blob: 7da6e9c3cb53eb7d28ab44f6d40dbcea63e2d486 [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>
Christoph Hellwig63fe0812009-04-14 12:01:53 +100020#include <linux/kobject.h>
21#include <linux/list.h>
22#include <linux/mm.h>
23#include <linux/mutex.h>
24#include <linux/timer.h>
25#include <linux/wait.h>
26#include <linux/workqueue.h>
Goldwyn Rodriguesedb39c92014-03-29 10:01:53 -050027#include "md-cluster.h"
David Howells93614012006-09-30 20:45:40 +020028
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#define MaxSector (~(sector_t)0)
Linus Torvalds1da177e2005-04-16 15:20:36 -070030
NeilBrown2230dfe2011-07-28 11:31:46 +100031/* Bad block numbers are stored sorted in a single page.
32 * 64bits is used for each block or extent.
33 * 54 bits are sector number, 9 bits are extent size,
34 * 1 bit is an 'acknowledged' flag.
35 */
36#define MD_MAX_BADBLOCKS (PAGE_SIZE/8)
37
Linus Torvalds1da177e2005-04-16 15:20:36 -070038/*
Linus Torvalds1da177e2005-04-16 15:20:36 -070039 * MD's 'extended' device
40 */
NeilBrown3cb03002011-10-11 16:45:26 +110041struct md_rdev {
Linus Torvalds1da177e2005-04-16 15:20:36 -070042 struct list_head same_set; /* RAID devices within the same set */
43
Andre Nolldd8ac332009-03-31 14:33:13 +110044 sector_t sectors; /* Device size (in 512bytes sectors) */
NeilBrownfd01b882011-10-11 16:47:53 +110045 struct mddev *mddev; /* RAID array if running */
NeilBrowneea1bf32009-03-31 14:27:02 +110046 int last_events; /* IO event timestamp */
Linus Torvalds1da177e2005-04-16 15:20:36 -070047
Jonathan Brassowa6ff7e02011-01-14 09:14:34 +110048 /*
49 * If meta_bdev is non-NULL, it means that a separate device is
50 * being used to store the metadata (superblock/bitmap) which
51 * would otherwise be contained on the same device as the data (bdev).
52 */
53 struct block_device *meta_bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -070054 struct block_device *bdev; /* block device handle */
55
NeilBrown2699b672011-07-28 11:31:47 +100056 struct page *sb_page, *bb_page;
Linus Torvalds1da177e2005-04-16 15:20:36 -070057 int sb_loaded;
NeilBrown42543762006-06-26 00:27:57 -070058 __u64 sb_events;
Linus Torvalds1da177e2005-04-16 15:20:36 -070059 sector_t data_offset; /* start of data in array */
NeilBrownc6563a82012-05-21 09:27:00 +100060 sector_t new_data_offset;/* only relevant while reshaping */
NeilBrownf72ffdd2014-09-30 14:23:59 +100061 sector_t sb_start; /* offset of the super block (in 512byte sectors) */
NeilBrown0002b272005-09-09 16:23:53 -070062 int sb_size; /* bytes in the superblock */
Linus Torvalds1da177e2005-04-16 15:20:36 -070063 int preferred_minor; /* autorun support */
64
NeilBrown86e6ffd2005-11-08 21:39:24 -080065 struct kobject kobj;
66
Linus Torvalds1da177e2005-04-16 15:20:36 -070067 /* A device can be in one of three states based on two flags:
68 * Not working: faulty==1 in_sync==0
69 * Fully working: faulty==0 in_sync==1
70 * Working, but not
71 * in sync with array
72 * faulty==0 in_sync==0
73 *
74 * It can never have faulty==1, in_sync==1
75 * This reduces the burden of testing multiple flags in many cases
76 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070077
NeilBrown2d78f8c2011-12-23 10:17:51 +110078 unsigned long flags; /* bit set of 'enum flag_bits' bits. */
Dan Williams6bfe0b42008-04-30 00:52:32 -070079 wait_queue_head_t blocked_wait;
NeilBrown8ddf9ef2005-09-09 16:23:45 -070080
Linus Torvalds1da177e2005-04-16 15:20:36 -070081 int desc_nr; /* descriptor index in the superblock */
82 int raid_disk; /* role of device in array */
NeilBrowne93f68a2010-06-15 09:36:03 +010083 int new_raid_disk; /* role that the device will have in
84 * the array after a level-change completes.
85 */
NeilBrown41158c72005-06-21 17:17:25 -070086 int saved_raid_disk; /* role that device used to have in the
87 * array and could again if we did a partial
88 * resync from the bitmap
89 */
NeilBrown5fd6c1d2006-06-26 00:27:40 -070090 sector_t recovery_offset;/* If this device has been partially
91 * recovered, this is where we were
92 * up to.
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 */
Robert Becker1e509152009-12-14 12:49:58 +1100102 struct timespec last_read_error; /* monotonic time since our
103 * 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
114 struct badblocks {
115 int count; /* count of bad blocks */
NeilBrownde393cd2011-07-28 11:31:48 +1000116 int unacked_exist; /* there probably are unacknowledged
117 * bad blocks. This is only cleared
118 * when a read discovers none
119 */
NeilBrown2230dfe2011-07-28 11:31:46 +1000120 int shift; /* shift from sectors to block size
121 * a -ve shift means badblocks are
122 * disabled.*/
123 u64 *page; /* badblock list */
124 int changed;
125 seqlock_t lock;
NeilBrown2699b672011-07-28 11:31:47 +1000126
127 sector_t sector;
128 sector_t size; /* in sectors */
NeilBrown2230dfe2011-07-28 11:31:46 +1000129 } badblocks;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130};
NeilBrown2d78f8c2011-12-23 10:17:51 +1100131enum flag_bits {
132 Faulty, /* device is known to have a fault */
133 In_sync, /* device is in_sync with rest of array */
NeilBrown8313b8e2013-12-12 10:13:33 +1100134 Bitmap_sync, /* ..actually, not quite In_sync. Need a
135 * bitmap-based recovery to get fully in sync
136 */
NeilBrown050b6612012-03-19 12:46:39 +1100137 Unmerged, /* device is being added to array and should
138 * be considerred for bvec_merge_fn but not
139 * yet for actual IO
140 */
NeilBrown2d78f8c2011-12-23 10:17:51 +1100141 WriteMostly, /* Avoid reading if at all possible */
142 AutoDetected, /* added by auto-detect */
143 Blocked, /* An error occurred but has not yet
144 * been acknowledged by the metadata
145 * handler, so don't allow writes
146 * until it is cleared */
147 WriteErrorSeen, /* A write error has been seen on this
148 * device
149 */
150 FaultRecorded, /* Intermediate state for clearing
151 * Blocked. The Fault is/will-be
152 * recorded in the metadata, but that
153 * metadata hasn't been stored safely
154 * on disk yet.
155 */
156 BlockedBadBlocks, /* A writer is blocked because they
157 * found an unacknowledged bad-block.
158 * This can safely be cleared at any
159 * time, and the writer will re-check.
160 * It may be set at any time, and at
161 * worst the writer will timeout and
162 * re-check. So setting it as
163 * accurately as possible is good, but
164 * not absolutely critical.
165 */
166 WantReplacement, /* This device is a candidate to be
167 * hot-replaced, either because it has
168 * reported some faults, or because
169 * of explicit request.
170 */
171 Replacement, /* This device is a replacement for
172 * a want_replacement device with same
173 * raid_disk number.
174 */
Goldwyn Rodrigues1aee41f2014-10-29 18:51:31 -0500175 Candidate, /* For clustered environments only:
176 * This device is seen locally but not
177 * by the whole cluster
178 */
NeilBrown2d78f8c2011-12-23 10:17:51 +1100179};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700180
NeilBrown2230dfe2011-07-28 11:31:46 +1000181#define BB_LEN_MASK (0x00000000000001FFULL)
182#define BB_OFFSET_MASK (0x7FFFFFFFFFFFFE00ULL)
183#define BB_ACK_MASK (0x8000000000000000ULL)
184#define BB_MAX_LEN 512
185#define BB_OFFSET(x) (((x) & BB_OFFSET_MASK) >> 9)
186#define BB_LEN(x) (((x) & BB_LEN_MASK) + 1)
187#define BB_ACK(x) (!!((x) & BB_ACK_MASK))
188#define BB_MAKE(a, l, ack) (((a)<<9) | ((l)-1) | ((u64)(!!(ack)) << 63))
189
190extern int md_is_badblock(struct badblocks *bb, sector_t s, int sectors,
191 sector_t *first_bad, int *bad_sectors);
NeilBrown3cb03002011-10-11 16:45:26 +1100192static inline int is_badblock(struct md_rdev *rdev, sector_t s, int sectors,
NeilBrown2230dfe2011-07-28 11:31:46 +1000193 sector_t *first_bad, int *bad_sectors)
194{
195 if (unlikely(rdev->badblocks.count)) {
196 int rv = md_is_badblock(&rdev->badblocks, rdev->data_offset + s,
197 sectors,
198 first_bad, bad_sectors);
199 if (rv)
200 *first_bad -= rdev->data_offset;
201 return rv;
202 }
203 return 0;
204}
NeilBrown3cb03002011-10-11 16:45:26 +1100205extern int rdev_set_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
NeilBrownc6563a82012-05-21 09:27:00 +1000206 int is_new);
207extern int rdev_clear_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
208 int is_new);
NeilBrown2230dfe2011-07-28 11:31:46 +1000209extern void md_ack_all_badblocks(struct badblocks *bb);
210
Goldwyn Rodriguesc4ce8672014-03-29 10:20:02 -0500211struct md_cluster_info;
212
NeilBrownfd01b882011-10-11 16:47:53 +1100213struct mddev {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214 void *private;
NeilBrown84fc4b52011-10-11 16:49:58 +1100215 struct md_personality *pers;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216 dev_t unit;
217 int md_minor;
NeilBrown7a0a5352013-08-27 16:28:23 +1000218 struct list_head disks;
NeilBrown850b2b42006-10-03 01:15:46 -0700219 unsigned long flags;
220#define MD_CHANGE_DEVS 0 /* Some device status has changed */
221#define MD_CHANGE_CLEAN 1 /* transition to or from 'clean' */
NeilBrown070dc6d2010-08-30 17:33:34 +1000222#define MD_CHANGE_PENDING 2 /* switch from 'clean' to 'active' in progress */
NeilBrown7a0a5352013-08-27 16:28:23 +1000223#define MD_UPDATE_SB_FLAGS (1 | 2 | 4) /* If these are set, md_update_sb needed */
Jonathan Brassow9c810752011-06-08 17:59:30 -0500224#define MD_ARRAY_FIRST_USE 3 /* First use of array, needs initialization */
NeilBrown260fa032013-08-27 16:44:13 +1000225#define MD_STILL_CLOSED 4 /* If set, then array has not been opened since
226 * md_ioctl checked on it.
227 */
NeilBrown850b2b42006-10-03 01:15:46 -0700228
NeilBrown409c57f2009-03-31 14:39:39 +1100229 int suspended;
230 atomic_t active_io;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231 int ro;
NeilBrownbb4f1e92010-08-08 21:18:03 +1000232 int sysfs_active; /* set when sysfs deletes
233 * are happening, so run/
234 * takeover/stop are not safe
235 */
NeilBrown7a0a5352013-08-27 16:28:23 +1000236 int ready; /* See when safe to pass
NeilBrown0ca69882011-01-14 09:14:33 +1100237 * IO requests down */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238 struct gendisk *gendisk;
239
NeilBrowneae17012005-11-08 21:39:23 -0800240 struct kobject kobj;
NeilBrownd3374822009-01-09 08:31:10 +1100241 int hold_active;
242#define UNTIL_IOCTL 1
NeilBrownefeb53c2009-01-09 08:31:10 +1100243#define UNTIL_STOP 2
NeilBrowneae17012005-11-08 21:39:23 -0800244
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245 /* Superblock information */
246 int major_version,
247 minor_version,
248 patch_version;
249 int persistent;
NeilBrownf72ffdd2014-09-30 14:23:59 +1000250 int external; /* metadata is
NeilBrowne6910632008-02-06 01:39:51 -0800251 * managed externally */
252 char metadata_type[17]; /* externally set*/
Andre Noll9d8f0362009-06-18 08:45:01 +1000253 int chunk_sectors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254 time_t ctime, utime;
255 int level, layout;
NeilBrownd9d166c2006-01-06 00:20:51 -0800256 char clevel[16];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257 int raid_disks;
258 int max_disks;
NeilBrownf72ffdd2014-09-30 14:23:59 +1000259 sector_t dev_sectors; /* used size of
Andre Noll58c0fed2009-03-31 14:33:13 +1100260 * component devices */
Andre Nollf233ea52008-07-21 17:05:22 +1000261 sector_t array_sectors; /* exported array size */
Dan Williamsb522adc2009-03-31 15:00:31 +1100262 int external_size; /* size managed
263 * externally */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264 __u64 events;
NeilBrowna8707c02010-05-18 09:28:43 +1000265 /* If the last 'event' was simply a clean->dirty transition, and
266 * we didn't write it to the spares, then it is safe and simple
267 * to just decrement the event count on a dirty->clean transition.
268 * So we record that possibility here.
269 */
270 int can_decrease_events;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271
272 char uuid[16];
273
NeilBrownf6705572006-03-27 01:18:11 -0800274 /* If the array is being reshaped, we need to record the
275 * new shape and an indication of where we are up to.
276 * This is written to the superblock.
277 * If reshape_position is MaxSector, then no reshape is happening (yet).
278 */
279 sector_t reshape_position;
Andre Noll664e7c42009-06-18 08:45:27 +1000280 int delta_disks, new_level, new_layout;
281 int new_chunk_sectors;
NeilBrown2c810cd2012-05-21 09:27:00 +1000282 int reshape_backwards;
NeilBrownf6705572006-03-27 01:18:11 -0800283
NeilBrown2b8bf342011-10-11 16:48:23 +1100284 struct md_thread *thread; /* management thread */
285 struct md_thread *sync_thread; /* doing resync or reconstruct */
Jonathan Brassowc4a39552013-06-25 01:23:59 -0500286
287 /* 'last_sync_action' is initialized to "none". It is set when a
288 * sync operation (i.e "data-check", "requested-resync", "resync",
289 * "recovery", or "reshape") is started. It holds this value even
290 * when the sync thread is "frozen" (interrupted) or "idle" (stopped
291 * or finished). It is overwritten when a new sync operation is begun.
292 */
293 char *last_sync_action;
NeilBrownff4e8d92006-07-10 04:44:16 -0700294 sector_t curr_resync; /* last block scheduled */
NeilBrown97e4f422009-03-31 14:33:13 +1100295 /* As resync requests can complete out of order, we cannot easily track
296 * how much resync has been completed. So we occasionally pause until
297 * everything completes, then set curr_resync_completed to curr_resync.
298 * As such it may be well behind the real resync mark, but it is a value
299 * we are certain of.
300 */
301 sector_t curr_resync_completed;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302 unsigned long resync_mark; /* a recent timestamp */
303 sector_t resync_mark_cnt;/* blocks written at resync_mark */
NeilBrownff4e8d92006-07-10 04:44:16 -0700304 sector_t curr_mark_cnt; /* blocks scheduled now */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305
306 sector_t resync_max_sectors; /* may be set by personality */
NeilBrown9d888832005-11-08 21:39:26 -0800307
Jianpeng Ma7f7583d2012-10-11 14:17:59 +1100308 atomic64_t resync_mismatches; /* count of sectors where
NeilBrown9d888832005-11-08 21:39:26 -0800309 * parity/replica mismatch found
310 */
NeilBrowne464eaf2006-03-27 01:18:14 -0800311
312 /* allow user-space to request suspension of IO to regions of the array */
313 sector_t suspend_lo;
314 sector_t suspend_hi;
NeilBrown88202a02006-01-06 00:21:36 -0800315 /* if zero, use the system-wide default */
316 int sync_speed_min;
317 int sync_speed_max;
318
Bernd Schubert90b08712008-05-23 13:04:38 -0700319 /* resync even though the same disks are shared among md-devices */
320 int parallel_resync;
321
NeilBrown6ff8d8ec2006-01-06 00:20:15 -0800322 int ok_start_degraded;
NeilBrownf72ffdd2014-09-30 14:23:59 +1000323 /* recovery/resync flags
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324 * NEEDED: we might need to start a resync/recover
325 * RUNNING: a thread is running, or about to be started
326 * SYNC: actually doing a resync, not a recovery
Neil Brown72a23c22008-06-28 08:31:41 +1000327 * RECOVER: doing recovery, or need to try it.
NeilBrowndfc70642008-05-23 13:04:39 -0700328 * INTR: resync needs to be aborted for some reason
Linus Torvalds1da177e2005-04-16 15:20:36 -0700329 * DONE: thread is done and is waiting to be reaped
NeilBrown24dd4692005-11-08 21:39:26 -0800330 * REQUEST: user-space has requested a sync (used with SYNC)
Anand Gadiyar411c9402009-07-07 15:24:23 +0530331 * CHECK: user-space request for check-only, no repair
NeilBrownccfcc3c2006-03-27 01:18:09 -0800332 * RESHAPE: A reshape is happening
majianpeng0a19caa2012-11-19 19:57:34 +0800333 * ERROR: sync-action interrupted because io-error
NeilBrownccfcc3c2006-03-27 01:18:09 -0800334 *
335 * If neither SYNC or RESHAPE are set, then it is a recovery.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336 */
337#define MD_RECOVERY_RUNNING 0
338#define MD_RECOVERY_SYNC 1
Neil Brown72a23c22008-06-28 08:31:41 +1000339#define MD_RECOVERY_RECOVER 2
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340#define MD_RECOVERY_INTR 3
341#define MD_RECOVERY_DONE 4
342#define MD_RECOVERY_NEEDED 5
NeilBrown24dd4692005-11-08 21:39:26 -0800343#define MD_RECOVERY_REQUESTED 6
344#define MD_RECOVERY_CHECK 7
NeilBrownccfcc3c2006-03-27 01:18:09 -0800345#define MD_RECOVERY_RESHAPE 8
NeilBrown5fd6c1d2006-06-26 00:27:40 -0700346#define MD_RECOVERY_FROZEN 9
majianpeng0a19caa2012-11-19 19:57:34 +0800347#define MD_RECOVERY_ERROR 10
NeilBrown5fd6c1d2006-06-26 00:27:40 -0700348
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349 unsigned long recovery;
NeilBrown53890422011-07-27 11:00:36 +1000350 /* If a RAID personality determines that recovery (of a particular
351 * device) will fail due to a read error on the source device, it
352 * takes a copy of this number and does not attempt recovery again
353 * until this number changes.
354 */
355 int recovery_disabled;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356
357 int in_sync; /* know to not need resync */
NeilBrownc8c00a62009-08-10 12:50:52 +1000358 /* 'open_mutex' avoids races between 'md_open' and 'do_md_stop', so
359 * that we are never stopping an array while it is open.
360 * 'reconfig_mutex' protects all other reconfiguration.
361 * These locks are separate due to conflicting interactions
362 * with bdev->bd_mutex.
363 * Lock ordering is:
364 * reconfig_mutex -> bd_mutex : e.g. do_md_run -> revalidate_disk
365 * bd_mutex -> open_mutex: e.g. __blkdev_get -> md_open
366 */
367 struct mutex open_mutex;
NeilBrowndf5b89b2006-03-27 01:18:20 -0800368 struct mutex reconfig_mutex;
NeilBrownf2ea68c2008-07-21 17:05:25 +1000369 atomic_t active; /* general refcount */
370 atomic_t openers; /* number of active opens */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371
NeilBrownf0b4f7e2011-02-24 17:26:41 +1100372 int changed; /* True if we might need to
373 * reread partition info */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374 int degraded; /* whether md should consider
375 * adding a spare
376 */
NeilBrown050b6612012-03-19 12:46:39 +1100377 int merge_check_needed; /* at least one
378 * member device
379 * has a
380 * merge_bvec_fn */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381
382 atomic_t recovery_active; /* blocks scheduled, but not written */
383 wait_queue_head_t recovery_wait;
384 sector_t recovery_cp;
Neil Brown5e96ee62008-06-28 08:31:24 +1000385 sector_t resync_min; /* user requested sync
386 * starts here */
NeilBrownc6207272008-02-06 01:39:52 -0800387 sector_t resync_max; /* resync should pause
388 * when it gets here */
NeilBrown06d91a52005-06-21 17:17:12 -0700389
Tejun Heo324a56e2013-12-11 14:11:53 -0500390 struct kernfs_node *sysfs_state; /* handle for 'array_state'
NeilBrownb62b7592008-10-21 13:25:21 +1100391 * file in sysfs.
392 */
Tejun Heo324a56e2013-12-11 14:11:53 -0500393 struct kernfs_node *sysfs_action; /* handle for 'sync_action' */
NeilBrownb62b7592008-10-21 13:25:21 +1100394
NeilBrownd3374822009-01-09 08:31:10 +1100395 struct work_struct del_work; /* used for delayed sysfs removal */
396
NeilBrown85572d72014-12-15 12:56:56 +1100397 /* "lock" protects:
398 * flush_bio transition from NULL to !NULL
399 * rdev superblocks, events
400 * clearing MD_CHANGE_*
401 * in_sync - and related safemode and MD_CHANGE changes
NeilBrown36d091f2014-12-15 12:56:58 +1100402 * pers (also protected by reconfig_mutex and pending IO).
NeilBrown978a7a42014-12-15 12:56:58 +1100403 * clearing ->bitmap
NeilBrown4af1a042014-12-15 12:57:00 +1100404 * clearing ->bitmap_info.file
NeilBrown23da4222014-12-15 12:57:01 +1100405 * changing ->resync_{min,max}
406 * setting MD_RECOVERY_RUNNING (which interacts with resync_{min,max})
NeilBrown85572d72014-12-15 12:56:56 +1100407 */
408 spinlock_t lock;
NeilBrown3d310eb2005-06-21 17:17:26 -0700409 wait_queue_head_t sb_wait; /* for waiting on superblock updates */
NeilBrown7bfa19f2005-06-21 17:17:28 -0700410 atomic_t pending_writes; /* number of active superblock writes */
NeilBrown06d91a52005-06-21 17:17:12 -0700411
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412 unsigned int safemode; /* if set, update "clean" superblock
413 * when no writes pending.
NeilBrownf72ffdd2014-09-30 14:23:59 +1000414 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 unsigned int safemode_delay;
416 struct timer_list safemode_timer;
NeilBrownf72ffdd2014-09-30 14:23:59 +1000417 atomic_t writes_pending;
Jens Axboe165125e2007-07-24 09:28:11 +0200418 struct request_queue *queue; /* for plugging ... */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419
NeilBrownf72ffdd2014-09-30 14:23:59 +1000420 struct bitmap *bitmap; /* the bitmap for the device */
NeilBrownc3d97142009-12-14 12:49:52 +1100421 struct {
422 struct file *file; /* the bitmap file */
NeilBrownf6af9492009-12-14 12:49:54 +1100423 loff_t offset; /* offset from superblock of
NeilBrownc3d97142009-12-14 12:49:52 +1100424 * start of bitmap. May be
425 * negative, but not '0'
NeilBrownf6af9492009-12-14 12:49:54 +1100426 * For external metadata, offset
NeilBrownf72ffdd2014-09-30 14:23:59 +1000427 * from start of device.
NeilBrownc3d97142009-12-14 12:49:52 +1100428 */
NeilBrown6409bb02012-05-22 13:55:07 +1000429 unsigned long space; /* space available at this offset */
NeilBrownf6af9492009-12-14 12:49:54 +1100430 loff_t default_offset; /* this is the offset to use when
NeilBrownc3d97142009-12-14 12:49:52 +1100431 * hot-adding a bitmap. It should
432 * eventually be settable by sysfs.
433 */
NeilBrown6409bb02012-05-22 13:55:07 +1000434 unsigned long default_space; /* space available at
435 * default offset */
NeilBrownc3d97142009-12-14 12:49:52 +1100436 struct mutex mutex;
NeilBrown42a04b52009-12-14 12:49:53 +1100437 unsigned long chunksize;
NeilBrownac2f40b2010-06-01 19:37:31 +1000438 unsigned long daemon_sleep; /* how many jiffies between updates? */
NeilBrown42a04b52009-12-14 12:49:53 +1100439 unsigned long max_write_behind; /* write-behind mode */
NeilBrownece5cff2009-12-14 12:49:56 +1100440 int external;
Goldwyn Rodriguesc4ce8672014-03-29 10:20:02 -0500441 int nodes; /* Maximum number of nodes in the cluster */
Goldwyn Rodriguescf921cc2014-03-30 00:42:49 -0500442 char cluster_name[64]; /* Name of the cluster */
NeilBrownc3d97142009-12-14 12:49:52 +1100443 } bitmap_info;
NeilBrown32a76272005-06-21 17:17:14 -0700444
NeilBrownf72ffdd2014-09-30 14:23:59 +1000445 atomic_t max_corr_read_errors; /* max read retries */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446 struct list_head all_mddevs;
NeilBrowna2826aa2009-12-14 12:49:49 +1100447
NeilBrowna64c8762010-04-14 17:15:37 +1000448 struct attribute_group *to_remove;
NeilBrown252ac522010-06-01 19:37:29 +1000449
NeilBrowna167f662010-10-26 18:31:13 +1100450 struct bio_set *bio_set;
451
Tejun Heoe9c74692010-09-03 11:56:18 +0200452 /* Generic flush handling.
453 * The last to finish preflush schedules a worker to submit
454 * the rest of the request (without the REQ_FLUSH flag).
NeilBrowna2826aa2009-12-14 12:49:49 +1100455 */
Tejun Heoe9c74692010-09-03 11:56:18 +0200456 struct bio *flush_bio;
NeilBrowna2826aa2009-12-14 12:49:49 +1100457 atomic_t flush_pending;
Tejun Heoe9c74692010-09-03 11:56:18 +0200458 struct work_struct flush_work;
NeilBrown768a4182010-07-26 11:49:55 +1000459 struct work_struct event_work; /* used by dm to report failure event */
NeilBrownfd01b882011-10-11 16:47:53 +1100460 void (*sync_super)(struct mddev *mddev, struct md_rdev *rdev);
Goldwyn Rodriguesc4ce8672014-03-29 10:20:02 -0500461 struct md_cluster_info *cluster_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462};
463
NeilBrown5c47daf2014-12-15 12:57:01 +1100464static inline int __must_check mddev_lock(struct mddev *mddev)
465{
466 return mutex_lock_interruptible(&mddev->reconfig_mutex);
467}
468
469/* Sometimes we need to take the lock in a situation where
470 * failure due to interrupts is not acceptable.
471 */
472static inline void mddev_lock_nointr(struct mddev *mddev)
473{
474 mutex_lock(&mddev->reconfig_mutex);
475}
476
477static inline int mddev_is_locked(struct mddev *mddev)
478{
479 return mutex_is_locked(&mddev->reconfig_mutex);
480}
481
482static inline int mddev_trylock(struct mddev *mddev)
483{
484 return mutex_trylock(&mddev->reconfig_mutex);
485}
486extern void mddev_unlock(struct mddev *mddev);
487
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors)
489{
NeilBrownf72ffdd2014-09-30 14:23:59 +1000490 atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491}
492
NeilBrown84fc4b52011-10-11 16:49:58 +1100493struct md_personality
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494{
495 char *name;
NeilBrown2604b702006-01-06 00:20:36 -0800496 int level;
497 struct list_head list;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498 struct module *owner;
Linus Torvaldsb4fdcb02011-11-04 17:06:58 -0700499 void (*make_request)(struct mddev *mddev, struct bio *bio);
NeilBrownfd01b882011-10-11 16:47:53 +1100500 int (*run)(struct mddev *mddev);
NeilBrownafa0f552014-12-15 12:56:58 +1100501 void (*free)(struct mddev *mddev, void *priv);
NeilBrownfd01b882011-10-11 16:47:53 +1100502 void (*status)(struct seq_file *seq, struct mddev *mddev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503 /* error_handler must set ->faulty and clear ->in_sync
NeilBrownf72ffdd2014-09-30 14:23:59 +1000504 * if appropriate, and should abort recovery if needed
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505 */
NeilBrownfd01b882011-10-11 16:47:53 +1100506 void (*error_handler)(struct mddev *mddev, struct md_rdev *rdev);
507 int (*hot_add_disk) (struct mddev *mddev, struct md_rdev *rdev);
NeilBrownb8321b62011-12-23 10:17:51 +1100508 int (*hot_remove_disk) (struct mddev *mddev, struct md_rdev *rdev);
NeilBrownfd01b882011-10-11 16:47:53 +1100509 int (*spare_active) (struct mddev *mddev);
NeilBrown09314792015-02-19 16:04:40 +1100510 sector_t (*sync_request)(struct mddev *mddev, sector_t sector_nr, int *skipped);
NeilBrownfd01b882011-10-11 16:47:53 +1100511 int (*resize) (struct mddev *mddev, sector_t sectors);
512 sector_t (*size) (struct mddev *mddev, sector_t sectors, int raid_disks);
513 int (*check_reshape) (struct mddev *mddev);
514 int (*start_reshape) (struct mddev *mddev);
515 void (*finish_reshape) (struct mddev *mddev);
NeilBrown36fa3062005-09-09 16:23:45 -0700516 /* quiesce moves between quiescence states
517 * 0 - fully active
518 * 1 - no new requests allowed
519 * others - reserved
520 */
NeilBrownfd01b882011-10-11 16:47:53 +1100521 void (*quiesce) (struct mddev *mddev, int state);
NeilBrown245f46c2009-03-31 14:39:39 +1100522 /* takeover is used to transition an array from one
523 * personality to another. The new personality must be able
524 * to handle the data in the current layout.
525 * e.g. 2drive raid1 -> 2drive raid5
526 * ndrive raid5 -> degraded n+1drive raid6 with special layout
527 * If the takeover succeeds, a new 'private' structure is returned.
528 * This needs to be installed and then ->run used to activate the
529 * array.
530 */
NeilBrownfd01b882011-10-11 16:47:53 +1100531 void *(*takeover) (struct mddev *mddev);
NeilBrown5c675f82014-12-15 12:56:56 +1100532 /* congested implements bdi.congested_fn().
533 * Will not be called while array is 'suspended' */
534 int (*congested)(struct mddev *mddev, int bits);
NeilBrown64590f42014-12-15 12:56:57 +1100535 /* mergeable_bvec is use to implement ->merge_bvec_fn */
536 int (*mergeable_bvec)(struct mddev *mddev,
537 struct bvec_merge_data *bvm,
538 struct bio_vec *biovec);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539};
540
NeilBrown007583c2005-11-08 21:39:30 -0800541struct md_sysfs_entry {
542 struct attribute attr;
NeilBrownfd01b882011-10-11 16:47:53 +1100543 ssize_t (*show)(struct mddev *, char *);
544 ssize_t (*store)(struct mddev *, const char *, size_t);
NeilBrown007583c2005-11-08 21:39:30 -0800545};
NeilBrown43a70502009-12-14 12:49:55 +1100546extern struct attribute_group md_bitmap_group;
NeilBrown007583c2005-11-08 21:39:30 -0800547
Tejun Heo324a56e2013-12-11 14:11:53 -0500548static inline struct kernfs_node *sysfs_get_dirent_safe(struct kernfs_node *sd, char *name)
NeilBrown00bcb4a2010-06-01 19:37:23 +1000549{
550 if (sd)
Tejun Heo388975c2013-09-11 23:19:13 -0400551 return sysfs_get_dirent(sd, name);
NeilBrown00bcb4a2010-06-01 19:37:23 +1000552 return sd;
553}
Tejun Heo324a56e2013-12-11 14:11:53 -0500554static inline void sysfs_notify_dirent_safe(struct kernfs_node *sd)
NeilBrown00bcb4a2010-06-01 19:37:23 +1000555{
556 if (sd)
557 sysfs_notify_dirent(sd);
558}
559
NeilBrownfd01b882011-10-11 16:47:53 +1100560static inline char * mdname (struct mddev * mddev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561{
562 return mddev->gendisk ? mddev->gendisk->disk_name : "mdX";
563}
564
NeilBrownfd01b882011-10-11 16:47:53 +1100565static inline int sysfs_link_rdev(struct mddev *mddev, struct md_rdev *rdev)
Namhyung Kim36fad852011-07-27 11:00:36 +1000566{
567 char nm[20];
Jonathan Brassow90584fc2013-03-07 16:24:26 -0600568 if (!test_bit(Replacement, &rdev->flags) && mddev->kobj.sd) {
NeilBrown2d78f8c2011-12-23 10:17:51 +1100569 sprintf(nm, "rd%d", rdev->raid_disk);
570 return sysfs_create_link(&mddev->kobj, &rdev->kobj, nm);
571 } else
572 return 0;
Namhyung Kim36fad852011-07-27 11:00:36 +1000573}
574
NeilBrownfd01b882011-10-11 16:47:53 +1100575static inline void sysfs_unlink_rdev(struct mddev *mddev, struct md_rdev *rdev)
Namhyung Kim36fad852011-07-27 11:00:36 +1000576{
577 char nm[20];
Jonathan Brassow90584fc2013-03-07 16:24:26 -0600578 if (!test_bit(Replacement, &rdev->flags) && mddev->kobj.sd) {
NeilBrown2d78f8c2011-12-23 10:17:51 +1100579 sprintf(nm, "rd%d", rdev->raid_disk);
580 sysfs_remove_link(&mddev->kobj, nm);
581 }
Namhyung Kim36fad852011-07-27 11:00:36 +1000582}
583
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584/*
585 * iterates through some rdev ringlist. It's safe to remove the
586 * current 'rdev'. Dont touch 'tmp' though.
587 */
Cheng Renquan159ec1f2009-01-09 08:31:08 +1100588#define rdev_for_each_list(rdev, tmp, head) \
589 list_for_each_entry_safe(rdev, tmp, head, same_set)
590
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591/*
592 * iterates through the 'same array disks' ringlist
593 */
NeilBrowndafb20f2012-03-19 12:46:39 +1100594#define rdev_for_each(rdev, mddev) \
595 list_for_each_entry(rdev, &((mddev)->disks), same_set)
596
597#define rdev_for_each_safe(rdev, tmp, mddev) \
Cheng Renquan159ec1f2009-01-09 08:31:08 +1100598 list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599
NeilBrown4b809912008-07-21 17:05:25 +1000600#define rdev_for_each_rcu(rdev, mddev) \
601 list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set)
602
NeilBrown2b8bf342011-10-11 16:48:23 +1100603struct md_thread {
Shaohua Li4ed87312012-10-11 13:34:00 +1100604 void (*run) (struct md_thread *thread);
NeilBrownfd01b882011-10-11 16:47:53 +1100605 struct mddev *mddev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606 wait_queue_head_t wqueue;
NeilBrownf72ffdd2014-09-30 14:23:59 +1000607 unsigned long flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608 struct task_struct *tsk;
NeilBrown32a76272005-06-21 17:17:14 -0700609 unsigned long timeout;
Shaohua Li4ed87312012-10-11 13:34:00 +1100610 void *private;
NeilBrown2b8bf342011-10-11 16:48:23 +1100611};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612
613#define THREAD_WAKEUP 0
614
NeilBrown1345b1d2006-01-06 00:20:40 -0800615static inline void safe_put_page(struct page *p)
616{
617 if (p) put_page(p);
618}
619
NeilBrown84fc4b52011-10-11 16:49:58 +1100620extern int register_md_personality(struct md_personality *p);
621extern int unregister_md_personality(struct md_personality *p);
Goldwyn Rodriguesedb39c92014-03-29 10:01:53 -0500622extern int register_md_cluster_operations(struct md_cluster_operations *ops,
623 struct module *module);
624extern int unregister_md_cluster_operations(void);
625extern int md_setup_cluster(struct mddev *mddev, int nodes);
626extern void md_cluster_stop(struct mddev *mddev);
NeilBrown2b8bf342011-10-11 16:48:23 +1100627extern struct md_thread *md_register_thread(
Shaohua Li4ed87312012-10-11 13:34:00 +1100628 void (*run)(struct md_thread *thread),
NeilBrown2b8bf342011-10-11 16:48:23 +1100629 struct mddev *mddev,
630 const char *name);
631extern void md_unregister_thread(struct md_thread **threadp);
632extern void md_wakeup_thread(struct md_thread *thread);
NeilBrownfd01b882011-10-11 16:47:53 +1100633extern void md_check_recovery(struct mddev *mddev);
Jonathan Brassowa91d5ac2013-04-24 11:42:43 +1000634extern void md_reap_sync_thread(struct mddev *mddev);
NeilBrownfd01b882011-10-11 16:47:53 +1100635extern void md_write_start(struct mddev *mddev, struct bio *bi);
636extern void md_write_end(struct mddev *mddev);
637extern void md_done_sync(struct mddev *mddev, int blocks, int ok);
638extern void md_error(struct mddev *mddev, struct md_rdev *rdev);
NeilBrownc6563a82012-05-21 09:27:00 +1000639extern void md_finish_reshape(struct mddev *mddev);
NeilBrown92022952009-03-31 14:33:13 +1100640
NeilBrownfd01b882011-10-11 16:47:53 +1100641extern int mddev_congested(struct mddev *mddev, int bits);
642extern void md_flush_request(struct mddev *mddev, struct bio *bio);
643extern void md_super_write(struct mddev *mddev, struct md_rdev *rdev,
NeilBrown92022952009-03-31 14:33:13 +1100644 sector_t sector, int size, struct page *page);
NeilBrownfd01b882011-10-11 16:47:53 +1100645extern void md_super_wait(struct mddev *mddev);
NeilBrownf72ffdd2014-09-30 14:23:59 +1000646extern int sync_page_io(struct md_rdev *rdev, sector_t sector, int size,
Jonathan Brassowccebd4c2011-01-14 09:14:33 +1100647 struct page *page, int rw, bool metadata_op);
Shaohua Li4ed87312012-10-11 13:34:00 +1100648extern void md_do_sync(struct md_thread *thread);
NeilBrownfd01b882011-10-11 16:47:53 +1100649extern void md_new_event(struct mddev *mddev);
650extern int md_allow_write(struct mddev *mddev);
651extern void md_wait_for_blocked_rdev(struct md_rdev *rdev, struct mddev *mddev);
652extern void md_set_array_sectors(struct mddev *mddev, sector_t array_sectors);
653extern int md_check_no_bitmap(struct mddev *mddev);
654extern int md_integrity_register(struct mddev *mddev);
655extern void md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev);
NeilBrown72e02072009-12-14 12:49:55 +1100656extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale);
Christoph Hellwig63fe0812009-04-14 12:01:53 +1000657
NeilBrownfd01b882011-10-11 16:47:53 +1100658extern void mddev_init(struct mddev *mddev);
659extern int md_run(struct mddev *mddev);
660extern void md_stop(struct mddev *mddev);
661extern void md_stop_writes(struct mddev *mddev);
NeilBrown3cb03002011-10-11 16:45:26 +1100662extern int md_rdev_init(struct md_rdev *rdev);
NeilBrown545c8792012-05-22 13:54:30 +1000663extern void md_rdev_clear(struct md_rdev *rdev);
NeilBrown390ee602010-06-01 19:37:27 +1000664
NeilBrownfd01b882011-10-11 16:47:53 +1100665extern void mddev_suspend(struct mddev *mddev);
666extern void mddev_resume(struct mddev *mddev);
NeilBrowna167f662010-10-26 18:31:13 +1100667extern struct bio *bio_clone_mddev(struct bio *bio, gfp_t gfp_mask,
NeilBrownfd01b882011-10-11 16:47:53 +1100668 struct mddev *mddev);
NeilBrowna167f662010-10-26 18:31:13 +1100669extern struct bio *bio_alloc_mddev(gfp_t gfp_mask, int nr_iovecs,
NeilBrownfd01b882011-10-11 16:47:53 +1100670 struct mddev *mddev);
NeilBrown9cbb1752012-07-31 09:08:14 +0200671
NeilBrown74018dc2012-07-31 09:08:15 +0200672extern void md_unplug(struct blk_plug_cb *cb, bool from_schedule);
Goldwyn Rodrigues1d7e3e92014-06-07 01:53:00 -0500673extern void md_reload_sb(struct mddev *mddev);
Goldwyn Rodrigues1aee41f2014-10-29 18:51:31 -0500674extern void md_update_sb(struct mddev *mddev, int force);
Goldwyn Rodriguesfb56dfe2015-04-14 10:43:24 -0500675extern void md_kick_rdev_from_array(struct md_rdev * rdev);
Goldwyn Rodrigues57d051d2015-04-14 10:43:55 -0500676struct md_rdev *md_find_rdev_nr_rcu(struct mddev *mddev, int nr);
NeilBrown9cbb1752012-07-31 09:08:14 +0200677static inline int mddev_check_plugged(struct mddev *mddev)
678{
679 return !!blk_check_plugged(md_unplug, mddev,
680 sizeof(struct blk_plug_cb));
681}
Hannes Reineckedfe15ac2012-07-26 11:12:18 +0200682
683static inline void rdev_dec_pending(struct md_rdev *rdev, struct mddev *mddev)
684{
685 int faulty = test_bit(Faulty, &rdev->flags);
686 if (atomic_dec_and_test(&rdev->nr_pending) && faulty) {
687 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
688 md_wakeup_thread(mddev->thread);
689 }
690}
691
Goldwyn Rodriguesedb39c92014-03-29 10:01:53 -0500692extern struct md_cluster_operations *md_cluster_ops;
Goldwyn Rodriguesc4ce8672014-03-29 10:20:02 -0500693static inline int mddev_is_clustered(struct mddev *mddev)
694{
695 return mddev->cluster_info && mddev->bitmap_info.nodes > 1;
696}
Christoph Hellwig63fe0812009-04-14 12:01:53 +1000697#endif /* _MD_MD_H */