Horton

Re: BUG #19438: segfault with temp_file_limit inside cursor

Tom Lane <tgl@sss.pgh.pa.us>

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: David Rowley <dgrowleyml@gmail.com>
Cc: kuzmin.db4@gmail.com, pgsql-bugs@lists.postgresql.org
Date: 2026-03-29T15:32:54Z
Lists: pgsql-bugs

Attachments

David Rowley <dgrowleyml@gmail.com> writes:
> For the switching MemoryContextMethodID patch, I applied the memory
> context benchmarking patch I used when writing that code to test out
> the overhead in a tight palloc/pfree loop (attached). I can see an
> overhead of a little over 6.5%.

Hm.  I got an overhead of about 2% on an Apple M4, which might be
argued to be acceptable, but 12% on an aging x86_64 platform.
Realistically, given that we failed to notice this omission at
all for more than three years, it's hard to argue that testing
for it in non-debug builds is worth any overhead.

Here's a fleshed-out version of the requested_size method.
I noted that AllocSetRealloc needs a defense too, and then
extended the patch to generation.c and slab.c.  bump.c
doesn't have an issue, and I don't think alignedalloc.c
needs its own defense either: it can rely on the underlying
context type.

			regards, tom lane

Commits

  1. Be more careful to preserve consistency of a tuplestore.

  2. Detect pfree or repalloc of a previously-freed memory chunk.

  3. Improve performance of and reduce overheads of memory management