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