raid6check.c

   1 /*
   2  * raid6check - extended consistency check for RAID-6
   3  *
   4  * Copyright (C) 2011 Piergiorgio Sartor
   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: Piergiorgio Sartor
  22  *    Based on "restripe.c" from "mdadm" codebase
  23  */
  24
  25 #include "mdadm.h"
  26 #include <stdint.h>
  27 #include <sys/mman.h>
  28
  29 #define CHECK_PAGE_BITS (12)
  30 #define CHECK_PAGE_SIZE (1 << CHECK_PAGE_BITS)
  31
  32 char const Name[] = "raid6check";
  33
  34 enum repair {
  35         NO_REPAIR = 0,
  36         MANUAL_REPAIR,
  37         AUTO_REPAIR
  38 };
  39
  40 int geo_map(int block, unsigned long long stripe, int raid_disks,
  41             int level, int layout);
  42 int is_ddf(int layout);
  43 void qsyndrome(uint8_t *p, uint8_t *q, uint8_t **sources, int disks, int size);
  44 void make_tables(void);
  45 void ensure_zero_has_size(int chunk_size);
  46 void raid6_datap_recov(int disks, size_t bytes, int faila, uint8_t **ptrs,
  47                        int neg_offset);
  48 void raid6_2data_recov(int disks, size_t bytes, int faila, int failb,
  49                        uint8_t **ptrs, int neg_offset);
  50 void xor_blocks(char *target, char **sources, int disks, int size);
  51
  52 /* Collect per stripe consistency information */
  53 void raid6_collect(int chunk_size, uint8_t *p, uint8_t *q,
  54                    char *chunkP, char *chunkQ, int *results)
  55 {
  56         int i;
  57         int data_id;
  58         uint8_t Px, Qx;
  59         extern uint8_t raid6_gflog[];
  60
  61         for(i = 0; i < chunk_size; i++) {
  62                 Px = (uint8_t)chunkP[i] ^ (uint8_t)p[i];
  63                 Qx = (uint8_t)chunkQ[i] ^ (uint8_t)q[i];
  64
  65                 if((Px != 0) && (Qx == 0))
  66                         results[i] = -1;
  67
  68                 if((Px == 0) && (Qx != 0))
  69                         results[i] = -2;
  70
  71                 if((Px != 0) && (Qx != 0)) {
  72                         data_id = (raid6_gflog[Qx] - raid6_gflog[Px]);
  73                         if(data_id < 0) data_id += 255;
  74                         results[i] = data_id;
  75                 }
  76
  77                 if((Px == 0) && (Qx == 0))
  78                         results[i] = -255;
  79         }
  80 }
  81
  82 /* Try to find out if a specific disk has problems in a CHECK_PAGE_SIZE page size */
  83 int raid6_stats_blk(int *results, int raid_disks)
  84 {
  85         int i;
  86         int curr_broken_disk = -255;
  87         int prev_broken_disk = -255;
  88         int broken_status = 0;
  89
  90         for(i = 0; i < CHECK_PAGE_SIZE; i++) {
  91
  92                 if(results[i] != -255)
  93                         curr_broken_disk = results[i];
  94
  95                 if(curr_broken_disk >= raid_disks)
  96                         broken_status = 2;
  97
  98                 switch(broken_status) {
  99                 case 0:
 100                         if(curr_broken_disk != -255) {
 101                                 prev_broken_disk = curr_broken_disk;
 102                                 broken_status = 1;
 103                         }
 104                         break;
 105
 106                 case 1:
 107                         if(curr_broken_disk != prev_broken_disk)
 108                                 broken_status = 2;
 109                         break;
 110
 111                 case 2:
 112                 default:
 113                         curr_broken_disk = prev_broken_disk = -65535;
 114                         break;
 115                 }
 116         }
 117
 118         return curr_broken_disk;
 119 }
 120
 121 /* Collect disks status for a strip in CHECK_PAGE_SIZE page size blocks */
 122 void raid6_stats(int *disk, int *results, int raid_disks, int chunk_size)
 123 {
 124         int i, j;
 125
 126         for(i = 0, j = 0; i < chunk_size; i += CHECK_PAGE_SIZE, j++) {
 127                 disk[j] = raid6_stats_blk(&results[i], raid_disks);
 128         }
 129 }
 130
 131 int lock_stripe(struct mdinfo *info, unsigned long long start,
 132                 int chunk_size, int data_disks, sighandler_t *sig)
 133 {
 134         int rv;
 135
 136         sig[0] = signal_s(SIGTERM, SIG_IGN);
 137         sig[1] = signal_s(SIGINT, SIG_IGN);
 138         sig[2] = signal_s(SIGQUIT, SIG_IGN);
 139
 140         if (sig[0] == SIG_ERR || sig[1] == SIG_ERR || sig[2] == SIG_ERR)
 141                 return 1;
 142
 143         if(mlockall(MCL_CURRENT | MCL_FUTURE) != 0) {
 144                 return 2;
 145         }
 146
 147         rv = sysfs_set_num(info, NULL, "suspend_lo", start * chunk_size * data_disks);
 148         rv |= sysfs_set_num(info, NULL, "suspend_hi", (start + 1) * chunk_size * data_disks);
 149         return rv * 256;
 150 }
 151
 152 int unlock_all_stripes(struct mdinfo *info, sighandler_t *sig)
 153 {
 154         int rv;
 155         rv = sysfs_set_num(info, NULL, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL);
 156         rv |= sysfs_set_num(info, NULL, "suspend_hi", 0);
 157         rv |= sysfs_set_num(info, NULL, "suspend_lo", 0);
 158
 159         signal_s(SIGQUIT, sig[2]);
 160         signal_s(SIGINT, sig[1]);
 161         signal_s(SIGTERM, sig[0]);
 162
 163         if(munlockall() != 0)
 164                 return 3;
 165         return rv * 256;
 166 }
 167
 168 /* Autorepair */
 169 int autorepair(int *disk, unsigned long long start, int chunk_size,
 170                 char *name[], int raid_disks, int syndrome_disks, char **blocks_page,
 171                 char **blocks, uint8_t *p, int *block_index_for_slot,
 172                 int *source, unsigned long long *offsets)
 173 {
 174         int i, j;
 175         int pages_to_write_count = 0;
 176         int page_to_write[chunk_size >> CHECK_PAGE_BITS];
 177         for(j = 0; j < (chunk_size >> CHECK_PAGE_BITS); j++) {
 178                 if (disk[j] >= -2 && block_index_for_slot[disk[j]] >= 0) {
 179                         int slot = block_index_for_slot[disk[j]];
 180                         printf("Auto-repairing slot %d (%s)\n", slot, name[slot]);
 181                         pages_to_write_count++;
 182                         page_to_write[j] = 1;
 183                         for(i = -2; i < syndrome_disks; i++) {
 184                                 blocks_page[i] = blocks[i] + j * CHECK_PAGE_SIZE;
 185                         }
 186                         if (disk[j] == -2) {
 187                                 qsyndrome(p, (uint8_t*)blocks_page[-2],
 188                                           (uint8_t**)blocks_page,
 189                                           syndrome_disks, CHECK_PAGE_SIZE);
 190                         }
 191                         else {
 192                                 char *all_but_failed_blocks[syndrome_disks];
 193                                 for(i = 0; i < syndrome_disks; i++) {
 194                                         if (i == disk[j])
 195                                                 all_but_failed_blocks[i] = blocks_page[-1];
 196                                         else
 197                                                 all_but_failed_blocks[i] = blocks_page[i];
 198                                 }
 199                                 xor_blocks(blocks_page[disk[j]],
 200                                            all_but_failed_blocks, syndrome_disks,
 201                                            CHECK_PAGE_SIZE);
 202                         }
 203                 }
 204                 else {
 205                         page_to_write[j] = 0;
 206                 }
 207         }
 208
 209         if(pages_to_write_count > 0) {
 210                 int write_res = 0;
 211                 for(j = 0; j < (chunk_size >> CHECK_PAGE_BITS); j++) {
 212                         if(page_to_write[j] == 1) {
 213                                 int slot = block_index_for_slot[disk[j]];
 214                                 lseek64(source[slot], offsets[slot] + start * chunk_size + j * CHECK_PAGE_SIZE, SEEK_SET);
 215                                 write_res += write(source[slot],
 216                                                    blocks[disk[j]] + j * CHECK_PAGE_SIZE,
 217                                                    CHECK_PAGE_SIZE);
 218                         }
 219                 }
 220
 221                 if (write_res != (CHECK_PAGE_SIZE * pages_to_write_count)) {
 222                         fprintf(stderr, "Failed to write a full chunk.\n");
 223                         return -1;
 224                 }
 225         }
 226
 227         return 0;
 228 }
 229
 230 /* Manual repair */
 231 int manual_repair(int chunk_size, int syndrome_disks,
 232                   int failed_slot1, int failed_slot2,
 233                   unsigned long long start, int *block_index_for_slot,
 234                   char *name[], char **stripes, char **blocks, uint8_t *p,
 235                   int *source, unsigned long long *offsets)
 236 {
 237         int i;
 238         int fd1 = block_index_for_slot[failed_slot1];
 239         int fd2 = block_index_for_slot[failed_slot2];
 240         printf("Repairing stripe %llu\n", start);
 241         printf("Assuming slots %d (%s) and %d (%s) are incorrect\n",
 242                fd1, name[fd1],
 243                fd2, name[fd2]);
 244
 245         if (failed_slot1 == -2 || failed_slot2 == -2) {
 246                 char *all_but_failed_blocks[syndrome_disks];
 247                 int failed_data_or_p;
 248
 249                 if (failed_slot1 == -2)
 250                         failed_data_or_p = failed_slot2;
 251                 else
 252                         failed_data_or_p = failed_slot1;
 253
 254                 printf("Repairing D/P(%d) and Q\n", failed_data_or_p);
 255
 256                 for (i = 0; i < syndrome_disks; i++) {
 257                         if (i == failed_data_or_p)
 258                                 all_but_failed_blocks[i] = blocks[-1];
 259                         else
 260                                 all_but_failed_blocks[i] = blocks[i];
 261                 }
 262                 xor_blocks(blocks[failed_data_or_p],
 263                            all_but_failed_blocks, syndrome_disks, chunk_size);
 264                 qsyndrome(p, (uint8_t*)blocks[-2], (uint8_t**)blocks,
 265                           syndrome_disks, chunk_size);
 266         } else {
 267                 ensure_zero_has_size(chunk_size);
 268                 if (failed_slot1 == -1 || failed_slot2 == -1) {
 269                         int failed_data;
 270                         if (failed_slot1 == -1)
 271                                 failed_data = failed_slot2;
 272                         else
 273                                 failed_data = failed_slot1;
 274
 275                         printf("Repairing D(%d) and P\n", failed_data);
 276                         raid6_datap_recov(syndrome_disks+2, chunk_size,
 277                                           failed_data, (uint8_t**)blocks, 1);
 278                 } else {
 279                         printf("Repairing D and D\n");
 280                         raid6_2data_recov(syndrome_disks+2, chunk_size,
 281                                           failed_slot1, failed_slot2,
 282                                           (uint8_t**)blocks, 1);
 283                 }
 284         }
 285
 286         int write_res1, write_res2;
 287         off64_t seek_res;
 288
 289         seek_res = lseek64(source[fd1],
 290                            offsets[fd1] + start * chunk_size, SEEK_SET);
 291         if (seek_res < 0) {
 292                 fprintf(stderr, "lseek failed for failed_disk1\n");
 293                 return -1;
 294         }
 295         write_res1 = write(source[fd1], blocks[failed_slot1], chunk_size);
 296
 297         seek_res = lseek64(source[fd2],
 298                            offsets[fd2] + start * chunk_size, SEEK_SET);
 299         if (seek_res < 0) {
 300                 fprintf(stderr, "lseek failed for failed_disk2\n");
 301                 return -1;
 302         }
 303         write_res2 = write(source[fd2], blocks[failed_slot2], chunk_size);
 304
 305         if (write_res1 != chunk_size || write_res2 != chunk_size) {
 306                 fprintf(stderr, "Failed to write a complete chunk.\n");
 307                 return -2;
 308         }
 309
 310         return 0;
 311 }
 312
 313 int check_stripes(struct mdinfo *info, int *source, unsigned long long *offsets,
 314                   int raid_disks, int chunk_size, int level, int layout,
 315                   unsigned long long start, unsigned long long length, char *name[],
 316                   enum repair repair, int failed_disk1, int failed_disk2)
 317 {
 318         /* read the data and p and q blocks, and check we got them right */
 319         int data_disks = raid_disks - 2;
 320         int syndrome_disks = data_disks + is_ddf(layout) * 2;
 321         char *stripe_buf;
 322
 323         /* stripes[] is indexed by raid_disk and holds chunks from each device */
 324         char **stripes = xmalloc(raid_disks * sizeof(char*));
 325
 326         /* blocks[] is indexed by syndrome number and points to either one of the
 327          * chunks from 'stripes[]', or to a chunk of zeros. -1 and -2 are
 328          * P and Q */
 329         char **blocks = xmalloc((syndrome_disks + 2) * sizeof(char*));
 330
 331         /* blocks_page[] is a temporary index to just one page of the chunks
 332          * that blocks[] points to. */
 333         char **blocks_page = xmalloc((syndrome_disks + 2) * sizeof(char*));
 334
 335         /* block_index_for_slot[] provides the reverse mapping from blocks to stripes.
 336          * The index is a syndrome position, the content is a raid_disk number.
 337          * indicies -1 and -2 work, and are P and Q disks */
 338         int *block_index_for_slot = xmalloc((syndrome_disks+2) * sizeof(int));
 339
 340         /* 'p' and 'q' contain calcualted P and Q, to be compared with
 341          * blocks[-1] and blocks[-2];
 342          */
 343         uint8_t *p = xmalloc(chunk_size);
 344         uint8_t *q = xmalloc(chunk_size);
 345         char *zero = xmalloc(chunk_size);
 346         int *results = xmalloc(chunk_size * sizeof(int));
 347         sighandler_t *sig = xmalloc(3 * sizeof(sighandler_t));
 348
 349         int i, j;
 350         int diskP, diskQ, diskD;
 351         int err = 0;
 352
 353         extern int tables_ready;
 354
 355         if (!tables_ready)
 356                 make_tables();
 357
 358         if (posix_memalign((void**)&stripe_buf, 4096, raid_disks * chunk_size) != 0)
 359                 exit(4);
 360         block_index_for_slot += 2;
 361         blocks += 2;
 362         blocks_page += 2;
 363
 364         memset(zero, 0, chunk_size);
 365         for ( i = 0 ; i < raid_disks ; i++)
 366                 stripes[i] = stripe_buf + i * chunk_size;
 367
 368         while (length > 0) {
 369                 /* The syndrome number of the broken disk is recorded
 370                  * in 'disk[]' which allows a different broken disk for
 371                  * each page.
 372                  */
 373                 int disk[chunk_size >> CHECK_PAGE_BITS];
 374
 375                 err = lock_stripe(info, start, chunk_size, data_disks, sig);
 376                 if(err != 0) {
 377                         if (err != 2)
 378                                 unlock_all_stripes(info, sig);
 379                         goto exitCheck;
 380                 }
 381                 for (i = 0 ; i < raid_disks ; i++) {
 382                         off64_t seek_res = lseek64(source[i], offsets[i] + start * chunk_size,
 383                                                    SEEK_SET);
 384                         if (seek_res < 0) {
 385                                 fprintf(stderr, "lseek to source %d failed\n", i);
 386                                 unlock_all_stripes(info, sig);
 387                                 err = -1;
 388                                 goto exitCheck;
 389                         }
 390                         int read_res = read(source[i], stripes[i], chunk_size);
 391                         if (read_res < chunk_size) {
 392                                 fprintf(stderr, "Failed to read complete chunk disk %d, aborting\n", i);
 393                                 unlock_all_stripes(info, sig);
 394                                 err = -1;
 395                                 goto exitCheck;
 396                         }
 397                 }
 398
 399                 diskP = geo_map(-1, start, raid_disks, level, layout);
 400                 block_index_for_slot[-1] = diskP;
 401                 blocks[-1] = stripes[diskP];
 402
 403                 diskQ = geo_map(-2, start, raid_disks, level, layout);
 404                 block_index_for_slot[-2] = diskQ;
 405                 blocks[-2] = stripes[diskQ];
 406
 407                 if (!is_ddf(layout)) {
 408                         /* The syndrome-order of disks starts immediately after 'Q',
 409                          * but skips P */
 410                         diskD = diskQ;
 411                         for (i = 0 ; i < data_disks ; i++) {
 412                                 diskD = diskD + 1;
 413                                 if (diskD >= raid_disks)
 414                                         diskD = 0;
 415                                 if (diskD == diskP)
 416                                         diskD += 1;
 417                                 if (diskD >= raid_disks)
 418                                         diskD = 0;
 419                                 blocks[i] = stripes[diskD];
 420                                 block_index_for_slot[i] = diskD;
 421                         }
 422                 } else {
 423                         /* The syndrome-order exactly follows raid-disk
 424                          * numbers, with ZERO in place of P and Q
 425                          */
 426                         for (i = 0 ; i < raid_disks; i++) {
 427                                 if (i == diskP || i == diskQ) {
 428                                         blocks[i] = zero;
 429                                         block_index_for_slot[i] = -1;
 430                                 } else {
 431                                         blocks[i] = stripes[i];
 432                                         block_index_for_slot[i] = i;
 433                                 }
 434                         }
 435                 }
 436
 437                 qsyndrome(p, q, (uint8_t**)blocks, syndrome_disks, chunk_size);
 438
 439                 raid6_collect(chunk_size, p, q, stripes[diskP], stripes[diskQ], results);
 440                 raid6_stats(disk, results, raid_disks, chunk_size);
 441
 442                 for(j = 0; j < (chunk_size >> CHECK_PAGE_BITS); j++) {
 443                         int role = disk[j];
 444                         if (role >= -2) {
 445                                 int slot = block_index_for_slot[role];
 446                                 if (slot >= 0)
 447                                         printf("Error detected at stripe %llu, page %d: possible failed disk slot %d: %d --> %s\n",
 448                                                start, j, role, slot, name[slot]);
 449                                 else
 450                                         printf("Error detected at stripe %llu, page %d: failed slot %d should be zeros\n",
 451                                                start, j, role);
 452                         } else if(disk[j] == -65535) {
 453                                 printf("Error detected at stripe %llu, page %d: disk slot unknown\n", start, j);
 454                         }
 455                 }
 456
 457                 if(repair == AUTO_REPAIR) {
 458                         err = autorepair(disk, start, chunk_size,
 459                                         name, raid_disks, syndrome_disks, blocks_page,
 460                                         blocks, p, block_index_for_slot,
 461                                         source, offsets);
 462                         if(err != 0) {
 463                                 unlock_all_stripes(info, sig);
 464                                 goto exitCheck;
 465                         }
 466                 }
 467
 468                 if(repair == MANUAL_REPAIR) {
 469                         int failed_slot1 = -1, failed_slot2 = -1;
 470                         for (i = -2; i < syndrome_disks; i++) {
 471                                 if (block_index_for_slot[i] == failed_disk1)
 472                                         failed_slot1 = i;
 473                                 if (block_index_for_slot[i] == failed_disk2)
 474                                         failed_slot2 = i;
 475                         }
 476                         err = manual_repair(chunk_size, syndrome_disks,
 477                                             failed_slot1, failed_slot2,
 478                                             start, block_index_for_slot,
 479                                             name, stripes, blocks, p,
 480                                             source, offsets);
 481                         if(err == -1) {
 482                                 unlock_all_stripes(info, sig);
 483                                 goto exitCheck;
 484                         }
 485                 }
 486
 487                 err = unlock_all_stripes(info, sig);
 488                 if(err != 0) {
 489                         goto exitCheck;
 490                 }
 491
 492                 length--;
 493                 start++;
 494         }
 495
 496 exitCheck:
 497
 498         free(stripe_buf);
 499         free(stripes);
 500         free(blocks-2);
 501         free(blocks_page-2);
 502         free(block_index_for_slot-2);
 503         free(p);
 504         free(q);
 505         free(results);
 506         free(sig);
 507
 508         return err;
 509 }
 510
 511 unsigned long long getnum(char *str, char **err)
 512 {
 513         char *e;
 514         unsigned long long rv = strtoull(str, &e, 10);
 515         if (e==str || *e) {
 516                 *err = str;
 517                 return 0;
 518         }
 519         return rv;
 520 }
 521
 522 int main(int argc, char *argv[])
 523 {
 524         /* md_device start length */
 525         int *fds = NULL;
 526         char *buf = NULL;
 527         char **disk_name = NULL;
 528         unsigned long long *offsets = NULL;
 529         int raid_disks = 0;
 530         int active_disks;
 531         int chunk_size = 0;
 532         int layout = -1;
 533         int level = 6;
 534         enum repair repair = NO_REPAIR;
 535         int failed_disk1 = -1;
 536         int failed_disk2 = -1;
 537         unsigned long long start, length;
 538         int i;
 539         int mdfd;
 540         struct mdinfo *info = NULL, *comp = NULL;
 541         char *err = NULL;
 542         int exit_err = 0;
 543         int close_flag = 0;
 544         char *prg = strrchr(argv[0], '/');
 545
 546         if (prg == NULL)
 547                 prg = argv[0];
 548         else
 549                 prg++;
 550
 551         if (argc < 4) {
 552                 fprintf(stderr, "Usage: %s md_device start_stripe length_stripes [autorepair]\n", prg);
 553                 fprintf(stderr, "   or: %s md_device repair stripe failed_slot_1 failed_slot_2\n", prg);
 554                 exit_err = 1;
 555                 goto exitHere;
 556         }
 557
 558         mdfd = open(argv[1], O_RDONLY);
 559         if(mdfd < 0) {
 560                 perror(argv[1]);
 561                 fprintf(stderr, "%s: cannot open %s\n", prg, argv[1]);
 562                 exit_err = 2;
 563                 goto exitHere;
 564         }
 565
 566         info = sysfs_read(mdfd, NULL,
 567                           GET_LEVEL|
 568                           GET_LAYOUT|
 569                           GET_DISKS|
 570                           GET_STATE |
 571                           GET_COMPONENT|
 572                           GET_CHUNK|
 573                           GET_DEVS|
 574                           GET_OFFSET|
 575                           GET_SIZE);
 576
 577         if(info == NULL) {
 578                 fprintf(stderr, "%s: Error reading sysfs information of %s\n", prg, argv[1]);
 579                 exit_err = 9;
 580                 goto exitHere;
 581         }
 582
 583         if(info->array.level != level) {
 584                 fprintf(stderr, "%s: %s not a RAID-6\n", prg, argv[1]);
 585                 exit_err = 3;
 586                 goto exitHere;
 587         }
 588
 589         if(info->array.failed_disks > 0) {
 590                 fprintf(stderr, "%s: %s degraded array\n", prg, argv[1]);
 591                 exit_err = 8;
 592                 goto exitHere;
 593         }
 594
 595         printf("layout: %d\n", info->array.layout);
 596         printf("disks: %d\n", info->array.raid_disks);
 597         printf("component size: %llu\n", info->component_size * 512);
 598         printf("total stripes: %llu\n", (info->component_size * 512) / info->array.chunk_size);
 599         printf("chunk size: %d\n", info->array.chunk_size);
 600         printf("\n");
 601
 602         comp = info->devs;
 603         for(i = 0, active_disks = 0; active_disks < info->array.raid_disks; i++) {
 604                 printf("disk: %d - offset: %llu - size: %llu - name: %s - slot: %d\n",
 605                         i, comp->data_offset * 512, comp->component_size * 512,
 606                         map_dev(comp->disk.major, comp->disk.minor, 0),
 607                         comp->disk.raid_disk);
 608                 if(comp->disk.raid_disk >= 0)
 609                         active_disks++;
 610                 comp = comp->next;
 611         }
 612         printf("\n");
 613
 614         close(mdfd);
 615
 616         raid_disks = info->array.raid_disks;
 617         chunk_size = info->array.chunk_size;
 618         layout = info->array.layout;
 619         if (strcmp(argv[2], "repair")==0) {
 620                 if (argc < 6) {
 621                         fprintf(stderr, "For repair mode, call %s md_device repair stripe failed_slot_1 failed_slot_2\n", prg);
 622                         exit_err = 1;
 623                         goto exitHere;
 624                 }
 625                 repair = MANUAL_REPAIR;
 626                 start = getnum(argv[3], &err);
 627                 length = 1;
 628                 failed_disk1 = getnum(argv[4], &err);
 629                 failed_disk2 = getnum(argv[5], &err);
 630
 631                 if(failed_disk1 >= info->array.raid_disks) {
 632                         fprintf(stderr, "%s: failed_slot_1 index is higher than number of devices in raid\n", prg);
 633                         exit_err = 4;
 634                         goto exitHere;
 635                 }
 636                 if(failed_disk2 >= info->array.raid_disks) {
 637                         fprintf(stderr, "%s: failed_slot_2 index is higher than number of devices in raid\n", prg);
 638                         exit_err = 4;
 639                         goto exitHere;
 640                 }
 641                 if(failed_disk1 == failed_disk2) {
 642                         fprintf(stderr, "%s: failed_slot_1 and failed_slot_2 are the same\n", prg);
 643                         exit_err = 4;
 644                         goto exitHere;
 645                 }
 646         }
 647         else {
 648                 start = getnum(argv[2], &err);
 649                 length = getnum(argv[3], &err);
 650                 if (argc >= 5 && strcmp(argv[4], "autorepair")==0)
 651                         repair = AUTO_REPAIR;
 652         }
 653
 654         if (err) {
 655                 fprintf(stderr, "%s: Bad number: %s\n", prg, err);
 656                 exit_err = 4;
 657                 goto exitHere;
 658         }
 659
 660         if(start > ((info->component_size * 512) / chunk_size)) {
 661                 start = (info->component_size * 512) / chunk_size;
 662                 fprintf(stderr, "%s: start beyond disks size\n", prg);
 663         }
 664
 665         if((length == 0) ||
 666            ((length + start) > ((info->component_size * 512) / chunk_size))) {
 667                 length = (info->component_size * 512) / chunk_size - start;
 668         }
 669
 670         disk_name = xmalloc(raid_disks * sizeof(*disk_name));
 671         fds = xmalloc(raid_disks * sizeof(*fds));
 672         offsets = xcalloc(raid_disks, sizeof(*offsets));
 673         buf = xmalloc(raid_disks * chunk_size);
 674
 675         for(i=0; i<raid_disks; i++) {
 676                 fds[i] = -1;
 677         }
 678         close_flag = 1;
 679
 680         comp = info->devs;
 681         for (i=0, active_disks=0; active_disks<raid_disks; i++) {
 682                 int disk_slot = comp->disk.raid_disk;
 683                 if(disk_slot >= 0) {
 684                         disk_name[disk_slot] = map_dev(comp->disk.major, comp->disk.minor, 0);
 685                         offsets[disk_slot] = comp->data_offset * 512;
 686                         fds[disk_slot] = open(disk_name[disk_slot], O_RDWR | O_DIRECT);
 687                         if (fds[disk_slot] < 0) {
 688                                 perror(disk_name[disk_slot]);
 689                                 fprintf(stderr,"%s: cannot open %s\n", prg, disk_name[disk_slot]);
 690                                 exit_err = 6;
 691                                 goto exitHere;
 692                         }
 693                         active_disks++;
 694                 }
 695                 comp = comp->next;
 696         }
 697
 698         int rv = check_stripes(info, fds, offsets,
 699                                raid_disks, chunk_size, level, layout,
 700                                start, length, disk_name, repair, failed_disk1, failed_disk2);
 701         if (rv != 0) {
 702                 fprintf(stderr, "%s: check_stripes returned %d\n", prg, rv);
 703                 exit_err = 7;
 704                 goto exitHere;
 705         }
 706
 707 exitHere:
 708
 709         if (close_flag)
 710                 for(i = 0; i < raid_disks; i++)
 711                         close(fds[i]);
 712
 713         free(disk_name);
 714         free(fds);
 715         free(offsets);
 716         free(buf);
 717
 718         exit(exit_err);
 719 }