[Intel-gfx] [PATCH 15/19] drm/i915/gt: Flush the requests after wedging on suspend

[Chris Wilson]

Quoting Tvrtko Ursulin (2019-11-19 16:12:18)
> 
> On 18/11/2019 23:02, Chris Wilson wrote:
> > Retire all requests if we resort to wedged the driver on suspend. They
> > will now be idle, so we might as we free them before shutting down.
> > 
> > Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> > ---
> >   drivers/gpu/drm/i915/gt/intel_gt_pm.c | 1 +
> >   1 file changed, 1 insertion(+)
> > 
> > diff --git a/drivers/gpu/drm/i915/gt/intel_gt_pm.c b/drivers/gpu/drm/i915/gt/intel_gt_pm.c
> > index 7a9044ac4b75..f6b5169d623f 100644
> > --- a/drivers/gpu/drm/i915/gt/intel_gt_pm.c
> > +++ b/drivers/gpu/drm/i915/gt/intel_gt_pm.c
> > @@ -256,6 +256,7 @@ static void wait_for_suspend(struct intel_gt *gt)
> >                * the gpu quiet.
> >                */
> >               intel_gt_set_wedged(gt);
> > +             intel_gt_retire_requests(gt);
> >       }
> >   
> >       intel_gt_pm_wait_for_idle(gt);
> > 
> 
> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
> 
> Or given that parking is now sync it could work to make 
> intel_gt_park_requests flush and then intel_gt_pm_wait_for_idle would 
> handle it. But that would keep the GPU alive for too long, given that 
> request retire can run independently. So perhaps this is better.

It's the unlikely path, so favours the simpler hammer.

It's what we used to do, dropped and then forgotten as the mutexes were
moved around. Hopefully, it still makes sense tomorrow.
-Chris