[kbl] mmio death (forcewake ack lost)
I experience random graphic freezes. The system locks up for about a minute and then falls back to a tty. dmesg shows a GPU hang with ecode 9:0:00000000. Sometimes the system remains locked up forever and a hard shutdown is necessary.
[ 770.017381] Asynchronous wait on fence 0000:00:02.0:gnome-shell[3575]:7d84 timed out (hint:intel_atomic_commit_ready [i915])
...
[ 821.556124] [drm:fw_domain_wait_ack_with_fallback [i915]] media had to use fallback to set ack, 0x0 (passes 11)
[ 821.606176] [drm:fw_domains_get_with_fallback [i915]] *ERROR* media: timed out waiting for forcewake ack request.
[ 821.606627] i915 0000:00:02.0: GPU HANG: ecode 9:0:00000000
[ 821.606634] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.
[ 821.606635] Please file a _new_ bug report at https://gitlab.freedesktop.org/drm/intel/issues/new.
[ 821.606636] Please see https://gitlab.freedesktop.org/drm/intel/-/wikis/How-to-file-i915-bugs for details.
[ 821.606638] drm/i915 developers can then reassign to the right component if it's not a kernel issue.
[ 821.606639] The GPU crash dump is required to analyze GPU hangs, so please always attach it.
[ 821.606640] GPU crash dump saved to /sys/class/drm/card0/error
[ 822.157232] [drm:fw_domain_wait_ack_with_fallback [i915]] render had to use fallback to set ack, 0x0 (passes 11)
Reproducing is difficult but most of the time building a project in CLion and rapidly invoking intellisense will cause a freeze after 10-20 minutes. I've also had frequent hangs with Microsoft Teams calls.
This issue occurs on 5.6 kernels as well as drm-tip.
System information:
- Architecture: x86_64
- Kernel version: 5.6.0-rc6+ (drm-tip: b732c91760678820ae86aacecac964b7f9a7eb97, 2020/03/23)
- Linux distribution: Fedora 32
- Machine: Laptop Dell XPS-15 9560 (i7-7700HQ)
Logs:
- dmesg dmesg.2.txt
- gpu crash dump gpu-dump.2.txt
Edited by Chris Wilson