Re: Missing notifications for front buffer status change(invalidate)

Re: Missing notifications for front buffer status change(invalidate)

[Ramalingam C]

cc += intel-gfx

On Wednesday 18 March 2015 02:12 PM, Ramalingam C wrote:
> Hi,
>
> Is there anyway I can test the front buffer tracking?
>
> On latest drm-intel-nightly, I am observing that when ubuntu bootsup 
> after few secs (like ~15Secs)
> there no call from intel_fb_obj_invalidate() to invalidate the drrs.
>
> But system is healthy and FBs are keep changing. But no notification 
> to DRRS for FB status change.
>
> I have verified this by observing the busy_frontbuffer_bits from 
> /sys/kernel/debug/dri/0/i915_drrs_status. This bit is not changing at 
> all.
> Hence DRRS is stuck at DRRS_LOW_RR itself.
>
> Is it known valid issue? Or am I doing something wrong.!?
>

-- 
Ram

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: Missing notifications for front buffer status change(invalidate)

[Daniel Vetter]

On Wed, Mar 18, 2015 at 02:23:01PM +0530, Ramalingam C wrote:
> cc += intel-gfx
> 
> On Wednesday 18 March 2015 02:12 PM, Ramalingam C wrote:
> >Hi,
> >
> >Is there anyway I can test the front buffer tracking?
> >
> >On latest drm-intel-nightly, I am observing that when ubuntu bootsup after
> >few secs (like ~15Secs)
> >there no call from intel_fb_obj_invalidate() to invalidate the drrs.
> >
> >But system is healthy and FBs are keep changing. But no notification to
> >DRRS for FB status change.
> >
> >I have verified this by observing the busy_frontbuffer_bits from
> >/sys/kernel/debug/dri/0/i915_drrs_status. This bit is not changing at all.
> >Hence DRRS is stuck at DRRS_LOW_RR itself.
> >
> >Is it known valid issue? Or am I doing something wrong.!?

Yeah that shouldn't happen indeed. Unfortunately we've created the
frontbuffer code for fbc/psr, and that code is still not yet enabled by
default. So testing for frontbuffer tracking is probably too thin. And we
still discover corner-cases in the frontbuffer tracking logic.

So "just" tracking down what's going and fix it. Plus perhaps we do indeed
need a DRRS testcase to exercise these corner-cases and check that the
kernel does correctly switch between low and high modes.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx