]>
Commit | Line | Data |
---|---|---|
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 | ExternalBbl, /* External metadata provides bad | |
172 | * block management for a disk | |
173 | */ | |
174 | }; | |
175 | ||
176 | static inline int is_badblock(struct md_rdev *rdev, sector_t s, int sectors, | |
177 | sector_t *first_bad, int *bad_sectors) | |
178 | { | |
179 | if (unlikely(rdev->badblocks.count)) { | |
180 | int rv = badblocks_check(&rdev->badblocks, rdev->data_offset + s, | |
181 | sectors, | |
182 | first_bad, bad_sectors); | |
183 | if (rv) | |
184 | *first_bad -= rdev->data_offset; | |
185 | return rv; | |
186 | } | |
187 | return 0; | |
188 | } | |
189 | extern int rdev_set_badblocks(struct md_rdev *rdev, sector_t s, int sectors, | |
190 | int is_new); | |
191 | extern int rdev_clear_badblocks(struct md_rdev *rdev, sector_t s, int sectors, | |
192 | int is_new); | |
193 | struct md_cluster_info; | |
194 | ||
195 | struct mddev { | |
196 | void *private; | |
197 | struct md_personality *pers; | |
198 | dev_t unit; | |
199 | int md_minor; | |
200 | struct list_head disks; | |
201 | unsigned long flags; | |
202 | #define MD_CHANGE_DEVS 0 /* Some device status has changed */ | |
203 | #define MD_CHANGE_CLEAN 1 /* transition to or from 'clean' */ | |
204 | #define MD_CHANGE_PENDING 2 /* switch from 'clean' to 'active' in progress */ | |
205 | #define MD_UPDATE_SB_FLAGS (1 | 2 | 4) /* If these are set, md_update_sb needed */ | |
206 | #define MD_ARRAY_FIRST_USE 3 /* First use of array, needs initialization */ | |
207 | #define MD_CLOSING 4 /* If set, we are closing the array, do not open | |
208 | * it then */ | |
209 | #define MD_JOURNAL_CLEAN 5 /* A raid with journal is already clean */ | |
210 | #define MD_HAS_JOURNAL 6 /* The raid array has journal feature set */ | |
211 | #define MD_RELOAD_SB 7 /* Reload the superblock because another node | |
212 | * updated it. | |
213 | */ | |
214 | #define MD_CLUSTER_RESYNC_LOCKED 8 /* cluster raid only, which means node | |
215 | * already took resync lock, need to | |
216 | * release the lock */ | |
217 | ||
218 | int suspended; | |
219 | atomic_t active_io; | |
220 | int ro; | |
221 | int sysfs_active; /* set when sysfs deletes | |
222 | * are happening, so run/ | |
223 | * takeover/stop are not safe | |
224 | */ | |
225 | struct gendisk *gendisk; | |
226 | ||
227 | struct kobject kobj; | |
228 | int hold_active; | |
229 | #define UNTIL_IOCTL 1 | |
230 | #define UNTIL_STOP 2 | |
231 | ||
232 | /* Superblock information */ | |
233 | int major_version, | |
234 | minor_version, | |
235 | patch_version; | |
236 | int persistent; | |
237 | int external; /* metadata is | |
238 | * managed externally */ | |
239 | char metadata_type[17]; /* externally set*/ | |
240 | int chunk_sectors; | |
241 | time64_t ctime, utime; | |
242 | int level, layout; | |
243 | char clevel[16]; | |
244 | int raid_disks; | |
245 | int max_disks; | |
246 | sector_t dev_sectors; /* used size of | |
247 | * component devices */ | |
248 | sector_t array_sectors; /* exported array size */ | |
249 | int external_size; /* size managed | |
250 | * externally */ | |
251 | __u64 events; | |
252 | /* If the last 'event' was simply a clean->dirty transition, and | |
253 | * we didn't write it to the spares, then it is safe and simple | |
254 | * to just decrement the event count on a dirty->clean transition. | |
255 | * So we record that possibility here. | |
256 | */ | |
257 | int can_decrease_events; | |
258 | ||
259 | char uuid[16]; | |
260 | ||
261 | /* If the array is being reshaped, we need to record the | |
262 | * new shape and an indication of where we are up to. | |
263 | * This is written to the superblock. | |
264 | * If reshape_position is MaxSector, then no reshape is happening (yet). | |
265 | */ | |
266 | sector_t reshape_position; | |
267 | int delta_disks, new_level, new_layout; | |
268 | int new_chunk_sectors; | |
269 | int reshape_backwards; | |
270 | ||
271 | struct md_thread *thread; /* management thread */ | |
272 | struct md_thread *sync_thread; /* doing resync or reconstruct */ | |
273 | ||
274 | /* 'last_sync_action' is initialized to "none". It is set when a | |
275 | * sync operation (i.e "data-check", "requested-resync", "resync", | |
276 | * "recovery", or "reshape") is started. It holds this value even | |
277 | * when the sync thread is "frozen" (interrupted) or "idle" (stopped | |
278 | * or finished). It is overwritten when a new sync operation is begun. | |
279 | */ | |
280 | char *last_sync_action; | |
281 | sector_t curr_resync; /* last block scheduled */ | |
282 | /* As resync requests can complete out of order, we cannot easily track | |
283 | * how much resync has been completed. So we occasionally pause until | |
284 | * everything completes, then set curr_resync_completed to curr_resync. | |
285 | * As such it may be well behind the real resync mark, but it is a value | |
286 | * we are certain of. | |
287 | */ | |
288 | sector_t curr_resync_completed; | |
289 | unsigned long resync_mark; /* a recent timestamp */ | |
290 | sector_t resync_mark_cnt;/* blocks written at resync_mark */ | |
291 | sector_t curr_mark_cnt; /* blocks scheduled now */ | |
292 | ||
293 | sector_t resync_max_sectors; /* may be set by personality */ | |
294 | ||
295 | atomic64_t resync_mismatches; /* count of sectors where | |
296 | * parity/replica mismatch found | |
297 | */ | |
298 | ||
299 | /* allow user-space to request suspension of IO to regions of the array */ | |
300 | sector_t suspend_lo; | |
301 | sector_t suspend_hi; | |
302 | /* if zero, use the system-wide default */ | |
303 | int sync_speed_min; | |
304 | int sync_speed_max; | |
305 | ||
306 | /* resync even though the same disks are shared among md-devices */ | |
307 | int parallel_resync; | |
308 | ||
309 | int ok_start_degraded; | |
310 | /* recovery/resync flags | |
311 | * NEEDED: we might need to start a resync/recover | |
312 | * RUNNING: a thread is running, or about to be started | |
313 | * SYNC: actually doing a resync, not a recovery | |
314 | * RECOVER: doing recovery, or need to try it. | |
315 | * INTR: resync needs to be aborted for some reason | |
316 | * DONE: thread is done and is waiting to be reaped | |
317 | * REQUEST: user-space has requested a sync (used with SYNC) | |
318 | * CHECK: user-space request for check-only, no repair | |
319 | * RESHAPE: A reshape is happening | |
320 | * ERROR: sync-action interrupted because io-error | |
321 | * | |
322 | * If neither SYNC or RESHAPE are set, then it is a recovery. | |
323 | */ | |
324 | #define MD_RECOVERY_RUNNING 0 | |
325 | #define MD_RECOVERY_SYNC 1 | |
326 | #define MD_RECOVERY_RECOVER 2 | |
327 | #define MD_RECOVERY_INTR 3 | |
328 | #define MD_RECOVERY_DONE 4 | |
329 | #define MD_RECOVERY_NEEDED 5 | |
330 | #define MD_RECOVERY_REQUESTED 6 | |
331 | #define MD_RECOVERY_CHECK 7 | |
332 | #define MD_RECOVERY_RESHAPE 8 | |
333 | #define MD_RECOVERY_FROZEN 9 | |
334 | #define MD_RECOVERY_ERROR 10 | |
335 | ||
336 | unsigned long recovery; | |
337 | /* If a RAID personality determines that recovery (of a particular | |
338 | * device) will fail due to a read error on the source device, it | |
339 | * takes a copy of this number and does not attempt recovery again | |
340 | * until this number changes. | |
341 | */ | |
342 | int recovery_disabled; | |
343 | ||
344 | int in_sync; /* know to not need resync */ | |
345 | /* 'open_mutex' avoids races between 'md_open' and 'do_md_stop', so | |
346 | * that we are never stopping an array while it is open. | |
347 | * 'reconfig_mutex' protects all other reconfiguration. | |
348 | * These locks are separate due to conflicting interactions | |
349 | * with bdev->bd_mutex. | |
350 | * Lock ordering is: | |
351 | * reconfig_mutex -> bd_mutex : e.g. do_md_run -> revalidate_disk | |
352 | * bd_mutex -> open_mutex: e.g. __blkdev_get -> md_open | |
353 | */ | |
354 | struct mutex open_mutex; | |
355 | struct mutex reconfig_mutex; | |
356 | atomic_t active; /* general refcount */ | |
357 | atomic_t openers; /* number of active opens */ | |
358 | ||
359 | int changed; /* True if we might need to | |
360 | * reread partition info */ | |
361 | int degraded; /* whether md should consider | |
362 | * adding a spare | |
363 | */ | |
364 | ||
365 | atomic_t recovery_active; /* blocks scheduled, but not written */ | |
366 | wait_queue_head_t recovery_wait; | |
367 | sector_t recovery_cp; | |
368 | sector_t resync_min; /* user requested sync | |
369 | * starts here */ | |
370 | sector_t resync_max; /* resync should pause | |
371 | * when it gets here */ | |
372 | ||
373 | struct kernfs_node *sysfs_state; /* handle for 'array_state' | |
374 | * file in sysfs. | |
375 | */ | |
376 | struct kernfs_node *sysfs_action; /* handle for 'sync_action' */ | |
377 | ||
378 | struct work_struct del_work; /* used for delayed sysfs removal */ | |
379 | ||
380 | /* "lock" protects: | |
381 | * flush_bio transition from NULL to !NULL | |
382 | * rdev superblocks, events | |
383 | * clearing MD_CHANGE_* | |
384 | * in_sync - and related safemode and MD_CHANGE changes | |
385 | * pers (also protected by reconfig_mutex and pending IO). | |
386 | * clearing ->bitmap | |
387 | * clearing ->bitmap_info.file | |
388 | * changing ->resync_{min,max} | |
389 | * setting MD_RECOVERY_RUNNING (which interacts with resync_{min,max}) | |
390 | */ | |
391 | spinlock_t lock; | |
392 | wait_queue_head_t sb_wait; /* for waiting on superblock updates */ | |
393 | atomic_t pending_writes; /* number of active superblock writes */ | |
394 | ||
395 | unsigned int safemode; /* if set, update "clean" superblock | |
396 | * when no writes pending. | |
397 | */ | |
398 | unsigned int safemode_delay; | |
399 | struct timer_list safemode_timer; | |
400 | atomic_t writes_pending; | |
401 | struct request_queue *queue; /* for plugging ... */ | |
402 | ||
403 | struct bitmap *bitmap; /* the bitmap for the device */ | |
404 | struct { | |
405 | struct file *file; /* the bitmap file */ | |
406 | loff_t offset; /* offset from superblock of | |
407 | * start of bitmap. May be | |
408 | * negative, but not '0' | |
409 | * For external metadata, offset | |
410 | * from start of device. | |
411 | */ | |
412 | unsigned long space; /* space available at this offset */ | |
413 | loff_t default_offset; /* this is the offset to use when | |
414 | * hot-adding a bitmap. It should | |
415 | * eventually be settable by sysfs. | |
416 | */ | |
417 | unsigned long default_space; /* space available at | |
418 | * default offset */ | |
419 | struct mutex mutex; | |
420 | unsigned long chunksize; | |
421 | unsigned long daemon_sleep; /* how many jiffies between updates? */ | |
422 | unsigned long max_write_behind; /* write-behind mode */ | |
423 | int external; | |
424 | int nodes; /* Maximum number of nodes in the cluster */ | |
425 | char cluster_name[64]; /* Name of the cluster */ | |
426 | } bitmap_info; | |
427 | ||
428 | atomic_t max_corr_read_errors; /* max read retries */ | |
429 | struct list_head all_mddevs; | |
430 | ||
431 | struct attribute_group *to_remove; | |
432 | ||
433 | struct bio_set *bio_set; | |
434 | ||
435 | /* Generic flush handling. | |
436 | * The last to finish preflush schedules a worker to submit | |
437 | * the rest of the request (without the REQ_PREFLUSH flag). | |
438 | */ | |
439 | struct bio *flush_bio; | |
440 | atomic_t flush_pending; | |
441 | struct work_struct flush_work; | |
442 | struct work_struct event_work; /* used by dm to report failure event */ | |
443 | void (*sync_super)(struct mddev *mddev, struct md_rdev *rdev); | |
444 | struct md_cluster_info *cluster_info; | |
445 | unsigned int good_device_nr; /* good device num within cluster raid */ | |
446 | }; | |
447 | ||
448 | static inline int __must_check mddev_lock(struct mddev *mddev) | |
449 | { | |
450 | return mutex_lock_interruptible(&mddev->reconfig_mutex); | |
451 | } | |
452 | ||
453 | /* Sometimes we need to take the lock in a situation where | |
454 | * failure due to interrupts is not acceptable. | |
455 | */ | |
456 | static inline void mddev_lock_nointr(struct mddev *mddev) | |
457 | { | |
458 | mutex_lock(&mddev->reconfig_mutex); | |
459 | } | |
460 | ||
461 | static inline int mddev_is_locked(struct mddev *mddev) | |
462 | { | |
463 | return mutex_is_locked(&mddev->reconfig_mutex); | |
464 | } | |
465 | ||
466 | static inline int mddev_trylock(struct mddev *mddev) | |
467 | { | |
468 | return mutex_trylock(&mddev->reconfig_mutex); | |
469 | } | |
470 | extern void mddev_unlock(struct mddev *mddev); | |
471 | ||
472 | static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors) | |
473 | { | |
474 | atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io); | |
475 | } | |
476 | ||
477 | struct md_personality | |
478 | { | |
479 | char *name; | |
480 | int level; | |
481 | struct list_head list; | |
482 | struct module *owner; | |
483 | void (*make_request)(struct mddev *mddev, struct bio *bio); | |
484 | int (*run)(struct mddev *mddev); | |
485 | void (*free)(struct mddev *mddev, void *priv); | |
486 | void (*status)(struct seq_file *seq, struct mddev *mddev); | |
487 | /* error_handler must set ->faulty and clear ->in_sync | |
488 | * if appropriate, and should abort recovery if needed | |
489 | */ | |
490 | void (*error_handler)(struct mddev *mddev, struct md_rdev *rdev); | |
491 | int (*hot_add_disk) (struct mddev *mddev, struct md_rdev *rdev); | |
492 | int (*hot_remove_disk) (struct mddev *mddev, struct md_rdev *rdev); | |
493 | int (*spare_active) (struct mddev *mddev); | |
494 | sector_t (*sync_request)(struct mddev *mddev, sector_t sector_nr, int *skipped); | |
495 | int (*resize) (struct mddev *mddev, sector_t sectors); | |
496 | sector_t (*size) (struct mddev *mddev, sector_t sectors, int raid_disks); | |
497 | int (*check_reshape) (struct mddev *mddev); | |
498 | int (*start_reshape) (struct mddev *mddev); | |
499 | void (*finish_reshape) (struct mddev *mddev); | |
500 | /* quiesce moves between quiescence states | |
501 | * 0 - fully active | |
502 | * 1 - no new requests allowed | |
503 | * others - reserved | |
504 | */ | |
505 | void (*quiesce) (struct mddev *mddev, int state); | |
506 | /* takeover is used to transition an array from one | |
507 | * personality to another. The new personality must be able | |
508 | * to handle the data in the current layout. | |
509 | * e.g. 2drive raid1 -> 2drive raid5 | |
510 | * ndrive raid5 -> degraded n+1drive raid6 with special layout | |
511 | * If the takeover succeeds, a new 'private' structure is returned. | |
512 | * This needs to be installed and then ->run used to activate the | |
513 | * array. | |
514 | */ | |
515 | void *(*takeover) (struct mddev *mddev); | |
516 | /* congested implements bdi.congested_fn(). | |
517 | * Will not be called while array is 'suspended' */ | |
518 | int (*congested)(struct mddev *mddev, int bits); | |
519 | }; | |
520 | ||
521 | struct md_sysfs_entry { | |
522 | struct attribute attr; | |
523 | ssize_t (*show)(struct mddev *, char *); | |
524 | ssize_t (*store)(struct mddev *, const char *, size_t); | |
525 | }; | |
526 | extern struct attribute_group md_bitmap_group; | |
527 | ||
528 | static inline struct kernfs_node *sysfs_get_dirent_safe(struct kernfs_node *sd, char *name) | |
529 | { | |
530 | if (sd) | |
531 | return sysfs_get_dirent(sd, name); | |
532 | return sd; | |
533 | } | |
534 | static inline void sysfs_notify_dirent_safe(struct kernfs_node *sd) | |
535 | { | |
536 | if (sd) | |
537 | sysfs_notify_dirent(sd); | |
538 | } | |
539 | ||
540 | static inline char * mdname (struct mddev * mddev) | |
541 | { | |
542 | return mddev->gendisk ? mddev->gendisk->disk_name : "mdX"; | |
543 | } | |
544 | ||
545 | static inline int sysfs_link_rdev(struct mddev *mddev, struct md_rdev *rdev) | |
546 | { | |
547 | char nm[20]; | |
548 | if (!test_bit(Replacement, &rdev->flags) && | |
549 | !test_bit(Journal, &rdev->flags) && | |
550 | mddev->kobj.sd) { | |
551 | sprintf(nm, "rd%d", rdev->raid_disk); | |
552 | return sysfs_create_link(&mddev->kobj, &rdev->kobj, nm); | |
553 | } else | |
554 | return 0; | |
555 | } | |
556 | ||
557 | static inline void sysfs_unlink_rdev(struct mddev *mddev, struct md_rdev *rdev) | |
558 | { | |
559 | char nm[20]; | |
560 | if (!test_bit(Replacement, &rdev->flags) && | |
561 | !test_bit(Journal, &rdev->flags) && | |
562 | mddev->kobj.sd) { | |
563 | sprintf(nm, "rd%d", rdev->raid_disk); | |
564 | sysfs_remove_link(&mddev->kobj, nm); | |
565 | } | |
566 | } | |
567 | ||
568 | /* | |
569 | * iterates through some rdev ringlist. It's safe to remove the | |
570 | * current 'rdev'. Dont touch 'tmp' though. | |
571 | */ | |
572 | #define rdev_for_each_list(rdev, tmp, head) \ | |
573 | list_for_each_entry_safe(rdev, tmp, head, same_set) | |
574 | ||
575 | /* | |
576 | * iterates through the 'same array disks' ringlist | |
577 | */ | |
578 | #define rdev_for_each(rdev, mddev) \ | |
579 | list_for_each_entry(rdev, &((mddev)->disks), same_set) | |
580 | ||
581 | #define rdev_for_each_safe(rdev, tmp, mddev) \ | |
582 | list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set) | |
583 | ||
584 | #define rdev_for_each_rcu(rdev, mddev) \ | |
585 | list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set) | |
586 | ||
587 | struct md_thread { | |
588 | void (*run) (struct md_thread *thread); | |
589 | struct mddev *mddev; | |
590 | wait_queue_head_t wqueue; | |
591 | unsigned long flags; | |
592 | struct task_struct *tsk; | |
593 | unsigned long timeout; | |
594 | void *private; | |
595 | }; | |
596 | ||
597 | #define THREAD_WAKEUP 0 | |
598 | ||
599 | static inline void safe_put_page(struct page *p) | |
600 | { | |
601 | if (p) put_page(p); | |
602 | } | |
603 | ||
604 | extern int register_md_personality(struct md_personality *p); | |
605 | extern int unregister_md_personality(struct md_personality *p); | |
606 | extern int register_md_cluster_operations(struct md_cluster_operations *ops, | |
607 | struct module *module); | |
608 | extern int unregister_md_cluster_operations(void); | |
609 | extern int md_setup_cluster(struct mddev *mddev, int nodes); | |
610 | extern void md_cluster_stop(struct mddev *mddev); | |
611 | extern struct md_thread *md_register_thread( | |
612 | void (*run)(struct md_thread *thread), | |
613 | struct mddev *mddev, | |
614 | const char *name); | |
615 | extern void md_unregister_thread(struct md_thread **threadp); | |
616 | extern void md_wakeup_thread(struct md_thread *thread); | |
617 | extern void md_check_recovery(struct mddev *mddev); | |
618 | extern void md_reap_sync_thread(struct mddev *mddev); | |
619 | extern void md_write_start(struct mddev *mddev, struct bio *bi); | |
620 | extern void md_write_end(struct mddev *mddev); | |
621 | extern void md_done_sync(struct mddev *mddev, int blocks, int ok); | |
622 | extern void md_error(struct mddev *mddev, struct md_rdev *rdev); | |
623 | extern void md_finish_reshape(struct mddev *mddev); | |
624 | ||
625 | extern int mddev_congested(struct mddev *mddev, int bits); | |
626 | extern void md_flush_request(struct mddev *mddev, struct bio *bio); | |
627 | extern void md_super_write(struct mddev *mddev, struct md_rdev *rdev, | |
628 | sector_t sector, int size, struct page *page); | |
629 | extern void md_super_wait(struct mddev *mddev); | |
630 | extern int sync_page_io(struct md_rdev *rdev, sector_t sector, int size, | |
631 | struct page *page, int op, int op_flags, | |
632 | bool metadata_op); | |
633 | extern void md_do_sync(struct md_thread *thread); | |
634 | extern void md_new_event(struct mddev *mddev); | |
635 | extern int md_allow_write(struct mddev *mddev); | |
636 | extern void md_wait_for_blocked_rdev(struct md_rdev *rdev, struct mddev *mddev); | |
637 | extern void md_set_array_sectors(struct mddev *mddev, sector_t array_sectors); | |
638 | extern int md_check_no_bitmap(struct mddev *mddev); | |
639 | extern int md_integrity_register(struct mddev *mddev); | |
640 | extern int md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev); | |
641 | extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale); | |
642 | ||
643 | extern void mddev_init(struct mddev *mddev); | |
644 | extern int md_run(struct mddev *mddev); | |
645 | extern void md_stop(struct mddev *mddev); | |
646 | extern void md_stop_writes(struct mddev *mddev); | |
647 | extern int md_rdev_init(struct md_rdev *rdev); | |
648 | extern void md_rdev_clear(struct md_rdev *rdev); | |
649 | ||
650 | extern void mddev_suspend(struct mddev *mddev); | |
651 | extern void mddev_resume(struct mddev *mddev); | |
652 | extern struct bio *bio_clone_mddev(struct bio *bio, gfp_t gfp_mask, | |
653 | struct mddev *mddev); | |
654 | extern struct bio *bio_alloc_mddev(gfp_t gfp_mask, int nr_iovecs, | |
655 | struct mddev *mddev); | |
656 | ||
657 | extern void md_unplug(struct blk_plug_cb *cb, bool from_schedule); | |
658 | extern void md_reload_sb(struct mddev *mddev, int raid_disk); | |
659 | extern void md_update_sb(struct mddev *mddev, int force); | |
660 | extern void md_kick_rdev_from_array(struct md_rdev * rdev); | |
661 | struct md_rdev *md_find_rdev_nr_rcu(struct mddev *mddev, int nr); | |
662 | static inline int mddev_check_plugged(struct mddev *mddev) | |
663 | { | |
664 | return !!blk_check_plugged(md_unplug, mddev, | |
665 | sizeof(struct blk_plug_cb)); | |
666 | } | |
667 | ||
668 | static inline void rdev_dec_pending(struct md_rdev *rdev, struct mddev *mddev) | |
669 | { | |
670 | int faulty = test_bit(Faulty, &rdev->flags); | |
671 | if (atomic_dec_and_test(&rdev->nr_pending) && faulty) { | |
672 | set_bit(MD_RECOVERY_NEEDED, &mddev->recovery); | |
673 | md_wakeup_thread(mddev->thread); | |
674 | } | |
675 | } | |
676 | ||
677 | extern struct md_cluster_operations *md_cluster_ops; | |
678 | static inline int mddev_is_clustered(struct mddev *mddev) | |
679 | { | |
680 | return mddev->cluster_info && mddev->bitmap_info.nodes > 1; | |
681 | } | |
682 | #endif /* _MD_MD_H */ |