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 | |
46533ff7 N |
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) | |
1da177e4 LT |
42 | /* |
43 | * MD's 'extended' device | |
44 | */ | |
3cb03002 | 45 | struct md_rdev { |
1da177e4 LT |
46 | struct list_head same_set; /* RAID devices within the same set */ |
47 | ||
dd8ac336 | 48 | sector_t sectors; /* Device size (in 512bytes sectors) */ |
fd01b88c | 49 | struct mddev *mddev; /* RAID array if running */ |
eea1bf38 | 50 | int last_events; /* IO event timestamp */ |
1da177e4 | 51 | |
a6ff7e08 JB |
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; | |
1da177e4 LT |
58 | struct block_device *bdev; /* block device handle */ |
59 | ||
2699b672 | 60 | struct page *sb_page, *bb_page; |
1da177e4 | 61 | int sb_loaded; |
42543769 | 62 | __u64 sb_events; |
1da177e4 | 63 | sector_t data_offset; /* start of data in array */ |
c6563a8c | 64 | sector_t new_data_offset;/* only relevant while reshaping */ |
f72ffdd6 | 65 | sector_t sb_start; /* offset of the super block (in 512byte sectors) */ |
0002b271 | 66 | int sb_size; /* bytes in the superblock */ |
1da177e4 LT |
67 | int preferred_minor; /* autorun support */ |
68 | ||
86e6ffdd N |
69 | struct kobject kobj; |
70 | ||
1da177e4 LT |
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 | */ | |
1da177e4 | 81 | |
2d78f8c4 | 82 | unsigned long flags; /* bit set of 'enum flag_bits' bits. */ |
6bfe0b49 | 83 | wait_queue_head_t blocked_wait; |
8ddf9efe | 84 | |
1da177e4 LT |
85 | int desc_nr; /* descriptor index in the superblock */ |
86 | int raid_disk; /* role of device in array */ | |
e93f68a1 N |
87 | int new_raid_disk; /* role that the device will have in |
88 | * the array after a level-change completes. | |
89 | */ | |
41158c7e N |
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 | */ | |
3069aa8d SL |
94 | union { |
95 | sector_t recovery_offset;/* If this device has been partially | |
5fd6c1dc N |
96 | * recovered, this is where we were |
97 | * up to. | |
98 | */ | |
3069aa8d SL |
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 | }; | |
1da177e4 LT |
104 | |
105 | atomic_t nr_pending; /* number of pending requests. | |
106 | * only maintained for arrays that | |
107 | * support hot removal | |
108 | */ | |
ba22dcbf N |
109 | atomic_t read_errors; /* number of consecutive read errors that |
110 | * we have tried to ignore. | |
111 | */ | |
0e3ef49e | 112 | time64_t last_read_error; /* monotonic time since our |
1e50915f RB |
113 | * last read error |
114 | */ | |
4dbcdc75 N |
115 | atomic_t corrected_errors; /* number of corrected read errors, |
116 | * for reporting to userspace and storing | |
117 | * in superblock. | |
118 | */ | |
5792a285 | 119 | struct work_struct del_work; /* used for delayed sysfs removal */ |
3c0ee63a | 120 | |
324a56e1 | 121 | struct kernfs_node *sysfs_state; /* handle for 'state' |
3c0ee63a | 122 | * sysfs entry */ |
2230dfe4 | 123 | |
fc974ee2 | 124 | struct badblocks badblocks; |
1da177e4 | 125 | }; |
2d78f8c4 N |
126 | enum flag_bits { |
127 | Faulty, /* device is known to have a fault */ | |
128 | In_sync, /* device is in_sync with rest of array */ | |
8313b8e5 N |
129 | Bitmap_sync, /* ..actually, not quite In_sync. Need a |
130 | * bitmap-based recovery to get fully in sync | |
131 | */ | |
2d78f8c4 N |
132 | WriteMostly, /* Avoid reading if at all possible */ |
133 | AutoDetected, /* added by auto-detect */ | |
134 | Blocked, /* An error occurred but has not yet | |
135 | * been acknowledged by the metadata | |
136 | * handler, so don't allow writes | |
137 | * until it is cleared */ | |
138 | WriteErrorSeen, /* A write error has been seen on this | |
139 | * device | |
140 | */ | |
141 | FaultRecorded, /* Intermediate state for clearing | |
142 | * Blocked. The Fault is/will-be | |
143 | * recorded in the metadata, but that | |
144 | * metadata hasn't been stored safely | |
145 | * on disk yet. | |
146 | */ | |
147 | BlockedBadBlocks, /* A writer is blocked because they | |
148 | * found an unacknowledged bad-block. | |
149 | * This can safely be cleared at any | |
150 | * time, and the writer will re-check. | |
151 | * It may be set at any time, and at | |
152 | * worst the writer will timeout and | |
153 | * re-check. So setting it as | |
154 | * accurately as possible is good, but | |
155 | * not absolutely critical. | |
156 | */ | |
157 | WantReplacement, /* This device is a candidate to be | |
158 | * hot-replaced, either because it has | |
159 | * reported some faults, or because | |
160 | * of explicit request. | |
161 | */ | |
162 | Replacement, /* This device is a replacement for | |
163 | * a want_replacement device with same | |
164 | * raid_disk number. | |
165 | */ | |
1aee41f6 GR |
166 | Candidate, /* For clustered environments only: |
167 | * This device is seen locally but not | |
168 | * by the whole cluster | |
169 | */ | |
bac624f3 SL |
170 | Journal, /* This device is used as journal for |
171 | * raid-5/6. | |
172 | * Usually, this device should be faster | |
173 | * than other devices in the array | |
174 | */ | |
659b254f | 175 | ClusterRemove, |
d787be40 N |
176 | RemoveSynchronized, /* synchronize_rcu() was called after |
177 | * this device was known to be faulty, | |
178 | * so it is safe to remove without | |
179 | * another synchronize_rcu() call. | |
180 | */ | |
35b785f7 TM |
181 | ExternalBbl, /* External metadata provides bad |
182 | * block management for a disk | |
183 | */ | |
688834e6 N |
184 | FailFast, /* Minimal retries should be attempted on |
185 | * this device, so use REQ_FAILFAST_DEV. | |
186 | * Also don't try to repair failed reads. | |
187 | * It is expects that no bad block log | |
188 | * is present. | |
189 | */ | |
46533ff7 N |
190 | LastDev, /* Seems to be the last working dev as |
191 | * it didn't fail, so don't use FailFast | |
192 | * any more for metadata | |
193 | */ | |
2d78f8c4 | 194 | }; |
1da177e4 | 195 | |
3cb03002 | 196 | static inline int is_badblock(struct md_rdev *rdev, sector_t s, int sectors, |
2230dfe4 N |
197 | sector_t *first_bad, int *bad_sectors) |
198 | { | |
199 | if (unlikely(rdev->badblocks.count)) { | |
fc974ee2 | 200 | int rv = badblocks_check(&rdev->badblocks, rdev->data_offset + s, |
2230dfe4 N |
201 | sectors, |
202 | first_bad, bad_sectors); | |
203 | if (rv) | |
204 | *first_bad -= rdev->data_offset; | |
205 | return rv; | |
206 | } | |
207 | return 0; | |
208 | } | |
3cb03002 | 209 | extern int rdev_set_badblocks(struct md_rdev *rdev, sector_t s, int sectors, |
c6563a8c N |
210 | int is_new); |
211 | extern int rdev_clear_badblocks(struct md_rdev *rdev, sector_t s, int sectors, | |
212 | int is_new); | |
c4ce867f GR |
213 | struct md_cluster_info; |
214 | ||
394ed8e4 | 215 | /* change UNSUPPORTED_MDDEV_FLAGS for each array type if new flag is added */ |
be306c29 | 216 | enum mddev_flags { |
be306c29 N |
217 | MD_ARRAY_FIRST_USE, /* First use of array, needs initialization */ |
218 | MD_CLOSING, /* If set, we are closing the array, do not open | |
219 | * it then */ | |
220 | MD_JOURNAL_CLEAN, /* A raid with journal is already clean */ | |
221 | MD_HAS_JOURNAL, /* The raid array has journal feature set */ | |
222 | MD_RELOAD_SB, /* Reload the superblock because another node | |
223 | * updated it. | |
224 | */ | |
225 | MD_CLUSTER_RESYNC_LOCKED, /* cluster raid only, which means node | |
226 | * already took resync lock, need to | |
227 | * release the lock */ | |
46533ff7 N |
228 | MD_FAILFAST_SUPPORTED, /* Using MD_FAILFAST on metadata writes is |
229 | * supported as calls to md_error() will | |
230 | * never cause the array to become failed. | |
231 | */ | |
be306c29 | 232 | }; |
2953079c SL |
233 | |
234 | enum mddev_sb_flags { | |
235 | MD_SB_CHANGE_DEVS, /* Some device status has changed */ | |
236 | MD_SB_CHANGE_CLEAN, /* transition to or from 'clean' */ | |
237 | MD_SB_CHANGE_PENDING, /* switch from 'clean' to 'active' in progress */ | |
238 | MD_SB_NEED_REWRITE, /* metadata write needs to be repeated */ | |
239 | }; | |
240 | ||
fd01b88c | 241 | struct mddev { |
1da177e4 | 242 | void *private; |
84fc4b56 | 243 | struct md_personality *pers; |
1da177e4 LT |
244 | dev_t unit; |
245 | int md_minor; | |
7a0a5355 | 246 | struct list_head disks; |
850b2b42 | 247 | unsigned long flags; |
2953079c | 248 | unsigned long sb_flags; |
850b2b42 | 249 | |
409c57f3 N |
250 | int suspended; |
251 | atomic_t active_io; | |
1da177e4 | 252 | int ro; |
bb4f1e9d N |
253 | int sysfs_active; /* set when sysfs deletes |
254 | * are happening, so run/ | |
255 | * takeover/stop are not safe | |
256 | */ | |
1da177e4 LT |
257 | struct gendisk *gendisk; |
258 | ||
eae1701f | 259 | struct kobject kobj; |
d3374825 N |
260 | int hold_active; |
261 | #define UNTIL_IOCTL 1 | |
efeb53c0 | 262 | #define UNTIL_STOP 2 |
eae1701f | 263 | |
1da177e4 LT |
264 | /* Superblock information */ |
265 | int major_version, | |
266 | minor_version, | |
267 | patch_version; | |
268 | int persistent; | |
f72ffdd6 | 269 | int external; /* metadata is |
e691063a N |
270 | * managed externally */ |
271 | char metadata_type[17]; /* externally set*/ | |
9d8f0363 | 272 | int chunk_sectors; |
9ebc6ef1 | 273 | time64_t ctime, utime; |
1da177e4 | 274 | int level, layout; |
d9d166c2 | 275 | char clevel[16]; |
1da177e4 LT |
276 | int raid_disks; |
277 | int max_disks; | |
f72ffdd6 | 278 | sector_t dev_sectors; /* used size of |
58c0fed4 | 279 | * component devices */ |
f233ea5c | 280 | sector_t array_sectors; /* exported array size */ |
b522adcd DW |
281 | int external_size; /* size managed |
282 | * externally */ | |
1da177e4 | 283 | __u64 events; |
a8707c08 N |
284 | /* If the last 'event' was simply a clean->dirty transition, and |
285 | * we didn't write it to the spares, then it is safe and simple | |
286 | * to just decrement the event count on a dirty->clean transition. | |
287 | * So we record that possibility here. | |
288 | */ | |
289 | int can_decrease_events; | |
1da177e4 LT |
290 | |
291 | char uuid[16]; | |
292 | ||
f6705578 N |
293 | /* If the array is being reshaped, we need to record the |
294 | * new shape and an indication of where we are up to. | |
295 | * This is written to the superblock. | |
296 | * If reshape_position is MaxSector, then no reshape is happening (yet). | |
297 | */ | |
298 | sector_t reshape_position; | |
664e7c41 AN |
299 | int delta_disks, new_level, new_layout; |
300 | int new_chunk_sectors; | |
2c810cdd | 301 | int reshape_backwards; |
f6705578 | 302 | |
2b8bf345 N |
303 | struct md_thread *thread; /* management thread */ |
304 | struct md_thread *sync_thread; /* doing resync or reconstruct */ | |
c4a39551 JB |
305 | |
306 | /* 'last_sync_action' is initialized to "none". It is set when a | |
307 | * sync operation (i.e "data-check", "requested-resync", "resync", | |
308 | * "recovery", or "reshape") is started. It holds this value even | |
309 | * when the sync thread is "frozen" (interrupted) or "idle" (stopped | |
310 | * or finished). It is overwritten when a new sync operation is begun. | |
311 | */ | |
312 | char *last_sync_action; | |
ff4e8d9a | 313 | sector_t curr_resync; /* last block scheduled */ |
97e4f42d N |
314 | /* As resync requests can complete out of order, we cannot easily track |
315 | * how much resync has been completed. So we occasionally pause until | |
316 | * everything completes, then set curr_resync_completed to curr_resync. | |
317 | * As such it may be well behind the real resync mark, but it is a value | |
318 | * we are certain of. | |
319 | */ | |
320 | sector_t curr_resync_completed; | |
1da177e4 LT |
321 | unsigned long resync_mark; /* a recent timestamp */ |
322 | sector_t resync_mark_cnt;/* blocks written at resync_mark */ | |
ff4e8d9a | 323 | sector_t curr_mark_cnt; /* blocks scheduled now */ |
1da177e4 LT |
324 | |
325 | sector_t resync_max_sectors; /* may be set by personality */ | |
9d88883e | 326 | |
7f7583d4 | 327 | atomic64_t resync_mismatches; /* count of sectors where |
9d88883e N |
328 | * parity/replica mismatch found |
329 | */ | |
e464eafd N |
330 | |
331 | /* allow user-space to request suspension of IO to regions of the array */ | |
332 | sector_t suspend_lo; | |
333 | sector_t suspend_hi; | |
88202a0c N |
334 | /* if zero, use the system-wide default */ |
335 | int sync_speed_min; | |
336 | int sync_speed_max; | |
337 | ||
90b08710 BS |
338 | /* resync even though the same disks are shared among md-devices */ |
339 | int parallel_resync; | |
340 | ||
6ff8d8ec | 341 | int ok_start_degraded; |
5fd6c1dc | 342 | |
1da177e4 | 343 | unsigned long recovery; |
5389042f N |
344 | /* If a RAID personality determines that recovery (of a particular |
345 | * device) will fail due to a read error on the source device, it | |
346 | * takes a copy of this number and does not attempt recovery again | |
347 | * until this number changes. | |
348 | */ | |
349 | int recovery_disabled; | |
1da177e4 LT |
350 | |
351 | int in_sync; /* know to not need resync */ | |
c8c00a69 N |
352 | /* 'open_mutex' avoids races between 'md_open' and 'do_md_stop', so |
353 | * that we are never stopping an array while it is open. | |
354 | * 'reconfig_mutex' protects all other reconfiguration. | |
355 | * These locks are separate due to conflicting interactions | |
356 | * with bdev->bd_mutex. | |
357 | * Lock ordering is: | |
358 | * reconfig_mutex -> bd_mutex : e.g. do_md_run -> revalidate_disk | |
359 | * bd_mutex -> open_mutex: e.g. __blkdev_get -> md_open | |
360 | */ | |
361 | struct mutex open_mutex; | |
df5b89b3 | 362 | struct mutex reconfig_mutex; |
f2ea68cf N |
363 | atomic_t active; /* general refcount */ |
364 | atomic_t openers; /* number of active opens */ | |
1da177e4 | 365 | |
f0b4f7e2 N |
366 | int changed; /* True if we might need to |
367 | * reread partition info */ | |
1da177e4 LT |
368 | int degraded; /* whether md should consider |
369 | * adding a spare | |
370 | */ | |
371 | ||
372 | atomic_t recovery_active; /* blocks scheduled, but not written */ | |
373 | wait_queue_head_t recovery_wait; | |
374 | sector_t recovery_cp; | |
5e96ee65 NB |
375 | sector_t resync_min; /* user requested sync |
376 | * starts here */ | |
c6207277 N |
377 | sector_t resync_max; /* resync should pause |
378 | * when it gets here */ | |
06d91a5f | 379 | |
324a56e1 | 380 | struct kernfs_node *sysfs_state; /* handle for 'array_state' |
b62b7590 N |
381 | * file in sysfs. |
382 | */ | |
324a56e1 | 383 | struct kernfs_node *sysfs_action; /* handle for 'sync_action' */ |
b62b7590 | 384 | |
d3374825 N |
385 | struct work_struct del_work; /* used for delayed sysfs removal */ |
386 | ||
85572d7c N |
387 | /* "lock" protects: |
388 | * flush_bio transition from NULL to !NULL | |
389 | * rdev superblocks, events | |
390 | * clearing MD_CHANGE_* | |
391 | * in_sync - and related safemode and MD_CHANGE changes | |
36d091f4 | 392 | * pers (also protected by reconfig_mutex and pending IO). |
978a7a47 | 393 | * clearing ->bitmap |
4af1a041 | 394 | * clearing ->bitmap_info.file |
23da422b N |
395 | * changing ->resync_{min,max} |
396 | * setting MD_RECOVERY_RUNNING (which interacts with resync_{min,max}) | |
85572d7c N |
397 | */ |
398 | spinlock_t lock; | |
3d310eb7 | 399 | wait_queue_head_t sb_wait; /* for waiting on superblock updates */ |
7bfa19f2 | 400 | atomic_t pending_writes; /* number of active superblock writes */ |
06d91a5f | 401 | |
1da177e4 LT |
402 | unsigned int safemode; /* if set, update "clean" superblock |
403 | * when no writes pending. | |
f72ffdd6 | 404 | */ |
1da177e4 LT |
405 | unsigned int safemode_delay; |
406 | struct timer_list safemode_timer; | |
f72ffdd6 | 407 | atomic_t writes_pending; |
165125e1 | 408 | struct request_queue *queue; /* for plugging ... */ |
1da177e4 | 409 | |
f72ffdd6 | 410 | struct bitmap *bitmap; /* the bitmap for the device */ |
c3d9714e N |
411 | struct { |
412 | struct file *file; /* the bitmap file */ | |
f6af949c | 413 | loff_t offset; /* offset from superblock of |
c3d9714e N |
414 | * start of bitmap. May be |
415 | * negative, but not '0' | |
f6af949c | 416 | * For external metadata, offset |
f72ffdd6 | 417 | * from start of device. |
c3d9714e | 418 | */ |
6409bb05 | 419 | unsigned long space; /* space available at this offset */ |
f6af949c | 420 | loff_t default_offset; /* this is the offset to use when |
c3d9714e N |
421 | * hot-adding a bitmap. It should |
422 | * eventually be settable by sysfs. | |
423 | */ | |
6409bb05 N |
424 | unsigned long default_space; /* space available at |
425 | * default offset */ | |
c3d9714e | 426 | struct mutex mutex; |
42a04b50 | 427 | unsigned long chunksize; |
ac2f40be | 428 | unsigned long daemon_sleep; /* how many jiffies between updates? */ |
42a04b50 | 429 | unsigned long max_write_behind; /* write-behind mode */ |
ece5cff0 | 430 | int external; |
c4ce867f | 431 | int nodes; /* Maximum number of nodes in the cluster */ |
cf921cc1 | 432 | char cluster_name[64]; /* Name of the cluster */ |
c3d9714e | 433 | } bitmap_info; |
32a7627c | 434 | |
f72ffdd6 | 435 | atomic_t max_corr_read_errors; /* max read retries */ |
1da177e4 | 436 | struct list_head all_mddevs; |
a2826aa9 | 437 | |
a64c876f | 438 | struct attribute_group *to_remove; |
252ac522 | 439 | |
a167f663 N |
440 | struct bio_set *bio_set; |
441 | ||
e9c7469b TH |
442 | /* Generic flush handling. |
443 | * The last to finish preflush schedules a worker to submit | |
28a8f0d3 | 444 | * the rest of the request (without the REQ_PREFLUSH flag). |
a2826aa9 | 445 | */ |
e9c7469b | 446 | struct bio *flush_bio; |
a2826aa9 | 447 | atomic_t flush_pending; |
e9c7469b | 448 | struct work_struct flush_work; |
768a418d | 449 | struct work_struct event_work; /* used by dm to report failure event */ |
fd01b88c | 450 | void (*sync_super)(struct mddev *mddev, struct md_rdev *rdev); |
c4ce867f | 451 | struct md_cluster_info *cluster_info; |
15858fa5 | 452 | unsigned int good_device_nr; /* good device num within cluster raid */ |
be306c29 N |
453 | }; |
454 | ||
455 | enum recovery_flags { | |
456 | /* | |
457 | * If neither SYNC or RESHAPE are set, then it is a recovery. | |
458 | */ | |
459 | MD_RECOVERY_RUNNING, /* a thread is running, or about to be started */ | |
460 | MD_RECOVERY_SYNC, /* actually doing a resync, not a recovery */ | |
461 | MD_RECOVERY_RECOVER, /* doing recovery, or need to try it. */ | |
462 | MD_RECOVERY_INTR, /* resync needs to be aborted for some reason */ | |
463 | MD_RECOVERY_DONE, /* thread is done and is waiting to be reaped */ | |
464 | MD_RECOVERY_NEEDED, /* we might need to start a resync/recover */ | |
465 | MD_RECOVERY_REQUESTED, /* user-space has requested a sync (used with SYNC) */ | |
466 | MD_RECOVERY_CHECK, /* user-space request for check-only, no repair */ | |
467 | MD_RECOVERY_RESHAPE, /* A reshape is happening */ | |
468 | MD_RECOVERY_FROZEN, /* User request to abort, and not restart, any action */ | |
469 | MD_RECOVERY_ERROR, /* sync-action interrupted because io-error */ | |
1da177e4 LT |
470 | }; |
471 | ||
5c47daf6 N |
472 | static inline int __must_check mddev_lock(struct mddev *mddev) |
473 | { | |
474 | return mutex_lock_interruptible(&mddev->reconfig_mutex); | |
475 | } | |
476 | ||
477 | /* Sometimes we need to take the lock in a situation where | |
478 | * failure due to interrupts is not acceptable. | |
479 | */ | |
480 | static inline void mddev_lock_nointr(struct mddev *mddev) | |
481 | { | |
482 | mutex_lock(&mddev->reconfig_mutex); | |
483 | } | |
484 | ||
485 | static inline int mddev_is_locked(struct mddev *mddev) | |
486 | { | |
487 | return mutex_is_locked(&mddev->reconfig_mutex); | |
488 | } | |
489 | ||
490 | static inline int mddev_trylock(struct mddev *mddev) | |
491 | { | |
492 | return mutex_trylock(&mddev->reconfig_mutex); | |
493 | } | |
494 | extern void mddev_unlock(struct mddev *mddev); | |
495 | ||
1da177e4 LT |
496 | static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors) |
497 | { | |
f72ffdd6 | 498 | atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io); |
1da177e4 LT |
499 | } |
500 | ||
84fc4b56 | 501 | struct md_personality |
1da177e4 LT |
502 | { |
503 | char *name; | |
2604b703 N |
504 | int level; |
505 | struct list_head list; | |
1da177e4 | 506 | struct module *owner; |
b4fdcb02 | 507 | void (*make_request)(struct mddev *mddev, struct bio *bio); |
fd01b88c | 508 | int (*run)(struct mddev *mddev); |
afa0f557 | 509 | void (*free)(struct mddev *mddev, void *priv); |
fd01b88c | 510 | void (*status)(struct seq_file *seq, struct mddev *mddev); |
1da177e4 | 511 | /* error_handler must set ->faulty and clear ->in_sync |
f72ffdd6 | 512 | * if appropriate, and should abort recovery if needed |
1da177e4 | 513 | */ |
fd01b88c N |
514 | void (*error_handler)(struct mddev *mddev, struct md_rdev *rdev); |
515 | int (*hot_add_disk) (struct mddev *mddev, struct md_rdev *rdev); | |
b8321b68 | 516 | int (*hot_remove_disk) (struct mddev *mddev, struct md_rdev *rdev); |
fd01b88c | 517 | int (*spare_active) (struct mddev *mddev); |
09314799 | 518 | sector_t (*sync_request)(struct mddev *mddev, sector_t sector_nr, int *skipped); |
fd01b88c N |
519 | int (*resize) (struct mddev *mddev, sector_t sectors); |
520 | sector_t (*size) (struct mddev *mddev, sector_t sectors, int raid_disks); | |
521 | int (*check_reshape) (struct mddev *mddev); | |
522 | int (*start_reshape) (struct mddev *mddev); | |
523 | void (*finish_reshape) (struct mddev *mddev); | |
36fa3063 N |
524 | /* quiesce moves between quiescence states |
525 | * 0 - fully active | |
526 | * 1 - no new requests allowed | |
527 | * others - reserved | |
528 | */ | |
fd01b88c | 529 | void (*quiesce) (struct mddev *mddev, int state); |
245f46c2 N |
530 | /* takeover is used to transition an array from one |
531 | * personality to another. The new personality must be able | |
532 | * to handle the data in the current layout. | |
533 | * e.g. 2drive raid1 -> 2drive raid5 | |
534 | * ndrive raid5 -> degraded n+1drive raid6 with special layout | |
535 | * If the takeover succeeds, a new 'private' structure is returned. | |
536 | * This needs to be installed and then ->run used to activate the | |
537 | * array. | |
538 | */ | |
fd01b88c | 539 | void *(*takeover) (struct mddev *mddev); |
5c675f83 N |
540 | /* congested implements bdi.congested_fn(). |
541 | * Will not be called while array is 'suspended' */ | |
542 | int (*congested)(struct mddev *mddev, int bits); | |
1da177e4 LT |
543 | }; |
544 | ||
007583c9 N |
545 | struct md_sysfs_entry { |
546 | struct attribute attr; | |
fd01b88c N |
547 | ssize_t (*show)(struct mddev *, char *); |
548 | ssize_t (*store)(struct mddev *, const char *, size_t); | |
007583c9 | 549 | }; |
43a70507 | 550 | extern struct attribute_group md_bitmap_group; |
007583c9 | 551 | |
324a56e1 | 552 | static inline struct kernfs_node *sysfs_get_dirent_safe(struct kernfs_node *sd, char *name) |
00bcb4ac N |
553 | { |
554 | if (sd) | |
388975cc | 555 | return sysfs_get_dirent(sd, name); |
00bcb4ac N |
556 | return sd; |
557 | } | |
324a56e1 | 558 | static inline void sysfs_notify_dirent_safe(struct kernfs_node *sd) |
00bcb4ac N |
559 | { |
560 | if (sd) | |
561 | sysfs_notify_dirent(sd); | |
562 | } | |
563 | ||
fd01b88c | 564 | static inline char * mdname (struct mddev * mddev) |
1da177e4 LT |
565 | { |
566 | return mddev->gendisk ? mddev->gendisk->disk_name : "mdX"; | |
567 | } | |
568 | ||
fd01b88c | 569 | static inline int sysfs_link_rdev(struct mddev *mddev, struct md_rdev *rdev) |
36fad858 NK |
570 | { |
571 | char nm[20]; | |
9b15603d SL |
572 | if (!test_bit(Replacement, &rdev->flags) && |
573 | !test_bit(Journal, &rdev->flags) && | |
574 | mddev->kobj.sd) { | |
2d78f8c4 N |
575 | sprintf(nm, "rd%d", rdev->raid_disk); |
576 | return sysfs_create_link(&mddev->kobj, &rdev->kobj, nm); | |
577 | } else | |
578 | return 0; | |
36fad858 NK |
579 | } |
580 | ||
fd01b88c | 581 | static inline void sysfs_unlink_rdev(struct mddev *mddev, struct md_rdev *rdev) |
36fad858 NK |
582 | { |
583 | char nm[20]; | |
9b15603d SL |
584 | if (!test_bit(Replacement, &rdev->flags) && |
585 | !test_bit(Journal, &rdev->flags) && | |
586 | mddev->kobj.sd) { | |
2d78f8c4 N |
587 | sprintf(nm, "rd%d", rdev->raid_disk); |
588 | sysfs_remove_link(&mddev->kobj, nm); | |
589 | } | |
36fad858 NK |
590 | } |
591 | ||
1da177e4 LT |
592 | /* |
593 | * iterates through some rdev ringlist. It's safe to remove the | |
594 | * current 'rdev'. Dont touch 'tmp' though. | |
595 | */ | |
159ec1fc CR |
596 | #define rdev_for_each_list(rdev, tmp, head) \ |
597 | list_for_each_entry_safe(rdev, tmp, head, same_set) | |
598 | ||
1da177e4 LT |
599 | /* |
600 | * iterates through the 'same array disks' ringlist | |
601 | */ | |
dafb20fa N |
602 | #define rdev_for_each(rdev, mddev) \ |
603 | list_for_each_entry(rdev, &((mddev)->disks), same_set) | |
604 | ||
605 | #define rdev_for_each_safe(rdev, tmp, mddev) \ | |
159ec1fc | 606 | list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set) |
1da177e4 | 607 | |
4b80991c N |
608 | #define rdev_for_each_rcu(rdev, mddev) \ |
609 | list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set) | |
610 | ||
2b8bf345 | 611 | struct md_thread { |
4ed8731d | 612 | void (*run) (struct md_thread *thread); |
fd01b88c | 613 | struct mddev *mddev; |
1da177e4 | 614 | wait_queue_head_t wqueue; |
f72ffdd6 | 615 | unsigned long flags; |
1da177e4 | 616 | struct task_struct *tsk; |
32a7627c | 617 | unsigned long timeout; |
4ed8731d | 618 | void *private; |
2b8bf345 | 619 | }; |
1da177e4 LT |
620 | |
621 | #define THREAD_WAKEUP 0 | |
622 | ||
1345b1d8 N |
623 | static inline void safe_put_page(struct page *p) |
624 | { | |
625 | if (p) put_page(p); | |
626 | } | |
627 | ||
84fc4b56 N |
628 | extern int register_md_personality(struct md_personality *p); |
629 | extern int unregister_md_personality(struct md_personality *p); | |
edb39c9d GR |
630 | extern int register_md_cluster_operations(struct md_cluster_operations *ops, |
631 | struct module *module); | |
632 | extern int unregister_md_cluster_operations(void); | |
633 | extern int md_setup_cluster(struct mddev *mddev, int nodes); | |
634 | extern void md_cluster_stop(struct mddev *mddev); | |
2b8bf345 | 635 | extern struct md_thread *md_register_thread( |
4ed8731d | 636 | void (*run)(struct md_thread *thread), |
2b8bf345 N |
637 | struct mddev *mddev, |
638 | const char *name); | |
639 | extern void md_unregister_thread(struct md_thread **threadp); | |
640 | extern void md_wakeup_thread(struct md_thread *thread); | |
fd01b88c | 641 | extern void md_check_recovery(struct mddev *mddev); |
a91d5ac0 | 642 | extern void md_reap_sync_thread(struct mddev *mddev); |
fd01b88c N |
643 | extern void md_write_start(struct mddev *mddev, struct bio *bi); |
644 | extern void md_write_end(struct mddev *mddev); | |
645 | extern void md_done_sync(struct mddev *mddev, int blocks, int ok); | |
646 | extern void md_error(struct mddev *mddev, struct md_rdev *rdev); | |
c6563a8c | 647 | extern void md_finish_reshape(struct mddev *mddev); |
fd01b88c N |
648 | |
649 | extern int mddev_congested(struct mddev *mddev, int bits); | |
650 | extern void md_flush_request(struct mddev *mddev, struct bio *bio); | |
651 | extern void md_super_write(struct mddev *mddev, struct md_rdev *rdev, | |
92022950 | 652 | sector_t sector, int size, struct page *page); |
46533ff7 | 653 | extern int md_super_wait(struct mddev *mddev); |
f72ffdd6 | 654 | extern int sync_page_io(struct md_rdev *rdev, sector_t sector, int size, |
796a5cf0 MC |
655 | struct page *page, int op, int op_flags, |
656 | bool metadata_op); | |
4ed8731d | 657 | extern void md_do_sync(struct md_thread *thread); |
fd01b88c N |
658 | extern void md_new_event(struct mddev *mddev); |
659 | extern int md_allow_write(struct mddev *mddev); | |
660 | extern void md_wait_for_blocked_rdev(struct md_rdev *rdev, struct mddev *mddev); | |
661 | extern void md_set_array_sectors(struct mddev *mddev, sector_t array_sectors); | |
662 | extern int md_check_no_bitmap(struct mddev *mddev); | |
663 | extern int md_integrity_register(struct mddev *mddev); | |
1501efad | 664 | extern int md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev); |
72e02075 | 665 | extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale); |
63fe0817 | 666 | |
fd01b88c N |
667 | extern void mddev_init(struct mddev *mddev); |
668 | extern int md_run(struct mddev *mddev); | |
669 | extern void md_stop(struct mddev *mddev); | |
670 | extern void md_stop_writes(struct mddev *mddev); | |
3cb03002 | 671 | extern int md_rdev_init(struct md_rdev *rdev); |
545c8795 | 672 | extern void md_rdev_clear(struct md_rdev *rdev); |
63fe0817 | 673 | |
fd01b88c N |
674 | extern void mddev_suspend(struct mddev *mddev); |
675 | extern void mddev_resume(struct mddev *mddev); | |
a167f663 | 676 | extern struct bio *bio_alloc_mddev(gfp_t gfp_mask, int nr_iovecs, |
fd01b88c | 677 | struct mddev *mddev); |
9cbb1750 | 678 | |
70bcecdb | 679 | extern void md_reload_sb(struct mddev *mddev, int raid_disk); |
1aee41f6 | 680 | extern void md_update_sb(struct mddev *mddev, int force); |
fb56dfef | 681 | extern void md_kick_rdev_from_array(struct md_rdev * rdev); |
57d051dc | 682 | struct md_rdev *md_find_rdev_nr_rcu(struct mddev *mddev, int nr); |
dfe15ac1 HR |
683 | |
684 | static inline void rdev_dec_pending(struct md_rdev *rdev, struct mddev *mddev) | |
685 | { | |
686 | int faulty = test_bit(Faulty, &rdev->flags); | |
687 | if (atomic_dec_and_test(&rdev->nr_pending) && faulty) { | |
688 | set_bit(MD_RECOVERY_NEEDED, &mddev->recovery); | |
689 | md_wakeup_thread(mddev->thread); | |
690 | } | |
691 | } | |
692 | ||
edb39c9d | 693 | extern struct md_cluster_operations *md_cluster_ops; |
c4ce867f GR |
694 | static inline int mddev_is_clustered(struct mddev *mddev) |
695 | { | |
696 | return mddev->cluster_info && mddev->bitmap_info.nodes > 1; | |
697 | } | |
394ed8e4 SL |
698 | |
699 | /* clear unsupported mddev_flags */ | |
700 | static inline void mddev_clear_unsupported_flags(struct mddev *mddev, | |
701 | unsigned long unsupported_flags) | |
702 | { | |
703 | mddev->flags &= ~unsupported_flags; | |
704 | } | |
26483819 SL |
705 | |
706 | static inline void mddev_check_writesame(struct mddev *mddev, struct bio *bio) | |
707 | { | |
708 | if (bio_op(bio) == REQ_OP_WRITE_SAME && | |
709 | !bdev_get_queue(bio->bi_bdev)->limits.max_write_same_sectors) | |
710 | mddev->queue->limits.max_write_same_sectors = 0; | |
711 | } | |
3deff1a7 CH |
712 | |
713 | static inline void mddev_check_write_zeroes(struct mddev *mddev, struct bio *bio) | |
714 | { | |
715 | if (bio_op(bio) == REQ_OP_WRITE_ZEROES && | |
716 | !bdev_get_queue(bio->bi_bdev)->limits.max_write_zeroes_sectors) | |
717 | mddev->queue->limits.max_write_zeroes_sectors = 0; | |
718 | } | |
63fe0817 | 719 | #endif /* _MD_MD_H */ |