mbox series

[0/3] Fix some races/bugs in GuC engine busyness

Message ID 20241118232223.53639-1-umesh.nerlige.ramappa@intel.com (mailing list archive)
Headers show
Series Fix some races/bugs in GuC engine busyness | expand

Message

Umesh Nerlige Ramappa Nov. 18, 2024, 11:22 p.m. UTC
A few races and bugs in PMU busyness implementation are resulting in a wide
range of IGT failures. This series addresses some failures that are easily
reproduced.

To repro the issues, we run a couple iterations of igt@perf_pmu@busy-hang
followed by igt@perf_pmu@most-busy-idle-check-all test.

Signed-off-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>

Umesh Nerlige Ramappa (3):
  i915/guc: Reset engine utilization buffer before registration
  i915/guc: Ensure busyness counter increases motonically
  i915/guc: Accumulate active runtime on gt reset

 drivers/gpu/drm/i915/gt/intel_engine_types.h  |  5 +++
 .../gpu/drm/i915/gt/uc/intel_guc_submission.c | 41 ++++++++++++++++++-
 2 files changed, 44 insertions(+), 2 deletions(-)

Comments

Umesh Nerlige Ramappa Nov. 19, 2024, 6:46 p.m. UTC | #1
On Tue, Nov 19, 2024 at 12:23:07AM +0000, Patchwork wrote:
>   Patch Details
>
>Series:  Fix some races/bugs in GuC engine busyness
>URL:     [1]https://patchwork.freedesktop.org/series/141522/
>State:   failure
>Details: [2]https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_141522v1/index.html
>
>          CI Bug Log - changes from CI_DRM_15717 -> Patchwork_141522v1
>
>Summary
>
>   FAILURE
>
>   Serious unknown changes coming with Patchwork_141522v1 absolutely need to
>   be
>   verified manually.
>
>   If you think the reported changes have nothing to do with the changes
>   introduced in Patchwork_141522v1, please notify your bug team
>   (I915-ci-infra@lists.freedesktop.org) to allow them
>   to document this new failure mode, which will reduce false positives in
>   CI.
>
>   External URL:
>   https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_141522v1/index.html
>
>Participating hosts (46 -> 45)
>
>   Missing (1): fi-snb-2520m
>
>Possible new issues
>
>   Here are the unknown changes that may have been introduced in
>   Patchwork_141522v1:
>
>  IGT changes
>
>    Possible regressions
>
>     * igt@i915_pm_rpm@module-reload:
>
>          * bat-dg2-11: [3]PASS -> [4]FAIL
>          * bat-adls-6: [5]PASS -> [6]FAIL
>
>     * igt@i915_selftest@live:
>
>          * bat-adlm-1: [7]PASS -> [8]INCOMPLETE
>
>     * igt@i915_selftest@live@workarounds:
>
>          * bat-mtlp-6: [9]PASS -> [10]ABORT
>
>     * igt@kms_pipe_crc_basic@compare-crc-sanitycheck-xr24@pipe-b-dp-2:
>
>          * fi-cfl-8109u: [11]PASS -> [12]DMESG-WARN +2 other tests
>            dmesg-warn

These are unrelated to fixes in this series and passing locally with the 
patches.  Queueing a rerun.

Umesh

>
>Known issues
>