super0.c

   1 /*
   2  * mdadm - manage Linux "md" devices aka RAID arrays.
   3  *
   4  * Copyright (C) 2001-2004 Neil Brown <neilb@cse.unsw.edu.au>
   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 #include "mdadm.h"
  31 #include <asm/byteorder.h>
  32
  33 /*
  34  * All handling for the 0.90.0 version superblock is in
  35  * this file.
  36  * This includes:
  37  *   - finding, loading, and writing the superblock.
  38  *   - initialising a new superblock
  39  *   - printing the superblock for --examine
  40  *   - printing part of the superblock for --detail
  41  * .. other stuff
  42  */
  43
  44
  45 static unsigned long calc_sb0_csum(mdp_super_t *super)
  46 {
  47         unsigned long csum = super->sb_csum;
  48         unsigned long newcsum;
  49         super->sb_csum= 0 ;
  50         newcsum = calc_csum(super, MD_SB_BYTES);
  51         super->sb_csum = csum;
  52         return newcsum;
  53 }
  54
  55
  56 void super0_swap_endian(struct mdp_superblock_s *sb)
  57 {
  58         /* as super0 superblocks are host-endian, it is sometimes
  59          * useful to be able to swap the endianness
  60          * as (almost) everything is u32's we byte-swap every 4byte
  61          * number.
  62          * We then also have to swap the events_hi and events_lo
  63          */
  64         char *sbc = (char *)sb;
  65         __u32 t32;
  66         int i;
  67
  68         for (i=0; i < MD_SB_BYTES ; i+=4) {
  69                 char t = sbc[i];
  70                 sbc[i] = sbc[i+3];
  71                 sbc[i+3] = t;
  72                 t=sbc[i+1];
  73                 sbc[i+1]=sbc[i+2];
  74                 sbc[i+2]=t;
  75         }
  76         t32 = sb->events_hi;
  77         sb->events_hi = sb->events_lo;
  78         sb->events_lo = t32;
  79
  80         t32 = sb->cp_events_hi;
  81         sb->cp_events_hi = sb->cp_events_lo;
  82         sb->cp_events_lo = t32;
  83
  84 }
  85
  86 #ifndef MDASSEMBLE
  87
  88 static void examine_super0(void *sbv)
  89 {
  90         mdp_super_t *sb = sbv;
  91         time_t atime;
  92         int d;
  93         char *c;
  94
  95         printf("          Magic : %08x\n", sb->md_magic);
  96         printf("        Version : %02d.%02d.%02d\n", sb->major_version, sb->minor_version,
  97                sb->patch_version);
  98         if (sb->minor_version >= 90)
  99                 printf("           UUID : %08x:%08x:%08x:%08x\n", sb->set_uuid0, sb->set_uuid1,
 100                        sb->set_uuid2, sb->set_uuid3);
 101         else
 102                 printf("           UUID : %08x\n", sb->set_uuid0);
 103
 104         atime = sb->ctime;
 105         printf("  Creation Time : %.24s\n", ctime(&atime));
 106         c=map_num(pers, sb->level);
 107         printf("     Raid Level : %s\n", c?c:"-unknown-");
 108         if ((int)sb->level >= 0)
 109                 printf("    Device Size : %d%s\n", sb->size, human_size((long long)sb->size<<10));
 110         printf("   Raid Devices : %d\n", sb->raid_disks);
 111         printf("  Total Devices : %d\n", sb->nr_disks);
 112         printf("Preferred Minor : %d\n", sb->md_minor);
 113         printf("\n");
 114         atime = sb->utime;
 115         printf("    Update Time : %.24s\n", ctime(&atime));
 116         printf("          State : %s\n",
 117                (sb->state&(1<<MD_SB_CLEAN))?"clean":"active");
 118         if (sb->state & (1<<MD_SB_BITMAP_PRESENT))
 119                 printf("Internal Bitmap : present\n");
 120         printf(" Active Devices : %d\n", sb->active_disks);
 121         printf("Working Devices : %d\n", sb->working_disks);
 122         printf(" Failed Devices : %d\n", sb->failed_disks);
 123         printf("  Spare Devices : %d\n", sb->spare_disks);
 124         if (calc_sb0_csum(sb) == sb->sb_csum)
 125                 printf("       Checksum : %x - correct\n", sb->sb_csum);
 126         else
 127                 printf("       Checksum : %x - expected %lx\n", sb->sb_csum, calc_sb0_csum(sb));
 128         printf("         Events : %d.%d\n", sb->events_hi, sb->events_lo);
 129         printf("\n");
 130         if (sb->level == 5) {
 131                 c = map_num(r5layout, sb->layout);
 132                 printf("         Layout : %s\n", c?c:"-unknown-");
 133         }
 134         switch(sb->level) {
 135         case 0:
 136         case 4:
 137         case 5:
 138                 printf("     Chunk Size : %dK\n", sb->chunk_size/1024);
 139                 break;
 140         case -1:
 141                 printf("       Rounding : %dK\n", sb->chunk_size/1024);
 142                 break;
 143         default: break;
 144         }
 145         printf("\n");
 146         printf("      Number   Major   Minor   RaidDevice State\n");
 147         for (d= -1; d<(signed int)(sb->raid_disks+sb->spare_disks); d++) {
 148                 mdp_disk_t *dp;
 149                 char *dv;
 150                 char nb[5];
 151                 int wonly;
 152                 if (d>=0) dp = &sb->disks[d];
 153                 else dp = &sb->this_disk;
 154                 snprintf(nb, sizeof(nb), "%4d", d);
 155                 printf("%4s %5d   %5d    %5d    %5d     ", d < 0 ? "this" :  nb,
 156                        dp->number, dp->major, dp->minor, dp->raid_disk);
 157                 wonly = dp->state & (1<<MD_DISK_WRITEMOSTLY);
 158                 dp->state &= ~(1<<MD_DISK_WRITEMOSTLY);
 159                 if (dp->state & (1<<MD_DISK_FAULTY)) printf(" faulty");
 160                 if (dp->state & (1<<MD_DISK_ACTIVE)) printf(" active");
 161                 if (dp->state & (1<<MD_DISK_SYNC)) printf(" sync");
 162                 if (dp->state & (1<<MD_DISK_REMOVED)) printf(" removed");
 163                 if (wonly) printf(" write-mostly");
 164                 if (dp->state == 0) printf(" spare");
 165                 if ((dv=map_dev(dp->major, dp->minor)))
 166                         printf("   %s", dv);
 167                 printf("\n");
 168                 if (d == -1) printf("\n");
 169         }
 170 }
 171
 172 static void brief_examine_super0(void *sbv)
 173 {
 174         mdp_super_t *sb = sbv;
 175         char *c=map_num(pers, sb->level);
 176
 177         printf("ARRAY %s level=%s num-devices=%d UUID=",
 178                get_md_name(sb->md_minor),
 179                c?c:"-unknown-", sb->raid_disks);
 180         if (sb->minor_version >= 90)
 181                 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
 182                        sb->set_uuid2, sb->set_uuid3);
 183         else
 184                 printf("%08x", sb->set_uuid0);
 185         printf("\n");
 186 }
 187
 188 static void detail_super0(void *sbv)
 189 {
 190         mdp_super_t *sb = sbv;
 191         printf("           UUID : ");
 192         if (sb->minor_version >= 90)
 193                 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
 194                        sb->set_uuid2, sb->set_uuid3);
 195         else
 196                 printf("%08x", sb->set_uuid0);
 197         printf("\n         Events : %d.%d\n\n", sb->events_hi, sb->events_lo);
 198 }
 199
 200 static void brief_detail_super0(void *sbv)
 201 {
 202         mdp_super_t *sb = sbv;
 203         printf(" UUID=");
 204         if (sb->minor_version >= 90)
 205                 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
 206                        sb->set_uuid2, sb->set_uuid3);
 207         else
 208                 printf("%08x", sb->set_uuid0);
 209 }
 210 #endif
 211 static void uuid_from_super0(int uuid[4], void * sbv)
 212 {
 213         mdp_super_t *super = sbv;
 214         uuid[0] = super->set_uuid0;
 215         if (super->minor_version >= 90) {
 216                 uuid[1] = super->set_uuid1;
 217                 uuid[2] = super->set_uuid2;
 218                 uuid[3] = super->set_uuid3;
 219         } else {
 220                 uuid[1] = 0;
 221                 uuid[2] = 0;
 222                 uuid[3] = 0;
 223         }
 224 }
 225
 226 static void getinfo_super0(struct mdinfo *info, mddev_ident_t ident, void *sbv)
 227 {
 228         mdp_super_t *sb = sbv;
 229         int working = 0;
 230         int i;
 231
 232         info->array.major_version = sb->major_version;
 233         info->array.minor_version = sb->minor_version;
 234         info->array.patch_version = sb->patch_version;
 235         info->array.raid_disks = sb->raid_disks;
 236         info->array.level = sb->level;
 237         info->array.md_minor = sb->md_minor;
 238         info->array.ctime = sb->ctime;
 239
 240         info->disk.state = sb->this_disk.state;
 241         info->disk.major = sb->this_disk.major;
 242         info->disk.minor = sb->this_disk.minor;
 243         info->disk.raid_disk = sb->this_disk.raid_disk;
 244         info->disk.number = sb->this_disk.number;
 245
 246         info->events = md_event(sb);
 247
 248         uuid_from_super0(info->uuid, sbv);
 249
 250         ident->name[0] = 0;
 251         /* work_disks is calculated rather than read directly */
 252         for (i=0; i < MD_SB_DISKS; i++)
 253                 if ((sb->disks[i].state & (1<<MD_DISK_SYNC)) &&
 254                     (sb->disks[i].state & (1<<MD_DISK_ACTIVE)) &&
 255                     !(sb->disks[i].state & (1<<MD_DISK_FAULTY)))
 256                         working ++;
 257         info->array.working_disks = working;
 258 }
 259
 260
 261 static int update_super0(struct mdinfo *info, void *sbv, char *update, char *devname, int verbose)
 262 {
 263         int rv = 0;
 264         mdp_super_t *sb = sbv;
 265         if (strcmp(update, "sparc2.2")==0 ) {
 266                 /* 2.2 sparc put the events in the wrong place
 267                  * So we copy the tail of the superblock
 268                  * up 4 bytes before continuing
 269                  */
 270                 __u32 *sb32 = (__u32*)sb;
 271                 memcpy(sb32+MD_SB_GENERIC_CONSTANT_WORDS+7,
 272                        sb32+MD_SB_GENERIC_CONSTANT_WORDS+7+1,
 273                        (MD_SB_WORDS - (MD_SB_GENERIC_CONSTANT_WORDS+7+1))*4);
 274                 fprintf (stderr, Name ": adjusting superblock of %s for 2.2/sparc compatability.\n",
 275                          devname);
 276         }
 277         if (strcmp(update, "super-minor") ==0) {
 278                 sb->md_minor = info->array.md_minor;
 279                 if (verbose)
 280                         fprintf(stderr, Name ": updating superblock of %s with minor number %d\n",
 281                                 devname, info->array.md_minor);
 282         }
 283         if (strcmp(update, "summaries") == 0) {
 284                 int i;
 285                 /* set nr_disks, active_disks, working_disks,
 286                  * failed_disks, spare_disks based on disks[]
 287                  * array in superblock.
 288                  * Also make sure extra slots aren't 'failed'
 289                  */
 290                 sb->nr_disks = sb->active_disks =
 291                         sb->working_disks = sb->failed_disks =
 292                         sb->spare_disks = 0;
 293                 for (i=0; i < MD_SB_DISKS ; i++)
 294                         if (sb->disks[i].major ||
 295                             sb->disks[i].minor) {
 296                                 int state = sb->disks[i].state;
 297                                 if (state & (1<<MD_DISK_REMOVED))
 298                                         continue;
 299                                 sb->nr_disks++;
 300                                 if (state & (1<<MD_DISK_ACTIVE))
 301                                         sb->active_disks++;
 302                                 if (state & (1<<MD_DISK_FAULTY))
 303                                         sb->failed_disks++;
 304                                 else
 305                                         sb->working_disks++;
 306                                 if (state == 0)
 307                                         sb->spare_disks++;
 308                         } else if (i >= sb->raid_disks && sb->disks[i].number == 0)
 309                                 sb->disks[i].state = 0;
 310         }
 311         if (strcmp(update, "force")==0) {
 312                 sb->events_hi = (info->events>>32) & 0xFFFFFFFF;
 313                 sb->events_lo = (info->events) & 0xFFFFFFFF;
 314                 if (sb->level == 5 || sb->level == 4 || sb->level == 6)
 315                         /* need to force clean */
 316                         sb->state |= (1 << MD_SB_CLEAN);
 317         }
 318         if (strcmp(update, "assemble")==0) {
 319                 int d = info->disk.number;
 320                 int wonly = sb->disks[d].state & (1<<MD_DISK_WRITEMOSTLY);
 321                 sb->disks[d].state &= ~(1<<MD_DISK_WRITEMOSTLY);
 322                 if (sb->disks[d].state != info->disk.state) {
 323                         sb->disks[d].state = info->disk.state & wonly;
 324                         rv = 1;
 325                 }
 326         }
 327         if (strcmp(update, "newdev") == 0) {
 328                 int d = info->disk.number;
 329                 memset(&sb->disks[d], 0, sizeof(sb->disks[d]));
 330                 sb->disks[d].number = d;
 331                 sb->disks[d].major = info->disk.major;
 332                 sb->disks[d].minor = info->disk.minor;
 333                 sb->disks[d].raid_disk = info->disk.raid_disk;
 334                 sb->disks[d].state = info->disk.state;
 335                 sb->this_disk = sb->disks[d];
 336         }
 337         if (strcmp(update, "grow") == 0) {
 338                 sb->raid_disks = info->array.raid_disks;
 339                 sb->nr_disks = info->array.nr_disks;
 340                 sb->active_disks = info->array.active_disks;
 341                 sb->working_disks = info->array.working_disks;
 342                 memset(&sb->disks[info->disk.number], 0, sizeof(sb->disks[0]));
 343                 sb->disks[info->disk.number].number = info->disk.number;
 344                 sb->disks[info->disk.number].major = info->disk.major;
 345                 sb->disks[info->disk.number].minor = info->disk.minor;
 346                 sb->disks[info->disk.number].raid_disk = info->disk.raid_disk;
 347                 sb->disks[info->disk.number].state = info->disk.state;
 348                 if (sb->this_disk.number == info->disk.number)
 349                         sb->this_disk = sb->disks[info->disk.number];
 350         }
 351         if (strcmp(update, "resync") == 0) {
 352                 /* make sure resync happens */
 353                 sb->state &= ~(1<<MD_SB_CLEAN);
 354                 sb->recovery_cp = 0;
 355         }
 356
 357         sb->sb_csum = calc_sb0_csum(sb);
 358         return rv;
 359 }
 360
 361 static __u64 event_super0(void *sbv)
 362 {
 363         mdp_super_t *sb = sbv;
 364         return md_event(sb);
 365 }
 366
 367
 368
 369 static int init_super0(struct supertype *st, void **sbp, mdu_array_info_t *info, char *ignored_name)
 370 {
 371         mdp_super_t *sb = malloc(MD_SB_BYTES + sizeof(bitmap_super_t));
 372         int spares;
 373         int rfd;
 374         memset(sb, 0, MD_SB_BYTES + sizeof(bitmap_super_t));
 375
 376         if (info->major_version == -1) {
 377                 /* zeroing the superblock */
 378                 return 0;
 379         }
 380
 381         spares = info->working_disks - info->active_disks;
 382         if (info->raid_disks + spares  > MD_SB_DISKS) {
 383                 fprintf(stderr, Name ": too many devices requested: %d+%d > %d\n",
 384                         info->raid_disks , spares, MD_SB_DISKS);
 385                 return 0;
 386         }
 387
 388         rfd = open("/dev/urandom", O_RDONLY);
 389         sb->md_magic = MD_SB_MAGIC;
 390         sb->major_version = 0;
 391         sb->minor_version = 90;
 392         sb->patch_version = 0;
 393         sb->gvalid_words = 0; /* ignored */
 394         if (rfd < 0 || read(rfd, &sb->set_uuid0, 4) != 4)
 395                 sb->set_uuid0 = random();
 396         sb->ctime = time(0);
 397         sb->level = info->level;
 398         sb->size = info->size;
 399         sb->nr_disks = info->nr_disks;
 400         sb->raid_disks = info->raid_disks;
 401         sb->md_minor = info->md_minor;
 402         sb->not_persistent = 0;
 403         if (rfd < 0 || read(rfd, &sb->set_uuid1, 12) != 12) {
 404                 sb->set_uuid1 = random();
 405                 sb->set_uuid2 = random();
 406                 sb->set_uuid3 = random();
 407         }
 408         if (rfd >= 0)
 409                 close(rfd);
 410
 411         sb->utime = sb->ctime;
 412         sb->state = info->state;
 413         sb->active_disks = info->active_disks;
 414         sb->working_disks = info->working_disks;
 415         sb->failed_disks = info->failed_disks;
 416         sb->spare_disks = info->spare_disks;
 417         sb->events_hi = 0;
 418         sb->events_lo = 1;
 419
 420         sb->layout = info->layout;
 421         sb->chunk_size = info->chunk_size;
 422
 423         *sbp = sb;
 424         return 1;
 425 }
 426
 427 /* Add a device to the superblock being created */
 428 static void add_to_super0(void *sbv, mdu_disk_info_t *dinfo)
 429 {
 430         mdp_super_t *sb = sbv;
 431         mdp_disk_t *dk = &sb->disks[dinfo->number];
 432
 433         dk->number = dinfo->number;
 434         dk->major = dinfo->major;
 435         dk->minor = dinfo->minor;
 436         dk->raid_disk = dinfo->raid_disk;
 437         dk->state = dinfo->state;
 438 }
 439
 440 static int store_super0(struct supertype *st, int fd, void *sbv)
 441 {
 442         unsigned long size;
 443         unsigned long long dsize;
 444         unsigned long long offset;
 445         mdp_super_t *super = sbv;
 446
 447 #ifdef BLKGETSIZE64
 448         if (ioctl(fd, BLKGETSIZE64, &dsize) != 0)
 449 #endif
 450         {
 451                 if (ioctl(fd, BLKGETSIZE, &size))
 452                         return 1;
 453                 else
 454                         dsize = ((unsigned long long)size)<<9;
 455         }
 456
 457         if (dsize < MD_RESERVED_SECTORS*2)
 458                 return 2;
 459
 460         offset = MD_NEW_SIZE_SECTORS(dsize>>9);
 461
 462         offset *= 512;
 463
 464         if (lseek64(fd, offset, 0)< 0LL)
 465                 return 3;
 466
 467         if (write(fd, super, sizeof(*super)) != sizeof(*super))
 468                 return 4;
 469
 470         fsync(fd);
 471         return 0;
 472 }
 473
 474 static int write_init_super0(struct supertype *st, void *sbv, mdu_disk_info_t *dinfo, char *devname)
 475 {
 476         mdp_super_t *sb = sbv;
 477         int fd = open(devname, O_RDWR|O_EXCL);
 478         int rv;
 479
 480         if (fd < 0) {
 481                 fprintf(stderr, Name ": Failed to open %s to write superblock\n", devname);
 482                 return -1;
 483         }
 484
 485         sb->disks[dinfo->number].state &= ~(1<<MD_DISK_FAULTY);
 486         sb->disks[dinfo->number].state |= (1<<MD_DISK_SYNC);
 487
 488         sb->this_disk = sb->disks[dinfo->number];
 489         sb->sb_csum = calc_sb0_csum(sb);
 490         rv = store_super0(st, fd, sb);
 491
 492         if (sb->state & (1<<MD_SB_BITMAP_PRESENT)) {
 493                 int towrite, n;
 494                 char buf[4096];
 495
 496                 write(fd, ((char*)sb)+MD_SB_BYTES, sizeof(bitmap_super_t));
 497                 towrite = 64*1024 - MD_SB_BYTES - sizeof(bitmap_super_t);
 498                 memset(buf, 0xff, sizeof(buf));
 499                 while (towrite > 0) {
 500                         n = towrite;
 501                         if (n > sizeof(buf))
 502                                 n = sizeof(buf);
 503                         n = write(fd, buf, n);
 504                         if (n > 0)
 505                                 towrite -= n;
 506                         else
 507                                 break;
 508                 }
 509                 fsync(fd);
 510                 if (towrite)
 511                         rv = -2;
 512         }
 513
 514         close(fd);
 515         if (rv)
 516                 fprintf(stderr, Name ": failed to write superblock to %s\n", devname);
 517         return rv;
 518 }
 519
 520 static int compare_super0(void **firstp, void *secondv)
 521 {
 522         /*
 523          * return:
 524          *  0 same, or first was empty, and second was copied
 525          *  1 second had wrong number
 526          *  2 wrong uuid
 527          *  3 wrong other info
 528          */
 529         mdp_super_t *first = *firstp;
 530         mdp_super_t *second = secondv;
 531
 532         int uuid1[4], uuid2[4];
 533         if (second->md_magic != MD_SB_MAGIC)
 534                 return 1;
 535         if (!first) {
 536                 first = malloc(MD_SB_BYTES);
 537                 memcpy(first, second, MD_SB_BYTES);
 538                 *firstp = first;
 539                 return 0;
 540         }
 541
 542         uuid_from_super0(uuid1, first);
 543         uuid_from_super0(uuid2, second);
 544         if (!same_uuid(uuid1, uuid2, 0))
 545                 return 2;
 546         if (first->major_version != second->major_version ||
 547             first->minor_version != second->minor_version ||
 548             first->patch_version != second->patch_version ||
 549             first->gvalid_words  != second->gvalid_words  ||
 550             first->ctime         != second->ctime         ||
 551             first->level         != second->level         ||
 552             first->size          != second->size          ||
 553             first->raid_disks    != second->raid_disks    )
 554                 return 3;
 555
 556         return 0;
 557 }
 558
 559
 560 static int load_super0(struct supertype *st, int fd, void **sbp, char *devname)
 561 {
 562         /* try to read in the superblock
 563          * Return:
 564          *  0 on success
 565          *  1 on cannot get superblock
 566          *  2 on superblock meaningless
 567          */
 568         unsigned long size;
 569         unsigned long long dsize;
 570         unsigned long long offset;
 571         mdp_super_t *super;
 572
 573 #ifdef BLKGETSIZE64
 574         if (ioctl(fd, BLKGETSIZE64, &dsize) != 0)
 575 #endif
 576         {
 577                 if (ioctl(fd, BLKGETSIZE, &size)) {
 578                         if (devname)
 579                                 fprintf(stderr, Name ": cannot find device size for %s: %s\n",
 580                                         devname, strerror(errno));
 581                         return 1;
 582                 } else
 583                         dsize = size << 9;
 584         }
 585
 586         if (dsize < MD_RESERVED_SECTORS*2) {
 587                 if (devname)
 588                         fprintf(stderr, Name ": %s is too small for md: size is %ld sectors.\n",
 589                                 devname, size);
 590                 return 1;
 591         }
 592
 593         offset = MD_NEW_SIZE_SECTORS(dsize>>9);
 594
 595         offset *= 512;
 596
 597         ioctl(fd, BLKFLSBUF, 0); /* make sure we read current data */
 598
 599         if (lseek64(fd, offset, 0)< 0LL) {
 600                 if (devname)
 601                         fprintf(stderr, Name ": Cannot seek to superblock on %s: %s\n",
 602                                 devname, strerror(errno));
 603                 return 1;
 604         }
 605
 606         super = malloc(MD_SB_BYTES + sizeof(bitmap_super_t));
 607
 608         if (read(fd, super, sizeof(*super)) != MD_SB_BYTES) {
 609                 if (devname)
 610                         fprintf(stderr, Name ": Cannot read superblock on %s\n",
 611                                 devname);
 612                 free(super);
 613                 return 1;
 614         }
 615
 616         if (st->ss && st->minor_version == 9)
 617                 super0_swap_endian(super);
 618
 619         if (super->md_magic != MD_SB_MAGIC) {
 620                 if (devname)
 621                         fprintf(stderr, Name ": No super block found on %s (Expected magic %08x, got %08x)\n",
 622                                 devname, MD_SB_MAGIC, super->md_magic);
 623                 free(super);
 624                 return 2;
 625         }
 626
 627         if (super->major_version != 0) {
 628                 if (devname)
 629                         fprintf(stderr, Name ": Cannot interpret superblock on %s - version is %d\n",
 630                                 devname, super->major_version);
 631                 free(super);
 632                 return 2;
 633         }
 634         *sbp = super;
 635         if (st->ss == NULL) {
 636                 st->ss = &super0;
 637                 st->minor_version = 90;
 638                 st->max_devs = MD_SB_DISKS;
 639         }
 640
 641         return 0;
 642 }
 643
 644 static struct supertype *match_metadata_desc0(char *arg)
 645 {
 646         struct supertype *st = malloc(sizeof(*st));
 647         if (!st) return st;
 648
 649         st->ss = &super0;
 650         st->minor_version = 90;
 651         st->max_devs = MD_SB_DISKS;
 652         if (strcmp(arg, "0") == 0 ||
 653             strcmp(arg, "0.90") == 0 ||
 654             strcmp(arg, "default") == 0
 655                 )
 656                 return st;
 657
 658         st->minor_version = 9; /* flag for 'byte-swapped' */
 659         if (strcmp(arg, "0.swap")==0)
 660                 return st;
 661
 662         free(st);
 663         return NULL;
 664 }
 665
 666 static __u64 avail_size0(__u64 devsize)
 667 {
 668         if (devsize < MD_RESERVED_SECTORS*2)
 669                 return 0ULL;
 670         return MD_NEW_SIZE_SECTORS(devsize);
 671 }
 672
 673 static int add_internal_bitmap0(void *sbv, int chunk, int delay, int write_behind, unsigned long long size)
 674 {
 675         /*
 676          * The bitmap comes immediately after the superblock and must be 60K in size
 677          * at most.  The default size is between 30K and 60K
 678          *
 679          * size is in K,  chunk is in bytes !!!
 680          */
 681
 682         unsigned long long bits = size;
 683         unsigned long long max_bits = 60*1024*8;
 684         unsigned long long min_chunk;
 685         mdp_super_t *sb = sbv;
 686         bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + MD_SB_BYTES);
 687
 688
 689         min_chunk = 4096; /* sub-page chunks don't work yet.. */
 690         while (bits > max_bits) {
 691                 min_chunk *= 2;
 692                 bits = (bits+1)/2;
 693         }
 694         if (chunk == UnSet)
 695                 chunk = min_chunk;
 696         else if (chunk < min_chunk)
 697                 return 0; /* chunk size too small */
 698
 699         sb->state |= (1<<MD_SB_BITMAP_PRESENT);
 700
 701         memset(bms, sizeof(*bms), 0);
 702         bms->magic = __cpu_to_le32(BITMAP_MAGIC);
 703         bms->version = __cpu_to_le32(BITMAP_MAJOR);
 704         uuid_from_super0((int*)bms->uuid, sb);
 705         bms->chunksize = __cpu_to_le32(chunk);
 706         bms->daemon_sleep = __cpu_to_le32(delay);
 707         bms->sync_size = __cpu_to_le64(size);
 708         bms->write_behind = __cpu_to_le32(write_behind);
 709
 710
 711
 712         return 1;
 713 }
 714
 715
 716 void locate_bitmap0(struct supertype *st, int fd)
 717 {
 718         unsigned long long dsize;
 719         unsigned long size;
 720         unsigned long long offset;
 721 #ifdef BLKGETSIZE64
 722         if (ioctl(fd, BLKGETSIZE64, &dsize) != 0)
 723 #endif
 724         {
 725                 if (ioctl(fd, BLKGETSIZE, &size))
 726                         return;
 727                 else
 728                         dsize = ((unsigned long long)size)<<9;
 729         }
 730
 731         if (dsize < MD_RESERVED_SECTORS*2)
 732                 return;
 733
 734         offset = MD_NEW_SIZE_SECTORS(dsize>>9);
 735
 736         offset *= 512;
 737
 738         offset += MD_SB_BYTES;
 739
 740         lseek64(fd, offset, 0);
 741 }
 742
 743 int write_bitmap0(struct supertype *st, int fd, void *sbv)
 744 {
 745         unsigned long size;
 746         unsigned long long dsize;
 747         unsigned long long offset;
 748         mdp_super_t *sb = sbv;
 749
 750         int rv = 0;
 751
 752         int towrite, n;
 753         char buf[4096];
 754
 755 #ifdef BLKGETSIZE64
 756         if (ioctl(fd, BLKGETSIZE64, &dsize) != 0)
 757 #endif
 758         {
 759                 if (ioctl(fd, BLKGETSIZE, &size))
 760                         return 1;
 761                 else
 762                         dsize = ((unsigned long long)size)<<9;
 763         }
 764
 765         if (dsize < MD_RESERVED_SECTORS*2)
 766                 return -1;
 767
 768         offset = MD_NEW_SIZE_SECTORS(dsize>>9);
 769
 770         offset *= 512;
 771
 772         if (lseek64(fd, offset + 4096, 0)< 0LL)
 773                 return 3;
 774
 775
 776         write(fd, ((char*)sb)+MD_SB_BYTES, sizeof(bitmap_super_t));
 777         towrite = 64*1024 - MD_SB_BYTES - sizeof(bitmap_super_t);
 778         memset(buf, 0xff, sizeof(buf));
 779         while (towrite > 0) {
 780                 n = towrite;
 781                 if (n > sizeof(buf))
 782                         n = sizeof(buf);
 783                 n = write(fd, buf, n);
 784                 if (n > 0)
 785                         towrite -= n;
 786                 else
 787                         break;
 788         }
 789         fsync(fd);
 790         if (towrite)
 791                 rv = -2;
 792
 793         return rv;
 794 }
 795
 796 struct superswitch super0 = {
 797 #ifndef MDASSEMBLE
 798         .examine_super = examine_super0,
 799         .brief_examine_super = brief_examine_super0,
 800         .detail_super = detail_super0,
 801         .brief_detail_super = brief_detail_super0,
 802 #endif
 803         .uuid_from_super = uuid_from_super0,
 804         .getinfo_super = getinfo_super0,
 805         .update_super = update_super0,
 806         .event_super = event_super0,
 807         .init_super = init_super0,
 808         .add_to_super = add_to_super0,
 809         .store_super = store_super0,
 810         .write_init_super = write_init_super0,
 811         .compare_super = compare_super0,
 812         .load_super = load_super0,
 813         .match_metadata_desc = match_metadata_desc0,
 814         .avail_size = avail_size0,
 815         .add_internal_bitmap = add_internal_bitmap0,
 816         .locate_bitmap = locate_bitmap0,
 817         .write_bitmap = write_bitmap0,
 818         .major = 0,
 819         .swapuuid = 0,
 820 };