at v4.19 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 DM_TYPE_NVME_BIO_BASED = 5, 32}; 33 34typedef enum { STATUSTYPE_INFO, STATUSTYPE_TABLE } status_type_t; 35 36union 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 */ 44typedef 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 */ 51typedef 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 */ 60typedef int (*dm_map_fn) (struct dm_target *ti, struct bio *bio); 61typedef 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); 65typedef void (*dm_release_clone_request_fn) (struct request *clone); 66 67/* 68 * Returns: 69 * < 0 : error (currently ignored) 70 * 0 : ended successfully 71 * 1 : for some reason the io has still not completed (eg, 72 * multipath target might want to requeue a failed io). 73 * 2 : The target wants to push back the io 74 */ 75typedef int (*dm_endio_fn) (struct dm_target *ti, 76 struct bio *bio, blk_status_t *error); 77typedef int (*dm_request_endio_fn) (struct dm_target *ti, 78 struct request *clone, blk_status_t error, 79 union map_info *map_context); 80 81typedef void (*dm_presuspend_fn) (struct dm_target *ti); 82typedef void (*dm_presuspend_undo_fn) (struct dm_target *ti); 83typedef void (*dm_postsuspend_fn) (struct dm_target *ti); 84typedef int (*dm_preresume_fn) (struct dm_target *ti); 85typedef void (*dm_resume_fn) (struct dm_target *ti); 86 87typedef void (*dm_status_fn) (struct dm_target *ti, status_type_t status_type, 88 unsigned status_flags, char *result, unsigned maxlen); 89 90typedef int (*dm_message_fn) (struct dm_target *ti, unsigned argc, char **argv, 91 char *result, unsigned maxlen); 92 93typedef int (*dm_prepare_ioctl_fn) (struct dm_target *ti, struct block_device **bdev); 94 95/* 96 * These iteration functions are typically used to check (and combine) 97 * properties of underlying devices. 98 * E.g. Does at least one underlying device support flush? 99 * Does any underlying device not support WRITE_SAME? 100 * 101 * The callout function is called once for each contiguous section of 102 * an underlying device. State can be maintained in *data. 103 * Return non-zero to stop iterating through any further devices. 104 */ 105typedef int (*iterate_devices_callout_fn) (struct dm_target *ti, 106 struct dm_dev *dev, 107 sector_t start, sector_t len, 108 void *data); 109 110/* 111 * This function must iterate through each section of device used by the 112 * target until it encounters a non-zero return code, which it then returns. 113 * Returns zero if no callout returned non-zero. 114 */ 115typedef int (*dm_iterate_devices_fn) (struct dm_target *ti, 116 iterate_devices_callout_fn fn, 117 void *data); 118 119typedef void (*dm_io_hints_fn) (struct dm_target *ti, 120 struct queue_limits *limits); 121 122/* 123 * Returns: 124 * 0: The target can handle the next I/O immediately. 125 * 1: The target can't handle the next I/O immediately. 126 */ 127typedef int (*dm_busy_fn) (struct dm_target *ti); 128 129/* 130 * Returns: 131 * < 0 : error 132 * >= 0 : the number of bytes accessible at the address 133 */ 134typedef long (*dm_dax_direct_access_fn) (struct dm_target *ti, pgoff_t pgoff, 135 long nr_pages, void **kaddr, pfn_t *pfn); 136typedef size_t (*dm_dax_copy_iter_fn)(struct dm_target *ti, pgoff_t pgoff, 137 void *addr, size_t bytes, struct iov_iter *i); 138#define PAGE_SECTORS (PAGE_SIZE / 512) 139 140void dm_error(const char *message); 141 142struct dm_dev { 143 struct block_device *bdev; 144 struct dax_device *dax_dev; 145 fmode_t mode; 146 char name[16]; 147}; 148 149dev_t dm_get_dev_t(const char *path); 150 151/* 152 * Constructors should call these functions to ensure destination devices 153 * are opened/closed correctly. 154 */ 155int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode, 156 struct dm_dev **result); 157void dm_put_device(struct dm_target *ti, struct dm_dev *d); 158 159/* 160 * Information about a target type 161 */ 162 163struct target_type { 164 uint64_t features; 165 const char *name; 166 struct module *module; 167 unsigned version[3]; 168 dm_ctr_fn ctr; 169 dm_dtr_fn dtr; 170 dm_map_fn map; 171 dm_clone_and_map_request_fn clone_and_map_rq; 172 dm_release_clone_request_fn release_clone_rq; 173 dm_endio_fn end_io; 174 dm_request_endio_fn rq_end_io; 175 dm_presuspend_fn presuspend; 176 dm_presuspend_undo_fn presuspend_undo; 177 dm_postsuspend_fn postsuspend; 178 dm_preresume_fn preresume; 179 dm_resume_fn resume; 180 dm_status_fn status; 181 dm_message_fn message; 182 dm_prepare_ioctl_fn prepare_ioctl; 183 dm_busy_fn busy; 184 dm_iterate_devices_fn iterate_devices; 185 dm_io_hints_fn io_hints; 186 dm_dax_direct_access_fn direct_access; 187 dm_dax_copy_iter_fn dax_copy_from_iter; 188 dm_dax_copy_iter_fn dax_copy_to_iter; 189 190 /* For internal device-mapper use. */ 191 struct list_head list; 192}; 193 194/* 195 * Target features 196 */ 197 198/* 199 * Any table that contains an instance of this target must have only one. 200 */ 201#define DM_TARGET_SINGLETON 0x00000001 202#define dm_target_needs_singleton(type) ((type)->features & DM_TARGET_SINGLETON) 203 204/* 205 * Indicates that a target does not support read-only devices. 206 */ 207#define DM_TARGET_ALWAYS_WRITEABLE 0x00000002 208#define dm_target_always_writeable(type) \ 209 ((type)->features & DM_TARGET_ALWAYS_WRITEABLE) 210 211/* 212 * Any device that contains a table with an instance of this target may never 213 * have tables containing any different target type. 214 */ 215#define DM_TARGET_IMMUTABLE 0x00000004 216#define dm_target_is_immutable(type) ((type)->features & DM_TARGET_IMMUTABLE) 217 218/* 219 * Indicates that a target may replace any target; even immutable targets. 220 * .map, .map_rq, .clone_and_map_rq and .release_clone_rq are all defined. 221 */ 222#define DM_TARGET_WILDCARD 0x00000008 223#define dm_target_is_wildcard(type) ((type)->features & DM_TARGET_WILDCARD) 224 225/* 226 * A target implements own bio data integrity. 227 */ 228#define DM_TARGET_INTEGRITY 0x00000010 229#define dm_target_has_integrity(type) ((type)->features & DM_TARGET_INTEGRITY) 230 231/* 232 * A target passes integrity data to the lower device. 233 */ 234#define DM_TARGET_PASSES_INTEGRITY 0x00000020 235#define dm_target_passes_integrity(type) ((type)->features & DM_TARGET_PASSES_INTEGRITY) 236 237/* 238 * Indicates that a target supports host-managed zoned block devices. 239 */ 240#define DM_TARGET_ZONED_HM 0x00000040 241#define dm_target_supports_zoned_hm(type) ((type)->features & DM_TARGET_ZONED_HM) 242 243struct dm_target { 244 struct dm_table *table; 245 struct target_type *type; 246 247 /* target limits */ 248 sector_t begin; 249 sector_t len; 250 251 /* If non-zero, maximum size of I/O submitted to a target. */ 252 uint32_t max_io_len; 253 254 /* 255 * A number of zero-length barrier bios that will be submitted 256 * to the target for the purpose of flushing cache. 257 * 258 * The bio number can be accessed with dm_bio_get_target_bio_nr. 259 * It is a responsibility of the target driver to remap these bios 260 * to the real underlying devices. 261 */ 262 unsigned num_flush_bios; 263 264 /* 265 * The number of discard bios that will be submitted to the target. 266 * The bio number can be accessed with dm_bio_get_target_bio_nr. 267 */ 268 unsigned num_discard_bios; 269 270 /* 271 * The number of secure erase bios that will be submitted to the target. 272 * The bio number can be accessed with dm_bio_get_target_bio_nr. 273 */ 274 unsigned num_secure_erase_bios; 275 276 /* 277 * The number of WRITE SAME 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_write_same_bios; 281 282 /* 283 * The number of WRITE ZEROES 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_write_zeroes_bios; 287 288 /* 289 * The minimum number of extra bytes allocated in each io for the 290 * target to use. 291 */ 292 unsigned per_io_data_size; 293 294 /* target specific data */ 295 void *private; 296 297 /* Used to provide an error string from the ctr */ 298 char *error; 299 300 /* 301 * Set if this target needs to receive flushes regardless of 302 * whether or not its underlying devices have support. 303 */ 304 bool flush_supported:1; 305 306 /* 307 * Set if this target needs to receive discards regardless of 308 * whether or not its underlying devices have support. 309 */ 310 bool discards_supported:1; 311 312 /* 313 * Set if the target required discard bios to be split 314 * on max_io_len boundary. 315 */ 316 bool split_discard_bios:1; 317}; 318 319/* Each target can link one of these into the table */ 320struct dm_target_callbacks { 321 struct list_head list; 322 int (*congested_fn) (struct dm_target_callbacks *, int); 323}; 324 325void *dm_per_bio_data(struct bio *bio, size_t data_size); 326struct bio *dm_bio_from_per_bio_data(void *data, size_t data_size); 327unsigned dm_bio_get_target_bio_nr(const struct bio *bio); 328 329int dm_register_target(struct target_type *t); 330void dm_unregister_target(struct target_type *t); 331 332/* 333 * Target argument parsing. 334 */ 335struct dm_arg_set { 336 unsigned argc; 337 char **argv; 338}; 339 340/* 341 * The minimum and maximum value of a numeric argument, together with 342 * the error message to use if the number is found to be outside that range. 343 */ 344struct dm_arg { 345 unsigned min; 346 unsigned max; 347 char *error; 348}; 349 350/* 351 * Validate the next argument, either returning it as *value or, if invalid, 352 * returning -EINVAL and setting *error. 353 */ 354int dm_read_arg(const struct dm_arg *arg, struct dm_arg_set *arg_set, 355 unsigned *value, char **error); 356 357/* 358 * Process the next argument as the start of a group containing between 359 * arg->min and arg->max further arguments. Either return the size as 360 * *num_args or, if invalid, return -EINVAL and set *error. 361 */ 362int dm_read_arg_group(const struct dm_arg *arg, struct dm_arg_set *arg_set, 363 unsigned *num_args, char **error); 364 365/* 366 * Return the current argument and shift to the next. 367 */ 368const char *dm_shift_arg(struct dm_arg_set *as); 369 370/* 371 * Move through num_args arguments. 372 */ 373void dm_consume_args(struct dm_arg_set *as, unsigned num_args); 374 375/*----------------------------------------------------------------- 376 * Functions for creating and manipulating mapped devices. 377 * Drop the reference with dm_put when you finish with the object. 378 *---------------------------------------------------------------*/ 379 380/* 381 * DM_ANY_MINOR chooses the next available minor number. 382 */ 383#define DM_ANY_MINOR (-1) 384int dm_create(int minor, struct mapped_device **md); 385 386/* 387 * Reference counting for md. 388 */ 389struct mapped_device *dm_get_md(dev_t dev); 390void dm_get(struct mapped_device *md); 391int dm_hold(struct mapped_device *md); 392void dm_put(struct mapped_device *md); 393 394/* 395 * An arbitrary pointer may be stored alongside a mapped device. 396 */ 397void dm_set_mdptr(struct mapped_device *md, void *ptr); 398void *dm_get_mdptr(struct mapped_device *md); 399 400/* 401 * A device can still be used while suspended, but I/O is deferred. 402 */ 403int dm_suspend(struct mapped_device *md, unsigned suspend_flags); 404int dm_resume(struct mapped_device *md); 405 406/* 407 * Event functions. 408 */ 409uint32_t dm_get_event_nr(struct mapped_device *md); 410int dm_wait_event(struct mapped_device *md, int event_nr); 411uint32_t dm_next_uevent_seq(struct mapped_device *md); 412void dm_uevent_add(struct mapped_device *md, struct list_head *elist); 413 414/* 415 * Info functions. 416 */ 417const char *dm_device_name(struct mapped_device *md); 418int dm_copy_name_and_uuid(struct mapped_device *md, char *name, char *uuid); 419struct gendisk *dm_disk(struct mapped_device *md); 420int dm_suspended(struct dm_target *ti); 421int dm_noflush_suspending(struct dm_target *ti); 422void dm_accept_partial_bio(struct bio *bio, unsigned n_sectors); 423void dm_remap_zone_report(struct dm_target *ti, struct bio *bio, 424 sector_t start); 425union map_info *dm_get_rq_mapinfo(struct request *rq); 426 427struct queue_limits *dm_get_queue_limits(struct mapped_device *md); 428 429/* 430 * Geometry functions. 431 */ 432int dm_get_geometry(struct mapped_device *md, struct hd_geometry *geo); 433int dm_set_geometry(struct mapped_device *md, struct hd_geometry *geo); 434 435/*----------------------------------------------------------------- 436 * Functions for manipulating device-mapper tables. 437 *---------------------------------------------------------------*/ 438 439/* 440 * First create an empty table. 441 */ 442int dm_table_create(struct dm_table **result, fmode_t mode, 443 unsigned num_targets, struct mapped_device *md); 444 445/* 446 * Then call this once for each target. 447 */ 448int dm_table_add_target(struct dm_table *t, const char *type, 449 sector_t start, sector_t len, char *params); 450 451/* 452 * Target_ctr should call this if it needs to add any callbacks. 453 */ 454void dm_table_add_target_callbacks(struct dm_table *t, struct dm_target_callbacks *cb); 455 456/* 457 * Target can use this to set the table's type. 458 * Can only ever be called from a target's ctr. 459 * Useful for "hybrid" target (supports both bio-based 460 * and request-based). 461 */ 462void dm_table_set_type(struct dm_table *t, enum dm_queue_mode type); 463 464/* 465 * Finally call this to make the table ready for use. 466 */ 467int dm_table_complete(struct dm_table *t); 468 469/* 470 * Destroy the table when finished. 471 */ 472void dm_table_destroy(struct dm_table *t); 473 474/* 475 * Target may require that it is never sent I/O larger than len. 476 */ 477int __must_check dm_set_target_max_io_len(struct dm_target *ti, sector_t len); 478 479/* 480 * Table reference counting. 481 */ 482struct dm_table *dm_get_live_table(struct mapped_device *md, int *srcu_idx); 483void dm_put_live_table(struct mapped_device *md, int srcu_idx); 484void dm_sync_table(struct mapped_device *md); 485 486/* 487 * Queries 488 */ 489sector_t dm_table_get_size(struct dm_table *t); 490unsigned int dm_table_get_num_targets(struct dm_table *t); 491fmode_t dm_table_get_mode(struct dm_table *t); 492struct mapped_device *dm_table_get_md(struct dm_table *t); 493 494/* 495 * Trigger an event. 496 */ 497void dm_table_event(struct dm_table *t); 498 499/* 500 * Run the queue for request-based targets. 501 */ 502void dm_table_run_md_queue_async(struct dm_table *t); 503 504/* 505 * The device must be suspended before calling this method. 506 * Returns the previous table, which the caller must destroy. 507 */ 508struct dm_table *dm_swap_table(struct mapped_device *md, 509 struct dm_table *t); 510 511/* 512 * A wrapper around vmalloc. 513 */ 514void *dm_vcalloc(unsigned long nmemb, unsigned long elem_size); 515 516/*----------------------------------------------------------------- 517 * Macros. 518 *---------------------------------------------------------------*/ 519#define DM_NAME "device-mapper" 520 521#define DM_RATELIMIT(pr_func, fmt, ...) \ 522do { \ 523 static DEFINE_RATELIMIT_STATE(rs, DEFAULT_RATELIMIT_INTERVAL, \ 524 DEFAULT_RATELIMIT_BURST); \ 525 \ 526 if (__ratelimit(&rs)) \ 527 pr_func(DM_FMT(fmt), ##__VA_ARGS__); \ 528} while (0) 529 530#define DM_FMT(fmt) DM_NAME ": " DM_MSG_PREFIX ": " fmt "\n" 531 532#define DMCRIT(fmt, ...) pr_crit(DM_FMT(fmt), ##__VA_ARGS__) 533 534#define DMERR(fmt, ...) pr_err(DM_FMT(fmt), ##__VA_ARGS__) 535#define DMERR_LIMIT(fmt, ...) DM_RATELIMIT(pr_err, fmt, ##__VA_ARGS__) 536#define DMWARN(fmt, ...) pr_warn(DM_FMT(fmt), ##__VA_ARGS__) 537#define DMWARN_LIMIT(fmt, ...) DM_RATELIMIT(pr_warn, fmt, ##__VA_ARGS__) 538#define DMINFO(fmt, ...) pr_info(DM_FMT(fmt), ##__VA_ARGS__) 539#define DMINFO_LIMIT(fmt, ...) DM_RATELIMIT(pr_info, fmt, ##__VA_ARGS__) 540 541#ifdef CONFIG_DM_DEBUG 542#define DMDEBUG(fmt, ...) printk(KERN_DEBUG DM_FMT(fmt), ##__VA_ARGS__) 543#define DMDEBUG_LIMIT(fmt, ...) DM_RATELIMIT(pr_debug, fmt, ##__VA_ARGS__) 544#else 545#define DMDEBUG(fmt, ...) no_printk(fmt, ##__VA_ARGS__) 546#define DMDEBUG_LIMIT(fmt, ...) no_printk(fmt, ##__VA_ARGS__) 547#endif 548 549#define DMEMIT(x...) sz += ((sz >= maxlen) ? \ 550 0 : scnprintf(result + sz, maxlen - sz, x)) 551 552/* 553 * Definitions of return values from target end_io function. 554 */ 555#define DM_ENDIO_DONE 0 556#define DM_ENDIO_INCOMPLETE 1 557#define DM_ENDIO_REQUEUE 2 558#define DM_ENDIO_DELAY_REQUEUE 3 559 560/* 561 * Definitions of return values from target map function. 562 */ 563#define DM_MAPIO_SUBMITTED 0 564#define DM_MAPIO_REMAPPED 1 565#define DM_MAPIO_REQUEUE DM_ENDIO_REQUEUE 566#define DM_MAPIO_DELAY_REQUEUE DM_ENDIO_DELAY_REQUEUE 567#define DM_MAPIO_KILL 4 568 569#define dm_sector_div64(x, y)( \ 570{ \ 571 u64 _res; \ 572 (x) = div64_u64_rem(x, y, &_res); \ 573 _res; \ 574} \ 575) 576 577/* 578 * Ceiling(n / sz) 579 */ 580#define dm_div_up(n, sz) (((n) + (sz) - 1) / (sz)) 581 582#define dm_sector_div_up(n, sz) ( \ 583{ \ 584 sector_t _r = ((n) + (sz) - 1); \ 585 sector_div(_r, (sz)); \ 586 _r; \ 587} \ 588) 589 590/* 591 * ceiling(n / size) * size 592 */ 593#define dm_round_up(n, sz) (dm_div_up((n), (sz)) * (sz)) 594 595#define dm_array_too_big(fixed, obj, num) \ 596 ((num) > (UINT_MAX - (fixed)) / (obj)) 597 598/* 599 * Sector offset taken relative to the start of the target instead of 600 * relative to the start of the device. 601 */ 602#define dm_target_offset(ti, sector) ((sector) - (ti)->begin) 603 604static inline sector_t to_sector(unsigned long n) 605{ 606 return (n >> SECTOR_SHIFT); 607} 608 609static inline unsigned long to_bytes(sector_t n) 610{ 611 return (n << SECTOR_SHIFT); 612} 613 614#endif /* _LINUX_DEVICE_MAPPER_H */