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 /* and dm-bio-list.h is not under include/linux because.... ??? */
19 #include "../../../drivers/md/dm-bio-list.h"
23 #define LEVEL_MULTIPATH (-4)
24 #define LEVEL_LINEAR (-1)
25 #define LEVEL_FAULTY (-5)
27 /* we need a value for 'no level specified' and 0
28 * means 'raid0', so we need something else. This is
29 * for internal use only
31 #define LEVEL_NONE (-1000000)
33 #define MaxSector (~(sector_t)0)
34 #define MD_THREAD_NAME_MAX 14
36 typedef struct mddev_s mddev_t
;
37 typedef struct mdk_rdev_s mdk_rdev_t
;
39 #define MAX_MD_DEVS 256 /* Max number of md dev */
42 * options passed in raidrun:
45 /* Currently this must fix in an 'int' */
46 #define MAX_CHUNK_SIZE (1<<30)
49 * MD's 'extended' device
53 struct list_head same_set
; /* RAID devices within the same set */
55 sector_t size
; /* Device size (in blocks) */
56 mddev_t
*mddev
; /* RAID array if running */
57 unsigned long last_events
; /* IO event timestamp */
59 struct block_device
*bdev
; /* block device handle */
64 sector_t data_offset
; /* start of data in array */
66 int sb_size
; /* bytes in the superblock */
67 int preferred_minor
; /* autorun support */
71 /* A device can be in one of three states based on two flags:
72 * Not working: faulty==1 in_sync==0
73 * Fully working: faulty==0 in_sync==1
76 * faulty==0 in_sync==0
78 * It can never have faulty==1, in_sync==1
79 * This reduces the burden of testing multiple flags in many cases
83 #define Faulty 1 /* device is known to have a fault */
84 #define In_sync 2 /* device is in_sync with rest of array */
85 #define WriteMostly 4 /* Avoid reading if at all possible */
86 #define BarriersNotsupp 5 /* BIO_RW_BARRIER is not supported */
88 int desc_nr
; /* descriptor index in the superblock */
89 int raid_disk
; /* role of device in array */
90 int saved_raid_disk
; /* role that device used to have in the
91 * array and could again if we did a partial
92 * resync from the bitmap
94 sector_t recovery_offset
;/* If this device has been partially
95 * recovered, this is where we were
99 atomic_t nr_pending
; /* number of pending requests.
100 * only maintained for arrays that
101 * support hot removal
103 atomic_t read_errors
; /* number of consecutive read errors that
104 * we have tried to ignore.
106 atomic_t corrected_errors
; /* number of corrected read errors,
107 * for reporting to userspace and storing
115 struct mdk_personality
*pers
;
118 struct list_head disks
;
122 struct gendisk
*gendisk
;
126 /* Superblock information */
137 sector_t size
; /* used size of component devices */
138 sector_t array_size
; /* exported array size */
143 /* If the array is being reshaped, we need to record the
144 * new shape and an indication of where we are up to.
145 * This is written to the superblock.
146 * If reshape_position is MaxSector, then no reshape is happening (yet).
148 sector_t reshape_position
;
149 int delta_disks
, new_level
, new_layout
, new_chunk
;
151 struct mdk_thread_s
*thread
; /* management thread */
152 struct mdk_thread_s
*sync_thread
; /* doing resync or reconstruct */
153 sector_t curr_resync
; /* last block scheduled */
154 unsigned long resync_mark
; /* a recent timestamp */
155 sector_t resync_mark_cnt
;/* blocks written at resync_mark */
156 sector_t curr_mark_cnt
; /* blocks scheduled now */
158 sector_t resync_max_sectors
; /* may be set by personality */
160 sector_t resync_mismatches
; /* count of sectors where
161 * parity/replica mismatch found
164 /* allow user-space to request suspension of IO to regions of the array */
167 /* if zero, use the system-wide default */
171 int ok_start_degraded
;
172 /* recovery/resync flags
173 * NEEDED: we might need to start a resync/recover
174 * RUNNING: a thread is running, or about to be started
175 * SYNC: actually doing a resync, not a recovery
176 * ERR: and IO error was detected - abort the resync/recovery
177 * INTR: someone requested a (clean) early abort.
178 * DONE: thread is done and is waiting to be reaped
179 * REQUEST: user-space has requested a sync (used with SYNC)
180 * CHECK: user-space request for for check-only, no repair
181 * RESHAPE: A reshape is happening
183 * If neither SYNC or RESHAPE are set, then it is a recovery.
185 #define MD_RECOVERY_RUNNING 0
186 #define MD_RECOVERY_SYNC 1
187 #define MD_RECOVERY_ERR 2
188 #define MD_RECOVERY_INTR 3
189 #define MD_RECOVERY_DONE 4
190 #define MD_RECOVERY_NEEDED 5
191 #define MD_RECOVERY_REQUESTED 6
192 #define MD_RECOVERY_CHECK 7
193 #define MD_RECOVERY_RESHAPE 8
194 #define MD_RECOVERY_FROZEN 9
196 unsigned long recovery
;
198 int in_sync
; /* know to not need resync */
199 struct mutex reconfig_mutex
;
202 int changed
; /* true if we might need to reread partition info */
203 int degraded
; /* whether md should consider
206 int barriers_work
; /* initialised to true, cleared as soon
207 * as a barrier request to slave
208 * fails. Only supported
210 struct bio
*biolist
; /* bios that need to be retried
211 * because BIO_RW_BARRIER is not supported
214 atomic_t recovery_active
; /* blocks scheduled, but not written */
215 wait_queue_head_t recovery_wait
;
216 sector_t recovery_cp
;
218 spinlock_t write_lock
;
219 wait_queue_head_t sb_wait
; /* for waiting on superblock updates */
220 atomic_t pending_writes
; /* number of active superblock writes */
222 unsigned int safemode
; /* if set, update "clean" superblock
223 * when no writes pending.
225 unsigned int safemode_delay
;
226 struct timer_list safemode_timer
;
227 atomic_t writes_pending
;
228 request_queue_t
*queue
; /* for plugging ... */
230 atomic_t write_behind
; /* outstanding async IO */
231 unsigned int max_write_behind
; /* 0 = sync */
233 struct bitmap
*bitmap
; /* the bitmap for the device */
234 struct file
*bitmap_file
; /* the bitmap file */
235 long bitmap_offset
; /* offset from superblock of
236 * start of bitmap. May be
237 * negative, but not '0'
239 long default_bitmap_offset
; /* this is the offset to use when
240 * hot-adding a bitmap. It should
241 * eventually be settable by sysfs.
244 struct list_head all_mddevs
;
248 static inline void rdev_dec_pending(mdk_rdev_t
*rdev
, mddev_t
*mddev
)
250 int faulty
= test_bit(Faulty
, &rdev
->flags
);
251 if (atomic_dec_and_test(&rdev
->nr_pending
) && faulty
)
252 set_bit(MD_RECOVERY_NEEDED
, &mddev
->recovery
);
255 static inline void md_sync_acct(struct block_device
*bdev
, unsigned long nr_sectors
)
257 atomic_add(nr_sectors
, &bdev
->bd_contains
->bd_disk
->sync_io
);
260 struct mdk_personality
264 struct list_head list
;
265 struct module
*owner
;
266 int (*make_request
)(request_queue_t
*q
, struct bio
*bio
);
267 int (*run
)(mddev_t
*mddev
);
268 int (*stop
)(mddev_t
*mddev
);
269 void (*status
)(struct seq_file
*seq
, mddev_t
*mddev
);
270 /* error_handler must set ->faulty and clear ->in_sync
271 * if appropriate, and should abort recovery if needed
273 void (*error_handler
)(mddev_t
*mddev
, mdk_rdev_t
*rdev
);
274 int (*hot_add_disk
) (mddev_t
*mddev
, mdk_rdev_t
*rdev
);
275 int (*hot_remove_disk
) (mddev_t
*mddev
, int number
);
276 int (*spare_active
) (mddev_t
*mddev
);
277 sector_t (*sync_request
)(mddev_t
*mddev
, sector_t sector_nr
, int *skipped
, int go_faster
);
278 int (*resize
) (mddev_t
*mddev
, sector_t sectors
);
279 int (*check_reshape
) (mddev_t
*mddev
);
280 int (*start_reshape
) (mddev_t
*mddev
);
281 int (*reconfig
) (mddev_t
*mddev
, int layout
, int chunk_size
);
282 /* quiesce moves between quiescence states
284 * 1 - no new requests allowed
287 void (*quiesce
) (mddev_t
*mddev
, int state
);
291 struct md_sysfs_entry
{
292 struct attribute attr
;
293 ssize_t (*show
)(mddev_t
*, char *);
294 ssize_t (*store
)(mddev_t
*, const char *, size_t);
298 static inline char * mdname (mddev_t
* mddev
)
300 return mddev
->gendisk
? mddev
->gendisk
->disk_name
: "mdX";
304 * iterates through some rdev ringlist. It's safe to remove the
305 * current 'rdev'. Dont touch 'tmp' though.
307 #define ITERATE_RDEV_GENERIC(head,rdev,tmp) \
309 for ((tmp) = (head).next; \
310 (rdev) = (list_entry((tmp), mdk_rdev_t, same_set)), \
311 (tmp) = (tmp)->next, (tmp)->prev != &(head) \
314 * iterates through the 'same array disks' ringlist
316 #define ITERATE_RDEV(mddev,rdev,tmp) \
317 ITERATE_RDEV_GENERIC((mddev)->disks,rdev,tmp)
320 * Iterates through 'pending RAID disks'
322 #define ITERATE_RDEV_PENDING(rdev,tmp) \
323 ITERATE_RDEV_GENERIC(pending_raid_disks,rdev,tmp)
325 typedef struct mdk_thread_s
{
326 void (*run
) (mddev_t
*mddev
);
328 wait_queue_head_t wqueue
;
330 struct task_struct
*tsk
;
331 unsigned long timeout
;
334 #define THREAD_WAKEUP 0
336 #define __wait_event_lock_irq(wq, condition, lock, cmd) \
338 wait_queue_t __wait; \
339 init_waitqueue_entry(&__wait, current); \
341 add_wait_queue(&wq, &__wait); \
343 set_current_state(TASK_UNINTERRUPTIBLE); \
346 spin_unlock_irq(&lock); \
349 spin_lock_irq(&lock); \
351 current->state = TASK_RUNNING; \
352 remove_wait_queue(&wq, &__wait); \
355 #define wait_event_lock_irq(wq, condition, lock, cmd) \
359 __wait_event_lock_irq(wq, condition, lock, cmd); \
362 static inline void safe_put_page(struct page
*p
)
367 #endif /* CONFIG_BLOCK */