libzfs_import.c revision 260339
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 (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 23 * Copyright 2011 Nexenta Systems, Inc. All rights reserved. 24 * Copyright (c) 2012 by Delphix. All rights reserved. 25 */ 26 27/* 28 * Pool import support functions. 29 * 30 * To import a pool, we rely on reading the configuration information from the 31 * ZFS label of each device. If we successfully read the label, then we 32 * organize the configuration information in the following hierarchy: 33 * 34 * pool guid -> toplevel vdev guid -> label txg 35 * 36 * Duplicate entries matching this same tuple will be discarded. Once we have 37 * examined every device, we pick the best label txg config for each toplevel 38 * vdev. We then arrange these toplevel vdevs into a complete pool config, and 39 * update any paths that have changed. Finally, we attempt to import the pool 40 * using our derived config, and record the results. 41 */ 42 43#include <ctype.h> 44#include <devid.h> 45#include <dirent.h> 46#include <errno.h> 47#include <libintl.h> 48#include <stddef.h> 49#include <stdlib.h> 50#include <string.h> 51#include <sys/stat.h> 52#include <unistd.h> 53#include <fcntl.h> 54#include <thread_pool.h> 55#include <libgeom.h> 56 57#include <sys/vdev_impl.h> 58 59#include "libzfs.h" 60#include "libzfs_impl.h" 61 62/* 63 * Intermediate structures used to gather configuration information. 64 */ 65typedef struct config_entry { 66 uint64_t ce_txg; 67 nvlist_t *ce_config; 68 struct config_entry *ce_next; 69} config_entry_t; 70 71typedef struct vdev_entry { 72 uint64_t ve_guid; 73 config_entry_t *ve_configs; 74 struct vdev_entry *ve_next; 75} vdev_entry_t; 76 77typedef struct pool_entry { 78 uint64_t pe_guid; 79 vdev_entry_t *pe_vdevs; 80 struct pool_entry *pe_next; 81} pool_entry_t; 82 83typedef struct name_entry { 84 char *ne_name; 85 uint64_t ne_guid; 86 struct name_entry *ne_next; 87} name_entry_t; 88 89typedef struct pool_list { 90 pool_entry_t *pools; 91 name_entry_t *names; 92} pool_list_t; 93 94static char * 95get_devid(const char *path) 96{ 97 int fd; 98 ddi_devid_t devid; 99 char *minor, *ret; 100 101 if ((fd = open(path, O_RDONLY)) < 0) 102 return (NULL); 103 104 minor = NULL; 105 ret = NULL; 106 if (devid_get(fd, &devid) == 0) { 107 if (devid_get_minor_name(fd, &minor) == 0) 108 ret = devid_str_encode(devid, minor); 109 if (minor != NULL) 110 devid_str_free(minor); 111 devid_free(devid); 112 } 113 (void) close(fd); 114 115 return (ret); 116} 117 118 119/* 120 * Go through and fix up any path and/or devid information for the given vdev 121 * configuration. 122 */ 123static int 124fix_paths(nvlist_t *nv, name_entry_t *names) 125{ 126 nvlist_t **child; 127 uint_t c, children; 128 uint64_t guid; 129 name_entry_t *ne, *best; 130 char *path, *devid; 131 int matched; 132 133 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, 134 &child, &children) == 0) { 135 for (c = 0; c < children; c++) 136 if (fix_paths(child[c], names) != 0) 137 return (-1); 138 return (0); 139 } 140 141 /* 142 * This is a leaf (file or disk) vdev. In either case, go through 143 * the name list and see if we find a matching guid. If so, replace 144 * the path and see if we can calculate a new devid. 145 * 146 * There may be multiple names associated with a particular guid, in 147 * which case we have overlapping slices or multiple paths to the same 148 * disk. If this is the case, then we want to pick the path that is 149 * the most similar to the original, where "most similar" is the number 150 * of matching characters starting from the end of the path. This will 151 * preserve slice numbers even if the disks have been reorganized, and 152 * will also catch preferred disk names if multiple paths exist. 153 */ 154 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0); 155 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) != 0) 156 path = NULL; 157 158 matched = 0; 159 best = NULL; 160 for (ne = names; ne != NULL; ne = ne->ne_next) { 161 if (ne->ne_guid == guid) { 162 const char *src, *dst; 163 int count; 164 165 if (path == NULL) { 166 best = ne; 167 break; 168 } 169 170 src = ne->ne_name + strlen(ne->ne_name) - 1; 171 dst = path + strlen(path) - 1; 172 for (count = 0; src >= ne->ne_name && dst >= path; 173 src--, dst--, count++) 174 if (*src != *dst) 175 break; 176 177 /* 178 * At this point, 'count' is the number of characters 179 * matched from the end. 180 */ 181 if (count > matched || best == NULL) { 182 best = ne; 183 matched = count; 184 } 185 } 186 } 187 188 if (best == NULL) 189 return (0); 190 191 if (nvlist_add_string(nv, ZPOOL_CONFIG_PATH, best->ne_name) != 0) 192 return (-1); 193 194 if ((devid = get_devid(best->ne_name)) == NULL) { 195 (void) nvlist_remove_all(nv, ZPOOL_CONFIG_DEVID); 196 } else { 197 if (nvlist_add_string(nv, ZPOOL_CONFIG_DEVID, devid) != 0) 198 return (-1); 199 devid_str_free(devid); 200 } 201 202 return (0); 203} 204 205/* 206 * Add the given configuration to the list of known devices. 207 */ 208static int 209add_config(libzfs_handle_t *hdl, pool_list_t *pl, const char *path, 210 nvlist_t *config) 211{ 212 uint64_t pool_guid, vdev_guid, top_guid, txg, state; 213 pool_entry_t *pe; 214 vdev_entry_t *ve; 215 config_entry_t *ce; 216 name_entry_t *ne; 217 218 /* 219 * If this is a hot spare not currently in use or level 2 cache 220 * device, add it to the list of names to translate, but don't do 221 * anything else. 222 */ 223 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE, 224 &state) == 0 && 225 (state == POOL_STATE_SPARE || state == POOL_STATE_L2CACHE) && 226 nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, &vdev_guid) == 0) { 227 if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL) 228 return (-1); 229 230 if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) { 231 free(ne); 232 return (-1); 233 } 234 ne->ne_guid = vdev_guid; 235 ne->ne_next = pl->names; 236 pl->names = ne; 237 return (0); 238 } 239 240 /* 241 * If we have a valid config but cannot read any of these fields, then 242 * it means we have a half-initialized label. In vdev_label_init() 243 * we write a label with txg == 0 so that we can identify the device 244 * in case the user refers to the same disk later on. If we fail to 245 * create the pool, we'll be left with a label in this state 246 * which should not be considered part of a valid pool. 247 */ 248 if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, 249 &pool_guid) != 0 || 250 nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, 251 &vdev_guid) != 0 || 252 nvlist_lookup_uint64(config, ZPOOL_CONFIG_TOP_GUID, 253 &top_guid) != 0 || 254 nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG, 255 &txg) != 0 || txg == 0) { 256 nvlist_free(config); 257 return (0); 258 } 259 260 /* 261 * First, see if we know about this pool. If not, then add it to the 262 * list of known pools. 263 */ 264 for (pe = pl->pools; pe != NULL; pe = pe->pe_next) { 265 if (pe->pe_guid == pool_guid) 266 break; 267 } 268 269 if (pe == NULL) { 270 if ((pe = zfs_alloc(hdl, sizeof (pool_entry_t))) == NULL) { 271 nvlist_free(config); 272 return (-1); 273 } 274 pe->pe_guid = pool_guid; 275 pe->pe_next = pl->pools; 276 pl->pools = pe; 277 } 278 279 /* 280 * Second, see if we know about this toplevel vdev. Add it if its 281 * missing. 282 */ 283 for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) { 284 if (ve->ve_guid == top_guid) 285 break; 286 } 287 288 if (ve == NULL) { 289 if ((ve = zfs_alloc(hdl, sizeof (vdev_entry_t))) == NULL) { 290 nvlist_free(config); 291 return (-1); 292 } 293 ve->ve_guid = top_guid; 294 ve->ve_next = pe->pe_vdevs; 295 pe->pe_vdevs = ve; 296 } 297 298 /* 299 * Third, see if we have a config with a matching transaction group. If 300 * so, then we do nothing. Otherwise, add it to the list of known 301 * configs. 302 */ 303 for (ce = ve->ve_configs; ce != NULL; ce = ce->ce_next) { 304 if (ce->ce_txg == txg) 305 break; 306 } 307 308 if (ce == NULL) { 309 if ((ce = zfs_alloc(hdl, sizeof (config_entry_t))) == NULL) { 310 nvlist_free(config); 311 return (-1); 312 } 313 ce->ce_txg = txg; 314 ce->ce_config = config; 315 ce->ce_next = ve->ve_configs; 316 ve->ve_configs = ce; 317 } else { 318 nvlist_free(config); 319 } 320 321 /* 322 * At this point we've successfully added our config to the list of 323 * known configs. The last thing to do is add the vdev guid -> path 324 * mappings so that we can fix up the configuration as necessary before 325 * doing the import. 326 */ 327 if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL) 328 return (-1); 329 330 if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) { 331 free(ne); 332 return (-1); 333 } 334 335 ne->ne_guid = vdev_guid; 336 ne->ne_next = pl->names; 337 pl->names = ne; 338 339 return (0); 340} 341 342/* 343 * Returns true if the named pool matches the given GUID. 344 */ 345static int 346pool_active(libzfs_handle_t *hdl, const char *name, uint64_t guid, 347 boolean_t *isactive) 348{ 349 zpool_handle_t *zhp; 350 uint64_t theguid; 351 352 if (zpool_open_silent(hdl, name, &zhp) != 0) 353 return (-1); 354 355 if (zhp == NULL) { 356 *isactive = B_FALSE; 357 return (0); 358 } 359 360 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_POOL_GUID, 361 &theguid) == 0); 362 363 zpool_close(zhp); 364 365 *isactive = (theguid == guid); 366 return (0); 367} 368 369static nvlist_t * 370refresh_config(libzfs_handle_t *hdl, nvlist_t *config) 371{ 372 nvlist_t *nvl; 373 zfs_cmd_t zc = { 0 }; 374 int err; 375 376 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) 377 return (NULL); 378 379 if (zcmd_alloc_dst_nvlist(hdl, &zc, 380 zc.zc_nvlist_conf_size * 2) != 0) { 381 zcmd_free_nvlists(&zc); 382 return (NULL); 383 } 384 385 while ((err = ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_TRYIMPORT, 386 &zc)) != 0 && errno == ENOMEM) { 387 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) { 388 zcmd_free_nvlists(&zc); 389 return (NULL); 390 } 391 } 392 393 if (err) { 394 zcmd_free_nvlists(&zc); 395 return (NULL); 396 } 397 398 if (zcmd_read_dst_nvlist(hdl, &zc, &nvl) != 0) { 399 zcmd_free_nvlists(&zc); 400 return (NULL); 401 } 402 403 zcmd_free_nvlists(&zc); 404 return (nvl); 405} 406 407/* 408 * Determine if the vdev id is a hole in the namespace. 409 */ 410boolean_t 411vdev_is_hole(uint64_t *hole_array, uint_t holes, uint_t id) 412{ 413 for (int c = 0; c < holes; c++) { 414 415 /* Top-level is a hole */ 416 if (hole_array[c] == id) 417 return (B_TRUE); 418 } 419 return (B_FALSE); 420} 421 422/* 423 * Convert our list of pools into the definitive set of configurations. We 424 * start by picking the best config for each toplevel vdev. Once that's done, 425 * we assemble the toplevel vdevs into a full config for the pool. We make a 426 * pass to fix up any incorrect paths, and then add it to the main list to 427 * return to the user. 428 */ 429static nvlist_t * 430get_configs(libzfs_handle_t *hdl, pool_list_t *pl, boolean_t active_ok) 431{ 432 pool_entry_t *pe; 433 vdev_entry_t *ve; 434 config_entry_t *ce; 435 nvlist_t *ret = NULL, *config = NULL, *tmp, *nvtop, *nvroot; 436 nvlist_t **spares, **l2cache; 437 uint_t i, nspares, nl2cache; 438 boolean_t config_seen; 439 uint64_t best_txg; 440 char *name, *hostname; 441 uint64_t guid; 442 uint_t children = 0; 443 nvlist_t **child = NULL; 444 uint_t holes; 445 uint64_t *hole_array, max_id; 446 uint_t c; 447 boolean_t isactive; 448 uint64_t hostid; 449 nvlist_t *nvl; 450 boolean_t found_one = B_FALSE; 451 boolean_t valid_top_config = B_FALSE; 452 453 if (nvlist_alloc(&ret, 0, 0) != 0) 454 goto nomem; 455 456 for (pe = pl->pools; pe != NULL; pe = pe->pe_next) { 457 uint64_t id, max_txg = 0; 458 459 if (nvlist_alloc(&config, NV_UNIQUE_NAME, 0) != 0) 460 goto nomem; 461 config_seen = B_FALSE; 462 463 /* 464 * Iterate over all toplevel vdevs. Grab the pool configuration 465 * from the first one we find, and then go through the rest and 466 * add them as necessary to the 'vdevs' member of the config. 467 */ 468 for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) { 469 470 /* 471 * Determine the best configuration for this vdev by 472 * selecting the config with the latest transaction 473 * group. 474 */ 475 best_txg = 0; 476 for (ce = ve->ve_configs; ce != NULL; 477 ce = ce->ce_next) { 478 479 if (ce->ce_txg > best_txg) { 480 tmp = ce->ce_config; 481 best_txg = ce->ce_txg; 482 } 483 } 484 485 /* 486 * We rely on the fact that the max txg for the 487 * pool will contain the most up-to-date information 488 * about the valid top-levels in the vdev namespace. 489 */ 490 if (best_txg > max_txg) { 491 (void) nvlist_remove(config, 492 ZPOOL_CONFIG_VDEV_CHILDREN, 493 DATA_TYPE_UINT64); 494 (void) nvlist_remove(config, 495 ZPOOL_CONFIG_HOLE_ARRAY, 496 DATA_TYPE_UINT64_ARRAY); 497 498 max_txg = best_txg; 499 hole_array = NULL; 500 holes = 0; 501 max_id = 0; 502 valid_top_config = B_FALSE; 503 504 if (nvlist_lookup_uint64(tmp, 505 ZPOOL_CONFIG_VDEV_CHILDREN, &max_id) == 0) { 506 verify(nvlist_add_uint64(config, 507 ZPOOL_CONFIG_VDEV_CHILDREN, 508 max_id) == 0); 509 valid_top_config = B_TRUE; 510 } 511 512 if (nvlist_lookup_uint64_array(tmp, 513 ZPOOL_CONFIG_HOLE_ARRAY, &hole_array, 514 &holes) == 0) { 515 verify(nvlist_add_uint64_array(config, 516 ZPOOL_CONFIG_HOLE_ARRAY, 517 hole_array, holes) == 0); 518 } 519 } 520 521 if (!config_seen) { 522 /* 523 * Copy the relevant pieces of data to the pool 524 * configuration: 525 * 526 * version 527 * pool guid 528 * name 529 * comment (if available) 530 * pool state 531 * hostid (if available) 532 * hostname (if available) 533 */ 534 uint64_t state, version; 535 char *comment = NULL; 536 537 version = fnvlist_lookup_uint64(tmp, 538 ZPOOL_CONFIG_VERSION); 539 fnvlist_add_uint64(config, 540 ZPOOL_CONFIG_VERSION, version); 541 guid = fnvlist_lookup_uint64(tmp, 542 ZPOOL_CONFIG_POOL_GUID); 543 fnvlist_add_uint64(config, 544 ZPOOL_CONFIG_POOL_GUID, guid); 545 name = fnvlist_lookup_string(tmp, 546 ZPOOL_CONFIG_POOL_NAME); 547 fnvlist_add_string(config, 548 ZPOOL_CONFIG_POOL_NAME, name); 549 550 if (nvlist_lookup_string(tmp, 551 ZPOOL_CONFIG_COMMENT, &comment) == 0) 552 fnvlist_add_string(config, 553 ZPOOL_CONFIG_COMMENT, comment); 554 555 state = fnvlist_lookup_uint64(tmp, 556 ZPOOL_CONFIG_POOL_STATE); 557 fnvlist_add_uint64(config, 558 ZPOOL_CONFIG_POOL_STATE, state); 559 560 hostid = 0; 561 if (nvlist_lookup_uint64(tmp, 562 ZPOOL_CONFIG_HOSTID, &hostid) == 0) { 563 fnvlist_add_uint64(config, 564 ZPOOL_CONFIG_HOSTID, hostid); 565 hostname = fnvlist_lookup_string(tmp, 566 ZPOOL_CONFIG_HOSTNAME); 567 fnvlist_add_string(config, 568 ZPOOL_CONFIG_HOSTNAME, hostname); 569 } 570 571 config_seen = B_TRUE; 572 } 573 574 /* 575 * Add this top-level vdev to the child array. 576 */ 577 verify(nvlist_lookup_nvlist(tmp, 578 ZPOOL_CONFIG_VDEV_TREE, &nvtop) == 0); 579 verify(nvlist_lookup_uint64(nvtop, ZPOOL_CONFIG_ID, 580 &id) == 0); 581 582 if (id >= children) { 583 nvlist_t **newchild; 584 585 newchild = zfs_alloc(hdl, (id + 1) * 586 sizeof (nvlist_t *)); 587 if (newchild == NULL) 588 goto nomem; 589 590 for (c = 0; c < children; c++) 591 newchild[c] = child[c]; 592 593 free(child); 594 child = newchild; 595 children = id + 1; 596 } 597 if (nvlist_dup(nvtop, &child[id], 0) != 0) 598 goto nomem; 599 600 } 601 602 /* 603 * If we have information about all the top-levels then 604 * clean up the nvlist which we've constructed. This 605 * means removing any extraneous devices that are 606 * beyond the valid range or adding devices to the end 607 * of our array which appear to be missing. 608 */ 609 if (valid_top_config) { 610 if (max_id < children) { 611 for (c = max_id; c < children; c++) 612 nvlist_free(child[c]); 613 children = max_id; 614 } else if (max_id > children) { 615 nvlist_t **newchild; 616 617 newchild = zfs_alloc(hdl, (max_id) * 618 sizeof (nvlist_t *)); 619 if (newchild == NULL) 620 goto nomem; 621 622 for (c = 0; c < children; c++) 623 newchild[c] = child[c]; 624 625 free(child); 626 child = newchild; 627 children = max_id; 628 } 629 } 630 631 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, 632 &guid) == 0); 633 634 /* 635 * The vdev namespace may contain holes as a result of 636 * device removal. We must add them back into the vdev 637 * tree before we process any missing devices. 638 */ 639 if (holes > 0) { 640 ASSERT(valid_top_config); 641 642 for (c = 0; c < children; c++) { 643 nvlist_t *holey; 644 645 if (child[c] != NULL || 646 !vdev_is_hole(hole_array, holes, c)) 647 continue; 648 649 if (nvlist_alloc(&holey, NV_UNIQUE_NAME, 650 0) != 0) 651 goto nomem; 652 653 /* 654 * Holes in the namespace are treated as 655 * "hole" top-level vdevs and have a 656 * special flag set on them. 657 */ 658 if (nvlist_add_string(holey, 659 ZPOOL_CONFIG_TYPE, 660 VDEV_TYPE_HOLE) != 0 || 661 nvlist_add_uint64(holey, 662 ZPOOL_CONFIG_ID, c) != 0 || 663 nvlist_add_uint64(holey, 664 ZPOOL_CONFIG_GUID, 0ULL) != 0) 665 goto nomem; 666 child[c] = holey; 667 } 668 } 669 670 /* 671 * Look for any missing top-level vdevs. If this is the case, 672 * create a faked up 'missing' vdev as a placeholder. We cannot 673 * simply compress the child array, because the kernel performs 674 * certain checks to make sure the vdev IDs match their location 675 * in the configuration. 676 */ 677 for (c = 0; c < children; c++) { 678 if (child[c] == NULL) { 679 nvlist_t *missing; 680 if (nvlist_alloc(&missing, NV_UNIQUE_NAME, 681 0) != 0) 682 goto nomem; 683 if (nvlist_add_string(missing, 684 ZPOOL_CONFIG_TYPE, 685 VDEV_TYPE_MISSING) != 0 || 686 nvlist_add_uint64(missing, 687 ZPOOL_CONFIG_ID, c) != 0 || 688 nvlist_add_uint64(missing, 689 ZPOOL_CONFIG_GUID, 0ULL) != 0) { 690 nvlist_free(missing); 691 goto nomem; 692 } 693 child[c] = missing; 694 } 695 } 696 697 /* 698 * Put all of this pool's top-level vdevs into a root vdev. 699 */ 700 if (nvlist_alloc(&nvroot, NV_UNIQUE_NAME, 0) != 0) 701 goto nomem; 702 if (nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE, 703 VDEV_TYPE_ROOT) != 0 || 704 nvlist_add_uint64(nvroot, ZPOOL_CONFIG_ID, 0ULL) != 0 || 705 nvlist_add_uint64(nvroot, ZPOOL_CONFIG_GUID, guid) != 0 || 706 nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, 707 child, children) != 0) { 708 nvlist_free(nvroot); 709 goto nomem; 710 } 711 712 for (c = 0; c < children; c++) 713 nvlist_free(child[c]); 714 free(child); 715 children = 0; 716 child = NULL; 717 718 /* 719 * Go through and fix up any paths and/or devids based on our 720 * known list of vdev GUID -> path mappings. 721 */ 722 if (fix_paths(nvroot, pl->names) != 0) { 723 nvlist_free(nvroot); 724 goto nomem; 725 } 726 727 /* 728 * Add the root vdev to this pool's configuration. 729 */ 730 if (nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 731 nvroot) != 0) { 732 nvlist_free(nvroot); 733 goto nomem; 734 } 735 nvlist_free(nvroot); 736 737 /* 738 * zdb uses this path to report on active pools that were 739 * imported or created using -R. 740 */ 741 if (active_ok) 742 goto add_pool; 743 744 /* 745 * Determine if this pool is currently active, in which case we 746 * can't actually import it. 747 */ 748 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, 749 &name) == 0); 750 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, 751 &guid) == 0); 752 753 if (pool_active(hdl, name, guid, &isactive) != 0) 754 goto error; 755 756 if (isactive) { 757 nvlist_free(config); 758 config = NULL; 759 continue; 760 } 761 762 if ((nvl = refresh_config(hdl, config)) == NULL) { 763 nvlist_free(config); 764 config = NULL; 765 continue; 766 } 767 768 nvlist_free(config); 769 config = nvl; 770 771 /* 772 * Go through and update the paths for spares, now that we have 773 * them. 774 */ 775 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 776 &nvroot) == 0); 777 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, 778 &spares, &nspares) == 0) { 779 for (i = 0; i < nspares; i++) { 780 if (fix_paths(spares[i], pl->names) != 0) 781 goto nomem; 782 } 783 } 784 785 /* 786 * Update the paths for l2cache devices. 787 */ 788 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, 789 &l2cache, &nl2cache) == 0) { 790 for (i = 0; i < nl2cache; i++) { 791 if (fix_paths(l2cache[i], pl->names) != 0) 792 goto nomem; 793 } 794 } 795 796 /* 797 * Restore the original information read from the actual label. 798 */ 799 (void) nvlist_remove(config, ZPOOL_CONFIG_HOSTID, 800 DATA_TYPE_UINT64); 801 (void) nvlist_remove(config, ZPOOL_CONFIG_HOSTNAME, 802 DATA_TYPE_STRING); 803 if (hostid != 0) { 804 verify(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID, 805 hostid) == 0); 806 verify(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME, 807 hostname) == 0); 808 } 809 810add_pool: 811 /* 812 * Add this pool to the list of configs. 813 */ 814 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, 815 &name) == 0); 816 if (nvlist_add_nvlist(ret, name, config) != 0) 817 goto nomem; 818 819 found_one = B_TRUE; 820 nvlist_free(config); 821 config = NULL; 822 } 823 824 if (!found_one) { 825 nvlist_free(ret); 826 ret = NULL; 827 } 828 829 return (ret); 830 831nomem: 832 (void) no_memory(hdl); 833error: 834 nvlist_free(config); 835 nvlist_free(ret); 836 for (c = 0; c < children; c++) 837 nvlist_free(child[c]); 838 free(child); 839 840 return (NULL); 841} 842 843/* 844 * Return the offset of the given label. 845 */ 846static uint64_t 847label_offset(uint64_t size, int l) 848{ 849 ASSERT(P2PHASE_TYPED(size, sizeof (vdev_label_t), uint64_t) == 0); 850 return (l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ? 851 0 : size - VDEV_LABELS * sizeof (vdev_label_t))); 852} 853 854/* 855 * Given a file descriptor, read the label information and return an nvlist 856 * describing the configuration, if there is one. 857 */ 858int 859zpool_read_label(int fd, nvlist_t **config) 860{ 861 struct stat64 statbuf; 862 int l; 863 vdev_label_t *label; 864 uint64_t state, txg, size; 865 866 *config = NULL; 867 868 if (fstat64(fd, &statbuf) == -1) 869 return (0); 870 size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t); 871 872 if ((label = malloc(sizeof (vdev_label_t))) == NULL) 873 return (-1); 874 875 for (l = 0; l < VDEV_LABELS; l++) { 876 if (pread64(fd, label, sizeof (vdev_label_t), 877 label_offset(size, l)) != sizeof (vdev_label_t)) 878 continue; 879 880 if (nvlist_unpack(label->vl_vdev_phys.vp_nvlist, 881 sizeof (label->vl_vdev_phys.vp_nvlist), config, 0) != 0) 882 continue; 883 884 if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE, 885 &state) != 0 || state > POOL_STATE_L2CACHE) { 886 nvlist_free(*config); 887 continue; 888 } 889 890 if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE && 891 (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_TXG, 892 &txg) != 0 || txg == 0)) { 893 nvlist_free(*config); 894 continue; 895 } 896 897 free(label); 898 return (0); 899 } 900 901 free(label); 902 *config = NULL; 903 return (0); 904} 905 906typedef struct rdsk_node { 907 char *rn_name; 908 int rn_dfd; 909 libzfs_handle_t *rn_hdl; 910 nvlist_t *rn_config; 911 avl_tree_t *rn_avl; 912 avl_node_t rn_node; 913 boolean_t rn_nozpool; 914} rdsk_node_t; 915 916static int 917slice_cache_compare(const void *arg1, const void *arg2) 918{ 919 const char *nm1 = ((rdsk_node_t *)arg1)->rn_name; 920 const char *nm2 = ((rdsk_node_t *)arg2)->rn_name; 921 char *nm1slice, *nm2slice; 922 int rv; 923 924 /* 925 * slices zero and two are the most likely to provide results, 926 * so put those first 927 */ 928 nm1slice = strstr(nm1, "s0"); 929 nm2slice = strstr(nm2, "s0"); 930 if (nm1slice && !nm2slice) { 931 return (-1); 932 } 933 if (!nm1slice && nm2slice) { 934 return (1); 935 } 936 nm1slice = strstr(nm1, "s2"); 937 nm2slice = strstr(nm2, "s2"); 938 if (nm1slice && !nm2slice) { 939 return (-1); 940 } 941 if (!nm1slice && nm2slice) { 942 return (1); 943 } 944 945 rv = strcmp(nm1, nm2); 946 if (rv == 0) 947 return (0); 948 return (rv > 0 ? 1 : -1); 949} 950 951#ifdef sun 952static void 953check_one_slice(avl_tree_t *r, char *diskname, uint_t partno, 954 diskaddr_t size, uint_t blksz) 955{ 956 rdsk_node_t tmpnode; 957 rdsk_node_t *node; 958 char sname[MAXNAMELEN]; 959 960 tmpnode.rn_name = &sname[0]; 961 (void) snprintf(tmpnode.rn_name, MAXNAMELEN, "%s%u", 962 diskname, partno); 963 /* 964 * protect against division by zero for disk labels that 965 * contain a bogus sector size 966 */ 967 if (blksz == 0) 968 blksz = DEV_BSIZE; 969 /* too small to contain a zpool? */ 970 if ((size < (SPA_MINDEVSIZE / blksz)) && 971 (node = avl_find(r, &tmpnode, NULL))) 972 node->rn_nozpool = B_TRUE; 973} 974#endif /* sun */ 975 976static void 977nozpool_all_slices(avl_tree_t *r, const char *sname) 978{ 979#ifdef sun 980 char diskname[MAXNAMELEN]; 981 char *ptr; 982 int i; 983 984 (void) strncpy(diskname, sname, MAXNAMELEN); 985 if (((ptr = strrchr(diskname, 's')) == NULL) && 986 ((ptr = strrchr(diskname, 'p')) == NULL)) 987 return; 988 ptr[0] = 's'; 989 ptr[1] = '\0'; 990 for (i = 0; i < NDKMAP; i++) 991 check_one_slice(r, diskname, i, 0, 1); 992 ptr[0] = 'p'; 993 for (i = 0; i <= FD_NUMPART; i++) 994 check_one_slice(r, diskname, i, 0, 1); 995#endif /* sun */ 996} 997 998#ifdef sun 999static void 1000check_slices(avl_tree_t *r, int fd, const char *sname) 1001{ 1002 struct extvtoc vtoc; 1003 struct dk_gpt *gpt; 1004 char diskname[MAXNAMELEN]; 1005 char *ptr; 1006 int i; 1007 1008 (void) strncpy(diskname, sname, MAXNAMELEN); 1009 if ((ptr = strrchr(diskname, 's')) == NULL || !isdigit(ptr[1])) 1010 return; 1011 ptr[1] = '\0'; 1012 1013 if (read_extvtoc(fd, &vtoc) >= 0) { 1014 for (i = 0; i < NDKMAP; i++) 1015 check_one_slice(r, diskname, i, 1016 vtoc.v_part[i].p_size, vtoc.v_sectorsz); 1017 } else if (efi_alloc_and_read(fd, &gpt) >= 0) { 1018 /* 1019 * on x86 we'll still have leftover links that point 1020 * to slices s[9-15], so use NDKMAP instead 1021 */ 1022 for (i = 0; i < NDKMAP; i++) 1023 check_one_slice(r, diskname, i, 1024 gpt->efi_parts[i].p_size, gpt->efi_lbasize); 1025 /* nodes p[1-4] are never used with EFI labels */ 1026 ptr[0] = 'p'; 1027 for (i = 1; i <= FD_NUMPART; i++) 1028 check_one_slice(r, diskname, i, 0, 1); 1029 efi_free(gpt); 1030 } 1031} 1032#endif /* sun */ 1033 1034static void 1035zpool_open_func(void *arg) 1036{ 1037 rdsk_node_t *rn = arg; 1038 struct stat64 statbuf; 1039 nvlist_t *config; 1040 int fd; 1041 1042 if (rn->rn_nozpool) 1043 return; 1044 if ((fd = openat64(rn->rn_dfd, rn->rn_name, O_RDONLY)) < 0) { 1045 /* symlink to a device that's no longer there */ 1046 if (errno == ENOENT) 1047 nozpool_all_slices(rn->rn_avl, rn->rn_name); 1048 return; 1049 } 1050 /* 1051 * Ignore failed stats. We only want regular 1052 * files, character devs and block devs. 1053 */ 1054 if (fstat64(fd, &statbuf) != 0 || 1055 (!S_ISREG(statbuf.st_mode) && 1056 !S_ISCHR(statbuf.st_mode) && 1057 !S_ISBLK(statbuf.st_mode))) { 1058 (void) close(fd); 1059 return; 1060 } 1061 /* this file is too small to hold a zpool */ 1062#ifdef sun 1063 if (S_ISREG(statbuf.st_mode) && 1064 statbuf.st_size < SPA_MINDEVSIZE) { 1065 (void) close(fd); 1066 return; 1067 } else if (!S_ISREG(statbuf.st_mode)) { 1068 /* 1069 * Try to read the disk label first so we don't have to 1070 * open a bunch of minor nodes that can't have a zpool. 1071 */ 1072 check_slices(rn->rn_avl, fd, rn->rn_name); 1073 } 1074#else /* !sun */ 1075 if (statbuf.st_size < SPA_MINDEVSIZE) { 1076 (void) close(fd); 1077 return; 1078 } 1079#endif /* sun */ 1080 1081 if ((zpool_read_label(fd, &config)) != 0) { 1082 (void) close(fd); 1083 (void) no_memory(rn->rn_hdl); 1084 return; 1085 } 1086 (void) close(fd); 1087 1088 1089 rn->rn_config = config; 1090 if (config != NULL) { 1091 assert(rn->rn_nozpool == B_FALSE); 1092 } 1093} 1094 1095/* 1096 * Given a file descriptor, clear (zero) the label information. This function 1097 * is used in the appliance stack as part of the ZFS sysevent module and 1098 * to implement the "zpool labelclear" command. 1099 */ 1100int 1101zpool_clear_label(int fd) 1102{ 1103 struct stat64 statbuf; 1104 int l; 1105 vdev_label_t *label; 1106 uint64_t size; 1107 1108 if (fstat64(fd, &statbuf) == -1) 1109 return (0); 1110 size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t); 1111 1112 if ((label = calloc(sizeof (vdev_label_t), 1)) == NULL) 1113 return (-1); 1114 1115 for (l = 0; l < VDEV_LABELS; l++) { 1116 if (pwrite64(fd, label, sizeof (vdev_label_t), 1117 label_offset(size, l)) != sizeof (vdev_label_t)) 1118 return (-1); 1119 } 1120 1121 free(label); 1122 return (0); 1123} 1124 1125/* 1126 * Given a list of directories to search, find all pools stored on disk. This 1127 * includes partial pools which are not available to import. If no args are 1128 * given (argc is 0), then the default directory (/dev/dsk) is searched. 1129 * poolname or guid (but not both) are provided by the caller when trying 1130 * to import a specific pool. 1131 */ 1132static nvlist_t * 1133zpool_find_import_impl(libzfs_handle_t *hdl, importargs_t *iarg) 1134{ 1135 int i, dirs = iarg->paths; 1136 DIR *dirp = NULL; 1137 struct dirent64 *dp; 1138 char path[MAXPATHLEN]; 1139 char *end, **dir = iarg->path; 1140 size_t pathleft; 1141 nvlist_t *ret = NULL; 1142 static char *default_dir = "/dev"; 1143 pool_list_t pools = { 0 }; 1144 pool_entry_t *pe, *penext; 1145 vdev_entry_t *ve, *venext; 1146 config_entry_t *ce, *cenext; 1147 name_entry_t *ne, *nenext; 1148 avl_tree_t slice_cache; 1149 rdsk_node_t *slice; 1150 void *cookie; 1151 1152 if (dirs == 0) { 1153 dirs = 1; 1154 dir = &default_dir; 1155 } 1156 1157 /* 1158 * Go through and read the label configuration information from every 1159 * possible device, organizing the information according to pool GUID 1160 * and toplevel GUID. 1161 */ 1162 for (i = 0; i < dirs; i++) { 1163 tpool_t *t; 1164 char *rdsk; 1165 int dfd; 1166 1167 /* use realpath to normalize the path */ 1168 if (realpath(dir[i], path) == 0) { 1169 (void) zfs_error_fmt(hdl, EZFS_BADPATH, 1170 dgettext(TEXT_DOMAIN, "cannot open '%s'"), dir[i]); 1171 goto error; 1172 } 1173 end = &path[strlen(path)]; 1174 *end++ = '/'; 1175 *end = 0; 1176 pathleft = &path[sizeof (path)] - end; 1177 1178 /* 1179 * Using raw devices instead of block devices when we're 1180 * reading the labels skips a bunch of slow operations during 1181 * close(2) processing, so we replace /dev/dsk with /dev/rdsk. 1182 */ 1183 if (strcmp(path, "/dev/dsk/") == 0) 1184 rdsk = "/dev/"; 1185 else 1186 rdsk = path; 1187 1188 if ((dfd = open64(rdsk, O_RDONLY)) < 0 || 1189 (dirp = fdopendir(dfd)) == NULL) { 1190 zfs_error_aux(hdl, strerror(errno)); 1191 (void) zfs_error_fmt(hdl, EZFS_BADPATH, 1192 dgettext(TEXT_DOMAIN, "cannot open '%s'"), 1193 rdsk); 1194 goto error; 1195 } 1196 1197 avl_create(&slice_cache, slice_cache_compare, 1198 sizeof (rdsk_node_t), offsetof(rdsk_node_t, rn_node)); 1199 1200 if (strcmp(rdsk, "/dev/") == 0) { 1201 struct gmesh mesh; 1202 struct gclass *mp; 1203 struct ggeom *gp; 1204 struct gprovider *pp; 1205 1206 errno = geom_gettree(&mesh); 1207 if (errno != 0) { 1208 zfs_error_aux(hdl, strerror(errno)); 1209 (void) zfs_error_fmt(hdl, EZFS_BADPATH, 1210 dgettext(TEXT_DOMAIN, "cannot get GEOM tree")); 1211 goto error; 1212 } 1213 1214 LIST_FOREACH(mp, &mesh.lg_class, lg_class) { 1215 LIST_FOREACH(gp, &mp->lg_geom, lg_geom) { 1216 LIST_FOREACH(pp, &gp->lg_provider, lg_provider) { 1217 slice = zfs_alloc(hdl, sizeof (rdsk_node_t)); 1218 slice->rn_name = zfs_strdup(hdl, pp->lg_name); 1219 slice->rn_avl = &slice_cache; 1220 slice->rn_dfd = dfd; 1221 slice->rn_hdl = hdl; 1222 slice->rn_nozpool = B_FALSE; 1223 avl_add(&slice_cache, slice); 1224 } 1225 } 1226 } 1227 1228 geom_deletetree(&mesh); 1229 goto skipdir; 1230 } 1231 1232 /* 1233 * This is not MT-safe, but we have no MT consumers of libzfs 1234 */ 1235 while ((dp = readdir64(dirp)) != NULL) { 1236 const char *name = dp->d_name; 1237 if (name[0] == '.' && 1238 (name[1] == 0 || (name[1] == '.' && name[2] == 0))) 1239 continue; 1240 1241 slice = zfs_alloc(hdl, sizeof (rdsk_node_t)); 1242 slice->rn_name = zfs_strdup(hdl, name); 1243 slice->rn_avl = &slice_cache; 1244 slice->rn_dfd = dfd; 1245 slice->rn_hdl = hdl; 1246 slice->rn_nozpool = B_FALSE; 1247 avl_add(&slice_cache, slice); 1248 } 1249skipdir: 1250 /* 1251 * create a thread pool to do all of this in parallel; 1252 * rn_nozpool is not protected, so this is racy in that 1253 * multiple tasks could decide that the same slice can 1254 * not hold a zpool, which is benign. Also choose 1255 * double the number of processors; we hold a lot of 1256 * locks in the kernel, so going beyond this doesn't 1257 * buy us much. 1258 */ 1259 t = tpool_create(1, 2 * sysconf(_SC_NPROCESSORS_ONLN), 1260 0, NULL); 1261 for (slice = avl_first(&slice_cache); slice; 1262 (slice = avl_walk(&slice_cache, slice, 1263 AVL_AFTER))) 1264 (void) tpool_dispatch(t, zpool_open_func, slice); 1265 tpool_wait(t); 1266 tpool_destroy(t); 1267 1268 cookie = NULL; 1269 while ((slice = avl_destroy_nodes(&slice_cache, 1270 &cookie)) != NULL) { 1271 if (slice->rn_config != NULL) { 1272 nvlist_t *config = slice->rn_config; 1273 boolean_t matched = B_TRUE; 1274 1275 if (iarg->poolname != NULL) { 1276 char *pname; 1277 1278 matched = nvlist_lookup_string(config, 1279 ZPOOL_CONFIG_POOL_NAME, 1280 &pname) == 0 && 1281 strcmp(iarg->poolname, pname) == 0; 1282 } else if (iarg->guid != 0) { 1283 uint64_t this_guid; 1284 1285 matched = nvlist_lookup_uint64(config, 1286 ZPOOL_CONFIG_POOL_GUID, 1287 &this_guid) == 0 && 1288 iarg->guid == this_guid; 1289 } 1290 if (!matched) { 1291 nvlist_free(config); 1292 config = NULL; 1293 continue; 1294 } 1295 /* use the non-raw path for the config */ 1296 (void) strlcpy(end, slice->rn_name, pathleft); 1297 if (add_config(hdl, &pools, path, config) != 0) 1298 goto error; 1299 } 1300 free(slice->rn_name); 1301 free(slice); 1302 } 1303 avl_destroy(&slice_cache); 1304 1305 (void) closedir(dirp); 1306 dirp = NULL; 1307 } 1308 1309 ret = get_configs(hdl, &pools, iarg->can_be_active); 1310 1311error: 1312 for (pe = pools.pools; pe != NULL; pe = penext) { 1313 penext = pe->pe_next; 1314 for (ve = pe->pe_vdevs; ve != NULL; ve = venext) { 1315 venext = ve->ve_next; 1316 for (ce = ve->ve_configs; ce != NULL; ce = cenext) { 1317 cenext = ce->ce_next; 1318 if (ce->ce_config) 1319 nvlist_free(ce->ce_config); 1320 free(ce); 1321 } 1322 free(ve); 1323 } 1324 free(pe); 1325 } 1326 1327 for (ne = pools.names; ne != NULL; ne = nenext) { 1328 nenext = ne->ne_next; 1329 if (ne->ne_name) 1330 free(ne->ne_name); 1331 free(ne); 1332 } 1333 1334 if (dirp) 1335 (void) closedir(dirp); 1336 1337 return (ret); 1338} 1339 1340nvlist_t * 1341zpool_find_import(libzfs_handle_t *hdl, int argc, char **argv) 1342{ 1343 importargs_t iarg = { 0 }; 1344 1345 iarg.paths = argc; 1346 iarg.path = argv; 1347 1348 return (zpool_find_import_impl(hdl, &iarg)); 1349} 1350 1351/* 1352 * Given a cache file, return the contents as a list of importable pools. 1353 * poolname or guid (but not both) are provided by the caller when trying 1354 * to import a specific pool. 1355 */ 1356nvlist_t * 1357zpool_find_import_cached(libzfs_handle_t *hdl, const char *cachefile, 1358 char *poolname, uint64_t guid) 1359{ 1360 char *buf; 1361 int fd; 1362 struct stat64 statbuf; 1363 nvlist_t *raw, *src, *dst; 1364 nvlist_t *pools; 1365 nvpair_t *elem; 1366 char *name; 1367 uint64_t this_guid; 1368 boolean_t active; 1369 1370 verify(poolname == NULL || guid == 0); 1371 1372 if ((fd = open(cachefile, O_RDONLY)) < 0) { 1373 zfs_error_aux(hdl, "%s", strerror(errno)); 1374 (void) zfs_error(hdl, EZFS_BADCACHE, 1375 dgettext(TEXT_DOMAIN, "failed to open cache file")); 1376 return (NULL); 1377 } 1378 1379 if (fstat64(fd, &statbuf) != 0) { 1380 zfs_error_aux(hdl, "%s", strerror(errno)); 1381 (void) close(fd); 1382 (void) zfs_error(hdl, EZFS_BADCACHE, 1383 dgettext(TEXT_DOMAIN, "failed to get size of cache file")); 1384 return (NULL); 1385 } 1386 1387 if ((buf = zfs_alloc(hdl, statbuf.st_size)) == NULL) { 1388 (void) close(fd); 1389 return (NULL); 1390 } 1391 1392 if (read(fd, buf, statbuf.st_size) != statbuf.st_size) { 1393 (void) close(fd); 1394 free(buf); 1395 (void) zfs_error(hdl, EZFS_BADCACHE, 1396 dgettext(TEXT_DOMAIN, 1397 "failed to read cache file contents")); 1398 return (NULL); 1399 } 1400 1401 (void) close(fd); 1402 1403 if (nvlist_unpack(buf, statbuf.st_size, &raw, 0) != 0) { 1404 free(buf); 1405 (void) zfs_error(hdl, EZFS_BADCACHE, 1406 dgettext(TEXT_DOMAIN, 1407 "invalid or corrupt cache file contents")); 1408 return (NULL); 1409 } 1410 1411 free(buf); 1412 1413 /* 1414 * Go through and get the current state of the pools and refresh their 1415 * state. 1416 */ 1417 if (nvlist_alloc(&pools, 0, 0) != 0) { 1418 (void) no_memory(hdl); 1419 nvlist_free(raw); 1420 return (NULL); 1421 } 1422 1423 elem = NULL; 1424 while ((elem = nvlist_next_nvpair(raw, elem)) != NULL) { 1425 verify(nvpair_value_nvlist(elem, &src) == 0); 1426 1427 verify(nvlist_lookup_string(src, ZPOOL_CONFIG_POOL_NAME, 1428 &name) == 0); 1429 if (poolname != NULL && strcmp(poolname, name) != 0) 1430 continue; 1431 1432 verify(nvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID, 1433 &this_guid) == 0); 1434 if (guid != 0) { 1435 verify(nvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID, 1436 &this_guid) == 0); 1437 if (guid != this_guid) 1438 continue; 1439 } 1440 1441 if (pool_active(hdl, name, this_guid, &active) != 0) { 1442 nvlist_free(raw); 1443 nvlist_free(pools); 1444 return (NULL); 1445 } 1446 1447 if (active) 1448 continue; 1449 1450 if ((dst = refresh_config(hdl, src)) == NULL) { 1451 nvlist_free(raw); 1452 nvlist_free(pools); 1453 return (NULL); 1454 } 1455 1456 if (nvlist_add_nvlist(pools, nvpair_name(elem), dst) != 0) { 1457 (void) no_memory(hdl); 1458 nvlist_free(dst); 1459 nvlist_free(raw); 1460 nvlist_free(pools); 1461 return (NULL); 1462 } 1463 nvlist_free(dst); 1464 } 1465 1466 nvlist_free(raw); 1467 return (pools); 1468} 1469 1470static int 1471name_or_guid_exists(zpool_handle_t *zhp, void *data) 1472{ 1473 importargs_t *import = data; 1474 int found = 0; 1475 1476 if (import->poolname != NULL) { 1477 char *pool_name; 1478 1479 verify(nvlist_lookup_string(zhp->zpool_config, 1480 ZPOOL_CONFIG_POOL_NAME, &pool_name) == 0); 1481 if (strcmp(pool_name, import->poolname) == 0) 1482 found = 1; 1483 } else { 1484 uint64_t pool_guid; 1485 1486 verify(nvlist_lookup_uint64(zhp->zpool_config, 1487 ZPOOL_CONFIG_POOL_GUID, &pool_guid) == 0); 1488 if (pool_guid == import->guid) 1489 found = 1; 1490 } 1491 1492 zpool_close(zhp); 1493 return (found); 1494} 1495 1496nvlist_t * 1497zpool_search_import(libzfs_handle_t *hdl, importargs_t *import) 1498{ 1499 verify(import->poolname == NULL || import->guid == 0); 1500 1501 if (import->unique) 1502 import->exists = zpool_iter(hdl, name_or_guid_exists, import); 1503 1504 if (import->cachefile != NULL) 1505 return (zpool_find_import_cached(hdl, import->cachefile, 1506 import->poolname, import->guid)); 1507 1508 return (zpool_find_import_impl(hdl, import)); 1509} 1510 1511boolean_t 1512find_guid(nvlist_t *nv, uint64_t guid) 1513{ 1514 uint64_t tmp; 1515 nvlist_t **child; 1516 uint_t c, children; 1517 1518 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &tmp) == 0); 1519 if (tmp == guid) 1520 return (B_TRUE); 1521 1522 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, 1523 &child, &children) == 0) { 1524 for (c = 0; c < children; c++) 1525 if (find_guid(child[c], guid)) 1526 return (B_TRUE); 1527 } 1528 1529 return (B_FALSE); 1530} 1531 1532typedef struct aux_cbdata { 1533 const char *cb_type; 1534 uint64_t cb_guid; 1535 zpool_handle_t *cb_zhp; 1536} aux_cbdata_t; 1537 1538static int 1539find_aux(zpool_handle_t *zhp, void *data) 1540{ 1541 aux_cbdata_t *cbp = data; 1542 nvlist_t **list; 1543 uint_t i, count; 1544 uint64_t guid; 1545 nvlist_t *nvroot; 1546 1547 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE, 1548 &nvroot) == 0); 1549 1550 if (nvlist_lookup_nvlist_array(nvroot, cbp->cb_type, 1551 &list, &count) == 0) { 1552 for (i = 0; i < count; i++) { 1553 verify(nvlist_lookup_uint64(list[i], 1554 ZPOOL_CONFIG_GUID, &guid) == 0); 1555 if (guid == cbp->cb_guid) { 1556 cbp->cb_zhp = zhp; 1557 return (1); 1558 } 1559 } 1560 } 1561 1562 zpool_close(zhp); 1563 return (0); 1564} 1565 1566/* 1567 * Determines if the pool is in use. If so, it returns true and the state of 1568 * the pool as well as the name of the pool. Both strings are allocated and 1569 * must be freed by the caller. 1570 */ 1571int 1572zpool_in_use(libzfs_handle_t *hdl, int fd, pool_state_t *state, char **namestr, 1573 boolean_t *inuse) 1574{ 1575 nvlist_t *config; 1576 char *name; 1577 boolean_t ret; 1578 uint64_t guid, vdev_guid; 1579 zpool_handle_t *zhp; 1580 nvlist_t *pool_config; 1581 uint64_t stateval, isspare; 1582 aux_cbdata_t cb = { 0 }; 1583 boolean_t isactive; 1584 1585 *inuse = B_FALSE; 1586 1587 if (zpool_read_label(fd, &config) != 0) { 1588 (void) no_memory(hdl); 1589 return (-1); 1590 } 1591 1592 if (config == NULL) 1593 return (0); 1594 1595 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE, 1596 &stateval) == 0); 1597 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, 1598 &vdev_guid) == 0); 1599 1600 if (stateval != POOL_STATE_SPARE && stateval != POOL_STATE_L2CACHE) { 1601 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, 1602 &name) == 0); 1603 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, 1604 &guid) == 0); 1605 } 1606 1607 switch (stateval) { 1608 case POOL_STATE_EXPORTED: 1609 /* 1610 * A pool with an exported state may in fact be imported 1611 * read-only, so check the in-core state to see if it's 1612 * active and imported read-only. If it is, set 1613 * its state to active. 1614 */ 1615 if (pool_active(hdl, name, guid, &isactive) == 0 && isactive && 1616 (zhp = zpool_open_canfail(hdl, name)) != NULL && 1617 zpool_get_prop_int(zhp, ZPOOL_PROP_READONLY, NULL)) 1618 stateval = POOL_STATE_ACTIVE; 1619 1620 ret = B_TRUE; 1621 break; 1622 1623 case POOL_STATE_ACTIVE: 1624 /* 1625 * For an active pool, we have to determine if it's really part 1626 * of a currently active pool (in which case the pool will exist 1627 * and the guid will be the same), or whether it's part of an 1628 * active pool that was disconnected without being explicitly 1629 * exported. 1630 */ 1631 if (pool_active(hdl, name, guid, &isactive) != 0) { 1632 nvlist_free(config); 1633 return (-1); 1634 } 1635 1636 if (isactive) { 1637 /* 1638 * Because the device may have been removed while 1639 * offlined, we only report it as active if the vdev is 1640 * still present in the config. Otherwise, pretend like 1641 * it's not in use. 1642 */ 1643 if ((zhp = zpool_open_canfail(hdl, name)) != NULL && 1644 (pool_config = zpool_get_config(zhp, NULL)) 1645 != NULL) { 1646 nvlist_t *nvroot; 1647 1648 verify(nvlist_lookup_nvlist(pool_config, 1649 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 1650 ret = find_guid(nvroot, vdev_guid); 1651 } else { 1652 ret = B_FALSE; 1653 } 1654 1655 /* 1656 * If this is an active spare within another pool, we 1657 * treat it like an unused hot spare. This allows the 1658 * user to create a pool with a hot spare that currently 1659 * in use within another pool. Since we return B_TRUE, 1660 * libdiskmgt will continue to prevent generic consumers 1661 * from using the device. 1662 */ 1663 if (ret && nvlist_lookup_uint64(config, 1664 ZPOOL_CONFIG_IS_SPARE, &isspare) == 0 && isspare) 1665 stateval = POOL_STATE_SPARE; 1666 1667 if (zhp != NULL) 1668 zpool_close(zhp); 1669 } else { 1670 stateval = POOL_STATE_POTENTIALLY_ACTIVE; 1671 ret = B_TRUE; 1672 } 1673 break; 1674 1675 case POOL_STATE_SPARE: 1676 /* 1677 * For a hot spare, it can be either definitively in use, or 1678 * potentially active. To determine if it's in use, we iterate 1679 * over all pools in the system and search for one with a spare 1680 * with a matching guid. 1681 * 1682 * Due to the shared nature of spares, we don't actually report 1683 * the potentially active case as in use. This means the user 1684 * can freely create pools on the hot spares of exported pools, 1685 * but to do otherwise makes the resulting code complicated, and 1686 * we end up having to deal with this case anyway. 1687 */ 1688 cb.cb_zhp = NULL; 1689 cb.cb_guid = vdev_guid; 1690 cb.cb_type = ZPOOL_CONFIG_SPARES; 1691 if (zpool_iter(hdl, find_aux, &cb) == 1) { 1692 name = (char *)zpool_get_name(cb.cb_zhp); 1693 ret = TRUE; 1694 } else { 1695 ret = FALSE; 1696 } 1697 break; 1698 1699 case POOL_STATE_L2CACHE: 1700 1701 /* 1702 * Check if any pool is currently using this l2cache device. 1703 */ 1704 cb.cb_zhp = NULL; 1705 cb.cb_guid = vdev_guid; 1706 cb.cb_type = ZPOOL_CONFIG_L2CACHE; 1707 if (zpool_iter(hdl, find_aux, &cb) == 1) { 1708 name = (char *)zpool_get_name(cb.cb_zhp); 1709 ret = TRUE; 1710 } else { 1711 ret = FALSE; 1712 } 1713 break; 1714 1715 default: 1716 ret = B_FALSE; 1717 } 1718 1719 1720 if (ret) { 1721 if ((*namestr = zfs_strdup(hdl, name)) == NULL) { 1722 if (cb.cb_zhp) 1723 zpool_close(cb.cb_zhp); 1724 nvlist_free(config); 1725 return (-1); 1726 } 1727 *state = (pool_state_t)stateval; 1728 } 1729 1730 if (cb.cb_zhp) 1731 zpool_close(cb.cb_zhp); 1732 1733 nvlist_free(config); 1734 *inuse = ret; 1735 return (0); 1736} 1737