Horton

Re: Is this a problem in GenericXLogFinish()?

Robert Haas <robertmhaas@gmail.com>

View thread View on postgr.es
From: Robert Haas <robertmhaas@gmail.com>
To: Jeff Davis <pgsql@j-davis.com>
Cc: Heikki Linnakangas <hlinnaka@iki.fi>, pgsql-hackers@postgresql.org
Date: 2023-10-23T14:14:27Z
Lists: pgsql-hackers 
On Fri, Oct 20, 2023 at 12:28 PM Jeff Davis <pgsql@j-davis.com> wrote:
> I still have a question though: if a buffer is exclusive-locked,
> unmodified and clean, and then the caller registers it and later does
> PageSetLSN (just as if it were dirty), is that a definite bug?
>
> There are a couple callsites where the control flow is complex enough
> that it's hard to be sure the buffer is always marked dirty before
> being registered (like in log_heap_visible(), as I mentioned upthread).
> But those callsites are all doing PageSetLSN, unlike the hash index
> case.

I think that would be a bug. I think it would be OK to just change the
page LSN and nothing else, but that requires calling MarkBufferDirty()
at some point. If you only call PageSetLSN() and never
MarkBufferDirty(), that sounds messed up: either the former should be
skipped, or the latter should be added. We shouldn't modify a shared
buffer without marking it dirty.

-- 
Robert Haas
EDB: http://www.enterprisedb.com

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.