2 md_k.h : kernel internal structure of the Linux MD driver
3 Copyright (C) 1996-98 Ingo Molnar, Gadi Oxman
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)
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.
18 #include <linux/blkdev.h>
19 #include <linux/kobject.h>
20 #include <linux/list.h>
22 #include <linux/mutex.h>
23 #include <linux/timer.h>
24 #include <linux/wait.h>
25 #include <linux/workqueue.h>
27 #define MaxSector (~(sector_t)0)
29 typedef struct mddev_s mddev_t
;
30 typedef struct mdk_rdev_s mdk_rdev_t
;
32 /* generic plugging support - like that provided with request_queue,
33 * but does not require a request_queue
36 void (*unplug_fn
)(struct plug_handle
*);
37 struct timer_list unplug_timer
;
38 struct work_struct unplug_work
;
39 unsigned long unplug_flag
;
41 #define PLUGGED_FLAG 1
42 void plugger_init(struct plug_handle
*plug
,
43 void (*unplug_fn
)(struct plug_handle
*));
44 void plugger_set_plug(struct plug_handle
*plug
);
45 int plugger_remove_plug(struct plug_handle
*plug
);
46 static inline void plugger_flush(struct plug_handle
*plug
)
48 del_timer_sync(&plug
->unplug_timer
);
49 cancel_work_sync(&plug
->unplug_work
);
53 * MD's 'extended' device
57 struct list_head same_set
; /* RAID devices within the same set */
59 sector_t sectors
; /* Device size (in 512bytes sectors) */
60 mddev_t
*mddev
; /* RAID array if running */
61 int last_events
; /* IO event timestamp */
63 struct block_device
*bdev
; /* block device handle */
68 sector_t data_offset
; /* start of data in array */
69 sector_t sb_start
; /* offset of the super block (in 512byte sectors) */
70 int sb_size
; /* bytes in the superblock */
71 int preferred_minor
; /* autorun support */
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
80 * faulty==0 in_sync==0
82 * It can never have faulty==1, in_sync==1
83 * This reduces the burden of testing multiple flags in many cases
87 #define Faulty 1 /* device is known to have a fault */
88 #define In_sync 2 /* device is in_sync with rest of array */
89 #define WriteMostly 4 /* Avoid reading if at all possible */
90 #define AllReserved 6 /* If whole device is reserved for
92 #define AutoDetected 7 /* added by auto-detect */
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
;
98 int desc_nr
; /* descriptor index in the superblock */
99 int raid_disk
; /* role of device in array */
100 int new_raid_disk
; /* role that the device will have in
101 * the array after a level-change completes.
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
107 sector_t recovery_offset
;/* If this device has been partially
108 * recovered, this is where we were
112 atomic_t nr_pending
; /* number of pending requests.
113 * only maintained for arrays that
114 * support hot removal
116 atomic_t read_errors
; /* number of consecutive read errors that
117 * we have tried to ignore.
119 struct timespec last_read_error
; /* monotonic time since our
122 atomic_t corrected_errors
; /* number of corrected read errors,
123 * for reporting to userspace and storing
126 struct work_struct del_work
; /* used for delayed sysfs removal */
128 struct sysfs_dirent
*sysfs_state
; /* handle for 'state'
135 struct mdk_personality
*pers
;
138 struct list_head disks
;
140 #define MD_CHANGE_DEVS 0 /* Some device status has changed */
141 #define MD_CHANGE_CLEAN 1 /* transition to or from 'clean' */
142 #define MD_CHANGE_PENDING 2 /* switch from 'clean' to 'active' in progress */
147 int sysfs_active
; /* set when sysfs deletes
148 * are happening, so run/
149 * takeover/stop are not safe
151 int ready
; /* See when safe to pass
152 * IO requests down */
153 struct gendisk
*gendisk
;
157 #define UNTIL_IOCTL 1
160 /* Superblock information */
165 int external
; /* metadata is
166 * managed externally */
167 char metadata_type
[17]; /* externally set*/
174 sector_t dev_sectors
; /* used size of
175 * component devices */
176 sector_t array_sectors
; /* exported array size */
177 int external_size
; /* size managed
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.
185 int can_decrease_events
;
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).
194 sector_t reshape_position
;
195 int delta_disks
, new_level
, new_layout
;
196 int new_chunk_sectors
;
198 struct mdk_thread_s
*thread
; /* management thread */
199 struct mdk_thread_s
*sync_thread
; /* doing resync or reconstruct */
200 sector_t curr_resync
; /* last block scheduled */
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
207 sector_t curr_resync_completed
;
208 unsigned long resync_mark
; /* a recent timestamp */
209 sector_t resync_mark_cnt
;/* blocks written at resync_mark */
210 sector_t curr_mark_cnt
; /* blocks scheduled now */
212 sector_t resync_max_sectors
; /* may be set by personality */
214 sector_t resync_mismatches
; /* count of sectors where
215 * parity/replica mismatch found
218 /* allow user-space to request suspension of IO to regions of the array */
221 /* if zero, use the system-wide default */
225 /* resync even though the same disks are shared among md-devices */
228 int ok_start_degraded
;
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
233 * RECOVER: doing recovery, or need to try it.
234 * INTR: resync needs to be aborted for some reason
235 * DONE: thread is done and is waiting to be reaped
236 * REQUEST: user-space has requested a sync (used with SYNC)
237 * CHECK: user-space request for check-only, no repair
238 * RESHAPE: A reshape is happening
240 * If neither SYNC or RESHAPE are set, then it is a recovery.
242 #define MD_RECOVERY_RUNNING 0
243 #define MD_RECOVERY_SYNC 1
244 #define MD_RECOVERY_RECOVER 2
245 #define MD_RECOVERY_INTR 3
246 #define MD_RECOVERY_DONE 4
247 #define MD_RECOVERY_NEEDED 5
248 #define MD_RECOVERY_REQUESTED 6
249 #define MD_RECOVERY_CHECK 7
250 #define MD_RECOVERY_RESHAPE 8
251 #define MD_RECOVERY_FROZEN 9
253 unsigned long recovery
;
254 int recovery_disabled
; /* if we detect that recovery
255 * will always fail, set this
256 * so we don't loop trying */
258 int in_sync
; /* know to not need resync */
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.
265 * reconfig_mutex -> bd_mutex : e.g. do_md_run -> revalidate_disk
266 * bd_mutex -> open_mutex: e.g. __blkdev_get -> md_open
268 struct mutex open_mutex
;
269 struct mutex reconfig_mutex
;
270 atomic_t active
; /* general refcount */
271 atomic_t openers
; /* number of active opens */
273 int changed
; /* True if we might need to
274 * reread partition info */
275 int degraded
; /* whether md should consider
279 atomic_t recovery_active
; /* blocks scheduled, but not written */
280 wait_queue_head_t recovery_wait
;
281 sector_t recovery_cp
;
282 sector_t resync_min
; /* user requested sync
284 sector_t resync_max
; /* resync should pause
285 * when it gets here */
287 struct sysfs_dirent
*sysfs_state
; /* handle for 'array_state'
290 struct sysfs_dirent
*sysfs_action
; /* handle for 'sync_action' */
292 struct work_struct del_work
; /* used for delayed sysfs removal */
294 spinlock_t write_lock
;
295 wait_queue_head_t sb_wait
; /* for waiting on superblock updates */
296 atomic_t pending_writes
; /* number of active superblock writes */
298 unsigned int safemode
; /* if set, update "clean" superblock
299 * when no writes pending.
301 unsigned int safemode_delay
;
302 struct timer_list safemode_timer
;
303 atomic_t writes_pending
;
304 struct request_queue
*queue
; /* for plugging ... */
306 struct bitmap
*bitmap
; /* the bitmap for the device */
308 struct file
*file
; /* the bitmap file */
309 loff_t offset
; /* offset from superblock of
310 * start of bitmap. May be
311 * negative, but not '0'
312 * For external metadata, offset
313 * from start of device.
315 loff_t default_offset
; /* this is the offset to use when
316 * hot-adding a bitmap. It should
317 * eventually be settable by sysfs.
319 /* When md is serving under dm, it might use a
320 * dirty_log to store the bits.
322 struct dm_dirty_log
*log
;
325 unsigned long chunksize
;
326 unsigned long daemon_sleep
; /* how many jiffies between updates? */
327 unsigned long max_write_behind
; /* write-behind mode */
331 atomic_t max_corr_read_errors
; /* max read retries */
332 struct list_head all_mddevs
;
334 struct attribute_group
*to_remove
;
335 struct plug_handle
*plug
; /* if used by personality */
337 struct bio_set
*bio_set
;
339 /* Generic flush handling.
340 * The last to finish preflush schedules a worker to submit
341 * the rest of the request (without the REQ_FLUSH flag).
343 struct bio
*flush_bio
;
344 atomic_t flush_pending
;
345 struct work_struct flush_work
;
346 struct work_struct event_work
; /* used by dm to report failure event */
350 static inline void rdev_dec_pending(mdk_rdev_t
*rdev
, mddev_t
*mddev
)
352 int faulty
= test_bit(Faulty
, &rdev
->flags
);
353 if (atomic_dec_and_test(&rdev
->nr_pending
) && faulty
)
354 set_bit(MD_RECOVERY_NEEDED
, &mddev
->recovery
);
357 static inline void md_sync_acct(struct block_device
*bdev
, unsigned long nr_sectors
)
359 atomic_add(nr_sectors
, &bdev
->bd_contains
->bd_disk
->sync_io
);
362 struct mdk_personality
366 struct list_head list
;
367 struct module
*owner
;
368 int (*make_request
)(mddev_t
*mddev
, struct bio
*bio
);
369 int (*run
)(mddev_t
*mddev
);
370 int (*stop
)(mddev_t
*mddev
);
371 void (*status
)(struct seq_file
*seq
, mddev_t
*mddev
);
372 /* error_handler must set ->faulty and clear ->in_sync
373 * if appropriate, and should abort recovery if needed
375 void (*error_handler
)(mddev_t
*mddev
, mdk_rdev_t
*rdev
);
376 int (*hot_add_disk
) (mddev_t
*mddev
, mdk_rdev_t
*rdev
);
377 int (*hot_remove_disk
) (mddev_t
*mddev
, int number
);
378 int (*spare_active
) (mddev_t
*mddev
);
379 sector_t (*sync_request
)(mddev_t
*mddev
, sector_t sector_nr
, int *skipped
, int go_faster
);
380 int (*resize
) (mddev_t
*mddev
, sector_t sectors
);
381 sector_t (*size
) (mddev_t
*mddev
, sector_t sectors
, int raid_disks
);
382 int (*check_reshape
) (mddev_t
*mddev
);
383 int (*start_reshape
) (mddev_t
*mddev
);
384 void (*finish_reshape
) (mddev_t
*mddev
);
385 /* quiesce moves between quiescence states
387 * 1 - no new requests allowed
390 void (*quiesce
) (mddev_t
*mddev
, int state
);
391 /* takeover is used to transition an array from one
392 * personality to another. The new personality must be able
393 * to handle the data in the current layout.
394 * e.g. 2drive raid1 -> 2drive raid5
395 * ndrive raid5 -> degraded n+1drive raid6 with special layout
396 * If the takeover succeeds, a new 'private' structure is returned.
397 * This needs to be installed and then ->run used to activate the
400 void *(*takeover
) (mddev_t
*mddev
);
404 struct md_sysfs_entry
{
405 struct attribute attr
;
406 ssize_t (*show
)(mddev_t
*, char *);
407 ssize_t (*store
)(mddev_t
*, const char *, size_t);
409 extern struct attribute_group md_bitmap_group
;
411 static inline struct sysfs_dirent
*sysfs_get_dirent_safe(struct sysfs_dirent
*sd
, char *name
)
414 return sysfs_get_dirent(sd
, NULL
, name
);
417 static inline void sysfs_notify_dirent_safe(struct sysfs_dirent
*sd
)
420 sysfs_notify_dirent(sd
);
423 static inline char * mdname (mddev_t
* mddev
)
425 return mddev
->gendisk
? mddev
->gendisk
->disk_name
: "mdX";
429 * iterates through some rdev ringlist. It's safe to remove the
430 * current 'rdev'. Dont touch 'tmp' though.
432 #define rdev_for_each_list(rdev, tmp, head) \
433 list_for_each_entry_safe(rdev, tmp, head, same_set)
436 * iterates through the 'same array disks' ringlist
438 #define rdev_for_each(rdev, tmp, mddev) \
439 list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set)
441 #define rdev_for_each_rcu(rdev, mddev) \
442 list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set)
444 typedef struct mdk_thread_s
{
445 void (*run
) (mddev_t
*mddev
);
447 wait_queue_head_t wqueue
;
449 struct task_struct
*tsk
;
450 unsigned long timeout
;
453 #define THREAD_WAKEUP 0
455 #define __wait_event_lock_irq(wq, condition, lock, cmd) \
457 wait_queue_t __wait; \
458 init_waitqueue_entry(&__wait, current); \
460 add_wait_queue(&wq, &__wait); \
462 set_current_state(TASK_UNINTERRUPTIBLE); \
465 spin_unlock_irq(&lock); \
468 spin_lock_irq(&lock); \
470 current->state = TASK_RUNNING; \
471 remove_wait_queue(&wq, &__wait); \
474 #define wait_event_lock_irq(wq, condition, lock, cmd) \
478 __wait_event_lock_irq(wq, condition, lock, cmd); \
481 static inline void safe_put_page(struct page
*p
)
486 extern int register_md_personality(struct mdk_personality
*p
);
487 extern int unregister_md_personality(struct mdk_personality
*p
);
488 extern mdk_thread_t
* md_register_thread(void (*run
) (mddev_t
*mddev
),
489 mddev_t
*mddev
, const char *name
);
490 extern void md_unregister_thread(mdk_thread_t
*thread
);
491 extern void md_wakeup_thread(mdk_thread_t
*thread
);
492 extern void md_check_recovery(mddev_t
*mddev
);
493 extern void md_write_start(mddev_t
*mddev
, struct bio
*bi
);
494 extern void md_write_end(mddev_t
*mddev
);
495 extern void md_done_sync(mddev_t
*mddev
, int blocks
, int ok
);
496 extern void md_error(mddev_t
*mddev
, mdk_rdev_t
*rdev
);
498 extern int mddev_congested(mddev_t
*mddev
, int bits
);
499 extern void md_flush_request(mddev_t
*mddev
, struct bio
*bio
);
500 extern void md_super_write(mddev_t
*mddev
, mdk_rdev_t
*rdev
,
501 sector_t sector
, int size
, struct page
*page
);
502 extern void md_super_wait(mddev_t
*mddev
);
503 extern int sync_page_io(mdk_rdev_t
*rdev
, sector_t sector
, int size
,
504 struct page
*page
, int rw
);
505 extern void md_do_sync(mddev_t
*mddev
);
506 extern void md_new_event(mddev_t
*mddev
);
507 extern int md_allow_write(mddev_t
*mddev
);
508 extern void md_wait_for_blocked_rdev(mdk_rdev_t
*rdev
, mddev_t
*mddev
);
509 extern void md_set_array_sectors(mddev_t
*mddev
, sector_t array_sectors
);
510 extern int md_check_no_bitmap(mddev_t
*mddev
);
511 extern int md_integrity_register(mddev_t
*mddev
);
512 extern void md_integrity_add_rdev(mdk_rdev_t
*rdev
, mddev_t
*mddev
);
513 extern int strict_strtoul_scaled(const char *cp
, unsigned long *res
, int scale
);
514 extern void restore_bitmap_write_access(struct file
*file
);
515 extern void md_unplug(mddev_t
*mddev
);
517 extern void mddev_init(mddev_t
*mddev
);
518 extern int md_run(mddev_t
*mddev
);
519 extern void md_stop(mddev_t
*mddev
);
520 extern void md_stop_writes(mddev_t
*mddev
);
521 extern void md_rdev_init(mdk_rdev_t
*rdev
);
523 extern void mddev_suspend(mddev_t
*mddev
);
524 extern void mddev_resume(mddev_t
*mddev
);
525 extern struct bio
*bio_clone_mddev(struct bio
*bio
, gfp_t gfp_mask
,
527 extern struct bio
*bio_alloc_mddev(gfp_t gfp_mask
, int nr_iovecs
,
529 #endif /* _MD_MD_H */