[Bug 47458] New: Sleeping while pthreads mutex held from the driver code in winepulse-PulseAudio_Support

https://bugs.winehq.org/show_bug.cgi?id=47458

Bug ID: 47458
           Summary: Sleeping while pthreads mutex held from the driver
                    code in winepulse-PulseAudio_Support
           Product: Wine-staging
           Version: unspecified
          Hardware: x86
                OS: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: -unknown
          Assignee: [e-mail]
          Reporter: [e-mail]
                CC: [e-mail], [e-mail]
      Distribution: ---

The SetEvent call in the function pulse_timer_cb() sleeps, calls into the
wineserver, while a pthread mutex is held. This can cause long delays in
releasing the "pulse_lock" mutex, anywhere from 1-2ms to 90ms that I can see
with my custom instrumentation.

It potentially can cause dropouts and other issues with audio handling. It
seems to be in the main core audio processing loop.

The general practice here is to avoid sleeping operations within critical
sections.

[Bug 47458] Sleeping while pthreads mutex held from the driver code in winepulse-PulseAudio_Support

https://bugs.winehq.org/show_bug.cgi?id=47458

--- Comment #1 from Bill Huey (hui)  ---
This is against master with patches from wine staging,
winepulse-PulseAudio_Support

[Bug 47458] Sleeping while pthreads mutex held from the driver code in winepulse-PulseAudio_Support

https://bugs.winehq.org/show_bug.cgi?id=47458

Bill Huey (hui)  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail],
                   |                            |[e-mail]
       Distribution|---                         |Ubuntu

[Bug 47458] Sleeping while pthreads mutex held from the driver code in winepulse-PulseAudio_Support

https://bugs.winehq.org/show_bug.cgi?id=47458

--- Comment #2 from Alistair Leslie-Hughes  ---
So moving this lines 
if (This->event)
      SetEvent(This->event);

after the pthread_mutex_unlock call would fix the issue you are experiencing.

[Bug 47458] Sleeping while pthreads mutex held from the driver code in winepulse-PulseAudio_Support

https://bugs.winehq.org/show_bug.cgi?id=47458

--- Comment #3 from Bill Huey (hui)  ---
I don't know the driver that well yet, but it seems like there has to be
consideration to AudioClient_SetEventHandle() as well. It looks like that
values is set there and it's also held within pulse_lock mutex for safely
setting those values.

Please, correct me if I'm wrong as I'm new to this code.

[Bug 47458] Sleeping while pthreads mutex held from the driver code in winepulse-PulseAudio_Support

https://bugs.winehq.org/show_bug.cgi?id=47458

--- Comment #4 from Zebediah Figura  ---
Do we know for a fact that calling SetEvent() within a mutex does cause
dropouts or other poor audio performance?

[Bug 47458] Sleeping while pthreads mutex held from the driver code in winepulse-PulseAudio_Support

https://bugs.winehq.org/show_bug.cgi?id=47458

--- Comment #5 from Bill Huey (hui)  ---
No, but in my test case for Ableton, it's glitching quite badly.

I instrumented the pthreads stuff with pthreads_mutex_trylock as the front end
and I'm getting mutex contentions with owners from multimedia DLLs as well as
the pulse driver which could be contentions. It's hard to tell what's going on
and if this is a normal behavior of the condition variable or not.

The case I'm working with is particularly demanding and having this delay in
processing is definitely not going to help things