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 blkdev_put(d
->dm_dev
.bdev
, d
->dm_dev
.mode
| FMODE_EXCL
);
354 d
->dm_dev
.bdev
= NULL
;
358 * If possible, this checks an area of a destination device is invalid.
360 static int device_area_is_invalid(struct dm_target
*ti
, struct dm_dev
*dev
,
361 sector_t start
, sector_t len
, void *data
)
363 struct queue_limits
*limits
= data
;
364 struct block_device
*bdev
= dev
->bdev
;
366 i_size_read(bdev
->bd_inode
) >> SECTOR_SHIFT
;
367 unsigned short logical_block_size_sectors
=
368 limits
->logical_block_size
>> SECTOR_SHIFT
;
369 char b
[BDEVNAME_SIZE
];
374 if ((start
>= dev_size
) || (start
+ len
> dev_size
)) {
375 DMWARN("%s: %s too small for target: "
376 "start=%llu, len=%llu, dev_size=%llu",
377 dm_device_name(ti
->table
->md
), bdevname(bdev
, b
),
378 (unsigned long long)start
,
379 (unsigned long long)len
,
380 (unsigned long long)dev_size
);
384 if (logical_block_size_sectors
<= 1)
387 if (start
& (logical_block_size_sectors
- 1)) {
388 DMWARN("%s: start=%llu not aligned to h/w "
389 "logical block size %u of %s",
390 dm_device_name(ti
->table
->md
),
391 (unsigned long long)start
,
392 limits
->logical_block_size
, bdevname(bdev
, b
));
396 if (len
& (logical_block_size_sectors
- 1)) {
397 DMWARN("%s: len=%llu not aligned to h/w "
398 "logical block size %u of %s",
399 dm_device_name(ti
->table
->md
),
400 (unsigned long long)len
,
401 limits
->logical_block_size
, bdevname(bdev
, b
));
409 * This upgrades the mode on an already open dm_dev, being
410 * careful to leave things as they were if we fail to reopen the
411 * device and not to touch the existing bdev field in case
412 * it is accessed concurrently inside dm_table_any_congested().
414 static int upgrade_mode(struct dm_dev_internal
*dd
, fmode_t new_mode
,
415 struct mapped_device
*md
)
418 struct dm_dev_internal dd_new
, dd_old
;
420 dd_new
= dd_old
= *dd
;
422 dd_new
.dm_dev
.mode
|= new_mode
;
423 dd_new
.dm_dev
.bdev
= NULL
;
425 r
= open_dev(&dd_new
, dd
->dm_dev
.bdev
->bd_dev
, md
);
429 dd
->dm_dev
.mode
|= new_mode
;
430 close_dev(&dd_old
, md
);
436 * Add a device to the list, or just increment the usage count if
437 * it's already present.
439 static int __table_get_device(struct dm_table
*t
, struct dm_target
*ti
,
440 const char *path
, fmode_t mode
, struct dm_dev
**result
)
443 dev_t
uninitialized_var(dev
);
444 struct dm_dev_internal
*dd
;
445 unsigned int major
, minor
;
449 if (sscanf(path
, "%u:%u", &major
, &minor
) == 2) {
450 /* Extract the major/minor numbers */
451 dev
= MKDEV(major
, minor
);
452 if (MAJOR(dev
) != major
|| MINOR(dev
) != minor
)
455 /* convert the path to a device */
456 struct block_device
*bdev
= lookup_bdev(path
);
459 return PTR_ERR(bdev
);
464 dd
= find_device(&t
->devices
, dev
);
466 dd
= kmalloc(sizeof(*dd
), GFP_KERNEL
);
470 dd
->dm_dev
.mode
= mode
;
471 dd
->dm_dev
.bdev
= NULL
;
473 if ((r
= open_dev(dd
, dev
, t
->md
))) {
478 format_dev_t(dd
->dm_dev
.name
, dev
);
480 atomic_set(&dd
->count
, 0);
481 list_add(&dd
->list
, &t
->devices
);
483 } else if (dd
->dm_dev
.mode
!= (mode
| dd
->dm_dev
.mode
)) {
484 r
= upgrade_mode(dd
, mode
, t
->md
);
488 atomic_inc(&dd
->count
);
490 *result
= &dd
->dm_dev
;
494 int dm_set_device_limits(struct dm_target
*ti
, struct dm_dev
*dev
,
495 sector_t start
, sector_t len
, void *data
)
497 struct queue_limits
*limits
= data
;
498 struct block_device
*bdev
= dev
->bdev
;
499 struct request_queue
*q
= bdev_get_queue(bdev
);
500 char b
[BDEVNAME_SIZE
];
503 DMWARN("%s: Cannot set limits for nonexistent device %s",
504 dm_device_name(ti
->table
->md
), bdevname(bdev
, b
));
508 if (bdev_stack_limits(limits
, bdev
, start
) < 0)
509 DMWARN("%s: adding target device %s caused an alignment inconsistency: "
510 "physical_block_size=%u, logical_block_size=%u, "
511 "alignment_offset=%u, start=%llu",
512 dm_device_name(ti
->table
->md
), bdevname(bdev
, b
),
513 q
->limits
.physical_block_size
,
514 q
->limits
.logical_block_size
,
515 q
->limits
.alignment_offset
,
516 (unsigned long long) start
<< SECTOR_SHIFT
);
519 * Check if merge fn is supported.
520 * If not we'll force DM to use PAGE_SIZE or
521 * smaller I/O, just to be safe.
524 if (q
->merge_bvec_fn
&& !ti
->type
->merge
)
525 blk_limits_max_hw_sectors(limits
,
526 (unsigned int) (PAGE_SIZE
>> 9));
529 EXPORT_SYMBOL_GPL(dm_set_device_limits
);
531 int dm_get_device(struct dm_target
*ti
, const char *path
, fmode_t mode
,
532 struct dm_dev
**result
)
534 return __table_get_device(ti
->table
, ti
, path
, mode
, result
);
539 * Decrement a devices use count and remove it if necessary.
541 void dm_put_device(struct dm_target
*ti
, struct dm_dev
*d
)
543 struct dm_dev_internal
*dd
= container_of(d
, struct dm_dev_internal
,
546 if (atomic_dec_and_test(&dd
->count
)) {
547 close_dev(dd
, ti
->table
->md
);
554 * Checks to see if the target joins onto the end of the table.
556 static int adjoin(struct dm_table
*table
, struct dm_target
*ti
)
558 struct dm_target
*prev
;
560 if (!table
->num_targets
)
563 prev
= &table
->targets
[table
->num_targets
- 1];
564 return (ti
->begin
== (prev
->begin
+ prev
->len
));
568 * Used to dynamically allocate the arg array.
570 static char **realloc_argv(unsigned *array_size
, char **old_argv
)
575 new_size
= *array_size
? *array_size
* 2 : 64;
576 argv
= kmalloc(new_size
* sizeof(*argv
), GFP_KERNEL
);
578 memcpy(argv
, old_argv
, *array_size
* sizeof(*argv
));
579 *array_size
= new_size
;
587 * Destructively splits up the argument list to pass to ctr.
589 int dm_split_args(int *argc
, char ***argvp
, char *input
)
591 char *start
, *end
= input
, *out
, **argv
= NULL
;
592 unsigned array_size
= 0;
601 argv
= realloc_argv(&array_size
, argv
);
606 /* Skip whitespace */
607 start
= skip_spaces(end
);
610 break; /* success, we hit the end */
612 /* 'out' is used to remove any back-quotes */
615 /* Everything apart from '\0' can be quoted */
616 if (*end
== '\\' && *(end
+ 1)) {
623 break; /* end of token */
628 /* have we already filled the array ? */
629 if ((*argc
+ 1) > array_size
) {
630 argv
= realloc_argv(&array_size
, argv
);
635 /* we know this is whitespace */
639 /* terminate the string and put it in the array */
650 * Impose necessary and sufficient conditions on a devices's table such
651 * that any incoming bio which respects its logical_block_size can be
652 * processed successfully. If it falls across the boundary between
653 * two or more targets, the size of each piece it gets split into must
654 * be compatible with the logical_block_size of the target processing it.
656 static int validate_hardware_logical_block_alignment(struct dm_table
*table
,
657 struct queue_limits
*limits
)
660 * This function uses arithmetic modulo the logical_block_size
661 * (in units of 512-byte sectors).
663 unsigned short device_logical_block_size_sects
=
664 limits
->logical_block_size
>> SECTOR_SHIFT
;
667 * Offset of the start of the next table entry, mod logical_block_size.
669 unsigned short next_target_start
= 0;
672 * Given an aligned bio that extends beyond the end of a
673 * target, how many sectors must the next target handle?
675 unsigned short remaining
= 0;
677 struct dm_target
*uninitialized_var(ti
);
678 struct queue_limits ti_limits
;
682 * Check each entry in the table in turn.
684 while (i
< dm_table_get_num_targets(table
)) {
685 ti
= dm_table_get_target(table
, i
++);
687 blk_set_default_limits(&ti_limits
);
689 /* combine all target devices' limits */
690 if (ti
->type
->iterate_devices
)
691 ti
->type
->iterate_devices(ti
, dm_set_device_limits
,
695 * If the remaining sectors fall entirely within this
696 * table entry are they compatible with its logical_block_size?
698 if (remaining
< ti
->len
&&
699 remaining
& ((ti_limits
.logical_block_size
>>
704 (unsigned short) ((next_target_start
+ ti
->len
) &
705 (device_logical_block_size_sects
- 1));
706 remaining
= next_target_start
?
707 device_logical_block_size_sects
- next_target_start
: 0;
711 DMWARN("%s: table line %u (start sect %llu len %llu) "
712 "not aligned to h/w logical block size %u",
713 dm_device_name(table
->md
), i
,
714 (unsigned long long) ti
->begin
,
715 (unsigned long long) ti
->len
,
716 limits
->logical_block_size
);
723 int dm_table_add_target(struct dm_table
*t
, const char *type
,
724 sector_t start
, sector_t len
, char *params
)
726 int r
= -EINVAL
, argc
;
728 struct dm_target
*tgt
;
730 if ((r
= check_space(t
)))
733 tgt
= t
->targets
+ t
->num_targets
;
734 memset(tgt
, 0, sizeof(*tgt
));
737 DMERR("%s: zero-length target", dm_device_name(t
->md
));
741 tgt
->type
= dm_get_target_type(type
);
743 DMERR("%s: %s: unknown target type", dm_device_name(t
->md
),
751 tgt
->error
= "Unknown error";
754 * Does this target adjoin the previous one ?
756 if (!adjoin(t
, tgt
)) {
757 tgt
->error
= "Gap in table";
762 r
= dm_split_args(&argc
, &argv
, params
);
764 tgt
->error
= "couldn't split parameters (insufficient memory)";
768 r
= tgt
->type
->ctr(tgt
, argc
, argv
);
773 t
->highs
[t
->num_targets
++] = tgt
->begin
+ tgt
->len
- 1;
775 if (!tgt
->num_discard_requests
)
776 t
->discards_supported
= 0;
781 DMERR("%s: %s: %s", dm_device_name(t
->md
), type
, tgt
->error
);
782 dm_put_target_type(tgt
->type
);
786 static int dm_table_set_type(struct dm_table
*t
)
789 unsigned bio_based
= 0, request_based
= 0;
790 struct dm_target
*tgt
;
791 struct dm_dev_internal
*dd
;
792 struct list_head
*devices
;
794 for (i
= 0; i
< t
->num_targets
; i
++) {
795 tgt
= t
->targets
+ i
;
796 if (dm_target_request_based(tgt
))
801 if (bio_based
&& request_based
) {
802 DMWARN("Inconsistent table: different target types"
803 " can't be mixed up");
809 /* We must use this table as bio-based */
810 t
->type
= DM_TYPE_BIO_BASED
;
814 BUG_ON(!request_based
); /* No targets in this table */
816 /* Non-request-stackable devices can't be used for request-based dm */
817 devices
= dm_table_get_devices(t
);
818 list_for_each_entry(dd
, devices
, list
) {
819 if (!blk_queue_stackable(bdev_get_queue(dd
->dm_dev
.bdev
))) {
820 DMWARN("table load rejected: including"
821 " non-request-stackable devices");
827 * Request-based dm supports only tables that have a single target now.
828 * To support multiple targets, request splitting support is needed,
829 * and that needs lots of changes in the block-layer.
830 * (e.g. request completion process for partial completion.)
832 if (t
->num_targets
> 1) {
833 DMWARN("Request-based dm doesn't support multiple targets yet");
837 t
->type
= DM_TYPE_REQUEST_BASED
;
842 unsigned dm_table_get_type(struct dm_table
*t
)
847 bool dm_table_request_based(struct dm_table
*t
)
849 return dm_table_get_type(t
) == DM_TYPE_REQUEST_BASED
;
852 int dm_table_alloc_md_mempools(struct dm_table
*t
)
854 unsigned type
= dm_table_get_type(t
);
856 if (unlikely(type
== DM_TYPE_NONE
)) {
857 DMWARN("no table type is set, can't allocate mempools");
861 t
->mempools
= dm_alloc_md_mempools(type
);
868 void dm_table_free_md_mempools(struct dm_table
*t
)
870 dm_free_md_mempools(t
->mempools
);
874 struct dm_md_mempools
*dm_table_get_md_mempools(struct dm_table
*t
)
879 static int setup_indexes(struct dm_table
*t
)
882 unsigned int total
= 0;
885 /* allocate the space for *all* the indexes */
886 for (i
= t
->depth
- 2; i
>= 0; i
--) {
887 t
->counts
[i
] = dm_div_up(t
->counts
[i
+ 1], CHILDREN_PER_NODE
);
888 total
+= t
->counts
[i
];
891 indexes
= (sector_t
*) dm_vcalloc(total
, (unsigned long) NODE_SIZE
);
895 /* set up internal nodes, bottom-up */
896 for (i
= t
->depth
- 2; i
>= 0; i
--) {
897 t
->index
[i
] = indexes
;
898 indexes
+= (KEYS_PER_NODE
* t
->counts
[i
]);
899 setup_btree_index(i
, t
);
906 * Builds the btree to index the map.
908 static int dm_table_build_index(struct dm_table
*t
)
911 unsigned int leaf_nodes
;
913 /* how many indexes will the btree have ? */
914 leaf_nodes
= dm_div_up(t
->num_targets
, KEYS_PER_NODE
);
915 t
->depth
= 1 + int_log(leaf_nodes
, CHILDREN_PER_NODE
);
917 /* leaf layer has already been set up */
918 t
->counts
[t
->depth
- 1] = leaf_nodes
;
919 t
->index
[t
->depth
- 1] = t
->highs
;
922 r
= setup_indexes(t
);
928 * Register the mapped device for blk_integrity support if
929 * the underlying devices support it.
931 static int dm_table_prealloc_integrity(struct dm_table
*t
, struct mapped_device
*md
)
933 struct list_head
*devices
= dm_table_get_devices(t
);
934 struct dm_dev_internal
*dd
;
936 list_for_each_entry(dd
, devices
, list
)
937 if (bdev_get_integrity(dd
->dm_dev
.bdev
))
938 return blk_integrity_register(dm_disk(md
), NULL
);
944 * Prepares the table for use by building the indices,
945 * setting the type, and allocating mempools.
947 int dm_table_complete(struct dm_table
*t
)
951 r
= dm_table_set_type(t
);
953 DMERR("unable to set table type");
957 r
= dm_table_build_index(t
);
959 DMERR("unable to build btrees");
963 r
= dm_table_prealloc_integrity(t
, t
->md
);
965 DMERR("could not register integrity profile.");
969 r
= dm_table_alloc_md_mempools(t
);
971 DMERR("unable to allocate mempools");
976 static DEFINE_MUTEX(_event_lock
);
977 void dm_table_event_callback(struct dm_table
*t
,
978 void (*fn
)(void *), void *context
)
980 mutex_lock(&_event_lock
);
982 t
->event_context
= context
;
983 mutex_unlock(&_event_lock
);
986 void dm_table_event(struct dm_table
*t
)
989 * You can no longer call dm_table_event() from interrupt
990 * context, use a bottom half instead.
992 BUG_ON(in_interrupt());
994 mutex_lock(&_event_lock
);
996 t
->event_fn(t
->event_context
);
997 mutex_unlock(&_event_lock
);
1000 sector_t
dm_table_get_size(struct dm_table
*t
)
1002 return t
->num_targets
? (t
->highs
[t
->num_targets
- 1] + 1) : 0;
1005 struct dm_target
*dm_table_get_target(struct dm_table
*t
, unsigned int index
)
1007 if (index
>= t
->num_targets
)
1010 return t
->targets
+ index
;
1014 * Search the btree for the correct target.
1016 * Caller should check returned pointer with dm_target_is_valid()
1017 * to trap I/O beyond end of device.
1019 struct dm_target
*dm_table_find_target(struct dm_table
*t
, sector_t sector
)
1021 unsigned int l
, n
= 0, k
= 0;
1024 for (l
= 0; l
< t
->depth
; l
++) {
1025 n
= get_child(n
, k
);
1026 node
= get_node(t
, l
, n
);
1028 for (k
= 0; k
< KEYS_PER_NODE
; k
++)
1029 if (node
[k
] >= sector
)
1033 return &t
->targets
[(KEYS_PER_NODE
* n
) + k
];
1037 * Establish the new table's queue_limits and validate them.
1039 int dm_calculate_queue_limits(struct dm_table
*table
,
1040 struct queue_limits
*limits
)
1042 struct dm_target
*uninitialized_var(ti
);
1043 struct queue_limits ti_limits
;
1046 blk_set_default_limits(limits
);
1048 while (i
< dm_table_get_num_targets(table
)) {
1049 blk_set_default_limits(&ti_limits
);
1051 ti
= dm_table_get_target(table
, i
++);
1053 if (!ti
->type
->iterate_devices
)
1054 goto combine_limits
;
1057 * Combine queue limits of all the devices this target uses.
1059 ti
->type
->iterate_devices(ti
, dm_set_device_limits
,
1062 /* Set I/O hints portion of queue limits */
1063 if (ti
->type
->io_hints
)
1064 ti
->type
->io_hints(ti
, &ti_limits
);
1067 * Check each device area is consistent with the target's
1068 * overall queue limits.
1070 if (ti
->type
->iterate_devices(ti
, device_area_is_invalid
,
1076 * Merge this target's queue limits into the overall limits
1079 if (blk_stack_limits(limits
, &ti_limits
, 0) < 0)
1080 DMWARN("%s: adding target device "
1081 "(start sect %llu len %llu) "
1082 "caused an alignment inconsistency",
1083 dm_device_name(table
->md
),
1084 (unsigned long long) ti
->begin
,
1085 (unsigned long long) ti
->len
);
1088 return validate_hardware_logical_block_alignment(table
, limits
);
1092 * Set the integrity profile for this device if all devices used have
1093 * matching profiles.
1095 static void dm_table_set_integrity(struct dm_table
*t
)
1097 struct list_head
*devices
= dm_table_get_devices(t
);
1098 struct dm_dev_internal
*prev
= NULL
, *dd
= NULL
;
1100 if (!blk_get_integrity(dm_disk(t
->md
)))
1103 list_for_each_entry(dd
, devices
, list
) {
1105 blk_integrity_compare(prev
->dm_dev
.bdev
->bd_disk
,
1106 dd
->dm_dev
.bdev
->bd_disk
) < 0) {
1107 DMWARN("%s: integrity not set: %s and %s mismatch",
1108 dm_device_name(t
->md
),
1109 prev
->dm_dev
.bdev
->bd_disk
->disk_name
,
1110 dd
->dm_dev
.bdev
->bd_disk
->disk_name
);
1116 if (!prev
|| !bdev_get_integrity(prev
->dm_dev
.bdev
))
1119 blk_integrity_register(dm_disk(t
->md
),
1120 bdev_get_integrity(prev
->dm_dev
.bdev
));
1125 blk_integrity_register(dm_disk(t
->md
), NULL
);
1130 void dm_table_set_restrictions(struct dm_table
*t
, struct request_queue
*q
,
1131 struct queue_limits
*limits
)
1134 * Copy table's limits to the DM device's request_queue
1136 q
->limits
= *limits
;
1138 if (!dm_table_supports_discards(t
))
1139 queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD
, q
);
1141 queue_flag_set_unlocked(QUEUE_FLAG_DISCARD
, q
);
1143 dm_table_set_integrity(t
);
1146 * QUEUE_FLAG_STACKABLE must be set after all queue settings are
1147 * visible to other CPUs because, once the flag is set, incoming bios
1148 * are processed by request-based dm, which refers to the queue
1150 * Until the flag set, bios are passed to bio-based dm and queued to
1151 * md->deferred where queue settings are not needed yet.
1152 * Those bios are passed to request-based dm at the resume time.
1155 if (dm_table_request_based(t
))
1156 queue_flag_set_unlocked(QUEUE_FLAG_STACKABLE
, q
);
1159 unsigned int dm_table_get_num_targets(struct dm_table
*t
)
1161 return t
->num_targets
;
1164 struct list_head
*dm_table_get_devices(struct dm_table
*t
)
1169 fmode_t
dm_table_get_mode(struct dm_table
*t
)
1174 static void suspend_targets(struct dm_table
*t
, unsigned postsuspend
)
1176 int i
= t
->num_targets
;
1177 struct dm_target
*ti
= t
->targets
;
1181 if (ti
->type
->postsuspend
)
1182 ti
->type
->postsuspend(ti
);
1183 } else if (ti
->type
->presuspend
)
1184 ti
->type
->presuspend(ti
);
1190 void dm_table_presuspend_targets(struct dm_table
*t
)
1195 suspend_targets(t
, 0);
1198 void dm_table_postsuspend_targets(struct dm_table
*t
)
1203 suspend_targets(t
, 1);
1206 int dm_table_resume_targets(struct dm_table
*t
)
1210 for (i
= 0; i
< t
->num_targets
; i
++) {
1211 struct dm_target
*ti
= t
->targets
+ i
;
1213 if (!ti
->type
->preresume
)
1216 r
= ti
->type
->preresume(ti
);
1221 for (i
= 0; i
< t
->num_targets
; i
++) {
1222 struct dm_target
*ti
= t
->targets
+ i
;
1224 if (ti
->type
->resume
)
1225 ti
->type
->resume(ti
);
1231 void dm_table_add_target_callbacks(struct dm_table
*t
, struct dm_target_callbacks
*cb
)
1233 list_add(&cb
->list
, &t
->target_callbacks
);
1235 EXPORT_SYMBOL_GPL(dm_table_add_target_callbacks
);
1237 int dm_table_any_congested(struct dm_table
*t
, int bdi_bits
)
1239 struct dm_dev_internal
*dd
;
1240 struct list_head
*devices
= dm_table_get_devices(t
);
1241 struct dm_target_callbacks
*cb
;
1244 list_for_each_entry(dd
, devices
, list
) {
1245 struct request_queue
*q
= bdev_get_queue(dd
->dm_dev
.bdev
);
1246 char b
[BDEVNAME_SIZE
];
1249 r
|= bdi_congested(&q
->backing_dev_info
, bdi_bits
);
1251 DMWARN_LIMIT("%s: any_congested: nonexistent device %s",
1252 dm_device_name(t
->md
),
1253 bdevname(dd
->dm_dev
.bdev
, b
));
1256 list_for_each_entry(cb
, &t
->target_callbacks
, list
)
1257 if (cb
->congested_fn
)
1258 r
|= cb
->congested_fn(cb
, bdi_bits
);
1263 int dm_table_any_busy_target(struct dm_table
*t
)
1266 struct dm_target
*ti
;
1268 for (i
= 0; i
< t
->num_targets
; i
++) {
1269 ti
= t
->targets
+ i
;
1270 if (ti
->type
->busy
&& ti
->type
->busy(ti
))
1277 void dm_table_unplug_all(struct dm_table
*t
)
1279 struct dm_dev_internal
*dd
;
1280 struct list_head
*devices
= dm_table_get_devices(t
);
1281 struct dm_target_callbacks
*cb
;
1283 list_for_each_entry(dd
, devices
, list
) {
1284 struct request_queue
*q
= bdev_get_queue(dd
->dm_dev
.bdev
);
1285 char b
[BDEVNAME_SIZE
];
1290 DMWARN_LIMIT("%s: Cannot unplug nonexistent device %s",
1291 dm_device_name(t
->md
),
1292 bdevname(dd
->dm_dev
.bdev
, b
));
1295 list_for_each_entry(cb
, &t
->target_callbacks
, list
)
1300 struct mapped_device
*dm_table_get_md(struct dm_table
*t
)
1305 static int device_discard_capable(struct dm_target
*ti
, struct dm_dev
*dev
,
1306 sector_t start
, sector_t len
, void *data
)
1308 struct request_queue
*q
= bdev_get_queue(dev
->bdev
);
1310 return q
&& blk_queue_discard(q
);
1313 bool dm_table_supports_discards(struct dm_table
*t
)
1315 struct dm_target
*ti
;
1318 if (!t
->discards_supported
)
1322 * Ensure that at least one underlying device supports discards.
1323 * t->devices includes internal dm devices such as mirror logs
1324 * so we need to use iterate_devices here, which targets
1325 * supporting discard must provide.
1327 while (i
< dm_table_get_num_targets(t
)) {
1328 ti
= dm_table_get_target(t
, i
++);
1330 if (ti
->type
->iterate_devices
&&
1331 ti
->type
->iterate_devices(ti
, device_discard_capable
, NULL
))
1338 EXPORT_SYMBOL(dm_vcalloc
);
1339 EXPORT_SYMBOL(dm_get_device
);
1340 EXPORT_SYMBOL(dm_put_device
);
1341 EXPORT_SYMBOL(dm_table_event
);
1342 EXPORT_SYMBOL(dm_table_get_size
);
1343 EXPORT_SYMBOL(dm_table_get_mode
);
1344 EXPORT_SYMBOL(dm_table_get_md
);
1345 EXPORT_SYMBOL(dm_table_put
);
1346 EXPORT_SYMBOL(dm_table_get
);
1347 EXPORT_SYMBOL(dm_table_unplug_all
);