Horton

Re: LogwrtResult contended spinlock

Alvaro Herrera <alvherre@alvh.no-ip.org>

View thread View on postgr.es
From: Alvaro Herrera <alvherre@alvh.no-ip.org>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Alexander Lakhin <exclusion@gmail.com>, Bharath Rupireddy <bharath.rupireddyforpostgres@gmail.com>, Andres Freund <andres@anarazel.de>, Jeff Davis <pgsql@j-davis.com>, pgsql-hackers@lists.postgresql.org, Jaime Casanova <jcasanov@systemguards.com.ec>
Date: 2024-07-01T14:35:02Z
Lists: pgsql-hackers 
On 2024-Jul-01, Tom Lane wrote:

> Alvaro Herrera <alvherre@alvh.no-ip.org> writes:
> >> because the failed assertion is:
> >> #ifndef PG_HAVE_ATOMIC_U64_SIMULATION
> >>     AssertPointerAlignment(&currval, 8);
> >> #endif
> 
> Perhaps this assertion is what is wrong?  If the platform has no
> native 8-byte alignment requirement, why do we think that atomics
> need it?

Oh, that's a good question.  TBH I just copied the assertion from the
other routines for 64-bit variables in the same file.  But I think
that's correct.  We're gating the assertion on _not_ having emulation,
which must mean we have native atomics; on MSVC, if I read the #ifdef
maze correctly, that's implemented using _InterlockedCompareExchange,
whose docs state:

: The variables for this function must be aligned on a 64-bit boundary;
: otherwise, this function will behave unpredictably on multiprocessor x86
: systems and any non-x86 systems. See _aligned_malloc.
https://learn.microsoft.com/en-us/windows/win32/api/winnt/nf-winnt-interlockedcompareexchange64

So I think the assertion is correct.

-- 
Álvaro Herrera         PostgreSQL Developer  —  https://www.EnterpriseDB.com/
"[PostgreSQL] is a great group; in my opinion it is THE best open source
development communities in existence anywhere."                (Lamar Owen)

Commits

  1. Remove bogus assertion in pg_atomic_monotonic_advance_u64

  2. Add XLogCtl->logInsertResult

  3. Operate XLogCtl->log{Write,Flush}Result with atomics

  4. Split XLogCtl->LogwrtResult into separate struct members

  5. Introduce atomic read/write functions with full barrier semantics.

  6. Reduce the number of GetFlushRecPtr() calls done by walsenders.

  7. Remove most volatile qualifiers from xlog.c