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 /* 23 * Copyright 2009 Sun Microsystems, Inc. All rights reserved. 24 * Use is subject to license terms. 25 */ 26 27 #include <ctype.h> 28 #include <errno.h> 29 #include <devid.h> 30 #include <fcntl.h> 31 #include <libintl.h> 32 #include <stdio.h> 33 #include <stdlib.h> 34 #include <strings.h> 35 #include <unistd.h> 36 #include <sys/efi_partition.h> 37 #include <sys/vtoc.h> 38 #include <sys/zfs_ioctl.h> 39 #include <dlfcn.h> 40 41 #include "zfs_namecheck.h" 42 #include "zfs_prop.h" 43 #include "libzfs_impl.h" 44 45 static int read_efi_label(nvlist_t *config, diskaddr_t *sb); 46 47 #if defined(__i386) || defined(__amd64) 48 #define BOOTCMD "installgrub(1M)" 49 #else 50 #define BOOTCMD "installboot(1M)" 51 #endif 52 53 #define DISK_ROOT "/dev/dsk" 54 #define RDISK_ROOT "/dev/rdsk" 55 #define BACKUP_SLICE "s2" 56 57 /* 58 * ==================================================================== 59 * zpool property functions 60 * ==================================================================== 61 */ 62 63 static int 64 zpool_get_all_props(zpool_handle_t *zhp) 65 { 66 zfs_cmd_t zc = { 0 }; 67 libzfs_handle_t *hdl = zhp->zpool_hdl; 68 69 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 70 71 if (zcmd_alloc_dst_nvlist(hdl, &zc, 0) != 0) 72 return (-1); 73 74 while (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_PROPS, &zc) != 0) { 75 if (errno == ENOMEM) { 76 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) { 77 zcmd_free_nvlists(&zc); 78 return (-1); 79 } 80 } else { 81 zcmd_free_nvlists(&zc); 82 return (-1); 83 } 84 } 85 86 if (zcmd_read_dst_nvlist(hdl, &zc, &zhp->zpool_props) != 0) { 87 zcmd_free_nvlists(&zc); 88 return (-1); 89 } 90 91 zcmd_free_nvlists(&zc); 92 93 return (0); 94 } 95 96 static int 97 zpool_props_refresh(zpool_handle_t *zhp) 98 { 99 nvlist_t *old_props; 100 101 old_props = zhp->zpool_props; 102 103 if (zpool_get_all_props(zhp) != 0) 104 return (-1); 105 106 nvlist_free(old_props); 107 return (0); 108 } 109 110 static char * 111 zpool_get_prop_string(zpool_handle_t *zhp, zpool_prop_t prop, 112 zprop_source_t *src) 113 { 114 nvlist_t *nv, *nvl; 115 uint64_t ival; 116 char *value; 117 zprop_source_t source; 118 119 nvl = zhp->zpool_props; 120 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) { 121 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &ival) == 0); 122 source = ival; 123 verify(nvlist_lookup_string(nv, ZPROP_VALUE, &value) == 0); 124 } else { 125 source = ZPROP_SRC_DEFAULT; 126 if ((value = (char *)zpool_prop_default_string(prop)) == NULL) 127 value = "-"; 128 } 129 130 if (src) 131 *src = source; 132 133 return (value); 134 } 135 136 uint64_t 137 zpool_get_prop_int(zpool_handle_t *zhp, zpool_prop_t prop, zprop_source_t *src) 138 { 139 nvlist_t *nv, *nvl; 140 uint64_t value; 141 zprop_source_t source; 142 143 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp)) { 144 /* 145 * zpool_get_all_props() has most likely failed because 146 * the pool is faulted, but if all we need is the top level 147 * vdev's guid then get it from the zhp config nvlist. 148 */ 149 if ((prop == ZPOOL_PROP_GUID) && 150 (nvlist_lookup_nvlist(zhp->zpool_config, 151 ZPOOL_CONFIG_VDEV_TREE, &nv) == 0) && 152 (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value) 153 == 0)) { 154 return (value); 155 } 156 return (zpool_prop_default_numeric(prop)); 157 } 158 159 nvl = zhp->zpool_props; 160 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) { 161 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &value) == 0); 162 source = value; 163 verify(nvlist_lookup_uint64(nv, ZPROP_VALUE, &value) == 0); 164 } else { 165 source = ZPROP_SRC_DEFAULT; 166 value = zpool_prop_default_numeric(prop); 167 } 168 169 if (src) 170 *src = source; 171 172 return (value); 173 } 174 175 /* 176 * Map VDEV STATE to printed strings. 177 */ 178 char * 179 zpool_state_to_name(vdev_state_t state, vdev_aux_t aux) 180 { 181 switch (state) { 182 case VDEV_STATE_CLOSED: 183 case VDEV_STATE_OFFLINE: 184 return (gettext("OFFLINE")); 185 case VDEV_STATE_REMOVED: 186 return (gettext("REMOVED")); 187 case VDEV_STATE_CANT_OPEN: 188 if (aux == VDEV_AUX_CORRUPT_DATA || aux == VDEV_AUX_BAD_LOG) 189 return (gettext("FAULTED")); 190 else 191 return (gettext("UNAVAIL")); 192 case VDEV_STATE_FAULTED: 193 return (gettext("FAULTED")); 194 case VDEV_STATE_DEGRADED: 195 return (gettext("DEGRADED")); 196 case VDEV_STATE_HEALTHY: 197 return (gettext("ONLINE")); 198 } 199 200 return (gettext("UNKNOWN")); 201 } 202 203 /* 204 * Get a zpool property value for 'prop' and return the value in 205 * a pre-allocated buffer. 206 */ 207 int 208 zpool_get_prop(zpool_handle_t *zhp, zpool_prop_t prop, char *buf, size_t len, 209 zprop_source_t *srctype) 210 { 211 uint64_t intval; 212 const char *strval; 213 zprop_source_t src = ZPROP_SRC_NONE; 214 nvlist_t *nvroot; 215 vdev_stat_t *vs; 216 uint_t vsc; 217 218 if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) { 219 switch (prop) { 220 case ZPOOL_PROP_NAME: 221 (void) strlcpy(buf, zpool_get_name(zhp), len); 222 break; 223 224 case ZPOOL_PROP_HEALTH: 225 (void) strlcpy(buf, "FAULTED", len); 226 break; 227 228 case ZPOOL_PROP_GUID: 229 intval = zpool_get_prop_int(zhp, prop, &src); 230 (void) snprintf(buf, len, "%llu", intval); 231 break; 232 233 case ZPOOL_PROP_ALTROOT: 234 case ZPOOL_PROP_CACHEFILE: 235 if (zhp->zpool_props != NULL || 236 zpool_get_all_props(zhp) == 0) { 237 (void) strlcpy(buf, 238 zpool_get_prop_string(zhp, prop, &src), 239 len); 240 if (srctype != NULL) 241 *srctype = src; 242 return (0); 243 } 244 /* FALLTHROUGH */ 245 default: 246 (void) strlcpy(buf, "-", len); 247 break; 248 } 249 250 if (srctype != NULL) 251 *srctype = src; 252 return (0); 253 } 254 255 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp) && 256 prop != ZPOOL_PROP_NAME) 257 return (-1); 258 259 switch (zpool_prop_get_type(prop)) { 260 case PROP_TYPE_STRING: 261 (void) strlcpy(buf, zpool_get_prop_string(zhp, prop, &src), 262 len); 263 break; 264 265 case PROP_TYPE_NUMBER: 266 intval = zpool_get_prop_int(zhp, prop, &src); 267 268 switch (prop) { 269 case ZPOOL_PROP_SIZE: 270 case ZPOOL_PROP_USED: 271 case ZPOOL_PROP_AVAILABLE: 272 (void) zfs_nicenum(intval, buf, len); 273 break; 274 275 case ZPOOL_PROP_CAPACITY: 276 (void) snprintf(buf, len, "%llu%%", 277 (u_longlong_t)intval); 278 break; 279 280 case ZPOOL_PROP_HEALTH: 281 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL), 282 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 283 verify(nvlist_lookup_uint64_array(nvroot, 284 ZPOOL_CONFIG_STATS, (uint64_t **)&vs, &vsc) == 0); 285 286 (void) strlcpy(buf, zpool_state_to_name(intval, 287 vs->vs_aux), len); 288 break; 289 default: 290 (void) snprintf(buf, len, "%llu", intval); 291 } 292 break; 293 294 case PROP_TYPE_INDEX: 295 intval = zpool_get_prop_int(zhp, prop, &src); 296 if (zpool_prop_index_to_string(prop, intval, &strval) 297 != 0) 298 return (-1); 299 (void) strlcpy(buf, strval, len); 300 break; 301 302 default: 303 abort(); 304 } 305 306 if (srctype) 307 *srctype = src; 308 309 return (0); 310 } 311 312 /* 313 * Check if the bootfs name has the same pool name as it is set to. 314 * Assuming bootfs is a valid dataset name. 315 */ 316 static boolean_t 317 bootfs_name_valid(const char *pool, char *bootfs) 318 { 319 int len = strlen(pool); 320 321 if (!zfs_name_valid(bootfs, ZFS_TYPE_FILESYSTEM|ZFS_TYPE_SNAPSHOT)) 322 return (B_FALSE); 323 324 if (strncmp(pool, bootfs, len) == 0 && 325 (bootfs[len] == '/' || bootfs[len] == '\0')) 326 return (B_TRUE); 327 328 return (B_FALSE); 329 } 330 331 /* 332 * Inspect the configuration to determine if any of the devices contain 333 * an EFI label. 334 */ 335 static boolean_t 336 pool_uses_efi(nvlist_t *config) 337 { 338 nvlist_t **child; 339 uint_t c, children; 340 341 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN, 342 &child, &children) != 0) 343 return (read_efi_label(config, NULL) >= 0); 344 345 for (c = 0; c < children; c++) { 346 if (pool_uses_efi(child[c])) 347 return (B_TRUE); 348 } 349 return (B_FALSE); 350 } 351 352 static boolean_t 353 pool_is_bootable(zpool_handle_t *zhp) 354 { 355 char bootfs[ZPOOL_MAXNAMELEN]; 356 357 return (zpool_get_prop(zhp, ZPOOL_PROP_BOOTFS, bootfs, 358 sizeof (bootfs), NULL) == 0 && strncmp(bootfs, "-", 359 sizeof (bootfs)) != 0); 360 } 361 362 363 /* 364 * Given an nvlist of zpool properties to be set, validate that they are 365 * correct, and parse any numeric properties (index, boolean, etc) if they are 366 * specified as strings. 367 */ 368 static nvlist_t * 369 zpool_valid_proplist(libzfs_handle_t *hdl, const char *poolname, 370 nvlist_t *props, uint64_t version, boolean_t create_or_import, char *errbuf) 371 { 372 nvpair_t *elem; 373 nvlist_t *retprops; 374 zpool_prop_t prop; 375 char *strval; 376 uint64_t intval; 377 char *slash; 378 struct stat64 statbuf; 379 zpool_handle_t *zhp; 380 nvlist_t *nvroot; 381 382 if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) { 383 (void) no_memory(hdl); 384 return (NULL); 385 } 386 387 elem = NULL; 388 while ((elem = nvlist_next_nvpair(props, elem)) != NULL) { 389 const char *propname = nvpair_name(elem); 390 391 /* 392 * Make sure this property is valid and applies to this type. 393 */ 394 if ((prop = zpool_name_to_prop(propname)) == ZPROP_INVAL) { 395 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 396 "invalid property '%s'"), propname); 397 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 398 goto error; 399 } 400 401 if (zpool_prop_readonly(prop)) { 402 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' " 403 "is readonly"), propname); 404 (void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf); 405 goto error; 406 } 407 408 if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_POOL, retprops, 409 &strval, &intval, errbuf) != 0) 410 goto error; 411 412 /* 413 * Perform additional checking for specific properties. 414 */ 415 switch (prop) { 416 case ZPOOL_PROP_VERSION: 417 if (intval < version || intval > SPA_VERSION) { 418 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 419 "property '%s' number %d is invalid."), 420 propname, intval); 421 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf); 422 goto error; 423 } 424 break; 425 426 case ZPOOL_PROP_BOOTFS: 427 if (create_or_import) { 428 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 429 "property '%s' cannot be set at creation " 430 "or import time"), propname); 431 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 432 goto error; 433 } 434 435 if (version < SPA_VERSION_BOOTFS) { 436 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 437 "pool must be upgraded to support " 438 "'%s' property"), propname); 439 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf); 440 goto error; 441 } 442 443 /* 444 * bootfs property value has to be a dataset name and 445 * the dataset has to be in the same pool as it sets to. 446 */ 447 if (strval[0] != '\0' && !bootfs_name_valid(poolname, 448 strval)) { 449 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' " 450 "is an invalid name"), strval); 451 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf); 452 goto error; 453 } 454 455 if ((zhp = zpool_open_canfail(hdl, poolname)) == NULL) { 456 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 457 "could not open pool '%s'"), poolname); 458 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf); 459 goto error; 460 } 461 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL), 462 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 463 464 /* 465 * bootfs property cannot be set on a disk which has 466 * been EFI labeled. 467 */ 468 if (pool_uses_efi(nvroot)) { 469 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 470 "property '%s' not supported on " 471 "EFI labeled devices"), propname); 472 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf); 473 zpool_close(zhp); 474 goto error; 475 } 476 zpool_close(zhp); 477 break; 478 479 case ZPOOL_PROP_ALTROOT: 480 if (!create_or_import) { 481 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 482 "property '%s' can only be set during pool " 483 "creation or import"), propname); 484 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 485 goto error; 486 } 487 488 if (strval[0] != '/') { 489 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 490 "bad alternate root '%s'"), strval); 491 (void) zfs_error(hdl, EZFS_BADPATH, errbuf); 492 goto error; 493 } 494 break; 495 496 case ZPOOL_PROP_CACHEFILE: 497 if (strval[0] == '\0') 498 break; 499 500 if (strcmp(strval, "none") == 0) 501 break; 502 503 if (strval[0] != '/') { 504 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 505 "property '%s' must be empty, an " 506 "absolute path, or 'none'"), propname); 507 (void) zfs_error(hdl, EZFS_BADPATH, errbuf); 508 goto error; 509 } 510 511 slash = strrchr(strval, '/'); 512 513 if (slash[1] == '\0' || strcmp(slash, "/.") == 0 || 514 strcmp(slash, "/..") == 0) { 515 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 516 "'%s' is not a valid file"), strval); 517 (void) zfs_error(hdl, EZFS_BADPATH, errbuf); 518 goto error; 519 } 520 521 *slash = '\0'; 522 523 if (strval[0] != '\0' && 524 (stat64(strval, &statbuf) != 0 || 525 !S_ISDIR(statbuf.st_mode))) { 526 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 527 "'%s' is not a valid directory"), 528 strval); 529 (void) zfs_error(hdl, EZFS_BADPATH, errbuf); 530 goto error; 531 } 532 533 *slash = '/'; 534 break; 535 } 536 } 537 538 return (retprops); 539 error: 540 nvlist_free(retprops); 541 return (NULL); 542 } 543 544 /* 545 * Set zpool property : propname=propval. 546 */ 547 int 548 zpool_set_prop(zpool_handle_t *zhp, const char *propname, const char *propval) 549 { 550 zfs_cmd_t zc = { 0 }; 551 int ret = -1; 552 char errbuf[1024]; 553 nvlist_t *nvl = NULL; 554 nvlist_t *realprops; 555 uint64_t version; 556 557 (void) snprintf(errbuf, sizeof (errbuf), 558 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"), 559 zhp->zpool_name); 560 561 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0) 562 return (no_memory(zhp->zpool_hdl)); 563 564 if (nvlist_add_string(nvl, propname, propval) != 0) { 565 nvlist_free(nvl); 566 return (no_memory(zhp->zpool_hdl)); 567 } 568 569 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL); 570 if ((realprops = zpool_valid_proplist(zhp->zpool_hdl, 571 zhp->zpool_name, nvl, version, B_FALSE, errbuf)) == NULL) { 572 nvlist_free(nvl); 573 return (-1); 574 } 575 576 nvlist_free(nvl); 577 nvl = realprops; 578 579 /* 580 * Execute the corresponding ioctl() to set this property. 581 */ 582 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 583 584 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) { 585 nvlist_free(nvl); 586 return (-1); 587 } 588 589 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc); 590 591 zcmd_free_nvlists(&zc); 592 nvlist_free(nvl); 593 594 if (ret) 595 (void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf); 596 else 597 (void) zpool_props_refresh(zhp); 598 599 return (ret); 600 } 601 602 int 603 zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp) 604 { 605 libzfs_handle_t *hdl = zhp->zpool_hdl; 606 zprop_list_t *entry; 607 char buf[ZFS_MAXPROPLEN]; 608 609 if (zprop_expand_list(hdl, plp, ZFS_TYPE_POOL) != 0) 610 return (-1); 611 612 for (entry = *plp; entry != NULL; entry = entry->pl_next) { 613 614 if (entry->pl_fixed) 615 continue; 616 617 if (entry->pl_prop != ZPROP_INVAL && 618 zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf), 619 NULL) == 0) { 620 if (strlen(buf) > entry->pl_width) 621 entry->pl_width = strlen(buf); 622 } 623 } 624 625 return (0); 626 } 627 628 629 /* 630 * Don't start the slice at the default block of 34; many storage 631 * devices will use a stripe width of 128k, so start there instead. 632 */ 633 #define NEW_START_BLOCK 256 634 635 /* 636 * Validate the given pool name, optionally putting an extended error message in 637 * 'buf'. 638 */ 639 boolean_t 640 zpool_name_valid(libzfs_handle_t *hdl, boolean_t isopen, const char *pool) 641 { 642 namecheck_err_t why; 643 char what; 644 int ret; 645 646 ret = pool_namecheck(pool, &why, &what); 647 648 /* 649 * The rules for reserved pool names were extended at a later point. 650 * But we need to support users with existing pools that may now be 651 * invalid. So we only check for this expanded set of names during a 652 * create (or import), and only in userland. 653 */ 654 if (ret == 0 && !isopen && 655 (strncmp(pool, "mirror", 6) == 0 || 656 strncmp(pool, "raidz", 5) == 0 || 657 strncmp(pool, "spare", 5) == 0 || 658 strcmp(pool, "log") == 0)) { 659 if (hdl != NULL) 660 zfs_error_aux(hdl, 661 dgettext(TEXT_DOMAIN, "name is reserved")); 662 return (B_FALSE); 663 } 664 665 666 if (ret != 0) { 667 if (hdl != NULL) { 668 switch (why) { 669 case NAME_ERR_TOOLONG: 670 zfs_error_aux(hdl, 671 dgettext(TEXT_DOMAIN, "name is too long")); 672 break; 673 674 case NAME_ERR_INVALCHAR: 675 zfs_error_aux(hdl, 676 dgettext(TEXT_DOMAIN, "invalid character " 677 "'%c' in pool name"), what); 678 break; 679 680 case NAME_ERR_NOLETTER: 681 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 682 "name must begin with a letter")); 683 break; 684 685 case NAME_ERR_RESERVED: 686 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 687 "name is reserved")); 688 break; 689 690 case NAME_ERR_DISKLIKE: 691 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 692 "pool name is reserved")); 693 break; 694 695 case NAME_ERR_LEADING_SLASH: 696 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 697 "leading slash in name")); 698 break; 699 700 case NAME_ERR_EMPTY_COMPONENT: 701 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 702 "empty component in name")); 703 break; 704 705 case NAME_ERR_TRAILING_SLASH: 706 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 707 "trailing slash in name")); 708 break; 709 710 case NAME_ERR_MULTIPLE_AT: 711 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 712 "multiple '@' delimiters in name")); 713 break; 714 715 } 716 } 717 return (B_FALSE); 718 } 719 720 return (B_TRUE); 721 } 722 723 /* 724 * Open a handle to the given pool, even if the pool is currently in the FAULTED 725 * state. 726 */ 727 zpool_handle_t * 728 zpool_open_canfail(libzfs_handle_t *hdl, const char *pool) 729 { 730 zpool_handle_t *zhp; 731 boolean_t missing; 732 733 /* 734 * Make sure the pool name is valid. 735 */ 736 if (!zpool_name_valid(hdl, B_TRUE, pool)) { 737 (void) zfs_error_fmt(hdl, EZFS_INVALIDNAME, 738 dgettext(TEXT_DOMAIN, "cannot open '%s'"), 739 pool); 740 return (NULL); 741 } 742 743 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL) 744 return (NULL); 745 746 zhp->zpool_hdl = hdl; 747 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name)); 748 749 if (zpool_refresh_stats(zhp, &missing) != 0) { 750 zpool_close(zhp); 751 return (NULL); 752 } 753 754 if (missing) { 755 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool")); 756 (void) zfs_error_fmt(hdl, EZFS_NOENT, 757 dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool); 758 zpool_close(zhp); 759 return (NULL); 760 } 761 762 return (zhp); 763 } 764 765 /* 766 * Like the above, but silent on error. Used when iterating over pools (because 767 * the configuration cache may be out of date). 768 */ 769 int 770 zpool_open_silent(libzfs_handle_t *hdl, const char *pool, zpool_handle_t **ret) 771 { 772 zpool_handle_t *zhp; 773 boolean_t missing; 774 775 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL) 776 return (-1); 777 778 zhp->zpool_hdl = hdl; 779 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name)); 780 781 if (zpool_refresh_stats(zhp, &missing) != 0) { 782 zpool_close(zhp); 783 return (-1); 784 } 785 786 if (missing) { 787 zpool_close(zhp); 788 *ret = NULL; 789 return (0); 790 } 791 792 *ret = zhp; 793 return (0); 794 } 795 796 /* 797 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted 798 * state. 799 */ 800 zpool_handle_t * 801 zpool_open(libzfs_handle_t *hdl, const char *pool) 802 { 803 zpool_handle_t *zhp; 804 805 if ((zhp = zpool_open_canfail(hdl, pool)) == NULL) 806 return (NULL); 807 808 if (zhp->zpool_state == POOL_STATE_UNAVAIL) { 809 (void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL, 810 dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name); 811 zpool_close(zhp); 812 return (NULL); 813 } 814 815 return (zhp); 816 } 817 818 /* 819 * Close the handle. Simply frees the memory associated with the handle. 820 */ 821 void 822 zpool_close(zpool_handle_t *zhp) 823 { 824 if (zhp->zpool_config) 825 nvlist_free(zhp->zpool_config); 826 if (zhp->zpool_old_config) 827 nvlist_free(zhp->zpool_old_config); 828 if (zhp->zpool_props) 829 nvlist_free(zhp->zpool_props); 830 free(zhp); 831 } 832 833 /* 834 * Return the name of the pool. 835 */ 836 const char * 837 zpool_get_name(zpool_handle_t *zhp) 838 { 839 return (zhp->zpool_name); 840 } 841 842 843 /* 844 * Return the state of the pool (ACTIVE or UNAVAILABLE) 845 */ 846 int 847 zpool_get_state(zpool_handle_t *zhp) 848 { 849 return (zhp->zpool_state); 850 } 851 852 /* 853 * Create the named pool, using the provided vdev list. It is assumed 854 * that the consumer has already validated the contents of the nvlist, so we 855 * don't have to worry about error semantics. 856 */ 857 int 858 zpool_create(libzfs_handle_t *hdl, const char *pool, nvlist_t *nvroot, 859 nvlist_t *props, nvlist_t *fsprops) 860 { 861 zfs_cmd_t zc = { 0 }; 862 nvlist_t *zc_fsprops = NULL; 863 nvlist_t *zc_props = NULL; 864 char msg[1024]; 865 char *altroot; 866 int ret = -1; 867 868 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 869 "cannot create '%s'"), pool); 870 871 if (!zpool_name_valid(hdl, B_FALSE, pool)) 872 return (zfs_error(hdl, EZFS_INVALIDNAME, msg)); 873 874 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0) 875 return (-1); 876 877 if (props) { 878 if ((zc_props = zpool_valid_proplist(hdl, pool, props, 879 SPA_VERSION_1, B_TRUE, msg)) == NULL) { 880 goto create_failed; 881 } 882 } 883 884 if (fsprops) { 885 uint64_t zoned; 886 char *zonestr; 887 888 zoned = ((nvlist_lookup_string(fsprops, 889 zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) && 890 strcmp(zonestr, "on") == 0); 891 892 if ((zc_fsprops = zfs_valid_proplist(hdl, 893 ZFS_TYPE_FILESYSTEM, fsprops, zoned, NULL, msg)) == NULL) { 894 goto create_failed; 895 } 896 if (!zc_props && 897 (nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) { 898 goto create_failed; 899 } 900 if (nvlist_add_nvlist(zc_props, 901 ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) { 902 goto create_failed; 903 } 904 } 905 906 if (zc_props && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0) 907 goto create_failed; 908 909 (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name)); 910 911 if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) { 912 913 zcmd_free_nvlists(&zc); 914 nvlist_free(zc_props); 915 nvlist_free(zc_fsprops); 916 917 switch (errno) { 918 case EBUSY: 919 /* 920 * This can happen if the user has specified the same 921 * device multiple times. We can't reliably detect this 922 * until we try to add it and see we already have a 923 * label. 924 */ 925 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 926 "one or more vdevs refer to the same device")); 927 return (zfs_error(hdl, EZFS_BADDEV, msg)); 928 929 case EOVERFLOW: 930 /* 931 * This occurs when one of the devices is below 932 * SPA_MINDEVSIZE. Unfortunately, we can't detect which 933 * device was the problem device since there's no 934 * reliable way to determine device size from userland. 935 */ 936 { 937 char buf[64]; 938 939 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf)); 940 941 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 942 "one or more devices is less than the " 943 "minimum size (%s)"), buf); 944 } 945 return (zfs_error(hdl, EZFS_BADDEV, msg)); 946 947 case ENOSPC: 948 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 949 "one or more devices is out of space")); 950 return (zfs_error(hdl, EZFS_BADDEV, msg)); 951 952 case ENOTBLK: 953 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 954 "cache device must be a disk or disk slice")); 955 return (zfs_error(hdl, EZFS_BADDEV, msg)); 956 957 default: 958 return (zpool_standard_error(hdl, errno, msg)); 959 } 960 } 961 962 /* 963 * If this is an alternate root pool, then we automatically set the 964 * mountpoint of the root dataset to be '/'. 965 */ 966 if (nvlist_lookup_string(props, zpool_prop_to_name(ZPOOL_PROP_ALTROOT), 967 &altroot) == 0) { 968 zfs_handle_t *zhp; 969 970 verify((zhp = zfs_open(hdl, pool, ZFS_TYPE_DATASET)) != NULL); 971 verify(zfs_prop_set(zhp, zfs_prop_to_name(ZFS_PROP_MOUNTPOINT), 972 "/") == 0); 973 974 zfs_close(zhp); 975 } 976 977 create_failed: 978 zcmd_free_nvlists(&zc); 979 nvlist_free(zc_props); 980 nvlist_free(zc_fsprops); 981 return (ret); 982 } 983 984 /* 985 * Destroy the given pool. It is up to the caller to ensure that there are no 986 * datasets left in the pool. 987 */ 988 int 989 zpool_destroy(zpool_handle_t *zhp) 990 { 991 zfs_cmd_t zc = { 0 }; 992 zfs_handle_t *zfp = NULL; 993 libzfs_handle_t *hdl = zhp->zpool_hdl; 994 char msg[1024]; 995 996 if (zhp->zpool_state == POOL_STATE_ACTIVE && 997 (zfp = zfs_open(zhp->zpool_hdl, zhp->zpool_name, 998 ZFS_TYPE_FILESYSTEM)) == NULL) 999 return (-1); 1000 1001 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1002 1003 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) { 1004 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1005 "cannot destroy '%s'"), zhp->zpool_name); 1006 1007 if (errno == EROFS) { 1008 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1009 "one or more devices is read only")); 1010 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1011 } else { 1012 (void) zpool_standard_error(hdl, errno, msg); 1013 } 1014 1015 if (zfp) 1016 zfs_close(zfp); 1017 return (-1); 1018 } 1019 1020 if (zfp) { 1021 remove_mountpoint(zfp); 1022 zfs_close(zfp); 1023 } 1024 1025 return (0); 1026 } 1027 1028 /* 1029 * Add the given vdevs to the pool. The caller must have already performed the 1030 * necessary verification to ensure that the vdev specification is well-formed. 1031 */ 1032 int 1033 zpool_add(zpool_handle_t *zhp, nvlist_t *nvroot) 1034 { 1035 zfs_cmd_t zc = { 0 }; 1036 int ret; 1037 libzfs_handle_t *hdl = zhp->zpool_hdl; 1038 char msg[1024]; 1039 nvlist_t **spares, **l2cache; 1040 uint_t nspares, nl2cache; 1041 1042 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1043 "cannot add to '%s'"), zhp->zpool_name); 1044 1045 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) < 1046 SPA_VERSION_SPARES && 1047 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, 1048 &spares, &nspares) == 0) { 1049 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be " 1050 "upgraded to add hot spares")); 1051 return (zfs_error(hdl, EZFS_BADVERSION, msg)); 1052 } 1053 1054 if (pool_is_bootable(zhp) && nvlist_lookup_nvlist_array(nvroot, 1055 ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0) { 1056 uint64_t s; 1057 1058 for (s = 0; s < nspares; s++) { 1059 char *path; 1060 1061 if (nvlist_lookup_string(spares[s], ZPOOL_CONFIG_PATH, 1062 &path) == 0 && pool_uses_efi(spares[s])) { 1063 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1064 "device '%s' contains an EFI label and " 1065 "cannot be used on root pools."), 1066 zpool_vdev_name(hdl, NULL, spares[s])); 1067 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg)); 1068 } 1069 } 1070 } 1071 1072 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) < 1073 SPA_VERSION_L2CACHE && 1074 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, 1075 &l2cache, &nl2cache) == 0) { 1076 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be " 1077 "upgraded to add cache devices")); 1078 return (zfs_error(hdl, EZFS_BADVERSION, msg)); 1079 } 1080 1081 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0) 1082 return (-1); 1083 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1084 1085 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) { 1086 switch (errno) { 1087 case EBUSY: 1088 /* 1089 * This can happen if the user has specified the same 1090 * device multiple times. We can't reliably detect this 1091 * until we try to add it and see we already have a 1092 * label. 1093 */ 1094 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1095 "one or more vdevs refer to the same device")); 1096 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1097 break; 1098 1099 case EOVERFLOW: 1100 /* 1101 * This occurrs when one of the devices is below 1102 * SPA_MINDEVSIZE. Unfortunately, we can't detect which 1103 * device was the problem device since there's no 1104 * reliable way to determine device size from userland. 1105 */ 1106 { 1107 char buf[64]; 1108 1109 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf)); 1110 1111 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1112 "device is less than the minimum " 1113 "size (%s)"), buf); 1114 } 1115 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1116 break; 1117 1118 case ENOTSUP: 1119 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1120 "pool must be upgraded to add these vdevs")); 1121 (void) zfs_error(hdl, EZFS_BADVERSION, msg); 1122 break; 1123 1124 case EDOM: 1125 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1126 "root pool can not have multiple vdevs" 1127 " or separate logs")); 1128 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, msg); 1129 break; 1130 1131 case ENOTBLK: 1132 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1133 "cache device must be a disk or disk slice")); 1134 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1135 break; 1136 1137 default: 1138 (void) zpool_standard_error(hdl, errno, msg); 1139 } 1140 1141 ret = -1; 1142 } else { 1143 ret = 0; 1144 } 1145 1146 zcmd_free_nvlists(&zc); 1147 1148 return (ret); 1149 } 1150 1151 /* 1152 * Exports the pool from the system. The caller must ensure that there are no 1153 * mounted datasets in the pool. 1154 */ 1155 int 1156 zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce) 1157 { 1158 zfs_cmd_t zc = { 0 }; 1159 char msg[1024]; 1160 1161 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1162 "cannot export '%s'"), zhp->zpool_name); 1163 1164 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1165 zc.zc_cookie = force; 1166 zc.zc_guid = hardforce; 1167 1168 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) { 1169 switch (errno) { 1170 case EXDEV: 1171 zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN, 1172 "use '-f' to override the following errors:\n" 1173 "'%s' has an active shared spare which could be" 1174 " used by other pools once '%s' is exported."), 1175 zhp->zpool_name, zhp->zpool_name); 1176 return (zfs_error(zhp->zpool_hdl, EZFS_ACTIVE_SPARE, 1177 msg)); 1178 default: 1179 return (zpool_standard_error_fmt(zhp->zpool_hdl, errno, 1180 msg)); 1181 } 1182 } 1183 1184 return (0); 1185 } 1186 1187 int 1188 zpool_export(zpool_handle_t *zhp, boolean_t force) 1189 { 1190 return (zpool_export_common(zhp, force, B_FALSE)); 1191 } 1192 1193 int 1194 zpool_export_force(zpool_handle_t *zhp) 1195 { 1196 return (zpool_export_common(zhp, B_TRUE, B_TRUE)); 1197 } 1198 1199 /* 1200 * zpool_import() is a contracted interface. Should be kept the same 1201 * if possible. 1202 * 1203 * Applications should use zpool_import_props() to import a pool with 1204 * new properties value to be set. 1205 */ 1206 int 1207 zpool_import(libzfs_handle_t *hdl, nvlist_t *config, const char *newname, 1208 char *altroot) 1209 { 1210 nvlist_t *props = NULL; 1211 int ret; 1212 1213 if (altroot != NULL) { 1214 if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) { 1215 return (zfs_error_fmt(hdl, EZFS_NOMEM, 1216 dgettext(TEXT_DOMAIN, "cannot import '%s'"), 1217 newname)); 1218 } 1219 1220 if (nvlist_add_string(props, 1221 zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 || 1222 nvlist_add_string(props, 1223 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) { 1224 nvlist_free(props); 1225 return (zfs_error_fmt(hdl, EZFS_NOMEM, 1226 dgettext(TEXT_DOMAIN, "cannot import '%s'"), 1227 newname)); 1228 } 1229 } 1230 1231 ret = zpool_import_props(hdl, config, newname, props, B_FALSE); 1232 if (props) 1233 nvlist_free(props); 1234 return (ret); 1235 } 1236 1237 /* 1238 * Import the given pool using the known configuration and a list of 1239 * properties to be set. The configuration should have come from 1240 * zpool_find_import(). The 'newname' parameters control whether the pool 1241 * is imported with a different name. 1242 */ 1243 int 1244 zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname, 1245 nvlist_t *props, boolean_t importfaulted) 1246 { 1247 zfs_cmd_t zc = { 0 }; 1248 char *thename; 1249 char *origname; 1250 int ret; 1251 char errbuf[1024]; 1252 1253 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, 1254 &origname) == 0); 1255 1256 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, 1257 "cannot import pool '%s'"), origname); 1258 1259 if (newname != NULL) { 1260 if (!zpool_name_valid(hdl, B_FALSE, newname)) 1261 return (zfs_error_fmt(hdl, EZFS_INVALIDNAME, 1262 dgettext(TEXT_DOMAIN, "cannot import '%s'"), 1263 newname)); 1264 thename = (char *)newname; 1265 } else { 1266 thename = origname; 1267 } 1268 1269 if (props) { 1270 uint64_t version; 1271 1272 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, 1273 &version) == 0); 1274 1275 if ((props = zpool_valid_proplist(hdl, origname, 1276 props, version, B_TRUE, errbuf)) == NULL) { 1277 return (-1); 1278 } else if (zcmd_write_src_nvlist(hdl, &zc, props) != 0) { 1279 nvlist_free(props); 1280 return (-1); 1281 } 1282 } 1283 1284 (void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name)); 1285 1286 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, 1287 &zc.zc_guid) == 0); 1288 1289 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) { 1290 nvlist_free(props); 1291 return (-1); 1292 } 1293 1294 zc.zc_cookie = (uint64_t)importfaulted; 1295 ret = 0; 1296 if (zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc) != 0) { 1297 char desc[1024]; 1298 if (newname == NULL) 1299 (void) snprintf(desc, sizeof (desc), 1300 dgettext(TEXT_DOMAIN, "cannot import '%s'"), 1301 thename); 1302 else 1303 (void) snprintf(desc, sizeof (desc), 1304 dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"), 1305 origname, thename); 1306 1307 switch (errno) { 1308 case ENOTSUP: 1309 /* 1310 * Unsupported version. 1311 */ 1312 (void) zfs_error(hdl, EZFS_BADVERSION, desc); 1313 break; 1314 1315 case EINVAL: 1316 (void) zfs_error(hdl, EZFS_INVALCONFIG, desc); 1317 break; 1318 1319 default: 1320 (void) zpool_standard_error(hdl, errno, desc); 1321 } 1322 1323 ret = -1; 1324 } else { 1325 zpool_handle_t *zhp; 1326 1327 /* 1328 * This should never fail, but play it safe anyway. 1329 */ 1330 if (zpool_open_silent(hdl, thename, &zhp) != 0) 1331 ret = -1; 1332 else if (zhp != NULL) 1333 zpool_close(zhp); 1334 } 1335 1336 zcmd_free_nvlists(&zc); 1337 nvlist_free(props); 1338 1339 return (ret); 1340 } 1341 1342 /* 1343 * Scrub the pool. 1344 */ 1345 int 1346 zpool_scrub(zpool_handle_t *zhp, pool_scrub_type_t type) 1347 { 1348 zfs_cmd_t zc = { 0 }; 1349 char msg[1024]; 1350 libzfs_handle_t *hdl = zhp->zpool_hdl; 1351 1352 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1353 zc.zc_cookie = type; 1354 1355 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SCRUB, &zc) == 0) 1356 return (0); 1357 1358 (void) snprintf(msg, sizeof (msg), 1359 dgettext(TEXT_DOMAIN, "cannot scrub %s"), zc.zc_name); 1360 1361 if (errno == EBUSY) 1362 return (zfs_error(hdl, EZFS_RESILVERING, msg)); 1363 else 1364 return (zpool_standard_error(hdl, errno, msg)); 1365 } 1366 1367 /* 1368 * Find a vdev that matches the search criteria specified. We use the 1369 * the nvpair name to determine how we should look for the device. 1370 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL 1371 * spare; but FALSE if its an INUSE spare. 1372 */ 1373 static nvlist_t * 1374 vdev_to_nvlist_iter(nvlist_t *nv, nvlist_t *search, boolean_t *avail_spare, 1375 boolean_t *l2cache, boolean_t *log) 1376 { 1377 uint_t c, children; 1378 nvlist_t **child; 1379 nvlist_t *ret; 1380 uint64_t is_log; 1381 char *srchkey; 1382 nvpair_t *pair = nvlist_next_nvpair(search, NULL); 1383 1384 /* Nothing to look for */ 1385 if (search == NULL || pair == NULL) 1386 return (NULL); 1387 1388 /* Obtain the key we will use to search */ 1389 srchkey = nvpair_name(pair); 1390 1391 switch (nvpair_type(pair)) { 1392 case DATA_TYPE_UINT64: { 1393 uint64_t srchval, theguid, present; 1394 1395 verify(nvpair_value_uint64(pair, &srchval) == 0); 1396 if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) { 1397 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, 1398 &present) == 0) { 1399 /* 1400 * If the device has never been present since 1401 * import, the only reliable way to match the 1402 * vdev is by GUID. 1403 */ 1404 verify(nvlist_lookup_uint64(nv, 1405 ZPOOL_CONFIG_GUID, &theguid) == 0); 1406 if (theguid == srchval) 1407 return (nv); 1408 } 1409 } 1410 break; 1411 } 1412 1413 case DATA_TYPE_STRING: { 1414 char *srchval, *val; 1415 1416 verify(nvpair_value_string(pair, &srchval) == 0); 1417 if (nvlist_lookup_string(nv, srchkey, &val) != 0) 1418 break; 1419 1420 /* 1421 * Search for the requested value. We special case the search 1422 * for ZPOOL_CONFIG_PATH when it's a wholedisk. Otherwise, 1423 * all other searches are simple string compares. 1424 */ 1425 if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0 && val) { 1426 uint64_t wholedisk = 0; 1427 1428 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, 1429 &wholedisk); 1430 if (wholedisk) { 1431 /* 1432 * For whole disks, the internal path has 's0', 1433 * but the path passed in by the user doesn't. 1434 */ 1435 if (strlen(srchval) == strlen(val) - 2 && 1436 strncmp(srchval, val, strlen(srchval)) == 0) 1437 return (nv); 1438 break; 1439 } 1440 } 1441 1442 /* 1443 * Common case 1444 */ 1445 if (strcmp(srchval, val) == 0) 1446 return (nv); 1447 break; 1448 } 1449 1450 default: 1451 break; 1452 } 1453 1454 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, 1455 &child, &children) != 0) 1456 return (NULL); 1457 1458 for (c = 0; c < children; c++) { 1459 if ((ret = vdev_to_nvlist_iter(child[c], search, 1460 avail_spare, l2cache, NULL)) != NULL) { 1461 /* 1462 * The 'is_log' value is only set for the toplevel 1463 * vdev, not the leaf vdevs. So we always lookup the 1464 * log device from the root of the vdev tree (where 1465 * 'log' is non-NULL). 1466 */ 1467 if (log != NULL && 1468 nvlist_lookup_uint64(child[c], 1469 ZPOOL_CONFIG_IS_LOG, &is_log) == 0 && 1470 is_log) { 1471 *log = B_TRUE; 1472 } 1473 return (ret); 1474 } 1475 } 1476 1477 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES, 1478 &child, &children) == 0) { 1479 for (c = 0; c < children; c++) { 1480 if ((ret = vdev_to_nvlist_iter(child[c], search, 1481 avail_spare, l2cache, NULL)) != NULL) { 1482 *avail_spare = B_TRUE; 1483 return (ret); 1484 } 1485 } 1486 } 1487 1488 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE, 1489 &child, &children) == 0) { 1490 for (c = 0; c < children; c++) { 1491 if ((ret = vdev_to_nvlist_iter(child[c], search, 1492 avail_spare, l2cache, NULL)) != NULL) { 1493 *l2cache = B_TRUE; 1494 return (ret); 1495 } 1496 } 1497 } 1498 1499 return (NULL); 1500 } 1501 1502 /* 1503 * Given a physical path (minus the "/devices" prefix), find the 1504 * associated vdev. 1505 */ 1506 nvlist_t * 1507 zpool_find_vdev_by_physpath(zpool_handle_t *zhp, const char *ppath, 1508 boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log) 1509 { 1510 nvlist_t *search, *nvroot, *ret; 1511 1512 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1513 verify(nvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH, ppath) == 0); 1514 1515 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE, 1516 &nvroot) == 0); 1517 1518 *avail_spare = B_FALSE; 1519 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log); 1520 nvlist_free(search); 1521 1522 return (ret); 1523 } 1524 1525 nvlist_t * 1526 zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare, 1527 boolean_t *l2cache, boolean_t *log) 1528 { 1529 char buf[MAXPATHLEN]; 1530 char *end; 1531 nvlist_t *nvroot, *search, *ret; 1532 uint64_t guid; 1533 1534 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1535 1536 guid = strtoull(path, &end, 10); 1537 if (guid != 0 && *end == '\0') { 1538 verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0); 1539 } else if (path[0] != '/') { 1540 (void) snprintf(buf, sizeof (buf), "%s%s", "/dev/dsk/", path); 1541 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, buf) == 0); 1542 } else { 1543 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, path) == 0); 1544 } 1545 1546 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE, 1547 &nvroot) == 0); 1548 1549 *avail_spare = B_FALSE; 1550 *l2cache = B_FALSE; 1551 if (log != NULL) 1552 *log = B_FALSE; 1553 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log); 1554 nvlist_free(search); 1555 1556 return (ret); 1557 } 1558 1559 static int 1560 vdev_online(nvlist_t *nv) 1561 { 1562 uint64_t ival; 1563 1564 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE, &ival) == 0 || 1565 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED, &ival) == 0 || 1566 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED, &ival) == 0) 1567 return (0); 1568 1569 return (1); 1570 } 1571 1572 /* 1573 * Helper function for zpool_get_physpaths(). 1574 */ 1575 static int 1576 vdev_get_one_physpath(nvlist_t *config, char *physpath, size_t physpath_size, 1577 size_t *bytes_written) 1578 { 1579 size_t bytes_left, pos, rsz; 1580 char *tmppath; 1581 const char *format; 1582 1583 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PHYS_PATH, 1584 &tmppath) != 0) 1585 return (EZFS_NODEVICE); 1586 1587 pos = *bytes_written; 1588 bytes_left = physpath_size - pos; 1589 format = (pos == 0) ? "%s" : " %s"; 1590 1591 rsz = snprintf(physpath + pos, bytes_left, format, tmppath); 1592 *bytes_written += rsz; 1593 1594 if (rsz >= bytes_left) { 1595 /* if physpath was not copied properly, clear it */ 1596 if (bytes_left != 0) { 1597 physpath[pos] = 0; 1598 } 1599 return (EZFS_NOSPC); 1600 } 1601 return (0); 1602 } 1603 1604 static int 1605 vdev_get_physpaths(nvlist_t *nv, char *physpath, size_t phypath_size, 1606 size_t *rsz, boolean_t is_spare) 1607 { 1608 char *type; 1609 int ret; 1610 1611 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0) 1612 return (EZFS_INVALCONFIG); 1613 1614 if (strcmp(type, VDEV_TYPE_DISK) == 0) { 1615 /* 1616 * An active spare device has ZPOOL_CONFIG_IS_SPARE set. 1617 * For a spare vdev, we only want to boot from the active 1618 * spare device. 1619 */ 1620 if (is_spare) { 1621 uint64_t spare = 0; 1622 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE, 1623 &spare); 1624 if (!spare) 1625 return (EZFS_INVALCONFIG); 1626 } 1627 1628 if (vdev_online(nv)) { 1629 if ((ret = vdev_get_one_physpath(nv, physpath, 1630 phypath_size, rsz)) != 0) 1631 return (ret); 1632 } 1633 } else if (strcmp(type, VDEV_TYPE_MIRROR) == 0 || 1634 strcmp(type, VDEV_TYPE_REPLACING) == 0 || 1635 (is_spare = (strcmp(type, VDEV_TYPE_SPARE) == 0))) { 1636 nvlist_t **child; 1637 uint_t count; 1638 int i, ret; 1639 1640 if (nvlist_lookup_nvlist_array(nv, 1641 ZPOOL_CONFIG_CHILDREN, &child, &count) != 0) 1642 return (EZFS_INVALCONFIG); 1643 1644 for (i = 0; i < count; i++) { 1645 ret = vdev_get_physpaths(child[i], physpath, 1646 phypath_size, rsz, is_spare); 1647 if (ret == EZFS_NOSPC) 1648 return (ret); 1649 } 1650 } 1651 1652 return (EZFS_POOL_INVALARG); 1653 } 1654 1655 /* 1656 * Get phys_path for a root pool config. 1657 * Return 0 on success; non-zero on failure. 1658 */ 1659 static int 1660 zpool_get_config_physpath(nvlist_t *config, char *physpath, size_t phypath_size) 1661 { 1662 size_t rsz; 1663 nvlist_t *vdev_root; 1664 nvlist_t **child; 1665 uint_t count; 1666 char *type; 1667 1668 rsz = 0; 1669 1670 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 1671 &vdev_root) != 0) 1672 return (EZFS_INVALCONFIG); 1673 1674 if (nvlist_lookup_string(vdev_root, ZPOOL_CONFIG_TYPE, &type) != 0 || 1675 nvlist_lookup_nvlist_array(vdev_root, ZPOOL_CONFIG_CHILDREN, 1676 &child, &count) != 0) 1677 return (EZFS_INVALCONFIG); 1678 1679 /* 1680 * root pool can not have EFI labeled disks and can only have 1681 * a single top-level vdev. 1682 */ 1683 if (strcmp(type, VDEV_TYPE_ROOT) != 0 || count != 1 || 1684 pool_uses_efi(vdev_root)) 1685 return (EZFS_POOL_INVALARG); 1686 1687 (void) vdev_get_physpaths(child[0], physpath, phypath_size, &rsz, 1688 B_FALSE); 1689 1690 /* No online devices */ 1691 if (rsz == 0) 1692 return (EZFS_NODEVICE); 1693 1694 return (0); 1695 } 1696 1697 /* 1698 * Get phys_path for a root pool 1699 * Return 0 on success; non-zero on failure. 1700 */ 1701 int 1702 zpool_get_physpath(zpool_handle_t *zhp, char *physpath, size_t phypath_size) 1703 { 1704 return (zpool_get_config_physpath(zhp->zpool_config, physpath, 1705 phypath_size)); 1706 } 1707 1708 /* 1709 * Returns TRUE if the given guid corresponds to the given type. 1710 * This is used to check for hot spares (INUSE or not), and level 2 cache 1711 * devices. 1712 */ 1713 static boolean_t 1714 is_guid_type(zpool_handle_t *zhp, uint64_t guid, const char *type) 1715 { 1716 uint64_t target_guid; 1717 nvlist_t *nvroot; 1718 nvlist_t **list; 1719 uint_t count; 1720 int i; 1721 1722 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE, 1723 &nvroot) == 0); 1724 if (nvlist_lookup_nvlist_array(nvroot, type, &list, &count) == 0) { 1725 for (i = 0; i < count; i++) { 1726 verify(nvlist_lookup_uint64(list[i], ZPOOL_CONFIG_GUID, 1727 &target_guid) == 0); 1728 if (guid == target_guid) 1729 return (B_TRUE); 1730 } 1731 } 1732 1733 return (B_FALSE); 1734 } 1735 1736 /* 1737 * If the device has being dynamically expanded then we need to relabel 1738 * the disk to use the new unallocated space. 1739 */ 1740 static int 1741 zpool_relabel_disk(libzfs_handle_t *hdl, const char *name) 1742 { 1743 char path[MAXPATHLEN]; 1744 char errbuf[1024]; 1745 int fd, error; 1746 int (*_efi_use_whole_disk)(int); 1747 1748 if ((_efi_use_whole_disk = (int (*)(int))dlsym(RTLD_DEFAULT, 1749 "efi_use_whole_disk")) == NULL) 1750 return (-1); 1751 1752 (void) snprintf(path, sizeof (path), "%s/%s", RDISK_ROOT, name); 1753 1754 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) { 1755 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot " 1756 "relabel '%s': unable to open device"), name); 1757 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf)); 1758 } 1759 1760 /* 1761 * It's possible that we might encounter an error if the device 1762 * does not have any unallocated space left. If so, we simply 1763 * ignore that error and continue on. 1764 */ 1765 error = _efi_use_whole_disk(fd); 1766 (void) close(fd); 1767 if (error && error != VT_ENOSPC) { 1768 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot " 1769 "relabel '%s': unable to read disk capacity"), name); 1770 return (zfs_error(hdl, EZFS_NOCAP, errbuf)); 1771 } 1772 return (0); 1773 } 1774 1775 /* 1776 * Bring the specified vdev online. The 'flags' parameter is a set of the 1777 * ZFS_ONLINE_* flags. 1778 */ 1779 int 1780 zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags, 1781 vdev_state_t *newstate) 1782 { 1783 zfs_cmd_t zc = { 0 }; 1784 char msg[1024]; 1785 nvlist_t *tgt; 1786 boolean_t avail_spare, l2cache, islog; 1787 libzfs_handle_t *hdl = zhp->zpool_hdl; 1788 1789 if (flags & ZFS_ONLINE_EXPAND) { 1790 (void) snprintf(msg, sizeof (msg), 1791 dgettext(TEXT_DOMAIN, "cannot expand %s"), path); 1792 } else { 1793 (void) snprintf(msg, sizeof (msg), 1794 dgettext(TEXT_DOMAIN, "cannot online %s"), path); 1795 } 1796 1797 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1798 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 1799 &islog)) == NULL) 1800 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 1801 1802 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 1803 1804 if (avail_spare || 1805 is_guid_type(zhp, zc.zc_guid, ZPOOL_CONFIG_SPARES) == B_TRUE) 1806 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 1807 1808 if (flags & ZFS_ONLINE_EXPAND || 1809 zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) { 1810 char *pathname = NULL; 1811 uint64_t wholedisk = 0; 1812 1813 (void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK, 1814 &wholedisk); 1815 verify(nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH, 1816 &pathname) == 0); 1817 1818 /* 1819 * XXX - L2ARC 1.0 devices can't support expansion. 1820 */ 1821 if (l2cache) { 1822 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1823 "cannot expand cache devices")); 1824 return (zfs_error(hdl, EZFS_VDEVNOTSUP, msg)); 1825 } 1826 1827 if (wholedisk) { 1828 pathname += strlen(DISK_ROOT) + 1; 1829 (void) zpool_relabel_disk(zhp->zpool_hdl, pathname); 1830 } 1831 } 1832 1833 zc.zc_cookie = VDEV_STATE_ONLINE; 1834 zc.zc_obj = flags; 1835 1836 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0) 1837 return (zpool_standard_error(hdl, errno, msg)); 1838 1839 *newstate = zc.zc_cookie; 1840 return (0); 1841 } 1842 1843 /* 1844 * Take the specified vdev offline 1845 */ 1846 int 1847 zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp) 1848 { 1849 zfs_cmd_t zc = { 0 }; 1850 char msg[1024]; 1851 nvlist_t *tgt; 1852 boolean_t avail_spare, l2cache; 1853 libzfs_handle_t *hdl = zhp->zpool_hdl; 1854 1855 (void) snprintf(msg, sizeof (msg), 1856 dgettext(TEXT_DOMAIN, "cannot offline %s"), path); 1857 1858 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1859 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 1860 NULL)) == NULL) 1861 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 1862 1863 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 1864 1865 if (avail_spare || 1866 is_guid_type(zhp, zc.zc_guid, ZPOOL_CONFIG_SPARES) == B_TRUE) 1867 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 1868 1869 zc.zc_cookie = VDEV_STATE_OFFLINE; 1870 zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0; 1871 1872 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0) 1873 return (0); 1874 1875 switch (errno) { 1876 case EBUSY: 1877 1878 /* 1879 * There are no other replicas of this device. 1880 */ 1881 return (zfs_error(hdl, EZFS_NOREPLICAS, msg)); 1882 1883 case EEXIST: 1884 /* 1885 * The log device has unplayed logs 1886 */ 1887 return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, msg)); 1888 1889 default: 1890 return (zpool_standard_error(hdl, errno, msg)); 1891 } 1892 } 1893 1894 /* 1895 * Mark the given vdev faulted. 1896 */ 1897 int 1898 zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid) 1899 { 1900 zfs_cmd_t zc = { 0 }; 1901 char msg[1024]; 1902 libzfs_handle_t *hdl = zhp->zpool_hdl; 1903 1904 (void) snprintf(msg, sizeof (msg), 1905 dgettext(TEXT_DOMAIN, "cannot fault %llu"), guid); 1906 1907 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1908 zc.zc_guid = guid; 1909 zc.zc_cookie = VDEV_STATE_FAULTED; 1910 1911 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0) 1912 return (0); 1913 1914 switch (errno) { 1915 case EBUSY: 1916 1917 /* 1918 * There are no other replicas of this device. 1919 */ 1920 return (zfs_error(hdl, EZFS_NOREPLICAS, msg)); 1921 1922 default: 1923 return (zpool_standard_error(hdl, errno, msg)); 1924 } 1925 1926 } 1927 1928 /* 1929 * Mark the given vdev degraded. 1930 */ 1931 int 1932 zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid) 1933 { 1934 zfs_cmd_t zc = { 0 }; 1935 char msg[1024]; 1936 libzfs_handle_t *hdl = zhp->zpool_hdl; 1937 1938 (void) snprintf(msg, sizeof (msg), 1939 dgettext(TEXT_DOMAIN, "cannot degrade %llu"), guid); 1940 1941 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1942 zc.zc_guid = guid; 1943 zc.zc_cookie = VDEV_STATE_DEGRADED; 1944 1945 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0) 1946 return (0); 1947 1948 return (zpool_standard_error(hdl, errno, msg)); 1949 } 1950 1951 /* 1952 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as 1953 * a hot spare. 1954 */ 1955 static boolean_t 1956 is_replacing_spare(nvlist_t *search, nvlist_t *tgt, int which) 1957 { 1958 nvlist_t **child; 1959 uint_t c, children; 1960 char *type; 1961 1962 if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child, 1963 &children) == 0) { 1964 verify(nvlist_lookup_string(search, ZPOOL_CONFIG_TYPE, 1965 &type) == 0); 1966 1967 if (strcmp(type, VDEV_TYPE_SPARE) == 0 && 1968 children == 2 && child[which] == tgt) 1969 return (B_TRUE); 1970 1971 for (c = 0; c < children; c++) 1972 if (is_replacing_spare(child[c], tgt, which)) 1973 return (B_TRUE); 1974 } 1975 1976 return (B_FALSE); 1977 } 1978 1979 /* 1980 * Attach new_disk (fully described by nvroot) to old_disk. 1981 * If 'replacing' is specified, the new disk will replace the old one. 1982 */ 1983 int 1984 zpool_vdev_attach(zpool_handle_t *zhp, 1985 const char *old_disk, const char *new_disk, nvlist_t *nvroot, int replacing) 1986 { 1987 zfs_cmd_t zc = { 0 }; 1988 char msg[1024]; 1989 int ret; 1990 nvlist_t *tgt; 1991 boolean_t avail_spare, l2cache, islog; 1992 uint64_t val; 1993 char *path, *newname; 1994 nvlist_t **child; 1995 uint_t children; 1996 nvlist_t *config_root; 1997 libzfs_handle_t *hdl = zhp->zpool_hdl; 1998 boolean_t rootpool = pool_is_bootable(zhp); 1999 2000 if (replacing) 2001 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 2002 "cannot replace %s with %s"), old_disk, new_disk); 2003 else 2004 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 2005 "cannot attach %s to %s"), new_disk, old_disk); 2006 2007 /* 2008 * If this is a root pool, make sure that we're not attaching an 2009 * EFI labeled device. 2010 */ 2011 if (rootpool && pool_uses_efi(nvroot)) { 2012 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2013 "EFI labeled devices are not supported on root pools.")); 2014 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg)); 2015 } 2016 2017 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2018 if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache, 2019 &islog)) == 0) 2020 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 2021 2022 if (avail_spare) 2023 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 2024 2025 if (l2cache) 2026 return (zfs_error(hdl, EZFS_ISL2CACHE, msg)); 2027 2028 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 2029 zc.zc_cookie = replacing; 2030 2031 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, 2032 &child, &children) != 0 || children != 1) { 2033 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2034 "new device must be a single disk")); 2035 return (zfs_error(hdl, EZFS_INVALCONFIG, msg)); 2036 } 2037 2038 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL), 2039 ZPOOL_CONFIG_VDEV_TREE, &config_root) == 0); 2040 2041 if ((newname = zpool_vdev_name(NULL, NULL, child[0])) == NULL) 2042 return (-1); 2043 2044 /* 2045 * If the target is a hot spare that has been swapped in, we can only 2046 * replace it with another hot spare. 2047 */ 2048 if (replacing && 2049 nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 && 2050 (zpool_find_vdev(zhp, newname, &avail_spare, &l2cache, 2051 NULL) == NULL || !avail_spare) && 2052 is_replacing_spare(config_root, tgt, 1)) { 2053 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2054 "can only be replaced by another hot spare")); 2055 free(newname); 2056 return (zfs_error(hdl, EZFS_BADTARGET, msg)); 2057 } 2058 2059 /* 2060 * If we are attempting to replace a spare, it canot be applied to an 2061 * already spared device. 2062 */ 2063 if (replacing && 2064 nvlist_lookup_string(child[0], ZPOOL_CONFIG_PATH, &path) == 0 && 2065 zpool_find_vdev(zhp, newname, &avail_spare, 2066 &l2cache, NULL) != NULL && avail_spare && 2067 is_replacing_spare(config_root, tgt, 0)) { 2068 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2069 "device has already been replaced with a spare")); 2070 free(newname); 2071 return (zfs_error(hdl, EZFS_BADTARGET, msg)); 2072 } 2073 2074 free(newname); 2075 2076 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0) 2077 return (-1); 2078 2079 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_VDEV_ATTACH, &zc); 2080 2081 zcmd_free_nvlists(&zc); 2082 2083 if (ret == 0) { 2084 if (rootpool) { 2085 /* 2086 * XXX - This should be removed once we can 2087 * automatically install the bootblocks on the 2088 * newly attached disk. 2089 */ 2090 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Please " 2091 "be sure to invoke %s to make '%s' bootable.\n"), 2092 BOOTCMD, new_disk); 2093 2094 /* 2095 * XXX need a better way to prevent user from 2096 * booting up a half-baked vdev. 2097 */ 2098 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Make " 2099 "sure to wait until resilver is done " 2100 "before rebooting.\n")); 2101 } 2102 return (0); 2103 } 2104 2105 switch (errno) { 2106 case ENOTSUP: 2107 /* 2108 * Can't attach to or replace this type of vdev. 2109 */ 2110 if (replacing) { 2111 if (islog) 2112 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2113 "cannot replace a log with a spare")); 2114 else 2115 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2116 "cannot replace a replacing device")); 2117 } else { 2118 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2119 "can only attach to mirrors and top-level " 2120 "disks")); 2121 } 2122 (void) zfs_error(hdl, EZFS_BADTARGET, msg); 2123 break; 2124 2125 case EINVAL: 2126 /* 2127 * The new device must be a single disk. 2128 */ 2129 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2130 "new device must be a single disk")); 2131 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg); 2132 break; 2133 2134 case EBUSY: 2135 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy"), 2136 new_disk); 2137 (void) zfs_error(hdl, EZFS_BADDEV, msg); 2138 break; 2139 2140 case EOVERFLOW: 2141 /* 2142 * The new device is too small. 2143 */ 2144 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2145 "device is too small")); 2146 (void) zfs_error(hdl, EZFS_BADDEV, msg); 2147 break; 2148 2149 case EDOM: 2150 /* 2151 * The new device has a different alignment requirement. 2152 */ 2153 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2154 "devices have different sector alignment")); 2155 (void) zfs_error(hdl, EZFS_BADDEV, msg); 2156 break; 2157 2158 case ENAMETOOLONG: 2159 /* 2160 * The resulting top-level vdev spec won't fit in the label. 2161 */ 2162 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, msg); 2163 break; 2164 2165 default: 2166 (void) zpool_standard_error(hdl, errno, msg); 2167 } 2168 2169 return (-1); 2170 } 2171 2172 /* 2173 * Detach the specified device. 2174 */ 2175 int 2176 zpool_vdev_detach(zpool_handle_t *zhp, const char *path) 2177 { 2178 zfs_cmd_t zc = { 0 }; 2179 char msg[1024]; 2180 nvlist_t *tgt; 2181 boolean_t avail_spare, l2cache; 2182 libzfs_handle_t *hdl = zhp->zpool_hdl; 2183 2184 (void) snprintf(msg, sizeof (msg), 2185 dgettext(TEXT_DOMAIN, "cannot detach %s"), path); 2186 2187 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2188 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 2189 NULL)) == 0) 2190 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 2191 2192 if (avail_spare) 2193 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 2194 2195 if (l2cache) 2196 return (zfs_error(hdl, EZFS_ISL2CACHE, msg)); 2197 2198 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 2199 2200 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0) 2201 return (0); 2202 2203 switch (errno) { 2204 2205 case ENOTSUP: 2206 /* 2207 * Can't detach from this type of vdev. 2208 */ 2209 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only " 2210 "applicable to mirror and replacing vdevs")); 2211 (void) zfs_error(zhp->zpool_hdl, EZFS_BADTARGET, msg); 2212 break; 2213 2214 case EBUSY: 2215 /* 2216 * There are no other replicas of this device. 2217 */ 2218 (void) zfs_error(hdl, EZFS_NOREPLICAS, msg); 2219 break; 2220 2221 default: 2222 (void) zpool_standard_error(hdl, errno, msg); 2223 } 2224 2225 return (-1); 2226 } 2227 2228 /* 2229 * Remove the given device. Currently, this is supported only for hot spares 2230 * and level 2 cache devices. 2231 */ 2232 int 2233 zpool_vdev_remove(zpool_handle_t *zhp, const char *path) 2234 { 2235 zfs_cmd_t zc = { 0 }; 2236 char msg[1024]; 2237 nvlist_t *tgt; 2238 boolean_t avail_spare, l2cache; 2239 libzfs_handle_t *hdl = zhp->zpool_hdl; 2240 2241 (void) snprintf(msg, sizeof (msg), 2242 dgettext(TEXT_DOMAIN, "cannot remove %s"), path); 2243 2244 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2245 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 2246 NULL)) == 0) 2247 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 2248 2249 if (!avail_spare && !l2cache) { 2250 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2251 "only inactive hot spares or cache devices " 2252 "can be removed")); 2253 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 2254 } 2255 2256 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 2257 2258 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0) 2259 return (0); 2260 2261 return (zpool_standard_error(hdl, errno, msg)); 2262 } 2263 2264 /* 2265 * Clear the errors for the pool, or the particular device if specified. 2266 */ 2267 int 2268 zpool_clear(zpool_handle_t *zhp, const char *path) 2269 { 2270 zfs_cmd_t zc = { 0 }; 2271 char msg[1024]; 2272 nvlist_t *tgt; 2273 boolean_t avail_spare, l2cache; 2274 libzfs_handle_t *hdl = zhp->zpool_hdl; 2275 2276 if (path) 2277 (void) snprintf(msg, sizeof (msg), 2278 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"), 2279 path); 2280 else 2281 (void) snprintf(msg, sizeof (msg), 2282 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"), 2283 zhp->zpool_name); 2284 2285 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2286 if (path) { 2287 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, 2288 &l2cache, NULL)) == 0) 2289 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 2290 2291 /* 2292 * Don't allow error clearing for hot spares. Do allow 2293 * error clearing for l2cache devices. 2294 */ 2295 if (avail_spare) 2296 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 2297 2298 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, 2299 &zc.zc_guid) == 0); 2300 } 2301 2302 if (zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc) == 0) 2303 return (0); 2304 2305 return (zpool_standard_error(hdl, errno, msg)); 2306 } 2307 2308 /* 2309 * Similar to zpool_clear(), but takes a GUID (used by fmd). 2310 */ 2311 int 2312 zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid) 2313 { 2314 zfs_cmd_t zc = { 0 }; 2315 char msg[1024]; 2316 libzfs_handle_t *hdl = zhp->zpool_hdl; 2317 2318 (void) snprintf(msg, sizeof (msg), 2319 dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"), 2320 guid); 2321 2322 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2323 zc.zc_guid = guid; 2324 2325 if (ioctl(hdl->libzfs_fd, ZFS_IOC_CLEAR, &zc) == 0) 2326 return (0); 2327 2328 return (zpool_standard_error(hdl, errno, msg)); 2329 } 2330 2331 /* 2332 * Convert from a devid string to a path. 2333 */ 2334 static char * 2335 devid_to_path(char *devid_str) 2336 { 2337 ddi_devid_t devid; 2338 char *minor; 2339 char *path; 2340 devid_nmlist_t *list = NULL; 2341 int ret; 2342 2343 if (devid_str_decode(devid_str, &devid, &minor) != 0) 2344 return (NULL); 2345 2346 ret = devid_deviceid_to_nmlist("/dev", devid, minor, &list); 2347 2348 devid_str_free(minor); 2349 devid_free(devid); 2350 2351 if (ret != 0) 2352 return (NULL); 2353 2354 if ((path = strdup(list[0].devname)) == NULL) 2355 return (NULL); 2356 2357 devid_free_nmlist(list); 2358 2359 return (path); 2360 } 2361 2362 /* 2363 * Convert from a path to a devid string. 2364 */ 2365 static char * 2366 path_to_devid(const char *path) 2367 { 2368 int fd; 2369 ddi_devid_t devid; 2370 char *minor, *ret; 2371 2372 if ((fd = open(path, O_RDONLY)) < 0) 2373 return (NULL); 2374 2375 minor = NULL; 2376 ret = NULL; 2377 if (devid_get(fd, &devid) == 0) { 2378 if (devid_get_minor_name(fd, &minor) == 0) 2379 ret = devid_str_encode(devid, minor); 2380 if (minor != NULL) 2381 devid_str_free(minor); 2382 devid_free(devid); 2383 } 2384 (void) close(fd); 2385 2386 return (ret); 2387 } 2388 2389 /* 2390 * Issue the necessary ioctl() to update the stored path value for the vdev. We 2391 * ignore any failure here, since a common case is for an unprivileged user to 2392 * type 'zpool status', and we'll display the correct information anyway. 2393 */ 2394 static void 2395 set_path(zpool_handle_t *zhp, nvlist_t *nv, const char *path) 2396 { 2397 zfs_cmd_t zc = { 0 }; 2398 2399 (void) strncpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2400 (void) strncpy(zc.zc_value, path, sizeof (zc.zc_value)); 2401 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, 2402 &zc.zc_guid) == 0); 2403 2404 (void) ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SETPATH, &zc); 2405 } 2406 2407 /* 2408 * Given a vdev, return the name to display in iostat. If the vdev has a path, 2409 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type. 2410 * We also check if this is a whole disk, in which case we strip off the 2411 * trailing 's0' slice name. 2412 * 2413 * This routine is also responsible for identifying when disks have been 2414 * reconfigured in a new location. The kernel will have opened the device by 2415 * devid, but the path will still refer to the old location. To catch this, we 2416 * first do a path -> devid translation (which is fast for the common case). If 2417 * the devid matches, we're done. If not, we do a reverse devid -> path 2418 * translation and issue the appropriate ioctl() to update the path of the vdev. 2419 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any 2420 * of these checks. 2421 */ 2422 char * 2423 zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv) 2424 { 2425 char *path, *devid; 2426 uint64_t value; 2427 char buf[64]; 2428 vdev_stat_t *vs; 2429 uint_t vsc; 2430 2431 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, 2432 &value) == 0) { 2433 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, 2434 &value) == 0); 2435 (void) snprintf(buf, sizeof (buf), "%llu", 2436 (u_longlong_t)value); 2437 path = buf; 2438 } else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) { 2439 2440 /* 2441 * If the device is dead (faulted, offline, etc) then don't 2442 * bother opening it. Otherwise we may be forcing the user to 2443 * open a misbehaving device, which can have undesirable 2444 * effects. 2445 */ 2446 if ((nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_STATS, 2447 (uint64_t **)&vs, &vsc) != 0 || 2448 vs->vs_state >= VDEV_STATE_DEGRADED) && 2449 zhp != NULL && 2450 nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &devid) == 0) { 2451 /* 2452 * Determine if the current path is correct. 2453 */ 2454 char *newdevid = path_to_devid(path); 2455 2456 if (newdevid == NULL || 2457 strcmp(devid, newdevid) != 0) { 2458 char *newpath; 2459 2460 if ((newpath = devid_to_path(devid)) != NULL) { 2461 /* 2462 * Update the path appropriately. 2463 */ 2464 set_path(zhp, nv, newpath); 2465 if (nvlist_add_string(nv, 2466 ZPOOL_CONFIG_PATH, newpath) == 0) 2467 verify(nvlist_lookup_string(nv, 2468 ZPOOL_CONFIG_PATH, 2469 &path) == 0); 2470 free(newpath); 2471 } 2472 } 2473 2474 if (newdevid) 2475 devid_str_free(newdevid); 2476 } 2477 2478 if (strncmp(path, "/dev/dsk/", 9) == 0) 2479 path += 9; 2480 2481 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, 2482 &value) == 0 && value) { 2483 char *tmp = zfs_strdup(hdl, path); 2484 if (tmp == NULL) 2485 return (NULL); 2486 tmp[strlen(path) - 2] = '\0'; 2487 return (tmp); 2488 } 2489 } else { 2490 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &path) == 0); 2491 2492 /* 2493 * If it's a raidz device, we need to stick in the parity level. 2494 */ 2495 if (strcmp(path, VDEV_TYPE_RAIDZ) == 0) { 2496 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY, 2497 &value) == 0); 2498 (void) snprintf(buf, sizeof (buf), "%s%llu", path, 2499 (u_longlong_t)value); 2500 path = buf; 2501 } 2502 } 2503 2504 return (zfs_strdup(hdl, path)); 2505 } 2506 2507 static int 2508 zbookmark_compare(const void *a, const void *b) 2509 { 2510 return (memcmp(a, b, sizeof (zbookmark_t))); 2511 } 2512 2513 /* 2514 * Retrieve the persistent error log, uniquify the members, and return to the 2515 * caller. 2516 */ 2517 int 2518 zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp) 2519 { 2520 zfs_cmd_t zc = { 0 }; 2521 uint64_t count; 2522 zbookmark_t *zb = NULL; 2523 int i; 2524 2525 /* 2526 * Retrieve the raw error list from the kernel. If the number of errors 2527 * has increased, allocate more space and continue until we get the 2528 * entire list. 2529 */ 2530 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_ERRCOUNT, 2531 &count) == 0); 2532 if (count == 0) 2533 return (0); 2534 if ((zc.zc_nvlist_dst = (uintptr_t)zfs_alloc(zhp->zpool_hdl, 2535 count * sizeof (zbookmark_t))) == (uintptr_t)NULL) 2536 return (-1); 2537 zc.zc_nvlist_dst_size = count; 2538 (void) strcpy(zc.zc_name, zhp->zpool_name); 2539 for (;;) { 2540 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_ERROR_LOG, 2541 &zc) != 0) { 2542 free((void *)(uintptr_t)zc.zc_nvlist_dst); 2543 if (errno == ENOMEM) { 2544 count = zc.zc_nvlist_dst_size; 2545 if ((zc.zc_nvlist_dst = (uintptr_t) 2546 zfs_alloc(zhp->zpool_hdl, count * 2547 sizeof (zbookmark_t))) == (uintptr_t)NULL) 2548 return (-1); 2549 } else { 2550 return (-1); 2551 } 2552 } else { 2553 break; 2554 } 2555 } 2556 2557 /* 2558 * Sort the resulting bookmarks. This is a little confusing due to the 2559 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last 2560 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks 2561 * _not_ copied as part of the process. So we point the start of our 2562 * array appropriate and decrement the total number of elements. 2563 */ 2564 zb = ((zbookmark_t *)(uintptr_t)zc.zc_nvlist_dst) + 2565 zc.zc_nvlist_dst_size; 2566 count -= zc.zc_nvlist_dst_size; 2567 2568 qsort(zb, count, sizeof (zbookmark_t), zbookmark_compare); 2569 2570 verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0); 2571 2572 /* 2573 * Fill in the nverrlistp with nvlist's of dataset and object numbers. 2574 */ 2575 for (i = 0; i < count; i++) { 2576 nvlist_t *nv; 2577 2578 /* ignoring zb_blkid and zb_level for now */ 2579 if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset && 2580 zb[i-1].zb_object == zb[i].zb_object) 2581 continue; 2582 2583 if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0) 2584 goto nomem; 2585 if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET, 2586 zb[i].zb_objset) != 0) { 2587 nvlist_free(nv); 2588 goto nomem; 2589 } 2590 if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT, 2591 zb[i].zb_object) != 0) { 2592 nvlist_free(nv); 2593 goto nomem; 2594 } 2595 if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) { 2596 nvlist_free(nv); 2597 goto nomem; 2598 } 2599 nvlist_free(nv); 2600 } 2601 2602 free((void *)(uintptr_t)zc.zc_nvlist_dst); 2603 return (0); 2604 2605 nomem: 2606 free((void *)(uintptr_t)zc.zc_nvlist_dst); 2607 return (no_memory(zhp->zpool_hdl)); 2608 } 2609 2610 /* 2611 * Upgrade a ZFS pool to the latest on-disk version. 2612 */ 2613 int 2614 zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version) 2615 { 2616 zfs_cmd_t zc = { 0 }; 2617 libzfs_handle_t *hdl = zhp->zpool_hdl; 2618 2619 (void) strcpy(zc.zc_name, zhp->zpool_name); 2620 zc.zc_cookie = new_version; 2621 2622 if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0) 2623 return (zpool_standard_error_fmt(hdl, errno, 2624 dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"), 2625 zhp->zpool_name)); 2626 return (0); 2627 } 2628 2629 void 2630 zpool_set_history_str(const char *subcommand, int argc, char **argv, 2631 char *history_str) 2632 { 2633 int i; 2634 2635 (void) strlcpy(history_str, subcommand, HIS_MAX_RECORD_LEN); 2636 for (i = 1; i < argc; i++) { 2637 if (strlen(history_str) + 1 + strlen(argv[i]) > 2638 HIS_MAX_RECORD_LEN) 2639 break; 2640 (void) strlcat(history_str, " ", HIS_MAX_RECORD_LEN); 2641 (void) strlcat(history_str, argv[i], HIS_MAX_RECORD_LEN); 2642 } 2643 } 2644 2645 /* 2646 * Stage command history for logging. 2647 */ 2648 int 2649 zpool_stage_history(libzfs_handle_t *hdl, const char *history_str) 2650 { 2651 if (history_str == NULL) 2652 return (EINVAL); 2653 2654 if (strlen(history_str) > HIS_MAX_RECORD_LEN) 2655 return (EINVAL); 2656 2657 if (hdl->libzfs_log_str != NULL) 2658 free(hdl->libzfs_log_str); 2659 2660 if ((hdl->libzfs_log_str = strdup(history_str)) == NULL) 2661 return (no_memory(hdl)); 2662 2663 return (0); 2664 } 2665 2666 /* 2667 * Perform ioctl to get some command history of a pool. 2668 * 2669 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the 2670 * logical offset of the history buffer to start reading from. 2671 * 2672 * Upon return, 'off' is the next logical offset to read from and 2673 * 'len' is the actual amount of bytes read into 'buf'. 2674 */ 2675 static int 2676 get_history(zpool_handle_t *zhp, char *buf, uint64_t *off, uint64_t *len) 2677 { 2678 zfs_cmd_t zc = { 0 }; 2679 libzfs_handle_t *hdl = zhp->zpool_hdl; 2680 2681 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2682 2683 zc.zc_history = (uint64_t)(uintptr_t)buf; 2684 zc.zc_history_len = *len; 2685 zc.zc_history_offset = *off; 2686 2687 if (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) { 2688 switch (errno) { 2689 case EPERM: 2690 return (zfs_error_fmt(hdl, EZFS_PERM, 2691 dgettext(TEXT_DOMAIN, 2692 "cannot show history for pool '%s'"), 2693 zhp->zpool_name)); 2694 case ENOENT: 2695 return (zfs_error_fmt(hdl, EZFS_NOHISTORY, 2696 dgettext(TEXT_DOMAIN, "cannot get history for pool " 2697 "'%s'"), zhp->zpool_name)); 2698 case ENOTSUP: 2699 return (zfs_error_fmt(hdl, EZFS_BADVERSION, 2700 dgettext(TEXT_DOMAIN, "cannot get history for pool " 2701 "'%s', pool must be upgraded"), zhp->zpool_name)); 2702 default: 2703 return (zpool_standard_error_fmt(hdl, errno, 2704 dgettext(TEXT_DOMAIN, 2705 "cannot get history for '%s'"), zhp->zpool_name)); 2706 } 2707 } 2708 2709 *len = zc.zc_history_len; 2710 *off = zc.zc_history_offset; 2711 2712 return (0); 2713 } 2714 2715 /* 2716 * Process the buffer of nvlists, unpacking and storing each nvlist record 2717 * into 'records'. 'leftover' is set to the number of bytes that weren't 2718 * processed as there wasn't a complete record. 2719 */ 2720 static int 2721 zpool_history_unpack(char *buf, uint64_t bytes_read, uint64_t *leftover, 2722 nvlist_t ***records, uint_t *numrecords) 2723 { 2724 uint64_t reclen; 2725 nvlist_t *nv; 2726 int i; 2727 2728 while (bytes_read > sizeof (reclen)) { 2729 2730 /* get length of packed record (stored as little endian) */ 2731 for (i = 0, reclen = 0; i < sizeof (reclen); i++) 2732 reclen += (uint64_t)(((uchar_t *)buf)[i]) << (8*i); 2733 2734 if (bytes_read < sizeof (reclen) + reclen) 2735 break; 2736 2737 /* unpack record */ 2738 if (nvlist_unpack(buf + sizeof (reclen), reclen, &nv, 0) != 0) 2739 return (ENOMEM); 2740 bytes_read -= sizeof (reclen) + reclen; 2741 buf += sizeof (reclen) + reclen; 2742 2743 /* add record to nvlist array */ 2744 (*numrecords)++; 2745 if (ISP2(*numrecords + 1)) { 2746 *records = realloc(*records, 2747 *numrecords * 2 * sizeof (nvlist_t *)); 2748 } 2749 (*records)[*numrecords - 1] = nv; 2750 } 2751 2752 *leftover = bytes_read; 2753 return (0); 2754 } 2755 2756 #define HIS_BUF_LEN (128*1024) 2757 2758 /* 2759 * Retrieve the command history of a pool. 2760 */ 2761 int 2762 zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp) 2763 { 2764 char buf[HIS_BUF_LEN]; 2765 uint64_t off = 0; 2766 nvlist_t **records = NULL; 2767 uint_t numrecords = 0; 2768 int err, i; 2769 2770 do { 2771 uint64_t bytes_read = sizeof (buf); 2772 uint64_t leftover; 2773 2774 if ((err = get_history(zhp, buf, &off, &bytes_read)) != 0) 2775 break; 2776 2777 /* if nothing else was read in, we're at EOF, just return */ 2778 if (!bytes_read) 2779 break; 2780 2781 if ((err = zpool_history_unpack(buf, bytes_read, 2782 &leftover, &records, &numrecords)) != 0) 2783 break; 2784 off -= leftover; 2785 2786 /* CONSTCOND */ 2787 } while (1); 2788 2789 if (!err) { 2790 verify(nvlist_alloc(nvhisp, NV_UNIQUE_NAME, 0) == 0); 2791 verify(nvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD, 2792 records, numrecords) == 0); 2793 } 2794 for (i = 0; i < numrecords; i++) 2795 nvlist_free(records[i]); 2796 free(records); 2797 2798 return (err); 2799 } 2800 2801 void 2802 zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj, 2803 char *pathname, size_t len) 2804 { 2805 zfs_cmd_t zc = { 0 }; 2806 boolean_t mounted = B_FALSE; 2807 char *mntpnt = NULL; 2808 char dsname[MAXNAMELEN]; 2809 2810 if (dsobj == 0) { 2811 /* special case for the MOS */ 2812 (void) snprintf(pathname, len, "<metadata>:<0x%llx>", obj); 2813 return; 2814 } 2815 2816 /* get the dataset's name */ 2817 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2818 zc.zc_obj = dsobj; 2819 if (ioctl(zhp->zpool_hdl->libzfs_fd, 2820 ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) { 2821 /* just write out a path of two object numbers */ 2822 (void) snprintf(pathname, len, "<0x%llx>:<0x%llx>", 2823 dsobj, obj); 2824 return; 2825 } 2826 (void) strlcpy(dsname, zc.zc_value, sizeof (dsname)); 2827 2828 /* find out if the dataset is mounted */ 2829 mounted = is_mounted(zhp->zpool_hdl, dsname, &mntpnt); 2830 2831 /* get the corrupted object's path */ 2832 (void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name)); 2833 zc.zc_obj = obj; 2834 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_OBJ_TO_PATH, 2835 &zc) == 0) { 2836 if (mounted) { 2837 (void) snprintf(pathname, len, "%s%s", mntpnt, 2838 zc.zc_value); 2839 } else { 2840 (void) snprintf(pathname, len, "%s:%s", 2841 dsname, zc.zc_value); 2842 } 2843 } else { 2844 (void) snprintf(pathname, len, "%s:<0x%llx>", dsname, obj); 2845 } 2846 free(mntpnt); 2847 } 2848 2849 /* 2850 * Read the EFI label from the config, if a label does not exist then 2851 * pass back the error to the caller. If the caller has passed a non-NULL 2852 * diskaddr argument then we set it to the starting address of the EFI 2853 * partition. 2854 */ 2855 static int 2856 read_efi_label(nvlist_t *config, diskaddr_t *sb) 2857 { 2858 char *path; 2859 int fd; 2860 char diskname[MAXPATHLEN]; 2861 int err = -1; 2862 2863 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PATH, &path) != 0) 2864 return (err); 2865 2866 (void) snprintf(diskname, sizeof (diskname), "%s%s", RDISK_ROOT, 2867 strrchr(path, '/')); 2868 if ((fd = open(diskname, O_RDONLY|O_NDELAY)) >= 0) { 2869 struct dk_gpt *vtoc; 2870 2871 if ((err = efi_alloc_and_read(fd, &vtoc)) >= 0) { 2872 if (sb != NULL) 2873 *sb = vtoc->efi_parts[0].p_start; 2874 efi_free(vtoc); 2875 } 2876 (void) close(fd); 2877 } 2878 return (err); 2879 } 2880 2881 /* 2882 * determine where a partition starts on a disk in the current 2883 * configuration 2884 */ 2885 static diskaddr_t 2886 find_start_block(nvlist_t *config) 2887 { 2888 nvlist_t **child; 2889 uint_t c, children; 2890 diskaddr_t sb = MAXOFFSET_T; 2891 uint64_t wholedisk; 2892 2893 if (nvlist_lookup_nvlist_array(config, 2894 ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) { 2895 if (nvlist_lookup_uint64(config, 2896 ZPOOL_CONFIG_WHOLE_DISK, 2897 &wholedisk) != 0 || !wholedisk) { 2898 return (MAXOFFSET_T); 2899 } 2900 if (read_efi_label(config, &sb) < 0) 2901 sb = MAXOFFSET_T; 2902 return (sb); 2903 } 2904 2905 for (c = 0; c < children; c++) { 2906 sb = find_start_block(child[c]); 2907 if (sb != MAXOFFSET_T) { 2908 return (sb); 2909 } 2910 } 2911 return (MAXOFFSET_T); 2912 } 2913 2914 /* 2915 * Label an individual disk. The name provided is the short name, 2916 * stripped of any leading /dev path. 2917 */ 2918 int 2919 zpool_label_disk(libzfs_handle_t *hdl, zpool_handle_t *zhp, char *name) 2920 { 2921 char path[MAXPATHLEN]; 2922 struct dk_gpt *vtoc; 2923 int fd; 2924 size_t resv = EFI_MIN_RESV_SIZE; 2925 uint64_t slice_size; 2926 diskaddr_t start_block; 2927 char errbuf[1024]; 2928 2929 /* prepare an error message just in case */ 2930 (void) snprintf(errbuf, sizeof (errbuf), 2931 dgettext(TEXT_DOMAIN, "cannot label '%s'"), name); 2932 2933 if (zhp) { 2934 nvlist_t *nvroot; 2935 2936 if (pool_is_bootable(zhp)) { 2937 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2938 "EFI labeled devices are not supported on root " 2939 "pools.")); 2940 return (zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf)); 2941 } 2942 2943 verify(nvlist_lookup_nvlist(zhp->zpool_config, 2944 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 2945 2946 if (zhp->zpool_start_block == 0) 2947 start_block = find_start_block(nvroot); 2948 else 2949 start_block = zhp->zpool_start_block; 2950 zhp->zpool_start_block = start_block; 2951 } else { 2952 /* new pool */ 2953 start_block = NEW_START_BLOCK; 2954 } 2955 2956 (void) snprintf(path, sizeof (path), "%s/%s%s", RDISK_ROOT, name, 2957 BACKUP_SLICE); 2958 2959 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) { 2960 /* 2961 * This shouldn't happen. We've long since verified that this 2962 * is a valid device. 2963 */ 2964 zfs_error_aux(hdl, 2965 dgettext(TEXT_DOMAIN, "unable to open device")); 2966 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf)); 2967 } 2968 2969 if (efi_alloc_and_init(fd, EFI_NUMPAR, &vtoc) != 0) { 2970 /* 2971 * The only way this can fail is if we run out of memory, or we 2972 * were unable to read the disk's capacity 2973 */ 2974 if (errno == ENOMEM) 2975 (void) no_memory(hdl); 2976 2977 (void) close(fd); 2978 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2979 "unable to read disk capacity"), name); 2980 2981 return (zfs_error(hdl, EZFS_NOCAP, errbuf)); 2982 } 2983 2984 slice_size = vtoc->efi_last_u_lba + 1; 2985 slice_size -= EFI_MIN_RESV_SIZE; 2986 if (start_block == MAXOFFSET_T) 2987 start_block = NEW_START_BLOCK; 2988 slice_size -= start_block; 2989 2990 vtoc->efi_parts[0].p_start = start_block; 2991 vtoc->efi_parts[0].p_size = slice_size; 2992 2993 /* 2994 * Why we use V_USR: V_BACKUP confuses users, and is considered 2995 * disposable by some EFI utilities (since EFI doesn't have a backup 2996 * slice). V_UNASSIGNED is supposed to be used only for zero size 2997 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT, 2998 * etc. were all pretty specific. V_USR is as close to reality as we 2999 * can get, in the absence of V_OTHER. 3000 */ 3001 vtoc->efi_parts[0].p_tag = V_USR; 3002 (void) strcpy(vtoc->efi_parts[0].p_name, "zfs"); 3003 3004 vtoc->efi_parts[8].p_start = slice_size + start_block; 3005 vtoc->efi_parts[8].p_size = resv; 3006 vtoc->efi_parts[8].p_tag = V_RESERVED; 3007 3008 if (efi_write(fd, vtoc) != 0) { 3009 /* 3010 * Some block drivers (like pcata) may not support EFI 3011 * GPT labels. Print out a helpful error message dir- 3012 * ecting the user to manually label the disk and give 3013 * a specific slice. 3014 */ 3015 (void) close(fd); 3016 efi_free(vtoc); 3017 3018 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3019 "try using fdisk(1M) and then provide a specific slice")); 3020 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf)); 3021 } 3022 3023 (void) close(fd); 3024 efi_free(vtoc); 3025 return (0); 3026 } 3027 3028 static boolean_t 3029 supported_dump_vdev_type(libzfs_handle_t *hdl, nvlist_t *config, char *errbuf) 3030 { 3031 char *type; 3032 nvlist_t **child; 3033 uint_t children, c; 3034 3035 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_TYPE, &type) == 0); 3036 if (strcmp(type, VDEV_TYPE_RAIDZ) == 0 || 3037 strcmp(type, VDEV_TYPE_FILE) == 0 || 3038 strcmp(type, VDEV_TYPE_LOG) == 0 || 3039 strcmp(type, VDEV_TYPE_MISSING) == 0) { 3040 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3041 "vdev type '%s' is not supported"), type); 3042 (void) zfs_error(hdl, EZFS_VDEVNOTSUP, errbuf); 3043 return (B_FALSE); 3044 } 3045 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN, 3046 &child, &children) == 0) { 3047 for (c = 0; c < children; c++) { 3048 if (!supported_dump_vdev_type(hdl, child[c], errbuf)) 3049 return (B_FALSE); 3050 } 3051 } 3052 return (B_TRUE); 3053 } 3054 3055 /* 3056 * check if this zvol is allowable for use as a dump device; zero if 3057 * it is, > 0 if it isn't, < 0 if it isn't a zvol 3058 */ 3059 int 3060 zvol_check_dump_config(char *arg) 3061 { 3062 zpool_handle_t *zhp = NULL; 3063 nvlist_t *config, *nvroot; 3064 char *p, *volname; 3065 nvlist_t **top; 3066 uint_t toplevels; 3067 libzfs_handle_t *hdl; 3068 char errbuf[1024]; 3069 char poolname[ZPOOL_MAXNAMELEN]; 3070 int pathlen = strlen(ZVOL_FULL_DEV_DIR); 3071 int ret = 1; 3072 3073 if (strncmp(arg, ZVOL_FULL_DEV_DIR, pathlen)) { 3074 return (-1); 3075 } 3076 3077 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, 3078 "dump is not supported on device '%s'"), arg); 3079 3080 if ((hdl = libzfs_init()) == NULL) 3081 return (1); 3082 libzfs_print_on_error(hdl, B_TRUE); 3083 3084 volname = arg + pathlen; 3085 3086 /* check the configuration of the pool */ 3087 if ((p = strchr(volname, '/')) == NULL) { 3088 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3089 "malformed dataset name")); 3090 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf); 3091 return (1); 3092 } else if (p - volname >= ZFS_MAXNAMELEN) { 3093 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3094 "dataset name is too long")); 3095 (void) zfs_error(hdl, EZFS_NAMETOOLONG, errbuf); 3096 return (1); 3097 } else { 3098 (void) strncpy(poolname, volname, p - volname); 3099 poolname[p - volname] = '\0'; 3100 } 3101 3102 if ((zhp = zpool_open(hdl, poolname)) == NULL) { 3103 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3104 "could not open pool '%s'"), poolname); 3105 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf); 3106 goto out; 3107 } 3108 config = zpool_get_config(zhp, NULL); 3109 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 3110 &nvroot) != 0) { 3111 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3112 "could not obtain vdev configuration for '%s'"), poolname); 3113 (void) zfs_error(hdl, EZFS_INVALCONFIG, errbuf); 3114 goto out; 3115 } 3116 3117 verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, 3118 &top, &toplevels) == 0); 3119 if (toplevels != 1) { 3120 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3121 "'%s' has multiple top level vdevs"), poolname); 3122 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, errbuf); 3123 goto out; 3124 } 3125 3126 if (!supported_dump_vdev_type(hdl, top[0], errbuf)) { 3127 goto out; 3128 } 3129 ret = 0; 3130 3131 out: 3132 if (zhp) 3133 zpool_close(zhp); 3134 libzfs_fini(hdl); 3135 return (ret); 3136 } 3137