mdadm.h

   1 /*
   2  * mdadm - manage Linux "md" devices aka RAID arrays.
   3  *
   4  * Copyright (C) 2001-2009 Neil Brown <neilb@suse.de>
   5  *
   6  *
   7  *    This program is free software; you can redistribute it and/or modify
   8  *    it under the terms of the GNU General Public License as published by
   9  *    the Free Software Foundation; either version 2 of the License, or
  10  *    (at your option) any later version.
  11  *
  12  *    This program is distributed in the hope that it will be useful,
  13  *    but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  *    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  15  *    GNU General Public License for more details.
  16  *
  17  *    You should have received a copy of the GNU General Public License
  18  *    along with this program; if not, write to the Free Software
  19  *    Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  20  *
  21  *    Author: Neil Brown
  22  *    Email: <neilb@suse.de>
  23  */
  24
  25 #define _GNU_SOURCE
  26 #define _FILE_OFFSET_BITS 64
  27 #include        <unistd.h>
  28 #if !defined(__dietlibc__) && !defined(__KLIBC__)
  29 extern __off64_t lseek64 __P ((int __fd, __off64_t __offset, int __whence));
  30 #else
  31 # if defined(__NO_STAT64) || __WORDSIZE != 32
  32 # define lseek64 lseek
  33 # endif
  34 #endif
  35
  36 #include        <sys/types.h>
  37 #include        <sys/stat.h>
  38 #include        <stdlib.h>
  39 #include        <time.h>
  40 #include        <sys/time.h>
  41 #include        <getopt.h>
  42 #include        <fcntl.h>
  43 #include        <stdio.h>
  44 #include        <errno.h>
  45 #include        <string.h>
  46 #include        <syslog.h>
  47 #ifdef __dietlibc__
  48 #include        <strings.h>
  49 /* dietlibc has deprecated random and srandom!! */
  50 #define random rand
  51 #define srandom srand
  52 #endif
  53
  54
  55 #include        <linux/kdev_t.h>
  56 /*#include      <linux/fs.h> */
  57 #include        <sys/mount.h>
  58 #include        <asm/types.h>
  59 #include        <sys/ioctl.h>
  60 #define MD_MAJOR 9
  61 #define MdpMinorShift 6
  62
  63 #ifndef BLKGETSIZE64
  64 #define BLKGETSIZE64 _IOR(0x12,114,size_t) /* return device size in bytes (u64 *arg) */
  65 #endif
  66
  67 #define DEFAULT_CHUNK 512
  68 #define DEFAULT_BITMAP_CHUNK 4096
  69 #define DEFAULT_BITMAP_DELAY 5
  70 #define DEFAULT_MAX_WRITE_BEHIND 256
  71
  72 /* MAP_DIR should be somewhere that persists across the pivotroot
  73  * from early boot to late boot.
  74  * /run  seems to have emerged as the best standard.
  75  */
  76 #ifndef MAP_DIR
  77 #define MAP_DIR "/run/mdadm"
  78 #endif /* MAP_DIR */
  79 /* MAP_FILE is what we name the map file we put in MAP_DIR, in case you
  80  * want something other than the default of "map"
  81  */
  82 #ifndef MAP_FILE
  83 #define MAP_FILE "map"
  84 #endif /* MAP_FILE */
  85 /* MDMON_DIR is where pid and socket files used for communicating
  86  * with mdmon normally live.  Best is /var/run/mdadm as
  87  * mdmon is needed at early boot then it needs to write there prior
  88  * to /var/run being mounted read/write, and it also then needs to
  89  * persist beyond when /var/run is mounter read-only.  So, to be
  90  * safe, the default is somewhere that is read/write early in the
  91  * boot process and stays up as long as possible during shutdown.
  92  */
  93 #ifndef MDMON_DIR
  94 #define MDMON_DIR "/run/mdadm"
  95 #endif /* MDMON_DIR */
  96
  97 /* FAILED_SLOTS is where to save files storing recent removal of array
  98  * member in order to allow future reuse of disk inserted in the same
  99  * slot for array recovery
 100  */
 101 #ifndef FAILED_SLOTS_DIR
 102 #define FAILED_SLOTS_DIR "/run/mdadm/failed-slots"
 103 #endif /* FAILED_SLOTS */
 104
 105 #include        "md_u.h"
 106 #include        "md_p.h"
 107 #include        "bitmap.h"
 108 #include        "msg.h"
 109
 110 #include <endian.h>
 111 /* Redhat don't like to #include <asm/byteorder.h>, and
 112  * some time include <linux/byteorder/xxx_endian.h> isn't enough,
 113  * and there is no standard conversion function so... */
 114 /* And dietlibc doesn't think byteswap is ok, so.. */
 115 /*  #include <byteswap.h> */
 116 #define bswap_16(x) (((x) & 0x00ffU) << 8 | \
 117                      ((x) & 0xff00U) >> 8)
 118 #define bswap_32(x) (((x) & 0x000000ffU) << 24 | \
 119                      ((x) & 0xff000000U) >> 24 | \
 120                      ((x) & 0x0000ff00U) << 8  | \
 121                      ((x) & 0x00ff0000U) >> 8)
 122 #define bswap_64(x) (((x) & 0x00000000000000ffULL) << 56 | \
 123                      ((x) & 0xff00000000000000ULL) >> 56 | \
 124                      ((x) & 0x000000000000ff00ULL) << 40 | \
 125                      ((x) & 0x00ff000000000000ULL) >> 40 | \
 126                      ((x) & 0x0000000000ff0000ULL) << 24 | \
 127                      ((x) & 0x0000ff0000000000ULL) >> 24 | \
 128                      ((x) & 0x00000000ff000000ULL) << 8 | \
 129                      ((x) & 0x000000ff00000000ULL) >> 8)
 130
 131 #if !defined(__KLIBC__)
 132 #if BYTE_ORDER == LITTLE_ENDIAN
 133 #define __cpu_to_le16(_x) (_x)
 134 #define __cpu_to_le32(_x) (_x)
 135 #define __cpu_to_le64(_x) (_x)
 136 #define __le16_to_cpu(_x) (_x)
 137 #define __le32_to_cpu(_x) (_x)
 138 #define __le64_to_cpu(_x) (_x)
 139
 140 #define __cpu_to_be16(_x) bswap_16(_x)
 141 #define __cpu_to_be32(_x) bswap_32(_x)
 142 #define __cpu_to_be64(_x) bswap_64(_x)
 143 #define __be16_to_cpu(_x) bswap_16(_x)
 144 #define __be32_to_cpu(_x) bswap_32(_x)
 145 #define __be64_to_cpu(_x) bswap_64(_x)
 146 #elif BYTE_ORDER == BIG_ENDIAN
 147 #define __cpu_to_le16(_x) bswap_16(_x)
 148 #define __cpu_to_le32(_x) bswap_32(_x)
 149 #define __cpu_to_le64(_x) bswap_64(_x)
 150 #define __le16_to_cpu(_x) bswap_16(_x)
 151 #define __le32_to_cpu(_x) bswap_32(_x)
 152 #define __le64_to_cpu(_x) bswap_64(_x)
 153
 154 #define __cpu_to_be16(_x) (_x)
 155 #define __cpu_to_be32(_x) (_x)
 156 #define __cpu_to_be64(_x) (_x)
 157 #define __be16_to_cpu(_x) (_x)
 158 #define __be32_to_cpu(_x) (_x)
 159 #define __be64_to_cpu(_x) (_x)
 160 #else
 161 #  error "unknown endianness."
 162 #endif
 163 #endif /* __KLIBC__ */
 164
 165
 166 /*
 167  * min()/max()/clamp() macros that also do
 168  * strict type-checking.. See the
 169  * "unnecessary" pointer comparison.
 170  */
 171 #define min(x, y) ({                            \
 172         typeof(x) _min1 = (x);                  \
 173         typeof(y) _min2 = (y);                  \
 174         (void) (&_min1 == &_min2);              \
 175         _min1 < _min2 ? _min1 : _min2; })
 176
 177 #define max(x, y) ({                            \
 178         typeof(x) _max1 = (x);                  \
 179         typeof(y) _max2 = (y);                  \
 180         (void) (&_max1 == &_max2);              \
 181         _max1 > _max2 ? _max1 : _max2; })
 182
 183 /* general information that might be extracted from a superblock */
 184 struct mdinfo {
 185         mdu_array_info_t        array;
 186         mdu_disk_info_t         disk;
 187         __u64                   events;
 188         int                     uuid[4];
 189         char                    name[33];
 190         unsigned long long      data_offset;
 191         unsigned long long      component_size; /* same as array.size, except in
 192                                                  * sectors and up to 64bits.
 193                                                  */
 194         unsigned long long      custom_array_size; /* size for non-default sized
 195                                                     * arrays (in sectors)
 196                                                     */
 197 #define NO_RESHAPE              0
 198 #define VOLUME_RESHAPE          1
 199 #define CONTAINER_RESHAPE       2
 200         int                     reshape_active;
 201         unsigned long long      reshape_progress;
 202         int                     recovery_blocked; /* for external metadata it
 203                                                    * indicates that there is
 204                                                    * reshape in progress in
 205                                                    * container,
 206                                                    * for native metadata it is
 207                                                    * reshape_active field mirror
 208                                                    */
 209         union {
 210                 unsigned long long resync_start; /* per-array resync position */
 211                 unsigned long long recovery_start; /* per-device rebuild position */
 212                 #define MaxSector  (~0ULL) /* resync/recovery complete position */
 213         };
 214         long                    bitmap_offset;  /* 0 == none, 1 == a file */
 215         unsigned long           safe_mode_delay; /* ms delay to mark clean */
 216         int                     new_level, delta_disks, new_layout, new_chunk;
 217         int                     errors;
 218         unsigned long           cache_size; /* size of raid456 stripe cache*/
 219         int                     mismatch_cnt;
 220         char                    text_version[50];
 221
 222         int container_member; /* for assembling external-metatdata arrays
 223                                * This is to be used internally by metadata
 224                                * handler only */
 225         int container_enough; /* flag external handlers can set to
 226                                * indicate that subarrays have not enough (-1),
 227                                * enough to start (0), or all expected disks (1) */
 228         char            sys_name[20];
 229         struct mdinfo *devs;
 230         struct mdinfo *next;
 231
 232         /* Device info for mdmon: */
 233         int recovery_fd;
 234         int state_fd;
 235         #define DS_FAULTY       1
 236         #define DS_INSYNC       2
 237         #define DS_WRITE_MOSTLY 4
 238         #define DS_SPARE        8
 239         #define DS_BLOCKED      16
 240         #define DS_REMOVE       1024
 241         #define DS_UNBLOCK      2048
 242         int prev_state, curr_state, next_state;
 243
 244 };
 245
 246 struct createinfo {
 247         int     uid;
 248         int     gid;
 249         int     autof;
 250         int     mode;
 251         int     symlinks;
 252         struct supertype *supertype;
 253 };
 254
 255 #define Name "mdadm"
 256
 257 enum mode {
 258         ASSEMBLE=1,
 259         BUILD,
 260         CREATE,
 261         MANAGE,
 262         MISC,
 263         MONITOR,
 264         GROW,
 265         INCREMENTAL,
 266         AUTODETECT,
 267         mode_count
 268 };
 269
 270 extern char short_options[];
 271 extern char short_bitmap_options[];
 272 extern char short_bitmap_auto_options[];
 273 extern struct option long_options[];
 274 extern char Version[], Usage[], Help[], OptionHelp[],
 275         *mode_help[],
 276         Help_create[], Help_build[], Help_assemble[], Help_grow[],
 277         Help_incr[],
 278         Help_manage[], Help_misc[], Help_monitor[], Help_config[];
 279
 280 /* for option that don't have short equivilents, we assign arbitrary
 281  * numbers later than any 'short' character option.
 282  */
 283 enum special_options {
 284         AssumeClean = 300,
 285         BitmapChunk,
 286         WriteBehind,
 287         ReAdd,
 288         NoDegraded,
 289         Sparc22,
 290         BackupFile,
 291         HomeHost,
 292         AutoHomeHost,
 293         Symlinks,
 294         AutoDetect,
 295         Waitclean,
 296         DetailPlatform,
 297         KillSubarray,
 298         UpdateSubarray,
 299         IncrementalPath,
 300         NoSharing,
 301         HelpOptions,
 302         Brief,
 303         ManageOpt,
 304         Add,
 305         Remove,
 306         Fail,
 307         MiscOpt,
 308         WaitOpt,
 309         ConfigFile,
 310         ChunkSize,
 311         WriteMostly,
 312         Layout,
 313         Auto,
 314         Force,
 315         SuperMinor,
 316         EMail,
 317         ProgramOpt,
 318         Increment,
 319         Fork,
 320         Bitmap,
 321         RebuildMapOpt,
 322         InvalidBackup,
 323         UdevRules,
 324         FreezeReshape,
 325         Continue,
 326         OffRootOpt,
 327         Prefer,
 328         KillOpt,
 329 };
 330
 331 /* structures read from config file */
 332 /* List of mddevice names and identifiers
 333  * Identifiers can be:
 334  *    uuid=128-hex-uuid
 335  *    super-minor=decimal-minor-number-from-superblock
 336  *    devices=comma,separated,list,of,device,names,with,wildcards
 337  *
 338  * If multiple fields are present, the intersection of all matching
 339  * devices is considered
 340  */
 341 #define UnSet (0xfffe)
 342 struct mddev_ident {
 343         char    *devname;
 344
 345         int     uuid_set;
 346         int     uuid[4];
 347         char    name[33];
 348
 349         int super_minor;
 350
 351         char    *devices;       /* comma separated list of device
 352                                  * names with wild cards
 353                                  */
 354         int     level;
 355         int raid_disks;
 356         int spare_disks;
 357         struct supertype *st;
 358         int     autof;          /* 1 for normal, 2 for partitioned */
 359         char    *spare_group;
 360         char    *bitmap_file;
 361         int     bitmap_fd;
 362
 363         char    *container;     /* /dev/whatever name of container, or
 364                                  * uuid of container.  You would expect
 365                                  * this to be the 'devname' or UUID
 366                                  * of some other entry.
 367                                  */
 368         char    *member;        /* subarray within a container */
 369
 370         struct mddev_ident *next;
 371         union {
 372                 /* fields needed by different users of this structure */
 373                 int assembled;  /* set when assembly succeeds */
 374         };
 375 };
 376
 377 struct context {
 378         int     readonly;
 379         int     runstop;
 380         int     verbose;
 381         int     brief;
 382         int     force;
 383         char    *homehost;
 384         int     require_homehost;
 385         char    *prefer;
 386         int     export;
 387         int     test;
 388         char    *subarray;
 389         char    *update;
 390         int     scan;
 391         int     SparcAdjust;
 392         int     autof;
 393         int     delay;
 394         int     freeze_reshape;
 395         char    *backup_file;
 396         int     invalid_backup;
 397 };
 398
 399 struct shape {
 400         int     raiddisks;
 401         int     sparedisks;
 402         int     level;
 403         int     layout;
 404         char    *layout_str;
 405         int     chunk;
 406         int     bitmap_chunk;
 407         char    *bitmap_file;
 408         int     assume_clean;
 409         int     write_behind;
 410         unsigned long long size;
 411 };
 412
 413 /* List of device names - wildcards expanded */
 414 struct mddev_dev {
 415         char *devname;
 416         int disposition;        /* 'a' for add, 'r' for remove, 'f' for fail,
 417                                  * 'A' for re_add.
 418                                  * Not set for names read from .config
 419                                  */
 420         char writemostly;       /* 1 for 'set writemostly', 2 for 'clear writemostly' */
 421         char used;              /* set when used */
 422         struct mddev_dev *next;
 423 };
 424
 425 typedef struct mapping {
 426         char *name;
 427         int num;
 428 } mapping_t;
 429
 430
 431 struct mdstat_ent {
 432         char            *dev;
 433         int             devnum;
 434         int             active;
 435         char            *level;
 436         char            *pattern; /* U or up, _ for down */
 437         int             percent; /* -1 if no resync */
 438         int             resync; /* 3 if check, 2 if reshape, 1 if resync, 0 if recovery */
 439         int             devcnt;
 440         int             raid_disks;
 441         char *          metadata_version;
 442         struct dev_member {
 443                 char                    *name;
 444                 struct dev_member       *next;
 445         }               *members;
 446         struct mdstat_ent *next;
 447 };
 448
 449 extern struct mdstat_ent *mdstat_read(int hold, int start);
 450 extern void free_mdstat(struct mdstat_ent *ms);
 451 extern void mdstat_wait(int seconds);
 452 extern void mdstat_wait_fd(int fd, const sigset_t *sigmask);
 453 extern int mddev_busy(int devnum);
 454 extern struct mdstat_ent *mdstat_by_component(char *name);
 455 extern struct mdstat_ent *mdstat_by_subdev(char *subdev, int container);
 456
 457 struct map_ent {
 458         struct map_ent *next;
 459         int     devnum;
 460         char    metadata[20];
 461         int     uuid[4];
 462         int     bad;
 463         char    *path;
 464 };
 465 extern int map_update(struct map_ent **mpp, int devnum, char *metadata,
 466                       int uuid[4], char *path);
 467 extern void map_remove(struct map_ent **map, int devnum);
 468 extern struct map_ent *map_by_uuid(struct map_ent **map, int uuid[4]);
 469 extern struct map_ent *map_by_devnum(struct map_ent **map, int devnum);
 470 extern struct map_ent *map_by_name(struct map_ent **map, char *name);
 471 extern void map_read(struct map_ent **melp);
 472 extern int map_write(struct map_ent *mel);
 473 extern void map_delete(struct map_ent **mapp, int devnum);
 474 extern void map_free(struct map_ent *map);
 475 extern void map_add(struct map_ent **melp,
 476                     int devnum, char *metadata, int uuid[4], char *path);
 477 extern int map_lock(struct map_ent **melp);
 478 extern void map_unlock(struct map_ent **melp);
 479 extern void map_fork(void);
 480
 481 /* various details can be requested */
 482 enum sysfs_read_flags {
 483         GET_LEVEL       = (1 << 0),
 484         GET_LAYOUT      = (1 << 1),
 485         GET_COMPONENT   = (1 << 2),
 486         GET_CHUNK       = (1 << 3),
 487         GET_CACHE       = (1 << 4),
 488         GET_MISMATCH    = (1 << 5),
 489         GET_VERSION     = (1 << 6),
 490         GET_DISKS       = (1 << 7),
 491         GET_DEGRADED    = (1 << 8),
 492         GET_SAFEMODE    = (1 << 9),
 493         GET_BITMAP_LOCATION = (1 << 10),
 494
 495         GET_DEVS        = (1 << 20), /* gets role, major, minor */
 496         GET_OFFSET      = (1 << 21),
 497         GET_SIZE        = (1 << 22),
 498         GET_STATE       = (1 << 23),
 499         GET_ERROR       = (1 << 24),
 500 };
 501
 502 /* If fd >= 0, get the array it is open on,
 503  * else use devnum. >=0 -> major9. <0.....
 504  */
 505 extern int sysfs_open(int devnum, char *devname, char *attr);
 506 extern void sysfs_init(struct mdinfo *mdi, int fd, int devnum);
 507 extern void sysfs_free(struct mdinfo *sra);
 508 extern struct mdinfo *sysfs_read(int fd, int devnum, unsigned long options);
 509 extern int sysfs_attr_match(const char *attr, const char *str);
 510 extern int sysfs_match_word(const char *word, char **list);
 511 extern int sysfs_set_str(struct mdinfo *sra, struct mdinfo *dev,
 512                          char *name, char *val);
 513 extern int sysfs_set_num(struct mdinfo *sra, struct mdinfo *dev,
 514                          char *name, unsigned long long val);
 515 extern int sysfs_set_num_signed(struct mdinfo *sra, struct mdinfo *dev,
 516                                 char *name, long long val);
 517 extern int sysfs_uevent(struct mdinfo *sra, char *event);
 518 extern int sysfs_get_fd(struct mdinfo *sra, struct mdinfo *dev,
 519                         char *name);
 520 extern int sysfs_fd_get_ll(int fd, unsigned long long *val);
 521 extern int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev,
 522                         char *name, unsigned long long *val);
 523 extern int sysfs_fd_get_str(int fd, char *val, int size);
 524 extern int sysfs_attribute_available(struct mdinfo *sra, struct mdinfo *dev,
 525                                      char *name);
 526 extern int sysfs_get_str(struct mdinfo *sra, struct mdinfo *dev,
 527                          char *name, char *val, int size);
 528 extern int sysfs_set_safemode(struct mdinfo *sra, unsigned long ms);
 529 extern int sysfs_set_array(struct mdinfo *info, int vers);
 530 extern int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd, int resume);
 531 extern int sysfs_disk_to_scsi_id(int fd, __u32 *id);
 532 extern int sysfs_unique_holder(int devnum, long rdev);
 533 extern int sysfs_freeze_array(struct mdinfo *sra);
 534 extern int load_sys(char *path, char *buf);
 535 extern int reshape_prepare_fdlist(char *devname,
 536                                   struct mdinfo *sra,
 537                                   int raid_disks,
 538                                   int nrdisks,
 539                                   unsigned long blocks,
 540                                   char *backup_file,
 541                                   int *fdlist,
 542                                   unsigned long long *offsets);
 543 extern void reshape_free_fdlist(int *fdlist,
 544                                 unsigned long long *offsets,
 545                                 int size);
 546 extern int reshape_open_backup_file(char *backup,
 547                                     int fd,
 548                                     char *devname,
 549                                     long blocks,
 550                                     int *fdlist,
 551                                     unsigned long long *offsets,
 552                                     int restart);
 553 extern unsigned long compute_backup_blocks(int nchunk, int ochunk,
 554                                            unsigned int ndata, unsigned int odata);
 555
 556 extern int save_stripes(int *source, unsigned long long *offsets,
 557                         int raid_disks, int chunk_size, int level, int layout,
 558                         int nwrites, int *dest,
 559                         unsigned long long start, unsigned long long length,
 560                         char *buf);
 561 extern int restore_stripes(int *dest, unsigned long long *offsets,
 562                            int raid_disks, int chunk_size, int level, int layout,
 563                            int source, unsigned long long read_offset,
 564                            unsigned long long start, unsigned long long length,
 565                            char *src_buf);
 566
 567 #ifndef Sendmail
 568 #define Sendmail "/usr/lib/sendmail -t"
 569 #endif
 570
 571 #define SYSLOG_FACILITY LOG_DAEMON
 572
 573 extern char *map_num(mapping_t *map, int num);
 574 extern int map_name(mapping_t *map, char *name);
 575 extern mapping_t r5layout[], r6layout[], pers[], modes[], faultylayout[];
 576
 577 extern char *map_dev_preferred(int major, int minor, int create,
 578                                char *prefer);
 579 static inline char *map_dev(int major, int minor, int create)
 580 {
 581         return map_dev_preferred(major, minor, create, NULL);
 582 }
 583
 584 struct active_array;
 585 struct metadata_update;
 586
 587
 588 /* 'struct reshape' records the intermediate states
 589  * a general reshape.
 590  * The starting geometry is converted to the 'before' geometry
 591  * by at most an atomic level change. They could be the same.
 592  * Similarly the 'after' geometry is converted to the final
 593  * geometry by at most a level change.
 594  * Note that 'before' and 'after' must have the same level.
 595  * 'blocks' is the minimum number of sectors for a reshape unit.
 596  * This will be a multiple of the stripe size in each of the
 597  * 'before' and 'after' geometries.
 598  * If 'blocks' is 0, no restriping is necessary.
 599  */
 600 struct reshape {
 601         int level;
 602         int parity; /* number of parity blocks/devices */
 603         struct {
 604                 int layout;
 605                 int data_disks;
 606         } before, after;
 607         unsigned long long backup_blocks;
 608         unsigned long long stripes; /* number of old stripes that comprise 'blocks'*/
 609         unsigned long long new_size; /* New size of array in sectors */
 610 };
 611
 612 /* A superswitch provides entry point the a metadata handler.
 613  *
 614  * The super_switch primarily operates on some "metadata" that
 615  * is accessed via the 'supertype'.
 616  * This metadata has one of three possible sources.
 617  * 1/ It is read from a single device.  In this case it may not completely
 618  *    describe the array or arrays as some information might be on other
 619  *    devices.
 620  * 2/ It is read from all devices in a container.  In this case all
 621  *    information is present.
 622  * 3/ It is created by ->init_super / ->add_to_super.  In this case it will
 623  *    be complete once enough ->add_to_super calls have completed.
 624  *
 625  * When creating an array inside a container, the metadata will be
 626  * formed by a combination of 2 and 3.  The metadata or the array is read,
 627  * then new information is added.
 628  *
 629  * The metadata must sometimes have a concept of a 'current' array
 630  * and a 'current' device.
 631  * The 'current' array is set by init_super to be the newly created array,
 632  * or is set by super_by_fd when it finds it is looking at an array inside
 633  * a container.
 634  *
 635  * The 'current' device is either the device that the metadata was read from
 636  * in case 1, or the last device added by add_to_super in case 3.
 637  * Case 2 does not identify a 'current' device.
 638  */
 639 extern struct superswitch {
 640
 641         /* Used to report details of metadata read from a component
 642          * device. ->load_super has been called.
 643          */
 644         void (*examine_super)(struct supertype *st, char *homehost);
 645         void (*brief_examine_super)(struct supertype *st, int verbose);
 646         void (*brief_examine_subarrays)(struct supertype *st, int verbose);
 647         void (*export_examine_super)(struct supertype *st);
 648
 649         /* Used to report details of an active array.
 650          * ->load_super was possibly given a 'component' string.
 651          */
 652         void (*detail_super)(struct supertype *st, char *homehost);
 653         void (*brief_detail_super)(struct supertype *st);
 654         void (*export_detail_super)(struct supertype *st);
 655
 656         /* Optional: platform hardware / firmware details */
 657         int (*detail_platform)(int verbose, int enumerate_only);
 658
 659         /* Used:
 660          *   to get uuid to storing in bitmap metadata
 661          *   and 'reshape' backup-data metadata
 662          *   To see if a device is being re-added to an array it was part of.
 663          */
 664         void (*uuid_from_super)(struct supertype *st, int uuid[4]);
 665
 666         /* Extract generic details from metadata.  This could be details about
 667          * the container, or about an individual array within the container.
 668          * The determination is made either by:
 669          *   load_super being given a 'component' string.
 670          *   validate_geometry determining what to create.
 671          * The info includes both array information and device information.
 672          * The particular device should be:
 673          *   The last device added by add_to_super
 674          *   The device the metadata was loaded from by load_super
 675          * If 'map' is present, then it is an array raid_disks long
 676          * (raid_disk must already be set and correct) and it is filled
 677          * with 1 for slots that are thought to be active and 0 for slots which
 678          * appear to be failed/missing.
 679          * *info is zeroed out before data is added.
 680          */
 681         void (*getinfo_super)(struct supertype *st, struct mdinfo *info, char *map);
 682         struct mdinfo *(*getinfo_super_disks)(struct supertype *st);
 683         /* Check if the given metadata is flagged as belonging to "this"
 684          * host.  0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
 685          */
 686         int (*match_home)(struct supertype *st, char *homehost);
 687
 688         /* Make one of several generic modifications to metadata
 689          * prior to assembly (or other times).
 690          *   sparc2.2  - first bug in early 0.90 metadata
 691          *   super-minor - change name of 0.90 metadata
 692          *   summaries - 'correct' any redundant data
 693          *   resync - mark array as dirty to trigger a resync.
 694          *   uuid - set new uuid - only 0.90 or 1.x
 695          *   name - change the name of the array (where supported)
 696          *   homehost - change which host this array is tied to.
 697          *   devicesize - If metadata is at start of device, change recorded
 698          *               device size to match actual device size
 699          *   byteorder - swap bytes for 0.90 metadata
 700          *
 701          *   force-one  - mark that device as uptodate, not old or failed.
 702          *   force-array - mark array as clean if it would not otherwise
 703          *               assemble
 704          *   assemble   - not sure how this is different from force-one...
 705          *   linear-grow-new - add a new device to a linear array, but don't
 706          *                   change the size: so superblock still matches
 707          *   linear-grow-update - now change the size of the array.
 708          *   writemostly - set the WriteMostly1 bit in the superblock devflags
 709          *   readwrite - clear the WriteMostly1 bit in the superblock devflags
 710          */
 711         int (*update_super)(struct supertype *st, struct mdinfo *info,
 712                             char *update,
 713                             char *devname, int verbose,
 714                             int uuid_set, char *homehost);
 715
 716         /* Create new metadata for new array as described.  This could
 717          * be a new container, or an array in a pre-existing container.
 718          * Also used to zero metadata prior to writing it to invalidate old
 719          * metadata.
 720          */
 721         int (*init_super)(struct supertype *st, mdu_array_info_t *info,
 722                           unsigned long long size, char *name,
 723                           char *homehost, int *uuid);
 724
 725         /* update the metadata to include new device, either at create or
 726          * when hot-adding a spare.
 727          */
 728         int (*add_to_super)(struct supertype *st, mdu_disk_info_t *dinfo,
 729                             int fd, char *devname);
 730         /* update the metadata to delete a device,
 731          * when hot-removing.
 732          */
 733         int (*remove_from_super)(struct supertype *st, mdu_disk_info_t *dinfo);
 734
 735         /* Write metadata to one device when fixing problems or adding
 736          * a new device.
 737          */
 738         int (*store_super)(struct supertype *st, int fd);
 739
 740         /*  Write all metadata for this array.
 741          */
 742         int (*write_init_super)(struct supertype *st);
 743         int (*compare_super)(struct supertype *st, struct supertype *tst);
 744         int (*load_super)(struct supertype *st, int fd, char *devname);
 745         int (*load_container)(struct supertype *st, int fd, char *devname);
 746         struct supertype * (*match_metadata_desc)(char *arg);
 747         __u64 (*avail_size)(struct supertype *st, __u64 size);
 748         unsigned long long (*min_acceptable_spare_size)(struct supertype *st);
 749         int (*add_internal_bitmap)(struct supertype *st, int *chunkp,
 750                                    int delay, int write_behind,
 751                                    unsigned long long size, int may_change, int major);
 752         void (*locate_bitmap)(struct supertype *st, int fd);
 753         int (*write_bitmap)(struct supertype *st, int fd);
 754         void (*free_super)(struct supertype *st);
 755
 756         /* validate_geometry is called with an st returned by
 757          * match_metadata_desc.
 758          * It should check that the geometry described in compatible with
 759          * the metadata type.  It will be called repeatedly as devices
 760          * added to validate changing size and new devices.  If there are
 761          * inter-device dependencies, it should record sufficient details
 762          * so these can be validated.
 763          * Both 'size' and '*freesize' are in sectors.  chunk is KiB.
 764          * Return value is:
 765          *  1: everything is OK
 766          *  0: not OK for some reason - if 'verbose', then error was reported.
 767          * -1: st->sb was NULL, 'subdev' is a member of a container of this
 768          *     types, but array is not acceptable for some reason
 769          *     message was reported even if verbose is 0.
 770          */
 771         int (*validate_geometry)(struct supertype *st, int level, int layout,
 772                                  int raiddisks,
 773                                  int *chunk, unsigned long long size,
 774                                  char *subdev, unsigned long long *freesize,
 775                                  int verbose);
 776
 777         struct mdinfo *(*container_content)(struct supertype *st, char *subarray);
 778         /* query the supertype for default geometry */
 779         void (*default_geometry)(struct supertype *st, int *level, int *layout, int *chunk); /* optional */
 780         /* Permit subarray's to be deleted from inactive containers */
 781         int (*kill_subarray)(struct supertype *st); /* optional */
 782         /* Permit subarray's to be modified */
 783         int (*update_subarray)(struct supertype *st, char *subarray,
 784                                char *update, struct mddev_ident *ident); /* optional */
 785         /* Check if reshape is supported for this external format.
 786          * st is obtained from super_by_fd() where st->subarray[0] is
 787          * initialized to indicate if reshape is being performed at the
 788          * container or subarray level
 789          */
 790 #define APPLY_METADATA_CHANGES          1
 791 #define ROLLBACK_METADATA_CHANGES       0
 792
 793         int (*reshape_super)(struct supertype *st,
 794                              unsigned long long size, int level,
 795                              int layout, int chunksize, int raid_disks,
 796                              int delta_disks, char *backup, char *dev,
 797                              int direction,
 798                              int verbose); /* optional */
 799         int (*manage_reshape)( /* optional */
 800                 int afd, struct mdinfo *sra, struct reshape *reshape,
 801                 struct supertype *st, unsigned long blocks,
 802                 int *fds, unsigned long long *offsets,
 803                 int dests, int *destfd, unsigned long long *destoffsets);
 804
 805 /* for mdmon */
 806         int (*open_new)(struct supertype *c, struct active_array *a,
 807                         char *inst);
 808
 809         /* Tell the metadata handler the current state of the array.
 810          * This covers whether it is known to be consistent (no pending writes)
 811          * and how far along a resync is known to have progressed
 812          * (in a->resync_start).
 813          * resync status is really irrelevant if the array is not consistent,
 814          * but some metadata (DDF!) have a place to record the distinction.
 815          * If 'consistent' is '2', then the array can mark it dirty if a
 816          * resync/recovery/whatever is required, or leave it clean if not.
 817          * Return value is 0 dirty (not consistent) and 1 if clean.
 818          * it is only really important if consistent is passed in as '2'.
 819          */
 820         int (*set_array_state)(struct active_array *a, int consistent);
 821
 822         /* When the state of a device might have changed, we call set_disk to
 823          * tell the metadata what the current state is.
 824          * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
 825          * transitions.
 826          * set_disk might be called when the state of the particular disk has
 827          * not in fact changed.
 828          */
 829         void (*set_disk)(struct active_array *a, int n, int state);
 830         void (*sync_metadata)(struct supertype *st);
 831         void (*process_update)(struct supertype *st,
 832                                struct metadata_update *update);
 833         void (*prepare_update)(struct supertype *st,
 834                                struct metadata_update *update);
 835
 836         /* activate_spare will check if the array is degraded and, if it
 837          * is, try to find some spare space in the container.
 838          * On success, it add appropriate updates (For process_update) to
 839          * to the 'updates' list and returns a list of 'mdinfo' identifying
 840          * the device, or devices as there might be multiple missing
 841          * devices and multiple spares available.
 842          */
 843         struct mdinfo *(*activate_spare)(struct active_array *a,
 844                                          struct metadata_update **updates);
 845         /*
 846          * Return statically allocated string that represents metadata specific
 847          * controller domain of the disk. The domain is used in disk domain
 848          * matching functions. Disks belong to the same domain if the they have
 849          * the same domain from mdadm.conf and belong the same metadata domain.
 850          * Returning NULL or not providing this handler means that metadata
 851          * does not distinguish the differences between disks that belong to
 852          * different controllers. They are in the domain specified by
 853          * configuration file (mdadm.conf).
 854          * In case when the metadata has the notion of domains based on disk
 855          * it shall return NULL for disks that do not belong to the controller
 856          * the supported domains. Such disks will form another domain and won't
 857          * be mixed with supported ones.
 858          */
 859         const char *(*get_disk_controller_domain)(const char *path);
 860
 861         /* for external backup area */
 862         int (*recover_backup)(struct supertype *st, struct mdinfo *info);
 863
 864         int swapuuid; /* true if uuid is bigending rather than hostendian */
 865         int external;
 866         const char *name; /* canonical metadata name */
 867 } *superlist[];
 868
 869 extern struct superswitch super0, super1;
 870 extern struct superswitch super_imsm, super_ddf;
 871 extern struct superswitch mbr, gpt;
 872
 873 struct metadata_update {
 874         int     len;
 875         char    *buf;
 876         void    *space; /* allocated space that monitor will use */
 877         void    **space_list; /* list of allocated spaces that monitor can
 878                                * use or that it returned.
 879                                */
 880         struct metadata_update *next;
 881 };
 882
 883 /* A supertype holds a particular collection of metadata.
 884  * It identifies the metadata type by the superswitch, and the particular
 885  * sub-version of that metadata type.
 886  * metadata read in or created is stored in 'sb' and 'info'.
 887  * There are also fields used by mdmon to track containers.
 888  *
 889  * A supertype may refer to:
 890  *   Just an array, possibly in a container
 891  *   A container, not identifying any particular array
 892  *   Info read from just one device, not yet fully describing the array/container.
 893  *
 894  *
 895  * A supertype is created by:
 896  *   super_by_fd
 897  *   guess_super
 898  *   dup_super
 899  */
 900 struct supertype {
 901         struct superswitch *ss;
 902         int minor_version;
 903         int max_devs;
 904         int container_dev;    /* devnum of container */
 905         void *sb;
 906         void *info;
 907         int ignore_hw_compat; /* used to inform metadata handlers that it should ignore
 908                                  HW/firmware related incompatability to load metadata.
 909                                  Used when examining metadata to display content of disk
 910                                  when user has no hw/firmare compatible system.
 911                               */
 912         struct metadata_update *updates;
 913         struct metadata_update **update_tail;
 914
 915         /* extra stuff used by mdmon */
 916         struct active_array *arrays;
 917         int sock; /* listen to external programs */
 918         int devnum;
 919         char *devname; /* e.g. md0.  This appears in metadata_verison:
 920                         *  external:/md0/12
 921                         */
 922         int devcnt;
 923         int retry_soon;
 924
 925         struct mdinfo *devs;
 926
 927 };
 928
 929 extern struct supertype *super_by_fd(int fd, char **subarray);
 930 enum guess_types { guess_any, guess_array, guess_partitions };
 931 extern struct supertype *guess_super_type(int fd, enum guess_types guess_type);
 932 static inline struct supertype *guess_super(int fd) {
 933         return guess_super_type(fd, guess_any);
 934 }
 935 extern struct supertype *dup_super(struct supertype *st);
 936 extern int get_dev_size(int fd, char *dname, unsigned long long *sizep);
 937 extern int must_be_container(int fd);
 938 extern int dev_size_from_id(dev_t id, unsigned long long *size);
 939 extern void get_one_disk(int mdfd, mdu_array_info_t *ainf,
 940                          mdu_disk_info_t *disk);
 941 void wait_for(char *dev, int fd);
 942
 943 /*
 944  * Data structures for policy management.
 945  * Each device can have a policy structure that lists
 946  * various name/value pairs each possibly with a metadata associated.
 947  * The policy list is sorted by name/value/metadata
 948  */
 949 struct dev_policy {
 950         struct dev_policy *next;
 951         char *name;     /* None of these strings are allocated.  They are
 952                          * all just references to strings which are known
 953                          * to exist elsewhere.
 954                          * name and metadata can be compared by address equality.
 955                          */
 956         const char *metadata;
 957         const char *value;
 958 };
 959
 960 extern char pol_act[], pol_domain[], pol_metadata[], pol_auto[];
 961
 962 /* iterate over the sublist starting at list, having the same
 963  * 'name' as 'list', and matching the given metadata (Where
 964  * NULL matches anything
 965  */
 966 #define pol_for_each(item, list, _metadata)                             \
 967         for (item = list;                                               \
 968              item && item->name == list->name;                          \
 969              item = item->next)                                         \
 970                 if (!(!_metadata || !item->metadata || _metadata == item->metadata)) \
 971                         ; else
 972
 973 /*
 974  * policy records read from mdadm are largely just name-value pairs.
 975  * The names are constants, not strdupped
 976  */
 977 struct pol_rule {
 978         struct pol_rule *next;
 979         char *type;     /* rule_policy or rule_part */
 980         struct rule {
 981                 struct rule *next;
 982                 char *name;
 983                 char *value;
 984                 char *dups; /* duplicates of 'value' with a partNN appended */
 985         } *rule;
 986 };
 987
 988 extern char rule_policy[], rule_part[];
 989 extern char rule_path[], rule_type[];
 990 extern char type_part[], type_disk[];
 991
 992 extern void policyline(char *line, char *type);
 993 extern void policy_add(char *type, ...);
 994 extern void policy_free(void);
 995
 996 extern struct dev_policy *path_policy(char *path, char *type);
 997 extern struct dev_policy *disk_policy(struct mdinfo *disk);
 998 extern struct dev_policy *devnum_policy(int dev);
 999 extern void dev_policy_free(struct dev_policy *p);
1000
1001 //extern void pol_new(struct dev_policy **pol, char *name, char *val, char *metadata);
1002 extern void pol_add(struct dev_policy **pol, char *name, char *val, char *metadata);
1003 extern struct dev_policy *pol_find(struct dev_policy *pol, char *name);
1004
1005 enum policy_action {
1006         act_default,
1007         act_include,
1008         act_re_add,
1009         act_spare,      /* This only applies to bare devices */
1010         act_spare_same_slot, /* this allows non-bare devices,
1011                               * but only if recent removal */
1012         act_force_spare, /* this allow non-bare devices in any case */
1013         act_err
1014 };
1015
1016 extern int policy_action_allows(struct dev_policy *plist, const char *metadata,
1017                                 enum policy_action want);
1018 extern int disk_action_allows(struct mdinfo *disk, const char *metadata,
1019                               enum policy_action want);
1020
1021 struct domainlist {
1022         struct domainlist *next;
1023         const char *dom;
1024 };
1025
1026 extern int domain_test(struct domainlist *dom, struct dev_policy *pol,
1027                        const char *metadata);
1028 extern struct domainlist *domain_from_array(struct mdinfo *mdi,
1029                                             const char *metadata);
1030 extern void domainlist_add_dev(struct domainlist **dom, int devnum,
1031                                const char *metadata);
1032 extern void domain_free(struct domainlist *dl);
1033 extern void domain_merge(struct domainlist **domp, struct dev_policy *pol,
1034                          const char *metadata);
1035 void domain_add(struct domainlist **domp, char *domain);
1036
1037 extern void policy_save_path(char *id_path, struct map_ent *array);
1038 extern int policy_check_path(struct mdinfo *disk, struct map_ent *array);
1039
1040 #if __GNUC__ < 3
1041 struct stat64;
1042 #endif
1043
1044 #define HAVE_NFTW  we assume
1045 #define HAVE_FTW
1046
1047 #ifdef __UCLIBC__
1048 # include <features.h>
1049 # ifndef __UCLIBC_HAS_LFS__
1050 #  define lseek64 lseek
1051 # endif
1052 # ifndef  __UCLIBC_HAS_FTW__
1053 #  undef HAVE_FTW
1054 #  undef HAVE_NFTW
1055 # endif
1056 #endif
1057
1058 #ifdef __dietlibc__
1059 # undef HAVE_NFTW
1060 #endif
1061
1062 #if defined(__KLIBC__)
1063 # undef HAVE_NFTW
1064 # undef HAVE_FTW
1065 #endif
1066
1067 #ifndef HAVE_NFTW
1068 # define FTW_PHYS 1
1069 # ifndef HAVE_FTW
1070   struct FTW {};
1071 # endif
1072 #endif
1073
1074 #ifdef HAVE_FTW
1075 # include <ftw.h>
1076 #endif
1077
1078 extern int add_dev(const char *name, const struct stat *stb, int flag, struct FTW *s);
1079
1080
1081 extern int Manage_ro(char *devname, int fd, int readonly);
1082 extern int Manage_runstop(char *devname, int fd, int runstop, int quiet,
1083                           int will_retry);
1084 extern int Manage_subdevs(char *devname, int fd,
1085                           struct mddev_dev *devlist, int verbose, int test,
1086                           char *update, int force);
1087 extern int autodetect(void);
1088 extern int Grow_Add_device(char *devname, int fd, char *newdev);
1089 extern int Grow_addbitmap(char *devname, int fd,
1090                           struct context *c, struct shape *s);
1091 extern int Grow_reshape(char *devname, int fd,
1092                         struct mddev_dev *devlist,
1093                         struct context *c, struct shape *s);
1094 extern int Grow_restart(struct supertype *st, struct mdinfo *info,
1095                         int *fdlist, int cnt, char *backup_file, int verbose);
1096 extern int Grow_continue(int mdfd, struct supertype *st,
1097                          struct mdinfo *info, char *backup_file,
1098                          int freeze_reshape);
1099
1100 extern int restore_backup(struct supertype *st,
1101                           struct mdinfo *content,
1102                           int working_disks,
1103                           int spares,
1104                           char *backup_file,
1105                           int verbose);
1106 extern int Grow_continue_command(char *devname, int fd,
1107                                  char *backup_file, int verbose);
1108
1109 extern int Assemble(struct supertype *st, char *mddev,
1110                     struct mddev_ident *ident,
1111                     struct mddev_dev *devlist,
1112                     struct context *c);
1113
1114 extern int Build(char *mddev, struct mddev_dev *devlist,
1115                  struct shape *s, struct context *c);
1116
1117 extern int Create(struct supertype *st, char *mddev,
1118                   char *name, int *uuid,
1119                   int subdevs, struct mddev_dev *devlist,
1120                   struct shape *s,
1121                   struct context *c);
1122
1123 extern int Detail(char *dev, struct context *c);
1124 extern int Detail_Platform(struct superswitch *ss, int scan, int verbose);
1125 extern int Query(char *dev);
1126 extern int Examine(struct mddev_dev *devlist, struct context *c,
1127                    struct supertype *forcest);
1128 extern int Monitor(struct mddev_dev *devlist,
1129                    char *mailaddr, char *alert_cmd,
1130                    struct context *c,
1131                    int daemonise, int oneshot,
1132                    int dosyslog, char *pidfile, int increments,
1133                    int share);
1134
1135 extern int Kill(char *dev, struct supertype *st, int force, int verbose, int noexcl);
1136 extern int Kill_subarray(char *dev, char *subarray, int verbose);
1137 extern int Update_subarray(char *dev, char *subarray, char *update, struct mddev_ident *ident, int quiet);
1138 extern int Wait(char *dev);
1139 extern int WaitClean(char *dev, int sock, int verbose);
1140
1141 extern int Incremental(char *devname, struct context *c,
1142                        struct supertype *st);
1143 extern void RebuildMap(void);
1144 extern int IncrementalScan(int verbose);
1145 extern int IncrementalRemove(char *devname, char *path, int verbose);
1146 extern int CreateBitmap(char *filename, int force, char uuid[16],
1147                         unsigned long chunksize, unsigned long daemon_sleep,
1148                         unsigned long write_behind,
1149                         unsigned long long array_size,
1150                         int major);
1151 extern int ExamineBitmap(char *filename, int brief, struct supertype *st);
1152 extern int Write_rules(char *rule_name);
1153 extern int bitmap_update_uuid(int fd, int *uuid, int swap);
1154 extern unsigned long bitmap_sectors(struct bitmap_super_s *bsb);
1155
1156 extern int md_get_version(int fd);
1157 extern int get_linux_version(void);
1158 extern int mdadm_version(char *version);
1159 extern unsigned long long parse_size(char *size);
1160 extern int parse_uuid(char *str, int uuid[4]);
1161 extern int parse_layout_10(char *layout);
1162 extern int parse_layout_faulty(char *layout);
1163 extern long parse_num(char *num);
1164 extern int check_ext2(int fd, char *name);
1165 extern int check_reiser(int fd, char *name);
1166 extern int check_raid(int fd, char *name);
1167 extern int check_partitions(int fd, char *dname,
1168                             unsigned long long freesize,
1169                             unsigned long long size);
1170
1171 extern int get_mdp_major(void);
1172 extern int dev_open(char *dev, int flags);
1173 extern int open_dev(int devnum);
1174 extern int open_dev_flags(int devnum, int flags);
1175 extern int open_dev_excl(int devnum);
1176 extern int is_standard(char *dev, int *nump);
1177 extern int same_dev(char *one, char *two);
1178
1179 extern int parse_auto(char *str, char *msg, int config);
1180 extern struct mddev_ident *conf_get_ident(char *dev);
1181 extern struct mddev_dev *conf_get_devs(void);
1182 extern int conf_test_dev(char *devname);
1183 extern int conf_test_metadata(const char *version, struct dev_policy *pol, int is_homehost);
1184 extern struct createinfo *conf_get_create_info(void);
1185 extern void set_conffile(char *file);
1186 extern char *conf_get_mailaddr(void);
1187 extern char *conf_get_mailfrom(void);
1188 extern char *conf_get_program(void);
1189 extern char *conf_get_homehost(int *require_homehostp);
1190 extern char *conf_line(FILE *file);
1191 extern char *conf_word(FILE *file, int allow_key);
1192 extern int conf_name_is_free(char *name);
1193 extern int conf_verify_devnames(struct mddev_ident *array_list);
1194 extern int devname_matches(char *name, char *match);
1195 extern struct mddev_ident *conf_match(struct supertype *st,
1196                                       struct mdinfo *info,
1197                                       char *devname,
1198                                       int verbose, int *rvp);
1199 extern int experimental(void);
1200
1201 extern void free_line(char *line);
1202 extern int match_oneof(char *devices, char *devname);
1203 extern void uuid_from_super(int uuid[4], mdp_super_t *super);
1204 extern const int uuid_zero[4];
1205 extern int same_uuid(int a[4], int b[4], int swapuuid);
1206 extern void copy_uuid(void *a, int b[4], int swapuuid);
1207 extern char *__fname_from_uuid(int id[4], int swap, char *buf, char sep);
1208 extern char *fname_from_uuid(struct supertype *st,
1209                              struct mdinfo *info, char *buf, char sep);
1210 extern unsigned long calc_csum(void *super, int bytes);
1211 extern int enough(int level, int raid_disks, int layout, int clean,
1212                    char *avail);
1213 extern int enough_fd(int fd);
1214 extern int ask(char *mesg);
1215 extern unsigned long long get_component_size(int fd);
1216 extern void remove_partitions(int fd);
1217 extern int test_partition(int fd);
1218 extern int test_partition_from_id(dev_t id);
1219 extern int get_data_disks(int level, int layout, int raid_disks);
1220 extern unsigned long long calc_array_size(int level, int raid_disks, int layout,
1221                                    int chunksize, unsigned long long devsize);
1222 extern int flush_metadata_updates(struct supertype *st);
1223 extern void append_metadata_update(struct supertype *st, void *buf, int len);
1224 extern int assemble_container_content(struct supertype *st, int mdfd,
1225                                       struct mdinfo *content,
1226                                       struct context *c,
1227                                       char *chosen_name);
1228 extern struct mdinfo *container_choose_spares(struct supertype *st,
1229                                               unsigned long long min_size,
1230                                               struct domainlist *domlist,
1231                                               char *spare_group,
1232                                               const char *metadata, int get_one);
1233 extern int move_spare(char *from_devname, char *to_devname, dev_t devid);
1234 extern int add_disk(int mdfd, struct supertype *st,
1235                     struct mdinfo *sra, struct mdinfo *info);
1236 extern int remove_disk(int mdfd, struct supertype *st,
1237                        struct mdinfo *sra, struct mdinfo *info);
1238 extern int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info);
1239 unsigned long long min_recovery_start(struct mdinfo *array);
1240
1241 extern char *human_size(long long bytes);
1242 extern char *human_size_brief(long long bytes);
1243 extern void print_r10_layout(int layout);
1244
1245 #define NoMdDev (1<<23)
1246 extern int find_free_devnum(int use_partitions);
1247
1248 extern void put_md_name(char *name);
1249 extern char *get_md_name(int dev);
1250
1251 extern char DefaultConfFile[];
1252
1253 extern int create_mddev(char *dev, char *name, int autof, int trustworthy,
1254                         char *chosen);
1255 /* values for 'trustworthy' */
1256 #define LOCAL   1
1257 #define LOCAL_ANY 10
1258 #define FOREIGN 2
1259 #define METADATA 3
1260 extern int open_mddev(char *dev, int report_errors);
1261 extern int open_container(int fd);
1262 extern int is_container_member(struct mdstat_ent *ent, char *devname);
1263 extern int is_subarray_active(char *subarray, char *devname);
1264 extern int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet);
1265 extern struct superswitch *version_to_superswitch(char *vers);
1266
1267 extern int mdmon_running(int devnum);
1268 extern int mdmon_pid(int devnum);
1269 extern int check_env(char *name);
1270 extern __u32 random32(void);
1271 extern int start_mdmon(int devnum);
1272
1273 extern int child_monitor(int afd, struct mdinfo *sra, struct reshape *reshape,
1274                          struct supertype *st, unsigned long stripes,
1275                          int *fds, unsigned long long *offsets,
1276                          int dests, int *destfd, unsigned long long *destoffsets);
1277 void abort_reshape(struct mdinfo *sra);
1278
1279 extern char *devnum2devname(int num);
1280 extern void fmt_devname(char *name, int num);
1281 extern int devname2devnum(char *name);
1282 extern int stat2devnum(struct stat *st);
1283 extern int fd2devnum(int fd);
1284
1285 static inline int dev2major(int d)
1286 {
1287         if (d >= 0)
1288                 return MD_MAJOR;
1289         else
1290                 return get_mdp_major();
1291 }
1292
1293 static inline int dev2minor(int d)
1294 {
1295         if (d >= 0)
1296                 return d;
1297         return (-1-d) << MdpMinorShift;
1298 }
1299
1300 #define _ROUND_UP(val, base)    (((val) + (base) - 1) & ~(base - 1))
1301 #define ROUND_UP(val, base)     _ROUND_UP(val, (typeof(val))(base))
1302 #define ROUND_UP_PTR(ptr, base) ((typeof(ptr)) \
1303                                  (ROUND_UP((unsigned long)(ptr), base)))
1304
1305 static inline int is_subarray(char *vers)
1306 {
1307         /* The version string for a 'subarray' (an array in a container)
1308          * is
1309          *    /containername/componentname    for normal read-write arrays
1310          *    -containername/componentname    for arrays which mdmon must not
1311          *                                    reconfigure.  They might be read-only
1312          *                                    or might be undergoing reshape etc.
1313          * containername is e.g. md0, md_d1
1314          * componentname is dependant on the metadata. e.g. '1' 'S1' ...
1315          */
1316         return (*vers == '/' || *vers == '-');
1317 }
1318
1319 static inline char *to_subarray(struct mdstat_ent *ent, char *container)
1320 {
1321         return &ent->metadata_version[10+strlen(container)+1];
1322 }
1323
1324 #ifdef DEBUG
1325 #define dprintf(fmt, arg...) \
1326         fprintf(stderr, fmt, ##arg)
1327 #else
1328 #define dprintf(fmt, arg...) \
1329         ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
1330 #endif
1331 #include <assert.h>
1332 #include <stdarg.h>
1333 static inline int xasprintf(char **strp, const char *fmt, ...) {
1334         va_list ap;
1335         int ret;
1336         va_start(ap, fmt);
1337         ret = vasprintf(strp, fmt, ap);
1338         va_end(ap);
1339         assert(ret >= 0);
1340         return ret;
1341 }
1342
1343 #define pr_err(fmt ...) fprintf(stderr, Name ": " fmt)
1344 #define cont_err(fmt ...) fprintf(stderr, "       " fmt)
1345
1346 void *xmalloc(size_t len);
1347 void *xrealloc(void *ptr, size_t len);
1348 void *xcalloc(size_t num, size_t size);
1349 char *xstrdup(const char *str);
1350
1351 #define LEVEL_MULTIPATH         (-4)
1352 #define LEVEL_LINEAR            (-1)
1353 #define LEVEL_FAULTY            (-5)
1354
1355 /* kernel module doesn't know about these */
1356 #define LEVEL_CONTAINER         (-100)
1357 #define LEVEL_UNSUPPORTED       (-200)
1358
1359
1360 /* faulty stuff */
1361
1362 #define WriteTransient  0
1363 #define ReadTransient   1
1364 #define WritePersistent 2
1365 #define ReadPersistent  3
1366 #define WriteAll        4 /* doesn't go to device */
1367 #define ReadFixable     5
1368 #define Modes   6
1369
1370 #define ClearErrors     31
1371 #define ClearFaults     30
1372
1373 #define AllPersist      100 /* internal use only */
1374 #define NoPersist       101
1375
1376 #define ModeMask        0x1f
1377 #define ModeShift       5
1378
1379
1380 #ifdef __TINYC__
1381 #undef minor
1382 #undef major
1383 #undef makedev
1384 #define minor(x) ((x)&0xff)
1385 #define major(x) (((x)>>8)&0xff)
1386 #define makedev(M,m) (((M)<<8) | (m))
1387 #endif
1388
1389 /* for raid4/5/6 */
1390 #define ALGORITHM_LEFT_ASYMMETRIC       0
1391 #define ALGORITHM_RIGHT_ASYMMETRIC      1
1392 #define ALGORITHM_LEFT_SYMMETRIC        2
1393 #define ALGORITHM_RIGHT_SYMMETRIC       3
1394
1395 /* Define non-rotating (raid4) algorithms.  These allow
1396  * conversion of raid4 to raid5.
1397  */
1398 #define ALGORITHM_PARITY_0              4 /* P or P,Q are initial devices */
1399 #define ALGORITHM_PARITY_N              5 /* P or P,Q are final devices. */
1400
1401 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1402  * Firstly, the exact positioning of the parity block is slightly
1403  * different between the 'LEFT_*' modes of md and the "_N_*" modes
1404  * of DDF.
1405  * Secondly, or order of datablocks over which the Q syndrome is computed
1406  * is different.
1407  * Consequently we have different layouts for DDF/raid6 than md/raid6.
1408  * These layouts are from the DDFv1.2 spec.
1409  * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1410  * leaves RLQ=3 as 'Vendor Specific'
1411  */
1412
1413 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1414 #define ALGORITHM_ROTATING_N_RESTART    9 /* DDF PRL=6 RLQ=2 */
1415 #define ALGORITHM_ROTATING_N_CONTINUE   10 /*DDF PRL=6 RLQ=3 */
1416
1417
1418 /* For every RAID5 algorithm we define a RAID6 algorithm
1419  * with exactly the same layout for data and parity, and
1420  * with the Q block always on the last device (N-1).
1421  * This allows trivial conversion from RAID5 to RAID6
1422  */
1423 #define ALGORITHM_LEFT_ASYMMETRIC_6     16
1424 #define ALGORITHM_RIGHT_ASYMMETRIC_6    17
1425 #define ALGORITHM_LEFT_SYMMETRIC_6      18
1426 #define ALGORITHM_RIGHT_SYMMETRIC_6     19
1427 #define ALGORITHM_PARITY_0_6            20
1428 #define ALGORITHM_PARITY_N_6            ALGORITHM_PARITY_N
1429
1430 /* Define PATH_MAX in case we don't use glibc or standard library does
1431  * not have PATH_MAX defined. Assume max path length is 4K characters.
1432  */
1433 #ifndef PATH_MAX
1434 #define PATH_MAX        4096
1435 #endif
1436
1437 #define RESYNC_NONE -1
1438 #define RESYNC_DELAYED -2
1439 #define RESYNC_PENDING -3
1440 #define RESYNC_UNKNOWN -4
1441
1442 /* When using "GET_DISK_INFO" it isn't certain how high
1443  * we need to check.  So we impose an absolute limit of
1444  * MAX_DISKS.  This needs to be much more than the largest
1445  * number of devices any metadata can support.  Currently
1446  * v1.x can support 1920
1447  */
1448 #define MAX_DISKS       4096
1449
1450 /* Sometimes the 'size' value passed needs to mean "Maximum".
1451  * In those cases with use MAX_SIZE
1452  */
1453 #define MAX_SIZE        1
1454 extern int __offroot;