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"
21 #define LEVEL_MULTIPATH (-4)
22 #define LEVEL_LINEAR (-1)
23 #define LEVEL_FAULTY (-5)
25 /* we need a value for 'no level specified' and 0
26 * means 'raid0', so we need something else. This is
27 * for internal use only
29 #define LEVEL_NONE (-1000000)
31 #define MaxSector (~(sector_t)0)
32 #define MD_THREAD_NAME_MAX 14
34 typedef struct mddev_s mddev_t;
35 typedef struct mdk_rdev_s mdk_rdev_t;
37 #define MAX_MD_DEVS 256 /* Max number of md dev */
40 * options passed in raidrun:
43 /* Currently this must fix in an 'int' */
44 #define MAX_CHUNK_SIZE (1<<30)
47 * MD's 'extended' device
51 struct list_head same_set; /* RAID devices within the same set */
53 sector_t size; /* Device size (in blocks) */
54 mddev_t *mddev; /* RAID array if running */
55 unsigned long last_events; /* IO event timestamp */
57 struct block_device *bdev; /* block device handle */
62 sector_t data_offset; /* start of data in array */
64 int sb_size; /* bytes in the superblock */
65 int preferred_minor; /* autorun support */
69 /* A device can be in one of three states based on two flags:
70 * Not working: faulty==1 in_sync==0
71 * Fully working: faulty==0 in_sync==1
74 * faulty==0 in_sync==0
76 * It can never have faulty==1, in_sync==1
77 * This reduces the burden of testing multiple flags in many cases
81 #define Faulty 1 /* device is known to have a fault */
82 #define In_sync 2 /* device is in_sync with rest of array */
83 #define WriteMostly 4 /* Avoid reading if at all possible */
84 #define BarriersNotsupp 5 /* BIO_RW_BARRIER is not supported */
86 int desc_nr; /* descriptor index in the superblock */
87 int raid_disk; /* role of device in array */
88 int saved_raid_disk; /* role that device used to have in the
89 * array and could again if we did a partial
90 * resync from the bitmap
92 sector_t recovery_offset;/* If this device has been partially
93 * recovered, this is where we were
97 atomic_t nr_pending; /* number of pending requests.
98 * only maintained for arrays that
101 atomic_t read_errors; /* number of consecutive read errors that
102 * we have tried to ignore.
104 atomic_t corrected_errors; /* number of corrected read errors,
105 * for reporting to userspace and storing
113 struct mdk_personality *pers;
116 struct list_head disks;
120 struct gendisk *gendisk;
124 /* Superblock information */
135 sector_t size; /* used size of component devices */
136 sector_t array_size; /* exported array size */
141 /* If the array is being reshaped, we need to record the
142 * new shape and an indication of where we are up to.
143 * This is written to the superblock.
144 * If reshape_position is MaxSector, then no reshape is happening (yet).
146 sector_t reshape_position;
147 int delta_disks, new_level, new_layout, new_chunk;
149 struct mdk_thread_s *thread; /* management thread */
150 struct mdk_thread_s *sync_thread; /* doing resync or reconstruct */
151 sector_t curr_resync; /* blocks scheduled */
152 unsigned long resync_mark; /* a recent timestamp */
153 sector_t resync_mark_cnt;/* blocks written at resync_mark */
155 sector_t resync_max_sectors; /* may be set by personality */
157 sector_t resync_mismatches; /* count of sectors where
158 * parity/replica mismatch found
161 /* allow user-space to request suspension of IO to regions of the array */
164 /* if zero, use the system-wide default */
168 int ok_start_degraded;
169 /* recovery/resync flags
170 * NEEDED: we might need to start a resync/recover
171 * RUNNING: a thread is running, or about to be started
172 * SYNC: actually doing a resync, not a recovery
173 * ERR: and IO error was detected - abort the resync/recovery
174 * INTR: someone requested a (clean) early abort.
175 * DONE: thread is done and is waiting to be reaped
176 * REQUEST: user-space has requested a sync (used with SYNC)
177 * CHECK: user-space request for for check-only, no repair
178 * RESHAPE: A reshape is happening
180 * If neither SYNC or RESHAPE are set, then it is a recovery.
182 #define MD_RECOVERY_RUNNING 0
183 #define MD_RECOVERY_SYNC 1
184 #define MD_RECOVERY_ERR 2
185 #define MD_RECOVERY_INTR 3
186 #define MD_RECOVERY_DONE 4
187 #define MD_RECOVERY_NEEDED 5
188 #define MD_RECOVERY_REQUESTED 6
189 #define MD_RECOVERY_CHECK 7
190 #define MD_RECOVERY_RESHAPE 8
191 #define MD_RECOVERY_FROZEN 9
193 unsigned long recovery;
195 int in_sync; /* know to not need resync */
196 struct mutex reconfig_mutex;
199 int changed; /* true if we might need to reread partition info */
200 int degraded; /* whether md should consider
203 int barriers_work; /* initialised to true, cleared as soon
204 * as a barrier request to slave
205 * fails. Only supported
207 struct bio *biolist; /* bios that need to be retried
208 * because BIO_RW_BARRIER is not supported
211 atomic_t recovery_active; /* blocks scheduled, but not written */
212 wait_queue_head_t recovery_wait;
213 sector_t recovery_cp;
215 spinlock_t write_lock;
216 wait_queue_head_t sb_wait; /* for waiting on superblock updates */
217 atomic_t pending_writes; /* number of active superblock writes */
219 unsigned int safemode; /* if set, update "clean" superblock
220 * when no writes pending.
222 unsigned int safemode_delay;
223 struct timer_list safemode_timer;
224 atomic_t writes_pending;
225 request_queue_t *queue; /* for plugging ... */
227 atomic_t write_behind; /* outstanding async IO */
228 unsigned int max_write_behind; /* 0 = sync */
230 struct bitmap *bitmap; /* the bitmap for the device */
231 struct file *bitmap_file; /* the bitmap file */
232 long bitmap_offset; /* offset from superblock of
233 * start of bitmap. May be
234 * negative, but not '0'
236 long default_bitmap_offset; /* this is the offset to use when
237 * hot-adding a bitmap. It should
238 * eventually be settable by sysfs.
241 struct list_head all_mddevs;
245 static inline void rdev_dec_pending(mdk_rdev_t *rdev, mddev_t *mddev)
247 int faulty = test_bit(Faulty, &rdev->flags);
248 if (atomic_dec_and_test(&rdev->nr_pending) && faulty)
249 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
252 static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors)
254 atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io);
257 struct mdk_personality
261 struct list_head list;
262 struct module *owner;
263 int (*make_request)(request_queue_t *q, struct bio *bio);
264 int (*run)(mddev_t *mddev);
265 int (*stop)(mddev_t *mddev);
266 void (*status)(struct seq_file *seq, mddev_t *mddev);
267 /* error_handler must set ->faulty and clear ->in_sync
268 * if appropriate, and should abort recovery if needed
270 void (*error_handler)(mddev_t *mddev, mdk_rdev_t *rdev);
271 int (*hot_add_disk) (mddev_t *mddev, mdk_rdev_t *rdev);
272 int (*hot_remove_disk) (mddev_t *mddev, int number);
273 int (*spare_active) (mddev_t *mddev);
274 sector_t (*sync_request)(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster);
275 int (*resize) (mddev_t *mddev, sector_t sectors);
276 int (*check_reshape) (mddev_t *mddev);
277 int (*start_reshape) (mddev_t *mddev);
278 int (*reconfig) (mddev_t *mddev, int layout, int chunk_size);
279 /* quiesce moves between quiescence states
281 * 1 - no new requests allowed
284 void (*quiesce) (mddev_t *mddev, int state);
288 struct md_sysfs_entry {
289 struct attribute attr;
290 ssize_t (*show)(mddev_t *, char *);
291 ssize_t (*store)(mddev_t *, const char *, size_t);
295 static inline char * mdname (mddev_t * mddev)
297 return mddev->gendisk ? mddev->gendisk->disk_name : "mdX";
301 * iterates through some rdev ringlist. It's safe to remove the
302 * current 'rdev'. Dont touch 'tmp' though.
304 #define ITERATE_RDEV_GENERIC(head,rdev,tmp) \
306 for ((tmp) = (head).next; \
307 (rdev) = (list_entry((tmp), mdk_rdev_t, same_set)), \
308 (tmp) = (tmp)->next, (tmp)->prev != &(head) \
311 * iterates through the 'same array disks' ringlist
313 #define ITERATE_RDEV(mddev,rdev,tmp) \
314 ITERATE_RDEV_GENERIC((mddev)->disks,rdev,tmp)
317 * Iterates through 'pending RAID disks'
319 #define ITERATE_RDEV_PENDING(rdev,tmp) \
320 ITERATE_RDEV_GENERIC(pending_raid_disks,rdev,tmp)
322 typedef struct mdk_thread_s {
323 void (*run) (mddev_t *mddev);
325 wait_queue_head_t wqueue;
327 struct task_struct *tsk;
328 unsigned long timeout;
331 #define THREAD_WAKEUP 0
333 #define __wait_event_lock_irq(wq, condition, lock, cmd) \
335 wait_queue_t __wait; \
336 init_waitqueue_entry(&__wait, current); \
338 add_wait_queue(&wq, &__wait); \
340 set_current_state(TASK_UNINTERRUPTIBLE); \
343 spin_unlock_irq(&lock); \
346 spin_lock_irq(&lock); \
348 current->state = TASK_RUNNING; \
349 remove_wait_queue(&wq, &__wait); \
352 #define wait_event_lock_irq(wq, condition, lock, cmd) \
356 __wait_event_lock_irq(wq, condition, lock, cmd); \
359 static inline void safe_put_page(struct page *p)