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 blk_limits_max_hw_sectors(limits
,
521 (unsigned int) (PAGE_SIZE
>> 9));
524 EXPORT_SYMBOL_GPL(dm_set_device_limits
);
526 int dm_get_device(struct dm_target
*ti
, const char *path
, fmode_t mode
,
527 struct dm_dev
**result
)
529 return __table_get_device(ti
->table
, ti
, path
, mode
, result
);
534 * Decrement a devices use count and remove it if necessary.
536 void dm_put_device(struct dm_target
*ti
, struct dm_dev
*d
)
538 struct dm_dev_internal
*dd
= container_of(d
, struct dm_dev_internal
,
541 if (atomic_dec_and_test(&dd
->count
)) {
542 close_dev(dd
, ti
->table
->md
);
549 * Checks to see if the target joins onto the end of the table.
551 static int adjoin(struct dm_table
*table
, struct dm_target
*ti
)
553 struct dm_target
*prev
;
555 if (!table
->num_targets
)
558 prev
= &table
->targets
[table
->num_targets
- 1];
559 return (ti
->begin
== (prev
->begin
+ prev
->len
));
563 * Used to dynamically allocate the arg array.
565 static char **realloc_argv(unsigned *array_size
, char **old_argv
)
570 new_size
= *array_size
? *array_size
* 2 : 64;
571 argv
= kmalloc(new_size
* sizeof(*argv
), GFP_KERNEL
);
573 memcpy(argv
, old_argv
, *array_size
* sizeof(*argv
));
574 *array_size
= new_size
;
582 * Destructively splits up the argument list to pass to ctr.
584 int dm_split_args(int *argc
, char ***argvp
, char *input
)
586 char *start
, *end
= input
, *out
, **argv
= NULL
;
587 unsigned array_size
= 0;
596 argv
= realloc_argv(&array_size
, argv
);
601 /* Skip whitespace */
602 start
= skip_spaces(end
);
605 break; /* success, we hit the end */
607 /* 'out' is used to remove any back-quotes */
610 /* Everything apart from '\0' can be quoted */
611 if (*end
== '\\' && *(end
+ 1)) {
618 break; /* end of token */
623 /* have we already filled the array ? */
624 if ((*argc
+ 1) > array_size
) {
625 argv
= realloc_argv(&array_size
, argv
);
630 /* we know this is whitespace */
634 /* terminate the string and put it in the array */
645 * Impose necessary and sufficient conditions on a devices's table such
646 * that any incoming bio which respects its logical_block_size can be
647 * processed successfully. If it falls across the boundary between
648 * two or more targets, the size of each piece it gets split into must
649 * be compatible with the logical_block_size of the target processing it.
651 static int validate_hardware_logical_block_alignment(struct dm_table
*table
,
652 struct queue_limits
*limits
)
655 * This function uses arithmetic modulo the logical_block_size
656 * (in units of 512-byte sectors).
658 unsigned short device_logical_block_size_sects
=
659 limits
->logical_block_size
>> SECTOR_SHIFT
;
662 * Offset of the start of the next table entry, mod logical_block_size.
664 unsigned short next_target_start
= 0;
667 * Given an aligned bio that extends beyond the end of a
668 * target, how many sectors must the next target handle?
670 unsigned short remaining
= 0;
672 struct dm_target
*uninitialized_var(ti
);
673 struct queue_limits ti_limits
;
677 * Check each entry in the table in turn.
679 while (i
< dm_table_get_num_targets(table
)) {
680 ti
= dm_table_get_target(table
, i
++);
682 blk_set_default_limits(&ti_limits
);
684 /* combine all target devices' limits */
685 if (ti
->type
->iterate_devices
)
686 ti
->type
->iterate_devices(ti
, dm_set_device_limits
,
690 * If the remaining sectors fall entirely within this
691 * table entry are they compatible with its logical_block_size?
693 if (remaining
< ti
->len
&&
694 remaining
& ((ti_limits
.logical_block_size
>>
699 (unsigned short) ((next_target_start
+ ti
->len
) &
700 (device_logical_block_size_sects
- 1));
701 remaining
= next_target_start
?
702 device_logical_block_size_sects
- next_target_start
: 0;
706 DMWARN("%s: table line %u (start sect %llu len %llu) "
707 "not aligned to h/w logical block size %u",
708 dm_device_name(table
->md
), i
,
709 (unsigned long long) ti
->begin
,
710 (unsigned long long) ti
->len
,
711 limits
->logical_block_size
);
718 int dm_table_add_target(struct dm_table
*t
, const char *type
,
719 sector_t start
, sector_t len
, char *params
)
721 int r
= -EINVAL
, argc
;
723 struct dm_target
*tgt
;
725 if ((r
= check_space(t
)))
728 tgt
= t
->targets
+ t
->num_targets
;
729 memset(tgt
, 0, sizeof(*tgt
));
732 DMERR("%s: zero-length target", dm_device_name(t
->md
));
736 tgt
->type
= dm_get_target_type(type
);
738 DMERR("%s: %s: unknown target type", dm_device_name(t
->md
),
746 tgt
->error
= "Unknown error";
749 * Does this target adjoin the previous one ?
751 if (!adjoin(t
, tgt
)) {
752 tgt
->error
= "Gap in table";
757 r
= dm_split_args(&argc
, &argv
, params
);
759 tgt
->error
= "couldn't split parameters (insufficient memory)";
763 r
= tgt
->type
->ctr(tgt
, argc
, argv
);
768 t
->highs
[t
->num_targets
++] = tgt
->begin
+ tgt
->len
- 1;
770 if (!tgt
->num_discard_requests
)
771 t
->discards_supported
= 0;
776 DMERR("%s: %s: %s", dm_device_name(t
->md
), type
, tgt
->error
);
777 dm_put_target_type(tgt
->type
);
781 static int dm_table_set_type(struct dm_table
*t
)
784 unsigned bio_based
= 0, request_based
= 0;
785 struct dm_target
*tgt
;
786 struct dm_dev_internal
*dd
;
787 struct list_head
*devices
;
789 for (i
= 0; i
< t
->num_targets
; i
++) {
790 tgt
= t
->targets
+ i
;
791 if (dm_target_request_based(tgt
))
796 if (bio_based
&& request_based
) {
797 DMWARN("Inconsistent table: different target types"
798 " can't be mixed up");
804 /* We must use this table as bio-based */
805 t
->type
= DM_TYPE_BIO_BASED
;
809 BUG_ON(!request_based
); /* No targets in this table */
811 /* Non-request-stackable devices can't be used for request-based dm */
812 devices
= dm_table_get_devices(t
);
813 list_for_each_entry(dd
, devices
, list
) {
814 if (!blk_queue_stackable(bdev_get_queue(dd
->dm_dev
.bdev
))) {
815 DMWARN("table load rejected: including"
816 " non-request-stackable devices");
822 * Request-based dm supports only tables that have a single target now.
823 * To support multiple targets, request splitting support is needed,
824 * and that needs lots of changes in the block-layer.
825 * (e.g. request completion process for partial completion.)
827 if (t
->num_targets
> 1) {
828 DMWARN("Request-based dm doesn't support multiple targets yet");
832 t
->type
= DM_TYPE_REQUEST_BASED
;
837 unsigned dm_table_get_type(struct dm_table
*t
)
842 bool dm_table_request_based(struct dm_table
*t
)
844 return dm_table_get_type(t
) == DM_TYPE_REQUEST_BASED
;
847 int dm_table_alloc_md_mempools(struct dm_table
*t
)
849 unsigned type
= dm_table_get_type(t
);
851 if (unlikely(type
== DM_TYPE_NONE
)) {
852 DMWARN("no table type is set, can't allocate mempools");
856 t
->mempools
= dm_alloc_md_mempools(type
);
863 void dm_table_free_md_mempools(struct dm_table
*t
)
865 dm_free_md_mempools(t
->mempools
);
869 struct dm_md_mempools
*dm_table_get_md_mempools(struct dm_table
*t
)
874 static int setup_indexes(struct dm_table
*t
)
877 unsigned int total
= 0;
880 /* allocate the space for *all* the indexes */
881 for (i
= t
->depth
- 2; i
>= 0; i
--) {
882 t
->counts
[i
] = dm_div_up(t
->counts
[i
+ 1], CHILDREN_PER_NODE
);
883 total
+= t
->counts
[i
];
886 indexes
= (sector_t
*) dm_vcalloc(total
, (unsigned long) NODE_SIZE
);
890 /* set up internal nodes, bottom-up */
891 for (i
= t
->depth
- 2; i
>= 0; i
--) {
892 t
->index
[i
] = indexes
;
893 indexes
+= (KEYS_PER_NODE
* t
->counts
[i
]);
894 setup_btree_index(i
, t
);
901 * Builds the btree to index the map.
903 static int dm_table_build_index(struct dm_table
*t
)
906 unsigned int leaf_nodes
;
908 /* how many indexes will the btree have ? */
909 leaf_nodes
= dm_div_up(t
->num_targets
, KEYS_PER_NODE
);
910 t
->depth
= 1 + int_log(leaf_nodes
, CHILDREN_PER_NODE
);
912 /* leaf layer has already been set up */
913 t
->counts
[t
->depth
- 1] = leaf_nodes
;
914 t
->index
[t
->depth
- 1] = t
->highs
;
917 r
= setup_indexes(t
);
923 * Register the mapped device for blk_integrity support if
924 * the underlying devices support it.
926 static int dm_table_prealloc_integrity(struct dm_table
*t
, struct mapped_device
*md
)
928 struct list_head
*devices
= dm_table_get_devices(t
);
929 struct dm_dev_internal
*dd
;
931 list_for_each_entry(dd
, devices
, list
)
932 if (bdev_get_integrity(dd
->dm_dev
.bdev
))
933 return blk_integrity_register(dm_disk(md
), NULL
);
939 * Prepares the table for use by building the indices,
940 * setting the type, and allocating mempools.
942 int dm_table_complete(struct dm_table
*t
)
946 r
= dm_table_set_type(t
);
948 DMERR("unable to set table type");
952 r
= dm_table_build_index(t
);
954 DMERR("unable to build btrees");
958 r
= dm_table_prealloc_integrity(t
, t
->md
);
960 DMERR("could not register integrity profile.");
964 r
= dm_table_alloc_md_mempools(t
);
966 DMERR("unable to allocate mempools");
971 static DEFINE_MUTEX(_event_lock
);
972 void dm_table_event_callback(struct dm_table
*t
,
973 void (*fn
)(void *), void *context
)
975 mutex_lock(&_event_lock
);
977 t
->event_context
= context
;
978 mutex_unlock(&_event_lock
);
981 void dm_table_event(struct dm_table
*t
)
984 * You can no longer call dm_table_event() from interrupt
985 * context, use a bottom half instead.
987 BUG_ON(in_interrupt());
989 mutex_lock(&_event_lock
);
991 t
->event_fn(t
->event_context
);
992 mutex_unlock(&_event_lock
);
995 sector_t
dm_table_get_size(struct dm_table
*t
)
997 return t
->num_targets
? (t
->highs
[t
->num_targets
- 1] + 1) : 0;
1000 struct dm_target
*dm_table_get_target(struct dm_table
*t
, unsigned int index
)
1002 if (index
>= t
->num_targets
)
1005 return t
->targets
+ index
;
1009 * Search the btree for the correct target.
1011 * Caller should check returned pointer with dm_target_is_valid()
1012 * to trap I/O beyond end of device.
1014 struct dm_target
*dm_table_find_target(struct dm_table
*t
, sector_t sector
)
1016 unsigned int l
, n
= 0, k
= 0;
1019 for (l
= 0; l
< t
->depth
; l
++) {
1020 n
= get_child(n
, k
);
1021 node
= get_node(t
, l
, n
);
1023 for (k
= 0; k
< KEYS_PER_NODE
; k
++)
1024 if (node
[k
] >= sector
)
1028 return &t
->targets
[(KEYS_PER_NODE
* n
) + k
];
1032 * Establish the new table's queue_limits and validate them.
1034 int dm_calculate_queue_limits(struct dm_table
*table
,
1035 struct queue_limits
*limits
)
1037 struct dm_target
*uninitialized_var(ti
);
1038 struct queue_limits ti_limits
;
1041 blk_set_default_limits(limits
);
1043 while (i
< dm_table_get_num_targets(table
)) {
1044 blk_set_default_limits(&ti_limits
);
1046 ti
= dm_table_get_target(table
, i
++);
1048 if (!ti
->type
->iterate_devices
)
1049 goto combine_limits
;
1052 * Combine queue limits of all the devices this target uses.
1054 ti
->type
->iterate_devices(ti
, dm_set_device_limits
,
1057 /* Set I/O hints portion of queue limits */
1058 if (ti
->type
->io_hints
)
1059 ti
->type
->io_hints(ti
, &ti_limits
);
1062 * Check each device area is consistent with the target's
1063 * overall queue limits.
1065 if (ti
->type
->iterate_devices(ti
, device_area_is_invalid
,
1071 * Merge this target's queue limits into the overall limits
1074 if (blk_stack_limits(limits
, &ti_limits
, 0) < 0)
1075 DMWARN("%s: adding target device "
1076 "(start sect %llu len %llu) "
1077 "caused an alignment inconsistency",
1078 dm_device_name(table
->md
),
1079 (unsigned long long) ti
->begin
,
1080 (unsigned long long) ti
->len
);
1083 return validate_hardware_logical_block_alignment(table
, limits
);
1087 * Set the integrity profile for this device if all devices used have
1088 * matching profiles.
1090 static void dm_table_set_integrity(struct dm_table
*t
)
1092 struct list_head
*devices
= dm_table_get_devices(t
);
1093 struct dm_dev_internal
*prev
= NULL
, *dd
= NULL
;
1095 if (!blk_get_integrity(dm_disk(t
->md
)))
1098 list_for_each_entry(dd
, devices
, list
) {
1100 blk_integrity_compare(prev
->dm_dev
.bdev
->bd_disk
,
1101 dd
->dm_dev
.bdev
->bd_disk
) < 0) {
1102 DMWARN("%s: integrity not set: %s and %s mismatch",
1103 dm_device_name(t
->md
),
1104 prev
->dm_dev
.bdev
->bd_disk
->disk_name
,
1105 dd
->dm_dev
.bdev
->bd_disk
->disk_name
);
1111 if (!prev
|| !bdev_get_integrity(prev
->dm_dev
.bdev
))
1114 blk_integrity_register(dm_disk(t
->md
),
1115 bdev_get_integrity(prev
->dm_dev
.bdev
));
1120 blk_integrity_register(dm_disk(t
->md
), NULL
);
1125 void dm_table_set_restrictions(struct dm_table
*t
, struct request_queue
*q
,
1126 struct queue_limits
*limits
)
1129 * Copy table's limits to the DM device's request_queue
1131 q
->limits
= *limits
;
1133 if (!dm_table_supports_discards(t
))
1134 queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD
, q
);
1136 queue_flag_set_unlocked(QUEUE_FLAG_DISCARD
, q
);
1138 dm_table_set_integrity(t
);
1141 * QUEUE_FLAG_STACKABLE must be set after all queue settings are
1142 * visible to other CPUs because, once the flag is set, incoming bios
1143 * are processed by request-based dm, which refers to the queue
1145 * Until the flag set, bios are passed to bio-based dm and queued to
1146 * md->deferred where queue settings are not needed yet.
1147 * Those bios are passed to request-based dm at the resume time.
1150 if (dm_table_request_based(t
))
1151 queue_flag_set_unlocked(QUEUE_FLAG_STACKABLE
, q
);
1154 unsigned int dm_table_get_num_targets(struct dm_table
*t
)
1156 return t
->num_targets
;
1159 struct list_head
*dm_table_get_devices(struct dm_table
*t
)
1164 fmode_t
dm_table_get_mode(struct dm_table
*t
)
1169 static void suspend_targets(struct dm_table
*t
, unsigned postsuspend
)
1171 int i
= t
->num_targets
;
1172 struct dm_target
*ti
= t
->targets
;
1176 if (ti
->type
->postsuspend
)
1177 ti
->type
->postsuspend(ti
);
1178 } else if (ti
->type
->presuspend
)
1179 ti
->type
->presuspend(ti
);
1185 void dm_table_presuspend_targets(struct dm_table
*t
)
1190 suspend_targets(t
, 0);
1193 void dm_table_postsuspend_targets(struct dm_table
*t
)
1198 suspend_targets(t
, 1);
1201 int dm_table_resume_targets(struct dm_table
*t
)
1205 for (i
= 0; i
< t
->num_targets
; i
++) {
1206 struct dm_target
*ti
= t
->targets
+ i
;
1208 if (!ti
->type
->preresume
)
1211 r
= ti
->type
->preresume(ti
);
1216 for (i
= 0; i
< t
->num_targets
; i
++) {
1217 struct dm_target
*ti
= t
->targets
+ i
;
1219 if (ti
->type
->resume
)
1220 ti
->type
->resume(ti
);
1226 int dm_table_any_congested(struct dm_table
*t
, int bdi_bits
)
1228 struct dm_dev_internal
*dd
;
1229 struct list_head
*devices
= dm_table_get_devices(t
);
1232 list_for_each_entry(dd
, devices
, list
) {
1233 struct request_queue
*q
= bdev_get_queue(dd
->dm_dev
.bdev
);
1234 char b
[BDEVNAME_SIZE
];
1237 r
|= bdi_congested(&q
->backing_dev_info
, bdi_bits
);
1239 DMWARN_LIMIT("%s: any_congested: nonexistent device %s",
1240 dm_device_name(t
->md
),
1241 bdevname(dd
->dm_dev
.bdev
, b
));
1247 int dm_table_any_busy_target(struct dm_table
*t
)
1250 struct dm_target
*ti
;
1252 for (i
= 0; i
< t
->num_targets
; i
++) {
1253 ti
= t
->targets
+ i
;
1254 if (ti
->type
->busy
&& ti
->type
->busy(ti
))
1261 void dm_table_unplug_all(struct dm_table
*t
)
1263 struct dm_dev_internal
*dd
;
1264 struct list_head
*devices
= dm_table_get_devices(t
);
1266 list_for_each_entry(dd
, devices
, list
) {
1267 struct request_queue
*q
= bdev_get_queue(dd
->dm_dev
.bdev
);
1268 char b
[BDEVNAME_SIZE
];
1273 DMWARN_LIMIT("%s: Cannot unplug nonexistent device %s",
1274 dm_device_name(t
->md
),
1275 bdevname(dd
->dm_dev
.bdev
, b
));
1279 struct mapped_device
*dm_table_get_md(struct dm_table
*t
)
1284 static int device_discard_capable(struct dm_target
*ti
, struct dm_dev
*dev
,
1285 sector_t start
, sector_t len
, void *data
)
1287 struct request_queue
*q
= bdev_get_queue(dev
->bdev
);
1289 return q
&& blk_queue_discard(q
);
1292 bool dm_table_supports_discards(struct dm_table
*t
)
1294 struct dm_target
*ti
;
1297 if (!t
->discards_supported
)
1301 * Ensure that at least one underlying device supports discards.
1302 * t->devices includes internal dm devices such as mirror logs
1303 * so we need to use iterate_devices here, which targets
1304 * supporting discard must provide.
1306 while (i
< dm_table_get_num_targets(t
)) {
1307 ti
= dm_table_get_target(t
, i
++);
1309 if (ti
->type
->iterate_devices
&&
1310 ti
->type
->iterate_devices(ti
, device_discard_capable
, NULL
))
1317 EXPORT_SYMBOL(dm_vcalloc
);
1318 EXPORT_SYMBOL(dm_get_device
);
1319 EXPORT_SYMBOL(dm_put_device
);
1320 EXPORT_SYMBOL(dm_table_event
);
1321 EXPORT_SYMBOL(dm_table_get_size
);
1322 EXPORT_SYMBOL(dm_table_get_mode
);
1323 EXPORT_SYMBOL(dm_table_get_md
);
1324 EXPORT_SYMBOL(dm_table_put
);
1325 EXPORT_SYMBOL(dm_table_get
);
1326 EXPORT_SYMBOL(dm_table_unplug_all
);