1 /*
2 * Copyright (C) 2001 Sistina Software (UK) Limited.
3 * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
4 *
5 * This file is released under the LGPL.
6 */
7
8 #ifndef _LINUX_DEVICE_MAPPER_H
9 #define _LINUX_DEVICE_MAPPER_H
10
11 #include <linux/bio.h>
12 #include <linux/blkdev.h>
13 #include <linux/dm-ioctl.h>
14 #include <linux/math64.h>
15 #include <linux/ratelimit.h>
16
17 struct dm_dev;
18 struct dm_target;
19 struct dm_table;
20 struct dm_report_zones_args;
21 struct mapped_device;
22 struct bio_vec;
23
24 /*
25 * Type of table, mapped_device's mempool and request_queue
26 */
27 enum dm_queue_mode {
28 DM_TYPE_NONE = 0,
29 DM_TYPE_BIO_BASED = 1,
30 DM_TYPE_REQUEST_BASED = 2,
31 DM_TYPE_DAX_BIO_BASED = 3,
32 };
33
34 typedef enum { STATUSTYPE_INFO, STATUSTYPE_TABLE } status_type_t;
35
36 union map_info {
37 void *ptr;
38 };
39
40 /*
41 * In the constructor the target parameter will already have the
42 * table, type, begin and len fields filled in.
43 */
44 typedef int (*dm_ctr_fn) (struct dm_target *target,
45 unsigned int argc, char **argv);
46
47 /*
48 * The destructor doesn't need to free the dm_target, just
49 * anything hidden ti->private.
50 */
51 typedef void (*dm_dtr_fn) (struct dm_target *ti);
52
53 /*
54 * The map function must return:
55 * < 0: error
56 * = 0: The target will handle the io by resubmitting it later
57 * = 1: simple remap complete
58 * = 2: The target wants to push back the io
59 */
60 typedef int (*dm_map_fn) (struct dm_target *ti, struct bio *bio);
61 typedef int (*dm_clone_and_map_request_fn) (struct dm_target *ti,
62 struct request *rq,
63 union map_info *map_context,
64 struct request **clone);
65 typedef void (*dm_release_clone_request_fn) (struct request *clone,
66 union map_info *map_context);
67
68 /*
69 * Returns:
70 * < 0 : error (currently ignored)
71 * 0 : ended successfully
72 * 1 : for some reason the io has still not completed (eg,
73 * multipath target might want to requeue a failed io).
74 * 2 : The target wants to push back the io
75 */
76 typedef int (*dm_endio_fn) (struct dm_target *ti,
77 struct bio *bio, blk_status_t *error);
78 typedef int (*dm_request_endio_fn) (struct dm_target *ti,
79 struct request *clone, blk_status_t error,
80 union map_info *map_context);
81
82 typedef void (*dm_presuspend_fn) (struct dm_target *ti);
83 typedef void (*dm_presuspend_undo_fn) (struct dm_target *ti);
84 typedef void (*dm_postsuspend_fn) (struct dm_target *ti);
85 typedef int (*dm_preresume_fn) (struct dm_target *ti);
86 typedef void (*dm_resume_fn) (struct dm_target *ti);
87
88 typedef void (*dm_status_fn) (struct dm_target *ti, status_type_t status_type,
89 unsigned status_flags, char *result, unsigned maxlen);
90
91 typedef int (*dm_message_fn) (struct dm_target *ti, unsigned argc, char **argv,
92 char *result, unsigned maxlen);
93
94 typedef int (*dm_prepare_ioctl_fn) (struct dm_target *ti, struct block_device **bdev);
95
96 typedef int (*dm_report_zones_fn) (struct dm_target *ti,
97 struct dm_report_zones_args *args,
98 unsigned int nr_zones);
99
100 /*
101 * These iteration functions are typically used to check (and combine)
102 * properties of underlying devices.
103 * E.g. Does at least one underlying device support flush?
104 * Does any underlying device not support WRITE_SAME?
105 *
106 * The callout function is called once for each contiguous section of
107 * an underlying device. State can be maintained in *data.
108 * Return non-zero to stop iterating through any further devices.
109 */
110 typedef int (*iterate_devices_callout_fn) (struct dm_target *ti,
111 struct dm_dev *dev,
112 sector_t start, sector_t len,
113 void *data);
114
115 /*
116 * This function must iterate through each section of device used by the
117 * target until it encounters a non-zero return code, which it then returns.
118 * Returns zero if no callout returned non-zero.
119 */
120 typedef int (*dm_iterate_devices_fn) (struct dm_target *ti,
121 iterate_devices_callout_fn fn,
122 void *data);
123
124 typedef void (*dm_io_hints_fn) (struct dm_target *ti,
125 struct queue_limits *limits);
126
127 /*
128 * Returns:
129 * 0: The target can handle the next I/O immediately.
130 * 1: The target can't handle the next I/O immediately.
131 */
132 typedef int (*dm_busy_fn) (struct dm_target *ti);
133
134 /*
135 * Returns:
136 * < 0 : error
137 * >= 0 : the number of bytes accessible at the address
138 */
139 typedef long (*dm_dax_direct_access_fn) (struct dm_target *ti, pgoff_t pgoff,
140 long nr_pages, void **kaddr, pfn_t *pfn);
141 typedef size_t (*dm_dax_copy_iter_fn)(struct dm_target *ti, pgoff_t pgoff,
142 void *addr, size_t bytes, struct iov_iter *i);
143 typedef int (*dm_dax_zero_page_range_fn)(struct dm_target *ti, pgoff_t pgoff,
144 size_t nr_pages);
145 #define PAGE_SECTORS (PAGE_SIZE / 512)
146
147 void dm_error(const char *message);
148
149 struct dm_dev {
150 struct block_device *bdev;
151 struct dax_device *dax_dev;
152 fmode_t mode;
153 char name[16];
154 };
155
156 dev_t dm_get_dev_t(const char *path);
157
158 /*
159 * Constructors should call these functions to ensure destination devices
160 * are opened/closed correctly.
161 */
162 int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode,
163 struct dm_dev **result);
164 void dm_put_device(struct dm_target *ti, struct dm_dev *d);
165
166 /*
167 * Information about a target type
168 */
169
170 struct target_type {
171 uint64_t features;
172 const char *name;
173 struct module *module;
174 unsigned version[3];
175 dm_ctr_fn ctr;
176 dm_dtr_fn dtr;
177 dm_map_fn map;
178 dm_clone_and_map_request_fn clone_and_map_rq;
179 dm_release_clone_request_fn release_clone_rq;
180 dm_endio_fn end_io;
181 dm_request_endio_fn rq_end_io;
182 dm_presuspend_fn presuspend;
183 dm_presuspend_undo_fn presuspend_undo;
184 dm_postsuspend_fn postsuspend;
185 dm_preresume_fn preresume;
186 dm_resume_fn resume;
187 dm_status_fn status;
188 dm_message_fn message;
189 dm_prepare_ioctl_fn prepare_ioctl;
190 #ifdef CONFIG_BLK_DEV_ZONED
191 dm_report_zones_fn report_zones;
192 #endif
193 dm_busy_fn busy;
194 dm_iterate_devices_fn iterate_devices;
195 dm_io_hints_fn io_hints;
196 dm_dax_direct_access_fn direct_access;
197 dm_dax_copy_iter_fn dax_copy_from_iter;
198 dm_dax_copy_iter_fn dax_copy_to_iter;
199 dm_dax_zero_page_range_fn dax_zero_page_range;
200
201 /* For internal device-mapper use. */
202 struct list_head list;
203 };
204
205 /*
206 * Target features
207 */
208
209 /*
210 * Any table that contains an instance of this target must have only one.
211 */
212 #define DM_TARGET_SINGLETON 0x00000001
213 #define dm_target_needs_singleton(type) ((type)->features & DM_TARGET_SINGLETON)
214
215 /*
216 * Indicates that a target does not support read-only devices.
217 */
218 #define DM_TARGET_ALWAYS_WRITEABLE 0x00000002
219 #define dm_target_always_writeable(type) \
220 ((type)->features & DM_TARGET_ALWAYS_WRITEABLE)
221
222 /*
223 * Any device that contains a table with an instance of this target may never
224 * have tables containing any different target type.
225 */
226 #define DM_TARGET_IMMUTABLE 0x00000004
227 #define dm_target_is_immutable(type) ((type)->features & DM_TARGET_IMMUTABLE)
228
229 /*
230 * Indicates that a target may replace any target; even immutable targets.
231 * .map, .map_rq, .clone_and_map_rq and .release_clone_rq are all defined.
232 */
233 #define DM_TARGET_WILDCARD 0x00000008
234 #define dm_target_is_wildcard(type) ((type)->features & DM_TARGET_WILDCARD)
235
236 /*
237 * A target implements own bio data integrity.
238 */
239 #define DM_TARGET_INTEGRITY 0x00000010
240 #define dm_target_has_integrity(type) ((type)->features & DM_TARGET_INTEGRITY)
241
242 /*
243 * A target passes integrity data to the lower device.
244 */
245 #define DM_TARGET_PASSES_INTEGRITY 0x00000020
246 #define dm_target_passes_integrity(type) ((type)->features & DM_TARGET_PASSES_INTEGRITY)
247
248 /*
249 * Indicates support for zoned block devices:
250 * - DM_TARGET_ZONED_HM: the target also supports host-managed zoned
251 * block devices but does not support combining different zoned models.
252 * - DM_TARGET_MIXED_ZONED_MODEL: the target supports combining multiple
253 * devices with different zoned models.
254 */
255 #define DM_TARGET_ZONED_HM 0x00000040
256 #define dm_target_supports_zoned_hm(type) ((type)->features & DM_TARGET_ZONED_HM)
257
258 /*
259 * A target handles REQ_NOWAIT
260 */
261 #define DM_TARGET_NOWAIT 0x00000080
262 #define dm_target_supports_nowait(type) ((type)->features & DM_TARGET_NOWAIT)
263
264 #ifdef CONFIG_BLK_DEV_ZONED
265 #define DM_TARGET_MIXED_ZONED_MODEL 0x00000200
266 #define dm_target_supports_mixed_zoned_model(type) \
267 ((type)->features & DM_TARGET_MIXED_ZONED_MODEL)
268 #else
269 #define DM_TARGET_MIXED_ZONED_MODEL 0x00000000
270 #define dm_target_supports_mixed_zoned_model(type) (false)
271 #endif
272
273 struct dm_target {
274 struct dm_table *table;
275 struct target_type *type;
276
277 /* target limits */
278 sector_t begin;
279 sector_t len;
280
281 /* If non-zero, maximum size of I/O submitted to a target. */
282 uint32_t max_io_len;
283
284 /*
285 * A number of zero-length barrier bios that will be submitted
286 * to the target for the purpose of flushing cache.
287 *
288 * The bio number can be accessed with dm_bio_get_target_bio_nr.
289 * It is a responsibility of the target driver to remap these bios
290 * to the real underlying devices.
291 */
292 unsigned num_flush_bios;
293
294 /*
295 * The number of discard bios that will be submitted to the target.
296 * The bio number can be accessed with dm_bio_get_target_bio_nr.
297 */
298 unsigned num_discard_bios;
299
300 /*
301 * The number of secure erase bios that will be submitted to the target.
302 * The bio number can be accessed with dm_bio_get_target_bio_nr.
303 */
304 unsigned num_secure_erase_bios;
305
306 /*
307 * The number of WRITE SAME bios that will be submitted to the target.
308 * The bio number can be accessed with dm_bio_get_target_bio_nr.
309 */
310 unsigned num_write_same_bios;
311
312 /*
313 * The number of WRITE ZEROES bios that will be submitted to the target.
314 * The bio number can be accessed with dm_bio_get_target_bio_nr.
315 */
316 unsigned num_write_zeroes_bios;
317
318 /*
319 * The minimum number of extra bytes allocated in each io for the
320 * target to use.
321 */
322 unsigned per_io_data_size;
323
324 /* target specific data */
325 void *private;
326
327 /* Used to provide an error string from the ctr */
328 char *error;
329
330 /*
331 * Set if this target needs to receive flushes regardless of
332 * whether or not its underlying devices have support.
333 */
334 bool flush_supported:1;
335
336 /*
337 * Set if this target needs to receive discards regardless of
338 * whether or not its underlying devices have support.
339 */
340 bool discards_supported:1;
341
342 /*
343 * Set if we need to limit the number of in-flight bios when swapping.
344 */
345 bool limit_swap_bios:1;
346 };
347
348 void *dm_per_bio_data(struct bio *bio, size_t data_size);
349 struct bio *dm_bio_from_per_bio_data(void *data, size_t data_size);
350 unsigned dm_bio_get_target_bio_nr(const struct bio *bio);
351
352 u64 dm_start_time_ns_from_clone(struct bio *bio);
353
354 int dm_register_target(struct target_type *t);
355 void dm_unregister_target(struct target_type *t);
356
357 /*
358 * Target argument parsing.
359 */
360 struct dm_arg_set {
361 unsigned argc;
362 char **argv;
363 };
364
365 /*
366 * The minimum and maximum value of a numeric argument, together with
367 * the error message to use if the number is found to be outside that range.
368 */
369 struct dm_arg {
370 unsigned min;
371 unsigned max;
372 char *error;
373 };
374
375 /*
376 * Validate the next argument, either returning it as *value or, if invalid,
377 * returning -EINVAL and setting *error.
378 */
379 int dm_read_arg(const struct dm_arg *arg, struct dm_arg_set *arg_set,
380 unsigned *value, char **error);
381
382 /*
383 * Process the next argument as the start of a group containing between
384 * arg->min and arg->max further arguments. Either return the size as
385 * *num_args or, if invalid, return -EINVAL and set *error.
386 */
387 int dm_read_arg_group(const struct dm_arg *arg, struct dm_arg_set *arg_set,
388 unsigned *num_args, char **error);
389
390 /*
391 * Return the current argument and shift to the next.
392 */
393 const char *dm_shift_arg(struct dm_arg_set *as);
394
395 /*
396 * Move through num_args arguments.
397 */
398 void dm_consume_args(struct dm_arg_set *as, unsigned num_args);
399
400 /*-----------------------------------------------------------------
401 * Functions for creating and manipulating mapped devices.
402 * Drop the reference with dm_put when you finish with the object.
403 *---------------------------------------------------------------*/
404
405 /*
406 * DM_ANY_MINOR chooses the next available minor number.
407 */
408 #define DM_ANY_MINOR (-1)
409 int dm_create(int minor, struct mapped_device **md);
410
411 /*
412 * Reference counting for md.
413 */
414 struct mapped_device *dm_get_md(dev_t dev);
415 void dm_get(struct mapped_device *md);
416 int dm_hold(struct mapped_device *md);
417 void dm_put(struct mapped_device *md);
418
419 /*
420 * An arbitrary pointer may be stored alongside a mapped device.
421 */
422 void dm_set_mdptr(struct mapped_device *md, void *ptr);
423 void *dm_get_mdptr(struct mapped_device *md);
424
425 /*
426 * A device can still be used while suspended, but I/O is deferred.
427 */
428 int dm_suspend(struct mapped_device *md, unsigned suspend_flags);
429 int dm_resume(struct mapped_device *md);
430
431 /*
432 * Event functions.
433 */
434 uint32_t dm_get_event_nr(struct mapped_device *md);
435 int dm_wait_event(struct mapped_device *md, int event_nr);
436 uint32_t dm_next_uevent_seq(struct mapped_device *md);
437 void dm_uevent_add(struct mapped_device *md, struct list_head *elist);
438
439 /*
440 * Info functions.
441 */
442 const char *dm_device_name(struct mapped_device *md);
443 int dm_copy_name_and_uuid(struct mapped_device *md, char *name, char *uuid);
444 struct gendisk *dm_disk(struct mapped_device *md);
445 int dm_suspended(struct dm_target *ti);
446 int dm_post_suspending(struct dm_target *ti);
447 int dm_noflush_suspending(struct dm_target *ti);
448 void dm_accept_partial_bio(struct bio *bio, unsigned n_sectors);
449 union map_info *dm_get_rq_mapinfo(struct request *rq);
450
451 #ifdef CONFIG_BLK_DEV_ZONED
452 struct dm_report_zones_args {
453 struct dm_target *tgt;
454 sector_t next_sector;
455
456 void *orig_data;
457 report_zones_cb orig_cb;
458 unsigned int zone_idx;
459
460 /* must be filled by ->report_zones before calling dm_report_zones_cb */
461 sector_t start;
462 };
463 int dm_report_zones_cb(struct blk_zone *zone, unsigned int idx, void *data);
464 #endif /* CONFIG_BLK_DEV_ZONED */
465
466 /*
467 * Device mapper functions to parse and create devices specified by the
468 * parameter "dm-mod.create="
469 */
470 int __init dm_early_create(struct dm_ioctl *dmi,
471 struct dm_target_spec **spec_array,
472 char **target_params_array);
473
474 struct queue_limits *dm_get_queue_limits(struct mapped_device *md);
475
476 /*
477 * Geometry functions.
478 */
479 int dm_get_geometry(struct mapped_device *md, struct hd_geometry *geo);
480 int dm_set_geometry(struct mapped_device *md, struct hd_geometry *geo);
481
482 /*-----------------------------------------------------------------
483 * Functions for manipulating device-mapper tables.
484 *---------------------------------------------------------------*/
485
486 /*
487 * First create an empty table.
488 */
489 int dm_table_create(struct dm_table **result, fmode_t mode,
490 unsigned num_targets, struct mapped_device *md);
491
492 /*
493 * Then call this once for each target.
494 */
495 int dm_table_add_target(struct dm_table *t, const char *type,
496 sector_t start, sector_t len, char *params);
497
498 /*
499 * Target can use this to set the table's type.
500 * Can only ever be called from a target's ctr.
501 * Useful for "hybrid" target (supports both bio-based
502 * and request-based).
503 */
504 void dm_table_set_type(struct dm_table *t, enum dm_queue_mode type);
505
506 /*
507 * Finally call this to make the table ready for use.
508 */
509 int dm_table_complete(struct dm_table *t);
510
511 /*
512 * Destroy the table when finished.
513 */
514 void dm_table_destroy(struct dm_table *t);
515
516 /*
517 * Target may require that it is never sent I/O larger than len.
518 */
519 int __must_check dm_set_target_max_io_len(struct dm_target *ti, sector_t len);
520
521 /*
522 * Table reference counting.
523 */
524 struct dm_table *dm_get_live_table(struct mapped_device *md, int *srcu_idx);
525 void dm_put_live_table(struct mapped_device *md, int srcu_idx);
526 void dm_sync_table(struct mapped_device *md);
527
528 /*
529 * Queries
530 */
531 sector_t dm_table_get_size(struct dm_table *t);
532 unsigned int dm_table_get_num_targets(struct dm_table *t);
533 fmode_t dm_table_get_mode(struct dm_table *t);
534 struct mapped_device *dm_table_get_md(struct dm_table *t);
535 const char *dm_table_device_name(struct dm_table *t);
536
537 /*
538 * Trigger an event.
539 */
540 void dm_table_event(struct dm_table *t);
541
542 /*
543 * Run the queue for request-based targets.
544 */
545 void dm_table_run_md_queue_async(struct dm_table *t);
546
547 /*
548 * The device must be suspended before calling this method.
549 * Returns the previous table, which the caller must destroy.
550 */
551 struct dm_table *dm_swap_table(struct mapped_device *md,
552 struct dm_table *t);
553
554 /*
555 * A wrapper around vmalloc.
556 */
557 void *dm_vcalloc(unsigned long nmemb, unsigned long elem_size);
558
559 /*-----------------------------------------------------------------
560 * Macros.
561 *---------------------------------------------------------------*/
562 #define DM_NAME "device-mapper"
563
564 #define DM_FMT(fmt) DM_NAME ": " DM_MSG_PREFIX ": " fmt "\n"
565
566 #define DMCRIT(fmt, ...) pr_crit(DM_FMT(fmt), ##__VA_ARGS__)
567
568 #define DMERR(fmt, ...) pr_err(DM_FMT(fmt), ##__VA_ARGS__)
569 #define DMERR_LIMIT(fmt, ...) pr_err_ratelimited(DM_FMT(fmt), ##__VA_ARGS__)
570 #define DMWARN(fmt, ...) pr_warn(DM_FMT(fmt), ##__VA_ARGS__)
571 #define DMWARN_LIMIT(fmt, ...) pr_warn_ratelimited(DM_FMT(fmt), ##__VA_ARGS__)
572 #define DMINFO(fmt, ...) pr_info(DM_FMT(fmt), ##__VA_ARGS__)
573 #define DMINFO_LIMIT(fmt, ...) pr_info_ratelimited(DM_FMT(fmt), ##__VA_ARGS__)
574
575 #define DMDEBUG(fmt, ...) pr_debug(DM_FMT(fmt), ##__VA_ARGS__)
576 #define DMDEBUG_LIMIT(fmt, ...) pr_debug_ratelimited(DM_FMT(fmt), ##__VA_ARGS__)
577
578 #define DMEMIT(x...) sz += ((sz >= maxlen) ? \
579 0 : scnprintf(result + sz, maxlen - sz, x))
580
581 /*
582 * Definitions of return values from target end_io function.
583 */
584 #define DM_ENDIO_DONE 0
585 #define DM_ENDIO_INCOMPLETE 1
586 #define DM_ENDIO_REQUEUE 2
587 #define DM_ENDIO_DELAY_REQUEUE 3
588
589 /*
590 * Definitions of return values from target map function.
591 */
592 #define DM_MAPIO_SUBMITTED 0
593 #define DM_MAPIO_REMAPPED 1
594 #define DM_MAPIO_REQUEUE DM_ENDIO_REQUEUE
595 #define DM_MAPIO_DELAY_REQUEUE DM_ENDIO_DELAY_REQUEUE
596 #define DM_MAPIO_KILL 4
597
598 #define dm_sector_div64(x, y)( \
599 { \
600 u64 _res; \
601 (x) = div64_u64_rem(x, y, &_res); \
602 _res; \
603 } \
604 )
605
606 /*
607 * Ceiling(n / sz)
608 */
609 #define dm_div_up(n, sz) (((n) + (sz) - 1) / (sz))
610
611 #define dm_sector_div_up(n, sz) ( \
612 { \
613 sector_t _r = ((n) + (sz) - 1); \
614 sector_div(_r, (sz)); \
615 _r; \
616 } \
617 )
618
619 /*
620 * ceiling(n / size) * size
621 */
622 #define dm_round_up(n, sz) (dm_div_up((n), (sz)) * (sz))
623
624 /*
625 * Sector offset taken relative to the start of the target instead of
626 * relative to the start of the device.
627 */
628 #define dm_target_offset(ti, sector) ((sector) - (ti)->begin)
629
to_sector(unsigned long long n)630 static inline sector_t to_sector(unsigned long long n)
631 {
632 return (n >> SECTOR_SHIFT);
633 }
634
to_bytes(sector_t n)635 static inline unsigned long to_bytes(sector_t n)
636 {
637 return (n << SECTOR_SHIFT);
638 }
639
640 #endif /* _LINUX_DEVICE_MAPPER_H */
641