Can you please test what happens if you don't enable the audio driver?

(In reply to comment #1)
> Can you please test what happens if you don't enable the audio driver?

Daniel, How can I disable the audio driver? Thanks.

Build a kernel with CONFIG_SOUND=n Of course you need to do that only temporarily and not for the default builds ;-)

(In reply to comment #1)
> Can you please test what happens if you don't enable the audio driver?

Disable audio driver, It still happens.

We get those interrupts because we're reading registers that are on the disabled power well when we are at the function that collects data for the error state file.

I sent a patch fixing this a long ago, but Daniel Denied it.

If we fix the GPU hang, the problem will go away.

(In reply to comment #5)
> We get those interrupts because we're reading registers that are on the
> disabled power well when we are at the function that collects data for the
> error state file.
> 
> I sent a patch fixing this a long ago, but Daniel Denied it.
> 
> If we fix the GPU hang, the problem will go away.

Hm, which register access where?

(In reply to comment #6)
> (In reply to comment #5)
> > We get those interrupts because we're reading registers that are on the
> > disabled power well when we are at the function that collects data for the
> > error state file.
> > 
> > I sent a patch fixing this a long ago, but Daniel Denied it.
> > 
> > If we fix the GPU hang, the problem will go away.
> 
> Hm, which register access where?

All the I915_READ(something) calls at intel_display_capture_error_state().

Changing the bug title because even if we fix the "unclaimed register" error message, the GPU hang will still bite us.

Also, I already sent a patch for the "unclaimed register" error a long time ago and it got denied, so I'm not sure what to do :)

(In reply to comment #0)
> Created attachment 86433 [details]
> dmesg
> 
> System Environment:
> --------------------------
> Platform:  Haswell ULT
> Kernel: (drm-intel-next-queued)1b068ee25776533074251f1c6276c5f720c0284b
> 
> Bug detailed description:
> -----------------------------
> It causes <3>[  326.885041] [drm:intel_uncore_check_errors] *ERROR*
> Unclaimed register before interrupt. It happens on haswell ULT with
> -queued,-fixes and -nightly kernel. I can't find out a good commit.
> 
> output:
> Using monotonic timestamps
> Beginning flip-vs-panning-vs-hang-interruptible on crtc 3, connector 10
>   1920x1080 60 1920 1966 1996 2080 1080 1082 1086 1112 0xa 0x48 138780
> ...
> flip-vs-panning-vs-hang-interruptible on crtc 3, connector 10: PASSED
> 
> Beginning flip-vs-panning-vs-hang-interruptible on crtc 5, connector 10
>   1920x1080 60 1920 1966 1996 2080 1080 1082 1086 1112 0xa 0x48 138780
> ...
> flip-vs-panning-vs-hang-interruptible on crtc 5, connector 10: PASSED
> 
> Beginning flip-vs-panning-vs-hang-interruptible on crtc 7, connector 10
>   1920x1080 60 1920 1966 1996 2080 1080 1082 1086 1112 0xa 0x48 138780
> ...
> flip-vs-panning-vs-hang-interruptible on crtc 7, connector 10: PASSED
> 
> Subtest flip-vs-panning-vs-hang-interruptible: SUCCESS
> 
> Reproduce steps:
> ----------------------------
> 1. ./kms_flip --run-subtest flip-vs-panning-vs-hang-interruptible

Can you please reproduce the bug again and attach the error state?

Whenever you see a dmesg with "[  158.798928] [drm] capturing error event; look for more information in /sys/class/drm/card0/error", please attach the error state.

Thanks,
Paulo

(In reply to comment #9)
> (In reply to comment #0)
> > Created attachment 86433 [details]
> > dmesg
> > 
> > System Environment:
> > --------------------------
> > Platform:  Haswell ULT
> > Kernel: (drm-intel-next-queued)1b068ee25776533074251f1c6276c5f720c0284b
> > 
> > Bug detailed description:
> > -----------------------------
> > It causes <3>[  326.885041] [drm:intel_uncore_check_errors] *ERROR*
> > Unclaimed register before interrupt. It happens on haswell ULT with
> > -queued,-fixes and -nightly kernel. I can't find out a good commit.
> > 
> > output:
> > Using monotonic timestamps
> > Beginning flip-vs-panning-vs-hang-interruptible on crtc 3, connector 10
> >   1920x1080 60 1920 1966 1996 2080 1080 1082 1086 1112 0xa 0x48 138780
> > ...
> > flip-vs-panning-vs-hang-interruptible on crtc 3, connector 10: PASSED
> > 
> > Beginning flip-vs-panning-vs-hang-interruptible on crtc 5, connector 10
> >   1920x1080 60 1920 1966 1996 2080 1080 1082 1086 1112 0xa 0x48 138780
> > ...
> > flip-vs-panning-vs-hang-interruptible on crtc 5, connector 10: PASSED
> > 
> > Beginning flip-vs-panning-vs-hang-interruptible on crtc 7, connector 10
> >   1920x1080 60 1920 1966 1996 2080 1080 1082 1086 1112 0xa 0x48 138780
> > ...
> > flip-vs-panning-vs-hang-interruptible on crtc 7, connector 10: PASSED
> > 
> > Subtest flip-vs-panning-vs-hang-interruptible: SUCCESS
> > 
> > Reproduce steps:
> > ----------------------------
> > 1. ./kms_flip --run-subtest flip-vs-panning-vs-hang-interruptible
> 
> Can you please reproduce the bug again and attach the error state?
> 
> Whenever you see a dmesg with "[  158.798928] [drm] capturing error event;
> look for more information in /sys/class/drm/card0/error", please attach the
> error state.

Ignore that. Daniel just explained this test is supposed to hang the gpu... My bad.

Created attachment 88483 [details] [review]
Fix

Hi

Can you please confirm this patch fixes the bug? It works for me.

Thanks,
Paulo

commit 9d1cb9147dbe45f6e94dc796518ecf67cb64b359
Author: Paulo Zanoni <paulo.r.zanoni@intel.com>
Date:   Fri Nov 1 13:32:08 2013 -0200

    drm/i915: avoid unclaimed registers when capturing the error state

Verified.Fixed.

Closing old verified.