Bug 108698

Summary: [CI][BAT] igt@drv_selftest@live_contexts - dmesg-fail - live_nop_switch failed with error -5
Product: DRI Reporter: Martin Peres <martin.peres>
Component: DRM/IntelAssignee: Andi <andi.shyti>
Status: RESOLVED WORKSFORME QA Contact: Intel GFX Bugs mailing list <intel-gfx-bugs>
Severity: normal    
Priority: high CC: intel-gfx-bugs
Version: XOrg git   
Hardware: Other   
OS: All   
Whiteboard: ReadyForDev
i915 platform: CFL i915 features: GEM/Other

Description Martin Peres 2018-11-09 09:39:42 UTC
https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_5106/fi-cfl-8109u/igt@drv_selftest@live_contexts.html

(drv_selftest:4567) igt_kmod-WARNING: live_nop_switch failed with error -5
Comment 1 Chris Wilson 2018-11-09 10:46:19 UTC
21:22            ickle : <3> [385.741372] Switching between 241 contexts timed out
21:22            ickle : ...
21:22            ickle : <7> [385.848580] i915_gem_set_wedged \x09Execlist CSB read 2, write 5 [mmio:5], tasklet queued? yes 
                         (enabled)
21:22            ickle : ksoftirqd strikes again
21:23            ickle : no, that maybe a little simplistic
21:23            ickle : it executed 1a06 and 1a07 whilst dumping the state
21:24            ickle : so maybe it just accumulated a lot of slowness
Comment 2 Francesco Balestrieri 2018-11-12 09:29:55 UTC
Do we need more occurrences of this to investigate? So far there has been one.
Comment 3 Chris Wilson 2018-11-12 10:53:51 UTC
This one has the hallmark of being an uncommon sw issue. I'm still betting on ksoftirqd being at the heart of it, proof left to the imagination.
Comment 4 Francesco Balestrieri 2018-11-14 15:12:46 UTC
Still no other occurrences. I'm going setting it to high priority.
Comment 5 Chris Wilson 2018-11-29 12:48:31 UTC
https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_10950/fi-cfl-8109u/igt@i915_selftest@live_contexts.html

Same again, running very slowly but progress is still being made. Still taking over 2s for a 100 context switches is bogus.
Comment 6 Francesco Balestrieri 2019-01-09 07:53:14 UTC
Occurred several times in a day, then never again for almost two months. Closing.

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.