]>
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 | ||
1da177e4 LT |
21 | #define LEVEL_MULTIPATH (-4) |
22 | #define LEVEL_LINEAR (-1) | |
23 | #define LEVEL_FAULTY (-5) | |
24 | ||
2604b703 N |
25 | /* we need a value for 'no level specified' and 0 |
26 | * means 'raid0', so we need something else. This is | |
27 | * for internal use only | |
28 | */ | |
29 | #define LEVEL_NONE (-1000000) | |
30 | ||
1da177e4 LT |
31 | #define MaxSector (~(sector_t)0) |
32 | #define MD_THREAD_NAME_MAX 14 | |
33 | ||
1da177e4 LT |
34 | typedef struct mddev_s mddev_t; |
35 | typedef struct mdk_rdev_s mdk_rdev_t; | |
36 | ||
37 | #define MAX_MD_DEVS 256 /* Max number of md dev */ | |
38 | ||
39 | /* | |
40 | * options passed in raidrun: | |
41 | */ | |
42 | ||
43 | #define MAX_CHUNK_SIZE (4096*1024) | |
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 */ | |
54 | unsigned long last_events; /* IO event timestamp */ | |
55 | ||
56 | struct block_device *bdev; /* block device handle */ | |
57 | ||
58 | struct page *sb_page; | |
59 | int sb_loaded; | |
60 | sector_t data_offset; /* start of data in array */ | |
61 | sector_t sb_offset; | |
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 | |
b2d444d7 N |
78 | unsigned long flags; |
79 | #define Faulty 1 /* device is known to have a fault */ | |
80 | #define In_sync 2 /* device is in_sync with rest of array */ | |
8ddf9efe | 81 | #define WriteMostly 4 /* Avoid reading if at all possible */ |
a9701a30 | 82 | #define BarriersNotsupp 5 /* BIO_RW_BARRIER is not supported */ |
8ddf9efe | 83 | |
1da177e4 LT |
84 | int desc_nr; /* descriptor index in the superblock */ |
85 | int raid_disk; /* role of device in array */ | |
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 | */ | |
1da177e4 LT |
90 | |
91 | atomic_t nr_pending; /* number of pending requests. | |
92 | * only maintained for arrays that | |
93 | * support hot removal | |
94 | */ | |
ba22dcbf N |
95 | atomic_t read_errors; /* number of consecutive read errors that |
96 | * we have tried to ignore. | |
97 | */ | |
1da177e4 LT |
98 | }; |
99 | ||
1da177e4 LT |
100 | struct mddev_s |
101 | { | |
102 | void *private; | |
2604b703 | 103 | struct mdk_personality *pers; |
1da177e4 LT |
104 | dev_t unit; |
105 | int md_minor; | |
106 | struct list_head disks; | |
107 | int sb_dirty; | |
108 | int ro; | |
109 | ||
110 | struct gendisk *gendisk; | |
111 | ||
eae1701f N |
112 | struct kobject kobj; |
113 | ||
1da177e4 LT |
114 | /* Superblock information */ |
115 | int major_version, | |
116 | minor_version, | |
117 | patch_version; | |
118 | int persistent; | |
119 | int chunk_size; | |
120 | time_t ctime, utime; | |
121 | int level, layout; | |
d9d166c2 | 122 | char clevel[16]; |
1da177e4 LT |
123 | int raid_disks; |
124 | int max_disks; | |
125 | sector_t size; /* used size of component devices */ | |
126 | sector_t array_size; /* exported array size */ | |
127 | __u64 events; | |
128 | ||
129 | char uuid[16]; | |
130 | ||
131 | struct mdk_thread_s *thread; /* management thread */ | |
132 | struct mdk_thread_s *sync_thread; /* doing resync or reconstruct */ | |
133 | sector_t curr_resync; /* blocks scheduled */ | |
134 | unsigned long resync_mark; /* a recent timestamp */ | |
135 | sector_t resync_mark_cnt;/* blocks written at resync_mark */ | |
136 | ||
137 | sector_t resync_max_sectors; /* may be set by personality */ | |
9d88883e N |
138 | |
139 | sector_t resync_mismatches; /* count of sectors where | |
140 | * parity/replica mismatch found | |
141 | */ | |
6ff8d8ec | 142 | int ok_start_degraded; |
1da177e4 LT |
143 | /* recovery/resync flags |
144 | * NEEDED: we might need to start a resync/recover | |
145 | * RUNNING: a thread is running, or about to be started | |
146 | * SYNC: actually doing a resync, not a recovery | |
147 | * ERR: and IO error was detected - abort the resync/recovery | |
148 | * INTR: someone requested a (clean) early abort. | |
149 | * DONE: thread is done and is waiting to be reaped | |
24dd469d N |
150 | * REQUEST: user-space has requested a sync (used with SYNC) |
151 | * CHECK: user-space request for for check-only, no repair | |
1da177e4 LT |
152 | */ |
153 | #define MD_RECOVERY_RUNNING 0 | |
154 | #define MD_RECOVERY_SYNC 1 | |
155 | #define MD_RECOVERY_ERR 2 | |
156 | #define MD_RECOVERY_INTR 3 | |
157 | #define MD_RECOVERY_DONE 4 | |
158 | #define MD_RECOVERY_NEEDED 5 | |
24dd469d N |
159 | #define MD_RECOVERY_REQUESTED 6 |
160 | #define MD_RECOVERY_CHECK 7 | |
1da177e4 LT |
161 | unsigned long recovery; |
162 | ||
163 | int in_sync; /* know to not need resync */ | |
164 | struct semaphore reconfig_sem; | |
165 | atomic_t active; | |
166 | ||
167 | int changed; /* true if we might need to reread partition info */ | |
168 | int degraded; /* whether md should consider | |
169 | * adding a spare | |
170 | */ | |
a9701a30 N |
171 | int barriers_work; /* initialised to true, cleared as soon |
172 | * as a barrier request to slave | |
173 | * fails. Only supported | |
174 | */ | |
175 | struct bio *biolist; /* bios that need to be retried | |
176 | * because BIO_RW_BARRIER is not supported | |
177 | */ | |
1da177e4 LT |
178 | |
179 | atomic_t recovery_active; /* blocks scheduled, but not written */ | |
180 | wait_queue_head_t recovery_wait; | |
181 | sector_t recovery_cp; | |
06d91a5f N |
182 | |
183 | spinlock_t write_lock; | |
3d310eb7 | 184 | wait_queue_head_t sb_wait; /* for waiting on superblock updates */ |
7bfa19f2 | 185 | atomic_t pending_writes; /* number of active superblock writes */ |
06d91a5f | 186 | |
1da177e4 LT |
187 | unsigned int safemode; /* if set, update "clean" superblock |
188 | * when no writes pending. | |
189 | */ | |
190 | unsigned int safemode_delay; | |
191 | struct timer_list safemode_timer; | |
192 | atomic_t writes_pending; | |
193 | request_queue_t *queue; /* for plugging ... */ | |
194 | ||
4b6d287f N |
195 | atomic_t write_behind; /* outstanding async IO */ |
196 | unsigned int max_write_behind; /* 0 = sync */ | |
197 | ||
32a7627c N |
198 | struct bitmap *bitmap; /* the bitmap for the device */ |
199 | struct file *bitmap_file; /* the bitmap file */ | |
a654b9d8 N |
200 | long bitmap_offset; /* offset from superblock of |
201 | * start of bitmap. May be | |
202 | * negative, but not '0' | |
203 | */ | |
36fa3063 N |
204 | long default_bitmap_offset; /* this is the offset to use when |
205 | * hot-adding a bitmap. It should | |
206 | * eventually be settable by sysfs. | |
207 | */ | |
32a7627c | 208 | |
1da177e4 LT |
209 | struct list_head all_mddevs; |
210 | }; | |
211 | ||
212 | ||
213 | static inline void rdev_dec_pending(mdk_rdev_t *rdev, mddev_t *mddev) | |
214 | { | |
b2d444d7 | 215 | int faulty = test_bit(Faulty, &rdev->flags); |
1da177e4 LT |
216 | if (atomic_dec_and_test(&rdev->nr_pending) && faulty) |
217 | set_bit(MD_RECOVERY_NEEDED, &mddev->recovery); | |
218 | } | |
219 | ||
220 | static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors) | |
221 | { | |
222 | atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io); | |
223 | } | |
224 | ||
2604b703 | 225 | struct mdk_personality |
1da177e4 LT |
226 | { |
227 | char *name; | |
2604b703 N |
228 | int level; |
229 | struct list_head list; | |
1da177e4 LT |
230 | struct module *owner; |
231 | int (*make_request)(request_queue_t *q, struct bio *bio); | |
232 | int (*run)(mddev_t *mddev); | |
233 | int (*stop)(mddev_t *mddev); | |
234 | void (*status)(struct seq_file *seq, mddev_t *mddev); | |
235 | /* error_handler must set ->faulty and clear ->in_sync | |
236 | * if appropriate, and should abort recovery if needed | |
237 | */ | |
238 | void (*error_handler)(mddev_t *mddev, mdk_rdev_t *rdev); | |
239 | int (*hot_add_disk) (mddev_t *mddev, mdk_rdev_t *rdev); | |
240 | int (*hot_remove_disk) (mddev_t *mddev, int number); | |
241 | int (*spare_active) (mddev_t *mddev); | |
57afd89f | 242 | sector_t (*sync_request)(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster); |
1da177e4 LT |
243 | int (*resize) (mddev_t *mddev, sector_t sectors); |
244 | int (*reshape) (mddev_t *mddev, int raid_disks); | |
245 | int (*reconfig) (mddev_t *mddev, int layout, int chunk_size); | |
36fa3063 N |
246 | /* quiesce moves between quiescence states |
247 | * 0 - fully active | |
248 | * 1 - no new requests allowed | |
249 | * others - reserved | |
250 | */ | |
251 | void (*quiesce) (mddev_t *mddev, int state); | |
1da177e4 LT |
252 | }; |
253 | ||
254 | ||
007583c9 N |
255 | struct md_sysfs_entry { |
256 | struct attribute attr; | |
257 | ssize_t (*show)(mddev_t *, char *); | |
258 | ssize_t (*store)(mddev_t *, const char *, size_t); | |
259 | }; | |
260 | ||
261 | ||
1da177e4 LT |
262 | static inline char * mdname (mddev_t * mddev) |
263 | { | |
264 | return mddev->gendisk ? mddev->gendisk->disk_name : "mdX"; | |
265 | } | |
266 | ||
1da177e4 LT |
267 | /* |
268 | * iterates through some rdev ringlist. It's safe to remove the | |
269 | * current 'rdev'. Dont touch 'tmp' though. | |
270 | */ | |
271 | #define ITERATE_RDEV_GENERIC(head,rdev,tmp) \ | |
272 | \ | |
273 | for ((tmp) = (head).next; \ | |
274 | (rdev) = (list_entry((tmp), mdk_rdev_t, same_set)), \ | |
275 | (tmp) = (tmp)->next, (tmp)->prev != &(head) \ | |
276 | ; ) | |
277 | /* | |
278 | * iterates through the 'same array disks' ringlist | |
279 | */ | |
280 | #define ITERATE_RDEV(mddev,rdev,tmp) \ | |
281 | ITERATE_RDEV_GENERIC((mddev)->disks,rdev,tmp) | |
282 | ||
283 | /* | |
284 | * Iterates through 'pending RAID disks' | |
285 | */ | |
286 | #define ITERATE_RDEV_PENDING(rdev,tmp) \ | |
287 | ITERATE_RDEV_GENERIC(pending_raid_disks,rdev,tmp) | |
288 | ||
289 | typedef struct mdk_thread_s { | |
290 | void (*run) (mddev_t *mddev); | |
291 | mddev_t *mddev; | |
292 | wait_queue_head_t wqueue; | |
293 | unsigned long flags; | |
1da177e4 | 294 | struct task_struct *tsk; |
32a7627c | 295 | unsigned long timeout; |
1da177e4 LT |
296 | } mdk_thread_t; |
297 | ||
298 | #define THREAD_WAKEUP 0 | |
299 | ||
300 | #define __wait_event_lock_irq(wq, condition, lock, cmd) \ | |
301 | do { \ | |
302 | wait_queue_t __wait; \ | |
303 | init_waitqueue_entry(&__wait, current); \ | |
304 | \ | |
305 | add_wait_queue(&wq, &__wait); \ | |
306 | for (;;) { \ | |
307 | set_current_state(TASK_UNINTERRUPTIBLE); \ | |
308 | if (condition) \ | |
309 | break; \ | |
310 | spin_unlock_irq(&lock); \ | |
311 | cmd; \ | |
312 | schedule(); \ | |
313 | spin_lock_irq(&lock); \ | |
314 | } \ | |
315 | current->state = TASK_RUNNING; \ | |
316 | remove_wait_queue(&wq, &__wait); \ | |
317 | } while (0) | |
318 | ||
319 | #define wait_event_lock_irq(wq, condition, lock, cmd) \ | |
320 | do { \ | |
321 | if (condition) \ | |
322 | break; \ | |
323 | __wait_event_lock_irq(wq, condition, lock, cmd); \ | |
324 | } while (0) | |
325 | ||
1345b1d8 N |
326 | static inline void safe_put_page(struct page *p) |
327 | { | |
328 | if (p) put_page(p); | |
329 | } | |
330 | ||
1da177e4 LT |
331 | #endif |
332 |