Quoting Tvrtko Ursulin (2018-02-05 09:47:52)
>
> On 05/02/2018 09:11, Chris Wilson wrote:
> > The pmu_event_read() callchain is in an atomic context which then
>
> Call-chain.
>
> > forbids us sleeping, such as required to wake the device up for rpm.
> > In this case, we can only read our registers iff the device is already
>
> s/iff/if/
iff == if and only if
> > awake. In the case of rc6 this means that we cannot report the latest
> > values when the whole system is asleep.
> > ---
> > drivers/gpu/drm/i915/i915_pmu.c | 29 +++++++++++++++++------------
> > drivers/gpu/drm/i915/i915_pmu.h | 1 +
> > 2 files changed, 18 insertions(+), 12 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/i915/i915_pmu.c b/drivers/gpu/drm/i915/i915_pmu.c
> > index ecb0198bfb7a..9c50b173e9a1 100644
> > --- a/drivers/gpu/drm/i915/i915_pmu.c
> > +++ b/drivers/gpu/drm/i915/i915_pmu.c
> > @@ -423,18 +423,23 @@ static u64 __i915_pmu_event_read(struct perf_event *event)
> > val = count_interrupts(i915);
> > break;
> > case I915_PMU_RC6_RESIDENCY:
> > - intel_runtime_pm_get(i915);
> > - val = intel_rc6_residency_ns(i915,
> > - IS_VALLEYVIEW(i915) ?
> > - VLV_GT_RENDER_RC6 :
> > - GEN6_GT_GFX_RC6);
> > - if (HAS_RC6p(i915))
> > - val += intel_rc6_residency_ns(i915,
> > - GEN6_GT_GFX_RC6p);
> > - if (HAS_RC6pp(i915))
> > - val += intel_rc6_residency_ns(i915,
> > - GEN6_GT_GFX_RC6pp);
> > - intel_runtime_pm_put(i915);
> > + if (intel_runtime_pm_get_if_in_use(i915)) {
> > + val = intel_rc6_residency_ns(i915,
> > + IS_VALLEYVIEW(i915) ?
> > + VLV_GT_RENDER_RC6 :
> > + GEN6_GT_GFX_RC6);
> > + if (HAS_RC6p(i915))
> > + val += intel_rc6_residency_ns(i915,
> > + GEN6_GT_GFX_RC6p);
> > + if (HAS_RC6pp(i915))
> > + val += intel_rc6_residency_ns(i915,
> > + GEN6_GT_GFX_RC6pp);
> > + intel_runtime_pm_put(i915);
> > +
> > + i915->pmu.sample[__I915_SAMPLE_RC6].cur = val;
> > + } else {
> > + val = i915->pmu.sample[__I915_SAMPLE_RC6].cur;
> > + }
>
> Downside is that idle system, when auto-suspended, will show 0% RC6. But
> I have no alternative ideas. :(
And when it wakes back up >100% RC6 for an interval. Unpleasant.
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
