]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - drivers/md/md.h
MD: change journal disk role to disk 0
[mirror_ubuntu-bionic-kernel.git] / drivers / md / md.h
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/kobject.h>
21 #include <linux/list.h>
22 #include <linux/mm.h>
23 #include <linux/mutex.h>
24 #include <linux/timer.h>
25 #include <linux/wait.h>
26 #include <linux/workqueue.h>
27 #include "md-cluster.h"
28
29 #define MaxSector (~(sector_t)0)
30
31 /* Bad block numbers are stored sorted in a single page.
32 * 64bits is used for each block or extent.
33 * 54 bits are sector number, 9 bits are extent size,
34 * 1 bit is an 'acknowledged' flag.
35 */
36 #define MD_MAX_BADBLOCKS (PAGE_SIZE/8)
37
38 /*
39 * MD's 'extended' device
40 */
41 struct md_rdev {
42 struct list_head same_set; /* RAID devices within the same set */
43
44 sector_t sectors; /* Device size (in 512bytes sectors) */
45 struct mddev *mddev; /* RAID array if running */
46 int last_events; /* IO event timestamp */
47
48 /*
49 * If meta_bdev is non-NULL, it means that a separate device is
50 * being used to store the metadata (superblock/bitmap) which
51 * would otherwise be contained on the same device as the data (bdev).
52 */
53 struct block_device *meta_bdev;
54 struct block_device *bdev; /* block device handle */
55
56 struct page *sb_page, *bb_page;
57 int sb_loaded;
58 __u64 sb_events;
59 sector_t data_offset; /* start of data in array */
60 sector_t new_data_offset;/* only relevant while reshaping */
61 sector_t sb_start; /* offset of the super block (in 512byte sectors) */
62 int sb_size; /* bytes in the superblock */
63 int preferred_minor; /* autorun support */
64
65 struct kobject kobj;
66
67 /* A device can be in one of three states based on two flags:
68 * Not working: faulty==1 in_sync==0
69 * Fully working: faulty==0 in_sync==1
70 * Working, but not
71 * in sync with array
72 * faulty==0 in_sync==0
73 *
74 * It can never have faulty==1, in_sync==1
75 * This reduces the burden of testing multiple flags in many cases
76 */
77
78 unsigned long flags; /* bit set of 'enum flag_bits' bits. */
79 wait_queue_head_t blocked_wait;
80
81 int desc_nr; /* descriptor index in the superblock */
82 int raid_disk; /* role of device in array */
83 int new_raid_disk; /* role that the device will have in
84 * the array after a level-change completes.
85 */
86 int saved_raid_disk; /* role that device used to have in the
87 * array and could again if we did a partial
88 * resync from the bitmap
89 */
90 union {
91 sector_t recovery_offset;/* If this device has been partially
92 * recovered, this is where we were
93 * up to.
94 */
95 sector_t journal_tail; /* If this device is a journal device,
96 * this is the journal tail (journal
97 * recovery start point)
98 */
99 };
100
101 atomic_t nr_pending; /* number of pending requests.
102 * only maintained for arrays that
103 * support hot removal
104 */
105 atomic_t read_errors; /* number of consecutive read errors that
106 * we have tried to ignore.
107 */
108 struct timespec last_read_error; /* monotonic time since our
109 * last read error
110 */
111 atomic_t corrected_errors; /* number of corrected read errors,
112 * for reporting to userspace and storing
113 * in superblock.
114 */
115 struct work_struct del_work; /* used for delayed sysfs removal */
116
117 struct kernfs_node *sysfs_state; /* handle for 'state'
118 * sysfs entry */
119
120 struct badblocks {
121 int count; /* count of bad blocks */
122 int unacked_exist; /* there probably are unacknowledged
123 * bad blocks. This is only cleared
124 * when a read discovers none
125 */
126 int shift; /* shift from sectors to block size
127 * a -ve shift means badblocks are
128 * disabled.*/
129 u64 *page; /* badblock list */
130 int changed;
131 seqlock_t lock;
132
133 sector_t sector;
134 sector_t size; /* in sectors */
135 } badblocks;
136 };
137 enum flag_bits {
138 Faulty, /* device is known to have a fault */
139 In_sync, /* device is in_sync with rest of array */
140 Bitmap_sync, /* ..actually, not quite In_sync. Need a
141 * bitmap-based recovery to get fully in sync
142 */
143 WriteMostly, /* Avoid reading if at all possible */
144 AutoDetected, /* added by auto-detect */
145 Blocked, /* An error occurred but has not yet
146 * been acknowledged by the metadata
147 * handler, so don't allow writes
148 * until it is cleared */
149 WriteErrorSeen, /* A write error has been seen on this
150 * device
151 */
152 FaultRecorded, /* Intermediate state for clearing
153 * Blocked. The Fault is/will-be
154 * recorded in the metadata, but that
155 * metadata hasn't been stored safely
156 * on disk yet.
157 */
158 BlockedBadBlocks, /* A writer is blocked because they
159 * found an unacknowledged bad-block.
160 * This can safely be cleared at any
161 * time, and the writer will re-check.
162 * It may be set at any time, and at
163 * worst the writer will timeout and
164 * re-check. So setting it as
165 * accurately as possible is good, but
166 * not absolutely critical.
167 */
168 WantReplacement, /* This device is a candidate to be
169 * hot-replaced, either because it has
170 * reported some faults, or because
171 * of explicit request.
172 */
173 Replacement, /* This device is a replacement for
174 * a want_replacement device with same
175 * raid_disk number.
176 */
177 Candidate, /* For clustered environments only:
178 * This device is seen locally but not
179 * by the whole cluster
180 */
181 Journal, /* This device is used as journal for
182 * raid-5/6.
183 * Usually, this device should be faster
184 * than other devices in the array
185 */
186 };
187
188 #define BB_LEN_MASK (0x00000000000001FFULL)
189 #define BB_OFFSET_MASK (0x7FFFFFFFFFFFFE00ULL)
190 #define BB_ACK_MASK (0x8000000000000000ULL)
191 #define BB_MAX_LEN 512
192 #define BB_OFFSET(x) (((x) & BB_OFFSET_MASK) >> 9)
193 #define BB_LEN(x) (((x) & BB_LEN_MASK) + 1)
194 #define BB_ACK(x) (!!((x) & BB_ACK_MASK))
195 #define BB_MAKE(a, l, ack) (((a)<<9) | ((l)-1) | ((u64)(!!(ack)) << 63))
196
197 extern int md_is_badblock(struct badblocks *bb, sector_t s, int sectors,
198 sector_t *first_bad, int *bad_sectors);
199 static inline int is_badblock(struct md_rdev *rdev, sector_t s, int sectors,
200 sector_t *first_bad, int *bad_sectors)
201 {
202 if (unlikely(rdev->badblocks.count)) {
203 int rv = md_is_badblock(&rdev->badblocks, rdev->data_offset + s,
204 sectors,
205 first_bad, bad_sectors);
206 if (rv)
207 *first_bad -= rdev->data_offset;
208 return rv;
209 }
210 return 0;
211 }
212 extern int rdev_set_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
213 int is_new);
214 extern int rdev_clear_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
215 int is_new);
216 extern void md_ack_all_badblocks(struct badblocks *bb);
217
218 struct md_cluster_info;
219
220 struct mddev {
221 void *private;
222 struct md_personality *pers;
223 dev_t unit;
224 int md_minor;
225 struct list_head disks;
226 unsigned long flags;
227 #define MD_CHANGE_DEVS 0 /* Some device status has changed */
228 #define MD_CHANGE_CLEAN 1 /* transition to or from 'clean' */
229 #define MD_CHANGE_PENDING 2 /* switch from 'clean' to 'active' in progress */
230 #define MD_UPDATE_SB_FLAGS (1 | 2 | 4) /* If these are set, md_update_sb needed */
231 #define MD_ARRAY_FIRST_USE 3 /* First use of array, needs initialization */
232 #define MD_STILL_CLOSED 4 /* If set, then array has not been opened since
233 * md_ioctl checked on it.
234 */
235 #define MD_JOURNAL_CLEAN 5 /* A raid with journal is already clean */
236 #define MD_HAS_JOURNAL 6 /* The raid array has journal feature set */
237
238 int suspended;
239 atomic_t active_io;
240 int ro;
241 int sysfs_active; /* set when sysfs deletes
242 * are happening, so run/
243 * takeover/stop are not safe
244 */
245 int ready; /* See when safe to pass
246 * IO requests down */
247 struct gendisk *gendisk;
248
249 struct kobject kobj;
250 int hold_active;
251 #define UNTIL_IOCTL 1
252 #define UNTIL_STOP 2
253
254 /* Superblock information */
255 int major_version,
256 minor_version,
257 patch_version;
258 int persistent;
259 int external; /* metadata is
260 * managed externally */
261 char metadata_type[17]; /* externally set*/
262 int chunk_sectors;
263 time_t ctime, utime;
264 int level, layout;
265 char clevel[16];
266 int raid_disks;
267 int max_disks;
268 sector_t dev_sectors; /* used size of
269 * component devices */
270 sector_t array_sectors; /* exported array size */
271 int external_size; /* size managed
272 * externally */
273 __u64 events;
274 /* If the last 'event' was simply a clean->dirty transition, and
275 * we didn't write it to the spares, then it is safe and simple
276 * to just decrement the event count on a dirty->clean transition.
277 * So we record that possibility here.
278 */
279 int can_decrease_events;
280
281 char uuid[16];
282
283 /* If the array is being reshaped, we need to record the
284 * new shape and an indication of where we are up to.
285 * This is written to the superblock.
286 * If reshape_position is MaxSector, then no reshape is happening (yet).
287 */
288 sector_t reshape_position;
289 int delta_disks, new_level, new_layout;
290 int new_chunk_sectors;
291 int reshape_backwards;
292
293 struct md_thread *thread; /* management thread */
294 struct md_thread *sync_thread; /* doing resync or reconstruct */
295
296 /* 'last_sync_action' is initialized to "none". It is set when a
297 * sync operation (i.e "data-check", "requested-resync", "resync",
298 * "recovery", or "reshape") is started. It holds this value even
299 * when the sync thread is "frozen" (interrupted) or "idle" (stopped
300 * or finished). It is overwritten when a new sync operation is begun.
301 */
302 char *last_sync_action;
303 sector_t curr_resync; /* last block scheduled */
304 /* As resync requests can complete out of order, we cannot easily track
305 * how much resync has been completed. So we occasionally pause until
306 * everything completes, then set curr_resync_completed to curr_resync.
307 * As such it may be well behind the real resync mark, but it is a value
308 * we are certain of.
309 */
310 sector_t curr_resync_completed;
311 unsigned long resync_mark; /* a recent timestamp */
312 sector_t resync_mark_cnt;/* blocks written at resync_mark */
313 sector_t curr_mark_cnt; /* blocks scheduled now */
314
315 sector_t resync_max_sectors; /* may be set by personality */
316
317 atomic64_t resync_mismatches; /* count of sectors where
318 * parity/replica mismatch found
319 */
320
321 /* allow user-space to request suspension of IO to regions of the array */
322 sector_t suspend_lo;
323 sector_t suspend_hi;
324 /* if zero, use the system-wide default */
325 int sync_speed_min;
326 int sync_speed_max;
327
328 /* resync even though the same disks are shared among md-devices */
329 int parallel_resync;
330
331 int ok_start_degraded;
332 /* recovery/resync flags
333 * NEEDED: we might need to start a resync/recover
334 * RUNNING: a thread is running, or about to be started
335 * SYNC: actually doing a resync, not a recovery
336 * RECOVER: doing recovery, or need to try it.
337 * INTR: resync needs to be aborted for some reason
338 * DONE: thread is done and is waiting to be reaped
339 * REQUEST: user-space has requested a sync (used with SYNC)
340 * CHECK: user-space request for check-only, no repair
341 * RESHAPE: A reshape is happening
342 * ERROR: sync-action interrupted because io-error
343 *
344 * If neither SYNC or RESHAPE are set, then it is a recovery.
345 */
346 #define MD_RECOVERY_RUNNING 0
347 #define MD_RECOVERY_SYNC 1
348 #define MD_RECOVERY_RECOVER 2
349 #define MD_RECOVERY_INTR 3
350 #define MD_RECOVERY_DONE 4
351 #define MD_RECOVERY_NEEDED 5
352 #define MD_RECOVERY_REQUESTED 6
353 #define MD_RECOVERY_CHECK 7
354 #define MD_RECOVERY_RESHAPE 8
355 #define MD_RECOVERY_FROZEN 9
356 #define MD_RECOVERY_ERROR 10
357
358 unsigned long recovery;
359 /* If a RAID personality determines that recovery (of a particular
360 * device) will fail due to a read error on the source device, it
361 * takes a copy of this number and does not attempt recovery again
362 * until this number changes.
363 */
364 int recovery_disabled;
365
366 int in_sync; /* know to not need resync */
367 /* 'open_mutex' avoids races between 'md_open' and 'do_md_stop', so
368 * that we are never stopping an array while it is open.
369 * 'reconfig_mutex' protects all other reconfiguration.
370 * These locks are separate due to conflicting interactions
371 * with bdev->bd_mutex.
372 * Lock ordering is:
373 * reconfig_mutex -> bd_mutex : e.g. do_md_run -> revalidate_disk
374 * bd_mutex -> open_mutex: e.g. __blkdev_get -> md_open
375 */
376 struct mutex open_mutex;
377 struct mutex reconfig_mutex;
378 atomic_t active; /* general refcount */
379 atomic_t openers; /* number of active opens */
380
381 int changed; /* True if we might need to
382 * reread partition info */
383 int degraded; /* whether md should consider
384 * adding a spare
385 */
386
387 atomic_t recovery_active; /* blocks scheduled, but not written */
388 wait_queue_head_t recovery_wait;
389 sector_t recovery_cp;
390 sector_t resync_min; /* user requested sync
391 * starts here */
392 sector_t resync_max; /* resync should pause
393 * when it gets here */
394
395 struct kernfs_node *sysfs_state; /* handle for 'array_state'
396 * file in sysfs.
397 */
398 struct kernfs_node *sysfs_action; /* handle for 'sync_action' */
399
400 struct work_struct del_work; /* used for delayed sysfs removal */
401
402 /* "lock" protects:
403 * flush_bio transition from NULL to !NULL
404 * rdev superblocks, events
405 * clearing MD_CHANGE_*
406 * in_sync - and related safemode and MD_CHANGE changes
407 * pers (also protected by reconfig_mutex and pending IO).
408 * clearing ->bitmap
409 * clearing ->bitmap_info.file
410 * changing ->resync_{min,max}
411 * setting MD_RECOVERY_RUNNING (which interacts with resync_{min,max})
412 */
413 spinlock_t lock;
414 wait_queue_head_t sb_wait; /* for waiting on superblock updates */
415 atomic_t pending_writes; /* number of active superblock writes */
416
417 unsigned int safemode; /* if set, update "clean" superblock
418 * when no writes pending.
419 */
420 unsigned int safemode_delay;
421 struct timer_list safemode_timer;
422 atomic_t writes_pending;
423 struct request_queue *queue; /* for plugging ... */
424
425 struct bitmap *bitmap; /* the bitmap for the device */
426 struct {
427 struct file *file; /* the bitmap file */
428 loff_t offset; /* offset from superblock of
429 * start of bitmap. May be
430 * negative, but not '0'
431 * For external metadata, offset
432 * from start of device.
433 */
434 unsigned long space; /* space available at this offset */
435 loff_t default_offset; /* this is the offset to use when
436 * hot-adding a bitmap. It should
437 * eventually be settable by sysfs.
438 */
439 unsigned long default_space; /* space available at
440 * default offset */
441 struct mutex mutex;
442 unsigned long chunksize;
443 unsigned long daemon_sleep; /* how many jiffies between updates? */
444 unsigned long max_write_behind; /* write-behind mode */
445 int external;
446 int nodes; /* Maximum number of nodes in the cluster */
447 char cluster_name[64]; /* Name of the cluster */
448 } bitmap_info;
449
450 atomic_t max_corr_read_errors; /* max read retries */
451 struct list_head all_mddevs;
452
453 struct attribute_group *to_remove;
454
455 struct bio_set *bio_set;
456
457 /* Generic flush handling.
458 * The last to finish preflush schedules a worker to submit
459 * the rest of the request (without the REQ_FLUSH flag).
460 */
461 struct bio *flush_bio;
462 atomic_t flush_pending;
463 struct work_struct flush_work;
464 struct work_struct event_work; /* used by dm to report failure event */
465 void (*sync_super)(struct mddev *mddev, struct md_rdev *rdev);
466 struct md_cluster_info *cluster_info;
467 };
468
469 static inline int __must_check mddev_lock(struct mddev *mddev)
470 {
471 return mutex_lock_interruptible(&mddev->reconfig_mutex);
472 }
473
474 /* Sometimes we need to take the lock in a situation where
475 * failure due to interrupts is not acceptable.
476 */
477 static inline void mddev_lock_nointr(struct mddev *mddev)
478 {
479 mutex_lock(&mddev->reconfig_mutex);
480 }
481
482 static inline int mddev_is_locked(struct mddev *mddev)
483 {
484 return mutex_is_locked(&mddev->reconfig_mutex);
485 }
486
487 static inline int mddev_trylock(struct mddev *mddev)
488 {
489 return mutex_trylock(&mddev->reconfig_mutex);
490 }
491 extern void mddev_unlock(struct mddev *mddev);
492
493 static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors)
494 {
495 atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io);
496 }
497
498 struct md_personality
499 {
500 char *name;
501 int level;
502 struct list_head list;
503 struct module *owner;
504 void (*make_request)(struct mddev *mddev, struct bio *bio);
505 int (*run)(struct mddev *mddev);
506 void (*free)(struct mddev *mddev, void *priv);
507 void (*status)(struct seq_file *seq, struct mddev *mddev);
508 /* error_handler must set ->faulty and clear ->in_sync
509 * if appropriate, and should abort recovery if needed
510 */
511 void (*error_handler)(struct mddev *mddev, struct md_rdev *rdev);
512 int (*hot_add_disk) (struct mddev *mddev, struct md_rdev *rdev);
513 int (*hot_remove_disk) (struct mddev *mddev, struct md_rdev *rdev);
514 int (*spare_active) (struct mddev *mddev);
515 sector_t (*sync_request)(struct mddev *mddev, sector_t sector_nr, int *skipped);
516 int (*resize) (struct mddev *mddev, sector_t sectors);
517 sector_t (*size) (struct mddev *mddev, sector_t sectors, int raid_disks);
518 int (*check_reshape) (struct mddev *mddev);
519 int (*start_reshape) (struct mddev *mddev);
520 void (*finish_reshape) (struct mddev *mddev);
521 /* quiesce moves between quiescence states
522 * 0 - fully active
523 * 1 - no new requests allowed
524 * others - reserved
525 */
526 void (*quiesce) (struct mddev *mddev, int state);
527 /* takeover is used to transition an array from one
528 * personality to another. The new personality must be able
529 * to handle the data in the current layout.
530 * e.g. 2drive raid1 -> 2drive raid5
531 * ndrive raid5 -> degraded n+1drive raid6 with special layout
532 * If the takeover succeeds, a new 'private' structure is returned.
533 * This needs to be installed and then ->run used to activate the
534 * array.
535 */
536 void *(*takeover) (struct mddev *mddev);
537 /* congested implements bdi.congested_fn().
538 * Will not be called while array is 'suspended' */
539 int (*congested)(struct mddev *mddev, int bits);
540 };
541
542 struct md_sysfs_entry {
543 struct attribute attr;
544 ssize_t (*show)(struct mddev *, char *);
545 ssize_t (*store)(struct mddev *, const char *, size_t);
546 };
547 extern struct attribute_group md_bitmap_group;
548
549 static inline struct kernfs_node *sysfs_get_dirent_safe(struct kernfs_node *sd, char *name)
550 {
551 if (sd)
552 return sysfs_get_dirent(sd, name);
553 return sd;
554 }
555 static inline void sysfs_notify_dirent_safe(struct kernfs_node *sd)
556 {
557 if (sd)
558 sysfs_notify_dirent(sd);
559 }
560
561 static inline char * mdname (struct mddev * mddev)
562 {
563 return mddev->gendisk ? mddev->gendisk->disk_name : "mdX";
564 }
565
566 static inline int sysfs_link_rdev(struct mddev *mddev, struct md_rdev *rdev)
567 {
568 char nm[20];
569 if (!test_bit(Replacement, &rdev->flags) &&
570 !test_bit(Journal, &rdev->flags) &&
571 mddev->kobj.sd) {
572 sprintf(nm, "rd%d", rdev->raid_disk);
573 return sysfs_create_link(&mddev->kobj, &rdev->kobj, nm);
574 } else
575 return 0;
576 }
577
578 static inline void sysfs_unlink_rdev(struct mddev *mddev, struct md_rdev *rdev)
579 {
580 char nm[20];
581 if (!test_bit(Replacement, &rdev->flags) &&
582 !test_bit(Journal, &rdev->flags) &&
583 mddev->kobj.sd) {
584 sprintf(nm, "rd%d", rdev->raid_disk);
585 sysfs_remove_link(&mddev->kobj, nm);
586 }
587 }
588
589 /*
590 * iterates through some rdev ringlist. It's safe to remove the
591 * current 'rdev'. Dont touch 'tmp' though.
592 */
593 #define rdev_for_each_list(rdev, tmp, head) \
594 list_for_each_entry_safe(rdev, tmp, head, same_set)
595
596 /*
597 * iterates through the 'same array disks' ringlist
598 */
599 #define rdev_for_each(rdev, mddev) \
600 list_for_each_entry(rdev, &((mddev)->disks), same_set)
601
602 #define rdev_for_each_safe(rdev, tmp, mddev) \
603 list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set)
604
605 #define rdev_for_each_rcu(rdev, mddev) \
606 list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set)
607
608 struct md_thread {
609 void (*run) (struct md_thread *thread);
610 struct mddev *mddev;
611 wait_queue_head_t wqueue;
612 unsigned long flags;
613 struct task_struct *tsk;
614 unsigned long timeout;
615 void *private;
616 };
617
618 #define THREAD_WAKEUP 0
619
620 static inline void safe_put_page(struct page *p)
621 {
622 if (p) put_page(p);
623 }
624
625 extern int register_md_personality(struct md_personality *p);
626 extern int unregister_md_personality(struct md_personality *p);
627 extern int register_md_cluster_operations(struct md_cluster_operations *ops,
628 struct module *module);
629 extern int unregister_md_cluster_operations(void);
630 extern int md_setup_cluster(struct mddev *mddev, int nodes);
631 extern void md_cluster_stop(struct mddev *mddev);
632 extern struct md_thread *md_register_thread(
633 void (*run)(struct md_thread *thread),
634 struct mddev *mddev,
635 const char *name);
636 extern void md_unregister_thread(struct md_thread **threadp);
637 extern void md_wakeup_thread(struct md_thread *thread);
638 extern void md_check_recovery(struct mddev *mddev);
639 extern void md_reap_sync_thread(struct mddev *mddev);
640 extern void md_write_start(struct mddev *mddev, struct bio *bi);
641 extern void md_write_end(struct mddev *mddev);
642 extern void md_done_sync(struct mddev *mddev, int blocks, int ok);
643 extern void md_error(struct mddev *mddev, struct md_rdev *rdev);
644 extern void md_finish_reshape(struct mddev *mddev);
645
646 extern int mddev_congested(struct mddev *mddev, int bits);
647 extern void md_flush_request(struct mddev *mddev, struct bio *bio);
648 extern void md_super_write(struct mddev *mddev, struct md_rdev *rdev,
649 sector_t sector, int size, struct page *page);
650 extern void md_super_wait(struct mddev *mddev);
651 extern int sync_page_io(struct md_rdev *rdev, sector_t sector, int size,
652 struct page *page, int rw, bool metadata_op);
653 extern void md_do_sync(struct md_thread *thread);
654 extern void md_new_event(struct mddev *mddev);
655 extern int md_allow_write(struct mddev *mddev);
656 extern void md_wait_for_blocked_rdev(struct md_rdev *rdev, struct mddev *mddev);
657 extern void md_set_array_sectors(struct mddev *mddev, sector_t array_sectors);
658 extern int md_check_no_bitmap(struct mddev *mddev);
659 extern int md_integrity_register(struct mddev *mddev);
660 extern void md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev);
661 extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale);
662
663 extern void mddev_init(struct mddev *mddev);
664 extern int md_run(struct mddev *mddev);
665 extern void md_stop(struct mddev *mddev);
666 extern void md_stop_writes(struct mddev *mddev);
667 extern int md_rdev_init(struct md_rdev *rdev);
668 extern void md_rdev_clear(struct md_rdev *rdev);
669
670 extern void mddev_suspend(struct mddev *mddev);
671 extern void mddev_resume(struct mddev *mddev);
672 extern struct bio *bio_clone_mddev(struct bio *bio, gfp_t gfp_mask,
673 struct mddev *mddev);
674 extern struct bio *bio_alloc_mddev(gfp_t gfp_mask, int nr_iovecs,
675 struct mddev *mddev);
676
677 extern void md_unplug(struct blk_plug_cb *cb, bool from_schedule);
678 extern void md_reload_sb(struct mddev *mddev, int raid_disk);
679 extern void md_update_sb(struct mddev *mddev, int force);
680 extern void md_kick_rdev_from_array(struct md_rdev * rdev);
681 struct md_rdev *md_find_rdev_nr_rcu(struct mddev *mddev, int nr);
682 static inline int mddev_check_plugged(struct mddev *mddev)
683 {
684 return !!blk_check_plugged(md_unplug, mddev,
685 sizeof(struct blk_plug_cb));
686 }
687
688 static inline void rdev_dec_pending(struct md_rdev *rdev, struct mddev *mddev)
689 {
690 int faulty = test_bit(Faulty, &rdev->flags);
691 if (atomic_dec_and_test(&rdev->nr_pending) && faulty) {
692 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
693 md_wakeup_thread(mddev->thread);
694 }
695 }
696
697 extern struct md_cluster_operations *md_cluster_ops;
698 static inline int mddev_is_clustered(struct mddev *mddev)
699 {
700 return mddev->cluster_info && mddev->bitmap_info.nodes > 1;
701 }
702 #endif /* _MD_MD_H */