Re: [Intel-gfx] [PATCH] drm/i915/guc: Remove racey GEM_BUG_ON

[Matthew Brost <matthew.brost@intel.com>]

On Thu, Dec 09, 2021 at 11:26:09AM -0800, Daniele Ceraolo Spurio wrote:
> 
> 
> On 12/9/2021 10:51 AM, Matthew Brost wrote:
> > A full GT can race with the last context put resulting in the context
> > ref count being zero but the destroyed bit not yet being set. Remove
> > GEM_BUG_ON in scrub_guc_desc_for_outstanding_g2h that asserts the
> > destroyed bit must be set in ref count is zero.
> > 
> > Signed-off-by: Matthew Brost <matthew.brost@intel.com>
> > ---
> >   drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c | 2 --
> >   1 file changed, 2 deletions(-)
> > 
> > diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c b/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c
> > index 9b7b4f4e0d91..0f99bb83293a 100644
> > --- a/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c
> > +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c
> > @@ -1040,8 +1040,6 @@ static void scrub_guc_desc_for_outstanding_g2h(struct intel_guc *guc)
> >   		spin_unlock(&ce->guc_state.lock);
> > -		GEM_BUG_ON(!do_put && !destroyed);
> > -
> 
> Do we need to re-queue/flush the destroyer work to make sure it runs before
> we reset, or is it ok for that to run in parallel?
> 

The code in the put path will either see the reset or that it isn't
registered and destroy the context without any interaction with the GuC.

Matt

> Daniele
> 
> >   		if (pending_enable || destroyed || deregister) {
> >   			decr_outstanding_submission_g2h(guc);
> >   			if (deregister)
>