mdadm.h

   1 /*
   2  * mdadm - manage Linux "md" devices aka RAID arrays.
   3  *
   4  * Copyright (C) 2001-2006 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@cse.unsw.edu.au>
  23  *    Paper: Neil Brown
  24  *           School of Computer Science and Engineering
  25  *           The University of New South Wales
  26  *           Sydney, 2052
  27  *           Australia
  28  */
  29
  30 #define _GNU_SOURCE
  31 #define _FILE_OFFSET_BITS 64
  32 #include        <unistd.h>
  33 #if !defined(__dietlibc__) && !defined(__KLIBC__)
  34 extern __off64_t lseek64 __P ((int __fd, __off64_t __offset, int __whence));
  35 #else
  36 # if defined(__NO_STAT64) || __WORDSIZE != 32
  37 # define lseek64 lseek
  38 # endif
  39 #endif
  40
  41 #include        <sys/types.h>
  42 #include        <sys/stat.h>
  43 #include        <stdlib.h>
  44 #include        <time.h>
  45 #include        <sys/time.h>
  46 #include        <getopt.h>
  47 #include        <fcntl.h>
  48 #include        <stdio.h>
  49 #include        <errno.h>
  50 #include        <string.h>
  51 #include        <syslog.h>
  52 #ifdef __dietlibc__
  53 #include        <strings.h>
  54 /* dietlibc has deprecated random and srandom!! */
  55 #define random rand
  56 #define srandom srand
  57 #endif
  58
  59
  60 #include        <linux/kdev_t.h>
  61 /*#include      <linux/fs.h> */
  62 #include        <sys/mount.h>
  63 #include        <asm/types.h>
  64 #include        <sys/ioctl.h>
  65 #define MD_MAJOR 9
  66 #define MdpMinorShift 6
  67
  68 #ifndef BLKGETSIZE64
  69 #define BLKGETSIZE64 _IOR(0x12,114,size_t) /* return device size in bytes (u64 *arg) */
  70 #endif
  71
  72 #define DEFAULT_BITMAP_CHUNK 4096
  73 #define DEFAULT_BITMAP_DELAY 5
  74 #define DEFAULT_MAX_WRITE_BEHIND 256
  75
  76 #include        "md_u.h"
  77 #include        "md_p.h"
  78 #include        "bitmap.h"
  79 #include        "msg.h"
  80
  81 #include <endian.h>
  82 /* Redhat don't like to #include <asm/byteorder.h>, and
  83  * some time include <linux/byteorder/xxx_endian.h> isn't enough,
  84  * and there is no standard conversion function so... */
  85 /* And dietlibc doesn't think byteswap is ok, so.. */
  86 /*  #include <byteswap.h> */
  87 #define bswap_16(x) (((x) & 0x00ffU) << 8 | \
  88                      ((x) & 0xff00U) >> 8)
  89 #define bswap_32(x) (((x) & 0x000000ffU) << 24 | \
  90                      ((x) & 0xff000000U) >> 24 | \
  91                      ((x) & 0x0000ff00U) << 8  | \
  92                      ((x) & 0x00ff0000U) >> 8)
  93 #define bswap_64(x) (((x) & 0x00000000000000ffULL) << 56 | \
  94                      ((x) & 0xff00000000000000ULL) >> 56 | \
  95                      ((x) & 0x000000000000ff00ULL) << 40 | \
  96                      ((x) & 0x00ff000000000000ULL) >> 40 | \
  97                      ((x) & 0x0000000000ff0000ULL) << 24 | \
  98                      ((x) & 0x0000ff0000000000ULL) >> 24 | \
  99                      ((x) & 0x00000000ff000000ULL) << 8 | \
 100                      ((x) & 0x000000ff00000000ULL) >> 8)
 101
 102 #if !defined(__KLIBC__)
 103 #if BYTE_ORDER == LITTLE_ENDIAN
 104 #define __cpu_to_le16(_x) (_x)
 105 #define __cpu_to_le32(_x) (_x)
 106 #define __cpu_to_le64(_x) (_x)
 107 #define __le16_to_cpu(_x) (_x)
 108 #define __le32_to_cpu(_x) (_x)
 109 #define __le64_to_cpu(_x) (_x)
 110
 111 #define __cpu_to_be16(_x) bswap_16(_x)
 112 #define __cpu_to_be32(_x) bswap_32(_x)
 113 #define __cpu_to_be64(_x) bswap_64(_x)
 114 #define __be16_to_cpu(_x) bswap_16(_x)
 115 #define __be32_to_cpu(_x) bswap_32(_x)
 116 #define __be64_to_cpu(_x) bswap_64(_x)
 117 #elif BYTE_ORDER == BIG_ENDIAN
 118 #define __cpu_to_le16(_x) bswap_16(_x)
 119 #define __cpu_to_le32(_x) bswap_32(_x)
 120 #define __cpu_to_le64(_x) bswap_64(_x)
 121 #define __le16_to_cpu(_x) bswap_16(_x)
 122 #define __le32_to_cpu(_x) bswap_32(_x)
 123 #define __le64_to_cpu(_x) bswap_64(_x)
 124
 125 #define __cpu_to_be16(_x) (_x)
 126 #define __cpu_to_be32(_x) (_x)
 127 #define __cpu_to_be64(_x) (_x)
 128 #define __be16_to_cpu(_x) (_x)
 129 #define __be32_to_cpu(_x) (_x)
 130 #define __be64_to_cpu(_x) (_x)
 131 #else
 132 #  error "unknown endianness."
 133 #endif
 134 #endif /* __KLIBC__ */
 135
 136
 137
 138 /* general information that might be extracted from a superblock */
 139 struct mdinfo {
 140         mdu_array_info_t        array;
 141         mdu_disk_info_t         disk;
 142         __u64                   events;
 143         int                     uuid[4];
 144         char                    name[33];
 145         unsigned long long      data_offset;
 146         unsigned long long      component_size; /* same as array.size, except in
 147                                                  * sectors and up to 64bits.
 148                                                  */
 149         unsigned long long      custom_array_size; /* size for non-default sized
 150                                                     * arrays (in sectors)
 151                                                     */
 152         int                     reshape_active;
 153         unsigned long long      reshape_progress;
 154         unsigned long long      resync_start;
 155         unsigned long           safe_mode_delay; /* ms delay to mark clean */
 156         int                     new_level, delta_disks, new_layout, new_chunk;
 157         int                     errors;
 158         int                     cache_size; /* size of raid456 stripe cache*/
 159         int                     mismatch_cnt;
 160         char                    text_version[50];
 161
 162         int container_member; /* for assembling external-metatdata arrays
 163                                * This is to be used internally by metadata
 164                                * handler only */
 165
 166         char            sys_name[20];
 167         struct mdinfo *devs;
 168         struct mdinfo *next;
 169
 170         /* Device info for mdmon: */
 171         int state_fd;
 172         #define DS_FAULTY       1
 173         #define DS_INSYNC       2
 174         #define DS_WRITE_MOSTLY 4
 175         #define DS_SPARE        8
 176         #define DS_BLOCKED      16
 177         #define DS_REMOVE       1024
 178         #define DS_UNBLOCK      2048
 179         int prev_state, curr_state, next_state;
 180
 181 };
 182
 183 struct createinfo {
 184         int     uid;
 185         int     gid;
 186         int     autof;
 187         int     mode;
 188         int     symlinks;
 189         struct supertype *supertype;
 190 };
 191
 192 #define Name "mdadm"
 193
 194 enum mode {
 195         ASSEMBLE=1,
 196         BUILD,
 197         CREATE,
 198         MANAGE,
 199         MISC,
 200         MONITOR,
 201         GROW,
 202         INCREMENTAL,
 203         AUTODETECT,
 204 };
 205
 206 extern char short_options[];
 207 extern char short_bitmap_options[];
 208 extern char short_bitmap_auto_options[];
 209 extern struct option long_options[];
 210 extern char Version[], Usage[], Help[], OptionHelp[],
 211         Help_create[], Help_build[], Help_assemble[], Help_grow[],
 212         Help_incr[],
 213         Help_manage[], Help_misc[], Help_monitor[], Help_config[];
 214
 215 /* for option that don't have short equivilents, we assign arbitrary
 216  * small numbers.  '1' means an undecorated option, so we start at '2'.
 217  */
 218 enum special_options {
 219         AssumeClean = 2,
 220         BitmapChunk,
 221         WriteBehind,
 222         ReAdd,
 223         NoDegraded,
 224         Sparc22,
 225         BackupFile,
 226         HomeHost,
 227         AutoHomeHost,
 228         Symlinks,
 229         AutoDetect,
 230         Waitclean,
 231         DetailPlatform,
 232 };
 233
 234 /* structures read from config file */
 235 /* List of mddevice names and identifiers
 236  * Identifiers can be:
 237  *    uuid=128-hex-uuid
 238  *    super-minor=decimal-minor-number-from-superblock
 239  *    devices=comma,separated,list,of,device,names,with,wildcards
 240  *
 241  * If multiple fields are present, the intersection of all matching
 242  * devices is considered
 243  */
 244 #define UnSet (0xfffe)
 245 typedef struct mddev_ident_s {
 246         char    *devname;
 247
 248         int     uuid_set;
 249         int     uuid[4];
 250         char    name[33];
 251
 252         unsigned int super_minor;
 253
 254         char    *devices;       /* comma separated list of device
 255                                  * names with wild cards
 256                                  */
 257         int     level;
 258         unsigned int raid_disks;
 259         unsigned int spare_disks;
 260         struct supertype *st;
 261         int     autof;          /* 1 for normal, 2 for partitioned */
 262         char    *spare_group;
 263         char    *bitmap_file;
 264         int     bitmap_fd;
 265
 266         char    *container;     /* /dev/whatever name of container, or
 267                                  * uuid of container.  You would expect
 268                                  * this to be the 'devname' or UUID
 269                                  * of some other entry.
 270                                  */
 271         char    *member;        /* subarray within a container */
 272
 273         struct mddev_ident_s *next;
 274 } *mddev_ident_t;
 275
 276 /* List of device names - wildcards expanded */
 277 typedef struct mddev_dev_s {
 278         char *devname;
 279         char disposition;       /* 'a' for add, 'r' for remove, 'f' for fail.
 280                                  * Not set for names read from .config
 281                                  */
 282         char writemostly;       /* 1 for 'set writemostly', 2 for 'clear writemostly' */
 283         char re_add;
 284         char used;              /* set when used */
 285         struct mdinfo *content; /* If devname is a container, this might list
 286                                  * the remaining member arrays. */
 287         struct mddev_dev_s *next;
 288 } *mddev_dev_t;
 289
 290 typedef struct mapping {
 291         char *name;
 292         int num;
 293 } mapping_t;
 294
 295
 296 struct mdstat_ent {
 297         char            *dev;
 298         int             devnum;
 299         int             active;
 300         char            *level;
 301         char            *pattern; /* U or up, _ for down */
 302         int             percent; /* -1 if no resync */
 303         int             resync; /* 1 if resync, 0 if recovery */
 304         int             devcnt;
 305         int             raid_disks;
 306         int             chunk_size;
 307         char *          metadata_version;
 308         struct mdstat_ent *next;
 309 };
 310
 311 extern struct mdstat_ent *mdstat_read(int hold, int start);
 312 extern void free_mdstat(struct mdstat_ent *ms);
 313 extern void mdstat_wait(int seconds);
 314 extern void mdstat_wait_fd(int fd, const sigset_t *sigmask);
 315 extern int mddev_busy(int devnum);
 316
 317 struct map_ent {
 318         struct map_ent *next;
 319         int     devnum;
 320         char    metadata[20];
 321         int     uuid[4];
 322         int     bad;
 323         char    *path;
 324 };
 325 extern int map_update(struct map_ent **mpp, int devnum, char *metadata,
 326                       int uuid[4], char *path);
 327 extern struct map_ent *map_by_uuid(struct map_ent **map, int uuid[4]);
 328 extern struct map_ent *map_by_devnum(struct map_ent **map, int devnum);
 329 extern struct map_ent *map_by_name(struct map_ent **map, char *name);
 330 extern void map_read(struct map_ent **melp);
 331 extern int map_write(struct map_ent *mel);
 332 extern void map_delete(struct map_ent **mapp, int devnum);
 333 extern void map_free(struct map_ent *map);
 334 extern void map_add(struct map_ent **melp,
 335                     int devnum, char *metadata, int uuid[4], char *path);
 336 extern int map_lock(struct map_ent **melp);
 337 extern void map_unlock(struct map_ent **melp);
 338
 339 /* various details can be requested */
 340 enum sysfs_read_flags {
 341         GET_LEVEL       = (1 << 0),
 342         GET_LAYOUT      = (1 << 1),
 343         GET_COMPONENT   = (1 << 2),
 344         GET_CHUNK       = (1 << 3),
 345         GET_CACHE       = (1 << 4),
 346         GET_MISMATCH    = (1 << 5),
 347         GET_VERSION     = (1 << 6),
 348         GET_DISKS       = (1 << 7),
 349         GET_DEGRADED    = (1 << 8),
 350         GET_SAFEMODE    = (1 << 9),
 351         GET_DEVS        = (1 << 10), /* gets role, major, minor */
 352         GET_OFFSET      = (1 << 11),
 353         GET_SIZE        = (1 << 12),
 354         GET_STATE       = (1 << 13),
 355         GET_ERROR       = (1 << 14),
 356         SKIP_GONE_DEVS  = (1 << 15),
 357 };
 358
 359 /* If fd >= 0, get the array it is open on,
 360  * else use devnum. >=0 -> major9. <0.....
 361  */
 362 extern int sysfs_open(int devnum, char *devname, char *attr);
 363 extern void sysfs_init(struct mdinfo *mdi, int fd, int devnum);
 364 extern void sysfs_free(struct mdinfo *sra);
 365 extern struct mdinfo *sysfs_read(int fd, int devnum, unsigned long options);
 366 extern int sysfs_attr_match(const char *attr, const char *str);
 367 extern int sysfs_match_word(const char *word, char **list);
 368 extern int sysfs_set_str(struct mdinfo *sra, struct mdinfo *dev,
 369                          char *name, char *val);
 370 extern int sysfs_set_num(struct mdinfo *sra, struct mdinfo *dev,
 371                          char *name, unsigned long long val);
 372 extern int sysfs_uevent(struct mdinfo *sra, char *event);
 373 extern int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev,
 374                         char *name, unsigned long long *val);
 375 extern int sysfs_set_safemode(struct mdinfo *sra, unsigned long ms);
 376 extern int sysfs_set_array(struct mdinfo *info, int vers);
 377 extern int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd);
 378 extern int sysfs_disk_to_scsi_id(int fd, __u32 *id);
 379 extern int sysfs_unique_holder(int devnum, long rdev);
 380 extern int load_sys(char *path, char *buf);
 381
 382
 383 extern int save_stripes(int *source, unsigned long long *offsets,
 384                         int raid_disks, int chunk_size, int level, int layout,
 385                         int nwrites, int *dest,
 386                         unsigned long long start, unsigned long long length);
 387 extern int restore_stripes(int *dest, unsigned long long *offsets,
 388                            int raid_disks, int chunk_size, int level, int layout,
 389                            int source, unsigned long long read_offset,
 390                            unsigned long long start, unsigned long long length);
 391
 392 #ifndef Sendmail
 393 #define Sendmail "/usr/lib/sendmail -t"
 394 #endif
 395
 396 #define SYSLOG_FACILITY LOG_DAEMON
 397
 398 extern char *map_num(mapping_t *map, int num);
 399 extern int map_name(mapping_t *map, char *name);
 400 extern mapping_t r5layout[], r6layout[], pers[], modes[], faultylayout[];
 401
 402 extern char *map_dev(int major, int minor, int create);
 403
 404 struct active_array;
 405 struct metadata_update;
 406
 407 /* A superswitch provides entry point the a metadata handler.
 408  *
 409  * The super_switch primarily operates on some "metadata" that
 410  * is accessed via the 'supertype'.
 411  * This metadata has one of three possible sources.
 412  * 1/ It is read from a single device.  In this case it may not completely
 413  *    describe the array or arrays as some information might be on other
 414  *    devices.
 415  * 2/ It is read from all devices in a container.  In this case all
 416  *    information is present.
 417  * 3/ It is created by ->init_super / ->add_to_super.  In this case it will
 418  *    be complete once enough ->add_to_super calls have completed.
 419  *
 420  * When creating an array inside a container, the metadata will be
 421  * formed by a combination of 2 and 3.  The metadata or the array is read,
 422  * then new information is added.
 423  *
 424  * The metadata must sometimes have a concept of a 'current' array
 425  * and a 'current' device.
 426  * The 'current' array is set by init_super to be the newly created array,
 427  * or is set by super_by_fd when it finds it is looking at an array inside
 428  * a container.
 429  *
 430  * The 'current' device is either the device that the metadata was read from
 431  * in case 1, or the last device added by add_to_super in case 3.
 432  * Case 2 does not identify a 'current' device.
 433  */
 434 extern struct superswitch {
 435
 436         /* Used to report details of metadata read from a component
 437          * device. ->load_super has been called.
 438          */
 439         void (*examine_super)(struct supertype *st, char *homehost);
 440         void (*brief_examine_super)(struct supertype *st);
 441         void (*export_examine_super)(struct supertype *st);
 442
 443         /* Used to report details of an active array.
 444          * ->load_super was possibly given a 'component' string.
 445          */
 446         void (*detail_super)(struct supertype *st, char *homehost);
 447         void (*brief_detail_super)(struct supertype *st);
 448         void (*export_detail_super)(struct supertype *st);
 449
 450         /* Optional: platform hardware / firmware details */
 451         int (*detail_platform)(int verbose, int enumerate_only);
 452
 453         /* Used:
 454          *   to get uuid to storing in bitmap metadata
 455          *   and 'reshape' backup-data metadata
 456          *   To see if a device is being re-added to an array it was part of.
 457          */
 458         void (*uuid_from_super)(struct supertype *st, int uuid[4]);
 459
 460         /* Extract generic details from metadata.  This could be details about
 461          * the container, or about an individual array within the container.
 462          * The determination is made either by:
 463          *   load_super being given a 'component' string.
 464          *   validate_geometry determining what to create.
 465          * The info includes both array information and device information.
 466          * The particular device should be:
 467          *   The last device added by add_to_super
 468          *   The device the metadata was loaded from by load_super
 469          */
 470         void (*getinfo_super)(struct supertype *st, struct mdinfo *info);
 471
 472         /* Check if the given metadata is flagged as belonging to "this"
 473          * host.  0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
 474          */
 475         int (*match_home)(struct supertype *st, char *homehost);
 476
 477         /* Make one of several generic modifications to metadata
 478          * prior to assembly (or other times).
 479          *   sparc2.2  - first bug in early 0.90 metadata
 480          *   super-minor - change name of 0.90 metadata
 481          *   summaries - 'correct' any redundant data
 482          *   resync - mark array as dirty to trigger a resync.
 483          *   uuid - set new uuid - only 0.90 or 1.x
 484          *   name - change the name of the array (where supported)
 485          *   homehost - change which host this array is tied to.
 486          *   devicesize - If metadata is at start of device, change recorded
 487          *               device size to match actual device size
 488          *   byteorder - swap bytes for 0.90 metadata
 489          *
 490          *   force-one  - mark that device as uptodate, not old or failed.
 491          *   force-array - mark array as clean if it would not otherwise
 492          *               assemble
 493          *   assemble   - not sure how this is different from force-one...
 494          *   linear-grow-new - add a new device to a linear array, but don't
 495          *                   change the size: so superblock still matches
 496          *   linear-grow-update - now change the size of the array.
 497          */
 498         int (*update_super)(struct supertype *st, struct mdinfo *info,
 499                             char *update,
 500                             char *devname, int verbose,
 501                             int uuid_set, char *homehost);
 502
 503         /* Create new metadata for new array as described.  This could
 504          * be a new container, or an array in a pre-existing container.
 505          * Also used to zero metadata prior to writing it to invalidate old
 506          * metadata.
 507          */
 508         int (*init_super)(struct supertype *st, mdu_array_info_t *info,
 509                           unsigned long long size, char *name,
 510                           char *homehost, int *uuid);
 511
 512         /* update the metadata to include new device, either at create or
 513          * when hot-adding a spare.
 514          */
 515         int (*add_to_super)(struct supertype *st, mdu_disk_info_t *dinfo,
 516                              int fd, char *devname);
 517
 518         /* Write metadata to one device when fixing problems or adding
 519          * a new device.
 520          */
 521         int (*store_super)(struct supertype *st, int fd);
 522
 523         /*  Write all metadata for this array.
 524          */
 525         int (*write_init_super)(struct supertype *st);
 526         int (*compare_super)(struct supertype *st, struct supertype *tst);
 527         int (*load_super)(struct supertype *st, int fd, char *devname);
 528         struct supertype * (*match_metadata_desc)(char *arg);
 529         __u64 (*avail_size)(struct supertype *st, __u64 size);
 530         int (*add_internal_bitmap)(struct supertype *st, int *chunkp,
 531                                    int delay, int write_behind,
 532                                    unsigned long long size, int may_change, int major);
 533         void (*locate_bitmap)(struct supertype *st, int fd);
 534         int (*write_bitmap)(struct supertype *st, int fd);
 535         void (*free_super)(struct supertype *st);
 536
 537         /* validate_geometry is called with an st returned by
 538          * match_metadata_desc.
 539          * It should check that the geometry described in compatible with
 540          * the metadata type.  It will be called repeatedly as devices
 541          * added to validate changing size and new devices.  If there are
 542          * inter-device dependencies, it should record sufficient details
 543          * so these can be validated.
 544          * Both 'size' and '*freesize' are in sectors.  chunk is bytes.
 545          */
 546         int (*validate_geometry)(struct supertype *st, int level, int layout,
 547                                  int raiddisks,
 548                                  int chunk, unsigned long long size,
 549                                  char *subdev, unsigned long long *freesize,
 550                                  int verbose);
 551
 552         struct mdinfo *(*container_content)(struct supertype *st);
 553         /* Allow a metadata handler to override mdadm's default layouts */
 554         int (*default_layout)(int level); /* optional */
 555
 556 /* for mdmon */
 557         int (*open_new)(struct supertype *c, struct active_array *a,
 558                         char *inst);
 559
 560         /* Tell the metadata handler the current state of the array.
 561          * This covers whether it is known to be consistent (no pending writes)
 562          * and how far along a resync is known to have progressed
 563          * (in a->resync_start).
 564          * resync status is really irrelevant if the array is not consistent,
 565          * but some metadata (DDF!) have a place to record the distinction.
 566          * If 'consistent' is '2', then the array can mark it dirty if a
 567          * resync/recovery/whatever is required, or leave it clean if not.
 568          * Return value is 0 dirty (not consistent) and 1 if clean.
 569          * it is only really important if consistent is passed in as '2'.
 570          */
 571         int (*set_array_state)(struct active_array *a, int consistent);
 572
 573         /* When the state of a device might have changed, we call set_disk to
 574          * tell the metadata what the current state is.
 575          * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
 576          * transitions.
 577          * set_disk might be called when the state of the particular disk has
 578          * not in fact changed.
 579          */
 580         void (*set_disk)(struct active_array *a, int n, int state);
 581         void (*sync_metadata)(struct supertype *st);
 582         void (*process_update)(struct supertype *st,
 583                                struct metadata_update *update);
 584         void (*prepare_update)(struct supertype *st,
 585                                struct metadata_update *update);
 586
 587         /* activate_spare will check if the array is degraded and, if it
 588          * is, try to find some spare space in the container.
 589          * On success, it add appropriate updates (For process_update) to
 590          * to the 'updates' list and returns a list of 'mdinfo' identifying
 591          * the device, or devices as there might be multiple missing
 592          * devices and multiple spares available.
 593          */
 594         struct mdinfo *(*activate_spare)(struct active_array *a,
 595                                          struct metadata_update **updates);
 596
 597         int swapuuid; /* true if uuid is bigending rather than hostendian */
 598         int external;
 599         const char *name; /* canonical metadata name */
 600 } super0, super1, super_ddf, *superlist[];
 601
 602 extern struct superswitch super_imsm;
 603
 604 struct metadata_update {
 605         int     len;
 606         char    *buf;
 607         void    *space; /* allocated space that monitor will use */
 608         struct metadata_update *next;
 609 };
 610
 611 /* A supertype holds a particular collection of metadata.
 612  * It identifies the metadata type by the superswitch, and the particular
 613  * sub-version of that metadata type.
 614  * metadata read in or created is stored in 'sb' and 'info'.
 615  * There are also fields used by mdmon to track containers.
 616  *
 617  * A supertype may refer to:
 618  *   Just an array, possibly in a container
 619  *   A container, not identifying any particular array
 620  *   Info read from just one device, not yet fully describing the array/container.
 621  *
 622  *
 623  * A supertype is created by:
 624  *   super_by_fd
 625  *   guess_super
 626  *   dup_super
 627  */
 628 struct supertype {
 629         struct superswitch *ss;
 630         int minor_version;
 631         int max_devs;
 632         int container_dev;    /* devnum of container */
 633         char subarray[32];      /* name of array inside container */
 634         void *sb;
 635         void *info;
 636         int loaded_container;   /* Set if load_super found a container,
 637                                  * not just one device */
 638
 639         struct metadata_update *updates;
 640         struct metadata_update **update_tail;
 641
 642         /* extra stuff used by mdmon */
 643         struct active_array *arrays;
 644         int sock; /* listen to external programs */
 645         int devnum;
 646         char *devname; /* e.g. md0.  This appears in metadata_verison:
 647                         *  external:/md0/12
 648                         */
 649         int devcnt;
 650
 651         struct mdinfo *devs;
 652
 653 };
 654
 655 extern struct supertype *super_by_fd(int fd);
 656 extern struct supertype *guess_super(int fd);
 657 extern struct supertype *dup_super(struct supertype *st);
 658 extern int get_dev_size(int fd, char *dname, unsigned long long *sizep);
 659 extern void get_one_disk(int mdfd, mdu_array_info_t *ainf,
 660                          mdu_disk_info_t *disk);
 661 void wait_for(char *dev);
 662
 663 #if __GNUC__ < 3
 664 struct stat64;
 665 #endif
 666
 667 #define HAVE_NFTW  we assume
 668 #define HAVE_FTW
 669
 670 #ifdef __UCLIBC__
 671 # include <features.h>
 672 # ifndef __UCLIBC_HAS_LFS__
 673 #  define lseek64 lseek
 674 # endif
 675 # ifndef  __UCLIBC_HAS_FTW__
 676 #  undef HAVE_FTW
 677 #  undef HAVE_NFTW
 678 # endif
 679 #endif
 680
 681 #ifdef __dietlibc__
 682 # undef HAVE_NFTW
 683 #endif
 684
 685 #if defined(__KLIBC__)
 686 # undef HAVE_NFTW
 687 # undef HAVE_FTW
 688 #endif
 689
 690 #ifndef HAVE_NFTW
 691 # define FTW_PHYS 1
 692 # ifndef HAVE_FTW
 693   struct FTW {};
 694 # endif
 695 #endif
 696
 697 #ifdef HAVE_FTW
 698 # include <ftw.h>
 699 #endif
 700
 701 extern int add_dev(const char *name, const struct stat *stb, int flag, struct FTW *s);
 702
 703
 704 extern int Manage_ro(char *devname, int fd, int readonly);
 705 extern int Manage_runstop(char *devname, int fd, int runstop, int quiet);
 706 extern int Manage_resize(char *devname, int fd, long long size, int raid_disks);
 707 extern int Manage_reconfig(char *devname, int fd, int layout);
 708 extern int Manage_subdevs(char *devname, int fd,
 709                           mddev_dev_t devlist, int verbose);
 710 extern int autodetect(void);
 711 extern int Grow_Add_device(char *devname, int fd, char *newdev);
 712 extern int Grow_addbitmap(char *devname, int fd, char *file, int chunk, int delay, int write_behind, int force);
 713 extern int Grow_reshape(char *devname, int fd, int quiet, char *backup_file,
 714                         long long size,
 715                         int level, int layout, int chunksize, int raid_disks);
 716 extern int Grow_restart(struct supertype *st, struct mdinfo *info,
 717                         int *fdlist, int cnt, char *backup_file);
 718
 719
 720 extern int Assemble(struct supertype *st, char *mddev,
 721                     mddev_ident_t ident,
 722                     mddev_dev_t devlist, char *backup_file,
 723                     int readonly, int runstop,
 724                     char *update, char *homehost,
 725                     int verbose, int force);
 726
 727 extern int Build(char *mddev, int chunk, int level, int layout,
 728                  int raiddisks, mddev_dev_t devlist, int assume_clean,
 729                  char *bitmap_file, int bitmap_chunk, int write_behind,
 730                  int delay, int verbose, int autof);
 731
 732
 733 extern int Create(struct supertype *st, char *mddev,
 734                   int chunk, int level, int layout, unsigned long long size, int raiddisks, int sparedisks,
 735                   char *name, char *homehost, int *uuid,
 736                   int subdevs, mddev_dev_t devlist,
 737                   int runstop, int verbose, int force, int assume_clean,
 738                   char *bitmap_file, int bitmap_chunk, int write_behind, int delay, int autof);
 739
 740 extern int Detail(char *dev, int brief, int export, int test, char *homehost);
 741 extern int Detail_Platform(struct superswitch *ss, int scan, int verbose);
 742 extern int Query(char *dev);
 743 extern int Examine(mddev_dev_t devlist, int brief, int export, int scan,
 744                    int SparcAdjust, struct supertype *forcest, char *homehost);
 745 extern int Monitor(mddev_dev_t devlist,
 746                    char *mailaddr, char *alert_cmd,
 747                    int period, int daemonise, int scan, int oneshot,
 748                    int dosyslog, int test, char *pidfile);
 749
 750 extern int Kill(char *dev, int force, int quiet, int noexcl);
 751 extern int Wait(char *dev);
 752 extern int WaitClean(char *dev, int verbose);
 753
 754 extern int Incremental(char *devname, int verbose, int runstop,
 755                        struct supertype *st, char *homehost, int autof);
 756 extern int Incremental_container(struct supertype *st, char *devname,
 757                                  int verbose, int runstop, int autof,
 758                                  int trustworthy);
 759 extern void RebuildMap(void);
 760 extern int IncrementalScan(int verbose);
 761
 762 extern int CreateBitmap(char *filename, int force, char uuid[16],
 763                         unsigned long chunksize, unsigned long daemon_sleep,
 764                         unsigned long write_behind,
 765                         unsigned long long array_size,
 766                         int major);
 767 extern int ExamineBitmap(char *filename, int brief, struct supertype *st);
 768 extern int bitmap_update_uuid(int fd, int *uuid, int swap);
 769 extern unsigned long bitmap_sectors(struct bitmap_super_s *bsb);
 770
 771 extern int md_get_version(int fd);
 772 extern int get_linux_version(void);
 773 extern int parse_uuid(char *str, int uuid[4]);
 774 extern int check_ext2(int fd, char *name);
 775 extern int check_reiser(int fd, char *name);
 776 extern int check_raid(int fd, char *name);
 777
 778 extern int get_mdp_major(void);
 779 extern int dev_open(char *dev, int flags);
 780 extern int open_dev(int devnum);
 781 extern int open_dev_excl(int devnum);
 782 extern int is_standard(char *dev, int *nump);
 783 extern int same_dev(char *one, char *two);
 784
 785 extern int parse_auto(char *str, char *msg, int config);
 786 extern mddev_ident_t conf_get_ident(char *dev);
 787 extern mddev_dev_t conf_get_devs(void);
 788 extern int conf_test_dev(char *devname);
 789 extern struct createinfo *conf_get_create_info(void);
 790 extern void set_conffile(char *file);
 791 extern char *conf_get_mailaddr(void);
 792 extern char *conf_get_mailfrom(void);
 793 extern char *conf_get_program(void);
 794 extern char *conf_get_homehost(void);
 795 extern char *conf_line(FILE *file);
 796 extern char *conf_word(FILE *file, int allow_key);
 797 extern void free_line(char *line);
 798 extern int match_oneof(char *devices, char *devname);
 799 extern void uuid_from_super(int uuid[4], mdp_super_t *super);
 800 extern const int uuid_match_any[4];
 801 extern int same_uuid(int a[4], int b[4], int swapuuid);
 802 extern void copy_uuid(void *a, int b[4], int swapuuid);
 803 extern char *fname_from_uuid(struct supertype *st,
 804                              struct mdinfo *info, char *buf, char sep);
 805 extern unsigned long calc_csum(void *super, int bytes);
 806 extern int enough(int level, int raid_disks, int layout, int clean,
 807                    char *avail, int avail_disks);
 808 extern int ask(char *mesg);
 809 extern unsigned long long get_component_size(int fd);
 810 extern void remove_partitions(int fd);
 811 extern unsigned long long calc_array_size(int level, int raid_disks, int layout,
 812                                    int chunksize, unsigned long long devsize);
 813 extern int flush_metadata_updates(struct supertype *st);
 814 extern void append_metadata_update(struct supertype *st, void *buf, int len);
 815 extern int assemble_container_content(struct supertype *st, int mdfd,
 816                                       struct mdinfo *content, int runstop,
 817                                       char *chosen_name, int verbose);
 818
 819 extern int add_disk(int mdfd, struct supertype *st,
 820                     struct mdinfo *sra, struct mdinfo *info);
 821 extern int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info);
 822
 823 extern char *human_size(long long bytes);
 824 extern char *human_size_brief(long long bytes);
 825 extern void print_r10_layout(int layout);
 826
 827 #define NoMdDev (1<<23)
 828 extern int find_free_devnum(int use_partitions);
 829
 830 extern void put_md_name(char *name);
 831 extern char *get_md_name(int dev);
 832
 833 extern char DefaultConfFile[];
 834
 835 extern int create_mddev(char *dev, char *name, int autof, int trustworthy,
 836                         char *chosen);
 837 /* values for 'trustworthy' */
 838 #define LOCAL   1
 839 #define FOREIGN 2
 840 #define METADATA 3
 841 extern int open_mddev(char *dev, int report_errors);
 842 extern int open_container(int fd);
 843
 844 extern int mdmon_running(int devnum);
 845 extern int signal_mdmon(int devnum);
 846 extern int check_env(char *name);
 847 extern int start_mdmon(int devnum);
 848
 849 extern char *devnum2devname(int num);
 850 extern int devname2devnum(char *name);
 851 extern int stat2devnum(struct stat *st);
 852 extern int fd2devnum(int fd);
 853
 854 static inline int dev2major(int d)
 855 {
 856         if (d >= 0)
 857                 return MD_MAJOR;
 858         else
 859                 return get_mdp_major();
 860 }
 861
 862 static inline int dev2minor(int d)
 863 {
 864         if (d >= 0)
 865                 return d;
 866         return (-1-d) << MdpMinorShift;
 867 }
 868
 869 static inline int ROUND_UP(int a, int base)
 870 {
 871         return ((a+base-1)/base)*base;
 872 }
 873
 874 static inline int is_subarray(char *vers)
 875 {
 876         /* The version string for a 'subarray' (an array in a container)
 877          * is
 878          *    /containername/componentname    for normal read-write arrays
 879          *    -containername/componentname    for read-only arrays.
 880          * containername is e.g. md0, md_d1
 881          * componentname is dependant on the metadata. e.g. '1' 'S1' ...
 882          */
 883         return (*vers == '/' || *vers == '-');
 884 }
 885
 886 #ifdef DEBUG
 887 #define dprintf(fmt, arg...) \
 888         fprintf(stderr, fmt, ##arg)
 889 #else
 890 #define dprintf(fmt, arg...) \
 891         ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
 892 #endif
 893 #include <assert.h>
 894 #include <stdarg.h>
 895 static inline int xasprintf(char **strp, const char *fmt, ...) {
 896         va_list ap;
 897         int ret;
 898         va_start(ap, fmt);
 899         ret = vasprintf(strp, fmt, ap);
 900         va_end(ap);
 901         assert(ret >= 0);
 902         return ret;
 903 }
 904
 905 #define LEVEL_MULTIPATH         (-4)
 906 #define LEVEL_LINEAR            (-1)
 907 #define LEVEL_FAULTY            (-5)
 908
 909 /* kernel module doesn't know about these */
 910 #define LEVEL_CONTAINER         (-100)
 911 #define LEVEL_UNSUPPORTED       (-200)
 912
 913
 914 /* faulty stuff */
 915
 916 #define WriteTransient  0
 917 #define ReadTransient   1
 918 #define WritePersistent 2
 919 #define ReadPersistent  3
 920 #define WriteAll        4 /* doesn't go to device */
 921 #define ReadFixable     5
 922 #define Modes   6
 923
 924 #define ClearErrors     31
 925 #define ClearFaults     30
 926
 927 #define AllPersist      100 /* internal use only */
 928 #define NoPersist       101
 929
 930 #define ModeMask        0x1f
 931 #define ModeShift       5
 932
 933
 934 #ifdef __TINYC__
 935 #undef minor
 936 #undef major
 937 #undef makedev
 938 #define minor(x) ((x)&0xff)
 939 #define major(x) (((x)>>8)&0xff)
 940 #define makedev(M,m) (((M)<<8) | (m))
 941 #endif
 942
 943 /* for raid4/5/6 */
 944 #define ALGORITHM_LEFT_ASYMMETRIC       0
 945 #define ALGORITHM_RIGHT_ASYMMETRIC      1
 946 #define ALGORITHM_LEFT_SYMMETRIC        2
 947 #define ALGORITHM_RIGHT_SYMMETRIC       3
 948
 949 /* Define non-rotating (raid4) algorithms.  These allow
 950  * conversion of raid4 to raid5.
 951  */
 952 #define ALGORITHM_PARITY_0              4 /* P or P,Q are initial devices */
 953 #define ALGORITHM_PARITY_N              5 /* P or P,Q are final devices. */
 954
 955 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
 956  * Firstly, the exact positioning of the parity block is slightly
 957  * different between the 'LEFT_*' modes of md and the "_N_*" modes
 958  * of DDF.
 959  * Secondly, or order of datablocks over which the Q syndrome is computed
 960  * is different.
 961  * Consequently we have different layouts for DDF/raid6 than md/raid6.
 962  * These layouts are from the DDFv1.2 spec.
 963  * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
 964  * leaves RLQ=3 as 'Vendor Specific'
 965  */
 966
 967 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
 968 #define ALGORITHM_ROTATING_N_RESTART    9 /* DDF PRL=6 RLQ=2 */
 969 #define ALGORITHM_ROTATING_N_CONTINUE   10 /*DDF PRL=6 RLQ=3 */
 970
 971
 972 /* For every RAID5 algorithm we define a RAID6 algorithm
 973  * with exactly the same layout for data and parity, and
 974  * with the Q block always on the last device (N-1).
 975  * This allows trivial conversion from RAID5 to RAID6
 976  */
 977 #define ALGORITHM_LEFT_ASYMMETRIC_6     16
 978 #define ALGORITHM_RIGHT_ASYMMETRIC_6    17
 979 #define ALGORITHM_LEFT_SYMMETRIC_6      18
 980 #define ALGORITHM_RIGHT_SYMMETRIC_6     19
 981 #define ALGORITHM_PARITY_0_6            20
 982 #define ALGORITHM_PARITY_N_6            ALGORITHM_PARITY_N
 983