Thomas Gleixner | af1a889 | 2019-05-20 19:08:12 +0200 | [diff] [blame] | 1 | /* SPDX-License-Identifier: GPL-2.0-or-later */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2 | /* |
Wang Sheng-Hui | 7e84152 | 2011-09-21 15:37:46 +1000 | [diff] [blame] | 3 | md.h : kernel internal structure of the Linux MD driver |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4 | Copyright (C) 1996-98 Ingo Molnar, Gadi Oxman |
NeilBrown | f72ffdd | 2014-09-30 14:23:59 +1000 | [diff] [blame] | 5 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 6 | */ |
| 7 | |
Christoph Hellwig | 63fe081 | 2009-04-14 12:01:53 +1000 | [diff] [blame] | 8 | #ifndef _MD_MD_H |
| 9 | #define _MD_MD_H |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 10 | |
Christoph Hellwig | 63fe081 | 2009-04-14 12:01:53 +1000 | [diff] [blame] | 11 | #include <linux/blkdev.h> |
Tejun Heo | 66114ca | 2015-05-22 17:13:32 -0400 | [diff] [blame] | 12 | #include <linux/backing-dev.h> |
Vishal Verma | fc974ee | 2015-12-24 19:20:34 -0700 | [diff] [blame] | 13 | #include <linux/badblocks.h> |
Christoph Hellwig | 63fe081 | 2009-04-14 12:01:53 +1000 | [diff] [blame] | 14 | #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 Rodrigues | edb39c9 | 2014-03-29 10:01:53 -0500 | [diff] [blame] | 21 | #include "md-cluster.h" |
David Howells | 9361401 | 2006-09-30 20:45:40 +0200 | [diff] [blame] | 22 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 23 | #define MaxSector (~(sector_t)0) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 24 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 25 | /* |
NeilBrown | 46533ff | 2016-11-18 16:16:11 +1100 | [diff] [blame] | 26 | * 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 Jiang | 69b00b5 | 2019-12-23 10:49:00 +0100 | [diff] [blame] | 35 | |
| 36 | /* |
| 37 | * The struct embedded in rdev is used to serialize IO. |
| 38 | */ |
| 39 | struct 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 | |
NeilBrown | 46533ff | 2016-11-18 16:16:11 +1100 | [diff] [blame] | 45 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 46 | * MD's 'extended' device |
| 47 | */ |
NeilBrown | 3cb0300 | 2011-10-11 16:45:26 +1100 | [diff] [blame] | 48 | struct md_rdev { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 49 | struct list_head same_set; /* RAID devices within the same set */ |
| 50 | |
Andre Noll | dd8ac33 | 2009-03-31 14:33:13 +1100 | [diff] [blame] | 51 | sector_t sectors; /* Device size (in 512bytes sectors) */ |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 52 | struct mddev *mddev; /* RAID array if running */ |
NeilBrown | eea1bf3 | 2009-03-31 14:27:02 +1100 | [diff] [blame] | 53 | int last_events; /* IO event timestamp */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 54 | |
Jonathan Brassow | a6ff7e0 | 2011-01-14 09:14:34 +1100 | [diff] [blame] | 55 | /* |
| 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 Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 61 | struct block_device *bdev; /* block device handle */ |
| 62 | |
NeilBrown | 2699b67 | 2011-07-28 11:31:47 +1000 | [diff] [blame] | 63 | struct page *sb_page, *bb_page; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 64 | int sb_loaded; |
NeilBrown | 4254376 | 2006-06-26 00:27:57 -0700 | [diff] [blame] | 65 | __u64 sb_events; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 66 | sector_t data_offset; /* start of data in array */ |
NeilBrown | c6563a8 | 2012-05-21 09:27:00 +1000 | [diff] [blame] | 67 | sector_t new_data_offset;/* only relevant while reshaping */ |
NeilBrown | f72ffdd | 2014-09-30 14:23:59 +1000 | [diff] [blame] | 68 | sector_t sb_start; /* offset of the super block (in 512byte sectors) */ |
NeilBrown | 0002b27 | 2005-09-09 16:23:53 -0700 | [diff] [blame] | 69 | int sb_size; /* bytes in the superblock */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 70 | int preferred_minor; /* autorun support */ |
| 71 | |
NeilBrown | 86e6ffd | 2005-11-08 21:39:24 -0800 | [diff] [blame] | 72 | struct kobject kobj; |
| 73 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 74 | /* 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 Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 84 | |
NeilBrown | 2d78f8c | 2011-12-23 10:17:51 +1100 | [diff] [blame] | 85 | unsigned long flags; /* bit set of 'enum flag_bits' bits. */ |
Dan Williams | 6bfe0b4 | 2008-04-30 00:52:32 -0700 | [diff] [blame] | 86 | wait_queue_head_t blocked_wait; |
NeilBrown | 8ddf9ef | 2005-09-09 16:23:45 -0700 | [diff] [blame] | 87 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 88 | int desc_nr; /* descriptor index in the superblock */ |
| 89 | int raid_disk; /* role of device in array */ |
NeilBrown | e93f68a | 2010-06-15 09:36:03 +0100 | [diff] [blame] | 90 | int new_raid_disk; /* role that the device will have in |
| 91 | * the array after a level-change completes. |
| 92 | */ |
NeilBrown | 41158c7 | 2005-06-21 17:17:25 -0700 | [diff] [blame] | 93 | 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 Li | 3069aa8 | 2015-08-13 14:31:56 -0700 | [diff] [blame] | 97 | union { |
| 98 | sector_t recovery_offset;/* If this device has been partially |
NeilBrown | 5fd6c1d | 2006-06-26 00:27:40 -0700 | [diff] [blame] | 99 | * recovered, this is where we were |
| 100 | * up to. |
| 101 | */ |
Shaohua Li | 3069aa8 | 2015-08-13 14:31:56 -0700 | [diff] [blame] | 102 | 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 Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 107 | |
| 108 | atomic_t nr_pending; /* number of pending requests. |
| 109 | * only maintained for arrays that |
| 110 | * support hot removal |
| 111 | */ |
NeilBrown | ba22dcb | 2005-11-08 21:39:31 -0800 | [diff] [blame] | 112 | atomic_t read_errors; /* number of consecutive read errors that |
| 113 | * we have tried to ignore. |
| 114 | */ |
Arnd Bergmann | 0e3ef49 | 2016-06-17 17:33:10 +0200 | [diff] [blame] | 115 | time64_t last_read_error; /* monotonic time since our |
Robert Becker | 1e50915 | 2009-12-14 12:49:58 +1100 | [diff] [blame] | 116 | * last read error |
| 117 | */ |
NeilBrown | 4dbcdc7 | 2006-01-06 00:20:52 -0800 | [diff] [blame] | 118 | atomic_t corrected_errors; /* number of corrected read errors, |
| 119 | * for reporting to userspace and storing |
| 120 | * in superblock. |
| 121 | */ |
Guoqing Jiang | 3e148a3 | 2019-06-19 17:30:46 +0800 | [diff] [blame] | 122 | |
Guoqing Jiang | 69b00b5 | 2019-12-23 10:49:00 +0100 | [diff] [blame] | 123 | struct serial_in_rdev *serial; /* used for raid1 io serialization */ |
Guoqing Jiang | 3e148a3 | 2019-06-19 17:30:46 +0800 | [diff] [blame] | 124 | |
NeilBrown | 5792a28 | 2007-04-04 19:08:18 -0700 | [diff] [blame] | 125 | struct work_struct del_work; /* used for delayed sysfs removal */ |
NeilBrown | 3c0ee63 | 2008-10-21 13:25:28 +1100 | [diff] [blame] | 126 | |
Tejun Heo | 324a56e | 2013-12-11 14:11:53 -0500 | [diff] [blame] | 127 | struct kernfs_node *sysfs_state; /* handle for 'state' |
NeilBrown | 3c0ee63 | 2008-10-21 13:25:28 +1100 | [diff] [blame] | 128 | * sysfs entry */ |
Junxiao Bi | e1a86db | 2020-07-14 16:10:26 -0700 | [diff] [blame] | 129 | /* 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 Verma | fc974ee | 2015-12-24 19:20:34 -0700 | [diff] [blame] | 133 | struct badblocks badblocks; |
Artur Paszkiewicz | ea0213e | 2017-03-09 09:59:57 +0100 | [diff] [blame] | 134 | |
| 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 Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 141 | }; |
NeilBrown | 2d78f8c | 2011-12-23 10:17:51 +1100 | [diff] [blame] | 142 | enum flag_bits { |
| 143 | Faulty, /* device is known to have a fault */ |
| 144 | In_sync, /* device is in_sync with rest of array */ |
NeilBrown | 8313b8e | 2013-12-12 10:13:33 +1100 | [diff] [blame] | 145 | Bitmap_sync, /* ..actually, not quite In_sync. Need a |
Guoqing Jiang | 4aaf7694 | 2017-07-04 11:20:30 +0800 | [diff] [blame] | 146 | * 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. |
NeilBrown | 8313b8e | 2013-12-12 10:13:33 +1100 | [diff] [blame] | 149 | */ |
NeilBrown | 2d78f8c | 2011-12-23 10:17:51 +1100 | [diff] [blame] | 150 | 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 Rodrigues | 1aee41f | 2014-10-29 18:51:31 -0500 | [diff] [blame] | 184 | Candidate, /* For clustered environments only: |
| 185 | * This device is seen locally but not |
| 186 | * by the whole cluster |
| 187 | */ |
Song Liu | bac624f | 2015-08-13 14:31:55 -0700 | [diff] [blame] | 188 | 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 Jiang | 659b254 | 2015-12-21 10:50:59 +1100 | [diff] [blame] | 193 | ClusterRemove, |
NeilBrown | d787be4 | 2016-06-02 16:19:53 +1000 | [diff] [blame] | 194 | 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 Majchrzak | 35b785f | 2016-10-21 16:26:57 +0200 | [diff] [blame] | 199 | ExternalBbl, /* External metadata provides bad |
| 200 | * block management for a disk |
| 201 | */ |
NeilBrown | 688834e | 2016-11-18 16:16:11 +1100 | [diff] [blame] | 202 | 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 | */ |
NeilBrown | 46533ff | 2016-11-18 16:16:11 +1100 | [diff] [blame] | 208 | 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 Jiang | 404659c | 2019-12-23 10:48:53 +0100 | [diff] [blame] | 212 | CollisionCheck, /* |
| 213 | * check if there is collision between raid1 |
| 214 | * serial bios. |
Guoqing Jiang | 3e148a3 | 2019-06-19 17:30:46 +0800 | [diff] [blame] | 215 | */ |
NeilBrown | 2d78f8c | 2011-12-23 10:17:51 +1100 | [diff] [blame] | 216 | }; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 217 | |
NeilBrown | 3cb0300 | 2011-10-11 16:45:26 +1100 | [diff] [blame] | 218 | static inline int is_badblock(struct md_rdev *rdev, sector_t s, int sectors, |
NeilBrown | 2230dfe | 2011-07-28 11:31:46 +1000 | [diff] [blame] | 219 | sector_t *first_bad, int *bad_sectors) |
| 220 | { |
| 221 | if (unlikely(rdev->badblocks.count)) { |
Vishal Verma | fc974ee | 2015-12-24 19:20:34 -0700 | [diff] [blame] | 222 | int rv = badblocks_check(&rdev->badblocks, rdev->data_offset + s, |
NeilBrown | 2230dfe | 2011-07-28 11:31:46 +1000 | [diff] [blame] | 223 | sectors, |
| 224 | first_bad, bad_sectors); |
| 225 | if (rv) |
| 226 | *first_bad -= rdev->data_offset; |
| 227 | return rv; |
| 228 | } |
| 229 | return 0; |
| 230 | } |
NeilBrown | 3cb0300 | 2011-10-11 16:45:26 +1100 | [diff] [blame] | 231 | extern int rdev_set_badblocks(struct md_rdev *rdev, sector_t s, int sectors, |
NeilBrown | c6563a8 | 2012-05-21 09:27:00 +1000 | [diff] [blame] | 232 | int is_new); |
| 233 | extern int rdev_clear_badblocks(struct md_rdev *rdev, sector_t s, int sectors, |
| 234 | int is_new); |
Goldwyn Rodrigues | c4ce867 | 2014-03-29 10:20:02 -0500 | [diff] [blame] | 235 | struct md_cluster_info; |
| 236 | |
Shaohua Li | 394ed8e | 2017-01-04 16:10:19 -0800 | [diff] [blame] | 237 | /* change UNSUPPORTED_MDDEV_FLAGS for each array type if new flag is added */ |
NeilBrown | be306c2 | 2016-11-09 10:21:33 +1100 | [diff] [blame] | 238 | enum mddev_flags { |
NeilBrown | be306c2 | 2016-11-09 10:21:33 +1100 | [diff] [blame] | 239 | 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 */ |
NeilBrown | be306c2 | 2016-11-09 10:21:33 +1100 | [diff] [blame] | 244 | MD_CLUSTER_RESYNC_LOCKED, /* cluster raid only, which means node |
| 245 | * already took resync lock, need to |
| 246 | * release the lock */ |
NeilBrown | 46533ff | 2016-11-18 16:16:11 +1100 | [diff] [blame] | 247 | 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 Paszkiewicz | ea0213e | 2017-03-09 09:59:57 +0100 | [diff] [blame] | 251 | MD_HAS_PPL, /* The raid array has PPL feature set */ |
Pawel Baldysiak | ddc0882 | 2017-08-16 17:13:45 +0200 | [diff] [blame] | 252 | MD_HAS_MULTIPLE_PPLS, /* The raid array has multiple PPLs feature set */ |
NeilBrown | 35bfc52 | 2017-10-17 13:46:43 +1100 | [diff] [blame] | 253 | 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 | */ |
NeilBrown | 9d4b45d | 2019-08-20 10:21:09 +1000 | [diff] [blame] | 259 | MD_NOT_READY, /* do_md_run() is active, so 'array_state' |
| 260 | * must not report that array is ready yet |
| 261 | */ |
Guilherme G. Piccoli | 62f7b19 | 2019-09-03 16:49:00 -0300 | [diff] [blame] | 262 | MD_BROKEN, /* This is used in RAID-0/LINEAR only, to stop |
| 263 | * I/O in case an array member is gone/failed. |
| 264 | */ |
NeilBrown | be306c2 | 2016-11-09 10:21:33 +1100 | [diff] [blame] | 265 | }; |
Shaohua Li | 2953079 | 2016-12-08 15:48:19 -0800 | [diff] [blame] | 266 | |
| 267 | enum 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 Jiang | 404659c | 2019-12-23 10:48:53 +0100 | [diff] [blame] | 274 | #define NR_SERIAL_INFOS 8 |
| 275 | /* record current range of serialize IOs */ |
| 276 | struct serial_info { |
Guoqing Jiang | 69b00b5 | 2019-12-23 10:49:00 +0100 | [diff] [blame] | 277 | 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 Jiang | 3e148a3 | 2019-06-19 17:30:46 +0800 | [diff] [blame] | 281 | }; |
| 282 | |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 283 | struct mddev { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 284 | void *private; |
NeilBrown | 84fc4b5 | 2011-10-11 16:49:58 +1100 | [diff] [blame] | 285 | struct md_personality *pers; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 286 | dev_t unit; |
| 287 | int md_minor; |
NeilBrown | 7a0a535 | 2013-08-27 16:28:23 +1000 | [diff] [blame] | 288 | struct list_head disks; |
NeilBrown | 850b2b42 | 2006-10-03 01:15:46 -0700 | [diff] [blame] | 289 | unsigned long flags; |
Shaohua Li | 2953079 | 2016-12-08 15:48:19 -0800 | [diff] [blame] | 290 | unsigned long sb_flags; |
NeilBrown | 850b2b42 | 2006-10-03 01:15:46 -0700 | [diff] [blame] | 291 | |
NeilBrown | 409c57f | 2009-03-31 14:39:39 +1100 | [diff] [blame] | 292 | int suspended; |
| 293 | atomic_t active_io; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 294 | int ro; |
NeilBrown | bb4f1e9 | 2010-08-08 21:18:03 +1000 | [diff] [blame] | 295 | int sysfs_active; /* set when sysfs deletes |
| 296 | * are happening, so run/ |
| 297 | * takeover/stop are not safe |
| 298 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 299 | struct gendisk *gendisk; |
| 300 | |
NeilBrown | eae1701 | 2005-11-08 21:39:23 -0800 | [diff] [blame] | 301 | struct kobject kobj; |
NeilBrown | d337482 | 2009-01-09 08:31:10 +1100 | [diff] [blame] | 302 | int hold_active; |
| 303 | #define UNTIL_IOCTL 1 |
NeilBrown | efeb53c | 2009-01-09 08:31:10 +1100 | [diff] [blame] | 304 | #define UNTIL_STOP 2 |
NeilBrown | eae1701 | 2005-11-08 21:39:23 -0800 | [diff] [blame] | 305 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 306 | /* Superblock information */ |
| 307 | int major_version, |
| 308 | minor_version, |
| 309 | patch_version; |
| 310 | int persistent; |
NeilBrown | f72ffdd | 2014-09-30 14:23:59 +1000 | [diff] [blame] | 311 | int external; /* metadata is |
NeilBrown | e691063 | 2008-02-06 01:39:51 -0800 | [diff] [blame] | 312 | * managed externally */ |
| 313 | char metadata_type[17]; /* externally set*/ |
Mike Snitzer | 77a6869 | 2020-12-14 12:08:48 -0500 | [diff] [blame] | 314 | int chunk_sectors; |
Deepa Dinamani | 9ebc6ef | 2015-12-21 10:51:01 +1100 | [diff] [blame] | 315 | time64_t ctime, utime; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 316 | int level, layout; |
NeilBrown | d9d166c | 2006-01-06 00:20:51 -0800 | [diff] [blame] | 317 | char clevel[16]; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 318 | int raid_disks; |
| 319 | int max_disks; |
NeilBrown | f72ffdd | 2014-09-30 14:23:59 +1000 | [diff] [blame] | 320 | sector_t dev_sectors; /* used size of |
Andre Noll | 58c0fed | 2009-03-31 14:33:13 +1100 | [diff] [blame] | 321 | * component devices */ |
Andre Noll | f233ea5 | 2008-07-21 17:05:22 +1000 | [diff] [blame] | 322 | sector_t array_sectors; /* exported array size */ |
Dan Williams | b522adc | 2009-03-31 15:00:31 +1100 | [diff] [blame] | 323 | int external_size; /* size managed |
| 324 | * externally */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 325 | __u64 events; |
NeilBrown | a8707c0 | 2010-05-18 09:28:43 +1000 | [diff] [blame] | 326 | /* 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 Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 332 | |
| 333 | char uuid[16]; |
| 334 | |
NeilBrown | f670557 | 2006-03-27 01:18:11 -0800 | [diff] [blame] | 335 | /* 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 Noll | 664e7c4 | 2009-06-18 08:45:27 +1000 | [diff] [blame] | 341 | int delta_disks, new_level, new_layout; |
Mike Snitzer | 77a6869 | 2020-12-14 12:08:48 -0500 | [diff] [blame] | 342 | int new_chunk_sectors; |
NeilBrown | 2c810cd | 2012-05-21 09:27:00 +1000 | [diff] [blame] | 343 | int reshape_backwards; |
NeilBrown | f670557 | 2006-03-27 01:18:11 -0800 | [diff] [blame] | 344 | |
NeilBrown | 2b8bf34 | 2011-10-11 16:48:23 +1100 | [diff] [blame] | 345 | struct md_thread *thread; /* management thread */ |
| 346 | struct md_thread *sync_thread; /* doing resync or reconstruct */ |
Jonathan Brassow | c4a3955 | 2013-06-25 01:23:59 -0500 | [diff] [blame] | 347 | |
| 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; |
NeilBrown | ff4e8d9 | 2006-07-10 04:44:16 -0700 | [diff] [blame] | 355 | sector_t curr_resync; /* last block scheduled */ |
NeilBrown | 97e4f42 | 2009-03-31 14:33:13 +1100 | [diff] [blame] | 356 | /* 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 Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 363 | unsigned long resync_mark; /* a recent timestamp */ |
| 364 | sector_t resync_mark_cnt;/* blocks written at resync_mark */ |
NeilBrown | ff4e8d9 | 2006-07-10 04:44:16 -0700 | [diff] [blame] | 365 | sector_t curr_mark_cnt; /* blocks scheduled now */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 366 | |
| 367 | sector_t resync_max_sectors; /* may be set by personality */ |
NeilBrown | 9d88883 | 2005-11-08 21:39:26 -0800 | [diff] [blame] | 368 | |
Jianpeng Ma | 7f7583d | 2012-10-11 14:17:59 +1100 | [diff] [blame] | 369 | atomic64_t resync_mismatches; /* count of sectors where |
NeilBrown | 9d88883 | 2005-11-08 21:39:26 -0800 | [diff] [blame] | 370 | * parity/replica mismatch found |
| 371 | */ |
NeilBrown | e464eaf | 2006-03-27 01:18:14 -0800 | [diff] [blame] | 372 | |
| 373 | /* allow user-space to request suspension of IO to regions of the array */ |
| 374 | sector_t suspend_lo; |
| 375 | sector_t suspend_hi; |
NeilBrown | 88202a0 | 2006-01-06 00:21:36 -0800 | [diff] [blame] | 376 | /* if zero, use the system-wide default */ |
| 377 | int sync_speed_min; |
| 378 | int sync_speed_max; |
| 379 | |
Bernd Schubert | 90b0871 | 2008-05-23 13:04:38 -0700 | [diff] [blame] | 380 | /* resync even though the same disks are shared among md-devices */ |
| 381 | int parallel_resync; |
| 382 | |
NeilBrown | 6ff8d8ec | 2006-01-06 00:20:15 -0800 | [diff] [blame] | 383 | int ok_start_degraded; |
NeilBrown | 5fd6c1d | 2006-06-26 00:27:40 -0700 | [diff] [blame] | 384 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 385 | unsigned long recovery; |
NeilBrown | 5389042 | 2011-07-27 11:00:36 +1000 | [diff] [blame] | 386 | /* 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 Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 392 | |
| 393 | int in_sync; /* know to not need resync */ |
NeilBrown | c8c00a6 | 2009-08-10 12:50:52 +1000 | [diff] [blame] | 394 | /* '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 Hellwig | de09077 | 2020-09-01 17:57:48 +0200 | [diff] [blame] | 400 | * reconfig_mutex -> bd_mutex |
NeilBrown | c8c00a6 | 2009-08-10 12:50:52 +1000 | [diff] [blame] | 401 | * bd_mutex -> open_mutex: e.g. __blkdev_get -> md_open |
| 402 | */ |
| 403 | struct mutex open_mutex; |
NeilBrown | df5b89b | 2006-03-27 01:18:20 -0800 | [diff] [blame] | 404 | struct mutex reconfig_mutex; |
NeilBrown | f2ea68c | 2008-07-21 17:05:25 +1000 | [diff] [blame] | 405 | atomic_t active; /* general refcount */ |
| 406 | atomic_t openers; /* number of active opens */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 407 | |
NeilBrown | f0b4f7e | 2011-02-24 17:26:41 +1100 | [diff] [blame] | 408 | int changed; /* True if we might need to |
| 409 | * reread partition info */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 410 | 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 Brown | 5e96ee6 | 2008-06-28 08:31:24 +1000 | [diff] [blame] | 417 | sector_t resync_min; /* user requested sync |
| 418 | * starts here */ |
NeilBrown | c620727 | 2008-02-06 01:39:52 -0800 | [diff] [blame] | 419 | sector_t resync_max; /* resync should pause |
| 420 | * when it gets here */ |
NeilBrown | 06d91a5 | 2005-06-21 17:17:12 -0700 | [diff] [blame] | 421 | |
Tejun Heo | 324a56e | 2013-12-11 14:11:53 -0500 | [diff] [blame] | 422 | struct kernfs_node *sysfs_state; /* handle for 'array_state' |
NeilBrown | b62b759 | 2008-10-21 13:25:21 +1100 | [diff] [blame] | 423 | * file in sysfs. |
| 424 | */ |
Tejun Heo | 324a56e | 2013-12-11 14:11:53 -0500 | [diff] [blame] | 425 | struct kernfs_node *sysfs_action; /* handle for 'sync_action' */ |
Junxiao Bi | e1a86db | 2020-07-14 16:10:26 -0700 | [diff] [blame] | 426 | 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' */ |
NeilBrown | b62b759 | 2008-10-21 13:25:21 +1100 | [diff] [blame] | 429 | |
NeilBrown | d337482 | 2009-01-09 08:31:10 +1100 | [diff] [blame] | 430 | struct work_struct del_work; /* used for delayed sysfs removal */ |
| 431 | |
NeilBrown | 85572d7 | 2014-12-15 12:56:56 +1100 | [diff] [blame] | 432 | /* "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 |
NeilBrown | 36d091f | 2014-12-15 12:56:58 +1100 | [diff] [blame] | 437 | * pers (also protected by reconfig_mutex and pending IO). |
NeilBrown | 978a7a4 | 2014-12-15 12:56:58 +1100 | [diff] [blame] | 438 | * clearing ->bitmap |
NeilBrown | 4af1a04 | 2014-12-15 12:57:00 +1100 | [diff] [blame] | 439 | * clearing ->bitmap_info.file |
NeilBrown | 23da422 | 2014-12-15 12:57:01 +1100 | [diff] [blame] | 440 | * changing ->resync_{min,max} |
| 441 | * setting MD_RECOVERY_RUNNING (which interacts with resync_{min,max}) |
NeilBrown | 85572d7 | 2014-12-15 12:56:56 +1100 | [diff] [blame] | 442 | */ |
| 443 | spinlock_t lock; |
NeilBrown | 3d310eb | 2005-06-21 17:17:26 -0700 | [diff] [blame] | 444 | wait_queue_head_t sb_wait; /* for waiting on superblock updates */ |
NeilBrown | 7bfa19f | 2005-06-21 17:17:28 -0700 | [diff] [blame] | 445 | atomic_t pending_writes; /* number of active superblock writes */ |
NeilBrown | 06d91a5 | 2005-06-21 17:17:12 -0700 | [diff] [blame] | 446 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 447 | unsigned int safemode; /* if set, update "clean" superblock |
| 448 | * when no writes pending. |
NeilBrown | f72ffdd | 2014-09-30 14:23:59 +1000 | [diff] [blame] | 449 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 450 | unsigned int safemode_delay; |
| 451 | struct timer_list safemode_timer; |
NeilBrown | 4ad23a97 | 2017-03-15 14:05:14 +1100 | [diff] [blame] | 452 | struct percpu_ref writes_pending; |
| 453 | int sync_checkers; /* # of threads checking writes_pending */ |
Jens Axboe | 165125e | 2007-07-24 09:28:11 +0200 | [diff] [blame] | 454 | struct request_queue *queue; /* for plugging ... */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 455 | |
NeilBrown | f72ffdd | 2014-09-30 14:23:59 +1000 | [diff] [blame] | 456 | struct bitmap *bitmap; /* the bitmap for the device */ |
NeilBrown | c3d9714 | 2009-12-14 12:49:52 +1100 | [diff] [blame] | 457 | struct { |
| 458 | struct file *file; /* the bitmap file */ |
NeilBrown | f6af949 | 2009-12-14 12:49:54 +1100 | [diff] [blame] | 459 | loff_t offset; /* offset from superblock of |
NeilBrown | c3d9714 | 2009-12-14 12:49:52 +1100 | [diff] [blame] | 460 | * start of bitmap. May be |
| 461 | * negative, but not '0' |
NeilBrown | f6af949 | 2009-12-14 12:49:54 +1100 | [diff] [blame] | 462 | * For external metadata, offset |
NeilBrown | f72ffdd | 2014-09-30 14:23:59 +1000 | [diff] [blame] | 463 | * from start of device. |
NeilBrown | c3d9714 | 2009-12-14 12:49:52 +1100 | [diff] [blame] | 464 | */ |
NeilBrown | 6409bb0 | 2012-05-22 13:55:07 +1000 | [diff] [blame] | 465 | unsigned long space; /* space available at this offset */ |
NeilBrown | f6af949 | 2009-12-14 12:49:54 +1100 | [diff] [blame] | 466 | loff_t default_offset; /* this is the offset to use when |
NeilBrown | c3d9714 | 2009-12-14 12:49:52 +1100 | [diff] [blame] | 467 | * hot-adding a bitmap. It should |
| 468 | * eventually be settable by sysfs. |
| 469 | */ |
NeilBrown | 6409bb0 | 2012-05-22 13:55:07 +1000 | [diff] [blame] | 470 | unsigned long default_space; /* space available at |
| 471 | * default offset */ |
NeilBrown | c3d9714 | 2009-12-14 12:49:52 +1100 | [diff] [blame] | 472 | struct mutex mutex; |
NeilBrown | 42a04b5 | 2009-12-14 12:49:53 +1100 | [diff] [blame] | 473 | unsigned long chunksize; |
NeilBrown | ac2f40b | 2010-06-01 19:37:31 +1000 | [diff] [blame] | 474 | unsigned long daemon_sleep; /* how many jiffies between updates? */ |
NeilBrown | 42a04b5 | 2009-12-14 12:49:53 +1100 | [diff] [blame] | 475 | unsigned long max_write_behind; /* write-behind mode */ |
NeilBrown | ece5cff | 2009-12-14 12:49:56 +1100 | [diff] [blame] | 476 | int external; |
Goldwyn Rodrigues | c4ce867 | 2014-03-29 10:20:02 -0500 | [diff] [blame] | 477 | int nodes; /* Maximum number of nodes in the cluster */ |
Goldwyn Rodrigues | cf921cc | 2014-03-30 00:42:49 -0500 | [diff] [blame] | 478 | char cluster_name[64]; /* Name of the cluster */ |
NeilBrown | c3d9714 | 2009-12-14 12:49:52 +1100 | [diff] [blame] | 479 | } bitmap_info; |
NeilBrown | 32a7627 | 2005-06-21 17:17:14 -0700 | [diff] [blame] | 480 | |
NeilBrown | f72ffdd | 2014-09-30 14:23:59 +1000 | [diff] [blame] | 481 | atomic_t max_corr_read_errors; /* max read retries */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 482 | struct list_head all_mddevs; |
NeilBrown | a2826aa | 2009-12-14 12:49:49 +1100 | [diff] [blame] | 483 | |
NeilBrown | a64c876 | 2010-04-14 17:15:37 +1000 | [diff] [blame] | 484 | struct attribute_group *to_remove; |
NeilBrown | 252ac52 | 2010-06-01 19:37:29 +1000 | [diff] [blame] | 485 | |
Kent Overstreet | afeee51 | 2018-05-20 18:25:52 -0400 | [diff] [blame] | 486 | struct bio_set bio_set; |
| 487 | struct bio_set sync_set; /* for sync operations like |
NeilBrown | 5a85071 | 2017-06-21 09:12:21 +1000 | [diff] [blame] | 488 | * metadata and bitmap writes |
| 489 | */ |
Artur Paszkiewicz | 41d2d84 | 2020-07-03 11:13:09 +0200 | [diff] [blame] | 490 | mempool_t md_io_pool; |
NeilBrown | a167f66 | 2010-10-26 18:31:13 +1100 | [diff] [blame] | 491 | |
NeilBrown | 4bc034d | 2019-03-29 10:46:16 -0700 | [diff] [blame] | 492 | /* 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 Gupta | 81ba3c2 | 2020-11-11 06:16:56 +0100 | [diff] [blame] | 498 | ktime_t start_flush, prev_flush_start; /* prev_flush_start is when the previous completed |
| 499 | * flush was started. |
| 500 | */ |
NeilBrown | 4bc034d | 2019-03-29 10:46:16 -0700 | [diff] [blame] | 501 | struct work_struct flush_work; |
NeilBrown | 768a418 | 2010-07-26 11:49:55 +1000 | [diff] [blame] | 502 | struct work_struct event_work; /* used by dm to report failure event */ |
Guoqing Jiang | 404659c | 2019-12-23 10:48:53 +0100 | [diff] [blame] | 503 | mempool_t *serial_info_pool; |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 504 | void (*sync_super)(struct mddev *mddev, struct md_rdev *rdev); |
Goldwyn Rodrigues | c4ce867 | 2014-03-29 10:20:02 -0500 | [diff] [blame] | 505 | struct md_cluster_info *cluster_info; |
Guoqing Jiang | 15858fa | 2015-12-21 10:51:00 +1100 | [diff] [blame] | 506 | unsigned int good_device_nr; /* good device num within cluster raid */ |
Coly Li | 78f57ef | 2020-04-09 22:17:20 +0800 | [diff] [blame] | 507 | unsigned int noio_flag; /* for memalloc scope API */ |
Heinz Mauelshagen | 4b6c106 | 2018-02-02 23:13:19 +0100 | [diff] [blame] | 508 | |
| 509 | bool has_superblocks:1; |
Guoqing Jiang | 9a56784 | 2019-07-24 11:09:19 +0200 | [diff] [blame] | 510 | bool fail_last_dev:1; |
Guoqing Jiang | 3938f5f | 2019-12-23 10:48:56 +0100 | [diff] [blame] | 511 | bool serialize_policy:1; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 512 | }; |
| 513 | |
NeilBrown | be306c2 | 2016-11-09 10:21:33 +1100 | [diff] [blame] | 514 | enum 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 Liu | d5d885f | 2017-11-19 22:17:01 -0800 | [diff] [blame] | 529 | MD_RECOVERY_WAIT, /* waiting for pers->start() to finish */ |
Guoqing Jiang | 0357ba2 | 2018-07-02 16:26:25 +0800 | [diff] [blame] | 530 | MD_RESYNCING_REMOTE, /* remote node is running resync thread */ |
NeilBrown | be306c2 | 2016-11-09 10:21:33 +1100 | [diff] [blame] | 531 | }; |
| 532 | |
NeilBrown | 5c47daf | 2014-12-15 12:57:01 +1100 | [diff] [blame] | 533 | static 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 | */ |
| 541 | static inline void mddev_lock_nointr(struct mddev *mddev) |
| 542 | { |
| 543 | mutex_lock(&mddev->reconfig_mutex); |
| 544 | } |
| 545 | |
NeilBrown | 5c47daf | 2014-12-15 12:57:01 +1100 | [diff] [blame] | 546 | static inline int mddev_trylock(struct mddev *mddev) |
| 547 | { |
| 548 | return mutex_trylock(&mddev->reconfig_mutex); |
| 549 | } |
| 550 | extern void mddev_unlock(struct mddev *mddev); |
| 551 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 552 | static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors) |
| 553 | { |
Christoph Hellwig | 4245e52 | 2020-09-03 07:40:59 +0200 | [diff] [blame] | 554 | atomic_add(nr_sectors, &bdev->bd_disk->sync_io); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 555 | } |
| 556 | |
Christoph Hellwig | 74d4699 | 2017-08-23 19:10:32 +0200 | [diff] [blame] | 557 | static inline void md_sync_acct_bio(struct bio *bio, unsigned long nr_sectors) |
| 558 | { |
Christoph Hellwig | 309dca30 | 2021-01-24 11:02:34 +0100 | [diff] [blame] | 559 | md_sync_acct(bio->bi_bdev, nr_sectors); |
Christoph Hellwig | 74d4699 | 2017-08-23 19:10:32 +0200 | [diff] [blame] | 560 | } |
| 561 | |
NeilBrown | 84fc4b5 | 2011-10-11 16:49:58 +1100 | [diff] [blame] | 562 | struct md_personality |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 563 | { |
| 564 | char *name; |
NeilBrown | 2604b70 | 2006-01-06 00:20:36 -0800 | [diff] [blame] | 565 | int level; |
| 566 | struct list_head list; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 567 | struct module *owner; |
David Jeffery | 775d783 | 2019-09-16 13:15:14 -0400 | [diff] [blame] | 568 | bool __must_check (*make_request)(struct mddev *mddev, struct bio *bio); |
Song Liu | d5d885f | 2017-11-19 22:17:01 -0800 | [diff] [blame] | 569 | /* |
| 570 | * start up works that do NOT require md_thread. tasks that |
| 571 | * requires md_thread should go into start() |
| 572 | */ |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 573 | int (*run)(struct mddev *mddev); |
Song Liu | d5d885f | 2017-11-19 22:17:01 -0800 | [diff] [blame] | 574 | /* start up works that require md threads */ |
| 575 | int (*start)(struct mddev *mddev); |
NeilBrown | afa0f55 | 2014-12-15 12:56:58 +1100 | [diff] [blame] | 576 | void (*free)(struct mddev *mddev, void *priv); |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 577 | void (*status)(struct seq_file *seq, struct mddev *mddev); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 578 | /* error_handler must set ->faulty and clear ->in_sync |
NeilBrown | f72ffdd | 2014-09-30 14:23:59 +1000 | [diff] [blame] | 579 | * if appropriate, and should abort recovery if needed |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 580 | */ |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 581 | void (*error_handler)(struct mddev *mddev, struct md_rdev *rdev); |
| 582 | int (*hot_add_disk) (struct mddev *mddev, struct md_rdev *rdev); |
NeilBrown | b8321b6 | 2011-12-23 10:17:51 +1100 | [diff] [blame] | 583 | int (*hot_remove_disk) (struct mddev *mddev, struct md_rdev *rdev); |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 584 | int (*spare_active) (struct mddev *mddev); |
NeilBrown | 0931479 | 2015-02-19 16:04:40 +1100 | [diff] [blame] | 585 | sector_t (*sync_request)(struct mddev *mddev, sector_t sector_nr, int *skipped); |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 586 | 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 Jiang | 7564bed | 2018-10-18 16:37:42 +0800 | [diff] [blame] | 591 | void (*update_reshape_pos) (struct mddev *mddev); |
NeilBrown | b03e0cc | 2017-10-19 12:49:15 +1100 | [diff] [blame] | 592 | /* quiesce suspends or resumes internal processing. |
| 593 | * 1 - stop new actions and wait for action io to complete |
| 594 | * 0 - return to normal behaviour |
NeilBrown | 36fa306 | 2005-09-09 16:23:45 -0700 | [diff] [blame] | 595 | */ |
NeilBrown | b03e0cc | 2017-10-19 12:49:15 +1100 | [diff] [blame] | 596 | void (*quiesce) (struct mddev *mddev, int quiesce); |
NeilBrown | 245f46c | 2009-03-31 14:39:39 +1100 | [diff] [blame] | 597 | /* 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 | */ |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 606 | void *(*takeover) (struct mddev *mddev); |
Artur Paszkiewicz | ba903a3 | 2017-03-09 10:00:03 +0100 | [diff] [blame] | 607 | /* Changes the consistency policy of an active array. */ |
| 608 | int (*change_consistency_policy)(struct mddev *mddev, const char *buf); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 609 | }; |
| 610 | |
NeilBrown | 007583c | 2005-11-08 21:39:30 -0800 | [diff] [blame] | 611 | struct md_sysfs_entry { |
| 612 | struct attribute attr; |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 613 | ssize_t (*show)(struct mddev *, char *); |
| 614 | ssize_t (*store)(struct mddev *, const char *, size_t); |
NeilBrown | 007583c | 2005-11-08 21:39:30 -0800 | [diff] [blame] | 615 | }; |
NeilBrown | 43a7050 | 2009-12-14 12:49:55 +1100 | [diff] [blame] | 616 | extern struct attribute_group md_bitmap_group; |
NeilBrown | 007583c | 2005-11-08 21:39:30 -0800 | [diff] [blame] | 617 | |
Tejun Heo | 324a56e | 2013-12-11 14:11:53 -0500 | [diff] [blame] | 618 | static inline struct kernfs_node *sysfs_get_dirent_safe(struct kernfs_node *sd, char *name) |
NeilBrown | 00bcb4a | 2010-06-01 19:37:23 +1000 | [diff] [blame] | 619 | { |
| 620 | if (sd) |
Tejun Heo | 388975c | 2013-09-11 23:19:13 -0400 | [diff] [blame] | 621 | return sysfs_get_dirent(sd, name); |
NeilBrown | 00bcb4a | 2010-06-01 19:37:23 +1000 | [diff] [blame] | 622 | return sd; |
| 623 | } |
Tejun Heo | 324a56e | 2013-12-11 14:11:53 -0500 | [diff] [blame] | 624 | static inline void sysfs_notify_dirent_safe(struct kernfs_node *sd) |
NeilBrown | 00bcb4a | 2010-06-01 19:37:23 +1000 | [diff] [blame] | 625 | { |
| 626 | if (sd) |
| 627 | sysfs_notify_dirent(sd); |
| 628 | } |
| 629 | |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 630 | static inline char * mdname (struct mddev * mddev) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 631 | { |
| 632 | return mddev->gendisk ? mddev->gendisk->disk_name : "mdX"; |
| 633 | } |
| 634 | |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 635 | static inline int sysfs_link_rdev(struct mddev *mddev, struct md_rdev *rdev) |
Namhyung Kim | 36fad85 | 2011-07-27 11:00:36 +1000 | [diff] [blame] | 636 | { |
| 637 | char nm[20]; |
Shaohua Li | 9b15603 | 2015-12-18 15:19:16 +1100 | [diff] [blame] | 638 | if (!test_bit(Replacement, &rdev->flags) && |
| 639 | !test_bit(Journal, &rdev->flags) && |
| 640 | mddev->kobj.sd) { |
NeilBrown | 2d78f8c | 2011-12-23 10:17:51 +1100 | [diff] [blame] | 641 | sprintf(nm, "rd%d", rdev->raid_disk); |
| 642 | return sysfs_create_link(&mddev->kobj, &rdev->kobj, nm); |
| 643 | } else |
| 644 | return 0; |
Namhyung Kim | 36fad85 | 2011-07-27 11:00:36 +1000 | [diff] [blame] | 645 | } |
| 646 | |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 647 | static inline void sysfs_unlink_rdev(struct mddev *mddev, struct md_rdev *rdev) |
Namhyung Kim | 36fad85 | 2011-07-27 11:00:36 +1000 | [diff] [blame] | 648 | { |
| 649 | char nm[20]; |
Shaohua Li | 9b15603 | 2015-12-18 15:19:16 +1100 | [diff] [blame] | 650 | if (!test_bit(Replacement, &rdev->flags) && |
| 651 | !test_bit(Journal, &rdev->flags) && |
| 652 | mddev->kobj.sd) { |
NeilBrown | 2d78f8c | 2011-12-23 10:17:51 +1100 | [diff] [blame] | 653 | sprintf(nm, "rd%d", rdev->raid_disk); |
| 654 | sysfs_remove_link(&mddev->kobj, nm); |
| 655 | } |
Namhyung Kim | 36fad85 | 2011-07-27 11:00:36 +1000 | [diff] [blame] | 656 | } |
| 657 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 658 | /* |
| 659 | * iterates through some rdev ringlist. It's safe to remove the |
| 660 | * current 'rdev'. Dont touch 'tmp' though. |
| 661 | */ |
Cheng Renquan | 159ec1f | 2009-01-09 08:31:08 +1100 | [diff] [blame] | 662 | #define rdev_for_each_list(rdev, tmp, head) \ |
| 663 | list_for_each_entry_safe(rdev, tmp, head, same_set) |
| 664 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 665 | /* |
| 666 | * iterates through the 'same array disks' ringlist |
| 667 | */ |
NeilBrown | dafb20f | 2012-03-19 12:46:39 +1100 | [diff] [blame] | 668 | #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 Renquan | 159ec1f | 2009-01-09 08:31:08 +1100 | [diff] [blame] | 672 | list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 673 | |
NeilBrown | 4b80991 | 2008-07-21 17:05:25 +1000 | [diff] [blame] | 674 | #define rdev_for_each_rcu(rdev, mddev) \ |
| 675 | list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set) |
| 676 | |
NeilBrown | 2b8bf34 | 2011-10-11 16:48:23 +1100 | [diff] [blame] | 677 | struct md_thread { |
Shaohua Li | 4ed8731 | 2012-10-11 13:34:00 +1100 | [diff] [blame] | 678 | void (*run) (struct md_thread *thread); |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 679 | struct mddev *mddev; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 680 | wait_queue_head_t wqueue; |
NeilBrown | f72ffdd | 2014-09-30 14:23:59 +1000 | [diff] [blame] | 681 | unsigned long flags; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 682 | struct task_struct *tsk; |
NeilBrown | 32a7627 | 2005-06-21 17:17:14 -0700 | [diff] [blame] | 683 | unsigned long timeout; |
Shaohua Li | 4ed8731 | 2012-10-11 13:34:00 +1100 | [diff] [blame] | 684 | void *private; |
NeilBrown | 2b8bf34 | 2011-10-11 16:48:23 +1100 | [diff] [blame] | 685 | }; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 686 | |
| 687 | #define THREAD_WAKEUP 0 |
| 688 | |
NeilBrown | 1345b1d | 2006-01-06 00:20:40 -0800 | [diff] [blame] | 689 | static inline void safe_put_page(struct page *p) |
| 690 | { |
| 691 | if (p) put_page(p); |
| 692 | } |
| 693 | |
NeilBrown | 84fc4b5 | 2011-10-11 16:49:58 +1100 | [diff] [blame] | 694 | extern int register_md_personality(struct md_personality *p); |
| 695 | extern int unregister_md_personality(struct md_personality *p); |
Goldwyn Rodrigues | edb39c9 | 2014-03-29 10:01:53 -0500 | [diff] [blame] | 696 | extern int register_md_cluster_operations(struct md_cluster_operations *ops, |
| 697 | struct module *module); |
| 698 | extern int unregister_md_cluster_operations(void); |
| 699 | extern int md_setup_cluster(struct mddev *mddev, int nodes); |
| 700 | extern void md_cluster_stop(struct mddev *mddev); |
NeilBrown | 2b8bf34 | 2011-10-11 16:48:23 +1100 | [diff] [blame] | 701 | extern struct md_thread *md_register_thread( |
Shaohua Li | 4ed8731 | 2012-10-11 13:34:00 +1100 | [diff] [blame] | 702 | void (*run)(struct md_thread *thread), |
NeilBrown | 2b8bf34 | 2011-10-11 16:48:23 +1100 | [diff] [blame] | 703 | struct mddev *mddev, |
| 704 | const char *name); |
| 705 | extern void md_unregister_thread(struct md_thread **threadp); |
| 706 | extern void md_wakeup_thread(struct md_thread *thread); |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 707 | extern void md_check_recovery(struct mddev *mddev); |
Jonathan Brassow | a91d5ac | 2013-04-24 11:42:43 +1000 | [diff] [blame] | 708 | extern void md_reap_sync_thread(struct mddev *mddev); |
NeilBrown | a415c0f | 2017-06-05 16:05:13 +1000 | [diff] [blame] | 709 | extern int mddev_init_writes_pending(struct mddev *mddev); |
NeilBrown | cc27b0c | 2017-06-05 16:49:39 +1000 | [diff] [blame] | 710 | extern bool md_write_start(struct mddev *mddev, struct bio *bi); |
NeilBrown | 4972805 | 2017-03-15 14:05:12 +1100 | [diff] [blame] | 711 | extern void md_write_inc(struct mddev *mddev, struct bio *bi); |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 712 | extern void md_write_end(struct mddev *mddev); |
| 713 | extern void md_done_sync(struct mddev *mddev, int blocks, int ok); |
| 714 | extern void md_error(struct mddev *mddev, struct md_rdev *rdev); |
NeilBrown | c6563a8 | 2012-05-21 09:27:00 +1000 | [diff] [blame] | 715 | extern void md_finish_reshape(struct mddev *mddev); |
Xiao Ni | cf78408 | 2021-02-04 15:50:43 +0800 | [diff] [blame] | 716 | void md_submit_discard_bio(struct mddev *mddev, struct md_rdev *rdev, |
| 717 | struct bio *bio, sector_t start, sector_t size); |
NeilBrown | 9202295 | 2009-03-31 14:33:13 +1100 | [diff] [blame] | 718 | |
David Jeffery | 775d783 | 2019-09-16 13:15:14 -0400 | [diff] [blame] | 719 | extern bool __must_check md_flush_request(struct mddev *mddev, struct bio *bio); |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 720 | extern void md_super_write(struct mddev *mddev, struct md_rdev *rdev, |
NeilBrown | 9202295 | 2009-03-31 14:33:13 +1100 | [diff] [blame] | 721 | sector_t sector, int size, struct page *page); |
NeilBrown | 46533ff | 2016-11-18 16:16:11 +1100 | [diff] [blame] | 722 | extern int md_super_wait(struct mddev *mddev); |
NeilBrown | f72ffdd | 2014-09-30 14:23:59 +1000 | [diff] [blame] | 723 | extern int sync_page_io(struct md_rdev *rdev, sector_t sector, int size, |
Mike Christie | 796a5cf | 2016-06-05 14:32:07 -0500 | [diff] [blame] | 724 | struct page *page, int op, int op_flags, |
| 725 | bool metadata_op); |
Shaohua Li | 4ed8731 | 2012-10-11 13:34:00 +1100 | [diff] [blame] | 726 | extern void md_do_sync(struct md_thread *thread); |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 727 | extern void md_new_event(struct mddev *mddev); |
Artur Paszkiewicz | 2214c26 | 2017-05-08 11:56:55 +0200 | [diff] [blame] | 728 | extern void md_allow_write(struct mddev *mddev); |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 729 | extern void md_wait_for_blocked_rdev(struct md_rdev *rdev, struct mddev *mddev); |
| 730 | extern void md_set_array_sectors(struct mddev *mddev, sector_t array_sectors); |
| 731 | extern int md_check_no_bitmap(struct mddev *mddev); |
| 732 | extern int md_integrity_register(struct mddev *mddev); |
Dan Williams | 1501efa | 2016-01-13 16:00:07 -0800 | [diff] [blame] | 733 | extern int md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev); |
NeilBrown | 72e0207 | 2009-12-14 12:49:55 +1100 | [diff] [blame] | 734 | extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale); |
Christoph Hellwig | 63fe081 | 2009-04-14 12:01:53 +1000 | [diff] [blame] | 735 | |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 736 | extern void mddev_init(struct mddev *mddev); |
| 737 | extern int md_run(struct mddev *mddev); |
Song Liu | d5d885f | 2017-11-19 22:17:01 -0800 | [diff] [blame] | 738 | extern int md_start(struct mddev *mddev); |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 739 | extern void md_stop(struct mddev *mddev); |
| 740 | extern void md_stop_writes(struct mddev *mddev); |
NeilBrown | 3cb0300 | 2011-10-11 16:45:26 +1100 | [diff] [blame] | 741 | extern int md_rdev_init(struct md_rdev *rdev); |
NeilBrown | 545c879 | 2012-05-22 13:54:30 +1000 | [diff] [blame] | 742 | extern void md_rdev_clear(struct md_rdev *rdev); |
NeilBrown | 390ee60 | 2010-06-01 19:37:27 +1000 | [diff] [blame] | 743 | |
Shaohua Li | 393debc | 2017-09-21 10:23:35 -0700 | [diff] [blame] | 744 | extern void md_handle_request(struct mddev *mddev, struct bio *bio); |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 745 | extern void mddev_suspend(struct mddev *mddev); |
| 746 | extern void mddev_resume(struct mddev *mddev); |
NeilBrown | 9cbb175 | 2012-07-31 09:08:14 +0200 | [diff] [blame] | 747 | |
Goldwyn Rodrigues | 70bcecd | 2015-08-21 10:33:39 -0500 | [diff] [blame] | 748 | extern void md_reload_sb(struct mddev *mddev, int raid_disk); |
Goldwyn Rodrigues | 1aee41f | 2014-10-29 18:51:31 -0500 | [diff] [blame] | 749 | extern void md_update_sb(struct mddev *mddev, int force); |
Goldwyn Rodrigues | fb56dfe | 2015-04-14 10:43:24 -0500 | [diff] [blame] | 750 | extern void md_kick_rdev_from_array(struct md_rdev * rdev); |
Guoqing Jiang | 404659c | 2019-12-23 10:48:53 +0100 | [diff] [blame] | 751 | extern void mddev_create_serial_pool(struct mddev *mddev, struct md_rdev *rdev, |
Guoqing Jiang | 11d3a9f | 2019-12-23 10:48:55 +0100 | [diff] [blame] | 752 | bool is_suspend); |
Guoqing Jiang | 69b00b5 | 2019-12-23 10:49:00 +0100 | [diff] [blame] | 753 | extern void mddev_destroy_serial_pool(struct mddev *mddev, struct md_rdev *rdev, |
| 754 | bool is_suspend); |
Goldwyn Rodrigues | 57d051d | 2015-04-14 10:43:55 -0500 | [diff] [blame] | 755 | struct md_rdev *md_find_rdev_nr_rcu(struct mddev *mddev, int nr); |
Tomasz Majchrzak | 1532d9e | 2017-12-27 10:31:40 +0100 | [diff] [blame] | 756 | struct md_rdev *md_find_rdev_rcu(struct mddev *mddev, dev_t dev); |
Hannes Reinecke | dfe15ac | 2012-07-26 11:12:18 +0200 | [diff] [blame] | 757 | |
Guilherme G. Piccoli | 62f7b19 | 2019-09-03 16:49:00 -0300 | [diff] [blame] | 758 | static 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 Reinecke | dfe15ac | 2012-07-26 11:12:18 +0200 | [diff] [blame] | 771 | static 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 Rodrigues | edb39c9 | 2014-03-29 10:01:53 -0500 | [diff] [blame] | 780 | extern struct md_cluster_operations *md_cluster_ops; |
Goldwyn Rodrigues | c4ce867 | 2014-03-29 10:20:02 -0500 | [diff] [blame] | 781 | static inline int mddev_is_clustered(struct mddev *mddev) |
| 782 | { |
| 783 | return mddev->cluster_info && mddev->bitmap_info.nodes > 1; |
| 784 | } |
Shaohua Li | 394ed8e | 2017-01-04 16:10:19 -0800 | [diff] [blame] | 785 | |
| 786 | /* clear unsupported mddev_flags */ |
| 787 | static inline void mddev_clear_unsupported_flags(struct mddev *mddev, |
| 788 | unsigned long unsupported_flags) |
| 789 | { |
| 790 | mddev->flags &= ~unsupported_flags; |
| 791 | } |
Shaohua Li | 2648381 | 2017-02-13 16:21:49 -0800 | [diff] [blame] | 792 | |
| 793 | static inline void mddev_check_writesame(struct mddev *mddev, struct bio *bio) |
| 794 | { |
| 795 | if (bio_op(bio) == REQ_OP_WRITE_SAME && |
Christoph Hellwig | 309dca30 | 2021-01-24 11:02:34 +0100 | [diff] [blame] | 796 | !bio->bi_bdev->bd_disk->queue->limits.max_write_same_sectors) |
Shaohua Li | 2648381 | 2017-02-13 16:21:49 -0800 | [diff] [blame] | 797 | mddev->queue->limits.max_write_same_sectors = 0; |
| 798 | } |
Christoph Hellwig | 3deff1a | 2017-04-05 19:21:03 +0200 | [diff] [blame] | 799 | |
| 800 | static inline void mddev_check_write_zeroes(struct mddev *mddev, struct bio *bio) |
| 801 | { |
| 802 | if (bio_op(bio) == REQ_OP_WRITE_ZEROES && |
Christoph Hellwig | 309dca30 | 2021-01-24 11:02:34 +0100 | [diff] [blame] | 803 | !bio->bi_bdev->bd_disk->queue->limits.max_write_zeroes_sectors) |
Christoph Hellwig | 3deff1a | 2017-04-05 19:21:03 +0200 | [diff] [blame] | 804 | mddev->queue->limits.max_write_zeroes_sectors = 0; |
| 805 | } |
Christoph Hellwig | d82fa81 | 2020-06-06 15:00:24 +0200 | [diff] [blame] | 806 | |
Christoph Hellwig | 7e0adbf | 2020-06-07 17:31:19 +0200 | [diff] [blame] | 807 | struct mdu_array_info_s; |
| 808 | struct mdu_disk_info_s; |
| 809 | |
Christoph Hellwig | 1a6a050 | 2020-06-07 16:33:01 +0200 | [diff] [blame] | 810 | extern int mdp_major; |
Christoph Hellwig | d82fa81 | 2020-06-06 15:00:24 +0200 | [diff] [blame] | 811 | void md_autostart_arrays(int part); |
Christoph Hellwig | 7e0adbf | 2020-06-07 17:31:19 +0200 | [diff] [blame] | 812 | int md_set_array_info(struct mddev *mddev, struct mdu_array_info_s *info); |
| 813 | int md_add_new_disk(struct mddev *mddev, struct mdu_disk_info_s *info); |
| 814 | int do_md_run(struct mddev *mddev); |
| 815 | |
| 816 | extern const struct block_device_operations md_fops; |
Christoph Hellwig | d82fa81 | 2020-06-06 15:00:24 +0200 | [diff] [blame] | 817 | |
Christoph Hellwig | 63fe081 | 2009-04-14 12:01:53 +1000 | [diff] [blame] | 818 | #endif /* _MD_MD_H */ |