On Mon, Jun 03, 2013 at 02:00:00PM +0200, Daniel Vetter wrote:
> On Fri, May 31, 2013 at 08:07:03PM +0300, ville.syrjala at linux.intel.com wrote:
> > From: Ville Syrj?l? <ville.syrjala at linux.intel.com>
> >
> > plane->enabled is never set, so this code didn't do anything.
> >
> > Fix the code for sprites by calling the new drm_plane_force_disable()
> > function. That means the plane will remain off until someone explicitly
> > turns it back on.
> >
> > And do the same for cursors and the old video overlays, since we only
> > want to see the primary plane for fbdev.
> >
> > v2: Disable sprites/cursors until explicitly re-enabled
> >
> > Signed-off-by: Ville Syrj?l? <ville.syrjala at linux.intel.com>
> > ---
> > drivers/gpu/drm/i915/intel_display.c | 16 ++++++++++++++++
> > drivers/gpu/drm/i915/intel_drv.h | 1 +
> > drivers/gpu/drm/i915/intel_fb.c | 9 ++-------
> > 3 files changed, 19 insertions(+), 7 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
> > index 944b6d5..1d1a3fd 100644
> > --- a/drivers/gpu/drm/i915/intel_display.c
> > +++ b/drivers/gpu/drm/i915/intel_display.c
> > @@ -9842,3 +9842,19 @@ intel_display_print_error_state(struct drm_i915_error_state_buf *m,
> > }
> > }
> > #endif
> > +
> > +void intel_disable_cursors_and_sprites(struct drm_device *dev)
> > +{
> > + struct drm_crtc *crtc;
> > + struct drm_plane *plane;
> > +
> > + list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
> > + intel_crtc_dpms_overlay(to_intel_crtc(crtc), false);
> > + intel_crtc_cursor_set(crtc, NULL, 0, 0, 0);
> > + }
>
> I still think we should just move this into drm_fb_helper_restore_fbdev_mode:
> - Cursors can be killed with crtc->funcs->cursor_set(crtc, NULL, 0, 0, 0);
> Or at least we should be able to do so, I've done a quick audit of all
> drivers and none really cares about file_priv if the handle is 0. Of
> course we should put that into a drm_cursor_force_disable helper to
> prepare for a better world (atomic modeset and all ...).
>
> vmwgfx needs to be fixed up slightly since it derefences file_priv
> outside of a handle != 0 check, but that one can be trivially fixed for
> both cursor_set callbacks by moving the driver_private dereference into
> the if block.
>
> - Legacy overlay stuff has been broken like that forever, and at least on
> i8xx those are used much more often. Imo whoever's offended by gen2
> being a bit broken can write a drm plane wrapper for the overlay stuff
> (or better, move the code to drm planes and convert the ioctl to be a
> shim around the real drm plane interfaces). I don't think we should add
> hacks for that.
OK I give up. I did a quick check of the private plane stuff, and it
turns out they're not on plane_list, so I think just walking the list
and disabling everything there should be OK. Since you're happy w/
leaving the video overlay out in the cold for now, I'll shovel it all
into drm_fb_helper...
--
Ville Syrj?l?
Intel OTC
