On Wed, Aug 13, 2014 at 01:39:22PM +0200, Daniel Vetter wrote: > On Wed, Aug 13, 2014 at 12:32:33PM +0100, Chris Wilson wrote: > > On Wed, Aug 13, 2014 at 01:26:40PM +0200, Daniel Vetter wrote: > > > This one here must be synchronous. > > > > Right, but notice that we synchronize the work afterwards. I thought if > > we were careful enough to call the waiter that waited for additional > > work items to be completed, it would be sufficient. > > Given that the work creates new work when it can't do it's job I don't > think that flush is sufficient. Iirc that only flush out pending stuff, > not newly queued work. I had a fancy while(flush_work()) ; all planned out. But that is effectively a busy-wait during suspend and the console_lock() is contended, so that got binned. :( -Chris -- Chris Wilson, Intel Open Source Technology Centre _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/intel-gfx
