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;
70 * Similar to ceiling(log_size(n))
72 static unsigned int int_log(unsigned int n, unsigned int base)
77 n = dm_div_up(n, base);
85 * Calculate the index of the child node of the n'th node k'th key.
87 static inline unsigned int get_child(unsigned int n, unsigned int k)
89 return (n * CHILDREN_PER_NODE) + k;
93 * Return the n'th node of level l from table t.
95 static inline sector_t *get_node(struct dm_table *t,
96 unsigned int l, unsigned int n)
98 return t->index[l] + (n * KEYS_PER_NODE);
102 * Return the highest key that you could lookup from the n'th
103 * node on level l of the btree.
105 static sector_t high(struct dm_table *t, unsigned int l, unsigned int n)
107 for (; l < t->depth - 1; l++)
108 n = get_child(n, CHILDREN_PER_NODE - 1);
110 if (n >= t->counts[l])
111 return (sector_t) - 1;
113 return get_node(t, l, n)[KEYS_PER_NODE - 1];
117 * Fills in a level of the btree based on the highs of the level
120 static int setup_btree_index(unsigned int l, struct dm_table *t)
125 for (n = 0U; n < t->counts[l]; n++) {
126 node = get_node(t, l, n);
128 for (k = 0U; k < KEYS_PER_NODE; k++)
129 node[k] = high(t, l + 1, get_child(n, k));
135 void *dm_vcalloc(unsigned long nmemb, unsigned long elem_size)
141 * Check that we're not going to overflow.
143 if (nmemb > (ULONG_MAX / elem_size))
146 size = nmemb * elem_size;
147 addr = vzalloc(size);
151 EXPORT_SYMBOL(dm_vcalloc);
154 * highs, and targets are managed as dynamic arrays during a
157 static int alloc_targets(struct dm_table *t, unsigned int num)
160 struct dm_target *n_targets;
163 * Allocate both the target array and offset array at once.
165 n_highs = (sector_t *) dm_vcalloc(num, sizeof(struct dm_target) +
170 n_targets = (struct dm_target *) (n_highs + num);
172 memset(n_highs, -1, sizeof(*n_highs) * num);
175 t->num_allocated = num;
177 t->targets = n_targets;
182 int dm_table_create(struct dm_table **result, fmode_t mode,
183 unsigned num_targets, struct mapped_device *md)
185 struct dm_table *t = kzalloc(sizeof(*t), GFP_KERNEL);
190 INIT_LIST_HEAD(&t->devices);
193 num_targets = KEYS_PER_NODE;
195 num_targets = dm_round_up(num_targets, KEYS_PER_NODE);
202 if (alloc_targets(t, num_targets)) {
207 t->type = DM_TYPE_NONE;
214 static void free_devices(struct list_head *devices, struct mapped_device *md)
216 struct list_head *tmp, *next;
218 list_for_each_safe(tmp, next, devices) {
219 struct dm_dev_internal *dd =
220 list_entry(tmp, struct dm_dev_internal, list);
221 DMWARN("%s: dm_table_destroy: dm_put_device call missing for %s",
222 dm_device_name(md), dd->dm_dev->name);
223 dm_put_table_device(md, dd->dm_dev);
228 void dm_table_destroy(struct dm_table *t)
235 /* free the indexes */
237 vfree(t->index[t->depth - 2]);
239 /* free the targets */
240 for (i = 0; i < t->num_targets; i++) {
241 struct dm_target *tgt = t->targets + i;
246 dm_put_target_type(tgt->type);
251 /* free the device list */
252 free_devices(&t->devices, t->md);
254 dm_free_md_mempools(t->mempools);
260 * See if we've already got a device in the list.
262 static struct dm_dev_internal *find_device(struct list_head *l, dev_t dev)
264 struct dm_dev_internal *dd;
266 list_for_each_entry (dd, l, list)
267 if (dd->dm_dev->bdev->bd_dev == dev)
274 * If possible, this checks an area of a destination device is invalid.
276 static int device_area_is_invalid(struct dm_target *ti, struct dm_dev *dev,
277 sector_t start, sector_t len, void *data)
279 struct queue_limits *limits = data;
280 struct block_device *bdev = dev->bdev;
282 i_size_read(bdev->bd_inode) >> SECTOR_SHIFT;
283 unsigned short logical_block_size_sectors =
284 limits->logical_block_size >> SECTOR_SHIFT;
285 char b[BDEVNAME_SIZE];
290 if ((start >= dev_size) || (start + len > dev_size)) {
291 DMWARN("%s: %s too small for target: "
292 "start=%llu, len=%llu, dev_size=%llu",
293 dm_device_name(ti->table->md), bdevname(bdev, b),
294 (unsigned long long)start,
295 (unsigned long long)len,
296 (unsigned long long)dev_size);
301 * If the target is mapped to zoned block device(s), check
302 * that the zones are not partially mapped.
304 if (bdev_zoned_model(bdev) != BLK_ZONED_NONE) {
305 unsigned int zone_sectors = bdev_zone_sectors(bdev);
307 if (start & (zone_sectors - 1)) {
308 DMWARN("%s: start=%llu not aligned to h/w zone size %u of %s",
309 dm_device_name(ti->table->md),
310 (unsigned long long)start,
311 zone_sectors, bdevname(bdev, b));
316 * Note: The last zone of a zoned block device may be smaller
317 * than other zones. So for a target mapping the end of a
318 * zoned block device with such a zone, len would not be zone
319 * aligned. We do not allow such last smaller zone to be part
320 * of the mapping here to ensure that mappings with multiple
321 * devices do not end up with a smaller zone in the middle of
324 if (len & (zone_sectors - 1)) {
325 DMWARN("%s: len=%llu not aligned to h/w zone size %u of %s",
326 dm_device_name(ti->table->md),
327 (unsigned long long)len,
328 zone_sectors, bdevname(bdev, b));
333 if (logical_block_size_sectors <= 1)
336 if (start & (logical_block_size_sectors - 1)) {
337 DMWARN("%s: start=%llu not aligned to h/w "
338 "logical block size %u of %s",
339 dm_device_name(ti->table->md),
340 (unsigned long long)start,
341 limits->logical_block_size, bdevname(bdev, b));
345 if (len & (logical_block_size_sectors - 1)) {
346 DMWARN("%s: len=%llu not aligned to h/w "
347 "logical block size %u of %s",
348 dm_device_name(ti->table->md),
349 (unsigned long long)len,
350 limits->logical_block_size, bdevname(bdev, b));
358 * This upgrades the mode on an already open dm_dev, being
359 * careful to leave things as they were if we fail to reopen the
360 * device and not to touch the existing bdev field in case
361 * it is accessed concurrently.
363 static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode,
364 struct mapped_device *md)
367 struct dm_dev *old_dev, *new_dev;
369 old_dev = dd->dm_dev;
371 r = dm_get_table_device(md, dd->dm_dev->bdev->bd_dev,
372 dd->dm_dev->mode | new_mode, &new_dev);
376 dd->dm_dev = new_dev;
377 dm_put_table_device(md, old_dev);
383 * Convert the path to a device
385 dev_t dm_get_dev_t(const char *path)
388 struct block_device *bdev;
390 bdev = lookup_bdev(path);
392 dev = name_to_dev_t(path);
400 EXPORT_SYMBOL_GPL(dm_get_dev_t);
403 * Add a device to the list, or just increment the usage count if
404 * it's already present.
406 int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode,
407 struct dm_dev **result)
411 struct dm_dev_internal *dd;
412 struct dm_table *t = ti->table;
416 dev = dm_get_dev_t(path);
420 dd = find_device(&t->devices, dev);
422 dd = kmalloc(sizeof(*dd), GFP_KERNEL);
426 if ((r = dm_get_table_device(t->md, dev, mode, &dd->dm_dev))) {
431 refcount_set(&dd->count, 1);
432 list_add(&dd->list, &t->devices);
435 } else if (dd->dm_dev->mode != (mode | dd->dm_dev->mode)) {
436 r = upgrade_mode(dd, mode, t->md);
440 refcount_inc(&dd->count);
442 *result = dd->dm_dev;
445 EXPORT_SYMBOL(dm_get_device);
447 static int dm_set_device_limits(struct dm_target *ti, struct dm_dev *dev,
448 sector_t start, sector_t len, void *data)
450 struct queue_limits *limits = data;
451 struct block_device *bdev = dev->bdev;
452 struct request_queue *q = bdev_get_queue(bdev);
453 char b[BDEVNAME_SIZE];
456 DMWARN("%s: Cannot set limits for nonexistent device %s",
457 dm_device_name(ti->table->md), bdevname(bdev, b));
461 if (bdev_stack_limits(limits, bdev, start) < 0)
462 DMWARN("%s: adding target device %s caused an alignment inconsistency: "
463 "physical_block_size=%u, logical_block_size=%u, "
464 "alignment_offset=%u, start=%llu",
465 dm_device_name(ti->table->md), bdevname(bdev, b),
466 q->limits.physical_block_size,
467 q->limits.logical_block_size,
468 q->limits.alignment_offset,
469 (unsigned long long) start << SECTOR_SHIFT);
471 limits->zoned = blk_queue_zoned_model(q);
477 * Decrement a device's use count and remove it if necessary.
479 void dm_put_device(struct dm_target *ti, struct dm_dev *d)
482 struct list_head *devices = &ti->table->devices;
483 struct dm_dev_internal *dd;
485 list_for_each_entry(dd, devices, list) {
486 if (dd->dm_dev == d) {
492 DMWARN("%s: device %s not in table devices list",
493 dm_device_name(ti->table->md), d->name);
496 if (refcount_dec_and_test(&dd->count)) {
497 dm_put_table_device(ti->table->md, d);
502 EXPORT_SYMBOL(dm_put_device);
505 * Checks to see if the target joins onto the end of the table.
507 static int adjoin(struct dm_table *table, struct dm_target *ti)
509 struct dm_target *prev;
511 if (!table->num_targets)
514 prev = &table->targets[table->num_targets - 1];
515 return (ti->begin == (prev->begin + prev->len));
519 * Used to dynamically allocate the arg array.
521 * We do first allocation with GFP_NOIO because dm-mpath and dm-thin must
522 * process messages even if some device is suspended. These messages have a
523 * small fixed number of arguments.
525 * On the other hand, dm-switch needs to process bulk data using messages and
526 * excessive use of GFP_NOIO could cause trouble.
528 static char **realloc_argv(unsigned *size, char **old_argv)
535 new_size = *size * 2;
541 argv = kmalloc_array(new_size, sizeof(*argv), gfp);
542 if (argv && old_argv) {
543 memcpy(argv, old_argv, *size * sizeof(*argv));
552 * Destructively splits up the argument list to pass to ctr.
554 int dm_split_args(int *argc, char ***argvp, char *input)
556 char *start, *end = input, *out, **argv = NULL;
557 unsigned array_size = 0;
566 argv = realloc_argv(&array_size, argv);
571 /* Skip whitespace */
572 start = skip_spaces(end);
575 break; /* success, we hit the end */
577 /* 'out' is used to remove any back-quotes */
580 /* Everything apart from '\0' can be quoted */
581 if (*end == '\\' && *(end + 1)) {
588 break; /* end of token */
593 /* have we already filled the array ? */
594 if ((*argc + 1) > array_size) {
595 argv = realloc_argv(&array_size, argv);
600 /* we know this is whitespace */
604 /* terminate the string and put it in the array */
615 * Impose necessary and sufficient conditions on a devices's table such
616 * that any incoming bio which respects its logical_block_size can be
617 * processed successfully. If it falls across the boundary between
618 * two or more targets, the size of each piece it gets split into must
619 * be compatible with the logical_block_size of the target processing it.
621 static int validate_hardware_logical_block_alignment(struct dm_table *table,
622 struct queue_limits *limits)
625 * This function uses arithmetic modulo the logical_block_size
626 * (in units of 512-byte sectors).
628 unsigned short device_logical_block_size_sects =
629 limits->logical_block_size >> SECTOR_SHIFT;
632 * Offset of the start of the next table entry, mod logical_block_size.
634 unsigned short next_target_start = 0;
637 * Given an aligned bio that extends beyond the end of a
638 * target, how many sectors must the next target handle?
640 unsigned short remaining = 0;
642 struct dm_target *ti;
643 struct queue_limits ti_limits;
647 * Check each entry in the table in turn.
649 for (i = 0; i < dm_table_get_num_targets(table); i++) {
650 ti = dm_table_get_target(table, i);
652 blk_set_stacking_limits(&ti_limits);
654 /* combine all target devices' limits */
655 if (ti->type->iterate_devices)
656 ti->type->iterate_devices(ti, dm_set_device_limits,
660 * If the remaining sectors fall entirely within this
661 * table entry are they compatible with its logical_block_size?
663 if (remaining < ti->len &&
664 remaining & ((ti_limits.logical_block_size >>
669 (unsigned short) ((next_target_start + ti->len) &
670 (device_logical_block_size_sects - 1));
671 remaining = next_target_start ?
672 device_logical_block_size_sects - next_target_start : 0;
676 DMWARN("%s: table line %u (start sect %llu len %llu) "
677 "not aligned to h/w logical block size %u",
678 dm_device_name(table->md), i,
679 (unsigned long long) ti->begin,
680 (unsigned long long) ti->len,
681 limits->logical_block_size);
688 int dm_table_add_target(struct dm_table *t, const char *type,
689 sector_t start, sector_t len, char *params)
691 int r = -EINVAL, argc;
693 struct dm_target *tgt;
696 DMERR("%s: target type %s must appear alone in table",
697 dm_device_name(t->md), t->targets->type->name);
701 BUG_ON(t->num_targets >= t->num_allocated);
703 tgt = t->targets + t->num_targets;
704 memset(tgt, 0, sizeof(*tgt));
707 DMERR("%s: zero-length target", dm_device_name(t->md));
711 tgt->type = dm_get_target_type(type);
713 DMERR("%s: %s: unknown target type", dm_device_name(t->md), type);
717 if (dm_target_needs_singleton(tgt->type)) {
718 if (t->num_targets) {
719 tgt->error = "singleton target type must appear alone in table";
725 if (dm_target_always_writeable(tgt->type) && !(t->mode & FMODE_WRITE)) {
726 tgt->error = "target type may not be included in a read-only table";
730 if (t->immutable_target_type) {
731 if (t->immutable_target_type != tgt->type) {
732 tgt->error = "immutable target type cannot be mixed with other target types";
735 } else if (dm_target_is_immutable(tgt->type)) {
736 if (t->num_targets) {
737 tgt->error = "immutable target type cannot be mixed with other target types";
740 t->immutable_target_type = tgt->type;
743 if (dm_target_has_integrity(tgt->type))
744 t->integrity_added = 1;
749 tgt->error = "Unknown error";
752 * Does this target adjoin the previous one ?
754 if (!adjoin(t, tgt)) {
755 tgt->error = "Gap in table";
759 r = dm_split_args(&argc, &argv, params);
761 tgt->error = "couldn't split parameters (insufficient memory)";
765 r = tgt->type->ctr(tgt, argc, argv);
770 t->highs[t->num_targets++] = tgt->begin + tgt->len - 1;
772 if (!tgt->num_discard_bios && tgt->discards_supported)
773 DMWARN("%s: %s: ignoring discards_supported because num_discard_bios is zero.",
774 dm_device_name(t->md), type);
779 DMERR("%s: %s: %s", dm_device_name(t->md), type, tgt->error);
780 dm_put_target_type(tgt->type);
785 * Target argument parsing helpers.
787 static int validate_next_arg(const struct dm_arg *arg,
788 struct dm_arg_set *arg_set,
789 unsigned *value, char **error, unsigned grouped)
791 const char *arg_str = dm_shift_arg(arg_set);
795 (sscanf(arg_str, "%u%c", value, &dummy) != 1) ||
796 (*value < arg->min) ||
797 (*value > arg->max) ||
798 (grouped && arg_set->argc < *value)) {
806 int dm_read_arg(const struct dm_arg *arg, struct dm_arg_set *arg_set,
807 unsigned *value, char **error)
809 return validate_next_arg(arg, arg_set, value, error, 0);
811 EXPORT_SYMBOL(dm_read_arg);
813 int dm_read_arg_group(const struct dm_arg *arg, struct dm_arg_set *arg_set,
814 unsigned *value, char **error)
816 return validate_next_arg(arg, arg_set, value, error, 1);
818 EXPORT_SYMBOL(dm_read_arg_group);
820 const char *dm_shift_arg(struct dm_arg_set *as)
833 EXPORT_SYMBOL(dm_shift_arg);
835 void dm_consume_args(struct dm_arg_set *as, unsigned num_args)
837 BUG_ON(as->argc < num_args);
838 as->argc -= num_args;
839 as->argv += num_args;
841 EXPORT_SYMBOL(dm_consume_args);
843 static bool __table_type_bio_based(enum dm_queue_mode table_type)
845 return (table_type == DM_TYPE_BIO_BASED ||
846 table_type == DM_TYPE_DAX_BIO_BASED ||
847 table_type == DM_TYPE_NVME_BIO_BASED);
850 static bool __table_type_request_based(enum dm_queue_mode table_type)
852 return table_type == DM_TYPE_REQUEST_BASED;
855 void dm_table_set_type(struct dm_table *t, enum dm_queue_mode type)
859 EXPORT_SYMBOL_GPL(dm_table_set_type);
861 /* validate the dax capability of the target device span */
862 int device_supports_dax(struct dm_target *ti, struct dm_dev *dev,
863 sector_t start, sector_t len, void *data)
865 int blocksize = *(int *) data;
867 return generic_fsdax_supported(dev->dax_dev, dev->bdev, blocksize,
871 /* Check devices support synchronous DAX */
872 static int device_dax_synchronous(struct dm_target *ti, struct dm_dev *dev,
873 sector_t start, sector_t len, void *data)
875 return dev->dax_dev && dax_synchronous(dev->dax_dev);
878 bool dm_table_supports_dax(struct dm_table *t,
879 iterate_devices_callout_fn iterate_fn, int *blocksize)
881 struct dm_target *ti;
884 /* Ensure that all targets support DAX. */
885 for (i = 0; i < dm_table_get_num_targets(t); i++) {
886 ti = dm_table_get_target(t, i);
888 if (!ti->type->direct_access)
891 if (!ti->type->iterate_devices ||
892 !ti->type->iterate_devices(ti, iterate_fn, blocksize))
899 static bool dm_table_does_not_support_partial_completion(struct dm_table *t);
901 static int device_is_rq_stackable(struct dm_target *ti, struct dm_dev *dev,
902 sector_t start, sector_t len, void *data)
904 struct block_device *bdev = dev->bdev;
905 struct request_queue *q = bdev_get_queue(bdev);
907 /* request-based cannot stack on partitions! */
908 if (bdev != bdev->bd_contains)
911 return queue_is_mq(q);
914 static int dm_table_determine_type(struct dm_table *t)
917 unsigned bio_based = 0, request_based = 0, hybrid = 0;
918 struct dm_target *tgt;
919 struct list_head *devices = dm_table_get_devices(t);
920 enum dm_queue_mode live_md_type = dm_get_md_type(t->md);
921 int page_size = PAGE_SIZE;
923 if (t->type != DM_TYPE_NONE) {
924 /* target already set the table's type */
925 if (t->type == DM_TYPE_BIO_BASED) {
926 /* possibly upgrade to a variant of bio-based */
927 goto verify_bio_based;
929 BUG_ON(t->type == DM_TYPE_DAX_BIO_BASED);
930 BUG_ON(t->type == DM_TYPE_NVME_BIO_BASED);
931 goto verify_rq_based;
934 for (i = 0; i < t->num_targets; i++) {
935 tgt = t->targets + i;
936 if (dm_target_hybrid(tgt))
938 else if (dm_target_request_based(tgt))
943 if (bio_based && request_based) {
944 DMERR("Inconsistent table: different target types"
945 " can't be mixed up");
950 if (hybrid && !bio_based && !request_based) {
952 * The targets can work either way.
953 * Determine the type from the live device.
954 * Default to bio-based if device is new.
956 if (__table_type_request_based(live_md_type))
964 /* We must use this table as bio-based */
965 t->type = DM_TYPE_BIO_BASED;
966 if (dm_table_supports_dax(t, device_supports_dax, &page_size) ||
967 (list_empty(devices) && live_md_type == DM_TYPE_DAX_BIO_BASED)) {
968 t->type = DM_TYPE_DAX_BIO_BASED;
970 /* Check if upgrading to NVMe bio-based is valid or required */
971 tgt = dm_table_get_immutable_target(t);
972 if (tgt && !tgt->max_io_len && dm_table_does_not_support_partial_completion(t)) {
973 t->type = DM_TYPE_NVME_BIO_BASED;
974 goto verify_rq_based; /* must be stacked directly on NVMe (blk-mq) */
975 } else if (list_empty(devices) && live_md_type == DM_TYPE_NVME_BIO_BASED) {
976 t->type = DM_TYPE_NVME_BIO_BASED;
982 BUG_ON(!request_based); /* No targets in this table */
984 t->type = DM_TYPE_REQUEST_BASED;
988 * Request-based dm supports only tables that have a single target now.
989 * To support multiple targets, request splitting support is needed,
990 * and that needs lots of changes in the block-layer.
991 * (e.g. request completion process for partial completion.)
993 if (t->num_targets > 1) {
994 DMERR("%s DM doesn't support multiple targets",
995 t->type == DM_TYPE_NVME_BIO_BASED ? "nvme bio-based" : "request-based");
999 if (list_empty(devices)) {
1001 struct dm_table *live_table = dm_get_live_table(t->md, &srcu_idx);
1003 /* inherit live table's type */
1005 t->type = live_table->type;
1006 dm_put_live_table(t->md, srcu_idx);
1010 tgt = dm_table_get_immutable_target(t);
1012 DMERR("table load rejected: immutable target is required");
1014 } else if (tgt->max_io_len) {
1015 DMERR("table load rejected: immutable target that splits IO is not supported");
1019 /* Non-request-stackable devices can't be used for request-based dm */
1020 if (!tgt->type->iterate_devices ||
1021 !tgt->type->iterate_devices(tgt, device_is_rq_stackable, NULL)) {
1022 DMERR("table load rejected: including non-request-stackable devices");
1029 enum dm_queue_mode dm_table_get_type(struct dm_table *t)
1034 struct target_type *dm_table_get_immutable_target_type(struct dm_table *t)
1036 return t->immutable_target_type;
1039 struct dm_target *dm_table_get_immutable_target(struct dm_table *t)
1041 /* Immutable target is implicitly a singleton */
1042 if (t->num_targets > 1 ||
1043 !dm_target_is_immutable(t->targets[0].type))
1049 struct dm_target *dm_table_get_wildcard_target(struct dm_table *t)
1051 struct dm_target *ti;
1054 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1055 ti = dm_table_get_target(t, i);
1056 if (dm_target_is_wildcard(ti->type))
1063 bool dm_table_bio_based(struct dm_table *t)
1065 return __table_type_bio_based(dm_table_get_type(t));
1068 bool dm_table_request_based(struct dm_table *t)
1070 return __table_type_request_based(dm_table_get_type(t));
1073 static int dm_table_alloc_md_mempools(struct dm_table *t, struct mapped_device *md)
1075 enum dm_queue_mode type = dm_table_get_type(t);
1076 unsigned per_io_data_size = 0;
1077 unsigned min_pool_size = 0;
1078 struct dm_target *ti;
1081 if (unlikely(type == DM_TYPE_NONE)) {
1082 DMWARN("no table type is set, can't allocate mempools");
1086 if (__table_type_bio_based(type))
1087 for (i = 0; i < t->num_targets; i++) {
1088 ti = t->targets + i;
1089 per_io_data_size = max(per_io_data_size, ti->per_io_data_size);
1090 min_pool_size = max(min_pool_size, ti->num_flush_bios);
1093 t->mempools = dm_alloc_md_mempools(md, type, t->integrity_supported,
1094 per_io_data_size, min_pool_size);
1101 void dm_table_free_md_mempools(struct dm_table *t)
1103 dm_free_md_mempools(t->mempools);
1107 struct dm_md_mempools *dm_table_get_md_mempools(struct dm_table *t)
1112 static int setup_indexes(struct dm_table *t)
1115 unsigned int total = 0;
1118 /* allocate the space for *all* the indexes */
1119 for (i = t->depth - 2; i >= 0; i--) {
1120 t->counts[i] = dm_div_up(t->counts[i + 1], CHILDREN_PER_NODE);
1121 total += t->counts[i];
1124 indexes = (sector_t *) dm_vcalloc(total, (unsigned long) NODE_SIZE);
1128 /* set up internal nodes, bottom-up */
1129 for (i = t->depth - 2; i >= 0; i--) {
1130 t->index[i] = indexes;
1131 indexes += (KEYS_PER_NODE * t->counts[i]);
1132 setup_btree_index(i, t);
1139 * Builds the btree to index the map.
1141 static int dm_table_build_index(struct dm_table *t)
1144 unsigned int leaf_nodes;
1146 /* how many indexes will the btree have ? */
1147 leaf_nodes = dm_div_up(t->num_targets, KEYS_PER_NODE);
1148 t->depth = 1 + int_log(leaf_nodes, CHILDREN_PER_NODE);
1150 /* leaf layer has already been set up */
1151 t->counts[t->depth - 1] = leaf_nodes;
1152 t->index[t->depth - 1] = t->highs;
1155 r = setup_indexes(t);
1160 static bool integrity_profile_exists(struct gendisk *disk)
1162 return !!blk_get_integrity(disk);
1166 * Get a disk whose integrity profile reflects the table's profile.
1167 * Returns NULL if integrity support was inconsistent or unavailable.
1169 static struct gendisk * dm_table_get_integrity_disk(struct dm_table *t)
1171 struct list_head *devices = dm_table_get_devices(t);
1172 struct dm_dev_internal *dd = NULL;
1173 struct gendisk *prev_disk = NULL, *template_disk = NULL;
1176 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1177 struct dm_target *ti = dm_table_get_target(t, i);
1178 if (!dm_target_passes_integrity(ti->type))
1182 list_for_each_entry(dd, devices, list) {
1183 template_disk = dd->dm_dev->bdev->bd_disk;
1184 if (!integrity_profile_exists(template_disk))
1186 else if (prev_disk &&
1187 blk_integrity_compare(prev_disk, template_disk) < 0)
1189 prev_disk = template_disk;
1192 return template_disk;
1196 DMWARN("%s: integrity not set: %s and %s profile mismatch",
1197 dm_device_name(t->md),
1198 prev_disk->disk_name,
1199 template_disk->disk_name);
1204 * Register the mapped device for blk_integrity support if the
1205 * underlying devices have an integrity profile. But all devices may
1206 * not have matching profiles (checking all devices isn't reliable
1207 * during table load because this table may use other DM device(s) which
1208 * must be resumed before they will have an initialized integity
1209 * profile). Consequently, stacked DM devices force a 2 stage integrity
1210 * profile validation: First pass during table load, final pass during
1213 static int dm_table_register_integrity(struct dm_table *t)
1215 struct mapped_device *md = t->md;
1216 struct gendisk *template_disk = NULL;
1218 /* If target handles integrity itself do not register it here. */
1219 if (t->integrity_added)
1222 template_disk = dm_table_get_integrity_disk(t);
1226 if (!integrity_profile_exists(dm_disk(md))) {
1227 t->integrity_supported = true;
1229 * Register integrity profile during table load; we can do
1230 * this because the final profile must match during resume.
1232 blk_integrity_register(dm_disk(md),
1233 blk_get_integrity(template_disk));
1238 * If DM device already has an initialized integrity
1239 * profile the new profile should not conflict.
1241 if (blk_integrity_compare(dm_disk(md), template_disk) < 0) {
1242 DMWARN("%s: conflict with existing integrity profile: "
1243 "%s profile mismatch",
1244 dm_device_name(t->md),
1245 template_disk->disk_name);
1249 /* Preserve existing integrity profile */
1250 t->integrity_supported = true;
1255 * Prepares the table for use by building the indices,
1256 * setting the type, and allocating mempools.
1258 int dm_table_complete(struct dm_table *t)
1262 r = dm_table_determine_type(t);
1264 DMERR("unable to determine table type");
1268 r = dm_table_build_index(t);
1270 DMERR("unable to build btrees");
1274 r = dm_table_register_integrity(t);
1276 DMERR("could not register integrity profile.");
1280 r = dm_table_alloc_md_mempools(t, t->md);
1282 DMERR("unable to allocate mempools");
1287 static DEFINE_MUTEX(_event_lock);
1288 void dm_table_event_callback(struct dm_table *t,
1289 void (*fn)(void *), void *context)
1291 mutex_lock(&_event_lock);
1293 t->event_context = context;
1294 mutex_unlock(&_event_lock);
1297 void dm_table_event(struct dm_table *t)
1300 * You can no longer call dm_table_event() from interrupt
1301 * context, use a bottom half instead.
1303 BUG_ON(in_interrupt());
1305 mutex_lock(&_event_lock);
1307 t->event_fn(t->event_context);
1308 mutex_unlock(&_event_lock);
1310 EXPORT_SYMBOL(dm_table_event);
1312 inline sector_t dm_table_get_size(struct dm_table *t)
1314 return t->num_targets ? (t->highs[t->num_targets - 1] + 1) : 0;
1316 EXPORT_SYMBOL(dm_table_get_size);
1318 struct dm_target *dm_table_get_target(struct dm_table *t, unsigned int index)
1320 if (index >= t->num_targets)
1323 return t->targets + index;
1327 * Search the btree for the correct target.
1329 * Caller should check returned pointer for NULL
1330 * to trap I/O beyond end of device.
1332 struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector)
1334 unsigned int l, n = 0, k = 0;
1337 if (unlikely(sector >= dm_table_get_size(t)))
1340 for (l = 0; l < t->depth; l++) {
1341 n = get_child(n, k);
1342 node = get_node(t, l, n);
1344 for (k = 0; k < KEYS_PER_NODE; k++)
1345 if (node[k] >= sector)
1349 return &t->targets[(KEYS_PER_NODE * n) + k];
1352 static int count_device(struct dm_target *ti, struct dm_dev *dev,
1353 sector_t start, sector_t len, void *data)
1355 unsigned *num_devices = data;
1363 * Check whether a table has no data devices attached using each
1364 * target's iterate_devices method.
1365 * Returns false if the result is unknown because a target doesn't
1366 * support iterate_devices.
1368 bool dm_table_has_no_data_devices(struct dm_table *table)
1370 struct dm_target *ti;
1371 unsigned i, num_devices;
1373 for (i = 0; i < dm_table_get_num_targets(table); i++) {
1374 ti = dm_table_get_target(table, i);
1376 if (!ti->type->iterate_devices)
1380 ti->type->iterate_devices(ti, count_device, &num_devices);
1388 static int device_is_zoned_model(struct dm_target *ti, struct dm_dev *dev,
1389 sector_t start, sector_t len, void *data)
1391 struct request_queue *q = bdev_get_queue(dev->bdev);
1392 enum blk_zoned_model *zoned_model = data;
1394 return q && blk_queue_zoned_model(q) == *zoned_model;
1397 static bool dm_table_supports_zoned_model(struct dm_table *t,
1398 enum blk_zoned_model zoned_model)
1400 struct dm_target *ti;
1403 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1404 ti = dm_table_get_target(t, i);
1406 if (zoned_model == BLK_ZONED_HM &&
1407 !dm_target_supports_zoned_hm(ti->type))
1410 if (!ti->type->iterate_devices ||
1411 !ti->type->iterate_devices(ti, device_is_zoned_model, &zoned_model))
1418 static int device_matches_zone_sectors(struct dm_target *ti, struct dm_dev *dev,
1419 sector_t start, sector_t len, void *data)
1421 struct request_queue *q = bdev_get_queue(dev->bdev);
1422 unsigned int *zone_sectors = data;
1424 return q && blk_queue_zone_sectors(q) == *zone_sectors;
1427 static bool dm_table_matches_zone_sectors(struct dm_table *t,
1428 unsigned int zone_sectors)
1430 struct dm_target *ti;
1433 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1434 ti = dm_table_get_target(t, i);
1436 if (!ti->type->iterate_devices ||
1437 !ti->type->iterate_devices(ti, device_matches_zone_sectors, &zone_sectors))
1444 static int validate_hardware_zoned_model(struct dm_table *table,
1445 enum blk_zoned_model zoned_model,
1446 unsigned int zone_sectors)
1448 if (zoned_model == BLK_ZONED_NONE)
1451 if (!dm_table_supports_zoned_model(table, zoned_model)) {
1452 DMERR("%s: zoned model is not consistent across all devices",
1453 dm_device_name(table->md));
1457 /* Check zone size validity and compatibility */
1458 if (!zone_sectors || !is_power_of_2(zone_sectors))
1461 if (!dm_table_matches_zone_sectors(table, zone_sectors)) {
1462 DMERR("%s: zone sectors is not consistent across all devices",
1463 dm_device_name(table->md));
1471 * Establish the new table's queue_limits and validate them.
1473 int dm_calculate_queue_limits(struct dm_table *table,
1474 struct queue_limits *limits)
1476 struct dm_target *ti;
1477 struct queue_limits ti_limits;
1479 enum blk_zoned_model zoned_model = BLK_ZONED_NONE;
1480 unsigned int zone_sectors = 0;
1482 blk_set_stacking_limits(limits);
1484 for (i = 0; i < dm_table_get_num_targets(table); i++) {
1485 blk_set_stacking_limits(&ti_limits);
1487 ti = dm_table_get_target(table, i);
1489 if (!ti->type->iterate_devices)
1490 goto combine_limits;
1493 * Combine queue limits of all the devices this target uses.
1495 ti->type->iterate_devices(ti, dm_set_device_limits,
1498 if (zoned_model == BLK_ZONED_NONE && ti_limits.zoned != BLK_ZONED_NONE) {
1500 * After stacking all limits, validate all devices
1501 * in table support this zoned model and zone sectors.
1503 zoned_model = ti_limits.zoned;
1504 zone_sectors = ti_limits.chunk_sectors;
1507 /* Set I/O hints portion of queue limits */
1508 if (ti->type->io_hints)
1509 ti->type->io_hints(ti, &ti_limits);
1512 * Check each device area is consistent with the target's
1513 * overall queue limits.
1515 if (ti->type->iterate_devices(ti, device_area_is_invalid,
1521 * Merge this target's queue limits into the overall limits
1524 if (blk_stack_limits(limits, &ti_limits, 0) < 0)
1525 DMWARN("%s: adding target device "
1526 "(start sect %llu len %llu) "
1527 "caused an alignment inconsistency",
1528 dm_device_name(table->md),
1529 (unsigned long long) ti->begin,
1530 (unsigned long long) ti->len);
1533 * FIXME: this should likely be moved to blk_stack_limits(), would
1534 * also eliminate limits->zoned stacking hack in dm_set_device_limits()
1536 if (limits->zoned == BLK_ZONED_NONE && ti_limits.zoned != BLK_ZONED_NONE) {
1538 * By default, the stacked limits zoned model is set to
1539 * BLK_ZONED_NONE in blk_set_stacking_limits(). Update
1540 * this model using the first target model reported
1541 * that is not BLK_ZONED_NONE. This will be either the
1542 * first target device zoned model or the model reported
1543 * by the target .io_hints.
1545 limits->zoned = ti_limits.zoned;
1550 * Verify that the zoned model and zone sectors, as determined before
1551 * any .io_hints override, are the same across all devices in the table.
1552 * - this is especially relevant if .io_hints is emulating a disk-managed
1553 * zoned model (aka BLK_ZONED_NONE) on host-managed zoned block devices.
1556 if (limits->zoned != BLK_ZONED_NONE) {
1558 * ...IF the above limits stacking determined a zoned model
1559 * validate that all of the table's devices conform to it.
1561 zoned_model = limits->zoned;
1562 zone_sectors = limits->chunk_sectors;
1564 if (validate_hardware_zoned_model(table, zoned_model, zone_sectors))
1567 return validate_hardware_logical_block_alignment(table, limits);
1571 * Verify that all devices have an integrity profile that matches the
1572 * DM device's registered integrity profile. If the profiles don't
1573 * match then unregister the DM device's integrity profile.
1575 static void dm_table_verify_integrity(struct dm_table *t)
1577 struct gendisk *template_disk = NULL;
1579 if (t->integrity_added)
1582 if (t->integrity_supported) {
1584 * Verify that the original integrity profile
1585 * matches all the devices in this table.
1587 template_disk = dm_table_get_integrity_disk(t);
1588 if (template_disk &&
1589 blk_integrity_compare(dm_disk(t->md), template_disk) >= 0)
1593 if (integrity_profile_exists(dm_disk(t->md))) {
1594 DMWARN("%s: unable to establish an integrity profile",
1595 dm_device_name(t->md));
1596 blk_integrity_unregister(dm_disk(t->md));
1600 static int device_flush_capable(struct dm_target *ti, struct dm_dev *dev,
1601 sector_t start, sector_t len, void *data)
1603 unsigned long flush = (unsigned long) data;
1604 struct request_queue *q = bdev_get_queue(dev->bdev);
1606 return q && (q->queue_flags & flush);
1609 static bool dm_table_supports_flush(struct dm_table *t, unsigned long flush)
1611 struct dm_target *ti;
1615 * Require at least one underlying device to support flushes.
1616 * t->devices includes internal dm devices such as mirror logs
1617 * so we need to use iterate_devices here, which targets
1618 * supporting flushes must provide.
1620 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1621 ti = dm_table_get_target(t, i);
1623 if (!ti->num_flush_bios)
1626 if (ti->flush_supported)
1629 if (ti->type->iterate_devices &&
1630 ti->type->iterate_devices(ti, device_flush_capable, (void *) flush))
1637 static int device_dax_write_cache_enabled(struct dm_target *ti,
1638 struct dm_dev *dev, sector_t start,
1639 sector_t len, void *data)
1641 struct dax_device *dax_dev = dev->dax_dev;
1646 if (dax_write_cache_enabled(dax_dev))
1651 static int dm_table_supports_dax_write_cache(struct dm_table *t)
1653 struct dm_target *ti;
1656 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1657 ti = dm_table_get_target(t, i);
1659 if (ti->type->iterate_devices &&
1660 ti->type->iterate_devices(ti,
1661 device_dax_write_cache_enabled, NULL))
1668 static int device_is_nonrot(struct dm_target *ti, struct dm_dev *dev,
1669 sector_t start, sector_t len, void *data)
1671 struct request_queue *q = bdev_get_queue(dev->bdev);
1673 return q && blk_queue_nonrot(q);
1676 static int device_is_not_random(struct dm_target *ti, struct dm_dev *dev,
1677 sector_t start, sector_t len, void *data)
1679 struct request_queue *q = bdev_get_queue(dev->bdev);
1681 return q && !blk_queue_add_random(q);
1684 static bool dm_table_all_devices_attribute(struct dm_table *t,
1685 iterate_devices_callout_fn func)
1687 struct dm_target *ti;
1690 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1691 ti = dm_table_get_target(t, i);
1693 if (!ti->type->iterate_devices ||
1694 !ti->type->iterate_devices(ti, func, NULL))
1701 static int device_no_partial_completion(struct dm_target *ti, struct dm_dev *dev,
1702 sector_t start, sector_t len, void *data)
1704 char b[BDEVNAME_SIZE];
1706 /* For now, NVMe devices are the only devices of this class */
1707 return (strncmp(bdevname(dev->bdev, b), "nvme", 4) == 0);
1710 static bool dm_table_does_not_support_partial_completion(struct dm_table *t)
1712 return dm_table_all_devices_attribute(t, device_no_partial_completion);
1715 static int device_not_write_same_capable(struct dm_target *ti, struct dm_dev *dev,
1716 sector_t start, sector_t len, void *data)
1718 struct request_queue *q = bdev_get_queue(dev->bdev);
1720 return q && !q->limits.max_write_same_sectors;
1723 static bool dm_table_supports_write_same(struct dm_table *t)
1725 struct dm_target *ti;
1728 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1729 ti = dm_table_get_target(t, i);
1731 if (!ti->num_write_same_bios)
1734 if (!ti->type->iterate_devices ||
1735 ti->type->iterate_devices(ti, device_not_write_same_capable, NULL))
1742 static int device_not_write_zeroes_capable(struct dm_target *ti, struct dm_dev *dev,
1743 sector_t start, sector_t len, void *data)
1745 struct request_queue *q = bdev_get_queue(dev->bdev);
1747 return q && !q->limits.max_write_zeroes_sectors;
1750 static bool dm_table_supports_write_zeroes(struct dm_table *t)
1752 struct dm_target *ti;
1755 while (i < dm_table_get_num_targets(t)) {
1756 ti = dm_table_get_target(t, i++);
1758 if (!ti->num_write_zeroes_bios)
1761 if (!ti->type->iterate_devices ||
1762 ti->type->iterate_devices(ti, device_not_write_zeroes_capable, NULL))
1769 static int device_not_discard_capable(struct dm_target *ti, struct dm_dev *dev,
1770 sector_t start, sector_t len, void *data)
1772 struct request_queue *q = bdev_get_queue(dev->bdev);
1774 return q && !blk_queue_discard(q);
1777 static bool dm_table_supports_discards(struct dm_table *t)
1779 struct dm_target *ti;
1782 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1783 ti = dm_table_get_target(t, i);
1785 if (!ti->num_discard_bios)
1789 * Either the target provides discard support (as implied by setting
1790 * 'discards_supported') or it relies on _all_ data devices having
1793 if (!ti->discards_supported &&
1794 (!ti->type->iterate_devices ||
1795 ti->type->iterate_devices(ti, device_not_discard_capable, NULL)))
1802 static int device_not_secure_erase_capable(struct dm_target *ti,
1803 struct dm_dev *dev, sector_t start,
1804 sector_t len, void *data)
1806 struct request_queue *q = bdev_get_queue(dev->bdev);
1808 return q && !blk_queue_secure_erase(q);
1811 static bool dm_table_supports_secure_erase(struct dm_table *t)
1813 struct dm_target *ti;
1816 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1817 ti = dm_table_get_target(t, i);
1819 if (!ti->num_secure_erase_bios)
1822 if (!ti->type->iterate_devices ||
1823 ti->type->iterate_devices(ti, device_not_secure_erase_capable, NULL))
1830 static int device_requires_stable_pages(struct dm_target *ti,
1831 struct dm_dev *dev, sector_t start,
1832 sector_t len, void *data)
1834 struct request_queue *q = bdev_get_queue(dev->bdev);
1836 return q && bdi_cap_stable_pages_required(q->backing_dev_info);
1840 * If any underlying device requires stable pages, a table must require
1841 * them as well. Only targets that support iterate_devices are considered:
1842 * don't want error, zero, etc to require stable pages.
1844 static bool dm_table_requires_stable_pages(struct dm_table *t)
1846 struct dm_target *ti;
1849 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1850 ti = dm_table_get_target(t, i);
1852 if (ti->type->iterate_devices &&
1853 ti->type->iterate_devices(ti, device_requires_stable_pages, NULL))
1860 void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q,
1861 struct queue_limits *limits)
1863 bool wc = false, fua = false;
1864 int page_size = PAGE_SIZE;
1867 * Copy table's limits to the DM device's request_queue
1869 q->limits = *limits;
1871 if (!dm_table_supports_discards(t)) {
1872 blk_queue_flag_clear(QUEUE_FLAG_DISCARD, q);
1873 /* Must also clear discard limits... */
1874 q->limits.max_discard_sectors = 0;
1875 q->limits.max_hw_discard_sectors = 0;
1876 q->limits.discard_granularity = 0;
1877 q->limits.discard_alignment = 0;
1878 q->limits.discard_misaligned = 0;
1880 blk_queue_flag_set(QUEUE_FLAG_DISCARD, q);
1882 if (dm_table_supports_secure_erase(t))
1883 blk_queue_flag_set(QUEUE_FLAG_SECERASE, q);
1885 if (dm_table_supports_flush(t, (1UL << QUEUE_FLAG_WC))) {
1887 if (dm_table_supports_flush(t, (1UL << QUEUE_FLAG_FUA)))
1890 blk_queue_write_cache(q, wc, fua);
1892 if (dm_table_supports_dax(t, device_supports_dax, &page_size)) {
1893 blk_queue_flag_set(QUEUE_FLAG_DAX, q);
1894 if (dm_table_supports_dax(t, device_dax_synchronous, NULL))
1895 set_dax_synchronous(t->md->dax_dev);
1898 blk_queue_flag_clear(QUEUE_FLAG_DAX, q);
1900 if (dm_table_supports_dax_write_cache(t))
1901 dax_write_cache(t->md->dax_dev, true);
1903 /* Ensure that all underlying devices are non-rotational. */
1904 if (dm_table_all_devices_attribute(t, device_is_nonrot))
1905 blk_queue_flag_set(QUEUE_FLAG_NONROT, q);
1907 blk_queue_flag_clear(QUEUE_FLAG_NONROT, q);
1909 if (!dm_table_supports_write_same(t))
1910 q->limits.max_write_same_sectors = 0;
1911 if (!dm_table_supports_write_zeroes(t))
1912 q->limits.max_write_zeroes_sectors = 0;
1914 dm_table_verify_integrity(t);
1917 * Some devices don't use blk_integrity but still want stable pages
1918 * because they do their own checksumming.
1920 if (dm_table_requires_stable_pages(t))
1921 q->backing_dev_info->capabilities |= BDI_CAP_STABLE_WRITES;
1923 q->backing_dev_info->capabilities &= ~BDI_CAP_STABLE_WRITES;
1926 * Determine whether or not this queue's I/O timings contribute
1927 * to the entropy pool, Only request-based targets use this.
1928 * Clear QUEUE_FLAG_ADD_RANDOM if any underlying device does not
1931 if (blk_queue_add_random(q) && dm_table_all_devices_attribute(t, device_is_not_random))
1932 blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, q);
1935 * For a zoned target, the number of zones should be updated for the
1936 * correct value to be exposed in sysfs queue/nr_zones. For a BIO based
1937 * target, this is all that is needed.
1939 #ifdef CONFIG_BLK_DEV_ZONED
1940 if (blk_queue_is_zoned(q)) {
1941 WARN_ON_ONCE(queue_is_mq(q));
1942 q->nr_zones = blkdev_nr_zones(t->md->disk);
1946 /* Allow reads to exceed readahead limits */
1947 q->backing_dev_info->io_pages = limits->max_sectors >> (PAGE_SHIFT - 9);
1950 unsigned int dm_table_get_num_targets(struct dm_table *t)
1952 return t->num_targets;
1955 struct list_head *dm_table_get_devices(struct dm_table *t)
1960 fmode_t dm_table_get_mode(struct dm_table *t)
1964 EXPORT_SYMBOL(dm_table_get_mode);
1972 static void suspend_targets(struct dm_table *t, enum suspend_mode mode)
1974 int i = t->num_targets;
1975 struct dm_target *ti = t->targets;
1977 lockdep_assert_held(&t->md->suspend_lock);
1982 if (ti->type->presuspend)
1983 ti->type->presuspend(ti);
1985 case PRESUSPEND_UNDO:
1986 if (ti->type->presuspend_undo)
1987 ti->type->presuspend_undo(ti);
1990 if (ti->type->postsuspend)
1991 ti->type->postsuspend(ti);
1998 void dm_table_presuspend_targets(struct dm_table *t)
2003 suspend_targets(t, PRESUSPEND);
2006 void dm_table_presuspend_undo_targets(struct dm_table *t)
2011 suspend_targets(t, PRESUSPEND_UNDO);
2014 void dm_table_postsuspend_targets(struct dm_table *t)
2019 suspend_targets(t, POSTSUSPEND);
2022 int dm_table_resume_targets(struct dm_table *t)
2026 lockdep_assert_held(&t->md->suspend_lock);
2028 for (i = 0; i < t->num_targets; i++) {
2029 struct dm_target *ti = t->targets + i;
2031 if (!ti->type->preresume)
2034 r = ti->type->preresume(ti);
2036 DMERR("%s: %s: preresume failed, error = %d",
2037 dm_device_name(t->md), ti->type->name, r);
2042 for (i = 0; i < t->num_targets; i++) {
2043 struct dm_target *ti = t->targets + i;
2045 if (ti->type->resume)
2046 ti->type->resume(ti);
2052 struct mapped_device *dm_table_get_md(struct dm_table *t)
2056 EXPORT_SYMBOL(dm_table_get_md);
2058 const char *dm_table_device_name(struct dm_table *t)
2060 return dm_device_name(t->md);
2062 EXPORT_SYMBOL_GPL(dm_table_device_name);
2064 void dm_table_run_md_queue_async(struct dm_table *t)
2066 struct mapped_device *md;
2067 struct request_queue *queue;
2069 if (!dm_table_request_based(t))
2072 md = dm_table_get_md(t);
2073 queue = dm_get_md_queue(md);
2075 blk_mq_run_hw_queues(queue, true);
2077 EXPORT_SYMBOL(dm_table_run_md_queue_async);