Re: Improve LWLock tranche name visibility across backends

From: Sami Imseih <samimseih(at)gmail(dot)com>
To: Bertrand Drouvot <bertranddrouvot(dot)pg(at)gmail(dot)com>
Cc: Nathan Bossart <nathandbossart(at)gmail(dot)com>, pgsql-hackers <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: Improve LWLock tranche name visibility across backends
Date: 2025-08-05 15:39:52
Message-ID: CAA5RZ0tU9j-jEPUCS0=K6PCzYY=291O9Uky9ndbqH_BDp4Hc=Q@mail.gmail.com
Views: Whole Thread | Raw Message | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Thanks for reviewing!

> Issue 1 --
>
> If I register enough tranches to go to:
>
> + /* Resize if needed */
> + if (LWLockTrancheNames.shmem->count >= LWLockTrancheNames.shmem->allocated)
> + {
> + newalloc = pg_nextpower2_32(Max(LWLOCK_TRANCHE_NAMES_INIT_SIZE, LWLockTrancheNames.shmem->allocated + 1));
> + new_list_ptr = dsa_allocate(LWLockTrancheNames.dsa, newalloc * sizeof(dsa_pointer));
> + new_name_ptrs = dsa_get_address(LWLockTrancheNames.dsa, new_list_ptr);
> + memcpy(new_name_ptrs, current_ptrs, LWLockTrancheNames.shmem->allocated * sizeof(dsa_pointer));
> + dsa_free(LWLockTrancheNames.dsa, *current_ptrs);
>
> then I get:
>
> ERROR: dsa_area could not attach to a segment that has been freed

Will investigate this one and correct in the next rev.

> Issue 2 --
>
> If an extension calls RequestNamedLWLockTranche() it will register the same
> tranche twice:
>
> (gdb) p LWLockTrancheNames.local[0]
> $1 = 0x7acf5a40c420 "pg_playmem"
> (gdb) p LWLockTrancheNames.local[97]
> $2 = 0x7acf5a40c420 "pg_playmem"

Thanks for catching. This one is clear, there is an extra call to
register inside RequestNamedLWLockTranche.
I'll fix this in the next rev.

--
Sami

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Christoph Berg 2025-08-05 15:44:15 Re: libpq-oauth: a mid-beta naming check
Previous Message Jacob Champion 2025-08-05 15:21:14 Re: libpq-oauth: a mid-beta naming check