Bug 102949

dmesg:

9/24 17:23:18 kernel: drm/i915: Resetting chip after gpu hang
9/24 17:23:26 kernel: drm/i915: Resetting chip after gpu hang
9/24 17:23:34 kernel: drm/i915: Resetting chip after gpu hang
9/24 17:25:38 kernel: drm/i915: Resetting chip after gpu hang
9/24 17:25:46 kernel: drm/i915: Resetting chip after gpu hang
9/24 17:25:54 kernel: drm/i915: Resetting chip after gpu hang
9/24 17:26:02 kernel: drm/i915: Resetting chip after gpu hang

more details:

[  +1.195806] [drm:intel_cpu_fifo_underrun_irq_handler [i915]] *ERROR* CPU pipe A FIFO underrun
[  +0.002587] [drm:ironlake_irq_handler [i915]] *ERROR* PCH transcoder A FIFO underrun

(In reply to nanericwang from comment #2)
> more details:
> 
> [  +1.195806] [drm:intel_cpu_fifo_underrun_irq_handler [i915]] *ERROR* CPU
> pipe A FIFO underrun
> [  +0.002587] [drm:ironlake_irq_handler [i915]] *ERROR* PCH transcoder A
> FIFO underrun

This is related to display / monitor problems - shouldn't affect this bug.  The GPU hangs are concerning.

Can you provide the output of:
$ lspci -nn | grep VGA

Thanks!

(In reply to Kenneth Graunke from comment #3)
> (In reply to nanericwang from comment #2)
> > more details:
> > 
> > [  +1.195806] [drm:intel_cpu_fifo_underrun_irq_handler [i915]] *ERROR* CPU
> > pipe A FIFO underrun
> > [  +0.002587] [drm:ironlake_irq_handler [i915]] *ERROR* PCH transcoder A
> > FIFO underrun
> 
> This is related to display / monitor problems - shouldn't affect this bug. 
> The GPU hangs are concerning.
> 
> Can you provide the output of:
> $ lspci -nn | grep VGA
> 
> Thanks!

$ lspci -nn | grep VGA
2:00:02.0 VGA compatible controller [0300]: Intel Corporation Core Processor Integrated Graphics Controller [8086:0046] (rev 02)

video playback was good for first few seconds before frozen. i'm guessing it was because of some resource leaks?

Created attachment 134750 [details]
i915 error output

Hi,

I added the output of /sys/class/drm/card0/error
This happens to me on my Dell XPS 13 9360. I'm using kernel 4.12.14 but it also happens on 4.13.5.

Mesa version is 17.2.2

$ lspci -nn | grep VGA
00:02.0 VGA compatible controller [0300]: Intel Corporation HD Graphics 620 [8086:5916] (rev 02)

Journal output looks like this :
Oct 08 17:13:13 richter-xps-13 kernel: [drm] GPU HANG: ecode 9:0:0x85dffffb, in chromium [5245], reason: Hang on rcs, action: reset
Oct 08 17:13:13 richter-xps-13 kernel: drm/i915: Resetting chip after gpu hang
Oct 08 17:13:13 richter-xps-13 kernel: [drm] RC6 on
Oct 08 17:13:13 richter-xps-13 kernel: [drm] GuC submission enabled (firmware i915/kbl_guc_ver9_14.bin [version 9.14])
Oct 08 17:13:34 richter-xps-13 kernel: [drm:intel_cpu_fifo_underrun_irq_handler [i915]] *ERROR* CPU pipe A FIFO underrun


This seems specific to chromium-vaapi, when playing a video everything freezes while the sound is playing then comes the reset and the video catches up to the sound by going in fast forward (all missed frames are played back at once it seems).

So going back to mesa-17.1.8 makes the problem disappear. No hangs, normal video playback. Looks like a regression.

(In reply to Philipp Richter from comment #7)
> So going back to mesa-17.1.8 makes the problem disappear. No hangs, normal
> video playback. Looks like a regression.

Philipp, your bug is https://bugs.freedesktop.org/show_bug.cgi?id=102774 - this one is about Ironlake.

(In reply to Kenneth Graunke from comment #8)
> (In reply to Philipp Richter from comment #7)
> > So going back to mesa-17.1.8 makes the problem disappear. No hangs, normal
> > video playback. Looks like a regression.
> 
> Philipp, your bug is https://bugs.freedesktop.org/show_bug.cgi?id=102774 -
> this one is about Ironlake.

Oh, sorry for that you're absolutely right. Thank you for the patch now I can use chromium with 17.2.2.

Created attachment 135637 [details]
GPU crash dump saved to /sys/class/drm/card0/error

Here is the GPU crash dump. Hopefully it helps.

-- GitLab Migration Automatic Message --

This bug has been migrated to freedesktop.org's GitLab instance and has been closed from further activity.

You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.freedesktop.org/mesa/mesa/issues/1633.