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/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  * These flags should really be called "NO_RETRY" rather than
  34  * "FAILFAST" because they don't make any promise about time lapse,
  35  * only about the number of retries, which will be zero.
  36  * REQ_FAILFAST_DRIVER is not included because
  37  * Commit: 4a27446f3e39 ("[SCSI] modify scsi to handle new fail fast flags.")
  38  * seems to suggest that the errors it avoids retrying should usually
  39  * be retried.
  40  */
  41 #define MD_FAILFAST     (REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT)
  42 /*
  43  * MD's 'extended' device
  44  */
  45 struct md_rdev {
  46         struct list_head same_set;      /* RAID devices within the same set */
  47
  48         sector_t sectors;               /* Device size (in 512bytes sectors) */
  49         struct mddev *mddev;            /* RAID array if running */
  50         int last_events;                /* IO event timestamp */
  51
  52         /*
  53          * If meta_bdev is non-NULL, it means that a separate device is
  54          * being used to store the metadata (superblock/bitmap) which
  55          * would otherwise be contained on the same device as the data (bdev).
  56          */
  57         struct block_device *meta_bdev;
  58         struct block_device *bdev;      /* block device handle */
  59
  60         struct page     *sb_page, *bb_page;
  61         int             sb_loaded;
  62         __u64           sb_events;
  63         sector_t        data_offset;    /* start of data in array */
  64         sector_t        new_data_offset;/* only relevant while reshaping */
  65         sector_t        sb_start;       /* offset of the super block (in 512byte sectors) */
  66         int             sb_size;        /* bytes in the superblock */
  67         int             preferred_minor;        /* autorun support */
  68
  69         struct kobject  kobj;
  70
  71         /* A device can be in one of three states based on two flags:
  72          * Not working:   faulty==1 in_sync==0
  73          * Fully working: faulty==0 in_sync==1
  74          * Working, but not
  75          * in sync with array
  76          *                faulty==0 in_sync==0
  77          *
  78          * It can never have faulty==1, in_sync==1
  79          * This reduces the burden of testing multiple flags in many cases
  80          */
  81
  82         unsigned long   flags;  /* bit set of 'enum flag_bits' bits. */
  83         wait_queue_head_t blocked_wait;
  84
  85         int desc_nr;                    /* descriptor index in the superblock */
  86         int raid_disk;                  /* role of device in array */
  87         int new_raid_disk;              /* role that the device will have in
  88                                          * the array after a level-change completes.
  89                                          */
  90         int saved_raid_disk;            /* role that device used to have in the
  91                                          * array and could again if we did a partial
  92                                          * resync from the bitmap
  93                                          */
  94         union {
  95                 sector_t recovery_offset;/* If this device has been partially
  96                                          * recovered, this is where we were
  97                                          * up to.
  98                                          */
  99                 sector_t journal_tail;  /* If this device is a journal device,
 100                                          * this is the journal tail (journal
 101                                          * recovery start point)
 102                                          */
 103         };
 104
 105         atomic_t        nr_pending;     /* number of pending requests.
 106                                          * only maintained for arrays that
 107                                          * support hot removal
 108                                          */
 109         atomic_t        read_errors;    /* number of consecutive read errors that
 110                                          * we have tried to ignore.
 111                                          */
 112         time64_t        last_read_error;        /* monotonic time since our
 113                                                  * last read error
 114                                                  */
 115         atomic_t        corrected_errors; /* number of corrected read errors,
 116                                            * for reporting to userspace and storing
 117                                            * in superblock.
 118                                            */
 119         struct work_struct del_work;    /* used for delayed sysfs removal */
 120
 121         struct kernfs_node *sysfs_state; /* handle for 'state'
 122                                            * sysfs entry */
 123
 124         struct badblocks badblocks;
 125
 126         struct {
 127                 short offset;   /* Offset from superblock to start of PPL.
 128                                  * Not used by external metadata. */
 129                 unsigned int size;      /* Size in sectors of the PPL space */
 130                 sector_t sector;        /* First sector of the PPL space */
 131         } ppl;
 132 };
 133 enum flag_bits {
 134         Faulty,                 /* device is known to have a fault */
 135         In_sync,                /* device is in_sync with rest of array */
 136         Bitmap_sync,            /* ..actually, not quite In_sync.  Need a
 137                                  * bitmap-based recovery to get fully in sync
 138                                  */
 139         WriteMostly,            /* Avoid reading if at all possible */
 140         AutoDetected,           /* added by auto-detect */
 141         Blocked,                /* An error occurred but has not yet
 142                                  * been acknowledged by the metadata
 143                                  * handler, so don't allow writes
 144                                  * until it is cleared */
 145         WriteErrorSeen,         /* A write error has been seen on this
 146                                  * device
 147                                  */
 148         FaultRecorded,          /* Intermediate state for clearing
 149                                  * Blocked.  The Fault is/will-be
 150                                  * recorded in the metadata, but that
 151                                  * metadata hasn't been stored safely
 152                                  * on disk yet.
 153                                  */
 154         BlockedBadBlocks,       /* A writer is blocked because they
 155                                  * found an unacknowledged bad-block.
 156                                  * This can safely be cleared at any
 157                                  * time, and the writer will re-check.
 158                                  * It may be set at any time, and at
 159                                  * worst the writer will timeout and
 160                                  * re-check.  So setting it as
 161                                  * accurately as possible is good, but
 162                                  * not absolutely critical.
 163                                  */
 164         WantReplacement,        /* This device is a candidate to be
 165                                  * hot-replaced, either because it has
 166                                  * reported some faults, or because
 167                                  * of explicit request.
 168                                  */
 169         Replacement,            /* This device is a replacement for
 170                                  * a want_replacement device with same
 171                                  * raid_disk number.
 172                                  */
 173         Candidate,              /* For clustered environments only:
 174                                  * This device is seen locally but not
 175                                  * by the whole cluster
 176                                  */
 177         Journal,                /* This device is used as journal for
 178                                  * raid-5/6.
 179                                  * Usually, this device should be faster
 180                                  * than other devices in the array
 181                                  */
 182         ClusterRemove,
 183         RemoveSynchronized,     /* synchronize_rcu() was called after
 184                                  * this device was known to be faulty,
 185                                  * so it is safe to remove without
 186                                  * another synchronize_rcu() call.
 187                                  */
 188         ExternalBbl,            /* External metadata provides bad
 189                                  * block management for a disk
 190                                  */
 191         FailFast,               /* Minimal retries should be attempted on
 192                                  * this device, so use REQ_FAILFAST_DEV.
 193                                  * Also don't try to repair failed reads.
 194                                  * It is expects that no bad block log
 195                                  * is present.
 196                                  */
 197         LastDev,                /* Seems to be the last working dev as
 198                                  * it didn't fail, so don't use FailFast
 199                                  * any more for metadata
 200                                  */
 201 };
 202
 203 static inline int is_badblock(struct md_rdev *rdev, sector_t s, int sectors,
 204                               sector_t *first_bad, int *bad_sectors)
 205 {
 206         if (unlikely(rdev->badblocks.count)) {
 207                 int rv = badblocks_check(&rdev->badblocks, rdev->data_offset + s,
 208                                         sectors,
 209                                         first_bad, bad_sectors);
 210                 if (rv)
 211                         *first_bad -= rdev->data_offset;
 212                 return rv;
 213         }
 214         return 0;
 215 }
 216 extern int rdev_set_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
 217                               int is_new);
 218 extern int rdev_clear_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
 219                                 int is_new);
 220 struct md_cluster_info;
 221
 222 /* change UNSUPPORTED_MDDEV_FLAGS for each array type if new flag is added */
 223 enum mddev_flags {
 224         MD_ARRAY_FIRST_USE,     /* First use of array, needs initialization */
 225         MD_CLOSING,             /* If set, we are closing the array, do not open
 226                                  * it then */
 227         MD_JOURNAL_CLEAN,       /* A raid with journal is already clean */
 228         MD_HAS_JOURNAL,         /* The raid array has journal feature set */
 229         MD_CLUSTER_RESYNC_LOCKED, /* cluster raid only, which means node
 230                                    * already took resync lock, need to
 231                                    * release the lock */
 232         MD_FAILFAST_SUPPORTED,  /* Using MD_FAILFAST on metadata writes is
 233                                  * supported as calls to md_error() will
 234                                  * never cause the array to become failed.
 235                                  */
 236         MD_HAS_PPL,             /* The raid array has PPL feature set */
 237 };
 238
 239 enum mddev_sb_flags {
 240         MD_SB_CHANGE_DEVS,              /* Some device status has changed */
 241         MD_SB_CHANGE_CLEAN,     /* transition to or from 'clean' */
 242         MD_SB_CHANGE_PENDING,   /* switch from 'clean' to 'active' in progress */
 243         MD_SB_NEED_REWRITE,     /* metadata write needs to be repeated */
 244 };
 245
 246 struct mddev {
 247         void                            *private;
 248         struct md_personality           *pers;
 249         dev_t                           unit;
 250         int                             md_minor;
 251         struct list_head                disks;
 252         unsigned long                   flags;
 253         unsigned long                   sb_flags;
 254
 255         int                             suspended;
 256         atomic_t                        active_io;
 257         int                             ro;
 258         int                             sysfs_active; /* set when sysfs deletes
 259                                                        * are happening, so run/
 260                                                        * takeover/stop are not safe
 261                                                        */
 262         struct gendisk                  *gendisk;
 263
 264         struct kobject                  kobj;
 265         int                             hold_active;
 266 #define UNTIL_IOCTL     1
 267 #define UNTIL_STOP      2
 268
 269         /* Superblock information */
 270         int                             major_version,
 271                                         minor_version,
 272                                         patch_version;
 273         int                             persistent;
 274         int                             external;       /* metadata is
 275                                                          * managed externally */
 276         char                            metadata_type[17]; /* externally set*/
 277         int                             chunk_sectors;
 278         time64_t                        ctime, utime;
 279         int                             level, layout;
 280         char                            clevel[16];
 281         int                             raid_disks;
 282         int                             max_disks;
 283         sector_t                        dev_sectors;    /* used size of
 284                                                          * component devices */
 285         sector_t                        array_sectors; /* exported array size */
 286         int                             external_size; /* size managed
 287                                                         * externally */
 288         __u64                           events;
 289         /* If the last 'event' was simply a clean->dirty transition, and
 290          * we didn't write it to the spares, then it is safe and simple
 291          * to just decrement the event count on a dirty->clean transition.
 292          * So we record that possibility here.
 293          */
 294         int                             can_decrease_events;
 295
 296         char                            uuid[16];
 297
 298         /* If the array is being reshaped, we need to record the
 299          * new shape and an indication of where we are up to.
 300          * This is written to the superblock.
 301          * If reshape_position is MaxSector, then no reshape is happening (yet).
 302          */
 303         sector_t                        reshape_position;
 304         int                             delta_disks, new_level, new_layout;
 305         int                             new_chunk_sectors;
 306         int                             reshape_backwards;
 307
 308         struct md_thread                *thread;        /* management thread */
 309         struct md_thread                *sync_thread;   /* doing resync or reconstruct */
 310
 311         /* 'last_sync_action' is initialized to "none".  It is set when a
 312          * sync operation (i.e "data-check", "requested-resync", "resync",
 313          * "recovery", or "reshape") is started.  It holds this value even
 314          * when the sync thread is "frozen" (interrupted) or "idle" (stopped
 315          * or finished).  It is overwritten when a new sync operation is begun.
 316          */
 317         char                            *last_sync_action;
 318         sector_t                        curr_resync;    /* last block scheduled */
 319         /* As resync requests can complete out of order, we cannot easily track
 320          * how much resync has been completed.  So we occasionally pause until
 321          * everything completes, then set curr_resync_completed to curr_resync.
 322          * As such it may be well behind the real resync mark, but it is a value
 323          * we are certain of.
 324          */
 325         sector_t                        curr_resync_completed;
 326         unsigned long                   resync_mark;    /* a recent timestamp */
 327         sector_t                        resync_mark_cnt;/* blocks written at resync_mark */
 328         sector_t                        curr_mark_cnt; /* blocks scheduled now */
 329
 330         sector_t                        resync_max_sectors; /* may be set by personality */
 331
 332         atomic64_t                      resync_mismatches; /* count of sectors where
 333                                                             * parity/replica mismatch found
 334                                                             */
 335
 336         /* allow user-space to request suspension of IO to regions of the array */
 337         sector_t                        suspend_lo;
 338         sector_t                        suspend_hi;
 339         /* if zero, use the system-wide default */
 340         int                             sync_speed_min;
 341         int                             sync_speed_max;
 342
 343         /* resync even though the same disks are shared among md-devices */
 344         int                             parallel_resync;
 345
 346         int                             ok_start_degraded;
 347
 348         unsigned long                   recovery;
 349         /* If a RAID personality determines that recovery (of a particular
 350          * device) will fail due to a read error on the source device, it
 351          * takes a copy of this number and does not attempt recovery again
 352          * until this number changes.
 353          */
 354         int                             recovery_disabled;
 355
 356         int                             in_sync;        /* know to not need resync */
 357         /* 'open_mutex' avoids races between 'md_open' and 'do_md_stop', so
 358          * that we are never stopping an array while it is open.
 359          * 'reconfig_mutex' protects all other reconfiguration.
 360          * These locks are separate due to conflicting interactions
 361          * with bdev->bd_mutex.
 362          * Lock ordering is:
 363          *  reconfig_mutex -> bd_mutex : e.g. do_md_run -> revalidate_disk
 364          *  bd_mutex -> open_mutex:  e.g. __blkdev_get -> md_open
 365          */
 366         struct mutex                    open_mutex;
 367         struct mutex                    reconfig_mutex;
 368         atomic_t                        active;         /* general refcount */
 369         atomic_t                        openers;        /* number of active opens */
 370
 371         int                             changed;        /* True if we might need to
 372                                                          * reread partition info */
 373         int                             degraded;       /* whether md should consider
 374                                                          * adding a spare
 375                                                          */
 376
 377         atomic_t                        recovery_active; /* blocks scheduled, but not written */
 378         wait_queue_head_t               recovery_wait;
 379         sector_t                        recovery_cp;
 380         sector_t                        resync_min;     /* user requested sync
 381                                                          * starts here */
 382         sector_t                        resync_max;     /* resync should pause
 383                                                          * when it gets here */
 384
 385         struct kernfs_node              *sysfs_state;   /* handle for 'array_state'
 386                                                          * file in sysfs.
 387                                                          */
 388         struct kernfs_node              *sysfs_action;  /* handle for 'sync_action' */
 389
 390         struct work_struct del_work;    /* used for delayed sysfs removal */
 391
 392         /* "lock" protects:
 393          *   flush_bio transition from NULL to !NULL
 394          *   rdev superblocks, events
 395          *   clearing MD_CHANGE_*
 396          *   in_sync - and related safemode and MD_CHANGE changes
 397          *   pers (also protected by reconfig_mutex and pending IO).
 398          *   clearing ->bitmap
 399          *   clearing ->bitmap_info.file
 400          *   changing ->resync_{min,max}
 401          *   setting MD_RECOVERY_RUNNING (which interacts with resync_{min,max})
 402          */
 403         spinlock_t                      lock;
 404         wait_queue_head_t               sb_wait;        /* for waiting on superblock updates */
 405         atomic_t                        pending_writes; /* number of active superblock writes */
 406
 407         unsigned int                    safemode;       /* if set, update "clean" superblock
 408                                                          * when no writes pending.
 409                                                          */
 410         unsigned int                    safemode_delay;
 411         struct timer_list               safemode_timer;
 412         struct percpu_ref               writes_pending;
 413         int                             sync_checkers;  /* # of threads checking writes_pending */
 414         struct request_queue            *queue; /* for plugging ... */
 415
 416         struct bitmap                   *bitmap; /* the bitmap for the device */
 417         struct {
 418                 struct file             *file; /* the bitmap file */
 419                 loff_t                  offset; /* offset from superblock of
 420                                                  * start of bitmap. May be
 421                                                  * negative, but not '0'
 422                                                  * For external metadata, offset
 423                                                  * from start of device.
 424                                                  */
 425                 unsigned long           space; /* space available at this offset */
 426                 loff_t                  default_offset; /* this is the offset to use when
 427                                                          * hot-adding a bitmap.  It should
 428                                                          * eventually be settable by sysfs.
 429                                                          */
 430                 unsigned long           default_space; /* space available at
 431                                                         * default offset */
 432                 struct mutex            mutex;
 433                 unsigned long           chunksize;
 434                 unsigned long           daemon_sleep; /* how many jiffies between updates? */
 435                 unsigned long           max_write_behind; /* write-behind mode */
 436                 int                     external;
 437                 int                     nodes; /* Maximum number of nodes in the cluster */
 438                 char                    cluster_name[64]; /* Name of the cluster */
 439         } bitmap_info;
 440
 441         atomic_t                        max_corr_read_errors; /* max read retries */
 442         struct list_head                all_mddevs;
 443
 444         struct attribute_group          *to_remove;
 445
 446         struct bio_set                  *bio_set;
 447
 448         /* Generic flush handling.
 449          * The last to finish preflush schedules a worker to submit
 450          * the rest of the request (without the REQ_PREFLUSH flag).
 451          */
 452         struct bio *flush_bio;
 453         atomic_t flush_pending;
 454         struct work_struct flush_work;
 455         struct work_struct event_work;  /* used by dm to report failure event */
 456         void (*sync_super)(struct mddev *mddev, struct md_rdev *rdev);
 457         struct md_cluster_info          *cluster_info;
 458         unsigned int                    good_device_nr; /* good device num within cluster raid */
 459 };
 460
 461 enum recovery_flags {
 462         /*
 463          * If neither SYNC or RESHAPE are set, then it is a recovery.
 464          */
 465         MD_RECOVERY_RUNNING,    /* a thread is running, or about to be started */
 466         MD_RECOVERY_SYNC,       /* actually doing a resync, not a recovery */
 467         MD_RECOVERY_RECOVER,    /* doing recovery, or need to try it. */
 468         MD_RECOVERY_INTR,       /* resync needs to be aborted for some reason */
 469         MD_RECOVERY_DONE,       /* thread is done and is waiting to be reaped */
 470         MD_RECOVERY_NEEDED,     /* we might need to start a resync/recover */
 471         MD_RECOVERY_REQUESTED,  /* user-space has requested a sync (used with SYNC) */
 472         MD_RECOVERY_CHECK,      /* user-space request for check-only, no repair */
 473         MD_RECOVERY_RESHAPE,    /* A reshape is happening */
 474         MD_RECOVERY_FROZEN,     /* User request to abort, and not restart, any action */
 475         MD_RECOVERY_ERROR,      /* sync-action interrupted because io-error */
 476 };
 477
 478 static inline int __must_check mddev_lock(struct mddev *mddev)
 479 {
 480         return mutex_lock_interruptible(&mddev->reconfig_mutex);
 481 }
 482
 483 /* Sometimes we need to take the lock in a situation where
 484  * failure due to interrupts is not acceptable.
 485  */
 486 static inline void mddev_lock_nointr(struct mddev *mddev)
 487 {
 488         mutex_lock(&mddev->reconfig_mutex);
 489 }
 490
 491 static inline int mddev_is_locked(struct mddev *mddev)
 492 {
 493         return mutex_is_locked(&mddev->reconfig_mutex);
 494 }
 495
 496 static inline int mddev_trylock(struct mddev *mddev)
 497 {
 498         return mutex_trylock(&mddev->reconfig_mutex);
 499 }
 500 extern void mddev_unlock(struct mddev *mddev);
 501
 502 static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors)
 503 {
 504         atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io);
 505 }
 506
 507 struct md_personality
 508 {
 509         char *name;
 510         int level;
 511         struct list_head list;
 512         struct module *owner;
 513         void (*make_request)(struct mddev *mddev, struct bio *bio);
 514         int (*run)(struct mddev *mddev);
 515         void (*free)(struct mddev *mddev, void *priv);
 516         void (*status)(struct seq_file *seq, struct mddev *mddev);
 517         /* error_handler must set ->faulty and clear ->in_sync
 518          * if appropriate, and should abort recovery if needed
 519          */
 520         void (*error_handler)(struct mddev *mddev, struct md_rdev *rdev);
 521         int (*hot_add_disk) (struct mddev *mddev, struct md_rdev *rdev);
 522         int (*hot_remove_disk) (struct mddev *mddev, struct md_rdev *rdev);
 523         int (*spare_active) (struct mddev *mddev);
 524         sector_t (*sync_request)(struct mddev *mddev, sector_t sector_nr, int *skipped);
 525         int (*resize) (struct mddev *mddev, sector_t sectors);
 526         sector_t (*size) (struct mddev *mddev, sector_t sectors, int raid_disks);
 527         int (*check_reshape) (struct mddev *mddev);
 528         int (*start_reshape) (struct mddev *mddev);
 529         void (*finish_reshape) (struct mddev *mddev);
 530         /* quiesce moves between quiescence states
 531          * 0 - fully active
 532          * 1 - no new requests allowed
 533          * others - reserved
 534          */
 535         void (*quiesce) (struct mddev *mddev, int state);
 536         /* takeover is used to transition an array from one
 537          * personality to another.  The new personality must be able
 538          * to handle the data in the current layout.
 539          * e.g. 2drive raid1 -> 2drive raid5
 540          *      ndrive raid5 -> degraded n+1drive raid6 with special layout
 541          * If the takeover succeeds, a new 'private' structure is returned.
 542          * This needs to be installed and then ->run used to activate the
 543          * array.
 544          */
 545         void *(*takeover) (struct mddev *mddev);
 546         /* congested implements bdi.congested_fn().
 547          * Will not be called while array is 'suspended' */
 548         int (*congested)(struct mddev *mddev, int bits);
 549         /* Changes the consistency policy of an active array. */
 550         int (*change_consistency_policy)(struct mddev *mddev, const char *buf);
 551 };
 552
 553 struct md_sysfs_entry {
 554         struct attribute attr;
 555         ssize_t (*show)(struct mddev *, char *);
 556         ssize_t (*store)(struct mddev *, const char *, size_t);
 557 };
 558 extern struct attribute_group md_bitmap_group;
 559
 560 static inline struct kernfs_node *sysfs_get_dirent_safe(struct kernfs_node *sd, char *name)
 561 {
 562         if (sd)
 563                 return sysfs_get_dirent(sd, name);
 564         return sd;
 565 }
 566 static inline void sysfs_notify_dirent_safe(struct kernfs_node *sd)
 567 {
 568         if (sd)
 569                 sysfs_notify_dirent(sd);
 570 }
 571
 572 static inline char * mdname (struct mddev * mddev)
 573 {
 574         return mddev->gendisk ? mddev->gendisk->disk_name : "mdX";
 575 }
 576
 577 static inline int sysfs_link_rdev(struct mddev *mddev, struct md_rdev *rdev)
 578 {
 579         char nm[20];
 580         if (!test_bit(Replacement, &rdev->flags) &&
 581             !test_bit(Journal, &rdev->flags) &&
 582             mddev->kobj.sd) {
 583                 sprintf(nm, "rd%d", rdev->raid_disk);
 584                 return sysfs_create_link(&mddev->kobj, &rdev->kobj, nm);
 585         } else
 586                 return 0;
 587 }
 588
 589 static inline void sysfs_unlink_rdev(struct mddev *mddev, struct md_rdev *rdev)
 590 {
 591         char nm[20];
 592         if (!test_bit(Replacement, &rdev->flags) &&
 593             !test_bit(Journal, &rdev->flags) &&
 594             mddev->kobj.sd) {
 595                 sprintf(nm, "rd%d", rdev->raid_disk);
 596                 sysfs_remove_link(&mddev->kobj, nm);
 597         }
 598 }
 599
 600 /*
 601  * iterates through some rdev ringlist. It's safe to remove the
 602  * current 'rdev'. Dont touch 'tmp' though.
 603  */
 604 #define rdev_for_each_list(rdev, tmp, head)                             \
 605         list_for_each_entry_safe(rdev, tmp, head, same_set)
 606
 607 /*
 608  * iterates through the 'same array disks' ringlist
 609  */
 610 #define rdev_for_each(rdev, mddev)                              \
 611         list_for_each_entry(rdev, &((mddev)->disks), same_set)
 612
 613 #define rdev_for_each_safe(rdev, tmp, mddev)                            \
 614         list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set)
 615
 616 #define rdev_for_each_rcu(rdev, mddev)                          \
 617         list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set)
 618
 619 struct md_thread {
 620         void                    (*run) (struct md_thread *thread);
 621         struct mddev            *mddev;
 622         wait_queue_head_t       wqueue;
 623         unsigned long           flags;
 624         struct task_struct      *tsk;
 625         unsigned long           timeout;
 626         void                    *private;
 627 };
 628
 629 #define THREAD_WAKEUP  0
 630
 631 static inline void safe_put_page(struct page *p)
 632 {
 633         if (p) put_page(p);
 634 }
 635
 636 extern int register_md_personality(struct md_personality *p);
 637 extern int unregister_md_personality(struct md_personality *p);
 638 extern int register_md_cluster_operations(struct md_cluster_operations *ops,
 639                 struct module *module);
 640 extern int unregister_md_cluster_operations(void);
 641 extern int md_setup_cluster(struct mddev *mddev, int nodes);
 642 extern void md_cluster_stop(struct mddev *mddev);
 643 extern struct md_thread *md_register_thread(
 644         void (*run)(struct md_thread *thread),
 645         struct mddev *mddev,
 646         const char *name);
 647 extern void md_unregister_thread(struct md_thread **threadp);
 648 extern void md_wakeup_thread(struct md_thread *thread);
 649 extern void md_check_recovery(struct mddev *mddev);
 650 extern void md_reap_sync_thread(struct mddev *mddev);
 651 extern int mddev_init_writes_pending(struct mddev *mddev);
 652 extern void md_write_start(struct mddev *mddev, struct bio *bi);
 653 extern void md_write_inc(struct mddev *mddev, struct bio *bi);
 654 extern void md_write_end(struct mddev *mddev);
 655 extern void md_done_sync(struct mddev *mddev, int blocks, int ok);
 656 extern void md_error(struct mddev *mddev, struct md_rdev *rdev);
 657 extern void md_finish_reshape(struct mddev *mddev);
 658
 659 extern int mddev_congested(struct mddev *mddev, int bits);
 660 extern void md_flush_request(struct mddev *mddev, struct bio *bio);
 661 extern void md_super_write(struct mddev *mddev, struct md_rdev *rdev,
 662                            sector_t sector, int size, struct page *page);
 663 extern int md_super_wait(struct mddev *mddev);
 664 extern int sync_page_io(struct md_rdev *rdev, sector_t sector, int size,
 665                         struct page *page, int op, int op_flags,
 666                         bool metadata_op);
 667 extern void md_do_sync(struct md_thread *thread);
 668 extern void md_new_event(struct mddev *mddev);
 669 extern void md_allow_write(struct mddev *mddev);
 670 extern void md_wait_for_blocked_rdev(struct md_rdev *rdev, struct mddev *mddev);
 671 extern void md_set_array_sectors(struct mddev *mddev, sector_t array_sectors);
 672 extern int md_check_no_bitmap(struct mddev *mddev);
 673 extern int md_integrity_register(struct mddev *mddev);
 674 extern int md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev);
 675 extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale);
 676
 677 extern void mddev_init(struct mddev *mddev);
 678 extern int md_run(struct mddev *mddev);
 679 extern void md_stop(struct mddev *mddev);
 680 extern void md_stop_writes(struct mddev *mddev);
 681 extern int md_rdev_init(struct md_rdev *rdev);
 682 extern void md_rdev_clear(struct md_rdev *rdev);
 683
 684 extern void mddev_suspend(struct mddev *mddev);
 685 extern void mddev_resume(struct mddev *mddev);
 686 extern struct bio *bio_alloc_mddev(gfp_t gfp_mask, int nr_iovecs,
 687                                    struct mddev *mddev);
 688
 689 extern void md_reload_sb(struct mddev *mddev, int raid_disk);
 690 extern void md_update_sb(struct mddev *mddev, int force);
 691 extern void md_kick_rdev_from_array(struct md_rdev * rdev);
 692 struct md_rdev *md_find_rdev_nr_rcu(struct mddev *mddev, int nr);
 693
 694 static inline void rdev_dec_pending(struct md_rdev *rdev, struct mddev *mddev)
 695 {
 696         int faulty = test_bit(Faulty, &rdev->flags);
 697         if (atomic_dec_and_test(&rdev->nr_pending) && faulty) {
 698                 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
 699                 md_wakeup_thread(mddev->thread);
 700         }
 701 }
 702
 703 extern struct md_cluster_operations *md_cluster_ops;
 704 static inline int mddev_is_clustered(struct mddev *mddev)
 705 {
 706         return mddev->cluster_info && mddev->bitmap_info.nodes > 1;
 707 }
 708
 709 /* clear unsupported mddev_flags */
 710 static inline void mddev_clear_unsupported_flags(struct mddev *mddev,
 711         unsigned long unsupported_flags)
 712 {
 713         mddev->flags &= ~unsupported_flags;
 714 }
 715
 716 static inline void mddev_check_writesame(struct mddev *mddev, struct bio *bio)
 717 {
 718         if (bio_op(bio) == REQ_OP_WRITE_SAME &&
 719             !bdev_get_queue(bio->bi_bdev)->limits.max_write_same_sectors)
 720                 mddev->queue->limits.max_write_same_sectors = 0;
 721 }
 722
 723 static inline void mddev_check_write_zeroes(struct mddev *mddev, struct bio *bio)
 724 {
 725         if (bio_op(bio) == REQ_OP_WRITE_ZEROES &&
 726             !bdev_get_queue(bio->bi_bdev)->limits.max_write_zeroes_sectors)
 727                 mddev->queue->limits.max_write_zeroes_sectors = 0;
 728 }
 729
 730 /* Maximum size of each resync request */
 731 #define RESYNC_BLOCK_SIZE (64*1024)
 732 #define RESYNC_PAGES ((RESYNC_BLOCK_SIZE + PAGE_SIZE-1) / PAGE_SIZE)
 733
 734 /* for managing resync I/O pages */
 735 struct resync_pages {
 736         unsigned        idx;    /* for get/put page from the pool */
 737         void            *raid_bio;
 738         struct page     *pages[RESYNC_PAGES];
 739 };
 740
 741 static inline int resync_alloc_pages(struct resync_pages *rp,
 742                                      gfp_t gfp_flags)
 743 {
 744         int i;
 745
 746         for (i = 0; i < RESYNC_PAGES; i++) {
 747                 rp->pages[i] = alloc_page(gfp_flags);
 748                 if (!rp->pages[i])
 749                         goto out_free;
 750         }
 751
 752         return 0;
 753
 754 out_free:
 755         while (--i >= 0)
 756                 put_page(rp->pages[i]);
 757         return -ENOMEM;
 758 }
 759
 760 static inline void resync_free_pages(struct resync_pages *rp)
 761 {
 762         int i;
 763
 764         for (i = 0; i < RESYNC_PAGES; i++)
 765                 put_page(rp->pages[i]);
 766 }
 767
 768 static inline void resync_get_all_pages(struct resync_pages *rp)
 769 {
 770         int i;
 771
 772         for (i = 0; i < RESYNC_PAGES; i++)
 773                 get_page(rp->pages[i]);
 774 }
 775
 776 static inline struct page *resync_fetch_page(struct resync_pages *rp,
 777                                              unsigned idx)
 778 {
 779         if (WARN_ON_ONCE(idx >= RESYNC_PAGES))
 780                 return NULL;
 781         return rp->pages[idx];
 782 }
 783 #endif /* _MD_MD_H */