Re: Memory leak in pg_stat_statements when qtext file contains invalid encoding

Hi Lukas,

Thank you for the correction on the LWLock. You are right, LWLockReleaseAll
on abort handles that. The leak is limited to the malloc'd qbuffer.

I thought about switching to palloc for the pg_stat_statements_internal
path, but I think it would change the existing OOM behavior in a way that
upstream may not want.

Currently, when qtext_load_file fails on OOM, it returns NULL and the
function continues gracefully, returning rows with NULL query text columns.
The user still gets their result set. With palloc, an OOM
would instead throw a hard ERROR, which changes the semantics from graceful
degradation to a failure.

Additionally, qtext_load_file is called from gc_qtexts (where an ERROR
during garbage collection would abort the user's actual in-flight query)
and from pgss_shmem_shutdown (where an ERROR could interfere with a clean
server stop). Creating a separate palloc-based variant just for
pg_stat_statements_internal would avoid those issues, but it would
still change the OOM behavior from silent degradation to a
visible error for that path.

The PG_TRY/PG_FINALLY approach preserves the existing malloc-based OOM
semantics exactly as they are today. The only thing it adds is
cleanup of the malloc'd buffer when pg_any_to_server throws an encoding
error. In terms of scope, it does not need to wrap the entire function. It
only needs to cover the section after LWLockAcquire where qbuffer is
live through the end of the hash iteration loop, which is where
pg_any_to_server can throw.

I can also scope the PG_FINALLY to just free(qbuffer) since
you confirmed LWLockReleaseAll already handles the lock on abort. That
would make it even more targeted. Happy to send a patch either way.

Apologies for the HTML formatting on the previous email.
I will use plain text going forward.

Thanks,

Gaurav

On Fri, Mar 27, 2026 at 1:52 PM Lukas Fittl <lukas(at)fittl(dot)com> wrote:

> Hi Gaurav,
>
> On Fri, Mar 27, 2026 at 12:54 AM Gaurav Singh <gaurav(dot)singh(at)yugabyte(dot)com>
> wrote:
> > If the qtext file contains an invalid encoding, pg_any_to_server calls
> ereport(ERROR) which longjmps out of the function.
> > The cleanup code at the bottom of the function is never reached.
> >
> > LWLockRelease(pgss->lock);
> > if (qbuffer)
> > free(qbuffer);
> > On every subsequent call, the malloc'd buffer (the entire file contents)
> is leaked, and the LWLock release is also skipped.
>
> I don't think the analysis is correct in regards to the LWLock release
> - that should be taken care of by LWLockReleaseAll on abort.
>
> But I think you're correct about qbuffer - because that buffer is
> using malloc (not palloc), its not part of any memory context, and so
> it will happily leak on abort.
>
> It appears our use of malloc in pg_stat_statements is so that we can
> fail on OOM and return NULL without a jump. I think that makes sense
> for when a GC cycle was triggered during regular query execution
> (since we don't want to error the original query), but it seems like
> just bubbling up the OOM if needed when querying the
> pg_stat_statements function seems fine.
>
> I wonder if its worth separating the two cases, since the issue you're
> describing (the call to pg_any_to_server failing) only happens when
> returning the query text file contents to the client. I think your
> PG_FINALLY suggestion could also work, but it feels a bit tedious to
> wrap the whole pg_stat_statements_internal function in it.
>
> Thanks,
> Lukas
>
> PS: I would recommend reviewing the use of a text format email client
> for posting to the Postgres mailing lists, or significantly reducing
> your formatting when sending HTML emails - your email has a lot of
> styling that is hard to read (even for me in Gmail), and even harder
> to quote in a plain text email response.
>
>
> --
> Lukas Fittl
>