| 1 | /* |
| 2 | md.h : kernel internal structure of the Linux MD driver |
| 3 | Copyright (C) 1996-98 Ingo Molnar, Gadi Oxman |
| 4 | |
| 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. |
| 9 | |
| 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 |
| 12 | Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. |
| 13 | */ |
| 14 | |
| 15 | #ifndef _MD_MD_H |
| 16 | #define _MD_MD_H |
| 17 | |
| 18 | #include <linux/blkdev.h> |
| 19 | #include <linux/backing-dev.h> |
| 20 | #include <linux/badblocks.h> |
| 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> |
| 28 | #include "md-cluster.h" |
| 29 | |
| 30 | #define MaxSector (~(sector_t)0) |
| 31 | |
| 32 | /* |
| 33 | * MD's 'extended' device |
| 34 | */ |
| 35 | struct md_rdev { |
| 36 | struct list_head same_set; /* RAID devices within the same set */ |
| 37 | |
| 38 | sector_t sectors; /* Device size (in 512bytes sectors) */ |
| 39 | struct mddev *mddev; /* RAID array if running */ |
| 40 | int last_events; /* IO event timestamp */ |
| 41 | |
| 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; |
| 48 | struct block_device *bdev; /* block device handle */ |
| 49 | |
| 50 | struct page *sb_page, *bb_page; |
| 51 | int sb_loaded; |
| 52 | __u64 sb_events; |
| 53 | sector_t data_offset; /* start of data in array */ |
| 54 | sector_t new_data_offset;/* only relevant while reshaping */ |
| 55 | sector_t sb_start; /* offset of the super block (in 512byte sectors) */ |
| 56 | int sb_size; /* bytes in the superblock */ |
| 57 | int preferred_minor; /* autorun support */ |
| 58 | |
| 59 | struct kobject kobj; |
| 60 | |
| 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 | */ |
| 71 | |
| 72 | unsigned long flags; /* bit set of 'enum flag_bits' bits. */ |
| 73 | wait_queue_head_t blocked_wait; |
| 74 | |
| 75 | int desc_nr; /* descriptor index in the superblock */ |
| 76 | int raid_disk; /* role of device in array */ |
| 77 | int new_raid_disk; /* role that the device will have in |
| 78 | * the array after a level-change completes. |
| 79 | */ |
| 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 | */ |
| 84 | union { |
| 85 | sector_t recovery_offset;/* If this device has been partially |
| 86 | * recovered, this is where we were |
| 87 | * up to. |
| 88 | */ |
| 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 | }; |
| 94 | |
| 95 | atomic_t nr_pending; /* number of pending requests. |
| 96 | * only maintained for arrays that |
| 97 | * support hot removal |
| 98 | */ |
| 99 | atomic_t read_errors; /* number of consecutive read errors that |
| 100 | * we have tried to ignore. |
| 101 | */ |
| 102 | struct timespec last_read_error; /* monotonic time since our |
| 103 | * last read error |
| 104 | */ |
| 105 | atomic_t corrected_errors; /* number of corrected read errors, |
| 106 | * for reporting to userspace and storing |
| 107 | * in superblock. |
| 108 | */ |
| 109 | struct work_struct del_work; /* used for delayed sysfs removal */ |
| 110 | |
| 111 | struct kernfs_node *sysfs_state; /* handle for 'state' |
| 112 | * sysfs entry */ |
| 113 | |
| 114 | struct badblocks badblocks; |
| 115 | }; |
| 116 | enum flag_bits { |
| 117 | Faulty, /* device is known to have a fault */ |
| 118 | In_sync, /* device is in_sync with rest of array */ |
| 119 | Bitmap_sync, /* ..actually, not quite In_sync. Need a |
| 120 | * bitmap-based recovery to get fully in sync |
| 121 | */ |
| 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 | */ |
| 156 | Candidate, /* For clustered environments only: |
| 157 | * This device is seen locally but not |
| 158 | * by the whole cluster |
| 159 | */ |
| 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 | */ |
| 165 | ClusterRemove, |
| 166 | }; |
| 167 | |
| 168 | static inline int is_badblock(struct md_rdev *rdev, sector_t s, int sectors, |
| 169 | sector_t *first_bad, int *bad_sectors) |
| 170 | { |
| 171 | if (unlikely(rdev->badblocks.count)) { |
| 172 | int rv = badblocks_check(&rdev->badblocks, rdev->data_offset + s, |
| 173 | sectors, |
| 174 | first_bad, bad_sectors); |
| 175 | if (rv) |
| 176 | *first_bad -= rdev->data_offset; |
| 177 | return rv; |
| 178 | } |
| 179 | return 0; |
| 180 | } |
| 181 | extern int rdev_set_badblocks(struct md_rdev *rdev, sector_t s, int sectors, |
| 182 | int is_new); |
| 183 | extern int rdev_clear_badblocks(struct md_rdev *rdev, sector_t s, int sectors, |
| 184 | int is_new); |
| 185 | struct md_cluster_info; |
| 186 | |
| 187 | struct mddev { |
| 188 | void *private; |
| 189 | struct md_personality *pers; |
| 190 | dev_t unit; |
| 191 | int md_minor; |
| 192 | struct list_head disks; |
| 193 | unsigned long flags; |
| 194 | #define MD_CHANGE_DEVS 0 /* Some device status has changed */ |
| 195 | #define MD_CHANGE_CLEAN 1 /* transition to or from 'clean' */ |
| 196 | #define MD_CHANGE_PENDING 2 /* switch from 'clean' to 'active' in progress */ |
| 197 | #define MD_UPDATE_SB_FLAGS (1 | 2 | 4) /* If these are set, md_update_sb needed */ |
| 198 | #define MD_ARRAY_FIRST_USE 3 /* First use of array, needs initialization */ |
| 199 | #define MD_STILL_CLOSED 4 /* If set, then array has not been opened since |
| 200 | * md_ioctl checked on it. |
| 201 | */ |
| 202 | #define MD_JOURNAL_CLEAN 5 /* A raid with journal is already clean */ |
| 203 | #define MD_HAS_JOURNAL 6 /* The raid array has journal feature set */ |
| 204 | #define MD_RELOAD_SB 7 /* Reload the superblock because another node |
| 205 | * updated it. |
| 206 | */ |
| 207 | #define MD_CLUSTER_RESYNC_LOCKED 8 /* cluster raid only, which means node |
| 208 | * already took resync lock, need to |
| 209 | * release the lock */ |
| 210 | |
| 211 | int suspended; |
| 212 | atomic_t active_io; |
| 213 | int ro; |
| 214 | int sysfs_active; /* set when sysfs deletes |
| 215 | * are happening, so run/ |
| 216 | * takeover/stop are not safe |
| 217 | */ |
| 218 | struct gendisk *gendisk; |
| 219 | |
| 220 | struct kobject kobj; |
| 221 | int hold_active; |
| 222 | #define UNTIL_IOCTL 1 |
| 223 | #define UNTIL_STOP 2 |
| 224 | |
| 225 | /* Superblock information */ |
| 226 | int major_version, |
| 227 | minor_version, |
| 228 | patch_version; |
| 229 | int persistent; |
| 230 | int external; /* metadata is |
| 231 | * managed externally */ |
| 232 | char metadata_type[17]; /* externally set*/ |
| 233 | int chunk_sectors; |
| 234 | time64_t ctime, utime; |
| 235 | int level, layout; |
| 236 | char clevel[16]; |
| 237 | int raid_disks; |
| 238 | int max_disks; |
| 239 | sector_t dev_sectors; /* used size of |
| 240 | * component devices */ |
| 241 | sector_t array_sectors; /* exported array size */ |
| 242 | int external_size; /* size managed |
| 243 | * externally */ |
| 244 | __u64 events; |
| 245 | /* If the last 'event' was simply a clean->dirty transition, and |
| 246 | * we didn't write it to the spares, then it is safe and simple |
| 247 | * to just decrement the event count on a dirty->clean transition. |
| 248 | * So we record that possibility here. |
| 249 | */ |
| 250 | int can_decrease_events; |
| 251 | |
| 252 | char uuid[16]; |
| 253 | |
| 254 | /* If the array is being reshaped, we need to record the |
| 255 | * new shape and an indication of where we are up to. |
| 256 | * This is written to the superblock. |
| 257 | * If reshape_position is MaxSector, then no reshape is happening (yet). |
| 258 | */ |
| 259 | sector_t reshape_position; |
| 260 | int delta_disks, new_level, new_layout; |
| 261 | int new_chunk_sectors; |
| 262 | int reshape_backwards; |
| 263 | |
| 264 | struct md_thread *thread; /* management thread */ |
| 265 | struct md_thread *sync_thread; /* doing resync or reconstruct */ |
| 266 | |
| 267 | /* 'last_sync_action' is initialized to "none". It is set when a |
| 268 | * sync operation (i.e "data-check", "requested-resync", "resync", |
| 269 | * "recovery", or "reshape") is started. It holds this value even |
| 270 | * when the sync thread is "frozen" (interrupted) or "idle" (stopped |
| 271 | * or finished). It is overwritten when a new sync operation is begun. |
| 272 | */ |
| 273 | char *last_sync_action; |
| 274 | sector_t curr_resync; /* last block scheduled */ |
| 275 | /* As resync requests can complete out of order, we cannot easily track |
| 276 | * how much resync has been completed. So we occasionally pause until |
| 277 | * everything completes, then set curr_resync_completed to curr_resync. |
| 278 | * As such it may be well behind the real resync mark, but it is a value |
| 279 | * we are certain of. |
| 280 | */ |
| 281 | sector_t curr_resync_completed; |
| 282 | unsigned long resync_mark; /* a recent timestamp */ |
| 283 | sector_t resync_mark_cnt;/* blocks written at resync_mark */ |
| 284 | sector_t curr_mark_cnt; /* blocks scheduled now */ |
| 285 | |
| 286 | sector_t resync_max_sectors; /* may be set by personality */ |
| 287 | |
| 288 | atomic64_t resync_mismatches; /* count of sectors where |
| 289 | * parity/replica mismatch found |
| 290 | */ |
| 291 | |
| 292 | /* allow user-space to request suspension of IO to regions of the array */ |
| 293 | sector_t suspend_lo; |
| 294 | sector_t suspend_hi; |
| 295 | /* if zero, use the system-wide default */ |
| 296 | int sync_speed_min; |
| 297 | int sync_speed_max; |
| 298 | |
| 299 | /* resync even though the same disks are shared among md-devices */ |
| 300 | int parallel_resync; |
| 301 | |
| 302 | int ok_start_degraded; |
| 303 | /* recovery/resync flags |
| 304 | * NEEDED: we might need to start a resync/recover |
| 305 | * RUNNING: a thread is running, or about to be started |
| 306 | * SYNC: actually doing a resync, not a recovery |
| 307 | * RECOVER: doing recovery, or need to try it. |
| 308 | * INTR: resync needs to be aborted for some reason |
| 309 | * DONE: thread is done and is waiting to be reaped |
| 310 | * REQUEST: user-space has requested a sync (used with SYNC) |
| 311 | * CHECK: user-space request for check-only, no repair |
| 312 | * RESHAPE: A reshape is happening |
| 313 | * ERROR: sync-action interrupted because io-error |
| 314 | * |
| 315 | * If neither SYNC or RESHAPE are set, then it is a recovery. |
| 316 | */ |
| 317 | #define MD_RECOVERY_RUNNING 0 |
| 318 | #define MD_RECOVERY_SYNC 1 |
| 319 | #define MD_RECOVERY_RECOVER 2 |
| 320 | #define MD_RECOVERY_INTR 3 |
| 321 | #define MD_RECOVERY_DONE 4 |
| 322 | #define MD_RECOVERY_NEEDED 5 |
| 323 | #define MD_RECOVERY_REQUESTED 6 |
| 324 | #define MD_RECOVERY_CHECK 7 |
| 325 | #define MD_RECOVERY_RESHAPE 8 |
| 326 | #define MD_RECOVERY_FROZEN 9 |
| 327 | #define MD_RECOVERY_ERROR 10 |
| 328 | |
| 329 | unsigned long recovery; |
| 330 | /* If a RAID personality determines that recovery (of a particular |
| 331 | * device) will fail due to a read error on the source device, it |
| 332 | * takes a copy of this number and does not attempt recovery again |
| 333 | * until this number changes. |
| 334 | */ |
| 335 | int recovery_disabled; |
| 336 | |
| 337 | int in_sync; /* know to not need resync */ |
| 338 | /* 'open_mutex' avoids races between 'md_open' and 'do_md_stop', so |
| 339 | * that we are never stopping an array while it is open. |
| 340 | * 'reconfig_mutex' protects all other reconfiguration. |
| 341 | * These locks are separate due to conflicting interactions |
| 342 | * with bdev->bd_mutex. |
| 343 | * Lock ordering is: |
| 344 | * reconfig_mutex -> bd_mutex : e.g. do_md_run -> revalidate_disk |
| 345 | * bd_mutex -> open_mutex: e.g. __blkdev_get -> md_open |
| 346 | */ |
| 347 | struct mutex open_mutex; |
| 348 | struct mutex reconfig_mutex; |
| 349 | atomic_t active; /* general refcount */ |
| 350 | atomic_t openers; /* number of active opens */ |
| 351 | |
| 352 | int changed; /* True if we might need to |
| 353 | * reread partition info */ |
| 354 | int degraded; /* whether md should consider |
| 355 | * adding a spare |
| 356 | */ |
| 357 | |
| 358 | atomic_t recovery_active; /* blocks scheduled, but not written */ |
| 359 | wait_queue_head_t recovery_wait; |
| 360 | sector_t recovery_cp; |
| 361 | sector_t resync_min; /* user requested sync |
| 362 | * starts here */ |
| 363 | sector_t resync_max; /* resync should pause |
| 364 | * when it gets here */ |
| 365 | |
| 366 | struct kernfs_node *sysfs_state; /* handle for 'array_state' |
| 367 | * file in sysfs. |
| 368 | */ |
| 369 | struct kernfs_node *sysfs_action; /* handle for 'sync_action' */ |
| 370 | |
| 371 | struct work_struct del_work; /* used for delayed sysfs removal */ |
| 372 | |
| 373 | /* "lock" protects: |
| 374 | * flush_bio transition from NULL to !NULL |
| 375 | * rdev superblocks, events |
| 376 | * clearing MD_CHANGE_* |
| 377 | * in_sync - and related safemode and MD_CHANGE changes |
| 378 | * pers (also protected by reconfig_mutex and pending IO). |
| 379 | * clearing ->bitmap |
| 380 | * clearing ->bitmap_info.file |
| 381 | * changing ->resync_{min,max} |
| 382 | * setting MD_RECOVERY_RUNNING (which interacts with resync_{min,max}) |
| 383 | */ |
| 384 | spinlock_t lock; |
| 385 | wait_queue_head_t sb_wait; /* for waiting on superblock updates */ |
| 386 | atomic_t pending_writes; /* number of active superblock writes */ |
| 387 | |
| 388 | unsigned int safemode; /* if set, update "clean" superblock |
| 389 | * when no writes pending. |
| 390 | */ |
| 391 | unsigned int safemode_delay; |
| 392 | struct timer_list safemode_timer; |
| 393 | atomic_t writes_pending; |
| 394 | struct request_queue *queue; /* for plugging ... */ |
| 395 | |
| 396 | struct bitmap *bitmap; /* the bitmap for the device */ |
| 397 | struct { |
| 398 | struct file *file; /* the bitmap file */ |
| 399 | loff_t offset; /* offset from superblock of |
| 400 | * start of bitmap. May be |
| 401 | * negative, but not '0' |
| 402 | * For external metadata, offset |
| 403 | * from start of device. |
| 404 | */ |
| 405 | unsigned long space; /* space available at this offset */ |
| 406 | loff_t default_offset; /* this is the offset to use when |
| 407 | * hot-adding a bitmap. It should |
| 408 | * eventually be settable by sysfs. |
| 409 | */ |
| 410 | unsigned long default_space; /* space available at |
| 411 | * default offset */ |
| 412 | struct mutex mutex; |
| 413 | unsigned long chunksize; |
| 414 | unsigned long daemon_sleep; /* how many jiffies between updates? */ |
| 415 | unsigned long max_write_behind; /* write-behind mode */ |
| 416 | int external; |
| 417 | int nodes; /* Maximum number of nodes in the cluster */ |
| 418 | char cluster_name[64]; /* Name of the cluster */ |
| 419 | } bitmap_info; |
| 420 | |
| 421 | atomic_t max_corr_read_errors; /* max read retries */ |
| 422 | struct list_head all_mddevs; |
| 423 | |
| 424 | struct attribute_group *to_remove; |
| 425 | |
| 426 | struct bio_set *bio_set; |
| 427 | |
| 428 | /* Generic flush handling. |
| 429 | * The last to finish preflush schedules a worker to submit |
| 430 | * the rest of the request (without the REQ_FLUSH flag). |
| 431 | */ |
| 432 | struct bio *flush_bio; |
| 433 | atomic_t flush_pending; |
| 434 | struct work_struct flush_work; |
| 435 | struct work_struct event_work; /* used by dm to report failure event */ |
| 436 | void (*sync_super)(struct mddev *mddev, struct md_rdev *rdev); |
| 437 | struct md_cluster_info *cluster_info; |
| 438 | unsigned int good_device_nr; /* good device num within cluster raid */ |
| 439 | }; |
| 440 | |
| 441 | static inline int __must_check mddev_lock(struct mddev *mddev) |
| 442 | { |
| 443 | return mutex_lock_interruptible(&mddev->reconfig_mutex); |
| 444 | } |
| 445 | |
| 446 | /* Sometimes we need to take the lock in a situation where |
| 447 | * failure due to interrupts is not acceptable. |
| 448 | */ |
| 449 | static inline void mddev_lock_nointr(struct mddev *mddev) |
| 450 | { |
| 451 | mutex_lock(&mddev->reconfig_mutex); |
| 452 | } |
| 453 | |
| 454 | static inline int mddev_is_locked(struct mddev *mddev) |
| 455 | { |
| 456 | return mutex_is_locked(&mddev->reconfig_mutex); |
| 457 | } |
| 458 | |
| 459 | static inline int mddev_trylock(struct mddev *mddev) |
| 460 | { |
| 461 | return mutex_trylock(&mddev->reconfig_mutex); |
| 462 | } |
| 463 | extern void mddev_unlock(struct mddev *mddev); |
| 464 | |
| 465 | static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors) |
| 466 | { |
| 467 | atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io); |
| 468 | } |
| 469 | |
| 470 | struct md_personality |
| 471 | { |
| 472 | char *name; |
| 473 | int level; |
| 474 | struct list_head list; |
| 475 | struct module *owner; |
| 476 | void (*make_request)(struct mddev *mddev, struct bio *bio); |
| 477 | int (*run)(struct mddev *mddev); |
| 478 | void (*free)(struct mddev *mddev, void *priv); |
| 479 | void (*status)(struct seq_file *seq, struct mddev *mddev); |
| 480 | /* error_handler must set ->faulty and clear ->in_sync |
| 481 | * if appropriate, and should abort recovery if needed |
| 482 | */ |
| 483 | void (*error_handler)(struct mddev *mddev, struct md_rdev *rdev); |
| 484 | int (*hot_add_disk) (struct mddev *mddev, struct md_rdev *rdev); |
| 485 | int (*hot_remove_disk) (struct mddev *mddev, struct md_rdev *rdev); |
| 486 | int (*spare_active) (struct mddev *mddev); |
| 487 | sector_t (*sync_request)(struct mddev *mddev, sector_t sector_nr, int *skipped); |
| 488 | int (*resize) (struct mddev *mddev, sector_t sectors); |
| 489 | sector_t (*size) (struct mddev *mddev, sector_t sectors, int raid_disks); |
| 490 | int (*check_reshape) (struct mddev *mddev); |
| 491 | int (*start_reshape) (struct mddev *mddev); |
| 492 | void (*finish_reshape) (struct mddev *mddev); |
| 493 | /* quiesce moves between quiescence states |
| 494 | * 0 - fully active |
| 495 | * 1 - no new requests allowed |
| 496 | * others - reserved |
| 497 | */ |
| 498 | void (*quiesce) (struct mddev *mddev, int state); |
| 499 | /* takeover is used to transition an array from one |
| 500 | * personality to another. The new personality must be able |
| 501 | * to handle the data in the current layout. |
| 502 | * e.g. 2drive raid1 -> 2drive raid5 |
| 503 | * ndrive raid5 -> degraded n+1drive raid6 with special layout |
| 504 | * If the takeover succeeds, a new 'private' structure is returned. |
| 505 | * This needs to be installed and then ->run used to activate the |
| 506 | * array. |
| 507 | */ |
| 508 | void *(*takeover) (struct mddev *mddev); |
| 509 | /* congested implements bdi.congested_fn(). |
| 510 | * Will not be called while array is 'suspended' */ |
| 511 | int (*congested)(struct mddev *mddev, int bits); |
| 512 | }; |
| 513 | |
| 514 | struct md_sysfs_entry { |
| 515 | struct attribute attr; |
| 516 | ssize_t (*show)(struct mddev *, char *); |
| 517 | ssize_t (*store)(struct mddev *, const char *, size_t); |
| 518 | }; |
| 519 | extern struct attribute_group md_bitmap_group; |
| 520 | |
| 521 | static inline struct kernfs_node *sysfs_get_dirent_safe(struct kernfs_node *sd, char *name) |
| 522 | { |
| 523 | if (sd) |
| 524 | return sysfs_get_dirent(sd, name); |
| 525 | return sd; |
| 526 | } |
| 527 | static inline void sysfs_notify_dirent_safe(struct kernfs_node *sd) |
| 528 | { |
| 529 | if (sd) |
| 530 | sysfs_notify_dirent(sd); |
| 531 | } |
| 532 | |
| 533 | static inline char * mdname (struct mddev * mddev) |
| 534 | { |
| 535 | return mddev->gendisk ? mddev->gendisk->disk_name : "mdX"; |
| 536 | } |
| 537 | |
| 538 | static inline int sysfs_link_rdev(struct mddev *mddev, struct md_rdev *rdev) |
| 539 | { |
| 540 | char nm[20]; |
| 541 | if (!test_bit(Replacement, &rdev->flags) && |
| 542 | !test_bit(Journal, &rdev->flags) && |
| 543 | mddev->kobj.sd) { |
| 544 | sprintf(nm, "rd%d", rdev->raid_disk); |
| 545 | return sysfs_create_link(&mddev->kobj, &rdev->kobj, nm); |
| 546 | } else |
| 547 | return 0; |
| 548 | } |
| 549 | |
| 550 | static inline void sysfs_unlink_rdev(struct mddev *mddev, struct md_rdev *rdev) |
| 551 | { |
| 552 | char nm[20]; |
| 553 | if (!test_bit(Replacement, &rdev->flags) && |
| 554 | !test_bit(Journal, &rdev->flags) && |
| 555 | mddev->kobj.sd) { |
| 556 | sprintf(nm, "rd%d", rdev->raid_disk); |
| 557 | sysfs_remove_link(&mddev->kobj, nm); |
| 558 | } |
| 559 | } |
| 560 | |
| 561 | /* |
| 562 | * iterates through some rdev ringlist. It's safe to remove the |
| 563 | * current 'rdev'. Dont touch 'tmp' though. |
| 564 | */ |
| 565 | #define rdev_for_each_list(rdev, tmp, head) \ |
| 566 | list_for_each_entry_safe(rdev, tmp, head, same_set) |
| 567 | |
| 568 | /* |
| 569 | * iterates through the 'same array disks' ringlist |
| 570 | */ |
| 571 | #define rdev_for_each(rdev, mddev) \ |
| 572 | list_for_each_entry(rdev, &((mddev)->disks), same_set) |
| 573 | |
| 574 | #define rdev_for_each_safe(rdev, tmp, mddev) \ |
| 575 | list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set) |
| 576 | |
| 577 | #define rdev_for_each_rcu(rdev, mddev) \ |
| 578 | list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set) |
| 579 | |
| 580 | struct md_thread { |
| 581 | void (*run) (struct md_thread *thread); |
| 582 | struct mddev *mddev; |
| 583 | wait_queue_head_t wqueue; |
| 584 | unsigned long flags; |
| 585 | struct task_struct *tsk; |
| 586 | unsigned long timeout; |
| 587 | void *private; |
| 588 | }; |
| 589 | |
| 590 | #define THREAD_WAKEUP 0 |
| 591 | |
| 592 | static inline void safe_put_page(struct page *p) |
| 593 | { |
| 594 | if (p) put_page(p); |
| 595 | } |
| 596 | |
| 597 | extern int register_md_personality(struct md_personality *p); |
| 598 | extern int unregister_md_personality(struct md_personality *p); |
| 599 | extern int register_md_cluster_operations(struct md_cluster_operations *ops, |
| 600 | struct module *module); |
| 601 | extern int unregister_md_cluster_operations(void); |
| 602 | extern int md_setup_cluster(struct mddev *mddev, int nodes); |
| 603 | extern void md_cluster_stop(struct mddev *mddev); |
| 604 | extern struct md_thread *md_register_thread( |
| 605 | void (*run)(struct md_thread *thread), |
| 606 | struct mddev *mddev, |
| 607 | const char *name); |
| 608 | extern void md_unregister_thread(struct md_thread **threadp); |
| 609 | extern void md_wakeup_thread(struct md_thread *thread); |
| 610 | extern void md_check_recovery(struct mddev *mddev); |
| 611 | extern void md_reap_sync_thread(struct mddev *mddev); |
| 612 | extern void md_write_start(struct mddev *mddev, struct bio *bi); |
| 613 | extern void md_write_end(struct mddev *mddev); |
| 614 | extern void md_done_sync(struct mddev *mddev, int blocks, int ok); |
| 615 | extern void md_error(struct mddev *mddev, struct md_rdev *rdev); |
| 616 | extern void md_finish_reshape(struct mddev *mddev); |
| 617 | |
| 618 | extern int mddev_congested(struct mddev *mddev, int bits); |
| 619 | extern void md_flush_request(struct mddev *mddev, struct bio *bio); |
| 620 | extern void md_super_write(struct mddev *mddev, struct md_rdev *rdev, |
| 621 | sector_t sector, int size, struct page *page); |
| 622 | extern void md_super_wait(struct mddev *mddev); |
| 623 | extern int sync_page_io(struct md_rdev *rdev, sector_t sector, int size, |
| 624 | struct page *page, int rw, bool metadata_op); |
| 625 | extern void md_do_sync(struct md_thread *thread); |
| 626 | extern void md_new_event(struct mddev *mddev); |
| 627 | extern int md_allow_write(struct mddev *mddev); |
| 628 | extern void md_wait_for_blocked_rdev(struct md_rdev *rdev, struct mddev *mddev); |
| 629 | extern void md_set_array_sectors(struct mddev *mddev, sector_t array_sectors); |
| 630 | extern int md_check_no_bitmap(struct mddev *mddev); |
| 631 | extern int md_integrity_register(struct mddev *mddev); |
| 632 | extern int md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev); |
| 633 | extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale); |
| 634 | |
| 635 | extern void mddev_init(struct mddev *mddev); |
| 636 | extern int md_run(struct mddev *mddev); |
| 637 | extern void md_stop(struct mddev *mddev); |
| 638 | extern void md_stop_writes(struct mddev *mddev); |
| 639 | extern int md_rdev_init(struct md_rdev *rdev); |
| 640 | extern void md_rdev_clear(struct md_rdev *rdev); |
| 641 | |
| 642 | extern void mddev_suspend(struct mddev *mddev); |
| 643 | extern void mddev_resume(struct mddev *mddev); |
| 644 | extern struct bio *bio_clone_mddev(struct bio *bio, gfp_t gfp_mask, |
| 645 | struct mddev *mddev); |
| 646 | extern struct bio *bio_alloc_mddev(gfp_t gfp_mask, int nr_iovecs, |
| 647 | struct mddev *mddev); |
| 648 | |
| 649 | extern void md_unplug(struct blk_plug_cb *cb, bool from_schedule); |
| 650 | extern void md_reload_sb(struct mddev *mddev, int raid_disk); |
| 651 | extern void md_update_sb(struct mddev *mddev, int force); |
| 652 | extern void md_kick_rdev_from_array(struct md_rdev * rdev); |
| 653 | struct md_rdev *md_find_rdev_nr_rcu(struct mddev *mddev, int nr); |
| 654 | static inline int mddev_check_plugged(struct mddev *mddev) |
| 655 | { |
| 656 | return !!blk_check_plugged(md_unplug, mddev, |
| 657 | sizeof(struct blk_plug_cb)); |
| 658 | } |
| 659 | |
| 660 | static inline void rdev_dec_pending(struct md_rdev *rdev, struct mddev *mddev) |
| 661 | { |
| 662 | int faulty = test_bit(Faulty, &rdev->flags); |
| 663 | if (atomic_dec_and_test(&rdev->nr_pending) && faulty) { |
| 664 | set_bit(MD_RECOVERY_NEEDED, &mddev->recovery); |
| 665 | md_wakeup_thread(mddev->thread); |
| 666 | } |
| 667 | } |
| 668 | |
| 669 | extern struct md_cluster_operations *md_cluster_ops; |
| 670 | static inline int mddev_is_clustered(struct mddev *mddev) |
| 671 | { |
| 672 | return mddev->cluster_info && mddev->bitmap_info.nodes > 1; |
| 673 | } |
| 674 | #endif /* _MD_MD_H */ |