On 6/16/26 13:37, SHANMUGAM, SRINIVASAN wrote: > AMD General > > Hi Christian, > >> -----Original Message----- >> From: Koenig, Christian <[email protected]> >> Sent: Tuesday, June 16, 2026 2:32 PM >> To: SHANMUGAM, SRINIVASAN <[email protected]>; >> Deucher, Alexander <[email protected]> >> Cc: [email protected] >> Subject: Re: [PATCH 3/5] drm/amdgpu: Signal SCRATCH EVENTFD notifications >> >> On 6/12/26 07:52, Srinivasan Shanmugam wrote: >>> Signal SCRATCH EVENTFD subscriptions when user queue restore detects >>> missing VA mappings. >> >> I think the scratch event is meant for something else. >> >> When a core VA to run the queues is missing (WPTR, RPTR, ring buffer etc...) >> we >> have a hard and unrecoverable error for this queue. >> >> But scratch handling is a different interface as far as I know. > > If a queue cannot be restored because its VA mappings are missing, what > notification would you expect userspace to receive? Is SCRATCH appropriate > here, or should we handle it differently?
Differently, we probably need some notification for per queue error handling. Regards, Christian. > > Thanks, > Srini > >> >> Regards, >> Christian. >>
