xl_tot_len comes first in a WAL record. Usually we don't trust it to be
the true length until we've validated the record header. If the record
header was split across two pages, previously we wouldn't do the
validation until after we'd already tried to allocate enough memory to
hold the record, which was bad because it might actually be garbage
bytes from a recycled WAL file, so we could try to allocate a lot of
memory. Release 15 made it worse.
Since 70b4f82a4b, we'd at least generate an end-of-WAL condition if the
garbage 4 byte value happened to be > 1GB, but we'd still try to
allocate up to 1GB of memory bogusly otherwise. That was an
improvement, but unfortunately release 15 tries to allocate another
object before that, so you could get a FATAL error and recovery could
fail.
We can fix both variants of the problem more fundamentally using
pre-existing page-level validation, if we just re-order some logic.
The new order of operations in the split-header case defers all memory
allocation based on xl_tot_len until we've read the following page. At
that point we know that its first few bytes are not recycled data, by
checking its xlp_pageaddr, and that its xlp_rem_len agrees with
xl_tot_len on the preceding page. That is strong evidence that
xl_tot_len was truly the start of a record that was logged.
This problem was most likely to occur on a standby, because
walreceiver.c recycles WAL files without zeroing out trailing regions of
each page. We could fix that too, but it wouldn't protect us from rare
crash scenarios where the trailing zeroes don't make it to disk.
With reliable xl_tot_len validation in place, the ancient policy of
considering malloc failure to indicate corruption at end-of-WAL seems
quite surprising, but changing that is left for later work.
Also included is a new TAP test to exercise various cases of end-of-WAL
detection by writing contrived data into the WAL from Perl.
Back-patch to 12. We decided not to put this change into the final
release of 11.
Author: Thomas Munro <thomas.munro@gmail.com>
Author: Michael Paquier <michael@paquier.xyz>
Reported-by: Alexander Lakhin <exclusion@gmail.com>
Reviewed-by: Noah Misch <noah@leadboat.com> (the idea, not the code)
Reviewed-by: Michael Paquier <michael@paquier.xyz>
Reviewed-by: Sergei Kornilov <sk@zsrv.org>
Reviewed-by: Alexander Lakhin <exclusion@gmail.com>
Discussion: https://postgr.es/m/17928-aa92416a70ff44a2%40postgresql.org
The previous wording had a faint archaic whiff to it, and more
importantly used "catalogs" as a verb, which while cutely
self-referential seems likely to provoke confusion in this
particular context. Also consistently use "kind" not "type" to
refer to the different kinds of relations distinguished by relkind.
Per gripe from Martin Nash. Back-patch to supported versions.
Discussion: https://postgr.es/m/169518739902.3727338.4793815593763320945@wrigleys.postgresql.org
The SSL tests for pg_ctl restart with an incorrect key passphrase used
the internal _update_pid method to set the pidfile after running pg_ctl
manually instead of using the supplied ->restart method. This refactors
the ->restart method to accept a fail_ok parameter like how ->start and
->stop does, and changes the SSL tests to use this instead. This removes
the need to call internal test module functions.
Reviewed-by: Melih Mutlu <m.melihmutlu@gmail.com>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/F81643C4-D7B8-4C6B-AF18-B73839966279@yesql.se
Guard against the pointer being NULL before pfreeing upon an error
returned from OpenSSL. Also handle errors from X509_NAME_print_ex
which can return -1 on memory allocation errors.
Backpatch down to v15 where the code was added.
Author: Sergey Shinderuk <s.shinderuk@postgrespro.ru>
Discussion: https://postgr.es/m/8db5374d-32e0-6abb-d402-40762511eff2@postgrespro.ru
Backpatch-through: v15
The computation of the column
information_schema.check_constraints.check_clause used
pg_get_constraintdef() plus some string manipulation to get the check
clause back out. This ended up with an extra pair of parentheses,
which is only an aesthetic problem, but also with suffixes like "NOT
VALID", which don't belong into that column. We can fix both of these
problems and simplify the code by just using pg_get_expr() instead.
Discussion: https://www.postgresql.org/message-id/799b59ef-3330-f0d2-ee23-8cdfa1740987@eisentraut.org
In older branches, COMMIT/ROLLBACK AND CHAIN failed to propagate
the current transaction's properties to the new transaction if
there was any open subtransaction (unreleased savepoint).
Instead, some previous transaction's properties would be restored.
This is because the "if (s->chain)" check in CommitTransactionCommand
examined the wrong instance of the "chain" flag and falsely
concluded that it didn't need to save transaction properties.
Our regression tests would have noticed this, except they used
identical transaction properties for multiple tests in a row,
so that the faulty behavior was not distinguishable from correct
behavior.
Commit 12d768e70 fixed the problem in v15 and later, but only rather
accidentally, because I removed the "if (s->chain)" test to avoid a
compiler warning, while not realizing that the warning was flagging a
real bug.
In v14 and before, remove the if-test and save transaction properties
unconditionally; just as in the newer branches, that's not expensive
enough to justify thinking harder.
Add the comment and extra regression test to v15 and later to
forestall any future recurrence, but there's no live bug in those
branches.
Patch by me, per bug #18118 from Liu Xiang. Back-patch to v12 where
the AND CHAIN feature was added.
Discussion: https://postgr.es/m/18118-4b72fcbb903aace6@postgresql.org
As physical replication work at the cluster level and not database
level, any dbname in the connection string is ignored. Proxies and
middleware used in connecting to the cluster might however need to
know the dbname in order to make the correct routing decision for
the connection.
With this the startup packet will include the dbname parameter.
Author: Jelte Fennema-Nio <me@jeltef.nl>
Reviewed-by: Tristen Raab <tristen.raab@highgo.ca>
Reviewed-by: Jim Jones <jim.jones@uni-muenster.de>
Discussion: https://postgr.es/m/CAGECzQTw-dZkVT_RELRzfWRzY714-VaTjoBATYfZq93R8C-auA@mail.gmail.com
The comment introduced by commit e7cb7ee14 was a bit too terse, which
could lead to extensions doing different things within the hook function
than we intend to allow. Extend the comment to explain what they can do
within the hook function.
Back-patch to all supported branches.
In passing, I rephrased a nearby comment that I recently added to the
back branches.
Reviewed by David Rowley and Andrei Lepikhov.
Discussion: https://postgr.es/m/CAPmGK15SBPA1nr3Aqsdm%2BYyS-ay0Ayo2BRYQ8_A2To9eLqwopQ%40mail.gmail.com
ae78cae3b added the --buffer-usage-limit to vacuumdb to allow it to
include the BUFFER_USAGE_LIMIT option in the VACUUM command.
Unfortunately, that commit forgot to adjust the code so the option was
added to the ANALYZE command when the -Z command line argument was
specified.
There were no issues with the -z command as that option just adds
ANALYZE to the VACUUM command.
In passing adjust the code to escape the --buffer-usage-limit option
before passing it to the server. It seems nothing beyond a confusing
error message could become this lack of escaping as VACUUM cannot be
specified in a multi-command string.
Reported-by: Ryoga Yoshida
Author: Ryoga Yoshida, David Rowley
Discussion: https://postgr.es/m/08930c0b541700a5264e5fbf3a685f5a%40oss.nttdata.com
Backpatch-through: 16, where ae78cae3b was introduced.
As reported in bug #18057, the extension unaccent removes in its rule
file whitespace characters that are intentionally specified when
building unaccent.rules from UnicodeData.txt, causing an incorrect
translation for some characters like numeric symbols. This is caused by
the fact that all whitespaces before and after the origin and target
characters are all discarded (this limitation is documented).
This commit makes possible the use of quotes around target characters,
so as whitespaces can be considered part of target characters. Some
target characters use a double quote, these require an extra double
quote.
The documentation is updated to show how to use quoted areas,
generate_unaccent_rules.py is updated to generate unaccent.rules and a
couple of tests are added for numeric symbols. While working on this
patch, I have implemented a fake rule file to test the parsing logic
implemented, which is not included here as it would just consume extra
cycles in the tests, and it requires the manipulation of an installation
tree to be able to work correctly.
As this requires a change of format in unaccent.rules, this cannot be
backpatched, unfortunately. The idea to use double quotes as escaped
characters comes from Tom Lane.
Reported-by: Martin Schlossarek
Author: Michael Paquier
Discussion: https://postgr.es/m/18057-62712cad01bd202c@postgresql.org
Thanks to commit 5af0263afd, binaryheap is available to frontend
code. This commit replaces the open-coded heap implementation in
pg_dump_sort.c with a binaryheap, saving a few lines of code.
Reviewed-by: Tom Lane
Discussion: https://postgr.es/m/3612876.1689443232%40sss.pgh.pa.us
If any of these commands fail during editing or pre-processing, the
command stored in the query buffer would remain around without being
executed immediately as PSQL_CMD_ERROR is returned as status. The next
command provided by the user would run it, likely causing failures as
this could include silently some of the contents generated automatically
for views or functions.
The problems would be different depending on the psql meta-command used:
- For \ev and \ef, some errors can happen in a predictable way while
doing an object lookup or while creating an object command. A failure
while editing is equally problematic, but the class of failures
happening in the code path of do_edit() are unlikely. The query reset
is kept in exec_command_ef_ev() as a query may be unchanged.
- For \e, error can happen while editing.
In both cases, the query buffer is reset on error for an incorrect file
number provided, whose value check is done before filling up the query
buffer.
This is a slight change of behavior compared to the past for some of the
predictable error patterns for \ev and \ef, so for now I have made the
choice to not backpatch this commit (argument particularly available for
v11 that's going to be EOL'd soon). Perhaps this could be revisited
later depending on the feedback of this new behavior.
Author: Ryoga Yoshida, Michael Paquier
Reviewed-by: Aleksander Alekseev, Kyotaro Horiguchi
Discussion: https://postgr.es/m/01419622d84ef093fd4fe585520bf03c@oss.nttdata.com
Presently, parallel restores spend a lot of time sorting this list
so that we pick the largest items first. With many tables, this
sorting can become a significant bottleneck. There are a couple of
reports from the field about this, and it is easy to reproduce.
This commit improves the performance of parallel pg_restore with
many tables by converting its ready_list to a priority queue, i.e.,
a binary heap. We will first try to run the highest priority item,
but if it cannot be chosen due to the lock heuristic, we'll do a
sequential scan through the heap nodes until we find one that is
runnable. This means that we might end up picking an item with a
much lower priority. However, we expect that we will typically be
able to pick one of the first few items, which should usually have
a relatively high priority.
Suggested-by: Tom Lane
Tested-by: Pierre Ducroquet
Reviewed-by: Tom Lane
Discussion: https://postgr.es/m/3612876.1689443232%40sss.pgh.pa.us
When a snapshot file fails to be read in ImportSnapshot(), it would
issue an ERROR as "invalid snapshot identifier" when opening a stream
for it in read-only mode. The error handling is improved to be more
talkative in failure cases:
- If a snapshot identifier uses incorrect characters, complain with the
same error as before this commit.
- If the snapshot file cannot be found in pg_snapshots/, complain with a
"snapshot \"foo\" does not exist" instead. This maps to the case where
AllocateFile() fails on ENOENT. Based on a suggestion from Andres
Freund.
- If AllocateFile() throws something else than ENOENT as errno, report
it with more details in %m instead, as these failures are never
expected.
b29504eeb489 was the first improvement take. The older error message
exists since bb446b689b that introduced snapshot imports. Two test
cases are added to cover the cases of an identifier with an incorrect
format and of a missing snapshot.
Author: Bharath Rupireddy
Reviewed-by: Andres Freund, Daniel Gustafsson, Michael Paquier
Discussion: https://postgr.es/m/CALj2ACWmr=3KdxDkm8h7Zn1XxBoF6hdzq8WQyMn2y1OL5RYFrg@mail.gmail.com
PLy_elog() was not able to handle correctly cases where a SPI called
failed, which would fill in a DETAIL string able to trigger an
assertion. We may want to improve this infrastructure so as it is able
to provide any extra detail information provided by an error stack, but
this is left as a future improvement as it could impact existing error
stacks and any applications that depend on them. For now, the assertion
is removed and a regression test is added to cover the case of a failure
with a detail string.
This problem exists since 2bd78eb8d5, so backpatch all the way down
with tweaks to the regression tests output added where required.
Author: Alexander Lakhin
Discussion: https://postgr.es/m/18070-ab9c171cbf4ebb0f@postgresql.org
Backpatch-through: 11
This commit introduces binaryheap_remove_node(), which can be used
to remove any node from a binary heap. The implementation is
straightforward. The target node is replaced with the last node in
the heap, and then we sift as needed to preserve the heap property.
This new function is intended for use in a follow-up commit that
will improve the performance of pg_restore.
Reviewed-by: Tom Lane
Discussion: https://postgr.es/m/3612876.1689443232%40sss.pgh.pa.us
There are a couple of places in frontend code that could make use
of this simple binary heap implementation. This commit makes
binaryheap usable in frontend code, much like commit 26aaf97b68 did
for StringInfo. Like StringInfo, the header file is left in lib/
to reduce the likelihood of unnecessary breakage.
The frontend version of binaryheap exposes a void *-based API since
frontend code does not have access to the Datum definitions. This
seemed like a better approach than switching all existing uses to
void * or making the Datum definitions available to frontend code.
Reviewed-by: Tom Lane, Alvaro Herrera
Discussion: https://postgr.es/m/3612876.1689443232%40sss.pgh.pa.us
cursor_to_xmlschema() assumed that any Portal must have a tupDesc,
which is not so. Add a defensive check.
It's plausible that this mistake occurred because of the rather
poorly chosen name of the lookup function SPI_cursor_find(),
which in such cases is returning something that isn't very much
like a cursor. Add some documentation to try to forestall future
errors of the same ilk.
Report and patch by Boyu Yang (docs changes by me). Back-patch
to all supported branches.
Discussion: https://postgr.es/m/dd343010-c637-434c-a8cb-418f53bda3b8.yangboyu.yby@alibaba-inc.com
It was reported as misaligned by Kyotaro, but it also needed to be
turned into a single translatable phrase (like the one for \g is), as
reported by Yugo.
This is a new issue (commit f347ec76e2), so no backpatch is needed.
Author: Kyotaro Horiguchi <horikyota.ntt@gmail.com>
Author: Yugo NAGATA <nagata@sraoss.co.jp>
Discussion: https://postgr.es/m/20230907.142956.2038600444404289870.horikyota.ntt@gmail.com
The documentation is pretty light on how to set column options
on foreign tables, and the file_fdw docs refer to COPY when
documenting force_null even though it's not used in the same
way. Add a small example to describe how to use it.
Reported-by: Boshomi Phenix <boshomi@gmail.com>
Discussion: https://postgr.es/m/CAJVkCUparn4_Oarernm=U6LWVsTkecKcALHtwGr5M3qJRj_czw@mail.gmail.com
Between Unicode 15.0.0 and 15.1.0, the whitespace in
EastAsianWidth.txt has changed a bit, such as from
0020;Na # Zs SPACE
to
0020 ; Na # Zs SPACE
with space around the semicolon. Adjust the script to be able to
parse that.
expandRecordVariable() failed to adjust the parse nesting structure
correctly when recursing to inspect an outer-level Var. This could
result in assertion failures or core dumps in corner cases.
Likewise, get_name_for_var_field() failed to adjust the deparse
namespace stack correctly when recursing to inspect an outer-level
Var. In this case the likely result was a "bogus varno" error
while deparsing a view.
Per bug #18077 from Jingzhou Fu. Back-patch to all supported
branches.
Richard Guo, with some adjustments by me
Discussion: https://postgr.es/m/18077-b9db97c6e0ab45d8@postgresql.org
When tracking IO timing for WAL, the duration is what we calculate
based on the start and end timestamps, it's not what the variable
contains. Rename the timestamp variable to end to better communicate
what it contains. Original patch by Krishnakumar with additional
hacking to fix another occurrence by me.
Author: Krishnakumar R <kksrcv001@gmail.com>
Discussion: https://postgr.es/m/CAPMWgZ9f9o8awrQpjo8oxnNQ=bMDVPx00NE0QcDzvHD_ZrdLPw@mail.gmail.com
This became safe after commit 4b4798e138. The smgrcreate() call will
now register the segment for syncing at the next checkpoint, so we
don't need to sync it here. If a checkpoint happens before the
creation is WAL-logged, the records will be replayed when starting
recovery from the checkpoint. If a checkpoint happens after the WAL
logging, the checkpoint will fsync() it.
In the passing, clarify a comment in smgrDoPendingSyncs().
Discussion: https://www.postgresql.org/message-id/6e5bbc08-cdfc-b2b3-9e23-1a914b9850a9%40iki.fi
Reviewed-by: Robert Haas
The majority of all filenames are quoted in user facing error and
log messages, but a few were still printed without quotes. While
these filenames do not risk causing any ambiguity as their format
is strict, quote them anyways to be consistent across all logs.
Also concatenate a message to keep it one line to make it easier
to grep for in the code.
Reviewed-by: Peter Eisentraut <peter@eisentraut.org>
Reviewed-by: Michael Paquier <michael@paquier.xyz>
Discussion: https://postgr.es/m/080EEABE-6645-4A46-AB20-6285ADAC44FE@yesql.se
This reverts commit a0d87bcd9b, following a remark from Andres Frend
that the new error can be triggered with an incorrect SET TRANSACTION
SNAPSHOT command without being really helpful for the user as it uses
the internal file name.
Discussion: https://postgr.es/m/20230914020724.hlks7vunitvtbbz4@awork3.anarazel.de
Backpatch-through: 11
It's entirely possible for a logical slot to have a confirmed_flush LSN
higher than the last value saved on disk while not being marked as dirty.
Currently, it is not a major problem but a later patch adding support for
the upgrade of slots relies on that value being properly flushed to disk.
It can also help avoid processing the same transactions again in some
boundary cases after the clean shutdown and restart. Say, we process
some transactions for which we didn't send anything downstream (the
changes got filtered) but the confirm_flush LSN is updated due to
keepalives. As we don't flush the latest value of confirm_flush LSN, it
may lead to processing the same changes again without this patch.
The approach taken by this patch has been suggested by Ashutosh Bapat.
Author: Vignesh C, Julien Rouhaud, Kuroda Hayato
Reviewed-by: Amit Kapila, Dilip Kumar, Michael Paquier, Ashutosh Bapat, Peter Smith, Hou Zhijie
Discussion: http://postgr.es/m/CAA4eK1JzJagMmb_E8D4au=GYQkxox0AfNBm1FbP7sy7t4YWXPQ@mail.gmail.com
Discussion: http://postgr.es/m/TYAPR01MB58664C81887B3AF2EB6B16E3F5939@TYAPR01MB5866.jpnprd01.prod.outlook.com
Karina figured out that I (Andres) confused BufferUsage.temp_blks_written with
BufferUsage.local_blks_written in fcdda1e4b5.
Tests in core PG can't easily test this, as BufferUsage is just used for
EXPLAIN (ANALYZE, BUFFERS) and pg_stat_statements. Thus this commit adds tests
for this to pg_stat_statements.
Reported-by: Karina Litskevich <litskevichkarina@gmail.com>
Author: Karina Litskevich <litskevichkarina@gmail.com>
Author: Andres Freund <andres@anarazel.de>
Discussion: https://postgr.es/m/CACiT8ibxXA6+0amGikbeFhm8B84XdQVo6D0Qfd1pQ1s8zpsnxQ@mail.gmail.com
Backpatch: 16-, where fcdda1e4b5 was merged
When a snapshot file fails to be read in ImportSnapshot(), it would
issue an ERROR as "invalid snapshot identifier" when opening a stream
for it in read-only mode. This error message is reworded to be the same
as all the other messages used in this case on failure, which is useful
when debugging this area.
Thinko introduced by bb446b689b where snapshot imports have been
added. A backpatch down to 11 is done as this can improve any work
related to snapshot imports in older branches.
Author: Bharath Rupireddy
Reviewed-by: Daniel Gustafsson
Discussion: https://postgr.es/m/CALj2ACWmr=3KdxDkm8h7Zn1XxBoF6hdzq8WQyMn2y1OL5RYFrg@mail.gmail.com
Backpatch-through: 11
These code paths should not be reached normally, but if they are an
error with "(null)" as information for the collation provider would show
up if no locale is set, while we can assume that we are referring to
libc.
This refactors the code so as the provider is always reported even if no
locale is set. The name of the function where the error happens is
added, while on it, as it can be helpful for debugging.
Issue introduced by d87d548cd0, so backpatch down to 16.
Author: Michael Paquier, Ranier Vilela
Reviewed-by: Jeff Davis, Kyotaro Horiguchi
Discussion: https://postgr.es/m/7073610042fcf97e1bea2ce08b7e0214b5e11094.camel@j-davis.com
Backpatch-through: 16
Both 50e17ad28 and 29f45e299 mistakenly tried to record a plan dependency
on a function but mistakenly inverted the OidIsValid test. This meant
that we'd record a dependency only when the function's Oid was
InvalidOid. Clearly this was meant to *not* record the dependency in
that case.
50e17ad28 made this mistake first, then in v15 29f45e299 copied the same
mistake.
Reported-by: Tom Lane
Backpatch-through: 14, where 50e17ad28 first made this mistake
Discussion: https://postgr.es/m/2277537.1694301772@sss.pgh.pa.us
If an out-of-memory error was thrown at an unfortunate time,
ensure_record_cache_typmod_slot_exists() could leak memory and leave
behind a global state that produced an infinite loop on the next call.
Fix by merging RecordCacheArray and RecordIdentifierArray into a single
array. With only one allocation or re-allocation, there is no
intermediate state.
Back-patch to all supported releases.
Reported-by: "James Pang (chaolpan)" <chaolpan@cisco.com>
Reviewed-by: Michael Paquier <michael@paquier.xyz>
Discussion: https://postgr.es/m/PH0PR11MB519113E738814BDDA702EDADD6EFA%40PH0PR11MB5191.namprd11.prod.outlook.com
Previously, the test relied on a trick with a shell to retrieve the PID
of the psql session to be stopped with SIGINT, that was skipped on
Windows. This commit changes the test to use IPC::Run::signal()
instead, which still does not work on Windows, but for a different
reason: SIGINT would stop the test before finishing.
This should allow the test to run on non-Windows platforms where PPID is
not supported (like NetBSD), spreading it a bit more across the
buildfarm. And the logic of the test is simpler.
It is the first time in the tree that IPC::Run::signal() is used, so, as
a matter of safety (or just call that as me having cold feet), no
backpatch is done, at least for now.
Author: Yugo NAGATA
Reviewed-by: Fabien Coelho
Discussion: https://postgr.es/m/20230810125935.22c2922ea5250ba79358965b@sraoss.co.jp
This changes 020_cancel.pl so as the test is entirely skipped on
Windows. This test was already doing nothing under WIN32, except
initializing and starting a node without using it so this shaves a few
test cycles.
Author: Yugo NAGATA
Reviewed-by: Fabien Coelho
Discussion: https://postgr.es/m/20230810125935.22c2922ea5250ba79358965b@sraoss.co.jp
Backpatch-through: 15
The new test added by commit 68a59f9e9 disables the subscription and
manually drops the associated replication slot. However, since
disabling the subsubscription doesn't wait for a walsender to release
the replication slot and exit, pg_drop_replication_slot() could
fail. Avoid failure by adding a wait for the replication slot to
become inactive.
Reported-by: Hou Zhijie, as per buildfarm
Reviewed-by: Hou Zhijie
Discussion: https://postgr.es/m/OS0PR01MB571682316378379AA34854F694E9A%40OS0PR01MB5716.jpnprd01.prod.outlook.com
Backpatch-through: 15
generation_counter includes time spent on both JIT:ing expressions
and tuple deforming which are configured independently via options
jit_expressions and jit_tuple_deforming. As they are combined in
the same counter it's not apparent what fraction of time the tuple
deforming takes.
This adds deform_counter dedicated to tuple deforming, which allows
seeing more directly the influence jit_tuple_deforming is having on
the query. The counter is exposed in EXPLAIN and pg_stat_statements
bumpin pg_stat_statements to 1.11.
Author: Dmitry Dolgov <9erthalion6@gmail.com>
Reviewed-by: Pavel Stehule <pavel.stehule@gmail.com>
Reviewed-by: Daniel Gustafsson <daniel@yesql.se>
Discussion: https://postgr.es/m/20220612091253.eegstkufdsu4kfls@erthalion.local
Thomas Munro