include/linux/raid/md_k.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_K_H
  16 #define _MD_K_H
  17
  18 #define MD_RESERVED       0UL
  19 #define LINEAR            1UL
  20 #define RAID0             2UL
  21 #define RAID1             3UL
  22 #define RAID5             4UL
  23 #define TRANSLUCENT       5UL
  24 #define HSM               6UL
  25 #define MULTIPATH         7UL
  26 #define RAID6             8UL
  27 #define RAID10            9UL
  28 #define FAULTY            10UL
  29 #define MAX_PERSONALITY   11UL
  30
  31 #define LEVEL_MULTIPATH         (-4)
  32 #define LEVEL_LINEAR            (-1)
  33 #define LEVEL_FAULTY            (-5)
  34
  35 #define MaxSector (~(sector_t)0)
  36 #define MD_THREAD_NAME_MAX 14
  37
  38 static inline int pers_to_level (int pers)
  39 {
  40         switch (pers) {
  41                 case FAULTY:            return LEVEL_FAULTY;
  42                 case MULTIPATH:         return LEVEL_MULTIPATH;
  43                 case HSM:               return -3;
  44                 case TRANSLUCENT:       return -2;
  45                 case LINEAR:            return LEVEL_LINEAR;
  46                 case RAID0:             return 0;
  47                 case RAID1:             return 1;
  48                 case RAID5:             return 5;
  49                 case RAID6:             return 6;
  50                 case RAID10:            return 10;
  51         }
  52         BUG();
  53         return MD_RESERVED;
  54 }
  55
  56 static inline int level_to_pers (int level)
  57 {
  58         switch (level) {
  59                 case LEVEL_FAULTY: return FAULTY;
  60                 case LEVEL_MULTIPATH: return MULTIPATH;
  61                 case -3: return HSM;
  62                 case -2: return TRANSLUCENT;
  63                 case LEVEL_LINEAR: return LINEAR;
  64                 case 0: return RAID0;
  65                 case 1: return RAID1;
  66                 case 4:
  67                 case 5: return RAID5;
  68                 case 6: return RAID6;
  69                 case 10: return RAID10;
  70         }
  71         return MD_RESERVED;
  72 }
  73
  74 typedef struct mddev_s mddev_t;
  75 typedef struct mdk_rdev_s mdk_rdev_t;
  76
  77 #define MAX_MD_DEVS  256        /* Max number of md dev */
  78
  79 /*
  80  * options passed in raidrun:
  81  */
  82
  83 #define MAX_CHUNK_SIZE (4096*1024)
  84
  85 /*
  86  * default readahead
  87  */
  88
  89 static inline int disk_faulty(mdp_disk_t * d)
  90 {
  91         return d->state & (1 << MD_DISK_FAULTY);
  92 }
  93
  94 static inline int disk_active(mdp_disk_t * d)
  95 {
  96         return d->state & (1 << MD_DISK_ACTIVE);
  97 }
  98
  99 static inline int disk_sync(mdp_disk_t * d)
 100 {
 101         return d->state & (1 << MD_DISK_SYNC);
 102 }
 103
 104 static inline int disk_spare(mdp_disk_t * d)
 105 {
 106         return !disk_sync(d) && !disk_active(d) && !disk_faulty(d);
 107 }
 108
 109 static inline int disk_removed(mdp_disk_t * d)
 110 {
 111         return d->state & (1 << MD_DISK_REMOVED);
 112 }
 113
 114 static inline void mark_disk_faulty(mdp_disk_t * d)
 115 {
 116         d->state |= (1 << MD_DISK_FAULTY);
 117 }
 118
 119 static inline void mark_disk_active(mdp_disk_t * d)
 120 {
 121         d->state |= (1 << MD_DISK_ACTIVE);
 122 }
 123
 124 static inline void mark_disk_sync(mdp_disk_t * d)
 125 {
 126         d->state |= (1 << MD_DISK_SYNC);
 127 }
 128
 129 static inline void mark_disk_spare(mdp_disk_t * d)
 130 {
 131         d->state = 0;
 132 }
 133
 134 static inline void mark_disk_removed(mdp_disk_t * d)
 135 {
 136         d->state = (1 << MD_DISK_FAULTY) | (1 << MD_DISK_REMOVED);
 137 }
 138
 139 static inline void mark_disk_inactive(mdp_disk_t * d)
 140 {
 141         d->state &= ~(1 << MD_DISK_ACTIVE);
 142 }
 143
 144 static inline void mark_disk_nonsync(mdp_disk_t * d)
 145 {
 146         d->state &= ~(1 << MD_DISK_SYNC);
 147 }
 148
 149 /*
 150  * MD's 'extended' device
 151  */
 152 struct mdk_rdev_s
 153 {
 154         struct list_head same_set;      /* RAID devices within the same set */
 155
 156         sector_t size;                  /* Device size (in blocks) */
 157         mddev_t *mddev;                 /* RAID array if running */
 158         unsigned long last_events;      /* IO event timestamp */
 159
 160         struct block_device *bdev;      /* block device handle */
 161
 162         struct page     *sb_page;
 163         int             sb_loaded;
 164         sector_t        data_offset;    /* start of data in array */
 165         sector_t        sb_offset;
 166         int             preferred_minor;        /* autorun support */
 167
 168         /* A device can be in one of three states based on two flags:
 169          * Not working:   faulty==1 in_sync==0
 170          * Fully working: faulty==0 in_sync==1
 171          * Working, but not
 172          * in sync with array
 173          *                faulty==0 in_sync==0
 174          *
 175          * It can never have faulty==1, in_sync==1
 176          * This reduces the burden of testing multiple flags in many cases
 177          */
 178         int faulty;                     /* if faulty do not issue IO requests */
 179         int in_sync;                    /* device is a full member of the array */
 180
 181         int desc_nr;                    /* descriptor index in the superblock */
 182         int raid_disk;                  /* role of device in array */
 183
 184         atomic_t        nr_pending;     /* number of pending requests.
 185                                          * only maintained for arrays that
 186                                          * support hot removal
 187                                          */
 188 };
 189
 190 typedef struct mdk_personality_s mdk_personality_t;
 191
 192 struct mddev_s
 193 {
 194         void                            *private;
 195         mdk_personality_t               *pers;
 196         dev_t                           unit;
 197         int                             md_minor;
 198         struct list_head                disks;
 199         int                             sb_dirty;
 200         int                             ro;
 201
 202         struct gendisk                  *gendisk;
 203
 204         /* Superblock information */
 205         int                             major_version,
 206                                         minor_version,
 207                                         patch_version;
 208         int                             persistent;
 209         int                             chunk_size;
 210         time_t                          ctime, utime;
 211         int                             level, layout;
 212         int                             raid_disks;
 213         int                             max_disks;
 214         sector_t                        size; /* used size of component devices */
 215         sector_t                        array_size; /* exported array size */
 216         __u64                           events;
 217
 218         char                            uuid[16];
 219
 220         struct mdk_thread_s             *thread;        /* management thread */
 221         struct mdk_thread_s             *sync_thread;   /* doing resync or reconstruct */
 222         sector_t                        curr_resync;    /* blocks scheduled */
 223         unsigned long                   resync_mark;    /* a recent timestamp */
 224         sector_t                        resync_mark_cnt;/* blocks written at resync_mark */
 225
 226         sector_t                        resync_max_sectors; /* may be set by personality */
 227         /* recovery/resync flags
 228          * NEEDED:   we might need to start a resync/recover
 229          * RUNNING:  a thread is running, or about to be started
 230          * SYNC:     actually doing a resync, not a recovery
 231          * ERR:      and IO error was detected - abort the resync/recovery
 232          * INTR:     someone requested a (clean) early abort.
 233          * DONE:     thread is done and is waiting to be reaped
 234          */
 235 #define MD_RECOVERY_RUNNING     0
 236 #define MD_RECOVERY_SYNC        1
 237 #define MD_RECOVERY_ERR         2
 238 #define MD_RECOVERY_INTR        3
 239 #define MD_RECOVERY_DONE        4
 240 #define MD_RECOVERY_NEEDED      5
 241         unsigned long                   recovery;
 242
 243         int                             in_sync;        /* know to not need resync */
 244         struct semaphore                reconfig_sem;
 245         atomic_t                        active;
 246
 247         int                             changed;        /* true if we might need to reread partition info */
 248         int                             degraded;       /* whether md should consider
 249                                                          * adding a spare
 250                                                          */
 251
 252         atomic_t                        recovery_active; /* blocks scheduled, but not written */
 253         wait_queue_head_t               recovery_wait;
 254         sector_t                        recovery_cp;
 255         unsigned int                    safemode;       /* if set, update "clean" superblock
 256                                                          * when no writes pending.
 257                                                          */
 258         unsigned int                    safemode_delay;
 259         struct timer_list               safemode_timer;
 260         atomic_t                        writes_pending;
 261         request_queue_t                 *queue; /* for plugging ... */
 262
 263         struct list_head                all_mddevs;
 264 };
 265
 266
 267 static inline void rdev_dec_pending(mdk_rdev_t *rdev, mddev_t *mddev)
 268 {
 269         int faulty = rdev->faulty;
 270         if (atomic_dec_and_test(&rdev->nr_pending) && faulty)
 271                 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
 272 }
 273
 274 static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors)
 275 {
 276         atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io);
 277 }
 278
 279 struct mdk_personality_s
 280 {
 281         char *name;
 282         struct module *owner;
 283         int (*make_request)(request_queue_t *q, struct bio *bio);
 284         int (*run)(mddev_t *mddev);
 285         int (*stop)(mddev_t *mddev);
 286         void (*status)(struct seq_file *seq, mddev_t *mddev);
 287         /* error_handler must set ->faulty and clear ->in_sync
 288          * if appropriate, and should abort recovery if needed
 289          */
 290         void (*error_handler)(mddev_t *mddev, mdk_rdev_t *rdev);
 291         int (*hot_add_disk) (mddev_t *mddev, mdk_rdev_t *rdev);
 292         int (*hot_remove_disk) (mddev_t *mddev, int number);
 293         int (*spare_active) (mddev_t *mddev);
 294         int (*sync_request)(mddev_t *mddev, sector_t sector_nr, int go_faster);
 295         int (*resize) (mddev_t *mddev, sector_t sectors);
 296         int (*reshape) (mddev_t *mddev, int raid_disks);
 297         int (*reconfig) (mddev_t *mddev, int layout, int chunk_size);
 298 };
 299
 300
 301 static inline char * mdname (mddev_t * mddev)
 302 {
 303         return mddev->gendisk ? mddev->gendisk->disk_name : "mdX";
 304 }
 305
 306 extern mdk_rdev_t * find_rdev_nr(mddev_t *mddev, int nr);
 307
 308 /*
 309  * iterates through some rdev ringlist. It's safe to remove the
 310  * current 'rdev'. Dont touch 'tmp' though.
 311  */
 312 #define ITERATE_RDEV_GENERIC(head,rdev,tmp)                             \
 313                                                                         \
 314         for ((tmp) = (head).next;                                       \
 315                 (rdev) = (list_entry((tmp), mdk_rdev_t, same_set)),     \
 316                         (tmp) = (tmp)->next, (tmp)->prev != &(head)     \
 317                 ; )
 318 /*
 319  * iterates through the 'same array disks' ringlist
 320  */
 321 #define ITERATE_RDEV(mddev,rdev,tmp)                                    \
 322         ITERATE_RDEV_GENERIC((mddev)->disks,rdev,tmp)
 323
 324 /*
 325  * Iterates through 'pending RAID disks'
 326  */
 327 #define ITERATE_RDEV_PENDING(rdev,tmp)                                  \
 328         ITERATE_RDEV_GENERIC(pending_raid_disks,rdev,tmp)
 329
 330 typedef struct mdk_thread_s {
 331         void                    (*run) (mddev_t *mddev);
 332         mddev_t                 *mddev;
 333         wait_queue_head_t       wqueue;
 334         unsigned long           flags;
 335         struct completion       *event;
 336         struct task_struct      *tsk;
 337         const char              *name;
 338 } mdk_thread_t;
 339
 340 #define THREAD_WAKEUP  0
 341
 342 #define __wait_event_lock_irq(wq, condition, lock, cmd)                 \
 343 do {                                                                    \
 344         wait_queue_t __wait;                                            \
 345         init_waitqueue_entry(&__wait, current);                         \
 346                                                                         \
 347         add_wait_queue(&wq, &__wait);                                   \
 348         for (;;) {                                                      \
 349                 set_current_state(TASK_UNINTERRUPTIBLE);                \
 350                 if (condition)                                          \
 351                         break;                                          \
 352                 spin_unlock_irq(&lock);                                 \
 353                 cmd;                                                    \
 354                 schedule();                                             \
 355                 spin_lock_irq(&lock);                                   \
 356         }                                                               \
 357         current->state = TASK_RUNNING;                                  \
 358         remove_wait_queue(&wq, &__wait);                                \
 359 } while (0)
 360
 361 #define wait_event_lock_irq(wq, condition, lock, cmd)                   \
 362 do {                                                                    \
 363         if (condition)                                                  \
 364                 break;                                                  \
 365         __wait_event_lock_irq(wq, condition, lock, cmd);                \
 366 } while (0)
 367
 368 #endif
 369