[Intel-gfx] [PATCH 05/14] drm/i915: Remove delay for idle_work

[Chris Wilson]

Quoting Tvrtko Ursulin (2019-05-02 14:51:31)
> 
> On 02/05/2019 14:22, Chris Wilson wrote:
> > Quoting Tvrtko Ursulin (2019-05-02 14:19:38)
> >>
> >> On 01/05/2019 12:45, Chris Wilson wrote:
> >>> diff --git a/drivers/gpu/drm/i915/i915_gem_pm.c b/drivers/gpu/drm/i915/i915_gem_pm.c
> >>> index 49b0ce594f20..ae91ad7cb31e 100644
> >>> --- a/drivers/gpu/drm/i915/i915_gem_pm.c
> >>> +++ b/drivers/gpu/drm/i915/i915_gem_pm.c
> >>> @@ -29,12 +29,12 @@ static void i915_gem_park(struct drm_i915_private *i915)
> >>>    static void idle_work_handler(struct work_struct *work)
> >>>    {
> >>>        struct drm_i915_private *i915 =
> >>> -             container_of(work, typeof(*i915), gem.idle_work.work);
> >>> +             container_of(work, typeof(*i915), gem.idle_work);
> >>>    
> >>>        mutex_lock(&i915->drm.struct_mutex);
> >>>    
> >>>        intel_wakeref_lock(&i915->gt.wakeref);
> >>> -     if (!intel_wakeref_active(&i915->gt.wakeref))
> >>> +     if (!intel_wakeref_active(&i915->gt.wakeref) && !work_pending(work))
> >>
> >> What is the reason for the !work_pending check?
> > 
> > Just that we are going to be called again, so wait until the next time to
> > see if we still need to park.
> 
> When does it get called again? If a whole new cycle of unpark-park 
> happened before the previous park was able to finish?

work_pending() implies that we've done at least one cycle while we
waited for the locks and the work is already queued to be rerun.
-Chris