Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* |
Wang Sheng-Hui | 7e84152 | 2011-09-21 15:37:46 +1000 | [diff] [blame] | 2 | md.h : kernel internal structure of the Linux MD driver |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 3 | Copyright (C) 1996-98 Ingo Molnar, Gadi Oxman |
NeilBrown | f72ffdd | 2014-09-30 14:23:59 +1000 | [diff] [blame] | 4 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5 | This program is free software; you can redistribute it and/or modify |
| 6 | it under the terms of the GNU General Public License as published by |
| 7 | the Free Software Foundation; either version 2, or (at your option) |
| 8 | any later version. |
NeilBrown | f72ffdd | 2014-09-30 14:23:59 +1000 | [diff] [blame] | 9 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 10 | You should have received a copy of the GNU General Public License |
| 11 | (for example /usr/src/linux/COPYING); if not, write to the Free |
NeilBrown | f72ffdd | 2014-09-30 14:23:59 +1000 | [diff] [blame] | 12 | Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 13 | */ |
| 14 | |
Christoph Hellwig | 63fe081 | 2009-04-14 12:01:53 +1000 | [diff] [blame] | 15 | #ifndef _MD_MD_H |
| 16 | #define _MD_MD_H |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 17 | |
Christoph Hellwig | 63fe081 | 2009-04-14 12:01:53 +1000 | [diff] [blame] | 18 | #include <linux/blkdev.h> |
Tejun Heo | 66114ca | 2015-05-22 17:13:32 -0400 | [diff] [blame] | 19 | #include <linux/backing-dev.h> |
Vishal Verma | fc974ee | 2015-12-24 19:20:34 -0700 | [diff] [blame] | 20 | #include <linux/badblocks.h> |
Christoph Hellwig | 63fe081 | 2009-04-14 12:01:53 +1000 | [diff] [blame] | 21 | #include <linux/kobject.h> |
| 22 | #include <linux/list.h> |
| 23 | #include <linux/mm.h> |
| 24 | #include <linux/mutex.h> |
| 25 | #include <linux/timer.h> |
| 26 | #include <linux/wait.h> |
| 27 | #include <linux/workqueue.h> |
Goldwyn Rodrigues | edb39c9 | 2014-03-29 10:01:53 -0500 | [diff] [blame] | 28 | #include "md-cluster.h" |
David Howells | 9361401 | 2006-09-30 20:45:40 +0200 | [diff] [blame] | 29 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 30 | #define MaxSector (~(sector_t)0) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 31 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 32 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 33 | * MD's 'extended' device |
| 34 | */ |
NeilBrown | 3cb0300 | 2011-10-11 16:45:26 +1100 | [diff] [blame] | 35 | struct md_rdev { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 36 | struct list_head same_set; /* RAID devices within the same set */ |
| 37 | |
Andre Noll | dd8ac33 | 2009-03-31 14:33:13 +1100 | [diff] [blame] | 38 | sector_t sectors; /* Device size (in 512bytes sectors) */ |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 39 | struct mddev *mddev; /* RAID array if running */ |
NeilBrown | eea1bf3 | 2009-03-31 14:27:02 +1100 | [diff] [blame] | 40 | int last_events; /* IO event timestamp */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 41 | |
Jonathan Brassow | a6ff7e0 | 2011-01-14 09:14:34 +1100 | [diff] [blame] | 42 | /* |
| 43 | * If meta_bdev is non-NULL, it means that a separate device is |
| 44 | * being used to store the metadata (superblock/bitmap) which |
| 45 | * would otherwise be contained on the same device as the data (bdev). |
| 46 | */ |
| 47 | struct block_device *meta_bdev; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 48 | struct block_device *bdev; /* block device handle */ |
| 49 | |
NeilBrown | 2699b67 | 2011-07-28 11:31:47 +1000 | [diff] [blame] | 50 | struct page *sb_page, *bb_page; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 51 | int sb_loaded; |
NeilBrown | 4254376 | 2006-06-26 00:27:57 -0700 | [diff] [blame] | 52 | __u64 sb_events; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 53 | sector_t data_offset; /* start of data in array */ |
NeilBrown | c6563a8 | 2012-05-21 09:27:00 +1000 | [diff] [blame] | 54 | sector_t new_data_offset;/* only relevant while reshaping */ |
NeilBrown | f72ffdd | 2014-09-30 14:23:59 +1000 | [diff] [blame] | 55 | sector_t sb_start; /* offset of the super block (in 512byte sectors) */ |
NeilBrown | 0002b27 | 2005-09-09 16:23:53 -0700 | [diff] [blame] | 56 | int sb_size; /* bytes in the superblock */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 57 | int preferred_minor; /* autorun support */ |
| 58 | |
NeilBrown | 86e6ffd | 2005-11-08 21:39:24 -0800 | [diff] [blame] | 59 | struct kobject kobj; |
| 60 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 61 | /* A device can be in one of three states based on two flags: |
| 62 | * Not working: faulty==1 in_sync==0 |
| 63 | * Fully working: faulty==0 in_sync==1 |
| 64 | * Working, but not |
| 65 | * in sync with array |
| 66 | * faulty==0 in_sync==0 |
| 67 | * |
| 68 | * It can never have faulty==1, in_sync==1 |
| 69 | * This reduces the burden of testing multiple flags in many cases |
| 70 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 71 | |
NeilBrown | 2d78f8c | 2011-12-23 10:17:51 +1100 | [diff] [blame] | 72 | unsigned long flags; /* bit set of 'enum flag_bits' bits. */ |
Dan Williams | 6bfe0b4 | 2008-04-30 00:52:32 -0700 | [diff] [blame] | 73 | wait_queue_head_t blocked_wait; |
NeilBrown | 8ddf9ef | 2005-09-09 16:23:45 -0700 | [diff] [blame] | 74 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 75 | int desc_nr; /* descriptor index in the superblock */ |
| 76 | int raid_disk; /* role of device in array */ |
NeilBrown | e93f68a | 2010-06-15 09:36:03 +0100 | [diff] [blame] | 77 | int new_raid_disk; /* role that the device will have in |
| 78 | * the array after a level-change completes. |
| 79 | */ |
NeilBrown | 41158c7 | 2005-06-21 17:17:25 -0700 | [diff] [blame] | 80 | int saved_raid_disk; /* role that device used to have in the |
| 81 | * array and could again if we did a partial |
| 82 | * resync from the bitmap |
| 83 | */ |
Shaohua Li | 3069aa8 | 2015-08-13 14:31:56 -0700 | [diff] [blame] | 84 | union { |
| 85 | sector_t recovery_offset;/* If this device has been partially |
NeilBrown | 5fd6c1d | 2006-06-26 00:27:40 -0700 | [diff] [blame] | 86 | * recovered, this is where we were |
| 87 | * up to. |
| 88 | */ |
Shaohua Li | 3069aa8 | 2015-08-13 14:31:56 -0700 | [diff] [blame] | 89 | sector_t journal_tail; /* If this device is a journal device, |
| 90 | * this is the journal tail (journal |
| 91 | * recovery start point) |
| 92 | */ |
| 93 | }; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 94 | |
| 95 | atomic_t nr_pending; /* number of pending requests. |
| 96 | * only maintained for arrays that |
| 97 | * support hot removal |
| 98 | */ |
NeilBrown | ba22dcb | 2005-11-08 21:39:31 -0800 | [diff] [blame] | 99 | atomic_t read_errors; /* number of consecutive read errors that |
| 100 | * we have tried to ignore. |
| 101 | */ |
Robert Becker | 1e50915 | 2009-12-14 12:49:58 +1100 | [diff] [blame] | 102 | struct timespec last_read_error; /* monotonic time since our |
| 103 | * last read error |
| 104 | */ |
NeilBrown | 4dbcdc7 | 2006-01-06 00:20:52 -0800 | [diff] [blame] | 105 | atomic_t corrected_errors; /* number of corrected read errors, |
| 106 | * for reporting to userspace and storing |
| 107 | * in superblock. |
| 108 | */ |
NeilBrown | 5792a28 | 2007-04-04 19:08:18 -0700 | [diff] [blame] | 109 | struct work_struct del_work; /* used for delayed sysfs removal */ |
NeilBrown | 3c0ee63 | 2008-10-21 13:25:28 +1100 | [diff] [blame] | 110 | |
Tejun Heo | 324a56e | 2013-12-11 14:11:53 -0500 | [diff] [blame] | 111 | struct kernfs_node *sysfs_state; /* handle for 'state' |
NeilBrown | 3c0ee63 | 2008-10-21 13:25:28 +1100 | [diff] [blame] | 112 | * sysfs entry */ |
NeilBrown | 2230dfe | 2011-07-28 11:31:46 +1000 | [diff] [blame] | 113 | |
Vishal Verma | fc974ee | 2015-12-24 19:20:34 -0700 | [diff] [blame] | 114 | struct badblocks badblocks; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 115 | }; |
NeilBrown | 2d78f8c | 2011-12-23 10:17:51 +1100 | [diff] [blame] | 116 | enum flag_bits { |
| 117 | Faulty, /* device is known to have a fault */ |
| 118 | In_sync, /* device is in_sync with rest of array */ |
NeilBrown | 8313b8e | 2013-12-12 10:13:33 +1100 | [diff] [blame] | 119 | Bitmap_sync, /* ..actually, not quite In_sync. Need a |
| 120 | * bitmap-based recovery to get fully in sync |
| 121 | */ |
NeilBrown | 2d78f8c | 2011-12-23 10:17:51 +1100 | [diff] [blame] | 122 | WriteMostly, /* Avoid reading if at all possible */ |
| 123 | AutoDetected, /* added by auto-detect */ |
| 124 | Blocked, /* An error occurred but has not yet |
| 125 | * been acknowledged by the metadata |
| 126 | * handler, so don't allow writes |
| 127 | * until it is cleared */ |
| 128 | WriteErrorSeen, /* A write error has been seen on this |
| 129 | * device |
| 130 | */ |
| 131 | FaultRecorded, /* Intermediate state for clearing |
| 132 | * Blocked. The Fault is/will-be |
| 133 | * recorded in the metadata, but that |
| 134 | * metadata hasn't been stored safely |
| 135 | * on disk yet. |
| 136 | */ |
| 137 | BlockedBadBlocks, /* A writer is blocked because they |
| 138 | * found an unacknowledged bad-block. |
| 139 | * This can safely be cleared at any |
| 140 | * time, and the writer will re-check. |
| 141 | * It may be set at any time, and at |
| 142 | * worst the writer will timeout and |
| 143 | * re-check. So setting it as |
| 144 | * accurately as possible is good, but |
| 145 | * not absolutely critical. |
| 146 | */ |
| 147 | WantReplacement, /* This device is a candidate to be |
| 148 | * hot-replaced, either because it has |
| 149 | * reported some faults, or because |
| 150 | * of explicit request. |
| 151 | */ |
| 152 | Replacement, /* This device is a replacement for |
| 153 | * a want_replacement device with same |
| 154 | * raid_disk number. |
| 155 | */ |
Goldwyn Rodrigues | 1aee41f | 2014-10-29 18:51:31 -0500 | [diff] [blame] | 156 | Candidate, /* For clustered environments only: |
| 157 | * This device is seen locally but not |
| 158 | * by the whole cluster |
| 159 | */ |
Song Liu | bac624f | 2015-08-13 14:31:55 -0700 | [diff] [blame] | 160 | Journal, /* This device is used as journal for |
| 161 | * raid-5/6. |
| 162 | * Usually, this device should be faster |
| 163 | * than other devices in the array |
| 164 | */ |
Guoqing Jiang | 659b254 | 2015-12-21 10:50:59 +1100 | [diff] [blame] | 165 | ClusterRemove, |
NeilBrown | d787be4 | 2016-06-02 16:19:53 +1000 | [diff] [blame^] | 166 | RemoveSynchronized, /* synchronize_rcu() was called after |
| 167 | * this device was known to be faulty, |
| 168 | * so it is safe to remove without |
| 169 | * another synchronize_rcu() call. |
| 170 | */ |
NeilBrown | 2d78f8c | 2011-12-23 10:17:51 +1100 | [diff] [blame] | 171 | }; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 172 | |
NeilBrown | 3cb0300 | 2011-10-11 16:45:26 +1100 | [diff] [blame] | 173 | 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] | 174 | sector_t *first_bad, int *bad_sectors) |
| 175 | { |
| 176 | if (unlikely(rdev->badblocks.count)) { |
Vishal Verma | fc974ee | 2015-12-24 19:20:34 -0700 | [diff] [blame] | 177 | int rv = badblocks_check(&rdev->badblocks, rdev->data_offset + s, |
NeilBrown | 2230dfe | 2011-07-28 11:31:46 +1000 | [diff] [blame] | 178 | sectors, |
| 179 | first_bad, bad_sectors); |
| 180 | if (rv) |
| 181 | *first_bad -= rdev->data_offset; |
| 182 | return rv; |
| 183 | } |
| 184 | return 0; |
| 185 | } |
NeilBrown | 3cb0300 | 2011-10-11 16:45:26 +1100 | [diff] [blame] | 186 | 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] | 187 | int is_new); |
| 188 | extern int rdev_clear_badblocks(struct md_rdev *rdev, sector_t s, int sectors, |
| 189 | int is_new); |
Goldwyn Rodrigues | c4ce867 | 2014-03-29 10:20:02 -0500 | [diff] [blame] | 190 | struct md_cluster_info; |
| 191 | |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 192 | struct mddev { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 193 | void *private; |
NeilBrown | 84fc4b5 | 2011-10-11 16:49:58 +1100 | [diff] [blame] | 194 | struct md_personality *pers; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 195 | dev_t unit; |
| 196 | int md_minor; |
NeilBrown | 7a0a535 | 2013-08-27 16:28:23 +1000 | [diff] [blame] | 197 | struct list_head disks; |
NeilBrown | 850b2b4 | 2006-10-03 01:15:46 -0700 | [diff] [blame] | 198 | unsigned long flags; |
| 199 | #define MD_CHANGE_DEVS 0 /* Some device status has changed */ |
| 200 | #define MD_CHANGE_CLEAN 1 /* transition to or from 'clean' */ |
NeilBrown | 070dc6d | 2010-08-30 17:33:34 +1000 | [diff] [blame] | 201 | #define MD_CHANGE_PENDING 2 /* switch from 'clean' to 'active' in progress */ |
NeilBrown | 7a0a535 | 2013-08-27 16:28:23 +1000 | [diff] [blame] | 202 | #define MD_UPDATE_SB_FLAGS (1 | 2 | 4) /* If these are set, md_update_sb needed */ |
Jonathan Brassow | 9c81075 | 2011-06-08 17:59:30 -0500 | [diff] [blame] | 203 | #define MD_ARRAY_FIRST_USE 3 /* First use of array, needs initialization */ |
NeilBrown | 260fa03 | 2013-08-27 16:44:13 +1000 | [diff] [blame] | 204 | #define MD_STILL_CLOSED 4 /* If set, then array has not been opened since |
| 205 | * md_ioctl checked on it. |
| 206 | */ |
Shaohua Li | bd18f64 | 2015-09-02 13:49:50 -0700 | [diff] [blame] | 207 | #define MD_JOURNAL_CLEAN 5 /* A raid with journal is already clean */ |
Song Liu | a97b789 | 2015-10-08 21:54:09 -0700 | [diff] [blame] | 208 | #define MD_HAS_JOURNAL 6 /* The raid array has journal feature set */ |
Guoqing Jiang | 15858fa | 2015-12-21 10:51:00 +1100 | [diff] [blame] | 209 | #define MD_RELOAD_SB 7 /* Reload the superblock because another node |
| 210 | * updated it. |
| 211 | */ |
Guoqing Jiang | bb8bf15 | 2016-06-02 23:32:04 -0400 | [diff] [blame] | 212 | #define MD_CLUSTER_RESYNC_LOCKED 8 /* cluster raid only, which means node |
| 213 | * already took resync lock, need to |
| 214 | * release the lock */ |
NeilBrown | 850b2b4 | 2006-10-03 01:15:46 -0700 | [diff] [blame] | 215 | |
NeilBrown | 409c57f | 2009-03-31 14:39:39 +1100 | [diff] [blame] | 216 | int suspended; |
| 217 | atomic_t active_io; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 218 | int ro; |
NeilBrown | bb4f1e9 | 2010-08-08 21:18:03 +1000 | [diff] [blame] | 219 | int sysfs_active; /* set when sysfs deletes |
| 220 | * are happening, so run/ |
| 221 | * takeover/stop are not safe |
| 222 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 223 | struct gendisk *gendisk; |
| 224 | |
NeilBrown | eae1701 | 2005-11-08 21:39:23 -0800 | [diff] [blame] | 225 | struct kobject kobj; |
NeilBrown | d337482 | 2009-01-09 08:31:10 +1100 | [diff] [blame] | 226 | int hold_active; |
| 227 | #define UNTIL_IOCTL 1 |
NeilBrown | efeb53c | 2009-01-09 08:31:10 +1100 | [diff] [blame] | 228 | #define UNTIL_STOP 2 |
NeilBrown | eae1701 | 2005-11-08 21:39:23 -0800 | [diff] [blame] | 229 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 230 | /* Superblock information */ |
| 231 | int major_version, |
| 232 | minor_version, |
| 233 | patch_version; |
| 234 | int persistent; |
NeilBrown | f72ffdd | 2014-09-30 14:23:59 +1000 | [diff] [blame] | 235 | int external; /* metadata is |
NeilBrown | e691063 | 2008-02-06 01:39:51 -0800 | [diff] [blame] | 236 | * managed externally */ |
| 237 | char metadata_type[17]; /* externally set*/ |
Andre Noll | 9d8f036 | 2009-06-18 08:45:01 +1000 | [diff] [blame] | 238 | int chunk_sectors; |
Deepa Dinamani | 9ebc6ef | 2015-12-21 10:51:01 +1100 | [diff] [blame] | 239 | time64_t ctime, utime; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 240 | int level, layout; |
NeilBrown | d9d166c | 2006-01-06 00:20:51 -0800 | [diff] [blame] | 241 | char clevel[16]; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 242 | int raid_disks; |
| 243 | int max_disks; |
NeilBrown | f72ffdd | 2014-09-30 14:23:59 +1000 | [diff] [blame] | 244 | sector_t dev_sectors; /* used size of |
Andre Noll | 58c0fed | 2009-03-31 14:33:13 +1100 | [diff] [blame] | 245 | * component devices */ |
Andre Noll | f233ea5 | 2008-07-21 17:05:22 +1000 | [diff] [blame] | 246 | sector_t array_sectors; /* exported array size */ |
Dan Williams | b522adc | 2009-03-31 15:00:31 +1100 | [diff] [blame] | 247 | int external_size; /* size managed |
| 248 | * externally */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 249 | __u64 events; |
NeilBrown | a8707c0 | 2010-05-18 09:28:43 +1000 | [diff] [blame] | 250 | /* If the last 'event' was simply a clean->dirty transition, and |
| 251 | * we didn't write it to the spares, then it is safe and simple |
| 252 | * to just decrement the event count on a dirty->clean transition. |
| 253 | * So we record that possibility here. |
| 254 | */ |
| 255 | int can_decrease_events; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 256 | |
| 257 | char uuid[16]; |
| 258 | |
NeilBrown | f670557 | 2006-03-27 01:18:11 -0800 | [diff] [blame] | 259 | /* If the array is being reshaped, we need to record the |
| 260 | * new shape and an indication of where we are up to. |
| 261 | * This is written to the superblock. |
| 262 | * If reshape_position is MaxSector, then no reshape is happening (yet). |
| 263 | */ |
| 264 | sector_t reshape_position; |
Andre Noll | 664e7c4 | 2009-06-18 08:45:27 +1000 | [diff] [blame] | 265 | int delta_disks, new_level, new_layout; |
| 266 | int new_chunk_sectors; |
NeilBrown | 2c810cd | 2012-05-21 09:27:00 +1000 | [diff] [blame] | 267 | int reshape_backwards; |
NeilBrown | f670557 | 2006-03-27 01:18:11 -0800 | [diff] [blame] | 268 | |
NeilBrown | 2b8bf34 | 2011-10-11 16:48:23 +1100 | [diff] [blame] | 269 | struct md_thread *thread; /* management thread */ |
| 270 | struct md_thread *sync_thread; /* doing resync or reconstruct */ |
Jonathan Brassow | c4a3955 | 2013-06-25 01:23:59 -0500 | [diff] [blame] | 271 | |
| 272 | /* 'last_sync_action' is initialized to "none". It is set when a |
| 273 | * sync operation (i.e "data-check", "requested-resync", "resync", |
| 274 | * "recovery", or "reshape") is started. It holds this value even |
| 275 | * when the sync thread is "frozen" (interrupted) or "idle" (stopped |
| 276 | * or finished). It is overwritten when a new sync operation is begun. |
| 277 | */ |
| 278 | char *last_sync_action; |
NeilBrown | ff4e8d9 | 2006-07-10 04:44:16 -0700 | [diff] [blame] | 279 | sector_t curr_resync; /* last block scheduled */ |
NeilBrown | 97e4f42 | 2009-03-31 14:33:13 +1100 | [diff] [blame] | 280 | /* As resync requests can complete out of order, we cannot easily track |
| 281 | * how much resync has been completed. So we occasionally pause until |
| 282 | * everything completes, then set curr_resync_completed to curr_resync. |
| 283 | * As such it may be well behind the real resync mark, but it is a value |
| 284 | * we are certain of. |
| 285 | */ |
| 286 | sector_t curr_resync_completed; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 287 | unsigned long resync_mark; /* a recent timestamp */ |
| 288 | sector_t resync_mark_cnt;/* blocks written at resync_mark */ |
NeilBrown | ff4e8d9 | 2006-07-10 04:44:16 -0700 | [diff] [blame] | 289 | sector_t curr_mark_cnt; /* blocks scheduled now */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 290 | |
| 291 | sector_t resync_max_sectors; /* may be set by personality */ |
NeilBrown | 9d88883 | 2005-11-08 21:39:26 -0800 | [diff] [blame] | 292 | |
Jianpeng Ma | 7f7583d | 2012-10-11 14:17:59 +1100 | [diff] [blame] | 293 | atomic64_t resync_mismatches; /* count of sectors where |
NeilBrown | 9d88883 | 2005-11-08 21:39:26 -0800 | [diff] [blame] | 294 | * parity/replica mismatch found |
| 295 | */ |
NeilBrown | e464eaf | 2006-03-27 01:18:14 -0800 | [diff] [blame] | 296 | |
| 297 | /* allow user-space to request suspension of IO to regions of the array */ |
| 298 | sector_t suspend_lo; |
| 299 | sector_t suspend_hi; |
NeilBrown | 88202a0 | 2006-01-06 00:21:36 -0800 | [diff] [blame] | 300 | /* if zero, use the system-wide default */ |
| 301 | int sync_speed_min; |
| 302 | int sync_speed_max; |
| 303 | |
Bernd Schubert | 90b0871 | 2008-05-23 13:04:38 -0700 | [diff] [blame] | 304 | /* resync even though the same disks are shared among md-devices */ |
| 305 | int parallel_resync; |
| 306 | |
NeilBrown | 6ff8d8ec | 2006-01-06 00:20:15 -0800 | [diff] [blame] | 307 | int ok_start_degraded; |
NeilBrown | f72ffdd | 2014-09-30 14:23:59 +1000 | [diff] [blame] | 308 | /* recovery/resync flags |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 309 | * NEEDED: we might need to start a resync/recover |
| 310 | * RUNNING: a thread is running, or about to be started |
| 311 | * SYNC: actually doing a resync, not a recovery |
Neil Brown | 72a23c2 | 2008-06-28 08:31:41 +1000 | [diff] [blame] | 312 | * RECOVER: doing recovery, or need to try it. |
NeilBrown | dfc7064 | 2008-05-23 13:04:39 -0700 | [diff] [blame] | 313 | * INTR: resync needs to be aborted for some reason |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 314 | * DONE: thread is done and is waiting to be reaped |
NeilBrown | 24dd469 | 2005-11-08 21:39:26 -0800 | [diff] [blame] | 315 | * REQUEST: user-space has requested a sync (used with SYNC) |
Anand Gadiyar | 411c940 | 2009-07-07 15:24:23 +0530 | [diff] [blame] | 316 | * CHECK: user-space request for check-only, no repair |
NeilBrown | ccfcc3c | 2006-03-27 01:18:09 -0800 | [diff] [blame] | 317 | * RESHAPE: A reshape is happening |
majianpeng | 0a19caa | 2012-11-19 19:57:34 +0800 | [diff] [blame] | 318 | * ERROR: sync-action interrupted because io-error |
NeilBrown | ccfcc3c | 2006-03-27 01:18:09 -0800 | [diff] [blame] | 319 | * |
| 320 | * If neither SYNC or RESHAPE are set, then it is a recovery. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 321 | */ |
| 322 | #define MD_RECOVERY_RUNNING 0 |
| 323 | #define MD_RECOVERY_SYNC 1 |
Neil Brown | 72a23c2 | 2008-06-28 08:31:41 +1000 | [diff] [blame] | 324 | #define MD_RECOVERY_RECOVER 2 |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 325 | #define MD_RECOVERY_INTR 3 |
| 326 | #define MD_RECOVERY_DONE 4 |
| 327 | #define MD_RECOVERY_NEEDED 5 |
NeilBrown | 24dd469 | 2005-11-08 21:39:26 -0800 | [diff] [blame] | 328 | #define MD_RECOVERY_REQUESTED 6 |
| 329 | #define MD_RECOVERY_CHECK 7 |
NeilBrown | ccfcc3c | 2006-03-27 01:18:09 -0800 | [diff] [blame] | 330 | #define MD_RECOVERY_RESHAPE 8 |
NeilBrown | 5fd6c1d | 2006-06-26 00:27:40 -0700 | [diff] [blame] | 331 | #define MD_RECOVERY_FROZEN 9 |
majianpeng | 0a19caa | 2012-11-19 19:57:34 +0800 | [diff] [blame] | 332 | #define MD_RECOVERY_ERROR 10 |
NeilBrown | 5fd6c1d | 2006-06-26 00:27:40 -0700 | [diff] [blame] | 333 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 334 | unsigned long recovery; |
NeilBrown | 5389042 | 2011-07-27 11:00:36 +1000 | [diff] [blame] | 335 | /* If a RAID personality determines that recovery (of a particular |
| 336 | * device) will fail due to a read error on the source device, it |
| 337 | * takes a copy of this number and does not attempt recovery again |
| 338 | * until this number changes. |
| 339 | */ |
| 340 | int recovery_disabled; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 341 | |
| 342 | int in_sync; /* know to not need resync */ |
NeilBrown | c8c00a6 | 2009-08-10 12:50:52 +1000 | [diff] [blame] | 343 | /* 'open_mutex' avoids races between 'md_open' and 'do_md_stop', so |
| 344 | * that we are never stopping an array while it is open. |
| 345 | * 'reconfig_mutex' protects all other reconfiguration. |
| 346 | * These locks are separate due to conflicting interactions |
| 347 | * with bdev->bd_mutex. |
| 348 | * Lock ordering is: |
| 349 | * reconfig_mutex -> bd_mutex : e.g. do_md_run -> revalidate_disk |
| 350 | * bd_mutex -> open_mutex: e.g. __blkdev_get -> md_open |
| 351 | */ |
| 352 | struct mutex open_mutex; |
NeilBrown | df5b89b | 2006-03-27 01:18:20 -0800 | [diff] [blame] | 353 | struct mutex reconfig_mutex; |
NeilBrown | f2ea68c | 2008-07-21 17:05:25 +1000 | [diff] [blame] | 354 | atomic_t active; /* general refcount */ |
| 355 | atomic_t openers; /* number of active opens */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 356 | |
NeilBrown | f0b4f7e | 2011-02-24 17:26:41 +1100 | [diff] [blame] | 357 | int changed; /* True if we might need to |
| 358 | * reread partition info */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 359 | int degraded; /* whether md should consider |
| 360 | * adding a spare |
| 361 | */ |
| 362 | |
| 363 | atomic_t recovery_active; /* blocks scheduled, but not written */ |
| 364 | wait_queue_head_t recovery_wait; |
| 365 | sector_t recovery_cp; |
Neil Brown | 5e96ee6 | 2008-06-28 08:31:24 +1000 | [diff] [blame] | 366 | sector_t resync_min; /* user requested sync |
| 367 | * starts here */ |
NeilBrown | c620727 | 2008-02-06 01:39:52 -0800 | [diff] [blame] | 368 | sector_t resync_max; /* resync should pause |
| 369 | * when it gets here */ |
NeilBrown | 06d91a5 | 2005-06-21 17:17:12 -0700 | [diff] [blame] | 370 | |
Tejun Heo | 324a56e | 2013-12-11 14:11:53 -0500 | [diff] [blame] | 371 | struct kernfs_node *sysfs_state; /* handle for 'array_state' |
NeilBrown | b62b759 | 2008-10-21 13:25:21 +1100 | [diff] [blame] | 372 | * file in sysfs. |
| 373 | */ |
Tejun Heo | 324a56e | 2013-12-11 14:11:53 -0500 | [diff] [blame] | 374 | struct kernfs_node *sysfs_action; /* handle for 'sync_action' */ |
NeilBrown | b62b759 | 2008-10-21 13:25:21 +1100 | [diff] [blame] | 375 | |
NeilBrown | d337482 | 2009-01-09 08:31:10 +1100 | [diff] [blame] | 376 | struct work_struct del_work; /* used for delayed sysfs removal */ |
| 377 | |
NeilBrown | 85572d7 | 2014-12-15 12:56:56 +1100 | [diff] [blame] | 378 | /* "lock" protects: |
| 379 | * flush_bio transition from NULL to !NULL |
| 380 | * rdev superblocks, events |
| 381 | * clearing MD_CHANGE_* |
| 382 | * in_sync - and related safemode and MD_CHANGE changes |
NeilBrown | 36d091f | 2014-12-15 12:56:58 +1100 | [diff] [blame] | 383 | * pers (also protected by reconfig_mutex and pending IO). |
NeilBrown | 978a7a4 | 2014-12-15 12:56:58 +1100 | [diff] [blame] | 384 | * clearing ->bitmap |
NeilBrown | 4af1a04 | 2014-12-15 12:57:00 +1100 | [diff] [blame] | 385 | * clearing ->bitmap_info.file |
NeilBrown | 23da422 | 2014-12-15 12:57:01 +1100 | [diff] [blame] | 386 | * changing ->resync_{min,max} |
| 387 | * setting MD_RECOVERY_RUNNING (which interacts with resync_{min,max}) |
NeilBrown | 85572d7 | 2014-12-15 12:56:56 +1100 | [diff] [blame] | 388 | */ |
| 389 | spinlock_t lock; |
NeilBrown | 3d310eb | 2005-06-21 17:17:26 -0700 | [diff] [blame] | 390 | wait_queue_head_t sb_wait; /* for waiting on superblock updates */ |
NeilBrown | 7bfa19f | 2005-06-21 17:17:28 -0700 | [diff] [blame] | 391 | atomic_t pending_writes; /* number of active superblock writes */ |
NeilBrown | 06d91a5 | 2005-06-21 17:17:12 -0700 | [diff] [blame] | 392 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 393 | unsigned int safemode; /* if set, update "clean" superblock |
| 394 | * when no writes pending. |
NeilBrown | f72ffdd | 2014-09-30 14:23:59 +1000 | [diff] [blame] | 395 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 396 | unsigned int safemode_delay; |
| 397 | struct timer_list safemode_timer; |
NeilBrown | f72ffdd | 2014-09-30 14:23:59 +1000 | [diff] [blame] | 398 | atomic_t writes_pending; |
Jens Axboe | 165125e | 2007-07-24 09:28:11 +0200 | [diff] [blame] | 399 | struct request_queue *queue; /* for plugging ... */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 400 | |
NeilBrown | f72ffdd | 2014-09-30 14:23:59 +1000 | [diff] [blame] | 401 | struct bitmap *bitmap; /* the bitmap for the device */ |
NeilBrown | c3d9714 | 2009-12-14 12:49:52 +1100 | [diff] [blame] | 402 | struct { |
| 403 | struct file *file; /* the bitmap file */ |
NeilBrown | f6af949 | 2009-12-14 12:49:54 +1100 | [diff] [blame] | 404 | loff_t offset; /* offset from superblock of |
NeilBrown | c3d9714 | 2009-12-14 12:49:52 +1100 | [diff] [blame] | 405 | * start of bitmap. May be |
| 406 | * negative, but not '0' |
NeilBrown | f6af949 | 2009-12-14 12:49:54 +1100 | [diff] [blame] | 407 | * For external metadata, offset |
NeilBrown | f72ffdd | 2014-09-30 14:23:59 +1000 | [diff] [blame] | 408 | * from start of device. |
NeilBrown | c3d9714 | 2009-12-14 12:49:52 +1100 | [diff] [blame] | 409 | */ |
NeilBrown | 6409bb0 | 2012-05-22 13:55:07 +1000 | [diff] [blame] | 410 | unsigned long space; /* space available at this offset */ |
NeilBrown | f6af949 | 2009-12-14 12:49:54 +1100 | [diff] [blame] | 411 | loff_t default_offset; /* this is the offset to use when |
NeilBrown | c3d9714 | 2009-12-14 12:49:52 +1100 | [diff] [blame] | 412 | * hot-adding a bitmap. It should |
| 413 | * eventually be settable by sysfs. |
| 414 | */ |
NeilBrown | 6409bb0 | 2012-05-22 13:55:07 +1000 | [diff] [blame] | 415 | unsigned long default_space; /* space available at |
| 416 | * default offset */ |
NeilBrown | c3d9714 | 2009-12-14 12:49:52 +1100 | [diff] [blame] | 417 | struct mutex mutex; |
NeilBrown | 42a04b5 | 2009-12-14 12:49:53 +1100 | [diff] [blame] | 418 | unsigned long chunksize; |
NeilBrown | ac2f40b | 2010-06-01 19:37:31 +1000 | [diff] [blame] | 419 | unsigned long daemon_sleep; /* how many jiffies between updates? */ |
NeilBrown | 42a04b5 | 2009-12-14 12:49:53 +1100 | [diff] [blame] | 420 | unsigned long max_write_behind; /* write-behind mode */ |
NeilBrown | ece5cff | 2009-12-14 12:49:56 +1100 | [diff] [blame] | 421 | int external; |
Goldwyn Rodrigues | c4ce867 | 2014-03-29 10:20:02 -0500 | [diff] [blame] | 422 | int nodes; /* Maximum number of nodes in the cluster */ |
Goldwyn Rodrigues | cf921cc | 2014-03-30 00:42:49 -0500 | [diff] [blame] | 423 | char cluster_name[64]; /* Name of the cluster */ |
NeilBrown | c3d9714 | 2009-12-14 12:49:52 +1100 | [diff] [blame] | 424 | } bitmap_info; |
NeilBrown | 32a7627 | 2005-06-21 17:17:14 -0700 | [diff] [blame] | 425 | |
NeilBrown | f72ffdd | 2014-09-30 14:23:59 +1000 | [diff] [blame] | 426 | atomic_t max_corr_read_errors; /* max read retries */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 427 | struct list_head all_mddevs; |
NeilBrown | a2826aa | 2009-12-14 12:49:49 +1100 | [diff] [blame] | 428 | |
NeilBrown | a64c876 | 2010-04-14 17:15:37 +1000 | [diff] [blame] | 429 | struct attribute_group *to_remove; |
NeilBrown | 252ac52 | 2010-06-01 19:37:29 +1000 | [diff] [blame] | 430 | |
NeilBrown | a167f66 | 2010-10-26 18:31:13 +1100 | [diff] [blame] | 431 | struct bio_set *bio_set; |
| 432 | |
Tejun Heo | e9c7469 | 2010-09-03 11:56:18 +0200 | [diff] [blame] | 433 | /* Generic flush handling. |
| 434 | * The last to finish preflush schedules a worker to submit |
| 435 | * the rest of the request (without the REQ_FLUSH flag). |
NeilBrown | a2826aa | 2009-12-14 12:49:49 +1100 | [diff] [blame] | 436 | */ |
Tejun Heo | e9c7469 | 2010-09-03 11:56:18 +0200 | [diff] [blame] | 437 | struct bio *flush_bio; |
NeilBrown | a2826aa | 2009-12-14 12:49:49 +1100 | [diff] [blame] | 438 | atomic_t flush_pending; |
Tejun Heo | e9c7469 | 2010-09-03 11:56:18 +0200 | [diff] [blame] | 439 | struct work_struct flush_work; |
NeilBrown | 768a418 | 2010-07-26 11:49:55 +1000 | [diff] [blame] | 440 | struct work_struct event_work; /* used by dm to report failure event */ |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 441 | void (*sync_super)(struct mddev *mddev, struct md_rdev *rdev); |
Goldwyn Rodrigues | c4ce867 | 2014-03-29 10:20:02 -0500 | [diff] [blame] | 442 | struct md_cluster_info *cluster_info; |
Guoqing Jiang | 15858fa | 2015-12-21 10:51:00 +1100 | [diff] [blame] | 443 | unsigned int good_device_nr; /* good device num within cluster raid */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 444 | }; |
| 445 | |
NeilBrown | 5c47daf | 2014-12-15 12:57:01 +1100 | [diff] [blame] | 446 | static inline int __must_check mddev_lock(struct mddev *mddev) |
| 447 | { |
| 448 | return mutex_lock_interruptible(&mddev->reconfig_mutex); |
| 449 | } |
| 450 | |
| 451 | /* Sometimes we need to take the lock in a situation where |
| 452 | * failure due to interrupts is not acceptable. |
| 453 | */ |
| 454 | static inline void mddev_lock_nointr(struct mddev *mddev) |
| 455 | { |
| 456 | mutex_lock(&mddev->reconfig_mutex); |
| 457 | } |
| 458 | |
| 459 | static inline int mddev_is_locked(struct mddev *mddev) |
| 460 | { |
| 461 | return mutex_is_locked(&mddev->reconfig_mutex); |
| 462 | } |
| 463 | |
| 464 | static inline int mddev_trylock(struct mddev *mddev) |
| 465 | { |
| 466 | return mutex_trylock(&mddev->reconfig_mutex); |
| 467 | } |
| 468 | extern void mddev_unlock(struct mddev *mddev); |
| 469 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 470 | static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors) |
| 471 | { |
NeilBrown | f72ffdd | 2014-09-30 14:23:59 +1000 | [diff] [blame] | 472 | atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 473 | } |
| 474 | |
NeilBrown | 84fc4b5 | 2011-10-11 16:49:58 +1100 | [diff] [blame] | 475 | struct md_personality |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 476 | { |
| 477 | char *name; |
NeilBrown | 2604b70 | 2006-01-06 00:20:36 -0800 | [diff] [blame] | 478 | int level; |
| 479 | struct list_head list; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 480 | struct module *owner; |
Linus Torvalds | b4fdcb0 | 2011-11-04 17:06:58 -0700 | [diff] [blame] | 481 | void (*make_request)(struct mddev *mddev, struct bio *bio); |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 482 | int (*run)(struct mddev *mddev); |
NeilBrown | afa0f55 | 2014-12-15 12:56:58 +1100 | [diff] [blame] | 483 | void (*free)(struct mddev *mddev, void *priv); |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 484 | void (*status)(struct seq_file *seq, struct mddev *mddev); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 485 | /* error_handler must set ->faulty and clear ->in_sync |
NeilBrown | f72ffdd | 2014-09-30 14:23:59 +1000 | [diff] [blame] | 486 | * if appropriate, and should abort recovery if needed |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 487 | */ |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 488 | void (*error_handler)(struct mddev *mddev, struct md_rdev *rdev); |
| 489 | int (*hot_add_disk) (struct mddev *mddev, struct md_rdev *rdev); |
NeilBrown | b8321b6 | 2011-12-23 10:17:51 +1100 | [diff] [blame] | 490 | int (*hot_remove_disk) (struct mddev *mddev, struct md_rdev *rdev); |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 491 | int (*spare_active) (struct mddev *mddev); |
NeilBrown | 0931479 | 2015-02-19 16:04:40 +1100 | [diff] [blame] | 492 | sector_t (*sync_request)(struct mddev *mddev, sector_t sector_nr, int *skipped); |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 493 | int (*resize) (struct mddev *mddev, sector_t sectors); |
| 494 | sector_t (*size) (struct mddev *mddev, sector_t sectors, int raid_disks); |
| 495 | int (*check_reshape) (struct mddev *mddev); |
| 496 | int (*start_reshape) (struct mddev *mddev); |
| 497 | void (*finish_reshape) (struct mddev *mddev); |
NeilBrown | 36fa306 | 2005-09-09 16:23:45 -0700 | [diff] [blame] | 498 | /* quiesce moves between quiescence states |
| 499 | * 0 - fully active |
| 500 | * 1 - no new requests allowed |
| 501 | * others - reserved |
| 502 | */ |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 503 | void (*quiesce) (struct mddev *mddev, int state); |
NeilBrown | 245f46c | 2009-03-31 14:39:39 +1100 | [diff] [blame] | 504 | /* takeover is used to transition an array from one |
| 505 | * personality to another. The new personality must be able |
| 506 | * to handle the data in the current layout. |
| 507 | * e.g. 2drive raid1 -> 2drive raid5 |
| 508 | * ndrive raid5 -> degraded n+1drive raid6 with special layout |
| 509 | * If the takeover succeeds, a new 'private' structure is returned. |
| 510 | * This needs to be installed and then ->run used to activate the |
| 511 | * array. |
| 512 | */ |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 513 | void *(*takeover) (struct mddev *mddev); |
NeilBrown | 5c675f8 | 2014-12-15 12:56:56 +1100 | [diff] [blame] | 514 | /* congested implements bdi.congested_fn(). |
| 515 | * Will not be called while array is 'suspended' */ |
| 516 | int (*congested)(struct mddev *mddev, int bits); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 517 | }; |
| 518 | |
NeilBrown | 007583c | 2005-11-08 21:39:30 -0800 | [diff] [blame] | 519 | struct md_sysfs_entry { |
| 520 | struct attribute attr; |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 521 | ssize_t (*show)(struct mddev *, char *); |
| 522 | ssize_t (*store)(struct mddev *, const char *, size_t); |
NeilBrown | 007583c | 2005-11-08 21:39:30 -0800 | [diff] [blame] | 523 | }; |
NeilBrown | 43a7050 | 2009-12-14 12:49:55 +1100 | [diff] [blame] | 524 | extern struct attribute_group md_bitmap_group; |
NeilBrown | 007583c | 2005-11-08 21:39:30 -0800 | [diff] [blame] | 525 | |
Tejun Heo | 324a56e | 2013-12-11 14:11:53 -0500 | [diff] [blame] | 526 | 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] | 527 | { |
| 528 | if (sd) |
Tejun Heo | 388975c | 2013-09-11 23:19:13 -0400 | [diff] [blame] | 529 | return sysfs_get_dirent(sd, name); |
NeilBrown | 00bcb4a | 2010-06-01 19:37:23 +1000 | [diff] [blame] | 530 | return sd; |
| 531 | } |
Tejun Heo | 324a56e | 2013-12-11 14:11:53 -0500 | [diff] [blame] | 532 | static inline void sysfs_notify_dirent_safe(struct kernfs_node *sd) |
NeilBrown | 00bcb4a | 2010-06-01 19:37:23 +1000 | [diff] [blame] | 533 | { |
| 534 | if (sd) |
| 535 | sysfs_notify_dirent(sd); |
| 536 | } |
| 537 | |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 538 | static inline char * mdname (struct mddev * mddev) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 539 | { |
| 540 | return mddev->gendisk ? mddev->gendisk->disk_name : "mdX"; |
| 541 | } |
| 542 | |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 543 | 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] | 544 | { |
| 545 | char nm[20]; |
Shaohua Li | 9b15603 | 2015-12-18 15:19:16 +1100 | [diff] [blame] | 546 | if (!test_bit(Replacement, &rdev->flags) && |
| 547 | !test_bit(Journal, &rdev->flags) && |
| 548 | mddev->kobj.sd) { |
NeilBrown | 2d78f8c | 2011-12-23 10:17:51 +1100 | [diff] [blame] | 549 | sprintf(nm, "rd%d", rdev->raid_disk); |
| 550 | return sysfs_create_link(&mddev->kobj, &rdev->kobj, nm); |
| 551 | } else |
| 552 | return 0; |
Namhyung Kim | 36fad85 | 2011-07-27 11:00:36 +1000 | [diff] [blame] | 553 | } |
| 554 | |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 555 | 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] | 556 | { |
| 557 | char nm[20]; |
Shaohua Li | 9b15603 | 2015-12-18 15:19:16 +1100 | [diff] [blame] | 558 | if (!test_bit(Replacement, &rdev->flags) && |
| 559 | !test_bit(Journal, &rdev->flags) && |
| 560 | mddev->kobj.sd) { |
NeilBrown | 2d78f8c | 2011-12-23 10:17:51 +1100 | [diff] [blame] | 561 | sprintf(nm, "rd%d", rdev->raid_disk); |
| 562 | sysfs_remove_link(&mddev->kobj, nm); |
| 563 | } |
Namhyung Kim | 36fad85 | 2011-07-27 11:00:36 +1000 | [diff] [blame] | 564 | } |
| 565 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 566 | /* |
| 567 | * iterates through some rdev ringlist. It's safe to remove the |
| 568 | * current 'rdev'. Dont touch 'tmp' though. |
| 569 | */ |
Cheng Renquan | 159ec1f | 2009-01-09 08:31:08 +1100 | [diff] [blame] | 570 | #define rdev_for_each_list(rdev, tmp, head) \ |
| 571 | list_for_each_entry_safe(rdev, tmp, head, same_set) |
| 572 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 573 | /* |
| 574 | * iterates through the 'same array disks' ringlist |
| 575 | */ |
NeilBrown | dafb20f | 2012-03-19 12:46:39 +1100 | [diff] [blame] | 576 | #define rdev_for_each(rdev, mddev) \ |
| 577 | list_for_each_entry(rdev, &((mddev)->disks), same_set) |
| 578 | |
| 579 | #define rdev_for_each_safe(rdev, tmp, mddev) \ |
Cheng Renquan | 159ec1f | 2009-01-09 08:31:08 +1100 | [diff] [blame] | 580 | list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 581 | |
NeilBrown | 4b80991 | 2008-07-21 17:05:25 +1000 | [diff] [blame] | 582 | #define rdev_for_each_rcu(rdev, mddev) \ |
| 583 | list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set) |
| 584 | |
NeilBrown | 2b8bf34 | 2011-10-11 16:48:23 +1100 | [diff] [blame] | 585 | struct md_thread { |
Shaohua Li | 4ed8731 | 2012-10-11 13:34:00 +1100 | [diff] [blame] | 586 | void (*run) (struct md_thread *thread); |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 587 | struct mddev *mddev; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 588 | wait_queue_head_t wqueue; |
NeilBrown | f72ffdd | 2014-09-30 14:23:59 +1000 | [diff] [blame] | 589 | unsigned long flags; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 590 | struct task_struct *tsk; |
NeilBrown | 32a7627 | 2005-06-21 17:17:14 -0700 | [diff] [blame] | 591 | unsigned long timeout; |
Shaohua Li | 4ed8731 | 2012-10-11 13:34:00 +1100 | [diff] [blame] | 592 | void *private; |
NeilBrown | 2b8bf34 | 2011-10-11 16:48:23 +1100 | [diff] [blame] | 593 | }; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 594 | |
| 595 | #define THREAD_WAKEUP 0 |
| 596 | |
NeilBrown | 1345b1d | 2006-01-06 00:20:40 -0800 | [diff] [blame] | 597 | static inline void safe_put_page(struct page *p) |
| 598 | { |
| 599 | if (p) put_page(p); |
| 600 | } |
| 601 | |
NeilBrown | 84fc4b5 | 2011-10-11 16:49:58 +1100 | [diff] [blame] | 602 | extern int register_md_personality(struct md_personality *p); |
| 603 | extern int unregister_md_personality(struct md_personality *p); |
Goldwyn Rodrigues | edb39c9 | 2014-03-29 10:01:53 -0500 | [diff] [blame] | 604 | extern int register_md_cluster_operations(struct md_cluster_operations *ops, |
| 605 | struct module *module); |
| 606 | extern int unregister_md_cluster_operations(void); |
| 607 | extern int md_setup_cluster(struct mddev *mddev, int nodes); |
| 608 | extern void md_cluster_stop(struct mddev *mddev); |
NeilBrown | 2b8bf34 | 2011-10-11 16:48:23 +1100 | [diff] [blame] | 609 | extern struct md_thread *md_register_thread( |
Shaohua Li | 4ed8731 | 2012-10-11 13:34:00 +1100 | [diff] [blame] | 610 | void (*run)(struct md_thread *thread), |
NeilBrown | 2b8bf34 | 2011-10-11 16:48:23 +1100 | [diff] [blame] | 611 | struct mddev *mddev, |
| 612 | const char *name); |
| 613 | extern void md_unregister_thread(struct md_thread **threadp); |
| 614 | extern void md_wakeup_thread(struct md_thread *thread); |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 615 | extern void md_check_recovery(struct mddev *mddev); |
Jonathan Brassow | a91d5ac | 2013-04-24 11:42:43 +1000 | [diff] [blame] | 616 | extern void md_reap_sync_thread(struct mddev *mddev); |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 617 | extern void md_write_start(struct mddev *mddev, struct bio *bi); |
| 618 | extern void md_write_end(struct mddev *mddev); |
| 619 | extern void md_done_sync(struct mddev *mddev, int blocks, int ok); |
| 620 | extern void md_error(struct mddev *mddev, struct md_rdev *rdev); |
NeilBrown | c6563a8 | 2012-05-21 09:27:00 +1000 | [diff] [blame] | 621 | extern void md_finish_reshape(struct mddev *mddev); |
NeilBrown | 9202295 | 2009-03-31 14:33:13 +1100 | [diff] [blame] | 622 | |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 623 | extern int mddev_congested(struct mddev *mddev, int bits); |
| 624 | extern void md_flush_request(struct mddev *mddev, struct bio *bio); |
| 625 | extern void md_super_write(struct mddev *mddev, struct md_rdev *rdev, |
NeilBrown | 9202295 | 2009-03-31 14:33:13 +1100 | [diff] [blame] | 626 | sector_t sector, int size, struct page *page); |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 627 | extern void md_super_wait(struct mddev *mddev); |
NeilBrown | f72ffdd | 2014-09-30 14:23:59 +1000 | [diff] [blame] | 628 | extern int sync_page_io(struct md_rdev *rdev, sector_t sector, int size, |
Jonathan Brassow | ccebd4c | 2011-01-14 09:14:33 +1100 | [diff] [blame] | 629 | struct page *page, int rw, bool metadata_op); |
Shaohua Li | 4ed8731 | 2012-10-11 13:34:00 +1100 | [diff] [blame] | 630 | extern void md_do_sync(struct md_thread *thread); |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 631 | extern void md_new_event(struct mddev *mddev); |
| 632 | extern int md_allow_write(struct mddev *mddev); |
| 633 | extern void md_wait_for_blocked_rdev(struct md_rdev *rdev, struct mddev *mddev); |
| 634 | extern void md_set_array_sectors(struct mddev *mddev, sector_t array_sectors); |
| 635 | extern int md_check_no_bitmap(struct mddev *mddev); |
| 636 | extern int md_integrity_register(struct mddev *mddev); |
Dan Williams | 1501efa | 2016-01-13 16:00:07 -0800 | [diff] [blame] | 637 | extern int md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev); |
NeilBrown | 72e0207 | 2009-12-14 12:49:55 +1100 | [diff] [blame] | 638 | 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] | 639 | |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 640 | extern void mddev_init(struct mddev *mddev); |
| 641 | extern int md_run(struct mddev *mddev); |
| 642 | extern void md_stop(struct mddev *mddev); |
| 643 | extern void md_stop_writes(struct mddev *mddev); |
NeilBrown | 3cb0300 | 2011-10-11 16:45:26 +1100 | [diff] [blame] | 644 | extern int md_rdev_init(struct md_rdev *rdev); |
NeilBrown | 545c879 | 2012-05-22 13:54:30 +1000 | [diff] [blame] | 645 | extern void md_rdev_clear(struct md_rdev *rdev); |
NeilBrown | 390ee60 | 2010-06-01 19:37:27 +1000 | [diff] [blame] | 646 | |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 647 | extern void mddev_suspend(struct mddev *mddev); |
| 648 | extern void mddev_resume(struct mddev *mddev); |
NeilBrown | a167f66 | 2010-10-26 18:31:13 +1100 | [diff] [blame] | 649 | extern struct bio *bio_clone_mddev(struct bio *bio, gfp_t gfp_mask, |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 650 | struct mddev *mddev); |
NeilBrown | a167f66 | 2010-10-26 18:31:13 +1100 | [diff] [blame] | 651 | extern struct bio *bio_alloc_mddev(gfp_t gfp_mask, int nr_iovecs, |
NeilBrown | fd01b88 | 2011-10-11 16:47:53 +1100 | [diff] [blame] | 652 | struct mddev *mddev); |
NeilBrown | 9cbb175 | 2012-07-31 09:08:14 +0200 | [diff] [blame] | 653 | |
NeilBrown | 74018dc | 2012-07-31 09:08:15 +0200 | [diff] [blame] | 654 | extern void md_unplug(struct blk_plug_cb *cb, bool from_schedule); |
Goldwyn Rodrigues | 70bcecd | 2015-08-21 10:33:39 -0500 | [diff] [blame] | 655 | extern void md_reload_sb(struct mddev *mddev, int raid_disk); |
Goldwyn Rodrigues | 1aee41f | 2014-10-29 18:51:31 -0500 | [diff] [blame] | 656 | extern void md_update_sb(struct mddev *mddev, int force); |
Goldwyn Rodrigues | fb56dfe | 2015-04-14 10:43:24 -0500 | [diff] [blame] | 657 | extern void md_kick_rdev_from_array(struct md_rdev * rdev); |
Goldwyn Rodrigues | 57d051d | 2015-04-14 10:43:55 -0500 | [diff] [blame] | 658 | struct md_rdev *md_find_rdev_nr_rcu(struct mddev *mddev, int nr); |
NeilBrown | 9cbb175 | 2012-07-31 09:08:14 +0200 | [diff] [blame] | 659 | static inline int mddev_check_plugged(struct mddev *mddev) |
| 660 | { |
| 661 | return !!blk_check_plugged(md_unplug, mddev, |
| 662 | sizeof(struct blk_plug_cb)); |
| 663 | } |
Hannes Reinecke | dfe15ac | 2012-07-26 11:12:18 +0200 | [diff] [blame] | 664 | |
| 665 | static inline void rdev_dec_pending(struct md_rdev *rdev, struct mddev *mddev) |
| 666 | { |
| 667 | int faulty = test_bit(Faulty, &rdev->flags); |
| 668 | if (atomic_dec_and_test(&rdev->nr_pending) && faulty) { |
| 669 | set_bit(MD_RECOVERY_NEEDED, &mddev->recovery); |
| 670 | md_wakeup_thread(mddev->thread); |
| 671 | } |
| 672 | } |
| 673 | |
Goldwyn Rodrigues | edb39c9 | 2014-03-29 10:01:53 -0500 | [diff] [blame] | 674 | extern struct md_cluster_operations *md_cluster_ops; |
Goldwyn Rodrigues | c4ce867 | 2014-03-29 10:20:02 -0500 | [diff] [blame] | 675 | static inline int mddev_is_clustered(struct mddev *mddev) |
| 676 | { |
| 677 | return mddev->cluster_info && mddev->bitmap_info.nodes > 1; |
| 678 | } |
Christoph Hellwig | 63fe081 | 2009-04-14 12:01:53 +1000 | [diff] [blame] | 679 | #endif /* _MD_MD_H */ |