[Intel-gfx] [PATCH] drm/i915/selftests: re-init the GT in live_gt_pm

[Daniele Ceraolo Spurio]

On 11/19/19 4:32 PM, Daniele Ceraolo Spurio wrote:
> 
> 
> On 11/19/19 4:21 PM, Chris Wilson wrote:
>> Quoting Daniele Ceraolo Spurio (2019-11-20 00:04:25)
>>> When GuC is in use we need to make sure it is re-loaded before the call
>>> to gt_resume, otherwise communication from the engines to the GuC will
>>> not be processed, which blocks the engines from ctx switching and from
>>> being reset.
>>>
>>> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=112205
>>> Cc: Andi Shyti <andi.shyti at intel.com>
>>> Cc: Chris Wilson <chris at chris-wilson.co.uk>
>>> Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio at intel.com>
>>> ---
>>> Â  drivers/gpu/drm/i915/gt/selftest_gt_pm.c | 9 +++++++++
>>> Â  1 file changed, 9 insertions(+)
>>>
>>> diff --git a/drivers/gpu/drm/i915/gt/selftest_gt_pm.c 
>>> b/drivers/gpu/drm/i915/gt/selftest_gt_pm.c
>>> index d1752f15702a..0bb17c806dfc 100644
>>> --- a/drivers/gpu/drm/i915/gt/selftest_gt_pm.c
>>> +++ b/drivers/gpu/drm/i915/gt/selftest_gt_pm.c
>>> @@ -11,8 +11,11 @@ static int live_gt_resume(void *arg)
>>> Â  {
>>> Â Â Â Â Â Â Â Â  struct intel_gt *gt = arg;
>>> Â Â Â Â Â Â Â Â  IGT_TIMEOUT(end_time);
>>> +Â Â Â Â Â Â  intel_wakeref_t wakeref;
>>> Â Â Â Â Â Â Â Â  int err;
>>> +Â Â Â Â Â Â  wakeref = intel_runtime_pm_get(gt->uncore->rpm);
>>
>> That defeats the point of gt pm, no?
>>
>>> +
>>> Â Â Â Â Â Â Â Â  /* Do several suspend/resume cycles to check we don't 
>>> explode! */
>>> Â Â Â Â Â Â Â Â  do {
>>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â  intel_gt_suspend_prepare(gt);
>>> @@ -25,6 +28,10 @@ static int live_gt_resume(void *arg)
>>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â  break;
>>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â  }
>>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â  err = intel_gt_init_hw(gt);
>>
>> Hmm. I have that as part of intel_gt_resume.  Which also pulls it into
>> the pm.
> 
> I also considered moving init_hw() inside resume(), but in the end opted 
> not to to keep the fix isolated to the test. But if you have already 
> done the work...
> 
>>
>> I think I prefer my plan/patches :)
> 
> Can you point me to them if they're already on the list?
> 

Hey Chris,

If your solution is still going to take a while, do you mind if we go 
ahead with this fix in the meantime (switching to gt pm), so we can fix 
the bug? More GuC code is going to slowly trickle in in the next few 
weeks, including attempting to turn HuC auth on by default again if 
things look good enough, so we need to get the GuC CI health under 
control before that.

Thanks,
Daniele

> Thanks,
> Daniele
> 
>> -Chris
>>
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/intel-gfx