dm-verity: support block number limits for different ioprio classes

Calling verity_verify_io in bh for IO of all sizes is not suitable for
embedded devices. From our tests, it can improve the performance of 4K
synchronise random reads.
For example:
./fio --name=rand_read --ioengine=psync --rw=randread --bs=4K \
 --direct=1 --numjobs=8 --runtime=60 --time_based --group_reporting \
 --filename=/dev/block/mapper/xx-verity

But it will degrade the performance of 512K synchronise sequential reads
on our devices.
For example:
./fio --name=read --ioengine=psync --rw=read --bs=512K --direct=1 \
 --numjobs=8 --runtime=60 --time_based --group_reporting \
 --filename=/dev/block/mapper/xx-verity

A parameter array is introduced by this change. And users can modify the
default config by /sys/module/dm_verity/parameters/use_bh_bytes.

The default limits for NONE/RT/BE is set to 8192.
The default limits for IDLE is set to 0.

Call verity_verify_io directly when verity_end_io is not in hardirq.

Signed-off-by: LongPing Wei <weilongping@oppo.com>
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>

Author: LongPing Wei

Documentation/admin-guide/device-mapper/verity.rst

@@ -151,8 +151,15 @@ root_hash_sig_key_desc <key_description>
     already in the secondary trusted keyring.
 
 try_verify_in_tasklet
-    If verity hashes are in cache, verify data blocks in kernel tasklet instead
-    of workqueue. This option can reduce IO latency.
+    If verity hashes are in cache and the IO size does not exceed the limit,
+    verify data blocks in bottom half instead of workqueue. This option can
+    reduce IO latency. The size limits can be configured via
+    /sys/module/dm_verity/parameters/use_bh_bytes. The four parameters
+    correspond to limits for IOPRIO_CLASS_NONE, IOPRIO_CLASS_RT,
+    IOPRIO_CLASS_BE and IOPRIO_CLASS_IDLE in turn.
+    For example:
+    <none>,<rt>,<be>,<idle>
+    4096,4096,4096,4096
 
 Theory of operation
 ===================

drivers/md/dm-verity-target.c

@@ -30,6 +30,7 @@
 #define DM_VERITY_ENV_VAR_NAME		"DM_VERITY_ERR_BLOCK_NR"
 
 #define DM_VERITY_DEFAULT_PREFETCH_SIZE	262144
+#define DM_VERITY_USE_BH_DEFAULT_BYTES	8192
 
 #define DM_VERITY_MAX_CORRUPTED_ERRS	100
 
@@ -49,6 +50,15 @@ static unsigned int dm_verity_prefetch_cluster = DM_VERITY_DEFAULT_PREFETCH_SIZE
 
 module_param_named(prefetch_cluster, dm_verity_prefetch_cluster, uint, 0644);
 
+static unsigned int dm_verity_use_bh_bytes[4] = {
+	DM_VERITY_USE_BH_DEFAULT_BYTES,	// IOPRIO_CLASS_NONE
+	DM_VERITY_USE_BH_DEFAULT_BYTES,	// IOPRIO_CLASS_RT
+	DM_VERITY_USE_BH_DEFAULT_BYTES,	// IOPRIO_CLASS_BE
+	0				// IOPRIO_CLASS_IDLE
+};
+
+module_param_array_named(use_bh_bytes, dm_verity_use_bh_bytes, uint, NULL, 0644);
+
 static DEFINE_STATIC_KEY_FALSE(use_bh_wq_enabled);
 
 /* Is at least one dm-verity instance using ahash_tfm instead of shash_tfm? */
@@ -669,9 +679,17 @@ static void verity_bh_work(struct work_struct *w)
 	verity_finish_io(io, errno_to_blk_status(err));
 }
 
+static inline bool verity_use_bh(unsigned int bytes, unsigned short ioprio)
+{
+	return ioprio <= IOPRIO_CLASS_IDLE &&
+		bytes <= READ_ONCE(dm_verity_use_bh_bytes[ioprio]);
+}
+
 static void verity_end_io(struct bio *bio)
 {
 	struct dm_verity_io *io = bio->bi_private;
+	unsigned short ioprio = IOPRIO_PRIO_CLASS(bio->bi_ioprio);
+	unsigned int bytes = io->n_blocks << io->v->data_dev_block_bits;
 
 	if (bio->bi_status &&
 	    (!verity_fec_is_enabled(io->v) ||
@@ -681,9 +699,14 @@ static void verity_end_io(struct bio *bio)
 		return;
 	}
 
-	if (static_branch_unlikely(&use_bh_wq_enabled) && io->v->use_bh_wq) {
-		INIT_WORK(&io->bh_work, verity_bh_work);
-		queue_work(system_bh_wq, &io->bh_work);
+	if (static_branch_unlikely(&use_bh_wq_enabled) && io->v->use_bh_wq &&
+		verity_use_bh(bytes, ioprio)) {
+		if (in_hardirq() || irqs_disabled()) {
+			INIT_WORK(&io->bh_work, verity_bh_work);
+			queue_work(system_bh_wq, &io->bh_work);
+		} else {
+			verity_bh_work(&io->bh_work);
+		}
 	} else {
 		INIT_WORK(&io->work, verity_work);
 		queue_work(io->v->verify_wq, &io->work);