1 Oct
2009
1 Oct
'09
2:08 p.m.
On Thu, Oct 01, 2009 at 02:01:09PM +0300, Eero Nurkkala wrote:
I still think the described scenario can happen. Or could you point where the mutex is taken earlier? If it was, I would've deadlocked every time....Maybe I'm missing some info.
Any of the control paths down from user space should be taking it.
BTW, what's the reasoning for codec mutex anyway?
It protects all the data on the card - we've got a lot of read/modify/write cycles going on, plus things like the power state transitions which need to be run single threaded otherwise they'll get therribly confused. We could do something finer grained but it's never been an issue.