The current definition of init_spin_delay (introduced recently in
48354581a) wasn't C89 compliant. It's not legal to refer to refer to
non-constant expressions, and the ptr argument was one. This, as
reported by Tom, lead to a failure on buildfarm animal pademelon.
The pointer, especially on system systems with ASLR, isn't super helpful
anyway, though. So instead of making init_spin_delay into an inline
function, make s_lock_stuck() report the function name in addition to
file:line and change init_spin_delay() accordingly. While not a direct
replacement, the function name is likely more useful anyway (line
numbers are often hard to interpret in third party reports).
This also fixes what file/line number is reported for waits via
s_lock().
As PG_FUNCNAME_MACRO is now used outside of elog.h, move it to c.h.
Reported-By: Tom Lane
Discussion: 4369.1460435533@sss.pgh.pa.us
The recent patch to make Pin/UnpinBuffer lockfree in the hot
path (48354581a), accidentally used pg_atomic_fetch_or_u32() in
MarkLocalBufferDirty(). Other code operating on local buffers was
careful to only use pg_atomic_read/write_u32 which just read/write from
memory; to avoid unnecessary overhead.
On its own that'd just make MarkLocalBufferDirty() slightly less
efficient, but in addition InitLocalBuffers() doesn't call
pg_atomic_init_u32() - thus the spinlock fallback for the atomic
operations isn't initialized. That in turn caused, as reported by Tom,
buildfarm animal gaur to fail. As those errors are actually useful
against this type of error, continue to omit - intentionally this time -
initialization of the atomic variable.
In addition, add an explicit note about only using pg_atomic_read/write
on local buffers's state to BufferDesc's description.
Reported-By: Tom Lane
Discussion: 1881.1460431476@sss.pgh.pa.us
It's silly to define these counts as narrower than they might someday
need to be. Also, I believe that the BLCKSZ * nflush calculation in
mdwriteback was capable of overflowing an int.
Commit 428b1d6b29 introduced the use of
msync() for flushing dirty data from the kernel's file buffers. Several
portability issues were overlooked, though:
* Not all implementations of mmap() think that nbytes == 0 means "map
the whole file". To fix, use lseek() to find out the true length.
Fix callers of pg_flush_data to be aware that nbytes == 0 may result
in trashing the file's seek position.
* Not all implementations of mmap() will accept partial-page mmap
requests. To fix, round down the length request to whatever sysconf()
says the page size is. (I think this is OK from a portability standpoint,
because sysconf() is required by SUS v2, and we aren't trying to compile
this part on Windows anyway. Buildfarm should let us know if not.)
* On 32-bit machines, the file size might exceed the available free
address space, or even exceed what will fit in size_t. Check for
the latter explicitly to avoid passing a false request size to mmap().
If mmap fails, silently fall through to the next implementation method,
rather than bleating to the postmaster log and giving up.
* mmap'ing directories fails on some platforms, and even if it works,
msync'ing the directory is quite unlikely to help, as for that matter are
the other flush implementations. In pre_sync_fname(), just skip flush
attempts on directories.
In passing, copy-edit the comments a bit.
Stas Kelvich and myself
On a big NUMA machine with 1000 connections in saturation load
there was a performance regression due to spinlock contention, for
acquiring values which were never used. Just fill with dummy
values if we're not going to use them.
This patch has not been benchmarked yet on a big NUMA machine, but
it seems like a good idea on general principle, and it seemed to
prevent an apparent 2.2% regression on a single-socket i7 box
running 200 connections at saturation load.
I was initially concerned that the some of the hundreds of
references to BufferGetPage() where the literal
BGP_NO_SNAPSHOT_TEST were passed might not optimize as well as a
macro, leading to some hard-to-find performance regressions in
corner cases. Inspection of disassembled code has shown identical
code at all inspected locations, and the size difference doesn't
amount to even one byte per such call. So make it readable.
Per gripes from Álvaro Herrera and Tom Lane
Previously we used a spinlock, in adition to the atomically manipulated
->state field, to protect the wait queue. But it's pretty simple to
instead perform the locking using a flag in state.
Due to 6150a1b0 BufferDescs, on platforms (like PPC) with > 1 byte
spinlocks, increased their size above 64byte. As 64 bytes are the size
we pad allocated BufferDescs to, this can increase false sharing;
causing performance problems in turn. Together with the previous commit
this reduces the size to <= 64 bytes on all common platforms.
Author: Andres Freund
Discussion: CAA4eK1+ZeB8PMwwktf+3bRS0Pt4Ux6Rs6Aom0uip8c6shJWmyg@mail.gmail.com20160327121858.zrmrjegmji2ymnvr@alap3.anarazel.de
Pinning/Unpinning a buffer is a very frequent operation; especially in
read-mostly cache resident workloads. Benchmarking shows that in various
scenarios the spinlock protecting a buffer header's state becomes a
significant bottleneck. The problem can be reproduced with pgbench -S on
larger machines, but can be considerably worse for queries which touch
the same buffers over and over at a high frequency (e.g. nested loops
over a small inner table).
To allow atomic operations to be used, cram BufferDesc's flags,
usage_count, buf_hdr_lock, refcount into a single 32bit atomic variable;
that allows to manipulate them together using 32bit compare-and-swap
operations. This requires reducing MAX_BACKENDS to 2^18-1 (which could
be lifted by using a 64bit field, but it's not a realistic configuration
atm).
As not all operations can easily implemented in a lockfree manner,
implement the previous buf_hdr_lock via a flag bit in the atomic
variable. That way we can continue to lock the header in places where
it's needed, but can get away without acquiring it in the more frequent
hot-paths. There's some additional operations which can be done without
the lock, but aren't in this patch; but the most important places are
covered.
As bufmgr.c now essentially re-implements spinlocks, abstract the delay
logic from s_lock.c into something more generic. It now has already two
users, and more are coming up; there's a follupw patch for lwlock.c at
least.
This patch is based on a proof-of-concept written by me, which Alexander
Korotkov made into a fully working patch; the committed version is again
revised by me. Benchmarking and testing has, amongst others, been
provided by Dilip Kumar, Alexander Korotkov, Robert Haas.
On a large x86 system improvements for readonly pgbench, with a high
client count, of a factor of 8 have been observed.
Author: Alexander Korotkov and Andres Freund
Discussion: 2400449.GjM57CE0Yg@dinodell
This feature is controlled by a new old_snapshot_threshold GUC. A
value of -1 disables the feature, and that is the default. The
value of 0 is just intended for testing. Above that it is the
number of minutes a snapshot can reach before pruning and vacuum
are allowed to remove dead tuples which the snapshot would
otherwise protect. The xmin associated with a transaction ID does
still protect dead tuples. A connection which is using an "old"
snapshot does not get an error unless it accesses a page modified
recently enough that it might not be able to produce accurate
results.
This is similar to the Oracle feature, and we use the same SQLSTATE
and error message for compatibility.
This patch is a no-op patch which is intended to reduce the chances
of failures of omission once the functional part of the "snapshot
too old" patch goes in. It adds parameters for snapshot, relation,
and an enum to specify whether the snapshot age check needs to be
done for the page at this point. This initial patch passes NULL
for the first two new parameters and BGP_NO_SNAPSHOT_TEST for the
third. The follow-on patch will change the places where the test
needs to be made.
Contention on the relation extension lock can become quite fierce when
multiple processes are inserting data into the same relation at the same
time at a high rate. Experimentation shows the extending the relation
multiple blocks at a time improves scalability.
Dilip Kumar, reviewed by Petr Jelinek, Amit Kapila, and me.
Experimentation shows this only costs about 6kB, which seems well
worth it given the major performance effects that can be caused
by insufficient alignment, especially on larger systems.
Discussion: 14166.1458924422@sss.pgh.pa.us
If archive_timeout > 0 we should avoid logging XLOG_RUNNING_XACTS if idle.
Bug 13685 reported by Laurence Rowe, investigated in detail by Michael Paquier,
though this is not his proposed fix.
20151016203031.3019.72930@wrigleys.postgresql.org
Simple non-invasive patch to allow later backpatch to 9.4 and 9.5
Commit ac1d794 ("Make idle backends exit if the postmaster dies.")
introduced a regression on, at least, large linux systems. Constantly
adding the same postmaster_alive_fds to the OSs internal datastructures
for implementing poll/select can cause significant contention; leading
to a performance regression of nearly 3x in one example.
This can be avoided by using e.g. linux' epoll, which avoids having to
add/remove file descriptors to the wait datastructures at a high rate.
Unfortunately the current latch interface makes it hard to allocate any
persistent per-backend resources.
Replace, with a backward compatibility layer, WaitLatchOrSocket with a
new WaitEventSet API. Users can allocate such a Set across multiple
calls, and add more than one file-descriptor to wait on. The latter has
been added because there's upcoming postgres features where that will be
helpful.
In addition to the previously existing poll(2), select(2),
WaitForMultipleObjects() implementations also provide an epoll_wait(2)
based implementation to address the aforementioned performance
problem. Epoll is only available on linux, but that is the most likely
OS for machines large enough (four sockets) to reproduce the problem.
To actually address the aforementioned regression, create and use a
long-lived WaitEventSet for FE/BE communication. There are additional
places that would benefit from a long-lived set, but that's a task for
another day.
Thanks to Amit Kapila, who helped make the windows code I blindly wrote
actually work.
Reported-By: Dmitry Vasilyev Discussion:
CAB-SwXZh44_2ybvS5Z67p_CDz=XFn4hNAD=CnMEF+QqkXwFrGg@mail.gmail.com20160114143931.GG10941@awork2.anarazel.de
Previously latches for windows and unix had been implemented in
different files. A later patch introduce an expanded wait
infrastructure, keeping the implementation separate would introduce too
much duplication.
This basically just moves the functions, without too much change. The
reason to keep this separate is that it allows blame to continue working
a little less badly; and to make review a tiny bit easier.
Discussion: 20160114143931.GG10941@awork2.anarazel.de
Buildfarm members gaur and pademelon are old enough not to know about
MAP_FAILED; which is used in 428b1d6. Include portability/mem.h to fix;
as already done in a bunch of other places.
The old code is wrong, because it returns a pointer to an automatic
variable. And it's also more clever than we really need to be
considering that the case it's worrying about should never happen.
Up to now checkpoints were written in the order they're in the
BufferDescriptors. That's nearly random in a lot of cases, which
performs badly on rotating media, but even on SSDs it causes slowdowns.
To avoid that, sort checkpoints before writing them out. We currently
sort by tablespace, relfilenode, fork and block number.
One of the major reasons that previously wasn't done, was fear of
imbalance between tablespaces. To address that balance writes between
tablespaces.
The other prime concern was that the relatively large allocation to sort
the buffers in might fail, preventing checkpoints from happening. Thus
pre-allocate the required memory in shared memory, at server startup.
This particularly makes it more efficient to have checkpoint flushing
enabled, because that'll often result in a lot of writes that can be
coalesced into one flush.
Discussion: alpine.DEB.2.10.1506011320000.28433@sto
Author: Fabien Coelho and Andres Freund
Currently writes to the main data files of postgres all go through the
OS page cache. This means that some operating systems can end up
collecting a large number of dirty buffers in their respective page
caches. When these dirty buffers are flushed to storage rapidly, be it
because of fsync(), timeouts, or dirty ratios, latency for other reads
and writes can increase massively. This is the primary reason for
regular massive stalls observed in real world scenarios and artificial
benchmarks; on rotating disks stalls on the order of hundreds of seconds
have been observed.
On linux it is possible to control this by reducing the global dirty
limits significantly, reducing the above problem. But global
configuration is rather problematic because it'll affect other
applications; also PostgreSQL itself doesn't always generally want this
behavior, e.g. for temporary files it's undesirable.
Several operating systems allow some control over the kernel page
cache. Linux has sync_file_range(2), several posix systems have msync(2)
and posix_fadvise(2). sync_file_range(2) is preferable because it
requires no special setup, whereas msync() requires the to-be-flushed
range to be mmap'ed. For the purpose of flushing dirty data
posix_fadvise(2) is the worst alternative, as flushing dirty data is
just a side-effect of POSIX_FADV_DONTNEED, which also removes the pages
from the page cache. Thus the feature is enabled by default only on
linux, but can be enabled on all systems that have any of the above
APIs.
While desirable and likely possible this patch does not contain an
implementation for windows.
With the infrastructure added, writes made via checkpointer, bgwriter
and normal user backends can be flushed after a configurable number of
writes. Each of these sources of writes controlled by a separate GUC,
checkpointer_flush_after, bgwriter_flush_after and backend_flush_after
respectively; they're separate because the number of flushes that are
good are separate, and because the performance considerations of
controlled flushing for each of these are different.
A later patch will add checkpoint sorting - after that flushes from the
ckeckpoint will almost always be desirable. Bgwriter flushes are most of
the time going to be random, which are slow on lots of storage hardware.
Flushing in backends works well if the storage and bgwriter can keep up,
but if not it can have negative consequences. This patch is likely to
have negative performance consequences without checkpoint sorting, but
unfortunately so has sorting without flush control.
Discussion: alpine.DEB.2.10.1506011320000.28433@sto
Author: Fabien Coelho and Andres Freund
Earlier version committed in 9.0 caused spurious waits in some cases.
New infrastructure for lock waits in 9.3 used to correct and improve this.
Jeff Janes based upon a proposal by Simon Riggs, who also reviewed
Additional review comments from Amit Kapila
When a process is waiting for a heavyweight lock, we will now indicate
the type of heavyweight lock for which it is waiting. Also, you can
now see when a process is waiting for a lightweight lock - in which
case we will indicate the individual lock name or the tranche, as
appropriate - or for a buffer pin.
Amit Kapila, Ildus Kurbangaliev, reviewed by me. Lots of helpful
discussion and suggestions by many others, including Alexander
Korotkov, Vladimir Borodin, and many others.
Renaming a file using rename(2) is not guaranteed to be durable in face
of crashes; especially on filesystems like xfs and ext4 when mounted
with data=writeback. To be certain that a rename() atomically replaces
the previous file contents in the face of crashes and different
filesystems, one has to fsync the old filename, rename the file, fsync
the new filename, fsync the containing directory. This sequence is not
generally adhered to currently; which exposes us to data loss risks. To
avoid having to repeat this arduous sequence, introduce
durable_rename(), which wraps all that.
Also add durable_link_or_rename(). Several places use link() (with a
fallback to rename()) to rename a file, trying to avoid replacing the
target file out of paranoia. Some of those rename sequences need to be
durable as well. There seems little reason extend several copies of the
same logic, so centralize the link() callers.
This commit does not yet make use of the new functions; they're used in
a followup commit.
Author: Michael Paquier, Andres Freund
Discussion: 56583BDD.9060302@2ndquadrant.com
Backpatch: All supported branches
For example, if you want to perform an ioctl() on a file descriptor
opened through the fd.c routines, there's no way to do that without
being able to get at the underlying fd.
KaiGai Kohei
The new bit indicates whether every tuple on the page is already frozen.
It is cleared only when the all-visible bit is cleared, and it can be
set only when we vacuum a page and find that every tuple on that page is
both visible to every transaction and in no need of any future
vacuuming.
A future commit will use this new bit to optimize away full-table scans
that would otherwise be triggered by XID wraparound considerations. A
page which is merely all-visible must still be scanned in that case, but
a page which is all-frozen need not be. This commit does not attempt
that optimization, although that optimization is the goal here. It
seems better to get the basic infrastructure in place first.
Per discussion, it's very desirable for pg_upgrade to automatically
migrate existing VM forks from the old format to the new format. That,
too, will be handled in a follow-on patch.
Masahiko Sawada, reviewed by Kyotaro Horiguchi, Fujii Masao, Amit
Kapila, Simon Riggs, Andres Freund, and others, and substantially
revised by me.
This patch introduces "pg_blocking_pids(int) returns int[]", which returns
the PIDs of any sessions that are blocking the session with the given PID.
Historically people have obtained such information using a self-join on
the pg_locks view, but it's unreasonably tedious to do it that way with any
modicum of correctness, and the addition of parallel queries has pretty
much broken that approach altogether. (Given some more columns in the view
than there are today, you could imagine handling parallel-query cases with
a 4-way join; but ugh.)
The new function has the following behaviors that are painful or impossible
to get right via pg_locks:
1. Correctly understands which lock modes block which other ones.
2. In soft-block situations (two processes both waiting for conflicting lock
modes), only the one that's in front in the wait queue is reported to
block the other.
3. In parallel-query cases, reports all sessions blocking any member of
the given PID's lock group, and reports a session by naming its leader
process's PID, which will be the pg_backend_pid() value visible to
clients.
The motivation for doing this right now is mostly to fix the isolation
tests. Commit 38f8bdcac4 lobotomized
isolationtester's is-it-waiting query by removing its ability to recognize
nonconflicting lock modes, as a crude workaround for the inability to
handle soft-block situations properly. But even without the lock mode
tests, the old query was excessively slow, particularly in
CLOBBER_CACHE_ALWAYS builds; some of our buildfarm animals fail the new
deadlock-hard test because the deadlock timeout elapses before they can
probe the waiting status of all eight sessions. Replacing the pg_locks
self-join with use of pg_blocking_pids() is not only much more correct, but
a lot faster: I measure it at about 9X faster in a typical dev build with
Asserts, and 3X faster in CLOBBER_CACHE_ALWAYS builds. That should provide
enough headroom for the slower CLOBBER_CACHE_ALWAYS animals to pass the
test, without having to lengthen deadlock_timeout yet more and thus slow
down the test for everyone else.
We don't really need this field, because it's either zero or redundant with
PGPROC.pid. The use of zero to mark "not a group leader" is not necessary
since we can just as well test whether lockGroupLeader is NULL. This does
not save very much, either as to code or data, but the simplification seems
worthwhile anyway.
In 4b4b680c3 I accidentally used sizeof(PrivateRefCountArray) instead of
sizeof(PrivateRefCountEntry) when creating the refcount overflow
hashtable. As the former is bigger than the latter, this luckily only
resulted in a slightly increased memory usage when many buffers are
pinned in a backend.
Reported-By: Takashi Horikawa
Discussion: 73FA3881462C614096F815F75628AFCD035A48C3@BPXM01GP.gisp.nec.co.jp
Backpatch: 9.5, where thew new ref count infrastructure was introduced
Reflow text in lock manager README so that it fits within 80 columns.
Correct some mistakes. Expand the README to explain not only why group
locking exists but also the data structures that support it. Improve
comments related to group locking several files. Change the name of a
macro argument for improved clarity.
Most of these problems were reported by Tom Lane, but I found a few
of them myself.
Robert Haas and Tom Lane
The API spec for this function was changed completely (and for the better)
by commit 3cba8999b3, but it didn't bother
with anything as mundane as updating the comments.
This finishes the work - spread across many commits over the last
several months - of putting each type of lock other than the named
individual locks into a separate tranche.
Amit Kapila
Commit 0e141c0fbb introduced a new
facility to reduce ProcArrayLock contention by clearing several XIDs
from the ProcArray under a single lock acquisition. The names
initially chosen were deemed not to be very good choices, so commit
4aec49899e renamed them. But now it
seems like we still didn't get it right. A pending patch wants to
add similar infrastructure for batching CLOG updates, so the names
need to be clear enough to allow a new set of structure members with
a related purpose.
Amit Kapila
As of commit c1772ad922, there's no
longer any way of requesting additional LWLocks in the main tranche,
so we don't need NumLWLocks() or LWLockAssign() any more. Also,
some of the allocation counters that we had previously aren't needed
any more either.
Amit Kapila
Further investigation says that there may be some slow operations after
we've finished ShutdownXLOG(), so add some more log messages to try to
isolate that. This is all temporary code too.
Early returns from the buildfarm show that there's a bit of a gap in the
logging I added in 3971f64843b02e4a: the portion of CreateCheckPoint()
after CheckPointGuts() can take a fair amount of time. Add a few more
log messages in that section of code. This too shall be reverted later.
This is a quick hack, due to be reverted when its purpose has been served,
to try to gather information about why some of the buildfarm critters
regularly fail with "postmaster does not shut down" complaints. Maybe they
are just really overloaded, but maybe something else is going on. Hence,
instrument pg_ctl to print the current time when it starts waiting for
postmaster shutdown and when it gives up, and add a lot of logging of the
current time in the server's checkpoint and shutdown code paths.
No attempt has been made to make this pretty. I'm not even totally sure
if it will build on Windows, but we'll soon find out.
For locking purposes, we now regard heavyweight locks as mutually
non-conflicting between cooperating parallel processes. There are some
possible pitfalls to this approach that are not to be taken lightly,
but it works OK for now and can be changed later if we find a better
approach. Without this, it's very easy for parallel queries to
silently self-deadlock if the user backend holds strong relation locks.
Robert Haas, with help from Amit Kapila. Thanks to Noah Misch and
Andres Freund for extensive discussion of possible issues with this
approach.
The previous RequestAddinLWLocks() method had several disadvantages.
First, the locks would be in the main tranche; we've recently decided
that it's useful for LWLocks used for separate purposes to have
separate tranche IDs. Second, there wasn't any correlation between
what code called RequestAddinLWLocks() and what code called
LWLockAssign(); when multiple modules are in use, it could become
quite difficult to troubleshoot problems where LWLockAssign() ran out
of locks. To fix, create a concept of named LWLock tranches which
can be used either by extension or by core code.
Amit Kapila and Robert Haas
This is following in a long train of similar changes and for the same
reasons - see b319356f0e and
fe702a7b3f inter alia.
Author: Amit Kapila
Reviewed-by: Alexander Korotkov, Robert Haas
Previously, each PGPROC's backendLock was part of the main tranche,
and the PGPROC just contained a pointer. Now, the actual LWLock is
part of the PGPROC.
As with previous, similar patches, this makes it significantly easier
to identify these lwlocks in LWLOCK_STATS or Trace_lwlocks output
and improves modularity.
Author: Ildus Kurbangaliev
Reviewed-by: Amit Kapila, Robert Haas
It's entirely surprising that mdnblocks() has the side effect of
creating new files on disk, so let's make it not do that. One
consequence of the old behavior is that, if running on a damaged
cluster that is missing a file, mdnblocks() can recreate the file
and allow a subsequent _mdfd_getseg() for a higher segment to succeed.
This happens because, while mdnblocks() stops when it finds a segment
that is shorter than 1GB, _mdfd_getseg() has no such check, and thus
the empty file created by mdnblocks() can allow it to continue its
traversal and find higher-numbered segments which remain.
It might be a good idea for _mdfd_getseg() to actually verify that
each segment it finds is exactly 1GB before proceeding to the next
one, but that would involve some additional system calls, so for
now I'm just doing this much.
Patch by me, per off-list analysis by Kevin Grittner and Rahila Syed.
Review by Andres Freund.
Move the content lock directly into the BufferDesc, so that locking and
pinning a buffer touches only one cache line rather than two. Adjust
the definition of BufferDesc slightly so that this doesn't make the
BufferDesc any larger than one cache line (at least on platforms where
a spinlock is only 1 or 2 bytes).
We can't fit the I/O locks into the BufferDesc and stay within one
cache line, so move those to a completely separate tranche. This
leaves a relatively limited number of LWLocks in the main tranche, so
increase the padding of those remaining locks to a full cache line,
rather than allowing adjacent locks to share a cache line, hopefully
reducing false sharing.
Performance testing shows that these changes make little difference
on laptop-class machines, but help significantly on larger servers,
especially those with more than 2 sockets.
Andres Freund, originally based on an earlier patch by Simon Riggs.
Review and cosmetic adjustments (including heavy rewriting of the
comments) by me.
It's a bit cumbersome to use LWLockNewTrancheId(), because the returned
value needs to be shared between backends so that each backend can call
LWLockRegisterTranche() with the correct ID. So, for built-in tranches,
use a hard-coded value instead.
This is motivated by an upcoming patch adding further built-in tranches.
Andres Freund and Robert Haas
e3f4cfc7 introduced a LWLockHeldByMe() call, without the corresponding
Assert() surrounding it.
Spotted by Coverity.
Backpatch: 9.1+, like the previous commit
At the end of crash recovery, unlogged relations are reset to the empty
state, using their init fork as the template. The init fork is copied to
the main fork without going through shared buffers. Unfortunately WAL
replay so far has not necessarily flushed writes from shared buffers to
disk at that point. In normal crash recovery, and before the
introduction of 'fast promotions' in fd4ced523 / 9.3, the
END_OF_RECOVERY checkpoint flushes the buffers out in time. But with
fast promotions that's not the case anymore.
To fix, force WAL writes targeting the init fork to be flushed
immediately (using the new FlushOneBuffer() function). In 9.5+ that
flush can centrally be triggered from the code dealing with restoring
full page writes (XLogReadBufferForRedoExtended), in earlier releases
that responsibility is in the hands of XLOG_HEAP_NEWPAGE's replay
function.
Backpatch to 9.1, even if this currently is only known to trigger in
9.3+. Flushing earlier is more robust, and it is advantageous to keep
the branches similar.
Typical symptoms of this bug are errors like
'ERROR: index "..." contains unexpected zero page at block 0'
shortly after promoting a node.
Reported-By: Thom Brown
Author: Andres Freund and Michael Paquier
Discussion: 20150326175024.GJ451@alap3.anarazel.de
Backpatch: 9.1-
Prior to commit 0709b7ee72, access to
variables within a spinlock-protected critical section had to be done
through a volatile pointer, but that should no longer be necessary.
Review by Andres Freund
This makes it significantly easier to identify these lwlocks in
LWLOCK_STATS or Trace_lwlocks output. It's also arguably better
from a modularity standpoint, since lwlock.c no longer needs to
know anything about the LWLock needs of the higher-level SLRU
facility.
Ildus Kurbangaliev, reviewd by Álvaro Herrera and by me.
Commit a1480ec1d3 purported to fix the
problems with commit b2ccb5f4e6, but it
didn't completely fix them. The problem is that the checks were
performed in the wrong order, leading to a race condition. If the
sender attached, sent a message, and detached after the receiver
called shm_mq_get_sender and before the receiver called
shm_mq_counterparty_gone, we'd incorrectly return SHM_MQ_DETACHED
before all messages were read. Repair by reversing the order of
operations, and add a long comment explaining why this new logic is
(hopefully) correct.
On insert the CheckForSerializableConflictIn() test was performed
before the page(s) which were going to be modified had been locked
(with an exclusive buffer content lock). If another process
acquired a relation SIReadLock on the heap and scanned to a page on
which an insert was going to occur before the page was so locked,
a rw-conflict would be missed, which could allow a serialization
anomaly to be missed. The window between the check and the page
lock was small, so the bug was generally not noticed unless there
was high concurrency with multiple processes inserting into the
same table.
This was reported by Peter Bailis as bug #11732, by Sean Chittenden
as bug #13667, and by others.
The race condition was eliminated in heap_insert() by moving the
check down below the acquisition of the buffer lock, which had been
the very next statement. Because of the loop locking and unlocking
multiple buffers in heap_multi_insert() a check was added after all
inserts were completed. The check before the start of the inserts
was left because it might avoid a large amount of work to detect a
serialization anomaly before performing the all of the inserts and
the related WAL logging.
While investigating this bug, other SSI bugs which were even harder
to hit in practice were noticed and fixed, an unnecessary check
(covered by another check, so redundant) was removed from
heap_update(), and comments were improved.
Back-patch to all supported branches.
Kevin Grittner and Thomas Munro
If the counterparty writes some data into the queue and then detaches,
it's wrong to return SHM_MQ_DETACHED right away. If we do that, we
fail to read whatever was written.
The shm_mq mechanism was intended to optionally notice when the process
on the other end of the queue fails to attach to the queue. It does
this by allowing the user to pass a BackgroundWorkerHandle; if the
background worker in question is launched and dies without attaching
to the queue, then we know it never will. This logic works OK in
blocking mode, but when called with nowait = true we fail to notice
that this has happened due to an asymmetry in the logic. Repair.
Reported off-list by Rushabh Lathia. Patch by me.
Prior to commit 0709b7ee72, access to
variables within a spinlock-protected critical section had to be done
through a volatile pointer, but that should no longer be necessary.
Michael Paquier
Prior to commit 0709b7ee72, access to
variables within a spinlock-protected critical section had to be done
through a volatile pointer, but that should no longer be necessary.
Thomas Munro
This flag has proven to be a recipe for bugs, and it doesn't seem like
it can really buy anything in terms of performance. So let's just
*always* set the process latch when we receive SIGUSR1 instead of
trying to do it only when needed.
Per my recent proposal on pgsql-hackers.
The fact that multixact truncations are not WAL logged has caused a fair
share of problems. Amongst others it requires to do computations during
recovery while the database is not in a consistent state, delaying
truncations till checkpoints, and handling members being truncated, but
offset not.
We tried to put bandaids on lots of these issues over the last years,
but it seems time to change course. Thus this patch introduces WAL
logging for multixact truncations.
This allows:
1) to perform the truncation directly during VACUUM, instead of delaying it
to the checkpoint.
2) to avoid looking at the offsets SLRU for truncation during recovery,
we can just use the master's values.
3) simplify a fair amount of logic to keep in memory limits straight,
this has gotten much easier
During the course of fixing this a bunch of additional bugs had to be
fixed:
1) Data was not purged from memory the member's SLRU before deleting
segments. This happened to be hard or impossible to hit due to the
interlock between checkpoints and truncation.
2) find_multixact_start() relied on SimpleLruDoesPhysicalPageExist - but
that doesn't work for offsets that haven't yet been flushed to
disk. Add code to flush the SLRUs to fix. Not pretty, but it feels
slightly safer to only make decisions based on actual on-disk state.
3) find_multixact_start() could be called concurrently with a truncation
and thus fail. Via SetOffsetVacuumLimit() that could lead to a round
of emergency vacuuming. The problem remains in
pg_get_multixact_members(), but that's quite harmless.
For now this is going to only get applied to 9.5+, leaving the issues in
the older branches in place. It is quite possible that we need to
backpatch at a later point though.
For the case this gets backpatched we need to handle that an updated
standby may be replaying WAL from a not-yet upgraded primary. We have to
recognize that situation and use "old style" truncation (i.e. looking at
the SLRUs) during WAL replay. In contrast to before, this now happens in
the startup process, when replaying a checkpoint record, instead of the
checkpointer. Doing truncation in the restartpoint is incorrect, they
can happen much later than the original checkpoint, thereby leading to
wraparound. To avoid "multixact_redo: unknown op code 48" errors
standbys would have to be upgraded before primaries.
A later patch will bump the WAL page magic, and remove the legacy
truncation codepaths. Legacy truncation support is just included to make
a possible future backpatch easier.
Discussion: 20150621192409.GA4797@alap3.anarazel.de
Reviewed-By: Robert Haas, Alvaro Herrera, Thomas Munro
Backpatch: 9.5 for now
RemoveLocalLock() must consider the possibility that LockAcquireExtended()
failed to palloc the initial space for a locallock's lockOwners array.
I had evidently meant to cope with this hazard when the code was originally
written (commit 1785acebf2), but missed that
the pfree needed to be protected with an if-test. Just to make sure things
are left in a clean state, reset numLockOwners as well.
Per low-memory testing by Andreas Seltenreich. Back-patch to all supported
branches.
The shm_mq mechanism was built to send error (and notice) messages and
tuples between backends. However, shm_mq itself only deals in raw
bytes. Since commit 2bd9e412f9, we have
had infrastructure for one message to redirect protocol messages to a
queue and for another backend to parse them and do useful things with
them. This commit introduces a somewhat analogous facility for tuples
by adding a new type of DestReceiver, DestTupleQueue, which writes
each tuple generated by a query into a shm_mq, and a new
TupleQueueFunnel facility which reads raw tuples out of the queue and
reconstructs the HeapTuple format expected by the executor.
The TupleQueueFunnel abstraction supports reading from multiple tuple
streams at the same time, but only in round-robin fashion. Someone
could imaginably want other policies, but this should be good enough
to meet our short-term needs related to parallel query, and we can
always extend it later.
This also makes one minor addition to the shm_mq API that didn'
seem worth breaking out as a separate patch.
Extracted from Amit Kapila's parallel sequential scan patch. This
code was originally written by me, and then it was revised by Amit,
and then it was revised some more by me.
Naming the individual lwlocks seems like something that may be useful
for other types of debugging, monitoring, or instrumentation output,
but this commit just implements it for the specific case of
trace_lwlocks.
Patch by me, reviewed by Amit Kapila and Kyotaro Horiguchi
Per discussion, nowadays it is possible to have tablespaces that have
wildly different I/O characteristics from others. Setting different
effective_io_concurrency parameters for those has been measured to
improve performance.
Author: Julien Rouhaud
Reviewed by: Andres Freund
Post-commit review by Andres Freund discovered a couple of concurrency
bugs in the original patch: specifically, if the leader cleared a
follower's XID before it reached PGSemaphoreLock, the semaphore would be
left in the wrong state; and if another process did PGSemaphoreUnlock
for some unrelated reason, we might resume execution before the fact
that our XID was cleared was globally visible.
Also, improve the wording of some comments, rename nextClearXidElem
to firstClearXidElem in PROC_HDR for clarity, and drop some volatile
qualifiers that aren't necessary.
Amit Kapila, reviewed and slightly revised by me.
This fixes a bunch of somewhat pedantic warnings with new
compilers. Since by far the majority of other functions definitions use
the (void) style it just seems to be consistent to do so as well in the
remaining few places.
In 4b4b680c I passed a buffer index number (starting from 0) instead of
a proper Buffer id (which start from 1 for shared buffers) in two
places.
This wasn't noticed so far as one of those locations isn't compiled at
all (PrintPinnedBufs) and the other one (InvalidBuffer) requires a
unlikely, but possible, set of circumstances to trigger a symptom.
To reduce the likelihood of such incidents a bit also convert existing
open coded mappings from buffer descriptors to buffer ids with
BufferDescriptorGetBuffer().
Author: Qingqing Zhou
Reported-By: Qingqing Zhou
Discussion: CAJjS0u2ai9ooUisKtkV8cuVUtEkMTsbK8c7juNAjv8K11zeCQg@mail.gmail.com
Backpatch: 9.5 where the private ref count infrastructure was introduced
If some, but not all, of the length word has already been read, and the
next attempt to read sees exactly the number of bytes needed to complete
the length word, or fewer, then we'll incorrectly read less than all of
the available data.
Antonin Houska
When a write transaction commits, it must clear its XID advertised via
the ProcArray, which requires that we hold ProcArrayLock in exclusive
mode in order to prevent concurrent processes running GetSnapshotData
from seeing inconsistent results. When many processes try to commit
at once, ProcArrayLock must change hands repeatedly, with each
concurrent process trying to commit waking up to acquire the lock in
turn. To make things more efficient, when more than one backend is
trying to commit a write transaction at the same time, have just one
of them acquire ProcArrayLock in exclusive mode and clear the XIDs of
all processes in the group. Benchmarking reveals that this is much
more efficient at very high client counts.
Amit Kapila, heavily revised by me, with some review also from Pavan
Deolasee.
LWLockAttemptLock pointlessly read the lock's state in every loop
iteration, even though pg_atomic_compare_exchange_u32() returns the old
value. Instead do that only once before the loop iteration.
Additionally there's no need to have the expected_state variable,
old_state mostly had the same value anyway.
Noticed-By: Heikki Linnakangas
Backpatch: 9.5, no reason to let the branches diverge at this point
The lwlock scalability work introduced two race conditions into the
lwlock variable support provided for xlog.c. First, and harmlessly on
most platforms, it set/read the variable without the spinlock in some
places. Secondly, due to the removal of the spinlock, it was possible
that a backend missed changes to the variable's state if it changed in
the wrong moment because checking the lock's state, the variable's state
and the queuing are not protected by a single spinlock acquisition
anymore.
To fix first move resetting the variable's from LWLockAcquireWithVar to
WALInsertLockRelease, via a new function LWLockReleaseClearVar. That
prevents issues around waiting for a variable's value to change when a
new locker has acquired the lock, but not yet set the value. Secondly
re-check that the variable hasn't changed after enqueing, that prevents
the issue that the lock has been released and already re-acquired by the
time the woken up backend checks for the lock's state.
Reported-By: Jeff Janes
Analyzed-By: Heikki Linnakangas
Reviewed-By: Heikki Linnakangas
Discussion: 5592DB35.2060401@iki.fi
Backpatch: 9.5, where the lwlock scalability went in
Don't print a WARNING if we get ESRCH from a kill() that's attempting
to cancel an autovacuum worker. It's possible (and has been seen in the
buildfarm) that the worker is already gone by the time we are able to
execute the kill, in which case the failure is harmless. About the only
plausible reason for reporting such cases would be to help debug corrupted
lock table contents, but this is hardly likely to be the most important
symptom if that happens. Moreover issuing a WARNING might scare users
more than is warranted.
Also, since sending a signal to an autovacuum worker is now entirely a
routine thing, and the worker will log the query cancel on its end anyway,
reduce the message saying we're doing that from LOG to DEBUG1 level.
Very minor cosmetic cleanup as well.
Since the main practical reason for doing this is to avoid unnecessary
buildfarm failures, back-patch to all active branches.
This code was originally written as part of parallel query effort, but
it seems to have independent value, because if we make one decision
about where to get a PGPROC when we allocate and then put it back on a
different list at backend-exit time, bad things happen. This isn't
just a theoretical risk; we fixed an actual problem of this type in
commit e280c630a8.
1. Replay of the WAL record for setting a bit in the visibility map
contained an assertion that a full-page image of that record type can only
occur with checksums enabled. But it can also happen with wal_log_hints, so
remove the assertion. Unlike checksums, wal_log_hints can be changed on the
fly, so it would be complicated to figure out if it was enabled at the time
that the WAL record was generated.
2. wal_log_hints has the same effect on the locking needed to read the LSN
of a page as data checksums. BufferGetLSNAtomic() didn't get the memo.
Backpatch to 9.4, where wal_log_hints was added.
The argument that this is a sufficiently-expected case to be silently
ignored seems pretty thin. Andres had brought it up back when we were
still considering that most fsync failures should be hard errors, and it
probably would be legit not to fail hard for ETXTBSY --- but the same is
true for EROFS and other cases, which is why we gave up on hard failures.
ETXTBSY is surely not a normal case, so logging the failure seems fine
from here.
Commit 2ce439f337 introduced a rather serious
regression, namely that if its scan of the data directory came across any
un-fsync-able files, it would fail and thereby prevent database startup.
Worse yet, symlinks to such files also caused the problem, which meant that
crash restart was guaranteed to fail on certain common installations such
as older Debian.
After discussion, we agreed that (1) failure to start is worse than any
consequence of not fsync'ing is likely to be, therefore treat all errors
in this code as nonfatal; (2) we should not chase symlinks other than
those that are expected to exist, namely pg_xlog/ and tablespace links
under pg_tblspc/. The latter restriction avoids possibly fsync'ing a
much larger part of the filesystem than intended, if the user has left
random symlinks hanging about in the data directory.
This commit takes care of that and also does some code beautification,
mainly moving the relevant code into fd.c, which seems a much better place
for it than xlog.c, and making sure that the conditional compilation for
the pre_sync_fname pass has something to do with whether pg_flush_data
works.
I also relocated the call site in xlog.c down a few lines; it seems a
bit silly to be doing this before ValidateXLOGDirectoryStructure().
The similar logic in initdb.c ought to be made to match this, but that
change is noncritical and will be dealt with separately.
Back-patch to all active branches, like the prior commit.
Abhijit Menon-Sen and Tom Lane
Ensure that we null-terminate the result string (one place in pg_rewind).
Be paranoid about out-of-range results from readlink() (should not happen,
but there is no good reason for some call sites to be careful about it and
others not). Consistently use the whole buffer, not sometimes one byte
less. Ensure we emit an appropriate errcode() in all cases. Spell the
error messages the same way.
The only serious bug here is the missing null-termination in pg_rewind,
which is new code, so no need for a back-patch.
Abhijit Menon-Sen and Tom Lane
We want this struct to be exactly a series of 3 int16 words, no more
and no less. Historically, at least, some ARM compilers preferred to
pad it to 8 bytes unless coerced. Our old way of doing that was just
to use __attribute__((packed)), but as pointed out by Piotr Stefaniak,
that does too much: it also licenses the compiler to give the struct
only byte-alignment. We don't want that because it adds access overhead,
possibly quite significant overhead. According to the GCC manual, what
we want requires also specifying __attribute__((align(2))). It's not
entirely clear if all the relevant compilers accept this pragma as well,
but we can hope the buildfarm will tell us if not. We can also add a
static assertion that should fire if the compiler padded the struct.
Since the combination of these pragmas should define exactly what we
want on any compiler that accepts them, let's try using them wherever
we think they exist, not only for __arm__. (This is likely to expose
that the conditional definitions in c.h are inadequate, but finding
that out would be a good thing.)
The immediate motivation for this is that the current definition of
ExecRowMark allows its curCtid field to be misaligned. It is not clear
whether there are any other uses of ItemPointerData with a similar hazard.
We could change the definition of ExecRowMark if this doesn't work, but
it would be far better to have a future-proof fix.
Piotr Stefaniak, some further hacking by me
Use "a" and "an" correctly, mostly in comments. Two error messages were
also fixed (they were just elogs, so no translation work required). Two
function comments in pg_proc.h were also fixed. Etsuro Fujita reported one
of these, but I found a lot more with grep.
Also fix a few other typos spotted while grepping for the a/an typos.
For example, "consists out of ..." -> "consists of ...". Plus a "though"/
"through" mixup reported by Euler Taveira.
Many of these typos were in old code, which would be nice to backpatch to
make future backpatching easier. But much of the code was new, and I didn't
feel like crafting separate patches for each branch. So no backpatching.
The point of the assertion is to ensure that the arrays allocated in stack
are large enough, but the check was one item short.
This won't matter in practice because MaxIndexTuplesPerPage is an
overestimate, so you can't have that many items on a page in reality.
But let's be tidy.
Spotted by Anastasia Lubennikova. Backpatch to all supported versions, like
the patch that added the assertion.
The newly added ON CONFLICT clause allows to specify an alternative to
raising a unique or exclusion constraint violation error when inserting.
ON CONFLICT refers to constraints that can either be specified using a
inference clause (by specifying the columns of a unique constraint) or
by naming a unique or exclusion constraint. DO NOTHING avoids the
constraint violation, without touching the pre-existing row. DO UPDATE
SET ... [WHERE ...] updates the pre-existing tuple, and has access to
both the tuple proposed for insertion and the existing tuple; the
optional WHERE clause can be used to prevent an update from being
executed. The UPDATE SET and WHERE clauses have access to the tuple
proposed for insertion using the "magic" EXCLUDED alias, and to the
pre-existing tuple using the table name or its alias.
This feature is often referred to as upsert.
This is implemented using a new infrastructure called "speculative
insertion". It is an optimistic variant of regular insertion that first
does a pre-check for existing tuples and then attempts an insert. If a
violating tuple was inserted concurrently, the speculatively inserted
tuple is deleted and a new attempt is made. If the pre-check finds a
matching tuple the alternative DO NOTHING or DO UPDATE action is taken.
If the insertion succeeds without detecting a conflict, the tuple is
deemed inserted.
To handle the possible ambiguity between the excluded alias and a table
named excluded, and for convenience with long relation names, INSERT
INTO now can alias its target table.
Bumps catversion as stored rules change.
Author: Peter Geoghegan, with significant contributions from Heikki
Linnakangas and Andres Freund. Testing infrastructure by Jeff Janes.
Reviewed-By: Heikki Linnakangas, Andres Freund, Robert Haas, Simon Riggs,
Dean Rasheed, Stephen Frost and many others.
pg_win32_is_junction() was a typo for pgwin32_is_junction(). open()
was used not only in a two-argument form, which breaks on Windows,
but also where BasicOpenFile() should have been used.
Per reports from Andrew Dunstan and David Rowley.
Otherwise, if there's another crash, some writes from after the first
crash might make it to disk while writes from before the crash fail
to make it to disk. This could lead to data corruption.
Back-patch to all supported versions.
Abhijit Menon-Sen, reviewed by Andres Freund and slightly revised
by me.
This does four basic things. First, it provides convenience routines
to coordinate the startup and shutdown of parallel workers. Second,
it synchronizes various pieces of state (e.g. GUCs, combo CID
mappings, transaction snapshot) from the parallel group leader to the
worker processes. Third, it prohibits various operations that would
result in unsafe changes to that state while parallelism is active.
Finally, it propagates events that would result in an ErrorResponse,
NoticeResponse, or NotifyResponse message being sent to the client
from the parallel workers back to the master, from which they can then
be sent on to the client.
Robert Haas, Amit Kapila, Noah Misch, Rushabh Lathia, Jeevan Chalke.
Suggestions and review from Andres Freund, Heikki Linnakangas, Noah
Misch, Simon Riggs, Euler Taveira, and Jim Nasby.
When implementing a replication solution ontop of logical decoding, two
related problems exist:
* How to safely keep track of replication progress
* How to change replication behavior, based on the origin of a row;
e.g. to avoid loops in bi-directional replication setups
The solution to these problems, as implemented here, consist out of
three parts:
1) 'replication origins', which identify nodes in a replication setup.
2) 'replication progress tracking', which remembers, for each
replication origin, how far replay has progressed in a efficient and
crash safe manner.
3) The ability to filter out changes performed on the behest of a
replication origin during logical decoding; this allows complex
replication topologies. E.g. by filtering all replayed changes out.
Most of this could also be implemented in "userspace", e.g. by inserting
additional rows contain origin information, but that ends up being much
less efficient and more complicated. We don't want to require various
replication solutions to reimplement logic for this independently. The
infrastructure is intended to be generic enough to be reusable.
This infrastructure also replaces the 'nodeid' infrastructure of commit
timestamps. It is intended to provide all the former capabilities,
except that there's only 2^16 different origins; but now they integrate
with logical decoding. Additionally more functionality is accessible via
SQL. Since the commit timestamp infrastructure has also been introduced
in 9.5 (commit 73c986add) changing the API is not a problem.
For now the number of origins for which the replication progress can be
tracked simultaneously is determined by the max_replication_slots
GUC. That GUC is not a perfect match to configure this, but there
doesn't seem to be sufficient reason to introduce a separate new one.
Bumps both catversion and wal page magic.
Author: Andres Freund, with contributions from Petr Jelinek and Craig Ringer
Reviewed-By: Heikki Linnakangas, Petr Jelinek, Robert Haas, Steve Singer
Discussion: 20150216002155.GI15326@awork2.anarazel.de,
20140923182422.GA15776@alap3.anarazel.de,
20131114172632.GE7522@alap2.anarazel.de
BackendIdGetTransactionIds() neglected the possibility that the PROC
pointer in a ProcState array entry is null. In current usage, this could
only crash if the other backend had exited since pgstat_read_current_status
saw it as active, which is a pretty narrow window. But it's reachable in
the field, per bug #12918 from Vladimir Borodin.
Back-patch to 9.4 where the faulty code was introduced.
Even though the main benefit of the Lehman and Yao algorithm for
btrees is that no locks need be held between page reads in an
index search, we were holding a buffer pin on each leaf page after
it was read until we were ready to read the next one. The reason
was so that we could treat this as a weak lock to create an
"interlock" with vacuum's deletion of heap line pointers, even
though our README file pointed out that this was not necessary for
a scan using an MVCC snapshot.
The main goal of this patch is to reduce the blocking of vacuum
processes by in-progress btree index scans (including a cursor
which is idle), but the code rearrangement also allows for one
less buffer content lock to be taken when a forward scan steps from
one page to the next, which results in a small but consistent
performance improvement in many workloads.
This patch leaves behavior unchanged for some cases, which can be
addressed separately so that each case can be evaluated on its own
merits. These unchanged cases are when a scan uses a non-MVCC
snapshot, an index-only scan, and a scan of a btree index for which
modifications are not WAL-logged. If later patches allow all of
these cases to drop the buffer pin after reading a leaf page, then
the btree vacuum process can be simplified; it will no longer need
the "super-exclusive" lock to delete tuples from a page.
Reviewed by Heikki Linnakangas and Kyotaro Horiguchi
dsm_control->nitems never decreases, so this is testing whether the
server has *ever* run out of DSM segments, not whether it is
*currently* out of DSM segments.
Reported off-list by Amit Kapila.
Right now, there's only one flag, DSM_CREATE_NULL_IF_MAXSEGMENTS,
which suppresses the error that would normally be thrown when the
maximum number of segments already exists, instead returning NULL.
It might be useful to add more flags in the future, such as one to
ignore allocation errors, but I haven't done that here.
When LockBufferForCleanup() has to wait for getting a cleanup lock on a
buffer it does so by setting a flag in the buffer header and then wait
for other backends to signal it using ProcWaitForSignal().
Unfortunately LockBufferForCleanup() missed that ProcWaitForSignal() can
return for other reasons than the signal it is hoping for. If such a
spurious signal arrives the wait flags on the buffer header will still
be set. That then triggers "ERROR: multiple backends attempting to wait
for pincount 1".
The fix is simple, unset the flag if still set when retrying. That
implies an additional spinlock acquisition/release, but that's unlikely
to matter given the cost of waiting for a cleanup lock. Alternatively
it'd have been possible to move responsibility for maintaining the
relevant flag to the waiter all together, but that might have had
negative consequences due to possible floods of signals. Besides being
more invasive.
This looks to be a very longstanding bug. The relevant code in
LockBufferForCleanup() hasn't changed materially since its introduction
and ProcWaitForSignal() was documented to return for unrelated reasons
since 8.2. The master only patch series removing ImmediateInterruptOK
made it much easier to hit though, as ProcSendSignal/ProcWaitForSignal
now uses a latch shared with other tasks.
Per discussion with Kevin Grittner, Tom Lane and me.
Backpatch to all supported branches.
Discussion: 11553.1423805224@sss.pgh.pa.us
The remaining caller (lwlocks) doesn't need that facility, and we plan
to remove ImmedidateInterruptOK entirely. That means that interrupts
can't be serviced race-free and portably anyway, so there's little
reason for keeping the feature.
Reviewed-By: Heikki Linnakangas
Deadlock checking was performed inside signal handlers up to
now. While it's a remarkable feat to have made this work reliably,
it's quite complex to understand why that is the case. Partially it
worked due to the assumption that semaphores are signal safe - which
is not actually documented to be the case for sysv semaphores.
The reason we had to rely on performing this work inside signal
handlers is that semaphores aren't guaranteed to be interruptable by
signals on all platforms. But now that latches provide a somewhat
similar API, which actually has the guarantee of being interruptible,
we can avoid doing so.
Signalling between ProcSleep, ProcWakeup, ProcWaitForSignal and
ProcSendSignal is now done using latches. This increases the
likelihood of spurious wakeups. As spurious wakeup already were
possible and aren't likely to be frequent enough to be an actual
problem, this seems acceptable.
This change would allow for further simplification of the deadlock
checking, now that it doesn't have to run in a signal handler. But
even if I were motivated to do so right now, it would still be better
to do that separately. Such a cleanup shouldn't have to be reviewed a
the same time as the more fundamental changes in this commit.
There is one possible usability regression due to this commit. Namely
it is more likely than before that log_lock_waits messages are output
more than once.
Reviewed-By: Heikki Linnakangas
Up to now large swathes of backend code ran inside signal handlers
while reading commands from the client, to allow for speedy reaction to
asynchronous events. Most prominently shared invalidation and NOTIFY
handling. That means that complex code like the starting/stopping of
transactions is run in signal handlers... The required code was
fragile and verbose, and is likely to contain bugs.
That approach also severely limited what could be done while
communicating with the client. As the read might be from within
openssl it wasn't safely possible to trigger an error, e.g. to cancel
a backend in idle-in-transaction state. We did that in some cases,
namely fatal errors, nonetheless.
Now that FE/BE communication in the backend employs non-blocking
sockets and latches to block, we can quite simply interrupt reads from
signal handlers by setting the latch. That allows us to signal an
interrupted read, which is supposed to be retried after returning from
within the ssl library.
As signal handlers now only need to set the latch to guarantee timely
interrupt processing, remove a fair amount of complicated & fragile
code from async.c and sinval.c.
We could now actually start to process some kinds of interrupts, like
sinval ones, more often that before, but that seems better done
separately.
This work will hopefully allow to handle cases like being blocked by
sending data, interrupting idle transactions and similar to be
implemented without too much effort. In addition to allowing getting
rid of ImmediateInterruptOK, that is.
Author: Andres Freund
Reviewed-By: Heikki Linnakangas
The logic to compact away removed tuples from page was duplicated with
small differences in PageRepairFragmentation, PageIndexMultiDelete, and
PageIndexDeleteNoCompact. Put it into a common function.
Reviewed by Peter Geoghegan.
If any error occurred while we were in the middle of reading a protocol
message from the client, we could lose sync, and incorrectly try to
interpret a part of another message as a new protocol message. That will
usually lead to an "invalid frontend message" error that terminates the
connection. However, this is a security issue because an attacker might
be able to deliberately cause an error, inject a Query message in what's
supposed to be just user data, and have the server execute it.
We were quite careful to not have CHECK_FOR_INTERRUPTS() calls or other
operations that could ereport(ERROR) in the middle of processing a message,
but a query cancel interrupt or statement timeout could nevertheless cause
it to happen. Also, the V2 fastpath and COPY handling were not so careful.
It's very difficult to recover in the V2 COPY protocol, so we will just
terminate the connection on error. In practice, that's what happened
previously anyway, as we lost protocol sync.
To fix, add a new variable in pqcomm.c, PqCommReadingMsg, that is set
whenever we're in the middle of reading a message. When it's set, we cannot
safely ERROR out and continue running, because we might've read only part
of a message. PqCommReadingMsg acts somewhat similarly to critical sections
in that if an error occurs while it's set, the error handler will force the
connection to be terminated, as if the error was FATAL. It's not
implemented by promoting ERROR to FATAL in elog.c, like ERROR is promoted
to PANIC in critical sections, because we want to be able to use
PG_TRY/CATCH to recover and regain protocol sync. pq_getmessage() takes
advantage of that to prevent an OOM error from terminating the connection.
To prevent unnecessary connection terminations, add a holdoff mechanism
similar to HOLD/RESUME_INTERRUPTS() that can be used hold off query cancel
interrupts, but still allow die interrupts. The rules on which interrupts
are processed when are now a bit more complicated, so refactor
ProcessInterrupts() and the calls to it in signal handlers so that the
signal handlers always call it if ImmediateInterruptOK is set, and
ProcessInterrupts() can decide to not do anything if the other conditions
are not met.
Reported by Emil Lenngren. Patch reviewed by Noah Misch and Andres Freund.
Backpatch to all supported versions.
Security: CVE-2015-0244
GetLockConflicts() has for a long time not properly terminated the
returned array. During normal processing the returned array is zero
initialized which, while not pretty, is sufficient to be recognized as
a invalid virtual transaction id. But the HotStandby case is more than
aesthetically broken: The allocated (and reused) array is neither
zeroed upon allocation, nor reinitialized, nor terminated.
Not having a terminating element means that the end of the array will
not be recognized and that recovery conflict handling will thus read
ahead into adjacent memory. Only terminating when hitting memory
content that looks like a invalid virtual transaction id. Luckily
this seems so far not have caused significant problems, besides making
recovery conflict more expensive.
Discussion: 20150127142713.GD29457@awork2.anarazel.de
Backpatch into all supported branches.
Benchmarks has shown that aligning the buffer descriptor array to
cache lines is important for scalability; especially on bigger,
multi-socket, machines.
Currently the array sometimes already happens to be aligned by
happenstance, depending how large previous shared memory allocations
were. That can lead to wildly varying performance results after minor
configuration changes.
In addition to aligning the start of descriptor array, also force the
size of individual descriptors to be of a common cache line size (64
bytes). That happens to already be the case on 64bit platforms, but
this way we can change the struct BufferDesc more easily.
As the alignment primarily matters in highly concurrent workloads
which probably all are 64bit these days, and the space wastage of
element alignment would be a bit more noticeable on 32bit systems, we
don't force the stride to be cacheline sized on 32bit platforms for
now. If somebody does actual performance testing, we can reevaluate
that decision by changing the definition of BUFFERDESC_PADDED_SIZE.
Discussion: 20140202151319.GD32123@awork2.anarazel.de
Per discussion with Bruce Momjan, Tom Lane, Robert Haas, and Peter
Geoghegan.
As noted by Tom Lane the improvements in 4b4b680c3d had the problem
that in some situations we searched, entered and modified entries in
the private refcount hash while holding a spinlock. I had tried to
keep the logic entirely local to PinBuffer_Locked(), but that's not
really possible given it's called with a spinlock held...
Besides being disadvantageous from a performance point of view, this
also has problems with error handling safety. If we failed inserting
an entry into the hashtable due to an out of memory error, we'd error
out with a held spinlock. Not good.
Change the way private refcounts are manipulated: Before a buffer can
be tracked an entry has to be reserved using
ReservePrivateRefCountEntry(); then, if a entry is not found using
GetPrivateRefCountEntry(), it can be entered with
NewPrivateRefCountEntry().
Also take advantage of the fact that PinBuffer_Locked() currently is
never called for buffers that already have been pinned by the current
backend and don't search the private refcount entries for preexisting
local pins. That results in a small, but measurable, performance
improvement.
Additionally make ReleaseBuffer() always call UnpinBuffer() for shared
buffers. That avoids duplicating work in an eventual UnpinBuffer()
call that already has been done in ReleaseBuffer() and also saves some
code.
Per discussion with Tom Lane.
Discussion: 15028.1418772313@sss.pgh.pa.us
To do so, move InitializeLatchSupport() into the new common process
initialization functions, and add a new global variable MyLatch.
MyLatch is usable as soon InitPostmasterChild() has been called
(i.e. very early during startup). Initially it points to a process
local latch that exists in all processes. InitProcess/InitAuxiliaryProcess
then replaces that local latch with PGPROC->procLatch. During shutdown
the reverse happens.
This is primarily advantageous for two reasons: For one it simplifies
dealing with the shared process latch, especially in signal handlers,
because instead of having to check for MyProc, MyLatch can be used
unconditionally. For another, a later patch that makes FEs/BE
communication use latches, now can rely on the existence of a latch,
even before having gone through InitProcess.
Discussion: 20140927191243.GD5423@alap3.anarazel.de
Since their introduction latches have required barriers in SetLatch
and ResetLatch - but when they were introduced there wasn't any
barrier abstraction. Instead latches were documented to rely on the
callsites to provide barrier semantics.
Now that the barrier support looks halfway complete, add the necessary
barriers to both latch implementations.
Also remove a now superflous lock acquisition from syncrep.c and a
superflous (and insufficient) barrier from freelist.c. There might be
other cases that can now be simplified, but those are the only ones
I've seen on a quick scan.
We might want to backpatch this at some later point, but right now the
barrier infrastructure in the backbranches isn't totally on par with
master.
Discussion: 20150112154026.GB2092@awork2.anarazel.de
Back in ed0b409, PGPROC was split and moved to static variables in
procarray.c, with procs in ProcArrayStruct replaced by an array of
integers representing process numbers (pgprocnos), with -1 indicating a
dead process which has yet to be removed. Access to procArray is
generally done under ProcArrayLock and therefore most code does not have
to concern itself with -1 entries.
However, MinimumActiveBackends intentionally does not take
ProcArrayLock, which means it has to be extra careful when accessing
procArray. Prior to ed0b409, this was handled by checking for a NULL
in the pointer array, but that check was no longer valid after the
split. Coverity pointed out that the check could never happen and so
it was removed in 5592eba. That didn't make anything worse, but it
didn't fix the issue either.
The correct fix is to check for pgprocno == -1 and skip over that entry
if it is encountered.
Back-patch to 9.2, since there can be attempts to access the arrays
prior to their start otherwise. Note that the changes prior to 9.4 will
look a bit different due to the change in 5592eba.
Note that MinimumActiveBackends only returns a bool for heuristic
purposes and any pre-array accesses are strictly read-only and so there
is no security implication and the lack of fields complaints indicates
it's very unlikely to run into issues due to this.
Pointed out by Noah.
StrategyGetBuffer() has proven to be a bottleneck in a number of
buffer acquisition heavy workloads. To some degree this has already
been alleviated by 5d7962c6, but it still can be quite a heavy
bottleneck. The problem is that in unfortunate usage patterns a
single StrategyGetBuffer() call will have to look at a large number of
buffers - in turn making it likely that the process will be put to
sleep while still holding the spinlock.
Replace most of the usage of the buffer_strategy_lock spinlock for the
clock sweep by a atomic nextVictimBuffer variable. That variable,
modulo NBuffers, is the current hand of the clock sweep. The buffer
clock-sweep then only needs to acquire the spinlock after a
wraparound. And even then only in the process that did the wrapping
around. That alleviates nearly all the contention on the relevant
spinlock, although significant contention on the cacheline can still
exist.
Reviewed-By: Robert Haas and Amit Kapila
Discussion: 20141010160020.GG6670@alap3.anarazel.de,
20141027133218.GA2639@awork2.anarazel.de
The old LWLock implementation had the problem that concurrent lock
acquisitions required exclusively acquiring a spinlock. Often that
could lead to acquirers waiting behind the spinlock, even if the
actual LWLock was free.
The new implementation doesn't acquire the spinlock when acquiring the
lock itself. Instead the new atomic operations are used to atomically
manipulate the state. Only the waitqueue, used solely in the slow
path, is still protected by the spinlock. Check lwlock.c's header for
an explanation about the used algorithm.
For some common workloads on larger machines this can yield
significant performance improvements. Particularly in read mostly
workloads.
Reviewed-By: Amit Kapila and Robert Haas
Author: Andres Freund
Discussion: 20130926225545.GB26663@awork2.anarazel.de
Besides being shorter and much easier to read it changes the logic in
LWLockRelease() to release all shared lockers when waking up any. This
can yield some significant performance improvements - and the fairness
isn't really much worse than before, as we always allowed new shared
lockers to jump the queue.
In LWLockRelease() (and in 9.4+ LWLockUpdateVar()) we release enqueued
waiters using PGSemaphoreUnlock(). As there are other sources of such
unlocks backends only wake up if MyProc->lwWaiting is set to false;
which is only done in the aforementioned functions.
Before this commit there were dangers because the store to lwWaitLink
could become visible before the store to lwWaitLink. This could both
happen due to compiler reordering (on most compilers) and on some
platforms due to the CPU reordering stores.
The possible consequence of this is that a backend stops waiting
before lwWaitLink is set to NULL. If that backend then tries to
acquire another lock and has to wait there the list could become
corrupted once the lwWaitLink store is finally performed.
Add a write memory barrier to prevent that issue.
Unfortunately the barrier support has been only added in 9.2. Given
that the issue has not knowingly been observed in praxis it seems
sufficient to prohibit compiler reordering using volatile for 9.0 and
9.1. Actual problems due to compiler reordering are more likely
anyway.
Discussion: 20140210134625.GA15246@awork2.anarazel.de
Previously, if you wanted anything besides C-string hash keys, you had to
specify a custom hashing function to hash_create(). Nearly all such
callers were specifying tag_hash or oid_hash; which is tedious, and rather
error-prone, since a caller could easily miss the opportunity to optimize
by using hash_uint32 when appropriate. Replace this with a design whereby
callers using simple binary-data keys just specify HASH_BLOBS and don't
need to mess with specific support functions. hash_create() itself will
take care of optimizing when the key size is four bytes.
This nets out saving a few hundred bytes of code space, and offers
a measurable performance improvement in tidbitmap.c (which was not
exploiting the opportunity to use hash_uint32 for its 4-byte keys).
There might be some wins elsewhere too, I didn't analyze closely.
In future we could look into offering a similar optimized hashing function
for 8-byte keys. Under this design that could be done in a centralized
and machine-independent fashion, whereas getting it right for keys of
platform-dependent sizes would've been notationally painful before.
For the moment, the old way still works fine, so as not to break source
code compatibility for loadable modules. Eventually we might want to
remove tag_hash and friends from the exported API altogether, since there's
no real need for them to be explicitly referenced from outside dynahash.c.
Teodor Sigaev and Tom Lane
Transactions can now set their commit timestamp directly as they commit,
or an external transaction commit timestamp can be fed from an outside
system using the new function TransactionTreeSetCommitTsData(). This
data is crash-safe, and truncated at Xid freeze point, same as pg_clog.
This module is disabled by default because it causes a performance hit,
but can be enabled in postgresql.conf requiring only a server restart.
A new test in src/test/modules is included.
Catalog version bumped due to the new subdirectory within PGDATA and a
couple of new SQL functions.
Authors: Álvaro Herrera and Petr Jelínek
Reviewed to varying degrees by Michael Paquier, Andres Freund, Robert
Haas, Amit Kapila, Fujii Masao, Jaime Casanova, Simon Riggs, Steven
Singer, Peter Eisentraut
Each WAL record now carries information about the modified relation and
block(s) in a standardized format. That makes it easier to write tools that
need that information, like pg_rewind, prefetching the blocks to speed up
recovery, etc.
There's a whole new API for building WAL records, replacing the XLogRecData
chains used previously. The new API consists of XLogRegister* functions,
which are called for each buffer and chunk of data that is added to the
record. The new API also gives more control over when a full-page image is
written, by passing flags to the XLogRegisterBuffer function.
This also simplifies the XLogReadBufferForRedo() calls. The function can dig
the relation and block number from the WAL record, so they no longer need to
be passed as arguments.
For the convenience of redo routines, XLogReader now disects each WAL record
after reading it, copying the main data part and the per-block data into
MAXALIGNed buffers. The data chunks are not aligned within the WAL record,
but the redo routines can assume that the pointers returned by XLogRecGet*
functions are. Redo routines are now passed the XLogReaderState, which
contains the record in the already-disected format, instead of the plain
XLogRecord.
The new record format also makes the fixed size XLogRecord header smaller,
by removing the xl_len field. The length of the "main data" portion is now
stored at the end of the WAL record, and there's a separate header after
XLogRecord for it. The alignment padding at the end of XLogRecord is also
removed. This compansates for the fact that the new format would otherwise
be more bulky than the old format.
Reviewed by Andres Freund, Amit Kapila, Michael Paquier, Alvaro Herrera,
Fujii Masao.
Unlogged relations are only reset when performing a unclean
restart. That means they have to be synced to disk during clean
shutdowns. During normal processing that's achieved by registering a
buffer's file to be fsynced at the next checkpoint when flushed. But
ResetUnloggedRelations() doesn't go through the buffer manager, so
nothing will force reset relations to disk before the next shutdown
checkpoint.
So just make ResetUnloggedRelations() fsync the newly created main
forks to disk.
Discussion: 20140912112246.GA4984@alap3.anarazel.de
Backpatch to 9.1 where unlogged tables were introduced.
Abhijit Menon-Sen and Andres Freund
There was a window in RestoreBackupBlock where a page would be zeroed out,
but not yet locked. If a backend pinned and locked the page in that window,
it saw the zeroed page instead of the old page or new page contents, which
could lead to missing rows in a result set, or errors.
To fix, replace RBM_ZERO with RBM_ZERO_AND_LOCK, which atomically pins,
zeroes, and locks the page, if it's not in the buffer cache already.
In stable branches, the old RBM_ZERO constant is renamed to RBM_DO_NOT_USE,
to avoid breaking any 3rd party extensions that might use RBM_ZERO. More
importantly, this avoids renumbering the other enum values, which would
cause even bigger confusion in extensions that use ReadBufferExtended, but
haven't been recompiled.
Backpatch to all supported versions; this has been racy since hot standby
was introduced.
BRIN is a new index access method intended to accelerate scans of very
large tables, without the maintenance overhead of btrees or other
traditional indexes. They work by maintaining "summary" data about
block ranges. Bitmap index scans work by reading each summary tuple and
comparing them with the query quals; all pages in the range are returned
in a lossy TID bitmap if the quals are consistent with the values in the
summary tuple, otherwise not. Normal index scans are not supported
because these indexes do not store TIDs.
As new tuples are added into the index, the summary information is
updated (if the block range in which the tuple is added is already
summarized) or not; in the latter case, a subsequent pass of VACUUM or
the brin_summarize_new_values() function will create the summary
information.
For data types with natural 1-D sort orders, the summary info consists
of the maximum and the minimum values of each indexed column within each
page range. This type of operator class we call "Minmax", and we
supply a bunch of them for most data types with B-tree opclasses.
Since the BRIN code is generalized, other approaches are possible for
things such as arrays, geometric types, ranges, etc; even for things
such as enum types we could do something different than minmax with
better results. In this commit I only include minmax.
Catalog version bumped due to new builtin catalog entries.
There's more that could be done here, but this is a good step forwards.
Loosely based on ideas from Simon Riggs; code mostly by Álvaro Herrera,
with contribution by Heikki Linnakangas.
Patch reviewed by: Amit Kapila, Heikki Linnakangas, Robert Haas.
Testing help from Jeff Janes, Erik Rijkers, Emanuel Calvo.
PS:
The research leading to these results has received funding from the
European Union's Seventh Framework Programme (FP7/2007-2013) under
grant agreement n° 318633.
xlog.c is huge, this makes it a little bit smaller, which is nice. Functions
related to putting together the WAL record are in xloginsert.c, and the
lower level stuff for managing WAL buffers and such are in xlog.c.
Also move the definition of XLogRecord to a separate header file. This
causes churn in the #includes of all the files that write WAL records, and
redo routines, but it avoids pulling in xlog.h into most places.
Reviewed by Michael Paquier, Alvaro Herrera, Andres Freund and Amit Kapila.
This reassociates a dynamic shared memory handle previous passed to
dsm_pin_mapping with the current resource owner, so that it will be
cleaned up at the end of the current query.
Patch by me. Review of the function name by Andres Freund, Amit
Kapila, Jim Nasby, Petr Jelinek, and Álvaro Herrera.
Nobody seemed concerned about this naming when it originally went in,
but there's a pending patch that implements the opposite of
dsm_keep_mapping, and the term "unkeep" was judged unpalatable.
"unpin" has existing precedent in the PostgreSQL code base, and the
English language, so use this terminology instead.
Per discussion, back-patch to 9.4.
CREATE DATABASE and ALTER DATABASE .. SET TABLESPACE copy the source
database directory on the filesystem level. To ensure the on disk
state is consistent they block out users of the affected database and
force a checkpoint to flush out all data to disk. Unfortunately, up to
now, that checkpoint didn't flush out dirty buffers from unlogged
relations.
That bug means there could be leftover dirty buffers in either the
template database, or the database in its old location. Leading to
problems when accessing relations in an inconsistent state; and to
possible problems during shutdown in the SET TABLESPACE case because
buffers belonging files that don't exist anymore are flushed.
This was reported in bug #10675 by Maxim Boguk.
Fix by Pavan Deolasee, modified somewhat by me. Reviewed by MauMau and
Fujii Masao.
Backpatch to 9.1 where unlogged tables were introduced.
LWLockRelease should release all backends waiting with LWLockWaitForVar,
even when another backend has already been woken up to acquire the lock,
i.e. when releaseOK is false. LWLockWaitForVar can return as soon as the
protected value changes, even if the other backend will acquire the lock.
Fix that by resetting releaseOK to true in LWLockWaitForVar, whenever
adding itself to the wait queue.
This should fix the bug reported by MauMau, where the system occasionally
hangs when there is a lot of concurrent WAL activity and a checkpoint.
Backpatch to 9.4, where this code was added.
shm_mq_sendv sends a message to the queue assembled from multiple
locations. This is expected to be used by forthcoming patches to
allow frontend/backend protocol messages to be sent via shm_mq, but
might be useful for other purposes as well.
shm_mq_set_handle associates a BackgroundWorkerHandle with an
already-existing shm_mq_handle. This solves a timing problem when
creating a shm_mq to communicate with a newly-launched background
worker: if you attach to the queue first, and the background worker
fails to start, you might block forever trying to do I/O on the queue;
but if you start the background worker first, but then die before
attaching to the queue, the background worrker might block forever
trying to do I/O on the queue. This lets you attach before starting
the worker (so that the worker is protected) and then associate the
BackgroundWorkerHandle later (so that you are also protected).
Patch by me, reviewed by Stephen Frost.
Testing by Amit Kapila, Andres Freund, and myself, with and without
other patches that also aim to improve scalability, seems to indicate
that this change is a significant win over the current value and over
smaller values such as 64. It's not clear how high we can push this
value before it starts to have negative side-effects elsewhere, but
going this far looks OK.
Several upcoming performance/scalability improvements require atomic
operations. This new API avoids the need to splatter compiler and
architecture dependent code over all the locations employing atomic
ops.
For several of the potential usages it'd be problematic to maintain
both, a atomics using implementation and one using spinlocks or
similar. In all likelihood one of the implementations would not get
tested regularly under concurrency. To avoid that scenario the new API
provides a automatic fallback of atomic operations to spinlocks. All
properties of atomic operations are maintained. This fallback -
obviously - isn't as fast as just using atomic ops, but it's not bad
either. For one of the future users the atomics ontop spinlocks
implementation was actually slightly faster than the old purely
spinlock using implementation. That's important because it reduces the
fear of regressing older platforms when improving the scalability for
new ones.
The API, loosely modeled after the C11 atomics support, currently
provides 'atomic flags' and 32 bit unsigned integers. If the platform
efficiently supports atomic 64 bit unsigned integers those are also
provided.
To implement atomics support for a platform/architecture/compiler for
a type of atomics 32bit compare and exchange needs to be
implemented. If available and more efficient native support for flags,
32 bit atomic addition, and corresponding 64 bit operations may also
be provided. Additional useful atomic operations are implemented
generically ontop of these.
The implementation for various versions of gcc, msvc and sun studio have
been tested. Additional existing stub implementations for
* Intel icc
* HUPX acc
* IBM xlc
are included but have never been tested. These will likely require
fixes based on buildfarm and user feedback.
As atomic operations also require barriers for some operations the
existing barrier support has been moved into the atomics code.
Author: Andres Freund with contributions from Oskari Saarenmaa
Reviewed-By: Amit Kapila, Robert Haas, Heikki Linnakangas and Álvaro Herrera
Discussion: CA+TgmoYBW+ux5-8Ja=Mcyuy8=VXAnVRHp3Kess6Pn3DMXAPAEA@mail.gmail.com,
20131015123303.GH5300@awork2.anarazel.de,
20131028205522.GI20248@awork2.anarazel.de
Previously, we used an lwlock that was held from the time we began
seeking a candidate buffer until the time when we found and pinned
one, which is disastrous for concurrency. Instead, use a spinlock
which is held just long enough to pop the freelist or advance the
clock sweep hand, and then released. If we need to advance the clock
sweep further, we reacquire the spinlock once per buffer.
This represents a significant increase in atomic operations around
buffer eviction, but it still wins on many workloads. On others, it
may result in no gain, or even cause a regression, unless the number
of buffer mapping locks is also increased. However, that seems like
material for a separate commit. We may also need to consider other
methods of mitigating contention on this spinlock, such as splitting
it into multiple locks or jumping the clock sweep hand more than one
buffer at a time, but those, too, seem like separate improvements.
Patch by me, inspired by a much larger patch from Amit Kapila.
Reviewed by Andres Freund.
Now that spinlocks (hopefully!) act as compiler barriers, as of commit
0709b7ee72, this should be safe. This
serves as a demonstration of the new coding style, and may be optimized
better on some machines as well.
Previously, they functioned as barriers against CPU reordering but not
compiler reordering, an odd API that required extensive use of volatile
everywhere that spinlocks are used. That's error-prone and has negative
implications for performance, so change it.
In theory, this makes it safe to remove many of the uses of volatile
that we currently have in our code base, but we may find that there are
some bugs in this effort when we do. In the long run, though, this
should make for much more maintainable code.
Patch by me. Review by Andres Freund.
68a2e52bba has introduced LWLockAcquireCommon() containing the
previous contents of LWLockAcquire() plus added functionality. The
latter then calls it, just like LWLockAcquireWithVar(). Because the
majority of callers don't need the added functionality, declare the
common code as inline. The compiler then can optimize away the unused
code. Doing so is also useful when looking at profiles, to
differentiate the users.
Backpatch to 9.4, the first branch to contain LWLockAcquireCommon().
Since the dawn of time (aka Postgres95) multiple pins of the same
buffer by one backend have been optimized not to modify the shared
refcount more than once. This optimization has always used a NBuffer
sized array in each backend keeping track of a backend's pins.
That array (PrivateRefCount) was one of the biggest per-backend memory
allocations, depending on the shared_buffers setting. Besides the
waste of memory it also has proven to be a performance bottleneck when
assertions are enabled as we make sure that there's no remaining pins
left at the end of transactions. Also, on servers with lots of memory
and a correspondingly high shared_buffers setting the amount of random
memory accesses can also lead to poor cpu cache efficiency.
Because of these reasons a backend's buffers pins are now kept track
of in a small statically sized array that overflows into a hash table
when necessary. Benchmarks have shown neutral to positive performance
results with considerably lower memory usage.
Patch by me, review by Robert Haas.
Discussion: 20140321182231.GA17111@alap3.anarazel.de
shm_mq_send_bytes didn't invariably initialize *bytes_written before
returning, which would cause shm_mq_send to read from uninitialized
memory and add the value it found there to mqh->mqh_partial_bytes.
This could cause the next attempt to send a message via the queue to
fail an assertion (if the queue was detached) or copy data from a
garbage pointer value into the queue (if non-blocking mode was in use).
get_raw_page tried to validate the supplied block number against
RelationGetNumberOfBlocks(), which of course is only right when
accessing the main fork. In most cases, the main fork is longer
than the others, so that the check was too weak (allowing a
lower-level error to be reported, but no real harm to be done).
However, very small tables could have an FSM larger than their heap,
in which case the mistake prevented access to some FSM pages.
Per report from Torsten Foertsch.
In passing, make the bad-block-number error into an ereport not elog
(since it's certainly not an internal error); and fix sloppily
maintained comment for RelationGetNumberOfBlocksInFork.
This has been wrong since we invented relation forks, so back-patch
to all supported branches.
The assertion failed if WAL_DEBUG or LWLOCK_STATS was enabled; fix that by
using separate memory contexts for the allocations made within those code
blocks.
This patch introduces a mechanism for marking any memory context as allowed
in a critical section. Previously ErrorContext was exempt as a special case.
Instead of a blanket exception of the checkpointer process, only exempt the
memory context used for the pending ops hash table.
The existance of the assert_enabled variable (backing the
debug_assertions GUC) reduced the amount of knowledge some static code
checkers (like coverity and various compilers) could infer from the
existance of the assertion. That could have been solved by optionally
removing the assertion_enabled variable from the Assert() et al macros
at compile time when some special macro is defined, but the resulting
complication doesn't seem to be worth the gain from having
debug_assertions. Recompiling is fast enough.
The debug_assertions GUC is still available, but readonly, as it's
useful when diagnosing problems. The commandline/client startup option
-A, which previously also allowed to enable/disable assertions, has
been removed as it doesn't serve a purpose anymore.
While at it, reduce code duplication in bufmgr.c and localbuf.c
assertions checking for spurious buffer pins. That code had to be
reindented anyway to cope with the assert_enabled removal.
It's critical that the backend's idea of LOBLKSIZE match the way data has
actually been divided up in pg_largeobject. While we don't provide any
direct way to adjust that value, doing so is a one-line source code change
and various people have expressed interest recently in changing it. So,
just as with TOAST_MAX_CHUNK_SIZE, it seems prudent to record the value in
pg_control and cross-check that the backend's compiled-in setting matches
the on-disk data.
Also tweak the code in inv_api.c so that fetches from pg_largeobject
explicitly verify that the length of the data field is not more than
LOBLKSIZE. Formerly we just had Asserts() for that, which is no protection
at all in production builds. In some of the call sites an overlength data
value would translate directly to a security-relevant stack clobber, so it
seems worth one extra runtime comparison to be sure.
In the back branches, we can't change the contents of pg_control; but we
can still make the extra checks in inv_api.c, which will offer some amount
of protection against running with the wrong value of LOBLKSIZE.
The previous coding would potentially cause attaching to segment A to
fail if segment B was at the same time in the process of going away.
Andres Freund, with a comment tweak by me
Commit fad153ec45 modified sinval.c to reduce
the number of calls into sinvaladt.c (which require taking a shared lock)
by keeping a local buffer of collected-but-not-yet-processed messages.
However, if processing of the last message in a batch resulted in a
recursive call to ReceiveSharedInvalidMessages, we could overwrite that
message with a new one while the outer invalidation function was still
working on it. This would be likely to lead to invalidation of the wrong
cache entry, allowing subsequent processing to use stale cache data.
The fix is just to make a local copy of each message while we're processing
it.
Spotted by Andres Freund. Back-patch to 8.4 where the bug was introduced.
This was intended to work always, but the previous code only allowed
it if at least one message was successfully read by the receiver
before the sender detached the queue.
Report by Petr Jelinek. Patch by me.
Commit a730183926 created rather a mess by
putting dependencies on backend-only include files into include/common.
We really shouldn't do that. To clean it up:
* Move TABLESPACE_VERSION_DIRECTORY back to its longtime home in
catalog/catalog.h. We won't consider this symbol part of the FE/BE API.
* Push enum ForkNumber from relfilenode.h into relpath.h. We'll consider
relpath.h as the source of truth for fork numbers, since relpath.c was
already partially serving that function, and anyway relfilenode.h was
kind of a random place for that enum.
* So, relfilenode.h now includes relpath.h rather than vice-versa. This
direction of dependency is fine. (That allows most, but not quite all,
of the existing explicit #includes of relpath.h to go away again.)
* Push forkname_to_number from catalog.c to relpath.c, just to centralize
fork number stuff a bit better.
* Push GetDatabasePath from catalog.c to relpath.c; it was rather odd
that the previous commit didn't keep this together with relpath().
* To avoid needing relfilenode.h in common/, redefine the underlying
function (now called GetRelationPath) as taking separate OID arguments,
and make the APIs using RelFileNode or RelFileNodeBackend into macro
wrappers. (The macros have a potential multiple-eval risk, but none of
the existing call sites have an issue with that; one of them had such a
risk already anyway.)
* Fix failure to follow the directions when "init" fork type was added;
specifically, the errhint in forkname_to_number wasn't updated, and neither
was the SGML documentation for pg_relation_size().
* Fix tablespace-path-too-long check in CreateTableSpace() to account for
fork-name component of maximum-length pathnames. This requires putting
FORKNAMECHARS into a header file, but it was rather useless (and
actually unreferenced) where it was.
The last couple of items are potentially back-patchable bug fixes,
if anyone is sufficiently excited about them; but personally I'm not.
Per a gripe from Christoph Berg about how include/common wasn't
self-contained.
Original coding failed to enlarge the array as required if
the requested tranche_id was equal to LWLockTranchesAllocated.
In passing, fix poor style of not casting the result of (re)palloc.
Apparently, Windows can sometimes return an error code even when the
operation actually worked just fine. Rearrange the order of checks
according to what appear to be the best practices in this area.
Amit Kapila
Instead of storing the ID of the dynamic shared memory control
segment in a file within the data directory, store it in the main
control segment. This avoids a number of nasty corner cases,
most seriously that doing an online backup and then using it on
the same machine (e.g. to fire up a standby) would result in the
standby clobbering all of the master's dynamic shared memory
segments.
Per complaints from Heikki Linnakangas, Fujii Masao, and Tom
Lane.
The one existing assertion of this type has tripped a few times in the
buildfarm lately, but it's not clear whether the problem is really
originating there or whether it's leftovers from a trip through one
of the other two paths that lack a matching assertion. So add one.
Since the same bug(s) most likely exist(s) in the back-branches also,
back-patch to 9.2, where the fast-path lock mechanism was added.
Otherwise, the compiler might decide to move modifications to data
within this structure outside the enclosing SpinLockAcquire /
SpinLockRelease pair, leading to shared memory corruption.
This may or may not explain a recent lmgr-related buildfarm failure
on prairiedog, but it needs to be fixed either way.
Previously, such buffers weren't counted, with the possible result that
EXPLAIN (BUFFERS) and pg_stat_statements would understate the true
number of blocks dirtied by an SQL statement.
Back-patch to 9.2, where this counter was introduced.
Amit Kapila
Also fix the name of the dtrace probe for LWLockAcquireOrWait(). The
function was renamed from LWLockWaitUntilFree to LWLockAqcuireOrWait, but
the dtrace probe was neglected.
Pointed out by Andres Freund and the buildfarm.
Clear errno before calling readdir() and handle old MinGW errno bug
while adding full test coverage for readdir/closedir failures.
Backpatch through 8.4.
The special feature the XLogInsert slots had over regular LWLocks is the
insertingAt value that was updated atomically with releasing backends
waiting on it. Add new functions to the LWLock API to do that, and replace
the slots with LWLocks. This reduces the amount of duplicated code.
(There's still some duplication, but at least it's all in lwlock.c now.)
Reviewed by Andres Freund.
With this in place, a session blocking behind another one because of
tuple locks will get a context line mentioning the relation name, tuple
TID, and operation being done on tuple. For example:
LOG: process 11367 still waiting for ShareLock on transaction 717 after 1000.108 ms
DETAIL: Process holding the lock: 11366. Wait queue: 11367.
CONTEXT: while updating tuple (0,2) in relation "foo"
STATEMENT: UPDATE foo SET value = 3;
Most usefully, the new line is displayed by log entries due to
log_lock_waits, although of course it will be printed by any other log
message as well.
Author: Christian Kruse, some tweaks by Álvaro Herrera
Reviewed-by: Amit Kapila, Andres Freund, Tom Lane, Robert Haas
Revise the original decision to expose a uint64-based interface and
use Size everywhere possible. Avoid assuming that MAXIMUM_ALIGNOF is
8, or making any assumption about the relationship between that value
and sizeof(Size). If MAXIMUM_ALIGNOF is bigger, we'll now insert
padding after the length word; if it's smaller, we are now prepared
to read and write the length word in chunks.
Per discussion with Tom Lane.
The new function dsm_detach_all() can be used either by postmaster
children that don't wish to take any risk of accidentally corrupting
shared memory; or by forked children of regular backends with
the same need. This patch also updates the postmaster children that
already do PGSharedMemoryDetach() to do dsm_detach_all() as well.
Per discussion with Tom Lane.
If a postmaster child invokes fork() and then calls on_exit_reset, that
should be sufficient to let it exit() without breaking anything, but
dynamic shared memory broke that by not updating on_exit_reset() to
discard callbacks registered with dynamic shared memory segments.
Per investigation of a complaint from Tom Lane.
Andres Freund
Tom Lane
Kevin Grittner
Robert Haas
Simon Riggs
Noah Misch
Bruce Momjian
Peter Eisentraut
Alvaro Herrera
Heikki Linnakangas
Stephen Frost
Fujii Masao