1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 /* 22 * Copyright 2007 Sun Microsystems, Inc. All rights reserved. 23 * Use is subject to license terms. 24 */ 25 26 #pragma ident "%Z%%M% %I% %E% SMI" 27 28 /* 29 * Pool import support functions. 30 * 31 * To import a pool, we rely on reading the configuration information from the 32 * ZFS label of each device. If we successfully read the label, then we 33 * organize the configuration information in the following hierarchy: 34 * 35 * pool guid -> toplevel vdev guid -> label txg 36 * 37 * Duplicate entries matching this same tuple will be discarded. Once we have 38 * examined every device, we pick the best label txg config for each toplevel 39 * vdev. We then arrange these toplevel vdevs into a complete pool config, and 40 * update any paths that have changed. Finally, we attempt to import the pool 41 * using our derived config, and record the results. 42 */ 43 44 #include <devid.h> 45 #include <dirent.h> 46 #include <errno.h> 47 #include <libintl.h> 48 #include <stdlib.h> 49 #include <string.h> 50 #include <sys/stat.h> 51 #include <unistd.h> 52 #include <fcntl.h> 53 54 #include <sys/vdev_impl.h> 55 56 #include "libzfs.h" 57 #include "libzfs_impl.h" 58 59 /* 60 * Intermediate structures used to gather configuration information. 61 */ 62 typedef struct config_entry { 63 uint64_t ce_txg; 64 nvlist_t *ce_config; 65 struct config_entry *ce_next; 66 } config_entry_t; 67 68 typedef struct vdev_entry { 69 uint64_t ve_guid; 70 config_entry_t *ve_configs; 71 struct vdev_entry *ve_next; 72 } vdev_entry_t; 73 74 typedef struct pool_entry { 75 uint64_t pe_guid; 76 vdev_entry_t *pe_vdevs; 77 struct pool_entry *pe_next; 78 } pool_entry_t; 79 80 typedef struct name_entry { 81 char *ne_name; 82 uint64_t ne_guid; 83 struct name_entry *ne_next; 84 } name_entry_t; 85 86 typedef struct pool_list { 87 pool_entry_t *pools; 88 name_entry_t *names; 89 } pool_list_t; 90 91 static char * 92 get_devid(const char *path) 93 { 94 int fd; 95 ddi_devid_t devid; 96 char *minor, *ret; 97 98 if ((fd = open(path, O_RDONLY)) < 0) 99 return (NULL); 100 101 minor = NULL; 102 ret = NULL; 103 if (devid_get(fd, &devid) == 0) { 104 if (devid_get_minor_name(fd, &minor) == 0) 105 ret = devid_str_encode(devid, minor); 106 if (minor != NULL) 107 devid_str_free(minor); 108 devid_free(devid); 109 } 110 (void) close(fd); 111 112 return (ret); 113 } 114 115 116 /* 117 * Go through and fix up any path and/or devid information for the given vdev 118 * configuration. 119 */ 120 static int 121 fix_paths(nvlist_t *nv, name_entry_t *names) 122 { 123 nvlist_t **child; 124 uint_t c, children; 125 uint64_t guid; 126 name_entry_t *ne, *best; 127 char *path, *devid; 128 int matched; 129 130 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, 131 &child, &children) == 0) { 132 for (c = 0; c < children; c++) 133 if (fix_paths(child[c], names) != 0) 134 return (-1); 135 return (0); 136 } 137 138 /* 139 * This is a leaf (file or disk) vdev. In either case, go through 140 * the name list and see if we find a matching guid. If so, replace 141 * the path and see if we can calculate a new devid. 142 * 143 * There may be multiple names associated with a particular guid, in 144 * which case we have overlapping slices or multiple paths to the same 145 * disk. If this is the case, then we want to pick the path that is 146 * the most similar to the original, where "most similar" is the number 147 * of matching characters starting from the end of the path. This will 148 * preserve slice numbers even if the disks have been reorganized, and 149 * will also catch preferred disk names if multiple paths exist. 150 */ 151 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0); 152 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) != 0) 153 path = NULL; 154 155 matched = 0; 156 best = NULL; 157 for (ne = names; ne != NULL; ne = ne->ne_next) { 158 if (ne->ne_guid == guid) { 159 const char *src, *dst; 160 int count; 161 162 if (path == NULL) { 163 best = ne; 164 break; 165 } 166 167 src = ne->ne_name + strlen(ne->ne_name) - 1; 168 dst = path + strlen(path) - 1; 169 for (count = 0; src >= ne->ne_name && dst >= path; 170 src--, dst--, count++) 171 if (*src != *dst) 172 break; 173 174 /* 175 * At this point, 'count' is the number of characters 176 * matched from the end. 177 */ 178 if (count > matched || best == NULL) { 179 best = ne; 180 matched = count; 181 } 182 } 183 } 184 185 if (best == NULL) 186 return (0); 187 188 if (nvlist_add_string(nv, ZPOOL_CONFIG_PATH, best->ne_name) != 0) 189 return (-1); 190 191 if ((devid = get_devid(best->ne_name)) == NULL) { 192 (void) nvlist_remove_all(nv, ZPOOL_CONFIG_DEVID); 193 } else { 194 if (nvlist_add_string(nv, ZPOOL_CONFIG_DEVID, devid) != 0) 195 return (-1); 196 devid_str_free(devid); 197 } 198 199 return (0); 200 } 201 202 /* 203 * Add the given configuration to the list of known devices. 204 */ 205 static int 206 add_config(libzfs_handle_t *hdl, pool_list_t *pl, const char *path, 207 nvlist_t *config) 208 { 209 uint64_t pool_guid, vdev_guid, top_guid, txg, state; 210 pool_entry_t *pe; 211 vdev_entry_t *ve; 212 config_entry_t *ce; 213 name_entry_t *ne; 214 215 /* 216 * If this is a hot spare not currently in use, add it to the list of 217 * names to translate, but don't do anything else. 218 */ 219 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE, 220 &state) == 0 && state == POOL_STATE_SPARE && 221 nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, &vdev_guid) == 0) { 222 if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL) 223 return (-1); 224 225 if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) { 226 free(ne); 227 return (-1); 228 } 229 ne->ne_guid = vdev_guid; 230 ne->ne_next = pl->names; 231 pl->names = ne; 232 return (0); 233 } 234 235 /* 236 * If we have a valid config but cannot read any of these fields, then 237 * it means we have a half-initialized label. In vdev_label_init() 238 * we write a label with txg == 0 so that we can identify the device 239 * in case the user refers to the same disk later on. If we fail to 240 * create the pool, we'll be left with a label in this state 241 * which should not be considered part of a valid pool. 242 */ 243 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, 244 &pool_guid) != 0 || 245 nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, 246 &vdev_guid) != 0 || 247 nvlist_lookup_uint64(config, ZPOOL_CONFIG_TOP_GUID, 248 &top_guid) != 0 || 249 nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG, 250 &txg) != 0 || txg == 0) { 251 nvlist_free(config); 252 return (0); 253 } 254 255 /* 256 * First, see if we know about this pool. If not, then add it to the 257 * list of known pools. 258 */ 259 for (pe = pl->pools; pe != NULL; pe = pe->pe_next) { 260 if (pe->pe_guid == pool_guid) 261 break; 262 } 263 264 if (pe == NULL) { 265 if ((pe = zfs_alloc(hdl, sizeof (pool_entry_t))) == NULL) { 266 nvlist_free(config); 267 return (-1); 268 } 269 pe->pe_guid = pool_guid; 270 pe->pe_next = pl->pools; 271 pl->pools = pe; 272 } 273 274 /* 275 * Second, see if we know about this toplevel vdev. Add it if its 276 * missing. 277 */ 278 for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) { 279 if (ve->ve_guid == top_guid) 280 break; 281 } 282 283 if (ve == NULL) { 284 if ((ve = zfs_alloc(hdl, sizeof (vdev_entry_t))) == NULL) { 285 nvlist_free(config); 286 return (-1); 287 } 288 ve->ve_guid = top_guid; 289 ve->ve_next = pe->pe_vdevs; 290 pe->pe_vdevs = ve; 291 } 292 293 /* 294 * Third, see if we have a config with a matching transaction group. If 295 * so, then we do nothing. Otherwise, add it to the list of known 296 * configs. 297 */ 298 for (ce = ve->ve_configs; ce != NULL; ce = ce->ce_next) { 299 if (ce->ce_txg == txg) 300 break; 301 } 302 303 if (ce == NULL) { 304 if ((ce = zfs_alloc(hdl, sizeof (config_entry_t))) == NULL) { 305 nvlist_free(config); 306 return (-1); 307 } 308 ce->ce_txg = txg; 309 ce->ce_config = config; 310 ce->ce_next = ve->ve_configs; 311 ve->ve_configs = ce; 312 } else { 313 nvlist_free(config); 314 } 315 316 /* 317 * At this point we've successfully added our config to the list of 318 * known configs. The last thing to do is add the vdev guid -> path 319 * mappings so that we can fix up the configuration as necessary before 320 * doing the import. 321 */ 322 if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL) 323 return (-1); 324 325 if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) { 326 free(ne); 327 return (-1); 328 } 329 330 ne->ne_guid = vdev_guid; 331 ne->ne_next = pl->names; 332 pl->names = ne; 333 334 return (0); 335 } 336 337 /* 338 * Returns true if the named pool matches the given GUID. 339 */ 340 static int 341 pool_active(libzfs_handle_t *hdl, const char *name, uint64_t guid, 342 boolean_t *isactive) 343 { 344 zpool_handle_t *zhp; 345 uint64_t theguid; 346 347 if (zpool_open_silent(hdl, name, &zhp) != 0) 348 return (-1); 349 350 if (zhp == NULL) { 351 *isactive = B_FALSE; 352 return (0); 353 } 354 355 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_POOL_GUID, 356 &theguid) == 0); 357 358 zpool_close(zhp); 359 360 *isactive = (theguid == guid); 361 return (0); 362 } 363 364 /* 365 * Convert our list of pools into the definitive set of configurations. We 366 * start by picking the best config for each toplevel vdev. Once that's done, 367 * we assemble the toplevel vdevs into a full config for the pool. We make a 368 * pass to fix up any incorrect paths, and then add it to the main list to 369 * return to the user. 370 */ 371 static nvlist_t * 372 get_configs(libzfs_handle_t *hdl, pool_list_t *pl) 373 { 374 pool_entry_t *pe; 375 vdev_entry_t *ve; 376 config_entry_t *ce; 377 nvlist_t *ret = NULL, *config = NULL, *tmp, *nvtop, *nvroot; 378 nvlist_t **spares; 379 uint_t i, nspares; 380 boolean_t config_seen; 381 uint64_t best_txg; 382 char *name, *hostname; 383 zfs_cmd_t zc = { 0 }; 384 uint64_t version, guid; 385 size_t len; 386 int err; 387 uint_t children = 0; 388 nvlist_t **child = NULL; 389 uint_t c; 390 boolean_t isactive; 391 uint64_t hostid; 392 393 if (nvlist_alloc(&ret, 0, 0) != 0) 394 goto nomem; 395 396 for (pe = pl->pools; pe != NULL; pe = pe->pe_next) { 397 uint64_t id; 398 399 if (nvlist_alloc(&config, NV_UNIQUE_NAME, 0) != 0) 400 goto nomem; 401 config_seen = B_FALSE; 402 403 /* 404 * Iterate over all toplevel vdevs. Grab the pool configuration 405 * from the first one we find, and then go through the rest and 406 * add them as necessary to the 'vdevs' member of the config. 407 */ 408 for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) { 409 410 /* 411 * Determine the best configuration for this vdev by 412 * selecting the config with the latest transaction 413 * group. 414 */ 415 best_txg = 0; 416 for (ce = ve->ve_configs; ce != NULL; 417 ce = ce->ce_next) { 418 419 if (ce->ce_txg > best_txg) { 420 tmp = ce->ce_config; 421 best_txg = ce->ce_txg; 422 } 423 } 424 425 if (!config_seen) { 426 /* 427 * Copy the relevant pieces of data to the pool 428 * configuration: 429 * 430 * version 431 * pool guid 432 * name 433 * pool state 434 * hostid (if available) 435 * hostname (if available) 436 */ 437 uint64_t state; 438 439 verify(nvlist_lookup_uint64(tmp, 440 ZPOOL_CONFIG_VERSION, &version) == 0); 441 if (nvlist_add_uint64(config, 442 ZPOOL_CONFIG_VERSION, version) != 0) 443 goto nomem; 444 verify(nvlist_lookup_uint64(tmp, 445 ZPOOL_CONFIG_POOL_GUID, &guid) == 0); 446 if (nvlist_add_uint64(config, 447 ZPOOL_CONFIG_POOL_GUID, guid) != 0) 448 goto nomem; 449 verify(nvlist_lookup_string(tmp, 450 ZPOOL_CONFIG_POOL_NAME, &name) == 0); 451 if (nvlist_add_string(config, 452 ZPOOL_CONFIG_POOL_NAME, name) != 0) 453 goto nomem; 454 verify(nvlist_lookup_uint64(tmp, 455 ZPOOL_CONFIG_POOL_STATE, &state) == 0); 456 if (nvlist_add_uint64(config, 457 ZPOOL_CONFIG_POOL_STATE, state) != 0) 458 goto nomem; 459 hostid = 0; 460 if (nvlist_lookup_uint64(tmp, 461 ZPOOL_CONFIG_HOSTID, &hostid) == 0) { 462 if (nvlist_add_uint64(config, 463 ZPOOL_CONFIG_HOSTID, hostid) != 0) 464 goto nomem; 465 verify(nvlist_lookup_string(tmp, 466 ZPOOL_CONFIG_HOSTNAME, 467 &hostname) == 0); 468 if (nvlist_add_string(config, 469 ZPOOL_CONFIG_HOSTNAME, 470 hostname) != 0) 471 goto nomem; 472 } 473 474 config_seen = B_TRUE; 475 } 476 477 /* 478 * Add this top-level vdev to the child array. 479 */ 480 verify(nvlist_lookup_nvlist(tmp, 481 ZPOOL_CONFIG_VDEV_TREE, &nvtop) == 0); 482 verify(nvlist_lookup_uint64(nvtop, ZPOOL_CONFIG_ID, 483 &id) == 0); 484 if (id >= children) { 485 nvlist_t **newchild; 486 487 newchild = zfs_alloc(hdl, (id + 1) * 488 sizeof (nvlist_t *)); 489 if (newchild == NULL) 490 goto nomem; 491 492 for (c = 0; c < children; c++) 493 newchild[c] = child[c]; 494 495 free(child); 496 child = newchild; 497 children = id + 1; 498 } 499 if (nvlist_dup(nvtop, &child[id], 0) != 0) 500 goto nomem; 501 502 } 503 504 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, 505 &guid) == 0); 506 507 /* 508 * Look for any missing top-level vdevs. If this is the case, 509 * create a faked up 'missing' vdev as a placeholder. We cannot 510 * simply compress the child array, because the kernel performs 511 * certain checks to make sure the vdev IDs match their location 512 * in the configuration. 513 */ 514 for (c = 0; c < children; c++) 515 if (child[c] == NULL) { 516 nvlist_t *missing; 517 if (nvlist_alloc(&missing, NV_UNIQUE_NAME, 518 0) != 0) 519 goto nomem; 520 if (nvlist_add_string(missing, 521 ZPOOL_CONFIG_TYPE, 522 VDEV_TYPE_MISSING) != 0 || 523 nvlist_add_uint64(missing, 524 ZPOOL_CONFIG_ID, c) != 0 || 525 nvlist_add_uint64(missing, 526 ZPOOL_CONFIG_GUID, 0ULL) != 0) { 527 nvlist_free(missing); 528 goto nomem; 529 } 530 child[c] = missing; 531 } 532 533 /* 534 * Put all of this pool's top-level vdevs into a root vdev. 535 */ 536 if (nvlist_alloc(&nvroot, NV_UNIQUE_NAME, 0) != 0) 537 goto nomem; 538 if (nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE, 539 VDEV_TYPE_ROOT) != 0 || 540 nvlist_add_uint64(nvroot, ZPOOL_CONFIG_ID, 0ULL) != 0 || 541 nvlist_add_uint64(nvroot, ZPOOL_CONFIG_GUID, guid) != 0 || 542 nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, 543 child, children) != 0) { 544 nvlist_free(nvroot); 545 goto nomem; 546 } 547 548 for (c = 0; c < children; c++) 549 nvlist_free(child[c]); 550 free(child); 551 children = 0; 552 child = NULL; 553 554 /* 555 * Go through and fix up any paths and/or devids based on our 556 * known list of vdev GUID -> path mappings. 557 */ 558 if (fix_paths(nvroot, pl->names) != 0) { 559 nvlist_free(nvroot); 560 goto nomem; 561 } 562 563 /* 564 * Add the root vdev to this pool's configuration. 565 */ 566 if (nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 567 nvroot) != 0) { 568 nvlist_free(nvroot); 569 goto nomem; 570 } 571 nvlist_free(nvroot); 572 573 /* 574 * Determine if this pool is currently active, in which case we 575 * can't actually import it. 576 */ 577 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, 578 &name) == 0); 579 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, 580 &guid) == 0); 581 582 if (pool_active(hdl, name, guid, &isactive) != 0) 583 goto error; 584 585 if (isactive) { 586 nvlist_free(config); 587 config = NULL; 588 continue; 589 } 590 591 /* 592 * Try to do the import in order to get vdev state. 593 */ 594 if (zcmd_write_src_nvlist(hdl, &zc, config, &len) != 0) 595 goto error; 596 597 nvlist_free(config); 598 config = NULL; 599 600 if (zcmd_alloc_dst_nvlist(hdl, &zc, len * 2) != 0) { 601 zcmd_free_nvlists(&zc); 602 goto error; 603 } 604 605 while ((err = ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_TRYIMPORT, 606 &zc)) != 0 && errno == ENOMEM) { 607 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) { 608 zcmd_free_nvlists(&zc); 609 goto error; 610 } 611 } 612 613 if (err) { 614 (void) zpool_standard_error(hdl, errno, 615 dgettext(TEXT_DOMAIN, "cannot discover pools")); 616 zcmd_free_nvlists(&zc); 617 goto error; 618 } 619 620 if (zcmd_read_dst_nvlist(hdl, &zc, &config) != 0) { 621 zcmd_free_nvlists(&zc); 622 goto error; 623 } 624 625 zcmd_free_nvlists(&zc); 626 627 /* 628 * Go through and update the paths for spares, now that we have 629 * them. 630 */ 631 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 632 &nvroot) == 0); 633 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, 634 &spares, &nspares) == 0) { 635 for (i = 0; i < nspares; i++) { 636 if (fix_paths(spares[i], pl->names) != 0) 637 goto nomem; 638 } 639 } 640 641 /* 642 * Restore the original information read from the actual label. 643 */ 644 (void) nvlist_remove(config, ZPOOL_CONFIG_HOSTID, 645 DATA_TYPE_UINT64); 646 (void) nvlist_remove(config, ZPOOL_CONFIG_HOSTNAME, 647 DATA_TYPE_STRING); 648 if (hostid != 0) { 649 verify(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID, 650 hostid) == 0); 651 verify(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME, 652 hostname) == 0); 653 } 654 655 /* 656 * Add this pool to the list of configs. 657 */ 658 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, 659 &name) == 0); 660 if (nvlist_add_nvlist(ret, name, config) != 0) 661 goto nomem; 662 663 nvlist_free(config); 664 config = NULL; 665 } 666 667 return (ret); 668 669 nomem: 670 (void) no_memory(hdl); 671 error: 672 nvlist_free(config); 673 nvlist_free(ret); 674 for (c = 0; c < children; c++) 675 nvlist_free(child[c]); 676 free(child); 677 678 return (NULL); 679 } 680 681 /* 682 * Return the offset of the given label. 683 */ 684 static uint64_t 685 label_offset(size_t size, int l) 686 { 687 return (l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ? 688 0 : size - VDEV_LABELS * sizeof (vdev_label_t))); 689 } 690 691 /* 692 * Given a file descriptor, read the label information and return an nvlist 693 * describing the configuration, if there is one. 694 */ 695 int 696 zpool_read_label(int fd, nvlist_t **config) 697 { 698 struct stat64 statbuf; 699 int l; 700 vdev_label_t *label; 701 uint64_t state, txg; 702 703 *config = NULL; 704 705 if (fstat64(fd, &statbuf) == -1) 706 return (0); 707 708 if ((label = malloc(sizeof (vdev_label_t))) == NULL) 709 return (-1); 710 711 for (l = 0; l < VDEV_LABELS; l++) { 712 if (pread(fd, label, sizeof (vdev_label_t), 713 label_offset(statbuf.st_size, l)) != sizeof (vdev_label_t)) 714 continue; 715 716 if (nvlist_unpack(label->vl_vdev_phys.vp_nvlist, 717 sizeof (label->vl_vdev_phys.vp_nvlist), config, 0) != 0) 718 continue; 719 720 if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE, 721 &state) != 0 || state > POOL_STATE_SPARE) { 722 nvlist_free(*config); 723 continue; 724 } 725 726 if (state != POOL_STATE_SPARE && 727 (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_TXG, 728 &txg) != 0 || txg == 0)) { 729 nvlist_free(*config); 730 continue; 731 } 732 733 free(label); 734 return (0); 735 } 736 737 free(label); 738 *config = NULL; 739 return (0); 740 } 741 742 /* 743 * Given a list of directories to search, find all pools stored on disk. This 744 * includes partial pools which are not available to import. If no args are 745 * given (argc is 0), then the default directory (/dev/dsk) is searched. 746 */ 747 nvlist_t * 748 zpool_find_import(libzfs_handle_t *hdl, int argc, char **argv) 749 { 750 int i; 751 DIR *dirp = NULL; 752 struct dirent64 *dp; 753 char path[MAXPATHLEN]; 754 struct stat64 statbuf; 755 nvlist_t *ret = NULL, *config; 756 static char *default_dir = "/dev/dsk"; 757 int fd; 758 pool_list_t pools = { 0 }; 759 pool_entry_t *pe, *penext; 760 vdev_entry_t *ve, *venext; 761 config_entry_t *ce, *cenext; 762 name_entry_t *ne, *nenext; 763 764 765 if (argc == 0) { 766 argc = 1; 767 argv = &default_dir; 768 } 769 770 /* 771 * Go through and read the label configuration information from every 772 * possible device, organizing the information according to pool GUID 773 * and toplevel GUID. 774 */ 775 for (i = 0; i < argc; i++) { 776 if (argv[i][0] != '/') { 777 (void) zfs_error_fmt(hdl, EZFS_BADPATH, 778 dgettext(TEXT_DOMAIN, "cannot open '%s'"), 779 argv[i]); 780 goto error; 781 } 782 783 if ((dirp = opendir(argv[i])) == NULL) { 784 zfs_error_aux(hdl, strerror(errno)); 785 (void) zfs_error_fmt(hdl, EZFS_BADPATH, 786 dgettext(TEXT_DOMAIN, "cannot open '%s'"), 787 argv[i]); 788 goto error; 789 } 790 791 /* 792 * This is not MT-safe, but we have no MT consumers of libzfs 793 */ 794 while ((dp = readdir64(dirp)) != NULL) { 795 796 (void) snprintf(path, sizeof (path), "%s/%s", 797 argv[i], dp->d_name); 798 799 if (stat64(path, &statbuf) != 0) 800 continue; 801 802 /* 803 * Ignore directories (which includes "." and ".."). 804 */ 805 if (S_ISDIR(statbuf.st_mode)) 806 continue; 807 808 /* 809 * Ignore special (non-character or non-block) files. 810 */ 811 if (!S_ISREG(statbuf.st_mode) && 812 !S_ISBLK(statbuf.st_mode)) 813 continue; 814 815 if ((fd = open64(path, O_RDONLY)) < 0) 816 continue; 817 818 if ((zpool_read_label(fd, &config)) != 0) { 819 (void) close(fd); 820 (void) no_memory(hdl); 821 goto error; 822 } 823 824 (void) close(fd); 825 826 if (config != NULL) 827 if (add_config(hdl, &pools, path, config) != 0) 828 goto error; 829 } 830 831 (void) closedir(dirp); 832 dirp = NULL; 833 } 834 835 ret = get_configs(hdl, &pools); 836 837 error: 838 for (pe = pools.pools; pe != NULL; pe = penext) { 839 penext = pe->pe_next; 840 for (ve = pe->pe_vdevs; ve != NULL; ve = venext) { 841 venext = ve->ve_next; 842 for (ce = ve->ve_configs; ce != NULL; ce = cenext) { 843 cenext = ce->ce_next; 844 if (ce->ce_config) 845 nvlist_free(ce->ce_config); 846 free(ce); 847 } 848 free(ve); 849 } 850 free(pe); 851 } 852 853 for (ne = pools.names; ne != NULL; ne = nenext) { 854 nenext = ne->ne_next; 855 if (ne->ne_name) 856 free(ne->ne_name); 857 free(ne); 858 } 859 860 if (dirp) 861 (void) closedir(dirp); 862 863 return (ret); 864 } 865 866 boolean_t 867 find_guid(nvlist_t *nv, uint64_t guid) 868 { 869 uint64_t tmp; 870 nvlist_t **child; 871 uint_t c, children; 872 873 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &tmp) == 0); 874 if (tmp == guid) 875 return (B_TRUE); 876 877 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, 878 &child, &children) == 0) { 879 for (c = 0; c < children; c++) 880 if (find_guid(child[c], guid)) 881 return (B_TRUE); 882 } 883 884 return (B_FALSE); 885 } 886 887 typedef struct spare_cbdata { 888 uint64_t cb_guid; 889 zpool_handle_t *cb_zhp; 890 } spare_cbdata_t; 891 892 static int 893 find_spare(zpool_handle_t *zhp, void *data) 894 { 895 spare_cbdata_t *cbp = data; 896 nvlist_t **spares; 897 uint_t i, nspares; 898 uint64_t guid; 899 nvlist_t *nvroot; 900 901 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE, 902 &nvroot) == 0); 903 904 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, 905 &spares, &nspares) == 0) { 906 for (i = 0; i < nspares; i++) { 907 verify(nvlist_lookup_uint64(spares[i], 908 ZPOOL_CONFIG_GUID, &guid) == 0); 909 if (guid == cbp->cb_guid) { 910 cbp->cb_zhp = zhp; 911 return (1); 912 } 913 } 914 } 915 916 zpool_close(zhp); 917 return (0); 918 } 919 920 /* 921 * Determines if the pool is in use. If so, it returns true and the state of 922 * the pool as well as the name of the pool. Both strings are allocated and 923 * must be freed by the caller. 924 */ 925 int 926 zpool_in_use(libzfs_handle_t *hdl, int fd, pool_state_t *state, char **namestr, 927 boolean_t *inuse) 928 { 929 nvlist_t *config; 930 char *name; 931 boolean_t ret; 932 uint64_t guid, vdev_guid; 933 zpool_handle_t *zhp; 934 nvlist_t *pool_config; 935 uint64_t stateval, isspare; 936 spare_cbdata_t cb = { 0 }; 937 boolean_t isactive; 938 939 *inuse = B_FALSE; 940 941 if (zpool_read_label(fd, &config) != 0) { 942 (void) no_memory(hdl); 943 return (-1); 944 } 945 946 if (config == NULL) 947 return (0); 948 949 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE, 950 &stateval) == 0); 951 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, 952 &vdev_guid) == 0); 953 954 if (stateval != POOL_STATE_SPARE) { 955 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, 956 &name) == 0); 957 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, 958 &guid) == 0); 959 } 960 961 switch (stateval) { 962 case POOL_STATE_EXPORTED: 963 ret = B_TRUE; 964 break; 965 966 case POOL_STATE_ACTIVE: 967 /* 968 * For an active pool, we have to determine if it's really part 969 * of a currently active pool (in which case the pool will exist 970 * and the guid will be the same), or whether it's part of an 971 * active pool that was disconnected without being explicitly 972 * exported. 973 */ 974 if (pool_active(hdl, name, guid, &isactive) != 0) { 975 nvlist_free(config); 976 return (-1); 977 } 978 979 if (isactive) { 980 /* 981 * Because the device may have been removed while 982 * offlined, we only report it as active if the vdev is 983 * still present in the config. Otherwise, pretend like 984 * it's not in use. 985 */ 986 if ((zhp = zpool_open_canfail(hdl, name)) != NULL && 987 (pool_config = zpool_get_config(zhp, NULL)) 988 != NULL) { 989 nvlist_t *nvroot; 990 991 verify(nvlist_lookup_nvlist(pool_config, 992 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 993 ret = find_guid(nvroot, vdev_guid); 994 } else { 995 ret = B_FALSE; 996 } 997 998 /* 999 * If this is an active spare within another pool, we 1000 * treat it like an unused hot spare. This allows the 1001 * user to create a pool with a hot spare that currently 1002 * in use within another pool. Since we return B_TRUE, 1003 * libdiskmgt will continue to prevent generic consumers 1004 * from using the device. 1005 */ 1006 if (ret && nvlist_lookup_uint64(config, 1007 ZPOOL_CONFIG_IS_SPARE, &isspare) == 0 && isspare) 1008 stateval = POOL_STATE_SPARE; 1009 1010 if (zhp != NULL) 1011 zpool_close(zhp); 1012 } else { 1013 stateval = POOL_STATE_POTENTIALLY_ACTIVE; 1014 ret = B_TRUE; 1015 } 1016 break; 1017 1018 case POOL_STATE_SPARE: 1019 /* 1020 * For a hot spare, it can be either definitively in use, or 1021 * potentially active. To determine if it's in use, we iterate 1022 * over all pools in the system and search for one with a spare 1023 * with a matching guid. 1024 * 1025 * Due to the shared nature of spares, we don't actually report 1026 * the potentially active case as in use. This means the user 1027 * can freely create pools on the hot spares of exported pools, 1028 * but to do otherwise makes the resulting code complicated, and 1029 * we end up having to deal with this case anyway. 1030 */ 1031 cb.cb_zhp = NULL; 1032 cb.cb_guid = vdev_guid; 1033 if (zpool_iter(hdl, find_spare, &cb) == 1) { 1034 name = (char *)zpool_get_name(cb.cb_zhp); 1035 ret = TRUE; 1036 } else { 1037 ret = FALSE; 1038 } 1039 break; 1040 1041 default: 1042 ret = B_FALSE; 1043 } 1044 1045 1046 if (ret) { 1047 if ((*namestr = zfs_strdup(hdl, name)) == NULL) { 1048 nvlist_free(config); 1049 return (-1); 1050 } 1051 *state = (pool_state_t)stateval; 1052 } 1053 1054 if (cb.cb_zhp) 1055 zpool_close(cb.cb_zhp); 1056 1057 nvlist_free(config); 1058 *inuse = ret; 1059 return (0); 1060 } 1061