Re: [Intel-gfx] [PATCH] drm/i915/xehpsdv/selftests: Flush all tiles on test exit

[Andi Shyti <andi.shyti@linux.intel.com>]

Hi Nirmoy,

> > > +	struct intel_gt *gt;
> > > +	unsigned int i;
> > > +	int ret = 0;
> > > -	cond_resched();
> > > +	for_each_gt(gt, i915, i) {
> > > +		if (intel_gt_is_wedged(gt))
> > > +			ret = -EIO;
> > I'm just wondering if it makes sense to check if the gt is
> > wedged.
> 
> Could you please expand more, what are your thoughts about this ?

if we are wedged, I do expect the wait_for_idle to fail and not
having any pending job.

But nevertheless it's not that important, it's just the way this
function is organized that makes me raise an eyebrow.

My r-b stands, still.

Andi

[...]

> > > +		cond_resched();
> > > -		GEM_TRACE("%pS timed out.\n",
> > > -			  __builtin_return_address(0));
> > > -		GEM_TRACE_DUMP();
> > > +		if (intel_gt_wait_for_idle(gt, HZ * 3) == -ETIME) {
> > > +			pr_err("%pS timed out, cancelling all further testing.\n",
> > > +			       __builtin_return_address(0));
> > > -		intel_gt_set_wedged(gt);
> > > -		ret = -EIO;
> > > +			GEM_TRACE("%pS timed out.\n",
> > > +				  __builtin_return_address(0));
> > > +			GEM_TRACE_DUMP();
> > > +
> > > +			intel_gt_set_wedged(gt);
> > > +			ret = -EIO;
> > > +		}
> > >   	}
> > >   	return ret;
> > > -- 
> > > 2.39.0