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