[Intel-gfx] [PATCH v2 14/14] drm/i915/skl: Do not allow scaling when crtc is disabled.
Maarten Lankhorst
maarten.lankhorst at linux.intel.com
Tue Nov 3 04:00:32 PST 2015
Op 03-11-15 om 11:40 schreef Ville Syrjälä:
> On Tue, Nov 03, 2015 at 11:06:53AM +0100, Maarten Lankhorst wrote:
>> Op 03-11-15 om 10:09 schreef Ville Syrjälä:
>>> On Tue, Nov 03, 2015 at 08:31:53AM +0100, Maarten Lankhorst wrote:
>>>> This fixes a warning when the crtc is turned off. In that case fb
>>>> will be NULL, and crtc_clock will be 0. Because the crtc is no longer
>>>> active this is not a bug, and shouldn't trigger the WARN_ON.
>>> Mm. We want to do scaling checks and whatnot during DPMS. So this should
>>> check .enabled, no?
>> Not sure what the right decision would be here..
>>
>> * skl max scale is lower of:
>> * close to 3 but not 3, -1 is for that purpose
>> * or
>> * cdclk/crtc_clock
>>
>> So when multiple pipes are enabled potentially 3x scaling is allowed, but if you dpms them all off
>> cdclk might get set to 0. This means a previous valid amount of scaling might suddenly become invalid.
>>
>> Maybe the fix is keeping 2 cdclk's, one for scaling and one for setting.
> I think we should keep around the min required cdclk in the crtc state.
> And we recalculate that one every time anything changes. Then we can
> just compare it against the current cdclk after plane/crtc states have
> otherwise been computed to see if we need to change the current cdclk.
>
What would the use be here? In that case the above formula reduces to 1<<16.
If you want to treat dpms off the same as dpms on then you need the separate cdclk's..
More information about the Intel-gfx
mailing list