Re: [PATCH v2 2/2] drm/xe/guc: Synchronize Dead CT worker with unbind

["Summers, Stuart" <stuart.summers@intel.com>]

On Fri, 2025-10-31 at 20:23 +0530, Vivekanandan, Balasubramani wrote:
> On 30.10.2025 23:08, Summers, Stuart wrote:
> > On Thu, 2025-10-30 at 20:41 +0530, Balasubramani Vivekanandan
> > wrote:
> > > Cancel and wait for any Dead CT worker to complete before
> > > continuing
> > > with device unbinding. Else the worker will end up using
> > > resources
> > > freed
> > > by the undind operation.
> > 
> > /s/undind/unbind/
> > 
> > And similar here - can you provide a stack trace?
> 
> Unfortunately I didn't save the stack trace. Kernel panic happened in
> Dead CT worker after I applied the fix from the first patch and
> retried
> the use case. Also the issue becomes sporadic after the first patch.

Makes sense. I do agree with the change here. We want to make sure the
state is captured, but by the time we get to the guc_ct_fini, the
assumption is this has already happened.

So with that minor spelling change in the commit message:
Reviewed-by: Stuart Summers <stuart.summers@intel.com>

Thanks,
Stuart

> 
> Regards,
> Bala
> 
> > 
> > Thanks,
> > Stuart
> > 
> > > 
> > > Cc: Zhanjun Dong <zhanjun.dong@intel.com>
> > > Fixes: d2c5a5a926f4 ("drm/xe/guc: Dead CT helper")
> > > Signed-off-by: Balasubramani Vivekanandan
> > > <balasubramani.vivekanandan@intel.com>
> > > ---
> > >  drivers/gpu/drm/xe/xe_guc_ct.c | 3 +++
> > >  1 file changed, 3 insertions(+)
> > > 
> > > diff --git a/drivers/gpu/drm/xe/xe_guc_ct.c
> > > b/drivers/gpu/drm/xe/xe_guc_ct.c
> > > index e68953ef3a00..536433b061fc 100644
> > > --- a/drivers/gpu/drm/xe/xe_guc_ct.c
> > > +++ b/drivers/gpu/drm/xe/xe_guc_ct.c
> > > @@ -199,6 +199,9 @@ static void guc_ct_fini(struct drm_device
> > > *drm,
> > > void *arg)
> > >  {
> > >         struct xe_guc_ct *ct = arg;
> > >  
> > > +#if IS_ENABLED(CONFIG_DRM_XE_DEBUG)
> > > +       cancel_work_sync(&ct->dead.worker);
> > > 
> > > +#endif
> > >         ct_exit_safe_mode(ct);
> > >         destroy_workqueue(ct->g2h_wq);
> > >         xa_destroy(&ct->fence_lookup);
> >