[pulseaudio-discuss] pa_once can run twice?

Dalleau, Frederic frederic.dalleau at intel.com
Mon Apr 23 08:03:37 PDT 2012

Hi David,

On Fri, Apr 20, 2012 at 4:44 PM, David Henningsson
<david.henningsson at canonical.com> wrote:
> While researching a bug I came across something that might be a bug in the
> pa_once logic, but this stuff is tricky, so I might also be missing
> something.
> Imagine this:
>  * Thread 1 runs pa_once_begin, succeeds and starts running the payload (i e
> the code that should only run once).
>  * Thread 2 starts running pa_once_begin, but only the first row. We're now
> right *before* pa_atomic_inc(&control->ref) but *after*
> pa_atomic_load(&control->done).
>  * Thread 1 finishes the payload, runs pa_once_done which sets control->done
> and frees the mutex.
>  * Thread 2 continues, pa_once_begin succeeds and the payload is now run a
> second time!

After reading your mail, I made some experiments by adding a usleep() call
in Thread 1 between pa_atomic_load(&control->done) and
and that failed once-test 100% of time.

I reverted the usleep and made another experiment using 50000 iterations in
once-test and it just failed.

Good catch !


More information about the pulseaudio-discuss mailing list