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>

Documentation/block/writeback_cache_control.rst

@@ -46,41 +46,50 @@ worry if the underlying devices need any explicit cache flushing and how
 the Forced Unit Access is implemented.  The REQ_PREFLUSH and REQ_FUA flags
 may both be set on a single bio.
 
-
-Implementation details for bio based block drivers
---------------------------------------------------------------
-
-These drivers will always see the REQ_PREFLUSH and REQ_FUA bits as they sit
-directly below the submit_bio interface.  For remapping drivers the REQ_FUA
-bits need to be propagated to underlying devices, and a global flush needs
-to be implemented for bios with the REQ_PREFLUSH bit set.  For real device
-drivers that do not have a volatile cache the REQ_PREFLUSH and REQ_FUA bits
-on non-empty bios can simply be ignored, and REQ_PREFLUSH requests without
-data can be completed successfully without doing any work.  Drivers for
-devices with volatile caches need to implement the support for these
-flags themselves without any help from the block layer.
-
-
-Implementation details for request_fn based block drivers
----------------------------------------------------------
+Feature settings for block drivers
+----------------------------------
 
 For devices that do not support volatile write caches there is no driver
 support required, the block layer completes empty REQ_PREFLUSH requests before
 entering the driver and strips off the REQ_PREFLUSH and REQ_FUA bits from
-requests that have a payload.  For devices with volatile write caches the
-driver needs to tell the block layer that it supports flushing caches by
-doing::
+requests that have a payload.
 
-	blk_queue_write_cache(sdkp->disk->queue, true, false);
+For devices with volatile write caches the driver needs to tell the block layer
+that it supports flushing caches by setting the
 
-and handle empty REQ_OP_FLUSH requests in its prep_fn/request_fn.  Note that
-REQ_PREFLUSH requests with a payload are automatically turned into a sequence
-of an empty REQ_OP_FLUSH request followed by the actual write by the block
-layer.  For devices that also support the FUA bit the block layer needs
-to be told to pass through the REQ_FUA bit using::
+   BLK_FEAT_WRITE_CACHE
 
-	blk_queue_write_cache(sdkp->disk->queue, true, true);
+flag in the queue_limits feature field.  For devices that also support the FUA
+bit the block layer needs to be told to pass on the REQ_FUA bit by also setting
+the
 
-and the driver must handle write requests that have the REQ_FUA bit set
-in prep_fn/request_fn.  If the FUA bit is not natively supported the block
-layer turns it into an empty REQ_OP_FLUSH request after the actual write.
+   BLK_FEAT_FUA
+
+flag in the features field of the queue_limits structure.
+
+Implementation details for bio based block drivers
+--------------------------------------------------
+
+For bio based drivers the REQ_PREFLUSH and REQ_FUA bit are simplify passed on
+to the driver if the drivers sets the BLK_FEAT_WRITE_CACHE flag and the drivers
+needs to handle them.
+
+*NOTE*: The REQ_FUA bit also gets passed on when the BLK_FEAT_FUA flags is
+_not_ set.  Any bio based driver that sets BLK_FEAT_WRITE_CACHE also needs to
+handle REQ_FUA.
+
+For remapping drivers the REQ_FUA bits need to be propagated to underlying
+devices, and a global flush needs to be implemented for bios with the
+REQ_PREFLUSH bit set.
+
+Implementation details for blk-mq drivers
+-----------------------------------------
+
+When the BLK_FEAT_WRITE_CACHE flag is set, REQ_OP_WRITE | REQ_PREFLUSH requests
+with a payload are automatically turned into a sequence of a REQ_OP_FLUSH
+request followed by the actual write by the block layer.
+
+When the BLK_FEAT_FUA flags is set, the REQ_FUA bit simplify passed on for the
+REQ_OP_WRITE request, else a REQ_OP_FLUSH request is sent by the block layer
+after the completion of the write request for bio submissions with the REQ_FUA
+bit set.

arch/um/drivers/ubd_kern.c

@@ -835,6 +835,7 @@ static int ubd_add(int n, char **error_out)
 	struct queue_limits lim = {
 		.max_segments		= MAX_SG,
 		.seg_boundary_mask	= PAGE_SIZE - 1,
+		.features		= BLK_FEAT_WRITE_CACHE,
 	};
 	struct gendisk *disk;
 	int err = 0;
@@ -882,7 +883,6 @@ static int ubd_add(int n, char **error_out)
 	}
 
 	blk_queue_flag_set(QUEUE_FLAG_NONROT, disk->queue);
-	blk_queue_write_cache(disk->queue, true, false);
 	disk->major = UBD_MAJOR;
 	disk->first_minor = n << UBD_SHIFT;
 	disk->minors = 1 << UBD_SHIFT;

block/blk-core.c

@@ -782,7 +782,7 @@ void submit_bio_noacct(struct bio *bio)
 		if (WARN_ON_ONCE(bio_op(bio) != REQ_OP_WRITE &&
 				 bio_op(bio) != REQ_OP_ZONE_APPEND))
 			goto end_io;
-		if (!test_bit(QUEUE_FLAG_WC, &q->queue_flags)) {
+		if (!bdev_write_cache(bdev)) {
 			bio->bi_opf &= ~(REQ_PREFLUSH | REQ_FUA);
 			if (!bio_sectors(bio)) {
 				status = BLK_STS_OK;

block/blk-flush.c

@@ -381,8 +381,8 @@ static void blk_rq_init_flush(struct request *rq)
 bool blk_insert_flush(struct request *rq)
 {
 	struct request_queue *q = rq->q;
-	unsigned long fflags = q->queue_flags;	/* may change, cache */
 	struct blk_flush_queue *fq = blk_get_flush_queue(q, rq->mq_ctx);
+	bool supports_fua = q->limits.features & BLK_FEAT_FUA;
 	unsigned int policy = 0;
 
 	/* FLUSH/FUA request must never be merged */
@@ -394,11 +394,10 @@ bool blk_insert_flush(struct request *rq)
 	/*
 	 * Check which flushes we need to sequence for this operation.
 	 */
-	if (fflags & (1UL << QUEUE_FLAG_WC)) {
+	if (blk_queue_write_cache(q)) {
 		if (rq->cmd_flags & REQ_PREFLUSH)
 			policy |= REQ_FSEQ_PREFLUSH;
-		if (!(fflags & (1UL << QUEUE_FLAG_FUA)) &&
-		    (rq->cmd_flags & REQ_FUA))
+		if ((rq->cmd_flags & REQ_FUA) && !supports_fua)
 			policy |= REQ_FSEQ_POSTFLUSH;
 	}
 
@@ -407,7 +406,7 @@ bool blk_insert_flush(struct request *rq)
 	 * REQ_PREFLUSH and FUA for the driver.
 	 */
 	rq->cmd_flags &= ~REQ_PREFLUSH;
-	if (!(fflags & (1UL << QUEUE_FLAG_FUA)))
+	if (!supports_fua)
 		rq->cmd_flags &= ~REQ_FUA;
 
 	/*

block/blk-mq-debugfs.c

@@ -93,8 +93,6 @@ static const char *const blk_queue_flag_name[] = {
 	QUEUE_FLAG_NAME(INIT_DONE),
 	QUEUE_FLAG_NAME(STABLE_WRITES),
 	QUEUE_FLAG_NAME(POLL),
-	QUEUE_FLAG_NAME(WC),
-	QUEUE_FLAG_NAME(FUA),
 	QUEUE_FLAG_NAME(DAX),
 	QUEUE_FLAG_NAME(STATS),
 	QUEUE_FLAG_NAME(REGISTERED),

block/blk-settings.c

@@ -261,6 +261,9 @@ static int blk_validate_limits(struct queue_limits *lim)
 		lim->misaligned = 0;
 	}
 
+	if (!(lim->features & BLK_FEAT_WRITE_CACHE))
+		lim->features &= ~BLK_FEAT_FUA;
+
 	err = blk_validate_integrity_limits(lim);
 	if (err)
 		return err;
@@ -454,6 +457,8 @@ int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
 {
 	unsigned int top, bottom, alignment, ret = 0;
 
+	t->features |= (b->features & BLK_FEAT_INHERIT_MASK);
+
 	t->max_sectors = min_not_zero(t->max_sectors, b->max_sectors);
 	t->max_user_sectors = min_not_zero(t->max_user_sectors,
 			b->max_user_sectors);
@@ -711,30 +716,6 @@ void blk_set_queue_depth(struct request_queue *q, unsigned int depth)
 }
 EXPORT_SYMBOL(blk_set_queue_depth);
 
-/**
- * blk_queue_write_cache - configure queue's write cache
- * @q:		the request queue for the device
- * @wc:		write back cache on or off
- * @fua:	device supports FUA writes, if true
- *
- * Tell the block layer about the write cache of @q.
- */
-void blk_queue_write_cache(struct request_queue *q, bool wc, bool fua)
-{
-	if (wc) {
-		blk_queue_flag_set(QUEUE_FLAG_HW_WC, q);
-		blk_queue_flag_set(QUEUE_FLAG_WC, q);
-	} else {
-		blk_queue_flag_clear(QUEUE_FLAG_HW_WC, q);
-		blk_queue_flag_clear(QUEUE_FLAG_WC, q);
-	}
-	if (fua)
-		blk_queue_flag_set(QUEUE_FLAG_FUA, q);
-	else
-		blk_queue_flag_clear(QUEUE_FLAG_FUA, q);
-}
-EXPORT_SYMBOL_GPL(blk_queue_write_cache);
-
 int bdev_alignment_offset(struct block_device *bdev)
 {
 	struct request_queue *q = bdev_get_queue(bdev);

block/blk-sysfs.c

@@ -423,32 +423,41 @@ static ssize_t queue_io_timeout_store(struct request_queue *q, const char *page,
 
 static ssize_t queue_wc_show(struct request_queue *q, char *page)
 {
-	if (test_bit(QUEUE_FLAG_WC, &q->queue_flags))
-		return sprintf(page, "write back\n");
-
-	return sprintf(page, "write through\n");
+	if (q->limits.features & BLK_FLAGS_WRITE_CACHE_DISABLED)
+		return sprintf(page, "write through\n");
+	return sprintf(page, "write back\n");
 }
 
 static ssize_t queue_wc_store(struct request_queue *q, const char *page,
 			      size_t count)
 {
+	struct queue_limits lim;
+	bool disable;
+	int err;
+
 	if (!strncmp(page, "write back", 10)) {
-		if (!test_bit(QUEUE_FLAG_HW_WC, &q->queue_flags))
-			return -EINVAL;
-		blk_queue_flag_set(QUEUE_FLAG_WC, q);
+		disable = false;
 	} else if (!strncmp(page, "write through", 13) ||
-		 !strncmp(page, "none", 4)) {
-		blk_queue_flag_clear(QUEUE_FLAG_WC, q);
+		   !strncmp(page, "none", 4)) {
+		disable = true;
 	} else {
 		return -EINVAL;
 	}
 
+	lim = queue_limits_start_update(q);
+	if (disable)
+		lim.flags |= BLK_FLAGS_WRITE_CACHE_DISABLED;
+	else
+		lim.flags &= ~BLK_FLAGS_WRITE_CACHE_DISABLED;
+	err = queue_limits_commit_update(q, &lim);
+	if (err)
+		return err;
 	return count;
 }
 
 static ssize_t queue_fua_show(struct request_queue *q, char *page)
 {
-	return sprintf(page, "%u\n", test_bit(QUEUE_FLAG_FUA, &q->queue_flags));
+	return sprintf(page, "%u\n", !!(q->limits.features & BLK_FEAT_FUA));
 }
 
 static ssize_t queue_dax_show(struct request_queue *q, char *page)

block/blk-wbt.c

@@ -206,8 +206,8 @@ static void wbt_rqw_done(struct rq_wb *rwb, struct rq_wait *rqw,
 	 */
 	if (wb_acct & WBT_DISCARD)
 		limit = rwb->wb_background;
-	else if (test_bit(QUEUE_FLAG_WC, &rwb->rqos.disk->queue->queue_flags) &&
-	         !wb_recent_wait(rwb))
+	else if (blk_queue_write_cache(rwb->rqos.disk->queue) &&
+		 !wb_recent_wait(rwb))
 		limit = 0;
 	else
 		limit = rwb->wb_normal;

drivers/block/drbd/drbd_main.c

@@ -2697,6 +2697,7 @@ enum drbd_ret_code drbd_create_device(struct drbd_config_context *adm_ctx, unsig
 		 * connect.
 		 */
 		.max_hw_sectors		= DRBD_MAX_BIO_SIZE_SAFE >> 8,
+		.features		= BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA,
 	};
 
 	device = minor_to_device(minor);
@@ -2736,7 +2737,6 @@ enum drbd_ret_code drbd_create_device(struct drbd_config_context *adm_ctx, unsig
 	disk->private_data = device;
 
 	blk_queue_flag_set(QUEUE_FLAG_STABLE_WRITES, disk->queue);
-	blk_queue_write_cache(disk->queue, true, true);
 
 	device->md_io.page = alloc_page(GFP_KERNEL);
 	if (!device->md_io.page)

drivers/block/loop.c

@@ -985,6 +985,9 @@ static int loop_reconfigure_limits(struct loop_device *lo, unsigned short bsize)
 	lim.logical_block_size = bsize;
 	lim.physical_block_size = bsize;
 	lim.io_min = bsize;
+	lim.features &= ~BLK_FEAT_WRITE_CACHE;
+	if (file->f_op->fsync && !(lo->lo_flags & LO_FLAGS_READ_ONLY))
+		lim.features |= BLK_FEAT_WRITE_CACHE;
 	if (!backing_bdev || bdev_nonrot(backing_bdev))
 		blk_queue_flag_set(QUEUE_FLAG_NONROT, lo->lo_queue);
 	else
@@ -1078,9 +1081,6 @@ static int loop_configure(struct loop_device *lo, blk_mode_t mode,
 	lo->old_gfp_mask = mapping_gfp_mask(mapping);
 	mapping_set_gfp_mask(mapping, lo->old_gfp_mask & ~(__GFP_IO|__GFP_FS));
 
-	if (!(lo->lo_flags & LO_FLAGS_READ_ONLY) && file->f_op->fsync)
-		blk_queue_write_cache(lo->lo_queue, true, false);
-
 	error = loop_reconfigure_limits(lo, config->block_size);
 	if (WARN_ON_ONCE(error))
 		goto out_unlock;
@@ -1131,9 +1131,6 @@ static void __loop_clr_fd(struct loop_device *lo, bool release)
 	struct file *filp;
 	gfp_t gfp = lo->old_gfp_mask;
 
-	if (test_bit(QUEUE_FLAG_WC, &lo->lo_queue->queue_flags))
-		blk_queue_write_cache(lo->lo_queue, false, false);
-
 	/*
 	 * Freeze the request queue when unbinding on a live file descriptor and
 	 * thus an open device.  When called from ->release we are guaranteed

drivers/block/nbd.c

@@ -342,12 +342,14 @@ static int __nbd_set_size(struct nbd_device *nbd, loff_t bytesize,
 		lim.max_hw_discard_sectors = UINT_MAX;
 	else
 		lim.max_hw_discard_sectors = 0;
-	if (!(nbd->config->flags & NBD_FLAG_SEND_FLUSH))
-		blk_queue_write_cache(nbd->disk->queue, false, false);
-	else if (nbd->config->flags & NBD_FLAG_SEND_FUA)
-		blk_queue_write_cache(nbd->disk->queue, true, true);
-	else
-		blk_queue_write_cache(nbd->disk->queue, true, false);
+	if (!(nbd->config->flags & NBD_FLAG_SEND_FLUSH)) {
+		lim.features &= ~(BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA);
+	} else if (nbd->config->flags & NBD_FLAG_SEND_FUA) {
+		lim.features |= BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA;
+	} else {
+		lim.features |= BLK_FEAT_WRITE_CACHE;
+		lim.features &= ~BLK_FEAT_FUA;
+	}
 	lim.logical_block_size = blksize;
 	lim.physical_block_size = blksize;
 	error = queue_limits_commit_update(nbd->disk->queue, &lim);

drivers/block/null_blk/main.c

@@ -1928,6 +1928,13 @@ static int null_add_dev(struct nullb_device *dev)
 			goto out_cleanup_tags;
 	}
 
+	if (dev->cache_size > 0) {
+		set_bit(NULLB_DEV_FL_CACHE, &nullb->dev->flags);
+		lim.features |= BLK_FEAT_WRITE_CACHE;
+		if (dev->fua)
+			lim.features |= BLK_FEAT_FUA;
+	}
+
 	nullb->disk = blk_mq_alloc_disk(nullb->tag_set, &lim, nullb);
 	if (IS_ERR(nullb->disk)) {
 		rv = PTR_ERR(nullb->disk);
@@ -1940,11 +1947,6 @@ static int null_add_dev(struct nullb_device *dev)
 		nullb_setup_bwtimer(nullb);
 	}
 
-	if (dev->cache_size > 0) {
-		set_bit(NULLB_DEV_FL_CACHE, &nullb->dev->flags);
-		blk_queue_write_cache(nullb->q, true, dev->fua);
-	}
-
 	nullb->q->queuedata = nullb;
 	blk_queue_flag_set(QUEUE_FLAG_NONROT, nullb->q);
 

drivers/block/ps3disk.c

@@ -388,9 +388,8 @@ static int ps3disk_probe(struct ps3_system_bus_device *_dev)
 		.max_segments		= -1,
 		.max_segment_size	= dev->bounce_size,
 		.dma_alignment		= dev->blk_size - 1,
+		.features		= BLK_FEAT_WRITE_CACHE,
 	};
-
-	struct request_queue *queue;
 	struct gendisk *gendisk;
 
 	if (dev->blk_size < 512) {
@@ -447,10 +446,6 @@ static int ps3disk_probe(struct ps3_system_bus_device *_dev)
 		goto fail_free_tag_set;
 	}
 
-	queue = gendisk->queue;
-
-	blk_queue_write_cache(queue, true, false);
-
 	priv->gendisk = gendisk;
 	gendisk->major = ps3disk_major;
 	gendisk->first_minor = devidx * PS3DISK_MINORS;

drivers/block/rnbd/rnbd-clt.c

@@ -1389,6 +1389,12 @@ static int rnbd_client_setup_device(struct rnbd_clt_dev *dev,
 			le32_to_cpu(rsp->max_discard_sectors);
 	}
 
+	if (rsp->cache_policy & RNBD_WRITEBACK) {
+		lim.features |= BLK_FEAT_WRITE_CACHE;
+		if (rsp->cache_policy & RNBD_FUA)
+			lim.features |= BLK_FEAT_FUA;
+	}
+
 	dev->gd = blk_mq_alloc_disk(&dev->sess->tag_set, &lim, dev);
 	if (IS_ERR(dev->gd))
 		return PTR_ERR(dev->gd);
@@ -1397,10 +1403,6 @@ static int rnbd_client_setup_device(struct rnbd_clt_dev *dev,
 
 	blk_queue_flag_set(QUEUE_FLAG_SAME_COMP, dev->queue);
 	blk_queue_flag_set(QUEUE_FLAG_SAME_FORCE, dev->queue);
-	blk_queue_write_cache(dev->queue,
-			      !!(rsp->cache_policy & RNBD_WRITEBACK),
-			      !!(rsp->cache_policy & RNBD_FUA));
-
 	return rnbd_clt_setup_gen_disk(dev, rsp, idx);
 }
 

drivers/block/ublk_drv.c

@@ -487,8 +487,6 @@ static void ublk_dev_param_basic_apply(struct ublk_device *ub)
 	struct request_queue *q = ub->ub_disk->queue;
 	const struct ublk_param_basic *p = &ub->params.basic;
 
-	blk_queue_write_cache(q, p->attrs & UBLK_ATTR_VOLATILE_CACHE,
-			p->attrs & UBLK_ATTR_FUA);
 	if (p->attrs & UBLK_ATTR_ROTATIONAL)
 		blk_queue_flag_clear(QUEUE_FLAG_NONROT, q);
 	else
@@ -2210,6 +2208,12 @@ static int ublk_ctrl_start_dev(struct ublk_device *ub, struct io_uring_cmd *cmd)
 		lim.max_zone_append_sectors = p->max_zone_append_sectors;
 	}
 
+	if (ub->params.basic.attrs & UBLK_ATTR_VOLATILE_CACHE) {
+		lim.features |= BLK_FEAT_WRITE_CACHE;
+		if (ub->params.basic.attrs & UBLK_ATTR_FUA)
+			lim.features |= BLK_FEAT_FUA;
+	}
+
 	if (wait_for_completion_interruptible(&ub->completion) != 0)
 		return -EINTR;
 

drivers/block/virtio_blk.c

@@ -1100,6 +1100,7 @@ cache_type_store(struct device *dev, struct device_attribute *attr,
 	struct gendisk *disk = dev_to_disk(dev);
 	struct virtio_blk *vblk = disk->private_data;
 	struct virtio_device *vdev = vblk->vdev;
+	struct queue_limits lim;
 	int i;
 
 	BUG_ON(!virtio_has_feature(vblk->vdev, VIRTIO_BLK_F_CONFIG_WCE));
@@ -1108,7 +1109,17 @@ cache_type_store(struct device *dev, struct device_attribute *attr,
 		return i;
 
 	virtio_cwrite8(vdev, offsetof(struct virtio_blk_config, wce), i);
-	blk_queue_write_cache(disk->queue, virtblk_get_cache_mode(vdev), false);
+
+	lim = queue_limits_start_update(disk->queue);
+	if (virtblk_get_cache_mode(vdev))
+		lim.features |= BLK_FEAT_WRITE_CACHE;
+	else
+		lim.features &= ~BLK_FEAT_WRITE_CACHE;
+	blk_mq_freeze_queue(disk->queue);
+	i = queue_limits_commit_update(disk->queue, &lim);
+	blk_mq_unfreeze_queue(disk->queue);
+	if (i)
+		return i;
 	return count;
 }
 
@@ -1504,6 +1515,9 @@ static int virtblk_probe(struct virtio_device *vdev)
 	if (err)
 		goto out_free_tags;
 
+	if (virtblk_get_cache_mode(vdev))
+		lim.features |= BLK_FEAT_WRITE_CACHE;
+
 	vblk->disk = blk_mq_alloc_disk(&vblk->tag_set, &lim, vblk);
 	if (IS_ERR(vblk->disk)) {
 		err = PTR_ERR(vblk->disk);
@@ -1519,10 +1533,6 @@ static int virtblk_probe(struct virtio_device *vdev)
 	vblk->disk->fops = &virtblk_fops;
 	vblk->index = index;
 
-	/* configure queue flush support */
-	blk_queue_write_cache(vblk->disk->queue, virtblk_get_cache_mode(vdev),
-			false);
-
 	/* If disk is read-only in the host, the guest should obey */
 	if (virtio_has_feature(vdev, VIRTIO_BLK_F_RO))
 		set_disk_ro(vblk->disk, 1);

drivers/block/xen-blkfront.c

@@ -959,6 +959,12 @@ static void blkif_set_queue_limits(const struct blkfront_info *info,
 			lim->max_secure_erase_sectors = UINT_MAX;
 	}
 
+	if (info->feature_flush) {
+		lim->features |= BLK_FEAT_WRITE_CACHE;
+		if (info->feature_fua)
+			lim->features |= BLK_FEAT_FUA;
+	}
+
 	/* Hard sector size and max sectors impersonate the equiv. hardware. */
 	lim->logical_block_size = info->sector_size;
 	lim->physical_block_size = info->physical_sector_size;
@@ -987,8 +993,6 @@ static const char *flush_info(struct blkfront_info *info)
 
 static void xlvbd_flush(struct blkfront_info *info)
 {
-	blk_queue_write_cache(info->rq, info->feature_flush ? true : false,
-			      info->feature_fua ? true : false);
 	pr_info("blkfront: %s: %s %s %s %s %s %s %s\n",
 		info->gd->disk_name, flush_info(info),
 		"persistent grants:", info->feature_persistent ?

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:

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)
+/*
+ * 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)
 {
-	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)
-{
-	/*
-	 * 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);

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);