Commit | Line | Data |
---|---|---|
af1a8899 | 1 | /* SPDX-License-Identifier: GPL-2.0-or-later */ |
1da177e4 | 2 | /* |
7e841526 | 3 | md.h : kernel internal structure of the Linux MD driver |
1da177e4 | 4 | Copyright (C) 1996-98 Ingo Molnar, Gadi Oxman |
f72ffdd6 | 5 | |
1da177e4 LT |
6 | */ |
7 | ||
63fe0817 CH |
8 | #ifndef _MD_MD_H |
9 | #define _MD_MD_H | |
10 | ||
11 | #include <linux/blkdev.h> | |
66114cad | 12 | #include <linux/backing-dev.h> |
fc974ee2 | 13 | #include <linux/badblocks.h> |
63fe0817 CH |
14 | #include <linux/kobject.h> |
15 | #include <linux/list.h> | |
16 | #include <linux/mm.h> | |
17 | #include <linux/mutex.h> | |
18 | #include <linux/timer.h> | |
19 | #include <linux/wait.h> | |
20 | #include <linux/workqueue.h> | |
edb39c9d | 21 | #include "md-cluster.h" |
9361401e | 22 | |
1da177e4 | 23 | #define MaxSector (~(sector_t)0) |
1da177e4 | 24 | |
46533ff7 N |
25 | /* |
26 | * These flags should really be called "NO_RETRY" rather than | |
27 | * "FAILFAST" because they don't make any promise about time lapse, | |
28 | * only about the number of retries, which will be zero. | |
29 | * REQ_FAILFAST_DRIVER is not included because | |
30 | * Commit: 4a27446f3e39 ("[SCSI] modify scsi to handle new fail fast flags.") | |
31 | * seems to suggest that the errors it avoids retrying should usually | |
32 | * be retried. | |
33 | */ | |
34 | #define MD_FAILFAST (REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT) | |
69b00b5b GJ |
35 | |
36 | /* | |
37 | * The struct embedded in rdev is used to serialize IO. | |
38 | */ | |
39 | struct serial_in_rdev { | |
40 | struct rb_root_cached serial_rb; | |
41 | spinlock_t serial_lock; | |
42 | wait_queue_head_t serial_io_wait; | |
43 | }; | |
44 | ||
1da177e4 LT |
45 | /* |
46 | * MD's 'extended' device | |
47 | */ | |
3cb03002 | 48 | struct md_rdev { |
1da177e4 LT |
49 | struct list_head same_set; /* RAID devices within the same set */ |
50 | ||
dd8ac336 | 51 | sector_t sectors; /* Device size (in 512bytes sectors) */ |
fd01b88c | 52 | struct mddev *mddev; /* RAID array if running */ |
eea1bf38 | 53 | int last_events; /* IO event timestamp */ |
1da177e4 | 54 | |
a6ff7e08 JB |
55 | /* |
56 | * If meta_bdev is non-NULL, it means that a separate device is | |
57 | * being used to store the metadata (superblock/bitmap) which | |
58 | * would otherwise be contained on the same device as the data (bdev). | |
59 | */ | |
60 | struct block_device *meta_bdev; | |
1da177e4 LT |
61 | struct block_device *bdev; /* block device handle */ |
62 | ||
2699b672 | 63 | struct page *sb_page, *bb_page; |
1da177e4 | 64 | int sb_loaded; |
42543769 | 65 | __u64 sb_events; |
1da177e4 | 66 | sector_t data_offset; /* start of data in array */ |
c6563a8c | 67 | sector_t new_data_offset;/* only relevant while reshaping */ |
f72ffdd6 | 68 | sector_t sb_start; /* offset of the super block (in 512byte sectors) */ |
0002b271 | 69 | int sb_size; /* bytes in the superblock */ |
1da177e4 LT |
70 | int preferred_minor; /* autorun support */ |
71 | ||
86e6ffdd N |
72 | struct kobject kobj; |
73 | ||
1da177e4 LT |
74 | /* A device can be in one of three states based on two flags: |
75 | * Not working: faulty==1 in_sync==0 | |
76 | * Fully working: faulty==0 in_sync==1 | |
77 | * Working, but not | |
78 | * in sync with array | |
79 | * faulty==0 in_sync==0 | |
80 | * | |
81 | * It can never have faulty==1, in_sync==1 | |
82 | * This reduces the burden of testing multiple flags in many cases | |
83 | */ | |
1da177e4 | 84 | |
2d78f8c4 | 85 | unsigned long flags; /* bit set of 'enum flag_bits' bits. */ |
6bfe0b49 | 86 | wait_queue_head_t blocked_wait; |
8ddf9efe | 87 | |
1da177e4 LT |
88 | int desc_nr; /* descriptor index in the superblock */ |
89 | int raid_disk; /* role of device in array */ | |
e93f68a1 N |
90 | int new_raid_disk; /* role that the device will have in |
91 | * the array after a level-change completes. | |
92 | */ | |
41158c7e N |
93 | int saved_raid_disk; /* role that device used to have in the |
94 | * array and could again if we did a partial | |
95 | * resync from the bitmap | |
96 | */ | |
3069aa8d SL |
97 | union { |
98 | sector_t recovery_offset;/* If this device has been partially | |
5fd6c1dc N |
99 | * recovered, this is where we were |
100 | * up to. | |
101 | */ | |
3069aa8d SL |
102 | sector_t journal_tail; /* If this device is a journal device, |
103 | * this is the journal tail (journal | |
104 | * recovery start point) | |
105 | */ | |
106 | }; | |
1da177e4 LT |
107 | |
108 | atomic_t nr_pending; /* number of pending requests. | |
109 | * only maintained for arrays that | |
110 | * support hot removal | |
111 | */ | |
ba22dcbf N |
112 | atomic_t read_errors; /* number of consecutive read errors that |
113 | * we have tried to ignore. | |
114 | */ | |
0e3ef49e | 115 | time64_t last_read_error; /* monotonic time since our |
1e50915f RB |
116 | * last read error |
117 | */ | |
4dbcdc75 N |
118 | atomic_t corrected_errors; /* number of corrected read errors, |
119 | * for reporting to userspace and storing | |
120 | * in superblock. | |
121 | */ | |
3e148a32 | 122 | |
69b00b5b | 123 | struct serial_in_rdev *serial; /* used for raid1 io serialization */ |
3e148a32 | 124 | |
324a56e1 | 125 | struct kernfs_node *sysfs_state; /* handle for 'state' |
3c0ee63a | 126 | * sysfs entry */ |
e1a86dbb JB |
127 | /* handle for 'unacknowledged_bad_blocks' sysfs dentry */ |
128 | struct kernfs_node *sysfs_unack_badblocks; | |
129 | /* handle for 'bad_blocks' sysfs dentry */ | |
130 | struct kernfs_node *sysfs_badblocks; | |
fc974ee2 | 131 | struct badblocks badblocks; |
ea0213e0 AP |
132 | |
133 | struct { | |
134 | short offset; /* Offset from superblock to start of PPL. | |
135 | * Not used by external metadata. */ | |
136 | unsigned int size; /* Size in sectors of the PPL space */ | |
137 | sector_t sector; /* First sector of the PPL space */ | |
138 | } ppl; | |
1da177e4 | 139 | }; |
2d78f8c4 N |
140 | enum flag_bits { |
141 | Faulty, /* device is known to have a fault */ | |
142 | In_sync, /* device is in_sync with rest of array */ | |
8313b8e5 | 143 | Bitmap_sync, /* ..actually, not quite In_sync. Need a |
4aaf7694 GJ |
144 | * bitmap-based recovery to get fully in sync. |
145 | * The bit is only meaningful before device | |
146 | * has been passed to pers->hot_add_disk. | |
8313b8e5 | 147 | */ |
2d78f8c4 N |
148 | WriteMostly, /* Avoid reading if at all possible */ |
149 | AutoDetected, /* added by auto-detect */ | |
150 | Blocked, /* An error occurred but has not yet | |
151 | * been acknowledged by the metadata | |
152 | * handler, so don't allow writes | |
153 | * until it is cleared */ | |
154 | WriteErrorSeen, /* A write error has been seen on this | |
155 | * device | |
156 | */ | |
157 | FaultRecorded, /* Intermediate state for clearing | |
158 | * Blocked. The Fault is/will-be | |
159 | * recorded in the metadata, but that | |
160 | * metadata hasn't been stored safely | |
161 | * on disk yet. | |
162 | */ | |
163 | BlockedBadBlocks, /* A writer is blocked because they | |
164 | * found an unacknowledged bad-block. | |
165 | * This can safely be cleared at any | |
166 | * time, and the writer will re-check. | |
167 | * It may be set at any time, and at | |
168 | * worst the writer will timeout and | |
169 | * re-check. So setting it as | |
170 | * accurately as possible is good, but | |
171 | * not absolutely critical. | |
172 | */ | |
173 | WantReplacement, /* This device is a candidate to be | |
174 | * hot-replaced, either because it has | |
175 | * reported some faults, or because | |
176 | * of explicit request. | |
177 | */ | |
178 | Replacement, /* This device is a replacement for | |
179 | * a want_replacement device with same | |
180 | * raid_disk number. | |
181 | */ | |
1aee41f6 GR |
182 | Candidate, /* For clustered environments only: |
183 | * This device is seen locally but not | |
184 | * by the whole cluster | |
185 | */ | |
bac624f3 SL |
186 | Journal, /* This device is used as journal for |
187 | * raid-5/6. | |
188 | * Usually, this device should be faster | |
189 | * than other devices in the array | |
190 | */ | |
659b254f | 191 | ClusterRemove, |
d787be40 N |
192 | RemoveSynchronized, /* synchronize_rcu() was called after |
193 | * this device was known to be faulty, | |
194 | * so it is safe to remove without | |
195 | * another synchronize_rcu() call. | |
196 | */ | |
35b785f7 TM |
197 | ExternalBbl, /* External metadata provides bad |
198 | * block management for a disk | |
199 | */ | |
688834e6 N |
200 | FailFast, /* Minimal retries should be attempted on |
201 | * this device, so use REQ_FAILFAST_DEV. | |
202 | * Also don't try to repair failed reads. | |
203 | * It is expects that no bad block log | |
204 | * is present. | |
205 | */ | |
46533ff7 N |
206 | LastDev, /* Seems to be the last working dev as |
207 | * it didn't fail, so don't use FailFast | |
208 | * any more for metadata | |
209 | */ | |
404659cf GJ |
210 | CollisionCheck, /* |
211 | * check if there is collision between raid1 | |
212 | * serial bios. | |
3e148a32 | 213 | */ |
99892147 YK |
214 | Holder, /* rdev is used as holder while opening |
215 | * underlying disk exclusively. | |
216 | */ | |
2d78f8c4 | 217 | }; |
1da177e4 | 218 | |
3cb03002 | 219 | static inline int is_badblock(struct md_rdev *rdev, sector_t s, int sectors, |
2230dfe4 N |
220 | sector_t *first_bad, int *bad_sectors) |
221 | { | |
222 | if (unlikely(rdev->badblocks.count)) { | |
fc974ee2 | 223 | int rv = badblocks_check(&rdev->badblocks, rdev->data_offset + s, |
2230dfe4 N |
224 | sectors, |
225 | first_bad, bad_sectors); | |
226 | if (rv) | |
227 | *first_bad -= rdev->data_offset; | |
228 | return rv; | |
229 | } | |
230 | return 0; | |
231 | } | |
3cb03002 | 232 | extern int rdev_set_badblocks(struct md_rdev *rdev, sector_t s, int sectors, |
c6563a8c N |
233 | int is_new); |
234 | extern int rdev_clear_badblocks(struct md_rdev *rdev, sector_t s, int sectors, | |
235 | int is_new); | |
c4ce867f GR |
236 | struct md_cluster_info; |
237 | ||
9631abdb MT |
238 | /** |
239 | * enum mddev_flags - md device flags. | |
240 | * @MD_ARRAY_FIRST_USE: First use of array, needs initialization. | |
241 | * @MD_CLOSING: If set, we are closing the array, do not open it then. | |
242 | * @MD_JOURNAL_CLEAN: A raid with journal is already clean. | |
243 | * @MD_HAS_JOURNAL: The raid array has journal feature set. | |
244 | * @MD_CLUSTER_RESYNC_LOCKED: cluster raid only, which means node, already took | |
245 | * resync lock, need to release the lock. | |
246 | * @MD_FAILFAST_SUPPORTED: Using MD_FAILFAST on metadata writes is supported as | |
247 | * calls to md_error() will never cause the array to | |
248 | * become failed. | |
249 | * @MD_HAS_PPL: The raid array has PPL feature set. | |
250 | * @MD_HAS_MULTIPLE_PPLS: The raid array has multiple PPLs feature set. | |
251 | * @MD_ALLOW_SB_UPDATE: md_check_recovery is allowed to update the metadata | |
252 | * without taking reconfig_mutex. | |
253 | * @MD_UPDATING_SB: md_check_recovery is updating the metadata without | |
254 | * explicitly holding reconfig_mutex. | |
255 | * @MD_NOT_READY: do_md_run() is active, so 'array_state', ust not report that | |
256 | * array is ready yet. | |
257 | * @MD_BROKEN: This is used to stop writes and mark array as failed. | |
12a6caf2 | 258 | * @MD_DELETED: This device is being deleted |
9631abdb MT |
259 | * |
260 | * change UNSUPPORTED_MDDEV_FLAGS for each array type if new flag is added | |
261 | */ | |
be306c29 | 262 | enum mddev_flags { |
9631abdb MT |
263 | MD_ARRAY_FIRST_USE, |
264 | MD_CLOSING, | |
265 | MD_JOURNAL_CLEAN, | |
266 | MD_HAS_JOURNAL, | |
267 | MD_CLUSTER_RESYNC_LOCKED, | |
268 | MD_FAILFAST_SUPPORTED, | |
269 | MD_HAS_PPL, | |
270 | MD_HAS_MULTIPLE_PPLS, | |
271 | MD_ALLOW_SB_UPDATE, | |
272 | MD_UPDATING_SB, | |
273 | MD_NOT_READY, | |
274 | MD_BROKEN, | |
12a6caf2 | 275 | MD_DELETED, |
be306c29 | 276 | }; |
2953079c SL |
277 | |
278 | enum mddev_sb_flags { | |
279 | MD_SB_CHANGE_DEVS, /* Some device status has changed */ | |
280 | MD_SB_CHANGE_CLEAN, /* transition to or from 'clean' */ | |
281 | MD_SB_CHANGE_PENDING, /* switch from 'clean' to 'active' in progress */ | |
282 | MD_SB_NEED_REWRITE, /* metadata write needs to be repeated */ | |
283 | }; | |
284 | ||
404659cf GJ |
285 | #define NR_SERIAL_INFOS 8 |
286 | /* record current range of serialize IOs */ | |
287 | struct serial_info { | |
69b00b5b GJ |
288 | struct rb_node node; |
289 | sector_t start; /* start sector of rb node */ | |
290 | sector_t last; /* end sector of rb node */ | |
291 | sector_t _subtree_last; /* highest sector in subtree of rb node */ | |
3e148a32 GJ |
292 | }; |
293 | ||
eac58d08 LG |
294 | /* |
295 | * mddev->curr_resync stores the current sector of the resync but | |
296 | * also has some overloaded values. | |
297 | */ | |
298 | enum { | |
299 | /* No resync in progress */ | |
300 | MD_RESYNC_NONE = 0, | |
301 | /* Yielded to allow another conflicting resync to commence */ | |
302 | MD_RESYNC_YIELDED = 1, | |
303 | /* Delayed to check that there is no conflict with another sync */ | |
304 | MD_RESYNC_DELAYED = 2, | |
305 | /* Any value greater than or equal to this is in an active resync */ | |
306 | MD_RESYNC_ACTIVE = 3, | |
307 | }; | |
308 | ||
fd01b88c | 309 | struct mddev { |
1da177e4 | 310 | void *private; |
84fc4b56 | 311 | struct md_personality *pers; |
1da177e4 LT |
312 | dev_t unit; |
313 | int md_minor; | |
7a0a5355 | 314 | struct list_head disks; |
850b2b42 | 315 | unsigned long flags; |
2953079c | 316 | unsigned long sb_flags; |
850b2b42 | 317 | |
409c57f3 | 318 | int suspended; |
72adae23 | 319 | struct percpu_ref active_io; |
1da177e4 | 320 | int ro; |
bb4f1e9d N |
321 | int sysfs_active; /* set when sysfs deletes |
322 | * are happening, so run/ | |
323 | * takeover/stop are not safe | |
324 | */ | |
1da177e4 LT |
325 | struct gendisk *gendisk; |
326 | ||
eae1701f | 327 | struct kobject kobj; |
d3374825 N |
328 | int hold_active; |
329 | #define UNTIL_IOCTL 1 | |
efeb53c0 | 330 | #define UNTIL_STOP 2 |
eae1701f | 331 | |
1da177e4 LT |
332 | /* Superblock information */ |
333 | int major_version, | |
334 | minor_version, | |
335 | patch_version; | |
336 | int persistent; | |
f72ffdd6 | 337 | int external; /* metadata is |
e691063a N |
338 | * managed externally */ |
339 | char metadata_type[17]; /* externally set*/ | |
77a68698 | 340 | int chunk_sectors; |
9ebc6ef1 | 341 | time64_t ctime, utime; |
1da177e4 | 342 | int level, layout; |
d9d166c2 | 343 | char clevel[16]; |
1da177e4 LT |
344 | int raid_disks; |
345 | int max_disks; | |
f72ffdd6 | 346 | sector_t dev_sectors; /* used size of |
58c0fed4 | 347 | * component devices */ |
f233ea5c | 348 | sector_t array_sectors; /* exported array size */ |
b522adcd DW |
349 | int external_size; /* size managed |
350 | * externally */ | |
1da177e4 | 351 | __u64 events; |
a8707c08 N |
352 | /* If the last 'event' was simply a clean->dirty transition, and |
353 | * we didn't write it to the spares, then it is safe and simple | |
354 | * to just decrement the event count on a dirty->clean transition. | |
355 | * So we record that possibility here. | |
356 | */ | |
357 | int can_decrease_events; | |
1da177e4 LT |
358 | |
359 | char uuid[16]; | |
360 | ||
f6705578 N |
361 | /* If the array is being reshaped, we need to record the |
362 | * new shape and an indication of where we are up to. | |
363 | * This is written to the superblock. | |
364 | * If reshape_position is MaxSector, then no reshape is happening (yet). | |
365 | */ | |
366 | sector_t reshape_position; | |
664e7c41 | 367 | int delta_disks, new_level, new_layout; |
77a68698 | 368 | int new_chunk_sectors; |
2c810cdd | 369 | int reshape_backwards; |
f6705578 | 370 | |
44693154 YK |
371 | struct md_thread __rcu *thread; /* management thread */ |
372 | struct md_thread __rcu *sync_thread; /* doing resync or reconstruct */ | |
c4a39551 JB |
373 | |
374 | /* 'last_sync_action' is initialized to "none". It is set when a | |
375 | * sync operation (i.e "data-check", "requested-resync", "resync", | |
376 | * "recovery", or "reshape") is started. It holds this value even | |
377 | * when the sync thread is "frozen" (interrupted) or "idle" (stopped | |
378 | * or finished). It is overwritten when a new sync operation is begun. | |
379 | */ | |
380 | char *last_sync_action; | |
ff4e8d9a | 381 | sector_t curr_resync; /* last block scheduled */ |
97e4f42d N |
382 | /* As resync requests can complete out of order, we cannot easily track |
383 | * how much resync has been completed. So we occasionally pause until | |
384 | * everything completes, then set curr_resync_completed to curr_resync. | |
385 | * As such it may be well behind the real resync mark, but it is a value | |
386 | * we are certain of. | |
387 | */ | |
388 | sector_t curr_resync_completed; | |
1da177e4 LT |
389 | unsigned long resync_mark; /* a recent timestamp */ |
390 | sector_t resync_mark_cnt;/* blocks written at resync_mark */ | |
ff4e8d9a | 391 | sector_t curr_mark_cnt; /* blocks scheduled now */ |
1da177e4 LT |
392 | |
393 | sector_t resync_max_sectors; /* may be set by personality */ | |
9d88883e | 394 | |
7f7583d4 | 395 | atomic64_t resync_mismatches; /* count of sectors where |
9d88883e N |
396 | * parity/replica mismatch found |
397 | */ | |
e464eafd N |
398 | |
399 | /* allow user-space to request suspension of IO to regions of the array */ | |
400 | sector_t suspend_lo; | |
401 | sector_t suspend_hi; | |
88202a0c N |
402 | /* if zero, use the system-wide default */ |
403 | int sync_speed_min; | |
404 | int sync_speed_max; | |
405 | ||
90b08710 BS |
406 | /* resync even though the same disks are shared among md-devices */ |
407 | int parallel_resync; | |
408 | ||
6ff8d8ec | 409 | int ok_start_degraded; |
5fd6c1dc | 410 | |
1da177e4 | 411 | unsigned long recovery; |
5389042f N |
412 | /* If a RAID personality determines that recovery (of a particular |
413 | * device) will fail due to a read error on the source device, it | |
414 | * takes a copy of this number and does not attempt recovery again | |
415 | * until this number changes. | |
416 | */ | |
417 | int recovery_disabled; | |
1da177e4 LT |
418 | |
419 | int in_sync; /* know to not need resync */ | |
c8c00a69 N |
420 | /* 'open_mutex' avoids races between 'md_open' and 'do_md_stop', so |
421 | * that we are never stopping an array while it is open. | |
422 | * 'reconfig_mutex' protects all other reconfiguration. | |
423 | * These locks are separate due to conflicting interactions | |
a8698707 | 424 | * with disk->open_mutex. |
c8c00a69 | 425 | * Lock ordering is: |
a8698707 CH |
426 | * reconfig_mutex -> disk->open_mutex |
427 | * disk->open_mutex -> open_mutex: e.g. __blkdev_get -> md_open | |
c8c00a69 N |
428 | */ |
429 | struct mutex open_mutex; | |
df5b89b3 | 430 | struct mutex reconfig_mutex; |
f2ea68cf N |
431 | atomic_t active; /* general refcount */ |
432 | atomic_t openers; /* number of active opens */ | |
1da177e4 | 433 | |
f0b4f7e2 N |
434 | int changed; /* True if we might need to |
435 | * reread partition info */ | |
1da177e4 LT |
436 | int degraded; /* whether md should consider |
437 | * adding a spare | |
438 | */ | |
439 | ||
440 | atomic_t recovery_active; /* blocks scheduled, but not written */ | |
441 | wait_queue_head_t recovery_wait; | |
442 | sector_t recovery_cp; | |
5e96ee65 NB |
443 | sector_t resync_min; /* user requested sync |
444 | * starts here */ | |
c6207277 N |
445 | sector_t resync_max; /* resync should pause |
446 | * when it gets here */ | |
06d91a5f | 447 | |
324a56e1 | 448 | struct kernfs_node *sysfs_state; /* handle for 'array_state' |
b62b7590 N |
449 | * file in sysfs. |
450 | */ | |
324a56e1 | 451 | struct kernfs_node *sysfs_action; /* handle for 'sync_action' */ |
e1a86dbb JB |
452 | struct kernfs_node *sysfs_completed; /*handle for 'sync_completed' */ |
453 | struct kernfs_node *sysfs_degraded; /*handle for 'degraded' */ | |
454 | struct kernfs_node *sysfs_level; /*handle for 'level' */ | |
b62b7590 | 455 | |
d3374825 N |
456 | struct work_struct del_work; /* used for delayed sysfs removal */ |
457 | ||
85572d7c N |
458 | /* "lock" protects: |
459 | * flush_bio transition from NULL to !NULL | |
460 | * rdev superblocks, events | |
461 | * clearing MD_CHANGE_* | |
462 | * in_sync - and related safemode and MD_CHANGE changes | |
36d091f4 | 463 | * pers (also protected by reconfig_mutex and pending IO). |
978a7a47 | 464 | * clearing ->bitmap |
4af1a041 | 465 | * clearing ->bitmap_info.file |
23da422b N |
466 | * changing ->resync_{min,max} |
467 | * setting MD_RECOVERY_RUNNING (which interacts with resync_{min,max}) | |
85572d7c N |
468 | */ |
469 | spinlock_t lock; | |
3d310eb7 | 470 | wait_queue_head_t sb_wait; /* for waiting on superblock updates */ |
7bfa19f2 | 471 | atomic_t pending_writes; /* number of active superblock writes */ |
06d91a5f | 472 | |
1da177e4 LT |
473 | unsigned int safemode; /* if set, update "clean" superblock |
474 | * when no writes pending. | |
f72ffdd6 | 475 | */ |
1da177e4 LT |
476 | unsigned int safemode_delay; |
477 | struct timer_list safemode_timer; | |
4ad23a97 N |
478 | struct percpu_ref writes_pending; |
479 | int sync_checkers; /* # of threads checking writes_pending */ | |
165125e1 | 480 | struct request_queue *queue; /* for plugging ... */ |
1da177e4 | 481 | |
f72ffdd6 | 482 | struct bitmap *bitmap; /* the bitmap for the device */ |
c3d9714e N |
483 | struct { |
484 | struct file *file; /* the bitmap file */ | |
f6af949c | 485 | loff_t offset; /* offset from superblock of |
c3d9714e N |
486 | * start of bitmap. May be |
487 | * negative, but not '0' | |
f6af949c | 488 | * For external metadata, offset |
f72ffdd6 | 489 | * from start of device. |
c3d9714e | 490 | */ |
6409bb05 | 491 | unsigned long space; /* space available at this offset */ |
f6af949c | 492 | loff_t default_offset; /* this is the offset to use when |
c3d9714e N |
493 | * hot-adding a bitmap. It should |
494 | * eventually be settable by sysfs. | |
495 | */ | |
6409bb05 N |
496 | unsigned long default_space; /* space available at |
497 | * default offset */ | |
c3d9714e | 498 | struct mutex mutex; |
42a04b50 | 499 | unsigned long chunksize; |
ac2f40be | 500 | unsigned long daemon_sleep; /* how many jiffies between updates? */ |
42a04b50 | 501 | unsigned long max_write_behind; /* write-behind mode */ |
ece5cff0 | 502 | int external; |
c4ce867f | 503 | int nodes; /* Maximum number of nodes in the cluster */ |
cf921cc1 | 504 | char cluster_name[64]; /* Name of the cluster */ |
c3d9714e | 505 | } bitmap_info; |
32a7627c | 506 | |
f72ffdd6 | 507 | atomic_t max_corr_read_errors; /* max read retries */ |
1da177e4 | 508 | struct list_head all_mddevs; |
a2826aa9 | 509 | |
c32dc040 | 510 | const struct attribute_group *to_remove; |
252ac522 | 511 | |
afeee514 KO |
512 | struct bio_set bio_set; |
513 | struct bio_set sync_set; /* for sync operations like | |
5a85071c N |
514 | * metadata and bitmap writes |
515 | */ | |
c687297b | 516 | struct bio_set io_clone_set; |
a167f663 | 517 | |
4bc034d3 N |
518 | /* Generic flush handling. |
519 | * The last to finish preflush schedules a worker to submit | |
520 | * the rest of the request (without the REQ_PREFLUSH flag). | |
521 | */ | |
522 | struct bio *flush_bio; | |
523 | atomic_t flush_pending; | |
81ba3c24 PG |
524 | ktime_t start_flush, prev_flush_start; /* prev_flush_start is when the previous completed |
525 | * flush was started. | |
526 | */ | |
4bc034d3 | 527 | struct work_struct flush_work; |
768a418d | 528 | struct work_struct event_work; /* used by dm to report failure event */ |
404659cf | 529 | mempool_t *serial_info_pool; |
fd01b88c | 530 | void (*sync_super)(struct mddev *mddev, struct md_rdev *rdev); |
c4ce867f | 531 | struct md_cluster_info *cluster_info; |
15858fa5 | 532 | unsigned int good_device_nr; /* good device num within cluster raid */ |
78f57ef9 | 533 | unsigned int noio_flag; /* for memalloc scope API */ |
4b6c1060 | 534 | |
3ce94ce5 YK |
535 | /* |
536 | * Temporarily store rdev that will be finally removed when | |
4934b640 | 537 | * reconfig_mutex is unlocked, protected by reconfig_mutex. |
3ce94ce5 YK |
538 | */ |
539 | struct list_head deleting; | |
3ce94ce5 | 540 | |
6f56f0c4 YK |
541 | /* Used to synchronize idle and frozen for action_store() */ |
542 | struct mutex sync_mutex; | |
130443d6 YK |
543 | /* The sequence number for sync thread */ |
544 | atomic_t sync_seq; | |
6f56f0c4 | 545 | |
4b6c1060 | 546 | bool has_superblocks:1; |
9a567843 | 547 | bool fail_last_dev:1; |
3938f5fb | 548 | bool serialize_policy:1; |
be306c29 N |
549 | }; |
550 | ||
551 | enum recovery_flags { | |
552 | /* | |
553 | * If neither SYNC or RESHAPE are set, then it is a recovery. | |
554 | */ | |
555 | MD_RECOVERY_RUNNING, /* a thread is running, or about to be started */ | |
556 | MD_RECOVERY_SYNC, /* actually doing a resync, not a recovery */ | |
557 | MD_RECOVERY_RECOVER, /* doing recovery, or need to try it. */ | |
558 | MD_RECOVERY_INTR, /* resync needs to be aborted for some reason */ | |
559 | MD_RECOVERY_DONE, /* thread is done and is waiting to be reaped */ | |
560 | MD_RECOVERY_NEEDED, /* we might need to start a resync/recover */ | |
561 | MD_RECOVERY_REQUESTED, /* user-space has requested a sync (used with SYNC) */ | |
562 | MD_RECOVERY_CHECK, /* user-space request for check-only, no repair */ | |
563 | MD_RECOVERY_RESHAPE, /* A reshape is happening */ | |
564 | MD_RECOVERY_FROZEN, /* User request to abort, and not restart, any action */ | |
565 | MD_RECOVERY_ERROR, /* sync-action interrupted because io-error */ | |
d5d885fd | 566 | MD_RECOVERY_WAIT, /* waiting for pers->start() to finish */ |
0357ba27 | 567 | MD_RESYNCING_REMOTE, /* remote node is running resync thread */ |
1da177e4 LT |
568 | }; |
569 | ||
431e6125 YK |
570 | enum md_ro_state { |
571 | MD_RDWR, | |
572 | MD_RDONLY, | |
573 | MD_AUTO_READ, | |
574 | MD_MAX_STATE | |
575 | }; | |
576 | ||
577 | static inline bool md_is_rdwr(struct mddev *mddev) | |
578 | { | |
579 | return (mddev->ro == MD_RDWR); | |
580 | } | |
581 | ||
582 | static inline bool is_md_suspended(struct mddev *mddev) | |
583 | { | |
584 | return percpu_ref_is_dying(&mddev->active_io); | |
585 | } | |
586 | ||
5c47daf6 N |
587 | static inline int __must_check mddev_lock(struct mddev *mddev) |
588 | { | |
589 | return mutex_lock_interruptible(&mddev->reconfig_mutex); | |
590 | } | |
591 | ||
592 | /* Sometimes we need to take the lock in a situation where | |
593 | * failure due to interrupts is not acceptable. | |
594 | */ | |
595 | static inline void mddev_lock_nointr(struct mddev *mddev) | |
596 | { | |
597 | mutex_lock(&mddev->reconfig_mutex); | |
598 | } | |
599 | ||
5c47daf6 N |
600 | static inline int mddev_trylock(struct mddev *mddev) |
601 | { | |
602 | return mutex_trylock(&mddev->reconfig_mutex); | |
603 | } | |
604 | extern void mddev_unlock(struct mddev *mddev); | |
605 | ||
1da177e4 LT |
606 | static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors) |
607 | { | |
4245e52d | 608 | atomic_add(nr_sectors, &bdev->bd_disk->sync_io); |
1da177e4 LT |
609 | } |
610 | ||
74d46992 CH |
611 | static inline void md_sync_acct_bio(struct bio *bio, unsigned long nr_sectors) |
612 | { | |
309dca30 | 613 | md_sync_acct(bio->bi_bdev, nr_sectors); |
74d46992 CH |
614 | } |
615 | ||
84fc4b56 | 616 | struct md_personality |
1da177e4 LT |
617 | { |
618 | char *name; | |
2604b703 N |
619 | int level; |
620 | struct list_head list; | |
1da177e4 | 621 | struct module *owner; |
775d7831 | 622 | bool __must_check (*make_request)(struct mddev *mddev, struct bio *bio); |
d5d885fd SL |
623 | /* |
624 | * start up works that do NOT require md_thread. tasks that | |
625 | * requires md_thread should go into start() | |
626 | */ | |
fd01b88c | 627 | int (*run)(struct mddev *mddev); |
d5d885fd SL |
628 | /* start up works that require md threads */ |
629 | int (*start)(struct mddev *mddev); | |
afa0f557 | 630 | void (*free)(struct mddev *mddev, void *priv); |
fd01b88c | 631 | void (*status)(struct seq_file *seq, struct mddev *mddev); |
1da177e4 | 632 | /* error_handler must set ->faulty and clear ->in_sync |
f72ffdd6 | 633 | * if appropriate, and should abort recovery if needed |
1da177e4 | 634 | */ |
fd01b88c N |
635 | void (*error_handler)(struct mddev *mddev, struct md_rdev *rdev); |
636 | int (*hot_add_disk) (struct mddev *mddev, struct md_rdev *rdev); | |
b8321b68 | 637 | int (*hot_remove_disk) (struct mddev *mddev, struct md_rdev *rdev); |
fd01b88c | 638 | int (*spare_active) (struct mddev *mddev); |
09314799 | 639 | sector_t (*sync_request)(struct mddev *mddev, sector_t sector_nr, int *skipped); |
fd01b88c N |
640 | int (*resize) (struct mddev *mddev, sector_t sectors); |
641 | sector_t (*size) (struct mddev *mddev, sector_t sectors, int raid_disks); | |
642 | int (*check_reshape) (struct mddev *mddev); | |
643 | int (*start_reshape) (struct mddev *mddev); | |
644 | void (*finish_reshape) (struct mddev *mddev); | |
7564beda | 645 | void (*update_reshape_pos) (struct mddev *mddev); |
3e00777d | 646 | void (*prepare_suspend) (struct mddev *mddev); |
b03e0ccb N |
647 | /* quiesce suspends or resumes internal processing. |
648 | * 1 - stop new actions and wait for action io to complete | |
649 | * 0 - return to normal behaviour | |
36fa3063 | 650 | */ |
b03e0ccb | 651 | void (*quiesce) (struct mddev *mddev, int quiesce); |
245f46c2 N |
652 | /* takeover is used to transition an array from one |
653 | * personality to another. The new personality must be able | |
654 | * to handle the data in the current layout. | |
655 | * e.g. 2drive raid1 -> 2drive raid5 | |
656 | * ndrive raid5 -> degraded n+1drive raid6 with special layout | |
657 | * If the takeover succeeds, a new 'private' structure is returned. | |
658 | * This needs to be installed and then ->run used to activate the | |
659 | * array. | |
660 | */ | |
fd01b88c | 661 | void *(*takeover) (struct mddev *mddev); |
ba903a3e AP |
662 | /* Changes the consistency policy of an active array. */ |
663 | int (*change_consistency_policy)(struct mddev *mddev, const char *buf); | |
1da177e4 LT |
664 | }; |
665 | ||
007583c9 N |
666 | struct md_sysfs_entry { |
667 | struct attribute attr; | |
fd01b88c N |
668 | ssize_t (*show)(struct mddev *, char *); |
669 | ssize_t (*store)(struct mddev *, const char *, size_t); | |
007583c9 | 670 | }; |
c32dc040 | 671 | extern const struct attribute_group md_bitmap_group; |
007583c9 | 672 | |
324a56e1 | 673 | static inline struct kernfs_node *sysfs_get_dirent_safe(struct kernfs_node *sd, char *name) |
00bcb4ac N |
674 | { |
675 | if (sd) | |
388975cc | 676 | return sysfs_get_dirent(sd, name); |
00bcb4ac N |
677 | return sd; |
678 | } | |
324a56e1 | 679 | static inline void sysfs_notify_dirent_safe(struct kernfs_node *sd) |
00bcb4ac N |
680 | { |
681 | if (sd) | |
682 | sysfs_notify_dirent(sd); | |
683 | } | |
684 | ||
fd01b88c | 685 | static inline char * mdname (struct mddev * mddev) |
1da177e4 LT |
686 | { |
687 | return mddev->gendisk ? mddev->gendisk->disk_name : "mdX"; | |
688 | } | |
689 | ||
fd01b88c | 690 | static inline int sysfs_link_rdev(struct mddev *mddev, struct md_rdev *rdev) |
36fad858 NK |
691 | { |
692 | char nm[20]; | |
9b15603d SL |
693 | if (!test_bit(Replacement, &rdev->flags) && |
694 | !test_bit(Journal, &rdev->flags) && | |
695 | mddev->kobj.sd) { | |
2d78f8c4 N |
696 | sprintf(nm, "rd%d", rdev->raid_disk); |
697 | return sysfs_create_link(&mddev->kobj, &rdev->kobj, nm); | |
698 | } else | |
699 | return 0; | |
36fad858 NK |
700 | } |
701 | ||
fd01b88c | 702 | static inline void sysfs_unlink_rdev(struct mddev *mddev, struct md_rdev *rdev) |
36fad858 NK |
703 | { |
704 | char nm[20]; | |
9b15603d SL |
705 | if (!test_bit(Replacement, &rdev->flags) && |
706 | !test_bit(Journal, &rdev->flags) && | |
707 | mddev->kobj.sd) { | |
2d78f8c4 N |
708 | sprintf(nm, "rd%d", rdev->raid_disk); |
709 | sysfs_remove_link(&mddev->kobj, nm); | |
710 | } | |
36fad858 NK |
711 | } |
712 | ||
1da177e4 LT |
713 | /* |
714 | * iterates through some rdev ringlist. It's safe to remove the | |
715 | * current 'rdev'. Dont touch 'tmp' though. | |
716 | */ | |
159ec1fc CR |
717 | #define rdev_for_each_list(rdev, tmp, head) \ |
718 | list_for_each_entry_safe(rdev, tmp, head, same_set) | |
719 | ||
1da177e4 LT |
720 | /* |
721 | * iterates through the 'same array disks' ringlist | |
722 | */ | |
dafb20fa N |
723 | #define rdev_for_each(rdev, mddev) \ |
724 | list_for_each_entry(rdev, &((mddev)->disks), same_set) | |
725 | ||
726 | #define rdev_for_each_safe(rdev, tmp, mddev) \ | |
159ec1fc | 727 | list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set) |
1da177e4 | 728 | |
4b80991c N |
729 | #define rdev_for_each_rcu(rdev, mddev) \ |
730 | list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set) | |
731 | ||
2b8bf345 | 732 | struct md_thread { |
4ed8731d | 733 | void (*run) (struct md_thread *thread); |
fd01b88c | 734 | struct mddev *mddev; |
1da177e4 | 735 | wait_queue_head_t wqueue; |
f72ffdd6 | 736 | unsigned long flags; |
1da177e4 | 737 | struct task_struct *tsk; |
32a7627c | 738 | unsigned long timeout; |
4ed8731d | 739 | void *private; |
2b8bf345 | 740 | }; |
1da177e4 | 741 | |
c687297b | 742 | struct md_io_clone { |
76fed014 XN |
743 | struct mddev *mddev; |
744 | struct bio *orig_bio; | |
745 | unsigned long start_time; | |
746 | struct bio bio_clone; | |
10764815 GJ |
747 | }; |
748 | ||
1da177e4 LT |
749 | #define THREAD_WAKEUP 0 |
750 | ||
1345b1d8 N |
751 | static inline void safe_put_page(struct page *p) |
752 | { | |
753 | if (p) put_page(p); | |
754 | } | |
755 | ||
84fc4b56 N |
756 | extern int register_md_personality(struct md_personality *p); |
757 | extern int unregister_md_personality(struct md_personality *p); | |
edb39c9d GR |
758 | extern int register_md_cluster_operations(struct md_cluster_operations *ops, |
759 | struct module *module); | |
760 | extern int unregister_md_cluster_operations(void); | |
761 | extern int md_setup_cluster(struct mddev *mddev, int nodes); | |
762 | extern void md_cluster_stop(struct mddev *mddev); | |
2b8bf345 | 763 | extern struct md_thread *md_register_thread( |
4ed8731d | 764 | void (*run)(struct md_thread *thread), |
2b8bf345 N |
765 | struct mddev *mddev, |
766 | const char *name); | |
7eb8ff02 | 767 | extern void md_unregister_thread(struct mddev *mddev, struct md_thread __rcu **threadp); |
44693154 | 768 | extern void md_wakeup_thread(struct md_thread __rcu *thread); |
fd01b88c | 769 | extern void md_check_recovery(struct mddev *mddev); |
d0a18034 | 770 | extern void md_reap_sync_thread(struct mddev *mddev); |
a415c0f1 | 771 | extern int mddev_init_writes_pending(struct mddev *mddev); |
cc27b0c7 | 772 | extern bool md_write_start(struct mddev *mddev, struct bio *bi); |
49728050 | 773 | extern void md_write_inc(struct mddev *mddev, struct bio *bi); |
fd01b88c N |
774 | extern void md_write_end(struct mddev *mddev); |
775 | extern void md_done_sync(struct mddev *mddev, int blocks, int ok); | |
776 | extern void md_error(struct mddev *mddev, struct md_rdev *rdev); | |
c6563a8c | 777 | extern void md_finish_reshape(struct mddev *mddev); |
cf78408f XN |
778 | void md_submit_discard_bio(struct mddev *mddev, struct md_rdev *rdev, |
779 | struct bio *bio, sector_t start, sector_t size); | |
10764815 | 780 | void md_account_bio(struct mddev *mddev, struct bio **bio); |
fd01b88c | 781 | |
775d7831 | 782 | extern bool __must_check md_flush_request(struct mddev *mddev, struct bio *bio); |
fd01b88c | 783 | extern void md_super_write(struct mddev *mddev, struct md_rdev *rdev, |
92022950 | 784 | sector_t sector, int size, struct page *page); |
46533ff7 | 785 | extern int md_super_wait(struct mddev *mddev); |
f72ffdd6 | 786 | extern int sync_page_io(struct md_rdev *rdev, sector_t sector, int size, |
4ce4c73f | 787 | struct page *page, blk_opf_t opf, bool metadata_op); |
4ed8731d | 788 | extern void md_do_sync(struct md_thread *thread); |
54679486 | 789 | extern void md_new_event(void); |
2214c260 | 790 | extern void md_allow_write(struct mddev *mddev); |
fd01b88c N |
791 | extern void md_wait_for_blocked_rdev(struct md_rdev *rdev, struct mddev *mddev); |
792 | extern void md_set_array_sectors(struct mddev *mddev, sector_t array_sectors); | |
793 | extern int md_check_no_bitmap(struct mddev *mddev); | |
794 | extern int md_integrity_register(struct mddev *mddev); | |
1501efad | 795 | extern int md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev); |
72e02075 | 796 | extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale); |
63fe0817 | 797 | |
fd01b88c | 798 | extern void mddev_init(struct mddev *mddev); |
34cb92c0 CH |
799 | struct mddev *md_alloc(dev_t dev, char *name); |
800 | void mddev_put(struct mddev *mddev); | |
fd01b88c | 801 | extern int md_run(struct mddev *mddev); |
d5d885fd | 802 | extern int md_start(struct mddev *mddev); |
fd01b88c N |
803 | extern void md_stop(struct mddev *mddev); |
804 | extern void md_stop_writes(struct mddev *mddev); | |
3cb03002 | 805 | extern int md_rdev_init(struct md_rdev *rdev); |
545c8795 | 806 | extern void md_rdev_clear(struct md_rdev *rdev); |
63fe0817 | 807 | |
393debc2 | 808 | extern void md_handle_request(struct mddev *mddev, struct bio *bio); |
fd01b88c N |
809 | extern void mddev_suspend(struct mddev *mddev); |
810 | extern void mddev_resume(struct mddev *mddev); | |
9cbb1750 | 811 | |
70bcecdb | 812 | extern void md_reload_sb(struct mddev *mddev, int raid_disk); |
1aee41f6 | 813 | extern void md_update_sb(struct mddev *mddev, int force); |
404659cf | 814 | extern void mddev_create_serial_pool(struct mddev *mddev, struct md_rdev *rdev, |
11d3a9f6 | 815 | bool is_suspend); |
69b00b5b GJ |
816 | extern void mddev_destroy_serial_pool(struct mddev *mddev, struct md_rdev *rdev, |
817 | bool is_suspend); | |
57d051dc | 818 | struct md_rdev *md_find_rdev_nr_rcu(struct mddev *mddev, int nr); |
1532d9e8 | 819 | struct md_rdev *md_find_rdev_rcu(struct mddev *mddev, dev_t dev); |
dfe15ac1 | 820 | |
c31fea2f | 821 | static inline bool is_rdev_broken(struct md_rdev *rdev) |
62f7b198 | 822 | { |
c31fea2f | 823 | return !disk_live(rdev->bdev->bd_disk); |
62f7b198 GP |
824 | } |
825 | ||
dfe15ac1 HR |
826 | static inline void rdev_dec_pending(struct md_rdev *rdev, struct mddev *mddev) |
827 | { | |
828 | int faulty = test_bit(Faulty, &rdev->flags); | |
829 | if (atomic_dec_and_test(&rdev->nr_pending) && faulty) { | |
830 | set_bit(MD_RECOVERY_NEEDED, &mddev->recovery); | |
831 | md_wakeup_thread(mddev->thread); | |
832 | } | |
833 | } | |
834 | ||
edb39c9d | 835 | extern struct md_cluster_operations *md_cluster_ops; |
c4ce867f GR |
836 | static inline int mddev_is_clustered(struct mddev *mddev) |
837 | { | |
838 | return mddev->cluster_info && mddev->bitmap_info.nodes > 1; | |
839 | } | |
394ed8e4 SL |
840 | |
841 | /* clear unsupported mddev_flags */ | |
842 | static inline void mddev_clear_unsupported_flags(struct mddev *mddev, | |
843 | unsigned long unsupported_flags) | |
844 | { | |
845 | mddev->flags &= ~unsupported_flags; | |
846 | } | |
26483819 | 847 | |
3deff1a7 CH |
848 | static inline void mddev_check_write_zeroes(struct mddev *mddev, struct bio *bio) |
849 | { | |
850 | if (bio_op(bio) == REQ_OP_WRITE_ZEROES && | |
309dca30 | 851 | !bio->bi_bdev->bd_disk->queue->limits.max_write_zeroes_sectors) |
3deff1a7 CH |
852 | mddev->queue->limits.max_write_zeroes_sectors = 0; |
853 | } | |
d82fa81c | 854 | |
7e0adbfc CH |
855 | struct mdu_array_info_s; |
856 | struct mdu_disk_info_s; | |
857 | ||
1a6a0506 | 858 | extern int mdp_major; |
a022325a | 859 | extern struct workqueue_struct *md_bitmap_wq; |
d82fa81c | 860 | void md_autostart_arrays(int part); |
7e0adbfc CH |
861 | int md_set_array_info(struct mddev *mddev, struct mdu_array_info_s *info); |
862 | int md_add_new_disk(struct mddev *mddev, struct mdu_disk_info_s *info); | |
863 | int do_md_run(struct mddev *mddev); | |
864 | ||
865 | extern const struct block_device_operations md_fops; | |
d82fa81c | 866 | |
63fe0817 | 867 | #endif /* _MD_MD_H */ |