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 | ||
15 | #ifndef _MD_K_H | |
16 | #define _MD_K_H | |
17 | ||
06d91a5f N |
18 | /* and dm-bio-list.h is not under include/linux because.... ??? */ |
19 | #include "../../../drivers/md/dm-bio-list.h" | |
20 | ||
9361401e DH |
21 | #ifdef CONFIG_BLOCK |
22 | ||
1da177e4 LT |
23 | #define LEVEL_MULTIPATH (-4) |
24 | #define LEVEL_LINEAR (-1) | |
25 | #define LEVEL_FAULTY (-5) | |
26 | ||
2604b703 N |
27 | /* we need a value for 'no level specified' and 0 |
28 | * means 'raid0', so we need something else. This is | |
29 | * for internal use only | |
30 | */ | |
31 | #define LEVEL_NONE (-1000000) | |
32 | ||
1da177e4 | 33 | #define MaxSector (~(sector_t)0) |
1da177e4 | 34 | |
1da177e4 LT |
35 | typedef struct mddev_s mddev_t; |
36 | typedef struct mdk_rdev_s mdk_rdev_t; | |
37 | ||
1da177e4 LT |
38 | /* |
39 | * options passed in raidrun: | |
40 | */ | |
41 | ||
e8703fe1 | 42 | /* Currently this must fit in an 'int' */ |
8932c2e0 | 43 | #define MAX_CHUNK_SIZE (1<<30) |
1da177e4 | 44 | |
1da177e4 LT |
45 | /* |
46 | * MD's 'extended' device | |
47 | */ | |
48 | struct mdk_rdev_s | |
49 | { | |
50 | struct list_head same_set; /* RAID devices within the same set */ | |
51 | ||
52 | sector_t size; /* Device size (in blocks) */ | |
53 | mddev_t *mddev; /* RAID array if running */ | |
713f6ab1 | 54 | long last_events; /* IO event timestamp */ |
1da177e4 LT |
55 | |
56 | struct block_device *bdev; /* block device handle */ | |
57 | ||
58 | struct page *sb_page; | |
59 | int sb_loaded; | |
42543769 | 60 | __u64 sb_events; |
1da177e4 LT |
61 | sector_t data_offset; /* start of data in array */ |
62 | sector_t sb_offset; | |
0002b271 | 63 | int sb_size; /* bytes in the superblock */ |
1da177e4 LT |
64 | int preferred_minor; /* autorun support */ |
65 | ||
86e6ffdd N |
66 | struct kobject kobj; |
67 | ||
1da177e4 LT |
68 | /* A device can be in one of three states based on two flags: |
69 | * Not working: faulty==1 in_sync==0 | |
70 | * Fully working: faulty==0 in_sync==1 | |
71 | * Working, but not | |
72 | * in sync with array | |
73 | * faulty==0 in_sync==0 | |
74 | * | |
75 | * It can never have faulty==1, in_sync==1 | |
76 | * This reduces the burden of testing multiple flags in many cases | |
77 | */ | |
1da177e4 | 78 | |
b2d444d7 N |
79 | unsigned long flags; |
80 | #define Faulty 1 /* device is known to have a fault */ | |
81 | #define In_sync 2 /* device is in_sync with rest of array */ | |
8ddf9efe | 82 | #define WriteMostly 4 /* Avoid reading if at all possible */ |
a9701a30 | 83 | #define BarriersNotsupp 5 /* BIO_RW_BARRIER is not supported */ |
8ddf9efe | 84 | |
1da177e4 LT |
85 | int desc_nr; /* descriptor index in the superblock */ |
86 | int raid_disk; /* role of device in array */ | |
41158c7e N |
87 | int saved_raid_disk; /* role that device used to have in the |
88 | * array and could again if we did a partial | |
89 | * resync from the bitmap | |
90 | */ | |
5fd6c1dc N |
91 | sector_t recovery_offset;/* If this device has been partially |
92 | * recovered, this is where we were | |
93 | * up to. | |
94 | */ | |
1da177e4 LT |
95 | |
96 | atomic_t nr_pending; /* number of pending requests. | |
97 | * only maintained for arrays that | |
98 | * support hot removal | |
99 | */ | |
ba22dcbf N |
100 | atomic_t read_errors; /* number of consecutive read errors that |
101 | * we have tried to ignore. | |
102 | */ | |
4dbcdc75 N |
103 | atomic_t corrected_errors; /* number of corrected read errors, |
104 | * for reporting to userspace and storing | |
105 | * in superblock. | |
106 | */ | |
5792a285 | 107 | struct work_struct del_work; /* used for delayed sysfs removal */ |
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 | ||
1da177e4 LT |
122 | int ro; |
123 | ||
124 | struct gendisk *gendisk; | |
125 | ||
eae1701f N |
126 | struct kobject kobj; |
127 | ||
1da177e4 LT |
128 | /* Superblock information */ |
129 | int major_version, | |
130 | minor_version, | |
131 | patch_version; | |
132 | int persistent; | |
e691063a N |
133 | int external; /* metadata is |
134 | * managed externally */ | |
135 | char metadata_type[17]; /* externally set*/ | |
1da177e4 LT |
136 | int chunk_size; |
137 | time_t ctime, utime; | |
138 | int level, layout; | |
d9d166c2 | 139 | char clevel[16]; |
1da177e4 LT |
140 | int raid_disks; |
141 | int max_disks; | |
142 | sector_t size; /* used size of component devices */ | |
143 | sector_t array_size; /* exported array size */ | |
144 | __u64 events; | |
145 | ||
146 | char uuid[16]; | |
147 | ||
f6705578 N |
148 | /* If the array is being reshaped, we need to record the |
149 | * new shape and an indication of where we are up to. | |
150 | * This is written to the superblock. | |
151 | * If reshape_position is MaxSector, then no reshape is happening (yet). | |
152 | */ | |
153 | sector_t reshape_position; | |
154 | int delta_disks, new_level, new_layout, new_chunk; | |
155 | ||
1da177e4 LT |
156 | struct mdk_thread_s *thread; /* management thread */ |
157 | struct mdk_thread_s *sync_thread; /* doing resync or reconstruct */ | |
ff4e8d9a | 158 | sector_t curr_resync; /* last block scheduled */ |
1da177e4 LT |
159 | unsigned long resync_mark; /* a recent timestamp */ |
160 | sector_t resync_mark_cnt;/* blocks written at resync_mark */ | |
ff4e8d9a | 161 | sector_t curr_mark_cnt; /* blocks scheduled now */ |
1da177e4 LT |
162 | |
163 | sector_t resync_max_sectors; /* may be set by personality */ | |
9d88883e N |
164 | |
165 | sector_t resync_mismatches; /* count of sectors where | |
166 | * parity/replica mismatch found | |
167 | */ | |
e464eafd N |
168 | |
169 | /* allow user-space to request suspension of IO to regions of the array */ | |
170 | sector_t suspend_lo; | |
171 | sector_t suspend_hi; | |
88202a0c N |
172 | /* if zero, use the system-wide default */ |
173 | int sync_speed_min; | |
174 | int sync_speed_max; | |
175 | ||
6ff8d8ec | 176 | int ok_start_degraded; |
1da177e4 LT |
177 | /* recovery/resync flags |
178 | * NEEDED: we might need to start a resync/recover | |
179 | * RUNNING: a thread is running, or about to be started | |
180 | * SYNC: actually doing a resync, not a recovery | |
181 | * ERR: and IO error was detected - abort the resync/recovery | |
182 | * INTR: someone requested a (clean) early abort. | |
183 | * DONE: thread is done and is waiting to be reaped | |
24dd469d N |
184 | * REQUEST: user-space has requested a sync (used with SYNC) |
185 | * CHECK: user-space request for for check-only, no repair | |
ccfcc3c1 N |
186 | * RESHAPE: A reshape is happening |
187 | * | |
188 | * If neither SYNC or RESHAPE are set, then it is a recovery. | |
1da177e4 LT |
189 | */ |
190 | #define MD_RECOVERY_RUNNING 0 | |
191 | #define MD_RECOVERY_SYNC 1 | |
192 | #define MD_RECOVERY_ERR 2 | |
193 | #define MD_RECOVERY_INTR 3 | |
194 | #define MD_RECOVERY_DONE 4 | |
195 | #define MD_RECOVERY_NEEDED 5 | |
24dd469d N |
196 | #define MD_RECOVERY_REQUESTED 6 |
197 | #define MD_RECOVERY_CHECK 7 | |
ccfcc3c1 | 198 | #define MD_RECOVERY_RESHAPE 8 |
5fd6c1dc N |
199 | #define MD_RECOVERY_FROZEN 9 |
200 | ||
1da177e4 LT |
201 | unsigned long recovery; |
202 | ||
203 | int in_sync; /* know to not need resync */ | |
df5b89b3 | 204 | struct mutex reconfig_mutex; |
1da177e4 LT |
205 | atomic_t active; |
206 | ||
44ce6294 | 207 | int changed; /* true if we might need to reread partition info */ |
1da177e4 LT |
208 | int degraded; /* whether md should consider |
209 | * adding a spare | |
210 | */ | |
a9701a30 N |
211 | int barriers_work; /* initialised to true, cleared as soon |
212 | * as a barrier request to slave | |
213 | * fails. Only supported | |
214 | */ | |
215 | struct bio *biolist; /* bios that need to be retried | |
216 | * because BIO_RW_BARRIER is not supported | |
217 | */ | |
1da177e4 LT |
218 | |
219 | atomic_t recovery_active; /* blocks scheduled, but not written */ | |
220 | wait_queue_head_t recovery_wait; | |
221 | sector_t recovery_cp; | |
06d91a5f N |
222 | |
223 | spinlock_t write_lock; | |
3d310eb7 | 224 | wait_queue_head_t sb_wait; /* for waiting on superblock updates */ |
7bfa19f2 | 225 | atomic_t pending_writes; /* number of active superblock writes */ |
06d91a5f | 226 | |
1da177e4 LT |
227 | unsigned int safemode; /* if set, update "clean" superblock |
228 | * when no writes pending. | |
229 | */ | |
230 | unsigned int safemode_delay; | |
231 | struct timer_list safemode_timer; | |
232 | atomic_t writes_pending; | |
165125e1 | 233 | struct request_queue *queue; /* for plugging ... */ |
1da177e4 | 234 | |
4b6d287f N |
235 | atomic_t write_behind; /* outstanding async IO */ |
236 | unsigned int max_write_behind; /* 0 = sync */ | |
237 | ||
32a7627c N |
238 | struct bitmap *bitmap; /* the bitmap for the device */ |
239 | struct file *bitmap_file; /* the bitmap file */ | |
a654b9d8 N |
240 | long bitmap_offset; /* offset from superblock of |
241 | * start of bitmap. May be | |
242 | * negative, but not '0' | |
243 | */ | |
36fa3063 N |
244 | long default_bitmap_offset; /* this is the offset to use when |
245 | * hot-adding a bitmap. It should | |
246 | * eventually be settable by sysfs. | |
247 | */ | |
32a7627c | 248 | |
1da177e4 LT |
249 | struct list_head all_mddevs; |
250 | }; | |
251 | ||
252 | ||
253 | static inline void rdev_dec_pending(mdk_rdev_t *rdev, mddev_t *mddev) | |
254 | { | |
b2d444d7 | 255 | int faulty = test_bit(Faulty, &rdev->flags); |
1da177e4 LT |
256 | if (atomic_dec_and_test(&rdev->nr_pending) && faulty) |
257 | set_bit(MD_RECOVERY_NEEDED, &mddev->recovery); | |
258 | } | |
259 | ||
260 | static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors) | |
261 | { | |
262 | atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io); | |
263 | } | |
264 | ||
2604b703 | 265 | struct mdk_personality |
1da177e4 LT |
266 | { |
267 | char *name; | |
2604b703 N |
268 | int level; |
269 | struct list_head list; | |
1da177e4 | 270 | struct module *owner; |
165125e1 | 271 | int (*make_request)(struct request_queue *q, struct bio *bio); |
1da177e4 LT |
272 | int (*run)(mddev_t *mddev); |
273 | int (*stop)(mddev_t *mddev); | |
274 | void (*status)(struct seq_file *seq, mddev_t *mddev); | |
275 | /* error_handler must set ->faulty and clear ->in_sync | |
276 | * if appropriate, and should abort recovery if needed | |
277 | */ | |
278 | void (*error_handler)(mddev_t *mddev, mdk_rdev_t *rdev); | |
279 | int (*hot_add_disk) (mddev_t *mddev, mdk_rdev_t *rdev); | |
280 | int (*hot_remove_disk) (mddev_t *mddev, int number); | |
281 | int (*spare_active) (mddev_t *mddev); | |
57afd89f | 282 | sector_t (*sync_request)(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster); |
1da177e4 | 283 | int (*resize) (mddev_t *mddev, sector_t sectors); |
63c70c4f N |
284 | int (*check_reshape) (mddev_t *mddev); |
285 | int (*start_reshape) (mddev_t *mddev); | |
1da177e4 | 286 | int (*reconfig) (mddev_t *mddev, int layout, int chunk_size); |
36fa3063 N |
287 | /* quiesce moves between quiescence states |
288 | * 0 - fully active | |
289 | * 1 - no new requests allowed | |
290 | * others - reserved | |
291 | */ | |
292 | void (*quiesce) (mddev_t *mddev, int state); | |
1da177e4 LT |
293 | }; |
294 | ||
295 | ||
007583c9 N |
296 | struct md_sysfs_entry { |
297 | struct attribute attr; | |
298 | ssize_t (*show)(mddev_t *, char *); | |
299 | ssize_t (*store)(mddev_t *, const char *, size_t); | |
300 | }; | |
301 | ||
302 | ||
1da177e4 LT |
303 | static inline char * mdname (mddev_t * mddev) |
304 | { | |
305 | return mddev->gendisk ? mddev->gendisk->disk_name : "mdX"; | |
306 | } | |
307 | ||
1da177e4 LT |
308 | /* |
309 | * iterates through some rdev ringlist. It's safe to remove the | |
310 | * current 'rdev'. Dont touch 'tmp' though. | |
311 | */ | |
312 | #define ITERATE_RDEV_GENERIC(head,rdev,tmp) \ | |
313 | \ | |
314 | for ((tmp) = (head).next; \ | |
315 | (rdev) = (list_entry((tmp), mdk_rdev_t, same_set)), \ | |
316 | (tmp) = (tmp)->next, (tmp)->prev != &(head) \ | |
317 | ; ) | |
318 | /* | |
319 | * iterates through the 'same array disks' ringlist | |
320 | */ | |
321 | #define ITERATE_RDEV(mddev,rdev,tmp) \ | |
322 | ITERATE_RDEV_GENERIC((mddev)->disks,rdev,tmp) | |
323 | ||
324 | /* | |
325 | * Iterates through 'pending RAID disks' | |
326 | */ | |
327 | #define ITERATE_RDEV_PENDING(rdev,tmp) \ | |
328 | ITERATE_RDEV_GENERIC(pending_raid_disks,rdev,tmp) | |
329 | ||
330 | typedef struct mdk_thread_s { | |
331 | void (*run) (mddev_t *mddev); | |
332 | mddev_t *mddev; | |
333 | wait_queue_head_t wqueue; | |
334 | unsigned long flags; | |
1da177e4 | 335 | struct task_struct *tsk; |
32a7627c | 336 | unsigned long timeout; |
1da177e4 LT |
337 | } mdk_thread_t; |
338 | ||
339 | #define THREAD_WAKEUP 0 | |
340 | ||
341 | #define __wait_event_lock_irq(wq, condition, lock, cmd) \ | |
342 | do { \ | |
343 | wait_queue_t __wait; \ | |
344 | init_waitqueue_entry(&__wait, current); \ | |
345 | \ | |
346 | add_wait_queue(&wq, &__wait); \ | |
347 | for (;;) { \ | |
348 | set_current_state(TASK_UNINTERRUPTIBLE); \ | |
349 | if (condition) \ | |
350 | break; \ | |
351 | spin_unlock_irq(&lock); \ | |
352 | cmd; \ | |
353 | schedule(); \ | |
354 | spin_lock_irq(&lock); \ | |
355 | } \ | |
356 | current->state = TASK_RUNNING; \ | |
357 | remove_wait_queue(&wq, &__wait); \ | |
358 | } while (0) | |
359 | ||
360 | #define wait_event_lock_irq(wq, condition, lock, cmd) \ | |
361 | do { \ | |
362 | if (condition) \ | |
363 | break; \ | |
364 | __wait_event_lock_irq(wq, condition, lock, cmd); \ | |
365 | } while (0) | |
366 | ||
1345b1d8 N |
367 | static inline void safe_put_page(struct page *p) |
368 | { | |
369 | if (p) put_page(p); | |
370 | } | |
371 | ||
9361401e | 372 | #endif /* CONFIG_BLOCK */ |
1da177e4 LT |
373 | #endif |
374 |