On Tue, 29 Jun 2021 10:50:36 +0200 Daniel Vetter <daniel@xxxxxxxx> wrote: > On Tue, Jun 29, 2021 at 09:34:56AM +0200, Boris Brezillon wrote: > > Mali Midgard/Bifrost GPUs have 3 hardware queues but only a global GPU > > reset. This leads to extra complexity when we need to synchronize timeout > > works with the reset work. One solution to address that is to have an > > ordered workqueue at the driver level that will be used by the different > > schedulers to queue their timeout work. Thanks to the serialization > > provided by the ordered workqueue we are guaranteed that timeout > > handlers are executed sequentially, and can thus easily reset the GPU > > from the timeout handler without extra synchronization. > > > > v5: > > * Add a new paragraph to the timedout_job() method > > > > v3: > > * New patch > > > > v4: > > * Actually use the timeout_wq to queue the timeout work > > > > Signed-off-by: Boris Brezillon <boris.brezillon@xxxxxxxxxxxxx> > > Reviewed-by: Steven Price <steven.price@xxxxxxx> > > Reviewed-by: Lucas Stach <l.stach@xxxxxxxxxxxxxx> > > Cc: Qiang Yu <yuq825@xxxxxxxxx> > > Cc: Emma Anholt <emma@xxxxxxxxxx> > > Cc: Alex Deucher <alexander.deucher@xxxxxxx> > > Cc: "Christian König" <christian.koenig@xxxxxxx> > > Acked-by: Daniel Vetter <daniel.vetter@xxxxxxxx> > > Also since I'm occasionally blinded by my own pride, add suggested-by: me? Duh, it's an oversight (I thought I had that 'Suggested-by: Daniel Vetter ...' already). > I did spend quite a bit pondering how to untangle your various lockdep > splats in the trd handler :-) And I'm grateful for your help ;-).
