[Intel-gfx] [PATCH 05/11] drm/i915/skl: Fail the flip if no FB for WM calculation

[Matt Roper]

On Mon, May 08, 2017 at 05:31:30PM +0530, Mahesh Kumar wrote:
> Hi,
> 
> 
> On Monday 08 May 2017 05:18 PM, Lankhorst, Maarten wrote:
> > Mahesh Kumar schreef op ma 08-05-2017 om 17:18 [+0530]:
> > > Fail the flip if no FB is present but plane_state is set as visible.
> > > Above is not a valid combination so instead of continue fail the
> > > flip.
> > Why is this patch necessary? drm_atomic_plane_check handles this.
> Ideally we should never get such combination here. But current WM code
> checks for this situation and even if it's true it proceeds further. This
> patch just corrects the WM code flow decision.
> I also think some of these checks are redundant here.

Yeah, WARN()'s are basically "I'm sure this could never happen" type
assertions.  Of course sometimes we restructure parts of the code and
forget about assumptions we've made elsewhere (or we just plain screw up
and add new bugs to existing code), so we wind up hitting the WARN()'s
anyway.  If proceeding on here could lead to a panic (which I think it
could since we dereference the fb in skl_compute_plane_wm()), then
adding a sensible bail-out here seems okay to me; the extra paranoia
only costs one extra line of code.


Matt

> 
> -Mahesh
> 
> > 
> > ~Maarten
> 

-- 
Matt Roper
Graphics Software Engineer
IoTG Platform Enabling & Development
Intel Corporation
(916) 356-2795