On 10/12/2017 09:22 AM, Ming Lei wrote: > On Thu, Oct 12, 2017 at 08:52:12AM -0600, Jens Axboe wrote: >> On 10/12/2017 04:01 AM, Ming Lei wrote: >>> On Tue, Oct 10, 2017 at 11:23:45AM -0700, Omar Sandoval wrote: >>>> On Mon, Oct 09, 2017 at 07:24:23PM +0800, Ming Lei wrote: >>>>> SCSI devices use host-wide tagset, and the shared driver tag space is >>>>> often quite big. Meantime there is also queue depth for each lun( >>>>> .cmd_per_lun), which is often small, for example, on both lpfc and >>>>> qla2xxx, .cmd_per_lun is just 3. >>>>> >>>>> So lots of requests may stay in sw queue, and we always flush all >>>>> belonging to same hw queue and dispatch them all to driver, unfortunately >>>>> it is easy to cause queue busy because of the small .cmd_per_lun. >>>>> Once these requests are flushed out, they have to stay in hctx->dispatch, >>>>> and no bio merge can participate into these requests, and sequential IO >>>>> performance is hurt a lot. >>>>> >>>>> This patch introduces blk_mq_dequeue_from_ctx for dequeuing request from >>>>> sw queue so that we can dispatch them in scheduler's way, then we can >>>>> avoid to dequeue too many requests from sw queue when ->dispatch isn't >>>>> flushed completely. >>>>> >>>>> This patch improves dispatching from sw queue when there is per-request-queue >>>>> queue depth by taking request one by one from sw queue, just like the way >>>>> of IO scheduler. >>>> >>>> This still didn't address Jens' concern about using q->queue_depth as >>>> the heuristic for whether to do the full sw queue flush or one-by-one >>>> dispatch. The EWMA approach is a bit too complex for now, can you please >>>> try the heuristic of whether the driver ever returned BLK_STS_RESOURCE? >>> >>> That can be done easily, but I am not sure if it is good. >>> >>> For example, inside queue rq path of NVMe, kmalloc(GFP_ATOMIC) is >>> often used, if kmalloc() returns NULL just once, BLK_STS_RESOURCE >>> will be returned to blk-mq, then blk-mq will never do full sw >>> queue flush even when kmalloc() always succeed from that time >>> on. >> >> Have it be a bit more than a single bit, then. Reset it every x IOs or >> something like that, that'll be more representative of transient busy >> conditions anyway. > > OK, that can be done via a simplified EWMA by considering > the dispatch result only. Yes, if it's kept simple enough, then that would be fine. I'm not totally against EWMA, I just don't want to have any of this over-engineered. Especially not when it's a pretty simple thing, we don't care about averages, basically only if we ever see BLK_STS_RESOURCE in any kind of recurring fashion. -- Jens Axboe
