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 <signal.h>
  28 #include <sys/mman.h>
  29
  30 int geo_map(int block, unsigned long long stripe, int raid_disks,
  31             int level, int layout);
  32 void qsyndrome(uint8_t *p, uint8_t *q, uint8_t **sources, int disks, int size);
  33 void make_tables(void);
  34 void ensure_zero_has_size(int chunk_size);
  35 void raid6_datap_recov(int disks, size_t bytes, int faila, uint8_t **ptrs);
  36 void raid6_2data_recov(int disks, size_t bytes, int faila, int failb,
  37                        uint8_t **ptrs);
  38 void xor_blocks(char *target, char **sources, int disks, int size);
  39
  40
  41 /* Collect per stripe consistency information */
  42 void raid6_collect(int chunk_size, uint8_t *p, uint8_t *q,
  43                    char *chunkP, char *chunkQ, int *results)
  44 {
  45         int i;
  46         int data_id;
  47         uint8_t Px, Qx;
  48         extern uint8_t raid6_gflog[];
  49
  50         for(i = 0; i < chunk_size; i++) {
  51                 Px = (uint8_t)chunkP[i] ^ (uint8_t)p[i];
  52                 Qx = (uint8_t)chunkQ[i] ^ (uint8_t)q[i];
  53
  54                 if((Px != 0) && (Qx == 0))
  55                         results[i] = -1;
  56
  57                 if((Px == 0) && (Qx != 0))
  58                         results[i] = -2;
  59
  60                 if((Px != 0) && (Qx != 0)) {
  61                         data_id = (raid6_gflog[Qx] - raid6_gflog[Px]);
  62                         if(data_id < 0) data_id += 255;
  63                         results[i] = data_id;
  64                 }
  65
  66                 if((Px == 0) && (Qx == 0))
  67                         results[i] = -255;
  68         }
  69 }
  70
  71 /* Try to find out if a specific disk has problems */
  72 int raid6_stats(int *results, int raid_disks, int chunk_size)
  73 {
  74         int i;
  75         int curr_broken_disk = -255;
  76         int prev_broken_disk = -255;
  77         int broken_status = 0;
  78
  79         for(i = 0; i < chunk_size; i++) {
  80
  81                 if(results[i] != -255)
  82                         curr_broken_disk = results[i];
  83
  84                 if(curr_broken_disk >= raid_disks)
  85                         broken_status = 2;
  86
  87                 switch(broken_status) {
  88                 case 0:
  89                         if(curr_broken_disk != -255) {
  90                                 prev_broken_disk = curr_broken_disk;
  91                                 broken_status = 1;
  92                         }
  93                         break;
  94
  95                 case 1:
  96                         if(curr_broken_disk != prev_broken_disk)
  97                                 broken_status = 2;
  98                         break;
  99
 100                 case 2:
 101                 default:
 102                         curr_broken_disk = prev_broken_disk = -65535;
 103                         break;
 104                 }
 105         }
 106
 107         return curr_broken_disk;
 108 }
 109
 110 int check_stripes(struct mdinfo *info, int *source, unsigned long long *offsets,
 111                   int raid_disks, int chunk_size, int level, int layout,
 112                   unsigned long long start, unsigned long long length, char *name[],
 113                   int repair, int failed_disk1, int failed_disk2)
 114 {
 115         /* read the data and p and q blocks, and check we got them right */
 116         char *stripe_buf = xmalloc(raid_disks * chunk_size);
 117         char **stripes = xmalloc(raid_disks * sizeof(char*));
 118         char **blocks = xmalloc(raid_disks * sizeof(char*));
 119         uint8_t *p = xmalloc(chunk_size);
 120         uint8_t *q = xmalloc(chunk_size);
 121         int *results = xmalloc(chunk_size * sizeof(int));
 122
 123         int i;
 124         int diskP, diskQ;
 125         int data_disks = raid_disks - 2;
 126         int err = 0;
 127         sighandler_t sig[3];
 128         int rv;
 129
 130         extern int tables_ready;
 131
 132         if (!tables_ready)
 133                 make_tables();
 134
 135         for ( i = 0 ; i < raid_disks ; i++)
 136                 stripes[i] = stripe_buf + i * chunk_size;
 137
 138         while (length > 0) {
 139                 int disk;
 140
 141                 printf("pos --> %llu\n", start);
 142
 143                 if(mlockall(MCL_CURRENT | MCL_FUTURE) != 0) {
 144                         err = 2;
 145                         goto exitCheck;
 146                 }
 147                 sig[0] = signal(SIGTERM, SIG_IGN);
 148                 sig[1] = signal(SIGINT, SIG_IGN);
 149                 sig[2] = signal(SIGQUIT, SIG_IGN);
 150                 rv = sysfs_set_num(info, NULL, "suspend_lo", start * chunk_size * data_disks);
 151                 rv |= sysfs_set_num(info, NULL, "suspend_hi", (start + 1) * chunk_size * data_disks);
 152                 for (i = 0 ; i < raid_disks ; i++) {
 153                         lseek64(source[i], offsets[i] + start * chunk_size, 0);
 154                         read(source[i], stripes[i], chunk_size);
 155                 }
 156                 rv |= sysfs_set_num(info, NULL, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL);
 157                 rv |= sysfs_set_num(info, NULL, "suspend_hi", 0);
 158                 rv |= sysfs_set_num(info, NULL, "suspend_lo", 0);
 159                 signal(SIGQUIT, sig[2]);
 160                 signal(SIGINT, sig[1]);
 161                 signal(SIGTERM, sig[0]);
 162                 if(munlockall() != 0) {
 163                         err = 3;
 164                         goto exitCheck;
 165                 }
 166
 167                 if(rv != 0) {
 168                         err = rv * 256;
 169                         goto exitCheck;
 170                 }
 171
 172                 for (i = 0 ; i < data_disks ; i++) {
 173                         int disk = geo_map(i, start, raid_disks, level, layout);
 174                         blocks[i] = stripes[disk];
 175                         printf("%d->%d\n", i, disk);
 176                 }
 177
 178                 qsyndrome(p, q, (uint8_t**)blocks, data_disks, chunk_size);
 179                 diskP = geo_map(-1, start, raid_disks, level, layout);
 180                 diskQ = geo_map(-2, start, raid_disks, level, layout);
 181                 blocks[data_disks] = stripes[diskP];
 182                 blocks[data_disks+1] = stripes[diskQ];
 183
 184                 if (memcmp(p, stripes[diskP], chunk_size) != 0) {
 185                         printf("P(%d) wrong at %llu\n", diskP, start);
 186                 }
 187                 if (memcmp(q, stripes[diskQ], chunk_size) != 0) {
 188                         printf("Q(%d) wrong at %llu\n", diskQ, start);
 189                 }
 190                 raid6_collect(chunk_size, p, q, stripes[diskP], stripes[diskQ], results);
 191                 disk = raid6_stats(results, raid_disks, chunk_size);
 192
 193                 if(disk >= -2) {
 194                         disk = geo_map(disk, start, raid_disks, level, layout);
 195                 }
 196                 if(disk >= 0) {
 197                         printf("Error detected at %llu: possible failed disk slot: %d --> %s\n",
 198                                 start, disk, name[disk]);
 199                 }
 200                 if(disk == -65535) {
 201                         printf("Error detected at %llu: disk slot unknown\n", start);
 202                 }
 203                 if(repair == 1) {
 204                         printf("Repairing stripe %llu\n", start);
 205                         printf("Assuming slots %d (%s) and %d (%s) are incorrect\n",
 206                                failed_disk1, name[failed_disk1],
 207                                failed_disk2, name[failed_disk2]);
 208
 209                         if (failed_disk1 == diskQ || failed_disk2 == diskQ) {
 210                                 char *all_but_failed_blocks[data_disks];
 211                                 int failed_data;
 212                                 int failed_block_index;
 213
 214                                 if (failed_disk1 == diskQ)
 215                                         failed_data = failed_disk2;
 216                                 else
 217                                         failed_data = failed_disk1;
 218                                 printf("Repairing D/P(%d) and Q\n", failed_data);
 219                                 failed_block_index = geo_map(
 220                                         failed_data, start, raid_disks,
 221                                         level, layout);
 222                                 for (i=0; i < data_disks; i++)
 223                                         if (failed_block_index == i)
 224                                                 all_but_failed_blocks[i] = stripes[diskP];
 225                                         else
 226                                                 all_but_failed_blocks[i] = blocks[i];
 227                                 xor_blocks(stripes[failed_data],
 228                                         all_but_failed_blocks, data_disks, chunk_size);
 229                                 qsyndrome(p, (uint8_t*)stripes[diskQ], (uint8_t**)blocks, data_disks, chunk_size);
 230                         } else {
 231                                 ensure_zero_has_size(chunk_size);
 232                                 if (failed_disk1 == diskP || failed_disk2 == diskP) {
 233                                         int failed_data, failed_block_index;
 234                                         if (failed_disk1 == diskP)
 235                                                 failed_data = failed_disk2;
 236                                         else
 237                                                 failed_data = failed_disk1;
 238                                         failed_block_index = geo_map(failed_data, start, raid_disks, level, layout);
 239                                         printf("Repairing D(%d) and P\n", failed_data);
 240                                         raid6_datap_recov(raid_disks, chunk_size, failed_block_index, (uint8_t**)blocks);
 241                                 } else {
 242                                         printf("Repairing D and D\n");
 243                                         int failed_block_index1 = geo_map(failed_disk1, start, raid_disks, level, layout);
 244                                         int failed_block_index2 = geo_map(failed_disk2, start, raid_disks, level, layout);
 245                                         if (failed_block_index1 > failed_block_index2) {
 246                                                 int t = failed_block_index1;
 247                                                 failed_block_index1 = failed_block_index2;
 248                                                 failed_block_index2 = t;
 249                                         }
 250                                         raid6_2data_recov(raid_disks, chunk_size, failed_block_index1, failed_block_index2, (uint8_t**)blocks);
 251                                 }
 252                         }
 253                         if(mlockall(MCL_CURRENT | MCL_FUTURE) != 0) {
 254                                 err = 2;
 255                                 goto exitCheck;
 256                         }
 257                         sig[0] = signal(SIGTERM, SIG_IGN);
 258                         sig[1] = signal(SIGINT, SIG_IGN);
 259                         sig[2] = signal(SIGQUIT, SIG_IGN);
 260                         rv = sysfs_set_num(info, NULL, "suspend_lo", start * chunk_size * data_disks);
 261                         rv |= sysfs_set_num(info, NULL, "suspend_hi", (start + 1) * chunk_size * data_disks);
 262                         lseek64(source[failed_disk1], offsets[failed_disk1] + start * chunk_size, 0);
 263                         write(source[failed_disk1], stripes[failed_disk1], chunk_size);
 264                         lseek64(source[failed_disk2], offsets[failed_disk2] + start * chunk_size, 0);
 265                         write(source[failed_disk2], stripes[failed_disk2], chunk_size);
 266                         rv |= sysfs_set_num(info, NULL, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL);
 267                         rv |= sysfs_set_num(info, NULL, "suspend_hi", 0);
 268                         rv |= sysfs_set_num(info, NULL, "suspend_lo", 0);
 269                         signal(SIGQUIT, sig[2]);
 270                         signal(SIGINT, sig[1]);
 271                         signal(SIGTERM, sig[0]);
 272                         if(munlockall() != 0) {
 273                                 err = 3;
 274                                 goto exitCheck;
 275                         }
 276
 277                         if(rv != 0) {
 278                                 err = rv * 256;
 279                                 goto exitCheck;
 280                         }
 281                 }
 282
 283
 284                 length--;
 285                 start++;
 286         }
 287
 288 exitCheck:
 289
 290         free(stripe_buf);
 291         free(stripes);
 292         free(blocks);
 293         free(p);
 294         free(q);
 295         free(results);
 296
 297         return err;
 298 }
 299
 300 unsigned long long getnum(char *str, char **err)
 301 {
 302         char *e;
 303         unsigned long long rv = strtoull(str, &e, 10);
 304         if (e==str || *e) {
 305                 *err = str;
 306                 return 0;
 307         }
 308         return rv;
 309 }
 310
 311 int main(int argc, char *argv[])
 312 {
 313         /* md_device start length */
 314         int *fds = NULL;
 315         char *buf = NULL;
 316         char **disk_name = NULL;
 317         unsigned long long *offsets = NULL;
 318         int raid_disks = 0;
 319         int active_disks;
 320         int chunk_size = 0;
 321         int layout = -1;
 322         int level = 6;
 323         int repair = 0;
 324         int failed_disk1, failed_disk2;
 325         unsigned long long start, length;
 326         int i;
 327         int mdfd;
 328         struct mdinfo *info = NULL, *comp = NULL;
 329         char *err = NULL;
 330         int exit_err = 0;
 331         int close_flag = 0;
 332         char *prg = strrchr(argv[0], '/');
 333
 334         if (prg == NULL)
 335                 prg = argv[0];
 336         else
 337                 prg++;
 338
 339         if (argc < 4) {
 340                 fprintf(stderr, "Usage: %s md_device start_stripe length_stripes\n", prg);
 341                 fprintf(stderr, "   or: %s md_device repair stripe failed_slot_1 failed_slot_2\n", prg);
 342                 exit_err = 1;
 343                 goto exitHere;
 344         }
 345
 346         mdfd = open(argv[1], O_RDONLY);
 347         if(mdfd < 0) {
 348                 perror(argv[1]);
 349                 fprintf(stderr, "%s: cannot open %s\n", prg, argv[1]);
 350                 exit_err = 2;
 351                 goto exitHere;
 352         }
 353
 354         info = sysfs_read(mdfd, -1,
 355                           GET_LEVEL|
 356                           GET_LAYOUT|
 357                           GET_DISKS|
 358                           GET_DEGRADED |
 359                           GET_COMPONENT|
 360                           GET_CHUNK|
 361                           GET_DEVS|
 362                           GET_OFFSET|
 363                           GET_SIZE);
 364
 365         if(info == NULL) {
 366                 fprintf(stderr, "%s: Error reading sysfs information of %s\n", prg, argv[1]);
 367                 exit_err = 9;
 368                 goto exitHere;
 369         }
 370
 371         if(info->array.level != level) {
 372                 fprintf(stderr, "%s: %s not a RAID-6\n", prg, argv[1]);
 373                 exit_err = 3;
 374                 goto exitHere;
 375         }
 376
 377         if(info->array.failed_disks > 0) {
 378                 fprintf(stderr, "%s: %s degraded array\n", prg, argv[1]);
 379                 exit_err = 8;
 380                 goto exitHere;
 381         }
 382
 383         printf("layout: %d\n", info->array.layout);
 384         printf("disks: %d\n", info->array.raid_disks);
 385         printf("component size: %llu\n", info->component_size * 512);
 386         printf("total stripes: %llu\n", (info->component_size * 512) / info->array.chunk_size);
 387         printf("chunk size: %d\n", info->array.chunk_size);
 388         printf("\n");
 389
 390         comp = info->devs;
 391         for(i = 0, active_disks = 0; active_disks < info->array.raid_disks; i++) {
 392                 printf("disk: %d - offset: %llu - size: %llu - name: %s - slot: %d\n",
 393                         i, comp->data_offset * 512, comp->component_size * 512,
 394                         map_dev(comp->disk.major, comp->disk.minor, 0),
 395                         comp->disk.raid_disk);
 396                 if(comp->disk.raid_disk >= 0)
 397                         active_disks++;
 398                 comp = comp->next;
 399         }
 400         printf("\n");
 401
 402         close(mdfd);
 403
 404         raid_disks = info->array.raid_disks;
 405         chunk_size = info->array.chunk_size;
 406         layout = info->array.layout;
 407         if (strcmp(argv[2], "repair")==0) {
 408                 if (argc < 6) {
 409                         fprintf(stderr, "For repair mode, call %s md_device repair stripe failed_slot_1 failed_slot_2\n", prg);
 410                         exit_err = 1;
 411                         goto exitHere;
 412                 }
 413                 repair = 1;
 414                 start = getnum(argv[3], &err);
 415                 length = 1;
 416                 failed_disk1 = getnum(argv[4], &err);
 417                 failed_disk2 = getnum(argv[5], &err);
 418
 419                 if(failed_disk1 > info->array.raid_disks) {
 420                         fprintf(stderr, "%s: failed_slot_1 index is higher than number of devices in raid\n", prg);
 421                         exit_err = 4;
 422                         goto exitHere;
 423                 }
 424                 if(failed_disk2 > info->array.raid_disks) {
 425                         fprintf(stderr, "%s: failed_slot_2 index is higher than number of devices in raid\n", prg);
 426                         exit_err = 4;
 427                         goto exitHere;
 428                 }
 429                 if(failed_disk1 == failed_disk2) {
 430                         fprintf(stderr, "%s: failed_slot_1 and failed_slot_2 are the same\n", prg);
 431                         exit_err = 4;
 432                         goto exitHere;
 433                 }
 434         }
 435         else {
 436                 start = getnum(argv[2], &err);
 437                 length = getnum(argv[3], &err);
 438         }
 439
 440         if (err) {
 441                 fprintf(stderr, "%s: Bad number: %s\n", prg, err);
 442                 exit_err = 4;
 443                 goto exitHere;
 444         }
 445
 446         if(start > ((info->component_size * 512) / chunk_size)) {
 447                 start = (info->component_size * 512) / chunk_size;
 448                 fprintf(stderr, "%s: start beyond disks size\n", prg);
 449         }
 450
 451         if((length == 0) ||
 452            ((length + start) > ((info->component_size * 512) / chunk_size))) {
 453                 length = (info->component_size * 512) / chunk_size - start;
 454         }
 455
 456         disk_name = xmalloc(raid_disks * sizeof(*disk_name));
 457         fds = xmalloc(raid_disks * sizeof(*fds));
 458         offsets = xcalloc(raid_disks, sizeof(*offsets));
 459         buf = xmalloc(raid_disks * chunk_size);
 460
 461         for(i=0; i<raid_disks; i++) {
 462                 fds[i] = -1;
 463         }
 464         close_flag = 1;
 465
 466         comp = info->devs;
 467         for (i=0, active_disks=0; active_disks<raid_disks; i++) {
 468                 int disk_slot = comp->disk.raid_disk;
 469                 if(disk_slot >= 0) {
 470                         disk_name[disk_slot] = map_dev(comp->disk.major, comp->disk.minor, 0);
 471                         offsets[disk_slot] = comp->data_offset * 512;
 472                         fds[disk_slot] = open(disk_name[disk_slot], O_RDWR);
 473                         if (fds[disk_slot] < 0) {
 474                                 perror(disk_name[disk_slot]);
 475                                 fprintf(stderr,"%s: cannot open %s\n", prg, disk_name[disk_slot]);
 476                                 exit_err = 6;
 477                                 goto exitHere;
 478                         }
 479                         active_disks++;
 480                 }
 481                 comp = comp->next;
 482         }
 483
 484         int rv = check_stripes(info, fds, offsets,
 485                                raid_disks, chunk_size, level, layout,
 486                                start, length, disk_name, repair, failed_disk1, failed_disk2);
 487         if (rv != 0) {
 488                 fprintf(stderr,
 489                         "%s: check_stripes returned %d\n", prg, rv);
 490                 exit_err = 7;
 491                 goto exitHere;
 492         }
 493
 494 exitHere:
 495
 496         if (close_flag)
 497                 for(i = 0; i < raid_disks; i++)
 498                         close(fds[i]);
 499
 500         free(disk_name);
 501         free(fds);
 502         free(offsets);
 503         free(buf);
 504
 505         exit(exit_err);
 506 }