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_BITMAP_CHUNK 4096
  68 #define DEFAULT_BITMAP_DELAY 5
  69 #define DEFAULT_MAX_WRITE_BEHIND 256
  70
  71 /* VAR_RUN is where pid and socket files used for communicating
  72  * with mdmon normally live.  It should be /var/run, but if
  73  * it is too hard to remount /var/run as read-only rather than
  74  * unmounting it at shutdown time, then it should be
  75  * redefined to some place that comfortably persists until
  76  * final shutdown, possibly in /dev if that is a tmpfs.
  77  * Note: VAR_RUN does not need to be writable at shutdown,
  78  * only during boot when "mdmon --takeover" is run.
  79  */
  80 #ifndef VAR_RUN
  81 #define VAR_RUN "/var/run/mdadm"
  82 #endif /* VAR_RUN */
  83 /* ALT_RUN should be somewhere that persists across the pivotroot
  84  * from early boot to late boot.
  85  * If you don't have /lib/init/rw you might want to use /dev/.something
  86  */
  87 #ifndef ALT_RUN
  88 #define ALT_RUN "/lib/init/rw/mdadm"
  89 #endif /* ALT_RUN */
  90 #ifndef ALT_MAPFILE
  91 #define ALT_MAPFILE "map"
  92 #endif /* ALT_MAPFILE */
  93
  94 #include        "md_u.h"
  95 #include        "md_p.h"
  96 #include        "bitmap.h"
  97 #include        "msg.h"
  98
  99 #include <endian.h>
 100 /* Redhat don't like to #include <asm/byteorder.h>, and
 101  * some time include <linux/byteorder/xxx_endian.h> isn't enough,
 102  * and there is no standard conversion function so... */
 103 /* And dietlibc doesn't think byteswap is ok, so.. */
 104 /*  #include <byteswap.h> */
 105 #define bswap_16(x) (((x) & 0x00ffU) << 8 | \
 106                      ((x) & 0xff00U) >> 8)
 107 #define bswap_32(x) (((x) & 0x000000ffU) << 24 | \
 108                      ((x) & 0xff000000U) >> 24 | \
 109                      ((x) & 0x0000ff00U) << 8  | \
 110                      ((x) & 0x00ff0000U) >> 8)
 111 #define bswap_64(x) (((x) & 0x00000000000000ffULL) << 56 | \
 112                      ((x) & 0xff00000000000000ULL) >> 56 | \
 113                      ((x) & 0x000000000000ff00ULL) << 40 | \
 114                      ((x) & 0x00ff000000000000ULL) >> 40 | \
 115                      ((x) & 0x0000000000ff0000ULL) << 24 | \
 116                      ((x) & 0x0000ff0000000000ULL) >> 24 | \
 117                      ((x) & 0x00000000ff000000ULL) << 8 | \
 118                      ((x) & 0x000000ff00000000ULL) >> 8)
 119
 120 #if !defined(__KLIBC__)
 121 #if BYTE_ORDER == LITTLE_ENDIAN
 122 #define __cpu_to_le16(_x) (_x)
 123 #define __cpu_to_le32(_x) (_x)
 124 #define __cpu_to_le64(_x) (_x)
 125 #define __le16_to_cpu(_x) (_x)
 126 #define __le32_to_cpu(_x) (_x)
 127 #define __le64_to_cpu(_x) (_x)
 128
 129 #define __cpu_to_be16(_x) bswap_16(_x)
 130 #define __cpu_to_be32(_x) bswap_32(_x)
 131 #define __cpu_to_be64(_x) bswap_64(_x)
 132 #define __be16_to_cpu(_x) bswap_16(_x)
 133 #define __be32_to_cpu(_x) bswap_32(_x)
 134 #define __be64_to_cpu(_x) bswap_64(_x)
 135 #elif BYTE_ORDER == BIG_ENDIAN
 136 #define __cpu_to_le16(_x) bswap_16(_x)
 137 #define __cpu_to_le32(_x) bswap_32(_x)
 138 #define __cpu_to_le64(_x) bswap_64(_x)
 139 #define __le16_to_cpu(_x) bswap_16(_x)
 140 #define __le32_to_cpu(_x) bswap_32(_x)
 141 #define __le64_to_cpu(_x) bswap_64(_x)
 142
 143 #define __cpu_to_be16(_x) (_x)
 144 #define __cpu_to_be32(_x) (_x)
 145 #define __cpu_to_be64(_x) (_x)
 146 #define __be16_to_cpu(_x) (_x)
 147 #define __be32_to_cpu(_x) (_x)
 148 #define __be64_to_cpu(_x) (_x)
 149 #else
 150 #  error "unknown endianness."
 151 #endif
 152 #endif /* __KLIBC__ */
 153
 154
 155 /*
 156  * min()/max()/clamp() macros that also do
 157  * strict type-checking.. See the
 158  * "unnecessary" pointer comparison.
 159  */
 160 #define min(x, y) ({                            \
 161         typeof(x) _min1 = (x);                  \
 162         typeof(y) _min2 = (y);                  \
 163         (void) (&_min1 == &_min2);              \
 164         _min1 < _min2 ? _min1 : _min2; })
 165
 166 #define max(x, y) ({                            \
 167         typeof(x) _max1 = (x);                  \
 168         typeof(y) _max2 = (y);                  \
 169         (void) (&_max1 == &_max2);              \
 170         _max1 > _max2 ? _max1 : _max2; })
 171
 172 /* general information that might be extracted from a superblock */
 173 struct mdinfo {
 174         mdu_array_info_t        array;
 175         mdu_disk_info_t         disk;
 176         __u64                   events;
 177         int                     uuid[4];
 178         char                    name[33];
 179         unsigned long long      data_offset;
 180         unsigned long long      component_size; /* same as array.size, except in
 181                                                  * sectors and up to 64bits.
 182                                                  */
 183         unsigned long long      custom_array_size; /* size for non-default sized
 184                                                     * arrays (in sectors)
 185                                                     */
 186         int                     reshape_active;
 187         unsigned long long      reshape_progress;
 188         union {
 189                 unsigned long long resync_start; /* per-array resync position */
 190                 unsigned long long recovery_start; /* per-device rebuild position */
 191                 #define MaxSector  (~0ULL) /* resync/recovery complete position */
 192         };
 193         unsigned long           safe_mode_delay; /* ms delay to mark clean */
 194         int                     new_level, delta_disks, new_layout, new_chunk;
 195         int                     errors;
 196         int                     cache_size; /* size of raid456 stripe cache*/
 197         int                     mismatch_cnt;
 198         char                    text_version[50];
 199         void                    *update_private; /* for passing metadata-format
 200                                                   * specific update data
 201                                                   * between successive calls to
 202                                                   * update_super()
 203                                                   */
 204
 205         int container_member; /* for assembling external-metatdata arrays
 206                                * This is to be used internally by metadata
 207                                * handler only */
 208         int container_enough; /* flag external handlers can set to
 209                                * indicate that subarrays have not enough (-1),
 210                                * enough to start (0), or all expected disks (1) */
 211         char            sys_name[20];
 212         struct mdinfo *devs;
 213         struct mdinfo *next;
 214
 215         /* Device info for mdmon: */
 216         int recovery_fd;
 217         int state_fd;
 218         #define DS_FAULTY       1
 219         #define DS_INSYNC       2
 220         #define DS_WRITE_MOSTLY 4
 221         #define DS_SPARE        8
 222         #define DS_BLOCKED      16
 223         #define DS_REMOVE       1024
 224         #define DS_UNBLOCK      2048
 225         int prev_state, curr_state, next_state;
 226
 227 };
 228
 229 struct createinfo {
 230         int     uid;
 231         int     gid;
 232         int     autof;
 233         int     mode;
 234         int     symlinks;
 235         struct supertype *supertype;
 236 };
 237
 238 #define Name "mdadm"
 239
 240 enum mode {
 241         ASSEMBLE=1,
 242         BUILD,
 243         CREATE,
 244         MANAGE,
 245         MISC,
 246         MONITOR,
 247         GROW,
 248         INCREMENTAL,
 249         AUTODETECT,
 250 };
 251
 252 extern char short_options[];
 253 extern char short_bitmap_options[];
 254 extern char short_bitmap_auto_options[];
 255 extern struct option long_options[];
 256 extern char Version[], Usage[], Help[], OptionHelp[],
 257         Help_create[], Help_build[], Help_assemble[], Help_grow[],
 258         Help_incr[],
 259         Help_manage[], Help_misc[], Help_monitor[], Help_config[];
 260
 261 /* for option that don't have short equivilents, we assign arbitrary
 262  * small numbers.  '1' means an undecorated option, so we start at '2'.
 263  * (note we must stop before we get to 65 i.e. 'A')
 264  */
 265 enum special_options {
 266         AssumeClean = 2,
 267         BitmapChunk,
 268         WriteBehind,
 269         ReAdd,
 270         NoDegraded,
 271         Sparc22,
 272         BackupFile, /* 8 */
 273         HomeHost,
 274         AutoHomeHost,
 275         Symlinks,
 276         AutoDetect,
 277         Waitclean,
 278         DetailPlatform,
 279         KillSubarray,
 280         UpdateSubarray, /* 16 */
 281 };
 282
 283 /* structures read from config file */
 284 /* List of mddevice names and identifiers
 285  * Identifiers can be:
 286  *    uuid=128-hex-uuid
 287  *    super-minor=decimal-minor-number-from-superblock
 288  *    devices=comma,separated,list,of,device,names,with,wildcards
 289  *
 290  * If multiple fields are present, the intersection of all matching
 291  * devices is considered
 292  */
 293 #define UnSet (0xfffe)
 294 typedef struct mddev_ident_s {
 295         char    *devname;
 296
 297         int     uuid_set;
 298         int     uuid[4];
 299         char    name[33];
 300
 301         unsigned int super_minor;
 302
 303         char    *devices;       /* comma separated list of device
 304                                  * names with wild cards
 305                                  */
 306         int     level;
 307         unsigned int raid_disks;
 308         unsigned int spare_disks;
 309         struct supertype *st;
 310         int     autof;          /* 1 for normal, 2 for partitioned */
 311         char    *spare_group;
 312         char    *bitmap_file;
 313         int     bitmap_fd;
 314
 315         char    *container;     /* /dev/whatever name of container, or
 316                                  * uuid of container.  You would expect
 317                                  * this to be the 'devname' or UUID
 318                                  * of some other entry.
 319                                  */
 320         char    *member;        /* subarray within a container */
 321
 322         struct mddev_ident_s *next;
 323         union {
 324                 /* fields needed by different users of this structure */
 325                 int assembled;  /* set when assembly succeeds */
 326         };
 327 } *mddev_ident_t;
 328
 329 /* List of device names - wildcards expanded */
 330 typedef struct mddev_dev_s {
 331         char *devname;
 332         char disposition;       /* 'a' for add, 'r' for remove, 'f' for fail.
 333                                  * Not set for names read from .config
 334                                  */
 335         char writemostly;       /* 1 for 'set writemostly', 2 for 'clear writemostly' */
 336         char re_add;
 337         char used;              /* set when used */
 338         struct mdinfo *content; /* If devname is a container, this might list
 339                                  * the remaining member arrays. */
 340         struct mddev_dev_s *next;
 341 } *mddev_dev_t;
 342
 343 typedef struct mapping {
 344         char *name;
 345         int num;
 346 } mapping_t;
 347
 348
 349 struct mdstat_ent {
 350         char            *dev;
 351         int             devnum;
 352         int             active;
 353         char            *level;
 354         char            *pattern; /* U or up, _ for down */
 355         int             percent; /* -1 if no resync */
 356         int             resync; /* 1 if resync, 0 if recovery */
 357         int             devcnt;
 358         int             raid_disks;
 359         int             chunk_size;
 360         char *          metadata_version;
 361         struct mdstat_ent *next;
 362 };
 363
 364 extern struct mdstat_ent *mdstat_read(int hold, int start);
 365 extern void free_mdstat(struct mdstat_ent *ms);
 366 extern void mdstat_wait(int seconds);
 367 extern void mdstat_wait_fd(int fd, const sigset_t *sigmask);
 368 extern int mddev_busy(int devnum);
 369
 370 struct map_ent {
 371         struct map_ent *next;
 372         int     devnum;
 373         char    metadata[20];
 374         int     uuid[4];
 375         int     bad;
 376         char    *path;
 377 };
 378 extern int map_update(struct map_ent **mpp, int devnum, char *metadata,
 379                       int uuid[4], char *path);
 380 extern void map_remove(struct map_ent **map, int devnum);
 381 extern struct map_ent *map_by_uuid(struct map_ent **map, int uuid[4]);
 382 extern struct map_ent *map_by_devnum(struct map_ent **map, int devnum);
 383 extern struct map_ent *map_by_name(struct map_ent **map, char *name);
 384 extern void map_read(struct map_ent **melp);
 385 extern int map_write(struct map_ent *mel);
 386 extern void map_delete(struct map_ent **mapp, int devnum);
 387 extern void map_free(struct map_ent *map);
 388 extern void map_add(struct map_ent **melp,
 389                     int devnum, char *metadata, int uuid[4], char *path);
 390 extern int map_lock(struct map_ent **melp);
 391 extern void map_unlock(struct map_ent **melp);
 392
 393 /* various details can be requested */
 394 enum sysfs_read_flags {
 395         GET_LEVEL       = (1 << 0),
 396         GET_LAYOUT      = (1 << 1),
 397         GET_COMPONENT   = (1 << 2),
 398         GET_CHUNK       = (1 << 3),
 399         GET_CACHE       = (1 << 4),
 400         GET_MISMATCH    = (1 << 5),
 401         GET_VERSION     = (1 << 6),
 402         GET_DISKS       = (1 << 7),
 403         GET_DEGRADED    = (1 << 8),
 404         GET_SAFEMODE    = (1 << 9),
 405         GET_DEVS        = (1 << 10), /* gets role, major, minor */
 406         GET_OFFSET      = (1 << 11),
 407         GET_SIZE        = (1 << 12),
 408         GET_STATE       = (1 << 13),
 409         GET_ERROR       = (1 << 14),
 410 };
 411
 412 /* If fd >= 0, get the array it is open on,
 413  * else use devnum. >=0 -> major9. <0.....
 414  */
 415 extern int sysfs_open(int devnum, char *devname, char *attr);
 416 extern void sysfs_init(struct mdinfo *mdi, int fd, int devnum);
 417 extern void sysfs_free(struct mdinfo *sra);
 418 extern struct mdinfo *sysfs_read(int fd, int devnum, unsigned long options);
 419 extern int sysfs_attr_match(const char *attr, const char *str);
 420 extern int sysfs_match_word(const char *word, char **list);
 421 extern int sysfs_set_str(struct mdinfo *sra, struct mdinfo *dev,
 422                          char *name, char *val);
 423 extern int sysfs_set_num(struct mdinfo *sra, struct mdinfo *dev,
 424                          char *name, unsigned long long val);
 425 extern int sysfs_uevent(struct mdinfo *sra, char *event);
 426 extern int sysfs_get_fd(struct mdinfo *sra, struct mdinfo *dev,
 427                         char *name);
 428 extern int sysfs_fd_get_ll(int fd, unsigned long long *val);
 429 extern int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev,
 430                         char *name, unsigned long long *val);
 431 extern int sysfs_fd_get_str(int fd, char *val, int size);
 432 extern int sysfs_get_str(struct mdinfo *sra, struct mdinfo *dev,
 433                          char *name, char *val, int size);
 434 extern int sysfs_set_safemode(struct mdinfo *sra, unsigned long ms);
 435 extern int sysfs_set_array(struct mdinfo *info, int vers);
 436 extern int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd, int resume);
 437 extern int sysfs_disk_to_scsi_id(int fd, __u32 *id);
 438 extern int sysfs_unique_holder(int devnum, long rdev);
 439 extern int load_sys(char *path, char *buf);
 440
 441
 442 extern int save_stripes(int *source, unsigned long long *offsets,
 443                         int raid_disks, int chunk_size, int level, int layout,
 444                         int nwrites, int *dest,
 445                         unsigned long long start, unsigned long long length,
 446                         char *buf);
 447 extern int restore_stripes(int *dest, unsigned long long *offsets,
 448                            int raid_disks, int chunk_size, int level, int layout,
 449                            int source, unsigned long long read_offset,
 450                            unsigned long long start, unsigned long long length);
 451
 452 #ifndef Sendmail
 453 #define Sendmail "/usr/lib/sendmail -t"
 454 #endif
 455
 456 #define SYSLOG_FACILITY LOG_DAEMON
 457
 458 extern char *map_num(mapping_t *map, int num);
 459 extern int map_name(mapping_t *map, char *name);
 460 extern mapping_t r5layout[], r6layout[], pers[], modes[], faultylayout[];
 461
 462 extern char *map_dev(int major, int minor, int create);
 463
 464 struct active_array;
 465 struct metadata_update;
 466
 467 /* A superswitch provides entry point the a metadata handler.
 468  *
 469  * The super_switch primarily operates on some "metadata" that
 470  * is accessed via the 'supertype'.
 471  * This metadata has one of three possible sources.
 472  * 1/ It is read from a single device.  In this case it may not completely
 473  *    describe the array or arrays as some information might be on other
 474  *    devices.
 475  * 2/ It is read from all devices in a container.  In this case all
 476  *    information is present.
 477  * 3/ It is created by ->init_super / ->add_to_super.  In this case it will
 478  *    be complete once enough ->add_to_super calls have completed.
 479  *
 480  * When creating an array inside a container, the metadata will be
 481  * formed by a combination of 2 and 3.  The metadata or the array is read,
 482  * then new information is added.
 483  *
 484  * The metadata must sometimes have a concept of a 'current' array
 485  * and a 'current' device.
 486  * The 'current' array is set by init_super to be the newly created array,
 487  * or is set by super_by_fd when it finds it is looking at an array inside
 488  * a container.
 489  *
 490  * The 'current' device is either the device that the metadata was read from
 491  * in case 1, or the last device added by add_to_super in case 3.
 492  * Case 2 does not identify a 'current' device.
 493  */
 494 extern struct superswitch {
 495
 496         /* Used to report details of metadata read from a component
 497          * device. ->load_super has been called.
 498          */
 499         void (*examine_super)(struct supertype *st, char *homehost);
 500         void (*brief_examine_super)(struct supertype *st, int verbose);
 501         void (*brief_examine_subarrays)(struct supertype *st, int verbose);
 502         void (*export_examine_super)(struct supertype *st);
 503
 504         /* Used to report details of an active array.
 505          * ->load_super was possibly given a 'component' string.
 506          */
 507         void (*detail_super)(struct supertype *st, char *homehost);
 508         void (*brief_detail_super)(struct supertype *st);
 509         void (*export_detail_super)(struct supertype *st);
 510
 511         /* Optional: platform hardware / firmware details */
 512         int (*detail_platform)(int verbose, int enumerate_only);
 513
 514         /* Used:
 515          *   to get uuid to storing in bitmap metadata
 516          *   and 'reshape' backup-data metadata
 517          *   To see if a device is being re-added to an array it was part of.
 518          */
 519         void (*uuid_from_super)(struct supertype *st, int uuid[4]);
 520
 521         /* Extract generic details from metadata.  This could be details about
 522          * the container, or about an individual array within the container.
 523          * The determination is made either by:
 524          *   load_super being given a 'component' string.
 525          *   validate_geometry determining what to create.
 526          * The info includes both array information and device information.
 527          * The particular device should be:
 528          *   The last device added by add_to_super
 529          *   The device the metadata was loaded from by load_super
 530          */
 531         void (*getinfo_super)(struct supertype *st, struct mdinfo *info);
 532
 533         /* Check if the given metadata is flagged as belonging to "this"
 534          * host.  0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
 535          */
 536         int (*match_home)(struct supertype *st, char *homehost);
 537
 538         /* Make one of several generic modifications to metadata
 539          * prior to assembly (or other times).
 540          *   sparc2.2  - first bug in early 0.90 metadata
 541          *   super-minor - change name of 0.90 metadata
 542          *   summaries - 'correct' any redundant data
 543          *   resync - mark array as dirty to trigger a resync.
 544          *   uuid - set new uuid - only 0.90 or 1.x
 545          *   name - change the name of the array (where supported)
 546          *   homehost - change which host this array is tied to.
 547          *   devicesize - If metadata is at start of device, change recorded
 548          *               device size to match actual device size
 549          *   byteorder - swap bytes for 0.90 metadata
 550          *
 551          *   force-one  - mark that device as uptodate, not old or failed.
 552          *   force-array - mark array as clean if it would not otherwise
 553          *               assemble
 554          *   assemble   - not sure how this is different from force-one...
 555          *   linear-grow-new - add a new device to a linear array, but don't
 556          *                   change the size: so superblock still matches
 557          *   linear-grow-update - now change the size of the array.
 558          */
 559         int (*update_super)(struct supertype *st, struct mdinfo *info,
 560                             char *update,
 561                             char *devname, int verbose,
 562                             int uuid_set, char *homehost);
 563
 564         /* Create new metadata for new array as described.  This could
 565          * be a new container, or an array in a pre-existing container.
 566          * Also used to zero metadata prior to writing it to invalidate old
 567          * metadata.
 568          */
 569         int (*init_super)(struct supertype *st, mdu_array_info_t *info,
 570                           unsigned long long size, char *name,
 571                           char *homehost, int *uuid);
 572
 573         /* update the metadata to include new device, either at create or
 574          * when hot-adding a spare.
 575          */
 576         int (*add_to_super)(struct supertype *st, mdu_disk_info_t *dinfo,
 577                              int fd, char *devname);
 578
 579         /* Write metadata to one device when fixing problems or adding
 580          * a new device.
 581          */
 582         int (*store_super)(struct supertype *st, int fd);
 583
 584         /*  Write all metadata for this array.
 585          */
 586         int (*write_init_super)(struct supertype *st);
 587         int (*compare_super)(struct supertype *st, struct supertype *tst);
 588         int (*load_super)(struct supertype *st, int fd, char *devname);
 589         struct supertype * (*match_metadata_desc)(char *arg);
 590         __u64 (*avail_size)(struct supertype *st, __u64 size);
 591         int (*add_internal_bitmap)(struct supertype *st, int *chunkp,
 592                                    int delay, int write_behind,
 593                                    unsigned long long size, int may_change, int major);
 594         void (*locate_bitmap)(struct supertype *st, int fd);
 595         int (*write_bitmap)(struct supertype *st, int fd);
 596         void (*free_super)(struct supertype *st);
 597
 598         /* validate_geometry is called with an st returned by
 599          * match_metadata_desc.
 600          * It should check that the geometry described in compatible with
 601          * the metadata type.  It will be called repeatedly as devices
 602          * added to validate changing size and new devices.  If there are
 603          * inter-device dependencies, it should record sufficient details
 604          * so these can be validated.
 605          * Both 'size' and '*freesize' are in sectors.  chunk is bytes.
 606          */
 607         int (*validate_geometry)(struct supertype *st, int level, int layout,
 608                                  int raiddisks,
 609                                  int chunk, unsigned long long size,
 610                                  char *subdev, unsigned long long *freesize,
 611                                  int verbose);
 612
 613         struct mdinfo *(*container_content)(struct supertype *st);
 614         /* Allow a metadata handler to override mdadm's default layouts */
 615         int (*default_layout)(int level); /* optional */
 616         /* query the supertype for default chunk size */
 617         int (*default_chunk)(struct supertype *st); /* optional */
 618         /* Permit subarray's to be deleted from inactive containers */
 619         int (*kill_subarray)(struct supertype *st); /* optional */
 620         /* Permit subarray's to be modified */
 621         int (*update_subarray)(struct supertype *st, char *update, mddev_ident_t ident); /* optional */
 622
 623 /* for mdmon */
 624         int (*open_new)(struct supertype *c, struct active_array *a,
 625                         char *inst);
 626
 627         /* Tell the metadata handler the current state of the array.
 628          * This covers whether it is known to be consistent (no pending writes)
 629          * and how far along a resync is known to have progressed
 630          * (in a->resync_start).
 631          * resync status is really irrelevant if the array is not consistent,
 632          * but some metadata (DDF!) have a place to record the distinction.
 633          * If 'consistent' is '2', then the array can mark it dirty if a
 634          * resync/recovery/whatever is required, or leave it clean if not.
 635          * Return value is 0 dirty (not consistent) and 1 if clean.
 636          * it is only really important if consistent is passed in as '2'.
 637          */
 638         int (*set_array_state)(struct active_array *a, int consistent);
 639
 640         /* When the state of a device might have changed, we call set_disk to
 641          * tell the metadata what the current state is.
 642          * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
 643          * transitions.
 644          * set_disk might be called when the state of the particular disk has
 645          * not in fact changed.
 646          */
 647         void (*set_disk)(struct active_array *a, int n, int state);
 648         void (*sync_metadata)(struct supertype *st);
 649         void (*process_update)(struct supertype *st,
 650                                struct metadata_update *update);
 651         void (*prepare_update)(struct supertype *st,
 652                                struct metadata_update *update);
 653
 654         /* activate_spare will check if the array is degraded and, if it
 655          * is, try to find some spare space in the container.
 656          * On success, it add appropriate updates (For process_update) to
 657          * to the 'updates' list and returns a list of 'mdinfo' identifying
 658          * the device, or devices as there might be multiple missing
 659          * devices and multiple spares available.
 660          */
 661         struct mdinfo *(*activate_spare)(struct active_array *a,
 662                                          struct metadata_update **updates);
 663
 664         int swapuuid; /* true if uuid is bigending rather than hostendian */
 665         int external;
 666         const char *name; /* canonical metadata name */
 667 } super0, super1, super_ddf, *superlist[];
 668
 669 extern struct superswitch super_imsm;
 670
 671 struct metadata_update {
 672         int     len;
 673         char    *buf;
 674         void    *space; /* allocated space that monitor will use */
 675         struct metadata_update *next;
 676 };
 677
 678 /* A supertype holds a particular collection of metadata.
 679  * It identifies the metadata type by the superswitch, and the particular
 680  * sub-version of that metadata type.
 681  * metadata read in or created is stored in 'sb' and 'info'.
 682  * There are also fields used by mdmon to track containers.
 683  *
 684  * A supertype may refer to:
 685  *   Just an array, possibly in a container
 686  *   A container, not identifying any particular array
 687  *   Info read from just one device, not yet fully describing the array/container.
 688  *
 689  *
 690  * A supertype is created by:
 691  *   super_by_fd
 692  *   guess_super
 693  *   dup_super
 694  */
 695 struct supertype {
 696         struct superswitch *ss;
 697         int minor_version;
 698         int max_devs;
 699         int container_dev;    /* devnum of container */
 700         char subarray[32];      /* name of array inside container */
 701         void *sb;
 702         void *info;
 703         int loaded_container;   /* Set if load_super found a container,
 704                                  * not just one device */
 705
 706         struct metadata_update *updates;
 707         struct metadata_update **update_tail;
 708
 709         /* extra stuff used by mdmon */
 710         struct active_array *arrays;
 711         int sock; /* listen to external programs */
 712         int devnum;
 713         char *devname; /* e.g. md0.  This appears in metadata_verison:
 714                         *  external:/md0/12
 715                         */
 716         int devcnt;
 717
 718         struct mdinfo *devs;
 719
 720 };
 721
 722 extern struct supertype *super_by_fd(int fd);
 723 extern struct supertype *guess_super(int fd);
 724 extern struct supertype *dup_super(struct supertype *st);
 725 extern int get_dev_size(int fd, char *dname, unsigned long long *sizep);
 726 extern void get_one_disk(int mdfd, mdu_array_info_t *ainf,
 727                          mdu_disk_info_t *disk);
 728 void wait_for(char *dev, int fd);
 729
 730 #if __GNUC__ < 3
 731 struct stat64;
 732 #endif
 733
 734 #define HAVE_NFTW  we assume
 735 #define HAVE_FTW
 736
 737 #ifdef __UCLIBC__
 738 # include <features.h>
 739 # ifndef __UCLIBC_HAS_LFS__
 740 #  define lseek64 lseek
 741 # endif
 742 # ifndef  __UCLIBC_HAS_FTW__
 743 #  undef HAVE_FTW
 744 #  undef HAVE_NFTW
 745 # endif
 746 #endif
 747
 748 #ifdef __dietlibc__
 749 # undef HAVE_NFTW
 750 #endif
 751
 752 #if defined(__KLIBC__)
 753 # undef HAVE_NFTW
 754 # undef HAVE_FTW
 755 #endif
 756
 757 #ifndef HAVE_NFTW
 758 # define FTW_PHYS 1
 759 # ifndef HAVE_FTW
 760   struct FTW {};
 761 # endif
 762 #endif
 763
 764 #ifdef HAVE_FTW
 765 # include <ftw.h>
 766 #endif
 767
 768 extern int add_dev(const char *name, const struct stat *stb, int flag, struct FTW *s);
 769
 770
 771 extern int Manage_ro(char *devname, int fd, int readonly);
 772 extern int Manage_runstop(char *devname, int fd, int runstop, int quiet);
 773 extern int Manage_resize(char *devname, int fd, long long size, int raid_disks);
 774 extern int Manage_subdevs(char *devname, int fd,
 775                           mddev_dev_t devlist, int verbose);
 776 extern int autodetect(void);
 777 extern int Grow_Add_device(char *devname, int fd, char *newdev);
 778 extern int Grow_addbitmap(char *devname, int fd, char *file, int chunk, int delay, int write_behind, int force);
 779 extern int Grow_reshape(char *devname, int fd, int quiet, char *backup_file,
 780                         long long size,
 781                         int level, char *layout_str, int chunksize, int raid_disks);
 782 extern int Grow_restart(struct supertype *st, struct mdinfo *info,
 783                         int *fdlist, int cnt, char *backup_file, int verbose);
 784 extern int Grow_continue(int mdfd, struct supertype *st,
 785                          struct mdinfo *info, char *backup_file);
 786
 787 extern int Assemble(struct supertype *st, char *mddev,
 788                     mddev_ident_t ident,
 789                     mddev_dev_t devlist, char *backup_file,
 790                     int readonly, int runstop,
 791                     char *update, char *homehost, int require_homehost,
 792                     int verbose, int force);
 793
 794 extern int Build(char *mddev, int chunk, int level, int layout,
 795                  int raiddisks, mddev_dev_t devlist, int assume_clean,
 796                  char *bitmap_file, int bitmap_chunk, int write_behind,
 797                  int delay, int verbose, int autof, unsigned long long size);
 798
 799
 800 extern int Create(struct supertype *st, char *mddev,
 801                   int chunk, int level, int layout, unsigned long long size, int raiddisks, int sparedisks,
 802                   char *name, char *homehost, int *uuid,
 803                   int subdevs, mddev_dev_t devlist,
 804                   int runstop, int verbose, int force, int assume_clean,
 805                   char *bitmap_file, int bitmap_chunk, int write_behind, int delay, int autof);
 806
 807 extern int Detail(char *dev, int brief, int export, int test, char *homehost);
 808 extern int Detail_Platform(struct superswitch *ss, int scan, int verbose);
 809 extern int Query(char *dev);
 810 extern int Examine(mddev_dev_t devlist, int brief, int export, int scan,
 811                    int SparcAdjust, struct supertype *forcest, char *homehost);
 812 extern int Monitor(mddev_dev_t devlist,
 813                    char *mailaddr, char *alert_cmd,
 814                    int period, int daemonise, int scan, int oneshot,
 815                    int dosyslog, int test, char *pidfile, int increments);
 816
 817 extern int Kill(char *dev, struct supertype *st, int force, int quiet, int noexcl);
 818 extern int Kill_subarray(char *dev, char *subarray, int quiet);
 819 extern int Update_subarray(char *dev, char *subarray, char *update, mddev_ident_t ident, int quiet);
 820 extern int Wait(char *dev);
 821 extern int WaitClean(char *dev, int sock, int verbose);
 822
 823 extern int Incremental(char *devname, int verbose, int runstop,
 824                        struct supertype *st, char *homehost, int require_homehost,
 825                        int autof);
 826 extern int Incremental_container(struct supertype *st, char *devname,
 827                                  int verbose, int runstop, int autof,
 828                                  int trustworthy);
 829 extern void RebuildMap(void);
 830 extern int IncrementalScan(int verbose);
 831
 832 extern int CreateBitmap(char *filename, int force, char uuid[16],
 833                         unsigned long chunksize, unsigned long daemon_sleep,
 834                         unsigned long write_behind,
 835                         unsigned long long array_size,
 836                         int major);
 837 extern int ExamineBitmap(char *filename, int brief, struct supertype *st);
 838 extern int bitmap_update_uuid(int fd, int *uuid, int swap);
 839 extern unsigned long bitmap_sectors(struct bitmap_super_s *bsb);
 840
 841 extern int md_get_version(int fd);
 842 extern int get_linux_version(void);
 843 extern long long parse_size(char *size);
 844 extern int parse_uuid(char *str, int uuid[4]);
 845 extern int parse_layout_10(char *layout);
 846 extern int parse_layout_faulty(char *layout);
 847 extern int check_ext2(int fd, char *name);
 848 extern int check_reiser(int fd, char *name);
 849 extern int check_raid(int fd, char *name);
 850 extern int check_partitions(int fd, char *dname, unsigned long long freesize);
 851
 852 extern int get_mdp_major(void);
 853 extern int dev_open(char *dev, int flags);
 854 extern int open_dev(int devnum);
 855 extern int open_dev_excl(int devnum);
 856 extern int is_standard(char *dev, int *nump);
 857 extern int same_dev(char *one, char *two);
 858
 859 extern int parse_auto(char *str, char *msg, int config);
 860 extern mddev_ident_t conf_get_ident(char *dev);
 861 extern mddev_dev_t conf_get_devs(void);
 862 extern int conf_test_dev(char *devname);
 863 extern int conf_test_metadata(const char *version, int is_homehost);
 864 extern struct createinfo *conf_get_create_info(void);
 865 extern void set_conffile(char *file);
 866 extern char *conf_get_mailaddr(void);
 867 extern char *conf_get_mailfrom(void);
 868 extern char *conf_get_program(void);
 869 extern char *conf_get_homehost(int *require_homehostp);
 870 extern char *conf_line(FILE *file);
 871 extern char *conf_word(FILE *file, int allow_key);
 872 extern int conf_name_is_free(char *name);
 873 extern int devname_matches(char *name, char *match);
 874 extern struct mddev_ident_s *conf_match(struct mdinfo *info, struct supertype *st);
 875
 876 extern void free_line(char *line);
 877 extern int match_oneof(char *devices, char *devname);
 878 extern void uuid_from_super(int uuid[4], mdp_super_t *super);
 879 extern const int uuid_match_any[4];
 880 extern int same_uuid(int a[4], int b[4], int swapuuid);
 881 extern void copy_uuid(void *a, int b[4], int swapuuid);
 882 extern char *__fname_from_uuid(int id[4], int swap, char *buf, char sep);
 883 extern char *fname_from_uuid(struct supertype *st,
 884                              struct mdinfo *info, char *buf, char sep);
 885 extern unsigned long calc_csum(void *super, int bytes);
 886 extern int enough(int level, int raid_disks, int layout, int clean,
 887                    char *avail, int avail_disks);
 888 extern int ask(char *mesg);
 889 extern unsigned long long get_component_size(int fd);
 890 extern void remove_partitions(int fd);
 891 extern int test_partition(int fd);
 892 extern unsigned long long calc_array_size(int level, int raid_disks, int layout,
 893                                    int chunksize, unsigned long long devsize);
 894 extern int flush_metadata_updates(struct supertype *st);
 895 extern void append_metadata_update(struct supertype *st, void *buf, int len);
 896 extern int assemble_container_content(struct supertype *st, int mdfd,
 897                                       struct mdinfo *content, int runstop,
 898                                       char *chosen_name, int verbose);
 899
 900 extern int add_disk(int mdfd, struct supertype *st,
 901                     struct mdinfo *sra, struct mdinfo *info);
 902 extern int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info);
 903 unsigned long long min_recovery_start(struct mdinfo *array);
 904
 905 extern char *human_size(long long bytes);
 906 extern char *human_size_brief(long long bytes);
 907 extern void print_r10_layout(int layout);
 908
 909 #define NoMdDev (1<<23)
 910 extern int find_free_devnum(int use_partitions);
 911
 912 extern void put_md_name(char *name);
 913 extern char *get_md_name(int dev);
 914
 915 extern char DefaultConfFile[];
 916
 917 extern int create_mddev(char *dev, char *name, int autof, int trustworthy,
 918                         char *chosen);
 919 /* values for 'trustworthy' */
 920 #define LOCAL   1
 921 #define LOCAL_ANY 10
 922 #define FOREIGN 2
 923 #define METADATA 3
 924 extern int open_mddev(char *dev, int report_errors);
 925 extern int open_container(int fd);
 926 extern int is_container_member(struct mdstat_ent *ent, char *devname);
 927 extern int is_subarray_active(char *subarray, char *devname);
 928 extern int open_subarray(char *dev, struct supertype *st, int quiet);
 929 extern struct superswitch *version_to_superswitch(char *vers);
 930
 931 extern char *pid_dir;
 932 extern int mdmon_running(int devnum);
 933 extern int mdmon_pid(int devnum);
 934 extern int check_env(char *name);
 935 extern __u32 random32(void);
 936 extern int start_mdmon(int devnum);
 937
 938 extern char *devnum2devname(int num);
 939 extern int devname2devnum(char *name);
 940 extern int stat2devnum(struct stat *st);
 941 extern int fd2devnum(int fd);
 942
 943 static inline int dev2major(int d)
 944 {
 945         if (d >= 0)
 946                 return MD_MAJOR;
 947         else
 948                 return get_mdp_major();
 949 }
 950
 951 static inline int dev2minor(int d)
 952 {
 953         if (d >= 0)
 954                 return d;
 955         return (-1-d) << MdpMinorShift;
 956 }
 957
 958 static inline int ROUND_UP(int a, int base)
 959 {
 960         return ((a+base-1)/base)*base;
 961 }
 962
 963 static inline int is_subarray(char *vers)
 964 {
 965         /* The version string for a 'subarray' (an array in a container)
 966          * is
 967          *    /containername/componentname    for normal read-write arrays
 968          *    -containername/componentname    for read-only arrays.
 969          * containername is e.g. md0, md_d1
 970          * componentname is dependant on the metadata. e.g. '1' 'S1' ...
 971          */
 972         return (*vers == '/' || *vers == '-');
 973 }
 974
 975 #ifdef DEBUG
 976 #define dprintf(fmt, arg...) \
 977         fprintf(stderr, fmt, ##arg)
 978 #else
 979 #define dprintf(fmt, arg...) \
 980         ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
 981 #endif
 982 #include <assert.h>
 983 #include <stdarg.h>
 984 static inline int xasprintf(char **strp, const char *fmt, ...) {
 985         va_list ap;
 986         int ret;
 987         va_start(ap, fmt);
 988         ret = vasprintf(strp, fmt, ap);
 989         va_end(ap);
 990         assert(ret >= 0);
 991         return ret;
 992 }
 993
 994 #define LEVEL_MULTIPATH         (-4)
 995 #define LEVEL_LINEAR            (-1)
 996 #define LEVEL_FAULTY            (-5)
 997
 998 /* kernel module doesn't know about these */
 999 #define LEVEL_CONTAINER         (-100)
1000 #define LEVEL_UNSUPPORTED       (-200)
1001
1002
1003 /* faulty stuff */
1004
1005 #define WriteTransient  0
1006 #define ReadTransient   1
1007 #define WritePersistent 2
1008 #define ReadPersistent  3
1009 #define WriteAll        4 /* doesn't go to device */
1010 #define ReadFixable     5
1011 #define Modes   6
1012
1013 #define ClearErrors     31
1014 #define ClearFaults     30
1015
1016 #define AllPersist      100 /* internal use only */
1017 #define NoPersist       101
1018
1019 #define ModeMask        0x1f
1020 #define ModeShift       5
1021
1022
1023 #ifdef __TINYC__
1024 #undef minor
1025 #undef major
1026 #undef makedev
1027 #define minor(x) ((x)&0xff)
1028 #define major(x) (((x)>>8)&0xff)
1029 #define makedev(M,m) (((M)<<8) | (m))
1030 #endif
1031
1032 /* for raid4/5/6 */
1033 #define ALGORITHM_LEFT_ASYMMETRIC       0
1034 #define ALGORITHM_RIGHT_ASYMMETRIC      1
1035 #define ALGORITHM_LEFT_SYMMETRIC        2
1036 #define ALGORITHM_RIGHT_SYMMETRIC       3
1037
1038 /* Define non-rotating (raid4) algorithms.  These allow
1039  * conversion of raid4 to raid5.
1040  */
1041 #define ALGORITHM_PARITY_0              4 /* P or P,Q are initial devices */
1042 #define ALGORITHM_PARITY_N              5 /* P or P,Q are final devices. */
1043
1044 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1045  * Firstly, the exact positioning of the parity block is slightly
1046  * different between the 'LEFT_*' modes of md and the "_N_*" modes
1047  * of DDF.
1048  * Secondly, or order of datablocks over which the Q syndrome is computed
1049  * is different.
1050  * Consequently we have different layouts for DDF/raid6 than md/raid6.
1051  * These layouts are from the DDFv1.2 spec.
1052  * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1053  * leaves RLQ=3 as 'Vendor Specific'
1054  */
1055
1056 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1057 #define ALGORITHM_ROTATING_N_RESTART    9 /* DDF PRL=6 RLQ=2 */
1058 #define ALGORITHM_ROTATING_N_CONTINUE   10 /*DDF PRL=6 RLQ=3 */
1059
1060
1061 /* For every RAID5 algorithm we define a RAID6 algorithm
1062  * with exactly the same layout for data and parity, and
1063  * with the Q block always on the last device (N-1).
1064  * This allows trivial conversion from RAID5 to RAID6
1065  */
1066 #define ALGORITHM_LEFT_ASYMMETRIC_6     16
1067 #define ALGORITHM_RIGHT_ASYMMETRIC_6    17
1068 #define ALGORITHM_LEFT_SYMMETRIC_6      18
1069 #define ALGORITHM_RIGHT_SYMMETRIC_6     19
1070 #define ALGORITHM_PARITY_0_6            20
1071 #define ALGORITHM_PARITY_N_6            ALGORITHM_PARITY_N
1072
1073 /* Define PATH_MAX in case we don't use glibc or standard library does
1074  * not have PATH_MAX defined. Assume max path length is 4K characters.
1075  */
1076 #ifndef PATH_MAX
1077 #define PATH_MAX        4096
1078 #endif
1079