Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
7e841526 | 2 | md.h : kernel internal structure of the Linux MD driver |
1da177e4 | 3 | Copyright (C) 1996-98 Ingo Molnar, Gadi Oxman |
f72ffdd6 | 4 | |
1da177e4 LT |
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. | |
f72ffdd6 | 9 | |
1da177e4 LT |
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 | |
f72ffdd6 | 12 | Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. |
1da177e4 LT |
13 | */ |
14 | ||
63fe0817 CH |
15 | #ifndef _MD_MD_H |
16 | #define _MD_MD_H | |
17 | ||
18 | #include <linux/blkdev.h> | |
66114cad | 19 | #include <linux/backing-dev.h> |
fc974ee2 | 20 | #include <linux/badblocks.h> |
63fe0817 CH |
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> | |
edb39c9d | 28 | #include "md-cluster.h" |
9361401e | 29 | |
1da177e4 | 30 | #define MaxSector (~(sector_t)0) |
1da177e4 | 31 | |
1da177e4 LT |
32 | /* |
33 | * MD's 'extended' device | |
34 | */ | |
3cb03002 | 35 | struct md_rdev { |
1da177e4 LT |
36 | struct list_head same_set; /* RAID devices within the same set */ |
37 | ||
dd8ac336 | 38 | sector_t sectors; /* Device size (in 512bytes sectors) */ |
fd01b88c | 39 | struct mddev *mddev; /* RAID array if running */ |
eea1bf38 | 40 | int last_events; /* IO event timestamp */ |
1da177e4 | 41 | |
a6ff7e08 JB |
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; | |
1da177e4 LT |
48 | struct block_device *bdev; /* block device handle */ |
49 | ||
2699b672 | 50 | struct page *sb_page, *bb_page; |
1da177e4 | 51 | int sb_loaded; |
42543769 | 52 | __u64 sb_events; |
1da177e4 | 53 | sector_t data_offset; /* start of data in array */ |
c6563a8c | 54 | sector_t new_data_offset;/* only relevant while reshaping */ |
f72ffdd6 | 55 | sector_t sb_start; /* offset of the super block (in 512byte sectors) */ |
0002b271 | 56 | int sb_size; /* bytes in the superblock */ |
1da177e4 LT |
57 | int preferred_minor; /* autorun support */ |
58 | ||
86e6ffdd N |
59 | struct kobject kobj; |
60 | ||
1da177e4 LT |
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 | */ | |
1da177e4 | 71 | |
2d78f8c4 | 72 | unsigned long flags; /* bit set of 'enum flag_bits' bits. */ |
6bfe0b49 | 73 | wait_queue_head_t blocked_wait; |
8ddf9efe | 74 | |
1da177e4 LT |
75 | int desc_nr; /* descriptor index in the superblock */ |
76 | int raid_disk; /* role of device in array */ | |
e93f68a1 N |
77 | int new_raid_disk; /* role that the device will have in |
78 | * the array after a level-change completes. | |
79 | */ | |
41158c7e N |
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 | */ | |
3069aa8d SL |
84 | union { |
85 | sector_t recovery_offset;/* If this device has been partially | |
5fd6c1dc N |
86 | * recovered, this is where we were |
87 | * up to. | |
88 | */ | |
3069aa8d SL |
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 | }; | |
1da177e4 LT |
94 | |
95 | atomic_t nr_pending; /* number of pending requests. | |
96 | * only maintained for arrays that | |
97 | * support hot removal | |
98 | */ | |
ba22dcbf N |
99 | atomic_t read_errors; /* number of consecutive read errors that |
100 | * we have tried to ignore. | |
101 | */ | |
0e3ef49e | 102 | time64_t last_read_error; /* monotonic time since our |
1e50915f RB |
103 | * last read error |
104 | */ | |
4dbcdc75 N |
105 | atomic_t corrected_errors; /* number of corrected read errors, |
106 | * for reporting to userspace and storing | |
107 | * in superblock. | |
108 | */ | |
5792a285 | 109 | struct work_struct del_work; /* used for delayed sysfs removal */ |
3c0ee63a | 110 | |
324a56e1 | 111 | struct kernfs_node *sysfs_state; /* handle for 'state' |
3c0ee63a | 112 | * sysfs entry */ |
2230dfe4 | 113 | |
fc974ee2 | 114 | struct badblocks badblocks; |
1da177e4 | 115 | }; |
2d78f8c4 N |
116 | enum flag_bits { |
117 | Faulty, /* device is known to have a fault */ | |
118 | In_sync, /* device is in_sync with rest of array */ | |
8313b8e5 N |
119 | Bitmap_sync, /* ..actually, not quite In_sync. Need a |
120 | * bitmap-based recovery to get fully in sync | |
121 | */ | |
2d78f8c4 N |
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 | */ | |
1aee41f6 GR |
156 | Candidate, /* For clustered environments only: |
157 | * This device is seen locally but not | |
158 | * by the whole cluster | |
159 | */ | |
bac624f3 SL |
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 | */ | |
659b254f | 165 | ClusterRemove, |
d787be40 N |
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 | */ | |
35b785f7 TM |
171 | ExternalBbl, /* External metadata provides bad |
172 | * block management for a disk | |
173 | */ | |
2d78f8c4 | 174 | }; |
1da177e4 | 175 | |
3cb03002 | 176 | static inline int is_badblock(struct md_rdev *rdev, sector_t s, int sectors, |
2230dfe4 N |
177 | sector_t *first_bad, int *bad_sectors) |
178 | { | |
179 | if (unlikely(rdev->badblocks.count)) { | |
fc974ee2 | 180 | int rv = badblocks_check(&rdev->badblocks, rdev->data_offset + s, |
2230dfe4 N |
181 | sectors, |
182 | first_bad, bad_sectors); | |
183 | if (rv) | |
184 | *first_bad -= rdev->data_offset; | |
185 | return rv; | |
186 | } | |
187 | return 0; | |
188 | } | |
3cb03002 | 189 | extern int rdev_set_badblocks(struct md_rdev *rdev, sector_t s, int sectors, |
c6563a8c N |
190 | int is_new); |
191 | extern int rdev_clear_badblocks(struct md_rdev *rdev, sector_t s, int sectors, | |
192 | int is_new); | |
c4ce867f GR |
193 | struct md_cluster_info; |
194 | ||
be306c29 N |
195 | enum mddev_flags { |
196 | MD_CHANGE_DEVS, /* Some device status has changed */ | |
197 | MD_CHANGE_CLEAN, /* transition to or from 'clean' */ | |
198 | MD_CHANGE_PENDING, /* switch from 'clean' to 'active' in progress */ | |
199 | MD_ARRAY_FIRST_USE, /* First use of array, needs initialization */ | |
200 | MD_CLOSING, /* If set, we are closing the array, do not open | |
201 | * it then */ | |
202 | MD_JOURNAL_CLEAN, /* A raid with journal is already clean */ | |
203 | MD_HAS_JOURNAL, /* The raid array has journal feature set */ | |
204 | MD_RELOAD_SB, /* Reload the superblock because another node | |
205 | * updated it. | |
206 | */ | |
207 | MD_CLUSTER_RESYNC_LOCKED, /* cluster raid only, which means node | |
208 | * already took resync lock, need to | |
209 | * release the lock */ | |
210 | }; | |
211 | #define MD_UPDATE_SB_FLAGS (BIT(MD_CHANGE_DEVS) | \ | |
212 | BIT(MD_CHANGE_CLEAN) | \ | |
213 | BIT(MD_CHANGE_PENDING)) /* If these are set, md_update_sb needed */ | |
fd01b88c | 214 | struct mddev { |
1da177e4 | 215 | void *private; |
84fc4b56 | 216 | struct md_personality *pers; |
1da177e4 LT |
217 | dev_t unit; |
218 | int md_minor; | |
7a0a5355 | 219 | struct list_head disks; |
850b2b42 | 220 | unsigned long flags; |
850b2b42 | 221 | |
409c57f3 N |
222 | int suspended; |
223 | atomic_t active_io; | |
1da177e4 | 224 | int ro; |
bb4f1e9d N |
225 | int sysfs_active; /* set when sysfs deletes |
226 | * are happening, so run/ | |
227 | * takeover/stop are not safe | |
228 | */ | |
1da177e4 LT |
229 | struct gendisk *gendisk; |
230 | ||
eae1701f | 231 | struct kobject kobj; |
d3374825 N |
232 | int hold_active; |
233 | #define UNTIL_IOCTL 1 | |
efeb53c0 | 234 | #define UNTIL_STOP 2 |
eae1701f | 235 | |
1da177e4 LT |
236 | /* Superblock information */ |
237 | int major_version, | |
238 | minor_version, | |
239 | patch_version; | |
240 | int persistent; | |
f72ffdd6 | 241 | int external; /* metadata is |
e691063a N |
242 | * managed externally */ |
243 | char metadata_type[17]; /* externally set*/ | |
9d8f0363 | 244 | int chunk_sectors; |
9ebc6ef1 | 245 | time64_t ctime, utime; |
1da177e4 | 246 | int level, layout; |
d9d166c2 | 247 | char clevel[16]; |
1da177e4 LT |
248 | int raid_disks; |
249 | int max_disks; | |
f72ffdd6 | 250 | sector_t dev_sectors; /* used size of |
58c0fed4 | 251 | * component devices */ |
f233ea5c | 252 | sector_t array_sectors; /* exported array size */ |
b522adcd DW |
253 | int external_size; /* size managed |
254 | * externally */ | |
1da177e4 | 255 | __u64 events; |
a8707c08 N |
256 | /* If the last 'event' was simply a clean->dirty transition, and |
257 | * we didn't write it to the spares, then it is safe and simple | |
258 | * to just decrement the event count on a dirty->clean transition. | |
259 | * So we record that possibility here. | |
260 | */ | |
261 | int can_decrease_events; | |
1da177e4 LT |
262 | |
263 | char uuid[16]; | |
264 | ||
f6705578 N |
265 | /* If the array is being reshaped, we need to record the |
266 | * new shape and an indication of where we are up to. | |
267 | * This is written to the superblock. | |
268 | * If reshape_position is MaxSector, then no reshape is happening (yet). | |
269 | */ | |
270 | sector_t reshape_position; | |
664e7c41 AN |
271 | int delta_disks, new_level, new_layout; |
272 | int new_chunk_sectors; | |
2c810cdd | 273 | int reshape_backwards; |
f6705578 | 274 | |
2b8bf345 N |
275 | struct md_thread *thread; /* management thread */ |
276 | struct md_thread *sync_thread; /* doing resync or reconstruct */ | |
c4a39551 JB |
277 | |
278 | /* 'last_sync_action' is initialized to "none". It is set when a | |
279 | * sync operation (i.e "data-check", "requested-resync", "resync", | |
280 | * "recovery", or "reshape") is started. It holds this value even | |
281 | * when the sync thread is "frozen" (interrupted) or "idle" (stopped | |
282 | * or finished). It is overwritten when a new sync operation is begun. | |
283 | */ | |
284 | char *last_sync_action; | |
ff4e8d9a | 285 | sector_t curr_resync; /* last block scheduled */ |
97e4f42d N |
286 | /* As resync requests can complete out of order, we cannot easily track |
287 | * how much resync has been completed. So we occasionally pause until | |
288 | * everything completes, then set curr_resync_completed to curr_resync. | |
289 | * As such it may be well behind the real resync mark, but it is a value | |
290 | * we are certain of. | |
291 | */ | |
292 | sector_t curr_resync_completed; | |
1da177e4 LT |
293 | unsigned long resync_mark; /* a recent timestamp */ |
294 | sector_t resync_mark_cnt;/* blocks written at resync_mark */ | |
ff4e8d9a | 295 | sector_t curr_mark_cnt; /* blocks scheduled now */ |
1da177e4 LT |
296 | |
297 | sector_t resync_max_sectors; /* may be set by personality */ | |
9d88883e | 298 | |
7f7583d4 | 299 | atomic64_t resync_mismatches; /* count of sectors where |
9d88883e N |
300 | * parity/replica mismatch found |
301 | */ | |
e464eafd N |
302 | |
303 | /* allow user-space to request suspension of IO to regions of the array */ | |
304 | sector_t suspend_lo; | |
305 | sector_t suspend_hi; | |
88202a0c N |
306 | /* if zero, use the system-wide default */ |
307 | int sync_speed_min; | |
308 | int sync_speed_max; | |
309 | ||
90b08710 BS |
310 | /* resync even though the same disks are shared among md-devices */ |
311 | int parallel_resync; | |
312 | ||
6ff8d8ec | 313 | int ok_start_degraded; |
5fd6c1dc | 314 | |
1da177e4 | 315 | unsigned long recovery; |
5389042f N |
316 | /* If a RAID personality determines that recovery (of a particular |
317 | * device) will fail due to a read error on the source device, it | |
318 | * takes a copy of this number and does not attempt recovery again | |
319 | * until this number changes. | |
320 | */ | |
321 | int recovery_disabled; | |
1da177e4 LT |
322 | |
323 | int in_sync; /* know to not need resync */ | |
c8c00a69 N |
324 | /* 'open_mutex' avoids races between 'md_open' and 'do_md_stop', so |
325 | * that we are never stopping an array while it is open. | |
326 | * 'reconfig_mutex' protects all other reconfiguration. | |
327 | * These locks are separate due to conflicting interactions | |
328 | * with bdev->bd_mutex. | |
329 | * Lock ordering is: | |
330 | * reconfig_mutex -> bd_mutex : e.g. do_md_run -> revalidate_disk | |
331 | * bd_mutex -> open_mutex: e.g. __blkdev_get -> md_open | |
332 | */ | |
333 | struct mutex open_mutex; | |
df5b89b3 | 334 | struct mutex reconfig_mutex; |
f2ea68cf N |
335 | atomic_t active; /* general refcount */ |
336 | atomic_t openers; /* number of active opens */ | |
1da177e4 | 337 | |
f0b4f7e2 N |
338 | int changed; /* True if we might need to |
339 | * reread partition info */ | |
1da177e4 LT |
340 | int degraded; /* whether md should consider |
341 | * adding a spare | |
342 | */ | |
343 | ||
344 | atomic_t recovery_active; /* blocks scheduled, but not written */ | |
345 | wait_queue_head_t recovery_wait; | |
346 | sector_t recovery_cp; | |
5e96ee65 NB |
347 | sector_t resync_min; /* user requested sync |
348 | * starts here */ | |
c6207277 N |
349 | sector_t resync_max; /* resync should pause |
350 | * when it gets here */ | |
06d91a5f | 351 | |
324a56e1 | 352 | struct kernfs_node *sysfs_state; /* handle for 'array_state' |
b62b7590 N |
353 | * file in sysfs. |
354 | */ | |
324a56e1 | 355 | struct kernfs_node *sysfs_action; /* handle for 'sync_action' */ |
b62b7590 | 356 | |
d3374825 N |
357 | struct work_struct del_work; /* used for delayed sysfs removal */ |
358 | ||
85572d7c N |
359 | /* "lock" protects: |
360 | * flush_bio transition from NULL to !NULL | |
361 | * rdev superblocks, events | |
362 | * clearing MD_CHANGE_* | |
363 | * in_sync - and related safemode and MD_CHANGE changes | |
36d091f4 | 364 | * pers (also protected by reconfig_mutex and pending IO). |
978a7a47 | 365 | * clearing ->bitmap |
4af1a041 | 366 | * clearing ->bitmap_info.file |
23da422b N |
367 | * changing ->resync_{min,max} |
368 | * setting MD_RECOVERY_RUNNING (which interacts with resync_{min,max}) | |
85572d7c N |
369 | */ |
370 | spinlock_t lock; | |
3d310eb7 | 371 | wait_queue_head_t sb_wait; /* for waiting on superblock updates */ |
7bfa19f2 | 372 | atomic_t pending_writes; /* number of active superblock writes */ |
06d91a5f | 373 | |
1da177e4 LT |
374 | unsigned int safemode; /* if set, update "clean" superblock |
375 | * when no writes pending. | |
f72ffdd6 | 376 | */ |
1da177e4 LT |
377 | unsigned int safemode_delay; |
378 | struct timer_list safemode_timer; | |
f72ffdd6 | 379 | atomic_t writes_pending; |
165125e1 | 380 | struct request_queue *queue; /* for plugging ... */ |
1da177e4 | 381 | |
f72ffdd6 | 382 | struct bitmap *bitmap; /* the bitmap for the device */ |
c3d9714e N |
383 | struct { |
384 | struct file *file; /* the bitmap file */ | |
f6af949c | 385 | loff_t offset; /* offset from superblock of |
c3d9714e N |
386 | * start of bitmap. May be |
387 | * negative, but not '0' | |
f6af949c | 388 | * For external metadata, offset |
f72ffdd6 | 389 | * from start of device. |
c3d9714e | 390 | */ |
6409bb05 | 391 | unsigned long space; /* space available at this offset */ |
f6af949c | 392 | loff_t default_offset; /* this is the offset to use when |
c3d9714e N |
393 | * hot-adding a bitmap. It should |
394 | * eventually be settable by sysfs. | |
395 | */ | |
6409bb05 N |
396 | unsigned long default_space; /* space available at |
397 | * default offset */ | |
c3d9714e | 398 | struct mutex mutex; |
42a04b50 | 399 | unsigned long chunksize; |
ac2f40be | 400 | unsigned long daemon_sleep; /* how many jiffies between updates? */ |
42a04b50 | 401 | unsigned long max_write_behind; /* write-behind mode */ |
ece5cff0 | 402 | int external; |
c4ce867f | 403 | int nodes; /* Maximum number of nodes in the cluster */ |
cf921cc1 | 404 | char cluster_name[64]; /* Name of the cluster */ |
c3d9714e | 405 | } bitmap_info; |
32a7627c | 406 | |
f72ffdd6 | 407 | atomic_t max_corr_read_errors; /* max read retries */ |
1da177e4 | 408 | struct list_head all_mddevs; |
a2826aa9 | 409 | |
a64c876f | 410 | struct attribute_group *to_remove; |
252ac522 | 411 | |
a167f663 N |
412 | struct bio_set *bio_set; |
413 | ||
e9c7469b TH |
414 | /* Generic flush handling. |
415 | * The last to finish preflush schedules a worker to submit | |
28a8f0d3 | 416 | * the rest of the request (without the REQ_PREFLUSH flag). |
a2826aa9 | 417 | */ |
e9c7469b | 418 | struct bio *flush_bio; |
a2826aa9 | 419 | atomic_t flush_pending; |
e9c7469b | 420 | struct work_struct flush_work; |
768a418d | 421 | struct work_struct event_work; /* used by dm to report failure event */ |
fd01b88c | 422 | void (*sync_super)(struct mddev *mddev, struct md_rdev *rdev); |
c4ce867f | 423 | struct md_cluster_info *cluster_info; |
15858fa5 | 424 | unsigned int good_device_nr; /* good device num within cluster raid */ |
be306c29 N |
425 | }; |
426 | ||
427 | enum recovery_flags { | |
428 | /* | |
429 | * If neither SYNC or RESHAPE are set, then it is a recovery. | |
430 | */ | |
431 | MD_RECOVERY_RUNNING, /* a thread is running, or about to be started */ | |
432 | MD_RECOVERY_SYNC, /* actually doing a resync, not a recovery */ | |
433 | MD_RECOVERY_RECOVER, /* doing recovery, or need to try it. */ | |
434 | MD_RECOVERY_INTR, /* resync needs to be aborted for some reason */ | |
435 | MD_RECOVERY_DONE, /* thread is done and is waiting to be reaped */ | |
436 | MD_RECOVERY_NEEDED, /* we might need to start a resync/recover */ | |
437 | MD_RECOVERY_REQUESTED, /* user-space has requested a sync (used with SYNC) */ | |
438 | MD_RECOVERY_CHECK, /* user-space request for check-only, no repair */ | |
439 | MD_RECOVERY_RESHAPE, /* A reshape is happening */ | |
440 | MD_RECOVERY_FROZEN, /* User request to abort, and not restart, any action */ | |
441 | MD_RECOVERY_ERROR, /* sync-action interrupted because io-error */ | |
1da177e4 LT |
442 | }; |
443 | ||
5c47daf6 N |
444 | static inline int __must_check mddev_lock(struct mddev *mddev) |
445 | { | |
446 | return mutex_lock_interruptible(&mddev->reconfig_mutex); | |
447 | } | |
448 | ||
449 | /* Sometimes we need to take the lock in a situation where | |
450 | * failure due to interrupts is not acceptable. | |
451 | */ | |
452 | static inline void mddev_lock_nointr(struct mddev *mddev) | |
453 | { | |
454 | mutex_lock(&mddev->reconfig_mutex); | |
455 | } | |
456 | ||
457 | static inline int mddev_is_locked(struct mddev *mddev) | |
458 | { | |
459 | return mutex_is_locked(&mddev->reconfig_mutex); | |
460 | } | |
461 | ||
462 | static inline int mddev_trylock(struct mddev *mddev) | |
463 | { | |
464 | return mutex_trylock(&mddev->reconfig_mutex); | |
465 | } | |
466 | extern void mddev_unlock(struct mddev *mddev); | |
467 | ||
1da177e4 LT |
468 | static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors) |
469 | { | |
f72ffdd6 | 470 | atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io); |
1da177e4 LT |
471 | } |
472 | ||
84fc4b56 | 473 | struct md_personality |
1da177e4 LT |
474 | { |
475 | char *name; | |
2604b703 N |
476 | int level; |
477 | struct list_head list; | |
1da177e4 | 478 | struct module *owner; |
b4fdcb02 | 479 | void (*make_request)(struct mddev *mddev, struct bio *bio); |
fd01b88c | 480 | int (*run)(struct mddev *mddev); |
afa0f557 | 481 | void (*free)(struct mddev *mddev, void *priv); |
fd01b88c | 482 | void (*status)(struct seq_file *seq, struct mddev *mddev); |
1da177e4 | 483 | /* error_handler must set ->faulty and clear ->in_sync |
f72ffdd6 | 484 | * if appropriate, and should abort recovery if needed |
1da177e4 | 485 | */ |
fd01b88c N |
486 | void (*error_handler)(struct mddev *mddev, struct md_rdev *rdev); |
487 | int (*hot_add_disk) (struct mddev *mddev, struct md_rdev *rdev); | |
b8321b68 | 488 | int (*hot_remove_disk) (struct mddev *mddev, struct md_rdev *rdev); |
fd01b88c | 489 | int (*spare_active) (struct mddev *mddev); |
09314799 | 490 | sector_t (*sync_request)(struct mddev *mddev, sector_t sector_nr, int *skipped); |
fd01b88c N |
491 | int (*resize) (struct mddev *mddev, sector_t sectors); |
492 | sector_t (*size) (struct mddev *mddev, sector_t sectors, int raid_disks); | |
493 | int (*check_reshape) (struct mddev *mddev); | |
494 | int (*start_reshape) (struct mddev *mddev); | |
495 | void (*finish_reshape) (struct mddev *mddev); | |
36fa3063 N |
496 | /* quiesce moves between quiescence states |
497 | * 0 - fully active | |
498 | * 1 - no new requests allowed | |
499 | * others - reserved | |
500 | */ | |
fd01b88c | 501 | void (*quiesce) (struct mddev *mddev, int state); |
245f46c2 N |
502 | /* takeover is used to transition an array from one |
503 | * personality to another. The new personality must be able | |
504 | * to handle the data in the current layout. | |
505 | * e.g. 2drive raid1 -> 2drive raid5 | |
506 | * ndrive raid5 -> degraded n+1drive raid6 with special layout | |
507 | * If the takeover succeeds, a new 'private' structure is returned. | |
508 | * This needs to be installed and then ->run used to activate the | |
509 | * array. | |
510 | */ | |
fd01b88c | 511 | void *(*takeover) (struct mddev *mddev); |
5c675f83 N |
512 | /* congested implements bdi.congested_fn(). |
513 | * Will not be called while array is 'suspended' */ | |
514 | int (*congested)(struct mddev *mddev, int bits); | |
1da177e4 LT |
515 | }; |
516 | ||
007583c9 N |
517 | struct md_sysfs_entry { |
518 | struct attribute attr; | |
fd01b88c N |
519 | ssize_t (*show)(struct mddev *, char *); |
520 | ssize_t (*store)(struct mddev *, const char *, size_t); | |
007583c9 | 521 | }; |
43a70507 | 522 | extern struct attribute_group md_bitmap_group; |
007583c9 | 523 | |
324a56e1 | 524 | static inline struct kernfs_node *sysfs_get_dirent_safe(struct kernfs_node *sd, char *name) |
00bcb4ac N |
525 | { |
526 | if (sd) | |
388975cc | 527 | return sysfs_get_dirent(sd, name); |
00bcb4ac N |
528 | return sd; |
529 | } | |
324a56e1 | 530 | static inline void sysfs_notify_dirent_safe(struct kernfs_node *sd) |
00bcb4ac N |
531 | { |
532 | if (sd) | |
533 | sysfs_notify_dirent(sd); | |
534 | } | |
535 | ||
fd01b88c | 536 | static inline char * mdname (struct mddev * mddev) |
1da177e4 LT |
537 | { |
538 | return mddev->gendisk ? mddev->gendisk->disk_name : "mdX"; | |
539 | } | |
540 | ||
fd01b88c | 541 | static inline int sysfs_link_rdev(struct mddev *mddev, struct md_rdev *rdev) |
36fad858 NK |
542 | { |
543 | char nm[20]; | |
9b15603d SL |
544 | if (!test_bit(Replacement, &rdev->flags) && |
545 | !test_bit(Journal, &rdev->flags) && | |
546 | mddev->kobj.sd) { | |
2d78f8c4 N |
547 | sprintf(nm, "rd%d", rdev->raid_disk); |
548 | return sysfs_create_link(&mddev->kobj, &rdev->kobj, nm); | |
549 | } else | |
550 | return 0; | |
36fad858 NK |
551 | } |
552 | ||
fd01b88c | 553 | static inline void sysfs_unlink_rdev(struct mddev *mddev, struct md_rdev *rdev) |
36fad858 NK |
554 | { |
555 | char nm[20]; | |
9b15603d SL |
556 | if (!test_bit(Replacement, &rdev->flags) && |
557 | !test_bit(Journal, &rdev->flags) && | |
558 | mddev->kobj.sd) { | |
2d78f8c4 N |
559 | sprintf(nm, "rd%d", rdev->raid_disk); |
560 | sysfs_remove_link(&mddev->kobj, nm); | |
561 | } | |
36fad858 NK |
562 | } |
563 | ||
1da177e4 LT |
564 | /* |
565 | * iterates through some rdev ringlist. It's safe to remove the | |
566 | * current 'rdev'. Dont touch 'tmp' though. | |
567 | */ | |
159ec1fc CR |
568 | #define rdev_for_each_list(rdev, tmp, head) \ |
569 | list_for_each_entry_safe(rdev, tmp, head, same_set) | |
570 | ||
1da177e4 LT |
571 | /* |
572 | * iterates through the 'same array disks' ringlist | |
573 | */ | |
dafb20fa N |
574 | #define rdev_for_each(rdev, mddev) \ |
575 | list_for_each_entry(rdev, &((mddev)->disks), same_set) | |
576 | ||
577 | #define rdev_for_each_safe(rdev, tmp, mddev) \ | |
159ec1fc | 578 | list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set) |
1da177e4 | 579 | |
4b80991c N |
580 | #define rdev_for_each_rcu(rdev, mddev) \ |
581 | list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set) | |
582 | ||
2b8bf345 | 583 | struct md_thread { |
4ed8731d | 584 | void (*run) (struct md_thread *thread); |
fd01b88c | 585 | struct mddev *mddev; |
1da177e4 | 586 | wait_queue_head_t wqueue; |
f72ffdd6 | 587 | unsigned long flags; |
1da177e4 | 588 | struct task_struct *tsk; |
32a7627c | 589 | unsigned long timeout; |
4ed8731d | 590 | void *private; |
2b8bf345 | 591 | }; |
1da177e4 LT |
592 | |
593 | #define THREAD_WAKEUP 0 | |
594 | ||
1345b1d8 N |
595 | static inline void safe_put_page(struct page *p) |
596 | { | |
597 | if (p) put_page(p); | |
598 | } | |
599 | ||
84fc4b56 N |
600 | extern int register_md_personality(struct md_personality *p); |
601 | extern int unregister_md_personality(struct md_personality *p); | |
edb39c9d GR |
602 | extern int register_md_cluster_operations(struct md_cluster_operations *ops, |
603 | struct module *module); | |
604 | extern int unregister_md_cluster_operations(void); | |
605 | extern int md_setup_cluster(struct mddev *mddev, int nodes); | |
606 | extern void md_cluster_stop(struct mddev *mddev); | |
2b8bf345 | 607 | extern struct md_thread *md_register_thread( |
4ed8731d | 608 | void (*run)(struct md_thread *thread), |
2b8bf345 N |
609 | struct mddev *mddev, |
610 | const char *name); | |
611 | extern void md_unregister_thread(struct md_thread **threadp); | |
612 | extern void md_wakeup_thread(struct md_thread *thread); | |
fd01b88c | 613 | extern void md_check_recovery(struct mddev *mddev); |
a91d5ac0 | 614 | extern void md_reap_sync_thread(struct mddev *mddev); |
fd01b88c N |
615 | extern void md_write_start(struct mddev *mddev, struct bio *bi); |
616 | extern void md_write_end(struct mddev *mddev); | |
617 | extern void md_done_sync(struct mddev *mddev, int blocks, int ok); | |
618 | extern void md_error(struct mddev *mddev, struct md_rdev *rdev); | |
c6563a8c | 619 | extern void md_finish_reshape(struct mddev *mddev); |
fd01b88c N |
620 | |
621 | extern int mddev_congested(struct mddev *mddev, int bits); | |
622 | extern void md_flush_request(struct mddev *mddev, struct bio *bio); | |
623 | extern void md_super_write(struct mddev *mddev, struct md_rdev *rdev, | |
92022950 | 624 | sector_t sector, int size, struct page *page); |
fd01b88c | 625 | extern void md_super_wait(struct mddev *mddev); |
f72ffdd6 | 626 | extern int sync_page_io(struct md_rdev *rdev, sector_t sector, int size, |
796a5cf0 MC |
627 | struct page *page, int op, int op_flags, |
628 | bool metadata_op); | |
4ed8731d | 629 | extern void md_do_sync(struct md_thread *thread); |
fd01b88c N |
630 | extern void md_new_event(struct mddev *mddev); |
631 | extern int md_allow_write(struct mddev *mddev); | |
632 | extern void md_wait_for_blocked_rdev(struct md_rdev *rdev, struct mddev *mddev); | |
633 | extern void md_set_array_sectors(struct mddev *mddev, sector_t array_sectors); | |
634 | extern int md_check_no_bitmap(struct mddev *mddev); | |
635 | extern int md_integrity_register(struct mddev *mddev); | |
1501efad | 636 | extern int md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev); |
72e02075 | 637 | extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale); |
63fe0817 | 638 | |
fd01b88c N |
639 | extern void mddev_init(struct mddev *mddev); |
640 | extern int md_run(struct mddev *mddev); | |
641 | extern void md_stop(struct mddev *mddev); | |
642 | extern void md_stop_writes(struct mddev *mddev); | |
3cb03002 | 643 | extern int md_rdev_init(struct md_rdev *rdev); |
545c8795 | 644 | extern void md_rdev_clear(struct md_rdev *rdev); |
63fe0817 | 645 | |
fd01b88c N |
646 | extern void mddev_suspend(struct mddev *mddev); |
647 | extern void mddev_resume(struct mddev *mddev); | |
a167f663 | 648 | extern struct bio *bio_clone_mddev(struct bio *bio, gfp_t gfp_mask, |
fd01b88c | 649 | struct mddev *mddev); |
a167f663 | 650 | extern struct bio *bio_alloc_mddev(gfp_t gfp_mask, int nr_iovecs, |
fd01b88c | 651 | struct mddev *mddev); |
9cbb1750 | 652 | |
74018dc3 | 653 | extern void md_unplug(struct blk_plug_cb *cb, bool from_schedule); |
70bcecdb | 654 | extern void md_reload_sb(struct mddev *mddev, int raid_disk); |
1aee41f6 | 655 | extern void md_update_sb(struct mddev *mddev, int force); |
fb56dfef | 656 | extern void md_kick_rdev_from_array(struct md_rdev * rdev); |
57d051dc | 657 | struct md_rdev *md_find_rdev_nr_rcu(struct mddev *mddev, int nr); |
9cbb1750 N |
658 | static inline int mddev_check_plugged(struct mddev *mddev) |
659 | { | |
660 | return !!blk_check_plugged(md_unplug, mddev, | |
661 | sizeof(struct blk_plug_cb)); | |
662 | } | |
dfe15ac1 HR |
663 | |
664 | static inline void rdev_dec_pending(struct md_rdev *rdev, struct mddev *mddev) | |
665 | { | |
666 | int faulty = test_bit(Faulty, &rdev->flags); | |
667 | if (atomic_dec_and_test(&rdev->nr_pending) && faulty) { | |
668 | set_bit(MD_RECOVERY_NEEDED, &mddev->recovery); | |
669 | md_wakeup_thread(mddev->thread); | |
670 | } | |
671 | } | |
672 | ||
edb39c9d | 673 | extern struct md_cluster_operations *md_cluster_ops; |
c4ce867f GR |
674 | static inline int mddev_is_clustered(struct mddev *mddev) |
675 | { | |
676 | return mddev->cluster_info && mddev->bitmap_info.nodes > 1; | |
677 | } | |
63fe0817 | 678 | #endif /* _MD_MD_H */ |