1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 md.h : kernel internal structure of the Linux MD driver 4 Copyright (C) 1996-98 Ingo Molnar, Gadi Oxman 5 6 */ 7 8 #ifndef _MD_MD_H 9 #define _MD_MD_H 10 11 #include <linux/blkdev.h> 12 #include <linux/backing-dev.h> 13 #include <linux/badblocks.h> 14 #include <linux/kobject.h> 15 #include <linux/list.h> 16 #include <linux/mm.h> 17 #include <linux/mutex.h> 18 #include <linux/timer.h> 19 #include <linux/wait.h> 20 #include <linux/workqueue.h> 21 #include "md-cluster.h" 22 23 #define MaxSector (~(sector_t)0) 24 25 /* 26 * These flags should really be called "NO_RETRY" rather than 27 * "FAILFAST" because they don't make any promise about time lapse, 28 * only about the number of retries, which will be zero. 29 * REQ_FAILFAST_DRIVER is not included because 30 * Commit: 4a27446f3e39 ("[SCSI] modify scsi to handle new fail fast flags.") 31 * seems to suggest that the errors it avoids retrying should usually 32 * be retried. 33 */ 34 #define MD_FAILFAST (REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT) 35 36 /* 37 * The struct embedded in rdev is used to serialize IO. 38 */ 39 struct serial_in_rdev { 40 struct rb_root_cached serial_rb; 41 spinlock_t serial_lock; 42 wait_queue_head_t serial_io_wait; 43 }; 44 45 /* 46 * MD's 'extended' device 47 */ 48 struct md_rdev { 49 struct list_head same_set; /* RAID devices within the same set */ 50 51 sector_t sectors; /* Device size (in 512bytes sectors) */ 52 struct mddev *mddev; /* RAID array if running */ 53 int last_events; /* IO event timestamp */ 54 55 /* 56 * If meta_bdev is non-NULL, it means that a separate device is 57 * being used to store the metadata (superblock/bitmap) which 58 * would otherwise be contained on the same device as the data (bdev). 59 */ 60 struct block_device *meta_bdev; 61 struct block_device *bdev; /* block device handle */ 62 struct bdev_handle *bdev_handle; /* Handle from open for bdev */ 63 64 struct page *sb_page, *bb_page; 65 int sb_loaded; 66 __u64 sb_events; 67 sector_t data_offset; /* start of data in array */ 68 sector_t new_data_offset;/* only relevant while reshaping */ 69 sector_t sb_start; /* offset of the super block (in 512byte sectors) */ 70 int sb_size; /* bytes in the superblock */ 71 int preferred_minor; /* autorun support */ 72 73 struct kobject kobj; 74 75 /* A device can be in one of three states based on two flags: 76 * Not working: faulty==1 in_sync==0 77 * Fully working: faulty==0 in_sync==1 78 * Working, but not 79 * in sync with array 80 * faulty==0 in_sync==0 81 * 82 * It can never have faulty==1, in_sync==1 83 * This reduces the burden of testing multiple flags in many cases 84 */ 85 86 unsigned long flags; /* bit set of 'enum flag_bits' bits. */ 87 wait_queue_head_t blocked_wait; 88 89 int desc_nr; /* descriptor index in the superblock */ 90 int raid_disk; /* role of device in array */ 91 int new_raid_disk; /* role that the device will have in 92 * the array after a level-change completes. 93 */ 94 int saved_raid_disk; /* role that device used to have in the 95 * array and could again if we did a partial 96 * resync from the bitmap 97 */ 98 union { 99 sector_t recovery_offset;/* If this device has been partially 100 * recovered, this is where we were 101 * up to. 102 */ 103 sector_t journal_tail; /* If this device is a journal device, 104 * this is the journal tail (journal 105 * recovery start point) 106 */ 107 }; 108 109 atomic_t nr_pending; /* number of pending requests. 110 * only maintained for arrays that 111 * support hot removal 112 */ 113 atomic_t read_errors; /* number of consecutive read errors that 114 * we have tried to ignore. 115 */ 116 time64_t last_read_error; /* monotonic time since our 117 * last read error 118 */ 119 atomic_t corrected_errors; /* number of corrected read errors, 120 * for reporting to userspace and storing 121 * in superblock. 122 */ 123 124 struct serial_in_rdev *serial; /* used for raid1 io serialization */ 125 126 struct kernfs_node *sysfs_state; /* handle for 'state' 127 * sysfs entry */ 128 /* handle for 'unacknowledged_bad_blocks' sysfs dentry */ 129 struct kernfs_node *sysfs_unack_badblocks; 130 /* handle for 'bad_blocks' sysfs dentry */ 131 struct kernfs_node *sysfs_badblocks; 132 struct badblocks badblocks; 133 134 struct { 135 short offset; /* Offset from superblock to start of PPL. 136 * Not used by external metadata. */ 137 unsigned int size; /* Size in sectors of the PPL space */ 138 sector_t sector; /* First sector of the PPL space */ 139 } ppl; 140 }; 141 enum flag_bits { 142 Faulty, /* device is known to have a fault */ 143 In_sync, /* device is in_sync with rest of array */ 144 Bitmap_sync, /* ..actually, not quite In_sync. Need a 145 * bitmap-based recovery to get fully in sync. 146 * The bit is only meaningful before device 147 * has been passed to pers->hot_add_disk. 148 */ 149 WriteMostly, /* Avoid reading if at all possible */ 150 AutoDetected, /* added by auto-detect */ 151 Blocked, /* An error occurred but has not yet 152 * been acknowledged by the metadata 153 * handler, so don't allow writes 154 * until it is cleared */ 155 WriteErrorSeen, /* A write error has been seen on this 156 * device 157 */ 158 FaultRecorded, /* Intermediate state for clearing 159 * Blocked. The Fault is/will-be 160 * recorded in the metadata, but that 161 * metadata hasn't been stored safely 162 * on disk yet. 163 */ 164 BlockedBadBlocks, /* A writer is blocked because they 165 * found an unacknowledged bad-block. 166 * This can safely be cleared at any 167 * time, and the writer will re-check. 168 * It may be set at any time, and at 169 * worst the writer will timeout and 170 * re-check. So setting it as 171 * accurately as possible is good, but 172 * not absolutely critical. 173 */ 174 WantReplacement, /* This device is a candidate to be 175 * hot-replaced, either because it has 176 * reported some faults, or because 177 * of explicit request. 178 */ 179 Replacement, /* This device is a replacement for 180 * a want_replacement device with same 181 * raid_disk number. 182 */ 183 Candidate, /* For clustered environments only: 184 * This device is seen locally but not 185 * by the whole cluster 186 */ 187 Journal, /* This device is used as journal for 188 * raid-5/6. 189 * Usually, this device should be faster 190 * than other devices in the array 191 */ 192 ClusterRemove, 193 RemoveSynchronized, /* synchronize_rcu() was called after 194 * this device was known to be faulty, 195 * so it is safe to remove without 196 * another synchronize_rcu() call. 197 */ 198 ExternalBbl, /* External metadata provides bad 199 * block management for a disk 200 */ 201 FailFast, /* Minimal retries should be attempted on 202 * this device, so use REQ_FAILFAST_DEV. 203 * Also don't try to repair failed reads. 204 * It is expects that no bad block log 205 * is present. 206 */ 207 LastDev, /* Seems to be the last working dev as 208 * it didn't fail, so don't use FailFast 209 * any more for metadata 210 */ 211 CollisionCheck, /* 212 * check if there is collision between raid1 213 * serial bios. 214 */ 215 }; 216 217 static inline int is_badblock(struct md_rdev *rdev, sector_t s, int sectors, 218 sector_t *first_bad, int *bad_sectors) 219 { 220 if (unlikely(rdev->badblocks.count)) { 221 int rv = badblocks_check(&rdev->badblocks, rdev->data_offset + s, 222 sectors, 223 first_bad, bad_sectors); 224 if (rv) 225 *first_bad -= rdev->data_offset; 226 return rv; 227 } 228 return 0; 229 } 230 extern int rdev_set_badblocks(struct md_rdev *rdev, sector_t s, int sectors, 231 int is_new); 232 extern int rdev_clear_badblocks(struct md_rdev *rdev, sector_t s, int sectors, 233 int is_new); 234 struct md_cluster_info; 235 236 /** 237 * enum mddev_flags - md device flags. 238 * @MD_ARRAY_FIRST_USE: First use of array, needs initialization. 239 * @MD_CLOSING: If set, we are closing the array, do not open it then. 240 * @MD_JOURNAL_CLEAN: A raid with journal is already clean. 241 * @MD_HAS_JOURNAL: The raid array has journal feature set. 242 * @MD_CLUSTER_RESYNC_LOCKED: cluster raid only, which means node, already took 243 * resync lock, need to release the lock. 244 * @MD_FAILFAST_SUPPORTED: Using MD_FAILFAST on metadata writes is supported as 245 * calls to md_error() will never cause the array to 246 * become failed. 247 * @MD_HAS_PPL: The raid array has PPL feature set. 248 * @MD_HAS_MULTIPLE_PPLS: The raid array has multiple PPLs feature set. 249 * @MD_ALLOW_SB_UPDATE: md_check_recovery is allowed to update the metadata 250 * without taking reconfig_mutex. 251 * @MD_UPDATING_SB: md_check_recovery is updating the metadata without 252 * explicitly holding reconfig_mutex. 253 * @MD_NOT_READY: do_md_run() is active, so 'array_state', ust not report that 254 * array is ready yet. 255 * @MD_BROKEN: This is used to stop writes and mark array as failed. 256 * @MD_DELETED: This device is being deleted 257 * 258 * change UNSUPPORTED_MDDEV_FLAGS for each array type if new flag is added 259 */ 260 enum mddev_flags { 261 MD_ARRAY_FIRST_USE, 262 MD_CLOSING, 263 MD_JOURNAL_CLEAN, 264 MD_HAS_JOURNAL, 265 MD_CLUSTER_RESYNC_LOCKED, 266 MD_FAILFAST_SUPPORTED, 267 MD_HAS_PPL, 268 MD_HAS_MULTIPLE_PPLS, 269 MD_ALLOW_SB_UPDATE, 270 MD_UPDATING_SB, 271 MD_NOT_READY, 272 MD_BROKEN, 273 MD_DELETED, 274 }; 275 276 enum mddev_sb_flags { 277 MD_SB_CHANGE_DEVS, /* Some device status has changed */ 278 MD_SB_CHANGE_CLEAN, /* transition to or from 'clean' */ 279 MD_SB_CHANGE_PENDING, /* switch from 'clean' to 'active' in progress */ 280 MD_SB_NEED_REWRITE, /* metadata write needs to be repeated */ 281 }; 282 283 #define NR_SERIAL_INFOS 8 284 /* record current range of serialize IOs */ 285 struct serial_info { 286 struct rb_node node; 287 sector_t start; /* start sector of rb node */ 288 sector_t last; /* end sector of rb node */ 289 sector_t _subtree_last; /* highest sector in subtree of rb node */ 290 }; 291 292 /* 293 * mddev->curr_resync stores the current sector of the resync but 294 * also has some overloaded values. 295 */ 296 enum { 297 /* No resync in progress */ 298 MD_RESYNC_NONE = 0, 299 /* Yielded to allow another conflicting resync to commence */ 300 MD_RESYNC_YIELDED = 1, 301 /* Delayed to check that there is no conflict with another sync */ 302 MD_RESYNC_DELAYED = 2, 303 /* Any value greater than or equal to this is in an active resync */ 304 MD_RESYNC_ACTIVE = 3, 305 }; 306 307 struct mddev { 308 void *private; 309 struct md_personality *pers; 310 dev_t unit; 311 int md_minor; 312 struct list_head disks; 313 unsigned long flags; 314 unsigned long sb_flags; 315 316 int suspended; 317 struct percpu_ref active_io; 318 int ro; 319 int sysfs_active; /* set when sysfs deletes 320 * are happening, so run/ 321 * takeover/stop are not safe 322 */ 323 struct gendisk *gendisk; 324 325 struct kobject kobj; 326 int hold_active; 327 #define UNTIL_IOCTL 1 328 #define UNTIL_STOP 2 329 330 /* Superblock information */ 331 int major_version, 332 minor_version, 333 patch_version; 334 int persistent; 335 int external; /* metadata is 336 * managed externally */ 337 char metadata_type[17]; /* externally set*/ 338 int chunk_sectors; 339 time64_t ctime, utime; 340 int level, layout; 341 char clevel[16]; 342 int raid_disks; 343 int max_disks; 344 sector_t dev_sectors; /* used size of 345 * component devices */ 346 sector_t array_sectors; /* exported array size */ 347 int external_size; /* size managed 348 * externally */ 349 __u64 events; 350 /* If the last 'event' was simply a clean->dirty transition, and 351 * we didn't write it to the spares, then it is safe and simple 352 * to just decrement the event count on a dirty->clean transition. 353 * So we record that possibility here. 354 */ 355 int can_decrease_events; 356 357 char uuid[16]; 358 359 /* If the array is being reshaped, we need to record the 360 * new shape and an indication of where we are up to. 361 * This is written to the superblock. 362 * If reshape_position is MaxSector, then no reshape is happening (yet). 363 */ 364 sector_t reshape_position; 365 int delta_disks, new_level, new_layout; 366 int new_chunk_sectors; 367 int reshape_backwards; 368 369 struct md_thread __rcu *thread; /* management thread */ 370 struct md_thread __rcu *sync_thread; /* doing resync or reconstruct */ 371 372 /* 'last_sync_action' is initialized to "none". It is set when a 373 * sync operation (i.e "data-check", "requested-resync", "resync", 374 * "recovery", or "reshape") is started. It holds this value even 375 * when the sync thread is "frozen" (interrupted) or "idle" (stopped 376 * or finished). It is overwritten when a new sync operation is begun. 377 */ 378 char *last_sync_action; 379 sector_t curr_resync; /* last block scheduled */ 380 /* As resync requests can complete out of order, we cannot easily track 381 * how much resync has been completed. So we occasionally pause until 382 * everything completes, then set curr_resync_completed to curr_resync. 383 * As such it may be well behind the real resync mark, but it is a value 384 * we are certain of. 385 */ 386 sector_t curr_resync_completed; 387 unsigned long resync_mark; /* a recent timestamp */ 388 sector_t resync_mark_cnt;/* blocks written at resync_mark */ 389 sector_t curr_mark_cnt; /* blocks scheduled now */ 390 391 sector_t resync_max_sectors; /* may be set by personality */ 392 393 atomic64_t resync_mismatches; /* count of sectors where 394 * parity/replica mismatch found 395 */ 396 397 /* allow user-space to request suspension of IO to regions of the array */ 398 sector_t suspend_lo; 399 sector_t suspend_hi; 400 /* if zero, use the system-wide default */ 401 int sync_speed_min; 402 int sync_speed_max; 403 404 /* resync even though the same disks are shared among md-devices */ 405 int parallel_resync; 406 407 int ok_start_degraded; 408 409 unsigned long recovery; 410 /* If a RAID personality determines that recovery (of a particular 411 * device) will fail due to a read error on the source device, it 412 * takes a copy of this number and does not attempt recovery again 413 * until this number changes. 414 */ 415 int recovery_disabled; 416 417 int in_sync; /* know to not need resync */ 418 /* 'open_mutex' avoids races between 'md_open' and 'do_md_stop', so 419 * that we are never stopping an array while it is open. 420 * 'reconfig_mutex' protects all other reconfiguration. 421 * These locks are separate due to conflicting interactions 422 * with disk->open_mutex. 423 * Lock ordering is: 424 * reconfig_mutex -> disk->open_mutex 425 * disk->open_mutex -> open_mutex: e.g. __blkdev_get -> md_open 426 */ 427 struct mutex open_mutex; 428 struct mutex reconfig_mutex; 429 atomic_t active; /* general refcount */ 430 atomic_t openers; /* number of active opens */ 431 432 int changed; /* True if we might need to 433 * reread partition info */ 434 int degraded; /* whether md should consider 435 * adding a spare 436 */ 437 438 atomic_t recovery_active; /* blocks scheduled, but not written */ 439 wait_queue_head_t recovery_wait; 440 sector_t recovery_cp; 441 sector_t resync_min; /* user requested sync 442 * starts here */ 443 sector_t resync_max; /* resync should pause 444 * when it gets here */ 445 446 struct kernfs_node *sysfs_state; /* handle for 'array_state' 447 * file in sysfs. 448 */ 449 struct kernfs_node *sysfs_action; /* handle for 'sync_action' */ 450 struct kernfs_node *sysfs_completed; /*handle for 'sync_completed' */ 451 struct kernfs_node *sysfs_degraded; /*handle for 'degraded' */ 452 struct kernfs_node *sysfs_level; /*handle for 'level' */ 453 454 struct work_struct del_work; /* used for delayed sysfs removal */ 455 456 /* "lock" protects: 457 * flush_bio transition from NULL to !NULL 458 * rdev superblocks, events 459 * clearing MD_CHANGE_* 460 * in_sync - and related safemode and MD_CHANGE changes 461 * pers (also protected by reconfig_mutex and pending IO). 462 * clearing ->bitmap 463 * clearing ->bitmap_info.file 464 * changing ->resync_{min,max} 465 * setting MD_RECOVERY_RUNNING (which interacts with resync_{min,max}) 466 */ 467 spinlock_t lock; 468 wait_queue_head_t sb_wait; /* for waiting on superblock updates */ 469 atomic_t pending_writes; /* number of active superblock writes */ 470 471 unsigned int safemode; /* if set, update "clean" superblock 472 * when no writes pending. 473 */ 474 unsigned int safemode_delay; 475 struct timer_list safemode_timer; 476 struct percpu_ref writes_pending; 477 int sync_checkers; /* # of threads checking writes_pending */ 478 struct request_queue *queue; /* for plugging ... */ 479 480 struct bitmap *bitmap; /* the bitmap for the device */ 481 struct { 482 struct file *file; /* the bitmap file */ 483 loff_t offset; /* offset from superblock of 484 * start of bitmap. May be 485 * negative, but not '0' 486 * For external metadata, offset 487 * from start of device. 488 */ 489 unsigned long space; /* space available at this offset */ 490 loff_t default_offset; /* this is the offset to use when 491 * hot-adding a bitmap. It should 492 * eventually be settable by sysfs. 493 */ 494 unsigned long default_space; /* space available at 495 * default offset */ 496 struct mutex mutex; 497 unsigned long chunksize; 498 unsigned long daemon_sleep; /* how many jiffies between updates? */ 499 unsigned long max_write_behind; /* write-behind mode */ 500 int external; 501 int nodes; /* Maximum number of nodes in the cluster */ 502 char cluster_name[64]; /* Name of the cluster */ 503 } bitmap_info; 504 505 atomic_t max_corr_read_errors; /* max read retries */ 506 struct list_head all_mddevs; 507 508 const struct attribute_group *to_remove; 509 510 struct bio_set bio_set; 511 struct bio_set sync_set; /* for sync operations like 512 * metadata and bitmap writes 513 */ 514 struct bio_set io_clone_set; 515 516 /* Generic flush handling. 517 * The last to finish preflush schedules a worker to submit 518 * the rest of the request (without the REQ_PREFLUSH flag). 519 */ 520 struct bio *flush_bio; 521 atomic_t flush_pending; 522 ktime_t start_flush, prev_flush_start; /* prev_flush_start is when the previous completed 523 * flush was started. 524 */ 525 struct work_struct flush_work; 526 struct work_struct event_work; /* used by dm to report failure event */ 527 mempool_t *serial_info_pool; 528 void (*sync_super)(struct mddev *mddev, struct md_rdev *rdev); 529 struct md_cluster_info *cluster_info; 530 unsigned int good_device_nr; /* good device num within cluster raid */ 531 unsigned int noio_flag; /* for memalloc scope API */ 532 533 /* 534 * Temporarily store rdev that will be finally removed when 535 * reconfig_mutex is unlocked, protected by reconfig_mutex. 536 */ 537 struct list_head deleting; 538 539 /* Used to synchronize idle and frozen for action_store() */ 540 struct mutex sync_mutex; 541 /* The sequence number for sync thread */ 542 atomic_t sync_seq; 543 544 bool has_superblocks:1; 545 bool fail_last_dev:1; 546 bool serialize_policy:1; 547 }; 548 549 enum recovery_flags { 550 /* 551 * If neither SYNC or RESHAPE are set, then it is a recovery. 552 */ 553 MD_RECOVERY_RUNNING, /* a thread is running, or about to be started */ 554 MD_RECOVERY_SYNC, /* actually doing a resync, not a recovery */ 555 MD_RECOVERY_RECOVER, /* doing recovery, or need to try it. */ 556 MD_RECOVERY_INTR, /* resync needs to be aborted for some reason */ 557 MD_RECOVERY_DONE, /* thread is done and is waiting to be reaped */ 558 MD_RECOVERY_NEEDED, /* we might need to start a resync/recover */ 559 MD_RECOVERY_REQUESTED, /* user-space has requested a sync (used with SYNC) */ 560 MD_RECOVERY_CHECK, /* user-space request for check-only, no repair */ 561 MD_RECOVERY_RESHAPE, /* A reshape is happening */ 562 MD_RECOVERY_FROZEN, /* User request to abort, and not restart, any action */ 563 MD_RECOVERY_ERROR, /* sync-action interrupted because io-error */ 564 MD_RECOVERY_WAIT, /* waiting for pers->start() to finish */ 565 MD_RESYNCING_REMOTE, /* remote node is running resync thread */ 566 }; 567 568 enum md_ro_state { 569 MD_RDWR, 570 MD_RDONLY, 571 MD_AUTO_READ, 572 MD_MAX_STATE 573 }; 574 575 static inline bool md_is_rdwr(struct mddev *mddev) 576 { 577 return (mddev->ro == MD_RDWR); 578 } 579 580 static inline bool is_md_suspended(struct mddev *mddev) 581 { 582 return percpu_ref_is_dying(&mddev->active_io); 583 } 584 585 static inline int __must_check mddev_lock(struct mddev *mddev) 586 { 587 return mutex_lock_interruptible(&mddev->reconfig_mutex); 588 } 589 590 /* Sometimes we need to take the lock in a situation where 591 * failure due to interrupts is not acceptable. 592 */ 593 static inline void mddev_lock_nointr(struct mddev *mddev) 594 { 595 mutex_lock(&mddev->reconfig_mutex); 596 } 597 598 static inline int mddev_trylock(struct mddev *mddev) 599 { 600 return mutex_trylock(&mddev->reconfig_mutex); 601 } 602 extern void mddev_unlock(struct mddev *mddev); 603 604 static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors) 605 { 606 atomic_add(nr_sectors, &bdev->bd_disk->sync_io); 607 } 608 609 static inline void md_sync_acct_bio(struct bio *bio, unsigned long nr_sectors) 610 { 611 md_sync_acct(bio->bi_bdev, nr_sectors); 612 } 613 614 struct md_personality 615 { 616 char *name; 617 int level; 618 struct list_head list; 619 struct module *owner; 620 bool __must_check (*make_request)(struct mddev *mddev, struct bio *bio); 621 /* 622 * start up works that do NOT require md_thread. tasks that 623 * requires md_thread should go into start() 624 */ 625 int (*run)(struct mddev *mddev); 626 /* start up works that require md threads */ 627 int (*start)(struct mddev *mddev); 628 void (*free)(struct mddev *mddev, void *priv); 629 void (*status)(struct seq_file *seq, struct mddev *mddev); 630 /* error_handler must set ->faulty and clear ->in_sync 631 * if appropriate, and should abort recovery if needed 632 */ 633 void (*error_handler)(struct mddev *mddev, struct md_rdev *rdev); 634 int (*hot_add_disk) (struct mddev *mddev, struct md_rdev *rdev); 635 int (*hot_remove_disk) (struct mddev *mddev, struct md_rdev *rdev); 636 int (*spare_active) (struct mddev *mddev); 637 sector_t (*sync_request)(struct mddev *mddev, sector_t sector_nr, int *skipped); 638 int (*resize) (struct mddev *mddev, sector_t sectors); 639 sector_t (*size) (struct mddev *mddev, sector_t sectors, int raid_disks); 640 int (*check_reshape) (struct mddev *mddev); 641 int (*start_reshape) (struct mddev *mddev); 642 void (*finish_reshape) (struct mddev *mddev); 643 void (*update_reshape_pos) (struct mddev *mddev); 644 void (*prepare_suspend) (struct mddev *mddev); 645 /* quiesce suspends or resumes internal processing. 646 * 1 - stop new actions and wait for action io to complete 647 * 0 - return to normal behaviour 648 */ 649 void (*quiesce) (struct mddev *mddev, int quiesce); 650 /* takeover is used to transition an array from one 651 * personality to another. The new personality must be able 652 * to handle the data in the current layout. 653 * e.g. 2drive raid1 -> 2drive raid5 654 * ndrive raid5 -> degraded n+1drive raid6 with special layout 655 * If the takeover succeeds, a new 'private' structure is returned. 656 * This needs to be installed and then ->run used to activate the 657 * array. 658 */ 659 void *(*takeover) (struct mddev *mddev); 660 /* Changes the consistency policy of an active array. */ 661 int (*change_consistency_policy)(struct mddev *mddev, const char *buf); 662 }; 663 664 struct md_sysfs_entry { 665 struct attribute attr; 666 ssize_t (*show)(struct mddev *, char *); 667 ssize_t (*store)(struct mddev *, const char *, size_t); 668 }; 669 extern const struct attribute_group md_bitmap_group; 670 671 static inline struct kernfs_node *sysfs_get_dirent_safe(struct kernfs_node *sd, char *name) 672 { 673 if (sd) 674 return sysfs_get_dirent(sd, name); 675 return sd; 676 } 677 static inline void sysfs_notify_dirent_safe(struct kernfs_node *sd) 678 { 679 if (sd) 680 sysfs_notify_dirent(sd); 681 } 682 683 static inline char * mdname (struct mddev * mddev) 684 { 685 return mddev->gendisk ? mddev->gendisk->disk_name : "mdX"; 686 } 687 688 static inline int sysfs_link_rdev(struct mddev *mddev, struct md_rdev *rdev) 689 { 690 char nm[20]; 691 if (!test_bit(Replacement, &rdev->flags) && 692 !test_bit(Journal, &rdev->flags) && 693 mddev->kobj.sd) { 694 sprintf(nm, "rd%d", rdev->raid_disk); 695 return sysfs_create_link(&mddev->kobj, &rdev->kobj, nm); 696 } else 697 return 0; 698 } 699 700 static inline void sysfs_unlink_rdev(struct mddev *mddev, struct md_rdev *rdev) 701 { 702 char nm[20]; 703 if (!test_bit(Replacement, &rdev->flags) && 704 !test_bit(Journal, &rdev->flags) && 705 mddev->kobj.sd) { 706 sprintf(nm, "rd%d", rdev->raid_disk); 707 sysfs_remove_link(&mddev->kobj, nm); 708 } 709 } 710 711 /* 712 * iterates through some rdev ringlist. It's safe to remove the 713 * current 'rdev'. Dont touch 'tmp' though. 714 */ 715 #define rdev_for_each_list(rdev, tmp, head) \ 716 list_for_each_entry_safe(rdev, tmp, head, same_set) 717 718 /* 719 * iterates through the 'same array disks' ringlist 720 */ 721 #define rdev_for_each(rdev, mddev) \ 722 list_for_each_entry(rdev, &((mddev)->disks), same_set) 723 724 #define rdev_for_each_safe(rdev, tmp, mddev) \ 725 list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set) 726 727 #define rdev_for_each_rcu(rdev, mddev) \ 728 list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set) 729 730 struct md_thread { 731 void (*run) (struct md_thread *thread); 732 struct mddev *mddev; 733 wait_queue_head_t wqueue; 734 unsigned long flags; 735 struct task_struct *tsk; 736 unsigned long timeout; 737 void *private; 738 }; 739 740 struct md_io_clone { 741 struct mddev *mddev; 742 struct bio *orig_bio; 743 unsigned long start_time; 744 struct bio bio_clone; 745 }; 746 747 #define THREAD_WAKEUP 0 748 749 static inline void safe_put_page(struct page *p) 750 { 751 if (p) put_page(p); 752 } 753 754 extern int register_md_personality(struct md_personality *p); 755 extern int unregister_md_personality(struct md_personality *p); 756 extern int register_md_cluster_operations(struct md_cluster_operations *ops, 757 struct module *module); 758 extern int unregister_md_cluster_operations(void); 759 extern int md_setup_cluster(struct mddev *mddev, int nodes); 760 extern void md_cluster_stop(struct mddev *mddev); 761 extern struct md_thread *md_register_thread( 762 void (*run)(struct md_thread *thread), 763 struct mddev *mddev, 764 const char *name); 765 extern void md_unregister_thread(struct mddev *mddev, struct md_thread __rcu **threadp); 766 extern void md_wakeup_thread(struct md_thread __rcu *thread); 767 extern void md_check_recovery(struct mddev *mddev); 768 extern void md_reap_sync_thread(struct mddev *mddev); 769 extern int mddev_init_writes_pending(struct mddev *mddev); 770 extern bool md_write_start(struct mddev *mddev, struct bio *bi); 771 extern void md_write_inc(struct mddev *mddev, struct bio *bi); 772 extern void md_write_end(struct mddev *mddev); 773 extern void md_done_sync(struct mddev *mddev, int blocks, int ok); 774 extern void md_error(struct mddev *mddev, struct md_rdev *rdev); 775 extern void md_finish_reshape(struct mddev *mddev); 776 void md_submit_discard_bio(struct mddev *mddev, struct md_rdev *rdev, 777 struct bio *bio, sector_t start, sector_t size); 778 void md_account_bio(struct mddev *mddev, struct bio **bio); 779 780 extern bool __must_check md_flush_request(struct mddev *mddev, struct bio *bio); 781 extern void md_super_write(struct mddev *mddev, struct md_rdev *rdev, 782 sector_t sector, int size, struct page *page); 783 extern int md_super_wait(struct mddev *mddev); 784 extern int sync_page_io(struct md_rdev *rdev, sector_t sector, int size, 785 struct page *page, blk_opf_t opf, bool metadata_op); 786 extern void md_do_sync(struct md_thread *thread); 787 extern void md_new_event(void); 788 extern void md_allow_write(struct mddev *mddev); 789 extern void md_wait_for_blocked_rdev(struct md_rdev *rdev, struct mddev *mddev); 790 extern void md_set_array_sectors(struct mddev *mddev, sector_t array_sectors); 791 extern int md_check_no_bitmap(struct mddev *mddev); 792 extern int md_integrity_register(struct mddev *mddev); 793 extern int md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev); 794 extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale); 795 796 extern void mddev_init(struct mddev *mddev); 797 struct mddev *md_alloc(dev_t dev, char *name); 798 void mddev_put(struct mddev *mddev); 799 extern int md_run(struct mddev *mddev); 800 extern int md_start(struct mddev *mddev); 801 extern void md_stop(struct mddev *mddev); 802 extern void md_stop_writes(struct mddev *mddev); 803 extern int md_rdev_init(struct md_rdev *rdev); 804 extern void md_rdev_clear(struct md_rdev *rdev); 805 806 extern void md_handle_request(struct mddev *mddev, struct bio *bio); 807 extern void mddev_suspend(struct mddev *mddev); 808 extern void mddev_resume(struct mddev *mddev); 809 810 extern void md_reload_sb(struct mddev *mddev, int raid_disk); 811 extern void md_update_sb(struct mddev *mddev, int force); 812 extern void mddev_create_serial_pool(struct mddev *mddev, struct md_rdev *rdev, 813 bool is_suspend); 814 extern void mddev_destroy_serial_pool(struct mddev *mddev, struct md_rdev *rdev, 815 bool is_suspend); 816 struct md_rdev *md_find_rdev_nr_rcu(struct mddev *mddev, int nr); 817 struct md_rdev *md_find_rdev_rcu(struct mddev *mddev, dev_t dev); 818 819 static inline bool is_rdev_broken(struct md_rdev *rdev) 820 { 821 return !disk_live(rdev->bdev->bd_disk); 822 } 823 824 static inline void rdev_dec_pending(struct md_rdev *rdev, struct mddev *mddev) 825 { 826 int faulty = test_bit(Faulty, &rdev->flags); 827 if (atomic_dec_and_test(&rdev->nr_pending) && faulty) { 828 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery); 829 md_wakeup_thread(mddev->thread); 830 } 831 } 832 833 extern struct md_cluster_operations *md_cluster_ops; 834 static inline int mddev_is_clustered(struct mddev *mddev) 835 { 836 return mddev->cluster_info && mddev->bitmap_info.nodes > 1; 837 } 838 839 /* clear unsupported mddev_flags */ 840 static inline void mddev_clear_unsupported_flags(struct mddev *mddev, 841 unsigned long unsupported_flags) 842 { 843 mddev->flags &= ~unsupported_flags; 844 } 845 846 static inline void mddev_check_write_zeroes(struct mddev *mddev, struct bio *bio) 847 { 848 if (bio_op(bio) == REQ_OP_WRITE_ZEROES && 849 !bio->bi_bdev->bd_disk->queue->limits.max_write_zeroes_sectors) 850 mddev->queue->limits.max_write_zeroes_sectors = 0; 851 } 852 853 struct mdu_array_info_s; 854 struct mdu_disk_info_s; 855 856 extern int mdp_major; 857 extern struct workqueue_struct *md_bitmap_wq; 858 void md_autostart_arrays(int part); 859 int md_set_array_info(struct mddev *mddev, struct mdu_array_info_s *info); 860 int md_add_new_disk(struct mddev *mddev, struct mdu_disk_info_s *info); 861 int do_md_run(struct mddev *mddev); 862 863 extern const struct block_device_operations md_fops; 864 865 #endif /* _MD_MD_H */ 866