block/mq-deadline: Prioritize high-priority requests

While one or more requests with a certain I/O priority are pending, do not
dispatch lower priority requests. Dispatch lower priority requests anyway
after the "aging" time has expired.

This patch has been tested as follows:

modprobe scsi_debug ndelay=1000000 max_queue=16 &&
sd='' &&
while [ -z "$sd" ]; do
  sd=/dev/$(basename /sys/bus/pseudo/drivers/scsi_debug/adapter*/host*/target*/*/block/*)
done &&
echo $((100*1000)) > /sys/block/$sd/queue/iosched/aging_expire &&
cd /sys/fs/cgroup/blkio/ &&
echo $$ >cgroup.procs &&
echo restrict-to-be >blkio.prio.class &&
mkdir -p hipri &&
cd hipri &&
echo none-to-rt >blkio.prio.class &&
{ max-iops -a1 -d32 -j1 -e mq-deadline $sd >& ~/low-pri.txt & } &&
echo $$ >cgroup.procs &&
max-iops -a1 -d32 -j1 -e mq-deadline $sd >& ~/hi-pri.txt

Result:
* 11000 IOPS for the high-priority job
*    40 IOPS for the low-priority job

If the aging expiry time is changed from 100s into 0, the IOPS results change
into 6712 and 6796 IOPS.

The max-iops script is a script that runs fio with the following arguments:
--bs=4K --gtod_reduce=1 --ioengine=libaio --ioscheduler=${arg_e} --runtime=60
--norandommap --rw=read --thread --buffered=0 --numjobs=${arg_j}
--iodepth=${arg_d} --iodepth_batch_submit=${arg_a}
--iodepth_batch_complete=$((arg_d / 2)) --name=${positional_argument_1}
--filename=${positional_argument_1}

Reviewed-by: Damien Le Moal <[email protected]>
Cc: Hannes Reinecke <[email protected]>
Cc: Christoph Hellwig <[email protected]>
Cc: Ming Lei <[email protected]>
Cc: Johannes Thumshirn <[email protected]>
Cc: Himanshu Madhani <[email protected]>
Signed-off-by: Bart Van Assche <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jens Axboe <[email protected]>

Author: bvanassche

block/mq-deadline-main.c

@@ -32,6 +32,11 @@
  */
 static const int read_expire = HZ / 2;  /* max time before a read is submitted. */
 static const int write_expire = 5 * HZ; /* ditto for writes, these limits are SOFT! */
+/*
+ * Time after which to dispatch lower priority requests even if higher
+ * priority requests are pending.
+ */
+static const int aging_expire = 10 * HZ;
 static const int writes_starved = 2;    /* max times reads can starve a write */
 static const int fifo_batch = 16;       /* # of sequential requests treated as one
 				     by the above parameters. For throughput. */
@@ -94,6 +99,7 @@ struct deadline_data {
 	int writes_starved;
 	int front_merges;
 	u32 async_depth;
+	int aging_expire;
 
 	spinlock_t lock;
 	spinlock_t zone_lock;
@@ -361,10 +367,11 @@ deadline_next_request(struct deadline_data *dd, struct dd_per_prio *per_prio,
 
 /*
  * deadline_dispatch_requests selects the best request according to
- * read/write expire, fifo_batch, etc
+ * read/write expire, fifo_batch, etc and with a start time <= @latest.
  */
 static struct request *__dd_dispatch_request(struct deadline_data *dd,
-					     struct dd_per_prio *per_prio)
+					     struct dd_per_prio *per_prio,
+					     u64 latest_start_ns)
 {
 	struct request *rq, *next_rq;
 	enum dd_data_dir data_dir;
@@ -377,6 +384,8 @@ static struct request *__dd_dispatch_request(struct deadline_data *dd,
 	if (!list_empty(&per_prio->dispatch)) {
 		rq = list_first_entry(&per_prio->dispatch, struct request,
 				      queuelist);
+		if (rq->start_time_ns > latest_start_ns)
+			return NULL;
 		list_del_init(&rq->queuelist);
 		goto done;
 	}
@@ -454,6 +463,8 @@ static struct request *__dd_dispatch_request(struct deadline_data *dd,
 	dd->batching = 0;
 
 dispatch_request:
+	if (rq->start_time_ns > latest_start_ns)
+		return NULL;
 	/*
 	 * rq is the selected appropriate request.
 	 */
@@ -484,15 +495,32 @@ static struct request *__dd_dispatch_request(struct deadline_data *dd,
 static struct request *dd_dispatch_request(struct blk_mq_hw_ctx *hctx)
 {
 	struct deadline_data *dd = hctx->queue->elevator->elevator_data;
-	struct request *rq;
+	const u64 now_ns = ktime_get_ns();
+	struct request *rq = NULL;
 	enum dd_prio prio;
 
 	spin_lock(&dd->lock);
-	for (prio = 0; prio <= DD_PRIO_MAX; prio++) {
-		rq = __dd_dispatch_request(dd, &dd->per_prio[prio]);
+	/*
+	 * Start with dispatching requests whose deadline expired more than
+	 * aging_expire jiffies ago.
+	 */
+	for (prio = DD_BE_PRIO; prio <= DD_PRIO_MAX; prio++) {
+		rq = __dd_dispatch_request(dd, &dd->per_prio[prio], now_ns -
+					   jiffies_to_nsecs(dd->aging_expire));
 		if (rq)
+			goto unlock;
+	}
+	/*
+	 * Next, dispatch requests in priority order. Ignore lower priority
+	 * requests if any higher priority requests are pending.
+	 */
+	for (prio = 0; prio <= DD_PRIO_MAX; prio++) {
+		rq = __dd_dispatch_request(dd, &dd->per_prio[prio], now_ns);
+		if (rq || dd_queued(dd, prio))
 			break;
 	}
+
+unlock:
 	spin_unlock(&dd->lock);
 
 	return rq;
@@ -603,6 +631,7 @@ static int dd_init_sched(struct request_queue *q, struct elevator_type *e)
 	dd->front_merges = 1;
 	dd->last_dir = DD_WRITE;
 	dd->fifo_batch = fifo_batch;
+	dd->aging_expire = aging_expire;
 	spin_lock_init(&dd->lock);
 	spin_lock_init(&dd->zone_lock);
 
@@ -835,6 +864,7 @@ static ssize_t __FUNC(struct elevator_queue *e, char *page)		\
 #define SHOW_JIFFIES(__FUNC, __VAR) SHOW_INT(__FUNC, jiffies_to_msecs(__VAR))
 SHOW_JIFFIES(deadline_read_expire_show, dd->fifo_expire[DD_READ]);
 SHOW_JIFFIES(deadline_write_expire_show, dd->fifo_expire[DD_WRITE]);
+SHOW_JIFFIES(deadline_aging_expire_show, dd->aging_expire);
 SHOW_INT(deadline_writes_starved_show, dd->writes_starved);
 SHOW_INT(deadline_front_merges_show, dd->front_merges);
 SHOW_INT(deadline_async_depth_show, dd->front_merges);
@@ -864,6 +894,7 @@ static ssize_t __FUNC(struct elevator_queue *e, const char *page, size_t count)
 	STORE_FUNCTION(__FUNC, __PTR, MIN, MAX, msecs_to_jiffies)
 STORE_JIFFIES(deadline_read_expire_store, &dd->fifo_expire[DD_READ], 0, INT_MAX);
 STORE_JIFFIES(deadline_write_expire_store, &dd->fifo_expire[DD_WRITE], 0, INT_MAX);
+STORE_JIFFIES(deadline_aging_expire_store, &dd->aging_expire, 0, INT_MAX);
 STORE_INT(deadline_writes_starved_store, &dd->writes_starved, INT_MIN, INT_MAX);
 STORE_INT(deadline_front_merges_store, &dd->front_merges, 0, 1);
 STORE_INT(deadline_async_depth_store, &dd->front_merges, 1, INT_MAX);
@@ -882,6 +913,7 @@ static struct elv_fs_entry deadline_attrs[] = {
 	DD_ATTR(front_merges),
 	DD_ATTR(async_depth),
 	DD_ATTR(fifo_batch),
+	DD_ATTR(aging_expire),
 	__ATTR_NULL
 };