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 <asm/atomic.h>
22 #define DM_MSG_PREFIX "table"
25 #define NODE_SIZE L1_CACHE_BYTES
26 #define KEYS_PER_NODE (NODE_SIZE / sizeof(sector_t))
27 #define CHILDREN_PER_NODE (KEYS_PER_NODE + 1)
30 * The table has always exactly one reference from either mapped_device->map
31 * or hash_cell->new_map. This reference is not counted in table->holders.
32 * A pair of dm_create_table/dm_destroy_table functions is used for table
33 * creation/destruction.
35 * Temporary references from the other code increase table->holders. A pair
36 * of dm_table_get/dm_table_put functions is used to manipulate it.
38 * When the table is about to be destroyed, we wait for table->holders to
43 struct mapped_device *md;
49 unsigned int counts[MAX_DEPTH]; /* in nodes */
50 sector_t *index[MAX_DEPTH];
52 unsigned int num_targets;
53 unsigned int num_allocated;
55 struct dm_target *targets;
57 unsigned discards_supported:1;
60 * Indicates the rw permissions for the new logical
61 * device. This should be a combination of FMODE_READ
66 /* a list of devices used by this table */
67 struct list_head devices;
69 /* events get handed up using this callback */
70 void (*event_fn)(void *);
73 struct dm_md_mempools *mempools;
77 * Similar to ceiling(log_size(n))
79 static unsigned int int_log(unsigned int n, unsigned int base)
84 n = dm_div_up(n, base);
92 * Calculate the index of the child node of the n'th node k'th key.
94 static inline unsigned int get_child(unsigned int n, unsigned int k)
96 return (n * CHILDREN_PER_NODE) + k;
100 * Return the n'th node of level l from table t.
102 static inline sector_t *get_node(struct dm_table *t,
103 unsigned int l, unsigned int n)
105 return t->index[l] + (n * KEYS_PER_NODE);
109 * Return the highest key that you could lookup from the n'th
110 * node on level l of the btree.
112 static sector_t high(struct dm_table *t, unsigned int l, unsigned int n)
114 for (; l < t->depth - 1; l++)
115 n = get_child(n, CHILDREN_PER_NODE - 1);
117 if (n >= t->counts[l])
118 return (sector_t) - 1;
120 return get_node(t, l, n)[KEYS_PER_NODE - 1];
124 * Fills in a level of the btree based on the highs of the level
127 static int setup_btree_index(unsigned int l, struct dm_table *t)
132 for (n = 0U; n < t->counts[l]; n++) {
133 node = get_node(t, l, n);
135 for (k = 0U; k < KEYS_PER_NODE; k++)
136 node[k] = high(t, l + 1, get_child(n, k));
142 void *dm_vcalloc(unsigned long nmemb, unsigned long elem_size)
148 * Check that we're not going to overflow.
150 if (nmemb > (ULONG_MAX / elem_size))
153 size = nmemb * elem_size;
154 addr = vmalloc(size);
156 memset(addr, 0, size);
162 * highs, and targets are managed as dynamic arrays during a
165 static int alloc_targets(struct dm_table *t, unsigned int num)
168 struct dm_target *n_targets;
169 int n = t->num_targets;
172 * Allocate both the target array and offset array at once.
173 * Append an empty entry to catch sectors beyond the end of
176 n_highs = (sector_t *) dm_vcalloc(num + 1, sizeof(struct dm_target) +
181 n_targets = (struct dm_target *) (n_highs + num);
184 memcpy(n_highs, t->highs, sizeof(*n_highs) * n);
185 memcpy(n_targets, t->targets, sizeof(*n_targets) * n);
188 memset(n_highs + n, -1, sizeof(*n_highs) * (num - n));
191 t->num_allocated = num;
193 t->targets = n_targets;
198 int dm_table_create(struct dm_table **result, fmode_t mode,
199 unsigned num_targets, struct mapped_device *md)
201 struct dm_table *t = kzalloc(sizeof(*t), GFP_KERNEL);
206 INIT_LIST_HEAD(&t->devices);
207 atomic_set(&t->holders, 0);
208 t->discards_supported = 1;
211 num_targets = KEYS_PER_NODE;
213 num_targets = dm_round_up(num_targets, KEYS_PER_NODE);
215 if (alloc_targets(t, num_targets)) {
227 static void free_devices(struct list_head *devices)
229 struct list_head *tmp, *next;
231 list_for_each_safe(tmp, next, devices) {
232 struct dm_dev_internal *dd =
233 list_entry(tmp, struct dm_dev_internal, list);
234 DMWARN("dm_table_destroy: dm_put_device call missing for %s",
240 void dm_table_destroy(struct dm_table *t)
247 while (atomic_read(&t->holders))
251 /* free the indexes */
253 vfree(t->index[t->depth - 2]);
255 /* free the targets */
256 for (i = 0; i < t->num_targets; i++) {
257 struct dm_target *tgt = t->targets + i;
262 dm_put_target_type(tgt->type);
267 /* free the device list */
268 if (t->devices.next != &t->devices)
269 free_devices(&t->devices);
271 dm_free_md_mempools(t->mempools);
276 void dm_table_get(struct dm_table *t)
278 atomic_inc(&t->holders);
281 void dm_table_put(struct dm_table *t)
286 smp_mb__before_atomic_dec();
287 atomic_dec(&t->holders);
291 * Checks to see if we need to extend highs or targets.
293 static inline int check_space(struct dm_table *t)
295 if (t->num_targets >= t->num_allocated)
296 return alloc_targets(t, t->num_allocated * 2);
302 * See if we've already got a device in the list.
304 static struct dm_dev_internal *find_device(struct list_head *l, dev_t dev)
306 struct dm_dev_internal *dd;
308 list_for_each_entry (dd, l, list)
309 if (dd->dm_dev.bdev->bd_dev == dev)
316 * Open a device so we can use it as a map destination.
318 static int open_dev(struct dm_dev_internal *d, dev_t dev,
319 struct mapped_device *md)
321 static char *_claim_ptr = "I belong to device-mapper";
322 struct block_device *bdev;
326 BUG_ON(d->dm_dev.bdev);
328 bdev = open_by_devnum(dev, d->dm_dev.mode);
330 return PTR_ERR(bdev);
331 r = bd_claim_by_disk(bdev, _claim_ptr, dm_disk(md));
333 blkdev_put(bdev, d->dm_dev.mode);
335 d->dm_dev.bdev = bdev;
340 * Close a device that we've been using.
342 static void close_dev(struct dm_dev_internal *d, struct mapped_device *md)
347 bd_release_from_disk(d->dm_dev.bdev, dm_disk(md));
348 blkdev_put(d->dm_dev.bdev, d->dm_dev.mode);
349 d->dm_dev.bdev = NULL;
353 * If possible, this checks an area of a destination device is invalid.
355 static int device_area_is_invalid(struct dm_target *ti, struct dm_dev *dev,
356 sector_t start, sector_t len, void *data)
358 struct queue_limits *limits = data;
359 struct block_device *bdev = dev->bdev;
361 i_size_read(bdev->bd_inode) >> SECTOR_SHIFT;
362 unsigned short logical_block_size_sectors =
363 limits->logical_block_size >> SECTOR_SHIFT;
364 char b[BDEVNAME_SIZE];
369 if ((start >= dev_size) || (start + len > dev_size)) {
370 DMWARN("%s: %s too small for target: "
371 "start=%llu, len=%llu, dev_size=%llu",
372 dm_device_name(ti->table->md), bdevname(bdev, b),
373 (unsigned long long)start,
374 (unsigned long long)len,
375 (unsigned long long)dev_size);
379 if (logical_block_size_sectors <= 1)
382 if (start & (logical_block_size_sectors - 1)) {
383 DMWARN("%s: start=%llu not aligned to h/w "
384 "logical block size %u of %s",
385 dm_device_name(ti->table->md),
386 (unsigned long long)start,
387 limits->logical_block_size, bdevname(bdev, b));
391 if (len & (logical_block_size_sectors - 1)) {
392 DMWARN("%s: len=%llu not aligned to h/w "
393 "logical block size %u of %s",
394 dm_device_name(ti->table->md),
395 (unsigned long long)len,
396 limits->logical_block_size, bdevname(bdev, b));
404 * This upgrades the mode on an already open dm_dev, being
405 * careful to leave things as they were if we fail to reopen the
406 * device and not to touch the existing bdev field in case
407 * it is accessed concurrently inside dm_table_any_congested().
409 static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode,
410 struct mapped_device *md)
413 struct dm_dev_internal dd_new, dd_old;
415 dd_new = dd_old = *dd;
417 dd_new.dm_dev.mode |= new_mode;
418 dd_new.dm_dev.bdev = NULL;
420 r = open_dev(&dd_new, dd->dm_dev.bdev->bd_dev, md);
424 dd->dm_dev.mode |= new_mode;
425 close_dev(&dd_old, md);
431 * Add a device to the list, or just increment the usage count if
432 * it's already present.
434 static int __table_get_device(struct dm_table *t, struct dm_target *ti,
435 const char *path, fmode_t mode, struct dm_dev **result)
438 dev_t uninitialized_var(dev);
439 struct dm_dev_internal *dd;
440 unsigned int major, minor;
444 if (sscanf(path, "%u:%u", &major, &minor) == 2) {
445 /* Extract the major/minor numbers */
446 dev = MKDEV(major, minor);
447 if (MAJOR(dev) != major || MINOR(dev) != minor)
450 /* convert the path to a device */
451 struct block_device *bdev = lookup_bdev(path);
454 return PTR_ERR(bdev);
459 dd = find_device(&t->devices, dev);
461 dd = kmalloc(sizeof(*dd), GFP_KERNEL);
465 dd->dm_dev.mode = mode;
466 dd->dm_dev.bdev = NULL;
468 if ((r = open_dev(dd, dev, t->md))) {
473 format_dev_t(dd->dm_dev.name, dev);
475 atomic_set(&dd->count, 0);
476 list_add(&dd->list, &t->devices);
478 } else if (dd->dm_dev.mode != (mode | dd->dm_dev.mode)) {
479 r = upgrade_mode(dd, mode, t->md);
483 atomic_inc(&dd->count);
485 *result = &dd->dm_dev;
489 int dm_set_device_limits(struct dm_target *ti, struct dm_dev *dev,
490 sector_t start, sector_t len, void *data)
492 struct queue_limits *limits = data;
493 struct block_device *bdev = dev->bdev;
494 struct request_queue *q = bdev_get_queue(bdev);
495 char b[BDEVNAME_SIZE];
498 DMWARN("%s: Cannot set limits for nonexistent device %s",
499 dm_device_name(ti->table->md), bdevname(bdev, b));
503 if (bdev_stack_limits(limits, bdev, start) < 0)
504 DMWARN("%s: adding target device %s caused an alignment inconsistency: "
505 "physical_block_size=%u, logical_block_size=%u, "
506 "alignment_offset=%u, start=%llu",
507 dm_device_name(ti->table->md), bdevname(bdev, b),
508 q->limits.physical_block_size,
509 q->limits.logical_block_size,
510 q->limits.alignment_offset,
511 (unsigned long long) start << SECTOR_SHIFT);
514 * Check if merge fn is supported.
515 * If not we'll force DM to use PAGE_SIZE or
516 * smaller I/O, just to be safe.
519 if (q->merge_bvec_fn && !ti->type->merge)
520 limits->max_sectors =
521 min_not_zero(limits->max_sectors,
522 (unsigned int) (PAGE_SIZE >> 9));
525 EXPORT_SYMBOL_GPL(dm_set_device_limits);
527 int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode,
528 struct dm_dev **result)
530 return __table_get_device(ti->table, ti, path, mode, result);
535 * Decrement a devices use count and remove it if necessary.
537 void dm_put_device(struct dm_target *ti, struct dm_dev *d)
539 struct dm_dev_internal *dd = container_of(d, struct dm_dev_internal,
542 if (atomic_dec_and_test(&dd->count)) {
543 close_dev(dd, ti->table->md);
550 * Checks to see if the target joins onto the end of the table.
552 static int adjoin(struct dm_table *table, struct dm_target *ti)
554 struct dm_target *prev;
556 if (!table->num_targets)
559 prev = &table->targets[table->num_targets - 1];
560 return (ti->begin == (prev->begin + prev->len));
564 * Used to dynamically allocate the arg array.
566 static char **realloc_argv(unsigned *array_size, char **old_argv)
571 new_size = *array_size ? *array_size * 2 : 64;
572 argv = kmalloc(new_size * sizeof(*argv), GFP_KERNEL);
574 memcpy(argv, old_argv, *array_size * sizeof(*argv));
575 *array_size = new_size;
583 * Destructively splits up the argument list to pass to ctr.
585 int dm_split_args(int *argc, char ***argvp, char *input)
587 char *start, *end = input, *out, **argv = NULL;
588 unsigned array_size = 0;
597 argv = realloc_argv(&array_size, argv);
602 /* Skip whitespace */
603 start = skip_spaces(end);
606 break; /* success, we hit the end */
608 /* 'out' is used to remove any back-quotes */
611 /* Everything apart from '\0' can be quoted */
612 if (*end == '\\' && *(end + 1)) {
619 break; /* end of token */
624 /* have we already filled the array ? */
625 if ((*argc + 1) > array_size) {
626 argv = realloc_argv(&array_size, argv);
631 /* we know this is whitespace */
635 /* terminate the string and put it in the array */
646 * Impose necessary and sufficient conditions on a devices's table such
647 * that any incoming bio which respects its logical_block_size can be
648 * processed successfully. If it falls across the boundary between
649 * two or more targets, the size of each piece it gets split into must
650 * be compatible with the logical_block_size of the target processing it.
652 static int validate_hardware_logical_block_alignment(struct dm_table *table,
653 struct queue_limits *limits)
656 * This function uses arithmetic modulo the logical_block_size
657 * (in units of 512-byte sectors).
659 unsigned short device_logical_block_size_sects =
660 limits->logical_block_size >> SECTOR_SHIFT;
663 * Offset of the start of the next table entry, mod logical_block_size.
665 unsigned short next_target_start = 0;
668 * Given an aligned bio that extends beyond the end of a
669 * target, how many sectors must the next target handle?
671 unsigned short remaining = 0;
673 struct dm_target *uninitialized_var(ti);
674 struct queue_limits ti_limits;
678 * Check each entry in the table in turn.
680 while (i < dm_table_get_num_targets(table)) {
681 ti = dm_table_get_target(table, i++);
683 blk_set_default_limits(&ti_limits);
685 /* combine all target devices' limits */
686 if (ti->type->iterate_devices)
687 ti->type->iterate_devices(ti, dm_set_device_limits,
691 * If the remaining sectors fall entirely within this
692 * table entry are they compatible with its logical_block_size?
694 if (remaining < ti->len &&
695 remaining & ((ti_limits.logical_block_size >>
700 (unsigned short) ((next_target_start + ti->len) &
701 (device_logical_block_size_sects - 1));
702 remaining = next_target_start ?
703 device_logical_block_size_sects - next_target_start : 0;
707 DMWARN("%s: table line %u (start sect %llu len %llu) "
708 "not aligned to h/w logical block size %u",
709 dm_device_name(table->md), i,
710 (unsigned long long) ti->begin,
711 (unsigned long long) ti->len,
712 limits->logical_block_size);
719 int dm_table_add_target(struct dm_table *t, const char *type,
720 sector_t start, sector_t len, char *params)
722 int r = -EINVAL, argc;
724 struct dm_target *tgt;
726 if ((r = check_space(t)))
729 tgt = t->targets + t->num_targets;
730 memset(tgt, 0, sizeof(*tgt));
733 DMERR("%s: zero-length target", dm_device_name(t->md));
737 tgt->type = dm_get_target_type(type);
739 DMERR("%s: %s: unknown target type", dm_device_name(t->md),
747 tgt->error = "Unknown error";
750 * Does this target adjoin the previous one ?
752 if (!adjoin(t, tgt)) {
753 tgt->error = "Gap in table";
758 r = dm_split_args(&argc, &argv, params);
760 tgt->error = "couldn't split parameters (insufficient memory)";
764 r = tgt->type->ctr(tgt, argc, argv);
769 t->highs[t->num_targets++] = tgt->begin + tgt->len - 1;
771 if (!tgt->num_discard_requests)
772 t->discards_supported = 0;
777 DMERR("%s: %s: %s", dm_device_name(t->md), type, tgt->error);
778 dm_put_target_type(tgt->type);
782 static int dm_table_set_type(struct dm_table *t)
785 unsigned bio_based = 0, request_based = 0;
786 struct dm_target *tgt;
787 struct dm_dev_internal *dd;
788 struct list_head *devices;
790 for (i = 0; i < t->num_targets; i++) {
791 tgt = t->targets + i;
792 if (dm_target_request_based(tgt))
797 if (bio_based && request_based) {
798 DMWARN("Inconsistent table: different target types"
799 " can't be mixed up");
805 /* We must use this table as bio-based */
806 t->type = DM_TYPE_BIO_BASED;
810 BUG_ON(!request_based); /* No targets in this table */
812 /* Non-request-stackable devices can't be used for request-based dm */
813 devices = dm_table_get_devices(t);
814 list_for_each_entry(dd, devices, list) {
815 if (!blk_queue_stackable(bdev_get_queue(dd->dm_dev.bdev))) {
816 DMWARN("table load rejected: including"
817 " non-request-stackable devices");
823 * Request-based dm supports only tables that have a single target now.
824 * To support multiple targets, request splitting support is needed,
825 * and that needs lots of changes in the block-layer.
826 * (e.g. request completion process for partial completion.)
828 if (t->num_targets > 1) {
829 DMWARN("Request-based dm doesn't support multiple targets yet");
833 t->type = DM_TYPE_REQUEST_BASED;
838 unsigned dm_table_get_type(struct dm_table *t)
843 bool dm_table_request_based(struct dm_table *t)
845 return dm_table_get_type(t) == DM_TYPE_REQUEST_BASED;
848 int dm_table_alloc_md_mempools(struct dm_table *t)
850 unsigned type = dm_table_get_type(t);
852 if (unlikely(type == DM_TYPE_NONE)) {
853 DMWARN("no table type is set, can't allocate mempools");
857 t->mempools = dm_alloc_md_mempools(type);
864 void dm_table_free_md_mempools(struct dm_table *t)
866 dm_free_md_mempools(t->mempools);
870 struct dm_md_mempools *dm_table_get_md_mempools(struct dm_table *t)
875 static int setup_indexes(struct dm_table *t)
878 unsigned int total = 0;
881 /* allocate the space for *all* the indexes */
882 for (i = t->depth - 2; i >= 0; i--) {
883 t->counts[i] = dm_div_up(t->counts[i + 1], CHILDREN_PER_NODE);
884 total += t->counts[i];
887 indexes = (sector_t *) dm_vcalloc(total, (unsigned long) NODE_SIZE);
891 /* set up internal nodes, bottom-up */
892 for (i = t->depth - 2; i >= 0; i--) {
893 t->index[i] = indexes;
894 indexes += (KEYS_PER_NODE * t->counts[i]);
895 setup_btree_index(i, t);
902 * Builds the btree to index the map.
904 static int dm_table_build_index(struct dm_table *t)
907 unsigned int leaf_nodes;
909 /* how many indexes will the btree have ? */
910 leaf_nodes = dm_div_up(t->num_targets, KEYS_PER_NODE);
911 t->depth = 1 + int_log(leaf_nodes, CHILDREN_PER_NODE);
913 /* leaf layer has already been set up */
914 t->counts[t->depth - 1] = leaf_nodes;
915 t->index[t->depth - 1] = t->highs;
918 r = setup_indexes(t);
924 * Register the mapped device for blk_integrity support if
925 * the underlying devices support it.
927 static int dm_table_prealloc_integrity(struct dm_table *t, struct mapped_device *md)
929 struct list_head *devices = dm_table_get_devices(t);
930 struct dm_dev_internal *dd;
932 list_for_each_entry(dd, devices, list)
933 if (bdev_get_integrity(dd->dm_dev.bdev))
934 return blk_integrity_register(dm_disk(md), NULL);
940 * Prepares the table for use by building the indices,
941 * setting the type, and allocating mempools.
943 int dm_table_complete(struct dm_table *t)
947 r = dm_table_set_type(t);
949 DMERR("unable to set table type");
953 r = dm_table_build_index(t);
955 DMERR("unable to build btrees");
959 r = dm_table_prealloc_integrity(t, t->md);
961 DMERR("could not register integrity profile.");
965 r = dm_table_alloc_md_mempools(t);
967 DMERR("unable to allocate mempools");
972 static DEFINE_MUTEX(_event_lock);
973 void dm_table_event_callback(struct dm_table *t,
974 void (*fn)(void *), void *context)
976 mutex_lock(&_event_lock);
978 t->event_context = context;
979 mutex_unlock(&_event_lock);
982 void dm_table_event(struct dm_table *t)
985 * You can no longer call dm_table_event() from interrupt
986 * context, use a bottom half instead.
988 BUG_ON(in_interrupt());
990 mutex_lock(&_event_lock);
992 t->event_fn(t->event_context);
993 mutex_unlock(&_event_lock);
996 sector_t dm_table_get_size(struct dm_table *t)
998 return t->num_targets ? (t->highs[t->num_targets - 1] + 1) : 0;
1001 struct dm_target *dm_table_get_target(struct dm_table *t, unsigned int index)
1003 if (index >= t->num_targets)
1006 return t->targets + index;
1010 * Search the btree for the correct target.
1012 * Caller should check returned pointer with dm_target_is_valid()
1013 * to trap I/O beyond end of device.
1015 struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector)
1017 unsigned int l, n = 0, k = 0;
1020 for (l = 0; l < t->depth; l++) {
1021 n = get_child(n, k);
1022 node = get_node(t, l, n);
1024 for (k = 0; k < KEYS_PER_NODE; k++)
1025 if (node[k] >= sector)
1029 return &t->targets[(KEYS_PER_NODE * n) + k];
1033 * Establish the new table's queue_limits and validate them.
1035 int dm_calculate_queue_limits(struct dm_table *table,
1036 struct queue_limits *limits)
1038 struct dm_target *uninitialized_var(ti);
1039 struct queue_limits ti_limits;
1042 blk_set_default_limits(limits);
1044 while (i < dm_table_get_num_targets(table)) {
1045 blk_set_default_limits(&ti_limits);
1047 ti = dm_table_get_target(table, i++);
1049 if (!ti->type->iterate_devices)
1050 goto combine_limits;
1053 * Combine queue limits of all the devices this target uses.
1055 ti->type->iterate_devices(ti, dm_set_device_limits,
1058 /* Set I/O hints portion of queue limits */
1059 if (ti->type->io_hints)
1060 ti->type->io_hints(ti, &ti_limits);
1063 * Check each device area is consistent with the target's
1064 * overall queue limits.
1066 if (ti->type->iterate_devices(ti, device_area_is_invalid,
1072 * Merge this target's queue limits into the overall limits
1075 if (blk_stack_limits(limits, &ti_limits, 0) < 0)
1076 DMWARN("%s: adding target device "
1077 "(start sect %llu len %llu) "
1078 "caused an alignment inconsistency",
1079 dm_device_name(table->md),
1080 (unsigned long long) ti->begin,
1081 (unsigned long long) ti->len);
1084 return validate_hardware_logical_block_alignment(table, limits);
1088 * Set the integrity profile for this device if all devices used have
1089 * matching profiles.
1091 static void dm_table_set_integrity(struct dm_table *t)
1093 struct list_head *devices = dm_table_get_devices(t);
1094 struct dm_dev_internal *prev = NULL, *dd = NULL;
1096 if (!blk_get_integrity(dm_disk(t->md)))
1099 list_for_each_entry(dd, devices, list) {
1101 blk_integrity_compare(prev->dm_dev.bdev->bd_disk,
1102 dd->dm_dev.bdev->bd_disk) < 0) {
1103 DMWARN("%s: integrity not set: %s and %s mismatch",
1104 dm_device_name(t->md),
1105 prev->dm_dev.bdev->bd_disk->disk_name,
1106 dd->dm_dev.bdev->bd_disk->disk_name);
1112 if (!prev || !bdev_get_integrity(prev->dm_dev.bdev))
1115 blk_integrity_register(dm_disk(t->md),
1116 bdev_get_integrity(prev->dm_dev.bdev));
1121 blk_integrity_register(dm_disk(t->md), NULL);
1126 void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q,
1127 struct queue_limits *limits)
1130 * Copy table's limits to the DM device's request_queue
1132 q->limits = *limits;
1134 if (limits->no_cluster)
1135 queue_flag_clear_unlocked(QUEUE_FLAG_CLUSTER, q);
1137 queue_flag_set_unlocked(QUEUE_FLAG_CLUSTER, q);
1139 if (!dm_table_supports_discards(t))
1140 queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD, q);
1142 queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, q);
1144 dm_table_set_integrity(t);
1147 * QUEUE_FLAG_STACKABLE must be set after all queue settings are
1148 * visible to other CPUs because, once the flag is set, incoming bios
1149 * are processed by request-based dm, which refers to the queue
1151 * Until the flag set, bios are passed to bio-based dm and queued to
1152 * md->deferred where queue settings are not needed yet.
1153 * Those bios are passed to request-based dm at the resume time.
1156 if (dm_table_request_based(t))
1157 queue_flag_set_unlocked(QUEUE_FLAG_STACKABLE, q);
1160 unsigned int dm_table_get_num_targets(struct dm_table *t)
1162 return t->num_targets;
1165 struct list_head *dm_table_get_devices(struct dm_table *t)
1170 fmode_t dm_table_get_mode(struct dm_table *t)
1175 static void suspend_targets(struct dm_table *t, unsigned postsuspend)
1177 int i = t->num_targets;
1178 struct dm_target *ti = t->targets;
1182 if (ti->type->postsuspend)
1183 ti->type->postsuspend(ti);
1184 } else if (ti->type->presuspend)
1185 ti->type->presuspend(ti);
1191 void dm_table_presuspend_targets(struct dm_table *t)
1196 suspend_targets(t, 0);
1199 void dm_table_postsuspend_targets(struct dm_table *t)
1204 suspend_targets(t, 1);
1207 int dm_table_resume_targets(struct dm_table *t)
1211 for (i = 0; i < t->num_targets; i++) {
1212 struct dm_target *ti = t->targets + i;
1214 if (!ti->type->preresume)
1217 r = ti->type->preresume(ti);
1222 for (i = 0; i < t->num_targets; i++) {
1223 struct dm_target *ti = t->targets + i;
1225 if (ti->type->resume)
1226 ti->type->resume(ti);
1232 int dm_table_any_congested(struct dm_table *t, int bdi_bits)
1234 struct dm_dev_internal *dd;
1235 struct list_head *devices = dm_table_get_devices(t);
1238 list_for_each_entry(dd, devices, list) {
1239 struct request_queue *q = bdev_get_queue(dd->dm_dev.bdev);
1240 char b[BDEVNAME_SIZE];
1243 r |= bdi_congested(&q->backing_dev_info, bdi_bits);
1245 DMWARN_LIMIT("%s: any_congested: nonexistent device %s",
1246 dm_device_name(t->md),
1247 bdevname(dd->dm_dev.bdev, b));
1253 int dm_table_any_busy_target(struct dm_table *t)
1256 struct dm_target *ti;
1258 for (i = 0; i < t->num_targets; i++) {
1259 ti = t->targets + i;
1260 if (ti->type->busy && ti->type->busy(ti))
1267 void dm_table_unplug_all(struct dm_table *t)
1269 struct dm_dev_internal *dd;
1270 struct list_head *devices = dm_table_get_devices(t);
1272 list_for_each_entry(dd, devices, list) {
1273 struct request_queue *q = bdev_get_queue(dd->dm_dev.bdev);
1274 char b[BDEVNAME_SIZE];
1279 DMWARN_LIMIT("%s: Cannot unplug nonexistent device %s",
1280 dm_device_name(t->md),
1281 bdevname(dd->dm_dev.bdev, b));
1285 struct mapped_device *dm_table_get_md(struct dm_table *t)
1290 static int device_discard_capable(struct dm_target *ti, struct dm_dev *dev,
1291 sector_t start, sector_t len, void *data)
1293 struct request_queue *q = bdev_get_queue(dev->bdev);
1295 return q && blk_queue_discard(q);
1298 bool dm_table_supports_discards(struct dm_table *t)
1300 struct dm_target *ti;
1303 if (!t->discards_supported)
1307 * Ensure that at least one underlying device supports discards.
1308 * t->devices includes internal dm devices such as mirror logs
1309 * so we need to use iterate_devices here, which targets
1310 * supporting discard must provide.
1312 while (i < dm_table_get_num_targets(t)) {
1313 ti = dm_table_get_target(t, i++);
1315 if (ti->type->iterate_devices &&
1316 ti->type->iterate_devices(ti, device_discard_capable, NULL))
1323 EXPORT_SYMBOL(dm_vcalloc);
1324 EXPORT_SYMBOL(dm_get_device);
1325 EXPORT_SYMBOL(dm_put_device);
1326 EXPORT_SYMBOL(dm_table_event);
1327 EXPORT_SYMBOL(dm_table_get_size);
1328 EXPORT_SYMBOL(dm_table_get_mode);
1329 EXPORT_SYMBOL(dm_table_get_md);
1330 EXPORT_SYMBOL(dm_table_put);
1331 EXPORT_SYMBOL(dm_table_get);
1332 EXPORT_SYMBOL(dm_table_unplug_all);