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 63 struct page *sb_page, *bb_page; 64 int sb_loaded; 65 __u64 sb_events; 66 sector_t data_offset; /* start of data in array */ 67 sector_t new_data_offset;/* only relevant while reshaping */ 68 sector_t sb_start; /* offset of the super block (in 512byte sectors) */ 69 int sb_size; /* bytes in the superblock */ 70 int preferred_minor; /* autorun support */ 71 72 struct kobject kobj; 73 74 /* A device can be in one of three states based on two flags: 75 * Not working: faulty==1 in_sync==0 76 * Fully working: faulty==0 in_sync==1 77 * Working, but not 78 * in sync with array 79 * faulty==0 in_sync==0 80 * 81 * It can never have faulty==1, in_sync==1 82 * This reduces the burden of testing multiple flags in many cases 83 */ 84 85 unsigned long flags; /* bit set of 'enum flag_bits' bits. */ 86 wait_queue_head_t blocked_wait; 87 88 int desc_nr; /* descriptor index in the superblock */ 89 int raid_disk; /* role of device in array */ 90 int new_raid_disk; /* role that the device will have in 91 * the array after a level-change completes. 92 */ 93 int saved_raid_disk; /* role that device used to have in the 94 * array and could again if we did a partial 95 * resync from the bitmap 96 */ 97 union { 98 sector_t recovery_offset;/* If this device has been partially 99 * recovered, this is where we were 100 * up to. 101 */ 102 sector_t journal_tail; /* If this device is a journal device, 103 * this is the journal tail (journal 104 * recovery start point) 105 */ 106 }; 107 108 atomic_t nr_pending; /* number of pending requests. 109 * only maintained for arrays that 110 * support hot removal 111 */ 112 atomic_t read_errors; /* number of consecutive read errors that 113 * we have tried to ignore. 114 */ 115 time64_t last_read_error; /* monotonic time since our 116 * last read error 117 */ 118 atomic_t corrected_errors; /* number of corrected read errors, 119 * for reporting to userspace and storing 120 * in superblock. 121 */ 122 123 struct serial_in_rdev *serial; /* used for raid1 io serialization */ 124 125 struct work_struct del_work; /* used for delayed sysfs removal */ 126 127 struct kernfs_node *sysfs_state; /* handle for 'state' 128 * sysfs entry */ 129 /* handle for 'unacknowledged_bad_blocks' sysfs dentry */ 130 struct kernfs_node *sysfs_unack_badblocks; 131 /* handle for 'bad_blocks' sysfs dentry */ 132 struct kernfs_node *sysfs_badblocks; 133 struct badblocks badblocks; 134 135 struct { 136 short offset; /* Offset from superblock to start of PPL. 137 * Not used by external metadata. */ 138 unsigned int size; /* Size in sectors of the PPL space */ 139 sector_t sector; /* First sector of the PPL space */ 140 } ppl; 141 }; 142 enum flag_bits { 143 Faulty, /* device is known to have a fault */ 144 In_sync, /* device is in_sync with rest of array */ 145 Bitmap_sync, /* ..actually, not quite In_sync. Need a 146 * bitmap-based recovery to get fully in sync. 147 * The bit is only meaningful before device 148 * has been passed to pers->hot_add_disk. 149 */ 150 WriteMostly, /* Avoid reading if at all possible */ 151 AutoDetected, /* added by auto-detect */ 152 Blocked, /* An error occurred but has not yet 153 * been acknowledged by the metadata 154 * handler, so don't allow writes 155 * until it is cleared */ 156 WriteErrorSeen, /* A write error has been seen on this 157 * device 158 */ 159 FaultRecorded, /* Intermediate state for clearing 160 * Blocked. The Fault is/will-be 161 * recorded in the metadata, but that 162 * metadata hasn't been stored safely 163 * on disk yet. 164 */ 165 BlockedBadBlocks, /* A writer is blocked because they 166 * found an unacknowledged bad-block. 167 * This can safely be cleared at any 168 * time, and the writer will re-check. 169 * It may be set at any time, and at 170 * worst the writer will timeout and 171 * re-check. So setting it as 172 * accurately as possible is good, but 173 * not absolutely critical. 174 */ 175 WantReplacement, /* This device is a candidate to be 176 * hot-replaced, either because it has 177 * reported some faults, or because 178 * of explicit request. 179 */ 180 Replacement, /* This device is a replacement for 181 * a want_replacement device with same 182 * raid_disk number. 183 */ 184 Candidate, /* For clustered environments only: 185 * This device is seen locally but not 186 * by the whole cluster 187 */ 188 Journal, /* This device is used as journal for 189 * raid-5/6. 190 * Usually, this device should be faster 191 * than other devices in the array 192 */ 193 ClusterRemove, 194 RemoveSynchronized, /* synchronize_rcu() was called after 195 * this device was known to be faulty, 196 * so it is safe to remove without 197 * another synchronize_rcu() call. 198 */ 199 ExternalBbl, /* External metadata provides bad 200 * block management for a disk 201 */ 202 FailFast, /* Minimal retries should be attempted on 203 * this device, so use REQ_FAILFAST_DEV. 204 * Also don't try to repair failed reads. 205 * It is expects that no bad block log 206 * is present. 207 */ 208 LastDev, /* Seems to be the last working dev as 209 * it didn't fail, so don't use FailFast 210 * any more for metadata 211 */ 212 CollisionCheck, /* 213 * check if there is collision between raid1 214 * serial bios. 215 */ 216 }; 217 218 static inline int is_badblock(struct md_rdev *rdev, sector_t s, int sectors, 219 sector_t *first_bad, int *bad_sectors) 220 { 221 if (unlikely(rdev->badblocks.count)) { 222 int rv = badblocks_check(&rdev->badblocks, rdev->data_offset + s, 223 sectors, 224 first_bad, bad_sectors); 225 if (rv) 226 *first_bad -= rdev->data_offset; 227 return rv; 228 } 229 return 0; 230 } 231 extern int rdev_set_badblocks(struct md_rdev *rdev, sector_t s, int sectors, 232 int is_new); 233 extern int rdev_clear_badblocks(struct md_rdev *rdev, sector_t s, int sectors, 234 int is_new); 235 struct md_cluster_info; 236 237 /** 238 * enum mddev_flags - md device flags. 239 * @MD_ARRAY_FIRST_USE: First use of array, needs initialization. 240 * @MD_CLOSING: If set, we are closing the array, do not open it then. 241 * @MD_JOURNAL_CLEAN: A raid with journal is already clean. 242 * @MD_HAS_JOURNAL: The raid array has journal feature set. 243 * @MD_CLUSTER_RESYNC_LOCKED: cluster raid only, which means node, already took 244 * resync lock, need to release the lock. 245 * @MD_FAILFAST_SUPPORTED: Using MD_FAILFAST on metadata writes is supported as 246 * calls to md_error() will never cause the array to 247 * become failed. 248 * @MD_HAS_PPL: The raid array has PPL feature set. 249 * @MD_HAS_MULTIPLE_PPLS: The raid array has multiple PPLs feature set. 250 * @MD_ALLOW_SB_UPDATE: md_check_recovery is allowed to update the metadata 251 * without taking reconfig_mutex. 252 * @MD_UPDATING_SB: md_check_recovery is updating the metadata without 253 * explicitly holding reconfig_mutex. 254 * @MD_NOT_READY: do_md_run() is active, so 'array_state', ust not report that 255 * array is ready yet. 256 * @MD_BROKEN: This is used to stop writes and mark array as failed. 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 }; 274 275 enum mddev_sb_flags { 276 MD_SB_CHANGE_DEVS, /* Some device status has changed */ 277 MD_SB_CHANGE_CLEAN, /* transition to or from 'clean' */ 278 MD_SB_CHANGE_PENDING, /* switch from 'clean' to 'active' in progress */ 279 MD_SB_NEED_REWRITE, /* metadata write needs to be repeated */ 280 }; 281 282 #define NR_SERIAL_INFOS 8 283 /* record current range of serialize IOs */ 284 struct serial_info { 285 struct rb_node node; 286 sector_t start; /* start sector of rb node */ 287 sector_t last; /* end sector of rb node */ 288 sector_t _subtree_last; /* highest sector in subtree of rb node */ 289 }; 290 291 struct mddev { 292 void *private; 293 struct md_personality *pers; 294 dev_t unit; 295 int md_minor; 296 struct list_head disks; 297 unsigned long flags; 298 unsigned long sb_flags; 299 300 int suspended; 301 atomic_t active_io; 302 int ro; 303 int sysfs_active; /* set when sysfs deletes 304 * are happening, so run/ 305 * takeover/stop are not safe 306 */ 307 struct gendisk *gendisk; 308 309 struct kobject kobj; 310 int hold_active; 311 #define UNTIL_IOCTL 1 312 #define UNTIL_STOP 2 313 314 /* Superblock information */ 315 int major_version, 316 minor_version, 317 patch_version; 318 int persistent; 319 int external; /* metadata is 320 * managed externally */ 321 char metadata_type[17]; /* externally set*/ 322 int chunk_sectors; 323 time64_t ctime, utime; 324 int level, layout; 325 char clevel[16]; 326 int raid_disks; 327 int max_disks; 328 sector_t dev_sectors; /* used size of 329 * component devices */ 330 sector_t array_sectors; /* exported array size */ 331 int external_size; /* size managed 332 * externally */ 333 __u64 events; 334 /* If the last 'event' was simply a clean->dirty transition, and 335 * we didn't write it to the spares, then it is safe and simple 336 * to just decrement the event count on a dirty->clean transition. 337 * So we record that possibility here. 338 */ 339 int can_decrease_events; 340 341 char uuid[16]; 342 343 /* If the array is being reshaped, we need to record the 344 * new shape and an indication of where we are up to. 345 * This is written to the superblock. 346 * If reshape_position is MaxSector, then no reshape is happening (yet). 347 */ 348 sector_t reshape_position; 349 int delta_disks, new_level, new_layout; 350 int new_chunk_sectors; 351 int reshape_backwards; 352 353 struct md_thread *thread; /* management thread */ 354 struct md_thread *sync_thread; /* doing resync or reconstruct */ 355 356 /* 'last_sync_action' is initialized to "none". It is set when a 357 * sync operation (i.e "data-check", "requested-resync", "resync", 358 * "recovery", or "reshape") is started. It holds this value even 359 * when the sync thread is "frozen" (interrupted) or "idle" (stopped 360 * or finished). It is overwritten when a new sync operation is begun. 361 */ 362 char *last_sync_action; 363 sector_t curr_resync; /* last block scheduled */ 364 /* As resync requests can complete out of order, we cannot easily track 365 * how much resync has been completed. So we occasionally pause until 366 * everything completes, then set curr_resync_completed to curr_resync. 367 * As such it may be well behind the real resync mark, but it is a value 368 * we are certain of. 369 */ 370 sector_t curr_resync_completed; 371 unsigned long resync_mark; /* a recent timestamp */ 372 sector_t resync_mark_cnt;/* blocks written at resync_mark */ 373 sector_t curr_mark_cnt; /* blocks scheduled now */ 374 375 sector_t resync_max_sectors; /* may be set by personality */ 376 377 atomic64_t resync_mismatches; /* count of sectors where 378 * parity/replica mismatch found 379 */ 380 381 /* allow user-space to request suspension of IO to regions of the array */ 382 sector_t suspend_lo; 383 sector_t suspend_hi; 384 /* if zero, use the system-wide default */ 385 int sync_speed_min; 386 int sync_speed_max; 387 388 /* resync even though the same disks are shared among md-devices */ 389 int parallel_resync; 390 391 int ok_start_degraded; 392 393 unsigned long recovery; 394 /* If a RAID personality determines that recovery (of a particular 395 * device) will fail due to a read error on the source device, it 396 * takes a copy of this number and does not attempt recovery again 397 * until this number changes. 398 */ 399 int recovery_disabled; 400 401 int in_sync; /* know to not need resync */ 402 /* 'open_mutex' avoids races between 'md_open' and 'do_md_stop', so 403 * that we are never stopping an array while it is open. 404 * 'reconfig_mutex' protects all other reconfiguration. 405 * These locks are separate due to conflicting interactions 406 * with disk->open_mutex. 407 * Lock ordering is: 408 * reconfig_mutex -> disk->open_mutex 409 * disk->open_mutex -> open_mutex: e.g. __blkdev_get -> md_open 410 */ 411 struct mutex open_mutex; 412 struct mutex reconfig_mutex; 413 atomic_t active; /* general refcount */ 414 atomic_t openers; /* number of active opens */ 415 416 int changed; /* True if we might need to 417 * reread partition info */ 418 int degraded; /* whether md should consider 419 * adding a spare 420 */ 421 422 atomic_t recovery_active; /* blocks scheduled, but not written */ 423 wait_queue_head_t recovery_wait; 424 sector_t recovery_cp; 425 sector_t resync_min; /* user requested sync 426 * starts here */ 427 sector_t resync_max; /* resync should pause 428 * when it gets here */ 429 430 struct kernfs_node *sysfs_state; /* handle for 'array_state' 431 * file in sysfs. 432 */ 433 struct kernfs_node *sysfs_action; /* handle for 'sync_action' */ 434 struct kernfs_node *sysfs_completed; /*handle for 'sync_completed' */ 435 struct kernfs_node *sysfs_degraded; /*handle for 'degraded' */ 436 struct kernfs_node *sysfs_level; /*handle for 'level' */ 437 438 struct work_struct del_work; /* used for delayed sysfs removal */ 439 440 /* "lock" protects: 441 * flush_bio transition from NULL to !NULL 442 * rdev superblocks, events 443 * clearing MD_CHANGE_* 444 * in_sync - and related safemode and MD_CHANGE changes 445 * pers (also protected by reconfig_mutex and pending IO). 446 * clearing ->bitmap 447 * clearing ->bitmap_info.file 448 * changing ->resync_{min,max} 449 * setting MD_RECOVERY_RUNNING (which interacts with resync_{min,max}) 450 */ 451 spinlock_t lock; 452 wait_queue_head_t sb_wait; /* for waiting on superblock updates */ 453 atomic_t pending_writes; /* number of active superblock writes */ 454 455 unsigned int safemode; /* if set, update "clean" superblock 456 * when no writes pending. 457 */ 458 unsigned int safemode_delay; 459 struct timer_list safemode_timer; 460 struct percpu_ref writes_pending; 461 int sync_checkers; /* # of threads checking writes_pending */ 462 struct request_queue *queue; /* for plugging ... */ 463 464 struct bitmap *bitmap; /* the bitmap for the device */ 465 struct { 466 struct file *file; /* the bitmap file */ 467 loff_t offset; /* offset from superblock of 468 * start of bitmap. May be 469 * negative, but not '0' 470 * For external metadata, offset 471 * from start of device. 472 */ 473 unsigned long space; /* space available at this offset */ 474 loff_t default_offset; /* this is the offset to use when 475 * hot-adding a bitmap. It should 476 * eventually be settable by sysfs. 477 */ 478 unsigned long default_space; /* space available at 479 * default offset */ 480 struct mutex mutex; 481 unsigned long chunksize; 482 unsigned long daemon_sleep; /* how many jiffies between updates? */ 483 unsigned long max_write_behind; /* write-behind mode */ 484 int external; 485 int nodes; /* Maximum number of nodes in the cluster */ 486 char cluster_name[64]; /* Name of the cluster */ 487 } bitmap_info; 488 489 atomic_t max_corr_read_errors; /* max read retries */ 490 struct list_head all_mddevs; 491 492 const struct attribute_group *to_remove; 493 494 struct bio_set bio_set; 495 struct bio_set sync_set; /* for sync operations like 496 * metadata and bitmap writes 497 */ 498 struct bio_set io_acct_set; /* for raid0 and raid5 io accounting */ 499 500 /* Generic flush handling. 501 * The last to finish preflush schedules a worker to submit 502 * the rest of the request (without the REQ_PREFLUSH flag). 503 */ 504 struct bio *flush_bio; 505 atomic_t flush_pending; 506 ktime_t start_flush, prev_flush_start; /* prev_flush_start is when the previous completed 507 * flush was started. 508 */ 509 struct work_struct flush_work; 510 struct work_struct event_work; /* used by dm to report failure event */ 511 mempool_t *serial_info_pool; 512 void (*sync_super)(struct mddev *mddev, struct md_rdev *rdev); 513 struct md_cluster_info *cluster_info; 514 unsigned int good_device_nr; /* good device num within cluster raid */ 515 unsigned int noio_flag; /* for memalloc scope API */ 516 517 bool has_superblocks:1; 518 bool fail_last_dev:1; 519 bool serialize_policy:1; 520 }; 521 522 enum recovery_flags { 523 /* 524 * If neither SYNC or RESHAPE are set, then it is a recovery. 525 */ 526 MD_RECOVERY_RUNNING, /* a thread is running, or about to be started */ 527 MD_RECOVERY_SYNC, /* actually doing a resync, not a recovery */ 528 MD_RECOVERY_RECOVER, /* doing recovery, or need to try it. */ 529 MD_RECOVERY_INTR, /* resync needs to be aborted for some reason */ 530 MD_RECOVERY_DONE, /* thread is done and is waiting to be reaped */ 531 MD_RECOVERY_NEEDED, /* we might need to start a resync/recover */ 532 MD_RECOVERY_REQUESTED, /* user-space has requested a sync (used with SYNC) */ 533 MD_RECOVERY_CHECK, /* user-space request for check-only, no repair */ 534 MD_RECOVERY_RESHAPE, /* A reshape is happening */ 535 MD_RECOVERY_FROZEN, /* User request to abort, and not restart, any action */ 536 MD_RECOVERY_ERROR, /* sync-action interrupted because io-error */ 537 MD_RECOVERY_WAIT, /* waiting for pers->start() to finish */ 538 MD_RESYNCING_REMOTE, /* remote node is running resync thread */ 539 }; 540 541 static inline int __must_check mddev_lock(struct mddev *mddev) 542 { 543 return mutex_lock_interruptible(&mddev->reconfig_mutex); 544 } 545 546 /* Sometimes we need to take the lock in a situation where 547 * failure due to interrupts is not acceptable. 548 */ 549 static inline void mddev_lock_nointr(struct mddev *mddev) 550 { 551 mutex_lock(&mddev->reconfig_mutex); 552 } 553 554 static inline int mddev_trylock(struct mddev *mddev) 555 { 556 return mutex_trylock(&mddev->reconfig_mutex); 557 } 558 extern void mddev_unlock(struct mddev *mddev); 559 560 static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors) 561 { 562 atomic_add(nr_sectors, &bdev->bd_disk->sync_io); 563 } 564 565 static inline void md_sync_acct_bio(struct bio *bio, unsigned long nr_sectors) 566 { 567 md_sync_acct(bio->bi_bdev, nr_sectors); 568 } 569 570 struct md_personality 571 { 572 char *name; 573 int level; 574 struct list_head list; 575 struct module *owner; 576 bool __must_check (*make_request)(struct mddev *mddev, struct bio *bio); 577 /* 578 * start up works that do NOT require md_thread. tasks that 579 * requires md_thread should go into start() 580 */ 581 int (*run)(struct mddev *mddev); 582 /* start up works that require md threads */ 583 int (*start)(struct mddev *mddev); 584 void (*free)(struct mddev *mddev, void *priv); 585 void (*status)(struct seq_file *seq, struct mddev *mddev); 586 /* error_handler must set ->faulty and clear ->in_sync 587 * if appropriate, and should abort recovery if needed 588 */ 589 void (*error_handler)(struct mddev *mddev, struct md_rdev *rdev); 590 int (*hot_add_disk) (struct mddev *mddev, struct md_rdev *rdev); 591 int (*hot_remove_disk) (struct mddev *mddev, struct md_rdev *rdev); 592 int (*spare_active) (struct mddev *mddev); 593 sector_t (*sync_request)(struct mddev *mddev, sector_t sector_nr, int *skipped); 594 int (*resize) (struct mddev *mddev, sector_t sectors); 595 sector_t (*size) (struct mddev *mddev, sector_t sectors, int raid_disks); 596 int (*check_reshape) (struct mddev *mddev); 597 int (*start_reshape) (struct mddev *mddev); 598 void (*finish_reshape) (struct mddev *mddev); 599 void (*update_reshape_pos) (struct mddev *mddev); 600 /* quiesce suspends or resumes internal processing. 601 * 1 - stop new actions and wait for action io to complete 602 * 0 - return to normal behaviour 603 */ 604 void (*quiesce) (struct mddev *mddev, int quiesce); 605 /* takeover is used to transition an array from one 606 * personality to another. The new personality must be able 607 * to handle the data in the current layout. 608 * e.g. 2drive raid1 -> 2drive raid5 609 * ndrive raid5 -> degraded n+1drive raid6 with special layout 610 * If the takeover succeeds, a new 'private' structure is returned. 611 * This needs to be installed and then ->run used to activate the 612 * array. 613 */ 614 void *(*takeover) (struct mddev *mddev); 615 /* Changes the consistency policy of an active array. */ 616 int (*change_consistency_policy)(struct mddev *mddev, const char *buf); 617 }; 618 619 struct md_sysfs_entry { 620 struct attribute attr; 621 ssize_t (*show)(struct mddev *, char *); 622 ssize_t (*store)(struct mddev *, const char *, size_t); 623 }; 624 extern const struct attribute_group md_bitmap_group; 625 626 static inline struct kernfs_node *sysfs_get_dirent_safe(struct kernfs_node *sd, char *name) 627 { 628 if (sd) 629 return sysfs_get_dirent(sd, name); 630 return sd; 631 } 632 static inline void sysfs_notify_dirent_safe(struct kernfs_node *sd) 633 { 634 if (sd) 635 sysfs_notify_dirent(sd); 636 } 637 638 static inline char * mdname (struct mddev * mddev) 639 { 640 return mddev->gendisk ? mddev->gendisk->disk_name : "mdX"; 641 } 642 643 static inline int sysfs_link_rdev(struct mddev *mddev, struct md_rdev *rdev) 644 { 645 char nm[20]; 646 if (!test_bit(Replacement, &rdev->flags) && 647 !test_bit(Journal, &rdev->flags) && 648 mddev->kobj.sd) { 649 sprintf(nm, "rd%d", rdev->raid_disk); 650 return sysfs_create_link(&mddev->kobj, &rdev->kobj, nm); 651 } else 652 return 0; 653 } 654 655 static inline void sysfs_unlink_rdev(struct mddev *mddev, struct md_rdev *rdev) 656 { 657 char nm[20]; 658 if (!test_bit(Replacement, &rdev->flags) && 659 !test_bit(Journal, &rdev->flags) && 660 mddev->kobj.sd) { 661 sprintf(nm, "rd%d", rdev->raid_disk); 662 sysfs_remove_link(&mddev->kobj, nm); 663 } 664 } 665 666 /* 667 * iterates through some rdev ringlist. It's safe to remove the 668 * current 'rdev'. Dont touch 'tmp' though. 669 */ 670 #define rdev_for_each_list(rdev, tmp, head) \ 671 list_for_each_entry_safe(rdev, tmp, head, same_set) 672 673 /* 674 * iterates through the 'same array disks' ringlist 675 */ 676 #define rdev_for_each(rdev, mddev) \ 677 list_for_each_entry(rdev, &((mddev)->disks), same_set) 678 679 #define rdev_for_each_safe(rdev, tmp, mddev) \ 680 list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set) 681 682 #define rdev_for_each_rcu(rdev, mddev) \ 683 list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set) 684 685 struct md_thread { 686 void (*run) (struct md_thread *thread); 687 struct mddev *mddev; 688 wait_queue_head_t wqueue; 689 unsigned long flags; 690 struct task_struct *tsk; 691 unsigned long timeout; 692 void *private; 693 }; 694 695 struct md_io_acct { 696 struct bio *orig_bio; 697 unsigned long start_time; 698 struct bio bio_clone; 699 }; 700 701 #define THREAD_WAKEUP 0 702 703 static inline void safe_put_page(struct page *p) 704 { 705 if (p) put_page(p); 706 } 707 708 extern int register_md_personality(struct md_personality *p); 709 extern int unregister_md_personality(struct md_personality *p); 710 extern int register_md_cluster_operations(struct md_cluster_operations *ops, 711 struct module *module); 712 extern int unregister_md_cluster_operations(void); 713 extern int md_setup_cluster(struct mddev *mddev, int nodes); 714 extern void md_cluster_stop(struct mddev *mddev); 715 extern struct md_thread *md_register_thread( 716 void (*run)(struct md_thread *thread), 717 struct mddev *mddev, 718 const char *name); 719 extern void md_unregister_thread(struct md_thread **threadp); 720 extern void md_wakeup_thread(struct md_thread *thread); 721 extern void md_check_recovery(struct mddev *mddev); 722 extern void md_reap_sync_thread(struct mddev *mddev); 723 extern int mddev_init_writes_pending(struct mddev *mddev); 724 extern bool md_write_start(struct mddev *mddev, struct bio *bi); 725 extern void md_write_inc(struct mddev *mddev, struct bio *bi); 726 extern void md_write_end(struct mddev *mddev); 727 extern void md_done_sync(struct mddev *mddev, int blocks, int ok); 728 extern void md_error(struct mddev *mddev, struct md_rdev *rdev); 729 extern void md_finish_reshape(struct mddev *mddev); 730 void md_submit_discard_bio(struct mddev *mddev, struct md_rdev *rdev, 731 struct bio *bio, sector_t start, sector_t size); 732 int acct_bioset_init(struct mddev *mddev); 733 void acct_bioset_exit(struct mddev *mddev); 734 void md_account_bio(struct mddev *mddev, struct bio **bio); 735 736 extern bool __must_check md_flush_request(struct mddev *mddev, struct bio *bio); 737 extern void md_super_write(struct mddev *mddev, struct md_rdev *rdev, 738 sector_t sector, int size, struct page *page); 739 extern int md_super_wait(struct mddev *mddev); 740 extern int sync_page_io(struct md_rdev *rdev, sector_t sector, int size, 741 struct page *page, int op, int op_flags, 742 bool metadata_op); 743 extern void md_do_sync(struct md_thread *thread); 744 extern void md_new_event(void); 745 extern void md_allow_write(struct mddev *mddev); 746 extern void md_wait_for_blocked_rdev(struct md_rdev *rdev, struct mddev *mddev); 747 extern void md_set_array_sectors(struct mddev *mddev, sector_t array_sectors); 748 extern int md_check_no_bitmap(struct mddev *mddev); 749 extern int md_integrity_register(struct mddev *mddev); 750 extern int md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev); 751 extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale); 752 753 extern void mddev_init(struct mddev *mddev); 754 extern int md_run(struct mddev *mddev); 755 extern int md_start(struct mddev *mddev); 756 extern void md_stop(struct mddev *mddev); 757 extern void md_stop_writes(struct mddev *mddev); 758 extern int md_rdev_init(struct md_rdev *rdev); 759 extern void md_rdev_clear(struct md_rdev *rdev); 760 761 extern void md_handle_request(struct mddev *mddev, struct bio *bio); 762 extern void mddev_suspend(struct mddev *mddev); 763 extern void mddev_resume(struct mddev *mddev); 764 765 extern void md_reload_sb(struct mddev *mddev, int raid_disk); 766 extern void md_update_sb(struct mddev *mddev, int force); 767 extern void md_kick_rdev_from_array(struct md_rdev * rdev); 768 extern void mddev_create_serial_pool(struct mddev *mddev, struct md_rdev *rdev, 769 bool is_suspend); 770 extern void mddev_destroy_serial_pool(struct mddev *mddev, struct md_rdev *rdev, 771 bool is_suspend); 772 struct md_rdev *md_find_rdev_nr_rcu(struct mddev *mddev, int nr); 773 struct md_rdev *md_find_rdev_rcu(struct mddev *mddev, dev_t dev); 774 775 static inline bool is_mddev_broken(struct md_rdev *rdev, const char *md_type) 776 { 777 if (!disk_live(rdev->bdev->bd_disk)) { 778 if (!test_and_set_bit(MD_BROKEN, &rdev->mddev->flags)) 779 pr_warn("md: %s: %s array has a missing/failed member\n", 780 mdname(rdev->mddev), md_type); 781 return true; 782 } 783 return false; 784 } 785 786 static inline void rdev_dec_pending(struct md_rdev *rdev, struct mddev *mddev) 787 { 788 int faulty = test_bit(Faulty, &rdev->flags); 789 if (atomic_dec_and_test(&rdev->nr_pending) && faulty) { 790 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery); 791 md_wakeup_thread(mddev->thread); 792 } 793 } 794 795 extern struct md_cluster_operations *md_cluster_ops; 796 static inline int mddev_is_clustered(struct mddev *mddev) 797 { 798 return mddev->cluster_info && mddev->bitmap_info.nodes > 1; 799 } 800 801 /* clear unsupported mddev_flags */ 802 static inline void mddev_clear_unsupported_flags(struct mddev *mddev, 803 unsigned long unsupported_flags) 804 { 805 mddev->flags &= ~unsupported_flags; 806 } 807 808 static inline void mddev_check_write_zeroes(struct mddev *mddev, struct bio *bio) 809 { 810 if (bio_op(bio) == REQ_OP_WRITE_ZEROES && 811 !bio->bi_bdev->bd_disk->queue->limits.max_write_zeroes_sectors) 812 mddev->queue->limits.max_write_zeroes_sectors = 0; 813 } 814 815 struct mdu_array_info_s; 816 struct mdu_disk_info_s; 817 818 extern int mdp_major; 819 void md_autostart_arrays(int part); 820 int md_set_array_info(struct mddev *mddev, struct mdu_array_info_s *info); 821 int md_add_new_disk(struct mddev *mddev, struct mdu_disk_info_s *info); 822 int do_md_run(struct mddev *mddev); 823 824 extern const struct block_device_operations md_fops; 825 826 #endif /* _MD_MD_H */ 827