agora inbox for [email protected]  
help / color / mirror / Atom feed
Race conditions with WAL sender PID lookups
160+ messages / 12 participants
[nested] [flat]

* Race conditions with WAL sender PID lookups
@ 2017-05-11 01:48  Michael Paquier <[email protected]>
  0 siblings, 3 replies; 160+ messages in thread

From: Michael Paquier @ 2017-05-11 01:48 UTC (permalink / raw)
  To: pgsql-hackers

Hi all,

I had my eyes on the WAL sender code this morning, and I have noticed
that walsender.c is not completely consistent with the PID lookups it
does in walsender.c. In two code paths, the PID value is checked
without holding the WAL sender spin lock (WalSndRqstFileReload and
pg_stat_get_wal_senders), which looks like a very bad idea contrary to
what the new WalSndWaitStopping() does and what InitWalSenderSlot() is
doing for ages.

Any thoughts about the patch attached to make things more consistent?
It seems to me that having some safeguards would be nice for
robustness.

That's an old bug, so I am adding that to the next CF.
Thanks,
-- 
Michael


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers


Attachments:

  [application/octet-stream] walsnd-pid-races.patch (1.3K, ../../CAB7nPqTWYqtzD=LN_oDaf9r-hAjUEPAy0B9yRkhcsLdRN8fzrw@mail.gmail.com/2-walsnd-pid-races.patch)
  download | inline diff:
diff --git a/src/backend/replication/walsender.c b/src/backend/replication/walsender.c
index 45d027803a..9bccaceb7e 100644
--- a/src/backend/replication/walsender.c
+++ b/src/backend/replication/walsender.c
@@ -2847,10 +2847,12 @@ WalSndRqstFileReload(void)
 	{
 		WalSnd	   *walsnd = &WalSndCtl->walsnds[i];
 
+		SpinLockAcquire(&walsnd->mutex);
 		if (walsnd->pid == 0)
+		{
+			SpinLockRelease(&walsnd->mutex);
 			continue;
-
-		SpinLockAcquire(&walsnd->mutex);
+		}
 		walsnd->needreload = true;
 		SpinLockRelease(&walsnd->mutex);
 	}
@@ -3163,14 +3165,18 @@ pg_stat_get_wal_senders(PG_FUNCTION_ARGS)
 		TimeOffset	flushLag;
 		TimeOffset	applyLag;
 		int			priority;
+		int			pid;
 		WalSndState state;
 		Datum		values[PG_STAT_GET_WAL_SENDERS_COLS];
 		bool		nulls[PG_STAT_GET_WAL_SENDERS_COLS];
 
+		SpinLockAcquire(&walsnd->mutex);
 		if (walsnd->pid == 0)
+		{
+			SpinLockRelease(&walsnd->mutex);
 			continue;
-
-		SpinLockAcquire(&walsnd->mutex);
+		}
+		pid = walsnd->pid;
 		sentPtr = walsnd->sentPtr;
 		state = walsnd->state;
 		write = walsnd->write;
@@ -3183,7 +3189,7 @@ pg_stat_get_wal_senders(PG_FUNCTION_ARGS)
 		SpinLockRelease(&walsnd->mutex);
 
 		memset(nulls, 0, sizeof(nulls));
-		values[0] = Int32GetDatum(walsnd->pid);
+		values[0] = Int32GetDatum(pid);
 
 		if (!superuser())
 		{


^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-05-12 02:44  Masahiko Sawada <[email protected]>
  parent: Michael Paquier <[email protected]>
  2 siblings, 1 reply; 160+ messages in thread

From: Masahiko Sawada @ 2017-05-12 02:44 UTC (permalink / raw)
  To: Michael Paquier <[email protected]>; +Cc: pgsql-hackers

On Thu, May 11, 2017 at 10:48 AM, Michael Paquier
<[email protected]> wrote:
> Hi all,
>
> I had my eyes on the WAL sender code this morning, and I have noticed
> that walsender.c is not completely consistent with the PID lookups it
> does in walsender.c. In two code paths, the PID value is checked
> without holding the WAL sender spin lock (WalSndRqstFileReload and
> pg_stat_get_wal_senders), which looks like a very bad idea contrary to
> what the new WalSndWaitStopping() does and what InitWalSenderSlot() is
> doing for ages.
>
> Any thoughts about the patch attached to make things more consistent?
> It seems to me that having some safeguards would be nice for
> robustness.

+1. I think this is a sensible change.

Regards,

--
Masahiko Sawada
NIPPON TELEGRAPH AND TELEPHONE CORPORATION
NTT Open Source Software Center


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-05-15 03:25  Kyotaro HORIGUCHI <[email protected]>
  parent: Masahiko Sawada <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro HORIGUCHI @ 2017-05-15 03:25 UTC (permalink / raw)
  To: [email protected]; +Cc: [email protected]; pgsql-hackers

At Fri, 12 May 2017 11:44:19 +0900, Masahiko Sawada <[email protected]> wrote in <CAD21AoA15xsu1gbOH=L1XU7g7zuKk1UACtOz+-mqOwP1_xBC_g@mail.gmail.com>
> On Thu, May 11, 2017 at 10:48 AM, Michael Paquier
> <[email protected]> wrote:
> > Hi all,
> >
> > I had my eyes on the WAL sender code this morning, and I have noticed
> > that walsender.c is not completely consistent with the PID lookups it
> > does in walsender.c. In two code paths, the PID value is checked
> > without holding the WAL sender spin lock (WalSndRqstFileReload and
> > pg_stat_get_wal_senders), which looks like a very bad idea contrary to
> > what the new WalSndWaitStopping() does and what InitWalSenderSlot() is
> > doing for ages.
> >
> > Any thoughts about the patch attached to make things more consistent?
> > It seems to me that having some safeguards would be nice for
> > robustness.
> 
> +1. I think this is a sensible change.

It intends to avoid exccesive locking during looking up stats
values. But we don't have so much vacant WanSnd slots in a
reasonable setup. Thus it seems reasonable to read the pid value
within the lock section since it adds practically no additional
cost. pg_stat_get_wal_receiver seems to need the same amendment
since the code is a parallel to that of wal receiver.

Or, if we were too sensitive to such locks for nothing, we could
use double-checked locking but I don't think we are so here.


In short, +1 too and walreceiver needs the same amendment.

regards,

-- 
Kyotaro Horiguchi
NTT Open Source Software Center



-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-05-16 15:14  Robert Haas <[email protected]>
  parent: Michael Paquier <[email protected]>
  2 siblings, 1 reply; 160+ messages in thread

From: Robert Haas @ 2017-05-16 15:14 UTC (permalink / raw)
  To: Michael Paquier <[email protected]>; +Cc: pgsql-hackers

On Wed, May 10, 2017 at 9:48 PM, Michael Paquier
<[email protected]> wrote:
> I had my eyes on the WAL sender code this morning, and I have noticed
> that walsender.c is not completely consistent with the PID lookups it
> does in walsender.c. In two code paths, the PID value is checked
> without holding the WAL sender spin lock (WalSndRqstFileReload and
> pg_stat_get_wal_senders), which looks like a very bad idea contrary to
> what the new WalSndWaitStopping() does and what InitWalSenderSlot() is
> doing for ages.

Well, it's probably worth changing for consistency, but I'm not sure
that it rises to the level of "a very bad idea".  It actually seems
almost entirely harmless.  Spuriously setting the needreload flag on a
just-deceased WAL sender will just result in some future WAL sender
doing a bit of unnecessary work, but I don't think it breaks anything
and the probability is vanishingly low.  The other change could result
a bogus 0 PID in pg_stat_get_wal_senders output, but I bet you
couldn't reproduce that more than once in a blue moon even with a test
rig designed to provoke it, and if it does happen it isn't really
anything more than a trivial annoyance.

So I'm in favor of committing this and maybe even back-patching it,
but I also don't think it's a big deal.

-- 
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-05-18 02:25  Michael Paquier <[email protected]>
  parent: Robert Haas <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Michael Paquier @ 2017-05-18 02:25 UTC (permalink / raw)
  To: Robert Haas <[email protected]>; +Cc: pgsql-hackers

On Wed, May 17, 2017 at 12:14 AM, Robert Haas <[email protected]> wrote:
> Well, it's probably worth changing for consistency, but I'm not sure
> that it rises to the level of "a very bad idea".  It actually seems
> almost entirely harmless.  Spuriously setting the needreload flag on a
> just-deceased WAL sender will just result in some future WAL sender
> doing a bit of unnecessary work, but I don't think it breaks anything
> and the probability is vanishingly low.  The other change could result
> a bogus 0 PID in pg_stat_get_wal_senders output, but I bet you
> couldn't reproduce that more than once in a blue moon even with a test
> rig designed to provoke it, and if it does happen it isn't really
> anything more than a trivial annoyance.

Well, the window is very low, so only tests with precisely taken
breakpoints would show problems.

> So I'm in favor of committing this and maybe even back-patching it,
> but I also don't think it's a big deal.

Thanks. I would not mind if this is seen as a HEAD-only improvement.
-- 
Michael


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-05-18 04:43  Thomas Munro <[email protected]>
  parent: Michael Paquier <[email protected]>
  2 siblings, 1 reply; 160+ messages in thread

From: Thomas Munro @ 2017-05-18 04:43 UTC (permalink / raw)
  To: Michael Paquier <[email protected]>; +Cc: pgsql-hackers

On Thu, May 11, 2017 at 1:48 PM, Michael Paquier
<[email protected]> wrote:
> I had my eyes on the WAL sender code this morning, and I have noticed
> that walsender.c is not completely consistent with the PID lookups it
> does in walsender.c. In two code paths, the PID value is checked
> without holding the WAL sender spin lock (WalSndRqstFileReload and
> pg_stat_get_wal_senders), which looks like a very bad idea contrary to
> what the new WalSndWaitStopping() does and what InitWalSenderSlot() is
> doing for ages.

There is also code that accesses shared walsender state without
spinlocks over in syncrep.c.  I think that file could use a few words
of explanation for why it's OK to access pid, state and flush without
synchronisation.

-- 
Thomas Munro
http://www.enterprisedb.com


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-05-19 02:05  Michael Paquier <[email protected]>
  parent: Thomas Munro <[email protected]>
  0 siblings, 2 replies; 160+ messages in thread

From: Michael Paquier @ 2017-05-19 02:05 UTC (permalink / raw)
  To: Thomas Munro <[email protected]>; +Cc: pgsql-hackers

On Thu, May 18, 2017 at 1:43 PM, Thomas Munro
<[email protected]> wrote:
> On Thu, May 11, 2017 at 1:48 PM, Michael Paquier
> <[email protected]> wrote:
>> I had my eyes on the WAL sender code this morning, and I have noticed
>> that walsender.c is not completely consistent with the PID lookups it
>> does in walsender.c. In two code paths, the PID value is checked
>> without holding the WAL sender spin lock (WalSndRqstFileReload and
>> pg_stat_get_wal_senders), which looks like a very bad idea contrary to
>> what the new WalSndWaitStopping() does and what InitWalSenderSlot() is
>> doing for ages.
>
> There is also code that accesses shared walsender state without
> spinlocks over in syncrep.c.  I think that file could use a few words
> of explanation for why it's OK to access pid, state and flush without
> synchronisation.

Yes, that is read during the quorum and priority sync evaluation.
Except sync_standby_priority, all the other variables should be
protected using the spin lock of the WAL sender. walsender_private.h
is clear regarding that. So the current coding is inconsistent even
there. Attached is an updated patch.
-- 
Michael


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers


Attachments:

  [application/octet-stream] walsnd-pid-races-v2.patch (4.0K, ../../CAB7nPqRa3EpFedbW8RuHpmSn8u_fdzV3oCRhRndFqWfELSvHVg@mail.gmail.com/2-walsnd-pid-races-v2.patch)
  download | inline diff:
diff --git a/src/backend/replication/syncrep.c b/src/backend/replication/syncrep.c
index ad213fc454..5ab5e95fa9 100644
--- a/src/backend/replication/syncrep.c
+++ b/src/backend/replication/syncrep.c
@@ -409,15 +409,23 @@ void
 SyncRepReleaseWaiters(void)
 {
 	volatile WalSndCtlData *walsndctl = WalSndCtl;
+	WalSnd	   *walsnd = MyWalSnd;
 	XLogRecPtr	writePtr;
 	XLogRecPtr	flushPtr;
 	XLogRecPtr	applyPtr;
+	XLogRecPtr	flush;
+	WalSndState	state;
 	bool		got_recptr;
 	bool		am_sync;
 	int			numwrite = 0;
 	int			numflush = 0;
 	int			numapply = 0;
 
+	SpinLockAcquire(&walsnd->mutex);
+	flush = walsnd->flush;
+	state = walsnd->state;
+	SpinLockRelease(&walsnd->mutex);
+
 	/*
 	 * If this WALSender is serving a standby that is not on the list of
 	 * potential sync standbys then we have nothing to do. If we are still
@@ -425,8 +433,8 @@ SyncRepReleaseWaiters(void)
 	 * still invalid, then leave quickly also.
 	 */
 	if (MyWalSnd->sync_standby_priority == 0 ||
-		MyWalSnd->state < WALSNDSTATE_STREAMING ||
-		XLogRecPtrIsInvalid(MyWalSnd->flush))
+		state < WALSNDSTATE_STREAMING ||
+		XLogRecPtrIsInvalid(flush))
 	{
 		announce_next_takeover = true;
 		return;
@@ -711,14 +719,24 @@ SyncRepGetSyncStandbysQuorum(bool *am_sync)
 
 	for (i = 0; i < max_wal_senders; i++)
 	{
+		XLogRecPtr	flush;
+		WalSndState	state;
+		int			pid;
+
 		walsnd = &WalSndCtl->walsnds[i];
 
+		SpinLockAcquire(&walsnd->mutex);
+		pid = walsnd->pid;
+		flush = walsnd->flush;
+		state = walsnd->state;
+		SpinLockRelease(&walsnd->mutex);
+
 		/* Must be active */
-		if (walsnd->pid == 0)
+		if (pid == 0)
 			continue;
 
 		/* Must be streaming */
-		if (walsnd->state != WALSNDSTATE_STREAMING)
+		if (state != WALSNDSTATE_STREAMING)
 			continue;
 
 		/* Must be synchronous */
@@ -726,7 +744,7 @@ SyncRepGetSyncStandbysQuorum(bool *am_sync)
 			continue;
 
 		/* Must have a valid flush position */
-		if (XLogRecPtrIsInvalid(walsnd->flush))
+		if (XLogRecPtrIsInvalid(flush))
 			continue;
 
 		/*
@@ -780,14 +798,24 @@ SyncRepGetSyncStandbysPriority(bool *am_sync)
 	 */
 	for (i = 0; i < max_wal_senders; i++)
 	{
+		XLogRecPtr	flush;
+		WalSndState	state;
+		int			pid;
+
 		walsnd = &WalSndCtl->walsnds[i];
 
+		SpinLockAcquire(&walsnd->mutex);
+		pid = walsnd->pid;
+		flush = walsnd->flush;
+		state = walsnd->state;
+		SpinLockRelease(&walsnd->mutex);
+
 		/* Must be active */
-		if (walsnd->pid == 0)
+		if (pid == 0)
 			continue;
 
 		/* Must be streaming */
-		if (walsnd->state != WALSNDSTATE_STREAMING)
+		if (state != WALSNDSTATE_STREAMING)
 			continue;
 
 		/* Must be synchronous */
@@ -796,7 +824,7 @@ SyncRepGetSyncStandbysPriority(bool *am_sync)
 			continue;
 
 		/* Must have a valid flush position */
-		if (XLogRecPtrIsInvalid(walsnd->flush))
+		if (XLogRecPtrIsInvalid(flush))
 			continue;
 
 		/*
diff --git a/src/backend/replication/walsender.c b/src/backend/replication/walsender.c
index 49cce38880..a04be8039d 100644
--- a/src/backend/replication/walsender.c
+++ b/src/backend/replication/walsender.c
@@ -2874,10 +2874,12 @@ WalSndRqstFileReload(void)
 	{
 		WalSnd	   *walsnd = &WalSndCtl->walsnds[i];
 
+		SpinLockAcquire(&walsnd->mutex);
 		if (walsnd->pid == 0)
+		{
+			SpinLockRelease(&walsnd->mutex);
 			continue;
-
-		SpinLockAcquire(&walsnd->mutex);
+		}
 		walsnd->needreload = true;
 		SpinLockRelease(&walsnd->mutex);
 	}
@@ -3190,14 +3192,18 @@ pg_stat_get_wal_senders(PG_FUNCTION_ARGS)
 		TimeOffset	flushLag;
 		TimeOffset	applyLag;
 		int			priority;
+		int			pid;
 		WalSndState state;
 		Datum		values[PG_STAT_GET_WAL_SENDERS_COLS];
 		bool		nulls[PG_STAT_GET_WAL_SENDERS_COLS];
 
+		SpinLockAcquire(&walsnd->mutex);
 		if (walsnd->pid == 0)
+		{
+			SpinLockRelease(&walsnd->mutex);
 			continue;
-
-		SpinLockAcquire(&walsnd->mutex);
+		}
+		pid = walsnd->pid;
 		sentPtr = walsnd->sentPtr;
 		state = walsnd->state;
 		write = walsnd->write;
@@ -3210,7 +3216,7 @@ pg_stat_get_wal_senders(PG_FUNCTION_ARGS)
 		SpinLockRelease(&walsnd->mutex);
 
 		memset(nulls, 0, sizeof(nulls));
-		values[0] = Int32GetDatum(walsnd->pid);
+		values[0] = Int32GetDatum(pid);
 
 		if (!superuser())
 		{


^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-05-19 06:01  Masahiko Sawada <[email protected]>
  parent: Michael Paquier <[email protected]>
  1 sibling, 1 reply; 160+ messages in thread

From: Masahiko Sawada @ 2017-05-19 06:01 UTC (permalink / raw)
  To: Michael Paquier <[email protected]>; +Cc: Thomas Munro <[email protected]>; pgsql-hackers

On Fri, May 19, 2017 at 11:05 AM, Michael Paquier
<[email protected]> wrote:
> On Thu, May 18, 2017 at 1:43 PM, Thomas Munro
> <[email protected]> wrote:
>> On Thu, May 11, 2017 at 1:48 PM, Michael Paquier
>> <[email protected]> wrote:
>>> I had my eyes on the WAL sender code this morning, and I have noticed
>>> that walsender.c is not completely consistent with the PID lookups it
>>> does in walsender.c. In two code paths, the PID value is checked
>>> without holding the WAL sender spin lock (WalSndRqstFileReload and
>>> pg_stat_get_wal_senders), which looks like a very bad idea contrary to
>>> what the new WalSndWaitStopping() does and what InitWalSenderSlot() is
>>> doing for ages.
>>
>> There is also code that accesses shared walsender state without
>> spinlocks over in syncrep.c.  I think that file could use a few words
>> of explanation for why it's OK to access pid, state and flush without
>> synchronisation.
>
> Yes, that is read during the quorum and priority sync evaluation.
> Except sync_standby_priority, all the other variables should be
> protected using the spin lock of the WAL sender. walsender_private.h
> is clear regarding that. So the current coding is inconsistent even
> there. Attached is an updated patch.

Also, as Horiguchi-san pointed out earlier, walreceiver seems need the
similar fix.

Regards,

--
Masahiko Sawada
NIPPON TELEGRAPH AND TELEPHONE CORPORATION
NTT Open Source Software Center


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-05-20 12:40  Michael Paquier <[email protected]>
  parent: Masahiko Sawada <[email protected]>
  0 siblings, 3 replies; 160+ messages in thread

From: Michael Paquier @ 2017-05-20 12:40 UTC (permalink / raw)
  To: Masahiko Sawada <[email protected]>; +Cc: Thomas Munro <[email protected]>; pgsql-hackers

On Fri, May 19, 2017 at 3:01 PM, Masahiko Sawada <[email protected]> wrote:
> Also, as Horiguchi-san pointed out earlier, walreceiver seems need the
> similar fix.

Actually, now that I look at it, ready_to_display should as well be
protected by the lock of the WAL receiver, so it is incorrectly placed
in walreceiver.h. As you are pointing out, pg_stat_get_wal_receiver()
is lazy as well, and that's new in 10, so we have an open item here
for both of them. And I am the author for both things. No issues
spotted in walreceiverfuncs.c after review.

I am adding an open item so as both issues are fixed in PG10. With the
WAL sender part, I think that this should be a group shot.

So what do you think about the attached?
-- 
Michael


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers


Attachments:

  [text/x-patch] walsnd-pid-races-v3.patch (6.6K, ../../CAB7nPqSTsnK+b53s-NZUuTPosJmjbV+zKHV3zHr7K6ZKFMVY_A@mail.gmail.com/2-walsnd-pid-races-v3.patch)
  download | inline diff:
diff --git a/src/backend/replication/syncrep.c b/src/backend/replication/syncrep.c
index ad213fc454..5ab5e95fa9 100644
--- a/src/backend/replication/syncrep.c
+++ b/src/backend/replication/syncrep.c
@@ -409,15 +409,23 @@ void
 SyncRepReleaseWaiters(void)
 {
 	volatile WalSndCtlData *walsndctl = WalSndCtl;
+	WalSnd	   *walsnd = MyWalSnd;
 	XLogRecPtr	writePtr;
 	XLogRecPtr	flushPtr;
 	XLogRecPtr	applyPtr;
+	XLogRecPtr	flush;
+	WalSndState	state;
 	bool		got_recptr;
 	bool		am_sync;
 	int			numwrite = 0;
 	int			numflush = 0;
 	int			numapply = 0;
 
+	SpinLockAcquire(&walsnd->mutex);
+	flush = walsnd->flush;
+	state = walsnd->state;
+	SpinLockRelease(&walsnd->mutex);
+
 	/*
 	 * If this WALSender is serving a standby that is not on the list of
 	 * potential sync standbys then we have nothing to do. If we are still
@@ -425,8 +433,8 @@ SyncRepReleaseWaiters(void)
 	 * still invalid, then leave quickly also.
 	 */
 	if (MyWalSnd->sync_standby_priority == 0 ||
-		MyWalSnd->state < WALSNDSTATE_STREAMING ||
-		XLogRecPtrIsInvalid(MyWalSnd->flush))
+		state < WALSNDSTATE_STREAMING ||
+		XLogRecPtrIsInvalid(flush))
 	{
 		announce_next_takeover = true;
 		return;
@@ -711,14 +719,24 @@ SyncRepGetSyncStandbysQuorum(bool *am_sync)
 
 	for (i = 0; i < max_wal_senders; i++)
 	{
+		XLogRecPtr	flush;
+		WalSndState	state;
+		int			pid;
+
 		walsnd = &WalSndCtl->walsnds[i];
 
+		SpinLockAcquire(&walsnd->mutex);
+		pid = walsnd->pid;
+		flush = walsnd->flush;
+		state = walsnd->state;
+		SpinLockRelease(&walsnd->mutex);
+
 		/* Must be active */
-		if (walsnd->pid == 0)
+		if (pid == 0)
 			continue;
 
 		/* Must be streaming */
-		if (walsnd->state != WALSNDSTATE_STREAMING)
+		if (state != WALSNDSTATE_STREAMING)
 			continue;
 
 		/* Must be synchronous */
@@ -726,7 +744,7 @@ SyncRepGetSyncStandbysQuorum(bool *am_sync)
 			continue;
 
 		/* Must have a valid flush position */
-		if (XLogRecPtrIsInvalid(walsnd->flush))
+		if (XLogRecPtrIsInvalid(flush))
 			continue;
 
 		/*
@@ -780,14 +798,24 @@ SyncRepGetSyncStandbysPriority(bool *am_sync)
 	 */
 	for (i = 0; i < max_wal_senders; i++)
 	{
+		XLogRecPtr	flush;
+		WalSndState	state;
+		int			pid;
+
 		walsnd = &WalSndCtl->walsnds[i];
 
+		SpinLockAcquire(&walsnd->mutex);
+		pid = walsnd->pid;
+		flush = walsnd->flush;
+		state = walsnd->state;
+		SpinLockRelease(&walsnd->mutex);
+
 		/* Must be active */
-		if (walsnd->pid == 0)
+		if (pid == 0)
 			continue;
 
 		/* Must be streaming */
-		if (walsnd->state != WALSNDSTATE_STREAMING)
+		if (state != WALSNDSTATE_STREAMING)
 			continue;
 
 		/* Must be synchronous */
@@ -796,7 +824,7 @@ SyncRepGetSyncStandbysPriority(bool *am_sync)
 			continue;
 
 		/* Must have a valid flush position */
-		if (XLogRecPtrIsInvalid(walsnd->flush))
+		if (XLogRecPtrIsInvalid(flush))
 			continue;
 
 		/*
diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c
index 2723612718..25f12e0706 100644
--- a/src/backend/replication/walreceiver.c
+++ b/src/backend/replication/walreceiver.c
@@ -1392,23 +1392,13 @@ pg_stat_get_wal_receiver(PG_FUNCTION_ARGS)
 	TimestampTz latest_end_time;
 	char	   *slotname;
 	char	   *conninfo;
-
-	/*
-	 * No WAL receiver (or not ready yet), just return a tuple with NULL
-	 * values
-	 */
-	if (walrcv->pid == 0 || !walrcv->ready_to_display)
-		PG_RETURN_NULL();
-
-	/* determine result type */
-	if (get_call_result_type(fcinfo, NULL, &tupdesc) != TYPEFUNC_COMPOSITE)
-		elog(ERROR, "return type must be a row type");
-
-	values = palloc0(sizeof(Datum) * tupdesc->natts);
-	nulls = palloc0(sizeof(bool) * tupdesc->natts);
+	int			pid;
+	bool		ready_to_display;
 
 	/* Take a lock to ensure value consistency */
 	SpinLockAcquire(&walrcv->mutex);
+	pid = walrcv->pid;
+	ready_to_display = walrcv->ready_to_display;
 	state = walrcv->walRcvState;
 	receive_start_lsn = walrcv->receiveStart;
 	receive_start_tli = walrcv->receiveStartTLI;
@@ -1422,8 +1412,22 @@ pg_stat_get_wal_receiver(PG_FUNCTION_ARGS)
 	conninfo = pstrdup(walrcv->conninfo);
 	SpinLockRelease(&walrcv->mutex);
 
+	/*
+	 * No WAL receiver (or not ready yet), just return a tuple with NULL
+	 * values
+	 */
+	if (pid == 0 || !ready_to_display)
+		PG_RETURN_NULL();
+
+	/* determine result type */
+	if (get_call_result_type(fcinfo, NULL, &tupdesc) != TYPEFUNC_COMPOSITE)
+		elog(ERROR, "return type must be a row type");
+
+	values = palloc0(sizeof(Datum) * tupdesc->natts);
+	nulls = palloc0(sizeof(bool) * tupdesc->natts);
+
 	/* Fetch values */
-	values[0] = Int32GetDatum(walrcv->pid);
+	values[0] = Int32GetDatum(pid);
 
 	if (!is_member_of_role(GetUserId(), DEFAULT_ROLE_READ_ALL_STATS))
 	{
diff --git a/src/backend/replication/walsender.c b/src/backend/replication/walsender.c
index 49cce38880..a04be8039d 100644
--- a/src/backend/replication/walsender.c
+++ b/src/backend/replication/walsender.c
@@ -2874,10 +2874,12 @@ WalSndRqstFileReload(void)
 	{
 		WalSnd	   *walsnd = &WalSndCtl->walsnds[i];
 
+		SpinLockAcquire(&walsnd->mutex);
 		if (walsnd->pid == 0)
+		{
+			SpinLockRelease(&walsnd->mutex);
 			continue;
-
-		SpinLockAcquire(&walsnd->mutex);
+		}
 		walsnd->needreload = true;
 		SpinLockRelease(&walsnd->mutex);
 	}
@@ -3190,14 +3192,18 @@ pg_stat_get_wal_senders(PG_FUNCTION_ARGS)
 		TimeOffset	flushLag;
 		TimeOffset	applyLag;
 		int			priority;
+		int			pid;
 		WalSndState state;
 		Datum		values[PG_STAT_GET_WAL_SENDERS_COLS];
 		bool		nulls[PG_STAT_GET_WAL_SENDERS_COLS];
 
+		SpinLockAcquire(&walsnd->mutex);
 		if (walsnd->pid == 0)
+		{
+			SpinLockRelease(&walsnd->mutex);
 			continue;
-
-		SpinLockAcquire(&walsnd->mutex);
+		}
+		pid = walsnd->pid;
 		sentPtr = walsnd->sentPtr;
 		state = walsnd->state;
 		write = walsnd->write;
@@ -3210,7 +3216,7 @@ pg_stat_get_wal_senders(PG_FUNCTION_ARGS)
 		SpinLockRelease(&walsnd->mutex);
 
 		memset(nulls, 0, sizeof(nulls));
-		values[0] = Int32GetDatum(walsnd->pid);
+		values[0] = Int32GetDatum(pid);
 
 		if (!superuser())
 		{
diff --git a/src/include/replication/walreceiver.h b/src/include/replication/walreceiver.h
index 31d090c99d..44199670b1 100644
--- a/src/include/replication/walreceiver.h
+++ b/src/include/replication/walreceiver.h
@@ -114,6 +114,9 @@ typedef struct
 	 */
 	char		slotname[NAMEDATALEN];
 
+	/* set true once conninfo is ready to display (obfuscated pwds etc) */
+	bool		ready_to_display;
+
 	slock_t		mutex;			/* locks shared variables shown above */
 
 	/*
@@ -122,9 +125,6 @@ typedef struct
 	 */
 	bool		force_reply;
 
-	/* set true once conninfo is ready to display (obfuscated pwds etc) */
-	bool		ready_to_display;
-
 	/*
 	 * Latch used by startup process to wake up walreceiver after telling it
 	 * where to start streaming (after setting receiveStart and


^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-05-21 04:37  Erik Rijkers <[email protected]>
  parent: Michael Paquier <[email protected]>
  2 siblings, 1 reply; 160+ messages in thread

From: Erik Rijkers @ 2017-05-21 04:37 UTC (permalink / raw)
  To: Michael Paquier <[email protected]>; +Cc: Masahiko Sawada <[email protected]>; Thomas Munro <[email protected]>; pgsql-hackers; [email protected]

On 2017-05-20 14:40, Michael Paquier wrote:
> On Fri, May 19, 2017 at 3:01 PM, Masahiko Sawada 
> <[email protected]> wrote:
>> Also, as Horiguchi-san pointed out earlier, walreceiver seems need the
>> similar fix.
> 
> Actually, now that I look at it, ready_to_display should as well be
> protected by the lock of the WAL receiver, so it is incorrectly placed
> in walreceiver.h. As you are pointing out, pg_stat_get_wal_receiver()
> is lazy as well, and that's new in 10, so we have an open item here
> for both of them. And I am the author for both things. No issues
> spotted in walreceiverfuncs.c after review.
> 
> I am adding an open item so as both issues are fixed in PG10. With the
> WAL sender part, I think that this should be a group shot.
> 
> So what do you think about the attached?

> [walsnd-pid-races-v3.patch]


With this patch on current master my logical replication tests 
(pgbench-over-logical-replication) run without errors for the first time 
in many days (even weeks).

I'll do still more and longer tests but I have gathered already a long 
streak of successful runs since you posted the patch so I am getting 
convinced this patch is solved the problem that I was experiencing.

Pity it didn't make the beta.


thanks,

Erik Rijkers



-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-05-21 06:19  Erik Rijkers <[email protected]>
  parent: Erik Rijkers <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Erik Rijkers @ 2017-05-21 06:19 UTC (permalink / raw)
  To: Michael Paquier <[email protected]>; +Cc: Masahiko Sawada <[email protected]>; Thomas Munro <[email protected]>; pgsql-hackers; [email protected]

On 2017-05-21 06:37, Erik Rijkers wrote:
> On 2017-05-20 14:40, Michael Paquier wrote:
>> On Fri, May 19, 2017 at 3:01 PM, Masahiko Sawada 
>> <[email protected]> wrote:
>>> Also, as Horiguchi-san pointed out earlier, walreceiver seems need 
>>> the
>>> similar fix.
>> 
>> Actually, now that I look at it, ready_to_display should as well be
>> protected by the lock of the WAL receiver, so it is incorrectly placed
>> in walreceiver.h. As you are pointing out, pg_stat_get_wal_receiver()
>> is lazy as well, and that's new in 10, so we have an open item here
>> for both of them. And I am the author for both things. No issues
>> spotted in walreceiverfuncs.c after review.
>> 
>> I am adding an open item so as both issues are fixed in PG10. With the
>> WAL sender part, I think that this should be a group shot.
>> 
>> So what do you think about the attached?
> 
>> [walsnd-pid-races-v3.patch]
> 
> 
> With this patch on current master my logical replication tests
> (pgbench-over-logical-replication) run without errors for the first
> time in many days (even weeks).

Unfortunately, just now another logical-replication failure occurred.  
The same as I have seen all along:

The symptom: after starting logical replication, there are no rows in 
pg_stat_replication and in the replica-log logical replication complains 
about max_replication_slots being too low. (from previous experience I 
know that making max_replication_slots higher does indeed 'help', but 
only until the next (same) error occurs, with renewed (same) complaint).

Also from previous experience of this failed state I know that it can be 
'cleaned up' by
manually emptying these tables:
   delete from pg_subscription_rel;
   delete from pg_subscription;
   delete from pg_replication_origin;
Then it becomes possible to start a new subscription without the above 
symptoms.

I'll do some more testing and hopefully get some information that's less 
vague...


Erik Rijkers



-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-05-30 01:38  Noah Misch <[email protected]>
  parent: Michael Paquier <[email protected]>
  2 siblings, 1 reply; 160+ messages in thread

From: Noah Misch @ 2017-05-30 01:38 UTC (permalink / raw)
  To: [email protected]; +Cc: Michael Paquier <[email protected]>; Masahiko Sawada <[email protected]>; Thomas Munro <[email protected]>; pgsql-hackers

On Sat, May 20, 2017 at 09:40:57PM +0900, Michael Paquier wrote:
> On Fri, May 19, 2017 at 3:01 PM, Masahiko Sawada <[email protected]> wrote:
> > Also, as Horiguchi-san pointed out earlier, walreceiver seems need the
> > similar fix.
> 
> Actually, now that I look at it, ready_to_display should as well be
> protected by the lock of the WAL receiver, so it is incorrectly placed
> in walreceiver.h. As you are pointing out, pg_stat_get_wal_receiver()
> is lazy as well, and that's new in 10, so we have an open item here
> for both of them. And I am the author for both things. No issues
> spotted in walreceiverfuncs.c after review.
> 
> I am adding an open item so as both issues are fixed in PG10. With the
> WAL sender part, I think that this should be a group shot.
> 
> So what do you think about the attached?

[Action required within three days.  This is a generic notification.]

The above-described topic is currently a PostgreSQL 10 open item.  Peter,
since you committed the patch believed to have created it, you own this open
item.  If some other commit is more relevant or if this does not belong as a
v10 open item, please let us know.  Otherwise, please observe the policy on
open item ownership[1] and send a status update within three calendar days of
this message.  Include a date for your subsequent status update.  Testers may
discover new open items at any time, and I want to plan to get them all fixed
well in advance of shipping v10.  Consequently, I will appreciate your efforts
toward speedy resolution.  Thanks.

[1] https://www.postgresql.org/message-id/20170404140717.GA2675809%40tornado.leadboat.com


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-05-30 18:09  Peter Eisentraut <[email protected]>
  parent: Noah Misch <[email protected]>
  0 siblings, 1 reply; 160+ messages in thread

From: Peter Eisentraut @ 2017-05-30 18:09 UTC (permalink / raw)
  To: Noah Misch <[email protected]>; +Cc: Michael Paquier <[email protected]>; Masahiko Sawada <[email protected]>; Thomas Munro <[email protected]>; pgsql-hackers

On 5/29/17 21:38, Noah Misch wrote:
>> Actually, now that I look at it, ready_to_display should as well be
>> protected by the lock of the WAL receiver, so it is incorrectly placed
>> in walreceiver.h. As you are pointing out, pg_stat_get_wal_receiver()
>> is lazy as well, and that's new in 10, so we have an open item here
>> for both of them. And I am the author for both things. No issues
>> spotted in walreceiverfuncs.c after review.
>>
>> I am adding an open item so as both issues are fixed in PG10. With the
>> WAL sender part, I think that this should be a group shot.
>>
>> So what do you think about the attached?
> 
> [Action required within three days.  This is a generic notification.]
> 
> The above-described topic is currently a PostgreSQL 10 open item.  Peter,
> since you committed the patch believed to have created it, you own this open
> item.

I don't think this is my item.  Most of the behavior is old, and
pg_stat_get_wal_receiver() is from commit
b1a9bad9e744857291c7d5516080527da8219854.

(The logical replication equivalent of that is
pg_stat_get_subscription(), which does appear to use appropriate locking.)

I would appreciate if another committer can take the lead on this.

-- 
Peter Eisentraut              http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-05-30 18:13  Michael Paquier <[email protected]>
  parent: Peter Eisentraut <[email protected]>
  0 siblings, 1 reply; 160+ messages in thread

From: Michael Paquier @ 2017-05-30 18:13 UTC (permalink / raw)
  To: Peter Eisentraut <[email protected]>; +Cc: Noah Misch <[email protected]>; Masahiko Sawada <[email protected]>; Thomas Munro <[email protected]>; pgsql-hackers

On Tue, May 30, 2017 at 11:09 AM, Peter Eisentraut
<[email protected]> wrote:
> I don't think this is my item.  Most of the behavior is old, and
> pg_stat_get_wal_receiver() is from commit
> b1a9bad9e744857291c7d5516080527da8219854.
>
> I would appreciate if another committer can take the lead on this.

Those things are on Alvaro's plate for the WAL receiver portion, and I
was the author of those patches. The WAL sender portion is older
though, but it seems crazy to me to not fix both things at the same
time per their similarities.
-- 
Michael


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-05-30 22:23  Thomas Munro <[email protected]>
  parent: Michael Paquier <[email protected]>
  1 sibling, 0 replies; 160+ messages in thread

From: Thomas Munro @ 2017-05-30 22:23 UTC (permalink / raw)
  To: Michael Paquier <[email protected]>; +Cc: pgsql-hackers

On Fri, May 19, 2017 at 2:05 PM, Michael Paquier
<[email protected]> wrote:
> On Thu, May 18, 2017 at 1:43 PM, Thomas Munro
> <[email protected]> wrote:
>> On Thu, May 11, 2017 at 1:48 PM, Michael Paquier
>> <[email protected]> wrote:
>>> I had my eyes on the WAL sender code this morning, and I have noticed
>>> that walsender.c is not completely consistent with the PID lookups it
>>> does in walsender.c. In two code paths, the PID value is checked
>>> without holding the WAL sender spin lock (WalSndRqstFileReload and
>>> pg_stat_get_wal_senders), which looks like a very bad idea contrary to
>>> what the new WalSndWaitStopping() does and what InitWalSenderSlot() is
>>> doing for ages.
>>
>> There is also code that accesses shared walsender state without
>> spinlocks over in syncrep.c.  I think that file could use a few words
>> of explanation for why it's OK to access pid, state and flush without
>> synchronisation.
>
> Yes, that is read during the quorum and priority sync evaluation.
> Except sync_standby_priority, all the other variables should be
> protected using the spin lock of the WAL sender. walsender_private.h
> is clear regarding that. So the current coding is inconsistent even
> there. Attached is an updated patch.

I don't claim that that stuff is wrong, just that it would be good to
hear an analysis.  I think the question is: is there a way for syncrep
to hang because of a perfectly timed walsender transition, however
vanishingly unlikely?  I'm thinking of something like: syncrep skips a
walsender slot because it's looking at arbitrarily old 'pid' from
before a walsender connected, or gets a torn read of 'flush' that
comes out as 0 but was actually non-0.

Incidentally, I suspect that a couple of places where 'volatile' is
used it's superfluous (accessing things protected by an LWLock that is
held).

I don't see any of this as 'open item' material, it's interesting to
look into but it's preexisting code.  As for unlocked reads used for
pg_stat_X views, it seems well established that we're OK with that (at
least for things that the project has decided can be read atomically,
to wit aligned 32 bit values).

-- 
Thomas Munro
http://www.enterprisedb.com


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-06-02 05:21  Noah Misch <[email protected]>
  parent: Michael Paquier <[email protected]>
  0 siblings, 1 reply; 160+ messages in thread

From: Noah Misch @ 2017-06-02 05:21 UTC (permalink / raw)
  To: Michael Paquier <[email protected]>; +Cc: Peter Eisentraut <[email protected]>; Masahiko Sawada <[email protected]>; Thomas Munro <[email protected]>; pgsql-hackers

On Sun, May 21, 2017 at 08:19:34AM +0200, Erik Rijkers wrote:
> On 2017-05-21 06:37, Erik Rijkers wrote:
> >With this patch on current master my logical replication tests
> >(pgbench-over-logical-replication) run without errors for the first
> >time in many days (even weeks).
> 
> Unfortunately, just now another logical-replication failure occurred.  The
> same as I have seen all along:

This creates a rebuttable presumption of logical replication being the cause
of this open item.  (I am not stating an opinion on whether this rebuttable
presumption is true or is false.)  As long as that stands and Alvaro has not
explicitly requested ownership of this open item, Peter owns it.

On Tue, May 30, 2017 at 11:13:37AM -0700, Michael Paquier wrote:
> On Tue, May 30, 2017 at 11:09 AM, Peter Eisentraut
> <[email protected]> wrote:
> > I don't think this is my item.  Most of the behavior is old, and
> > pg_stat_get_wal_receiver() is from commit
> > b1a9bad9e744857291c7d5516080527da8219854.
> >
> > I would appreciate if another committer can take the lead on this.
> 
> Those things are on Alvaro's plate for the WAL receiver portion, and I
> was the author of those patches. The WAL sender portion is older
> though, but it seems crazy to me to not fix both things at the same
> time per their similarities.

As a 9.6 commit, b1a9bad cannot be the cause of a v10 open item.  If a v10
commit expanded the consequences of a pre-existing bug, the committer of that
v10 work owns this open item.  If the bug's consequences are the same in v9.6
and v10, this is ineligible to be an open item.  Which applies?


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-06-02 06:11  Michael Paquier <[email protected]>
  parent: Noah Misch <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Michael Paquier @ 2017-06-02 06:11 UTC (permalink / raw)
  To: Noah Misch <[email protected]>; +Cc: Peter Eisentraut <[email protected]>; Masahiko Sawada <[email protected]>; Thomas Munro <[email protected]>; pgsql-hackers

On Fri, Jun 2, 2017 at 2:21 PM, Noah Misch <[email protected]> wrote:
> On Tue, May 30, 2017 at 11:13:37AM -0700, Michael Paquier wrote:
>> On Tue, May 30, 2017 at 11:09 AM, Peter Eisentraut
>> <[email protected]> wrote:
>> > I don't think this is my item.  Most of the behavior is old, and
>> > pg_stat_get_wal_receiver() is from commit
>> > b1a9bad9e744857291c7d5516080527da8219854.
>> >
>> > I would appreciate if another committer can take the lead on this.
>>
>> Those things are on Alvaro's plate for the WAL receiver portion, and I
>> was the author of those patches. The WAL sender portion is older
>> though, but it seems crazy to me to not fix both things at the same
>> time per their similarities.
>
> As a 9.6 commit, b1a9bad cannot be the cause of a v10 open item.  If a v10
> commit expanded the consequences of a pre-existing bug, the committer of that
> v10 work owns this open item.  If the bug's consequences are the same in v9.6
> and v10, this is ineligible to be an open item.  Which applies?

You are right. Even 1bdae16f is from 9.6. Mea culpa. I have moved that
into the section of older bugs.
-- 
Michael


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-06-07 18:31  Robert Haas <[email protected]>
  parent: Michael Paquier <[email protected]>
  2 siblings, 2 replies; 160+ messages in thread

From: Robert Haas @ 2017-06-07 18:31 UTC (permalink / raw)
  To: Michael Paquier <[email protected]>; +Cc: Masahiko Sawada <[email protected]>; Thomas Munro <[email protected]>; pgsql-hackers

On Sat, May 20, 2017 at 8:40 AM, Michael Paquier
<[email protected]> wrote:
> On Fri, May 19, 2017 at 3:01 PM, Masahiko Sawada <[email protected]> wrote:
>> Also, as Horiguchi-san pointed out earlier, walreceiver seems need the
>> similar fix.
>
> Actually, now that I look at it, ready_to_display should as well be
> protected by the lock of the WAL receiver, so it is incorrectly placed
> in walreceiver.h. As you are pointing out, pg_stat_get_wal_receiver()
> is lazy as well, and that's new in 10, so we have an open item here
> for both of them. And I am the author for both things. No issues
> spotted in walreceiverfuncs.c after review.
>
> I am adding an open item so as both issues are fixed in PG10. With the
> WAL sender part, I think that this should be a group shot.
>
> So what do you think about the attached?

I think if you're going to fix it so that we take spinlocks on
MyWalSnd in a bunch of places that we didn't take them before, it
would make sense to fix all the places where we're accessing those
fields without a spinlock instead of only some of them, unless there
are good reasons why we only need it in some cases and not others, in
which case I think the patch should add comments to each place where
the lock was not taken explaining why it's OK.  It didn't take me and
my copy of vi very long to find instances that you didn't change.

I also think that should provide some analysis regarding the question
Thomas asked - are these actually bugs, or are they OK for some
reason?  We shouldn't just plaster the code with spinlock acquisitions
unless it's really necessary.  It seems at least possible that these
changes could cause performance regressions, and it doesn't look like
you've done any testing or provided any analysis indicating whether
that's likely to happen or not.

Basically, I don't think this patch is ready to commit.  We have
neither evidence that it is necessary nor evidence that it is
complete.  I don't want to be unfair, here, but it seems to me you
ought to do a little more legwork before throwing this over the wall
to the pool of committers.

[ Side note: Erik's report on this thread initially seemed to suggest
that we needed this patch to make logical decoding stable.  But my
impression is that this is belied by subsequent developments on other
threads, so my theory is that this patch was never really related to
the problem, but rather than by the time Erik got around to testing
this patch, other fixes had made the problems relatively rare, and the
apparently-improved results with this patch were just chance.  If that
theory is wrong, it would be good to hear about it. ]

-- 
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-06-07 21:00  Erik Rijkers <[email protected]>
  parent: Robert Haas <[email protected]>
  1 sibling, 0 replies; 160+ messages in thread

From: Erik Rijkers @ 2017-06-07 21:00 UTC (permalink / raw)
  To: Robert Haas <[email protected]>; +Cc: Michael Paquier <[email protected]>; Masahiko Sawada <[email protected]>; Thomas Munro <[email protected]>; pgsql-hackers; [email protected]

On 2017-06-07 20:31, Robert Haas wrote:

> [...]
> 
> [ Side note: Erik's report on this thread initially seemed to suggest
> that we needed this patch to make logical decoding stable.  But my
> impression is that this is belied by subsequent developments on other
> threads, so my theory is that this patch was never really related to
> the problem, but rather than by the time Erik got around to testing
> this patch, other fixes had made the problems relatively rare, and the
> apparently-improved results with this patch were just chance.  If that
> theory is wrong, it would be good to hear about it. ]

Yes, agreed; I was probably mistaken.


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-06-08 04:15  Michael Paquier <[email protected]>
  parent: Robert Haas <[email protected]>
  1 sibling, 3 replies; 160+ messages in thread

From: Michael Paquier @ 2017-06-08 04:15 UTC (permalink / raw)
  To: Robert Haas <[email protected]>; +Cc: Masahiko Sawada <[email protected]>; Thomas Munro <[email protected]>; pgsql-hackers

v

On Thu, Jun 8, 2017 at 3:31 AM, Robert Haas <[email protected]> wrote:
> I think if you're going to fix it so that we take spinlocks on
> MyWalSnd in a bunch of places that we didn't take them before, it
> would make sense to fix all the places where we're accessing those
> fields without a spinlock instead of only some of them, unless there
> are good reasons why we only need it in some cases and not others, in
> which case I think the patch should add comments to each place where
> the lock was not taken explaining why it's OK.  It didn't take me and
> my copy of vi very long to find instances that you didn't change.

I take that you are referring to the two lookups in
WalSndWaitForWal(), one in exec_replication_command(), one in
WalSndLoop() and one in WalSndDone(). First let's remember that all
the fields protected by the spinlock are only updated in a WAL sender
process, so reading them directly is safe in this context: we know
that no other processes will write them. And all those functions are
called only in the context of a WAL sender process. So the copies of
MyWalSnd that you are referring to here don't need a spinlock. Is
there something I am missing?

Actually, perhaps it would make sense to add some Assert(am_walsender)
in this code?

> I also think that should provide some analysis regarding the question
> Thomas asked - are these actually bugs, or are they OK for some
> reason?  We shouldn't just plaster the code with spinlock acquisitions
> unless it's really necessary.  It seems at least possible that these
> changes could cause performance regressions, and it doesn't look like
> you've done any testing or provided any analysis indicating whether
> that's likely to happen or not.

Now taking a step back on the patch, v3 that I sent 3 weeks ago.
Additional spinlocks are taken in:
1) SyncRepReleaseWaiters(), which is called in a WAL sender. So indeed
no spinlocks are needed here. The last patch is incorrect here.
2) SyncRepGetSyncStandbysPriority() and SyncRepGetSyncStandbysQuorum()
as of HEAD via SyncRepGetSyncStandbys(), something that's not good if
done lock-less as this gets used for pg_stat_replication for
pg_stat_get_wal_senders(). This can return a garbage list of sync
standbys to the client as WAL senders update their flush, write and
apply positions in parallel to that, and don't care about SyncRepLock
as this gets calculated with the standby feedback messages. The
problem here is the gap between walsnd->sync_standby_priority and
walsnd->flush:
-- With a primary, and more or more sync standbys, take a checkpoint
in SyncRepGetSyncStandbysPriority() after the first check is passed.
-- Shutdown the standby and restart it.
-- The standby reconnects, initializes the WAL sender slot with
InitWalSenderSlot().
-- Let the process of SyncRepGetSyncStandbysPriority() continue, the
sync standby does not show up.
That would be really unlikely to happen, but the code is written in
such a way that it could happen. One could also say that this gives a
closer idea that the standby disconnected but it looks wrong to me to
do this value lookup without a required lock.
3) In walreceiver.c's pg_stat_get_wal_receiver's:
- Launch pg_stat_get_wal_receiver and take a checkpoint on it.
- Pass the lookups of pid and ready_to_display.
- Make the WAL receiver stop.
- The view reports inconsistent data. If the WAL receiver data was
just initialized, the caller would get back PID values or similar 0.
Particularly a WAL receiver marked with !ready_to_display could show
data to the caller. That's not cool.
4) KeepFileRestoredFromArchive() needs a lock, as this is called from
the startup process. That's harmless, the worst that can happen is
that needreload is switched to true for a process that has been marked
with a PID of 0 because of a WAL sender restart (slot taken again and
initialized). But that will just process an extra reload in a worst
case.

> Basically, I don't think this patch is ready to commit.  We have
> neither evidence that it is necessary nor evidence that it is
> complete.  I don't want to be unfair, here, but it seems to me you
> ought to do a little more legwork before throwing this over the wall
> to the pool of committers.

Fair enough. Is the analysis written above more convincing?
-- 
Michael


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers


Attachments:

  [application/octet-stream] walsnd-pid-races-v4.patch (5.6K, ../../CAB7nPqTfu2P1QCZB22LJG+P4ET-pS0yW+qSTWO9JcVcCtaPZAA@mail.gmail.com/2-walsnd-pid-races-v4.patch)
  download | inline diff:
diff --git a/src/backend/replication/syncrep.c b/src/backend/replication/syncrep.c
index ad213fc454..e4ddbf2c44 100644
--- a/src/backend/replication/syncrep.c
+++ b/src/backend/replication/syncrep.c
@@ -711,14 +711,24 @@ SyncRepGetSyncStandbysQuorum(bool *am_sync)
 
 	for (i = 0; i < max_wal_senders; i++)
 	{
+		XLogRecPtr	flush;
+		WalSndState	state;
+		int			pid;
+
 		walsnd = &WalSndCtl->walsnds[i];
 
+		SpinLockAcquire(&walsnd->mutex);
+		pid = walsnd->pid;
+		flush = walsnd->flush;
+		state = walsnd->state;
+		SpinLockRelease(&walsnd->mutex);
+
 		/* Must be active */
-		if (walsnd->pid == 0)
+		if (pid == 0)
 			continue;
 
 		/* Must be streaming */
-		if (walsnd->state != WALSNDSTATE_STREAMING)
+		if (state != WALSNDSTATE_STREAMING)
 			continue;
 
 		/* Must be synchronous */
@@ -726,7 +736,7 @@ SyncRepGetSyncStandbysQuorum(bool *am_sync)
 			continue;
 
 		/* Must have a valid flush position */
-		if (XLogRecPtrIsInvalid(walsnd->flush))
+		if (XLogRecPtrIsInvalid(flush))
 			continue;
 
 		/*
@@ -780,14 +790,24 @@ SyncRepGetSyncStandbysPriority(bool *am_sync)
 	 */
 	for (i = 0; i < max_wal_senders; i++)
 	{
+		XLogRecPtr	flush;
+		WalSndState	state;
+		int			pid;
+
 		walsnd = &WalSndCtl->walsnds[i];
 
+		SpinLockAcquire(&walsnd->mutex);
+		pid = walsnd->pid;
+		flush = walsnd->flush;
+		state = walsnd->state;
+		SpinLockRelease(&walsnd->mutex);
+
 		/* Must be active */
-		if (walsnd->pid == 0)
+		if (pid == 0)
 			continue;
 
 		/* Must be streaming */
-		if (walsnd->state != WALSNDSTATE_STREAMING)
+		if (state != WALSNDSTATE_STREAMING)
 			continue;
 
 		/* Must be synchronous */
@@ -796,7 +816,7 @@ SyncRepGetSyncStandbysPriority(bool *am_sync)
 			continue;
 
 		/* Must have a valid flush position */
-		if (XLogRecPtrIsInvalid(walsnd->flush))
+		if (XLogRecPtrIsInvalid(flush))
 			continue;
 
 		/*
diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c
index 2723612718..25f12e0706 100644
--- a/src/backend/replication/walreceiver.c
+++ b/src/backend/replication/walreceiver.c
@@ -1392,23 +1392,13 @@ pg_stat_get_wal_receiver(PG_FUNCTION_ARGS)
 	TimestampTz latest_end_time;
 	char	   *slotname;
 	char	   *conninfo;
-
-	/*
-	 * No WAL receiver (or not ready yet), just return a tuple with NULL
-	 * values
-	 */
-	if (walrcv->pid == 0 || !walrcv->ready_to_display)
-		PG_RETURN_NULL();
-
-	/* determine result type */
-	if (get_call_result_type(fcinfo, NULL, &tupdesc) != TYPEFUNC_COMPOSITE)
-		elog(ERROR, "return type must be a row type");
-
-	values = palloc0(sizeof(Datum) * tupdesc->natts);
-	nulls = palloc0(sizeof(bool) * tupdesc->natts);
+	int			pid;
+	bool		ready_to_display;
 
 	/* Take a lock to ensure value consistency */
 	SpinLockAcquire(&walrcv->mutex);
+	pid = walrcv->pid;
+	ready_to_display = walrcv->ready_to_display;
 	state = walrcv->walRcvState;
 	receive_start_lsn = walrcv->receiveStart;
 	receive_start_tli = walrcv->receiveStartTLI;
@@ -1422,8 +1412,22 @@ pg_stat_get_wal_receiver(PG_FUNCTION_ARGS)
 	conninfo = pstrdup(walrcv->conninfo);
 	SpinLockRelease(&walrcv->mutex);
 
+	/*
+	 * No WAL receiver (or not ready yet), just return a tuple with NULL
+	 * values
+	 */
+	if (pid == 0 || !ready_to_display)
+		PG_RETURN_NULL();
+
+	/* determine result type */
+	if (get_call_result_type(fcinfo, NULL, &tupdesc) != TYPEFUNC_COMPOSITE)
+		elog(ERROR, "return type must be a row type");
+
+	values = palloc0(sizeof(Datum) * tupdesc->natts);
+	nulls = palloc0(sizeof(bool) * tupdesc->natts);
+
 	/* Fetch values */
-	values[0] = Int32GetDatum(walrcv->pid);
+	values[0] = Int32GetDatum(pid);
 
 	if (!is_member_of_role(GetUserId(), DEFAULT_ROLE_READ_ALL_STATS))
 	{
diff --git a/src/backend/replication/walsender.c b/src/backend/replication/walsender.c
index 976a42f86d..13d55826f9 100644
--- a/src/backend/replication/walsender.c
+++ b/src/backend/replication/walsender.c
@@ -2877,10 +2877,12 @@ WalSndRqstFileReload(void)
 	{
 		WalSnd	   *walsnd = &WalSndCtl->walsnds[i];
 
+		SpinLockAcquire(&walsnd->mutex);
 		if (walsnd->pid == 0)
+		{
+			SpinLockRelease(&walsnd->mutex);
 			continue;
-
-		SpinLockAcquire(&walsnd->mutex);
+		}
 		walsnd->needreload = true;
 		SpinLockRelease(&walsnd->mutex);
 	}
@@ -3195,14 +3197,18 @@ pg_stat_get_wal_senders(PG_FUNCTION_ARGS)
 		TimeOffset	flushLag;
 		TimeOffset	applyLag;
 		int			priority;
+		int			pid;
 		WalSndState state;
 		Datum		values[PG_STAT_GET_WAL_SENDERS_COLS];
 		bool		nulls[PG_STAT_GET_WAL_SENDERS_COLS];
 
+		SpinLockAcquire(&walsnd->mutex);
 		if (walsnd->pid == 0)
+		{
+			SpinLockRelease(&walsnd->mutex);
 			continue;
-
-		SpinLockAcquire(&walsnd->mutex);
+		}
+		pid = walsnd->pid;
 		sentPtr = walsnd->sentPtr;
 		state = walsnd->state;
 		write = walsnd->write;
@@ -3215,7 +3221,7 @@ pg_stat_get_wal_senders(PG_FUNCTION_ARGS)
 		SpinLockRelease(&walsnd->mutex);
 
 		memset(nulls, 0, sizeof(nulls));
-		values[0] = Int32GetDatum(walsnd->pid);
+		values[0] = Int32GetDatum(pid);
 
 		if (!superuser())
 		{
diff --git a/src/include/replication/walreceiver.h b/src/include/replication/walreceiver.h
index 31d090c99d..44199670b1 100644
--- a/src/include/replication/walreceiver.h
+++ b/src/include/replication/walreceiver.h
@@ -114,6 +114,9 @@ typedef struct
 	 */
 	char		slotname[NAMEDATALEN];
 
+	/* set true once conninfo is ready to display (obfuscated pwds etc) */
+	bool		ready_to_display;
+
 	slock_t		mutex;			/* locks shared variables shown above */
 
 	/*
@@ -122,9 +125,6 @@ typedef struct
 	 */
 	bool		force_reply;
 
-	/* set true once conninfo is ready to display (obfuscated pwds etc) */
-	bool		ready_to_display;
-
 	/*
 	 * Latch used by startup process to wake up walreceiver after telling it
 	 * where to start streaming (after setting receiveStart and


^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-06-13 02:16  Michael Paquier <[email protected]>
  parent: Michael Paquier <[email protected]>
  2 siblings, 1 reply; 160+ messages in thread

From: Michael Paquier @ 2017-06-13 02:16 UTC (permalink / raw)
  To: Robert Haas <[email protected]>; +Cc: Masahiko Sawada <[email protected]>; Thomas Munro <[email protected]>; pgsql-hackers

On Thu, Jun 8, 2017 at 1:15 PM, Michael Paquier
<[email protected]> wrote:
> 3) In walreceiver.c's pg_stat_get_wal_receiver's:
> - Launch pg_stat_get_wal_receiver and take a checkpoint on it.
> - Pass the lookups of pid and ready_to_display.
> - Make the WAL receiver stop.
> - The view reports inconsistent data. If the WAL receiver data was
> just initialized, the caller would get back PID values or similar 0.
> Particularly a WAL receiver marked with !ready_to_display could show
> data to the caller. That's not cool.

This can actually leak password information to any user who is a
member of the group DEFAULT_ROLE_READ_ALL_STATS, though the window to
put hands on this password information is very small, it can be
possible if the WAL receiver gets down and restarted for a reason or
another during a maintenance window for example:
1) The user queries pg_stat_wal_receiver, for example take a
breakpoint just after the check on walrcv->ready_to_display.
2) Restart the primary once, forcing the standby to spawn a new WAL receiver.
3) Breakpoint on the WAL receiver process, with something like that to help:
--- a/src/backend/replication/walreceiver.c
+++ b/src/backend/replication/walreceiver.c
@@ -243,6 +243,7 @@ WalReceiverMain(void)

    /* Fetch information required to start streaming */
    walrcv->ready_to_display = false;
+   pg_usleep(10000000L); /* 10s */
    strlcpy(conninfo, (char *) walrcv->conninfo, MAXCONNINFO);
    strlcpy(slotname, (char *) walrcv->slotname, NAMEDATALEN);
    startpoint = walrcv->receiveStart;
4) continue the session querying pg_stat_wal_receiver, at this stage
it has information with the WAL receiver data set as
!ready_to_display. If the connection string includes a password, this
becomes visible as well.

If queried at high frequency, pg_stat_wal_receiver may show up some
information. Postgres 9.6 includes this leak as well, but there is no real
leak non-superusers will just see NULL fields for this view. In Postgres
10 though, any member of this group for statistics can see leaking
information. Based on that, this is an open item, so I have added it back
now to the list, moved from the section for older bugs.
-- 
Michael


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-06-15 01:59  Noah Misch <[email protected]>
  parent: Michael Paquier <[email protected]>
  0 siblings, 1 reply; 160+ messages in thread

From: Noah Misch @ 2017-06-15 01:59 UTC (permalink / raw)
  To: [email protected]; +Cc: Michael Paquier <[email protected]>; Robert Haas <[email protected]>; Masahiko Sawada <[email protected]>; Thomas Munro <[email protected]>; pgsql-hackers

On Tue, Jun 13, 2017 at 11:16:54AM +0900, Michael Paquier wrote:
> On Thu, Jun 8, 2017 at 1:15 PM, Michael Paquier
> <[email protected]> wrote:
> > 3) In walreceiver.c's pg_stat_get_wal_receiver's:
> > - Launch pg_stat_get_wal_receiver and take a checkpoint on it.
> > - Pass the lookups of pid and ready_to_display.
> > - Make the WAL receiver stop.
> > - The view reports inconsistent data. If the WAL receiver data was
> > just initialized, the caller would get back PID values or similar 0.
> > Particularly a WAL receiver marked with !ready_to_display could show
> > data to the caller. That's not cool.
> 
> This can actually leak password information to any user who is a
> member of the group DEFAULT_ROLE_READ_ALL_STATS, though the window to
> put hands on this password information is very small, it can be
> possible if the WAL receiver gets down and restarted for a reason or
> another during a maintenance window for example:
> 1) The user queries pg_stat_wal_receiver, for example take a
> breakpoint just after the check on walrcv->ready_to_display.
> 2) Restart the primary once, forcing the standby to spawn a new WAL receiver.
> 3) Breakpoint on the WAL receiver process, with something like that to help:
> --- a/src/backend/replication/walreceiver.c
> +++ b/src/backend/replication/walreceiver.c
> @@ -243,6 +243,7 @@ WalReceiverMain(void)
> 
>     /* Fetch information required to start streaming */
>     walrcv->ready_to_display = false;
> +   pg_usleep(10000000L); /* 10s */
>     strlcpy(conninfo, (char *) walrcv->conninfo, MAXCONNINFO);
>     strlcpy(slotname, (char *) walrcv->slotname, NAMEDATALEN);
>     startpoint = walrcv->receiveStart;
> 4) continue the session querying pg_stat_wal_receiver, at this stage
> it has information with the WAL receiver data set as
> !ready_to_display. If the connection string includes a password, this
> becomes visible as well.
> 
> If queried at high frequency, pg_stat_wal_receiver may show up some
> information. Postgres 9.6 includes this leak as well, but there is no real
> leak non-superusers will just see NULL fields for this view. In Postgres
> 10 though, any member of this group for statistics can see leaking
> information. Based on that, this is an open item, so I have added it back
> now to the list, moved from the section for older bugs.

Formally, the causative commit is the one that removed the superuser() test,
namely 25fff40.

[Action required within three days.  This is a generic notification.]

The above-described topic is currently a PostgreSQL 10 open item.  Simon,
since you committed the patch believed to have created it, you own this open
item.  If some other commit is more relevant or if this does not belong as a
v10 open item, please let us know.  Otherwise, please observe the policy on
open item ownership[1] and send a status update within three calendar days of
this message.  Include a date for your subsequent status update.  Testers may
discover new open items at any time, and I want to plan to get them all fixed
well in advance of shipping v10.  Consequently, I will appreciate your efforts
toward speedy resolution.  Thanks.

[1] https://www.postgresql.org/message-id/20170404140717.GA2675809%40tornado.leadboat.com


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-06-15 07:51  Kyotaro HORIGUCHI <[email protected]>
  parent: Michael Paquier <[email protected]>
  2 siblings, 0 replies; 160+ messages in thread

From: Kyotaro HORIGUCHI @ 2017-06-15 07:51 UTC (permalink / raw)
  To: [email protected]; +Cc: [email protected]; [email protected]; [email protected]; pgsql-hackers

Hi,

At Thu, 8 Jun 2017 13:15:02 +0900, Michael Paquier <[email protected]> wrote in <CAB7nPqTfu2P1QCZB22LJG+P4ET-pS0yW+qSTWO9JcVcCtaPZAA@mail.gmail.com>
> On Thu, Jun 8, 2017 at 3:31 AM, Robert Haas <[email protected]> wrote:
> > I think if you're going to fix it so that we take spinlocks on
> > MyWalSnd in a bunch of places that we didn't take them before, it
> > would make sense to fix all the places where we're accessing those
> > fields without a spinlock instead of only some of them, unless there
> > are good reasons why we only need it in some cases and not others, in
> > which case I think the patch should add comments to each place where
> > the lock was not taken explaining why it's OK.  It didn't take me and
> > my copy of vi very long to find instances that you didn't change.
> 
> I take that you are referring to the two lookups in
> WalSndWaitForWal(), one in exec_replication_command(), one in
> WalSndLoop() and one in WalSndDone(). First let's remember that all
> the fields protected by the spinlock are only updated in a WAL sender
> process, so reading them directly is safe in this context: we know
> that no other processes will write them. And all those functions are
> called only in the context of a WAL sender process. So the copies of
> MyWalSnd that you are referring to here don't need a spinlock. Is
> there something I am missing?
> 
> Actually, perhaps it would make sense to add some Assert(am_walsender)
> in this code?

It is not obvious that reading MyWalSnd without spinlock is safe
without knowing that it is modified only by itself. Based on that
I see two ways to go.

1. Put spinlock there (WalSndWaitForWal and..) even though it is
   not required. It ensures safety even if it is modified by
   other processes for possible additional conflict with other
   readers. (Unless someone careless forgets lock for new code).

2. Put a comment there instead (maybe assertion is not verbose
   enough), containing something like the mention above.


By the way, this this discussion make me think of
SyncRepInitConfig. Taking SyncRepLock seems unnecessary there to
me. 

> > I also think that should provide some analysis regarding the question
> > Thomas asked - are these actually bugs, or are they OK for some
> > reason?  We shouldn't just plaster the code with spinlock acquisitions
> > unless it's really necessary.  It seems at least possible that these
> > changes could cause performance regressions, and it doesn't look like
> > you've done any testing or provided any analysis indicating whether
> > that's likely to happen or not.
> 
> Now taking a step back on the patch, v3 that I sent 3 weeks ago.
> Additional spinlocks are taken in:
> 1) SyncRepReleaseWaiters(), which is called in a WAL sender. So indeed
> no spinlocks are needed here. The last patch is incorrect here.

Agreed that it is not necessary, but we might should put one
depending on the conclusion of the discussion.

> 2) SyncRepGetSyncStandbysPriority() and SyncRepGetSyncStandbysQuorum()
> as of HEAD via SyncRepGetSyncStandbys(), something that's not good if
> done lock-less as this gets used for pg_stat_replication for
> pg_stat_get_wal_senders(). This can return a garbage list of sync
> standbys to the client as WAL senders update their flush, write and
> apply positions in parallel to that, and don't care about SyncRepLock
> as this gets calculated with the standby feedback messages. The
> problem here is the gap between walsnd->sync_standby_priority and
> walsnd->flush:
> -- With a primary, and more or more sync standbys, take a checkpoint
> in SyncRepGetSyncStandbysPriority() after the first check is passed.
> -- Shutdown the standby and restart it.
> -- The standby reconnects, initializes the WAL sender slot with
> InitWalSenderSlot().
> -- Let the process of SyncRepGetSyncStandbysPriority() continue, the
> sync standby does not show up.
> That would be really unlikely to happen, but the code is written in
> such a way that it could happen. One could also say that this gives a
> closer idea that the standby disconnected but it looks wrong to me to
> do this value lookup without a required lock.

It doesn't seem to be a problem for
pg_stat_get_wal_senders(). For SyncRepReleaseWaiters, also it
doesn't seem to be a problem because next reply message runs the
process again.

> 3) In walreceiver.c's pg_stat_get_wal_receiver's:
> - Launch pg_stat_get_wal_receiver and take a checkpoint on it.
> - Pass the lookups of pid and ready_to_display.
> - Make the WAL receiver stop.
> - The view reports inconsistent data. If the WAL receiver data was
> just initialized, the caller would get back PID values or similar 0.
> Particularly a WAL receiver marked with !ready_to_display could show
> data to the caller. That's not cool.

Totally agreed.

> 4) KeepFileRestoredFromArchive() needs a lock, as this is called from
> the startup process. That's harmless, the worst that can happen is
> that needreload is switched to true for a process that has been marked
> with a PID of 0 because of a WAL sender restart (slot taken again and
> initialized). But that will just process an extra reload in a worst
> case.

Agreed to the disucussion but it doesn't seem to me to be
necessary, however, seems better to have.

By the way InitWalSenderSlot seems fogetting to initialize
needreload and sync_standby_priority. The initialization code
would be better to be arranged in the same order with its
defintion. Initializing needreload is quite natural from its
definition. (Still it is possible that an extra reload happens if
a reload is requested just after the initialization...)

> > Basically, I don't think this patch is ready to commit.  We have
> > neither evidence that it is necessary nor evidence that it is
> > complete.  I don't want to be unfair, here, but it seems to me you
> > ought to do a little more legwork before throwing this over the wall
> > to the pool of committers.
> 
> Fair enough. Is the analysis written above more convincing?

regards,

-- 
Kyotaro Horiguchi
NTT Open Source Software Center



-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-06-15 08:04  Simon Riggs <[email protected]>
  parent: Noah Misch <[email protected]>
  0 siblings, 1 reply; 160+ messages in thread

From: Simon Riggs @ 2017-06-15 08:04 UTC (permalink / raw)
  To: Noah Misch <[email protected]>; +Cc: Michael Paquier <[email protected]>; Robert Haas <[email protected]>; Masahiko Sawada <[email protected]>; Thomas Munro <[email protected]>; pgsql-hackers

On 15 June 2017 at 02:59, Noah Misch <[email protected]> wrote:

> Formally, the causative commit is the one that removed the superuser() test,
> namely 25fff40.
>
> [Action required within three days.  This is a generic notification.]
>
> The above-described topic is currently a PostgreSQL 10 open item.  Simon,
> since you committed the patch believed to have created it, you own this open
> item.  If some other commit is more relevant or if this does not belong as a
> v10 open item, please let us know.  Otherwise, please observe the policy on
> open item ownership[1] and send a status update within three calendar days of
> this message.  Include a date for your subsequent status update.  Testers may
> discover new open items at any time, and I want to plan to get them all fixed
> well in advance of shipping v10.  Consequently, I will appreciate your efforts
> toward speedy resolution.  Thanks.
>
> [1] https://www.postgresql.org/message-id/20170404140717.GA2675809%40tornado.leadboat.com

Thanks for letting me know. I'm on leave at present, so won't be able
to get to this until June 20, though will make it a priority for then.

-- 
Simon Riggs                http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-06-22 05:52  Noah Misch <[email protected]>
  parent: Simon Riggs <[email protected]>
  0 siblings, 1 reply; 160+ messages in thread

From: Noah Misch @ 2017-06-22 05:52 UTC (permalink / raw)
  To: Simon Riggs <[email protected]>; +Cc: Michael Paquier <[email protected]>; Robert Haas <[email protected]>; Masahiko Sawada <[email protected]>; Thomas Munro <[email protected]>; pgsql-hackers

On Thu, Jun 15, 2017 at 09:04:44AM +0100, Simon Riggs wrote:
> On 15 June 2017 at 02:59, Noah Misch <[email protected]> wrote:
> 
> > Formally, the causative commit is the one that removed the superuser() test,
> > namely 25fff40.
> >
> > [Action required within three days.  This is a generic notification.]
> >
> > The above-described topic is currently a PostgreSQL 10 open item.  Simon,
> > since you committed the patch believed to have created it, you own this open
> > item.  If some other commit is more relevant or if this does not belong as a
> > v10 open item, please let us know.  Otherwise, please observe the policy on
> > open item ownership[1] and send a status update within three calendar days of
> > this message.  Include a date for your subsequent status update.  Testers may
> > discover new open items at any time, and I want to plan to get them all fixed
> > well in advance of shipping v10.  Consequently, I will appreciate your efforts
> > toward speedy resolution.  Thanks.
> >
> > [1] https://www.postgresql.org/message-id/20170404140717.GA2675809%40tornado.leadboat.com
> 
> Thanks for letting me know. I'm on leave at present, so won't be able
> to get to this until June 20, though will make it a priority for then.

This PostgreSQL 10 open item is past due for your status update.  Kindly send
a status update within 24 hours, and include a date for your subsequent status
update.  Refer to the policy on open item ownership:
https://www.postgresql.org/message-id/20170404140717.GA2675809%40tornado.leadboat.com


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-06-24 05:21  Noah Misch <[email protected]>
  parent: Noah Misch <[email protected]>
  0 siblings, 1 reply; 160+ messages in thread

From: Noah Misch @ 2017-06-24 05:21 UTC (permalink / raw)
  To: Simon Riggs <[email protected]>; +Cc: Michael Paquier <[email protected]>; Robert Haas <[email protected]>; Masahiko Sawada <[email protected]>; Thomas Munro <[email protected]>; pgsql-hackers

On Wed, Jun 21, 2017 at 10:52:50PM -0700, Noah Misch wrote:
> On Thu, Jun 15, 2017 at 09:04:44AM +0100, Simon Riggs wrote:
> > On 15 June 2017 at 02:59, Noah Misch <[email protected]> wrote:
> > 
> > > Formally, the causative commit is the one that removed the superuser() test,
> > > namely 25fff40.
> > >
> > > [Action required within three days.  This is a generic notification.]
> > >
> > > The above-described topic is currently a PostgreSQL 10 open item.  Simon,
> > > since you committed the patch believed to have created it, you own this open
> > > item.  If some other commit is more relevant or if this does not belong as a
> > > v10 open item, please let us know.  Otherwise, please observe the policy on
> > > open item ownership[1] and send a status update within three calendar days of
> > > this message.  Include a date for your subsequent status update.  Testers may
> > > discover new open items at any time, and I want to plan to get them all fixed
> > > well in advance of shipping v10.  Consequently, I will appreciate your efforts
> > > toward speedy resolution.  Thanks.
> > >
> > > [1] https://www.postgresql.org/message-id/20170404140717.GA2675809%40tornado.leadboat.com
> > 
> > Thanks for letting me know. I'm on leave at present, so won't be able
> > to get to this until June 20, though will make it a priority for then.
> 
> This PostgreSQL 10 open item is past due for your status update.  Kindly send
> a status update within 24 hours, and include a date for your subsequent status
> update.  Refer to the policy on open item ownership:
> https://www.postgresql.org/message-id/20170404140717.GA2675809%40tornado.leadboat.com

IMMEDIATE ATTENTION REQUIRED.  This PostgreSQL 10 open item is long past due
for your status update.  Please reacquaint yourself with the policy on open
item ownership[1] and then reply immediately.  If I do not hear from you by
2017-06-25 06:00 UTC, I will transfer this item to release management team
ownership without further notice.

[1] https://www.postgresql.org/message-id/20170404140717.GA2675809%40tornado.leadboat.com


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-06-25 05:11  Alvaro Herrera <[email protected]>
  parent: Noah Misch <[email protected]>
  0 siblings, 1 reply; 160+ messages in thread

From: Alvaro Herrera @ 2017-06-25 05:11 UTC (permalink / raw)
  To: Noah Misch <[email protected]>; +Cc: Simon Riggs <[email protected]>; Michael Paquier <[email protected]>; Robert Haas <[email protected]>; Masahiko Sawada <[email protected]>; Thomas Munro <[email protected]>; pgsql-hackers

Noah Misch wrote:

> IMMEDIATE ATTENTION REQUIRED.  This PostgreSQL 10 open item is long past due
> for your status update.  Please reacquaint yourself with the policy on open
> item ownership[1] and then reply immediately.  If I do not hear from you by
> 2017-06-25 06:00 UTC, I will transfer this item to release management team
> ownership without further notice.

I volunteer to own this item.  Next update before Wednesday 28th 19:00 CLT.

-- 
Álvaro Herrera                https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-06-25 08:38  Michael Paquier <[email protected]>
  parent: Alvaro Herrera <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Michael Paquier @ 2017-06-25 08:38 UTC (permalink / raw)
  To: Alvaro Herrera <[email protected]>; +Cc: Noah Misch <[email protected]>; Simon Riggs <[email protected]>; Robert Haas <[email protected]>; Masahiko Sawada <[email protected]>; Thomas Munro <[email protected]>; pgsql-hackers

On Sun, Jun 25, 2017 at 2:11 PM, Alvaro Herrera
<[email protected]> wrote:
> Noah Misch wrote:
>
>> IMMEDIATE ATTENTION REQUIRED.  This PostgreSQL 10 open item is long past due
>> for your status update.  Please reacquaint yourself with the policy on open
>> item ownership[1] and then reply immediately.  If I do not hear from you by
>> 2017-06-25 06:00 UTC, I will transfer this item to release management team
>> ownership without further notice.
>
> I volunteer to own this item.  Next update before Wednesday 28th 19:00 CLT.

Thanks Álvaro.
-- 
Michael


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-06-28 22:52  Alvaro Herrera <[email protected]>
  parent: Michael Paquier <[email protected]>
  2 siblings, 2 replies; 160+ messages in thread

From: Alvaro Herrera @ 2017-06-28 22:52 UTC (permalink / raw)
  To: Michael Paquier <[email protected]>; +Cc: Robert Haas <[email protected]>; Masahiko Sawada <[email protected]>; Thomas Munro <[email protected]>; pgsql-hackers

Michael Paquier wrote:

> On Thu, Jun 8, 2017 at 3:31 AM, Robert Haas <[email protected]> wrote:
> > I think if you're going to fix it so that we take spinlocks on
> > MyWalSnd in a bunch of places that we didn't take them before, it
> > would make sense to fix all the places where we're accessing those
> > fields without a spinlock instead of only some of them, unless there
> > are good reasons why we only need it in some cases and not others, in
> > which case I think the patch should add comments to each place where
> > the lock was not taken explaining why it's OK.  It didn't take me and
> > my copy of vi very long to find instances that you didn't change.
> 
> I take that you are referring to the two lookups in
> WalSndWaitForWal(), one in exec_replication_command(), one in
> WalSndLoop() and one in WalSndDone(). First let's remember that all
> the fields protected by the spinlock are only updated in a WAL sender
> process, so reading them directly is safe in this context: we know
> that no other processes will write them. And all those functions are
> called only in the context of a WAL sender process. So the copies of
> MyWalSnd that you are referring to here don't need a spinlock. Is
> there something I am missing?

I assume you mean "reading them unlocked" when you write "reading them
directly".  If so, I agree with that rationale; I verified that it
applies to members state, sentPtr, write, sent, flush, writeLag,
sentLag, flushLag.  I wonder if there's a maintainability danger: as
soon as we add a write to those members in a process other than the
walsender itself, we have a bug.  I don't yet have an opinion on the
severity of that problem.

Other struct members such as needreload are written by other processes,
so they should be always accessed with mutex held.  Regarding pid, it
seems easiest if we use the rule that it must also be always accessed
with spinlock held.  I propose updating the comment on WalSnd to this:

/*
 * Each walsender has a WalSnd struct in shared memory.
 *
 * This struct is protected by 'mutex', with two exceptions; one is
 * sync_standby_priority as noted below.  The other exception is that some
 * members are only written by the walsender process itself, and thus that
 * process is free to read those members without holding spinlock.  pid and
 * needreload always require the spinlock to be held for all accesses.
 */


> Actually, perhaps it would make sense to add some Assert(am_walsender)
> in this code?

I don't think that's needed, since MyWalSnd will only be valid in a
walsender.

I think I'm done with the walsender half of this patch; I still need to
review the walreceiver part.  I will report back tomorrow 19:00 CLT.

Currently, I'm considering not to backpatch any of this.

-- 
Álvaro Herrera                https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-06-29 00:34  Michael Paquier <[email protected]>
  parent: Alvaro Herrera <[email protected]>
  1 sibling, 1 reply; 160+ messages in thread

From: Michael Paquier @ 2017-06-29 00:34 UTC (permalink / raw)
  To: Alvaro Herrera <[email protected]>; +Cc: Robert Haas <[email protected]>; Masahiko Sawada <[email protected]>; Thomas Munro <[email protected]>; pgsql-hackers

On Thu, Jun 29, 2017 at 7:52 AM, Alvaro Herrera
<[email protected]> wrote:
> Michael Paquier wrote:
>> On Thu, Jun 8, 2017 at 3:31 AM, Robert Haas <[email protected]> wrote:
>> I take that you are referring to the two lookups in
>> WalSndWaitForWal(), one in exec_replication_command(), one in
>> WalSndLoop() and one in WalSndDone(). First let's remember that all
>> the fields protected by the spinlock are only updated in a WAL sender
>> process, so reading them directly is safe in this context: we know
>> that no other processes will write them. And all those functions are
>> called only in the context of a WAL sender process. So the copies of
>> MyWalSnd that you are referring to here don't need a spinlock. Is
>> there something I am missing?
>
> I assume you mean "reading them unlocked" when you write "reading them
> directly".

Both expressions sound the same to me. So yes I meant that.

> If so, I agree with that rationale; I verified that it
> applies to members state, sentPtr, write, sent, flush, writeLag,
> sentLag, flushLag.  I wonder if there's a maintainability danger: as
> soon as we add a write to those members in a process other than the
> walsender itself, we have a bug.  I don't yet have an opinion on the
> severity of that problem.
>
> Other struct members such as needreload are written by other processes,
> so they should be always accessed with mutex held.  Regarding pid, it
> seems easiest if we use the rule that it must also be always accessed
> with spinlock held.  I propose updating the comment on WalSnd to this:
>
> /*
>  * Each walsender has a WalSnd struct in shared memory.
>  *
>  * This struct is protected by 'mutex', with two exceptions; one is
>  * sync_standby_priority as noted below.  The other exception is that some
>  * members are only written by the walsender process itself, and thus that
>  * process is free to read those members without holding spinlock.  pid and
>  * needreload always require the spinlock to be held for all accesses.
>  */

Agreed, a comment seems appropriate to me. That's in line with what
Horiguchi-san has mentioned upthread.

> I think I'm done with the walsender half of this patch; I still need to
> review the walreceiver part.  I will report back tomorrow 19:00 CLT.

Thanks!

> Currently, I'm considering not to backpatch any of this.

Considering how crazy the conditions to make the information fetched
by users inconsistent are met, I agree with that.
-- 
Michael


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-06-30 03:43  Noah Misch <[email protected]>
  parent: Alvaro Herrera <[email protected]>
  1 sibling, 1 reply; 160+ messages in thread

From: Noah Misch @ 2017-06-30 03:43 UTC (permalink / raw)
  To: Alvaro Herrera <[email protected]>; +Cc: Michael Paquier <[email protected]>; Robert Haas <[email protected]>; Masahiko Sawada <[email protected]>; Thomas Munro <[email protected]>; pgsql-hackers

On Wed, Jun 28, 2017 at 06:52:14PM -0400, Alvaro Herrera wrote:
> I think I'm done with the walsender half of this patch; I still need to
> review the walreceiver part.  I will report back tomorrow 19:00 CLT.

This PostgreSQL 10 open item is past due for your status update.  Kindly send
a status update within 24 hours, and include a date for your subsequent status
update.  Refer to the policy on open item ownership:
https://www.postgresql.org/message-id/20170404140717.GA2675809%40tornado.leadboat.com


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-06-30 04:52  Alvaro Herrera <[email protected]>
  parent: Noah Misch <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Alvaro Herrera @ 2017-06-30 04:52 UTC (permalink / raw)
  To: Noah Misch <[email protected]>; +Cc: Michael Paquier <[email protected]>; Robert Haas <[email protected]>; Masahiko Sawada <[email protected]>; Thomas Munro <[email protected]>; pgsql-hackers

Noah Misch wrote:
> On Wed, Jun 28, 2017 at 06:52:14PM -0400, Alvaro Herrera wrote:
> > I think I'm done with the walsender half of this patch; I still need to
> > review the walreceiver part.  I will report back tomorrow 19:00 CLT.
> 
> This PostgreSQL 10 open item is past due for your status update.  Kindly send
> a status update within 24 hours, and include a date for your subsequent status
> update.  Refer to the policy on open item ownership:
> https://www.postgresql.org/message-id/20170404140717.GA2675809%40tornado.leadboat.com

Real life took over.  I'll report before tomorrow same time.

-- 
Álvaro Herrera                https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: Race conditions with WAL sender PID lookups
@ 2017-06-30 22:20  Alvaro Herrera <[email protected]>
  parent: Michael Paquier <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Alvaro Herrera @ 2017-06-30 22:20 UTC (permalink / raw)
  To: Michael Paquier <[email protected]>; +Cc: Robert Haas <[email protected]>; Masahiko Sawada <[email protected]>; Thomas Munro <[email protected]>; pgsql-hackers

Michael Paquier wrote:
> On Thu, Jun 29, 2017 at 7:52 AM, Alvaro Herrera
> <[email protected]> wrote:

> > I think I'm done with the walsender half of this patch; I still need to
> > review the walreceiver part.  I will report back tomorrow 19:00 CLT.
> 
> Thanks!
> 
> > Currently, I'm considering not to backpatch any of this.
> 
> Considering how crazy the conditions to make the information fetched
> by users inconsistent are met, I agree with that.

Pushed.

-- 
Álvaro Herrera                https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services


-- 
Sent via pgsql-hackers mailing list ([email protected])
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--XsQoSWH+UP9D9v3l
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--jozmn01XJZjDjM3N
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH v1 2/2] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 4e6a3df6b8..85b1c21ee0 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -1996,7 +2031,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 377afb8732..1a503c1447 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1159,7 +1159,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index f49a57b35e..536a5e653b 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.27.0


----Next_Part(Mon_Mar__8_17_32_42_2021_572)----





^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 2/8] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e04250f4e9..e3128564e1 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1161,7 +1161,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--f0KYrhQ4vYSV2aJu
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0003-Allow-alternate-compression-methods-for-wal_compress.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 04192b7add..0183589b4d 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1171,7 +1171,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--0qVF/w3MHQqLSynd
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--XsQoSWH+UP9D9v3l
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--jozmn01XJZjDjM3N
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH v1 2/2] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 4e6a3df6b8..85b1c21ee0 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -1996,7 +2031,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 377afb8732..1a503c1447 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1159,7 +1159,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index f49a57b35e..536a5e653b 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.27.0


----Next_Part(Mon_Mar__8_17_32_42_2021_572)----





^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 2/8] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e04250f4e9..e3128564e1 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1161,7 +1161,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--f0KYrhQ4vYSV2aJu
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0003-Allow-alternate-compression-methods-for-wal_compress.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 04192b7add..0183589b4d 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1171,7 +1171,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--0qVF/w3MHQqLSynd
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--XsQoSWH+UP9D9v3l
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--jozmn01XJZjDjM3N
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH v1 2/2] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 4e6a3df6b8..85b1c21ee0 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -1996,7 +2031,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 377afb8732..1a503c1447 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1159,7 +1159,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index f49a57b35e..536a5e653b 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.27.0


----Next_Part(Mon_Mar__8_17_32_42_2021_572)----





^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 2/8] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e04250f4e9..e3128564e1 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1161,7 +1161,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--f0KYrhQ4vYSV2aJu
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0003-Allow-alternate-compression-methods-for-wal_compress.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 04192b7add..0183589b4d 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1171,7 +1171,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--0qVF/w3MHQqLSynd
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--XsQoSWH+UP9D9v3l
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--jozmn01XJZjDjM3N
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH v1 2/2] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 4e6a3df6b8..85b1c21ee0 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -1996,7 +2031,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 377afb8732..1a503c1447 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1159,7 +1159,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index f49a57b35e..536a5e653b 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.27.0


----Next_Part(Mon_Mar__8_17_32_42_2021_572)----





^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 2/8] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e04250f4e9..e3128564e1 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1161,7 +1161,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--f0KYrhQ4vYSV2aJu
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0003-Allow-alternate-compression-methods-for-wal_compress.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 04192b7add..0183589b4d 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1171,7 +1171,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--0qVF/w3MHQqLSynd
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--XsQoSWH+UP9D9v3l
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--jozmn01XJZjDjM3N
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH v1 2/2] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 4e6a3df6b8..85b1c21ee0 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -1996,7 +2031,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 377afb8732..1a503c1447 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1159,7 +1159,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index f49a57b35e..536a5e653b 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.27.0


----Next_Part(Mon_Mar__8_17_32_42_2021_572)----





^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 2/8] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e04250f4e9..e3128564e1 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1161,7 +1161,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--f0KYrhQ4vYSV2aJu
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0003-Allow-alternate-compression-methods-for-wal_compress.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 04192b7add..0183589b4d 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1171,7 +1171,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--0qVF/w3MHQqLSynd
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--XsQoSWH+UP9D9v3l
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--jozmn01XJZjDjM3N
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH v1 2/2] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 4e6a3df6b8..85b1c21ee0 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -1996,7 +2031,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 377afb8732..1a503c1447 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1159,7 +1159,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index f49a57b35e..536a5e653b 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.27.0


----Next_Part(Mon_Mar__8_17_32_42_2021_572)----





^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 2/8] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e04250f4e9..e3128564e1 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1161,7 +1161,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--f0KYrhQ4vYSV2aJu
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0003-Allow-alternate-compression-methods-for-wal_compress.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 04192b7add..0183589b4d 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1171,7 +1171,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--0qVF/w3MHQqLSynd
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--XsQoSWH+UP9D9v3l
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--jozmn01XJZjDjM3N
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH v1 2/2] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 4e6a3df6b8..85b1c21ee0 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -1996,7 +2031,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 377afb8732..1a503c1447 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1159,7 +1159,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index f49a57b35e..536a5e653b 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.27.0


----Next_Part(Mon_Mar__8_17_32_42_2021_572)----





^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 2/8] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e04250f4e9..e3128564e1 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1161,7 +1161,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--f0KYrhQ4vYSV2aJu
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0003-Allow-alternate-compression-methods-for-wal_compress.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 04192b7add..0183589b4d 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1171,7 +1171,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--0qVF/w3MHQqLSynd
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--XsQoSWH+UP9D9v3l
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--jozmn01XJZjDjM3N
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH v1 2/2] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 4e6a3df6b8..85b1c21ee0 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -1996,7 +2031,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 377afb8732..1a503c1447 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1159,7 +1159,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index f49a57b35e..536a5e653b 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.27.0


----Next_Part(Mon_Mar__8_17_32_42_2021_572)----





^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 2/8] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e04250f4e9..e3128564e1 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1161,7 +1161,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--f0KYrhQ4vYSV2aJu
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0003-Allow-alternate-compression-methods-for-wal_compress.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 04192b7add..0183589b4d 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1171,7 +1171,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--0qVF/w3MHQqLSynd
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--XsQoSWH+UP9D9v3l
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--jozmn01XJZjDjM3N
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH v1 2/2] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 4e6a3df6b8..85b1c21ee0 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -1996,7 +2031,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 377afb8732..1a503c1447 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1159,7 +1159,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index f49a57b35e..536a5e653b 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.27.0


----Next_Part(Mon_Mar__8_17_32_42_2021_572)----





^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 2/8] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e04250f4e9..e3128564e1 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1161,7 +1161,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--f0KYrhQ4vYSV2aJu
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0003-Allow-alternate-compression-methods-for-wal_compress.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 04192b7add..0183589b4d 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1171,7 +1171,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--0qVF/w3MHQqLSynd
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--XsQoSWH+UP9D9v3l
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--jozmn01XJZjDjM3N
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH v1 2/2] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 4e6a3df6b8..85b1c21ee0 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -1996,7 +2031,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 377afb8732..1a503c1447 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1159,7 +1159,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index f49a57b35e..536a5e653b 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.27.0


----Next_Part(Mon_Mar__8_17_32_42_2021_572)----





^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 2/8] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e04250f4e9..e3128564e1 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1161,7 +1161,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--f0KYrhQ4vYSV2aJu
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0003-Allow-alternate-compression-methods-for-wal_compress.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 04192b7add..0183589b4d 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1171,7 +1171,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--0qVF/w3MHQqLSynd
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--XsQoSWH+UP9D9v3l
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--jozmn01XJZjDjM3N
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH v1 2/2] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 4e6a3df6b8..85b1c21ee0 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -1996,7 +2031,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 377afb8732..1a503c1447 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1159,7 +1159,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index f49a57b35e..536a5e653b 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.27.0


----Next_Part(Mon_Mar__8_17_32_42_2021_572)----





^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 2/8] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e04250f4e9..e3128564e1 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1161,7 +1161,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--f0KYrhQ4vYSV2aJu
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0003-Allow-alternate-compression-methods-for-wal_compress.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 04192b7add..0183589b4d 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1171,7 +1171,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--0qVF/w3MHQqLSynd
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--XsQoSWH+UP9D9v3l
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--jozmn01XJZjDjM3N
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH v1 2/2] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 4e6a3df6b8..85b1c21ee0 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -1996,7 +2031,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 377afb8732..1a503c1447 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1159,7 +1159,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index f49a57b35e..536a5e653b 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.27.0


----Next_Part(Mon_Mar__8_17_32_42_2021_572)----





^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 2/8] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e04250f4e9..e3128564e1 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1161,7 +1161,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--f0KYrhQ4vYSV2aJu
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0003-Allow-alternate-compression-methods-for-wal_compress.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--XsQoSWH+UP9D9v3l
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--jozmn01XJZjDjM3N
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH v1 2/2] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 4e6a3df6b8..85b1c21ee0 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -1996,7 +2031,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 377afb8732..1a503c1447 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1159,7 +1159,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index f49a57b35e..536a5e653b 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.27.0


----Next_Part(Mon_Mar__8_17_32_42_2021_572)----





^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 2/8] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e04250f4e9..e3128564e1 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1161,7 +1161,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--f0KYrhQ4vYSV2aJu
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0003-Allow-alternate-compression-methods-for-wal_compress.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 04192b7add..0183589b4d 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1171,7 +1171,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--0qVF/w3MHQqLSynd
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--XsQoSWH+UP9D9v3l
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--jozmn01XJZjDjM3N
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH v1 2/2] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 4e6a3df6b8..85b1c21ee0 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -1996,7 +2031,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 377afb8732..1a503c1447 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1159,7 +1159,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index f49a57b35e..536a5e653b 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.27.0


----Next_Part(Mon_Mar__8_17_32_42_2021_572)----





^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 2/8] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e04250f4e9..e3128564e1 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1161,7 +1161,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--f0KYrhQ4vYSV2aJu
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0003-Allow-alternate-compression-methods-for-wal_compress.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 04192b7add..0183589b4d 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1171,7 +1171,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--0qVF/w3MHQqLSynd
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--XsQoSWH+UP9D9v3l
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--jozmn01XJZjDjM3N
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH v1 2/2] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 4e6a3df6b8..85b1c21ee0 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -1996,7 +2031,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 377afb8732..1a503c1447 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1159,7 +1159,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index f49a57b35e..536a5e653b 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.27.0


----Next_Part(Mon_Mar__8_17_32_42_2021_572)----





^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 2/8] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e04250f4e9..e3128564e1 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1161,7 +1161,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--f0KYrhQ4vYSV2aJu
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0003-Allow-alternate-compression-methods-for-wal_compress.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 04192b7add..0183589b4d 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1171,7 +1171,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--0qVF/w3MHQqLSynd
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--XsQoSWH+UP9D9v3l
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--jozmn01XJZjDjM3N
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH v1 2/2] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 4e6a3df6b8..85b1c21ee0 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -1996,7 +2031,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 377afb8732..1a503c1447 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1159,7 +1159,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index f49a57b35e..536a5e653b 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.27.0


----Next_Part(Mon_Mar__8_17_32_42_2021_572)----





^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 2/8] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e04250f4e9..e3128564e1 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1161,7 +1161,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--f0KYrhQ4vYSV2aJu
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0003-Allow-alternate-compression-methods-for-wal_compress.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 04192b7add..0183589b4d 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1171,7 +1171,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--0qVF/w3MHQqLSynd
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--XsQoSWH+UP9D9v3l
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--jozmn01XJZjDjM3N
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH v1 2/2] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 4e6a3df6b8..85b1c21ee0 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -1996,7 +2031,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 377afb8732..1a503c1447 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1159,7 +1159,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index f49a57b35e..536a5e653b 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.27.0


----Next_Part(Mon_Mar__8_17_32_42_2021_572)----





^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 2/8] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e04250f4e9..e3128564e1 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1161,7 +1161,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--f0KYrhQ4vYSV2aJu
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0003-Allow-alternate-compression-methods-for-wal_compress.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 04192b7add..0183589b4d 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1171,7 +1171,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--0qVF/w3MHQqLSynd
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--XsQoSWH+UP9D9v3l
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--jozmn01XJZjDjM3N
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/12] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 5b657ec724..8db3727502 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--45Z9DzgjV8m4Oswq
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH v1 2/2] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 4e6a3df6b8..85b1c21ee0 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -1996,7 +2031,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 377afb8732..1a503c1447 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1159,7 +1159,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index f49a57b35e..536a5e653b 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.27.0


----Next_Part(Mon_Mar__8_17_32_42_2021_572)----





^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 2/8] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e04250f4e9..e3128564e1 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1161,7 +1161,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--f0KYrhQ4vYSV2aJu
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0003-Allow-alternate-compression-methods-for-wal_compress.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 04192b7add..0183589b4d 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1171,7 +1171,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--0qVF/w3MHQqLSynd
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--XsQoSWH+UP9D9v3l
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--jozmn01XJZjDjM3N
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/12] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 5b657ec724..8db3727502 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--45Z9DzgjV8m4Oswq
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 04192b7add..0183589b4d 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1171,7 +1171,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--0qVF/w3MHQqLSynd
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 6/6] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 56 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 60 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index bb1f106946..f0ba9c33ed 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -204,7 +204,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		topXidLogged;	/* for a subxact: is top-level XID logged? */
@@ -523,12 +523,16 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
 /*
@@ -582,6 +586,35 @@ MarkSubxactTopXidLogged(void)
 	CurrentTransactionState->topXidLogged = true;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
+
 /*
  *	GetStableLatestTransactionId
  *
@@ -669,14 +702,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -746,6 +779,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -760,11 +794,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2049,7 +2085,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 198e6906ee..7c7bea768c 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -915,7 +915,7 @@ XLogInsertRecord(XLogRecData *rdata,
 
 	END_CRIT_SECTION();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	/*
 	 * Mark top transaction id is logged (if needed) so that we should not try
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index 60e57876de..47bae59898 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 17a6fa4abd..33d03e1913 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.1


--pAwQNkOnpTn9IO2O--





^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH v9 3/9] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 441445927e..da8a460722 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -1996,7 +2031,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 1ccc51575a..f7909aa1d7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1162,7 +1162,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 134f6862da..593a4140df 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--neYutvxvOLaeuPCA
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="v9-0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH v1 2/2] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 4e6a3df6b8..85b1c21ee0 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -1996,7 +2031,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 377afb8732..1a503c1447 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1159,7 +1159,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index f49a57b35e..536a5e653b 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.27.0


----Next_Part(Mon_Mar__8_17_32_42_2021_572)----





^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 2/8] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e04250f4e9..e3128564e1 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1161,7 +1161,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--f0KYrhQ4vYSV2aJu
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0003-Allow-alternate-compression-methods-for-wal_compress.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 04192b7add..0183589b4d 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1171,7 +1171,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--0qVF/w3MHQqLSynd
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--XsQoSWH+UP9D9v3l
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--jozmn01XJZjDjM3N
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH v7 3/9] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 441445927e..da8a460722 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -1996,7 +2031,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index b7f8e12aea..18dd4ff1eb 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 134f6862da..593a4140df 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--JYK4vJDZwFMowpUq
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="v7-0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH v8 3/9] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 441445927e..da8a460722 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -1996,7 +2031,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 64094e7175..18d8743715 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1162,7 +1162,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 134f6862da..593a4140df 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--P+33d92oIH25kiaB
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="v8-0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH v1 2/2] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 4e6a3df6b8..85b1c21ee0 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -1996,7 +2031,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 377afb8732..1a503c1447 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1159,7 +1159,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index f49a57b35e..536a5e653b 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.27.0


----Next_Part(Mon_Mar__8_17_32_42_2021_572)----





^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 2/8] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e04250f4e9..e3128564e1 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1161,7 +1161,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--f0KYrhQ4vYSV2aJu
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0003-Allow-alternate-compression-methods-for-wal_compress.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 04192b7add..0183589b4d 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1171,7 +1171,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--0qVF/w3MHQqLSynd
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--XsQoSWH+UP9D9v3l
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--jozmn01XJZjDjM3N
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH v1 2/2] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 4e6a3df6b8..85b1c21ee0 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -1996,7 +2031,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 377afb8732..1a503c1447 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1159,7 +1159,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index f49a57b35e..536a5e653b 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.27.0


----Next_Part(Mon_Mar__8_17_32_42_2021_572)----





^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 2/8] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e04250f4e9..e3128564e1 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1161,7 +1161,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--f0KYrhQ4vYSV2aJu
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0003-Allow-alternate-compression-methods-for-wal_compress.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 04192b7add..0183589b4d 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1171,7 +1171,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--0qVF/w3MHQqLSynd
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--XsQoSWH+UP9D9v3l
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--jozmn01XJZjDjM3N
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH v1 2/2] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 4e6a3df6b8..85b1c21ee0 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -1996,7 +2031,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 377afb8732..1a503c1447 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1159,7 +1159,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index f49a57b35e..536a5e653b 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.27.0


----Next_Part(Mon_Mar__8_17_32_42_2021_572)----





^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 2/8] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e04250f4e9..e3128564e1 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1161,7 +1161,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--f0KYrhQ4vYSV2aJu
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0003-Allow-alternate-compression-methods-for-wal_compress.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 04192b7add..0183589b4d 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1171,7 +1171,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--0qVF/w3MHQqLSynd
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--XsQoSWH+UP9D9v3l
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--jozmn01XJZjDjM3N
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH v1 2/2] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 4e6a3df6b8..85b1c21ee0 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -1996,7 +2031,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 377afb8732..1a503c1447 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1159,7 +1159,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index f49a57b35e..536a5e653b 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.27.0


----Next_Part(Mon_Mar__8_17_32_42_2021_572)----





^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 2/8] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e04250f4e9..e3128564e1 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1161,7 +1161,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--f0KYrhQ4vYSV2aJu
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0003-Allow-alternate-compression-methods-for-wal_compress.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 04192b7add..0183589b4d 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1171,7 +1171,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--0qVF/w3MHQqLSynd
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--XsQoSWH+UP9D9v3l
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 15da91a8dd..6eb46ea8a7 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1172,7 +1172,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--jozmn01XJZjDjM3N
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH v1 2/2] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 4e6a3df6b8..85b1c21ee0 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -1996,7 +2031,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 377afb8732..1a503c1447 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1159,7 +1159,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index f49a57b35e..536a5e653b 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.27.0


----Next_Part(Mon_Mar__8_17_32_42_2021_572)----





^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 2/8] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index e04250f4e9..e3128564e1 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1161,7 +1161,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--f0KYrhQ4vYSV2aJu
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0003-Allow-alternate-compression-methods-for-wal_compress.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* [PATCH 03/10] Make sure published XIDs are persistent
@ 2021-03-08 06:43  Kyotaro Horiguchi <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Kyotaro Horiguchi @ 2021-03-08 06:43 UTC (permalink / raw)

pg_xact_status() premises that XIDs obtained by
pg_current_xact_id(_if_assigned)() are persistent beyond a crash. But
XIDs are not guaranteed to go beyond WAL buffers before commit and
thus XIDs may vanish if server crashes before commit. This patch
guarantees the XID shown by the functions to be flushed out to disk.

Copied from: https://www.postgresql.org/message-id/20210308.173242.463790587797836129.horikyota.ntt%40gmail.com
---
 src/backend/access/transam/xact.c | 55 +++++++++++++++++++++++++------
 src/backend/access/transam/xlog.c |  2 +-
 src/backend/utils/adt/xid8funcs.c | 12 ++++++-
 src/include/access/xact.h         |  3 +-
 4 files changed, 59 insertions(+), 13 deletions(-)

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index 6395a9b240..38e978d238 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -201,7 +201,7 @@ typedef struct TransactionStateData
 	int			prevSecContext; /* previous SecurityRestrictionContext */
 	bool		prevXactReadOnly;	/* entry-time xact r/o state */
 	bool		startedInRecovery;	/* did we start in recovery? */
-	bool		didLogXid;		/* has xid been included in WAL record? */
+	XLogRecPtr 	minLSN;			/* LSN needed to reach to record the xid */
 	int			parallelModeLevel;	/* Enter/ExitParallelMode counter */
 	bool		chain;			/* start a new block after this one */
 	bool		assigned;		/* assigned to top-level XID */
@@ -520,14 +520,46 @@ GetCurrentFullTransactionIdIfAny(void)
  *	MarkCurrentTransactionIdLoggedIfAny
  *
  * Remember that the current xid - if it is assigned - now has been wal logged.
+ *
+ * upto is the LSN up to which we need to flush WAL to ensure the current xid
+ * to be persistent. See EnsureCurrentTransactionIdLogged().
  */
 void
-MarkCurrentTransactionIdLoggedIfAny(void)
+MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto)
 {
-	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId))
-		CurrentTransactionState->didLogXid = true;
+	if (FullTransactionIdIsValid(CurrentTransactionState->fullTransactionId) &&
+		XLogRecPtrIsInvalid(CurrentTransactionState->minLSN))
+		CurrentTransactionState->minLSN = upto;
 }
 
+/*
+ *	EnsureCurrentTransactionIdLogged
+ *
+ * Make sure that the current top XID is WAL-logged.
+ */
+void
+EnsureTopTransactionIdLogged(void)
+{
+	/*
+	 * We need at least one WAL record for the current top transaction to be
+	 * flushed out.  Write one if we don't have one yet.
+	 */
+	if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+	{
+		xl_xact_assignment xlrec;
+
+		xlrec.xtop = XidFromFullTransactionId(XactTopFullTransactionId);
+		Assert(TransactionIdIsValid(xlrec.xtop));
+		xlrec.nsubxacts = 0;
+
+		XLogBeginInsert();
+		XLogRegisterData((char *) &xlrec, MinSizeOfXactAssignment);
+		TopTransactionStateData.minLSN =
+			XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+	}
+
+	XLogFlush(TopTransactionStateData.minLSN);
+}
 
 /*
  *	GetStableLatestTransactionId
@@ -616,14 +648,14 @@ AssignTransactionId(TransactionState s)
 	 * When wal_level=logical, guarantee that a subtransaction's xid can only
 	 * be seen in the WAL stream if its toplevel xid has been logged before.
 	 * If necessary we log an xact_assignment record with fewer than
-	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if didLogXid isn't set
+	 * PGPROC_MAX_CACHED_SUBXIDS. Note that it is fine if minLSN isn't set
 	 * for a transaction even though it appears in a WAL record, we just might
 	 * superfluously log something. That can happen when an xid is included
 	 * somewhere inside a wal record, but not in XLogRecord->xl_xid, like in
 	 * xl_standby_locks.
 	 */
 	if (isSubXact && XLogLogicalInfoActive() &&
-		!TopTransactionStateData.didLogXid)
+		XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
 		log_unknown_top = true;
 
 	/*
@@ -693,6 +725,7 @@ AssignTransactionId(TransactionState s)
 			log_unknown_top)
 		{
 			xl_xact_assignment xlrec;
+			XLogRecPtr		   endptr;
 
 			/*
 			 * xtop is always set by now because we recurse up transaction
@@ -707,11 +740,13 @@ AssignTransactionId(TransactionState s)
 			XLogRegisterData((char *) unreportedXids,
 							 nUnreportedXids * sizeof(TransactionId));
 
-			(void) XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
+			endptr = XLogInsert(RM_XACT_ID, XLOG_XACT_ASSIGNMENT);
 
 			nUnreportedXids = 0;
-			/* mark top, not current xact as having been logged */
-			TopTransactionStateData.didLogXid = true;
+
+			/* set minLSN of top, not of current xact if not yet */
+			if (XLogRecPtrIsInvalid(TopTransactionStateData.minLSN))
+				TopTransactionStateData.minLSN = endptr;
 		}
 	}
 }
@@ -2022,7 +2057,7 @@ StartTransaction(void)
 	 * initialize reported xid accounting
 	 */
 	nUnreportedXids = 0;
-	s->didLogXid = false;
+	s->minLSN = InvalidXLogRecPtr;
 
 	/*
 	 * must initialize resource-management stuff first
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 04192b7add..0183589b4d 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -1171,7 +1171,7 @@ XLogInsertRecord(XLogRecData *rdata,
 	 */
 	WALInsertLockRelease();
 
-	MarkCurrentTransactionIdLoggedIfAny();
+	MarkCurrentTransactionIdLoggedIfAny(EndPos);
 
 	END_CRIT_SECTION();
 
diff --git a/src/backend/utils/adt/xid8funcs.c b/src/backend/utils/adt/xid8funcs.c
index cc2b4ac797..992482f8c8 100644
--- a/src/backend/utils/adt/xid8funcs.c
+++ b/src/backend/utils/adt/xid8funcs.c
@@ -357,6 +357,8 @@ bad_format:
 Datum
 pg_current_xact_id(PG_FUNCTION_ARGS)
 {
+	FullTransactionId xid;
+
 	/*
 	 * Must prevent during recovery because if an xid is not assigned we try
 	 * to assign one, which would fail. Programs already rely on this function
@@ -365,7 +367,12 @@ pg_current_xact_id(PG_FUNCTION_ARGS)
 	 */
 	PreventCommandDuringRecovery("pg_current_xact_id()");
 
-	PG_RETURN_FULLTRANSACTIONID(GetTopFullTransactionId());
+	xid = GetTopFullTransactionId();
+
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
+	PG_RETURN_FULLTRANSACTIONID(xid);
 }
 
 /*
@@ -380,6 +387,9 @@ pg_current_xact_id_if_assigned(PG_FUNCTION_ARGS)
 	if (!FullTransactionIdIsValid(topfxid))
 		PG_RETURN_NULL();
 
+	/* the XID is going to be published, make sure it is psersistent */
+	EnsureTopTransactionIdLogged();
+
 	PG_RETURN_FULLTRANSACTIONID(topfxid);
 }
 
diff --git a/src/include/access/xact.h b/src/include/access/xact.h
index 34cfaf542c..a61e4d6da5 100644
--- a/src/include/access/xact.h
+++ b/src/include/access/xact.h
@@ -386,7 +386,8 @@ extern FullTransactionId GetTopFullTransactionId(void);
 extern FullTransactionId GetTopFullTransactionIdIfAny(void);
 extern FullTransactionId GetCurrentFullTransactionId(void);
 extern FullTransactionId GetCurrentFullTransactionIdIfAny(void);
-extern void MarkCurrentTransactionIdLoggedIfAny(void);
+extern void MarkCurrentTransactionIdLoggedIfAny(XLogRecPtr upto);
+extern void EnsureTopTransactionIdLogged(void);
 extern bool SubTransactionIsActive(SubTransactionId subxid);
 extern CommandId GetCurrentCommandId(bool used);
 extern void SetParallelStartTimestamps(TimestampTz xact_ts, TimestampTz stmt_ts);
-- 
2.17.0


--0qVF/w3MHQqLSynd
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
 filename="0004-wal_compression_method-default-to-zlib.patch"



^ permalink  raw  reply  [nested|flat] 160+ messages in thread

* Re: [PATCH] Support reading large objects with pg_read_all_data
@ 2026-02-13 12:05  Nitin Motiani <[email protected]>
  0 siblings, 0 replies; 160+ messages in thread

From: Nitin Motiani @ 2026-02-13 12:05 UTC (permalink / raw)
  To: Nathan Bossart <[email protected]>; +Cc: Dilip Kumar <[email protected]>; pgsql-hackers

Thanks Nathan for the feedback.

On Fri, Feb 13, 2026 at 2:51 AM Nathan Bossart <[email protected]> wrote:
>
> Thanks for the new patch.  Please create a commitfest entry so that this
> patch 1) isn't forgotten and 2) gets tested by cfbot.
>

I have created the commitfest entry
https://commitfest.postgresql.org/patch/6485/.

> nitpick: I usually try to make small doc updates in a way that doesn't
> disturb the surrounding lines so that we retain as much git history as
> possible.
>

Changed the formatting in the doc.

>
> I'd suggest simplifying this a bit by borrowing some lines from the
> surrounding tests.

Changed the test based on your suggestions.

Attaching the latest patch.

Regards,
Nitin Motiani
Google


Attachments:

  [application/x-patch] v4-0001-Support-large-object-functions-with-pg_read_all_d.patch (6.5K, ../../CAH5HC97zMa-ku8B-J4BDSedyad5FT3uRqPLYwnXPHgOWfeGW-A@mail.gmail.com/2-v4-0001-Support-large-object-functions-with-pg_read_all_d.patch)
  download | inline diff:
From 7943be732780b56e0c693986a3eabcb87023ed95 Mon Sep 17 00:00:00 2001
From: Nitin Motiani <[email protected]>
Date: Thu, 5 Feb 2026 09:08:57 +0000
Subject: [PATCH v4] Support large object functions with pg_read_all_data

Allow members of the pg_read_all_data predefined role to access large
objects via the large object functional API (e.g., lo_get, loread).

Previously, while pg_read_all_data permitted direct SELECT queries
on the pg_largeobject catalog table, it did not grant the necessary
permissions to use the logical large object functions. This created an
inconsistency in how the role accessed data stored in large objects
versus standard relations.

This change updates the ACL check for large objects to recognize
pg_read_all_data as having SELECT privileges. Note that this
support is intentionally omitted for pg_write_all_data, as granting
write access would imply write permissions on a system catalog, a
privilege level that pg_write_all_data does not currently provide
for other objects.
---
 doc/src/sgml/user-manag.sgml             |  2 +-
 src/backend/catalog/aclchk.c             | 11 ++++++
 src/test/regress/expected/privileges.out | 44 +++++++++++++++++++++++-
 src/test/regress/sql/privileges.sql      | 21 ++++++++++-
 4 files changed, 75 insertions(+), 3 deletions(-)

diff --git a/doc/src/sgml/user-manag.sgml b/doc/src/sgml/user-manag.sgml
index ed18704a9c2..b268b7b046a 100644
--- a/doc/src/sgml/user-manag.sgml
+++ b/doc/src/sgml/user-manag.sgml
@@ -713,7 +713,7 @@ GRANT pg_signal_backend TO admin_user;
      <listitem>
       <para>
        <literal>pg_read_all_data</literal> allows reading all data (tables,
-       views, sequences), as if having <command>SELECT</command> rights on
+       views, sequences, and large objects), as if having <command>SELECT</command> rights on
        those objects and <literal>USAGE</literal> rights on all schemas.  This
        role does not bypass row-level security (RLS) policies.  If RLS is being
        used, an administrator may wish to set <literal>BYPASSRLS</literal> on
diff --git a/src/backend/catalog/aclchk.c b/src/backend/catalog/aclchk.c
index a431fc0926f..36d9207672f 100644
--- a/src/backend/catalog/aclchk.c
+++ b/src/backend/catalog/aclchk.c
@@ -3598,6 +3598,17 @@ pg_largeobject_aclmask_snapshot(Oid lobj_oid, Oid roleid,
 
 	table_close(pg_lo_meta, AccessShareLock);
 
+	/*
+	 * Check if ACL_SELECT is being checked and, if so, and not set already as
+	 * part of the result, then check if the user is a member of the
+	 * pg_read_all_data role, which allows read access to all relations. We
+	 * don't provide any write access to PG_WRITE_ALL_DATA as that would be
+	 * equivalent to providing write access to a system catalog.
+	 */
+	if (mask & ACL_SELECT && !(result & ACL_SELECT) &&
+		has_privs_of_role(roleid, ROLE_PG_READ_ALL_DATA))
+		result |= ACL_SELECT;
+
 	return result;
 }
 
diff --git a/src/test/regress/expected/privileges.out b/src/test/regress/expected/privileges.out
index 84c1c1ca38d..f07b118f943 100644
--- a/src/test/regress/expected/privileges.out
+++ b/src/test/regress/expected/privileges.out
@@ -2175,6 +2175,42 @@ SELECT lo_truncate(lo_open(2001, x'20000'::int), 10);
            0
 (1 row)
 
+\c -
+SELECT lowrite(lo_open(1002, x'20000'::int), 'hello world');
+ lowrite 
+---------
+      11
+(1 row)
+
+-- pg_read_all_data implies read access to large objects
+SET SESSION AUTHORIZATION regress_priv_user6;
+SELECT loread(lo_open(1002, x'40000'::int), 32); -- ok
+          loread          
+--------------------------
+ \x68656c6c6f20776f726c64
+(1 row)
+
+SELECT lo_get(1002); -- ok
+          lo_get          
+--------------------------
+ \x68656c6c6f20776f726c64
+(1 row)
+
+SELECT lowrite(lo_open(1002, x'20000'::int), 'abcd');   -- to be denied
+ERROR:  permission denied for large object 1002
+SELECT lo_unlink(1002);                                 -- to be denied
+ERROR:  must be owner of large object 1002
+SELECT lo_truncate(lo_open(1002, x'20000'::int), 10);   -- to be denied
+ERROR:  permission denied for large object 1002
+SELECT lo_put(1002, 1, 'abcd');                         -- to be denied
+ERROR:  permission denied for large object 1002
+RESET SESSION AUTHORIZATION;
+SELECT lo_truncate(lo_open(1002, x'20000'::int), 0);	-- ok
+ lo_truncate 
+-------------
+           0
+(1 row)
+
 -- has_largeobject_privilege function
 -- superuser
 \c -
@@ -2727,7 +2763,13 @@ SELECT has_largeobject_privilege('regress_priv_user2', 1008, 'SELECT'); -- yes
  t
 (1 row)
 
-SELECT has_largeobject_privilege('regress_priv_user6', 1008, 'SELECT'); -- no
+SELECT has_largeobject_privilege('regress_priv_user6', 1008, 'SELECT'); -- yes
+ has_largeobject_privilege 
+---------------------------
+ t
+(1 row)
+
+SELECT has_largeobject_privilege('regress_priv_user3', 1008, 'SELECT'); -- no
  has_largeobject_privilege 
 ---------------------------
  f
diff --git a/src/test/regress/sql/privileges.sql b/src/test/regress/sql/privileges.sql
index 66e06d91a41..cc1947ef7df 100644
--- a/src/test/regress/sql/privileges.sql
+++ b/src/test/regress/sql/privileges.sql
@@ -1384,6 +1384,24 @@ SELECT loread(lo_open(1005, x'40000'::int), 32);
 SELECT lo_truncate(lo_open(1005, x'20000'::int), 10);	-- to be denied
 SELECT lo_truncate(lo_open(2001, x'20000'::int), 10);
 
+\c -
+
+SELECT lowrite(lo_open(1002, x'20000'::int), 'hello world');
+
+-- pg_read_all_data implies read access to large objects
+SET SESSION AUTHORIZATION regress_priv_user6;
+
+SELECT loread(lo_open(1002, x'40000'::int), 32); -- ok
+SELECT lo_get(1002); -- ok
+SELECT lowrite(lo_open(1002, x'20000'::int), 'abcd');   -- to be denied
+SELECT lo_unlink(1002);                                 -- to be denied
+SELECT lo_truncate(lo_open(1002, x'20000'::int), 10);   -- to be denied
+SELECT lo_put(1002, 1, 'abcd');                         -- to be denied
+
+RESET SESSION AUTHORIZATION;
+
+SELECT lo_truncate(lo_open(1002, x'20000'::int), 0);	-- ok
+
 -- has_largeobject_privilege function
 
 -- superuser
@@ -1619,7 +1637,8 @@ ALTER DEFAULT PRIVILEGES GRANT SELECT ON LARGE OBJECTS TO regress_priv_user2;
 
 SELECT lo_create(1008);
 SELECT has_largeobject_privilege('regress_priv_user2', 1008, 'SELECT'); -- yes
-SELECT has_largeobject_privilege('regress_priv_user6', 1008, 'SELECT'); -- no
+SELECT has_largeobject_privilege('regress_priv_user6', 1008, 'SELECT'); -- yes
+SELECT has_largeobject_privilege('regress_priv_user3', 1008, 'SELECT'); -- no
 SELECT has_largeobject_privilege('regress_priv_user2', 1008, 'UPDATE'); -- no
 
 ALTER DEFAULT PRIVILEGES GRANT ALL ON LARGE OBJECTS TO regress_priv_user2;
-- 
2.53.0.310.g728cabbaf7-goog



^ permalink  raw  reply  [nested|flat] 160+ messages in thread


end of thread, other threads:[~2026-02-13 12:05 UTC | newest]

Thread overview: 160+ messages (download: mbox mbox.gz follow: Atom feed)
-- links below jump to the message on this page --
2017-05-11 01:48 Race conditions with WAL sender PID lookups Michael Paquier <[email protected]>
2017-05-12 02:44 ` Masahiko Sawada <[email protected]>
2017-05-15 03:25   ` Kyotaro HORIGUCHI <[email protected]>
2017-05-16 15:14 ` Robert Haas <[email protected]>
2017-05-18 02:25   ` Michael Paquier <[email protected]>
2017-05-18 04:43 ` Thomas Munro <[email protected]>
2017-05-19 02:05   ` Michael Paquier <[email protected]>
2017-05-19 06:01     ` Masahiko Sawada <[email protected]>
2017-05-20 12:40       ` Michael Paquier <[email protected]>
2017-05-21 04:37         ` Erik Rijkers <[email protected]>
2017-05-21 06:19           ` Erik Rijkers <[email protected]>
2017-05-30 01:38         ` Noah Misch <[email protected]>
2017-05-30 18:09           ` Peter Eisentraut <[email protected]>
2017-05-30 18:13             ` Michael Paquier <[email protected]>
2017-06-02 05:21               ` Noah Misch <[email protected]>
2017-06-02 06:11                 ` Michael Paquier <[email protected]>
2017-06-07 18:31         ` Robert Haas <[email protected]>
2017-06-07 21:00           ` Erik Rijkers <[email protected]>
2017-06-08 04:15           ` Michael Paquier <[email protected]>
2017-06-13 02:16             ` Michael Paquier <[email protected]>
2017-06-15 01:59               ` Noah Misch <[email protected]>
2017-06-15 08:04                 ` Simon Riggs <[email protected]>
2017-06-22 05:52                   ` Noah Misch <[email protected]>
2017-06-24 05:21                     ` Noah Misch <[email protected]>
2017-06-25 05:11                       ` Alvaro Herrera <[email protected]>
2017-06-25 08:38                         ` Michael Paquier <[email protected]>
2017-06-15 07:51             ` Kyotaro HORIGUCHI <[email protected]>
2017-06-28 22:52             ` Alvaro Herrera <[email protected]>
2017-06-29 00:34               ` Michael Paquier <[email protected]>
2017-06-30 22:20                 ` Alvaro Herrera <[email protected]>
2017-06-30 03:43               ` Noah Misch <[email protected]>
2017-06-30 04:52                 ` Alvaro Herrera <[email protected]>
2017-05-30 22:23     ` Thomas Munro <[email protected]>
2021-03-08 06:43 [PATCH v1 2/2] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 2/8] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH v1 2/2] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 2/8] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH v1 2/2] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH v1 2/2] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 2/8] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/12] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH v9 3/9] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH v1 2/2] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH v1 2/2] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/12] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH v1 2/2] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH v1 2/2] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 2/8] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH v1 2/2] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 2/8] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH v1 2/2] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 2/8] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 2/8] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 2/8] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH v1 2/2] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH v1 2/2] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH v1 2/2] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 2/8] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH v1 2/2] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH v1 2/2] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 2/8] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH v1 2/2] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH v8 3/9] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH v1 2/2] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 2/8] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH v1 2/2] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH v1 2/2] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH v1 2/2] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH v1 2/2] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 2/8] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 2/8] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 6/6] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 2/8] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 2/8] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH v1 2/2] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 2/8] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 2/8] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 2/8] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH v7 3/9] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 2/8] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH v1 2/2] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 2/8] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH v1 2/2] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 2/8] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 2/8] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 2/8] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 2/8] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2021-03-08 06:43 [PATCH 03/10] Make sure published XIDs are persistent Kyotaro Horiguchi <[email protected]>
2026-02-13 12:05 Re: [PATCH] Support reading large objects with pg_read_all_data Nitin Motiani <[email protected]>

This inbox is served by agora; see mirroring instructions
for how to clone and mirror all data and code used for this inbox