Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | md_k.h : kernel internal structure of the Linux MD driver | |
3 | Copyright (C) 1996-98 Ingo Molnar, Gadi Oxman | |
4 | ||
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. | |
9 | ||
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 | |
12 | Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. | |
13 | */ | |
14 | ||
63fe0817 CH |
15 | #ifndef _MD_MD_H |
16 | #define _MD_MD_H | |
17 | ||
18 | #include <linux/blkdev.h> | |
19 | #include <linux/kobject.h> | |
20 | #include <linux/list.h> | |
21 | #include <linux/mm.h> | |
22 | #include <linux/mutex.h> | |
23 | #include <linux/timer.h> | |
24 | #include <linux/wait.h> | |
25 | #include <linux/workqueue.h> | |
9361401e | 26 | |
1da177e4 | 27 | #define MaxSector (~(sector_t)0) |
1da177e4 | 28 | |
1da177e4 LT |
29 | typedef struct mddev_s mddev_t; |
30 | typedef struct mdk_rdev_s mdk_rdev_t; | |
31 | ||
1da177e4 LT |
32 | /* |
33 | * MD's 'extended' device | |
34 | */ | |
35 | struct mdk_rdev_s | |
36 | { | |
37 | struct list_head same_set; /* RAID devices within the same set */ | |
38 | ||
dd8ac336 | 39 | sector_t sectors; /* Device size (in 512bytes sectors) */ |
1da177e4 | 40 | mddev_t *mddev; /* RAID array if running */ |
eea1bf38 | 41 | int last_events; /* IO event timestamp */ |
1da177e4 LT |
42 | |
43 | struct block_device *bdev; /* block device handle */ | |
44 | ||
45 | struct page *sb_page; | |
46 | int sb_loaded; | |
42543769 | 47 | __u64 sb_events; |
1da177e4 | 48 | sector_t data_offset; /* start of data in array */ |
0f420358 | 49 | sector_t sb_start; /* offset of the super block (in 512byte sectors) */ |
0002b271 | 50 | int sb_size; /* bytes in the superblock */ |
1da177e4 LT |
51 | int preferred_minor; /* autorun support */ |
52 | ||
86e6ffdd N |
53 | struct kobject kobj; |
54 | ||
1da177e4 LT |
55 | /* A device can be in one of three states based on two flags: |
56 | * Not working: faulty==1 in_sync==0 | |
57 | * Fully working: faulty==0 in_sync==1 | |
58 | * Working, but not | |
59 | * in sync with array | |
60 | * faulty==0 in_sync==0 | |
61 | * | |
62 | * It can never have faulty==1, in_sync==1 | |
63 | * This reduces the burden of testing multiple flags in many cases | |
64 | */ | |
1da177e4 | 65 | |
b2d444d7 N |
66 | unsigned long flags; |
67 | #define Faulty 1 /* device is known to have a fault */ | |
68 | #define In_sync 2 /* device is in_sync with rest of array */ | |
8ddf9efe | 69 | #define WriteMostly 4 /* Avoid reading if at all possible */ |
a9701a30 | 70 | #define BarriersNotsupp 5 /* BIO_RW_BARRIER is not supported */ |
c5d79adb N |
71 | #define AllReserved 6 /* If whole device is reserved for |
72 | * one array */ | |
d0fae18f | 73 | #define AutoDetected 7 /* added by auto-detect */ |
6bfe0b49 DW |
74 | #define Blocked 8 /* An error occured on an externally |
75 | * managed array, don't allow writes | |
76 | * until it is cleared */ | |
77 | wait_queue_head_t blocked_wait; | |
8ddf9efe | 78 | |
1da177e4 LT |
79 | int desc_nr; /* descriptor index in the superblock */ |
80 | int raid_disk; /* role of device in array */ | |
41158c7e N |
81 | int saved_raid_disk; /* role that device used to have in the |
82 | * array and could again if we did a partial | |
83 | * resync from the bitmap | |
84 | */ | |
5fd6c1dc N |
85 | sector_t recovery_offset;/* If this device has been partially |
86 | * recovered, this is where we were | |
87 | * up to. | |
88 | */ | |
1da177e4 LT |
89 | |
90 | atomic_t nr_pending; /* number of pending requests. | |
91 | * only maintained for arrays that | |
92 | * support hot removal | |
93 | */ | |
ba22dcbf N |
94 | atomic_t read_errors; /* number of consecutive read errors that |
95 | * we have tried to ignore. | |
96 | */ | |
1e50915f RB |
97 | struct timespec last_read_error; /* monotonic time since our |
98 | * last read error | |
99 | */ | |
4dbcdc75 N |
100 | atomic_t corrected_errors; /* number of corrected read errors, |
101 | * for reporting to userspace and storing | |
102 | * in superblock. | |
103 | */ | |
5792a285 | 104 | struct work_struct del_work; /* used for delayed sysfs removal */ |
3c0ee63a N |
105 | |
106 | struct sysfs_dirent *sysfs_state; /* handle for 'state' | |
107 | * sysfs entry */ | |
1da177e4 LT |
108 | }; |
109 | ||
1da177e4 LT |
110 | struct mddev_s |
111 | { | |
112 | void *private; | |
2604b703 | 113 | struct mdk_personality *pers; |
1da177e4 LT |
114 | dev_t unit; |
115 | int md_minor; | |
116 | struct list_head disks; | |
850b2b42 N |
117 | unsigned long flags; |
118 | #define MD_CHANGE_DEVS 0 /* Some device status has changed */ | |
119 | #define MD_CHANGE_CLEAN 1 /* transition to or from 'clean' */ | |
120 | #define MD_CHANGE_PENDING 2 /* superblock update in progress */ | |
121 | ||
409c57f3 N |
122 | int suspended; |
123 | atomic_t active_io; | |
1da177e4 LT |
124 | int ro; |
125 | ||
126 | struct gendisk *gendisk; | |
127 | ||
eae1701f | 128 | struct kobject kobj; |
d3374825 N |
129 | int hold_active; |
130 | #define UNTIL_IOCTL 1 | |
efeb53c0 | 131 | #define UNTIL_STOP 2 |
eae1701f | 132 | |
1da177e4 LT |
133 | /* Superblock information */ |
134 | int major_version, | |
135 | minor_version, | |
136 | patch_version; | |
137 | int persistent; | |
e691063a N |
138 | int external; /* metadata is |
139 | * managed externally */ | |
140 | char metadata_type[17]; /* externally set*/ | |
9d8f0363 | 141 | int chunk_sectors; |
1da177e4 LT |
142 | time_t ctime, utime; |
143 | int level, layout; | |
d9d166c2 | 144 | char clevel[16]; |
1da177e4 LT |
145 | int raid_disks; |
146 | int max_disks; | |
58c0fed4 AN |
147 | sector_t dev_sectors; /* used size of |
148 | * component devices */ | |
f233ea5c | 149 | sector_t array_sectors; /* exported array size */ |
b522adcd DW |
150 | int external_size; /* size managed |
151 | * externally */ | |
1da177e4 LT |
152 | __u64 events; |
153 | ||
154 | char uuid[16]; | |
155 | ||
f6705578 N |
156 | /* If the array is being reshaped, we need to record the |
157 | * new shape and an indication of where we are up to. | |
158 | * This is written to the superblock. | |
159 | * If reshape_position is MaxSector, then no reshape is happening (yet). | |
160 | */ | |
161 | sector_t reshape_position; | |
664e7c41 AN |
162 | int delta_disks, new_level, new_layout; |
163 | int new_chunk_sectors; | |
f6705578 | 164 | |
1da177e4 LT |
165 | struct mdk_thread_s *thread; /* management thread */ |
166 | struct mdk_thread_s *sync_thread; /* doing resync or reconstruct */ | |
ff4e8d9a | 167 | sector_t curr_resync; /* last block scheduled */ |
97e4f42d N |
168 | /* As resync requests can complete out of order, we cannot easily track |
169 | * how much resync has been completed. So we occasionally pause until | |
170 | * everything completes, then set curr_resync_completed to curr_resync. | |
171 | * As such it may be well behind the real resync mark, but it is a value | |
172 | * we are certain of. | |
173 | */ | |
174 | sector_t curr_resync_completed; | |
1da177e4 LT |
175 | unsigned long resync_mark; /* a recent timestamp */ |
176 | sector_t resync_mark_cnt;/* blocks written at resync_mark */ | |
ff4e8d9a | 177 | sector_t curr_mark_cnt; /* blocks scheduled now */ |
1da177e4 LT |
178 | |
179 | sector_t resync_max_sectors; /* may be set by personality */ | |
9d88883e N |
180 | |
181 | sector_t resync_mismatches; /* count of sectors where | |
182 | * parity/replica mismatch found | |
183 | */ | |
e464eafd N |
184 | |
185 | /* allow user-space to request suspension of IO to regions of the array */ | |
186 | sector_t suspend_lo; | |
187 | sector_t suspend_hi; | |
88202a0c N |
188 | /* if zero, use the system-wide default */ |
189 | int sync_speed_min; | |
190 | int sync_speed_max; | |
191 | ||
90b08710 BS |
192 | /* resync even though the same disks are shared among md-devices */ |
193 | int parallel_resync; | |
194 | ||
6ff8d8ec | 195 | int ok_start_degraded; |
1da177e4 LT |
196 | /* recovery/resync flags |
197 | * NEEDED: we might need to start a resync/recover | |
198 | * RUNNING: a thread is running, or about to be started | |
199 | * SYNC: actually doing a resync, not a recovery | |
72a23c21 | 200 | * RECOVER: doing recovery, or need to try it. |
dfc70645 | 201 | * INTR: resync needs to be aborted for some reason |
1da177e4 | 202 | * DONE: thread is done and is waiting to be reaped |
24dd469d | 203 | * REQUEST: user-space has requested a sync (used with SYNC) |
411c9403 | 204 | * CHECK: user-space request for check-only, no repair |
ccfcc3c1 N |
205 | * RESHAPE: A reshape is happening |
206 | * | |
207 | * If neither SYNC or RESHAPE are set, then it is a recovery. | |
1da177e4 LT |
208 | */ |
209 | #define MD_RECOVERY_RUNNING 0 | |
210 | #define MD_RECOVERY_SYNC 1 | |
72a23c21 | 211 | #define MD_RECOVERY_RECOVER 2 |
1da177e4 LT |
212 | #define MD_RECOVERY_INTR 3 |
213 | #define MD_RECOVERY_DONE 4 | |
214 | #define MD_RECOVERY_NEEDED 5 | |
24dd469d N |
215 | #define MD_RECOVERY_REQUESTED 6 |
216 | #define MD_RECOVERY_CHECK 7 | |
ccfcc3c1 | 217 | #define MD_RECOVERY_RESHAPE 8 |
5fd6c1dc N |
218 | #define MD_RECOVERY_FROZEN 9 |
219 | ||
1da177e4 | 220 | unsigned long recovery; |
4044ba58 N |
221 | int recovery_disabled; /* if we detect that recovery |
222 | * will always fail, set this | |
223 | * so we don't loop trying */ | |
1da177e4 LT |
224 | |
225 | int in_sync; /* know to not need resync */ | |
c8c00a69 N |
226 | /* 'open_mutex' avoids races between 'md_open' and 'do_md_stop', so |
227 | * that we are never stopping an array while it is open. | |
228 | * 'reconfig_mutex' protects all other reconfiguration. | |
229 | * These locks are separate due to conflicting interactions | |
230 | * with bdev->bd_mutex. | |
231 | * Lock ordering is: | |
232 | * reconfig_mutex -> bd_mutex : e.g. do_md_run -> revalidate_disk | |
233 | * bd_mutex -> open_mutex: e.g. __blkdev_get -> md_open | |
234 | */ | |
235 | struct mutex open_mutex; | |
df5b89b3 | 236 | struct mutex reconfig_mutex; |
f2ea68cf N |
237 | atomic_t active; /* general refcount */ |
238 | atomic_t openers; /* number of active opens */ | |
1da177e4 | 239 | |
1da177e4 LT |
240 | int degraded; /* whether md should consider |
241 | * adding a spare | |
242 | */ | |
a9701a30 N |
243 | int barriers_work; /* initialised to true, cleared as soon |
244 | * as a barrier request to slave | |
245 | * fails. Only supported | |
246 | */ | |
247 | struct bio *biolist; /* bios that need to be retried | |
248 | * because BIO_RW_BARRIER is not supported | |
249 | */ | |
1da177e4 LT |
250 | |
251 | atomic_t recovery_active; /* blocks scheduled, but not written */ | |
252 | wait_queue_head_t recovery_wait; | |
253 | sector_t recovery_cp; | |
5e96ee65 NB |
254 | sector_t resync_min; /* user requested sync |
255 | * starts here */ | |
c6207277 N |
256 | sector_t resync_max; /* resync should pause |
257 | * when it gets here */ | |
06d91a5f | 258 | |
b62b7590 N |
259 | struct sysfs_dirent *sysfs_state; /* handle for 'array_state' |
260 | * file in sysfs. | |
261 | */ | |
0c3573f1 | 262 | struct sysfs_dirent *sysfs_action; /* handle for 'sync_action' */ |
b62b7590 | 263 | |
d3374825 N |
264 | struct work_struct del_work; /* used for delayed sysfs removal */ |
265 | ||
06d91a5f | 266 | spinlock_t write_lock; |
3d310eb7 | 267 | wait_queue_head_t sb_wait; /* for waiting on superblock updates */ |
7bfa19f2 | 268 | atomic_t pending_writes; /* number of active superblock writes */ |
06d91a5f | 269 | |
1da177e4 LT |
270 | unsigned int safemode; /* if set, update "clean" superblock |
271 | * when no writes pending. | |
272 | */ | |
273 | unsigned int safemode_delay; | |
274 | struct timer_list safemode_timer; | |
275 | atomic_t writes_pending; | |
165125e1 | 276 | struct request_queue *queue; /* for plugging ... */ |
1da177e4 | 277 | |
32a7627c | 278 | struct bitmap *bitmap; /* the bitmap for the device */ |
c3d9714e N |
279 | struct { |
280 | struct file *file; /* the bitmap file */ | |
f6af949c | 281 | loff_t offset; /* offset from superblock of |
c3d9714e N |
282 | * start of bitmap. May be |
283 | * negative, but not '0' | |
f6af949c N |
284 | * For external metadata, offset |
285 | * from start of device. | |
c3d9714e | 286 | */ |
f6af949c | 287 | loff_t default_offset; /* this is the offset to use when |
c3d9714e N |
288 | * hot-adding a bitmap. It should |
289 | * eventually be settable by sysfs. | |
290 | */ | |
291 | struct mutex mutex; | |
42a04b50 N |
292 | unsigned long chunksize; |
293 | unsigned long daemon_sleep; /* how many seconds between updates? */ | |
294 | unsigned long max_write_behind; /* write-behind mode */ | |
ece5cff0 | 295 | int external; |
c3d9714e | 296 | } bitmap_info; |
32a7627c | 297 | |
1e50915f | 298 | atomic_t max_corr_read_errors; /* max read retries */ |
1da177e4 | 299 | struct list_head all_mddevs; |
a2826aa9 | 300 | |
a64c876f | 301 | struct attribute_group *to_remove; |
a2826aa9 N |
302 | /* Generic barrier handling. |
303 | * If there is a pending barrier request, all other | |
304 | * writes are blocked while the devices are flushed. | |
305 | * The last to finish a flush schedules a worker to | |
306 | * submit the barrier request (without the barrier flag), | |
307 | * then submit more flush requests. | |
308 | */ | |
309 | struct bio *barrier; | |
310 | atomic_t flush_pending; | |
311 | struct work_struct barrier_work; | |
1da177e4 LT |
312 | }; |
313 | ||
314 | ||
315 | static inline void rdev_dec_pending(mdk_rdev_t *rdev, mddev_t *mddev) | |
316 | { | |
b2d444d7 | 317 | int faulty = test_bit(Faulty, &rdev->flags); |
1da177e4 LT |
318 | if (atomic_dec_and_test(&rdev->nr_pending) && faulty) |
319 | set_bit(MD_RECOVERY_NEEDED, &mddev->recovery); | |
320 | } | |
321 | ||
322 | static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors) | |
323 | { | |
324 | atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io); | |
325 | } | |
326 | ||
2604b703 | 327 | struct mdk_personality |
1da177e4 LT |
328 | { |
329 | char *name; | |
2604b703 N |
330 | int level; |
331 | struct list_head list; | |
1da177e4 | 332 | struct module *owner; |
21a52c6d | 333 | int (*make_request)(mddev_t *mddev, struct bio *bio); |
1da177e4 LT |
334 | int (*run)(mddev_t *mddev); |
335 | int (*stop)(mddev_t *mddev); | |
336 | void (*status)(struct seq_file *seq, mddev_t *mddev); | |
337 | /* error_handler must set ->faulty and clear ->in_sync | |
338 | * if appropriate, and should abort recovery if needed | |
339 | */ | |
340 | void (*error_handler)(mddev_t *mddev, mdk_rdev_t *rdev); | |
341 | int (*hot_add_disk) (mddev_t *mddev, mdk_rdev_t *rdev); | |
342 | int (*hot_remove_disk) (mddev_t *mddev, int number); | |
343 | int (*spare_active) (mddev_t *mddev); | |
57afd89f | 344 | sector_t (*sync_request)(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster); |
1da177e4 | 345 | int (*resize) (mddev_t *mddev, sector_t sectors); |
80c3a6ce | 346 | sector_t (*size) (mddev_t *mddev, sector_t sectors, int raid_disks); |
63c70c4f N |
347 | int (*check_reshape) (mddev_t *mddev); |
348 | int (*start_reshape) (mddev_t *mddev); | |
cea9c228 | 349 | void (*finish_reshape) (mddev_t *mddev); |
36fa3063 N |
350 | /* quiesce moves between quiescence states |
351 | * 0 - fully active | |
352 | * 1 - no new requests allowed | |
353 | * others - reserved | |
354 | */ | |
355 | void (*quiesce) (mddev_t *mddev, int state); | |
245f46c2 N |
356 | /* takeover is used to transition an array from one |
357 | * personality to another. The new personality must be able | |
358 | * to handle the data in the current layout. | |
359 | * e.g. 2drive raid1 -> 2drive raid5 | |
360 | * ndrive raid5 -> degraded n+1drive raid6 with special layout | |
361 | * If the takeover succeeds, a new 'private' structure is returned. | |
362 | * This needs to be installed and then ->run used to activate the | |
363 | * array. | |
364 | */ | |
365 | void *(*takeover) (mddev_t *mddev); | |
1da177e4 LT |
366 | }; |
367 | ||
368 | ||
007583c9 N |
369 | struct md_sysfs_entry { |
370 | struct attribute attr; | |
371 | ssize_t (*show)(mddev_t *, char *); | |
372 | ssize_t (*store)(mddev_t *, const char *, size_t); | |
373 | }; | |
43a70507 | 374 | extern struct attribute_group md_bitmap_group; |
007583c9 | 375 | |
1da177e4 LT |
376 | static inline char * mdname (mddev_t * mddev) |
377 | { | |
378 | return mddev->gendisk ? mddev->gendisk->disk_name : "mdX"; | |
379 | } | |
380 | ||
1da177e4 LT |
381 | /* |
382 | * iterates through some rdev ringlist. It's safe to remove the | |
383 | * current 'rdev'. Dont touch 'tmp' though. | |
384 | */ | |
159ec1fc CR |
385 | #define rdev_for_each_list(rdev, tmp, head) \ |
386 | list_for_each_entry_safe(rdev, tmp, head, same_set) | |
387 | ||
1da177e4 LT |
388 | /* |
389 | * iterates through the 'same array disks' ringlist | |
390 | */ | |
d089c6af | 391 | #define rdev_for_each(rdev, tmp, mddev) \ |
159ec1fc | 392 | list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set) |
1da177e4 | 393 | |
4b80991c N |
394 | #define rdev_for_each_rcu(rdev, mddev) \ |
395 | list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set) | |
396 | ||
1da177e4 LT |
397 | typedef struct mdk_thread_s { |
398 | void (*run) (mddev_t *mddev); | |
399 | mddev_t *mddev; | |
400 | wait_queue_head_t wqueue; | |
401 | unsigned long flags; | |
1da177e4 | 402 | struct task_struct *tsk; |
32a7627c | 403 | unsigned long timeout; |
1da177e4 LT |
404 | } mdk_thread_t; |
405 | ||
406 | #define THREAD_WAKEUP 0 | |
407 | ||
408 | #define __wait_event_lock_irq(wq, condition, lock, cmd) \ | |
409 | do { \ | |
410 | wait_queue_t __wait; \ | |
411 | init_waitqueue_entry(&__wait, current); \ | |
412 | \ | |
413 | add_wait_queue(&wq, &__wait); \ | |
414 | for (;;) { \ | |
415 | set_current_state(TASK_UNINTERRUPTIBLE); \ | |
416 | if (condition) \ | |
417 | break; \ | |
418 | spin_unlock_irq(&lock); \ | |
419 | cmd; \ | |
420 | schedule(); \ | |
421 | spin_lock_irq(&lock); \ | |
422 | } \ | |
423 | current->state = TASK_RUNNING; \ | |
424 | remove_wait_queue(&wq, &__wait); \ | |
425 | } while (0) | |
426 | ||
427 | #define wait_event_lock_irq(wq, condition, lock, cmd) \ | |
428 | do { \ | |
429 | if (condition) \ | |
430 | break; \ | |
431 | __wait_event_lock_irq(wq, condition, lock, cmd); \ | |
432 | } while (0) | |
433 | ||
1345b1d8 N |
434 | static inline void safe_put_page(struct page *p) |
435 | { | |
436 | if (p) put_page(p); | |
437 | } | |
438 | ||
92022950 N |
439 | extern int register_md_personality(struct mdk_personality *p); |
440 | extern int unregister_md_personality(struct mdk_personality *p); | |
441 | extern mdk_thread_t * md_register_thread(void (*run) (mddev_t *mddev), | |
442 | mddev_t *mddev, const char *name); | |
443 | extern void md_unregister_thread(mdk_thread_t *thread); | |
444 | extern void md_wakeup_thread(mdk_thread_t *thread); | |
445 | extern void md_check_recovery(mddev_t *mddev); | |
446 | extern void md_write_start(mddev_t *mddev, struct bio *bi); | |
447 | extern void md_write_end(mddev_t *mddev); | |
448 | extern void md_done_sync(mddev_t *mddev, int blocks, int ok); | |
449 | extern void md_error(mddev_t *mddev, mdk_rdev_t *rdev); | |
450 | ||
3fa841d7 | 451 | extern int mddev_congested(mddev_t *mddev, int bits); |
a2826aa9 | 452 | extern void md_barrier_request(mddev_t *mddev, struct bio *bio); |
92022950 N |
453 | extern void md_super_write(mddev_t *mddev, mdk_rdev_t *rdev, |
454 | sector_t sector, int size, struct page *page); | |
455 | extern void md_super_wait(mddev_t *mddev); | |
456 | extern int sync_page_io(struct block_device *bdev, sector_t sector, int size, | |
457 | struct page *page, int rw); | |
458 | extern void md_do_sync(mddev_t *mddev); | |
459 | extern void md_new_event(mddev_t *mddev); | |
460 | extern int md_allow_write(mddev_t *mddev); | |
461 | extern void md_wait_for_blocked_rdev(mdk_rdev_t *rdev, mddev_t *mddev); | |
1f403624 | 462 | extern void md_set_array_sectors(mddev_t *mddev, sector_t array_sectors); |
0894cc30 | 463 | extern int md_check_no_bitmap(mddev_t *mddev); |
ac5e7113 | 464 | extern int md_integrity_register(mddev_t *mddev); |
72e02075 N |
465 | extern void md_integrity_add_rdev(mdk_rdev_t *rdev, mddev_t *mddev); |
466 | extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale); | |
43a70507 | 467 | extern void restore_bitmap_write_access(struct file *file); |
63fe0817 CH |
468 | |
469 | #endif /* _MD_MD_H */ |