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 (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 24 */ 25 26 #include <ctype.h> 27 #include <errno.h> 28 #include <devid.h> 29 #include <fcntl.h> 30 #include <libintl.h> 31 #include <stdio.h> 32 #include <stdlib.h> 33 #include <strings.h> 34 #include <unistd.h> 35 #include <sys/efi_partition.h> 36 #include <sys/vtoc.h> 37 #include <sys/zfs_ioctl.h> 38 #include <dlfcn.h> 39 40 #include "zfs_namecheck.h" 41 #include "zfs_prop.h" 42 #include "libzfs_impl.h" 43 #include "zfs_comutil.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 if (aux == VDEV_AUX_SPLIT_POOL) 191 return (gettext("SPLIT")); 192 else 193 return (gettext("UNAVAIL")); 194 case VDEV_STATE_FAULTED: 195 return (gettext("FAULTED")); 196 case VDEV_STATE_DEGRADED: 197 return (gettext("DEGRADED")); 198 case VDEV_STATE_HEALTHY: 199 return (gettext("ONLINE")); 200 } 201 202 return (gettext("UNKNOWN")); 203 } 204 205 /* 206 * Get a zpool property value for 'prop' and return the value in 207 * a pre-allocated buffer. 208 */ 209 int 210 zpool_get_prop(zpool_handle_t *zhp, zpool_prop_t prop, char *buf, size_t len, 211 zprop_source_t *srctype) 212 { 213 uint64_t intval; 214 const char *strval; 215 zprop_source_t src = ZPROP_SRC_NONE; 216 nvlist_t *nvroot; 217 vdev_stat_t *vs; 218 uint_t vsc; 219 220 if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) { 221 switch (prop) { 222 case ZPOOL_PROP_NAME: 223 (void) strlcpy(buf, zpool_get_name(zhp), len); 224 break; 225 226 case ZPOOL_PROP_HEALTH: 227 (void) strlcpy(buf, "FAULTED", len); 228 break; 229 230 case ZPOOL_PROP_GUID: 231 intval = zpool_get_prop_int(zhp, prop, &src); 232 (void) snprintf(buf, len, "%llu", intval); 233 break; 234 235 case ZPOOL_PROP_ALTROOT: 236 case ZPOOL_PROP_CACHEFILE: 237 if (zhp->zpool_props != NULL || 238 zpool_get_all_props(zhp) == 0) { 239 (void) strlcpy(buf, 240 zpool_get_prop_string(zhp, prop, &src), 241 len); 242 if (srctype != NULL) 243 *srctype = src; 244 return (0); 245 } 246 /* FALLTHROUGH */ 247 default: 248 (void) strlcpy(buf, "-", len); 249 break; 250 } 251 252 if (srctype != NULL) 253 *srctype = src; 254 return (0); 255 } 256 257 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp) && 258 prop != ZPOOL_PROP_NAME) 259 return (-1); 260 261 switch (zpool_prop_get_type(prop)) { 262 case PROP_TYPE_STRING: 263 (void) strlcpy(buf, zpool_get_prop_string(zhp, prop, &src), 264 len); 265 break; 266 267 case PROP_TYPE_NUMBER: 268 intval = zpool_get_prop_int(zhp, prop, &src); 269 270 switch (prop) { 271 case ZPOOL_PROP_SIZE: 272 case ZPOOL_PROP_ALLOCATED: 273 case ZPOOL_PROP_FREE: 274 (void) zfs_nicenum(intval, buf, len); 275 break; 276 277 case ZPOOL_PROP_CAPACITY: 278 (void) snprintf(buf, len, "%llu%%", 279 (u_longlong_t)intval); 280 break; 281 282 case ZPOOL_PROP_DEDUPRATIO: 283 (void) snprintf(buf, len, "%llu.%02llux", 284 (u_longlong_t)(intval / 100), 285 (u_longlong_t)(intval % 100)); 286 break; 287 288 case ZPOOL_PROP_HEALTH: 289 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL), 290 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 291 verify(nvlist_lookup_uint64_array(nvroot, 292 ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc) 293 == 0); 294 295 (void) strlcpy(buf, zpool_state_to_name(intval, 296 vs->vs_aux), len); 297 break; 298 default: 299 (void) snprintf(buf, len, "%llu", intval); 300 } 301 break; 302 303 case PROP_TYPE_INDEX: 304 intval = zpool_get_prop_int(zhp, prop, &src); 305 if (zpool_prop_index_to_string(prop, intval, &strval) 306 != 0) 307 return (-1); 308 (void) strlcpy(buf, strval, len); 309 break; 310 311 default: 312 abort(); 313 } 314 315 if (srctype) 316 *srctype = src; 317 318 return (0); 319 } 320 321 /* 322 * Check if the bootfs name has the same pool name as it is set to. 323 * Assuming bootfs is a valid dataset name. 324 */ 325 static boolean_t 326 bootfs_name_valid(const char *pool, char *bootfs) 327 { 328 int len = strlen(pool); 329 330 if (!zfs_name_valid(bootfs, ZFS_TYPE_FILESYSTEM|ZFS_TYPE_SNAPSHOT)) 331 return (B_FALSE); 332 333 if (strncmp(pool, bootfs, len) == 0 && 334 (bootfs[len] == '/' || bootfs[len] == '\0')) 335 return (B_TRUE); 336 337 return (B_FALSE); 338 } 339 340 /* 341 * Inspect the configuration to determine if any of the devices contain 342 * an EFI label. 343 */ 344 static boolean_t 345 pool_uses_efi(nvlist_t *config) 346 { 347 nvlist_t **child; 348 uint_t c, children; 349 350 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN, 351 &child, &children) != 0) 352 return (read_efi_label(config, NULL) >= 0); 353 354 for (c = 0; c < children; c++) { 355 if (pool_uses_efi(child[c])) 356 return (B_TRUE); 357 } 358 return (B_FALSE); 359 } 360 361 static boolean_t 362 pool_is_bootable(zpool_handle_t *zhp) 363 { 364 char bootfs[ZPOOL_MAXNAMELEN]; 365 366 return (zpool_get_prop(zhp, ZPOOL_PROP_BOOTFS, bootfs, 367 sizeof (bootfs), NULL) == 0 && strncmp(bootfs, "-", 368 sizeof (bootfs)) != 0); 369 } 370 371 372 /* 373 * Given an nvlist of zpool properties to be set, validate that they are 374 * correct, and parse any numeric properties (index, boolean, etc) if they are 375 * specified as strings. 376 */ 377 static nvlist_t * 378 zpool_valid_proplist(libzfs_handle_t *hdl, const char *poolname, 379 nvlist_t *props, uint64_t version, boolean_t create_or_import, char *errbuf) 380 { 381 nvpair_t *elem; 382 nvlist_t *retprops; 383 zpool_prop_t prop; 384 char *strval; 385 uint64_t intval; 386 char *slash; 387 struct stat64 statbuf; 388 zpool_handle_t *zhp; 389 nvlist_t *nvroot; 390 391 if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) { 392 (void) no_memory(hdl); 393 return (NULL); 394 } 395 396 elem = NULL; 397 while ((elem = nvlist_next_nvpair(props, elem)) != NULL) { 398 const char *propname = nvpair_name(elem); 399 400 /* 401 * Make sure this property is valid and applies to this type. 402 */ 403 if ((prop = zpool_name_to_prop(propname)) == ZPROP_INVAL) { 404 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 405 "invalid property '%s'"), propname); 406 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 407 goto error; 408 } 409 410 if (zpool_prop_readonly(prop)) { 411 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' " 412 "is readonly"), propname); 413 (void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf); 414 goto error; 415 } 416 417 if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_POOL, retprops, 418 &strval, &intval, errbuf) != 0) 419 goto error; 420 421 /* 422 * Perform additional checking for specific properties. 423 */ 424 switch (prop) { 425 case ZPOOL_PROP_VERSION: 426 if (intval < version || intval > SPA_VERSION) { 427 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 428 "property '%s' number %d is invalid."), 429 propname, intval); 430 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf); 431 goto error; 432 } 433 break; 434 435 case ZPOOL_PROP_BOOTFS: 436 if (create_or_import) { 437 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 438 "property '%s' cannot be set at creation " 439 "or import time"), propname); 440 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 441 goto error; 442 } 443 444 if (version < SPA_VERSION_BOOTFS) { 445 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 446 "pool must be upgraded to support " 447 "'%s' property"), propname); 448 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf); 449 goto error; 450 } 451 452 /* 453 * bootfs property value has to be a dataset name and 454 * the dataset has to be in the same pool as it sets to. 455 */ 456 if (strval[0] != '\0' && !bootfs_name_valid(poolname, 457 strval)) { 458 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' " 459 "is an invalid name"), strval); 460 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf); 461 goto error; 462 } 463 464 if ((zhp = zpool_open_canfail(hdl, poolname)) == NULL) { 465 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 466 "could not open pool '%s'"), poolname); 467 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf); 468 goto error; 469 } 470 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL), 471 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 472 473 /* 474 * bootfs property cannot be set on a disk which has 475 * been EFI labeled. 476 */ 477 if (pool_uses_efi(nvroot)) { 478 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 479 "property '%s' not supported on " 480 "EFI labeled devices"), propname); 481 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf); 482 zpool_close(zhp); 483 goto error; 484 } 485 zpool_close(zhp); 486 break; 487 488 case ZPOOL_PROP_ALTROOT: 489 if (!create_or_import) { 490 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 491 "property '%s' can only be set during pool " 492 "creation or import"), propname); 493 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 494 goto error; 495 } 496 497 if (strval[0] != '/') { 498 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 499 "bad alternate root '%s'"), strval); 500 (void) zfs_error(hdl, EZFS_BADPATH, errbuf); 501 goto error; 502 } 503 break; 504 505 case ZPOOL_PROP_CACHEFILE: 506 if (strval[0] == '\0') 507 break; 508 509 if (strcmp(strval, "none") == 0) 510 break; 511 512 if (strval[0] != '/') { 513 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 514 "property '%s' must be empty, an " 515 "absolute path, or 'none'"), propname); 516 (void) zfs_error(hdl, EZFS_BADPATH, errbuf); 517 goto error; 518 } 519 520 slash = strrchr(strval, '/'); 521 522 if (slash[1] == '\0' || strcmp(slash, "/.") == 0 || 523 strcmp(slash, "/..") == 0) { 524 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 525 "'%s' is not a valid file"), strval); 526 (void) zfs_error(hdl, EZFS_BADPATH, errbuf); 527 goto error; 528 } 529 530 *slash = '\0'; 531 532 if (strval[0] != '\0' && 533 (stat64(strval, &statbuf) != 0 || 534 !S_ISDIR(statbuf.st_mode))) { 535 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 536 "'%s' is not a valid directory"), 537 strval); 538 (void) zfs_error(hdl, EZFS_BADPATH, errbuf); 539 goto error; 540 } 541 542 *slash = '/'; 543 break; 544 } 545 } 546 547 return (retprops); 548 error: 549 nvlist_free(retprops); 550 return (NULL); 551 } 552 553 /* 554 * Set zpool property : propname=propval. 555 */ 556 int 557 zpool_set_prop(zpool_handle_t *zhp, const char *propname, const char *propval) 558 { 559 zfs_cmd_t zc = { 0 }; 560 int ret = -1; 561 char errbuf[1024]; 562 nvlist_t *nvl = NULL; 563 nvlist_t *realprops; 564 uint64_t version; 565 566 (void) snprintf(errbuf, sizeof (errbuf), 567 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"), 568 zhp->zpool_name); 569 570 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0) 571 return (no_memory(zhp->zpool_hdl)); 572 573 if (nvlist_add_string(nvl, propname, propval) != 0) { 574 nvlist_free(nvl); 575 return (no_memory(zhp->zpool_hdl)); 576 } 577 578 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL); 579 if ((realprops = zpool_valid_proplist(zhp->zpool_hdl, 580 zhp->zpool_name, nvl, version, B_FALSE, errbuf)) == NULL) { 581 nvlist_free(nvl); 582 return (-1); 583 } 584 585 nvlist_free(nvl); 586 nvl = realprops; 587 588 /* 589 * Execute the corresponding ioctl() to set this property. 590 */ 591 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 592 593 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) { 594 nvlist_free(nvl); 595 return (-1); 596 } 597 598 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc); 599 600 zcmd_free_nvlists(&zc); 601 nvlist_free(nvl); 602 603 if (ret) 604 (void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf); 605 else 606 (void) zpool_props_refresh(zhp); 607 608 return (ret); 609 } 610 611 int 612 zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp) 613 { 614 libzfs_handle_t *hdl = zhp->zpool_hdl; 615 zprop_list_t *entry; 616 char buf[ZFS_MAXPROPLEN]; 617 618 if (zprop_expand_list(hdl, plp, ZFS_TYPE_POOL) != 0) 619 return (-1); 620 621 for (entry = *plp; entry != NULL; entry = entry->pl_next) { 622 623 if (entry->pl_fixed) 624 continue; 625 626 if (entry->pl_prop != ZPROP_INVAL && 627 zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf), 628 NULL) == 0) { 629 if (strlen(buf) > entry->pl_width) 630 entry->pl_width = strlen(buf); 631 } 632 } 633 634 return (0); 635 } 636 637 638 /* 639 * Don't start the slice at the default block of 34; many storage 640 * devices will use a stripe width of 128k, so start there instead. 641 */ 642 #define NEW_START_BLOCK 256 643 644 /* 645 * Validate the given pool name, optionally putting an extended error message in 646 * 'buf'. 647 */ 648 boolean_t 649 zpool_name_valid(libzfs_handle_t *hdl, boolean_t isopen, const char *pool) 650 { 651 namecheck_err_t why; 652 char what; 653 int ret; 654 655 ret = pool_namecheck(pool, &why, &what); 656 657 /* 658 * The rules for reserved pool names were extended at a later point. 659 * But we need to support users with existing pools that may now be 660 * invalid. So we only check for this expanded set of names during a 661 * create (or import), and only in userland. 662 */ 663 if (ret == 0 && !isopen && 664 (strncmp(pool, "mirror", 6) == 0 || 665 strncmp(pool, "raidz", 5) == 0 || 666 strncmp(pool, "spare", 5) == 0 || 667 strcmp(pool, "log") == 0)) { 668 if (hdl != NULL) 669 zfs_error_aux(hdl, 670 dgettext(TEXT_DOMAIN, "name is reserved")); 671 return (B_FALSE); 672 } 673 674 675 if (ret != 0) { 676 if (hdl != NULL) { 677 switch (why) { 678 case NAME_ERR_TOOLONG: 679 zfs_error_aux(hdl, 680 dgettext(TEXT_DOMAIN, "name is too long")); 681 break; 682 683 case NAME_ERR_INVALCHAR: 684 zfs_error_aux(hdl, 685 dgettext(TEXT_DOMAIN, "invalid character " 686 "'%c' in pool name"), what); 687 break; 688 689 case NAME_ERR_NOLETTER: 690 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 691 "name must begin with a letter")); 692 break; 693 694 case NAME_ERR_RESERVED: 695 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 696 "name is reserved")); 697 break; 698 699 case NAME_ERR_DISKLIKE: 700 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 701 "pool name is reserved")); 702 break; 703 704 case NAME_ERR_LEADING_SLASH: 705 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 706 "leading slash in name")); 707 break; 708 709 case NAME_ERR_EMPTY_COMPONENT: 710 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 711 "empty component in name")); 712 break; 713 714 case NAME_ERR_TRAILING_SLASH: 715 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 716 "trailing slash in name")); 717 break; 718 719 case NAME_ERR_MULTIPLE_AT: 720 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 721 "multiple '@' delimiters in name")); 722 break; 723 724 } 725 } 726 return (B_FALSE); 727 } 728 729 return (B_TRUE); 730 } 731 732 /* 733 * Open a handle to the given pool, even if the pool is currently in the FAULTED 734 * state. 735 */ 736 zpool_handle_t * 737 zpool_open_canfail(libzfs_handle_t *hdl, const char *pool) 738 { 739 zpool_handle_t *zhp; 740 boolean_t missing; 741 742 /* 743 * Make sure the pool name is valid. 744 */ 745 if (!zpool_name_valid(hdl, B_TRUE, pool)) { 746 (void) zfs_error_fmt(hdl, EZFS_INVALIDNAME, 747 dgettext(TEXT_DOMAIN, "cannot open '%s'"), 748 pool); 749 return (NULL); 750 } 751 752 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL) 753 return (NULL); 754 755 zhp->zpool_hdl = hdl; 756 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name)); 757 758 if (zpool_refresh_stats(zhp, &missing) != 0) { 759 zpool_close(zhp); 760 return (NULL); 761 } 762 763 if (missing) { 764 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool")); 765 (void) zfs_error_fmt(hdl, EZFS_NOENT, 766 dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool); 767 zpool_close(zhp); 768 return (NULL); 769 } 770 771 return (zhp); 772 } 773 774 /* 775 * Like the above, but silent on error. Used when iterating over pools (because 776 * the configuration cache may be out of date). 777 */ 778 int 779 zpool_open_silent(libzfs_handle_t *hdl, const char *pool, zpool_handle_t **ret) 780 { 781 zpool_handle_t *zhp; 782 boolean_t missing; 783 784 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL) 785 return (-1); 786 787 zhp->zpool_hdl = hdl; 788 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name)); 789 790 if (zpool_refresh_stats(zhp, &missing) != 0) { 791 zpool_close(zhp); 792 return (-1); 793 } 794 795 if (missing) { 796 zpool_close(zhp); 797 *ret = NULL; 798 return (0); 799 } 800 801 *ret = zhp; 802 return (0); 803 } 804 805 /* 806 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted 807 * state. 808 */ 809 zpool_handle_t * 810 zpool_open(libzfs_handle_t *hdl, const char *pool) 811 { 812 zpool_handle_t *zhp; 813 814 if ((zhp = zpool_open_canfail(hdl, pool)) == NULL) 815 return (NULL); 816 817 if (zhp->zpool_state == POOL_STATE_UNAVAIL) { 818 (void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL, 819 dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name); 820 zpool_close(zhp); 821 return (NULL); 822 } 823 824 return (zhp); 825 } 826 827 /* 828 * Close the handle. Simply frees the memory associated with the handle. 829 */ 830 void 831 zpool_close(zpool_handle_t *zhp) 832 { 833 if (zhp->zpool_config) 834 nvlist_free(zhp->zpool_config); 835 if (zhp->zpool_old_config) 836 nvlist_free(zhp->zpool_old_config); 837 if (zhp->zpool_props) 838 nvlist_free(zhp->zpool_props); 839 free(zhp); 840 } 841 842 /* 843 * Return the name of the pool. 844 */ 845 const char * 846 zpool_get_name(zpool_handle_t *zhp) 847 { 848 return (zhp->zpool_name); 849 } 850 851 852 /* 853 * Return the state of the pool (ACTIVE or UNAVAILABLE) 854 */ 855 int 856 zpool_get_state(zpool_handle_t *zhp) 857 { 858 return (zhp->zpool_state); 859 } 860 861 /* 862 * Create the named pool, using the provided vdev list. It is assumed 863 * that the consumer has already validated the contents of the nvlist, so we 864 * don't have to worry about error semantics. 865 */ 866 int 867 zpool_create(libzfs_handle_t *hdl, const char *pool, nvlist_t *nvroot, 868 nvlist_t *props, nvlist_t *fsprops) 869 { 870 zfs_cmd_t zc = { 0 }; 871 nvlist_t *zc_fsprops = NULL; 872 nvlist_t *zc_props = NULL; 873 char msg[1024]; 874 char *altroot; 875 int ret = -1; 876 877 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 878 "cannot create '%s'"), pool); 879 880 if (!zpool_name_valid(hdl, B_FALSE, pool)) 881 return (zfs_error(hdl, EZFS_INVALIDNAME, msg)); 882 883 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0) 884 return (-1); 885 886 if (props) { 887 if ((zc_props = zpool_valid_proplist(hdl, pool, props, 888 SPA_VERSION_1, B_TRUE, msg)) == NULL) { 889 goto create_failed; 890 } 891 } 892 893 if (fsprops) { 894 uint64_t zoned; 895 char *zonestr; 896 897 zoned = ((nvlist_lookup_string(fsprops, 898 zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) && 899 strcmp(zonestr, "on") == 0); 900 901 if ((zc_fsprops = zfs_valid_proplist(hdl, 902 ZFS_TYPE_FILESYSTEM, fsprops, zoned, NULL, msg)) == NULL) { 903 goto create_failed; 904 } 905 if (!zc_props && 906 (nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) { 907 goto create_failed; 908 } 909 if (nvlist_add_nvlist(zc_props, 910 ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) { 911 goto create_failed; 912 } 913 } 914 915 if (zc_props && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0) 916 goto create_failed; 917 918 (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name)); 919 920 if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) { 921 922 zcmd_free_nvlists(&zc); 923 nvlist_free(zc_props); 924 nvlist_free(zc_fsprops); 925 926 switch (errno) { 927 case EBUSY: 928 /* 929 * This can happen if the user has specified the same 930 * device multiple times. We can't reliably detect this 931 * until we try to add it and see we already have a 932 * label. 933 */ 934 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 935 "one or more vdevs refer to the same device")); 936 return (zfs_error(hdl, EZFS_BADDEV, msg)); 937 938 case EOVERFLOW: 939 /* 940 * This occurs when one of the devices is below 941 * SPA_MINDEVSIZE. Unfortunately, we can't detect which 942 * device was the problem device since there's no 943 * reliable way to determine device size from userland. 944 */ 945 { 946 char buf[64]; 947 948 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf)); 949 950 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 951 "one or more devices is less than the " 952 "minimum size (%s)"), buf); 953 } 954 return (zfs_error(hdl, EZFS_BADDEV, msg)); 955 956 case ENOSPC: 957 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 958 "one or more devices is out of space")); 959 return (zfs_error(hdl, EZFS_BADDEV, msg)); 960 961 case ENOTBLK: 962 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 963 "cache device must be a disk or disk slice")); 964 return (zfs_error(hdl, EZFS_BADDEV, msg)); 965 966 default: 967 return (zpool_standard_error(hdl, errno, msg)); 968 } 969 } 970 971 /* 972 * If this is an alternate root pool, then we automatically set the 973 * mountpoint of the root dataset to be '/'. 974 */ 975 if (nvlist_lookup_string(props, zpool_prop_to_name(ZPOOL_PROP_ALTROOT), 976 &altroot) == 0) { 977 zfs_handle_t *zhp; 978 979 verify((zhp = zfs_open(hdl, pool, ZFS_TYPE_DATASET)) != NULL); 980 verify(zfs_prop_set(zhp, zfs_prop_to_name(ZFS_PROP_MOUNTPOINT), 981 "/") == 0); 982 983 zfs_close(zhp); 984 } 985 986 create_failed: 987 zcmd_free_nvlists(&zc); 988 nvlist_free(zc_props); 989 nvlist_free(zc_fsprops); 990 return (ret); 991 } 992 993 /* 994 * Destroy the given pool. It is up to the caller to ensure that there are no 995 * datasets left in the pool. 996 */ 997 int 998 zpool_destroy(zpool_handle_t *zhp) 999 { 1000 zfs_cmd_t zc = { 0 }; 1001 zfs_handle_t *zfp = NULL; 1002 libzfs_handle_t *hdl = zhp->zpool_hdl; 1003 char msg[1024]; 1004 1005 if (zhp->zpool_state == POOL_STATE_ACTIVE && 1006 (zfp = zfs_open(zhp->zpool_hdl, zhp->zpool_name, 1007 ZFS_TYPE_FILESYSTEM)) == NULL) 1008 return (-1); 1009 1010 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1011 1012 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) { 1013 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1014 "cannot destroy '%s'"), zhp->zpool_name); 1015 1016 if (errno == EROFS) { 1017 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1018 "one or more devices is read only")); 1019 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1020 } else { 1021 (void) zpool_standard_error(hdl, errno, msg); 1022 } 1023 1024 if (zfp) 1025 zfs_close(zfp); 1026 return (-1); 1027 } 1028 1029 if (zfp) { 1030 remove_mountpoint(zfp); 1031 zfs_close(zfp); 1032 } 1033 1034 return (0); 1035 } 1036 1037 /* 1038 * Add the given vdevs to the pool. The caller must have already performed the 1039 * necessary verification to ensure that the vdev specification is well-formed. 1040 */ 1041 int 1042 zpool_add(zpool_handle_t *zhp, nvlist_t *nvroot) 1043 { 1044 zfs_cmd_t zc = { 0 }; 1045 int ret; 1046 libzfs_handle_t *hdl = zhp->zpool_hdl; 1047 char msg[1024]; 1048 nvlist_t **spares, **l2cache; 1049 uint_t nspares, nl2cache; 1050 1051 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1052 "cannot add to '%s'"), zhp->zpool_name); 1053 1054 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) < 1055 SPA_VERSION_SPARES && 1056 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, 1057 &spares, &nspares) == 0) { 1058 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be " 1059 "upgraded to add hot spares")); 1060 return (zfs_error(hdl, EZFS_BADVERSION, msg)); 1061 } 1062 1063 if (pool_is_bootable(zhp) && nvlist_lookup_nvlist_array(nvroot, 1064 ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0) { 1065 uint64_t s; 1066 1067 for (s = 0; s < nspares; s++) { 1068 char *path; 1069 1070 if (nvlist_lookup_string(spares[s], ZPOOL_CONFIG_PATH, 1071 &path) == 0 && pool_uses_efi(spares[s])) { 1072 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1073 "device '%s' contains an EFI label and " 1074 "cannot be used on root pools."), 1075 zpool_vdev_name(hdl, NULL, spares[s], 1076 B_FALSE)); 1077 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg)); 1078 } 1079 } 1080 } 1081 1082 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) < 1083 SPA_VERSION_L2CACHE && 1084 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, 1085 &l2cache, &nl2cache) == 0) { 1086 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be " 1087 "upgraded to add cache devices")); 1088 return (zfs_error(hdl, EZFS_BADVERSION, msg)); 1089 } 1090 1091 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0) 1092 return (-1); 1093 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1094 1095 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) { 1096 switch (errno) { 1097 case EBUSY: 1098 /* 1099 * This can happen if the user has specified the same 1100 * device multiple times. We can't reliably detect this 1101 * until we try to add it and see we already have a 1102 * label. 1103 */ 1104 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1105 "one or more vdevs refer to the same device")); 1106 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1107 break; 1108 1109 case EOVERFLOW: 1110 /* 1111 * This occurrs when one of the devices is below 1112 * SPA_MINDEVSIZE. Unfortunately, we can't detect which 1113 * device was the problem device since there's no 1114 * reliable way to determine device size from userland. 1115 */ 1116 { 1117 char buf[64]; 1118 1119 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf)); 1120 1121 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1122 "device is less than the minimum " 1123 "size (%s)"), buf); 1124 } 1125 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1126 break; 1127 1128 case ENOTSUP: 1129 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1130 "pool must be upgraded to add these vdevs")); 1131 (void) zfs_error(hdl, EZFS_BADVERSION, msg); 1132 break; 1133 1134 case EDOM: 1135 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1136 "root pool can not have multiple vdevs" 1137 " or separate logs")); 1138 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, msg); 1139 break; 1140 1141 case ENOTBLK: 1142 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1143 "cache device must be a disk or disk slice")); 1144 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1145 break; 1146 1147 default: 1148 (void) zpool_standard_error(hdl, errno, msg); 1149 } 1150 1151 ret = -1; 1152 } else { 1153 ret = 0; 1154 } 1155 1156 zcmd_free_nvlists(&zc); 1157 1158 return (ret); 1159 } 1160 1161 /* 1162 * Exports the pool from the system. The caller must ensure that there are no 1163 * mounted datasets in the pool. 1164 */ 1165 int 1166 zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce) 1167 { 1168 zfs_cmd_t zc = { 0 }; 1169 char msg[1024]; 1170 1171 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1172 "cannot export '%s'"), zhp->zpool_name); 1173 1174 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1175 zc.zc_cookie = force; 1176 zc.zc_guid = hardforce; 1177 1178 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) { 1179 switch (errno) { 1180 case EXDEV: 1181 zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN, 1182 "use '-f' to override the following errors:\n" 1183 "'%s' has an active shared spare which could be" 1184 " used by other pools once '%s' is exported."), 1185 zhp->zpool_name, zhp->zpool_name); 1186 return (zfs_error(zhp->zpool_hdl, EZFS_ACTIVE_SPARE, 1187 msg)); 1188 default: 1189 return (zpool_standard_error_fmt(zhp->zpool_hdl, errno, 1190 msg)); 1191 } 1192 } 1193 1194 return (0); 1195 } 1196 1197 int 1198 zpool_export(zpool_handle_t *zhp, boolean_t force) 1199 { 1200 return (zpool_export_common(zhp, force, B_FALSE)); 1201 } 1202 1203 int 1204 zpool_export_force(zpool_handle_t *zhp) 1205 { 1206 return (zpool_export_common(zhp, B_TRUE, B_TRUE)); 1207 } 1208 1209 static void 1210 zpool_rewind_exclaim(libzfs_handle_t *hdl, const char *name, boolean_t dryrun, 1211 nvlist_t *rbi) 1212 { 1213 uint64_t rewindto; 1214 int64_t loss = -1; 1215 struct tm t; 1216 char timestr[128]; 1217 1218 if (!hdl->libzfs_printerr || rbi == NULL) 1219 return; 1220 1221 if (nvlist_lookup_uint64(rbi, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0) 1222 return; 1223 (void) nvlist_lookup_int64(rbi, ZPOOL_CONFIG_REWIND_TIME, &loss); 1224 1225 if (localtime_r((time_t *)&rewindto, &t) != NULL && 1226 strftime(timestr, 128, 0, &t) != 0) { 1227 if (dryrun) { 1228 (void) printf(dgettext(TEXT_DOMAIN, 1229 "Would be able to return %s " 1230 "to its state as of %s.\n"), 1231 name, timestr); 1232 } else { 1233 (void) printf(dgettext(TEXT_DOMAIN, 1234 "Pool %s returned to its state as of %s.\n"), 1235 name, timestr); 1236 } 1237 if (loss > 120) { 1238 (void) printf(dgettext(TEXT_DOMAIN, 1239 "%s approximately %lld "), 1240 dryrun ? "Would discard" : "Discarded", 1241 (loss + 30) / 60); 1242 (void) printf(dgettext(TEXT_DOMAIN, 1243 "minutes of transactions.\n")); 1244 } else if (loss > 0) { 1245 (void) printf(dgettext(TEXT_DOMAIN, 1246 "%s approximately %lld "), 1247 dryrun ? "Would discard" : "Discarded", loss); 1248 (void) printf(dgettext(TEXT_DOMAIN, 1249 "seconds of transactions.\n")); 1250 } 1251 } 1252 } 1253 1254 void 1255 zpool_explain_recover(libzfs_handle_t *hdl, const char *name, int reason, 1256 nvlist_t *config) 1257 { 1258 int64_t loss = -1; 1259 uint64_t edata = UINT64_MAX; 1260 uint64_t rewindto; 1261 struct tm t; 1262 char timestr[128]; 1263 1264 if (!hdl->libzfs_printerr) 1265 return; 1266 1267 if (reason >= 0) 1268 (void) printf(dgettext(TEXT_DOMAIN, "action: ")); 1269 else 1270 (void) printf(dgettext(TEXT_DOMAIN, "\t")); 1271 1272 /* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */ 1273 if (nvlist_lookup_uint64(config, 1274 ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0) 1275 goto no_info; 1276 1277 (void) nvlist_lookup_int64(config, ZPOOL_CONFIG_REWIND_TIME, &loss); 1278 (void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_LOAD_DATA_ERRORS, 1279 &edata); 1280 1281 (void) printf(dgettext(TEXT_DOMAIN, 1282 "Recovery is possible, but will result in some data loss.\n")); 1283 1284 if (localtime_r((time_t *)&rewindto, &t) != NULL && 1285 strftime(timestr, 128, 0, &t) != 0) { 1286 (void) printf(dgettext(TEXT_DOMAIN, 1287 "\tReturning the pool to its state as of %s\n" 1288 "\tshould correct the problem. "), 1289 timestr); 1290 } else { 1291 (void) printf(dgettext(TEXT_DOMAIN, 1292 "\tReverting the pool to an earlier state " 1293 "should correct the problem.\n\t")); 1294 } 1295 1296 if (loss > 120) { 1297 (void) printf(dgettext(TEXT_DOMAIN, 1298 "Approximately %lld minutes of data\n" 1299 "\tmust be discarded, irreversibly. "), (loss + 30) / 60); 1300 } else if (loss > 0) { 1301 (void) printf(dgettext(TEXT_DOMAIN, 1302 "Approximately %lld seconds of data\n" 1303 "\tmust be discarded, irreversibly. "), loss); 1304 } 1305 if (edata != 0 && edata != UINT64_MAX) { 1306 if (edata == 1) { 1307 (void) printf(dgettext(TEXT_DOMAIN, 1308 "After rewind, at least\n" 1309 "\tone persistent user-data error will remain. ")); 1310 } else { 1311 (void) printf(dgettext(TEXT_DOMAIN, 1312 "After rewind, several\n" 1313 "\tpersistent user-data errors will remain. ")); 1314 } 1315 } 1316 (void) printf(dgettext(TEXT_DOMAIN, 1317 "Recovery can be attempted\n\tby executing 'zpool %s -F %s'. "), 1318 reason >= 0 ? "clear" : "import", name); 1319 1320 (void) printf(dgettext(TEXT_DOMAIN, 1321 "A scrub of the pool\n" 1322 "\tis strongly recommended after recovery.\n")); 1323 return; 1324 1325 no_info: 1326 (void) printf(dgettext(TEXT_DOMAIN, 1327 "Destroy and re-create the pool from\n\ta backup source.\n")); 1328 } 1329 1330 /* 1331 * zpool_import() is a contracted interface. Should be kept the same 1332 * if possible. 1333 * 1334 * Applications should use zpool_import_props() to import a pool with 1335 * new properties value to be set. 1336 */ 1337 int 1338 zpool_import(libzfs_handle_t *hdl, nvlist_t *config, const char *newname, 1339 char *altroot) 1340 { 1341 nvlist_t *props = NULL; 1342 int ret; 1343 1344 if (altroot != NULL) { 1345 if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) { 1346 return (zfs_error_fmt(hdl, EZFS_NOMEM, 1347 dgettext(TEXT_DOMAIN, "cannot import '%s'"), 1348 newname)); 1349 } 1350 1351 if (nvlist_add_string(props, 1352 zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 || 1353 nvlist_add_string(props, 1354 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) { 1355 nvlist_free(props); 1356 return (zfs_error_fmt(hdl, EZFS_NOMEM, 1357 dgettext(TEXT_DOMAIN, "cannot import '%s'"), 1358 newname)); 1359 } 1360 } 1361 1362 ret = zpool_import_props(hdl, config, newname, props, B_FALSE); 1363 if (props) 1364 nvlist_free(props); 1365 return (ret); 1366 } 1367 1368 /* 1369 * Import the given pool using the known configuration and a list of 1370 * properties to be set. The configuration should have come from 1371 * zpool_find_import(). The 'newname' parameters control whether the pool 1372 * is imported with a different name. 1373 */ 1374 int 1375 zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname, 1376 nvlist_t *props, boolean_t importfaulted) 1377 { 1378 zfs_cmd_t zc = { 0 }; 1379 zpool_rewind_policy_t policy; 1380 nvlist_t *nvi = NULL; 1381 char *thename; 1382 char *origname; 1383 uint64_t returned_size; 1384 int ret; 1385 char errbuf[1024]; 1386 1387 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, 1388 &origname) == 0); 1389 1390 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, 1391 "cannot import pool '%s'"), origname); 1392 1393 if (newname != NULL) { 1394 if (!zpool_name_valid(hdl, B_FALSE, newname)) 1395 return (zfs_error_fmt(hdl, EZFS_INVALIDNAME, 1396 dgettext(TEXT_DOMAIN, "cannot import '%s'"), 1397 newname)); 1398 thename = (char *)newname; 1399 } else { 1400 thename = origname; 1401 } 1402 1403 if (props) { 1404 uint64_t version; 1405 1406 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, 1407 &version) == 0); 1408 1409 if ((props = zpool_valid_proplist(hdl, origname, 1410 props, version, B_TRUE, errbuf)) == NULL) { 1411 return (-1); 1412 } else if (zcmd_write_src_nvlist(hdl, &zc, props) != 0) { 1413 nvlist_free(props); 1414 return (-1); 1415 } 1416 } 1417 1418 (void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name)); 1419 1420 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, 1421 &zc.zc_guid) == 0); 1422 1423 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) { 1424 nvlist_free(props); 1425 return (-1); 1426 } 1427 returned_size = zc.zc_nvlist_conf_size + 512; 1428 if (zcmd_alloc_dst_nvlist(hdl, &zc, returned_size) != 0) { 1429 nvlist_free(props); 1430 return (-1); 1431 } 1432 1433 zc.zc_cookie = (uint64_t)importfaulted; 1434 ret = 0; 1435 if (zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc) != 0) { 1436 char desc[1024]; 1437 1438 (void) zcmd_read_dst_nvlist(hdl, &zc, &nvi); 1439 zpool_get_rewind_policy(config, &policy); 1440 /* 1441 * Dry-run failed, but we print out what success 1442 * looks like if we found a best txg 1443 */ 1444 if ((policy.zrp_request & ZPOOL_TRY_REWIND) && nvi) { 1445 zpool_rewind_exclaim(hdl, newname ? origname : thename, 1446 B_TRUE, nvi); 1447 nvlist_free(nvi); 1448 return (-1); 1449 } 1450 1451 if (newname == NULL) 1452 (void) snprintf(desc, sizeof (desc), 1453 dgettext(TEXT_DOMAIN, "cannot import '%s'"), 1454 thename); 1455 else 1456 (void) snprintf(desc, sizeof (desc), 1457 dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"), 1458 origname, thename); 1459 1460 switch (errno) { 1461 case ENOTSUP: 1462 /* 1463 * Unsupported version. 1464 */ 1465 (void) zfs_error(hdl, EZFS_BADVERSION, desc); 1466 break; 1467 1468 case EINVAL: 1469 (void) zfs_error(hdl, EZFS_INVALCONFIG, desc); 1470 break; 1471 1472 case EROFS: 1473 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1474 "one or more devices is read only")); 1475 (void) zfs_error(hdl, EZFS_BADDEV, desc); 1476 break; 1477 1478 default: 1479 (void) zcmd_read_dst_nvlist(hdl, &zc, &nvi); 1480 (void) zpool_standard_error(hdl, errno, desc); 1481 zpool_explain_recover(hdl, 1482 newname ? origname : thename, -errno, nvi); 1483 nvlist_free(nvi); 1484 break; 1485 } 1486 1487 ret = -1; 1488 } else { 1489 zpool_handle_t *zhp; 1490 1491 /* 1492 * This should never fail, but play it safe anyway. 1493 */ 1494 if (zpool_open_silent(hdl, thename, &zhp) != 0) 1495 ret = -1; 1496 else if (zhp != NULL) 1497 zpool_close(zhp); 1498 (void) zcmd_read_dst_nvlist(hdl, &zc, &nvi); 1499 zpool_get_rewind_policy(config, &policy); 1500 if (policy.zrp_request & 1501 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) { 1502 zpool_rewind_exclaim(hdl, newname ? origname : thename, 1503 ((policy.zrp_request & ZPOOL_TRY_REWIND) != 0), 1504 nvi); 1505 } 1506 nvlist_free(nvi); 1507 return (0); 1508 } 1509 1510 zcmd_free_nvlists(&zc); 1511 nvlist_free(props); 1512 1513 return (ret); 1514 } 1515 1516 /* 1517 * Scan the pool. 1518 */ 1519 int 1520 zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func) 1521 { 1522 zfs_cmd_t zc = { 0 }; 1523 char msg[1024]; 1524 libzfs_handle_t *hdl = zhp->zpool_hdl; 1525 1526 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1527 zc.zc_cookie = func; 1528 1529 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SCAN, &zc) == 0 || 1530 (errno == ENOENT && func != POOL_SCAN_NONE)) 1531 return (0); 1532 1533 if (func == POOL_SCAN_SCRUB) { 1534 (void) snprintf(msg, sizeof (msg), 1535 dgettext(TEXT_DOMAIN, "cannot scrub %s"), zc.zc_name); 1536 } else if (func == POOL_SCAN_NONE) { 1537 (void) snprintf(msg, sizeof (msg), 1538 dgettext(TEXT_DOMAIN, "cannot cancel scrubbing %s"), 1539 zc.zc_name); 1540 } else { 1541 assert(!"unexpected result"); 1542 } 1543 1544 if (errno == EBUSY) { 1545 nvlist_t *nvroot; 1546 pool_scan_stat_t *ps = NULL; 1547 uint_t psc; 1548 1549 verify(nvlist_lookup_nvlist(zhp->zpool_config, 1550 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 1551 (void) nvlist_lookup_uint64_array(nvroot, 1552 ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &psc); 1553 if (ps && ps->pss_func == POOL_SCAN_SCRUB) 1554 return (zfs_error(hdl, EZFS_SCRUBBING, msg)); 1555 else 1556 return (zfs_error(hdl, EZFS_RESILVERING, msg)); 1557 } else if (errno == ENOENT) { 1558 return (zfs_error(hdl, EZFS_NO_SCRUB, msg)); 1559 } else { 1560 return (zpool_standard_error(hdl, errno, msg)); 1561 } 1562 } 1563 1564 /* 1565 * Find a vdev that matches the search criteria specified. We use the 1566 * the nvpair name to determine how we should look for the device. 1567 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL 1568 * spare; but FALSE if its an INUSE spare. 1569 */ 1570 static nvlist_t * 1571 vdev_to_nvlist_iter(nvlist_t *nv, nvlist_t *search, boolean_t *avail_spare, 1572 boolean_t *l2cache, boolean_t *log) 1573 { 1574 uint_t c, children; 1575 nvlist_t **child; 1576 nvlist_t *ret; 1577 uint64_t is_log; 1578 char *srchkey; 1579 nvpair_t *pair = nvlist_next_nvpair(search, NULL); 1580 1581 /* Nothing to look for */ 1582 if (search == NULL || pair == NULL) 1583 return (NULL); 1584 1585 /* Obtain the key we will use to search */ 1586 srchkey = nvpair_name(pair); 1587 1588 switch (nvpair_type(pair)) { 1589 case DATA_TYPE_UINT64: { 1590 uint64_t srchval, theguid, present; 1591 1592 verify(nvpair_value_uint64(pair, &srchval) == 0); 1593 if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) { 1594 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, 1595 &present) == 0) { 1596 /* 1597 * If the device has never been present since 1598 * import, the only reliable way to match the 1599 * vdev is by GUID. 1600 */ 1601 verify(nvlist_lookup_uint64(nv, 1602 ZPOOL_CONFIG_GUID, &theguid) == 0); 1603 if (theguid == srchval) 1604 return (nv); 1605 } 1606 } 1607 break; 1608 } 1609 1610 case DATA_TYPE_STRING: { 1611 char *srchval, *val; 1612 1613 verify(nvpair_value_string(pair, &srchval) == 0); 1614 if (nvlist_lookup_string(nv, srchkey, &val) != 0) 1615 break; 1616 1617 /* 1618 * Search for the requested value. We special case the search 1619 * for ZPOOL_CONFIG_PATH when it's a wholedisk and when 1620 * Looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE). 1621 * Otherwise, all other searches are simple string compares. 1622 */ 1623 if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0 && val) { 1624 uint64_t wholedisk = 0; 1625 1626 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, 1627 &wholedisk); 1628 if (wholedisk) { 1629 /* 1630 * For whole disks, the internal path has 's0', 1631 * but the path passed in by the user doesn't. 1632 */ 1633 if (strlen(srchval) == strlen(val) - 2 && 1634 strncmp(srchval, val, strlen(srchval)) == 0) 1635 return (nv); 1636 break; 1637 } 1638 } else if (strcmp(srchkey, ZPOOL_CONFIG_TYPE) == 0 && val) { 1639 char *type, *idx, *end, *p; 1640 uint64_t id, vdev_id; 1641 1642 /* 1643 * Determine our vdev type, keeping in mind 1644 * that the srchval is composed of a type and 1645 * vdev id pair (i.e. mirror-4). 1646 */ 1647 if ((type = strdup(srchval)) == NULL) 1648 return (NULL); 1649 1650 if ((p = strrchr(type, '-')) == NULL) { 1651 free(type); 1652 break; 1653 } 1654 idx = p + 1; 1655 *p = '\0'; 1656 1657 /* 1658 * If the types don't match then keep looking. 1659 */ 1660 if (strncmp(val, type, strlen(val)) != 0) { 1661 free(type); 1662 break; 1663 } 1664 1665 verify(strncmp(type, VDEV_TYPE_RAIDZ, 1666 strlen(VDEV_TYPE_RAIDZ)) == 0 || 1667 strncmp(type, VDEV_TYPE_MIRROR, 1668 strlen(VDEV_TYPE_MIRROR)) == 0); 1669 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID, 1670 &id) == 0); 1671 1672 errno = 0; 1673 vdev_id = strtoull(idx, &end, 10); 1674 1675 free(type); 1676 if (errno != 0) 1677 return (NULL); 1678 1679 /* 1680 * Now verify that we have the correct vdev id. 1681 */ 1682 if (vdev_id == id) 1683 return (nv); 1684 } 1685 1686 /* 1687 * Common case 1688 */ 1689 if (strcmp(srchval, val) == 0) 1690 return (nv); 1691 break; 1692 } 1693 1694 default: 1695 break; 1696 } 1697 1698 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, 1699 &child, &children) != 0) 1700 return (NULL); 1701 1702 for (c = 0; c < children; c++) { 1703 if ((ret = vdev_to_nvlist_iter(child[c], search, 1704 avail_spare, l2cache, NULL)) != NULL) { 1705 /* 1706 * The 'is_log' value is only set for the toplevel 1707 * vdev, not the leaf vdevs. So we always lookup the 1708 * log device from the root of the vdev tree (where 1709 * 'log' is non-NULL). 1710 */ 1711 if (log != NULL && 1712 nvlist_lookup_uint64(child[c], 1713 ZPOOL_CONFIG_IS_LOG, &is_log) == 0 && 1714 is_log) { 1715 *log = B_TRUE; 1716 } 1717 return (ret); 1718 } 1719 } 1720 1721 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES, 1722 &child, &children) == 0) { 1723 for (c = 0; c < children; c++) { 1724 if ((ret = vdev_to_nvlist_iter(child[c], search, 1725 avail_spare, l2cache, NULL)) != NULL) { 1726 *avail_spare = B_TRUE; 1727 return (ret); 1728 } 1729 } 1730 } 1731 1732 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE, 1733 &child, &children) == 0) { 1734 for (c = 0; c < children; c++) { 1735 if ((ret = vdev_to_nvlist_iter(child[c], search, 1736 avail_spare, l2cache, NULL)) != NULL) { 1737 *l2cache = B_TRUE; 1738 return (ret); 1739 } 1740 } 1741 } 1742 1743 return (NULL); 1744 } 1745 1746 /* 1747 * Given a physical path (minus the "/devices" prefix), find the 1748 * associated vdev. 1749 */ 1750 nvlist_t * 1751 zpool_find_vdev_by_physpath(zpool_handle_t *zhp, const char *ppath, 1752 boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log) 1753 { 1754 nvlist_t *search, *nvroot, *ret; 1755 1756 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1757 verify(nvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH, ppath) == 0); 1758 1759 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE, 1760 &nvroot) == 0); 1761 1762 *avail_spare = B_FALSE; 1763 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log); 1764 nvlist_free(search); 1765 1766 return (ret); 1767 } 1768 1769 /* 1770 * Determine if we have an "interior" top-level vdev (i.e mirror/raidz). 1771 */ 1772 boolean_t 1773 zpool_vdev_is_interior(const char *name) 1774 { 1775 if (strncmp(name, VDEV_TYPE_RAIDZ, strlen(VDEV_TYPE_RAIDZ)) == 0 || 1776 strncmp(name, VDEV_TYPE_MIRROR, strlen(VDEV_TYPE_MIRROR)) == 0) 1777 return (B_TRUE); 1778 return (B_FALSE); 1779 } 1780 1781 nvlist_t * 1782 zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare, 1783 boolean_t *l2cache, boolean_t *log) 1784 { 1785 char buf[MAXPATHLEN]; 1786 char *end; 1787 nvlist_t *nvroot, *search, *ret; 1788 uint64_t guid; 1789 1790 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1791 1792 guid = strtoull(path, &end, 10); 1793 if (guid != 0 && *end == '\0') { 1794 verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0); 1795 } else if (zpool_vdev_is_interior(path)) { 1796 verify(nvlist_add_string(search, ZPOOL_CONFIG_TYPE, path) == 0); 1797 } else if (path[0] != '/') { 1798 (void) snprintf(buf, sizeof (buf), "%s%s", "/dev/dsk/", path); 1799 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, buf) == 0); 1800 } else { 1801 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, path) == 0); 1802 } 1803 1804 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE, 1805 &nvroot) == 0); 1806 1807 *avail_spare = B_FALSE; 1808 *l2cache = B_FALSE; 1809 if (log != NULL) 1810 *log = B_FALSE; 1811 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log); 1812 nvlist_free(search); 1813 1814 return (ret); 1815 } 1816 1817 static int 1818 vdev_online(nvlist_t *nv) 1819 { 1820 uint64_t ival; 1821 1822 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE, &ival) == 0 || 1823 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED, &ival) == 0 || 1824 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED, &ival) == 0) 1825 return (0); 1826 1827 return (1); 1828 } 1829 1830 /* 1831 * Helper function for zpool_get_physpaths(). 1832 */ 1833 static int 1834 vdev_get_one_physpath(nvlist_t *config, char *physpath, size_t physpath_size, 1835 size_t *bytes_written) 1836 { 1837 size_t bytes_left, pos, rsz; 1838 char *tmppath; 1839 const char *format; 1840 1841 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PHYS_PATH, 1842 &tmppath) != 0) 1843 return (EZFS_NODEVICE); 1844 1845 pos = *bytes_written; 1846 bytes_left = physpath_size - pos; 1847 format = (pos == 0) ? "%s" : " %s"; 1848 1849 rsz = snprintf(physpath + pos, bytes_left, format, tmppath); 1850 *bytes_written += rsz; 1851 1852 if (rsz >= bytes_left) { 1853 /* if physpath was not copied properly, clear it */ 1854 if (bytes_left != 0) { 1855 physpath[pos] = 0; 1856 } 1857 return (EZFS_NOSPC); 1858 } 1859 return (0); 1860 } 1861 1862 static int 1863 vdev_get_physpaths(nvlist_t *nv, char *physpath, size_t phypath_size, 1864 size_t *rsz, boolean_t is_spare) 1865 { 1866 char *type; 1867 int ret; 1868 1869 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0) 1870 return (EZFS_INVALCONFIG); 1871 1872 if (strcmp(type, VDEV_TYPE_DISK) == 0) { 1873 /* 1874 * An active spare device has ZPOOL_CONFIG_IS_SPARE set. 1875 * For a spare vdev, we only want to boot from the active 1876 * spare device. 1877 */ 1878 if (is_spare) { 1879 uint64_t spare = 0; 1880 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE, 1881 &spare); 1882 if (!spare) 1883 return (EZFS_INVALCONFIG); 1884 } 1885 1886 if (vdev_online(nv)) { 1887 if ((ret = vdev_get_one_physpath(nv, physpath, 1888 phypath_size, rsz)) != 0) 1889 return (ret); 1890 } 1891 } else if (strcmp(type, VDEV_TYPE_MIRROR) == 0 || 1892 strcmp(type, VDEV_TYPE_REPLACING) == 0 || 1893 (is_spare = (strcmp(type, VDEV_TYPE_SPARE) == 0))) { 1894 nvlist_t **child; 1895 uint_t count; 1896 int i, ret; 1897 1898 if (nvlist_lookup_nvlist_array(nv, 1899 ZPOOL_CONFIG_CHILDREN, &child, &count) != 0) 1900 return (EZFS_INVALCONFIG); 1901 1902 for (i = 0; i < count; i++) { 1903 ret = vdev_get_physpaths(child[i], physpath, 1904 phypath_size, rsz, is_spare); 1905 if (ret == EZFS_NOSPC) 1906 return (ret); 1907 } 1908 } 1909 1910 return (EZFS_POOL_INVALARG); 1911 } 1912 1913 /* 1914 * Get phys_path for a root pool config. 1915 * Return 0 on success; non-zero on failure. 1916 */ 1917 static int 1918 zpool_get_config_physpath(nvlist_t *config, char *physpath, size_t phypath_size) 1919 { 1920 size_t rsz; 1921 nvlist_t *vdev_root; 1922 nvlist_t **child; 1923 uint_t count; 1924 char *type; 1925 1926 rsz = 0; 1927 1928 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 1929 &vdev_root) != 0) 1930 return (EZFS_INVALCONFIG); 1931 1932 if (nvlist_lookup_string(vdev_root, ZPOOL_CONFIG_TYPE, &type) != 0 || 1933 nvlist_lookup_nvlist_array(vdev_root, ZPOOL_CONFIG_CHILDREN, 1934 &child, &count) != 0) 1935 return (EZFS_INVALCONFIG); 1936 1937 /* 1938 * root pool can not have EFI labeled disks and can only have 1939 * a single top-level vdev. 1940 */ 1941 if (strcmp(type, VDEV_TYPE_ROOT) != 0 || count != 1 || 1942 pool_uses_efi(vdev_root)) 1943 return (EZFS_POOL_INVALARG); 1944 1945 (void) vdev_get_physpaths(child[0], physpath, phypath_size, &rsz, 1946 B_FALSE); 1947 1948 /* No online devices */ 1949 if (rsz == 0) 1950 return (EZFS_NODEVICE); 1951 1952 return (0); 1953 } 1954 1955 /* 1956 * Get phys_path for a root pool 1957 * Return 0 on success; non-zero on failure. 1958 */ 1959 int 1960 zpool_get_physpath(zpool_handle_t *zhp, char *physpath, size_t phypath_size) 1961 { 1962 return (zpool_get_config_physpath(zhp->zpool_config, physpath, 1963 phypath_size)); 1964 } 1965 1966 /* 1967 * If the device has being dynamically expanded then we need to relabel 1968 * the disk to use the new unallocated space. 1969 */ 1970 static int 1971 zpool_relabel_disk(libzfs_handle_t *hdl, const char *name) 1972 { 1973 char path[MAXPATHLEN]; 1974 char errbuf[1024]; 1975 int fd, error; 1976 int (*_efi_use_whole_disk)(int); 1977 1978 if ((_efi_use_whole_disk = (int (*)(int))dlsym(RTLD_DEFAULT, 1979 "efi_use_whole_disk")) == NULL) 1980 return (-1); 1981 1982 (void) snprintf(path, sizeof (path), "%s/%s", RDISK_ROOT, name); 1983 1984 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) { 1985 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot " 1986 "relabel '%s': unable to open device"), name); 1987 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf)); 1988 } 1989 1990 /* 1991 * It's possible that we might encounter an error if the device 1992 * does not have any unallocated space left. If so, we simply 1993 * ignore that error and continue on. 1994 */ 1995 error = _efi_use_whole_disk(fd); 1996 (void) close(fd); 1997 if (error && error != VT_ENOSPC) { 1998 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot " 1999 "relabel '%s': unable to read disk capacity"), name); 2000 return (zfs_error(hdl, EZFS_NOCAP, errbuf)); 2001 } 2002 return (0); 2003 } 2004 2005 /* 2006 * Bring the specified vdev online. The 'flags' parameter is a set of the 2007 * ZFS_ONLINE_* flags. 2008 */ 2009 int 2010 zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags, 2011 vdev_state_t *newstate) 2012 { 2013 zfs_cmd_t zc = { 0 }; 2014 char msg[1024]; 2015 nvlist_t *tgt; 2016 boolean_t avail_spare, l2cache, islog; 2017 libzfs_handle_t *hdl = zhp->zpool_hdl; 2018 2019 if (flags & ZFS_ONLINE_EXPAND) { 2020 (void) snprintf(msg, sizeof (msg), 2021 dgettext(TEXT_DOMAIN, "cannot expand %s"), path); 2022 } else { 2023 (void) snprintf(msg, sizeof (msg), 2024 dgettext(TEXT_DOMAIN, "cannot online %s"), path); 2025 } 2026 2027 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2028 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 2029 &islog)) == NULL) 2030 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 2031 2032 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 2033 2034 if (avail_spare) 2035 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 2036 2037 if (flags & ZFS_ONLINE_EXPAND || 2038 zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) { 2039 char *pathname = NULL; 2040 uint64_t wholedisk = 0; 2041 2042 (void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK, 2043 &wholedisk); 2044 verify(nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH, 2045 &pathname) == 0); 2046 2047 /* 2048 * XXX - L2ARC 1.0 devices can't support expansion. 2049 */ 2050 if (l2cache) { 2051 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2052 "cannot expand cache devices")); 2053 return (zfs_error(hdl, EZFS_VDEVNOTSUP, msg)); 2054 } 2055 2056 if (wholedisk) { 2057 pathname += strlen(DISK_ROOT) + 1; 2058 (void) zpool_relabel_disk(zhp->zpool_hdl, pathname); 2059 } 2060 } 2061 2062 zc.zc_cookie = VDEV_STATE_ONLINE; 2063 zc.zc_obj = flags; 2064 2065 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0) { 2066 if (errno == EINVAL) { 2067 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "was split " 2068 "from this pool into a new one. Use '%s' " 2069 "instead"), "zpool detach"); 2070 return (zfs_error(hdl, EZFS_POSTSPLIT_ONLINE, msg)); 2071 } 2072 return (zpool_standard_error(hdl, errno, msg)); 2073 } 2074 2075 *newstate = zc.zc_cookie; 2076 return (0); 2077 } 2078 2079 /* 2080 * Take the specified vdev offline 2081 */ 2082 int 2083 zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp) 2084 { 2085 zfs_cmd_t zc = { 0 }; 2086 char msg[1024]; 2087 nvlist_t *tgt; 2088 boolean_t avail_spare, l2cache; 2089 libzfs_handle_t *hdl = zhp->zpool_hdl; 2090 2091 (void) snprintf(msg, sizeof (msg), 2092 dgettext(TEXT_DOMAIN, "cannot offline %s"), path); 2093 2094 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2095 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 2096 NULL)) == NULL) 2097 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 2098 2099 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 2100 2101 if (avail_spare) 2102 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 2103 2104 zc.zc_cookie = VDEV_STATE_OFFLINE; 2105 zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0; 2106 2107 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0) 2108 return (0); 2109 2110 switch (errno) { 2111 case EBUSY: 2112 2113 /* 2114 * There are no other replicas of this device. 2115 */ 2116 return (zfs_error(hdl, EZFS_NOREPLICAS, msg)); 2117 2118 case EEXIST: 2119 /* 2120 * The log device has unplayed logs 2121 */ 2122 return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, msg)); 2123 2124 default: 2125 return (zpool_standard_error(hdl, errno, msg)); 2126 } 2127 } 2128 2129 /* 2130 * Mark the given vdev faulted. 2131 */ 2132 int 2133 zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux) 2134 { 2135 zfs_cmd_t zc = { 0 }; 2136 char msg[1024]; 2137 libzfs_handle_t *hdl = zhp->zpool_hdl; 2138 2139 (void) snprintf(msg, sizeof (msg), 2140 dgettext(TEXT_DOMAIN, "cannot fault %llu"), guid); 2141 2142 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2143 zc.zc_guid = guid; 2144 zc.zc_cookie = VDEV_STATE_FAULTED; 2145 zc.zc_obj = aux; 2146 2147 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0) 2148 return (0); 2149 2150 switch (errno) { 2151 case EBUSY: 2152 2153 /* 2154 * There are no other replicas of this device. 2155 */ 2156 return (zfs_error(hdl, EZFS_NOREPLICAS, msg)); 2157 2158 default: 2159 return (zpool_standard_error(hdl, errno, msg)); 2160 } 2161 2162 } 2163 2164 /* 2165 * Mark the given vdev degraded. 2166 */ 2167 int 2168 zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux) 2169 { 2170 zfs_cmd_t zc = { 0 }; 2171 char msg[1024]; 2172 libzfs_handle_t *hdl = zhp->zpool_hdl; 2173 2174 (void) snprintf(msg, sizeof (msg), 2175 dgettext(TEXT_DOMAIN, "cannot degrade %llu"), guid); 2176 2177 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2178 zc.zc_guid = guid; 2179 zc.zc_cookie = VDEV_STATE_DEGRADED; 2180 zc.zc_obj = aux; 2181 2182 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0) 2183 return (0); 2184 2185 return (zpool_standard_error(hdl, errno, msg)); 2186 } 2187 2188 /* 2189 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as 2190 * a hot spare. 2191 */ 2192 static boolean_t 2193 is_replacing_spare(nvlist_t *search, nvlist_t *tgt, int which) 2194 { 2195 nvlist_t **child; 2196 uint_t c, children; 2197 char *type; 2198 2199 if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child, 2200 &children) == 0) { 2201 verify(nvlist_lookup_string(search, ZPOOL_CONFIG_TYPE, 2202 &type) == 0); 2203 2204 if (strcmp(type, VDEV_TYPE_SPARE) == 0 && 2205 children == 2 && child[which] == tgt) 2206 return (B_TRUE); 2207 2208 for (c = 0; c < children; c++) 2209 if (is_replacing_spare(child[c], tgt, which)) 2210 return (B_TRUE); 2211 } 2212 2213 return (B_FALSE); 2214 } 2215 2216 /* 2217 * Attach new_disk (fully described by nvroot) to old_disk. 2218 * If 'replacing' is specified, the new disk will replace the old one. 2219 */ 2220 int 2221 zpool_vdev_attach(zpool_handle_t *zhp, 2222 const char *old_disk, const char *new_disk, nvlist_t *nvroot, int replacing) 2223 { 2224 zfs_cmd_t zc = { 0 }; 2225 char msg[1024]; 2226 int ret; 2227 nvlist_t *tgt; 2228 boolean_t avail_spare, l2cache, islog; 2229 uint64_t val; 2230 char *path, *newname; 2231 nvlist_t **child; 2232 uint_t children; 2233 nvlist_t *config_root; 2234 libzfs_handle_t *hdl = zhp->zpool_hdl; 2235 boolean_t rootpool = pool_is_bootable(zhp); 2236 2237 if (replacing) 2238 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 2239 "cannot replace %s with %s"), old_disk, new_disk); 2240 else 2241 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 2242 "cannot attach %s to %s"), new_disk, old_disk); 2243 2244 /* 2245 * If this is a root pool, make sure that we're not attaching an 2246 * EFI labeled device. 2247 */ 2248 if (rootpool && pool_uses_efi(nvroot)) { 2249 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2250 "EFI labeled devices are not supported on root pools.")); 2251 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg)); 2252 } 2253 2254 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2255 if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache, 2256 &islog)) == 0) 2257 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 2258 2259 if (avail_spare) 2260 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 2261 2262 if (l2cache) 2263 return (zfs_error(hdl, EZFS_ISL2CACHE, msg)); 2264 2265 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 2266 zc.zc_cookie = replacing; 2267 2268 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, 2269 &child, &children) != 0 || children != 1) { 2270 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2271 "new device must be a single disk")); 2272 return (zfs_error(hdl, EZFS_INVALCONFIG, msg)); 2273 } 2274 2275 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL), 2276 ZPOOL_CONFIG_VDEV_TREE, &config_root) == 0); 2277 2278 if ((newname = zpool_vdev_name(NULL, NULL, child[0], B_FALSE)) == NULL) 2279 return (-1); 2280 2281 /* 2282 * If the target is a hot spare that has been swapped in, we can only 2283 * replace it with another hot spare. 2284 */ 2285 if (replacing && 2286 nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 && 2287 (zpool_find_vdev(zhp, newname, &avail_spare, &l2cache, 2288 NULL) == NULL || !avail_spare) && 2289 is_replacing_spare(config_root, tgt, 1)) { 2290 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2291 "can only be replaced by another hot spare")); 2292 free(newname); 2293 return (zfs_error(hdl, EZFS_BADTARGET, msg)); 2294 } 2295 2296 /* 2297 * If we are attempting to replace a spare, it canot be applied to an 2298 * already spared device. 2299 */ 2300 if (replacing && 2301 nvlist_lookup_string(child[0], ZPOOL_CONFIG_PATH, &path) == 0 && 2302 zpool_find_vdev(zhp, newname, &avail_spare, 2303 &l2cache, NULL) != NULL && avail_spare && 2304 is_replacing_spare(config_root, tgt, 0)) { 2305 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2306 "device has already been replaced with a spare")); 2307 free(newname); 2308 return (zfs_error(hdl, EZFS_BADTARGET, msg)); 2309 } 2310 2311 free(newname); 2312 2313 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0) 2314 return (-1); 2315 2316 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_VDEV_ATTACH, &zc); 2317 2318 zcmd_free_nvlists(&zc); 2319 2320 if (ret == 0) { 2321 if (rootpool) { 2322 /* 2323 * XXX - This should be removed once we can 2324 * automatically install the bootblocks on the 2325 * newly attached disk. 2326 */ 2327 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Please " 2328 "be sure to invoke %s to make '%s' bootable.\n"), 2329 BOOTCMD, new_disk); 2330 2331 /* 2332 * XXX need a better way to prevent user from 2333 * booting up a half-baked vdev. 2334 */ 2335 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Make " 2336 "sure to wait until resilver is done " 2337 "before rebooting.\n")); 2338 } 2339 return (0); 2340 } 2341 2342 switch (errno) { 2343 case ENOTSUP: 2344 /* 2345 * Can't attach to or replace this type of vdev. 2346 */ 2347 if (replacing) { 2348 if (islog) 2349 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2350 "cannot replace a log with a spare")); 2351 else 2352 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2353 "cannot replace a replacing device")); 2354 } else { 2355 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2356 "can only attach to mirrors and top-level " 2357 "disks")); 2358 } 2359 (void) zfs_error(hdl, EZFS_BADTARGET, msg); 2360 break; 2361 2362 case EINVAL: 2363 /* 2364 * The new device must be a single disk. 2365 */ 2366 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2367 "new device must be a single disk")); 2368 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg); 2369 break; 2370 2371 case EBUSY: 2372 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy"), 2373 new_disk); 2374 (void) zfs_error(hdl, EZFS_BADDEV, msg); 2375 break; 2376 2377 case EOVERFLOW: 2378 /* 2379 * The new device is too small. 2380 */ 2381 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2382 "device is too small")); 2383 (void) zfs_error(hdl, EZFS_BADDEV, msg); 2384 break; 2385 2386 case EDOM: 2387 /* 2388 * The new device has a different alignment requirement. 2389 */ 2390 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2391 "devices have different sector alignment")); 2392 (void) zfs_error(hdl, EZFS_BADDEV, msg); 2393 break; 2394 2395 case ENAMETOOLONG: 2396 /* 2397 * The resulting top-level vdev spec won't fit in the label. 2398 */ 2399 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, msg); 2400 break; 2401 2402 default: 2403 (void) zpool_standard_error(hdl, errno, msg); 2404 } 2405 2406 return (-1); 2407 } 2408 2409 /* 2410 * Detach the specified device. 2411 */ 2412 int 2413 zpool_vdev_detach(zpool_handle_t *zhp, const char *path) 2414 { 2415 zfs_cmd_t zc = { 0 }; 2416 char msg[1024]; 2417 nvlist_t *tgt; 2418 boolean_t avail_spare, l2cache; 2419 libzfs_handle_t *hdl = zhp->zpool_hdl; 2420 2421 (void) snprintf(msg, sizeof (msg), 2422 dgettext(TEXT_DOMAIN, "cannot detach %s"), path); 2423 2424 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2425 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 2426 NULL)) == 0) 2427 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 2428 2429 if (avail_spare) 2430 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 2431 2432 if (l2cache) 2433 return (zfs_error(hdl, EZFS_ISL2CACHE, msg)); 2434 2435 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 2436 2437 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0) 2438 return (0); 2439 2440 switch (errno) { 2441 2442 case ENOTSUP: 2443 /* 2444 * Can't detach from this type of vdev. 2445 */ 2446 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only " 2447 "applicable to mirror and replacing vdevs")); 2448 (void) zfs_error(zhp->zpool_hdl, EZFS_BADTARGET, msg); 2449 break; 2450 2451 case EBUSY: 2452 /* 2453 * There are no other replicas of this device. 2454 */ 2455 (void) zfs_error(hdl, EZFS_NOREPLICAS, msg); 2456 break; 2457 2458 default: 2459 (void) zpool_standard_error(hdl, errno, msg); 2460 } 2461 2462 return (-1); 2463 } 2464 2465 /* 2466 * Find a mirror vdev in the source nvlist. 2467 * 2468 * The mchild array contains a list of disks in one of the top-level mirrors 2469 * of the source pool. The schild array contains a list of disks that the 2470 * user specified on the command line. We loop over the mchild array to 2471 * see if any entry in the schild array matches. 2472 * 2473 * If a disk in the mchild array is found in the schild array, we return 2474 * the index of that entry. Otherwise we return -1. 2475 */ 2476 static int 2477 find_vdev_entry(zpool_handle_t *zhp, nvlist_t **mchild, uint_t mchildren, 2478 nvlist_t **schild, uint_t schildren) 2479 { 2480 uint_t mc; 2481 2482 for (mc = 0; mc < mchildren; mc++) { 2483 uint_t sc; 2484 char *mpath = zpool_vdev_name(zhp->zpool_hdl, zhp, 2485 mchild[mc], B_FALSE); 2486 2487 for (sc = 0; sc < schildren; sc++) { 2488 char *spath = zpool_vdev_name(zhp->zpool_hdl, zhp, 2489 schild[sc], B_FALSE); 2490 boolean_t result = (strcmp(mpath, spath) == 0); 2491 2492 free(spath); 2493 if (result) { 2494 free(mpath); 2495 return (mc); 2496 } 2497 } 2498 2499 free(mpath); 2500 } 2501 2502 return (-1); 2503 } 2504 2505 /* 2506 * Split a mirror pool. If newroot points to null, then a new nvlist 2507 * is generated and it is the responsibility of the caller to free it. 2508 */ 2509 int 2510 zpool_vdev_split(zpool_handle_t *zhp, char *newname, nvlist_t **newroot, 2511 nvlist_t *props, splitflags_t flags) 2512 { 2513 zfs_cmd_t zc = { 0 }; 2514 char msg[1024]; 2515 nvlist_t *tree, *config, **child, **newchild, *newconfig = NULL; 2516 nvlist_t **varray = NULL, *zc_props = NULL; 2517 uint_t c, children, newchildren, lastlog = 0, vcount, found = 0; 2518 libzfs_handle_t *hdl = zhp->zpool_hdl; 2519 uint64_t vers; 2520 boolean_t freelist = B_FALSE, memory_err = B_TRUE; 2521 int retval = 0; 2522 2523 (void) snprintf(msg, sizeof (msg), 2524 dgettext(TEXT_DOMAIN, "Unable to split %s"), zhp->zpool_name); 2525 2526 if (!zpool_name_valid(hdl, B_FALSE, newname)) 2527 return (zfs_error(hdl, EZFS_INVALIDNAME, msg)); 2528 2529 if ((config = zpool_get_config(zhp, NULL)) == NULL) { 2530 (void) fprintf(stderr, gettext("Internal error: unable to " 2531 "retrieve pool configuration\n")); 2532 return (-1); 2533 } 2534 2535 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree) 2536 == 0); 2537 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &vers) == 0); 2538 2539 if (props) { 2540 if ((zc_props = zpool_valid_proplist(hdl, zhp->zpool_name, 2541 props, vers, B_TRUE, msg)) == NULL) 2542 return (-1); 2543 } 2544 2545 if (nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child, 2546 &children) != 0) { 2547 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2548 "Source pool is missing vdev tree")); 2549 if (zc_props) 2550 nvlist_free(zc_props); 2551 return (-1); 2552 } 2553 2554 varray = zfs_alloc(hdl, children * sizeof (nvlist_t *)); 2555 vcount = 0; 2556 2557 if (*newroot == NULL || 2558 nvlist_lookup_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, 2559 &newchild, &newchildren) != 0) 2560 newchildren = 0; 2561 2562 for (c = 0; c < children; c++) { 2563 uint64_t is_log = B_FALSE, is_hole = B_FALSE; 2564 char *type; 2565 nvlist_t **mchild, *vdev; 2566 uint_t mchildren; 2567 int entry; 2568 2569 /* 2570 * Unlike cache & spares, slogs are stored in the 2571 * ZPOOL_CONFIG_CHILDREN array. We filter them out here. 2572 */ 2573 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG, 2574 &is_log); 2575 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE, 2576 &is_hole); 2577 if (is_log || is_hole) { 2578 /* 2579 * Create a hole vdev and put it in the config. 2580 */ 2581 if (nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) != 0) 2582 goto out; 2583 if (nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE, 2584 VDEV_TYPE_HOLE) != 0) 2585 goto out; 2586 if (nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_HOLE, 2587 1) != 0) 2588 goto out; 2589 if (lastlog == 0) 2590 lastlog = vcount; 2591 varray[vcount++] = vdev; 2592 continue; 2593 } 2594 lastlog = 0; 2595 verify(nvlist_lookup_string(child[c], ZPOOL_CONFIG_TYPE, &type) 2596 == 0); 2597 if (strcmp(type, VDEV_TYPE_MIRROR) != 0) { 2598 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2599 "Source pool must be composed only of mirrors\n")); 2600 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg); 2601 goto out; 2602 } 2603 2604 verify(nvlist_lookup_nvlist_array(child[c], 2605 ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0); 2606 2607 /* find or add an entry for this top-level vdev */ 2608 if (newchildren > 0 && 2609 (entry = find_vdev_entry(zhp, mchild, mchildren, 2610 newchild, newchildren)) >= 0) { 2611 /* We found a disk that the user specified. */ 2612 vdev = mchild[entry]; 2613 ++found; 2614 } else { 2615 /* User didn't specify a disk for this vdev. */ 2616 vdev = mchild[mchildren - 1]; 2617 } 2618 2619 if (nvlist_dup(vdev, &varray[vcount++], 0) != 0) 2620 goto out; 2621 } 2622 2623 /* did we find every disk the user specified? */ 2624 if (found != newchildren) { 2625 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "Device list must " 2626 "include at most one disk from each mirror")); 2627 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg); 2628 goto out; 2629 } 2630 2631 /* Prepare the nvlist for populating. */ 2632 if (*newroot == NULL) { 2633 if (nvlist_alloc(newroot, NV_UNIQUE_NAME, 0) != 0) 2634 goto out; 2635 freelist = B_TRUE; 2636 if (nvlist_add_string(*newroot, ZPOOL_CONFIG_TYPE, 2637 VDEV_TYPE_ROOT) != 0) 2638 goto out; 2639 } else { 2640 verify(nvlist_remove_all(*newroot, ZPOOL_CONFIG_CHILDREN) == 0); 2641 } 2642 2643 /* Add all the children we found */ 2644 if (nvlist_add_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, varray, 2645 lastlog == 0 ? vcount : lastlog) != 0) 2646 goto out; 2647 2648 /* 2649 * If we're just doing a dry run, exit now with success. 2650 */ 2651 if (flags.dryrun) { 2652 memory_err = B_FALSE; 2653 freelist = B_FALSE; 2654 goto out; 2655 } 2656 2657 /* now build up the config list & call the ioctl */ 2658 if (nvlist_alloc(&newconfig, NV_UNIQUE_NAME, 0) != 0) 2659 goto out; 2660 2661 if (nvlist_add_nvlist(newconfig, 2662 ZPOOL_CONFIG_VDEV_TREE, *newroot) != 0 || 2663 nvlist_add_string(newconfig, 2664 ZPOOL_CONFIG_POOL_NAME, newname) != 0 || 2665 nvlist_add_uint64(newconfig, ZPOOL_CONFIG_VERSION, vers) != 0) 2666 goto out; 2667 2668 /* 2669 * The new pool is automatically part of the namespace unless we 2670 * explicitly export it. 2671 */ 2672 if (!flags.import) 2673 zc.zc_cookie = ZPOOL_EXPORT_AFTER_SPLIT; 2674 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2675 (void) strlcpy(zc.zc_string, newname, sizeof (zc.zc_string)); 2676 if (zcmd_write_conf_nvlist(hdl, &zc, newconfig) != 0) 2677 goto out; 2678 if (zc_props != NULL && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0) 2679 goto out; 2680 2681 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SPLIT, &zc) != 0) { 2682 retval = zpool_standard_error(hdl, errno, msg); 2683 goto out; 2684 } 2685 2686 freelist = B_FALSE; 2687 memory_err = B_FALSE; 2688 2689 out: 2690 if (varray != NULL) { 2691 int v; 2692 2693 for (v = 0; v < vcount; v++) 2694 nvlist_free(varray[v]); 2695 free(varray); 2696 } 2697 zcmd_free_nvlists(&zc); 2698 if (zc_props) 2699 nvlist_free(zc_props); 2700 if (newconfig) 2701 nvlist_free(newconfig); 2702 if (freelist) { 2703 nvlist_free(*newroot); 2704 *newroot = NULL; 2705 } 2706 2707 if (retval != 0) 2708 return (retval); 2709 2710 if (memory_err) 2711 return (no_memory(hdl)); 2712 2713 return (0); 2714 } 2715 2716 /* 2717 * Remove the given device. Currently, this is supported only for hot spares 2718 * and level 2 cache devices. 2719 */ 2720 int 2721 zpool_vdev_remove(zpool_handle_t *zhp, const char *path) 2722 { 2723 zfs_cmd_t zc = { 0 }; 2724 char msg[1024]; 2725 nvlist_t *tgt; 2726 boolean_t avail_spare, l2cache, islog; 2727 libzfs_handle_t *hdl = zhp->zpool_hdl; 2728 uint64_t version; 2729 2730 (void) snprintf(msg, sizeof (msg), 2731 dgettext(TEXT_DOMAIN, "cannot remove %s"), path); 2732 2733 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2734 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 2735 &islog)) == 0) 2736 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 2737 /* 2738 * XXX - this should just go away. 2739 */ 2740 if (!avail_spare && !l2cache && !islog) { 2741 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2742 "only inactive hot spares, cache, top-level, " 2743 "or log devices can be removed")); 2744 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 2745 } 2746 2747 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL); 2748 if (islog && version < SPA_VERSION_HOLES) { 2749 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2750 "pool must be upgrade to support log removal")); 2751 return (zfs_error(hdl, EZFS_BADVERSION, msg)); 2752 } 2753 2754 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 2755 2756 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0) 2757 return (0); 2758 2759 return (zpool_standard_error(hdl, errno, msg)); 2760 } 2761 2762 /* 2763 * Clear the errors for the pool, or the particular device if specified. 2764 */ 2765 int 2766 zpool_clear(zpool_handle_t *zhp, const char *path, nvlist_t *rewindnvl) 2767 { 2768 zfs_cmd_t zc = { 0 }; 2769 char msg[1024]; 2770 nvlist_t *tgt; 2771 zpool_rewind_policy_t policy; 2772 boolean_t avail_spare, l2cache; 2773 libzfs_handle_t *hdl = zhp->zpool_hdl; 2774 nvlist_t *nvi = NULL; 2775 2776 if (path) 2777 (void) snprintf(msg, sizeof (msg), 2778 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"), 2779 path); 2780 else 2781 (void) snprintf(msg, sizeof (msg), 2782 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"), 2783 zhp->zpool_name); 2784 2785 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2786 if (path) { 2787 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, 2788 &l2cache, NULL)) == 0) 2789 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 2790 2791 /* 2792 * Don't allow error clearing for hot spares. Do allow 2793 * error clearing for l2cache devices. 2794 */ 2795 if (avail_spare) 2796 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 2797 2798 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, 2799 &zc.zc_guid) == 0); 2800 } 2801 2802 zpool_get_rewind_policy(rewindnvl, &policy); 2803 zc.zc_cookie = policy.zrp_request; 2804 2805 if (zcmd_alloc_dst_nvlist(hdl, &zc, 8192) != 0) 2806 return (-1); 2807 2808 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, rewindnvl) != 0) 2809 return (-1); 2810 2811 if (zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc) == 0 || 2812 ((policy.zrp_request & ZPOOL_TRY_REWIND) && 2813 errno != EPERM && errno != EACCES)) { 2814 if (policy.zrp_request & 2815 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) { 2816 (void) zcmd_read_dst_nvlist(hdl, &zc, &nvi); 2817 zpool_rewind_exclaim(hdl, zc.zc_name, 2818 ((policy.zrp_request & ZPOOL_TRY_REWIND) != 0), 2819 nvi); 2820 nvlist_free(nvi); 2821 } 2822 zcmd_free_nvlists(&zc); 2823 return (0); 2824 } 2825 2826 zcmd_free_nvlists(&zc); 2827 return (zpool_standard_error(hdl, errno, msg)); 2828 } 2829 2830 /* 2831 * Similar to zpool_clear(), but takes a GUID (used by fmd). 2832 */ 2833 int 2834 zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid) 2835 { 2836 zfs_cmd_t zc = { 0 }; 2837 char msg[1024]; 2838 libzfs_handle_t *hdl = zhp->zpool_hdl; 2839 2840 (void) snprintf(msg, sizeof (msg), 2841 dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"), 2842 guid); 2843 2844 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2845 zc.zc_guid = guid; 2846 2847 if (ioctl(hdl->libzfs_fd, ZFS_IOC_CLEAR, &zc) == 0) 2848 return (0); 2849 2850 return (zpool_standard_error(hdl, errno, msg)); 2851 } 2852 2853 /* 2854 * Convert from a devid string to a path. 2855 */ 2856 static char * 2857 devid_to_path(char *devid_str) 2858 { 2859 ddi_devid_t devid; 2860 char *minor; 2861 char *path; 2862 devid_nmlist_t *list = NULL; 2863 int ret; 2864 2865 if (devid_str_decode(devid_str, &devid, &minor) != 0) 2866 return (NULL); 2867 2868 ret = devid_deviceid_to_nmlist("/dev", devid, minor, &list); 2869 2870 devid_str_free(minor); 2871 devid_free(devid); 2872 2873 if (ret != 0) 2874 return (NULL); 2875 2876 if ((path = strdup(list[0].devname)) == NULL) 2877 return (NULL); 2878 2879 devid_free_nmlist(list); 2880 2881 return (path); 2882 } 2883 2884 /* 2885 * Convert from a path to a devid string. 2886 */ 2887 static char * 2888 path_to_devid(const char *path) 2889 { 2890 int fd; 2891 ddi_devid_t devid; 2892 char *minor, *ret; 2893 2894 if ((fd = open(path, O_RDONLY)) < 0) 2895 return (NULL); 2896 2897 minor = NULL; 2898 ret = NULL; 2899 if (devid_get(fd, &devid) == 0) { 2900 if (devid_get_minor_name(fd, &minor) == 0) 2901 ret = devid_str_encode(devid, minor); 2902 if (minor != NULL) 2903 devid_str_free(minor); 2904 devid_free(devid); 2905 } 2906 (void) close(fd); 2907 2908 return (ret); 2909 } 2910 2911 /* 2912 * Issue the necessary ioctl() to update the stored path value for the vdev. We 2913 * ignore any failure here, since a common case is for an unprivileged user to 2914 * type 'zpool status', and we'll display the correct information anyway. 2915 */ 2916 static void 2917 set_path(zpool_handle_t *zhp, nvlist_t *nv, const char *path) 2918 { 2919 zfs_cmd_t zc = { 0 }; 2920 2921 (void) strncpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2922 (void) strncpy(zc.zc_value, path, sizeof (zc.zc_value)); 2923 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, 2924 &zc.zc_guid) == 0); 2925 2926 (void) ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SETPATH, &zc); 2927 } 2928 2929 /* 2930 * Given a vdev, return the name to display in iostat. If the vdev has a path, 2931 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type. 2932 * We also check if this is a whole disk, in which case we strip off the 2933 * trailing 's0' slice name. 2934 * 2935 * This routine is also responsible for identifying when disks have been 2936 * reconfigured in a new location. The kernel will have opened the device by 2937 * devid, but the path will still refer to the old location. To catch this, we 2938 * first do a path -> devid translation (which is fast for the common case). If 2939 * the devid matches, we're done. If not, we do a reverse devid -> path 2940 * translation and issue the appropriate ioctl() to update the path of the vdev. 2941 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any 2942 * of these checks. 2943 */ 2944 char * 2945 zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv, 2946 boolean_t verbose) 2947 { 2948 char *path, *devid; 2949 uint64_t value; 2950 char buf[64]; 2951 vdev_stat_t *vs; 2952 uint_t vsc; 2953 2954 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, 2955 &value) == 0) { 2956 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, 2957 &value) == 0); 2958 (void) snprintf(buf, sizeof (buf), "%llu", 2959 (u_longlong_t)value); 2960 path = buf; 2961 } else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) { 2962 2963 /* 2964 * If the device is dead (faulted, offline, etc) then don't 2965 * bother opening it. Otherwise we may be forcing the user to 2966 * open a misbehaving device, which can have undesirable 2967 * effects. 2968 */ 2969 if ((nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS, 2970 (uint64_t **)&vs, &vsc) != 0 || 2971 vs->vs_state >= VDEV_STATE_DEGRADED) && 2972 zhp != NULL && 2973 nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &devid) == 0) { 2974 /* 2975 * Determine if the current path is correct. 2976 */ 2977 char *newdevid = path_to_devid(path); 2978 2979 if (newdevid == NULL || 2980 strcmp(devid, newdevid) != 0) { 2981 char *newpath; 2982 2983 if ((newpath = devid_to_path(devid)) != NULL) { 2984 /* 2985 * Update the path appropriately. 2986 */ 2987 set_path(zhp, nv, newpath); 2988 if (nvlist_add_string(nv, 2989 ZPOOL_CONFIG_PATH, newpath) == 0) 2990 verify(nvlist_lookup_string(nv, 2991 ZPOOL_CONFIG_PATH, 2992 &path) == 0); 2993 free(newpath); 2994 } 2995 } 2996 2997 if (newdevid) 2998 devid_str_free(newdevid); 2999 } 3000 3001 if (strncmp(path, "/dev/dsk/", 9) == 0) 3002 path += 9; 3003 3004 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, 3005 &value) == 0 && value) { 3006 char *tmp = zfs_strdup(hdl, path); 3007 if (tmp == NULL) 3008 return (NULL); 3009 tmp[strlen(path) - 2] = '\0'; 3010 return (tmp); 3011 } 3012 } else { 3013 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &path) == 0); 3014 3015 /* 3016 * If it's a raidz device, we need to stick in the parity level. 3017 */ 3018 if (strcmp(path, VDEV_TYPE_RAIDZ) == 0) { 3019 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY, 3020 &value) == 0); 3021 (void) snprintf(buf, sizeof (buf), "%s%llu", path, 3022 (u_longlong_t)value); 3023 path = buf; 3024 } 3025 3026 /* 3027 * We identify each top-level vdev by using a <type-id> 3028 * naming convention. 3029 */ 3030 if (verbose) { 3031 uint64_t id; 3032 3033 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID, 3034 &id) == 0); 3035 (void) snprintf(buf, sizeof (buf), "%s-%llu", path, 3036 (u_longlong_t)id); 3037 path = buf; 3038 } 3039 } 3040 3041 return (zfs_strdup(hdl, path)); 3042 } 3043 3044 static int 3045 zbookmark_compare(const void *a, const void *b) 3046 { 3047 return (memcmp(a, b, sizeof (zbookmark_t))); 3048 } 3049 3050 /* 3051 * Retrieve the persistent error log, uniquify the members, and return to the 3052 * caller. 3053 */ 3054 int 3055 zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp) 3056 { 3057 zfs_cmd_t zc = { 0 }; 3058 uint64_t count; 3059 zbookmark_t *zb = NULL; 3060 int i; 3061 3062 /* 3063 * Retrieve the raw error list from the kernel. If the number of errors 3064 * has increased, allocate more space and continue until we get the 3065 * entire list. 3066 */ 3067 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_ERRCOUNT, 3068 &count) == 0); 3069 if (count == 0) 3070 return (0); 3071 if ((zc.zc_nvlist_dst = (uintptr_t)zfs_alloc(zhp->zpool_hdl, 3072 count * sizeof (zbookmark_t))) == (uintptr_t)NULL) 3073 return (-1); 3074 zc.zc_nvlist_dst_size = count; 3075 (void) strcpy(zc.zc_name, zhp->zpool_name); 3076 for (;;) { 3077 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_ERROR_LOG, 3078 &zc) != 0) { 3079 free((void *)(uintptr_t)zc.zc_nvlist_dst); 3080 if (errno == ENOMEM) { 3081 count = zc.zc_nvlist_dst_size; 3082 if ((zc.zc_nvlist_dst = (uintptr_t) 3083 zfs_alloc(zhp->zpool_hdl, count * 3084 sizeof (zbookmark_t))) == (uintptr_t)NULL) 3085 return (-1); 3086 } else { 3087 return (-1); 3088 } 3089 } else { 3090 break; 3091 } 3092 } 3093 3094 /* 3095 * Sort the resulting bookmarks. This is a little confusing due to the 3096 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last 3097 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks 3098 * _not_ copied as part of the process. So we point the start of our 3099 * array appropriate and decrement the total number of elements. 3100 */ 3101 zb = ((zbookmark_t *)(uintptr_t)zc.zc_nvlist_dst) + 3102 zc.zc_nvlist_dst_size; 3103 count -= zc.zc_nvlist_dst_size; 3104 3105 qsort(zb, count, sizeof (zbookmark_t), zbookmark_compare); 3106 3107 verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0); 3108 3109 /* 3110 * Fill in the nverrlistp with nvlist's of dataset and object numbers. 3111 */ 3112 for (i = 0; i < count; i++) { 3113 nvlist_t *nv; 3114 3115 /* ignoring zb_blkid and zb_level for now */ 3116 if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset && 3117 zb[i-1].zb_object == zb[i].zb_object) 3118 continue; 3119 3120 if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0) 3121 goto nomem; 3122 if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET, 3123 zb[i].zb_objset) != 0) { 3124 nvlist_free(nv); 3125 goto nomem; 3126 } 3127 if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT, 3128 zb[i].zb_object) != 0) { 3129 nvlist_free(nv); 3130 goto nomem; 3131 } 3132 if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) { 3133 nvlist_free(nv); 3134 goto nomem; 3135 } 3136 nvlist_free(nv); 3137 } 3138 3139 free((void *)(uintptr_t)zc.zc_nvlist_dst); 3140 return (0); 3141 3142 nomem: 3143 free((void *)(uintptr_t)zc.zc_nvlist_dst); 3144 return (no_memory(zhp->zpool_hdl)); 3145 } 3146 3147 /* 3148 * Upgrade a ZFS pool to the latest on-disk version. 3149 */ 3150 int 3151 zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version) 3152 { 3153 zfs_cmd_t zc = { 0 }; 3154 libzfs_handle_t *hdl = zhp->zpool_hdl; 3155 3156 (void) strcpy(zc.zc_name, zhp->zpool_name); 3157 zc.zc_cookie = new_version; 3158 3159 if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0) 3160 return (zpool_standard_error_fmt(hdl, errno, 3161 dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"), 3162 zhp->zpool_name)); 3163 return (0); 3164 } 3165 3166 void 3167 zpool_set_history_str(const char *subcommand, int argc, char **argv, 3168 char *history_str) 3169 { 3170 int i; 3171 3172 (void) strlcpy(history_str, subcommand, HIS_MAX_RECORD_LEN); 3173 for (i = 1; i < argc; i++) { 3174 if (strlen(history_str) + 1 + strlen(argv[i]) > 3175 HIS_MAX_RECORD_LEN) 3176 break; 3177 (void) strlcat(history_str, " ", HIS_MAX_RECORD_LEN); 3178 (void) strlcat(history_str, argv[i], HIS_MAX_RECORD_LEN); 3179 } 3180 } 3181 3182 /* 3183 * Stage command history for logging. 3184 */ 3185 int 3186 zpool_stage_history(libzfs_handle_t *hdl, const char *history_str) 3187 { 3188 if (history_str == NULL) 3189 return (EINVAL); 3190 3191 if (strlen(history_str) > HIS_MAX_RECORD_LEN) 3192 return (EINVAL); 3193 3194 if (hdl->libzfs_log_str != NULL) 3195 free(hdl->libzfs_log_str); 3196 3197 if ((hdl->libzfs_log_str = strdup(history_str)) == NULL) 3198 return (no_memory(hdl)); 3199 3200 return (0); 3201 } 3202 3203 /* 3204 * Perform ioctl to get some command history of a pool. 3205 * 3206 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the 3207 * logical offset of the history buffer to start reading from. 3208 * 3209 * Upon return, 'off' is the next logical offset to read from and 3210 * 'len' is the actual amount of bytes read into 'buf'. 3211 */ 3212 static int 3213 get_history(zpool_handle_t *zhp, char *buf, uint64_t *off, uint64_t *len) 3214 { 3215 zfs_cmd_t zc = { 0 }; 3216 libzfs_handle_t *hdl = zhp->zpool_hdl; 3217 3218 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3219 3220 zc.zc_history = (uint64_t)(uintptr_t)buf; 3221 zc.zc_history_len = *len; 3222 zc.zc_history_offset = *off; 3223 3224 if (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) { 3225 switch (errno) { 3226 case EPERM: 3227 return (zfs_error_fmt(hdl, EZFS_PERM, 3228 dgettext(TEXT_DOMAIN, 3229 "cannot show history for pool '%s'"), 3230 zhp->zpool_name)); 3231 case ENOENT: 3232 return (zfs_error_fmt(hdl, EZFS_NOHISTORY, 3233 dgettext(TEXT_DOMAIN, "cannot get history for pool " 3234 "'%s'"), zhp->zpool_name)); 3235 case ENOTSUP: 3236 return (zfs_error_fmt(hdl, EZFS_BADVERSION, 3237 dgettext(TEXT_DOMAIN, "cannot get history for pool " 3238 "'%s', pool must be upgraded"), zhp->zpool_name)); 3239 default: 3240 return (zpool_standard_error_fmt(hdl, errno, 3241 dgettext(TEXT_DOMAIN, 3242 "cannot get history for '%s'"), zhp->zpool_name)); 3243 } 3244 } 3245 3246 *len = zc.zc_history_len; 3247 *off = zc.zc_history_offset; 3248 3249 return (0); 3250 } 3251 3252 /* 3253 * Process the buffer of nvlists, unpacking and storing each nvlist record 3254 * into 'records'. 'leftover' is set to the number of bytes that weren't 3255 * processed as there wasn't a complete record. 3256 */ 3257 int 3258 zpool_history_unpack(char *buf, uint64_t bytes_read, uint64_t *leftover, 3259 nvlist_t ***records, uint_t *numrecords) 3260 { 3261 uint64_t reclen; 3262 nvlist_t *nv; 3263 int i; 3264 3265 while (bytes_read > sizeof (reclen)) { 3266 3267 /* get length of packed record (stored as little endian) */ 3268 for (i = 0, reclen = 0; i < sizeof (reclen); i++) 3269 reclen += (uint64_t)(((uchar_t *)buf)[i]) << (8*i); 3270 3271 if (bytes_read < sizeof (reclen) + reclen) 3272 break; 3273 3274 /* unpack record */ 3275 if (nvlist_unpack(buf + sizeof (reclen), reclen, &nv, 0) != 0) 3276 return (ENOMEM); 3277 bytes_read -= sizeof (reclen) + reclen; 3278 buf += sizeof (reclen) + reclen; 3279 3280 /* add record to nvlist array */ 3281 (*numrecords)++; 3282 if (ISP2(*numrecords + 1)) { 3283 *records = realloc(*records, 3284 *numrecords * 2 * sizeof (nvlist_t *)); 3285 } 3286 (*records)[*numrecords - 1] = nv; 3287 } 3288 3289 *leftover = bytes_read; 3290 return (0); 3291 } 3292 3293 #define HIS_BUF_LEN (128*1024) 3294 3295 /* 3296 * Retrieve the command history of a pool. 3297 */ 3298 int 3299 zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp) 3300 { 3301 char buf[HIS_BUF_LEN]; 3302 uint64_t off = 0; 3303 nvlist_t **records = NULL; 3304 uint_t numrecords = 0; 3305 int err, i; 3306 3307 do { 3308 uint64_t bytes_read = sizeof (buf); 3309 uint64_t leftover; 3310 3311 if ((err = get_history(zhp, buf, &off, &bytes_read)) != 0) 3312 break; 3313 3314 /* if nothing else was read in, we're at EOF, just return */ 3315 if (!bytes_read) 3316 break; 3317 3318 if ((err = zpool_history_unpack(buf, bytes_read, 3319 &leftover, &records, &numrecords)) != 0) 3320 break; 3321 off -= leftover; 3322 3323 /* CONSTCOND */ 3324 } while (1); 3325 3326 if (!err) { 3327 verify(nvlist_alloc(nvhisp, NV_UNIQUE_NAME, 0) == 0); 3328 verify(nvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD, 3329 records, numrecords) == 0); 3330 } 3331 for (i = 0; i < numrecords; i++) 3332 nvlist_free(records[i]); 3333 free(records); 3334 3335 return (err); 3336 } 3337 3338 void 3339 zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj, 3340 char *pathname, size_t len) 3341 { 3342 zfs_cmd_t zc = { 0 }; 3343 boolean_t mounted = B_FALSE; 3344 char *mntpnt = NULL; 3345 char dsname[MAXNAMELEN]; 3346 3347 if (dsobj == 0) { 3348 /* special case for the MOS */ 3349 (void) snprintf(pathname, len, "<metadata>:<0x%llx>", obj); 3350 return; 3351 } 3352 3353 /* get the dataset's name */ 3354 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3355 zc.zc_obj = dsobj; 3356 if (ioctl(zhp->zpool_hdl->libzfs_fd, 3357 ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) { 3358 /* just write out a path of two object numbers */ 3359 (void) snprintf(pathname, len, "<0x%llx>:<0x%llx>", 3360 dsobj, obj); 3361 return; 3362 } 3363 (void) strlcpy(dsname, zc.zc_value, sizeof (dsname)); 3364 3365 /* find out if the dataset is mounted */ 3366 mounted = is_mounted(zhp->zpool_hdl, dsname, &mntpnt); 3367 3368 /* get the corrupted object's path */ 3369 (void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name)); 3370 zc.zc_obj = obj; 3371 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_OBJ_TO_PATH, 3372 &zc) == 0) { 3373 if (mounted) { 3374 (void) snprintf(pathname, len, "%s%s", mntpnt, 3375 zc.zc_value); 3376 } else { 3377 (void) snprintf(pathname, len, "%s:%s", 3378 dsname, zc.zc_value); 3379 } 3380 } else { 3381 (void) snprintf(pathname, len, "%s:<0x%llx>", dsname, obj); 3382 } 3383 free(mntpnt); 3384 } 3385 3386 /* 3387 * Read the EFI label from the config, if a label does not exist then 3388 * pass back the error to the caller. If the caller has passed a non-NULL 3389 * diskaddr argument then we set it to the starting address of the EFI 3390 * partition. 3391 */ 3392 static int 3393 read_efi_label(nvlist_t *config, diskaddr_t *sb) 3394 { 3395 char *path; 3396 int fd; 3397 char diskname[MAXPATHLEN]; 3398 int err = -1; 3399 3400 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PATH, &path) != 0) 3401 return (err); 3402 3403 (void) snprintf(diskname, sizeof (diskname), "%s%s", RDISK_ROOT, 3404 strrchr(path, '/')); 3405 if ((fd = open(diskname, O_RDONLY|O_NDELAY)) >= 0) { 3406 struct dk_gpt *vtoc; 3407 3408 if ((err = efi_alloc_and_read(fd, &vtoc)) >= 0) { 3409 if (sb != NULL) 3410 *sb = vtoc->efi_parts[0].p_start; 3411 efi_free(vtoc); 3412 } 3413 (void) close(fd); 3414 } 3415 return (err); 3416 } 3417 3418 /* 3419 * determine where a partition starts on a disk in the current 3420 * configuration 3421 */ 3422 static diskaddr_t 3423 find_start_block(nvlist_t *config) 3424 { 3425 nvlist_t **child; 3426 uint_t c, children; 3427 diskaddr_t sb = MAXOFFSET_T; 3428 uint64_t wholedisk; 3429 3430 if (nvlist_lookup_nvlist_array(config, 3431 ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) { 3432 if (nvlist_lookup_uint64(config, 3433 ZPOOL_CONFIG_WHOLE_DISK, 3434 &wholedisk) != 0 || !wholedisk) { 3435 return (MAXOFFSET_T); 3436 } 3437 if (read_efi_label(config, &sb) < 0) 3438 sb = MAXOFFSET_T; 3439 return (sb); 3440 } 3441 3442 for (c = 0; c < children; c++) { 3443 sb = find_start_block(child[c]); 3444 if (sb != MAXOFFSET_T) { 3445 return (sb); 3446 } 3447 } 3448 return (MAXOFFSET_T); 3449 } 3450 3451 /* 3452 * Label an individual disk. The name provided is the short name, 3453 * stripped of any leading /dev path. 3454 */ 3455 int 3456 zpool_label_disk(libzfs_handle_t *hdl, zpool_handle_t *zhp, char *name) 3457 { 3458 char path[MAXPATHLEN]; 3459 struct dk_gpt *vtoc; 3460 int fd; 3461 size_t resv = EFI_MIN_RESV_SIZE; 3462 uint64_t slice_size; 3463 diskaddr_t start_block; 3464 char errbuf[1024]; 3465 3466 /* prepare an error message just in case */ 3467 (void) snprintf(errbuf, sizeof (errbuf), 3468 dgettext(TEXT_DOMAIN, "cannot label '%s'"), name); 3469 3470 if (zhp) { 3471 nvlist_t *nvroot; 3472 3473 if (pool_is_bootable(zhp)) { 3474 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3475 "EFI labeled devices are not supported on root " 3476 "pools.")); 3477 return (zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf)); 3478 } 3479 3480 verify(nvlist_lookup_nvlist(zhp->zpool_config, 3481 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 3482 3483 if (zhp->zpool_start_block == 0) 3484 start_block = find_start_block(nvroot); 3485 else 3486 start_block = zhp->zpool_start_block; 3487 zhp->zpool_start_block = start_block; 3488 } else { 3489 /* new pool */ 3490 start_block = NEW_START_BLOCK; 3491 } 3492 3493 (void) snprintf(path, sizeof (path), "%s/%s%s", RDISK_ROOT, name, 3494 BACKUP_SLICE); 3495 3496 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) { 3497 /* 3498 * This shouldn't happen. We've long since verified that this 3499 * is a valid device. 3500 */ 3501 zfs_error_aux(hdl, 3502 dgettext(TEXT_DOMAIN, "unable to open device")); 3503 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf)); 3504 } 3505 3506 if (efi_alloc_and_init(fd, EFI_NUMPAR, &vtoc) != 0) { 3507 /* 3508 * The only way this can fail is if we run out of memory, or we 3509 * were unable to read the disk's capacity 3510 */ 3511 if (errno == ENOMEM) 3512 (void) no_memory(hdl); 3513 3514 (void) close(fd); 3515 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3516 "unable to read disk capacity"), name); 3517 3518 return (zfs_error(hdl, EZFS_NOCAP, errbuf)); 3519 } 3520 3521 slice_size = vtoc->efi_last_u_lba + 1; 3522 slice_size -= EFI_MIN_RESV_SIZE; 3523 if (start_block == MAXOFFSET_T) 3524 start_block = NEW_START_BLOCK; 3525 slice_size -= start_block; 3526 3527 vtoc->efi_parts[0].p_start = start_block; 3528 vtoc->efi_parts[0].p_size = slice_size; 3529 3530 /* 3531 * Why we use V_USR: V_BACKUP confuses users, and is considered 3532 * disposable by some EFI utilities (since EFI doesn't have a backup 3533 * slice). V_UNASSIGNED is supposed to be used only for zero size 3534 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT, 3535 * etc. were all pretty specific. V_USR is as close to reality as we 3536 * can get, in the absence of V_OTHER. 3537 */ 3538 vtoc->efi_parts[0].p_tag = V_USR; 3539 (void) strcpy(vtoc->efi_parts[0].p_name, "zfs"); 3540 3541 vtoc->efi_parts[8].p_start = slice_size + start_block; 3542 vtoc->efi_parts[8].p_size = resv; 3543 vtoc->efi_parts[8].p_tag = V_RESERVED; 3544 3545 if (efi_write(fd, vtoc) != 0) { 3546 /* 3547 * Some block drivers (like pcata) may not support EFI 3548 * GPT labels. Print out a helpful error message dir- 3549 * ecting the user to manually label the disk and give 3550 * a specific slice. 3551 */ 3552 (void) close(fd); 3553 efi_free(vtoc); 3554 3555 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3556 "try using fdisk(1M) and then provide a specific slice")); 3557 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf)); 3558 } 3559 3560 (void) close(fd); 3561 efi_free(vtoc); 3562 return (0); 3563 } 3564 3565 static boolean_t 3566 supported_dump_vdev_type(libzfs_handle_t *hdl, nvlist_t *config, char *errbuf) 3567 { 3568 char *type; 3569 nvlist_t **child; 3570 uint_t children, c; 3571 3572 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_TYPE, &type) == 0); 3573 if (strcmp(type, VDEV_TYPE_RAIDZ) == 0 || 3574 strcmp(type, VDEV_TYPE_FILE) == 0 || 3575 strcmp(type, VDEV_TYPE_LOG) == 0 || 3576 strcmp(type, VDEV_TYPE_HOLE) == 0 || 3577 strcmp(type, VDEV_TYPE_MISSING) == 0) { 3578 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3579 "vdev type '%s' is not supported"), type); 3580 (void) zfs_error(hdl, EZFS_VDEVNOTSUP, errbuf); 3581 return (B_FALSE); 3582 } 3583 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN, 3584 &child, &children) == 0) { 3585 for (c = 0; c < children; c++) { 3586 if (!supported_dump_vdev_type(hdl, child[c], errbuf)) 3587 return (B_FALSE); 3588 } 3589 } 3590 return (B_TRUE); 3591 } 3592 3593 /* 3594 * check if this zvol is allowable for use as a dump device; zero if 3595 * it is, > 0 if it isn't, < 0 if it isn't a zvol 3596 */ 3597 int 3598 zvol_check_dump_config(char *arg) 3599 { 3600 zpool_handle_t *zhp = NULL; 3601 nvlist_t *config, *nvroot; 3602 char *p, *volname; 3603 nvlist_t **top; 3604 uint_t toplevels; 3605 libzfs_handle_t *hdl; 3606 char errbuf[1024]; 3607 char poolname[ZPOOL_MAXNAMELEN]; 3608 int pathlen = strlen(ZVOL_FULL_DEV_DIR); 3609 int ret = 1; 3610 3611 if (strncmp(arg, ZVOL_FULL_DEV_DIR, pathlen)) { 3612 return (-1); 3613 } 3614 3615 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, 3616 "dump is not supported on device '%s'"), arg); 3617 3618 if ((hdl = libzfs_init()) == NULL) 3619 return (1); 3620 libzfs_print_on_error(hdl, B_TRUE); 3621 3622 volname = arg + pathlen; 3623 3624 /* check the configuration of the pool */ 3625 if ((p = strchr(volname, '/')) == NULL) { 3626 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3627 "malformed dataset name")); 3628 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf); 3629 return (1); 3630 } else if (p - volname >= ZFS_MAXNAMELEN) { 3631 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3632 "dataset name is too long")); 3633 (void) zfs_error(hdl, EZFS_NAMETOOLONG, errbuf); 3634 return (1); 3635 } else { 3636 (void) strncpy(poolname, volname, p - volname); 3637 poolname[p - volname] = '\0'; 3638 } 3639 3640 if ((zhp = zpool_open(hdl, poolname)) == NULL) { 3641 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3642 "could not open pool '%s'"), poolname); 3643 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf); 3644 goto out; 3645 } 3646 config = zpool_get_config(zhp, NULL); 3647 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 3648 &nvroot) != 0) { 3649 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3650 "could not obtain vdev configuration for '%s'"), poolname); 3651 (void) zfs_error(hdl, EZFS_INVALCONFIG, errbuf); 3652 goto out; 3653 } 3654 3655 verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, 3656 &top, &toplevels) == 0); 3657 if (toplevels != 1) { 3658 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3659 "'%s' has multiple top level vdevs"), poolname); 3660 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, errbuf); 3661 goto out; 3662 } 3663 3664 if (!supported_dump_vdev_type(hdl, top[0], errbuf)) { 3665 goto out; 3666 } 3667 ret = 0; 3668 3669 out: 3670 if (zhp) 3671 zpool_close(zhp); 3672 libzfs_fini(hdl); 3673 return (ret); 3674 } 3675