Create.c

   1 /*
   2  * mdadm - manage Linux "md" devices aka RAID arrays.
   3  *
   4  * Copyright (C) 2001-2006 Neil Brown <neilb@suse.de>
   5  *
   6  *
   7  *    This program is free software; you can redistribute it and/or modify
   8  *    it under the terms of the GNU General Public License as published by
   9  *    the Free Software Foundation; either version 2 of the License, or
  10  *    (at your option) any later version.
  11  *
  12  *    This program is distributed in the hope that it will be useful,
  13  *    but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  *    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  15  *    GNU General Public License for more details.
  16  *
  17  *    You should have received a copy of the GNU General Public License
  18  *    along with this program; if not, write to the Free Software
  19  *    Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  20  *
  21  *    Author: Neil Brown
  22  *    Email: <neilb@cse.unsw.edu.au>
  23  *    Paper: Neil Brown
  24  *           School of Computer Science and Engineering
  25  *           The University of New South Wales
  26  *           Sydney, 2052
  27  *           Australia
  28  */
  29
  30 #include "mdadm.h"
  31 #include        "md_u.h"
  32 #include        "md_p.h"
  33 #include        <ctype.h>
  34
  35 int Create(struct supertype *st, char *mddev, int mdfd,
  36            int chunk, int level, int layout, unsigned long long size, int raiddisks, int sparedisks,
  37            char *name, char *homehost, int *uuid,
  38            int subdevs, mddev_dev_t devlist,
  39            int runstop, int verbose, int force, int assume_clean,
  40            char *bitmap_file, int bitmap_chunk, int write_behind, int delay)
  41 {
  42         /*
  43          * Create a new raid array.
  44          *
  45          * First check that necessary details are available
  46          * (i.e. level, raid-disks)
  47          *
  48          * Then check each disk to see what might be on it
  49          * and report anything interesting.
  50          *
  51          * If anything looks odd, and runstop not set,
  52          * abort.
  53          *
  54          * SET_ARRAY_INFO and ADD_NEW_DISK, and
  55          * if runstop==run, or raiddisks disks were used,
  56          * RUN_ARRAY
  57          */
  58         unsigned long long minsize=0, maxsize=0;
  59         char *mindisc = NULL;
  60         char *maxdisc = NULL;
  61         int dnum;
  62         mddev_dev_t dv;
  63         int fail=0, warn=0;
  64         struct stat stb;
  65         int first_missing = subdevs * 2;
  66         int second_missing = subdevs * 2;
  67         int missing_disks = 0;
  68         int insert_point = subdevs * 2; /* where to insert a missing drive */
  69         int pass;
  70         int vers;
  71         int rv;
  72         int bitmap_fd;
  73         unsigned long long bitmapsize;
  74         struct mdinfo info;
  75
  76         int major_num = BITMAP_MAJOR_HI;
  77
  78         memset(&info, 0, sizeof(info));
  79
  80         vers = md_get_version(mdfd);
  81         if (vers < 9000) {
  82                 fprintf(stderr, Name ": Create requires md driver version 0.90.0 or later\n");
  83                 return 1;
  84         } else {
  85                 mdu_array_info_t inf;
  86                 memset(&inf, 0, sizeof(inf));
  87                 ioctl(mdfd, GET_ARRAY_INFO, &inf);
  88                 if (inf.working_disks != 0) {
  89                         fprintf(stderr, Name ": another array by this name"
  90                                 " is already running.\n");
  91                         return 1;
  92                 }
  93         }
  94         if (level == UnSet) {
  95                 fprintf(stderr,
  96                         Name ": a RAID level is needed to create an array.\n");
  97                 return 1;
  98         }
  99         if (raiddisks < 4 && level == 6) {
 100                 fprintf(stderr,
 101                         Name ": at least 4 raid-devices needed for level 6\n");
 102                 return 1;
 103         }
 104         if (raiddisks > 256 && level == 6) {
 105                 fprintf(stderr,
 106                         Name ": no more than 256 raid-devices supported for level 6\n");
 107                 return 1;
 108         }
 109         if (raiddisks < 2 && level >= 4) {
 110                 fprintf(stderr,
 111                         Name ": at least 2 raid-devices needed for level 4 or 5\n");
 112                 return 1;
 113         }
 114         if (level <= 0 && sparedisks) {
 115                 fprintf(stderr,
 116                         Name ": This level does not support spare devices\n");
 117                 return 1;
 118         }
 119         if (subdevs > raiddisks+sparedisks) {
 120                 fprintf(stderr, Name ": You have listed more devices (%d) than are in the array(%d)!\n", subdevs, raiddisks+sparedisks);
 121                 return 1;
 122         }
 123         if (subdevs < raiddisks+sparedisks) {
 124                 fprintf(stderr, Name ": You haven't given enough devices (real or missing) to create this array\n");
 125                 return 1;
 126         }
 127         if (bitmap_file && level <= 0) {
 128                 fprintf(stderr, Name ": bitmaps not meaningful with level %s\n",
 129                         map_num(pers, level)?:"given");
 130                 return 1;
 131         }
 132
 133         /* now set some defaults */
 134         if (layout == UnSet)
 135                 switch(level) {
 136                 default: /* no layout */
 137                         layout = 0;
 138                         break;
 139                 case 10:
 140                         layout = 0x102; /* near=2, far=1 */
 141                         if (verbose > 0)
 142                                 fprintf(stderr,
 143                                         Name ": layout defaults to n1\n");
 144                         break;
 145                 case 5:
 146                 case 6:
 147                         layout = map_name(r5layout, "default");
 148                         if (verbose > 0)
 149                                 fprintf(stderr,
 150                                         Name ": layout defaults to %s\n", map_num(r5layout, layout));
 151                         break;
 152                 case LEVEL_FAULTY:
 153                         layout = map_name(faultylayout, "default");
 154
 155                         if (verbose > 0)
 156                                 fprintf(stderr,
 157                                         Name ": layout defaults to %s\n", map_num(faultylayout, layout));
 158                         break;
 159                 }
 160
 161         if (level == 10)
 162                 /* check layout fits in array*/
 163                 if ((layout&255) * ((layout>>8)&255) > raiddisks) {
 164                         fprintf(stderr, Name ": that layout requires at least %d devices\n",
 165                                 (layout&255) * ((layout>>8)&255));
 166                         return 1;
 167                 }
 168
 169         switch(level) {
 170         case 4:
 171         case 5:
 172         case 10:
 173         case 6:
 174         case 0:
 175         case LEVEL_LINEAR: /* linear */
 176                 if (chunk == 0) {
 177                         chunk = 64;
 178                         if (verbose > 0)
 179                                 fprintf(stderr, Name ": chunk size defaults to 64K\n");
 180                 }
 181                 break;
 182         case 1:
 183         case LEVEL_FAULTY:
 184         case LEVEL_MULTIPATH:
 185                 if (chunk) {
 186                         chunk = 0;
 187                         if (verbose > 0)
 188                                 fprintf(stderr, Name ": chunk size ignored for this level\n");
 189                 }
 190                 break;
 191         default:
 192                 fprintf(stderr, Name ": unknown level %d\n", level);
 193                 return 1;
 194         }
 195
 196         /* now look at the subdevs */
 197         info.array.active_disks = 0;
 198         info.array.working_disks = 0;
 199         dnum = 0;
 200         for (dv=devlist; dv; dv=dv->next, dnum++) {
 201                 char *dname = dv->devname;
 202                 unsigned long long ldsize, freesize;
 203                 int fd;
 204                 if (strcasecmp(dname, "missing")==0) {
 205                         if (first_missing > dnum)
 206                                 first_missing = dnum;
 207                         if (second_missing > dnum && dnum > first_missing)
 208                                 second_missing = dnum;
 209                         missing_disks ++;
 210                         continue;
 211                 }
 212                 info.array.working_disks++;
 213                 if (dnum < raiddisks)
 214                         info.array.active_disks++;
 215                 fd = open(dname, O_RDONLY|O_EXCL);
 216                 if (fd <0 ) {
 217                         fprintf(stderr, Name ": Cannot open %s: %s\n",
 218                                 dname, strerror(errno));
 219                         fail=1;
 220                         continue;
 221                 }
 222                 if (!get_dev_size(fd, dname, &ldsize)) {
 223                         fail = 1;
 224                         close(fd);
 225                         continue;
 226                 }
 227                 if (st == NULL) {
 228                         struct createinfo *ci = conf_get_create_info();
 229                         if (ci)
 230                                 st = ci->supertype;
 231                 }
 232                 if (st == NULL) {
 233                         /* Need to choose a default metadata, which is different
 234                          * depending on the sizes of devices
 235                          */
 236                         int i;
 237                         char *name = "default";
 238                         if (level >= 1 && ldsize > (0x7fffffffULL<<10))
 239                                 name = "default/large";
 240                         for(i=0; !st && superlist[i]; i++)
 241                                 st = superlist[i]->match_metadata_desc(name);
 242
 243                         if (!st) {
 244                                 fprintf(stderr, Name ": internal error - no default metadata style\n");
 245                                 exit(2);
 246                         }
 247                         if (st->ss->major != 0 ||
 248                             st->minor_version != 90)
 249                                 fprintf(stderr, Name ": Defaulting to version"
 250                                         " %d.%d metadata\n",
 251                                         st->ss->major,
 252                                         st->minor_version);
 253                 }
 254                 freesize = st->ss->avail_size(st, ldsize >> 9);
 255                 if (freesize == 0) {
 256                         fprintf(stderr, Name ": %s is too small: %luK\n",
 257                                 dname, (unsigned long)(ldsize>>10));
 258                         fail = 1;
 259                         close(fd);
 260                         continue;
 261                 }
 262
 263                 freesize /= 2; /* convert to K */
 264                 if (chunk) {
 265                         /* round to chunk size */
 266                         freesize = freesize & ~(chunk-1);
 267                 }
 268
 269                 if (size && freesize < size) {
 270                         fprintf(stderr, Name ": %s is smaller that given size."
 271                                 " %lluK < %lluK + superblock\n", dname, freesize, size);
 272                         fail = 1;
 273                         close(fd);
 274                         continue;
 275                 }
 276                 if (maxdisc == NULL || (maxdisc && freesize > maxsize)) {
 277                         maxdisc = dname;
 278                         maxsize = freesize;
 279                 }
 280                 if (mindisc ==NULL || (mindisc && freesize < minsize)) {
 281                         mindisc = dname;
 282                         minsize = freesize;
 283                 }
 284                 if (runstop != 1 || verbose >= 0) {
 285                         warn |= check_ext2(fd, dname);
 286                         warn |= check_reiser(fd, dname);
 287                         warn |= check_raid(fd, dname);
 288                 }
 289                 close(fd);
 290         }
 291         if (fail) {
 292                 fprintf(stderr, Name ": create aborted\n");
 293                 return 1;
 294         }
 295         if (size == 0) {
 296                 if (mindisc == NULL) {
 297                         fprintf(stderr, Name ": no size and no drives given - aborting create.\n");
 298                         return 1;
 299                 }
 300                 if (level > 0 || level == LEVEL_MULTIPATH || level == LEVEL_FAULTY) {
 301                         /* size is meaningful */
 302                         if (minsize > 0x100000000ULL && st->ss->major == 0) {
 303                                 fprintf(stderr, Name ": devices too large for RAID level %d\n", level);
 304                                 return 1;
 305                         }
 306                         size = minsize;
 307                         if (verbose > 0)
 308                                 fprintf(stderr, Name ": size set to %lluK\n", size);
 309                 }
 310         }
 311         if (level > 0 && ((maxsize-size)*100 > maxsize)) {
 312                 if (runstop != 1 || verbose >= 0)
 313                         fprintf(stderr, Name ": largest drive (%s) exceed size (%lluK) by more than 1%%\n",
 314                                 maxdisc, size);
 315                 warn = 1;
 316         }
 317
 318         if (warn) {
 319                 if (runstop!= 1) {
 320                         if (!ask("Continue creating array? ")) {
 321                                 fprintf(stderr, Name ": create aborted.\n");
 322                                 return 1;
 323                         }
 324                 } else {
 325                         if (verbose > 0)
 326                                 fprintf(stderr, Name ": creation continuing despite oddities due to --run\n");
 327                 }
 328         }
 329
 330         /* If this is raid4/5, we want to configure the last active slot
 331          * as missing, so that a reconstruct happens (faster than re-parity)
 332          * FIX: Can we do this for raid6 as well?
 333          */
 334         if (assume_clean==0 && force == 0 && first_missing >= raiddisks) {
 335                 switch ( level ) {
 336                 case 4:
 337                 case 5:
 338                         insert_point = raiddisks-1;
 339                         sparedisks++;
 340                         info.array.active_disks--;
 341                         missing_disks++;
 342                         break;
 343                 default:
 344                         break;
 345                 }
 346         }
 347         /* For raid6, if creating with 1 missing drive, make a good drive
 348          * into a spare, else the create will fail
 349          */
 350         if (assume_clean == 0 && force == 0 && first_missing < raiddisks &&
 351             second_missing >= raiddisks && level == 6) {
 352                 insert_point = raiddisks - 1;
 353                 if (insert_point == first_missing)
 354                         insert_point--;
 355                 sparedisks ++;
 356                 info.array.active_disks--;
 357                 missing_disks++;
 358         }
 359
 360         if (level <= 0 && first_missing != subdevs * 2) {
 361                 fprintf(stderr,
 362                         Name ": This level does not support missing devices\n");
 363                 return 1;
 364         }
 365
 366         /* Ok, lets try some ioctls */
 367
 368         info.array.level = level;
 369         info.array.size = size;
 370         info.array.raid_disks = raiddisks;
 371         /* The kernel should *know* what md_minor we are dealing
 372          * with, but it chooses to trust me instead. Sigh
 373          */
 374         info.array.md_minor = 0;
 375         if (fstat(mdfd, &stb)==0)
 376                 info.array.md_minor = minor(stb.st_rdev);
 377         info.array.not_persistent = 0;
 378
 379         if ( ( (level == 4 || level == 5) &&
 380                (insert_point < raiddisks || first_missing < raiddisks) )
 381              ||
 382              ( level == 6 && (insert_point < raiddisks
 383                               || second_missing < raiddisks))
 384              ||
 385              assume_clean
 386                 )
 387                 info.array.state = 1; /* clean, but one+ drive will be missing*/
 388         else
 389                 info.array.state = 0; /* not clean, but no errors */
 390
 391         if (level == 10) {
 392                 /* for raid10, the bitmap size is the capacity of the array,
 393                  * which is array.size * raid_disks / ncopies;
 394                  * .. but convert to sectors.
 395                  */
 396                 int ncopies = ((layout>>8) & 255) * (layout & 255);
 397                 bitmapsize = (unsigned long long)size * raiddisks / ncopies * 2;
 398 /*              printf("bms=%llu as=%d rd=%d nc=%d\n", bitmapsize, size, raiddisks, ncopies);*/
 399         } else
 400                 bitmapsize = (unsigned long long)size * 2;
 401
 402         /* There is lots of redundancy in these disk counts,
 403          * raid_disks is the most meaningful value
 404          *          it describes the geometry of the array
 405          *          it is constant
 406          * nr_disks is total number of used slots.
 407          *          it should be raid_disks+spare_disks
 408          * spare_disks is the number of extra disks present
 409          *          see above
 410          * active_disks is the number of working disks in
 411          *          active slots. (With raid_disks)
 412          * working_disks is the total number of working disks,
 413          *          including spares
 414          * failed_disks is the number of disks marked failed
 415          *
 416          * Ideally, the kernel would keep these (except raid_disks)
 417          * up-to-date as we ADD_NEW_DISK, but it doesn't (yet).
 418          * So for now, we assume that all raid and spare
 419          * devices will be given.
 420          */
 421         info.array.spare_disks=sparedisks;
 422         info.array.failed_disks=missing_disks;
 423         info.array.nr_disks = info.array.working_disks
 424                 + info.array.failed_disks;
 425         info.array.layout = layout;
 426         info.array.chunk_size = chunk*1024;
 427         info.array.major_version = st->ss->major;
 428
 429         if (name == NULL || *name == 0) {
 430                 /* base name on mddev */
 431                 /*  /dev/md0 -> 0
 432                  *  /dev/md_d0 -> d0
 433                  *  /dev/md/1 -> 1
 434                  *  /dev/md/d1 -> d1
 435                  *  /dev/md/home -> home
 436                  *  /dev/mdhome -> home
 437                  */
 438                 name = strrchr(mddev, '/');
 439                 if (name) {
 440                         name++;
 441                         if (strncmp(name, "md_d", 4)==0 &&
 442                             strlen(name) > 4 &&
 443                             isdigit(name[4]) &&
 444                             (name-mddev) == 5 /* /dev/ */)
 445                                 name += 3;
 446                         else if (strncmp(name, "md", 2)==0 &&
 447                                  strlen(name) > 2 &&
 448                                  isdigit(name[2]) &&
 449                                  (name-mddev) == 5 /* /dev/ */)
 450                                 name += 2;
 451                 }
 452         }
 453         if (!st->ss->init_super(st, &info.array, size, name, homehost, uuid))
 454                 return 1;
 455
 456         if (bitmap_file && vers < 9003) {
 457                 major_num = BITMAP_MAJOR_HOSTENDIAN;
 458 #ifdef __BIG_ENDIAN
 459                 fprintf(stderr, Name ": Warning - bitmaps created on this kernel are not portable\n"
 460                         "  between different architectured.  Consider upgrading the Linux kernel.\n");
 461 #endif
 462         }
 463
 464         if (bitmap_file && strcmp(bitmap_file, "internal")==0) {
 465                 if ((vers%100) < 2) {
 466                         fprintf(stderr, Name ": internal bitmaps not supported by this kernel.\n");
 467                         return 1;
 468                 }
 469                 if (!st->ss->add_internal_bitmap(st, &bitmap_chunk,
 470                                                  delay, write_behind,
 471                                                  bitmapsize, 1, major_num)) {
 472                         fprintf(stderr, Name ": Given bitmap chunk size not supported.\n");
 473                         return 1;
 474                 }
 475                 bitmap_file = NULL;
 476         }
 477
 478
 479
 480         if ((vers % 100) >= 1) { /* can use different versions */
 481                 mdu_array_info_t inf;
 482                 memset(&inf, 0, sizeof(inf));
 483                 inf.major_version = st->ss->major;
 484                 inf.minor_version = st->minor_version;
 485                 rv = ioctl(mdfd, SET_ARRAY_INFO, &inf);
 486         } else
 487                 rv = ioctl(mdfd, SET_ARRAY_INFO, NULL);
 488         if (rv) {
 489                 fprintf(stderr, Name ": SET_ARRAY_INFO failed for %s: %s\n",
 490                         mddev, strerror(errno));
 491                 return 1;
 492         }
 493
 494         if (bitmap_file) {
 495                 int uuid[4];
 496
 497                 st->ss->uuid_from_super(st, uuid);
 498                 if (CreateBitmap(bitmap_file, force, (char*)uuid, bitmap_chunk,
 499                                  delay, write_behind,
 500                                  bitmapsize,
 501                                  major_num)) {
 502                         return 1;
 503                 }
 504                 bitmap_fd = open(bitmap_file, O_RDWR);
 505                 if (bitmap_fd < 0) {
 506                         fprintf(stderr, Name ": weird: %s cannot be openned\n",
 507                                 bitmap_file);
 508                         return 1;
 509                 }
 510                 if (ioctl(mdfd, SET_BITMAP_FILE, bitmap_fd) < 0) {
 511                         fprintf(stderr, Name ": Cannot set bitmap file for %s: %s\n",
 512                                 mddev, strerror(errno));
 513                         return 1;
 514                 }
 515         }
 516
 517
 518
 519         for (pass=1; pass <=2 ; pass++) {
 520                 mddev_dev_t moved_disk = NULL; /* the disk that was moved out of the insert point */
 521
 522                 for (dnum=0, dv = devlist ; dv ;
 523                      dv=(dv->next)?(dv->next):moved_disk, dnum++) {
 524                         int fd;
 525                         struct stat stb;
 526
 527                         info.disk.number = dnum;
 528                         if (dnum == insert_point) {
 529                                 moved_disk = dv;
 530                         }
 531                         info.disk.raid_disk = info.disk.number;
 532                         if (info.disk.raid_disk < raiddisks)
 533                                 info.disk.state = (1<<MD_DISK_ACTIVE) |
 534                                                 (1<<MD_DISK_SYNC);
 535                         else
 536                                 info.disk.state = 0;
 537                         if (dv->writemostly == 1)
 538                                 info.disk.state |= (1<<MD_DISK_WRITEMOSTLY);
 539
 540                         if (dnum == insert_point ||
 541                             strcasecmp(dv->devname, "missing")==0) {
 542                                 info.disk.major = 0;
 543                                 info.disk.minor = 0;
 544                                 info.disk.state = (1<<MD_DISK_FAULTY);
 545                         } else {
 546                                 fd = open(dv->devname, O_RDONLY|O_EXCL);
 547                                 if (fd < 0) {
 548                                         fprintf(stderr, Name ": failed to open %s after earlier success - aborting\n",
 549                                                 dv->devname);
 550                                         return 1;
 551                                 }
 552                                 fstat(fd, &stb);
 553                                 info.disk.major = major(stb.st_rdev);
 554                                 info.disk.minor = minor(stb.st_rdev);
 555                                 remove_partitions(fd);
 556                                 close(fd);
 557                         }
 558                         switch(pass){
 559                         case 1:
 560                                 st->ss->add_to_super(st, &info.disk);
 561                                 break;
 562                         case 2:
 563                                 if (info.disk.state == 1) break;
 564                                 Kill(dv->devname, 0, 1); /* Just be sure it is clean */
 565                                 Kill(dv->devname, 0, 1); /* and again, there could be two superblocks */
 566                                 st->ss->write_init_super(st, &info.disk,
 567                                                          dv->devname);
 568
 569                                 if (ioctl(mdfd, ADD_NEW_DISK, &info.disk)) {
 570                                         fprintf(stderr, Name ": ADD_NEW_DISK for %s failed: %s\n",
 571                                                 dv->devname, strerror(errno));
 572                                         st->ss->free_super(st);
 573                                         return 1;
 574                                 }
 575
 576                                 break;
 577                         }
 578                         if (dv == moved_disk && dnum != insert_point) break;
 579                 }
 580         }
 581         st->ss->free_super(st);
 582
 583         /* param is not actually used */
 584         if (runstop == 1 || subdevs >= raiddisks) {
 585                 mdu_param_t param;
 586                 if (ioctl(mdfd, RUN_ARRAY, &param)) {
 587                         fprintf(stderr, Name ": RUN_ARRAY failed: %s\n",
 588                                 strerror(errno));
 589                         Manage_runstop(mddev, mdfd, -1, 0);
 590                         return 1;
 591                 }
 592                 if (verbose >= 0)
 593                         fprintf(stderr, Name ": array %s started.\n", mddev);
 594         } else {
 595                 fprintf(stderr, Name ": not starting array - not enough devices.\n");
 596         }
 597         return 0;
 598 }