[Intel-gfx] [PATCH 06/62] drm/i915: Flush the RPS bottom-half when the GPU idles

[Chris Wilson]

On Thu, Jun 16, 2016 at 10:49:17AM +0200, Michał Winiarski wrote:
> On Fri, Jun 03, 2016 at 05:36:31PM +0100, Chris Wilson wrote:
> > Make sure that the RPS bottom-half is flushed before we set the idle
> > frequency when we decide the GPU is idle. This should prevent any races
> > with the bottom-half and setting the idle frequency, and ensures that
> > the bottom-half is bounded by the GPU's rpm reference taken for when it
> > is active (i.e. between gen6_rps_busy() and gen6_rps_idle()).
> > 
> > v2: Avoid recursively using the i915->wq - RPS does not touch the
> > struct_mutex so has no place being on the ordered i915->wq.
> > v3: Enable/disable interrupts for RPS busy/idle in order to prevent
> > further HW access from RPS outside of the wakeref.
> 
> The race can be easily observed since:
> 
> commit aed242ff7ebb697e4dff912bd4dc7ec7192f7581
> Author: Chris Wilson <chris at chris-wilson.co.uk>
> Date:   Wed Mar 18 09:48:21 2015 +0000
> 
>     drm/i915: Relax RPS contraints to allows setting minfreq on idle
> 
> Because idle_freq != min_freq_softlimit for BDW and HSW - we see a failure in
> pm_rps. Flushing RPS bottom-half partially fixes that. We need to either modify
> the test to match the current behaviour, or switch back to min_freq_softlimit
> as soon as we transition idle->active.

Ensuring we are at or above min_freq_softlimit from gen6_rps_busy() is
not a bad plan, that matches the user expectations.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre