blob: 318ca8fd430f1f9cd422d84cde77ac6a1fff435a [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
NeilBrownf72ffdd2014-09-30 14:23:59 +10004
Linus Torvalds1da177e2005-04-16 15:20:36 -07005 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.
NeilBrownf72ffdd2014-09-30 14:23:59 +10009
Linus Torvalds1da177e2005-04-16 15:20:36 -070010 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
NeilBrownf72ffdd2014-09-30 14:23:59 +100012 Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
Linus Torvalds1da177e2005-04-16 15:20:36 -070013*/
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 */
NeilBrownf72ffdd2014-09-30 14:23:59 +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
Tejun Heo324a56e2013-12-11 14:11:53 -0500109 struct kernfs_node *sysfs_state; /* handle for 'state'
NeilBrown3c0ee632008-10-21 13:25:28 +1100110 * 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 */
NeilBrown8313b8e2013-12-12 10:13:33 +1100132 Bitmap_sync, /* ..actually, not quite In_sync. Need a
133 * bitmap-based recovery to get fully in sync
134 */
NeilBrown050b6612012-03-19 12:46:39 +1100135 Unmerged, /* device is being added to array and should
136 * be considerred for bvec_merge_fn but not
137 * yet for actual IO
138 */
NeilBrown2d78f8c2011-12-23 10:17:51 +1100139 WriteMostly, /* Avoid reading if at all possible */
140 AutoDetected, /* added by auto-detect */
141 Blocked, /* An error occurred but has not yet
142 * been acknowledged by the metadata
143 * handler, so don't allow writes
144 * until it is cleared */
145 WriteErrorSeen, /* A write error has been seen on this
146 * device
147 */
148 FaultRecorded, /* Intermediate state for clearing
149 * Blocked. The Fault is/will-be
150 * recorded in the metadata, but that
151 * metadata hasn't been stored safely
152 * on disk yet.
153 */
154 BlockedBadBlocks, /* A writer is blocked because they
155 * found an unacknowledged bad-block.
156 * This can safely be cleared at any
157 * time, and the writer will re-check.
158 * It may be set at any time, and at
159 * worst the writer will timeout and
160 * re-check. So setting it as
161 * accurately as possible is good, but
162 * not absolutely critical.
163 */
164 WantReplacement, /* This device is a candidate to be
165 * hot-replaced, either because it has
166 * reported some faults, or because
167 * of explicit request.
168 */
169 Replacement, /* This device is a replacement for
170 * a want_replacement device with same
171 * raid_disk number.
172 */
173};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174
NeilBrown2230dfe2011-07-28 11:31:46 +1000175#define BB_LEN_MASK (0x00000000000001FFULL)
176#define BB_OFFSET_MASK (0x7FFFFFFFFFFFFE00ULL)
177#define BB_ACK_MASK (0x8000000000000000ULL)
178#define BB_MAX_LEN 512
179#define BB_OFFSET(x) (((x) & BB_OFFSET_MASK) >> 9)
180#define BB_LEN(x) (((x) & BB_LEN_MASK) + 1)
181#define BB_ACK(x) (!!((x) & BB_ACK_MASK))
182#define BB_MAKE(a, l, ack) (((a)<<9) | ((l)-1) | ((u64)(!!(ack)) << 63))
183
184extern int md_is_badblock(struct badblocks *bb, sector_t s, int sectors,
185 sector_t *first_bad, int *bad_sectors);
NeilBrown3cb03002011-10-11 16:45:26 +1100186static inline int is_badblock(struct md_rdev *rdev, sector_t s, int sectors,
NeilBrown2230dfe2011-07-28 11:31:46 +1000187 sector_t *first_bad, int *bad_sectors)
188{
189 if (unlikely(rdev->badblocks.count)) {
190 int rv = md_is_badblock(&rdev->badblocks, rdev->data_offset + s,
191 sectors,
192 first_bad, bad_sectors);
193 if (rv)
194 *first_bad -= rdev->data_offset;
195 return rv;
196 }
197 return 0;
198}
NeilBrown3cb03002011-10-11 16:45:26 +1100199extern int rdev_set_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
NeilBrownc6563a82012-05-21 09:27:00 +1000200 int is_new);
201extern int rdev_clear_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
202 int is_new);
NeilBrown2230dfe2011-07-28 11:31:46 +1000203extern void md_ack_all_badblocks(struct badblocks *bb);
204
NeilBrownfd01b882011-10-11 16:47:53 +1100205struct mddev {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206 void *private;
NeilBrown84fc4b52011-10-11 16:49:58 +1100207 struct md_personality *pers;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208 dev_t unit;
209 int md_minor;
NeilBrown7a0a5352013-08-27 16:28:23 +1000210 struct list_head disks;
NeilBrown850b2b42006-10-03 01:15:46 -0700211 unsigned long flags;
212#define MD_CHANGE_DEVS 0 /* Some device status has changed */
213#define MD_CHANGE_CLEAN 1 /* transition to or from 'clean' */
NeilBrown070dc6d2010-08-30 17:33:34 +1000214#define MD_CHANGE_PENDING 2 /* switch from 'clean' to 'active' in progress */
NeilBrown7a0a5352013-08-27 16:28:23 +1000215#define MD_UPDATE_SB_FLAGS (1 | 2 | 4) /* If these are set, md_update_sb needed */
Jonathan Brassow9c810752011-06-08 17:59:30 -0500216#define MD_ARRAY_FIRST_USE 3 /* First use of array, needs initialization */
NeilBrown260fa032013-08-27 16:44:13 +1000217#define MD_STILL_CLOSED 4 /* If set, then array has not been opened since
218 * md_ioctl checked on it.
219 */
NeilBrown850b2b42006-10-03 01:15:46 -0700220
NeilBrown409c57f2009-03-31 14:39:39 +1100221 int suspended;
222 atomic_t active_io;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223 int ro;
NeilBrownbb4f1e92010-08-08 21:18:03 +1000224 int sysfs_active; /* set when sysfs deletes
225 * are happening, so run/
226 * takeover/stop are not safe
227 */
NeilBrown7a0a5352013-08-27 16:28:23 +1000228 int ready; /* See when safe to pass
NeilBrown0ca69882011-01-14 09:14:33 +1100229 * IO requests down */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230 struct gendisk *gendisk;
231
NeilBrowneae17012005-11-08 21:39:23 -0800232 struct kobject kobj;
NeilBrownd3374822009-01-09 08:31:10 +1100233 int hold_active;
234#define UNTIL_IOCTL 1
NeilBrownefeb53c2009-01-09 08:31:10 +1100235#define UNTIL_STOP 2
NeilBrowneae17012005-11-08 21:39:23 -0800236
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237 /* Superblock information */
238 int major_version,
239 minor_version,
240 patch_version;
241 int persistent;
NeilBrownf72ffdd2014-09-30 14:23:59 +1000242 int external; /* metadata is
NeilBrowne6910632008-02-06 01:39:51 -0800243 * managed externally */
244 char metadata_type[17]; /* externally set*/
Andre Noll9d8f0362009-06-18 08:45:01 +1000245 int chunk_sectors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246 time_t ctime, utime;
247 int level, layout;
NeilBrownd9d166c2006-01-06 00:20:51 -0800248 char clevel[16];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249 int raid_disks;
250 int max_disks;
NeilBrownf72ffdd2014-09-30 14:23:59 +1000251 sector_t dev_sectors; /* used size of
Andre Noll58c0fed2009-03-31 14:33:13 +1100252 * component devices */
Andre Nollf233ea52008-07-21 17:05:22 +1000253 sector_t array_sectors; /* exported array size */
Dan Williamsb522adc2009-03-31 15:00:31 +1100254 int external_size; /* size managed
255 * externally */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256 __u64 events;
NeilBrowna8707c02010-05-18 09:28:43 +1000257 /* If the last 'event' was simply a clean->dirty transition, and
258 * we didn't write it to the spares, then it is safe and simple
259 * to just decrement the event count on a dirty->clean transition.
260 * So we record that possibility here.
261 */
262 int can_decrease_events;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263
264 char uuid[16];
265
NeilBrownf6705572006-03-27 01:18:11 -0800266 /* If the array is being reshaped, we need to record the
267 * new shape and an indication of where we are up to.
268 * This is written to the superblock.
269 * If reshape_position is MaxSector, then no reshape is happening (yet).
270 */
271 sector_t reshape_position;
Andre Noll664e7c42009-06-18 08:45:27 +1000272 int delta_disks, new_level, new_layout;
273 int new_chunk_sectors;
NeilBrown2c810cd2012-05-21 09:27:00 +1000274 int reshape_backwards;
NeilBrownf6705572006-03-27 01:18:11 -0800275
NeilBrown2b8bf342011-10-11 16:48:23 +1100276 struct md_thread *thread; /* management thread */
277 struct md_thread *sync_thread; /* doing resync or reconstruct */
Jonathan Brassowc4a39552013-06-25 01:23:59 -0500278
279 /* 'last_sync_action' is initialized to "none". It is set when a
280 * sync operation (i.e "data-check", "requested-resync", "resync",
281 * "recovery", or "reshape") is started. It holds this value even
282 * when the sync thread is "frozen" (interrupted) or "idle" (stopped
283 * or finished). It is overwritten when a new sync operation is begun.
284 */
285 char *last_sync_action;
NeilBrownff4e8d92006-07-10 04:44:16 -0700286 sector_t curr_resync; /* last block scheduled */
NeilBrown97e4f422009-03-31 14:33:13 +1100287 /* As resync requests can complete out of order, we cannot easily track
288 * how much resync has been completed. So we occasionally pause until
289 * everything completes, then set curr_resync_completed to curr_resync.
290 * As such it may be well behind the real resync mark, but it is a value
291 * we are certain of.
292 */
293 sector_t curr_resync_completed;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 unsigned long resync_mark; /* a recent timestamp */
295 sector_t resync_mark_cnt;/* blocks written at resync_mark */
NeilBrownff4e8d92006-07-10 04:44:16 -0700296 sector_t curr_mark_cnt; /* blocks scheduled now */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297
298 sector_t resync_max_sectors; /* may be set by personality */
NeilBrown9d888832005-11-08 21:39:26 -0800299
Jianpeng Ma7f7583d2012-10-11 14:17:59 +1100300 atomic64_t resync_mismatches; /* count of sectors where
NeilBrown9d888832005-11-08 21:39:26 -0800301 * parity/replica mismatch found
302 */
NeilBrowne464eaf2006-03-27 01:18:14 -0800303
304 /* allow user-space to request suspension of IO to regions of the array */
305 sector_t suspend_lo;
306 sector_t suspend_hi;
NeilBrown88202a02006-01-06 00:21:36 -0800307 /* if zero, use the system-wide default */
308 int sync_speed_min;
309 int sync_speed_max;
310
Bernd Schubert90b08712008-05-23 13:04:38 -0700311 /* resync even though the same disks are shared among md-devices */
312 int parallel_resync;
313
NeilBrown6ff8d8ec2006-01-06 00:20:15 -0800314 int ok_start_degraded;
NeilBrownf72ffdd2014-09-30 14:23:59 +1000315 /* recovery/resync flags
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316 * NEEDED: we might need to start a resync/recover
317 * RUNNING: a thread is running, or about to be started
318 * SYNC: actually doing a resync, not a recovery
Neil Brown72a23c22008-06-28 08:31:41 +1000319 * RECOVER: doing recovery, or need to try it.
NeilBrowndfc70642008-05-23 13:04:39 -0700320 * INTR: resync needs to be aborted for some reason
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321 * DONE: thread is done and is waiting to be reaped
NeilBrown24dd4692005-11-08 21:39:26 -0800322 * REQUEST: user-space has requested a sync (used with SYNC)
Anand Gadiyar411c9402009-07-07 15:24:23 +0530323 * CHECK: user-space request for check-only, no repair
NeilBrownccfcc3c2006-03-27 01:18:09 -0800324 * RESHAPE: A reshape is happening
majianpeng0a19caa2012-11-19 19:57:34 +0800325 * ERROR: sync-action interrupted because io-error
NeilBrownccfcc3c2006-03-27 01:18:09 -0800326 *
327 * If neither SYNC or RESHAPE are set, then it is a recovery.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328 */
329#define MD_RECOVERY_RUNNING 0
330#define MD_RECOVERY_SYNC 1
Neil Brown72a23c22008-06-28 08:31:41 +1000331#define MD_RECOVERY_RECOVER 2
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332#define MD_RECOVERY_INTR 3
333#define MD_RECOVERY_DONE 4
334#define MD_RECOVERY_NEEDED 5
NeilBrown24dd4692005-11-08 21:39:26 -0800335#define MD_RECOVERY_REQUESTED 6
336#define MD_RECOVERY_CHECK 7
NeilBrownccfcc3c2006-03-27 01:18:09 -0800337#define MD_RECOVERY_RESHAPE 8
NeilBrown5fd6c1d2006-06-26 00:27:40 -0700338#define MD_RECOVERY_FROZEN 9
majianpeng0a19caa2012-11-19 19:57:34 +0800339#define MD_RECOVERY_ERROR 10
NeilBrown5fd6c1d2006-06-26 00:27:40 -0700340
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341 unsigned long recovery;
NeilBrown53890422011-07-27 11:00:36 +1000342 /* If a RAID personality determines that recovery (of a particular
343 * device) will fail due to a read error on the source device, it
344 * takes a copy of this number and does not attempt recovery again
345 * until this number changes.
346 */
347 int recovery_disabled;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348
349 int in_sync; /* know to not need resync */
NeilBrownc8c00a62009-08-10 12:50:52 +1000350 /* 'open_mutex' avoids races between 'md_open' and 'do_md_stop', so
351 * that we are never stopping an array while it is open.
352 * 'reconfig_mutex' protects all other reconfiguration.
353 * These locks are separate due to conflicting interactions
354 * with bdev->bd_mutex.
355 * Lock ordering is:
356 * reconfig_mutex -> bd_mutex : e.g. do_md_run -> revalidate_disk
357 * bd_mutex -> open_mutex: e.g. __blkdev_get -> md_open
358 */
359 struct mutex open_mutex;
NeilBrowndf5b89b2006-03-27 01:18:20 -0800360 struct mutex reconfig_mutex;
NeilBrownf2ea68c2008-07-21 17:05:25 +1000361 atomic_t active; /* general refcount */
362 atomic_t openers; /* number of active opens */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363
NeilBrownf0b4f7e2011-02-24 17:26:41 +1100364 int changed; /* True if we might need to
365 * reread partition info */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366 int degraded; /* whether md should consider
367 * adding a spare
368 */
NeilBrown050b6612012-03-19 12:46:39 +1100369 int merge_check_needed; /* at least one
370 * member device
371 * has a
372 * merge_bvec_fn */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373
374 atomic_t recovery_active; /* blocks scheduled, but not written */
375 wait_queue_head_t recovery_wait;
376 sector_t recovery_cp;
Neil Brown5e96ee62008-06-28 08:31:24 +1000377 sector_t resync_min; /* user requested sync
378 * starts here */
NeilBrownc6207272008-02-06 01:39:52 -0800379 sector_t resync_max; /* resync should pause
380 * when it gets here */
NeilBrown06d91a52005-06-21 17:17:12 -0700381
Tejun Heo324a56e2013-12-11 14:11:53 -0500382 struct kernfs_node *sysfs_state; /* handle for 'array_state'
NeilBrownb62b7592008-10-21 13:25:21 +1100383 * file in sysfs.
384 */
Tejun Heo324a56e2013-12-11 14:11:53 -0500385 struct kernfs_node *sysfs_action; /* handle for 'sync_action' */
NeilBrownb62b7592008-10-21 13:25:21 +1100386
NeilBrownd3374822009-01-09 08:31:10 +1100387 struct work_struct del_work; /* used for delayed sysfs removal */
388
NeilBrown85572d72014-12-15 12:56:56 +1100389 /* "lock" protects:
390 * flush_bio transition from NULL to !NULL
391 * rdev superblocks, events
392 * clearing MD_CHANGE_*
393 * in_sync - and related safemode and MD_CHANGE changes
NeilBrown36d091f2014-12-15 12:56:58 +1100394 * pers (also protected by reconfig_mutex and pending IO).
NeilBrown978a7a42014-12-15 12:56:58 +1100395 * clearing ->bitmap
NeilBrown4af1a042014-12-15 12:57:00 +1100396 * clearing ->bitmap_info.file
NeilBrown23da4222014-12-15 12:57:01 +1100397 * changing ->resync_{min,max}
398 * setting MD_RECOVERY_RUNNING (which interacts with resync_{min,max})
NeilBrown85572d72014-12-15 12:56:56 +1100399 */
400 spinlock_t lock;
NeilBrown3d310eb2005-06-21 17:17:26 -0700401 wait_queue_head_t sb_wait; /* for waiting on superblock updates */
NeilBrown7bfa19f2005-06-21 17:17:28 -0700402 atomic_t pending_writes; /* number of active superblock writes */
NeilBrown06d91a52005-06-21 17:17:12 -0700403
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404 unsigned int safemode; /* if set, update "clean" superblock
405 * when no writes pending.
NeilBrownf72ffdd2014-09-30 14:23:59 +1000406 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407 unsigned int safemode_delay;
408 struct timer_list safemode_timer;
NeilBrownf72ffdd2014-09-30 14:23:59 +1000409 atomic_t writes_pending;
Jens Axboe165125e2007-07-24 09:28:11 +0200410 struct request_queue *queue; /* for plugging ... */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411
NeilBrownf72ffdd2014-09-30 14:23:59 +1000412 struct bitmap *bitmap; /* the bitmap for the device */
NeilBrownc3d97142009-12-14 12:49:52 +1100413 struct {
414 struct file *file; /* the bitmap file */
NeilBrownf6af9492009-12-14 12:49:54 +1100415 loff_t offset; /* offset from superblock of
NeilBrownc3d97142009-12-14 12:49:52 +1100416 * start of bitmap. May be
417 * negative, but not '0'
NeilBrownf6af9492009-12-14 12:49:54 +1100418 * For external metadata, offset
NeilBrownf72ffdd2014-09-30 14:23:59 +1000419 * from start of device.
NeilBrownc3d97142009-12-14 12:49:52 +1100420 */
NeilBrown6409bb02012-05-22 13:55:07 +1000421 unsigned long space; /* space available at this offset */
NeilBrownf6af9492009-12-14 12:49:54 +1100422 loff_t default_offset; /* this is the offset to use when
NeilBrownc3d97142009-12-14 12:49:52 +1100423 * hot-adding a bitmap. It should
424 * eventually be settable by sysfs.
425 */
NeilBrown6409bb02012-05-22 13:55:07 +1000426 unsigned long default_space; /* space available at
427 * default offset */
NeilBrownc3d97142009-12-14 12:49:52 +1100428 struct mutex mutex;
NeilBrown42a04b52009-12-14 12:49:53 +1100429 unsigned long chunksize;
NeilBrownac2f40b2010-06-01 19:37:31 +1000430 unsigned long daemon_sleep; /* how many jiffies between updates? */
NeilBrown42a04b52009-12-14 12:49:53 +1100431 unsigned long max_write_behind; /* write-behind mode */
NeilBrownece5cff2009-12-14 12:49:56 +1100432 int external;
NeilBrownc3d97142009-12-14 12:49:52 +1100433 } bitmap_info;
NeilBrown32a76272005-06-21 17:17:14 -0700434
NeilBrownf72ffdd2014-09-30 14:23:59 +1000435 atomic_t max_corr_read_errors; /* max read retries */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436 struct list_head all_mddevs;
NeilBrowna2826aa2009-12-14 12:49:49 +1100437
NeilBrowna64c8762010-04-14 17:15:37 +1000438 struct attribute_group *to_remove;
NeilBrown252ac522010-06-01 19:37:29 +1000439
NeilBrowna167f662010-10-26 18:31:13 +1100440 struct bio_set *bio_set;
441
Tejun Heoe9c74692010-09-03 11:56:18 +0200442 /* Generic flush handling.
443 * The last to finish preflush schedules a worker to submit
444 * the rest of the request (without the REQ_FLUSH flag).
NeilBrowna2826aa2009-12-14 12:49:49 +1100445 */
Tejun Heoe9c74692010-09-03 11:56:18 +0200446 struct bio *flush_bio;
NeilBrowna2826aa2009-12-14 12:49:49 +1100447 atomic_t flush_pending;
Tejun Heoe9c74692010-09-03 11:56:18 +0200448 struct work_struct flush_work;
NeilBrown768a4182010-07-26 11:49:55 +1000449 struct work_struct event_work; /* used by dm to report failure event */
NeilBrownfd01b882011-10-11 16:47:53 +1100450 void (*sync_super)(struct mddev *mddev, struct md_rdev *rdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451};
452
NeilBrown5c47daf2014-12-15 12:57:01 +1100453static inline int __must_check mddev_lock(struct mddev *mddev)
454{
455 return mutex_lock_interruptible(&mddev->reconfig_mutex);
456}
457
458/* Sometimes we need to take the lock in a situation where
459 * failure due to interrupts is not acceptable.
460 */
461static inline void mddev_lock_nointr(struct mddev *mddev)
462{
463 mutex_lock(&mddev->reconfig_mutex);
464}
465
466static inline int mddev_is_locked(struct mddev *mddev)
467{
468 return mutex_is_locked(&mddev->reconfig_mutex);
469}
470
471static inline int mddev_trylock(struct mddev *mddev)
472{
473 return mutex_trylock(&mddev->reconfig_mutex);
474}
475extern void mddev_unlock(struct mddev *mddev);
476
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors)
478{
NeilBrownf72ffdd2014-09-30 14:23:59 +1000479 atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480}
481
NeilBrown84fc4b52011-10-11 16:49:58 +1100482struct md_personality
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483{
484 char *name;
NeilBrown2604b702006-01-06 00:20:36 -0800485 int level;
486 struct list_head list;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487 struct module *owner;
Linus Torvaldsb4fdcb02011-11-04 17:06:58 -0700488 void (*make_request)(struct mddev *mddev, struct bio *bio);
NeilBrownfd01b882011-10-11 16:47:53 +1100489 int (*run)(struct mddev *mddev);
NeilBrownafa0f552014-12-15 12:56:58 +1100490 void (*free)(struct mddev *mddev, void *priv);
NeilBrownfd01b882011-10-11 16:47:53 +1100491 void (*status)(struct seq_file *seq, struct mddev *mddev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492 /* error_handler must set ->faulty and clear ->in_sync
NeilBrownf72ffdd2014-09-30 14:23:59 +1000493 * if appropriate, and should abort recovery if needed
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 */
NeilBrownfd01b882011-10-11 16:47:53 +1100495 void (*error_handler)(struct mddev *mddev, struct md_rdev *rdev);
496 int (*hot_add_disk) (struct mddev *mddev, struct md_rdev *rdev);
NeilBrownb8321b62011-12-23 10:17:51 +1100497 int (*hot_remove_disk) (struct mddev *mddev, struct md_rdev *rdev);
NeilBrownfd01b882011-10-11 16:47:53 +1100498 int (*spare_active) (struct mddev *mddev);
499 sector_t (*sync_request)(struct mddev *mddev, sector_t sector_nr, int *skipped, int go_faster);
500 int (*resize) (struct mddev *mddev, sector_t sectors);
501 sector_t (*size) (struct mddev *mddev, sector_t sectors, int raid_disks);
502 int (*check_reshape) (struct mddev *mddev);
503 int (*start_reshape) (struct mddev *mddev);
504 void (*finish_reshape) (struct mddev *mddev);
NeilBrown36fa3062005-09-09 16:23:45 -0700505 /* quiesce moves between quiescence states
506 * 0 - fully active
507 * 1 - no new requests allowed
508 * others - reserved
509 */
NeilBrownfd01b882011-10-11 16:47:53 +1100510 void (*quiesce) (struct mddev *mddev, int state);
NeilBrown245f46c2009-03-31 14:39:39 +1100511 /* takeover is used to transition an array from one
512 * personality to another. The new personality must be able
513 * to handle the data in the current layout.
514 * e.g. 2drive raid1 -> 2drive raid5
515 * ndrive raid5 -> degraded n+1drive raid6 with special layout
516 * If the takeover succeeds, a new 'private' structure is returned.
517 * This needs to be installed and then ->run used to activate the
518 * array.
519 */
NeilBrownfd01b882011-10-11 16:47:53 +1100520 void *(*takeover) (struct mddev *mddev);
NeilBrown5c675f82014-12-15 12:56:56 +1100521 /* congested implements bdi.congested_fn().
522 * Will not be called while array is 'suspended' */
523 int (*congested)(struct mddev *mddev, int bits);
NeilBrown64590f42014-12-15 12:56:57 +1100524 /* mergeable_bvec is use to implement ->merge_bvec_fn */
525 int (*mergeable_bvec)(struct mddev *mddev,
526 struct bvec_merge_data *bvm,
527 struct bio_vec *biovec);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528};
529
NeilBrown007583c2005-11-08 21:39:30 -0800530struct md_sysfs_entry {
531 struct attribute attr;
NeilBrownfd01b882011-10-11 16:47:53 +1100532 ssize_t (*show)(struct mddev *, char *);
533 ssize_t (*store)(struct mddev *, const char *, size_t);
NeilBrown007583c2005-11-08 21:39:30 -0800534};
NeilBrown43a70502009-12-14 12:49:55 +1100535extern struct attribute_group md_bitmap_group;
NeilBrown007583c2005-11-08 21:39:30 -0800536
Tejun Heo324a56e2013-12-11 14:11:53 -0500537static inline struct kernfs_node *sysfs_get_dirent_safe(struct kernfs_node *sd, char *name)
NeilBrown00bcb4a2010-06-01 19:37:23 +1000538{
539 if (sd)
Tejun Heo388975c2013-09-11 23:19:13 -0400540 return sysfs_get_dirent(sd, name);
NeilBrown00bcb4a2010-06-01 19:37:23 +1000541 return sd;
542}
Tejun Heo324a56e2013-12-11 14:11:53 -0500543static inline void sysfs_notify_dirent_safe(struct kernfs_node *sd)
NeilBrown00bcb4a2010-06-01 19:37:23 +1000544{
545 if (sd)
546 sysfs_notify_dirent(sd);
547}
548
NeilBrownfd01b882011-10-11 16:47:53 +1100549static inline char * mdname (struct mddev * mddev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550{
551 return mddev->gendisk ? mddev->gendisk->disk_name : "mdX";
552}
553
NeilBrownfd01b882011-10-11 16:47:53 +1100554static inline int sysfs_link_rdev(struct mddev *mddev, struct md_rdev *rdev)
Namhyung Kim36fad852011-07-27 11:00:36 +1000555{
556 char nm[20];
Jonathan Brassow90584fc2013-03-07 16:24:26 -0600557 if (!test_bit(Replacement, &rdev->flags) && mddev->kobj.sd) {
NeilBrown2d78f8c2011-12-23 10:17:51 +1100558 sprintf(nm, "rd%d", rdev->raid_disk);
559 return sysfs_create_link(&mddev->kobj, &rdev->kobj, nm);
560 } else
561 return 0;
Namhyung Kim36fad852011-07-27 11:00:36 +1000562}
563
NeilBrownfd01b882011-10-11 16:47:53 +1100564static inline void sysfs_unlink_rdev(struct mddev *mddev, struct md_rdev *rdev)
Namhyung Kim36fad852011-07-27 11:00:36 +1000565{
566 char nm[20];
Jonathan Brassow90584fc2013-03-07 16:24:26 -0600567 if (!test_bit(Replacement, &rdev->flags) && mddev->kobj.sd) {
NeilBrown2d78f8c2011-12-23 10:17:51 +1100568 sprintf(nm, "rd%d", rdev->raid_disk);
569 sysfs_remove_link(&mddev->kobj, nm);
570 }
Namhyung Kim36fad852011-07-27 11:00:36 +1000571}
572
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573/*
574 * iterates through some rdev ringlist. It's safe to remove the
575 * current 'rdev'. Dont touch 'tmp' though.
576 */
Cheng Renquan159ec1f2009-01-09 08:31:08 +1100577#define rdev_for_each_list(rdev, tmp, head) \
578 list_for_each_entry_safe(rdev, tmp, head, same_set)
579
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580/*
581 * iterates through the 'same array disks' ringlist
582 */
NeilBrowndafb20f2012-03-19 12:46:39 +1100583#define rdev_for_each(rdev, mddev) \
584 list_for_each_entry(rdev, &((mddev)->disks), same_set)
585
586#define rdev_for_each_safe(rdev, tmp, mddev) \
Cheng Renquan159ec1f2009-01-09 08:31:08 +1100587 list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588
NeilBrown4b809912008-07-21 17:05:25 +1000589#define rdev_for_each_rcu(rdev, mddev) \
590 list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set)
591
NeilBrown2b8bf342011-10-11 16:48:23 +1100592struct md_thread {
Shaohua Li4ed87312012-10-11 13:34:00 +1100593 void (*run) (struct md_thread *thread);
NeilBrownfd01b882011-10-11 16:47:53 +1100594 struct mddev *mddev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595 wait_queue_head_t wqueue;
NeilBrownf72ffdd2014-09-30 14:23:59 +1000596 unsigned long flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597 struct task_struct *tsk;
NeilBrown32a76272005-06-21 17:17:14 -0700598 unsigned long timeout;
Shaohua Li4ed87312012-10-11 13:34:00 +1100599 void *private;
NeilBrown2b8bf342011-10-11 16:48:23 +1100600};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601
602#define THREAD_WAKEUP 0
603
NeilBrown1345b1d2006-01-06 00:20:40 -0800604static inline void safe_put_page(struct page *p)
605{
606 if (p) put_page(p);
607}
608
NeilBrown84fc4b52011-10-11 16:49:58 +1100609extern int register_md_personality(struct md_personality *p);
610extern int unregister_md_personality(struct md_personality *p);
NeilBrown2b8bf342011-10-11 16:48:23 +1100611extern struct md_thread *md_register_thread(
Shaohua Li4ed87312012-10-11 13:34:00 +1100612 void (*run)(struct md_thread *thread),
NeilBrown2b8bf342011-10-11 16:48:23 +1100613 struct mddev *mddev,
614 const char *name);
615extern void md_unregister_thread(struct md_thread **threadp);
616extern void md_wakeup_thread(struct md_thread *thread);
NeilBrownfd01b882011-10-11 16:47:53 +1100617extern void md_check_recovery(struct mddev *mddev);
Jonathan Brassowa91d5ac2013-04-24 11:42:43 +1000618extern void md_reap_sync_thread(struct mddev *mddev);
NeilBrownfd01b882011-10-11 16:47:53 +1100619extern void md_write_start(struct mddev *mddev, struct bio *bi);
620extern void md_write_end(struct mddev *mddev);
621extern void md_done_sync(struct mddev *mddev, int blocks, int ok);
622extern void md_error(struct mddev *mddev, struct md_rdev *rdev);
NeilBrownc6563a82012-05-21 09:27:00 +1000623extern void md_finish_reshape(struct mddev *mddev);
NeilBrown92022952009-03-31 14:33:13 +1100624
NeilBrownfd01b882011-10-11 16:47:53 +1100625extern int mddev_congested(struct mddev *mddev, int bits);
626extern void md_flush_request(struct mddev *mddev, struct bio *bio);
627extern void md_super_write(struct mddev *mddev, struct md_rdev *rdev,
NeilBrown92022952009-03-31 14:33:13 +1100628 sector_t sector, int size, struct page *page);
NeilBrownfd01b882011-10-11 16:47:53 +1100629extern void md_super_wait(struct mddev *mddev);
NeilBrownf72ffdd2014-09-30 14:23:59 +1000630extern int sync_page_io(struct md_rdev *rdev, sector_t sector, int size,
Jonathan Brassowccebd4c2011-01-14 09:14:33 +1100631 struct page *page, int rw, bool metadata_op);
Shaohua Li4ed87312012-10-11 13:34:00 +1100632extern void md_do_sync(struct md_thread *thread);
NeilBrownfd01b882011-10-11 16:47:53 +1100633extern void md_new_event(struct mddev *mddev);
634extern int md_allow_write(struct mddev *mddev);
635extern void md_wait_for_blocked_rdev(struct md_rdev *rdev, struct mddev *mddev);
636extern void md_set_array_sectors(struct mddev *mddev, sector_t array_sectors);
637extern int md_check_no_bitmap(struct mddev *mddev);
638extern int md_integrity_register(struct mddev *mddev);
639extern void md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev);
NeilBrown72e02072009-12-14 12:49:55 +1100640extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale);
Christoph Hellwig63fe0812009-04-14 12:01:53 +1000641
NeilBrownfd01b882011-10-11 16:47:53 +1100642extern void mddev_init(struct mddev *mddev);
643extern int md_run(struct mddev *mddev);
644extern void md_stop(struct mddev *mddev);
645extern void md_stop_writes(struct mddev *mddev);
NeilBrown3cb03002011-10-11 16:45:26 +1100646extern int md_rdev_init(struct md_rdev *rdev);
NeilBrown545c8792012-05-22 13:54:30 +1000647extern void md_rdev_clear(struct md_rdev *rdev);
NeilBrown390ee602010-06-01 19:37:27 +1000648
NeilBrownfd01b882011-10-11 16:47:53 +1100649extern void mddev_suspend(struct mddev *mddev);
650extern void mddev_resume(struct mddev *mddev);
NeilBrowna167f662010-10-26 18:31:13 +1100651extern struct bio *bio_clone_mddev(struct bio *bio, gfp_t gfp_mask,
NeilBrownfd01b882011-10-11 16:47:53 +1100652 struct mddev *mddev);
NeilBrowna167f662010-10-26 18:31:13 +1100653extern struct bio *bio_alloc_mddev(gfp_t gfp_mask, int nr_iovecs,
NeilBrownfd01b882011-10-11 16:47:53 +1100654 struct mddev *mddev);
NeilBrown9cbb1752012-07-31 09:08:14 +0200655
NeilBrown74018dc2012-07-31 09:08:15 +0200656extern void md_unplug(struct blk_plug_cb *cb, bool from_schedule);
NeilBrown9cbb1752012-07-31 09:08:14 +0200657static inline int mddev_check_plugged(struct mddev *mddev)
658{
659 return !!blk_check_plugged(md_unplug, mddev,
660 sizeof(struct blk_plug_cb));
661}
Hannes Reineckedfe15ac2012-07-26 11:12:18 +0200662
663static inline void rdev_dec_pending(struct md_rdev *rdev, struct mddev *mddev)
664{
665 int faulty = test_bit(Faulty, &rdev->flags);
666 if (atomic_dec_and_test(&rdev->nr_pending) && faulty) {
667 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
668 md_wakeup_thread(mddev->thread);
669 }
670}
671
Christoph Hellwig63fe0812009-04-14 12:01:53 +1000672#endif /* _MD_MD_H */