GPU hang: i915, Braswell, errcodes 0x85dffff9 and 85dffffb
Description:
- Usual scenario: Screen freezes, randomly or e.g. when starting an application or opening a menu, then after several seconds recovers. When there's an application playing music in the background, the playback is not interrupted by this. Happens 0-5 times a day. Happens on all kernels tested.
- Other scenario: Screen freezes as above, but everything stops irreversibly. Happens 0-2 times a day. Happened on Arch's kernel; didn't reproduce on drm-tip kernel yet (short uptime so far).
- Artifacts: Twice or thrice it happened while a new screen-element was being created and then that element (terminal window, desktop menu) was partially drawn on the screen.
Tried and reproduced under kernels:
- stock Arch Linux x86_64 kernel '5.5.10-arch1-1' and '5.5.11-arch1-1'
- compiled DRM-TIP kernel from https://aur.archlinux.org/packages/linux-drm-tip-git/ (modified config) (git src v. 5.7.903997.c7fa35c8bab0-1)
When it started:
- after an update that included the kernel '5.5.10-arch1-1', but i can't trace it exactly, because i haven't been updating very diligently.
journalctl -k
reports:
under kernel '5.5.10-arch1-1' without drm.debug kernel params:
...
i915 0000:00:02.0: GPU HANG: ecode 8:1:0x85dffff9, in systemd-logind [418], stopped heartbeat on rcs0
GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.
Please file a _new_ bug report at https://gitlab.freedesktop.org/drm/intel/issues/new.
Please see https://gitlab.freedesktop.org/drm/intel/-/wikis/How-to-file-i915-bugs for details.
drm/i915 developers can then reassign to the right component if it's not a kernel issue.
The GPU crash dump is required to analyze GPU hangs, so please always attach it.
GPU crash dump saved to /sys/class/drm/card0/error
Resetting rcs0 for stopped heartbeat on rcs0
...
under kernel '5.7.903997.c7fa35c8bab0-1' with drm.debug kernel params:
...
i915 0000:00:02.0: [drm:intel_plane_atomic_calc_changes [i915]] [CRTC:51:pipe A] with [PLANE:47:cursor A] visible 1 -> 1, off 0, on 0, ms 0
[drm:drm_atomic_set_fb_for_plane [drm]] Set [FB:121] for [PLANE:47:cursor A] state 000000000b126c0c
i915 0000:00:02.0: [drm:intel_plane_atomic_calc_changes [i915]] [CRTC:51:pipe A] with [PLANE:47:cursor A] visible 1 -> 1, off 0, on 0, ms 0
i915 0000:00:02.0: GPU HANG: ecode 8:1:85dffffb, in systemd-logind [372]
GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.
Please file a _new_ bug report at https://gitlab.freedesktop.org/drm/intel/issues/new.
Please see https://gitlab.freedesktop.org/drm/intel/-/wikis/How-to-file-i915-bugs for details.
drm/i915 developers can then reassign to the right component if it's not a kernel issue.
The GPU crash dump is required to analyze GPU hangs, so please always attach it.
GPU crash dump saved to /sys/class/drm/card0/error
i915 0000:00:02.0: Resetting rcs0 for stopped heartbeat on rcs0
i915 0000:00:02.0: systemd-logind[372] context reset due to GPU hang
[drm:drm_atomic_set_fb_for_plane [drm]] Set [FB:121] for [PLANE:47:cursor A] state 00000000cd5adeb7
i915 0000:00:02.0: [drm:intel_plane_atomic_calc_changes [i915]] [CRTC:51:pipe A] with [PLANE:47:cursor A] visible 1 -> 1, off 0, on 0, ms 0
[drm:drm_atomic_set_fb_for_plane [drm]] Set [FB:121] for [PLANE:47:cursor A] state 000000008dec2456
i915 0000:00:02.0: [drm:intel_plane_atomic_calc_changes [i915]] [CRTC:51:pipe A] with [PLANE:47:cursor A] visible 1 -> 1, off 0, on 0, ms 0
[drm:drm_atomic_set_fb_for_plane [drm]] Set [FB:121] for [PLANE:47:cursor A] state 00000000cd5adeb7
i915 0000:00:02.0: [drm:intel_plane_atomic_calc_changes [i915]] [CRTC:51:pipe A] with [PLANE:47:cursor A] visible 1 -> 1, off 0, on 0, ms 0
[drm:drm_atomic_set_fb_for_plane [drm]] Set [FB:121] for [PLANE:47:cursor A] state 000000008dec2456
i915 0000:00:02.0: [drm:intel_plane_atomic_calc_changes [i915]] [CRTC:51:pipe A] with [PLANE:47:cursor A] visible 1 -> 1, off 0, on 0, ms 0
[drm:drm_atomic_set_fb_for_plane [drm]] Set [FB:121] for [PLANE:47:cursor A] state 00000000cd5adeb7
i915 0000:00:02.0: [drm:intel_plane_atomic_calc_changes [i915]] [CRTC:51:pipe A] with [PLANE:47:cursor A] visible 1 -> 1, off 0, on 0, ms 0
...
Attached:
-
dumps.tar.gz
- dumps/dmesg-pruned
- dumps/dmidecode
- dumps/error
- dumps/error-2
System info:
- system architecture: x86_64
- kernel version: (arch) '5.5.10-arch1-1'; (drm-tip) '5.7.903997.c7fa35c8bab0-1' or so
- Linux distribution: Arch
- Machine or mother board model: CPU Intel Celeron N3150 (Braswell); more in dmidecode
- Display connector: VGA
- GPU crash dump: in error (under arch), error-2 (under drm-tip)
Note on kern. version:
There is a mix-up with the drm-tip kernel, where the 'config' comes from an AUR-distributed package, and has a different version from the kernel sources fetched from https://anongit.freedesktop.org/git/drm-tip.git#branch=drm-tip. Here 'uname -r' reports '5.6.0-rc7-1-drm-tip-gitcstm-gc7fa35c8bab0' and the git-version is marked as '5.7.903997.c7fa35c8bab0-1'.