[PATCH 4/9] drm/panfrost: Handle HW_ISSUE_TTRX_3076

[Alyssa Rosenzweig]

On Mon, Feb 14, 2022 at 04:23:18PM +0000, Steven Price wrote:
> On 11/02/2022 20:27, alyssa.rosenzweig at collabora.com wrote:
> > From: Alyssa Rosenzweig <alyssa.rosenzweig at collabora.com>
> > 
> > Some Valhall GPUs require resets when encountering bus faults due to
> > occlusion query writes. Add the issue bit for this and handle it.
> > 
> > Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig at collabora.com>
> 
> Reviewed-by: Steven Price <steven.price at arm.com>
> (although one nit below)
> 
> Just in case any one is wondering - these bus faults occur when
> switching the GPU's MMU to unmapped - it's not a normal "bus fault" from
> the external bus. This is triggered by an attempt to read unmapped
> memory which is completed by the driver by switching the entire MMU to
> unmapped.

Ouch, that's subtle.

> > diff --git a/drivers/gpu/drm/panfrost/panfrost_issues.h b/drivers/gpu/drm/panfrost/panfrost_issues.h
> > index a66692663833..058f6a4c8435 100644
> > --- a/drivers/gpu/drm/panfrost/panfrost_issues.h
> > +++ b/drivers/gpu/drm/panfrost/panfrost_issues.h
> > @@ -128,6 +128,10 @@ enum panfrost_hw_issue {
> >  	/* Must set SC_VAR_ALGORITHM */
> >  	HW_ISSUE_TTRX_2968_TTRX_3162,
> >  
> > +	/* Bus fault from occlusion query write may cause future fragment jobs
> > +	 * to hang */
> 
> NIT: Kernel comment style has the "/*" and "*/" on lines by themselves
> for multi-line comments. checkpatch will complain!

Yes, I am aware (and checkpatch did complain). The existing multi-line
comments in that file do not have the extra lines. Consistency within
the file seemed like the lesser evil. If you think it's better to
appease checkpatch, I can reformat for v2.

(I can also throw in a patch fixing the rest of that file's multiline
comments but that seems a bit extra.)