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 #define MAX_CHUNK_SIZE (4096*1024)
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 unsigned long last_events; /* IO event timestamp */
56 struct block_device *bdev; /* block device handle */
60 sector_t data_offset; /* start of data in array */
62 int sb_size; /* bytes in the superblock */
63 int preferred_minor; /* autorun support */
67 /* A device can be in one of three states based on two flags:
68 * Not working: faulty==1 in_sync==0
69 * Fully working: faulty==0 in_sync==1
72 * faulty==0 in_sync==0
74 * It can never have faulty==1, in_sync==1
75 * This reduces the burden of testing multiple flags in many cases
79 #define Faulty 1 /* device is known to have a fault */
80 #define In_sync 2 /* device is in_sync with rest of array */
81 #define WriteMostly 4 /* Avoid reading if at all possible */
82 #define BarriersNotsupp 5 /* BIO_RW_BARRIER is not supported */
84 int desc_nr; /* descriptor index in the superblock */
85 int raid_disk; /* role of device in array */
86 int saved_raid_disk; /* role that device used to have in the
87 * array and could again if we did a partial
88 * resync from the bitmap
91 atomic_t nr_pending; /* number of pending requests.
92 * only maintained for arrays that
95 atomic_t read_errors; /* number of consecutive read errors that
96 * we have tried to ignore.
98 atomic_t corrected_errors; /* number of corrected read errors,
99 * for reporting to userspace and storing
107 struct mdk_personality *pers;
110 struct list_head disks;
114 struct gendisk *gendisk;
118 /* Superblock information */
129 sector_t size; /* used size of component devices */
130 sector_t array_size; /* exported array size */
135 /* If the array is being reshaped, we need to record the
136 * new shape and an indication of where we are up to.
137 * This is written to the superblock.
138 * If reshape_position is MaxSector, then no reshape is happening (yet).
140 sector_t reshape_position;
141 int delta_disks, new_level, new_layout, new_chunk;
143 struct mdk_thread_s *thread; /* management thread */
144 struct mdk_thread_s *sync_thread; /* doing resync or reconstruct */
145 sector_t curr_resync; /* blocks scheduled */
146 unsigned long resync_mark; /* a recent timestamp */
147 sector_t resync_mark_cnt;/* blocks written at resync_mark */
149 sector_t resync_max_sectors; /* may be set by personality */
151 sector_t resync_mismatches; /* count of sectors where
152 * parity/replica mismatch found
154 /* if zero, use the system-wide default */
158 int ok_start_degraded;
159 /* recovery/resync flags
160 * NEEDED: we might need to start a resync/recover
161 * RUNNING: a thread is running, or about to be started
162 * SYNC: actually doing a resync, not a recovery
163 * ERR: and IO error was detected - abort the resync/recovery
164 * INTR: someone requested a (clean) early abort.
165 * DONE: thread is done and is waiting to be reaped
166 * REQUEST: user-space has requested a sync (used with SYNC)
167 * CHECK: user-space request for for check-only, no repair
168 * RESHAPE: A reshape is happening
170 * If neither SYNC or RESHAPE are set, then it is a recovery.
172 #define MD_RECOVERY_RUNNING 0
173 #define MD_RECOVERY_SYNC 1
174 #define MD_RECOVERY_ERR 2
175 #define MD_RECOVERY_INTR 3
176 #define MD_RECOVERY_DONE 4
177 #define MD_RECOVERY_NEEDED 5
178 #define MD_RECOVERY_REQUESTED 6
179 #define MD_RECOVERY_CHECK 7
180 #define MD_RECOVERY_RESHAPE 8
181 unsigned long recovery;
183 int in_sync; /* know to not need resync */
184 struct semaphore reconfig_sem;
187 int changed; /* true if we might need to reread partition info */
188 int degraded; /* whether md should consider
191 int barriers_work; /* initialised to true, cleared as soon
192 * as a barrier request to slave
193 * fails. Only supported
195 struct bio *biolist; /* bios that need to be retried
196 * because BIO_RW_BARRIER is not supported
199 atomic_t recovery_active; /* blocks scheduled, but not written */
200 wait_queue_head_t recovery_wait;
201 sector_t recovery_cp;
203 spinlock_t write_lock;
204 wait_queue_head_t sb_wait; /* for waiting on superblock updates */
205 atomic_t pending_writes; /* number of active superblock writes */
207 unsigned int safemode; /* if set, update "clean" superblock
208 * when no writes pending.
210 unsigned int safemode_delay;
211 struct timer_list safemode_timer;
212 atomic_t writes_pending;
213 request_queue_t *queue; /* for plugging ... */
215 atomic_t write_behind; /* outstanding async IO */
216 unsigned int max_write_behind; /* 0 = sync */
218 struct bitmap *bitmap; /* the bitmap for the device */
219 struct file *bitmap_file; /* the bitmap file */
220 long bitmap_offset; /* offset from superblock of
221 * start of bitmap. May be
222 * negative, but not '0'
224 long default_bitmap_offset; /* this is the offset to use when
225 * hot-adding a bitmap. It should
226 * eventually be settable by sysfs.
229 struct list_head all_mddevs;
233 static inline void rdev_dec_pending(mdk_rdev_t *rdev, mddev_t *mddev)
235 int faulty = test_bit(Faulty, &rdev->flags);
236 if (atomic_dec_and_test(&rdev->nr_pending) && faulty)
237 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
240 static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors)
242 atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io);
245 struct mdk_personality
249 struct list_head list;
250 struct module *owner;
251 int (*make_request)(request_queue_t *q, struct bio *bio);
252 int (*run)(mddev_t *mddev);
253 int (*stop)(mddev_t *mddev);
254 void (*status)(struct seq_file *seq, mddev_t *mddev);
255 /* error_handler must set ->faulty and clear ->in_sync
256 * if appropriate, and should abort recovery if needed
258 void (*error_handler)(mddev_t *mddev, mdk_rdev_t *rdev);
259 int (*hot_add_disk) (mddev_t *mddev, mdk_rdev_t *rdev);
260 int (*hot_remove_disk) (mddev_t *mddev, int number);
261 int (*spare_active) (mddev_t *mddev);
262 sector_t (*sync_request)(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster);
263 int (*resize) (mddev_t *mddev, sector_t sectors);
264 int (*reshape) (mddev_t *mddev, int raid_disks);
265 int (*reconfig) (mddev_t *mddev, int layout, int chunk_size);
266 /* quiesce moves between quiescence states
268 * 1 - no new requests allowed
271 void (*quiesce) (mddev_t *mddev, int state);
275 struct md_sysfs_entry {
276 struct attribute attr;
277 ssize_t (*show)(mddev_t *, char *);
278 ssize_t (*store)(mddev_t *, const char *, size_t);
282 static inline char * mdname (mddev_t * mddev)
284 return mddev->gendisk ? mddev->gendisk->disk_name : "mdX";
288 * iterates through some rdev ringlist. It's safe to remove the
289 * current 'rdev'. Dont touch 'tmp' though.
291 #define ITERATE_RDEV_GENERIC(head,rdev,tmp) \
293 for ((tmp) = (head).next; \
294 (rdev) = (list_entry((tmp), mdk_rdev_t, same_set)), \
295 (tmp) = (tmp)->next, (tmp)->prev != &(head) \
298 * iterates through the 'same array disks' ringlist
300 #define ITERATE_RDEV(mddev,rdev,tmp) \
301 ITERATE_RDEV_GENERIC((mddev)->disks,rdev,tmp)
304 * Iterates through 'pending RAID disks'
306 #define ITERATE_RDEV_PENDING(rdev,tmp) \
307 ITERATE_RDEV_GENERIC(pending_raid_disks,rdev,tmp)
309 typedef struct mdk_thread_s {
310 void (*run) (mddev_t *mddev);
312 wait_queue_head_t wqueue;
314 struct task_struct *tsk;
315 unsigned long timeout;
318 #define THREAD_WAKEUP 0
320 #define __wait_event_lock_irq(wq, condition, lock, cmd) \
322 wait_queue_t __wait; \
323 init_waitqueue_entry(&__wait, current); \
325 add_wait_queue(&wq, &__wait); \
327 set_current_state(TASK_UNINTERRUPTIBLE); \
330 spin_unlock_irq(&lock); \
333 spin_lock_irq(&lock); \
335 current->state = TASK_RUNNING; \
336 remove_wait_queue(&wq, &__wait); \
339 #define wait_event_lock_irq(wq, condition, lock, cmd) \
343 __wait_event_lock_irq(wq, condition, lock, cmd); \
346 static inline void safe_put_page(struct page *p)