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 superswitch *ss, char *mddev, int mdfd,
  35            int chunk, int level, int layout, unsigned long size, int raiddisks, int sparedisks,
  36            int subdevs, mddev_dev_t devlist,
  37            int runstop, int verbose, int force)
  38 {
  39         /*
  40          * Create a new raid array.
  41          *
  42          * First check that necessary details are available
  43          * (i.e. level, raid-disks)
  44          *
  45          * Then check each disk to see what might be on it
  46          * and report anything interesting.
  47          *
  48          * If anything looks odd, and runstop not set,
  49          * abort.
  50          *
  51          * SET_ARRAY_INFO and ADD_NEW_DISK, and
  52          * if runstop==run, or raiddisks diskswere used,
  53          * RUN_ARRAY
  54          */
  55         unsigned long long minsize=0, maxsize=0;
  56         char *mindisc = NULL;
  57         char *maxdisc = NULL;
  58         int dnum;
  59         mddev_dev_t dv;
  60         int fail=0, warn=0;
  61         struct stat stb;
  62         int first_missing = MD_SB_DISKS*2;
  63         int missing_disks = 0;
  64         int insert_point = MD_SB_DISKS*2; /* where to insert a missing drive */
  65         void *super;
  66         int pass;
  67
  68         mdu_array_info_t array;
  69
  70
  71         if (md_get_version(mdfd) < 9000) {
  72                 fprintf(stderr, Name ": Create requires md driver version 0.90.0 or later\n");
  73                 return 1;
  74         }
  75         if (level == UnSet) {
  76                 fprintf(stderr,
  77                         Name ": a RAID level is needed to create an array.\n");
  78                 return 1;
  79         }
  80         if (raiddisks < 1) {
  81                 fprintf(stderr,
  82                         Name ": a number of --raid-devices must be given to create an array\n");
  83                 return 1;
  84         }
  85         if (raiddisks < 4 && level == 6) {
  86                 fprintf(stderr,
  87                         Name ": at least 4 raid-devices needed for level 6\n");
  88                 return 1;
  89         }
  90         if (raiddisks > 256 && level == 6) {
  91                 fprintf(stderr,
  92                         Name ": no more than 256 raid-devices supported for level 6\n");
  93                 return 1;
  94         }
  95         if (raiddisks < 2 && level >= 4) {
  96                 fprintf(stderr,
  97                         Name ": at least 2 raid-devices needed for level 4 or 5\n");
  98                 return 1;
  99         }
 100         if (raiddisks+sparedisks > MD_SB_DISKS) {
 101                 fprintf(stderr,
 102                         Name ": too many devices requested: %d+%d > %d\n",
 103                         raiddisks, sparedisks, MD_SB_DISKS);
 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)
 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)
 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)
 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)
 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)
 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                         dsize <<= 9;
 213                 }
 214                 if (ldsize < MD_RESERVED_SECTORS*2LL*512LL) {
 215                         fprintf(stderr, Name ": %s is too small: %luK\n",
 216                                 dname, (unsigned long)(ldsize>>10));
 217                         fail = 1;
 218                         close(fd);
 219                         continue;
 220                 }
 221                 freesize = MD_NEW_SIZE_SECTORS((ldsize>>9));
 222                 freesize /= 2;
 223
 224                 if (size && freesize < size) {
 225                         fprintf(stderr, Name ": %s is smaller that given size."
 226                                 " %lluK < %luK + superblock\n", dname, freesize, size);
 227                         fail = 1;
 228                         close(fd);
 229                         continue;
 230                 }
 231                 if (maxdisc == NULL || (maxdisc && freesize > maxsize)) {
 232                         maxdisc = dname;
 233                         maxsize = freesize;
 234                 }
 235                 if (mindisc ==NULL || (mindisc && freesize < minsize)) {
 236                         mindisc = dname;
 237                         minsize = freesize;
 238                 }
 239                 warn |= check_ext2(fd, dname);
 240                 warn |= check_reiser(fd, dname);
 241                 warn |= check_raid(fd, dname);
 242                 close(fd);
 243         }
 244         if (fail) {
 245                 fprintf(stderr, Name ": create aborted\n");
 246                 return 1;
 247         }
 248         if (size == 0) {
 249                 if (mindisc == NULL) {
 250                         fprintf(stderr, Name ": no size and no drives given - aborting create.\n");
 251                         return 1;
 252                 }
 253                 if (level > 0) {
 254                         /* size is meaningful */
 255                         if (minsize > 0x100000000ULL) {
 256                                 fprintf(stderr, Name ": devices too large for RAID level %d\n", level);
 257                                 return 1;
 258                         }
 259                         size = minsize;
 260                         if (verbose)
 261                                 fprintf(stderr, Name ": size set to %luK\n", size);
 262                 }
 263         }
 264         if (level > 0 && ((maxsize-size)*100 > maxsize)) {
 265                 fprintf(stderr, Name ": largest drive (%s) exceed size (%luK) by more than 1%%\n",
 266                         maxdisc, size);
 267                 warn = 1;
 268         }
 269
 270         if (warn) {
 271                 if (runstop!= 1) {
 272                         if (!ask("Continue creating array? ")) {
 273                                 fprintf(stderr, Name ": create aborted.\n");
 274                                 return 1;
 275                         }
 276                 } else {
 277                         if (verbose)
 278                                 fprintf(stderr, Name ": creation continuing despite oddities due to --run\n");
 279                 }
 280         }
 281
 282         /* If this is  raid5, we want to configure the last active slot
 283          * as missing, so that a reconstruct happens (faster than re-parity)
 284          * FIX: Can we do this for raid6 as well?
 285          */
 286         if (force == 0 && first_missing >= raiddisks) {
 287                 switch ( level ) {
 288                 case 5:
 289                         insert_point = raiddisks-1;
 290                         sparedisks++;
 291                         array.active_disks--;
 292                         missing_disks++;
 293                         break;
 294                 default:
 295                         break;
 296                 }
 297         }
 298
 299         /* Ok, lets try some ioctls */
 300
 301         array.level = level;
 302         array.size = size;
 303         array.raid_disks = raiddisks;
 304         /* The kernel should *know* what md_minor we are dealing
 305          * with, but it chooses to trust me instead. Sigh
 306          */
 307         array.md_minor = 0;
 308         if (fstat(mdfd, &stb)==0)
 309                 array.md_minor = minor(stb.st_rdev);
 310         array.not_persistent = 0;
 311         /*** FIX: Need to do something about RAID-6 here ***/
 312         if ( ( (level == 5) &&
 313                (insert_point < raiddisks || first_missing < raiddisks) )
 314              ||
 315              ( level == 6 && missing_disks == 2)
 316                 )
 317                 array.state = 1; /* clean, but one+ drive will be missing */
 318         else
 319                 array.state = 0; /* not clean, but no errors */
 320
 321         /* There is lots of redundancy in these disk counts,
 322          * raid_disks is the most meaningful value
 323          *          it describes the geometry of the array
 324          *          it is constant
 325          * nr_disks is total number of used slots.
 326          *          it should be raid_disks+spare_disks
 327          * spare_disks is the number of extra disks present
 328          *          see above
 329          * active_disks is the number of working disks in
 330          *          active slots. (With raid_disks)
 331          * working_disks is the total number of working disks,
 332          *          including spares
 333          * failed_disks is the number of disks marked failed
 334          *
 335          * Ideally, the kernel would keep these (except raid_disks)
 336          * up-to-date as we ADD_NEW_DISK, but it doesn't (yet).
 337          * So for now, we assume that all raid and spare
 338          * devices will be given.
 339          */
 340         array.spare_disks=sparedisks;
 341         array.failed_disks=missing_disks;
 342         array.nr_disks = array.working_disks + array.failed_disks;
 343         array.layout = layout;
 344         array.chunk_size = chunk*1024;
 345
 346         if (ioctl(mdfd, SET_ARRAY_INFO, NULL)) {
 347                 fprintf(stderr, Name ": SET_ARRAY_INFO failed for %s: %s\n",
 348                         mddev, strerror(errno));
 349                 return 1;
 350         }
 351
 352         ss->init_super(&super, &array);
 353
 354
 355         for (pass=1; pass <=2 ; pass++) {
 356                 mddev_dev_t moved_disk = NULL; /* the disk that was moved out of the insert point */
 357
 358                 for (dnum=0, dv = devlist ; dv ;
 359                      dv=(dv->next)?(dv->next):moved_disk, dnum++) {
 360                         int fd;
 361                         struct stat stb;
 362                         mdu_disk_info_t disk;
 363
 364                         disk.number = dnum;
 365                         if (dnum == insert_point) {
 366                                 moved_disk = dv;
 367                         }
 368                         disk.raid_disk = disk.number;
 369                         if (disk.raid_disk < raiddisks)
 370                                 disk.state = 6; /* active and in sync */
 371                         else
 372                                 disk.state = 0;
 373                         if (dnum == insert_point ||
 374                             strcasecmp(dv->devname, "missing")==0) {
 375                                 disk.major = 0;
 376                                 disk.minor = 0;
 377                                 disk.state = 1; /* faulty */
 378                         } else {
 379                                 fd = open(dv->devname, O_RDONLY|O_EXCL, 0);
 380                                 if (fd < 0) {
 381                                         fprintf(stderr, Name ": failed to open %s after earlier success - aborting\n",
 382                                                 dv->devname);
 383                                         return 1;
 384                                 }
 385                                 fstat(fd, &stb);
 386                                 disk.major = major(stb.st_rdev);
 387                                 disk.minor = minor(stb.st_rdev);
 388                                 close(fd);
 389                         }
 390                         switch(pass){
 391                         case 1:
 392                                 ss->add_to_super(super, &disk);
 393                                 break;
 394                         case 2:
 395                                 ss->write_init_super(super, &disk, dv->devname);
 396
 397                                 if (ioctl(mdfd, ADD_NEW_DISK, &disk)) {
 398                                         fprintf(stderr, Name ": ADD_NEW_DISK for %s failed: %s\n",
 399                                                 dv->devname, strerror(errno));
 400                                         free(super);
 401                                         return 1;
 402                                 }
 403
 404                                 break;
 405                         }
 406                         if (dv == moved_disk && dnum != insert_point) break;
 407                 }
 408         }
 409         free(super);
 410
 411         /* param is not actually used */
 412         if (runstop == 1 || subdevs >= raiddisks) {
 413                 mdu_param_t param;
 414                 if (ioctl(mdfd, RUN_ARRAY, &param)) {
 415                         fprintf(stderr, Name ": RUN_ARRAY failed: %s\n",
 416                                 strerror(errno));
 417                         Manage_runstop(mddev, mdfd, -1);
 418                         return 1;
 419                 }
 420                 fprintf(stderr, Name ": array %s started.\n", mddev);
 421         } else {
 422                 fprintf(stderr, Name ": not starting array - not enough devices.\n");
 423         }
 424         return 0;
 425 }