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.
167 n_highs = (sector_t *) dm_vcalloc(num, sizeof(struct dm_target) +
172 n_targets = (struct dm_target *) (n_highs + num);
174 memset(n_highs, -1, sizeof(*n_highs) * num);
177 t->num_allocated = num;
179 t->targets = n_targets;
184 int dm_table_create(struct dm_table **result, fmode_t mode,
185 unsigned num_targets, struct mapped_device *md)
187 struct dm_table *t = kzalloc(sizeof(*t), GFP_KERNEL);
192 INIT_LIST_HEAD(&t->devices);
193 INIT_LIST_HEAD(&t->target_callbacks);
196 num_targets = KEYS_PER_NODE;
198 num_targets = dm_round_up(num_targets, KEYS_PER_NODE);
205 if (alloc_targets(t, num_targets)) {
210 t->type = DM_TYPE_NONE;
217 static void free_devices(struct list_head *devices, struct mapped_device *md)
219 struct list_head *tmp, *next;
221 list_for_each_safe(tmp, next, devices) {
222 struct dm_dev_internal *dd =
223 list_entry(tmp, struct dm_dev_internal, list);
224 DMWARN("%s: dm_table_destroy: dm_put_device call missing for %s",
225 dm_device_name(md), dd->dm_dev->name);
226 dm_put_table_device(md, dd->dm_dev);
231 void dm_table_destroy(struct dm_table *t)
238 /* free the indexes */
240 vfree(t->index[t->depth - 2]);
242 /* free the targets */
243 for (i = 0; i < t->num_targets; i++) {
244 struct dm_target *tgt = t->targets + i;
249 dm_put_target_type(tgt->type);
254 /* free the device list */
255 free_devices(&t->devices, t->md);
257 dm_free_md_mempools(t->mempools);
263 * See if we've already got a device in the list.
265 static struct dm_dev_internal *find_device(struct list_head *l, dev_t dev)
267 struct dm_dev_internal *dd;
269 list_for_each_entry (dd, l, list)
270 if (dd->dm_dev->bdev->bd_dev == dev)
277 * If possible, this checks an area of a destination device is invalid.
279 static int device_area_is_invalid(struct dm_target *ti, struct dm_dev *dev,
280 sector_t start, sector_t len, void *data)
282 struct request_queue *q;
283 struct queue_limits *limits = data;
284 struct block_device *bdev = dev->bdev;
286 i_size_read(bdev->bd_inode) >> SECTOR_SHIFT;
287 unsigned short logical_block_size_sectors =
288 limits->logical_block_size >> SECTOR_SHIFT;
289 char b[BDEVNAME_SIZE];
292 * Some devices exist without request functions,
293 * such as loop devices not yet bound to backing files.
294 * Forbid the use of such devices.
296 q = bdev_get_queue(bdev);
297 if (!q || !q->make_request_fn) {
298 DMWARN("%s: %s is not yet initialised: "
299 "start=%llu, len=%llu, dev_size=%llu",
300 dm_device_name(ti->table->md), bdevname(bdev, b),
301 (unsigned long long)start,
302 (unsigned long long)len,
303 (unsigned long long)dev_size);
310 if ((start >= dev_size) || (start + len > dev_size)) {
311 DMWARN("%s: %s too small for target: "
312 "start=%llu, len=%llu, dev_size=%llu",
313 dm_device_name(ti->table->md), bdevname(bdev, b),
314 (unsigned long long)start,
315 (unsigned long long)len,
316 (unsigned long long)dev_size);
321 * If the target is mapped to zoned block device(s), check
322 * that the zones are not partially mapped.
324 if (bdev_zoned_model(bdev) != BLK_ZONED_NONE) {
325 unsigned int zone_sectors = bdev_zone_sectors(bdev);
327 if (start & (zone_sectors - 1)) {
328 DMWARN("%s: start=%llu not aligned to h/w zone size %u of %s",
329 dm_device_name(ti->table->md),
330 (unsigned long long)start,
331 zone_sectors, bdevname(bdev, b));
336 * Note: The last zone of a zoned block device may be smaller
337 * than other zones. So for a target mapping the end of a
338 * zoned block device with such a zone, len would not be zone
339 * aligned. We do not allow such last smaller zone to be part
340 * of the mapping here to ensure that mappings with multiple
341 * devices do not end up with a smaller zone in the middle of
344 if (len & (zone_sectors - 1)) {
345 DMWARN("%s: len=%llu not aligned to h/w zone size %u of %s",
346 dm_device_name(ti->table->md),
347 (unsigned long long)len,
348 zone_sectors, bdevname(bdev, b));
353 if (logical_block_size_sectors <= 1)
356 if (start & (logical_block_size_sectors - 1)) {
357 DMWARN("%s: start=%llu not aligned to h/w "
358 "logical block size %u of %s",
359 dm_device_name(ti->table->md),
360 (unsigned long long)start,
361 limits->logical_block_size, bdevname(bdev, b));
365 if (len & (logical_block_size_sectors - 1)) {
366 DMWARN("%s: len=%llu not aligned to h/w "
367 "logical block size %u of %s",
368 dm_device_name(ti->table->md),
369 (unsigned long long)len,
370 limits->logical_block_size, bdevname(bdev, b));
378 * This upgrades the mode on an already open dm_dev, being
379 * careful to leave things as they were if we fail to reopen the
380 * device and not to touch the existing bdev field in case
381 * it is accessed concurrently inside dm_table_any_congested().
383 static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode,
384 struct mapped_device *md)
387 struct dm_dev *old_dev, *new_dev;
389 old_dev = dd->dm_dev;
391 r = dm_get_table_device(md, dd->dm_dev->bdev->bd_dev,
392 dd->dm_dev->mode | new_mode, &new_dev);
396 dd->dm_dev = new_dev;
397 dm_put_table_device(md, old_dev);
403 * Convert the path to a device
405 dev_t dm_get_dev_t(const char *path)
408 struct block_device *bdev;
410 bdev = lookup_bdev(path);
412 dev = name_to_dev_t(path);
420 EXPORT_SYMBOL_GPL(dm_get_dev_t);
423 * Add a device to the list, or just increment the usage count if
424 * it's already present.
426 int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode,
427 struct dm_dev **result)
431 unsigned int major, minor;
433 struct dm_dev_internal *dd;
434 struct dm_table *t = ti->table;
438 if (sscanf(path, "%u:%u%c", &major, &minor, &dummy) == 2) {
439 /* Extract the major/minor numbers */
440 dev = MKDEV(major, minor);
441 if (MAJOR(dev) != major || MINOR(dev) != minor)
444 dev = dm_get_dev_t(path);
449 dd = find_device(&t->devices, dev);
451 dd = kmalloc(sizeof(*dd), GFP_KERNEL);
455 if ((r = dm_get_table_device(t->md, dev, mode, &dd->dm_dev))) {
460 refcount_set(&dd->count, 1);
461 list_add(&dd->list, &t->devices);
464 } else if (dd->dm_dev->mode != (mode | dd->dm_dev->mode)) {
465 r = upgrade_mode(dd, mode, t->md);
469 refcount_inc(&dd->count);
471 *result = dd->dm_dev;
474 EXPORT_SYMBOL(dm_get_device);
476 static int dm_set_device_limits(struct dm_target *ti, struct dm_dev *dev,
477 sector_t start, sector_t len, void *data)
479 struct queue_limits *limits = data;
480 struct block_device *bdev = dev->bdev;
481 struct request_queue *q = bdev_get_queue(bdev);
482 char b[BDEVNAME_SIZE];
485 DMWARN("%s: Cannot set limits for nonexistent device %s",
486 dm_device_name(ti->table->md), bdevname(bdev, b));
490 if (bdev_stack_limits(limits, bdev, start) < 0)
491 DMWARN("%s: adding target device %s caused an alignment inconsistency: "
492 "physical_block_size=%u, logical_block_size=%u, "
493 "alignment_offset=%u, start=%llu",
494 dm_device_name(ti->table->md), bdevname(bdev, b),
495 q->limits.physical_block_size,
496 q->limits.logical_block_size,
497 q->limits.alignment_offset,
498 (unsigned long long) start << SECTOR_SHIFT);
500 limits->zoned = blk_queue_zoned_model(q);
506 * Decrement a device's use count and remove it if necessary.
508 void dm_put_device(struct dm_target *ti, struct dm_dev *d)
511 struct list_head *devices = &ti->table->devices;
512 struct dm_dev_internal *dd;
514 list_for_each_entry(dd, devices, list) {
515 if (dd->dm_dev == d) {
521 DMWARN("%s: device %s not in table devices list",
522 dm_device_name(ti->table->md), d->name);
525 if (refcount_dec_and_test(&dd->count)) {
526 dm_put_table_device(ti->table->md, d);
531 EXPORT_SYMBOL(dm_put_device);
534 * Checks to see if the target joins onto the end of the table.
536 static int adjoin(struct dm_table *table, struct dm_target *ti)
538 struct dm_target *prev;
540 if (!table->num_targets)
543 prev = &table->targets[table->num_targets - 1];
544 return (ti->begin == (prev->begin + prev->len));
548 * Used to dynamically allocate the arg array.
550 * We do first allocation with GFP_NOIO because dm-mpath and dm-thin must
551 * process messages even if some device is suspended. These messages have a
552 * small fixed number of arguments.
554 * On the other hand, dm-switch needs to process bulk data using messages and
555 * excessive use of GFP_NOIO could cause trouble.
557 static char **realloc_argv(unsigned *size, char **old_argv)
564 new_size = *size * 2;
570 argv = kmalloc_array(new_size, sizeof(*argv), gfp);
571 if (argv && old_argv) {
572 memcpy(argv, old_argv, *size * sizeof(*argv));
581 * Destructively splits up the argument list to pass to ctr.
583 int dm_split_args(int *argc, char ***argvp, char *input)
585 char *start, *end = input, *out, **argv = NULL;
586 unsigned array_size = 0;
595 argv = realloc_argv(&array_size, argv);
600 /* Skip whitespace */
601 start = skip_spaces(end);
604 break; /* success, we hit the end */
606 /* 'out' is used to remove any back-quotes */
609 /* Everything apart from '\0' can be quoted */
610 if (*end == '\\' && *(end + 1)) {
617 break; /* end of token */
622 /* have we already filled the array ? */
623 if ((*argc + 1) > array_size) {
624 argv = realloc_argv(&array_size, argv);
629 /* we know this is whitespace */
633 /* terminate the string and put it in the array */
644 * Impose necessary and sufficient conditions on a devices's table such
645 * that any incoming bio which respects its logical_block_size can be
646 * processed successfully. If it falls across the boundary between
647 * two or more targets, the size of each piece it gets split into must
648 * be compatible with the logical_block_size of the target processing it.
650 static int validate_hardware_logical_block_alignment(struct dm_table *table,
651 struct queue_limits *limits)
654 * This function uses arithmetic modulo the logical_block_size
655 * (in units of 512-byte sectors).
657 unsigned short device_logical_block_size_sects =
658 limits->logical_block_size >> SECTOR_SHIFT;
661 * Offset of the start of the next table entry, mod logical_block_size.
663 unsigned short next_target_start = 0;
666 * Given an aligned bio that extends beyond the end of a
667 * target, how many sectors must the next target handle?
669 unsigned short remaining = 0;
671 struct dm_target *uninitialized_var(ti);
672 struct queue_limits ti_limits;
676 * Check each entry in the table in turn.
678 for (i = 0; i < dm_table_get_num_targets(table); i++) {
679 ti = dm_table_get_target(table, i);
681 blk_set_stacking_limits(&ti_limits);
683 /* combine all target devices' limits */
684 if (ti->type->iterate_devices)
685 ti->type->iterate_devices(ti, dm_set_device_limits,
689 * If the remaining sectors fall entirely within this
690 * table entry are they compatible with its logical_block_size?
692 if (remaining < ti->len &&
693 remaining & ((ti_limits.logical_block_size >>
698 (unsigned short) ((next_target_start + ti->len) &
699 (device_logical_block_size_sects - 1));
700 remaining = next_target_start ?
701 device_logical_block_size_sects - next_target_start : 0;
705 DMWARN("%s: table line %u (start sect %llu len %llu) "
706 "not aligned to h/w logical block size %u",
707 dm_device_name(table->md), i,
708 (unsigned long long) ti->begin,
709 (unsigned long long) ti->len,
710 limits->logical_block_size);
717 int dm_table_add_target(struct dm_table *t, const char *type,
718 sector_t start, sector_t len, char *params)
720 int r = -EINVAL, argc;
722 struct dm_target *tgt;
725 DMERR("%s: target type %s must appear alone in table",
726 dm_device_name(t->md), t->targets->type->name);
730 BUG_ON(t->num_targets >= t->num_allocated);
732 tgt = t->targets + t->num_targets;
733 memset(tgt, 0, sizeof(*tgt));
736 DMERR("%s: zero-length target", dm_device_name(t->md));
740 tgt->type = dm_get_target_type(type);
742 DMERR("%s: %s: unknown target type", dm_device_name(t->md), type);
746 if (dm_target_needs_singleton(tgt->type)) {
747 if (t->num_targets) {
748 tgt->error = "singleton target type must appear alone in table";
754 if (dm_target_always_writeable(tgt->type) && !(t->mode & FMODE_WRITE)) {
755 tgt->error = "target type may not be included in a read-only table";
759 if (t->immutable_target_type) {
760 if (t->immutable_target_type != tgt->type) {
761 tgt->error = "immutable target type cannot be mixed with other target types";
764 } else if (dm_target_is_immutable(tgt->type)) {
765 if (t->num_targets) {
766 tgt->error = "immutable target type cannot be mixed with other target types";
769 t->immutable_target_type = tgt->type;
772 if (dm_target_has_integrity(tgt->type))
773 t->integrity_added = 1;
778 tgt->error = "Unknown error";
781 * Does this target adjoin the previous one ?
783 if (!adjoin(t, tgt)) {
784 tgt->error = "Gap in table";
788 r = dm_split_args(&argc, &argv, params);
790 tgt->error = "couldn't split parameters (insufficient memory)";
794 r = tgt->type->ctr(tgt, argc, argv);
799 t->highs[t->num_targets++] = tgt->begin + tgt->len - 1;
801 if (!tgt->num_discard_bios && tgt->discards_supported)
802 DMWARN("%s: %s: ignoring discards_supported because num_discard_bios is zero.",
803 dm_device_name(t->md), type);
808 DMERR("%s: %s: %s", dm_device_name(t->md), type, tgt->error);
809 dm_put_target_type(tgt->type);
814 * Target argument parsing helpers.
816 static int validate_next_arg(const struct dm_arg *arg,
817 struct dm_arg_set *arg_set,
818 unsigned *value, char **error, unsigned grouped)
820 const char *arg_str = dm_shift_arg(arg_set);
824 (sscanf(arg_str, "%u%c", value, &dummy) != 1) ||
825 (*value < arg->min) ||
826 (*value > arg->max) ||
827 (grouped && arg_set->argc < *value)) {
835 int dm_read_arg(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, 0);
840 EXPORT_SYMBOL(dm_read_arg);
842 int dm_read_arg_group(const struct dm_arg *arg, struct dm_arg_set *arg_set,
843 unsigned *value, char **error)
845 return validate_next_arg(arg, arg_set, value, error, 1);
847 EXPORT_SYMBOL(dm_read_arg_group);
849 const char *dm_shift_arg(struct dm_arg_set *as)
862 EXPORT_SYMBOL(dm_shift_arg);
864 void dm_consume_args(struct dm_arg_set *as, unsigned num_args)
866 BUG_ON(as->argc < num_args);
867 as->argc -= num_args;
868 as->argv += num_args;
870 EXPORT_SYMBOL(dm_consume_args);
872 static bool __table_type_bio_based(enum dm_queue_mode table_type)
874 return (table_type == DM_TYPE_BIO_BASED ||
875 table_type == DM_TYPE_DAX_BIO_BASED);
878 static bool __table_type_request_based(enum dm_queue_mode table_type)
880 return table_type == DM_TYPE_REQUEST_BASED;
883 void dm_table_set_type(struct dm_table *t, enum dm_queue_mode type)
887 EXPORT_SYMBOL_GPL(dm_table_set_type);
889 /* validate the dax capability of the target device span */
890 int device_not_dax_capable(struct dm_target *ti, struct dm_dev *dev,
891 sector_t start, sector_t len, void *data)
893 int blocksize = *(int *) data, id;
896 id = dax_read_lock();
897 rc = !dax_supported(dev->dax_dev, dev->bdev, blocksize, start, len);
903 /* Check devices support synchronous DAX */
904 static int device_not_dax_synchronous_capable(struct dm_target *ti, struct dm_dev *dev,
905 sector_t start, sector_t len, void *data)
907 return !dev->dax_dev || !dax_synchronous(dev->dax_dev);
910 bool dm_table_supports_dax(struct dm_table *t,
911 iterate_devices_callout_fn iterate_fn, int *blocksize)
913 struct dm_target *ti;
916 /* Ensure that all targets support DAX. */
917 for (i = 0; i < dm_table_get_num_targets(t); i++) {
918 ti = dm_table_get_target(t, i);
920 if (!ti->type->direct_access)
923 if (!ti->type->iterate_devices ||
924 ti->type->iterate_devices(ti, iterate_fn, blocksize))
931 static int device_is_rq_stackable(struct dm_target *ti, struct dm_dev *dev,
932 sector_t start, sector_t len, void *data)
934 struct block_device *bdev = dev->bdev;
935 struct request_queue *q = bdev_get_queue(bdev);
937 /* request-based cannot stack on partitions! */
938 if (bdev != bdev->bd_contains)
941 return queue_is_mq(q);
944 static int dm_table_determine_type(struct dm_table *t)
947 unsigned bio_based = 0, request_based = 0, hybrid = 0;
948 struct dm_target *tgt;
949 struct list_head *devices = dm_table_get_devices(t);
950 enum dm_queue_mode live_md_type = dm_get_md_type(t->md);
951 int page_size = PAGE_SIZE;
953 if (t->type != DM_TYPE_NONE) {
954 /* target already set the table's type */
955 if (t->type == DM_TYPE_BIO_BASED) {
956 /* possibly upgrade to a variant of bio-based */
957 goto verify_bio_based;
959 BUG_ON(t->type == DM_TYPE_DAX_BIO_BASED);
960 goto verify_rq_based;
963 for (i = 0; i < t->num_targets; i++) {
964 tgt = t->targets + i;
965 if (dm_target_hybrid(tgt))
967 else if (dm_target_request_based(tgt))
972 if (bio_based && request_based) {
973 DMERR("Inconsistent table: different target types"
974 " can't be mixed up");
979 if (hybrid && !bio_based && !request_based) {
981 * The targets can work either way.
982 * Determine the type from the live device.
983 * Default to bio-based if device is new.
985 if (__table_type_request_based(live_md_type))
993 /* We must use this table as bio-based */
994 t->type = DM_TYPE_BIO_BASED;
995 if (dm_table_supports_dax(t, device_not_dax_capable, &page_size) ||
996 (list_empty(devices) && live_md_type == DM_TYPE_DAX_BIO_BASED)) {
997 t->type = DM_TYPE_DAX_BIO_BASED;
1002 BUG_ON(!request_based); /* No targets in this table */
1004 t->type = DM_TYPE_REQUEST_BASED;
1008 * Request-based dm supports only tables that have a single target now.
1009 * To support multiple targets, request splitting support is needed,
1010 * and that needs lots of changes in the block-layer.
1011 * (e.g. request completion process for partial completion.)
1013 if (t->num_targets > 1) {
1014 DMERR("request-based DM doesn't support multiple targets");
1018 if (list_empty(devices)) {
1020 struct dm_table *live_table = dm_get_live_table(t->md, &srcu_idx);
1022 /* inherit live table's type */
1024 t->type = live_table->type;
1025 dm_put_live_table(t->md, srcu_idx);
1029 tgt = dm_table_get_immutable_target(t);
1031 DMERR("table load rejected: immutable target is required");
1033 } else if (tgt->max_io_len) {
1034 DMERR("table load rejected: immutable target that splits IO is not supported");
1038 /* Non-request-stackable devices can't be used for request-based dm */
1039 if (!tgt->type->iterate_devices ||
1040 !tgt->type->iterate_devices(tgt, device_is_rq_stackable, NULL)) {
1041 DMERR("table load rejected: including non-request-stackable devices");
1048 enum dm_queue_mode dm_table_get_type(struct dm_table *t)
1053 struct target_type *dm_table_get_immutable_target_type(struct dm_table *t)
1055 return t->immutable_target_type;
1058 struct dm_target *dm_table_get_immutable_target(struct dm_table *t)
1060 /* Immutable target is implicitly a singleton */
1061 if (t->num_targets > 1 ||
1062 !dm_target_is_immutable(t->targets[0].type))
1068 struct dm_target *dm_table_get_wildcard_target(struct dm_table *t)
1070 struct dm_target *ti;
1073 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1074 ti = dm_table_get_target(t, i);
1075 if (dm_target_is_wildcard(ti->type))
1082 bool dm_table_bio_based(struct dm_table *t)
1084 return __table_type_bio_based(dm_table_get_type(t));
1087 bool dm_table_request_based(struct dm_table *t)
1089 return __table_type_request_based(dm_table_get_type(t));
1092 static int dm_table_alloc_md_mempools(struct dm_table *t, struct mapped_device *md)
1094 enum dm_queue_mode type = dm_table_get_type(t);
1095 unsigned per_io_data_size = 0;
1096 unsigned min_pool_size = 0;
1097 struct dm_target *ti;
1100 if (unlikely(type == DM_TYPE_NONE)) {
1101 DMWARN("no table type is set, can't allocate mempools");
1105 if (__table_type_bio_based(type))
1106 for (i = 0; i < t->num_targets; i++) {
1107 ti = t->targets + i;
1108 per_io_data_size = max(per_io_data_size, ti->per_io_data_size);
1109 min_pool_size = max(min_pool_size, ti->num_flush_bios);
1112 t->mempools = dm_alloc_md_mempools(md, type, t->integrity_supported,
1113 per_io_data_size, min_pool_size);
1120 void dm_table_free_md_mempools(struct dm_table *t)
1122 dm_free_md_mempools(t->mempools);
1126 struct dm_md_mempools *dm_table_get_md_mempools(struct dm_table *t)
1131 static int setup_indexes(struct dm_table *t)
1134 unsigned int total = 0;
1137 /* allocate the space for *all* the indexes */
1138 for (i = t->depth - 2; i >= 0; i--) {
1139 t->counts[i] = dm_div_up(t->counts[i + 1], CHILDREN_PER_NODE);
1140 total += t->counts[i];
1143 indexes = (sector_t *) dm_vcalloc(total, (unsigned long) NODE_SIZE);
1147 /* set up internal nodes, bottom-up */
1148 for (i = t->depth - 2; i >= 0; i--) {
1149 t->index[i] = indexes;
1150 indexes += (KEYS_PER_NODE * t->counts[i]);
1151 setup_btree_index(i, t);
1158 * Builds the btree to index the map.
1160 static int dm_table_build_index(struct dm_table *t)
1163 unsigned int leaf_nodes;
1165 /* how many indexes will the btree have ? */
1166 leaf_nodes = dm_div_up(t->num_targets, KEYS_PER_NODE);
1167 t->depth = 1 + int_log(leaf_nodes, CHILDREN_PER_NODE);
1169 /* leaf layer has already been set up */
1170 t->counts[t->depth - 1] = leaf_nodes;
1171 t->index[t->depth - 1] = t->highs;
1174 r = setup_indexes(t);
1179 static bool integrity_profile_exists(struct gendisk *disk)
1181 return !!blk_get_integrity(disk);
1185 * Get a disk whose integrity profile reflects the table's profile.
1186 * Returns NULL if integrity support was inconsistent or unavailable.
1188 static struct gendisk * dm_table_get_integrity_disk(struct dm_table *t)
1190 struct list_head *devices = dm_table_get_devices(t);
1191 struct dm_dev_internal *dd = NULL;
1192 struct gendisk *prev_disk = NULL, *template_disk = NULL;
1195 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1196 struct dm_target *ti = dm_table_get_target(t, i);
1197 if (!dm_target_passes_integrity(ti->type))
1201 list_for_each_entry(dd, devices, list) {
1202 template_disk = dd->dm_dev->bdev->bd_disk;
1203 if (!integrity_profile_exists(template_disk))
1205 else if (prev_disk &&
1206 blk_integrity_compare(prev_disk, template_disk) < 0)
1208 prev_disk = template_disk;
1211 return template_disk;
1215 DMWARN("%s: integrity not set: %s and %s profile mismatch",
1216 dm_device_name(t->md),
1217 prev_disk->disk_name,
1218 template_disk->disk_name);
1223 * Register the mapped device for blk_integrity support if the
1224 * underlying devices have an integrity profile. But all devices may
1225 * not have matching profiles (checking all devices isn't reliable
1226 * during table load because this table may use other DM device(s) which
1227 * must be resumed before they will have an initialized integity
1228 * profile). Consequently, stacked DM devices force a 2 stage integrity
1229 * profile validation: First pass during table load, final pass during
1232 static int dm_table_register_integrity(struct dm_table *t)
1234 struct mapped_device *md = t->md;
1235 struct gendisk *template_disk = NULL;
1237 /* If target handles integrity itself do not register it here. */
1238 if (t->integrity_added)
1241 template_disk = dm_table_get_integrity_disk(t);
1245 if (!integrity_profile_exists(dm_disk(md))) {
1246 t->integrity_supported = true;
1248 * Register integrity profile during table load; we can do
1249 * this because the final profile must match during resume.
1251 blk_integrity_register(dm_disk(md),
1252 blk_get_integrity(template_disk));
1257 * If DM device already has an initialized integrity
1258 * profile the new profile should not conflict.
1260 if (blk_integrity_compare(dm_disk(md), template_disk) < 0) {
1261 DMWARN("%s: conflict with existing integrity profile: "
1262 "%s profile mismatch",
1263 dm_device_name(t->md),
1264 template_disk->disk_name);
1268 /* Preserve existing integrity profile */
1269 t->integrity_supported = true;
1274 * Prepares the table for use by building the indices,
1275 * setting the type, and allocating mempools.
1277 int dm_table_complete(struct dm_table *t)
1281 r = dm_table_determine_type(t);
1283 DMERR("unable to determine table type");
1287 r = dm_table_build_index(t);
1289 DMERR("unable to build btrees");
1293 r = dm_table_register_integrity(t);
1295 DMERR("could not register integrity profile.");
1299 r = dm_table_alloc_md_mempools(t, t->md);
1301 DMERR("unable to allocate mempools");
1306 static DEFINE_MUTEX(_event_lock);
1307 void dm_table_event_callback(struct dm_table *t,
1308 void (*fn)(void *), void *context)
1310 mutex_lock(&_event_lock);
1312 t->event_context = context;
1313 mutex_unlock(&_event_lock);
1316 void dm_table_event(struct dm_table *t)
1318 mutex_lock(&_event_lock);
1320 t->event_fn(t->event_context);
1321 mutex_unlock(&_event_lock);
1323 EXPORT_SYMBOL(dm_table_event);
1325 inline sector_t dm_table_get_size(struct dm_table *t)
1327 return t->num_targets ? (t->highs[t->num_targets - 1] + 1) : 0;
1329 EXPORT_SYMBOL(dm_table_get_size);
1331 struct dm_target *dm_table_get_target(struct dm_table *t, unsigned int index)
1333 if (index >= t->num_targets)
1336 return t->targets + index;
1340 * Search the btree for the correct target.
1342 * Caller should check returned pointer for NULL
1343 * to trap I/O beyond end of device.
1345 struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector)
1347 unsigned int l, n = 0, k = 0;
1350 if (unlikely(sector >= dm_table_get_size(t)))
1353 for (l = 0; l < t->depth; l++) {
1354 n = get_child(n, k);
1355 node = get_node(t, l, n);
1357 for (k = 0; k < KEYS_PER_NODE; k++)
1358 if (node[k] >= sector)
1362 return &t->targets[(KEYS_PER_NODE * n) + k];
1366 * type->iterate_devices() should be called when the sanity check needs to
1367 * iterate and check all underlying data devices. iterate_devices() will
1368 * iterate all underlying data devices until it encounters a non-zero return
1369 * code, returned by whether the input iterate_devices_callout_fn, or
1370 * iterate_devices() itself internally.
1372 * For some target type (e.g. dm-stripe), one call of iterate_devices() may
1373 * iterate multiple underlying devices internally, in which case a non-zero
1374 * return code returned by iterate_devices_callout_fn will stop the iteration
1377 * Cases requiring _any_ underlying device supporting some kind of attribute,
1378 * should use the iteration structure like dm_table_any_dev_attr(), or call
1379 * it directly. @func should handle semantics of positive examples, e.g.
1380 * capable of something.
1382 * Cases requiring _all_ underlying devices supporting some kind of attribute,
1383 * should use the iteration structure like dm_table_supports_nowait() or
1384 * dm_table_supports_discards(). Or introduce dm_table_all_devs_attr() that
1385 * uses an @anti_func that handle semantics of counter examples, e.g. not
1386 * capable of something. So: return !dm_table_any_dev_attr(t, anti_func, data);
1388 static bool dm_table_any_dev_attr(struct dm_table *t,
1389 iterate_devices_callout_fn func, void *data)
1391 struct dm_target *ti;
1394 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1395 ti = dm_table_get_target(t, i);
1397 if (ti->type->iterate_devices &&
1398 ti->type->iterate_devices(ti, func, data))
1405 static int count_device(struct dm_target *ti, struct dm_dev *dev,
1406 sector_t start, sector_t len, void *data)
1408 unsigned *num_devices = data;
1416 * Check whether a table has no data devices attached using each
1417 * target's iterate_devices method.
1418 * Returns false if the result is unknown because a target doesn't
1419 * support iterate_devices.
1421 bool dm_table_has_no_data_devices(struct dm_table *table)
1423 struct dm_target *ti;
1424 unsigned i, num_devices;
1426 for (i = 0; i < dm_table_get_num_targets(table); i++) {
1427 ti = dm_table_get_target(table, i);
1429 if (!ti->type->iterate_devices)
1433 ti->type->iterate_devices(ti, count_device, &num_devices);
1441 static int device_not_zoned_model(struct dm_target *ti, struct dm_dev *dev,
1442 sector_t start, sector_t len, void *data)
1444 struct request_queue *q = bdev_get_queue(dev->bdev);
1445 enum blk_zoned_model *zoned_model = data;
1447 return !q || blk_queue_zoned_model(q) != *zoned_model;
1450 static bool dm_table_supports_zoned_model(struct dm_table *t,
1451 enum blk_zoned_model zoned_model)
1453 struct dm_target *ti;
1456 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1457 ti = dm_table_get_target(t, i);
1459 if (zoned_model == BLK_ZONED_HM &&
1460 !dm_target_supports_zoned_hm(ti->type))
1463 if (!ti->type->iterate_devices ||
1464 ti->type->iterate_devices(ti, device_not_zoned_model, &zoned_model))
1471 static int device_not_matches_zone_sectors(struct dm_target *ti, struct dm_dev *dev,
1472 sector_t start, sector_t len, void *data)
1474 struct request_queue *q = bdev_get_queue(dev->bdev);
1475 unsigned int *zone_sectors = data;
1477 return !q || blk_queue_zone_sectors(q) != *zone_sectors;
1480 static int validate_hardware_zoned_model(struct dm_table *table,
1481 enum blk_zoned_model zoned_model,
1482 unsigned int zone_sectors)
1484 if (zoned_model == BLK_ZONED_NONE)
1487 if (!dm_table_supports_zoned_model(table, zoned_model)) {
1488 DMERR("%s: zoned model is not consistent across all devices",
1489 dm_device_name(table->md));
1493 /* Check zone size validity and compatibility */
1494 if (!zone_sectors || !is_power_of_2(zone_sectors))
1497 if (dm_table_any_dev_attr(table, device_not_matches_zone_sectors, &zone_sectors)) {
1498 DMERR("%s: zone sectors is not consistent across all devices",
1499 dm_device_name(table->md));
1507 * Establish the new table's queue_limits and validate them.
1509 int dm_calculate_queue_limits(struct dm_table *table,
1510 struct queue_limits *limits)
1512 struct dm_target *ti;
1513 struct queue_limits ti_limits;
1515 enum blk_zoned_model zoned_model = BLK_ZONED_NONE;
1516 unsigned int zone_sectors = 0;
1518 blk_set_stacking_limits(limits);
1520 for (i = 0; i < dm_table_get_num_targets(table); i++) {
1521 blk_set_stacking_limits(&ti_limits);
1523 ti = dm_table_get_target(table, i);
1525 if (!ti->type->iterate_devices)
1526 goto combine_limits;
1529 * Combine queue limits of all the devices this target uses.
1531 ti->type->iterate_devices(ti, dm_set_device_limits,
1534 if (zoned_model == BLK_ZONED_NONE && ti_limits.zoned != BLK_ZONED_NONE) {
1536 * After stacking all limits, validate all devices
1537 * in table support this zoned model and zone sectors.
1539 zoned_model = ti_limits.zoned;
1540 zone_sectors = ti_limits.chunk_sectors;
1543 /* Set I/O hints portion of queue limits */
1544 if (ti->type->io_hints)
1545 ti->type->io_hints(ti, &ti_limits);
1548 * Check each device area is consistent with the target's
1549 * overall queue limits.
1551 if (ti->type->iterate_devices(ti, device_area_is_invalid,
1557 * Merge this target's queue limits into the overall limits
1560 if (blk_stack_limits(limits, &ti_limits, 0) < 0)
1561 DMWARN("%s: adding target device "
1562 "(start sect %llu len %llu) "
1563 "caused an alignment inconsistency",
1564 dm_device_name(table->md),
1565 (unsigned long long) ti->begin,
1566 (unsigned long long) ti->len);
1569 * FIXME: this should likely be moved to blk_stack_limits(), would
1570 * also eliminate limits->zoned stacking hack in dm_set_device_limits()
1572 if (limits->zoned == BLK_ZONED_NONE && ti_limits.zoned != BLK_ZONED_NONE) {
1574 * By default, the stacked limits zoned model is set to
1575 * BLK_ZONED_NONE in blk_set_stacking_limits(). Update
1576 * this model using the first target model reported
1577 * that is not BLK_ZONED_NONE. This will be either the
1578 * first target device zoned model or the model reported
1579 * by the target .io_hints.
1581 limits->zoned = ti_limits.zoned;
1586 * Verify that the zoned model and zone sectors, as determined before
1587 * any .io_hints override, are the same across all devices in the table.
1588 * - this is especially relevant if .io_hints is emulating a disk-managed
1589 * zoned model (aka BLK_ZONED_NONE) on host-managed zoned block devices.
1592 if (limits->zoned != BLK_ZONED_NONE) {
1594 * ...IF the above limits stacking determined a zoned model
1595 * validate that all of the table's devices conform to it.
1597 zoned_model = limits->zoned;
1598 zone_sectors = limits->chunk_sectors;
1600 if (validate_hardware_zoned_model(table, zoned_model, zone_sectors))
1603 return validate_hardware_logical_block_alignment(table, limits);
1607 * Verify that all devices have an integrity profile that matches the
1608 * DM device's registered integrity profile. If the profiles don't
1609 * match then unregister the DM device's integrity profile.
1611 static void dm_table_verify_integrity(struct dm_table *t)
1613 struct gendisk *template_disk = NULL;
1615 if (t->integrity_added)
1618 if (t->integrity_supported) {
1620 * Verify that the original integrity profile
1621 * matches all the devices in this table.
1623 template_disk = dm_table_get_integrity_disk(t);
1624 if (template_disk &&
1625 blk_integrity_compare(dm_disk(t->md), template_disk) >= 0)
1629 if (integrity_profile_exists(dm_disk(t->md))) {
1630 DMWARN("%s: unable to establish an integrity profile",
1631 dm_device_name(t->md));
1632 blk_integrity_unregister(dm_disk(t->md));
1636 static int device_flush_capable(struct dm_target *ti, struct dm_dev *dev,
1637 sector_t start, sector_t len, void *data)
1639 unsigned long flush = (unsigned long) data;
1640 struct request_queue *q = bdev_get_queue(dev->bdev);
1642 return q && (q->queue_flags & flush);
1645 static bool dm_table_supports_flush(struct dm_table *t, unsigned long flush)
1647 struct dm_target *ti;
1651 * Require at least one underlying device to support flushes.
1652 * t->devices includes internal dm devices such as mirror logs
1653 * so we need to use iterate_devices here, which targets
1654 * supporting flushes must provide.
1656 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1657 ti = dm_table_get_target(t, i);
1659 if (!ti->num_flush_bios)
1662 if (ti->flush_supported)
1665 if (ti->type->iterate_devices &&
1666 ti->type->iterate_devices(ti, device_flush_capable, (void *) flush))
1673 static int device_dax_write_cache_enabled(struct dm_target *ti,
1674 struct dm_dev *dev, sector_t start,
1675 sector_t len, void *data)
1677 struct dax_device *dax_dev = dev->dax_dev;
1682 if (dax_write_cache_enabled(dax_dev))
1687 static int device_is_rotational(struct dm_target *ti, struct dm_dev *dev,
1688 sector_t start, sector_t len, void *data)
1690 struct request_queue *q = bdev_get_queue(dev->bdev);
1692 return q && !blk_queue_nonrot(q);
1695 static int device_is_not_random(struct dm_target *ti, struct dm_dev *dev,
1696 sector_t start, sector_t len, void *data)
1698 struct request_queue *q = bdev_get_queue(dev->bdev);
1700 return q && !blk_queue_add_random(q);
1703 static int device_not_write_same_capable(struct dm_target *ti, struct dm_dev *dev,
1704 sector_t start, sector_t len, void *data)
1706 struct request_queue *q = bdev_get_queue(dev->bdev);
1708 return q && !q->limits.max_write_same_sectors;
1711 static bool dm_table_supports_write_same(struct dm_table *t)
1713 struct dm_target *ti;
1716 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1717 ti = dm_table_get_target(t, i);
1719 if (!ti->num_write_same_bios)
1722 if (!ti->type->iterate_devices ||
1723 ti->type->iterate_devices(ti, device_not_write_same_capable, NULL))
1730 static int device_not_write_zeroes_capable(struct dm_target *ti, struct dm_dev *dev,
1731 sector_t start, sector_t len, void *data)
1733 struct request_queue *q = bdev_get_queue(dev->bdev);
1735 return q && !q->limits.max_write_zeroes_sectors;
1738 static bool dm_table_supports_write_zeroes(struct dm_table *t)
1740 struct dm_target *ti;
1743 while (i < dm_table_get_num_targets(t)) {
1744 ti = dm_table_get_target(t, i++);
1746 if (!ti->num_write_zeroes_bios)
1749 if (!ti->type->iterate_devices ||
1750 ti->type->iterate_devices(ti, device_not_write_zeroes_capable, NULL))
1757 static int device_not_discard_capable(struct dm_target *ti, struct dm_dev *dev,
1758 sector_t start, sector_t len, void *data)
1760 struct request_queue *q = bdev_get_queue(dev->bdev);
1762 return q && !blk_queue_discard(q);
1765 static bool dm_table_supports_discards(struct dm_table *t)
1767 struct dm_target *ti;
1770 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1771 ti = dm_table_get_target(t, i);
1773 if (!ti->num_discard_bios)
1777 * Either the target provides discard support (as implied by setting
1778 * 'discards_supported') or it relies on _all_ data devices having
1781 if (!ti->discards_supported &&
1782 (!ti->type->iterate_devices ||
1783 ti->type->iterate_devices(ti, device_not_discard_capable, NULL)))
1790 static int device_not_secure_erase_capable(struct dm_target *ti,
1791 struct dm_dev *dev, sector_t start,
1792 sector_t len, void *data)
1794 struct request_queue *q = bdev_get_queue(dev->bdev);
1796 return q && !blk_queue_secure_erase(q);
1799 static bool dm_table_supports_secure_erase(struct dm_table *t)
1801 struct dm_target *ti;
1804 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1805 ti = dm_table_get_target(t, i);
1807 if (!ti->num_secure_erase_bios)
1810 if (!ti->type->iterate_devices ||
1811 ti->type->iterate_devices(ti, device_not_secure_erase_capable, NULL))
1818 static int device_requires_stable_pages(struct dm_target *ti,
1819 struct dm_dev *dev, sector_t start,
1820 sector_t len, void *data)
1822 struct request_queue *q = bdev_get_queue(dev->bdev);
1824 return q && bdi_cap_stable_pages_required(q->backing_dev_info);
1827 void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q,
1828 struct queue_limits *limits)
1830 bool wc = false, fua = false;
1831 int page_size = PAGE_SIZE;
1834 * Copy table's limits to the DM device's request_queue
1836 q->limits = *limits;
1838 if (!dm_table_supports_discards(t)) {
1839 blk_queue_flag_clear(QUEUE_FLAG_DISCARD, q);
1840 /* Must also clear discard limits... */
1841 q->limits.max_discard_sectors = 0;
1842 q->limits.max_hw_discard_sectors = 0;
1843 q->limits.discard_granularity = 0;
1844 q->limits.discard_alignment = 0;
1845 q->limits.discard_misaligned = 0;
1847 blk_queue_flag_set(QUEUE_FLAG_DISCARD, q);
1849 if (dm_table_supports_secure_erase(t))
1850 blk_queue_flag_set(QUEUE_FLAG_SECERASE, q);
1852 if (dm_table_supports_flush(t, (1UL << QUEUE_FLAG_WC))) {
1854 if (dm_table_supports_flush(t, (1UL << QUEUE_FLAG_FUA)))
1857 blk_queue_write_cache(q, wc, fua);
1859 if (dm_table_supports_dax(t, device_not_dax_capable, &page_size)) {
1860 blk_queue_flag_set(QUEUE_FLAG_DAX, q);
1861 if (dm_table_supports_dax(t, device_not_dax_synchronous_capable, NULL))
1862 set_dax_synchronous(t->md->dax_dev);
1865 blk_queue_flag_clear(QUEUE_FLAG_DAX, q);
1867 if (dm_table_any_dev_attr(t, device_dax_write_cache_enabled, NULL))
1868 dax_write_cache(t->md->dax_dev, true);
1870 /* Ensure that all underlying devices are non-rotational. */
1871 if (dm_table_any_dev_attr(t, device_is_rotational, NULL))
1872 blk_queue_flag_clear(QUEUE_FLAG_NONROT, q);
1874 blk_queue_flag_set(QUEUE_FLAG_NONROT, q);
1876 if (!dm_table_supports_write_same(t))
1877 q->limits.max_write_same_sectors = 0;
1878 if (!dm_table_supports_write_zeroes(t))
1879 q->limits.max_write_zeroes_sectors = 0;
1881 dm_table_verify_integrity(t);
1884 * Some devices don't use blk_integrity but still want stable pages
1885 * because they do their own checksumming.
1886 * If any underlying device requires stable pages, a table must require
1887 * them as well. Only targets that support iterate_devices are considered:
1888 * don't want error, zero, etc to require stable pages.
1890 if (dm_table_any_dev_attr(t, device_requires_stable_pages, NULL))
1891 q->backing_dev_info->capabilities |= BDI_CAP_STABLE_WRITES;
1893 q->backing_dev_info->capabilities &= ~BDI_CAP_STABLE_WRITES;
1896 * Determine whether or not this queue's I/O timings contribute
1897 * to the entropy pool, Only request-based targets use this.
1898 * Clear QUEUE_FLAG_ADD_RANDOM if any underlying device does not
1901 if (blk_queue_add_random(q) &&
1902 dm_table_any_dev_attr(t, device_is_not_random, NULL))
1903 blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, q);
1906 * For a zoned target, the number of zones should be updated for the
1907 * correct value to be exposed in sysfs queue/nr_zones. For a BIO based
1908 * target, this is all that is needed. For a request based target, the
1909 * queue zone bitmaps must also be updated.
1910 * Use blk_revalidate_disk_zones() to handle this.
1912 if (blk_queue_is_zoned(q))
1913 blk_revalidate_disk_zones(t->md->disk);
1915 /* Allow reads to exceed readahead limits */
1916 q->backing_dev_info->io_pages = limits->max_sectors >> (PAGE_SHIFT - 9);
1919 unsigned int dm_table_get_num_targets(struct dm_table *t)
1921 return t->num_targets;
1924 struct list_head *dm_table_get_devices(struct dm_table *t)
1929 fmode_t dm_table_get_mode(struct dm_table *t)
1933 EXPORT_SYMBOL(dm_table_get_mode);
1941 static void suspend_targets(struct dm_table *t, enum suspend_mode mode)
1943 int i = t->num_targets;
1944 struct dm_target *ti = t->targets;
1946 lockdep_assert_held(&t->md->suspend_lock);
1951 if (ti->type->presuspend)
1952 ti->type->presuspend(ti);
1954 case PRESUSPEND_UNDO:
1955 if (ti->type->presuspend_undo)
1956 ti->type->presuspend_undo(ti);
1959 if (ti->type->postsuspend)
1960 ti->type->postsuspend(ti);
1967 void dm_table_presuspend_targets(struct dm_table *t)
1972 suspend_targets(t, PRESUSPEND);
1975 void dm_table_presuspend_undo_targets(struct dm_table *t)
1980 suspend_targets(t, PRESUSPEND_UNDO);
1983 void dm_table_postsuspend_targets(struct dm_table *t)
1988 suspend_targets(t, POSTSUSPEND);
1991 int dm_table_resume_targets(struct dm_table *t)
1995 lockdep_assert_held(&t->md->suspend_lock);
1997 for (i = 0; i < t->num_targets; i++) {
1998 struct dm_target *ti = t->targets + i;
2000 if (!ti->type->preresume)
2003 r = ti->type->preresume(ti);
2005 DMERR("%s: %s: preresume failed, error = %d",
2006 dm_device_name(t->md), ti->type->name, r);
2011 for (i = 0; i < t->num_targets; i++) {
2012 struct dm_target *ti = t->targets + i;
2014 if (ti->type->resume)
2015 ti->type->resume(ti);
2021 void dm_table_add_target_callbacks(struct dm_table *t, struct dm_target_callbacks *cb)
2023 list_add(&cb->list, &t->target_callbacks);
2025 EXPORT_SYMBOL_GPL(dm_table_add_target_callbacks);
2027 int dm_table_any_congested(struct dm_table *t, int bdi_bits)
2029 struct dm_dev_internal *dd;
2030 struct list_head *devices = dm_table_get_devices(t);
2031 struct dm_target_callbacks *cb;
2034 list_for_each_entry(dd, devices, list) {
2035 struct request_queue *q = bdev_get_queue(dd->dm_dev->bdev);
2036 char b[BDEVNAME_SIZE];
2039 r |= bdi_congested(q->backing_dev_info, bdi_bits);
2041 DMWARN_LIMIT("%s: any_congested: nonexistent device %s",
2042 dm_device_name(t->md),
2043 bdevname(dd->dm_dev->bdev, b));
2046 list_for_each_entry(cb, &t->target_callbacks, list)
2047 if (cb->congested_fn)
2048 r |= cb->congested_fn(cb, bdi_bits);
2053 struct mapped_device *dm_table_get_md(struct dm_table *t)
2057 EXPORT_SYMBOL(dm_table_get_md);
2059 const char *dm_table_device_name(struct dm_table *t)
2061 return dm_device_name(t->md);
2063 EXPORT_SYMBOL_GPL(dm_table_device_name);
2065 void dm_table_run_md_queue_async(struct dm_table *t)
2067 struct mapped_device *md;
2068 struct request_queue *queue;
2070 if (!dm_table_request_based(t))
2073 md = dm_table_get_md(t);
2074 queue = dm_get_md_queue(md);
2076 blk_mq_run_hw_queues(queue, true);
2078 EXPORT_SYMBOL(dm_table_run_md_queue_async);