1 /* SPDX-License-Identifier: GPL-2.0-or-later */
3 md.h : kernel internal structure of the Linux MD driver
4 Copyright (C) 1996-98 Ingo Molnar, Gadi Oxman
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>
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"
23 #define MaxSector (~(sector_t)0)
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
34 #define MD_FAILFAST (REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT)
37 * The struct embedded in rdev is used to serialize IO.
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;
46 * MD's 'extended' device
49 struct list_head same_set; /* RAID devices within the same set */
51 sector_t sectors; /* Device size (in 512bytes sectors) */
52 struct mddev *mddev; /* RAID array if running */
53 int last_events; /* IO event timestamp */
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).
60 struct block_device *meta_bdev;
61 struct block_device *bdev; /* block device handle */
63 struct page *sb_page, *bb_page;
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 */
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
79 * faulty==0 in_sync==0
81 * It can never have faulty==1, in_sync==1
82 * This reduces the burden of testing multiple flags in many cases
85 unsigned long flags; /* bit set of 'enum flag_bits' bits. */
86 wait_queue_head_t blocked_wait;
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.
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
98 sector_t recovery_offset;/* If this device has been partially
99 * recovered, this is where we were
102 sector_t journal_tail; /* If this device is a journal device,
103 * this is the journal tail (journal
104 * recovery start point)
108 atomic_t nr_pending; /* number of pending requests.
109 * only maintained for arrays that
110 * support hot removal
112 atomic_t read_errors; /* number of consecutive read errors that
113 * we have tried to ignore.
115 time64_t last_read_error; /* monotonic time since our
118 atomic_t corrected_errors; /* number of corrected read errors,
119 * for reporting to userspace and storing
123 struct serial_in_rdev *serial; /* used for raid1 io serialization */
125 struct kernfs_node *sysfs_state; /* handle for 'state'
127 /* handle for 'unacknowledged_bad_blocks' sysfs dentry */
128 struct kernfs_node *sysfs_unack_badblocks;
129 /* handle for 'bad_blocks' sysfs dentry */
130 struct kernfs_node *sysfs_badblocks;
131 struct badblocks badblocks;
134 short offset; /* Offset from superblock to start of PPL.
135 * Not used by external metadata. */
136 unsigned int size; /* Size in sectors of the PPL space */
137 sector_t sector; /* First sector of the PPL space */
141 Faulty, /* device is known to have a fault */
142 In_sync, /* device is in_sync with rest of array */
143 Bitmap_sync, /* ..actually, not quite In_sync. Need a
144 * bitmap-based recovery to get fully in sync.
145 * The bit is only meaningful before device
146 * has been passed to pers->hot_add_disk.
148 WriteMostly, /* Avoid reading if at all possible */
149 AutoDetected, /* added by auto-detect */
150 Blocked, /* An error occurred but has not yet
151 * been acknowledged by the metadata
152 * handler, so don't allow writes
153 * until it is cleared */
154 WriteErrorSeen, /* A write error has been seen on this
157 FaultRecorded, /* Intermediate state for clearing
158 * Blocked. The Fault is/will-be
159 * recorded in the metadata, but that
160 * metadata hasn't been stored safely
163 BlockedBadBlocks, /* A writer is blocked because they
164 * found an unacknowledged bad-block.
165 * This can safely be cleared at any
166 * time, and the writer will re-check.
167 * It may be set at any time, and at
168 * worst the writer will timeout and
169 * re-check. So setting it as
170 * accurately as possible is good, but
171 * not absolutely critical.
173 WantReplacement, /* This device is a candidate to be
174 * hot-replaced, either because it has
175 * reported some faults, or because
176 * of explicit request.
178 Replacement, /* This device is a replacement for
179 * a want_replacement device with same
182 Candidate, /* For clustered environments only:
183 * This device is seen locally but not
184 * by the whole cluster
186 Journal, /* This device is used as journal for
188 * Usually, this device should be faster
189 * than other devices in the array
192 RemoveSynchronized, /* synchronize_rcu() was called after
193 * this device was known to be faulty,
194 * so it is safe to remove without
195 * another synchronize_rcu() call.
197 ExternalBbl, /* External metadata provides bad
198 * block management for a disk
200 FailFast, /* Minimal retries should be attempted on
201 * this device, so use REQ_FAILFAST_DEV.
202 * Also don't try to repair failed reads.
203 * It is expects that no bad block log
206 LastDev, /* Seems to be the last working dev as
207 * it didn't fail, so don't use FailFast
208 * any more for metadata
211 * check if there is collision between raid1
216 static inline int is_badblock(struct md_rdev *rdev, sector_t s, int sectors,
217 sector_t *first_bad, int *bad_sectors)
219 if (unlikely(rdev->badblocks.count)) {
220 int rv = badblocks_check(&rdev->badblocks, rdev->data_offset + s,
222 first_bad, bad_sectors);
224 *first_bad -= rdev->data_offset;
229 extern int rdev_set_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
231 extern int rdev_clear_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
233 struct md_cluster_info;
236 * enum mddev_flags - md device flags.
237 * @MD_ARRAY_FIRST_USE: First use of array, needs initialization.
238 * @MD_CLOSING: If set, we are closing the array, do not open it then.
239 * @MD_JOURNAL_CLEAN: A raid with journal is already clean.
240 * @MD_HAS_JOURNAL: The raid array has journal feature set.
241 * @MD_CLUSTER_RESYNC_LOCKED: cluster raid only, which means node, already took
242 * resync lock, need to release the lock.
243 * @MD_FAILFAST_SUPPORTED: Using MD_FAILFAST on metadata writes is supported as
244 * calls to md_error() will never cause the array to
246 * @MD_HAS_PPL: The raid array has PPL feature set.
247 * @MD_HAS_MULTIPLE_PPLS: The raid array has multiple PPLs feature set.
248 * @MD_ALLOW_SB_UPDATE: md_check_recovery is allowed to update the metadata
249 * without taking reconfig_mutex.
250 * @MD_UPDATING_SB: md_check_recovery is updating the metadata without
251 * explicitly holding reconfig_mutex.
252 * @MD_NOT_READY: do_md_run() is active, so 'array_state', ust not report that
253 * array is ready yet.
254 * @MD_BROKEN: This is used to stop writes and mark array as failed.
255 * @MD_DELETED: This device is being deleted
257 * change UNSUPPORTED_MDDEV_FLAGS for each array type if new flag is added
264 MD_CLUSTER_RESYNC_LOCKED,
265 MD_FAILFAST_SUPPORTED,
267 MD_HAS_MULTIPLE_PPLS,
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 */
282 #define NR_SERIAL_INFOS 8
283 /* record current range of serialize IOs */
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 */
292 * mddev->curr_resync stores the current sector of the resync but
293 * also has some overloaded values.
296 /* No resync in progress */
298 /* Yielded to allow another conflicting resync to commence */
299 MD_RESYNC_YIELDED = 1,
300 /* Delayed to check that there is no conflict with another sync */
301 MD_RESYNC_DELAYED = 2,
302 /* Any value greater than or equal to this is in an active resync */
303 MD_RESYNC_ACTIVE = 3,
308 struct md_personality *pers;
311 struct list_head disks;
313 unsigned long sb_flags;
316 struct percpu_ref active_io;
318 int sysfs_active; /* set when sysfs deletes
319 * are happening, so run/
320 * takeover/stop are not safe
322 struct gendisk *gendisk;
326 #define UNTIL_IOCTL 1
329 /* Superblock information */
334 int external; /* metadata is
335 * managed externally */
336 char metadata_type[17]; /* externally set*/
338 time64_t ctime, utime;
343 sector_t dev_sectors; /* used size of
344 * component devices */
345 sector_t array_sectors; /* exported array size */
346 int external_size; /* size managed
349 /* If the last 'event' was simply a clean->dirty transition, and
350 * we didn't write it to the spares, then it is safe and simple
351 * to just decrement the event count on a dirty->clean transition.
352 * So we record that possibility here.
354 int can_decrease_events;
358 /* If the array is being reshaped, we need to record the
359 * new shape and an indication of where we are up to.
360 * This is written to the superblock.
361 * If reshape_position is MaxSector, then no reshape is happening (yet).
363 sector_t reshape_position;
364 int delta_disks, new_level, new_layout;
365 int new_chunk_sectors;
366 int reshape_backwards;
368 struct md_thread __rcu *thread; /* management thread */
369 struct md_thread __rcu *sync_thread; /* doing resync or reconstruct */
371 /* 'last_sync_action' is initialized to "none". It is set when a
372 * sync operation (i.e "data-check", "requested-resync", "resync",
373 * "recovery", or "reshape") is started. It holds this value even
374 * when the sync thread is "frozen" (interrupted) or "idle" (stopped
375 * or finished). It is overwritten when a new sync operation is begun.
377 char *last_sync_action;
378 sector_t curr_resync; /* last block scheduled */
379 /* As resync requests can complete out of order, we cannot easily track
380 * how much resync has been completed. So we occasionally pause until
381 * everything completes, then set curr_resync_completed to curr_resync.
382 * As such it may be well behind the real resync mark, but it is a value
385 sector_t curr_resync_completed;
386 unsigned long resync_mark; /* a recent timestamp */
387 sector_t resync_mark_cnt;/* blocks written at resync_mark */
388 sector_t curr_mark_cnt; /* blocks scheduled now */
390 sector_t resync_max_sectors; /* may be set by personality */
392 atomic64_t resync_mismatches; /* count of sectors where
393 * parity/replica mismatch found
396 /* allow user-space to request suspension of IO to regions of the array */
399 /* if zero, use the system-wide default */
403 /* resync even though the same disks are shared among md-devices */
406 int ok_start_degraded;
408 unsigned long recovery;
409 /* If a RAID personality determines that recovery (of a particular
410 * device) will fail due to a read error on the source device, it
411 * takes a copy of this number and does not attempt recovery again
412 * until this number changes.
414 int recovery_disabled;
416 int in_sync; /* know to not need resync */
417 /* 'open_mutex' avoids races between 'md_open' and 'do_md_stop', so
418 * that we are never stopping an array while it is open.
419 * 'reconfig_mutex' protects all other reconfiguration.
420 * These locks are separate due to conflicting interactions
421 * with disk->open_mutex.
423 * reconfig_mutex -> disk->open_mutex
424 * disk->open_mutex -> open_mutex: e.g. __blkdev_get -> md_open
426 struct mutex open_mutex;
427 struct mutex reconfig_mutex;
428 atomic_t active; /* general refcount */
429 atomic_t openers; /* number of active opens */
431 int changed; /* True if we might need to
432 * reread partition info */
433 int degraded; /* whether md should consider
437 atomic_t recovery_active; /* blocks scheduled, but not written */
438 wait_queue_head_t recovery_wait;
439 sector_t recovery_cp;
440 sector_t resync_min; /* user requested sync
442 sector_t resync_max; /* resync should pause
443 * when it gets here */
445 struct kernfs_node *sysfs_state; /* handle for 'array_state'
448 struct kernfs_node *sysfs_action; /* handle for 'sync_action' */
449 struct kernfs_node *sysfs_completed; /*handle for 'sync_completed' */
450 struct kernfs_node *sysfs_degraded; /*handle for 'degraded' */
451 struct kernfs_node *sysfs_level; /*handle for 'level' */
453 struct work_struct del_work; /* used for delayed sysfs removal */
456 * flush_bio transition from NULL to !NULL
457 * rdev superblocks, events
458 * clearing MD_CHANGE_*
459 * in_sync - and related safemode and MD_CHANGE changes
460 * pers (also protected by reconfig_mutex and pending IO).
462 * clearing ->bitmap_info.file
463 * changing ->resync_{min,max}
464 * setting MD_RECOVERY_RUNNING (which interacts with resync_{min,max})
467 wait_queue_head_t sb_wait; /* for waiting on superblock updates */
468 atomic_t pending_writes; /* number of active superblock writes */
470 unsigned int safemode; /* if set, update "clean" superblock
471 * when no writes pending.
473 unsigned int safemode_delay;
474 struct timer_list safemode_timer;
475 struct percpu_ref writes_pending;
476 int sync_checkers; /* # of threads checking writes_pending */
477 struct request_queue *queue; /* for plugging ... */
479 struct bitmap *bitmap; /* the bitmap for the device */
481 struct file *file; /* the bitmap file */
482 loff_t offset; /* offset from superblock of
483 * start of bitmap. May be
484 * negative, but not '0'
485 * For external metadata, offset
486 * from start of device.
488 unsigned long space; /* space available at this offset */
489 loff_t default_offset; /* this is the offset to use when
490 * hot-adding a bitmap. It should
491 * eventually be settable by sysfs.
493 unsigned long default_space; /* space available at
496 unsigned long chunksize;
497 unsigned long daemon_sleep; /* how many jiffies between updates? */
498 unsigned long max_write_behind; /* write-behind mode */
500 int nodes; /* Maximum number of nodes in the cluster */
501 char cluster_name[64]; /* Name of the cluster */
504 atomic_t max_corr_read_errors; /* max read retries */
505 struct list_head all_mddevs;
507 const struct attribute_group *to_remove;
509 struct bio_set bio_set;
510 struct bio_set sync_set; /* for sync operations like
511 * metadata and bitmap writes
513 struct bio_set io_acct_set; /* for raid0 and raid5 io accounting */
515 /* Generic flush handling.
516 * The last to finish preflush schedules a worker to submit
517 * the rest of the request (without the REQ_PREFLUSH flag).
519 struct bio *flush_bio;
520 atomic_t flush_pending;
521 ktime_t start_flush, prev_flush_start; /* prev_flush_start is when the previous completed
524 struct work_struct flush_work;
525 struct work_struct event_work; /* used by dm to report failure event */
526 mempool_t *serial_info_pool;
527 void (*sync_super)(struct mddev *mddev, struct md_rdev *rdev);
528 struct md_cluster_info *cluster_info;
529 unsigned int good_device_nr; /* good device num within cluster raid */
530 unsigned int noio_flag; /* for memalloc scope API */
533 * Temporarily store rdev that will be finally removed when
534 * reconfig_mutex is unlocked, protected by reconfig_mutex.
536 struct list_head deleting;
538 /* Used to synchronize idle and frozen for action_store() */
539 struct mutex sync_mutex;
541 bool has_superblocks:1;
542 bool fail_last_dev:1;
543 bool serialize_policy:1;
546 enum recovery_flags {
548 * If neither SYNC or RESHAPE are set, then it is a recovery.
550 MD_RECOVERY_RUNNING, /* a thread is running, or about to be started */
551 MD_RECOVERY_SYNC, /* actually doing a resync, not a recovery */
552 MD_RECOVERY_RECOVER, /* doing recovery, or need to try it. */
553 MD_RECOVERY_INTR, /* resync needs to be aborted for some reason */
554 MD_RECOVERY_DONE, /* thread is done and is waiting to be reaped */
555 MD_RECOVERY_NEEDED, /* we might need to start a resync/recover */
556 MD_RECOVERY_REQUESTED, /* user-space has requested a sync (used with SYNC) */
557 MD_RECOVERY_CHECK, /* user-space request for check-only, no repair */
558 MD_RECOVERY_RESHAPE, /* A reshape is happening */
559 MD_RECOVERY_FROZEN, /* User request to abort, and not restart, any action */
560 MD_RECOVERY_ERROR, /* sync-action interrupted because io-error */
561 MD_RECOVERY_WAIT, /* waiting for pers->start() to finish */
562 MD_RESYNCING_REMOTE, /* remote node is running resync thread */
572 static inline bool md_is_rdwr(struct mddev *mddev)
574 return (mddev->ro == MD_RDWR);
577 static inline bool is_md_suspended(struct mddev *mddev)
579 return percpu_ref_is_dying(&mddev->active_io);
582 static inline int __must_check mddev_lock(struct mddev *mddev)
584 return mutex_lock_interruptible(&mddev->reconfig_mutex);
587 /* Sometimes we need to take the lock in a situation where
588 * failure due to interrupts is not acceptable.
590 static inline void mddev_lock_nointr(struct mddev *mddev)
592 mutex_lock(&mddev->reconfig_mutex);
595 static inline int mddev_trylock(struct mddev *mddev)
597 return mutex_trylock(&mddev->reconfig_mutex);
599 extern void mddev_unlock(struct mddev *mddev);
601 static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors)
603 atomic_add(nr_sectors, &bdev->bd_disk->sync_io);
606 static inline void md_sync_acct_bio(struct bio *bio, unsigned long nr_sectors)
608 md_sync_acct(bio->bi_bdev, nr_sectors);
611 struct md_personality
615 struct list_head list;
616 struct module *owner;
617 bool __must_check (*make_request)(struct mddev *mddev, struct bio *bio);
619 * start up works that do NOT require md_thread. tasks that
620 * requires md_thread should go into start()
622 int (*run)(struct mddev *mddev);
623 /* start up works that require md threads */
624 int (*start)(struct mddev *mddev);
625 void (*free)(struct mddev *mddev, void *priv);
626 void (*status)(struct seq_file *seq, struct mddev *mddev);
627 /* error_handler must set ->faulty and clear ->in_sync
628 * if appropriate, and should abort recovery if needed
630 void (*error_handler)(struct mddev *mddev, struct md_rdev *rdev);
631 int (*hot_add_disk) (struct mddev *mddev, struct md_rdev *rdev);
632 int (*hot_remove_disk) (struct mddev *mddev, struct md_rdev *rdev);
633 int (*spare_active) (struct mddev *mddev);
634 sector_t (*sync_request)(struct mddev *mddev, sector_t sector_nr, int *skipped);
635 int (*resize) (struct mddev *mddev, sector_t sectors);
636 sector_t (*size) (struct mddev *mddev, sector_t sectors, int raid_disks);
637 int (*check_reshape) (struct mddev *mddev);
638 int (*start_reshape) (struct mddev *mddev);
639 void (*finish_reshape) (struct mddev *mddev);
640 void (*update_reshape_pos) (struct mddev *mddev);
641 void (*prepare_suspend) (struct mddev *mddev);
642 /* quiesce suspends or resumes internal processing.
643 * 1 - stop new actions and wait for action io to complete
644 * 0 - return to normal behaviour
646 void (*quiesce) (struct mddev *mddev, int quiesce);
647 /* takeover is used to transition an array from one
648 * personality to another. The new personality must be able
649 * to handle the data in the current layout.
650 * e.g. 2drive raid1 -> 2drive raid5
651 * ndrive raid5 -> degraded n+1drive raid6 with special layout
652 * If the takeover succeeds, a new 'private' structure is returned.
653 * This needs to be installed and then ->run used to activate the
656 void *(*takeover) (struct mddev *mddev);
657 /* Changes the consistency policy of an active array. */
658 int (*change_consistency_policy)(struct mddev *mddev, const char *buf);
661 struct md_sysfs_entry {
662 struct attribute attr;
663 ssize_t (*show)(struct mddev *, char *);
664 ssize_t (*store)(struct mddev *, const char *, size_t);
666 extern const struct attribute_group md_bitmap_group;
668 static inline struct kernfs_node *sysfs_get_dirent_safe(struct kernfs_node *sd, char *name)
671 return sysfs_get_dirent(sd, name);
674 static inline void sysfs_notify_dirent_safe(struct kernfs_node *sd)
677 sysfs_notify_dirent(sd);
680 static inline char * mdname (struct mddev * mddev)
682 return mddev->gendisk ? mddev->gendisk->disk_name : "mdX";
685 static inline int sysfs_link_rdev(struct mddev *mddev, struct md_rdev *rdev)
688 if (!test_bit(Replacement, &rdev->flags) &&
689 !test_bit(Journal, &rdev->flags) &&
691 sprintf(nm, "rd%d", rdev->raid_disk);
692 return sysfs_create_link(&mddev->kobj, &rdev->kobj, nm);
697 static inline void sysfs_unlink_rdev(struct mddev *mddev, struct md_rdev *rdev)
700 if (!test_bit(Replacement, &rdev->flags) &&
701 !test_bit(Journal, &rdev->flags) &&
703 sprintf(nm, "rd%d", rdev->raid_disk);
704 sysfs_remove_link(&mddev->kobj, nm);
709 * iterates through some rdev ringlist. It's safe to remove the
710 * current 'rdev'. Dont touch 'tmp' though.
712 #define rdev_for_each_list(rdev, tmp, head) \
713 list_for_each_entry_safe(rdev, tmp, head, same_set)
716 * iterates through the 'same array disks' ringlist
718 #define rdev_for_each(rdev, mddev) \
719 list_for_each_entry(rdev, &((mddev)->disks), same_set)
721 #define rdev_for_each_safe(rdev, tmp, mddev) \
722 list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set)
724 #define rdev_for_each_rcu(rdev, mddev) \
725 list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set)
728 void (*run) (struct md_thread *thread);
730 wait_queue_head_t wqueue;
732 struct task_struct *tsk;
733 unsigned long timeout;
739 struct bio *orig_bio;
740 unsigned long start_time;
741 struct bio bio_clone;
744 #define THREAD_WAKEUP 0
746 static inline void safe_put_page(struct page *p)
751 extern int register_md_personality(struct md_personality *p);
752 extern int unregister_md_personality(struct md_personality *p);
753 extern int register_md_cluster_operations(struct md_cluster_operations *ops,
754 struct module *module);
755 extern int unregister_md_cluster_operations(void);
756 extern int md_setup_cluster(struct mddev *mddev, int nodes);
757 extern void md_cluster_stop(struct mddev *mddev);
758 extern struct md_thread *md_register_thread(
759 void (*run)(struct md_thread *thread),
762 extern void md_unregister_thread(struct md_thread __rcu **threadp);
763 extern void md_wakeup_thread(struct md_thread __rcu *thread);
764 extern void md_check_recovery(struct mddev *mddev);
765 extern void md_reap_sync_thread(struct mddev *mddev);
766 extern int mddev_init_writes_pending(struct mddev *mddev);
767 extern bool md_write_start(struct mddev *mddev, struct bio *bi);
768 extern void md_write_inc(struct mddev *mddev, struct bio *bi);
769 extern void md_write_end(struct mddev *mddev);
770 extern void md_done_sync(struct mddev *mddev, int blocks, int ok);
771 extern void md_error(struct mddev *mddev, struct md_rdev *rdev);
772 extern void md_finish_reshape(struct mddev *mddev);
773 void md_submit_discard_bio(struct mddev *mddev, struct md_rdev *rdev,
774 struct bio *bio, sector_t start, sector_t size);
775 int acct_bioset_init(struct mddev *mddev);
776 void acct_bioset_exit(struct mddev *mddev);
777 void md_account_bio(struct mddev *mddev, struct bio **bio);
779 extern bool __must_check md_flush_request(struct mddev *mddev, struct bio *bio);
780 extern void md_super_write(struct mddev *mddev, struct md_rdev *rdev,
781 sector_t sector, int size, struct page *page);
782 extern int md_super_wait(struct mddev *mddev);
783 extern int sync_page_io(struct md_rdev *rdev, sector_t sector, int size,
784 struct page *page, blk_opf_t opf, bool metadata_op);
785 extern void md_do_sync(struct md_thread *thread);
786 extern void md_new_event(void);
787 extern void md_allow_write(struct mddev *mddev);
788 extern void md_wait_for_blocked_rdev(struct md_rdev *rdev, struct mddev *mddev);
789 extern void md_set_array_sectors(struct mddev *mddev, sector_t array_sectors);
790 extern int md_check_no_bitmap(struct mddev *mddev);
791 extern int md_integrity_register(struct mddev *mddev);
792 extern int md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev);
793 extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale);
795 extern void mddev_init(struct mddev *mddev);
796 struct mddev *md_alloc(dev_t dev, char *name);
797 void mddev_put(struct mddev *mddev);
798 extern int md_run(struct mddev *mddev);
799 extern int md_start(struct mddev *mddev);
800 extern void md_stop(struct mddev *mddev);
801 extern void md_stop_writes(struct mddev *mddev);
802 extern int md_rdev_init(struct md_rdev *rdev);
803 extern void md_rdev_clear(struct md_rdev *rdev);
805 extern void md_handle_request(struct mddev *mddev, struct bio *bio);
806 extern void mddev_suspend(struct mddev *mddev);
807 extern void mddev_resume(struct mddev *mddev);
809 extern void md_reload_sb(struct mddev *mddev, int raid_disk);
810 extern void md_update_sb(struct mddev *mddev, int force);
811 extern void mddev_create_serial_pool(struct mddev *mddev, struct md_rdev *rdev,
813 extern void mddev_destroy_serial_pool(struct mddev *mddev, struct md_rdev *rdev,
815 struct md_rdev *md_find_rdev_nr_rcu(struct mddev *mddev, int nr);
816 struct md_rdev *md_find_rdev_rcu(struct mddev *mddev, dev_t dev);
818 static inline bool is_rdev_broken(struct md_rdev *rdev)
820 return !disk_live(rdev->bdev->bd_disk);
823 static inline void rdev_dec_pending(struct md_rdev *rdev, struct mddev *mddev)
825 int faulty = test_bit(Faulty, &rdev->flags);
826 if (atomic_dec_and_test(&rdev->nr_pending) && faulty) {
827 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
828 md_wakeup_thread(mddev->thread);
832 extern struct md_cluster_operations *md_cluster_ops;
833 static inline int mddev_is_clustered(struct mddev *mddev)
835 return mddev->cluster_info && mddev->bitmap_info.nodes > 1;
838 /* clear unsupported mddev_flags */
839 static inline void mddev_clear_unsupported_flags(struct mddev *mddev,
840 unsigned long unsupported_flags)
842 mddev->flags &= ~unsupported_flags;
845 static inline void mddev_check_write_zeroes(struct mddev *mddev, struct bio *bio)
847 if (bio_op(bio) == REQ_OP_WRITE_ZEROES &&
848 !bio->bi_bdev->bd_disk->queue->limits.max_write_zeroes_sectors)
849 mddev->queue->limits.max_write_zeroes_sectors = 0;
852 struct mdu_array_info_s;
853 struct mdu_disk_info_s;
855 extern int mdp_major;
856 extern struct workqueue_struct *md_bitmap_wq;
857 void md_autostart_arrays(int part);
858 int md_set_array_info(struct mddev *mddev, struct mdu_array_info_s *info);
859 int md_add_new_disk(struct mddev *mddev, struct mdu_disk_info_s *info);
860 int do_md_run(struct mddev *mddev);
862 extern const struct block_device_operations md_fops;
864 #endif /* _MD_MD_H */