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