On Thu, Nov 13, 2014 at 05:23:05PM -0800, Matt Roper wrote: > On Thu, Nov 13, 2014 at 02:51:32PM -0800, Matt Roper wrote: > > Add the new driver entrypoints that will be called by the atomic plane > > helpers. > > > > This patch does not actually switch over to the new plane helpers yet, > > so there should be no functional change here. Also note that although > > plane programming was already split into check/prepare/commit steps, > > some of the semantics of those individual functions will need to change > > slightly when we do make the jump so that they match the behavior the > > plane helpers expect. > > > > v2: > > - Renamed file from intel_atomic.c to intel_atomic_plane.c (Daniel) > > - Fix a copy/paste comment mistake (Bob) > > > > Signed-off-by: Matt Roper <matthew.d.ro...@intel.com> > > Actually, I don't think this is quite ready to go yet. I must have had > too much debug spam turned on before and missed the important messages > in the kernel log, but it looks like we're still calling something that > can sleep (in the frontbuffer tracking code?) while irqs are disabled > for vblank evasion.
Yeah the frontbuffer tracking code will sleep since it can grab mutexes. We also want to eventually grow async support out of this, so I think the best approach might be to simply use frontbuffer_flip_prepare/complete around the entire sequence. Well the important part is that we call prepare before we start waiting for outstanding rendering and complete after pipe_update_end. Since this loops over abstract plane objects I think we need to store the correct frontbuffer tracking bit somewhere in intel_plane so that we can get at them. -Daniel -- Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch _______________________________________________ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
