Horton

Re: Is this a problem in GenericXLogFinish()?

Jeff Davis <pgsql@j-davis.com>

View thread View on postgr.es
From: Jeff Davis <pgsql@j-davis.com>
To: Heikki Linnakangas <hlinnaka@iki.fi>, Robert Haas <robertmhaas@gmail.com>
Cc: pgsql-hackers@postgresql.org
Date: 2023-09-28T19:05:00Z
Lists: pgsql-hackers 
On Wed, 2023-09-27 at 18:57 +0300, Heikki Linnakangas wrote:
> We could define a new REGBUF_NO_CHANGE flag to signal that the buffer
> is 
> registered just for locking purposes, and not actually modified by
> the 
> WAL record.

Out of curiosity I also added a trial assert (not intending to actually
add this):

  Assert(!(flags & REGBUF_NO_CHANGE) || !BufferIsExclusiveLocked());

I expected that to fail, but it didn't -- perhaps that buffer is only
locked during replay or something? I don't think we want that specific
Assert; I was just experimenting with some cross-checks we could do to
verify that REGBUF_NO_CHANGE is used properly.

Also, I ran into some problems with GIN that might require a bit more
refactoring in ginPlaceToPage(). Perhaps we could consider
REGBUF_NO_CHANGE a general bypass of the Assert(BufferIsDirty()), and
use it temporarily until we have a chance to analyze/refactor each of
the callers that need it.

Regards,
        Jeff Davis

Commits

  1. Fix inconsistency with replay of hash squeeze record for clean buffers

  2. Set LSN for wbuf in _hash_freeovflpage() iff wbuf is modified.

  3. Fix an uninitialized access in hash_xlog_squeeze_page().

  4. Use REGBUF_NO_CHANGE at one more place in the hash index.

  5. Assert that buffers are marked dirty before XLogRegisterBuffer().

  6. Fix bug in GenericXLogFinish().

  7. Improve hash index bucket split behavior.