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 LGPL.
8 #ifndef _LINUX_DEVICE_MAPPER_H
9 #define _LINUX_DEVICE_MAPPER_H
11 #include <linux/bio.h>
12 #include <linux/blkdev.h>
13 #include <linux/math64.h>
14 #include <linux/ratelimit.h>
23 * Type of table, mapped_device's mempool and request_queue
27 DM_TYPE_BIO_BASED
= 1,
28 DM_TYPE_REQUEST_BASED
= 2,
29 DM_TYPE_MQ_REQUEST_BASED
= 3,
30 DM_TYPE_DAX_BIO_BASED
= 4,
33 typedef enum { STATUSTYPE_INFO
, STATUSTYPE_TABLE
} status_type_t
;
40 * In the constructor the target parameter will already have the
41 * table, type, begin and len fields filled in.
43 typedef int (*dm_ctr_fn
) (struct dm_target
*target
,
44 unsigned int argc
, char **argv
);
47 * The destructor doesn't need to free the dm_target, just
48 * anything hidden ti->private.
50 typedef void (*dm_dtr_fn
) (struct dm_target
*ti
);
53 * The map function must return:
55 * = 0: The target will handle the io by resubmitting it later
56 * = 1: simple remap complete
57 * = 2: The target wants to push back the io
59 typedef int (*dm_map_fn
) (struct dm_target
*ti
, struct bio
*bio
);
60 typedef int (*dm_clone_and_map_request_fn
) (struct dm_target
*ti
,
62 union map_info
*map_context
,
63 struct request
**clone
);
64 typedef void (*dm_release_clone_request_fn
) (struct request
*clone
);
68 * < 0 : error (currently ignored)
69 * 0 : ended successfully
70 * 1 : for some reason the io has still not completed (eg,
71 * multipath target might want to requeue a failed io).
72 * 2 : The target wants to push back the io
74 typedef int (*dm_endio_fn
) (struct dm_target
*ti
,
75 struct bio
*bio
, blk_status_t
*error
);
76 typedef int (*dm_request_endio_fn
) (struct dm_target
*ti
,
77 struct request
*clone
, blk_status_t error
,
78 union map_info
*map_context
);
80 typedef void (*dm_presuspend_fn
) (struct dm_target
*ti
);
81 typedef void (*dm_presuspend_undo_fn
) (struct dm_target
*ti
);
82 typedef void (*dm_postsuspend_fn
) (struct dm_target
*ti
);
83 typedef int (*dm_preresume_fn
) (struct dm_target
*ti
);
84 typedef void (*dm_resume_fn
) (struct dm_target
*ti
);
86 typedef void (*dm_status_fn
) (struct dm_target
*ti
, status_type_t status_type
,
87 unsigned status_flags
, char *result
, unsigned maxlen
);
89 typedef int (*dm_message_fn
) (struct dm_target
*ti
, unsigned argc
, char **argv
);
91 typedef int (*dm_prepare_ioctl_fn
) (struct dm_target
*ti
,
92 struct block_device
**bdev
, fmode_t
*mode
);
95 * These iteration functions are typically used to check (and combine)
96 * properties of underlying devices.
97 * E.g. Does at least one underlying device support flush?
98 * Does any underlying device not support WRITE_SAME?
100 * The callout function is called once for each contiguous section of
101 * an underlying device. State can be maintained in *data.
102 * Return non-zero to stop iterating through any further devices.
104 typedef int (*iterate_devices_callout_fn
) (struct dm_target
*ti
,
106 sector_t start
, sector_t len
,
110 * This function must iterate through each section of device used by the
111 * target until it encounters a non-zero return code, which it then returns.
112 * Returns zero if no callout returned non-zero.
114 typedef int (*dm_iterate_devices_fn
) (struct dm_target
*ti
,
115 iterate_devices_callout_fn fn
,
118 typedef void (*dm_io_hints_fn
) (struct dm_target
*ti
,
119 struct queue_limits
*limits
);
123 * 0: The target can handle the next I/O immediately.
124 * 1: The target can't handle the next I/O immediately.
126 typedef int (*dm_busy_fn
) (struct dm_target
*ti
);
131 * >= 0 : the number of bytes accessible at the address
133 typedef long (*dm_dax_direct_access_fn
) (struct dm_target
*ti
, pgoff_t pgoff
,
134 long nr_pages
, void **kaddr
, pfn_t
*pfn
);
135 typedef size_t (*dm_dax_copy_from_iter_fn
)(struct dm_target
*ti
, pgoff_t pgoff
,
136 void *addr
, size_t bytes
, struct iov_iter
*i
);
137 typedef void (*dm_dax_flush_fn
)(struct dm_target
*ti
, pgoff_t pgoff
, void *addr
,
139 #define PAGE_SECTORS (PAGE_SIZE / 512)
141 void dm_error(const char *message
);
144 struct block_device
*bdev
;
145 struct dax_device
*dax_dev
;
150 dev_t
dm_get_dev_t(const char *path
);
153 * Constructors should call these functions to ensure destination devices
154 * are opened/closed correctly.
156 int dm_get_device(struct dm_target
*ti
, const char *path
, fmode_t mode
,
157 struct dm_dev
**result
);
158 void dm_put_device(struct dm_target
*ti
, struct dm_dev
*d
);
161 * Information about a target type
167 struct module
*module
;
172 dm_clone_and_map_request_fn clone_and_map_rq
;
173 dm_release_clone_request_fn release_clone_rq
;
175 dm_request_endio_fn rq_end_io
;
176 dm_presuspend_fn presuspend
;
177 dm_presuspend_undo_fn presuspend_undo
;
178 dm_postsuspend_fn postsuspend
;
179 dm_preresume_fn preresume
;
182 dm_message_fn message
;
183 dm_prepare_ioctl_fn prepare_ioctl
;
185 dm_iterate_devices_fn iterate_devices
;
186 dm_io_hints_fn io_hints
;
187 dm_dax_direct_access_fn direct_access
;
188 dm_dax_copy_from_iter_fn dax_copy_from_iter
;
189 dm_dax_flush_fn dax_flush
;
191 /* For internal device-mapper use. */
192 struct list_head list
;
200 * Any table that contains an instance of this target must have only one.
202 #define DM_TARGET_SINGLETON 0x00000001
203 #define dm_target_needs_singleton(type) ((type)->features & DM_TARGET_SINGLETON)
206 * Indicates that a target does not support read-only devices.
208 #define DM_TARGET_ALWAYS_WRITEABLE 0x00000002
209 #define dm_target_always_writeable(type) \
210 ((type)->features & DM_TARGET_ALWAYS_WRITEABLE)
213 * Any device that contains a table with an instance of this target may never
214 * have tables containing any different target type.
216 #define DM_TARGET_IMMUTABLE 0x00000004
217 #define dm_target_is_immutable(type) ((type)->features & DM_TARGET_IMMUTABLE)
220 * Indicates that a target may replace any target; even immutable targets.
221 * .map, .map_rq, .clone_and_map_rq and .release_clone_rq are all defined.
223 #define DM_TARGET_WILDCARD 0x00000008
224 #define dm_target_is_wildcard(type) ((type)->features & DM_TARGET_WILDCARD)
227 * Some targets need to be sent the same WRITE bio severals times so
228 * that they can send copies of it to different devices. This function
229 * examines any supplied bio and returns the number of copies of it the
232 typedef unsigned (*dm_num_write_bios_fn
) (struct dm_target
*ti
, struct bio
*bio
);
235 * A target implements own bio data integrity.
237 #define DM_TARGET_INTEGRITY 0x00000010
238 #define dm_target_has_integrity(type) ((type)->features & DM_TARGET_INTEGRITY)
241 * A target passes integrity data to the lower device.
243 #define DM_TARGET_PASSES_INTEGRITY 0x00000020
244 #define dm_target_passes_integrity(type) ((type)->features & DM_TARGET_PASSES_INTEGRITY)
247 * Indicates that a target supports host-managed zoned block devices.
249 #define DM_TARGET_ZONED_HM 0x00000040
250 #define dm_target_supports_zoned_hm(type) ((type)->features & DM_TARGET_ZONED_HM)
253 struct dm_table
*table
;
254 struct target_type
*type
;
260 /* If non-zero, maximum size of I/O submitted to a target. */
264 * A number of zero-length barrier bios that will be submitted
265 * to the target for the purpose of flushing cache.
267 * The bio number can be accessed with dm_bio_get_target_bio_nr.
268 * It is a responsibility of the target driver to remap these bios
269 * to the real underlying devices.
271 unsigned num_flush_bios
;
274 * The number of discard bios that will be submitted to the target.
275 * The bio number can be accessed with dm_bio_get_target_bio_nr.
277 unsigned num_discard_bios
;
280 * The number of WRITE SAME bios that will be submitted to the target.
281 * The bio number can be accessed with dm_bio_get_target_bio_nr.
283 unsigned num_write_same_bios
;
286 * The number of WRITE ZEROES bios that will be submitted to the target.
287 * The bio number can be accessed with dm_bio_get_target_bio_nr.
289 unsigned num_write_zeroes_bios
;
292 * The minimum number of extra bytes allocated in each io for the
295 unsigned per_io_data_size
;
298 * If defined, this function is called to find out how many
299 * duplicate bios should be sent to the target when writing
302 dm_num_write_bios_fn num_write_bios
;
304 /* target specific data */
307 /* Used to provide an error string from the ctr */
311 * Set if this target needs to receive flushes regardless of
312 * whether or not its underlying devices have support.
314 bool flush_supported
:1;
317 * Set if this target needs to receive discards regardless of
318 * whether or not its underlying devices have support.
320 bool discards_supported
:1;
323 * Set if the target required discard bios to be split
324 * on max_io_len boundary.
326 bool split_discard_bios
:1;
329 /* Each target can link one of these into the table */
330 struct dm_target_callbacks
{
331 struct list_head list
;
332 int (*congested_fn
) (struct dm_target_callbacks
*, int);
337 * One of these is allocated for each bio.
338 * This structure shouldn't be touched directly by target drivers.
339 * It is here so that we can inline dm_per_bio_data and
340 * dm_bio_from_per_bio_data
342 struct dm_target_io
{
344 struct dm_target
*ti
;
345 unsigned target_bio_nr
;
350 static inline void *dm_per_bio_data(struct bio
*bio
, size_t data_size
)
352 return (char *)bio
- offsetof(struct dm_target_io
, clone
) - data_size
;
355 static inline struct bio
*dm_bio_from_per_bio_data(void *data
, size_t data_size
)
357 return (struct bio
*)((char *)data
+ data_size
+ offsetof(struct dm_target_io
, clone
));
360 static inline unsigned dm_bio_get_target_bio_nr(const struct bio
*bio
)
362 return container_of(bio
, struct dm_target_io
, clone
)->target_bio_nr
;
365 int dm_register_target(struct target_type
*t
);
366 void dm_unregister_target(struct target_type
*t
);
369 * Target argument parsing.
377 * The minimum and maximum value of a numeric argument, together with
378 * the error message to use if the number is found to be outside that range.
387 * Validate the next argument, either returning it as *value or, if invalid,
388 * returning -EINVAL and setting *error.
390 int dm_read_arg(struct dm_arg
*arg
, struct dm_arg_set
*arg_set
,
391 unsigned *value
, char **error
);
394 * Process the next argument as the start of a group containing between
395 * arg->min and arg->max further arguments. Either return the size as
396 * *num_args or, if invalid, return -EINVAL and set *error.
398 int dm_read_arg_group(struct dm_arg
*arg
, struct dm_arg_set
*arg_set
,
399 unsigned *num_args
, char **error
);
402 * Return the current argument and shift to the next.
404 const char *dm_shift_arg(struct dm_arg_set
*as
);
407 * Move through num_args arguments.
409 void dm_consume_args(struct dm_arg_set
*as
, unsigned num_args
);
411 /*-----------------------------------------------------------------
412 * Functions for creating and manipulating mapped devices.
413 * Drop the reference with dm_put when you finish with the object.
414 *---------------------------------------------------------------*/
417 * DM_ANY_MINOR chooses the next available minor number.
419 #define DM_ANY_MINOR (-1)
420 int dm_create(int minor
, struct mapped_device
**md
);
423 * Reference counting for md.
425 struct mapped_device
*dm_get_md(dev_t dev
);
426 void dm_get(struct mapped_device
*md
);
427 int dm_hold(struct mapped_device
*md
);
428 void dm_put(struct mapped_device
*md
);
431 * An arbitrary pointer may be stored alongside a mapped device.
433 void dm_set_mdptr(struct mapped_device
*md
, void *ptr
);
434 void *dm_get_mdptr(struct mapped_device
*md
);
437 * A device can still be used while suspended, but I/O is deferred.
439 int dm_suspend(struct mapped_device
*md
, unsigned suspend_flags
);
440 int dm_resume(struct mapped_device
*md
);
445 uint32_t dm_get_event_nr(struct mapped_device
*md
);
446 int dm_wait_event(struct mapped_device
*md
, int event_nr
);
447 uint32_t dm_next_uevent_seq(struct mapped_device
*md
);
448 void dm_uevent_add(struct mapped_device
*md
, struct list_head
*elist
);
453 const char *dm_device_name(struct mapped_device
*md
);
454 int dm_copy_name_and_uuid(struct mapped_device
*md
, char *name
, char *uuid
);
455 struct gendisk
*dm_disk(struct mapped_device
*md
);
456 int dm_suspended(struct dm_target
*ti
);
457 int dm_noflush_suspending(struct dm_target
*ti
);
458 void dm_accept_partial_bio(struct bio
*bio
, unsigned n_sectors
);
459 void dm_remap_zone_report(struct dm_target
*ti
, struct bio
*bio
,
461 union map_info
*dm_get_rq_mapinfo(struct request
*rq
);
463 struct queue_limits
*dm_get_queue_limits(struct mapped_device
*md
);
466 * Geometry functions.
468 int dm_get_geometry(struct mapped_device
*md
, struct hd_geometry
*geo
);
469 int dm_set_geometry(struct mapped_device
*md
, struct hd_geometry
*geo
);
471 /*-----------------------------------------------------------------
472 * Functions for manipulating device-mapper tables.
473 *---------------------------------------------------------------*/
476 * First create an empty table.
478 int dm_table_create(struct dm_table
**result
, fmode_t mode
,
479 unsigned num_targets
, struct mapped_device
*md
);
482 * Then call this once for each target.
484 int dm_table_add_target(struct dm_table
*t
, const char *type
,
485 sector_t start
, sector_t len
, char *params
);
488 * Target_ctr should call this if it needs to add any callbacks.
490 void dm_table_add_target_callbacks(struct dm_table
*t
, struct dm_target_callbacks
*cb
);
493 * Target can use this to set the table's type.
494 * Can only ever be called from a target's ctr.
495 * Useful for "hybrid" target (supports both bio-based
496 * and request-based).
498 void dm_table_set_type(struct dm_table
*t
, enum dm_queue_mode type
);
501 * Finally call this to make the table ready for use.
503 int dm_table_complete(struct dm_table
*t
);
506 * Target may require that it is never sent I/O larger than len.
508 int __must_check
dm_set_target_max_io_len(struct dm_target
*ti
, sector_t len
);
511 * Table reference counting.
513 struct dm_table
*dm_get_live_table(struct mapped_device
*md
, int *srcu_idx
);
514 void dm_put_live_table(struct mapped_device
*md
, int srcu_idx
);
515 void dm_sync_table(struct mapped_device
*md
);
520 sector_t
dm_table_get_size(struct dm_table
*t
);
521 unsigned int dm_table_get_num_targets(struct dm_table
*t
);
522 fmode_t
dm_table_get_mode(struct dm_table
*t
);
523 struct mapped_device
*dm_table_get_md(struct dm_table
*t
);
528 void dm_table_event(struct dm_table
*t
);
531 * Run the queue for request-based targets.
533 void dm_table_run_md_queue_async(struct dm_table
*t
);
536 * The device must be suspended before calling this method.
537 * Returns the previous table, which the caller must destroy.
539 struct dm_table
*dm_swap_table(struct mapped_device
*md
,
543 * A wrapper around vmalloc.
545 void *dm_vcalloc(unsigned long nmemb
, unsigned long elem_size
);
547 /*-----------------------------------------------------------------
549 *---------------------------------------------------------------*/
550 #define DM_NAME "device-mapper"
553 extern struct ratelimit_state dm_ratelimit_state
;
555 #define dm_ratelimit() __ratelimit(&dm_ratelimit_state)
557 #define dm_ratelimit() 0
560 #define DM_FMT(fmt) DM_NAME ": " DM_MSG_PREFIX ": " fmt "\n"
562 #define DMCRIT(fmt, ...) pr_crit(DM_FMT(fmt), ##__VA_ARGS__)
564 #define DMERR(fmt, ...) pr_err(DM_FMT(fmt), ##__VA_ARGS__)
565 #define DMERR_LIMIT(fmt, ...) \
567 if (dm_ratelimit()) \
568 DMERR(fmt, ##__VA_ARGS__); \
571 #define DMWARN(fmt, ...) pr_warn(DM_FMT(fmt), ##__VA_ARGS__)
572 #define DMWARN_LIMIT(fmt, ...) \
574 if (dm_ratelimit()) \
575 DMWARN(fmt, ##__VA_ARGS__); \
578 #define DMINFO(fmt, ...) pr_info(DM_FMT(fmt), ##__VA_ARGS__)
579 #define DMINFO_LIMIT(fmt, ...) \
581 if (dm_ratelimit()) \
582 DMINFO(fmt, ##__VA_ARGS__); \
585 #ifdef CONFIG_DM_DEBUG
586 #define DMDEBUG(fmt, ...) printk(KERN_DEBUG DM_FMT(fmt), ##__VA_ARGS__)
587 #define DMDEBUG_LIMIT(fmt, ...) \
589 if (dm_ratelimit()) \
590 DMDEBUG(fmt, ##__VA_ARGS__); \
593 #define DMDEBUG(fmt, ...) no_printk(fmt, ##__VA_ARGS__)
594 #define DMDEBUG_LIMIT(fmt, ...) no_printk(fmt, ##__VA_ARGS__)
597 #define DMEMIT(x...) sz += ((sz >= maxlen) ? \
598 0 : scnprintf(result + sz, maxlen - sz, x))
600 #define SECTOR_SHIFT 9
603 * Definitions of return values from target end_io function.
605 #define DM_ENDIO_DONE 0
606 #define DM_ENDIO_INCOMPLETE 1
607 #define DM_ENDIO_REQUEUE 2
610 * Definitions of return values from target map function.
612 #define DM_MAPIO_SUBMITTED 0
613 #define DM_MAPIO_REMAPPED 1
614 #define DM_MAPIO_REQUEUE DM_ENDIO_REQUEUE
615 #define DM_MAPIO_DELAY_REQUEUE 3
616 #define DM_MAPIO_KILL 4
618 #define dm_sector_div64(x, y)( \
621 (x) = div64_u64_rem(x, y, &_res); \
629 #define dm_div_up(n, sz) (((n) + (sz) - 1) / (sz))
631 #define dm_sector_div_up(n, sz) ( \
633 sector_t _r = ((n) + (sz) - 1); \
634 sector_div(_r, (sz)); \
640 * ceiling(n / size) * size
642 #define dm_round_up(n, sz) (dm_div_up((n), (sz)) * (sz))
644 #define dm_array_too_big(fixed, obj, num) \
645 ((num) > (UINT_MAX - (fixed)) / (obj))
648 * Sector offset taken relative to the start of the target instead of
649 * relative to the start of the device.
651 #define dm_target_offset(ti, sector) ((sector) - (ti)->begin)
653 static inline sector_t
to_sector(unsigned long n
)
655 return (n
>> SECTOR_SHIFT
);
658 static inline unsigned long to_bytes(sector_t n
)
660 return (n
<< SECTOR_SHIFT
);
663 #endif /* _LINUX_DEVICE_MAPPER_H */