On 4/5/22 4:04 AM, Stefan Metzmacher wrote: > Hi Jens, > >> Most of this is prep patches, but the purpose is to make sure that we >> treat file assignment for links appropriately. If not, then we cannot >> use direct open/accept with links while avoiding separate submit+wait >> cycles. >> >> v4: >> - Drop merged patch for msg-ring >> - Drop inflight tracking completely, pointless now >> - Fix locking issue around file assignment > > If the behavior change is backported to 5.15 stable, > don't we need a flag to indicate that a newer kernel supports it? > > Otherwise how could userspace know it can rely on the new behavior? Yes, I have been thinking that too, we should add a FEAT flag for this. -- Jens Axboe
