at v4.13 18 kB view raw
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/math64.h> 14#include <linux/ratelimit.h> 15 16struct dm_dev; 17struct dm_target; 18struct dm_table; 19struct mapped_device; 20struct bio_vec; 21 22/* 23 * Type of table, mapped_device's mempool and request_queue 24 */ 25enum dm_queue_mode { 26 DM_TYPE_NONE = 0, 27 DM_TYPE_BIO_BASED = 1, 28 DM_TYPE_REQUEST_BASED = 2, 29 DM_TYPE_MQ_REQUEST_BASED = 3, 30 DM_TYPE_DAX_BIO_BASED = 4, 31}; 32 33typedef enum { STATUSTYPE_INFO, STATUSTYPE_TABLE } status_type_t; 34 35union map_info { 36 void *ptr; 37}; 38 39/* 40 * In the constructor the target parameter will already have the 41 * table, type, begin and len fields filled in. 42 */ 43typedef int (*dm_ctr_fn) (struct dm_target *target, 44 unsigned int argc, char **argv); 45 46/* 47 * The destructor doesn't need to free the dm_target, just 48 * anything hidden ti->private. 49 */ 50typedef void (*dm_dtr_fn) (struct dm_target *ti); 51 52/* 53 * The map function must return: 54 * < 0: error 55 * = 0: The target will handle the io by resubmitting it later 56 * = 1: simple remap complete 57 * = 2: The target wants to push back the io 58 */ 59typedef int (*dm_map_fn) (struct dm_target *ti, struct bio *bio); 60typedef int (*dm_clone_and_map_request_fn) (struct dm_target *ti, 61 struct request *rq, 62 union map_info *map_context, 63 struct request **clone); 64typedef void (*dm_release_clone_request_fn) (struct request *clone); 65 66/* 67 * Returns: 68 * < 0 : error (currently ignored) 69 * 0 : ended successfully 70 * 1 : for some reason the io has still not completed (eg, 71 * multipath target might want to requeue a failed io). 72 * 2 : The target wants to push back the io 73 */ 74typedef int (*dm_endio_fn) (struct dm_target *ti, 75 struct bio *bio, blk_status_t *error); 76typedef int (*dm_request_endio_fn) (struct dm_target *ti, 77 struct request *clone, blk_status_t error, 78 union map_info *map_context); 79 80typedef void (*dm_presuspend_fn) (struct dm_target *ti); 81typedef void (*dm_presuspend_undo_fn) (struct dm_target *ti); 82typedef void (*dm_postsuspend_fn) (struct dm_target *ti); 83typedef int (*dm_preresume_fn) (struct dm_target *ti); 84typedef void (*dm_resume_fn) (struct dm_target *ti); 85 86typedef void (*dm_status_fn) (struct dm_target *ti, status_type_t status_type, 87 unsigned status_flags, char *result, unsigned maxlen); 88 89typedef int (*dm_message_fn) (struct dm_target *ti, unsigned argc, char **argv); 90 91typedef int (*dm_prepare_ioctl_fn) (struct dm_target *ti, 92 struct block_device **bdev, fmode_t *mode); 93 94/* 95 * These iteration functions are typically used to check (and combine) 96 * properties of underlying devices. 97 * E.g. Does at least one underlying device support flush? 98 * Does any underlying device not support WRITE_SAME? 99 * 100 * The callout function is called once for each contiguous section of 101 * an underlying device. State can be maintained in *data. 102 * Return non-zero to stop iterating through any further devices. 103 */ 104typedef int (*iterate_devices_callout_fn) (struct dm_target *ti, 105 struct dm_dev *dev, 106 sector_t start, sector_t len, 107 void *data); 108 109/* 110 * This function must iterate through each section of device used by the 111 * target until it encounters a non-zero return code, which it then returns. 112 * Returns zero if no callout returned non-zero. 113 */ 114typedef int (*dm_iterate_devices_fn) (struct dm_target *ti, 115 iterate_devices_callout_fn fn, 116 void *data); 117 118typedef void (*dm_io_hints_fn) (struct dm_target *ti, 119 struct queue_limits *limits); 120 121/* 122 * Returns: 123 * 0: The target can handle the next I/O immediately. 124 * 1: The target can't handle the next I/O immediately. 125 */ 126typedef int (*dm_busy_fn) (struct dm_target *ti); 127 128/* 129 * Returns: 130 * < 0 : error 131 * >= 0 : the number of bytes accessible at the address 132 */ 133typedef long (*dm_dax_direct_access_fn) (struct dm_target *ti, pgoff_t pgoff, 134 long nr_pages, void **kaddr, pfn_t *pfn); 135typedef size_t (*dm_dax_copy_from_iter_fn)(struct dm_target *ti, pgoff_t pgoff, 136 void *addr, size_t bytes, struct iov_iter *i); 137typedef void (*dm_dax_flush_fn)(struct dm_target *ti, pgoff_t pgoff, void *addr, 138 size_t size); 139#define PAGE_SECTORS (PAGE_SIZE / 512) 140 141void dm_error(const char *message); 142 143struct dm_dev { 144 struct block_device *bdev; 145 struct dax_device *dax_dev; 146 fmode_t mode; 147 char name[16]; 148}; 149 150dev_t dm_get_dev_t(const char *path); 151 152/* 153 * Constructors should call these functions to ensure destination devices 154 * are opened/closed correctly. 155 */ 156int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode, 157 struct dm_dev **result); 158void dm_put_device(struct dm_target *ti, struct dm_dev *d); 159 160/* 161 * Information about a target type 162 */ 163 164struct target_type { 165 uint64_t features; 166 const char *name; 167 struct module *module; 168 unsigned version[3]; 169 dm_ctr_fn ctr; 170 dm_dtr_fn dtr; 171 dm_map_fn map; 172 dm_clone_and_map_request_fn clone_and_map_rq; 173 dm_release_clone_request_fn release_clone_rq; 174 dm_endio_fn end_io; 175 dm_request_endio_fn rq_end_io; 176 dm_presuspend_fn presuspend; 177 dm_presuspend_undo_fn presuspend_undo; 178 dm_postsuspend_fn postsuspend; 179 dm_preresume_fn preresume; 180 dm_resume_fn resume; 181 dm_status_fn status; 182 dm_message_fn message; 183 dm_prepare_ioctl_fn prepare_ioctl; 184 dm_busy_fn busy; 185 dm_iterate_devices_fn iterate_devices; 186 dm_io_hints_fn io_hints; 187 dm_dax_direct_access_fn direct_access; 188 dm_dax_copy_from_iter_fn dax_copy_from_iter; 189 dm_dax_flush_fn dax_flush; 190 191 /* For internal device-mapper use. */ 192 struct list_head list; 193}; 194 195/* 196 * Target features 197 */ 198 199/* 200 * Any table that contains an instance of this target must have only one. 201 */ 202#define DM_TARGET_SINGLETON 0x00000001 203#define dm_target_needs_singleton(type) ((type)->features & DM_TARGET_SINGLETON) 204 205/* 206 * Indicates that a target does not support read-only devices. 207 */ 208#define DM_TARGET_ALWAYS_WRITEABLE 0x00000002 209#define dm_target_always_writeable(type) \ 210 ((type)->features & DM_TARGET_ALWAYS_WRITEABLE) 211 212/* 213 * Any device that contains a table with an instance of this target may never 214 * have tables containing any different target type. 215 */ 216#define DM_TARGET_IMMUTABLE 0x00000004 217#define dm_target_is_immutable(type) ((type)->features & DM_TARGET_IMMUTABLE) 218 219/* 220 * Indicates that a target may replace any target; even immutable targets. 221 * .map, .map_rq, .clone_and_map_rq and .release_clone_rq are all defined. 222 */ 223#define DM_TARGET_WILDCARD 0x00000008 224#define dm_target_is_wildcard(type) ((type)->features & DM_TARGET_WILDCARD) 225 226/* 227 * Some targets need to be sent the same WRITE bio severals times so 228 * that they can send copies of it to different devices. This function 229 * examines any supplied bio and returns the number of copies of it the 230 * target requires. 231 */ 232typedef unsigned (*dm_num_write_bios_fn) (struct dm_target *ti, struct bio *bio); 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 252struct 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 WRITE SAME 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_write_same_bios; 284 285 /* 286 * The number of WRITE ZEROES 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_zeroes_bios; 290 291 /* 292 * The minimum number of extra bytes allocated in each io for the 293 * target to use. 294 */ 295 unsigned per_io_data_size; 296 297 /* 298 * If defined, this function is called to find out how many 299 * duplicate bios should be sent to the target when writing 300 * data. 301 */ 302 dm_num_write_bios_fn num_write_bios; 303 304 /* target specific data */ 305 void *private; 306 307 /* Used to provide an error string from the ctr */ 308 char *error; 309 310 /* 311 * Set if this target needs to receive flushes regardless of 312 * whether or not its underlying devices have support. 313 */ 314 bool flush_supported:1; 315 316 /* 317 * Set if this target needs to receive discards regardless of 318 * whether or not its underlying devices have support. 319 */ 320 bool discards_supported:1; 321 322 /* 323 * Set if the target required discard bios to be split 324 * on max_io_len boundary. 325 */ 326 bool split_discard_bios:1; 327}; 328 329/* Each target can link one of these into the table */ 330struct dm_target_callbacks { 331 struct list_head list; 332 int (*congested_fn) (struct dm_target_callbacks *, int); 333}; 334 335/* 336 * For bio-based dm. 337 * One of these is allocated for each bio. 338 * This structure shouldn't be touched directly by target drivers. 339 * It is here so that we can inline dm_per_bio_data and 340 * dm_bio_from_per_bio_data 341 */ 342struct dm_target_io { 343 struct dm_io *io; 344 struct dm_target *ti; 345 unsigned target_bio_nr; 346 unsigned *len_ptr; 347 struct bio clone; 348}; 349 350static inline void *dm_per_bio_data(struct bio *bio, size_t data_size) 351{ 352 return (char *)bio - offsetof(struct dm_target_io, clone) - data_size; 353} 354 355static inline struct bio *dm_bio_from_per_bio_data(void *data, size_t data_size) 356{ 357 return (struct bio *)((char *)data + data_size + offsetof(struct dm_target_io, clone)); 358} 359 360static inline unsigned dm_bio_get_target_bio_nr(const struct bio *bio) 361{ 362 return container_of(bio, struct dm_target_io, clone)->target_bio_nr; 363} 364 365int dm_register_target(struct target_type *t); 366void dm_unregister_target(struct target_type *t); 367 368/* 369 * Target argument parsing. 370 */ 371struct dm_arg_set { 372 unsigned argc; 373 char **argv; 374}; 375 376/* 377 * The minimum and maximum value of a numeric argument, together with 378 * the error message to use if the number is found to be outside that range. 379 */ 380struct dm_arg { 381 unsigned min; 382 unsigned max; 383 char *error; 384}; 385 386/* 387 * Validate the next argument, either returning it as *value or, if invalid, 388 * returning -EINVAL and setting *error. 389 */ 390int dm_read_arg(struct dm_arg *arg, struct dm_arg_set *arg_set, 391 unsigned *value, char **error); 392 393/* 394 * Process the next argument as the start of a group containing between 395 * arg->min and arg->max further arguments. Either return the size as 396 * *num_args or, if invalid, return -EINVAL and set *error. 397 */ 398int dm_read_arg_group(struct dm_arg *arg, struct dm_arg_set *arg_set, 399 unsigned *num_args, char **error); 400 401/* 402 * Return the current argument and shift to the next. 403 */ 404const char *dm_shift_arg(struct dm_arg_set *as); 405 406/* 407 * Move through num_args arguments. 408 */ 409void dm_consume_args(struct dm_arg_set *as, unsigned num_args); 410 411/*----------------------------------------------------------------- 412 * Functions for creating and manipulating mapped devices. 413 * Drop the reference with dm_put when you finish with the object. 414 *---------------------------------------------------------------*/ 415 416/* 417 * DM_ANY_MINOR chooses the next available minor number. 418 */ 419#define DM_ANY_MINOR (-1) 420int dm_create(int minor, struct mapped_device **md); 421 422/* 423 * Reference counting for md. 424 */ 425struct mapped_device *dm_get_md(dev_t dev); 426void dm_get(struct mapped_device *md); 427int dm_hold(struct mapped_device *md); 428void dm_put(struct mapped_device *md); 429 430/* 431 * An arbitrary pointer may be stored alongside a mapped device. 432 */ 433void dm_set_mdptr(struct mapped_device *md, void *ptr); 434void *dm_get_mdptr(struct mapped_device *md); 435 436/* 437 * A device can still be used while suspended, but I/O is deferred. 438 */ 439int dm_suspend(struct mapped_device *md, unsigned suspend_flags); 440int dm_resume(struct mapped_device *md); 441 442/* 443 * Event functions. 444 */ 445uint32_t dm_get_event_nr(struct mapped_device *md); 446int dm_wait_event(struct mapped_device *md, int event_nr); 447uint32_t dm_next_uevent_seq(struct mapped_device *md); 448void dm_uevent_add(struct mapped_device *md, struct list_head *elist); 449 450/* 451 * Info functions. 452 */ 453const char *dm_device_name(struct mapped_device *md); 454int dm_copy_name_and_uuid(struct mapped_device *md, char *name, char *uuid); 455struct gendisk *dm_disk(struct mapped_device *md); 456int dm_suspended(struct dm_target *ti); 457int dm_noflush_suspending(struct dm_target *ti); 458void dm_accept_partial_bio(struct bio *bio, unsigned n_sectors); 459void dm_remap_zone_report(struct dm_target *ti, struct bio *bio, 460 sector_t start); 461union map_info *dm_get_rq_mapinfo(struct request *rq); 462 463struct queue_limits *dm_get_queue_limits(struct mapped_device *md); 464 465/* 466 * Geometry functions. 467 */ 468int dm_get_geometry(struct mapped_device *md, struct hd_geometry *geo); 469int dm_set_geometry(struct mapped_device *md, struct hd_geometry *geo); 470 471/*----------------------------------------------------------------- 472 * Functions for manipulating device-mapper tables. 473 *---------------------------------------------------------------*/ 474 475/* 476 * First create an empty table. 477 */ 478int dm_table_create(struct dm_table **result, fmode_t mode, 479 unsigned num_targets, struct mapped_device *md); 480 481/* 482 * Then call this once for each target. 483 */ 484int dm_table_add_target(struct dm_table *t, const char *type, 485 sector_t start, sector_t len, char *params); 486 487/* 488 * Target_ctr should call this if it needs to add any callbacks. 489 */ 490void dm_table_add_target_callbacks(struct dm_table *t, struct dm_target_callbacks *cb); 491 492/* 493 * Target can use this to set the table's type. 494 * Can only ever be called from a target's ctr. 495 * Useful for "hybrid" target (supports both bio-based 496 * and request-based). 497 */ 498void dm_table_set_type(struct dm_table *t, enum dm_queue_mode type); 499 500/* 501 * Finally call this to make the table ready for use. 502 */ 503int dm_table_complete(struct dm_table *t); 504 505/* 506 * Target may require that it is never sent I/O larger than len. 507 */ 508int __must_check dm_set_target_max_io_len(struct dm_target *ti, sector_t len); 509 510/* 511 * Table reference counting. 512 */ 513struct dm_table *dm_get_live_table(struct mapped_device *md, int *srcu_idx); 514void dm_put_live_table(struct mapped_device *md, int srcu_idx); 515void dm_sync_table(struct mapped_device *md); 516 517/* 518 * Queries 519 */ 520sector_t dm_table_get_size(struct dm_table *t); 521unsigned int dm_table_get_num_targets(struct dm_table *t); 522fmode_t dm_table_get_mode(struct dm_table *t); 523struct mapped_device *dm_table_get_md(struct dm_table *t); 524 525/* 526 * Trigger an event. 527 */ 528void dm_table_event(struct dm_table *t); 529 530/* 531 * Run the queue for request-based targets. 532 */ 533void 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 */ 539struct dm_table *dm_swap_table(struct mapped_device *md, 540 struct dm_table *t); 541 542/* 543 * A wrapper around vmalloc. 544 */ 545void *dm_vcalloc(unsigned long nmemb, unsigned long elem_size); 546 547/*----------------------------------------------------------------- 548 * Macros. 549 *---------------------------------------------------------------*/ 550#define DM_NAME "device-mapper" 551 552#define DM_RATELIMIT(pr_func, fmt, ...) \ 553do { \ 554 static DEFINE_RATELIMIT_STATE(rs, DEFAULT_RATELIMIT_INTERVAL, \ 555 DEFAULT_RATELIMIT_BURST); \ 556 \ 557 if (__ratelimit(&rs)) \ 558 pr_func(DM_FMT(fmt), ##__VA_ARGS__); \ 559} while (0) 560 561#define DM_FMT(fmt) DM_NAME ": " DM_MSG_PREFIX ": " fmt "\n" 562 563#define DMCRIT(fmt, ...) pr_crit(DM_FMT(fmt), ##__VA_ARGS__) 564 565#define DMERR(fmt, ...) pr_err(DM_FMT(fmt), ##__VA_ARGS__) 566#define DMERR_LIMIT(fmt, ...) DM_RATELIMIT(pr_err, fmt, ##__VA_ARGS__) 567#define DMWARN(fmt, ...) pr_warn(DM_FMT(fmt), ##__VA_ARGS__) 568#define DMWARN_LIMIT(fmt, ...) DM_RATELIMIT(pr_warn, fmt, ##__VA_ARGS__) 569#define DMINFO(fmt, ...) pr_info(DM_FMT(fmt), ##__VA_ARGS__) 570#define DMINFO_LIMIT(fmt, ...) DM_RATELIMIT(pr_info, fmt, ##__VA_ARGS__) 571 572#ifdef CONFIG_DM_DEBUG 573#define DMDEBUG(fmt, ...) printk(KERN_DEBUG DM_FMT(fmt), ##__VA_ARGS__) 574#define DMDEBUG_LIMIT(fmt, ...) DM_RATELIMIT(pr_debug, fmt, ##__VA_ARGS__) 575#else 576#define DMDEBUG(fmt, ...) no_printk(fmt, ##__VA_ARGS__) 577#define DMDEBUG_LIMIT(fmt, ...) no_printk(fmt, ##__VA_ARGS__) 578#endif 579 580#define DMEMIT(x...) sz += ((sz >= maxlen) ? \ 581 0 : scnprintf(result + sz, maxlen - sz, x)) 582 583#define SECTOR_SHIFT 9 584 585/* 586 * Definitions of return values from target end_io function. 587 */ 588#define DM_ENDIO_DONE 0 589#define DM_ENDIO_INCOMPLETE 1 590#define DM_ENDIO_REQUEUE 2 591 592/* 593 * Definitions of return values from target map function. 594 */ 595#define DM_MAPIO_SUBMITTED 0 596#define DM_MAPIO_REMAPPED 1 597#define DM_MAPIO_REQUEUE DM_ENDIO_REQUEUE 598#define DM_MAPIO_DELAY_REQUEUE 3 599#define DM_MAPIO_KILL 4 600 601#define dm_sector_div64(x, y)( \ 602{ \ 603 u64 _res; \ 604 (x) = div64_u64_rem(x, y, &_res); \ 605 _res; \ 606} \ 607) 608 609/* 610 * Ceiling(n / sz) 611 */ 612#define dm_div_up(n, sz) (((n) + (sz) - 1) / (sz)) 613 614#define dm_sector_div_up(n, sz) ( \ 615{ \ 616 sector_t _r = ((n) + (sz) - 1); \ 617 sector_div(_r, (sz)); \ 618 _r; \ 619} \ 620) 621 622/* 623 * ceiling(n / size) * size 624 */ 625#define dm_round_up(n, sz) (dm_div_up((n), (sz)) * (sz)) 626 627#define dm_array_too_big(fixed, obj, num) \ 628 ((num) > (UINT_MAX - (fixed)) / (obj)) 629 630/* 631 * Sector offset taken relative to the start of the target instead of 632 * relative to the start of the device. 633 */ 634#define dm_target_offset(ti, sector) ((sector) - (ti)->begin) 635 636static inline sector_t to_sector(unsigned long n) 637{ 638 return (n >> SECTOR_SHIFT); 639} 640 641static inline unsigned long to_bytes(sector_t n) 642{ 643 return (n << SECTOR_SHIFT); 644} 645 646#endif /* _LINUX_DEVICE_MAPPER_H */