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 * Copyright (c) 2011, 2020 by Delphix. All rights reserved. 25 * Copyright 2019 Joyent, Inc. 26 * Copyright 2016 Nexenta Systems, Inc. 27 * Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com> 28 * Copyright (c) 2017 Datto Inc. 29 * Copyright (c) 2017, Intel Corporation. 30 * Copyright 2022 OmniOS Community Edition (OmniOSce) Association. 31 */ 32 33 #include <ctype.h> 34 #include <errno.h> 35 #include <devid.h> 36 #include <fcntl.h> 37 #include <libintl.h> 38 #include <stdio.h> 39 #include <stdlib.h> 40 #include <strings.h> 41 #include <unistd.h> 42 #include <libgen.h> 43 #include <sys/dkio.h> 44 #include <sys/efi_partition.h> 45 #include <sys/vtoc.h> 46 #include <sys/zfs_ioctl.h> 47 #include <dlfcn.h> 48 #include <libzutil.h> 49 50 #include "zfs_namecheck.h" 51 #include "zfs_prop.h" 52 #include "libzfs_impl.h" 53 #include "zfs_comutil.h" 54 #include "zfeature_common.h" 55 56 static int read_efi_label(nvlist_t *, diskaddr_t *, boolean_t *); 57 static boolean_t zpool_vdev_is_interior(const char *name); 58 59 #define BACKUP_SLICE "s2" 60 61 typedef struct prop_flags { 62 int create:1; /* Validate property on creation */ 63 int import:1; /* Validate property on import */ 64 } prop_flags_t; 65 66 /* 67 * ==================================================================== 68 * zpool property functions 69 * ==================================================================== 70 */ 71 72 static int 73 zpool_get_all_props(zpool_handle_t *zhp) 74 { 75 zfs_cmd_t zc = { 0 }; 76 libzfs_handle_t *hdl = zhp->zpool_hdl; 77 78 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 79 80 if (zcmd_alloc_dst_nvlist(hdl, &zc, 0) != 0) 81 return (-1); 82 83 while (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_PROPS, &zc) != 0) { 84 if (errno == ENOMEM) { 85 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) { 86 zcmd_free_nvlists(&zc); 87 return (-1); 88 } 89 } else { 90 zcmd_free_nvlists(&zc); 91 return (-1); 92 } 93 } 94 95 if (zcmd_read_dst_nvlist(hdl, &zc, &zhp->zpool_props) != 0) { 96 zcmd_free_nvlists(&zc); 97 return (-1); 98 } 99 100 zcmd_free_nvlists(&zc); 101 102 return (0); 103 } 104 105 static int 106 zpool_props_refresh(zpool_handle_t *zhp) 107 { 108 nvlist_t *old_props; 109 110 old_props = zhp->zpool_props; 111 112 if (zpool_get_all_props(zhp) != 0) 113 return (-1); 114 115 nvlist_free(old_props); 116 return (0); 117 } 118 119 static char * 120 zpool_get_prop_string(zpool_handle_t *zhp, zpool_prop_t prop, 121 zprop_source_t *src) 122 { 123 nvlist_t *nv, *nvl; 124 uint64_t ival; 125 char *value; 126 zprop_source_t source; 127 128 nvl = zhp->zpool_props; 129 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) { 130 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &ival) == 0); 131 source = ival; 132 verify(nvlist_lookup_string(nv, ZPROP_VALUE, &value) == 0); 133 } else { 134 source = ZPROP_SRC_DEFAULT; 135 if ((value = (char *)zpool_prop_default_string(prop)) == NULL) 136 value = "-"; 137 } 138 139 if (src) 140 *src = source; 141 142 return (value); 143 } 144 145 uint64_t 146 zpool_get_prop_int(zpool_handle_t *zhp, zpool_prop_t prop, zprop_source_t *src) 147 { 148 nvlist_t *nv, *nvl; 149 uint64_t value; 150 zprop_source_t source; 151 152 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp)) { 153 /* 154 * zpool_get_all_props() has most likely failed because 155 * the pool is faulted, but if all we need is the top level 156 * vdev's guid then get it from the zhp config nvlist. 157 */ 158 if ((prop == ZPOOL_PROP_GUID) && 159 (nvlist_lookup_nvlist(zhp->zpool_config, 160 ZPOOL_CONFIG_VDEV_TREE, &nv) == 0) && 161 (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value) 162 == 0)) { 163 return (value); 164 } 165 return (zpool_prop_default_numeric(prop)); 166 } 167 168 nvl = zhp->zpool_props; 169 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) { 170 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &value) == 0); 171 source = value; 172 verify(nvlist_lookup_uint64(nv, ZPROP_VALUE, &value) == 0); 173 } else { 174 source = ZPROP_SRC_DEFAULT; 175 value = zpool_prop_default_numeric(prop); 176 } 177 178 if (src) 179 *src = source; 180 181 return (value); 182 } 183 184 /* 185 * Map VDEV STATE to printed strings. 186 */ 187 const char * 188 zpool_state_to_name(vdev_state_t state, vdev_aux_t aux) 189 { 190 switch (state) { 191 case VDEV_STATE_CLOSED: 192 case VDEV_STATE_OFFLINE: 193 return (gettext("OFFLINE")); 194 case VDEV_STATE_REMOVED: 195 return (gettext("REMOVED")); 196 case VDEV_STATE_CANT_OPEN: 197 if (aux == VDEV_AUX_CORRUPT_DATA || aux == VDEV_AUX_BAD_LOG) 198 return (gettext("FAULTED")); 199 else if (aux == VDEV_AUX_SPLIT_POOL) 200 return (gettext("SPLIT")); 201 else 202 return (gettext("UNAVAIL")); 203 case VDEV_STATE_FAULTED: 204 return (gettext("FAULTED")); 205 case VDEV_STATE_DEGRADED: 206 return (gettext("DEGRADED")); 207 case VDEV_STATE_HEALTHY: 208 return (gettext("ONLINE")); 209 210 default: 211 break; 212 } 213 214 return (gettext("UNKNOWN")); 215 } 216 217 /* 218 * Map POOL STATE to printed strings. 219 */ 220 const char * 221 zpool_pool_state_to_name(pool_state_t state) 222 { 223 switch (state) { 224 case POOL_STATE_ACTIVE: 225 return (gettext("ACTIVE")); 226 case POOL_STATE_EXPORTED: 227 return (gettext("EXPORTED")); 228 case POOL_STATE_DESTROYED: 229 return (gettext("DESTROYED")); 230 case POOL_STATE_SPARE: 231 return (gettext("SPARE")); 232 case POOL_STATE_L2CACHE: 233 return (gettext("L2CACHE")); 234 case POOL_STATE_UNINITIALIZED: 235 return (gettext("UNINITIALIZED")); 236 case POOL_STATE_UNAVAIL: 237 return (gettext("UNAVAIL")); 238 case POOL_STATE_POTENTIALLY_ACTIVE: 239 return (gettext("POTENTIALLY_ACTIVE")); 240 } 241 242 return (gettext("UNKNOWN")); 243 } 244 245 /* 246 * Get a zpool property value for 'prop' and return the value in 247 * a pre-allocated buffer. 248 */ 249 int 250 zpool_get_prop(zpool_handle_t *zhp, zpool_prop_t prop, char *buf, size_t len, 251 zprop_source_t *srctype, boolean_t literal) 252 { 253 uint64_t intval; 254 const char *strval; 255 zprop_source_t src = ZPROP_SRC_NONE; 256 nvlist_t *nvroot; 257 vdev_stat_t *vs; 258 uint_t vsc; 259 260 if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) { 261 switch (prop) { 262 case ZPOOL_PROP_NAME: 263 (void) strlcpy(buf, zpool_get_name(zhp), len); 264 break; 265 266 case ZPOOL_PROP_HEALTH: 267 (void) strlcpy(buf, "FAULTED", len); 268 break; 269 270 case ZPOOL_PROP_GUID: 271 intval = zpool_get_prop_int(zhp, prop, &src); 272 (void) snprintf(buf, len, "%llu", intval); 273 break; 274 275 case ZPOOL_PROP_ALTROOT: 276 case ZPOOL_PROP_CACHEFILE: 277 case ZPOOL_PROP_COMMENT: 278 if (zhp->zpool_props != NULL || 279 zpool_get_all_props(zhp) == 0) { 280 (void) strlcpy(buf, 281 zpool_get_prop_string(zhp, prop, &src), 282 len); 283 break; 284 } 285 /* FALLTHROUGH */ 286 default: 287 (void) strlcpy(buf, "-", len); 288 break; 289 } 290 291 if (srctype != NULL) 292 *srctype = src; 293 return (0); 294 } 295 296 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp) && 297 prop != ZPOOL_PROP_NAME) 298 return (-1); 299 300 switch (zpool_prop_get_type(prop)) { 301 case PROP_TYPE_STRING: 302 (void) strlcpy(buf, zpool_get_prop_string(zhp, prop, &src), 303 len); 304 break; 305 306 case PROP_TYPE_NUMBER: 307 intval = zpool_get_prop_int(zhp, prop, &src); 308 309 switch (prop) { 310 case ZPOOL_PROP_SIZE: 311 case ZPOOL_PROP_ALLOCATED: 312 case ZPOOL_PROP_FREE: 313 case ZPOOL_PROP_FREEING: 314 case ZPOOL_PROP_LEAKED: 315 case ZPOOL_PROP_ASHIFT: 316 if (literal) { 317 (void) snprintf(buf, len, "%llu", 318 (u_longlong_t)intval); 319 } else { 320 (void) zfs_nicenum(intval, buf, len); 321 } 322 break; 323 case ZPOOL_PROP_BOOTSIZE: 324 case ZPOOL_PROP_EXPANDSZ: 325 case ZPOOL_PROP_CHECKPOINT: 326 if (intval == 0) { 327 (void) strlcpy(buf, "-", len); 328 } else if (literal) { 329 (void) snprintf(buf, len, "%llu", 330 (u_longlong_t)intval); 331 } else { 332 (void) zfs_nicenum(intval, buf, len); 333 } 334 break; 335 case ZPOOL_PROP_CAPACITY: 336 if (literal) { 337 (void) snprintf(buf, len, "%llu", 338 (u_longlong_t)intval); 339 } else { 340 (void) snprintf(buf, len, "%llu%%", 341 (u_longlong_t)intval); 342 } 343 break; 344 case ZPOOL_PROP_FRAGMENTATION: 345 if (intval == UINT64_MAX) { 346 (void) strlcpy(buf, "-", len); 347 } else if (literal) { 348 (void) snprintf(buf, len, "%llu", 349 (u_longlong_t)intval); 350 } else { 351 (void) snprintf(buf, len, "%llu%%", 352 (u_longlong_t)intval); 353 } 354 break; 355 case ZPOOL_PROP_DEDUPRATIO: 356 if (literal) 357 (void) snprintf(buf, len, "%llu.%02llu", 358 (u_longlong_t)(intval / 100), 359 (u_longlong_t)(intval % 100)); 360 else 361 (void) snprintf(buf, len, "%llu.%02llux", 362 (u_longlong_t)(intval / 100), 363 (u_longlong_t)(intval % 100)); 364 break; 365 case ZPOOL_PROP_HEALTH: 366 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL), 367 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 368 verify(nvlist_lookup_uint64_array(nvroot, 369 ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc) 370 == 0); 371 372 (void) strlcpy(buf, zpool_state_to_name(intval, 373 vs->vs_aux), len); 374 break; 375 case ZPOOL_PROP_VERSION: 376 if (intval >= SPA_VERSION_FEATURES) { 377 (void) snprintf(buf, len, "-"); 378 break; 379 } 380 /* FALLTHROUGH */ 381 default: 382 (void) snprintf(buf, len, "%llu", intval); 383 } 384 break; 385 386 case PROP_TYPE_INDEX: 387 intval = zpool_get_prop_int(zhp, prop, &src); 388 if (zpool_prop_index_to_string(prop, intval, &strval) 389 != 0) 390 return (-1); 391 (void) strlcpy(buf, strval, len); 392 break; 393 394 default: 395 abort(); 396 } 397 398 if (srctype) 399 *srctype = src; 400 401 return (0); 402 } 403 404 /* 405 * Check if the bootfs name has the same pool name as it is set to. 406 * Assuming bootfs is a valid dataset name. 407 */ 408 static boolean_t 409 bootfs_name_valid(const char *pool, const char *bootfs) 410 { 411 int len = strlen(pool); 412 if (bootfs[0] == '\0') 413 return (B_TRUE); 414 415 if (!zfs_name_valid(bootfs, ZFS_TYPE_FILESYSTEM|ZFS_TYPE_SNAPSHOT)) 416 return (B_FALSE); 417 418 if (strncmp(pool, bootfs, len) == 0 && 419 (bootfs[len] == '/' || bootfs[len] == '\0')) 420 return (B_TRUE); 421 422 return (B_FALSE); 423 } 424 425 boolean_t 426 zpool_is_bootable(zpool_handle_t *zhp) 427 { 428 char bootfs[ZFS_MAX_DATASET_NAME_LEN]; 429 430 return (zpool_get_prop(zhp, ZPOOL_PROP_BOOTFS, bootfs, 431 sizeof (bootfs), NULL, B_FALSE) == 0 && strncmp(bootfs, "-", 432 sizeof (bootfs)) != 0); 433 } 434 435 436 /* 437 * Given an nvlist of zpool properties to be set, validate that they are 438 * correct, and parse any numeric properties (index, boolean, etc) if they are 439 * specified as strings. 440 */ 441 static nvlist_t * 442 zpool_valid_proplist(libzfs_handle_t *hdl, const char *poolname, 443 nvlist_t *props, uint64_t version, prop_flags_t flags, char *errbuf) 444 { 445 nvpair_t *elem; 446 nvlist_t *retprops; 447 zpool_prop_t prop; 448 char *strval; 449 uint64_t intval; 450 char *slash, *check; 451 struct stat64 statbuf; 452 zpool_handle_t *zhp; 453 454 if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) { 455 (void) no_memory(hdl); 456 return (NULL); 457 } 458 459 elem = NULL; 460 while ((elem = nvlist_next_nvpair(props, elem)) != NULL) { 461 const char *propname = nvpair_name(elem); 462 463 prop = zpool_name_to_prop(propname); 464 if (prop == ZPOOL_PROP_INVAL && zpool_prop_feature(propname)) { 465 int err; 466 char *fname = strchr(propname, '@') + 1; 467 468 err = zfeature_lookup_name(fname, NULL); 469 if (err != 0) { 470 ASSERT3U(err, ==, ENOENT); 471 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 472 "invalid feature '%s', '%s'"), fname, 473 propname); 474 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 475 goto error; 476 } 477 478 if (nvpair_type(elem) != DATA_TYPE_STRING) { 479 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 480 "'%s' must be a string"), propname); 481 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 482 goto error; 483 } 484 485 (void) nvpair_value_string(elem, &strval); 486 if (strcmp(strval, ZFS_FEATURE_ENABLED) != 0 && 487 strcmp(strval, ZFS_FEATURE_DISABLED) != 0) { 488 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 489 "property '%s' can only be set to " 490 "'enabled' or 'disabled'"), propname); 491 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 492 goto error; 493 } 494 495 if (nvlist_add_uint64(retprops, propname, 0) != 0) { 496 (void) no_memory(hdl); 497 goto error; 498 } 499 continue; 500 } 501 502 /* 503 * Make sure this property is valid and applies to this type. 504 */ 505 if (prop == ZPOOL_PROP_INVAL) { 506 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 507 "invalid property '%s'"), propname); 508 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 509 goto error; 510 } 511 512 if (zpool_prop_readonly(prop)) { 513 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' " 514 "is readonly"), propname); 515 (void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf); 516 goto error; 517 } 518 519 if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_POOL, retprops, 520 &strval, &intval, errbuf) != 0) 521 goto error; 522 523 /* 524 * Perform additional checking for specific properties. 525 */ 526 switch (prop) { 527 case ZPOOL_PROP_VERSION: 528 if (intval < version || 529 !SPA_VERSION_IS_SUPPORTED(intval)) { 530 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 531 "property '%s' number %d is invalid."), 532 propname, intval); 533 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf); 534 goto error; 535 } 536 break; 537 538 case ZPOOL_PROP_BOOTSIZE: 539 if (!flags.create) { 540 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 541 "property '%s' can only be set during pool " 542 "creation"), propname); 543 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 544 goto error; 545 } 546 break; 547 548 case ZPOOL_PROP_ASHIFT: 549 if (intval != 0 && 550 (intval < ASHIFT_MIN || intval > ASHIFT_MAX)) { 551 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 552 "invalid '%s=%d' property: only values " 553 "between %" PRId32 " and %" PRId32 " " 554 "are allowed.\n"), 555 propname, intval, ASHIFT_MIN, ASHIFT_MAX); 556 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 557 goto error; 558 } 559 break; 560 561 case ZPOOL_PROP_BOOTFS: 562 if (flags.create || flags.import) { 563 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 564 "property '%s' cannot be set at creation " 565 "or import time"), propname); 566 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 567 goto error; 568 } 569 570 if (version < SPA_VERSION_BOOTFS) { 571 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 572 "pool must be upgraded to support " 573 "'%s' property"), propname); 574 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf); 575 goto error; 576 } 577 578 /* 579 * bootfs property value has to be a dataset name and 580 * the dataset has to be in the same pool as it sets to. 581 */ 582 if (!bootfs_name_valid(poolname, strval)) { 583 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' " 584 "is an invalid name"), strval); 585 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf); 586 goto error; 587 } 588 589 if ((zhp = zpool_open_canfail(hdl, poolname)) == NULL) { 590 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 591 "could not open pool '%s'"), poolname); 592 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf); 593 goto error; 594 } 595 zpool_close(zhp); 596 break; 597 598 case ZPOOL_PROP_ALTROOT: 599 if (!flags.create && !flags.import) { 600 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 601 "property '%s' can only be set during pool " 602 "creation or import"), propname); 603 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 604 goto error; 605 } 606 607 if (strval[0] != '/') { 608 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 609 "bad alternate root '%s'"), strval); 610 (void) zfs_error(hdl, EZFS_BADPATH, errbuf); 611 goto error; 612 } 613 break; 614 615 case ZPOOL_PROP_CACHEFILE: 616 if (strval[0] == '\0') 617 break; 618 619 if (strcmp(strval, "none") == 0) 620 break; 621 622 if (strval[0] != '/') { 623 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 624 "property '%s' must be empty, an " 625 "absolute path, or 'none'"), propname); 626 (void) zfs_error(hdl, EZFS_BADPATH, errbuf); 627 goto error; 628 } 629 630 slash = strrchr(strval, '/'); 631 632 if (slash[1] == '\0' || strcmp(slash, "/.") == 0 || 633 strcmp(slash, "/..") == 0) { 634 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 635 "'%s' is not a valid file"), strval); 636 (void) zfs_error(hdl, EZFS_BADPATH, errbuf); 637 goto error; 638 } 639 640 *slash = '\0'; 641 642 if (strval[0] != '\0' && 643 (stat64(strval, &statbuf) != 0 || 644 !S_ISDIR(statbuf.st_mode))) { 645 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 646 "'%s' is not a valid directory"), 647 strval); 648 (void) zfs_error(hdl, EZFS_BADPATH, errbuf); 649 goto error; 650 } 651 652 *slash = '/'; 653 break; 654 655 case ZPOOL_PROP_COMMENT: 656 for (check = strval; *check != '\0'; check++) { 657 if (!isprint(*check)) { 658 zfs_error_aux(hdl, 659 dgettext(TEXT_DOMAIN, 660 "comment may only have printable " 661 "characters")); 662 (void) zfs_error(hdl, EZFS_BADPROP, 663 errbuf); 664 goto error; 665 } 666 } 667 if (strlen(strval) > ZPROP_MAX_COMMENT) { 668 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 669 "comment must not exceed %d characters"), 670 ZPROP_MAX_COMMENT); 671 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 672 goto error; 673 } 674 break; 675 676 case ZPOOL_PROP_READONLY: 677 if (!flags.import) { 678 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 679 "property '%s' can only be set at " 680 "import time"), propname); 681 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 682 goto error; 683 } 684 break; 685 686 case ZPOOL_PROP_TNAME: 687 if (!flags.create) { 688 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 689 "property '%s' can only be set at " 690 "creation time"), propname); 691 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 692 goto error; 693 } 694 break; 695 696 case ZPOOL_PROP_MULTIHOST: 697 if (get_system_hostid() == 0) { 698 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 699 "requires a non-zero system hostid")); 700 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 701 goto error; 702 } 703 break; 704 705 default: 706 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 707 "property '%s'(%d) not defined"), propname, prop); 708 break; 709 } 710 } 711 712 return (retprops); 713 error: 714 nvlist_free(retprops); 715 return (NULL); 716 } 717 718 /* 719 * Set zpool property : propname=propval. 720 */ 721 int 722 zpool_set_prop(zpool_handle_t *zhp, const char *propname, const char *propval) 723 { 724 zfs_cmd_t zc = { 0 }; 725 int ret = -1; 726 char errbuf[1024]; 727 nvlist_t *nvl = NULL; 728 nvlist_t *realprops; 729 uint64_t version; 730 prop_flags_t flags = { 0 }; 731 732 (void) snprintf(errbuf, sizeof (errbuf), 733 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"), 734 zhp->zpool_name); 735 736 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0) 737 return (no_memory(zhp->zpool_hdl)); 738 739 if (nvlist_add_string(nvl, propname, propval) != 0) { 740 nvlist_free(nvl); 741 return (no_memory(zhp->zpool_hdl)); 742 } 743 744 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL); 745 if ((realprops = zpool_valid_proplist(zhp->zpool_hdl, 746 zhp->zpool_name, nvl, version, flags, errbuf)) == NULL) { 747 nvlist_free(nvl); 748 return (-1); 749 } 750 751 nvlist_free(nvl); 752 nvl = realprops; 753 754 /* 755 * Execute the corresponding ioctl() to set this property. 756 */ 757 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 758 759 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) { 760 nvlist_free(nvl); 761 return (-1); 762 } 763 764 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc); 765 766 zcmd_free_nvlists(&zc); 767 nvlist_free(nvl); 768 769 if (ret) 770 (void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf); 771 else 772 (void) zpool_props_refresh(zhp); 773 774 return (ret); 775 } 776 777 int 778 zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp) 779 { 780 libzfs_handle_t *hdl = zhp->zpool_hdl; 781 zprop_list_t *entry; 782 char buf[ZFS_MAXPROPLEN]; 783 nvlist_t *features = NULL; 784 zprop_list_t **last; 785 boolean_t firstexpand = (NULL == *plp); 786 787 if (zprop_expand_list(hdl, plp, ZFS_TYPE_POOL) != 0) 788 return (-1); 789 790 last = plp; 791 while (*last != NULL) 792 last = &(*last)->pl_next; 793 794 if ((*plp)->pl_all) 795 features = zpool_get_features(zhp); 796 797 if ((*plp)->pl_all && firstexpand) { 798 for (int i = 0; i < SPA_FEATURES; i++) { 799 zprop_list_t *entry = zfs_alloc(hdl, 800 sizeof (zprop_list_t)); 801 entry->pl_prop = ZPROP_INVAL; 802 entry->pl_user_prop = zfs_asprintf(hdl, "feature@%s", 803 spa_feature_table[i].fi_uname); 804 entry->pl_width = strlen(entry->pl_user_prop); 805 entry->pl_all = B_TRUE; 806 807 *last = entry; 808 last = &entry->pl_next; 809 } 810 } 811 812 /* add any unsupported features */ 813 for (nvpair_t *nvp = nvlist_next_nvpair(features, NULL); 814 nvp != NULL; nvp = nvlist_next_nvpair(features, nvp)) { 815 char *propname; 816 boolean_t found; 817 zprop_list_t *entry; 818 819 if (zfeature_is_supported(nvpair_name(nvp))) 820 continue; 821 822 propname = zfs_asprintf(hdl, "unsupported@%s", 823 nvpair_name(nvp)); 824 825 /* 826 * Before adding the property to the list make sure that no 827 * other pool already added the same property. 828 */ 829 found = B_FALSE; 830 entry = *plp; 831 while (entry != NULL) { 832 if (entry->pl_user_prop != NULL && 833 strcmp(propname, entry->pl_user_prop) == 0) { 834 found = B_TRUE; 835 break; 836 } 837 entry = entry->pl_next; 838 } 839 if (found) { 840 free(propname); 841 continue; 842 } 843 844 entry = zfs_alloc(hdl, sizeof (zprop_list_t)); 845 entry->pl_prop = ZPROP_INVAL; 846 entry->pl_user_prop = propname; 847 entry->pl_width = strlen(entry->pl_user_prop); 848 entry->pl_all = B_TRUE; 849 850 *last = entry; 851 last = &entry->pl_next; 852 } 853 854 for (entry = *plp; entry != NULL; entry = entry->pl_next) { 855 856 if (entry->pl_fixed) 857 continue; 858 859 if (entry->pl_prop != ZPROP_INVAL && 860 zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf), 861 NULL, B_FALSE) == 0) { 862 if (strlen(buf) > entry->pl_width) 863 entry->pl_width = strlen(buf); 864 } 865 } 866 867 return (0); 868 } 869 870 /* 871 * Get the state for the given feature on the given ZFS pool. 872 */ 873 int 874 zpool_prop_get_feature(zpool_handle_t *zhp, const char *propname, char *buf, 875 size_t len) 876 { 877 uint64_t refcount; 878 boolean_t found = B_FALSE; 879 nvlist_t *features = zpool_get_features(zhp); 880 boolean_t supported; 881 const char *feature = strchr(propname, '@') + 1; 882 883 supported = zpool_prop_feature(propname); 884 ASSERT(supported || zpool_prop_unsupported(propname)); 885 886 /* 887 * Convert from feature name to feature guid. This conversion is 888 * unecessary for unsupported@... properties because they already 889 * use guids. 890 */ 891 if (supported) { 892 int ret; 893 spa_feature_t fid; 894 895 ret = zfeature_lookup_name(feature, &fid); 896 if (ret != 0) { 897 (void) strlcpy(buf, "-", len); 898 return (ENOTSUP); 899 } 900 feature = spa_feature_table[fid].fi_guid; 901 } 902 903 if (nvlist_lookup_uint64(features, feature, &refcount) == 0) 904 found = B_TRUE; 905 906 if (supported) { 907 if (!found) { 908 (void) strlcpy(buf, ZFS_FEATURE_DISABLED, len); 909 } else { 910 if (refcount == 0) 911 (void) strlcpy(buf, ZFS_FEATURE_ENABLED, len); 912 else 913 (void) strlcpy(buf, ZFS_FEATURE_ACTIVE, len); 914 } 915 } else { 916 if (found) { 917 if (refcount == 0) { 918 (void) strcpy(buf, ZFS_UNSUPPORTED_INACTIVE); 919 } else { 920 (void) strcpy(buf, ZFS_UNSUPPORTED_READONLY); 921 } 922 } else { 923 (void) strlcpy(buf, "-", len); 924 return (ENOTSUP); 925 } 926 } 927 928 return (0); 929 } 930 931 /* 932 * Don't start the slice at the default block of 34; many storage 933 * devices will use a stripe width of 128k, so start there instead. 934 */ 935 #define NEW_START_BLOCK 256 936 937 /* 938 * Validate the given pool name, optionally putting an extended error message in 939 * 'buf'. 940 */ 941 boolean_t 942 zpool_name_valid(libzfs_handle_t *hdl, boolean_t isopen, const char *pool) 943 { 944 namecheck_err_t why; 945 char what; 946 int ret; 947 948 ret = pool_namecheck(pool, &why, &what); 949 950 /* 951 * The rules for reserved pool names were extended at a later point. 952 * But we need to support users with existing pools that may now be 953 * invalid. So we only check for this expanded set of names during a 954 * create (or import), and only in userland. 955 */ 956 if (ret == 0 && !isopen && 957 (strncmp(pool, "mirror", 6) == 0 || 958 strncmp(pool, "raidz", 5) == 0 || 959 strncmp(pool, "spare", 5) == 0 || 960 strcmp(pool, "log") == 0)) { 961 if (hdl != NULL) 962 zfs_error_aux(hdl, 963 dgettext(TEXT_DOMAIN, "name is reserved")); 964 return (B_FALSE); 965 } 966 967 968 if (ret != 0) { 969 if (hdl != NULL) { 970 switch (why) { 971 case NAME_ERR_TOOLONG: 972 zfs_error_aux(hdl, 973 dgettext(TEXT_DOMAIN, "name is too long")); 974 break; 975 976 case NAME_ERR_INVALCHAR: 977 zfs_error_aux(hdl, 978 dgettext(TEXT_DOMAIN, "invalid character " 979 "'%c' in pool name"), what); 980 break; 981 982 case NAME_ERR_NOLETTER: 983 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 984 "name must begin with a letter")); 985 break; 986 987 case NAME_ERR_RESERVED: 988 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 989 "name is reserved")); 990 break; 991 992 case NAME_ERR_DISKLIKE: 993 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 994 "pool name is reserved")); 995 break; 996 997 case NAME_ERR_LEADING_SLASH: 998 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 999 "leading slash in name")); 1000 break; 1001 1002 case NAME_ERR_EMPTY_COMPONENT: 1003 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1004 "empty component in name")); 1005 break; 1006 1007 case NAME_ERR_TRAILING_SLASH: 1008 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1009 "trailing slash in name")); 1010 break; 1011 1012 case NAME_ERR_MULTIPLE_DELIMITERS: 1013 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1014 "multiple '@' and/or '#' delimiters in " 1015 "name")); 1016 break; 1017 1018 default: 1019 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1020 "(%d) not defined"), why); 1021 break; 1022 } 1023 } 1024 return (B_FALSE); 1025 } 1026 1027 return (B_TRUE); 1028 } 1029 1030 /* 1031 * Open a handle to the given pool, even if the pool is currently in the FAULTED 1032 * state. 1033 */ 1034 zpool_handle_t * 1035 zpool_open_canfail(libzfs_handle_t *hdl, const char *pool) 1036 { 1037 zpool_handle_t *zhp; 1038 boolean_t missing; 1039 1040 /* 1041 * Make sure the pool name is valid. 1042 */ 1043 if (!zpool_name_valid(hdl, B_TRUE, pool)) { 1044 (void) zfs_error_fmt(hdl, EZFS_INVALIDNAME, 1045 dgettext(TEXT_DOMAIN, "cannot open '%s'"), 1046 pool); 1047 return (NULL); 1048 } 1049 1050 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL) 1051 return (NULL); 1052 1053 zhp->zpool_hdl = hdl; 1054 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name)); 1055 1056 if (zpool_refresh_stats(zhp, &missing) != 0) { 1057 zpool_close(zhp); 1058 return (NULL); 1059 } 1060 1061 if (missing) { 1062 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool")); 1063 (void) zfs_error_fmt(hdl, EZFS_NOENT, 1064 dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool); 1065 zpool_close(zhp); 1066 return (NULL); 1067 } 1068 1069 return (zhp); 1070 } 1071 1072 /* 1073 * Like the above, but silent on error. Used when iterating over pools (because 1074 * the configuration cache may be out of date). 1075 */ 1076 int 1077 zpool_open_silent(libzfs_handle_t *hdl, const char *pool, zpool_handle_t **ret) 1078 { 1079 zpool_handle_t *zhp; 1080 boolean_t missing; 1081 1082 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL) 1083 return (-1); 1084 1085 zhp->zpool_hdl = hdl; 1086 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name)); 1087 1088 if (zpool_refresh_stats(zhp, &missing) != 0) { 1089 zpool_close(zhp); 1090 return (-1); 1091 } 1092 1093 if (missing) { 1094 zpool_close(zhp); 1095 *ret = NULL; 1096 return (0); 1097 } 1098 1099 *ret = zhp; 1100 return (0); 1101 } 1102 1103 /* 1104 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted 1105 * state. 1106 */ 1107 zpool_handle_t * 1108 zpool_open(libzfs_handle_t *hdl, const char *pool) 1109 { 1110 zpool_handle_t *zhp; 1111 1112 if ((zhp = zpool_open_canfail(hdl, pool)) == NULL) 1113 return (NULL); 1114 1115 if (zhp->zpool_state == POOL_STATE_UNAVAIL) { 1116 (void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL, 1117 dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name); 1118 zpool_close(zhp); 1119 return (NULL); 1120 } 1121 1122 return (zhp); 1123 } 1124 1125 /* 1126 * Close the handle. Simply frees the memory associated with the handle. 1127 */ 1128 void 1129 zpool_close(zpool_handle_t *zhp) 1130 { 1131 nvlist_free(zhp->zpool_config); 1132 nvlist_free(zhp->zpool_old_config); 1133 nvlist_free(zhp->zpool_props); 1134 free(zhp); 1135 } 1136 1137 /* 1138 * Return the name of the pool. 1139 */ 1140 const char * 1141 zpool_get_name(zpool_handle_t *zhp) 1142 { 1143 return (zhp->zpool_name); 1144 } 1145 1146 1147 /* 1148 * Return the state of the pool (ACTIVE or UNAVAILABLE) 1149 */ 1150 int 1151 zpool_get_state(zpool_handle_t *zhp) 1152 { 1153 return (zhp->zpool_state); 1154 } 1155 1156 /* 1157 * Check if vdev list contains a special vdev 1158 */ 1159 static boolean_t 1160 zpool_has_special_vdev(nvlist_t *nvroot) 1161 { 1162 nvlist_t **child; 1163 uint_t children; 1164 1165 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, &child, 1166 &children) == 0) { 1167 for (uint_t c = 0; c < children; c++) { 1168 char *bias; 1169 1170 if (nvlist_lookup_string(child[c], 1171 ZPOOL_CONFIG_ALLOCATION_BIAS, &bias) == 0 && 1172 strcmp(bias, VDEV_ALLOC_BIAS_SPECIAL) == 0) { 1173 return (B_TRUE); 1174 } 1175 } 1176 } 1177 return (B_FALSE); 1178 } 1179 1180 /* 1181 * Create the named pool, using the provided vdev list. It is assumed 1182 * that the consumer has already validated the contents of the nvlist, so we 1183 * don't have to worry about error semantics. 1184 */ 1185 int 1186 zpool_create(libzfs_handle_t *hdl, const char *pool, nvlist_t *nvroot, 1187 nvlist_t *props, nvlist_t *fsprops) 1188 { 1189 zfs_cmd_t zc = { 0 }; 1190 nvlist_t *zc_fsprops = NULL; 1191 nvlist_t *zc_props = NULL; 1192 nvlist_t *hidden_args = NULL; 1193 uint8_t *wkeydata = NULL; 1194 uint_t wkeylen = 0; 1195 char msg[1024]; 1196 int ret = -1; 1197 1198 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1199 "cannot create '%s'"), pool); 1200 1201 if (!zpool_name_valid(hdl, B_FALSE, pool)) 1202 return (zfs_error(hdl, EZFS_INVALIDNAME, msg)); 1203 1204 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0) 1205 return (-1); 1206 1207 if (props) { 1208 prop_flags_t flags = { .create = B_TRUE, .import = B_FALSE }; 1209 1210 if ((zc_props = zpool_valid_proplist(hdl, pool, props, 1211 SPA_VERSION_1, flags, msg)) == NULL) { 1212 goto create_failed; 1213 } 1214 } 1215 1216 if (fsprops) { 1217 uint64_t zoned; 1218 char *zonestr; 1219 1220 zoned = ((nvlist_lookup_string(fsprops, 1221 zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) && 1222 strcmp(zonestr, "on") == 0); 1223 1224 if ((zc_fsprops = zfs_valid_proplist(hdl, ZFS_TYPE_FILESYSTEM, 1225 fsprops, zoned, NULL, NULL, B_TRUE, msg)) == NULL) { 1226 goto create_failed; 1227 } 1228 1229 if (nvlist_exists(zc_fsprops, 1230 zfs_prop_to_name(ZFS_PROP_SPECIAL_SMALL_BLOCKS)) && 1231 !zpool_has_special_vdev(nvroot)) { 1232 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1233 "%s property requires a special vdev"), 1234 zfs_prop_to_name(ZFS_PROP_SPECIAL_SMALL_BLOCKS)); 1235 (void) zfs_error(hdl, EZFS_BADPROP, msg); 1236 goto create_failed; 1237 } 1238 1239 if (!zc_props && 1240 (nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) { 1241 goto create_failed; 1242 } 1243 if (zfs_crypto_create(hdl, NULL, zc_fsprops, props, B_TRUE, 1244 &wkeydata, &wkeylen) != 0) { 1245 (void) zfs_error(hdl, EZFS_CRYPTOFAILED, msg); 1246 goto create_failed; 1247 } 1248 if (nvlist_add_nvlist(zc_props, 1249 ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) { 1250 goto create_failed; 1251 } 1252 if (wkeydata != NULL) { 1253 if (nvlist_alloc(&hidden_args, NV_UNIQUE_NAME, 0) != 0) 1254 goto create_failed; 1255 1256 if (nvlist_add_uint8_array(hidden_args, "wkeydata", 1257 wkeydata, wkeylen) != 0) 1258 goto create_failed; 1259 1260 if (nvlist_add_nvlist(zc_props, ZPOOL_HIDDEN_ARGS, 1261 hidden_args) != 0) 1262 goto create_failed; 1263 } 1264 } 1265 1266 if (zc_props && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0) 1267 goto create_failed; 1268 1269 (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name)); 1270 1271 if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) { 1272 1273 zcmd_free_nvlists(&zc); 1274 nvlist_free(zc_props); 1275 nvlist_free(zc_fsprops); 1276 nvlist_free(hidden_args); 1277 if (wkeydata != NULL) 1278 free(wkeydata); 1279 1280 switch (errno) { 1281 case EBUSY: 1282 /* 1283 * This can happen if the user has specified the same 1284 * device multiple times. We can't reliably detect this 1285 * until we try to add it and see we already have a 1286 * label. 1287 */ 1288 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1289 "one or more vdevs refer to the same device")); 1290 return (zfs_error(hdl, EZFS_BADDEV, msg)); 1291 1292 case ERANGE: 1293 /* 1294 * This happens if the record size is smaller or larger 1295 * than the allowed size range, or not a power of 2. 1296 * 1297 * NOTE: although zfs_valid_proplist is called earlier, 1298 * this case may have slipped through since the 1299 * pool does not exist yet and it is therefore 1300 * impossible to read properties e.g. max blocksize 1301 * from the pool. 1302 */ 1303 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1304 "record size invalid")); 1305 return (zfs_error(hdl, EZFS_BADPROP, msg)); 1306 1307 case EOVERFLOW: 1308 /* 1309 * This occurs when one of the devices is below 1310 * SPA_MINDEVSIZE. Unfortunately, we can't detect which 1311 * device was the problem device since there's no 1312 * reliable way to determine device size from userland. 1313 */ 1314 { 1315 char buf[64]; 1316 1317 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf)); 1318 1319 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1320 "one or more devices is less than the " 1321 "minimum size (%s)"), buf); 1322 } 1323 return (zfs_error(hdl, EZFS_BADDEV, msg)); 1324 1325 case ENOSPC: 1326 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1327 "one or more devices is out of space")); 1328 return (zfs_error(hdl, EZFS_BADDEV, msg)); 1329 1330 default: 1331 return (zpool_standard_error(hdl, errno, msg)); 1332 } 1333 } 1334 1335 create_failed: 1336 zcmd_free_nvlists(&zc); 1337 nvlist_free(zc_props); 1338 nvlist_free(zc_fsprops); 1339 nvlist_free(hidden_args); 1340 if (wkeydata != NULL) 1341 free(wkeydata); 1342 return (ret); 1343 } 1344 1345 /* 1346 * Destroy the given pool. It is up to the caller to ensure that there are no 1347 * datasets left in the pool. 1348 */ 1349 int 1350 zpool_destroy(zpool_handle_t *zhp, const char *log_str) 1351 { 1352 zfs_cmd_t zc = { 0 }; 1353 zfs_handle_t *zfp = NULL; 1354 libzfs_handle_t *hdl = zhp->zpool_hdl; 1355 char msg[1024]; 1356 1357 if (zhp->zpool_state == POOL_STATE_ACTIVE && 1358 (zfp = zfs_open(hdl, zhp->zpool_name, ZFS_TYPE_FILESYSTEM)) == NULL) 1359 return (-1); 1360 1361 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1362 zc.zc_history = (uint64_t)(uintptr_t)log_str; 1363 1364 if (zfs_ioctl(hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) { 1365 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1366 "cannot destroy '%s'"), zhp->zpool_name); 1367 1368 if (errno == EROFS) { 1369 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1370 "one or more devices is read only")); 1371 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1372 } else { 1373 (void) zpool_standard_error(hdl, errno, msg); 1374 } 1375 1376 if (zfp) 1377 zfs_close(zfp); 1378 return (-1); 1379 } 1380 1381 if (zfp) { 1382 remove_mountpoint(zfp); 1383 zfs_close(zfp); 1384 } 1385 1386 return (0); 1387 } 1388 1389 /* 1390 * Create a checkpoint in the given pool. 1391 */ 1392 int 1393 zpool_checkpoint(zpool_handle_t *zhp) 1394 { 1395 libzfs_handle_t *hdl = zhp->zpool_hdl; 1396 char msg[1024]; 1397 int error; 1398 1399 error = lzc_pool_checkpoint(zhp->zpool_name); 1400 if (error != 0) { 1401 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1402 "cannot checkpoint '%s'"), zhp->zpool_name); 1403 (void) zpool_standard_error(hdl, error, msg); 1404 return (-1); 1405 } 1406 1407 return (0); 1408 } 1409 1410 /* 1411 * Discard the checkpoint from the given pool. 1412 */ 1413 int 1414 zpool_discard_checkpoint(zpool_handle_t *zhp) 1415 { 1416 libzfs_handle_t *hdl = zhp->zpool_hdl; 1417 char msg[1024]; 1418 int error; 1419 1420 error = lzc_pool_checkpoint_discard(zhp->zpool_name); 1421 if (error != 0) { 1422 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1423 "cannot discard checkpoint in '%s'"), zhp->zpool_name); 1424 (void) zpool_standard_error(hdl, error, msg); 1425 return (-1); 1426 } 1427 1428 return (0); 1429 } 1430 1431 /* 1432 * Add the given vdevs to the pool. The caller must have already performed the 1433 * necessary verification to ensure that the vdev specification is well-formed. 1434 */ 1435 int 1436 zpool_add(zpool_handle_t *zhp, nvlist_t *nvroot) 1437 { 1438 zfs_cmd_t zc = { 0 }; 1439 int ret; 1440 libzfs_handle_t *hdl = zhp->zpool_hdl; 1441 char msg[1024]; 1442 nvlist_t **spares, **l2cache; 1443 uint_t nspares, nl2cache; 1444 1445 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1446 "cannot add to '%s'"), zhp->zpool_name); 1447 1448 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) < 1449 SPA_VERSION_SPARES && 1450 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, 1451 &spares, &nspares) == 0) { 1452 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be " 1453 "upgraded to add hot spares")); 1454 return (zfs_error(hdl, EZFS_BADVERSION, msg)); 1455 } 1456 1457 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) < 1458 SPA_VERSION_L2CACHE && 1459 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, 1460 &l2cache, &nl2cache) == 0) { 1461 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be " 1462 "upgraded to add cache devices")); 1463 return (zfs_error(hdl, EZFS_BADVERSION, msg)); 1464 } 1465 1466 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0) 1467 return (-1); 1468 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1469 1470 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) { 1471 switch (errno) { 1472 case EBUSY: 1473 /* 1474 * This can happen if the user has specified the same 1475 * device multiple times. We can't reliably detect this 1476 * until we try to add it and see we already have a 1477 * label. 1478 */ 1479 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1480 "one or more vdevs refer to the same device")); 1481 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1482 break; 1483 1484 case EINVAL: 1485 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1486 "invalid config; a pool with removing/removed " 1487 "vdevs does not support adding raidz vdevs")); 1488 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1489 break; 1490 1491 case EOVERFLOW: 1492 /* 1493 * This occurrs when one of the devices is below 1494 * SPA_MINDEVSIZE. Unfortunately, we can't detect which 1495 * device was the problem device since there's no 1496 * reliable way to determine device size from userland. 1497 */ 1498 { 1499 char buf[64]; 1500 1501 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf)); 1502 1503 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1504 "device is less than the minimum " 1505 "size (%s)"), buf); 1506 } 1507 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1508 break; 1509 1510 case ENOTSUP: 1511 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1512 "pool must be upgraded to add these vdevs")); 1513 (void) zfs_error(hdl, EZFS_BADVERSION, msg); 1514 break; 1515 1516 default: 1517 (void) zpool_standard_error(hdl, errno, msg); 1518 } 1519 1520 ret = -1; 1521 } else { 1522 ret = 0; 1523 } 1524 1525 zcmd_free_nvlists(&zc); 1526 1527 return (ret); 1528 } 1529 1530 /* 1531 * Exports the pool from the system. The caller must ensure that there are no 1532 * mounted datasets in the pool. 1533 */ 1534 static int 1535 zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce, 1536 const char *log_str) 1537 { 1538 zfs_cmd_t zc = { 0 }; 1539 char msg[1024]; 1540 1541 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1542 "cannot export '%s'"), zhp->zpool_name); 1543 1544 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1545 zc.zc_cookie = force; 1546 zc.zc_guid = hardforce; 1547 zc.zc_history = (uint64_t)(uintptr_t)log_str; 1548 1549 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) { 1550 switch (errno) { 1551 case EXDEV: 1552 zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN, 1553 "use '-f' to override the following errors:\n" 1554 "'%s' has an active shared spare which could be" 1555 " used by other pools once '%s' is exported."), 1556 zhp->zpool_name, zhp->zpool_name); 1557 return (zfs_error(zhp->zpool_hdl, EZFS_ACTIVE_SPARE, 1558 msg)); 1559 default: 1560 return (zpool_standard_error_fmt(zhp->zpool_hdl, errno, 1561 msg)); 1562 } 1563 } 1564 1565 return (0); 1566 } 1567 1568 int 1569 zpool_export(zpool_handle_t *zhp, boolean_t force, const char *log_str) 1570 { 1571 return (zpool_export_common(zhp, force, B_FALSE, log_str)); 1572 } 1573 1574 int 1575 zpool_export_force(zpool_handle_t *zhp, const char *log_str) 1576 { 1577 return (zpool_export_common(zhp, B_TRUE, B_TRUE, log_str)); 1578 } 1579 1580 static void 1581 zpool_rewind_exclaim(libzfs_handle_t *hdl, const char *name, boolean_t dryrun, 1582 nvlist_t *config) 1583 { 1584 nvlist_t *nv = NULL; 1585 uint64_t rewindto; 1586 int64_t loss = -1; 1587 struct tm t; 1588 char timestr[128]; 1589 1590 if (!hdl->libzfs_printerr || config == NULL) 1591 return; 1592 1593 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 || 1594 nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0) { 1595 return; 1596 } 1597 1598 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0) 1599 return; 1600 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss); 1601 1602 if (localtime_r((time_t *)&rewindto, &t) != NULL && 1603 strftime(timestr, 128, 0, &t) != 0) { 1604 if (dryrun) { 1605 (void) printf(dgettext(TEXT_DOMAIN, 1606 "Would be able to return %s " 1607 "to its state as of %s.\n"), 1608 name, timestr); 1609 } else { 1610 (void) printf(dgettext(TEXT_DOMAIN, 1611 "Pool %s returned to its state as of %s.\n"), 1612 name, timestr); 1613 } 1614 if (loss > 120) { 1615 (void) printf(dgettext(TEXT_DOMAIN, 1616 "%s approximately %lld "), 1617 dryrun ? "Would discard" : "Discarded", 1618 (loss + 30) / 60); 1619 (void) printf(dgettext(TEXT_DOMAIN, 1620 "minutes of transactions.\n")); 1621 } else if (loss > 0) { 1622 (void) printf(dgettext(TEXT_DOMAIN, 1623 "%s approximately %lld "), 1624 dryrun ? "Would discard" : "Discarded", loss); 1625 (void) printf(dgettext(TEXT_DOMAIN, 1626 "seconds of transactions.\n")); 1627 } 1628 } 1629 } 1630 1631 void 1632 zpool_explain_recover(libzfs_handle_t *hdl, const char *name, int reason, 1633 nvlist_t *config) 1634 { 1635 nvlist_t *nv = NULL; 1636 int64_t loss = -1; 1637 uint64_t edata = UINT64_MAX; 1638 uint64_t rewindto; 1639 struct tm t; 1640 char timestr[128]; 1641 1642 if (!hdl->libzfs_printerr) 1643 return; 1644 1645 if (reason >= 0) 1646 (void) printf(dgettext(TEXT_DOMAIN, "action: ")); 1647 else 1648 (void) printf(dgettext(TEXT_DOMAIN, "\t")); 1649 1650 /* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */ 1651 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 || 1652 nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0 || 1653 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0) 1654 goto no_info; 1655 1656 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss); 1657 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_DATA_ERRORS, 1658 &edata); 1659 1660 (void) printf(dgettext(TEXT_DOMAIN, 1661 "Recovery is possible, but will result in some data loss.\n")); 1662 1663 if (localtime_r((time_t *)&rewindto, &t) != NULL && 1664 strftime(timestr, 128, 0, &t) != 0) { 1665 (void) printf(dgettext(TEXT_DOMAIN, 1666 "\tReturning the pool to its state as of %s\n" 1667 "\tshould correct the problem. "), 1668 timestr); 1669 } else { 1670 (void) printf(dgettext(TEXT_DOMAIN, 1671 "\tReverting the pool to an earlier state " 1672 "should correct the problem.\n\t")); 1673 } 1674 1675 if (loss > 120) { 1676 (void) printf(dgettext(TEXT_DOMAIN, 1677 "Approximately %lld minutes of data\n" 1678 "\tmust be discarded, irreversibly. "), (loss + 30) / 60); 1679 } else if (loss > 0) { 1680 (void) printf(dgettext(TEXT_DOMAIN, 1681 "Approximately %lld seconds of data\n" 1682 "\tmust be discarded, irreversibly. "), loss); 1683 } 1684 if (edata != 0 && edata != UINT64_MAX) { 1685 if (edata == 1) { 1686 (void) printf(dgettext(TEXT_DOMAIN, 1687 "After rewind, at least\n" 1688 "\tone persistent user-data error will remain. ")); 1689 } else { 1690 (void) printf(dgettext(TEXT_DOMAIN, 1691 "After rewind, several\n" 1692 "\tpersistent user-data errors will remain. ")); 1693 } 1694 } 1695 (void) printf(dgettext(TEXT_DOMAIN, 1696 "Recovery can be attempted\n\tby executing 'zpool %s -F %s'. "), 1697 reason >= 0 ? "clear" : "import", name); 1698 1699 (void) printf(dgettext(TEXT_DOMAIN, 1700 "A scrub of the pool\n" 1701 "\tis strongly recommended after recovery.\n")); 1702 return; 1703 1704 no_info: 1705 (void) printf(dgettext(TEXT_DOMAIN, 1706 "Destroy and re-create the pool from\n\ta backup source.\n")); 1707 } 1708 1709 /* 1710 * zpool_import() is a contracted interface. Should be kept the same 1711 * if possible. 1712 * 1713 * Applications should use zpool_import_props() to import a pool with 1714 * new properties value to be set. 1715 */ 1716 int 1717 zpool_import(libzfs_handle_t *hdl, nvlist_t *config, const char *newname, 1718 char *altroot) 1719 { 1720 nvlist_t *props = NULL; 1721 int ret; 1722 1723 if (altroot != NULL) { 1724 if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) { 1725 return (zfs_error_fmt(hdl, EZFS_NOMEM, 1726 dgettext(TEXT_DOMAIN, "cannot import '%s'"), 1727 newname)); 1728 } 1729 1730 if (nvlist_add_string(props, 1731 zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 || 1732 nvlist_add_string(props, 1733 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) { 1734 nvlist_free(props); 1735 return (zfs_error_fmt(hdl, EZFS_NOMEM, 1736 dgettext(TEXT_DOMAIN, "cannot import '%s'"), 1737 newname)); 1738 } 1739 } 1740 1741 ret = zpool_import_props(hdl, config, newname, props, 1742 ZFS_IMPORT_NORMAL); 1743 nvlist_free(props); 1744 return (ret); 1745 } 1746 1747 static void 1748 print_vdev_tree(libzfs_handle_t *hdl, const char *name, nvlist_t *nv, 1749 int indent) 1750 { 1751 nvlist_t **child; 1752 uint_t c, children; 1753 char *vname; 1754 uint64_t is_log = 0; 1755 1756 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG, 1757 &is_log); 1758 1759 if (name != NULL) 1760 (void) printf("\t%*s%s%s\n", indent, "", name, 1761 is_log ? " [log]" : ""); 1762 1763 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, 1764 &child, &children) != 0) 1765 return; 1766 1767 for (c = 0; c < children; c++) { 1768 vname = zpool_vdev_name(hdl, NULL, child[c], VDEV_NAME_TYPE_ID); 1769 print_vdev_tree(hdl, vname, child[c], indent + 2); 1770 free(vname); 1771 } 1772 } 1773 1774 void 1775 zpool_print_unsup_feat(nvlist_t *config) 1776 { 1777 nvlist_t *nvinfo, *unsup_feat; 1778 1779 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 1780 0); 1781 verify(nvlist_lookup_nvlist(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT, 1782 &unsup_feat) == 0); 1783 1784 for (nvpair_t *nvp = nvlist_next_nvpair(unsup_feat, NULL); nvp != NULL; 1785 nvp = nvlist_next_nvpair(unsup_feat, nvp)) { 1786 char *desc; 1787 1788 verify(nvpair_type(nvp) == DATA_TYPE_STRING); 1789 verify(nvpair_value_string(nvp, &desc) == 0); 1790 1791 if (strlen(desc) > 0) 1792 (void) printf("\t%s (%s)\n", nvpair_name(nvp), desc); 1793 else 1794 (void) printf("\t%s\n", nvpair_name(nvp)); 1795 } 1796 } 1797 1798 /* 1799 * Import the given pool using the known configuration and a list of 1800 * properties to be set. The configuration should have come from 1801 * zpool_find_import(). The 'newname' parameters control whether the pool 1802 * is imported with a different name. 1803 */ 1804 int 1805 zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname, 1806 nvlist_t *props, int flags) 1807 { 1808 zfs_cmd_t zc = { 0 }; 1809 zpool_load_policy_t policy; 1810 nvlist_t *nv = NULL; 1811 nvlist_t *nvinfo = NULL; 1812 nvlist_t *missing = NULL; 1813 char *thename; 1814 char *origname; 1815 int ret; 1816 int error = 0; 1817 char errbuf[1024]; 1818 1819 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, 1820 &origname) == 0); 1821 1822 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, 1823 "cannot import pool '%s'"), origname); 1824 1825 if (newname != NULL) { 1826 if (!zpool_name_valid(hdl, B_FALSE, newname)) 1827 return (zfs_error_fmt(hdl, EZFS_INVALIDNAME, 1828 dgettext(TEXT_DOMAIN, "cannot import '%s'"), 1829 newname)); 1830 thename = (char *)newname; 1831 } else { 1832 thename = origname; 1833 } 1834 1835 if (props != NULL) { 1836 uint64_t version; 1837 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE }; 1838 1839 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, 1840 &version) == 0); 1841 1842 if ((props = zpool_valid_proplist(hdl, origname, 1843 props, version, flags, errbuf)) == NULL) 1844 return (-1); 1845 if (zcmd_write_src_nvlist(hdl, &zc, props) != 0) { 1846 nvlist_free(props); 1847 return (-1); 1848 } 1849 nvlist_free(props); 1850 } 1851 1852 (void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name)); 1853 1854 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, 1855 &zc.zc_guid) == 0); 1856 1857 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) { 1858 zcmd_free_nvlists(&zc); 1859 return (-1); 1860 } 1861 if (zcmd_alloc_dst_nvlist(hdl, &zc, zc.zc_nvlist_conf_size * 2) != 0) { 1862 zcmd_free_nvlists(&zc); 1863 return (-1); 1864 } 1865 1866 zc.zc_cookie = flags; 1867 while ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc)) != 0 && 1868 errno == ENOMEM) { 1869 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) { 1870 zcmd_free_nvlists(&zc); 1871 return (-1); 1872 } 1873 } 1874 if (ret != 0) 1875 error = errno; 1876 1877 (void) zcmd_read_dst_nvlist(hdl, &zc, &nv); 1878 1879 zcmd_free_nvlists(&zc); 1880 1881 zpool_get_load_policy(config, &policy); 1882 1883 if (error) { 1884 char desc[1024]; 1885 char aux[256]; 1886 1887 /* 1888 * Dry-run failed, but we print out what success 1889 * looks like if we found a best txg 1890 */ 1891 if (policy.zlp_rewind & ZPOOL_TRY_REWIND) { 1892 zpool_rewind_exclaim(hdl, newname ? origname : thename, 1893 B_TRUE, nv); 1894 nvlist_free(nv); 1895 return (-1); 1896 } 1897 1898 if (newname == NULL) 1899 (void) snprintf(desc, sizeof (desc), 1900 dgettext(TEXT_DOMAIN, "cannot import '%s'"), 1901 thename); 1902 else 1903 (void) snprintf(desc, sizeof (desc), 1904 dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"), 1905 origname, thename); 1906 1907 switch (error) { 1908 case ENOTSUP: 1909 if (nv != NULL && nvlist_lookup_nvlist(nv, 1910 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 && 1911 nvlist_exists(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT)) { 1912 (void) printf(dgettext(TEXT_DOMAIN, "This " 1913 "pool uses the following feature(s) not " 1914 "supported by this system:\n")); 1915 zpool_print_unsup_feat(nv); 1916 if (nvlist_exists(nvinfo, 1917 ZPOOL_CONFIG_CAN_RDONLY)) { 1918 (void) printf(dgettext(TEXT_DOMAIN, 1919 "All unsupported features are only " 1920 "required for writing to the pool." 1921 "\nThe pool can be imported using " 1922 "'-o readonly=on'.\n")); 1923 } 1924 } 1925 /* 1926 * Unsupported version. 1927 */ 1928 (void) zfs_error(hdl, EZFS_BADVERSION, desc); 1929 break; 1930 1931 case EREMOTEIO: 1932 if (nv != NULL && nvlist_lookup_nvlist(nv, 1933 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0) { 1934 char *hostname = "<unknown>"; 1935 uint64_t hostid = 0; 1936 mmp_state_t mmp_state; 1937 1938 mmp_state = fnvlist_lookup_uint64(nvinfo, 1939 ZPOOL_CONFIG_MMP_STATE); 1940 1941 if (nvlist_exists(nvinfo, 1942 ZPOOL_CONFIG_MMP_HOSTNAME)) 1943 hostname = fnvlist_lookup_string(nvinfo, 1944 ZPOOL_CONFIG_MMP_HOSTNAME); 1945 1946 if (nvlist_exists(nvinfo, 1947 ZPOOL_CONFIG_MMP_HOSTID)) 1948 hostid = fnvlist_lookup_uint64(nvinfo, 1949 ZPOOL_CONFIG_MMP_HOSTID); 1950 1951 if (mmp_state == MMP_STATE_ACTIVE) { 1952 (void) snprintf(aux, sizeof (aux), 1953 dgettext(TEXT_DOMAIN, "pool is imp" 1954 "orted on host '%s' (hostid=%lx).\n" 1955 "Export the pool on the other " 1956 "system, then run 'zpool import'."), 1957 hostname, (unsigned long) hostid); 1958 } else if (mmp_state == MMP_STATE_NO_HOSTID) { 1959 (void) snprintf(aux, sizeof (aux), 1960 dgettext(TEXT_DOMAIN, "pool has " 1961 "the multihost property on and " 1962 "the\nsystem's hostid is not " 1963 "set.\n")); 1964 } 1965 1966 (void) zfs_error_aux(hdl, aux); 1967 } 1968 (void) zfs_error(hdl, EZFS_ACTIVE_POOL, desc); 1969 break; 1970 1971 case EINVAL: 1972 (void) zfs_error(hdl, EZFS_INVALCONFIG, desc); 1973 break; 1974 1975 case EROFS: 1976 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1977 "one or more devices is read only")); 1978 (void) zfs_error(hdl, EZFS_BADDEV, desc); 1979 break; 1980 1981 case ENXIO: 1982 if (nv && nvlist_lookup_nvlist(nv, 1983 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 && 1984 nvlist_lookup_nvlist(nvinfo, 1985 ZPOOL_CONFIG_MISSING_DEVICES, &missing) == 0) { 1986 (void) printf(dgettext(TEXT_DOMAIN, 1987 "The devices below are missing or " 1988 "corrupted, use '-m' to import the pool " 1989 "anyway:\n")); 1990 print_vdev_tree(hdl, NULL, missing, 2); 1991 (void) printf("\n"); 1992 } 1993 (void) zpool_standard_error(hdl, error, desc); 1994 break; 1995 1996 case EEXIST: 1997 (void) zpool_standard_error(hdl, error, desc); 1998 break; 1999 case ENAMETOOLONG: 2000 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2001 "new name of at least one dataset is longer than " 2002 "the maximum allowable length")); 2003 (void) zfs_error(hdl, EZFS_NAMETOOLONG, desc); 2004 break; 2005 default: 2006 (void) zpool_standard_error(hdl, error, desc); 2007 zpool_explain_recover(hdl, 2008 newname ? origname : thename, -error, nv); 2009 break; 2010 } 2011 2012 nvlist_free(nv); 2013 ret = -1; 2014 } else { 2015 zpool_handle_t *zhp; 2016 2017 /* 2018 * This should never fail, but play it safe anyway. 2019 */ 2020 if (zpool_open_silent(hdl, thename, &zhp) != 0) 2021 ret = -1; 2022 else if (zhp != NULL) 2023 zpool_close(zhp); 2024 if (policy.zlp_rewind & 2025 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) { 2026 zpool_rewind_exclaim(hdl, newname ? origname : thename, 2027 ((policy.zlp_rewind & ZPOOL_TRY_REWIND) != 0), nv); 2028 } 2029 nvlist_free(nv); 2030 return (0); 2031 } 2032 2033 return (ret); 2034 } 2035 2036 /* 2037 * Translate vdev names to guids. If a vdev_path is determined to be 2038 * unsuitable then a vd_errlist is allocated and the vdev path and errno 2039 * are added to it. 2040 */ 2041 static int 2042 zpool_translate_vdev_guids(zpool_handle_t *zhp, nvlist_t *vds, 2043 nvlist_t *vdev_guids, nvlist_t *guids_to_paths, nvlist_t **vd_errlist) 2044 { 2045 nvlist_t *errlist = NULL; 2046 int error = 0; 2047 2048 for (nvpair_t *elem = nvlist_next_nvpair(vds, NULL); elem != NULL; 2049 elem = nvlist_next_nvpair(vds, elem)) { 2050 boolean_t spare, cache; 2051 2052 char *vd_path = nvpair_name(elem); 2053 nvlist_t *tgt = zpool_find_vdev(zhp, vd_path, &spare, &cache, 2054 NULL); 2055 2056 if ((tgt == NULL) || cache || spare) { 2057 if (errlist == NULL) { 2058 errlist = fnvlist_alloc(); 2059 error = EINVAL; 2060 } 2061 2062 uint64_t err = (tgt == NULL) ? EZFS_NODEVICE : 2063 (spare ? EZFS_ISSPARE : EZFS_ISL2CACHE); 2064 fnvlist_add_int64(errlist, vd_path, err); 2065 continue; 2066 } 2067 2068 uint64_t guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID); 2069 fnvlist_add_uint64(vdev_guids, vd_path, guid); 2070 2071 char msg[MAXNAMELEN]; 2072 (void) snprintf(msg, sizeof (msg), "%llu", (u_longlong_t)guid); 2073 fnvlist_add_string(guids_to_paths, msg, vd_path); 2074 } 2075 2076 if (error != 0) { 2077 verify(errlist != NULL); 2078 if (vd_errlist != NULL) 2079 *vd_errlist = errlist; 2080 else 2081 fnvlist_free(errlist); 2082 } 2083 2084 return (error); 2085 } 2086 2087 /* 2088 * Scan the pool. 2089 */ 2090 int 2091 zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func, pool_scrub_cmd_t cmd) 2092 { 2093 zfs_cmd_t zc = { 0 }; 2094 char msg[1024]; 2095 int err; 2096 libzfs_handle_t *hdl = zhp->zpool_hdl; 2097 2098 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2099 zc.zc_cookie = func; 2100 zc.zc_flags = cmd; 2101 2102 if (zfs_ioctl(hdl, ZFS_IOC_POOL_SCAN, &zc) == 0) 2103 return (0); 2104 2105 err = errno; 2106 2107 /* ECANCELED on a scrub means we resumed a paused scrub */ 2108 if (err == ECANCELED && func == POOL_SCAN_SCRUB && 2109 cmd == POOL_SCRUB_NORMAL) 2110 return (0); 2111 2112 if (err == ENOENT && func != POOL_SCAN_NONE && cmd == POOL_SCRUB_NORMAL) 2113 return (0); 2114 2115 if (func == POOL_SCAN_SCRUB) { 2116 if (cmd == POOL_SCRUB_PAUSE) { 2117 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 2118 "cannot pause scrubbing %s"), zc.zc_name); 2119 } else { 2120 assert(cmd == POOL_SCRUB_NORMAL); 2121 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 2122 "cannot scrub %s"), zc.zc_name); 2123 } 2124 } else if (func == POOL_SCAN_RESILVER) { 2125 assert(cmd == POOL_SCRUB_NORMAL); 2126 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 2127 "cannot restart resilver on %s"), zc.zc_name); 2128 } else if (func == POOL_SCAN_NONE) { 2129 (void) snprintf(msg, sizeof (msg), 2130 dgettext(TEXT_DOMAIN, "cannot cancel scrubbing %s"), 2131 zc.zc_name); 2132 } else { 2133 assert(!"unexpected result"); 2134 } 2135 2136 if (err == EBUSY) { 2137 nvlist_t *nvroot; 2138 pool_scan_stat_t *ps = NULL; 2139 uint_t psc; 2140 2141 verify(nvlist_lookup_nvlist(zhp->zpool_config, 2142 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 2143 (void) nvlist_lookup_uint64_array(nvroot, 2144 ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &psc); 2145 if (ps && ps->pss_func == POOL_SCAN_SCRUB) { 2146 if (cmd == POOL_SCRUB_PAUSE) 2147 return (zfs_error(hdl, EZFS_SCRUB_PAUSED, msg)); 2148 else 2149 return (zfs_error(hdl, EZFS_SCRUBBING, msg)); 2150 } else { 2151 return (zfs_error(hdl, EZFS_RESILVERING, msg)); 2152 } 2153 } else if (err == ENOENT) { 2154 return (zfs_error(hdl, EZFS_NO_SCRUB, msg)); 2155 } else if (err == ENOTSUP && func == POOL_SCAN_RESILVER) { 2156 return (zfs_error(hdl, EZFS_NO_RESILVER_DEFER, msg)); 2157 } else { 2158 return (zpool_standard_error(hdl, err, msg)); 2159 } 2160 } 2161 2162 static int 2163 xlate_init_err(int err) 2164 { 2165 switch (err) { 2166 case ENODEV: 2167 return (EZFS_NODEVICE); 2168 case EINVAL: 2169 case EROFS: 2170 return (EZFS_BADDEV); 2171 case EBUSY: 2172 return (EZFS_INITIALIZING); 2173 case ESRCH: 2174 return (EZFS_NO_INITIALIZE); 2175 } 2176 return (err); 2177 } 2178 2179 /* 2180 * Begin, suspend, or cancel the initialization (initializing of all free 2181 * blocks) for the given vdevs in the given pool. 2182 */ 2183 int 2184 zpool_initialize(zpool_handle_t *zhp, pool_initialize_func_t cmd_type, 2185 nvlist_t *vds) 2186 { 2187 char msg[1024]; 2188 int err; 2189 2190 nvlist_t *vdev_guids = fnvlist_alloc(); 2191 nvlist_t *guids_to_paths = fnvlist_alloc(); 2192 nvlist_t *vd_errlist = NULL; 2193 nvlist_t *errlist; 2194 nvpair_t *elem; 2195 2196 err = zpool_translate_vdev_guids(zhp, vds, vdev_guids, 2197 guids_to_paths, &vd_errlist); 2198 2199 if (err == 0) { 2200 err = lzc_initialize(zhp->zpool_name, cmd_type, 2201 vdev_guids, &errlist); 2202 if (err == 0) { 2203 fnvlist_free(vdev_guids); 2204 fnvlist_free(guids_to_paths); 2205 return (0); 2206 } 2207 2208 if (errlist != NULL) { 2209 vd_errlist = fnvlist_lookup_nvlist(errlist, 2210 ZPOOL_INITIALIZE_VDEVS); 2211 } 2212 2213 (void) snprintf(msg, sizeof (msg), 2214 dgettext(TEXT_DOMAIN, "operation failed")); 2215 } else { 2216 verify(vd_errlist != NULL); 2217 } 2218 2219 for (elem = nvlist_next_nvpair(vd_errlist, NULL); elem != NULL; 2220 elem = nvlist_next_nvpair(vd_errlist, elem)) { 2221 int64_t vd_error = xlate_init_err(fnvpair_value_int64(elem)); 2222 char *path; 2223 2224 if (nvlist_lookup_string(guids_to_paths, nvpair_name(elem), 2225 &path) != 0) 2226 path = nvpair_name(elem); 2227 2228 (void) zfs_error_fmt(zhp->zpool_hdl, vd_error, 2229 "cannot initialize '%s'", path); 2230 } 2231 2232 fnvlist_free(vdev_guids); 2233 fnvlist_free(guids_to_paths); 2234 2235 if (vd_errlist != NULL) { 2236 fnvlist_free(vd_errlist); 2237 return (-1); 2238 } 2239 2240 return (zpool_standard_error(zhp->zpool_hdl, err, msg)); 2241 } 2242 2243 static int 2244 xlate_trim_err(int err) 2245 { 2246 switch (err) { 2247 case ENODEV: 2248 return (EZFS_NODEVICE); 2249 case EINVAL: 2250 case EROFS: 2251 return (EZFS_BADDEV); 2252 case EBUSY: 2253 return (EZFS_TRIMMING); 2254 case ESRCH: 2255 return (EZFS_NO_TRIM); 2256 case EOPNOTSUPP: 2257 return (EZFS_TRIM_NOTSUP); 2258 } 2259 return (err); 2260 } 2261 2262 /* 2263 * Begin, suspend, or cancel the TRIM (discarding of all free blocks) for 2264 * the given vdevs in the given pool. 2265 */ 2266 int 2267 zpool_trim(zpool_handle_t *zhp, pool_trim_func_t cmd_type, nvlist_t *vds, 2268 trimflags_t *trim_flags) 2269 { 2270 char msg[1024]; 2271 int err; 2272 2273 nvlist_t *vdev_guids = fnvlist_alloc(); 2274 nvlist_t *guids_to_paths = fnvlist_alloc(); 2275 nvlist_t *vd_errlist = NULL; 2276 nvlist_t *errlist; 2277 nvpair_t *elem; 2278 2279 err = zpool_translate_vdev_guids(zhp, vds, vdev_guids, 2280 guids_to_paths, &vd_errlist); 2281 if (err == 0) { 2282 err = lzc_trim(zhp->zpool_name, cmd_type, trim_flags->rate, 2283 trim_flags->secure, vdev_guids, &errlist); 2284 if (err == 0) { 2285 fnvlist_free(vdev_guids); 2286 fnvlist_free(guids_to_paths); 2287 return (0); 2288 } 2289 2290 if (errlist != NULL) { 2291 vd_errlist = fnvlist_lookup_nvlist(errlist, 2292 ZPOOL_TRIM_VDEVS); 2293 } 2294 2295 (void) snprintf(msg, sizeof (msg), 2296 dgettext(TEXT_DOMAIN, "operation failed")); 2297 } else { 2298 verify(vd_errlist != NULL); 2299 } 2300 2301 for (elem = nvlist_next_nvpair(vd_errlist, NULL); 2302 elem != NULL; elem = nvlist_next_nvpair(vd_errlist, elem)) { 2303 int64_t vd_error = xlate_trim_err(fnvpair_value_int64(elem)); 2304 char *path; 2305 /* 2306 * If only the pool was specified, and it was not a secure 2307 * trim then suppress warnings for individual vdevs which 2308 * do not support trimming. 2309 */ 2310 if (vd_error == EZFS_TRIM_NOTSUP && 2311 trim_flags->fullpool && 2312 !trim_flags->secure) { 2313 continue; 2314 } 2315 2316 if (nvlist_lookup_string(guids_to_paths, nvpair_name(elem), 2317 &path) != 0) 2318 path = nvpair_name(elem); 2319 2320 (void) zfs_error_fmt(zhp->zpool_hdl, vd_error, 2321 "cannot trim '%s'", path); 2322 } 2323 2324 fnvlist_free(vdev_guids); 2325 fnvlist_free(guids_to_paths); 2326 2327 if (vd_errlist != NULL) { 2328 fnvlist_free(vd_errlist); 2329 return (-1); 2330 } 2331 2332 return (zpool_standard_error(zhp->zpool_hdl, err, msg)); 2333 } 2334 2335 /* 2336 * This provides a very minimal check whether a given string is likely a 2337 * c#t#d# style string. Users of this are expected to do their own 2338 * verification of the s# part. 2339 */ 2340 #define CTD_CHECK(str) (str && str[0] == 'c' && isdigit(str[1])) 2341 2342 /* 2343 * More elaborate version for ones which may start with "/dev/dsk/" 2344 * and the like. 2345 */ 2346 static int 2347 ctd_check_path(char *str) 2348 { 2349 /* 2350 * If it starts with a slash, check the last component. 2351 */ 2352 if (str && str[0] == '/') { 2353 char *tmp = strrchr(str, '/'); 2354 2355 /* 2356 * If it ends in "/old", check the second-to-last 2357 * component of the string instead. 2358 */ 2359 if (tmp != str && strcmp(tmp, "/old") == 0) { 2360 for (tmp--; *tmp != '/'; tmp--) 2361 ; 2362 } 2363 str = tmp + 1; 2364 } 2365 return (CTD_CHECK(str)); 2366 } 2367 2368 /* 2369 * Find a vdev that matches the search criteria specified. We use the 2370 * the nvpair name to determine how we should look for the device. 2371 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL 2372 * spare; but FALSE if its an INUSE spare. 2373 */ 2374 static nvlist_t * 2375 vdev_to_nvlist_iter(nvlist_t *nv, nvlist_t *search, boolean_t *avail_spare, 2376 boolean_t *l2cache, boolean_t *log) 2377 { 2378 uint_t c, children; 2379 nvlist_t **child; 2380 nvlist_t *ret; 2381 uint64_t is_log; 2382 char *srchkey; 2383 nvpair_t *pair = nvlist_next_nvpair(search, NULL); 2384 2385 /* Nothing to look for */ 2386 if (search == NULL || pair == NULL) 2387 return (NULL); 2388 2389 /* Obtain the key we will use to search */ 2390 srchkey = nvpair_name(pair); 2391 2392 switch (nvpair_type(pair)) { 2393 case DATA_TYPE_UINT64: 2394 if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) { 2395 uint64_t srchval, theguid; 2396 2397 verify(nvpair_value_uint64(pair, &srchval) == 0); 2398 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, 2399 &theguid) == 0); 2400 if (theguid == srchval) 2401 return (nv); 2402 } 2403 break; 2404 2405 case DATA_TYPE_STRING: { 2406 char *srchval, *val; 2407 2408 verify(nvpair_value_string(pair, &srchval) == 0); 2409 if (nvlist_lookup_string(nv, srchkey, &val) != 0) 2410 break; 2411 2412 /* 2413 * Search for the requested value. Special cases: 2414 * 2415 * - ZPOOL_CONFIG_PATH for whole disk entries. To support 2416 * UEFI boot, these end in "s0" or "s0/old" or "s1" or 2417 * "s1/old". The "s0" or "s1" part is hidden from the user, 2418 * but included in the string, so this matches around it. 2419 * - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE). 2420 * 2421 * Otherwise, all other searches are simple string compares. 2422 */ 2423 if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0 && 2424 ctd_check_path(val)) { 2425 uint64_t wholedisk = 0; 2426 2427 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, 2428 &wholedisk); 2429 if (wholedisk) { 2430 int slen = strlen(srchval); 2431 int vlen = strlen(val); 2432 2433 if (slen != vlen - 2) 2434 break; 2435 2436 /* 2437 * make_leaf_vdev() should only set 2438 * wholedisk for ZPOOL_CONFIG_PATHs which 2439 * will include "/dev/dsk/", giving plenty of 2440 * room for the indices used next. 2441 */ 2442 ASSERT(vlen >= 6); 2443 2444 /* 2445 * strings identical except trailing "s0" 2446 */ 2447 if ((strcmp(&val[vlen - 2], "s0") == 0 || 2448 strcmp(&val[vlen - 2], "s1") == 0) && 2449 strncmp(srchval, val, slen) == 0) 2450 return (nv); 2451 2452 /* 2453 * strings identical except trailing "s0/old" 2454 */ 2455 if ((strcmp(&val[vlen - 6], "s0/old") == 0 || 2456 strcmp(&val[vlen - 6], "s1/old") == 0) && 2457 strcmp(&srchval[slen - 4], "/old") == 0 && 2458 strncmp(srchval, val, slen - 4) == 0) 2459 return (nv); 2460 2461 break; 2462 } 2463 } else if (strcmp(srchkey, ZPOOL_CONFIG_TYPE) == 0 && val) { 2464 char *type, *idx, *end, *p; 2465 uint64_t id, vdev_id; 2466 2467 /* 2468 * Determine our vdev type, keeping in mind 2469 * that the srchval is composed of a type and 2470 * vdev id pair (i.e. mirror-4). 2471 */ 2472 if ((type = strdup(srchval)) == NULL) 2473 return (NULL); 2474 2475 if ((p = strrchr(type, '-')) == NULL) { 2476 free(type); 2477 break; 2478 } 2479 idx = p + 1; 2480 *p = '\0'; 2481 2482 /* 2483 * If the types don't match then keep looking. 2484 */ 2485 if (strncmp(val, type, strlen(val)) != 0) { 2486 free(type); 2487 break; 2488 } 2489 2490 verify(zpool_vdev_is_interior(type)); 2491 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID, 2492 &id) == 0); 2493 2494 errno = 0; 2495 vdev_id = strtoull(idx, &end, 10); 2496 2497 free(type); 2498 if (errno != 0) 2499 return (NULL); 2500 2501 /* 2502 * Now verify that we have the correct vdev id. 2503 */ 2504 if (vdev_id == id) 2505 return (nv); 2506 } 2507 2508 /* 2509 * Common case 2510 */ 2511 if (strcmp(srchval, val) == 0) 2512 return (nv); 2513 break; 2514 } 2515 2516 default: 2517 break; 2518 } 2519 2520 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, 2521 &child, &children) != 0) 2522 return (NULL); 2523 2524 for (c = 0; c < children; c++) { 2525 if ((ret = vdev_to_nvlist_iter(child[c], search, 2526 avail_spare, l2cache, NULL)) != NULL) { 2527 /* 2528 * The 'is_log' value is only set for the toplevel 2529 * vdev, not the leaf vdevs. So we always lookup the 2530 * log device from the root of the vdev tree (where 2531 * 'log' is non-NULL). 2532 */ 2533 if (log != NULL && 2534 nvlist_lookup_uint64(child[c], 2535 ZPOOL_CONFIG_IS_LOG, &is_log) == 0 && 2536 is_log) { 2537 *log = B_TRUE; 2538 } 2539 return (ret); 2540 } 2541 } 2542 2543 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES, 2544 &child, &children) == 0) { 2545 for (c = 0; c < children; c++) { 2546 if ((ret = vdev_to_nvlist_iter(child[c], search, 2547 avail_spare, l2cache, NULL)) != NULL) { 2548 *avail_spare = B_TRUE; 2549 return (ret); 2550 } 2551 } 2552 } 2553 2554 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE, 2555 &child, &children) == 0) { 2556 for (c = 0; c < children; c++) { 2557 if ((ret = vdev_to_nvlist_iter(child[c], search, 2558 avail_spare, l2cache, NULL)) != NULL) { 2559 *l2cache = B_TRUE; 2560 return (ret); 2561 } 2562 } 2563 } 2564 2565 return (NULL); 2566 } 2567 2568 /* 2569 * Given a physical path (minus the "/devices" prefix), find the 2570 * associated vdev. 2571 */ 2572 nvlist_t * 2573 zpool_find_vdev_by_physpath(zpool_handle_t *zhp, const char *ppath, 2574 boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log) 2575 { 2576 nvlist_t *search, *nvroot, *ret; 2577 2578 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2579 verify(nvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH, ppath) == 0); 2580 2581 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE, 2582 &nvroot) == 0); 2583 2584 *avail_spare = B_FALSE; 2585 *l2cache = B_FALSE; 2586 if (log != NULL) 2587 *log = B_FALSE; 2588 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log); 2589 nvlist_free(search); 2590 2591 return (ret); 2592 } 2593 2594 /* 2595 * Determine if we have an "interior" top-level vdev (i.e mirror/raidz). 2596 */ 2597 static boolean_t 2598 zpool_vdev_is_interior(const char *name) 2599 { 2600 if (strncmp(name, VDEV_TYPE_RAIDZ, strlen(VDEV_TYPE_RAIDZ)) == 0 || 2601 strncmp(name, VDEV_TYPE_SPARE, strlen(VDEV_TYPE_SPARE)) == 0 || 2602 strncmp(name, 2603 VDEV_TYPE_REPLACING, strlen(VDEV_TYPE_REPLACING)) == 0 || 2604 strncmp(name, VDEV_TYPE_MIRROR, strlen(VDEV_TYPE_MIRROR)) == 0) 2605 return (B_TRUE); 2606 return (B_FALSE); 2607 } 2608 2609 nvlist_t * 2610 zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare, 2611 boolean_t *l2cache, boolean_t *log) 2612 { 2613 char buf[MAXPATHLEN]; 2614 char *end; 2615 nvlist_t *nvroot, *search, *ret; 2616 uint64_t guid; 2617 2618 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2619 2620 guid = strtoull(path, &end, 10); 2621 if (guid != 0 && *end == '\0') { 2622 verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0); 2623 } else if (zpool_vdev_is_interior(path)) { 2624 verify(nvlist_add_string(search, ZPOOL_CONFIG_TYPE, path) == 0); 2625 } else if (path[0] != '/') { 2626 (void) snprintf(buf, sizeof (buf), "%s/%s", ZFS_DISK_ROOT, 2627 path); 2628 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, buf) == 0); 2629 } else { 2630 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, path) == 0); 2631 } 2632 2633 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE, 2634 &nvroot) == 0); 2635 2636 *avail_spare = B_FALSE; 2637 *l2cache = B_FALSE; 2638 if (log != NULL) 2639 *log = B_FALSE; 2640 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log); 2641 nvlist_free(search); 2642 2643 return (ret); 2644 } 2645 2646 static int 2647 vdev_is_online(nvlist_t *nv) 2648 { 2649 uint64_t ival; 2650 2651 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE, &ival) == 0 || 2652 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED, &ival) == 0 || 2653 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED, &ival) == 0) 2654 return (0); 2655 2656 return (1); 2657 } 2658 2659 /* 2660 * Helper function for zpool_get_physpaths(). 2661 */ 2662 static int 2663 vdev_get_one_physpath(nvlist_t *config, char *physpath, size_t physpath_size, 2664 size_t *bytes_written) 2665 { 2666 size_t bytes_left, pos, rsz; 2667 char *tmppath; 2668 const char *format; 2669 2670 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PHYS_PATH, 2671 &tmppath) != 0) 2672 return (EZFS_NODEVICE); 2673 2674 pos = *bytes_written; 2675 bytes_left = physpath_size - pos; 2676 format = (pos == 0) ? "%s" : " %s"; 2677 2678 rsz = snprintf(physpath + pos, bytes_left, format, tmppath); 2679 *bytes_written += rsz; 2680 2681 if (rsz >= bytes_left) { 2682 /* if physpath was not copied properly, clear it */ 2683 if (bytes_left != 0) { 2684 physpath[pos] = 0; 2685 } 2686 return (EZFS_NOSPC); 2687 } 2688 return (0); 2689 } 2690 2691 static int 2692 vdev_get_physpaths(nvlist_t *nv, char *physpath, size_t phypath_size, 2693 size_t *rsz, boolean_t is_spare) 2694 { 2695 char *type; 2696 int ret; 2697 2698 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0) 2699 return (EZFS_INVALCONFIG); 2700 2701 if (strcmp(type, VDEV_TYPE_DISK) == 0) { 2702 /* 2703 * An active spare device has ZPOOL_CONFIG_IS_SPARE set. 2704 * For a spare vdev, we only want to boot from the active 2705 * spare device. 2706 */ 2707 if (is_spare) { 2708 uint64_t spare = 0; 2709 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE, 2710 &spare); 2711 if (!spare) 2712 return (EZFS_INVALCONFIG); 2713 } 2714 2715 if (vdev_is_online(nv)) { 2716 if ((ret = vdev_get_one_physpath(nv, physpath, 2717 phypath_size, rsz)) != 0) 2718 return (ret); 2719 } 2720 } else if (strcmp(type, VDEV_TYPE_MIRROR) == 0 || 2721 strcmp(type, VDEV_TYPE_RAIDZ) == 0 || 2722 strcmp(type, VDEV_TYPE_REPLACING) == 0 || 2723 (is_spare = (strcmp(type, VDEV_TYPE_SPARE) == 0))) { 2724 nvlist_t **child; 2725 uint_t count; 2726 int i, ret; 2727 2728 if (nvlist_lookup_nvlist_array(nv, 2729 ZPOOL_CONFIG_CHILDREN, &child, &count) != 0) 2730 return (EZFS_INVALCONFIG); 2731 2732 for (i = 0; i < count; i++) { 2733 ret = vdev_get_physpaths(child[i], physpath, 2734 phypath_size, rsz, is_spare); 2735 if (ret == EZFS_NOSPC) 2736 return (ret); 2737 } 2738 } 2739 2740 return (EZFS_POOL_INVALARG); 2741 } 2742 2743 /* 2744 * Get phys_path for a root pool config. 2745 * Return 0 on success; non-zero on failure. 2746 */ 2747 static int 2748 zpool_get_config_physpath(nvlist_t *config, char *physpath, size_t phypath_size) 2749 { 2750 size_t rsz; 2751 nvlist_t *vdev_root; 2752 nvlist_t **child; 2753 uint_t count; 2754 char *type; 2755 2756 rsz = 0; 2757 2758 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 2759 &vdev_root) != 0) 2760 return (EZFS_INVALCONFIG); 2761 2762 if (nvlist_lookup_string(vdev_root, ZPOOL_CONFIG_TYPE, &type) != 0 || 2763 nvlist_lookup_nvlist_array(vdev_root, ZPOOL_CONFIG_CHILDREN, 2764 &child, &count) != 0) 2765 return (EZFS_INVALCONFIG); 2766 2767 /* 2768 * root pool can only have a single top-level vdev. 2769 */ 2770 if (strcmp(type, VDEV_TYPE_ROOT) != 0 || count != 1) 2771 return (EZFS_POOL_INVALARG); 2772 2773 (void) vdev_get_physpaths(child[0], physpath, phypath_size, &rsz, 2774 B_FALSE); 2775 2776 /* No online devices */ 2777 if (rsz == 0) 2778 return (EZFS_NODEVICE); 2779 2780 return (0); 2781 } 2782 2783 /* 2784 * Get phys_path for a root pool 2785 * Return 0 on success; non-zero on failure. 2786 */ 2787 int 2788 zpool_get_physpath(zpool_handle_t *zhp, char *physpath, size_t phypath_size) 2789 { 2790 return (zpool_get_config_physpath(zhp->zpool_config, physpath, 2791 phypath_size)); 2792 } 2793 2794 /* 2795 * If the device has being dynamically expanded then we need to relabel 2796 * the disk to use the new unallocated space. 2797 */ 2798 static int 2799 zpool_relabel_disk(libzfs_handle_t *hdl, const char *name, const char *msg) 2800 { 2801 char path[MAXPATHLEN]; 2802 enum dkio_state st; 2803 int fd, error; 2804 int (*_efi_use_whole_disk)(int); 2805 2806 if ((_efi_use_whole_disk = (int (*)(int))dlsym(RTLD_DEFAULT, 2807 "efi_use_whole_disk")) == NULL) 2808 return (-1); 2809 2810 (void) snprintf(path, sizeof (path), "%s/%s", ZFS_RDISK_ROOT, name); 2811 2812 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) { 2813 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot " 2814 "relabel '%s': unable to open device"), name); 2815 return (zfs_error(hdl, EZFS_OPENFAILED, msg)); 2816 } 2817 2818 /* 2819 * It's possible that we might encounter an error if the device 2820 * does not have any unallocated space left. If so, we simply 2821 * ignore that error and continue on. 2822 */ 2823 error = _efi_use_whole_disk(fd); 2824 if (error && error != VT_ENOSPC) { 2825 (void) close(fd); 2826 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot " 2827 "relabel '%s': unable to read disk capacity"), name); 2828 return (zfs_error(hdl, EZFS_NOCAP, msg)); 2829 } 2830 2831 /* 2832 * Writing a new EFI partition table to the disk will have marked 2833 * the geometry as needing re-validation. Before returning, force 2834 * it to be checked by querying the device state, otherwise the 2835 * subsequent vdev_reopen() will very likely fail to read the device 2836 * size, faulting the pool. 2837 */ 2838 st = DKIO_NONE; 2839 (void) ioctl(fd, DKIOCSTATE, (caddr_t)&st); 2840 2841 (void) close(fd); 2842 2843 return (0); 2844 } 2845 2846 /* 2847 * Bring the specified vdev online. The 'flags' parameter is a set of the 2848 * ZFS_ONLINE_* flags. 2849 */ 2850 int 2851 zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags, 2852 vdev_state_t *newstate) 2853 { 2854 zfs_cmd_t zc = { 0 }; 2855 char msg[1024]; 2856 char *pathname; 2857 nvlist_t *tgt; 2858 boolean_t avail_spare, l2cache, islog; 2859 libzfs_handle_t *hdl = zhp->zpool_hdl; 2860 int error; 2861 2862 if (flags & ZFS_ONLINE_EXPAND) { 2863 (void) snprintf(msg, sizeof (msg), 2864 dgettext(TEXT_DOMAIN, "cannot expand %s"), path); 2865 } else { 2866 (void) snprintf(msg, sizeof (msg), 2867 dgettext(TEXT_DOMAIN, "cannot online %s"), path); 2868 } 2869 2870 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2871 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 2872 &islog)) == NULL) 2873 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 2874 2875 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 2876 2877 if (avail_spare) 2878 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 2879 2880 if ((flags & ZFS_ONLINE_EXPAND || 2881 zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) && 2882 nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH, &pathname) == 0) { 2883 uint64_t wholedisk = 0; 2884 2885 (void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK, 2886 &wholedisk); 2887 2888 /* 2889 * XXX - L2ARC 1.0 devices can't support expansion. 2890 */ 2891 if (l2cache) { 2892 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2893 "cannot expand cache devices")); 2894 return (zfs_error(hdl, EZFS_VDEVNOTSUP, msg)); 2895 } 2896 2897 if (wholedisk) { 2898 pathname += strlen(ZFS_DISK_ROOT) + 1; 2899 error = zpool_relabel_disk(hdl, pathname, msg); 2900 if (error != 0) 2901 return (error); 2902 } 2903 } 2904 2905 zc.zc_cookie = VDEV_STATE_ONLINE; 2906 zc.zc_obj = flags; 2907 2908 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0) { 2909 if (errno == EINVAL) { 2910 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "was split " 2911 "from this pool into a new one. Use '%s' " 2912 "instead"), "zpool detach"); 2913 return (zfs_error(hdl, EZFS_POSTSPLIT_ONLINE, msg)); 2914 } 2915 return (zpool_standard_error(hdl, errno, msg)); 2916 } 2917 2918 *newstate = zc.zc_cookie; 2919 return (0); 2920 } 2921 2922 /* 2923 * Take the specified vdev offline 2924 */ 2925 int 2926 zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp) 2927 { 2928 zfs_cmd_t zc = { 0 }; 2929 char msg[1024]; 2930 nvlist_t *tgt; 2931 boolean_t avail_spare, l2cache; 2932 libzfs_handle_t *hdl = zhp->zpool_hdl; 2933 2934 (void) snprintf(msg, sizeof (msg), 2935 dgettext(TEXT_DOMAIN, "cannot offline %s"), path); 2936 2937 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2938 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 2939 NULL)) == NULL) 2940 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 2941 2942 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 2943 2944 if (avail_spare) 2945 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 2946 2947 zc.zc_cookie = VDEV_STATE_OFFLINE; 2948 zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0; 2949 2950 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0) 2951 return (0); 2952 2953 switch (errno) { 2954 case EBUSY: 2955 2956 /* 2957 * There are no other replicas of this device. 2958 */ 2959 return (zfs_error(hdl, EZFS_NOREPLICAS, msg)); 2960 2961 case EEXIST: 2962 /* 2963 * The log device has unplayed logs 2964 */ 2965 return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, msg)); 2966 2967 default: 2968 return (zpool_standard_error(hdl, errno, msg)); 2969 } 2970 } 2971 2972 /* 2973 * Mark the given vdev faulted. 2974 */ 2975 int 2976 zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux) 2977 { 2978 zfs_cmd_t zc = { 0 }; 2979 char msg[1024]; 2980 libzfs_handle_t *hdl = zhp->zpool_hdl; 2981 2982 (void) snprintf(msg, sizeof (msg), 2983 dgettext(TEXT_DOMAIN, "cannot fault %llu"), guid); 2984 2985 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2986 zc.zc_guid = guid; 2987 zc.zc_cookie = VDEV_STATE_FAULTED; 2988 zc.zc_obj = aux; 2989 2990 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0) 2991 return (0); 2992 2993 switch (errno) { 2994 case EBUSY: 2995 2996 /* 2997 * There are no other replicas of this device. 2998 */ 2999 return (zfs_error(hdl, EZFS_NOREPLICAS, msg)); 3000 3001 default: 3002 return (zpool_standard_error(hdl, errno, msg)); 3003 } 3004 3005 } 3006 3007 /* 3008 * Mark the given vdev degraded. 3009 */ 3010 int 3011 zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux) 3012 { 3013 zfs_cmd_t zc = { 0 }; 3014 char msg[1024]; 3015 libzfs_handle_t *hdl = zhp->zpool_hdl; 3016 3017 (void) snprintf(msg, sizeof (msg), 3018 dgettext(TEXT_DOMAIN, "cannot degrade %llu"), guid); 3019 3020 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3021 zc.zc_guid = guid; 3022 zc.zc_cookie = VDEV_STATE_DEGRADED; 3023 zc.zc_obj = aux; 3024 3025 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0) 3026 return (0); 3027 3028 return (zpool_standard_error(hdl, errno, msg)); 3029 } 3030 3031 /* 3032 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as 3033 * a hot spare. 3034 */ 3035 static boolean_t 3036 is_replacing_spare(nvlist_t *search, nvlist_t *tgt, int which) 3037 { 3038 nvlist_t **child; 3039 uint_t c, children; 3040 char *type; 3041 3042 if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child, 3043 &children) == 0) { 3044 verify(nvlist_lookup_string(search, ZPOOL_CONFIG_TYPE, 3045 &type) == 0); 3046 3047 if (strcmp(type, VDEV_TYPE_SPARE) == 0 && 3048 children == 2 && child[which] == tgt) 3049 return (B_TRUE); 3050 3051 for (c = 0; c < children; c++) 3052 if (is_replacing_spare(child[c], tgt, which)) 3053 return (B_TRUE); 3054 } 3055 3056 return (B_FALSE); 3057 } 3058 3059 /* 3060 * Attach new_disk (fully described by nvroot) to old_disk. 3061 * If 'replacing' is specified, the new disk will replace the old one. 3062 */ 3063 int 3064 zpool_vdev_attach(zpool_handle_t *zhp, 3065 const char *old_disk, const char *new_disk, nvlist_t *nvroot, int replacing) 3066 { 3067 zfs_cmd_t zc = { 0 }; 3068 char msg[1024]; 3069 int ret; 3070 nvlist_t *tgt, *newvd; 3071 boolean_t avail_spare, l2cache, islog; 3072 uint64_t val; 3073 char *newname; 3074 nvlist_t **child; 3075 uint_t children; 3076 nvlist_t *config_root; 3077 libzfs_handle_t *hdl = zhp->zpool_hdl; 3078 boolean_t rootpool = zpool_is_bootable(zhp); 3079 3080 if (replacing) 3081 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 3082 "cannot replace %s with %s"), old_disk, new_disk); 3083 else 3084 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 3085 "cannot attach %s to %s"), new_disk, old_disk); 3086 3087 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3088 if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache, 3089 &islog)) == NULL) 3090 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 3091 3092 if (avail_spare) 3093 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 3094 3095 if (l2cache) 3096 return (zfs_error(hdl, EZFS_ISL2CACHE, msg)); 3097 3098 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 3099 zc.zc_cookie = replacing; 3100 3101 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, 3102 &child, &children) != 0 || children != 1) { 3103 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3104 "new device must be a single disk")); 3105 return (zfs_error(hdl, EZFS_INVALCONFIG, msg)); 3106 } 3107 3108 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL), 3109 ZPOOL_CONFIG_VDEV_TREE, &config_root) == 0); 3110 3111 if ((newname = zpool_vdev_name(NULL, NULL, child[0], 0)) == NULL) 3112 return (-1); 3113 3114 newvd = zpool_find_vdev(zhp, newname, &avail_spare, &l2cache, NULL); 3115 /* 3116 * If the target is a hot spare that has been swapped in, we can only 3117 * replace it with another hot spare. 3118 */ 3119 if (replacing && 3120 nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 && 3121 (newvd == NULL || !avail_spare) && 3122 is_replacing_spare(config_root, tgt, 1)) { 3123 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3124 "can only be replaced by another hot spare")); 3125 free(newname); 3126 return (zfs_error(hdl, EZFS_BADTARGET, msg)); 3127 } 3128 3129 free(newname); 3130 3131 if (replacing && avail_spare && !vdev_is_online(newvd)) { 3132 (void) zpool_standard_error(hdl, ENXIO, msg); 3133 return (-1); 3134 } 3135 3136 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0) 3137 return (-1); 3138 3139 ret = zfs_ioctl(hdl, ZFS_IOC_VDEV_ATTACH, &zc); 3140 3141 zcmd_free_nvlists(&zc); 3142 3143 if (ret == 0) { 3144 if (rootpool) { 3145 /* 3146 * XXX need a better way to prevent user from 3147 * booting up a half-baked vdev. 3148 */ 3149 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Make " 3150 "sure to wait until resilver is done " 3151 "before rebooting.\n")); 3152 } 3153 return (0); 3154 } 3155 3156 switch (errno) { 3157 case ENOTSUP: 3158 /* 3159 * Can't attach to or replace this type of vdev. 3160 */ 3161 if (replacing) { 3162 uint64_t version = zpool_get_prop_int(zhp, 3163 ZPOOL_PROP_VERSION, NULL); 3164 3165 if (islog) 3166 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3167 "cannot replace a log with a spare")); 3168 else if (version >= SPA_VERSION_MULTI_REPLACE) 3169 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3170 "already in replacing/spare config; wait " 3171 "for completion or use 'zpool detach'")); 3172 else 3173 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3174 "cannot replace a replacing device")); 3175 } else { 3176 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3177 "can only attach to mirrors and top-level " 3178 "disks")); 3179 } 3180 (void) zfs_error(hdl, EZFS_BADTARGET, msg); 3181 break; 3182 3183 case EINVAL: 3184 /* 3185 * The new device must be a single disk. 3186 */ 3187 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3188 "new device must be a single disk")); 3189 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg); 3190 break; 3191 3192 case EBUSY: 3193 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy, " 3194 "or device removal is in progress"), 3195 new_disk); 3196 (void) zfs_error(hdl, EZFS_BADDEV, msg); 3197 break; 3198 3199 case EOVERFLOW: 3200 /* 3201 * The new device is too small. 3202 */ 3203 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3204 "device is too small")); 3205 (void) zfs_error(hdl, EZFS_BADDEV, msg); 3206 break; 3207 3208 case EDOM: 3209 /* 3210 * The new device has a different optimal sector size. 3211 */ 3212 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3213 "new device has a different optimal sector size; use the " 3214 "option '-o ashift=N' to override the optimal size")); 3215 (void) zfs_error(hdl, EZFS_BADDEV, msg); 3216 break; 3217 3218 case ENAMETOOLONG: 3219 /* 3220 * The resulting top-level vdev spec won't fit in the label. 3221 */ 3222 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, msg); 3223 break; 3224 3225 default: 3226 (void) zpool_standard_error(hdl, errno, msg); 3227 } 3228 3229 return (-1); 3230 } 3231 3232 /* 3233 * Detach the specified device. 3234 */ 3235 int 3236 zpool_vdev_detach(zpool_handle_t *zhp, const char *path) 3237 { 3238 zfs_cmd_t zc = { 0 }; 3239 char msg[1024]; 3240 nvlist_t *tgt; 3241 boolean_t avail_spare, l2cache; 3242 libzfs_handle_t *hdl = zhp->zpool_hdl; 3243 3244 (void) snprintf(msg, sizeof (msg), 3245 dgettext(TEXT_DOMAIN, "cannot detach %s"), path); 3246 3247 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3248 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 3249 NULL)) == NULL) 3250 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 3251 3252 if (avail_spare) 3253 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 3254 3255 if (l2cache) 3256 return (zfs_error(hdl, EZFS_ISL2CACHE, msg)); 3257 3258 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 3259 3260 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0) 3261 return (0); 3262 3263 switch (errno) { 3264 3265 case ENOTSUP: 3266 /* 3267 * Can't detach from this type of vdev. 3268 */ 3269 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only " 3270 "applicable to mirror and replacing vdevs")); 3271 (void) zfs_error(hdl, EZFS_BADTARGET, msg); 3272 break; 3273 3274 case EBUSY: 3275 /* 3276 * There are no other replicas of this device. 3277 */ 3278 (void) zfs_error(hdl, EZFS_NOREPLICAS, msg); 3279 break; 3280 3281 default: 3282 (void) zpool_standard_error(hdl, errno, msg); 3283 } 3284 3285 return (-1); 3286 } 3287 3288 /* 3289 * Find a mirror vdev in the source nvlist. 3290 * 3291 * The mchild array contains a list of disks in one of the top-level mirrors 3292 * of the source pool. The schild array contains a list of disks that the 3293 * user specified on the command line. We loop over the mchild array to 3294 * see if any entry in the schild array matches. 3295 * 3296 * If a disk in the mchild array is found in the schild array, we return 3297 * the index of that entry. Otherwise we return -1. 3298 */ 3299 static int 3300 find_vdev_entry(zpool_handle_t *zhp, nvlist_t **mchild, uint_t mchildren, 3301 nvlist_t **schild, uint_t schildren) 3302 { 3303 uint_t mc; 3304 3305 for (mc = 0; mc < mchildren; mc++) { 3306 uint_t sc; 3307 char *mpath = zpool_vdev_name(zhp->zpool_hdl, zhp, 3308 mchild[mc], 0); 3309 3310 for (sc = 0; sc < schildren; sc++) { 3311 char *spath = zpool_vdev_name(zhp->zpool_hdl, zhp, 3312 schild[sc], 0); 3313 boolean_t result = (strcmp(mpath, spath) == 0); 3314 3315 free(spath); 3316 if (result) { 3317 free(mpath); 3318 return (mc); 3319 } 3320 } 3321 3322 free(mpath); 3323 } 3324 3325 return (-1); 3326 } 3327 3328 /* 3329 * Split a mirror pool. If newroot points to null, then a new nvlist 3330 * is generated and it is the responsibility of the caller to free it. 3331 */ 3332 int 3333 zpool_vdev_split(zpool_handle_t *zhp, char *newname, nvlist_t **newroot, 3334 nvlist_t *props, splitflags_t flags) 3335 { 3336 zfs_cmd_t zc = { 0 }; 3337 char msg[1024]; 3338 nvlist_t *tree, *config, **child, **newchild, *newconfig = NULL; 3339 nvlist_t **varray = NULL, *zc_props = NULL; 3340 uint_t c, children, newchildren, lastlog = 0, vcount, found = 0; 3341 libzfs_handle_t *hdl = zhp->zpool_hdl; 3342 uint64_t vers; 3343 boolean_t freelist = B_FALSE, memory_err = B_TRUE; 3344 int retval = 0; 3345 3346 (void) snprintf(msg, sizeof (msg), 3347 dgettext(TEXT_DOMAIN, "Unable to split %s"), zhp->zpool_name); 3348 3349 if (!zpool_name_valid(hdl, B_FALSE, newname)) 3350 return (zfs_error(hdl, EZFS_INVALIDNAME, msg)); 3351 3352 if ((config = zpool_get_config(zhp, NULL)) == NULL) { 3353 (void) fprintf(stderr, gettext("Internal error: unable to " 3354 "retrieve pool configuration\n")); 3355 return (-1); 3356 } 3357 3358 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree) 3359 == 0); 3360 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &vers) == 0); 3361 3362 if (props) { 3363 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE }; 3364 if ((zc_props = zpool_valid_proplist(hdl, zhp->zpool_name, 3365 props, vers, flags, msg)) == NULL) 3366 return (-1); 3367 } 3368 3369 if (nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child, 3370 &children) != 0) { 3371 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3372 "Source pool is missing vdev tree")); 3373 nvlist_free(zc_props); 3374 return (-1); 3375 } 3376 3377 varray = zfs_alloc(hdl, children * sizeof (nvlist_t *)); 3378 vcount = 0; 3379 3380 if (*newroot == NULL || 3381 nvlist_lookup_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, 3382 &newchild, &newchildren) != 0) 3383 newchildren = 0; 3384 3385 for (c = 0; c < children; c++) { 3386 uint64_t is_log = B_FALSE, is_hole = B_FALSE; 3387 char *type; 3388 nvlist_t **mchild, *vdev; 3389 uint_t mchildren; 3390 int entry; 3391 3392 /* 3393 * Unlike cache & spares, slogs are stored in the 3394 * ZPOOL_CONFIG_CHILDREN array. We filter them out here. 3395 */ 3396 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG, 3397 &is_log); 3398 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE, 3399 &is_hole); 3400 if (is_log || is_hole) { 3401 /* 3402 * Create a hole vdev and put it in the config. 3403 */ 3404 if (nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) != 0) 3405 goto out; 3406 if (nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE, 3407 VDEV_TYPE_HOLE) != 0) 3408 goto out; 3409 if (nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_HOLE, 3410 1) != 0) 3411 goto out; 3412 if (lastlog == 0) 3413 lastlog = vcount; 3414 varray[vcount++] = vdev; 3415 continue; 3416 } 3417 lastlog = 0; 3418 verify(nvlist_lookup_string(child[c], ZPOOL_CONFIG_TYPE, &type) 3419 == 0); 3420 if (strcmp(type, VDEV_TYPE_MIRROR) != 0) { 3421 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3422 "Source pool must be composed only of mirrors\n")); 3423 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg); 3424 goto out; 3425 } 3426 3427 verify(nvlist_lookup_nvlist_array(child[c], 3428 ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0); 3429 3430 /* find or add an entry for this top-level vdev */ 3431 if (newchildren > 0 && 3432 (entry = find_vdev_entry(zhp, mchild, mchildren, 3433 newchild, newchildren)) >= 0) { 3434 /* We found a disk that the user specified. */ 3435 vdev = mchild[entry]; 3436 ++found; 3437 } else { 3438 /* User didn't specify a disk for this vdev. */ 3439 vdev = mchild[mchildren - 1]; 3440 } 3441 3442 if (nvlist_dup(vdev, &varray[vcount++], 0) != 0) 3443 goto out; 3444 } 3445 3446 /* did we find every disk the user specified? */ 3447 if (found != newchildren) { 3448 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "Device list must " 3449 "include at most one disk from each mirror")); 3450 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg); 3451 goto out; 3452 } 3453 3454 /* Prepare the nvlist for populating. */ 3455 if (*newroot == NULL) { 3456 if (nvlist_alloc(newroot, NV_UNIQUE_NAME, 0) != 0) 3457 goto out; 3458 freelist = B_TRUE; 3459 if (nvlist_add_string(*newroot, ZPOOL_CONFIG_TYPE, 3460 VDEV_TYPE_ROOT) != 0) 3461 goto out; 3462 } else { 3463 verify(nvlist_remove_all(*newroot, ZPOOL_CONFIG_CHILDREN) == 0); 3464 } 3465 3466 /* Add all the children we found */ 3467 if (nvlist_add_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, varray, 3468 lastlog == 0 ? vcount : lastlog) != 0) 3469 goto out; 3470 3471 /* 3472 * If we're just doing a dry run, exit now with success. 3473 */ 3474 if (flags.dryrun) { 3475 memory_err = B_FALSE; 3476 freelist = B_FALSE; 3477 goto out; 3478 } 3479 3480 /* now build up the config list & call the ioctl */ 3481 if (nvlist_alloc(&newconfig, NV_UNIQUE_NAME, 0) != 0) 3482 goto out; 3483 3484 if (nvlist_add_nvlist(newconfig, 3485 ZPOOL_CONFIG_VDEV_TREE, *newroot) != 0 || 3486 nvlist_add_string(newconfig, 3487 ZPOOL_CONFIG_POOL_NAME, newname) != 0 || 3488 nvlist_add_uint64(newconfig, ZPOOL_CONFIG_VERSION, vers) != 0) 3489 goto out; 3490 3491 /* 3492 * The new pool is automatically part of the namespace unless we 3493 * explicitly export it. 3494 */ 3495 if (!flags.import) 3496 zc.zc_cookie = ZPOOL_EXPORT_AFTER_SPLIT; 3497 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3498 (void) strlcpy(zc.zc_string, newname, sizeof (zc.zc_string)); 3499 if (zcmd_write_conf_nvlist(hdl, &zc, newconfig) != 0) 3500 goto out; 3501 if (zc_props != NULL && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0) 3502 goto out; 3503 3504 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SPLIT, &zc) != 0) { 3505 retval = zpool_standard_error(hdl, errno, msg); 3506 goto out; 3507 } 3508 3509 freelist = B_FALSE; 3510 memory_err = B_FALSE; 3511 3512 out: 3513 if (varray != NULL) { 3514 int v; 3515 3516 for (v = 0; v < vcount; v++) 3517 nvlist_free(varray[v]); 3518 free(varray); 3519 } 3520 zcmd_free_nvlists(&zc); 3521 nvlist_free(zc_props); 3522 nvlist_free(newconfig); 3523 if (freelist) { 3524 nvlist_free(*newroot); 3525 *newroot = NULL; 3526 } 3527 3528 if (retval != 0) 3529 return (retval); 3530 3531 if (memory_err) 3532 return (no_memory(hdl)); 3533 3534 return (0); 3535 } 3536 3537 /* 3538 * Remove the given device. 3539 */ 3540 int 3541 zpool_vdev_remove(zpool_handle_t *zhp, const char *path) 3542 { 3543 zfs_cmd_t zc = { 0 }; 3544 char msg[1024]; 3545 nvlist_t *tgt; 3546 boolean_t avail_spare, l2cache, islog; 3547 libzfs_handle_t *hdl = zhp->zpool_hdl; 3548 uint64_t version; 3549 3550 (void) snprintf(msg, sizeof (msg), 3551 dgettext(TEXT_DOMAIN, "cannot remove %s"), path); 3552 3553 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3554 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 3555 &islog)) == NULL) 3556 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 3557 3558 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL); 3559 if (islog && version < SPA_VERSION_HOLES) { 3560 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3561 "pool must be upgraded to support log removal")); 3562 return (zfs_error(hdl, EZFS_BADVERSION, msg)); 3563 } 3564 3565 zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID); 3566 3567 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0) 3568 return (0); 3569 3570 switch (errno) { 3571 3572 case EINVAL: 3573 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3574 "invalid config; all top-level vdevs must " 3575 "have the same sector size and not be raidz.")); 3576 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg); 3577 break; 3578 3579 case EBUSY: 3580 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3581 "Pool busy; removal may already be in progress")); 3582 (void) zfs_error(hdl, EZFS_BUSY, msg); 3583 break; 3584 3585 default: 3586 (void) zpool_standard_error(hdl, errno, msg); 3587 } 3588 return (-1); 3589 } 3590 3591 int 3592 zpool_vdev_remove_cancel(zpool_handle_t *zhp) 3593 { 3594 zfs_cmd_t zc = { 0 }; 3595 char msg[1024]; 3596 libzfs_handle_t *hdl = zhp->zpool_hdl; 3597 3598 (void) snprintf(msg, sizeof (msg), 3599 dgettext(TEXT_DOMAIN, "cannot cancel removal")); 3600 3601 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3602 zc.zc_cookie = 1; 3603 3604 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0) 3605 return (0); 3606 3607 return (zpool_standard_error(hdl, errno, msg)); 3608 } 3609 3610 int 3611 zpool_vdev_indirect_size(zpool_handle_t *zhp, const char *path, 3612 uint64_t *sizep) 3613 { 3614 char msg[1024]; 3615 nvlist_t *tgt; 3616 boolean_t avail_spare, l2cache, islog; 3617 libzfs_handle_t *hdl = zhp->zpool_hdl; 3618 3619 (void) snprintf(msg, sizeof (msg), 3620 dgettext(TEXT_DOMAIN, "cannot determine indirect size of %s"), 3621 path); 3622 3623 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 3624 &islog)) == NULL) 3625 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 3626 3627 if (avail_spare || l2cache || islog) { 3628 *sizep = 0; 3629 return (0); 3630 } 3631 3632 if (nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_INDIRECT_SIZE, sizep) != 0) { 3633 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3634 "indirect size not available")); 3635 return (zfs_error(hdl, EINVAL, msg)); 3636 } 3637 return (0); 3638 } 3639 3640 /* 3641 * Clear the errors for the pool, or the particular device if specified. 3642 */ 3643 int 3644 zpool_clear(zpool_handle_t *zhp, const char *path, nvlist_t *rewindnvl) 3645 { 3646 zfs_cmd_t zc = { 0 }; 3647 char msg[1024]; 3648 nvlist_t *tgt; 3649 zpool_load_policy_t policy; 3650 boolean_t avail_spare, l2cache; 3651 libzfs_handle_t *hdl = zhp->zpool_hdl; 3652 nvlist_t *nvi = NULL; 3653 int error; 3654 3655 if (path) 3656 (void) snprintf(msg, sizeof (msg), 3657 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"), 3658 path); 3659 else 3660 (void) snprintf(msg, sizeof (msg), 3661 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"), 3662 zhp->zpool_name); 3663 3664 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3665 if (path) { 3666 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, 3667 &l2cache, NULL)) == NULL) 3668 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 3669 3670 /* 3671 * Don't allow error clearing for hot spares. Do allow 3672 * error clearing for l2cache devices. 3673 */ 3674 if (avail_spare) 3675 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 3676 3677 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, 3678 &zc.zc_guid) == 0); 3679 } 3680 3681 zpool_get_load_policy(rewindnvl, &policy); 3682 zc.zc_cookie = policy.zlp_rewind; 3683 3684 if (zcmd_alloc_dst_nvlist(hdl, &zc, zhp->zpool_config_size * 2) != 0) 3685 return (-1); 3686 3687 if (zcmd_write_src_nvlist(hdl, &zc, rewindnvl) != 0) 3688 return (-1); 3689 3690 while ((error = zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc)) != 0 && 3691 errno == ENOMEM) { 3692 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) { 3693 zcmd_free_nvlists(&zc); 3694 return (-1); 3695 } 3696 } 3697 3698 if (!error || ((policy.zlp_rewind & ZPOOL_TRY_REWIND) && 3699 errno != EPERM && errno != EACCES)) { 3700 if (policy.zlp_rewind & 3701 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) { 3702 (void) zcmd_read_dst_nvlist(hdl, &zc, &nvi); 3703 zpool_rewind_exclaim(hdl, zc.zc_name, 3704 ((policy.zlp_rewind & ZPOOL_TRY_REWIND) != 0), 3705 nvi); 3706 nvlist_free(nvi); 3707 } 3708 zcmd_free_nvlists(&zc); 3709 return (0); 3710 } 3711 3712 zcmd_free_nvlists(&zc); 3713 return (zpool_standard_error(hdl, errno, msg)); 3714 } 3715 3716 /* 3717 * Similar to zpool_clear(), but takes a GUID (used by fmd). 3718 */ 3719 int 3720 zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid) 3721 { 3722 zfs_cmd_t zc = { 0 }; 3723 char msg[1024]; 3724 libzfs_handle_t *hdl = zhp->zpool_hdl; 3725 3726 (void) snprintf(msg, sizeof (msg), 3727 dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"), 3728 guid); 3729 3730 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3731 zc.zc_guid = guid; 3732 zc.zc_cookie = ZPOOL_NO_REWIND; 3733 3734 if (ioctl(hdl->libzfs_fd, ZFS_IOC_CLEAR, &zc) == 0) 3735 return (0); 3736 3737 return (zpool_standard_error(hdl, errno, msg)); 3738 } 3739 3740 /* 3741 * Change the GUID for a pool. 3742 */ 3743 int 3744 zpool_reguid(zpool_handle_t *zhp) 3745 { 3746 char msg[1024]; 3747 libzfs_handle_t *hdl = zhp->zpool_hdl; 3748 zfs_cmd_t zc = { 0 }; 3749 3750 (void) snprintf(msg, sizeof (msg), 3751 dgettext(TEXT_DOMAIN, "cannot reguid '%s'"), zhp->zpool_name); 3752 3753 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3754 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REGUID, &zc) == 0) 3755 return (0); 3756 3757 return (zpool_standard_error(hdl, errno, msg)); 3758 } 3759 3760 /* 3761 * Reopen the pool. 3762 */ 3763 int 3764 zpool_reopen(zpool_handle_t *zhp) 3765 { 3766 zfs_cmd_t zc = { 0 }; 3767 char msg[1024]; 3768 libzfs_handle_t *hdl = zhp->zpool_hdl; 3769 3770 (void) snprintf(msg, sizeof (msg), 3771 dgettext(TEXT_DOMAIN, "cannot reopen '%s'"), 3772 zhp->zpool_name); 3773 3774 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3775 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REOPEN, &zc) == 0) 3776 return (0); 3777 return (zpool_standard_error(hdl, errno, msg)); 3778 } 3779 3780 /* call into libzfs_core to execute the sync IOCTL per pool */ 3781 int 3782 zpool_sync_one(zpool_handle_t *zhp, void *data) 3783 { 3784 int ret; 3785 libzfs_handle_t *hdl = zpool_get_handle(zhp); 3786 const char *pool_name = zpool_get_name(zhp); 3787 boolean_t *force = data; 3788 nvlist_t *innvl = fnvlist_alloc(); 3789 3790 fnvlist_add_boolean_value(innvl, "force", *force); 3791 if ((ret = lzc_sync(pool_name, innvl, NULL)) != 0) { 3792 nvlist_free(innvl); 3793 return (zpool_standard_error_fmt(hdl, ret, 3794 dgettext(TEXT_DOMAIN, "sync '%s' failed"), pool_name)); 3795 } 3796 nvlist_free(innvl); 3797 3798 return (0); 3799 } 3800 3801 /* 3802 * Convert from a devid string to a path. 3803 */ 3804 static char * 3805 devid_to_path(char *devid_str) 3806 { 3807 ddi_devid_t devid; 3808 char *minor; 3809 char *path; 3810 devid_nmlist_t *list = NULL; 3811 int ret; 3812 3813 if (devid_str_decode(devid_str, &devid, &minor) != 0) 3814 return (NULL); 3815 3816 ret = devid_deviceid_to_nmlist("/dev", devid, minor, &list); 3817 3818 devid_str_free(minor); 3819 devid_free(devid); 3820 3821 if (ret != 0) 3822 return (NULL); 3823 3824 /* 3825 * In a case the strdup() fails, we will just return NULL below. 3826 */ 3827 path = strdup(list[0].devname); 3828 3829 devid_free_nmlist(list); 3830 3831 return (path); 3832 } 3833 3834 /* 3835 * Convert from a path to a devid string. 3836 */ 3837 static char * 3838 path_to_devid(const char *path) 3839 { 3840 int fd; 3841 ddi_devid_t devid; 3842 char *minor, *ret; 3843 3844 if ((fd = open(path, O_RDONLY)) < 0) 3845 return (NULL); 3846 3847 minor = NULL; 3848 ret = NULL; 3849 if (devid_get(fd, &devid) == 0) { 3850 if (devid_get_minor_name(fd, &minor) == 0) 3851 ret = devid_str_encode(devid, minor); 3852 if (minor != NULL) 3853 devid_str_free(minor); 3854 devid_free(devid); 3855 } 3856 (void) close(fd); 3857 3858 return (ret); 3859 } 3860 3861 struct path_from_physpath_walker_args { 3862 char *pfpwa_path; 3863 }; 3864 3865 /* 3866 * Walker for use with di_devlink_walk(). Stores the "/dev" path of the first 3867 * primary devlink (i.e., the first devlink which refers to our "/devices" 3868 * node) and stops walking. 3869 */ 3870 static int 3871 path_from_physpath_walker(di_devlink_t devlink, void *arg) 3872 { 3873 struct path_from_physpath_walker_args *pfpwa = arg; 3874 3875 if (di_devlink_type(devlink) != DI_PRIMARY_LINK) { 3876 return (DI_WALK_CONTINUE); 3877 } 3878 3879 verify(pfpwa->pfpwa_path == NULL); 3880 if ((pfpwa->pfpwa_path = strdup(di_devlink_path(devlink))) != NULL) { 3881 return (DI_WALK_TERMINATE); 3882 } 3883 3884 return (DI_WALK_CONTINUE); 3885 } 3886 3887 /* 3888 * Search for a "/dev" path that refers to our physical path. Returns the new 3889 * path if one is found and it does not match the existing "path" value. If 3890 * the value is unchanged, or one could not be found, returns NULL. 3891 */ 3892 static char * 3893 path_from_physpath(libzfs_handle_t *hdl, const char *path, 3894 const char *physpath) 3895 { 3896 struct path_from_physpath_walker_args pfpwa; 3897 3898 if (physpath == NULL) { 3899 return (NULL); 3900 } 3901 3902 if (hdl->libzfs_devlink == NULL) { 3903 if ((hdl->libzfs_devlink = di_devlink_init(NULL, 0)) == 3904 DI_LINK_NIL) { 3905 /* 3906 * We may not be able to open a handle if this process 3907 * is insufficiently privileged, or we are too early in 3908 * boot for devfsadm to be ready. Ignore this error 3909 * and defer the path check to a subsequent run. 3910 */ 3911 return (NULL); 3912 } 3913 } 3914 3915 pfpwa.pfpwa_path = NULL; 3916 (void) di_devlink_walk(hdl->libzfs_devlink, NULL, physpath, 3917 DI_PRIMARY_LINK, &pfpwa, path_from_physpath_walker); 3918 3919 if (path != NULL && pfpwa.pfpwa_path != NULL && 3920 strcmp(path, pfpwa.pfpwa_path) == 0) { 3921 /* 3922 * If the path is already correct, no change is required. 3923 */ 3924 free(pfpwa.pfpwa_path); 3925 return (NULL); 3926 } 3927 3928 return (pfpwa.pfpwa_path); 3929 } 3930 3931 /* 3932 * Issue the necessary ioctl() to update the stored path value for the vdev. We 3933 * ignore any failure here, since a common case is for an unprivileged user to 3934 * type 'zpool status', and we'll display the correct information anyway. 3935 */ 3936 static void 3937 set_path(zpool_handle_t *zhp, nvlist_t *nv, const char *path) 3938 { 3939 zfs_cmd_t zc = { 0 }; 3940 3941 (void) strncpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3942 (void) strncpy(zc.zc_value, path, sizeof (zc.zc_value)); 3943 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, 3944 &zc.zc_guid) == 0); 3945 3946 (void) ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SETPATH, &zc); 3947 } 3948 3949 /* 3950 * Given a vdev, return the name to display in iostat. If the vdev has a path, 3951 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type. 3952 * We also check if this is a whole disk, in which case we strip off the 3953 * trailing 's0' slice name. 3954 * 3955 * This routine is also responsible for identifying when disks have been 3956 * reconfigured in a new location. The kernel will have opened the device by 3957 * devid, but the path will still refer to the old location. To catch this, we 3958 * first do a path -> devid translation (which is fast for the common case). If 3959 * the devid matches, we're done. If not, we do a reverse devid -> path 3960 * translation and issue the appropriate ioctl() to update the path of the vdev. 3961 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any 3962 * of these checks. 3963 */ 3964 char * 3965 zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv, 3966 int name_flags) 3967 { 3968 char *path, *type, *env; 3969 uint64_t value; 3970 char buf[64]; 3971 3972 /* 3973 * vdev_name will be "root"/"root-0" for the root vdev, but it is the 3974 * zpool name that will be displayed to the user. 3975 */ 3976 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0); 3977 if (zhp != NULL && strcmp(type, "root") == 0) 3978 return (zfs_strdup(hdl, zpool_get_name(zhp))); 3979 3980 env = getenv("ZPOOL_VDEV_NAME_PATH"); 3981 if (env && (strtoul(env, NULL, 0) > 0 || 3982 !strncasecmp(env, "YES", 3) || !strncasecmp(env, "ON", 2))) 3983 name_flags |= VDEV_NAME_PATH; 3984 3985 env = getenv("ZPOOL_VDEV_NAME_GUID"); 3986 if (env && (strtoul(env, NULL, 0) > 0 || 3987 !strncasecmp(env, "YES", 3) || !strncasecmp(env, "ON", 2))) 3988 name_flags |= VDEV_NAME_GUID; 3989 3990 env = getenv("ZPOOL_VDEV_NAME_FOLLOW_LINKS"); 3991 if (env && (strtoul(env, NULL, 0) > 0 || 3992 !strncasecmp(env, "YES", 3) || !strncasecmp(env, "ON", 2))) 3993 name_flags |= VDEV_NAME_FOLLOW_LINKS; 3994 3995 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, &value) == 0 || 3996 name_flags & VDEV_NAME_GUID) { 3997 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value); 3998 (void) snprintf(buf, sizeof (buf), "%llu", (u_longlong_t)value); 3999 path = buf; 4000 } else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) { 4001 vdev_stat_t *vs; 4002 uint_t vsc; 4003 char *newpath = NULL; 4004 char *physpath = NULL; 4005 char *devid = NULL; 4006 4007 /* 4008 * If the device is dead (faulted, offline, etc) then don't 4009 * bother opening it. Otherwise we may be forcing the user to 4010 * open a misbehaving device, which can have undesirable 4011 * effects. 4012 */ 4013 if (nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS, 4014 (uint64_t **)&vs, &vsc) != 0 || 4015 vs->vs_state < VDEV_STATE_DEGRADED || 4016 zhp == NULL) { 4017 goto after_open; 4018 } 4019 4020 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &devid) == 0) { 4021 /* 4022 * This vdev has a devid. We can use it to check the 4023 * current path. 4024 */ 4025 char *newdevid = path_to_devid(path); 4026 4027 if (newdevid == NULL || strcmp(devid, newdevid) != 0) { 4028 newpath = devid_to_path(devid); 4029 } 4030 4031 if (newdevid != NULL) 4032 devid_str_free(newdevid); 4033 4034 } else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PHYS_PATH, 4035 &physpath) == 0) { 4036 /* 4037 * This vdev does not have a devid, but it does have a 4038 * physical path. Attempt to translate this to a /dev 4039 * path. 4040 */ 4041 newpath = path_from_physpath(hdl, path, physpath); 4042 } 4043 4044 if (newpath != NULL) { 4045 /* 4046 * Update the path appropriately. 4047 */ 4048 set_path(zhp, nv, newpath); 4049 if (nvlist_add_string(nv, ZPOOL_CONFIG_PATH, 4050 newpath) == 0) { 4051 verify(nvlist_lookup_string(nv, 4052 ZPOOL_CONFIG_PATH, &path) == 0); 4053 } 4054 free(newpath); 4055 } 4056 4057 if (name_flags & VDEV_NAME_FOLLOW_LINKS) { 4058 char *rp = realpath(path, NULL); 4059 if (rp) { 4060 strlcpy(buf, rp, sizeof (buf)); 4061 path = buf; 4062 free(rp); 4063 } 4064 } 4065 4066 after_open: 4067 if (strncmp(path, ZFS_DISK_ROOTD, strlen(ZFS_DISK_ROOTD)) == 0) 4068 path += strlen(ZFS_DISK_ROOTD); 4069 4070 /* 4071 * Remove the partition from the path it this is a whole disk. 4072 */ 4073 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, &value) 4074 == 0 && value && !(name_flags & VDEV_NAME_PATH)) { 4075 int pathlen = strlen(path); 4076 char *tmp = zfs_strdup(hdl, path); 4077 4078 /* 4079 * If it starts with c#, and ends with "s0" or "s1", 4080 * chop the slice off, or if it ends with "s0/old" or 4081 * "s1/old", remove the slice from the middle. 4082 */ 4083 if (CTD_CHECK(tmp)) { 4084 if (strcmp(&tmp[pathlen - 2], "s0") == 0 || 4085 strcmp(&tmp[pathlen - 2], "s1") == 0) { 4086 tmp[pathlen - 2] = '\0'; 4087 } else if (pathlen > 6 && 4088 (strcmp(&tmp[pathlen - 6], "s0/old") == 0 || 4089 strcmp(&tmp[pathlen - 6], "s1/old") == 0)) { 4090 (void) strcpy(&tmp[pathlen - 6], 4091 "/old"); 4092 } 4093 } 4094 return (tmp); 4095 } 4096 } else { 4097 path = type; 4098 4099 /* 4100 * If it's a raidz device, we need to stick in the parity level. 4101 */ 4102 if (strcmp(path, VDEV_TYPE_RAIDZ) == 0) { 4103 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY, 4104 &value) == 0); 4105 (void) snprintf(buf, sizeof (buf), "%s%llu", path, 4106 (u_longlong_t)value); 4107 path = buf; 4108 } 4109 4110 /* 4111 * We identify each top-level vdev by using a <type-id> 4112 * naming convention. 4113 */ 4114 if (name_flags & VDEV_NAME_TYPE_ID) { 4115 uint64_t id; 4116 4117 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID, 4118 &id) == 0); 4119 (void) snprintf(buf, sizeof (buf), "%s-%llu", path, 4120 (u_longlong_t)id); 4121 path = buf; 4122 } 4123 } 4124 4125 return (zfs_strdup(hdl, path)); 4126 } 4127 4128 static int 4129 zbookmark_mem_compare(const void *a, const void *b) 4130 { 4131 return (memcmp(a, b, sizeof (zbookmark_phys_t))); 4132 } 4133 4134 /* 4135 * Retrieve the persistent error log, uniquify the members, and return to the 4136 * caller. 4137 */ 4138 int 4139 zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp) 4140 { 4141 zfs_cmd_t zc = { 0 }; 4142 uint64_t count; 4143 zbookmark_phys_t *zb = NULL; 4144 int i; 4145 4146 /* 4147 * Retrieve the raw error list from the kernel. If the number of errors 4148 * has increased, allocate more space and continue until we get the 4149 * entire list. 4150 */ 4151 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_ERRCOUNT, 4152 &count) == 0); 4153 if (count == 0) 4154 return (0); 4155 if ((zc.zc_nvlist_dst = (uintptr_t)zfs_alloc(zhp->zpool_hdl, 4156 count * sizeof (zbookmark_phys_t))) == (uintptr_t)NULL) 4157 return (-1); 4158 zc.zc_nvlist_dst_size = count; 4159 (void) strcpy(zc.zc_name, zhp->zpool_name); 4160 for (;;) { 4161 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_ERROR_LOG, 4162 &zc) != 0) { 4163 free((void *)(uintptr_t)zc.zc_nvlist_dst); 4164 if (errno == ENOMEM) { 4165 void *dst; 4166 4167 count = zc.zc_nvlist_dst_size; 4168 dst = zfs_alloc(zhp->zpool_hdl, count * 4169 sizeof (zbookmark_phys_t)); 4170 if (dst == NULL) 4171 return (-1); 4172 zc.zc_nvlist_dst = (uintptr_t)dst; 4173 } else { 4174 return (-1); 4175 } 4176 } else { 4177 break; 4178 } 4179 } 4180 4181 /* 4182 * Sort the resulting bookmarks. This is a little confusing due to the 4183 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last 4184 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks 4185 * _not_ copied as part of the process. So we point the start of our 4186 * array appropriate and decrement the total number of elements. 4187 */ 4188 zb = ((zbookmark_phys_t *)(uintptr_t)zc.zc_nvlist_dst) + 4189 zc.zc_nvlist_dst_size; 4190 count -= zc.zc_nvlist_dst_size; 4191 4192 qsort(zb, count, sizeof (zbookmark_phys_t), zbookmark_mem_compare); 4193 4194 verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0); 4195 4196 /* 4197 * Fill in the nverrlistp with nvlist's of dataset and object numbers. 4198 */ 4199 for (i = 0; i < count; i++) { 4200 nvlist_t *nv; 4201 4202 /* ignoring zb_blkid and zb_level for now */ 4203 if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset && 4204 zb[i-1].zb_object == zb[i].zb_object) 4205 continue; 4206 4207 if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0) 4208 goto nomem; 4209 if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET, 4210 zb[i].zb_objset) != 0) { 4211 nvlist_free(nv); 4212 goto nomem; 4213 } 4214 if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT, 4215 zb[i].zb_object) != 0) { 4216 nvlist_free(nv); 4217 goto nomem; 4218 } 4219 if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) { 4220 nvlist_free(nv); 4221 goto nomem; 4222 } 4223 nvlist_free(nv); 4224 } 4225 4226 free((void *)(uintptr_t)zc.zc_nvlist_dst); 4227 return (0); 4228 4229 nomem: 4230 free((void *)(uintptr_t)zc.zc_nvlist_dst); 4231 return (no_memory(zhp->zpool_hdl)); 4232 } 4233 4234 /* 4235 * Upgrade a ZFS pool to the latest on-disk version. 4236 */ 4237 int 4238 zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version) 4239 { 4240 zfs_cmd_t zc = { 0 }; 4241 libzfs_handle_t *hdl = zhp->zpool_hdl; 4242 4243 (void) strcpy(zc.zc_name, zhp->zpool_name); 4244 zc.zc_cookie = new_version; 4245 4246 if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0) 4247 return (zpool_standard_error_fmt(hdl, errno, 4248 dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"), 4249 zhp->zpool_name)); 4250 return (0); 4251 } 4252 4253 void 4254 zfs_save_arguments(int argc, char **argv, char *string, int len) 4255 { 4256 (void) strlcpy(string, basename(argv[0]), len); 4257 for (int i = 1; i < argc; i++) { 4258 (void) strlcat(string, " ", len); 4259 (void) strlcat(string, argv[i], len); 4260 } 4261 } 4262 4263 int 4264 zpool_log_history(libzfs_handle_t *hdl, const char *message) 4265 { 4266 zfs_cmd_t zc = { 0 }; 4267 nvlist_t *args; 4268 int err; 4269 4270 args = fnvlist_alloc(); 4271 fnvlist_add_string(args, "message", message); 4272 err = zcmd_write_src_nvlist(hdl, &zc, args); 4273 if (err == 0) 4274 err = ioctl(hdl->libzfs_fd, ZFS_IOC_LOG_HISTORY, &zc); 4275 nvlist_free(args); 4276 zcmd_free_nvlists(&zc); 4277 return (err); 4278 } 4279 4280 /* 4281 * Perform ioctl to get some command history of a pool. 4282 * 4283 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the 4284 * logical offset of the history buffer to start reading from. 4285 * 4286 * Upon return, 'off' is the next logical offset to read from and 4287 * 'len' is the actual amount of bytes read into 'buf'. 4288 */ 4289 static int 4290 get_history(zpool_handle_t *zhp, char *buf, uint64_t *off, uint64_t *len) 4291 { 4292 zfs_cmd_t zc = { 0 }; 4293 libzfs_handle_t *hdl = zhp->zpool_hdl; 4294 4295 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 4296 4297 zc.zc_history = (uint64_t)(uintptr_t)buf; 4298 zc.zc_history_len = *len; 4299 zc.zc_history_offset = *off; 4300 4301 if (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) { 4302 switch (errno) { 4303 case EPERM: 4304 return (zfs_error_fmt(hdl, EZFS_PERM, 4305 dgettext(TEXT_DOMAIN, 4306 "cannot show history for pool '%s'"), 4307 zhp->zpool_name)); 4308 case ENOENT: 4309 return (zfs_error_fmt(hdl, EZFS_NOHISTORY, 4310 dgettext(TEXT_DOMAIN, "cannot get history for pool " 4311 "'%s'"), zhp->zpool_name)); 4312 case ENOTSUP: 4313 return (zfs_error_fmt(hdl, EZFS_BADVERSION, 4314 dgettext(TEXT_DOMAIN, "cannot get history for pool " 4315 "'%s', pool must be upgraded"), zhp->zpool_name)); 4316 default: 4317 return (zpool_standard_error_fmt(hdl, errno, 4318 dgettext(TEXT_DOMAIN, 4319 "cannot get history for '%s'"), zhp->zpool_name)); 4320 } 4321 } 4322 4323 *len = zc.zc_history_len; 4324 *off = zc.zc_history_offset; 4325 4326 return (0); 4327 } 4328 4329 /* 4330 * Retrieve the command history of a pool. 4331 */ 4332 int 4333 zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp, uint64_t *off, 4334 boolean_t *eof) 4335 { 4336 char *buf; 4337 int buflen = 128 * 1024; 4338 nvlist_t **records = NULL; 4339 uint_t numrecords = 0; 4340 int err = 0, i; 4341 uint64_t start = *off; 4342 4343 buf = malloc(buflen); 4344 if (buf == NULL) 4345 return (ENOMEM); 4346 /* process about 1MB a time */ 4347 while (*off - start < 1024 * 1024) { 4348 uint64_t bytes_read = buflen; 4349 uint64_t leftover; 4350 4351 if ((err = get_history(zhp, buf, off, &bytes_read)) != 0) 4352 break; 4353 4354 /* if nothing else was read in, we're at EOF, just return */ 4355 if (!bytes_read) { 4356 *eof = B_TRUE; 4357 break; 4358 } 4359 4360 if ((err = zpool_history_unpack(buf, bytes_read, 4361 &leftover, &records, &numrecords)) != 0) 4362 break; 4363 *off -= leftover; 4364 if (leftover == bytes_read) { 4365 /* 4366 * no progress made, because buffer is not big enough 4367 * to hold this record; resize and retry. 4368 */ 4369 buflen *= 2; 4370 free(buf); 4371 buf = malloc(buflen); 4372 if (buf == NULL) 4373 return (ENOMEM); 4374 } 4375 } 4376 4377 free(buf); 4378 4379 if (!err) { 4380 verify(nvlist_alloc(nvhisp, NV_UNIQUE_NAME, 0) == 0); 4381 verify(nvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD, 4382 records, numrecords) == 0); 4383 } 4384 for (i = 0; i < numrecords; i++) 4385 nvlist_free(records[i]); 4386 free(records); 4387 4388 return (err); 4389 } 4390 4391 void 4392 zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj, 4393 char *pathname, size_t len) 4394 { 4395 zfs_cmd_t zc = { 0 }; 4396 boolean_t mounted = B_FALSE; 4397 char *mntpnt = NULL; 4398 char dsname[ZFS_MAX_DATASET_NAME_LEN]; 4399 4400 if (dsobj == 0) { 4401 /* special case for the MOS */ 4402 (void) snprintf(pathname, len, "<metadata>:<0x%llx>", obj); 4403 return; 4404 } 4405 4406 /* get the dataset's name */ 4407 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 4408 zc.zc_obj = dsobj; 4409 if (ioctl(zhp->zpool_hdl->libzfs_fd, 4410 ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) { 4411 /* just write out a path of two object numbers */ 4412 (void) snprintf(pathname, len, "<0x%llx>:<0x%llx>", 4413 dsobj, obj); 4414 return; 4415 } 4416 (void) strlcpy(dsname, zc.zc_value, sizeof (dsname)); 4417 4418 /* find out if the dataset is mounted */ 4419 mounted = is_mounted(zhp->zpool_hdl, dsname, &mntpnt); 4420 4421 /* get the corrupted object's path */ 4422 (void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name)); 4423 zc.zc_obj = obj; 4424 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_OBJ_TO_PATH, 4425 &zc) == 0) { 4426 if (mounted) { 4427 (void) snprintf(pathname, len, "%s%s", mntpnt, 4428 zc.zc_value); 4429 } else { 4430 (void) snprintf(pathname, len, "%s:%s", 4431 dsname, zc.zc_value); 4432 } 4433 } else { 4434 (void) snprintf(pathname, len, "%s:<0x%llx>", dsname, obj); 4435 } 4436 free(mntpnt); 4437 } 4438 4439 int 4440 zpool_set_bootenv(zpool_handle_t *zhp, const nvlist_t *envmap) 4441 { 4442 int error = lzc_set_bootenv(zhp->zpool_name, envmap); 4443 if (error != 0) { 4444 (void) zpool_standard_error_fmt(zhp->zpool_hdl, error, 4445 dgettext(TEXT_DOMAIN, 4446 "error setting bootenv in pool '%s'"), zhp->zpool_name); 4447 } 4448 4449 return (error); 4450 } 4451 4452 int 4453 zpool_get_bootenv(zpool_handle_t *zhp, nvlist_t **nvlp) 4454 { 4455 nvlist_t *nvl; 4456 int error; 4457 4458 nvl = NULL; 4459 error = lzc_get_bootenv(zhp->zpool_name, &nvl); 4460 if (error != 0) { 4461 (void) zpool_standard_error_fmt(zhp->zpool_hdl, error, 4462 dgettext(TEXT_DOMAIN, 4463 "error getting bootenv in pool '%s'"), zhp->zpool_name); 4464 } else { 4465 *nvlp = nvl; 4466 } 4467 4468 return (error); 4469 } 4470 4471 /* 4472 * Read the EFI label from the config, if a label does not exist then 4473 * pass back the error to the caller. If the caller has passed a non-NULL 4474 * diskaddr argument then we set it to the starting address of the EFI 4475 * partition. If the caller has passed a non-NULL boolean argument, then 4476 * we set it to indicate if the disk does have efi system partition. 4477 */ 4478 static int 4479 read_efi_label(nvlist_t *config, diskaddr_t *sb, boolean_t *system) 4480 { 4481 char *path; 4482 int fd; 4483 char diskname[MAXPATHLEN]; 4484 boolean_t boot = B_FALSE; 4485 int err = -1; 4486 int slice; 4487 4488 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PATH, &path) != 0) 4489 return (err); 4490 4491 (void) snprintf(diskname, sizeof (diskname), "%s%s", ZFS_RDISK_ROOT, 4492 strrchr(path, '/')); 4493 if ((fd = open(diskname, O_RDONLY|O_NDELAY)) >= 0) { 4494 struct dk_gpt *vtoc; 4495 4496 if ((err = efi_alloc_and_read(fd, &vtoc)) >= 0) { 4497 for (slice = 0; slice < vtoc->efi_nparts; slice++) { 4498 if (vtoc->efi_parts[slice].p_tag == V_SYSTEM) 4499 boot = B_TRUE; 4500 if (vtoc->efi_parts[slice].p_tag == V_USR) 4501 break; 4502 } 4503 if (sb != NULL && vtoc->efi_parts[slice].p_tag == V_USR) 4504 *sb = vtoc->efi_parts[slice].p_start; 4505 if (system != NULL) 4506 *system = boot; 4507 efi_free(vtoc); 4508 } 4509 (void) close(fd); 4510 } 4511 return (err); 4512 } 4513 4514 /* 4515 * determine where a partition starts on a disk in the current 4516 * configuration 4517 */ 4518 static diskaddr_t 4519 find_start_block(nvlist_t *config) 4520 { 4521 nvlist_t **child; 4522 uint_t c, children; 4523 diskaddr_t sb = MAXOFFSET_T; 4524 uint64_t wholedisk; 4525 4526 if (nvlist_lookup_nvlist_array(config, 4527 ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) { 4528 if (nvlist_lookup_uint64(config, 4529 ZPOOL_CONFIG_WHOLE_DISK, 4530 &wholedisk) != 0 || !wholedisk) { 4531 return (MAXOFFSET_T); 4532 } 4533 if (read_efi_label(config, &sb, NULL) < 0) 4534 sb = MAXOFFSET_T; 4535 return (sb); 4536 } 4537 4538 for (c = 0; c < children; c++) { 4539 sb = find_start_block(child[c]); 4540 if (sb != MAXOFFSET_T) { 4541 return (sb); 4542 } 4543 } 4544 return (MAXOFFSET_T); 4545 } 4546 4547 /* 4548 * Label an individual disk. The name provided is the short name, 4549 * stripped of any leading /dev path. 4550 */ 4551 int 4552 zpool_label_disk(libzfs_handle_t *hdl, zpool_handle_t *zhp, const char *name, 4553 zpool_boot_label_t boot_type, uint64_t boot_size, int *slice) 4554 { 4555 char path[MAXPATHLEN]; 4556 struct dk_gpt *vtoc; 4557 int fd; 4558 size_t resv; 4559 uint64_t slice_size; 4560 diskaddr_t start_block; 4561 char errbuf[1024]; 4562 4563 /* prepare an error message just in case */ 4564 (void) snprintf(errbuf, sizeof (errbuf), 4565 dgettext(TEXT_DOMAIN, "cannot label '%s'"), name); 4566 4567 if (zhp) { 4568 nvlist_t *nvroot; 4569 4570 verify(nvlist_lookup_nvlist(zhp->zpool_config, 4571 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 4572 4573 if (zhp->zpool_start_block == 0) 4574 start_block = find_start_block(nvroot); 4575 else 4576 start_block = zhp->zpool_start_block; 4577 zhp->zpool_start_block = start_block; 4578 } else { 4579 /* new pool */ 4580 start_block = NEW_START_BLOCK; 4581 } 4582 4583 (void) snprintf(path, sizeof (path), "%s/%s%s", ZFS_RDISK_ROOT, name, 4584 BACKUP_SLICE); 4585 4586 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) { 4587 /* 4588 * This shouldn't happen. We've long since verified that this 4589 * is a valid device. 4590 */ 4591 zfs_error_aux(hdl, 4592 dgettext(TEXT_DOMAIN, "unable to open device")); 4593 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf)); 4594 } 4595 4596 if (efi_alloc_and_init(fd, EFI_NUMPAR, &vtoc) != 0) { 4597 /* 4598 * The only way this can fail is if we run out of memory, or we 4599 * were unable to read the disk's capacity 4600 */ 4601 if (errno == ENOMEM) 4602 (void) no_memory(hdl); 4603 4604 (void) close(fd); 4605 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4606 "unable to read disk capacity"), name); 4607 4608 return (zfs_error(hdl, EZFS_NOCAP, errbuf)); 4609 } 4610 resv = efi_reserved_sectors(vtoc); 4611 4612 /* 4613 * Why we use V_USR: V_BACKUP confuses users, and is considered 4614 * disposable by some EFI utilities (since EFI doesn't have a backup 4615 * slice). V_UNASSIGNED is supposed to be used only for zero size 4616 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT, 4617 * etc. were all pretty specific. V_USR is as close to reality as we 4618 * can get, in the absence of V_OTHER. 4619 */ 4620 /* first fix the partition start block */ 4621 if (start_block == MAXOFFSET_T) 4622 start_block = NEW_START_BLOCK; 4623 4624 /* 4625 * EFI System partition is using slice 0. 4626 * ZFS is on slice 1 and slice 8 is reserved. 4627 * We assume the GPT partition table without system 4628 * partition has zfs p_start == NEW_START_BLOCK. 4629 * If start_block != NEW_START_BLOCK, it means we have 4630 * system partition. Correct solution would be to query/cache vtoc 4631 * from existing vdev member. 4632 */ 4633 if (boot_type == ZPOOL_CREATE_BOOT_LABEL) { 4634 if (boot_size % vtoc->efi_lbasize != 0) { 4635 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4636 "boot partition size must be a multiple of %d"), 4637 vtoc->efi_lbasize); 4638 (void) close(fd); 4639 efi_free(vtoc); 4640 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf)); 4641 } 4642 /* 4643 * System partition size checks. 4644 * Note the 1MB is quite arbitrary value, since we 4645 * are creating dedicated pool, it should be enough 4646 * to hold fat + efi bootloader. May need to be 4647 * adjusted if the bootloader size will grow. 4648 */ 4649 if (boot_size < 1024 * 1024) { 4650 char buf[64]; 4651 zfs_nicenum(boot_size, buf, sizeof (buf)); 4652 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4653 "Specified size %s for EFI System partition is too " 4654 "small, the minimum size is 1MB."), buf); 4655 (void) close(fd); 4656 efi_free(vtoc); 4657 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf)); 4658 } 4659 /* 33MB is tested with mkfs -F pcfs */ 4660 if (hdl->libzfs_printerr && 4661 ((vtoc->efi_lbasize == 512 && 4662 boot_size < 33 * 1024 * 1024) || 4663 (vtoc->efi_lbasize == 4096 && 4664 boot_size < 256 * 1024 * 1024))) { 4665 char buf[64]; 4666 zfs_nicenum(boot_size, buf, sizeof (buf)); 4667 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, 4668 "Warning: EFI System partition size %s is " 4669 "not allowing to create FAT32 file\nsystem, which " 4670 "may result in unbootable system.\n"), buf); 4671 } 4672 /* Adjust zfs partition start by size of system partition. */ 4673 start_block += boot_size / vtoc->efi_lbasize; 4674 } 4675 4676 if (start_block == NEW_START_BLOCK) { 4677 /* 4678 * Use default layout. 4679 * ZFS is on slice 0 and slice 8 is reserved. 4680 */ 4681 slice_size = vtoc->efi_last_u_lba + 1; 4682 slice_size -= resv; 4683 slice_size -= start_block; 4684 if (slice != NULL) 4685 *slice = 0; 4686 4687 vtoc->efi_parts[0].p_start = start_block; 4688 vtoc->efi_parts[0].p_size = slice_size; 4689 4690 vtoc->efi_parts[0].p_tag = V_USR; 4691 (void) strcpy(vtoc->efi_parts[0].p_name, "zfs"); 4692 4693 vtoc->efi_parts[8].p_start = slice_size + start_block; 4694 vtoc->efi_parts[8].p_size = resv; 4695 vtoc->efi_parts[8].p_tag = V_RESERVED; 4696 } else { 4697 slice_size = start_block - NEW_START_BLOCK; 4698 vtoc->efi_parts[0].p_start = NEW_START_BLOCK; 4699 vtoc->efi_parts[0].p_size = slice_size; 4700 vtoc->efi_parts[0].p_tag = V_SYSTEM; 4701 (void) strcpy(vtoc->efi_parts[0].p_name, "loader"); 4702 if (slice != NULL) 4703 *slice = 1; 4704 /* prepare slice 1 */ 4705 slice_size = vtoc->efi_last_u_lba + 1 - slice_size; 4706 slice_size -= resv; 4707 slice_size -= NEW_START_BLOCK; 4708 vtoc->efi_parts[1].p_start = start_block; 4709 vtoc->efi_parts[1].p_size = slice_size; 4710 vtoc->efi_parts[1].p_tag = V_USR; 4711 (void) strcpy(vtoc->efi_parts[1].p_name, "zfs"); 4712 4713 vtoc->efi_parts[8].p_start = slice_size + start_block; 4714 vtoc->efi_parts[8].p_size = resv; 4715 vtoc->efi_parts[8].p_tag = V_RESERVED; 4716 } 4717 4718 if (efi_write(fd, vtoc) != 0) { 4719 /* 4720 * Some block drivers (like pcata) may not support EFI 4721 * GPT labels. Print out a helpful error message dir- 4722 * ecting the user to manually label the disk and give 4723 * a specific slice. 4724 */ 4725 (void) close(fd); 4726 efi_free(vtoc); 4727 4728 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4729 "try using fdisk(8) and then provide a specific slice")); 4730 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf)); 4731 } 4732 4733 (void) close(fd); 4734 efi_free(vtoc); 4735 return (0); 4736 } 4737 4738 static boolean_t 4739 supported_dump_vdev_type(libzfs_handle_t *hdl, nvlist_t *config, char *errbuf) 4740 { 4741 char *type; 4742 nvlist_t **child; 4743 uint_t children, c; 4744 4745 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_TYPE, &type) == 0); 4746 if (strcmp(type, VDEV_TYPE_FILE) == 0 || 4747 strcmp(type, VDEV_TYPE_HOLE) == 0 || 4748 strcmp(type, VDEV_TYPE_MISSING) == 0) { 4749 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4750 "vdev type '%s' is not supported"), type); 4751 (void) zfs_error(hdl, EZFS_VDEVNOTSUP, errbuf); 4752 return (B_FALSE); 4753 } 4754 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN, 4755 &child, &children) == 0) { 4756 for (c = 0; c < children; c++) { 4757 if (!supported_dump_vdev_type(hdl, child[c], errbuf)) 4758 return (B_FALSE); 4759 } 4760 } 4761 return (B_TRUE); 4762 } 4763 4764 /* 4765 * Check if this zvol is allowable for use as a dump device; zero if 4766 * it is, > 0 if it isn't, < 0 if it isn't a zvol. 4767 * 4768 * Allowable storage configurations include mirrors, all raidz variants, and 4769 * pools with log, cache, and spare devices. Pools which are backed by files or 4770 * have missing/hole vdevs are not suitable. 4771 */ 4772 int 4773 zvol_check_dump_config(char *arg) 4774 { 4775 zpool_handle_t *zhp = NULL; 4776 nvlist_t *config, *nvroot; 4777 char *p, *volname; 4778 nvlist_t **top; 4779 uint_t toplevels; 4780 libzfs_handle_t *hdl; 4781 char errbuf[1024]; 4782 char poolname[ZFS_MAX_DATASET_NAME_LEN]; 4783 int pathlen = strlen(ZVOL_FULL_DEV_DIR); 4784 int ret = 1; 4785 4786 if (strncmp(arg, ZVOL_FULL_DEV_DIR, pathlen)) { 4787 return (-1); 4788 } 4789 4790 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, 4791 "dump is not supported on device '%s'"), arg); 4792 4793 if ((hdl = libzfs_init()) == NULL) 4794 return (1); 4795 libzfs_print_on_error(hdl, B_TRUE); 4796 4797 volname = arg + pathlen; 4798 4799 /* check the configuration of the pool */ 4800 if ((p = strchr(volname, '/')) == NULL) { 4801 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4802 "malformed dataset name")); 4803 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf); 4804 return (1); 4805 } else if (p - volname >= ZFS_MAX_DATASET_NAME_LEN) { 4806 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4807 "dataset name is too long")); 4808 (void) zfs_error(hdl, EZFS_NAMETOOLONG, errbuf); 4809 return (1); 4810 } else { 4811 (void) strncpy(poolname, volname, p - volname); 4812 poolname[p - volname] = '\0'; 4813 } 4814 4815 if ((zhp = zpool_open(hdl, poolname)) == NULL) { 4816 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4817 "could not open pool '%s'"), poolname); 4818 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf); 4819 goto out; 4820 } 4821 config = zpool_get_config(zhp, NULL); 4822 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 4823 &nvroot) != 0) { 4824 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4825 "could not obtain vdev configuration for '%s'"), poolname); 4826 (void) zfs_error(hdl, EZFS_INVALCONFIG, errbuf); 4827 goto out; 4828 } 4829 4830 verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, 4831 &top, &toplevels) == 0); 4832 4833 if (!supported_dump_vdev_type(hdl, top[0], errbuf)) { 4834 goto out; 4835 } 4836 ret = 0; 4837 4838 out: 4839 if (zhp) 4840 zpool_close(zhp); 4841 libzfs_fini(hdl); 4842 return (ret); 4843 } 4844