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
  40 static unsigned long calc_sb0_csum(mdp_super_t *super)
  41 {
  42         unsigned long csum = super->sb_csum;
  43         unsigned long newcsum;
  44         super->sb_csum= 0 ;
  45         newcsum = calc_csum(super, MD_SB_BYTES);
  46         super->sb_csum = csum;
  47         return newcsum;
  48 }
  49
  50
  51 static void super0_swap_endian(struct mdp_superblock_s *sb)
  52 {
  53         /* as super0 superblocks are host-endian, it is sometimes
  54          * useful to be able to swap the endianness
  55          * as (almost) everything is u32's we byte-swap every 4byte
  56          * number.
  57          * We then also have to swap the events_hi and events_lo
  58          */
  59         char *sbc = (char *)sb;
  60         __u32 t32;
  61         int i;
  62
  63         for (i=0; i < MD_SB_BYTES ; i+=4) {
  64                 char t = sbc[i];
  65                 sbc[i] = sbc[i+3];
  66                 sbc[i+3] = t;
  67                 t=sbc[i+1];
  68                 sbc[i+1]=sbc[i+2];
  69                 sbc[i+2]=t;
  70         }
  71         t32 = sb->events_hi;
  72         sb->events_hi = sb->events_lo;
  73         sb->events_lo = t32;
  74
  75         t32 = sb->cp_events_hi;
  76         sb->cp_events_hi = sb->cp_events_lo;
  77         sb->cp_events_lo = t32;
  78
  79 }
  80
  81 #ifndef MDASSEMBLE
  82
  83 static void examine_super0(struct supertype *st, char *homehost)
  84 {
  85         mdp_super_t *sb = st->sb;
  86         time_t atime;
  87         int d;
  88         int delta_extra = 0;
  89         char *c;
  90
  91         printf("          Magic : %08x\n", sb->md_magic);
  92         printf("        Version : %d.%02d.%02d\n", sb->major_version, sb->minor_version,
  93                sb->patch_version);
  94         if (sb->minor_version >= 90) {
  95                 printf("           UUID : %08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
  96                        sb->set_uuid2, sb->set_uuid3);
  97                 if (homehost) {
  98                         char buf[20];
  99                         void *hash = sha1_buffer(homehost,
 100                                                  strlen(homehost),
 101                                                  buf);
 102                         if (memcmp(&sb->set_uuid2, hash, 8)==0)
 103                                 printf(" (local to host %s)", homehost);
 104                 }
 105                 printf("\n");
 106         } else
 107                 printf("           UUID : %08x\n", sb->set_uuid0);
 108
 109         if (sb->not_persistent)
 110                 printf("           Eedk : not persistent\n");
 111
 112         atime = sb->ctime;
 113         printf("  Creation Time : %.24s\n", ctime(&atime));
 114         c=map_num(pers, sb->level);
 115         printf("     Raid Level : %s\n", c?c:"-unknown-");
 116         if ((int)sb->level > 0) {
 117                 int ddsks=0;
 118                 printf("  Used Dev Size : %d%s\n", sb->size,
 119                        human_size((long long)sb->size<<10));
 120                 switch(sb->level) {
 121                 case 1: ddsks=1;break;
 122                 case 4:
 123                 case 5: ddsks = sb->raid_disks-1; break;
 124                 case 6: ddsks = sb->raid_disks-2; break;
 125                 case 10: ddsks = sb->raid_disks / (sb->layout&255) / ((sb->layout>>8)&255);
 126                 }
 127                 if (ddsks)
 128                         printf("     Array Size : %llu%s\n", (unsigned long long)ddsks * sb->size,
 129                                human_size(ddsks*(long long)sb->size<<10));
 130         }
 131         printf("   Raid Devices : %d\n", sb->raid_disks);
 132         printf("  Total Devices : %d\n", sb->nr_disks);
 133         printf("Preferred Minor : %d\n", sb->md_minor);
 134         printf("\n");
 135         if (sb->minor_version > 90 && (sb->reshape_position+1) != 0) {
 136                 printf("  Reshape pos'n : %llu%s\n", (unsigned long long)sb->reshape_position/2, human_size((long long)sb->reshape_position<<9));
 137                 if (sb->delta_disks) {
 138                         printf("  Delta Devices : %d", sb->delta_disks);
 139                         printf(" (%d->%d)\n", sb->raid_disks-sb->delta_disks, sb->raid_disks);
 140                         if (((int)sb->delta_disks) < 0)
 141                                 delta_extra = - sb->delta_disks;
 142                 }
 143                 if (sb->new_level != sb->level) {
 144                         c = map_num(pers, sb->new_level);
 145                         printf("      New Level : %s\n", c?c:"-unknown-");
 146                 }
 147                 if (sb->new_layout != sb->layout) {
 148                         if (sb->level == 5) {
 149                                 c = map_num(r5layout, sb->new_layout);
 150                                 printf("     New Layout : %s\n", c?c:"-unknown-");
 151                         }
 152                         if (sb->level == 6) {
 153                                 c = map_num(r6layout, sb->new_layout);
 154                                 printf("     New Layout : %s\n", c?c:"-unknown-");
 155                         }
 156                         if (sb->level == 10) {
 157                                 printf("     New Layout : near=%d, %s=%d\n",
 158                                        sb->new_layout&255,
 159                                        (sb->new_layout&0x10000)?"offset":"far",
 160                                        (sb->new_layout>>8)&255);
 161                         }
 162                 }
 163                 if (sb->new_chunk != sb->chunk_size)
 164                         printf("  New Chunksize : %d\n", sb->new_chunk);
 165                 printf("\n");
 166         }
 167         atime = sb->utime;
 168         printf("    Update Time : %.24s\n", ctime(&atime));
 169         printf("          State : %s\n",
 170                (sb->state&(1<<MD_SB_CLEAN))?"clean":"active");
 171         if (sb->state & (1<<MD_SB_BITMAP_PRESENT))
 172                 printf("Internal Bitmap : present\n");
 173         printf(" Active Devices : %d\n", sb->active_disks);
 174         printf("Working Devices : %d\n", sb->working_disks);
 175         printf(" Failed Devices : %d\n", sb->failed_disks);
 176         printf("  Spare Devices : %d\n", sb->spare_disks);
 177         if (calc_sb0_csum(sb) == sb->sb_csum)
 178                 printf("       Checksum : %x - correct\n", sb->sb_csum);
 179         else
 180                 printf("       Checksum : %x - expected %lx\n", sb->sb_csum, calc_sb0_csum(sb));
 181         printf("         Events : %llu\n",
 182                ((unsigned long long)sb->events_hi << 32)
 183                + sb->events_lo);
 184         printf("\n");
 185         if (sb->level == 5) {
 186                 c = map_num(r5layout, sb->layout);
 187                 printf("         Layout : %s\n", c?c:"-unknown-");
 188         }
 189         if (sb->level == 6) {
 190                 c = map_num(r6layout, sb->layout);
 191                 printf("         Layout : %s\n", c?c:"-unknown-");
 192         }
 193         if (sb->level == 10) {
 194                 printf("         Layout :");
 195                 print_r10_layout(sb->layout);
 196                 printf("\n");
 197         }
 198         switch(sb->level) {
 199         case 0:
 200         case 4:
 201         case 5:
 202         case 6:
 203         case 10:
 204                 printf("     Chunk Size : %dK\n", sb->chunk_size/1024);
 205                 break;
 206         case -1:
 207                 printf("       Rounding : %dK\n", sb->chunk_size/1024);
 208                 break;
 209         default: break;
 210         }
 211         printf("\n");
 212         printf("      Number   Major   Minor   RaidDevice State\n");
 213         for (d= -1; d<(signed int)(sb->raid_disks+delta_extra + sb->spare_disks); d++) {
 214                 mdp_disk_t *dp;
 215                 char *dv;
 216                 char nb[5];
 217                 int wonly;
 218                 if (d>=0) dp = &sb->disks[d];
 219                 else dp = &sb->this_disk;
 220                 snprintf(nb, sizeof(nb), "%4d", d);
 221                 printf("%4s %5d   %5d    %5d    %5d     ", d < 0 ? "this" :  nb,
 222                        dp->number, dp->major, dp->minor, dp->raid_disk);
 223                 wonly = dp->state & (1<<MD_DISK_WRITEMOSTLY);
 224                 dp->state &= ~(1<<MD_DISK_WRITEMOSTLY);
 225                 if (dp->state & (1<<MD_DISK_FAULTY)) printf(" faulty");
 226                 if (dp->state & (1<<MD_DISK_ACTIVE)) printf(" active");
 227                 if (dp->state & (1<<MD_DISK_SYNC)) printf(" sync");
 228                 if (dp->state & (1<<MD_DISK_REMOVED)) printf(" removed");
 229                 if (wonly) printf(" write-mostly");
 230                 if (dp->state == 0) printf(" spare");
 231                 if ((dv=map_dev(dp->major, dp->minor, 0)))
 232                         printf("   %s", dv);
 233                 printf("\n");
 234                 if (d == -1) printf("\n");
 235         }
 236 }
 237
 238 static void brief_examine_super0(struct supertype *st, int verbose)
 239 {
 240         mdp_super_t *sb = st->sb;
 241         char *c=map_num(pers, sb->level);
 242         char devname[20];
 243
 244         sprintf(devname, "/dev/md%d", sb->md_minor);
 245
 246         if (verbose) {
 247                 printf("ARRAY %s level=%s num-devices=%d",
 248                        devname,
 249                        c?c:"-unknown-", sb->raid_disks);
 250         } else
 251                 printf("ARRAY %s", devname);
 252
 253         if (sb->minor_version >= 90)
 254                 printf(" UUID=%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
 255                        sb->set_uuid2, sb->set_uuid3);
 256         else
 257                 printf(" UUID=%08x", sb->set_uuid0);
 258         printf("\n");
 259 }
 260
 261 static void export_examine_super0(struct supertype *st)
 262 {
 263         mdp_super_t *sb = st->sb;
 264
 265         printf("MD_LEVEL=%s\n", map_num(pers, sb->level));
 266         printf("MD_DEVICES=%d\n", sb->raid_disks);
 267         if (sb->minor_version >= 90)
 268                 printf("MD_UUID=%08x:%08x:%08x:%08x\n",
 269                        sb->set_uuid0, sb->set_uuid1,
 270                        sb->set_uuid2, sb->set_uuid3);
 271         else
 272                 printf("MD_UUID=%08x\n", sb->set_uuid0);
 273         printf("MD_UPDATE_TIME=%llu\n",
 274                __le64_to_cpu(sb->ctime) & 0xFFFFFFFFFFULL);
 275         printf("MD_EVENTS=%llu\n",
 276                ((unsigned long long)sb->events_hi << 32)
 277                + sb->events_lo);
 278 }
 279
 280 static void detail_super0(struct supertype *st, char *homehost)
 281 {
 282         mdp_super_t *sb = st->sb;
 283         printf("           UUID : ");
 284         if (sb->minor_version >= 90)
 285                 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
 286                        sb->set_uuid2, sb->set_uuid3);
 287         else
 288                 printf("%08x", sb->set_uuid0);
 289         if (homehost) {
 290                 char buf[20];
 291                 void *hash = sha1_buffer(homehost,
 292                                          strlen(homehost),
 293                                          buf);
 294                 if (memcmp(&sb->set_uuid2, hash, 8)==0)
 295                         printf(" (local to host %s)", homehost);
 296         }
 297         printf("\n         Events : %d.%d\n\n", sb->events_hi, sb->events_lo);
 298 }
 299
 300 static void brief_detail_super0(struct supertype *st)
 301 {
 302         mdp_super_t *sb = st->sb;
 303         printf(" UUID=");
 304         if (sb->minor_version >= 90)
 305                 printf("%08x:%08x:%08x:%08x", sb->set_uuid0, sb->set_uuid1,
 306                        sb->set_uuid2, sb->set_uuid3);
 307         else
 308                 printf("%08x", sb->set_uuid0);
 309 }
 310 #endif
 311
 312 static int match_home0(struct supertype *st, char *homehost)
 313 {
 314         mdp_super_t *sb = st->sb;
 315         char buf[20];
 316         char *hash;
 317
 318         if (!homehost)
 319                 return 0;
 320         hash = sha1_buffer(homehost,
 321                            strlen(homehost),
 322                            buf);
 323
 324         return (memcmp(&sb->set_uuid2, hash, 8)==0);
 325 }
 326
 327 static void uuid_from_super0(struct supertype *st, int uuid[4])
 328 {
 329         mdp_super_t *super = st->sb;
 330         uuid[0] = super->set_uuid0;
 331         if (super->minor_version >= 90) {
 332                 uuid[1] = super->set_uuid1;
 333                 uuid[2] = super->set_uuid2;
 334                 uuid[3] = super->set_uuid3;
 335         } else {
 336                 uuid[1] = 0;
 337                 uuid[2] = 0;
 338                 uuid[3] = 0;
 339         }
 340 }
 341
 342 static void getinfo_super0(struct supertype *st, struct mdinfo *info, char *map)
 343 {
 344         mdp_super_t *sb = st->sb;
 345         int working = 0;
 346         int i;
 347         int map_disks = info->array.raid_disks;
 348
 349         info->array.major_version = sb->major_version;
 350         info->array.minor_version = sb->minor_version;
 351         info->array.patch_version = sb->patch_version;
 352         info->array.raid_disks = sb->raid_disks;
 353         info->array.level = sb->level;
 354         info->array.layout = sb->layout;
 355         info->array.md_minor = sb->md_minor;
 356         info->array.ctime = sb->ctime;
 357         info->array.utime = sb->utime;
 358         info->array.chunk_size = sb->chunk_size;
 359         info->array.state = sb->state;
 360         info->component_size = sb->size*2;
 361
 362         info->disk.state = sb->this_disk.state;
 363         info->disk.major = sb->this_disk.major;
 364         info->disk.minor = sb->this_disk.minor;
 365         info->disk.raid_disk = sb->this_disk.raid_disk;
 366         info->disk.number = sb->this_disk.number;
 367
 368         info->events = md_event(sb);
 369         info->data_offset = 0;
 370
 371         sprintf(info->text_version, "0.%d", sb->minor_version);
 372         info->safe_mode_delay = 200;
 373
 374         uuid_from_super0(st, info->uuid);
 375
 376         info->recovery_start = MaxSector;
 377         if (sb->minor_version > 90 && (sb->reshape_position+1) != 0) {
 378                 info->reshape_active = 1;
 379                 info->reshape_progress = sb->reshape_position;
 380                 info->new_level = sb->new_level;
 381                 info->delta_disks = sb->delta_disks;
 382                 info->new_layout = sb->new_layout;
 383                 info->new_chunk = sb->new_chunk;
 384                 if (info->delta_disks < 0)
 385                         info->array.raid_disks -= info->delta_disks;
 386         } else
 387                 info->reshape_active = 0;
 388
 389         sprintf(info->name, "%d", sb->md_minor);
 390         /* work_disks is calculated rather than read directly */
 391         for (i=0; i < MD_SB_DISKS; i++)
 392                 if ((sb->disks[i].state & (1<<MD_DISK_SYNC)) &&
 393                     (sb->disks[i].raid_disk < (unsigned)info->array.raid_disks) &&
 394                     (sb->disks[i].state & (1<<MD_DISK_ACTIVE)) &&
 395                     !(sb->disks[i].state & (1<<MD_DISK_FAULTY))) {
 396                         working ++;
 397                         if (map && i < map_disks)
 398                                 map[i] = 1;
 399                 } else if (map && i < map_disks)
 400                         map[i] = 0;
 401         info->array.working_disks = working;
 402 }
 403
 404
 405 static int update_super0(struct supertype *st, struct mdinfo *info,
 406                          char *update,
 407                          char *devname, int verbose,
 408                          int uuid_set, char *homehost)
 409 {
 410         /* NOTE: for 'assemble' and 'force' we need to return non-zero if any change was made.
 411          * For others, the return value is ignored.
 412          */
 413         int rv = 0;
 414         mdp_super_t *sb = st->sb;
 415         if (strcmp(update, "sparc2.2")==0 ) {
 416                 /* 2.2 sparc put the events in the wrong place
 417                  * So we copy the tail of the superblock
 418                  * up 4 bytes before continuing
 419                  */
 420                 __u32 *sb32 = (__u32*)sb;
 421                 memcpy(sb32+MD_SB_GENERIC_CONSTANT_WORDS+7,
 422                        sb32+MD_SB_GENERIC_CONSTANT_WORDS+7+1,
 423                        (MD_SB_WORDS - (MD_SB_GENERIC_CONSTANT_WORDS+7+1))*4);
 424                 if (verbose >= 0)
 425                         fprintf (stderr, Name ": adjusting superblock of %s for 2.2/sparc compatability.\n",
 426                                  devname);
 427         } else if (strcmp(update, "super-minor") ==0) {
 428                 sb->md_minor = info->array.md_minor;
 429                 if (verbose > 0)
 430                         fprintf(stderr, Name ": updating superblock of %s with minor number %d\n",
 431                                 devname, info->array.md_minor);
 432         } else if (strcmp(update, "summaries") == 0) {
 433                 unsigned int i;
 434                 /* set nr_disks, active_disks, working_disks,
 435                  * failed_disks, spare_disks based on disks[]
 436                  * array in superblock.
 437                  * Also make sure extra slots aren't 'failed'
 438                  */
 439                 sb->nr_disks = sb->active_disks =
 440                         sb->working_disks = sb->failed_disks =
 441                         sb->spare_disks = 0;
 442                 for (i=0; i < MD_SB_DISKS ; i++)
 443                         if (sb->disks[i].major ||
 444                             sb->disks[i].minor) {
 445                                 int state = sb->disks[i].state;
 446                                 if (state & (1<<MD_DISK_REMOVED))
 447                                         continue;
 448                                 sb->nr_disks++;
 449                                 if (state & (1<<MD_DISK_ACTIVE))
 450                                         sb->active_disks++;
 451                                 if (state & (1<<MD_DISK_FAULTY))
 452                                         sb->failed_disks++;
 453                                 else
 454                                         sb->working_disks++;
 455                                 if (state == 0)
 456                                         sb->spare_disks++;
 457                         } else if (i >= sb->raid_disks && sb->disks[i].number == 0)
 458                                 sb->disks[i].state = 0;
 459         } else if (strcmp(update, "force-one")==0) {
 460                 /* Not enough devices for a working array, so
 461                  * bring this one up-to-date.
 462                  */
 463                 __u32 ehi = sb->events_hi, elo = sb->events_lo;
 464                 sb->events_hi = (info->events>>32) & 0xFFFFFFFF;
 465                 sb->events_lo = (info->events) & 0xFFFFFFFF;
 466                 if (sb->events_hi != ehi ||
 467                     sb->events_lo != elo)
 468                         rv = 1;
 469         } else if (strcmp(update, "force-array")==0) {
 470                 /* degraded array and 'force' requested, so
 471                  * maybe need to mark it 'clean'
 472                  */
 473                 if ((sb->level == 5 || sb->level == 4 || sb->level == 6) &&
 474                     (sb->state & (1 << MD_SB_CLEAN)) == 0) {
 475                         /* need to force clean */
 476                         sb->state |= (1 << MD_SB_CLEAN);
 477                         rv = 1;
 478                 }
 479         } else if (strcmp(update, "assemble")==0) {
 480                 int d = info->disk.number;
 481                 int wonly = sb->disks[d].state & (1<<MD_DISK_WRITEMOSTLY);
 482                 int mask = (1<<MD_DISK_WRITEMOSTLY);
 483                 int add = 0;
 484                 if (sb->minor_version >= 91)
 485                         /* During reshape we don't insist on everything
 486                          * being marked 'sync'
 487                          */
 488                         add = (1<<MD_DISK_SYNC);
 489                 if (((sb->disks[d].state & ~mask) | add)
 490                     != (unsigned)info->disk.state) {
 491                         sb->disks[d].state = info->disk.state | wonly;
 492                         rv = 1;
 493                 }
 494         } else if (strcmp(update, "linear-grow-new") == 0) {
 495                 memset(&sb->disks[info->disk.number], 0, sizeof(sb->disks[0]));
 496                 sb->disks[info->disk.number].number = info->disk.number;
 497                 sb->disks[info->disk.number].major = info->disk.major;
 498                 sb->disks[info->disk.number].minor = info->disk.minor;
 499                 sb->disks[info->disk.number].raid_disk = info->disk.raid_disk;
 500                 sb->disks[info->disk.number].state = info->disk.state;
 501                 sb->this_disk = sb->disks[info->disk.number];
 502         } else if (strcmp(update, "linear-grow-update") == 0) {
 503                 sb->raid_disks = info->array.raid_disks;
 504                 sb->nr_disks = info->array.nr_disks;
 505                 sb->active_disks = info->array.active_disks;
 506                 sb->working_disks = info->array.working_disks;
 507                 memset(&sb->disks[info->disk.number], 0, sizeof(sb->disks[0]));
 508                 sb->disks[info->disk.number].number = info->disk.number;
 509                 sb->disks[info->disk.number].major = info->disk.major;
 510                 sb->disks[info->disk.number].minor = info->disk.minor;
 511                 sb->disks[info->disk.number].raid_disk = info->disk.raid_disk;
 512                 sb->disks[info->disk.number].state = info->disk.state;
 513         } else if (strcmp(update, "resync") == 0) {
 514                 /* make sure resync happens */
 515                 sb->state &= ~(1<<MD_SB_CLEAN);
 516                 sb->recovery_cp = 0;
 517         } else if (strcmp(update, "homehost") == 0 &&
 518             homehost) {
 519                 uuid_set = 0;
 520                 update = "uuid";
 521                 info->uuid[0] = sb->set_uuid0;
 522                 info->uuid[1] = sb->set_uuid1;
 523         } else if (strcmp(update, "uuid") == 0) {
 524                 if (!uuid_set && homehost) {
 525                         char buf[20];
 526                         char *hash = sha1_buffer(homehost,
 527                                                  strlen(homehost),
 528                                                  buf);
 529                         memcpy(info->uuid+2, hash, 8);
 530                 }
 531                 sb->set_uuid0 = info->uuid[0];
 532                 sb->set_uuid1 = info->uuid[1];
 533                 sb->set_uuid2 = info->uuid[2];
 534                 sb->set_uuid3 = info->uuid[3];
 535                 if (sb->state & (1<<MD_SB_BITMAP_PRESENT)) {
 536                         struct bitmap_super_s *bm;
 537                         bm = (struct bitmap_super_s*)(sb+1);
 538                         uuid_from_super0(st, (int*)bm->uuid);
 539                 }
 540         } else if (strcmp(update, "_reshape_progress")==0)
 541                 sb->reshape_position = info->reshape_progress;
 542         else
 543                 rv = -1;
 544
 545         sb->sb_csum = calc_sb0_csum(sb);
 546         return rv;
 547 }
 548
 549 /*
 550  * For verion-0 superblock, the homehost is 'stored' in the
 551  * uuid.  8 bytes for a hash of the host leaving 8 bytes
 552  * of random material.
 553  * We use the first 8 bytes (64bits) of the sha1 of the
 554  * host name
 555  */
 556
 557
 558 static int init_super0(struct supertype *st, mdu_array_info_t *info,
 559                        unsigned long long size, char *ignored_name, char *homehost,
 560                        int *uuid)
 561 {
 562         mdp_super_t *sb;
 563         int spares;
 564
 565         if (posix_memalign((void**)&sb, 4096,
 566                            MD_SB_BYTES + ROUND_UP(sizeof(bitmap_super_t), 4096)) != 0) {
 567                 fprintf(stderr, Name ": %s could not allocate superblock\n", __func__);
 568                 return 0;
 569         }
 570         memset(sb, 0, MD_SB_BYTES + sizeof(bitmap_super_t));
 571
 572         st->sb = sb;
 573         if (info == NULL) {
 574                 /* zeroing the superblock */
 575                 return 0;
 576         }
 577
 578         spares = info->working_disks - info->active_disks;
 579         if (info->raid_disks + spares  > MD_SB_DISKS) {
 580                 fprintf(stderr, Name ": too many devices requested: %d+%d > %d\n",
 581                         info->raid_disks , spares, MD_SB_DISKS);
 582                 return 0;
 583         }
 584
 585         sb->md_magic = MD_SB_MAGIC;
 586         sb->major_version = 0;
 587         sb->minor_version = 90;
 588         sb->patch_version = 0;
 589         sb->gvalid_words = 0; /* ignored */
 590         sb->ctime = time(0);
 591         sb->level = info->level;
 592         if (size != (unsigned long long)info->size)
 593                 return 0;
 594         sb->size = info->size;
 595         sb->nr_disks = info->nr_disks;
 596         sb->raid_disks = info->raid_disks;
 597         sb->md_minor = info->md_minor;
 598         sb->not_persistent = 0;
 599         if (uuid) {
 600                 sb->set_uuid0 = uuid[0];
 601                 sb->set_uuid1 = uuid[1];
 602                 sb->set_uuid2 = uuid[2];
 603                 sb->set_uuid3 = uuid[3];
 604         } else {
 605                 int rfd = open("/dev/urandom", O_RDONLY);
 606                 if (rfd < 0 || read(rfd, &sb->set_uuid0, 4) != 4)
 607                         sb->set_uuid0 = random();
 608                 if (rfd < 0 || read(rfd, &sb->set_uuid1, 12) != 12) {
 609                         sb->set_uuid1 = random();
 610                         sb->set_uuid2 = random();
 611                         sb->set_uuid3 = random();
 612                 }
 613                 if (rfd >= 0)
 614                         close(rfd);
 615         }
 616         if (homehost) {
 617                 char buf[20];
 618                 char *hash = sha1_buffer(homehost,
 619                                          strlen(homehost),
 620                                          buf);
 621                 memcpy(&sb->set_uuid2, hash, 8);
 622         }
 623
 624         sb->utime = sb->ctime;
 625         sb->state = info->state;
 626         sb->active_disks = info->active_disks;
 627         sb->working_disks = info->working_disks;
 628         sb->failed_disks = info->failed_disks;
 629         sb->spare_disks = info->spare_disks;
 630         sb->events_hi = 0;
 631         sb->events_lo = 1;
 632
 633         sb->layout = info->layout;
 634         sb->chunk_size = info->chunk_size;
 635
 636         return 1;
 637 }
 638
 639 struct devinfo {
 640         int fd;
 641         char *devname;
 642         mdu_disk_info_t disk;
 643         struct devinfo *next;
 644 };
 645
 646 #ifndef MDASSEMBLE
 647 /* Add a device to the superblock being created */
 648 static int add_to_super0(struct supertype *st, mdu_disk_info_t *dinfo,
 649                           int fd, char *devname)
 650 {
 651         mdp_super_t *sb = st->sb;
 652         mdp_disk_t *dk = &sb->disks[dinfo->number];
 653         struct devinfo *di, **dip;
 654
 655         dk->number = dinfo->number;
 656         dk->major = dinfo->major;
 657         dk->minor = dinfo->minor;
 658         dk->raid_disk = dinfo->raid_disk;
 659         dk->state = dinfo->state;
 660
 661         sb->this_disk = sb->disks[dinfo->number];
 662         sb->sb_csum = calc_sb0_csum(sb);
 663
 664         dip = (struct devinfo **)&st->info;
 665         while (*dip)
 666                 dip = &(*dip)->next;
 667         di = malloc(sizeof(struct devinfo));
 668         di->fd = fd;
 669         di->devname = devname;
 670         di->disk = *dinfo;
 671         di->next = NULL;
 672         *dip = di;
 673
 674         return 0;
 675 }
 676 #endif
 677
 678 static int store_super0(struct supertype *st, int fd)
 679 {
 680         unsigned long long dsize;
 681         unsigned long long offset;
 682         mdp_super_t *super = st->sb;
 683
 684         if (!get_dev_size(fd, NULL, &dsize))
 685                 return 1;
 686
 687         if (dsize < MD_RESERVED_SECTORS*512)
 688                 return 2;
 689
 690         offset = MD_NEW_SIZE_SECTORS(dsize>>9);
 691
 692         offset *= 512;
 693
 694         if (lseek64(fd, offset, 0)< 0LL)
 695                 return 3;
 696
 697         if (write(fd, super, sizeof(*super)) != sizeof(*super))
 698                 return 4;
 699
 700         if (super->state & (1<<MD_SB_BITMAP_PRESENT)) {
 701                 struct bitmap_super_s * bm = (struct bitmap_super_s*)(super+1);
 702                 if (__le32_to_cpu(bm->magic) == BITMAP_MAGIC)
 703                         if (write(fd, bm, ROUND_UP(sizeof(*bm),4096)) !=
 704                             ROUND_UP(sizeof(*bm),4096))
 705                             return 5;
 706         }
 707
 708         fsync(fd);
 709         return 0;
 710 }
 711
 712 #ifndef MDASSEMBLE
 713 static int write_init_super0(struct supertype *st)
 714 {
 715         mdp_super_t *sb = st->sb;
 716         int rv = 0;
 717         struct devinfo *di;
 718
 719         for (di = st->info ; di && ! rv ; di = di->next) {
 720
 721                 if (di->disk.state == 1)
 722                         continue;
 723                 if (di->fd == -1)
 724                         continue;
 725                 while (Kill(di->devname, NULL, 0, 1, 1) == 0)
 726                         ;
 727
 728                 sb->disks[di->disk.number].state &= ~(1<<MD_DISK_FAULTY);
 729
 730                 sb->this_disk = sb->disks[di->disk.number];
 731                 sb->sb_csum = calc_sb0_csum(sb);
 732                 rv = store_super0(st, di->fd);
 733
 734                 if (rv == 0 && (sb->state & (1<<MD_SB_BITMAP_PRESENT)))
 735                         rv = st->ss->write_bitmap(st, di->fd);
 736
 737                 if (rv)
 738                         fprintf(stderr,
 739                                 Name ": failed to write superblock to %s\n",
 740                                 di->devname);
 741                 close(di->fd);
 742                 di->fd = -1;
 743         }
 744         return rv;
 745 }
 746 #endif
 747
 748 static int compare_super0(struct supertype *st, struct supertype *tst)
 749 {
 750         /*
 751          * return:
 752          *  0 same, or first was empty, and second was copied
 753          *  1 second had wrong number
 754          *  2 wrong uuid
 755          *  3 wrong other info
 756          */
 757         mdp_super_t *first = st->sb;
 758         mdp_super_t *second = tst->sb;
 759         int uuid1[4], uuid2[4];
 760
 761         if (second->md_magic != MD_SB_MAGIC)
 762                 return 1;
 763         if (!first) {
 764                 if (posix_memalign((void**)&first, 4096,
 765                              MD_SB_BYTES +
 766                              ROUND_UP(sizeof(struct bitmap_super_s), 4096)) != 0) {
 767                         fprintf(stderr, Name
 768                                 ": %s could not allocate superblock\n", __func__);
 769                         return 1;
 770                 }
 771                 memcpy(first, second, MD_SB_BYTES + sizeof(struct bitmap_super_s));
 772                 st->sb = first;
 773                 return 0;
 774         }
 775
 776         uuid_from_super0(st, uuid1);
 777         uuid_from_super0(tst, uuid2);
 778         if (!same_uuid(uuid1, uuid2, 0))
 779                 return 2;
 780         if (first->major_version != second->major_version ||
 781             first->minor_version != second->minor_version ||
 782             first->patch_version != second->patch_version ||
 783             first->gvalid_words  != second->gvalid_words  ||
 784             first->ctime         != second->ctime         ||
 785             first->level         != second->level         ||
 786             first->size          != second->size          ||
 787             first->raid_disks    != second->raid_disks    )
 788                 return 3;
 789
 790         return 0;
 791 }
 792
 793
 794 static void free_super0(struct supertype *st);
 795
 796 static int load_super0(struct supertype *st, int fd, char *devname)
 797 {
 798         /* try to read in the superblock
 799          * Return:
 800          *  0 on success
 801          *  1 on cannot get superblock
 802          *  2 on superblock meaningless
 803          */
 804         unsigned long long dsize;
 805         unsigned long long offset;
 806         mdp_super_t *super;
 807         int uuid[4];
 808         struct bitmap_super_s *bsb;
 809
 810         free_super0(st);
 811
 812         if (st->subarray[0])
 813                 return 1;
 814
 815         if (!get_dev_size(fd, devname, &dsize))
 816                 return 1;
 817
 818         if (dsize < MD_RESERVED_SECTORS*512) {
 819                 if (devname)
 820                         fprintf(stderr, Name
 821                             ": %s is too small for md: size is %llu sectors.\n",
 822                                 devname, dsize);
 823                 return 1;
 824         }
 825
 826         offset = MD_NEW_SIZE_SECTORS(dsize>>9);
 827
 828         offset *= 512;
 829
 830         ioctl(fd, BLKFLSBUF, 0); /* make sure we read current data */
 831
 832         if (lseek64(fd, offset, 0)< 0LL) {
 833                 if (devname)
 834                         fprintf(stderr, Name ": Cannot seek to superblock on %s: %s\n",
 835                                 devname, strerror(errno));
 836                 return 1;
 837         }
 838
 839         if (posix_memalign((void**)&super, 4096,
 840                            MD_SB_BYTES +
 841                            ROUND_UP(sizeof(bitmap_super_t), 4096)) != 0) {
 842                 fprintf(stderr, Name
 843                         ": %s could not allocate superblock\n", __func__);
 844                 return 1;
 845         }
 846
 847         if (read(fd, super, sizeof(*super)) != MD_SB_BYTES) {
 848                 if (devname)
 849                         fprintf(stderr, Name ": Cannot read superblock on %s\n",
 850                                 devname);
 851                 free(super);
 852                 return 1;
 853         }
 854
 855         if (st->ss && st->minor_version == 9)
 856                 super0_swap_endian(super);
 857
 858         if (super->md_magic != MD_SB_MAGIC) {
 859                 if (devname)
 860                         fprintf(stderr, Name ": No super block found on %s (Expected magic %08x, got %08x)\n",
 861                                 devname, MD_SB_MAGIC, super->md_magic);
 862                 free(super);
 863                 return 2;
 864         }
 865
 866         if (super->major_version != 0) {
 867                 if (devname)
 868                         fprintf(stderr, Name ": Cannot interpret superblock on %s - version is %d\n",
 869                                 devname, super->major_version);
 870                 free(super);
 871                 return 2;
 872         }
 873         st->sb = super;
 874
 875         if (st->ss == NULL) {
 876                 st->ss = &super0;
 877                 st->minor_version = super->minor_version;
 878                 st->max_devs = MD_SB_DISKS;
 879                 st->info = NULL;
 880         }
 881
 882         /* Now check on the bitmap superblock */
 883         if ((super->state & (1<<MD_SB_BITMAP_PRESENT)) == 0)
 884                 return 0;
 885         /* Read the bitmap superblock and make sure it looks
 886          * valid.  If it doesn't clear the bit.  An --assemble --force
 887          * should get that written out.
 888          */
 889         if (read(fd, super+1, ROUND_UP(sizeof(struct bitmap_super_s),4096))
 890             != ROUND_UP(sizeof(struct bitmap_super_s),4096))
 891                 goto no_bitmap;
 892
 893         uuid_from_super0(st, uuid);
 894         bsb = (struct bitmap_super_s *)(super+1);
 895         if (__le32_to_cpu(bsb->magic) != BITMAP_MAGIC ||
 896             memcmp(bsb->uuid, uuid, 16) != 0)
 897                 goto no_bitmap;
 898         return 0;
 899
 900  no_bitmap:
 901         super->state &= ~(1<<MD_SB_BITMAP_PRESENT);
 902
 903         return 0;
 904 }
 905
 906 static struct supertype *match_metadata_desc0(char *arg)
 907 {
 908         struct supertype *st = malloc(sizeof(*st));
 909         if (!st) return st;
 910
 911         memset(st, 0, sizeof(*st));
 912         st->ss = &super0;
 913         st->info = NULL;
 914         st->minor_version = 90;
 915         st->max_devs = MD_SB_DISKS;
 916         st->sb = NULL;
 917         /* we sometimes get 00.90 */
 918         while (arg[0] == '0' && arg[1] == '0')
 919                 arg++;
 920         if (strcmp(arg, "0") == 0 ||
 921 #ifdef DEFAULT_OLD_METADATA /* ifndef in super1.c */
 922             strcmp(arg, "default") == 0 ||
 923 #endif /* DEFAULT_OLD_METADATA */
 924             strcmp(arg, "0.90") == 0 ||
 925             strcmp(arg, "") == 0 /* no metadata  - i.e. non_persistent */
 926                 )
 927                 return st;
 928
 929         st->minor_version = 91; /* reshape in progress */
 930         if (strcmp(arg, "0.91") == 0) /* For dup_super support */
 931                 return st;
 932
 933         st->minor_version = 9; /* flag for 'byte-swapped' */
 934         if (strcmp(arg, "0.swap")==0 ||
 935             strcmp(arg, "0.9") == 0) /* For dup_super support */
 936                 return st;
 937
 938         free(st);
 939         return NULL;
 940 }
 941
 942 static __u64 avail_size0(struct supertype *st, __u64 devsize)
 943 {
 944         if (devsize < MD_RESERVED_SECTORS)
 945                 return 0ULL;
 946         return MD_NEW_SIZE_SECTORS(devsize);
 947 }
 948
 949 static int add_internal_bitmap0(struct supertype *st, int *chunkp,
 950                                 int delay, int write_behind,
 951                                 unsigned long long size, int may_change,
 952                                 int major)
 953 {
 954         /*
 955          * The bitmap comes immediately after the superblock and must be 60K in size
 956          * at most.  The default size is between 30K and 60K
 957          *
 958          * size is in sectors,  chunk is in bytes !!!
 959          */
 960         unsigned long long bits;
 961         unsigned long long max_bits = (60*1024 - sizeof(bitmap_super_t))*8;
 962         unsigned long long min_chunk;
 963         int chunk = *chunkp;
 964         mdp_super_t *sb = st->sb;
 965         bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + MD_SB_BYTES);
 966
 967
 968         min_chunk = 4096; /* sub-page chunks don't work yet.. */
 969         bits = (size * 512) / min_chunk + 1;
 970         while (bits > max_bits) {
 971                 min_chunk *= 2;
 972                 bits = (bits+1)/2;
 973         }
 974         if (chunk == UnSet) {
 975                 /* A chunk size less than a few Megabytes gives poor
 976                  * performance without increasing resync noticeably
 977                  */
 978                 chunk = min_chunk;
 979                 if (chunk < 64*1024*1024)
 980                         chunk = 64*1024*1024;
 981         } else if ((unsigned long long)chunk < min_chunk)
 982                 return 0; /* chunk size too small */
 983
 984         sb->state |= (1<<MD_SB_BITMAP_PRESENT);
 985
 986         memset(bms, 0, sizeof(*bms));
 987         bms->magic = __cpu_to_le32(BITMAP_MAGIC);
 988         bms->version = __cpu_to_le32(major);
 989         uuid_from_super0(st, (int*)bms->uuid);
 990         bms->chunksize = __cpu_to_le32(chunk);
 991         bms->daemon_sleep = __cpu_to_le32(delay);
 992         bms->sync_size = __cpu_to_le64(size);
 993         bms->write_behind = __cpu_to_le32(write_behind);
 994         *chunkp = chunk;
 995         return 1;
 996 }
 997
 998
 999 static void locate_bitmap0(struct supertype *st, int fd)
1000 {
1001         unsigned long long dsize;
1002         unsigned long long offset;
1003
1004         if (!get_dev_size(fd, NULL, &dsize))
1005                 return;
1006
1007         if (dsize < MD_RESERVED_SECTORS*512)
1008                 return;
1009
1010         offset = MD_NEW_SIZE_SECTORS(dsize>>9);
1011
1012         offset *= 512;
1013
1014         offset += MD_SB_BYTES;
1015
1016         lseek64(fd, offset, 0);
1017 }
1018
1019 static int write_bitmap0(struct supertype *st, int fd)
1020 {
1021         unsigned long long dsize;
1022         unsigned long long offset;
1023         mdp_super_t *sb = st->sb;
1024
1025         int rv = 0;
1026
1027         int towrite, n;
1028         char abuf[4096+4096];
1029         char *buf = (char*)(((long)(abuf+4096))&~4095L);
1030
1031         if (!get_dev_size(fd, NULL, &dsize))
1032                 return 1;
1033
1034
1035         if (dsize < MD_RESERVED_SECTORS*512)
1036                 return -1;
1037
1038         offset = MD_NEW_SIZE_SECTORS(dsize>>9);
1039
1040         offset *= 512;
1041
1042         if (lseek64(fd, offset + 4096, 0)< 0LL)
1043                 return 3;
1044
1045         memset(buf, 0xff, 4096);
1046         memcpy(buf,  ((char*)sb)+MD_SB_BYTES, sizeof(bitmap_super_t));
1047         towrite = 60*1024;
1048         while (towrite > 0) {
1049                 n = towrite;
1050                 if (n > 4096)
1051                         n = 4096;
1052                 n = write(fd, buf, n);
1053                 if (n > 0)
1054                         towrite -= n;
1055                 else
1056                         break;
1057                 memset(buf, 0xff, 4096);
1058         }
1059         fsync(fd);
1060         if (towrite)
1061                 rv = -2;
1062
1063         return rv;
1064 }
1065
1066 static void free_super0(struct supertype *st)
1067 {
1068         if (st->sb)
1069                 free(st->sb);
1070         st->sb = NULL;
1071 }
1072
1073 #ifndef MDASSEMBLE
1074 static int validate_geometry0(struct supertype *st, int level,
1075                               int layout, int raiddisks,
1076                               int chunk, unsigned long long size,
1077                               char *subdev, unsigned long long *freesize,
1078                               int verbose)
1079 {
1080         unsigned long long ldsize;
1081         int fd;
1082
1083         if (level == LEVEL_CONTAINER) {
1084                 if (verbose)
1085                         fprintf(stderr, Name ": 0.90 metadata does not support containers\n");
1086                 return 0;
1087         }
1088         if (raiddisks > MD_SB_DISKS) {
1089                 if (verbose)
1090                         fprintf(stderr, Name ": 0.90 metadata supports at most %d devices per array\n",
1091                                 MD_SB_DISKS);
1092                 return 0;
1093         }
1094         if (size > (0x7fffffffULL<<9)) {
1095                 if (verbose)
1096                         fprintf(stderr, Name ": 0.90 metadata supports at most 2 terrabytes per device\n");
1097                 return 0;
1098         }
1099         if (!subdev)
1100                 return 1;
1101
1102         fd = open(subdev, O_RDONLY|O_EXCL, 0);
1103         if (fd < 0) {
1104                 if (verbose)
1105                         fprintf(stderr, Name ": super0.90 cannot open %s: %s\n",
1106                                 subdev, strerror(errno));
1107                 return 0;
1108         }
1109
1110         if (!get_dev_size(fd, subdev, &ldsize)) {
1111                 close(fd);
1112                 return 0;
1113         }
1114         close(fd);
1115
1116         if (ldsize < MD_RESERVED_SECTORS * 512)
1117                 return 0;
1118         if (size > (0x7fffffffULL<<9))
1119                 return 0;
1120         *freesize = MD_NEW_SIZE_SECTORS(ldsize >> 9);
1121         return 1;
1122 }
1123 #endif /* MDASSEMBLE */
1124
1125 struct superswitch super0 = {
1126 #ifndef MDASSEMBLE
1127         .examine_super = examine_super0,
1128         .brief_examine_super = brief_examine_super0,
1129         .export_examine_super = export_examine_super0,
1130         .detail_super = detail_super0,
1131         .brief_detail_super = brief_detail_super0,
1132         .write_init_super = write_init_super0,
1133         .validate_geometry = validate_geometry0,
1134         .add_to_super = add_to_super0,
1135 #endif
1136         .match_home = match_home0,
1137         .uuid_from_super = uuid_from_super0,
1138         .getinfo_super = getinfo_super0,
1139         .update_super = update_super0,
1140         .init_super = init_super0,
1141         .store_super = store_super0,
1142         .compare_super = compare_super0,
1143         .load_super = load_super0,
1144         .match_metadata_desc = match_metadata_desc0,
1145         .avail_size = avail_size0,
1146         .add_internal_bitmap = add_internal_bitmap0,
1147         .locate_bitmap = locate_bitmap0,
1148         .write_bitmap = write_bitmap0,
1149         .free_super = free_super0,
1150         .name = "0.90",
1151 };