super0.c

   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 HAVE_STDINT_H 1
  26 #include "mdadm.h"
  27 #include "sha1.h"
  28 /*
  29  * All handling for the 0.90.0 version superblock is in
  30  * this file.
  31  * This includes:
  32  *   - finding, loading, and writing the superblock.
  33  *   - initialising a new superblock
  34  *   - printing the superblock for --examine
  35  *   - printing part of the superblock for --detail
  36  * .. other stuff
  37  */
  38
  39 static unsigned long calc_sb0_csum(mdp_super_t *super)
  40 {
  41         unsigned long csum = super->sb_csum;
  42         unsigned long newcsum;
  43         super->sb_csum= 0 ;
  44         newcsum = calc_csum(super, MD_SB_BYTES);
  45         super->sb_csum = csum;
  46         return newcsum;
  47 }
  48
  49 static void super0_swap_endian(struct mdp_superblock_s *sb)
  50 {
  51         /* as super0 superblocks are host-endian, it is sometimes
  52          * useful to be able to swap the endianness
  53          * as (almost) everything is u32's we byte-swap every 4byte
  54          * number.
  55          * We then also have to swap the events_hi and events_lo
  56          */
  57         char *sbc = (char *)sb;
  58         __u32 t32;
  59         int i;
  60
  61         for (i=0; i < MD_SB_BYTES ; i+=4) {
  62                 char t = sbc[i];
  63                 sbc[i] = sbc[i+3];
  64                 sbc[i+3] = t;
  65                 t=sbc[i+1];
  66                 sbc[i+1]=sbc[i+2];
  67                 sbc[i+2]=t;
  68         }
  69         t32 = sb->events_hi;
  70         sb->events_hi = sb->events_lo;
  71         sb->events_lo = t32;
  72
  73         t32 = sb->cp_events_hi;
  74         sb->cp_events_hi = sb->cp_events_lo;
  75         sb->cp_events_lo = t32;
  76
  77 }
  78
  79 static void examine_super0(struct supertype *st, char *homehost)
  80 {
  81         mdp_super_t *sb = st->sb;
  82         time_t atime;
  83         int d;
  84         int delta_extra = 0;
  85         char *c;
  86
  87         printf("          Magic : %08x\n", sb->md_magic);
  88         printf("        Version : %d.%02d.%02d\n",
  89                sb->major_version, sb->minor_version, sb->patch_version);
  90         if (sb->minor_version >= 90) {
  91                 printf("           UUID : %08x:%08x:%08x:%08x", sb->set_uuid0,
  92                        sb->set_uuid1, sb->set_uuid2, sb->set_uuid3);
  93                 if (homehost) {
  94                         char buf[20];
  95                         void *hash;
  96
  97                         hash = sha1_buffer(homehost, strlen(homehost), buf);
  98                         if (memcmp(&sb->set_uuid2, hash, 8) == 0)
  99                                 printf(" (local to host %s)", homehost);
 100                 }
 101                 printf("\n");
 102         } else
 103                 printf("           UUID : %08x\n", sb->set_uuid0);
 104
 105         if (sb->not_persistent)
 106                 printf("           Eedk : not persistent\n");
 107
 108         atime = sb->ctime;
 109         printf("  Creation Time : %.24s\n", ctime(&atime));
 110         c = map_num(pers, sb->level);
 111         printf("     Raid Level : %s\n", c?c:"-unknown-");
 112         if ((int)sb->level > 0) {
 113                 int ddsks = 0, ddsks_denom = 1;
 114                 printf("  Used Dev Size : %d%s\n", sb->size,
 115                        human_size((long long)sb->size<<10));
 116                 switch(sb->level) {
 117                 case 1:
 118                         ddsks=1;
 119                         break;
 120                 case 4:
 121                 case 5:
 122                         ddsks = sb->raid_disks - 1;
 123                         break;
 124                 case 6:
 125                         ddsks = sb->raid_disks - 2;
 126                         break;
 127                 case 10:
 128                         ddsks = sb->raid_disks;
 129                         ddsks_denom =
 130                                 (sb->layout & 255) * ((sb->layout >> 8) & 255);
 131                 }
 132                 if (ddsks) {
 133                         long long asize = sb->size;
 134                         asize = (asize << 10) * ddsks / ddsks_denom;
 135                         printf("     Array Size : %llu%s\n",
 136                                asize >> 10,  human_size(asize));
 137                 }
 138         }
 139         printf("   Raid Devices : %d\n", sb->raid_disks);
 140         printf("  Total Devices : %d\n", sb->nr_disks);
 141         printf("Preferred Minor : %d\n", sb->md_minor);
 142         printf("\n");
 143         if (sb->minor_version > 90 && (sb->reshape_position + 1) != 0) {
 144                 printf("  Reshape pos'n : %llu%s\n",
 145                        (unsigned long long)sb->reshape_position / 2,
 146                        human_size((long long)sb->reshape_position << 9));
 147                 if (sb->delta_disks) {
 148                         printf("  Delta Devices : %d", sb->delta_disks);
 149                         printf(" (%d->%d)\n", sb->raid_disks-sb->delta_disks,
 150                                sb->raid_disks);
 151                         if (((int)sb->delta_disks) < 0)
 152                                 delta_extra = - sb->delta_disks;
 153                 }
 154                 if (sb->new_level != sb->level) {
 155                         c = map_num(pers, sb->new_level);
 156                         printf("      New Level : %s\n", c?c:"-unknown-");
 157                 }
 158                 if (sb->new_layout != sb->layout) {
 159                         if (sb->level == 5) {
 160                                 c = map_num(r5layout, sb->new_layout);
 161                                 printf("     New Layout : %s\n",
 162                                        c?c:"-unknown-");
 163                         }
 164                         if (sb->level == 6) {
 165                                 c = map_num(r6layout, sb->new_layout);
 166                                 printf("     New Layout : %s\n",
 167                                        c?c:"-unknown-");
 168                         }
 169                         if (sb->level == 10) {
 170                                 printf("     New Layout : near=%d, %s=%d\n",
 171                                        sb->new_layout&255,
 172                                        (sb->new_layout&0x10000)?"offset":"far",
 173                                        (sb->new_layout>>8)&255);
 174                         }
 175                 }
 176                 if (sb->new_chunk != sb->chunk_size)
 177                         printf("  New Chunksize : %d\n", sb->new_chunk);
 178                 printf("\n");
 179         }
 180         atime = sb->utime;
 181         printf("    Update Time : %.24s\n", ctime(&atime));
 182         printf("          State : %s\n",
 183                (sb->state&(1 << MD_SB_CLEAN)) ? "clean":"active");
 184         if (sb->state & (1 << MD_SB_BITMAP_PRESENT))
 185                 printf("Internal Bitmap : present\n");
 186         printf(" Active Devices : %d\n", sb->active_disks);
 187         printf("Working Devices : %d\n", sb->working_disks);
 188         printf(" Failed Devices : %d\n", sb->failed_disks);
 189         printf("  Spare Devices : %d\n", sb->spare_disks);
 190         if (calc_sb0_csum(sb) == sb->sb_csum)
 191                 printf("       Checksum : %x - correct\n", sb->sb_csum);
 192         else
 193                 printf("       Checksum : %x - expected %lx\n",
 194                        sb->sb_csum, calc_sb0_csum(sb));
 195         printf("         Events : %llu\n",
 196                ((unsigned long long)sb->events_hi << 32) + sb->events_lo);
 197         printf("\n");
 198         if (sb->level == 5) {
 199                 c = map_num(r5layout, sb->layout);
 200                 printf("         Layout : %s\n", c?c:"-unknown-");
 201         }
 202         if (sb->level == 6) {
 203                 c = map_num(r6layout, sb->layout);
 204                 printf("         Layout : %s\n", c?c:"-unknown-");
 205         }
 206         if (sb->level == 10) {
 207                 printf("         Layout :");
 208                 print_r10_layout(sb->layout);
 209                 printf("\n");
 210         }
 211         switch(sb->level) {
 212         case 0:
 213         case 4:
 214         case 5:
 215         case 6:
 216         case 10:
 217                 printf("     Chunk Size : %dK\n", sb->chunk_size / 1024);
 218                 break;
 219         case -1:
 220                 printf("       Rounding : %dK\n", sb->chunk_size / 1024);
 221                 break;
 222         default:
 223                 break;
 224         }
 225         printf("\n");
 226         printf("      Number   Major   Minor   RaidDevice State\n");
 227         for (d = -1;
 228              d < (signed int)(sb->raid_disks + delta_extra + sb->spare_disks);
 229              d++) {
 230                 mdp_disk_t *dp;
 231                 char *dv;
 232                 char nb[11];
 233                 int wonly, failfast;
 234                 if (d>=0) dp = &sb->disks[d];
 235                 else dp = &sb->this_disk;
 236                 snprintf(nb, sizeof(nb), "%4d", d);
 237                 printf("%4s %5d   %5d    %5d    %5d     ", d < 0 ? "this" : nb,
 238                        dp->number, dp->major, dp->minor, dp->raid_disk);
 239                 wonly = dp->state & (1 << MD_DISK_WRITEMOSTLY);
 240                 failfast = dp->state & (1<<MD_DISK_FAILFAST);
 241                 dp->state &= ~(wonly | failfast);
 242                 if (dp->state & (1 << MD_DISK_FAULTY))
 243                         printf(" faulty");
 244                 if (dp->state & (1 << MD_DISK_ACTIVE))
 245                         printf(" active");
 246                 if (dp->state & (1 << MD_DISK_SYNC))
 247                         printf(" sync");
 248                 if (dp->state & (1 << MD_DISK_REMOVED))
 249                         printf(" removed");
 250                 if (wonly)
 251                         printf(" write-mostly");
 252                 if (failfast)
 253                         printf(" failfast");
 254                 if (dp->state == 0)
 255                         printf(" spare");
 256                 if ((dv = map_dev(dp->major, dp->minor, 0)))
 257                         printf("   %s", dv);
 258                 printf("\n");
 259                 if (d == -1)
 260                         printf("\n");
 261         }
 262 }
 263
 264 static void brief_examine_super0(struct supertype *st, int verbose)
 265 {
 266         mdp_super_t *sb = st->sb;
 267         char *c=map_num(pers, sb->level);
 268         char devname[20];
 269
 270         sprintf(devname, "/dev/md%d", sb->md_minor);
 271
 272         if (verbose) {
 273                 printf("ARRAY %s level=%s num-devices=%d",
 274                        devname,
 275                        c?c:"-unknown-", sb->raid_disks);
 276         } else
 277                 printf("ARRAY %s", devname);
 278
 279         if (sb->minor_version >= 90)
 280                 printf(" UUID=%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
 281                        sb->set_uuid2, sb->set_uuid3);
 282         else
 283                 printf(" UUID=%08x", sb->set_uuid0);
 284         printf("\n");
 285 }
 286
 287 static void export_examine_super0(struct supertype *st)
 288 {
 289         mdp_super_t *sb = st->sb;
 290
 291         printf("MD_LEVEL=%s\n", map_num(pers, sb->level));
 292         printf("MD_DEVICES=%d\n", sb->raid_disks);
 293         if (sb->minor_version >= 90)
 294                 printf("MD_UUID=%08x:%08x:%08x:%08x\n",
 295                        sb->set_uuid0, sb->set_uuid1,
 296                        sb->set_uuid2, sb->set_uuid3);
 297         else
 298                 printf("MD_UUID=%08x\n", sb->set_uuid0);
 299         printf("MD_UPDATE_TIME=%llu\n",
 300                __le64_to_cpu(sb->ctime) & 0xFFFFFFFFFFULL);
 301         printf("MD_EVENTS=%llu\n",
 302                ((unsigned long long)sb->events_hi << 32)
 303                + sb->events_lo);
 304 }
 305
 306 static int copy_metadata0(struct supertype *st, int from, int to)
 307 {
 308         /* Read 64K from the appropriate offset of 'from'
 309          * and if it looks a little like a 0.90 superblock,
 310          * write it to the same offset of 'to'
 311          */
 312         void *buf;
 313         unsigned long long dsize, offset;
 314         const int bufsize = 64*1024;
 315         mdp_super_t *super;
 316
 317         if (posix_memalign(&buf, 4096, bufsize) != 0)
 318                 return 1;
 319
 320         if (!get_dev_size(from, NULL, &dsize))
 321                 goto err;
 322
 323         if (dsize < MD_RESERVED_SECTORS*512)
 324                 goto err;
 325
 326         offset = MD_NEW_SIZE_SECTORS(dsize>>9);
 327
 328         offset *= 512;
 329
 330         if (lseek64(from, offset, 0) < 0LL)
 331                 goto err;
 332         if (read(from, buf, bufsize) != bufsize)
 333                 goto err;
 334
 335         if (lseek64(to, offset, 0) < 0LL)
 336                 goto err;
 337         super = buf;
 338         if (super->md_magic != MD_SB_MAGIC ||
 339             super->major_version != 0 ||
 340             calc_sb0_csum(super) != super->sb_csum)
 341                 goto err;
 342         if (write(to, buf, bufsize) != bufsize)
 343                 goto err;
 344         free(buf);
 345         return 0;
 346 err:
 347         free(buf);
 348         return 1;
 349 }
 350
 351 static void detail_super0(struct supertype *st, char *homehost)
 352 {
 353         mdp_super_t *sb = st->sb;
 354         printf("              UUID : ");
 355         if (sb->minor_version >= 90)
 356                 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
 357                        sb->set_uuid2, sb->set_uuid3);
 358         else
 359                 printf("%08x", sb->set_uuid0);
 360         if (homehost) {
 361                 char buf[20];
 362                 void *hash = sha1_buffer(homehost,
 363                                          strlen(homehost),
 364                                          buf);
 365                 if (memcmp(&sb->set_uuid2, hash, 8)==0)
 366                         printf(" (local to host %s)", homehost);
 367         }
 368         printf("\n            Events : %d.%d\n\n", sb->events_hi, sb->events_lo);
 369 }
 370
 371 static void brief_detail_super0(struct supertype *st)
 372 {
 373         mdp_super_t *sb = st->sb;
 374         printf(" UUID=");
 375         if (sb->minor_version >= 90)
 376                 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
 377                        sb->set_uuid2, sb->set_uuid3);
 378         else
 379                 printf("%08x", sb->set_uuid0);
 380 }
 381
 382 static int match_home0(struct supertype *st, char *homehost)
 383 {
 384         mdp_super_t *sb = st->sb;
 385         char buf[20];
 386         char *hash;
 387
 388         if (!homehost)
 389                 return 0;
 390         hash = sha1_buffer(homehost,
 391                            strlen(homehost),
 392                            buf);
 393
 394         return (memcmp(&sb->set_uuid2, hash, 8)==0);
 395 }
 396
 397 static void uuid_from_super0(struct supertype *st, int uuid[4])
 398 {
 399         mdp_super_t *super = st->sb;
 400         uuid[0] = super->set_uuid0;
 401         if (super->minor_version >= 90) {
 402                 uuid[1] = super->set_uuid1;
 403                 uuid[2] = super->set_uuid2;
 404                 uuid[3] = super->set_uuid3;
 405         } else {
 406                 uuid[1] = 0;
 407                 uuid[2] = 0;
 408                 uuid[3] = 0;
 409         }
 410 }
 411
 412 static void getinfo_super0(struct supertype *st, struct mdinfo *info, char *map)
 413 {
 414         mdp_super_t *sb = st->sb;
 415         int working = 0;
 416         int i;
 417         int map_disks = info->array.raid_disks;
 418
 419         memset(info, 0, sizeof(*info));
 420         info->array.major_version = sb->major_version;
 421         info->array.minor_version = sb->minor_version;
 422         info->array.patch_version = sb->patch_version;
 423         info->array.raid_disks = sb->raid_disks;
 424         info->array.level = sb->level;
 425         info->array.layout = sb->layout;
 426         info->array.md_minor = sb->md_minor;
 427         info->array.ctime = sb->ctime;
 428         info->array.utime = sb->utime;
 429         info->array.chunk_size = sb->chunk_size;
 430         info->array.state = sb->state;
 431         info->component_size = sb->size;
 432         info->component_size *= 2;
 433
 434         if (sb->state & (1<<MD_SB_BITMAP_PRESENT))
 435                 info->bitmap_offset = 8;
 436
 437         info->disk.state = sb->this_disk.state;
 438         info->disk.major = sb->this_disk.major;
 439         info->disk.minor = sb->this_disk.minor;
 440         info->disk.raid_disk = sb->this_disk.raid_disk;
 441         info->disk.number = sb->this_disk.number;
 442
 443         info->events = md_event(sb);
 444         info->data_offset = 0;
 445
 446         sprintf(info->text_version, "0.%d", sb->minor_version);
 447         info->safe_mode_delay = 200;
 448
 449         uuid_from_super0(st, info->uuid);
 450
 451         info->recovery_start = MaxSector;
 452         if (sb->minor_version > 90 && (sb->reshape_position+1) != 0) {
 453                 info->reshape_active = 1;
 454                 info->reshape_progress = sb->reshape_position;
 455                 info->new_level = sb->new_level;
 456                 info->delta_disks = sb->delta_disks;
 457                 info->new_layout = sb->new_layout;
 458                 info->new_chunk = sb->new_chunk;
 459                 if (info->delta_disks < 0)
 460                         info->array.raid_disks -= info->delta_disks;
 461         } else
 462                 info->reshape_active = 0;
 463
 464         info->recovery_blocked = info->reshape_active;
 465
 466         sprintf(info->name, "%d", sb->md_minor);
 467         /* work_disks is calculated rather than read directly */
 468         for (i=0; i < MD_SB_DISKS; i++)
 469                 if ((sb->disks[i].state & (1<<MD_DISK_SYNC)) &&
 470                     (sb->disks[i].raid_disk < (unsigned)info->array.raid_disks) &&
 471                     (sb->disks[i].state & (1<<MD_DISK_ACTIVE)) &&
 472                     !(sb->disks[i].state & (1<<MD_DISK_FAULTY))) {
 473                         working ++;
 474                         if (map && i < map_disks)
 475                                 map[i] = 1;
 476                 } else if (map && i < map_disks)
 477                         map[i] = 0;
 478         info->array.working_disks = working;
 479 }
 480
 481 static struct mdinfo *container_content0(struct supertype *st, char *subarray)
 482 {
 483         struct mdinfo *info;
 484
 485         if (subarray)
 486                 return NULL;
 487
 488         info = xmalloc(sizeof(*info));
 489         getinfo_super0(st, info, NULL);
 490         return info;
 491 }
 492
 493 static int update_super0(struct supertype *st, struct mdinfo *info,
 494                          char *update,
 495                          char *devname, int verbose,
 496                          int uuid_set, char *homehost)
 497 {
 498         /* NOTE: for 'assemble' and 'force' we need to return non-zero
 499          * if any change was made.  For others, the return value is
 500          * ignored.
 501          */
 502         int rv = 0;
 503         int uuid[4];
 504         mdp_super_t *sb = st->sb;
 505
 506         if (strcmp(update, "homehost") == 0 &&
 507             homehost) {
 508                 /* note that 'homehost' is special as it is really
 509                  * a "uuid" update.
 510                  */
 511                 uuid_set = 0;
 512                 update = "uuid";
 513                 info->uuid[0] = sb->set_uuid0;
 514                 info->uuid[1] = sb->set_uuid1;
 515         }
 516
 517         if (strcmp(update, "sparc2.2")==0 ) {
 518                 /* 2.2 sparc put the events in the wrong place
 519                  * So we copy the tail of the superblock
 520                  * up 4 bytes before continuing
 521                  */
 522                 __u32 *sb32 = (__u32*)sb;
 523                 memcpy(sb32+MD_SB_GENERIC_CONSTANT_WORDS+7,
 524                        sb32+MD_SB_GENERIC_CONSTANT_WORDS+7+1,
 525                        (MD_SB_WORDS - (MD_SB_GENERIC_CONSTANT_WORDS+7+1))*4);
 526                 if (verbose >= 0)
 527                         pr_err("adjusting superblock of %s for 2.2/sparc compatibility.\n",
 528                                devname);
 529         } else if (strcmp(update, "super-minor") ==0) {
 530                 sb->md_minor = info->array.md_minor;
 531                 if (verbose > 0)
 532                         pr_err("updating superblock of %s with minor number %d\n",
 533                                 devname, info->array.md_minor);
 534         } else if (strcmp(update, "summaries") == 0) {
 535                 unsigned int i;
 536                 /* set nr_disks, active_disks, working_disks,
 537                  * failed_disks, spare_disks based on disks[]
 538                  * array in superblock.
 539                  * Also make sure extra slots aren't 'failed'
 540                  */
 541                 sb->nr_disks = sb->active_disks =
 542                         sb->working_disks = sb->failed_disks =
 543                         sb->spare_disks = 0;
 544                 for (i=0; i < MD_SB_DISKS ; i++)
 545                         if (sb->disks[i].major ||
 546                             sb->disks[i].minor) {
 547                                 int state = sb->disks[i].state;
 548                                 if (state & (1<<MD_DISK_REMOVED))
 549                                         continue;
 550                                 sb->nr_disks++;
 551                                 if (state & (1<<MD_DISK_ACTIVE))
 552                                         sb->active_disks++;
 553                                 if (state & (1<<MD_DISK_FAULTY))
 554                                         sb->failed_disks++;
 555                                 else
 556                                         sb->working_disks++;
 557                                 if (state == 0)
 558                                         sb->spare_disks++;
 559                         } else if (i >= sb->raid_disks && sb->disks[i].number == 0)
 560                                 sb->disks[i].state = 0;
 561         } else if (strcmp(update, "force-one")==0) {
 562                 /* Not enough devices for a working array, so
 563                  * bring this one up-to-date.
 564                  */
 565                 __u32 ehi = sb->events_hi, elo = sb->events_lo;
 566                 sb->events_hi = (info->events>>32) & 0xFFFFFFFF;
 567                 sb->events_lo = (info->events) & 0xFFFFFFFF;
 568                 if (sb->events_hi != ehi ||
 569                     sb->events_lo != elo)
 570                         rv = 1;
 571         } else if (strcmp(update, "force-array")==0) {
 572                 /* degraded array and 'force' requested, so
 573                  * maybe need to mark it 'clean'
 574                  */
 575                 if ((sb->level == 5 || sb->level == 4 || sb->level == 6) &&
 576                     (sb->state & (1 << MD_SB_CLEAN)) == 0) {
 577                         /* need to force clean */
 578                         sb->state |= (1 << MD_SB_CLEAN);
 579                         rv = 1;
 580                 }
 581         } else if (strcmp(update, "assemble")==0) {
 582                 int d = info->disk.number;
 583                 int wonly = sb->disks[d].state & (1<<MD_DISK_WRITEMOSTLY);
 584                 int failfast = sb->disks[d].state & (1<<MD_DISK_FAILFAST);
 585                 int mask = (1<<MD_DISK_WRITEMOSTLY)|(1<<MD_DISK_FAILFAST);
 586                 int add = 0;
 587                 if (sb->minor_version >= 91)
 588                         /* During reshape we don't insist on everything
 589                          * being marked 'sync'
 590                          */
 591                         add = (1<<MD_DISK_SYNC);
 592                 if (((sb->disks[d].state & ~mask) | add) !=
 593                     (unsigned)info->disk.state) {
 594                         sb->disks[d].state = info->disk.state | wonly |failfast;
 595                         rv = 1;
 596                 }
 597                 if (info->reshape_active &&
 598                     sb->minor_version > 90 && (sb->reshape_position+1) != 0 &&
 599                     info->delta_disks >= 0 &&
 600                     info->reshape_progress < sb->reshape_position) {
 601                         sb->reshape_position = info->reshape_progress;
 602                         rv = 1;
 603                 }
 604                 if (info->reshape_active &&
 605                     sb->minor_version > 90 && (sb->reshape_position+1) != 0 &&
 606                     info->delta_disks < 0 &&
 607                     info->reshape_progress > sb->reshape_position) {
 608                         sb->reshape_position = info->reshape_progress;
 609                         rv = 1;
 610                 }
 611         } else if (strcmp(update, "linear-grow-new") == 0) {
 612                 memset(&sb->disks[info->disk.number], 0, sizeof(sb->disks[0]));
 613                 sb->disks[info->disk.number].number = info->disk.number;
 614                 sb->disks[info->disk.number].major = info->disk.major;
 615                 sb->disks[info->disk.number].minor = info->disk.minor;
 616                 sb->disks[info->disk.number].raid_disk = info->disk.raid_disk;
 617                 sb->disks[info->disk.number].state = info->disk.state;
 618                 sb->this_disk = sb->disks[info->disk.number];
 619         } else if (strcmp(update, "linear-grow-update") == 0) {
 620                 sb->raid_disks = info->array.raid_disks;
 621                 sb->nr_disks = info->array.nr_disks;
 622                 sb->active_disks = info->array.active_disks;
 623                 sb->working_disks = info->array.working_disks;
 624                 memset(&sb->disks[info->disk.number], 0, sizeof(sb->disks[0]));
 625                 sb->disks[info->disk.number].number = info->disk.number;
 626                 sb->disks[info->disk.number].major = info->disk.major;
 627                 sb->disks[info->disk.number].minor = info->disk.minor;
 628                 sb->disks[info->disk.number].raid_disk = info->disk.raid_disk;
 629                 sb->disks[info->disk.number].state = info->disk.state;
 630         } else if (strcmp(update, "resync") == 0) {
 631                 /* make sure resync happens */
 632                 sb->state &= ~(1<<MD_SB_CLEAN);
 633                 sb->recovery_cp = 0;
 634         } else if (strcmp(update, "uuid") == 0) {
 635                 if (!uuid_set && homehost) {
 636                         char buf[20];
 637                         char *hash = sha1_buffer(homehost,
 638                                                  strlen(homehost),
 639                                                  buf);
 640                         memcpy(info->uuid+2, hash, 8);
 641                 }
 642                 sb->set_uuid0 = info->uuid[0];
 643                 sb->set_uuid1 = info->uuid[1];
 644                 sb->set_uuid2 = info->uuid[2];
 645                 sb->set_uuid3 = info->uuid[3];
 646                 if (sb->state & (1<<MD_SB_BITMAP_PRESENT)) {
 647                         struct bitmap_super_s *bm;
 648                         bm = (struct bitmap_super_s*)(sb+1);
 649                         uuid_from_super0(st, uuid);
 650                         memcpy(bm->uuid, uuid, 16);
 651                 }
 652         } else if (strcmp(update, "metadata") == 0) {
 653                 /* Create some v1.0 metadata to match ours but make the
 654                  * ctime bigger.  Also update info->array.*_version.
 655                  * We need to arrange that store_super writes out
 656                  * the v1.0 metadata.
 657                  * Not permitted for unclean array, or array with
 658                  * bitmap.
 659                  */
 660                 if (info->bitmap_offset) {
 661                         pr_err("Cannot update metadata when bitmap is present\n");
 662                         rv = -2;
 663                 } else if (info->array.state != 1) {
 664                         pr_err("Cannot update metadata on unclean array\n");
 665                         rv = -2;
 666                 } else {
 667                         info->array.major_version = 1;
 668                         info->array.minor_version = 0;
 669                         uuid_from_super0(st, info->uuid);
 670                         st->other = super1_make_v0(st, info, st->sb);
 671                 }
 672         } else if (strcmp(update, "revert-reshape") == 0) {
 673                 rv = -2;
 674                 if (sb->minor_version <= 90)
 675                         pr_err("No active reshape to revert on %s\n",
 676                                devname);
 677                 else if (sb->delta_disks == 0)
 678                         pr_err("%s: Can only revert reshape which changes number of devices\n",
 679                                devname);
 680                 else {
 681                         int tmp;
 682                         int parity = sb->level == 6 ? 2 : 1;
 683                         rv = 0;
 684
 685                         if (sb->level >= 4 && sb->level <= 6 &&
 686                             sb->reshape_position % (
 687                                     sb->new_chunk/512 *
 688                                     (sb->raid_disks - sb->delta_disks - parity))) {
 689                                 pr_err("Reshape position is not suitably aligned.\n");
 690                                 pr_err("Try normal assembly and stop again\n");
 691                                 return -2;
 692                         }
 693                         sb->raid_disks -= sb->delta_disks;
 694                         sb->delta_disks = -sb->delta_disks;
 695
 696                         tmp = sb->new_layout;
 697                         sb->new_layout = sb->layout;
 698                         sb->layout = tmp;
 699
 700                         tmp = sb->new_chunk;
 701                         sb->new_chunk = sb->chunk_size;
 702                         sb->chunk_size = tmp;
 703                 }
 704         } else if (strcmp(update, "no-bitmap") == 0) {
 705                 sb->state &= ~(1<<MD_SB_BITMAP_PRESENT);
 706         } else if (strcmp(update, "_reshape_progress")==0)
 707                 sb->reshape_position = info->reshape_progress;
 708         else if (strcmp(update, "writemostly")==0)
 709                 sb->state |= (1<<MD_DISK_WRITEMOSTLY);
 710         else if (strcmp(update, "readwrite")==0)
 711                 sb->state &= ~(1<<MD_DISK_WRITEMOSTLY);
 712         else
 713                 rv = -1;
 714
 715         sb->sb_csum = calc_sb0_csum(sb);
 716         return rv;
 717 }
 718
 719 /*
 720  * For version-0 superblock, the homehost is 'stored' in the uuid.
 721  * 8 bytes for a hash of the host leaving 8 bytes of random material.
 722  * We use the first 8 bytes (64bits) of the sha1 of the host name
 723  */
 724 static int init_super0(struct supertype *st, mdu_array_info_t *info,
 725                        struct shape *s, char *ignored_name,
 726                        char *homehost, int *uuid,
 727                        unsigned long long data_offset)
 728 {
 729         mdp_super_t *sb;
 730         int spares;
 731
 732         if (data_offset != INVALID_SECTORS) {
 733                 pr_err("data-offset not support for 0.90\n");
 734                 return 0;
 735         }
 736
 737         if (posix_memalign((void**)&sb, 4096,
 738                            MD_SB_BYTES + ROUND_UP(sizeof(bitmap_super_t), 4096)) != 0) {
 739                 pr_err("could not allocate superblock\n");
 740                 return 0;
 741         }
 742         memset(sb, 0, MD_SB_BYTES + sizeof(bitmap_super_t));
 743
 744         st->sb = sb;
 745         if (info == NULL) {
 746                 /* zeroing the superblock */
 747                 return 0;
 748         }
 749
 750         spares = info->working_disks - info->active_disks;
 751         if (info->raid_disks + spares  > MD_SB_DISKS) {
 752                 pr_err("too many devices requested: %d+%d > %d\n",
 753                         info->raid_disks , spares, MD_SB_DISKS);
 754                 return 0;
 755         }
 756
 757         sb->md_magic = MD_SB_MAGIC;
 758         sb->major_version = 0;
 759         sb->minor_version = 90;
 760         sb->patch_version = 0;
 761         sb->gvalid_words = 0; /* ignored */
 762         sb->ctime = time(0);
 763         sb->level = info->level;
 764         sb->size = s->size;
 765         if (s->size != (unsigned long long)sb->size)
 766                 return 0;
 767         sb->nr_disks = info->nr_disks;
 768         sb->raid_disks = info->raid_disks;
 769         sb->md_minor = info->md_minor;
 770         sb->not_persistent = 0;
 771         if (uuid) {
 772                 sb->set_uuid0 = uuid[0];
 773                 sb->set_uuid1 = uuid[1];
 774                 sb->set_uuid2 = uuid[2];
 775                 sb->set_uuid3 = uuid[3];
 776         } else {
 777                 __u32 r[4];
 778                 random_uuid((__u8 *)r);
 779                 sb->set_uuid0 = r[0];
 780                 sb->set_uuid1 = r[1];
 781                 sb->set_uuid2 = r[2];
 782                 sb->set_uuid3 = r[3];
 783         }
 784         if (homehost && !uuid) {
 785                 char buf[20];
 786                 char *hash = sha1_buffer(homehost,
 787                                          strlen(homehost),
 788                                          buf);
 789                 memcpy(&sb->set_uuid2, hash, 8);
 790         }
 791
 792         sb->utime = sb->ctime;
 793         sb->state = info->state;
 794         sb->active_disks = info->active_disks;
 795         sb->working_disks = info->working_disks;
 796         sb->failed_disks = info->failed_disks;
 797         sb->spare_disks = info->spare_disks;
 798         sb->events_hi = 0;
 799         sb->events_lo = 1;
 800
 801         sb->layout = info->layout;
 802         sb->chunk_size = info->chunk_size;
 803
 804         return 1;
 805 }
 806
 807 struct devinfo {
 808         int fd;
 809         char *devname;
 810         mdu_disk_info_t disk;
 811         struct devinfo *next;
 812 };
 813
 814 /* Add a device to the superblock being created */
 815 static int add_to_super0(struct supertype *st, mdu_disk_info_t *dinfo,
 816                          int fd, char *devname, unsigned long long data_offset)
 817 {
 818         mdp_super_t *sb = st->sb;
 819         mdp_disk_t *dk = &sb->disks[dinfo->number];
 820         struct devinfo *di, **dip;
 821
 822         dk->number = dinfo->number;
 823         dk->major = dinfo->major;
 824         dk->minor = dinfo->minor;
 825         dk->raid_disk = dinfo->raid_disk;
 826         dk->state = dinfo->state & ((1<<MD_DISK_ACTIVE) |
 827                                     (1<<MD_DISK_SYNC));
 828
 829         sb->this_disk = sb->disks[dinfo->number];
 830         sb->sb_csum = calc_sb0_csum(sb);
 831
 832         dip = (struct devinfo **)&st->info;
 833         while (*dip)
 834                 dip = &(*dip)->next;
 835         di = xmalloc(sizeof(struct devinfo));
 836         di->fd = fd;
 837         di->devname = devname;
 838         di->disk = *dinfo;
 839         di->next = NULL;
 840         *dip = di;
 841
 842         return 0;
 843 }
 844
 845 static int store_super0(struct supertype *st, int fd)
 846 {
 847         unsigned long long dsize;
 848         unsigned long long offset;
 849         mdp_super_t *super = st->sb;
 850
 851         if (!get_dev_size(fd, NULL, &dsize))
 852                 return 1;
 853
 854         if (dsize < MD_RESERVED_SECTORS*512)
 855                 return 2;
 856
 857         if (st->other) {
 858                 /* Writing out v1.0 metadata for --update=metadata */
 859                 int ret = 0;
 860
 861                 offset = dsize/512 - 8*2;
 862                 offset &= ~(4*2-1);
 863                 offset *= 512;
 864                 if (lseek64(fd, offset, 0)< 0LL)
 865                         ret = 3;
 866                 else if (write(fd, st->other, 1024) != 1024)
 867                         ret = 4;
 868                 else
 869                         fsync(fd);
 870                 free(st->other);
 871                 st->other = NULL;
 872                 return ret;
 873         }
 874
 875         offset = MD_NEW_SIZE_SECTORS(dsize>>9);
 876
 877         offset *= 512;
 878
 879         if (lseek64(fd, offset, 0)< 0LL)
 880                 return 3;
 881
 882         if (write(fd, super, sizeof(*super)) != sizeof(*super))
 883                 return 4;
 884
 885         if (super->state & (1<<MD_SB_BITMAP_PRESENT)) {
 886                 struct bitmap_super_s * bm = (struct bitmap_super_s*)(super+1);
 887                 if (__le32_to_cpu(bm->magic) == BITMAP_MAGIC)
 888                         if (write(fd, bm, ROUND_UP(sizeof(*bm),4096)) !=
 889                             ROUND_UP(sizeof(*bm),4096))
 890                             return 5;
 891         }
 892
 893         fsync(fd);
 894         return 0;
 895 }
 896
 897 static int write_init_super0(struct supertype *st)
 898 {
 899         mdp_super_t *sb = st->sb;
 900         int rv = 0;
 901         struct devinfo *di;
 902
 903         for (di = st->info ; di && ! rv ; di = di->next) {
 904
 905                 if (di->disk.state & (1 << MD_DISK_FAULTY))
 906                         continue;
 907                 if (di->fd == -1)
 908                         continue;
 909                 while (Kill(di->devname, NULL, 0, -1, 1) == 0)
 910                         ;
 911
 912                 sb->disks[di->disk.number].state &= ~(1<<MD_DISK_FAULTY);
 913
 914                 sb->this_disk = sb->disks[di->disk.number];
 915                 sb->sb_csum = calc_sb0_csum(sb);
 916                 rv = store_super0(st, di->fd);
 917
 918                 if (rv == 0 && (sb->state & (1<<MD_SB_BITMAP_PRESENT)))
 919                         rv = st->ss->write_bitmap(st, di->fd, NoUpdate);
 920
 921                 if (rv)
 922                         pr_err("failed to write superblock to %s\n",
 923                                di->devname);
 924         }
 925         return rv;
 926 }
 927
 928 static int compare_super0(struct supertype *st, struct supertype *tst)
 929 {
 930         /*
 931          * return:
 932          *  0 same, or first was empty, and second was copied
 933          *  1 second had wrong number
 934          *  2 wrong uuid
 935          *  3 wrong other info
 936          */
 937         mdp_super_t *first = st->sb;
 938         mdp_super_t *second = tst->sb;
 939         int uuid1[4], uuid2[4];
 940
 941         if (second->md_magic != MD_SB_MAGIC)
 942                 return 1;
 943         if (!first) {
 944                 if (posix_memalign((void**)&first, 4096,
 945                              MD_SB_BYTES +
 946                              ROUND_UP(sizeof(struct bitmap_super_s), 4096)) != 0) {
 947                         pr_err("could not allocate superblock\n");
 948                         return 1;
 949                 }
 950                 memcpy(first, second, MD_SB_BYTES + sizeof(struct bitmap_super_s));
 951                 st->sb = first;
 952                 return 0;
 953         }
 954
 955         uuid_from_super0(st, uuid1);
 956         uuid_from_super0(tst, uuid2);
 957         if (!same_uuid(uuid1, uuid2, 0))
 958                 return 2;
 959         if (first->major_version != second->major_version ||
 960             first->minor_version != second->minor_version ||
 961             first->patch_version != second->patch_version ||
 962             first->gvalid_words  != second->gvalid_words  ||
 963             first->ctime         != second->ctime         ||
 964             first->level         != second->level         ||
 965             first->size          != second->size          ||
 966             first->raid_disks    != second->raid_disks    )
 967                 return 3;
 968
 969         return 0;
 970 }
 971
 972 static void free_super0(struct supertype *st);
 973
 974 static int load_super0(struct supertype *st, int fd, char *devname)
 975 {
 976         /* try to read in the superblock
 977          * Return:
 978          *  0 on success
 979          *  1 on cannot get superblock
 980          *  2 on superblock meaningless
 981          */
 982         unsigned long long dsize;
 983         unsigned long long offset;
 984         mdp_super_t *super;
 985         int uuid[4];
 986         struct bitmap_super_s *bsb;
 987
 988         free_super0(st);
 989
 990         if (!get_dev_size(fd, devname, &dsize))
 991                 return 1;
 992
 993         if (dsize < MD_RESERVED_SECTORS*512) {
 994                 if (devname)
 995                         pr_err("%s is too small for md: size is %llu sectors.\n",
 996                                devname, dsize);
 997                 return 1;
 998         }
 999         st->devsize = dsize;
1000
1001         offset = MD_NEW_SIZE_SECTORS(dsize>>9);
1002
1003         offset *= 512;
1004
1005         if (lseek64(fd, offset, 0)< 0LL) {
1006                 if (devname)
1007                         pr_err("Cannot seek to superblock on %s: %s\n",
1008                                 devname, strerror(errno));
1009                 return 1;
1010         }
1011
1012         if (posix_memalign((void**)&super, 4096,
1013                            MD_SB_BYTES +
1014                            ROUND_UP(sizeof(bitmap_super_t), 4096)) != 0) {
1015                 pr_err("could not allocate superblock\n");
1016                 return 1;
1017         }
1018
1019         if (read(fd, super, sizeof(*super)) != MD_SB_BYTES) {
1020                 if (devname)
1021                         pr_err("Cannot read superblock on %s\n",
1022                                 devname);
1023                 free(super);
1024                 return 1;
1025         }
1026
1027         if (st->ss && st->minor_version == 9)
1028                 super0_swap_endian(super);
1029
1030         if (super->md_magic != MD_SB_MAGIC) {
1031                 if (devname)
1032                         pr_err("No super block found on %s (Expected magic %08x, got %08x)\n",
1033                                 devname, MD_SB_MAGIC, super->md_magic);
1034                 free(super);
1035                 return 2;
1036         }
1037
1038         if (super->major_version != 0) {
1039                 if (devname)
1040                         pr_err("Cannot interpret superblock on %s - version is %d\n",
1041                                 devname, super->major_version);
1042                 free(super);
1043                 return 2;
1044         }
1045         st->sb = super;
1046
1047         if (st->ss == NULL) {
1048                 st->ss = &super0;
1049                 st->minor_version = super->minor_version;
1050                 st->max_devs = MD_SB_DISKS;
1051                 st->info = NULL;
1052         }
1053
1054         /* Now check on the bitmap superblock */
1055         if ((super->state & (1<<MD_SB_BITMAP_PRESENT)) == 0)
1056                 return 0;
1057         /* Read the bitmap superblock and make sure it looks
1058          * valid.  If it doesn't clear the bit.  An --assemble --force
1059          * should get that written out.
1060          */
1061         if (read(fd, super+1, ROUND_UP(sizeof(struct bitmap_super_s),4096)) !=
1062             ROUND_UP(sizeof(struct bitmap_super_s), 4096))
1063                 goto no_bitmap;
1064
1065         uuid_from_super0(st, uuid);
1066         bsb = (struct bitmap_super_s *)(super+1);
1067         if (__le32_to_cpu(bsb->magic) != BITMAP_MAGIC ||
1068             memcmp(bsb->uuid, uuid, 16) != 0)
1069                 goto no_bitmap;
1070         return 0;
1071
1072  no_bitmap:
1073         super->state &= ~(1<<MD_SB_BITMAP_PRESENT);
1074
1075         return 0;
1076 }
1077
1078 static struct supertype *match_metadata_desc0(char *arg)
1079 {
1080         struct supertype *st = xcalloc(1, sizeof(*st));
1081
1082         st->container_devnm[0] = 0;
1083         st->ss = &super0;
1084         st->info = NULL;
1085         st->minor_version = 90;
1086         st->max_devs = MD_SB_DISKS;
1087         st->sb = NULL;
1088         /* we sometimes get 00.90 */
1089         while (arg[0] == '0' && arg[1] == '0')
1090                 arg++;
1091         if (strcmp(arg, "0") == 0 ||
1092 #ifdef DEFAULT_OLD_METADATA /* ifndef in super1.c */
1093             strcmp(arg, "default") == 0 ||
1094 #endif /* DEFAULT_OLD_METADATA */
1095             strcmp(arg, "0.90") == 0 ||
1096             strcmp(arg, "") == 0 /* no metadata  - i.e. non_persistent */
1097                 )
1098                 return st;
1099
1100         st->minor_version = 91; /* reshape in progress */
1101         if (strcmp(arg, "0.91") == 0) /* For dup_super support */
1102                 return st;
1103
1104         st->minor_version = 9; /* flag for 'byte-swapped' */
1105         if (strcmp(arg, "0.swap")==0 ||
1106             strcmp(arg, "0.9") == 0) /* For dup_super support */
1107                 return st;
1108
1109         free(st);
1110         return NULL;
1111 }
1112
1113 static __u64 avail_size0(struct supertype *st, __u64 devsize,
1114                          unsigned long long data_offset)
1115 {
1116         if (data_offset != 0 && data_offset != INVALID_SECTORS)
1117                 return 0ULL;
1118         if (devsize < MD_RESERVED_SECTORS)
1119                 return 0ULL;
1120         return MD_NEW_SIZE_SECTORS(devsize);
1121 }
1122
1123 static int add_internal_bitmap0(struct supertype *st, int *chunkp,
1124                                 int delay, int write_behind,
1125                                 unsigned long long size, int may_change,
1126                                 int major)
1127 {
1128         /*
1129          * The bitmap comes immediately after the superblock and must be 60K in size
1130          * at most.  The default size is between 30K and 60K
1131          *
1132          * size is in sectors,  chunk is in bytes !!!
1133          */
1134         unsigned long long bits;
1135         unsigned long long max_bits = (60*1024 - sizeof(bitmap_super_t))*8;
1136         unsigned long long min_chunk;
1137         int chunk = *chunkp;
1138         mdp_super_t *sb = st->sb;
1139         bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + MD_SB_BYTES);
1140         int uuid[4];
1141
1142         min_chunk = 4096; /* sub-page chunks don't work yet.. */
1143         bits = (size * 512) / min_chunk + 1;
1144         while (bits > max_bits) {
1145                 min_chunk *= 2;
1146                 bits = (bits+1)/2;
1147         }
1148         if (chunk == UnSet) {
1149                 /* A chunk size less than a few Megabytes gives poor
1150                  * performance without increasing resync noticeably
1151                  */
1152                 chunk = min_chunk;
1153                 if (chunk < 64*1024*1024)
1154                         chunk = 64*1024*1024;
1155         } else if ((unsigned long long)chunk < min_chunk)
1156                 return -EINVAL; /* chunk size too small */
1157
1158         sb->state |= (1<<MD_SB_BITMAP_PRESENT);
1159
1160         memset(bms, 0, sizeof(*bms));
1161         bms->magic = __cpu_to_le32(BITMAP_MAGIC);
1162         bms->version = __cpu_to_le32(major);
1163         uuid_from_super0(st, uuid);
1164         memcpy(bms->uuid, uuid, 16);
1165         bms->chunksize = __cpu_to_le32(chunk);
1166         bms->daemon_sleep = __cpu_to_le32(delay);
1167         bms->sync_size = __cpu_to_le64(size);
1168         bms->write_behind = __cpu_to_le32(write_behind);
1169         *chunkp = chunk;
1170         return 0;
1171 }
1172
1173 static int locate_bitmap0(struct supertype *st, int fd, int node_num)
1174 {
1175         unsigned long long dsize;
1176         unsigned long long offset;
1177
1178         if (!get_dev_size(fd, NULL, &dsize))
1179                 return -1;
1180
1181         if (dsize < MD_RESERVED_SECTORS*512)
1182                 return -1;
1183
1184         offset = MD_NEW_SIZE_SECTORS(dsize>>9);
1185
1186         offset *= 512;
1187
1188         offset += MD_SB_BYTES;
1189
1190         lseek64(fd, offset, 0);
1191         return 0;
1192 }
1193
1194 static int write_bitmap0(struct supertype *st, int fd, enum bitmap_update update)
1195 {
1196         unsigned long long dsize;
1197         unsigned long long offset;
1198         mdp_super_t *sb = st->sb;
1199
1200         int rv = 0;
1201
1202         int towrite, n;
1203         void *buf;
1204
1205         if (!get_dev_size(fd, NULL, &dsize))
1206                 return 1;
1207
1208         if (dsize < MD_RESERVED_SECTORS*512)
1209                 return -1;
1210
1211         offset = MD_NEW_SIZE_SECTORS(dsize>>9);
1212
1213         offset *= 512;
1214
1215         if (lseek64(fd, offset + 4096, 0)< 0LL)
1216                 return 3;
1217
1218         if (posix_memalign(&buf, 4096, 4096))
1219                 return -ENOMEM;
1220
1221         memset(buf, 0xff, 4096);
1222         memcpy(buf,  ((char*)sb)+MD_SB_BYTES, sizeof(bitmap_super_t));
1223         towrite = 60*1024;
1224         while (towrite > 0) {
1225                 n = towrite;
1226                 if (n > 4096)
1227                         n = 4096;
1228                 n = write(fd, buf, n);
1229                 if (n > 0)
1230                         towrite -= n;
1231                 else
1232                         break;
1233                 memset(buf, 0xff, 4096);
1234         }
1235         fsync(fd);
1236         if (towrite)
1237                 rv = -2;
1238
1239         free(buf);
1240         return rv;
1241 }
1242
1243 static void free_super0(struct supertype *st)
1244 {
1245         if (st->sb)
1246                 free(st->sb);
1247         while (st->info) {
1248                 struct devinfo *di = st->info;
1249                 st->info = di->next;
1250                 if (di->fd >= 0)
1251                         close(di->fd);
1252                 free(di);
1253         }
1254         st->sb = NULL;
1255 }
1256
1257 static int validate_geometry0(struct supertype *st, int level,
1258                               int layout, int raiddisks,
1259                               int *chunk, unsigned long long size,
1260                               unsigned long long data_offset,
1261                               char *subdev, unsigned long long *freesize,
1262                               int consistency_policy, int verbose)
1263 {
1264         unsigned long long ldsize;
1265         int fd;
1266         unsigned int tbmax = 4;
1267
1268         /* prior to linux 3.1, a but limits usable device size to 2TB.
1269          * It was introduced in 2.6.29, but we won't worry about that detail
1270          */
1271         if (get_linux_version() < 3001000)
1272                 tbmax = 2;
1273
1274         if (level == LEVEL_CONTAINER) {
1275                 if (verbose)
1276                         pr_err("0.90 metadata does not support containers\n");
1277                 return 0;
1278         }
1279         if (raiddisks > MD_SB_DISKS) {
1280                 if (verbose)
1281                         pr_err("0.90 metadata supports at most %d devices per array\n",
1282                                 MD_SB_DISKS);
1283                 return 0;
1284         }
1285         if (size >= tbmax * 2ULL*1024*1024*1024) {
1286                 if (verbose)
1287                         pr_err("0.90 metadata supports at most %d terabytes per device\n", tbmax);
1288                 return 0;
1289         }
1290         if (*chunk == UnSet)
1291                 *chunk = DEFAULT_CHUNK;
1292
1293         if (!subdev)
1294                 return 1;
1295
1296         fd = open(subdev, O_RDONLY|O_EXCL, 0);
1297         if (fd < 0) {
1298                 if (verbose)
1299                         pr_err("super0.90 cannot open %s: %s\n",
1300                                 subdev, strerror(errno));
1301                 return 0;
1302         }
1303
1304         if (!get_dev_size(fd, subdev, &ldsize)) {
1305                 close(fd);
1306                 return 0;
1307         }
1308         close(fd);
1309
1310         if (ldsize < MD_RESERVED_SECTORS * 512)
1311                 return 0;
1312         *freesize = MD_NEW_SIZE_SECTORS(ldsize >> 9);
1313         return 1;
1314 }
1315
1316 struct superswitch super0 = {
1317         .examine_super = examine_super0,
1318         .brief_examine_super = brief_examine_super0,
1319         .export_examine_super = export_examine_super0,
1320         .detail_super = detail_super0,
1321         .brief_detail_super = brief_detail_super0,
1322         .write_init_super = write_init_super0,
1323         .validate_geometry = validate_geometry0,
1324         .add_to_super = add_to_super0,
1325         .copy_metadata = copy_metadata0,
1326         .match_home = match_home0,
1327         .uuid_from_super = uuid_from_super0,
1328         .getinfo_super = getinfo_super0,
1329         .container_content = container_content0,
1330         .update_super = update_super0,
1331         .init_super = init_super0,
1332         .store_super = store_super0,
1333         .compare_super = compare_super0,
1334         .load_super = load_super0,
1335         .match_metadata_desc = match_metadata_desc0,
1336         .avail_size = avail_size0,
1337         .add_internal_bitmap = add_internal_bitmap0,
1338         .locate_bitmap = locate_bitmap0,
1339         .write_bitmap = write_bitmap0,
1340         .free_super = free_super0,
1341         .name = "0.90",
1342 };