Most GUC check hooks that inspect database state have special checks
that prevent them from throwing hard errors for state-dependent issues
when source == PGC_S_TEST. This allows, for example,
"ALTER DATABASE d SET default_text_search_config = foo" when the "foo"
configuration hasn't been created yet. Without this, we have problems
during dump/reload or pg_upgrade, because pg_dump has no idea about
possible dependencies of GUC values and can't ensure a safe restore
ordering.
However, check_role() and check_session_authorization() hadn't gotten
the memo about that, and would throw hard errors anyway. It's not
entirely clear what is the use-case for "ALTER ROLE x SET role = y",
but we've now heard two independent complaints about that bollixing
an upgrade, so apparently some people are doing it.
Hence, fix these two functions to act more like other check hooks
with similar needs. (But I did not change their insistence on
being inside a transaction, as it's still not apparent that setting
either GUC from the configuration file would be wise.)
Also fix check_temp_buffers, which had a different form of the disease
of making state-dependent checks without any exception for PGC_S_TEST.
A cursory survey of other GUC check hooks did not find any more issues
of this ilk. (There are a lot of interdependencies among
PGC_POSTMASTER and PGC_SIGHUP GUCs, which may be a bad idea, but
they're not relevant to the immediate concern because they can't be
set via ALTER ROLE/DATABASE.)
Per reports from Charlie Hornsby and Nathan Bossart. Back-patch
to all supported branches.
Discussion: https://postgr.es/m/HE1P189MB0523B31598B0C772C908088DB7709@HE1P189MB0523.EURP189.PROD.OUTLOOK.COM
Discussion: https://postgr.es/m/20160711223641.1426.86096@wrigleys.postgresql.org
Previously, get_cached_rowtype() cached a pointer to a reference-counted
tuple descriptor from the typcache, relying on the ExprContextCallback
mechanism to release the tupdesc refcount when the expression tree
using the tupdesc was destroyed. This worked fine when it was designed,
but the introduction of within-DO-block COMMITs broke it. The refcount
is logged in a transaction-lifespan resource owner, but plpgsql won't
destroy simple expressions made within the DO block (before its first
commit) until the DO block is exited. That results in a warning about
a leaked tupdesc refcount when the COMMIT destroys the original resource
owner, and then an error about the active resource owner not holding a
matching refcount when the expression is destroyed.
To fix, get rid of the need to have a shutdown callback at all, by
instead caching a pointer to the relevant typcache entry. Those
survive for the life of the backend, so we needn't worry about the
pointer becoming stale. (For registered RECORD types, we can still
cache a pointer to the tupdesc, knowing that it won't change for the
life of the backend.) This mechanism has been in use in plpgsql
and expandedrecord.c since commit 4b93f5799, and seems to work well.
This change requires modifying the ExprEvalStep structs used by the
relevant expression step types, which is slightly worrisome for
back-patching. However, there seems no good reason for extensions
to be familiar with the details of these particular sub-structs.
Per report from Rohit Bhogate. Back-patch to v11 where within-DO-block
COMMITs became a thing.
Discussion: https://postgr.es/m/CAAV6ZkQRCVBh8qAY+SZiHnz+U+FqAGBBDaDTjF2yiKa2nJSLKg@mail.gmail.com
heap_update needs to clear any existing "all visible" flag on
the old tuple's page (and on the new page too, if different).
Per coding rules, to do this it must acquire pin on the appropriate
visibility-map page while not holding exclusive buffer lock;
which creates a race condition since someone else could set the
flag whenever we're not holding the buffer lock. The code is
supposed to handle that by re-checking the flag after acquiring
buffer lock and retrying if it became set. However, one code
path through heap_update itself, as well as one in its subroutine
RelationGetBufferForTuple, failed to do this. The end result,
in the unlikely event that a concurrent VACUUM did set the flag
while we're transiently not holding lock, is a non-recurring
"PANIC: wrong buffer passed to visibilitymap_clear" failure.
This has been seen a few times in the buildfarm since recent VACUUM
changes that added code paths that could set the all-visible flag
while holding only exclusive buffer lock. Previously, the flag
was (usually?) set only after doing LockBufferForCleanup, which
would insist on buffer pin count zero, thus preventing the flag
from becoming set partway through heap_update. However, it's
clear that it's heap_update not VACUUM that's at fault here.
What's less clear is whether there is any hazard from these bugs
in released branches. heap_update is certainly violating API
expectations, but if there is no code path that can set all-visible
without a cleanup lock then it's only a latent bug. That's not
100% certain though, besides which we should worry about extensions
or future back-patch fixes that could introduce such code paths.
I chose to back-patch to v12. Fixing RelationGetBufferForTuple
before that would require also back-patching portions of older
fixes (notably 0d1fe9f74), which is more code churn than seems
prudent to fix a hypothetical issue.
Discussion: https://postgr.es/m/2247102.1618008027@sss.pgh.pa.us
With the Oracle Developer Studio 12.6 compiler, #line directives alter
the current source file location for purposes of #include "..."
directives. Hence, a VPATH build failed with 'cannot find include file:
"specscanner.c"'. With two exceptions, parser-containing directories
already add "-I. -I$(srcdir)"; eliminate the exceptions. Back-patch to
9.6 (all supported versions).
Commit 6f38d4dac3 failed to heed a warning about the stability of the
value pointed to by "otid". The caller is allowed to pass in a pointer to
newtup->t_self, which will be updated during the execution of the
function. Instead, the SSI check should use the value we copy into
oldtup.t_self near the top of the function.
Not a live bug, because newtup->t_self doesn't really get updated until
a bit later, but it was confusing and broke the rule established by the
comment.
Back-patch to 13.
Reported-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/2689164.1618160085%40sss.pgh.pa.us
There are hacks in parse_coerce.c to push down a requested coercion
to below any CollateExpr that may appear. However, we did that even
if the requested data type is non-collatable, leading to an invalid
expression tree in which CollateExpr is applied to a non-collatable
type. The fix is just to drop the CollateExpr altogether, reasoning
that it's useless.
This bug is ten years old, dating to the original addition of
COLLATE support. The lack of field complaints suggests that there
aren't a lot of user-visible consequences. We noticed the problem
because it would trigger an assertion in DefineVirtualRelation if
the invalid structure appears as an output column of a view; however,
in a non-assert build, you don't see a crash just a (subtly incorrect)
complaint about applying collation to a non-collatable type. I found
that by putting the incorrect structure further down in a view, I could
make a view definition that would fail dump/reload, per the added
regression test case. But CollateExpr doesn't do anything at run-time,
so this likely doesn't lead to any really exciting consequences.
Per report from Yulin Pei. Back-patch to all supported branches.
Discussion: https://postgr.es/m/HK0PR01MB22744393C474D503E16C8509F4709@HK0PR01MB2274.apcprd01.prod.exchangelabs.com
Currently, parallel vacuum workers don't use any buffer access strategy.
We can fix it either by propagating the access strategy from a leader or
allow each worker to use BAS_VACUUM access strategy type. We don't see
much use in propagating this information from leader as both leader and
workers are supposed to use the same strategy. We might want to use a
different access strategy for leader and workers but that would be a
separate optimization not suitable for back-branches. This has been fixed
in HEAD as commit f6b8f19a08.
Author: Amit Kapila
Reviewed-by: Sawada Masahiko, Bharath Rupireddy
Discussion: https://postgr.es/m/CAA4eK1KbmJgRV2W3BbzRnKUSrukN7SbqBBriC4RDB5KBhopkGQ@mail.gmail.com
Using Roman numbers (via "RM" or "rm") for a conversion to calculate a
number of months has never considered the case of negative numbers,
where a conversion could easily cause out-of-bound memory accesses. The
conversions in themselves were not completely consistent either, as
specifying 12 would result in NULL, but it should mean XII.
This commit reworks the conversion calculation to have a more
consistent behavior:
- If the number of months and years is 0, return NULL.
- If the number of months is positive, return the exact month number.
- If the number of months is negative, do a backward calculation, with
-1 meaning December, -2 November, etc.
Reported-by: Theodor Arsenij Larionov-Trichkin
Author: Julien Rouhaud
Discussion: https://postgr.es/m/16953-f255a18f8c51f1d5@postgresql.org
backpatch-through: 9.6
Comment fixes are applied on HEAD, and documentation improvements are
applied on back-branches where needed.
Author: Justin Pryzby
Discussion: https://postgr.es/m/20210408164008.GJ6592@telsasoft.com
Backpatch-through: 9.6
The code in bringetbitmap() simply added the whole matching page range
to the TID bitmap, as determined by pages_per_range, even if some of the
pages were beyond the end of the heap. The query then might fail with
an error like this:
ERROR: could not open file "base/20176/20228.2" (target block
262144): previous segment is only 131021 blocks
In this case, the relation has 262093 pages (131072 and 131021 pages),
but we're trying to acess block 262144, i.e. first block of the 3rd
segment. At that point _mdfd_getseg() notices the preceding segment is
incomplete, and fails.
Hitting this in practice is rather unlikely, because:
* Most indexes use power-of-two ranges, so segments and page ranges
align perfectly (segment end is also a page range end).
* The table size has to be just right, with the last segment being
almost full - less than one page range from full segment, so that the
last page range actually crosses the segment boundary.
* Prefetch has to be enabled. The regular page access checks that
pages are not beyond heap end, but prefetch does not. On older
releases (before 12) the execution stops after hitting the first
non-existent page, so the prefetch distance has to be sufficient
to reach the first page in the next segment to trigger the issue.
Since 12 it's enough to just have prefetch enabled, the prefetch
distance does not matter.
Fixed by not adding non-existent pages to the TID bitmap. Backpatch
all the way back to 9.6 (BRIN indexes were introduced in 9.5, but that
release is EOL).
Backpatch-through: 9.6
Instead of writing a query to psql's stdin, which can cause a failure
where psql exits before writing, reporting a write failure with a broken
pipe, this changes the logic to use -c. This was not seen in the
buildfarm as no animals with a sensitive environment are running the
kerberos tests, but let's be safe.
HEAD is able to handle the situation as of 6d41dd0 for all the test
suites doing connection checks. f44b9b6 has fixed the same problem for
the LDAP tests.
Discussion: https://postgr.es/m/YGu7ceWAiSNQDgH5@paquier.xyz
Backpatch-through: 11
Maxim Orlov reported that the shutdown of standby server could result in
the following assertion failure. The cause of this issue was that,
when the shutdown caused the startup process to exit, recovery-time
transaction tracking was not shut down even if it's already initialized,
and some locks the tracked transactions were holding could not be released.
At this situation, if other process was invoked and the PGPROC entry that
the startup process used was assigned to it, it found such unreleased locks
and caused the assertion failure, during the initialization of it.
TRAP: FailedAssertion("SHMQueueEmpty(&(MyProc->myProcLocks[i]))"
This commit fixes this issue by making the startup process shut down
transaction tracking and release all locks, at the exit of it.
Back-patch to all supported branches.
Reported-by: Maxim Orlov
Author: Fujii Masao
Reviewed-by: Maxim Orlov
Discussion: https://postgr.es/m/ad4ce692cc1d89a093b471ab1d969b0b@postgrespro.ru
spg_box_quad_leaf_consistent unconditionally returned the leaf
datum as leafValue, even though in its usage for poly_ops that
value is of completely the wrong type.
In versions before 12, that was harmless because the core code did
nothing with leafValue in non-index-only scans ... but since commit
2a6368343, if we were doing a KNN-style scan, spgNewHeapItem would
unconditionally try to copy the value using the wrong datatype
parameters. Said copying is a waste of time and space if we're not
going to return the data, but it accidentally failed to fail until
I fixed the datatype confusion in ac9099fc1.
Hence, change spgNewHeapItem to not copy the datum unless we're
actually going to return it later. This saves cycles and dodges
the question of whether lossy opclasses are returning the right
type. Also change spg_box_quad_leaf_consistent to not return
data that might be of the wrong type, as insurance against
somebody introducing a similar bug into the core code in future.
It seems like a good idea to back-patch these two changes into
v12 and v13, although I'm afraid to change spgNewHeapItem's
mistaken idea of which datatype to use in those branches.
Per buildfarm results from ac9099fc1.
Discussion: https://postgr.es/m/3728741.1617381471@sss.pgh.pa.us
Command-line options, or previous "ALTER (ROLE|DATABASE) ...
SET ROLE ..." commands, can change the value of the default role
for a session. In the presence of one of these, RESET ROLE will
change the current user identifier to the default role rather
than the session user identifier. Fix the documentation to
reflect this reality. Backpatch to all supported versions.
Author: Nathan Bossart
Reviewed-By: Laurenz Albe, David G. Johnston, Joe Conway
Reported by: Nathan Bossart
Discussion: https://postgr.es/m/flat/925134DB-8212-4F60-8AB1-B1231D750CB4%40amazon.com
Backpatch-through: 9.6
pg_checksums uses two counters, total size and current size,
to calculate the progress. Previously the progress that
pg_checksums reported could not reach 100% at the end.
The cause of this issue was that the sizes of only pages excluding
new ones in each file were counted as the current size
while the size of each file is counted as the total size.
That is, the total size of all new pages could be reported
as the difference between the total size and current size.
This commit fixes this issue by making pg_checksums count
the sizes of all pages including new ones in each file as
the current size.
Back-patch to v12 where progress reporting was added to pg_checksums.
Reported-by: Shinya Kato
Author: Shinya Kato
Reviewed-by: Fujii Masao
Discussion: https://postgr.es/m/TYAPR01MB289656B1ACA0A5E7CAD07BE3C47A9@TYAPR01MB2896.jpnprd01.prod.outlook.com
The current instructions describing how to write the backup_label and
tablespace_map files are confusing. For example, opening a file in text
mode on Windows and copy-pasting the file's contents would result in a
failure at recovery because of the extra CRLF characters generated. The
documentation was not stating that clearly, and per discussion this is
not considered as a supported scenario.
This commit extends a bit the documentation to mention that it may be
required to open the file in binary mode before writing its data.
Reported-by: Wang Shenhao
Author: David Steele
Reviewed-by: Andrew Dunstan, Magnus Hagander
Discussion: https://postgr.es/m/8373f61426074f2cb6be92e02f838389@G08CNEXMBPEKD06.g08.fujitsu.local
Backpatch-through: 9.6
This test has been using a simple VACUUM with pg_relation_size() to
check if a relation gets physically truncated or not, but forgot the
fact that some concurrent activity, like checkpoint buffer writes, could
cause some pages to be skipped. The second test enabling
vacuum_truncate could fail, seeing a non-empty relation. The first test
would not have failed, but could finish by testing a behavior different
than the one aimed for. Both tests gain a FREEZE option, to make the
vacuums more aggressive and prevent page skips.
This is similar to the issues fixed in c2dc1a7.
Author: Arseny Sher
Reviewed-by: Masahiko Sawada
Discussion: https://postgr.es/m/87tuotr2hh.fsf@ars-thinkpad
backpatch-through: 12
Despite the clear comments pointing out that the duplicative code
segments in ReadHead() and _discoverArchiveFormat() needed to be
in sync, they were not: the latter did not bother to apply any of
the sanity checks in the former. We'd missed noticing this partly
because none of those checks would fail in scenarios we customarily
test, and partly because the oversight would be masked if both
segments execute, which they would in cases other than needing to
autodetect the format of a non-seekable stdin source. However,
in a case meeting all these requirements --- for example, trying
to read a newer-than-supported archive format from non-seekable
stdin --- pg_restore missed applying the version check and would
likely dump core or otherwise misbehave.
The whole thing is silly anyway, because there seems little reason
to duplicate the logic beyond the one-line verification that the
file starts with "PGDMP". There seems to have been an undocumented
assumption that multiple major formats (major enough to require
separate reader modules) would nonetheless share the first half-dozen
fields of the custom-format header. This seems unlikely, so let's
fix it by just nuking the duplicate logic in _discoverArchiveFormat().
Also get rid of the pointless attempt to seek back to the start of
the file after successful autodetection. That wastes cycles and
it means we have four behaviors to verify not two.
Per bug #16951 from Sergey Koposov. This has been broken for
decades, so back-patch to all supported versions.
Discussion: https://postgr.es/m/16951-a4dd68cf0de23048@postgresql.org
Some sections of the documentation used "exclusive lock" to describe
that an ACCESS EXCLUSIVE lock is taken during a given operation. This
can be confusing to the reader as ACCESS SHARE is allowed with an
EXCLUSIVE lock is used, but that would not be the case with what is
described on those parts of the documentation.
Author: Greg Rychlewski
Discussion: https://postgr.es/m/CAKemG7VptD=7fNWckFMsMVZL_zzvgDO6v2yVmQ+ZiBfc_06kCQ@mail.gmail.com
Backpatch-through: 9.6
The new appendix groups information on renamed or removed settings,
commands, etc into an out-of-the-way part of the docs.
The original id elements are retained in each subsection to ensure that
the same filenames are produced for HTML docs. This prevents /current/
links on the web from breaking, and allows users of the web docs
to follow links from old version pages to info on the changes in the
new version. Prior to this change, a link to /current/ for renamed
sections like the recovery.conf docs would just 404. Similarly if
someone searched for recovery.conf they would find the pg11 docs,
but there would be no /12/ or /current/ link, so they couldn't easily
find out that it was removed in pg12 or how to adapt.
Index entries are also added so that there's a breadcrumb trail for
users to follow when they know the old name, but not what we changed it
to. So a user who is trying to find out how to set standby_mode in
PostgreSQL 12+, or where pg_resetxlog went, now has more chance of
finding that information.
Craig Ringer and Stephen Frost
Reviewed-by: Euler Taveira
Discussion: https://postgr.es/m/CAGRY4nzPNOyYQ_1-pWYToUVqQ0ThqP5jdURnJMZPm539fdizOg%40mail.gmail.com
Backpatch-through: 10
We were never doing clearerr() on the output stream, which results in a
message being printed after each result once an EOF is seen:
could not print result table: Success
This message was added by commit b03436994b (in the pg13 era); before
that, the error indicator would never be examined. So backpatch only
that far back, even though the actual bug (to wit: the fact that the
error indicator is never cleared) is older.
When estimating the number of groups using extended statistics, the code
was discarding information about system attributes. This led to strange
situation that
SELECT 1 FROM t GROUP BY ctid;
could have produced higher estimate (equal to pg_class.reltuples) than
SELECT 1 FROM t GROUP BY a, b, ctid;
with extended statistics on (a,b). Fixed by retaining information about
the system attribute.
Backpatch all the way to 10, where extended statistics were introduced.
Author: Tomas Vondra
Backpatch-through: 10
On partition detach, we acquire a SHARE lock on all tables that
reference the partitioned table that we're detaching a partition from,
but failed to document this fact. My oversight in commit f56f8f8da6.
Repair. Backpatch to 12.
Author: Álvaro Herrera <alvherre@alvh.no-ip.org>
Discussion: https://postgr.es/m/20210325180244.GA12738@alvherre.pgsql
I introduced this duplicate code in commit 8b08f7d482 for no good
reason. Remove it, and backpatch to 11 where it was introduced.
Author: Álvaro Herrera <alvherre@alvh.no-ip.org>
Previously the WAL replay of COMMIT_TS_SETTS record called
TransactionTreeSetCommitTsData() with the argument write_xlog=true,
which generated and wrote new COMMIT_TS_SETTS record.
This should not be acceptable because it's during recovery.
This commit fixes the WAL replay of COMMIT_TS_SETTS record
so that it calls TransactionTreeSetCommitTsData() with write_xlog=false
and doesn't generate new WAL during recovery.
Back-patch to all supported branches.
Reported-by: lx zou <zoulx1982@163.com>
Author: Fujii Masao
Reviewed-by: Alvaro Herrera
Discussion: https://postgr.es/m/16931-620d0f2fdc6108f1@postgresql.org
Jasen Betts reported yet another unintended side effect of commit
85c54287a: reconnecting with "\c service=whatever" did not have the
expected results. The reason is that starting from the output of
PQconndefaults() effectively allows environment variables (such
as PGPORT) to override entries in the service file, whereas the
normal priority is the other way around.
Not using PQconndefaults at all would require yet a third main code
path in do_connect's parameter setup, so I don't really want to fix
it that way. But we can have the logic effectively ignore all the
default values for just a couple more lines of code.
This patch doesn't change the behavior for "\c -reuse-previous=on
service=whatever". That remains significantly different from before
85c54287a, because many more parameters will be re-used, and thus
not be possible for service entries to replace. But I think this
is (mostly?) intentional. In any case, since libpq does not report
where it got parameter values from, it's hard to do differently.
Per bug #16936 from Jasen Betts. As with the previous patches,
back-patch to all supported branches. (9.5 is unfortunately now
out of support, so this won't get fixed there.)
Discussion: https://postgr.es/m/16936-3f524322a53a29f0@postgresql.org
end_heap_rewrite was not careful to ensure that the target relation
is open at the smgr level before performing its final smgrimmedsync.
In ordinary cases this is no problem, because it would have been
opened earlier during the rewrite. However a crash can be reproduced
by re-clustering an empty table with CLOBBER_CACHE_ALWAYS enabled.
Although that exact scenario does not crash in v13, I think that's
a chance result of unrelated planner changes, and the problem is
likely still reachable with other test cases. The true proximate
cause of this failure is commit c6b92041d, which replaced a call to
heap_sync (which was careful about opening smgr) with a direct call
to smgrimmedsync. Hence, back-patch to v13.
Amul Sul, per report from Neha Sharma; cosmetic changes
and test case by me.
Discussion: https://postgr.es/m/CANiYTQsU7yMFpQYnv=BrcRVqK_3U3mtAzAsJCaqtzsDHfsUbdQ@mail.gmail.com
A couple error messages and comments used 'statistic kind', not the
correct 'statistics kind'. Fix and backpatch all the way back to 10,
where extended statistics were introduced.
Backpatch-through: 10
pg_waldump --stats=record identifies a record by a combination
of the RmgrId and the four bits of the xl_info field of the record.
But XACT records use the first bit of those four bits for an optional
flag variable, and the following three bits for the opcode to
identify a record. So previously the same type of XACT record
could have different four bits (three bits are the same but the
first one bit is different), and which could cause
pg_waldump --stats=record to show two lines of per-record statistics
for the same XACT record. This is a bug.
This commit changes pg_waldump --stats=record so that it processes
only XACT record differently, i.e., filters the opcode out of xl_info
and uses a combination of the RmgrId and those three bits as
the identifier of a record, only for XACT record. For other records,
the four bits of the xl_info field are still used.
Back-patch to all supported branches.
Author: Kyotaro Horiguchi
Reviewed-by: Shinya Kato, Fujii Masao
Discussion: https://postgr.es/m/2020100913412132258847@highgo.ca
This commit is mostly a revert of aaa3aed, that switched the routine
doing the internal renaming of recycled WAL segments to use on Windows a
combination of CreateHardLinkA() plus unlink() instead of rename(). As
reported by several users of Postgres 13, this is causing concurrency
issues when manipulating WAL segments, mostly in the shape of the
following error:
LOG: could not rename file "pg_wal/000000XX000000YY000000ZZ":
Permission denied
This moves back to a logic where a single rename() (well, pgrename() for
Windows) is used. This issue has proved to be hard to hit when I tested
it, facing it only once with an archive_command that was not able to do
its work, so it is environment-sensitive. The reporters of this issue
have been able to confirm that the situation improved once we switched
back to a single rename(). In order to check things, I have provided to
the reporters a patched build based on 13.2 with aaa3aed reverted, to
test if the error goes away, and an unpatched build of 13.2 to test if
the error still showed up (just to make sure that I did not mess up my
build process).
Extra thanks to Fujii Masao for pointing out what looked like the
culprit commit, and to all the reporters for taking the time to test
what I have sent them.
Reported-by: Andrus, Guy Burgess, Yaroslav Pashinsky, Thomas Trenz
Reviewed-by: Tom Lane, Andres Freund
Discussion: https://postgr.es/m/3861ff1e-0923-7838-e826-094cc9bef737@hot.ee
Discussion: https://postgr.es/m/16874-c3eecd319e36a2bf@postgresql.org
Discussion: https://postgr.es/m/095ccf8d-7f58-d928-427c-b17ace23cae6@burgess.co.nz
Discussion: https://postgr.es/m/16927-67c570d968c99567%40postgresql.org
Discussion: https://postgr.es/m/YFBcRbnBiPdGZvfW@paquier.xyz
Backpatch-through: 13
The test added by 595b9cb forgot that on Windows it is necessary to set
up pg_hba.conf (see PostgresNode::set_replication_conf) with a specific
entry or base backups fail. Any node that requires to support
replication just needs to pass down allows_streaming at initialization.
This updates the test to do so. Simplify things a bit while on it.
Per buildfarm member fairywren. Any Windows hosts running this test
would have failed, and I have reproduced the problem as well.
Backpatch-through: 10
Any transactions found as still prepared by a checkpoint have their
state data read from the WAL records generated by PREPARE TRANSACTION
before being moved into their new location within pg_twophase/. While
reading such records, the WAL reader uses the callback
read_local_xlog_page() to read a page, that is shared across various
parts of the system. This callback, since 1148e22a, has introduced an
update of ThisTimeLineID when reading a record while in recovery, which
is potentially helpful in the context of cascading WAL senders.
This update of ThisTimeLineID interacts badly with the checkpointer if a
promotion happens while some 2PC data is read from its record, as, by
changing ThisTimeLineID, any follow-up WAL records would be written to
an timeline older than the promoted one. This results in consistency
issues. For instance, a subsequent server restart would cause a failure
in finding a valid checkpoint record, resulting in a PANIC, for
instance.
This commit changes the code reading the 2PC data to reset the timeline
once the 2PC record has been read, to prevent messing up with the static
state of the checkpointer. It would be tempting to do the same thing
directly in read_local_xlog_page(). However, based on the discussion
that has led to 1148e22a, users may rely on the updates of
ThisTimeLineID when a WAL record page is read in recovery, so changing
this callback could break some cases that are working currently.
A TAP test reproducing the issue is added, relying on a PITR to
precisely trigger a promotion with a prepared transaction still
tracked.
Per discussion with Heikki Linnakangas, Kyotaro Horiguchi, Fujii Masao
and myself.
Author: Soumyadeep Chakraborty, Jimmy Yih, Kevin Yeap
Discussion: https://postgr.es/m/CAE-ML+_EjH_fzfq1F3RJ1=XaaNG=-Jz-i3JqkNhXiLAsM3z-Ew@mail.gmail.com
Backpatch-through: 10
While back-patching e0e569e1d, I noted that there were some other
places where we ought to be applying DH_free(); namely, where we
load some DH parameters from a file and then reject them as not
being sufficiently secure. While it seems really unlikely that
anybody would hit these code paths in production, let alone do
so repeatedly, let's fix it for consistency.
Back-patch to v10 where this code was introduced.
Discussion: https://postgr.es/m/16160-18367e56e9a28264@postgresql.org
We leaked the error report from PQconninfoParse, when there was
one. It seems unlikely that real usage patterns would repeat
the failure often enough to create serious bloat, but let's
back-patch anyway to keep the code similar in all branches.
Found via valgrind testing.
Back-patch to v10 where this code was added.
Discussion: https://postgr.es/m/3816764.1616104288@sss.pgh.pa.us
Because guc.c prefers to keep all its string values in malloc'd
not palloc'd storage, it has to be more careful than usual to
avoid leaks. Error exits out of string GUC hook checks failed
to clear the proposed value string, and error exits out of
ProcessGUCArray() failed to clear the malloc'd results of
ParseLongOption().
Found via valgrind testing.
This problem is ancient, so back-patch to all supported branches.
Discussion: https://postgr.es/m/3816764.1616104288@sss.pgh.pa.us
The text search cache mechanisms assume that we can clean up
an invalidated dictionary cache entry simply by resetting the
associated long-lived memory context. However, that does not work
for ispell affixes that make use of regular expressions, because
the regex library deals in plain old malloc. Hence, we leaked
compiled regex(es) any time we dropped such a cache entry. That
could quickly add up, since even a fairly trivial regex can use up
tens of kB, and a large one can eat megabytes. Add a memory context
callback to ensure that a regex gets freed when its owning cache
entry is cleared.
Found via valgrind testing.
This problem is ancient, so back-patch to all supported branches.
Discussion: https://postgr.es/m/3816764.1616104288@sss.pgh.pa.us
Some code paths in this function perform syscache lookups, which
can lead to table accesses and possibly leakage of cruft into
the caller's context. If said context is CacheMemoryContext,
we eventually will have visible bloat. But fixing this is no
harder than moving one memory context switch step. (The other
callers don't have a problem.)
Andres Freund and I independently found this via valgrind testing.
Back-patch to v12 where this code was added.
Discussion: https://postgr.es/m/20210317023101.anvejcfotwka6gaa@alap3.anarazel.de
Discussion: https://postgr.es/m/3816764.1616104288@sss.pgh.pa.us
Although these lists are usually NIL, and even when not empty
are unlikely to be large, constant relcache update traffic could
eventually result in visible bloat of CacheMemoryContext.
Found via valgrind testing.
Back-patch to v10 where this field was added.
Discussion: https://postgr.es/m/3816764.1616104288@sss.pgh.pa.us
pg_read_file() is the function that's in core, pg_file_read() is in
adminpack. But when using pg_file_read() in adminpack it calls the *C*
level function pg_read_file() in core, which probably threw the original
author off. But the error hint should be about the SQL function.
Reported-By: Sergei Kornilov
Backpatch-through: 11
Discussion: https://postgr.es/m/373021616060475@mail.yandex.ru
Robert Foggia of Trustwave reported that read_tablespace_map()
fails to prevent an overrun of its on-stack input buffer.
Since the tablespace map file is presumed trustworthy, this does
not seem like an interesting security vulnerability, but still
we should fix it just in the name of robustness.
While here, document that pg_basebackup's --tablespace-mapping option
doesn't work with tar-format output, because it doesn't. To make it
work, we'd have to modify the tablespace_map file within the tarball
sent by the server, which might be possible but I'm not volunteering.
(Less-painful solutions would require changing the basebackup protocol
so that the source server could adjust the map. That's not very
appetizing either.)
With very unlucky timing and parallel_leader_participation off, PHJ
could attempt to access per-batch state just as it was being freed.
There was code intended to prevent that by checking for a cleared
pointer, but it was buggy.
Fix, by introducing an extra barrier phase. The new phase
PHJ_BUILD_RUNNING means that it's safe to access the per-batch state to
find a batch to help with, and PHJ_BUILD_DONE means that it is too late.
The last to detach will free the array of per-batch state as before, but
now it will also atomically advance the phase at the same time, so that
late attachers can avoid the hazard, without the data race. This
mirrors the way per-batch hash tables are freed (see phases
PHJ_BATCH_PROBING and PHJ_BATCH_DONE).
Revealed by a one-off build farm failure, where BarrierAttach() failed a
sanity check assertion, because the memory had been clobbered by
dsa_free().
Back-patch to 11, where the code arrived.
Reported-by: Michael Paquier <michael@paquier.xyz>
Discussion: https://postgr.es/m/20200929061142.GA29096%40paquier.xyz