On Thu, Dec 04, 2014 at 09:00:35AM +0000, Andrew Jackson wrote:
On 12/03/14 17:25, Mark Brown wrote:
On Wed, Dec 03, 2014 at 04:38:55PM +0000, Andrew Jackson wrote:
case SNDRV_PCM_TRIGGER_STOP: case SNDRV_PCM_TRIGGER_SUSPEND: case SNDRV_PCM_TRIGGER_PAUSE_PUSH:
dev->active--;
if (dev->active > 0)dev->active--;
How is this triggering - this sounds like you're papering over some other bug somewhere?
When I looked at the code paths I couldn't convince myself that STOP wouldn't be called more than once. Then actuve would be negative and the device might not be restartable. I didn't have a problem per se, it was just that it seemed to be something of a loophole.
If you're just adding the check on the off chance that it might fire you need to add a warning message as well - what your change does is make the code look like it's supposed to have broken reference counting since it has a check to silently fix up and ignore problems.