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 #define PAGE_SECTORS (PAGE_SIZE / 512)
144
145 void dm_error(const char *message);
146
147 struct dm_dev {
148 struct block_device *bdev;
149 struct dax_device *dax_dev;
150 fmode_t mode;
151 char name[16];
152 };
153
154 dev_t dm_get_dev_t(const char *path);
155
156 /*
157 * Constructors should call these functions to ensure destination devices
158 * are opened/closed correctly.
159 */
160 int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode,
161 struct dm_dev **result);
162 void dm_put_device(struct dm_target *ti, struct dm_dev *d);
163
164 /*
165 * Information about a target type
166 */
167
168 struct target_type {
169 uint64_t features;
170 const char *name;
171 struct module *module;
172 unsigned version[3];
173 dm_ctr_fn ctr;
174 dm_dtr_fn dtr;
175 dm_map_fn map;
176 dm_clone_and_map_request_fn clone_and_map_rq;
177 dm_release_clone_request_fn release_clone_rq;
178 dm_endio_fn end_io;
179 dm_request_endio_fn rq_end_io;
180 dm_presuspend_fn presuspend;
181 dm_presuspend_undo_fn presuspend_undo;
182 dm_postsuspend_fn postsuspend;
183 dm_preresume_fn preresume;
184 dm_resume_fn resume;
185 dm_status_fn status;
186 dm_message_fn message;
187 dm_prepare_ioctl_fn prepare_ioctl;
188 #ifdef CONFIG_BLK_DEV_ZONED
189 dm_report_zones_fn report_zones;
190 #endif
191 dm_busy_fn busy;
192 dm_iterate_devices_fn iterate_devices;
193 dm_io_hints_fn io_hints;
194 dm_dax_direct_access_fn direct_access;
195 dm_dax_copy_iter_fn dax_copy_from_iter;
196 dm_dax_copy_iter_fn dax_copy_to_iter;
197
198 /* For internal device-mapper use. */
199 struct list_head list;
200 };
201
202 /*
203 * Target features
204 */
205
206 /*
207 * Any table that contains an instance of this target must have only one.
208 */
209 #define DM_TARGET_SINGLETON 0x00000001
210 #define dm_target_needs_singleton(type) ((type)->features & DM_TARGET_SINGLETON)
211
212 /*
213 * Indicates that a target does not support read-only devices.
214 */
215 #define DM_TARGET_ALWAYS_WRITEABLE 0x00000002
216 #define dm_target_always_writeable(type) \
217 ((type)->features & DM_TARGET_ALWAYS_WRITEABLE)
218
219 /*
220 * Any device that contains a table with an instance of this target may never
221 * have tables containing any different target type.
222 */
223 #define DM_TARGET_IMMUTABLE 0x00000004
224 #define dm_target_is_immutable(type) ((type)->features & DM_TARGET_IMMUTABLE)
225
226 /*
227 * Indicates that a target may replace any target; even immutable targets.
228 * .map, .map_rq, .clone_and_map_rq and .release_clone_rq are all defined.
229 */
230 #define DM_TARGET_WILDCARD 0x00000008
231 #define dm_target_is_wildcard(type) ((type)->features & DM_TARGET_WILDCARD)
232
233 /*
234 * A target implements own bio data integrity.
235 */
236 #define DM_TARGET_INTEGRITY 0x00000010
237 #define dm_target_has_integrity(type) ((type)->features & DM_TARGET_INTEGRITY)
238
239 /*
240 * A target passes integrity data to the lower device.
241 */
242 #define DM_TARGET_PASSES_INTEGRITY 0x00000020
243 #define dm_target_passes_integrity(type) ((type)->features & DM_TARGET_PASSES_INTEGRITY)
244
245 /*
246 * Indicates that a target supports host-managed zoned block devices.
247 */
248 #define DM_TARGET_ZONED_HM 0x00000040
249 #define dm_target_supports_zoned_hm(type) ((type)->features & DM_TARGET_ZONED_HM)
250
251 struct dm_target {
252 struct dm_table *table;
253 struct target_type *type;
254
255 /* target limits */
256 sector_t begin;
257 sector_t len;
258
259 /* If non-zero, maximum size of I/O submitted to a target. */
260 uint32_t max_io_len;
261
262 /*
263 * A number of zero-length barrier bios that will be submitted
264 * to the target for the purpose of flushing cache.
265 *
266 * The bio number can be accessed with dm_bio_get_target_bio_nr.
267 * It is a responsibility of the target driver to remap these bios
268 * to the real underlying devices.
269 */
270 unsigned num_flush_bios;
271
272 /*
273 * The number of discard bios that will be submitted to the target.
274 * The bio number can be accessed with dm_bio_get_target_bio_nr.
275 */
276 unsigned num_discard_bios;
277
278 /*
279 * The number of secure erase bios that will be submitted to the target.
280 * The bio number can be accessed with dm_bio_get_target_bio_nr.
281 */
282 unsigned num_secure_erase_bios;
283
284 /*
285 * The number of WRITE SAME bios that will be submitted to the target.
286 * The bio number can be accessed with dm_bio_get_target_bio_nr.
287 */
288 unsigned num_write_same_bios;
289
290 /*
291 * The number of WRITE ZEROES bios that will be submitted to the target.
292 * The bio number can be accessed with dm_bio_get_target_bio_nr.
293 */
294 unsigned num_write_zeroes_bios;
295
296 /*
297 * The minimum number of extra bytes allocated in each io for the
298 * target to use.
299 */
300 unsigned per_io_data_size;
301
302 /* target specific data */
303 void *private;
304
305 /* Used to provide an error string from the ctr */
306 char *error;
307
308 /*
309 * Set if this target needs to receive flushes regardless of
310 * whether or not its underlying devices have support.
311 */
312 bool flush_supported:1;
313
314 /*
315 * Set if this target needs to receive discards regardless of
316 * whether or not its underlying devices have support.
317 */
318 bool discards_supported:1;
319
320 /*
321 * Set if we need to limit the number of in-flight bios when swapping.
322 */
323 bool limit_swap_bios:1;
324
325 /*
326 * Set if inline crypto capabilities from this target's underlying
327 * device(s) can be exposed via the device-mapper device.
328 */
329 bool may_passthrough_inline_crypto:1;
330 };
331
332 /* Each target can link one of these into the table */
333 struct dm_target_callbacks {
334 struct list_head list;
335 int (*congested_fn) (struct dm_target_callbacks *, int);
336 };
337
338 void *dm_per_bio_data(struct bio *bio, size_t data_size);
339 struct bio *dm_bio_from_per_bio_data(void *data, size_t data_size);
340 unsigned dm_bio_get_target_bio_nr(const struct bio *bio);
341
342 int dm_register_target(struct target_type *t);
343 void dm_unregister_target(struct target_type *t);
344
345 /*
346 * Target argument parsing.
347 */
348 struct dm_arg_set {
349 unsigned argc;
350 char **argv;
351 };
352
353 /*
354 * The minimum and maximum value of a numeric argument, together with
355 * the error message to use if the number is found to be outside that range.
356 */
357 struct dm_arg {
358 unsigned min;
359 unsigned max;
360 char *error;
361 };
362
363 /*
364 * Validate the next argument, either returning it as *value or, if invalid,
365 * returning -EINVAL and setting *error.
366 */
367 int dm_read_arg(const struct dm_arg *arg, struct dm_arg_set *arg_set,
368 unsigned *value, char **error);
369
370 /*
371 * Process the next argument as the start of a group containing between
372 * arg->min and arg->max further arguments. Either return the size as
373 * *num_args or, if invalid, return -EINVAL and set *error.
374 */
375 int dm_read_arg_group(const struct dm_arg *arg, struct dm_arg_set *arg_set,
376 unsigned *num_args, char **error);
377
378 /*
379 * Return the current argument and shift to the next.
380 */
381 const char *dm_shift_arg(struct dm_arg_set *as);
382
383 /*
384 * Move through num_args arguments.
385 */
386 void dm_consume_args(struct dm_arg_set *as, unsigned num_args);
387
388 /*-----------------------------------------------------------------
389 * Functions for creating and manipulating mapped devices.
390 * Drop the reference with dm_put when you finish with the object.
391 *---------------------------------------------------------------*/
392
393 /*
394 * DM_ANY_MINOR chooses the next available minor number.
395 */
396 #define DM_ANY_MINOR (-1)
397 int dm_create(int minor, struct mapped_device **md);
398
399 /*
400 * Reference counting for md.
401 */
402 struct mapped_device *dm_get_md(dev_t dev);
403 void dm_get(struct mapped_device *md);
404 int dm_hold(struct mapped_device *md);
405 void dm_put(struct mapped_device *md);
406
407 /*
408 * An arbitrary pointer may be stored alongside a mapped device.
409 */
410 void dm_set_mdptr(struct mapped_device *md, void *ptr);
411 void *dm_get_mdptr(struct mapped_device *md);
412
413 /*
414 * A device can still be used while suspended, but I/O is deferred.
415 */
416 int dm_suspend(struct mapped_device *md, unsigned suspend_flags);
417 int dm_resume(struct mapped_device *md);
418
419 /*
420 * Event functions.
421 */
422 uint32_t dm_get_event_nr(struct mapped_device *md);
423 int dm_wait_event(struct mapped_device *md, int event_nr);
424 uint32_t dm_next_uevent_seq(struct mapped_device *md);
425 void dm_uevent_add(struct mapped_device *md, struct list_head *elist);
426
427 /*
428 * Info functions.
429 */
430 const char *dm_device_name(struct mapped_device *md);
431 int dm_copy_name_and_uuid(struct mapped_device *md, char *name, char *uuid);
432 struct gendisk *dm_disk(struct mapped_device *md);
433 int dm_suspended(struct dm_target *ti);
434 int dm_post_suspending(struct dm_target *ti);
435 int dm_noflush_suspending(struct dm_target *ti);
436 void dm_accept_partial_bio(struct bio *bio, unsigned n_sectors);
437 union map_info *dm_get_rq_mapinfo(struct request *rq);
438
439 #ifdef CONFIG_BLK_DEV_ZONED
440 struct dm_report_zones_args {
441 struct dm_target *tgt;
442 sector_t next_sector;
443
444 void *orig_data;
445 report_zones_cb orig_cb;
446 unsigned int zone_idx;
447
448 /* must be filled by ->report_zones before calling dm_report_zones_cb */
449 sector_t start;
450 };
451 int dm_report_zones_cb(struct blk_zone *zone, unsigned int idx, void *data);
452 #endif /* CONFIG_BLK_DEV_ZONED */
453
454 /*
455 * Device mapper functions to parse and create devices specified by the
456 * parameter "dm-mod.create="
457 */
458 int __init dm_early_create(struct dm_ioctl *dmi,
459 struct dm_target_spec **spec_array,
460 char **target_params_array);
461
462 struct queue_limits *dm_get_queue_limits(struct mapped_device *md);
463
464 /*
465 * Geometry functions.
466 */
467 int dm_get_geometry(struct mapped_device *md, struct hd_geometry *geo);
468 int dm_set_geometry(struct mapped_device *md, struct hd_geometry *geo);
469
470 /*-----------------------------------------------------------------
471 * Functions for manipulating device-mapper tables.
472 *---------------------------------------------------------------*/
473
474 /*
475 * First create an empty table.
476 */
477 int dm_table_create(struct dm_table **result, fmode_t mode,
478 unsigned num_targets, struct mapped_device *md);
479
480 /*
481 * Then call this once for each target.
482 */
483 int dm_table_add_target(struct dm_table *t, const char *type,
484 sector_t start, sector_t len, char *params);
485
486 /*
487 * Target_ctr should call this if it needs to add any callbacks.
488 */
489 void dm_table_add_target_callbacks(struct dm_table *t, struct dm_target_callbacks *cb);
490
491 /*
492 * Target can use this to set the table's type.
493 * Can only ever be called from a target's ctr.
494 * Useful for "hybrid" target (supports both bio-based
495 * and request-based).
496 */
497 void dm_table_set_type(struct dm_table *t, enum dm_queue_mode type);
498
499 /*
500 * Finally call this to make the table ready for use.
501 */
502 int dm_table_complete(struct dm_table *t);
503
504 /*
505 * Destroy the table when finished.
506 */
507 void dm_table_destroy(struct dm_table *t);
508
509 /*
510 * Target may require that it is never sent I/O larger than len.
511 */
512 int __must_check dm_set_target_max_io_len(struct dm_target *ti, sector_t len);
513
514 /*
515 * Table reference counting.
516 */
517 struct dm_table *dm_get_live_table(struct mapped_device *md, int *srcu_idx);
518 void dm_put_live_table(struct mapped_device *md, int srcu_idx);
519 void dm_sync_table(struct mapped_device *md);
520
521 /*
522 * Queries
523 */
524 sector_t dm_table_get_size(struct dm_table *t);
525 unsigned int dm_table_get_num_targets(struct dm_table *t);
526 fmode_t dm_table_get_mode(struct dm_table *t);
527 struct mapped_device *dm_table_get_md(struct dm_table *t);
528 const char *dm_table_device_name(struct dm_table *t);
529
530 /*
531 * Trigger an event.
532 */
533 void dm_table_event(struct dm_table *t);
534
535 /*
536 * Run the queue for request-based targets.
537 */
538 void dm_table_run_md_queue_async(struct dm_table *t);
539
540 /*
541 * The device must be suspended before calling this method.
542 * Returns the previous table, which the caller must destroy.
543 */
544 struct dm_table *dm_swap_table(struct mapped_device *md,
545 struct dm_table *t);
546
547 /*
548 * A wrapper around vmalloc.
549 */
550 void *dm_vcalloc(unsigned long nmemb, unsigned long elem_size);
551
552 /*-----------------------------------------------------------------
553 * Macros.
554 *---------------------------------------------------------------*/
555 #define DM_NAME "device-mapper"
556
557 #define DM_FMT(fmt) DM_NAME ": " DM_MSG_PREFIX ": " fmt "\n"
558
559 #define DMCRIT(fmt, ...) pr_crit(DM_FMT(fmt), ##__VA_ARGS__)
560
561 #define DMERR(fmt, ...) pr_err(DM_FMT(fmt), ##__VA_ARGS__)
562 #define DMERR_LIMIT(fmt, ...) pr_err_ratelimited(DM_FMT(fmt), ##__VA_ARGS__)
563 #define DMWARN(fmt, ...) pr_warn(DM_FMT(fmt), ##__VA_ARGS__)
564 #define DMWARN_LIMIT(fmt, ...) pr_warn_ratelimited(DM_FMT(fmt), ##__VA_ARGS__)
565 #define DMINFO(fmt, ...) pr_info(DM_FMT(fmt), ##__VA_ARGS__)
566 #define DMINFO_LIMIT(fmt, ...) pr_info_ratelimited(DM_FMT(fmt), ##__VA_ARGS__)
567
568 #ifdef CONFIG_DM_DEBUG
569 #define DMDEBUG(fmt, ...) printk(KERN_DEBUG DM_FMT(fmt), ##__VA_ARGS__)
570 #define DMDEBUG_LIMIT(fmt, ...) pr_debug_ratelimited(DM_FMT(fmt), ##__VA_ARGS__)
571 #else
572 #define DMDEBUG(fmt, ...) no_printk(fmt, ##__VA_ARGS__)
573 #define DMDEBUG_LIMIT(fmt, ...) no_printk(fmt, ##__VA_ARGS__)
574 #endif
575
576 #define DMEMIT(x...) sz += ((sz >= maxlen) ? \
577 0 : scnprintf(result + sz, maxlen - sz, x))
578
579 /*
580 * Definitions of return values from target end_io function.
581 */
582 #define DM_ENDIO_DONE 0
583 #define DM_ENDIO_INCOMPLETE 1
584 #define DM_ENDIO_REQUEUE 2
585 #define DM_ENDIO_DELAY_REQUEUE 3
586
587 /*
588 * Definitions of return values from target map function.
589 */
590 #define DM_MAPIO_SUBMITTED 0
591 #define DM_MAPIO_REMAPPED 1
592 #define DM_MAPIO_REQUEUE DM_ENDIO_REQUEUE
593 #define DM_MAPIO_DELAY_REQUEUE DM_ENDIO_DELAY_REQUEUE
594 #define DM_MAPIO_KILL 4
595
596 #define dm_sector_div64(x, y)( \
597 { \
598 u64 _res; \
599 (x) = div64_u64_rem(x, y, &_res); \
600 _res; \
601 } \
602 )
603
604 /*
605 * Ceiling(n / sz)
606 */
607 #define dm_div_up(n, sz) (((n) + (sz) - 1) / (sz))
608
609 #define dm_sector_div_up(n, sz) ( \
610 { \
611 sector_t _r = ((n) + (sz) - 1); \
612 sector_div(_r, (sz)); \
613 _r; \
614 } \
615 )
616
617 /*
618 * ceiling(n / size) * size
619 */
620 #define dm_round_up(n, sz) (dm_div_up((n), (sz)) * (sz))
621
622 #define dm_array_too_big(fixed, obj, num) \
623 ((num) > (UINT_MAX - (fixed)) / (obj))
624
625 /*
626 * Sector offset taken relative to the start of the target instead of
627 * relative to the start of the device.
628 */
629 #define dm_target_offset(ti, sector) ((sector) - (ti)->begin)
630
to_sector(unsigned long long n)631 static inline sector_t to_sector(unsigned long long n)
632 {
633 return (n >> SECTOR_SHIFT);
634 }
635
to_bytes(sector_t n)636 static inline unsigned long to_bytes(sector_t n)
637 {
638 return (n << SECTOR_SHIFT);
639 }
640
641 #endif /* _LINUX_DEVICE_MAPPER_H */
642