]> bbs.cooldavid.org Git - net-next-2.6.git/blame - include/linux/raid/md_k.h
[PATCH] md: remove MAX_MD_DEVS which is an arbitrary limit
[net-next-2.6.git] / include / linux / raid / md_k.h
CommitLineData
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
35typedef struct mddev_s mddev_t;
36typedef 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 */
48struct 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;
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 */
1da177e4
LT
107};
108
1da177e4
LT
109struct mddev_s
110{
111 void *private;
2604b703 112 struct mdk_personality *pers;
1da177e4
LT
113 dev_t unit;
114 int md_minor;
115 struct list_head disks;
850b2b42
N
116 unsigned long flags;
117#define MD_CHANGE_DEVS 0 /* Some device status has changed */
118#define MD_CHANGE_CLEAN 1 /* transition to or from 'clean' */
119#define MD_CHANGE_PENDING 2 /* superblock update in progress */
120
1da177e4
LT
121 int ro;
122
123 struct gendisk *gendisk;
124
eae1701f
N
125 struct kobject kobj;
126
1da177e4
LT
127 /* Superblock information */
128 int major_version,
129 minor_version,
130 patch_version;
131 int persistent;
132 int chunk_size;
133 time_t ctime, utime;
134 int level, layout;
d9d166c2 135 char clevel[16];
1da177e4
LT
136 int raid_disks;
137 int max_disks;
138 sector_t size; /* used size of component devices */
139 sector_t array_size; /* exported array size */
140 __u64 events;
141
142 char uuid[16];
143
f6705578
N
144 /* If the array is being reshaped, we need to record the
145 * new shape and an indication of where we are up to.
146 * This is written to the superblock.
147 * If reshape_position is MaxSector, then no reshape is happening (yet).
148 */
149 sector_t reshape_position;
150 int delta_disks, new_level, new_layout, new_chunk;
151
1da177e4
LT
152 struct mdk_thread_s *thread; /* management thread */
153 struct mdk_thread_s *sync_thread; /* doing resync or reconstruct */
ff4e8d9a 154 sector_t curr_resync; /* last block scheduled */
1da177e4
LT
155 unsigned long resync_mark; /* a recent timestamp */
156 sector_t resync_mark_cnt;/* blocks written at resync_mark */
ff4e8d9a 157 sector_t curr_mark_cnt; /* blocks scheduled now */
1da177e4
LT
158
159 sector_t resync_max_sectors; /* may be set by personality */
9d88883e
N
160
161 sector_t resync_mismatches; /* count of sectors where
162 * parity/replica mismatch found
163 */
e464eafd
N
164
165 /* allow user-space to request suspension of IO to regions of the array */
166 sector_t suspend_lo;
167 sector_t suspend_hi;
88202a0c
N
168 /* if zero, use the system-wide default */
169 int sync_speed_min;
170 int sync_speed_max;
171
6ff8d8ec 172 int ok_start_degraded;
1da177e4
LT
173 /* recovery/resync flags
174 * NEEDED: we might need to start a resync/recover
175 * RUNNING: a thread is running, or about to be started
176 * SYNC: actually doing a resync, not a recovery
177 * ERR: and IO error was detected - abort the resync/recovery
178 * INTR: someone requested a (clean) early abort.
179 * DONE: thread is done and is waiting to be reaped
24dd469d
N
180 * REQUEST: user-space has requested a sync (used with SYNC)
181 * CHECK: user-space request for for check-only, no repair
ccfcc3c1
N
182 * RESHAPE: A reshape is happening
183 *
184 * If neither SYNC or RESHAPE are set, then it is a recovery.
1da177e4
LT
185 */
186#define MD_RECOVERY_RUNNING 0
187#define MD_RECOVERY_SYNC 1
188#define MD_RECOVERY_ERR 2
189#define MD_RECOVERY_INTR 3
190#define MD_RECOVERY_DONE 4
191#define MD_RECOVERY_NEEDED 5
24dd469d
N
192#define MD_RECOVERY_REQUESTED 6
193#define MD_RECOVERY_CHECK 7
ccfcc3c1 194#define MD_RECOVERY_RESHAPE 8
5fd6c1dc
N
195#define MD_RECOVERY_FROZEN 9
196
1da177e4
LT
197 unsigned long recovery;
198
199 int in_sync; /* know to not need resync */
df5b89b3 200 struct mutex reconfig_mutex;
1da177e4
LT
201 atomic_t active;
202
203 int changed; /* true if we might need to reread partition info */
204 int degraded; /* whether md should consider
205 * adding a spare
206 */
a9701a30
N
207 int barriers_work; /* initialised to true, cleared as soon
208 * as a barrier request to slave
209 * fails. Only supported
210 */
211 struct bio *biolist; /* bios that need to be retried
212 * because BIO_RW_BARRIER is not supported
213 */
1da177e4
LT
214
215 atomic_t recovery_active; /* blocks scheduled, but not written */
216 wait_queue_head_t recovery_wait;
217 sector_t recovery_cp;
06d91a5f
N
218
219 spinlock_t write_lock;
3d310eb7 220 wait_queue_head_t sb_wait; /* for waiting on superblock updates */
7bfa19f2 221 atomic_t pending_writes; /* number of active superblock writes */
06d91a5f 222
1da177e4
LT
223 unsigned int safemode; /* if set, update "clean" superblock
224 * when no writes pending.
225 */
226 unsigned int safemode_delay;
227 struct timer_list safemode_timer;
228 atomic_t writes_pending;
229 request_queue_t *queue; /* for plugging ... */
230
4b6d287f
N
231 atomic_t write_behind; /* outstanding async IO */
232 unsigned int max_write_behind; /* 0 = sync */
233
32a7627c
N
234 struct bitmap *bitmap; /* the bitmap for the device */
235 struct file *bitmap_file; /* the bitmap file */
a654b9d8
N
236 long bitmap_offset; /* offset from superblock of
237 * start of bitmap. May be
238 * negative, but not '0'
239 */
36fa3063
N
240 long default_bitmap_offset; /* this is the offset to use when
241 * hot-adding a bitmap. It should
242 * eventually be settable by sysfs.
243 */
32a7627c 244
1da177e4
LT
245 struct list_head all_mddevs;
246};
247
248
249static inline void rdev_dec_pending(mdk_rdev_t *rdev, mddev_t *mddev)
250{
b2d444d7 251 int faulty = test_bit(Faulty, &rdev->flags);
1da177e4
LT
252 if (atomic_dec_and_test(&rdev->nr_pending) && faulty)
253 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
254}
255
256static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors)
257{
258 atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io);
259}
260
2604b703 261struct mdk_personality
1da177e4
LT
262{
263 char *name;
2604b703
N
264 int level;
265 struct list_head list;
1da177e4
LT
266 struct module *owner;
267 int (*make_request)(request_queue_t *q, struct bio *bio);
268 int (*run)(mddev_t *mddev);
269 int (*stop)(mddev_t *mddev);
270 void (*status)(struct seq_file *seq, mddev_t *mddev);
271 /* error_handler must set ->faulty and clear ->in_sync
272 * if appropriate, and should abort recovery if needed
273 */
274 void (*error_handler)(mddev_t *mddev, mdk_rdev_t *rdev);
275 int (*hot_add_disk) (mddev_t *mddev, mdk_rdev_t *rdev);
276 int (*hot_remove_disk) (mddev_t *mddev, int number);
277 int (*spare_active) (mddev_t *mddev);
57afd89f 278 sector_t (*sync_request)(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster);
1da177e4 279 int (*resize) (mddev_t *mddev, sector_t sectors);
63c70c4f
N
280 int (*check_reshape) (mddev_t *mddev);
281 int (*start_reshape) (mddev_t *mddev);
1da177e4 282 int (*reconfig) (mddev_t *mddev, int layout, int chunk_size);
36fa3063
N
283 /* quiesce moves between quiescence states
284 * 0 - fully active
285 * 1 - no new requests allowed
286 * others - reserved
287 */
288 void (*quiesce) (mddev_t *mddev, int state);
1da177e4
LT
289};
290
291
007583c9
N
292struct md_sysfs_entry {
293 struct attribute attr;
294 ssize_t (*show)(mddev_t *, char *);
295 ssize_t (*store)(mddev_t *, const char *, size_t);
296};
297
298
1da177e4
LT
299static inline char * mdname (mddev_t * mddev)
300{
301 return mddev->gendisk ? mddev->gendisk->disk_name : "mdX";
302}
303
1da177e4
LT
304/*
305 * iterates through some rdev ringlist. It's safe to remove the
306 * current 'rdev'. Dont touch 'tmp' though.
307 */
308#define ITERATE_RDEV_GENERIC(head,rdev,tmp) \
309 \
310 for ((tmp) = (head).next; \
311 (rdev) = (list_entry((tmp), mdk_rdev_t, same_set)), \
312 (tmp) = (tmp)->next, (tmp)->prev != &(head) \
313 ; )
314/*
315 * iterates through the 'same array disks' ringlist
316 */
317#define ITERATE_RDEV(mddev,rdev,tmp) \
318 ITERATE_RDEV_GENERIC((mddev)->disks,rdev,tmp)
319
320/*
321 * Iterates through 'pending RAID disks'
322 */
323#define ITERATE_RDEV_PENDING(rdev,tmp) \
324 ITERATE_RDEV_GENERIC(pending_raid_disks,rdev,tmp)
325
326typedef struct mdk_thread_s {
327 void (*run) (mddev_t *mddev);
328 mddev_t *mddev;
329 wait_queue_head_t wqueue;
330 unsigned long flags;
1da177e4 331 struct task_struct *tsk;
32a7627c 332 unsigned long timeout;
1da177e4
LT
333} mdk_thread_t;
334
335#define THREAD_WAKEUP 0
336
337#define __wait_event_lock_irq(wq, condition, lock, cmd) \
338do { \
339 wait_queue_t __wait; \
340 init_waitqueue_entry(&__wait, current); \
341 \
342 add_wait_queue(&wq, &__wait); \
343 for (;;) { \
344 set_current_state(TASK_UNINTERRUPTIBLE); \
345 if (condition) \
346 break; \
347 spin_unlock_irq(&lock); \
348 cmd; \
349 schedule(); \
350 spin_lock_irq(&lock); \
351 } \
352 current->state = TASK_RUNNING; \
353 remove_wait_queue(&wq, &__wait); \
354} while (0)
355
356#define wait_event_lock_irq(wq, condition, lock, cmd) \
357do { \
358 if (condition) \
359 break; \
360 __wait_event_lock_irq(wq, condition, lock, cmd); \
361} while (0)
362
1345b1d8
N
363static inline void safe_put_page(struct page *p)
364{
365 if (p) put_page(p);
366}
367
9361401e 368#endif /* CONFIG_BLOCK */
1da177e4
LT
369#endif
370