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