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
;
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
= vmalloc(size
);
158 memset(addr
, 0, size
);
164 * highs, and targets are managed as dynamic arrays during a
167 static int alloc_targets(struct dm_table
*t
, unsigned int num
)
170 struct dm_target
*n_targets
;
171 int n
= t
->num_targets
;
174 * Allocate both the target array and offset array at once.
175 * Append an empty entry to catch sectors beyond the end of
178 n_highs
= (sector_t
*) dm_vcalloc(num
+ 1, sizeof(struct dm_target
) +
183 n_targets
= (struct dm_target
*) (n_highs
+ num
);
186 memcpy(n_highs
, t
->highs
, sizeof(*n_highs
) * n
);
187 memcpy(n_targets
, t
->targets
, sizeof(*n_targets
) * n
);
190 memset(n_highs
+ n
, -1, sizeof(*n_highs
) * (num
- n
));
193 t
->num_allocated
= num
;
195 t
->targets
= n_targets
;
200 int dm_table_create(struct dm_table
**result
, fmode_t mode
,
201 unsigned num_targets
, struct mapped_device
*md
)
203 struct dm_table
*t
= kzalloc(sizeof(*t
), GFP_KERNEL
);
208 INIT_LIST_HEAD(&t
->devices
);
209 INIT_LIST_HEAD(&t
->target_callbacks
);
210 atomic_set(&t
->holders
, 0);
211 t
->discards_supported
= 1;
214 num_targets
= KEYS_PER_NODE
;
216 num_targets
= dm_round_up(num_targets
, KEYS_PER_NODE
);
218 if (alloc_targets(t
, num_targets
)) {
230 static void free_devices(struct list_head
*devices
)
232 struct list_head
*tmp
, *next
;
234 list_for_each_safe(tmp
, next
, devices
) {
235 struct dm_dev_internal
*dd
=
236 list_entry(tmp
, struct dm_dev_internal
, list
);
237 DMWARN("dm_table_destroy: dm_put_device call missing for %s",
243 void dm_table_destroy(struct dm_table
*t
)
250 while (atomic_read(&t
->holders
))
254 /* free the indexes */
256 vfree(t
->index
[t
->depth
- 2]);
258 /* free the targets */
259 for (i
= 0; i
< t
->num_targets
; i
++) {
260 struct dm_target
*tgt
= t
->targets
+ i
;
265 dm_put_target_type(tgt
->type
);
270 /* free the device list */
271 if (t
->devices
.next
!= &t
->devices
)
272 free_devices(&t
->devices
);
274 dm_free_md_mempools(t
->mempools
);
279 void dm_table_get(struct dm_table
*t
)
281 atomic_inc(&t
->holders
);
284 void dm_table_put(struct dm_table
*t
)
289 smp_mb__before_atomic_dec();
290 atomic_dec(&t
->holders
);
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 static int __table_get_device(struct dm_table
*t
, struct dm_target
*ti
,
458 const char *path
, fmode_t mode
, struct dm_dev
**result
)
461 dev_t
uninitialized_var(dev
);
462 struct dm_dev_internal
*dd
;
463 unsigned int major
, minor
;
467 if (sscanf(path
, "%u:%u", &major
, &minor
) == 2) {
468 /* Extract the major/minor numbers */
469 dev
= MKDEV(major
, minor
);
470 if (MAJOR(dev
) != major
|| MINOR(dev
) != minor
)
473 /* convert the path to a device */
474 struct block_device
*bdev
= lookup_bdev(path
);
477 return PTR_ERR(bdev
);
482 dd
= find_device(&t
->devices
, dev
);
484 dd
= kmalloc(sizeof(*dd
), GFP_KERNEL
);
488 dd
->dm_dev
.mode
= mode
;
489 dd
->dm_dev
.bdev
= NULL
;
491 if ((r
= open_dev(dd
, dev
, t
->md
))) {
496 format_dev_t(dd
->dm_dev
.name
, dev
);
498 atomic_set(&dd
->count
, 0);
499 list_add(&dd
->list
, &t
->devices
);
501 } else if (dd
->dm_dev
.mode
!= (mode
| dd
->dm_dev
.mode
)) {
502 r
= upgrade_mode(dd
, mode
, t
->md
);
506 atomic_inc(&dd
->count
);
508 *result
= &dd
->dm_dev
;
512 int dm_set_device_limits(struct dm_target
*ti
, struct dm_dev
*dev
,
513 sector_t start
, sector_t len
, void *data
)
515 struct queue_limits
*limits
= data
;
516 struct block_device
*bdev
= dev
->bdev
;
517 struct request_queue
*q
= bdev_get_queue(bdev
);
518 char b
[BDEVNAME_SIZE
];
521 DMWARN("%s: Cannot set limits for nonexistent device %s",
522 dm_device_name(ti
->table
->md
), bdevname(bdev
, b
));
526 if (bdev_stack_limits(limits
, bdev
, start
) < 0)
527 DMWARN("%s: adding target device %s caused an alignment inconsistency: "
528 "physical_block_size=%u, logical_block_size=%u, "
529 "alignment_offset=%u, start=%llu",
530 dm_device_name(ti
->table
->md
), bdevname(bdev
, b
),
531 q
->limits
.physical_block_size
,
532 q
->limits
.logical_block_size
,
533 q
->limits
.alignment_offset
,
534 (unsigned long long) start
<< SECTOR_SHIFT
);
537 * Check if merge fn is supported.
538 * If not we'll force DM to use PAGE_SIZE or
539 * smaller I/O, just to be safe.
542 if (q
->merge_bvec_fn
&& !ti
->type
->merge
)
543 blk_limits_max_hw_sectors(limits
,
544 (unsigned int) (PAGE_SIZE
>> 9));
547 EXPORT_SYMBOL_GPL(dm_set_device_limits
);
549 int dm_get_device(struct dm_target
*ti
, const char *path
, fmode_t mode
,
550 struct dm_dev
**result
)
552 return __table_get_device(ti
->table
, ti
, path
, mode
, result
);
557 * Decrement a devices use count and remove it if necessary.
559 void dm_put_device(struct dm_target
*ti
, struct dm_dev
*d
)
561 struct dm_dev_internal
*dd
= container_of(d
, struct dm_dev_internal
,
564 if (atomic_dec_and_test(&dd
->count
)) {
565 close_dev(dd
, ti
->table
->md
);
572 * Checks to see if the target joins onto the end of the table.
574 static int adjoin(struct dm_table
*table
, struct dm_target
*ti
)
576 struct dm_target
*prev
;
578 if (!table
->num_targets
)
581 prev
= &table
->targets
[table
->num_targets
- 1];
582 return (ti
->begin
== (prev
->begin
+ prev
->len
));
586 * Used to dynamically allocate the arg array.
588 static char **realloc_argv(unsigned *array_size
, char **old_argv
)
593 new_size
= *array_size
? *array_size
* 2 : 64;
594 argv
= kmalloc(new_size
* sizeof(*argv
), GFP_KERNEL
);
596 memcpy(argv
, old_argv
, *array_size
* sizeof(*argv
));
597 *array_size
= new_size
;
605 * Destructively splits up the argument list to pass to ctr.
607 int dm_split_args(int *argc
, char ***argvp
, char *input
)
609 char *start
, *end
= input
, *out
, **argv
= NULL
;
610 unsigned array_size
= 0;
619 argv
= realloc_argv(&array_size
, argv
);
624 /* Skip whitespace */
625 start
= skip_spaces(end
);
628 break; /* success, we hit the end */
630 /* 'out' is used to remove any back-quotes */
633 /* Everything apart from '\0' can be quoted */
634 if (*end
== '\\' && *(end
+ 1)) {
641 break; /* end of token */
646 /* have we already filled the array ? */
647 if ((*argc
+ 1) > array_size
) {
648 argv
= realloc_argv(&array_size
, argv
);
653 /* we know this is whitespace */
657 /* terminate the string and put it in the array */
668 * Impose necessary and sufficient conditions on a devices's table such
669 * that any incoming bio which respects its logical_block_size can be
670 * processed successfully. If it falls across the boundary between
671 * two or more targets, the size of each piece it gets split into must
672 * be compatible with the logical_block_size of the target processing it.
674 static int validate_hardware_logical_block_alignment(struct dm_table
*table
,
675 struct queue_limits
*limits
)
678 * This function uses arithmetic modulo the logical_block_size
679 * (in units of 512-byte sectors).
681 unsigned short device_logical_block_size_sects
=
682 limits
->logical_block_size
>> SECTOR_SHIFT
;
685 * Offset of the start of the next table entry, mod logical_block_size.
687 unsigned short next_target_start
= 0;
690 * Given an aligned bio that extends beyond the end of a
691 * target, how many sectors must the next target handle?
693 unsigned short remaining
= 0;
695 struct dm_target
*uninitialized_var(ti
);
696 struct queue_limits ti_limits
;
700 * Check each entry in the table in turn.
702 while (i
< dm_table_get_num_targets(table
)) {
703 ti
= dm_table_get_target(table
, i
++);
705 blk_set_default_limits(&ti_limits
);
707 /* combine all target devices' limits */
708 if (ti
->type
->iterate_devices
)
709 ti
->type
->iterate_devices(ti
, dm_set_device_limits
,
713 * If the remaining sectors fall entirely within this
714 * table entry are they compatible with its logical_block_size?
716 if (remaining
< ti
->len
&&
717 remaining
& ((ti_limits
.logical_block_size
>>
722 (unsigned short) ((next_target_start
+ ti
->len
) &
723 (device_logical_block_size_sects
- 1));
724 remaining
= next_target_start
?
725 device_logical_block_size_sects
- next_target_start
: 0;
729 DMWARN("%s: table line %u (start sect %llu len %llu) "
730 "not aligned to h/w logical block size %u",
731 dm_device_name(table
->md
), i
,
732 (unsigned long long) ti
->begin
,
733 (unsigned long long) ti
->len
,
734 limits
->logical_block_size
);
741 int dm_table_add_target(struct dm_table
*t
, const char *type
,
742 sector_t start
, sector_t len
, char *params
)
744 int r
= -EINVAL
, argc
;
746 struct dm_target
*tgt
;
748 if ((r
= check_space(t
)))
751 tgt
= t
->targets
+ t
->num_targets
;
752 memset(tgt
, 0, sizeof(*tgt
));
755 DMERR("%s: zero-length target", dm_device_name(t
->md
));
759 tgt
->type
= dm_get_target_type(type
);
761 DMERR("%s: %s: unknown target type", dm_device_name(t
->md
),
769 tgt
->error
= "Unknown error";
772 * Does this target adjoin the previous one ?
774 if (!adjoin(t
, tgt
)) {
775 tgt
->error
= "Gap in table";
780 r
= dm_split_args(&argc
, &argv
, params
);
782 tgt
->error
= "couldn't split parameters (insufficient memory)";
786 r
= tgt
->type
->ctr(tgt
, argc
, argv
);
791 t
->highs
[t
->num_targets
++] = tgt
->begin
+ tgt
->len
- 1;
793 if (!tgt
->num_discard_requests
)
794 t
->discards_supported
= 0;
799 DMERR("%s: %s: %s", dm_device_name(t
->md
), type
, tgt
->error
);
800 dm_put_target_type(tgt
->type
);
804 static int dm_table_set_type(struct dm_table
*t
)
807 unsigned bio_based
= 0, request_based
= 0;
808 struct dm_target
*tgt
;
809 struct dm_dev_internal
*dd
;
810 struct list_head
*devices
;
812 for (i
= 0; i
< t
->num_targets
; i
++) {
813 tgt
= t
->targets
+ i
;
814 if (dm_target_request_based(tgt
))
819 if (bio_based
&& request_based
) {
820 DMWARN("Inconsistent table: different target types"
821 " can't be mixed up");
827 /* We must use this table as bio-based */
828 t
->type
= DM_TYPE_BIO_BASED
;
832 BUG_ON(!request_based
); /* No targets in this table */
834 /* Non-request-stackable devices can't be used for request-based dm */
835 devices
= dm_table_get_devices(t
);
836 list_for_each_entry(dd
, devices
, list
) {
837 if (!blk_queue_stackable(bdev_get_queue(dd
->dm_dev
.bdev
))) {
838 DMWARN("table load rejected: including"
839 " non-request-stackable devices");
845 * Request-based dm supports only tables that have a single target now.
846 * To support multiple targets, request splitting support is needed,
847 * and that needs lots of changes in the block-layer.
848 * (e.g. request completion process for partial completion.)
850 if (t
->num_targets
> 1) {
851 DMWARN("Request-based dm doesn't support multiple targets yet");
855 t
->type
= DM_TYPE_REQUEST_BASED
;
860 unsigned dm_table_get_type(struct dm_table
*t
)
865 bool dm_table_request_based(struct dm_table
*t
)
867 return dm_table_get_type(t
) == DM_TYPE_REQUEST_BASED
;
870 int dm_table_alloc_md_mempools(struct dm_table
*t
)
872 unsigned type
= dm_table_get_type(t
);
874 if (unlikely(type
== DM_TYPE_NONE
)) {
875 DMWARN("no table type is set, can't allocate mempools");
879 t
->mempools
= dm_alloc_md_mempools(type
);
886 void dm_table_free_md_mempools(struct dm_table
*t
)
888 dm_free_md_mempools(t
->mempools
);
892 struct dm_md_mempools
*dm_table_get_md_mempools(struct dm_table
*t
)
897 static int setup_indexes(struct dm_table
*t
)
900 unsigned int total
= 0;
903 /* allocate the space for *all* the indexes */
904 for (i
= t
->depth
- 2; i
>= 0; i
--) {
905 t
->counts
[i
] = dm_div_up(t
->counts
[i
+ 1], CHILDREN_PER_NODE
);
906 total
+= t
->counts
[i
];
909 indexes
= (sector_t
*) dm_vcalloc(total
, (unsigned long) NODE_SIZE
);
913 /* set up internal nodes, bottom-up */
914 for (i
= t
->depth
- 2; i
>= 0; i
--) {
915 t
->index
[i
] = indexes
;
916 indexes
+= (KEYS_PER_NODE
* t
->counts
[i
]);
917 setup_btree_index(i
, t
);
924 * Builds the btree to index the map.
926 static int dm_table_build_index(struct dm_table
*t
)
929 unsigned int leaf_nodes
;
931 /* how many indexes will the btree have ? */
932 leaf_nodes
= dm_div_up(t
->num_targets
, KEYS_PER_NODE
);
933 t
->depth
= 1 + int_log(leaf_nodes
, CHILDREN_PER_NODE
);
935 /* leaf layer has already been set up */
936 t
->counts
[t
->depth
- 1] = leaf_nodes
;
937 t
->index
[t
->depth
- 1] = t
->highs
;
940 r
= setup_indexes(t
);
946 * Register the mapped device for blk_integrity support if
947 * the underlying devices support it.
949 static int dm_table_prealloc_integrity(struct dm_table
*t
, struct mapped_device
*md
)
951 struct list_head
*devices
= dm_table_get_devices(t
);
952 struct dm_dev_internal
*dd
;
954 list_for_each_entry(dd
, devices
, list
)
955 if (bdev_get_integrity(dd
->dm_dev
.bdev
))
956 return blk_integrity_register(dm_disk(md
), NULL
);
962 * Prepares the table for use by building the indices,
963 * setting the type, and allocating mempools.
965 int dm_table_complete(struct dm_table
*t
)
969 r
= dm_table_set_type(t
);
971 DMERR("unable to set table type");
975 r
= dm_table_build_index(t
);
977 DMERR("unable to build btrees");
981 r
= dm_table_prealloc_integrity(t
, t
->md
);
983 DMERR("could not register integrity profile.");
987 r
= dm_table_alloc_md_mempools(t
);
989 DMERR("unable to allocate mempools");
994 static DEFINE_MUTEX(_event_lock
);
995 void dm_table_event_callback(struct dm_table
*t
,
996 void (*fn
)(void *), void *context
)
998 mutex_lock(&_event_lock
);
1000 t
->event_context
= context
;
1001 mutex_unlock(&_event_lock
);
1004 void dm_table_event(struct dm_table
*t
)
1007 * You can no longer call dm_table_event() from interrupt
1008 * context, use a bottom half instead.
1010 BUG_ON(in_interrupt());
1012 mutex_lock(&_event_lock
);
1014 t
->event_fn(t
->event_context
);
1015 mutex_unlock(&_event_lock
);
1018 sector_t
dm_table_get_size(struct dm_table
*t
)
1020 return t
->num_targets
? (t
->highs
[t
->num_targets
- 1] + 1) : 0;
1023 struct dm_target
*dm_table_get_target(struct dm_table
*t
, unsigned int index
)
1025 if (index
>= t
->num_targets
)
1028 return t
->targets
+ index
;
1032 * Search the btree for the correct target.
1034 * Caller should check returned pointer with dm_target_is_valid()
1035 * to trap I/O beyond end of device.
1037 struct dm_target
*dm_table_find_target(struct dm_table
*t
, sector_t sector
)
1039 unsigned int l
, n
= 0, k
= 0;
1042 for (l
= 0; l
< t
->depth
; l
++) {
1043 n
= get_child(n
, k
);
1044 node
= get_node(t
, l
, n
);
1046 for (k
= 0; k
< KEYS_PER_NODE
; k
++)
1047 if (node
[k
] >= sector
)
1051 return &t
->targets
[(KEYS_PER_NODE
* n
) + k
];
1055 * Establish the new table's queue_limits and validate them.
1057 int dm_calculate_queue_limits(struct dm_table
*table
,
1058 struct queue_limits
*limits
)
1060 struct dm_target
*uninitialized_var(ti
);
1061 struct queue_limits ti_limits
;
1064 blk_set_default_limits(limits
);
1066 while (i
< dm_table_get_num_targets(table
)) {
1067 blk_set_default_limits(&ti_limits
);
1069 ti
= dm_table_get_target(table
, i
++);
1071 if (!ti
->type
->iterate_devices
)
1072 goto combine_limits
;
1075 * Combine queue limits of all the devices this target uses.
1077 ti
->type
->iterate_devices(ti
, dm_set_device_limits
,
1080 /* Set I/O hints portion of queue limits */
1081 if (ti
->type
->io_hints
)
1082 ti
->type
->io_hints(ti
, &ti_limits
);
1085 * Check each device area is consistent with the target's
1086 * overall queue limits.
1088 if (ti
->type
->iterate_devices(ti
, device_area_is_invalid
,
1094 * Merge this target's queue limits into the overall limits
1097 if (blk_stack_limits(limits
, &ti_limits
, 0) < 0)
1098 DMWARN("%s: adding target device "
1099 "(start sect %llu len %llu) "
1100 "caused an alignment inconsistency",
1101 dm_device_name(table
->md
),
1102 (unsigned long long) ti
->begin
,
1103 (unsigned long long) ti
->len
);
1106 return validate_hardware_logical_block_alignment(table
, limits
);
1110 * Set the integrity profile for this device if all devices used have
1111 * matching profiles.
1113 static void dm_table_set_integrity(struct dm_table
*t
)
1115 struct list_head
*devices
= dm_table_get_devices(t
);
1116 struct dm_dev_internal
*prev
= NULL
, *dd
= NULL
;
1118 if (!blk_get_integrity(dm_disk(t
->md
)))
1121 list_for_each_entry(dd
, devices
, list
) {
1123 blk_integrity_compare(prev
->dm_dev
.bdev
->bd_disk
,
1124 dd
->dm_dev
.bdev
->bd_disk
) < 0) {
1125 DMWARN("%s: integrity not set: %s and %s mismatch",
1126 dm_device_name(t
->md
),
1127 prev
->dm_dev
.bdev
->bd_disk
->disk_name
,
1128 dd
->dm_dev
.bdev
->bd_disk
->disk_name
);
1134 if (!prev
|| !bdev_get_integrity(prev
->dm_dev
.bdev
))
1137 blk_integrity_register(dm_disk(t
->md
),
1138 bdev_get_integrity(prev
->dm_dev
.bdev
));
1143 blk_integrity_register(dm_disk(t
->md
), NULL
);
1148 void dm_table_set_restrictions(struct dm_table
*t
, struct request_queue
*q
,
1149 struct queue_limits
*limits
)
1152 * Copy table's limits to the DM device's request_queue
1154 q
->limits
= *limits
;
1156 if (!dm_table_supports_discards(t
))
1157 queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD
, q
);
1159 queue_flag_set_unlocked(QUEUE_FLAG_DISCARD
, q
);
1161 dm_table_set_integrity(t
);
1164 * QUEUE_FLAG_STACKABLE must be set after all queue settings are
1165 * visible to other CPUs because, once the flag is set, incoming bios
1166 * are processed by request-based dm, which refers to the queue
1168 * Until the flag set, bios are passed to bio-based dm and queued to
1169 * md->deferred where queue settings are not needed yet.
1170 * Those bios are passed to request-based dm at the resume time.
1173 if (dm_table_request_based(t
))
1174 queue_flag_set_unlocked(QUEUE_FLAG_STACKABLE
, q
);
1177 unsigned int dm_table_get_num_targets(struct dm_table
*t
)
1179 return t
->num_targets
;
1182 struct list_head
*dm_table_get_devices(struct dm_table
*t
)
1187 fmode_t
dm_table_get_mode(struct dm_table
*t
)
1192 static void suspend_targets(struct dm_table
*t
, unsigned postsuspend
)
1194 int i
= t
->num_targets
;
1195 struct dm_target
*ti
= t
->targets
;
1199 if (ti
->type
->postsuspend
)
1200 ti
->type
->postsuspend(ti
);
1201 } else if (ti
->type
->presuspend
)
1202 ti
->type
->presuspend(ti
);
1208 void dm_table_presuspend_targets(struct dm_table
*t
)
1213 suspend_targets(t
, 0);
1216 void dm_table_postsuspend_targets(struct dm_table
*t
)
1221 suspend_targets(t
, 1);
1224 int dm_table_resume_targets(struct dm_table
*t
)
1228 for (i
= 0; i
< t
->num_targets
; i
++) {
1229 struct dm_target
*ti
= t
->targets
+ i
;
1231 if (!ti
->type
->preresume
)
1234 r
= ti
->type
->preresume(ti
);
1239 for (i
= 0; i
< t
->num_targets
; i
++) {
1240 struct dm_target
*ti
= t
->targets
+ i
;
1242 if (ti
->type
->resume
)
1243 ti
->type
->resume(ti
);
1249 void dm_table_add_target_callbacks(struct dm_table
*t
, struct dm_target_callbacks
*cb
)
1251 list_add(&cb
->list
, &t
->target_callbacks
);
1253 EXPORT_SYMBOL_GPL(dm_table_add_target_callbacks
);
1255 int dm_table_any_congested(struct dm_table
*t
, int bdi_bits
)
1257 struct dm_dev_internal
*dd
;
1258 struct list_head
*devices
= dm_table_get_devices(t
);
1259 struct dm_target_callbacks
*cb
;
1262 list_for_each_entry(dd
, devices
, list
) {
1263 struct request_queue
*q
= bdev_get_queue(dd
->dm_dev
.bdev
);
1264 char b
[BDEVNAME_SIZE
];
1267 r
|= bdi_congested(&q
->backing_dev_info
, bdi_bits
);
1269 DMWARN_LIMIT("%s: any_congested: nonexistent device %s",
1270 dm_device_name(t
->md
),
1271 bdevname(dd
->dm_dev
.bdev
, b
));
1274 list_for_each_entry(cb
, &t
->target_callbacks
, list
)
1275 if (cb
->congested_fn
)
1276 r
|= cb
->congested_fn(cb
, bdi_bits
);
1281 int dm_table_any_busy_target(struct dm_table
*t
)
1284 struct dm_target
*ti
;
1286 for (i
= 0; i
< t
->num_targets
; i
++) {
1287 ti
= t
->targets
+ i
;
1288 if (ti
->type
->busy
&& ti
->type
->busy(ti
))
1295 void dm_table_unplug_all(struct dm_table
*t
)
1297 struct dm_dev_internal
*dd
;
1298 struct list_head
*devices
= dm_table_get_devices(t
);
1299 struct dm_target_callbacks
*cb
;
1301 list_for_each_entry(dd
, devices
, list
) {
1302 struct request_queue
*q
= bdev_get_queue(dd
->dm_dev
.bdev
);
1303 char b
[BDEVNAME_SIZE
];
1308 DMWARN_LIMIT("%s: Cannot unplug nonexistent device %s",
1309 dm_device_name(t
->md
),
1310 bdevname(dd
->dm_dev
.bdev
, b
));
1313 list_for_each_entry(cb
, &t
->target_callbacks
, list
)
1318 struct mapped_device
*dm_table_get_md(struct dm_table
*t
)
1323 static int device_discard_capable(struct dm_target
*ti
, struct dm_dev
*dev
,
1324 sector_t start
, sector_t len
, void *data
)
1326 struct request_queue
*q
= bdev_get_queue(dev
->bdev
);
1328 return q
&& blk_queue_discard(q
);
1331 bool dm_table_supports_discards(struct dm_table
*t
)
1333 struct dm_target
*ti
;
1336 if (!t
->discards_supported
)
1340 * Ensure that at least one underlying device supports discards.
1341 * t->devices includes internal dm devices such as mirror logs
1342 * so we need to use iterate_devices here, which targets
1343 * supporting discard must provide.
1345 while (i
< dm_table_get_num_targets(t
)) {
1346 ti
= dm_table_get_target(t
, i
++);
1348 if (ti
->type
->iterate_devices
&&
1349 ti
->type
->iterate_devices(ti
, device_discard_capable
, NULL
))
1356 EXPORT_SYMBOL(dm_vcalloc
);
1357 EXPORT_SYMBOL(dm_get_device
);
1358 EXPORT_SYMBOL(dm_put_device
);
1359 EXPORT_SYMBOL(dm_table_event
);
1360 EXPORT_SYMBOL(dm_table_get_size
);
1361 EXPORT_SYMBOL(dm_table_get_mode
);
1362 EXPORT_SYMBOL(dm_table_get_md
);
1363 EXPORT_SYMBOL(dm_table_put
);
1364 EXPORT_SYMBOL(dm_table_get
);
1365 EXPORT_SYMBOL(dm_table_unplug_all
);