]> 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://oss.sgi.com/xfs/xfs
[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 */
7b6d91da 90#define BarriersNotsupp 5 /* REQ_HARDBARRIER is not supported */
c5d79adb
N		 91#define AllReserved 6 /* If whole device is reserved for
92 * one array */
d0fae18f 93#define AutoDetected 7 /* added by auto-detect */
6bfe0b49
DW		 94#define Blocked 8 /* An error occured on an externally
95 * managed array, don't allow writes
96 * until it is cleared */
97 wait_queue_head_t blocked_wait;
8ddf9efe 98
1da177e4
LT		 99 int desc_nr; /* descriptor index in the superblock */
100 int raid_disk; /* role of device in array */
e93f68a1
N		 101 int new_raid_disk; /* role that the device will have in
102 * the array after a level-change completes.
103 */
41158c7e
N		 104 int saved_raid_disk; /* role that device used to have in the
105 * array and could again if we did a partial
106 * resync from the bitmap
107 */
5fd6c1dc
N		 108 sector_t recovery_offset;/* If this device has been partially
109 * recovered, this is where we were
110 * up to.
111 */
1da177e4
LT		 112
113 atomic_t nr_pending; /* number of pending requests.
114 * only maintained for arrays that
115 * support hot removal
116 */
ba22dcbf
N		 117 atomic_t read_errors; /* number of consecutive read errors that
118 * we have tried to ignore.
119 */
1e50915f
RB		 120 struct timespec last_read_error; /* monotonic time since our
121 * last read error
122 */
4dbcdc75
N		 123 atomic_t corrected_errors; /* number of corrected read errors,
124 * for reporting to userspace and storing
125 * in superblock.
126 */
5792a285 127 struct work_struct del_work; /* used for delayed sysfs removal */
3c0ee63a
N		 128
129 struct sysfs_dirent *sysfs_state; /* handle for 'state'
130 * sysfs entry */
1da177e4
LT		 131};
132
1da177e4
LT		 133struct mddev_s
134{
135 void *private;
2604b703 136 struct mdk_personality *pers;
1da177e4
LT		 137 dev_t unit;
138 int md_minor;
139 struct list_head disks;
850b2b42
N		 140 unsigned long flags;
141#define MD_CHANGE_DEVS 0 /* Some device status has changed */
142#define MD_CHANGE_CLEAN 1 /* transition to or from 'clean' */
070dc6dd 143#define MD_CHANGE_PENDING 2 /* switch from 'clean' to 'active' in progress */
850b2b42 144
409c57f3
N		 145 int suspended;
146 atomic_t active_io;
1da177e4 147 int ro;
bb4f1e9d
N		 148 int sysfs_active; /* set when sysfs deletes
149 * are happening, so run/
150 * takeover/stop are not safe
151 */
1da177e4
LT		 152
153 struct gendisk *gendisk;
154
eae1701f 155 struct kobject kobj;
d3374825
N		 156 int hold_active;
157#define UNTIL_IOCTL 1
efeb53c0 158#define UNTIL_STOP 2
eae1701f 159
1da177e4
LT		 160 /* Superblock information */
161 int major_version,
162 minor_version,
163 patch_version;
164 int persistent;
e691063a
N		 165 int external; /* metadata is
166 * managed externally */
167 char metadata_type[17]; /* externally set*/
9d8f0363 168 int chunk_sectors;
1da177e4
LT		 169 time_t ctime, utime;
170 int level, layout;
d9d166c2 171 char clevel[16];
1da177e4
LT		 172 int raid_disks;
173 int max_disks;
58c0fed4
AN		 174 sector_t dev_sectors; /* used size of
175 * component devices */
f233ea5c 176 sector_t array_sectors; /* exported array size */
b522adcd
DW		 177 int external_size; /* size managed
178 * externally */
1da177e4 179 __u64 events;
a8707c08
N		 180 /* If the last 'event' was simply a clean->dirty transition, and
181 * we didn't write it to the spares, then it is safe and simple
182 * to just decrement the event count on a dirty->clean transition.
183 * So we record that possibility here.
184 */
185 int can_decrease_events;
1da177e4
LT		 186
187 char uuid[16];
188
f6705578
N		 189 /* If the array is being reshaped, we need to record the
190 * new shape and an indication of where we are up to.
191 * This is written to the superblock.
192 * If reshape_position is MaxSector, then no reshape is happening (yet).
193 */
194 sector_t reshape_position;
664e7c41
AN		 195 int delta_disks, new_level, new_layout;
196 int new_chunk_sectors;
f6705578 197
1da177e4
LT		 198 struct mdk_thread_s *thread; /* management thread */
199 struct mdk_thread_s *sync_thread; /* doing resync or reconstruct */
ff4e8d9a 200 sector_t curr_resync; /* last block scheduled */
97e4f42d
N		 201 /* As resync requests can complete out of order, we cannot easily track
202 * how much resync has been completed. So we occasionally pause until
203 * everything completes, then set curr_resync_completed to curr_resync.
204 * As such it may be well behind the real resync mark, but it is a value
205 * we are certain of.
206 */
207 sector_t curr_resync_completed;
1da177e4
LT		 208 unsigned long resync_mark; /* a recent timestamp */
209 sector_t resync_mark_cnt;/* blocks written at resync_mark */
ff4e8d9a 210 sector_t curr_mark_cnt; /* blocks scheduled now */
1da177e4
LT		 211
212 sector_t resync_max_sectors; /* may be set by personality */
9d88883e
N		 213
214 sector_t resync_mismatches; /* count of sectors where
215 * parity/replica mismatch found
216 */
e464eafd
N		 217
218 /* allow user-space to request suspension of IO to regions of the array */
219 sector_t suspend_lo;
220 sector_t suspend_hi;
88202a0c
N		 221 /* if zero, use the system-wide default */
222 int sync_speed_min;
223 int sync_speed_max;
224
90b08710
BS		 225 /* resync even though the same disks are shared among md-devices */
226 int parallel_resync;
227
6ff8d8ec 228 int ok_start_degraded;
1da177e4
LT		 229 /* recovery/resync flags
230 * NEEDED: we might need to start a resync/recover
231 * RUNNING: a thread is running, or about to be started
232 * SYNC: actually doing a resync, not a recovery
72a23c21 233 * RECOVER: doing recovery, or need to try it.
dfc70645 234 * INTR: resync needs to be aborted for some reason
1da177e4 235 * DONE: thread is done and is waiting to be reaped
24dd469d 236 * REQUEST: user-space has requested a sync (used with SYNC)
411c9403 237 * CHECK: user-space request for check-only, no repair
ccfcc3c1
N		 238 * RESHAPE: A reshape is happening
239 *
240 * If neither SYNC or RESHAPE are set, then it is a recovery.
1da177e4
LT		 241 */
242#define MD_RECOVERY_RUNNING 0
243#define MD_RECOVERY_SYNC 1
72a23c21 244#define MD_RECOVERY_RECOVER 2
1da177e4
LT		 245#define MD_RECOVERY_INTR 3
246#define MD_RECOVERY_DONE 4
247#define MD_RECOVERY_NEEDED 5
24dd469d
N		 248#define MD_RECOVERY_REQUESTED 6
249#define MD_RECOVERY_CHECK 7
ccfcc3c1 250#define MD_RECOVERY_RESHAPE 8
5fd6c1dc
N		 251#define MD_RECOVERY_FROZEN 9
252
1da177e4 253 unsigned long recovery;
4044ba58
N		 254 int recovery_disabled; /* if we detect that recovery
255 * will always fail, set this
256 * so we don't loop trying */
1da177e4
LT		 257
258 int in_sync; /* know to not need resync */
c8c00a69
N		 259 /* 'open_mutex' avoids races between 'md_open' and 'do_md_stop', so
260 * that we are never stopping an array while it is open.
261 * 'reconfig_mutex' protects all other reconfiguration.
262 * These locks are separate due to conflicting interactions
263 * with bdev->bd_mutex.
264 * Lock ordering is:
265 * reconfig_mutex -> bd_mutex : e.g. do_md_run -> revalidate_disk
266 * bd_mutex -> open_mutex: e.g. __blkdev_get -> md_open
267 */
268 struct mutex open_mutex;
df5b89b3 269 struct mutex reconfig_mutex;
f2ea68cf
N		 270 atomic_t active; /* general refcount */
271 atomic_t openers; /* number of active opens */
1da177e4 272
1da177e4
LT		 273 int degraded; /* whether md should consider
274 * adding a spare
275 */
a9701a30
N		 276 int barriers_work; /* initialised to true, cleared as soon
277 * as a barrier request to slave
278 * fails. Only supported
279 */
280 struct bio *biolist; /* bios that need to be retried
7b6d91da 281 * because REQ_HARDBARRIER is not supported
a9701a30 282 */
1da177e4
LT		 283
284 atomic_t recovery_active; /* blocks scheduled, but not written */
285 wait_queue_head_t recovery_wait;
286 sector_t recovery_cp;
5e96ee65
NB		 287 sector_t resync_min; /* user requested sync
288 * starts here */
c6207277
N		 289 sector_t resync_max; /* resync should pause
290 * when it gets here */
06d91a5f 291
b62b7590
N		 292 struct sysfs_dirent *sysfs_state; /* handle for 'array_state'
293 * file in sysfs.
294 */
0c3573f1 295 struct sysfs_dirent *sysfs_action; /* handle for 'sync_action' */
b62b7590 296
d3374825
N		 297 struct work_struct del_work; /* used for delayed sysfs removal */
298
06d91a5f 299 spinlock_t write_lock;
3d310eb7 300 wait_queue_head_t sb_wait; /* for waiting on superblock updates */
7bfa19f2 301 atomic_t pending_writes; /* number of active superblock writes */
06d91a5f 302
1da177e4
LT		 303 unsigned int safemode; /* if set, update "clean" superblock
304 * when no writes pending.
305 */
306 unsigned int safemode_delay;
307 struct timer_list safemode_timer;
308 atomic_t writes_pending;
165125e1 309 struct request_queue *queue; /* for plugging ... */
1da177e4 310
32a7627c 311 struct bitmap *bitmap; /* the bitmap for the device */
c3d9714e
N		 312 struct {
313 struct file *file; /* the bitmap file */
f6af949c 314 loff_t offset; /* offset from superblock of
c3d9714e
N		 315 * start of bitmap. May be
316 * negative, but not '0'
f6af949c
N		 317 * For external metadata, offset
318 * from start of device.
c3d9714e 319 */
f6af949c 320 loff_t default_offset; /* this is the offset to use when
c3d9714e
N		 321 * hot-adding a bitmap. It should
322 * eventually be settable by sysfs.
323 */
e384e585
N		 324 /* When md is serving under dm, it might use a
325 * dirty_log to store the bits.
326 */
327 struct dm_dirty_log *log;
328
c3d9714e 329 struct mutex mutex;
42a04b50 330 unsigned long chunksize;
ac2f40be 331 unsigned long daemon_sleep; /* how many jiffies between updates? */
42a04b50 332 unsigned long max_write_behind; /* write-behind mode */
ece5cff0 333 int external;
c3d9714e 334 } bitmap_info;
32a7627c 335
1e50915f 336 atomic_t max_corr_read_errors; /* max read retries */
1da177e4 337 struct list_head all_mddevs;
a2826aa9 338
a64c876f 339 struct attribute_group *to_remove;
252ac522
N		 340 struct plug_handle *plug; /* if used by personality */
341
a2826aa9
N		 342 /* Generic barrier handling.
343 * If there is a pending barrier request, all other
344 * writes are blocked while the devices are flushed.
345 * The last to finish a flush schedules a worker to
346 * submit the barrier request (without the barrier flag),
347 * then submit more flush requests.
348 */
349 struct bio *barrier;
350 atomic_t flush_pending;
351 struct work_struct barrier_work;
768a418d 352 struct work_struct event_work; /* used by dm to report failure event */
1da177e4
LT		 353};
354
355
356static inline void rdev_dec_pending(mdk_rdev_t *rdev, mddev_t *mddev)
357{
b2d444d7 358 int faulty = test_bit(Faulty, &rdev->flags);
1da177e4
LT		 359 if (atomic_dec_and_test(&rdev->nr_pending) && faulty)
360 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
361}
362
363static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors)
364{
365 atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io);
366}
367
2604b703 368struct mdk_personality
1da177e4
LT		 369{
370 char *name;
2604b703
N		 371 int level;
372 struct list_head list;
1da177e4 373 struct module *owner;
21a52c6d 374 int (*make_request)(mddev_t *mddev, struct bio *bio);
1da177e4
LT		 375 int (*run)(mddev_t *mddev);
376 int (*stop)(mddev_t *mddev);
377 void (*status)(struct seq_file *seq, mddev_t *mddev);
378 /* error_handler must set ->faulty and clear ->in_sync
379 * if appropriate, and should abort recovery if needed
380 */
381 void (*error_handler)(mddev_t *mddev, mdk_rdev_t *rdev);
382 int (*hot_add_disk) (mddev_t *mddev, mdk_rdev_t *rdev);
383 int (*hot_remove_disk) (mddev_t *mddev, int number);
384 int (*spare_active) (mddev_t *mddev);
57afd89f 385 sector_t (*sync_request)(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster);
1da177e4 386 int (*resize) (mddev_t *mddev, sector_t sectors);
80c3a6ce 387 sector_t (*size) (mddev_t *mddev, sector_t sectors, int raid_disks);
63c70c4f
N		 388 int (*check_reshape) (mddev_t *mddev);
389 int (*start_reshape) (mddev_t *mddev);
cea9c228 390 void (*finish_reshape) (mddev_t *mddev);
36fa3063
N		 391 /* quiesce moves between quiescence states
392 * 0 - fully active
393 * 1 - no new requests allowed
394 * others - reserved
395 */
396 void (*quiesce) (mddev_t *mddev, int state);
245f46c2
N		 397 /* takeover is used to transition an array from one
398 * personality to another. The new personality must be able
399 * to handle the data in the current layout.
400 * e.g. 2drive raid1 -> 2drive raid5
401 * ndrive raid5 -> degraded n+1drive raid6 with special layout
402 * If the takeover succeeds, a new 'private' structure is returned.
403 * This needs to be installed and then ->run used to activate the
404 * array.
405 */
406 void *(*takeover) (mddev_t *mddev);
1da177e4
LT		 407};
408
409
007583c9
N		 410struct md_sysfs_entry {
411 struct attribute attr;
412 ssize_t (*show)(mddev_t *, char *);
413 ssize_t (*store)(mddev_t *, const char *, size_t);
414};
43a70507 415extern struct attribute_group md_bitmap_group;
007583c9 416
00bcb4ac
N		 417static inline struct sysfs_dirent *sysfs_get_dirent_safe(struct sysfs_dirent *sd, char *name)
418{
419 if (sd)
420 return sysfs_get_dirent(sd, NULL, name);
421 return sd;
422}
423static inline void sysfs_notify_dirent_safe(struct sysfs_dirent *sd)
424{
425 if (sd)
426 sysfs_notify_dirent(sd);
427}
428
1da177e4
LT		 429static inline char * mdname (mddev_t * mddev)
430{
431 return mddev->gendisk ? mddev->gendisk->disk_name : "mdX";
432}
433
1da177e4
LT		 434/*
435 * iterates through some rdev ringlist. It's safe to remove the
436 * current 'rdev'. Dont touch 'tmp' though.
437 */
159ec1fc
CR		 438#define rdev_for_each_list(rdev, tmp, head) \
439 list_for_each_entry_safe(rdev, tmp, head, same_set)
440
1da177e4
LT		 441/*
442 * iterates through the 'same array disks' ringlist
443 */
d089c6af 444#define rdev_for_each(rdev, tmp, mddev) \
159ec1fc 445 list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set)
1da177e4 446
4b80991c
N		 447#define rdev_for_each_rcu(rdev, mddev) \
448 list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set)
449
1da177e4
LT		 450typedef struct mdk_thread_s {
451 void (*run) (mddev_t *mddev);
452 mddev_t *mddev;
453 wait_queue_head_t wqueue;
454 unsigned long flags;
1da177e4 455 struct task_struct *tsk;
32a7627c 456 unsigned long timeout;
1da177e4
LT		 457} mdk_thread_t;
458
459#define THREAD_WAKEUP 0
460
461#define __wait_event_lock_irq(wq, condition, lock, cmd) \
462do { \
463 wait_queue_t __wait; \
464 init_waitqueue_entry(&__wait, current); \
465 \
466 add_wait_queue(&wq, &__wait); \
467 for (;;) { \
468 set_current_state(TASK_UNINTERRUPTIBLE); \
469 if (condition) \
470 break; \
471 spin_unlock_irq(&lock); \
472 cmd; \
473 schedule(); \
474 spin_lock_irq(&lock); \
475 } \
476 current->state = TASK_RUNNING; \
477 remove_wait_queue(&wq, &__wait); \
478} while (0)
479
480#define wait_event_lock_irq(wq, condition, lock, cmd) \
481do { \
482 if (condition) \
483 break; \
484 __wait_event_lock_irq(wq, condition, lock, cmd); \
485} while (0)
486
1345b1d8
N		 487static inline void safe_put_page(struct page *p)
488{
489 if (p) put_page(p);
490}
491
92022950
N		 492extern int register_md_personality(struct mdk_personality *p);
493extern int unregister_md_personality(struct mdk_personality *p);
494extern mdk_thread_t * md_register_thread(void (*run) (mddev_t *mddev),
495 mddev_t *mddev, const char *name);
496extern void md_unregister_thread(mdk_thread_t *thread);
497extern void md_wakeup_thread(mdk_thread_t *thread);
498extern void md_check_recovery(mddev_t *mddev);
499extern void md_write_start(mddev_t *mddev, struct bio *bi);
500extern void md_write_end(mddev_t *mddev);
501extern void md_done_sync(mddev_t *mddev, int blocks, int ok);
502extern void md_error(mddev_t *mddev, mdk_rdev_t *rdev);
503
3fa841d7 504extern int mddev_congested(mddev_t *mddev, int bits);
a2826aa9 505extern void md_barrier_request(mddev_t *mddev, struct bio *bio);
92022950
N		 506extern void md_super_write(mddev_t *mddev, mdk_rdev_t *rdev,
507 sector_t sector, int size, struct page *page);
508extern void md_super_wait(mddev_t *mddev);
509extern int sync_page_io(struct block_device *bdev, sector_t sector, int size,
510 struct page *page, int rw);
511extern void md_do_sync(mddev_t *mddev);
512extern void md_new_event(mddev_t *mddev);
513extern int md_allow_write(mddev_t *mddev);
514extern void md_wait_for_blocked_rdev(mdk_rdev_t *rdev, mddev_t *mddev);
1f403624 515extern void md_set_array_sectors(mddev_t *mddev, sector_t array_sectors);
0894cc30 516extern int md_check_no_bitmap(mddev_t *mddev);
ac5e7113 517extern int md_integrity_register(mddev_t *mddev);
72e02075
N		 518extern void md_integrity_add_rdev(mdk_rdev_t *rdev, mddev_t *mddev);
519extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale);
43a70507 520extern void restore_bitmap_write_access(struct file *file);
b63d7c2e 521extern void md_unplug(mddev_t *mddev);
63fe0817 522
390ee602
N		 523extern void mddev_init(mddev_t *mddev);
524extern int md_run(mddev_t *mddev);
525extern void md_stop(mddev_t *mddev);
526extern void md_stop_writes(mddev_t *mddev);
e8bb9a83 527extern void md_rdev_init(mdk_rdev_t *rdev);
63fe0817 528
390ee602
N		 529extern void mddev_suspend(mddev_t *mddev);
530extern void mddev_resume(mddev_t *mddev);
63fe0817 531#endif /* _MD_MD_H */
Clone of davem's net-next-2.6 tree