Horton

Re: Improve LWLock tranche name visibility across backends

Sami Imseih <samimseih@gmail.com>

View thread View on postgr.es
From: Sami Imseih <samimseih@gmail.com>
To: Nathan Bossart <nathandbossart@gmail.com>
Cc: Tom Lane <tgl@sss.pgh.pa.us>, Andres Freund <andres@anarazel.de>, Bertrand Drouvot <bertranddrouvot.pg@gmail.com>, Rahila Syed <rahilasyed90@gmail.com>, pgsql-hackers <pgsql-hackers@postgresql.org>
Date: 2025-08-26T00:37:05Z
Lists: pgsql-hackers 
> On Mon, Aug 25, 2025 at 04:59:41PM -0500, Sami Imseih wrote:
> > hmm, can we really avoid a shared lock when reading from shared memory?
> > considering access for both reads and writes can be concurrent to shared
> > memory. We are also taking an exclusive lock when writing a new tranche.
>
> We probably want to hold a lock while we 1) increment LWLockCounter and
> copy a new tranche name to memory and

In the last rev, I removed the spinlock acquired on ShmemLock in-lieu of
a LWLock. This is because I wanted a single LWLock acquisition while
both incrementing LWLockCounter and writing to shared memory, and
doing this much work, particularly writing to shared memory,
with a spinlock seemed inappropriate. With that said, this is not high
concurrency of performance sensitive activity at all, so perhaps I was
being overly paranoid.

--
Sami

Commits

  1. test_dsa: Avoid leaking LWLock tranches.

  2. Teach DSM registry to ERROR if attaching to an uninitialized entry.

  3. Add a test harness for the LWLock tranche code.

  4. Revert recent change to RequestNamedLWLockTranche().

  5. Move dynamically-allocated LWLock tranche names to shared memory.

  6. Prepare DSM registry for upcoming changes to LWLock tranche names.

  7. Add GetNamedDSA() and GetNamedDSHash().