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 | 61 | struct block_device *bdev; /* block device handle */ |
9f0f5a30 | 62 | struct bdev_handle *bdev_handle; /* Handle from open for bdev */ |
1da177e4 | 63 | |
2699b672 | 64 | struct page *sb_page, *bb_page; |
1da177e4 | 65 | int sb_loaded; |
42543769 | 66 | __u64 sb_events; |
1da177e4 | 67 | sector_t data_offset; /* start of data in array */ |
c6563a8c | 68 | sector_t new_data_offset;/* only relevant while reshaping */ |
f72ffdd6 | 69 | sector_t sb_start; /* offset of the super block (in 512byte sectors) */ |
0002b271 | 70 | int sb_size; /* bytes in the superblock */ |
1da177e4 LT |
71 | int preferred_minor; /* autorun support */ |
72 | ||
86e6ffdd N |
73 | struct kobject kobj; |
74 | ||
1da177e4 LT |
75 | /* A device can be in one of three states based on two flags: |
76 | * Not working: faulty==1 in_sync==0 | |
77 | * Fully working: faulty==0 in_sync==1 | |
78 | * Working, but not | |
79 | * in sync with array | |
80 | * faulty==0 in_sync==0 | |
81 | * | |
82 | * It can never have faulty==1, in_sync==1 | |
83 | * This reduces the burden of testing multiple flags in many cases | |
84 | */ | |
1da177e4 | 85 | |
2d78f8c4 | 86 | unsigned long flags; /* bit set of 'enum flag_bits' bits. */ |
6bfe0b49 | 87 | wait_queue_head_t blocked_wait; |
8ddf9efe | 88 | |
1da177e4 LT |
89 | int desc_nr; /* descriptor index in the superblock */ |
90 | int raid_disk; /* role of device in array */ | |
e93f68a1 N |
91 | int new_raid_disk; /* role that the device will have in |
92 | * the array after a level-change completes. | |
93 | */ | |
41158c7e N |
94 | int saved_raid_disk; /* role that device used to have in the |
95 | * array and could again if we did a partial | |
96 | * resync from the bitmap | |
97 | */ | |
3069aa8d SL |
98 | union { |
99 | sector_t recovery_offset;/* If this device has been partially | |
5fd6c1dc N |
100 | * recovered, this is where we were |
101 | * up to. | |
102 | */ | |
3069aa8d SL |
103 | sector_t journal_tail; /* If this device is a journal device, |
104 | * this is the journal tail (journal | |
105 | * recovery start point) | |
106 | */ | |
107 | }; | |
1da177e4 LT |
108 | |
109 | atomic_t nr_pending; /* number of pending requests. | |
110 | * only maintained for arrays that | |
111 | * support hot removal | |
112 | */ | |
ba22dcbf N |
113 | atomic_t read_errors; /* number of consecutive read errors that |
114 | * we have tried to ignore. | |
115 | */ | |
0e3ef49e | 116 | time64_t last_read_error; /* monotonic time since our |
1e50915f RB |
117 | * last read error |
118 | */ | |
4dbcdc75 N |
119 | atomic_t corrected_errors; /* number of corrected read errors, |
120 | * for reporting to userspace and storing | |
121 | * in superblock. | |
122 | */ | |
3e148a32 | 123 | |
69b00b5b | 124 | struct serial_in_rdev *serial; /* used for raid1 io serialization */ |
3e148a32 | 125 | |
324a56e1 | 126 | struct kernfs_node *sysfs_state; /* handle for 'state' |
3c0ee63a | 127 | * sysfs entry */ |
e1a86dbb JB |
128 | /* handle for 'unacknowledged_bad_blocks' sysfs dentry */ |
129 | struct kernfs_node *sysfs_unack_badblocks; | |
130 | /* handle for 'bad_blocks' sysfs dentry */ | |
131 | struct kernfs_node *sysfs_badblocks; | |
fc974ee2 | 132 | struct badblocks badblocks; |
ea0213e0 AP |
133 | |
134 | struct { | |
135 | short offset; /* Offset from superblock to start of PPL. | |
136 | * Not used by external metadata. */ | |
137 | unsigned int size; /* Size in sectors of the PPL space */ | |
138 | sector_t sector; /* First sector of the PPL space */ | |
139 | } ppl; | |
1da177e4 | 140 | }; |
2d78f8c4 N |
141 | enum flag_bits { |
142 | Faulty, /* device is known to have a fault */ | |
143 | In_sync, /* device is in_sync with rest of array */ | |
8313b8e5 | 144 | Bitmap_sync, /* ..actually, not quite In_sync. Need a |
4aaf7694 GJ |
145 | * bitmap-based recovery to get fully in sync. |
146 | * The bit is only meaningful before device | |
147 | * has been passed to pers->hot_add_disk. | |
8313b8e5 | 148 | */ |
2d78f8c4 N |
149 | WriteMostly, /* Avoid reading if at all possible */ |
150 | AutoDetected, /* added by auto-detect */ | |
151 | Blocked, /* An error occurred but has not yet | |
152 | * been acknowledged by the metadata | |
153 | * handler, so don't allow writes | |
154 | * until it is cleared */ | |
155 | WriteErrorSeen, /* A write error has been seen on this | |
156 | * device | |
157 | */ | |
158 | FaultRecorded, /* Intermediate state for clearing | |
159 | * Blocked. The Fault is/will-be | |
160 | * recorded in the metadata, but that | |
161 | * metadata hasn't been stored safely | |
162 | * on disk yet. | |
163 | */ | |
164 | BlockedBadBlocks, /* A writer is blocked because they | |
165 | * found an unacknowledged bad-block. | |
166 | * This can safely be cleared at any | |
167 | * time, and the writer will re-check. | |
168 | * It may be set at any time, and at | |
169 | * worst the writer will timeout and | |
170 | * re-check. So setting it as | |
171 | * accurately as possible is good, but | |
172 | * not absolutely critical. | |
173 | */ | |
174 | WantReplacement, /* This device is a candidate to be | |
175 | * hot-replaced, either because it has | |
176 | * reported some faults, or because | |
177 | * of explicit request. | |
178 | */ | |
179 | Replacement, /* This device is a replacement for | |
180 | * a want_replacement device with same | |
181 | * raid_disk number. | |
182 | */ | |
1aee41f6 GR |
183 | Candidate, /* For clustered environments only: |
184 | * This device is seen locally but not | |
185 | * by the whole cluster | |
186 | */ | |
bac624f3 SL |
187 | Journal, /* This device is used as journal for |
188 | * raid-5/6. | |
189 | * Usually, this device should be faster | |
190 | * than other devices in the array | |
191 | */ | |
659b254f | 192 | ClusterRemove, |
35b785f7 TM |
193 | ExternalBbl, /* External metadata provides bad |
194 | * block management for a disk | |
195 | */ | |
688834e6 N |
196 | FailFast, /* Minimal retries should be attempted on |
197 | * this device, so use REQ_FAILFAST_DEV. | |
198 | * Also don't try to repair failed reads. | |
199 | * It is expects that no bad block log | |
200 | * is present. | |
201 | */ | |
46533ff7 N |
202 | LastDev, /* Seems to be the last working dev as |
203 | * it didn't fail, so don't use FailFast | |
204 | * any more for metadata | |
205 | */ | |
404659cf GJ |
206 | CollisionCheck, /* |
207 | * check if there is collision between raid1 | |
208 | * serial bios. | |
3e148a32 | 209 | */ |
2c27d09d | 210 | Nonrot, /* non-rotational device (SSD) */ |
2d78f8c4 | 211 | }; |
1da177e4 | 212 | |
3cb03002 | 213 | static inline int is_badblock(struct md_rdev *rdev, sector_t s, int sectors, |
2230dfe4 N |
214 | sector_t *first_bad, int *bad_sectors) |
215 | { | |
216 | if (unlikely(rdev->badblocks.count)) { | |
fc974ee2 | 217 | int rv = badblocks_check(&rdev->badblocks, rdev->data_offset + s, |
2230dfe4 N |
218 | sectors, |
219 | first_bad, bad_sectors); | |
220 | if (rv) | |
221 | *first_bad -= rdev->data_offset; | |
222 | return rv; | |
223 | } | |
224 | return 0; | |
225 | } | |
3a0f007b YK |
226 | |
227 | static inline int rdev_has_badblock(struct md_rdev *rdev, sector_t s, | |
228 | int sectors) | |
229 | { | |
230 | sector_t first_bad; | |
231 | int bad_sectors; | |
232 | ||
233 | return is_badblock(rdev, s, sectors, &first_bad, &bad_sectors); | |
234 | } | |
235 | ||
3cb03002 | 236 | extern int rdev_set_badblocks(struct md_rdev *rdev, sector_t s, int sectors, |
c6563a8c N |
237 | int is_new); |
238 | extern int rdev_clear_badblocks(struct md_rdev *rdev, sector_t s, int sectors, | |
239 | int is_new); | |
c4ce867f GR |
240 | struct md_cluster_info; |
241 | ||
9631abdb MT |
242 | /** |
243 | * enum mddev_flags - md device flags. | |
244 | * @MD_ARRAY_FIRST_USE: First use of array, needs initialization. | |
245 | * @MD_CLOSING: If set, we are closing the array, do not open it then. | |
246 | * @MD_JOURNAL_CLEAN: A raid with journal is already clean. | |
247 | * @MD_HAS_JOURNAL: The raid array has journal feature set. | |
248 | * @MD_CLUSTER_RESYNC_LOCKED: cluster raid only, which means node, already took | |
249 | * resync lock, need to release the lock. | |
250 | * @MD_FAILFAST_SUPPORTED: Using MD_FAILFAST on metadata writes is supported as | |
251 | * calls to md_error() will never cause the array to | |
252 | * become failed. | |
253 | * @MD_HAS_PPL: The raid array has PPL feature set. | |
254 | * @MD_HAS_MULTIPLE_PPLS: The raid array has multiple PPLs feature set. | |
9631abdb MT |
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, | |
9631abdb MT |
271 | MD_NOT_READY, |
272 | MD_BROKEN, | |
12a6caf2 | 273 | MD_DELETED, |
be306c29 | 274 | }; |
2953079c SL |
275 | |
276 | enum mddev_sb_flags { | |
277 | MD_SB_CHANGE_DEVS, /* Some device status has changed */ | |
278 | MD_SB_CHANGE_CLEAN, /* transition to or from 'clean' */ | |
279 | MD_SB_CHANGE_PENDING, /* switch from 'clean' to 'active' in progress */ | |
280 | MD_SB_NEED_REWRITE, /* metadata write needs to be repeated */ | |
281 | }; | |
282 | ||
404659cf GJ |
283 | #define NR_SERIAL_INFOS 8 |
284 | /* record current range of serialize IOs */ | |
285 | struct serial_info { | |
69b00b5b GJ |
286 | struct rb_node node; |
287 | sector_t start; /* start sector of rb node */ | |
288 | sector_t last; /* end sector of rb node */ | |
289 | sector_t _subtree_last; /* highest sector in subtree of rb node */ | |
3e148a32 GJ |
290 | }; |
291 | ||
eac58d08 LG |
292 | /* |
293 | * mddev->curr_resync stores the current sector of the resync but | |
294 | * also has some overloaded values. | |
295 | */ | |
296 | enum { | |
297 | /* No resync in progress */ | |
298 | MD_RESYNC_NONE = 0, | |
299 | /* Yielded to allow another conflicting resync to commence */ | |
300 | MD_RESYNC_YIELDED = 1, | |
301 | /* Delayed to check that there is no conflict with another sync */ | |
302 | MD_RESYNC_DELAYED = 2, | |
303 | /* Any value greater than or equal to this is in an active resync */ | |
304 | MD_RESYNC_ACTIVE = 3, | |
305 | }; | |
306 | ||
fd01b88c | 307 | struct mddev { |
1da177e4 | 308 | void *private; |
84fc4b56 | 309 | struct md_personality *pers; |
1da177e4 LT |
310 | dev_t unit; |
311 | int md_minor; | |
7a0a5355 | 312 | struct list_head disks; |
850b2b42 | 313 | unsigned long flags; |
2953079c | 314 | unsigned long sb_flags; |
850b2b42 | 315 | |
409c57f3 | 316 | int suspended; |
714d2015 | 317 | struct mutex suspend_mutex; |
72adae23 | 318 | struct percpu_ref active_io; |
1da177e4 | 319 | int ro; |
bb4f1e9d N |
320 | int sysfs_active; /* set when sysfs deletes |
321 | * are happening, so run/ | |
322 | * takeover/stop are not safe | |
323 | */ | |
1da177e4 LT |
324 | struct gendisk *gendisk; |
325 | ||
eae1701f | 326 | struct kobject kobj; |
d3374825 N |
327 | int hold_active; |
328 | #define UNTIL_IOCTL 1 | |
efeb53c0 | 329 | #define UNTIL_STOP 2 |
eae1701f | 330 | |
1da177e4 LT |
331 | /* Superblock information */ |
332 | int major_version, | |
333 | minor_version, | |
334 | patch_version; | |
335 | int persistent; | |
f72ffdd6 | 336 | int external; /* metadata is |
e691063a N |
337 | * managed externally */ |
338 | char metadata_type[17]; /* externally set*/ | |
77a68698 | 339 | int chunk_sectors; |
9ebc6ef1 | 340 | time64_t ctime, utime; |
1da177e4 | 341 | int level, layout; |
d9d166c2 | 342 | char clevel[16]; |
1da177e4 LT |
343 | int raid_disks; |
344 | int max_disks; | |
f72ffdd6 | 345 | sector_t dev_sectors; /* used size of |
58c0fed4 | 346 | * component devices */ |
f233ea5c | 347 | sector_t array_sectors; /* exported array size */ |
b522adcd DW |
348 | int external_size; /* size managed |
349 | * externally */ | |
1da177e4 | 350 | __u64 events; |
a8707c08 N |
351 | /* If the last 'event' was simply a clean->dirty transition, and |
352 | * we didn't write it to the spares, then it is safe and simple | |
353 | * to just decrement the event count on a dirty->clean transition. | |
354 | * So we record that possibility here. | |
355 | */ | |
356 | int can_decrease_events; | |
1da177e4 LT |
357 | |
358 | char uuid[16]; | |
359 | ||
f6705578 N |
360 | /* If the array is being reshaped, we need to record the |
361 | * new shape and an indication of where we are up to. | |
362 | * This is written to the superblock. | |
363 | * If reshape_position is MaxSector, then no reshape is happening (yet). | |
364 | */ | |
365 | sector_t reshape_position; | |
664e7c41 | 366 | int delta_disks, new_level, new_layout; |
77a68698 | 367 | int new_chunk_sectors; |
2c810cdd | 368 | int reshape_backwards; |
f6705578 | 369 | |
44693154 YK |
370 | struct md_thread __rcu *thread; /* management thread */ |
371 | struct md_thread __rcu *sync_thread; /* doing resync or reconstruct */ | |
c4a39551 JB |
372 | |
373 | /* 'last_sync_action' is initialized to "none". It is set when a | |
374 | * sync operation (i.e "data-check", "requested-resync", "resync", | |
375 | * "recovery", or "reshape") is started. It holds this value even | |
376 | * when the sync thread is "frozen" (interrupted) or "idle" (stopped | |
377 | * or finished). It is overwritten when a new sync operation is begun. | |
378 | */ | |
379 | char *last_sync_action; | |
ff4e8d9a | 380 | sector_t curr_resync; /* last block scheduled */ |
97e4f42d N |
381 | /* As resync requests can complete out of order, we cannot easily track |
382 | * how much resync has been completed. So we occasionally pause until | |
383 | * everything completes, then set curr_resync_completed to curr_resync. | |
384 | * As such it may be well behind the real resync mark, but it is a value | |
385 | * we are certain of. | |
386 | */ | |
387 | sector_t curr_resync_completed; | |
1da177e4 LT |
388 | unsigned long resync_mark; /* a recent timestamp */ |
389 | sector_t resync_mark_cnt;/* blocks written at resync_mark */ | |
ff4e8d9a | 390 | sector_t curr_mark_cnt; /* blocks scheduled now */ |
1da177e4 LT |
391 | |
392 | sector_t resync_max_sectors; /* may be set by personality */ | |
9d88883e | 393 | |
7f7583d4 | 394 | atomic64_t resync_mismatches; /* count of sectors where |
9d88883e N |
395 | * parity/replica mismatch found |
396 | */ | |
e464eafd N |
397 | |
398 | /* allow user-space to request suspension of IO to regions of the array */ | |
399 | sector_t suspend_lo; | |
400 | sector_t suspend_hi; | |
88202a0c N |
401 | /* if zero, use the system-wide default */ |
402 | int sync_speed_min; | |
403 | int sync_speed_max; | |
404 | ||
90b08710 BS |
405 | /* resync even though the same disks are shared among md-devices */ |
406 | int parallel_resync; | |
407 | ||
6ff8d8ec | 408 | int ok_start_degraded; |
5fd6c1dc | 409 | |
1da177e4 | 410 | unsigned long recovery; |
5389042f N |
411 | /* If a RAID personality determines that recovery (of a particular |
412 | * device) will fail due to a read error on the source device, it | |
413 | * takes a copy of this number and does not attempt recovery again | |
414 | * until this number changes. | |
415 | */ | |
416 | int recovery_disabled; | |
1da177e4 LT |
417 | |
418 | int in_sync; /* know to not need resync */ | |
c8c00a69 N |
419 | /* 'open_mutex' avoids races between 'md_open' and 'do_md_stop', so |
420 | * that we are never stopping an array while it is open. | |
421 | * 'reconfig_mutex' protects all other reconfiguration. | |
422 | * These locks are separate due to conflicting interactions | |
a8698707 | 423 | * with disk->open_mutex. |
c8c00a69 | 424 | * Lock ordering is: |
a8698707 CH |
425 | * reconfig_mutex -> disk->open_mutex |
426 | * disk->open_mutex -> open_mutex: e.g. __blkdev_get -> md_open | |
c8c00a69 N |
427 | */ |
428 | struct mutex open_mutex; | |
df5b89b3 | 429 | struct mutex reconfig_mutex; |
f2ea68cf N |
430 | atomic_t active; /* general refcount */ |
431 | atomic_t openers; /* number of active opens */ | |
1da177e4 | 432 | |
f0b4f7e2 N |
433 | int changed; /* True if we might need to |
434 | * reread partition info */ | |
1da177e4 LT |
435 | int degraded; /* whether md should consider |
436 | * adding a spare | |
437 | */ | |
438 | ||
439 | atomic_t recovery_active; /* blocks scheduled, but not written */ | |
440 | wait_queue_head_t recovery_wait; | |
441 | sector_t recovery_cp; | |
5e96ee65 NB |
442 | sector_t resync_min; /* user requested sync |
443 | * starts here */ | |
c6207277 N |
444 | sector_t resync_max; /* resync should pause |
445 | * when it gets here */ | |
06d91a5f | 446 | |
324a56e1 | 447 | struct kernfs_node *sysfs_state; /* handle for 'array_state' |
b62b7590 N |
448 | * file in sysfs. |
449 | */ | |
324a56e1 | 450 | struct kernfs_node *sysfs_action; /* handle for 'sync_action' */ |
e1a86dbb JB |
451 | struct kernfs_node *sysfs_completed; /*handle for 'sync_completed' */ |
452 | struct kernfs_node *sysfs_degraded; /*handle for 'degraded' */ | |
453 | struct kernfs_node *sysfs_level; /*handle for 'level' */ | |
b62b7590 | 454 | |
ac619781 YK |
455 | /* used for delayed sysfs removal */ |
456 | struct work_struct del_work; | |
457 | /* used for register new sync thread */ | |
458 | struct work_struct sync_work; | |
d3374825 | 459 | |
85572d7c N |
460 | /* "lock" protects: |
461 | * flush_bio transition from NULL to !NULL | |
462 | * rdev superblocks, events | |
463 | * clearing MD_CHANGE_* | |
464 | * in_sync - and related safemode and MD_CHANGE changes | |
36d091f4 | 465 | * pers (also protected by reconfig_mutex and pending IO). |
978a7a47 | 466 | * clearing ->bitmap |
4af1a041 | 467 | * clearing ->bitmap_info.file |
23da422b N |
468 | * changing ->resync_{min,max} |
469 | * setting MD_RECOVERY_RUNNING (which interacts with resync_{min,max}) | |
85572d7c N |
470 | */ |
471 | spinlock_t lock; | |
3d310eb7 | 472 | wait_queue_head_t sb_wait; /* for waiting on superblock updates */ |
7bfa19f2 | 473 | atomic_t pending_writes; /* number of active superblock writes */ |
06d91a5f | 474 | |
1da177e4 LT |
475 | unsigned int safemode; /* if set, update "clean" superblock |
476 | * when no writes pending. | |
f72ffdd6 | 477 | */ |
1da177e4 LT |
478 | unsigned int safemode_delay; |
479 | struct timer_list safemode_timer; | |
4ad23a97 N |
480 | struct percpu_ref writes_pending; |
481 | int sync_checkers; /* # of threads checking writes_pending */ | |
165125e1 | 482 | struct request_queue *queue; /* for plugging ... */ |
1da177e4 | 483 | |
f72ffdd6 | 484 | struct bitmap *bitmap; /* the bitmap for the device */ |
c3d9714e N |
485 | struct { |
486 | struct file *file; /* the bitmap file */ | |
f6af949c | 487 | loff_t offset; /* offset from superblock of |
c3d9714e N |
488 | * start of bitmap. May be |
489 | * negative, but not '0' | |
f6af949c | 490 | * For external metadata, offset |
f72ffdd6 | 491 | * from start of device. |
c3d9714e | 492 | */ |
6409bb05 | 493 | unsigned long space; /* space available at this offset */ |
f6af949c | 494 | loff_t default_offset; /* this is the offset to use when |
c3d9714e N |
495 | * hot-adding a bitmap. It should |
496 | * eventually be settable by sysfs. | |
497 | */ | |
6409bb05 N |
498 | unsigned long default_space; /* space available at |
499 | * default offset */ | |
c3d9714e | 500 | struct mutex mutex; |
42a04b50 | 501 | unsigned long chunksize; |
ac2f40be | 502 | unsigned long daemon_sleep; /* how many jiffies between updates? */ |
42a04b50 | 503 | unsigned long max_write_behind; /* write-behind mode */ |
ece5cff0 | 504 | int external; |
c4ce867f | 505 | int nodes; /* Maximum number of nodes in the cluster */ |
cf921cc1 | 506 | char cluster_name[64]; /* Name of the cluster */ |
c3d9714e | 507 | } bitmap_info; |
32a7627c | 508 | |
f72ffdd6 | 509 | atomic_t max_corr_read_errors; /* max read retries */ |
1da177e4 | 510 | struct list_head all_mddevs; |
a2826aa9 | 511 | |
c32dc040 | 512 | const struct attribute_group *to_remove; |
252ac522 | 513 | |
afeee514 KO |
514 | struct bio_set bio_set; |
515 | struct bio_set sync_set; /* for sync operations like | |
5a85071c N |
516 | * metadata and bitmap writes |
517 | */ | |
c687297b | 518 | struct bio_set io_clone_set; |
a167f663 | 519 | |
4bc034d3 N |
520 | /* Generic flush handling. |
521 | * The last to finish preflush schedules a worker to submit | |
522 | * the rest of the request (without the REQ_PREFLUSH flag). | |
523 | */ | |
524 | struct bio *flush_bio; | |
525 | atomic_t flush_pending; | |
81ba3c24 PG |
526 | ktime_t start_flush, prev_flush_start; /* prev_flush_start is when the previous completed |
527 | * flush was started. | |
528 | */ | |
4bc034d3 | 529 | struct work_struct flush_work; |
768a418d | 530 | struct work_struct event_work; /* used by dm to report failure event */ |
404659cf | 531 | mempool_t *serial_info_pool; |
fd01b88c | 532 | void (*sync_super)(struct mddev *mddev, struct md_rdev *rdev); |
c4ce867f | 533 | struct md_cluster_info *cluster_info; |
15858fa5 | 534 | unsigned int good_device_nr; /* good device num within cluster raid */ |
78f57ef9 | 535 | unsigned int noio_flag; /* for memalloc scope API */ |
4b6c1060 | 536 | |
3ce94ce5 YK |
537 | /* |
538 | * Temporarily store rdev that will be finally removed when | |
4934b640 | 539 | * reconfig_mutex is unlocked, protected by reconfig_mutex. |
3ce94ce5 YK |
540 | */ |
541 | struct list_head deleting; | |
3ce94ce5 | 542 | |
6f56f0c4 YK |
543 | /* Used to synchronize idle and frozen for action_store() */ |
544 | struct mutex sync_mutex; | |
130443d6 YK |
545 | /* The sequence number for sync thread */ |
546 | atomic_t sync_seq; | |
6f56f0c4 | 547 | |
4b6c1060 | 548 | bool has_superblocks:1; |
9a567843 | 549 | bool fail_last_dev:1; |
3938f5fb | 550 | bool serialize_policy:1; |
be306c29 N |
551 | }; |
552 | ||
553 | enum recovery_flags { | |
554 | /* | |
555 | * If neither SYNC or RESHAPE are set, then it is a recovery. | |
556 | */ | |
557 | MD_RECOVERY_RUNNING, /* a thread is running, or about to be started */ | |
558 | MD_RECOVERY_SYNC, /* actually doing a resync, not a recovery */ | |
559 | MD_RECOVERY_RECOVER, /* doing recovery, or need to try it. */ | |
560 | MD_RECOVERY_INTR, /* resync needs to be aborted for some reason */ | |
561 | MD_RECOVERY_DONE, /* thread is done and is waiting to be reaped */ | |
562 | MD_RECOVERY_NEEDED, /* we might need to start a resync/recover */ | |
563 | MD_RECOVERY_REQUESTED, /* user-space has requested a sync (used with SYNC) */ | |
564 | MD_RECOVERY_CHECK, /* user-space request for check-only, no repair */ | |
565 | MD_RECOVERY_RESHAPE, /* A reshape is happening */ | |
566 | MD_RECOVERY_FROZEN, /* User request to abort, and not restart, any action */ | |
567 | MD_RECOVERY_ERROR, /* sync-action interrupted because io-error */ | |
d5d885fd | 568 | MD_RECOVERY_WAIT, /* waiting for pers->start() to finish */ |
0357ba27 | 569 | MD_RESYNCING_REMOTE, /* remote node is running resync thread */ |
1da177e4 LT |
570 | }; |
571 | ||
314e9af0 YK |
572 | enum md_ro_state { |
573 | MD_RDWR, | |
574 | MD_RDONLY, | |
575 | MD_AUTO_READ, | |
576 | MD_MAX_STATE | |
577 | }; | |
578 | ||
579 | static inline bool md_is_rdwr(struct mddev *mddev) | |
580 | { | |
581 | return (mddev->ro == MD_RDWR); | |
582 | } | |
583 | ||
5c47daf6 N |
584 | static inline int __must_check mddev_lock(struct mddev *mddev) |
585 | { | |
586 | return mutex_lock_interruptible(&mddev->reconfig_mutex); | |
587 | } | |
588 | ||
589 | /* Sometimes we need to take the lock in a situation where | |
590 | * failure due to interrupts is not acceptable. | |
591 | */ | |
592 | static inline void mddev_lock_nointr(struct mddev *mddev) | |
593 | { | |
594 | mutex_lock(&mddev->reconfig_mutex); | |
595 | } | |
596 | ||
5c47daf6 N |
597 | static inline int mddev_trylock(struct mddev *mddev) |
598 | { | |
599 | return mutex_trylock(&mddev->reconfig_mutex); | |
600 | } | |
601 | extern void mddev_unlock(struct mddev *mddev); | |
602 | ||
1da177e4 LT |
603 | static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors) |
604 | { | |
4245e52d | 605 | atomic_add(nr_sectors, &bdev->bd_disk->sync_io); |
1da177e4 LT |
606 | } |
607 | ||
74d46992 CH |
608 | static inline void md_sync_acct_bio(struct bio *bio, unsigned long nr_sectors) |
609 | { | |
309dca30 | 610 | md_sync_acct(bio->bi_bdev, nr_sectors); |
74d46992 CH |
611 | } |
612 | ||
84fc4b56 | 613 | struct md_personality |
1da177e4 LT |
614 | { |
615 | char *name; | |
2604b703 N |
616 | int level; |
617 | struct list_head list; | |
1da177e4 | 618 | struct module *owner; |
775d7831 | 619 | bool __must_check (*make_request)(struct mddev *mddev, struct bio *bio); |
d5d885fd SL |
620 | /* |
621 | * start up works that do NOT require md_thread. tasks that | |
622 | * requires md_thread should go into start() | |
623 | */ | |
fd01b88c | 624 | int (*run)(struct mddev *mddev); |
d5d885fd SL |
625 | /* start up works that require md threads */ |
626 | int (*start)(struct mddev *mddev); | |
afa0f557 | 627 | void (*free)(struct mddev *mddev, void *priv); |
fd01b88c | 628 | void (*status)(struct seq_file *seq, struct mddev *mddev); |
1da177e4 | 629 | /* error_handler must set ->faulty and clear ->in_sync |
f72ffdd6 | 630 | * if appropriate, and should abort recovery if needed |
1da177e4 | 631 | */ |
fd01b88c N |
632 | void (*error_handler)(struct mddev *mddev, struct md_rdev *rdev); |
633 | int (*hot_add_disk) (struct mddev *mddev, struct md_rdev *rdev); | |
b8321b68 | 634 | int (*hot_remove_disk) (struct mddev *mddev, struct md_rdev *rdev); |
fd01b88c | 635 | int (*spare_active) (struct mddev *mddev); |
09314799 | 636 | sector_t (*sync_request)(struct mddev *mddev, sector_t sector_nr, int *skipped); |
fd01b88c N |
637 | int (*resize) (struct mddev *mddev, sector_t sectors); |
638 | sector_t (*size) (struct mddev *mddev, sector_t sectors, int raid_disks); | |
639 | int (*check_reshape) (struct mddev *mddev); | |
640 | int (*start_reshape) (struct mddev *mddev); | |
641 | void (*finish_reshape) (struct mddev *mddev); | |
7564beda | 642 | void (*update_reshape_pos) (struct mddev *mddev); |
b03e0ccb N |
643 | /* quiesce suspends or resumes internal processing. |
644 | * 1 - stop new actions and wait for action io to complete | |
645 | * 0 - return to normal behaviour | |
36fa3063 | 646 | */ |
b03e0ccb | 647 | void (*quiesce) (struct mddev *mddev, int quiesce); |
245f46c2 N |
648 | /* takeover is used to transition an array from one |
649 | * personality to another. The new personality must be able | |
650 | * to handle the data in the current layout. | |
651 | * e.g. 2drive raid1 -> 2drive raid5 | |
652 | * ndrive raid5 -> degraded n+1drive raid6 with special layout | |
653 | * If the takeover succeeds, a new 'private' structure is returned. | |
654 | * This needs to be installed and then ->run used to activate the | |
655 | * array. | |
656 | */ | |
fd01b88c | 657 | void *(*takeover) (struct mddev *mddev); |
ba903a3e AP |
658 | /* Changes the consistency policy of an active array. */ |
659 | int (*change_consistency_policy)(struct mddev *mddev, const char *buf); | |
1da177e4 LT |
660 | }; |
661 | ||
007583c9 N |
662 | struct md_sysfs_entry { |
663 | struct attribute attr; | |
fd01b88c N |
664 | ssize_t (*show)(struct mddev *, char *); |
665 | ssize_t (*store)(struct mddev *, const char *, size_t); | |
007583c9 | 666 | }; |
c32dc040 | 667 | extern const struct attribute_group md_bitmap_group; |
007583c9 | 668 | |
324a56e1 | 669 | static inline struct kernfs_node *sysfs_get_dirent_safe(struct kernfs_node *sd, char *name) |
00bcb4ac N |
670 | { |
671 | if (sd) | |
388975cc | 672 | return sysfs_get_dirent(sd, name); |
00bcb4ac N |
673 | return sd; |
674 | } | |
324a56e1 | 675 | static inline void sysfs_notify_dirent_safe(struct kernfs_node *sd) |
00bcb4ac N |
676 | { |
677 | if (sd) | |
678 | sysfs_notify_dirent(sd); | |
679 | } | |
680 | ||
fd01b88c | 681 | static inline char * mdname (struct mddev * mddev) |
1da177e4 LT |
682 | { |
683 | return mddev->gendisk ? mddev->gendisk->disk_name : "mdX"; | |
684 | } | |
685 | ||
fd01b88c | 686 | static inline int sysfs_link_rdev(struct mddev *mddev, struct md_rdev *rdev) |
36fad858 NK |
687 | { |
688 | char nm[20]; | |
9b15603d SL |
689 | if (!test_bit(Replacement, &rdev->flags) && |
690 | !test_bit(Journal, &rdev->flags) && | |
691 | mddev->kobj.sd) { | |
2d78f8c4 N |
692 | sprintf(nm, "rd%d", rdev->raid_disk); |
693 | return sysfs_create_link(&mddev->kobj, &rdev->kobj, nm); | |
694 | } else | |
695 | return 0; | |
36fad858 NK |
696 | } |
697 | ||
fd01b88c | 698 | static inline void sysfs_unlink_rdev(struct mddev *mddev, struct md_rdev *rdev) |
36fad858 NK |
699 | { |
700 | char nm[20]; | |
9b15603d SL |
701 | if (!test_bit(Replacement, &rdev->flags) && |
702 | !test_bit(Journal, &rdev->flags) && | |
703 | mddev->kobj.sd) { | |
2d78f8c4 N |
704 | sprintf(nm, "rd%d", rdev->raid_disk); |
705 | sysfs_remove_link(&mddev->kobj, nm); | |
706 | } | |
36fad858 NK |
707 | } |
708 | ||
1da177e4 LT |
709 | /* |
710 | * iterates through some rdev ringlist. It's safe to remove the | |
711 | * current 'rdev'. Dont touch 'tmp' though. | |
712 | */ | |
159ec1fc CR |
713 | #define rdev_for_each_list(rdev, tmp, head) \ |
714 | list_for_each_entry_safe(rdev, tmp, head, same_set) | |
715 | ||
1da177e4 LT |
716 | /* |
717 | * iterates through the 'same array disks' ringlist | |
718 | */ | |
dafb20fa N |
719 | #define rdev_for_each(rdev, mddev) \ |
720 | list_for_each_entry(rdev, &((mddev)->disks), same_set) | |
721 | ||
722 | #define rdev_for_each_safe(rdev, tmp, mddev) \ | |
159ec1fc | 723 | list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set) |
1da177e4 | 724 | |
4b80991c N |
725 | #define rdev_for_each_rcu(rdev, mddev) \ |
726 | list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set) | |
727 | ||
2b8bf345 | 728 | struct md_thread { |
4ed8731d | 729 | void (*run) (struct md_thread *thread); |
fd01b88c | 730 | struct mddev *mddev; |
1da177e4 | 731 | wait_queue_head_t wqueue; |
f72ffdd6 | 732 | unsigned long flags; |
1da177e4 | 733 | struct task_struct *tsk; |
32a7627c | 734 | unsigned long timeout; |
4ed8731d | 735 | void *private; |
2b8bf345 | 736 | }; |
1da177e4 | 737 | |
c687297b | 738 | struct md_io_clone { |
76fed014 XN |
739 | struct mddev *mddev; |
740 | struct bio *orig_bio; | |
741 | unsigned long start_time; | |
742 | struct bio bio_clone; | |
10764815 GJ |
743 | }; |
744 | ||
1da177e4 LT |
745 | #define THREAD_WAKEUP 0 |
746 | ||
1345b1d8 N |
747 | static inline void safe_put_page(struct page *p) |
748 | { | |
749 | if (p) put_page(p); | |
750 | } | |
751 | ||
84fc4b56 N |
752 | extern int register_md_personality(struct md_personality *p); |
753 | extern int unregister_md_personality(struct md_personality *p); | |
edb39c9d GR |
754 | extern int register_md_cluster_operations(struct md_cluster_operations *ops, |
755 | struct module *module); | |
756 | extern int unregister_md_cluster_operations(void); | |
757 | extern int md_setup_cluster(struct mddev *mddev, int nodes); | |
758 | extern void md_cluster_stop(struct mddev *mddev); | |
2b8bf345 | 759 | extern struct md_thread *md_register_thread( |
4ed8731d | 760 | void (*run)(struct md_thread *thread), |
2b8bf345 N |
761 | struct mddev *mddev, |
762 | const char *name); | |
7eb8ff02 | 763 | extern void md_unregister_thread(struct mddev *mddev, struct md_thread __rcu **threadp); |
44693154 | 764 | extern void md_wakeup_thread(struct md_thread __rcu *thread); |
fd01b88c | 765 | extern void md_check_recovery(struct mddev *mddev); |
d0a18034 | 766 | extern void md_reap_sync_thread(struct mddev *mddev); |
cc27b0c7 | 767 | extern bool md_write_start(struct mddev *mddev, struct bio *bi); |
49728050 | 768 | extern void md_write_inc(struct mddev *mddev, struct bio *bi); |
fd01b88c N |
769 | extern void md_write_end(struct mddev *mddev); |
770 | extern void md_done_sync(struct mddev *mddev, int blocks, int ok); | |
771 | extern void md_error(struct mddev *mddev, struct md_rdev *rdev); | |
c6563a8c | 772 | extern void md_finish_reshape(struct mddev *mddev); |
cf78408f XN |
773 | void md_submit_discard_bio(struct mddev *mddev, struct md_rdev *rdev, |
774 | struct bio *bio, sector_t start, sector_t size); | |
10764815 | 775 | void md_account_bio(struct mddev *mddev, struct bio **bio); |
fd01b88c | 776 | |
775d7831 | 777 | extern bool __must_check md_flush_request(struct mddev *mddev, struct bio *bio); |
fd01b88c | 778 | extern void md_super_write(struct mddev *mddev, struct md_rdev *rdev, |
92022950 | 779 | sector_t sector, int size, struct page *page); |
46533ff7 | 780 | extern int md_super_wait(struct mddev *mddev); |
f72ffdd6 | 781 | extern int sync_page_io(struct md_rdev *rdev, sector_t sector, int size, |
4ce4c73f | 782 | struct page *page, blk_opf_t opf, bool metadata_op); |
4ed8731d | 783 | extern void md_do_sync(struct md_thread *thread); |
54679486 | 784 | extern void md_new_event(void); |
2214c260 | 785 | extern void md_allow_write(struct mddev *mddev); |
fd01b88c N |
786 | extern void md_wait_for_blocked_rdev(struct md_rdev *rdev, struct mddev *mddev); |
787 | extern void md_set_array_sectors(struct mddev *mddev, sector_t array_sectors); | |
788 | extern int md_check_no_bitmap(struct mddev *mddev); | |
789 | extern int md_integrity_register(struct mddev *mddev); | |
1501efad | 790 | extern int md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev); |
72e02075 | 791 | extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale); |
63fe0817 | 792 | |
d58eff83 YK |
793 | extern int mddev_init(struct mddev *mddev); |
794 | extern void mddev_destroy(struct mddev *mddev); | |
34cb92c0 CH |
795 | struct mddev *md_alloc(dev_t dev, char *name); |
796 | void mddev_put(struct mddev *mddev); | |
fd01b88c | 797 | extern int md_run(struct mddev *mddev); |
d5d885fd | 798 | extern int md_start(struct mddev *mddev); |
fd01b88c N |
799 | extern void md_stop(struct mddev *mddev); |
800 | extern void md_stop_writes(struct mddev *mddev); | |
3cb03002 | 801 | extern int md_rdev_init(struct md_rdev *rdev); |
545c8795 | 802 | extern void md_rdev_clear(struct md_rdev *rdev); |
63fe0817 | 803 | |
393debc2 | 804 | extern void md_handle_request(struct mddev *mddev, struct bio *bio); |
2b16a525 | 805 | extern int mddev_suspend(struct mddev *mddev, bool interruptible); |
fd01b88c | 806 | extern void mddev_resume(struct mddev *mddev); |
7a2347e2 YK |
807 | extern void md_idle_sync_thread(struct mddev *mddev); |
808 | extern void md_frozen_sync_thread(struct mddev *mddev); | |
809 | extern void md_unfrozen_sync_thread(struct mddev *mddev); | |
9cbb1750 | 810 | |
70bcecdb | 811 | extern void md_reload_sb(struct mddev *mddev, int raid_disk); |
1aee41f6 | 812 | extern void md_update_sb(struct mddev *mddev, int force); |
b4128c00 YK |
813 | extern void mddev_create_serial_pool(struct mddev *mddev, struct md_rdev *rdev); |
814 | extern void mddev_destroy_serial_pool(struct mddev *mddev, | |
815 | struct md_rdev *rdev); | |
57d051dc | 816 | struct md_rdev *md_find_rdev_nr_rcu(struct mddev *mddev, int nr); |
1532d9e8 | 817 | struct md_rdev *md_find_rdev_rcu(struct mddev *mddev, dev_t dev); |
dfe15ac1 | 818 | |
c31fea2f | 819 | static inline bool is_rdev_broken(struct md_rdev *rdev) |
62f7b198 | 820 | { |
c31fea2f | 821 | return !disk_live(rdev->bdev->bd_disk); |
62f7b198 GP |
822 | } |
823 | ||
dfe15ac1 HR |
824 | static inline void rdev_dec_pending(struct md_rdev *rdev, struct mddev *mddev) |
825 | { | |
826 | int faulty = test_bit(Faulty, &rdev->flags); | |
827 | if (atomic_dec_and_test(&rdev->nr_pending) && faulty) { | |
828 | set_bit(MD_RECOVERY_NEEDED, &mddev->recovery); | |
829 | md_wakeup_thread(mddev->thread); | |
830 | } | |
831 | } | |
832 | ||
edb39c9d | 833 | extern struct md_cluster_operations *md_cluster_ops; |
c4ce867f GR |
834 | static inline int mddev_is_clustered(struct mddev *mddev) |
835 | { | |
836 | return mddev->cluster_info && mddev->bitmap_info.nodes > 1; | |
837 | } | |
394ed8e4 SL |
838 | |
839 | /* clear unsupported mddev_flags */ | |
840 | static inline void mddev_clear_unsupported_flags(struct mddev *mddev, | |
841 | unsigned long unsupported_flags) | |
842 | { | |
843 | mddev->flags &= ~unsupported_flags; | |
844 | } | |
26483819 | 845 | |
3deff1a7 CH |
846 | static inline void mddev_check_write_zeroes(struct mddev *mddev, struct bio *bio) |
847 | { | |
848 | if (bio_op(bio) == REQ_OP_WRITE_ZEROES && | |
309dca30 | 849 | !bio->bi_bdev->bd_disk->queue->limits.max_write_zeroes_sectors) |
3deff1a7 CH |
850 | mddev->queue->limits.max_write_zeroes_sectors = 0; |
851 | } | |
d82fa81c | 852 | |
f45461e2 YK |
853 | static inline int mddev_suspend_and_lock(struct mddev *mddev) |
854 | { | |
855 | int ret; | |
856 | ||
2b16a525 | 857 | ret = mddev_suspend(mddev, true); |
f45461e2 YK |
858 | if (ret) |
859 | return ret; | |
860 | ||
861 | ret = mddev_lock(mddev); | |
862 | if (ret) | |
2b16a525 | 863 | mddev_resume(mddev); |
f45461e2 YK |
864 | |
865 | return ret; | |
866 | } | |
867 | ||
868 | static inline void mddev_suspend_and_lock_nointr(struct mddev *mddev) | |
869 | { | |
2b16a525 | 870 | mddev_suspend(mddev, false); |
f45461e2 YK |
871 | mutex_lock(&mddev->reconfig_mutex); |
872 | } | |
873 | ||
874 | static inline void mddev_unlock_and_resume(struct mddev *mddev) | |
875 | { | |
876 | mddev_unlock(mddev); | |
2b16a525 | 877 | mddev_resume(mddev); |
f45461e2 YK |
878 | } |
879 | ||
7e0adbfc CH |
880 | struct mdu_array_info_s; |
881 | struct mdu_disk_info_s; | |
882 | ||
1a6a0506 | 883 | extern int mdp_major; |
a022325a | 884 | extern struct workqueue_struct *md_bitmap_wq; |
d82fa81c | 885 | void md_autostart_arrays(int part); |
7e0adbfc CH |
886 | int md_set_array_info(struct mddev *mddev, struct mdu_array_info_s *info); |
887 | int md_add_new_disk(struct mddev *mddev, struct mdu_disk_info_s *info); | |
888 | int do_md_run(struct mddev *mddev); | |
889 | ||
890 | extern const struct block_device_operations md_fops; | |
d82fa81c | 891 | |
63fe0817 | 892 | #endif /* _MD_MD_H */ |