1 /*
2 md.h : kernel internal structure of the Linux MD driver
3 Copyright (C) 1996-98 Ingo Molnar, Gadi Oxman
4
5 This program is free software; you can redistribute it and/or modify
6 it under the terms of the GNU General Public License as published by
7 the Free Software Foundation; either version 2, or (at your option)
8 any later version.
9
10 You should have received a copy of the GNU General Public License
11 (for example /usr/src/linux/COPYING); if not, write to the Free
12 Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
13 */
14
15 #ifndef _MD_MD_H
16 #define _MD_MD_H
17
18 #include <linux/blkdev.h>
19 #include <linux/backing-dev.h>
20 #include <linux/badblocks.h>
21 #include <linux/kobject.h>
22 #include <linux/list.h>
23 #include <linux/mm.h>
24 #include <linux/mutex.h>
25 #include <linux/timer.h>
26 #include <linux/wait.h>
27 #include <linux/workqueue.h>
28 #include "md-cluster.h"
29
30 #define MaxSector (~(sector_t)0)
31
32 /*
33 * MD's 'extended' device
34 */
35 struct md_rdev {
36 struct list_head same_set; /* RAID devices within the same set */
37
38 sector_t sectors; /* Device size (in 512bytes sectors) */
39 struct mddev *mddev; /* RAID array if running */
40 int last_events; /* IO event timestamp */
41
42 /*
43 * If meta_bdev is non-NULL, it means that a separate device is
44 * being used to store the metadata (superblock/bitmap) which
45 * would otherwise be contained on the same device as the data (bdev).
46 */
47 struct block_device *meta_bdev;
48 struct block_device *bdev; /* block device handle */
49
50 struct page *sb_page, *bb_page;
51 int sb_loaded;
52 __u64 sb_events;
53 sector_t data_offset; /* start of data in array */
54 sector_t new_data_offset;/* only relevant while reshaping */
55 sector_t sb_start; /* offset of the super block (in 512byte sectors) */
56 int sb_size; /* bytes in the superblock */
57 int preferred_minor; /* autorun support */
58
59 struct kobject kobj;
60
61 /* A device can be in one of three states based on two flags:
62 * Not working: faulty==1 in_sync==0
63 * Fully working: faulty==0 in_sync==1
64 * Working, but not
65 * in sync with array
66 * faulty==0 in_sync==0
67 *
68 * It can never have faulty==1, in_sync==1
69 * This reduces the burden of testing multiple flags in many cases
70 */
71
72 unsigned long flags; /* bit set of 'enum flag_bits' bits. */
73 wait_queue_head_t blocked_wait;
74
75 int desc_nr; /* descriptor index in the superblock */
76 int raid_disk; /* role of device in array */
77 int new_raid_disk; /* role that the device will have in
78 * the array after a level-change completes.
79 */
80 int saved_raid_disk; /* role that device used to have in the
81 * array and could again if we did a partial
82 * resync from the bitmap
83 */
84 union {
85 sector_t recovery_offset;/* If this device has been partially
86 * recovered, this is where we were
87 * up to.
88 */
89 sector_t journal_tail; /* If this device is a journal device,
90 * this is the journal tail (journal
91 * recovery start point)
92 */
93 };
94
95 atomic_t nr_pending; /* number of pending requests.
96 * only maintained for arrays that
97 * support hot removal
98 */
99 atomic_t read_errors; /* number of consecutive read errors that
100 * we have tried to ignore.
101 */
102 time64_t last_read_error; /* monotonic time since our
103 * last read error
104 */
105 atomic_t corrected_errors; /* number of corrected read errors,
106 * for reporting to userspace and storing
107 * in superblock.
108 */
109 struct work_struct del_work; /* used for delayed sysfs removal */
110
111 struct kernfs_node *sysfs_state; /* handle for 'state'
112 * sysfs entry */
113
114 struct badblocks badblocks;
115 };
116 enum flag_bits {
117 Faulty, /* device is known to have a fault */
118 In_sync, /* device is in_sync with rest of array */
119 Bitmap_sync, /* ..actually, not quite In_sync. Need a
120 * bitmap-based recovery to get fully in sync
121 */
122 WriteMostly, /* Avoid reading if at all possible */
123 AutoDetected, /* added by auto-detect */
124 Blocked, /* An error occurred but has not yet
125 * been acknowledged by the metadata
126 * handler, so don't allow writes
127 * until it is cleared */
128 WriteErrorSeen, /* A write error has been seen on this
129 * device
130 */
131 FaultRecorded, /* Intermediate state for clearing
132 * Blocked. The Fault is/will-be
133 * recorded in the metadata, but that
134 * metadata hasn't been stored safely
135 * on disk yet.
136 */
137 BlockedBadBlocks, /* A writer is blocked because they
138 * found an unacknowledged bad-block.
139 * This can safely be cleared at any
140 * time, and the writer will re-check.
141 * It may be set at any time, and at
142 * worst the writer will timeout and
143 * re-check. So setting it as
144 * accurately as possible is good, but
145 * not absolutely critical.
146 */
147 WantReplacement, /* This device is a candidate to be
148 * hot-replaced, either because it has
149 * reported some faults, or because
150 * of explicit request.
151 */
152 Replacement, /* This device is a replacement for
153 * a want_replacement device with same
154 * raid_disk number.
155 */
156 Candidate, /* For clustered environments only:
157 * This device is seen locally but not
158 * by the whole cluster
159 */
160 Journal, /* This device is used as journal for
161 * raid-5/6.
162 * Usually, this device should be faster
163 * than other devices in the array
164 */
165 ClusterRemove,
166 RemoveSynchronized, /* synchronize_rcu() was called after
167 * this device was known to be faulty,
168 * so it is safe to remove without
169 * another synchronize_rcu() call.
170 */
171 };
172
is_badblock(struct md_rdev * rdev,sector_t s,int sectors,sector_t * first_bad,int * bad_sectors)173 static inline int is_badblock(struct md_rdev *rdev, sector_t s, int sectors,
174 sector_t *first_bad, int *bad_sectors)
175 {
176 if (unlikely(rdev->badblocks.count)) {
177 int rv = badblocks_check(&rdev->badblocks, rdev->data_offset + s,
178 sectors,
179 first_bad, bad_sectors);
180 if (rv)
181 *first_bad -= rdev->data_offset;
182 return rv;
183 }
184 return 0;
185 }
186 extern int rdev_set_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
187 int is_new);
188 extern int rdev_clear_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
189 int is_new);
190 struct md_cluster_info;
191
192 struct mddev {
193 void *private;
194 struct md_personality *pers;
195 dev_t unit;
196 int md_minor;
197 struct list_head disks;
198 unsigned long flags;
199 #define MD_CHANGE_DEVS 0 /* Some device status has changed */
200 #define MD_CHANGE_CLEAN 1 /* transition to or from 'clean' */
201 #define MD_CHANGE_PENDING 2 /* switch from 'clean' to 'active' in progress */
202 #define MD_UPDATE_SB_FLAGS (1 | 2 | 4) /* If these are set, md_update_sb needed */
203 #define MD_ARRAY_FIRST_USE 3 /* First use of array, needs initialization */
204 #define MD_CLOSING 4 /* If set, we are closing the array, do not open
205 * it then */
206 #define MD_JOURNAL_CLEAN 5 /* A raid with journal is already clean */
207 #define MD_HAS_JOURNAL 6 /* The raid array has journal feature set */
208 #define MD_RELOAD_SB 7 /* Reload the superblock because another node
209 * updated it.
210 */
211 #define MD_CLUSTER_RESYNC_LOCKED 8 /* cluster raid only, which means node
212 * already took resync lock, need to
213 * release the lock */
214
215 int suspended;
216 atomic_t active_io;
217 int ro;
218 int sysfs_active; /* set when sysfs deletes
219 * are happening, so run/
220 * takeover/stop are not safe
221 */
222 struct gendisk *gendisk;
223
224 struct kobject kobj;
225 int hold_active;
226 #define UNTIL_IOCTL 1
227 #define UNTIL_STOP 2
228
229 /* Superblock information */
230 int major_version,
231 minor_version,
232 patch_version;
233 int persistent;
234 int external; /* metadata is
235 * managed externally */
236 char metadata_type[17]; /* externally set*/
237 int chunk_sectors;
238 time64_t ctime, utime;
239 int level, layout;
240 char clevel[16];
241 int raid_disks;
242 int max_disks;
243 sector_t dev_sectors; /* used size of
244 * component devices */
245 sector_t array_sectors; /* exported array size */
246 int external_size; /* size managed
247 * externally */
248 __u64 events;
249 /* If the last 'event' was simply a clean->dirty transition, and
250 * we didn't write it to the spares, then it is safe and simple
251 * to just decrement the event count on a dirty->clean transition.
252 * So we record that possibility here.
253 */
254 int can_decrease_events;
255
256 char uuid[16];
257
258 /* If the array is being reshaped, we need to record the
259 * new shape and an indication of where we are up to.
260 * This is written to the superblock.
261 * If reshape_position is MaxSector, then no reshape is happening (yet).
262 */
263 sector_t reshape_position;
264 int delta_disks, new_level, new_layout;
265 int new_chunk_sectors;
266 int reshape_backwards;
267
268 struct md_thread *thread; /* management thread */
269 struct md_thread *sync_thread; /* doing resync or reconstruct */
270
271 /* 'last_sync_action' is initialized to "none". It is set when a
272 * sync operation (i.e "data-check", "requested-resync", "resync",
273 * "recovery", or "reshape") is started. It holds this value even
274 * when the sync thread is "frozen" (interrupted) or "idle" (stopped
275 * or finished). It is overwritten when a new sync operation is begun.
276 */
277 char *last_sync_action;
278 sector_t curr_resync; /* last block scheduled */
279 /* As resync requests can complete out of order, we cannot easily track
280 * how much resync has been completed. So we occasionally pause until
281 * everything completes, then set curr_resync_completed to curr_resync.
282 * As such it may be well behind the real resync mark, but it is a value
283 * we are certain of.
284 */
285 sector_t curr_resync_completed;
286 unsigned long resync_mark; /* a recent timestamp */
287 sector_t resync_mark_cnt;/* blocks written at resync_mark */
288 sector_t curr_mark_cnt; /* blocks scheduled now */
289
290 sector_t resync_max_sectors; /* may be set by personality */
291
292 atomic64_t resync_mismatches; /* count of sectors where
293 * parity/replica mismatch found
294 */
295
296 /* allow user-space to request suspension of IO to regions of the array */
297 sector_t suspend_lo;
298 sector_t suspend_hi;
299 /* if zero, use the system-wide default */
300 int sync_speed_min;
301 int sync_speed_max;
302
303 /* resync even though the same disks are shared among md-devices */
304 int parallel_resync;
305
306 int ok_start_degraded;
307 /* recovery/resync flags
308 * NEEDED: we might need to start a resync/recover
309 * RUNNING: a thread is running, or about to be started
310 * SYNC: actually doing a resync, not a recovery
311 * RECOVER: doing recovery, or need to try it.
312 * INTR: resync needs to be aborted for some reason
313 * DONE: thread is done and is waiting to be reaped
314 * REQUEST: user-space has requested a sync (used with SYNC)
315 * CHECK: user-space request for check-only, no repair
316 * RESHAPE: A reshape is happening
317 * ERROR: sync-action interrupted because io-error
318 *
319 * If neither SYNC or RESHAPE are set, then it is a recovery.
320 */
321 #define MD_RECOVERY_RUNNING 0
322 #define MD_RECOVERY_SYNC 1
323 #define MD_RECOVERY_RECOVER 2
324 #define MD_RECOVERY_INTR 3
325 #define MD_RECOVERY_DONE 4
326 #define MD_RECOVERY_NEEDED 5
327 #define MD_RECOVERY_REQUESTED 6
328 #define MD_RECOVERY_CHECK 7
329 #define MD_RECOVERY_RESHAPE 8
330 #define MD_RECOVERY_FROZEN 9
331 #define MD_RECOVERY_ERROR 10
332
333 unsigned long recovery;
334 /* If a RAID personality determines that recovery (of a particular
335 * device) will fail due to a read error on the source device, it
336 * takes a copy of this number and does not attempt recovery again
337 * until this number changes.
338 */
339 int recovery_disabled;
340
341 int in_sync; /* know to not need resync */
342 /* 'open_mutex' avoids races between 'md_open' and 'do_md_stop', so
343 * that we are never stopping an array while it is open.
344 * 'reconfig_mutex' protects all other reconfiguration.
345 * These locks are separate due to conflicting interactions
346 * with bdev->bd_mutex.
347 * Lock ordering is:
348 * reconfig_mutex -> bd_mutex : e.g. do_md_run -> revalidate_disk
349 * bd_mutex -> open_mutex: e.g. __blkdev_get -> md_open
350 */
351 struct mutex open_mutex;
352 struct mutex reconfig_mutex;
353 atomic_t active; /* general refcount */
354 atomic_t openers; /* number of active opens */
355
356 int changed; /* True if we might need to
357 * reread partition info */
358 int degraded; /* whether md should consider
359 * adding a spare
360 */
361
362 atomic_t recovery_active; /* blocks scheduled, but not written */
363 wait_queue_head_t recovery_wait;
364 sector_t recovery_cp;
365 sector_t resync_min; /* user requested sync
366 * starts here */
367 sector_t resync_max; /* resync should pause
368 * when it gets here */
369
370 struct kernfs_node *sysfs_state; /* handle for 'array_state'
371 * file in sysfs.
372 */
373 struct kernfs_node *sysfs_action; /* handle for 'sync_action' */
374
375 struct work_struct del_work; /* used for delayed sysfs removal */
376
377 /* "lock" protects:
378 * flush_bio transition from NULL to !NULL
379 * rdev superblocks, events
380 * clearing MD_CHANGE_*
381 * in_sync - and related safemode and MD_CHANGE changes
382 * pers (also protected by reconfig_mutex and pending IO).
383 * clearing ->bitmap
384 * clearing ->bitmap_info.file
385 * changing ->resync_{min,max}
386 * setting MD_RECOVERY_RUNNING (which interacts with resync_{min,max})
387 */
388 spinlock_t lock;
389 wait_queue_head_t sb_wait; /* for waiting on superblock updates */
390 atomic_t pending_writes; /* number of active superblock writes */
391
392 unsigned int safemode; /* if set, update "clean" superblock
393 * when no writes pending.
394 */
395 unsigned int safemode_delay;
396 struct timer_list safemode_timer;
397 atomic_t writes_pending;
398 struct request_queue *queue; /* for plugging ... */
399
400 struct bitmap *bitmap; /* the bitmap for the device */
401 struct {
402 struct file *file; /* the bitmap file */
403 loff_t offset; /* offset from superblock of
404 * start of bitmap. May be
405 * negative, but not '0'
406 * For external metadata, offset
407 * from start of device.
408 */
409 unsigned long space; /* space available at this offset */
410 loff_t default_offset; /* this is the offset to use when
411 * hot-adding a bitmap. It should
412 * eventually be settable by sysfs.
413 */
414 unsigned long default_space; /* space available at
415 * default offset */
416 struct mutex mutex;
417 unsigned long chunksize;
418 unsigned long daemon_sleep; /* how many jiffies between updates? */
419 unsigned long max_write_behind; /* write-behind mode */
420 int external;
421 int nodes; /* Maximum number of nodes in the cluster */
422 char cluster_name[64]; /* Name of the cluster */
423 } bitmap_info;
424
425 atomic_t max_corr_read_errors; /* max read retries */
426 struct list_head all_mddevs;
427
428 struct attribute_group *to_remove;
429
430 struct bio_set *bio_set;
431
432 /* Generic flush handling.
433 * The last to finish preflush schedules a worker to submit
434 * the rest of the request (without the REQ_PREFLUSH flag).
435 */
436 struct bio *flush_bio;
437 atomic_t flush_pending;
438 struct work_struct flush_work;
439 struct work_struct event_work; /* used by dm to report failure event */
440 void (*sync_super)(struct mddev *mddev, struct md_rdev *rdev);
441 struct md_cluster_info *cluster_info;
442 unsigned int good_device_nr; /* good device num within cluster raid */
443 };
444
mddev_lock(struct mddev * mddev)445 static inline int __must_check mddev_lock(struct mddev *mddev)
446 {
447 return mutex_lock_interruptible(&mddev->reconfig_mutex);
448 }
449
450 /* Sometimes we need to take the lock in a situation where
451 * failure due to interrupts is not acceptable.
452 */
mddev_lock_nointr(struct mddev * mddev)453 static inline void mddev_lock_nointr(struct mddev *mddev)
454 {
455 mutex_lock(&mddev->reconfig_mutex);
456 }
457
mddev_is_locked(struct mddev * mddev)458 static inline int mddev_is_locked(struct mddev *mddev)
459 {
460 return mutex_is_locked(&mddev->reconfig_mutex);
461 }
462
mddev_trylock(struct mddev * mddev)463 static inline int mddev_trylock(struct mddev *mddev)
464 {
465 return mutex_trylock(&mddev->reconfig_mutex);
466 }
467 extern void mddev_unlock(struct mddev *mddev);
468
md_sync_acct(struct block_device * bdev,unsigned long nr_sectors)469 static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors)
470 {
471 atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io);
472 }
473
474 struct md_personality
475 {
476 char *name;
477 int level;
478 struct list_head list;
479 struct module *owner;
480 void (*make_request)(struct mddev *mddev, struct bio *bio);
481 int (*run)(struct mddev *mddev);
482 void (*free)(struct mddev *mddev, void *priv);
483 void (*status)(struct seq_file *seq, struct mddev *mddev);
484 /* error_handler must set ->faulty and clear ->in_sync
485 * if appropriate, and should abort recovery if needed
486 */
487 void (*error_handler)(struct mddev *mddev, struct md_rdev *rdev);
488 int (*hot_add_disk) (struct mddev *mddev, struct md_rdev *rdev);
489 int (*hot_remove_disk) (struct mddev *mddev, struct md_rdev *rdev);
490 int (*spare_active) (struct mddev *mddev);
491 sector_t (*sync_request)(struct mddev *mddev, sector_t sector_nr, int *skipped);
492 int (*resize) (struct mddev *mddev, sector_t sectors);
493 sector_t (*size) (struct mddev *mddev, sector_t sectors, int raid_disks);
494 int (*check_reshape) (struct mddev *mddev);
495 int (*start_reshape) (struct mddev *mddev);
496 void (*finish_reshape) (struct mddev *mddev);
497 /* quiesce moves between quiescence states
498 * 0 - fully active
499 * 1 - no new requests allowed
500 * others - reserved
501 */
502 void (*quiesce) (struct mddev *mddev, int state);
503 /* takeover is used to transition an array from one
504 * personality to another. The new personality must be able
505 * to handle the data in the current layout.
506 * e.g. 2drive raid1 -> 2drive raid5
507 * ndrive raid5 -> degraded n+1drive raid6 with special layout
508 * If the takeover succeeds, a new 'private' structure is returned.
509 * This needs to be installed and then ->run used to activate the
510 * array.
511 */
512 void *(*takeover) (struct mddev *mddev);
513 /* congested implements bdi.congested_fn().
514 * Will not be called while array is 'suspended' */
515 int (*congested)(struct mddev *mddev, int bits);
516 };
517
518 struct md_sysfs_entry {
519 struct attribute attr;
520 ssize_t (*show)(struct mddev *, char *);
521 ssize_t (*store)(struct mddev *, const char *, size_t);
522 };
523 extern struct attribute_group md_bitmap_group;
524
sysfs_get_dirent_safe(struct kernfs_node * sd,char * name)525 static inline struct kernfs_node *sysfs_get_dirent_safe(struct kernfs_node *sd, char *name)
526 {
527 if (sd)
528 return sysfs_get_dirent(sd, name);
529 return sd;
530 }
sysfs_notify_dirent_safe(struct kernfs_node * sd)531 static inline void sysfs_notify_dirent_safe(struct kernfs_node *sd)
532 {
533 if (sd)
534 sysfs_notify_dirent(sd);
535 }
536
mdname(struct mddev * mddev)537 static inline char * mdname (struct mddev * mddev)
538 {
539 return mddev->gendisk ? mddev->gendisk->disk_name : "mdX";
540 }
541
sysfs_link_rdev(struct mddev * mddev,struct md_rdev * rdev)542 static inline int sysfs_link_rdev(struct mddev *mddev, struct md_rdev *rdev)
543 {
544 char nm[20];
545 if (!test_bit(Replacement, &rdev->flags) &&
546 !test_bit(Journal, &rdev->flags) &&
547 mddev->kobj.sd) {
548 sprintf(nm, "rd%d", rdev->raid_disk);
549 return sysfs_create_link(&mddev->kobj, &rdev->kobj, nm);
550 } else
551 return 0;
552 }
553
sysfs_unlink_rdev(struct mddev * mddev,struct md_rdev * rdev)554 static inline void sysfs_unlink_rdev(struct mddev *mddev, struct md_rdev *rdev)
555 {
556 char nm[20];
557 if (!test_bit(Replacement, &rdev->flags) &&
558 !test_bit(Journal, &rdev->flags) &&
559 mddev->kobj.sd) {
560 sprintf(nm, "rd%d", rdev->raid_disk);
561 sysfs_remove_link(&mddev->kobj, nm);
562 }
563 }
564
565 /*
566 * iterates through some rdev ringlist. It's safe to remove the
567 * current 'rdev'. Dont touch 'tmp' though.
568 */
569 #define rdev_for_each_list(rdev, tmp, head) \
570 list_for_each_entry_safe(rdev, tmp, head, same_set)
571
572 /*
573 * iterates through the 'same array disks' ringlist
574 */
575 #define rdev_for_each(rdev, mddev) \
576 list_for_each_entry(rdev, &((mddev)->disks), same_set)
577
578 #define rdev_for_each_safe(rdev, tmp, mddev) \
579 list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set)
580
581 #define rdev_for_each_rcu(rdev, mddev) \
582 list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set)
583
584 struct md_thread {
585 void (*run) (struct md_thread *thread);
586 struct mddev *mddev;
587 wait_queue_head_t wqueue;
588 unsigned long flags;
589 struct task_struct *tsk;
590 unsigned long timeout;
591 void *private;
592 };
593
594 #define THREAD_WAKEUP 0
595
safe_put_page(struct page * p)596 static inline void safe_put_page(struct page *p)
597 {
598 if (p) put_page(p);
599 }
600
601 extern int register_md_personality(struct md_personality *p);
602 extern int unregister_md_personality(struct md_personality *p);
603 extern int register_md_cluster_operations(struct md_cluster_operations *ops,
604 struct module *module);
605 extern int unregister_md_cluster_operations(void);
606 extern int md_setup_cluster(struct mddev *mddev, int nodes);
607 extern void md_cluster_stop(struct mddev *mddev);
608 extern struct md_thread *md_register_thread(
609 void (*run)(struct md_thread *thread),
610 struct mddev *mddev,
611 const char *name);
612 extern void md_unregister_thread(struct md_thread **threadp);
613 extern void md_wakeup_thread(struct md_thread *thread);
614 extern void md_check_recovery(struct mddev *mddev);
615 extern void md_reap_sync_thread(struct mddev *mddev);
616 extern void md_write_start(struct mddev *mddev, struct bio *bi);
617 extern void md_write_end(struct mddev *mddev);
618 extern void md_done_sync(struct mddev *mddev, int blocks, int ok);
619 extern void md_error(struct mddev *mddev, struct md_rdev *rdev);
620 extern void md_finish_reshape(struct mddev *mddev);
621
622 extern int mddev_congested(struct mddev *mddev, int bits);
623 extern void md_flush_request(struct mddev *mddev, struct bio *bio);
624 extern void md_super_write(struct mddev *mddev, struct md_rdev *rdev,
625 sector_t sector, int size, struct page *page);
626 extern void md_super_wait(struct mddev *mddev);
627 extern int sync_page_io(struct md_rdev *rdev, sector_t sector, int size,
628 struct page *page, int op, int op_flags,
629 bool metadata_op);
630 extern void md_do_sync(struct md_thread *thread);
631 extern void md_new_event(struct mddev *mddev);
632 extern int md_allow_write(struct mddev *mddev);
633 extern void md_wait_for_blocked_rdev(struct md_rdev *rdev, struct mddev *mddev);
634 extern void md_set_array_sectors(struct mddev *mddev, sector_t array_sectors);
635 extern int md_check_no_bitmap(struct mddev *mddev);
636 extern int md_integrity_register(struct mddev *mddev);
637 extern int md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev);
638 extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale);
639
640 extern void mddev_init(struct mddev *mddev);
641 extern int md_run(struct mddev *mddev);
642 extern void md_stop(struct mddev *mddev);
643 extern void md_stop_writes(struct mddev *mddev);
644 extern int md_rdev_init(struct md_rdev *rdev);
645 extern void md_rdev_clear(struct md_rdev *rdev);
646
647 extern void mddev_suspend(struct mddev *mddev);
648 extern void mddev_resume(struct mddev *mddev);
649 extern struct bio *bio_clone_mddev(struct bio *bio, gfp_t gfp_mask,
650 struct mddev *mddev);
651 extern struct bio *bio_alloc_mddev(gfp_t gfp_mask, int nr_iovecs,
652 struct mddev *mddev);
653
654 extern void md_unplug(struct blk_plug_cb *cb, bool from_schedule);
655 extern void md_reload_sb(struct mddev *mddev, int raid_disk);
656 extern void md_update_sb(struct mddev *mddev, int force);
657 extern void md_kick_rdev_from_array(struct md_rdev * rdev);
658 struct md_rdev *md_find_rdev_nr_rcu(struct mddev *mddev, int nr);
mddev_check_plugged(struct mddev * mddev)659 static inline int mddev_check_plugged(struct mddev *mddev)
660 {
661 return !!blk_check_plugged(md_unplug, mddev,
662 sizeof(struct blk_plug_cb));
663 }
664
rdev_dec_pending(struct md_rdev * rdev,struct mddev * mddev)665 static inline void rdev_dec_pending(struct md_rdev *rdev, struct mddev *mddev)
666 {
667 int faulty = test_bit(Faulty, &rdev->flags);
668 if (atomic_dec_and_test(&rdev->nr_pending) && faulty) {
669 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
670 md_wakeup_thread(mddev->thread);
671 }
672 }
673
674 extern struct md_cluster_operations *md_cluster_ops;
mddev_is_clustered(struct mddev * mddev)675 static inline int mddev_is_clustered(struct mddev *mddev)
676 {
677 return mddev->cluster_info && mddev->bitmap_info.nodes > 1;
678 }
679 #endif /* _MD_MD_H */
680