1 /* SPDX-License-Identifier: GPL-2.0-only */
2 /*
3 * Copyright (C) 2001 Sistina Software (UK) Limited.
4 * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
5 *
6 * This file is released under the LGPL.
7 */
8
9 #ifndef _LINUX_DEVICE_MAPPER_H
10 #define _LINUX_DEVICE_MAPPER_H
11
12 #include <linux/bio.h>
13 #include <linux/blkdev.h>
14 #include <linux/dm-ioctl.h>
15 #include <linux/math64.h>
16 #include <linux/ratelimit.h>
17
18 struct dm_dev;
19 struct dm_target;
20 struct dm_table;
21 struct dm_report_zones_args;
22 struct mapped_device;
23 struct bio_vec;
24 enum dax_access_mode;
25
26 /*
27 * Type of table, mapped_device's mempool and request_queue
28 */
29 enum dm_queue_mode {
30 DM_TYPE_NONE = 0,
31 DM_TYPE_BIO_BASED = 1,
32 DM_TYPE_REQUEST_BASED = 2,
33 DM_TYPE_DAX_BIO_BASED = 3,
34 };
35
36 typedef enum { STATUSTYPE_INFO, STATUSTYPE_TABLE, STATUSTYPE_IMA } status_type_t;
37
38 union map_info {
39 void *ptr;
40 };
41
42 /*
43 * In the constructor the target parameter will already have the
44 * table, type, begin and len fields filled in.
45 */
46 typedef int (*dm_ctr_fn) (struct dm_target *target,
47 unsigned int argc, char **argv);
48
49 /*
50 * The destructor doesn't need to free the dm_target, just
51 * anything hidden ti->private.
52 */
53 typedef void (*dm_dtr_fn) (struct dm_target *ti);
54
55 /*
56 * The map function must return:
57 * < 0: error
58 * = 0: The target will handle the io by resubmitting it later
59 * = 1: simple remap complete
60 * = 2: The target wants to push back the io
61 */
62 typedef int (*dm_map_fn) (struct dm_target *ti, struct bio *bio);
63 typedef int (*dm_clone_and_map_request_fn) (struct dm_target *ti,
64 struct request *rq,
65 union map_info *map_context,
66 struct request **clone);
67 typedef void (*dm_release_clone_request_fn) (struct request *clone,
68 union map_info *map_context);
69
70 /*
71 * Returns:
72 * < 0 : error (currently ignored)
73 * 0 : ended successfully
74 * 1 : for some reason the io has still not completed (eg,
75 * multipath target might want to requeue a failed io).
76 * 2 : The target wants to push back the io
77 */
78 typedef int (*dm_endio_fn) (struct dm_target *ti,
79 struct bio *bio, blk_status_t *error);
80 typedef int (*dm_request_endio_fn) (struct dm_target *ti,
81 struct request *clone, blk_status_t error,
82 union map_info *map_context);
83
84 typedef void (*dm_presuspend_fn) (struct dm_target *ti);
85 typedef void (*dm_presuspend_undo_fn) (struct dm_target *ti);
86 typedef void (*dm_postsuspend_fn) (struct dm_target *ti);
87 typedef int (*dm_preresume_fn) (struct dm_target *ti);
88 typedef void (*dm_resume_fn) (struct dm_target *ti);
89
90 typedef void (*dm_status_fn) (struct dm_target *ti, status_type_t status_type,
91 unsigned int status_flags, char *result, unsigned int maxlen);
92
93 typedef int (*dm_message_fn) (struct dm_target *ti, unsigned int argc, char **argv,
94 char *result, unsigned int maxlen);
95
96 typedef int (*dm_prepare_ioctl_fn) (struct dm_target *ti, struct block_device **bdev);
97
98 #ifdef CONFIG_BLK_DEV_ZONED
99 typedef int (*dm_report_zones_fn) (struct dm_target *ti,
100 struct dm_report_zones_args *args,
101 unsigned int nr_zones);
102 #else
103 /*
104 * Define dm_report_zones_fn so that targets can assign to NULL if
105 * CONFIG_BLK_DEV_ZONED disabled. Otherwise each target needs to do
106 * awkward #ifdefs in their target_type, etc.
107 */
108 typedef int (*dm_report_zones_fn) (struct dm_target *dummy);
109 #endif
110
111 /*
112 * These iteration functions are typically used to check (and combine)
113 * properties of underlying devices.
114 * E.g. Does at least one underlying device support flush?
115 * Does any underlying device not support WRITE_SAME?
116 *
117 * The callout function is called once for each contiguous section of
118 * an underlying device. State can be maintained in *data.
119 * Return non-zero to stop iterating through any further devices.
120 */
121 typedef int (*iterate_devices_callout_fn) (struct dm_target *ti,
122 struct dm_dev *dev,
123 sector_t start, sector_t len,
124 void *data);
125
126 /*
127 * This function must iterate through each section of device used by the
128 * target until it encounters a non-zero return code, which it then returns.
129 * Returns zero if no callout returned non-zero.
130 */
131 typedef int (*dm_iterate_devices_fn) (struct dm_target *ti,
132 iterate_devices_callout_fn fn,
133 void *data);
134
135 typedef void (*dm_io_hints_fn) (struct dm_target *ti,
136 struct queue_limits *limits);
137
138 /*
139 * Returns:
140 * 0: The target can handle the next I/O immediately.
141 * 1: The target can't handle the next I/O immediately.
142 */
143 typedef int (*dm_busy_fn) (struct dm_target *ti);
144
145 /*
146 * Returns:
147 * < 0 : error
148 * >= 0 : the number of bytes accessible at the address
149 */
150 typedef long (*dm_dax_direct_access_fn) (struct dm_target *ti, pgoff_t pgoff,
151 long nr_pages, enum dax_access_mode node, void **kaddr,
152 pfn_t *pfn);
153 typedef int (*dm_dax_zero_page_range_fn)(struct dm_target *ti, pgoff_t pgoff,
154 size_t nr_pages);
155
156 /*
157 * Returns:
158 * != 0 : number of bytes transferred
159 * 0 : recovery write failed
160 */
161 typedef size_t (*dm_dax_recovery_write_fn)(struct dm_target *ti, pgoff_t pgoff,
162 void *addr, size_t bytes, struct iov_iter *i);
163
164 void dm_error(const char *message);
165
166 struct dm_dev {
167 struct block_device *bdev;
168 struct dax_device *dax_dev;
169 fmode_t mode;
170 char name[16];
171 };
172
173 dev_t dm_get_dev_t(const char *path);
174
175 /*
176 * Constructors should call these functions to ensure destination devices
177 * are opened/closed correctly.
178 */
179 int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode,
180 struct dm_dev **result);
181 void dm_put_device(struct dm_target *ti, struct dm_dev *d);
182
183 /*
184 * Information about a target type
185 */
186
187 struct target_type {
188 uint64_t features;
189 const char *name;
190 struct module *module;
191 unsigned int version[3];
192 dm_ctr_fn ctr;
193 dm_dtr_fn dtr;
194 dm_map_fn map;
195 dm_clone_and_map_request_fn clone_and_map_rq;
196 dm_release_clone_request_fn release_clone_rq;
197 dm_endio_fn end_io;
198 dm_request_endio_fn rq_end_io;
199 dm_presuspend_fn presuspend;
200 dm_presuspend_undo_fn presuspend_undo;
201 dm_postsuspend_fn postsuspend;
202 dm_preresume_fn preresume;
203 dm_resume_fn resume;
204 dm_status_fn status;
205 dm_message_fn message;
206 dm_prepare_ioctl_fn prepare_ioctl;
207 dm_report_zones_fn report_zones;
208 dm_busy_fn busy;
209 dm_iterate_devices_fn iterate_devices;
210 dm_io_hints_fn io_hints;
211 dm_dax_direct_access_fn direct_access;
212 dm_dax_zero_page_range_fn dax_zero_page_range;
213 dm_dax_recovery_write_fn dax_recovery_write;
214
215 /* For internal device-mapper use. */
216 struct list_head list;
217 };
218
219 /*
220 * Target features
221 */
222
223 /*
224 * Any table that contains an instance of this target must have only one.
225 */
226 #define DM_TARGET_SINGLETON 0x00000001
227 #define dm_target_needs_singleton(type) ((type)->features & DM_TARGET_SINGLETON)
228
229 /*
230 * Indicates that a target does not support read-only devices.
231 */
232 #define DM_TARGET_ALWAYS_WRITEABLE 0x00000002
233 #define dm_target_always_writeable(type) \
234 ((type)->features & DM_TARGET_ALWAYS_WRITEABLE)
235
236 /*
237 * Any device that contains a table with an instance of this target may never
238 * have tables containing any different target type.
239 */
240 #define DM_TARGET_IMMUTABLE 0x00000004
241 #define dm_target_is_immutable(type) ((type)->features & DM_TARGET_IMMUTABLE)
242
243 /*
244 * Indicates that a target may replace any target; even immutable targets.
245 * .map, .map_rq, .clone_and_map_rq and .release_clone_rq are all defined.
246 */
247 #define DM_TARGET_WILDCARD 0x00000008
248 #define dm_target_is_wildcard(type) ((type)->features & DM_TARGET_WILDCARD)
249
250 /*
251 * A target implements own bio data integrity.
252 */
253 #define DM_TARGET_INTEGRITY 0x00000010
254 #define dm_target_has_integrity(type) ((type)->features & DM_TARGET_INTEGRITY)
255
256 /*
257 * A target passes integrity data to the lower device.
258 */
259 #define DM_TARGET_PASSES_INTEGRITY 0x00000020
260 #define dm_target_passes_integrity(type) ((type)->features & DM_TARGET_PASSES_INTEGRITY)
261
262 /*
263 * Indicates support for zoned block devices:
264 * - DM_TARGET_ZONED_HM: the target also supports host-managed zoned
265 * block devices but does not support combining different zoned models.
266 * - DM_TARGET_MIXED_ZONED_MODEL: the target supports combining multiple
267 * devices with different zoned models.
268 */
269 #ifdef CONFIG_BLK_DEV_ZONED
270 #define DM_TARGET_ZONED_HM 0x00000040
271 #define dm_target_supports_zoned_hm(type) ((type)->features & DM_TARGET_ZONED_HM)
272 #else
273 #define DM_TARGET_ZONED_HM 0x00000000
274 #define dm_target_supports_zoned_hm(type) (false)
275 #endif
276
277 /*
278 * A target handles REQ_NOWAIT
279 */
280 #define DM_TARGET_NOWAIT 0x00000080
281 #define dm_target_supports_nowait(type) ((type)->features & DM_TARGET_NOWAIT)
282
283 /*
284 * A target supports passing through inline crypto support.
285 */
286 #define DM_TARGET_PASSES_CRYPTO 0x00000100
287 #define dm_target_passes_crypto(type) ((type)->features & DM_TARGET_PASSES_CRYPTO)
288
289 #ifdef CONFIG_BLK_DEV_ZONED
290 #define DM_TARGET_MIXED_ZONED_MODEL 0x00000200
291 #define dm_target_supports_mixed_zoned_model(type) \
292 ((type)->features & DM_TARGET_MIXED_ZONED_MODEL)
293 #else
294 #define DM_TARGET_MIXED_ZONED_MODEL 0x00000000
295 #define dm_target_supports_mixed_zoned_model(type) (false)
296 #endif
297
298 struct dm_target {
299 struct dm_table *table;
300 struct target_type *type;
301
302 /* target limits */
303 sector_t begin;
304 sector_t len;
305
306 /* If non-zero, maximum size of I/O submitted to a target. */
307 uint32_t max_io_len;
308
309 /*
310 * A number of zero-length barrier bios that will be submitted
311 * to the target for the purpose of flushing cache.
312 *
313 * The bio number can be accessed with dm_bio_get_target_bio_nr.
314 * It is a responsibility of the target driver to remap these bios
315 * to the real underlying devices.
316 */
317 unsigned int num_flush_bios;
318
319 /*
320 * The number of discard bios that will be submitted to the target.
321 * The bio number can be accessed with dm_bio_get_target_bio_nr.
322 */
323 unsigned int num_discard_bios;
324
325 /*
326 * The number of secure erase bios that will be submitted to the target.
327 * The bio number can be accessed with dm_bio_get_target_bio_nr.
328 */
329 unsigned int num_secure_erase_bios;
330
331 /*
332 * The number of WRITE ZEROES bios that will be submitted to the target.
333 * The bio number can be accessed with dm_bio_get_target_bio_nr.
334 */
335 unsigned int num_write_zeroes_bios;
336
337 /*
338 * The minimum number of extra bytes allocated in each io for the
339 * target to use.
340 */
341 unsigned int per_io_data_size;
342
343 /* target specific data */
344 void *private;
345
346 /* Used to provide an error string from the ctr */
347 char *error;
348
349 /*
350 * Set if this target needs to receive flushes regardless of
351 * whether or not its underlying devices have support.
352 */
353 bool flush_supported:1;
354
355 /*
356 * Set if this target needs to receive discards regardless of
357 * whether or not its underlying devices have support.
358 */
359 bool discards_supported:1;
360
361 /*
362 * Set if we need to limit the number of in-flight bios when swapping.
363 */
364 bool limit_swap_bios:1;
365
366 /*
367 * Set if this target implements a zoned device and needs emulation of
368 * zone append operations using regular writes.
369 */
370 bool emulate_zone_append:1;
371
372 /*
373 * Set if the target will submit IO using dm_submit_bio_remap()
374 * after returning DM_MAPIO_SUBMITTED from its map function.
375 */
376 bool accounts_remapped_io:1;
377
378 /*
379 * Set if the target will submit the DM bio without first calling
380 * bio_set_dev(). NOTE: ideally a target should _not_ need this.
381 */
382 bool needs_bio_set_dev:1;
383 };
384
385 void *dm_per_bio_data(struct bio *bio, size_t data_size);
386 struct bio *dm_bio_from_per_bio_data(void *data, size_t data_size);
387 unsigned int dm_bio_get_target_bio_nr(const struct bio *bio);
388
389 u64 dm_start_time_ns_from_clone(struct bio *bio);
390
391 int dm_register_target(struct target_type *t);
392 void dm_unregister_target(struct target_type *t);
393
394 /*
395 * Target argument parsing.
396 */
397 struct dm_arg_set {
398 unsigned int argc;
399 char **argv;
400 };
401
402 /*
403 * The minimum and maximum value of a numeric argument, together with
404 * the error message to use if the number is found to be outside that range.
405 */
406 struct dm_arg {
407 unsigned int min;
408 unsigned int max;
409 char *error;
410 };
411
412 /*
413 * Validate the next argument, either returning it as *value or, if invalid,
414 * returning -EINVAL and setting *error.
415 */
416 int dm_read_arg(const struct dm_arg *arg, struct dm_arg_set *arg_set,
417 unsigned int *value, char **error);
418
419 /*
420 * Process the next argument as the start of a group containing between
421 * arg->min and arg->max further arguments. Either return the size as
422 * *num_args or, if invalid, return -EINVAL and set *error.
423 */
424 int dm_read_arg_group(const struct dm_arg *arg, struct dm_arg_set *arg_set,
425 unsigned int *num_args, char **error);
426
427 /*
428 * Return the current argument and shift to the next.
429 */
430 const char *dm_shift_arg(struct dm_arg_set *as);
431
432 /*
433 * Move through num_args arguments.
434 */
435 void dm_consume_args(struct dm_arg_set *as, unsigned int num_args);
436
437 /*
438 *----------------------------------------------------------------
439 * Functions for creating and manipulating mapped devices.
440 * Drop the reference with dm_put when you finish with the object.
441 *----------------------------------------------------------------
442 */
443
444 /*
445 * DM_ANY_MINOR chooses the next available minor number.
446 */
447 #define DM_ANY_MINOR (-1)
448 int dm_create(int minor, struct mapped_device **md);
449
450 /*
451 * Reference counting for md.
452 */
453 struct mapped_device *dm_get_md(dev_t dev);
454 void dm_get(struct mapped_device *md);
455 int dm_hold(struct mapped_device *md);
456 void dm_put(struct mapped_device *md);
457
458 /*
459 * An arbitrary pointer may be stored alongside a mapped device.
460 */
461 void dm_set_mdptr(struct mapped_device *md, void *ptr);
462 void *dm_get_mdptr(struct mapped_device *md);
463
464 /*
465 * A device can still be used while suspended, but I/O is deferred.
466 */
467 int dm_suspend(struct mapped_device *md, unsigned int suspend_flags);
468 int dm_resume(struct mapped_device *md);
469
470 /*
471 * Event functions.
472 */
473 uint32_t dm_get_event_nr(struct mapped_device *md);
474 int dm_wait_event(struct mapped_device *md, int event_nr);
475 uint32_t dm_next_uevent_seq(struct mapped_device *md);
476 void dm_uevent_add(struct mapped_device *md, struct list_head *elist);
477
478 /*
479 * Info functions.
480 */
481 const char *dm_device_name(struct mapped_device *md);
482 int dm_copy_name_and_uuid(struct mapped_device *md, char *name, char *uuid);
483 struct gendisk *dm_disk(struct mapped_device *md);
484 int dm_suspended(struct dm_target *ti);
485 int dm_post_suspending(struct dm_target *ti);
486 int dm_noflush_suspending(struct dm_target *ti);
487 void dm_accept_partial_bio(struct bio *bio, unsigned int n_sectors);
488 void dm_submit_bio_remap(struct bio *clone, struct bio *tgt_clone);
489 union map_info *dm_get_rq_mapinfo(struct request *rq);
490
491 #ifdef CONFIG_BLK_DEV_ZONED
492 struct dm_report_zones_args {
493 struct dm_target *tgt;
494 sector_t next_sector;
495
496 void *orig_data;
497 report_zones_cb orig_cb;
498 unsigned int zone_idx;
499
500 /* must be filled by ->report_zones before calling dm_report_zones_cb */
501 sector_t start;
502 };
503 int dm_report_zones(struct block_device *bdev, sector_t start, sector_t sector,
504 struct dm_report_zones_args *args, unsigned int nr_zones);
505 #endif /* CONFIG_BLK_DEV_ZONED */
506
507 /*
508 * Device mapper functions to parse and create devices specified by the
509 * parameter "dm-mod.create="
510 */
511 int __init dm_early_create(struct dm_ioctl *dmi,
512 struct dm_target_spec **spec_array,
513 char **target_params_array);
514
515 struct queue_limits *dm_get_queue_limits(struct mapped_device *md);
516
517 /*
518 * Geometry functions.
519 */
520 int dm_get_geometry(struct mapped_device *md, struct hd_geometry *geo);
521 int dm_set_geometry(struct mapped_device *md, struct hd_geometry *geo);
522
523 /*
524 *---------------------------------------------------------------
525 * Functions for manipulating device-mapper tables.
526 *---------------------------------------------------------------
527 */
528
529 /*
530 * First create an empty table.
531 */
532 int dm_table_create(struct dm_table **result, fmode_t mode,
533 unsigned int num_targets, struct mapped_device *md);
534
535 /*
536 * Then call this once for each target.
537 */
538 int dm_table_add_target(struct dm_table *t, const char *type,
539 sector_t start, sector_t len, char *params);
540
541 /*
542 * Target can use this to set the table's type.
543 * Can only ever be called from a target's ctr.
544 * Useful for "hybrid" target (supports both bio-based
545 * and request-based).
546 */
547 void dm_table_set_type(struct dm_table *t, enum dm_queue_mode type);
548
549 /*
550 * Finally call this to make the table ready for use.
551 */
552 int dm_table_complete(struct dm_table *t);
553
554 /*
555 * Destroy the table when finished.
556 */
557 void dm_table_destroy(struct dm_table *t);
558
559 /*
560 * Target may require that it is never sent I/O larger than len.
561 */
562 int __must_check dm_set_target_max_io_len(struct dm_target *ti, sector_t len);
563
564 /*
565 * Table reference counting.
566 */
567 struct dm_table *dm_get_live_table(struct mapped_device *md, int *srcu_idx);
568 void dm_put_live_table(struct mapped_device *md, int srcu_idx);
569 void dm_sync_table(struct mapped_device *md);
570
571 /*
572 * Queries
573 */
574 sector_t dm_table_get_size(struct dm_table *t);
575 fmode_t dm_table_get_mode(struct dm_table *t);
576 struct mapped_device *dm_table_get_md(struct dm_table *t);
577 const char *dm_table_device_name(struct dm_table *t);
578
579 /*
580 * Trigger an event.
581 */
582 void dm_table_event(struct dm_table *t);
583
584 /*
585 * Run the queue for request-based targets.
586 */
587 void dm_table_run_md_queue_async(struct dm_table *t);
588
589 /*
590 * The device must be suspended before calling this method.
591 * Returns the previous table, which the caller must destroy.
592 */
593 struct dm_table *dm_swap_table(struct mapped_device *md,
594 struct dm_table *t);
595
596 /*
597 * Table blk_crypto_profile functions
598 */
599 void dm_destroy_crypto_profile(struct blk_crypto_profile *profile);
600
601 /*
602 *---------------------------------------------------------------
603 * Macros.
604 *---------------------------------------------------------------
605 */
606 #define DM_NAME "device-mapper"
607
608 #define DM_FMT(fmt) DM_NAME ": " DM_MSG_PREFIX ": " fmt "\n"
609
610 #define DMCRIT(fmt, ...) pr_crit(DM_FMT(fmt), ##__VA_ARGS__)
611
612 #define DMERR(fmt, ...) pr_err(DM_FMT(fmt), ##__VA_ARGS__)
613 #define DMERR_LIMIT(fmt, ...) pr_err_ratelimited(DM_FMT(fmt), ##__VA_ARGS__)
614 #define DMWARN(fmt, ...) pr_warn(DM_FMT(fmt), ##__VA_ARGS__)
615 #define DMWARN_LIMIT(fmt, ...) pr_warn_ratelimited(DM_FMT(fmt), ##__VA_ARGS__)
616 #define DMINFO(fmt, ...) pr_info(DM_FMT(fmt), ##__VA_ARGS__)
617 #define DMINFO_LIMIT(fmt, ...) pr_info_ratelimited(DM_FMT(fmt), ##__VA_ARGS__)
618
619 #define DMDEBUG(fmt, ...) pr_debug(DM_FMT(fmt), ##__VA_ARGS__)
620 #define DMDEBUG_LIMIT(fmt, ...) pr_debug_ratelimited(DM_FMT(fmt), ##__VA_ARGS__)
621
622 #define DMEMIT(x...) (sz += ((sz >= maxlen) ? 0 : scnprintf(result + sz, maxlen - sz, x)))
623
624 #define DMEMIT_TARGET_NAME_VERSION(y) \
625 DMEMIT("target_name=%s,target_version=%u.%u.%u", \
626 (y)->name, (y)->version[0], (y)->version[1], (y)->version[2])
627
628 /*
629 * Definitions of return values from target end_io function.
630 */
631 #define DM_ENDIO_DONE 0
632 #define DM_ENDIO_INCOMPLETE 1
633 #define DM_ENDIO_REQUEUE 2
634 #define DM_ENDIO_DELAY_REQUEUE 3
635
636 /*
637 * Definitions of return values from target map function.
638 */
639 #define DM_MAPIO_SUBMITTED 0
640 #define DM_MAPIO_REMAPPED 1
641 #define DM_MAPIO_REQUEUE DM_ENDIO_REQUEUE
642 #define DM_MAPIO_DELAY_REQUEUE DM_ENDIO_DELAY_REQUEUE
643 #define DM_MAPIO_KILL 4
644
645 #define dm_sector_div64(x, y)( \
646 { \
647 u64 _res; \
648 (x) = div64_u64_rem(x, y, &_res); \
649 _res; \
650 } \
651 )
652
653 /*
654 * Ceiling(n / sz)
655 */
656 #define dm_div_up(n, sz) (((n) + (sz) - 1) / (sz))
657
658 #define dm_sector_div_up(n, sz) ( \
659 { \
660 sector_t _r = ((n) + (sz) - 1); \
661 sector_div(_r, (sz)); \
662 _r; \
663 } \
664 )
665
666 /*
667 * ceiling(n / size) * size
668 */
669 #define dm_round_up(n, sz) (dm_div_up((n), (sz)) * (sz))
670
671 /*
672 * Sector offset taken relative to the start of the target instead of
673 * relative to the start of the device.
674 */
675 #define dm_target_offset(ti, sector) ((sector) - (ti)->begin)
676
to_sector(unsigned long long n)677 static inline sector_t to_sector(unsigned long long n)
678 {
679 return (n >> SECTOR_SHIFT);
680 }
681
to_bytes(sector_t n)682 static inline unsigned long to_bytes(sector_t n)
683 {
684 return (n << SECTOR_SHIFT);
685 }
686
687 #endif /* _LINUX_DEVICE_MAPPER_H */
688