Intel i915 GPU hang after kernel update to 5.8 (Pop OS 20.04)

[gerazo]

I am using Pop!_OS 20.04 LTS on a Haswell Intel(R) Core(TM) i7-4710HQ CPU.

After updating to kernel 5.8 a couple of days ago, the system constantly crashes on any heavier GPU load. As soon as a race starts in SuperTuxKart, the following happens 100% of the time:

[  288.066977] i915 0000:00:02.0: [drm] GPU HANG: ecode 7:1:85ddfffd, in Xorg [1571]
[  288.066978] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.
[  288.066979] Please file a _new_ bug report at https://gitlab.freedesktop.org/drm/intel/issues/new.
[  288.066979] Please see https://gitlab.freedesktop.org/drm/intel/-/wikis/How-to-file-i915-bugs for details.
[  288.066979] drm/i915 developers can then reassign to the right component if it's not a kernel issue.
[  288.066980] The GPU crash dump is required to analyze GPU hangs, so please always attach it.
[  288.066980] GPU crash dump saved to /sys/class/drm/card0/error
[  288.067604] i915 0000:00:02.0: [drm] Resetting chip for stopped heartbeat on rcs0
[  288.169353] i915 0000:00:02.0: [drm] Xorg[1571] context reset due to GPU hang
[  291.138874] i915 0000:00:02.0: [drm] GPU HANG: ecode 7:1:85ddfffd, in Xorg [1571]
[  291.138955] i915 0000:00:02.0: [drm] Resetting chip for stopped heartbeat on rcs0
[  291.242212] i915 0000:00:02.0: [drm] Xorg[1571] context reset due to GPU hang
[  297.030397] i915 0000:00:02.0: [drm] GPU HANG: ecode 7:1:85ddfffd, in Xorg [1571]
[  297.030448] i915 0000:00:02.0: [drm] Resetting chip for stopped heartbeat on rcs0
[  297.131629] i915 0000:00:02.0: [drm] Xorg[1571] context reset due to GPU hang

After this, the GPU is reset, all apps are closed and the login screen is started again. This crash can also happen randomly anywhere, but very rarely. However, full load always triggers it.

This problem is exactly the same as shown in this bug report:
https://gitlab.freedesktop.org/drm/intel/-/issues/2024
It seems that this problem was introduced by the newer kernel in Pop. There is a proposed fix (& 0) in the thread. It seems that more and more people will be affected as kernels of 5.7 and above are rolling out in various distributions.

[thomas-zimmerman]

In upstream 5.10, I just saw this get posed: I wonder if this is something that we can pull back to our 5.8 kernel.

---

commit be33805
Author: Chris Wilson chris@chris-wilson.co.uk
Date: Thu Oct 15 13:21:38 2020 +0100

drm/i915/gt: Fixup tgl mocs for PTE tracking

Forcing mocs:1 [used for our winsys follows-pte mode] to be cached
caused display glitches. Though it is documented as deprecated (and so
likely behaves as uncached) use the follow-pte bit and force it out of
L3 cache.

Testcase: igt/kms_frontbuffer_tracking
Testcase: igt/kms_big_fb
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ayaz A Siddiqui <ayaz.siddiqui@intel.com>
Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20201015122138.30161-4-chris@chris-wilson.co.uk
(cherry picked from commit a04ac827366594c7244f60e9be79fcb404af69f0)
Fixes: 849c0fe9e831 ("drm/i915/gt: Initialize reserved and unspecified MOCS indices")
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
[Rodrigo: Updated Fixes tag]

[bflanagin]

@gerazo: Now that we've moved on to Kernel 5.11 are you still seeing this issue?

[gerazo]

@bflanagin I've freshly updated today to the new kernel. The issue is gone. Switching works fine. The very same scenarios which used to trigger the problem now work perfectly with the integrated GPU. Thanks a lot.