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
 128         int                             suspended;
 129         atomic_t                        active_io;
 130         int                             ro;
 131         int                             sysfs_active; /* set when sysfs deletes
 132                                                        * are happening, so run/
 133                                                        * takeover/stop are not safe
 134                                                        */
 135         int                             ready; /* See when safe to pass
 136                                                 * IO requests down */
 137         struct gendisk                  *gendisk;
 138
 139         struct kobject                  kobj;
 140         int                             hold_active;
 141 #define UNTIL_IOCTL     1
 142 #define UNTIL_STOP      2
 143
 144         /* Superblock information */
 145         int                             major_version,
 146                                         minor_version,
 147                                         patch_version;
 148         int                             persistent;
 149         int                             external;       /* metadata is
 150                                                          * managed externally */
 151         char                            metadata_type[17]; /* externally set*/
 152         int                             chunk_sectors;
 153         time_t                          ctime, utime;
 154         int                             level, layout;
 155         char                            clevel[16];
 156         int                             raid_disks;
 157         int                             max_disks;
 158         sector_t                        dev_sectors;    /* used size of
 159                                                          * component devices */
 160         sector_t                        array_sectors; /* exported array size */
 161         int                             external_size; /* size managed
 162                                                         * externally */
 163         __u64                           events;
 164         /* If the last 'event' was simply a clean->dirty transition, and
 165          * we didn't write it to the spares, then it is safe and simple
 166          * to just decrement the event count on a dirty->clean transition.
 167          * So we record that possibility here.
 168          */
 169         int                             can_decrease_events;
 170
 171         char                            uuid[16];
 172
 173         /* If the array is being reshaped, we need to record the
 174          * new shape and an indication of where we are up to.
 175          * This is written to the superblock.
 176          * If reshape_position is MaxSector, then no reshape is happening (yet).
 177          */
 178         sector_t                        reshape_position;
 179         int                             delta_disks, new_level, new_layout;
 180         int                             new_chunk_sectors;
 181
 182         atomic_t                        plug_cnt;       /* If device is expecting
 183                                                          * more bios soon.
 184                                                          */
 185         struct mdk_thread_s             *thread;        /* management thread */
 186         struct mdk_thread_s             *sync_thread;   /* doing resync or reconstruct */
 187         sector_t                        curr_resync;    /* last block scheduled */
 188         /* As resync requests can complete out of order, we cannot easily track
 189          * how much resync has been completed.  So we occasionally pause until
 190          * everything completes, then set curr_resync_completed to curr_resync.
 191          * As such it may be well behind the real resync mark, but it is a value
 192          * we are certain of.
 193          */
 194         sector_t                        curr_resync_completed;
 195         unsigned long                   resync_mark;    /* a recent timestamp */
 196         sector_t                        resync_mark_cnt;/* blocks written at resync_mark */
 197         sector_t                        curr_mark_cnt; /* blocks scheduled now */
 198
 199         sector_t                        resync_max_sectors; /* may be set by personality */
 200
 201         sector_t                        resync_mismatches; /* count of sectors where
 202                                                             * parity/replica mismatch found
 203                                                             */
 204
 205         /* allow user-space to request suspension of IO to regions of the array */
 206         sector_t                        suspend_lo;
 207         sector_t                        suspend_hi;
 208         /* if zero, use the system-wide default */
 209         int                             sync_speed_min;
 210         int                             sync_speed_max;
 211
 212         /* resync even though the same disks are shared among md-devices */
 213         int                             parallel_resync;
 214
 215         int                             ok_start_degraded;
 216         /* recovery/resync flags
 217          * NEEDED:   we might need to start a resync/recover
 218          * RUNNING:  a thread is running, or about to be started
 219          * SYNC:     actually doing a resync, not a recovery
 220          * RECOVER:  doing recovery, or need to try it.
 221          * INTR:     resync needs to be aborted for some reason
 222          * DONE:     thread is done and is waiting to be reaped
 223          * REQUEST:  user-space has requested a sync (used with SYNC)
 224          * CHECK:    user-space request for check-only, no repair
 225          * RESHAPE:  A reshape is happening
 226          *
 227          * If neither SYNC or RESHAPE are set, then it is a recovery.
 228          */
 229 #define MD_RECOVERY_RUNNING     0
 230 #define MD_RECOVERY_SYNC        1
 231 #define MD_RECOVERY_RECOVER     2
 232 #define MD_RECOVERY_INTR        3
 233 #define MD_RECOVERY_DONE        4
 234 #define MD_RECOVERY_NEEDED      5
 235 #define MD_RECOVERY_REQUESTED   6
 236 #define MD_RECOVERY_CHECK       7
 237 #define MD_RECOVERY_RESHAPE     8
 238 #define MD_RECOVERY_FROZEN      9
 239
 240         unsigned long                   recovery;
 241         int                             recovery_disabled; /* if we detect that recovery
 242                                                             * will always fail, set this
 243                                                             * so we don't loop trying */
 244
 245         int                             in_sync;        /* know to not need resync */
 246         /* 'open_mutex' avoids races between 'md_open' and 'do_md_stop', so
 247          * that we are never stopping an array while it is open.
 248          * 'reconfig_mutex' protects all other reconfiguration.
 249          * These locks are separate due to conflicting interactions
 250          * with bdev->bd_mutex.
 251          * Lock ordering is:
 252          *  reconfig_mutex -> bd_mutex : e.g. do_md_run -> revalidate_disk
 253          *  bd_mutex -> open_mutex:  e.g. __blkdev_get -> md_open
 254          */
 255         struct mutex                    open_mutex;
 256         struct mutex                    reconfig_mutex;
 257         atomic_t                        active;         /* general refcount */
 258         atomic_t                        openers;        /* number of active opens */
 259
 260         int                             changed;        /* True if we might need to
 261                                                          * reread partition info */
 262         int                             degraded;       /* whether md should consider
 263                                                          * adding a spare
 264                                                          */
 265
 266         atomic_t                        recovery_active; /* blocks scheduled, but not written */
 267         wait_queue_head_t               recovery_wait;
 268         sector_t                        recovery_cp;
 269         sector_t                        resync_min;     /* user requested sync
 270                                                          * starts here */
 271         sector_t                        resync_max;     /* resync should pause
 272                                                          * when it gets here */
 273
 274         struct sysfs_dirent             *sysfs_state;   /* handle for 'array_state'
 275                                                          * file in sysfs.
 276                                                          */
 277         struct sysfs_dirent             *sysfs_action;  /* handle for 'sync_action' */
 278
 279         struct work_struct del_work;    /* used for delayed sysfs removal */
 280
 281         spinlock_t                      write_lock;
 282         wait_queue_head_t               sb_wait;        /* for waiting on superblock updates */
 283         atomic_t                        pending_writes; /* number of active superblock writes */
 284
 285         unsigned int                    safemode;       /* if set, update "clean" superblock
 286                                                          * when no writes pending.
 287                                                          */
 288         unsigned int                    safemode_delay;
 289         struct timer_list               safemode_timer;
 290         atomic_t                        writes_pending;
 291         struct request_queue            *queue; /* for plugging ... */
 292
 293         struct bitmap                   *bitmap; /* the bitmap for the device */
 294         struct {
 295                 struct file             *file; /* the bitmap file */
 296                 loff_t                  offset; /* offset from superblock of
 297                                                  * start of bitmap. May be
 298                                                  * negative, but not '0'
 299                                                  * For external metadata, offset
 300                                                  * from start of device.
 301                                                  */
 302                 loff_t                  default_offset; /* this is the offset to use when
 303                                                          * hot-adding a bitmap.  It should
 304                                                          * eventually be settable by sysfs.
 305                                                          */
 306                 /* When md is serving under dm, it might use a
 307                  * dirty_log to store the bits.
 308                  */
 309                 struct dm_dirty_log *log;
 310
 311                 struct mutex            mutex;
 312                 unsigned long           chunksize;
 313                 unsigned long           daemon_sleep; /* how many jiffies between updates? */
 314                 unsigned long           max_write_behind; /* write-behind mode */
 315                 int                     external;
 316         } bitmap_info;
 317
 318         atomic_t                        max_corr_read_errors; /* max read retries */
 319         struct list_head                all_mddevs;
 320
 321         struct attribute_group          *to_remove;
 322
 323         struct bio_set                  *bio_set;
 324
 325         /* Generic flush handling.
 326          * The last to finish preflush schedules a worker to submit
 327          * the rest of the request (without the REQ_FLUSH flag).
 328          */
 329         struct bio *flush_bio;
 330         atomic_t flush_pending;
 331         struct work_struct flush_work;
 332         struct work_struct event_work;  /* used by dm to report failure event */
 333 };
 334
 335
 336 static inline void rdev_dec_pending(mdk_rdev_t *rdev, mddev_t *mddev)
 337 {
 338         int faulty = test_bit(Faulty, &rdev->flags);
 339         if (atomic_dec_and_test(&rdev->nr_pending) && faulty)
 340                 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
 341 }
 342
 343 static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors)
 344 {
 345         atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io);
 346 }
 347
 348 struct mdk_personality
 349 {
 350         char *name;
 351         int level;
 352         struct list_head list;
 353         struct module *owner;
 354         int (*make_request)(mddev_t *mddev, struct bio *bio);
 355         int (*run)(mddev_t *mddev);
 356         int (*stop)(mddev_t *mddev);
 357         void (*status)(struct seq_file *seq, mddev_t *mddev);
 358         /* error_handler must set ->faulty and clear ->in_sync
 359          * if appropriate, and should abort recovery if needed
 360          */
 361         void (*error_handler)(mddev_t *mddev, mdk_rdev_t *rdev);
 362         int (*hot_add_disk) (mddev_t *mddev, mdk_rdev_t *rdev);
 363         int (*hot_remove_disk) (mddev_t *mddev, int number);
 364         int (*spare_active) (mddev_t *mddev);
 365         sector_t (*sync_request)(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster);
 366         int (*resize) (mddev_t *mddev, sector_t sectors);
 367         sector_t (*size) (mddev_t *mddev, sector_t sectors, int raid_disks);
 368         int (*check_reshape) (mddev_t *mddev);
 369         int (*start_reshape) (mddev_t *mddev);
 370         void (*finish_reshape) (mddev_t *mddev);
 371         /* quiesce moves between quiescence states
 372          * 0 - fully active
 373          * 1 - no new requests allowed
 374          * others - reserved
 375          */
 376         void (*quiesce) (mddev_t *mddev, int state);
 377         /* takeover is used to transition an array from one
 378          * personality to another.  The new personality must be able
 379          * to handle the data in the current layout.
 380          * e.g. 2drive raid1 -> 2drive raid5
 381          *      ndrive raid5 -> degraded n+1drive raid6 with special layout
 382          * If the takeover succeeds, a new 'private' structure is returned.
 383          * This needs to be installed and then ->run used to activate the
 384          * array.
 385          */
 386         void *(*takeover) (mddev_t *mddev);
 387 };
 388
 389
 390 struct md_sysfs_entry {
 391         struct attribute attr;
 392         ssize_t (*show)(mddev_t *, char *);
 393         ssize_t (*store)(mddev_t *, const char *, size_t);
 394 };
 395 extern struct attribute_group md_bitmap_group;
 396
 397 static inline struct sysfs_dirent *sysfs_get_dirent_safe(struct sysfs_dirent *sd, char *name)
 398 {
 399         if (sd)
 400                 return sysfs_get_dirent(sd, NULL, name);
 401         return sd;
 402 }
 403 static inline void sysfs_notify_dirent_safe(struct sysfs_dirent *sd)
 404 {
 405         if (sd)
 406                 sysfs_notify_dirent(sd);
 407 }
 408
 409 static inline char * mdname (mddev_t * mddev)
 410 {
 411         return mddev->gendisk ? mddev->gendisk->disk_name : "mdX";
 412 }
 413
 414 /*
 415  * iterates through some rdev ringlist. It's safe to remove the
 416  * current 'rdev'. Dont touch 'tmp' though.
 417  */
 418 #define rdev_for_each_list(rdev, tmp, head)                             \
 419         list_for_each_entry_safe(rdev, tmp, head, same_set)
 420
 421 /*
 422  * iterates through the 'same array disks' ringlist
 423  */
 424 #define rdev_for_each(rdev, tmp, mddev)                         \
 425         list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set)
 426
 427 #define rdev_for_each_rcu(rdev, mddev)                          \
 428         list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set)
 429
 430 typedef struct mdk_thread_s {
 431         void                    (*run) (mddev_t *mddev);
 432         mddev_t                 *mddev;
 433         wait_queue_head_t       wqueue;
 434         unsigned long           flags;
 435         struct task_struct      *tsk;
 436         unsigned long           timeout;
 437 } mdk_thread_t;
 438
 439 #define THREAD_WAKEUP  0
 440
 441 #define __wait_event_lock_irq(wq, condition, lock, cmd)                 \
 442 do {                                                                    \
 443         wait_queue_t __wait;                                            \
 444         init_waitqueue_entry(&__wait, current);                         \
 445                                                                         \
 446         add_wait_queue(&wq, &__wait);                                   \
 447         for (;;) {                                                      \
 448                 set_current_state(TASK_UNINTERRUPTIBLE);                \
 449                 if (condition)                                          \
 450                         break;                                          \
 451                 spin_unlock_irq(&lock);                                 \
 452                 cmd;                                                    \
 453                 schedule();                                             \
 454                 spin_lock_irq(&lock);                                   \
 455         }                                                               \
 456         current->state = TASK_RUNNING;                                  \
 457         remove_wait_queue(&wq, &__wait);                                \
 458 } while (0)
 459
 460 #define wait_event_lock_irq(wq, condition, lock, cmd)                   \
 461 do {                                                                    \
 462         if (condition)                                                  \
 463                 break;                                                  \
 464         __wait_event_lock_irq(wq, condition, lock, cmd);                \
 465 } while (0)
 466
 467 static inline void safe_put_page(struct page *p)
 468 {
 469         if (p) put_page(p);
 470 }
 471
 472 extern int register_md_personality(struct mdk_personality *p);
 473 extern int unregister_md_personality(struct mdk_personality *p);
 474 extern mdk_thread_t * md_register_thread(void (*run) (mddev_t *mddev),
 475                                 mddev_t *mddev, const char *name);
 476 extern void md_unregister_thread(mdk_thread_t *thread);
 477 extern void md_wakeup_thread(mdk_thread_t *thread);
 478 extern void md_check_recovery(mddev_t *mddev);
 479 extern void md_write_start(mddev_t *mddev, struct bio *bi);
 480 extern void md_write_end(mddev_t *mddev);
 481 extern void md_done_sync(mddev_t *mddev, int blocks, int ok);
 482 extern void md_error(mddev_t *mddev, mdk_rdev_t *rdev);
 483
 484 extern int mddev_congested(mddev_t *mddev, int bits);
 485 extern void md_flush_request(mddev_t *mddev, struct bio *bio);
 486 extern void md_super_write(mddev_t *mddev, mdk_rdev_t *rdev,
 487                            sector_t sector, int size, struct page *page);
 488 extern void md_super_wait(mddev_t *mddev);
 489 extern int sync_page_io(mdk_rdev_t *rdev, sector_t sector, int size,
 490                         struct page *page, int rw, bool metadata_op);
 491 extern void md_do_sync(mddev_t *mddev);
 492 extern void md_new_event(mddev_t *mddev);
 493 extern int md_allow_write(mddev_t *mddev);
 494 extern void md_wait_for_blocked_rdev(mdk_rdev_t *rdev, mddev_t *mddev);
 495 extern void md_set_array_sectors(mddev_t *mddev, sector_t array_sectors);
 496 extern int md_check_no_bitmap(mddev_t *mddev);
 497 extern int md_integrity_register(mddev_t *mddev);
 498 extern void md_integrity_add_rdev(mdk_rdev_t *rdev, mddev_t *mddev);
 499 extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale);
 500 extern void restore_bitmap_write_access(struct file *file);
 501
 502 extern void mddev_init(mddev_t *mddev);
 503 extern int md_run(mddev_t *mddev);
 504 extern void md_stop(mddev_t *mddev);
 505 extern void md_stop_writes(mddev_t *mddev);
 506 extern void md_rdev_init(mdk_rdev_t *rdev);
 507
 508 extern void mddev_suspend(mddev_t *mddev);
 509 extern void mddev_resume(mddev_t *mddev);
 510 extern struct bio *bio_clone_mddev(struct bio *bio, gfp_t gfp_mask,
 511                                    mddev_t *mddev);
 512 extern struct bio *bio_alloc_mddev(gfp_t gfp_mask, int nr_iovecs,
 513                                    mddev_t *mddev);
 514 extern int mddev_check_plugged(mddev_t *mddev);
 515 #endif /* _MD_MD_H */