Create.c

   1 /*
   2  * mdadm - manage Linux "md" devices aka RAID arrays.
   3  *
   4  * Copyright (C) 2001-2002 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        "md_u.h"
  32 #include        "md_p.h"
  33
  34 int Create(struct supertype *st, char *mddev, int mdfd,
  35            int chunk, int level, int layout, unsigned long size, int raiddisks, int sparedisks,
  36            char *name,
  37            int subdevs, mddev_dev_t devlist,
  38            int runstop, int verbose, int force,
  39            char *bitmap_file, int bitmap_chunk, int write_behind, int delay)
  40 {
  41         /*
  42          * Create a new raid array.
  43          *
  44          * First check that necessary details are available
  45          * (i.e. level, raid-disks)
  46          *
  47          * Then check each disk to see what might be on it
  48          * and report anything interesting.
  49          *
  50          * If anything looks odd, and runstop not set,
  51          * abort.
  52          *
  53          * SET_ARRAY_INFO and ADD_NEW_DISK, and
  54          * if runstop==run, or raiddisks disks were used,
  55          * RUN_ARRAY
  56          */
  57         unsigned long long minsize=0, maxsize=0;
  58         char *mindisc = NULL;
  59         char *maxdisc = NULL;
  60         int dnum;
  61         mddev_dev_t dv;
  62         int fail=0, warn=0;
  63         struct stat stb;
  64         int first_missing = subdevs * 2;
  65         int missing_disks = 0;
  66         int insert_point = subdevs * 2; /* where to insert a missing drive */
  67         void *super;
  68         int pass;
  69         int vers;
  70         int rv;
  71         int bitmap_fd;
  72
  73         mdu_array_info_t array;
  74         int major = BITMAP_MAJOR_HI;
  75
  76         vers = md_get_version(mdfd);
  77         if (vers < 9000) {
  78                 fprintf(stderr, Name ": Create requires md driver version 0.90.0 or later\n");
  79                 return 1;
  80         }
  81         if (level == UnSet) {
  82                 fprintf(stderr,
  83                         Name ": a RAID level is needed to create an array.\n");
  84                 return 1;
  85         }
  86         if (raiddisks < 1) {
  87                 fprintf(stderr,
  88                         Name ": a number of --raid-devices must be given to create an array\n");
  89                 return 1;
  90         }
  91         if (raiddisks < 4 && level == 6) {
  92                 fprintf(stderr,
  93                         Name ": at least 4 raid-devices needed for level 6\n");
  94                 return 1;
  95         }
  96         if (raiddisks > 256 && level == 6) {
  97                 fprintf(stderr,
  98                         Name ": no more than 256 raid-devices supported for level 6\n");
  99                 return 1;
 100         }
 101         if (raiddisks < 2 && level >= 4) {
 102                 fprintf(stderr,
 103                         Name ": at least 2 raid-devices needed for level 4 or 5\n");
 104                 return 1;
 105         }
 106         if (subdevs > raiddisks+sparedisks) {
 107                 fprintf(stderr, Name ": You have listed more devices (%d) than are in the array(%d)!\n", subdevs, raiddisks+sparedisks);
 108                 return 1;
 109         }
 110         if (subdevs < raiddisks+sparedisks) {
 111                 fprintf(stderr, Name ": You haven't given enough devices (real or missing) to create this array\n");
 112                 return 1;
 113         }
 114
 115         /* now set some defaults */
 116         if (layout == UnSet)
 117                 switch(level) {
 118                 default: /* no layout */
 119                         layout = 0;
 120                         break;
 121                 case 10:
 122                         layout = 0x102; /* near=2, far=1 */
 123                         if (verbose > 0)
 124                                 fprintf(stderr,
 125                                         Name ": layout defaults to n1\n");
 126                         break;
 127                 case 5:
 128                 case 6:
 129                         layout = map_name(r5layout, "default");
 130                         if (verbose > 0)
 131                                 fprintf(stderr,
 132                                         Name ": layout defaults to %s\n", map_num(r5layout, layout));
 133                         break;
 134                 case LEVEL_FAULTY:
 135                         layout = map_name(faultylayout, "default");
 136
 137                         if (verbose > 0)
 138                                 fprintf(stderr,
 139                                         Name ": layout defaults to %s\n", map_num(faultylayout, layout));
 140                         break;
 141                 }
 142
 143         if (level == 10)
 144                 /* check layout fits in array*/
 145                 if ((layout&255) * ((layout>>8)&255) > raiddisks) {
 146                         fprintf(stderr, Name ": that layout requires at least %d devices\n",
 147                                 (layout&255) * ((layout>>8)&255));
 148                         return 1;
 149                 }
 150
 151         switch(level) {
 152         case 4:
 153         case 5:
 154         case 10:
 155         case 6:
 156         case 0:
 157         case -1: /* linear */
 158                 if (chunk == 0) {
 159                         chunk = 64;
 160                         if (verbose > 0)
 161                                 fprintf(stderr, Name ": chunk size defaults to 64K\n");
 162                 }
 163                 break;
 164         default: /* raid1, multipath */
 165                 if (chunk) {
 166                         chunk = 0;
 167                         if (verbose > 0)
 168                                 fprintf(stderr, Name ": chunk size ignored for this level\n");
 169                 }
 170                 break;
 171         }
 172
 173         /* now look at the subdevs */
 174         array.active_disks = 0;
 175         array.working_disks = 0;
 176         dnum = 0;
 177         for (dv=devlist; dv; dv=dv->next, dnum++) {
 178                 char *dname = dv->devname;
 179                 unsigned long dsize;
 180                 unsigned long long ldsize, freesize;
 181                 int fd;
 182                 if (strcasecmp(dname, "missing")==0) {
 183                         if (first_missing > dnum)
 184                                 first_missing = dnum;
 185                         missing_disks ++;
 186                         continue;
 187                 }
 188                 array.working_disks++;
 189                 if (dnum < raiddisks)
 190                         array.active_disks++;
 191                 fd = open(dname, O_RDONLY|O_EXCL, 0);
 192                 if (fd <0 ) {
 193                         fprintf(stderr, Name ": Cannot open %s: %s\n",
 194                                 dname, strerror(errno));
 195                         fail=1;
 196                         continue;
 197                 }
 198 #ifdef BLKGETSIZE64
 199                 if (ioctl(fd, BLKGETSIZE64, &ldsize)==0)
 200                         ;
 201                 else
 202 #endif
 203                 if (ioctl(fd, BLKGETSIZE, &dsize)) {
 204                         fprintf(stderr, Name ": Cannot get size of %s: %s\n",
 205                                 dname, strerror(errno));
 206                         fail = 1;
 207                         close(fd);
 208                         continue;
 209                 }
 210                 else {
 211                         ldsize = dsize;
 212                         ldsize <<= 9;
 213                 }
 214                 freesize = st->ss->avail_size(st, ldsize >> 9);
 215                 if (freesize == 0) {
 216                         fprintf(stderr, Name ": %s is too small: %luK\n",
 217                                 dname, (unsigned long)(ldsize>>10));
 218                         fail = 1;
 219                         close(fd);
 220                         continue;
 221                 }
 222
 223                 freesize /= 2; /* convert to K */
 224
 225                 if (size && freesize < size) {
 226                         fprintf(stderr, Name ": %s is smaller that given size."
 227                                 " %lluK < %luK + superblock\n", dname, freesize, size);
 228                         fail = 1;
 229                         close(fd);
 230                         continue;
 231                 }
 232                 if (maxdisc == NULL || (maxdisc && freesize > maxsize)) {
 233                         maxdisc = dname;
 234                         maxsize = freesize;
 235                 }
 236                 if (mindisc ==NULL || (mindisc && freesize < minsize)) {
 237                         mindisc = dname;
 238                         minsize = freesize;
 239                 }
 240                 if (runstop != 1 || verbose >= 0) {
 241                         warn |= check_ext2(fd, dname);
 242                         warn |= check_reiser(fd, dname);
 243                         warn |= check_raid(fd, dname);
 244                 }
 245                 close(fd);
 246         }
 247         if (fail) {
 248                 fprintf(stderr, Name ": create aborted\n");
 249                 return 1;
 250         }
 251         if (size == 0) {
 252                 if (mindisc == NULL) {
 253                         fprintf(stderr, Name ": no size and no drives given - aborting create.\n");
 254                         return 1;
 255                 }
 256                 if (level > 0 || level == LEVEL_MULTIPATH || level == LEVEL_FAULTY) {
 257                         /* size is meaningful */
 258                         if (minsize > 0x100000000ULL) {
 259                                 fprintf(stderr, Name ": devices too large for RAID level %d\n", level);
 260                                 return 1;
 261                         }
 262                         size = minsize;
 263                         if (verbose > 0)
 264                                 fprintf(stderr, Name ": size set to %luK\n", size);
 265                 }
 266         }
 267         if (level > 0 && ((maxsize-size)*100 > maxsize)) {
 268                 if (runstop != 1 || verbose >= 0)
 269                         fprintf(stderr, Name ": largest drive (%s) exceed size (%luK) by more than 1%%\n",
 270                                 maxdisc, size);
 271                 warn = 1;
 272         }
 273
 274         if (warn) {
 275                 if (runstop!= 1) {
 276                         if (!ask("Continue creating array? ")) {
 277                                 fprintf(stderr, Name ": create aborted.\n");
 278                                 return 1;
 279                         }
 280                 } else {
 281                         if (verbose > 0)
 282                                 fprintf(stderr, Name ": creation continuing despite oddities due to --run\n");
 283                 }
 284         }
 285
 286         /* If this is  raid5, we want to configure the last active slot
 287          * as missing, so that a reconstruct happens (faster than re-parity)
 288          * FIX: Can we do this for raid6 as well?
 289          */
 290         if (force == 0 && first_missing >= raiddisks) {
 291                 switch ( level ) {
 292                 case 5:
 293                         insert_point = raiddisks-1;
 294                         sparedisks++;
 295                         array.active_disks--;
 296                         missing_disks++;
 297                         break;
 298                 default:
 299                         break;
 300                 }
 301         }
 302
 303         /* Ok, lets try some ioctls */
 304
 305         array.level = level;
 306         array.size = size;
 307         array.raid_disks = raiddisks;
 308         /* The kernel should *know* what md_minor we are dealing
 309          * with, but it chooses to trust me instead. Sigh
 310          */
 311         array.md_minor = 0;
 312         if (fstat(mdfd, &stb)==0)
 313                 array.md_minor = minor(stb.st_rdev);
 314         array.not_persistent = 0;
 315         /*** FIX: Need to do something about RAID-6 here ***/
 316         if ( ( (level == 5) &&
 317                (insert_point < raiddisks || first_missing < raiddisks) )
 318              ||
 319              ( level == 6 && missing_disks == 2)
 320                 )
 321                 array.state = 1; /* clean, but one+ drive will be missing */
 322         else
 323                 array.state = 0; /* not clean, but no errors */
 324
 325         /* There is lots of redundancy in these disk counts,
 326          * raid_disks is the most meaningful value
 327          *          it describes the geometry of the array
 328          *          it is constant
 329          * nr_disks is total number of used slots.
 330          *          it should be raid_disks+spare_disks
 331          * spare_disks is the number of extra disks present
 332          *          see above
 333          * active_disks is the number of working disks in
 334          *          active slots. (With raid_disks)
 335          * working_disks is the total number of working disks,
 336          *          including spares
 337          * failed_disks is the number of disks marked failed
 338          *
 339          * Ideally, the kernel would keep these (except raid_disks)
 340          * up-to-date as we ADD_NEW_DISK, but it doesn't (yet).
 341          * So for now, we assume that all raid and spare
 342          * devices will be given.
 343          */
 344         array.spare_disks=sparedisks;
 345         array.failed_disks=missing_disks;
 346         array.nr_disks = array.working_disks + array.failed_disks;
 347         array.layout = layout;
 348         array.chunk_size = chunk*1024;
 349
 350
 351         if (!st->ss->init_super(st, &super, &array, name))
 352                 return 1;
 353
 354         if (bitmap_file && vers < 9003) {
 355                 major = BITMAP_MAJOR_HOSTENDIAN;
 356 #ifdef __BIG_ENDIAN
 357                 fprintf(stderr, Name ": Warning - bitmaps created on this kernel are not portable\n"
 358                         "  between different architectured.  Consider upgrading the Linux kernel.\n");
 359 #endif
 360         }
 361
 362         if (bitmap_file && strcmp(bitmap_file, "internal")==0) {
 363                 if ((vers%100) < 2) {
 364                         fprintf(stderr, Name ": internal bitmaps not supported by this kernel.\n");
 365                         return 1;
 366                 }
 367                 if (!st->ss->add_internal_bitmap(st, super, bitmap_chunk, delay, write_behind,
 368                                                  &array.size, 1, major)) {
 369                         fprintf(stderr, Name ": Given bitmap chunk size not supported.\n");
 370                         return 1;
 371                 }
 372                 bitmap_file = NULL;
 373         }
 374
 375
 376
 377         if ((vers % 100) >= 1) { /* can use different versions */
 378                 mdu_array_info_t inf;
 379                 memset(&inf, 0, sizeof(inf));
 380                 inf.major_version = st->ss->major;
 381                 inf.minor_version = st->minor_version;
 382                 rv = ioctl(mdfd, SET_ARRAY_INFO, &inf);
 383         } else
 384                 rv = ioctl(mdfd, SET_ARRAY_INFO, NULL);
 385         if (rv) {
 386                 fprintf(stderr, Name ": SET_ARRAY_INFO failed for %s: %s\n",
 387                         mddev, strerror(errno));
 388                 return 1;
 389         }
 390
 391         if (bitmap_file) {
 392                 int uuid[4];
 393
 394                 if (bitmap_chunk == UnSet)
 395                         bitmap_chunk = DEFAULT_BITMAP_CHUNK;
 396
 397                 st->ss->uuid_from_super(uuid, super);
 398                 if (CreateBitmap(bitmap_file, force, (char*)uuid, bitmap_chunk,
 399                                  delay, write_behind,
 400                                  array.size*2ULL /* FIXME wrong for raid10 */,
 401                                  major)) {
 402                         return 1;
 403                 }
 404                 bitmap_fd = open(bitmap_file, O_RDWR);
 405                 if (bitmap_fd < 0) {
 406                         fprintf(stderr, Name ": weird: %s cannot be openned\n",
 407                                 bitmap_file);
 408                         return 1;
 409                 }
 410                 if (ioctl(mdfd, SET_BITMAP_FILE, bitmap_fd) < 0) {
 411                         fprintf(stderr, Name ": Cannot set bitmap file for %s: %s\n",
 412                                 mddev, strerror(errno));
 413                         return 1;
 414                 }
 415         }
 416
 417
 418
 419         for (pass=1; pass <=2 ; pass++) {
 420                 mddev_dev_t moved_disk = NULL; /* the disk that was moved out of the insert point */
 421
 422                 for (dnum=0, dv = devlist ; dv ;
 423                      dv=(dv->next)?(dv->next):moved_disk, dnum++) {
 424                         int fd;
 425                         struct stat stb;
 426                         mdu_disk_info_t disk;
 427
 428                         disk.number = dnum;
 429                         if (dnum == insert_point) {
 430                                 moved_disk = dv;
 431                         }
 432                         disk.raid_disk = disk.number;
 433                         if (disk.raid_disk < raiddisks)
 434                                 disk.state = (1<<MD_DISK_ACTIVE) |
 435                                                 (1<<MD_DISK_SYNC);
 436                         else
 437                                 disk.state = 0;
 438                         if (dv->writemostly)
 439                                 disk.state |= (1<<MD_DISK_WRITEMOSTLY);
 440
 441                         if (dnum == insert_point ||
 442                             strcasecmp(dv->devname, "missing")==0) {
 443                                 disk.major = 0;
 444                                 disk.minor = 0;
 445                                 disk.state = (1<<MD_DISK_FAULTY);
 446                         } else {
 447                                 fd = open(dv->devname, O_RDONLY|O_EXCL, 0);
 448                                 if (fd < 0) {
 449                                         fprintf(stderr, Name ": failed to open %s after earlier success - aborting\n",
 450                                                 dv->devname);
 451                                         return 1;
 452                                 }
 453                                 fstat(fd, &stb);
 454                                 disk.major = major(stb.st_rdev);
 455                                 disk.minor = minor(stb.st_rdev);
 456                                 close(fd);
 457                         }
 458                         switch(pass){
 459                         case 1:
 460                                 st->ss->add_to_super(super, &disk);
 461                                 break;
 462                         case 2:
 463                                 if (disk.state == 1) break;
 464                                 st->ss->write_init_super(st, super, &disk, dv->devname);
 465
 466                                 if (ioctl(mdfd, ADD_NEW_DISK, &disk)) {
 467                                         fprintf(stderr, Name ": ADD_NEW_DISK for %s failed: %s\n",
 468                                                 dv->devname, strerror(errno));
 469                                         free(super);
 470                                         return 1;
 471                                 }
 472
 473                                 break;
 474                         }
 475                         if (dv == moved_disk && dnum != insert_point) break;
 476                 }
 477         }
 478         free(super);
 479
 480         /* param is not actually used */
 481         if (runstop == 1 || subdevs >= raiddisks) {
 482                 mdu_param_t param;
 483                 if (ioctl(mdfd, RUN_ARRAY, &param)) {
 484                         fprintf(stderr, Name ": RUN_ARRAY failed: %s\n",
 485                                 strerror(errno));
 486                         Manage_runstop(mddev, mdfd, -1, 0);
 487                         return 1;
 488                 }
 489                 if (verbose >= 0)
 490                         fprintf(stderr, Name ": array %s started.\n", mddev);
 491         } else {
 492                 fprintf(stderr, Name ": not starting array - not enough devices.\n");
 493         }
 494         return 0;
 495 }