Starting new thread for this thing that Matthias noticed in my
work-in-progress patch at
https://www.postgresql.org/message-id/CAEze2WjgCROMMXY0+j8FFdm3iFcr7By-+6Mwiz=PgGSEydiW3A@mail.gmail.com.
On 05/04/2026 02:17, Matthias van de Meent wrote:
> 0006: I don't think it is a great idea to make the LwLock machinery
> the first to get allocation requests:
> It has the RequestNamedLWLockTranche infrastructure, which can only
> register new requests while process_shmem_requests_in_progress, and
> making it request its memory ahead of everything else is likely to
> cause an undersized tranche to be allocated. You could make sure that
> this isn't an issue by maintaining a flag in lwlock.c that's set when
> the shmem request is made (and reset on shmem exit), which must be
> false when RequestNamedLWLockTranche() is called, and if not then it
> should throw an error.
Good catch, RequestNamedLWLocktranche() was quite broken with the patch.
I'm surprised it didn't cause test failures. We even have unit tests for
that at src/test/modules/test_lwlock_tranches.
Looking at src/test/modules/test_lwlock_tranches, I realized that we
don't currently check that the tranche name registered with
RequestNamedLWLocktranche() is unique. If two extensions registered a
tranche with same name, we'd allocate two separate tranches for them,
but GetNamedLWLockTranche() would always return the first one.
Attached patches add a uniqueness check, and improves
test_lwlock_tranches so that it actually uses the requested LWLocks. And
I couldn't resist doing some more refactoring of the test while I was at
it; IMO it's more readable now.
Barring objections, I will commit these shortly.
- Heikki
