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;
33 * options passed in raidrun:
36 /* Currently this must fit in an 'int' */
37 #define MAX_CHUNK_SIZE (1<<30)
40 * MD's 'extended' device
44 struct list_head same_set; /* RAID devices within the same set */
46 sector_t sectors; /* Device size (in 512bytes sectors) */
47 mddev_t *mddev; /* RAID array if running */
48 int last_events; /* IO event timestamp */
50 struct block_device *bdev; /* block device handle */
55 sector_t data_offset; /* start of data in array */
56 sector_t sb_start; /* offset of the super block (in 512byte sectors) */
57 int sb_size; /* bytes in the superblock */
58 int preferred_minor; /* autorun support */
62 /* A device can be in one of three states based on two flags:
63 * Not working: faulty==1 in_sync==0
64 * Fully working: faulty==0 in_sync==1
67 * faulty==0 in_sync==0
69 * It can never have faulty==1, in_sync==1
70 * This reduces the burden of testing multiple flags in many cases
74 #define Faulty 1 /* device is known to have a fault */
75 #define In_sync 2 /* device is in_sync with rest of array */
76 #define WriteMostly 4 /* Avoid reading if at all possible */
77 #define BarriersNotsupp 5 /* BIO_RW_BARRIER is not supported */
78 #define AllReserved 6 /* If whole device is reserved for
80 #define AutoDetected 7 /* added by auto-detect */
81 #define Blocked 8 /* An error occured on an externally
82 * managed array, don't allow writes
83 * until it is cleared */
84 #define StateChanged 9 /* Faulty or Blocked has changed during
85 * interrupt, so it needs to be
86 * notified by the thread */
87 wait_queue_head_t blocked_wait;
89 int desc_nr; /* descriptor index in the superblock */
90 int raid_disk; /* role of device in array */
91 int saved_raid_disk; /* role that device used to have in the
92 * array and could again if we did a partial
93 * resync from the bitmap
95 sector_t recovery_offset;/* If this device has been partially
96 * recovered, this is where we were
100 atomic_t nr_pending; /* number of pending requests.
101 * only maintained for arrays that
102 * support hot removal
104 atomic_t read_errors; /* number of consecutive read errors that
105 * we have tried to ignore.
107 atomic_t corrected_errors; /* number of corrected read errors,
108 * for reporting to userspace and storing
111 struct work_struct del_work; /* used for delayed sysfs removal */
113 struct sysfs_dirent *sysfs_state; /* handle for 'state'
120 struct mdk_personality *pers;
123 struct list_head disks;
125 #define MD_CHANGE_DEVS 0 /* Some device status has changed */
126 #define MD_CHANGE_CLEAN 1 /* transition to or from 'clean' */
127 #define MD_CHANGE_PENDING 2 /* superblock update in progress */
133 struct gendisk *gendisk;
137 #define UNTIL_IOCTL 1
140 /* Superblock information */
145 int external; /* metadata is
146 * managed externally */
147 char metadata_type[17]; /* externally set*/
154 sector_t dev_sectors; /* used size of
155 * component devices */
156 sector_t array_sectors; /* exported array size */
157 int external_size; /* size managed
163 /* If the array is being reshaped, we need to record the
164 * new shape and an indication of where we are up to.
165 * This is written to the superblock.
166 * If reshape_position is MaxSector, then no reshape is happening (yet).
168 sector_t reshape_position;
169 int delta_disks, new_level, new_layout;
170 int new_chunk_sectors;
172 struct mdk_thread_s *thread; /* management thread */
173 struct mdk_thread_s *sync_thread; /* doing resync or reconstruct */
174 sector_t curr_resync; /* last block scheduled */
175 /* As resync requests can complete out of order, we cannot easily track
176 * how much resync has been completed. So we occasionally pause until
177 * everything completes, then set curr_resync_completed to curr_resync.
178 * As such it may be well behind the real resync mark, but it is a value
181 sector_t curr_resync_completed;
182 unsigned long resync_mark; /* a recent timestamp */
183 sector_t resync_mark_cnt;/* blocks written at resync_mark */
184 sector_t curr_mark_cnt; /* blocks scheduled now */
186 sector_t resync_max_sectors; /* may be set by personality */
188 sector_t resync_mismatches; /* count of sectors where
189 * parity/replica mismatch found
192 /* allow user-space to request suspension of IO to regions of the array */
195 /* if zero, use the system-wide default */
199 /* resync even though the same disks are shared among md-devices */
202 int ok_start_degraded;
203 /* recovery/resync flags
204 * NEEDED: we might need to start a resync/recover
205 * RUNNING: a thread is running, or about to be started
206 * SYNC: actually doing a resync, not a recovery
207 * RECOVER: doing recovery, or need to try it.
208 * INTR: resync needs to be aborted for some reason
209 * DONE: thread is done and is waiting to be reaped
210 * REQUEST: user-space has requested a sync (used with SYNC)
211 * CHECK: user-space request for for check-only, no repair
212 * RESHAPE: A reshape is happening
214 * If neither SYNC or RESHAPE are set, then it is a recovery.
216 #define MD_RECOVERY_RUNNING 0
217 #define MD_RECOVERY_SYNC 1
218 #define MD_RECOVERY_RECOVER 2
219 #define MD_RECOVERY_INTR 3
220 #define MD_RECOVERY_DONE 4
221 #define MD_RECOVERY_NEEDED 5
222 #define MD_RECOVERY_REQUESTED 6
223 #define MD_RECOVERY_CHECK 7
224 #define MD_RECOVERY_RESHAPE 8
225 #define MD_RECOVERY_FROZEN 9
227 unsigned long recovery;
228 int recovery_disabled; /* if we detect that recovery
229 * will always fail, set this
230 * so we don't loop trying */
232 int in_sync; /* know to not need resync */
233 struct mutex reconfig_mutex;
234 atomic_t active; /* general refcount */
235 atomic_t openers; /* number of active opens */
237 int changed; /* true if we might need to reread partition info */
238 int degraded; /* whether md should consider
241 int barriers_work; /* initialised to true, cleared as soon
242 * as a barrier request to slave
243 * fails. Only supported
245 struct bio *biolist; /* bios that need to be retried
246 * because BIO_RW_BARRIER is not supported
249 atomic_t recovery_active; /* blocks scheduled, but not written */
250 wait_queue_head_t recovery_wait;
251 sector_t recovery_cp;
252 sector_t resync_min; /* user requested sync
254 sector_t resync_max; /* resync should pause
255 * when it gets here */
257 struct sysfs_dirent *sysfs_state; /* handle for 'array_state'
260 struct sysfs_dirent *sysfs_action; /* handle for 'sync_action' */
262 struct work_struct del_work; /* used for delayed sysfs removal */
264 spinlock_t write_lock;
265 wait_queue_head_t sb_wait; /* for waiting on superblock updates */
266 atomic_t pending_writes; /* number of active superblock writes */
268 unsigned int safemode; /* if set, update "clean" superblock
269 * when no writes pending.
271 unsigned int safemode_delay;
272 struct timer_list safemode_timer;
273 atomic_t writes_pending;
274 struct request_queue *queue; /* for plugging ... */
276 atomic_t write_behind; /* outstanding async IO */
277 unsigned int max_write_behind; /* 0 = sync */
279 struct bitmap *bitmap; /* the bitmap for the device */
280 struct file *bitmap_file; /* the bitmap file */
281 long bitmap_offset; /* offset from superblock of
282 * start of bitmap. May be
283 * negative, but not '0'
285 long default_bitmap_offset; /* this is the offset to use when
286 * hot-adding a bitmap. It should
287 * eventually be settable by sysfs.
290 struct list_head all_mddevs;
294 static inline void rdev_dec_pending(mdk_rdev_t *rdev, mddev_t *mddev)
296 int faulty = test_bit(Faulty, &rdev->flags);
297 if (atomic_dec_and_test(&rdev->nr_pending) && faulty)
298 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
301 static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors)
303 atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io);
306 struct mdk_personality
310 struct list_head list;
311 struct module *owner;
312 int (*make_request)(struct request_queue *q, struct bio *bio);
313 int (*run)(mddev_t *mddev);
314 int (*stop)(mddev_t *mddev);
315 void (*status)(struct seq_file *seq, mddev_t *mddev);
316 /* error_handler must set ->faulty and clear ->in_sync
317 * if appropriate, and should abort recovery if needed
319 void (*error_handler)(mddev_t *mddev, mdk_rdev_t *rdev);
320 int (*hot_add_disk) (mddev_t *mddev, mdk_rdev_t *rdev);
321 int (*hot_remove_disk) (mddev_t *mddev, int number);
322 int (*spare_active) (mddev_t *mddev);
323 sector_t (*sync_request)(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster);
324 int (*resize) (mddev_t *mddev, sector_t sectors);
325 sector_t (*size) (mddev_t *mddev, sector_t sectors, int raid_disks);
326 int (*check_reshape) (mddev_t *mddev);
327 int (*start_reshape) (mddev_t *mddev);
328 void (*finish_reshape) (mddev_t *mddev);
329 int (*reconfig) (mddev_t *mddev);
330 /* quiesce moves between quiescence states
332 * 1 - no new requests allowed
335 void (*quiesce) (mddev_t *mddev, int state);
336 /* takeover is used to transition an array from one
337 * personality to another. The new personality must be able
338 * to handle the data in the current layout.
339 * e.g. 2drive raid1 -> 2drive raid5
340 * ndrive raid5 -> degraded n+1drive raid6 with special layout
341 * If the takeover succeeds, a new 'private' structure is returned.
342 * This needs to be installed and then ->run used to activate the
345 void *(*takeover) (mddev_t *mddev);
349 struct md_sysfs_entry {
350 struct attribute attr;
351 ssize_t (*show)(mddev_t *, char *);
352 ssize_t (*store)(mddev_t *, const char *, size_t);
356 static inline char * mdname (mddev_t * mddev)
358 return mddev->gendisk ? mddev->gendisk->disk_name : "mdX";
362 * iterates through some rdev ringlist. It's safe to remove the
363 * current 'rdev'. Dont touch 'tmp' though.
365 #define rdev_for_each_list(rdev, tmp, head) \
366 list_for_each_entry_safe(rdev, tmp, head, same_set)
369 * iterates through the 'same array disks' ringlist
371 #define rdev_for_each(rdev, tmp, mddev) \
372 list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set)
374 #define rdev_for_each_rcu(rdev, mddev) \
375 list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set)
377 typedef struct mdk_thread_s {
378 void (*run) (mddev_t *mddev);
380 wait_queue_head_t wqueue;
382 struct task_struct *tsk;
383 unsigned long timeout;
386 #define THREAD_WAKEUP 0
388 #define __wait_event_lock_irq(wq, condition, lock, cmd) \
390 wait_queue_t __wait; \
391 init_waitqueue_entry(&__wait, current); \
393 add_wait_queue(&wq, &__wait); \
395 set_current_state(TASK_UNINTERRUPTIBLE); \
398 spin_unlock_irq(&lock); \
401 spin_lock_irq(&lock); \
403 current->state = TASK_RUNNING; \
404 remove_wait_queue(&wq, &__wait); \
407 #define wait_event_lock_irq(wq, condition, lock, cmd) \
411 __wait_event_lock_irq(wq, condition, lock, cmd); \
414 static inline void safe_put_page(struct page *p)
419 extern int register_md_personality(struct mdk_personality *p);
420 extern int unregister_md_personality(struct mdk_personality *p);
421 extern mdk_thread_t * md_register_thread(void (*run) (mddev_t *mddev),
422 mddev_t *mddev, const char *name);
423 extern void md_unregister_thread(mdk_thread_t *thread);
424 extern void md_wakeup_thread(mdk_thread_t *thread);
425 extern void md_check_recovery(mddev_t *mddev);
426 extern void md_write_start(mddev_t *mddev, struct bio *bi);
427 extern void md_write_end(mddev_t *mddev);
428 extern void md_done_sync(mddev_t *mddev, int blocks, int ok);
429 extern void md_error(mddev_t *mddev, mdk_rdev_t *rdev);
431 extern void md_super_write(mddev_t *mddev, mdk_rdev_t *rdev,
432 sector_t sector, int size, struct page *page);
433 extern void md_super_wait(mddev_t *mddev);
434 extern int sync_page_io(struct block_device *bdev, sector_t sector, int size,
435 struct page *page, int rw);
436 extern void md_do_sync(mddev_t *mddev);
437 extern void md_new_event(mddev_t *mddev);
438 extern int md_allow_write(mddev_t *mddev);
439 extern void md_wait_for_blocked_rdev(mdk_rdev_t *rdev, mddev_t *mddev);
440 extern void md_set_array_sectors(mddev_t *mddev, sector_t array_sectors);
442 #endif /* _MD_MD_H */