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 #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         unsigned 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 };
 268
 269 extern char short_options[];
 270 extern char short_bitmap_options[];
 271 extern char short_bitmap_auto_options[];
 272 extern struct option long_options[];
 273 extern char Version[], Usage[], Help[], OptionHelp[],
 274         Help_create[], Help_build[], Help_assemble[], Help_grow[],
 275         Help_incr[],
 276         Help_manage[], Help_misc[], Help_monitor[], Help_config[];
 277
 278 /* for option that don't have short equivilents, we assign arbitrary
 279  * numbers later than any 'short' character option.
 280  */
 281 enum special_options {
 282         AssumeClean = 300,
 283         BitmapChunk,
 284         WriteBehind,
 285         ReAdd,
 286         NoDegraded,
 287         Sparc22,
 288         BackupFile,
 289         HomeHost,
 290         AutoHomeHost,
 291         Symlinks,
 292         AutoDetect,
 293         Waitclean,
 294         DetailPlatform,
 295         KillSubarray,
 296         UpdateSubarray,
 297         IncrementalPath,
 298         NoSharing,
 299         HelpOptions,
 300         Brief,
 301         ManageOpt,
 302         Add,
 303         Remove,
 304         Fail,
 305         MiscOpt,
 306         WaitOpt,
 307         ConfigFile,
 308         ChunkSize,
 309         WriteMostly,
 310         Layout,
 311         Auto,
 312         Force,
 313         SuperMinor,
 314         EMail,
 315         ProgramOpt,
 316         Increment,
 317         Fork,
 318         Bitmap,
 319         RebuildMapOpt,
 320         InvalidBackup,
 321         UdevRules,
 322         FreezeReshape,
 323         Continue,
 324 };
 325
 326 /* structures read from config file */
 327 /* List of mddevice names and identifiers
 328  * Identifiers can be:
 329  *    uuid=128-hex-uuid
 330  *    super-minor=decimal-minor-number-from-superblock
 331  *    devices=comma,separated,list,of,device,names,with,wildcards
 332  *
 333  * If multiple fields are present, the intersection of all matching
 334  * devices is considered
 335  */
 336 #define UnSet (0xfffe)
 337 struct mddev_ident {
 338         char    *devname;
 339
 340         int     uuid_set;
 341         int     uuid[4];
 342         char    name[33];
 343
 344         int super_minor;
 345
 346         char    *devices;       /* comma separated list of device
 347                                  * names with wild cards
 348                                  */
 349         int     level;
 350         int raid_disks;
 351         int spare_disks;
 352         struct supertype *st;
 353         int     autof;          /* 1 for normal, 2 for partitioned */
 354         char    *spare_group;
 355         char    *bitmap_file;
 356         int     bitmap_fd;
 357
 358         char    *container;     /* /dev/whatever name of container, or
 359                                  * uuid of container.  You would expect
 360                                  * this to be the 'devname' or UUID
 361                                  * of some other entry.
 362                                  */
 363         char    *member;        /* subarray within a container */
 364
 365         struct mddev_ident *next;
 366         union {
 367                 /* fields needed by different users of this structure */
 368                 int assembled;  /* set when assembly succeeds */
 369         };
 370 };
 371
 372 /* List of device names - wildcards expanded */
 373 struct mddev_dev {
 374         char *devname;
 375         int disposition;        /* 'a' for add, 'r' for remove, 'f' for fail.
 376                                  * Not set for names read from .config
 377                                  */
 378         char writemostly;       /* 1 for 'set writemostly', 2 for 'clear writemostly' */
 379         char re_add;
 380         char used;              /* set when used */
 381         struct mddev_dev *next;
 382 };
 383
 384 typedef struct mapping {
 385         char *name;
 386         int num;
 387 } mapping_t;
 388
 389
 390 struct mdstat_ent {
 391         char            *dev;
 392         int             devnum;
 393         int             active;
 394         char            *level;
 395         char            *pattern; /* U or up, _ for down */
 396         int             percent; /* -1 if no resync */
 397         int             resync; /* 3 if check, 2 if reshape, 1 if resync, 0 if recovery */
 398         int             devcnt;
 399         int             raid_disks;
 400         char *          metadata_version;
 401         struct dev_member {
 402                 char                    *name;
 403                 struct dev_member       *next;
 404         }               *members;
 405         struct mdstat_ent *next;
 406 };
 407
 408 extern struct mdstat_ent *mdstat_read(int hold, int start);
 409 extern void free_mdstat(struct mdstat_ent *ms);
 410 extern void mdstat_wait(int seconds);
 411 extern void mdstat_wait_fd(int fd, const sigset_t *sigmask);
 412 extern int mddev_busy(int devnum);
 413 extern struct mdstat_ent *mdstat_by_component(char *name);
 414 extern struct mdstat_ent *mdstat_by_subdev(char *subdev, int container);
 415
 416 struct map_ent {
 417         struct map_ent *next;
 418         int     devnum;
 419         char    metadata[20];
 420         int     uuid[4];
 421         int     bad;
 422         char    *path;
 423 };
 424 extern int map_update(struct map_ent **mpp, int devnum, char *metadata,
 425                       int uuid[4], char *path);
 426 extern void map_remove(struct map_ent **map, int devnum);
 427 extern struct map_ent *map_by_uuid(struct map_ent **map, int uuid[4]);
 428 extern struct map_ent *map_by_devnum(struct map_ent **map, int devnum);
 429 extern struct map_ent *map_by_name(struct map_ent **map, char *name);
 430 extern void map_read(struct map_ent **melp);
 431 extern int map_write(struct map_ent *mel);
 432 extern void map_delete(struct map_ent **mapp, int devnum);
 433 extern void map_free(struct map_ent *map);
 434 extern void map_add(struct map_ent **melp,
 435                     int devnum, char *metadata, int uuid[4], char *path);
 436 extern int map_lock(struct map_ent **melp);
 437 extern void map_unlock(struct map_ent **melp);
 438 extern void map_fork(void);
 439
 440 /* various details can be requested */
 441 enum sysfs_read_flags {
 442         GET_LEVEL       = (1 << 0),
 443         GET_LAYOUT      = (1 << 1),
 444         GET_COMPONENT   = (1 << 2),
 445         GET_CHUNK       = (1 << 3),
 446         GET_CACHE       = (1 << 4),
 447         GET_MISMATCH    = (1 << 5),
 448         GET_VERSION     = (1 << 6),
 449         GET_DISKS       = (1 << 7),
 450         GET_DEGRADED    = (1 << 8),
 451         GET_SAFEMODE    = (1 << 9),
 452         GET_BITMAP_LOCATION = (1 << 10),
 453
 454         GET_DEVS        = (1 << 20), /* gets role, major, minor */
 455         GET_OFFSET      = (1 << 21),
 456         GET_SIZE        = (1 << 22),
 457         GET_STATE       = (1 << 23),
 458         GET_ERROR       = (1 << 24),
 459 };
 460
 461 /* If fd >= 0, get the array it is open on,
 462  * else use devnum. >=0 -> major9. <0.....
 463  */
 464 extern int sysfs_open(int devnum, char *devname, char *attr);
 465 extern void sysfs_init(struct mdinfo *mdi, int fd, int devnum);
 466 extern void sysfs_free(struct mdinfo *sra);
 467 extern struct mdinfo *sysfs_read(int fd, int devnum, unsigned long options);
 468 extern int sysfs_attr_match(const char *attr, const char *str);
 469 extern int sysfs_match_word(const char *word, char **list);
 470 extern int sysfs_set_str(struct mdinfo *sra, struct mdinfo *dev,
 471                          char *name, char *val);
 472 extern int sysfs_set_num(struct mdinfo *sra, struct mdinfo *dev,
 473                          char *name, unsigned long long val);
 474 extern int sysfs_uevent(struct mdinfo *sra, char *event);
 475 extern int sysfs_get_fd(struct mdinfo *sra, struct mdinfo *dev,
 476                         char *name);
 477 extern int sysfs_fd_get_ll(int fd, unsigned long long *val);
 478 extern int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev,
 479                         char *name, unsigned long long *val);
 480 extern int sysfs_fd_get_str(int fd, char *val, int size);
 481 extern int sysfs_attribute_available(struct mdinfo *sra, struct mdinfo *dev,
 482                                      char *name);
 483 extern int sysfs_get_str(struct mdinfo *sra, struct mdinfo *dev,
 484                          char *name, char *val, int size);
 485 extern int sysfs_set_safemode(struct mdinfo *sra, unsigned long ms);
 486 extern int sysfs_set_array(struct mdinfo *info, int vers);
 487 extern int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd, int resume);
 488 extern int sysfs_disk_to_scsi_id(int fd, __u32 *id);
 489 extern int sysfs_unique_holder(int devnum, long rdev);
 490 extern int sysfs_freeze_array(struct mdinfo *sra);
 491 extern int load_sys(char *path, char *buf);
 492 extern int reshape_prepare_fdlist(char *devname,
 493                                   struct mdinfo *sra,
 494                                   int raid_disks,
 495                                   int nrdisks,
 496                                   unsigned long blocks,
 497                                   char *backup_file,
 498                                   int *fdlist,
 499                                   unsigned long long *offsets);
 500 extern void reshape_free_fdlist(int *fdlist,
 501                                 unsigned long long *offsets,
 502                                 int size);
 503 extern int reshape_open_backup_file(char *backup,
 504                                     int fd,
 505                                     char *devname,
 506                                     long blocks,
 507                                     int *fdlist,
 508                                     unsigned long long *offsets,
 509                                     int restart);
 510 extern unsigned long compute_backup_blocks(int nchunk, int ochunk,
 511                                            unsigned int ndata, unsigned int odata);
 512
 513 extern int save_stripes(int *source, unsigned long long *offsets,
 514                         int raid_disks, int chunk_size, int level, int layout,
 515                         int nwrites, int *dest,
 516                         unsigned long long start, unsigned long long length,
 517                         char *buf);
 518 extern int restore_stripes(int *dest, unsigned long long *offsets,
 519                            int raid_disks, int chunk_size, int level, int layout,
 520                            int source, unsigned long long read_offset,
 521                            unsigned long long start, unsigned long long length,
 522                            char *src_buf);
 523
 524 #ifndef Sendmail
 525 #define Sendmail "/usr/lib/sendmail -t"
 526 #endif
 527
 528 #define SYSLOG_FACILITY LOG_DAEMON
 529
 530 extern char *map_num(mapping_t *map, int num);
 531 extern int map_name(mapping_t *map, char *name);
 532 extern mapping_t r5layout[], r6layout[], pers[], modes[], faultylayout[];
 533
 534 extern char *map_dev(int major, int minor, int create);
 535
 536 struct active_array;
 537 struct metadata_update;
 538
 539
 540 /* 'struct reshape' records the intermediate states
 541  * a general reshape.
 542  * The starting geometry is converted to the 'before' geometry
 543  * by at most an atomic level change. They could be the same.
 544  * Similarly the 'after' geometry is converted to the final
 545  * geometry by at most a level change.
 546  * Note that 'before' and 'after' must have the same level.
 547  * 'blocks' is the minimum number of sectors for a reshape unit.
 548  * This will be a multiple of the stripe size in each of the
 549  * 'before' and 'after' geometries.
 550  * If 'blocks' is 0, no restriping is necessary.
 551  */
 552 struct reshape {
 553         int level;
 554         int parity; /* number of parity blocks/devices */
 555         struct {
 556                 int layout;
 557                 int data_disks;
 558         } before, after;
 559         unsigned long long backup_blocks;
 560         unsigned long long stripes; /* number of old stripes that comprise 'blocks'*/
 561         unsigned long long new_size; /* New size of array in sectors */
 562 };
 563
 564 /* A superswitch provides entry point the a metadata handler.
 565  *
 566  * The super_switch primarily operates on some "metadata" that
 567  * is accessed via the 'supertype'.
 568  * This metadata has one of three possible sources.
 569  * 1/ It is read from a single device.  In this case it may not completely
 570  *    describe the array or arrays as some information might be on other
 571  *    devices.
 572  * 2/ It is read from all devices in a container.  In this case all
 573  *    information is present.
 574  * 3/ It is created by ->init_super / ->add_to_super.  In this case it will
 575  *    be complete once enough ->add_to_super calls have completed.
 576  *
 577  * When creating an array inside a container, the metadata will be
 578  * formed by a combination of 2 and 3.  The metadata or the array is read,
 579  * then new information is added.
 580  *
 581  * The metadata must sometimes have a concept of a 'current' array
 582  * and a 'current' device.
 583  * The 'current' array is set by init_super to be the newly created array,
 584  * or is set by super_by_fd when it finds it is looking at an array inside
 585  * a container.
 586  *
 587  * The 'current' device is either the device that the metadata was read from
 588  * in case 1, or the last device added by add_to_super in case 3.
 589  * Case 2 does not identify a 'current' device.
 590  */
 591 extern struct superswitch {
 592
 593         /* Used to report details of metadata read from a component
 594          * device. ->load_super has been called.
 595          */
 596         void (*examine_super)(struct supertype *st, char *homehost);
 597         void (*brief_examine_super)(struct supertype *st, int verbose);
 598         void (*brief_examine_subarrays)(struct supertype *st, int verbose);
 599         void (*export_examine_super)(struct supertype *st);
 600
 601         /* Used to report details of an active array.
 602          * ->load_super was possibly given a 'component' string.
 603          */
 604         void (*detail_super)(struct supertype *st, char *homehost);
 605         void (*brief_detail_super)(struct supertype *st);
 606         void (*export_detail_super)(struct supertype *st);
 607
 608         /* Optional: platform hardware / firmware details */
 609         int (*detail_platform)(int verbose, int enumerate_only);
 610
 611         /* Used:
 612          *   to get uuid to storing in bitmap metadata
 613          *   and 'reshape' backup-data metadata
 614          *   To see if a device is being re-added to an array it was part of.
 615          */
 616         void (*uuid_from_super)(struct supertype *st, int uuid[4]);
 617
 618         /* Extract generic details from metadata.  This could be details about
 619          * the container, or about an individual array within the container.
 620          * The determination is made either by:
 621          *   load_super being given a 'component' string.
 622          *   validate_geometry determining what to create.
 623          * The info includes both array information and device information.
 624          * The particular device should be:
 625          *   The last device added by add_to_super
 626          *   The device the metadata was loaded from by load_super
 627          * If 'map' is present, then it is an array raid_disks long
 628          * (raid_disk must already be set and correct) and it is filled
 629          * with 1 for slots that are thought to be active and 0 for slots which
 630          * appear to be failed/missing.
 631          * *info is zeroed out before data is added.
 632          */
 633         void (*getinfo_super)(struct supertype *st, struct mdinfo *info, char *map);
 634         struct mdinfo *(*getinfo_super_disks)(struct supertype *st);
 635         /* Check if the given metadata is flagged as belonging to "this"
 636          * host.  0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
 637          */
 638         int (*match_home)(struct supertype *st, char *homehost);
 639
 640         /* Make one of several generic modifications to metadata
 641          * prior to assembly (or other times).
 642          *   sparc2.2  - first bug in early 0.90 metadata
 643          *   super-minor - change name of 0.90 metadata
 644          *   summaries - 'correct' any redundant data
 645          *   resync - mark array as dirty to trigger a resync.
 646          *   uuid - set new uuid - only 0.90 or 1.x
 647          *   name - change the name of the array (where supported)
 648          *   homehost - change which host this array is tied to.
 649          *   devicesize - If metadata is at start of device, change recorded
 650          *               device size to match actual device size
 651          *   byteorder - swap bytes for 0.90 metadata
 652          *
 653          *   force-one  - mark that device as uptodate, not old or failed.
 654          *   force-array - mark array as clean if it would not otherwise
 655          *               assemble
 656          *   assemble   - not sure how this is different from force-one...
 657          *   linear-grow-new - add a new device to a linear array, but don't
 658          *                   change the size: so superblock still matches
 659          *   linear-grow-update - now change the size of the array.
 660          *   writemostly - set the WriteMostly1 bit in the superblock devflags
 661          *   readwrite - clear the WriteMostly1 bit in the superblock devflags
 662          */
 663         int (*update_super)(struct supertype *st, struct mdinfo *info,
 664                             char *update,
 665                             char *devname, int verbose,
 666                             int uuid_set, char *homehost);
 667
 668         /* Create new metadata for new array as described.  This could
 669          * be a new container, or an array in a pre-existing container.
 670          * Also used to zero metadata prior to writing it to invalidate old
 671          * metadata.
 672          */
 673         int (*init_super)(struct supertype *st, mdu_array_info_t *info,
 674                           unsigned long long size, char *name,
 675                           char *homehost, int *uuid);
 676
 677         /* update the metadata to include new device, either at create or
 678          * when hot-adding a spare.
 679          */
 680         int (*add_to_super)(struct supertype *st, mdu_disk_info_t *dinfo,
 681                             int fd, char *devname);
 682         /* update the metadata to delete a device,
 683          * when hot-removing.
 684          */
 685         int (*remove_from_super)(struct supertype *st, mdu_disk_info_t *dinfo);
 686
 687         /* Write metadata to one device when fixing problems or adding
 688          * a new device.
 689          */
 690         int (*store_super)(struct supertype *st, int fd);
 691
 692         /*  Write all metadata for this array.
 693          */
 694         int (*write_init_super)(struct supertype *st);
 695         int (*compare_super)(struct supertype *st, struct supertype *tst);
 696         int (*load_super)(struct supertype *st, int fd, char *devname);
 697         int (*load_container)(struct supertype *st, int fd, char *devname);
 698         struct supertype * (*match_metadata_desc)(char *arg);
 699         __u64 (*avail_size)(struct supertype *st, __u64 size);
 700         unsigned long long (*min_acceptable_spare_size)(struct supertype *st);
 701         int (*add_internal_bitmap)(struct supertype *st, int *chunkp,
 702                                    int delay, int write_behind,
 703                                    unsigned long long size, int may_change, int major);
 704         void (*locate_bitmap)(struct supertype *st, int fd);
 705         int (*write_bitmap)(struct supertype *st, int fd);
 706         void (*free_super)(struct supertype *st);
 707
 708         /* validate_geometry is called with an st returned by
 709          * match_metadata_desc.
 710          * It should check that the geometry described in compatible with
 711          * the metadata type.  It will be called repeatedly as devices
 712          * added to validate changing size and new devices.  If there are
 713          * inter-device dependencies, it should record sufficient details
 714          * so these can be validated.
 715          * Both 'size' and '*freesize' are in sectors.  chunk is KiB.
 716          * Return value is:
 717          *  1: everything is OK
 718          *  0: not OK for some reason - if 'verbose', then error was reported.
 719          * -1: st->sb was NULL, 'subdev' is a member of a container of this
 720          *     types, but array is not acceptable for some reason
 721          *     message was reported even if verbose is 0.
 722          */
 723         int (*validate_geometry)(struct supertype *st, int level, int layout,
 724                                  int raiddisks,
 725                                  int *chunk, unsigned long long size,
 726                                  char *subdev, unsigned long long *freesize,
 727                                  int verbose);
 728
 729         struct mdinfo *(*container_content)(struct supertype *st, char *subarray);
 730         /* query the supertype for default geometry */
 731         void (*default_geometry)(struct supertype *st, int *level, int *layout, int *chunk); /* optional */
 732         /* Permit subarray's to be deleted from inactive containers */
 733         int (*kill_subarray)(struct supertype *st); /* optional */
 734         /* Permit subarray's to be modified */
 735         int (*update_subarray)(struct supertype *st, char *subarray,
 736                                char *update, struct mddev_ident *ident); /* optional */
 737         /* Check if reshape is supported for this external format.
 738          * st is obtained from super_by_fd() where st->subarray[0] is
 739          * initialized to indicate if reshape is being performed at the
 740          * container or subarray level
 741          */
 742         int (*reshape_super)(struct supertype *st, long long size, int level,
 743                              int layout, int chunksize, int raid_disks,
 744                              int delta_disks, char *backup, char *dev,
 745                              int verbose); /* optional */
 746         int (*manage_reshape)( /* optional */
 747                 int afd, struct mdinfo *sra, struct reshape *reshape,
 748                 struct supertype *st, unsigned long blocks,
 749                 int *fds, unsigned long long *offsets,
 750                 int dests, int *destfd, unsigned long long *destoffsets);
 751
 752 /* for mdmon */
 753         int (*open_new)(struct supertype *c, struct active_array *a,
 754                         char *inst);
 755
 756         /* Tell the metadata handler the current state of the array.
 757          * This covers whether it is known to be consistent (no pending writes)
 758          * and how far along a resync is known to have progressed
 759          * (in a->resync_start).
 760          * resync status is really irrelevant if the array is not consistent,
 761          * but some metadata (DDF!) have a place to record the distinction.
 762          * If 'consistent' is '2', then the array can mark it dirty if a
 763          * resync/recovery/whatever is required, or leave it clean if not.
 764          * Return value is 0 dirty (not consistent) and 1 if clean.
 765          * it is only really important if consistent is passed in as '2'.
 766          */
 767         int (*set_array_state)(struct active_array *a, int consistent);
 768
 769         /* When the state of a device might have changed, we call set_disk to
 770          * tell the metadata what the current state is.
 771          * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
 772          * transitions.
 773          * set_disk might be called when the state of the particular disk has
 774          * not in fact changed.
 775          */
 776         void (*set_disk)(struct active_array *a, int n, int state);
 777         void (*sync_metadata)(struct supertype *st);
 778         void (*process_update)(struct supertype *st,
 779                                struct metadata_update *update);
 780         void (*prepare_update)(struct supertype *st,
 781                                struct metadata_update *update);
 782
 783         /* activate_spare will check if the array is degraded and, if it
 784          * is, try to find some spare space in the container.
 785          * On success, it add appropriate updates (For process_update) to
 786          * to the 'updates' list and returns a list of 'mdinfo' identifying
 787          * the device, or devices as there might be multiple missing
 788          * devices and multiple spares available.
 789          */
 790         struct mdinfo *(*activate_spare)(struct active_array *a,
 791                                          struct metadata_update **updates);
 792         /*
 793          * Return statically allocated string that represents metadata specific
 794          * controller domain of the disk. The domain is used in disk domain
 795          * matching functions. Disks belong to the same domain if the they have
 796          * the same domain from mdadm.conf and belong the same metadata domain.
 797          * Returning NULL or not providing this handler means that metadata
 798          * does not distinguish the differences between disks that belong to
 799          * different controllers. They are in the domain specified by
 800          * configuration file (mdadm.conf).
 801          * In case when the metadata has the notion of domains based on disk
 802          * it shall return NULL for disks that do not belong to the controller
 803          * the supported domains. Such disks will form another domain and won't
 804          * be mixed with supported ones.
 805          */
 806         const char *(*get_disk_controller_domain)(const char *path);
 807
 808         /* for external backup area */
 809         int (*recover_backup)(struct supertype *st, struct mdinfo *info);
 810
 811         int swapuuid; /* true if uuid is bigending rather than hostendian */
 812         int external;
 813         const char *name; /* canonical metadata name */
 814 } *superlist[];
 815
 816 extern struct superswitch super0, super1;
 817 extern struct superswitch super_imsm, super_ddf;
 818 extern struct superswitch mbr, gpt;
 819
 820 struct metadata_update {
 821         int     len;
 822         char    *buf;
 823         void    *space; /* allocated space that monitor will use */
 824         void    **space_list; /* list of allocated spaces that monitor can
 825                                * use or that it returned.
 826                                */
 827         struct metadata_update *next;
 828 };
 829
 830 /* A supertype holds a particular collection of metadata.
 831  * It identifies the metadata type by the superswitch, and the particular
 832  * sub-version of that metadata type.
 833  * metadata read in or created is stored in 'sb' and 'info'.
 834  * There are also fields used by mdmon to track containers.
 835  *
 836  * A supertype may refer to:
 837  *   Just an array, possibly in a container
 838  *   A container, not identifying any particular array
 839  *   Info read from just one device, not yet fully describing the array/container.
 840  *
 841  *
 842  * A supertype is created by:
 843  *   super_by_fd
 844  *   guess_super
 845  *   dup_super
 846  */
 847 struct supertype {
 848         struct superswitch *ss;
 849         int minor_version;
 850         int max_devs;
 851         int container_dev;    /* devnum of container */
 852         void *sb;
 853         void *info;
 854         int ignore_hw_compat; /* used to inform metadata handlers that it should ignore
 855                                  HW/firmware related incompatability to load metadata.
 856                                  Used when examining metadata to display content of disk
 857                                  when user has no hw/firmare compatible system.
 858                               */
 859         struct metadata_update *updates;
 860         struct metadata_update **update_tail;
 861
 862         /* extra stuff used by mdmon */
 863         struct active_array *arrays;
 864         int sock; /* listen to external programs */
 865         int devnum;
 866         char *devname; /* e.g. md0.  This appears in metadata_verison:
 867                         *  external:/md0/12
 868                         */
 869         int devcnt;
 870
 871         struct mdinfo *devs;
 872
 873 };
 874
 875 extern struct supertype *super_by_fd(int fd, char **subarray);
 876 enum guess_types { guess_any, guess_array, guess_partitions };
 877 extern struct supertype *guess_super_type(int fd, enum guess_types guess_type);
 878 static inline struct supertype *guess_super(int fd) {
 879         return guess_super_type(fd, guess_any);
 880 }
 881 extern struct supertype *dup_super(struct supertype *st);
 882 extern int get_dev_size(int fd, char *dname, unsigned long long *sizep);
 883 extern int must_be_container(int fd);
 884 extern int dev_size_from_id(dev_t id, unsigned long long *size);
 885 extern void get_one_disk(int mdfd, mdu_array_info_t *ainf,
 886                          mdu_disk_info_t *disk);
 887 void wait_for(char *dev, int fd);
 888
 889 /*
 890  * Data structures for policy management.
 891  * Each device can have a policy structure that lists
 892  * various name/value pairs each possibly with a metadata associated.
 893  * The policy list is sorted by name/value/metadata
 894  */
 895 struct dev_policy {
 896         struct dev_policy *next;
 897         char *name;     /* None of these strings are allocated.  They are
 898                          * all just references to strings which are known
 899                          * to exist elsewhere.
 900                          * name and metadata can be compared by address equality.
 901                          */
 902         const char *metadata;
 903         const char *value;
 904 };
 905
 906 extern char pol_act[], pol_domain[], pol_metadata[], pol_auto[];
 907
 908 /* iterate over the sublist starting at list, having the same
 909  * 'name' as 'list', and matching the given metadata (Where
 910  * NULL matches anything
 911  */
 912 #define pol_for_each(item, list, _metadata)                             \
 913         for (item = list;                                               \
 914              item && item->name == list->name;                          \
 915              item = item->next)                                         \
 916                 if (!(!_metadata || !item->metadata || _metadata == item->metadata)) \
 917                         ; else
 918
 919 /*
 920  * policy records read from mdadm are largely just name-value pairs.
 921  * The names are constants, not strdupped
 922  */
 923 struct pol_rule {
 924         struct pol_rule *next;
 925         char *type;     /* rule_policy or rule_part */
 926         struct rule {
 927                 struct rule *next;
 928                 char *name;
 929                 char *value;
 930                 char *dups; /* duplicates of 'value' with a partNN appended */
 931         } *rule;
 932 };
 933
 934 extern char rule_policy[], rule_part[];
 935 extern char rule_path[], rule_type[];
 936 extern char type_part[], type_disk[];
 937
 938 extern void policyline(char *line, char *type);
 939 extern void policy_add(char *type, ...);
 940 extern void policy_free(void);
 941
 942 extern struct dev_policy *path_policy(char *path, char *type);
 943 extern struct dev_policy *disk_policy(struct mdinfo *disk);
 944 extern struct dev_policy *devnum_policy(int dev);
 945 extern void dev_policy_free(struct dev_policy *p);
 946
 947 //extern void pol_new(struct dev_policy **pol, char *name, char *val, char *metadata);
 948 extern void pol_add(struct dev_policy **pol, char *name, char *val, char *metadata);
 949 extern struct dev_policy *pol_find(struct dev_policy *pol, char *name);
 950
 951 enum policy_action {
 952         act_default,
 953         act_include,
 954         act_re_add,
 955         act_spare,      /* This only applies to bare devices */
 956         act_spare_same_slot, /* this allows non-bare devices,
 957                               * but only if recent removal */
 958         act_force_spare, /* this allow non-bare devices in any case */
 959         act_err
 960 };
 961
 962 extern int policy_action_allows(struct dev_policy *plist, const char *metadata,
 963                                 enum policy_action want);
 964 extern int disk_action_allows(struct mdinfo *disk, const char *metadata,
 965                               enum policy_action want);
 966
 967 struct domainlist {
 968         struct domainlist *next;
 969         const char *dom;
 970 };
 971
 972 extern int domain_test(struct domainlist *dom, struct dev_policy *pol,
 973                        const char *metadata);
 974 extern struct domainlist *domain_from_array(struct mdinfo *mdi,
 975                                             const char *metadata);
 976 extern void domainlist_add_dev(struct domainlist **dom, int devnum,
 977                                const char *metadata);
 978 extern void domain_free(struct domainlist *dl);
 979 extern void domain_merge(struct domainlist **domp, struct dev_policy *pol,
 980                          const char *metadata);
 981 void domain_add(struct domainlist **domp, char *domain);
 982
 983 extern void policy_save_path(char *id_path, struct map_ent *array);
 984 extern int policy_check_path(struct mdinfo *disk, struct map_ent *array);
 985
 986 #if __GNUC__ < 3
 987 struct stat64;
 988 #endif
 989
 990 #define HAVE_NFTW  we assume
 991 #define HAVE_FTW
 992
 993 #ifdef __UCLIBC__
 994 # include <features.h>
 995 # ifndef __UCLIBC_HAS_LFS__
 996 #  define lseek64 lseek
 997 # endif
 998 # ifndef  __UCLIBC_HAS_FTW__
 999 #  undef HAVE_FTW
1000 #  undef HAVE_NFTW
1001 # endif
1002 #endif
1003
1004 #ifdef __dietlibc__
1005 # undef HAVE_NFTW
1006 #endif
1007
1008 #if defined(__KLIBC__)
1009 # undef HAVE_NFTW
1010 # undef HAVE_FTW
1011 #endif
1012
1013 #ifndef HAVE_NFTW
1014 # define FTW_PHYS 1
1015 # ifndef HAVE_FTW
1016   struct FTW {};
1017 # endif
1018 #endif
1019
1020 #ifdef HAVE_FTW
1021 # include <ftw.h>
1022 #endif
1023
1024 extern int add_dev(const char *name, const struct stat *stb, int flag, struct FTW *s);
1025
1026
1027 extern int Manage_ro(char *devname, int fd, int readonly);
1028 extern int Manage_runstop(char *devname, int fd, int runstop, int quiet);
1029 extern int Manage_resize(char *devname, int fd, long long size, int raid_disks);
1030 extern int Manage_subdevs(char *devname, int fd,
1031                           struct mddev_dev *devlist, int verbose, int test,
1032                           char *update, int force);
1033 extern int autodetect(void);
1034 extern int Grow_Add_device(char *devname, int fd, char *newdev);
1035 extern int Grow_addbitmap(char *devname, int fd, char *file, int chunk, int delay, int write_behind, int force);
1036 extern int Grow_reshape(char *devname, int fd, int quiet, char *backup_file,
1037                         long long size,
1038                         int level, char *layout_str, int chunksize, int raid_disks,
1039                         struct mddev_dev *devlist,
1040                         int assume_clean, int force);
1041 extern int Grow_restart(struct supertype *st, struct mdinfo *info,
1042                         int *fdlist, int cnt, char *backup_file, int verbose);
1043 extern int Grow_continue(int mdfd, struct supertype *st,
1044                          struct mdinfo *info, char *backup_file,
1045                          int freeze_reshape);
1046
1047 extern int restore_backup(struct supertype *st,
1048                           struct mdinfo *content,
1049                           int working_disks,
1050                           int spares,
1051                           char *backup_file,
1052                           int verbose);
1053 extern int Grow_continue_command(char *devname, int fd,
1054                                  char *backup_file, int verbose);
1055
1056 extern int Assemble(struct supertype *st, char *mddev,
1057                     struct mddev_ident *ident,
1058                     struct mddev_dev *devlist,
1059                     char *backup_file, int invalid_backup,
1060                     int readonly, int runstop,
1061                     char *update, char *homehost, int require_homehost,
1062                     int verbose, int force, int freeze_reshape);
1063
1064 extern int Build(char *mddev, int chunk, int level, int layout,
1065                  int raiddisks, struct mddev_dev *devlist, int assume_clean,
1066                  char *bitmap_file, int bitmap_chunk, int write_behind,
1067                  int delay, int verbose, int autof, unsigned long long size);
1068
1069
1070 extern int Create(struct supertype *st, char *mddev,
1071                   int chunk, int level, int layout, unsigned long long size, int raiddisks, int sparedisks,
1072                   char *name, char *homehost, int *uuid,
1073                   int subdevs, struct mddev_dev *devlist,
1074                   int runstop, int verbose, int force, int assume_clean,
1075                   char *bitmap_file, int bitmap_chunk, int write_behind, int delay, int autof);
1076
1077 extern int Detail(char *dev, int brief, int export, int test, char *homehost);
1078 extern int Detail_Platform(struct superswitch *ss, int scan, int verbose);
1079 extern int Query(char *dev);
1080 extern int Examine(struct mddev_dev *devlist, int brief, int export, int scan,
1081                    int SparcAdjust, struct supertype *forcest, char *homehost);
1082 extern int Monitor(struct mddev_dev *devlist,
1083                    char *mailaddr, char *alert_cmd,
1084                    int period, int daemonise, int scan, int oneshot,
1085                    int dosyslog, int test, char *pidfile, int increments,
1086                    int share);
1087
1088 extern int Kill(char *dev, struct supertype *st, int force, int quiet, int noexcl);
1089 extern int Kill_subarray(char *dev, char *subarray, int quiet);
1090 extern int Update_subarray(char *dev, char *subarray, char *update, struct mddev_ident *ident, int quiet);
1091 extern int Wait(char *dev);
1092 extern int WaitClean(char *dev, int sock, int verbose);
1093
1094 extern int Incremental(char *devname, int verbose, int runstop,
1095                        struct supertype *st, char *homehost, int require_homehost,
1096                        int autof, int freeze_reshape);
1097 extern void RebuildMap(void);
1098 extern int IncrementalScan(int verbose);
1099 extern int IncrementalRemove(char *devname, char *path, int verbose);
1100 extern int CreateBitmap(char *filename, int force, char uuid[16],
1101                         unsigned long chunksize, unsigned long daemon_sleep,
1102                         unsigned long write_behind,
1103                         unsigned long long array_size,
1104                         int major);
1105 extern int ExamineBitmap(char *filename, int brief, struct supertype *st);
1106 extern int Write_rules(char *rule_name);
1107 extern int bitmap_update_uuid(int fd, int *uuid, int swap);
1108 extern unsigned long bitmap_sectors(struct bitmap_super_s *bsb);
1109
1110 extern int md_get_version(int fd);
1111 extern int get_linux_version(void);
1112 extern int mdadm_version(char *version);
1113 extern long long parse_size(char *size);
1114 extern int parse_uuid(char *str, int uuid[4]);
1115 extern int parse_layout_10(char *layout);
1116 extern int parse_layout_faulty(char *layout);
1117 extern int check_ext2(int fd, char *name);
1118 extern int check_reiser(int fd, char *name);
1119 extern int check_raid(int fd, char *name);
1120 extern int check_partitions(int fd, char *dname,
1121                             unsigned long long freesize,
1122                             unsigned long long size);
1123
1124 extern int get_mdp_major(void);
1125 extern int dev_open(char *dev, int flags);
1126 extern int open_dev(int devnum);
1127 extern int open_dev_flags(int devnum, int flags);
1128 extern int open_dev_excl(int devnum);
1129 extern int is_standard(char *dev, int *nump);
1130 extern int same_dev(char *one, char *two);
1131
1132 extern int parse_auto(char *str, char *msg, int config);
1133 extern struct mddev_ident *conf_get_ident(char *dev);
1134 extern struct mddev_dev *conf_get_devs(void);
1135 extern int conf_test_dev(char *devname);
1136 extern int conf_test_metadata(const char *version, struct dev_policy *pol, int is_homehost);
1137 extern struct createinfo *conf_get_create_info(void);
1138 extern void set_conffile(char *file);
1139 extern char *conf_get_mailaddr(void);
1140 extern char *conf_get_mailfrom(void);
1141 extern char *conf_get_program(void);
1142 extern char *conf_get_homehost(int *require_homehostp);
1143 extern char *conf_line(FILE *file);
1144 extern char *conf_word(FILE *file, int allow_key);
1145 extern int conf_name_is_free(char *name);
1146 extern int conf_verify_devnames(struct mddev_ident *array_list);
1147 extern int devname_matches(char *name, char *match);
1148 extern struct mddev_ident *conf_match(struct supertype *st,
1149                                       struct mdinfo *info,
1150                                       char *devname,
1151                                       int verbose, int *rvp);
1152 extern int experimental(void);
1153
1154 extern void free_line(char *line);
1155 extern int match_oneof(char *devices, char *devname);
1156 extern void uuid_from_super(int uuid[4], mdp_super_t *super);
1157 extern const int uuid_zero[4];
1158 extern int same_uuid(int a[4], int b[4], int swapuuid);
1159 extern void copy_uuid(void *a, int b[4], int swapuuid);
1160 extern char *__fname_from_uuid(int id[4], int swap, char *buf, char sep);
1161 extern char *fname_from_uuid(struct supertype *st,
1162                              struct mdinfo *info, char *buf, char sep);
1163 extern unsigned long calc_csum(void *super, int bytes);
1164 extern int enough(int level, int raid_disks, int layout, int clean,
1165                    char *avail, int avail_disks);
1166 extern int enough_fd(int fd);
1167 extern int ask(char *mesg);
1168 extern unsigned long long get_component_size(int fd);
1169 extern void remove_partitions(int fd);
1170 extern int test_partition(int fd);
1171 extern int test_partition_from_id(dev_t id);
1172 extern int get_data_disks(int level, int layout, int raid_disks);
1173 extern unsigned long long calc_array_size(int level, int raid_disks, int layout,
1174                                    int chunksize, unsigned long long devsize);
1175 extern int flush_metadata_updates(struct supertype *st);
1176 extern void append_metadata_update(struct supertype *st, void *buf, int len);
1177 extern int assemble_container_content(struct supertype *st, int mdfd,
1178                                       struct mdinfo *content, int runstop,
1179                                       char *chosen_name, int verbose,
1180                                       char *backup_file, int freeze_reshape);
1181 extern struct mdinfo *container_choose_spares(struct supertype *st,
1182                                               unsigned long long min_size,
1183                                               struct domainlist *domlist,
1184                                               char *spare_group,
1185                                               const char *metadata, int get_one);
1186 extern int move_spare(char *from_devname, char *to_devname, dev_t devid);
1187 extern int add_disk(int mdfd, struct supertype *st,
1188                     struct mdinfo *sra, struct mdinfo *info);
1189 extern int remove_disk(int mdfd, struct supertype *st,
1190                        struct mdinfo *sra, struct mdinfo *info);
1191 extern int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info);
1192 unsigned long long min_recovery_start(struct mdinfo *array);
1193
1194 extern char *human_size(long long bytes);
1195 extern char *human_size_brief(long long bytes);
1196 extern void print_r10_layout(int layout);
1197
1198 #define NoMdDev (1<<23)
1199 extern int find_free_devnum(int use_partitions);
1200
1201 extern void put_md_name(char *name);
1202 extern char *get_md_name(int dev);
1203
1204 extern char DefaultConfFile[];
1205
1206 extern int create_mddev(char *dev, char *name, int autof, int trustworthy,
1207                         char *chosen);
1208 /* values for 'trustworthy' */
1209 #define LOCAL   1
1210 #define LOCAL_ANY 10
1211 #define FOREIGN 2
1212 #define METADATA 3
1213 extern int open_mddev(char *dev, int report_errors);
1214 extern int open_container(int fd);
1215 extern int is_container_member(struct mdstat_ent *ent, char *devname);
1216 extern int is_subarray_active(char *subarray, char *devname);
1217 extern int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet);
1218 extern struct superswitch *version_to_superswitch(char *vers);
1219
1220 extern int mdmon_running(int devnum);
1221 extern int mdmon_pid(int devnum);
1222 extern int check_env(char *name);
1223 extern __u32 random32(void);
1224 extern int start_mdmon(int devnum);
1225
1226 extern int child_monitor(int afd, struct mdinfo *sra, struct reshape *reshape,
1227                          struct supertype *st, unsigned long stripes,
1228                          int *fds, unsigned long long *offsets,
1229                          int dests, int *destfd, unsigned long long *destoffsets);
1230 void abort_reshape(struct mdinfo *sra);
1231
1232 extern char *devnum2devname(int num);
1233 extern void fmt_devname(char *name, int num);
1234 extern int devname2devnum(char *name);
1235 extern int stat2devnum(struct stat *st);
1236 extern int fd2devnum(int fd);
1237
1238 static inline int dev2major(int d)
1239 {
1240         if (d >= 0)
1241                 return MD_MAJOR;
1242         else
1243                 return get_mdp_major();
1244 }
1245
1246 static inline int dev2minor(int d)
1247 {
1248         if (d >= 0)
1249                 return d;
1250         return (-1-d) << MdpMinorShift;
1251 }
1252
1253 static inline int ROUND_UP(int a, int base)
1254 {
1255         return ((a+base-1)/base)*base;
1256 }
1257
1258 static inline int is_subarray(char *vers)
1259 {
1260         /* The version string for a 'subarray' (an array in a container)
1261          * is
1262          *    /containername/componentname    for normal read-write arrays
1263          *    -containername/componentname    for arrays which mdmon must not
1264          *                                    reconfigure.  They might be read-only
1265          *                                    or might be undergoing reshape etc.
1266          * containername is e.g. md0, md_d1
1267          * componentname is dependant on the metadata. e.g. '1' 'S1' ...
1268          */
1269         return (*vers == '/' || *vers == '-');
1270 }
1271
1272 static inline char *to_subarray(struct mdstat_ent *ent, char *container)
1273 {
1274         return &ent->metadata_version[10+strlen(container)+1];
1275 }
1276
1277 #ifdef DEBUG
1278 #define dprintf(fmt, arg...) \
1279         fprintf(stderr, fmt, ##arg)
1280 #else
1281 #define dprintf(fmt, arg...) \
1282         ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
1283 #endif
1284 #include <assert.h>
1285 #include <stdarg.h>
1286 static inline int xasprintf(char **strp, const char *fmt, ...) {
1287         va_list ap;
1288         int ret;
1289         va_start(ap, fmt);
1290         ret = vasprintf(strp, fmt, ap);
1291         va_end(ap);
1292         assert(ret >= 0);
1293         return ret;
1294 }
1295
1296 #define LEVEL_MULTIPATH         (-4)
1297 #define LEVEL_LINEAR            (-1)
1298 #define LEVEL_FAULTY            (-5)
1299
1300 /* kernel module doesn't know about these */
1301 #define LEVEL_CONTAINER         (-100)
1302 #define LEVEL_UNSUPPORTED       (-200)
1303
1304
1305 /* faulty stuff */
1306
1307 #define WriteTransient  0
1308 #define ReadTransient   1
1309 #define WritePersistent 2
1310 #define ReadPersistent  3
1311 #define WriteAll        4 /* doesn't go to device */
1312 #define ReadFixable     5
1313 #define Modes   6
1314
1315 #define ClearErrors     31
1316 #define ClearFaults     30
1317
1318 #define AllPersist      100 /* internal use only */
1319 #define NoPersist       101
1320
1321 #define ModeMask        0x1f
1322 #define ModeShift       5
1323
1324
1325 #ifdef __TINYC__
1326 #undef minor
1327 #undef major
1328 #undef makedev
1329 #define minor(x) ((x)&0xff)
1330 #define major(x) (((x)>>8)&0xff)
1331 #define makedev(M,m) (((M)<<8) | (m))
1332 #endif
1333
1334 /* for raid4/5/6 */
1335 #define ALGORITHM_LEFT_ASYMMETRIC       0
1336 #define ALGORITHM_RIGHT_ASYMMETRIC      1
1337 #define ALGORITHM_LEFT_SYMMETRIC        2
1338 #define ALGORITHM_RIGHT_SYMMETRIC       3
1339
1340 /* Define non-rotating (raid4) algorithms.  These allow
1341  * conversion of raid4 to raid5.
1342  */
1343 #define ALGORITHM_PARITY_0              4 /* P or P,Q are initial devices */
1344 #define ALGORITHM_PARITY_N              5 /* P or P,Q are final devices. */
1345
1346 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1347  * Firstly, the exact positioning of the parity block is slightly
1348  * different between the 'LEFT_*' modes of md and the "_N_*" modes
1349  * of DDF.
1350  * Secondly, or order of datablocks over which the Q syndrome is computed
1351  * is different.
1352  * Consequently we have different layouts for DDF/raid6 than md/raid6.
1353  * These layouts are from the DDFv1.2 spec.
1354  * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1355  * leaves RLQ=3 as 'Vendor Specific'
1356  */
1357
1358 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1359 #define ALGORITHM_ROTATING_N_RESTART    9 /* DDF PRL=6 RLQ=2 */
1360 #define ALGORITHM_ROTATING_N_CONTINUE   10 /*DDF PRL=6 RLQ=3 */
1361
1362
1363 /* For every RAID5 algorithm we define a RAID6 algorithm
1364  * with exactly the same layout for data and parity, and
1365  * with the Q block always on the last device (N-1).
1366  * This allows trivial conversion from RAID5 to RAID6
1367  */
1368 #define ALGORITHM_LEFT_ASYMMETRIC_6     16
1369 #define ALGORITHM_RIGHT_ASYMMETRIC_6    17
1370 #define ALGORITHM_LEFT_SYMMETRIC_6      18
1371 #define ALGORITHM_RIGHT_SYMMETRIC_6     19
1372 #define ALGORITHM_PARITY_0_6            20
1373 #define ALGORITHM_PARITY_N_6            ALGORITHM_PARITY_N
1374
1375 /* Define PATH_MAX in case we don't use glibc or standard library does
1376  * not have PATH_MAX defined. Assume max path length is 4K characters.
1377  */
1378 #ifndef PATH_MAX
1379 #define PATH_MAX        4096
1380 #endif
1381
1382 #define PROCESS_DELAYED -2
1383 #define PROCESS_PENDING -3