[Intel-gfx] [PATCH v3 0/2] Fix live busy stats selftest failure

[Umesh Nerlige Ramappa]

On Mon, Nov 21, 2022 at 08:13:15PM +0000, Umesh Nerlige Ramappa wrote:
>Reg: Fi.CI.IGT
>
>https://patchwork.freedesktop.org/series/110557/#rev3 shows the below 
>CI failures:
>
>igt at i915_module_load@reload-with-fault-injection:
>shard-snb: PASS -> INCOMPLETE

Looks more like a stack corruption, but again this is fault injection, 
so I don't see any relation to this series.

>
>igt at i915_selftest@live at evict:
>shard-skl: NOTRUN -> INCOMPLETE

Looks like the test just causes a reset as I don't see the test in the 
dmesg log, but see it in the igt_runner log. I don't see any way this 
accesses the code path in this series. see (2) below.

>
>igt at kms_cursor_crc@cursor-alpha-transparent at pipe-d-edp-1:
>shard-tglb: PASS -> INCOMPLETE

Not sure why this shows up as incomplete since I see the test exiting in 
dmesg.

[329.896238] [IGT] kms_cursor_crc: starting dynamic subtest pipe-D-eDP-1
[332.499005] [IGT] kms_cursor_crc: exiting, ret=0

This doesn't seem to fall in the code path modified in this series as 
well.

>
>These failures are not related to the changes in this series.

Note that the change in this series is localized to 2 areas:

(1) selftest at gt-engines@live_engine_busy_stats
This cannot be causing the above failures since it is test specific 
change.

(2) A refactor of intel_uncore_read64_2x32
This is used only by guc_update_pm_timestamp() and 
i915_reg_read_ioctl(). I don't see these paths coming into play in the 
above tests.

Thanks,
Umesh

>
>Thanks,
>Umesh
>
>On Thu, Nov 10, 2022 at 05:19:11PM +0000, Umesh Nerlige Ramappa wrote:
>>Fix live busy stats selftest failure in BAT. The issue is seen more easily on
>>DG1 runs.
>>
>>v2: (Tvrtko)
>>In addition refactor intel_uncore_read64_2x32 to obtain the forcewake
>>once before reading upper and lower register dwords.
>>
>>v3: (Ashutosh)
>>Update commit msg
>>
>>Signed-off-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa at intel.com>
>>
>>Umesh Nerlige Ramappa (2):
>> i915/uncore: Acquire fw before loop in intel_uncore_read64_2x32
>> drm/i915/selftest: Bump up sample period for busy stats selftest
>>
>>drivers/gpu/drm/i915/gt/selftest_engine_pm.c |  2 +-
>>drivers/gpu/drm/i915/intel_uncore.h          | 44 +++++++++++++-------
>>2 files changed, 31 insertions(+), 15 deletions(-)
>>
>>-- 
>>2.25.1
>>