Re: [PATCH 1/2] drm/i915: Add a partial instruction shootdown workaround on Broadwell.

[Daniel Vetter <daniel@ffwll.ch>]

On Thu, Feb 27, 2014 at 04:05:04PM -0800, Ben Widawsky wrote:
> On Thu, Feb 27, 2014 at 10:43:24AM +0200, Ville Syrjälä wrote:
> > On Wed, Feb 26, 2014 at 11:59:30PM -0800, Kenneth Graunke wrote:
> > > I believe this will be necessary on production hardware.
> > > 
> > > Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
> > > ---
> > >  drivers/gpu/drm/i915/i915_reg.h | 3 +++
> > >  drivers/gpu/drm/i915/intel_pm.c | 4 ++++
> > >  2 files changed, 7 insertions(+)
> > > 
> > > I just realized tonight that my workarounds series never got merged.
> > > 
> > > After reviewing Ben and Chris's comments, I agree with all of them, so
> > > I've dropped the unnecessary patches.  I also believe I shouldn't need
> > > the pre-production workarounds I sent last time, so I've dropped those.
> > > 
> > > These two should apply to production hardware, so we probably want them.
> > > 
> > > diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
> > > index ad75ff7..f36d5e0 100644
> > > --- a/drivers/gpu/drm/i915/i915_reg.h
> > > +++ b/drivers/gpu/drm/i915/i915_reg.h
> > > @@ -5050,6 +5050,9 @@
> > >  #define   GEN7_SINGLE_SUBSCAN_DISPATCH_ENABLE	(1<<10)
> > >  #define   GEN7_PSD_SINGLE_PORT_DISPATCH_ENABLE	(1<<3)
> > >  
> > > +#define GEN8_ROW_CHICKEN		0xe4f0
> > > +#define   PARTIAL_INSTRUCTION_SHOOTDOWN_DISABLE	(1<<8)
> > > +
> > >  #define GEN7_ROW_CHICKEN2		0xe4f4
> > >  #define GEN7_ROW_CHICKEN2_GT2		0xf4f4
> > >  #define   DOP_CLOCK_GATING_DISABLE	(1<<0)
> > > diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c
> > > index 4f01b04..df8ad21 100644
> > > --- a/drivers/gpu/drm/i915/intel_pm.c
> > > +++ b/drivers/gpu/drm/i915/intel_pm.c
> > > @@ -4838,6 +4838,10 @@ static void gen8_init_clock_gating(struct drm_device *dev)
> > >  	/* FIXME(BDW): Check all the w/a, some might only apply to
> > >  	 * pre-production hw. */
> > >  
> > > +	/* WaDisablePartialInstShootdown */

out of curiousity: Does ROW_CHICKEN survive a gpu reset? We've had fun
with this kind of stuff before ...

> > 
> > +:bdw
> > 
> > Apart from that:
> > Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
> > 
> 
> I think they're both stable material.
> And:
> Reviewed-by: Ben Widawsky <ben@bwidawsk.net>

Both merged to dinq. bdw in 3.13 is prelim, so would only need to go into
3.14-fixes. But there's too much bdw fixes floating (iirc 10 patches) that
some time to settle everything looks like the right approach.

I guess as the platform owner you're then signed up to assemeble a
backports branch for 3.14-bdw.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch