Merge tag 'for-5.1/dm-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm

Pull device mapper updates from Mike Snitzer:

 - Update bio-based DM core to always call blk_queue_split() and update
   DM targets to properly advertise discard limits that
   blk_queue_split() looks at when dtermining to split discard. Whereby
   allowing DM core's own 'split_discard_bios' to be removed.

 - Improve DM cache target to provide support for discard passdown to
   the origin device.

 - Introduce support to directly boot to a DM mapped device from init by
   using dm-mod.create= module param. This eliminates the need for an
   elaborate initramfs that is otherwise needed to create DM devices.

   This feature's implementation has been worked on for quite some time
   (got up to v12) and is of particular interest to Android and other
   more embedded platforms (e.g. ARM).

 - Rate limit errors from the DM integrity target that were identified
   as the cause for recent NMI hangs due to console limitations.

 - Add sanity checks for user input to thin-pool and external snapshot
   creation.

 - Remove some unused leftover kmem caches from when old .request_fn
   request-based support was removed.

 - Various small cleanups and fixes to targets (e.g. typos, needless
   unlikely() annotations, use struct_size(), remove needless
   .direct_access method from dm-snapshot)

* tag 'for-5.1/dm-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
  dm integrity: limit the rate of error messages
  dm snapshot: don't define direct_access if we don't support it
  dm cache: add support for discard passdown to the origin device
  dm writecache: fix typo in name for writeback_wq
  dm: add support to directly boot to a mapped device
  dm thin: add sanity checks to thin-pool and external snapshot creation
  dm block manager: remove redundant unlikely annotation
  dm verity fec: remove redundant unlikely annotation
  dm integrity: remove redundant unlikely annotation
  dm: always call blk_queue_split() in dm_process_bio()
  dm: fix to_sector() for 32bit
  dm switch: use struct_size() in kzalloc()
  dm: remove unused _rq_tio_cache and _rq_cache
  dm: eliminate 'split_discard_bios' flag from DM target interface
  dm: update dm_process_bio() to split bio if in ->make_request_fn()

1 files changed, 73 insertions, 66 deletions

diff --git a/drivers/md/dm.c b/drivers/md/dm.c
index 515e6af9bed2..68d24056d0b1 100644
--- a/drivers/md/dm.c
+++ b/drivers/md/dm.c
@@ -158,9 +158,6 @@ struct table_device {
 	struct dm_dev dm_dev;
 };
 
-static struct kmem_cache *_rq_tio_cache;
-static struct kmem_cache *_rq_cache;
-
 /*
  * Bio-based DM's mempools' reserved IOs set by the user.
  */
@@ -222,20 +219,11 @@ static unsigned dm_get_numa_node(void)
 
 static int __init local_init(void)
 {
-	int r = -ENOMEM;
-
-	_rq_tio_cache = KMEM_CACHE(dm_rq_target_io, 0);
-	if (!_rq_tio_cache)
-		return r;
-
-	_rq_cache = kmem_cache_create("dm_old_clone_request", sizeof(struct request),
-				      __alignof__(struct request), 0, NULL);
-	if (!_rq_cache)
-		goto out_free_rq_tio_cache;
+	int r;
 
 	r = dm_uevent_init();
 	if (r)
-		goto out_free_rq_cache;
+		return r;
 
 	deferred_remove_workqueue = alloc_workqueue("kdmremove", WQ_UNBOUND, 1);
 	if (!deferred_remove_workqueue) {
@@ -257,10 +245,6 @@ out_free_workqueue:
 	destroy_workqueue(deferred_remove_workqueue);
 out_uevent_exit:
 	dm_uevent_exit();
-out_free_rq_cache:
-	kmem_cache_destroy(_rq_cache);
-out_free_rq_tio_cache:
-	kmem_cache_destroy(_rq_tio_cache);
 
 	return r;
 }
@@ -270,8 +254,6 @@ static void local_exit(void)
 	flush_scheduled_work();
 	destroy_workqueue(deferred_remove_workqueue);
 
-	kmem_cache_destroy(_rq_cache);
-	kmem_cache_destroy(_rq_tio_cache);
 	unregister_blkdev(_major, _name);
 	dm_uevent_exit();
 
@@ -1478,17 +1460,10 @@ static unsigned get_num_write_zeroes_bios(struct dm_target *ti)
 	return ti->num_write_zeroes_bios;
 }
 
-typedef bool (*is_split_required_fn)(struct dm_target *ti);
-
-static bool is_split_required_for_discard(struct dm_target *ti)
-{
-	return ti->split_discard_bios;
-}
-
 static int __send_changing_extent_only(struct clone_info *ci, struct dm_target *ti,
-				       unsigned num_bios, bool is_split_required)
+				       unsigned num_bios)
 {
-	unsigned len;
+	unsigned len = ci->sector_count;
 
 	/*
 	 * Even though the device advertised support for this type of
@@ -1499,11 +1474,6 @@ static int __send_changing_extent_only(struct clone_info *ci, struct dm_target *
 	if (!num_bios)
 		return -EOPNOTSUPP;
 
-	if (!is_split_required)
-		len = min((sector_t)ci->sector_count, max_io_len_target_boundary(ci->sector, ti));
-	else
-		len = min((sector_t)ci->sector_count, max_io_len(ci->sector, ti));
-
 	__send_duplicate_bios(ci, ti, num_bios, &len);
 
 	ci->sector += len;
@@ -1514,23 +1484,38 @@ static int __send_changing_extent_only(struct clone_info *ci, struct dm_target *
 
 static int __send_discard(struct clone_info *ci, struct dm_target *ti)
 {
-	return __send_changing_extent_only(ci, ti, get_num_discard_bios(ti),
-					   is_split_required_for_discard(ti));
+	return __send_changing_extent_only(ci, ti, get_num_discard_bios(ti));
 }
 
 static int __send_secure_erase(struct clone_info *ci, struct dm_target *ti)
 {
-	return __send_changing_extent_only(ci, ti, get_num_secure_erase_bios(ti), false);
+	return __send_changing_extent_only(ci, ti, get_num_secure_erase_bios(ti));
 }
 
 static int __send_write_same(struct clone_info *ci, struct dm_target *ti)
 {
-	return __send_changing_extent_only(ci, ti, get_num_write_same_bios(ti), false);
+	return __send_changing_extent_only(ci, ti, get_num_write_same_bios(ti));
 }
 
 static int __send_write_zeroes(struct clone_info *ci, struct dm_target *ti)
 {
-	return __send_changing_extent_only(ci, ti, get_num_write_zeroes_bios(ti), false);
+	return __send_changing_extent_only(ci, ti, get_num_write_zeroes_bios(ti));
+}
+
+static bool is_abnormal_io(struct bio *bio)
+{
+	bool r = false;
+
+	switch (bio_op(bio)) {
+	case REQ_OP_DISCARD:
+	case REQ_OP_SECURE_ERASE:
+	case REQ_OP_WRITE_SAME:
+	case REQ_OP_WRITE_ZEROES:
+		r = true;
+		break;
+	}
+
+	return r;
 }
 
 static bool __process_abnormal_io(struct clone_info *ci, struct dm_target *ti,
@@ -1565,7 +1550,7 @@ static int __split_and_process_non_flush(struct clone_info *ci)
 	if (!dm_target_is_valid(ti))
 		return -EIO;
 
-	if (unlikely(__process_abnormal_io(ci, ti, &r)))
+	if (__process_abnormal_io(ci, ti, &r))
 		return r;
 
 	len = min_t(sector_t, max_io_len(ci->sector, ti), ci->sector_count);
@@ -1601,13 +1586,6 @@ static blk_qc_t __split_and_process_bio(struct mapped_device *md,
 	blk_qc_t ret = BLK_QC_T_NONE;
 	int error = 0;
 
-	if (unlikely(!map)) {
-		bio_io_error(bio);
-		return ret;
-	}
-
-	blk_queue_split(md->queue, &bio);
-
 	init_clone_info(&ci, md, map, bio);
 
 	if (bio->bi_opf & REQ_PREFLUSH) {
@@ -1675,18 +1653,13 @@ static blk_qc_t __split_and_process_bio(struct mapped_device *md,
  * Optimized variant of __split_and_process_bio that leverages the
  * fact that targets that use it do _not_ have a need to split bios.
  */
-static blk_qc_t __process_bio(struct mapped_device *md,
-			      struct dm_table *map, struct bio *bio)
+static blk_qc_t __process_bio(struct mapped_device *md, struct dm_table *map,
+			      struct bio *bio, struct dm_target *ti)
 {
 	struct clone_info ci;
 	blk_qc_t ret = BLK_QC_T_NONE;
 	int error = 0;
 
-	if (unlikely(!map)) {
-		bio_io_error(bio);
-		return ret;
-	}
-
 	init_clone_info(&ci, md, map, bio);
 
 	if (bio->bi_opf & REQ_PREFLUSH) {
@@ -1704,21 +1677,11 @@ static blk_qc_t __process_bio(struct mapped_device *md,
 		error = __send_empty_flush(&ci);
 		/* dec_pending submits any data associated with flush */
 	} else {
-		struct dm_target *ti = md->immutable_target;
 		struct dm_target_io *tio;
 
-		/*
-		 * Defend against IO still getting in during teardown
-		 * - as was seen for a time with nvme-fcloop
-		 */
-		if (WARN_ON_ONCE(!ti || !dm_target_is_valid(ti))) {
-			error = -EIO;
-			goto out;
-		}
-
 		ci.bio = bio;
 		ci.sector_count = bio_sectors(bio);
-		if (unlikely(__process_abnormal_io(&ci, ti, &error)))
+		if (__process_abnormal_io(&ci, ti, &error))
 			goto out;
 
 		tio = alloc_tio(&ci, ti, 0, GFP_NOIO);
@@ -1730,11 +1693,55 @@ out:
 	return ret;
 }
 
+static void dm_queue_split(struct mapped_device *md, struct dm_target *ti, struct bio **bio)
+{
+	unsigned len, sector_count;
+
+	sector_count = bio_sectors(*bio);
+	len = min_t(sector_t, max_io_len((*bio)->bi_iter.bi_sector, ti), sector_count);
+
+	if (sector_count > len) {
+		struct bio *split = bio_split(*bio, len, GFP_NOIO, &md->queue->bio_split);
+
+		bio_chain(split, *bio);
+		trace_block_split(md->queue, split, (*bio)->bi_iter.bi_sector);
+		generic_make_request(*bio);
+		*bio = split;
+	}
+}
+
 static blk_qc_t dm_process_bio(struct mapped_device *md,
 			       struct dm_table *map, struct bio *bio)
 {
+	blk_qc_t ret = BLK_QC_T_NONE;
+	struct dm_target *ti = md->immutable_target;
+
+	if (unlikely(!map)) {
+		bio_io_error(bio);
+		return ret;
+	}
+
+	if (!ti) {
+		ti = dm_table_find_target(map, bio->bi_iter.bi_sector);
+		if (unlikely(!ti || !dm_target_is_valid(ti))) {
+			bio_io_error(bio);
+			return ret;
+		}
+	}
+
+	/*
+	 * If in ->make_request_fn we need to use blk_queue_split(), otherwise
+	 * queue_limits for abnormal requests (e.g. discard, writesame, etc)
+	 * won't be imposed.
+	 */
+	if (current->bio_list) {
+		blk_queue_split(md->queue, &bio);
+		if (!is_abnormal_io(bio))
+			dm_queue_split(md, ti, &bio);
+	}
+
 	if (dm_get_md_type(md) == DM_TYPE_NVME_BIO_BASED)
-		return __process_bio(md, map, bio);
+		return __process_bio(md, map, bio, ti);
 	else
 		return __split_and_process_bio(md, map, bio);
 }