On Tue, 2017-10-17 at 08:33 +0200, Hannes Reinecke wrote: > How do you ensure that PREEMPT requests are not stuck in the queue > _behind_ non-PREEMPT requests? > Once they are in the queue the request are already allocated, so your > deferred allocation won't work. > _And_ deferred requests will be re-inserted at the head of the queue. > Consequently the PREEMPT request will never ever scheduled during quiesce. > How do you avoid such a scenario? Hello Hannes, The blk_mq_freeze_queue() / blk_mq_unfreeze_queue() calls that have been added through patch 9/10 to scsi_device_quiesce() will only succeed after all outstanding requests have finished. That includes non-preempt requests and requests that are about to be requeued. The changes in scsi_device_quiesce() are such that blk_queue_enter() will fail for non-preempt requests after the queue has been unfrozen. In other words, if a scsi_device_quiesce() call succeeds, it is guaranteed that there are no non-preempt requests in the queue and also that no new non-preempt requests will enter the queue until scsi_device_resume() is called. Bart.
