ACK: [PATCH 0/1][SRU][E] Fix i915 power use regression in 5.3

[Alex Hung]

On 2020-02-24 4:47 p.m., Colin Ian King wrote:
> On 24/02/2020 21:06, Seth Forshee wrote:
>> SRU Justification
>>
>> BugLink: https://bugs.launchpad.net/bugs/1853044
>>
>> Impact: "drm/i915/gen8+: Add RC6 CTX corruption WA" makes it effectively
>> impossible to enter RC6 for some Intel GPUs when a display server is running.
>> This results in increased energy usage and temperature.
>>
>> Fix: Upstream has changed too much to make the fixes there applicable, but
>> Chris Wilson has provided a patch for 5.3, here:
>> https://gitlab.freedesktop.org/drm/intel/issues/614#note_366057
>>
>> Test Case: See test results on the bug. On an affected machine powertop will
>> show increased RC6 usage on an idle desktop after the patch, and other symptoms
>> of excessive power use should also subside.
>>
>> Regression Potential: The 5.3 patch is pretty straightforward, and only adds
>> running of an existing delayed worker to retire GPU requests. No regressions
>> have been reported in testing so far.
>>
>> Thanks,
>> Seth
>>
>> ---
>>
>> Chris Wilson (1):
>>   drm/i915: Schedue request retirement in execlists_user_end()
>>
>>  drivers/gpu/drm/i915/gt/intel_lrc.c | 4 ++++
>>  1 file changed, 4 insertions(+)
>>
> 
> Excellent! I'm glad a suitable fix was found for this issue. Thanks Seth.
> 
> This looks like the regression potential is really very low, so..
> 
> Acked-by: Colin Ian King <colin.king at canonical.com>
> 

I tested this patch, and RC6 can be 99% (was 0%).

Acked-by: Alex Hung <alex.hung at canonical.com>