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.
20 #define MaxSector (~(sector_t)0)
22 typedef struct mddev_s mddev_t;
23 typedef struct mdk_rdev_s mdk_rdev_t;
26 * options passed in raidrun:
29 /* Currently this must fit in an 'int' */
30 #define MAX_CHUNK_SIZE (1<<30)
33 * MD's 'extended' device
37 struct list_head same_set; /* RAID devices within the same set */
39 sector_t sectors; /* Device size (in 512bytes sectors) */
40 mddev_t *mddev; /* RAID array if running */
41 int last_events; /* IO event timestamp */
43 struct block_device *bdev; /* block device handle */
48 sector_t data_offset; /* start of data in array */
49 sector_t sb_start; /* offset of the super block (in 512byte sectors) */
50 int sb_size; /* bytes in the superblock */
51 int preferred_minor; /* autorun support */
55 /* A device can be in one of three states based on two flags:
56 * Not working: faulty==1 in_sync==0
57 * Fully working: faulty==0 in_sync==1
60 * faulty==0 in_sync==0
62 * It can never have faulty==1, in_sync==1
63 * This reduces the burden of testing multiple flags in many cases
67 #define Faulty 1 /* device is known to have a fault */
68 #define In_sync 2 /* device is in_sync with rest of array */
69 #define WriteMostly 4 /* Avoid reading if at all possible */
70 #define BarriersNotsupp 5 /* BIO_RW_BARRIER is not supported */
71 #define AllReserved 6 /* If whole device is reserved for
73 #define AutoDetected 7 /* added by auto-detect */
74 #define Blocked 8 /* An error occured on an externally
75 * managed array, don't allow writes
76 * until it is cleared */
77 #define StateChanged 9 /* Faulty or Blocked has changed during
78 * interrupt, so it needs to be
79 * notified by the thread */
80 wait_queue_head_t blocked_wait;
82 int desc_nr; /* descriptor index in the superblock */
83 int raid_disk; /* role of device in array */
84 int saved_raid_disk; /* role that device used to have in the
85 * array and could again if we did a partial
86 * resync from the bitmap
88 sector_t recovery_offset;/* If this device has been partially
89 * recovered, this is where we were
93 atomic_t nr_pending; /* number of pending requests.
94 * only maintained for arrays that
97 atomic_t read_errors; /* number of consecutive read errors that
98 * we have tried to ignore.
100 atomic_t corrected_errors; /* number of corrected read errors,
101 * for reporting to userspace and storing
104 struct work_struct del_work; /* used for delayed sysfs removal */
106 struct sysfs_dirent *sysfs_state; /* handle for 'state'
113 struct mdk_personality *pers;
116 struct list_head disks;
118 #define MD_CHANGE_DEVS 0 /* Some device status has changed */
119 #define MD_CHANGE_CLEAN 1 /* transition to or from 'clean' */
120 #define MD_CHANGE_PENDING 2 /* superblock update in progress */
126 struct gendisk *gendisk;
130 #define UNTIL_IOCTL 1
133 /* Superblock information */
138 int external; /* metadata is
139 * managed externally */
140 char metadata_type[17]; /* externally set*/
147 sector_t dev_sectors; /* used size of
148 * component devices */
149 sector_t array_sectors; /* exported array size */
154 /* If the array is being reshaped, we need to record the
155 * new shape and an indication of where we are up to.
156 * This is written to the superblock.
157 * If reshape_position is MaxSector, then no reshape is happening (yet).
159 sector_t reshape_position;
160 int delta_disks, new_level, new_layout, new_chunk;
162 struct mdk_thread_s *thread; /* management thread */
163 struct mdk_thread_s *sync_thread; /* doing resync or reconstruct */
164 sector_t curr_resync; /* last block scheduled */
165 /* As resync requests can complete out of order, we cannot easily track
166 * how much resync has been completed. So we occasionally pause until
167 * everything completes, then set curr_resync_completed to curr_resync.
168 * As such it may be well behind the real resync mark, but it is a value
171 sector_t curr_resync_completed;
172 unsigned long resync_mark; /* a recent timestamp */
173 sector_t resync_mark_cnt;/* blocks written at resync_mark */
174 sector_t curr_mark_cnt; /* blocks scheduled now */
176 sector_t resync_max_sectors; /* may be set by personality */
178 sector_t resync_mismatches; /* count of sectors where
179 * parity/replica mismatch found
182 /* allow user-space to request suspension of IO to regions of the array */
185 /* if zero, use the system-wide default */
189 /* resync even though the same disks are shared among md-devices */
192 int ok_start_degraded;
193 /* recovery/resync flags
194 * NEEDED: we might need to start a resync/recover
195 * RUNNING: a thread is running, or about to be started
196 * SYNC: actually doing a resync, not a recovery
197 * RECOVER: doing recovery, or need to try it.
198 * INTR: resync needs to be aborted for some reason
199 * DONE: thread is done and is waiting to be reaped
200 * REQUEST: user-space has requested a sync (used with SYNC)
201 * CHECK: user-space request for for check-only, no repair
202 * RESHAPE: A reshape is happening
204 * If neither SYNC or RESHAPE are set, then it is a recovery.
206 #define MD_RECOVERY_RUNNING 0
207 #define MD_RECOVERY_SYNC 1
208 #define MD_RECOVERY_RECOVER 2
209 #define MD_RECOVERY_INTR 3
210 #define MD_RECOVERY_DONE 4
211 #define MD_RECOVERY_NEEDED 5
212 #define MD_RECOVERY_REQUESTED 6
213 #define MD_RECOVERY_CHECK 7
214 #define MD_RECOVERY_RESHAPE 8
215 #define MD_RECOVERY_FROZEN 9
217 unsigned long recovery;
218 int recovery_disabled; /* if we detect that recovery
219 * will always fail, set this
220 * so we don't loop trying */
222 int in_sync; /* know to not need resync */
223 struct mutex reconfig_mutex;
224 atomic_t active; /* general refcount */
225 atomic_t openers; /* number of active opens */
227 int changed; /* true if we might need to reread partition info */
228 int degraded; /* whether md should consider
231 int barriers_work; /* initialised to true, cleared as soon
232 * as a barrier request to slave
233 * fails. Only supported
235 struct bio *biolist; /* bios that need to be retried
236 * because BIO_RW_BARRIER is not supported
239 atomic_t recovery_active; /* blocks scheduled, but not written */
240 wait_queue_head_t recovery_wait;
241 sector_t recovery_cp;
242 sector_t resync_min; /* user requested sync
244 sector_t resync_max; /* resync should pause
245 * when it gets here */
247 struct sysfs_dirent *sysfs_state; /* handle for 'array_state'
250 struct sysfs_dirent *sysfs_action; /* handle for 'sync_action' */
252 struct work_struct del_work; /* used for delayed sysfs removal */
254 spinlock_t write_lock;
255 wait_queue_head_t sb_wait; /* for waiting on superblock updates */
256 atomic_t pending_writes; /* number of active superblock writes */
258 unsigned int safemode; /* if set, update "clean" superblock
259 * when no writes pending.
261 unsigned int safemode_delay;
262 struct timer_list safemode_timer;
263 atomic_t writes_pending;
264 struct request_queue *queue; /* for plugging ... */
266 atomic_t write_behind; /* outstanding async IO */
267 unsigned int max_write_behind; /* 0 = sync */
269 struct bitmap *bitmap; /* the bitmap for the device */
270 struct file *bitmap_file; /* the bitmap file */
271 long bitmap_offset; /* offset from superblock of
272 * start of bitmap. May be
273 * negative, but not '0'
275 long default_bitmap_offset; /* this is the offset to use when
276 * hot-adding a bitmap. It should
277 * eventually be settable by sysfs.
280 struct list_head all_mddevs;
284 static inline void rdev_dec_pending(mdk_rdev_t *rdev, mddev_t *mddev)
286 int faulty = test_bit(Faulty, &rdev->flags);
287 if (atomic_dec_and_test(&rdev->nr_pending) && faulty)
288 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
291 static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors)
293 atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io);
296 struct mdk_personality
300 struct list_head list;
301 struct module *owner;
302 int (*make_request)(struct request_queue *q, struct bio *bio);
303 int (*run)(mddev_t *mddev);
304 int (*stop)(mddev_t *mddev);
305 void (*status)(struct seq_file *seq, mddev_t *mddev);
306 /* error_handler must set ->faulty and clear ->in_sync
307 * if appropriate, and should abort recovery if needed
309 void (*error_handler)(mddev_t *mddev, mdk_rdev_t *rdev);
310 int (*hot_add_disk) (mddev_t *mddev, mdk_rdev_t *rdev);
311 int (*hot_remove_disk) (mddev_t *mddev, int number);
312 int (*spare_active) (mddev_t *mddev);
313 sector_t (*sync_request)(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster);
314 int (*resize) (mddev_t *mddev, sector_t sectors);
315 int (*check_reshape) (mddev_t *mddev);
316 int (*start_reshape) (mddev_t *mddev);
317 int (*reconfig) (mddev_t *mddev, int layout, int chunk_size);
318 /* quiesce moves between quiescence states
320 * 1 - no new requests allowed
323 void (*quiesce) (mddev_t *mddev, int state);
324 /* takeover is used to transition an array from one
325 * personality to another. The new personality must be able
326 * to handle the data in the current layout.
327 * e.g. 2drive raid1 -> 2drive raid5
328 * ndrive raid5 -> degraded n+1drive raid6 with special layout
329 * If the takeover succeeds, a new 'private' structure is returned.
330 * This needs to be installed and then ->run used to activate the
333 void *(*takeover) (mddev_t *mddev);
337 struct md_sysfs_entry {
338 struct attribute attr;
339 ssize_t (*show)(mddev_t *, char *);
340 ssize_t (*store)(mddev_t *, const char *, size_t);
344 static inline char * mdname (mddev_t * mddev)
346 return mddev->gendisk ? mddev->gendisk->disk_name : "mdX";
350 * iterates through some rdev ringlist. It's safe to remove the
351 * current 'rdev'. Dont touch 'tmp' though.
353 #define rdev_for_each_list(rdev, tmp, head) \
354 list_for_each_entry_safe(rdev, tmp, head, same_set)
357 * iterates through the 'same array disks' ringlist
359 #define rdev_for_each(rdev, tmp, mddev) \
360 list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set)
362 #define rdev_for_each_rcu(rdev, mddev) \
363 list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set)
365 typedef struct mdk_thread_s {
366 void (*run) (mddev_t *mddev);
368 wait_queue_head_t wqueue;
370 struct task_struct *tsk;
371 unsigned long timeout;
374 #define THREAD_WAKEUP 0
376 #define __wait_event_lock_irq(wq, condition, lock, cmd) \
378 wait_queue_t __wait; \
379 init_waitqueue_entry(&__wait, current); \
381 add_wait_queue(&wq, &__wait); \
383 set_current_state(TASK_UNINTERRUPTIBLE); \
386 spin_unlock_irq(&lock); \
389 spin_lock_irq(&lock); \
391 current->state = TASK_RUNNING; \
392 remove_wait_queue(&wq, &__wait); \
395 #define wait_event_lock_irq(wq, condition, lock, cmd) \
399 __wait_event_lock_irq(wq, condition, lock, cmd); \
402 static inline void safe_put_page(struct page *p)
407 #endif /* CONFIG_BLOCK */
411 extern int register_md_personality(struct mdk_personality *p);
412 extern int unregister_md_personality(struct mdk_personality *p);
413 extern mdk_thread_t * md_register_thread(void (*run) (mddev_t *mddev),
414 mddev_t *mddev, const char *name);
415 extern void md_unregister_thread(mdk_thread_t *thread);
416 extern void md_wakeup_thread(mdk_thread_t *thread);
417 extern void md_check_recovery(mddev_t *mddev);
418 extern void md_write_start(mddev_t *mddev, struct bio *bi);
419 extern void md_write_end(mddev_t *mddev);
420 extern void md_done_sync(mddev_t *mddev, int blocks, int ok);
421 extern void md_error(mddev_t *mddev, mdk_rdev_t *rdev);
423 extern void md_super_write(mddev_t *mddev, mdk_rdev_t *rdev,
424 sector_t sector, int size, struct page *page);
425 extern void md_super_wait(mddev_t *mddev);
426 extern int sync_page_io(struct block_device *bdev, sector_t sector, int size,
427 struct page *page, int rw);
428 extern void md_do_sync(mddev_t *mddev);
429 extern void md_new_event(mddev_t *mddev);
430 extern int md_allow_write(mddev_t *mddev);
431 extern void md_wait_for_blocked_rdev(mdk_rdev_t *rdev, mddev_t *mddev);