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)
35 typedef struct mddev_s mddev_t;
36 typedef struct mdk_rdev_s mdk_rdev_t;
39 * options passed in raidrun:
42 /* Currently this must fit in an 'int' */
43 #define MAX_CHUNK_SIZE (1<<30)
46 * MD's 'extended' device
50 struct list_head same_set; /* RAID devices within the same set */
52 sector_t size; /* Device size (in blocks) */
53 mddev_t *mddev; /* RAID array if running */
54 long last_events; /* IO event timestamp */
56 struct block_device *bdev; /* block device handle */
61 sector_t data_offset; /* start of data in array */
63 int sb_size; /* bytes in the superblock */
64 int preferred_minor; /* autorun support */
68 /* A device can be in one of three states based on two flags:
69 * Not working: faulty==1 in_sync==0
70 * Fully working: faulty==0 in_sync==1
73 * faulty==0 in_sync==0
75 * It can never have faulty==1, in_sync==1
76 * This reduces the burden of testing multiple flags in many cases
80 #define Faulty 1 /* device is known to have a fault */
81 #define In_sync 2 /* device is in_sync with rest of array */
82 #define WriteMostly 4 /* Avoid reading if at all possible */
83 #define BarriersNotsupp 5 /* BIO_RW_BARRIER is not supported */
85 int desc_nr; /* descriptor index in the superblock */
86 int raid_disk; /* role of device in array */
87 int saved_raid_disk; /* role that device used to have in the
88 * array and could again if we did a partial
89 * resync from the bitmap
91 sector_t recovery_offset;/* If this device has been partially
92 * recovered, this is where we were
96 atomic_t nr_pending; /* number of pending requests.
97 * only maintained for arrays that
100 atomic_t read_errors; /* number of consecutive read errors that
101 * we have tried to ignore.
103 atomic_t corrected_errors; /* number of corrected read errors,
104 * for reporting to userspace and storing
107 struct work_struct del_work; /* used for delayed sysfs removal */
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 */
124 struct gendisk *gendisk;
128 /* Superblock information */
133 int external; /* metadata is
134 * managed externally */
135 char metadata_type[17]; /* externally set*/
142 sector_t size; /* used size of component devices */
143 sector_t array_size; /* exported array size */
148 /* If the array is being reshaped, we need to record the
149 * new shape and an indication of where we are up to.
150 * This is written to the superblock.
151 * If reshape_position is MaxSector, then no reshape is happening (yet).
153 sector_t reshape_position;
154 int delta_disks, new_level, new_layout, new_chunk;
156 struct mdk_thread_s *thread; /* management thread */
157 struct mdk_thread_s *sync_thread; /* doing resync or reconstruct */
158 sector_t curr_resync; /* last block scheduled */
159 unsigned long resync_mark; /* a recent timestamp */
160 sector_t resync_mark_cnt;/* blocks written at resync_mark */
161 sector_t curr_mark_cnt; /* blocks scheduled now */
163 sector_t resync_max_sectors; /* may be set by personality */
165 sector_t resync_mismatches; /* count of sectors where
166 * parity/replica mismatch found
169 /* allow user-space to request suspension of IO to regions of the array */
172 /* if zero, use the system-wide default */
176 int ok_start_degraded;
177 /* recovery/resync flags
178 * NEEDED: we might need to start a resync/recover
179 * RUNNING: a thread is running, or about to be started
180 * SYNC: actually doing a resync, not a recovery
181 * ERR: and IO error was detected - abort the resync/recovery
182 * INTR: someone requested a (clean) early abort.
183 * DONE: thread is done and is waiting to be reaped
184 * REQUEST: user-space has requested a sync (used with SYNC)
185 * CHECK: user-space request for for check-only, no repair
186 * RESHAPE: A reshape is happening
188 * If neither SYNC or RESHAPE are set, then it is a recovery.
190 #define MD_RECOVERY_RUNNING 0
191 #define MD_RECOVERY_SYNC 1
192 #define MD_RECOVERY_ERR 2
193 #define MD_RECOVERY_INTR 3
194 #define MD_RECOVERY_DONE 4
195 #define MD_RECOVERY_NEEDED 5
196 #define MD_RECOVERY_REQUESTED 6
197 #define MD_RECOVERY_CHECK 7
198 #define MD_RECOVERY_RESHAPE 8
199 #define MD_RECOVERY_FROZEN 9
201 unsigned long recovery;
203 int in_sync; /* know to not need resync */
204 struct mutex reconfig_mutex;
207 int changed; /* true if we might need to reread partition info */
208 int degraded; /* whether md should consider
211 int barriers_work; /* initialised to true, cleared as soon
212 * as a barrier request to slave
213 * fails. Only supported
215 struct bio *biolist; /* bios that need to be retried
216 * because BIO_RW_BARRIER is not supported
219 atomic_t recovery_active; /* blocks scheduled, but not written */
220 wait_queue_head_t recovery_wait;
221 sector_t recovery_cp;
223 spinlock_t write_lock;
224 wait_queue_head_t sb_wait; /* for waiting on superblock updates */
225 atomic_t pending_writes; /* number of active superblock writes */
227 unsigned int safemode; /* if set, update "clean" superblock
228 * when no writes pending.
230 unsigned int safemode_delay;
231 struct timer_list safemode_timer;
232 atomic_t writes_pending;
233 struct request_queue *queue; /* for plugging ... */
235 atomic_t write_behind; /* outstanding async IO */
236 unsigned int max_write_behind; /* 0 = sync */
238 struct bitmap *bitmap; /* the bitmap for the device */
239 struct file *bitmap_file; /* the bitmap file */
240 long bitmap_offset; /* offset from superblock of
241 * start of bitmap. May be
242 * negative, but not '0'
244 long default_bitmap_offset; /* this is the offset to use when
245 * hot-adding a bitmap. It should
246 * eventually be settable by sysfs.
249 struct list_head all_mddevs;
253 static inline void rdev_dec_pending(mdk_rdev_t *rdev, mddev_t *mddev)
255 int faulty = test_bit(Faulty, &rdev->flags);
256 if (atomic_dec_and_test(&rdev->nr_pending) && faulty)
257 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
260 static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors)
262 atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io);
265 struct mdk_personality
269 struct list_head list;
270 struct module *owner;
271 int (*make_request)(struct request_queue *q, struct bio *bio);
272 int (*run)(mddev_t *mddev);
273 int (*stop)(mddev_t *mddev);
274 void (*status)(struct seq_file *seq, mddev_t *mddev);
275 /* error_handler must set ->faulty and clear ->in_sync
276 * if appropriate, and should abort recovery if needed
278 void (*error_handler)(mddev_t *mddev, mdk_rdev_t *rdev);
279 int (*hot_add_disk) (mddev_t *mddev, mdk_rdev_t *rdev);
280 int (*hot_remove_disk) (mddev_t *mddev, int number);
281 int (*spare_active) (mddev_t *mddev);
282 sector_t (*sync_request)(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster);
283 int (*resize) (mddev_t *mddev, sector_t sectors);
284 int (*check_reshape) (mddev_t *mddev);
285 int (*start_reshape) (mddev_t *mddev);
286 int (*reconfig) (mddev_t *mddev, int layout, int chunk_size);
287 /* quiesce moves between quiescence states
289 * 1 - no new requests allowed
292 void (*quiesce) (mddev_t *mddev, int state);
296 struct md_sysfs_entry {
297 struct attribute attr;
298 ssize_t (*show)(mddev_t *, char *);
299 ssize_t (*store)(mddev_t *, const char *, size_t);
303 static inline char * mdname (mddev_t * mddev)
305 return mddev->gendisk ? mddev->gendisk->disk_name : "mdX";
309 * iterates through some rdev ringlist. It's safe to remove the
310 * current 'rdev'. Dont touch 'tmp' though.
312 #define ITERATE_RDEV_GENERIC(head,rdev,tmp) \
314 for ((tmp) = (head).next; \
315 (rdev) = (list_entry((tmp), mdk_rdev_t, same_set)), \
316 (tmp) = (tmp)->next, (tmp)->prev != &(head) \
319 * iterates through the 'same array disks' ringlist
321 #define ITERATE_RDEV(mddev,rdev,tmp) \
322 ITERATE_RDEV_GENERIC((mddev)->disks,rdev,tmp)
325 * Iterates through 'pending RAID disks'
327 #define ITERATE_RDEV_PENDING(rdev,tmp) \
328 ITERATE_RDEV_GENERIC(pending_raid_disks,rdev,tmp)
330 typedef struct mdk_thread_s {
331 void (*run) (mddev_t *mddev);
333 wait_queue_head_t wqueue;
335 struct task_struct *tsk;
336 unsigned long timeout;
339 #define THREAD_WAKEUP 0
341 #define __wait_event_lock_irq(wq, condition, lock, cmd) \
343 wait_queue_t __wait; \
344 init_waitqueue_entry(&__wait, current); \
346 add_wait_queue(&wq, &__wait); \
348 set_current_state(TASK_UNINTERRUPTIBLE); \
351 spin_unlock_irq(&lock); \
354 spin_lock_irq(&lock); \
356 current->state = TASK_RUNNING; \
357 remove_wait_queue(&wq, &__wait); \
360 #define wait_event_lock_irq(wq, condition, lock, cmd) \
364 __wait_event_lock_irq(wq, condition, lock, cmd); \
367 static inline void safe_put_page(struct page *p)
372 #endif /* CONFIG_BLOCK */