sysfs.c

   1 /*
   2  * sysfs - extract md related information from sysfs.  Part of:
   3  * mdadm - manage Linux "md" devices aka RAID arrays.
   4  *
   5  * Copyright (C) 2006-2009 Neil Brown <neilb@suse.de>
   6  *
   7  *
   8  *    This program is free software; you can redistribute it and/or modify
   9  *    it under the terms of the GNU General Public License as published by
  10  *    the Free Software Foundation; either version 2 of the License, or
  11  *    (at your option) any later version.
  12  *
  13  *    This program is distributed in the hope that it will be useful,
  14  *    but WITHOUT ANY WARRANTY; without even the implied warranty of
  15  *    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  16  *    GNU General Public License for more details.
  17  *
  18  *    You should have received a copy of the GNU General Public License
  19  *    along with this program; if not, write to the Free Software
  20  *    Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  21  *
  22  *    Author: Neil Brown
  23  *    Email: <neilb@suse.de>
  24  */
  25
  26 #include        "mdadm.h"
  27 #include        <dirent.h>
  28 #include        <ctype.h>
  29 #include        "dlink.h"
  30
  31 #define MAX_SYSFS_PATH_LEN      120
  32
  33 struct dev_sysfs_rule {
  34         struct dev_sysfs_rule *next;
  35         char *devname;
  36         int uuid[4];
  37         int uuid_set;
  38         struct sysfs_entry {
  39                 struct sysfs_entry *next;
  40                 char *name;
  41                 char *value;
  42         } *entry;
  43 };
  44
  45 int load_sys(char *path, char *buf, int len)
  46 {
  47         int fd = open(path, O_RDONLY);
  48         int n;
  49         if (fd < 0)
  50                 return -1;
  51         n = read(fd, buf, len);
  52         close(fd);
  53         if (n <0 || n >= len)
  54                 return -1;
  55         buf[n] = 0;
  56         if (n && buf[n-1] == '\n')
  57                 buf[n-1] = 0;
  58         return 0;
  59 }
  60
  61 void sysfs_free(struct mdinfo *sra)
  62 {
  63         while (sra) {
  64                 struct mdinfo *sra2 = sra->next;
  65                 while (sra->devs) {
  66                         struct mdinfo *d = sra->devs;
  67                         sra->devs = d->next;
  68                         free(d->bb.entries);
  69                         free(d);
  70                 }
  71                 free(sra->bb.entries);
  72                 free(sra);
  73                 sra = sra2;
  74         }
  75 }
  76
  77 /**
  78  * sysfs_get_container_devnm() - extract container device name.
  79  * @mdi: md_info describes member array, with GET_VERSION option.
  80  * @buf: buf to fill, must be MD_NAME_MAX.
  81  *
  82  * External array version is in format {/,-}<container_devnm>/<array_index>
  83  * Extract container_devnm from it and safe it in @buf.
  84  */
  85 void sysfs_get_container_devnm(struct mdinfo *mdi, char *buf)
  86 {
  87         char *p;
  88
  89         assert(is_subarray(mdi->text_version));
  90
  91         /* Skip first special sign */
  92         snprintf(buf, MD_NAME_MAX, "%s", mdi->text_version + 1);
  93
  94         /* Remove array index */
  95         p = strchr(buf, '/');
  96         if (p)
  97                 *p = 0;
  98 }
  99
 100 int sysfs_open(char *devnm, char *devname, char *attr)
 101 {
 102         char fname[MAX_SYSFS_PATH_LEN];
 103         int fd;
 104
 105         snprintf(fname, MAX_SYSFS_PATH_LEN, "/sys/block/%s/md/", devnm);
 106         if (devname) {
 107                 strncat(fname, devname, MAX_SYSFS_PATH_LEN - strlen(fname));
 108                 strncat(fname, "/", MAX_SYSFS_PATH_LEN - strlen(fname));
 109         }
 110         strncat(fname, attr, MAX_SYSFS_PATH_LEN - strlen(fname));
 111         fd = open(fname, O_RDWR);
 112         if (fd < 0 && errno == EACCES)
 113                 fd = open(fname, O_RDONLY);
 114         return fd;
 115 }
 116
 117 void sysfs_init_dev(struct mdinfo *mdi, dev_t devid)
 118 {
 119         snprintf(mdi->sys_name,
 120                  sizeof(mdi->sys_name), "dev-%s", devid2kname(devid));
 121 }
 122
 123 int sysfs_init(struct mdinfo *mdi, int fd, char *devnm)
 124 {
 125         struct stat stb;
 126         char fname[MAX_SYSFS_PATH_LEN];
 127         int retval = -ENODEV;
 128
 129         mdi->sys_name[0] = 0;
 130         if (fd >= 0)
 131                 devnm = fd2devnm(fd);
 132
 133         if (devnm == NULL)
 134                 goto out;
 135
 136         snprintf(fname, MAX_SYSFS_PATH_LEN, "/sys/block/%s/md", devnm);
 137
 138         if (stat(fname, &stb))
 139                 goto out;
 140         if (!S_ISDIR(stb.st_mode))
 141                 goto out;
 142         strcpy(mdi->sys_name, devnm);
 143
 144         retval = 0;
 145 out:
 146         return retval;
 147 }
 148
 149 struct mdinfo *sysfs_read(int fd, char *devnm, unsigned long options)
 150 {
 151         char fname[PATH_MAX];
 152         char buf[PATH_MAX];
 153         char *base;
 154         char *dbase;
 155         struct mdinfo *sra;
 156         struct mdinfo *dev, **devp;
 157         DIR *dir = NULL;
 158         struct dirent *de;
 159
 160         sra = xcalloc(1, sizeof(*sra));
 161         if (sysfs_init(sra, fd, devnm)) {
 162                 free(sra);
 163                 return NULL;
 164         }
 165
 166         sprintf(fname, "/sys/block/%s/md/", sra->sys_name);
 167         base = fname + strlen(fname);
 168
 169         sra->devs = NULL;
 170         if (options & GET_VERSION) {
 171                 strcpy(base, "metadata_version");
 172                 if (load_sys(fname, buf, sizeof(buf)))
 173                         goto abort;
 174                 if (str_is_none(buf) == true) {
 175                         sra->array.major_version =
 176                                 sra->array.minor_version = -1;
 177                         strcpy(sra->text_version, "");
 178                 } else if (strncmp(buf, "external:", 9) == 0) {
 179                         sra->array.major_version = -1;
 180                         sra->array.minor_version = -2;
 181                         strcpy(sra->text_version, buf+9);
 182                 } else {
 183                         sscanf(buf, "%d.%d",
 184                                &sra->array.major_version,
 185                                &sra->array.minor_version);
 186                         strcpy(sra->text_version, buf);
 187                 }
 188         }
 189         if (options & GET_LEVEL) {
 190                 strcpy(base, "level");
 191                 if (load_sys(fname, buf, sizeof(buf)))
 192                         goto abort;
 193                 sra->array.level = map_name(pers, buf);
 194         }
 195         if (options & GET_LAYOUT) {
 196                 strcpy(base, "layout");
 197                 if (load_sys(fname, buf, sizeof(buf)))
 198                         goto abort;
 199                 sra->array.layout = strtoul(buf, NULL, 0);
 200         }
 201         if (options & (GET_DISKS|GET_STATE)) {
 202                 strcpy(base, "raid_disks");
 203                 if (load_sys(fname, buf, sizeof(buf)))
 204                         goto abort;
 205                 sra->array.raid_disks = strtoul(buf, NULL, 0);
 206         }
 207         if (options & GET_COMPONENT) {
 208                 strcpy(base, "component_size");
 209                 if (load_sys(fname, buf, sizeof(buf)))
 210                         goto abort;
 211                 sra->component_size = strtoull(buf, NULL, 0);
 212                 /* sysfs reports "K", but we want sectors */
 213                 sra->component_size *= 2;
 214         }
 215         if (options & GET_CHUNK) {
 216                 strcpy(base, "chunk_size");
 217                 if (load_sys(fname, buf, sizeof(buf)))
 218                         goto abort;
 219                 sra->array.chunk_size = strtoul(buf, NULL, 0);
 220         }
 221         if (options & GET_CACHE) {
 222                 strcpy(base, "stripe_cache_size");
 223                 if (load_sys(fname, buf, sizeof(buf)))
 224                         /* Probably level doesn't support it */
 225                         sra->cache_size = 0;
 226                 else
 227                         sra->cache_size = strtoul(buf, NULL, 0);
 228         }
 229         if (options & GET_MISMATCH) {
 230                 strcpy(base, "mismatch_cnt");
 231                 if (load_sys(fname, buf, sizeof(buf)))
 232                         goto abort;
 233                 sra->mismatch_cnt = strtoul(buf, NULL, 0);
 234         }
 235         if (options & GET_SAFEMODE) {
 236                 int scale = 1;
 237                 int dot = 0;
 238                 unsigned i;
 239                 unsigned long msec;
 240                 size_t len;
 241
 242                 strcpy(base, "safe_mode_delay");
 243                 if (load_sys(fname, buf, sizeof(buf)))
 244                         goto abort;
 245
 246                 /* remove a period, and count digits after it */
 247                 len = strlen(buf);
 248                 for (i = 0; i < len; i++) {
 249                         if (dot) {
 250                                 if (isdigit(buf[i])) {
 251                                         buf[i-1] = buf[i];
 252                                         scale *= 10;
 253                                 }
 254                                 buf[i] = 0;
 255                         } else if (buf[i] == '.') {
 256                                 dot=1;
 257                                 buf[i] = 0;
 258                         }
 259                 }
 260                 msec = strtoul(buf, NULL, 10);
 261                 msec = (msec * 1000) / scale;
 262                 sra->safe_mode_delay = msec;
 263         }
 264         if (options & GET_BITMAP_LOCATION) {
 265                 strcpy(base, "bitmap/location");
 266                 if (load_sys(fname, buf, sizeof(buf)))
 267                         goto abort;
 268                 if (strncmp(buf, "file", 4) == 0)
 269                         sra->bitmap_offset = 1;
 270                 else if (str_is_none(buf) == true)
 271                         sra->bitmap_offset = 0;
 272                 else if (buf[0] == '+')
 273                         sra->bitmap_offset = strtol(buf+1, NULL, 10);
 274                 else
 275                         goto abort;
 276         }
 277
 278         if (options & GET_ARRAY_STATE) {
 279                 strcpy(base, "array_state");
 280                 if (load_sys(fname, buf, sizeof(buf)))
 281                         goto abort;
 282                 sra->array_state = map_name(sysfs_array_states, buf);
 283         }
 284
 285         if (options & GET_CONSISTENCY_POLICY) {
 286                 strcpy(base, "consistency_policy");
 287                 if (load_sys(fname, buf, sizeof(buf)))
 288                         sra->consistency_policy = CONSISTENCY_POLICY_UNKNOWN;
 289                 else
 290                         sra->consistency_policy = map_name(consistency_policies,
 291                                                            buf);
 292         }
 293
 294         if (! (options & GET_DEVS))
 295                 return sra;
 296
 297         /* Get all the devices as well */
 298         *base = 0;
 299         dir = opendir(fname);
 300         if (!dir)
 301                 goto abort;
 302         sra->array.spare_disks = 0;
 303         sra->array.active_disks = 0;
 304         sra->array.failed_disks = 0;
 305         sra->array.working_disks = 0;
 306
 307         devp = &sra->devs;
 308         sra->devs = NULL;
 309         while ((de = readdir(dir)) != NULL) {
 310                 char *ep;
 311                 if (de->d_ino == 0 ||
 312                     strncmp(de->d_name, "dev-", 4) != 0)
 313                         continue;
 314                 strcpy(base, de->d_name);
 315                 dbase = base + strlen(base);
 316                 *dbase++ = '/';
 317
 318                 dev = xcalloc(1, sizeof(*dev));
 319
 320                 /* Always get slot, major, minor */
 321                 strcpy(dbase, "slot");
 322                 if (load_sys(fname, buf, sizeof(buf))) {
 323                         /* hmm... unable to read 'slot' maybe the device
 324                          * is going away?
 325                          */
 326                         strcpy(dbase, "block");
 327                         if (readlink(fname, buf, sizeof(buf)) < 0 &&
 328                             errno != ENAMETOOLONG) {
 329                                 /* ...yup device is gone */
 330                                 free(dev);
 331                                 continue;
 332                         } else {
 333                                 /* slot is unreadable but 'block' link
 334                                  * still intact... something bad is happening
 335                                  * so abort
 336                                  */
 337                                 free(dev);
 338                                 goto abort;
 339                         }
 340
 341                 }
 342                 strcpy(dev->sys_name, de->d_name);
 343                 dev->disk.raid_disk = strtoul(buf, &ep, 10);
 344                 if (*ep) dev->disk.raid_disk = -1;
 345
 346                 sra->array.nr_disks++;
 347                 strcpy(dbase, "block/dev");
 348                 if (load_sys(fname, buf, sizeof(buf))) {
 349                         /* assume this is a stale reference to a hot
 350                          * removed device
 351                          */
 352                         if (!(options & GET_DEVS_ALL)) {
 353                                 free(dev);
 354                                 continue;
 355                         }
 356                 } else {
 357                         sscanf(buf, "%d:%d", &dev->disk.major, &dev->disk.minor);
 358                 }
 359
 360                 if (!(options & GET_DEVS_ALL)) {
 361                         /* special case check for block devices that can go 'offline' */
 362                         strcpy(dbase, "block/device/state");
 363                         if (load_sys(fname, buf, sizeof(buf)) == 0 &&
 364                             strncmp(buf, "offline", 7) == 0) {
 365                                 free(dev);
 366                                 continue;
 367                         }
 368                 }
 369
 370                 /* finally add this disk to the array */
 371                 *devp = dev;
 372                 devp = & dev->next;
 373                 dev->next = NULL;
 374
 375                 if (options & GET_OFFSET) {
 376                         strcpy(dbase, "offset");
 377                         if (load_sys(fname, buf, sizeof(buf)))
 378                                 goto abort;
 379                         dev->data_offset = strtoull(buf, NULL, 0);
 380                         strcpy(dbase, "new_offset");
 381                         if (load_sys(fname, buf, sizeof(buf)) == 0)
 382                                 dev->new_data_offset = strtoull(buf, NULL, 0);
 383                         else
 384                                 dev->new_data_offset = dev->data_offset;
 385                 }
 386                 if (options & GET_SIZE) {
 387                         strcpy(dbase, "size");
 388                         if (load_sys(fname, buf, sizeof(buf)))
 389                                 goto abort;
 390                         dev->component_size = strtoull(buf, NULL, 0) * 2;
 391                 }
 392                 if (options & GET_STATE) {
 393                         dev->disk.state = 0;
 394                         strcpy(dbase, "state");
 395                         if (load_sys(fname, buf, sizeof(buf)))
 396                                 goto abort;
 397                         if (strstr(buf, "faulty"))
 398                                 dev->disk.state |= (1<<MD_DISK_FAULTY);
 399                         else {
 400                                 sra->array.working_disks++;
 401                                 if (strstr(buf, "in_sync")) {
 402                                         dev->disk.state |= (1<<MD_DISK_SYNC);
 403                                         sra->array.active_disks++;
 404                                 }
 405                                 if (dev->disk.state == 0)
 406                                         sra->array.spare_disks++;
 407                         }
 408                 }
 409                 if (options & GET_ERROR) {
 410                         strcpy(buf, "errors");
 411                         if (load_sys(fname, buf, sizeof(buf)))
 412                                 goto abort;
 413                         dev->errors = strtoul(buf, NULL, 0);
 414                 }
 415         }
 416
 417         if ((options & GET_STATE) && sra->array.raid_disks)
 418                 sra->array.failed_disks = sra->array.raid_disks -
 419                         sra->array.active_disks - sra->array.spare_disks;
 420
 421         closedir(dir);
 422         return sra;
 423
 424  abort:
 425         if (dir)
 426                 closedir(dir);
 427         sysfs_free(sra);
 428         return NULL;
 429 }
 430
 431 int sysfs_attr_match(const char *attr, const char *str)
 432 {
 433         /* See if attr, read from a sysfs file, matches
 434          * str.  They must either be the same, or attr can
 435          * have a trailing newline or comma
 436          */
 437         while (*attr && *str && *attr == *str) {
 438                 attr++;
 439                 str++;
 440         }
 441
 442         if (*str || (*attr && *attr != ',' && *attr != '\n'))
 443                 return 0;
 444         return 1;
 445 }
 446
 447 int sysfs_match_word(const char *word, char **list)
 448 {
 449         int n;
 450         for (n=0; list[n]; n++)
 451                 if (sysfs_attr_match(word, list[n]))
 452                         break;
 453         return n;
 454 }
 455
 456 unsigned long long get_component_size(int fd)
 457 {
 458         /* Find out the component size of the array.
 459          * We cannot trust GET_ARRAY_INFO ioctl as it's
 460          * size field is only 32bits.
 461          * So look in /sys/block/mdXXX/md/component_size
 462          *
 463          * This returns in units of sectors.
 464          */
 465         struct stat stb;
 466         char fname[MAX_SYSFS_PATH_LEN];
 467         int n;
 468         if (fstat(fd, &stb))
 469                 return 0;
 470         snprintf(fname, MAX_SYSFS_PATH_LEN,
 471                  "/sys/block/%s/md/component_size", stat2devnm(&stb));
 472         fd = open(fname, O_RDONLY);
 473         if (fd < 0)
 474                 return 0;
 475         n = read(fd, fname, sizeof(fname));
 476         close(fd);
 477         if (n < 0 || n == sizeof(fname))
 478                 return 0;
 479         fname[n] = 0;
 480         return strtoull(fname, NULL, 10) * 2;
 481 }
 482
 483 int sysfs_set_str(struct mdinfo *sra, struct mdinfo *dev,
 484                   char *name, char *val)
 485 {
 486         char fname[MAX_SYSFS_PATH_LEN];
 487         unsigned int n;
 488         int fd;
 489
 490         snprintf(fname, MAX_SYSFS_PATH_LEN, "/sys/block/%s/md/%s/%s",
 491                 sra->sys_name, dev?dev->sys_name:"", name);
 492         fd = open(fname, O_WRONLY);
 493         if (fd < 0)
 494                 return -1;
 495         n = write(fd, val, strlen(val));
 496         close(fd);
 497         if (n != strlen(val)) {
 498                 dprintf("failed to write '%s' to '%s' (%s)\n",
 499                         val, fname, strerror(errno));
 500                 return -1;
 501         }
 502         return 0;
 503 }
 504
 505 int sysfs_set_num(struct mdinfo *sra, struct mdinfo *dev,
 506                   char *name, unsigned long long val)
 507 {
 508         char valstr[50];
 509         sprintf(valstr, "%llu", val);
 510         return sysfs_set_str(sra, dev, name, valstr);
 511 }
 512
 513 int sysfs_set_num_signed(struct mdinfo *sra, struct mdinfo *dev,
 514                          char *name, long long val)
 515 {
 516         char valstr[50];
 517         sprintf(valstr, "%lli", val);
 518         return sysfs_set_str(sra, dev, name, valstr);
 519 }
 520
 521 int sysfs_uevent(struct mdinfo *sra, char *event)
 522 {
 523         char fname[MAX_SYSFS_PATH_LEN];
 524         int n;
 525         int fd;
 526
 527         snprintf(fname, MAX_SYSFS_PATH_LEN, "/sys/block/%s/uevent",
 528                 sra->sys_name);
 529         fd = open(fname, O_WRONLY);
 530         if (fd < 0)
 531                 return -1;
 532         n = write(fd, event, strlen(event));
 533         close(fd);
 534         if (n != (int)strlen(event)) {
 535                 dprintf("failed to write '%s' to '%s' (%s)\n",
 536                         event, fname, strerror(errno));
 537                 return -1;
 538         }
 539         return 0;
 540 }
 541
 542 int sysfs_attribute_available(struct mdinfo *sra, struct mdinfo *dev, char *name)
 543 {
 544         char fname[MAX_SYSFS_PATH_LEN];
 545         struct stat st;
 546
 547         snprintf(fname, MAX_SYSFS_PATH_LEN, "/sys/block/%s/md/%s/%s",
 548                 sra->sys_name, dev?dev->sys_name:"", name);
 549
 550         return stat(fname, &st) == 0;
 551 }
 552
 553 int sysfs_get_fd(struct mdinfo *sra, struct mdinfo *dev,
 554                        char *name)
 555 {
 556         char fname[MAX_SYSFS_PATH_LEN];
 557         int fd;
 558
 559         snprintf(fname, MAX_SYSFS_PATH_LEN, "/sys/block/%s/md/%s/%s",
 560                 sra->sys_name, dev?dev->sys_name:"", name);
 561         fd = open(fname, O_RDWR);
 562         if (fd < 0)
 563                 fd = open(fname, O_RDONLY);
 564         return fd;
 565 }
 566
 567 int sysfs_fd_get_ll(int fd, unsigned long long *val)
 568 {
 569         char buf[50];
 570         int n;
 571         char *ep;
 572
 573         lseek(fd, 0, 0);
 574         n = read(fd, buf, sizeof(buf));
 575         if (n <= 0 || n == sizeof(buf))
 576                 return -2;
 577         buf[n] = 0;
 578         *val = strtoull(buf, &ep, 0);
 579         if (ep == buf || (*ep != 0 && *ep != '\n' && *ep != ' '))
 580                 return -1;
 581         return 0;
 582 }
 583
 584 int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev,
 585                        char *name, unsigned long long *val)
 586 {
 587         int n;
 588         int fd;
 589
 590         fd = sysfs_get_fd(sra, dev, name);
 591         if (fd < 0)
 592                 return -1;
 593         n = sysfs_fd_get_ll(fd, val);
 594         close(fd);
 595         return n;
 596 }
 597
 598 int sysfs_fd_get_two(int fd, unsigned long long *v1, unsigned long long *v2)
 599 {
 600         /* two numbers in this sysfs file, either
 601          *  NNN (NNN)
 602          * or
 603          *  NNN / NNN
 604          */
 605         char buf[80];
 606         int n;
 607         char *ep, *ep2;
 608
 609         lseek(fd, 0, 0);
 610         n = read(fd, buf, sizeof(buf));
 611         if (n <= 0 || n == sizeof(buf))
 612                 return -2;
 613         buf[n] = 0;
 614         *v1 = strtoull(buf, &ep, 0);
 615         if (ep == buf || (*ep != 0 && *ep != '\n' && *ep != ' '))
 616                 return -1;
 617         while (*ep == ' ' || *ep == '/' || *ep == '(')
 618                 ep++;
 619         *v2 = strtoull(ep, &ep2, 0);
 620         if (ep2 == ep || (*ep2 != 0 && *ep2 != '\n' && *ep2 != ' ' && *ep2 != ')')) {
 621                 *v2 = *v1;
 622                 return 1;
 623         }
 624         return 2;
 625 }
 626
 627 int sysfs_get_two(struct mdinfo *sra, struct mdinfo *dev,
 628                   char *name, unsigned long long *v1, unsigned long long *v2)
 629 {
 630         int n;
 631         int fd;
 632
 633         fd = sysfs_get_fd(sra, dev, name);
 634         if (fd < 0)
 635                 return -1;
 636         n = sysfs_fd_get_two(fd, v1, v2);
 637         close(fd);
 638         return n;
 639 }
 640
 641 int sysfs_fd_get_str(int fd, char *val, int size)
 642 {
 643         int n;
 644
 645         lseek(fd, 0, 0);
 646         n = read(fd, val, size);
 647         if (n <= 0 || n == size)
 648                 return -1;
 649         val[n] = 0;
 650         return n;
 651 }
 652
 653 int sysfs_get_str(struct mdinfo *sra, struct mdinfo *dev,
 654                        char *name, char *val, int size)
 655 {
 656         int n;
 657         int fd;
 658
 659         fd = sysfs_get_fd(sra, dev, name);
 660         if (fd < 0)
 661                 return -1;
 662         n = sysfs_fd_get_str(fd, val, size);
 663         close(fd);
 664         return n;
 665 }
 666
 667 int sysfs_set_safemode(struct mdinfo *sra, unsigned long ms)
 668 {
 669         unsigned long sec;
 670         unsigned long msec;
 671         char delay[30];
 672
 673         sec = ms / 1000;
 674         msec = ms % 1000;
 675
 676         sprintf(delay, "%ld.%03ld\n", sec, msec);
 677         /*             this '\n' ^ needed for kernels older than 2.6.28 */
 678         return sysfs_set_str(sra, NULL, "safe_mode_delay", delay);
 679 }
 680
 681 int sysfs_set_array(struct mdinfo *info)
 682 {
 683         int rv = 0;
 684         char ver[100];
 685         int raid_disks = info->array.raid_disks;
 686
 687         ver[0] = 0;
 688         if (info->array.major_version == -1 &&
 689             info->array.minor_version == -2) {
 690                 char buf[SYSFS_MAX_BUF_SIZE];
 691
 692                 strcat(strcpy(ver, "external:"), info->text_version);
 693
 694                 /* meta version might already be set if we are setting
 695                  * new geometry for a reshape.  In that case we don't
 696                  * want to over-write the 'readonly' flag that is
 697                  * stored in the metadata version.  So read the current
 698                  * version first, and preserve the flag
 699                  */
 700                 if (sysfs_get_str(info, NULL, "metadata_version",
 701                                   buf, sizeof(buf)) > 0)
 702                         if (strlen(buf) >= 9 && buf[9] == '-')
 703                                 ver[9] = '-';
 704
 705                 if (sysfs_set_str(info, NULL, "metadata_version", ver) < 0) {
 706                         pr_err("This kernel does not support external metadata.\n");
 707                         return 1;
 708                 }
 709         }
 710         if (info->array.level < 0)
 711                 return 0; /* FIXME */
 712         rv |= sysfs_set_str(info, NULL, "level",
 713                             map_num_s(pers, info->array.level));
 714         if (info->reshape_active && info->delta_disks != UnSet)
 715                 raid_disks -= info->delta_disks;
 716         rv |= sysfs_set_num(info, NULL, "raid_disks", raid_disks);
 717         rv |= sysfs_set_num(info, NULL, "chunk_size", info->array.chunk_size);
 718         rv |= sysfs_set_num(info, NULL, "layout", info->array.layout);
 719         rv |= sysfs_set_num(info, NULL, "component_size", info->component_size/2);
 720         if (info->custom_array_size) {
 721                 int rc;
 722
 723                 rc = sysfs_set_num(info, NULL, "array_size",
 724                                    info->custom_array_size/2);
 725                 if (rc && errno == ENOENT) {
 726                         pr_err("This kernel does not have the md/array_size attribute, the array may be larger than expected\n");
 727                         rc = 0;
 728                 }
 729                 rv |= rc;
 730         }
 731
 732         if (info->array.level > 0)
 733                 rv |= sysfs_set_num(info, NULL, "resync_start", info->resync_start);
 734
 735         if (info->reshape_active) {
 736                 rv |= sysfs_set_num(info, NULL, "reshape_position",
 737                                     info->reshape_progress);
 738                 rv |= sysfs_set_num(info, NULL, "chunk_size", info->new_chunk);
 739                 rv |= sysfs_set_num(info, NULL, "layout", info->new_layout);
 740                 rv |= sysfs_set_num(info, NULL, "raid_disks",
 741                                     info->array.raid_disks);
 742                 /* We don't set 'new_level' here.  That can only happen
 743                  * once the reshape completes.
 744                  */
 745         }
 746
 747         if (info->consistency_policy == CONSISTENCY_POLICY_PPL) {
 748                 char *policy = map_num_s(consistency_policies,
 749                                             info->consistency_policy);
 750
 751                 if (sysfs_set_str(info, NULL, "consistency_policy", policy)) {
 752                         pr_err("This kernel does not support PPL. Falling back to consistency-policy=resync.\n");
 753                         info->consistency_policy = CONSISTENCY_POLICY_RESYNC;
 754                 }
 755         }
 756
 757         return rv;
 758 }
 759
 760 int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd, int resume)
 761 {
 762         char dv[PATH_MAX];
 763         char nm[PATH_MAX];
 764         char *dname;
 765         int rv;
 766         int i;
 767
 768         sprintf(dv, "%d:%d", sd->disk.major, sd->disk.minor);
 769         rv = sysfs_set_str(sra, NULL, "new_dev", dv);
 770         if (rv)
 771                 return rv;
 772
 773         memset(nm, 0, sizeof(nm));
 774         dname = devid2kname(makedev(sd->disk.major, sd->disk.minor));
 775         strcpy(sd->sys_name, "dev-");
 776         strcpy(sd->sys_name+4, dname);
 777
 778         /* test write to see if 'recovery_start' is available */
 779         if (resume && sd->recovery_start < MaxSector &&
 780             sysfs_set_num(sra, sd, "recovery_start", 0)) {
 781                 sysfs_set_str(sra, sd, "state", "remove");
 782                 return -1;
 783         }
 784
 785         rv = sysfs_set_num(sra, sd, "offset", sd->data_offset);
 786         rv |= sysfs_set_num(sra, sd, "size", (sd->component_size+1) / 2);
 787         if (!is_container(sra->array.level)) {
 788                 if (sra->consistency_policy == CONSISTENCY_POLICY_PPL) {
 789                         rv |= sysfs_set_num(sra, sd, "ppl_sector", sd->ppl_sector);
 790                         rv |= sysfs_set_num(sra, sd, "ppl_size", sd->ppl_size);
 791                 }
 792                 if (sd->recovery_start == MaxSector)
 793                         /* This can correctly fail if array isn't started,
 794                          * yet, so just ignore status for now.
 795                          */
 796                         sysfs_set_str(sra, sd, "state", "insync");
 797                 if (sd->disk.raid_disk >= 0)
 798                         rv |= sysfs_set_num(sra, sd, "slot", sd->disk.raid_disk);
 799                 if (resume)
 800                         sysfs_set_num(sra, sd, "recovery_start", sd->recovery_start);
 801         }
 802         if (sd->bb.supported) {
 803                 if (sysfs_set_str(sra, sd, "state", "external_bbl")) {
 804                         /*
 805                          * backward compatibility - if kernel doesn't support
 806                          * bad blocks for external metadata, let it continue
 807                          * as long as there are none known so far
 808                          */
 809                         if (sd->bb.count) {
 810                                 pr_err("The kernel has no support for bad blocks in external metadata\n");
 811                                 return -1;
 812                         }
 813                 }
 814
 815                 for (i = 0; i < sd->bb.count; i++) {
 816                         char s[30];
 817                         const struct md_bb_entry *entry = &sd->bb.entries[i];
 818
 819                         snprintf(s, sizeof(s) - 1, "%llu %d\n", entry->sector,
 820                                  entry->length);
 821                         rv |= sysfs_set_str(sra, sd, "bad_blocks", s);
 822                 }
 823         }
 824         return rv;
 825 }
 826
 827 int sysfs_disk_to_scsi_id(int fd, __u32 *id)
 828 {
 829         /* from an open block device, try to retrieve it scsi_id */
 830         struct stat st;
 831         char path[256];
 832         DIR *dir;
 833         struct dirent *de;
 834         int host, bus, target, lun;
 835
 836         if (fstat(fd, &st))
 837                 return 1;
 838
 839         snprintf(path, sizeof(path), "/sys/dev/block/%d:%d/device/scsi_device",
 840                  major(st.st_rdev), minor(st.st_rdev));
 841
 842         dir = opendir(path);
 843         if (!dir)
 844                 return 1;
 845
 846         for (de = readdir(dir); de; de = readdir(dir)) {
 847                 int count;
 848
 849                 if (de->d_type != DT_DIR)
 850                         continue;
 851
 852                 count = sscanf(de->d_name, "%d:%d:%d:%d", &host, &bus, &target, &lun);
 853                 if (count == 4)
 854                         break;
 855         }
 856         closedir(dir);
 857
 858         if (!de)
 859                 return 1;
 860
 861         *id = (host << 24) | (bus << 16) | (target << 8) | (lun << 0);
 862         return 0;
 863 }
 864
 865 int sysfs_unique_holder(char *devnm, long rdev)
 866 {
 867         /* Check that devnm is a holder of rdev,
 868          * and is the only holder.
 869          * we should be locked against races by
 870          * an O_EXCL on devnm
 871          * Return values:
 872          *  0 - not unique, not even a holder
 873          *  1 - unique, this is the only holder.
 874          *  2/3 - not unique, there is another holder
 875          * -1 - error, cannot find the holders
 876          */
 877         DIR *dir;
 878         struct dirent *de;
 879         char dirname[100];
 880         char l;
 881         int ret = 0;
 882         sprintf(dirname, "/sys/dev/block/%d:%d/holders",
 883                 major(rdev), minor(rdev));
 884         dir = opendir(dirname);
 885         if (!dir)
 886                 return -1;
 887         l = strlen(dirname);
 888         while ((de = readdir(dir)) != NULL) {
 889                 char buf[100];
 890                 char *sl;
 891                 int n;
 892
 893                 if (de->d_ino == 0)
 894                         continue;
 895                 if (de->d_name[0] == '.')
 896                         continue;
 897                 strcpy(dirname+l, "/");
 898                 strcat(dirname+l, de->d_name);
 899                 n = readlink(dirname, buf, sizeof(buf)-1);
 900                 if (n <= 0)
 901                         continue;
 902                 buf[n] = 0;
 903                 sl = strrchr(buf, '/');
 904                 if (!sl)
 905                         continue;
 906                 sl++;
 907
 908                 if (strcmp(devnm, sl) == 0)
 909                         ret |= 1;
 910                 else
 911                         ret |= 2;
 912         }
 913         closedir(dir);
 914         return ret;
 915 }
 916
 917 int sysfs_freeze_array(struct mdinfo *sra)
 918 {
 919         /* Try to freeze resync/rebuild on this array/container.
 920          * Return -1 if the array is busy,
 921          * return 0 if this kernel doesn't support 'frozen'
 922          * return 1 if it worked.
 923          */
 924         char buf[SYSFS_MAX_BUF_SIZE];
 925
 926         if (!sysfs_attribute_available(sra, NULL, "sync_action"))
 927                 return 1; /* no sync_action == frozen */
 928         if (sysfs_get_str(sra, NULL, "sync_action", buf, sizeof(buf)) <= 0)
 929                 return 0;
 930         if (strcmp(buf, "frozen\n") == 0)
 931                 /* Already frozen */
 932                 return 0;
 933         if (strcmp(buf, "idle\n") != 0 && strcmp(buf, "recover\n") != 0)
 934                 return -1;
 935         if (sysfs_set_str(sra, NULL, "sync_action", "frozen") < 0)
 936                 return 0;
 937         return 1;
 938 }
 939
 940 int sysfs_wait(int fd, int *msec)
 941 {
 942         /* Wait up to '*msec' for fd to have an exception condition.
 943          * if msec == NULL, wait indefinitely.
 944          */
 945         fd_set fds;
 946         int n;
 947         FD_ZERO(&fds);
 948         FD_SET(fd, &fds);
 949         if (msec == NULL)
 950                 n = select(fd+1, NULL, NULL, &fds, NULL);
 951         else if (*msec < 0)
 952                 n = 0;
 953         else {
 954                 struct timeval start, end, tv;
 955                 gettimeofday(&start, NULL);
 956                 if (*msec < 1000) {
 957                         tv.tv_sec = 0;
 958                         tv.tv_usec = (*msec)*1000;
 959                 } else {
 960                         tv.tv_sec = (*msec)/1000;
 961                         tv.tv_usec = 0;
 962                 }
 963                 n = select(fd+1, NULL, NULL, &fds, &tv);
 964                 gettimeofday(&end, NULL);
 965                 end.tv_sec -= start.tv_sec;
 966                 *msec -= (end.tv_sec * 1000 + end.tv_usec/1000
 967                           - start.tv_usec/1000) + 1;
 968         }
 969         return n;
 970 }
 971
 972 int sysfs_rules_apply_check(const struct mdinfo *sra,
 973                             const struct sysfs_entry *ent)
 974 {
 975         /* Check whether parameter is regular file,
 976          * exists and is under specified directory.
 977          */
 978         char fname[MAX_SYSFS_PATH_LEN];
 979         char dname[MAX_SYSFS_PATH_LEN];
 980         char resolved_path[PATH_MAX];
 981         char resolved_dir[PATH_MAX];
 982         int result;
 983
 984         if (sra == NULL || ent == NULL)
 985                 return -1;
 986
 987         result = snprintf(dname, MAX_SYSFS_PATH_LEN,
 988                           "/sys/block/%s/md/", sra->sys_name);
 989         if (result < 0 || result >= MAX_SYSFS_PATH_LEN)
 990                 return -1;
 991
 992         result = snprintf(fname, MAX_SYSFS_PATH_LEN,
 993                           "%s/%s", dname, ent->name);
 994         if (result < 0 || result >= MAX_SYSFS_PATH_LEN)
 995                 return -1;
 996
 997         if (realpath(fname, resolved_path) == NULL ||
 998             realpath(dname, resolved_dir) == NULL)
 999                 return -1;
1000
1001         if (strncmp(resolved_dir, resolved_path,
1002                     strnlen(resolved_dir, PATH_MAX)) != 0)
1003                 return -1;
1004
1005         return 0;
1006 }
1007
1008 static struct dev_sysfs_rule *sysfs_rules;
1009
1010 void sysfs_rules_apply(char *devnm, struct mdinfo *dev)
1011 {
1012         struct dev_sysfs_rule *rules = sysfs_rules;
1013
1014         while (rules) {
1015                 struct sysfs_entry *ent = rules->entry;
1016                 int match  = 0;
1017
1018                 if (!rules->uuid_set) {
1019                         if (rules->devname)
1020                                 match = strcmp(devnm, rules->devname) == 0;
1021                 } else {
1022                         match = memcmp(dev->uuid, rules->uuid,
1023                                        sizeof(int[4])) == 0;
1024                 }
1025
1026                 while (match && ent) {
1027                         if (sysfs_rules_apply_check(dev, ent) < 0)
1028                                 pr_err("SYSFS: failed to write '%s' to '%s'\n",
1029                                         ent->value, ent->name);
1030                         else
1031                                 sysfs_set_str(dev, NULL, ent->name, ent->value);
1032                         ent = ent->next;
1033                 }
1034                 rules = rules->next;
1035         }
1036 }
1037
1038 static void sysfs_rule_free(struct dev_sysfs_rule *rule)
1039 {
1040         struct sysfs_entry *entry;
1041
1042         while (rule) {
1043                 struct dev_sysfs_rule *tmp = rule->next;
1044
1045                 entry = rule->entry;
1046                 while (entry) {
1047                         struct sysfs_entry *tmp = entry->next;
1048
1049                         free(entry->name);
1050                         free(entry->value);
1051                         free(entry);
1052                         entry = tmp;
1053                 }
1054
1055                 if (rule->devname)
1056                         free(rule->devname);
1057                 free(rule);
1058                 rule = tmp;
1059         }
1060 }
1061
1062 void sysfsline(char *line)
1063 {
1064         struct dev_sysfs_rule *sr;
1065         char *w;
1066
1067         sr = xcalloc(1, sizeof(*sr));
1068         for (w = dl_next(line); w != line ; w = dl_next(w)) {
1069                 if (strncasecmp(w, "name=", 5) == 0) {
1070                         char *devname = w + 5;
1071
1072                         if (strncmp(devname, DEV_MD_DIR, DEV_MD_DIR_LEN) == 0) {
1073                                 if (sr->devname)
1074                                         pr_err("Only give one device per SYSFS line: %s\n",
1075                                                 devname);
1076                                 else
1077                                         sr->devname = xstrdup(devname);
1078                         } else {
1079                                 pr_err("%s is an invalid name for an md device - ignored.\n",
1080                                        devname);
1081                         }
1082                 } else if (strncasecmp(w, "uuid=", 5) == 0) {
1083                         char *uuid = w + 5;
1084
1085                         if (sr->uuid_set) {
1086                                 pr_err("Only give one uuid per SYSFS line: %s\n",
1087                                         uuid);
1088                         } else {
1089                                 if (parse_uuid(w + 5, sr->uuid) &&
1090                                     memcmp(sr->uuid, uuid_zero,
1091                                            sizeof(int[4])) != 0)
1092                                         sr->uuid_set = 1;
1093                                 else
1094                                         pr_err("Invalid uuid: %s\n", uuid);
1095                         }
1096                 } else {
1097                         struct sysfs_entry *prop;
1098
1099                         char *sep = strchr(w, '=');
1100
1101                         if (sep == NULL || *(sep + 1) == 0) {
1102                                 pr_err("Cannot parse \"%s\" - ignoring.\n", w);
1103                                 continue;
1104                         }
1105
1106                         prop = xmalloc(sizeof(*prop));
1107                         prop->value = xstrdup(sep + 1);
1108                         *sep = 0;
1109                         prop->name = xstrdup(w);
1110                         prop->next = sr->entry;
1111                         sr->entry = prop;
1112                 }
1113         }
1114
1115         if (!sr->devname && !sr->uuid_set) {
1116                 pr_err("Device name not found in sysfs config entry - ignoring.\n");
1117                 sysfs_rule_free(sr);
1118                 return;
1119         }
1120
1121         sr->next = sysfs_rules;
1122         sysfs_rules = sr;
1123 }
1124
1125 /**
1126  * sysfs_is_libata_allow_tpm_enabled() - check if libata allow_tmp is enabled.
1127  * @verbose: verbose flag.
1128  *
1129  * Check if libata allow_tmp flag is set, this is required for SATA Opal Security commands to work.
1130  *
1131  * Return: true if allow_tpm enable, false otherwise.
1132  */
1133 bool sysfs_is_libata_allow_tpm_enabled(const int verbose)
1134 {
1135         const char *path = "/sys/module/libata/parameters/allow_tpm";
1136         const char *expected_value = "1";
1137         int fd = open(path, O_RDONLY);
1138         char buf[3];
1139
1140         if (!is_fd_valid(fd)) {
1141                 pr_vrb("Failed open file descriptor to %s. Cannot check libata allow_tpm param.\n",
1142                        path);
1143                 return false;
1144         }
1145
1146         sysfs_fd_get_str(fd, buf, sizeof(buf));
1147         close(fd);
1148
1149         if (strncmp(buf, expected_value, 1) == 0)
1150                 return true;
1151         return false;
1152 }