block: move cache control settings out of queue->flags

Move the cache control settings into the queue_limits so that the flags
can be set atomically with the device queue frozen.

Add new features and flags field for the driver set flags, and internal
(usually sysfs-controlled) flags in the block layer.  Note that we'll
eventually remove enough field from queue_limits to bring it back to the
previous size.

The disable flag is inverted compared to the previous meaning, which
means it now survives a rescan, similar to the max_sectors and
max_discard_sectors user limits.

The FLUSH and FUA flags are now inherited by blk_stack_limits, which
simplified the code in dm a lot, but also causes a slight behavior
change in that dm-switch and dm-unstripe now advertise a write cache
despite setting num_flush_bios to 0.  The I/O path will handle this
gracefully, but as far as I can tell the lack of num_flush_bios
and thus flush support is a pre-existing data integrity bug in those
targets that really needs fixing, after which a non-zero num_flush_bios
should be required in dm for targets that map to underlying devices.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Ulf Hansson <ulf.hansson@linaro.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240617060532.127975-14-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

3 files changed, 14 insertions, 40 deletions

diff --git a/drivers/md/bcache/super.c b/drivers/md/bcache/super.c
index 4d11fc664cb0..cb6595c8b551 100644
--- a/drivers/md/bcache/super.c
+++ b/drivers/md/bcache/super.c
@@ -897,7 +897,6 @@ static int bcache_device_init(struct bcache_device *d, unsigned int block_size,
 		sector_t sectors, struct block_device *cached_bdev,
 		const struct block_device_operations *ops)
 {
-	struct request_queue *q;
 	const size_t max_stripes = min_t(size_t, INT_MAX,
 					 SIZE_MAX / sizeof(atomic_t));
 	struct queue_limits lim = {
@@ -909,6 +908,7 @@ static int bcache_device_init(struct bcache_device *d, unsigned int block_size,
 		.io_min			= block_size,
 		.logical_block_size	= block_size,
 		.physical_block_size	= block_size,
+		.features		= BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA,
 	};
 	uint64_t n;
 	int idx;
@@ -975,12 +975,7 @@ static int bcache_device_init(struct bcache_device *d, unsigned int block_size,
 	d->disk->fops		= ops;
 	d->disk->private_data	= d;
 
-	q = d->disk->queue;
-
 	blk_queue_flag_set(QUEUE_FLAG_NONROT, d->disk->queue);
-
-	blk_queue_write_cache(q, true, true);
-
 	return 0;
 
 out_bioset_exit:
diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c
index fd789eeb62d9..03abdae64682 100644
--- a/drivers/md/dm-table.c
+++ b/drivers/md/dm-table.c
@@ -1686,34 +1686,16 @@ combine_limits:
 	return validate_hardware_logical_block_alignment(t, limits);
 }
 
-static int device_flush_capable(struct dm_target *ti, struct dm_dev *dev,
-				sector_t start, sector_t len, void *data)
-{
-	unsigned long flush = (unsigned long) data;
-	struct request_queue *q = bdev_get_queue(dev->bdev);
-
-	return (q->queue_flags & flush);
-}
-
-static bool dm_table_supports_flush(struct dm_table *t, unsigned long flush)
+/*
+ * Check if a target requires flush support even if none of the underlying
+ * devices need it (e.g. to persist target-specific metadata).
+ */
+static bool dm_table_supports_flush(struct dm_table *t)
 {
-	/*
-	 * Require at least one underlying device to support flushes.
-	 * t->devices includes internal dm devices such as mirror logs
-	 * so we need to use iterate_devices here, which targets
-	 * supporting flushes must provide.
-	 */
 	for (unsigned int i = 0; i < t->num_targets; i++) {
 		struct dm_target *ti = dm_table_get_target(t, i);
 
-		if (!ti->num_flush_bios)
-			continue;
-
-		if (ti->flush_supported)
-			return true;
-
-		if (ti->type->iterate_devices &&
-		    ti->type->iterate_devices(ti, device_flush_capable, (void *) flush))
+		if (ti->num_flush_bios && ti->flush_supported)
 			return true;
 	}
 
@@ -1855,7 +1837,6 @@ static int device_requires_stable_pages(struct dm_target *ti,
 int dm_table_set_restrictions(struct dm_table *t, struct request_queue *q,
 			      struct queue_limits *limits)
 {
-	bool wc = false, fua = false;
 	int r;
 
 	if (dm_table_supports_nowait(t))
@@ -1876,12 +1857,8 @@ int dm_table_set_restrictions(struct dm_table *t, struct request_queue *q,
 	if (!dm_table_supports_secure_erase(t))
 		limits->max_secure_erase_sectors = 0;
 
-	if (dm_table_supports_flush(t, (1UL << QUEUE_FLAG_WC))) {
-		wc = true;
-		if (dm_table_supports_flush(t, (1UL << QUEUE_FLAG_FUA)))
-			fua = true;
-	}
-	blk_queue_write_cache(q, wc, fua);
+	if (dm_table_supports_flush(t))
+		limits->features |= BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA;
 
 	if (dm_table_supports_dax(t, device_not_dax_capable)) {
 		blk_queue_flag_set(QUEUE_FLAG_DAX, q);
diff --git a/drivers/md/md.c b/drivers/md/md.c
index 67ece2cd725f..2f4c5d1755d8 100644
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -5785,7 +5785,10 @@ struct mddev *md_alloc(dev_t dev, char *name)
 	int partitioned;
 	int shift;
 	int unit;
-	int error ;
+	int error;
+	struct queue_limits lim = {
+		.features		= BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA,
+	};
 
 	/*
 	 * Wait for any previous instance of this device to be completely
@@ -5825,7 +5828,7 @@ struct mddev *md_alloc(dev_t dev, char *name)
 		 */
 		mddev->hold_active = UNTIL_STOP;
 
-	disk = blk_alloc_disk(NULL, NUMA_NO_NODE);
+	disk = blk_alloc_disk(&lim, NUMA_NO_NODE);
 	if (IS_ERR(disk)) {
 		error = PTR_ERR(disk);
 		goto out_free_mddev;
@@ -5843,7 +5846,6 @@ struct mddev *md_alloc(dev_t dev, char *name)
 	disk->fops = &md_fops;
 	disk->private_data = mddev;
 
-	blk_queue_write_cache(disk->queue, true, true);
 	disk->events |= DISK_EVENT_MEDIA_CHANGE;
 	mddev->gendisk = disk;
 	error = add_disk(disk);