blob: fb7eab58cfd517b3f52e7de0898104df2a71b2a0 [file] [log] [blame]
Thomas Gleixneraf1a8892019-05-20 19:08:12 +02001/* SPDX-License-Identifier: GPL-2.0-or-later */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002/*
Wang Sheng-Hui7e841522011-09-21 15:37:46 +10003 md.h : kernel internal structure of the Linux MD driver
Linus Torvalds1da177e2005-04-16 15:20:36 -07004 Copyright (C) 1996-98 Ingo Molnar, Gadi Oxman
NeilBrownf72ffdd2014-09-30 14:23:59 +10005
Linus Torvalds1da177e2005-04-16 15:20:36 -07006*/
7
Christoph Hellwig63fe0812009-04-14 12:01:53 +10008#ifndef _MD_MD_H
9#define _MD_MD_H
Linus Torvalds1da177e2005-04-16 15:20:36 -070010
Christoph Hellwig63fe0812009-04-14 12:01:53 +100011#include <linux/blkdev.h>
Tejun Heo66114ca2015-05-22 17:13:32 -040012#include <linux/backing-dev.h>
Vishal Vermafc974ee2015-12-24 19:20:34 -070013#include <linux/badblocks.h>
Christoph Hellwig63fe0812009-04-14 12:01:53 +100014#include <linux/kobject.h>
15#include <linux/list.h>
16#include <linux/mm.h>
17#include <linux/mutex.h>
18#include <linux/timer.h>
19#include <linux/wait.h>
20#include <linux/workqueue.h>
Goldwyn Rodriguesedb39c92014-03-29 10:01:53 -050021#include "md-cluster.h"
David Howells93614012006-09-30 20:45:40 +020022
Linus Torvalds1da177e2005-04-16 15:20:36 -070023#define MaxSector (~(sector_t)0)
Linus Torvalds1da177e2005-04-16 15:20:36 -070024
Linus Torvalds1da177e2005-04-16 15:20:36 -070025/*
NeilBrown46533ff2016-11-18 16:16:11 +110026 * These flags should really be called "NO_RETRY" rather than
27 * "FAILFAST" because they don't make any promise about time lapse,
28 * only about the number of retries, which will be zero.
29 * REQ_FAILFAST_DRIVER is not included because
30 * Commit: 4a27446f3e39 ("[SCSI] modify scsi to handle new fail fast flags.")
31 * seems to suggest that the errors it avoids retrying should usually
32 * be retried.
33 */
34#define MD_FAILFAST (REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT)
Guoqing Jiang69b00b52019-12-23 10:49:00 +010035
36/*
37 * The struct embedded in rdev is used to serialize IO.
38 */
39struct serial_in_rdev {
40 struct rb_root_cached serial_rb;
41 spinlock_t serial_lock;
42 wait_queue_head_t serial_io_wait;
43};
44
NeilBrown46533ff2016-11-18 16:16:11 +110045/*
Linus Torvalds1da177e2005-04-16 15:20:36 -070046 * MD's 'extended' device
47 */
NeilBrown3cb03002011-10-11 16:45:26 +110048struct md_rdev {
Linus Torvalds1da177e2005-04-16 15:20:36 -070049 struct list_head same_set; /* RAID devices within the same set */
50
Andre Nolldd8ac332009-03-31 14:33:13 +110051 sector_t sectors; /* Device size (in 512bytes sectors) */
NeilBrownfd01b882011-10-11 16:47:53 +110052 struct mddev *mddev; /* RAID array if running */
NeilBrowneea1bf32009-03-31 14:27:02 +110053 int last_events; /* IO event timestamp */
Linus Torvalds1da177e2005-04-16 15:20:36 -070054
Jonathan Brassowa6ff7e02011-01-14 09:14:34 +110055 /*
56 * If meta_bdev is non-NULL, it means that a separate device is
57 * being used to store the metadata (superblock/bitmap) which
58 * would otherwise be contained on the same device as the data (bdev).
59 */
60 struct block_device *meta_bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -070061 struct block_device *bdev; /* block device handle */
62
NeilBrown2699b672011-07-28 11:31:47 +100063 struct page *sb_page, *bb_page;
Linus Torvalds1da177e2005-04-16 15:20:36 -070064 int sb_loaded;
NeilBrown42543762006-06-26 00:27:57 -070065 __u64 sb_events;
Linus Torvalds1da177e2005-04-16 15:20:36 -070066 sector_t data_offset; /* start of data in array */
NeilBrownc6563a82012-05-21 09:27:00 +100067 sector_t new_data_offset;/* only relevant while reshaping */
NeilBrownf72ffdd2014-09-30 14:23:59 +100068 sector_t sb_start; /* offset of the super block (in 512byte sectors) */
NeilBrown0002b272005-09-09 16:23:53 -070069 int sb_size; /* bytes in the superblock */
Linus Torvalds1da177e2005-04-16 15:20:36 -070070 int preferred_minor; /* autorun support */
71
NeilBrown86e6ffd2005-11-08 21:39:24 -080072 struct kobject kobj;
73
Linus Torvalds1da177e2005-04-16 15:20:36 -070074 /* A device can be in one of three states based on two flags:
75 * Not working: faulty==1 in_sync==0
76 * Fully working: faulty==0 in_sync==1
77 * Working, but not
78 * in sync with array
79 * faulty==0 in_sync==0
80 *
81 * It can never have faulty==1, in_sync==1
82 * This reduces the burden of testing multiple flags in many cases
83 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070084
NeilBrown2d78f8c2011-12-23 10:17:51 +110085 unsigned long flags; /* bit set of 'enum flag_bits' bits. */
Dan Williams6bfe0b42008-04-30 00:52:32 -070086 wait_queue_head_t blocked_wait;
NeilBrown8ddf9ef2005-09-09 16:23:45 -070087
Linus Torvalds1da177e2005-04-16 15:20:36 -070088 int desc_nr; /* descriptor index in the superblock */
89 int raid_disk; /* role of device in array */
NeilBrowne93f68a2010-06-15 09:36:03 +010090 int new_raid_disk; /* role that the device will have in
91 * the array after a level-change completes.
92 */
NeilBrown41158c72005-06-21 17:17:25 -070093 int saved_raid_disk; /* role that device used to have in the
94 * array and could again if we did a partial
95 * resync from the bitmap
96 */
Shaohua Li3069aa82015-08-13 14:31:56 -070097 union {
98 sector_t recovery_offset;/* If this device has been partially
NeilBrown5fd6c1d2006-06-26 00:27:40 -070099 * recovered, this is where we were
100 * up to.
101 */
Shaohua Li3069aa82015-08-13 14:31:56 -0700102 sector_t journal_tail; /* If this device is a journal device,
103 * this is the journal tail (journal
104 * recovery start point)
105 */
106 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107
108 atomic_t nr_pending; /* number of pending requests.
109 * only maintained for arrays that
110 * support hot removal
111 */
NeilBrownba22dcb2005-11-08 21:39:31 -0800112 atomic_t read_errors; /* number of consecutive read errors that
113 * we have tried to ignore.
114 */
Arnd Bergmann0e3ef492016-06-17 17:33:10 +0200115 time64_t last_read_error; /* monotonic time since our
Robert Becker1e509152009-12-14 12:49:58 +1100116 * last read error
117 */
NeilBrown4dbcdc72006-01-06 00:20:52 -0800118 atomic_t corrected_errors; /* number of corrected read errors,
119 * for reporting to userspace and storing
120 * in superblock.
121 */
Guoqing Jiang3e148a32019-06-19 17:30:46 +0800122
Guoqing Jiang69b00b52019-12-23 10:49:00 +0100123 struct serial_in_rdev *serial; /* used for raid1 io serialization */
Guoqing Jiang3e148a32019-06-19 17:30:46 +0800124
NeilBrown5792a282007-04-04 19:08:18 -0700125 struct work_struct del_work; /* used for delayed sysfs removal */
NeilBrown3c0ee632008-10-21 13:25:28 +1100126
Tejun Heo324a56e2013-12-11 14:11:53 -0500127 struct kernfs_node *sysfs_state; /* handle for 'state'
NeilBrown3c0ee632008-10-21 13:25:28 +1100128 * sysfs entry */
Junxiao Bie1a86db2020-07-14 16:10:26 -0700129 /* handle for 'unacknowledged_bad_blocks' sysfs dentry */
130 struct kernfs_node *sysfs_unack_badblocks;
131 /* handle for 'bad_blocks' sysfs dentry */
132 struct kernfs_node *sysfs_badblocks;
Vishal Vermafc974ee2015-12-24 19:20:34 -0700133 struct badblocks badblocks;
Artur Paszkiewiczea0213e2017-03-09 09:59:57 +0100134
135 struct {
136 short offset; /* Offset from superblock to start of PPL.
137 * Not used by external metadata. */
138 unsigned int size; /* Size in sectors of the PPL space */
139 sector_t sector; /* First sector of the PPL space */
140 } ppl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141};
NeilBrown2d78f8c2011-12-23 10:17:51 +1100142enum flag_bits {
143 Faulty, /* device is known to have a fault */
144 In_sync, /* device is in_sync with rest of array */
NeilBrown8313b8e2013-12-12 10:13:33 +1100145 Bitmap_sync, /* ..actually, not quite In_sync. Need a
Guoqing Jiang4aaf76942017-07-04 11:20:30 +0800146 * bitmap-based recovery to get fully in sync.
147 * The bit is only meaningful before device
148 * has been passed to pers->hot_add_disk.
NeilBrown8313b8e2013-12-12 10:13:33 +1100149 */
NeilBrown2d78f8c2011-12-23 10:17:51 +1100150 WriteMostly, /* Avoid reading if at all possible */
151 AutoDetected, /* added by auto-detect */
152 Blocked, /* An error occurred but has not yet
153 * been acknowledged by the metadata
154 * handler, so don't allow writes
155 * until it is cleared */
156 WriteErrorSeen, /* A write error has been seen on this
157 * device
158 */
159 FaultRecorded, /* Intermediate state for clearing
160 * Blocked. The Fault is/will-be
161 * recorded in the metadata, but that
162 * metadata hasn't been stored safely
163 * on disk yet.
164 */
165 BlockedBadBlocks, /* A writer is blocked because they
166 * found an unacknowledged bad-block.
167 * This can safely be cleared at any
168 * time, and the writer will re-check.
169 * It may be set at any time, and at
170 * worst the writer will timeout and
171 * re-check. So setting it as
172 * accurately as possible is good, but
173 * not absolutely critical.
174 */
175 WantReplacement, /* This device is a candidate to be
176 * hot-replaced, either because it has
177 * reported some faults, or because
178 * of explicit request.
179 */
180 Replacement, /* This device is a replacement for
181 * a want_replacement device with same
182 * raid_disk number.
183 */
Goldwyn Rodrigues1aee41f2014-10-29 18:51:31 -0500184 Candidate, /* For clustered environments only:
185 * This device is seen locally but not
186 * by the whole cluster
187 */
Song Liubac624f2015-08-13 14:31:55 -0700188 Journal, /* This device is used as journal for
189 * raid-5/6.
190 * Usually, this device should be faster
191 * than other devices in the array
192 */
Guoqing Jiang659b2542015-12-21 10:50:59 +1100193 ClusterRemove,
NeilBrownd787be42016-06-02 16:19:53 +1000194 RemoveSynchronized, /* synchronize_rcu() was called after
195 * this device was known to be faulty,
196 * so it is safe to remove without
197 * another synchronize_rcu() call.
198 */
Tomasz Majchrzak35b785f2016-10-21 16:26:57 +0200199 ExternalBbl, /* External metadata provides bad
200 * block management for a disk
201 */
NeilBrown688834e2016-11-18 16:16:11 +1100202 FailFast, /* Minimal retries should be attempted on
203 * this device, so use REQ_FAILFAST_DEV.
204 * Also don't try to repair failed reads.
205 * It is expects that no bad block log
206 * is present.
207 */
NeilBrown46533ff2016-11-18 16:16:11 +1100208 LastDev, /* Seems to be the last working dev as
209 * it didn't fail, so don't use FailFast
210 * any more for metadata
211 */
Guoqing Jiang404659c2019-12-23 10:48:53 +0100212 CollisionCheck, /*
213 * check if there is collision between raid1
214 * serial bios.
Guoqing Jiang3e148a32019-06-19 17:30:46 +0800215 */
NeilBrown2d78f8c2011-12-23 10:17:51 +1100216};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217
NeilBrown3cb03002011-10-11 16:45:26 +1100218static inline int is_badblock(struct md_rdev *rdev, sector_t s, int sectors,
NeilBrown2230dfe2011-07-28 11:31:46 +1000219 sector_t *first_bad, int *bad_sectors)
220{
221 if (unlikely(rdev->badblocks.count)) {
Vishal Vermafc974ee2015-12-24 19:20:34 -0700222 int rv = badblocks_check(&rdev->badblocks, rdev->data_offset + s,
NeilBrown2230dfe2011-07-28 11:31:46 +1000223 sectors,
224 first_bad, bad_sectors);
225 if (rv)
226 *first_bad -= rdev->data_offset;
227 return rv;
228 }
229 return 0;
230}
NeilBrown3cb03002011-10-11 16:45:26 +1100231extern int rdev_set_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
NeilBrownc6563a82012-05-21 09:27:00 +1000232 int is_new);
233extern int rdev_clear_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
234 int is_new);
Goldwyn Rodriguesc4ce8672014-03-29 10:20:02 -0500235struct md_cluster_info;
236
Shaohua Li394ed8e2017-01-04 16:10:19 -0800237/* change UNSUPPORTED_MDDEV_FLAGS for each array type if new flag is added */
NeilBrownbe306c22016-11-09 10:21:33 +1100238enum mddev_flags {
NeilBrownbe306c22016-11-09 10:21:33 +1100239 MD_ARRAY_FIRST_USE, /* First use of array, needs initialization */
240 MD_CLOSING, /* If set, we are closing the array, do not open
241 * it then */
242 MD_JOURNAL_CLEAN, /* A raid with journal is already clean */
243 MD_HAS_JOURNAL, /* The raid array has journal feature set */
NeilBrownbe306c22016-11-09 10:21:33 +1100244 MD_CLUSTER_RESYNC_LOCKED, /* cluster raid only, which means node
245 * already took resync lock, need to
246 * release the lock */
NeilBrown46533ff2016-11-18 16:16:11 +1100247 MD_FAILFAST_SUPPORTED, /* Using MD_FAILFAST on metadata writes is
248 * supported as calls to md_error() will
249 * never cause the array to become failed.
250 */
Artur Paszkiewiczea0213e2017-03-09 09:59:57 +0100251 MD_HAS_PPL, /* The raid array has PPL feature set */
Pawel Baldysiakddc08822017-08-16 17:13:45 +0200252 MD_HAS_MULTIPLE_PPLS, /* The raid array has multiple PPLs feature set */
NeilBrown35bfc522017-10-17 13:46:43 +1100253 MD_ALLOW_SB_UPDATE, /* md_check_recovery is allowed to update
254 * the metadata without taking reconfig_mutex.
255 */
256 MD_UPDATING_SB, /* md_check_recovery is updating the metadata
257 * without explicitly holding reconfig_mutex.
258 */
NeilBrown9d4b45d2019-08-20 10:21:09 +1000259 MD_NOT_READY, /* do_md_run() is active, so 'array_state'
260 * must not report that array is ready yet
261 */
Guilherme G. Piccoli62f7b192019-09-03 16:49:00 -0300262 MD_BROKEN, /* This is used in RAID-0/LINEAR only, to stop
263 * I/O in case an array member is gone/failed.
264 */
NeilBrownbe306c22016-11-09 10:21:33 +1100265};
Shaohua Li29530792016-12-08 15:48:19 -0800266
267enum mddev_sb_flags {
268 MD_SB_CHANGE_DEVS, /* Some device status has changed */
269 MD_SB_CHANGE_CLEAN, /* transition to or from 'clean' */
270 MD_SB_CHANGE_PENDING, /* switch from 'clean' to 'active' in progress */
271 MD_SB_NEED_REWRITE, /* metadata write needs to be repeated */
272};
273
Guoqing Jiang404659c2019-12-23 10:48:53 +0100274#define NR_SERIAL_INFOS 8
275/* record current range of serialize IOs */
276struct serial_info {
Guoqing Jiang69b00b52019-12-23 10:49:00 +0100277 struct rb_node node;
278 sector_t start; /* start sector of rb node */
279 sector_t last; /* end sector of rb node */
280 sector_t _subtree_last; /* highest sector in subtree of rb node */
Guoqing Jiang3e148a32019-06-19 17:30:46 +0800281};
282
NeilBrownfd01b882011-10-11 16:47:53 +1100283struct mddev {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284 void *private;
NeilBrown84fc4b52011-10-11 16:49:58 +1100285 struct md_personality *pers;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286 dev_t unit;
287 int md_minor;
NeilBrown7a0a5352013-08-27 16:28:23 +1000288 struct list_head disks;
NeilBrown850b2b422006-10-03 01:15:46 -0700289 unsigned long flags;
Shaohua Li29530792016-12-08 15:48:19 -0800290 unsigned long sb_flags;
NeilBrown850b2b422006-10-03 01:15:46 -0700291
NeilBrown409c57f2009-03-31 14:39:39 +1100292 int suspended;
293 atomic_t active_io;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 int ro;
NeilBrownbb4f1e92010-08-08 21:18:03 +1000295 int sysfs_active; /* set when sysfs deletes
296 * are happening, so run/
297 * takeover/stop are not safe
298 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299 struct gendisk *gendisk;
300
NeilBrowneae17012005-11-08 21:39:23 -0800301 struct kobject kobj;
NeilBrownd3374822009-01-09 08:31:10 +1100302 int hold_active;
303#define UNTIL_IOCTL 1
NeilBrownefeb53c2009-01-09 08:31:10 +1100304#define UNTIL_STOP 2
NeilBrowneae17012005-11-08 21:39:23 -0800305
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306 /* Superblock information */
307 int major_version,
308 minor_version,
309 patch_version;
310 int persistent;
NeilBrownf72ffdd2014-09-30 14:23:59 +1000311 int external; /* metadata is
NeilBrowne6910632008-02-06 01:39:51 -0800312 * managed externally */
313 char metadata_type[17]; /* externally set*/
Mike Snitzer77a68692020-12-14 12:08:48 -0500314 int chunk_sectors;
Deepa Dinamani9ebc6ef2015-12-21 10:51:01 +1100315 time64_t ctime, utime;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316 int level, layout;
NeilBrownd9d166c2006-01-06 00:20:51 -0800317 char clevel[16];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318 int raid_disks;
319 int max_disks;
NeilBrownf72ffdd2014-09-30 14:23:59 +1000320 sector_t dev_sectors; /* used size of
Andre Noll58c0fed2009-03-31 14:33:13 +1100321 * component devices */
Andre Nollf233ea52008-07-21 17:05:22 +1000322 sector_t array_sectors; /* exported array size */
Dan Williamsb522adc2009-03-31 15:00:31 +1100323 int external_size; /* size managed
324 * externally */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325 __u64 events;
NeilBrowna8707c02010-05-18 09:28:43 +1000326 /* If the last 'event' was simply a clean->dirty transition, and
327 * we didn't write it to the spares, then it is safe and simple
328 * to just decrement the event count on a dirty->clean transition.
329 * So we record that possibility here.
330 */
331 int can_decrease_events;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332
333 char uuid[16];
334
NeilBrownf6705572006-03-27 01:18:11 -0800335 /* If the array is being reshaped, we need to record the
336 * new shape and an indication of where we are up to.
337 * This is written to the superblock.
338 * If reshape_position is MaxSector, then no reshape is happening (yet).
339 */
340 sector_t reshape_position;
Andre Noll664e7c42009-06-18 08:45:27 +1000341 int delta_disks, new_level, new_layout;
Mike Snitzer77a68692020-12-14 12:08:48 -0500342 int new_chunk_sectors;
NeilBrown2c810cd2012-05-21 09:27:00 +1000343 int reshape_backwards;
NeilBrownf6705572006-03-27 01:18:11 -0800344
NeilBrown2b8bf342011-10-11 16:48:23 +1100345 struct md_thread *thread; /* management thread */
346 struct md_thread *sync_thread; /* doing resync or reconstruct */
Jonathan Brassowc4a39552013-06-25 01:23:59 -0500347
348 /* 'last_sync_action' is initialized to "none". It is set when a
349 * sync operation (i.e "data-check", "requested-resync", "resync",
350 * "recovery", or "reshape") is started. It holds this value even
351 * when the sync thread is "frozen" (interrupted) or "idle" (stopped
352 * or finished). It is overwritten when a new sync operation is begun.
353 */
354 char *last_sync_action;
NeilBrownff4e8d92006-07-10 04:44:16 -0700355 sector_t curr_resync; /* last block scheduled */
NeilBrown97e4f422009-03-31 14:33:13 +1100356 /* As resync requests can complete out of order, we cannot easily track
357 * how much resync has been completed. So we occasionally pause until
358 * everything completes, then set curr_resync_completed to curr_resync.
359 * As such it may be well behind the real resync mark, but it is a value
360 * we are certain of.
361 */
362 sector_t curr_resync_completed;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363 unsigned long resync_mark; /* a recent timestamp */
364 sector_t resync_mark_cnt;/* blocks written at resync_mark */
NeilBrownff4e8d92006-07-10 04:44:16 -0700365 sector_t curr_mark_cnt; /* blocks scheduled now */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366
367 sector_t resync_max_sectors; /* may be set by personality */
NeilBrown9d888832005-11-08 21:39:26 -0800368
Jianpeng Ma7f7583d2012-10-11 14:17:59 +1100369 atomic64_t resync_mismatches; /* count of sectors where
NeilBrown9d888832005-11-08 21:39:26 -0800370 * parity/replica mismatch found
371 */
NeilBrowne464eaf2006-03-27 01:18:14 -0800372
373 /* allow user-space to request suspension of IO to regions of the array */
374 sector_t suspend_lo;
375 sector_t suspend_hi;
NeilBrown88202a02006-01-06 00:21:36 -0800376 /* if zero, use the system-wide default */
377 int sync_speed_min;
378 int sync_speed_max;
379
Bernd Schubert90b08712008-05-23 13:04:38 -0700380 /* resync even though the same disks are shared among md-devices */
381 int parallel_resync;
382
NeilBrown6ff8d8ec2006-01-06 00:20:15 -0800383 int ok_start_degraded;
NeilBrown5fd6c1d2006-06-26 00:27:40 -0700384
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385 unsigned long recovery;
NeilBrown53890422011-07-27 11:00:36 +1000386 /* If a RAID personality determines that recovery (of a particular
387 * device) will fail due to a read error on the source device, it
388 * takes a copy of this number and does not attempt recovery again
389 * until this number changes.
390 */
391 int recovery_disabled;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392
393 int in_sync; /* know to not need resync */
NeilBrownc8c00a62009-08-10 12:50:52 +1000394 /* 'open_mutex' avoids races between 'md_open' and 'do_md_stop', so
395 * that we are never stopping an array while it is open.
396 * 'reconfig_mutex' protects all other reconfiguration.
397 * These locks are separate due to conflicting interactions
398 * with bdev->bd_mutex.
399 * Lock ordering is:
Christoph Hellwigde090772020-09-01 17:57:48 +0200400 * reconfig_mutex -> bd_mutex
NeilBrownc8c00a62009-08-10 12:50:52 +1000401 * bd_mutex -> open_mutex: e.g. __blkdev_get -> md_open
402 */
403 struct mutex open_mutex;
NeilBrowndf5b89b2006-03-27 01:18:20 -0800404 struct mutex reconfig_mutex;
NeilBrownf2ea68c2008-07-21 17:05:25 +1000405 atomic_t active; /* general refcount */
406 atomic_t openers; /* number of active opens */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407
NeilBrownf0b4f7e2011-02-24 17:26:41 +1100408 int changed; /* True if we might need to
409 * reread partition info */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410 int degraded; /* whether md should consider
411 * adding a spare
412 */
413
414 atomic_t recovery_active; /* blocks scheduled, but not written */
415 wait_queue_head_t recovery_wait;
416 sector_t recovery_cp;
Neil Brown5e96ee62008-06-28 08:31:24 +1000417 sector_t resync_min; /* user requested sync
418 * starts here */
NeilBrownc6207272008-02-06 01:39:52 -0800419 sector_t resync_max; /* resync should pause
420 * when it gets here */
NeilBrown06d91a52005-06-21 17:17:12 -0700421
Tejun Heo324a56e2013-12-11 14:11:53 -0500422 struct kernfs_node *sysfs_state; /* handle for 'array_state'
NeilBrownb62b7592008-10-21 13:25:21 +1100423 * file in sysfs.
424 */
Tejun Heo324a56e2013-12-11 14:11:53 -0500425 struct kernfs_node *sysfs_action; /* handle for 'sync_action' */
Junxiao Bie1a86db2020-07-14 16:10:26 -0700426 struct kernfs_node *sysfs_completed; /*handle for 'sync_completed' */
427 struct kernfs_node *sysfs_degraded; /*handle for 'degraded' */
428 struct kernfs_node *sysfs_level; /*handle for 'level' */
NeilBrownb62b7592008-10-21 13:25:21 +1100429
NeilBrownd3374822009-01-09 08:31:10 +1100430 struct work_struct del_work; /* used for delayed sysfs removal */
431
NeilBrown85572d72014-12-15 12:56:56 +1100432 /* "lock" protects:
433 * flush_bio transition from NULL to !NULL
434 * rdev superblocks, events
435 * clearing MD_CHANGE_*
436 * in_sync - and related safemode and MD_CHANGE changes
NeilBrown36d091f2014-12-15 12:56:58 +1100437 * pers (also protected by reconfig_mutex and pending IO).
NeilBrown978a7a42014-12-15 12:56:58 +1100438 * clearing ->bitmap
NeilBrown4af1a042014-12-15 12:57:00 +1100439 * clearing ->bitmap_info.file
NeilBrown23da4222014-12-15 12:57:01 +1100440 * changing ->resync_{min,max}
441 * setting MD_RECOVERY_RUNNING (which interacts with resync_{min,max})
NeilBrown85572d72014-12-15 12:56:56 +1100442 */
443 spinlock_t lock;
NeilBrown3d310eb2005-06-21 17:17:26 -0700444 wait_queue_head_t sb_wait; /* for waiting on superblock updates */
NeilBrown7bfa19f2005-06-21 17:17:28 -0700445 atomic_t pending_writes; /* number of active superblock writes */
NeilBrown06d91a52005-06-21 17:17:12 -0700446
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 unsigned int safemode; /* if set, update "clean" superblock
448 * when no writes pending.
NeilBrownf72ffdd2014-09-30 14:23:59 +1000449 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450 unsigned int safemode_delay;
451 struct timer_list safemode_timer;
NeilBrown4ad23a972017-03-15 14:05:14 +1100452 struct percpu_ref writes_pending;
453 int sync_checkers; /* # of threads checking writes_pending */
Jens Axboe165125e2007-07-24 09:28:11 +0200454 struct request_queue *queue; /* for plugging ... */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455
NeilBrownf72ffdd2014-09-30 14:23:59 +1000456 struct bitmap *bitmap; /* the bitmap for the device */
NeilBrownc3d97142009-12-14 12:49:52 +1100457 struct {
458 struct file *file; /* the bitmap file */
NeilBrownf6af9492009-12-14 12:49:54 +1100459 loff_t offset; /* offset from superblock of
NeilBrownc3d97142009-12-14 12:49:52 +1100460 * start of bitmap. May be
461 * negative, but not '0'
NeilBrownf6af9492009-12-14 12:49:54 +1100462 * For external metadata, offset
NeilBrownf72ffdd2014-09-30 14:23:59 +1000463 * from start of device.
NeilBrownc3d97142009-12-14 12:49:52 +1100464 */
NeilBrown6409bb02012-05-22 13:55:07 +1000465 unsigned long space; /* space available at this offset */
NeilBrownf6af9492009-12-14 12:49:54 +1100466 loff_t default_offset; /* this is the offset to use when
NeilBrownc3d97142009-12-14 12:49:52 +1100467 * hot-adding a bitmap. It should
468 * eventually be settable by sysfs.
469 */
NeilBrown6409bb02012-05-22 13:55:07 +1000470 unsigned long default_space; /* space available at
471 * default offset */
NeilBrownc3d97142009-12-14 12:49:52 +1100472 struct mutex mutex;
NeilBrown42a04b52009-12-14 12:49:53 +1100473 unsigned long chunksize;
NeilBrownac2f40b2010-06-01 19:37:31 +1000474 unsigned long daemon_sleep; /* how many jiffies between updates? */
NeilBrown42a04b52009-12-14 12:49:53 +1100475 unsigned long max_write_behind; /* write-behind mode */
NeilBrownece5cff2009-12-14 12:49:56 +1100476 int external;
Goldwyn Rodriguesc4ce8672014-03-29 10:20:02 -0500477 int nodes; /* Maximum number of nodes in the cluster */
Goldwyn Rodriguescf921cc2014-03-30 00:42:49 -0500478 char cluster_name[64]; /* Name of the cluster */
NeilBrownc3d97142009-12-14 12:49:52 +1100479 } bitmap_info;
NeilBrown32a76272005-06-21 17:17:14 -0700480
NeilBrownf72ffdd2014-09-30 14:23:59 +1000481 atomic_t max_corr_read_errors; /* max read retries */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482 struct list_head all_mddevs;
NeilBrowna2826aa2009-12-14 12:49:49 +1100483
NeilBrowna64c8762010-04-14 17:15:37 +1000484 struct attribute_group *to_remove;
NeilBrown252ac522010-06-01 19:37:29 +1000485
Kent Overstreetafeee512018-05-20 18:25:52 -0400486 struct bio_set bio_set;
487 struct bio_set sync_set; /* for sync operations like
NeilBrown5a850712017-06-21 09:12:21 +1000488 * metadata and bitmap writes
489 */
Artur Paszkiewicz41d2d842020-07-03 11:13:09 +0200490 mempool_t md_io_pool;
NeilBrowna167f662010-10-26 18:31:13 +1100491
NeilBrown4bc034d2019-03-29 10:46:16 -0700492 /* Generic flush handling.
493 * The last to finish preflush schedules a worker to submit
494 * the rest of the request (without the REQ_PREFLUSH flag).
495 */
496 struct bio *flush_bio;
497 atomic_t flush_pending;
Pankaj Gupta81ba3c22020-11-11 06:16:56 +0100498 ktime_t start_flush, prev_flush_start; /* prev_flush_start is when the previous completed
499 * flush was started.
500 */
NeilBrown4bc034d2019-03-29 10:46:16 -0700501 struct work_struct flush_work;
NeilBrown768a4182010-07-26 11:49:55 +1000502 struct work_struct event_work; /* used by dm to report failure event */
Guoqing Jiang404659c2019-12-23 10:48:53 +0100503 mempool_t *serial_info_pool;
NeilBrownfd01b882011-10-11 16:47:53 +1100504 void (*sync_super)(struct mddev *mddev, struct md_rdev *rdev);
Goldwyn Rodriguesc4ce8672014-03-29 10:20:02 -0500505 struct md_cluster_info *cluster_info;
Guoqing Jiang15858fa2015-12-21 10:51:00 +1100506 unsigned int good_device_nr; /* good device num within cluster raid */
Coly Li78f57ef2020-04-09 22:17:20 +0800507 unsigned int noio_flag; /* for memalloc scope API */
Heinz Mauelshagen4b6c1062018-02-02 23:13:19 +0100508
509 bool has_superblocks:1;
Guoqing Jiang9a567842019-07-24 11:09:19 +0200510 bool fail_last_dev:1;
Guoqing Jiang3938f5f2019-12-23 10:48:56 +0100511 bool serialize_policy:1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512};
513
NeilBrownbe306c22016-11-09 10:21:33 +1100514enum recovery_flags {
515 /*
516 * If neither SYNC or RESHAPE are set, then it is a recovery.
517 */
518 MD_RECOVERY_RUNNING, /* a thread is running, or about to be started */
519 MD_RECOVERY_SYNC, /* actually doing a resync, not a recovery */
520 MD_RECOVERY_RECOVER, /* doing recovery, or need to try it. */
521 MD_RECOVERY_INTR, /* resync needs to be aborted for some reason */
522 MD_RECOVERY_DONE, /* thread is done and is waiting to be reaped */
523 MD_RECOVERY_NEEDED, /* we might need to start a resync/recover */
524 MD_RECOVERY_REQUESTED, /* user-space has requested a sync (used with SYNC) */
525 MD_RECOVERY_CHECK, /* user-space request for check-only, no repair */
526 MD_RECOVERY_RESHAPE, /* A reshape is happening */
527 MD_RECOVERY_FROZEN, /* User request to abort, and not restart, any action */
528 MD_RECOVERY_ERROR, /* sync-action interrupted because io-error */
Song Liud5d885f2017-11-19 22:17:01 -0800529 MD_RECOVERY_WAIT, /* waiting for pers->start() to finish */
Guoqing Jiang0357ba22018-07-02 16:26:25 +0800530 MD_RESYNCING_REMOTE, /* remote node is running resync thread */
NeilBrownbe306c22016-11-09 10:21:33 +1100531};
532
NeilBrown5c47daf2014-12-15 12:57:01 +1100533static inline int __must_check mddev_lock(struct mddev *mddev)
534{
535 return mutex_lock_interruptible(&mddev->reconfig_mutex);
536}
537
538/* Sometimes we need to take the lock in a situation where
539 * failure due to interrupts is not acceptable.
540 */
541static inline void mddev_lock_nointr(struct mddev *mddev)
542{
543 mutex_lock(&mddev->reconfig_mutex);
544}
545
NeilBrown5c47daf2014-12-15 12:57:01 +1100546static inline int mddev_trylock(struct mddev *mddev)
547{
548 return mutex_trylock(&mddev->reconfig_mutex);
549}
550extern void mddev_unlock(struct mddev *mddev);
551
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors)
553{
Christoph Hellwig4245e522020-09-03 07:40:59 +0200554 atomic_add(nr_sectors, &bdev->bd_disk->sync_io);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555}
556
Christoph Hellwig74d46992017-08-23 19:10:32 +0200557static inline void md_sync_acct_bio(struct bio *bio, unsigned long nr_sectors)
558{
Christoph Hellwig309dca302021-01-24 11:02:34 +0100559 md_sync_acct(bio->bi_bdev, nr_sectors);
Christoph Hellwig74d46992017-08-23 19:10:32 +0200560}
561
NeilBrown84fc4b52011-10-11 16:49:58 +1100562struct md_personality
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563{
564 char *name;
NeilBrown2604b702006-01-06 00:20:36 -0800565 int level;
566 struct list_head list;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567 struct module *owner;
David Jeffery775d7832019-09-16 13:15:14 -0400568 bool __must_check (*make_request)(struct mddev *mddev, struct bio *bio);
Song Liud5d885f2017-11-19 22:17:01 -0800569 /*
570 * start up works that do NOT require md_thread. tasks that
571 * requires md_thread should go into start()
572 */
NeilBrownfd01b882011-10-11 16:47:53 +1100573 int (*run)(struct mddev *mddev);
Song Liud5d885f2017-11-19 22:17:01 -0800574 /* start up works that require md threads */
575 int (*start)(struct mddev *mddev);
NeilBrownafa0f552014-12-15 12:56:58 +1100576 void (*free)(struct mddev *mddev, void *priv);
NeilBrownfd01b882011-10-11 16:47:53 +1100577 void (*status)(struct seq_file *seq, struct mddev *mddev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578 /* error_handler must set ->faulty and clear ->in_sync
NeilBrownf72ffdd2014-09-30 14:23:59 +1000579 * if appropriate, and should abort recovery if needed
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580 */
NeilBrownfd01b882011-10-11 16:47:53 +1100581 void (*error_handler)(struct mddev *mddev, struct md_rdev *rdev);
582 int (*hot_add_disk) (struct mddev *mddev, struct md_rdev *rdev);
NeilBrownb8321b62011-12-23 10:17:51 +1100583 int (*hot_remove_disk) (struct mddev *mddev, struct md_rdev *rdev);
NeilBrownfd01b882011-10-11 16:47:53 +1100584 int (*spare_active) (struct mddev *mddev);
NeilBrown09314792015-02-19 16:04:40 +1100585 sector_t (*sync_request)(struct mddev *mddev, sector_t sector_nr, int *skipped);
NeilBrownfd01b882011-10-11 16:47:53 +1100586 int (*resize) (struct mddev *mddev, sector_t sectors);
587 sector_t (*size) (struct mddev *mddev, sector_t sectors, int raid_disks);
588 int (*check_reshape) (struct mddev *mddev);
589 int (*start_reshape) (struct mddev *mddev);
590 void (*finish_reshape) (struct mddev *mddev);
Guoqing Jiang7564bed2018-10-18 16:37:42 +0800591 void (*update_reshape_pos) (struct mddev *mddev);
NeilBrownb03e0cc2017-10-19 12:49:15 +1100592 /* quiesce suspends or resumes internal processing.
593 * 1 - stop new actions and wait for action io to complete
594 * 0 - return to normal behaviour
NeilBrown36fa3062005-09-09 16:23:45 -0700595 */
NeilBrownb03e0cc2017-10-19 12:49:15 +1100596 void (*quiesce) (struct mddev *mddev, int quiesce);
NeilBrown245f46c2009-03-31 14:39:39 +1100597 /* takeover is used to transition an array from one
598 * personality to another. The new personality must be able
599 * to handle the data in the current layout.
600 * e.g. 2drive raid1 -> 2drive raid5
601 * ndrive raid5 -> degraded n+1drive raid6 with special layout
602 * If the takeover succeeds, a new 'private' structure is returned.
603 * This needs to be installed and then ->run used to activate the
604 * array.
605 */
NeilBrownfd01b882011-10-11 16:47:53 +1100606 void *(*takeover) (struct mddev *mddev);
Artur Paszkiewiczba903a32017-03-09 10:00:03 +0100607 /* Changes the consistency policy of an active array. */
608 int (*change_consistency_policy)(struct mddev *mddev, const char *buf);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609};
610
NeilBrown007583c2005-11-08 21:39:30 -0800611struct md_sysfs_entry {
612 struct attribute attr;
NeilBrownfd01b882011-10-11 16:47:53 +1100613 ssize_t (*show)(struct mddev *, char *);
614 ssize_t (*store)(struct mddev *, const char *, size_t);
NeilBrown007583c2005-11-08 21:39:30 -0800615};
NeilBrown43a70502009-12-14 12:49:55 +1100616extern struct attribute_group md_bitmap_group;
NeilBrown007583c2005-11-08 21:39:30 -0800617
Tejun Heo324a56e2013-12-11 14:11:53 -0500618static inline struct kernfs_node *sysfs_get_dirent_safe(struct kernfs_node *sd, char *name)
NeilBrown00bcb4a2010-06-01 19:37:23 +1000619{
620 if (sd)
Tejun Heo388975c2013-09-11 23:19:13 -0400621 return sysfs_get_dirent(sd, name);
NeilBrown00bcb4a2010-06-01 19:37:23 +1000622 return sd;
623}
Tejun Heo324a56e2013-12-11 14:11:53 -0500624static inline void sysfs_notify_dirent_safe(struct kernfs_node *sd)
NeilBrown00bcb4a2010-06-01 19:37:23 +1000625{
626 if (sd)
627 sysfs_notify_dirent(sd);
628}
629
NeilBrownfd01b882011-10-11 16:47:53 +1100630static inline char * mdname (struct mddev * mddev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631{
632 return mddev->gendisk ? mddev->gendisk->disk_name : "mdX";
633}
634
NeilBrownfd01b882011-10-11 16:47:53 +1100635static inline int sysfs_link_rdev(struct mddev *mddev, struct md_rdev *rdev)
Namhyung Kim36fad852011-07-27 11:00:36 +1000636{
637 char nm[20];
Shaohua Li9b156032015-12-18 15:19:16 +1100638 if (!test_bit(Replacement, &rdev->flags) &&
639 !test_bit(Journal, &rdev->flags) &&
640 mddev->kobj.sd) {
NeilBrown2d78f8c2011-12-23 10:17:51 +1100641 sprintf(nm, "rd%d", rdev->raid_disk);
642 return sysfs_create_link(&mddev->kobj, &rdev->kobj, nm);
643 } else
644 return 0;
Namhyung Kim36fad852011-07-27 11:00:36 +1000645}
646
NeilBrownfd01b882011-10-11 16:47:53 +1100647static inline void sysfs_unlink_rdev(struct mddev *mddev, struct md_rdev *rdev)
Namhyung Kim36fad852011-07-27 11:00:36 +1000648{
649 char nm[20];
Shaohua Li9b156032015-12-18 15:19:16 +1100650 if (!test_bit(Replacement, &rdev->flags) &&
651 !test_bit(Journal, &rdev->flags) &&
652 mddev->kobj.sd) {
NeilBrown2d78f8c2011-12-23 10:17:51 +1100653 sprintf(nm, "rd%d", rdev->raid_disk);
654 sysfs_remove_link(&mddev->kobj, nm);
655 }
Namhyung Kim36fad852011-07-27 11:00:36 +1000656}
657
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658/*
659 * iterates through some rdev ringlist. It's safe to remove the
660 * current 'rdev'. Dont touch 'tmp' though.
661 */
Cheng Renquan159ec1f2009-01-09 08:31:08 +1100662#define rdev_for_each_list(rdev, tmp, head) \
663 list_for_each_entry_safe(rdev, tmp, head, same_set)
664
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665/*
666 * iterates through the 'same array disks' ringlist
667 */
NeilBrowndafb20f2012-03-19 12:46:39 +1100668#define rdev_for_each(rdev, mddev) \
669 list_for_each_entry(rdev, &((mddev)->disks), same_set)
670
671#define rdev_for_each_safe(rdev, tmp, mddev) \
Cheng Renquan159ec1f2009-01-09 08:31:08 +1100672 list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673
NeilBrown4b809912008-07-21 17:05:25 +1000674#define rdev_for_each_rcu(rdev, mddev) \
675 list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set)
676
NeilBrown2b8bf342011-10-11 16:48:23 +1100677struct md_thread {
Shaohua Li4ed87312012-10-11 13:34:00 +1100678 void (*run) (struct md_thread *thread);
NeilBrownfd01b882011-10-11 16:47:53 +1100679 struct mddev *mddev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680 wait_queue_head_t wqueue;
NeilBrownf72ffdd2014-09-30 14:23:59 +1000681 unsigned long flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682 struct task_struct *tsk;
NeilBrown32a76272005-06-21 17:17:14 -0700683 unsigned long timeout;
Shaohua Li4ed87312012-10-11 13:34:00 +1100684 void *private;
NeilBrown2b8bf342011-10-11 16:48:23 +1100685};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686
687#define THREAD_WAKEUP 0
688
NeilBrown1345b1d2006-01-06 00:20:40 -0800689static inline void safe_put_page(struct page *p)
690{
691 if (p) put_page(p);
692}
693
NeilBrown84fc4b52011-10-11 16:49:58 +1100694extern int register_md_personality(struct md_personality *p);
695extern int unregister_md_personality(struct md_personality *p);
Goldwyn Rodriguesedb39c92014-03-29 10:01:53 -0500696extern int register_md_cluster_operations(struct md_cluster_operations *ops,
697 struct module *module);
698extern int unregister_md_cluster_operations(void);
699extern int md_setup_cluster(struct mddev *mddev, int nodes);
700extern void md_cluster_stop(struct mddev *mddev);
NeilBrown2b8bf342011-10-11 16:48:23 +1100701extern struct md_thread *md_register_thread(
Shaohua Li4ed87312012-10-11 13:34:00 +1100702 void (*run)(struct md_thread *thread),
NeilBrown2b8bf342011-10-11 16:48:23 +1100703 struct mddev *mddev,
704 const char *name);
705extern void md_unregister_thread(struct md_thread **threadp);
706extern void md_wakeup_thread(struct md_thread *thread);
NeilBrownfd01b882011-10-11 16:47:53 +1100707extern void md_check_recovery(struct mddev *mddev);
Jonathan Brassowa91d5ac2013-04-24 11:42:43 +1000708extern void md_reap_sync_thread(struct mddev *mddev);
NeilBrowna415c0f2017-06-05 16:05:13 +1000709extern int mddev_init_writes_pending(struct mddev *mddev);
NeilBrowncc27b0c2017-06-05 16:49:39 +1000710extern bool md_write_start(struct mddev *mddev, struct bio *bi);
NeilBrown49728052017-03-15 14:05:12 +1100711extern void md_write_inc(struct mddev *mddev, struct bio *bi);
NeilBrownfd01b882011-10-11 16:47:53 +1100712extern void md_write_end(struct mddev *mddev);
713extern void md_done_sync(struct mddev *mddev, int blocks, int ok);
714extern void md_error(struct mddev *mddev, struct md_rdev *rdev);
NeilBrownc6563a82012-05-21 09:27:00 +1000715extern void md_finish_reshape(struct mddev *mddev);
Xiao Nicf784082021-02-04 15:50:43 +0800716void md_submit_discard_bio(struct mddev *mddev, struct md_rdev *rdev,
717 struct bio *bio, sector_t start, sector_t size);
NeilBrown92022952009-03-31 14:33:13 +1100718
David Jeffery775d7832019-09-16 13:15:14 -0400719extern bool __must_check md_flush_request(struct mddev *mddev, struct bio *bio);
NeilBrownfd01b882011-10-11 16:47:53 +1100720extern void md_super_write(struct mddev *mddev, struct md_rdev *rdev,
NeilBrown92022952009-03-31 14:33:13 +1100721 sector_t sector, int size, struct page *page);
NeilBrown46533ff2016-11-18 16:16:11 +1100722extern int md_super_wait(struct mddev *mddev);
NeilBrownf72ffdd2014-09-30 14:23:59 +1000723extern int sync_page_io(struct md_rdev *rdev, sector_t sector, int size,
Mike Christie796a5cf2016-06-05 14:32:07 -0500724 struct page *page, int op, int op_flags,
725 bool metadata_op);
Shaohua Li4ed87312012-10-11 13:34:00 +1100726extern void md_do_sync(struct md_thread *thread);
NeilBrownfd01b882011-10-11 16:47:53 +1100727extern void md_new_event(struct mddev *mddev);
Artur Paszkiewicz2214c262017-05-08 11:56:55 +0200728extern void md_allow_write(struct mddev *mddev);
NeilBrownfd01b882011-10-11 16:47:53 +1100729extern void md_wait_for_blocked_rdev(struct md_rdev *rdev, struct mddev *mddev);
730extern void md_set_array_sectors(struct mddev *mddev, sector_t array_sectors);
731extern int md_check_no_bitmap(struct mddev *mddev);
732extern int md_integrity_register(struct mddev *mddev);
Dan Williams1501efa2016-01-13 16:00:07 -0800733extern int md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev);
NeilBrown72e02072009-12-14 12:49:55 +1100734extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale);
Christoph Hellwig63fe0812009-04-14 12:01:53 +1000735
NeilBrownfd01b882011-10-11 16:47:53 +1100736extern void mddev_init(struct mddev *mddev);
737extern int md_run(struct mddev *mddev);
Song Liud5d885f2017-11-19 22:17:01 -0800738extern int md_start(struct mddev *mddev);
NeilBrownfd01b882011-10-11 16:47:53 +1100739extern void md_stop(struct mddev *mddev);
740extern void md_stop_writes(struct mddev *mddev);
NeilBrown3cb03002011-10-11 16:45:26 +1100741extern int md_rdev_init(struct md_rdev *rdev);
NeilBrown545c8792012-05-22 13:54:30 +1000742extern void md_rdev_clear(struct md_rdev *rdev);
NeilBrown390ee602010-06-01 19:37:27 +1000743
Shaohua Li393debc2017-09-21 10:23:35 -0700744extern void md_handle_request(struct mddev *mddev, struct bio *bio);
NeilBrownfd01b882011-10-11 16:47:53 +1100745extern void mddev_suspend(struct mddev *mddev);
746extern void mddev_resume(struct mddev *mddev);
NeilBrown9cbb1752012-07-31 09:08:14 +0200747
Goldwyn Rodrigues70bcecd2015-08-21 10:33:39 -0500748extern void md_reload_sb(struct mddev *mddev, int raid_disk);
Goldwyn Rodrigues1aee41f2014-10-29 18:51:31 -0500749extern void md_update_sb(struct mddev *mddev, int force);
Goldwyn Rodriguesfb56dfe2015-04-14 10:43:24 -0500750extern void md_kick_rdev_from_array(struct md_rdev * rdev);
Guoqing Jiang404659c2019-12-23 10:48:53 +0100751extern void mddev_create_serial_pool(struct mddev *mddev, struct md_rdev *rdev,
Guoqing Jiang11d3a9f2019-12-23 10:48:55 +0100752 bool is_suspend);
Guoqing Jiang69b00b52019-12-23 10:49:00 +0100753extern void mddev_destroy_serial_pool(struct mddev *mddev, struct md_rdev *rdev,
754 bool is_suspend);
Goldwyn Rodrigues57d051d2015-04-14 10:43:55 -0500755struct md_rdev *md_find_rdev_nr_rcu(struct mddev *mddev, int nr);
Tomasz Majchrzak1532d9e2017-12-27 10:31:40 +0100756struct md_rdev *md_find_rdev_rcu(struct mddev *mddev, dev_t dev);
Hannes Reineckedfe15ac2012-07-26 11:12:18 +0200757
Guilherme G. Piccoli62f7b192019-09-03 16:49:00 -0300758static inline bool is_mddev_broken(struct md_rdev *rdev, const char *md_type)
759{
760 int flags = rdev->bdev->bd_disk->flags;
761
762 if (!(flags & GENHD_FL_UP)) {
763 if (!test_and_set_bit(MD_BROKEN, &rdev->mddev->flags))
764 pr_warn("md: %s: %s array has a missing/failed member\n",
765 mdname(rdev->mddev), md_type);
766 return true;
767 }
768 return false;
769}
770
Hannes Reineckedfe15ac2012-07-26 11:12:18 +0200771static inline void rdev_dec_pending(struct md_rdev *rdev, struct mddev *mddev)
772{
773 int faulty = test_bit(Faulty, &rdev->flags);
774 if (atomic_dec_and_test(&rdev->nr_pending) && faulty) {
775 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
776 md_wakeup_thread(mddev->thread);
777 }
778}
779
Goldwyn Rodriguesedb39c92014-03-29 10:01:53 -0500780extern struct md_cluster_operations *md_cluster_ops;
Goldwyn Rodriguesc4ce8672014-03-29 10:20:02 -0500781static inline int mddev_is_clustered(struct mddev *mddev)
782{
783 return mddev->cluster_info && mddev->bitmap_info.nodes > 1;
784}
Shaohua Li394ed8e2017-01-04 16:10:19 -0800785
786/* clear unsupported mddev_flags */
787static inline void mddev_clear_unsupported_flags(struct mddev *mddev,
788 unsigned long unsupported_flags)
789{
790 mddev->flags &= ~unsupported_flags;
791}
Shaohua Li26483812017-02-13 16:21:49 -0800792
793static inline void mddev_check_writesame(struct mddev *mddev, struct bio *bio)
794{
795 if (bio_op(bio) == REQ_OP_WRITE_SAME &&
Christoph Hellwig309dca302021-01-24 11:02:34 +0100796 !bio->bi_bdev->bd_disk->queue->limits.max_write_same_sectors)
Shaohua Li26483812017-02-13 16:21:49 -0800797 mddev->queue->limits.max_write_same_sectors = 0;
798}
Christoph Hellwig3deff1a2017-04-05 19:21:03 +0200799
800static inline void mddev_check_write_zeroes(struct mddev *mddev, struct bio *bio)
801{
802 if (bio_op(bio) == REQ_OP_WRITE_ZEROES &&
Christoph Hellwig309dca302021-01-24 11:02:34 +0100803 !bio->bi_bdev->bd_disk->queue->limits.max_write_zeroes_sectors)
Christoph Hellwig3deff1a2017-04-05 19:21:03 +0200804 mddev->queue->limits.max_write_zeroes_sectors = 0;
805}
Christoph Hellwigd82fa812020-06-06 15:00:24 +0200806
Christoph Hellwig7e0adbf2020-06-07 17:31:19 +0200807struct mdu_array_info_s;
808struct mdu_disk_info_s;
809
Christoph Hellwig1a6a0502020-06-07 16:33:01 +0200810extern int mdp_major;
Christoph Hellwigd82fa812020-06-06 15:00:24 +0200811void md_autostart_arrays(int part);
Christoph Hellwig7e0adbf2020-06-07 17:31:19 +0200812int md_set_array_info(struct mddev *mddev, struct mdu_array_info_s *info);
813int md_add_new_disk(struct mddev *mddev, struct mdu_disk_info_s *info);
814int do_md_run(struct mddev *mddev);
815
816extern const struct block_device_operations md_fops;
Christoph Hellwigd82fa812020-06-06 15:00:24 +0200817
Christoph Hellwig63fe0812009-04-14 12:01:53 +1000818#endif /* _MD_MD_H */