drivers/md/md.h

   1 /* SPDX-License-Identifier: GPL-2.0-or-later */
   2 /*
   3    md.h : kernel internal structure of the Linux MD driver
   4           Copyright (C) 1996-98 Ingo Molnar, Gadi Oxman
   5
   6 */
   7
   8 #ifndef _MD_MD_H
   9 #define _MD_MD_H
  10
  11 #include <linux/blkdev.h>
  12 #include <linux/backing-dev.h>
  13 #include <linux/badblocks.h>
  14 #include <linux/kobject.h>
  15 #include <linux/list.h>
  16 #include <linux/mm.h>
  17 #include <linux/mutex.h>
  18 #include <linux/timer.h>
  19 #include <linux/wait.h>
  20 #include <linux/workqueue.h>
  21 #include "md-cluster.h"
  22
  23 #define MaxSector (~(sector_t)0)
  24
  25 /*
  26  * These flags should really be called "NO_RETRY" rather than
  27  * "FAILFAST" because they don't make any promise about time lapse,
  28  * only about the number of retries, which will be zero.
  29  * REQ_FAILFAST_DRIVER is not included because
  30  * Commit: 4a27446f3e39 ("[SCSI] modify scsi to handle new fail fast flags.")
  31  * seems to suggest that the errors it avoids retrying should usually
  32  * be retried.
  33  */
  34 #define MD_FAILFAST     (REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT)
  35
  36 /*
  37  * The struct embedded in rdev is used to serialize IO.
  38  */
  39 struct serial_in_rdev {
  40         struct rb_root_cached serial_rb;
  41         spinlock_t serial_lock;
  42         wait_queue_head_t serial_io_wait;
  43 };
  44
  45 /*
  46  * MD's 'extended' device
  47  */
  48 struct md_rdev {
  49         struct list_head same_set;      /* RAID devices within the same set */
  50
  51         sector_t sectors;               /* Device size (in 512bytes sectors) */
  52         struct mddev *mddev;            /* RAID array if running */
  53         int last_events;                /* IO event timestamp */
  54
  55         /*
  56          * If meta_bdev is non-NULL, it means that a separate device is
  57          * being used to store the metadata (superblock/bitmap) which
  58          * would otherwise be contained on the same device as the data (bdev).
  59          */
  60         struct block_device *meta_bdev;
  61         struct block_device *bdev;      /* block device handle */
  62
  63         struct page     *sb_page, *bb_page;
  64         int             sb_loaded;
  65         __u64           sb_events;
  66         sector_t        data_offset;    /* start of data in array */
  67         sector_t        new_data_offset;/* only relevant while reshaping */
  68         sector_t        sb_start;       /* offset of the super block (in 512byte sectors) */
  69         int             sb_size;        /* bytes in the superblock */
  70         int             preferred_minor;        /* autorun support */
  71
  72         struct kobject  kobj;
  73
  74         /* A device can be in one of three states based on two flags:
  75          * Not working:   faulty==1 in_sync==0
  76          * Fully working: faulty==0 in_sync==1
  77          * Working, but not
  78          * in sync with array
  79          *                faulty==0 in_sync==0
  80          *
  81          * It can never have faulty==1, in_sync==1
  82          * This reduces the burden of testing multiple flags in many cases
  83          */
  84
  85         unsigned long   flags;  /* bit set of 'enum flag_bits' bits. */
  86         wait_queue_head_t blocked_wait;
  87
  88         int desc_nr;                    /* descriptor index in the superblock */
  89         int raid_disk;                  /* role of device in array */
  90         int new_raid_disk;              /* role that the device will have in
  91                                          * the array after a level-change completes.
  92                                          */
  93         int saved_raid_disk;            /* role that device used to have in the
  94                                          * array and could again if we did a partial
  95                                          * resync from the bitmap
  96                                          */
  97         union {
  98                 sector_t recovery_offset;/* If this device has been partially
  99                                          * recovered, this is where we were
 100                                          * up to.
 101                                          */
 102                 sector_t journal_tail;  /* If this device is a journal device,
 103                                          * this is the journal tail (journal
 104                                          * recovery start point)
 105                                          */
 106         };
 107
 108         atomic_t        nr_pending;     /* number of pending requests.
 109                                          * only maintained for arrays that
 110                                          * support hot removal
 111                                          */
 112         atomic_t        read_errors;    /* number of consecutive read errors that
 113                                          * we have tried to ignore.
 114                                          */
 115         time64_t        last_read_error;        /* monotonic time since our
 116                                                  * last read error
 117                                                  */
 118         atomic_t        corrected_errors; /* number of corrected read errors,
 119                                            * for reporting to userspace and storing
 120                                            * in superblock.
 121                                            */
 122
 123         struct serial_in_rdev *serial;  /* used for raid1 io serialization */
 124
 125         struct work_struct del_work;    /* used for delayed sysfs removal */
 126
 127         struct kernfs_node *sysfs_state; /* handle for 'state'
 128                                            * sysfs entry */
 129
 130         struct badblocks badblocks;
 131
 132         struct {
 133                 short offset;   /* Offset from superblock to start of PPL.
 134                                  * Not used by external metadata. */
 135                 unsigned int size;      /* Size in sectors of the PPL space */
 136                 sector_t sector;        /* First sector of the PPL space */
 137         } ppl;
 138 };
 139 enum flag_bits {
 140         Faulty,                 /* device is known to have a fault */
 141         In_sync,                /* device is in_sync with rest of array */
 142         Bitmap_sync,            /* ..actually, not quite In_sync.  Need a
 143                                  * bitmap-based recovery to get fully in sync.
 144                                  * The bit is only meaningful before device
 145                                  * has been passed to pers->hot_add_disk.
 146                                  */
 147         WriteMostly,            /* Avoid reading if at all possible */
 148         AutoDetected,           /* added by auto-detect */
 149         Blocked,                /* An error occurred but has not yet
 150                                  * been acknowledged by the metadata
 151                                  * handler, so don't allow writes
 152                                  * until it is cleared */
 153         WriteErrorSeen,         /* A write error has been seen on this
 154                                  * device
 155                                  */
 156         FaultRecorded,          /* Intermediate state for clearing
 157                                  * Blocked.  The Fault is/will-be
 158                                  * recorded in the metadata, but that
 159                                  * metadata hasn't been stored safely
 160                                  * on disk yet.
 161                                  */
 162         BlockedBadBlocks,       /* A writer is blocked because they
 163                                  * found an unacknowledged bad-block.
 164                                  * This can safely be cleared at any
 165                                  * time, and the writer will re-check.
 166                                  * It may be set at any time, and at
 167                                  * worst the writer will timeout and
 168                                  * re-check.  So setting it as
 169                                  * accurately as possible is good, but
 170                                  * not absolutely critical.
 171                                  */
 172         WantReplacement,        /* This device is a candidate to be
 173                                  * hot-replaced, either because it has
 174                                  * reported some faults, or because
 175                                  * of explicit request.
 176                                  */
 177         Replacement,            /* This device is a replacement for
 178                                  * a want_replacement device with same
 179                                  * raid_disk number.
 180                                  */
 181         Candidate,              /* For clustered environments only:
 182                                  * This device is seen locally but not
 183                                  * by the whole cluster
 184                                  */
 185         Journal,                /* This device is used as journal for
 186                                  * raid-5/6.
 187                                  * Usually, this device should be faster
 188                                  * than other devices in the array
 189                                  */
 190         ClusterRemove,
 191         RemoveSynchronized,     /* synchronize_rcu() was called after
 192                                  * this device was known to be faulty,
 193                                  * so it is safe to remove without
 194                                  * another synchronize_rcu() call.
 195                                  */
 196         ExternalBbl,            /* External metadata provides bad
 197                                  * block management for a disk
 198                                  */
 199         FailFast,               /* Minimal retries should be attempted on
 200                                  * this device, so use REQ_FAILFAST_DEV.
 201                                  * Also don't try to repair failed reads.
 202                                  * It is expects that no bad block log
 203                                  * is present.
 204                                  */
 205         LastDev,                /* Seems to be the last working dev as
 206                                  * it didn't fail, so don't use FailFast
 207                                  * any more for metadata
 208                                  */
 209         CollisionCheck,         /*
 210                                  * check if there is collision between raid1
 211                                  * serial bios.
 212                                  */
 213 };
 214
 215 static inline int is_badblock(struct md_rdev *rdev, sector_t s, int sectors,
 216                               sector_t *first_bad, int *bad_sectors)
 217 {
 218         if (unlikely(rdev->badblocks.count)) {
 219                 int rv = badblocks_check(&rdev->badblocks, rdev->data_offset + s,
 220                                         sectors,
 221                                         first_bad, bad_sectors);
 222                 if (rv)
 223                         *first_bad -= rdev->data_offset;
 224                 return rv;
 225         }
 226         return 0;
 227 }
 228 extern int rdev_set_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
 229                               int is_new);
 230 extern int rdev_clear_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
 231                                 int is_new);
 232 struct md_cluster_info;
 233
 234 /* change UNSUPPORTED_MDDEV_FLAGS for each array type if new flag is added */
 235 enum mddev_flags {
 236         MD_ARRAY_FIRST_USE,     /* First use of array, needs initialization */
 237         MD_CLOSING,             /* If set, we are closing the array, do not open
 238                                  * it then */
 239         MD_JOURNAL_CLEAN,       /* A raid with journal is already clean */
 240         MD_HAS_JOURNAL,         /* The raid array has journal feature set */
 241         MD_CLUSTER_RESYNC_LOCKED, /* cluster raid only, which means node
 242                                    * already took resync lock, need to
 243                                    * release the lock */
 244         MD_FAILFAST_SUPPORTED,  /* Using MD_FAILFAST on metadata writes is
 245                                  * supported as calls to md_error() will
 246                                  * never cause the array to become failed.
 247                                  */
 248         MD_HAS_PPL,             /* The raid array has PPL feature set */
 249         MD_HAS_MULTIPLE_PPLS,   /* The raid array has multiple PPLs feature set */
 250         MD_ALLOW_SB_UPDATE,     /* md_check_recovery is allowed to update
 251                                  * the metadata without taking reconfig_mutex.
 252                                  */
 253         MD_UPDATING_SB,         /* md_check_recovery is updating the metadata
 254                                  * without explicitly holding reconfig_mutex.
 255                                  */
 256         MD_NOT_READY,           /* do_md_run() is active, so 'array_state'
 257                                  * must not report that array is ready yet
 258                                  */
 259         MD_BROKEN,              /* This is used in RAID-0/LINEAR only, to stop
 260                                  * I/O in case an array member is gone/failed.
 261                                  */
 262 };
 263
 264 enum mddev_sb_flags {
 265         MD_SB_CHANGE_DEVS,              /* Some device status has changed */
 266         MD_SB_CHANGE_CLEAN,     /* transition to or from 'clean' */
 267         MD_SB_CHANGE_PENDING,   /* switch from 'clean' to 'active' in progress */
 268         MD_SB_NEED_REWRITE,     /* metadata write needs to be repeated */
 269 };
 270
 271 #define NR_SERIAL_INFOS         8
 272 /* record current range of serialize IOs */
 273 struct serial_info {
 274         struct rb_node node;
 275         sector_t start;         /* start sector of rb node */
 276         sector_t last;          /* end sector of rb node */
 277         sector_t _subtree_last; /* highest sector in subtree of rb node */
 278 };
 279
 280 struct mddev {
 281         void                            *private;
 282         struct md_personality           *pers;
 283         dev_t                           unit;
 284         int                             md_minor;
 285         struct list_head                disks;
 286         unsigned long                   flags;
 287         unsigned long                   sb_flags;
 288
 289         int                             suspended;
 290         atomic_t                        active_io;
 291         int                             ro;
 292         int                             sysfs_active; /* set when sysfs deletes
 293                                                        * are happening, so run/
 294                                                        * takeover/stop are not safe
 295                                                        */
 296         struct gendisk                  *gendisk;
 297
 298         struct kobject                  kobj;
 299         int                             hold_active;
 300 #define UNTIL_IOCTL     1
 301 #define UNTIL_STOP      2
 302
 303         /* Superblock information */
 304         int                             major_version,
 305                                         minor_version,
 306                                         patch_version;
 307         int                             persistent;
 308         int                             external;       /* metadata is
 309                                                          * managed externally */
 310         char                            metadata_type[17]; /* externally set*/
 311         int                             chunk_sectors;
 312         time64_t                        ctime, utime;
 313         int                             level, layout;
 314         char                            clevel[16];
 315         int                             raid_disks;
 316         int                             max_disks;
 317         sector_t                        dev_sectors;    /* used size of
 318                                                          * component devices */
 319         sector_t                        array_sectors; /* exported array size */
 320         int                             external_size; /* size managed
 321                                                         * externally */
 322         __u64                           events;
 323         /* If the last 'event' was simply a clean->dirty transition, and
 324          * we didn't write it to the spares, then it is safe and simple
 325          * to just decrement the event count on a dirty->clean transition.
 326          * So we record that possibility here.
 327          */
 328         int                             can_decrease_events;
 329
 330         char                            uuid[16];
 331
 332         /* If the array is being reshaped, we need to record the
 333          * new shape and an indication of where we are up to.
 334          * This is written to the superblock.
 335          * If reshape_position is MaxSector, then no reshape is happening (yet).
 336          */
 337         sector_t                        reshape_position;
 338         int                             delta_disks, new_level, new_layout;
 339         int                             new_chunk_sectors;
 340         int                             reshape_backwards;
 341
 342         struct md_thread                *thread;        /* management thread */
 343         struct md_thread                *sync_thread;   /* doing resync or reconstruct */
 344
 345         /* 'last_sync_action' is initialized to "none".  It is set when a
 346          * sync operation (i.e "data-check", "requested-resync", "resync",
 347          * "recovery", or "reshape") is started.  It holds this value even
 348          * when the sync thread is "frozen" (interrupted) or "idle" (stopped
 349          * or finished).  It is overwritten when a new sync operation is begun.
 350          */
 351         char                            *last_sync_action;
 352         sector_t                        curr_resync;    /* last block scheduled */
 353         /* As resync requests can complete out of order, we cannot easily track
 354          * how much resync has been completed.  So we occasionally pause until
 355          * everything completes, then set curr_resync_completed to curr_resync.
 356          * As such it may be well behind the real resync mark, but it is a value
 357          * we are certain of.
 358          */
 359         sector_t                        curr_resync_completed;
 360         unsigned long                   resync_mark;    /* a recent timestamp */
 361         sector_t                        resync_mark_cnt;/* blocks written at resync_mark */
 362         sector_t                        curr_mark_cnt; /* blocks scheduled now */
 363
 364         sector_t                        resync_max_sectors; /* may be set by personality */
 365
 366         atomic64_t                      resync_mismatches; /* count of sectors where
 367                                                             * parity/replica mismatch found
 368                                                             */
 369
 370         /* allow user-space to request suspension of IO to regions of the array */
 371         sector_t                        suspend_lo;
 372         sector_t                        suspend_hi;
 373         /* if zero, use the system-wide default */
 374         int                             sync_speed_min;
 375         int                             sync_speed_max;
 376
 377         /* resync even though the same disks are shared among md-devices */
 378         int                             parallel_resync;
 379
 380         int                             ok_start_degraded;
 381
 382         unsigned long                   recovery;
 383         /* If a RAID personality determines that recovery (of a particular
 384          * device) will fail due to a read error on the source device, it
 385          * takes a copy of this number and does not attempt recovery again
 386          * until this number changes.
 387          */
 388         int                             recovery_disabled;
 389
 390         int                             in_sync;        /* know to not need resync */
 391         /* 'open_mutex' avoids races between 'md_open' and 'do_md_stop', so
 392          * that we are never stopping an array while it is open.
 393          * 'reconfig_mutex' protects all other reconfiguration.
 394          * These locks are separate due to conflicting interactions
 395          * with bdev->bd_mutex.
 396          * Lock ordering is:
 397          *  reconfig_mutex -> bd_mutex : e.g. do_md_run -> revalidate_disk
 398          *  bd_mutex -> open_mutex:  e.g. __blkdev_get -> md_open
 399          */
 400         struct mutex                    open_mutex;
 401         struct mutex                    reconfig_mutex;
 402         atomic_t                        active;         /* general refcount */
 403         atomic_t                        openers;        /* number of active opens */
 404
 405         int                             changed;        /* True if we might need to
 406                                                          * reread partition info */
 407         int                             degraded;       /* whether md should consider
 408                                                          * adding a spare
 409                                                          */
 410
 411         atomic_t                        recovery_active; /* blocks scheduled, but not written */
 412         wait_queue_head_t               recovery_wait;
 413         sector_t                        recovery_cp;
 414         sector_t                        resync_min;     /* user requested sync
 415                                                          * starts here */
 416         sector_t                        resync_max;     /* resync should pause
 417                                                          * when it gets here */
 418
 419         struct kernfs_node              *sysfs_state;   /* handle for 'array_state'
 420                                                          * file in sysfs.
 421                                                          */
 422         struct kernfs_node              *sysfs_action;  /* handle for 'sync_action' */
 423
 424         struct work_struct del_work;    /* used for delayed sysfs removal */
 425
 426         /* "lock" protects:
 427          *   flush_bio transition from NULL to !NULL
 428          *   rdev superblocks, events
 429          *   clearing MD_CHANGE_*
 430          *   in_sync - and related safemode and MD_CHANGE changes
 431          *   pers (also protected by reconfig_mutex and pending IO).
 432          *   clearing ->bitmap
 433          *   clearing ->bitmap_info.file
 434          *   changing ->resync_{min,max}
 435          *   setting MD_RECOVERY_RUNNING (which interacts with resync_{min,max})
 436          */
 437         spinlock_t                      lock;
 438         wait_queue_head_t               sb_wait;        /* for waiting on superblock updates */
 439         atomic_t                        pending_writes; /* number of active superblock writes */
 440
 441         unsigned int                    safemode;       /* if set, update "clean" superblock
 442                                                          * when no writes pending.
 443                                                          */
 444         unsigned int                    safemode_delay;
 445         struct timer_list               safemode_timer;
 446         struct percpu_ref               writes_pending;
 447         int                             sync_checkers;  /* # of threads checking writes_pending */
 448         struct request_queue            *queue; /* for plugging ... */
 449
 450         struct bitmap                   *bitmap; /* the bitmap for the device */
 451         struct {
 452                 struct file             *file; /* the bitmap file */
 453                 loff_t                  offset; /* offset from superblock of
 454                                                  * start of bitmap. May be
 455                                                  * negative, but not '0'
 456                                                  * For external metadata, offset
 457                                                  * from start of device.
 458                                                  */
 459                 unsigned long           space; /* space available at this offset */
 460                 loff_t                  default_offset; /* this is the offset to use when
 461                                                          * hot-adding a bitmap.  It should
 462                                                          * eventually be settable by sysfs.
 463                                                          */
 464                 unsigned long           default_space; /* space available at
 465                                                         * default offset */
 466                 struct mutex            mutex;
 467                 unsigned long           chunksize;
 468                 unsigned long           daemon_sleep; /* how many jiffies between updates? */
 469                 unsigned long           max_write_behind; /* write-behind mode */
 470                 int                     external;
 471                 int                     nodes; /* Maximum number of nodes in the cluster */
 472                 char                    cluster_name[64]; /* Name of the cluster */
 473         } bitmap_info;
 474
 475         atomic_t                        max_corr_read_errors; /* max read retries */
 476         struct list_head                all_mddevs;
 477
 478         struct attribute_group          *to_remove;
 479
 480         struct bio_set                  bio_set;
 481         struct bio_set                  sync_set; /* for sync operations like
 482                                                    * metadata and bitmap writes
 483                                                    */
 484
 485         /* Generic flush handling.
 486          * The last to finish preflush schedules a worker to submit
 487          * the rest of the request (without the REQ_PREFLUSH flag).
 488          */
 489         struct bio *flush_bio;
 490         atomic_t flush_pending;
 491         ktime_t start_flush, last_flush; /* last_flush is when the last completed
 492                                           * flush was started.
 493                                           */
 494         struct work_struct flush_work;
 495         struct work_struct event_work;  /* used by dm to report failure event */
 496         mempool_t *serial_info_pool;
 497         void (*sync_super)(struct mddev *mddev, struct md_rdev *rdev);
 498         struct md_cluster_info          *cluster_info;
 499         unsigned int                    good_device_nr; /* good device num within cluster raid */
 500
 501         bool    has_superblocks:1;
 502         bool    fail_last_dev:1;
 503         bool    serialize_policy:1;
 504 };
 505
 506 enum recovery_flags {
 507         /*
 508          * If neither SYNC or RESHAPE are set, then it is a recovery.
 509          */
 510         MD_RECOVERY_RUNNING,    /* a thread is running, or about to be started */
 511         MD_RECOVERY_SYNC,       /* actually doing a resync, not a recovery */
 512         MD_RECOVERY_RECOVER,    /* doing recovery, or need to try it. */
 513         MD_RECOVERY_INTR,       /* resync needs to be aborted for some reason */
 514         MD_RECOVERY_DONE,       /* thread is done and is waiting to be reaped */
 515         MD_RECOVERY_NEEDED,     /* we might need to start a resync/recover */
 516         MD_RECOVERY_REQUESTED,  /* user-space has requested a sync (used with SYNC) */
 517         MD_RECOVERY_CHECK,      /* user-space request for check-only, no repair */
 518         MD_RECOVERY_RESHAPE,    /* A reshape is happening */
 519         MD_RECOVERY_FROZEN,     /* User request to abort, and not restart, any action */
 520         MD_RECOVERY_ERROR,      /* sync-action interrupted because io-error */
 521         MD_RECOVERY_WAIT,       /* waiting for pers->start() to finish */
 522         MD_RESYNCING_REMOTE,    /* remote node is running resync thread */
 523 };
 524
 525 static inline int __must_check mddev_lock(struct mddev *mddev)
 526 {
 527         return mutex_lock_interruptible(&mddev->reconfig_mutex);
 528 }
 529
 530 /* Sometimes we need to take the lock in a situation where
 531  * failure due to interrupts is not acceptable.
 532  */
 533 static inline void mddev_lock_nointr(struct mddev *mddev)
 534 {
 535         mutex_lock(&mddev->reconfig_mutex);
 536 }
 537
 538 static inline int mddev_trylock(struct mddev *mddev)
 539 {
 540         return mutex_trylock(&mddev->reconfig_mutex);
 541 }
 542 extern void mddev_unlock(struct mddev *mddev);
 543
 544 static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors)
 545 {
 546         atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io);
 547 }
 548
 549 static inline void md_sync_acct_bio(struct bio *bio, unsigned long nr_sectors)
 550 {
 551         atomic_add(nr_sectors, &bio->bi_disk->sync_io);
 552 }
 553
 554 struct md_personality
 555 {
 556         char *name;
 557         int level;
 558         struct list_head list;
 559         struct module *owner;
 560         bool __must_check (*make_request)(struct mddev *mddev, struct bio *bio);
 561         /*
 562          * start up works that do NOT require md_thread. tasks that
 563          * requires md_thread should go into start()
 564          */
 565         int (*run)(struct mddev *mddev);
 566         /* start up works that require md threads */
 567         int (*start)(struct mddev *mddev);
 568         void (*free)(struct mddev *mddev, void *priv);
 569         void (*status)(struct seq_file *seq, struct mddev *mddev);
 570         /* error_handler must set ->faulty and clear ->in_sync
 571          * if appropriate, and should abort recovery if needed
 572          */
 573         void (*error_handler)(struct mddev *mddev, struct md_rdev *rdev);
 574         int (*hot_add_disk) (struct mddev *mddev, struct md_rdev *rdev);
 575         int (*hot_remove_disk) (struct mddev *mddev, struct md_rdev *rdev);
 576         int (*spare_active) (struct mddev *mddev);
 577         sector_t (*sync_request)(struct mddev *mddev, sector_t sector_nr, int *skipped);
 578         int (*resize) (struct mddev *mddev, sector_t sectors);
 579         sector_t (*size) (struct mddev *mddev, sector_t sectors, int raid_disks);
 580         int (*check_reshape) (struct mddev *mddev);
 581         int (*start_reshape) (struct mddev *mddev);
 582         void (*finish_reshape) (struct mddev *mddev);
 583         void (*update_reshape_pos) (struct mddev *mddev);
 584         /* quiesce suspends or resumes internal processing.
 585          * 1 - stop new actions and wait for action io to complete
 586          * 0 - return to normal behaviour
 587          */
 588         void (*quiesce) (struct mddev *mddev, int quiesce);
 589         /* takeover is used to transition an array from one
 590          * personality to another.  The new personality must be able
 591          * to handle the data in the current layout.
 592          * e.g. 2drive raid1 -> 2drive raid5
 593          *      ndrive raid5 -> degraded n+1drive raid6 with special layout
 594          * If the takeover succeeds, a new 'private' structure is returned.
 595          * This needs to be installed and then ->run used to activate the
 596          * array.
 597          */
 598         void *(*takeover) (struct mddev *mddev);
 599         /* congested implements bdi.congested_fn().
 600          * Will not be called while array is 'suspended' */
 601         int (*congested)(struct mddev *mddev, int bits);
 602         /* Changes the consistency policy of an active array. */
 603         int (*change_consistency_policy)(struct mddev *mddev, const char *buf);
 604 };
 605
 606 struct md_sysfs_entry {
 607         struct attribute attr;
 608         ssize_t (*show)(struct mddev *, char *);
 609         ssize_t (*store)(struct mddev *, const char *, size_t);
 610 };
 611 extern struct attribute_group md_bitmap_group;
 612
 613 static inline struct kernfs_node *sysfs_get_dirent_safe(struct kernfs_node *sd, char *name)
 614 {
 615         if (sd)
 616                 return sysfs_get_dirent(sd, name);
 617         return sd;
 618 }
 619 static inline void sysfs_notify_dirent_safe(struct kernfs_node *sd)
 620 {
 621         if (sd)
 622                 sysfs_notify_dirent(sd);
 623 }
 624
 625 static inline char * mdname (struct mddev * mddev)
 626 {
 627         return mddev->gendisk ? mddev->gendisk->disk_name : "mdX";
 628 }
 629
 630 static inline int sysfs_link_rdev(struct mddev *mddev, struct md_rdev *rdev)
 631 {
 632         char nm[20];
 633         if (!test_bit(Replacement, &rdev->flags) &&
 634             !test_bit(Journal, &rdev->flags) &&
 635             mddev->kobj.sd) {
 636                 sprintf(nm, "rd%d", rdev->raid_disk);
 637                 return sysfs_create_link(&mddev->kobj, &rdev->kobj, nm);
 638         } else
 639                 return 0;
 640 }
 641
 642 static inline void sysfs_unlink_rdev(struct mddev *mddev, struct md_rdev *rdev)
 643 {
 644         char nm[20];
 645         if (!test_bit(Replacement, &rdev->flags) &&
 646             !test_bit(Journal, &rdev->flags) &&
 647             mddev->kobj.sd) {
 648                 sprintf(nm, "rd%d", rdev->raid_disk);
 649                 sysfs_remove_link(&mddev->kobj, nm);
 650         }
 651 }
 652
 653 /*
 654  * iterates through some rdev ringlist. It's safe to remove the
 655  * current 'rdev'. Dont touch 'tmp' though.
 656  */
 657 #define rdev_for_each_list(rdev, tmp, head)                             \
 658         list_for_each_entry_safe(rdev, tmp, head, same_set)
 659
 660 /*
 661  * iterates through the 'same array disks' ringlist
 662  */
 663 #define rdev_for_each(rdev, mddev)                              \
 664         list_for_each_entry(rdev, &((mddev)->disks), same_set)
 665
 666 #define rdev_for_each_safe(rdev, tmp, mddev)                            \
 667         list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set)
 668
 669 #define rdev_for_each_rcu(rdev, mddev)                          \
 670         list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set)
 671
 672 struct md_thread {
 673         void                    (*run) (struct md_thread *thread);
 674         struct mddev            *mddev;
 675         wait_queue_head_t       wqueue;
 676         unsigned long           flags;
 677         struct task_struct      *tsk;
 678         unsigned long           timeout;
 679         void                    *private;
 680 };
 681
 682 #define THREAD_WAKEUP  0
 683
 684 static inline void safe_put_page(struct page *p)
 685 {
 686         if (p) put_page(p);
 687 }
 688
 689 extern int register_md_personality(struct md_personality *p);
 690 extern int unregister_md_personality(struct md_personality *p);
 691 extern int register_md_cluster_operations(struct md_cluster_operations *ops,
 692                 struct module *module);
 693 extern int unregister_md_cluster_operations(void);
 694 extern int md_setup_cluster(struct mddev *mddev, int nodes);
 695 extern void md_cluster_stop(struct mddev *mddev);
 696 extern struct md_thread *md_register_thread(
 697         void (*run)(struct md_thread *thread),
 698         struct mddev *mddev,
 699         const char *name);
 700 extern void md_unregister_thread(struct md_thread **threadp);
 701 extern void md_wakeup_thread(struct md_thread *thread);
 702 extern void md_check_recovery(struct mddev *mddev);
 703 extern void md_reap_sync_thread(struct mddev *mddev);
 704 extern int mddev_init_writes_pending(struct mddev *mddev);
 705 extern bool md_write_start(struct mddev *mddev, struct bio *bi);
 706 extern void md_write_inc(struct mddev *mddev, struct bio *bi);
 707 extern void md_write_end(struct mddev *mddev);
 708 extern void md_done_sync(struct mddev *mddev, int blocks, int ok);
 709 extern void md_error(struct mddev *mddev, struct md_rdev *rdev);
 710 extern void md_finish_reshape(struct mddev *mddev);
 711
 712 extern int mddev_congested(struct mddev *mddev, int bits);
 713 extern bool __must_check md_flush_request(struct mddev *mddev, struct bio *bio);
 714 extern void md_super_write(struct mddev *mddev, struct md_rdev *rdev,
 715                            sector_t sector, int size, struct page *page);
 716 extern int md_super_wait(struct mddev *mddev);
 717 extern int sync_page_io(struct md_rdev *rdev, sector_t sector, int size,
 718                         struct page *page, int op, int op_flags,
 719                         bool metadata_op);
 720 extern void md_do_sync(struct md_thread *thread);
 721 extern void md_new_event(struct mddev *mddev);
 722 extern void md_allow_write(struct mddev *mddev);
 723 extern void md_wait_for_blocked_rdev(struct md_rdev *rdev, struct mddev *mddev);
 724 extern void md_set_array_sectors(struct mddev *mddev, sector_t array_sectors);
 725 extern int md_check_no_bitmap(struct mddev *mddev);
 726 extern int md_integrity_register(struct mddev *mddev);
 727 extern int md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev);
 728 extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale);
 729
 730 extern void mddev_init(struct mddev *mddev);
 731 extern int md_run(struct mddev *mddev);
 732 extern int md_start(struct mddev *mddev);
 733 extern void md_stop(struct mddev *mddev);
 734 extern void md_stop_writes(struct mddev *mddev);
 735 extern int md_rdev_init(struct md_rdev *rdev);
 736 extern void md_rdev_clear(struct md_rdev *rdev);
 737
 738 extern void md_handle_request(struct mddev *mddev, struct bio *bio);
 739 extern void mddev_suspend(struct mddev *mddev);
 740 extern void mddev_resume(struct mddev *mddev);
 741 extern struct bio *bio_alloc_mddev(gfp_t gfp_mask, int nr_iovecs,
 742                                    struct mddev *mddev);
 743
 744 extern void md_reload_sb(struct mddev *mddev, int raid_disk);
 745 extern void md_update_sb(struct mddev *mddev, int force);
 746 extern void md_kick_rdev_from_array(struct md_rdev * rdev);
 747 extern void mddev_create_serial_pool(struct mddev *mddev, struct md_rdev *rdev,
 748                                      bool is_suspend);
 749 extern void mddev_destroy_serial_pool(struct mddev *mddev, struct md_rdev *rdev,
 750                                       bool is_suspend);
 751 struct md_rdev *md_find_rdev_nr_rcu(struct mddev *mddev, int nr);
 752 struct md_rdev *md_find_rdev_rcu(struct mddev *mddev, dev_t dev);
 753
 754 static inline bool is_mddev_broken(struct md_rdev *rdev, const char *md_type)
 755 {
 756         int flags = rdev->bdev->bd_disk->flags;
 757
 758         if (!(flags & GENHD_FL_UP)) {
 759                 if (!test_and_set_bit(MD_BROKEN, &rdev->mddev->flags))
 760                         pr_warn("md: %s: %s array has a missing/failed member\n",
 761                                 mdname(rdev->mddev), md_type);
 762                 return true;
 763         }
 764         return false;
 765 }
 766
 767 static inline void rdev_dec_pending(struct md_rdev *rdev, struct mddev *mddev)
 768 {
 769         int faulty = test_bit(Faulty, &rdev->flags);
 770         if (atomic_dec_and_test(&rdev->nr_pending) && faulty) {
 771                 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
 772                 md_wakeup_thread(mddev->thread);
 773         }
 774 }
 775
 776 extern struct md_cluster_operations *md_cluster_ops;
 777 static inline int mddev_is_clustered(struct mddev *mddev)
 778 {
 779         return mddev->cluster_info && mddev->bitmap_info.nodes > 1;
 780 }
 781
 782 /* clear unsupported mddev_flags */
 783 static inline void mddev_clear_unsupported_flags(struct mddev *mddev,
 784         unsigned long unsupported_flags)
 785 {
 786         mddev->flags &= ~unsupported_flags;
 787 }
 788
 789 static inline void mddev_check_writesame(struct mddev *mddev, struct bio *bio)
 790 {
 791         if (bio_op(bio) == REQ_OP_WRITE_SAME &&
 792             !bio->bi_disk->queue->limits.max_write_same_sectors)
 793                 mddev->queue->limits.max_write_same_sectors = 0;
 794 }
 795
 796 static inline void mddev_check_write_zeroes(struct mddev *mddev, struct bio *bio)
 797 {
 798         if (bio_op(bio) == REQ_OP_WRITE_ZEROES &&
 799             !bio->bi_disk->queue->limits.max_write_zeroes_sectors)
 800                 mddev->queue->limits.max_write_zeroes_sectors = 0;
 801 }
 802 #endif /* _MD_MD_H */