Horton

Re: `pg_ctl init` crashes when run concurrently; semget(2) suspected

Thomas Munro <thomas.munro@gmail.com>

View thread View on postgr.es
From: Thomas Munro <thomas.munro@gmail.com>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Gavin Panella <gavinpanella@gmail.com>, pgsql-hackers@postgresql.org
Date: 2025-08-13T00:45:19Z
Lists: pgsql-hackers 
On Wed, Aug 13, 2025 at 11:29 AM Thomas Munro <thomas.munro@gmail.com> wrote:
> FWIW in early prototype multithreading patches you can just use
> sem_init() on all these systems since you don't need pshared=1.

Oops, I misremembered that.  It works on NetBSD and OpenBSD, but not
macOS :-(.

Hmm, but it looks like there has been a new development in macOS 15.
They have finally made the futex API public:

https://developer.apple.com/documentation/os/os_sync_wait_on_address

That was one of the sticking points for my patch that added our own
replacement sem_init() implementation.  At the time the only "proper"
way to reach the undocumented and undeclared
__ulock_wait()/__ulock_wake() functions was to write a module in C++
using std::atomic<int>::wait()/notify(), which I had started to wonder
about, but here it is with a shiny new C interface and documentation.

Maybe I should have another go at that!

Commits

  1. Don't treat EINVAL from semget() as a hard failure.

  2. Give up on running with NetBSD/OpenBSD's default semaphore settings.