]> bbs.cooldavid.org Git - net-next-2.6.git/blame - drivers/md/md.h
git://bbs.cooldavid.org / net-next-2.6.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge branch 'for-linus' of git://git.infradead.org/users/eparis/notify
[net-next-2.6.git] / drivers / md / md.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
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		 29typedef struct mddev_s mddev_t;
30typedef struct mdk_rdev_s mdk_rdev_t;
31
2ac87401
N		 32/* generic plugging support - like that provided with request_queue,
33 * but does not require a request_queue
34 */
35struct plug_handle {
36 void (*unplug_fn)(struct plug_handle *);
37 struct timer_list unplug_timer;
38 struct work_struct unplug_work;
39 unsigned long unplug_flag;
40};
41#define PLUGGED_FLAG 1
42void plugger_init(struct plug_handle *plug,
43 void (*unplug_fn)(struct plug_handle *));
44void plugger_set_plug(struct plug_handle *plug);
45int plugger_remove_plug(struct plug_handle *plug);
46static inline void plugger_flush(struct plug_handle *plug)
47{
48 del_timer_sync(&plug->unplug_timer);
49 cancel_work_sync(&plug->unplug_work);
50}
51
1da177e4
LT		 52/*
53 * MD's 'extended' device
54 */
55struct mdk_rdev_s
56{
57 struct list_head same_set; /* RAID devices within the same set */
58
dd8ac336 59 sector_t sectors; /* Device size (in 512bytes sectors) */
1da177e4 60 mddev_t *mddev; /* RAID array if running */
eea1bf38 61 int last_events; /* IO event timestamp */
1da177e4
LT		 62
63 struct block_device *bdev; /* block device handle */
64
65 struct page *sb_page;
66 int sb_loaded;
42543769 67 __u64 sb_events;
1da177e4 68 sector_t data_offset; /* start of data in array */
0f420358 69 sector_t sb_start; /* offset of the super block (in 512byte sectors) */
0002b271 70 int sb_size; /* bytes in the superblock */
1da177e4
LT		 71 int preferred_minor; /* autorun support */
72
86e6ffdd
N		 73 struct kobject kobj;
74
1da177e4
LT		 75 /* A device can be in one of three states based on two flags:
76 * Not working: faulty==1 in_sync==0
77 * Fully working: faulty==0 in_sync==1
78 * Working, but not
79 * in sync with array
80 * faulty==0 in_sync==0
81 *
82 * It can never have faulty==1, in_sync==1
83 * This reduces the burden of testing multiple flags in many cases
84 */
1da177e4 85
b2d444d7
N		 86 unsigned long flags;
87#define Faulty 1 /* device is known to have a fault */
88#define In_sync 2 /* device is in_sync with rest of array */
8ddf9efe 89#define WriteMostly 4 /* Avoid reading if at all possible */
c5d79adb
N		 90#define AllReserved 6 /* If whole device is reserved for
91 * one array */
d0fae18f 92#define AutoDetected 7 /* added by auto-detect */
6bfe0b49
DW		 93#define Blocked 8 /* An error occured on an externally
94 * managed array, don't allow writes
95 * until it is cleared */
96 wait_queue_head_t blocked_wait;
8ddf9efe 97
1da177e4
LT		 98 int desc_nr; /* descriptor index in the superblock */
99 int raid_disk; /* role of device in array */
e93f68a1
N		 100 int new_raid_disk; /* role that the device will have in
101 * the array after a level-change completes.
102 */
41158c7e
N		 103 int saved_raid_disk; /* role that device used to have in the
104 * array and could again if we did a partial
105 * resync from the bitmap
106 */
5fd6c1dc
N		 107 sector_t recovery_offset;/* If this device has been partially
108 * recovered, this is where we were
109 * up to.
110 */
1da177e4
LT		 111
112 atomic_t nr_pending; /* number of pending requests.
113 * only maintained for arrays that
114 * support hot removal
115 */
ba22dcbf
N		 116 atomic_t read_errors; /* number of consecutive read errors that
117 * we have tried to ignore.
118 */
1e50915f
RB		 119 struct timespec last_read_error; /* monotonic time since our
120 * last read error
121 */
4dbcdc75
N		 122 atomic_t corrected_errors; /* number of corrected read errors,
123 * for reporting to userspace and storing
124 * in superblock.
125 */
5792a285 126 struct work_struct del_work; /* used for delayed sysfs removal */
3c0ee63a
N		 127
128 struct sysfs_dirent *sysfs_state; /* handle for 'state'
129 * sysfs entry */
1da177e4
LT		 130};
131
1da177e4
LT		 132struct mddev_s
133{
134 void *private;
2604b703 135 struct mdk_personality *pers;
1da177e4
LT		 136 dev_t unit;
137 int md_minor;
138 struct list_head disks;
850b2b42
N		 139 unsigned long flags;
140#define MD_CHANGE_DEVS 0 /* Some device status has changed */
141#define MD_CHANGE_CLEAN 1 /* transition to or from 'clean' */
070dc6dd 142#define MD_CHANGE_PENDING 2 /* switch from 'clean' to 'active' in progress */
850b2b42 143
409c57f3
N		 144 int suspended;
145 atomic_t active_io;
1da177e4 146 int ro;
bb4f1e9d
N		 147 int sysfs_active; /* set when sysfs deletes
148 * are happening, so run/
149 * takeover/stop are not safe
150 */
1da177e4
LT		 151
152 struct gendisk *gendisk;
153
eae1701f 154 struct kobject kobj;
d3374825
N		 155 int hold_active;
156#define UNTIL_IOCTL 1
efeb53c0 157#define UNTIL_STOP 2
eae1701f 158
1da177e4
LT		 159 /* Superblock information */
160 int major_version,
161 minor_version,
162 patch_version;
163 int persistent;
e691063a
N		 164 int external; /* metadata is
165 * managed externally */
166 char metadata_type[17]; /* externally set*/
9d8f0363 167 int chunk_sectors;
1da177e4
LT		 168 time_t ctime, utime;
169 int level, layout;
d9d166c2 170 char clevel[16];
1da177e4
LT		 171 int raid_disks;
172 int max_disks;
58c0fed4
AN		 173 sector_t dev_sectors; /* used size of
174 * component devices */
f233ea5c 175 sector_t array_sectors; /* exported array size */
b522adcd
DW		 176 int external_size; /* size managed
177 * externally */
1da177e4 178 __u64 events;
a8707c08
N		 179 /* If the last 'event' was simply a clean->dirty transition, and
180 * we didn't write it to the spares, then it is safe and simple
181 * to just decrement the event count on a dirty->clean transition.
182 * So we record that possibility here.
183 */
184 int can_decrease_events;
1da177e4
LT		 185
186 char uuid[16];
187
f6705578
N		 188 /* If the array is being reshaped, we need to record the
189 * new shape and an indication of where we are up to.
190 * This is written to the superblock.
191 * If reshape_position is MaxSector, then no reshape is happening (yet).
192 */
193 sector_t reshape_position;
664e7c41
AN		 194 int delta_disks, new_level, new_layout;
195 int new_chunk_sectors;
f6705578 196
1da177e4
LT		 197 struct mdk_thread_s *thread; /* management thread */
198 struct mdk_thread_s *sync_thread; /* doing resync or reconstruct */
ff4e8d9a 199 sector_t curr_resync; /* last block scheduled */
97e4f42d
N		 200 /* As resync requests can complete out of order, we cannot easily track
201 * how much resync has been completed. So we occasionally pause until
202 * everything completes, then set curr_resync_completed to curr_resync.
203 * As such it may be well behind the real resync mark, but it is a value
204 * we are certain of.
205 */
206 sector_t curr_resync_completed;
1da177e4
LT		 207 unsigned long resync_mark; /* a recent timestamp */
208 sector_t resync_mark_cnt;/* blocks written at resync_mark */
ff4e8d9a 209 sector_t curr_mark_cnt; /* blocks scheduled now */
1da177e4
LT		 210
211 sector_t resync_max_sectors; /* may be set by personality */
9d88883e
N		 212
213 sector_t resync_mismatches; /* count of sectors where
214 * parity/replica mismatch found
215 */
e464eafd
N		 216
217 /* allow user-space to request suspension of IO to regions of the array */
218 sector_t suspend_lo;
219 sector_t suspend_hi;
88202a0c
N		 220 /* if zero, use the system-wide default */
221 int sync_speed_min;
222 int sync_speed_max;
223
90b08710
BS		 224 /* resync even though the same disks are shared among md-devices */
225 int parallel_resync;
226
6ff8d8ec 227 int ok_start_degraded;
1da177e4
LT		 228 /* recovery/resync flags
229 * NEEDED: we might need to start a resync/recover
230 * RUNNING: a thread is running, or about to be started
231 * SYNC: actually doing a resync, not a recovery
72a23c21 232 * RECOVER: doing recovery, or need to try it.
dfc70645 233 * INTR: resync needs to be aborted for some reason
1da177e4 234 * DONE: thread is done and is waiting to be reaped
24dd469d 235 * REQUEST: user-space has requested a sync (used with SYNC)
411c9403 236 * CHECK: user-space request for check-only, no repair
ccfcc3c1
N		 237 * RESHAPE: A reshape is happening
238 *
239 * If neither SYNC or RESHAPE are set, then it is a recovery.
1da177e4
LT		 240 */
241#define MD_RECOVERY_RUNNING 0
242#define MD_RECOVERY_SYNC 1
72a23c21 243#define MD_RECOVERY_RECOVER 2
1da177e4
LT		 244#define MD_RECOVERY_INTR 3
245#define MD_RECOVERY_DONE 4
246#define MD_RECOVERY_NEEDED 5
24dd469d
N		 247#define MD_RECOVERY_REQUESTED 6
248#define MD_RECOVERY_CHECK 7
ccfcc3c1 249#define MD_RECOVERY_RESHAPE 8
5fd6c1dc
N		 250#define MD_RECOVERY_FROZEN 9
251
1da177e4 252 unsigned long recovery;
4044ba58
N		 253 int recovery_disabled; /* if we detect that recovery
254 * will always fail, set this
255 * so we don't loop trying */
1da177e4
LT		 256
257 int in_sync; /* know to not need resync */
c8c00a69
N		 258 /* 'open_mutex' avoids races between 'md_open' and 'do_md_stop', so
259 * that we are never stopping an array while it is open.
260 * 'reconfig_mutex' protects all other reconfiguration.
261 * These locks are separate due to conflicting interactions
262 * with bdev->bd_mutex.
263 * Lock ordering is:
264 * reconfig_mutex -> bd_mutex : e.g. do_md_run -> revalidate_disk
265 * bd_mutex -> open_mutex: e.g. __blkdev_get -> md_open
266 */
267 struct mutex open_mutex;
df5b89b3 268 struct mutex reconfig_mutex;
f2ea68cf
N		 269 atomic_t active; /* general refcount */
270 atomic_t openers; /* number of active opens */
1da177e4 271
1da177e4
LT		 272 int degraded; /* whether md should consider
273 * adding a spare
274 */
275
276 atomic_t recovery_active; /* blocks scheduled, but not written */
277 wait_queue_head_t recovery_wait;
278 sector_t recovery_cp;
5e96ee65
NB		 279 sector_t resync_min; /* user requested sync
280 * starts here */
c6207277
N		 281 sector_t resync_max; /* resync should pause
282 * when it gets here */
06d91a5f 283
b62b7590
N		 284 struct sysfs_dirent *sysfs_state; /* handle for 'array_state'
285 * file in sysfs.
286 */
0c3573f1 287 struct sysfs_dirent *sysfs_action; /* handle for 'sync_action' */
b62b7590 288
d3374825
N		 289 struct work_struct del_work; /* used for delayed sysfs removal */
290
06d91a5f 291 spinlock_t write_lock;
3d310eb7 292 wait_queue_head_t sb_wait; /* for waiting on superblock updates */
7bfa19f2 293 atomic_t pending_writes; /* number of active superblock writes */
06d91a5f 294
1da177e4
LT		 295 unsigned int safemode; /* if set, update "clean" superblock
296 * when no writes pending.
297 */
298 unsigned int safemode_delay;
299 struct timer_list safemode_timer;
300 atomic_t writes_pending;
165125e1 301 struct request_queue *queue; /* for plugging ... */
1da177e4 302
32a7627c 303 struct bitmap *bitmap; /* the bitmap for the device */
c3d9714e
N		 304 struct {
305 struct file *file; /* the bitmap file */
f6af949c 306 loff_t offset; /* offset from superblock of
c3d9714e
N		 307 * start of bitmap. May be
308 * negative, but not '0'
f6af949c
N		 309 * For external metadata, offset
310 * from start of device.
c3d9714e 311 */
f6af949c 312 loff_t default_offset; /* this is the offset to use when
c3d9714e
N		 313 * hot-adding a bitmap. It should
314 * eventually be settable by sysfs.
315 */
e384e585
N		 316 /* When md is serving under dm, it might use a
317 * dirty_log to store the bits.
318 */
319 struct dm_dirty_log *log;
320
c3d9714e 321 struct mutex mutex;
42a04b50 322 unsigned long chunksize;
ac2f40be 323 unsigned long daemon_sleep; /* how many jiffies between updates? */
42a04b50 324 unsigned long max_write_behind; /* write-behind mode */
ece5cff0 325 int external;
c3d9714e 326 } bitmap_info;
32a7627c 327
1e50915f 328 atomic_t max_corr_read_errors; /* max read retries */
1da177e4 329 struct list_head all_mddevs;
a2826aa9 330
a64c876f 331 struct attribute_group *to_remove;
252ac522
N		 332 struct plug_handle *plug; /* if used by personality */
333
a167f663
N		 334 struct bio_set *bio_set;
335
e9c7469b
TH		 336 /* Generic flush handling.
337 * The last to finish preflush schedules a worker to submit
338 * the rest of the request (without the REQ_FLUSH flag).
a2826aa9 339 */
e9c7469b 340 struct bio *flush_bio;
a2826aa9 341 atomic_t flush_pending;
e9c7469b 342 struct work_struct flush_work;
768a418d 343 struct work_struct event_work; /* used by dm to report failure event */
1da177e4
LT		 344};
345
346
347static inline void rdev_dec_pending(mdk_rdev_t *rdev, mddev_t *mddev)
348{
b2d444d7 349 int faulty = test_bit(Faulty, &rdev->flags);
1da177e4
LT		 350 if (atomic_dec_and_test(&rdev->nr_pending) && faulty)
351 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
352}
353
354static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors)
355{
356 atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io);
357}
358
2604b703 359struct mdk_personality
1da177e4
LT		 360{
361 char *name;
2604b703
N		 362 int level;
363 struct list_head list;
1da177e4 364 struct module *owner;
21a52c6d 365 int (*make_request)(mddev_t *mddev, struct bio *bio);
1da177e4
LT		 366 int (*run)(mddev_t *mddev);
367 int (*stop)(mddev_t *mddev);
368 void (*status)(struct seq_file *seq, mddev_t *mddev);
369 /* error_handler must set ->faulty and clear ->in_sync
370 * if appropriate, and should abort recovery if needed
371 */
372 void (*error_handler)(mddev_t *mddev, mdk_rdev_t *rdev);
373 int (*hot_add_disk) (mddev_t *mddev, mdk_rdev_t *rdev);
374 int (*hot_remove_disk) (mddev_t *mddev, int number);
375 int (*spare_active) (mddev_t *mddev);
57afd89f 376 sector_t (*sync_request)(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster);
1da177e4 377 int (*resize) (mddev_t *mddev, sector_t sectors);
80c3a6ce 378 sector_t (*size) (mddev_t *mddev, sector_t sectors, int raid_disks);
63c70c4f
N		 379 int (*check_reshape) (mddev_t *mddev);
380 int (*start_reshape) (mddev_t *mddev);
cea9c228 381 void (*finish_reshape) (mddev_t *mddev);
36fa3063
N		 382 /* quiesce moves between quiescence states
383 * 0 - fully active
384 * 1 - no new requests allowed
385 * others - reserved
386 */
387 void (*quiesce) (mddev_t *mddev, int state);
245f46c2
N		 388 /* takeover is used to transition an array from one
389 * personality to another. The new personality must be able
390 * to handle the data in the current layout.
391 * e.g. 2drive raid1 -> 2drive raid5
392 * ndrive raid5 -> degraded n+1drive raid6 with special layout
393 * If the takeover succeeds, a new 'private' structure is returned.
394 * This needs to be installed and then ->run used to activate the
395 * array.
396 */
397 void *(*takeover) (mddev_t *mddev);
1da177e4
LT		 398};
399
400
007583c9
N		 401struct md_sysfs_entry {
402 struct attribute attr;
403 ssize_t (*show)(mddev_t *, char *);
404 ssize_t (*store)(mddev_t *, const char *, size_t);
405};
43a70507 406extern struct attribute_group md_bitmap_group;
007583c9 407
00bcb4ac
N		 408static inline struct sysfs_dirent *sysfs_get_dirent_safe(struct sysfs_dirent *sd, char *name)
409{
410 if (sd)
411 return sysfs_get_dirent(sd, NULL, name);
412 return sd;
413}
414static inline void sysfs_notify_dirent_safe(struct sysfs_dirent *sd)
415{
416 if (sd)
417 sysfs_notify_dirent(sd);
418}
419
1da177e4
LT		 420static inline char * mdname (mddev_t * mddev)
421{
422 return mddev->gendisk ? mddev->gendisk->disk_name : "mdX";
423}
424
1da177e4
LT		 425/*
426 * iterates through some rdev ringlist. It's safe to remove the
427 * current 'rdev'. Dont touch 'tmp' though.
428 */
159ec1fc
CR		 429#define rdev_for_each_list(rdev, tmp, head) \
430 list_for_each_entry_safe(rdev, tmp, head, same_set)
431
1da177e4
LT		 432/*
433 * iterates through the 'same array disks' ringlist
434 */
d089c6af 435#define rdev_for_each(rdev, tmp, mddev) \
159ec1fc 436 list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set)
1da177e4 437
4b80991c
N		 438#define rdev_for_each_rcu(rdev, mddev) \
439 list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set)
440
1da177e4
LT		 441typedef struct mdk_thread_s {
442 void (*run) (mddev_t *mddev);
443 mddev_t *mddev;
444 wait_queue_head_t wqueue;
445 unsigned long flags;
1da177e4 446 struct task_struct *tsk;
32a7627c 447 unsigned long timeout;
1da177e4
LT		 448} mdk_thread_t;
449
450#define THREAD_WAKEUP 0
451
452#define __wait_event_lock_irq(wq, condition, lock, cmd) \
453do { \
454 wait_queue_t __wait; \
455 init_waitqueue_entry(&__wait, current); \
456 \
457 add_wait_queue(&wq, &__wait); \
458 for (;;) { \
459 set_current_state(TASK_UNINTERRUPTIBLE); \
460 if (condition) \
461 break; \
462 spin_unlock_irq(&lock); \
463 cmd; \
464 schedule(); \
465 spin_lock_irq(&lock); \
466 } \
467 current->state = TASK_RUNNING; \
468 remove_wait_queue(&wq, &__wait); \
469} while (0)
470
471#define wait_event_lock_irq(wq, condition, lock, cmd) \
472do { \
473 if (condition) \
474 break; \
475 __wait_event_lock_irq(wq, condition, lock, cmd); \
476} while (0)
477
1345b1d8
N		 478static inline void safe_put_page(struct page *p)
479{
480 if (p) put_page(p);
481}
482
92022950
N		 483extern int register_md_personality(struct mdk_personality *p);
484extern int unregister_md_personality(struct mdk_personality *p);
485extern mdk_thread_t * md_register_thread(void (*run) (mddev_t *mddev),
486 mddev_t *mddev, const char *name);
487extern void md_unregister_thread(mdk_thread_t *thread);
488extern void md_wakeup_thread(mdk_thread_t *thread);
489extern void md_check_recovery(mddev_t *mddev);
490extern void md_write_start(mddev_t *mddev, struct bio *bi);
491extern void md_write_end(mddev_t *mddev);
492extern void md_done_sync(mddev_t *mddev, int blocks, int ok);
493extern void md_error(mddev_t *mddev, mdk_rdev_t *rdev);
494
3fa841d7 495extern int mddev_congested(mddev_t *mddev, int bits);
e9c7469b 496extern void md_flush_request(mddev_t *mddev, struct bio *bio);
92022950
N		 497extern void md_super_write(mddev_t *mddev, mdk_rdev_t *rdev,
498 sector_t sector, int size, struct page *page);
499extern void md_super_wait(mddev_t *mddev);
2b193363 500extern int sync_page_io(mdk_rdev_t *rdev, sector_t sector, int size,
92022950
N		 501 struct page *page, int rw);
502extern void md_do_sync(mddev_t *mddev);
503extern void md_new_event(mddev_t *mddev);
504extern int md_allow_write(mddev_t *mddev);
505extern void md_wait_for_blocked_rdev(mdk_rdev_t *rdev, mddev_t *mddev);
1f403624 506extern void md_set_array_sectors(mddev_t *mddev, sector_t array_sectors);
0894cc30 507extern int md_check_no_bitmap(mddev_t *mddev);
ac5e7113 508extern int md_integrity_register(mddev_t *mddev);
72e02075
N		 509extern void md_integrity_add_rdev(mdk_rdev_t *rdev, mddev_t *mddev);
510extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale);
43a70507 511extern void restore_bitmap_write_access(struct file *file);
b63d7c2e 512extern void md_unplug(mddev_t *mddev);
63fe0817 513
390ee602
N		 514extern void mddev_init(mddev_t *mddev);
515extern int md_run(mddev_t *mddev);
516extern void md_stop(mddev_t *mddev);
517extern void md_stop_writes(mddev_t *mddev);
e8bb9a83 518extern void md_rdev_init(mdk_rdev_t *rdev);
63fe0817 519
390ee602
N		 520extern void mddev_suspend(mddev_t *mddev);
521extern void mddev_resume(mddev_t *mddev);
a167f663
N		 522extern struct bio *bio_clone_mddev(struct bio *bio, gfp_t gfp_mask,
523 mddev_t *mddev);
524extern struct bio *bio_alloc_mddev(gfp_t gfp_mask, int nr_iovecs,
525 mddev_t *mddev);
63fe0817 526#endif /* _MD_MD_H */
Clone of davem's net-next-2.6 tree