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