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> |
63fe0817 CH |
20 | #include <linux/kobject.h> |
21 | #include <linux/list.h> | |
22 | #include <linux/mm.h> | |
23 | #include <linux/mutex.h> | |
24 | #include <linux/timer.h> | |
25 | #include <linux/wait.h> | |
26 | #include <linux/workqueue.h> | |
edb39c9d | 27 | #include "md-cluster.h" |
9361401e | 28 | |
1da177e4 | 29 | #define MaxSector (~(sector_t)0) |
1da177e4 | 30 | |
2230dfe4 N |
31 | /* Bad block numbers are stored sorted in a single page. |
32 | * 64bits is used for each block or extent. | |
33 | * 54 bits are sector number, 9 bits are extent size, | |
34 | * 1 bit is an 'acknowledged' flag. | |
35 | */ | |
36 | #define MD_MAX_BADBLOCKS (PAGE_SIZE/8) | |
37 | ||
1da177e4 LT |
38 | /* |
39 | * MD's 'extended' device | |
40 | */ | |
3cb03002 | 41 | struct md_rdev { |
1da177e4 LT |
42 | struct list_head same_set; /* RAID devices within the same set */ |
43 | ||
dd8ac336 | 44 | sector_t sectors; /* Device size (in 512bytes sectors) */ |
fd01b88c | 45 | struct mddev *mddev; /* RAID array if running */ |
eea1bf38 | 46 | int last_events; /* IO event timestamp */ |
1da177e4 | 47 | |
a6ff7e08 JB |
48 | /* |
49 | * If meta_bdev is non-NULL, it means that a separate device is | |
50 | * being used to store the metadata (superblock/bitmap) which | |
51 | * would otherwise be contained on the same device as the data (bdev). | |
52 | */ | |
53 | struct block_device *meta_bdev; | |
1da177e4 LT |
54 | struct block_device *bdev; /* block device handle */ |
55 | ||
2699b672 | 56 | struct page *sb_page, *bb_page; |
1da177e4 | 57 | int sb_loaded; |
42543769 | 58 | __u64 sb_events; |
1da177e4 | 59 | sector_t data_offset; /* start of data in array */ |
c6563a8c | 60 | sector_t new_data_offset;/* only relevant while reshaping */ |
f72ffdd6 | 61 | sector_t sb_start; /* offset of the super block (in 512byte sectors) */ |
0002b271 | 62 | int sb_size; /* bytes in the superblock */ |
1da177e4 LT |
63 | int preferred_minor; /* autorun support */ |
64 | ||
86e6ffdd N |
65 | struct kobject kobj; |
66 | ||
1da177e4 LT |
67 | /* A device can be in one of three states based on two flags: |
68 | * Not working: faulty==1 in_sync==0 | |
69 | * Fully working: faulty==0 in_sync==1 | |
70 | * Working, but not | |
71 | * in sync with array | |
72 | * faulty==0 in_sync==0 | |
73 | * | |
74 | * It can never have faulty==1, in_sync==1 | |
75 | * This reduces the burden of testing multiple flags in many cases | |
76 | */ | |
1da177e4 | 77 | |
2d78f8c4 | 78 | unsigned long flags; /* bit set of 'enum flag_bits' bits. */ |
6bfe0b49 | 79 | wait_queue_head_t blocked_wait; |
8ddf9efe | 80 | |
1da177e4 LT |
81 | int desc_nr; /* descriptor index in the superblock */ |
82 | int raid_disk; /* role of device in array */ | |
e93f68a1 N |
83 | int new_raid_disk; /* role that the device will have in |
84 | * the array after a level-change completes. | |
85 | */ | |
41158c7e N |
86 | int saved_raid_disk; /* role that device used to have in the |
87 | * array and could again if we did a partial | |
88 | * resync from the bitmap | |
89 | */ | |
3069aa8d SL |
90 | union { |
91 | sector_t recovery_offset;/* If this device has been partially | |
5fd6c1dc N |
92 | * recovered, this is where we were |
93 | * up to. | |
94 | */ | |
3069aa8d SL |
95 | sector_t journal_tail; /* If this device is a journal device, |
96 | * this is the journal tail (journal | |
97 | * recovery start point) | |
98 | */ | |
99 | }; | |
1da177e4 LT |
100 | |
101 | atomic_t nr_pending; /* number of pending requests. | |
102 | * only maintained for arrays that | |
103 | * support hot removal | |
104 | */ | |
ba22dcbf N |
105 | atomic_t read_errors; /* number of consecutive read errors that |
106 | * we have tried to ignore. | |
107 | */ | |
1e50915f RB |
108 | struct timespec last_read_error; /* monotonic time since our |
109 | * last read error | |
110 | */ | |
4dbcdc75 N |
111 | atomic_t corrected_errors; /* number of corrected read errors, |
112 | * for reporting to userspace and storing | |
113 | * in superblock. | |
114 | */ | |
5792a285 | 115 | struct work_struct del_work; /* used for delayed sysfs removal */ |
3c0ee63a | 116 | |
324a56e1 | 117 | struct kernfs_node *sysfs_state; /* handle for 'state' |
3c0ee63a | 118 | * sysfs entry */ |
2230dfe4 N |
119 | |
120 | struct badblocks { | |
121 | int count; /* count of bad blocks */ | |
de393cde N |
122 | int unacked_exist; /* there probably are unacknowledged |
123 | * bad blocks. This is only cleared | |
124 | * when a read discovers none | |
125 | */ | |
2230dfe4 N |
126 | int shift; /* shift from sectors to block size |
127 | * a -ve shift means badblocks are | |
128 | * disabled.*/ | |
129 | u64 *page; /* badblock list */ | |
130 | int changed; | |
131 | seqlock_t lock; | |
2699b672 N |
132 | |
133 | sector_t sector; | |
134 | sector_t size; /* in sectors */ | |
2230dfe4 | 135 | } badblocks; |
1da177e4 | 136 | }; |
2d78f8c4 N |
137 | enum flag_bits { |
138 | Faulty, /* device is known to have a fault */ | |
139 | In_sync, /* device is in_sync with rest of array */ | |
8313b8e5 N |
140 | Bitmap_sync, /* ..actually, not quite In_sync. Need a |
141 | * bitmap-based recovery to get fully in sync | |
142 | */ | |
2d78f8c4 N |
143 | WriteMostly, /* Avoid reading if at all possible */ |
144 | AutoDetected, /* added by auto-detect */ | |
145 | Blocked, /* An error occurred but has not yet | |
146 | * been acknowledged by the metadata | |
147 | * handler, so don't allow writes | |
148 | * until it is cleared */ | |
149 | WriteErrorSeen, /* A write error has been seen on this | |
150 | * device | |
151 | */ | |
152 | FaultRecorded, /* Intermediate state for clearing | |
153 | * Blocked. The Fault is/will-be | |
154 | * recorded in the metadata, but that | |
155 | * metadata hasn't been stored safely | |
156 | * on disk yet. | |
157 | */ | |
158 | BlockedBadBlocks, /* A writer is blocked because they | |
159 | * found an unacknowledged bad-block. | |
160 | * This can safely be cleared at any | |
161 | * time, and the writer will re-check. | |
162 | * It may be set at any time, and at | |
163 | * worst the writer will timeout and | |
164 | * re-check. So setting it as | |
165 | * accurately as possible is good, but | |
166 | * not absolutely critical. | |
167 | */ | |
168 | WantReplacement, /* This device is a candidate to be | |
169 | * hot-replaced, either because it has | |
170 | * reported some faults, or because | |
171 | * of explicit request. | |
172 | */ | |
173 | Replacement, /* This device is a replacement for | |
174 | * a want_replacement device with same | |
175 | * raid_disk number. | |
176 | */ | |
1aee41f6 GR |
177 | Candidate, /* For clustered environments only: |
178 | * This device is seen locally but not | |
179 | * by the whole cluster | |
180 | */ | |
bac624f3 SL |
181 | Journal, /* This device is used as journal for |
182 | * raid-5/6. | |
183 | * Usually, this device should be faster | |
184 | * than other devices in the array | |
185 | */ | |
659b254f | 186 | ClusterRemove, |
2d78f8c4 | 187 | }; |
1da177e4 | 188 | |
2230dfe4 N |
189 | #define BB_LEN_MASK (0x00000000000001FFULL) |
190 | #define BB_OFFSET_MASK (0x7FFFFFFFFFFFFE00ULL) | |
191 | #define BB_ACK_MASK (0x8000000000000000ULL) | |
192 | #define BB_MAX_LEN 512 | |
193 | #define BB_OFFSET(x) (((x) & BB_OFFSET_MASK) >> 9) | |
194 | #define BB_LEN(x) (((x) & BB_LEN_MASK) + 1) | |
195 | #define BB_ACK(x) (!!((x) & BB_ACK_MASK)) | |
196 | #define BB_MAKE(a, l, ack) (((a)<<9) | ((l)-1) | ((u64)(!!(ack)) << 63)) | |
197 | ||
198 | extern int md_is_badblock(struct badblocks *bb, sector_t s, int sectors, | |
199 | sector_t *first_bad, int *bad_sectors); | |
3cb03002 | 200 | static inline int is_badblock(struct md_rdev *rdev, sector_t s, int sectors, |
2230dfe4 N |
201 | sector_t *first_bad, int *bad_sectors) |
202 | { | |
203 | if (unlikely(rdev->badblocks.count)) { | |
204 | int rv = md_is_badblock(&rdev->badblocks, rdev->data_offset + s, | |
205 | sectors, | |
206 | first_bad, bad_sectors); | |
207 | if (rv) | |
208 | *first_bad -= rdev->data_offset; | |
209 | return rv; | |
210 | } | |
211 | return 0; | |
212 | } | |
3cb03002 | 213 | extern int rdev_set_badblocks(struct md_rdev *rdev, sector_t s, int sectors, |
c6563a8c N |
214 | int is_new); |
215 | extern int rdev_clear_badblocks(struct md_rdev *rdev, sector_t s, int sectors, | |
216 | int is_new); | |
2230dfe4 N |
217 | extern void md_ack_all_badblocks(struct badblocks *bb); |
218 | ||
c4ce867f GR |
219 | struct md_cluster_info; |
220 | ||
fd01b88c | 221 | struct mddev { |
1da177e4 | 222 | void *private; |
84fc4b56 | 223 | struct md_personality *pers; |
1da177e4 LT |
224 | dev_t unit; |
225 | int md_minor; | |
7a0a5355 | 226 | struct list_head disks; |
850b2b42 N |
227 | unsigned long flags; |
228 | #define MD_CHANGE_DEVS 0 /* Some device status has changed */ | |
229 | #define MD_CHANGE_CLEAN 1 /* transition to or from 'clean' */ | |
070dc6dd | 230 | #define MD_CHANGE_PENDING 2 /* switch from 'clean' to 'active' in progress */ |
7a0a5355 | 231 | #define MD_UPDATE_SB_FLAGS (1 | 2 | 4) /* If these are set, md_update_sb needed */ |
9c81075f | 232 | #define MD_ARRAY_FIRST_USE 3 /* First use of array, needs initialization */ |
260fa034 N |
233 | #define MD_STILL_CLOSED 4 /* If set, then array has not been opened since |
234 | * md_ioctl checked on it. | |
235 | */ | |
bd18f646 | 236 | #define MD_JOURNAL_CLEAN 5 /* A raid with journal is already clean */ |
a97b7896 | 237 | #define MD_HAS_JOURNAL 6 /* The raid array has journal feature set */ |
850b2b42 | 238 | |
409c57f3 N |
239 | int suspended; |
240 | atomic_t active_io; | |
1da177e4 | 241 | int ro; |
bb4f1e9d N |
242 | int sysfs_active; /* set when sysfs deletes |
243 | * are happening, so run/ | |
244 | * takeover/stop are not safe | |
245 | */ | |
7a0a5355 | 246 | int ready; /* See when safe to pass |
0ca69886 | 247 | * IO requests down */ |
1da177e4 LT |
248 | struct gendisk *gendisk; |
249 | ||
eae1701f | 250 | struct kobject kobj; |
d3374825 N |
251 | int hold_active; |
252 | #define UNTIL_IOCTL 1 | |
efeb53c0 | 253 | #define UNTIL_STOP 2 |
eae1701f | 254 | |
1da177e4 LT |
255 | /* Superblock information */ |
256 | int major_version, | |
257 | minor_version, | |
258 | patch_version; | |
259 | int persistent; | |
f72ffdd6 | 260 | int external; /* metadata is |
e691063a N |
261 | * managed externally */ |
262 | char metadata_type[17]; /* externally set*/ | |
9d8f0363 | 263 | int chunk_sectors; |
1da177e4 LT |
264 | time_t ctime, utime; |
265 | int level, layout; | |
d9d166c2 | 266 | char clevel[16]; |
1da177e4 LT |
267 | int raid_disks; |
268 | int max_disks; | |
f72ffdd6 | 269 | sector_t dev_sectors; /* used size of |
58c0fed4 | 270 | * component devices */ |
f233ea5c | 271 | sector_t array_sectors; /* exported array size */ |
b522adcd DW |
272 | int external_size; /* size managed |
273 | * externally */ | |
1da177e4 | 274 | __u64 events; |
a8707c08 N |
275 | /* If the last 'event' was simply a clean->dirty transition, and |
276 | * we didn't write it to the spares, then it is safe and simple | |
277 | * to just decrement the event count on a dirty->clean transition. | |
278 | * So we record that possibility here. | |
279 | */ | |
280 | int can_decrease_events; | |
1da177e4 LT |
281 | |
282 | char uuid[16]; | |
283 | ||
f6705578 N |
284 | /* If the array is being reshaped, we need to record the |
285 | * new shape and an indication of where we are up to. | |
286 | * This is written to the superblock. | |
287 | * If reshape_position is MaxSector, then no reshape is happening (yet). | |
288 | */ | |
289 | sector_t reshape_position; | |
664e7c41 AN |
290 | int delta_disks, new_level, new_layout; |
291 | int new_chunk_sectors; | |
2c810cdd | 292 | int reshape_backwards; |
f6705578 | 293 | |
2b8bf345 N |
294 | struct md_thread *thread; /* management thread */ |
295 | struct md_thread *sync_thread; /* doing resync or reconstruct */ | |
c4a39551 JB |
296 | |
297 | /* 'last_sync_action' is initialized to "none". It is set when a | |
298 | * sync operation (i.e "data-check", "requested-resync", "resync", | |
299 | * "recovery", or "reshape") is started. It holds this value even | |
300 | * when the sync thread is "frozen" (interrupted) or "idle" (stopped | |
301 | * or finished). It is overwritten when a new sync operation is begun. | |
302 | */ | |
303 | char *last_sync_action; | |
ff4e8d9a | 304 | sector_t curr_resync; /* last block scheduled */ |
97e4f42d N |
305 | /* As resync requests can complete out of order, we cannot easily track |
306 | * how much resync has been completed. So we occasionally pause until | |
307 | * everything completes, then set curr_resync_completed to curr_resync. | |
308 | * As such it may be well behind the real resync mark, but it is a value | |
309 | * we are certain of. | |
310 | */ | |
311 | sector_t curr_resync_completed; | |
1da177e4 LT |
312 | unsigned long resync_mark; /* a recent timestamp */ |
313 | sector_t resync_mark_cnt;/* blocks written at resync_mark */ | |
ff4e8d9a | 314 | sector_t curr_mark_cnt; /* blocks scheduled now */ |
1da177e4 LT |
315 | |
316 | sector_t resync_max_sectors; /* may be set by personality */ | |
9d88883e | 317 | |
7f7583d4 | 318 | atomic64_t resync_mismatches; /* count of sectors where |
9d88883e N |
319 | * parity/replica mismatch found |
320 | */ | |
e464eafd N |
321 | |
322 | /* allow user-space to request suspension of IO to regions of the array */ | |
323 | sector_t suspend_lo; | |
324 | sector_t suspend_hi; | |
88202a0c N |
325 | /* if zero, use the system-wide default */ |
326 | int sync_speed_min; | |
327 | int sync_speed_max; | |
328 | ||
90b08710 BS |
329 | /* resync even though the same disks are shared among md-devices */ |
330 | int parallel_resync; | |
331 | ||
6ff8d8ec | 332 | int ok_start_degraded; |
f72ffdd6 | 333 | /* recovery/resync flags |
1da177e4 LT |
334 | * NEEDED: we might need to start a resync/recover |
335 | * RUNNING: a thread is running, or about to be started | |
336 | * SYNC: actually doing a resync, not a recovery | |
72a23c21 | 337 | * RECOVER: doing recovery, or need to try it. |
dfc70645 | 338 | * INTR: resync needs to be aborted for some reason |
1da177e4 | 339 | * DONE: thread is done and is waiting to be reaped |
24dd469d | 340 | * REQUEST: user-space has requested a sync (used with SYNC) |
411c9403 | 341 | * CHECK: user-space request for check-only, no repair |
ccfcc3c1 | 342 | * RESHAPE: A reshape is happening |
0a19caab | 343 | * ERROR: sync-action interrupted because io-error |
ccfcc3c1 N |
344 | * |
345 | * If neither SYNC or RESHAPE are set, then it is a recovery. | |
1da177e4 LT |
346 | */ |
347 | #define MD_RECOVERY_RUNNING 0 | |
348 | #define MD_RECOVERY_SYNC 1 | |
72a23c21 | 349 | #define MD_RECOVERY_RECOVER 2 |
1da177e4 LT |
350 | #define MD_RECOVERY_INTR 3 |
351 | #define MD_RECOVERY_DONE 4 | |
352 | #define MD_RECOVERY_NEEDED 5 | |
24dd469d N |
353 | #define MD_RECOVERY_REQUESTED 6 |
354 | #define MD_RECOVERY_CHECK 7 | |
ccfcc3c1 | 355 | #define MD_RECOVERY_RESHAPE 8 |
5fd6c1dc | 356 | #define MD_RECOVERY_FROZEN 9 |
0a19caab | 357 | #define MD_RECOVERY_ERROR 10 |
5fd6c1dc | 358 | |
1da177e4 | 359 | unsigned long recovery; |
5389042f N |
360 | /* If a RAID personality determines that recovery (of a particular |
361 | * device) will fail due to a read error on the source device, it | |
362 | * takes a copy of this number and does not attempt recovery again | |
363 | * until this number changes. | |
364 | */ | |
365 | int recovery_disabled; | |
1da177e4 LT |
366 | |
367 | int in_sync; /* know to not need resync */ | |
c8c00a69 N |
368 | /* 'open_mutex' avoids races between 'md_open' and 'do_md_stop', so |
369 | * that we are never stopping an array while it is open. | |
370 | * 'reconfig_mutex' protects all other reconfiguration. | |
371 | * These locks are separate due to conflicting interactions | |
372 | * with bdev->bd_mutex. | |
373 | * Lock ordering is: | |
374 | * reconfig_mutex -> bd_mutex : e.g. do_md_run -> revalidate_disk | |
375 | * bd_mutex -> open_mutex: e.g. __blkdev_get -> md_open | |
376 | */ | |
377 | struct mutex open_mutex; | |
df5b89b3 | 378 | struct mutex reconfig_mutex; |
f2ea68cf N |
379 | atomic_t active; /* general refcount */ |
380 | atomic_t openers; /* number of active opens */ | |
1da177e4 | 381 | |
f0b4f7e2 N |
382 | int changed; /* True if we might need to |
383 | * reread partition info */ | |
1da177e4 LT |
384 | int degraded; /* whether md should consider |
385 | * adding a spare | |
386 | */ | |
387 | ||
388 | atomic_t recovery_active; /* blocks scheduled, but not written */ | |
389 | wait_queue_head_t recovery_wait; | |
390 | sector_t recovery_cp; | |
5e96ee65 NB |
391 | sector_t resync_min; /* user requested sync |
392 | * starts here */ | |
c6207277 N |
393 | sector_t resync_max; /* resync should pause |
394 | * when it gets here */ | |
06d91a5f | 395 | |
324a56e1 | 396 | struct kernfs_node *sysfs_state; /* handle for 'array_state' |
b62b7590 N |
397 | * file in sysfs. |
398 | */ | |
324a56e1 | 399 | struct kernfs_node *sysfs_action; /* handle for 'sync_action' */ |
b62b7590 | 400 | |
d3374825 N |
401 | struct work_struct del_work; /* used for delayed sysfs removal */ |
402 | ||
85572d7c N |
403 | /* "lock" protects: |
404 | * flush_bio transition from NULL to !NULL | |
405 | * rdev superblocks, events | |
406 | * clearing MD_CHANGE_* | |
407 | * in_sync - and related safemode and MD_CHANGE changes | |
36d091f4 | 408 | * pers (also protected by reconfig_mutex and pending IO). |
978a7a47 | 409 | * clearing ->bitmap |
4af1a041 | 410 | * clearing ->bitmap_info.file |
23da422b N |
411 | * changing ->resync_{min,max} |
412 | * setting MD_RECOVERY_RUNNING (which interacts with resync_{min,max}) | |
85572d7c N |
413 | */ |
414 | spinlock_t lock; | |
3d310eb7 | 415 | wait_queue_head_t sb_wait; /* for waiting on superblock updates */ |
7bfa19f2 | 416 | atomic_t pending_writes; /* number of active superblock writes */ |
06d91a5f | 417 | |
1da177e4 LT |
418 | unsigned int safemode; /* if set, update "clean" superblock |
419 | * when no writes pending. | |
f72ffdd6 | 420 | */ |
1da177e4 LT |
421 | unsigned int safemode_delay; |
422 | struct timer_list safemode_timer; | |
f72ffdd6 | 423 | atomic_t writes_pending; |
165125e1 | 424 | struct request_queue *queue; /* for plugging ... */ |
1da177e4 | 425 | |
f72ffdd6 | 426 | struct bitmap *bitmap; /* the bitmap for the device */ |
c3d9714e N |
427 | struct { |
428 | struct file *file; /* the bitmap file */ | |
f6af949c | 429 | loff_t offset; /* offset from superblock of |
c3d9714e N |
430 | * start of bitmap. May be |
431 | * negative, but not '0' | |
f6af949c | 432 | * For external metadata, offset |
f72ffdd6 | 433 | * from start of device. |
c3d9714e | 434 | */ |
6409bb05 | 435 | unsigned long space; /* space available at this offset */ |
f6af949c | 436 | loff_t default_offset; /* this is the offset to use when |
c3d9714e N |
437 | * hot-adding a bitmap. It should |
438 | * eventually be settable by sysfs. | |
439 | */ | |
6409bb05 N |
440 | unsigned long default_space; /* space available at |
441 | * default offset */ | |
c3d9714e | 442 | struct mutex mutex; |
42a04b50 | 443 | unsigned long chunksize; |
ac2f40be | 444 | unsigned long daemon_sleep; /* how many jiffies between updates? */ |
42a04b50 | 445 | unsigned long max_write_behind; /* write-behind mode */ |
ece5cff0 | 446 | int external; |
c4ce867f | 447 | int nodes; /* Maximum number of nodes in the cluster */ |
cf921cc1 | 448 | char cluster_name[64]; /* Name of the cluster */ |
c3d9714e | 449 | } bitmap_info; |
32a7627c | 450 | |
f72ffdd6 | 451 | atomic_t max_corr_read_errors; /* max read retries */ |
1da177e4 | 452 | struct list_head all_mddevs; |
a2826aa9 | 453 | |
a64c876f | 454 | struct attribute_group *to_remove; |
252ac522 | 455 | |
a167f663 N |
456 | struct bio_set *bio_set; |
457 | ||
e9c7469b TH |
458 | /* Generic flush handling. |
459 | * The last to finish preflush schedules a worker to submit | |
460 | * the rest of the request (without the REQ_FLUSH flag). | |
a2826aa9 | 461 | */ |
e9c7469b | 462 | struct bio *flush_bio; |
a2826aa9 | 463 | atomic_t flush_pending; |
e9c7469b | 464 | struct work_struct flush_work; |
768a418d | 465 | struct work_struct event_work; /* used by dm to report failure event */ |
fd01b88c | 466 | void (*sync_super)(struct mddev *mddev, struct md_rdev *rdev); |
c4ce867f | 467 | struct md_cluster_info *cluster_info; |
1da177e4 LT |
468 | }; |
469 | ||
5c47daf6 N |
470 | static inline int __must_check mddev_lock(struct mddev *mddev) |
471 | { | |
472 | return mutex_lock_interruptible(&mddev->reconfig_mutex); | |
473 | } | |
474 | ||
475 | /* Sometimes we need to take the lock in a situation where | |
476 | * failure due to interrupts is not acceptable. | |
477 | */ | |
478 | static inline void mddev_lock_nointr(struct mddev *mddev) | |
479 | { | |
480 | mutex_lock(&mddev->reconfig_mutex); | |
481 | } | |
482 | ||
483 | static inline int mddev_is_locked(struct mddev *mddev) | |
484 | { | |
485 | return mutex_is_locked(&mddev->reconfig_mutex); | |
486 | } | |
487 | ||
488 | static inline int mddev_trylock(struct mddev *mddev) | |
489 | { | |
490 | return mutex_trylock(&mddev->reconfig_mutex); | |
491 | } | |
492 | extern void mddev_unlock(struct mddev *mddev); | |
493 | ||
1da177e4 LT |
494 | static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors) |
495 | { | |
f72ffdd6 | 496 | atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io); |
1da177e4 LT |
497 | } |
498 | ||
84fc4b56 | 499 | struct md_personality |
1da177e4 LT |
500 | { |
501 | char *name; | |
2604b703 N |
502 | int level; |
503 | struct list_head list; | |
1da177e4 | 504 | struct module *owner; |
b4fdcb02 | 505 | void (*make_request)(struct mddev *mddev, struct bio *bio); |
fd01b88c | 506 | int (*run)(struct mddev *mddev); |
afa0f557 | 507 | void (*free)(struct mddev *mddev, void *priv); |
fd01b88c | 508 | void (*status)(struct seq_file *seq, struct mddev *mddev); |
1da177e4 | 509 | /* error_handler must set ->faulty and clear ->in_sync |
f72ffdd6 | 510 | * if appropriate, and should abort recovery if needed |
1da177e4 | 511 | */ |
fd01b88c N |
512 | void (*error_handler)(struct mddev *mddev, struct md_rdev *rdev); |
513 | int (*hot_add_disk) (struct mddev *mddev, struct md_rdev *rdev); | |
b8321b68 | 514 | int (*hot_remove_disk) (struct mddev *mddev, struct md_rdev *rdev); |
fd01b88c | 515 | int (*spare_active) (struct mddev *mddev); |
09314799 | 516 | sector_t (*sync_request)(struct mddev *mddev, sector_t sector_nr, int *skipped); |
fd01b88c N |
517 | int (*resize) (struct mddev *mddev, sector_t sectors); |
518 | sector_t (*size) (struct mddev *mddev, sector_t sectors, int raid_disks); | |
519 | int (*check_reshape) (struct mddev *mddev); | |
520 | int (*start_reshape) (struct mddev *mddev); | |
521 | void (*finish_reshape) (struct mddev *mddev); | |
36fa3063 N |
522 | /* quiesce moves between quiescence states |
523 | * 0 - fully active | |
524 | * 1 - no new requests allowed | |
525 | * others - reserved | |
526 | */ | |
fd01b88c | 527 | void (*quiesce) (struct mddev *mddev, int state); |
245f46c2 N |
528 | /* takeover is used to transition an array from one |
529 | * personality to another. The new personality must be able | |
530 | * to handle the data in the current layout. | |
531 | * e.g. 2drive raid1 -> 2drive raid5 | |
532 | * ndrive raid5 -> degraded n+1drive raid6 with special layout | |
533 | * If the takeover succeeds, a new 'private' structure is returned. | |
534 | * This needs to be installed and then ->run used to activate the | |
535 | * array. | |
536 | */ | |
fd01b88c | 537 | void *(*takeover) (struct mddev *mddev); |
5c675f83 N |
538 | /* congested implements bdi.congested_fn(). |
539 | * Will not be called while array is 'suspended' */ | |
540 | int (*congested)(struct mddev *mddev, int bits); | |
1da177e4 LT |
541 | }; |
542 | ||
007583c9 N |
543 | struct md_sysfs_entry { |
544 | struct attribute attr; | |
fd01b88c N |
545 | ssize_t (*show)(struct mddev *, char *); |
546 | ssize_t (*store)(struct mddev *, const char *, size_t); | |
007583c9 | 547 | }; |
43a70507 | 548 | extern struct attribute_group md_bitmap_group; |
007583c9 | 549 | |
324a56e1 | 550 | static inline struct kernfs_node *sysfs_get_dirent_safe(struct kernfs_node *sd, char *name) |
00bcb4ac N |
551 | { |
552 | if (sd) | |
388975cc | 553 | return sysfs_get_dirent(sd, name); |
00bcb4ac N |
554 | return sd; |
555 | } | |
324a56e1 | 556 | static inline void sysfs_notify_dirent_safe(struct kernfs_node *sd) |
00bcb4ac N |
557 | { |
558 | if (sd) | |
559 | sysfs_notify_dirent(sd); | |
560 | } | |
561 | ||
fd01b88c | 562 | static inline char * mdname (struct mddev * mddev) |
1da177e4 LT |
563 | { |
564 | return mddev->gendisk ? mddev->gendisk->disk_name : "mdX"; | |
565 | } | |
566 | ||
fd01b88c | 567 | static inline int sysfs_link_rdev(struct mddev *mddev, struct md_rdev *rdev) |
36fad858 NK |
568 | { |
569 | char nm[20]; | |
9b15603d SL |
570 | if (!test_bit(Replacement, &rdev->flags) && |
571 | !test_bit(Journal, &rdev->flags) && | |
572 | mddev->kobj.sd) { | |
2d78f8c4 N |
573 | sprintf(nm, "rd%d", rdev->raid_disk); |
574 | return sysfs_create_link(&mddev->kobj, &rdev->kobj, nm); | |
575 | } else | |
576 | return 0; | |
36fad858 NK |
577 | } |
578 | ||
fd01b88c | 579 | static inline void sysfs_unlink_rdev(struct mddev *mddev, struct md_rdev *rdev) |
36fad858 NK |
580 | { |
581 | char nm[20]; | |
9b15603d SL |
582 | if (!test_bit(Replacement, &rdev->flags) && |
583 | !test_bit(Journal, &rdev->flags) && | |
584 | mddev->kobj.sd) { | |
2d78f8c4 N |
585 | sprintf(nm, "rd%d", rdev->raid_disk); |
586 | sysfs_remove_link(&mddev->kobj, nm); | |
587 | } | |
36fad858 NK |
588 | } |
589 | ||
1da177e4 LT |
590 | /* |
591 | * iterates through some rdev ringlist. It's safe to remove the | |
592 | * current 'rdev'. Dont touch 'tmp' though. | |
593 | */ | |
159ec1fc CR |
594 | #define rdev_for_each_list(rdev, tmp, head) \ |
595 | list_for_each_entry_safe(rdev, tmp, head, same_set) | |
596 | ||
1da177e4 LT |
597 | /* |
598 | * iterates through the 'same array disks' ringlist | |
599 | */ | |
dafb20fa N |
600 | #define rdev_for_each(rdev, mddev) \ |
601 | list_for_each_entry(rdev, &((mddev)->disks), same_set) | |
602 | ||
603 | #define rdev_for_each_safe(rdev, tmp, mddev) \ | |
159ec1fc | 604 | list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set) |
1da177e4 | 605 | |
4b80991c N |
606 | #define rdev_for_each_rcu(rdev, mddev) \ |
607 | list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set) | |
608 | ||
2b8bf345 | 609 | struct md_thread { |
4ed8731d | 610 | void (*run) (struct md_thread *thread); |
fd01b88c | 611 | struct mddev *mddev; |
1da177e4 | 612 | wait_queue_head_t wqueue; |
f72ffdd6 | 613 | unsigned long flags; |
1da177e4 | 614 | struct task_struct *tsk; |
32a7627c | 615 | unsigned long timeout; |
4ed8731d | 616 | void *private; |
2b8bf345 | 617 | }; |
1da177e4 LT |
618 | |
619 | #define THREAD_WAKEUP 0 | |
620 | ||
1345b1d8 N |
621 | static inline void safe_put_page(struct page *p) |
622 | { | |
623 | if (p) put_page(p); | |
624 | } | |
625 | ||
84fc4b56 N |
626 | extern int register_md_personality(struct md_personality *p); |
627 | extern int unregister_md_personality(struct md_personality *p); | |
edb39c9d GR |
628 | extern int register_md_cluster_operations(struct md_cluster_operations *ops, |
629 | struct module *module); | |
630 | extern int unregister_md_cluster_operations(void); | |
631 | extern int md_setup_cluster(struct mddev *mddev, int nodes); | |
632 | extern void md_cluster_stop(struct mddev *mddev); | |
2b8bf345 | 633 | extern struct md_thread *md_register_thread( |
4ed8731d | 634 | void (*run)(struct md_thread *thread), |
2b8bf345 N |
635 | struct mddev *mddev, |
636 | const char *name); | |
637 | extern void md_unregister_thread(struct md_thread **threadp); | |
638 | extern void md_wakeup_thread(struct md_thread *thread); | |
fd01b88c | 639 | extern void md_check_recovery(struct mddev *mddev); |
a91d5ac0 | 640 | extern void md_reap_sync_thread(struct mddev *mddev); |
fd01b88c N |
641 | extern void md_write_start(struct mddev *mddev, struct bio *bi); |
642 | extern void md_write_end(struct mddev *mddev); | |
643 | extern void md_done_sync(struct mddev *mddev, int blocks, int ok); | |
644 | extern void md_error(struct mddev *mddev, struct md_rdev *rdev); | |
c6563a8c | 645 | extern void md_finish_reshape(struct mddev *mddev); |
fd01b88c N |
646 | |
647 | extern int mddev_congested(struct mddev *mddev, int bits); | |
648 | extern void md_flush_request(struct mddev *mddev, struct bio *bio); | |
649 | extern void md_super_write(struct mddev *mddev, struct md_rdev *rdev, | |
92022950 | 650 | sector_t sector, int size, struct page *page); |
fd01b88c | 651 | extern void md_super_wait(struct mddev *mddev); |
f72ffdd6 | 652 | extern int sync_page_io(struct md_rdev *rdev, sector_t sector, int size, |
ccebd4c4 | 653 | struct page *page, int rw, bool metadata_op); |
4ed8731d | 654 | extern void md_do_sync(struct md_thread *thread); |
fd01b88c N |
655 | extern void md_new_event(struct mddev *mddev); |
656 | extern int md_allow_write(struct mddev *mddev); | |
657 | extern void md_wait_for_blocked_rdev(struct md_rdev *rdev, struct mddev *mddev); | |
658 | extern void md_set_array_sectors(struct mddev *mddev, sector_t array_sectors); | |
659 | extern int md_check_no_bitmap(struct mddev *mddev); | |
660 | extern int md_integrity_register(struct mddev *mddev); | |
661 | extern void md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev); | |
72e02075 | 662 | extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale); |
63fe0817 | 663 | |
fd01b88c N |
664 | extern void mddev_init(struct mddev *mddev); |
665 | extern int md_run(struct mddev *mddev); | |
666 | extern void md_stop(struct mddev *mddev); | |
667 | extern void md_stop_writes(struct mddev *mddev); | |
3cb03002 | 668 | extern int md_rdev_init(struct md_rdev *rdev); |
545c8795 | 669 | extern void md_rdev_clear(struct md_rdev *rdev); |
63fe0817 | 670 | |
fd01b88c N |
671 | extern void mddev_suspend(struct mddev *mddev); |
672 | extern void mddev_resume(struct mddev *mddev); | |
a167f663 | 673 | extern struct bio *bio_clone_mddev(struct bio *bio, gfp_t gfp_mask, |
fd01b88c | 674 | struct mddev *mddev); |
a167f663 | 675 | extern struct bio *bio_alloc_mddev(gfp_t gfp_mask, int nr_iovecs, |
fd01b88c | 676 | struct mddev *mddev); |
9cbb1750 | 677 | |
74018dc3 | 678 | extern void md_unplug(struct blk_plug_cb *cb, bool from_schedule); |
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); |
9cbb1750 N |
683 | static inline int mddev_check_plugged(struct mddev *mddev) |
684 | { | |
685 | return !!blk_check_plugged(md_unplug, mddev, | |
686 | sizeof(struct blk_plug_cb)); | |
687 | } | |
dfe15ac1 HR |
688 | |
689 | static inline void rdev_dec_pending(struct md_rdev *rdev, struct mddev *mddev) | |
690 | { | |
691 | int faulty = test_bit(Faulty, &rdev->flags); | |
692 | if (atomic_dec_and_test(&rdev->nr_pending) && faulty) { | |
693 | set_bit(MD_RECOVERY_NEEDED, &mddev->recovery); | |
694 | md_wakeup_thread(mddev->thread); | |
695 | } | |
696 | } | |
697 | ||
edb39c9d | 698 | extern struct md_cluster_operations *md_cluster_ops; |
c4ce867f GR |
699 | static inline int mddev_is_clustered(struct mddev *mddev) |
700 | { | |
701 | return mddev->cluster_info && mddev->bitmap_info.nodes > 1; | |
702 | } | |
63fe0817 | 703 | #endif /* _MD_MD_H */ |