mdadm.h

   1 /*
   2  * mdadm - manage Linux "md" devices aka RAID arrays.
   3  *
   4  * Copyright (C) 2001-2009 Neil Brown <neilb@suse.de>
   5  *
   6  *
   7  *    This program is free software; you can redistribute it and/or modify
   8  *    it under the terms of the GNU General Public License as published by
   9  *    the Free Software Foundation; either version 2 of the License, or
  10  *    (at your option) any later version.
  11  *
  12  *    This program is distributed in the hope that it will be useful,
  13  *    but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  *    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  15  *    GNU General Public License for more details.
  16  *
  17  *    You should have received a copy of the GNU General Public License
  18  *    along with this program; if not, write to the Free Software
  19  *    Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  20  *
  21  *    Author: Neil Brown
  22  *    Email: <neilb@suse.de>
  23  */
  24
  25 #define _GNU_SOURCE
  26 #define _FILE_OFFSET_BITS 64
  27 #include        <unistd.h>
  28 #if !defined(__dietlibc__) && !defined(__KLIBC__)
  29 extern __off64_t lseek64 __P ((int __fd, __off64_t __offset, int __whence));
  30 #else
  31 # if defined(__NO_STAT64) || __WORDSIZE != 32
  32 # define lseek64 lseek
  33 # endif
  34 #endif
  35
  36 #include        <sys/types.h>
  37 #include        <sys/stat.h>
  38 #include        <stdlib.h>
  39 #include        <time.h>
  40 #include        <sys/time.h>
  41 #include        <getopt.h>
  42 #include        <fcntl.h>
  43 #include        <stdio.h>
  44 #include        <errno.h>
  45 #include        <string.h>
  46 #include        <syslog.h>
  47 #ifdef __dietlibc__
  48 #include        <strings.h>
  49 /* dietlibc has deprecated random and srandom!! */
  50 #define random rand
  51 #define srandom srand
  52 #endif
  53
  54
  55 #include        <linux/kdev_t.h>
  56 /*#include      <linux/fs.h> */
  57 #include        <sys/mount.h>
  58 #include        <asm/types.h>
  59 #include        <sys/ioctl.h>
  60 #define MD_MAJOR 9
  61 #define MdpMinorShift 6
  62
  63 #ifndef BLKGETSIZE64
  64 #define BLKGETSIZE64 _IOR(0x12,114,size_t) /* return device size in bytes (u64 *arg) */
  65 #endif
  66
  67 #define DEFAULT_CHUNK 512
  68 #define DEFAULT_BITMAP_CHUNK 4096
  69 #define DEFAULT_BITMAP_DELAY 5
  70 #define DEFAULT_MAX_WRITE_BEHIND 256
  71
  72 /* MAP_DIR should be somewhere that persists across the pivotroot
  73  * from early boot to late boot.
  74  * /run  seems to have emerged as the best standard.
  75  */
  76 #ifndef MAP_DIR
  77 #define MAP_DIR "/run/mdadm"
  78 #endif /* MAP_DIR */
  79 /* MAP_FILE is what we name the map file we put in MAP_DIR, in case you
  80  * want something other than the default of "map"
  81  */
  82 #ifndef MAP_FILE
  83 #define MAP_FILE "map"
  84 #endif /* MAP_FILE */
  85 /* MDMON_DIR is where pid and socket files used for communicating
  86  * with mdmon normally live.  Best is /var/run/mdadm as
  87  * mdmon is needed at early boot then it needs to write there prior
  88  * to /var/run being mounted read/write, and it also then needs to
  89  * persist beyond when /var/run is mounter read-only.  So, to be
  90  * safe, the default is somewhere that is read/write early in the
  91  * boot process and stays up as long as possible during shutdown.
  92  */
  93 #ifndef MDMON_DIR
  94 #define MDMON_DIR "/run/mdadm"
  95 #endif /* MDMON_DIR */
  96
  97 /* FAILED_SLOTS is where to save files storing recent removal of array
  98  * member in order to allow future reuse of disk inserted in the same
  99  * slot for array recovery
 100  */
 101 #ifndef FAILED_SLOTS_DIR
 102 #define FAILED_SLOTS_DIR "/run/mdadm/failed-slots"
 103 #endif /* FAILED_SLOTS */
 104
 105 #include        "md_u.h"
 106 #include        "md_p.h"
 107 #include        "bitmap.h"
 108 #include        "msg.h"
 109
 110 #include <endian.h>
 111 /* Redhat don't like to #include <asm/byteorder.h>, and
 112  * some time include <linux/byteorder/xxx_endian.h> isn't enough,
 113  * and there is no standard conversion function so... */
 114 /* And dietlibc doesn't think byteswap is ok, so.. */
 115 /*  #include <byteswap.h> */
 116 #define bswap_16(x) (((x) & 0x00ffU) << 8 | \
 117                      ((x) & 0xff00U) >> 8)
 118 #define bswap_32(x) (((x) & 0x000000ffU) << 24 | \
 119                      ((x) & 0xff000000U) >> 24 | \
 120                      ((x) & 0x0000ff00U) << 8  | \
 121                      ((x) & 0x00ff0000U) >> 8)
 122 #define bswap_64(x) (((x) & 0x00000000000000ffULL) << 56 | \
 123                      ((x) & 0xff00000000000000ULL) >> 56 | \
 124                      ((x) & 0x000000000000ff00ULL) << 40 | \
 125                      ((x) & 0x00ff000000000000ULL) >> 40 | \
 126                      ((x) & 0x0000000000ff0000ULL) << 24 | \
 127                      ((x) & 0x0000ff0000000000ULL) >> 24 | \
 128                      ((x) & 0x00000000ff000000ULL) << 8 | \
 129                      ((x) & 0x000000ff00000000ULL) >> 8)
 130
 131 #if !defined(__KLIBC__)
 132 #if BYTE_ORDER == LITTLE_ENDIAN
 133 #define __cpu_to_le16(_x) (_x)
 134 #define __cpu_to_le32(_x) (_x)
 135 #define __cpu_to_le64(_x) (_x)
 136 #define __le16_to_cpu(_x) (_x)
 137 #define __le32_to_cpu(_x) (_x)
 138 #define __le64_to_cpu(_x) (_x)
 139
 140 #define __cpu_to_be16(_x) bswap_16(_x)
 141 #define __cpu_to_be32(_x) bswap_32(_x)
 142 #define __cpu_to_be64(_x) bswap_64(_x)
 143 #define __be16_to_cpu(_x) bswap_16(_x)
 144 #define __be32_to_cpu(_x) bswap_32(_x)
 145 #define __be64_to_cpu(_x) bswap_64(_x)
 146 #elif BYTE_ORDER == BIG_ENDIAN
 147 #define __cpu_to_le16(_x) bswap_16(_x)
 148 #define __cpu_to_le32(_x) bswap_32(_x)
 149 #define __cpu_to_le64(_x) bswap_64(_x)
 150 #define __le16_to_cpu(_x) bswap_16(_x)
 151 #define __le32_to_cpu(_x) bswap_32(_x)
 152 #define __le64_to_cpu(_x) bswap_64(_x)
 153
 154 #define __cpu_to_be16(_x) (_x)
 155 #define __cpu_to_be32(_x) (_x)
 156 #define __cpu_to_be64(_x) (_x)
 157 #define __be16_to_cpu(_x) (_x)
 158 #define __be32_to_cpu(_x) (_x)
 159 #define __be64_to_cpu(_x) (_x)
 160 #else
 161 #  error "unknown endianness."
 162 #endif
 163 #endif /* __KLIBC__ */
 164
 165
 166 /*
 167  * min()/max()/clamp() macros that also do
 168  * strict type-checking.. See the
 169  * "unnecessary" pointer comparison.
 170  */
 171 #define min(x, y) ({                            \
 172         typeof(x) _min1 = (x);                  \
 173         typeof(y) _min2 = (y);                  \
 174         (void) (&_min1 == &_min2);              \
 175         _min1 < _min2 ? _min1 : _min2; })
 176
 177 #define max(x, y) ({                            \
 178         typeof(x) _max1 = (x);                  \
 179         typeof(y) _max2 = (y);                  \
 180         (void) (&_max1 == &_max2);              \
 181         _max1 > _max2 ? _max1 : _max2; })
 182
 183 /* general information that might be extracted from a superblock */
 184 struct mdinfo {
 185         mdu_array_info_t        array;
 186         mdu_disk_info_t         disk;
 187         __u64                   events;
 188         int                     uuid[4];
 189         char                    name[33];
 190         unsigned long long      data_offset;
 191         unsigned long long      component_size; /* same as array.size, except in
 192                                                  * sectors and up to 64bits.
 193                                                  */
 194         unsigned long long      custom_array_size; /* size for non-default sized
 195                                                     * arrays (in sectors)
 196                                                     */
 197 #define NO_RESHAPE              0
 198 #define VOLUME_RESHAPE          1
 199 #define CONTAINER_RESHAPE       2
 200         int                     reshape_active;
 201         unsigned long long      reshape_progress;
 202         int                     recovery_blocked; /* for external metadata it
 203                                                    * indicates that there is
 204                                                    * reshape in progress in
 205                                                    * container,
 206                                                    * for native metadata it is
 207                                                    * reshape_active field mirror
 208                                                    */
 209         union {
 210                 unsigned long long resync_start; /* per-array resync position */
 211                 unsigned long long recovery_start; /* per-device rebuild position */
 212                 #define MaxSector  (~0ULL) /* resync/recovery complete position */
 213         };
 214         long                    bitmap_offset;  /* 0 == none, 1 == a file */
 215         unsigned long           safe_mode_delay; /* ms delay to mark clean */
 216         int                     new_level, delta_disks, new_layout, new_chunk;
 217         int                     errors;
 218         unsigned long           cache_size; /* size of raid456 stripe cache*/
 219         int                     mismatch_cnt;
 220         char                    text_version[50];
 221
 222         int container_member; /* for assembling external-metatdata arrays
 223                                * This is to be used internally by metadata
 224                                * handler only */
 225         int container_enough; /* flag external handlers can set to
 226                                * indicate that subarrays have not enough (-1),
 227                                * enough to start (0), or all expected disks (1) */
 228         char            sys_name[20];
 229         struct mdinfo *devs;
 230         struct mdinfo *next;
 231
 232         /* Device info for mdmon: */
 233         int recovery_fd;
 234         int state_fd;
 235         #define DS_FAULTY       1
 236         #define DS_INSYNC       2
 237         #define DS_WRITE_MOSTLY 4
 238         #define DS_SPARE        8
 239         #define DS_BLOCKED      16
 240         #define DS_REMOVE       1024
 241         #define DS_UNBLOCK      2048
 242         int prev_state, curr_state, next_state;
 243
 244 };
 245
 246 struct createinfo {
 247         int     uid;
 248         int     gid;
 249         int     autof;
 250         int     mode;
 251         int     symlinks;
 252         struct supertype *supertype;
 253 };
 254
 255 #define Name "mdadm"
 256
 257 enum mode {
 258         ASSEMBLE=1,
 259         BUILD,
 260         CREATE,
 261         MANAGE,
 262         MISC,
 263         MONITOR,
 264         GROW,
 265         INCREMENTAL,
 266         AUTODETECT,
 267         mode_count
 268 };
 269
 270 extern char short_options[];
 271 extern char short_bitmap_options[];
 272 extern char short_bitmap_auto_options[];
 273 extern struct option long_options[];
 274 extern char Version[], Usage[], Help[], OptionHelp[],
 275         *mode_help[],
 276         Help_create[], Help_build[], Help_assemble[], Help_grow[],
 277         Help_incr[],
 278         Help_manage[], Help_misc[], Help_monitor[], Help_config[];
 279
 280 /* for option that don't have short equivilents, we assign arbitrary
 281  * numbers later than any 'short' character option.
 282  */
 283 enum special_options {
 284         AssumeClean = 300,
 285         BitmapChunk,
 286         WriteBehind,
 287         ReAdd,
 288         NoDegraded,
 289         Sparc22,
 290         BackupFile,
 291         HomeHost,
 292         AutoHomeHost,
 293         Symlinks,
 294         AutoDetect,
 295         Waitclean,
 296         DetailPlatform,
 297         KillSubarray,
 298         UpdateSubarray,
 299         IncrementalPath,
 300         NoSharing,
 301         HelpOptions,
 302         Brief,
 303         ManageOpt,
 304         Add,
 305         Remove,
 306         Fail,
 307         MiscOpt,
 308         WaitOpt,
 309         ConfigFile,
 310         ChunkSize,
 311         WriteMostly,
 312         Layout,
 313         Auto,
 314         Force,
 315         SuperMinor,
 316         EMail,
 317         ProgramOpt,
 318         Increment,
 319         Fork,
 320         Bitmap,
 321         RebuildMapOpt,
 322         InvalidBackup,
 323         UdevRules,
 324         FreezeReshape,
 325         Continue,
 326         OffRootOpt,
 327         Prefer,
 328         KillOpt,
 329 };
 330
 331 /* structures read from config file */
 332 /* List of mddevice names and identifiers
 333  * Identifiers can be:
 334  *    uuid=128-hex-uuid
 335  *    super-minor=decimal-minor-number-from-superblock
 336  *    devices=comma,separated,list,of,device,names,with,wildcards
 337  *
 338  * If multiple fields are present, the intersection of all matching
 339  * devices is considered
 340  */
 341 #define UnSet (0xfffe)
 342 struct mddev_ident {
 343         char    *devname;
 344
 345         int     uuid_set;
 346         int     uuid[4];
 347         char    name[33];
 348
 349         int super_minor;
 350
 351         char    *devices;       /* comma separated list of device
 352                                  * names with wild cards
 353                                  */
 354         int     level;
 355         int raid_disks;
 356         int spare_disks;
 357         struct supertype *st;
 358         int     autof;          /* 1 for normal, 2 for partitioned */
 359         char    *spare_group;
 360         char    *bitmap_file;
 361         int     bitmap_fd;
 362
 363         char    *container;     /* /dev/whatever name of container, or
 364                                  * uuid of container.  You would expect
 365                                  * this to be the 'devname' or UUID
 366                                  * of some other entry.
 367                                  */
 368         char    *member;        /* subarray within a container */
 369
 370         struct mddev_ident *next;
 371         union {
 372                 /* fields needed by different users of this structure */
 373                 int assembled;  /* set when assembly succeeds */
 374         };
 375 };
 376
 377 struct context {
 378         int     readonly;
 379         int     runstop;
 380         int     verbose;
 381         int     brief;
 382         int     force;
 383         char    *homehost;
 384         int     require_homehost;
 385         char    *prefer;
 386         int     export;
 387         int     test;
 388         char    *subarray;
 389         char    *update;
 390         int     scan;
 391         int     SparcAdjust;
 392         int     autof;
 393         int     delay;
 394         int     freeze_reshape;
 395         char    *backup_file;
 396         int     invalid_backup;
 397 };
 398
 399 /* List of device names - wildcards expanded */
 400 struct mddev_dev {
 401         char *devname;
 402         int disposition;        /* 'a' for add, 'r' for remove, 'f' for fail,
 403                                  * 'A' for re_add.
 404                                  * Not set for names read from .config
 405                                  */
 406         char writemostly;       /* 1 for 'set writemostly', 2 for 'clear writemostly' */
 407         char used;              /* set when used */
 408         struct mddev_dev *next;
 409 };
 410
 411 typedef struct mapping {
 412         char *name;
 413         int num;
 414 } mapping_t;
 415
 416
 417 struct mdstat_ent {
 418         char            *dev;
 419         int             devnum;
 420         int             active;
 421         char            *level;
 422         char            *pattern; /* U or up, _ for down */
 423         int             percent; /* -1 if no resync */
 424         int             resync; /* 3 if check, 2 if reshape, 1 if resync, 0 if recovery */
 425         int             devcnt;
 426         int             raid_disks;
 427         char *          metadata_version;
 428         struct dev_member {
 429                 char                    *name;
 430                 struct dev_member       *next;
 431         }               *members;
 432         struct mdstat_ent *next;
 433 };
 434
 435 extern struct mdstat_ent *mdstat_read(int hold, int start);
 436 extern void free_mdstat(struct mdstat_ent *ms);
 437 extern void mdstat_wait(int seconds);
 438 extern void mdstat_wait_fd(int fd, const sigset_t *sigmask);
 439 extern int mddev_busy(int devnum);
 440 extern struct mdstat_ent *mdstat_by_component(char *name);
 441 extern struct mdstat_ent *mdstat_by_subdev(char *subdev, int container);
 442
 443 struct map_ent {
 444         struct map_ent *next;
 445         int     devnum;
 446         char    metadata[20];
 447         int     uuid[4];
 448         int     bad;
 449         char    *path;
 450 };
 451 extern int map_update(struct map_ent **mpp, int devnum, char *metadata,
 452                       int uuid[4], char *path);
 453 extern void map_remove(struct map_ent **map, int devnum);
 454 extern struct map_ent *map_by_uuid(struct map_ent **map, int uuid[4]);
 455 extern struct map_ent *map_by_devnum(struct map_ent **map, int devnum);
 456 extern struct map_ent *map_by_name(struct map_ent **map, char *name);
 457 extern void map_read(struct map_ent **melp);
 458 extern int map_write(struct map_ent *mel);
 459 extern void map_delete(struct map_ent **mapp, int devnum);
 460 extern void map_free(struct map_ent *map);
 461 extern void map_add(struct map_ent **melp,
 462                     int devnum, char *metadata, int uuid[4], char *path);
 463 extern int map_lock(struct map_ent **melp);
 464 extern void map_unlock(struct map_ent **melp);
 465 extern void map_fork(void);
 466
 467 /* various details can be requested */
 468 enum sysfs_read_flags {
 469         GET_LEVEL       = (1 << 0),
 470         GET_LAYOUT      = (1 << 1),
 471         GET_COMPONENT   = (1 << 2),
 472         GET_CHUNK       = (1 << 3),
 473         GET_CACHE       = (1 << 4),
 474         GET_MISMATCH    = (1 << 5),
 475         GET_VERSION     = (1 << 6),
 476         GET_DISKS       = (1 << 7),
 477         GET_DEGRADED    = (1 << 8),
 478         GET_SAFEMODE    = (1 << 9),
 479         GET_BITMAP_LOCATION = (1 << 10),
 480
 481         GET_DEVS        = (1 << 20), /* gets role, major, minor */
 482         GET_OFFSET      = (1 << 21),
 483         GET_SIZE        = (1 << 22),
 484         GET_STATE       = (1 << 23),
 485         GET_ERROR       = (1 << 24),
 486 };
 487
 488 /* If fd >= 0, get the array it is open on,
 489  * else use devnum. >=0 -> major9. <0.....
 490  */
 491 extern int sysfs_open(int devnum, char *devname, char *attr);
 492 extern void sysfs_init(struct mdinfo *mdi, int fd, int devnum);
 493 extern void sysfs_free(struct mdinfo *sra);
 494 extern struct mdinfo *sysfs_read(int fd, int devnum, unsigned long options);
 495 extern int sysfs_attr_match(const char *attr, const char *str);
 496 extern int sysfs_match_word(const char *word, char **list);
 497 extern int sysfs_set_str(struct mdinfo *sra, struct mdinfo *dev,
 498                          char *name, char *val);
 499 extern int sysfs_set_num(struct mdinfo *sra, struct mdinfo *dev,
 500                          char *name, unsigned long long val);
 501 extern int sysfs_set_num_signed(struct mdinfo *sra, struct mdinfo *dev,
 502                                 char *name, long long val);
 503 extern int sysfs_uevent(struct mdinfo *sra, char *event);
 504 extern int sysfs_get_fd(struct mdinfo *sra, struct mdinfo *dev,
 505                         char *name);
 506 extern int sysfs_fd_get_ll(int fd, unsigned long long *val);
 507 extern int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev,
 508                         char *name, unsigned long long *val);
 509 extern int sysfs_fd_get_str(int fd, char *val, int size);
 510 extern int sysfs_attribute_available(struct mdinfo *sra, struct mdinfo *dev,
 511                                      char *name);
 512 extern int sysfs_get_str(struct mdinfo *sra, struct mdinfo *dev,
 513                          char *name, char *val, int size);
 514 extern int sysfs_set_safemode(struct mdinfo *sra, unsigned long ms);
 515 extern int sysfs_set_array(struct mdinfo *info, int vers);
 516 extern int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd, int resume);
 517 extern int sysfs_disk_to_scsi_id(int fd, __u32 *id);
 518 extern int sysfs_unique_holder(int devnum, long rdev);
 519 extern int sysfs_freeze_array(struct mdinfo *sra);
 520 extern int load_sys(char *path, char *buf);
 521 extern int reshape_prepare_fdlist(char *devname,
 522                                   struct mdinfo *sra,
 523                                   int raid_disks,
 524                                   int nrdisks,
 525                                   unsigned long blocks,
 526                                   char *backup_file,
 527                                   int *fdlist,
 528                                   unsigned long long *offsets);
 529 extern void reshape_free_fdlist(int *fdlist,
 530                                 unsigned long long *offsets,
 531                                 int size);
 532 extern int reshape_open_backup_file(char *backup,
 533                                     int fd,
 534                                     char *devname,
 535                                     long blocks,
 536                                     int *fdlist,
 537                                     unsigned long long *offsets,
 538                                     int restart);
 539 extern unsigned long compute_backup_blocks(int nchunk, int ochunk,
 540                                            unsigned int ndata, unsigned int odata);
 541
 542 extern int save_stripes(int *source, unsigned long long *offsets,
 543                         int raid_disks, int chunk_size, int level, int layout,
 544                         int nwrites, int *dest,
 545                         unsigned long long start, unsigned long long length,
 546                         char *buf);
 547 extern int restore_stripes(int *dest, unsigned long long *offsets,
 548                            int raid_disks, int chunk_size, int level, int layout,
 549                            int source, unsigned long long read_offset,
 550                            unsigned long long start, unsigned long long length,
 551                            char *src_buf);
 552
 553 #ifndef Sendmail
 554 #define Sendmail "/usr/lib/sendmail -t"
 555 #endif
 556
 557 #define SYSLOG_FACILITY LOG_DAEMON
 558
 559 extern char *map_num(mapping_t *map, int num);
 560 extern int map_name(mapping_t *map, char *name);
 561 extern mapping_t r5layout[], r6layout[], pers[], modes[], faultylayout[];
 562
 563 extern char *map_dev_preferred(int major, int minor, int create,
 564                                char *prefer);
 565 static inline char *map_dev(int major, int minor, int create)
 566 {
 567         return map_dev_preferred(major, minor, create, NULL);
 568 }
 569
 570 struct active_array;
 571 struct metadata_update;
 572
 573
 574 /* 'struct reshape' records the intermediate states
 575  * a general reshape.
 576  * The starting geometry is converted to the 'before' geometry
 577  * by at most an atomic level change. They could be the same.
 578  * Similarly the 'after' geometry is converted to the final
 579  * geometry by at most a level change.
 580  * Note that 'before' and 'after' must have the same level.
 581  * 'blocks' is the minimum number of sectors for a reshape unit.
 582  * This will be a multiple of the stripe size in each of the
 583  * 'before' and 'after' geometries.
 584  * If 'blocks' is 0, no restriping is necessary.
 585  */
 586 struct reshape {
 587         int level;
 588         int parity; /* number of parity blocks/devices */
 589         struct {
 590                 int layout;
 591                 int data_disks;
 592         } before, after;
 593         unsigned long long backup_blocks;
 594         unsigned long long stripes; /* number of old stripes that comprise 'blocks'*/
 595         unsigned long long new_size; /* New size of array in sectors */
 596 };
 597
 598 /* A superswitch provides entry point the a metadata handler.
 599  *
 600  * The super_switch primarily operates on some "metadata" that
 601  * is accessed via the 'supertype'.
 602  * This metadata has one of three possible sources.
 603  * 1/ It is read from a single device.  In this case it may not completely
 604  *    describe the array or arrays as some information might be on other
 605  *    devices.
 606  * 2/ It is read from all devices in a container.  In this case all
 607  *    information is present.
 608  * 3/ It is created by ->init_super / ->add_to_super.  In this case it will
 609  *    be complete once enough ->add_to_super calls have completed.
 610  *
 611  * When creating an array inside a container, the metadata will be
 612  * formed by a combination of 2 and 3.  The metadata or the array is read,
 613  * then new information is added.
 614  *
 615  * The metadata must sometimes have a concept of a 'current' array
 616  * and a 'current' device.
 617  * The 'current' array is set by init_super to be the newly created array,
 618  * or is set by super_by_fd when it finds it is looking at an array inside
 619  * a container.
 620  *
 621  * The 'current' device is either the device that the metadata was read from
 622  * in case 1, or the last device added by add_to_super in case 3.
 623  * Case 2 does not identify a 'current' device.
 624  */
 625 extern struct superswitch {
 626
 627         /* Used to report details of metadata read from a component
 628          * device. ->load_super has been called.
 629          */
 630         void (*examine_super)(struct supertype *st, char *homehost);
 631         void (*brief_examine_super)(struct supertype *st, int verbose);
 632         void (*brief_examine_subarrays)(struct supertype *st, int verbose);
 633         void (*export_examine_super)(struct supertype *st);
 634
 635         /* Used to report details of an active array.
 636          * ->load_super was possibly given a 'component' string.
 637          */
 638         void (*detail_super)(struct supertype *st, char *homehost);
 639         void (*brief_detail_super)(struct supertype *st);
 640         void (*export_detail_super)(struct supertype *st);
 641
 642         /* Optional: platform hardware / firmware details */
 643         int (*detail_platform)(int verbose, int enumerate_only);
 644
 645         /* Used:
 646          *   to get uuid to storing in bitmap metadata
 647          *   and 'reshape' backup-data metadata
 648          *   To see if a device is being re-added to an array it was part of.
 649          */
 650         void (*uuid_from_super)(struct supertype *st, int uuid[4]);
 651
 652         /* Extract generic details from metadata.  This could be details about
 653          * the container, or about an individual array within the container.
 654          * The determination is made either by:
 655          *   load_super being given a 'component' string.
 656          *   validate_geometry determining what to create.
 657          * The info includes both array information and device information.
 658          * The particular device should be:
 659          *   The last device added by add_to_super
 660          *   The device the metadata was loaded from by load_super
 661          * If 'map' is present, then it is an array raid_disks long
 662          * (raid_disk must already be set and correct) and it is filled
 663          * with 1 for slots that are thought to be active and 0 for slots which
 664          * appear to be failed/missing.
 665          * *info is zeroed out before data is added.
 666          */
 667         void (*getinfo_super)(struct supertype *st, struct mdinfo *info, char *map);
 668         struct mdinfo *(*getinfo_super_disks)(struct supertype *st);
 669         /* Check if the given metadata is flagged as belonging to "this"
 670          * host.  0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
 671          */
 672         int (*match_home)(struct supertype *st, char *homehost);
 673
 674         /* Make one of several generic modifications to metadata
 675          * prior to assembly (or other times).
 676          *   sparc2.2  - first bug in early 0.90 metadata
 677          *   super-minor - change name of 0.90 metadata
 678          *   summaries - 'correct' any redundant data
 679          *   resync - mark array as dirty to trigger a resync.
 680          *   uuid - set new uuid - only 0.90 or 1.x
 681          *   name - change the name of the array (where supported)
 682          *   homehost - change which host this array is tied to.
 683          *   devicesize - If metadata is at start of device, change recorded
 684          *               device size to match actual device size
 685          *   byteorder - swap bytes for 0.90 metadata
 686          *
 687          *   force-one  - mark that device as uptodate, not old or failed.
 688          *   force-array - mark array as clean if it would not otherwise
 689          *               assemble
 690          *   assemble   - not sure how this is different from force-one...
 691          *   linear-grow-new - add a new device to a linear array, but don't
 692          *                   change the size: so superblock still matches
 693          *   linear-grow-update - now change the size of the array.
 694          *   writemostly - set the WriteMostly1 bit in the superblock devflags
 695          *   readwrite - clear the WriteMostly1 bit in the superblock devflags
 696          */
 697         int (*update_super)(struct supertype *st, struct mdinfo *info,
 698                             char *update,
 699                             char *devname, int verbose,
 700                             int uuid_set, char *homehost);
 701
 702         /* Create new metadata for new array as described.  This could
 703          * be a new container, or an array in a pre-existing container.
 704          * Also used to zero metadata prior to writing it to invalidate old
 705          * metadata.
 706          */
 707         int (*init_super)(struct supertype *st, mdu_array_info_t *info,
 708                           unsigned long long size, char *name,
 709                           char *homehost, int *uuid);
 710
 711         /* update the metadata to include new device, either at create or
 712          * when hot-adding a spare.
 713          */
 714         int (*add_to_super)(struct supertype *st, mdu_disk_info_t *dinfo,
 715                             int fd, char *devname);
 716         /* update the metadata to delete a device,
 717          * when hot-removing.
 718          */
 719         int (*remove_from_super)(struct supertype *st, mdu_disk_info_t *dinfo);
 720
 721         /* Write metadata to one device when fixing problems or adding
 722          * a new device.
 723          */
 724         int (*store_super)(struct supertype *st, int fd);
 725
 726         /*  Write all metadata for this array.
 727          */
 728         int (*write_init_super)(struct supertype *st);
 729         int (*compare_super)(struct supertype *st, struct supertype *tst);
 730         int (*load_super)(struct supertype *st, int fd, char *devname);
 731         int (*load_container)(struct supertype *st, int fd, char *devname);
 732         struct supertype * (*match_metadata_desc)(char *arg);
 733         __u64 (*avail_size)(struct supertype *st, __u64 size);
 734         unsigned long long (*min_acceptable_spare_size)(struct supertype *st);
 735         int (*add_internal_bitmap)(struct supertype *st, int *chunkp,
 736                                    int delay, int write_behind,
 737                                    unsigned long long size, int may_change, int major);
 738         void (*locate_bitmap)(struct supertype *st, int fd);
 739         int (*write_bitmap)(struct supertype *st, int fd);
 740         void (*free_super)(struct supertype *st);
 741
 742         /* validate_geometry is called with an st returned by
 743          * match_metadata_desc.
 744          * It should check that the geometry described in compatible with
 745          * the metadata type.  It will be called repeatedly as devices
 746          * added to validate changing size and new devices.  If there are
 747          * inter-device dependencies, it should record sufficient details
 748          * so these can be validated.
 749          * Both 'size' and '*freesize' are in sectors.  chunk is KiB.
 750          * Return value is:
 751          *  1: everything is OK
 752          *  0: not OK for some reason - if 'verbose', then error was reported.
 753          * -1: st->sb was NULL, 'subdev' is a member of a container of this
 754          *     types, but array is not acceptable for some reason
 755          *     message was reported even if verbose is 0.
 756          */
 757         int (*validate_geometry)(struct supertype *st, int level, int layout,
 758                                  int raiddisks,
 759                                  int *chunk, unsigned long long size,
 760                                  char *subdev, unsigned long long *freesize,
 761                                  int verbose);
 762
 763         struct mdinfo *(*container_content)(struct supertype *st, char *subarray);
 764         /* query the supertype for default geometry */
 765         void (*default_geometry)(struct supertype *st, int *level, int *layout, int *chunk); /* optional */
 766         /* Permit subarray's to be deleted from inactive containers */
 767         int (*kill_subarray)(struct supertype *st); /* optional */
 768         /* Permit subarray's to be modified */
 769         int (*update_subarray)(struct supertype *st, char *subarray,
 770                                char *update, struct mddev_ident *ident); /* optional */
 771         /* Check if reshape is supported for this external format.
 772          * st is obtained from super_by_fd() where st->subarray[0] is
 773          * initialized to indicate if reshape is being performed at the
 774          * container or subarray level
 775          */
 776 #define APPLY_METADATA_CHANGES          1
 777 #define ROLLBACK_METADATA_CHANGES       0
 778
 779         int (*reshape_super)(struct supertype *st,
 780                              unsigned long long size, int level,
 781                              int layout, int chunksize, int raid_disks,
 782                              int delta_disks, char *backup, char *dev,
 783                              int direction,
 784                              int verbose); /* optional */
 785         int (*manage_reshape)( /* optional */
 786                 int afd, struct mdinfo *sra, struct reshape *reshape,
 787                 struct supertype *st, unsigned long blocks,
 788                 int *fds, unsigned long long *offsets,
 789                 int dests, int *destfd, unsigned long long *destoffsets);
 790
 791 /* for mdmon */
 792         int (*open_new)(struct supertype *c, struct active_array *a,
 793                         char *inst);
 794
 795         /* Tell the metadata handler the current state of the array.
 796          * This covers whether it is known to be consistent (no pending writes)
 797          * and how far along a resync is known to have progressed
 798          * (in a->resync_start).
 799          * resync status is really irrelevant if the array is not consistent,
 800          * but some metadata (DDF!) have a place to record the distinction.
 801          * If 'consistent' is '2', then the array can mark it dirty if a
 802          * resync/recovery/whatever is required, or leave it clean if not.
 803          * Return value is 0 dirty (not consistent) and 1 if clean.
 804          * it is only really important if consistent is passed in as '2'.
 805          */
 806         int (*set_array_state)(struct active_array *a, int consistent);
 807
 808         /* When the state of a device might have changed, we call set_disk to
 809          * tell the metadata what the current state is.
 810          * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
 811          * transitions.
 812          * set_disk might be called when the state of the particular disk has
 813          * not in fact changed.
 814          */
 815         void (*set_disk)(struct active_array *a, int n, int state);
 816         void (*sync_metadata)(struct supertype *st);
 817         void (*process_update)(struct supertype *st,
 818                                struct metadata_update *update);
 819         void (*prepare_update)(struct supertype *st,
 820                                struct metadata_update *update);
 821
 822         /* activate_spare will check if the array is degraded and, if it
 823          * is, try to find some spare space in the container.
 824          * On success, it add appropriate updates (For process_update) to
 825          * to the 'updates' list and returns a list of 'mdinfo' identifying
 826          * the device, or devices as there might be multiple missing
 827          * devices and multiple spares available.
 828          */
 829         struct mdinfo *(*activate_spare)(struct active_array *a,
 830                                          struct metadata_update **updates);
 831         /*
 832          * Return statically allocated string that represents metadata specific
 833          * controller domain of the disk. The domain is used in disk domain
 834          * matching functions. Disks belong to the same domain if the they have
 835          * the same domain from mdadm.conf and belong the same metadata domain.
 836          * Returning NULL or not providing this handler means that metadata
 837          * does not distinguish the differences between disks that belong to
 838          * different controllers. They are in the domain specified by
 839          * configuration file (mdadm.conf).
 840          * In case when the metadata has the notion of domains based on disk
 841          * it shall return NULL for disks that do not belong to the controller
 842          * the supported domains. Such disks will form another domain and won't
 843          * be mixed with supported ones.
 844          */
 845         const char *(*get_disk_controller_domain)(const char *path);
 846
 847         /* for external backup area */
 848         int (*recover_backup)(struct supertype *st, struct mdinfo *info);
 849
 850         int swapuuid; /* true if uuid is bigending rather than hostendian */
 851         int external;
 852         const char *name; /* canonical metadata name */
 853 } *superlist[];
 854
 855 extern struct superswitch super0, super1;
 856 extern struct superswitch super_imsm, super_ddf;
 857 extern struct superswitch mbr, gpt;
 858
 859 struct metadata_update {
 860         int     len;
 861         char    *buf;
 862         void    *space; /* allocated space that monitor will use */
 863         void    **space_list; /* list of allocated spaces that monitor can
 864                                * use or that it returned.
 865                                */
 866         struct metadata_update *next;
 867 };
 868
 869 /* A supertype holds a particular collection of metadata.
 870  * It identifies the metadata type by the superswitch, and the particular
 871  * sub-version of that metadata type.
 872  * metadata read in or created is stored in 'sb' and 'info'.
 873  * There are also fields used by mdmon to track containers.
 874  *
 875  * A supertype may refer to:
 876  *   Just an array, possibly in a container
 877  *   A container, not identifying any particular array
 878  *   Info read from just one device, not yet fully describing the array/container.
 879  *
 880  *
 881  * A supertype is created by:
 882  *   super_by_fd
 883  *   guess_super
 884  *   dup_super
 885  */
 886 struct supertype {
 887         struct superswitch *ss;
 888         int minor_version;
 889         int max_devs;
 890         int container_dev;    /* devnum of container */
 891         void *sb;
 892         void *info;
 893         int ignore_hw_compat; /* used to inform metadata handlers that it should ignore
 894                                  HW/firmware related incompatability to load metadata.
 895                                  Used when examining metadata to display content of disk
 896                                  when user has no hw/firmare compatible system.
 897                               */
 898         struct metadata_update *updates;
 899         struct metadata_update **update_tail;
 900
 901         /* extra stuff used by mdmon */
 902         struct active_array *arrays;
 903         int sock; /* listen to external programs */
 904         int devnum;
 905         char *devname; /* e.g. md0.  This appears in metadata_verison:
 906                         *  external:/md0/12
 907                         */
 908         int devcnt;
 909         int retry_soon;
 910
 911         struct mdinfo *devs;
 912
 913 };
 914
 915 extern struct supertype *super_by_fd(int fd, char **subarray);
 916 enum guess_types { guess_any, guess_array, guess_partitions };
 917 extern struct supertype *guess_super_type(int fd, enum guess_types guess_type);
 918 static inline struct supertype *guess_super(int fd) {
 919         return guess_super_type(fd, guess_any);
 920 }
 921 extern struct supertype *dup_super(struct supertype *st);
 922 extern int get_dev_size(int fd, char *dname, unsigned long long *sizep);
 923 extern int must_be_container(int fd);
 924 extern int dev_size_from_id(dev_t id, unsigned long long *size);
 925 extern void get_one_disk(int mdfd, mdu_array_info_t *ainf,
 926                          mdu_disk_info_t *disk);
 927 void wait_for(char *dev, int fd);
 928
 929 /*
 930  * Data structures for policy management.
 931  * Each device can have a policy structure that lists
 932  * various name/value pairs each possibly with a metadata associated.
 933  * The policy list is sorted by name/value/metadata
 934  */
 935 struct dev_policy {
 936         struct dev_policy *next;
 937         char *name;     /* None of these strings are allocated.  They are
 938                          * all just references to strings which are known
 939                          * to exist elsewhere.
 940                          * name and metadata can be compared by address equality.
 941                          */
 942         const char *metadata;
 943         const char *value;
 944 };
 945
 946 extern char pol_act[], pol_domain[], pol_metadata[], pol_auto[];
 947
 948 /* iterate over the sublist starting at list, having the same
 949  * 'name' as 'list', and matching the given metadata (Where
 950  * NULL matches anything
 951  */
 952 #define pol_for_each(item, list, _metadata)                             \
 953         for (item = list;                                               \
 954              item && item->name == list->name;                          \
 955              item = item->next)                                         \
 956                 if (!(!_metadata || !item->metadata || _metadata == item->metadata)) \
 957                         ; else
 958
 959 /*
 960  * policy records read from mdadm are largely just name-value pairs.
 961  * The names are constants, not strdupped
 962  */
 963 struct pol_rule {
 964         struct pol_rule *next;
 965         char *type;     /* rule_policy or rule_part */
 966         struct rule {
 967                 struct rule *next;
 968                 char *name;
 969                 char *value;
 970                 char *dups; /* duplicates of 'value' with a partNN appended */
 971         } *rule;
 972 };
 973
 974 extern char rule_policy[], rule_part[];
 975 extern char rule_path[], rule_type[];
 976 extern char type_part[], type_disk[];
 977
 978 extern void policyline(char *line, char *type);
 979 extern void policy_add(char *type, ...);
 980 extern void policy_free(void);
 981
 982 extern struct dev_policy *path_policy(char *path, char *type);
 983 extern struct dev_policy *disk_policy(struct mdinfo *disk);
 984 extern struct dev_policy *devnum_policy(int dev);
 985 extern void dev_policy_free(struct dev_policy *p);
 986
 987 //extern void pol_new(struct dev_policy **pol, char *name, char *val, char *metadata);
 988 extern void pol_add(struct dev_policy **pol, char *name, char *val, char *metadata);
 989 extern struct dev_policy *pol_find(struct dev_policy *pol, char *name);
 990
 991 enum policy_action {
 992         act_default,
 993         act_include,
 994         act_re_add,
 995         act_spare,      /* This only applies to bare devices */
 996         act_spare_same_slot, /* this allows non-bare devices,
 997                               * but only if recent removal */
 998         act_force_spare, /* this allow non-bare devices in any case */
 999         act_err
1000 };
1001
1002 extern int policy_action_allows(struct dev_policy *plist, const char *metadata,
1003                                 enum policy_action want);
1004 extern int disk_action_allows(struct mdinfo *disk, const char *metadata,
1005                               enum policy_action want);
1006
1007 struct domainlist {
1008         struct domainlist *next;
1009         const char *dom;
1010 };
1011
1012 extern int domain_test(struct domainlist *dom, struct dev_policy *pol,
1013                        const char *metadata);
1014 extern struct domainlist *domain_from_array(struct mdinfo *mdi,
1015                                             const char *metadata);
1016 extern void domainlist_add_dev(struct domainlist **dom, int devnum,
1017                                const char *metadata);
1018 extern void domain_free(struct domainlist *dl);
1019 extern void domain_merge(struct domainlist **domp, struct dev_policy *pol,
1020                          const char *metadata);
1021 void domain_add(struct domainlist **domp, char *domain);
1022
1023 extern void policy_save_path(char *id_path, struct map_ent *array);
1024 extern int policy_check_path(struct mdinfo *disk, struct map_ent *array);
1025
1026 #if __GNUC__ < 3
1027 struct stat64;
1028 #endif
1029
1030 #define HAVE_NFTW  we assume
1031 #define HAVE_FTW
1032
1033 #ifdef __UCLIBC__
1034 # include <features.h>
1035 # ifndef __UCLIBC_HAS_LFS__
1036 #  define lseek64 lseek
1037 # endif
1038 # ifndef  __UCLIBC_HAS_FTW__
1039 #  undef HAVE_FTW
1040 #  undef HAVE_NFTW
1041 # endif
1042 #endif
1043
1044 #ifdef __dietlibc__
1045 # undef HAVE_NFTW
1046 #endif
1047
1048 #if defined(__KLIBC__)
1049 # undef HAVE_NFTW
1050 # undef HAVE_FTW
1051 #endif
1052
1053 #ifndef HAVE_NFTW
1054 # define FTW_PHYS 1
1055 # ifndef HAVE_FTW
1056   struct FTW {};
1057 # endif
1058 #endif
1059
1060 #ifdef HAVE_FTW
1061 # include <ftw.h>
1062 #endif
1063
1064 extern int add_dev(const char *name, const struct stat *stb, int flag, struct FTW *s);
1065
1066
1067 extern int Manage_ro(char *devname, int fd, int readonly);
1068 extern int Manage_runstop(char *devname, int fd, int runstop, int quiet,
1069                           int will_retry);
1070 extern int Manage_resize(char *devname, int fd, long long size, int raid_disks);
1071 extern int Manage_subdevs(char *devname, int fd,
1072                           struct mddev_dev *devlist, int verbose, int test,
1073                           char *update, int force);
1074 extern int autodetect(void);
1075 extern int Grow_Add_device(char *devname, int fd, char *newdev);
1076 extern int Grow_addbitmap(char *devname, int fd, char *file, int chunk, int delay, int write_behind, int force);
1077 extern int Grow_reshape(char *devname, int fd, int quiet, char *backup_file,
1078                         unsigned long long size,
1079                         int level, char *layout_str, int chunksize, int raid_disks,
1080                         struct mddev_dev *devlist,
1081                         int assume_clean, int force);
1082 extern int Grow_restart(struct supertype *st, struct mdinfo *info,
1083                         int *fdlist, int cnt, char *backup_file, int verbose);
1084 extern int Grow_continue(int mdfd, struct supertype *st,
1085                          struct mdinfo *info, char *backup_file,
1086                          int freeze_reshape);
1087
1088 extern int restore_backup(struct supertype *st,
1089                           struct mdinfo *content,
1090                           int working_disks,
1091                           int spares,
1092                           char *backup_file,
1093                           int verbose);
1094 extern int Grow_continue_command(char *devname, int fd,
1095                                  char *backup_file, int verbose);
1096
1097 extern int Assemble(struct supertype *st, char *mddev,
1098                     struct mddev_ident *ident,
1099                     struct mddev_dev *devlist,
1100                     struct context *c);
1101
1102 extern int Build(char *mddev, int chunk, int level, int layout,
1103                  int raiddisks, struct mddev_dev *devlist, int assume_clean,
1104                  char *bitmap_file, int bitmap_chunk, int write_behind,
1105                  struct context *c, unsigned long long size);
1106
1107
1108 extern int Create(struct supertype *st, char *mddev,
1109                   int chunk, int level, int layout, unsigned long long size,
1110                   int raiddisks, int sparedisks,
1111                   char *name, int *uuid,
1112                   int subdevs, struct mddev_dev *devlist,
1113                   int assume_clean,
1114                   char *bitmap_file, int bitmap_chunk, int write_behind,
1115                   struct context *c);
1116
1117 extern int Detail(char *dev, struct context *c);
1118 extern int Detail_Platform(struct superswitch *ss, int scan, int verbose);
1119 extern int Query(char *dev);
1120 extern int Examine(struct mddev_dev *devlist, struct context *c,
1121                    struct supertype *forcest);
1122 extern int Monitor(struct mddev_dev *devlist,
1123                    char *mailaddr, char *alert_cmd,
1124                    struct context *c,
1125                    int daemonise, int oneshot,
1126                    int dosyslog, char *pidfile, int increments,
1127                    int share);
1128
1129 extern int Kill(char *dev, struct supertype *st, int force, int verbose, int noexcl);
1130 extern int Kill_subarray(char *dev, char *subarray, int verbose);
1131 extern int Update_subarray(char *dev, char *subarray, char *update, struct mddev_ident *ident, int quiet);
1132 extern int Wait(char *dev);
1133 extern int WaitClean(char *dev, int sock, int verbose);
1134
1135 extern int Incremental(char *devname, struct context *c,
1136                        struct supertype *st);
1137 extern void RebuildMap(void);
1138 extern int IncrementalScan(int verbose);
1139 extern int IncrementalRemove(char *devname, char *path, int verbose);
1140 extern int CreateBitmap(char *filename, int force, char uuid[16],
1141                         unsigned long chunksize, unsigned long daemon_sleep,
1142                         unsigned long write_behind,
1143                         unsigned long long array_size,
1144                         int major);
1145 extern int ExamineBitmap(char *filename, int brief, struct supertype *st);
1146 extern int Write_rules(char *rule_name);
1147 extern int bitmap_update_uuid(int fd, int *uuid, int swap);
1148 extern unsigned long bitmap_sectors(struct bitmap_super_s *bsb);
1149
1150 extern int md_get_version(int fd);
1151 extern int get_linux_version(void);
1152 extern int mdadm_version(char *version);
1153 extern long long parse_size(char *size);
1154 extern int parse_uuid(char *str, int uuid[4]);
1155 extern int parse_layout_10(char *layout);
1156 extern int parse_layout_faulty(char *layout);
1157 extern long parse_num(char *num);
1158 extern int check_ext2(int fd, char *name);
1159 extern int check_reiser(int fd, char *name);
1160 extern int check_raid(int fd, char *name);
1161 extern int check_partitions(int fd, char *dname,
1162                             unsigned long long freesize,
1163                             unsigned long long size);
1164
1165 extern int get_mdp_major(void);
1166 extern int dev_open(char *dev, int flags);
1167 extern int open_dev(int devnum);
1168 extern int open_dev_flags(int devnum, int flags);
1169 extern int open_dev_excl(int devnum);
1170 extern int is_standard(char *dev, int *nump);
1171 extern int same_dev(char *one, char *two);
1172
1173 extern int parse_auto(char *str, char *msg, int config);
1174 extern struct mddev_ident *conf_get_ident(char *dev);
1175 extern struct mddev_dev *conf_get_devs(void);
1176 extern int conf_test_dev(char *devname);
1177 extern int conf_test_metadata(const char *version, struct dev_policy *pol, int is_homehost);
1178 extern struct createinfo *conf_get_create_info(void);
1179 extern void set_conffile(char *file);
1180 extern char *conf_get_mailaddr(void);
1181 extern char *conf_get_mailfrom(void);
1182 extern char *conf_get_program(void);
1183 extern char *conf_get_homehost(int *require_homehostp);
1184 extern char *conf_line(FILE *file);
1185 extern char *conf_word(FILE *file, int allow_key);
1186 extern int conf_name_is_free(char *name);
1187 extern int conf_verify_devnames(struct mddev_ident *array_list);
1188 extern int devname_matches(char *name, char *match);
1189 extern struct mddev_ident *conf_match(struct supertype *st,
1190                                       struct mdinfo *info,
1191                                       char *devname,
1192                                       int verbose, int *rvp);
1193 extern int experimental(void);
1194
1195 extern void free_line(char *line);
1196 extern int match_oneof(char *devices, char *devname);
1197 extern void uuid_from_super(int uuid[4], mdp_super_t *super);
1198 extern const int uuid_zero[4];
1199 extern int same_uuid(int a[4], int b[4], int swapuuid);
1200 extern void copy_uuid(void *a, int b[4], int swapuuid);
1201 extern char *__fname_from_uuid(int id[4], int swap, char *buf, char sep);
1202 extern char *fname_from_uuid(struct supertype *st,
1203                              struct mdinfo *info, char *buf, char sep);
1204 extern unsigned long calc_csum(void *super, int bytes);
1205 extern int enough(int level, int raid_disks, int layout, int clean,
1206                    char *avail);
1207 extern int enough_fd(int fd);
1208 extern int ask(char *mesg);
1209 extern unsigned long long get_component_size(int fd);
1210 extern void remove_partitions(int fd);
1211 extern int test_partition(int fd);
1212 extern int test_partition_from_id(dev_t id);
1213 extern int get_data_disks(int level, int layout, int raid_disks);
1214 extern unsigned long long calc_array_size(int level, int raid_disks, int layout,
1215                                    int chunksize, unsigned long long devsize);
1216 extern int flush_metadata_updates(struct supertype *st);
1217 extern void append_metadata_update(struct supertype *st, void *buf, int len);
1218 extern int assemble_container_content(struct supertype *st, int mdfd,
1219                                       struct mdinfo *content,
1220                                       struct context *c,
1221                                       char *chosen_name);
1222 extern struct mdinfo *container_choose_spares(struct supertype *st,
1223                                               unsigned long long min_size,
1224                                               struct domainlist *domlist,
1225                                               char *spare_group,
1226                                               const char *metadata, int get_one);
1227 extern int move_spare(char *from_devname, char *to_devname, dev_t devid);
1228 extern int add_disk(int mdfd, struct supertype *st,
1229                     struct mdinfo *sra, struct mdinfo *info);
1230 extern int remove_disk(int mdfd, struct supertype *st,
1231                        struct mdinfo *sra, struct mdinfo *info);
1232 extern int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info);
1233 unsigned long long min_recovery_start(struct mdinfo *array);
1234
1235 extern char *human_size(long long bytes);
1236 extern char *human_size_brief(long long bytes);
1237 extern void print_r10_layout(int layout);
1238
1239 #define NoMdDev (1<<23)
1240 extern int find_free_devnum(int use_partitions);
1241
1242 extern void put_md_name(char *name);
1243 extern char *get_md_name(int dev);
1244
1245 extern char DefaultConfFile[];
1246
1247 extern int create_mddev(char *dev, char *name, int autof, int trustworthy,
1248                         char *chosen);
1249 /* values for 'trustworthy' */
1250 #define LOCAL   1
1251 #define LOCAL_ANY 10
1252 #define FOREIGN 2
1253 #define METADATA 3
1254 extern int open_mddev(char *dev, int report_errors);
1255 extern int open_container(int fd);
1256 extern int is_container_member(struct mdstat_ent *ent, char *devname);
1257 extern int is_subarray_active(char *subarray, char *devname);
1258 extern int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet);
1259 extern struct superswitch *version_to_superswitch(char *vers);
1260
1261 extern int mdmon_running(int devnum);
1262 extern int mdmon_pid(int devnum);
1263 extern int check_env(char *name);
1264 extern __u32 random32(void);
1265 extern int start_mdmon(int devnum);
1266
1267 extern int child_monitor(int afd, struct mdinfo *sra, struct reshape *reshape,
1268                          struct supertype *st, unsigned long stripes,
1269                          int *fds, unsigned long long *offsets,
1270                          int dests, int *destfd, unsigned long long *destoffsets);
1271 void abort_reshape(struct mdinfo *sra);
1272
1273 extern char *devnum2devname(int num);
1274 extern void fmt_devname(char *name, int num);
1275 extern int devname2devnum(char *name);
1276 extern int stat2devnum(struct stat *st);
1277 extern int fd2devnum(int fd);
1278
1279 static inline int dev2major(int d)
1280 {
1281         if (d >= 0)
1282                 return MD_MAJOR;
1283         else
1284                 return get_mdp_major();
1285 }
1286
1287 static inline int dev2minor(int d)
1288 {
1289         if (d >= 0)
1290                 return d;
1291         return (-1-d) << MdpMinorShift;
1292 }
1293
1294 #define _ROUND_UP(val, base)    (((val) + (base) - 1) & ~(base - 1))
1295 #define ROUND_UP(val, base)     _ROUND_UP(val, (typeof(val))(base))
1296 #define ROUND_UP_PTR(ptr, base) ((typeof(ptr)) \
1297                                  (ROUND_UP((unsigned long)(ptr), base)))
1298
1299 static inline int is_subarray(char *vers)
1300 {
1301         /* The version string for a 'subarray' (an array in a container)
1302          * is
1303          *    /containername/componentname    for normal read-write arrays
1304          *    -containername/componentname    for arrays which mdmon must not
1305          *                                    reconfigure.  They might be read-only
1306          *                                    or might be undergoing reshape etc.
1307          * containername is e.g. md0, md_d1
1308          * componentname is dependant on the metadata. e.g. '1' 'S1' ...
1309          */
1310         return (*vers == '/' || *vers == '-');
1311 }
1312
1313 static inline char *to_subarray(struct mdstat_ent *ent, char *container)
1314 {
1315         return &ent->metadata_version[10+strlen(container)+1];
1316 }
1317
1318 #ifdef DEBUG
1319 #define dprintf(fmt, arg...) \
1320         fprintf(stderr, fmt, ##arg)
1321 #else
1322 #define dprintf(fmt, arg...) \
1323         ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
1324 #endif
1325 #include <assert.h>
1326 #include <stdarg.h>
1327 static inline int xasprintf(char **strp, const char *fmt, ...) {
1328         va_list ap;
1329         int ret;
1330         va_start(ap, fmt);
1331         ret = vasprintf(strp, fmt, ap);
1332         va_end(ap);
1333         assert(ret >= 0);
1334         return ret;
1335 }
1336
1337 #define pr_err(fmt ...) fprintf(stderr, Name ": " fmt)
1338 #define cont_err(fmt ...) fprintf(stderr, "       " fmt)
1339
1340 void *xmalloc(size_t len);
1341 void *xrealloc(void *ptr, size_t len);
1342 void *xcalloc(size_t num, size_t size);
1343 char *xstrdup(const char *str);
1344
1345 #define LEVEL_MULTIPATH         (-4)
1346 #define LEVEL_LINEAR            (-1)
1347 #define LEVEL_FAULTY            (-5)
1348
1349 /* kernel module doesn't know about these */
1350 #define LEVEL_CONTAINER         (-100)
1351 #define LEVEL_UNSUPPORTED       (-200)
1352
1353
1354 /* faulty stuff */
1355
1356 #define WriteTransient  0
1357 #define ReadTransient   1
1358 #define WritePersistent 2
1359 #define ReadPersistent  3
1360 #define WriteAll        4 /* doesn't go to device */
1361 #define ReadFixable     5
1362 #define Modes   6
1363
1364 #define ClearErrors     31
1365 #define ClearFaults     30
1366
1367 #define AllPersist      100 /* internal use only */
1368 #define NoPersist       101
1369
1370 #define ModeMask        0x1f
1371 #define ModeShift       5
1372
1373
1374 #ifdef __TINYC__
1375 #undef minor
1376 #undef major
1377 #undef makedev
1378 #define minor(x) ((x)&0xff)
1379 #define major(x) (((x)>>8)&0xff)
1380 #define makedev(M,m) (((M)<<8) | (m))
1381 #endif
1382
1383 /* for raid4/5/6 */
1384 #define ALGORITHM_LEFT_ASYMMETRIC       0
1385 #define ALGORITHM_RIGHT_ASYMMETRIC      1
1386 #define ALGORITHM_LEFT_SYMMETRIC        2
1387 #define ALGORITHM_RIGHT_SYMMETRIC       3
1388
1389 /* Define non-rotating (raid4) algorithms.  These allow
1390  * conversion of raid4 to raid5.
1391  */
1392 #define ALGORITHM_PARITY_0              4 /* P or P,Q are initial devices */
1393 #define ALGORITHM_PARITY_N              5 /* P or P,Q are final devices. */
1394
1395 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1396  * Firstly, the exact positioning of the parity block is slightly
1397  * different between the 'LEFT_*' modes of md and the "_N_*" modes
1398  * of DDF.
1399  * Secondly, or order of datablocks over which the Q syndrome is computed
1400  * is different.
1401  * Consequently we have different layouts for DDF/raid6 than md/raid6.
1402  * These layouts are from the DDFv1.2 spec.
1403  * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1404  * leaves RLQ=3 as 'Vendor Specific'
1405  */
1406
1407 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1408 #define ALGORITHM_ROTATING_N_RESTART    9 /* DDF PRL=6 RLQ=2 */
1409 #define ALGORITHM_ROTATING_N_CONTINUE   10 /*DDF PRL=6 RLQ=3 */
1410
1411
1412 /* For every RAID5 algorithm we define a RAID6 algorithm
1413  * with exactly the same layout for data and parity, and
1414  * with the Q block always on the last device (N-1).
1415  * This allows trivial conversion from RAID5 to RAID6
1416  */
1417 #define ALGORITHM_LEFT_ASYMMETRIC_6     16
1418 #define ALGORITHM_RIGHT_ASYMMETRIC_6    17
1419 #define ALGORITHM_LEFT_SYMMETRIC_6      18
1420 #define ALGORITHM_RIGHT_SYMMETRIC_6     19
1421 #define ALGORITHM_PARITY_0_6            20
1422 #define ALGORITHM_PARITY_N_6            ALGORITHM_PARITY_N
1423
1424 /* Define PATH_MAX in case we don't use glibc or standard library does
1425  * not have PATH_MAX defined. Assume max path length is 4K characters.
1426  */
1427 #ifndef PATH_MAX
1428 #define PATH_MAX        4096
1429 #endif
1430
1431 #define RESYNC_NONE -1
1432 #define RESYNC_DELAYED -2
1433 #define RESYNC_PENDING -3
1434 #define RESYNC_UNKNOWN -4
1435
1436 /* When using "GET_DISK_INFO" it isn't certain how high
1437  * we need to check.  So we impose an absolute limit of
1438  * MAX_DISKS.  This needs to be much more than the largest
1439  * number of devices any metadata can support.  Currently
1440  * v1.x can support 1920
1441  */
1442 #define MAX_DISKS       4096
1443
1444 /* Sometimes the 'size' value passed needs to mean "Maximum".
1445  * In those cases with use MAX_SIZE
1446  */
1447 #define MAX_SIZE        1
1448 extern int __offroot;