It's possible that external code is calling smgrtruncate(). Any
external callers might like to consider the recent changes to
RelationTruncate(), but commit 38c579b0 should not have changed the
function prototype in the back-branches, per ABI stability policy.
Restore smgrtruncate()'s traditional argument list in the back-branches,
but make it a wrapper for a new function smgrtruncate2(). The three
callers in core can use smgrtruncate2() directly. In master (18-to-be),
smgrtruncate2() is effectively renamed to smgrtruncate(), so this wart
is cleaned up.
Reviewed-by: Michael Paquier <michael@paquier.xyz>
Discussion: https://postgr.es/m/CA%2BhUKG%2BThae6x6%2BjmQiuALQBT2Ae1ChjMh1%3DkMvJ8y_SBJZrvA%40mail.gmail.com
Cause parallel workers to not check datallowconn, rolcanlogin, and
ACL_CONNECT privileges. The leader already checked these things
(except for rolcanlogin which might have been checked for a different
role). Re-checking can accomplish little except to induce unexpected
failures in applications that might not even be aware that their query
has been parallelized. We already had the principle that parallel
workers rely on their leader to pass a valid set of authorization
information, so this change just extends that a bit further.
Also, modify the ReservedConnections, datconnlimit and rolconnlimit
logic so that these limits are only enforced against regular backends,
and only regular backends are counted while checking if the limits
were already reached. Previously, background processes that had an
assigned database or role were subject to these limits (with rather
random exclusions for autovac workers and walsenders), and the set of
existing processes that counted against each limit was quite haphazard
as well. The point of these limits, AFAICS, is to ensure the
availability of PGPROC slots for regular backends. Since all other
types of processes have their own separate pools of PGPROC slots, it
makes no sense either to enforce these limits against them or to count
them while enforcing the limit.
While edge-case failures of these sorts have been possible for a
long time, the problem got a good deal worse with commit 5a2fed911
(CVE-2024-10978), which caused parallel workers to make some of these
checks using the leader's current role where before we had used its
AuthenticatedUserId, thus allowing parallel queries to fail after
SET ROLE. The previous behavior was fairly accidental and I have
no desire to return to it.
This patch includes reverting 73c9f91a1, which was an emergency hack
to suppress these same checks in some cases. It wasn't complete,
as shown by a recent bug report from Laurenz Albe. We can also revert
fd4d93d26 and 492217301, which hacked around the same problems in one
regression test.
In passing, remove the special case for autovac workers in
CheckMyDatabase; it seems cleaner to have AutoVacWorkerMain pass
the INIT_PG_OVERRIDE_ALLOW_CONNS flag, now that that does what's
needed.
Like 5a2fed911, back-patch to supported branches (which sadly no
longer includes v12).
Discussion: https://postgr.es/m/1808397.1735156190@sss.pgh.pa.us
The need for this was missed in commit 93db6cbda, with the result
being that if we launch a slotsync worker it would consume one of
the PGPROCs in the max_connections pool. That could lead to inability
to launch the worker, or to subsequent failures of connection requests
that should have succeeded according to the configured settings.
Rather than create some one-off infrastructure to support this,
let's group the slotsync worker with the existing autovac launcher
in a new category of "special worker" processes. These are kind of
like auxiliary processes, but they cannot use that infrastructure
because they need to be able to run transactions.
For the moment, make these processes share the PGPROC freelist
used for autovac workers (which previously supplied the autovac
launcher too). This is partly to avoid an ABI change in v17,
and partly because it seems silly to have a freelist with
at most two members. This might be worth revisiting if we grow
enough workers in this category.
Tom Lane and Hou Zhijie. Back-patch to v17.
Discussion: https://postgr.es/m/1808397.1735156190@sss.pgh.pa.us
RelationTruncate() does three things, while holding an
AccessExclusiveLock and preventing checkpoints:
1. Logs the truncation.
2. Drops buffers, even if they're dirty.
3. Truncates some number of files.
Step 2 could previously be canceled if it had to wait for I/O, and step
3 could and still can fail in file APIs. All orderings of these
operations have data corruption hazards if interrupted, so we can't give
up until the whole operation is done. When dirty pages were discarded
but the corresponding blocks were left on disk due to ERROR, old page
versions could come back from disk, reviving deleted data (see
pgsql-bugs #18146 and several like it). When primary and standby were
allowed to disagree on relation size, standbys could panic (see
pgsql-bugs #18426) or revive data unknown to visibility management on
the primary (theorized).
Changes:
* WAL is now unconditionally flushed first
* smgrtruncate() is now called in a critical section, preventing
interrupts and causing PANIC on file API failure
* smgrtruncate() has a new parameter for existing fork sizes,
because it can't call smgrnblocks() itself inside a critical section
The changes apply to RelationTruncate(), smgr_redo() and
pg_truncate_visibility_map(). That last is also brought up to date with
other evolutions of the truncation protocol.
The VACUUM FileTruncate() failure mode had been discussed in older
reports than the ones referenced below, with independent analysis from
many people, but earlier theories on how to fix it were too complicated
to back-patch. The more recently invented cancellation bug was
diagnosed by Alexander Lakhin. Other corruption scenarios were spotted
by me while iterating on this patch and earlier commit 75818b3a.
Back-patch to all supported releases.
Reviewed-by: Michael Paquier <michael@paquier.xyz>
Reviewed-by: Robert Haas <robertmhaas@gmail.com>
Reported-by: rootcause000@gmail.com
Reported-by: Alexander Lakhin <exclusion@gmail.com>
Discussion: https://postgr.es/m/18146-04e908c662113ad5%40postgresql.org
Discussion: https://postgr.es/m/18426-2d18da6586f152d6%40postgresql.org
An inplace update's invalidation messages are part of its transaction's
commit record. However, the update survives even if its transaction
aborts or we stop recovery before replaying its transaction commit.
After recovery, a backend that started in recovery could update the row
without incorporating the inplace update. That could result in a table
with an index, yet relhasindex=f. That is a source of index corruption.
This bulk invalidation avoids the functional consequences. A future
change can fix the !RecoveryInProgress() scenario without changing the
WAL format. Back-patch to v17 - v12 (all supported versions). v18 will
instead add invalidations to WAL.
Discussion: https://postgr.es/m/20240618152349.7f.nmisch@google.com
The previous commit fixed some ways of losing an inplace update. It
remained possible to lose one when a backend working toward a
heap_update() copied a tuple into memory just before inplace update of
that tuple. In catalogs eligible for inplace update, use LOCKTAG_TUPLE
to govern admission to the steps of copying an old tuple, modifying it,
and issuing heap_update(). This includes MERGE commands. To avoid
changing most of the pg_class DDL, don't require LOCKTAG_TUPLE when
holding a relation lock sufficient to exclude inplace updaters.
Back-patch to v12 (all supported versions). In v13 and v12, "UPDATE
pg_class" or "UPDATE pg_database" can still lose an inplace update. The
v14+ UPDATE fix needs commit 86dc90056d,
and it wasn't worth reimplementing that fix without such infrastructure.
Reviewed by Nitin Motiani and (in earlier versions) Heikki Linnakangas.
Discussion: https://postgr.es/m/20231027214946.79.nmisch@google.com
Commits 041b9680 and 6377e12a changed the interface of
scan_analyze_next_block() to take a ReadStream instead of a BlockNumber
and a BufferAccessStrategy, and to return a value to indicate when the
stream has run out of blocks.
This caused integration problems for at least one known extension that
uses specially encoded BlockNumber values that map to different
underlying storage, because acquire_sample_rows() sets up the stream so
that read_stream_next_buffer() reads blocks from the main fork of the
relation's SMgrRelation.
Provide read_stream_next_block(), as a way for such an extension to
access the stream of raw BlockNumbers directly and forward them to its
own ReadBuffer() calls after decoding, as it could in earlier releases.
The new function returns the BlockNumber and BufferAccessStrategy that
were previously passed directly to scan_analyze_next_block().
Alternatively, an extension could wrap the stream of BlockNumbers in
another ReadStream with a callback that performs any decoding required
to arrive at real storage manager BlockNumber values, so that it could
benefit from the I/O combining and concurrency provided by
read_stream.c.
Another class of table access method that does nothing in
scan_analyze_next_block() because it is not block-oriented could use
this function to control the number of block sampling loops. It could
match the previous behavior with "return read_stream_next_block(stream,
&bas) != InvalidBlockNumber".
Ongoing work is expected to provide better ANALYZE support for table
access methods that don't behave like heapam with respect to storage
blocks, but that will be for future releases.
Back-patch to 17.
Reported-by: Mats Kindahl <mats@timescale.com>
Reviewed-by: Mats Kindahl <mats@timescale.com>
Discussion: https://postgr.es/m/CA%2B14425%2BCcm07ocG97Fp%2BFrD9xUXqmBKFvecp0p%2BgV2YYR258Q%40mail.gmail.com
Commit 5b562644fe added a comment that
SetRelationHasSubclass() callers must hold this lock. When commit
17f206fbc8 extended use of this column to
partitioned indexes, it didn't take the lock. As the latter commit
message mentioned, we currently never reset a partitioned index to
relhassubclass=f. That largely avoids harm from the lock omission. The
cause for fixing this now is to unblock introducing a rule about locks
required to heap_update() a pg_class row. This might cause more
deadlocks. It gives minor user-visible benefits:
- If an ALTER INDEX SET TABLESPACE runs concurrently with ALTER TABLE
ATTACH PARTITION or CREATE PARTITION OF, one transaction blocks
instead of failing with "tuple concurrently updated". (Many cases of
DDL concurrency still fail that way.)
- Match ALTER INDEX ATTACH PARTITION in choosing to lock the index.
While not user-visible today, we'll need this if we ever make something
set the flag to false for a partitioned index, like ANALYZE does today
for tables. Back-patch to v12 (all supported versions), the plan for
the commit relying on the new rule. In back branches, add
LockOrStrongerHeldByMe() instead of adding a LockHeldByMe() parameter.
Reviewed (in an earlier version) by Robert Haas.
Discussion: https://postgr.es/m/20240611024525.9f.nmisch@google.com
We did not recover the subtransaction IDs of prepared transactions
when starting a hot standby from a shutdown checkpoint. As a result,
such subtransactions were considered as aborted, rather than
in-progress. That would lead to hint bits being set incorrectly, and
the subtransactions suddenly becoming visible to old snapshots when
the prepared transaction was committed.
To fix, update pg_subtrans with prepared transactions's subxids when
starting hot standby from a shutdown checkpoint. The snapshots taken
from that state need to be marked as "suboverflowed", so that we also
check the pg_subtrans.
Backport to all supported versions.
Discussion: https://www.postgresql.org/message-id/6b852e98-2d49-4ca1-9e95-db419a2696e0@iki.fi
Make sure that function declarations use names that exactly match the
corresponding names from function definitions in a few places. These
inconsistencies were all introduced during Postgres 17 development.
pg_bsd_indent still has a couple of similar inconsistencies, which I
(pgeoghegan) have left untouched for now.
This commit was written with help from clang-tidy, by mechanically
applying the same rules as similar clean-up commits (the earliest such
commit was commit 035ce1fe).
There were a few typedefs that were never used to define a variable or
field. This had the effect that the buildfarm's typedefs.list would
not include them, and so they would need to be re-added manually to
keep the overall pgindent result perfectly clean.
We can easily get rid of these typedefs to avoid the issue. In a few
cases, we just let the enum or struct stand on its own without a
typedef around it. In one case, an enum was abused to define flag
bits; that's better done with macro definitions.
This fixes all the remaining issues with missing entries in the
buildfarm's typedefs.list.
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://www.postgresql.org/message-id/1919000.1715815925@sss.pgh.pa.us
While pinning extra buffers to look ahead, users of strategies are in
danger of using too many buffers. For some strategies, that means
"escaping" from the ring, and in others it means forcing dirty data to
disk very frequently with associated WAL flushing. Since external code
has no insight into any of that, allow individual strategy types to
expose a clamp that should be applied when deciding how many buffers to
pin at once.
Reviewed-by: Andres Freund <andres@anarazel.de>
Reviewed-by: Melanie Plageman <melanieplageman@gmail.com>
Discussion: https://postgr.es/m/CAAKRu_aJXnqsyZt6HwFLnxYEBgE17oypkxbKbT1t1geE_wvH2Q%40mail.gmail.com
If temp tables have dependencies (such as sequences) then it's
possible for autovacuum's cleanup of orphan temp tables to deadlock
against an incoming backend that's trying to clean out the temp
namespace for its own use. That can happen because RemoveTempRelations'
performDeletion call can visit objects within the namespace in
an order different from the order in which a per-table deletion
will visit them.
To fix, observe that performDeletion will begin by taking an exclusive
lock on the temp namespace (even though it won't actually delete it).
So, if we can get a shared lock on the namespace, we can be sure we're
not running concurrently with RemoveTempRelations, while also not
conflicting with ordinary use of the namespace. This requires
introducing a conditional version of LockDatabaseObject, but that's no
big deal. (It's surprising we've got along without that this long.)
Report and patch by Mikhail Zhilin. Back-patch to all supported
branches.
Discussion: https://postgr.es/m/c43ce028-2bc2-4865-9b89-3f706246eed5@postgrespro.ru
Introduce an abstraction allowing relation data to be accessed as a
stream of buffers, with an implementation that is more efficient than
the equivalent sequence of ReadBuffer() calls.
Client code supplies a callback that can say which block number it wants
next, and then consumes individual buffers one at a time from the
stream. This division puts read_stream.c in control of how far ahead it
can see and allows it to read clusters of neighboring blocks with
StartReadBuffers(). It also issues POSIX_FADV_WILLNEED advice ahead of
time when random access is detected.
Other variants of I/O stream will be proposed in future work (for
example to support recovery, whose LsnReadQueue device in
xlogprefetcher.c is a distant cousin of this code and should eventually
be replaced by this), but this basic API is sufficient for many common
executor usage patterns involving predictable access to a single fork of
a single relation.
Several patches using this API are proposed separately.
This stream concept is loosely based on ideas from Andres Freund on how
we should pave the way for later work on asynchronous I/O.
Author: Thomas Munro <thomas.munro@gmail.com>
Author: Heikki Linnakangas <hlinnaka@iki.fi> (contributions)
Author: Melanie Plageman <melanieplageman@gmail.com> (contributions)
Suggested-by: Andres Freund <andres@anarazel.de>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Melanie Plageman <melanieplageman@gmail.com>
Reviewed-by: Nazir Bilal Yavuz <byavuz81@gmail.com>
Reviewed-by: Andres Freund <andres@anarazel.de>
Tested-by: Tomas Vondra <tomas.vondra@enterprisedb.com>
Discussion: https://postgr.es/m/CA+hUKGJkOiOCa+mag4BF+zHo7qo=o9CFheB8=g6uT5TUm2gkvA@mail.gmail.com
Break ReadBuffer() up into two steps. StartReadBuffers() and
WaitReadBuffers() give us two main advantages:
1. Multiple consecutive blocks can be read with one system call.
2. Advice (hints of future reads) can optionally be issued to the
kernel ahead of time.
The traditional ReadBuffer() function is now implemented in terms of
those functions, to avoid duplication.
A new GUC io_combine_limit is defined, and the functions for limiting
per-backend pin counts are made into public APIs. Those are provided
for use by callers of StartReadBuffers(), when deciding how many buffers
to read at once. The following commit will add a higher level mechanism
for doing that automatically with a practical interface.
With some more infrastructure in later work, StartReadBuffers() could
be extended to start real asynchronous I/O instead of just issuing
advice and leaving WaitReadBuffers() to do the work synchronously.
Author: Thomas Munro <thomas.munro@gmail.com>
Author: Andres Freund <andres@anarazel.de> (some optimization tweaks)
Reviewed-by: Melanie Plageman <melanieplageman@gmail.com>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Nazir Bilal Yavuz <byavuz81@gmail.com>
Reviewed-by: Dilip Kumar <dilipbalaut@gmail.com>
Reviewed-by: Andres Freund <andres@anarazel.de>
Tested-by: Tomas Vondra <tomas.vondra@enterprisedb.com>
Discussion: https://postgr.es/m/CA+hUKGJkOiOCa+mag4BF+zHo7qo=o9CFheB8=g6uT5TUm2gkvA@mail.gmail.com
Previously, we used a simple array for storing dead tuple IDs during
lazy vacuum, which had a number of problems:
* The array used a single allocation and so was limited to 1GB.
* The allocation was pessimistically sized according to table size.
* Lookup with binary search was slow because of poor CPU cache and
branch prediction behavior.
This commit replaces that array with the TID store from commit
30e144287a.
Since the backing radix tree makes small allocations as needed, the
1GB limit is now gone. Further, the total memory used is now often
smaller by an order of magnitude or more, depending on the
distribution of blocks and offsets. These two features should make
multiple rounds of heap scanning and index cleanup an extremely rare
event. TID lookup during index cleanup is also several times faster,
even more so when index order is correlated with heap tuple order.
Since there is no longer a predictable relationship between the number
of dead tuples vacuumed and the space taken up by their TIDs, the
number of tuples no longer provides any meaningful insights for users,
nor is the maximum number predictable. For that reason this commit
also changes to byte-based progress reporting, with the relevant
columns of pg_stat_progress_vacuum renamed accordingly to
max_dead_tuple_bytes and dead_tuple_bytes.
For parallel vacuum, both the TID store and supplemental information
specific to vacuum are shared among the parallel vacuum workers. As
with the previous array, we don't take any locks on TidStore during
parallel vacuum since writes are still only done by the leader
process.
Bump catalog version.
Reviewed-by: John Naylor, (in an earlier version) Dilip Kumar
Discussion: https://postgr.es/m/CAD21AoAfOZvmfR0j8VmZorZjL7RhTiQdVttNuC4W-Shdc2a-AA%40mail.gmail.com
This way, we can fold the list of lock names to occur in
BuiltinTrancheNames instead of having its own separate array. This
saves two lines of code in GetLWTrancheName and some space in
BuiltinTrancheNames, as foreseen in commit 74a7306310, as well as
removing the need for a separate lwlocknames.c file.
We still have to build lwlocknames.h using Perl code, which initially I
wanted to avoid, but it gives us the chance to cross-check
wait_event_names.txt.
Discussion: https://postgr.es/m/202401231025.gbv4nnte5fmm@alvherre.pgsql
We don't determine the position at which a process waiting for a lock
should insert itself into the wait queue until we reach ProcSleep(),
and we may at that point discover that we must insert ourselves ahead
of everyone who wants a conflicting lock, in which case we obtain the
lock immediately. Up until now, a no-wait lock acquisition would fail
in such cases, erroneously claiming that the lock couldn't be obtained
immediately. Fix that by trying ProcSleep even in the no-wait case.
No back-patch for now, because I'm treating this as an improvement to
the existing no-wait feature. It could instead be argued that it's a
bug fix, on the theory that there should never be any case whatsoever
where no-wait fails to obtain a lock that would have been obtained
immediately without no-wait, but I'm reluctant to interpret the
semantics of no-wait that strictly.
Robert Haas and Jingxian Li
Discussion: http://postgr.es/m/CA+TgmobCH-kMXGVpb0BB-iNMdtcNkTvcZ4JBxDJows3kYM+GDg@mail.gmail.com
Currently, pg_visibility computes its xid horizon using the
GetOldestNonRemovableTransactionId(). The problem is that this horizon can
sometimes go backward. That can lead to reporting false errors.
In order to fix that, this commit implements a new function
GetStrictOldestNonRemovableTransactionId(). This function computes the xid
horizon, which would be guaranteed to be newer or equal to any xid horizon
computed before.
We have to do the following to achieve this.
1. Ignore processes xmin's, because they consider connection to other databases
that were ignored before.
2. Ignore KnownAssignedXids, because they are not database-aware. At the same
time, the primary could compute its horizons database-aware.
3. Ignore walsender xmin, because it could go backward if some replication
connections don't use replication slots.
As a result, we're using only currently running xids to compute the horizon.
Surely these would significantly sacrifice accuracy. But we have to do so to
avoid reporting false errors.
Inspired by earlier patch by Daniel Shelepanov and the following discussion
with Robert Haas and Tom Lane.
Discussion: https://postgr.es/m/1649062270.289865713%40f403.i.mail.ru
Reviewed-by: Alexander Lakhin, Dmitry Koval
Now that BackendId was just another index into the proc array, it was
redundant with the 0-based proc numbers used in other places. Replace
all usage of backend IDs with proc numbers.
The only place where the term "backend id" remains is in a few pgstat
functions that expose backend IDs at the SQL level. Those IDs are now
in fact 0-based ProcNumbers too, but the documentation still calls
them "backend ids". That term still seems appropriate to describe what
the numbers are, so I let it be.
One user-visible effect is that pg_temp_0 is now a valid temp schema
name, for backend with ProcNumber 0.
Reviewed-by: Andres Freund
Discussion: https://www.postgresql.org/message-id/8171f1aa-496f-46a6-afc3-c46fe7a9b407@iki.fi
Previously, backend ID was an index into the ProcState array, in the
shared cache invalidation manager (sinvaladt.c). The entry in the
ProcState array was reserved at backend startup by scanning the array
for a free entry, and that was also when the backend got its backend
ID. Things become slightly simpler if we redefine backend ID to be the
index into the PGPROC array, and directly use it also as an index to
the ProcState array. This uses a little more memory, as we reserve a
few extra slots in the ProcState array for aux processes that don't
need them, but the simplicity is worth it.
Aux processes now also have a backend ID. This simplifies the
reservation of BackendStatusArray and ProcSignal slots.
You can now convert a backend ID into an index into the PGPROC array
simply by subtracting 1. We still use 0-based "pgprocnos" in various
places, for indexes into the PGPROC array, but the only difference now
is that backend IDs start at 1 while pgprocnos start at 0. (The next
commmit will get rid of the term "backend ID" altogether and make
everything 0-based.)
There is still a 'backendId' field in PGPROC, now part of 'vxid' which
encapsulates the backend ID and local transaction ID together. It's
needed for prepared xacts. For regular backends, the backendId is
always equal to pgprocno + 1, but for prepared xact PGPROC entries,
it's the ID of the original backend that processed the transaction.
Reviewed-by: Andres Freund, Reid Thompson
Discussion: https://www.postgresql.org/message-id/8171f1aa-496f-46a6-afc3-c46fe7a9b407@iki.fi
In the past, FileRead() and FileWrite() used types based on the Unix
read() and write() functions from before C and POSIX standardization,
though not exactly (we had int for amount instead of unsigned). In
commit 2d4f1ba6 we changed to the appropriate standard C types, just
like the modern POSIX functions they wrap, but again not exactly: the
return type stayed as int. In theory, a ssize_t value could be returned
by the underlying call that is too large for an int.
That wasn't really a live bug, because we don't expect PostgreSQL code
to perform reads or writes of gigabytes, and OSes probably apply
internal caps smaller than that anyway. This change is done on the
principle that the return might as well follow the standard interfaces
consistently.
Reported-by: Tom Lane <tgl@sss.pgh.pa.us>
Reviewed-by: Peter Eisentraut <peter@eisentraut.org>
Discussion: https://postgr.es/m/1672202.1703441340%40sss.pgh.pa.us
More precisely, what we do here is make the SLRU cache sizes
configurable with new GUCs, so that sites with high concurrency and big
ranges of transactions in flight (resp. multixacts/subtransactions) can
benefit from bigger caches. In order for this to work with good
performance, two additional changes are made:
1. the cache is divided in "banks" (to borrow terminology from CPU
caches), and algorithms such as eviction buffer search only affect
one specific bank. This forestalls the problem that linear searching
for a specific buffer across the whole cache takes too long: we only
have to search the specific bank, whose size is small. This work is
authored by Andrey Borodin.
2. Change the locking regime for the SLRU banks, so that each bank uses
a separate LWLock. This allows for increased scalability. This work
is authored by Dilip Kumar. (A part of this was previously committed as
d172b717c6f4.)
Special care is taken so that the algorithms that can potentially
traverse more than one bank release one bank's lock before acquiring the
next. This should happen rarely, but particularly clog.c's group commit
feature needed code adjustment to cope with this. I (Álvaro) also added
lots of comments to make sure the design is sound.
The new GUCs match the names introduced by bcdfa5f2e2 in the
pg_stat_slru view.
The default values for these parameters are similar to the previous
sizes of each SLRU. commit_ts, clog and subtrans accept value 0, which
means to adjust by dividing shared_buffers by 512 (so 2MB for every 1GB
of shared_buffers), with a cap of 8MB. (A new slru.c function
SimpleLruAutotuneBuffers() was added to support this.) The cap was
previously 1MB for clog, so for sites with more than 512MB of shared
memory the total memory used increases, which is likely a good tradeoff.
However, other SLRUs (notably multixact ones) retain smaller sizes and
don't support a configured value of 0. These values based on
shared_buffers may need to be revisited, but that's an easy change.
There was some resistance to adding these new GUCs: it would be better
to adjust to memory pressure automatically somehow, for example by
stealing memory from shared_buffers (where the caches can grow and
shrink naturally). However, doing that seems to be a much larger
project and one which has made virtually no progress in several years,
and because this is such a pain point for so many users, here we take
the pragmatic approach.
Author: Andrey Borodin <x4mmm@yandex-team.ru>
Author: Dilip Kumar <dilipbalaut@gmail.com>
Reviewed-by: Amul Sul, Gilles Darold, Anastasia Lubennikova,
Ivan Lazarev, Robert Haas, Thomas Munro, Tomas Vondra,
Yura Sokolov, Васильев Дмитрий (Dmitry Vasiliev).
Discussion: https://postgr.es/m/2BEC2B3F-9B61-4C1D-9FB5-5FAB0F05EF86@yandex-team.ru
Discussion: https://postgr.es/m/CAFiTN-vzDvNz=ExGXz6gdyjtzGixKSqs0mKHMmaQ8sOSEFZ33A@mail.gmail.com
There isn't a lot of user demand for AIX support, we have a bunch of
hacks to work around AIX-specific compiler bugs and idiosyncrasies,
and no one has stepped up to the plate to properly maintain it.
Remove support for AIX to get rid of that maintenance overhead. It's
still supported for stable versions.
The acute issue that triggered this decision was that after commit
8af2565248, the AIX buildfarm members have been hitting this
assertion:
TRAP: failed Assert("(uintptr_t) buffer == TYPEALIGN(PG_IO_ALIGN_SIZE, buffer)"), File: "md.c", Line: 472, PID: 2949728
Apperently the "pg_attribute_aligned(a)" attribute doesn't work on AIX
for values larger than PG_IO_ALIGN_SIZE, for a static const variable.
That could be worked around, but we decided to just drop the AIX support
instead.
Discussion: https://www.postgresql.org/message-id/20240224172345.32@rfd.leadboat.com
Reviewed-by: Andres Freund, Noah Misch, Thomas Munro
The new facility makes it easier to optimize bulk loading, as the
logic for buffering, WAL-logging, and syncing the relation only needs
to be implemented once. It's also less error-prone: We have had a
number of bugs in how a relation is fsync'd - or not - at the end of a
bulk loading operation. By centralizing that logic to one place, we
only need to write it correctly once.
The new facility is faster for small relations: Instead of of calling
smgrimmedsync(), we register the fsync to happen at next checkpoint,
which avoids the fsync latency. That can make a big difference if you
are e.g. restoring a schema-only dump with lots of relations.
It is also slightly more efficient with large relations, as the WAL
logging is performed multiple pages at a time. That avoids some WAL
header overhead. The sorted GiST index build did that already, this
moves the buffering to the new facility.
The changes to pageinspect GiST test needs an explanation: Before this
patch, the sorted GiST index build set the LSN on every page to the
special GistBuildLSN value, not the LSN of the WAL record, even though
they were WAL-logged. There was no particular need for it, it just
happened naturally when we wrote out the pages before WAL-logging
them. Now we WAL-log the pages first, like in B-tree build, so the
pages are stamped with the record's real LSN. When the build is not
WAL-logged, we still use GistBuildLSN. To make the test output
predictable, use an unlogged index.
Reviewed-by: Andres Freund
Discussion: https://www.postgresql.org/message-id/30e8f366-58b3-b239-c521-422122dd5150%40iki.fi
After calling smgropen(), it was not clear how long you could continue
to use the result, because various code paths including cache
invalidation could call smgrclose(), which freed the memory.
Guarantee that the object won't be destroyed until the end of the
current transaction, or in recovery, the commit/abort record that
destroys the underlying storage.
smgrclose() is now just an alias for smgrrelease(). It closes files
and forgets all state except the rlocator, but keeps the SMgrRelation
object valid.
A new smgrdestroy() function is used by rare places that know there
should be no other references to the SMgrRelation.
The short version:
* smgrclose() is now just an alias for smgrrelease(). It releases
resources, but doesn't destroy until EOX
* smgrdestroy() now frees memory, and should rarely be used.
Existing code should be unaffected, but it is now possible for code that
has an SMgrRelation object to use it repeatedly during a transaction as
long as the storage hasn't been physically dropped. Such code would
normally hold a lock on the relation.
This also replaces the "ownership" mechanism of SMgrRelations with a
pin counter. An SMgrRelation can now be "pinned", which prevents it
from being destroyed at end of transaction. There can be multiple pins
on the same SMgrRelation. In practice, the pin mechanism is only used
by the relcache, so there cannot be more than one pin on the same
SMgrRelation. Except with swap_relation_files XXX
Author: Thomas Munro, Heikki Linnakangas
Reviewed-by: Robert Haas <robertmhaas@gmail.com>
Discussion: https://www.postgresql.org/message-id/CA%2BhUKGJ8NTvqLHz6dqbQnt2c8XCki4r2QvXjBQcXpVwxTY_pvA@mail.gmail.com
Presently, the most straightforward way for a shared library to use
shared memory is to request it at server startup via a
shmem_request_hook, which requires specifying the library in
shared_preload_libraries. Alternatively, the library can create a
dynamic shared memory (DSM) segment, but absent a shared location
to store the segment's handle, other backends cannot use it. This
commit introduces a registry for DSM segments so that these other
backends can look up existing segments with a library-specified
string. This allows libraries to easily use shared memory without
needing to request it at server startup.
The registry is accessed via the new GetNamedDSMSegment() function.
This function handles allocating the segment and initializing it
via a provided callback. If another backend already created and
initialized the segment, it simply attaches the segment.
GetNamedDSMSegment() locks the registry appropriately to ensure
that only one backend initializes the segment and that all other
backends just attach it.
The registry itself is comprised of a dshash table that stores the
DSM segment handles keyed by a library-specified string.
Reviewed-by: Michael Paquier, Andrei Lepikhov, Nikita Malakhov, Robert Haas, Bharath Rupireddy, Zhang Mingli, Amul Sul
Discussion: https://postgr.es/m/20231205034647.GA2705267%40nathanxps13
jit.c and dfgr.c had a copy of the same code to check if a file exists
or not, with a twist: jit.c did not check for EACCES when failing the
stat() call for the path whose existence is tested. This refactored
routine will be used by an upcoming patch.
Reviewed-by: Ashutosh Bapat
Discussion: https://postgr.es/m/ZTiV8tn_MIb_H2rE@paquier.xyz
Both lwlocknames.txt and wait_event_names.txt contain a list of all
the predefined LWLocks, i.e., those with predefined positions
within MainLWLockArray. It is easy to miss one or the other,
especially since the list in wait_event_names.txt omits the "Lock"
suffix from all the LWLock wait events. This commit adds a cross-
check of these lists to the script that generates lwlocknames.h.
If the lists do not match exactly, building will fail.
Suggested-by: Robert Haas
Reviewed-by: Robert Haas, Michael Paquier, Bertrand Drouvot
Discussion: https://postgr.es/m/20240102173120.GA1061678%40nathanxps13
When active, this process writes WAL summary files to
$PGDATA/pg_wal/summaries. Each summary file contains information for a
certain range of LSNs on a certain TLI. For each relation, it stores a
"limit block" which is 0 if a relation is created or destroyed within
a certain range of WAL records, or otherwise the shortest length to
which the relation was truncated during that range of WAL records, or
otherwise InvalidBlockNumber. In addition, it stores a list of blocks
which have been modified during that range of WAL records, but
excluding blocks which were removed by truncation after they were
modified and never subsequently modified again.
In other words, it tells us which blocks need to copied in case of an
incremental backup covering that range of WAL records. But this
doesn't yet add the capability to actually perform an incremental
backup; the next patch will do that.
A new parameter summarize_wal enables or disables this new background
process. The background process also automatically deletes summary
files that are older than wal_summarize_keep_time, if that parameter
has a non-zero value and the summarizer is configured to run.
Patch by me, with some design help from Dilip Kumar and Andres Freund.
Reviewed by Matthias van de Meent, Dilip Kumar, Jakub Wartak, Peter
Eisentraut, and Álvaro Herrera.
Discussion: http://postgr.es/m/CA+TgmoYOYZfMCyOXFyC-P+-mdrZqm5pP2N7S-r0z3_402h9rsA@mail.gmail.com
smgrreadv() and smgrwritev() and their md.c implementations call
FileReadV() and FileWriteV(). A range of disk blocks beginning at
'blocknum' and extending for 'nblocks' can be scattered to or gathered
from multiple buffers with a single system call. The traditional
smgrread() and smgrwrite() functions are implemented in terms of the new
functions.
Later commits will introduce calls with nblocks > 1, but the following
behavioral changes can be seen already:
* After a short transfer we'll now retry until we eventually read 0
bytes (= EOF) or get ENOSPC, EDQUOT, EFBIG etc, where previously we
would infer the reason. Retrying is consistent with xlog.c's
treatment of large WAL writes, and arguably also xlog.c and fd.c's
treatment of EINTR. Arbitrary short returns for larger transfers have
been observed on several OSes, and might in theory also happen for
transient reasons with our own pg_p*v() fallback code.
* After unexpected EOF or -1, the error thrown now talks about
a range even for the single block case, eg "blocks 42..42".
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/CA+hUKGJkOiOCa+mag4BF+zHo7qo=o9CFheB8=g6uT5TUm2gkvA@mail.gmail.com
Previously smgrprefetch() could issue POSIX_FADV_WILLNEED advice for a
single block at a time. Add an nblocks argument so that we can do the
same for a range of blocks. This usually produces a single system call,
but might need to loop if it crosses a segment boundary. Initially it
is only called with nblocks == 1, but proposed patches will make wider
calls.
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi> (earlier version)
Discussion: https://postgr.es/m/CA+hUKGJkOiOCa+mag4BF+zHo7qo=o9CFheB8=g6uT5TUm2gkvA@mail.gmail.com
FileReadV() and FileWriteV() adapt pg_preadv() and pg_pwritev() for
fd.c's virtual file descriptors. The simple FileRead() and FileWrite()
functions are now implemented in terms of the vectored functions, to
avoid code duplication, and they are converted back to the corresponding
simple system calls further down (commit 15c9ac36). Later work will
make more interesting multi-iovec calls.
The traditional behavior of reporting a "fake" ENOSPC error is
simplified. It's now always set for non-failing writes, for the benefit
of callers that expect to log a meaningful "%m" if they determine that
the write was short. (Perhaps we should consider getting rid of that
expectation one day.)
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/CA+hUKGJkOiOCa+mag4BF+zHo7qo=o9CFheB8=g6uT5TUm2gkvA@mail.gmail.com
For clarity, have separate functions for *creating* the shared memory
and semaphores at postmaster or single-user backend startup, and
for *attaching* to existing shared memory structures in EXEC_BACKEND
case. CreateSharedMemoryAndSemaphores() is now called only at
postmaster startup, and a new AttachSharedMemoryStructs() function is
called at backend startup in EXEC_BACKEND mode.
Reviewed-by: Tristan Partin, Andres Freund
Discussion: https://www.postgresql.org/message-id/7a59b073-5b5b-151e-7ed3-8b01ff7ce9ef@iki.fi
We've had repeated bugs in the area of handling SLRU wraparound in the past,
some of which have caused data loss. Switching to an indexing system for SLRUs
that does not wrap around should allow us to get rid of a whole bunch
of problems and improve the overall reliability of the system.
This particular patch however only changes the indexing and doesn't address
the wraparound per se. This is going to be done in the following patches.
Author: Maxim Orlov, Aleksander Alekseev, Alexander Korotkov, Teodor Sigaev
Author: Nikita Glukhov, Pavel Borisov, Yura Sokolov
Reviewed-by: Jacob Champion, Heikki Linnakangas, Alexander Korotkov
Reviewed-by: Japin Li, Pavel Borisov, Tom Lane, Peter Eisentraut, Andres Freund
Reviewed-by: Andrey Borodin, Dilip Kumar, Aleksander Alekseev
Discussion: https://postgr.es/m/CACG%3DezZe1NQSCnfHOr78AtAZxJZeCvxrts0ygrxYwe%3DpyyjVWA%40mail.gmail.com
Discussion: https://postgr.es/m/CAJ7c6TPDOYBYrnCAeyndkBktO0WG2xSdYduTF0nxq%2BvfkmTF5Q%40mail.gmail.com
A WaitEventSet holds file descriptors or event handles (on Windows).
If FreeWaitEventSet is not called, those fds or handles are leaked.
Use ResourceOwners to track WaitEventSets, to clean those up
automatically on error.
This was a live bug in async Append nodes, if a FDW's
ForeignAsyncRequest function failed. (In back branches, I will apply a
more localized fix for that based on PG_TRY-PG_FINALLY.)
The added test doesn't check for leaking resources, so it passed even
before this commit. But at least it covers the code path.
In the passing, fix misleading comment on what the 'nevents' argument
to WaitEventSetWait means.
Report by Alexander Lakhin, analysis and suggestion for the fix by
Tom Lane. Fixes bug #17828.
Reviewed-by: Alexander Lakhin, Thomas Munro
Discussion: https://www.postgresql.org/message-id/472235.1678387869@sss.pgh.pa.us
The code previously relied on "long" as type to track block numbers,
which would be 4 bytes in all Windows builds or any 32-bit builds. This
limited the code to be able to handle up to 16TB of data with the
default block size of 8kB, like during a CLUSTER. This code now relies
on a more portable int64, which should be more than enough for at least
the next 20 years to come.
This issue has been reported back in 2017, but nothing was done about it
back then, so here we go now.
Reported-by: Peter Geoghegan
Reviewed-by: Heikki Linnakangas
Discussion: https://postgr.es/m/CAH2-WznCscXnWmnj=STC0aSa7QG+BRedDnZsP=Jo_R9GUZvUrg@mail.gmail.com
