]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blame - include/linux/raid/md_k.h
md: allow parallel resync of md-devices.
[mirror_ubuntu-zesty-kernel.git] / include / linux / raid / md_k.h
CommitLineData
1da177e4
LT
1/*
2 md_k.h : kernel internal structure of the Linux MD driver
3 Copyright (C) 1996-98 Ingo Molnar, Gadi Oxman
4
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)
8 any later version.
9
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.
13*/
14
15#ifndef _MD_K_H
16#define _MD_K_H
17
06d91a5f
N
18/* and dm-bio-list.h is not under include/linux because.... ??? */
19#include "../../../drivers/md/dm-bio-list.h"
20
9361401e
DH
21#ifdef CONFIG_BLOCK
22
1da177e4
LT
23#define LEVEL_MULTIPATH (-4)
24#define LEVEL_LINEAR (-1)
25#define LEVEL_FAULTY (-5)
26
2604b703
N
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
30 */
31#define LEVEL_NONE (-1000000)
32
1da177e4 33#define MaxSector (~(sector_t)0)
1da177e4 34
1da177e4
LT
35typedef struct mddev_s mddev_t;
36typedef struct mdk_rdev_s mdk_rdev_t;
37
1da177e4
LT
38/*
39 * options passed in raidrun:
40 */
41
e8703fe1 42/* Currently this must fit in an 'int' */
8932c2e0 43#define MAX_CHUNK_SIZE (1<<30)
1da177e4 44
1da177e4
LT
45/*
46 * MD's 'extended' device
47 */
48struct mdk_rdev_s
49{
50 struct list_head same_set; /* RAID devices within the same set */
51
52 sector_t size; /* Device size (in blocks) */
53 mddev_t *mddev; /* RAID array if running */
713f6ab1 54 long last_events; /* IO event timestamp */
1da177e4
LT
55
56 struct block_device *bdev; /* block device handle */
57
58 struct page *sb_page;
59 int sb_loaded;
42543769 60 __u64 sb_events;
1da177e4
LT
61 sector_t data_offset; /* start of data in array */
62 sector_t sb_offset;
0002b271 63 int sb_size; /* bytes in the superblock */
1da177e4
LT
64 int preferred_minor; /* autorun support */
65
86e6ffdd
N
66 struct kobject kobj;
67
1da177e4
LT
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
71 * Working, but not
72 * in sync with array
73 * faulty==0 in_sync==0
74 *
75 * It can never have faulty==1, in_sync==1
76 * This reduces the burden of testing multiple flags in many cases
77 */
1da177e4 78
b2d444d7
N
79 unsigned long flags;
80#define Faulty 1 /* device is known to have a fault */
81#define In_sync 2 /* device is in_sync with rest of array */
8ddf9efe 82#define WriteMostly 4 /* Avoid reading if at all possible */
a9701a30 83#define BarriersNotsupp 5 /* BIO_RW_BARRIER is not supported */
c5d79adb
N
84#define AllReserved 6 /* If whole device is reserved for
85 * one array */
d0fae18f 86#define AutoDetected 7 /* added by auto-detect */
6bfe0b49
DW
87#define Blocked 8 /* An error occured on an externally
88 * managed array, don't allow writes
89 * until it is cleared */
90 wait_queue_head_t blocked_wait;
8ddf9efe 91
1da177e4
LT
92 int desc_nr; /* descriptor index in the superblock */
93 int raid_disk; /* role of device in array */
41158c7e
N
94 int saved_raid_disk; /* role that device used to have in the
95 * array and could again if we did a partial
96 * resync from the bitmap
97 */
5fd6c1dc
N
98 sector_t recovery_offset;/* If this device has been partially
99 * recovered, this is where we were
100 * up to.
101 */
1da177e4
LT
102
103 atomic_t nr_pending; /* number of pending requests.
104 * only maintained for arrays that
105 * support hot removal
106 */
ba22dcbf
N
107 atomic_t read_errors; /* number of consecutive read errors that
108 * we have tried to ignore.
109 */
4dbcdc75
N
110 atomic_t corrected_errors; /* number of corrected read errors,
111 * for reporting to userspace and storing
112 * in superblock.
113 */
5792a285 114 struct work_struct del_work; /* used for delayed sysfs removal */
1da177e4
LT
115};
116
1da177e4
LT
117struct mddev_s
118{
119 void *private;
2604b703 120 struct mdk_personality *pers;
1da177e4
LT
121 dev_t unit;
122 int md_minor;
123 struct list_head disks;
850b2b42
N
124 unsigned long flags;
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 */
128
1da177e4
LT
129 int ro;
130
131 struct gendisk *gendisk;
132
eae1701f
N
133 struct kobject kobj;
134
1da177e4
LT
135 /* Superblock information */
136 int major_version,
137 minor_version,
138 patch_version;
139 int persistent;
e691063a
N
140 int external; /* metadata is
141 * managed externally */
142 char metadata_type[17]; /* externally set*/
1da177e4
LT
143 int chunk_size;
144 time_t ctime, utime;
145 int level, layout;
d9d166c2 146 char clevel[16];
1da177e4
LT
147 int raid_disks;
148 int max_disks;
149 sector_t size; /* used size of component devices */
150 sector_t array_size; /* exported array size */
151 __u64 events;
152
153 char uuid[16];
154
f6705578
N
155 /* If the array is being reshaped, we need to record the
156 * new shape and an indication of where we are up to.
157 * This is written to the superblock.
158 * If reshape_position is MaxSector, then no reshape is happening (yet).
159 */
160 sector_t reshape_position;
161 int delta_disks, new_level, new_layout, new_chunk;
162
1da177e4
LT
163 struct mdk_thread_s *thread; /* management thread */
164 struct mdk_thread_s *sync_thread; /* doing resync or reconstruct */
ff4e8d9a 165 sector_t curr_resync; /* last block scheduled */
1da177e4
LT
166 unsigned long resync_mark; /* a recent timestamp */
167 sector_t resync_mark_cnt;/* blocks written at resync_mark */
ff4e8d9a 168 sector_t curr_mark_cnt; /* blocks scheduled now */
1da177e4
LT
169
170 sector_t resync_max_sectors; /* may be set by personality */
9d88883e
N
171
172 sector_t resync_mismatches; /* count of sectors where
173 * parity/replica mismatch found
174 */
e464eafd
N
175
176 /* allow user-space to request suspension of IO to regions of the array */
177 sector_t suspend_lo;
178 sector_t suspend_hi;
88202a0c
N
179 /* if zero, use the system-wide default */
180 int sync_speed_min;
181 int sync_speed_max;
182
90b08710
BS
183 /* resync even though the same disks are shared among md-devices */
184 int parallel_resync;
185
6ff8d8ec 186 int ok_start_degraded;
1da177e4
LT
187 /* recovery/resync flags
188 * NEEDED: we might need to start a resync/recover
189 * RUNNING: a thread is running, or about to be started
190 * SYNC: actually doing a resync, not a recovery
191 * ERR: and IO error was detected - abort the resync/recovery
192 * INTR: someone requested a (clean) early abort.
193 * DONE: thread is done and is waiting to be reaped
24dd469d
N
194 * REQUEST: user-space has requested a sync (used with SYNC)
195 * CHECK: user-space request for for check-only, no repair
ccfcc3c1
N
196 * RESHAPE: A reshape is happening
197 *
198 * If neither SYNC or RESHAPE are set, then it is a recovery.
1da177e4
LT
199 */
200#define MD_RECOVERY_RUNNING 0
201#define MD_RECOVERY_SYNC 1
202#define MD_RECOVERY_ERR 2
203#define MD_RECOVERY_INTR 3
204#define MD_RECOVERY_DONE 4
205#define MD_RECOVERY_NEEDED 5
24dd469d
N
206#define MD_RECOVERY_REQUESTED 6
207#define MD_RECOVERY_CHECK 7
ccfcc3c1 208#define MD_RECOVERY_RESHAPE 8
5fd6c1dc
N
209#define MD_RECOVERY_FROZEN 9
210
1da177e4
LT
211 unsigned long recovery;
212
213 int in_sync; /* know to not need resync */
df5b89b3 214 struct mutex reconfig_mutex;
1da177e4
LT
215 atomic_t active;
216
44ce6294 217 int changed; /* true if we might need to reread partition info */
1da177e4
LT
218 int degraded; /* whether md should consider
219 * adding a spare
220 */
a9701a30
N
221 int barriers_work; /* initialised to true, cleared as soon
222 * as a barrier request to slave
223 * fails. Only supported
224 */
225 struct bio *biolist; /* bios that need to be retried
226 * because BIO_RW_BARRIER is not supported
227 */
1da177e4
LT
228
229 atomic_t recovery_active; /* blocks scheduled, but not written */
230 wait_queue_head_t recovery_wait;
231 sector_t recovery_cp;
c6207277
N
232 sector_t resync_max; /* resync should pause
233 * when it gets here */
06d91a5f
N
234
235 spinlock_t write_lock;
3d310eb7 236 wait_queue_head_t sb_wait; /* for waiting on superblock updates */
7bfa19f2 237 atomic_t pending_writes; /* number of active superblock writes */
06d91a5f 238
1da177e4
LT
239 unsigned int safemode; /* if set, update "clean" superblock
240 * when no writes pending.
241 */
242 unsigned int safemode_delay;
243 struct timer_list safemode_timer;
244 atomic_t writes_pending;
165125e1 245 struct request_queue *queue; /* for plugging ... */
1da177e4 246
4b6d287f
N
247 atomic_t write_behind; /* outstanding async IO */
248 unsigned int max_write_behind; /* 0 = sync */
249
32a7627c
N
250 struct bitmap *bitmap; /* the bitmap for the device */
251 struct file *bitmap_file; /* the bitmap file */
a654b9d8
N
252 long bitmap_offset; /* offset from superblock of
253 * start of bitmap. May be
254 * negative, but not '0'
255 */
36fa3063
N
256 long default_bitmap_offset; /* this is the offset to use when
257 * hot-adding a bitmap. It should
258 * eventually be settable by sysfs.
259 */
32a7627c 260
1da177e4
LT
261 struct list_head all_mddevs;
262};
263
264
265static inline void rdev_dec_pending(mdk_rdev_t *rdev, mddev_t *mddev)
266{
b2d444d7 267 int faulty = test_bit(Faulty, &rdev->flags);
1da177e4
LT
268 if (atomic_dec_and_test(&rdev->nr_pending) && faulty)
269 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
270}
271
272static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors)
273{
274 atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io);
275}
276
2604b703 277struct mdk_personality
1da177e4
LT
278{
279 char *name;
2604b703
N
280 int level;
281 struct list_head list;
1da177e4 282 struct module *owner;
165125e1 283 int (*make_request)(struct request_queue *q, struct bio *bio);
1da177e4
LT
284 int (*run)(mddev_t *mddev);
285 int (*stop)(mddev_t *mddev);
286 void (*status)(struct seq_file *seq, mddev_t *mddev);
287 /* error_handler must set ->faulty and clear ->in_sync
288 * if appropriate, and should abort recovery if needed
289 */
290 void (*error_handler)(mddev_t *mddev, mdk_rdev_t *rdev);
291 int (*hot_add_disk) (mddev_t *mddev, mdk_rdev_t *rdev);
292 int (*hot_remove_disk) (mddev_t *mddev, int number);
293 int (*spare_active) (mddev_t *mddev);
57afd89f 294 sector_t (*sync_request)(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster);
1da177e4 295 int (*resize) (mddev_t *mddev, sector_t sectors);
63c70c4f
N
296 int (*check_reshape) (mddev_t *mddev);
297 int (*start_reshape) (mddev_t *mddev);
1da177e4 298 int (*reconfig) (mddev_t *mddev, int layout, int chunk_size);
36fa3063
N
299 /* quiesce moves between quiescence states
300 * 0 - fully active
301 * 1 - no new requests allowed
302 * others - reserved
303 */
304 void (*quiesce) (mddev_t *mddev, int state);
1da177e4
LT
305};
306
307
007583c9
N
308struct md_sysfs_entry {
309 struct attribute attr;
310 ssize_t (*show)(mddev_t *, char *);
311 ssize_t (*store)(mddev_t *, const char *, size_t);
312};
313
314
1da177e4
LT
315static inline char * mdname (mddev_t * mddev)
316{
317 return mddev->gendisk ? mddev->gendisk->disk_name : "mdX";
318}
319
1da177e4
LT
320/*
321 * iterates through some rdev ringlist. It's safe to remove the
322 * current 'rdev'. Dont touch 'tmp' though.
323 */
73c34431 324#define rdev_for_each_list(rdev, tmp, list) \
1da177e4 325 \
73c34431 326 for ((tmp) = (list).next; \
1da177e4 327 (rdev) = (list_entry((tmp), mdk_rdev_t, same_set)), \
73c34431 328 (tmp) = (tmp)->next, (tmp)->prev != &(list) \
1da177e4
LT
329 ; )
330/*
331 * iterates through the 'same array disks' ringlist
332 */
d089c6af 333#define rdev_for_each(rdev, tmp, mddev) \
73c34431 334 rdev_for_each_list(rdev, tmp, (mddev)->disks)
1da177e4
LT
335
336typedef struct mdk_thread_s {
337 void (*run) (mddev_t *mddev);
338 mddev_t *mddev;
339 wait_queue_head_t wqueue;
340 unsigned long flags;
1da177e4 341 struct task_struct *tsk;
32a7627c 342 unsigned long timeout;
1da177e4
LT
343} mdk_thread_t;
344
345#define THREAD_WAKEUP 0
346
347#define __wait_event_lock_irq(wq, condition, lock, cmd) \
348do { \
349 wait_queue_t __wait; \
350 init_waitqueue_entry(&__wait, current); \
351 \
352 add_wait_queue(&wq, &__wait); \
353 for (;;) { \
354 set_current_state(TASK_UNINTERRUPTIBLE); \
355 if (condition) \
356 break; \
357 spin_unlock_irq(&lock); \
358 cmd; \
359 schedule(); \
360 spin_lock_irq(&lock); \
361 } \
362 current->state = TASK_RUNNING; \
363 remove_wait_queue(&wq, &__wait); \
364} while (0)
365
366#define wait_event_lock_irq(wq, condition, lock, cmd) \
367do { \
368 if (condition) \
369 break; \
370 __wait_event_lock_irq(wq, condition, lock, cmd); \
371} while (0)
372
1345b1d8
N
373static inline void safe_put_page(struct page *p)
374{
375 if (p) put_page(p);
376}
377
9361401e 378#endif /* CONFIG_BLOCK */
1da177e4
LT
379#endif
380