blob: 2b2041773e7904cf94e06ba5e141c8b5581863c3 [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>
Tejun Heo66114ca2015-05-22 17:13:32 -040019#include <linux/backing-dev.h>
Vishal Vermafc974ee2015-12-24 19:20:34 -070020#include <linux/badblocks.h>
Christoph Hellwig63fe0812009-04-14 12:01:53 +100021#include <linux/kobject.h>
22#include <linux/list.h>
23#include <linux/mm.h>
24#include <linux/mutex.h>
25#include <linux/timer.h>
26#include <linux/wait.h>
27#include <linux/workqueue.h>
Goldwyn Rodriguesedb39c92014-03-29 10:01:53 -050028#include "md-cluster.h"
David Howells93614012006-09-30 20:45:40 +020029
Linus Torvalds1da177e2005-04-16 15:20:36 -070030#define MaxSector (~(sector_t)0)
Linus Torvalds1da177e2005-04-16 15:20:36 -070031
Linus Torvalds1da177e2005-04-16 15:20:36 -070032/*
Linus Torvalds1da177e2005-04-16 15:20:36 -070033 * MD's 'extended' device
34 */
NeilBrown3cb03002011-10-11 16:45:26 +110035struct md_rdev {
Linus Torvalds1da177e2005-04-16 15:20:36 -070036 struct list_head same_set; /* RAID devices within the same set */
37
Andre Nolldd8ac332009-03-31 14:33:13 +110038 sector_t sectors; /* Device size (in 512bytes sectors) */
NeilBrownfd01b882011-10-11 16:47:53 +110039 struct mddev *mddev; /* RAID array if running */
NeilBrowneea1bf32009-03-31 14:27:02 +110040 int last_events; /* IO event timestamp */
Linus Torvalds1da177e2005-04-16 15:20:36 -070041
Jonathan Brassowa6ff7e02011-01-14 09:14:34 +110042 /*
43 * If meta_bdev is non-NULL, it means that a separate device is
44 * being used to store the metadata (superblock/bitmap) which
45 * would otherwise be contained on the same device as the data (bdev).
46 */
47 struct block_device *meta_bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -070048 struct block_device *bdev; /* block device handle */
49
NeilBrown2699b672011-07-28 11:31:47 +100050 struct page *sb_page, *bb_page;
Linus Torvalds1da177e2005-04-16 15:20:36 -070051 int sb_loaded;
NeilBrown42543762006-06-26 00:27:57 -070052 __u64 sb_events;
Linus Torvalds1da177e2005-04-16 15:20:36 -070053 sector_t data_offset; /* start of data in array */
NeilBrownc6563a82012-05-21 09:27:00 +100054 sector_t new_data_offset;/* only relevant while reshaping */
NeilBrownf72ffdd2014-09-30 14:23:59 +100055 sector_t sb_start; /* offset of the super block (in 512byte sectors) */
NeilBrown0002b272005-09-09 16:23:53 -070056 int sb_size; /* bytes in the superblock */
Linus Torvalds1da177e2005-04-16 15:20:36 -070057 int preferred_minor; /* autorun support */
58
NeilBrown86e6ffd2005-11-08 21:39:24 -080059 struct kobject kobj;
60
Linus Torvalds1da177e2005-04-16 15:20:36 -070061 /* A device can be in one of three states based on two flags:
62 * Not working: faulty==1 in_sync==0
63 * Fully working: faulty==0 in_sync==1
64 * Working, but not
65 * in sync with array
66 * faulty==0 in_sync==0
67 *
68 * It can never have faulty==1, in_sync==1
69 * This reduces the burden of testing multiple flags in many cases
70 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070071
NeilBrown2d78f8c2011-12-23 10:17:51 +110072 unsigned long flags; /* bit set of 'enum flag_bits' bits. */
Dan Williams6bfe0b42008-04-30 00:52:32 -070073 wait_queue_head_t blocked_wait;
NeilBrown8ddf9ef2005-09-09 16:23:45 -070074
Linus Torvalds1da177e2005-04-16 15:20:36 -070075 int desc_nr; /* descriptor index in the superblock */
76 int raid_disk; /* role of device in array */
NeilBrowne93f68a2010-06-15 09:36:03 +010077 int new_raid_disk; /* role that the device will have in
78 * the array after a level-change completes.
79 */
NeilBrown41158c72005-06-21 17:17:25 -070080 int saved_raid_disk; /* role that device used to have in the
81 * array and could again if we did a partial
82 * resync from the bitmap
83 */
Shaohua Li3069aa82015-08-13 14:31:56 -070084 union {
85 sector_t recovery_offset;/* If this device has been partially
NeilBrown5fd6c1d2006-06-26 00:27:40 -070086 * recovered, this is where we were
87 * up to.
88 */
Shaohua Li3069aa82015-08-13 14:31:56 -070089 sector_t journal_tail; /* If this device is a journal device,
90 * this is the journal tail (journal
91 * recovery start point)
92 */
93 };
Linus Torvalds1da177e2005-04-16 15:20:36 -070094
95 atomic_t nr_pending; /* number of pending requests.
96 * only maintained for arrays that
97 * support hot removal
98 */
NeilBrownba22dcb2005-11-08 21:39:31 -080099 atomic_t read_errors; /* number of consecutive read errors that
100 * we have tried to ignore.
101 */
Arnd Bergmann0e3ef492016-06-17 17:33:10 +0200102 time64_t last_read_error; /* monotonic time since our
Robert Becker1e509152009-12-14 12:49:58 +1100103 * last read error
104 */
NeilBrown4dbcdc72006-01-06 00:20:52 -0800105 atomic_t corrected_errors; /* number of corrected read errors,
106 * for reporting to userspace and storing
107 * in superblock.
108 */
NeilBrown5792a282007-04-04 19:08:18 -0700109 struct work_struct del_work; /* used for delayed sysfs removal */
NeilBrown3c0ee632008-10-21 13:25:28 +1100110
Tejun Heo324a56e2013-12-11 14:11:53 -0500111 struct kernfs_node *sysfs_state; /* handle for 'state'
NeilBrown3c0ee632008-10-21 13:25:28 +1100112 * sysfs entry */
NeilBrown2230dfe2011-07-28 11:31:46 +1000113
Vishal Vermafc974ee2015-12-24 19:20:34 -0700114 struct badblocks badblocks;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115};
NeilBrown2d78f8c2011-12-23 10:17:51 +1100116enum flag_bits {
117 Faulty, /* device is known to have a fault */
118 In_sync, /* device is in_sync with rest of array */
NeilBrown8313b8e2013-12-12 10:13:33 +1100119 Bitmap_sync, /* ..actually, not quite In_sync. Need a
120 * bitmap-based recovery to get fully in sync
121 */
NeilBrown2d78f8c2011-12-23 10:17:51 +1100122 WriteMostly, /* Avoid reading if at all possible */
123 AutoDetected, /* added by auto-detect */
124 Blocked, /* An error occurred but has not yet
125 * been acknowledged by the metadata
126 * handler, so don't allow writes
127 * until it is cleared */
128 WriteErrorSeen, /* A write error has been seen on this
129 * device
130 */
131 FaultRecorded, /* Intermediate state for clearing
132 * Blocked. The Fault is/will-be
133 * recorded in the metadata, but that
134 * metadata hasn't been stored safely
135 * on disk yet.
136 */
137 BlockedBadBlocks, /* A writer is blocked because they
138 * found an unacknowledged bad-block.
139 * This can safely be cleared at any
140 * time, and the writer will re-check.
141 * It may be set at any time, and at
142 * worst the writer will timeout and
143 * re-check. So setting it as
144 * accurately as possible is good, but
145 * not absolutely critical.
146 */
147 WantReplacement, /* This device is a candidate to be
148 * hot-replaced, either because it has
149 * reported some faults, or because
150 * of explicit request.
151 */
152 Replacement, /* This device is a replacement for
153 * a want_replacement device with same
154 * raid_disk number.
155 */
Goldwyn Rodrigues1aee41f2014-10-29 18:51:31 -0500156 Candidate, /* For clustered environments only:
157 * This device is seen locally but not
158 * by the whole cluster
159 */
Song Liubac624f2015-08-13 14:31:55 -0700160 Journal, /* This device is used as journal for
161 * raid-5/6.
162 * Usually, this device should be faster
163 * than other devices in the array
164 */
Guoqing Jiang659b2542015-12-21 10:50:59 +1100165 ClusterRemove,
NeilBrownd787be42016-06-02 16:19:53 +1000166 RemoveSynchronized, /* synchronize_rcu() was called after
167 * this device was known to be faulty,
168 * so it is safe to remove without
169 * another synchronize_rcu() call.
170 */
NeilBrown2d78f8c2011-12-23 10:17:51 +1100171};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172
NeilBrown3cb03002011-10-11 16:45:26 +1100173static inline int is_badblock(struct md_rdev *rdev, sector_t s, int sectors,
NeilBrown2230dfe2011-07-28 11:31:46 +1000174 sector_t *first_bad, int *bad_sectors)
175{
176 if (unlikely(rdev->badblocks.count)) {
Vishal Vermafc974ee2015-12-24 19:20:34 -0700177 int rv = badblocks_check(&rdev->badblocks, rdev->data_offset + s,
NeilBrown2230dfe2011-07-28 11:31:46 +1000178 sectors,
179 first_bad, bad_sectors);
180 if (rv)
181 *first_bad -= rdev->data_offset;
182 return rv;
183 }
184 return 0;
185}
NeilBrown3cb03002011-10-11 16:45:26 +1100186extern int rdev_set_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
NeilBrownc6563a82012-05-21 09:27:00 +1000187 int is_new);
188extern int rdev_clear_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
189 int is_new);
Goldwyn Rodriguesc4ce8672014-03-29 10:20:02 -0500190struct md_cluster_info;
191
NeilBrownfd01b882011-10-11 16:47:53 +1100192struct mddev {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193 void *private;
NeilBrown84fc4b52011-10-11 16:49:58 +1100194 struct md_personality *pers;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195 dev_t unit;
196 int md_minor;
NeilBrown7a0a5352013-08-27 16:28:23 +1000197 struct list_head disks;
NeilBrown850b2b42006-10-03 01:15:46 -0700198 unsigned long flags;
199#define MD_CHANGE_DEVS 0 /* Some device status has changed */
200#define MD_CHANGE_CLEAN 1 /* transition to or from 'clean' */
NeilBrown070dc6d2010-08-30 17:33:34 +1000201#define MD_CHANGE_PENDING 2 /* switch from 'clean' to 'active' in progress */
NeilBrown7a0a5352013-08-27 16:28:23 +1000202#define MD_UPDATE_SB_FLAGS (1 | 2 | 4) /* If these are set, md_update_sb needed */
Jonathan Brassow9c810752011-06-08 17:59:30 -0500203#define MD_ARRAY_FIRST_USE 3 /* First use of array, needs initialization */
Guoqing Jiangaf8d8e62016-08-12 13:42:37 +0800204#define MD_CLOSING 4 /* If set, we are closing the array, do not open
205 * it then */
Shaohua Libd18f642015-09-02 13:49:50 -0700206#define MD_JOURNAL_CLEAN 5 /* A raid with journal is already clean */
Song Liua97b7892015-10-08 21:54:09 -0700207#define MD_HAS_JOURNAL 6 /* The raid array has journal feature set */
Guoqing Jiang15858fa2015-12-21 10:51:00 +1100208#define MD_RELOAD_SB 7 /* Reload the superblock because another node
209 * updated it.
210 */
Guoqing Jiangbb8bf152016-06-02 23:32:04 -0400211#define MD_CLUSTER_RESYNC_LOCKED 8 /* cluster raid only, which means node
212 * already took resync lock, need to
213 * release the lock */
NeilBrown850b2b42006-10-03 01:15:46 -0700214
NeilBrown409c57f2009-03-31 14:39:39 +1100215 int suspended;
216 atomic_t active_io;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217 int ro;
NeilBrownbb4f1e92010-08-08 21:18:03 +1000218 int sysfs_active; /* set when sysfs deletes
219 * are happening, so run/
220 * takeover/stop are not safe
221 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222 struct gendisk *gendisk;
223
NeilBrowneae17012005-11-08 21:39:23 -0800224 struct kobject kobj;
NeilBrownd3374822009-01-09 08:31:10 +1100225 int hold_active;
226#define UNTIL_IOCTL 1
NeilBrownefeb53c2009-01-09 08:31:10 +1100227#define UNTIL_STOP 2
NeilBrowneae17012005-11-08 21:39:23 -0800228
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229 /* Superblock information */
230 int major_version,
231 minor_version,
232 patch_version;
233 int persistent;
NeilBrownf72ffdd2014-09-30 14:23:59 +1000234 int external; /* metadata is
NeilBrowne6910632008-02-06 01:39:51 -0800235 * managed externally */
236 char metadata_type[17]; /* externally set*/
Andre Noll9d8f0362009-06-18 08:45:01 +1000237 int chunk_sectors;
Deepa Dinamani9ebc6ef2015-12-21 10:51:01 +1100238 time64_t ctime, utime;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239 int level, layout;
NeilBrownd9d166c2006-01-06 00:20:51 -0800240 char clevel[16];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241 int raid_disks;
242 int max_disks;
NeilBrownf72ffdd2014-09-30 14:23:59 +1000243 sector_t dev_sectors; /* used size of
Andre Noll58c0fed2009-03-31 14:33:13 +1100244 * component devices */
Andre Nollf233ea52008-07-21 17:05:22 +1000245 sector_t array_sectors; /* exported array size */
Dan Williamsb522adc2009-03-31 15:00:31 +1100246 int external_size; /* size managed
247 * externally */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248 __u64 events;
NeilBrowna8707c02010-05-18 09:28:43 +1000249 /* If the last 'event' was simply a clean->dirty transition, and
250 * we didn't write it to the spares, then it is safe and simple
251 * to just decrement the event count on a dirty->clean transition.
252 * So we record that possibility here.
253 */
254 int can_decrease_events;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255
256 char uuid[16];
257
NeilBrownf6705572006-03-27 01:18:11 -0800258 /* If the array is being reshaped, we need to record the
259 * new shape and an indication of where we are up to.
260 * This is written to the superblock.
261 * If reshape_position is MaxSector, then no reshape is happening (yet).
262 */
263 sector_t reshape_position;
Andre Noll664e7c42009-06-18 08:45:27 +1000264 int delta_disks, new_level, new_layout;
265 int new_chunk_sectors;
NeilBrown2c810cd2012-05-21 09:27:00 +1000266 int reshape_backwards;
NeilBrownf6705572006-03-27 01:18:11 -0800267
NeilBrown2b8bf342011-10-11 16:48:23 +1100268 struct md_thread *thread; /* management thread */
269 struct md_thread *sync_thread; /* doing resync or reconstruct */
Jonathan Brassowc4a39552013-06-25 01:23:59 -0500270
271 /* 'last_sync_action' is initialized to "none". It is set when a
272 * sync operation (i.e "data-check", "requested-resync", "resync",
273 * "recovery", or "reshape") is started. It holds this value even
274 * when the sync thread is "frozen" (interrupted) or "idle" (stopped
275 * or finished). It is overwritten when a new sync operation is begun.
276 */
277 char *last_sync_action;
NeilBrownff4e8d92006-07-10 04:44:16 -0700278 sector_t curr_resync; /* last block scheduled */
NeilBrown97e4f422009-03-31 14:33:13 +1100279 /* As resync requests can complete out of order, we cannot easily track
280 * how much resync has been completed. So we occasionally pause until
281 * everything completes, then set curr_resync_completed to curr_resync.
282 * As such it may be well behind the real resync mark, but it is a value
283 * we are certain of.
284 */
285 sector_t curr_resync_completed;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286 unsigned long resync_mark; /* a recent timestamp */
287 sector_t resync_mark_cnt;/* blocks written at resync_mark */
NeilBrownff4e8d92006-07-10 04:44:16 -0700288 sector_t curr_mark_cnt; /* blocks scheduled now */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289
290 sector_t resync_max_sectors; /* may be set by personality */
NeilBrown9d888832005-11-08 21:39:26 -0800291
Jianpeng Ma7f7583d2012-10-11 14:17:59 +1100292 atomic64_t resync_mismatches; /* count of sectors where
NeilBrown9d888832005-11-08 21:39:26 -0800293 * parity/replica mismatch found
294 */
NeilBrowne464eaf2006-03-27 01:18:14 -0800295
296 /* allow user-space to request suspension of IO to regions of the array */
297 sector_t suspend_lo;
298 sector_t suspend_hi;
NeilBrown88202a02006-01-06 00:21:36 -0800299 /* if zero, use the system-wide default */
300 int sync_speed_min;
301 int sync_speed_max;
302
Bernd Schubert90b08712008-05-23 13:04:38 -0700303 /* resync even though the same disks are shared among md-devices */
304 int parallel_resync;
305
NeilBrown6ff8d8ec2006-01-06 00:20:15 -0800306 int ok_start_degraded;
NeilBrownf72ffdd2014-09-30 14:23:59 +1000307 /* recovery/resync flags
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308 * NEEDED: we might need to start a resync/recover
309 * RUNNING: a thread is running, or about to be started
310 * SYNC: actually doing a resync, not a recovery
Neil Brown72a23c22008-06-28 08:31:41 +1000311 * RECOVER: doing recovery, or need to try it.
NeilBrowndfc70642008-05-23 13:04:39 -0700312 * INTR: resync needs to be aborted for some reason
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313 * DONE: thread is done and is waiting to be reaped
NeilBrown24dd4692005-11-08 21:39:26 -0800314 * REQUEST: user-space has requested a sync (used with SYNC)
Anand Gadiyar411c9402009-07-07 15:24:23 +0530315 * CHECK: user-space request for check-only, no repair
NeilBrownccfcc3c2006-03-27 01:18:09 -0800316 * RESHAPE: A reshape is happening
majianpeng0a19caa2012-11-19 19:57:34 +0800317 * ERROR: sync-action interrupted because io-error
NeilBrownccfcc3c2006-03-27 01:18:09 -0800318 *
319 * If neither SYNC or RESHAPE are set, then it is a recovery.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320 */
321#define MD_RECOVERY_RUNNING 0
322#define MD_RECOVERY_SYNC 1
Neil Brown72a23c22008-06-28 08:31:41 +1000323#define MD_RECOVERY_RECOVER 2
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324#define MD_RECOVERY_INTR 3
325#define MD_RECOVERY_DONE 4
326#define MD_RECOVERY_NEEDED 5
NeilBrown24dd4692005-11-08 21:39:26 -0800327#define MD_RECOVERY_REQUESTED 6
328#define MD_RECOVERY_CHECK 7
NeilBrownccfcc3c2006-03-27 01:18:09 -0800329#define MD_RECOVERY_RESHAPE 8
NeilBrown5fd6c1d2006-06-26 00:27:40 -0700330#define MD_RECOVERY_FROZEN 9
majianpeng0a19caa2012-11-19 19:57:34 +0800331#define MD_RECOVERY_ERROR 10
NeilBrown5fd6c1d2006-06-26 00:27:40 -0700332
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333 unsigned long recovery;
NeilBrown53890422011-07-27 11:00:36 +1000334 /* If a RAID personality determines that recovery (of a particular
335 * device) will fail due to a read error on the source device, it
336 * takes a copy of this number and does not attempt recovery again
337 * until this number changes.
338 */
339 int recovery_disabled;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340
341 int in_sync; /* know to not need resync */
NeilBrownc8c00a62009-08-10 12:50:52 +1000342 /* 'open_mutex' avoids races between 'md_open' and 'do_md_stop', so
343 * that we are never stopping an array while it is open.
344 * 'reconfig_mutex' protects all other reconfiguration.
345 * These locks are separate due to conflicting interactions
346 * with bdev->bd_mutex.
347 * Lock ordering is:
348 * reconfig_mutex -> bd_mutex : e.g. do_md_run -> revalidate_disk
349 * bd_mutex -> open_mutex: e.g. __blkdev_get -> md_open
350 */
351 struct mutex open_mutex;
NeilBrowndf5b89b2006-03-27 01:18:20 -0800352 struct mutex reconfig_mutex;
NeilBrownf2ea68c2008-07-21 17:05:25 +1000353 atomic_t active; /* general refcount */
354 atomic_t openers; /* number of active opens */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355
NeilBrownf0b4f7e2011-02-24 17:26:41 +1100356 int changed; /* True if we might need to
357 * reread partition info */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358 int degraded; /* whether md should consider
359 * adding a spare
360 */
361
362 atomic_t recovery_active; /* blocks scheduled, but not written */
363 wait_queue_head_t recovery_wait;
364 sector_t recovery_cp;
Neil Brown5e96ee62008-06-28 08:31:24 +1000365 sector_t resync_min; /* user requested sync
366 * starts here */
NeilBrownc6207272008-02-06 01:39:52 -0800367 sector_t resync_max; /* resync should pause
368 * when it gets here */
NeilBrown06d91a52005-06-21 17:17:12 -0700369
Tejun Heo324a56e2013-12-11 14:11:53 -0500370 struct kernfs_node *sysfs_state; /* handle for 'array_state'
NeilBrownb62b7592008-10-21 13:25:21 +1100371 * file in sysfs.
372 */
Tejun Heo324a56e2013-12-11 14:11:53 -0500373 struct kernfs_node *sysfs_action; /* handle for 'sync_action' */
NeilBrownb62b7592008-10-21 13:25:21 +1100374
NeilBrownd3374822009-01-09 08:31:10 +1100375 struct work_struct del_work; /* used for delayed sysfs removal */
376
NeilBrown85572d72014-12-15 12:56:56 +1100377 /* "lock" protects:
378 * flush_bio transition from NULL to !NULL
379 * rdev superblocks, events
380 * clearing MD_CHANGE_*
381 * in_sync - and related safemode and MD_CHANGE changes
NeilBrown36d091f2014-12-15 12:56:58 +1100382 * pers (also protected by reconfig_mutex and pending IO).
NeilBrown978a7a42014-12-15 12:56:58 +1100383 * clearing ->bitmap
NeilBrown4af1a042014-12-15 12:57:00 +1100384 * clearing ->bitmap_info.file
NeilBrown23da4222014-12-15 12:57:01 +1100385 * changing ->resync_{min,max}
386 * setting MD_RECOVERY_RUNNING (which interacts with resync_{min,max})
NeilBrown85572d72014-12-15 12:56:56 +1100387 */
388 spinlock_t lock;
NeilBrown3d310eb2005-06-21 17:17:26 -0700389 wait_queue_head_t sb_wait; /* for waiting on superblock updates */
NeilBrown7bfa19f2005-06-21 17:17:28 -0700390 atomic_t pending_writes; /* number of active superblock writes */
NeilBrown06d91a52005-06-21 17:17:12 -0700391
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392 unsigned int safemode; /* if set, update "clean" superblock
393 * when no writes pending.
NeilBrownf72ffdd2014-09-30 14:23:59 +1000394 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395 unsigned int safemode_delay;
396 struct timer_list safemode_timer;
NeilBrownf72ffdd2014-09-30 14:23:59 +1000397 atomic_t writes_pending;
Jens Axboe165125e2007-07-24 09:28:11 +0200398 struct request_queue *queue; /* for plugging ... */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399
NeilBrownf72ffdd2014-09-30 14:23:59 +1000400 struct bitmap *bitmap; /* the bitmap for the device */
NeilBrownc3d97142009-12-14 12:49:52 +1100401 struct {
402 struct file *file; /* the bitmap file */
NeilBrownf6af9492009-12-14 12:49:54 +1100403 loff_t offset; /* offset from superblock of
NeilBrownc3d97142009-12-14 12:49:52 +1100404 * start of bitmap. May be
405 * negative, but not '0'
NeilBrownf6af9492009-12-14 12:49:54 +1100406 * For external metadata, offset
NeilBrownf72ffdd2014-09-30 14:23:59 +1000407 * from start of device.
NeilBrownc3d97142009-12-14 12:49:52 +1100408 */
NeilBrown6409bb02012-05-22 13:55:07 +1000409 unsigned long space; /* space available at this offset */
NeilBrownf6af9492009-12-14 12:49:54 +1100410 loff_t default_offset; /* this is the offset to use when
NeilBrownc3d97142009-12-14 12:49:52 +1100411 * hot-adding a bitmap. It should
412 * eventually be settable by sysfs.
413 */
NeilBrown6409bb02012-05-22 13:55:07 +1000414 unsigned long default_space; /* space available at
415 * default offset */
NeilBrownc3d97142009-12-14 12:49:52 +1100416 struct mutex mutex;
NeilBrown42a04b52009-12-14 12:49:53 +1100417 unsigned long chunksize;
NeilBrownac2f40b2010-06-01 19:37:31 +1000418 unsigned long daemon_sleep; /* how many jiffies between updates? */
NeilBrown42a04b52009-12-14 12:49:53 +1100419 unsigned long max_write_behind; /* write-behind mode */
NeilBrownece5cff2009-12-14 12:49:56 +1100420 int external;
Goldwyn Rodriguesc4ce8672014-03-29 10:20:02 -0500421 int nodes; /* Maximum number of nodes in the cluster */
Goldwyn Rodriguescf921cc2014-03-30 00:42:49 -0500422 char cluster_name[64]; /* Name of the cluster */
NeilBrownc3d97142009-12-14 12:49:52 +1100423 } bitmap_info;
NeilBrown32a76272005-06-21 17:17:14 -0700424
NeilBrownf72ffdd2014-09-30 14:23:59 +1000425 atomic_t max_corr_read_errors; /* max read retries */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426 struct list_head all_mddevs;
NeilBrowna2826aa2009-12-14 12:49:49 +1100427
NeilBrowna64c8762010-04-14 17:15:37 +1000428 struct attribute_group *to_remove;
NeilBrown252ac522010-06-01 19:37:29 +1000429
NeilBrowna167f662010-10-26 18:31:13 +1100430 struct bio_set *bio_set;
431
Tejun Heoe9c74692010-09-03 11:56:18 +0200432 /* Generic flush handling.
433 * The last to finish preflush schedules a worker to submit
Mike Christie28a8f0d2016-06-05 14:32:25 -0500434 * the rest of the request (without the REQ_PREFLUSH flag).
NeilBrowna2826aa2009-12-14 12:49:49 +1100435 */
Tejun Heoe9c74692010-09-03 11:56:18 +0200436 struct bio *flush_bio;
NeilBrowna2826aa2009-12-14 12:49:49 +1100437 atomic_t flush_pending;
Tejun Heoe9c74692010-09-03 11:56:18 +0200438 struct work_struct flush_work;
NeilBrown768a4182010-07-26 11:49:55 +1000439 struct work_struct event_work; /* used by dm to report failure event */
NeilBrownfd01b882011-10-11 16:47:53 +1100440 void (*sync_super)(struct mddev *mddev, struct md_rdev *rdev);
Goldwyn Rodriguesc4ce8672014-03-29 10:20:02 -0500441 struct md_cluster_info *cluster_info;
Guoqing Jiang15858fa2015-12-21 10:51:00 +1100442 unsigned int good_device_nr; /* good device num within cluster raid */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443};
444
NeilBrown5c47daf2014-12-15 12:57:01 +1100445static inline int __must_check mddev_lock(struct mddev *mddev)
446{
447 return mutex_lock_interruptible(&mddev->reconfig_mutex);
448}
449
450/* Sometimes we need to take the lock in a situation where
451 * failure due to interrupts is not acceptable.
452 */
453static inline void mddev_lock_nointr(struct mddev *mddev)
454{
455 mutex_lock(&mddev->reconfig_mutex);
456}
457
458static inline int mddev_is_locked(struct mddev *mddev)
459{
460 return mutex_is_locked(&mddev->reconfig_mutex);
461}
462
463static inline int mddev_trylock(struct mddev *mddev)
464{
465 return mutex_trylock(&mddev->reconfig_mutex);
466}
467extern void mddev_unlock(struct mddev *mddev);
468
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors)
470{
NeilBrownf72ffdd2014-09-30 14:23:59 +1000471 atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472}
473
NeilBrown84fc4b52011-10-11 16:49:58 +1100474struct md_personality
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475{
476 char *name;
NeilBrown2604b702006-01-06 00:20:36 -0800477 int level;
478 struct list_head list;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479 struct module *owner;
Linus Torvaldsb4fdcb02011-11-04 17:06:58 -0700480 void (*make_request)(struct mddev *mddev, struct bio *bio);
NeilBrownfd01b882011-10-11 16:47:53 +1100481 int (*run)(struct mddev *mddev);
NeilBrownafa0f552014-12-15 12:56:58 +1100482 void (*free)(struct mddev *mddev, void *priv);
NeilBrownfd01b882011-10-11 16:47:53 +1100483 void (*status)(struct seq_file *seq, struct mddev *mddev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484 /* error_handler must set ->faulty and clear ->in_sync
NeilBrownf72ffdd2014-09-30 14:23:59 +1000485 * if appropriate, and should abort recovery if needed
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486 */
NeilBrownfd01b882011-10-11 16:47:53 +1100487 void (*error_handler)(struct mddev *mddev, struct md_rdev *rdev);
488 int (*hot_add_disk) (struct mddev *mddev, struct md_rdev *rdev);
NeilBrownb8321b62011-12-23 10:17:51 +1100489 int (*hot_remove_disk) (struct mddev *mddev, struct md_rdev *rdev);
NeilBrownfd01b882011-10-11 16:47:53 +1100490 int (*spare_active) (struct mddev *mddev);
NeilBrown09314792015-02-19 16:04:40 +1100491 sector_t (*sync_request)(struct mddev *mddev, sector_t sector_nr, int *skipped);
NeilBrownfd01b882011-10-11 16:47:53 +1100492 int (*resize) (struct mddev *mddev, sector_t sectors);
493 sector_t (*size) (struct mddev *mddev, sector_t sectors, int raid_disks);
494 int (*check_reshape) (struct mddev *mddev);
495 int (*start_reshape) (struct mddev *mddev);
496 void (*finish_reshape) (struct mddev *mddev);
NeilBrown36fa3062005-09-09 16:23:45 -0700497 /* quiesce moves between quiescence states
498 * 0 - fully active
499 * 1 - no new requests allowed
500 * others - reserved
501 */
NeilBrownfd01b882011-10-11 16:47:53 +1100502 void (*quiesce) (struct mddev *mddev, int state);
NeilBrown245f46c2009-03-31 14:39:39 +1100503 /* takeover is used to transition an array from one
504 * personality to another. The new personality must be able
505 * to handle the data in the current layout.
506 * e.g. 2drive raid1 -> 2drive raid5
507 * ndrive raid5 -> degraded n+1drive raid6 with special layout
508 * If the takeover succeeds, a new 'private' structure is returned.
509 * This needs to be installed and then ->run used to activate the
510 * array.
511 */
NeilBrownfd01b882011-10-11 16:47:53 +1100512 void *(*takeover) (struct mddev *mddev);
NeilBrown5c675f82014-12-15 12:56:56 +1100513 /* congested implements bdi.congested_fn().
514 * Will not be called while array is 'suspended' */
515 int (*congested)(struct mddev *mddev, int bits);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516};
517
NeilBrown007583c2005-11-08 21:39:30 -0800518struct md_sysfs_entry {
519 struct attribute attr;
NeilBrownfd01b882011-10-11 16:47:53 +1100520 ssize_t (*show)(struct mddev *, char *);
521 ssize_t (*store)(struct mddev *, const char *, size_t);
NeilBrown007583c2005-11-08 21:39:30 -0800522};
NeilBrown43a70502009-12-14 12:49:55 +1100523extern struct attribute_group md_bitmap_group;
NeilBrown007583c2005-11-08 21:39:30 -0800524
Tejun Heo324a56e2013-12-11 14:11:53 -0500525static inline struct kernfs_node *sysfs_get_dirent_safe(struct kernfs_node *sd, char *name)
NeilBrown00bcb4a2010-06-01 19:37:23 +1000526{
527 if (sd)
Tejun Heo388975c2013-09-11 23:19:13 -0400528 return sysfs_get_dirent(sd, name);
NeilBrown00bcb4a2010-06-01 19:37:23 +1000529 return sd;
530}
Tejun Heo324a56e2013-12-11 14:11:53 -0500531static inline void sysfs_notify_dirent_safe(struct kernfs_node *sd)
NeilBrown00bcb4a2010-06-01 19:37:23 +1000532{
533 if (sd)
534 sysfs_notify_dirent(sd);
535}
536
NeilBrownfd01b882011-10-11 16:47:53 +1100537static inline char * mdname (struct mddev * mddev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538{
539 return mddev->gendisk ? mddev->gendisk->disk_name : "mdX";
540}
541
NeilBrownfd01b882011-10-11 16:47:53 +1100542static inline int sysfs_link_rdev(struct mddev *mddev, struct md_rdev *rdev)
Namhyung Kim36fad852011-07-27 11:00:36 +1000543{
544 char nm[20];
Shaohua Li9b156032015-12-18 15:19:16 +1100545 if (!test_bit(Replacement, &rdev->flags) &&
546 !test_bit(Journal, &rdev->flags) &&
547 mddev->kobj.sd) {
NeilBrown2d78f8c2011-12-23 10:17:51 +1100548 sprintf(nm, "rd%d", rdev->raid_disk);
549 return sysfs_create_link(&mddev->kobj, &rdev->kobj, nm);
550 } else
551 return 0;
Namhyung Kim36fad852011-07-27 11:00:36 +1000552}
553
NeilBrownfd01b882011-10-11 16:47:53 +1100554static inline void sysfs_unlink_rdev(struct mddev *mddev, struct md_rdev *rdev)
Namhyung Kim36fad852011-07-27 11:00:36 +1000555{
556 char nm[20];
Shaohua Li9b156032015-12-18 15:19:16 +1100557 if (!test_bit(Replacement, &rdev->flags) &&
558 !test_bit(Journal, &rdev->flags) &&
559 mddev->kobj.sd) {
NeilBrown2d78f8c2011-12-23 10:17:51 +1100560 sprintf(nm, "rd%d", rdev->raid_disk);
561 sysfs_remove_link(&mddev->kobj, nm);
562 }
Namhyung Kim36fad852011-07-27 11:00:36 +1000563}
564
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565/*
566 * iterates through some rdev ringlist. It's safe to remove the
567 * current 'rdev'. Dont touch 'tmp' though.
568 */
Cheng Renquan159ec1f2009-01-09 08:31:08 +1100569#define rdev_for_each_list(rdev, tmp, head) \
570 list_for_each_entry_safe(rdev, tmp, head, same_set)
571
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572/*
573 * iterates through the 'same array disks' ringlist
574 */
NeilBrowndafb20f2012-03-19 12:46:39 +1100575#define rdev_for_each(rdev, mddev) \
576 list_for_each_entry(rdev, &((mddev)->disks), same_set)
577
578#define rdev_for_each_safe(rdev, tmp, mddev) \
Cheng Renquan159ec1f2009-01-09 08:31:08 +1100579 list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580
NeilBrown4b809912008-07-21 17:05:25 +1000581#define rdev_for_each_rcu(rdev, mddev) \
582 list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set)
583
NeilBrown2b8bf342011-10-11 16:48:23 +1100584struct md_thread {
Shaohua Li4ed87312012-10-11 13:34:00 +1100585 void (*run) (struct md_thread *thread);
NeilBrownfd01b882011-10-11 16:47:53 +1100586 struct mddev *mddev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587 wait_queue_head_t wqueue;
NeilBrownf72ffdd2014-09-30 14:23:59 +1000588 unsigned long flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589 struct task_struct *tsk;
NeilBrown32a76272005-06-21 17:17:14 -0700590 unsigned long timeout;
Shaohua Li4ed87312012-10-11 13:34:00 +1100591 void *private;
NeilBrown2b8bf342011-10-11 16:48:23 +1100592};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593
594#define THREAD_WAKEUP 0
595
NeilBrown1345b1d2006-01-06 00:20:40 -0800596static inline void safe_put_page(struct page *p)
597{
598 if (p) put_page(p);
599}
600
NeilBrown84fc4b52011-10-11 16:49:58 +1100601extern int register_md_personality(struct md_personality *p);
602extern int unregister_md_personality(struct md_personality *p);
Goldwyn Rodriguesedb39c92014-03-29 10:01:53 -0500603extern int register_md_cluster_operations(struct md_cluster_operations *ops,
604 struct module *module);
605extern int unregister_md_cluster_operations(void);
606extern int md_setup_cluster(struct mddev *mddev, int nodes);
607extern void md_cluster_stop(struct mddev *mddev);
NeilBrown2b8bf342011-10-11 16:48:23 +1100608extern struct md_thread *md_register_thread(
Shaohua Li4ed87312012-10-11 13:34:00 +1100609 void (*run)(struct md_thread *thread),
NeilBrown2b8bf342011-10-11 16:48:23 +1100610 struct mddev *mddev,
611 const char *name);
612extern void md_unregister_thread(struct md_thread **threadp);
613extern void md_wakeup_thread(struct md_thread *thread);
NeilBrownfd01b882011-10-11 16:47:53 +1100614extern void md_check_recovery(struct mddev *mddev);
Jonathan Brassowa91d5ac2013-04-24 11:42:43 +1000615extern void md_reap_sync_thread(struct mddev *mddev);
NeilBrownfd01b882011-10-11 16:47:53 +1100616extern void md_write_start(struct mddev *mddev, struct bio *bi);
617extern void md_write_end(struct mddev *mddev);
618extern void md_done_sync(struct mddev *mddev, int blocks, int ok);
619extern void md_error(struct mddev *mddev, struct md_rdev *rdev);
NeilBrownc6563a82012-05-21 09:27:00 +1000620extern void md_finish_reshape(struct mddev *mddev);
NeilBrown92022952009-03-31 14:33:13 +1100621
NeilBrownfd01b882011-10-11 16:47:53 +1100622extern int mddev_congested(struct mddev *mddev, int bits);
623extern void md_flush_request(struct mddev *mddev, struct bio *bio);
624extern void md_super_write(struct mddev *mddev, struct md_rdev *rdev,
NeilBrown92022952009-03-31 14:33:13 +1100625 sector_t sector, int size, struct page *page);
NeilBrownfd01b882011-10-11 16:47:53 +1100626extern void md_super_wait(struct mddev *mddev);
NeilBrownf72ffdd2014-09-30 14:23:59 +1000627extern int sync_page_io(struct md_rdev *rdev, sector_t sector, int size,
Mike Christie796a5cf2016-06-05 14:32:07 -0500628 struct page *page, int op, int op_flags,
629 bool metadata_op);
Shaohua Li4ed87312012-10-11 13:34:00 +1100630extern void md_do_sync(struct md_thread *thread);
NeilBrownfd01b882011-10-11 16:47:53 +1100631extern void md_new_event(struct mddev *mddev);
632extern int md_allow_write(struct mddev *mddev);
633extern void md_wait_for_blocked_rdev(struct md_rdev *rdev, struct mddev *mddev);
634extern void md_set_array_sectors(struct mddev *mddev, sector_t array_sectors);
635extern int md_check_no_bitmap(struct mddev *mddev);
636extern int md_integrity_register(struct mddev *mddev);
Dan Williams1501efa2016-01-13 16:00:07 -0800637extern int md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev);
NeilBrown72e02072009-12-14 12:49:55 +1100638extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale);
Christoph Hellwig63fe0812009-04-14 12:01:53 +1000639
NeilBrownfd01b882011-10-11 16:47:53 +1100640extern void mddev_init(struct mddev *mddev);
641extern int md_run(struct mddev *mddev);
642extern void md_stop(struct mddev *mddev);
643extern void md_stop_writes(struct mddev *mddev);
NeilBrown3cb03002011-10-11 16:45:26 +1100644extern int md_rdev_init(struct md_rdev *rdev);
NeilBrown545c8792012-05-22 13:54:30 +1000645extern void md_rdev_clear(struct md_rdev *rdev);
NeilBrown390ee602010-06-01 19:37:27 +1000646
NeilBrownfd01b882011-10-11 16:47:53 +1100647extern void mddev_suspend(struct mddev *mddev);
648extern void mddev_resume(struct mddev *mddev);
NeilBrowna167f662010-10-26 18:31:13 +1100649extern struct bio *bio_clone_mddev(struct bio *bio, gfp_t gfp_mask,
NeilBrownfd01b882011-10-11 16:47:53 +1100650 struct mddev *mddev);
NeilBrowna167f662010-10-26 18:31:13 +1100651extern struct bio *bio_alloc_mddev(gfp_t gfp_mask, int nr_iovecs,
NeilBrownfd01b882011-10-11 16:47:53 +1100652 struct mddev *mddev);
NeilBrown9cbb1752012-07-31 09:08:14 +0200653
NeilBrown74018dc2012-07-31 09:08:15 +0200654extern void md_unplug(struct blk_plug_cb *cb, bool from_schedule);
Goldwyn Rodrigues70bcecd2015-08-21 10:33:39 -0500655extern void md_reload_sb(struct mddev *mddev, int raid_disk);
Goldwyn Rodrigues1aee41f2014-10-29 18:51:31 -0500656extern void md_update_sb(struct mddev *mddev, int force);
Goldwyn Rodriguesfb56dfe2015-04-14 10:43:24 -0500657extern void md_kick_rdev_from_array(struct md_rdev * rdev);
Goldwyn Rodrigues57d051d2015-04-14 10:43:55 -0500658struct md_rdev *md_find_rdev_nr_rcu(struct mddev *mddev, int nr);
NeilBrown9cbb1752012-07-31 09:08:14 +0200659static inline int mddev_check_plugged(struct mddev *mddev)
660{
661 return !!blk_check_plugged(md_unplug, mddev,
662 sizeof(struct blk_plug_cb));
663}
Hannes Reineckedfe15ac2012-07-26 11:12:18 +0200664
665static inline void rdev_dec_pending(struct md_rdev *rdev, struct mddev *mddev)
666{
667 int faulty = test_bit(Faulty, &rdev->flags);
668 if (atomic_dec_and_test(&rdev->nr_pending) && faulty) {
669 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
670 md_wakeup_thread(mddev->thread);
671 }
672}
673
Goldwyn Rodriguesedb39c92014-03-29 10:01:53 -0500674extern struct md_cluster_operations *md_cluster_ops;
Goldwyn Rodriguesc4ce8672014-03-29 10:20:02 -0500675static inline int mddev_is_clustered(struct mddev *mddev)
676{
677 return mddev->cluster_info && mddev->bitmap_info.nodes > 1;
678}
Christoph Hellwig63fe0812009-04-14 12:01:53 +1000679#endif /* _MD_MD_H */