drivers/md/md.h

   1 /*
   2    md_k.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/kobject.h>
  20 #include <linux/list.h>
  21 #include <linux/mm.h>
  22 #include <linux/mutex.h>
  23 #include <linux/timer.h>
  24 #include <linux/wait.h>
  25 #include <linux/workqueue.h>
  26
  27 #define MaxSector (~(sector_t)0)
  28
  29 typedef struct mddev_s mddev_t;
  30 typedef struct mdk_rdev_s mdk_rdev_t;
  31
  32 /*
  33  * MD's 'extended' device
  34  */
  35 struct mdk_rdev_s
  36 {
  37         struct list_head same_set;      /* RAID devices within the same set */
  38
  39         sector_t sectors;               /* Device size (in 512bytes sectors) */
  40         mddev_t *mddev;                 /* RAID array if running */
  41         int last_events;                /* IO event timestamp */
  42
  43         /*
  44          * If meta_bdev is non-NULL, it means that a separate device is
  45          * being used to store the metadata (superblock/bitmap) which
  46          * would otherwise be contained on the same device as the data (bdev).
  47          */
  48         struct block_device *meta_bdev;
  49         struct block_device *bdev;      /* block device handle */
  50
  51         struct page     *sb_page;
  52         int             sb_loaded;
  53         __u64           sb_events;
  54         sector_t        data_offset;    /* start of data in array */
  55         sector_t        sb_start;       /* offset of the super block (in 512byte sectors) */
  56         int             sb_size;        /* bytes in the superblock */
  57         int             preferred_minor;        /* autorun support */
  58
  59         struct kobject  kobj;
  60
  61         /* A device can be in one of three states based on two flags:
  62          * Not working:   faulty==1 in_sync==0
  63          * Fully working: faulty==0 in_sync==1
  64          * Working, but not
  65          * in sync with array
  66          *                faulty==0 in_sync==0
  67          *
  68          * It can never have faulty==1, in_sync==1
  69          * This reduces the burden of testing multiple flags in many cases
  70          */
  71
  72         unsigned long   flags;
  73 #define Faulty          1               /* device is known to have a fault */
  74 #define In_sync         2               /* device is in_sync with rest of array */
  75 #define WriteMostly     4               /* Avoid reading if at all possible */
  76 #define AutoDetected    7               /* added by auto-detect */
  77 #define Blocked         8               /* An error occurred on an externally
  78                                          * managed array, don't allow writes
  79                                          * until it is cleared */
  80         wait_queue_head_t blocked_wait;
  81
  82         int desc_nr;                    /* descriptor index in the superblock */
  83         int raid_disk;                  /* role of device in array */
  84         int new_raid_disk;              /* role that the device will have in
  85                                          * the array after a level-change completes.
  86                                          */
  87         int saved_raid_disk;            /* role that device used to have in the
  88                                          * array and could again if we did a partial
  89                                          * resync from the bitmap
  90                                          */
  91         sector_t        recovery_offset;/* If this device has been partially
  92                                          * recovered, this is where we were
  93                                          * up to.
  94                                          */
  95
  96         atomic_t        nr_pending;     /* number of pending requests.
  97                                          * only maintained for arrays that
  98                                          * support hot removal
  99                                          */
 100         atomic_t        read_errors;    /* number of consecutive read errors that
 101                                          * we have tried to ignore.
 102                                          */
 103         struct timespec last_read_error;        /* monotonic time since our
 104                                                  * last read error
 105                                                  */
 106         atomic_t        corrected_errors; /* number of corrected read errors,
 107                                            * for reporting to userspace and storing
 108                                            * in superblock.
 109                                            */
 110         struct work_struct del_work;    /* used for delayed sysfs removal */
 111
 112         struct sysfs_dirent *sysfs_state; /* handle for 'state'
 113                                            * sysfs entry */
 114 };
 115
 116 struct mddev_s
 117 {
 118         void                            *private;
 119         struct mdk_personality          *pers;
 120         dev_t                           unit;
 121         int                             md_minor;
 122         struct list_head                disks;
 123         unsigned long                   flags;
 124 #define MD_CHANGE_DEVS  0       /* Some device status has changed */
 125 #define MD_CHANGE_CLEAN 1       /* transition to or from 'clean' */
 126 #define MD_CHANGE_PENDING 2     /* switch from 'clean' to 'active' in progress */
 127 #define MD_ARRAY_FIRST_USE 3    /* First use of array, needs initialization */
 128
 129         int                             suspended;
 130         atomic_t                        active_io;
 131         int                             ro;
 132         int                             sysfs_active; /* set when sysfs deletes
 133                                                        * are happening, so run/
 134                                                        * takeover/stop are not safe
 135                                                        */
 136         int                             ready; /* See when safe to pass
 137                                                 * IO requests down */
 138         struct gendisk                  *gendisk;
 139
 140         struct kobject                  kobj;
 141         int                             hold_active;
 142 #define UNTIL_IOCTL     1
 143 #define UNTIL_STOP      2
 144
 145         /* Superblock information */
 146         int                             major_version,
 147                                         minor_version,
 148                                         patch_version;
 149         int                             persistent;
 150         int                             external;       /* metadata is
 151                                                          * managed externally */
 152         char                            metadata_type[17]; /* externally set*/
 153         int                             chunk_sectors;
 154         time_t                          ctime, utime;
 155         int                             level, layout;
 156         char                            clevel[16];
 157         int                             raid_disks;
 158         int                             max_disks;
 159         sector_t                        dev_sectors;    /* used size of
 160                                                          * component devices */
 161         sector_t                        array_sectors; /* exported array size */
 162         int                             external_size; /* size managed
 163                                                         * externally */
 164         __u64                           events;
 165         /* If the last 'event' was simply a clean->dirty transition, and
 166          * we didn't write it to the spares, then it is safe and simple
 167          * to just decrement the event count on a dirty->clean transition.
 168          * So we record that possibility here.
 169          */
 170         int                             can_decrease_events;
 171
 172         char                            uuid[16];
 173
 174         /* If the array is being reshaped, we need to record the
 175          * new shape and an indication of where we are up to.
 176          * This is written to the superblock.
 177          * If reshape_position is MaxSector, then no reshape is happening (yet).
 178          */
 179         sector_t                        reshape_position;
 180         int                             delta_disks, new_level, new_layout;
 181         int                             new_chunk_sectors;
 182
 183         atomic_t                        plug_cnt;       /* If device is expecting
 184                                                          * more bios soon.
 185                                                          */
 186         struct mdk_thread_s             *thread;        /* management thread */
 187         struct mdk_thread_s             *sync_thread;   /* doing resync or reconstruct */
 188         sector_t                        curr_resync;    /* last block scheduled */
 189         /* As resync requests can complete out of order, we cannot easily track
 190          * how much resync has been completed.  So we occasionally pause until
 191          * everything completes, then set curr_resync_completed to curr_resync.
 192          * As such it may be well behind the real resync mark, but it is a value
 193          * we are certain of.
 194          */
 195         sector_t                        curr_resync_completed;
 196         unsigned long                   resync_mark;    /* a recent timestamp */
 197         sector_t                        resync_mark_cnt;/* blocks written at resync_mark */
 198         sector_t                        curr_mark_cnt; /* blocks scheduled now */
 199
 200         sector_t                        resync_max_sectors; /* may be set by personality */
 201
 202         sector_t                        resync_mismatches; /* count of sectors where
 203                                                             * parity/replica mismatch found
 204                                                             */
 205
 206         /* allow user-space to request suspension of IO to regions of the array */
 207         sector_t                        suspend_lo;
 208         sector_t                        suspend_hi;
 209         /* if zero, use the system-wide default */
 210         int                             sync_speed_min;
 211         int                             sync_speed_max;
 212
 213         /* resync even though the same disks are shared among md-devices */
 214         int                             parallel_resync;
 215
 216         int                             ok_start_degraded;
 217         /* recovery/resync flags
 218          * NEEDED:   we might need to start a resync/recover
 219          * RUNNING:  a thread is running, or about to be started
 220          * SYNC:     actually doing a resync, not a recovery
 221          * RECOVER:  doing recovery, or need to try it.
 222          * INTR:     resync needs to be aborted for some reason
 223          * DONE:     thread is done and is waiting to be reaped
 224          * REQUEST:  user-space has requested a sync (used with SYNC)
 225          * CHECK:    user-space request for check-only, no repair
 226          * RESHAPE:  A reshape is happening
 227          *
 228          * If neither SYNC or RESHAPE are set, then it is a recovery.
 229          */
 230 #define MD_RECOVERY_RUNNING     0
 231 #define MD_RECOVERY_SYNC        1
 232 #define MD_RECOVERY_RECOVER     2
 233 #define MD_RECOVERY_INTR        3
 234 #define MD_RECOVERY_DONE        4
 235 #define MD_RECOVERY_NEEDED      5
 236 #define MD_RECOVERY_REQUESTED   6
 237 #define MD_RECOVERY_CHECK       7
 238 #define MD_RECOVERY_RESHAPE     8
 239 #define MD_RECOVERY_FROZEN      9
 240
 241         unsigned long                   recovery;
 242         /* If a RAID personality determines that recovery (of a particular
 243          * device) will fail due to a read error on the source device, it
 244          * takes a copy of this number and does not attempt recovery again
 245          * until this number changes.
 246          */
 247         int                             recovery_disabled;
 248
 249         int                             in_sync;        /* know to not need resync */
 250         /* 'open_mutex' avoids races between 'md_open' and 'do_md_stop', so
 251          * that we are never stopping an array while it is open.
 252          * 'reconfig_mutex' protects all other reconfiguration.
 253          * These locks are separate due to conflicting interactions
 254          * with bdev->bd_mutex.
 255          * Lock ordering is:
 256          *  reconfig_mutex -> bd_mutex : e.g. do_md_run -> revalidate_disk
 257          *  bd_mutex -> open_mutex:  e.g. __blkdev_get -> md_open
 258          */
 259         struct mutex                    open_mutex;
 260         struct mutex                    reconfig_mutex;
 261         atomic_t                        active;         /* general refcount */
 262         atomic_t                        openers;        /* number of active opens */
 263
 264         int                             changed;        /* True if we might need to
 265                                                          * reread partition info */
 266         int                             degraded;       /* whether md should consider
 267                                                          * adding a spare
 268                                                          */
 269
 270         atomic_t                        recovery_active; /* blocks scheduled, but not written */
 271         wait_queue_head_t               recovery_wait;
 272         sector_t                        recovery_cp;
 273         sector_t                        resync_min;     /* user requested sync
 274                                                          * starts here */
 275         sector_t                        resync_max;     /* resync should pause
 276                                                          * when it gets here */
 277
 278         struct sysfs_dirent             *sysfs_state;   /* handle for 'array_state'
 279                                                          * file in sysfs.
 280                                                          */
 281         struct sysfs_dirent             *sysfs_action;  /* handle for 'sync_action' */
 282
 283         struct work_struct del_work;    /* used for delayed sysfs removal */
 284
 285         spinlock_t                      write_lock;
 286         wait_queue_head_t               sb_wait;        /* for waiting on superblock updates */
 287         atomic_t                        pending_writes; /* number of active superblock writes */
 288
 289         unsigned int                    safemode;       /* if set, update "clean" superblock
 290                                                          * when no writes pending.
 291                                                          */
 292         unsigned int                    safemode_delay;
 293         struct timer_list               safemode_timer;
 294         atomic_t                        writes_pending;
 295         struct request_queue            *queue; /* for plugging ... */
 296
 297         struct bitmap                   *bitmap; /* the bitmap for the device */
 298         struct {
 299                 struct file             *file; /* the bitmap file */
 300                 loff_t                  offset; /* offset from superblock of
 301                                                  * start of bitmap. May be
 302                                                  * negative, but not '0'
 303                                                  * For external metadata, offset
 304                                                  * from start of device.
 305                                                  */
 306                 loff_t                  default_offset; /* this is the offset to use when
 307                                                          * hot-adding a bitmap.  It should
 308                                                          * eventually be settable by sysfs.
 309                                                          */
 310                 /* When md is serving under dm, it might use a
 311                  * dirty_log to store the bits.
 312                  */
 313                 struct dm_dirty_log *log;
 314
 315                 struct mutex            mutex;
 316                 unsigned long           chunksize;
 317                 unsigned long           daemon_sleep; /* how many jiffies between updates? */
 318                 unsigned long           max_write_behind; /* write-behind mode */
 319                 int                     external;
 320         } bitmap_info;
 321
 322         atomic_t                        max_corr_read_errors; /* max read retries */
 323         struct list_head                all_mddevs;
 324
 325         struct attribute_group          *to_remove;
 326
 327         struct bio_set                  *bio_set;
 328
 329         /* Generic flush handling.
 330          * The last to finish preflush schedules a worker to submit
 331          * the rest of the request (without the REQ_FLUSH flag).
 332          */
 333         struct bio *flush_bio;
 334         atomic_t flush_pending;
 335         struct work_struct flush_work;
 336         struct work_struct event_work;  /* used by dm to report failure event */
 337         void (*sync_super)(mddev_t *mddev, mdk_rdev_t *rdev);
 338 };
 339
 340
 341 static inline void rdev_dec_pending(mdk_rdev_t *rdev, mddev_t *mddev)
 342 {
 343         int faulty = test_bit(Faulty, &rdev->flags);
 344         if (atomic_dec_and_test(&rdev->nr_pending) && faulty)
 345                 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
 346 }
 347
 348 static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors)
 349 {
 350         atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io);
 351 }
 352
 353 struct mdk_personality
 354 {
 355         char *name;
 356         int level;
 357         struct list_head list;
 358         struct module *owner;
 359         int (*make_request)(mddev_t *mddev, struct bio *bio);
 360         int (*run)(mddev_t *mddev);
 361         int (*stop)(mddev_t *mddev);
 362         void (*status)(struct seq_file *seq, mddev_t *mddev);
 363         /* error_handler must set ->faulty and clear ->in_sync
 364          * if appropriate, and should abort recovery if needed
 365          */
 366         void (*error_handler)(mddev_t *mddev, mdk_rdev_t *rdev);
 367         int (*hot_add_disk) (mddev_t *mddev, mdk_rdev_t *rdev);
 368         int (*hot_remove_disk) (mddev_t *mddev, int number);
 369         int (*spare_active) (mddev_t *mddev);
 370         sector_t (*sync_request)(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster);
 371         int (*resize) (mddev_t *mddev, sector_t sectors);
 372         sector_t (*size) (mddev_t *mddev, sector_t sectors, int raid_disks);
 373         int (*check_reshape) (mddev_t *mddev);
 374         int (*start_reshape) (mddev_t *mddev);
 375         void (*finish_reshape) (mddev_t *mddev);
 376         /* quiesce moves between quiescence states
 377          * 0 - fully active
 378          * 1 - no new requests allowed
 379          * others - reserved
 380          */
 381         void (*quiesce) (mddev_t *mddev, int state);
 382         /* takeover is used to transition an array from one
 383          * personality to another.  The new personality must be able
 384          * to handle the data in the current layout.
 385          * e.g. 2drive raid1 -> 2drive raid5
 386          *      ndrive raid5 -> degraded n+1drive raid6 with special layout
 387          * If the takeover succeeds, a new 'private' structure is returned.
 388          * This needs to be installed and then ->run used to activate the
 389          * array.
 390          */
 391         void *(*takeover) (mddev_t *mddev);
 392 };
 393
 394
 395 struct md_sysfs_entry {
 396         struct attribute attr;
 397         ssize_t (*show)(mddev_t *, char *);
 398         ssize_t (*store)(mddev_t *, const char *, size_t);
 399 };
 400 extern struct attribute_group md_bitmap_group;
 401
 402 static inline struct sysfs_dirent *sysfs_get_dirent_safe(struct sysfs_dirent *sd, char *name)
 403 {
 404         if (sd)
 405                 return sysfs_get_dirent(sd, NULL, name);
 406         return sd;
 407 }
 408 static inline void sysfs_notify_dirent_safe(struct sysfs_dirent *sd)
 409 {
 410         if (sd)
 411                 sysfs_notify_dirent(sd);
 412 }
 413
 414 static inline char * mdname (mddev_t * mddev)
 415 {
 416         return mddev->gendisk ? mddev->gendisk->disk_name : "mdX";
 417 }
 418
 419 static inline int sysfs_link_rdev(mddev_t *mddev, mdk_rdev_t *rdev)
 420 {
 421         char nm[20];
 422         sprintf(nm, "rd%d", rdev->raid_disk);
 423         return sysfs_create_link(&mddev->kobj, &rdev->kobj, nm);
 424 }
 425
 426 static inline void sysfs_unlink_rdev(mddev_t *mddev, mdk_rdev_t *rdev)
 427 {
 428         char nm[20];
 429         sprintf(nm, "rd%d", rdev->raid_disk);
 430         sysfs_remove_link(&mddev->kobj, nm);
 431 }
 432
 433 /*
 434  * iterates through some rdev ringlist. It's safe to remove the
 435  * current 'rdev'. Dont touch 'tmp' though.
 436  */
 437 #define rdev_for_each_list(rdev, tmp, head)                             \
 438         list_for_each_entry_safe(rdev, tmp, head, same_set)
 439
 440 /*
 441  * iterates through the 'same array disks' ringlist
 442  */
 443 #define rdev_for_each(rdev, tmp, mddev)                         \
 444         list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set)
 445
 446 #define rdev_for_each_rcu(rdev, mddev)                          \
 447         list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set)
 448
 449 typedef struct mdk_thread_s {
 450         void                    (*run) (mddev_t *mddev);
 451         mddev_t                 *mddev;
 452         wait_queue_head_t       wqueue;
 453         unsigned long           flags;
 454         struct task_struct      *tsk;
 455         unsigned long           timeout;
 456 } mdk_thread_t;
 457
 458 #define THREAD_WAKEUP  0
 459
 460 #define __wait_event_lock_irq(wq, condition, lock, cmd)                 \
 461 do {                                                                    \
 462         wait_queue_t __wait;                                            \
 463         init_waitqueue_entry(&__wait, current);                         \
 464                                                                         \
 465         add_wait_queue(&wq, &__wait);                                   \
 466         for (;;) {                                                      \
 467                 set_current_state(TASK_UNINTERRUPTIBLE);                \
 468                 if (condition)                                          \
 469                         break;                                          \
 470                 spin_unlock_irq(&lock);                                 \
 471                 cmd;                                                    \
 472                 schedule();                                             \
 473                 spin_lock_irq(&lock);                                   \
 474         }                                                               \
 475         current->state = TASK_RUNNING;                                  \
 476         remove_wait_queue(&wq, &__wait);                                \
 477 } while (0)
 478
 479 #define wait_event_lock_irq(wq, condition, lock, cmd)                   \
 480 do {                                                                    \
 481         if (condition)                                                  \
 482                 break;                                                  \
 483         __wait_event_lock_irq(wq, condition, lock, cmd);                \
 484 } while (0)
 485
 486 static inline void safe_put_page(struct page *p)
 487 {
 488         if (p) put_page(p);
 489 }
 490
 491 extern int register_md_personality(struct mdk_personality *p);
 492 extern int unregister_md_personality(struct mdk_personality *p);
 493 extern mdk_thread_t * md_register_thread(void (*run) (mddev_t *mddev),
 494                                 mddev_t *mddev, const char *name);
 495 extern void md_unregister_thread(mdk_thread_t *thread);
 496 extern void md_wakeup_thread(mdk_thread_t *thread);
 497 extern void md_check_recovery(mddev_t *mddev);
 498 extern void md_write_start(mddev_t *mddev, struct bio *bi);
 499 extern void md_write_end(mddev_t *mddev);
 500 extern void md_done_sync(mddev_t *mddev, int blocks, int ok);
 501 extern void md_error(mddev_t *mddev, mdk_rdev_t *rdev);
 502
 503 extern int mddev_congested(mddev_t *mddev, int bits);
 504 extern void md_flush_request(mddev_t *mddev, struct bio *bio);
 505 extern void md_super_write(mddev_t *mddev, mdk_rdev_t *rdev,
 506                            sector_t sector, int size, struct page *page);
 507 extern void md_super_wait(mddev_t *mddev);
 508 extern int sync_page_io(mdk_rdev_t *rdev, sector_t sector, int size,
 509                         struct page *page, int rw, bool metadata_op);
 510 extern void md_do_sync(mddev_t *mddev);
 511 extern void md_new_event(mddev_t *mddev);
 512 extern int md_allow_write(mddev_t *mddev);
 513 extern void md_wait_for_blocked_rdev(mdk_rdev_t *rdev, mddev_t *mddev);
 514 extern void md_set_array_sectors(mddev_t *mddev, sector_t array_sectors);
 515 extern int md_check_no_bitmap(mddev_t *mddev);
 516 extern int md_integrity_register(mddev_t *mddev);
 517 extern void md_integrity_add_rdev(mdk_rdev_t *rdev, mddev_t *mddev);
 518 extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale);
 519 extern void restore_bitmap_write_access(struct file *file);
 520
 521 extern void mddev_init(mddev_t *mddev);
 522 extern int md_run(mddev_t *mddev);
 523 extern void md_stop(mddev_t *mddev);
 524 extern void md_stop_writes(mddev_t *mddev);
 525 extern void md_rdev_init(mdk_rdev_t *rdev);
 526
 527 extern void mddev_suspend(mddev_t *mddev);
 528 extern void mddev_resume(mddev_t *mddev);
 529 extern struct bio *bio_clone_mddev(struct bio *bio, gfp_t gfp_mask,
 530                                    mddev_t *mddev);
 531 extern struct bio *bio_alloc_mddev(gfp_t gfp_mask, int nr_iovecs,
 532                                    mddev_t *mddev);
 533 extern int mddev_check_plugged(mddev_t *mddev);
 534 #endif /* _MD_MD_H */