igt@i915_pm_rc6_residency@rc6-idle - fail - 'idle - prev' != 'ts[1] - ts[0]' / '(rc6)' != '(ts[1] - ts[0])'

https://intel-gfx-ci.01.org/tree/drm-tip/IGT_5363/shard-apl3/igt@i915_pm_rc6_residency@rc6-idle.html

Starting subtest: rc6-idle
(i915_pm_rc6_residency:1260) CRITICAL: Test assertion failure function rc6_idle, file ../tests/i915/i915_pm_rc6_residency.c:336:
(i915_pm_rc6_residency:1260) CRITICAL: Failed assertion: (double)(idle - prev) <= (1.0 + (5)) * (double)(ts[1] - ts[0]) && (double)(idle - prev) >= (1.0 - (5)) * (double)(ts[1] - ts[0])
(i915_pm_rc6_residency:1260) CRITICAL: 'idle - prev' != 'ts[1] - ts[0]' (131856333252.000000 not within +500.0%/-500.0% tolerance of 3000179278.000000)
Subtest rc6-idle failed.

added CI platform: ALL labels

The funny part here is the sample reports 130s of rc6 during a 3s period. It's attributing all the rc6 since boot to the first report.

commit f4e9894b6952a2819937f363cd42e7cd7894a1e4
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Tue Jan 14 10:56:47 2020 +0000

    drm/i915/pmu: Correct the rc6 offset upon enabling
    
    The rc6 residency starts ticking from 0 from BIOS POST, but the kernel
    starts measuring the time from its boot. If we start measuruing
    I915_PMU_RC6_RESIDENCY while the GT is idle, we start our sampling from
    0 and then upon first activity (park/unpark) add in all the rc6
    residency since boot. After the first park with the sampler engaged, the
    sleep/active counters are aligned.
    
    v2: With a wakeref to be sure
    
    Closes: https://gitlab.freedesktop.org/drm/intel/issues/973
    Fixes: df6a42053513 ("drm/i915/pmu: Ensure monotonic rc6")
    Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
    Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
    Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
    Link: https://patchwork.freedesktop.org/patch/msgid/20200114105648.2172026-1-chris@chris-wilson.co.uk

After we fix the sampling, I expect there to be a couple of HW bugs to investigate. On byt/bsw related to the PCU, and on gen6/gen7 related to coarse evaluation intervals for HW rc6.

closed

A CI Bug Log filter associated to this bug has been updated:

igt@i915_pm_rc6_residency@rc6-idle - fail - 'idle - prev' != 'ts[1] - ts[0]' / '(rc6)' != '(ts[1] - ts[0])' All machines: igt@i915_pm_rc6_residency@rc6-idle - fail - Failed assertion: (rc6) <= (ts[1] - ts[0]) * (1.0 + (5)/100.) && (rc6) >= (ts[1] - ts[0]) * (1.0 - (5)/100.)

New failures caught by the filter:

• https://intel-gfx-ci.01.org/tree/drm-tip/IGT_5365/shard-apl2/igt@i915_pm_rc6_residency@rc6-idle.html
• https://intel-gfx-ci.01.org/tree/drm-tip/IGT_5365/shard-glk5/igt@i915_pm_rc6_residency@rc6-idle.html
• https://intel-gfx-ci.01.org/tree/drm-tip/IGT_5365/shard-iclb4/igt@i915_pm_rc6_residency@rc6-idle.html
• https://intel-gfx-ci.01.org/tree/drm-tip/IGT_5365/shard-kbl4/igt@i915_pm_rc6_residency@rc6-idle.html
• https://intel-gfx-ci.01.org/tree/drm-tip/IGT_5365/shard-skl5/igt@i915_pm_rc6_residency@rc6-idle.html
• https://intel-gfx-ci.01.org/tree/drm-tip/IGT_5365/shard-snb4/igt@i915_pm_rc6_residency@rc6-idle.html
• https://intel-gfx-ci.01.org/tree/drm-tip/IGT_5365/shard-tglb3/igt@i915_pm_rc6_residency@rc6-idle.html
• https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_7738/shard-apl4/igt@i915_pm_rc6_residency@rc6-idle.html
• https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_7738/shard-glk9/igt@i915_pm_rc6_residency@rc6-idle.html
• https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_7738/shard-iclb2/igt@i915_pm_rc6_residency@rc6-idle.html
• https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_7738/shard-kbl6/igt@i915_pm_rc6_residency@rc6-idle.html
• https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_7738/shard-skl10/igt@i915_pm_rc6_residency@rc6-idle.html
• https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_7738/shard-snb1/igt@i915_pm_rc6_residency@rc6-idle.html
• https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_7739/shard-snb5/igt@i915_pm_rc6_residency@rc6-idle.html
• https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_7740/shard-snb4/igt@i915_pm_rc6_residency@rc6-idle.html
• https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_7742/shard-snb5/igt@i915_pm_rc6_residency@rc6-idle.html
• https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_7743/shard-snb4/igt@i915_pm_rc6_residency@rc6-idle.html
• https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_7744/shard-snb1/igt@i915_pm_rc6_residency@rc6-idle.html
• https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_7745/shard-hsw2/igt@i915_pm_rc6_residency@rc6-idle.html
• https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_7745/shard-snb1/igt@i915_pm_rc6_residency@rc6-idle.html
• https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_7746/shard-snb6/igt@i915_pm_rc6_residency@rc6-idle.html
• https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_7741/shard-snb1/igt@i915_pm_rc6_residency@rc6-idle.html

Note that snb on CI_DRM_7736+ is a new bug. That's the case where HW rc6 is not as fast as soft-rc6 and so while we do enter rc6, it's significantly worse than expected. If we can't find a way to idle faster, we may just have to make the test more lenient -- some rc6 is better than none!

mentioned in commit lyudess/linux@b8604e83

The CI Bug Log issue associated to this bug has been archived.

New failures matching the above filters will not be associated to this bug anymore.