dm verity: enable WQ_HIGHPRI on verify_wq

WQ_HIGHPRI increases throughput and decreases disk latency when using
dm-verity. This is important in Android for camera startup speed.

The following tests were run by doing 60 seconds of random reads using
a dm-verity device backed by two ramdisks.

Without WQ_HIGHPRI
lat (usec): min=13, max=3947, avg=69.53, stdev=50.55
READ: bw=51.1MiB/s (53.6MB/s), 51.1MiB/s-51.1MiB/s (53.6MB/s-53.6MB/s)

With WQ_HIGHPRI:
lat (usec): min=13, max=7854, avg=31.15, stdev=30.42
READ: bw=116MiB/s (121MB/s), 116MiB/s-116MiB/s (121MB/s-121MB/s)

Further testing was done by measuring how long it takes to open a
camera on an Android device.

Without WQ_HIGHPRI
Total verity work queue wait times (ms):
880.960, 789.517, 898.852

With WQ_HIGHPRI:
Total verity work queue wait times (ms):
528.824, 439.191, 433.300

The average time to open the camera is reduced by 350ms (or 40-50%).

Signed-off-by: Nathan Huckleberry <nhuck@google.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>

1 files changed, 10 insertions, 8 deletions

diff --git a/drivers/md/dm-verity-target.c b/drivers/md/dm-verity-target.c
index 8a00cc42e498..ccf5b852fbf7 100644
--- a/drivers/md/dm-verity-target.c
+++ b/drivers/md/dm-verity-target.c
@@ -1401,14 +1401,16 @@ static int verity_ctr(struct dm_target *ti, unsigned argc, char **argv)
 
 	/* WQ_UNBOUND greatly improves performance when running on ramdisk */
 	wq_flags = WQ_MEM_RECLAIM | WQ_UNBOUND;
-	if (v->use_tasklet) {
-		/*
-		 * Allow verify_wq to preempt softirq since verification in
-		 * tasklet will fall-back to using it for error handling
-		 * (or if the bufio cache doesn't have required hashes).
-		 */
-		wq_flags |= WQ_HIGHPRI;
-	}
+	/*
+	 * Using WQ_HIGHPRI improves throughput and completion latency by
+	 * reducing wait times when reading from a dm-verity device.
+	 *
+	 * Also as required for the "try_verify_in_tasklet" feature: WQ_HIGHPRI
+	 * allows verify_wq to preempt softirq since verification in tasklet
+	 * will fall-back to using it for error handling (or if the bufio cache
+	 * doesn't have required hashes).
+	 */
+	wq_flags |= WQ_HIGHPRI;
 	v->verify_wq = alloc_workqueue("kverityd", wq_flags, num_online_cpus());
 	if (!v->verify_wq) {
 		ti->error = "Cannot allocate workqueue";