On Mon, Jun 22, 2015 at 12:08:36PM +0200, Maarten Lankhorst wrote: > >>> Why are we still commiting planes on a disabled pipes? > >>> > >>> The way I envisioned the modeset path looking is something like this: > >>> > >>> 1. atomically disable all planes for pipes getting disabled > >>> 2. disable all pipes that need disabling > >>> 3. enable all pipes that need enabling > >>> 4. atomically update all planes that require it on now active pipes > >>> > >>> > >> I looked for other ways to do this, but they were ugly. Only updating sw > >> state in commit_planes for disabled crtc's made things a lot cleaner > >> without worrying about special cases. > > The commit hooks shouldn't update any sw state.
I looked at that a bit and the plane disabling (i.e. step 2 above) doesn't touch state objects. plane->fb is still a bit a sorry state since apparently page_flip won't update this properly. There's even a TODO about this in drm_atomic_helper_page_flip, I guess I should finally address this one asap. -Daniel -- Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
