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 queue_limits *limits = data;
283 struct block_device *bdev = dev->bdev;
285 i_size_read(bdev->bd_inode) >> SECTOR_SHIFT;
286 unsigned short logical_block_size_sectors =
287 limits->logical_block_size >> SECTOR_SHIFT;
288 char b[BDEVNAME_SIZE];
293 if ((start >= dev_size) || (start + len > dev_size)) {
294 DMWARN("%s: %s too small for target: "
295 "start=%llu, len=%llu, dev_size=%llu",
296 dm_device_name(ti->table->md), bdevname(bdev, b),
297 (unsigned long long)start,
298 (unsigned long long)len,
299 (unsigned long long)dev_size);
304 * If the target is mapped to zoned block device(s), check
305 * that the zones are not partially mapped.
307 if (bdev_zoned_model(bdev) != BLK_ZONED_NONE) {
308 unsigned int zone_sectors = bdev_zone_sectors(bdev);
310 if (start & (zone_sectors - 1)) {
311 DMWARN("%s: start=%llu not aligned to h/w zone size %u of %s",
312 dm_device_name(ti->table->md),
313 (unsigned long long)start,
314 zone_sectors, bdevname(bdev, b));
319 * Note: The last zone of a zoned block device may be smaller
320 * than other zones. So for a target mapping the end of a
321 * zoned block device with such a zone, len would not be zone
322 * aligned. We do not allow such last smaller zone to be part
323 * of the mapping here to ensure that mappings with multiple
324 * devices do not end up with a smaller zone in the middle of
327 if (len & (zone_sectors - 1)) {
328 DMWARN("%s: len=%llu not aligned to h/w zone size %u of %s",
329 dm_device_name(ti->table->md),
330 (unsigned long long)len,
331 zone_sectors, bdevname(bdev, b));
336 if (logical_block_size_sectors <= 1)
339 if (start & (logical_block_size_sectors - 1)) {
340 DMWARN("%s: start=%llu not aligned to h/w "
341 "logical block size %u of %s",
342 dm_device_name(ti->table->md),
343 (unsigned long long)start,
344 limits->logical_block_size, bdevname(bdev, b));
348 if (len & (logical_block_size_sectors - 1)) {
349 DMWARN("%s: len=%llu not aligned to h/w "
350 "logical block size %u of %s",
351 dm_device_name(ti->table->md),
352 (unsigned long long)len,
353 limits->logical_block_size, bdevname(bdev, b));
361 * This upgrades the mode on an already open dm_dev, being
362 * careful to leave things as they were if we fail to reopen the
363 * device and not to touch the existing bdev field in case
364 * it is accessed concurrently inside dm_table_any_congested().
366 static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode,
367 struct mapped_device *md)
370 struct dm_dev *old_dev, *new_dev;
372 old_dev = dd->dm_dev;
374 r = dm_get_table_device(md, dd->dm_dev->bdev->bd_dev,
375 dd->dm_dev->mode | new_mode, &new_dev);
379 dd->dm_dev = new_dev;
380 dm_put_table_device(md, old_dev);
386 * Convert the path to a device
388 dev_t dm_get_dev_t(const char *path)
391 struct block_device *bdev;
393 bdev = lookup_bdev(path);
395 dev = name_to_dev_t(path);
403 EXPORT_SYMBOL_GPL(dm_get_dev_t);
406 * Add a device to the list, or just increment the usage count if
407 * it's already present.
409 int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode,
410 struct dm_dev **result)
414 struct dm_dev_internal *dd;
415 struct dm_table *t = ti->table;
419 dev = dm_get_dev_t(path);
423 dd = find_device(&t->devices, dev);
425 dd = kmalloc(sizeof(*dd), GFP_KERNEL);
429 if ((r = dm_get_table_device(t->md, dev, mode, &dd->dm_dev))) {
434 refcount_set(&dd->count, 1);
435 list_add(&dd->list, &t->devices);
438 } else if (dd->dm_dev->mode != (mode | dd->dm_dev->mode)) {
439 r = upgrade_mode(dd, mode, t->md);
443 refcount_inc(&dd->count);
445 *result = dd->dm_dev;
448 EXPORT_SYMBOL(dm_get_device);
450 static int dm_set_device_limits(struct dm_target *ti, struct dm_dev *dev,
451 sector_t start, sector_t len, void *data)
453 struct queue_limits *limits = data;
454 struct block_device *bdev = dev->bdev;
455 struct request_queue *q = bdev_get_queue(bdev);
456 char b[BDEVNAME_SIZE];
459 DMWARN("%s: Cannot set limits for nonexistent device %s",
460 dm_device_name(ti->table->md), bdevname(bdev, b));
464 if (bdev_stack_limits(limits, bdev, start) < 0)
465 DMWARN("%s: adding target device %s caused an alignment inconsistency: "
466 "physical_block_size=%u, logical_block_size=%u, "
467 "alignment_offset=%u, start=%llu",
468 dm_device_name(ti->table->md), bdevname(bdev, b),
469 q->limits.physical_block_size,
470 q->limits.logical_block_size,
471 q->limits.alignment_offset,
472 (unsigned long long) start << SECTOR_SHIFT);
474 limits->zoned = blk_queue_zoned_model(q);
480 * Decrement a device's use count and remove it if necessary.
482 void dm_put_device(struct dm_target *ti, struct dm_dev *d)
485 struct list_head *devices = &ti->table->devices;
486 struct dm_dev_internal *dd;
488 list_for_each_entry(dd, devices, list) {
489 if (dd->dm_dev == d) {
495 DMWARN("%s: device %s not in table devices list",
496 dm_device_name(ti->table->md), d->name);
499 if (refcount_dec_and_test(&dd->count)) {
500 dm_put_table_device(ti->table->md, d);
505 EXPORT_SYMBOL(dm_put_device);
508 * Checks to see if the target joins onto the end of the table.
510 static int adjoin(struct dm_table *table, struct dm_target *ti)
512 struct dm_target *prev;
514 if (!table->num_targets)
517 prev = &table->targets[table->num_targets - 1];
518 return (ti->begin == (prev->begin + prev->len));
522 * Used to dynamically allocate the arg array.
524 * We do first allocation with GFP_NOIO because dm-mpath and dm-thin must
525 * process messages even if some device is suspended. These messages have a
526 * small fixed number of arguments.
528 * On the other hand, dm-switch needs to process bulk data using messages and
529 * excessive use of GFP_NOIO could cause trouble.
531 static char **realloc_argv(unsigned *size, char **old_argv)
538 new_size = *size * 2;
544 argv = kmalloc_array(new_size, sizeof(*argv), gfp);
545 if (argv && old_argv) {
546 memcpy(argv, old_argv, *size * sizeof(*argv));
555 * Destructively splits up the argument list to pass to ctr.
557 int dm_split_args(int *argc, char ***argvp, char *input)
559 char *start, *end = input, *out, **argv = NULL;
560 unsigned array_size = 0;
569 argv = realloc_argv(&array_size, argv);
574 /* Skip whitespace */
575 start = skip_spaces(end);
578 break; /* success, we hit the end */
580 /* 'out' is used to remove any back-quotes */
583 /* Everything apart from '\0' can be quoted */
584 if (*end == '\\' && *(end + 1)) {
591 break; /* end of token */
596 /* have we already filled the array ? */
597 if ((*argc + 1) > array_size) {
598 argv = realloc_argv(&array_size, argv);
603 /* we know this is whitespace */
607 /* terminate the string and put it in the array */
618 * Impose necessary and sufficient conditions on a devices's table such
619 * that any incoming bio which respects its logical_block_size can be
620 * processed successfully. If it falls across the boundary between
621 * two or more targets, the size of each piece it gets split into must
622 * be compatible with the logical_block_size of the target processing it.
624 static int validate_hardware_logical_block_alignment(struct dm_table *table,
625 struct queue_limits *limits)
628 * This function uses arithmetic modulo the logical_block_size
629 * (in units of 512-byte sectors).
631 unsigned short device_logical_block_size_sects =
632 limits->logical_block_size >> SECTOR_SHIFT;
635 * Offset of the start of the next table entry, mod logical_block_size.
637 unsigned short next_target_start = 0;
640 * Given an aligned bio that extends beyond the end of a
641 * target, how many sectors must the next target handle?
643 unsigned short remaining = 0;
645 struct dm_target *uninitialized_var(ti);
646 struct queue_limits ti_limits;
650 * Check each entry in the table in turn.
652 for (i = 0; i < dm_table_get_num_targets(table); i++) {
653 ti = dm_table_get_target(table, i);
655 blk_set_stacking_limits(&ti_limits);
657 /* combine all target devices' limits */
658 if (ti->type->iterate_devices)
659 ti->type->iterate_devices(ti, dm_set_device_limits,
663 * If the remaining sectors fall entirely within this
664 * table entry are they compatible with its logical_block_size?
666 if (remaining < ti->len &&
667 remaining & ((ti_limits.logical_block_size >>
672 (unsigned short) ((next_target_start + ti->len) &
673 (device_logical_block_size_sects - 1));
674 remaining = next_target_start ?
675 device_logical_block_size_sects - next_target_start : 0;
679 DMWARN("%s: table line %u (start sect %llu len %llu) "
680 "not aligned to h/w logical block size %u",
681 dm_device_name(table->md), i,
682 (unsigned long long) ti->begin,
683 (unsigned long long) ti->len,
684 limits->logical_block_size);
691 int dm_table_add_target(struct dm_table *t, const char *type,
692 sector_t start, sector_t len, char *params)
694 int r = -EINVAL, argc;
696 struct dm_target *tgt;
699 DMERR("%s: target type %s must appear alone in table",
700 dm_device_name(t->md), t->targets->type->name);
704 BUG_ON(t->num_targets >= t->num_allocated);
706 tgt = t->targets + t->num_targets;
707 memset(tgt, 0, sizeof(*tgt));
710 DMERR("%s: zero-length target", dm_device_name(t->md));
714 tgt->type = dm_get_target_type(type);
716 DMERR("%s: %s: unknown target type", dm_device_name(t->md), type);
720 if (dm_target_needs_singleton(tgt->type)) {
721 if (t->num_targets) {
722 tgt->error = "singleton target type must appear alone in table";
728 if (dm_target_always_writeable(tgt->type) && !(t->mode & FMODE_WRITE)) {
729 tgt->error = "target type may not be included in a read-only table";
733 if (t->immutable_target_type) {
734 if (t->immutable_target_type != tgt->type) {
735 tgt->error = "immutable target type cannot be mixed with other target types";
738 } else if (dm_target_is_immutable(tgt->type)) {
739 if (t->num_targets) {
740 tgt->error = "immutable target type cannot be mixed with other target types";
743 t->immutable_target_type = tgt->type;
746 if (dm_target_has_integrity(tgt->type))
747 t->integrity_added = 1;
752 tgt->error = "Unknown error";
755 * Does this target adjoin the previous one ?
757 if (!adjoin(t, tgt)) {
758 tgt->error = "Gap in table";
762 r = dm_split_args(&argc, &argv, params);
764 tgt->error = "couldn't split parameters (insufficient memory)";
768 r = tgt->type->ctr(tgt, argc, argv);
773 t->highs[t->num_targets++] = tgt->begin + tgt->len - 1;
775 if (!tgt->num_discard_bios && tgt->discards_supported)
776 DMWARN("%s: %s: ignoring discards_supported because num_discard_bios is zero.",
777 dm_device_name(t->md), type);
782 DMERR("%s: %s: %s", dm_device_name(t->md), type, tgt->error);
783 dm_put_target_type(tgt->type);
788 * Target argument parsing helpers.
790 static int validate_next_arg(const struct dm_arg *arg,
791 struct dm_arg_set *arg_set,
792 unsigned *value, char **error, unsigned grouped)
794 const char *arg_str = dm_shift_arg(arg_set);
798 (sscanf(arg_str, "%u%c", value, &dummy) != 1) ||
799 (*value < arg->min) ||
800 (*value > arg->max) ||
801 (grouped && arg_set->argc < *value)) {
809 int dm_read_arg(const struct dm_arg *arg, struct dm_arg_set *arg_set,
810 unsigned *value, char **error)
812 return validate_next_arg(arg, arg_set, value, error, 0);
814 EXPORT_SYMBOL(dm_read_arg);
816 int dm_read_arg_group(const struct dm_arg *arg, struct dm_arg_set *arg_set,
817 unsigned *value, char **error)
819 return validate_next_arg(arg, arg_set, value, error, 1);
821 EXPORT_SYMBOL(dm_read_arg_group);
823 const char *dm_shift_arg(struct dm_arg_set *as)
836 EXPORT_SYMBOL(dm_shift_arg);
838 void dm_consume_args(struct dm_arg_set *as, unsigned num_args)
840 BUG_ON(as->argc < num_args);
841 as->argc -= num_args;
842 as->argv += num_args;
844 EXPORT_SYMBOL(dm_consume_args);
846 static bool __table_type_bio_based(enum dm_queue_mode table_type)
848 return (table_type == DM_TYPE_BIO_BASED ||
849 table_type == DM_TYPE_DAX_BIO_BASED ||
850 table_type == DM_TYPE_NVME_BIO_BASED);
853 static bool __table_type_request_based(enum dm_queue_mode table_type)
855 return table_type == DM_TYPE_REQUEST_BASED;
858 void dm_table_set_type(struct dm_table *t, enum dm_queue_mode type)
862 EXPORT_SYMBOL_GPL(dm_table_set_type);
864 /* validate the dax capability of the target device span */
865 int device_supports_dax(struct dm_target *ti, struct dm_dev *dev,
866 sector_t start, sector_t len, void *data)
868 int blocksize = *(int *) data;
870 return generic_fsdax_supported(dev->dax_dev, dev->bdev, blocksize,
874 /* Check devices support synchronous DAX */
875 static int device_dax_synchronous(struct dm_target *ti, struct dm_dev *dev,
876 sector_t start, sector_t len, void *data)
878 return dev->dax_dev && dax_synchronous(dev->dax_dev);
881 bool dm_table_supports_dax(struct dm_table *t,
882 iterate_devices_callout_fn iterate_fn, int *blocksize)
884 struct dm_target *ti;
887 /* Ensure that all targets support DAX. */
888 for (i = 0; i < dm_table_get_num_targets(t); i++) {
889 ti = dm_table_get_target(t, i);
891 if (!ti->type->direct_access)
894 if (!ti->type->iterate_devices ||
895 !ti->type->iterate_devices(ti, iterate_fn, blocksize))
902 static bool dm_table_does_not_support_partial_completion(struct dm_table *t);
904 static int device_is_rq_stackable(struct dm_target *ti, struct dm_dev *dev,
905 sector_t start, sector_t len, void *data)
907 struct block_device *bdev = dev->bdev;
908 struct request_queue *q = bdev_get_queue(bdev);
910 /* request-based cannot stack on partitions! */
911 if (bdev != bdev->bd_contains)
914 return queue_is_mq(q);
917 static int dm_table_determine_type(struct dm_table *t)
920 unsigned bio_based = 0, request_based = 0, hybrid = 0;
921 struct dm_target *tgt;
922 struct list_head *devices = dm_table_get_devices(t);
923 enum dm_queue_mode live_md_type = dm_get_md_type(t->md);
924 int page_size = PAGE_SIZE;
926 if (t->type != DM_TYPE_NONE) {
927 /* target already set the table's type */
928 if (t->type == DM_TYPE_BIO_BASED) {
929 /* possibly upgrade to a variant of bio-based */
930 goto verify_bio_based;
932 BUG_ON(t->type == DM_TYPE_DAX_BIO_BASED);
933 BUG_ON(t->type == DM_TYPE_NVME_BIO_BASED);
934 goto verify_rq_based;
937 for (i = 0; i < t->num_targets; i++) {
938 tgt = t->targets + i;
939 if (dm_target_hybrid(tgt))
941 else if (dm_target_request_based(tgt))
946 if (bio_based && request_based) {
947 DMERR("Inconsistent table: different target types"
948 " can't be mixed up");
953 if (hybrid && !bio_based && !request_based) {
955 * The targets can work either way.
956 * Determine the type from the live device.
957 * Default to bio-based if device is new.
959 if (__table_type_request_based(live_md_type))
967 /* We must use this table as bio-based */
968 t->type = DM_TYPE_BIO_BASED;
969 if (dm_table_supports_dax(t, device_supports_dax, &page_size) ||
970 (list_empty(devices) && live_md_type == DM_TYPE_DAX_BIO_BASED)) {
971 t->type = DM_TYPE_DAX_BIO_BASED;
973 /* Check if upgrading to NVMe bio-based is valid or required */
974 tgt = dm_table_get_immutable_target(t);
975 if (tgt && !tgt->max_io_len && dm_table_does_not_support_partial_completion(t)) {
976 t->type = DM_TYPE_NVME_BIO_BASED;
977 goto verify_rq_based; /* must be stacked directly on NVMe (blk-mq) */
978 } else if (list_empty(devices) && live_md_type == DM_TYPE_NVME_BIO_BASED) {
979 t->type = DM_TYPE_NVME_BIO_BASED;
985 BUG_ON(!request_based); /* No targets in this table */
987 t->type = DM_TYPE_REQUEST_BASED;
991 * Request-based dm supports only tables that have a single target now.
992 * To support multiple targets, request splitting support is needed,
993 * and that needs lots of changes in the block-layer.
994 * (e.g. request completion process for partial completion.)
996 if (t->num_targets > 1) {
997 DMERR("%s DM doesn't support multiple targets",
998 t->type == DM_TYPE_NVME_BIO_BASED ? "nvme bio-based" : "request-based");
1002 if (list_empty(devices)) {
1004 struct dm_table *live_table = dm_get_live_table(t->md, &srcu_idx);
1006 /* inherit live table's type */
1008 t->type = live_table->type;
1009 dm_put_live_table(t->md, srcu_idx);
1013 tgt = dm_table_get_immutable_target(t);
1015 DMERR("table load rejected: immutable target is required");
1017 } else if (tgt->max_io_len) {
1018 DMERR("table load rejected: immutable target that splits IO is not supported");
1022 /* Non-request-stackable devices can't be used for request-based dm */
1023 if (!tgt->type->iterate_devices ||
1024 !tgt->type->iterate_devices(tgt, device_is_rq_stackable, NULL)) {
1025 DMERR("table load rejected: including non-request-stackable devices");
1032 enum dm_queue_mode dm_table_get_type(struct dm_table *t)
1037 struct target_type *dm_table_get_immutable_target_type(struct dm_table *t)
1039 return t->immutable_target_type;
1042 struct dm_target *dm_table_get_immutable_target(struct dm_table *t)
1044 /* Immutable target is implicitly a singleton */
1045 if (t->num_targets > 1 ||
1046 !dm_target_is_immutable(t->targets[0].type))
1052 struct dm_target *dm_table_get_wildcard_target(struct dm_table *t)
1054 struct dm_target *ti;
1057 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1058 ti = dm_table_get_target(t, i);
1059 if (dm_target_is_wildcard(ti->type))
1066 bool dm_table_bio_based(struct dm_table *t)
1068 return __table_type_bio_based(dm_table_get_type(t));
1071 bool dm_table_request_based(struct dm_table *t)
1073 return __table_type_request_based(dm_table_get_type(t));
1076 static int dm_table_alloc_md_mempools(struct dm_table *t, struct mapped_device *md)
1078 enum dm_queue_mode type = dm_table_get_type(t);
1079 unsigned per_io_data_size = 0;
1080 unsigned min_pool_size = 0;
1081 struct dm_target *ti;
1084 if (unlikely(type == DM_TYPE_NONE)) {
1085 DMWARN("no table type is set, can't allocate mempools");
1089 if (__table_type_bio_based(type))
1090 for (i = 0; i < t->num_targets; i++) {
1091 ti = t->targets + i;
1092 per_io_data_size = max(per_io_data_size, ti->per_io_data_size);
1093 min_pool_size = max(min_pool_size, ti->num_flush_bios);
1096 t->mempools = dm_alloc_md_mempools(md, type, t->integrity_supported,
1097 per_io_data_size, min_pool_size);
1104 void dm_table_free_md_mempools(struct dm_table *t)
1106 dm_free_md_mempools(t->mempools);
1110 struct dm_md_mempools *dm_table_get_md_mempools(struct dm_table *t)
1115 static int setup_indexes(struct dm_table *t)
1118 unsigned int total = 0;
1121 /* allocate the space for *all* the indexes */
1122 for (i = t->depth - 2; i >= 0; i--) {
1123 t->counts[i] = dm_div_up(t->counts[i + 1], CHILDREN_PER_NODE);
1124 total += t->counts[i];
1127 indexes = (sector_t *) dm_vcalloc(total, (unsigned long) NODE_SIZE);
1131 /* set up internal nodes, bottom-up */
1132 for (i = t->depth - 2; i >= 0; i--) {
1133 t->index[i] = indexes;
1134 indexes += (KEYS_PER_NODE * t->counts[i]);
1135 setup_btree_index(i, t);
1142 * Builds the btree to index the map.
1144 static int dm_table_build_index(struct dm_table *t)
1147 unsigned int leaf_nodes;
1149 /* how many indexes will the btree have ? */
1150 leaf_nodes = dm_div_up(t->num_targets, KEYS_PER_NODE);
1151 t->depth = 1 + int_log(leaf_nodes, CHILDREN_PER_NODE);
1153 /* leaf layer has already been set up */
1154 t->counts[t->depth - 1] = leaf_nodes;
1155 t->index[t->depth - 1] = t->highs;
1158 r = setup_indexes(t);
1163 static bool integrity_profile_exists(struct gendisk *disk)
1165 return !!blk_get_integrity(disk);
1169 * Get a disk whose integrity profile reflects the table's profile.
1170 * Returns NULL if integrity support was inconsistent or unavailable.
1172 static struct gendisk * dm_table_get_integrity_disk(struct dm_table *t)
1174 struct list_head *devices = dm_table_get_devices(t);
1175 struct dm_dev_internal *dd = NULL;
1176 struct gendisk *prev_disk = NULL, *template_disk = NULL;
1179 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1180 struct dm_target *ti = dm_table_get_target(t, i);
1181 if (!dm_target_passes_integrity(ti->type))
1185 list_for_each_entry(dd, devices, list) {
1186 template_disk = dd->dm_dev->bdev->bd_disk;
1187 if (!integrity_profile_exists(template_disk))
1189 else if (prev_disk &&
1190 blk_integrity_compare(prev_disk, template_disk) < 0)
1192 prev_disk = template_disk;
1195 return template_disk;
1199 DMWARN("%s: integrity not set: %s and %s profile mismatch",
1200 dm_device_name(t->md),
1201 prev_disk->disk_name,
1202 template_disk->disk_name);
1207 * Register the mapped device for blk_integrity support if the
1208 * underlying devices have an integrity profile. But all devices may
1209 * not have matching profiles (checking all devices isn't reliable
1210 * during table load because this table may use other DM device(s) which
1211 * must be resumed before they will have an initialized integity
1212 * profile). Consequently, stacked DM devices force a 2 stage integrity
1213 * profile validation: First pass during table load, final pass during
1216 static int dm_table_register_integrity(struct dm_table *t)
1218 struct mapped_device *md = t->md;
1219 struct gendisk *template_disk = NULL;
1221 /* If target handles integrity itself do not register it here. */
1222 if (t->integrity_added)
1225 template_disk = dm_table_get_integrity_disk(t);
1229 if (!integrity_profile_exists(dm_disk(md))) {
1230 t->integrity_supported = true;
1232 * Register integrity profile during table load; we can do
1233 * this because the final profile must match during resume.
1235 blk_integrity_register(dm_disk(md),
1236 blk_get_integrity(template_disk));
1241 * If DM device already has an initialized integrity
1242 * profile the new profile should not conflict.
1244 if (blk_integrity_compare(dm_disk(md), template_disk) < 0) {
1245 DMWARN("%s: conflict with existing integrity profile: "
1246 "%s profile mismatch",
1247 dm_device_name(t->md),
1248 template_disk->disk_name);
1252 /* Preserve existing integrity profile */
1253 t->integrity_supported = true;
1258 * Prepares the table for use by building the indices,
1259 * setting the type, and allocating mempools.
1261 int dm_table_complete(struct dm_table *t)
1265 r = dm_table_determine_type(t);
1267 DMERR("unable to determine table type");
1271 r = dm_table_build_index(t);
1273 DMERR("unable to build btrees");
1277 r = dm_table_register_integrity(t);
1279 DMERR("could not register integrity profile.");
1283 r = dm_table_alloc_md_mempools(t, t->md);
1285 DMERR("unable to allocate mempools");
1290 static DEFINE_MUTEX(_event_lock);
1291 void dm_table_event_callback(struct dm_table *t,
1292 void (*fn)(void *), void *context)
1294 mutex_lock(&_event_lock);
1296 t->event_context = context;
1297 mutex_unlock(&_event_lock);
1300 void dm_table_event(struct dm_table *t)
1303 * You can no longer call dm_table_event() from interrupt
1304 * context, use a bottom half instead.
1306 BUG_ON(in_interrupt());
1308 mutex_lock(&_event_lock);
1310 t->event_fn(t->event_context);
1311 mutex_unlock(&_event_lock);
1313 EXPORT_SYMBOL(dm_table_event);
1315 inline sector_t dm_table_get_size(struct dm_table *t)
1317 return t->num_targets ? (t->highs[t->num_targets - 1] + 1) : 0;
1319 EXPORT_SYMBOL(dm_table_get_size);
1321 struct dm_target *dm_table_get_target(struct dm_table *t, unsigned int index)
1323 if (index >= t->num_targets)
1326 return t->targets + index;
1330 * Search the btree for the correct target.
1332 * Caller should check returned pointer for NULL
1333 * to trap I/O beyond end of device.
1335 struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector)
1337 unsigned int l, n = 0, k = 0;
1340 if (unlikely(sector >= dm_table_get_size(t)))
1343 for (l = 0; l < t->depth; l++) {
1344 n = get_child(n, k);
1345 node = get_node(t, l, n);
1347 for (k = 0; k < KEYS_PER_NODE; k++)
1348 if (node[k] >= sector)
1352 return &t->targets[(KEYS_PER_NODE * n) + k];
1355 static int count_device(struct dm_target *ti, struct dm_dev *dev,
1356 sector_t start, sector_t len, void *data)
1358 unsigned *num_devices = data;
1366 * Check whether a table has no data devices attached using each
1367 * target's iterate_devices method.
1368 * Returns false if the result is unknown because a target doesn't
1369 * support iterate_devices.
1371 bool dm_table_has_no_data_devices(struct dm_table *table)
1373 struct dm_target *ti;
1374 unsigned i, num_devices;
1376 for (i = 0; i < dm_table_get_num_targets(table); i++) {
1377 ti = dm_table_get_target(table, i);
1379 if (!ti->type->iterate_devices)
1383 ti->type->iterate_devices(ti, count_device, &num_devices);
1391 static int device_is_zoned_model(struct dm_target *ti, struct dm_dev *dev,
1392 sector_t start, sector_t len, void *data)
1394 struct request_queue *q = bdev_get_queue(dev->bdev);
1395 enum blk_zoned_model *zoned_model = data;
1397 return q && blk_queue_zoned_model(q) == *zoned_model;
1400 static bool dm_table_supports_zoned_model(struct dm_table *t,
1401 enum blk_zoned_model zoned_model)
1403 struct dm_target *ti;
1406 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1407 ti = dm_table_get_target(t, i);
1409 if (zoned_model == BLK_ZONED_HM &&
1410 !dm_target_supports_zoned_hm(ti->type))
1413 if (!ti->type->iterate_devices ||
1414 !ti->type->iterate_devices(ti, device_is_zoned_model, &zoned_model))
1421 static int device_matches_zone_sectors(struct dm_target *ti, struct dm_dev *dev,
1422 sector_t start, sector_t len, void *data)
1424 struct request_queue *q = bdev_get_queue(dev->bdev);
1425 unsigned int *zone_sectors = data;
1427 return q && blk_queue_zone_sectors(q) == *zone_sectors;
1430 static bool dm_table_matches_zone_sectors(struct dm_table *t,
1431 unsigned int zone_sectors)
1433 struct dm_target *ti;
1436 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1437 ti = dm_table_get_target(t, i);
1439 if (!ti->type->iterate_devices ||
1440 !ti->type->iterate_devices(ti, device_matches_zone_sectors, &zone_sectors))
1447 static int validate_hardware_zoned_model(struct dm_table *table,
1448 enum blk_zoned_model zoned_model,
1449 unsigned int zone_sectors)
1451 if (zoned_model == BLK_ZONED_NONE)
1454 if (!dm_table_supports_zoned_model(table, zoned_model)) {
1455 DMERR("%s: zoned model is not consistent across all devices",
1456 dm_device_name(table->md));
1460 /* Check zone size validity and compatibility */
1461 if (!zone_sectors || !is_power_of_2(zone_sectors))
1464 if (!dm_table_matches_zone_sectors(table, zone_sectors)) {
1465 DMERR("%s: zone sectors is not consistent across all devices",
1466 dm_device_name(table->md));
1474 * Establish the new table's queue_limits and validate them.
1476 int dm_calculate_queue_limits(struct dm_table *table,
1477 struct queue_limits *limits)
1479 struct dm_target *ti;
1480 struct queue_limits ti_limits;
1482 enum blk_zoned_model zoned_model = BLK_ZONED_NONE;
1483 unsigned int zone_sectors = 0;
1485 blk_set_stacking_limits(limits);
1487 for (i = 0; i < dm_table_get_num_targets(table); i++) {
1488 blk_set_stacking_limits(&ti_limits);
1490 ti = dm_table_get_target(table, i);
1492 if (!ti->type->iterate_devices)
1493 goto combine_limits;
1496 * Combine queue limits of all the devices this target uses.
1498 ti->type->iterate_devices(ti, dm_set_device_limits,
1501 if (zoned_model == BLK_ZONED_NONE && ti_limits.zoned != BLK_ZONED_NONE) {
1503 * After stacking all limits, validate all devices
1504 * in table support this zoned model and zone sectors.
1506 zoned_model = ti_limits.zoned;
1507 zone_sectors = ti_limits.chunk_sectors;
1510 /* Set I/O hints portion of queue limits */
1511 if (ti->type->io_hints)
1512 ti->type->io_hints(ti, &ti_limits);
1515 * Check each device area is consistent with the target's
1516 * overall queue limits.
1518 if (ti->type->iterate_devices(ti, device_area_is_invalid,
1524 * Merge this target's queue limits into the overall limits
1527 if (blk_stack_limits(limits, &ti_limits, 0) < 0)
1528 DMWARN("%s: adding target device "
1529 "(start sect %llu len %llu) "
1530 "caused an alignment inconsistency",
1531 dm_device_name(table->md),
1532 (unsigned long long) ti->begin,
1533 (unsigned long long) ti->len);
1536 * FIXME: this should likely be moved to blk_stack_limits(), would
1537 * also eliminate limits->zoned stacking hack in dm_set_device_limits()
1539 if (limits->zoned == BLK_ZONED_NONE && ti_limits.zoned != BLK_ZONED_NONE) {
1541 * By default, the stacked limits zoned model is set to
1542 * BLK_ZONED_NONE in blk_set_stacking_limits(). Update
1543 * this model using the first target model reported
1544 * that is not BLK_ZONED_NONE. This will be either the
1545 * first target device zoned model or the model reported
1546 * by the target .io_hints.
1548 limits->zoned = ti_limits.zoned;
1553 * Verify that the zoned model and zone sectors, as determined before
1554 * any .io_hints override, are the same across all devices in the table.
1555 * - this is especially relevant if .io_hints is emulating a disk-managed
1556 * zoned model (aka BLK_ZONED_NONE) on host-managed zoned block devices.
1559 if (limits->zoned != BLK_ZONED_NONE) {
1561 * ...IF the above limits stacking determined a zoned model
1562 * validate that all of the table's devices conform to it.
1564 zoned_model = limits->zoned;
1565 zone_sectors = limits->chunk_sectors;
1567 if (validate_hardware_zoned_model(table, zoned_model, zone_sectors))
1570 return validate_hardware_logical_block_alignment(table, limits);
1574 * Verify that all devices have an integrity profile that matches the
1575 * DM device's registered integrity profile. If the profiles don't
1576 * match then unregister the DM device's integrity profile.
1578 static void dm_table_verify_integrity(struct dm_table *t)
1580 struct gendisk *template_disk = NULL;
1582 if (t->integrity_added)
1585 if (t->integrity_supported) {
1587 * Verify that the original integrity profile
1588 * matches all the devices in this table.
1590 template_disk = dm_table_get_integrity_disk(t);
1591 if (template_disk &&
1592 blk_integrity_compare(dm_disk(t->md), template_disk) >= 0)
1596 if (integrity_profile_exists(dm_disk(t->md))) {
1597 DMWARN("%s: unable to establish an integrity profile",
1598 dm_device_name(t->md));
1599 blk_integrity_unregister(dm_disk(t->md));
1603 static int device_flush_capable(struct dm_target *ti, struct dm_dev *dev,
1604 sector_t start, sector_t len, void *data)
1606 unsigned long flush = (unsigned long) data;
1607 struct request_queue *q = bdev_get_queue(dev->bdev);
1609 return q && (q->queue_flags & flush);
1612 static bool dm_table_supports_flush(struct dm_table *t, unsigned long flush)
1614 struct dm_target *ti;
1618 * Require at least one underlying device to support flushes.
1619 * t->devices includes internal dm devices such as mirror logs
1620 * so we need to use iterate_devices here, which targets
1621 * supporting flushes must provide.
1623 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1624 ti = dm_table_get_target(t, i);
1626 if (!ti->num_flush_bios)
1629 if (ti->flush_supported)
1632 if (ti->type->iterate_devices &&
1633 ti->type->iterate_devices(ti, device_flush_capable, (void *) flush))
1640 static int device_dax_write_cache_enabled(struct dm_target *ti,
1641 struct dm_dev *dev, sector_t start,
1642 sector_t len, void *data)
1644 struct dax_device *dax_dev = dev->dax_dev;
1649 if (dax_write_cache_enabled(dax_dev))
1654 static int dm_table_supports_dax_write_cache(struct dm_table *t)
1656 struct dm_target *ti;
1659 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1660 ti = dm_table_get_target(t, i);
1662 if (ti->type->iterate_devices &&
1663 ti->type->iterate_devices(ti,
1664 device_dax_write_cache_enabled, NULL))
1671 static int device_is_nonrot(struct dm_target *ti, struct dm_dev *dev,
1672 sector_t start, sector_t len, void *data)
1674 struct request_queue *q = bdev_get_queue(dev->bdev);
1676 return q && blk_queue_nonrot(q);
1679 static int device_is_not_random(struct dm_target *ti, struct dm_dev *dev,
1680 sector_t start, sector_t len, void *data)
1682 struct request_queue *q = bdev_get_queue(dev->bdev);
1684 return q && !blk_queue_add_random(q);
1687 static bool dm_table_all_devices_attribute(struct dm_table *t,
1688 iterate_devices_callout_fn func)
1690 struct dm_target *ti;
1693 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1694 ti = dm_table_get_target(t, i);
1696 if (!ti->type->iterate_devices ||
1697 !ti->type->iterate_devices(ti, func, NULL))
1704 static int device_no_partial_completion(struct dm_target *ti, struct dm_dev *dev,
1705 sector_t start, sector_t len, void *data)
1707 char b[BDEVNAME_SIZE];
1709 /* For now, NVMe devices are the only devices of this class */
1710 return (strncmp(bdevname(dev->bdev, b), "nvme", 4) == 0);
1713 static bool dm_table_does_not_support_partial_completion(struct dm_table *t)
1715 return dm_table_all_devices_attribute(t, device_no_partial_completion);
1718 static int device_not_write_same_capable(struct dm_target *ti, struct dm_dev *dev,
1719 sector_t start, sector_t len, void *data)
1721 struct request_queue *q = bdev_get_queue(dev->bdev);
1723 return q && !q->limits.max_write_same_sectors;
1726 static bool dm_table_supports_write_same(struct dm_table *t)
1728 struct dm_target *ti;
1731 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1732 ti = dm_table_get_target(t, i);
1734 if (!ti->num_write_same_bios)
1737 if (!ti->type->iterate_devices ||
1738 ti->type->iterate_devices(ti, device_not_write_same_capable, NULL))
1745 static int device_not_write_zeroes_capable(struct dm_target *ti, struct dm_dev *dev,
1746 sector_t start, sector_t len, void *data)
1748 struct request_queue *q = bdev_get_queue(dev->bdev);
1750 return q && !q->limits.max_write_zeroes_sectors;
1753 static bool dm_table_supports_write_zeroes(struct dm_table *t)
1755 struct dm_target *ti;
1758 while (i < dm_table_get_num_targets(t)) {
1759 ti = dm_table_get_target(t, i++);
1761 if (!ti->num_write_zeroes_bios)
1764 if (!ti->type->iterate_devices ||
1765 ti->type->iterate_devices(ti, device_not_write_zeroes_capable, NULL))
1772 static int device_not_discard_capable(struct dm_target *ti, struct dm_dev *dev,
1773 sector_t start, sector_t len, void *data)
1775 struct request_queue *q = bdev_get_queue(dev->bdev);
1777 return q && !blk_queue_discard(q);
1780 static bool dm_table_supports_discards(struct dm_table *t)
1782 struct dm_target *ti;
1785 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1786 ti = dm_table_get_target(t, i);
1788 if (!ti->num_discard_bios)
1792 * Either the target provides discard support (as implied by setting
1793 * 'discards_supported') or it relies on _all_ data devices having
1796 if (!ti->discards_supported &&
1797 (!ti->type->iterate_devices ||
1798 ti->type->iterate_devices(ti, device_not_discard_capable, NULL)))
1805 static int device_not_secure_erase_capable(struct dm_target *ti,
1806 struct dm_dev *dev, sector_t start,
1807 sector_t len, void *data)
1809 struct request_queue *q = bdev_get_queue(dev->bdev);
1811 return q && !blk_queue_secure_erase(q);
1814 static bool dm_table_supports_secure_erase(struct dm_table *t)
1816 struct dm_target *ti;
1819 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1820 ti = dm_table_get_target(t, i);
1822 if (!ti->num_secure_erase_bios)
1825 if (!ti->type->iterate_devices ||
1826 ti->type->iterate_devices(ti, device_not_secure_erase_capable, NULL))
1833 static int device_requires_stable_pages(struct dm_target *ti,
1834 struct dm_dev *dev, sector_t start,
1835 sector_t len, void *data)
1837 struct request_queue *q = bdev_get_queue(dev->bdev);
1839 return q && bdi_cap_stable_pages_required(q->backing_dev_info);
1843 * If any underlying device requires stable pages, a table must require
1844 * them as well. Only targets that support iterate_devices are considered:
1845 * don't want error, zero, etc to require stable pages.
1847 static bool dm_table_requires_stable_pages(struct dm_table *t)
1849 struct dm_target *ti;
1852 for (i = 0; i < dm_table_get_num_targets(t); i++) {
1853 ti = dm_table_get_target(t, i);
1855 if (ti->type->iterate_devices &&
1856 ti->type->iterate_devices(ti, device_requires_stable_pages, NULL))
1863 void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q,
1864 struct queue_limits *limits)
1866 bool wc = false, fua = false;
1867 int page_size = PAGE_SIZE;
1870 * Copy table's limits to the DM device's request_queue
1872 q->limits = *limits;
1874 if (!dm_table_supports_discards(t)) {
1875 blk_queue_flag_clear(QUEUE_FLAG_DISCARD, q);
1876 /* Must also clear discard limits... */
1877 q->limits.max_discard_sectors = 0;
1878 q->limits.max_hw_discard_sectors = 0;
1879 q->limits.discard_granularity = 0;
1880 q->limits.discard_alignment = 0;
1881 q->limits.discard_misaligned = 0;
1883 blk_queue_flag_set(QUEUE_FLAG_DISCARD, q);
1885 if (dm_table_supports_secure_erase(t))
1886 blk_queue_flag_set(QUEUE_FLAG_SECERASE, q);
1888 if (dm_table_supports_flush(t, (1UL << QUEUE_FLAG_WC))) {
1890 if (dm_table_supports_flush(t, (1UL << QUEUE_FLAG_FUA)))
1893 blk_queue_write_cache(q, wc, fua);
1895 if (dm_table_supports_dax(t, device_supports_dax, &page_size)) {
1896 blk_queue_flag_set(QUEUE_FLAG_DAX, q);
1897 if (dm_table_supports_dax(t, device_dax_synchronous, NULL))
1898 set_dax_synchronous(t->md->dax_dev);
1901 blk_queue_flag_clear(QUEUE_FLAG_DAX, q);
1903 if (dm_table_supports_dax_write_cache(t))
1904 dax_write_cache(t->md->dax_dev, true);
1906 /* Ensure that all underlying devices are non-rotational. */
1907 if (dm_table_all_devices_attribute(t, device_is_nonrot))
1908 blk_queue_flag_set(QUEUE_FLAG_NONROT, q);
1910 blk_queue_flag_clear(QUEUE_FLAG_NONROT, q);
1912 if (!dm_table_supports_write_same(t))
1913 q->limits.max_write_same_sectors = 0;
1914 if (!dm_table_supports_write_zeroes(t))
1915 q->limits.max_write_zeroes_sectors = 0;
1917 dm_table_verify_integrity(t);
1920 * Some devices don't use blk_integrity but still want stable pages
1921 * because they do their own checksumming.
1923 if (dm_table_requires_stable_pages(t))
1924 q->backing_dev_info->capabilities |= BDI_CAP_STABLE_WRITES;
1926 q->backing_dev_info->capabilities &= ~BDI_CAP_STABLE_WRITES;
1929 * Determine whether or not this queue's I/O timings contribute
1930 * to the entropy pool, Only request-based targets use this.
1931 * Clear QUEUE_FLAG_ADD_RANDOM if any underlying device does not
1934 if (blk_queue_add_random(q) && dm_table_all_devices_attribute(t, device_is_not_random))
1935 blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, q);
1938 * For a zoned target, the number of zones should be updated for the
1939 * correct value to be exposed in sysfs queue/nr_zones. For a BIO based
1940 * target, this is all that is needed.
1942 #ifdef CONFIG_BLK_DEV_ZONED
1943 if (blk_queue_is_zoned(q)) {
1944 WARN_ON_ONCE(queue_is_mq(q));
1945 q->nr_zones = blkdev_nr_zones(t->md->disk);
1949 /* Allow reads to exceed readahead limits */
1950 q->backing_dev_info->io_pages = limits->max_sectors >> (PAGE_SHIFT - 9);
1953 unsigned int dm_table_get_num_targets(struct dm_table *t)
1955 return t->num_targets;
1958 struct list_head *dm_table_get_devices(struct dm_table *t)
1963 fmode_t dm_table_get_mode(struct dm_table *t)
1967 EXPORT_SYMBOL(dm_table_get_mode);
1975 static void suspend_targets(struct dm_table *t, enum suspend_mode mode)
1977 int i = t->num_targets;
1978 struct dm_target *ti = t->targets;
1980 lockdep_assert_held(&t->md->suspend_lock);
1985 if (ti->type->presuspend)
1986 ti->type->presuspend(ti);
1988 case PRESUSPEND_UNDO:
1989 if (ti->type->presuspend_undo)
1990 ti->type->presuspend_undo(ti);
1993 if (ti->type->postsuspend)
1994 ti->type->postsuspend(ti);
2001 void dm_table_presuspend_targets(struct dm_table *t)
2006 suspend_targets(t, PRESUSPEND);
2009 void dm_table_presuspend_undo_targets(struct dm_table *t)
2014 suspend_targets(t, PRESUSPEND_UNDO);
2017 void dm_table_postsuspend_targets(struct dm_table *t)
2022 suspend_targets(t, POSTSUSPEND);
2025 int dm_table_resume_targets(struct dm_table *t)
2029 lockdep_assert_held(&t->md->suspend_lock);
2031 for (i = 0; i < t->num_targets; i++) {
2032 struct dm_target *ti = t->targets + i;
2034 if (!ti->type->preresume)
2037 r = ti->type->preresume(ti);
2039 DMERR("%s: %s: preresume failed, error = %d",
2040 dm_device_name(t->md), ti->type->name, r);
2045 for (i = 0; i < t->num_targets; i++) {
2046 struct dm_target *ti = t->targets + i;
2048 if (ti->type->resume)
2049 ti->type->resume(ti);
2055 void dm_table_add_target_callbacks(struct dm_table *t, struct dm_target_callbacks *cb)
2057 list_add(&cb->list, &t->target_callbacks);
2059 EXPORT_SYMBOL_GPL(dm_table_add_target_callbacks);
2061 int dm_table_any_congested(struct dm_table *t, int bdi_bits)
2063 struct dm_dev_internal *dd;
2064 struct list_head *devices = dm_table_get_devices(t);
2065 struct dm_target_callbacks *cb;
2068 list_for_each_entry(dd, devices, list) {
2069 struct request_queue *q = bdev_get_queue(dd->dm_dev->bdev);
2070 char b[BDEVNAME_SIZE];
2073 r |= bdi_congested(q->backing_dev_info, bdi_bits);
2075 DMWARN_LIMIT("%s: any_congested: nonexistent device %s",
2076 dm_device_name(t->md),
2077 bdevname(dd->dm_dev->bdev, b));
2080 list_for_each_entry(cb, &t->target_callbacks, list)
2081 if (cb->congested_fn)
2082 r |= cb->congested_fn(cb, bdi_bits);
2087 struct mapped_device *dm_table_get_md(struct dm_table *t)
2091 EXPORT_SYMBOL(dm_table_get_md);
2093 const char *dm_table_device_name(struct dm_table *t)
2095 return dm_device_name(t->md);
2097 EXPORT_SYMBOL_GPL(dm_table_device_name);
2099 void dm_table_run_md_queue_async(struct dm_table *t)
2101 struct mapped_device *md;
2102 struct request_queue *queue;
2104 if (!dm_table_request_based(t))
2107 md = dm_table_get_md(t);
2108 queue = dm_get_md_queue(md);
2110 blk_mq_run_hw_queues(queue, true);
2112 EXPORT_SYMBOL(dm_table_run_md_queue_async);