Dropping twice a pgstats entry should not happen, and the error report
generated was missing the "generation" counter (tracking when an entry
is reused) that has been added in 818119afcc.
Like d92573adcb, backpatch down to v15 where this information is
useful to have, to gather more information from instances where the
problem shows up. A report has shown that this error path has been
reached on a standby based on 17.3, for a relation stats entry and an
OID close to wraparound.
Author: Bertrand Drouvot <bertranddrouvot.pg@gmail.com>
Discussion: https://postgr.es/m/CAN4RuQvYth942J2+FcLmJKgdpq6fE5eqyFvb_PuskxF2eL=Wzg@mail.gmail.com
Backpatch-through: 15
Use Min(NBuffers, MAX_CHECKPOINT_REQUESTS) instead of NBuffers in
CheckpointerShmemSize() to match the actual array size limit set in
CheckpointerShmemInit(). This prevents wasting shared memory when
NBuffers > MAX_CHECKPOINT_REQUESTS. Also, fix the comment.
Reported-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/1439188.1754506714%40sss.pgh.pa.us
Author: Xuneng Zhou <xunengzhou@gmail.com>
Co-authored-by: Alexander Korotkov <aekorotkov@gmail.com>
The result of "DirectFunctionCall1(numeric_float8, d)" is already in
Datum form, but the code was incorrectly applying PG_RETURN_FLOAT8()
to it. On machines where float8 is pass-by-reference, this would
result in complete garbage, since an unpredictable pointer value
would be treated as an integer and then converted to float. It's not
entirely clear how much of a problem would ensue on 64-bit hardware,
but certainly interpreting a float8 bitpattern as uint64 and then
converting that to float isn't the intended behavior.
As luck would have it, even the complete-garbage case doesn't break
BRIN indexes, since the results are only used to make choices about
how to merge values into ranges: at worst, we'd make poor choices
resulting in an inefficient index. Doubtless that explains the lack
of field complaints. However, users with BRIN indexes that use the
numeric_minmax_multi_ops opclass may wish to reindex in hopes of
making their indexes more efficient.
Author: Peter Eisentraut <peter@eisentraut.org>
Co-authored-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/2093712.1753983215@sss.pgh.pa.us
Backpatch-through: 14
Currently, ALTER DATABASE/ROLE/SYSTEM RESET [ALL] with an unknown
custom GUC with a prefix reserved by MarkGUCPrefixReserved() errors
(unless a superuser runs a RESET ALL variant). This is problematic
for cases such as an extension library upgrade that removes a GUC.
To fix, simply make sure the relevant code paths explicitly allow
it. Note that we require superuser or privileges on the parameter
to reset it. This is perhaps a bit more restrictive than is
necessary, but it's not clear whether further relaxing the
requirements is safe.
Oversight in commit 88103567cb. The ALTER SYSTEM fix is dependent
on commit 2d870b4aef, which first appeared in v17. Unfortunately,
back-patching that commit would introduce ABI breakage, and while
that breakage seems unlikely to bother anyone, it doesn't seem
worth the risk. Hence, the ALTER SYSTEM part of this commit is
omitted on v15 and v16.
Reported-by: Mert Alev <mert@futo.org>
Reviewed-by: Laurenz Albe <laurenz.albe@cybertec.at>
Discussion: https://postgr.es/m/18964-ba09dea8c98fccd6%40postgresql.org
Backpatch-through: 15
A deadlock can occur when the DDL command and the apply worker acquire
catalog locks in different orders while dropping replication origins.
The issue is rare in PG16 and higher branches because, in most cases, the
tablesync worker performs the origin drop in those branches, and its
locking sequence does not conflict with DDL operations.
This patch ensures consistent lock acquisition to prevent such deadlocks.
As per buildfarm.
Reported-by: Alexander Lakhin <exclusion@gmail.com>
Author: Ajin Cherian <itsajin@gmail.com>
Reviewed-by: Hayato Kuroda <kuroda.hayato@fujitsu.com>
Reviewed-by: vignesh C <vignesh21@gmail.com>
Reviewed-by: Amit Kapila <amit.kapila16@gmail.com>
Backpatch-through: 14, where it was introduced
Discussion: https://postgr.es/m/bab95e12-6cc5-4ebb-80a8-3e41956aa297@gmail.com
For many optional libraries, we extract the -L and -l switches needed
to link the library from a helper program such as llvm-config. In
some cases we put the resulting -L switches into LDFLAGS ahead of
-L switches specified via --with-libraries. That risks breaking
the user's intention for --with-libraries.
It's not such a problem if the library's -L switch points to a
directory containing only that library, but on some platforms a
library helper may "helpfully" offer a switch such as -L/usr/lib
that points to a directory holding all standard libraries. If the
user specified --with-libraries in hopes of overriding the standard
build of some library, the -L/usr/lib switch prevents that from
happening since it will come before the user-specified directory.
To fix, avoid inserting these switches directly into LDFLAGS during
configure, instead adding them to LIBDIRS or SHLIB_LINK. They will
still eventually get added to LDFLAGS, but only after the switches
coming from --with-libraries.
The same problem exists for -I switches: those coming from
--with-includes should appear before any coming from helper programs
such as llvm-config. We have not heard field complaints about this
case, but it seems certain that a user attempting to override a
standard library could have issues.
The changes for this go well beyond configure itself, however,
because many Makefiles have occasion to manipulate CPPFLAGS to
insert locally-desirable -I switches, and some of them got it wrong.
The correct ordering is any -I switches pointing at within-the-
source-tree-or-build-tree directories, then those from the tree-wide
CPPFLAGS, then those from helper programs. There were several places
that risked pulling in a system-supplied copy of libpq headers, for
example, instead of the in-tree files. (Commit cb36f8ec2 fixed one
instance of that a few months ago, but this exercise found more.)
The Meson build scripts may or may not have any comparable problems,
but I'll leave it to someone else to investigate that.
Reported-by: Charles Samborski <demurgos@demurgos.net>
Author: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/70f2155f-27ca-4534-b33d-7750e20633d7@demurgos.net
Backpatch-through: 13
When I prepared 71c0921b6 et al yesterday, I was thinking that the
logic involving explicitly freeing the node_list output was still
needed to dodge leakage bugs in libxml2. But I was misremembering:
we introduced that only because with early 2.13.x releases we could
not trust xmlParseBalancedChunkMemory's result code, so we had to
look to see if a node list was returned or not. There's no reason
to believe that xmlParseBalancedChunkMemory will fail to clean up
the node list when required, so simplify. (This essentially
completes reverting all the non-cosmetic changes in 6082b3d5d.)
Reported-by: Jim Jones <jim.jones@uni-muenster.de>
Author: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/997668.1753802857@sss.pgh.pa.us
Backpatch-through: 13
This mostly reverts commit 6082b3d5d, "Use xmlParseInNodeContext
not xmlParseBalancedChunkMemory". It turns out that
xmlParseInNodeContext will reject text chunks exceeding 10MB, while
(in most libxml2 versions) xmlParseBalancedChunkMemory will not.
The bleeding-edge libxml2 bug that we needed to work around a year
ago is presumably no longer a factor, and the argument that
xmlParseBalancedChunkMemory is semi-deprecated is not enough to
justify a functionality regression. Hence, go back to doing it
the old way.
Reported-by: Michael Paquier <michael@paquier.xyz>
Author: Michael Paquier <michael@paquier.xyz>
Co-authored-by: Erik Wienhold <ewie@ewie.name>
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/aIGknLuc8b8ega2X@paquier.xyz
Backpatch-through: 13
If the number of sync requests is big enough, the palloc() call in
AbsorbSyncRequests() will attempt to allocate more than 1 GB of memory,
resulting in failure. This can lead to an infinite loop in the checkpointer
process, as it repeatedly fails to absorb the pending requests.
This commit limits the checkpointer requests queue size to 10M items. In
addition to preventing the palloc() failure, this change helps to avoid long
queue processing time.
Also, this commit is for backpathing only. The master branch receives
a more invasive yet comprehensive fix for this problem.
Discussion: https://postgr.es/m/db4534f83a22a29ab5ee2566ad86ca92%40postgrespro.ru
Backpatch-through: 13
Solaris has never bothered to add "const" to the second argument
of PAM conversation procs, as all other Unixen did decades ago.
This resulted in an "incompatible pointer" compiler warning when
building --with-pam, but had no more serious effect than that,
so we never did anything about it. However, as of GCC 14 the
case is an error not warning by default.
To complicate matters, recent OpenIndiana (and maybe illumos
in general?) *does* supply the "const" by default, so we can't
just assume that platforms using our solaris template need help.
What we can do, short of building a configure-time probe,
is to make solaris.h #define _PAM_LEGACY_NONCONST, which
causes OpenIndiana's pam_appl.h to revert to the traditional
definition, and hopefully will have no effect anywhere else.
Then we can use that same symbol to control whether we include
"const" in the declaration of pam_passwd_conv_proc().
Bug: #18995
Reported-by: Andrew Watkins <awatkins1966@gmail.com>
Author: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/18995-82058da9ab4337a7@postgresql.org
Backpatch-through: 13
If a crash occurs while writing a WAL record that spans multiple pages, the
recovery process marks the page with the XLP_FIRST_IS_OVERWRITE_CONTRECORD
flag. However, logical decoding currently attempts to read the full WAL
record based on its expected size before checking this flag, which can lead
to an infinite wait if the remaining data is never written (e.g., no activity
after crash).
This patch updates the logic first to read the page header and check for
the XLP_FIRST_IS_OVERWRITE_CONTRECORD flag before attempting to reconstruct
the full WAL record. If the flag is set, decoding correctly identifies
the record as incomplete and avoids waiting for WAL data that will never
arrive.
Discussion: https://postgr.es/m/CAAKRu_ZCOzQpEumLFgG_%2Biw3FTa%2BhJ4SRpxzaQBYxxM_ZAzWcA%40mail.gmail.com
Discussion: https://postgr.es/m/CALDaNm34m36PDHzsU_GdcNXU0gLTfFY5rzh9GSQv%3Dw6B%2BQVNRQ%40mail.gmail.com
Author: Vignesh C <vignesh21@gmail.com>
Reviewed-by: Hayato Kuroda <kuroda.hayato@fujitsu.com>
Reviewed-by: Dilip Kumar <dilipbalaut@gmail.com>
Reviewed-by: Michael Paquier <michael@paquier.xyz>
Reviewed-by: Alexander Korotkov <aekorotkov@gmail.com>
Backpatch-through: 13
We have an assertion to ensure that a command tag has been assigned by
the time we're done executing, but if we happen to execute a command
with no queries, the assertion would fail. Per discussion, rather than
contort things to get a tag assigned, just remove the assertion.
Oversight in 2f9661311b. That commit also retained a comment that
explained logic that had been adjacent to it but diffused into various
places, leaving none apt to keep part of the comment. Remove that part,
and rewrite what remains for extra clarity.
Bug: #18984
Backpatch-through: 13
Reported-by: Aleksander Alekseev <aleksander@tigerdata.com>
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Reviewed-by: Michaël Paquier <michael@paquier.xyz>
Discussion: https://postgr.es/m/18984-0f4778a6599ac3ae@postgresql.org
Because not every path through JsonbIteratorNext() sets val->type,
some compilers complain that compareJsonbContainers() is comparing
possibly-uninitialized values. The paths that don't set it return
WJB_DONE, WJB_END_ARRAY, or WJB_END_OBJECT, so it's clear by
manual inspection that the "(ra == rb)" code path is safe, and
indeed we aren't seeing warnings about that. But the (ra != rb)
case is much less obviously safe. In Assert-enabled builds it
seems that the asserts rejecting WJB_END_ARRAY and WJB_END_OBJECT
persuade gcc 15.x not to warn, which makes little sense because
it's impossible to believe that the compiler can prove of its
own accord that ra/rb aren't WJB_DONE here. (In fact they never
will be, so the code isn't wrong, but why is there no warning?)
Without Asserts, the appearance of warnings is quite unsurprising.
We discussed fixing this by converting those two Asserts into
pg_assume, but that seems not very satisfactory when it's so unclear
why the compiler is or isn't warning: the warning could easily
reappear with some other compiler version. Let's fix it in a less
magical, more future-proof way by changing JsonbIteratorNext()
so that it always does set val->type. The cost of that should be
pretty negligible, and it makes the function's API spec less squishy.
Reported-by: Erik Rijkers <er@xs4all.nl>
Author: Tom Lane <tgl@sss.pgh.pa.us>
Reviewed-by: Andres Freund <andres@anarazel.de>
Discussion: https://postgr.es/m/988bf1bc-3f1f-99f3-bf98-222f1cd9dc5e@xs4all.nl
Discussion: https://postgr.es/m/0c623e8a204187b87b4736792398eaf1@postgrespro.ru
Backpatch-through: 13
getid() and putid(), which parse and deparse role names within ACL
input/output, applied isalnum() to see if a character within a role
name requires quoting. They did this even for non-ASCII characters,
which is problematic because the results would depend on encoding,
locale, and perhaps even platform. So it's possible that putid()
could elect not to quote some string that, later in some other
environment, getid() will decide is not a valid identifier, causing
dump/reload or similar failures.
To fix this in a way that won't risk interoperability problems
with unpatched versions, make getid() treat any non-ASCII as a
legitimate identifier character (hence not requiring quotes),
while making putid() treat any non-ASCII as requiring quoting.
We could remove the resulting excess quoting once we feel that
no unpatched servers remain in the wild, but that'll be years.
A lesser problem is that getid() did the wrong thing with an input
consisting of just two double quotes (""). That has to represent an
empty string, but getid() read it as a single double quote instead.
The case cannot arise in the normal course of events, since we don't
allow empty-string role names. But let's fix it while we're here.
Although we've not heard field reports of problems with non-ASCII
role names, there's clearly a hazard there, so back-patch to all
supported versions.
Reported-by: Peter Eisentraut <peter@eisentraut.org>
Author: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/3792884.1751492172@sss.pgh.pa.us
Backpatch-through: 13
Attempting to use commit timestamps during bootstrapping leads to an
assertion failure, that can be reached for example with an initdb -c
that enables track_commit_timestamp. It makes little sense to register
a commit timestamp for a BootstrapTransactionId, so let's disable the
activation of the module in this case.
This problem has been independently reported once by each author of this
commit. Each author has proposed basically the same patch, relying on
IsBootstrapProcessingMode() to skip the use of commit_ts during
bootstrap. The test addition is a suggestion by me, and is applied down
to v16.
Author: Hayato Kuroda <kuroda.hayato@fujitsu.com>
Author: Andy Fan <zhihuifan1213@163.com>
Reviewed-by: Bertrand Drouvot <bertranddrouvot.pg@gmail.com>
Reviewed-by: Fujii Masao <masao.fujii@oss.nttdata.com>
Reviewed-by: Michael Paquier <michael@paquier.xyz>
Discussion: https://postgr.es/m/OSCPR01MB14966FF9E4C4145F37B937E52F5102@OSCPR01MB14966.jpnprd01.prod.outlook.com
Discussion: https://postgr.es/m/87plejmnpy.fsf@163.com
Backpatch-through: 13
Commits 8319e5cb5 et al missed the fact that ATPostAlterTypeCleanup
contains three calls to ATPostAlterTypeParse, and the other two
also need protection against passing a relid that we don't yet
have lock on. Add similar logic to those code paths, and add
some test cases demonstrating the need for it.
In v18 and master, the test cases demonstrate that there's a
behavioral discrepancy between stored generated columns and virtual
generated columns: we disallow changing the expression of a stored
column if it's used in any composite-type columns, but not that of
a virtual column. Since the expression isn't actually relevant to
either sort of composite-type usage, this prohibition seems
unnecessary; but changing it is a matter for separate discussion.
For now we are just documenting the existing behavior.
Reported-by: jian he <jian.universality@gmail.com>
Author: jian he <jian.universality@gmail.com>
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: CACJufxGKJtGNRRSXfwMW9SqVOPEMdP17BJ7DsBf=tNsv9pWU9g@mail.gmail.com
Backpatch-through: 13
Sometimes a table's constraint may depend on a column of another
table, so that we have to update the constraint when changing the
referenced column's type. We need to have lock on the constraint's
table to do that. ATPostAlterTypeCleanup believed that this case
was only possible for FOREIGN KEY constraints, but it's wrong at
least for CHECK and EXCLUDE constraints; and in general, we'd
probably need exclusive lock to alter any sort of constraint.
So just remove the contype check and acquire lock for any other
table. This prevents a "you don't have lock" assertion failure,
though no ill effect is observed in production builds.
We'll error out later anyway because we don't presently support
physically altering column types within stored composite columns.
But the catalog-munging is basically all there, so we may as well
make that part work.
Bug: #18970
Reported-by: Alexander Lakhin <exclusion@gmail.com>
Diagnosed-by: jian he <jian.universality@gmail.com>
Author: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/18970-a7d1cfe1f8d5d8d9@postgresql.org
Backpatch-through: 13
This fixes two issues with the handling of VacuumParams in vacuum_rel().
This code path has the idea to change the passed-in pointer of
VacuumParams for the "truncate" and "index_cleanup" options for the
relation worked on, impacting the two following scenarios where
incorrect options may be used because a VacuumParams pointer is shared
across multiple relations:
- Multiple relations in a single VACUUM command.
- TOAST relations vacuumed with their main relation.
The problem is avoided by providing to the two callers of vacuum_rel()
copies of VacuumParams, before the pointer is updated for the "truncate"
and "index_cleanup" options.
The refactoring of the VACUUM option and parameters done in 0d83138974
did not introduce an issue, but it has encouraged the problem we are
dealing with in this commit, with b84dbc8eb8 for "truncate" and
a96c41feec for "index_cleanup" that have been added a couple of years
after the initial refactoring. HEAD will be improved with a different
patch that hardens the uses of VacuumParams across the tree. This
cannot be backpatched as it introduces an ABI breakage.
The backend portion of the patch has been authored by Nathan, while I
have implemented the tests. The tests rely on injection points to check
the option values, making them faster, more reliable than the tests
originally proposed by Shihao, and they also provide more coverage.
This part can only be backpatched down to v17.
Reported-by: Shihao Zhong <zhong950419@gmail.com>
Author: Nathan Bossart <nathandbossart@gmail.com>
Co-authored-by: Michael Paquier <michael@paquier.xyz>
Discussion: https://postgr.es/m/CAGRkXqTo+aK=GTy5pSc-9cy8H2F2TJvcrZ-zXEiNJj93np1UUw@mail.gmail.com
Backpatch-through: 13
Specify whether the bucket bounds are inclusive or exclusive,
and improve some other vague language. Explain the behavior that
occurs when the "low" bound is greater than the "high" bound.
Make width_bucket_numeric's comment more like that for
width_bucket_float8, in particular noting that infinite
bounds are rejected (since they became possible in v14).
Reported-by: Ben Peachey Higdon <bpeacheyhigdon@gmail.com>
Author: Robert Treat <rob@xzilla.net>
Co-authored-by: Tom Lane <tgl@sss.pgh.pa.us>
Reviewed-by: Dean Rasheed <dean.a.rasheed@gmail.com>
Discussion: https://postgr.es/m/2BD74F86-5B89-4AC1-8F13-23CED3546AC1@gmail.com
Backpatch-through: 13
While choosing an autogenerated name for an index, look for
pre-existing relations using a SnapshotDirty snapshot, instead of the
previous behavior that considered only committed-good pg_class rows.
This allows us to detect and avoid conflicts against indexes that are
still being built.
It's still possible to fail due to a race condition, but the window
is now just the amount of time that it takes DefineIndex to validate
all its parameters, call smgrcreate(), and enter the index's pg_class
row. Formerly the race window covered the entire time needed to
create and fill an index, which could be very long if the table is
large. Worse, if the conflicting index creation is part of a larger
transaction, it wouldn't be visible till COMMIT.
So this isn't a complete solution, but it should greatly ameliorate
the problem, and the patch is simple enough to be back-patchable.
It might at some point be useful to do the same for pg_constraint
entries (cf. ChooseConstraintName, ConstraintNameExists, and related
functions). However, in the absence of field complaints, I'll leave
that alone for now. The relation-name test should be good enough for
index-based constraints, while foreign-key constraints seem to be okay
since they require exclusive locks to create.
Bug: #18959
Reported-by: Maximilian Chrzan <maximilian.chrzan@here.com>
Author: Tom Lane <tgl@sss.pgh.pa.us>
Reviewed-by: Dilip Kumar <dilipbalaut@gmail.com>
Discussion: https://postgr.es/m/18959-f63b53b864bb1417@postgresql.org
Backpatch-through: 13
Commit 4909b38af0 introduced logic to distribute invalidation messages
from catalog-modifying transactions to all concurrent in-progress
transactions. However, since each transaction distributes not only its
original invalidation messages but also previously distributed
messages to other transactions, this leads to an exponential increase
in allocation request size for invalidation messages, ultimately
causing memory allocation failure.
This commit fixes this issue by tracking distributed invalidation
messages separately per decoded transaction and not redistributing
these messages to other in-progress transactions. The maximum size of
distributed invalidation messages that one transaction can store is
limited to MAX_DISTR_INVAL_MSG_PER_TXN (8MB). Once the size of the
distributed invalidation messages exceeds this threshold, we
invalidate all caches in locations where distributed invalidation
messages need to be executed.
Back-patch to all supported versions where we introduced the fix by
commit 4909b38af0.
Note that this commit adds two new fields to ReorderBufferTXN to store
the distributed transactions. This change breaks ABI compatibility in
back branches, affecting third-party extensions that depend on the
size of the ReorderBufferTXN struct, though this scenario seems
unlikely.
Additionally, it adds a new flag to the txn_flags field of
ReorderBufferTXN to indicate distributed invalidation message
overflow. This should not affect existing implementations, as it is
unlikely that third-party extensions use unused bits in the txn_flags
field.
Bug: #18938#18942
Author: vignesh C <vignesh21@gmail.com>
Reported-by: Duncan Sands <duncan.sands@deepbluecap.com>
Reported-by: John Hutchins <john.hutchins@wicourts.gov>
Reported-by: Laurence Parry <greenreaper@hotmail.com>
Reported-by: Max Madden <maxmmadden@gmail.com>
Reported-by: Braulio Fdo Gonzalez <brauliofg@gmail.com>
Reviewed-by: Masahiko Sawada <sawada.mshk@gmail.com>
Reviewed-by: Amit Kapila <amit.kapila16@gmail.com>
Reviewed-by: Hayato Kuroda <kuroda.hayato@fujitsu.com>
Discussion: https://postgr.es/m/680bdaf6-f7d1-4536-b580-05c2760c67c6@deepbluecap.com
Discussion: https://postgr.es/m/18942-0ab1e5ae156613ad@postgresql.org
Discussion: https://postgr.es/m/18938-57c9a1c463b68ce0@postgresql.org
Discussion: https://postgr.es/m/CAD1FGCT2sYrP_70RTuo56QTizyc+J3wJdtn2gtO3VttQFpdMZg@mail.gmail.com
Discussion: https://postgr.es/m/CANO2=B=2BT1hSYCE=nuuTnVTnjidMg0+-FfnRnqM6kd23qoygg@mail.gmail.com
Backpatch-through: 13
The patch fixes the issue with the unexpected removal of old WAL segments
after checkpoint, followed by an immediate restart. The issue occurs when
a slot is advanced after the start of the checkpoint and before old WAL
segments are removed at the end of the checkpoint.
The idea of the patch is to get the minimal restart_lsn at the beginning
of checkpoint (or restart point) creation and use this value when calculating
the oldest LSN for WAL segments removal at the end of checkpoint. This idea
was proposed by Tomas Vondra in the discussion. Unlike 291221c46575, this
fix doesn't affect ABI and is intended for back branches.
Discussion: https://postgr.es/m/flat/1d12d2-67235980-35-19a406a0%4063439497
Author: Vitaly Davydov <v.davydov@postgrespro.ru>
Reviewed-by: Tomas Vondra <tomas@vondra.me>
Reviewed-by: Alexander Korotkov <aekorotkov@gmail.com>
Reviewed-by: Amit Kapila <amit.kapila16@gmail.com>
Backpatch-through: 13
Teach nbtree's _bt_killitems to leave the so->currPos page that it sets
LP_DEAD items on in whatever state it was in when _bt_killitems was
called. In particular, make sure that so->dropPin scans don't acquire a
pin whose reference is saved in so->currPos.buf.
Allowing _bt_killitems to change so->currPos.buf like this is wrong.
The immediate consequence of allowing it is that code in _bt_steppage
(that copies so->currPos into so->markPos) will behave as if the scan is
a !so->dropPin scan. so->markPos will therefore retain the buffer pin
indefinitely, even though _bt_killitems only needs to acquire a pin
(along with a lock) for long enough to mark known-dead items LP_DEAD.
This issue came to light following a report of a failure of an assertion
from recent commit e6eed40e. The test case in question involves the use
of mark and restore. An initial call to _bt_killitems takes place that
leaves so->currPos.buf in a state that is inconsistent with the scan
being so->dropPin. A subsequent call to _bt_killitems for the same
position (following so->currPos being saved in so->markPos, and then
restored as so->currPos) resulted in the failure of an assertion that
tests that so->currPos.buf is InvalidBuffer when the scan is so->dropPin
(non-assert builds got a "resource was not closed" WARNING instead).
The same problem exists on earlier releases, though the issue is far
more subtle there. Recent commit e6eed40e introduced the so->dropPin
field as a partial replacement for testing so->currPos.buf directly.
Earlier releases won't get an assertion failure (or buffer pin leak),
but they will allow the second _bt_killitems call from the test case to
behave as if a buffer pin was consistently held since the original call
to _bt_readpage. This is wrong; there will have been an initial window
during which no pin was held on the so->currPos page, and yet the second
_bt_killitems call will neglect to check if so->currPos.lsn continues to
match the page's now-current LSN.
As a result of all this, it's just about possible that _bt_killitems
will set the wrong items LP_DEAD (on release branches). This could only
happen with merge joins (the sole user of nbtree mark/restore support),
when a concurrently inserted index tuple used a recently-recycled TID
(and only when the new tuple was inserted onto the same page as a
distinct concurrently-removed tuple with the same TID). This is exactly
the scenario that _bt_killitems' check of the page's now-current LSN
against the LSN stashed in currPos was supposed to prevent.
A follow-up commit will make nbtree completely stop conditioning whether
or not a position's pin needs to be dropped on whether the 'buf' field
is set. All call sites that might need to drop a still-held pin will be
taught to rely on the scan-level so->dropPin field recently introduced
by commit e6eed40e. That will make bugs of the same general nature as
this one impossible (or make them much easier to detect, at least).
Author: Peter Geoghegan <pg@bowt.ie>
Reported-By: Alexander Lakhin <exclusion@gmail.com>
Discussion: https://postgr.es/m/545be1e5-3786-439a-9257-a90d30f8b849@gmail.com
Backpatch-through: 13
Traditionally, libpq's pqPutMsgEnd has rounded down the amount-to-send
to be a multiple of 8K when it is eagerly writing some data. This
still seems like a good idea when sending through a Unix socket, as
pipes typically have a buffer size of 8K or some fraction/multiple of
that. But there's not much argument for it on a TCP connection, since
(a) standard MTU values are not commensurate with that, and (b) the
kernel typically applies its own packet splitting/merging logic.
Worse, our SSL and GSSAPI code paths both have API stipulations that
if they fail to send all the data that was offered in the previous
write attempt, we mustn't offer less data in the next attempt; else
we may get "SSL error: bad length" or "GSSAPI caller failed to
retransmit all data needing to be retried". The previous write
attempt might've been pqFlush attempting to send everything in the
buffer, so pqPutMsgEnd can't safely write less than the full buffer
contents. (Well, we could add some more state to track exactly how
much the previous write attempt was, but there's little value evident
in such extra complication.) Hence, apply the round-down only on
AF_UNIX sockets, where we never use SSL or GSSAPI.
Interestingly, we had a very closely related bug report before,
which I attempted to fix in commit d053a879b. But the test case
we had then seemingly didn't trigger this pqFlush-then-pqPutMsgEnd
scenario, or at least we failed to recognize this variant of the bug.
Bug: #18907
Reported-by: Dorjpalam Batbaatar <htgn.dbat.95@gmail.com>
Author: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/18907-d41b9bcf6f29edda@postgresql.org
Backpatch-through: 13
We store values for these options as array elements with the syntax
"name=value", hence a name containing "=" confuses matters when
it's time to read the array back in. Since validation of the
options is often done (long) after this conversion to array format,
that leads to confusing and off-point error messages. We can
improve matters by rejecting names containing "=" up-front.
(Probably a better design would have involved pairs of array
elements, but it's too late now --- and anyway, there's no
evident use-case for option names like this. We already
reject such names in some other contexts such as GUCs.)
Reported-by: Chapman Flack <jcflack@acm.org>
Author: Tom Lane <tgl@sss.pgh.pa.us>
Reviewed-by: Chapman Flack <jcflack@acm.org>
Discussion: https://postgr.es/m/6830EB30.8090904@acm.org
Backpatch-through: 13
When a MERGE's target table is the parent of an inheritance tree, any
INSERT actions insert into the parent table using ModifyTableState's
rootResultRelInfo. However, there are two bugs in the way this is
initialized:
1. ExecInitMerge() incorrectly uses a different ResultRelInfo entry
from ModifyTableState's resultRelInfo array to build the insert
projection, which may not be compatible with rootResultRelInfo.
2. ExecInitModifyTable() does not fully initialize rootResultRelInfo.
Specifically, ri_WithCheckOptions, ri_WithCheckOptionExprs,
ri_returningList, and ri_projectReturning are not initialized.
This can lead to crashes, or incorrect query results due to failing to
check WCO's or process the RETURNING list for INSERT actions.
Fix both these bugs in ExecInitMerge(), noting that it is only
necessary to fully initialize rootResultRelInfo if the MERGE has
INSERT actions and the target table is a plain inheritance parent.
Backpatch to v15, where MERGE was introduced.
Reported-by: Andres Freund <andres@anarazel.de>
Author: Dean Rasheed <dean.a.rasheed@gmail.com>
Reviewed-by: Jian He <jian.universality@gmail.com>
Reviewed-by: Tender Wang <tndrwang@gmail.com>
Discussion: https://postgr.es/m/4rlmjfniiyffp6b3kv4pfy4jw3pciy6mq72rdgnedsnbsx7qe5@j5hlpiwdguvc
Backpatch-through: 15
A few places that access system catalogs don't set up an active
snapshot before potentially accessing their TOAST tables. To fix,
push an active snapshot just before each section of code that might
require accessing one of these TOAST tables, and pop it shortly
afterwards. While at it, this commit adds some rather strict
assertions in an attempt to prevent such issues in the future.
Commit 16bf24e0e4 recently removed pg_replication_origin's TOAST
table in order to fix the same problem for that catalog. On the
back-branches, those bugs are left in place. We cannot easily
remove a catalog's TOAST table on released major versions, and only
replication origins with extremely long names are affected. Given
the low severity of the issue, fixing older versions doesn't seem
worth the trouble of significantly modifying the patch.
Also, on v13 and v14, the aforementioned strict assertions have
been omitted because commit 2776922201, which added
HaveRegisteredOrActiveSnapshot(), was not back-patched. While we
could probably back-patch it now, I've opted against it because it
seems unlikely that new TOAST snapshot issues will be introduced in
the oldest supported versions.
Reported-by: Alexander Lakhin <exclusion@gmail.com>
Reviewed-by: Michael Paquier <michael@paquier.xyz>
Discussion: https://postgr.es/m/18127-fe54b6a667f29658%40postgresql.org
Discussion: https://postgr.es/m/18309-c0bf914950c46692%40postgresql.org
Discussion: https://postgr.es/m/ZvMSUPOqUU-VNADN%40nathan
Backpatch-through: 13
Our GSSAPI code only allows packet sizes up to 16kB. However it
emerges that during authentication, larger packets might be needed;
various authorities suggest 48kB or 64kB as the maximum packet size.
This limitation caused login failure for AD users who belong to many
AD groups. To add insult to injury, we gave an unintelligible error
message, typically "GSSAPI context establishment error: The routine
must be called again to complete its function: Unknown error".
As noted in code comments, the 16kB packet limit is effectively a
protocol constant once we are doing normal data transmission: the
GSSAPI code splits the data stream at those points, and if we change
the limit then we will have cross-version compatibility problems
due to the receiver's buffer being too small in some combinations.
However, during the authentication exchange the packet sizes are
not determined by us, but by the underlying GSSAPI library. So we
might as well just try to send what the library tells us to.
An unpatched recipient will fail on a packet larger than 16kB,
but that's not worse than the sender failing without even trying.
So this doesn't introduce any meaningful compatibility problem.
We still need a buffer size limit, but we can easily make it be
64kB rather than 16kB until transport negotiation is complete.
(Larger values were discussed, but don't seem likely to add
anything.)
Reported-by: Chris Gooch <cgooch@bamfunds.com>
Fix-suggested-by: Jacob Champion <jacob.champion@enterprisedb.com>
Author: Tom Lane <tgl@sss.pgh.pa.us>
Reviewed-by: Jacob Champion <jacob.champion@enterprisedb.com>
Discussion: https://postgr.es/m/DS0PR22MB5971A9C8A3F44BCC6293C4DABE99A@DS0PR22MB5971.namprd22.prod.outlook.com
Backpatch-through: 13
Previously, XactLockTableWait() and ConditionalXactLockTableWait() could enter
a non-interruptible loop when they successfully acquired a lock on a transaction
but the transaction still appeared to be running. Since this loop continued
until the transaction completed, it could result in long, uninterruptible waits.
Although this scenario is generally unlikely since XactLockTableWait() and
ConditionalXactLockTableWait() can basically acquire a transaction lock
only when the transaction is not running, it can occur in a hot standby.
In such cases, the transaction may still appear active due to
the KnownAssignedXids list, even while no lock on the transaction exists.
For example, this situation can happen when creating a logical replication
slot on a standby.
The cause of the non-interruptible loop was the absence of CHECK_FOR_INTERRUPTS()
within it. This commit adds CHECK_FOR_INTERRUPTS() to the loop in both functions,
ensuring they can be interrupted safely.
Back-patch to all supported branches.
Author: Kevin K Biju <kevinkbiju@gmail.com>
Reviewed-by: Fujii Masao <masao.fujii@gmail.com>
Discussion: https://postgr.es/m/CAM45KeELdjhS-rGuvN=ZLJ_asvZACucZ9LZWVzH7bGcD12DDwg@mail.gmail.com
Backpatch-through: 13
The code that translates SIMILAR TO pattern matching expressions to
POSIX-style regular expressions did not consider that square brackets
can be nested. For example, in an expression like [[:alpha:]%_], the
logic replaced the placeholders '_' and '%' but it should not.
This commit fixes the conversion logic by tracking the nesting level of
square brackets marking character class areas, while considering that
in expressions like []] or [^]] the first closing square bracket is a
regular character. Multiple tests are added to show how the conversions
should or should not apply applied while in a character class area, with
specific cases added for all the characters converted outside character
classes like an opening parenthesis '(', dollar sign '$', etc.
Author: Laurenz Albe <laurenz.albe@cybertec.at>
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Reviewed-by: Michael Paquier <michael@paquier.xyz>
Discussion: https://postgr.es/m/16ab039d1af455652bdf4173402ddda145f2c73b.camel@cybertec.at
Backpatch-through: 13
PgStat_StatTabEntry and AutoVacOpts structs were leaked until
the end of the autovacuum worker's run, which is bad news if
there are a lot of relations in the database.
Note: pfree'ing the PgStat_StatTabEntry structs here seems a bit
risky, because pgstat_fetch_stat_tabentry_ext does not guarantee
anything about whether its result is long-lived. It appears okay
so long as autovacuum forces PGSTAT_FETCH_CONSISTENCY_NONE, but
I think that API could use a re-think.
Also ensure that the VacuumRelation structure passed to
vacuum() is in recoverable storage.
Back-patch to v15 where we started to manage table statistics
this way. (The AutoVacOpts leakage is probably older, but
I'm not excited enough to worry about just that part.)
Author: Tom Lane <tgl@sss.pgh.pa.us>
Reviewed-by: Andres Freund <andres@anarazel.de>
Discussion: https://postgr.es/m/285483.1746756246@sss.pgh.pa.us
Backpatch-through: 15
Since commit 8b1dccd37c, the PREPARE TRANSACTION WAL record includes
information about dropped statistics entries. However, the WAL resource
manager description function for PREPARE TRANSACTION record failed to
parse this information correctly and always assumed there were
no such entries.
As a result, for example, pg_waldump could not display the dropped
statistics entries stored in PREPARE TRANSACTION records.
The root cause was that ParsePrepareRecord() did not set the number of
statistics entries to drop on commit or abort. These values remained
zero-initialized and were never updated from the parsed record.
This commit fixes the issue by properly setting those values during parsing.
With this fix, pg_waldump can now correctly report dropped statistics
entries in PREPARE TRANSACTION records.
Back-patch to v15, where commit 8b1dccd37c was introduced.
Author: Daniil Davydov <3danissimo@gmail.com>
Reviewed-by: Fujii Masao <masao.fujii@gmail.com>
Discussion: https://postgr.es/m/CAJDiXgh-6Epb2XiJe4uL0zF-cf0_s_7Lw1TfEHDMLzYjEmfGOw@mail.gmail.com
Backpatch-through: 15
In the grammar, <expr> is a c_expr, which accepts only a limited set
of integer literals and simple expressions without parens. The
deparsing logic didn't quite match the grammar rule, and failed to use
parens e.g. for "5::bigint".
To fix, always surround the expression with parens. Would be nice to
omit the parens in simple cases, but unfortunately it's non-trivial to
detect such simple cases. Even if the expression is a simple literal
123 in the original query, after parse analysis it becomes a FuncExpr
with COERCE_IMPLICIT_CAST rather than a simple Const.
Reported-by: yonghao lee
Backpatch-through: 13
Discussion: https://www.postgresql.org/message-id/18929-077d6b7093b176e2@postgresql.org
In an XMLTABLE expression, columns can be marked NOT NULL, and the
parser internally fabricates an option named "is_not_null" to
represent this. However, the parser also allows users to specify
arbitrary option names. This creates a conflict: a user can
explicitly use "is_not_null" as an option name and assign it a
non-Boolean value, which violates internal assumptions and triggers an
assertion failure.
To fix, this patch checks whether a user-supplied name collides with
the internally reserved option name and raises an error if so.
Additionally, the internal name is renamed to "__pg__is_not_null" to
further reduce the risk of collision with user-defined names.
Reported-by: Евгений Горбанев <gorbanyoves@basealt.ru>
Author: Richard Guo <guofenglinux@gmail.com>
Reviewed-by: Alvaro Herrera <alvherre@kurilemu.de>
Discussion: https://postgr.es/m/6bac9886-65bf-4cec-96bd-e304159f28db@basealt.ru
Backpatch-through: 15
With GB18030 as source encoding, applications could crash the server via
SQL functions convert() or convert_from(). Applications themselves
could crash after passing unterminated GB18030 input to libpq functions
PQescapeLiteral(), PQescapeIdentifier(), PQescapeStringConn(), or
PQescapeString(). Extension code could crash by passing unterminated
GB18030 input to jsonapi.h functions. All those functions have been
intended to handle untrusted, unterminated input safely.
A crash required allocating the input such that the last byte of the
allocation was the last byte of a virtual memory page. Some malloc()
implementations take measures against that, making the SIGSEGV hard to
reach. Back-patch to v13 (all supported versions).
Author: Noah Misch <noah@leadboat.com>
Author: Andres Freund <andres@anarazel.de>
Reviewed-by: Masahiko Sawada <sawada.mshk@gmail.com>
Backpatch-through: 13
Security: CVE-2025-4207
For self-referencing foreign keys in partitioned tables, we weren't
handling creation of pg_constraint rows during CREATE TABLE PARTITION AS
as well as ALTER TABLE ATTACH PARTITION. This is an old bug -- mostly,
we broke this in 614a406b4f while trying to fix it (so 12.13, 13.9,
14.6 and 15.0 and up all behave incorrectly). This commit reverts part
of that with additional fixes for full correctness, and installs more
tests to verify the parts we broke, not just the catalog contents but
also the user-visible behavior.
Backpatch to all live branches. In branches 13 and 14, commit
46a8c27a72 changed the behavior during DETACH to drop a FK
constraint rather than trying to repair it, because the complete fix of
repairing catalog constraints was problematic due to lack of previous
fixes. For this reason, the test behavior in those branches is a bit
different. However, as best as I can tell, the fix works correctly
there.
In release notes we have to recommend that all self-referencing foreign
keys on partitioned tables be recreated if partitions have been created
or attached after the FK was created, keeping in mind that violating
rows might already be present on the referencing side.
Reported-by: Guillaume Lelarge <guillaume@lelarge.info>
Reported-by: Matthew Gabeler-Lee <fastcat@gmail.com>
Reported-by: Luca Vallisa <luca.vallisa@gmail.com>
Discussion: https://postgr.es/m/CAECtzeWHCA+6tTcm2Oh2+g7fURUJpLZb-=pRXgeWJ-Pi+VU=_w@mail.gmail.com
Discussion: https://postgr.es/m/18156-a44bc7096f0683e6@postgresql.org
Discussion: https://postgr.es/m/CAAT=myvsiF-Attja5DcWoUWh21R12R-sfXECY2-3ynt8kaOqjw@mail.gmail.com
Clear any potential stale next_phase_at value from the snapshot
builder which otherwise may trip an assertion check ensuring
that there is no next_phase_at value.
This can be reproduced by running 80 concurrent sessions like
the below where $c is a loop counter (assumes there has been
1..$c databases created) :
echo "
CREATE TABLE replication_example(id SERIAL PRIMARY KEY,
somedata int,
text varchar(120));
SELECT 'init' FROM
pg_create_logical_replication_slot('regression_slot_$c',
'test_decoding');
SELECT data FROM
pg_logical_slot_get_changes('regression_slot_$c', NULL,
NULL, 'include-xids', '0',
'skip-empty-xacts', '1');
" | psql -d regress_$c >>psql.log &
This was originally committed as 48efb23 and backpatched down to
v16, but since then there have been reports of this happening on
v14 and v15 as well so this is a backpatch of 48efb23 down to 14.
Bug: #17695
Author: Masahiko Sawada <sawada.mshk@gmail.com>
Reviewed-by: Alexander Lakhin <exclusion@gmail.com>
Reported-by: bowenshi <zxwsbg@qq.com>
Reported-by: Alexander Pyhalov <a.pyhalov@postgrespro.ru>
Reported-by: Teja Mupparti
Discussion: https://postgr.es/m/17695-6be9277c9295985f@postgresql.org
Backpatch-through: v14
During logical decoding, we advance catalog_xmin of logical too early in
fast_forward mode, resulting in required catalog data being removed by
vacuum. This mode is normally used to advance the slot without processing
the changes, but we still can't let the slot's xmin to advance to an
incorrect value.
Commit f49a80c481 fixed a similar issue where the logical slot's
catalog_xmin was getting advanced prematurely during non-fast-forward
mode. During xl_running_xacts processing, instead of directly advancing
the slot's xmin to the oldest running xid in the record, it allowed the
xmin to be held back for snapshots that can be used for
not-yet-replayed transactions, as those might consider older txns as
running too. However, it missed the fact that the same problem can happen
during fast_forward mode decoding, as we won't build a base snapshot in
that mode, and the future call to get_changes from the same slot can miss
seeing the required catalog changes leading to incorrect reslts.
This commit allows building the base snapshot even in fast_forward mode to
prevent the early advancement of xmin.
Reported-by: Amit Kapila <amit.kapila16@gmail.com>
Author: Zhijie Hou <houzj.fnst@fujitsu.com>
Reviewed-by: Masahiko Sawada <sawada.mshk@gmail.com>
Reviewed-by: shveta malik <shveta.malik@gmail.com>
Reviewed-by: Amit Kapila <amit.kapila16@gmail.com>
Backpatch-through: 13
Discussion: https://postgr.es/m/CAA4eK1LqWncUOqKijiafe+Ypt1gQAQRjctKLMY953J79xDBgAg@mail.gmail.com
Discussion: https://postgr.es/m/OS0PR01MB57163087F86621D44D9A72BF94BB2@OS0PR01MB5716.jpnprd01.prod.outlook.com
This assertion, based on pending_since (timestamp used to prevent stats
reports to be too frequent or should a partial flush happen), is reached
when it is found that no data can be flushed but a previous call of
pgstat_report_stat() determined that some stats data has been found as
in need of a flush. So pending_since is set when some stats data is
pending (in non-force mode) or if report attempts are too frequent, and
reset to 0 once all stats have been flushed.
Since 5cbbe70a9c, WAL senders have begun to report their stats on a
periodic basis for IO stats in v16~ and backend stats on HEAD, creating
some friction with the concurrent pgstat_report_stat() calls that can
happen in the context of a WAL sender (shutdown callback doing a final
report or backend-related code paths). This problem is the cause of
spurious failures in the TAP tests.
In theory, this assertion can be also reached in v15, even if that's
very unlikely. For example, a process, say a background worker, could
do periodic and direct stats flushes with concurrent calls of
pgstat_report_stat() that could cause conflicting values of
pending_since. This can be done with WAL or SLRU stats flushes using
pgstat_flush_wal() or pgstat_slru_flush(). HEAD makes this situation
easier to happen with custom cumulative stats.
This commit removes the assertion altogether, per discussion, as it is
more useful to keep the state of things as they are for the WAL sender.
The assertion could use a special state based on for example
am_walsender, but I doubt that this would be meaningful in the long run
based on the other arguments raised while discussing this issue.
Reported-by: Tom Lane <tgl@sss.pgh.pa.us>
Reported-by: Andres Freund <andres@anarazel.de>
Discussion: https://postgr.es/m/1489124.1744685908@sss.pgh.pa.us
Discussion: https://postgr.es/m/dwrkeszz6czvtkxzr5mqlciy652zau5qqnm3cp5f3p2po74ppk@omg4g3cc6dgq
Backpatch-through: 15
Commit 7102070329 fixed a similar bug, but
it missed the case of database-wide ANALYZE ("use_own_xacts" mode).
Commit a07e03fd8f changed consequences
from silent discard of a pg_class stats (relpages et al.) update to
ERROR "tuple to be updated was already modified". Losing a relpages
update of an ON COMMIT DELETE ROWS table was negligible, but a
COMMIT-time error isn't negligible. Back-patch to v13 (all supported
versions).
Reported-by: Richard Guo <guofenglinux@gmail.com
Reported-by: Robins Tharakan <tharakan@gmail.com>
Discussion: https://postgr.es/m/CAMbWs4-XwMKMKJ_GT=p3_-_=j9rQSEs1FbDFUnW9zHuKPsPNEQ@mail.gmail.com
Backpatch-through: 13
If a GENERATED column is declared to have a domain data type where
the domain's constraints disallow null values, INSERT commands failed
because we built a targetlist that included coercing a null constant
to the domain's type. The failure occurred even when the generated
value would have been perfectly OK. This is adjacent to the issues
fixed in 0da39aa76, but we didn't notice for lack of testing a domain
with such a constraint.
We aren't going to use the result of the targetlist entry for the
generated column --- ExecComputeStoredGenerated will overwrite it.
So it's not really necessary that it have the exact datatype of
the generated column. This patch fixes the problem by changing
the targetlist entry to be a null Const of the domain's base type,
which should be sufficiently legal. (We do have to tweak
ExecCheckPlanOutput to accept the situation, though.)
This has been broken since we implemented generated columns.
However, this patch only applies easily as far back as v14, partly
because I (tgl) only carried 0da39aa76 back that far, but mostly
because v14 significantly refactored the handling of INSERT/UPDATE
targetlists. Given the lack of field complaints and the short
remaining support lifetime of v13, I judge the cost-benefit ratio
not good for devising a version that would work in v13.
Reported-by: jian he <jian.universality@gmail.com>
Author: jian he <jian.universality@gmail.com>
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/CACJufxG59tip2+9h=rEv-ykOFjt0cbsPVchhi0RTij8bABBA0Q@mail.gmail.com
Backpatch-through: 14
Commit 0f21db36d made an assumption that GIN triConsistentFns
would not modify their input entryRes[] arrays. But in fact,
the "shim" triConsistentFn that we use for opclasses that don't
supply their own did exactly that, potentially leading to wrong
answers from a GIN index search. Through bad luck, none of the
test cases that we have for such opclasses exposed the bug.
One response to this could be that the assumption of consistency check
functions not modifying entryRes[] arrays is a bad one, but it still
seems reasonable to me. Notably, shimTriConsistentFn is itself
assuming that with respect to the underlying boolean consistentFn,
so it's sure being self-centered in supposing that it gets to do so.
Fortunately, it's quite simple to fix shimTriConsistentFn to restore
the entry-time state of entryRes[], so let's do that instead.
This issue doesn't affect any core GIN opclasses, since they all
supply their own triConsistentFns. It does affect contrib modules
btree_gin, hstore, and intarray.
Along the way, I (tgl) noticed that shimTriConsistentFn failed to
pick up on a "recheck" flag returned by its first call to the boolean
consistentFn. This may be only a latent problem, since it would be
unlikely for a consistentFn to set recheck for the all-false case
and not any other cases. (Indeed, none of our contrib modules do
that.) Nonetheless, it's formally wrong.
Reported-by: Vinod Sridharan <vsridh90@gmail.com>
Author: Vinod Sridharan <vsridh90@gmail.com>
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/CAFMdLD7XzsXfi1+DpTqTgrD8XU0i2C99KuF=5VHLWjx4C1pkcg@mail.gmail.com
Backpatch-through: 13
synchronous_standby_names cannot be reloaded safely by backends, and the
checkpointer is in charge of updating a state in shared memory if the
GUC is enabled in WalSndCtl, to let the backends know if they should
wait or not for a given LSN. This provides a strict control on the
timing of the waiting queues if the GUC is enabled or disabled, then
reloaded. The checkpointer is also in charge of waking up the backends
that could be waiting for a LSN when the GUC is disabled.
This logic had a race condition at startup, where it would be possible
for backends to not wait for a LSN even if synchronous_standby_names is
enabled. This would cause visibility issues with transactions that we
should be waiting for but they were not. The problem lasts until the
checkpointer does its initial update of the shared memory state when it
loads synchronous_standby_names.
In order to take care of this problem, the shared memory state in
WalSndCtl is extended to detect if it has been initialized by the
checkpointer, and not only check if synchronous_standby_names is
defined. In WalSndCtlData, sync_standbys_defined is renamed to
sync_standbys_status, a bits8 able to know about two states:
- If the shared memory state has been initialized. This flag is set by
the checkpointer at startup once, and never removed.
- If synchronous_standby_names is known as defined in the shared memory
state. This is the same as the previous sync_standbys_defined in
WalSndCtl.
This method gives a way for backends to decide what they should do until
the shared memory area is initialized, and they now ultimately fall back
to a check on the GUC value in this case, which is the best thing that
can be done.
Fortunately, SyncRepUpdateSyncStandbysDefined() is called immediately by
the checkpointer when this process starts, so the window is very narrow.
It is possible to enlarge the problematic window by making the
checkpointer wait at the beginning of SyncRepUpdateSyncStandbysDefined()
with a hardcoded sleep for example, and doing so has showed that a 2PC
visibility test is indeed failing. On machines slow enough, this bug
would cause spurious failures.
In 17~, we have looked at the possibility of adding an injection point
to have a reproducible test, but as the problematic window happens at
early startup, we would need to invent a way to make an injection point
optionally persistent across restarts when attached, something that
would be fine for this case as it would involve the checkpointer. This
issue is quite old, and can be reproduced on all the stable branches.
Author: Melnikov Maksim <m.melnikov@postgrespro.ru>
Co-authored-by: Michael Paquier <michael@paquier.xyz>
Discussion: https://postgr.es/m/163fcbec-900b-4b07-beaa-d2ead8634bec@postgrespro.ru
Backpatch-through: 13
Data loss can happen when the DDLs like ALTER PUBLICATION ... ADD TABLE ...
or ALTER TYPE ... that don't take a strong lock on table happens
concurrently to DMLs on the tables involved in the DDL. This happens
because logical decoding doesn't distribute invalidations to concurrent
transactions and those transactions use stale cache data to decode the
changes. The problem becomes bigger because we keep using the stale cache
even after those in-progress transactions are finished and skip the
changes required to be sent to the client.
This commit fixes the issue by distributing invalidation messages from
catalog-modifying transactions to all concurrent in-progress transactions.
This allows the necessary rebuild of the catalog cache when decoding new
changes after concurrent DDL.
We observed performance regression primarily during frequent execution of
*publication DDL* statements that modify the published tables. The
regression is minor or nearly nonexistent for DDLs that do not affect the
published tables or occur infrequently, making this a worthwhile cost to
resolve a longstanding data loss issue.
An alternative approach considered was to take a strong lock on each
affected table during publication modification. However, this would only
address issues related to publication DDLs (but not the ALTER TYPE ...)
and require locking every relation in the database for publications
created as FOR ALL TABLES, which is impractical.
The bug exists in all supported branches, but we are backpatching till 14.
The fix for 13 requires somewhat bigger changes than this fix, so the fix
for that branch is still under discussion.
Reported-by: hubert depesz lubaczewski <depesz@depesz.com>
Reported-by: Tomas Vondra <tomas.vondra@enterprisedb.com>
Author: Shlok Kyal <shlok.kyal.oss@gmail.com>
Author: Hayato Kuroda <kuroda.hayato@fujitsu.com>
Reviewed-by: Zhijie Hou <houzj.fnst@fujitsu.com>
Reviewed-by: Masahiko Sawada <sawada.mshk@gmail.com>
Reviewed-by: Amit Kapila <amit.kapila16@gmail.com>
Tested-by: Benoit Lobréau <benoit.lobreau@dalibo.com>
Backpatch-through: 14
Discussion: https://postgr.es/m/de52b282-1166-1180-45a2-8d8917ca74c6@enterprisedb.com
Discussion: https://postgr.es/m/CAD21AoAenVqiMjpN-PvGHL1N9DWnHSq673bfgr6phmBUzx=kLQ@mail.gmail.com
