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