buffer hit is counted for the table.
Author: Melanie Plageman <[email protected]>
Reviewed-by: Chao Li <[email protected]>
Discussion: https://postgr.es/m/flat/CAAKRu_ZMw6Npd_qm2KM%2BFwQ3cMOMx1Dh3VMhp8-V7SOLxdK9-g%40mail.gmail.com
---
src/backend/access/heap/heapam.c | 40 ++++++++++++-------
src/backend/access/heap/heapam_xlog.c | 19 ++++++++-
src/backend/access/heap/pruneheap.c | 17 ++++----
.../modules/index/expected/killtuples.out | 8 ++--
4 files changed, 56 insertions(+), 28 deletions(-)
diff --git a/src/backend/access/heap/heapam.c b/src/backend/access/heap/heapam.c
index dbdf6521c42..c199646b25d 100644
--- a/src/backend/access/heap/heapam.c
+++ b/src/backend/access/heap/heapam.c
@@ -2156,6 +2156,7 @@ heap_insert(Relation relation, HeapTuple tup, CommandId cid,
TransactionId xid = GetCurrentTransactionId();
HeapTuple heaptup;
Buffer buffer;
+ Page page;
Buffer vmbuffer = InvalidBuffer;
bool all_visible_cleared = false;
@@ -2182,6 +2183,8 @@ heap_insert(Relation relation, HeapTuple tup, CommandId cid,
&vmbuffer, NULL,
0);
+ page = BufferGetPage(buffer);
+
/*
* We're about to do the actual insert -- but check for conflict first, to
* avoid possibly having to roll back work we've just done.
@@ -2205,25 +2208,30 @@ heap_insert(Relation relation, HeapTuple tup, CommandId cid,
RelationPutHeapTuple(relation, buffer, heaptup,
(options & HEAP_INSERT_SPECULATIVE) != 0);
- if (PageIsAllVisible(BufferGetPage(buffer)))
+ if (PageIsAllVisible(page))
{
all_visible_cleared = true;
- PageClearAllVisible(BufferGetPage(buffer));
+ PageClearAllVisible(page);
visibilitymap_clear(relation,
ItemPointerGetBlockNumber(&(heaptup->t_self)),
vmbuffer, VISIBILITYMAP_VALID_BITS);
}
/*
- * XXX Should we set PageSetPrunable on this page ?
+ * Set pd_prune_xid to trigger heap_page_prune_and_freeze() once the page
+ * is full so that we can set the page all-visible in the VM on the next
+ * page access.
*
- * The inserting transaction may eventually abort thus making this tuple
- * DEAD and hence available for pruning. Though we don't want to optimize
- * for aborts, if no other tuple in this page is UPDATEd/DELETEd, the
- * aborted tuple will never be pruned until next vacuum is triggered.
+ * Setting pd_prune_xid is also handy if the inserting transaction
+ * eventually aborts making this tuple DEAD and hence available for
+ * pruning. If no other tuple in this page is UPDATEd/DELETEd, the aborted
+ * tuple would never otherwise be pruned until next vacuum is triggered.
*
- * If you do add PageSetPrunable here, add it in heap_xlog_insert too.
+ * Don't set it if we are in bootstrap mode or we are inserting a frozen
+ * tuple.
*/
+ if (TransactionIdIsNormal(xid) && !(options & HEAP_INSERT_FROZEN))
+ PageSetPrunable(page, xid);
MarkBufferDirty(buffer);
@@ -2233,7 +2241,6 @@ heap_insert(Relation relation, HeapTuple tup, CommandId cid,
xl_heap_insert xlrec;
xl_heap_header xlhdr;
XLogRecPtr recptr;
- Page page = BufferGetPage(buffer);
uint8 info = XLOG_HEAP_INSERT;
int bufflags = 0;
@@ -2598,8 +2605,13 @@ heap_multi_insert(Relation relation, TupleTableSlot **slots, int ntuples,
}
/*
- * XXX Should we set PageSetPrunable on this page ? See heap_insert()
+ * Set pd_prune_xid. See heap_insert() for more on why we do this when
+ * inserting tuples. This only makes sense if we aren't already
+ * setting the page frozen in the VM. We also don't set it in
+ * bootstrap mode.
*/
+ if (!all_frozen_set && TransactionIdIsNormal(xid))
+ PageSetPrunable(page, xid);
MarkBufferDirty(buffer);
@@ -4141,12 +4153,12 @@ l2:
* the subsequent page pruning will be a no-op and the hint will be
* cleared.
*
- * XXX Should we set hint on newbuf as well? If the transaction aborts,
- * there would be a prunable tuple in the newbuf; but for now we choose
- * not to optimize for aborts. Note that heap_xlog_update must be kept in
- * sync if this decision changes.
+ * We set the new page prunable as well. See heap_insert() for more on why
+ * we do this when inserting tuples.
*/
PageSetPrunable(page, xid);
+ if (newbuf != buffer)
+ PageSetPrunable(newpage, xid);
if (use_hot_update)
{
diff --git a/src/backend/access/heap/heapam_xlog.c b/src/backend/access/heap/heapam_xlog.c
index 1302bb13e18..f3f419d3dc1 100644
--- a/src/backend/access/heap/heapam_xlog.c
+++ b/src/backend/access/heap/heapam_xlog.c
@@ -450,6 +450,14 @@ heap_xlog_insert(XLogReaderState *record)
freespace = PageGetHeapFreeSpace(page); /* needed to update FSM below */
+ /*
+ * Set the page prunable to trigger on-access pruning later, which may
+ * set the page all-visible in the VM. See comments in heap_insert().
+ */
+ if (TransactionIdIsNormal(XLogRecGetXid(record)) &&
+ !HeapTupleHeaderXminFrozen(htup))
+ PageSetPrunable(page, XLogRecGetXid(record));
+
PageSetLSN(page, lsn);
if (xlrec->flags & XLH_INSERT_ALL_VISIBLE_CLEARED)
@@ -599,12 +607,19 @@ heap_xlog_multi_insert(XLogReaderState *record)
if (xlrec->flags & XLH_INSERT_ALL_VISIBLE_CLEARED)
PageClearAllVisible(page);
- /* XLH_INSERT_ALL_FROZEN_SET implies that all tuples are visible */
+ /*
+ * XLH_INSERT_ALL_FROZEN_SET implies that all tuples are visible. If
+ * we are not setting the page frozen, then set the page's prunable
+ * hint so that we trigger on-access pruning later which may set the
+ * page all-visible in the VM.
+ */
if (xlrec->flags & XLH_INSERT_ALL_FROZEN_SET)
{
PageSetAllVisible(page);
PageClearPrunable(page);
}
+ else
+ PageSetPrunable(page, XLogRecGetXid(record));
MarkBufferDirty(buffer);
}
@@ -921,6 +936,8 @@ heap_xlog_update(XLogReaderState *record, bool hot_update)
freespace = PageGetHeapFreeSpace(npage);
PageSetLSN(npage, lsn);
+ /* See heap_insert() for why we set pd_prune_xid on insert */
+ PageSetPrunable(npage, XLogRecGetXid(record));
MarkBufferDirty(nbuffer);
}
diff --git a/src/backend/access/heap/pruneheap.c b/src/backend/access/heap/pruneheap.c
index 3cdc1a36441..7cb9e1e2aac 100644
--- a/src/backend/access/heap/pruneheap.c
+++ b/src/backend/access/heap/pruneheap.c
@@ -255,7 +255,8 @@ heap_page_prune_opt(Relation relation, Buffer buffer, Buffer *vmbuffer,
/*
* First check whether there's any chance there's something to prune,
* determining the appropriate horizon is a waste if there's no prune_xid
- * (i.e. no updates/deletes left potentially dead tuples around).
+ * (i.e. no updates/deletes left potentially dead tuples around and no
+ * inserts inserted new tuples that may be visible to all).
*/
prune_xid = PageGetPruneXid(page);
if (!TransactionIdIsValid(prune_xid))
@@ -1848,16 +1849,14 @@ heap_prune_record_unchanged_lp_normal(PruneState *prstate, OffsetNumber offnum)
prstate->set_all_visible = false;
prstate->set_all_frozen = false;
- /* The page should not be marked all-visible */
- if (PageIsAllVisible(page))
- heap_fix_vm_corruption(prstate, offnum);
-
/*
- * If we wanted to optimize for aborts, we might consider marking
- * the page prunable when we see INSERT_IN_PROGRESS. But we
- * don't. See related decisions about when to mark the page
- * prunable in heapam.c.
+ * Though there is nothing "prunable" on the page, we maintain
+ * pd_prune_xid for inserts so that we have the opportunity to
+ * mark them all-visible during the next round of pruning.
*/
+ heap_prune_record_prunable(prstate,
+ HeapTupleHeaderGetXmin(htup),
+ offnum);
break;
case HEAPTUPLE_DELETE_IN_PROGRESS:
diff --git a/src/test/modules/index/expected/killtuples.out b/src/test/modules/index/expected/killtuples.out
index be7ddd756ef..700144d6783 100644
--- a/src/test/modules/index/expected/killtuples.out
+++ b/src/test/modules/index/expected/killtuples.out
@@ -54,7 +54,7 @@ step flush: SELECT FROM pg_stat_force_next_flush();
step result: SELECT heap_blks_read + heap_blks_hit - counter.heap_accesses AS new_heap_accesses FROM counter, pg_statio_all_tables WHERE relname = 'kill_prior_tuple';
new_heap_accesses
-----------------
- 1
+ 2
(1 row)
step measure: UPDATE counter SET heap_accesses = (SELECT heap_blks_read + heap_blks_hit FROM pg_statio_all_tables WHERE relname = 'kill_prior_tuple');
@@ -130,7 +130,7 @@ step flush: SELECT FROM pg_stat_force_next_flush();
step result: SELECT heap_blks_read + heap_blks_hit - counter.heap_accesses AS new_heap_accesses FROM counter, pg_statio_all_tables WHERE relname = 'kill_prior_tuple';
new_heap_accesses
-----------------
- 1
+ 2
(1 row)
step measure: UPDATE counter SET heap_accesses = (SELECT heap_blks_read + heap_blks_hit FROM pg_statio_all_tables WHERE relname = 'kill_prior_tuple');
@@ -283,7 +283,7 @@ step flush: SELECT FROM pg_stat_force_next_flush();
step result: SELECT heap_blks_read + heap_blks_hit - counter.heap_accesses AS new_heap_accesses FROM counter, pg_statio_all_tables WHERE relname = 'kill_prior_tuple';
new_heap_accesses
-----------------
- 1
+ 2
(1 row)
step measure: UPDATE counter SET heap_accesses = (SELECT heap_blks_read + heap_blks_hit FROM pg_statio_all_tables WHERE relname = 'kill_prior_tuple');
@@ -329,7 +329,7 @@ step flush: SELECT FROM pg_stat_force_next_flush();
step result: SELECT heap_blks_read + heap_blks_hit - counter.heap_accesses AS new_heap_accesses FROM counter, pg_statio_all_tables WHERE relname = 'kill_prior_tuple';
new_heap_accesses
-----------------
- 1
+ 2
(1 row)
step measure: UPDATE counter SET heap_accesses = (SELECT heap_blks_read + heap_blks_hit FROM pg_statio_all_tables WHERE relname = 'kill_prior_tuple');
--
2.43.0
view thread (144+ messages) latest in thread
reply
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Reply to all the recipients using the --to and --cc options:
reply via email
To: [email protected]
Cc: [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected]
Subject: Re: eliminate xl_heap_visible to reduce WAL (and eventually set VM on-access)
In-Reply-To: <CAAKRu_ZbOp52rnkjf63h5mf94raEKBH7AAbz6QTx-xdH9yLfmQ@mail.gmail.com>
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
This inbox is served by agora; see mirroring instructions
for how to clone and mirror all data and code used for this inbox