2 * Copyright (C) 2001 Sistina Software (UK) Limited.
3 * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
5 * This file is released under the GPL.
10 #include <linux/module.h>
11 #include <linux/vmalloc.h>
12 #include <linux/blkdev.h>
13 #include <linux/namei.h>
14 #include <linux/ctype.h>
15 #include <linux/string.h>
16 #include <linux/slab.h>
17 #include <linux/interrupt.h>
18 #include <linux/mutex.h>
19 #include <linux/delay.h>
20 #include <linux/atomic.h>
21 #include <linux/blk-mq.h>
22 #include <linux/mount.h>
23 #include <linux/dax.h>
25 #define DM_MSG_PREFIX "table"
28 #define NODE_SIZE L1_CACHE_BYTES
29 #define KEYS_PER_NODE (NODE_SIZE / sizeof(sector_t))
30 #define CHILDREN_PER_NODE (KEYS_PER_NODE + 1)
33 struct mapped_device *md;
34 enum dm_queue_mode type;
38 unsigned int counts[MAX_DEPTH]; /* in nodes */
39 sector_t *index[MAX_DEPTH];
41 unsigned int num_targets;
42 unsigned int num_allocated;
44 struct dm_target *targets;
46 struct target_type *immutable_target_type;
48 bool integrity_supported:1;
50 unsigned integrity_added:1;
53 * Indicates the rw permissions for the new logical
54 * device. This should be a combination of FMODE_READ
59 /* a list of devices used by this table */
60 struct list_head devices;
62 /* events get handed up using this callback */
63 void (*event_fn)(void *);
66 struct dm_md_mempools *mempools;
68 struct list_head target_callbacks;
72 * Similar to ceiling(log_size(n))
74 static unsigned int int_log(unsigned int n, unsigned int base)
79 n = dm_div_up(n, base);
87 * Calculate the index of the child node of the n'th node k'th key.
89 static inline unsigned int get_child(unsigned int n, unsigned int k)
91 return (n * CHILDREN_PER_NODE) + k;
95 * Return the n'th node of level l from table t.
97 static inline sector_t *get_node(struct dm_table *t,
98 unsigned int l, unsigned int n)
100 return t->index[l] + (n * KEYS_PER_NODE);
104 * Return the highest key that you could lookup from the n'th
105 * node on level l of the btree.
107 static sector_t high(struct dm_table *t, unsigned int l, unsigned int n)
109 for (; l < t->depth - 1; l++)
110 n = get_child(n, CHILDREN_PER_NODE - 1);
112 if (n >= t->counts[l])
113 return (sector_t) - 1;
115 return get_node(t, l, n)[KEYS_PER_NODE - 1];
119 * Fills in a level of the btree based on the highs of the level
122 static int setup_btree_index(unsigned int l, struct dm_table *t)
127 for (n = 0U; n < t->counts[l]; n++) {
128 node = get_node(t, l, n);
130 for (k = 0U; k < KEYS_PER_NODE; k++)
131 node[k] = high(t, l + 1, get_child(n, k));
137 void *dm_vcalloc(unsigned long nmemb, unsigned long elem_size)
143 * Check that we're not going to overflow.
145 if (nmemb > (ULONG_MAX / elem_size))
148 size = nmemb * elem_size;
149 addr = vzalloc(size);
153 EXPORT_SYMBOL(dm_vcalloc);
156 * highs, and targets are managed as dynamic arrays during a
159 static int alloc_targets(struct dm_table *t, unsigned int num)
162 struct dm_target *n_targets;
165 * Allocate both the target array and offset array at once.
166 * Append an empty entry to catch sectors beyond the end of
169 n_highs = (sector_t *) dm_vcalloc(num + 1, sizeof(struct dm_target) +
174 n_targets = (struct dm_target *) (n_highs + num);
176 memset(n_highs, -1, sizeof(*n_highs) * num);
179 t->num_allocated = num;
181 t->targets = n_targets;
186 int dm_table_create(struct dm_table **result, fmode_t mode,
187 unsigned num_targets, struct mapped_device *md)
189 struct dm_table *t = kzalloc(sizeof(*t), GFP_KERNEL);
194 INIT_LIST_HEAD(&t->devices);
195 INIT_LIST_HEAD(&t->target_callbacks);
198 num_targets = KEYS_PER_NODE;
200 num_targets = dm_round_up(num_targets, KEYS_PER_NODE);
207 if (alloc_targets(t, num_targets)) {
212 t->type = DM_TYPE_NONE;
219 static void free_devices(struct list_head *devices, struct mapped_device *md)
221 struct list_head *tmp, *next;
223 list_for_each_safe(tmp, next, devices) {
224 struct dm_dev_internal *dd =
225 list_entry(tmp, struct dm_dev_internal, list);
226 DMWARN("%s: dm_table_destroy: dm_put_device call missing for %s",
227 dm_device_name(md), dd->dm_dev->name);
228 dm_put_table_device(md, dd->dm_dev);
233 void dm_table_destroy(struct dm_table *t)
240 /* free the indexes */
242 vfree(t->index[t->depth - 2]);
244 /* free the targets */
245 for (i = 0; i < t->num_targets; i++) {
246 struct dm_target *tgt = t->targets + i;
251 dm_put_target_type(tgt->type);
256 /* free the device list */
257 free_devices(&t->devices, t->md);
259 dm_free_md_mempools(t->mempools);
265 * See if we've already got a device in the list.
267 static struct dm_dev_internal *find_device(struct list_head *l, dev_t dev)
269 struct dm_dev_internal *dd;
271 list_for_each_entry (dd, l, list)
272 if (dd->dm_dev->bdev->bd_dev == dev)
279 * If possible, this checks an area of a destination device is invalid.
281 static int device_area_is_invalid(struct dm_target *ti, struct dm_dev *dev,
282 sector_t start, sector_t len, void *data)
284 struct request_queue *q;
285 struct queue_limits *limits = data;
286 struct block_device *bdev = dev->bdev;
288 i_size_read(bdev->bd_inode) >> SECTOR_SHIFT;
289 unsigned short logical_block_size_sectors =
290 limits->logical_block_size >> SECTOR_SHIFT;
291 char b[BDEVNAME_SIZE];
294 * Some devices exist without request functions,
295 * such as loop devices not yet bound to backing files.
296 * Forbid the use of such devices.
298 q = bdev_get_queue(bdev);
299 if (!q || !q->make_request_fn) {
300 DMWARN("%s: %s is not yet initialised: "
301 "start=%llu, len=%llu, dev_size=%llu",
302 dm_device_name(ti->table->md), bdevname(bdev, b),
303 (unsigned long long)start,
304 (unsigned long long)len,
305 (unsigned long long)dev_size);
312 if ((start >= dev_size) || (start + len > dev_size)) {
313 DMWARN("%s: %s too small for target: "
314 "start=%llu, len=%llu, dev_size=%llu",
315 dm_device_name(ti->table->md), bdevname(bdev, b),
316 (unsigned long long)start,
317 (unsigned long long)len,
318 (unsigned long long)dev_size);
323 * If the target is mapped to zoned block device(s), check
324 * that the zones are not partially mapped.
326 if (bdev_zoned_model(bdev) != BLK_ZONED_NONE) {
327 unsigned int zone_sectors = bdev_zone_sectors(bdev);
329 if (start & (zone_sectors - 1)) {
330 DMWARN("%s: start=%llu not aligned to h/w zone size %u of %s",
331 dm_device_name(ti->table->md),
332 (unsigned long long)start,
333 zone_sectors, bdevname(bdev, b));
338 * Note: The last zone of a zoned block device may be smaller
339 * than other zones. So for a target mapping the end of a
340 * zoned block device with such a zone, len would not be zone
341 * aligned. We do not allow such last smaller zone to be part
342 * of the mapping here to ensure that mappings with multiple
343 * devices do not end up with a smaller zone in the middle of
346 if (len & (zone_sectors - 1)) {
347 DMWARN("%s: len=%llu not aligned to h/w zone size %u of %s",
348 dm_device_name(ti->table->md),
349 (unsigned long long)len,
350 zone_sectors, bdevname(bdev, b));
355 if (logical_block_size_sectors <= 1)
358 if (start & (logical_block_size_sectors - 1)) {
359 DMWARN("%s: start=%llu not aligned to h/w "
360 "logical block size %u of %s",
361 dm_device_name(ti->table->md),
362 (unsigned long long)start,
363 limits->logical_block_size, bdevname(bdev, b));
367 if (len & (logical_block_size_sectors - 1)) {
368 DMWARN("%s: len=%llu not aligned to h/w "
369 "logical block size %u of %s",
370 dm_device_name(ti->table->md),
371 (unsigned long long)len,
372 limits->logical_block_size, bdevname(bdev, b));
380 * This upgrades the mode on an already open dm_dev, being
381 * careful to leave things as they were if we fail to reopen the
382 * device and not to touch the existing bdev field in case
383 * it is accessed concurrently inside dm_table_any_congested().
385 static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode,
386 struct mapped_device *md)
389 struct dm_dev *old_dev, *new_dev;
391 old_dev = dd->dm_dev;
393 r = dm_get_table_device(md, dd->dm_dev->bdev->bd_dev,
394 dd->dm_dev->mode | new_mode, &new_dev);
398 dd->dm_dev = new_dev;
399 dm_put_table_device(md, old_dev);
405 * Convert the path to a device
407 dev_t dm_get_dev_t(const char *path)
410 struct block_device *bdev;
412 bdev = lookup_bdev(path);
414 dev = name_to_dev_t(path);
422 EXPORT_SYMBOL_GPL(dm_get_dev_t);
425 * Add a device to the list, or just increment the usage count if
426 * it's already present.
428 int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode,
429 struct dm_dev **result)
433 struct dm_dev_internal *dd;
434 struct dm_table *t = ti->table;
438 dev = dm_get_dev_t(path);
442 dd = find_device(&t->devices, dev);
444 dd = kmalloc(sizeof(*dd), GFP_KERNEL);
448 if ((r = dm_get_table_device(t->md, dev, mode, &dd->dm_dev))) {
453 refcount_set(&dd->count, 1);
454 list_add(&dd->list, &t->devices);
457 } else if (dd->dm_dev->mode != (mode | dd->dm_dev->mode)) {
458 r = upgrade_mode(dd, mode, t->md);
462 refcount_inc(&dd->count);
464 *result = dd->dm_dev;
467 EXPORT_SYMBOL(dm_get_device);
469 static int dm_set_device_limits(struct dm_target *ti, struct dm_dev *dev,
470 sector_t start, sector_t len, void *data)
472 struct queue_limits *limits = data;
473 struct block_device *bdev = dev->bdev;
474 struct request_queue *q = bdev_get_queue(bdev);
475 char b[BDEVNAME_SIZE];
478 DMWARN("%s: Cannot set limits for nonexistent device %s",
479 dm_device_name(ti->table->md), bdevname(bdev, b));
483 if (bdev_stack_limits(limits, bdev, start) < 0)
484 DMWARN("%s: adding target device %s caused an alignment inconsistency: "
485 "physical_block_size=%u, logical_block_size=%u, "
486 "alignment_offset=%u, start=%llu",
487 dm_device_name(ti->table->md), bdevname(bdev, b),
488 q->limits.physical_block_size,
489 q->limits.logical_block_size,
490 q->limits.alignment_offset,
491 (unsigned long long) start << SECTOR_SHIFT);
493 limits->zoned = blk_queue_zoned_model(q);
499 * Decrement a device's use count and remove it if necessary.
501 void dm_put_device(struct dm_target *ti, struct dm_dev *d)
504 struct list_head *devices = &ti->table->devices;
505 struct dm_dev_internal *dd;
507 list_for_each_entry(dd, devices, list) {
508 if (dd->dm_dev == d) {
514 DMWARN("%s: device %s not in table devices list",
515 dm_device_name(ti->table->md), d->name);
518 if (refcount_dec_and_test(&dd->count)) {
519 dm_put_table_device(ti->table->md, d);
524 EXPORT_SYMBOL(dm_put_device);
527 * Checks to see if the target joins onto the end of the table.
529 static int adjoin(struct dm_table *table, struct dm_target *ti)
531 struct dm_target *prev;
533 if (!table->num_targets)
536 prev = &table->targets[table->num_targets - 1];
537 return (ti->begin == (prev->begin + prev->len));
541 * Used to dynamically allocate the arg array.
543 * We do first allocation with GFP_NOIO because dm-mpath and dm-thin must
544 * process messages even if some device is suspended. These messages have a
545 * small fixed number of arguments.
547 * On the other hand, dm-switch needs to process bulk data using messages and
548 * excessive use of GFP_NOIO could cause trouble.
550 static char **realloc_argv(unsigned *size, char **old_argv)
557 new_size = *size * 2;
563 argv = kmalloc_array(new_size, sizeof(*argv), gfp);
565 memcpy(argv, old_argv, *size * sizeof(*argv));
574 * Destructively splits up the argument list to pass to ctr.
576 int dm_split_args(int *argc, char ***argvp, char *input)
578 char *start, *end = input, *out, **argv = NULL;
579 unsigned array_size = 0;
588 argv = realloc_argv(&array_size, argv);
593 /* Skip whitespace */
594 start = skip_spaces(end);
597 break; /* success, we hit the end */
599 /* 'out' is used to remove any back-quotes */
602 /* Everything apart from '\0' can be quoted */
603 if (*end == '\\' && *(end + 1)) {
610 break; /* end of token */
615 /* have we already filled the array ? */
616 if ((*argc + 1) > array_size) {
617 argv = realloc_argv(&array_size, argv);
622 /* we know this is whitespace */
626 /* terminate the string and put it in the array */
637 * Impose necessary and sufficient conditions on a devices's table such
638 * that any incoming bio which respects its logical_block_size can be
639 * processed successfully. If it falls across the boundary between
640 * two or more targets, the size of each piece it gets split into must
641 * be compatible with the logical_block_size of the target processing it.
643 static int validate_hardware_logical_block_alignment(struct dm_table *table,
644 struct queue_limits *limits)
647 * This function uses arithmetic modulo the logical_block_size
648 * (in units of 512-byte sectors).
650 unsigned short device_logical_block_size_sects =
651 limits->logical_block_size >> SECTOR_SHIFT;
654 * Offset of the start of the next table entry, mod logical_block_size.
656 unsigned short next_target_start = 0;
659 * Given an aligned bio that extends beyond the end of a
660 * target, how many sectors must the next target handle?
662 unsigned short remaining = 0;
664 struct dm_target *uninitialized_var(ti);
665 struct queue_limits ti_limits;
669 * Check each entry in the table in turn.
671 for (i = 0; i < dm_table_get_num_targets(table); i++) {
672 ti = dm_table_get_target(table, i);
674 blk_set_stacking_limits(&ti_limits);
676 /* combine all target devices' limits */
677 if (ti->type->iterate_devices)
678 ti->type->iterate_devices(ti, dm_set_device_limits,
682 * If the remaining sectors fall entirely within this
683 * table entry are they compatible with its logical_block_size?
685 if (remaining < ti->len &&
686 remaining & ((ti_limits.logical_block_size >>
691 (unsigned short) ((next_target_start + ti->len) &
692 (device_logical_block_size_sects - 1));
693 remaining = next_target_start ?
694 device_logical_block_size_sects - next_target_start : 0;
698 DMWARN("%s: table line %u (start sect %llu len %llu) "
699 "not aligned to h/w logical block size %u",
700 dm_device_name(table->md), i,
701 (unsigned long long) ti->begin,
702 (unsigned long long) ti->len,
703 limits->logical_block_size);
710 int dm_table_add_target(struct dm_table *t, const char *type,
711 sector_t start, sector_t len, char *params)
713 int r = -EINVAL, argc;
715 struct dm_target *tgt;
718 DMERR("%s: target type %s must appear alone in table",
719 dm_device_name(t->md), t->targets->type->name);
723 BUG_ON(t->num_targets >= t->num_allocated);
725 tgt = t->targets + t->num_targets;
726 memset(tgt, 0, sizeof(*tgt));
729 DMERR("%s: zero-length target", dm_device_name(t->md));
733 tgt->type = dm_get_target_type(type);
735 DMERR("%s: %s: unknown target type", dm_device_name(t->md), type);
739 if (dm_target_needs_singleton(tgt->type)) {
740 if (t->num_targets) {
741 tgt->error = "singleton target type must appear alone in table";
747 if (dm_target_always_writeable(tgt->type) && !(t->mode & FMODE_WRITE)) {
748 tgt->error = "target type may not be included in a read-only table";
752 if (t->immutable_target_type) {
753 if (t->immutable_target_type != tgt->type) {
754 tgt->error = "immutable target type cannot be mixed with other target types";
757 } else if (dm_target_is_immutable(tgt->type)) {
758 if (t->num_targets) {
759 tgt->error = "immutable target type cannot be mixed with other target types";
762 t->immutable_target_type = tgt->type;
765 if (dm_target_has_integrity(tgt->type))
766 t->integrity_added = 1;
771 tgt->error = "Unknown error";
774 * Does this target adjoin the previous one ?
776 if (!adjoin(t, tgt)) {
777 tgt->error = "Gap in table";
781 r = dm_split_args(&argc, &argv, params);
783 tgt->error = "couldn't split parameters (insufficient memory)";
787 r = tgt->type->ctr(tgt, argc, argv);
792 t->highs[t->num_targets++] = tgt->begin + tgt->len - 1;
794 if (!tgt->num_discard_bios && tgt->discards_supported)
795 DMWARN("%s: %s: ignoring discards_supported because num_discard_bios is zero.",
796 dm_device_name(t->md), type);
801 DMERR("%s: %s: %s", dm_device_name(t->md), type, tgt->error);
802 dm_put_target_type(tgt->type);
807 * Target argument parsing helpers.
809 static int validate_next_arg(const struct dm_arg *arg,
810 struct dm_arg_set *arg_set,
811 unsigned *value, char **error, unsigned grouped)
813 const char *arg_str = dm_shift_arg(arg_set);
817 (sscanf(arg_str, "%u%c", value, &dummy) != 1) ||
818 (*value < arg->min) ||
819 (*value > arg->max) ||
820 (grouped && arg_set->argc < *value)) {
828 int dm_read_arg(const struct dm_arg *arg, struct dm_arg_set *arg_set,
829 unsigned *value, char **error)
831 return validate_next_arg(arg, arg_set, value, error, 0);
833 EXPORT_SYMBOL(dm_read_arg);
835 int dm_read_arg_group(const struct dm_arg *arg, struct dm_arg_set *arg_set,
836 unsigned *value, char **error)
838 return validate_next_arg(arg, arg_set, value, error, 1);
840 EXPORT_SYMBOL(dm_read_arg_group);
842 const char *dm_shift_arg(struct dm_arg_set *as)
855 EXPORT_SYMBOL(dm_shift_arg);
857 void dm_consume_args(struct dm_arg_set *as, unsigned num_args)
859 BUG_ON(as->argc < num_args);
860 as->argc -= num_args;
861 as->argv += num_args;
863 EXPORT_SYMBOL(dm_consume_args);
865 static bool __table_type_bio_based(enum dm_queue_mode table_type)
867 return (table_type == DM_TYPE_BIO_BASED ||
868 table_type == DM_TYPE_DAX_BIO_BASED ||
869 table_type == DM_TYPE_NVME_BIO_BASED);
872 static bool __table_type_request_based(enum dm_queue_mode table_type)
874 return table_type == DM_TYPE_REQUEST_BASED;
877 void dm_table_set_type(struct dm_table *t, enum dm_queue_mode type)
881 EXPORT_SYMBOL_GPL(dm_table_set_type);
883 static int device_supports_dax(struct dm_target *ti, struct dm_dev *dev,
884 sector_t start, sector_t len, void *data)
886 return bdev_dax_supported(dev->bdev, PAGE_SIZE);
889 static bool dm_table_supports_dax(struct dm_table *t)
891 struct dm_target *ti;
894 /* Ensure that all targets support DAX. */
895 for (i = 0; i < dm_table_get_num_targets(t); i++) {
896 ti = dm_table_get_target(t, i);
898 if (!ti->type->direct_access)
901 if (!ti->type->iterate_devices ||
902 !ti->type->iterate_devices(ti, device_supports_dax, NULL))
909 static bool dm_table_does_not_support_partial_completion(struct dm_table *t);
911 static int device_is_rq_based(struct dm_target *ti, struct dm_dev *dev,
912 sector_t start, sector_t len, void *data)
914 struct request_queue *q = bdev_get_queue(dev->bdev);
916 return queue_is_rq_based(q);
919 static int dm_table_determine_type(struct dm_table *t)
922 unsigned bio_based = 0, request_based = 0, hybrid = 0;
923 struct dm_target *tgt;
924 struct list_head *devices = dm_table_get_devices(t);
925 enum dm_queue_mode live_md_type = dm_get_md_type(t->md);
927 if (t->type != DM_TYPE_NONE) {
928 /* target already set the table's type */
929 if (t->type == DM_TYPE_BIO_BASED) {
930 /* possibly upgrade to a variant of bio-based */
931 goto verify_bio_based;
933 BUG_ON(t->type == DM_TYPE_DAX_BIO_BASED);
934 BUG_ON(t->type == DM_TYPE_NVME_BIO_BASED);
935 goto verify_rq_based;
938 for (i = 0; i < t->num_targets; i++) {
939 tgt = t->targets + i;
940 if (dm_target_hybrid(tgt))
942 else if (dm_target_request_based(tgt))
947 if (bio_based && request_based) {
948 DMERR("Inconsistent table: different target types"
949 " can't be mixed up");
954 if (hybrid && !bio_based && !request_based) {
956 * The targets can work either way.
957 * Determine the type from the live device.
958 * Default to bio-based if device is new.
960 if (__table_type_request_based(live_md_type))
968 /* We must use this table as bio-based */
969 t->type = DM_TYPE_BIO_BASED;
970 if (dm_table_supports_dax(t) ||
971 (list_empty(devices) && live_md_type == DM_TYPE_DAX_BIO_BASED)) {
972 t->type = DM_TYPE_DAX_BIO_BASED;
974 /* Check if upgrading to NVMe bio-based is valid or required */
975 tgt = dm_table_get_immutable_target(t);
976 if (tgt && !tgt->max_io_len && dm_table_does_not_support_partial_completion(t)) {
977 t->type = DM_TYPE_NVME_BIO_BASED;
978 goto verify_rq_based; /* must be stacked directly on NVMe (blk-mq) */
979 } else if (list_empty(devices) && live_md_type == DM_TYPE_NVME_BIO_BASED) {
980 t->type = DM_TYPE_NVME_BIO_BASED;
986 BUG_ON(!request_based); /* No targets in this table */
988 t->type = DM_TYPE_REQUEST_BASED;
992 * Request-based dm supports only tables that have a single target now.
993 * To support multiple targets, request splitting support is needed,
994 * and that needs lots of changes in the block-layer.
995 * (e.g. request completion process for partial completion.)
997 if (t->num_targets > 1) {
998 DMERR("%s DM doesn't support multiple targets",
999 t->type == DM_TYPE_NVME_BIO_BASED ? "nvme bio-based" : "request-based");
1003 if (list_empty(devices)) {
1005 struct dm_table *live_table = dm_get_live_table(t->md, &srcu_idx);
1007 /* inherit live table's type */
1009 t->type = live_table->type;
1010 dm_put_live_table(t->md, srcu_idx);
1014 tgt = dm_table_get_immutable_target(t);
1016 DMERR("table load rejected: immutable target is required");
1018 } else if (tgt->max_io_len) {
1019 DMERR("table load rejected: immutable target that splits IO is not supported");
1023 /* Non-request-stackable devices can't be used for request-based dm */
1024 if (!tgt->type->iterate_devices ||
1025 !tgt->type->iterate_devices(tgt, device_is_rq_based, NULL)) {
1026 DMERR("table load rejected: including non-request-stackable devices");
1033 enum dm_queue_mode dm_table_get_type(struct dm_table *t)
1038 struct target_type *dm_table_get_immutable_target_type(struct dm_table *t)
1040 return t->immutable_target_type;
1043 struct dm_target *dm_table_get_immutable_target(struct dm_table *t)
1045 /* Immutable target is implicitly a singleton */
1046 if (t->num_targets > 1 ||
1047 !dm_target_is_immutable(t->targets[0].type))
1053 struct dm_target *dm_table_get_wildcard_target(struct dm_table *t)
1055 struct dm_target *ti;
1058 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1059 ti = dm_table_get_target(t, i);
1060 if (dm_target_is_wildcard(ti->type))
1067 bool dm_table_bio_based(struct dm_table *t)
1069 return __table_type_bio_based(dm_table_get_type(t));
1072 bool dm_table_request_based(struct dm_table *t)
1074 return __table_type_request_based(dm_table_get_type(t));
1077 static int dm_table_alloc_md_mempools(struct dm_table *t, struct mapped_device *md)
1079 enum dm_queue_mode type = dm_table_get_type(t);
1080 unsigned per_io_data_size = 0;
1081 unsigned min_pool_size = 0;
1082 struct dm_target *ti;
1085 if (unlikely(type == DM_TYPE_NONE)) {
1086 DMWARN("no table type is set, can't allocate mempools");
1090 if (__table_type_bio_based(type))
1091 for (i = 0; i < t->num_targets; i++) {
1092 ti = t->targets + i;
1093 per_io_data_size = max(per_io_data_size, ti->per_io_data_size);
1094 min_pool_size = max(min_pool_size, ti->num_flush_bios);
1097 t->mempools = dm_alloc_md_mempools(md, type, t->integrity_supported,
1098 per_io_data_size, min_pool_size);
1105 void dm_table_free_md_mempools(struct dm_table *t)
1107 dm_free_md_mempools(t->mempools);
1111 struct dm_md_mempools *dm_table_get_md_mempools(struct dm_table *t)
1116 static int setup_indexes(struct dm_table *t)
1119 unsigned int total = 0;
1122 /* allocate the space for *all* the indexes */
1123 for (i = t->depth - 2; i >= 0; i--) {
1124 t->counts[i] = dm_div_up(t->counts[i + 1], CHILDREN_PER_NODE);
1125 total += t->counts[i];
1128 indexes = (sector_t *) dm_vcalloc(total, (unsigned long) NODE_SIZE);
1132 /* set up internal nodes, bottom-up */
1133 for (i = t->depth - 2; i >= 0; i--) {
1134 t->index[i] = indexes;
1135 indexes += (KEYS_PER_NODE * t->counts[i]);
1136 setup_btree_index(i, t);
1143 * Builds the btree to index the map.
1145 static int dm_table_build_index(struct dm_table *t)
1148 unsigned int leaf_nodes;
1150 /* how many indexes will the btree have ? */
1151 leaf_nodes = dm_div_up(t->num_targets, KEYS_PER_NODE);
1152 t->depth = 1 + int_log(leaf_nodes, CHILDREN_PER_NODE);
1154 /* leaf layer has already been set up */
1155 t->counts[t->depth - 1] = leaf_nodes;
1156 t->index[t->depth - 1] = t->highs;
1159 r = setup_indexes(t);
1164 static bool integrity_profile_exists(struct gendisk *disk)
1166 return !!blk_get_integrity(disk);
1170 * Get a disk whose integrity profile reflects the table's profile.
1171 * Returns NULL if integrity support was inconsistent or unavailable.
1173 static struct gendisk * dm_table_get_integrity_disk(struct dm_table *t)
1175 struct list_head *devices = dm_table_get_devices(t);
1176 struct dm_dev_internal *dd = NULL;
1177 struct gendisk *prev_disk = NULL, *template_disk = NULL;
1180 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1181 struct dm_target *ti = dm_table_get_target(t, i);
1182 if (!dm_target_passes_integrity(ti->type))
1186 list_for_each_entry(dd, devices, list) {
1187 template_disk = dd->dm_dev->bdev->bd_disk;
1188 if (!integrity_profile_exists(template_disk))
1190 else if (prev_disk &&
1191 blk_integrity_compare(prev_disk, template_disk) < 0)
1193 prev_disk = template_disk;
1196 return template_disk;
1200 DMWARN("%s: integrity not set: %s and %s profile mismatch",
1201 dm_device_name(t->md),
1202 prev_disk->disk_name,
1203 template_disk->disk_name);
1208 * Register the mapped device for blk_integrity support if the
1209 * underlying devices have an integrity profile. But all devices may
1210 * not have matching profiles (checking all devices isn't reliable
1211 * during table load because this table may use other DM device(s) which
1212 * must be resumed before they will have an initialized integity
1213 * profile). Consequently, stacked DM devices force a 2 stage integrity
1214 * profile validation: First pass during table load, final pass during
1217 static int dm_table_register_integrity(struct dm_table *t)
1219 struct mapped_device *md = t->md;
1220 struct gendisk *template_disk = NULL;
1222 /* If target handles integrity itself do not register it here. */
1223 if (t->integrity_added)
1226 template_disk = dm_table_get_integrity_disk(t);
1230 if (!integrity_profile_exists(dm_disk(md))) {
1231 t->integrity_supported = true;
1233 * Register integrity profile during table load; we can do
1234 * this because the final profile must match during resume.
1236 blk_integrity_register(dm_disk(md),
1237 blk_get_integrity(template_disk));
1242 * If DM device already has an initialized integrity
1243 * profile the new profile should not conflict.
1245 if (blk_integrity_compare(dm_disk(md), template_disk) < 0) {
1246 DMWARN("%s: conflict with existing integrity profile: "
1247 "%s profile mismatch",
1248 dm_device_name(t->md),
1249 template_disk->disk_name);
1253 /* Preserve existing integrity profile */
1254 t->integrity_supported = true;
1259 * Prepares the table for use by building the indices,
1260 * setting the type, and allocating mempools.
1262 int dm_table_complete(struct dm_table *t)
1266 r = dm_table_determine_type(t);
1268 DMERR("unable to determine table type");
1272 r = dm_table_build_index(t);
1274 DMERR("unable to build btrees");
1278 r = dm_table_register_integrity(t);
1280 DMERR("could not register integrity profile.");
1284 r = dm_table_alloc_md_mempools(t, t->md);
1286 DMERR("unable to allocate mempools");
1291 static DEFINE_MUTEX(_event_lock);
1292 void dm_table_event_callback(struct dm_table *t,
1293 void (*fn)(void *), void *context)
1295 mutex_lock(&_event_lock);
1297 t->event_context = context;
1298 mutex_unlock(&_event_lock);
1301 void dm_table_event(struct dm_table *t)
1304 * You can no longer call dm_table_event() from interrupt
1305 * context, use a bottom half instead.
1307 BUG_ON(in_interrupt());
1309 mutex_lock(&_event_lock);
1311 t->event_fn(t->event_context);
1312 mutex_unlock(&_event_lock);
1314 EXPORT_SYMBOL(dm_table_event);
1316 sector_t dm_table_get_size(struct dm_table *t)
1318 return t->num_targets ? (t->highs[t->num_targets - 1] + 1) : 0;
1320 EXPORT_SYMBOL(dm_table_get_size);
1322 struct dm_target *dm_table_get_target(struct dm_table *t, unsigned int index)
1324 if (index >= t->num_targets)
1327 return t->targets + index;
1331 * Search the btree for the correct target.
1333 * Caller should check returned pointer with dm_target_is_valid()
1334 * to trap I/O beyond end of device.
1336 struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector)
1338 unsigned int l, n = 0, k = 0;
1341 for (l = 0; l < t->depth; l++) {
1342 n = get_child(n, k);
1343 node = get_node(t, l, n);
1345 for (k = 0; k < KEYS_PER_NODE; k++)
1346 if (node[k] >= sector)
1350 return &t->targets[(KEYS_PER_NODE * n) + k];
1353 static int count_device(struct dm_target *ti, struct dm_dev *dev,
1354 sector_t start, sector_t len, void *data)
1356 unsigned *num_devices = data;
1364 * Check whether a table has no data devices attached using each
1365 * target's iterate_devices method.
1366 * Returns false if the result is unknown because a target doesn't
1367 * support iterate_devices.
1369 bool dm_table_has_no_data_devices(struct dm_table *table)
1371 struct dm_target *ti;
1372 unsigned i, num_devices;
1374 for (i = 0; i < dm_table_get_num_targets(table); i++) {
1375 ti = dm_table_get_target(table, i);
1377 if (!ti->type->iterate_devices)
1381 ti->type->iterate_devices(ti, count_device, &num_devices);
1389 static int device_is_zoned_model(struct dm_target *ti, struct dm_dev *dev,
1390 sector_t start, sector_t len, void *data)
1392 struct request_queue *q = bdev_get_queue(dev->bdev);
1393 enum blk_zoned_model *zoned_model = data;
1395 return q && blk_queue_zoned_model(q) == *zoned_model;
1398 static bool dm_table_supports_zoned_model(struct dm_table *t,
1399 enum blk_zoned_model zoned_model)
1401 struct dm_target *ti;
1404 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1405 ti = dm_table_get_target(t, i);
1407 if (zoned_model == BLK_ZONED_HM &&
1408 !dm_target_supports_zoned_hm(ti->type))
1411 if (!ti->type->iterate_devices ||
1412 !ti->type->iterate_devices(ti, device_is_zoned_model, &zoned_model))
1419 static int device_matches_zone_sectors(struct dm_target *ti, struct dm_dev *dev,
1420 sector_t start, sector_t len, void *data)
1422 struct request_queue *q = bdev_get_queue(dev->bdev);
1423 unsigned int *zone_sectors = data;
1425 return q && blk_queue_zone_sectors(q) == *zone_sectors;
1428 static bool dm_table_matches_zone_sectors(struct dm_table *t,
1429 unsigned int zone_sectors)
1431 struct dm_target *ti;
1434 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1435 ti = dm_table_get_target(t, i);
1437 if (!ti->type->iterate_devices ||
1438 !ti->type->iterate_devices(ti, device_matches_zone_sectors, &zone_sectors))
1445 static int validate_hardware_zoned_model(struct dm_table *table,
1446 enum blk_zoned_model zoned_model,
1447 unsigned int zone_sectors)
1449 if (zoned_model == BLK_ZONED_NONE)
1452 if (!dm_table_supports_zoned_model(table, zoned_model)) {
1453 DMERR("%s: zoned model is not consistent across all devices",
1454 dm_device_name(table->md));
1458 /* Check zone size validity and compatibility */
1459 if (!zone_sectors || !is_power_of_2(zone_sectors))
1462 if (!dm_table_matches_zone_sectors(table, zone_sectors)) {
1463 DMERR("%s: zone sectors is not consistent across all devices",
1464 dm_device_name(table->md));
1472 * Establish the new table's queue_limits and validate them.
1474 int dm_calculate_queue_limits(struct dm_table *table,
1475 struct queue_limits *limits)
1477 struct dm_target *ti;
1478 struct queue_limits ti_limits;
1480 enum blk_zoned_model zoned_model = BLK_ZONED_NONE;
1481 unsigned int zone_sectors = 0;
1483 blk_set_stacking_limits(limits);
1485 for (i = 0; i < dm_table_get_num_targets(table); i++) {
1486 blk_set_stacking_limits(&ti_limits);
1488 ti = dm_table_get_target(table, i);
1490 if (!ti->type->iterate_devices)
1491 goto combine_limits;
1494 * Combine queue limits of all the devices this target uses.
1496 ti->type->iterate_devices(ti, dm_set_device_limits,
1499 if (zoned_model == BLK_ZONED_NONE && ti_limits.zoned != BLK_ZONED_NONE) {
1501 * After stacking all limits, validate all devices
1502 * in table support this zoned model and zone sectors.
1504 zoned_model = ti_limits.zoned;
1505 zone_sectors = ti_limits.chunk_sectors;
1508 /* Set I/O hints portion of queue limits */
1509 if (ti->type->io_hints)
1510 ti->type->io_hints(ti, &ti_limits);
1513 * Check each device area is consistent with the target's
1514 * overall queue limits.
1516 if (ti->type->iterate_devices(ti, device_area_is_invalid,
1522 * Merge this target's queue limits into the overall limits
1525 if (blk_stack_limits(limits, &ti_limits, 0) < 0)
1526 DMWARN("%s: adding target device "
1527 "(start sect %llu len %llu) "
1528 "caused an alignment inconsistency",
1529 dm_device_name(table->md),
1530 (unsigned long long) ti->begin,
1531 (unsigned long long) ti->len);
1534 * FIXME: this should likely be moved to blk_stack_limits(), would
1535 * also eliminate limits->zoned stacking hack in dm_set_device_limits()
1537 if (limits->zoned == BLK_ZONED_NONE && ti_limits.zoned != BLK_ZONED_NONE) {
1539 * By default, the stacked limits zoned model is set to
1540 * BLK_ZONED_NONE in blk_set_stacking_limits(). Update
1541 * this model using the first target model reported
1542 * that is not BLK_ZONED_NONE. This will be either the
1543 * first target device zoned model or the model reported
1544 * by the target .io_hints.
1546 limits->zoned = ti_limits.zoned;
1551 * Verify that the zoned model and zone sectors, as determined before
1552 * any .io_hints override, are the same across all devices in the table.
1553 * - this is especially relevant if .io_hints is emulating a disk-managed
1554 * zoned model (aka BLK_ZONED_NONE) on host-managed zoned block devices.
1557 if (limits->zoned != BLK_ZONED_NONE) {
1559 * ...IF the above limits stacking determined a zoned model
1560 * validate that all of the table's devices conform to it.
1562 zoned_model = limits->zoned;
1563 zone_sectors = limits->chunk_sectors;
1565 if (validate_hardware_zoned_model(table, zoned_model, zone_sectors))
1568 return validate_hardware_logical_block_alignment(table, limits);
1572 * Verify that all devices have an integrity profile that matches the
1573 * DM device's registered integrity profile. If the profiles don't
1574 * match then unregister the DM device's integrity profile.
1576 static void dm_table_verify_integrity(struct dm_table *t)
1578 struct gendisk *template_disk = NULL;
1580 if (t->integrity_added)
1583 if (t->integrity_supported) {
1585 * Verify that the original integrity profile
1586 * matches all the devices in this table.
1588 template_disk = dm_table_get_integrity_disk(t);
1589 if (template_disk &&
1590 blk_integrity_compare(dm_disk(t->md), template_disk) >= 0)
1594 if (integrity_profile_exists(dm_disk(t->md))) {
1595 DMWARN("%s: unable to establish an integrity profile",
1596 dm_device_name(t->md));
1597 blk_integrity_unregister(dm_disk(t->md));
1601 static int device_flush_capable(struct dm_target *ti, struct dm_dev *dev,
1602 sector_t start, sector_t len, void *data)
1604 unsigned long flush = (unsigned long) data;
1605 struct request_queue *q = bdev_get_queue(dev->bdev);
1607 return q && (q->queue_flags & flush);
1610 static bool dm_table_supports_flush(struct dm_table *t, unsigned long flush)
1612 struct dm_target *ti;
1616 * Require at least one underlying device to support flushes.
1617 * t->devices includes internal dm devices such as mirror logs
1618 * so we need to use iterate_devices here, which targets
1619 * supporting flushes must provide.
1621 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1622 ti = dm_table_get_target(t, i);
1624 if (!ti->num_flush_bios)
1627 if (ti->flush_supported)
1630 if (ti->type->iterate_devices &&
1631 ti->type->iterate_devices(ti, device_flush_capable, (void *) flush))
1638 static int device_dax_write_cache_enabled(struct dm_target *ti,
1639 struct dm_dev *dev, sector_t start,
1640 sector_t len, void *data)
1642 struct dax_device *dax_dev = dev->dax_dev;
1647 if (dax_write_cache_enabled(dax_dev))
1652 static int dm_table_supports_dax_write_cache(struct dm_table *t)
1654 struct dm_target *ti;
1657 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1658 ti = dm_table_get_target(t, i);
1660 if (ti->type->iterate_devices &&
1661 ti->type->iterate_devices(ti,
1662 device_dax_write_cache_enabled, NULL))
1669 static int device_is_nonrot(struct dm_target *ti, struct dm_dev *dev,
1670 sector_t start, sector_t len, void *data)
1672 struct request_queue *q = bdev_get_queue(dev->bdev);
1674 return q && blk_queue_nonrot(q);
1677 static int device_is_not_random(struct dm_target *ti, struct dm_dev *dev,
1678 sector_t start, sector_t len, void *data)
1680 struct request_queue *q = bdev_get_queue(dev->bdev);
1682 return q && !blk_queue_add_random(q);
1685 static int queue_supports_sg_merge(struct dm_target *ti, struct dm_dev *dev,
1686 sector_t start, sector_t len, void *data)
1688 struct request_queue *q = bdev_get_queue(dev->bdev);
1690 return q && !test_bit(QUEUE_FLAG_NO_SG_MERGE, &q->queue_flags);
1693 static bool dm_table_all_devices_attribute(struct dm_table *t,
1694 iterate_devices_callout_fn func)
1696 struct dm_target *ti;
1699 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1700 ti = dm_table_get_target(t, i);
1702 if (!ti->type->iterate_devices ||
1703 !ti->type->iterate_devices(ti, func, NULL))
1710 static int device_no_partial_completion(struct dm_target *ti, struct dm_dev *dev,
1711 sector_t start, sector_t len, void *data)
1713 char b[BDEVNAME_SIZE];
1715 /* For now, NVMe devices are the only devices of this class */
1716 return (strncmp(bdevname(dev->bdev, b), "nvme", 4) == 0);
1719 static bool dm_table_does_not_support_partial_completion(struct dm_table *t)
1721 return dm_table_all_devices_attribute(t, device_no_partial_completion);
1724 static int device_not_write_same_capable(struct dm_target *ti, struct dm_dev *dev,
1725 sector_t start, sector_t len, void *data)
1727 struct request_queue *q = bdev_get_queue(dev->bdev);
1729 return q && !q->limits.max_write_same_sectors;
1732 static bool dm_table_supports_write_same(struct dm_table *t)
1734 struct dm_target *ti;
1737 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1738 ti = dm_table_get_target(t, i);
1740 if (!ti->num_write_same_bios)
1743 if (!ti->type->iterate_devices ||
1744 ti->type->iterate_devices(ti, device_not_write_same_capable, NULL))
1751 static int device_not_write_zeroes_capable(struct dm_target *ti, struct dm_dev *dev,
1752 sector_t start, sector_t len, void *data)
1754 struct request_queue *q = bdev_get_queue(dev->bdev);
1756 return q && !q->limits.max_write_zeroes_sectors;
1759 static bool dm_table_supports_write_zeroes(struct dm_table *t)
1761 struct dm_target *ti;
1764 while (i < dm_table_get_num_targets(t)) {
1765 ti = dm_table_get_target(t, i++);
1767 if (!ti->num_write_zeroes_bios)
1770 if (!ti->type->iterate_devices ||
1771 ti->type->iterate_devices(ti, device_not_write_zeroes_capable, NULL))
1778 static int device_not_discard_capable(struct dm_target *ti, struct dm_dev *dev,
1779 sector_t start, sector_t len, void *data)
1781 struct request_queue *q = bdev_get_queue(dev->bdev);
1783 return q && !blk_queue_discard(q);
1786 static bool dm_table_supports_discards(struct dm_table *t)
1788 struct dm_target *ti;
1791 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1792 ti = dm_table_get_target(t, i);
1794 if (!ti->num_discard_bios)
1798 * Either the target provides discard support (as implied by setting
1799 * 'discards_supported') or it relies on _all_ data devices having
1802 if (!ti->discards_supported &&
1803 (!ti->type->iterate_devices ||
1804 ti->type->iterate_devices(ti, device_not_discard_capable, NULL)))
1811 static int device_not_secure_erase_capable(struct dm_target *ti,
1812 struct dm_dev *dev, sector_t start,
1813 sector_t len, void *data)
1815 struct request_queue *q = bdev_get_queue(dev->bdev);
1817 return q && !blk_queue_secure_erase(q);
1820 static bool dm_table_supports_secure_erase(struct dm_table *t)
1822 struct dm_target *ti;
1825 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1826 ti = dm_table_get_target(t, i);
1828 if (!ti->num_secure_erase_bios)
1831 if (!ti->type->iterate_devices ||
1832 ti->type->iterate_devices(ti, device_not_secure_erase_capable, NULL))
1839 void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q,
1840 struct queue_limits *limits)
1842 bool wc = false, fua = false;
1845 * Copy table's limits to the DM device's request_queue
1847 q->limits = *limits;
1849 if (!dm_table_supports_discards(t)) {
1850 blk_queue_flag_clear(QUEUE_FLAG_DISCARD, q);
1851 /* Must also clear discard limits... */
1852 q->limits.max_discard_sectors = 0;
1853 q->limits.max_hw_discard_sectors = 0;
1854 q->limits.discard_granularity = 0;
1855 q->limits.discard_alignment = 0;
1856 q->limits.discard_misaligned = 0;
1858 blk_queue_flag_set(QUEUE_FLAG_DISCARD, q);
1860 if (dm_table_supports_secure_erase(t))
1861 blk_queue_flag_set(QUEUE_FLAG_SECERASE, q);
1863 if (dm_table_supports_flush(t, (1UL << QUEUE_FLAG_WC))) {
1865 if (dm_table_supports_flush(t, (1UL << QUEUE_FLAG_FUA)))
1868 blk_queue_write_cache(q, wc, fua);
1870 if (dm_table_supports_dax(t))
1871 blk_queue_flag_set(QUEUE_FLAG_DAX, q);
1873 blk_queue_flag_clear(QUEUE_FLAG_DAX, q);
1875 if (dm_table_supports_dax_write_cache(t))
1876 dax_write_cache(t->md->dax_dev, true);
1878 /* Ensure that all underlying devices are non-rotational. */
1879 if (dm_table_all_devices_attribute(t, device_is_nonrot))
1880 blk_queue_flag_set(QUEUE_FLAG_NONROT, q);
1882 blk_queue_flag_clear(QUEUE_FLAG_NONROT, q);
1884 if (!dm_table_supports_write_same(t))
1885 q->limits.max_write_same_sectors = 0;
1886 if (!dm_table_supports_write_zeroes(t))
1887 q->limits.max_write_zeroes_sectors = 0;
1889 if (dm_table_all_devices_attribute(t, queue_supports_sg_merge))
1890 blk_queue_flag_clear(QUEUE_FLAG_NO_SG_MERGE, q);
1892 blk_queue_flag_set(QUEUE_FLAG_NO_SG_MERGE, q);
1894 dm_table_verify_integrity(t);
1897 * Determine whether or not this queue's I/O timings contribute
1898 * to the entropy pool, Only request-based targets use this.
1899 * Clear QUEUE_FLAG_ADD_RANDOM if any underlying device does not
1902 if (blk_queue_add_random(q) && dm_table_all_devices_attribute(t, device_is_not_random))
1903 blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, q);
1906 unsigned int dm_table_get_num_targets(struct dm_table *t)
1908 return t->num_targets;
1911 struct list_head *dm_table_get_devices(struct dm_table *t)
1916 fmode_t dm_table_get_mode(struct dm_table *t)
1920 EXPORT_SYMBOL(dm_table_get_mode);
1928 static void suspend_targets(struct dm_table *t, enum suspend_mode mode)
1930 int i = t->num_targets;
1931 struct dm_target *ti = t->targets;
1933 lockdep_assert_held(&t->md->suspend_lock);
1938 if (ti->type->presuspend)
1939 ti->type->presuspend(ti);
1941 case PRESUSPEND_UNDO:
1942 if (ti->type->presuspend_undo)
1943 ti->type->presuspend_undo(ti);
1946 if (ti->type->postsuspend)
1947 ti->type->postsuspend(ti);
1954 void dm_table_presuspend_targets(struct dm_table *t)
1959 suspend_targets(t, PRESUSPEND);
1962 void dm_table_presuspend_undo_targets(struct dm_table *t)
1967 suspend_targets(t, PRESUSPEND_UNDO);
1970 void dm_table_postsuspend_targets(struct dm_table *t)
1975 suspend_targets(t, POSTSUSPEND);
1978 int dm_table_resume_targets(struct dm_table *t)
1982 lockdep_assert_held(&t->md->suspend_lock);
1984 for (i = 0; i < t->num_targets; i++) {
1985 struct dm_target *ti = t->targets + i;
1987 if (!ti->type->preresume)
1990 r = ti->type->preresume(ti);
1992 DMERR("%s: %s: preresume failed, error = %d",
1993 dm_device_name(t->md), ti->type->name, r);
1998 for (i = 0; i < t->num_targets; i++) {
1999 struct dm_target *ti = t->targets + i;
2001 if (ti->type->resume)
2002 ti->type->resume(ti);
2008 void dm_table_add_target_callbacks(struct dm_table *t, struct dm_target_callbacks *cb)
2010 list_add(&cb->list, &t->target_callbacks);
2012 EXPORT_SYMBOL_GPL(dm_table_add_target_callbacks);
2014 int dm_table_any_congested(struct dm_table *t, int bdi_bits)
2016 struct dm_dev_internal *dd;
2017 struct list_head *devices = dm_table_get_devices(t);
2018 struct dm_target_callbacks *cb;
2021 list_for_each_entry(dd, devices, list) {
2022 struct request_queue *q = bdev_get_queue(dd->dm_dev->bdev);
2023 char b[BDEVNAME_SIZE];
2026 r |= bdi_congested(q->backing_dev_info, bdi_bits);
2028 DMWARN_LIMIT("%s: any_congested: nonexistent device %s",
2029 dm_device_name(t->md),
2030 bdevname(dd->dm_dev->bdev, b));
2033 list_for_each_entry(cb, &t->target_callbacks, list)
2034 if (cb->congested_fn)
2035 r |= cb->congested_fn(cb, bdi_bits);
2040 struct mapped_device *dm_table_get_md(struct dm_table *t)
2044 EXPORT_SYMBOL(dm_table_get_md);
2046 void dm_table_run_md_queue_async(struct dm_table *t)
2048 struct mapped_device *md;
2049 struct request_queue *queue;
2051 if (!dm_table_request_based(t))
2054 md = dm_table_get_md(t);
2055 queue = dm_get_md_queue(md);
2057 blk_mq_run_hw_queues(queue, true);
2059 EXPORT_SYMBOL(dm_table_run_md_queue_async);