Documentation says it cannot signal "a backend owned by a superuser".
On the contrary, it could signal background workers, including the
logical replication launcher. It could signal autovacuum workers and
the autovacuum launcher. Block all that. Signaling autovacuum workers
and those two launchers doesn't stall progress beyond what one could
achieve other ways. If a cluster uses a non-core extension with a
background worker that does not auto-restart, this could create a denial
of service with respect to that background worker. A background worker
with bugs in its code for responding to terminations or cancellations
could experience those bugs at a time the pg_signal_backend member
chooses. Back-patch to v11 (all supported versions).
Reviewed by Jelte Fennema-Nio. Reported by Hemanth Sandrana and
Mahendrakar Srinivasarao.
Security: CVE-2023-5870
This function implements the standard XMLTest function, which
converts text into xml text nodes. It uses the libxml2 function
xmlEncodeSpecialChars to escape predefined entities (&"<>), so
that those do not cause any conflict when concatenating the text
node output with existing xml documents.
This also adds a note in features.sgml about not supporting
XML(SEQUENCE). The SQL specification defines a RETURNING clause
to a set of XML functions, where RETURNING CONTENT or RETURNING
SEQUENCE can be defined. Since PostgreSQL doesn't support
XML(SEQUENCE) all of these functions operate with an
implicit RETURNING CONTENT.
Author: Jim Jones <jim.jones@uni-muenster.de>
Reviewed-by: Vik Fearing <vik@postgresfriends.org>
Discussion: https://postgr.es/m/86617a66-ec95-581f-8d54-08059cca8885@uni-muenster.de
recurse_set_operations() uses the parse tree for the group number estimation,
because of the "varno 0" hack. At the same time 2489d76c49 made root->parse
and corresponding parent_root->simple_rte_array[]->subquery distinct copies
of the parse tree, while d3d55ce571 introduced self-join removal replacing
relid of removed relation only in one of the copies.
The present commit fixes this bug by making recurse_set_operations() call
estimate_num_groups() with the copy of the parse tree processed by self-join
removal.
In future, we may think about maintaining just one copy of the parse tree
and/or keeping removed relids as aliases.
Reported-by: Zuming Jiang
Bug: #18170
Discussion: https://postgr.es/m/flat/18170-f1d17bf9a0d58b24%40postgresql.org
Author: Richard Guo, Alexander Korotkov
Reviewed-by: Andrei Lepikhov
As per the preceding commit, GUC APIs generally expose NULL-valued
string variables as empty strings. Extend that policy to
GetConfigOption() and GetConfigOptionResetString(), eliminating
a crash hazard for unwary callers, as well as a fundamental
ambiguity in GetConfigOption()'s API.
No back-patch, since this is an API change and conceivably somebody
somewhere is depending on this corner case.
Xing Guo, Aleksander Alekseev, Tom Lane
Discussion: https://postgr.es/m/CACpMh+AyDx5YUpPaAgzVwC1d8zfOL4JoD-uyFDnNSa1z0EsDQQ@mail.gmail.com
get_explain_guc_options() crashed if a string GUC marked GUC_EXPLAIN
has a NULL boot_val. Nosing around found a couple of other places
that seemed insufficiently cautious about NULL string values, although
those are likely unreachable in practice. Add some commentary
defining the expectations for NULL values of string variables,
in hopes of forestalling future additions of more such bugs.
Xing Guo, Aleksander Alekseev, Tom Lane
Discussion: https://postgr.es/m/CACpMh+AyDx5YUpPaAgzVwC1d8zfOL4JoD-uyFDnNSa1z0EsDQQ@mail.gmail.com
Introduce unicode_version(), icu_unicode_version(), and
unicode_assigned().
The latter requires introducing a new lookup table for the Unicode
General Category, which is generated along with the other Unicode
lookup tables.
Discussion: https://postgr.es/m/CA+TgmoYzYR-yhU6k1XFCADeyj=Oyz2PkVsa3iKv+keM8wp-F_A@mail.gmail.com
Reviewed-by: Peter Eisentraut
pgstatindex failed with ERRCODE_DATA_CORRUPTED, of the "can't-happen"
class XX. The other functions succeeded on an empty index; they might
have malfunctioned if the failed index build left torn I/O or other
complex state. Report an ERROR in statistics functions pgstatindex,
pgstatginindex, pgstathashindex, and pgstattuple. Report DEBUG1 and
skip all index I/O in maintenance functions brin_desummarize_range,
brin_summarize_new_values, brin_summarize_range, and
gin_clean_pending_list. Back-patch to v11 (all supported versions).
Discussion: https://postgr.es/m/20231001195309.a3@google.com
When beginning recovery, a LOG is displayed by the startup process to
show which recovery mode will be used depending on the .signal file(s)
set in the data folder, like "standby mode", recovery up to a given
target type and value, or archive recovery.
A different patch is under discussion to simplify the startup code by
requiring the presence of recovery.signal and/or standby.signal when a
backup_label file is read. Delaying a bit this LOG ensures that the
correct recovery mode would be reported, and putting it at this position
does not make it lose its value.
While on it, this commit adds a few comments documenting a bit more the
initial recovery steps and their dependencies, and fixes an incorrect
comment format. This introduces no behavior changes.
Extracted from a larger patch by me.
Reviewed-by: David Steele, Bowen Shi
Discussion: https://postgr.es/m/ZArVOMifjzE7f8W7@paquier.xyz
When beginning recovery from a base backup by reading a backup_label
file, it may be possible that no checkpoint record is available
depending on the method used when the case backup was taken, which would
prevent recovery from beginning. In this case, the FATAL messages
issued, initially added by c900c15269, mentioned recovery.signal as
an option to do recovery but not standby.signal. Let's add it as an
available option, for clarity.
Per suggestion from Bowen Shi, extracted from a larger patch by me.
Author: Michael Paquier
Discussion: https://postgr.es/m/CAM_vCudkSjr7NsNKSdjwtfAm9dbzepY6beZ5DP177POKy8=2aw@mail.gmail.com
Historically, the statistics of the checkpointer have been always part
of pg_stat_bgwriter. This commit removes a few columns from
pg_stat_bgwriter, and introduces pg_stat_checkpointer with equivalent,
renamed columns (plus a new one for the reset timestamp):
- checkpoints_timed -> num_timed
- checkpoints_req -> num_requested
- checkpoint_write_time -> write_time
- checkpoint_sync_time -> sync_time
- buffers_checkpoint -> buffers_written
The fields of PgStat_CheckpointerStats and its SQL functions are renamed
to match with the new field names, for consistency. Note that
background writer and checkpointer have been split into two different
processes in commits 806a2aee37 and bf405ba8e4. The pgstat
structures were already split, making this change straight-forward.
Bump catalog version.
Author: Bharath Rupireddy
Reviewed-by: Bertrand Drouvot, Andres Freund, Michael Paquier
Discussion: https://postgr.es/m/CALj2ACVxX2ii=66RypXRweZe2EsBRiPMj0aHfRfHUeXJcC7kHg@mail.gmail.com
This commit refactors find_tabstat_entry() so as transaction counters
for inserted, updated and deleted tuples are included in the result
returned. If a shared entry is found for a relation, its result is now
a copy of the PgStat_TableStatus entry retrieved from shared memory.
This idea has been proposed by Andres Freund.
While on it, the following SQL functions, used in system views, are
refactored with macros, in the same spirit as 83a1a1b566, reducing the
amount of code:
- pg_stat_get_xact_tuples_deleted()
- pg_stat_get_xact_tuples_inserted()
- pg_stat_get_xact_tuples_updated()
There is now only one caller of find_tabstat_entry() in the tree.
Author: Bertrand Drouvot
Discussion: https://postgr.es/m/b9e1f543-ee93-8168-d530-d961708ad9d3@gmail.com
The original code did very little to guard against integer or floating
point overflow when computing the interval's fields. Detect any such
overflows and error out, rather than silently returning bogus results.
Joseph Koshakow, reviewed by Ashutosh Bapat and me.
Discussion: https://postgr.es/m/CAAvxfHcm1TPwH_zaGWuFoL8pZBestbRZTU6Z%3D-RvAdSXTPbKfg%40mail.gmail.com
When looping around after finding that the set-returning function
returned zero rows for the current input tuple, ExecProjectSet
neglected to reset either of the two memory contexts it's
responsible for cleaning out. Typically this wouldn't cause much
problem, because once the SRF does return at least one row, the
contexts would get reset on the next call. However, if the SRF
returns no rows for many input tuples in succession, quite a lot
of memory could be transiently consumed.
To fix, make sure we reset both contexts while looping around.
Per bug #18172 from Sergei Kornilov. Back-patch to all supported
branches.
Discussion: https://postgr.es/m/18172-9b8c5fc1d676ded3@postgresql.org
When calculating distance for interval values, the code mostly mimicked
interval_mi, i.e. it built a new interval value for the difference.
That however does not work for sufficiently distant interval values,
when the difference overflows the interval range.
Instead, we can calculate the distance directly, without constructing
the intermediate (and unnecessary) interval value.
Backpatch to 14, where minmax-multi indexes were introduced.
Reported-by: Dean Rasheed
Reviewed-by: Ashutosh Bapat, Dean Rasheed
Backpatch-through: 14
Discussion: https://postgr.es/m/eef0ea8c-4aaa-8d0d-027f-58b1f35dd170@enterprisedb.com
Make sure that infinite values in date/timestamp columns are treated as
if in infinite distance. Infinite values should not be merged with other
values, leaving them as outliers. The code however returned distance 0
in this case, so that infinite values were merged first. While this does
not break the index (i.e. it still produces correct query results), it
may make it much less efficient.
We don't need explicit handling of infinite date/timestamp values when
calculating distances, because those values are represented as extreme
but regular values (e.g. INT64_MIN/MAX for the timestamp type).
We don't need an exact distance, just a value that is much larger than
distanced between regular values. With the added cast to double values,
we can simply subtract the values.
The regression test queries a value in the "gap" and checks the range
was properly eliminated by the BRIN index.
This only affects minmax-multi indexes on timestamp/date columns with
infinite values, which is not very common in practice. The affected
indexes may need to be rebuilt.
Backpatch to 14, where minmax-multi indexes were introduced.
Reported-by: Ashutosh Bapat
Reviewed-by: Ashutosh Bapat, Dean Rasheed
Backpatch-through: 14
Discussion: https://postgr.es/m/eef0ea8c-4aaa-8d0d-027f-58b1f35dd170@enterprisedb.com
When calculating the distance between date values, make sure to subtract
them in the right order, i.e. (larger - smaller).
The distance is used to determine which values to merge, and is expected
to be a positive value. The code unfortunately did the subtraction in
the opposite order, i.e. (smaller - larger), thus producing negative
values and merging values the most distant values first.
The resulting index is correct (i.e. produces correct results), but may
be significantly less efficient. This affects all minmax-multi indexes
on date columns.
Backpatch to 14, where minmax-multi indexes were introduced.
Reported-by: Ashutosh Bapat
Reviewed-by: Ashutosh Bapat, Dean Rasheed
Backpatch-through: 14
Discussion: https://postgr.es/m/eef0ea8c-4aaa-8d0d-027f-58b1f35dd170@enterprisedb.com
When calculating distances for timestamp values for BRIN minmax-multi
indexes, we need to be careful about overflows for extreme values. If
the value overflows into a negative value, the index may be inefficient.
The new regression test checks this for the timestamp type by adding a
table with enough values to force range compaction/merging. The values
are close to min/max, which means a risk of overflow.
Fixed by converting the int64 values to double first, before calculating
the distance. This prevents the overflow. We may lose some precision, of
course, but that's good enough. In the worst case we build a slightly
less efficient index, but for large distances this won't matter.
This only affects minmax-multi indexes on timestamp columns, with ranges
containing values sufficiently distant to cause an overflow. That seems
like a fairly rare case in practice.
Backpatch to 14, where minmax-multi indexes were introduced.
Reported-by: Ashutosh Bapat
Reviewed-by: Ashutosh Bapat, Dean Rasheed
Backpatch-through: 14
Discussion: https://postgr.es/m/eef0ea8c-4aaa-8d0d-027f-58b1f35dd170@enterprisedb.com
d3d55ce571 changed RelOptInfo.unique_for_rels from the list of Relid sets to
the list of UniqueRelInfo's. But it didn't make UniqueRelInfo a node.
This commit makes UniqueRelInfo a node. Also this commit revises some
comments related to RelOptInfo.unique_for_rels.
Reported-by: Tom Lane
Discussion: https://postgr.es/m/flat/1189851.1698340331%40sss.pgh.pa.us
Two attributes related to checkpointer statistics are removed in this
commit:
- buffers_backend, that counts the number of buffers written directly by
a backend.
- buffers_backend_fsync, that counts the number of times a backend had
to do fsync() by its own.
These are actually not checkpointer properties but backend properties.
Also, pg_stat_io provides a more accurate and equivalent report of these
numbers, by tracking all the I/O stats related to backends, including
writes and fsyncs, so storing them in pg_stat_checkpointer was
redundant.
Thanks also to Robert Haas and Amit Kapila for their input.
Bump catalog version.
Author: Bharath Rupireddy
Reviewed-by: Bertrand Drouvot, Andres Freund
Discussion: https://postgr.es/m/20230210004604.mcszbscsqs3bc5nx@awork3.anarazel.de
f0efa5aec added initReadOnlyStringInfo to allow a StringInfo to be
initialized from an existing buffer and also relaxed the requirement
that a StringInfo's buffer must be NUL terminated at data[len]. Now
that we have that, there's no need for these aggregate deserial
functions to use appendBinaryStringInfo() as that rather wastefully
palloc'd a new buffer and memcpy'd in the bytea's buffer. Instead, we can
just use the bytea's buffer and point the StringInfo directly to that
using the new initializer function.
In Amdahl's law, this speeds up the serial portion of parallel
aggregates and makes sum(numeric), avg(numeric), var_pop(numeric),
var_samp(numeric), variance(numeric), stddev_pop(numeric),
stddev_samp(numeric), stddev(numeric), array_agg(anyarray),
string_agg(text) and string_agg(bytea) scale better in parallel queries.
Author: David Rowley
Discussion: https://postgr.es/m/CAApHDvr%3De-YOigriSHHm324a40HPqcUhSp6pWWgjz5WwegR%3DcQ%40mail.gmail.com
Since C99, there can be a trailing comma after the last value in an
enum definition. A lot of new code has been introducing this style on
the fly. Some new patches are now taking an inconsistent approach to
this. Some add the last comma on the fly if they add a new last
value, some are trying to preserve the existing style in each place,
some are even dropping the last comma if there was one. We could
nudge this all in a consistent direction if we just add the trailing
commas everywhere once.
I omitted a few places where there was a fixed "last" value that will
always stay last. I also skipped the header files of libpq and ecpg,
in case people want to use those with older compilers. There were
also a small number of cases where the enum type wasn't used anywhere
(but the enum values were), which ended up confusing pgindent a bit,
so I left those alone.
Discussion: https://www.postgresql.org/message-id/flat/386f8c45-c8ac-4681-8add-e3b0852c1620%40eisentraut.org
There were various places in our codebase which conjured up a StringInfo
by manually assigning the StringInfo fields and setting the data field
to point to some existing buffer. There wasn't much consistency here as
to what fields like maxlen got set to and in one location we didn't
correctly ensure that the buffer was correctly NUL terminated at len
bytes, as per what was documented as required in stringinfo.h
Here we introduce 2 new functions to initialize StringInfos. One allows
callers to initialize a StringInfo passing along a buffer that is
already allocated by palloc. Here the StringInfo code uses this buffer
directly rather than doing any memcpying into a new allocation. Having
this as a function allows us to verify the buffer is correctly NUL
terminated. StringInfos initialized this way can be appended to and
reset just like any other normal StringInfo.
The other new initialization function also accepts an existing buffer,
but the given buffer does not need to be a pointer to a palloc'd chunk.
This buffer could be a pointer pointing partway into some palloc'd chunk
or may not even be palloc'd at all. StringInfos initialized this way
are deemed as "read-only". This means that it's not possible to
append to them or reset them.
For the latter of the two new initialization functions mentioned above,
we relax the requirement that the data buffer must be NUL terminated.
Relaxing this requirement is convenient in a few places as it can save
us from having to allocate an entire new buffer just to add the NUL
terminator or save us from having to temporarily add a NUL only to have to
put the original char back again later.
Incompatibility note:
Here we also forego adding the NUL in a few places where it does not
seem to be required. These locations are passing the given StringInfo
into a type's receive function. It does not seem like any of our
built-in receive functions require this, but perhaps there's some UDT
out there in the wild which does require this. It is likely worthy of
a mention in the release notes that a UDT's receive function mustn't rely
on the input StringInfo being NUL terminated.
Author: David Rowley
Reviewed-by: Tom Lane
Discussion: https://postgr.es/m/CAApHDvorfO3iBZ%3DxpiZvp3uHtJVLyFaPBSvcAhAq2HPLnaNSwQ%40mail.gmail.com
Currently, expand_single_inheritance_child() doesn't reset
perminfoindex in a plain-inheritance parent's child RTE, because
prior to 387f9ed0a0, the executor would use the first child RTE to
locate the parent's RTEPermissionInfo. That in turn causes
add_rte_to_flat_rtable() to create an extra RTEPermissionInfo
belonging to the parent's child RTE with the same content as the one
belonging to the parent's original ("root") RTE.
In 387f9ed0a0, we changed things so that the executor can now use the
parent's "root" RTE for locating its RTEPermissionInfo instead of the
child RTE, so the latter's perminfoindex need not be set anymore, so
make it so.
Reported-by: Tom Lane
Discussion: https://postgr.es/m/839708.1698174464@sss.pgh.pa.us
Backpatch-through: 16
While reading information from the old cluster, a list of logical
slots is fetched. At the later part of upgrading, pg_upgrade revisits the
list and restores slots by executing pg_create_logical_replication_slot()
on the new cluster. Migration of logical replication slots is only
supported when the old cluster is version 17.0 or later.
If the old node has invalid slots or slots with unconsumed WAL records,
the pg_upgrade fails. These checks are needed to prevent data loss.
The significant advantage of this commit is that it makes it easy to
continue logical replication even after upgrading the publisher node.
Previously, pg_upgrade allowed copying publications to a new node. With
this patch, adjusting the connection string to the new publisher will
cause the apply worker on the subscriber to connect to the new publisher
automatically. This enables seamless continuation of logical replication,
even after an upgrade.
Author: Hayato Kuroda, Hou Zhijie
Reviewed-by: Peter Smith, Bharath Rupireddy, Dilip Kumar, Vignesh C, Shlok Kyal
Discussion: http://postgr.es/m/TYAPR01MB58664C81887B3AF2EB6B16E3F5939@TYAPR01MB5866.jpnprd01.prod.outlook.com
Discussion: http://postgr.es/m/CAA4eK1+t7xYcfa0rEQw839=b2MzsfvYDPz3xbD+ZqOdP3zpKYg@mail.gmail.com
The Self Join Elimination (SJE) feature removes an inner join of a plain table
to itself in the query tree if is proved that the join can be replaced with
a scan without impacting the query result. Self join and inner relation are
replaced with the outer in query, equivalence classes, and planner info
structures. Also, inner restrictlist moves to the outer one with removing
duplicated clauses. Thus, this optimization reduces the length of the range
table list (this especially makes sense for partitioned relations), reduces
the number of restriction clauses === selectivity estimations, and potentially
can improve total planner prediction for the query.
The SJE proof is based on innerrel_is_unique machinery.
We can remove a self-join when for each outer row:
1. At most one inner row matches the join clause.
2. Each matched inner row must be (physically) the same row as the outer one.
In this patch we use the next approach to identify a self-join:
1. Collect all merge-joinable join quals which look like a.x = b.x
2. Add to the list above the baseretrictinfo of the inner table.
3. Check innerrel_is_unique() for the qual list. If it returns false, skip
this pair of joining tables.
4. Check uniqueness, proved by the baserestrictinfo clauses. To prove
the possibility of self-join elimination inner and outer clauses must have
an exact match.
The relation replacement procedure is not trivial and it is partly combined
with the one, used to remove useless left joins. Tests, covering this feature,
were added to join.sql. Some regression tests changed due to self-join removal
logic.
Discussion: https://postgr.es/m/flat/64486b0b-0404-e39e-322d-0801154901f3%40postgrespro.ru
Author: Andrey Lepikhov, Alexander Kuzmenkov
Reviewed-by: Tom Lane, Robert Haas, Andres Freund, Simon Riggs, Jonathan S. Katz
Reviewed-by: David Rowley, Thomas Munro, Konstantin Knizhnik, Heikki Linnakangas
Reviewed-by: Hywel Carver, Laurenz Albe, Ronan Dunklau, vignesh C, Zhihong Yu
Reviewed-by: Greg Stark, Jaime Casanova, Michał Kłeczek, Alena Rybakina
Reviewed-by: Alexander Korotkov
When an UPDATE/DELETE/MERGE's target table is an old-style
inheritance tree, it's possible for the parent to get excluded
from the plan while some children are not. (I believe this is
only possible if we can prove that a CHECK ... NO INHERIT
constraint on the parent contradicts the query WHERE clause,
so it's a very unusual case.) In such a case, ExecInitModifyTable
mistakenly concluded that the first surviving child is the target
table, leading to at least two bugs:
1. The wrong table's statement-level triggers would get fired.
2. In v16 and up, it was possible to fail with "invalid perminfoindex
0 in RTE with relid nnnn" due to the child RTE not having permissions
data included in the query plan. This was hard to reproduce reliably
because it did not occur unless the update triggered some non-HOT
index updates.
In v14 and up, this is easy to fix by defining ModifyTable.rootRelation
to be the parent RTE in plain inheritance as well as partitioned cases.
While the wrong-triggers bug also appears in older branches, the
relevant code in both the planner and executor is quite a bit
different, so it would take a good deal of effort to develop and
test a suitable patch. Given the lack of field complaints about the
trigger issue, I'll desist for now. (Patching v11 for this seems
unwise anyway, given that it will have no more releases after next
month.)
Per bug #18147 from Hans Buschmann.
Amit Langote and Tom Lane
Discussion: https://postgr.es/m/18147-6fc796538913ee88@postgresql.org
Enforce the rule from transam/README in XLogRegisterBuffer(), and
update callers to follow the rule.
Hash indexes sometimes register clean pages as a part of the locking
protocol, so provide a REGBUF_NO_CHANGE flag to support that use.
Discussion: https://postgr.es/m/c84114f8-c7f1-5b57-f85a-3adc31e1a904@iki.fi
Reviewed-by: Heikki Linnakangas
This shouldn't change behavior except in the unusual case where
there are file in the tablespace directory that have entirely
numeric names but are nevertheless not possible names for a
tablespace directory, either because their names have leading zeroes
that shouldn't be there, or the value is actually zero, or because
the value is too large to represent as an OID.
In those cases, the directory would previously have made it into
the list of tablespaceinfo objects and no longer will. Thus, base
backups will now ignore such directories, instead of treating them
as legitimate tablespace directories. Similarly, if entries for
such tablespaces occur in a tablespace_map file, they will now
be rejected as erroneous, instead of being honored.
This is infrastructure for future work that wants to be able to
know the tablespace of each relation that is part of a backup
*as an OID*. By strengthening the up-front validation, we don't
have to worry about weird cases later, and can more easily avoid
repeated string->integer conversions.
Patch by me, reviewed by David Steele.
Discussion: http://postgr.es/m/CA+TgmoZNVeBzoqDL8xvr-nkaepq815jtDR4nJzPew7=3iEuM1g@mail.gmail.com
Instead of returning the number of characters in the RelFileNumber,
return the RelFileNumber itself. Continue to return the fork number,
as before, and additionally return the segment number.
parse_filename_for_nontemp_relation now rejects a RelFileNumber or
segment number that begins with a leading zero. Before, we accepted
such cases as relation filenames, but if we continued to do so after
this change, the function might return the same values for two
different files (e.g. 1234.5 and 001234.5 or 1234.005) which could be
annoying for callers. Since we don't actually ever generate filenames
with leading zeroes in the names, any such files that we find must
have been created by something other than PostgreSQL, and it is
therefore reasonable to treat them as non-relation files.
Along the way, change unlogged_relation_entry to store a RelFileNumber
rather than an OID. This update should have been made in
851f4cc75c, but it was overlooked.
It's trivial to make the update as part of this commit, perhaps more
trivial than it would have been without it, so do that.
Patch by me, reviewed by David Steele.
Discussion: http://postgr.es/m/CA+TgmoZNVeBzoqDL8xvr-nkaepq815jtDR4nJzPew7=3iEuM1g@mail.gmail.com
pg_logical_replication_slot_advance() included a break condition to stop
when a targeted LSN is reached, when processing a series of WAL records
with XLogReadRecord(). Since 38a957316d, it matched with the check of
its main while loop. This condition saved from an extra CFI check,
actually pointless, so let's remove this condition and simplify the
code.
In passing, fix an incorrect comment.
Author: Bharath Rupireddy
Reviewed-by: Tom Lane, Gurjeet Singh
Discussion: https://postgr.es/m/CALj2ACWfGDLQ2cy7ZKwxnJqbDkO6Yvqqrqxne5ZN4HYm=PRTGg@mail.gmail.com
When min_dynamic_shared_memory is set above 0, we try to find space in a
pre-allocated region of the main shared memory area instead of calling
dsm_impl_XXX() routines to allocate more. The dsm_pin_segment() and
dsm_unpin_segment() routines had a bug: they called dsm_impl_XXX()
routines even for main region segments. Nobody noticed before now
because those routines do nothing on Unix, but on Windows they'd fail
while attempting to duplicate an invalid Windows HANDLE. Add the
missing gating.
Back-patch to 14, where commit 84b1c63a added this feature. Fixes
pgsql-bugs bug #18165.
Reported-by: Maxime Boyer <maxime.boyer@cra-arc.gc.ca>
Tested-by: Alexander Lakhin <exclusion@gmail.com>
Discussion: https://postgr.es/m/18165-bf4f525cea6e51de%40postgresql.org
Previously, ALTER SYSTEM failed if the target GUC wasn't present in
the session's GUC hashtable. That is a reasonable behavior for core
(single-part) GUC names, and for custom GUCs for which we have loaded
an extension that's reserved the prefix. But it's unnecessarily
restrictive otherwise, and it also causes inconsistent behavior:
you can "ALTER SYSTEM SET foo.bar" only if you did "SET foo.bar"
earlier in the session. That's fairly silly.
Hence, refactor things so that we can execute ALTER SYSTEM even
if the variable doesn't have a GUC hashtable entry, as long as the
name meets the custom-variable naming requirements and does not
have a reserved prefix. (It's safe to do this even if the
variable belongs to an extension we currently don't have loaded.
A bad value will at worst cause a WARNING when the extension
does get loaded.)
Also, adjust GRANT ON PARAMETER to have the same opinions about
whether to allow an unrecognized GUC name, and to throw the
same errors if not (it previously used a one-size-fits-all
message for several distinguishable conditions). By default,
only a superuser will be allowed to do ALTER SYSTEM SET on an
unrecognized name, but it's possible to GRANT the ability to
do it.
Patch by me, pursuant to a documentation complaint from
Gavin Panella. Arguably this is a bug fix, but given the
lack of other complaints I'll refrain from back-patching.
Discussion: https://postgr.es/m/2617358.1697501956@sss.pgh.pa.us
Discussion: https://postgr.es/m/169746329791.169914.16613647309012285391@wrigleys.postgresql.org
Allow the COMMUTATOR, NEGATOR, MERGES, and HASHES attributes to be set
by ALTER OPERATOR. However, we don't allow COMMUTATOR/NEGATOR to be
changed once set, nor allow the MERGES/HASHES flags to be unset once
set. Changes like that might invalidate plans already made, and
dealing with the consequences seems like more trouble than it's worth.
The main use-case we foresee for this is to allow addition of missed
properties in extension update scripts, such as extending an existing
operator to support hashing. So only transitions from not-set to set
states seem very useful.
This patch also causes us to reject some incorrect cases that formerly
resulted in inconsistent catalog state, such as trying to set the
commutator of an operator to be some other operator that already has a
(different) commutator.
While at it, move the InvokeObjectPostCreateHook call for CREATE
OPERATOR to not occur until after we've fixed up commutator or negator
links as needed. The previous ordering could only be justified by
thinking of the OperatorUpd call as a kind of ALTER OPERATOR step;
but we don't call InvokeObjectPostAlterHook therein. It seems better
to let the hook see the final state of the operator object.
In the documentation, move the discussion of how to establish
commutator pairs from xoper.sgml to the CREATE OPERATOR ref page.
Tommy Pavlicek, reviewed and editorialized a bit by me
Discussion: https://postgr.es/m/CAEhP-W-vGVzf4udhR5M8Bdv88UYnPrhoSkj3ieR3QNrsGQoqdg@mail.gmail.com
This allows tools that read the WAL sequentially to identify (possible)
redo points when they're reached, rather than only being able to
detect them in retrospect when XLOG_CHECKPOINT_ONLINE is found, possibly
much later in the WAL stream. There are other possible applications as
well; see the discussion links below.
Any redo location that precedes the checkpoint location should now point
to an XLOG_CHECKPOINT_REDO record, so add a cross-check to verify this.
While adjusting the code in CreateCheckPoint() for this patch, I made it
call WALInsertLockAcquireExclusive a bit later than before, since there
appears to be no need for it to be held while checking whether the system
is idle, whether this is an end-of-recovery checkpoint, or what the current
timeline is.
Bump XLOG_PAGE_MAGIC.
Patch by me, based in part on earlier work from Dilip Kumar. Review by
Dilip Kumar, Amit Kapila, Andres Freund, and Michael Paquier.
Discussion: http://postgr.es/m/CA+TgmoYy-Vc6G9QKcAKNksCa29cv__czr+N9X_QCxEfQVpp_8w@mail.gmail.com
Discussion: http://postgr.es/m/20230614194717.jyuw3okxup4cvtbt%40awork3.anarazel.de
Discussion: http://postgr.es/m/CA+hUKG+b2ego8=YNW2Ohe9QmSiReh1-ogrv8V_WZpJTqP3O+2w@mail.gmail.com
There was no I/O timing statistics for counting read and write timings
on local blocks, contrary to the counterparts for temp and shared
blocks. This information is available when track_io_timing is enabled.
The output of EXPLAIN is updated to show this information. An update of
pg_stat_statements is planned next.
Author: Nazir Bilal Yavuz
Reviewed-by: Robert Haas, Melanie Plageman
Discussion: https://postgr.es/m/CAN55FZ19Ss279mZuqGbuUNxka0iPbLgYuOQXqAKewrjNrp27VA@mail.gmail.com
These two counters, defined in BufferUsage to track respectively the
time spent while reading and writing blocks have historically only
tracked data related to shared buffers, when track_io_timing is enabled.
An upcoming patch to add specific counters for local buffers will take
advantage of this rename as it has come up that no data is currently
tracked for local buffers, and tracking local and shared buffers using
the same fields would be inconsistent with the treatment done for temp
buffers. Renaming the existing fields clarifies what the block type of
each stats field is.
pg_stat_statement is updated to reflect the rename. No extension
version bump is required as 5a3423ad8e has done one, affecting v17~.
Author: Nazir Bilal Yavuz
Reviewed-by: Robert Haas, Melanie Plageman
Discussion: https://postgr.es/m/CAN55FZ19Ss279mZuqGbuUNxka0iPbLgYuOQXqAKewrjNrp27VA@mail.gmail.com
Remove use of LLVMGetElementType() and provide the type of all pointers
to LLVMBuildXXX() functions when emitting IR, as required by modern LLVM
versions[1].
* For LLVM <= 14, we'll still use the old LLVMBuildXXX() functions.
* For LLVM == 15, we'll continue to do the same, explicitly opting
out of opaque pointer mode.
* For LLVM >= 16, we'll use the new LLVMBuildXXX2() functions that take
the extra type argument.
The difference is hidden behind some new IR emitting wrapper functions
l_load(), l_gep(), l_call() etc. The change is mostly mechanical,
except that at each site the correct type had to be provided.
In some places we needed to do some extra work to get functions types,
including some new wrappers for C++ APIs that are not yet exposed by in
LLVM's C API, and some new "example" functions in llvmjit_types.c
because it's no longer possible to start from the function pointer type
and ask for the function type.
Back-patch to 12, because it's a little tricker in 11 and we agreed not
to put the latest LLVM support into the upcoming final release of 11.
[1] https://llvm.org/docs/OpaquePointers.html
Reviewed-by: Dmitry Dolgov <9erthalion6@gmail.com>
Reviewed-by: Ronan Dunklau <ronan.dunklau@aiven.io>
Reviewed-by: Andres Freund <andres@anarazel.de>
Discussion: https://postgr.es/m/CA%2BhUKGKNX_%3Df%2B1C4r06WETKTq0G4Z_7q4L4Fxn5WWpMycDj9Fw%40mail.gmail.com
Relation files extended by multiple blocks at a time have been counting
the number of blocks written, but forgot to increment the write time in
this case, as single-block write and relation extension are treated as
two different I/O operations in the shared stats: IOOP_EXTEND vs
IOOP_WRITE. In this case IOOP_EXTEND was forgotten for normal
(non-temporary) relations, still the number of blocks written was
incremented according to the relation extend done.
Write times are tracked when track_io_timing is enabled, which is not
the case by default.
Author: Nazir Bilal Yavuz
Reviewed-by: Robert Haas, Melanie Plageman
Discussion: https://postgr.es/m/CAN55FZ19Ss279mZuqGbuUNxka0iPbLgYuOQXqAKewrjNrp27VA@mail.gmail.com
Backpatch-through: 16
Since its introduction, LogLogicalMessage() (via the SQL interface
pg_logical_emit_message()) has never included a call to XLogFlush(),
causing it to potentially lose messages on a crash when used in
non-transactional mode. This has come up to me as a problem while
playing with ideas to design a test suite for what has become
039_end_of_wal.pl introduced in bae868caf2 by Thomas Munro, because
there are no direct ways to force a WAL flush via SQL.
The default is false, to not flush messages and influence existing
use-cases where this function could be used. If set to true, the
message emitted is flushed before returning back to the caller, making
the message durable on crash. This new option has no effect when using
pg_logical_emit_message() in transactional mode, as the record's flush
is guaranteed by the WAL record generated by the transaction committed.
Two queries of test_decoding are tweaked to cover the new code path for
the flush.
Bump catalog version.
Author: Michael Paquier
Reviewed-by: Andres Freund, Amit Kapila, Fujii Masao, Tung Nguyen, Tomas
Vondra
Discussion: https://postgr.es/m/ZNsdThSe2qgsfs7R@paquier.xyz
Use a modulo operator instead of implementing the same behavior
with a loop. The loop solution is doubtless microscopically
faster for the typical case of only wrapping into the very next
day, but maybe not so much for large interval values. In any
case, timetz is such a backwater that it's doubtful anybody
would notice any performance change anyway.
This avoids a compiler bug occurring in AIX's xlc, even in pretty
late-model revisions.
We did not have test coverage for the case where the initial
result->time value is negative, so add that.
For the moment, install this only in HEAD. My plan is to
back-patch the test case, and then the code change assuming that
buildfarm testing proves the bug occurs in the back branches.
(That seems pretty likely, but let's find out for sure.)
Per buildfarm results from commits 97957fdba and 2f0472030.
Thanks to Michael Paquier for the idea to use a modulo operation
to replace the faulty loop.
Discussion: https://postgr.es/m/CA+hUKGK=DOC+hE-62FKfZy=Ybt5uLkrg3zCZD-jFykM-iPn8yw@mail.gmail.com
The SIGTERM handler for the startup process immediately calls
proc_exit() for the duration of the restore_command, i.e., a call
to system(). This system() call forks a new process to execute the
shell command, and this child process inherits the parent's signal
handlers. If both the parent and child processes receive SIGTERM,
both will attempt to call proc_exit(). This can end badly. For
example, both processes will try to remove themselves from the
PGPROC shared array.
To fix this problem, this commit adds a check in
StartupProcShutdownHandler() to see whether MyProcPid == getpid().
If they match, this is the parent process, and we can proc_exit()
like before. If they do not match, this is a child process, and we
just emit a message to STDERR (in a signal safe manner) and
_exit(), thereby skipping any problematic exit callbacks.
This commit also adds checks in proc_exit(), ProcKill(), and
AuxiliaryProcKill() that verify they are not being called within
such child processes.
Suggested-by: Andres Freund
Reviewed-by: Thomas Munro, Andres Freund
Discussion: https://postgr.es/m/Y9nGDSgIm83FHcad%40paquier.xyz
Discussion: https://postgr.es/m/20230223231503.GA743455%40nathanxps13
Backpatch-through: 11
First, we shouldn't recommend switching to single-user mode, because
that's terrible advice. Especially on newer versions where VACUUM
will enter emergency mode when nearing (M)XID exhaustion, it's
perfectly fine to just VACUUM in multi-user mode. Doing it that way
is less disruptive and avoids disabling the safeguards that prevent
actual wraparound, so recommend that instead.
Second, be more precise about what is going to happen (when we're
nearing the limits) or what is happening (when we actually hit them).
The database doesn't shut down, nor does it refuse all commands. It
refuses commands that assign whichever of XIDs and MXIDs are nearly
exhausted.
No back-patch. The existing hint that advises going to single-user
mode is sufficiently awful advice that removing it or changing it
might be justifiable even though we normally avoid changing
user-facing messages in back-branches, but I (rhaas) felt that it
was better to be more conservative and limit this fix to master
only. Aside from the usual risk of breaking translations, people
might be used to the existing message, or even have monitoring
scripts that look for it.
Alexander Alekseev, John Naylor, Robert Haas, reviewed at various
times by Peter Geoghegan, Hannu Krosing, and Andres Freund.
Discussion: http://postgr.es/m/CA+TgmoZBg95FiR9wVQPAXpGPRkacSt2okVge+PKPPFppN7sfnQ@mail.gmail.com
The word "assign" is used in various places internally to describe what
GetNewMultiXactId does, but the user-facing messages have previously
said "generate". For consistency, standardize on "assign," which seems
(at least to me) to be slightly clearer.
Discussion: http://postgr.es/m/CA+TgmoaoE1_i3=4-7GCTtKLVZVQ2Gh6qESW2VG1OprtycxOHMA@mail.gmail.com
It is possible that the tail XID of pg_serial/ gets ahead of its head
XID, which would cause the truncation of pg_serial/ done during
checkpoints to show up as a "wraparound" LOG in SimpleLruTruncate(),
which is confusing. This also wastes a bit of disk space until the head
page is reclaimed again.
CheckPointPredicate() is changed so as the cutoff page for the
truncation is switched to the head page if the tail XID has advanced
beyond the head XID, rather than the tail page. This prevents the
confusing LOG message about a wraparound while allowing some truncation
to be done to cut in disk space.
This could be considered as a bug fix, but the original behavior is
harmless as well, resulting only in disk space temporarily wasted, so
no backpatch is done.
Author: Sami Imseih
Reviewed-by: Heikki Linnakangas, Michael Paquier
Discussion: https://postgr.es/m/755E19CA-D02C-4A4C-80D3-74F775410C48@amazon.com
Restart the apply worker if the subscription owner's superuser privileges
have been revoked. This is required so that the subscription connection
string gets revalidated and use the password option to connect to the
publisher for non-superusers, if required.
Author: Vignesh C
Reviewed-by: Amit Kapila
Discussion: http://postgr.es/m/CALDaNm2Dxmhq08nr4P6G+24QvdBo_GAVyZ_Q1TcGYK+8NHs9xw@mail.gmail.com
Dropping a temp table could entail TOAST table access to clean out
toasted catalog entries, such as large pg_constraint.conbin strings
for complex CHECK constraints. If we did that via ON COMMIT DROP,
we triggered the assertion in init_toast_snapshot(), because
there was no provision for setting up a snapshot for the drop
actions. Fix that.
(I assume here that the adjacent truncation actions for ON COMMIT
DELETE ROWS don't have a similar problem: it doesn't seem like
nontransactional truncations would need to touch any toasted fields.
If that proves wrong, we could refactor a bit to have the same
snapshot acquisition cover that too.)
The test case added here does not fail before v15, because that
assertion was added in 277692220 which was not back-patched.
However, the race condition the assertion warns of surely
exists further back, so back-patch to all supported branches.
Per report from Richard Guo.
Discussion: https://postgr.es/m/CAMbWs4-x26=_QxxgdJyNbiCDzvtr2WV5ZDso_v-CukKEe6cBZw@mail.gmail.com
If SIGTERM is received within this section, the startup process
will immediately proc_exit() in the signal handler, so it is
inadvisable to include any more code than is required there (as
such code is unlikely to be compatible with doing proc_exit() in a
signal handler). This commit moves the code recently added to this
section (see 1b06d7bac9 and 7fed801135) to outside of the section.
This ensures that the startup process only calls proc_exit() in its
SIGTERM handler for the duration of the system() call, which is how
this code worked from v8.4 to v14.
Reported-by: Michael Paquier, Thomas Munro
Analyzed-by: Andres Freund
Suggested-by: Tom Lane
Reviewed-by: Michael Paquier, Robert Haas, Thomas Munro, Andres Freund
Discussion: https://postgr.es/m/Y9nGDSgIm83FHcad%40paquier.xyz
Discussion: https://postgr.es/m/20230223231503.GA743455%40nathanxps13
Backpatch-through: 15
This commit introduces trigger on login event, allowing to fire some actions
right on the user connection. This can be useful for logging or connection
check purposes as well as for some personalization of environment. Usage
details are described in the documentation included, but shortly usage is
the same as for other triggers: create function returning event_trigger and
then create event trigger on login event.
In order to prevent the connection time overhead when there are no triggers
the commit introduces pg_database.dathasloginevt flag, which indicates database
has active login triggers. This flag is set by CREATE/ALTER EVENT TRIGGER
command, and unset at connection time when no active triggers found.
Author: Konstantin Knizhnik, Mikhail Gribkov
Discussion: https://postgr.es/m/0d46d29f-4558-3af9-9c85-7774e14a7709%40postgrespro.ru
Reviewed-by: Pavel Stehule, Takayuki Tsunakawa, Greg Nancarrow, Ivan Panchenko
Reviewed-by: Daniel Gustafsson, Teodor Sigaev, Robert Haas, Andres Freund
Reviewed-by: Tom Lane, Andrey Sokolov, Zhihong Yu, Sergey Shinderuk
Reviewed-by: Gregory Stark, Nikita Malakhov, Ted Yu
Commit dc7d70ea added functions that read the control file, but didn't
acquire ControlFileLock. With unlucky timing, file systems that have
weak interlocking like ext4 and ntfs could expose partially overwritten
contents, and the checksum would fail.
Back-patch to all supported releases.
Reviewed-by: David Steele <david@pgmasters.net>
Reviewed-by: Anton A. Melnikov <aamelnikov@inbox.ru>
Reviewed-by: Michael Paquier <michael@paquier.xyz>
Discussion: https://postgr.es/m/20221123014224.xisi44byq3cf5psi%40awork3.anarazel.de
For the same reasons given in commit 403ac226d, adjust these
functions to not assume that checking SearchSysCacheExists can
guarantee success of a later fetch.
This follows the same internal API choices made in the earlier commit:
add a function XXXExt(oid, is_missing) and use that to eliminate
the need for a separate existence check. The changes are very
straightforward, though tedious. For the moment I just made the new
functions static in namespace.c, but we could export them if a need
emerges.
Per bug #18014 from Alexander Lakhin. Given the lack of hard evidence
that there's a bug in non-debug builds, I'm content to fix this only
in HEAD.
Discussion: https://postgr.es/m/18014-28c81cb79d44295d@postgresql.org
The versions of these functions that accept object OIDs are supposed
to return NULL, rather than failing, if the target object has been
dropped. This makes it safe(r) to use them in queries that scan
catalogs, since the functions will be applied to objects that are
visible in the query's snapshot but might now be gone according to
the catalog snapshot. In most cases we implemented this by doing
a SearchSysCacheExists test and assuming that if that succeeds, we
can safely invoke the appropriate aclchk.c function, which will
immediately re-fetch the same syscache entry. It was argued that
if the existence test succeeds then the followup fetch must succeed
as well, for lack of any intervening AcceptInvalidationMessages call.
Alexander Lakhin demonstrated that this is not so when
CATCACHE_FORCE_RELEASE is enabled: the syscache entry will be forcibly
dropped at the end of SearchSysCacheExists, and then it is possible
for the catalog snapshot to get advanced while re-fetching the entry.
Alexander's test case requires the operation to happen inside a
parallel worker, but that seems incidental to the fundamental problem.
What remains obscure is whether there is a way for this to happen in a
non-debug build. Nonetheless, CATCACHE_FORCE_RELEASE is a very useful
test methodology, so we'd better make the code safe for it.
After some discussion we concluded that the most future-proof fix
is to give up the assumption that checking SearchSysCacheExists can
guarantee success of a later fetch. At best that assumption leads
to fragile code --- for example, has_type_privilege appears broken
for array types even if you believe the assumption holds. And it's
not even particularly efficient.
There had already been some work towards extending the aclchk.c
APIs to include "is_missing" output flags, so this patch extends
that work to cover all the aclchk.c functions that are used by the
has_xxx_privilege() functions. (This allows getting rid of some
ad-hoc decisions about not throwing errors in certain places in
aclchk.c.)
In passing, this fixes the has_sequence_privilege() functions to
provide the same guarantees as their cousins: for some reason the
SearchSysCacheExists tests never got added to those.
There is more work to do to remove the unsafe coding pattern with
SearchSysCacheExists in other places, but this is a pretty
self-contained patch so I'll commit it separately.
Per bug #18014 from Alexander Lakhin. Given the lack of hard evidence
that there's a bug in non-debug builds, I'm content to fix this only
in HEAD. (Perhaps we should clean up the has_sequence_privilege()
oversight in the back branches, but in the absence of field complaints
I'm not too excited about that either.)
Discussion: https://postgr.es/m/18014-28c81cb79d44295d@postgresql.org
When COPYing into a partitioned table that does now permit the use of
table_multi_insert(), we could error out with
ERROR: could not read block NN in file "base/...": read only 0 of 8192 bytes
because BulkInsertState->next_free was not reset between partitions. This
problem occurred only when not able to use table_multi_insert(), as a
dedicated BulkInsertState for each partition is used in that case.
The bug was introduced in 00d1e02be2, but it was hard to hit at that point,
as commonly bulk relation extension is not used when not using
table_multi_insert(). It became more likely after 82a4edabd2, which expanded
the use of bulk extension.
To fix the bug, reset the bulk relation extension state in BulkInsertState in
ReleaseBulkInsertStatePin(). That was added (in b1ecb9b3fc) to tackle a very
similar issue. Obviously the name is not quite correct, but there might be
external callers, and bulk insert state needs to be reset in precisely in the
situations that ReleaseBulkInsertStatePin() already needed to be called.
Medium term the better fix likely is to disallow reusing BulkInsertState
across relations.
Add a test that, without the fix, reproduces #18130 in most
configurations. The test also catches the problem fixed in b1ecb9b3fc when
run with small shared_buffers.
Reported-by: Ivan Kolombet <enderstd@gmail.com>
Analyzed-by: Tom Lane <tgl@sss.pgh.pa.us>
Analyzed-by: Andres Freund <andres@anarazel.de>
Bug: #18130
Discussion: https://postgr.es/m/18130-7a86a7356a75209d%40postgresql.org
Discussion: https://postgr.es/m/257696.1695670946%40sss.pgh.pa.us
Backpatch: 16-
* sync_method is renamed to wal_sync_method.
* sync_method_options[] is renamed to wal_sync_method_options[].
* assign_xlog_sync_method() is renamed to assign_wal_sync_method().
* The names of the available synchronization methods are now
prefixed with "WAL_SYNC_METHOD_" and have been moved into a
WalSyncMethod enum.
* PLATFORM_DEFAULT_SYNC_METHOD is renamed to
PLATFORM_DEFAULT_WAL_SYNC_METHOD, and DEFAULT_SYNC_METHOD is
renamed to DEFAULT_WAL_SYNC_METHOD.
These more descriptive names help distinguish the code for
wal_sync_method from the code for DataDirSyncMethod (e.g., the
recovery_init_sync_method configuration parameter and the
--sync-method option provided by several frontend utilities). This
change also prevents name collisions between the aforementioned
sets of code. Since this only improves the naming of internal
identifiers, there should be no behavior change.
Author: Maxim Orlov
Discussion: https://postgr.es/m/CACG%3DezbL1gwE7_K7sr9uqaCGkWhmvRTcTEnm3%2BX1xsRNwbXULQ%40mail.gmail.com
When converting a timestamp to/from with/without time zone, the SQL
Standard specifies an AT LOCAL variant of AT TIME ZONE which uses the
session's time zone. This includes three system functions able to do
the work in the same way as the existing flavors for AT TIME ZONE,
except that these need to be marked as stable as they depend on the
session's TimeZone GUC.
Bump catalog version.
Author: Vik Fearing
Reviewed-by: Laurenz Albe, Cary Huang, Michael Paquier
Discussion: https://postgr.es/m/8e25dec4-5667-c1a5-6581-167d710c2182@postgresfriends.org
When MyProc->delayChkptFlags is set to temporarily block phase
transitions in a concurrent checkpoint, the checkpointer enters a
sleep-poll loop to wait for the flag to be cleared. We should show that
as a wait event in the pg_stat_activity view.
Reviewed-by: Robert Haas <robertmhaas@gmail.com>
Reviewed-by: Michael Paquier <michael@paquier.xyz>
Discussion: https://postgr.es/m/CA%2BhUKGL7Whi8iwKbzkbn_1fixH3Yy8aAPz7mfq6Hpj7FeJrKMg%40mail.gmail.com
An upcoming patch wants to introduce an additional special case in
this function. To keep that as cheap as possible, minimize the amount
of branching that we do based on whether this is an XLOG_SWITCH
record.
Additionally, and also in the interest of keeping the overhead of
special-case code paths as low as possible, apply likely() to the
non-XLOG_SWITCH case, since only a very tiny fraction of WAL records
will be XLOG_SWITCH records.
Patch by me, reviewed by Dilip Kumar, Amit Kapila, Andres Freund,
and Michael Paquier.
Discussion: http://postgr.es/m/CA+TgmoYy-Vc6G9QKcAKNksCa29cv__czr+N9X_QCxEfQVpp_8w@mail.gmail.com
This could only affect HASH partitioned tables with at least 2 partition
key columns.
If partition pruning was delayed until execution and the query contained
an IS NULL qual on one of the partitioned keys, and some subsequent
partitioned key was being compared to a non-Const, then this could result
in a crash due to the incorrect keyno being used to calculate the
stateidx for the expression evaluation code.
Here we fix this by properly skipping partitioned keys which have a
nullkey set. Effectively, this must be the same as what's going on
inside perform_pruning_base_step().
Sergei Glukhov also provided a patch, but that's not what's being used
here.
Reported-by: Sergei Glukhov
Reviewed-by: tender wang, Sergei Glukhov
Discussion: https://postgr.es/m/d05b26fa-af54-27e1-f693-6c31590802fa@postgrespro.ru
Backpatch-through: 11, where runtime partition pruning was added.
get_steps_using_prefix_recurse() incorrectly assumed that it could stop
recursive processing of the 'prefix' list when cur_keyno was one before
the step_lastkeyno. Since hash partition pruning can prune using IS
NULL quals, and these IS NULL quals are not present in the 'prefix'
list, then that logic could cause more levels of recursion than what is
needed and lead to there being no more items in the 'prefix' list to
process. This would manifest itself as a crash in some code that
expected the 'start' ListCell not to be NULL.
Here we adjust the logic so that instead of stopping recursion at 1 key
before the step_lastkeyno, we just look at the llast(prefix) item and
ensure we only recursively process up until just before whichever the last
key is. This effectively allows keys to be missing in the 'prefix' list.
This change does mean that step_lastkeyno is no longer needed, so we
remove that from the static functions. I also spent quite some time
reading this code and testing it to try to convince myself that there
are no other issues. That resulted in the irresistible temptation of
rewriting some comments, many of which were just not true or inconcise.
Reported-by: Sergei Glukhov
Reviewed-by: Sergei Glukhov, tender wang
Discussion: https://postgr.es/m/2f09ce72-315e-2a33-589a-8519ada8df61@postgrespro.ru
Backpatch-through: 11, where partition pruning was introduced.
This adds a new option called BGWORKER_BYPASS_ROLELOGINCHECK to the
flags available to BackgroundWorkerInitializeConnection() and
BackgroundWorkerInitializeConnectionByOid().
This gives the possibility to bgworkers to bypass the role login check,
making possible the use of a role that has no login rights while not
being a superuser. PostgresInit() gains a new flag called
INIT_PG_OVERRIDE_ROLE_LOGIN, taking advantage of the refactoring done in
4800a5dfb4.
Regression tests are added to worker_spi to check the behavior of this
new option with bgworkers.
Author: Bertrand Drouvot
Reviewed-by: Nathan Bossart, Michael Paquier, Bharath Rupireddy
Discussion: https://postgr.es/m/bcc36259-7850-4882-97ef-d6b905d2fc51@gmail.com
In commit 3fc6e2d7f, I (tgl) argued that we only need to check for
a constant-FALSE restriction clause when there's exactly one
restriction clause, on the grounds that const-folding would have
thrown away anything ANDed with a Const FALSE. That's true just after
const-folding has been applied, but subsequent processing such as
equivalence class expansion could result in cases where a Const FALSE
is ANDed with some other stuff. (Compare for instance joinrels.c's
restriction_is_constant_false.) Hence, tweak this logic to check all
the elements of the baserestrictinfo list, not just one; that's cheap
enough to not be worth worrying about.
There is one existing test case where this visibly improves the plan.
There would not be any savings in runtime, but the planner effort and
executor startup effort will be reduced, and anyway it's odd that
we can detect related cases but not this one.
Richard Guo (independently discovered by David Rowley)
Discussion: https://postgr.es/m/CAMbWs4_x3-CnVVrCboS1LkEhB5V+W7sLSCabsRiG+n7+5_kqbg@mail.gmail.com
The check is not about processing the startup packet, so the calling
function seems like a more natural place. I'm also working on a patch
that moves 'canAcceptConnections' out of the Port struct, and this
makes that refactoring more convenient.
Reviewed-by: Tristan Partin
Discussion: https://www.postgresql.org/message-id/7a59b073-5b5b-151e-7ed3-8b01ff7ce9ef@iki.fi
InitPostgres() has been using a set of boolean arguments to control its
behavior, and a patch under discussion was aiming at expanding it with a
third one. In preparation for expanding this area, this commit switches
all the current boolean arguments of this routine to a single bits32
argument instead. Two values are currently supported for the flags:
- INIT_PG_LOAD_SESSION_LIBS to load [session|local]_preload_libraries at
startup.
- INIT_PG_OVERRIDE_ALLOW_CONNS to allow connection to a database even if
it has !datallowconn. This is used by bgworkers.
Reviewed-by: Bertrand Drouvot
Discussion: https://postgr.es/m/ZSTn66_BXRZCeaqS@paquier.xyz
Since we added the PlannerInfo.all_baserels set, it's not really
necessary to grovel over the rangetable to count baserels in the
current query. So let's drop has_multiple_baserels() in favor
of a bms_membership() test. This might be microscopically
faster, but the main point is to remove some unnecessary code.
Richard Guo
Discussion: https://postgr.es/m/CAMbWs4_8RcSbbfs1ASZLrMuL0c0EQgXWcoLTQD8swBRY_pQQiA@mail.gmail.com
While working on a8a968a82, I failed to consider that
cheapest_startup_path can be NULL when there is no non-parameterized
path in the pathlist. This is well documented in set_cheapest(), I just
failed to notice.
Here we adjust the code to just check if the RelOptInfo has a
cheapest_startup_path set before adding it to the startup_subpaths list.
Reported-by: Richard Guo
Author: Richard Guo
Discussion: https://postgr.es/m/CAMbWs49w3t03V69XhdCuw+GDwivny4uQUxrkVp6Gejaspt0wMQ@mail.gmail.com
The comment claimed that it is "called from postmaster", but it is
actually called in the child process, pretty early in the process
initialization. I guess you could interpret "called from postmaster"
to mean that, but it seems wrong to me. Rename the function to be
consistent with other functions with similar role.
Reviewed-by: Thomas Munro
Discussion: https://www.postgresql.org/message-id/4f95c1fc-ad3c-7974-3a8c-6faa3931804c@iki.fi
In EXEC_BACKEND or single-user mode, we process
shared_preload_libraries at postmaster startup as usual, but also at
backend startup. When a library calls RegisterBackgroundWorker() when
being loaded into a backend process, we go through the motions to add
the worker to BackgroundWorkerList, even though that is a
postmaster-private data structure. Make it return early when called in
a backend process, without changing BackgroundWorkerList.
You could argue that it was intentional: In non-EXEC_BACKEND mode, the
backend processes inherit BackgroundWorkerList at fork(), so it does
make some sense to initialize it to the same state in EXEC_BACKEND
mode, too. It's clearly a postmaster-private structure, though, and
all the functions that use it are clearly marked as "should only be
called in postmaster".
You could also argue that libraries should not call
RegisterBackgroundWorker() during backend startup. It's too late to
correctly register any static background workers at that stage. But
it's a common pattern in extensions, and it doesn't seem worth the
churn to require all extensions to change it.
Another sloppiness was the exception for "internal" background
workers. We checked that RegisterBackgroundWorker() was called during
shared_preload_libraries processing, or the background worker was an
internal one. That exception was made in commit 665d1fad99 to allow
postmaster to register the logical apply launcher in
ApplyLauncherRegister(). The way the check was written, it would not
complain if you registered an internal background worker in a regular
backend process. But it would complain if postmaster registered a
background worker defined in a shared library, outside
shared_preload_libraries processing. I think the correct rule is that
you can only register static background workers in the postmaster
process, and only before the bgworker shared memory array has been
initialized. Check for that more directly.
Reviewed-by: Thomas Munro
Discussion: https://www.postgresql.org/message-id/4f95c1fc-ad3c-7974-3a8c-6faa3931804c@iki.fi
The serialized representation of an internal aggregate state is a bytea
value. In each deserial function, in order to "receive" the bytea value
we appended it onto a short-lived StringInfoData using
appendBinaryStringInfo. This was a little wasteful as it meant having to
palloc memory, copy a (possibly long) series of bytes then later pfree
that memory. Instead of going to this extra trouble, we can just fake up
a StringInfoData and point the data directly at the bytea's payload. This
should help increase the performance of internal aggregate
deserialization.
Reviewed-by: Michael Paquier
Discussion: https://postgr.es/m/CAApHDvr=e-YOigriSHHm324a40HPqcUhSp6pWWgjz5WwegR=cQ@mail.gmail.com
1349d2790 added code to adjust the PlannerInfo.group_pathkeys so that
ORDER BY / DISTINCT aggregate functions could obtain pre-sorted inputs
to allow faster execution. That commit forgot to adjust the pathkeys in
create_agg_path(). Some code in there assumed that it was always fine
to make the AggPath's pathkeys the same as its subpath's. That seems to
have been ok up until 1349d2790, but since that commit adds pathkeys for
columns which are within the aggregate function, those columns won't be
available above the aggregate node. This can result in "could not find
pathkey item to sort" during create_plan().
The fix here is to strip off the additional pathkeys added by
adjust_group_pathkeys_for_groupagg(). It seems that the pathkeys here
will only ever be group_pathkeys, so all we need to do is check if the
length of the pathkey list is longer than the num_groupby_pathkeys and
get rid of the additional ones only if we see extras.
Reported-by: Justin Pryzby
Reviewed-by: Richard Guo
Discussion: https://postgr.es/m/ZQhYYRhUxpW3PSf9%40telsasoft.com
Backpatch-through: 16, where 1349d2790 was introduced
Going by c4a1933b4, b33ef397a and 05893712c (to name just a few), it seems
that maintaining debug_print_rel() is often forgotten. In the case of
c4a1933b4, it was several years before anyone noticed that a path type
was not handled by debug_print_rel(). (debug_print_rel() is only
compiled when building with OPTIMIZER_DEBUG).
After a quick survey on the pgsql-hackers mailing list, nobody came
forward to admit that they use OPTIMIZER_DEBUG. So to prevent any future
maintenance neglect, let's just remove debug_print_rel() and have
OPTIMIZER_DEBUG make use of pprint() instead (as suggested by Tom Lane).
If anyone wants to come forward to claim they make use of
OPTIMIZER_DEBUG in a way that they need debug_print_rel() then they have
around 10 months remaining in the v17 cycle where we could revert this.
If nobody comes forward in that time, then we can likely safely declare
debug_print_rel() as not worth keeping.
Discussion: https://postgr.es/m/CAApHDvoCdjo8Cu2zEZF4-AxWG-90S+pYXAnoDDa9J3xH-OrczQ@mail.gmail.com
Currently, B-tree code matches every scan key to every item on the page.
Imagine the ordered B-tree scan for the query like this.
SELECT * FROM tbl WHERE col > 'a' AND col < 'b' ORDER BY col;
The (col > 'a') scan key will be always matched once we find the location to
start the scan. The (col < 'b') scan key will match every item on the page
as long as it matches the last item on the page.
This patch implements prechecking of the scan keys required for directional
scan on beginning of page scan. If precheck is successful we can skip this
scan keys check for the items on the page. That could lead to significant
acceleration especially if the comparison operator is expensive.
Idea from patch by Konstantin Knizhnik.
Discussion: https://postgr.es/m/079c3f8e-3371-abe2-e93c-fc8a0ae3f571%40garret.ru
Reviewed-by: Peter Geoghegan, Pavel Borisov
BuildDescForRelation() main job is to convert ColumnDef lists to
pg_attribute/tuple descriptor arrays, which is really mostly an
internal subroutine of DefineRelation() and some related functions,
which is more the remit of tablecmds.c and doesn't have much to do
with the basic tuple descriptor interfaces in tupdesc.c. This is also
supported by observing the header includes we can remove in tupdesc.c.
By moving it over, we can also (in the future) make
BuildDescForRelation() use more internals of tablecmds.c that are not
sensible to be exposed in tupdesc.c.
Discussion: https://www.postgresql.org/message-id/flat/52a125e4-ff9a-95f5-9f61-b87cf447e4da@eisentraut.org
Keep track of the used size of the array. That avoids looping through
the whole array in a few places. It doesn't matter from a performance
point of view since the array is small anyway, but this feels less
surprising and is a little less code. Now that we have an explicit
NumListenSockets variable that is statically initialized to 0, we
don't need the loop to initialize the array.
Allocate the array in PostmasterContext. The array isn't needed in
child processes, so this allows reusing that memory. We could easily
make the array resizable now, but we haven't heard any complaints
about the current 64 sockets limit.
Discussion: https://www.postgresql.org/message-id/7bb7ad65-a018-2419-742f-fa5fd877d338@iki.fi
Previously, the JSON code didn't have to worry too much about freeing
JsonLexContext, because it was never too long-lived. With new features
being added for SQL/JSON this is no longer the case. Add a routine
that knows how to free this struct and apply that to a few places, to
prevent this from becoming problematic.
At the same time, we change the API of makeJsonLexContextCstringLen to
make it receive a pointer to JsonLexContext for callers that want it to
be stack-allocated; it can also be passed as NULL to get the original
behavior of a palloc'ed one.
This also causes an ABI break due to the addition of flags to
JsonLexContext, so we can't easily backpatch it. AFAICS that's not much
of a problem; apparently some leaks might exist in JSON usage of
text-search, for example via json_to_tsvector, but I haven't seen any
complaints about that.
Per Coverity complaint about datum_to_jsonb_internal().
Discussion: https://postgr.es/m/20230808174110.oq3iymllsv6amkih@alvherre.pgsql
6b94e7a6d did this for ordered append paths to allow fast startup
MergeAppends, however, nothing was done for the Append case.
Here we adjust add_paths_to_append_rel() to have it build an AppendPath
containing the cheapest startup paths from each of the child relations
when the append rel has "consider_startup" set.
Author: Andy Fan, David Rowley
Discussion: https://www.postgresql.org/message-id/CAKU4AWrXSkUV=Pt-gRxQT7EbfUeNssprGyNsB=5mJibFZ6S3ww@mail.gmail.com
Ensure we switch to the per-tuple memory context to prevent any memory
leaks of detoasted Datums in MemoizeHash_hash() and MemoizeHash_equal().
Reported-by: Orlov Aleksej
Author: Orlov Aleksej, David Rowley
Discussion: https://postgr.es/m/83281eed63c74e4f940317186372abfd%40cft.ru
Backpatch-through: 14, where Memoize was added
It is unnecessary to include this field in IndexInfo. It is only used
by DDL code, not during execution. It is really only used to pass
local information around between functions in index.c and indexcmds.c,
for which it is clearer to use local variables, like in similar cases.
Discussion: https://www.postgresql.org/message-id/flat/f84640e3-00d3-5abd-3f41-e6a19d33c40b@eisentraut.org
In hopes of putting these where any would-be implementer is sure to
find them, make a placeholder grammar production for ALTER DROP VALUE
and put them there. This is really just a docs patch, though.
Vik Fearing, with a bit more wordsmithing by me
Discussion: https://postgr.es/m/9fffd149-da0f-0c9c-6745-731fb688642a@postgresfriends.org
This further reduces the length and complexity of sendFile(),
hopefully make it easier to understand and modify. In addition
to moving some logic into a new function, I took this opportunity
to make a few slight adjustments to sendFile() itself, including
renaming the 'len' variable to 'bytes_done', since we use it to represent
the number of bytes we've already handled so far, not the total
length of the file.
Patch by me, reviewed by David Steele.
Discussion: http://postgr.es/m/CA+TgmoYt5jXH4U6cu1dm9Oe2FTn1aae6hBNhZzJJjyjbE_zYig@mail.gmail.com
If checksum verification fails for a particular page, we reread the
page and try one more time. The code that does this somewhat complex
and difficult to follow. Move some of the logic into a new function
and rearrange the code a bit to try to make it clearer. This way,
we don't need the block_retry Boolean, a couple of other variables
move from sendFile() into the new function, and some code is now less
deeply indented.
Patch by me, reviewed by David Steele.
Discussion: http://postgr.es/m/CA+TgmoYt5jXH4U6cu1dm9Oe2FTn1aae6hBNhZzJJjyjbE_zYig@mail.gmail.com
The code in charge of copying the contents of PgBackendStatus to local
memory could fail on memory allocation because of an overflow on the
amount of memory to use. The overflow can happen when combining a high
value track_activity_query_size (max at 1MB) with a large
max_connections, when both multiplied get higher than INT32_MAX as both
parameters treated as signed integers. This could for example trigger
with the following functions, all calling pgstat_read_current_status():
- pg_stat_get_backend_subxact()
- pg_stat_get_backend_idset()
- pg_stat_get_progress_info()
- pg_stat_get_activity()
- pg_stat_get_db_numbackends()
The change to use MemoryContextAllocHuge() has been introduced in
8d0ddccec6, so backpatch down to 12.
Author: Jakub Wartak
Discussion: https://postgr.es/m/CAKZiRmw8QSNVw2qNK-dznsatQqz+9DkCquxP0GHbbv1jMkGHMA@mail.gmail.com
Backpatch-through: 12
Make a few newish calls to appendStringInfo() which have no special
formatting use appendStringInfoString() instead. Also, adjust usages of
appendStringInfoString() which only append a string containing a single
character to make use of appendStringInfoChar() instead.
This makes the code marginally faster, but primarily this change is so
we use the StringInfo type as it was intended to be used.
Discussion: https://postgr.es/m/CAApHDvpXKQmL+r=VDNS98upqhr9yGBhv2Jw3GBFFk_wKHcB39A@mail.gmail.com
This commit changes the WAL reader routines so as a FATAL for the
backend or exit(FAILURE) for the frontend is triggered if an allocation
for a WAL record decode fails in walreader.c, rather than treating this
case as bogus data, which would be equivalent to the end of WAL. The
key is to avoid palloc_extended(MCXT_ALLOC_NO_OOM) in walreader.c,
relying on plain palloc() calls.
The previous behavior could make WAL replay finish too early than it
should. For example, crash recovery finishing earlier may corrupt
clusters because not all the WAL available locally was replayed to
ensure a consistent state. Out-of-memory failures would show up
randomly depending on the memory pressure on the host, but one simple
case would be to generate a large record, then replay this record after
downsizing a host, as Ethan Mertz originally reported.
This relies on bae868caf2, as the WAL reader routines now do the
memory allocation required for a record only once its header has been
fully read and validated, making xl_tot_len trustable. Making the WAL
reader react differently on out-of-memory or bogus record data would
require ABI changes, so this is the safest choice for stable branches.
Also, it is worth noting that 3f1ce97346 has been using a plain
palloc() in this code for some time now.
Thanks to Noah Misch and Thomas Munro for the discussion.
Like the other commit, backpatch down to 12, leaving out v11 that will
be EOL'd soon. The behavior of considering a failed allocation as bogus
data comes originally from 0ffe11abd3, where the record length
retrieved from its header was not entirely trustable.
Reported-by: Ethan Mertz
Discussion: https://postgr.es/m/ZRKKdI5-RRlta3aF@paquier.xyz
Backpatch-through: 12
The retry loop is needed because heap_page_prune() calls
HeapTupleSatisfiesVacuum() and then lazy_scan_prune() does the same
thing again, and they might get different answers due to concurrent
clog updates. But this patch makes heap_page_prune() return the
HeapTupleSatisfiesVacuum() results that it computed back to the
caller, which allows lazy_scan_prune() to avoid needing to recompute
those values in the first place. That's nice both because it eliminates
the need for a retry loop and also because it's cheaper.
Melanie Plageman, reviewed by David Geier, Andres Freund, and me.
Discussion: https://postgr.es/m/CAAKRu_br124qsGJieuYA0nGjywEukhK1dKBfRdby_4yY3E9SXA%40mail.gmail.com
This adjusts the expression evaluation code for CoerceViaIO and
CoerceToDomain to handle errors softly if needed.
For CoerceViaIo, this means using InputFunctionCallSafe(), which
provides the option to handle errors softly, instead of calling the
type input function directly.
For CoerceToDomain, this simply entails replacing the ereport() in
ExecEvalConstraintCheck() by errsave().
In both cases, the ErrorSaveContext to be used when evaluating the
expression is stored by ExecInitExprRec() in the expression's struct
in the expression's ExprEvalStep. The ErrorSaveContext is passed by
setting ExprState.escontext to point to it when calling
ExecInitExprRec() on the expression whose errors are to be handled
softly.
Note that no call site of ExecInitExprRec() has been changed in this
commit, so there's no functional change. This is intended for
implementing new SQL/JSON expression nodes in future commits that
will use to it suppress errors that may occur during type coercions.
Reviewed-by: Álvaro Herrera
Discussion: https://postgr.es/m/CA+HiwqE4XTdfb1nW=Ojoy_tQSRhYt-q_kb6i5d4xcKyrLC1Nbg@mail.gmail.com
After receiving position data for a lexeme, tsvectorrecv()
advanced its "datalen" value by (npos+1)*sizeof(WordEntry)
where the correct calculation is (npos+1)*sizeof(WordEntryPos).
This accidentally failed to render the constructed tsvector
invalid, but it did result in leaving some wasted space
approximately equal to the space consumed by the position data.
That could have several bad effects:
* Disk space is wasted if the received tsvector is stored into a
table as-is.
* A legal tsvector could get rejected with "maximum total lexeme
length exceeded" if the extra space pushes it over the MAXSTRPOS
limit.
* In edge cases, the finished tsvector could be assigned a length
larger than the allocated size of its palloc chunk, conceivably
leading to SIGSEGV when the tsvector gets copied somewhere else.
The odds of a field failure of this sort seem low, though valgrind
testing could probably have found this.
While we're here, let's express the calculation as
"sizeof(uint16) + npos * sizeof(WordEntryPos)" to avoid the type
pun implicit in the "npos + 1" formulation. It's not wrong
given that WordEntryPos had better be 2 bytes to avoid padding
problems, but it seems clearer this way.
Report and patch by Denis Erokhin. Back-patch to all supported
versions.
Discussion: https://postgr.es/m/009801d9f2d9$f29730c0$d7c59240$@datagile.ru
In recent releases, such cases fail with "cache lookup failed for
function 0" rather than complaining that the conversion function
doesn't exist as prior versions did. Seems to be a consequence of
sloppy refactoring in commit f82de5c46. Add the missing error check.
Per report from Pierre Fortin. Back-patch to v14 where the
oversight crept in.
Discussion: https://postgr.es/m/20230929163739.3bea46e5.pfortin@pfortin.com
Commit 9f8377f7a2 was a little too eager in fetching default values.
Normally this would not matter, but if the default value is not valid
for the type (e.g. a varchar that's too long) it caused an unnecessary
error.
Complaint and fix from Laurenz Albe
Backpatch to release 16.
Discussion: https://postgr.es/m/75a7b7483aeb331aa017328d606d568fc715b90d.camel@cybertec.at
These options already exist, but you need to specify a column list for
them, which can be cumbersome. We already have the possibility of all
columns for FORCE QUOTE, so this is simply extending that facility to
FORCE_NULL and FORCE_NOT_NULL.
Author: Zhang Mingli
Reviewed-By: Richard Guo, Kyatoro Horiguchi, Michael Paquier.
Discussion: https://postgr.es/m/CACJufxEnVqzOFtqhexF2+AwOKFrV8zHOY3y=p+gPK6eB14pn_w@mail.gmail.com
ANALYZE on a table with inheritance children analyzes all the child
tables in a loop. When stepping to next child table, it updated the
child rel ID value in the command progress stats, but did not reset
the 'sample_blks_total' and 'sample_blks_scanned' counters.
acquire_sample_rows() updates 'sample_blks_total' as soon as the scan
starts and 'sample_blks_scanned' after processing the first block, but
until then, pg_stat_progress_analyze would display a bogus combination
of the new child table relid with old counter values from the
previously processed child table. Fix by resetting 'sample_blks_total'
and 'sample_blks_scanned' to zero at the same time that
'current_child_table_relid' is updated.
Backpatch to v13, where pg_stat_progress_analyze view was introduced.
Reported-by: Justin Pryzby
Discussion: https://www.postgresql.org/message-id/20230122162345.GP13860%40telsasoft.com
Under some circumstances, concurrent MERGE operations could lead to
inconsistent results, that varied according the plan chosen. This was
caused by a lack of rowmarks on the source relation, which meant that
EvalPlanQual rechecking was not guaranteed to return the same source
tuples when re-running the join query.
Fix by ensuring that preprocess_rowmarks() sets up PlanRowMarks for
all non-target relations used in MERGE, in the same way that it does
for UPDATE and DELETE.
Per bug #18103. Back-patch to v15, where MERGE was introduced.
Dean Rasheed, reviewed by Richard Guo.
Discussion: https://postgr.es/m/18103-c4386baab8e355e3%40postgresql.org
Improve find_base_rel() and find_base_rel_ignore_join() so that they
raise an ERROR if they ever receive a negative relid value in
non-cassert builds. If either of these functions had ever received a
negative relid then they'd have attempted to access memory that does not
belong to simple_rel_array.
Because no evidence has been presented of actual cases where bugs have
caused this to happen, here we take a lightweight approach to checking
for negative values and simply cast both values to uint32 before
performing the comparison. This will cause any negative relids to be
seen as greater than simple_rel_array_size which will ERROR rather than
attempt to access a negative simple_rel_array element. Obviously, the
run-time error is better than a crash, so it makes sense to protect
against this, especially when it can be done without adding any
additional run-time overhead.
There is a slight change here if the functions are ever called with a
relid of 0. This will pass the bounds check, but that array entry
should be NULL (along with the corresponding simple_rte_array entry), so
won't pass the "if (rel)" condition and still fall through and raise an
ERROR.
Author: Ranier Vilela
Reviewed-by: Ashutosh Bapat, David Rowley
Discussion: https://postgr.es/m/CAEudQArQSghBu2gLojg4o_tnHj_x2HcS%3D%2BwewL3NJS8z0VnK%2Bg%40mail.gmail.com
nbtree's mark/restore processing failed to correctly handle an edge case
involving array key advancement and related search-type scan key state.
Scans with ScalarArrayScalarArrayOpExpr quals requiring mark/restore
processing (for a merge join) could incorrectly conclude that an
affected array/scan key must not have advanced during the time between
marking and restoring the scan's position.
As a result of all this, array key handling within btrestrpos could skip
a required call to _bt_preprocess_keys(). This confusion allowed later
primitive index scans to overlook tuples matching the true current array
keys. The scan's search-type scan keys would still have spurious values
corresponding to the final array element(s) -- not values matching the
first/now-current array element(s).
To fix, remember that "array key wraparound" has taken place during the
ongoing btrescan in a flag variable stored in the scan's state, and use
that information at the point where btrestrpos decides if another call
to _bt_preprocess_keys is required.
Oversight in commit 70bc5833, which taught nbtree to handle array keys
during mark/restore processing, but missed this subtlety. That commit
was itself a bug fix for an issue in commit 9e8da0f7, which taught
nbtree to handle ScalarArrayOpExpr quals natively.
Author: Peter Geoghegan <pg@bowt.ie>
Discussion: https://postgr.es/m/CAH2-WzkgP3DDRJxw6DgjCxo-cu-DKrvjEv_ArkP2ctBJatDCYg@mail.gmail.com
Backpatch: 11- (all supported branches).
This code was sloppy about comparison of index columns that
are expressions. It didn't reliably reject cases where one
index has an expression where the other has a plain column,
and it could index off the start of the attmap array, leading
to a Valgrind complaint (though an actual crash seems unlikely).
I'm not sure that the expression-vs-column sloppiness leads
to any visible problem in practice, because the subsequent
comparison of the two expression lists would reject cases
where the indexes have different numbers of expressions
overall. Maybe we could falsely match indexes having the
same expressions in different column positions, but it'd
require unlucky contents of the word before the attmap array.
It's not too surprising that no problem has been reported
from the field. Nonetheless, this code is clearly wrong.
Per bug #18135 from Alexander Lakhin. Back-patch to all
supported branches.
Discussion: https://postgr.es/m/18135-532f4a755e71e4d2@postgresql.org
Previously, one of the values in the struct was returned as the return
value, and another was returned via an output parameter. In
preparation for returning more stuff, consolidate both values into a
struct returned via an output parameter.
Melanie Plageman, reviewed by Andres Freund and by me.
Discussion: https://postgr.es/m/CAAKRu_br124qsGJieuYA0nGjywEukhK1dKBfRdby_4yY3E9SXA%40mail.gmail.com
Tid Range scans were added back in bb437f995. That commit forgot to add
handling for TidRangePaths in print_path().
Only people building with OPTIMIZER_DEBUG might have noticed this, which
likely is the reason it's taken 4 years for anyone to notice.
Author: Andrey Lepikhov
Reported-by: Andrey Lepikhov
Discussion: https://postgr.es/m/379082d6-1b6a-4cd6-9ecf-7157d8c08635@postgrespro.ru
Backpatch-through: 14, where bb437f995 was introduced
This commit removes unnecessary ExecExprFreeContext() calls in
ExecEnd* routines because the actual cleanup is managed by
FreeExecutorState(). With no callers remaining for
ExecExprFreeContext(), this commit also removes the function.
This commit also drops redundant ExecClearTuple() calls, because
ExecResetTupleTable() in ExecEndPlan() already takes care of
resetting and dropping all TupleTableSlots initialized with
ExecInitScanTupleSlot() and ExecInitExtraTupleSlot().
After these modifications, the ExecEnd*() routines for ValuesScan,
NamedTuplestoreScan, and WorkTableScan became redundant. So, this
commit removes them.
Reviewed-by: Robert Haas
Discussion: https://postgr.es/m/CA+HiwqFGkMSge6TgC9KQzde0ohpAycLQuV7ooitEEpbKB0O_mg@mail.gmail.com
"in_streaming" is a flag used to track if an instance of pgoutput is
streaming changes. When pgoutput is started, the flag was always reset,
switched it back and forth in the stream start/stop callbacks.
Before this commit, it was a global variable, which is confusing as it
is actually attached to a state of PGOutputData. Per my analysis, using
a global variable did not lead to an active bug like in 54ccfd6586,
but it makes the code more consistent. Note that we cannot backpatch
this change anyway as it requires the addition of a new field to
PGOutputData, exposed in pgoutput.h.
Author: Hou Zhijie
Reviewed-by: Amit Kapila, Michael Paquier, Peter Smith
Discussion: https://postgr.es/m/OS0PR01MB571690EF24F51F51EFFCBB0E94FAA@OS0PR01MB5716.jpnprd01.prod.outlook.com
This unifies some repetitive code.
Note: I didn't push the "not found" error message into the new
function, even though all existing callers would be able to make use
of it. Using the existing error handling as-is would probably require
exposing the Relation type via tupdesc.h, which doesn't seem
desirable. (Or even if we changed it to just report the OID, it would
inject the concept of a relation containing the tuple descriptor into
tupdesc.h, which might be a layering violation. Perhaps some further
improvements could be considered here separately.)
Discussion: https://www.postgresql.org/message-id/flat/52a125e4-ff9a-95f5-9f61-b87cf447e4da%40eisentraut.org
When performing inlining LLVM unfortunately "leaks" types (the
types survive and are usable, but a new round of inlining will
recreate new structurally equivalent types). This accumulation
will over time amount to a memory leak which for some queries
can be large enough to trigger the OOM process killer.
To avoid accumulation of types, all IR related data is stored
in an LLVMContextRef which is dropped and recreated in order
to release all types. Dropping and recreating incurs overhead,
so it will be done only after 100 queries. This is a heuristic
which might be revisited, but until we can get the size of the
context from LLVM we are flying a bit blind.
This issue has been reported several times, there may be more
references to it in the archives on top of the threads linked
below.
Backpatching of this fix will be handled once it has matured
in master for a bit.
Reported-By: Justin Pryzby <pryzby@telsasoft.com>
Reported-By: Kurt Roeckx <kurt@roeckx.be>
Reported-By: Jaime Casanova <jcasanov@systemguards.com.ec>
Reported-By: Lauri Laanmets <pcspets@gmail.com>
Author: Andres Freund and Daniel Gustafsson
Discussion: https://postgr.es/m/7acc8678-df5f-4923-9cf6-e843131ae89d@www.fastmail.com
Discussion: https://postgr.es/m/20201218235607.GC30237@telsasoft.com
Discussion: https://postgr.es/m/CAPH-tTxLf44s3CvUUtQpkDr1D8Hxqc2NGDzGXS1ODsfiJ6WSqA@mail.gmail.com
Commit b059d2f456 introduced llvm_types_module and accidentally
exported it. As there is no usecase for accessing this variable
externally, this makes it static.
Author: Andres Freund <andres@anarazel.de>
Reviewed-by: Daniel Gustafsson <daniel@yesql.se>
Discussion: https://postgr.es/m/20221101055132.pjjsvlkeo4stbjkq@awork3.anarazel.de
The pgoutput module uses a global variable (publish_no_origin) to cache
the action for the origin filter, but we didn't reset the flag when
shutting down the output plugin, so subsequent retries may access the
previous publish_no_origin value.
We fix this by storing the flag in the output plugin's private data.
Additionally, the patch removes the currently unused origin string from the
structure.
For the back branch, to avoid changing the exposed structure, we eliminated the
global variable and instead directly used the origin string for change
filtering.
Author: Hou Zhijie
Reviewed-by: Amit Kapila, Michael Paquier
Backpatch-through: 16
Discussion: http://postgr.es/m/OS0PR01MB571690EF24F51F51EFFCBB0E94FAA@OS0PR01MB5716.jpnprd01.prod.outlook.com
Yet another bug in the ilk of commits a7ee7c851 and 741b88435. In
741b88435, we took care to clear the memorized location of the
downlink when we split the parent page, because splitting the parent
page can move the downlink. But we missed that even *updating* a tuple
on the parent can move it, because updating a tuple on a gist page is
implemented as a delete+insert, so the updated tuple gets moved to the
end of the page.
This commit fixes the bug in two different ways (belt and suspenders):
1. Clear the downlink when we update a tuple on the parent page, even
if it's not split. This the same approach as in commits a7ee7c851
and 741b88435.
I also noticed that gistFindCorrectParent did not clear the
'downlinkoffnum' when it stepped to the right sibling. Fix that
too, as it seems like a clear bug even though I haven't been able
to find a test case to hit that.
2. Change gistFindCorrectParent so that it treats 'downlinkoffnum'
merely as a hint. It now always first checks if the downlink is
still at that location, and if not, it scans the page like before.
That's more robust if there are still more cases where we fail to
clear 'downlinkoffnum' that we haven't yet uncovered. With this,
it's no longer necessary to meticulously clear 'downlinkoffnum',
so this makes the previous fixes unnecessary, but I didn't revert
them because it still seems nice to clear it when we know that the
downlink has moved.
Also add the test case using the same test data that Alexander
posted. I tried to reduce it to a smaller test, and I also tried to
reproduce this with different test data, but I was not able to, so
let's just include what we have.
Backpatch to v12, like the previous fixes.
Reported-by: Alexander Lakhin
Discussion: https://www.postgresql.org/message-id/18129-caca016eaf0c3702@postgresql.org
There was a mismatch between the const qualifiers for
excludeDirContents in src/backend/backup/basebackup.c and
src/bin/pg_rewind/filemap.c, which led to a quick search for similar
cases. We should make excludeDirContents match, but the rest of the
changes seem like a good idea as well.
Author: David Steele <david@pgmasters.net>
Discussion: https://www.postgresql.org/message-id/flat/669a035c-d23d-2f38-7ff0-0cb93e01d610@pgmasters.net
As implemented in 5891c7a8ed, setting "force" to true in
pgstat_report_wal() causes the routine to not wait for the pgstat
shmem lock if it cannot be acquired, in which case the WAL and I/O
statistics finish by not being flushed. The origin of the confusion
comes from pgstat_flush_wal() and pgstat_flush_io(), that use "nowait"
as sole argument. The I/O stats are new in v16.
This is the opposite behavior of what has been used in
pgstat_report_stat(), where "force" is the opposite of "nowait". In
this case, when "force" is true, the routine sets "nowait" to false,
which would cause the routine to wait for the pgstat shmem lock,
ensuring that the stats are always flushed. When "force" is false,
"nowait" is set to true, and the stats would only not be flushed if the
pgstat shmem lock can be acquired, returning immediately without
flushing the stats if the lock cannot be acquired.
This commit changes pgstat_report_wal() so as "force" has the same
behavior as in pgstat_report_stat(). There are currently three callers
of pgstat_report_wal():
- Two in the checkpointer where force=true during a shutdown and the
main checkpointer loop. Now the code behaves so as the stats are always
flushed.
- One in the main loop of the bgwriter, where force=false. Now the code
behaves so as the stats would not be flushed if the pgstat shmem lock
could not be acquired.
Before this commit, some stats on WAL and I/O could have been lost after
a shutdown, for example.
Reported-by: Ryoga Yoshida
Author: Ryoga Yoshida, Michael Paquier
Discussion: https://postgr.es/m/f87a4d7be70530606b864fd1df91718c@oss.nttdata.com
Backpatch-through: 15
bae868ca removed a check that was still needed. If you had an
xl_tot_len at the end of a page that was too small for a record header,
but not big enough to span onto the next page, we'd immediately perform
the CRC check using a bogus large length. Because of arbitrary coding
differences between the CRC implementations on different platforms,
nothing very bad happened on common modern systems. On systems using
the _sb8.c fallback we could segfault.
Restore that check, add a new assertion and supply a test for that case.
Back-patch to 12, like bae868ca.
Tested-by: Tom Lane <tgl@sss.pgh.pa.us>
Tested-by: Alexander Lakhin <exclusion@gmail.com>
Discussion: https://postgr.es/m/CA%2BhUKGLCkTT7zYjzOxuLGahBdQ%3DMcF%3Dz5ZvrjSOnW4EDhVjT-g%40mail.gmail.com
Thanks to commit 2a8b40e368, the logical replication worker type is
easily determined. The worker type could already be deduced via
other columns such as leader_pid and relid, but that is unnecessary
complexity for users.
Bumps catversion.
Author: Peter Smith
Reviewed-by: Michael Paquier, Maxim Orlov, Amit Kapila
Discussion: https://postgr.es/m/CAHut%2BPtmbSMfErSk0S7xxVdZJ9XVE3xVLhqBTmT91kf57BeKDQ%40mail.gmail.com
Parse analysis of a CallStmt will inject mutable information,
for instance the OID of the called procedure, so that subsequent
DDL may create a need to re-parse the CALL. We failed to detect
this for CALLs in plpgsql routines, because no dependency information
was collected when putting a CallStmt into the plan cache. That
could lead to misbehavior or strange errors such as "cache lookup
failed".
Before commit ee895a655, the issue would only manifest for CALLs
appearing in atomic contexts, because we re-planned non-atomic
CALLs every time through anyway.
It is now apparent that extract_query_dependencies() probably
needs a special case for every utility statement type for which
stmt_requires_parse_analysis() returns true. I wanted to add
something like Assert(!stmt_requires_parse_analysis(...)) when
falling out of extract_query_dependencies_walker without doing
anything, but there are API issues as well as a more fundamental
point: stmt_requires_parse_analysis is supposed to be applied to
raw parser output, so it'd be cheating to assume it will give the
correct answer for post-parse-analysis trees. I contented myself
with adding a comment.
Per bug #18131 from Christian Stork. Back-patch to all supported
branches.
Discussion: https://postgr.es/m/18131-576854e79c5cd264@postgresql.org
The initial estimate of the number of distinct ParsedWords is just
that: an estimate. Don't let it exceed what palloc is willing to
allocate. If in fact we need more entries, we'll eventually fail
trying to enlarge the array. But if we don't, this allows success on
inputs that currently draw "invalid memory alloc request size".
Per bug #18080 from Uwe Binder. Back-patch to all supported branches.
Discussion: https://postgr.es/m/18080-d5c5e58fef8c99b7@postgresql.org