blob: 7b4a3c318cae437ecb2041c33b2bdcee2c66bf15 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
Wang Sheng-Hui7e841522011-09-21 15:37:46 +10002 md.h : kernel internal structure of the Linux MD driver
Linus Torvalds1da177e2005-04-16 15:20:36 -07003 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
Christoph Hellwig63fe0812009-04-14 12:01:53 +100015#ifndef _MD_MD_H
16#define _MD_MD_H
Linus Torvalds1da177e2005-04-16 15:20:36 -070017
Christoph Hellwig63fe0812009-04-14 12:01:53 +100018#include <linux/blkdev.h>
19#include <linux/kobject.h>
20#include <linux/list.h>
21#include <linux/mm.h>
22#include <linux/mutex.h>
23#include <linux/timer.h>
24#include <linux/wait.h>
25#include <linux/workqueue.h>
David Howells93614012006-09-30 20:45:40 +020026
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#define MaxSector (~(sector_t)0)
Linus Torvalds1da177e2005-04-16 15:20:36 -070028
NeilBrown2230dfe2011-07-28 11:31:46 +100029/* Bad block numbers are stored sorted in a single page.
30 * 64bits is used for each block or extent.
31 * 54 bits are sector number, 9 bits are extent size,
32 * 1 bit is an 'acknowledged' flag.
33 */
34#define MD_MAX_BADBLOCKS (PAGE_SIZE/8)
35
Linus Torvalds1da177e2005-04-16 15:20:36 -070036/*
Linus Torvalds1da177e2005-04-16 15:20:36 -070037 * MD's 'extended' device
38 */
NeilBrown3cb03002011-10-11 16:45:26 +110039struct md_rdev {
Linus Torvalds1da177e2005-04-16 15:20:36 -070040 struct list_head same_set; /* RAID devices within the same set */
41
Andre Nolldd8ac332009-03-31 14:33:13 +110042 sector_t sectors; /* Device size (in 512bytes sectors) */
NeilBrownfd01b882011-10-11 16:47:53 +110043 struct mddev *mddev; /* RAID array if running */
NeilBrowneea1bf32009-03-31 14:27:02 +110044 int last_events; /* IO event timestamp */
Linus Torvalds1da177e2005-04-16 15:20:36 -070045
Jonathan Brassowa6ff7e02011-01-14 09:14:34 +110046 /*
47 * If meta_bdev is non-NULL, it means that a separate device is
48 * being used to store the metadata (superblock/bitmap) which
49 * would otherwise be contained on the same device as the data (bdev).
50 */
51 struct block_device *meta_bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -070052 struct block_device *bdev; /* block device handle */
53
NeilBrown2699b672011-07-28 11:31:47 +100054 struct page *sb_page, *bb_page;
Linus Torvalds1da177e2005-04-16 15:20:36 -070055 int sb_loaded;
NeilBrown42543762006-06-26 00:27:57 -070056 __u64 sb_events;
Linus Torvalds1da177e2005-04-16 15:20:36 -070057 sector_t data_offset; /* start of data in array */
NeilBrownc6563a82012-05-21 09:27:00 +100058 sector_t new_data_offset;/* only relevant while reshaping */
Andre Noll0f420352008-07-11 22:02:23 +100059 sector_t sb_start; /* offset of the super block (in 512byte sectors) */
NeilBrown0002b272005-09-09 16:23:53 -070060 int sb_size; /* bytes in the superblock */
Linus Torvalds1da177e2005-04-16 15:20:36 -070061 int preferred_minor; /* autorun support */
62
NeilBrown86e6ffd2005-11-08 21:39:24 -080063 struct kobject kobj;
64
Linus Torvalds1da177e2005-04-16 15:20:36 -070065 /* A device can be in one of three states based on two flags:
66 * Not working: faulty==1 in_sync==0
67 * Fully working: faulty==0 in_sync==1
68 * Working, but not
69 * in sync with array
70 * faulty==0 in_sync==0
71 *
72 * It can never have faulty==1, in_sync==1
73 * This reduces the burden of testing multiple flags in many cases
74 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070075
NeilBrown2d78f8c2011-12-23 10:17:51 +110076 unsigned long flags; /* bit set of 'enum flag_bits' bits. */
Dan Williams6bfe0b42008-04-30 00:52:32 -070077 wait_queue_head_t blocked_wait;
NeilBrown8ddf9ef2005-09-09 16:23:45 -070078
Linus Torvalds1da177e2005-04-16 15:20:36 -070079 int desc_nr; /* descriptor index in the superblock */
80 int raid_disk; /* role of device in array */
NeilBrowne93f68a2010-06-15 09:36:03 +010081 int new_raid_disk; /* role that the device will have in
82 * the array after a level-change completes.
83 */
NeilBrown41158c72005-06-21 17:17:25 -070084 int saved_raid_disk; /* role that device used to have in the
85 * array and could again if we did a partial
86 * resync from the bitmap
87 */
NeilBrown5fd6c1d2006-06-26 00:27:40 -070088 sector_t recovery_offset;/* If this device has been partially
89 * recovered, this is where we were
90 * up to.
91 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070092
93 atomic_t nr_pending; /* number of pending requests.
94 * only maintained for arrays that
95 * support hot removal
96 */
NeilBrownba22dcb2005-11-08 21:39:31 -080097 atomic_t read_errors; /* number of consecutive read errors that
98 * we have tried to ignore.
99 */
Robert Becker1e509152009-12-14 12:49:58 +1100100 struct timespec last_read_error; /* monotonic time since our
101 * last read error
102 */
NeilBrown4dbcdc72006-01-06 00:20:52 -0800103 atomic_t corrected_errors; /* number of corrected read errors,
104 * for reporting to userspace and storing
105 * in superblock.
106 */
NeilBrown5792a282007-04-04 19:08:18 -0700107 struct work_struct del_work; /* used for delayed sysfs removal */
NeilBrown3c0ee632008-10-21 13:25:28 +1100108
109 struct sysfs_dirent *sysfs_state; /* handle for 'state'
110 * sysfs entry */
NeilBrown2230dfe2011-07-28 11:31:46 +1000111
112 struct badblocks {
113 int count; /* count of bad blocks */
NeilBrownde393cd2011-07-28 11:31:48 +1000114 int unacked_exist; /* there probably are unacknowledged
115 * bad blocks. This is only cleared
116 * when a read discovers none
117 */
NeilBrown2230dfe2011-07-28 11:31:46 +1000118 int shift; /* shift from sectors to block size
119 * a -ve shift means badblocks are
120 * disabled.*/
121 u64 *page; /* badblock list */
122 int changed;
123 seqlock_t lock;
NeilBrown2699b672011-07-28 11:31:47 +1000124
125 sector_t sector;
126 sector_t size; /* in sectors */
NeilBrown2230dfe2011-07-28 11:31:46 +1000127 } badblocks;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128};
NeilBrown2d78f8c2011-12-23 10:17:51 +1100129enum flag_bits {
130 Faulty, /* device is known to have a fault */
131 In_sync, /* device is in_sync with rest of array */
NeilBrown050b6612012-03-19 12:46:39 +1100132 Unmerged, /* device is being added to array and should
133 * be considerred for bvec_merge_fn but not
134 * yet for actual IO
135 */
NeilBrown2d78f8c2011-12-23 10:17:51 +1100136 WriteMostly, /* Avoid reading if at all possible */
137 AutoDetected, /* added by auto-detect */
138 Blocked, /* An error occurred but has not yet
139 * been acknowledged by the metadata
140 * handler, so don't allow writes
141 * until it is cleared */
142 WriteErrorSeen, /* A write error has been seen on this
143 * device
144 */
145 FaultRecorded, /* Intermediate state for clearing
146 * Blocked. The Fault is/will-be
147 * recorded in the metadata, but that
148 * metadata hasn't been stored safely
149 * on disk yet.
150 */
151 BlockedBadBlocks, /* A writer is blocked because they
152 * found an unacknowledged bad-block.
153 * This can safely be cleared at any
154 * time, and the writer will re-check.
155 * It may be set at any time, and at
156 * worst the writer will timeout and
157 * re-check. So setting it as
158 * accurately as possible is good, but
159 * not absolutely critical.
160 */
161 WantReplacement, /* This device is a candidate to be
162 * hot-replaced, either because it has
163 * reported some faults, or because
164 * of explicit request.
165 */
166 Replacement, /* This device is a replacement for
167 * a want_replacement device with same
168 * raid_disk number.
169 */
170};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171
NeilBrown2230dfe2011-07-28 11:31:46 +1000172#define BB_LEN_MASK (0x00000000000001FFULL)
173#define BB_OFFSET_MASK (0x7FFFFFFFFFFFFE00ULL)
174#define BB_ACK_MASK (0x8000000000000000ULL)
175#define BB_MAX_LEN 512
176#define BB_OFFSET(x) (((x) & BB_OFFSET_MASK) >> 9)
177#define BB_LEN(x) (((x) & BB_LEN_MASK) + 1)
178#define BB_ACK(x) (!!((x) & BB_ACK_MASK))
179#define BB_MAKE(a, l, ack) (((a)<<9) | ((l)-1) | ((u64)(!!(ack)) << 63))
180
181extern int md_is_badblock(struct badblocks *bb, sector_t s, int sectors,
182 sector_t *first_bad, int *bad_sectors);
NeilBrown3cb03002011-10-11 16:45:26 +1100183static inline int is_badblock(struct md_rdev *rdev, sector_t s, int sectors,
NeilBrown2230dfe2011-07-28 11:31:46 +1000184 sector_t *first_bad, int *bad_sectors)
185{
186 if (unlikely(rdev->badblocks.count)) {
187 int rv = md_is_badblock(&rdev->badblocks, rdev->data_offset + s,
188 sectors,
189 first_bad, bad_sectors);
190 if (rv)
191 *first_bad -= rdev->data_offset;
192 return rv;
193 }
194 return 0;
195}
NeilBrown3cb03002011-10-11 16:45:26 +1100196extern int rdev_set_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
NeilBrownc6563a82012-05-21 09:27:00 +1000197 int is_new);
198extern int rdev_clear_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
199 int is_new);
NeilBrown2230dfe2011-07-28 11:31:46 +1000200extern void md_ack_all_badblocks(struct badblocks *bb);
201
NeilBrownfd01b882011-10-11 16:47:53 +1100202struct mddev {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203 void *private;
NeilBrown84fc4b52011-10-11 16:49:58 +1100204 struct md_personality *pers;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205 dev_t unit;
206 int md_minor;
207 struct list_head disks;
NeilBrown850b2b42006-10-03 01:15:46 -0700208 unsigned long flags;
209#define MD_CHANGE_DEVS 0 /* Some device status has changed */
210#define MD_CHANGE_CLEAN 1 /* transition to or from 'clean' */
NeilBrown070dc6d2010-08-30 17:33:34 +1000211#define MD_CHANGE_PENDING 2 /* switch from 'clean' to 'active' in progress */
Jonathan Brassow9c810752011-06-08 17:59:30 -0500212#define MD_ARRAY_FIRST_USE 3 /* First use of array, needs initialization */
NeilBrown850b2b42006-10-03 01:15:46 -0700213
NeilBrown409c57f2009-03-31 14:39:39 +1100214 int suspended;
215 atomic_t active_io;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216 int ro;
NeilBrownbb4f1e92010-08-08 21:18:03 +1000217 int sysfs_active; /* set when sysfs deletes
218 * are happening, so run/
219 * takeover/stop are not safe
220 */
NeilBrown0ca69882011-01-14 09:14:33 +1100221 int ready; /* See when safe to pass
222 * IO requests down */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223 struct gendisk *gendisk;
224
NeilBrowneae17012005-11-08 21:39:23 -0800225 struct kobject kobj;
NeilBrownd3374822009-01-09 08:31:10 +1100226 int hold_active;
227#define UNTIL_IOCTL 1
NeilBrownefeb53c2009-01-09 08:31:10 +1100228#define UNTIL_STOP 2
NeilBrowneae17012005-11-08 21:39:23 -0800229
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230 /* Superblock information */
231 int major_version,
232 minor_version,
233 patch_version;
234 int persistent;
NeilBrowne6910632008-02-06 01:39:51 -0800235 int external; /* metadata is
236 * managed externally */
237 char metadata_type[17]; /* externally set*/
Andre Noll9d8f0362009-06-18 08:45:01 +1000238 int chunk_sectors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239 time_t ctime, utime;
240 int level, layout;
NeilBrownd9d166c2006-01-06 00:20:51 -0800241 char clevel[16];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242 int raid_disks;
243 int max_disks;
Andre Noll58c0fed2009-03-31 14:33:13 +1100244 sector_t dev_sectors; /* used size of
245 * component devices */
Andre Nollf233ea52008-07-21 17:05:22 +1000246 sector_t array_sectors; /* exported array size */
Dan Williamsb522adc2009-03-31 15:00:31 +1100247 int external_size; /* size managed
248 * externally */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249 __u64 events;
NeilBrowna8707c02010-05-18 09:28:43 +1000250 /* If the last 'event' was simply a clean->dirty transition, and
251 * we didn't write it to the spares, then it is safe and simple
252 * to just decrement the event count on a dirty->clean transition.
253 * So we record that possibility here.
254 */
255 int can_decrease_events;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256
257 char uuid[16];
258
NeilBrownf6705572006-03-27 01:18:11 -0800259 /* If the array is being reshaped, we need to record the
260 * new shape and an indication of where we are up to.
261 * This is written to the superblock.
262 * If reshape_position is MaxSector, then no reshape is happening (yet).
263 */
264 sector_t reshape_position;
Andre Noll664e7c42009-06-18 08:45:27 +1000265 int delta_disks, new_level, new_layout;
266 int new_chunk_sectors;
NeilBrown2c810cd2012-05-21 09:27:00 +1000267 int reshape_backwards;
NeilBrownf6705572006-03-27 01:18:11 -0800268
NeilBrown97658cd2011-04-18 18:25:42 +1000269 atomic_t plug_cnt; /* If device is expecting
270 * more bios soon.
271 */
NeilBrown2b8bf342011-10-11 16:48:23 +1100272 struct md_thread *thread; /* management thread */
273 struct md_thread *sync_thread; /* doing resync or reconstruct */
NeilBrownff4e8d92006-07-10 04:44:16 -0700274 sector_t curr_resync; /* last block scheduled */
NeilBrown97e4f422009-03-31 14:33:13 +1100275 /* As resync requests can complete out of order, we cannot easily track
276 * how much resync has been completed. So we occasionally pause until
277 * everything completes, then set curr_resync_completed to curr_resync.
278 * As such it may be well behind the real resync mark, but it is a value
279 * we are certain of.
280 */
281 sector_t curr_resync_completed;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282 unsigned long resync_mark; /* a recent timestamp */
283 sector_t resync_mark_cnt;/* blocks written at resync_mark */
NeilBrownff4e8d92006-07-10 04:44:16 -0700284 sector_t curr_mark_cnt; /* blocks scheduled now */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285
286 sector_t resync_max_sectors; /* may be set by personality */
NeilBrown9d888832005-11-08 21:39:26 -0800287
288 sector_t resync_mismatches; /* count of sectors where
289 * parity/replica mismatch found
290 */
NeilBrowne464eaf2006-03-27 01:18:14 -0800291
292 /* allow user-space to request suspension of IO to regions of the array */
293 sector_t suspend_lo;
294 sector_t suspend_hi;
NeilBrown88202a02006-01-06 00:21:36 -0800295 /* if zero, use the system-wide default */
296 int sync_speed_min;
297 int sync_speed_max;
298
Bernd Schubert90b08712008-05-23 13:04:38 -0700299 /* resync even though the same disks are shared among md-devices */
300 int parallel_resync;
301
NeilBrown6ff8d8ec2006-01-06 00:20:15 -0800302 int ok_start_degraded;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303 /* recovery/resync flags
304 * NEEDED: we might need to start a resync/recover
305 * RUNNING: a thread is running, or about to be started
306 * SYNC: actually doing a resync, not a recovery
Neil Brown72a23c22008-06-28 08:31:41 +1000307 * RECOVER: doing recovery, or need to try it.
NeilBrowndfc70642008-05-23 13:04:39 -0700308 * INTR: resync needs to be aborted for some reason
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309 * DONE: thread is done and is waiting to be reaped
NeilBrown24dd4692005-11-08 21:39:26 -0800310 * REQUEST: user-space has requested a sync (used with SYNC)
Anand Gadiyar411c9402009-07-07 15:24:23 +0530311 * CHECK: user-space request for check-only, no repair
NeilBrownccfcc3c2006-03-27 01:18:09 -0800312 * RESHAPE: A reshape is happening
313 *
314 * If neither SYNC or RESHAPE are set, then it is a recovery.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315 */
316#define MD_RECOVERY_RUNNING 0
317#define MD_RECOVERY_SYNC 1
Neil Brown72a23c22008-06-28 08:31:41 +1000318#define MD_RECOVERY_RECOVER 2
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319#define MD_RECOVERY_INTR 3
320#define MD_RECOVERY_DONE 4
321#define MD_RECOVERY_NEEDED 5
NeilBrown24dd4692005-11-08 21:39:26 -0800322#define MD_RECOVERY_REQUESTED 6
323#define MD_RECOVERY_CHECK 7
NeilBrownccfcc3c2006-03-27 01:18:09 -0800324#define MD_RECOVERY_RESHAPE 8
NeilBrown5fd6c1d2006-06-26 00:27:40 -0700325#define MD_RECOVERY_FROZEN 9
326
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327 unsigned long recovery;
NeilBrown53890422011-07-27 11:00:36 +1000328 /* If a RAID personality determines that recovery (of a particular
329 * device) will fail due to a read error on the source device, it
330 * takes a copy of this number and does not attempt recovery again
331 * until this number changes.
332 */
333 int recovery_disabled;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334
335 int in_sync; /* know to not need resync */
NeilBrownc8c00a62009-08-10 12:50:52 +1000336 /* 'open_mutex' avoids races between 'md_open' and 'do_md_stop', so
337 * that we are never stopping an array while it is open.
338 * 'reconfig_mutex' protects all other reconfiguration.
339 * These locks are separate due to conflicting interactions
340 * with bdev->bd_mutex.
341 * Lock ordering is:
342 * reconfig_mutex -> bd_mutex : e.g. do_md_run -> revalidate_disk
343 * bd_mutex -> open_mutex: e.g. __blkdev_get -> md_open
344 */
345 struct mutex open_mutex;
NeilBrowndf5b89b2006-03-27 01:18:20 -0800346 struct mutex reconfig_mutex;
NeilBrownf2ea68c2008-07-21 17:05:25 +1000347 atomic_t active; /* general refcount */
348 atomic_t openers; /* number of active opens */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349
NeilBrownf0b4f7e2011-02-24 17:26:41 +1100350 int changed; /* True if we might need to
351 * reread partition info */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352 int degraded; /* whether md should consider
353 * adding a spare
354 */
NeilBrown050b6612012-03-19 12:46:39 +1100355 int merge_check_needed; /* at least one
356 * member device
357 * has a
358 * merge_bvec_fn */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359
360 atomic_t recovery_active; /* blocks scheduled, but not written */
361 wait_queue_head_t recovery_wait;
362 sector_t recovery_cp;
Neil Brown5e96ee62008-06-28 08:31:24 +1000363 sector_t resync_min; /* user requested sync
364 * starts here */
NeilBrownc6207272008-02-06 01:39:52 -0800365 sector_t resync_max; /* resync should pause
366 * when it gets here */
NeilBrown06d91a52005-06-21 17:17:12 -0700367
NeilBrownb62b7592008-10-21 13:25:21 +1100368 struct sysfs_dirent *sysfs_state; /* handle for 'array_state'
369 * file in sysfs.
370 */
NeilBrown0c3573f2009-01-09 08:31:05 +1100371 struct sysfs_dirent *sysfs_action; /* handle for 'sync_action' */
NeilBrownb62b7592008-10-21 13:25:21 +1100372
NeilBrownd3374822009-01-09 08:31:10 +1100373 struct work_struct del_work; /* used for delayed sysfs removal */
374
NeilBrown06d91a52005-06-21 17:17:12 -0700375 spinlock_t write_lock;
NeilBrown3d310eb2005-06-21 17:17:26 -0700376 wait_queue_head_t sb_wait; /* for waiting on superblock updates */
NeilBrown7bfa19f2005-06-21 17:17:28 -0700377 atomic_t pending_writes; /* number of active superblock writes */
NeilBrown06d91a52005-06-21 17:17:12 -0700378
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379 unsigned int safemode; /* if set, update "clean" superblock
380 * when no writes pending.
381 */
382 unsigned int safemode_delay;
383 struct timer_list safemode_timer;
384 atomic_t writes_pending;
Jens Axboe165125e2007-07-24 09:28:11 +0200385 struct request_queue *queue; /* for plugging ... */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386
NeilBrown32a76272005-06-21 17:17:14 -0700387 struct bitmap *bitmap; /* the bitmap for the device */
NeilBrownc3d97142009-12-14 12:49:52 +1100388 struct {
389 struct file *file; /* the bitmap file */
NeilBrownf6af9492009-12-14 12:49:54 +1100390 loff_t offset; /* offset from superblock of
NeilBrownc3d97142009-12-14 12:49:52 +1100391 * start of bitmap. May be
392 * negative, but not '0'
NeilBrownf6af9492009-12-14 12:49:54 +1100393 * For external metadata, offset
394 * from start of device.
NeilBrownc3d97142009-12-14 12:49:52 +1100395 */
NeilBrown6409bb02012-05-22 13:55:07 +1000396 unsigned long space; /* space available at this offset */
NeilBrownf6af9492009-12-14 12:49:54 +1100397 loff_t default_offset; /* this is the offset to use when
NeilBrownc3d97142009-12-14 12:49:52 +1100398 * hot-adding a bitmap. It should
399 * eventually be settable by sysfs.
400 */
NeilBrown6409bb02012-05-22 13:55:07 +1000401 unsigned long default_space; /* space available at
402 * default offset */
NeilBrownc3d97142009-12-14 12:49:52 +1100403 struct mutex mutex;
NeilBrown42a04b52009-12-14 12:49:53 +1100404 unsigned long chunksize;
NeilBrownac2f40b2010-06-01 19:37:31 +1000405 unsigned long daemon_sleep; /* how many jiffies between updates? */
NeilBrown42a04b52009-12-14 12:49:53 +1100406 unsigned long max_write_behind; /* write-behind mode */
NeilBrownece5cff2009-12-14 12:49:56 +1100407 int external;
NeilBrownc3d97142009-12-14 12:49:52 +1100408 } bitmap_info;
NeilBrown32a76272005-06-21 17:17:14 -0700409
Robert Becker1e509152009-12-14 12:49:58 +1100410 atomic_t max_corr_read_errors; /* max read retries */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411 struct list_head all_mddevs;
NeilBrowna2826aa2009-12-14 12:49:49 +1100412
NeilBrowna64c8762010-04-14 17:15:37 +1000413 struct attribute_group *to_remove;
NeilBrown252ac522010-06-01 19:37:29 +1000414
NeilBrowna167f662010-10-26 18:31:13 +1100415 struct bio_set *bio_set;
416
Tejun Heoe9c74692010-09-03 11:56:18 +0200417 /* Generic flush handling.
418 * The last to finish preflush schedules a worker to submit
419 * the rest of the request (without the REQ_FLUSH flag).
NeilBrowna2826aa2009-12-14 12:49:49 +1100420 */
Tejun Heoe9c74692010-09-03 11:56:18 +0200421 struct bio *flush_bio;
NeilBrowna2826aa2009-12-14 12:49:49 +1100422 atomic_t flush_pending;
Tejun Heoe9c74692010-09-03 11:56:18 +0200423 struct work_struct flush_work;
NeilBrown768a4182010-07-26 11:49:55 +1000424 struct work_struct event_work; /* used by dm to report failure event */
NeilBrownfd01b882011-10-11 16:47:53 +1100425 void (*sync_super)(struct mddev *mddev, struct md_rdev *rdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426};
427
428
NeilBrownfd01b882011-10-11 16:47:53 +1100429static inline void rdev_dec_pending(struct md_rdev *rdev, struct mddev *mddev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430{
NeilBrownb2d444d2005-11-08 21:39:31 -0800431 int faulty = test_bit(Faulty, &rdev->flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 if (atomic_dec_and_test(&rdev->nr_pending) && faulty)
433 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
434}
435
436static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors)
437{
438 atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io);
439}
440
NeilBrown84fc4b52011-10-11 16:49:58 +1100441struct md_personality
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442{
443 char *name;
NeilBrown2604b702006-01-06 00:20:36 -0800444 int level;
445 struct list_head list;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446 struct module *owner;
Linus Torvaldsb4fdcb02011-11-04 17:06:58 -0700447 void (*make_request)(struct mddev *mddev, struct bio *bio);
NeilBrownfd01b882011-10-11 16:47:53 +1100448 int (*run)(struct mddev *mddev);
449 int (*stop)(struct mddev *mddev);
450 void (*status)(struct seq_file *seq, struct mddev *mddev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451 /* error_handler must set ->faulty and clear ->in_sync
452 * if appropriate, and should abort recovery if needed
453 */
NeilBrownfd01b882011-10-11 16:47:53 +1100454 void (*error_handler)(struct mddev *mddev, struct md_rdev *rdev);
455 int (*hot_add_disk) (struct mddev *mddev, struct md_rdev *rdev);
NeilBrownb8321b62011-12-23 10:17:51 +1100456 int (*hot_remove_disk) (struct mddev *mddev, struct md_rdev *rdev);
NeilBrownfd01b882011-10-11 16:47:53 +1100457 int (*spare_active) (struct mddev *mddev);
458 sector_t (*sync_request)(struct mddev *mddev, sector_t sector_nr, int *skipped, int go_faster);
459 int (*resize) (struct mddev *mddev, sector_t sectors);
460 sector_t (*size) (struct mddev *mddev, sector_t sectors, int raid_disks);
461 int (*check_reshape) (struct mddev *mddev);
462 int (*start_reshape) (struct mddev *mddev);
463 void (*finish_reshape) (struct mddev *mddev);
NeilBrown36fa3062005-09-09 16:23:45 -0700464 /* quiesce moves between quiescence states
465 * 0 - fully active
466 * 1 - no new requests allowed
467 * others - reserved
468 */
NeilBrownfd01b882011-10-11 16:47:53 +1100469 void (*quiesce) (struct mddev *mddev, int state);
NeilBrown245f46c2009-03-31 14:39:39 +1100470 /* takeover is used to transition an array from one
471 * personality to another. The new personality must be able
472 * to handle the data in the current layout.
473 * e.g. 2drive raid1 -> 2drive raid5
474 * ndrive raid5 -> degraded n+1drive raid6 with special layout
475 * If the takeover succeeds, a new 'private' structure is returned.
476 * This needs to be installed and then ->run used to activate the
477 * array.
478 */
NeilBrownfd01b882011-10-11 16:47:53 +1100479 void *(*takeover) (struct mddev *mddev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480};
481
482
NeilBrown007583c2005-11-08 21:39:30 -0800483struct md_sysfs_entry {
484 struct attribute attr;
NeilBrownfd01b882011-10-11 16:47:53 +1100485 ssize_t (*show)(struct mddev *, char *);
486 ssize_t (*store)(struct mddev *, const char *, size_t);
NeilBrown007583c2005-11-08 21:39:30 -0800487};
NeilBrown43a70502009-12-14 12:49:55 +1100488extern struct attribute_group md_bitmap_group;
NeilBrown007583c2005-11-08 21:39:30 -0800489
NeilBrown00bcb4a2010-06-01 19:37:23 +1000490static inline struct sysfs_dirent *sysfs_get_dirent_safe(struct sysfs_dirent *sd, char *name)
491{
492 if (sd)
493 return sysfs_get_dirent(sd, NULL, name);
494 return sd;
495}
496static inline void sysfs_notify_dirent_safe(struct sysfs_dirent *sd)
497{
498 if (sd)
499 sysfs_notify_dirent(sd);
500}
501
NeilBrownfd01b882011-10-11 16:47:53 +1100502static inline char * mdname (struct mddev * mddev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503{
504 return mddev->gendisk ? mddev->gendisk->disk_name : "mdX";
505}
506
NeilBrownfd01b882011-10-11 16:47:53 +1100507static inline int sysfs_link_rdev(struct mddev *mddev, struct md_rdev *rdev)
Namhyung Kim36fad852011-07-27 11:00:36 +1000508{
509 char nm[20];
NeilBrown2d78f8c2011-12-23 10:17:51 +1100510 if (!test_bit(Replacement, &rdev->flags)) {
511 sprintf(nm, "rd%d", rdev->raid_disk);
512 return sysfs_create_link(&mddev->kobj, &rdev->kobj, nm);
513 } else
514 return 0;
Namhyung Kim36fad852011-07-27 11:00:36 +1000515}
516
NeilBrownfd01b882011-10-11 16:47:53 +1100517static inline void sysfs_unlink_rdev(struct mddev *mddev, struct md_rdev *rdev)
Namhyung Kim36fad852011-07-27 11:00:36 +1000518{
519 char nm[20];
NeilBrown2d78f8c2011-12-23 10:17:51 +1100520 if (!test_bit(Replacement, &rdev->flags)) {
521 sprintf(nm, "rd%d", rdev->raid_disk);
522 sysfs_remove_link(&mddev->kobj, nm);
523 }
Namhyung Kim36fad852011-07-27 11:00:36 +1000524}
525
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526/*
527 * iterates through some rdev ringlist. It's safe to remove the
528 * current 'rdev'. Dont touch 'tmp' though.
529 */
Cheng Renquan159ec1f2009-01-09 08:31:08 +1100530#define rdev_for_each_list(rdev, tmp, head) \
531 list_for_each_entry_safe(rdev, tmp, head, same_set)
532
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533/*
534 * iterates through the 'same array disks' ringlist
535 */
NeilBrowndafb20f2012-03-19 12:46:39 +1100536#define rdev_for_each(rdev, mddev) \
537 list_for_each_entry(rdev, &((mddev)->disks), same_set)
538
539#define rdev_for_each_safe(rdev, tmp, mddev) \
Cheng Renquan159ec1f2009-01-09 08:31:08 +1100540 list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541
NeilBrown4b809912008-07-21 17:05:25 +1000542#define rdev_for_each_rcu(rdev, mddev) \
543 list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set)
544
NeilBrown2b8bf342011-10-11 16:48:23 +1100545struct md_thread {
NeilBrownfd01b882011-10-11 16:47:53 +1100546 void (*run) (struct mddev *mddev);
547 struct mddev *mddev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700548 wait_queue_head_t wqueue;
549 unsigned long flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550 struct task_struct *tsk;
NeilBrown32a76272005-06-21 17:17:14 -0700551 unsigned long timeout;
NeilBrown2b8bf342011-10-11 16:48:23 +1100552};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553
554#define THREAD_WAKEUP 0
555
556#define __wait_event_lock_irq(wq, condition, lock, cmd) \
557do { \
558 wait_queue_t __wait; \
559 init_waitqueue_entry(&__wait, current); \
560 \
561 add_wait_queue(&wq, &__wait); \
562 for (;;) { \
563 set_current_state(TASK_UNINTERRUPTIBLE); \
564 if (condition) \
565 break; \
566 spin_unlock_irq(&lock); \
567 cmd; \
568 schedule(); \
569 spin_lock_irq(&lock); \
570 } \
571 current->state = TASK_RUNNING; \
572 remove_wait_queue(&wq, &__wait); \
573} while (0)
574
575#define wait_event_lock_irq(wq, condition, lock, cmd) \
576do { \
577 if (condition) \
578 break; \
579 __wait_event_lock_irq(wq, condition, lock, cmd); \
580} while (0)
581
NeilBrown1345b1d2006-01-06 00:20:40 -0800582static inline void safe_put_page(struct page *p)
583{
584 if (p) put_page(p);
585}
586
NeilBrown84fc4b52011-10-11 16:49:58 +1100587extern int register_md_personality(struct md_personality *p);
588extern int unregister_md_personality(struct md_personality *p);
NeilBrown2b8bf342011-10-11 16:48:23 +1100589extern struct md_thread *md_register_thread(
590 void (*run)(struct mddev *mddev),
591 struct mddev *mddev,
592 const char *name);
593extern void md_unregister_thread(struct md_thread **threadp);
594extern void md_wakeup_thread(struct md_thread *thread);
NeilBrownfd01b882011-10-11 16:47:53 +1100595extern void md_check_recovery(struct mddev *mddev);
596extern void md_write_start(struct mddev *mddev, struct bio *bi);
597extern void md_write_end(struct mddev *mddev);
598extern void md_done_sync(struct mddev *mddev, int blocks, int ok);
599extern void md_error(struct mddev *mddev, struct md_rdev *rdev);
NeilBrownc6563a82012-05-21 09:27:00 +1000600extern void md_finish_reshape(struct mddev *mddev);
NeilBrown92022952009-03-31 14:33:13 +1100601
NeilBrownfd01b882011-10-11 16:47:53 +1100602extern int mddev_congested(struct mddev *mddev, int bits);
603extern void md_flush_request(struct mddev *mddev, struct bio *bio);
604extern void md_super_write(struct mddev *mddev, struct md_rdev *rdev,
NeilBrown92022952009-03-31 14:33:13 +1100605 sector_t sector, int size, struct page *page);
NeilBrownfd01b882011-10-11 16:47:53 +1100606extern void md_super_wait(struct mddev *mddev);
NeilBrown3cb03002011-10-11 16:45:26 +1100607extern int sync_page_io(struct md_rdev *rdev, sector_t sector, int size,
Jonathan Brassowccebd4c2011-01-14 09:14:33 +1100608 struct page *page, int rw, bool metadata_op);
NeilBrownfd01b882011-10-11 16:47:53 +1100609extern void md_do_sync(struct mddev *mddev);
610extern void md_new_event(struct mddev *mddev);
611extern int md_allow_write(struct mddev *mddev);
612extern void md_wait_for_blocked_rdev(struct md_rdev *rdev, struct mddev *mddev);
613extern void md_set_array_sectors(struct mddev *mddev, sector_t array_sectors);
614extern int md_check_no_bitmap(struct mddev *mddev);
615extern int md_integrity_register(struct mddev *mddev);
616extern void md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev);
NeilBrown72e02072009-12-14 12:49:55 +1100617extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale);
NeilBrown43a70502009-12-14 12:49:55 +1100618extern void restore_bitmap_write_access(struct file *file);
Christoph Hellwig63fe0812009-04-14 12:01:53 +1000619
NeilBrownfd01b882011-10-11 16:47:53 +1100620extern void mddev_init(struct mddev *mddev);
621extern int md_run(struct mddev *mddev);
622extern void md_stop(struct mddev *mddev);
623extern void md_stop_writes(struct mddev *mddev);
NeilBrown3cb03002011-10-11 16:45:26 +1100624extern int md_rdev_init(struct md_rdev *rdev);
NeilBrown545c8792012-05-22 13:54:30 +1000625extern void md_rdev_clear(struct md_rdev *rdev);
NeilBrown390ee602010-06-01 19:37:27 +1000626
NeilBrownfd01b882011-10-11 16:47:53 +1100627extern void mddev_suspend(struct mddev *mddev);
628extern void mddev_resume(struct mddev *mddev);
NeilBrowna167f662010-10-26 18:31:13 +1100629extern struct bio *bio_clone_mddev(struct bio *bio, gfp_t gfp_mask,
NeilBrownfd01b882011-10-11 16:47:53 +1100630 struct mddev *mddev);
NeilBrowna167f662010-10-26 18:31:13 +1100631extern struct bio *bio_alloc_mddev(gfp_t gfp_mask, int nr_iovecs,
NeilBrownfd01b882011-10-11 16:47:53 +1100632 struct mddev *mddev);
633extern int mddev_check_plugged(struct mddev *mddev);
NeilBrownd2eb35a2011-07-28 11:31:48 +1000634extern void md_trim_bio(struct bio *bio, int offset, int size);
Christoph Hellwig63fe0812009-04-14 12:01:53 +1000635#endif /* _MD_MD_H */