other: 0.774
performance: 0.744
files: 0.739
semantic: 0.675
graphic: 0.647
device: 0.590
socket: 0.479
network: 0.462
PID: 0.418
permissions: 0.394
boot: 0.375
vnc: 0.307
debug: 0.265
KVM: 0.196

AC97 sound card crashes QEMU

The AC97 sound card does not work. It stops QEMU on startup. The cause appears to be some kind of deadlock. 

Steps to reproduce:
Just add -soundhw ac97 to QEMU's arguments. Example: qemu-system-ppc -soundhw ac97

The example above is all it takes to reproduce the problem.

This problem has been observed on Mac OS X and Debian Linux. 

I question whether the ac97 support ever worked. It is a file that was taken from VirtualBox and added to QEMU. I do know that VirtualBox's support for the ac97 sound card works perfectly. 

The exact line of code that stops QEMU in its tracks is located in the file main-loop.c, in the function os_host_main_loop_wait(), the call made to qemu_mutex_lock_iothread(). The is where QEMU stops under Mac OS X.

This problem also affects qemu-system-i386. 

On 2013-10-31 21:48, John Arbuckle wrote:
> Public bug reported:
> 
> The AC97 sound card does not work. It stops QEMU on startup. The cause
> appears to be some kind of deadlock.
> 
> Steps to reproduce:
> Just add -soundhw ac97 to QEMU's arguments. Example: qemu-system-ppc -soundhw ac97
> 
> The example above is all it takes to reproduce the problem.
> 
> This problem has been observed on Mac OS X and Debian Linux.
> 
> I question whether the ac97 support ever worked. It is a file that was
> taken from VirtualBox and added to QEMU. I do know that VirtualBox's
> support for the ac97 sound card works perfectly.
> 
> The exact line of code that stops QEMU in its tracks is located in the
> file main-loop.c, in the function os_host_main_loop_wait(), the call
> made to qemu_mutex_lock_iothread(). The is where QEMU stops under Mac OS
> X.
> 
> ** Affects: qemu
>      Importance: Undecided
>          Status: New
> 

Maybe this is just a regression: I'm using ac97 for a win7 guest for a
while, and since recently (don't recall when precisely, some weeks
maybe) I'm getting "main-loop: WARNING: I/O thread spun for 1000
iterations", a temporarily stuck guest and broken sound output. This
used to work fine. Someone has to bisect, I didn't find the time yet.

Jan



Turns out this problem was with another program called AudioJack. It interferes with QEMU and prevents sound from working. Uninstalling AudioJack made the AC97 sound card work again under Mac OS 10.6.