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>
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 integrity_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
;
75 struct list_head target_callbacks
;
79 * Similar to ceiling(log_size(n))
81 static unsigned int int_log(unsigned int n
, unsigned int base
)
86 n
= dm_div_up(n
, base
);
94 * Calculate the index of the child node of the n'th node k'th key.
96 static inline unsigned int get_child(unsigned int n
, unsigned int k
)
98 return (n
* CHILDREN_PER_NODE
) + k
;
102 * Return the n'th node of level l from table t.
104 static inline sector_t
*get_node(struct dm_table
*t
,
105 unsigned int l
, unsigned int n
)
107 return t
->index
[l
] + (n
* KEYS_PER_NODE
);
111 * Return the highest key that you could lookup from the n'th
112 * node on level l of the btree.
114 static sector_t
high(struct dm_table
*t
, unsigned int l
, unsigned int n
)
116 for (; l
< t
->depth
- 1; l
++)
117 n
= get_child(n
, CHILDREN_PER_NODE
- 1);
119 if (n
>= t
->counts
[l
])
120 return (sector_t
) - 1;
122 return get_node(t
, l
, n
)[KEYS_PER_NODE
- 1];
126 * Fills in a level of the btree based on the highs of the level
129 static int setup_btree_index(unsigned int l
, struct dm_table
*t
)
134 for (n
= 0U; n
< t
->counts
[l
]; n
++) {
135 node
= get_node(t
, l
, n
);
137 for (k
= 0U; k
< KEYS_PER_NODE
; k
++)
138 node
[k
] = high(t
, l
+ 1, get_child(n
, k
));
144 void *dm_vcalloc(unsigned long nmemb
, unsigned long elem_size
)
150 * Check that we're not going to overflow.
152 if (nmemb
> (ULONG_MAX
/ elem_size
))
155 size
= nmemb
* elem_size
;
156 addr
= vzalloc(size
);
160 EXPORT_SYMBOL(dm_vcalloc
);
163 * highs, and targets are managed as dynamic arrays during a
166 static int alloc_targets(struct dm_table
*t
, unsigned int num
)
169 struct dm_target
*n_targets
;
170 int n
= t
->num_targets
;
173 * Allocate both the target array and offset array at once.
174 * Append an empty entry to catch sectors beyond the end of
177 n_highs
= (sector_t
*) dm_vcalloc(num
+ 1, sizeof(struct dm_target
) +
182 n_targets
= (struct dm_target
*) (n_highs
+ num
);
185 memcpy(n_highs
, t
->highs
, sizeof(*n_highs
) * n
);
186 memcpy(n_targets
, t
->targets
, sizeof(*n_targets
) * n
);
189 memset(n_highs
+ n
, -1, sizeof(*n_highs
) * (num
- n
));
192 t
->num_allocated
= num
;
194 t
->targets
= n_targets
;
199 int dm_table_create(struct dm_table
**result
, fmode_t mode
,
200 unsigned num_targets
, struct mapped_device
*md
)
202 struct dm_table
*t
= kzalloc(sizeof(*t
), GFP_KERNEL
);
207 INIT_LIST_HEAD(&t
->devices
);
208 INIT_LIST_HEAD(&t
->target_callbacks
);
209 atomic_set(&t
->holders
, 0);
212 num_targets
= KEYS_PER_NODE
;
214 num_targets
= dm_round_up(num_targets
, KEYS_PER_NODE
);
216 if (alloc_targets(t
, num_targets
)) {
228 static void free_devices(struct list_head
*devices
)
230 struct list_head
*tmp
, *next
;
232 list_for_each_safe(tmp
, next
, devices
) {
233 struct dm_dev_internal
*dd
=
234 list_entry(tmp
, struct dm_dev_internal
, list
);
235 DMWARN("dm_table_destroy: dm_put_device call missing for %s",
241 void dm_table_destroy(struct dm_table
*t
)
248 while (atomic_read(&t
->holders
))
252 /* free the indexes */
254 vfree(t
->index
[t
->depth
- 2]);
256 /* free the targets */
257 for (i
= 0; i
< t
->num_targets
; i
++) {
258 struct dm_target
*tgt
= t
->targets
+ i
;
263 dm_put_target_type(tgt
->type
);
268 /* free the device list */
269 if (t
->devices
.next
!= &t
->devices
)
270 free_devices(&t
->devices
);
272 dm_free_md_mempools(t
->mempools
);
277 void dm_table_get(struct dm_table
*t
)
279 atomic_inc(&t
->holders
);
281 EXPORT_SYMBOL(dm_table_get
);
283 void dm_table_put(struct dm_table
*t
)
288 smp_mb__before_atomic_dec();
289 atomic_dec(&t
->holders
);
291 EXPORT_SYMBOL(dm_table_put
);
294 * Checks to see if we need to extend highs or targets.
296 static inline int check_space(struct dm_table
*t
)
298 if (t
->num_targets
>= t
->num_allocated
)
299 return alloc_targets(t
, t
->num_allocated
* 2);
305 * See if we've already got a device in the list.
307 static struct dm_dev_internal
*find_device(struct list_head
*l
, dev_t dev
)
309 struct dm_dev_internal
*dd
;
311 list_for_each_entry (dd
, l
, list
)
312 if (dd
->dm_dev
.bdev
->bd_dev
== dev
)
319 * Open a device so we can use it as a map destination.
321 static int open_dev(struct dm_dev_internal
*d
, dev_t dev
,
322 struct mapped_device
*md
)
324 static char *_claim_ptr
= "I belong to device-mapper";
325 struct block_device
*bdev
;
329 BUG_ON(d
->dm_dev
.bdev
);
331 bdev
= blkdev_get_by_dev(dev
, d
->dm_dev
.mode
| FMODE_EXCL
, _claim_ptr
);
333 return PTR_ERR(bdev
);
335 r
= bd_link_disk_holder(bdev
, dm_disk(md
));
337 blkdev_put(bdev
, d
->dm_dev
.mode
| FMODE_EXCL
);
341 d
->dm_dev
.bdev
= bdev
;
346 * Close a device that we've been using.
348 static void close_dev(struct dm_dev_internal
*d
, struct mapped_device
*md
)
353 bd_unlink_disk_holder(d
->dm_dev
.bdev
, dm_disk(md
));
354 blkdev_put(d
->dm_dev
.bdev
, d
->dm_dev
.mode
| FMODE_EXCL
);
355 d
->dm_dev
.bdev
= NULL
;
359 * If possible, this checks an area of a destination device is invalid.
361 static int device_area_is_invalid(struct dm_target
*ti
, struct dm_dev
*dev
,
362 sector_t start
, sector_t len
, void *data
)
364 struct request_queue
*q
;
365 struct queue_limits
*limits
= data
;
366 struct block_device
*bdev
= dev
->bdev
;
368 i_size_read(bdev
->bd_inode
) >> SECTOR_SHIFT
;
369 unsigned short logical_block_size_sectors
=
370 limits
->logical_block_size
>> SECTOR_SHIFT
;
371 char b
[BDEVNAME_SIZE
];
374 * Some devices exist without request functions,
375 * such as loop devices not yet bound to backing files.
376 * Forbid the use of such devices.
378 q
= bdev_get_queue(bdev
);
379 if (!q
|| !q
->make_request_fn
) {
380 DMWARN("%s: %s is not yet initialised: "
381 "start=%llu, len=%llu, dev_size=%llu",
382 dm_device_name(ti
->table
->md
), bdevname(bdev
, b
),
383 (unsigned long long)start
,
384 (unsigned long long)len
,
385 (unsigned long long)dev_size
);
392 if ((start
>= dev_size
) || (start
+ len
> dev_size
)) {
393 DMWARN("%s: %s too small for target: "
394 "start=%llu, len=%llu, dev_size=%llu",
395 dm_device_name(ti
->table
->md
), bdevname(bdev
, b
),
396 (unsigned long long)start
,
397 (unsigned long long)len
,
398 (unsigned long long)dev_size
);
402 if (logical_block_size_sectors
<= 1)
405 if (start
& (logical_block_size_sectors
- 1)) {
406 DMWARN("%s: start=%llu not aligned to h/w "
407 "logical block size %u of %s",
408 dm_device_name(ti
->table
->md
),
409 (unsigned long long)start
,
410 limits
->logical_block_size
, bdevname(bdev
, b
));
414 if (len
& (logical_block_size_sectors
- 1)) {
415 DMWARN("%s: len=%llu not aligned to h/w "
416 "logical block size %u of %s",
417 dm_device_name(ti
->table
->md
),
418 (unsigned long long)len
,
419 limits
->logical_block_size
, bdevname(bdev
, b
));
427 * This upgrades the mode on an already open dm_dev, being
428 * careful to leave things as they were if we fail to reopen the
429 * device and not to touch the existing bdev field in case
430 * it is accessed concurrently inside dm_table_any_congested().
432 static int upgrade_mode(struct dm_dev_internal
*dd
, fmode_t new_mode
,
433 struct mapped_device
*md
)
436 struct dm_dev_internal dd_new
, dd_old
;
438 dd_new
= dd_old
= *dd
;
440 dd_new
.dm_dev
.mode
|= new_mode
;
441 dd_new
.dm_dev
.bdev
= NULL
;
443 r
= open_dev(&dd_new
, dd
->dm_dev
.bdev
->bd_dev
, md
);
447 dd
->dm_dev
.mode
|= new_mode
;
448 close_dev(&dd_old
, md
);
454 * Add a device to the list, or just increment the usage count if
455 * it's already present.
457 int dm_get_device(struct dm_target
*ti
, const char *path
, fmode_t mode
,
458 struct dm_dev
**result
)
461 dev_t
uninitialized_var(dev
);
462 struct dm_dev_internal
*dd
;
463 unsigned int major
, minor
;
464 struct dm_table
*t
= ti
->table
;
468 if (sscanf(path
, "%u:%u", &major
, &minor
) == 2) {
469 /* Extract the major/minor numbers */
470 dev
= MKDEV(major
, minor
);
471 if (MAJOR(dev
) != major
|| MINOR(dev
) != minor
)
474 /* convert the path to a device */
475 struct block_device
*bdev
= lookup_bdev(path
);
478 return PTR_ERR(bdev
);
483 dd
= find_device(&t
->devices
, dev
);
485 dd
= kmalloc(sizeof(*dd
), GFP_KERNEL
);
489 dd
->dm_dev
.mode
= mode
;
490 dd
->dm_dev
.bdev
= NULL
;
492 if ((r
= open_dev(dd
, dev
, t
->md
))) {
497 format_dev_t(dd
->dm_dev
.name
, dev
);
499 atomic_set(&dd
->count
, 0);
500 list_add(&dd
->list
, &t
->devices
);
502 } else if (dd
->dm_dev
.mode
!= (mode
| dd
->dm_dev
.mode
)) {
503 r
= upgrade_mode(dd
, mode
, t
->md
);
507 atomic_inc(&dd
->count
);
509 *result
= &dd
->dm_dev
;
512 EXPORT_SYMBOL(dm_get_device
);
514 int dm_set_device_limits(struct dm_target
*ti
, struct dm_dev
*dev
,
515 sector_t start
, sector_t len
, void *data
)
517 struct queue_limits
*limits
= data
;
518 struct block_device
*bdev
= dev
->bdev
;
519 struct request_queue
*q
= bdev_get_queue(bdev
);
520 char b
[BDEVNAME_SIZE
];
523 DMWARN("%s: Cannot set limits for nonexistent device %s",
524 dm_device_name(ti
->table
->md
), bdevname(bdev
, b
));
528 if (bdev_stack_limits(limits
, bdev
, start
) < 0)
529 DMWARN("%s: adding target device %s caused an alignment inconsistency: "
530 "physical_block_size=%u, logical_block_size=%u, "
531 "alignment_offset=%u, start=%llu",
532 dm_device_name(ti
->table
->md
), bdevname(bdev
, b
),
533 q
->limits
.physical_block_size
,
534 q
->limits
.logical_block_size
,
535 q
->limits
.alignment_offset
,
536 (unsigned long long) start
<< SECTOR_SHIFT
);
539 * Check if merge fn is supported.
540 * If not we'll force DM to use PAGE_SIZE or
541 * smaller I/O, just to be safe.
544 if (q
->merge_bvec_fn
&& !ti
->type
->merge
)
545 blk_limits_max_hw_sectors(limits
,
546 (unsigned int) (PAGE_SIZE
>> 9));
549 EXPORT_SYMBOL_GPL(dm_set_device_limits
);
552 * Decrement a device's use count and remove it if necessary.
554 void dm_put_device(struct dm_target
*ti
, struct dm_dev
*d
)
556 struct dm_dev_internal
*dd
= container_of(d
, struct dm_dev_internal
,
559 if (atomic_dec_and_test(&dd
->count
)) {
560 close_dev(dd
, ti
->table
->md
);
565 EXPORT_SYMBOL(dm_put_device
);
568 * Checks to see if the target joins onto the end of the table.
570 static int adjoin(struct dm_table
*table
, struct dm_target
*ti
)
572 struct dm_target
*prev
;
574 if (!table
->num_targets
)
577 prev
= &table
->targets
[table
->num_targets
- 1];
578 return (ti
->begin
== (prev
->begin
+ prev
->len
));
582 * Used to dynamically allocate the arg array.
584 static char **realloc_argv(unsigned *array_size
, char **old_argv
)
589 new_size
= *array_size
? *array_size
* 2 : 64;
590 argv
= kmalloc(new_size
* sizeof(*argv
), GFP_KERNEL
);
592 memcpy(argv
, old_argv
, *array_size
* sizeof(*argv
));
593 *array_size
= new_size
;
601 * Destructively splits up the argument list to pass to ctr.
603 int dm_split_args(int *argc
, char ***argvp
, char *input
)
605 char *start
, *end
= input
, *out
, **argv
= NULL
;
606 unsigned array_size
= 0;
615 argv
= realloc_argv(&array_size
, argv
);
620 /* Skip whitespace */
621 start
= skip_spaces(end
);
624 break; /* success, we hit the end */
626 /* 'out' is used to remove any back-quotes */
629 /* Everything apart from '\0' can be quoted */
630 if (*end
== '\\' && *(end
+ 1)) {
637 break; /* end of token */
642 /* have we already filled the array ? */
643 if ((*argc
+ 1) > array_size
) {
644 argv
= realloc_argv(&array_size
, argv
);
649 /* we know this is whitespace */
653 /* terminate the string and put it in the array */
664 * Impose necessary and sufficient conditions on a devices's table such
665 * that any incoming bio which respects its logical_block_size can be
666 * processed successfully. If it falls across the boundary between
667 * two or more targets, the size of each piece it gets split into must
668 * be compatible with the logical_block_size of the target processing it.
670 static int validate_hardware_logical_block_alignment(struct dm_table
*table
,
671 struct queue_limits
*limits
)
674 * This function uses arithmetic modulo the logical_block_size
675 * (in units of 512-byte sectors).
677 unsigned short device_logical_block_size_sects
=
678 limits
->logical_block_size
>> SECTOR_SHIFT
;
681 * Offset of the start of the next table entry, mod logical_block_size.
683 unsigned short next_target_start
= 0;
686 * Given an aligned bio that extends beyond the end of a
687 * target, how many sectors must the next target handle?
689 unsigned short remaining
= 0;
691 struct dm_target
*uninitialized_var(ti
);
692 struct queue_limits ti_limits
;
696 * Check each entry in the table in turn.
698 while (i
< dm_table_get_num_targets(table
)) {
699 ti
= dm_table_get_target(table
, i
++);
701 blk_set_default_limits(&ti_limits
);
703 /* combine all target devices' limits */
704 if (ti
->type
->iterate_devices
)
705 ti
->type
->iterate_devices(ti
, dm_set_device_limits
,
709 * If the remaining sectors fall entirely within this
710 * table entry are they compatible with its logical_block_size?
712 if (remaining
< ti
->len
&&
713 remaining
& ((ti_limits
.logical_block_size
>>
718 (unsigned short) ((next_target_start
+ ti
->len
) &
719 (device_logical_block_size_sects
- 1));
720 remaining
= next_target_start
?
721 device_logical_block_size_sects
- next_target_start
: 0;
725 DMWARN("%s: table line %u (start sect %llu len %llu) "
726 "not aligned to h/w logical block size %u",
727 dm_device_name(table
->md
), i
,
728 (unsigned long long) ti
->begin
,
729 (unsigned long long) ti
->len
,
730 limits
->logical_block_size
);
737 int dm_table_add_target(struct dm_table
*t
, const char *type
,
738 sector_t start
, sector_t len
, char *params
)
740 int r
= -EINVAL
, argc
;
742 struct dm_target
*tgt
;
744 if ((r
= check_space(t
)))
747 tgt
= t
->targets
+ t
->num_targets
;
748 memset(tgt
, 0, sizeof(*tgt
));
751 DMERR("%s: zero-length target", dm_device_name(t
->md
));
755 tgt
->type
= dm_get_target_type(type
);
757 DMERR("%s: %s: unknown target type", dm_device_name(t
->md
),
765 tgt
->error
= "Unknown error";
768 * Does this target adjoin the previous one ?
770 if (!adjoin(t
, tgt
)) {
771 tgt
->error
= "Gap in table";
776 r
= dm_split_args(&argc
, &argv
, params
);
778 tgt
->error
= "couldn't split parameters (insufficient memory)";
782 r
= tgt
->type
->ctr(tgt
, argc
, argv
);
787 t
->highs
[t
->num_targets
++] = tgt
->begin
+ tgt
->len
- 1;
789 if (!tgt
->num_discard_requests
&& tgt
->discards_supported
)
790 DMWARN("%s: %s: ignoring discards_supported because num_discard_requests is zero.",
791 dm_device_name(t
->md
), type
);
796 DMERR("%s: %s: %s", dm_device_name(t
->md
), type
, tgt
->error
);
797 dm_put_target_type(tgt
->type
);
801 static int dm_table_set_type(struct dm_table
*t
)
804 unsigned bio_based
= 0, request_based
= 0;
805 struct dm_target
*tgt
;
806 struct dm_dev_internal
*dd
;
807 struct list_head
*devices
;
809 for (i
= 0; i
< t
->num_targets
; i
++) {
810 tgt
= t
->targets
+ i
;
811 if (dm_target_request_based(tgt
))
816 if (bio_based
&& request_based
) {
817 DMWARN("Inconsistent table: different target types"
818 " can't be mixed up");
824 /* We must use this table as bio-based */
825 t
->type
= DM_TYPE_BIO_BASED
;
829 BUG_ON(!request_based
); /* No targets in this table */
831 /* Non-request-stackable devices can't be used for request-based dm */
832 devices
= dm_table_get_devices(t
);
833 list_for_each_entry(dd
, devices
, list
) {
834 if (!blk_queue_stackable(bdev_get_queue(dd
->dm_dev
.bdev
))) {
835 DMWARN("table load rejected: including"
836 " non-request-stackable devices");
842 * Request-based dm supports only tables that have a single target now.
843 * To support multiple targets, request splitting support is needed,
844 * and that needs lots of changes in the block-layer.
845 * (e.g. request completion process for partial completion.)
847 if (t
->num_targets
> 1) {
848 DMWARN("Request-based dm doesn't support multiple targets yet");
852 t
->type
= DM_TYPE_REQUEST_BASED
;
857 unsigned dm_table_get_type(struct dm_table
*t
)
862 bool dm_table_request_based(struct dm_table
*t
)
864 return dm_table_get_type(t
) == DM_TYPE_REQUEST_BASED
;
867 int dm_table_alloc_md_mempools(struct dm_table
*t
)
869 unsigned type
= dm_table_get_type(t
);
871 if (unlikely(type
== DM_TYPE_NONE
)) {
872 DMWARN("no table type is set, can't allocate mempools");
876 t
->mempools
= dm_alloc_md_mempools(type
, t
->integrity_supported
);
883 void dm_table_free_md_mempools(struct dm_table
*t
)
885 dm_free_md_mempools(t
->mempools
);
889 struct dm_md_mempools
*dm_table_get_md_mempools(struct dm_table
*t
)
894 static int setup_indexes(struct dm_table
*t
)
897 unsigned int total
= 0;
900 /* allocate the space for *all* the indexes */
901 for (i
= t
->depth
- 2; i
>= 0; i
--) {
902 t
->counts
[i
] = dm_div_up(t
->counts
[i
+ 1], CHILDREN_PER_NODE
);
903 total
+= t
->counts
[i
];
906 indexes
= (sector_t
*) dm_vcalloc(total
, (unsigned long) NODE_SIZE
);
910 /* set up internal nodes, bottom-up */
911 for (i
= t
->depth
- 2; i
>= 0; i
--) {
912 t
->index
[i
] = indexes
;
913 indexes
+= (KEYS_PER_NODE
* t
->counts
[i
]);
914 setup_btree_index(i
, t
);
921 * Builds the btree to index the map.
923 static int dm_table_build_index(struct dm_table
*t
)
926 unsigned int leaf_nodes
;
928 /* how many indexes will the btree have ? */
929 leaf_nodes
= dm_div_up(t
->num_targets
, KEYS_PER_NODE
);
930 t
->depth
= 1 + int_log(leaf_nodes
, CHILDREN_PER_NODE
);
932 /* leaf layer has already been set up */
933 t
->counts
[t
->depth
- 1] = leaf_nodes
;
934 t
->index
[t
->depth
- 1] = t
->highs
;
937 r
= setup_indexes(t
);
943 * Get a disk whose integrity profile reflects the table's profile.
944 * If %match_all is true, all devices' profiles must match.
945 * If %match_all is false, all devices must at least have an
946 * allocated integrity profile; but uninitialized is ok.
947 * Returns NULL if integrity support was inconsistent or unavailable.
949 static struct gendisk
* dm_table_get_integrity_disk(struct dm_table
*t
,
952 struct list_head
*devices
= dm_table_get_devices(t
);
953 struct dm_dev_internal
*dd
= NULL
;
954 struct gendisk
*prev_disk
= NULL
, *template_disk
= NULL
;
956 list_for_each_entry(dd
, devices
, list
) {
957 template_disk
= dd
->dm_dev
.bdev
->bd_disk
;
958 if (!blk_get_integrity(template_disk
))
960 if (!match_all
&& !blk_integrity_is_initialized(template_disk
))
961 continue; /* skip uninitialized profiles */
962 else if (prev_disk
&&
963 blk_integrity_compare(prev_disk
, template_disk
) < 0)
965 prev_disk
= template_disk
;
968 return template_disk
;
972 DMWARN("%s: integrity not set: %s and %s profile mismatch",
973 dm_device_name(t
->md
),
974 prev_disk
->disk_name
,
975 template_disk
->disk_name
);
980 * Register the mapped device for blk_integrity support if
981 * the underlying devices have an integrity profile. But all devices
982 * may not have matching profiles (checking all devices isn't reliable
983 * during table load because this table may use other DM device(s) which
984 * must be resumed before they will have an initialized integity profile).
985 * Stacked DM devices force a 2 stage integrity profile validation:
986 * 1 - during load, validate all initialized integrity profiles match
987 * 2 - during resume, validate all integrity profiles match
989 static int dm_table_prealloc_integrity(struct dm_table
*t
, struct mapped_device
*md
)
991 struct gendisk
*template_disk
= NULL
;
993 template_disk
= dm_table_get_integrity_disk(t
, false);
997 if (!blk_integrity_is_initialized(dm_disk(md
))) {
998 t
->integrity_supported
= 1;
999 return blk_integrity_register(dm_disk(md
), NULL
);
1003 * If DM device already has an initalized integrity
1004 * profile the new profile should not conflict.
1006 if (blk_integrity_is_initialized(template_disk
) &&
1007 blk_integrity_compare(dm_disk(md
), template_disk
) < 0) {
1008 DMWARN("%s: conflict with existing integrity profile: "
1009 "%s profile mismatch",
1010 dm_device_name(t
->md
),
1011 template_disk
->disk_name
);
1015 /* Preserve existing initialized integrity profile */
1016 t
->integrity_supported
= 1;
1021 * Prepares the table for use by building the indices,
1022 * setting the type, and allocating mempools.
1024 int dm_table_complete(struct dm_table
*t
)
1028 r
= dm_table_set_type(t
);
1030 DMERR("unable to set table type");
1034 r
= dm_table_build_index(t
);
1036 DMERR("unable to build btrees");
1040 r
= dm_table_prealloc_integrity(t
, t
->md
);
1042 DMERR("could not register integrity profile.");
1046 r
= dm_table_alloc_md_mempools(t
);
1048 DMERR("unable to allocate mempools");
1053 static DEFINE_MUTEX(_event_lock
);
1054 void dm_table_event_callback(struct dm_table
*t
,
1055 void (*fn
)(void *), void *context
)
1057 mutex_lock(&_event_lock
);
1059 t
->event_context
= context
;
1060 mutex_unlock(&_event_lock
);
1063 void dm_table_event(struct dm_table
*t
)
1066 * You can no longer call dm_table_event() from interrupt
1067 * context, use a bottom half instead.
1069 BUG_ON(in_interrupt());
1071 mutex_lock(&_event_lock
);
1073 t
->event_fn(t
->event_context
);
1074 mutex_unlock(&_event_lock
);
1076 EXPORT_SYMBOL(dm_table_event
);
1078 sector_t
dm_table_get_size(struct dm_table
*t
)
1080 return t
->num_targets
? (t
->highs
[t
->num_targets
- 1] + 1) : 0;
1082 EXPORT_SYMBOL(dm_table_get_size
);
1084 struct dm_target
*dm_table_get_target(struct dm_table
*t
, unsigned int index
)
1086 if (index
>= t
->num_targets
)
1089 return t
->targets
+ index
;
1093 * Search the btree for the correct target.
1095 * Caller should check returned pointer with dm_target_is_valid()
1096 * to trap I/O beyond end of device.
1098 struct dm_target
*dm_table_find_target(struct dm_table
*t
, sector_t sector
)
1100 unsigned int l
, n
= 0, k
= 0;
1103 for (l
= 0; l
< t
->depth
; l
++) {
1104 n
= get_child(n
, k
);
1105 node
= get_node(t
, l
, n
);
1107 for (k
= 0; k
< KEYS_PER_NODE
; k
++)
1108 if (node
[k
] >= sector
)
1112 return &t
->targets
[(KEYS_PER_NODE
* n
) + k
];
1116 * Establish the new table's queue_limits and validate them.
1118 int dm_calculate_queue_limits(struct dm_table
*table
,
1119 struct queue_limits
*limits
)
1121 struct dm_target
*uninitialized_var(ti
);
1122 struct queue_limits ti_limits
;
1125 blk_set_default_limits(limits
);
1127 while (i
< dm_table_get_num_targets(table
)) {
1128 blk_set_default_limits(&ti_limits
);
1130 ti
= dm_table_get_target(table
, i
++);
1132 if (!ti
->type
->iterate_devices
)
1133 goto combine_limits
;
1136 * Combine queue limits of all the devices this target uses.
1138 ti
->type
->iterate_devices(ti
, dm_set_device_limits
,
1141 /* Set I/O hints portion of queue limits */
1142 if (ti
->type
->io_hints
)
1143 ti
->type
->io_hints(ti
, &ti_limits
);
1146 * Check each device area is consistent with the target's
1147 * overall queue limits.
1149 if (ti
->type
->iterate_devices(ti
, device_area_is_invalid
,
1155 * Merge this target's queue limits into the overall limits
1158 if (blk_stack_limits(limits
, &ti_limits
, 0) < 0)
1159 DMWARN("%s: adding target device "
1160 "(start sect %llu len %llu) "
1161 "caused an alignment inconsistency",
1162 dm_device_name(table
->md
),
1163 (unsigned long long) ti
->begin
,
1164 (unsigned long long) ti
->len
);
1167 return validate_hardware_logical_block_alignment(table
, limits
);
1171 * Set the integrity profile for this device if all devices used have
1172 * matching profiles. We're quite deep in the resume path but still
1173 * don't know if all devices (particularly DM devices this device
1174 * may be stacked on) have matching profiles. Even if the profiles
1175 * don't match we have no way to fail (to resume) at this point.
1177 static void dm_table_set_integrity(struct dm_table
*t
)
1179 struct gendisk
*template_disk
= NULL
;
1181 if (!blk_get_integrity(dm_disk(t
->md
)))
1184 template_disk
= dm_table_get_integrity_disk(t
, true);
1185 if (!template_disk
&&
1186 blk_integrity_is_initialized(dm_disk(t
->md
))) {
1187 DMWARN("%s: device no longer has a valid integrity profile",
1188 dm_device_name(t
->md
));
1191 blk_integrity_register(dm_disk(t
->md
),
1192 blk_get_integrity(template_disk
));
1195 void dm_table_set_restrictions(struct dm_table
*t
, struct request_queue
*q
,
1196 struct queue_limits
*limits
)
1199 * Copy table's limits to the DM device's request_queue
1201 q
->limits
= *limits
;
1203 if (!dm_table_supports_discards(t
))
1204 queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD
, q
);
1206 queue_flag_set_unlocked(QUEUE_FLAG_DISCARD
, q
);
1208 dm_table_set_integrity(t
);
1211 * QUEUE_FLAG_STACKABLE must be set after all queue settings are
1212 * visible to other CPUs because, once the flag is set, incoming bios
1213 * are processed by request-based dm, which refers to the queue
1215 * Until the flag set, bios are passed to bio-based dm and queued to
1216 * md->deferred where queue settings are not needed yet.
1217 * Those bios are passed to request-based dm at the resume time.
1220 if (dm_table_request_based(t
))
1221 queue_flag_set_unlocked(QUEUE_FLAG_STACKABLE
, q
);
1224 unsigned int dm_table_get_num_targets(struct dm_table
*t
)
1226 return t
->num_targets
;
1229 struct list_head
*dm_table_get_devices(struct dm_table
*t
)
1234 fmode_t
dm_table_get_mode(struct dm_table
*t
)
1238 EXPORT_SYMBOL(dm_table_get_mode
);
1240 static void suspend_targets(struct dm_table
*t
, unsigned postsuspend
)
1242 int i
= t
->num_targets
;
1243 struct dm_target
*ti
= t
->targets
;
1247 if (ti
->type
->postsuspend
)
1248 ti
->type
->postsuspend(ti
);
1249 } else if (ti
->type
->presuspend
)
1250 ti
->type
->presuspend(ti
);
1256 void dm_table_presuspend_targets(struct dm_table
*t
)
1261 suspend_targets(t
, 0);
1264 void dm_table_postsuspend_targets(struct dm_table
*t
)
1269 suspend_targets(t
, 1);
1272 int dm_table_resume_targets(struct dm_table
*t
)
1276 for (i
= 0; i
< t
->num_targets
; i
++) {
1277 struct dm_target
*ti
= t
->targets
+ i
;
1279 if (!ti
->type
->preresume
)
1282 r
= ti
->type
->preresume(ti
);
1287 for (i
= 0; i
< t
->num_targets
; i
++) {
1288 struct dm_target
*ti
= t
->targets
+ i
;
1290 if (ti
->type
->resume
)
1291 ti
->type
->resume(ti
);
1297 void dm_table_add_target_callbacks(struct dm_table
*t
, struct dm_target_callbacks
*cb
)
1299 list_add(&cb
->list
, &t
->target_callbacks
);
1301 EXPORT_SYMBOL_GPL(dm_table_add_target_callbacks
);
1303 int dm_table_any_congested(struct dm_table
*t
, int bdi_bits
)
1305 struct dm_dev_internal
*dd
;
1306 struct list_head
*devices
= dm_table_get_devices(t
);
1307 struct dm_target_callbacks
*cb
;
1310 list_for_each_entry(dd
, devices
, list
) {
1311 struct request_queue
*q
= bdev_get_queue(dd
->dm_dev
.bdev
);
1312 char b
[BDEVNAME_SIZE
];
1315 r
|= bdi_congested(&q
->backing_dev_info
, bdi_bits
);
1317 DMWARN_LIMIT("%s: any_congested: nonexistent device %s",
1318 dm_device_name(t
->md
),
1319 bdevname(dd
->dm_dev
.bdev
, b
));
1322 list_for_each_entry(cb
, &t
->target_callbacks
, list
)
1323 if (cb
->congested_fn
)
1324 r
|= cb
->congested_fn(cb
, bdi_bits
);
1329 int dm_table_any_busy_target(struct dm_table
*t
)
1332 struct dm_target
*ti
;
1334 for (i
= 0; i
< t
->num_targets
; i
++) {
1335 ti
= t
->targets
+ i
;
1336 if (ti
->type
->busy
&& ti
->type
->busy(ti
))
1343 struct mapped_device
*dm_table_get_md(struct dm_table
*t
)
1347 EXPORT_SYMBOL(dm_table_get_md
);
1349 static int device_discard_capable(struct dm_target
*ti
, struct dm_dev
*dev
,
1350 sector_t start
, sector_t len
, void *data
)
1352 struct request_queue
*q
= bdev_get_queue(dev
->bdev
);
1354 return q
&& blk_queue_discard(q
);
1357 bool dm_table_supports_discards(struct dm_table
*t
)
1359 struct dm_target
*ti
;
1363 * Unless any target used by the table set discards_supported,
1364 * require at least one underlying device to support discards.
1365 * t->devices includes internal dm devices such as mirror logs
1366 * so we need to use iterate_devices here, which targets
1367 * supporting discard selectively must provide.
1369 while (i
< dm_table_get_num_targets(t
)) {
1370 ti
= dm_table_get_target(t
, i
++);
1372 if (!ti
->num_discard_requests
)
1375 if (ti
->discards_supported
)
1378 if (ti
->type
->iterate_devices
&&
1379 ti
->type
->iterate_devices(ti
, device_discard_capable
, NULL
))