drivers/md/md.h

   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/kobject.h>
  21 #include <linux/list.h>
  22 #include <linux/mm.h>
  23 #include <linux/mutex.h>
  24 #include <linux/timer.h>
  25 #include <linux/wait.h>
  26 #include <linux/workqueue.h>
  27 #include "md-cluster.h"
  28
  29 #define MaxSector (~(sector_t)0)
  30
  31 /* Bad block numbers are stored sorted in a single page.
  32  * 64bits is used for each block or extent.
  33  * 54 bits are sector number, 9 bits are extent size,
  34  * 1 bit is an 'acknowledged' flag.
  35  */
  36 #define MD_MAX_BADBLOCKS        (PAGE_SIZE/8)
  37
  38 /*
  39  * MD's 'extended' device
  40  */
  41 struct md_rdev {
  42         struct list_head same_set;      /* RAID devices within the same set */
  43
  44         sector_t sectors;               /* Device size (in 512bytes sectors) */
  45         struct mddev *mddev;            /* RAID array if running */
  46         int last_events;                /* IO event timestamp */
  47
  48         /*
  49          * If meta_bdev is non-NULL, it means that a separate device is
  50          * being used to store the metadata (superblock/bitmap) which
  51          * would otherwise be contained on the same device as the data (bdev).
  52          */
  53         struct block_device *meta_bdev;
  54         struct block_device *bdev;      /* block device handle */
  55
  56         struct page     *sb_page, *bb_page;
  57         int             sb_loaded;
  58         __u64           sb_events;
  59         sector_t        data_offset;    /* start of data in array */
  60         sector_t        new_data_offset;/* only relevant while reshaping */
  61         sector_t        sb_start;       /* offset of the super block (in 512byte sectors) */
  62         int             sb_size;        /* bytes in the superblock */
  63         int             preferred_minor;        /* autorun support */
  64
  65         struct kobject  kobj;
  66
  67         /* A device can be in one of three states based on two flags:
  68          * Not working:   faulty==1 in_sync==0
  69          * Fully working: faulty==0 in_sync==1
  70          * Working, but not
  71          * in sync with array
  72          *                faulty==0 in_sync==0
  73          *
  74          * It can never have faulty==1, in_sync==1
  75          * This reduces the burden of testing multiple flags in many cases
  76          */
  77
  78         unsigned long   flags;  /* bit set of 'enum flag_bits' bits. */
  79         wait_queue_head_t blocked_wait;
  80
  81         int desc_nr;                    /* descriptor index in the superblock */
  82         int raid_disk;                  /* role of device in array */
  83         int new_raid_disk;              /* role that the device will have in
  84                                          * the array after a level-change completes.
  85                                          */
  86         int saved_raid_disk;            /* role that device used to have in the
  87                                          * array and could again if we did a partial
  88                                          * resync from the bitmap
  89                                          */
  90         union {
  91                 sector_t recovery_offset;/* If this device has been partially
  92                                          * recovered, this is where we were
  93                                          * up to.
  94                                          */
  95                 sector_t journal_tail;  /* If this device is a journal device,
  96                                          * this is the journal tail (journal
  97                                          * recovery start point)
  98                                          */
  99         };
 100
 101         atomic_t        nr_pending;     /* number of pending requests.
 102                                          * only maintained for arrays that
 103                                          * support hot removal
 104                                          */
 105         atomic_t        read_errors;    /* number of consecutive read errors that
 106                                          * we have tried to ignore.
 107                                          */
 108         struct timespec last_read_error;        /* monotonic time since our
 109                                                  * last read error
 110                                                  */
 111         atomic_t        corrected_errors; /* number of corrected read errors,
 112                                            * for reporting to userspace and storing
 113                                            * in superblock.
 114                                            */
 115         struct work_struct del_work;    /* used for delayed sysfs removal */
 116
 117         struct kernfs_node *sysfs_state; /* handle for 'state'
 118                                            * sysfs entry */
 119
 120         struct badblocks {
 121                 int     count;          /* count of bad blocks */
 122                 int     unacked_exist;  /* there probably are unacknowledged
 123                                          * bad blocks.  This is only cleared
 124                                          * when a read discovers none
 125                                          */
 126                 int     shift;          /* shift from sectors to block size
 127                                          * a -ve shift means badblocks are
 128                                          * disabled.*/
 129                 u64     *page;          /* badblock list */
 130                 int     changed;
 131                 seqlock_t lock;
 132
 133                 sector_t sector;
 134                 sector_t size;          /* in sectors */
 135         } badblocks;
 136 };
 137 enum flag_bits {
 138         Faulty,                 /* device is known to have a fault */
 139         In_sync,                /* device is in_sync with rest of array */
 140         Bitmap_sync,            /* ..actually, not quite In_sync.  Need a
 141                                  * bitmap-based recovery to get fully in sync
 142                                  */
 143         WriteMostly,            /* Avoid reading if at all possible */
 144         AutoDetected,           /* added by auto-detect */
 145         Blocked,                /* An error occurred but has not yet
 146                                  * been acknowledged by the metadata
 147                                  * handler, so don't allow writes
 148                                  * until it is cleared */
 149         WriteErrorSeen,         /* A write error has been seen on this
 150                                  * device
 151                                  */
 152         FaultRecorded,          /* Intermediate state for clearing
 153                                  * Blocked.  The Fault is/will-be
 154                                  * recorded in the metadata, but that
 155                                  * metadata hasn't been stored safely
 156                                  * on disk yet.
 157                                  */
 158         BlockedBadBlocks,       /* A writer is blocked because they
 159                                  * found an unacknowledged bad-block.
 160                                  * This can safely be cleared at any
 161                                  * time, and the writer will re-check.
 162                                  * It may be set at any time, and at
 163                                  * worst the writer will timeout and
 164                                  * re-check.  So setting it as
 165                                  * accurately as possible is good, but
 166                                  * not absolutely critical.
 167                                  */
 168         WantReplacement,        /* This device is a candidate to be
 169                                  * hot-replaced, either because it has
 170                                  * reported some faults, or because
 171                                  * of explicit request.
 172                                  */
 173         Replacement,            /* This device is a replacement for
 174                                  * a want_replacement device with same
 175                                  * raid_disk number.
 176                                  */
 177         Candidate,              /* For clustered environments only:
 178                                  * This device is seen locally but not
 179                                  * by the whole cluster
 180                                  */
 181         Journal,                /* This device is used as journal for
 182                                  * raid-5/6.
 183                                  * Usually, this device should be faster
 184                                  * than other devices in the array
 185                                  */
 186         ClusterRemove,
 187 };
 188
 189 #define BB_LEN_MASK     (0x00000000000001FFULL)
 190 #define BB_OFFSET_MASK  (0x7FFFFFFFFFFFFE00ULL)
 191 #define BB_ACK_MASK     (0x8000000000000000ULL)
 192 #define BB_MAX_LEN      512
 193 #define BB_OFFSET(x)    (((x) & BB_OFFSET_MASK) >> 9)
 194 #define BB_LEN(x)       (((x) & BB_LEN_MASK) + 1)
 195 #define BB_ACK(x)       (!!((x) & BB_ACK_MASK))
 196 #define BB_MAKE(a, l, ack) (((a)<<9) | ((l)-1) | ((u64)(!!(ack)) << 63))
 197
 198 extern int md_is_badblock(struct badblocks *bb, sector_t s, int sectors,
 199                           sector_t *first_bad, int *bad_sectors);
 200 static inline int is_badblock(struct md_rdev *rdev, sector_t s, int sectors,
 201                               sector_t *first_bad, int *bad_sectors)
 202 {
 203         if (unlikely(rdev->badblocks.count)) {
 204                 int rv = md_is_badblock(&rdev->badblocks, rdev->data_offset + s,
 205                                         sectors,
 206                                         first_bad, bad_sectors);
 207                 if (rv)
 208                         *first_bad -= rdev->data_offset;
 209                 return rv;
 210         }
 211         return 0;
 212 }
 213 extern int rdev_set_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
 214                               int is_new);
 215 extern int rdev_clear_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
 216                                 int is_new);
 217 extern void md_ack_all_badblocks(struct badblocks *bb);
 218
 219 struct md_cluster_info;
 220
 221 struct mddev {
 222         void                            *private;
 223         struct md_personality           *pers;
 224         dev_t                           unit;
 225         int                             md_minor;
 226         struct list_head                disks;
 227         unsigned long                   flags;
 228 #define MD_CHANGE_DEVS  0       /* Some device status has changed */
 229 #define MD_CHANGE_CLEAN 1       /* transition to or from 'clean' */
 230 #define MD_CHANGE_PENDING 2     /* switch from 'clean' to 'active' in progress */
 231 #define MD_UPDATE_SB_FLAGS (1 | 2 | 4)  /* If these are set, md_update_sb needed */
 232 #define MD_ARRAY_FIRST_USE 3    /* First use of array, needs initialization */
 233 #define MD_STILL_CLOSED 4       /* If set, then array has not been opened since
 234                                  * md_ioctl checked on it.
 235                                  */
 236 #define MD_JOURNAL_CLEAN 5      /* A raid with journal is already clean */
 237 #define MD_HAS_JOURNAL  6       /* The raid array has journal feature set */
 238 #define MD_RELOAD_SB    7       /* Reload the superblock because another node
 239                                  * updated it.
 240                                  */
 241
 242         int                             suspended;
 243         atomic_t                        active_io;
 244         int                             ro;
 245         int                             sysfs_active; /* set when sysfs deletes
 246                                                        * are happening, so run/
 247                                                        * takeover/stop are not safe
 248                                                        */
 249         struct gendisk                  *gendisk;
 250
 251         struct kobject                  kobj;
 252         int                             hold_active;
 253 #define UNTIL_IOCTL     1
 254 #define UNTIL_STOP      2
 255
 256         /* Superblock information */
 257         int                             major_version,
 258                                         minor_version,
 259                                         patch_version;
 260         int                             persistent;
 261         int                             external;       /* metadata is
 262                                                          * managed externally */
 263         char                            metadata_type[17]; /* externally set*/
 264         int                             chunk_sectors;
 265         time64_t                        ctime, utime;
 266         int                             level, layout;
 267         char                            clevel[16];
 268         int                             raid_disks;
 269         int                             max_disks;
 270         sector_t                        dev_sectors;    /* used size of
 271                                                          * component devices */
 272         sector_t                        array_sectors; /* exported array size */
 273         int                             external_size; /* size managed
 274                                                         * externally */
 275         __u64                           events;
 276         /* If the last 'event' was simply a clean->dirty transition, and
 277          * we didn't write it to the spares, then it is safe and simple
 278          * to just decrement the event count on a dirty->clean transition.
 279          * So we record that possibility here.
 280          */
 281         int                             can_decrease_events;
 282
 283         char                            uuid[16];
 284
 285         /* If the array is being reshaped, we need to record the
 286          * new shape and an indication of where we are up to.
 287          * This is written to the superblock.
 288          * If reshape_position is MaxSector, then no reshape is happening (yet).
 289          */
 290         sector_t                        reshape_position;
 291         int                             delta_disks, new_level, new_layout;
 292         int                             new_chunk_sectors;
 293         int                             reshape_backwards;
 294
 295         struct md_thread                *thread;        /* management thread */
 296         struct md_thread                *sync_thread;   /* doing resync or reconstruct */
 297
 298         /* 'last_sync_action' is initialized to "none".  It is set when a
 299          * sync operation (i.e "data-check", "requested-resync", "resync",
 300          * "recovery", or "reshape") is started.  It holds this value even
 301          * when the sync thread is "frozen" (interrupted) or "idle" (stopped
 302          * or finished).  It is overwritten when a new sync operation is begun.
 303          */
 304         char                            *last_sync_action;
 305         sector_t                        curr_resync;    /* last block scheduled */
 306         /* As resync requests can complete out of order, we cannot easily track
 307          * how much resync has been completed.  So we occasionally pause until
 308          * everything completes, then set curr_resync_completed to curr_resync.
 309          * As such it may be well behind the real resync mark, but it is a value
 310          * we are certain of.
 311          */
 312         sector_t                        curr_resync_completed;
 313         unsigned long                   resync_mark;    /* a recent timestamp */
 314         sector_t                        resync_mark_cnt;/* blocks written at resync_mark */
 315         sector_t                        curr_mark_cnt; /* blocks scheduled now */
 316
 317         sector_t                        resync_max_sectors; /* may be set by personality */
 318
 319         atomic64_t                      resync_mismatches; /* count of sectors where
 320                                                             * parity/replica mismatch found
 321                                                             */
 322
 323         /* allow user-space to request suspension of IO to regions of the array */
 324         sector_t                        suspend_lo;
 325         sector_t                        suspend_hi;
 326         /* if zero, use the system-wide default */
 327         int                             sync_speed_min;
 328         int                             sync_speed_max;
 329
 330         /* resync even though the same disks are shared among md-devices */
 331         int                             parallel_resync;
 332
 333         int                             ok_start_degraded;
 334         /* recovery/resync flags
 335          * NEEDED:   we might need to start a resync/recover
 336          * RUNNING:  a thread is running, or about to be started
 337          * SYNC:     actually doing a resync, not a recovery
 338          * RECOVER:  doing recovery, or need to try it.
 339          * INTR:     resync needs to be aborted for some reason
 340          * DONE:     thread is done and is waiting to be reaped
 341          * REQUEST:  user-space has requested a sync (used with SYNC)
 342          * CHECK:    user-space request for check-only, no repair
 343          * RESHAPE:  A reshape is happening
 344          * ERROR:    sync-action interrupted because io-error
 345          *
 346          * If neither SYNC or RESHAPE are set, then it is a recovery.
 347          */
 348 #define MD_RECOVERY_RUNNING     0
 349 #define MD_RECOVERY_SYNC        1
 350 #define MD_RECOVERY_RECOVER     2
 351 #define MD_RECOVERY_INTR        3
 352 #define MD_RECOVERY_DONE        4
 353 #define MD_RECOVERY_NEEDED      5
 354 #define MD_RECOVERY_REQUESTED   6
 355 #define MD_RECOVERY_CHECK       7
 356 #define MD_RECOVERY_RESHAPE     8
 357 #define MD_RECOVERY_FROZEN      9
 358 #define MD_RECOVERY_ERROR       10
 359
 360         unsigned long                   recovery;
 361         /* If a RAID personality determines that recovery (of a particular
 362          * device) will fail due to a read error on the source device, it
 363          * takes a copy of this number and does not attempt recovery again
 364          * until this number changes.
 365          */
 366         int                             recovery_disabled;
 367
 368         int                             in_sync;        /* know to not need resync */
 369         /* 'open_mutex' avoids races between 'md_open' and 'do_md_stop', so
 370          * that we are never stopping an array while it is open.
 371          * 'reconfig_mutex' protects all other reconfiguration.
 372          * These locks are separate due to conflicting interactions
 373          * with bdev->bd_mutex.
 374          * Lock ordering is:
 375          *  reconfig_mutex -> bd_mutex : e.g. do_md_run -> revalidate_disk
 376          *  bd_mutex -> open_mutex:  e.g. __blkdev_get -> md_open
 377          */
 378         struct mutex                    open_mutex;
 379         struct mutex                    reconfig_mutex;
 380         atomic_t                        active;         /* general refcount */
 381         atomic_t                        openers;        /* number of active opens */
 382
 383         int                             changed;        /* True if we might need to
 384                                                          * reread partition info */
 385         int                             degraded;       /* whether md should consider
 386                                                          * adding a spare
 387                                                          */
 388
 389         atomic_t                        recovery_active; /* blocks scheduled, but not written */
 390         wait_queue_head_t               recovery_wait;
 391         sector_t                        recovery_cp;
 392         sector_t                        resync_min;     /* user requested sync
 393                                                          * starts here */
 394         sector_t                        resync_max;     /* resync should pause
 395                                                          * when it gets here */
 396
 397         struct kernfs_node              *sysfs_state;   /* handle for 'array_state'
 398                                                          * file in sysfs.
 399                                                          */
 400         struct kernfs_node              *sysfs_action;  /* handle for 'sync_action' */
 401
 402         struct work_struct del_work;    /* used for delayed sysfs removal */
 403
 404         /* "lock" protects:
 405          *   flush_bio transition from NULL to !NULL
 406          *   rdev superblocks, events
 407          *   clearing MD_CHANGE_*
 408          *   in_sync - and related safemode and MD_CHANGE changes
 409          *   pers (also protected by reconfig_mutex and pending IO).
 410          *   clearing ->bitmap
 411          *   clearing ->bitmap_info.file
 412          *   changing ->resync_{min,max}
 413          *   setting MD_RECOVERY_RUNNING (which interacts with resync_{min,max})
 414          */
 415         spinlock_t                      lock;
 416         wait_queue_head_t               sb_wait;        /* for waiting on superblock updates */
 417         atomic_t                        pending_writes; /* number of active superblock writes */
 418
 419         unsigned int                    safemode;       /* if set, update "clean" superblock
 420                                                          * when no writes pending.
 421                                                          */
 422         unsigned int                    safemode_delay;
 423         struct timer_list               safemode_timer;
 424         atomic_t                        writes_pending;
 425         struct request_queue            *queue; /* for plugging ... */
 426
 427         struct bitmap                   *bitmap; /* the bitmap for the device */
 428         struct {
 429                 struct file             *file; /* the bitmap file */
 430                 loff_t                  offset; /* offset from superblock of
 431                                                  * start of bitmap. May be
 432                                                  * negative, but not '0'
 433                                                  * For external metadata, offset
 434                                                  * from start of device.
 435                                                  */
 436                 unsigned long           space; /* space available at this offset */
 437                 loff_t                  default_offset; /* this is the offset to use when
 438                                                          * hot-adding a bitmap.  It should
 439                                                          * eventually be settable by sysfs.
 440                                                          */
 441                 unsigned long           default_space; /* space available at
 442                                                         * default offset */
 443                 struct mutex            mutex;
 444                 unsigned long           chunksize;
 445                 unsigned long           daemon_sleep; /* how many jiffies between updates? */
 446                 unsigned long           max_write_behind; /* write-behind mode */
 447                 int                     external;
 448                 int                     nodes; /* Maximum number of nodes in the cluster */
 449                 char                    cluster_name[64]; /* Name of the cluster */
 450         } bitmap_info;
 451
 452         atomic_t                        max_corr_read_errors; /* max read retries */
 453         struct list_head                all_mddevs;
 454
 455         struct attribute_group          *to_remove;
 456
 457         struct bio_set                  *bio_set;
 458
 459         /* Generic flush handling.
 460          * The last to finish preflush schedules a worker to submit
 461          * the rest of the request (without the REQ_FLUSH flag).
 462          */
 463         struct bio *flush_bio;
 464         atomic_t flush_pending;
 465         struct work_struct flush_work;
 466         struct work_struct event_work;  /* used by dm to report failure event */
 467         void (*sync_super)(struct mddev *mddev, struct md_rdev *rdev);
 468         struct md_cluster_info          *cluster_info;
 469         unsigned int                    good_device_nr; /* good device num within cluster raid */
 470 };
 471
 472 static inline int __must_check mddev_lock(struct mddev *mddev)
 473 {
 474         return mutex_lock_interruptible(&mddev->reconfig_mutex);
 475 }
 476
 477 /* Sometimes we need to take the lock in a situation where
 478  * failure due to interrupts is not acceptable.
 479  */
 480 static inline void mddev_lock_nointr(struct mddev *mddev)
 481 {
 482         mutex_lock(&mddev->reconfig_mutex);
 483 }
 484
 485 static inline int mddev_is_locked(struct mddev *mddev)
 486 {
 487         return mutex_is_locked(&mddev->reconfig_mutex);
 488 }
 489
 490 static inline int mddev_trylock(struct mddev *mddev)
 491 {
 492         return mutex_trylock(&mddev->reconfig_mutex);
 493 }
 494 extern void mddev_unlock(struct mddev *mddev);
 495
 496 static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors)
 497 {
 498         atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io);
 499 }
 500
 501 struct md_personality
 502 {
 503         char *name;
 504         int level;
 505         struct list_head list;
 506         struct module *owner;
 507         void (*make_request)(struct mddev *mddev, struct bio *bio);
 508         int (*run)(struct mddev *mddev);
 509         void (*free)(struct mddev *mddev, void *priv);
 510         void (*status)(struct seq_file *seq, struct mddev *mddev);
 511         /* error_handler must set ->faulty and clear ->in_sync
 512          * if appropriate, and should abort recovery if needed
 513          */
 514         void (*error_handler)(struct mddev *mddev, struct md_rdev *rdev);
 515         int (*hot_add_disk) (struct mddev *mddev, struct md_rdev *rdev);
 516         int (*hot_remove_disk) (struct mddev *mddev, struct md_rdev *rdev);
 517         int (*spare_active) (struct mddev *mddev);
 518         sector_t (*sync_request)(struct mddev *mddev, sector_t sector_nr, int *skipped);
 519         int (*resize) (struct mddev *mddev, sector_t sectors);
 520         sector_t (*size) (struct mddev *mddev, sector_t sectors, int raid_disks);
 521         int (*check_reshape) (struct mddev *mddev);
 522         int (*start_reshape) (struct mddev *mddev);
 523         void (*finish_reshape) (struct mddev *mddev);
 524         /* quiesce moves between quiescence states
 525          * 0 - fully active
 526          * 1 - no new requests allowed
 527          * others - reserved
 528          */
 529         void (*quiesce) (struct mddev *mddev, int state);
 530         /* takeover is used to transition an array from one
 531          * personality to another.  The new personality must be able
 532          * to handle the data in the current layout.
 533          * e.g. 2drive raid1 -> 2drive raid5
 534          *      ndrive raid5 -> degraded n+1drive raid6 with special layout
 535          * If the takeover succeeds, a new 'private' structure is returned.
 536          * This needs to be installed and then ->run used to activate the
 537          * array.
 538          */
 539         void *(*takeover) (struct mddev *mddev);
 540         /* congested implements bdi.congested_fn().
 541          * Will not be called while array is 'suspended' */
 542         int (*congested)(struct mddev *mddev, int bits);
 543 };
 544
 545 struct md_sysfs_entry {
 546         struct attribute attr;
 547         ssize_t (*show)(struct mddev *, char *);
 548         ssize_t (*store)(struct mddev *, const char *, size_t);
 549 };
 550 extern struct attribute_group md_bitmap_group;
 551
 552 static inline struct kernfs_node *sysfs_get_dirent_safe(struct kernfs_node *sd, char *name)
 553 {
 554         if (sd)
 555                 return sysfs_get_dirent(sd, name);
 556         return sd;
 557 }
 558 static inline void sysfs_notify_dirent_safe(struct kernfs_node *sd)
 559 {
 560         if (sd)
 561                 sysfs_notify_dirent(sd);
 562 }
 563
 564 static inline char * mdname (struct mddev * mddev)
 565 {
 566         return mddev->gendisk ? mddev->gendisk->disk_name : "mdX";
 567 }
 568
 569 static inline int sysfs_link_rdev(struct mddev *mddev, struct md_rdev *rdev)
 570 {
 571         char nm[20];
 572         if (!test_bit(Replacement, &rdev->flags) &&
 573             !test_bit(Journal, &rdev->flags) &&
 574             mddev->kobj.sd) {
 575                 sprintf(nm, "rd%d", rdev->raid_disk);
 576                 return sysfs_create_link(&mddev->kobj, &rdev->kobj, nm);
 577         } else
 578                 return 0;
 579 }
 580
 581 static inline void sysfs_unlink_rdev(struct mddev *mddev, struct md_rdev *rdev)
 582 {
 583         char nm[20];
 584         if (!test_bit(Replacement, &rdev->flags) &&
 585             !test_bit(Journal, &rdev->flags) &&
 586             mddev->kobj.sd) {
 587                 sprintf(nm, "rd%d", rdev->raid_disk);
 588                 sysfs_remove_link(&mddev->kobj, nm);
 589         }
 590 }
 591
 592 /*
 593  * iterates through some rdev ringlist. It's safe to remove the
 594  * current 'rdev'. Dont touch 'tmp' though.
 595  */
 596 #define rdev_for_each_list(rdev, tmp, head)                             \
 597         list_for_each_entry_safe(rdev, tmp, head, same_set)
 598
 599 /*
 600  * iterates through the 'same array disks' ringlist
 601  */
 602 #define rdev_for_each(rdev, mddev)                              \
 603         list_for_each_entry(rdev, &((mddev)->disks), same_set)
 604
 605 #define rdev_for_each_safe(rdev, tmp, mddev)                            \
 606         list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set)
 607
 608 #define rdev_for_each_rcu(rdev, mddev)                          \
 609         list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set)
 610
 611 struct md_thread {
 612         void                    (*run) (struct md_thread *thread);
 613         struct mddev            *mddev;
 614         wait_queue_head_t       wqueue;
 615         unsigned long           flags;
 616         struct task_struct      *tsk;
 617         unsigned long           timeout;
 618         void                    *private;
 619 };
 620
 621 #define THREAD_WAKEUP  0
 622
 623 static inline void safe_put_page(struct page *p)
 624 {
 625         if (p) put_page(p);
 626 }
 627
 628 extern int register_md_personality(struct md_personality *p);
 629 extern int unregister_md_personality(struct md_personality *p);
 630 extern int register_md_cluster_operations(struct md_cluster_operations *ops,
 631                 struct module *module);
 632 extern int unregister_md_cluster_operations(void);
 633 extern int md_setup_cluster(struct mddev *mddev, int nodes);
 634 extern void md_cluster_stop(struct mddev *mddev);
 635 extern struct md_thread *md_register_thread(
 636         void (*run)(struct md_thread *thread),
 637         struct mddev *mddev,
 638         const char *name);
 639 extern void md_unregister_thread(struct md_thread **threadp);
 640 extern void md_wakeup_thread(struct md_thread *thread);
 641 extern void md_check_recovery(struct mddev *mddev);
 642 extern void md_reap_sync_thread(struct mddev *mddev);
 643 extern void md_write_start(struct mddev *mddev, struct bio *bi);
 644 extern void md_write_end(struct mddev *mddev);
 645 extern void md_done_sync(struct mddev *mddev, int blocks, int ok);
 646 extern void md_error(struct mddev *mddev, struct md_rdev *rdev);
 647 extern void md_finish_reshape(struct mddev *mddev);
 648
 649 extern int mddev_congested(struct mddev *mddev, int bits);
 650 extern void md_flush_request(struct mddev *mddev, struct bio *bio);
 651 extern void md_super_write(struct mddev *mddev, struct md_rdev *rdev,
 652                            sector_t sector, int size, struct page *page);
 653 extern void md_super_wait(struct mddev *mddev);
 654 extern int sync_page_io(struct md_rdev *rdev, sector_t sector, int size,
 655                         struct page *page, int rw, bool metadata_op);
 656 extern void md_do_sync(struct md_thread *thread);
 657 extern void md_new_event(struct mddev *mddev);
 658 extern int md_allow_write(struct mddev *mddev);
 659 extern void md_wait_for_blocked_rdev(struct md_rdev *rdev, struct mddev *mddev);
 660 extern void md_set_array_sectors(struct mddev *mddev, sector_t array_sectors);
 661 extern int md_check_no_bitmap(struct mddev *mddev);
 662 extern int md_integrity_register(struct mddev *mddev);
 663 extern int md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev);
 664 extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale);
 665
 666 extern void mddev_init(struct mddev *mddev);
 667 extern int md_run(struct mddev *mddev);
 668 extern void md_stop(struct mddev *mddev);
 669 extern void md_stop_writes(struct mddev *mddev);
 670 extern int md_rdev_init(struct md_rdev *rdev);
 671 extern void md_rdev_clear(struct md_rdev *rdev);
 672
 673 extern void mddev_suspend(struct mddev *mddev);
 674 extern void mddev_resume(struct mddev *mddev);
 675 extern struct bio *bio_clone_mddev(struct bio *bio, gfp_t gfp_mask,
 676                                    struct mddev *mddev);
 677 extern struct bio *bio_alloc_mddev(gfp_t gfp_mask, int nr_iovecs,
 678                                    struct mddev *mddev);
 679
 680 extern void md_unplug(struct blk_plug_cb *cb, bool from_schedule);
 681 extern void md_reload_sb(struct mddev *mddev, int raid_disk);
 682 extern void md_update_sb(struct mddev *mddev, int force);
 683 extern void md_kick_rdev_from_array(struct md_rdev * rdev);
 684 struct md_rdev *md_find_rdev_nr_rcu(struct mddev *mddev, int nr);
 685 static inline int mddev_check_plugged(struct mddev *mddev)
 686 {
 687         return !!blk_check_plugged(md_unplug, mddev,
 688                                    sizeof(struct blk_plug_cb));
 689 }
 690
 691 static inline void rdev_dec_pending(struct md_rdev *rdev, struct mddev *mddev)
 692 {
 693         int faulty = test_bit(Faulty, &rdev->flags);
 694         if (atomic_dec_and_test(&rdev->nr_pending) && faulty) {
 695                 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
 696                 md_wakeup_thread(mddev->thread);
 697         }
 698 }
 699
 700 extern struct md_cluster_operations *md_cluster_ops;
 701 static inline int mddev_is_clustered(struct mddev *mddev)
 702 {
 703         return mddev->cluster_info && mddev->bitmap_info.nodes > 1;
 704 }
 705 #endif /* _MD_MD_H */