On Wed, May 02, 2012 at 11:36:15PM +0200, Daniel Vetter wrote: > On Wed, May 02, 2012 at 10:22:33PM +0100, Chris Wilson wrote: > > On Wed, 2 May 2012 23:12:36 +0200, Daniel Vetter <daniel at ffwll.ch> wrote: > > > For consistency I guess we can ditch the dev parameter (and even then, the > > > ring would uniquely identify the device). Also, I guess you need to > > > explicitly pass in blocking, because mutex_is_locked is rather racy - > > > someone else could hold the mutex while we're waiting in a non-blocking > > > fashion. > > > > Meh, I suggested the race - I'd rather have a moment of confusion > > reading the trace than reading the code in 6 months time. > > Ok, I see the problem and agree, let's have it slightly racy ... I've forgotten to add: Ben, please add the reasons why we decided to go with the racy variant to the commit messages, that kind of stuff is really important (otherwise the patch simply looks buggy). -Daniel -- Daniel Vetter Mail: daniel at ffwll.ch Mobile: +41 (0)79 365 57 48
