block: enable passthrough command statistics

Applications using the passthrough interfaces for IO want to continue
seeing the disk stats. These requests had been fenced off from this
block layer feature. While the block layer doesn't necessarily know what
a passthrough command does, we do know the data size and direction,
which is enough to account for the command's stats.

Since tracking these has the potential to produce unexpected results,
the passthrough stats are locked behind a new queue flag that needs to
be enabled with the /sys/block/<dev>/queue/iostats_passthrough
attribute.

Signed-off-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20241007153236.2818562-1-kbusch@meta.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

Documentation/ABI/stable/sysfs-block

@@ -424,6 +424,13 @@ Description:
 		[RW] This file is used to control (on/off) the iostats
 		accounting of the disk.
 
+What:		/sys/block/<disk>/queue/iostats_passthrough
+Date:		October 2024
+Contact:	linux-block@vger.kernel.org
+Description:
+		[RW] This file is used to control (on/off) the iostats
+		accounting of the disk for passthrough commands.
+
 
 What:		/sys/block/<disk>/queue/logical_block_size
 Date:		May 2009

block/blk-mq.c

@@ -988,13 +988,43 @@ static inline void blk_account_io_done(struct request *req, u64 now)
 	}
 }
 
+static inline bool blk_rq_passthrough_stats(struct request *req)
+{
+	struct bio *bio = req->bio;
+
+	if (!blk_queue_passthrough_stat(req->q))
+		return false;
+
+	/* Requests without a bio do not transfer data. */
+	if (!bio)
+		return false;
+
+	/*
+	 * Stats are accumulated in the bdev, so must have one attached to a
+	 * bio to track stats. Most drivers do not set the bdev for passthrough
+	 * requests, but nvme is one that will set it.
+	 */
+	if (!bio->bi_bdev)
+		return false;
+
+	/*
+	 * We don't know what a passthrough command does, but we know the
+	 * payload size and data direction. Ensuring the size is aligned to the
+	 * block size filters out most commands with payloads that don't
+	 * represent sector access.
+	 */
+	if (blk_rq_bytes(req) & (bdev_logical_block_size(bio->bi_bdev) - 1))
+		return false;
+	return true;
+}
+
 static inline void blk_account_io_start(struct request *req)
 {
 	trace_block_io_start(req);
 
 	if (!blk_queue_io_stat(req->q))
 		return;
-	if (blk_rq_is_passthrough(req))
+	if (blk_rq_is_passthrough(req) && !blk_rq_passthrough_stats(req))
 		return;
 
 	req->rq_flags |= RQF_IO_STAT;

block/blk-sysfs.c

@@ -272,6 +272,34 @@ static ssize_t queue_nr_zones_show(struct gendisk *disk, char *page)
 	return queue_var_show(disk_nr_zones(disk), page);
 }
 
+static ssize_t queue_iostats_passthrough_show(struct gendisk *disk, char *page)
+{
+	return queue_var_show(blk_queue_passthrough_stat(disk->queue), page);
+}
+
+static ssize_t queue_iostats_passthrough_store(struct gendisk *disk,
+					       const char *page, size_t count)
+{
+	struct queue_limits lim;
+	unsigned long ios;
+	ssize_t ret;
+
+	ret = queue_var_store(&ios, page, count);
+	if (ret < 0)
+		return ret;
+
+	lim = queue_limits_start_update(disk->queue);
+	if (ios)
+		lim.flags |= BLK_FLAG_IOSTATS_PASSTHROUGH;
+	else
+		lim.flags &= ~BLK_FLAG_IOSTATS_PASSTHROUGH;
+
+	ret = queue_limits_commit_update(disk->queue, &lim);
+	if (ret)
+		return ret;
+
+	return count;
+}
 static ssize_t queue_nomerges_show(struct gendisk *disk, char *page)
 {
 	return queue_var_show((blk_queue_nomerges(disk->queue) << 1) |
@@ -460,6 +488,7 @@ QUEUE_RO_ENTRY(queue_max_open_zones, "max_open_zones");
 QUEUE_RO_ENTRY(queue_max_active_zones, "max_active_zones");
 
 QUEUE_RW_ENTRY(queue_nomerges, "nomerges");
+QUEUE_RW_ENTRY(queue_iostats_passthrough, "iostats_passthrough");
 QUEUE_RW_ENTRY(queue_rq_affinity, "rq_affinity");
 QUEUE_RW_ENTRY(queue_poll, "io_poll");
 QUEUE_RW_ENTRY(queue_poll_delay, "io_poll_delay");
@@ -586,6 +615,7 @@ static struct attribute *queue_attrs[] = {
 	&queue_max_open_zones_entry.attr,
 	&queue_max_active_zones_entry.attr,
 	&queue_nomerges_entry.attr,
+	&queue_iostats_passthrough_entry.attr,
 	&queue_iostats_entry.attr,
 	&queue_stable_writes_entry.attr,
 	&queue_add_random_entry.attr,

include/linux/blkdev.h

@@ -349,6 +349,9 @@ typedef unsigned int __bitwise blk_flags_t;
 /* I/O topology is misaligned */
 #define BLK_FLAG_MISALIGNED		((__force blk_flags_t)(1u << 1))
 
+/* passthrough command IO accounting */
+#define BLK_FLAG_IOSTATS_PASSTHROUGH	((__force blk_flags_t)(1u << 2))
+
 struct queue_limits {
 	blk_features_t		features;
 	blk_flags_t		flags;
@@ -617,6 +620,8 @@ void blk_queue_flag_clear(unsigned int flag, struct request_queue *q);
 	test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
 #define blk_queue_nonrot(q)	(!((q)->limits.features & BLK_FEAT_ROTATIONAL))
 #define blk_queue_io_stat(q)	((q)->limits.features & BLK_FEAT_IO_STAT)
+#define blk_queue_passthrough_stat(q)	\
+	((q)->limits.flags & BLK_FLAG_IOSTATS_PASSTHROUGH)
 #define blk_queue_dax(q)	((q)->limits.features & BLK_FEAT_DAX)
 #define blk_queue_pci_p2pdma(q)	((q)->limits.features & BLK_FEAT_PCI_P2PDMA)
 #ifdef CONFIG_BLK_RQ_ALLOC_TIME