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, 2017 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 */ 31 32 #include <ctype.h> 33 #include <errno.h> 34 #include <devid.h> 35 #include <fcntl.h> 36 #include <libintl.h> 37 #include <stdio.h> 38 #include <stdlib.h> 39 #include <strings.h> 40 #include <unistd.h> 41 #include <libgen.h> 42 #include <sys/efi_partition.h> 43 #include <sys/vtoc.h> 44 #include <sys/zfs_ioctl.h> 45 #include <dlfcn.h> 46 47 #include "zfs_namecheck.h" 48 #include "zfs_prop.h" 49 #include "libzfs_impl.h" 50 #include "zfs_comutil.h" 51 #include "zfeature_common.h" 52 53 static int read_efi_label(nvlist_t *, diskaddr_t *, boolean_t *); 54 static boolean_t zpool_vdev_is_interior(const char *name); 55 56 #define BACKUP_SLICE "s2" 57 58 typedef struct prop_flags { 59 int create:1; /* Validate property on creation */ 60 int import:1; /* Validate property on import */ 61 } prop_flags_t; 62 63 /* 64 * ==================================================================== 65 * zpool property functions 66 * ==================================================================== 67 */ 68 69 static int 70 zpool_get_all_props(zpool_handle_t *zhp) 71 { 72 zfs_cmd_t zc = { 0 }; 73 libzfs_handle_t *hdl = zhp->zpool_hdl; 74 75 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 76 77 if (zcmd_alloc_dst_nvlist(hdl, &zc, 0) != 0) 78 return (-1); 79 80 while (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_PROPS, &zc) != 0) { 81 if (errno == ENOMEM) { 82 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) { 83 zcmd_free_nvlists(&zc); 84 return (-1); 85 } 86 } else { 87 zcmd_free_nvlists(&zc); 88 return (-1); 89 } 90 } 91 92 if (zcmd_read_dst_nvlist(hdl, &zc, &zhp->zpool_props) != 0) { 93 zcmd_free_nvlists(&zc); 94 return (-1); 95 } 96 97 zcmd_free_nvlists(&zc); 98 99 return (0); 100 } 101 102 static int 103 zpool_props_refresh(zpool_handle_t *zhp) 104 { 105 nvlist_t *old_props; 106 107 old_props = zhp->zpool_props; 108 109 if (zpool_get_all_props(zhp) != 0) 110 return (-1); 111 112 nvlist_free(old_props); 113 return (0); 114 } 115 116 static char * 117 zpool_get_prop_string(zpool_handle_t *zhp, zpool_prop_t prop, 118 zprop_source_t *src) 119 { 120 nvlist_t *nv, *nvl; 121 uint64_t ival; 122 char *value; 123 zprop_source_t source; 124 125 nvl = zhp->zpool_props; 126 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) { 127 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &ival) == 0); 128 source = ival; 129 verify(nvlist_lookup_string(nv, ZPROP_VALUE, &value) == 0); 130 } else { 131 source = ZPROP_SRC_DEFAULT; 132 if ((value = (char *)zpool_prop_default_string(prop)) == NULL) 133 value = "-"; 134 } 135 136 if (src) 137 *src = source; 138 139 return (value); 140 } 141 142 uint64_t 143 zpool_get_prop_int(zpool_handle_t *zhp, zpool_prop_t prop, zprop_source_t *src) 144 { 145 nvlist_t *nv, *nvl; 146 uint64_t value; 147 zprop_source_t source; 148 149 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp)) { 150 /* 151 * zpool_get_all_props() has most likely failed because 152 * the pool is faulted, but if all we need is the top level 153 * vdev's guid then get it from the zhp config nvlist. 154 */ 155 if ((prop == ZPOOL_PROP_GUID) && 156 (nvlist_lookup_nvlist(zhp->zpool_config, 157 ZPOOL_CONFIG_VDEV_TREE, &nv) == 0) && 158 (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value) 159 == 0)) { 160 return (value); 161 } 162 return (zpool_prop_default_numeric(prop)); 163 } 164 165 nvl = zhp->zpool_props; 166 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) { 167 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &value) == 0); 168 source = value; 169 verify(nvlist_lookup_uint64(nv, ZPROP_VALUE, &value) == 0); 170 } else { 171 source = ZPROP_SRC_DEFAULT; 172 value = zpool_prop_default_numeric(prop); 173 } 174 175 if (src) 176 *src = source; 177 178 return (value); 179 } 180 181 /* 182 * Map VDEV STATE to printed strings. 183 */ 184 const char * 185 zpool_state_to_name(vdev_state_t state, vdev_aux_t aux) 186 { 187 switch (state) { 188 case VDEV_STATE_CLOSED: 189 case VDEV_STATE_OFFLINE: 190 return (gettext("OFFLINE")); 191 case VDEV_STATE_REMOVED: 192 return (gettext("REMOVED")); 193 case VDEV_STATE_CANT_OPEN: 194 if (aux == VDEV_AUX_CORRUPT_DATA || aux == VDEV_AUX_BAD_LOG) 195 return (gettext("FAULTED")); 196 else if (aux == VDEV_AUX_SPLIT_POOL) 197 return (gettext("SPLIT")); 198 else 199 return (gettext("UNAVAIL")); 200 case VDEV_STATE_FAULTED: 201 return (gettext("FAULTED")); 202 case VDEV_STATE_DEGRADED: 203 return (gettext("DEGRADED")); 204 case VDEV_STATE_HEALTHY: 205 return (gettext("ONLINE")); 206 207 default: 208 break; 209 } 210 211 return (gettext("UNKNOWN")); 212 } 213 214 /* 215 * Map POOL STATE to printed strings. 216 */ 217 const char * 218 zpool_pool_state_to_name(pool_state_t state) 219 { 220 switch (state) { 221 case POOL_STATE_ACTIVE: 222 return (gettext("ACTIVE")); 223 case POOL_STATE_EXPORTED: 224 return (gettext("EXPORTED")); 225 case POOL_STATE_DESTROYED: 226 return (gettext("DESTROYED")); 227 case POOL_STATE_SPARE: 228 return (gettext("SPARE")); 229 case POOL_STATE_L2CACHE: 230 return (gettext("L2CACHE")); 231 case POOL_STATE_UNINITIALIZED: 232 return (gettext("UNINITIALIZED")); 233 case POOL_STATE_UNAVAIL: 234 return (gettext("UNAVAIL")); 235 case POOL_STATE_POTENTIALLY_ACTIVE: 236 return (gettext("POTENTIALLY_ACTIVE")); 237 } 238 239 return (gettext("UNKNOWN")); 240 } 241 242 /* 243 * Get a zpool property value for 'prop' and return the value in 244 * a pre-allocated buffer. 245 */ 246 int 247 zpool_get_prop(zpool_handle_t *zhp, zpool_prop_t prop, char *buf, size_t len, 248 zprop_source_t *srctype, boolean_t literal) 249 { 250 uint64_t intval; 251 const char *strval; 252 zprop_source_t src = ZPROP_SRC_NONE; 253 nvlist_t *nvroot; 254 vdev_stat_t *vs; 255 uint_t vsc; 256 257 if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) { 258 switch (prop) { 259 case ZPOOL_PROP_NAME: 260 (void) strlcpy(buf, zpool_get_name(zhp), len); 261 break; 262 263 case ZPOOL_PROP_HEALTH: 264 (void) strlcpy(buf, "FAULTED", len); 265 break; 266 267 case ZPOOL_PROP_GUID: 268 intval = zpool_get_prop_int(zhp, prop, &src); 269 (void) snprintf(buf, len, "%llu", intval); 270 break; 271 272 case ZPOOL_PROP_ALTROOT: 273 case ZPOOL_PROP_CACHEFILE: 274 case ZPOOL_PROP_COMMENT: 275 if (zhp->zpool_props != NULL || 276 zpool_get_all_props(zhp) == 0) { 277 (void) strlcpy(buf, 278 zpool_get_prop_string(zhp, prop, &src), 279 len); 280 break; 281 } 282 /* FALLTHROUGH */ 283 default: 284 (void) strlcpy(buf, "-", len); 285 break; 286 } 287 288 if (srctype != NULL) 289 *srctype = src; 290 return (0); 291 } 292 293 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp) && 294 prop != ZPOOL_PROP_NAME) 295 return (-1); 296 297 switch (zpool_prop_get_type(prop)) { 298 case PROP_TYPE_STRING: 299 (void) strlcpy(buf, zpool_get_prop_string(zhp, prop, &src), 300 len); 301 break; 302 303 case PROP_TYPE_NUMBER: 304 intval = zpool_get_prop_int(zhp, prop, &src); 305 306 switch (prop) { 307 case ZPOOL_PROP_SIZE: 308 case ZPOOL_PROP_ALLOCATED: 309 case ZPOOL_PROP_FREE: 310 case ZPOOL_PROP_FREEING: 311 case ZPOOL_PROP_LEAKED: 312 case ZPOOL_PROP_ASHIFT: 313 if (literal) { 314 (void) snprintf(buf, len, "%llu", 315 (u_longlong_t)intval); 316 } else { 317 (void) zfs_nicenum(intval, buf, len); 318 } 319 break; 320 case ZPOOL_PROP_BOOTSIZE: 321 case ZPOOL_PROP_EXPANDSZ: 322 case ZPOOL_PROP_CHECKPOINT: 323 if (intval == 0) { 324 (void) strlcpy(buf, "-", len); 325 } else if (literal) { 326 (void) snprintf(buf, len, "%llu", 327 (u_longlong_t)intval); 328 } else { 329 (void) zfs_nicenum(intval, buf, len); 330 } 331 break; 332 case ZPOOL_PROP_CAPACITY: 333 if (literal) { 334 (void) snprintf(buf, len, "%llu", 335 (u_longlong_t)intval); 336 } else { 337 (void) snprintf(buf, len, "%llu%%", 338 (u_longlong_t)intval); 339 } 340 break; 341 case ZPOOL_PROP_FRAGMENTATION: 342 if (intval == UINT64_MAX) { 343 (void) strlcpy(buf, "-", len); 344 } else { 345 (void) snprintf(buf, len, "%llu%%", 346 (u_longlong_t)intval); 347 } 348 break; 349 case ZPOOL_PROP_DEDUPRATIO: 350 (void) snprintf(buf, len, "%llu.%02llux", 351 (u_longlong_t)(intval / 100), 352 (u_longlong_t)(intval % 100)); 353 break; 354 case ZPOOL_PROP_HEALTH: 355 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL), 356 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 357 verify(nvlist_lookup_uint64_array(nvroot, 358 ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc) 359 == 0); 360 361 (void) strlcpy(buf, zpool_state_to_name(intval, 362 vs->vs_aux), len); 363 break; 364 case ZPOOL_PROP_VERSION: 365 if (intval >= SPA_VERSION_FEATURES) { 366 (void) snprintf(buf, len, "-"); 367 break; 368 } 369 /* FALLTHROUGH */ 370 default: 371 (void) snprintf(buf, len, "%llu", intval); 372 } 373 break; 374 375 case PROP_TYPE_INDEX: 376 intval = zpool_get_prop_int(zhp, prop, &src); 377 if (zpool_prop_index_to_string(prop, intval, &strval) 378 != 0) 379 return (-1); 380 (void) strlcpy(buf, strval, len); 381 break; 382 383 default: 384 abort(); 385 } 386 387 if (srctype) 388 *srctype = src; 389 390 return (0); 391 } 392 393 /* 394 * Check if the bootfs name has the same pool name as it is set to. 395 * Assuming bootfs is a valid dataset name. 396 */ 397 static boolean_t 398 bootfs_name_valid(const char *pool, char *bootfs) 399 { 400 int len = strlen(pool); 401 if (bootfs[0] == '\0') 402 return (B_TRUE); 403 404 if (!zfs_name_valid(bootfs, ZFS_TYPE_FILESYSTEM|ZFS_TYPE_SNAPSHOT)) 405 return (B_FALSE); 406 407 if (strncmp(pool, bootfs, len) == 0 && 408 (bootfs[len] == '/' || bootfs[len] == '\0')) 409 return (B_TRUE); 410 411 return (B_FALSE); 412 } 413 414 boolean_t 415 zpool_is_bootable(zpool_handle_t *zhp) 416 { 417 char bootfs[ZFS_MAX_DATASET_NAME_LEN]; 418 419 return (zpool_get_prop(zhp, ZPOOL_PROP_BOOTFS, bootfs, 420 sizeof (bootfs), NULL, B_FALSE) == 0 && strncmp(bootfs, "-", 421 sizeof (bootfs)) != 0); 422 } 423 424 425 /* 426 * Given an nvlist of zpool properties to be set, validate that they are 427 * correct, and parse any numeric properties (index, boolean, etc) if they are 428 * specified as strings. 429 */ 430 static nvlist_t * 431 zpool_valid_proplist(libzfs_handle_t *hdl, const char *poolname, 432 nvlist_t *props, uint64_t version, prop_flags_t flags, char *errbuf) 433 { 434 nvpair_t *elem; 435 nvlist_t *retprops; 436 zpool_prop_t prop; 437 char *strval; 438 uint64_t intval; 439 char *slash, *check; 440 struct stat64 statbuf; 441 zpool_handle_t *zhp; 442 443 if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) { 444 (void) no_memory(hdl); 445 return (NULL); 446 } 447 448 elem = NULL; 449 while ((elem = nvlist_next_nvpair(props, elem)) != NULL) { 450 const char *propname = nvpair_name(elem); 451 452 prop = zpool_name_to_prop(propname); 453 if (prop == ZPOOL_PROP_INVAL && zpool_prop_feature(propname)) { 454 int err; 455 char *fname = strchr(propname, '@') + 1; 456 457 err = zfeature_lookup_name(fname, NULL); 458 if (err != 0) { 459 ASSERT3U(err, ==, ENOENT); 460 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 461 "invalid feature '%s', '%s'"), fname, 462 propname); 463 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 464 goto error; 465 } 466 467 if (nvpair_type(elem) != DATA_TYPE_STRING) { 468 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 469 "'%s' must be a string"), propname); 470 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 471 goto error; 472 } 473 474 (void) nvpair_value_string(elem, &strval); 475 if (strcmp(strval, ZFS_FEATURE_ENABLED) != 0) { 476 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 477 "property '%s' can only be set to " 478 "'enabled'"), propname); 479 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 480 goto error; 481 } 482 483 if (nvlist_add_uint64(retprops, propname, 0) != 0) { 484 (void) no_memory(hdl); 485 goto error; 486 } 487 continue; 488 } 489 490 /* 491 * Make sure this property is valid and applies to this type. 492 */ 493 if (prop == ZPOOL_PROP_INVAL) { 494 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 495 "invalid property '%s'"), propname); 496 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 497 goto error; 498 } 499 500 if (zpool_prop_readonly(prop)) { 501 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' " 502 "is readonly"), propname); 503 (void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf); 504 goto error; 505 } 506 507 if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_POOL, retprops, 508 &strval, &intval, errbuf) != 0) 509 goto error; 510 511 /* 512 * Perform additional checking for specific properties. 513 */ 514 switch (prop) { 515 case ZPOOL_PROP_VERSION: 516 if (intval < version || 517 !SPA_VERSION_IS_SUPPORTED(intval)) { 518 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 519 "property '%s' number %d is invalid."), 520 propname, intval); 521 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf); 522 goto error; 523 } 524 break; 525 526 case ZPOOL_PROP_BOOTSIZE: 527 if (!flags.create) { 528 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 529 "property '%s' can only be set during pool " 530 "creation"), propname); 531 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 532 goto error; 533 } 534 break; 535 536 case ZPOOL_PROP_ASHIFT: 537 if (intval != 0 && 538 (intval < ASHIFT_MIN || intval > ASHIFT_MAX)) { 539 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 540 "invalid '%s=%d' property: only values " 541 "between %" PRId32 " and %" PRId32 " " 542 "are allowed.\n"), 543 propname, intval, ASHIFT_MIN, ASHIFT_MAX); 544 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 545 goto error; 546 } 547 break; 548 549 case ZPOOL_PROP_BOOTFS: 550 if (flags.create || flags.import) { 551 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 552 "property '%s' cannot be set at creation " 553 "or import time"), propname); 554 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 555 goto error; 556 } 557 558 if (version < SPA_VERSION_BOOTFS) { 559 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 560 "pool must be upgraded to support " 561 "'%s' property"), propname); 562 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf); 563 goto error; 564 } 565 566 /* 567 * bootfs property value has to be a dataset name and 568 * the dataset has to be in the same pool as it sets to. 569 */ 570 if (!bootfs_name_valid(poolname, strval)) { 571 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' " 572 "is an invalid name"), strval); 573 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf); 574 goto error; 575 } 576 577 if ((zhp = zpool_open_canfail(hdl, poolname)) == NULL) { 578 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 579 "could not open pool '%s'"), poolname); 580 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf); 581 goto error; 582 } 583 zpool_close(zhp); 584 break; 585 586 case ZPOOL_PROP_ALTROOT: 587 if (!flags.create && !flags.import) { 588 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 589 "property '%s' can only be set during pool " 590 "creation or import"), propname); 591 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 592 goto error; 593 } 594 595 if (strval[0] != '/') { 596 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 597 "bad alternate root '%s'"), strval); 598 (void) zfs_error(hdl, EZFS_BADPATH, errbuf); 599 goto error; 600 } 601 break; 602 603 case ZPOOL_PROP_CACHEFILE: 604 if (strval[0] == '\0') 605 break; 606 607 if (strcmp(strval, "none") == 0) 608 break; 609 610 if (strval[0] != '/') { 611 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 612 "property '%s' must be empty, an " 613 "absolute path, or 'none'"), propname); 614 (void) zfs_error(hdl, EZFS_BADPATH, errbuf); 615 goto error; 616 } 617 618 slash = strrchr(strval, '/'); 619 620 if (slash[1] == '\0' || strcmp(slash, "/.") == 0 || 621 strcmp(slash, "/..") == 0) { 622 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 623 "'%s' is not a valid file"), strval); 624 (void) zfs_error(hdl, EZFS_BADPATH, errbuf); 625 goto error; 626 } 627 628 *slash = '\0'; 629 630 if (strval[0] != '\0' && 631 (stat64(strval, &statbuf) != 0 || 632 !S_ISDIR(statbuf.st_mode))) { 633 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 634 "'%s' is not a valid directory"), 635 strval); 636 (void) zfs_error(hdl, EZFS_BADPATH, errbuf); 637 goto error; 638 } 639 640 *slash = '/'; 641 break; 642 643 case ZPOOL_PROP_COMMENT: 644 for (check = strval; *check != '\0'; check++) { 645 if (!isprint(*check)) { 646 zfs_error_aux(hdl, 647 dgettext(TEXT_DOMAIN, 648 "comment may only have printable " 649 "characters")); 650 (void) zfs_error(hdl, EZFS_BADPROP, 651 errbuf); 652 goto error; 653 } 654 } 655 if (strlen(strval) > ZPROP_MAX_COMMENT) { 656 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 657 "comment must not exceed %d characters"), 658 ZPROP_MAX_COMMENT); 659 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 660 goto error; 661 } 662 break; 663 664 case ZPOOL_PROP_READONLY: 665 if (!flags.import) { 666 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 667 "property '%s' can only be set at " 668 "import time"), propname); 669 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 670 goto error; 671 } 672 break; 673 674 case ZPOOL_PROP_TNAME: 675 if (!flags.create) { 676 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 677 "property '%s' can only be set at " 678 "creation time"), propname); 679 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 680 goto error; 681 } 682 break; 683 684 case ZPOOL_PROP_MULTIHOST: 685 if (get_system_hostid() == 0) { 686 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 687 "requires a non-zero system hostid")); 688 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 689 goto error; 690 } 691 break; 692 693 default: 694 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 695 "property '%s'(%d) not defined"), propname, prop); 696 break; 697 } 698 } 699 700 return (retprops); 701 error: 702 nvlist_free(retprops); 703 return (NULL); 704 } 705 706 /* 707 * Set zpool property : propname=propval. 708 */ 709 int 710 zpool_set_prop(zpool_handle_t *zhp, const char *propname, const char *propval) 711 { 712 zfs_cmd_t zc = { 0 }; 713 int ret = -1; 714 char errbuf[1024]; 715 nvlist_t *nvl = NULL; 716 nvlist_t *realprops; 717 uint64_t version; 718 prop_flags_t flags = { 0 }; 719 720 (void) snprintf(errbuf, sizeof (errbuf), 721 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"), 722 zhp->zpool_name); 723 724 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0) 725 return (no_memory(zhp->zpool_hdl)); 726 727 if (nvlist_add_string(nvl, propname, propval) != 0) { 728 nvlist_free(nvl); 729 return (no_memory(zhp->zpool_hdl)); 730 } 731 732 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL); 733 if ((realprops = zpool_valid_proplist(zhp->zpool_hdl, 734 zhp->zpool_name, nvl, version, flags, errbuf)) == NULL) { 735 nvlist_free(nvl); 736 return (-1); 737 } 738 739 nvlist_free(nvl); 740 nvl = realprops; 741 742 /* 743 * Execute the corresponding ioctl() to set this property. 744 */ 745 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 746 747 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) { 748 nvlist_free(nvl); 749 return (-1); 750 } 751 752 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc); 753 754 zcmd_free_nvlists(&zc); 755 nvlist_free(nvl); 756 757 if (ret) 758 (void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf); 759 else 760 (void) zpool_props_refresh(zhp); 761 762 return (ret); 763 } 764 765 int 766 zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp) 767 { 768 libzfs_handle_t *hdl = zhp->zpool_hdl; 769 zprop_list_t *entry; 770 char buf[ZFS_MAXPROPLEN]; 771 nvlist_t *features = NULL; 772 zprop_list_t **last; 773 boolean_t firstexpand = (NULL == *plp); 774 775 if (zprop_expand_list(hdl, plp, ZFS_TYPE_POOL) != 0) 776 return (-1); 777 778 last = plp; 779 while (*last != NULL) 780 last = &(*last)->pl_next; 781 782 if ((*plp)->pl_all) 783 features = zpool_get_features(zhp); 784 785 if ((*plp)->pl_all && firstexpand) { 786 for (int i = 0; i < SPA_FEATURES; i++) { 787 zprop_list_t *entry = zfs_alloc(hdl, 788 sizeof (zprop_list_t)); 789 entry->pl_prop = ZPROP_INVAL; 790 entry->pl_user_prop = zfs_asprintf(hdl, "feature@%s", 791 spa_feature_table[i].fi_uname); 792 entry->pl_width = strlen(entry->pl_user_prop); 793 entry->pl_all = B_TRUE; 794 795 *last = entry; 796 last = &entry->pl_next; 797 } 798 } 799 800 /* add any unsupported features */ 801 for (nvpair_t *nvp = nvlist_next_nvpair(features, NULL); 802 nvp != NULL; nvp = nvlist_next_nvpair(features, nvp)) { 803 char *propname; 804 boolean_t found; 805 zprop_list_t *entry; 806 807 if (zfeature_is_supported(nvpair_name(nvp))) 808 continue; 809 810 propname = zfs_asprintf(hdl, "unsupported@%s", 811 nvpair_name(nvp)); 812 813 /* 814 * Before adding the property to the list make sure that no 815 * other pool already added the same property. 816 */ 817 found = B_FALSE; 818 entry = *plp; 819 while (entry != NULL) { 820 if (entry->pl_user_prop != NULL && 821 strcmp(propname, entry->pl_user_prop) == 0) { 822 found = B_TRUE; 823 break; 824 } 825 entry = entry->pl_next; 826 } 827 if (found) { 828 free(propname); 829 continue; 830 } 831 832 entry = zfs_alloc(hdl, sizeof (zprop_list_t)); 833 entry->pl_prop = ZPROP_INVAL; 834 entry->pl_user_prop = propname; 835 entry->pl_width = strlen(entry->pl_user_prop); 836 entry->pl_all = B_TRUE; 837 838 *last = entry; 839 last = &entry->pl_next; 840 } 841 842 for (entry = *plp; entry != NULL; entry = entry->pl_next) { 843 844 if (entry->pl_fixed) 845 continue; 846 847 if (entry->pl_prop != ZPROP_INVAL && 848 zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf), 849 NULL, B_FALSE) == 0) { 850 if (strlen(buf) > entry->pl_width) 851 entry->pl_width = strlen(buf); 852 } 853 } 854 855 return (0); 856 } 857 858 /* 859 * Get the state for the given feature on the given ZFS pool. 860 */ 861 int 862 zpool_prop_get_feature(zpool_handle_t *zhp, const char *propname, char *buf, 863 size_t len) 864 { 865 uint64_t refcount; 866 boolean_t found = B_FALSE; 867 nvlist_t *features = zpool_get_features(zhp); 868 boolean_t supported; 869 const char *feature = strchr(propname, '@') + 1; 870 871 supported = zpool_prop_feature(propname); 872 ASSERT(supported || zpool_prop_unsupported(propname)); 873 874 /* 875 * Convert from feature name to feature guid. This conversion is 876 * unecessary for unsupported@... properties because they already 877 * use guids. 878 */ 879 if (supported) { 880 int ret; 881 spa_feature_t fid; 882 883 ret = zfeature_lookup_name(feature, &fid); 884 if (ret != 0) { 885 (void) strlcpy(buf, "-", len); 886 return (ENOTSUP); 887 } 888 feature = spa_feature_table[fid].fi_guid; 889 } 890 891 if (nvlist_lookup_uint64(features, feature, &refcount) == 0) 892 found = B_TRUE; 893 894 if (supported) { 895 if (!found) { 896 (void) strlcpy(buf, ZFS_FEATURE_DISABLED, len); 897 } else { 898 if (refcount == 0) 899 (void) strlcpy(buf, ZFS_FEATURE_ENABLED, len); 900 else 901 (void) strlcpy(buf, ZFS_FEATURE_ACTIVE, len); 902 } 903 } else { 904 if (found) { 905 if (refcount == 0) { 906 (void) strcpy(buf, ZFS_UNSUPPORTED_INACTIVE); 907 } else { 908 (void) strcpy(buf, ZFS_UNSUPPORTED_READONLY); 909 } 910 } else { 911 (void) strlcpy(buf, "-", len); 912 return (ENOTSUP); 913 } 914 } 915 916 return (0); 917 } 918 919 /* 920 * Don't start the slice at the default block of 34; many storage 921 * devices will use a stripe width of 128k, so start there instead. 922 */ 923 #define NEW_START_BLOCK 256 924 925 /* 926 * Validate the given pool name, optionally putting an extended error message in 927 * 'buf'. 928 */ 929 boolean_t 930 zpool_name_valid(libzfs_handle_t *hdl, boolean_t isopen, const char *pool) 931 { 932 namecheck_err_t why; 933 char what; 934 int ret; 935 936 ret = pool_namecheck(pool, &why, &what); 937 938 /* 939 * The rules for reserved pool names were extended at a later point. 940 * But we need to support users with existing pools that may now be 941 * invalid. So we only check for this expanded set of names during a 942 * create (or import), and only in userland. 943 */ 944 if (ret == 0 && !isopen && 945 (strncmp(pool, "mirror", 6) == 0 || 946 strncmp(pool, "raidz", 5) == 0 || 947 strncmp(pool, "spare", 5) == 0 || 948 strcmp(pool, "log") == 0)) { 949 if (hdl != NULL) 950 zfs_error_aux(hdl, 951 dgettext(TEXT_DOMAIN, "name is reserved")); 952 return (B_FALSE); 953 } 954 955 956 if (ret != 0) { 957 if (hdl != NULL) { 958 switch (why) { 959 case NAME_ERR_TOOLONG: 960 zfs_error_aux(hdl, 961 dgettext(TEXT_DOMAIN, "name is too long")); 962 break; 963 964 case NAME_ERR_INVALCHAR: 965 zfs_error_aux(hdl, 966 dgettext(TEXT_DOMAIN, "invalid character " 967 "'%c' in pool name"), what); 968 break; 969 970 case NAME_ERR_NOLETTER: 971 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 972 "name must begin with a letter")); 973 break; 974 975 case NAME_ERR_RESERVED: 976 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 977 "name is reserved")); 978 break; 979 980 case NAME_ERR_DISKLIKE: 981 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 982 "pool name is reserved")); 983 break; 984 985 case NAME_ERR_LEADING_SLASH: 986 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 987 "leading slash in name")); 988 break; 989 990 case NAME_ERR_EMPTY_COMPONENT: 991 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 992 "empty component in name")); 993 break; 994 995 case NAME_ERR_TRAILING_SLASH: 996 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 997 "trailing slash in name")); 998 break; 999 1000 case NAME_ERR_MULTIPLE_DELIMITERS: 1001 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1002 "multiple '@' and/or '#' delimiters in " 1003 "name")); 1004 break; 1005 1006 default: 1007 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1008 "(%d) not defined"), why); 1009 break; 1010 } 1011 } 1012 return (B_FALSE); 1013 } 1014 1015 return (B_TRUE); 1016 } 1017 1018 /* 1019 * Open a handle to the given pool, even if the pool is currently in the FAULTED 1020 * state. 1021 */ 1022 zpool_handle_t * 1023 zpool_open_canfail(libzfs_handle_t *hdl, const char *pool) 1024 { 1025 zpool_handle_t *zhp; 1026 boolean_t missing; 1027 1028 /* 1029 * Make sure the pool name is valid. 1030 */ 1031 if (!zpool_name_valid(hdl, B_TRUE, pool)) { 1032 (void) zfs_error_fmt(hdl, EZFS_INVALIDNAME, 1033 dgettext(TEXT_DOMAIN, "cannot open '%s'"), 1034 pool); 1035 return (NULL); 1036 } 1037 1038 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL) 1039 return (NULL); 1040 1041 zhp->zpool_hdl = hdl; 1042 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name)); 1043 1044 if (zpool_refresh_stats(zhp, &missing) != 0) { 1045 zpool_close(zhp); 1046 return (NULL); 1047 } 1048 1049 if (missing) { 1050 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool")); 1051 (void) zfs_error_fmt(hdl, EZFS_NOENT, 1052 dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool); 1053 zpool_close(zhp); 1054 return (NULL); 1055 } 1056 1057 return (zhp); 1058 } 1059 1060 /* 1061 * Like the above, but silent on error. Used when iterating over pools (because 1062 * the configuration cache may be out of date). 1063 */ 1064 int 1065 zpool_open_silent(libzfs_handle_t *hdl, const char *pool, zpool_handle_t **ret) 1066 { 1067 zpool_handle_t *zhp; 1068 boolean_t missing; 1069 1070 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL) 1071 return (-1); 1072 1073 zhp->zpool_hdl = hdl; 1074 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name)); 1075 1076 if (zpool_refresh_stats(zhp, &missing) != 0) { 1077 zpool_close(zhp); 1078 return (-1); 1079 } 1080 1081 if (missing) { 1082 zpool_close(zhp); 1083 *ret = NULL; 1084 return (0); 1085 } 1086 1087 *ret = zhp; 1088 return (0); 1089 } 1090 1091 /* 1092 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted 1093 * state. 1094 */ 1095 zpool_handle_t * 1096 zpool_open(libzfs_handle_t *hdl, const char *pool) 1097 { 1098 zpool_handle_t *zhp; 1099 1100 if ((zhp = zpool_open_canfail(hdl, pool)) == NULL) 1101 return (NULL); 1102 1103 if (zhp->zpool_state == POOL_STATE_UNAVAIL) { 1104 (void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL, 1105 dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name); 1106 zpool_close(zhp); 1107 return (NULL); 1108 } 1109 1110 return (zhp); 1111 } 1112 1113 /* 1114 * Close the handle. Simply frees the memory associated with the handle. 1115 */ 1116 void 1117 zpool_close(zpool_handle_t *zhp) 1118 { 1119 nvlist_free(zhp->zpool_config); 1120 nvlist_free(zhp->zpool_old_config); 1121 nvlist_free(zhp->zpool_props); 1122 free(zhp); 1123 } 1124 1125 /* 1126 * Return the name of the pool. 1127 */ 1128 const char * 1129 zpool_get_name(zpool_handle_t *zhp) 1130 { 1131 return (zhp->zpool_name); 1132 } 1133 1134 1135 /* 1136 * Return the state of the pool (ACTIVE or UNAVAILABLE) 1137 */ 1138 int 1139 zpool_get_state(zpool_handle_t *zhp) 1140 { 1141 return (zhp->zpool_state); 1142 } 1143 1144 /* 1145 * Check if vdev list contains a special vdev 1146 */ 1147 static boolean_t 1148 zpool_has_special_vdev(nvlist_t *nvroot) 1149 { 1150 nvlist_t **child; 1151 uint_t children; 1152 1153 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, &child, 1154 &children) == 0) { 1155 for (uint_t c = 0; c < children; c++) { 1156 char *bias; 1157 1158 if (nvlist_lookup_string(child[c], 1159 ZPOOL_CONFIG_ALLOCATION_BIAS, &bias) == 0 && 1160 strcmp(bias, VDEV_ALLOC_BIAS_SPECIAL) == 0) { 1161 return (B_TRUE); 1162 } 1163 } 1164 } 1165 return (B_FALSE); 1166 } 1167 1168 /* 1169 * Create the named pool, using the provided vdev list. It is assumed 1170 * that the consumer has already validated the contents of the nvlist, so we 1171 * don't have to worry about error semantics. 1172 */ 1173 int 1174 zpool_create(libzfs_handle_t *hdl, const char *pool, nvlist_t *nvroot, 1175 nvlist_t *props, nvlist_t *fsprops) 1176 { 1177 zfs_cmd_t zc = { 0 }; 1178 nvlist_t *zc_fsprops = NULL; 1179 nvlist_t *zc_props = NULL; 1180 nvlist_t *hidden_args = NULL; 1181 uint8_t *wkeydata = NULL; 1182 uint_t wkeylen = 0; 1183 char msg[1024]; 1184 int ret = -1; 1185 1186 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1187 "cannot create '%s'"), pool); 1188 1189 if (!zpool_name_valid(hdl, B_FALSE, pool)) 1190 return (zfs_error(hdl, EZFS_INVALIDNAME, msg)); 1191 1192 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0) 1193 return (-1); 1194 1195 if (props) { 1196 prop_flags_t flags = { .create = B_TRUE, .import = B_FALSE }; 1197 1198 if ((zc_props = zpool_valid_proplist(hdl, pool, props, 1199 SPA_VERSION_1, flags, msg)) == NULL) { 1200 goto create_failed; 1201 } 1202 } 1203 1204 if (fsprops) { 1205 uint64_t zoned; 1206 char *zonestr; 1207 1208 zoned = ((nvlist_lookup_string(fsprops, 1209 zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) && 1210 strcmp(zonestr, "on") == 0); 1211 1212 if ((zc_fsprops = zfs_valid_proplist(hdl, ZFS_TYPE_FILESYSTEM, 1213 fsprops, zoned, NULL, NULL, B_TRUE, msg)) == NULL) { 1214 goto create_failed; 1215 } 1216 1217 if (nvlist_exists(zc_fsprops, 1218 zfs_prop_to_name(ZFS_PROP_SPECIAL_SMALL_BLOCKS)) && 1219 !zpool_has_special_vdev(nvroot)) { 1220 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1221 "%s property requires a special vdev"), 1222 zfs_prop_to_name(ZFS_PROP_SPECIAL_SMALL_BLOCKS)); 1223 (void) zfs_error(hdl, EZFS_BADPROP, msg); 1224 goto create_failed; 1225 } 1226 1227 if (!zc_props && 1228 (nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) { 1229 goto create_failed; 1230 } 1231 if (zfs_crypto_create(hdl, NULL, zc_fsprops, props, B_TRUE, 1232 &wkeydata, &wkeylen) != 0) { 1233 (void) zfs_error(hdl, EZFS_CRYPTOFAILED, msg); 1234 goto create_failed; 1235 } 1236 if (nvlist_add_nvlist(zc_props, 1237 ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) { 1238 goto create_failed; 1239 } 1240 if (wkeydata != NULL) { 1241 if (nvlist_alloc(&hidden_args, NV_UNIQUE_NAME, 0) != 0) 1242 goto create_failed; 1243 1244 if (nvlist_add_uint8_array(hidden_args, "wkeydata", 1245 wkeydata, wkeylen) != 0) 1246 goto create_failed; 1247 1248 if (nvlist_add_nvlist(zc_props, ZPOOL_HIDDEN_ARGS, 1249 hidden_args) != 0) 1250 goto create_failed; 1251 } 1252 } 1253 1254 if (zc_props && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0) 1255 goto create_failed; 1256 1257 (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name)); 1258 1259 if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) { 1260 1261 zcmd_free_nvlists(&zc); 1262 nvlist_free(zc_props); 1263 nvlist_free(zc_fsprops); 1264 nvlist_free(hidden_args); 1265 if (wkeydata != NULL) 1266 free(wkeydata); 1267 1268 switch (errno) { 1269 case EBUSY: 1270 /* 1271 * This can happen if the user has specified the same 1272 * device multiple times. We can't reliably detect this 1273 * until we try to add it and see we already have a 1274 * label. 1275 */ 1276 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1277 "one or more vdevs refer to the same device")); 1278 return (zfs_error(hdl, EZFS_BADDEV, msg)); 1279 1280 case ERANGE: 1281 /* 1282 * This happens if the record size is smaller or larger 1283 * than the allowed size range, or not a power of 2. 1284 * 1285 * NOTE: although zfs_valid_proplist is called earlier, 1286 * this case may have slipped through since the 1287 * pool does not exist yet and it is therefore 1288 * impossible to read properties e.g. max blocksize 1289 * from the pool. 1290 */ 1291 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1292 "record size invalid")); 1293 return (zfs_error(hdl, EZFS_BADPROP, msg)); 1294 1295 case EOVERFLOW: 1296 /* 1297 * This occurs when one of the devices is below 1298 * SPA_MINDEVSIZE. Unfortunately, we can't detect which 1299 * device was the problem device since there's no 1300 * reliable way to determine device size from userland. 1301 */ 1302 { 1303 char buf[64]; 1304 1305 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf)); 1306 1307 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1308 "one or more devices is less than the " 1309 "minimum size (%s)"), buf); 1310 } 1311 return (zfs_error(hdl, EZFS_BADDEV, msg)); 1312 1313 case ENOSPC: 1314 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1315 "one or more devices is out of space")); 1316 return (zfs_error(hdl, EZFS_BADDEV, msg)); 1317 1318 default: 1319 return (zpool_standard_error(hdl, errno, msg)); 1320 } 1321 } 1322 1323 create_failed: 1324 zcmd_free_nvlists(&zc); 1325 nvlist_free(zc_props); 1326 nvlist_free(zc_fsprops); 1327 nvlist_free(hidden_args); 1328 if (wkeydata != NULL) 1329 free(wkeydata); 1330 return (ret); 1331 } 1332 1333 /* 1334 * Destroy the given pool. It is up to the caller to ensure that there are no 1335 * datasets left in the pool. 1336 */ 1337 int 1338 zpool_destroy(zpool_handle_t *zhp, const char *log_str) 1339 { 1340 zfs_cmd_t zc = { 0 }; 1341 zfs_handle_t *zfp = NULL; 1342 libzfs_handle_t *hdl = zhp->zpool_hdl; 1343 char msg[1024]; 1344 1345 if (zhp->zpool_state == POOL_STATE_ACTIVE && 1346 (zfp = zfs_open(hdl, zhp->zpool_name, ZFS_TYPE_FILESYSTEM)) == NULL) 1347 return (-1); 1348 1349 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1350 zc.zc_history = (uint64_t)(uintptr_t)log_str; 1351 1352 if (zfs_ioctl(hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) { 1353 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1354 "cannot destroy '%s'"), zhp->zpool_name); 1355 1356 if (errno == EROFS) { 1357 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1358 "one or more devices is read only")); 1359 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1360 } else { 1361 (void) zpool_standard_error(hdl, errno, msg); 1362 } 1363 1364 if (zfp) 1365 zfs_close(zfp); 1366 return (-1); 1367 } 1368 1369 if (zfp) { 1370 remove_mountpoint(zfp); 1371 zfs_close(zfp); 1372 } 1373 1374 return (0); 1375 } 1376 1377 /* 1378 * Create a checkpoint in the given pool. 1379 */ 1380 int 1381 zpool_checkpoint(zpool_handle_t *zhp) 1382 { 1383 libzfs_handle_t *hdl = zhp->zpool_hdl; 1384 char msg[1024]; 1385 int error; 1386 1387 error = lzc_pool_checkpoint(zhp->zpool_name); 1388 if (error != 0) { 1389 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1390 "cannot checkpoint '%s'"), zhp->zpool_name); 1391 (void) zpool_standard_error(hdl, error, msg); 1392 return (-1); 1393 } 1394 1395 return (0); 1396 } 1397 1398 /* 1399 * Discard the checkpoint from the given pool. 1400 */ 1401 int 1402 zpool_discard_checkpoint(zpool_handle_t *zhp) 1403 { 1404 libzfs_handle_t *hdl = zhp->zpool_hdl; 1405 char msg[1024]; 1406 int error; 1407 1408 error = lzc_pool_checkpoint_discard(zhp->zpool_name); 1409 if (error != 0) { 1410 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1411 "cannot discard checkpoint in '%s'"), zhp->zpool_name); 1412 (void) zpool_standard_error(hdl, error, msg); 1413 return (-1); 1414 } 1415 1416 return (0); 1417 } 1418 1419 /* 1420 * Add the given vdevs to the pool. The caller must have already performed the 1421 * necessary verification to ensure that the vdev specification is well-formed. 1422 */ 1423 int 1424 zpool_add(zpool_handle_t *zhp, nvlist_t *nvroot) 1425 { 1426 zfs_cmd_t zc = { 0 }; 1427 int ret; 1428 libzfs_handle_t *hdl = zhp->zpool_hdl; 1429 char msg[1024]; 1430 nvlist_t **spares, **l2cache; 1431 uint_t nspares, nl2cache; 1432 1433 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1434 "cannot add to '%s'"), zhp->zpool_name); 1435 1436 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) < 1437 SPA_VERSION_SPARES && 1438 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, 1439 &spares, &nspares) == 0) { 1440 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be " 1441 "upgraded to add hot spares")); 1442 return (zfs_error(hdl, EZFS_BADVERSION, msg)); 1443 } 1444 1445 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) < 1446 SPA_VERSION_L2CACHE && 1447 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, 1448 &l2cache, &nl2cache) == 0) { 1449 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be " 1450 "upgraded to add cache devices")); 1451 return (zfs_error(hdl, EZFS_BADVERSION, msg)); 1452 } 1453 1454 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0) 1455 return (-1); 1456 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1457 1458 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) { 1459 switch (errno) { 1460 case EBUSY: 1461 /* 1462 * This can happen if the user has specified the same 1463 * device multiple times. We can't reliably detect this 1464 * until we try to add it and see we already have a 1465 * label. 1466 */ 1467 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1468 "one or more vdevs refer to the same device")); 1469 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1470 break; 1471 1472 case EINVAL: 1473 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1474 "invalid config; a pool with removing/removed " 1475 "vdevs does not support adding raidz vdevs")); 1476 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1477 break; 1478 1479 case EOVERFLOW: 1480 /* 1481 * This occurrs when one of the devices is below 1482 * SPA_MINDEVSIZE. Unfortunately, we can't detect which 1483 * device was the problem device since there's no 1484 * reliable way to determine device size from userland. 1485 */ 1486 { 1487 char buf[64]; 1488 1489 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf)); 1490 1491 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1492 "device is less than the minimum " 1493 "size (%s)"), buf); 1494 } 1495 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1496 break; 1497 1498 case ENOTSUP: 1499 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1500 "pool must be upgraded to add these vdevs")); 1501 (void) zfs_error(hdl, EZFS_BADVERSION, msg); 1502 break; 1503 1504 case EDOM: 1505 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1506 "root pool can not have multiple vdevs" 1507 " or separate logs")); 1508 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, msg); 1509 break; 1510 1511 default: 1512 (void) zpool_standard_error(hdl, errno, msg); 1513 } 1514 1515 ret = -1; 1516 } else { 1517 ret = 0; 1518 } 1519 1520 zcmd_free_nvlists(&zc); 1521 1522 return (ret); 1523 } 1524 1525 /* 1526 * Exports the pool from the system. The caller must ensure that there are no 1527 * mounted datasets in the pool. 1528 */ 1529 static int 1530 zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce, 1531 const char *log_str) 1532 { 1533 zfs_cmd_t zc = { 0 }; 1534 char msg[1024]; 1535 1536 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1537 "cannot export '%s'"), zhp->zpool_name); 1538 1539 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1540 zc.zc_cookie = force; 1541 zc.zc_guid = hardforce; 1542 zc.zc_history = (uint64_t)(uintptr_t)log_str; 1543 1544 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) { 1545 switch (errno) { 1546 case EXDEV: 1547 zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN, 1548 "use '-f' to override the following errors:\n" 1549 "'%s' has an active shared spare which could be" 1550 " used by other pools once '%s' is exported."), 1551 zhp->zpool_name, zhp->zpool_name); 1552 return (zfs_error(zhp->zpool_hdl, EZFS_ACTIVE_SPARE, 1553 msg)); 1554 default: 1555 return (zpool_standard_error_fmt(zhp->zpool_hdl, errno, 1556 msg)); 1557 } 1558 } 1559 1560 return (0); 1561 } 1562 1563 int 1564 zpool_export(zpool_handle_t *zhp, boolean_t force, const char *log_str) 1565 { 1566 return (zpool_export_common(zhp, force, B_FALSE, log_str)); 1567 } 1568 1569 int 1570 zpool_export_force(zpool_handle_t *zhp, const char *log_str) 1571 { 1572 return (zpool_export_common(zhp, B_TRUE, B_TRUE, log_str)); 1573 } 1574 1575 static void 1576 zpool_rewind_exclaim(libzfs_handle_t *hdl, const char *name, boolean_t dryrun, 1577 nvlist_t *config) 1578 { 1579 nvlist_t *nv = NULL; 1580 uint64_t rewindto; 1581 int64_t loss = -1; 1582 struct tm t; 1583 char timestr[128]; 1584 1585 if (!hdl->libzfs_printerr || config == NULL) 1586 return; 1587 1588 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 || 1589 nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0) { 1590 return; 1591 } 1592 1593 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0) 1594 return; 1595 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss); 1596 1597 if (localtime_r((time_t *)&rewindto, &t) != NULL && 1598 strftime(timestr, 128, 0, &t) != 0) { 1599 if (dryrun) { 1600 (void) printf(dgettext(TEXT_DOMAIN, 1601 "Would be able to return %s " 1602 "to its state as of %s.\n"), 1603 name, timestr); 1604 } else { 1605 (void) printf(dgettext(TEXT_DOMAIN, 1606 "Pool %s returned to its state as of %s.\n"), 1607 name, timestr); 1608 } 1609 if (loss > 120) { 1610 (void) printf(dgettext(TEXT_DOMAIN, 1611 "%s approximately %lld "), 1612 dryrun ? "Would discard" : "Discarded", 1613 (loss + 30) / 60); 1614 (void) printf(dgettext(TEXT_DOMAIN, 1615 "minutes of transactions.\n")); 1616 } else if (loss > 0) { 1617 (void) printf(dgettext(TEXT_DOMAIN, 1618 "%s approximately %lld "), 1619 dryrun ? "Would discard" : "Discarded", loss); 1620 (void) printf(dgettext(TEXT_DOMAIN, 1621 "seconds of transactions.\n")); 1622 } 1623 } 1624 } 1625 1626 void 1627 zpool_explain_recover(libzfs_handle_t *hdl, const char *name, int reason, 1628 nvlist_t *config) 1629 { 1630 nvlist_t *nv = NULL; 1631 int64_t loss = -1; 1632 uint64_t edata = UINT64_MAX; 1633 uint64_t rewindto; 1634 struct tm t; 1635 char timestr[128]; 1636 1637 if (!hdl->libzfs_printerr) 1638 return; 1639 1640 if (reason >= 0) 1641 (void) printf(dgettext(TEXT_DOMAIN, "action: ")); 1642 else 1643 (void) printf(dgettext(TEXT_DOMAIN, "\t")); 1644 1645 /* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */ 1646 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 || 1647 nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0 || 1648 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0) 1649 goto no_info; 1650 1651 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss); 1652 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_DATA_ERRORS, 1653 &edata); 1654 1655 (void) printf(dgettext(TEXT_DOMAIN, 1656 "Recovery is possible, but will result in some data loss.\n")); 1657 1658 if (localtime_r((time_t *)&rewindto, &t) != NULL && 1659 strftime(timestr, 128, 0, &t) != 0) { 1660 (void) printf(dgettext(TEXT_DOMAIN, 1661 "\tReturning the pool to its state as of %s\n" 1662 "\tshould correct the problem. "), 1663 timestr); 1664 } else { 1665 (void) printf(dgettext(TEXT_DOMAIN, 1666 "\tReverting the pool to an earlier state " 1667 "should correct the problem.\n\t")); 1668 } 1669 1670 if (loss > 120) { 1671 (void) printf(dgettext(TEXT_DOMAIN, 1672 "Approximately %lld minutes of data\n" 1673 "\tmust be discarded, irreversibly. "), (loss + 30) / 60); 1674 } else if (loss > 0) { 1675 (void) printf(dgettext(TEXT_DOMAIN, 1676 "Approximately %lld seconds of data\n" 1677 "\tmust be discarded, irreversibly. "), loss); 1678 } 1679 if (edata != 0 && edata != UINT64_MAX) { 1680 if (edata == 1) { 1681 (void) printf(dgettext(TEXT_DOMAIN, 1682 "After rewind, at least\n" 1683 "\tone persistent user-data error will remain. ")); 1684 } else { 1685 (void) printf(dgettext(TEXT_DOMAIN, 1686 "After rewind, several\n" 1687 "\tpersistent user-data errors will remain. ")); 1688 } 1689 } 1690 (void) printf(dgettext(TEXT_DOMAIN, 1691 "Recovery can be attempted\n\tby executing 'zpool %s -F %s'. "), 1692 reason >= 0 ? "clear" : "import", name); 1693 1694 (void) printf(dgettext(TEXT_DOMAIN, 1695 "A scrub of the pool\n" 1696 "\tis strongly recommended after recovery.\n")); 1697 return; 1698 1699 no_info: 1700 (void) printf(dgettext(TEXT_DOMAIN, 1701 "Destroy and re-create the pool from\n\ta backup source.\n")); 1702 } 1703 1704 /* 1705 * zpool_import() is a contracted interface. Should be kept the same 1706 * if possible. 1707 * 1708 * Applications should use zpool_import_props() to import a pool with 1709 * new properties value to be set. 1710 */ 1711 int 1712 zpool_import(libzfs_handle_t *hdl, nvlist_t *config, const char *newname, 1713 char *altroot) 1714 { 1715 nvlist_t *props = NULL; 1716 int ret; 1717 1718 if (altroot != NULL) { 1719 if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) { 1720 return (zfs_error_fmt(hdl, EZFS_NOMEM, 1721 dgettext(TEXT_DOMAIN, "cannot import '%s'"), 1722 newname)); 1723 } 1724 1725 if (nvlist_add_string(props, 1726 zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 || 1727 nvlist_add_string(props, 1728 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) { 1729 nvlist_free(props); 1730 return (zfs_error_fmt(hdl, EZFS_NOMEM, 1731 dgettext(TEXT_DOMAIN, "cannot import '%s'"), 1732 newname)); 1733 } 1734 } 1735 1736 ret = zpool_import_props(hdl, config, newname, props, 1737 ZFS_IMPORT_NORMAL); 1738 nvlist_free(props); 1739 return (ret); 1740 } 1741 1742 static void 1743 print_vdev_tree(libzfs_handle_t *hdl, const char *name, nvlist_t *nv, 1744 int indent) 1745 { 1746 nvlist_t **child; 1747 uint_t c, children; 1748 char *vname; 1749 uint64_t is_log = 0; 1750 1751 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG, 1752 &is_log); 1753 1754 if (name != NULL) 1755 (void) printf("\t%*s%s%s\n", indent, "", name, 1756 is_log ? " [log]" : ""); 1757 1758 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, 1759 &child, &children) != 0) 1760 return; 1761 1762 for (c = 0; c < children; c++) { 1763 vname = zpool_vdev_name(hdl, NULL, child[c], VDEV_NAME_TYPE_ID); 1764 print_vdev_tree(hdl, vname, child[c], indent + 2); 1765 free(vname); 1766 } 1767 } 1768 1769 void 1770 zpool_print_unsup_feat(nvlist_t *config) 1771 { 1772 nvlist_t *nvinfo, *unsup_feat; 1773 1774 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 1775 0); 1776 verify(nvlist_lookup_nvlist(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT, 1777 &unsup_feat) == 0); 1778 1779 for (nvpair_t *nvp = nvlist_next_nvpair(unsup_feat, NULL); nvp != NULL; 1780 nvp = nvlist_next_nvpair(unsup_feat, nvp)) { 1781 char *desc; 1782 1783 verify(nvpair_type(nvp) == DATA_TYPE_STRING); 1784 verify(nvpair_value_string(nvp, &desc) == 0); 1785 1786 if (strlen(desc) > 0) 1787 (void) printf("\t%s (%s)\n", nvpair_name(nvp), desc); 1788 else 1789 (void) printf("\t%s\n", nvpair_name(nvp)); 1790 } 1791 } 1792 1793 /* 1794 * Import the given pool using the known configuration and a list of 1795 * properties to be set. The configuration should have come from 1796 * zpool_find_import(). The 'newname' parameters control whether the pool 1797 * is imported with a different name. 1798 */ 1799 int 1800 zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname, 1801 nvlist_t *props, int flags) 1802 { 1803 zfs_cmd_t zc = { 0 }; 1804 zpool_load_policy_t policy; 1805 nvlist_t *nv = NULL; 1806 nvlist_t *nvinfo = NULL; 1807 nvlist_t *missing = NULL; 1808 char *thename; 1809 char *origname; 1810 int ret; 1811 int error = 0; 1812 char errbuf[1024]; 1813 1814 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, 1815 &origname) == 0); 1816 1817 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, 1818 "cannot import pool '%s'"), origname); 1819 1820 if (newname != NULL) { 1821 if (!zpool_name_valid(hdl, B_FALSE, newname)) 1822 return (zfs_error_fmt(hdl, EZFS_INVALIDNAME, 1823 dgettext(TEXT_DOMAIN, "cannot import '%s'"), 1824 newname)); 1825 thename = (char *)newname; 1826 } else { 1827 thename = origname; 1828 } 1829 1830 if (props != NULL) { 1831 uint64_t version; 1832 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE }; 1833 1834 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, 1835 &version) == 0); 1836 1837 if ((props = zpool_valid_proplist(hdl, origname, 1838 props, version, flags, errbuf)) == NULL) 1839 return (-1); 1840 if (zcmd_write_src_nvlist(hdl, &zc, props) != 0) { 1841 nvlist_free(props); 1842 return (-1); 1843 } 1844 nvlist_free(props); 1845 } 1846 1847 (void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name)); 1848 1849 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, 1850 &zc.zc_guid) == 0); 1851 1852 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) { 1853 zcmd_free_nvlists(&zc); 1854 return (-1); 1855 } 1856 if (zcmd_alloc_dst_nvlist(hdl, &zc, zc.zc_nvlist_conf_size * 2) != 0) { 1857 zcmd_free_nvlists(&zc); 1858 return (-1); 1859 } 1860 1861 zc.zc_cookie = flags; 1862 while ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc)) != 0 && 1863 errno == ENOMEM) { 1864 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) { 1865 zcmd_free_nvlists(&zc); 1866 return (-1); 1867 } 1868 } 1869 if (ret != 0) 1870 error = errno; 1871 1872 (void) zcmd_read_dst_nvlist(hdl, &zc, &nv); 1873 1874 zcmd_free_nvlists(&zc); 1875 1876 zpool_get_load_policy(config, &policy); 1877 1878 if (error) { 1879 char desc[1024]; 1880 char aux[256]; 1881 1882 /* 1883 * Dry-run failed, but we print out what success 1884 * looks like if we found a best txg 1885 */ 1886 if (policy.zlp_rewind & ZPOOL_TRY_REWIND) { 1887 zpool_rewind_exclaim(hdl, newname ? origname : thename, 1888 B_TRUE, nv); 1889 nvlist_free(nv); 1890 return (-1); 1891 } 1892 1893 if (newname == NULL) 1894 (void) snprintf(desc, sizeof (desc), 1895 dgettext(TEXT_DOMAIN, "cannot import '%s'"), 1896 thename); 1897 else 1898 (void) snprintf(desc, sizeof (desc), 1899 dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"), 1900 origname, thename); 1901 1902 switch (error) { 1903 case ENOTSUP: 1904 if (nv != NULL && nvlist_lookup_nvlist(nv, 1905 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 && 1906 nvlist_exists(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT)) { 1907 (void) printf(dgettext(TEXT_DOMAIN, "This " 1908 "pool uses the following feature(s) not " 1909 "supported by this system:\n")); 1910 zpool_print_unsup_feat(nv); 1911 if (nvlist_exists(nvinfo, 1912 ZPOOL_CONFIG_CAN_RDONLY)) { 1913 (void) printf(dgettext(TEXT_DOMAIN, 1914 "All unsupported features are only " 1915 "required for writing to the pool." 1916 "\nThe pool can be imported using " 1917 "'-o readonly=on'.\n")); 1918 } 1919 } 1920 /* 1921 * Unsupported version. 1922 */ 1923 (void) zfs_error(hdl, EZFS_BADVERSION, desc); 1924 break; 1925 1926 case EREMOTEIO: 1927 if (nv != NULL && nvlist_lookup_nvlist(nv, 1928 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0) { 1929 char *hostname = "<unknown>"; 1930 uint64_t hostid = 0; 1931 mmp_state_t mmp_state; 1932 1933 mmp_state = fnvlist_lookup_uint64(nvinfo, 1934 ZPOOL_CONFIG_MMP_STATE); 1935 1936 if (nvlist_exists(nvinfo, 1937 ZPOOL_CONFIG_MMP_HOSTNAME)) 1938 hostname = fnvlist_lookup_string(nvinfo, 1939 ZPOOL_CONFIG_MMP_HOSTNAME); 1940 1941 if (nvlist_exists(nvinfo, 1942 ZPOOL_CONFIG_MMP_HOSTID)) 1943 hostid = fnvlist_lookup_uint64(nvinfo, 1944 ZPOOL_CONFIG_MMP_HOSTID); 1945 1946 if (mmp_state == MMP_STATE_ACTIVE) { 1947 (void) snprintf(aux, sizeof (aux), 1948 dgettext(TEXT_DOMAIN, "pool is imp" 1949 "orted on host '%s' (hostid=%lx).\n" 1950 "Export the pool on the other " 1951 "system, then run 'zpool import'."), 1952 hostname, (unsigned long) hostid); 1953 } else if (mmp_state == MMP_STATE_NO_HOSTID) { 1954 (void) snprintf(aux, sizeof (aux), 1955 dgettext(TEXT_DOMAIN, "pool has " 1956 "the multihost property on and " 1957 "the\nsystem's hostid is not " 1958 "set.\n")); 1959 } 1960 1961 (void) zfs_error_aux(hdl, aux); 1962 } 1963 (void) zfs_error(hdl, EZFS_ACTIVE_POOL, desc); 1964 break; 1965 1966 case EINVAL: 1967 (void) zfs_error(hdl, EZFS_INVALCONFIG, desc); 1968 break; 1969 1970 case EROFS: 1971 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1972 "one or more devices is read only")); 1973 (void) zfs_error(hdl, EZFS_BADDEV, desc); 1974 break; 1975 1976 case ENXIO: 1977 if (nv && nvlist_lookup_nvlist(nv, 1978 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 && 1979 nvlist_lookup_nvlist(nvinfo, 1980 ZPOOL_CONFIG_MISSING_DEVICES, &missing) == 0) { 1981 (void) printf(dgettext(TEXT_DOMAIN, 1982 "The devices below are missing or " 1983 "corrupted, use '-m' to import the pool " 1984 "anyway:\n")); 1985 print_vdev_tree(hdl, NULL, missing, 2); 1986 (void) printf("\n"); 1987 } 1988 (void) zpool_standard_error(hdl, error, desc); 1989 break; 1990 1991 case EEXIST: 1992 (void) zpool_standard_error(hdl, error, desc); 1993 break; 1994 case ENAMETOOLONG: 1995 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1996 "new name of at least one dataset is longer than " 1997 "the maximum allowable length")); 1998 (void) zfs_error(hdl, EZFS_NAMETOOLONG, desc); 1999 break; 2000 default: 2001 (void) zpool_standard_error(hdl, error, desc); 2002 zpool_explain_recover(hdl, 2003 newname ? origname : thename, -error, nv); 2004 break; 2005 } 2006 2007 nvlist_free(nv); 2008 ret = -1; 2009 } else { 2010 zpool_handle_t *zhp; 2011 2012 /* 2013 * This should never fail, but play it safe anyway. 2014 */ 2015 if (zpool_open_silent(hdl, thename, &zhp) != 0) 2016 ret = -1; 2017 else if (zhp != NULL) 2018 zpool_close(zhp); 2019 if (policy.zlp_rewind & 2020 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) { 2021 zpool_rewind_exclaim(hdl, newname ? origname : thename, 2022 ((policy.zlp_rewind & ZPOOL_TRY_REWIND) != 0), nv); 2023 } 2024 nvlist_free(nv); 2025 return (0); 2026 } 2027 2028 return (ret); 2029 } 2030 2031 /* 2032 * Scan the pool. 2033 */ 2034 int 2035 zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func, pool_scrub_cmd_t cmd) 2036 { 2037 zfs_cmd_t zc = { 0 }; 2038 char msg[1024]; 2039 int err; 2040 libzfs_handle_t *hdl = zhp->zpool_hdl; 2041 2042 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2043 zc.zc_cookie = func; 2044 zc.zc_flags = cmd; 2045 2046 if (zfs_ioctl(hdl, ZFS_IOC_POOL_SCAN, &zc) == 0) 2047 return (0); 2048 2049 err = errno; 2050 2051 /* ECANCELED on a scrub means we resumed a paused scrub */ 2052 if (err == ECANCELED && func == POOL_SCAN_SCRUB && 2053 cmd == POOL_SCRUB_NORMAL) 2054 return (0); 2055 2056 if (err == ENOENT && func != POOL_SCAN_NONE && cmd == POOL_SCRUB_NORMAL) 2057 return (0); 2058 2059 if (func == POOL_SCAN_SCRUB) { 2060 if (cmd == POOL_SCRUB_PAUSE) { 2061 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 2062 "cannot pause scrubbing %s"), zc.zc_name); 2063 } else { 2064 assert(cmd == POOL_SCRUB_NORMAL); 2065 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 2066 "cannot scrub %s"), zc.zc_name); 2067 } 2068 } else if (func == POOL_SCAN_RESILVER) { 2069 assert(cmd == POOL_SCRUB_NORMAL); 2070 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 2071 "cannot restart resilver on %s"), zc.zc_name); 2072 } else if (func == POOL_SCAN_NONE) { 2073 (void) snprintf(msg, sizeof (msg), 2074 dgettext(TEXT_DOMAIN, "cannot cancel scrubbing %s"), 2075 zc.zc_name); 2076 } else { 2077 assert(!"unexpected result"); 2078 } 2079 2080 if (err == EBUSY) { 2081 nvlist_t *nvroot; 2082 pool_scan_stat_t *ps = NULL; 2083 uint_t psc; 2084 2085 verify(nvlist_lookup_nvlist(zhp->zpool_config, 2086 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 2087 (void) nvlist_lookup_uint64_array(nvroot, 2088 ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &psc); 2089 if (ps && ps->pss_func == POOL_SCAN_SCRUB) { 2090 if (cmd == POOL_SCRUB_PAUSE) 2091 return (zfs_error(hdl, EZFS_SCRUB_PAUSED, msg)); 2092 else 2093 return (zfs_error(hdl, EZFS_SCRUBBING, msg)); 2094 } else { 2095 return (zfs_error(hdl, EZFS_RESILVERING, msg)); 2096 } 2097 } else if (err == ENOENT) { 2098 return (zfs_error(hdl, EZFS_NO_SCRUB, msg)); 2099 } else if (err == ENOTSUP && func == POOL_SCAN_RESILVER) { 2100 return (zfs_error(hdl, EZFS_NO_RESILVER_DEFER, msg)); 2101 } else { 2102 return (zpool_standard_error(hdl, err, msg)); 2103 } 2104 } 2105 2106 static int 2107 xlate_init_err(int err) 2108 { 2109 switch (err) { 2110 case ENODEV: 2111 return (EZFS_NODEVICE); 2112 case EINVAL: 2113 case EROFS: 2114 return (EZFS_BADDEV); 2115 case EBUSY: 2116 return (EZFS_INITIALIZING); 2117 case ESRCH: 2118 return (EZFS_NO_INITIALIZE); 2119 } 2120 return (err); 2121 } 2122 2123 /* 2124 * Begin, suspend, or cancel the initialization (initializing of all free 2125 * blocks) for the given vdevs in the given pool. 2126 */ 2127 int 2128 zpool_initialize(zpool_handle_t *zhp, pool_initialize_func_t cmd_type, 2129 nvlist_t *vds) 2130 { 2131 char msg[1024]; 2132 libzfs_handle_t *hdl = zhp->zpool_hdl; 2133 2134 nvlist_t *errlist; 2135 2136 /* translate vdev names to guids */ 2137 nvlist_t *vdev_guids = fnvlist_alloc(); 2138 nvlist_t *guids_to_paths = fnvlist_alloc(); 2139 boolean_t spare, cache; 2140 nvlist_t *tgt; 2141 nvpair_t *elem; 2142 2143 for (elem = nvlist_next_nvpair(vds, NULL); elem != NULL; 2144 elem = nvlist_next_nvpair(vds, elem)) { 2145 char *vd_path = nvpair_name(elem); 2146 tgt = zpool_find_vdev(zhp, vd_path, &spare, &cache, NULL); 2147 2148 if ((tgt == NULL) || cache || spare) { 2149 (void) snprintf(msg, sizeof (msg), 2150 dgettext(TEXT_DOMAIN, "cannot initialize '%s'"), 2151 vd_path); 2152 int err = (tgt == NULL) ? EZFS_NODEVICE : 2153 (spare ? EZFS_ISSPARE : EZFS_ISL2CACHE); 2154 fnvlist_free(vdev_guids); 2155 fnvlist_free(guids_to_paths); 2156 return (zfs_error(hdl, err, msg)); 2157 } 2158 2159 uint64_t guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID); 2160 fnvlist_add_uint64(vdev_guids, vd_path, guid); 2161 2162 (void) snprintf(msg, sizeof (msg), "%llu", guid); 2163 fnvlist_add_string(guids_to_paths, msg, vd_path); 2164 } 2165 2166 int err = lzc_initialize(zhp->zpool_name, cmd_type, vdev_guids, 2167 &errlist); 2168 fnvlist_free(vdev_guids); 2169 2170 if (err == 0) { 2171 fnvlist_free(guids_to_paths); 2172 return (0); 2173 } 2174 2175 nvlist_t *vd_errlist = NULL; 2176 if (errlist != NULL) { 2177 vd_errlist = fnvlist_lookup_nvlist(errlist, 2178 ZPOOL_INITIALIZE_VDEVS); 2179 } 2180 2181 (void) snprintf(msg, sizeof (msg), 2182 dgettext(TEXT_DOMAIN, "operation failed")); 2183 2184 for (elem = nvlist_next_nvpair(vd_errlist, NULL); elem != NULL; 2185 elem = nvlist_next_nvpair(vd_errlist, elem)) { 2186 int64_t vd_error = xlate_init_err(fnvpair_value_int64(elem)); 2187 char *path = fnvlist_lookup_string(guids_to_paths, 2188 nvpair_name(elem)); 2189 (void) zfs_error_fmt(hdl, vd_error, "cannot initialize '%s'", 2190 path); 2191 } 2192 2193 fnvlist_free(guids_to_paths); 2194 if (vd_errlist != NULL) 2195 return (-1); 2196 2197 return (zpool_standard_error(hdl, err, msg)); 2198 } 2199 2200 /* 2201 * This provides a very minimal check whether a given string is likely a 2202 * c#t#d# style string. Users of this are expected to do their own 2203 * verification of the s# part. 2204 */ 2205 #define CTD_CHECK(str) (str && str[0] == 'c' && isdigit(str[1])) 2206 2207 /* 2208 * More elaborate version for ones which may start with "/dev/dsk/" 2209 * and the like. 2210 */ 2211 static int 2212 ctd_check_path(char *str) 2213 { 2214 /* 2215 * If it starts with a slash, check the last component. 2216 */ 2217 if (str && str[0] == '/') { 2218 char *tmp = strrchr(str, '/'); 2219 2220 /* 2221 * If it ends in "/old", check the second-to-last 2222 * component of the string instead. 2223 */ 2224 if (tmp != str && strcmp(tmp, "/old") == 0) { 2225 for (tmp--; *tmp != '/'; tmp--) 2226 ; 2227 } 2228 str = tmp + 1; 2229 } 2230 return (CTD_CHECK(str)); 2231 } 2232 2233 /* 2234 * Find a vdev that matches the search criteria specified. We use the 2235 * the nvpair name to determine how we should look for the device. 2236 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL 2237 * spare; but FALSE if its an INUSE spare. 2238 */ 2239 static nvlist_t * 2240 vdev_to_nvlist_iter(nvlist_t *nv, nvlist_t *search, boolean_t *avail_spare, 2241 boolean_t *l2cache, boolean_t *log) 2242 { 2243 uint_t c, children; 2244 nvlist_t **child; 2245 nvlist_t *ret; 2246 uint64_t is_log; 2247 char *srchkey; 2248 nvpair_t *pair = nvlist_next_nvpair(search, NULL); 2249 2250 /* Nothing to look for */ 2251 if (search == NULL || pair == NULL) 2252 return (NULL); 2253 2254 /* Obtain the key we will use to search */ 2255 srchkey = nvpair_name(pair); 2256 2257 switch (nvpair_type(pair)) { 2258 case DATA_TYPE_UINT64: 2259 if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) { 2260 uint64_t srchval, theguid; 2261 2262 verify(nvpair_value_uint64(pair, &srchval) == 0); 2263 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, 2264 &theguid) == 0); 2265 if (theguid == srchval) 2266 return (nv); 2267 } 2268 break; 2269 2270 case DATA_TYPE_STRING: { 2271 char *srchval, *val; 2272 2273 verify(nvpair_value_string(pair, &srchval) == 0); 2274 if (nvlist_lookup_string(nv, srchkey, &val) != 0) 2275 break; 2276 2277 /* 2278 * Search for the requested value. Special cases: 2279 * 2280 * - ZPOOL_CONFIG_PATH for whole disk entries. To support 2281 * UEFI boot, these end in "s0" or "s0/old" or "s1" or 2282 * "s1/old". The "s0" or "s1" part is hidden from the user, 2283 * but included in the string, so this matches around it. 2284 * - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE). 2285 * 2286 * Otherwise, all other searches are simple string compares. 2287 */ 2288 if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0 && 2289 ctd_check_path(val)) { 2290 uint64_t wholedisk = 0; 2291 2292 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, 2293 &wholedisk); 2294 if (wholedisk) { 2295 int slen = strlen(srchval); 2296 int vlen = strlen(val); 2297 2298 if (slen != vlen - 2) 2299 break; 2300 2301 /* 2302 * make_leaf_vdev() should only set 2303 * wholedisk for ZPOOL_CONFIG_PATHs which 2304 * will include "/dev/dsk/", giving plenty of 2305 * room for the indices used next. 2306 */ 2307 ASSERT(vlen >= 6); 2308 2309 /* 2310 * strings identical except trailing "s0" 2311 */ 2312 if ((strcmp(&val[vlen - 2], "s0") == 0 || 2313 strcmp(&val[vlen - 2], "s1") == 0) && 2314 strncmp(srchval, val, slen) == 0) 2315 return (nv); 2316 2317 /* 2318 * strings identical except trailing "s0/old" 2319 */ 2320 if ((strcmp(&val[vlen - 6], "s0/old") == 0 || 2321 strcmp(&val[vlen - 6], "s1/old") == 0) && 2322 strcmp(&srchval[slen - 4], "/old") == 0 && 2323 strncmp(srchval, val, slen - 4) == 0) 2324 return (nv); 2325 2326 break; 2327 } 2328 } else if (strcmp(srchkey, ZPOOL_CONFIG_TYPE) == 0 && val) { 2329 char *type, *idx, *end, *p; 2330 uint64_t id, vdev_id; 2331 2332 /* 2333 * Determine our vdev type, keeping in mind 2334 * that the srchval is composed of a type and 2335 * vdev id pair (i.e. mirror-4). 2336 */ 2337 if ((type = strdup(srchval)) == NULL) 2338 return (NULL); 2339 2340 if ((p = strrchr(type, '-')) == NULL) { 2341 free(type); 2342 break; 2343 } 2344 idx = p + 1; 2345 *p = '\0'; 2346 2347 /* 2348 * If the types don't match then keep looking. 2349 */ 2350 if (strncmp(val, type, strlen(val)) != 0) { 2351 free(type); 2352 break; 2353 } 2354 2355 verify(zpool_vdev_is_interior(type)); 2356 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID, 2357 &id) == 0); 2358 2359 errno = 0; 2360 vdev_id = strtoull(idx, &end, 10); 2361 2362 free(type); 2363 if (errno != 0) 2364 return (NULL); 2365 2366 /* 2367 * Now verify that we have the correct vdev id. 2368 */ 2369 if (vdev_id == id) 2370 return (nv); 2371 } 2372 2373 /* 2374 * Common case 2375 */ 2376 if (strcmp(srchval, val) == 0) 2377 return (nv); 2378 break; 2379 } 2380 2381 default: 2382 break; 2383 } 2384 2385 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, 2386 &child, &children) != 0) 2387 return (NULL); 2388 2389 for (c = 0; c < children; c++) { 2390 if ((ret = vdev_to_nvlist_iter(child[c], search, 2391 avail_spare, l2cache, NULL)) != NULL) { 2392 /* 2393 * The 'is_log' value is only set for the toplevel 2394 * vdev, not the leaf vdevs. So we always lookup the 2395 * log device from the root of the vdev tree (where 2396 * 'log' is non-NULL). 2397 */ 2398 if (log != NULL && 2399 nvlist_lookup_uint64(child[c], 2400 ZPOOL_CONFIG_IS_LOG, &is_log) == 0 && 2401 is_log) { 2402 *log = B_TRUE; 2403 } 2404 return (ret); 2405 } 2406 } 2407 2408 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES, 2409 &child, &children) == 0) { 2410 for (c = 0; c < children; c++) { 2411 if ((ret = vdev_to_nvlist_iter(child[c], search, 2412 avail_spare, l2cache, NULL)) != NULL) { 2413 *avail_spare = B_TRUE; 2414 return (ret); 2415 } 2416 } 2417 } 2418 2419 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE, 2420 &child, &children) == 0) { 2421 for (c = 0; c < children; c++) { 2422 if ((ret = vdev_to_nvlist_iter(child[c], search, 2423 avail_spare, l2cache, NULL)) != NULL) { 2424 *l2cache = B_TRUE; 2425 return (ret); 2426 } 2427 } 2428 } 2429 2430 return (NULL); 2431 } 2432 2433 /* 2434 * Given a physical path (minus the "/devices" prefix), find the 2435 * associated vdev. 2436 */ 2437 nvlist_t * 2438 zpool_find_vdev_by_physpath(zpool_handle_t *zhp, const char *ppath, 2439 boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log) 2440 { 2441 nvlist_t *search, *nvroot, *ret; 2442 2443 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2444 verify(nvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH, ppath) == 0); 2445 2446 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE, 2447 &nvroot) == 0); 2448 2449 *avail_spare = B_FALSE; 2450 *l2cache = B_FALSE; 2451 if (log != NULL) 2452 *log = B_FALSE; 2453 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log); 2454 nvlist_free(search); 2455 2456 return (ret); 2457 } 2458 2459 /* 2460 * Determine if we have an "interior" top-level vdev (i.e mirror/raidz). 2461 */ 2462 static boolean_t 2463 zpool_vdev_is_interior(const char *name) 2464 { 2465 if (strncmp(name, VDEV_TYPE_RAIDZ, strlen(VDEV_TYPE_RAIDZ)) == 0 || 2466 strncmp(name, VDEV_TYPE_SPARE, strlen(VDEV_TYPE_SPARE)) == 0 || 2467 strncmp(name, 2468 VDEV_TYPE_REPLACING, strlen(VDEV_TYPE_REPLACING)) == 0 || 2469 strncmp(name, VDEV_TYPE_MIRROR, strlen(VDEV_TYPE_MIRROR)) == 0) 2470 return (B_TRUE); 2471 return (B_FALSE); 2472 } 2473 2474 nvlist_t * 2475 zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare, 2476 boolean_t *l2cache, boolean_t *log) 2477 { 2478 char buf[MAXPATHLEN]; 2479 char *end; 2480 nvlist_t *nvroot, *search, *ret; 2481 uint64_t guid; 2482 2483 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2484 2485 guid = strtoull(path, &end, 10); 2486 if (guid != 0 && *end == '\0') { 2487 verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0); 2488 } else if (zpool_vdev_is_interior(path)) { 2489 verify(nvlist_add_string(search, ZPOOL_CONFIG_TYPE, path) == 0); 2490 } else if (path[0] != '/') { 2491 (void) snprintf(buf, sizeof (buf), "%s/%s", ZFS_DISK_ROOT, 2492 path); 2493 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, buf) == 0); 2494 } else { 2495 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, path) == 0); 2496 } 2497 2498 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE, 2499 &nvroot) == 0); 2500 2501 *avail_spare = B_FALSE; 2502 *l2cache = B_FALSE; 2503 if (log != NULL) 2504 *log = B_FALSE; 2505 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log); 2506 nvlist_free(search); 2507 2508 return (ret); 2509 } 2510 2511 static int 2512 vdev_is_online(nvlist_t *nv) 2513 { 2514 uint64_t ival; 2515 2516 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE, &ival) == 0 || 2517 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED, &ival) == 0 || 2518 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED, &ival) == 0) 2519 return (0); 2520 2521 return (1); 2522 } 2523 2524 /* 2525 * Helper function for zpool_get_physpaths(). 2526 */ 2527 static int 2528 vdev_get_one_physpath(nvlist_t *config, char *physpath, size_t physpath_size, 2529 size_t *bytes_written) 2530 { 2531 size_t bytes_left, pos, rsz; 2532 char *tmppath; 2533 const char *format; 2534 2535 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PHYS_PATH, 2536 &tmppath) != 0) 2537 return (EZFS_NODEVICE); 2538 2539 pos = *bytes_written; 2540 bytes_left = physpath_size - pos; 2541 format = (pos == 0) ? "%s" : " %s"; 2542 2543 rsz = snprintf(physpath + pos, bytes_left, format, tmppath); 2544 *bytes_written += rsz; 2545 2546 if (rsz >= bytes_left) { 2547 /* if physpath was not copied properly, clear it */ 2548 if (bytes_left != 0) { 2549 physpath[pos] = 0; 2550 } 2551 return (EZFS_NOSPC); 2552 } 2553 return (0); 2554 } 2555 2556 static int 2557 vdev_get_physpaths(nvlist_t *nv, char *physpath, size_t phypath_size, 2558 size_t *rsz, boolean_t is_spare) 2559 { 2560 char *type; 2561 int ret; 2562 2563 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0) 2564 return (EZFS_INVALCONFIG); 2565 2566 if (strcmp(type, VDEV_TYPE_DISK) == 0) { 2567 /* 2568 * An active spare device has ZPOOL_CONFIG_IS_SPARE set. 2569 * For a spare vdev, we only want to boot from the active 2570 * spare device. 2571 */ 2572 if (is_spare) { 2573 uint64_t spare = 0; 2574 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE, 2575 &spare); 2576 if (!spare) 2577 return (EZFS_INVALCONFIG); 2578 } 2579 2580 if (vdev_is_online(nv)) { 2581 if ((ret = vdev_get_one_physpath(nv, physpath, 2582 phypath_size, rsz)) != 0) 2583 return (ret); 2584 } 2585 } else if (strcmp(type, VDEV_TYPE_MIRROR) == 0 || 2586 strcmp(type, VDEV_TYPE_RAIDZ) == 0 || 2587 strcmp(type, VDEV_TYPE_REPLACING) == 0 || 2588 (is_spare = (strcmp(type, VDEV_TYPE_SPARE) == 0))) { 2589 nvlist_t **child; 2590 uint_t count; 2591 int i, ret; 2592 2593 if (nvlist_lookup_nvlist_array(nv, 2594 ZPOOL_CONFIG_CHILDREN, &child, &count) != 0) 2595 return (EZFS_INVALCONFIG); 2596 2597 for (i = 0; i < count; i++) { 2598 ret = vdev_get_physpaths(child[i], physpath, 2599 phypath_size, rsz, is_spare); 2600 if (ret == EZFS_NOSPC) 2601 return (ret); 2602 } 2603 } 2604 2605 return (EZFS_POOL_INVALARG); 2606 } 2607 2608 /* 2609 * Get phys_path for a root pool config. 2610 * Return 0 on success; non-zero on failure. 2611 */ 2612 static int 2613 zpool_get_config_physpath(nvlist_t *config, char *physpath, size_t phypath_size) 2614 { 2615 size_t rsz; 2616 nvlist_t *vdev_root; 2617 nvlist_t **child; 2618 uint_t count; 2619 char *type; 2620 2621 rsz = 0; 2622 2623 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 2624 &vdev_root) != 0) 2625 return (EZFS_INVALCONFIG); 2626 2627 if (nvlist_lookup_string(vdev_root, ZPOOL_CONFIG_TYPE, &type) != 0 || 2628 nvlist_lookup_nvlist_array(vdev_root, ZPOOL_CONFIG_CHILDREN, 2629 &child, &count) != 0) 2630 return (EZFS_INVALCONFIG); 2631 2632 /* 2633 * root pool can only have a single top-level vdev. 2634 */ 2635 if (strcmp(type, VDEV_TYPE_ROOT) != 0 || count != 1) 2636 return (EZFS_POOL_INVALARG); 2637 2638 (void) vdev_get_physpaths(child[0], physpath, phypath_size, &rsz, 2639 B_FALSE); 2640 2641 /* No online devices */ 2642 if (rsz == 0) 2643 return (EZFS_NODEVICE); 2644 2645 return (0); 2646 } 2647 2648 /* 2649 * Get phys_path for a root pool 2650 * Return 0 on success; non-zero on failure. 2651 */ 2652 int 2653 zpool_get_physpath(zpool_handle_t *zhp, char *physpath, size_t phypath_size) 2654 { 2655 return (zpool_get_config_physpath(zhp->zpool_config, physpath, 2656 phypath_size)); 2657 } 2658 2659 /* 2660 * If the device has being dynamically expanded then we need to relabel 2661 * the disk to use the new unallocated space. 2662 */ 2663 static int 2664 zpool_relabel_disk(libzfs_handle_t *hdl, const char *name) 2665 { 2666 char path[MAXPATHLEN]; 2667 char errbuf[1024]; 2668 int fd, error; 2669 int (*_efi_use_whole_disk)(int); 2670 2671 if ((_efi_use_whole_disk = (int (*)(int))dlsym(RTLD_DEFAULT, 2672 "efi_use_whole_disk")) == NULL) 2673 return (-1); 2674 2675 (void) snprintf(path, sizeof (path), "%s/%s", ZFS_RDISK_ROOT, name); 2676 2677 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) { 2678 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot " 2679 "relabel '%s': unable to open device"), name); 2680 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf)); 2681 } 2682 2683 /* 2684 * It's possible that we might encounter an error if the device 2685 * does not have any unallocated space left. If so, we simply 2686 * ignore that error and continue on. 2687 */ 2688 error = _efi_use_whole_disk(fd); 2689 (void) close(fd); 2690 if (error && error != VT_ENOSPC) { 2691 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot " 2692 "relabel '%s': unable to read disk capacity"), name); 2693 return (zfs_error(hdl, EZFS_NOCAP, errbuf)); 2694 } 2695 return (0); 2696 } 2697 2698 /* 2699 * Bring the specified vdev online. The 'flags' parameter is a set of the 2700 * ZFS_ONLINE_* flags. 2701 */ 2702 int 2703 zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags, 2704 vdev_state_t *newstate) 2705 { 2706 zfs_cmd_t zc = { 0 }; 2707 char msg[1024]; 2708 char *pathname; 2709 nvlist_t *tgt; 2710 boolean_t avail_spare, l2cache, islog; 2711 libzfs_handle_t *hdl = zhp->zpool_hdl; 2712 2713 if (flags & ZFS_ONLINE_EXPAND) { 2714 (void) snprintf(msg, sizeof (msg), 2715 dgettext(TEXT_DOMAIN, "cannot expand %s"), path); 2716 } else { 2717 (void) snprintf(msg, sizeof (msg), 2718 dgettext(TEXT_DOMAIN, "cannot online %s"), path); 2719 } 2720 2721 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2722 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 2723 &islog)) == NULL) 2724 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 2725 2726 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 2727 2728 if (avail_spare) 2729 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 2730 2731 if ((flags & ZFS_ONLINE_EXPAND || 2732 zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) && 2733 nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH, &pathname) == 0) { 2734 uint64_t wholedisk = 0; 2735 2736 (void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK, 2737 &wholedisk); 2738 2739 /* 2740 * XXX - L2ARC 1.0 devices can't support expansion. 2741 */ 2742 if (l2cache) { 2743 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2744 "cannot expand cache devices")); 2745 return (zfs_error(hdl, EZFS_VDEVNOTSUP, msg)); 2746 } 2747 2748 if (wholedisk) { 2749 pathname += strlen(ZFS_DISK_ROOT) + 1; 2750 (void) zpool_relabel_disk(hdl, pathname); 2751 } 2752 } 2753 2754 zc.zc_cookie = VDEV_STATE_ONLINE; 2755 zc.zc_obj = flags; 2756 2757 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0) { 2758 if (errno == EINVAL) { 2759 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "was split " 2760 "from this pool into a new one. Use '%s' " 2761 "instead"), "zpool detach"); 2762 return (zfs_error(hdl, EZFS_POSTSPLIT_ONLINE, msg)); 2763 } 2764 return (zpool_standard_error(hdl, errno, msg)); 2765 } 2766 2767 *newstate = zc.zc_cookie; 2768 return (0); 2769 } 2770 2771 /* 2772 * Take the specified vdev offline 2773 */ 2774 int 2775 zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp) 2776 { 2777 zfs_cmd_t zc = { 0 }; 2778 char msg[1024]; 2779 nvlist_t *tgt; 2780 boolean_t avail_spare, l2cache; 2781 libzfs_handle_t *hdl = zhp->zpool_hdl; 2782 2783 (void) snprintf(msg, sizeof (msg), 2784 dgettext(TEXT_DOMAIN, "cannot offline %s"), path); 2785 2786 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2787 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 2788 NULL)) == NULL) 2789 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 2790 2791 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 2792 2793 if (avail_spare) 2794 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 2795 2796 zc.zc_cookie = VDEV_STATE_OFFLINE; 2797 zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0; 2798 2799 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0) 2800 return (0); 2801 2802 switch (errno) { 2803 case EBUSY: 2804 2805 /* 2806 * There are no other replicas of this device. 2807 */ 2808 return (zfs_error(hdl, EZFS_NOREPLICAS, msg)); 2809 2810 case EEXIST: 2811 /* 2812 * The log device has unplayed logs 2813 */ 2814 return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, msg)); 2815 2816 default: 2817 return (zpool_standard_error(hdl, errno, msg)); 2818 } 2819 } 2820 2821 /* 2822 * Mark the given vdev faulted. 2823 */ 2824 int 2825 zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux) 2826 { 2827 zfs_cmd_t zc = { 0 }; 2828 char msg[1024]; 2829 libzfs_handle_t *hdl = zhp->zpool_hdl; 2830 2831 (void) snprintf(msg, sizeof (msg), 2832 dgettext(TEXT_DOMAIN, "cannot fault %llu"), guid); 2833 2834 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2835 zc.zc_guid = guid; 2836 zc.zc_cookie = VDEV_STATE_FAULTED; 2837 zc.zc_obj = aux; 2838 2839 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0) 2840 return (0); 2841 2842 switch (errno) { 2843 case EBUSY: 2844 2845 /* 2846 * There are no other replicas of this device. 2847 */ 2848 return (zfs_error(hdl, EZFS_NOREPLICAS, msg)); 2849 2850 default: 2851 return (zpool_standard_error(hdl, errno, msg)); 2852 } 2853 2854 } 2855 2856 /* 2857 * Mark the given vdev degraded. 2858 */ 2859 int 2860 zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux) 2861 { 2862 zfs_cmd_t zc = { 0 }; 2863 char msg[1024]; 2864 libzfs_handle_t *hdl = zhp->zpool_hdl; 2865 2866 (void) snprintf(msg, sizeof (msg), 2867 dgettext(TEXT_DOMAIN, "cannot degrade %llu"), guid); 2868 2869 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2870 zc.zc_guid = guid; 2871 zc.zc_cookie = VDEV_STATE_DEGRADED; 2872 zc.zc_obj = aux; 2873 2874 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0) 2875 return (0); 2876 2877 return (zpool_standard_error(hdl, errno, msg)); 2878 } 2879 2880 /* 2881 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as 2882 * a hot spare. 2883 */ 2884 static boolean_t 2885 is_replacing_spare(nvlist_t *search, nvlist_t *tgt, int which) 2886 { 2887 nvlist_t **child; 2888 uint_t c, children; 2889 char *type; 2890 2891 if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child, 2892 &children) == 0) { 2893 verify(nvlist_lookup_string(search, ZPOOL_CONFIG_TYPE, 2894 &type) == 0); 2895 2896 if (strcmp(type, VDEV_TYPE_SPARE) == 0 && 2897 children == 2 && child[which] == tgt) 2898 return (B_TRUE); 2899 2900 for (c = 0; c < children; c++) 2901 if (is_replacing_spare(child[c], tgt, which)) 2902 return (B_TRUE); 2903 } 2904 2905 return (B_FALSE); 2906 } 2907 2908 /* 2909 * Attach new_disk (fully described by nvroot) to old_disk. 2910 * If 'replacing' is specified, the new disk will replace the old one. 2911 */ 2912 int 2913 zpool_vdev_attach(zpool_handle_t *zhp, 2914 const char *old_disk, const char *new_disk, nvlist_t *nvroot, int replacing) 2915 { 2916 zfs_cmd_t zc = { 0 }; 2917 char msg[1024]; 2918 int ret; 2919 nvlist_t *tgt; 2920 boolean_t avail_spare, l2cache, islog; 2921 uint64_t val; 2922 char *newname; 2923 nvlist_t **child; 2924 uint_t children; 2925 nvlist_t *config_root; 2926 libzfs_handle_t *hdl = zhp->zpool_hdl; 2927 boolean_t rootpool = zpool_is_bootable(zhp); 2928 2929 if (replacing) 2930 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 2931 "cannot replace %s with %s"), old_disk, new_disk); 2932 else 2933 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 2934 "cannot attach %s to %s"), new_disk, old_disk); 2935 2936 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2937 if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache, 2938 &islog)) == NULL) 2939 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 2940 2941 if (avail_spare) 2942 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 2943 2944 if (l2cache) 2945 return (zfs_error(hdl, EZFS_ISL2CACHE, msg)); 2946 2947 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 2948 zc.zc_cookie = replacing; 2949 2950 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, 2951 &child, &children) != 0 || children != 1) { 2952 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2953 "new device must be a single disk")); 2954 return (zfs_error(hdl, EZFS_INVALCONFIG, msg)); 2955 } 2956 2957 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL), 2958 ZPOOL_CONFIG_VDEV_TREE, &config_root) == 0); 2959 2960 if ((newname = zpool_vdev_name(NULL, NULL, child[0], 0)) == NULL) 2961 return (-1); 2962 2963 /* 2964 * If the target is a hot spare that has been swapped in, we can only 2965 * replace it with another hot spare. 2966 */ 2967 if (replacing && 2968 nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 && 2969 (zpool_find_vdev(zhp, newname, &avail_spare, &l2cache, 2970 NULL) == NULL || !avail_spare) && 2971 is_replacing_spare(config_root, tgt, 1)) { 2972 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2973 "can only be replaced by another hot spare")); 2974 free(newname); 2975 return (zfs_error(hdl, EZFS_BADTARGET, msg)); 2976 } 2977 2978 free(newname); 2979 2980 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0) 2981 return (-1); 2982 2983 ret = zfs_ioctl(hdl, ZFS_IOC_VDEV_ATTACH, &zc); 2984 2985 zcmd_free_nvlists(&zc); 2986 2987 if (ret == 0) { 2988 if (rootpool) { 2989 /* 2990 * XXX need a better way to prevent user from 2991 * booting up a half-baked vdev. 2992 */ 2993 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Make " 2994 "sure to wait until resilver is done " 2995 "before rebooting.\n")); 2996 } 2997 return (0); 2998 } 2999 3000 switch (errno) { 3001 case ENOTSUP: 3002 /* 3003 * Can't attach to or replace this type of vdev. 3004 */ 3005 if (replacing) { 3006 uint64_t version = zpool_get_prop_int(zhp, 3007 ZPOOL_PROP_VERSION, NULL); 3008 3009 if (islog) 3010 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3011 "cannot replace a log with a spare")); 3012 else if (version >= SPA_VERSION_MULTI_REPLACE) 3013 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3014 "already in replacing/spare config; wait " 3015 "for completion or use 'zpool detach'")); 3016 else 3017 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3018 "cannot replace a replacing device")); 3019 } else { 3020 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3021 "can only attach to mirrors and top-level " 3022 "disks")); 3023 } 3024 (void) zfs_error(hdl, EZFS_BADTARGET, msg); 3025 break; 3026 3027 case EINVAL: 3028 /* 3029 * The new device must be a single disk. 3030 */ 3031 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3032 "new device must be a single disk")); 3033 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg); 3034 break; 3035 3036 case EBUSY: 3037 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy, " 3038 "or device removal is in progress"), 3039 new_disk); 3040 (void) zfs_error(hdl, EZFS_BADDEV, msg); 3041 break; 3042 3043 case EOVERFLOW: 3044 /* 3045 * The new device is too small. 3046 */ 3047 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3048 "device is too small")); 3049 (void) zfs_error(hdl, EZFS_BADDEV, msg); 3050 break; 3051 3052 case EDOM: 3053 /* 3054 * The new device has a different optimal sector size. 3055 */ 3056 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3057 "new device has a different optimal sector size; use the " 3058 "option '-o ashift=N' to override the optimal size")); 3059 (void) zfs_error(hdl, EZFS_BADDEV, msg); 3060 break; 3061 3062 case ENAMETOOLONG: 3063 /* 3064 * The resulting top-level vdev spec won't fit in the label. 3065 */ 3066 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, msg); 3067 break; 3068 3069 default: 3070 (void) zpool_standard_error(hdl, errno, msg); 3071 } 3072 3073 return (-1); 3074 } 3075 3076 /* 3077 * Detach the specified device. 3078 */ 3079 int 3080 zpool_vdev_detach(zpool_handle_t *zhp, const char *path) 3081 { 3082 zfs_cmd_t zc = { 0 }; 3083 char msg[1024]; 3084 nvlist_t *tgt; 3085 boolean_t avail_spare, l2cache; 3086 libzfs_handle_t *hdl = zhp->zpool_hdl; 3087 3088 (void) snprintf(msg, sizeof (msg), 3089 dgettext(TEXT_DOMAIN, "cannot detach %s"), path); 3090 3091 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3092 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 3093 NULL)) == NULL) 3094 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 3095 3096 if (avail_spare) 3097 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 3098 3099 if (l2cache) 3100 return (zfs_error(hdl, EZFS_ISL2CACHE, msg)); 3101 3102 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 3103 3104 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0) 3105 return (0); 3106 3107 switch (errno) { 3108 3109 case ENOTSUP: 3110 /* 3111 * Can't detach from this type of vdev. 3112 */ 3113 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only " 3114 "applicable to mirror and replacing vdevs")); 3115 (void) zfs_error(hdl, EZFS_BADTARGET, msg); 3116 break; 3117 3118 case EBUSY: 3119 /* 3120 * There are no other replicas of this device. 3121 */ 3122 (void) zfs_error(hdl, EZFS_NOREPLICAS, msg); 3123 break; 3124 3125 default: 3126 (void) zpool_standard_error(hdl, errno, msg); 3127 } 3128 3129 return (-1); 3130 } 3131 3132 /* 3133 * Find a mirror vdev in the source nvlist. 3134 * 3135 * The mchild array contains a list of disks in one of the top-level mirrors 3136 * of the source pool. The schild array contains a list of disks that the 3137 * user specified on the command line. We loop over the mchild array to 3138 * see if any entry in the schild array matches. 3139 * 3140 * If a disk in the mchild array is found in the schild array, we return 3141 * the index of that entry. Otherwise we return -1. 3142 */ 3143 static int 3144 find_vdev_entry(zpool_handle_t *zhp, nvlist_t **mchild, uint_t mchildren, 3145 nvlist_t **schild, uint_t schildren) 3146 { 3147 uint_t mc; 3148 3149 for (mc = 0; mc < mchildren; mc++) { 3150 uint_t sc; 3151 char *mpath = zpool_vdev_name(zhp->zpool_hdl, zhp, 3152 mchild[mc], 0); 3153 3154 for (sc = 0; sc < schildren; sc++) { 3155 char *spath = zpool_vdev_name(zhp->zpool_hdl, zhp, 3156 schild[sc], 0); 3157 boolean_t result = (strcmp(mpath, spath) == 0); 3158 3159 free(spath); 3160 if (result) { 3161 free(mpath); 3162 return (mc); 3163 } 3164 } 3165 3166 free(mpath); 3167 } 3168 3169 return (-1); 3170 } 3171 3172 /* 3173 * Split a mirror pool. If newroot points to null, then a new nvlist 3174 * is generated and it is the responsibility of the caller to free it. 3175 */ 3176 int 3177 zpool_vdev_split(zpool_handle_t *zhp, char *newname, nvlist_t **newroot, 3178 nvlist_t *props, splitflags_t flags) 3179 { 3180 zfs_cmd_t zc = { 0 }; 3181 char msg[1024]; 3182 nvlist_t *tree, *config, **child, **newchild, *newconfig = NULL; 3183 nvlist_t **varray = NULL, *zc_props = NULL; 3184 uint_t c, children, newchildren, lastlog = 0, vcount, found = 0; 3185 libzfs_handle_t *hdl = zhp->zpool_hdl; 3186 uint64_t vers; 3187 boolean_t freelist = B_FALSE, memory_err = B_TRUE; 3188 int retval = 0; 3189 3190 (void) snprintf(msg, sizeof (msg), 3191 dgettext(TEXT_DOMAIN, "Unable to split %s"), zhp->zpool_name); 3192 3193 if (!zpool_name_valid(hdl, B_FALSE, newname)) 3194 return (zfs_error(hdl, EZFS_INVALIDNAME, msg)); 3195 3196 if ((config = zpool_get_config(zhp, NULL)) == NULL) { 3197 (void) fprintf(stderr, gettext("Internal error: unable to " 3198 "retrieve pool configuration\n")); 3199 return (-1); 3200 } 3201 3202 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree) 3203 == 0); 3204 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &vers) == 0); 3205 3206 if (props) { 3207 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE }; 3208 if ((zc_props = zpool_valid_proplist(hdl, zhp->zpool_name, 3209 props, vers, flags, msg)) == NULL) 3210 return (-1); 3211 } 3212 3213 if (nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child, 3214 &children) != 0) { 3215 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3216 "Source pool is missing vdev tree")); 3217 nvlist_free(zc_props); 3218 return (-1); 3219 } 3220 3221 varray = zfs_alloc(hdl, children * sizeof (nvlist_t *)); 3222 vcount = 0; 3223 3224 if (*newroot == NULL || 3225 nvlist_lookup_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, 3226 &newchild, &newchildren) != 0) 3227 newchildren = 0; 3228 3229 for (c = 0; c < children; c++) { 3230 uint64_t is_log = B_FALSE, is_hole = B_FALSE; 3231 char *type; 3232 nvlist_t **mchild, *vdev; 3233 uint_t mchildren; 3234 int entry; 3235 3236 /* 3237 * Unlike cache & spares, slogs are stored in the 3238 * ZPOOL_CONFIG_CHILDREN array. We filter them out here. 3239 */ 3240 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG, 3241 &is_log); 3242 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE, 3243 &is_hole); 3244 if (is_log || is_hole) { 3245 /* 3246 * Create a hole vdev and put it in the config. 3247 */ 3248 if (nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) != 0) 3249 goto out; 3250 if (nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE, 3251 VDEV_TYPE_HOLE) != 0) 3252 goto out; 3253 if (nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_HOLE, 3254 1) != 0) 3255 goto out; 3256 if (lastlog == 0) 3257 lastlog = vcount; 3258 varray[vcount++] = vdev; 3259 continue; 3260 } 3261 lastlog = 0; 3262 verify(nvlist_lookup_string(child[c], ZPOOL_CONFIG_TYPE, &type) 3263 == 0); 3264 if (strcmp(type, VDEV_TYPE_MIRROR) != 0) { 3265 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3266 "Source pool must be composed only of mirrors\n")); 3267 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg); 3268 goto out; 3269 } 3270 3271 verify(nvlist_lookup_nvlist_array(child[c], 3272 ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0); 3273 3274 /* find or add an entry for this top-level vdev */ 3275 if (newchildren > 0 && 3276 (entry = find_vdev_entry(zhp, mchild, mchildren, 3277 newchild, newchildren)) >= 0) { 3278 /* We found a disk that the user specified. */ 3279 vdev = mchild[entry]; 3280 ++found; 3281 } else { 3282 /* User didn't specify a disk for this vdev. */ 3283 vdev = mchild[mchildren - 1]; 3284 } 3285 3286 if (nvlist_dup(vdev, &varray[vcount++], 0) != 0) 3287 goto out; 3288 } 3289 3290 /* did we find every disk the user specified? */ 3291 if (found != newchildren) { 3292 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "Device list must " 3293 "include at most one disk from each mirror")); 3294 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg); 3295 goto out; 3296 } 3297 3298 /* Prepare the nvlist for populating. */ 3299 if (*newroot == NULL) { 3300 if (nvlist_alloc(newroot, NV_UNIQUE_NAME, 0) != 0) 3301 goto out; 3302 freelist = B_TRUE; 3303 if (nvlist_add_string(*newroot, ZPOOL_CONFIG_TYPE, 3304 VDEV_TYPE_ROOT) != 0) 3305 goto out; 3306 } else { 3307 verify(nvlist_remove_all(*newroot, ZPOOL_CONFIG_CHILDREN) == 0); 3308 } 3309 3310 /* Add all the children we found */ 3311 if (nvlist_add_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, varray, 3312 lastlog == 0 ? vcount : lastlog) != 0) 3313 goto out; 3314 3315 /* 3316 * If we're just doing a dry run, exit now with success. 3317 */ 3318 if (flags.dryrun) { 3319 memory_err = B_FALSE; 3320 freelist = B_FALSE; 3321 goto out; 3322 } 3323 3324 /* now build up the config list & call the ioctl */ 3325 if (nvlist_alloc(&newconfig, NV_UNIQUE_NAME, 0) != 0) 3326 goto out; 3327 3328 if (nvlist_add_nvlist(newconfig, 3329 ZPOOL_CONFIG_VDEV_TREE, *newroot) != 0 || 3330 nvlist_add_string(newconfig, 3331 ZPOOL_CONFIG_POOL_NAME, newname) != 0 || 3332 nvlist_add_uint64(newconfig, ZPOOL_CONFIG_VERSION, vers) != 0) 3333 goto out; 3334 3335 /* 3336 * The new pool is automatically part of the namespace unless we 3337 * explicitly export it. 3338 */ 3339 if (!flags.import) 3340 zc.zc_cookie = ZPOOL_EXPORT_AFTER_SPLIT; 3341 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3342 (void) strlcpy(zc.zc_string, newname, sizeof (zc.zc_string)); 3343 if (zcmd_write_conf_nvlist(hdl, &zc, newconfig) != 0) 3344 goto out; 3345 if (zc_props != NULL && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0) 3346 goto out; 3347 3348 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SPLIT, &zc) != 0) { 3349 retval = zpool_standard_error(hdl, errno, msg); 3350 goto out; 3351 } 3352 3353 freelist = B_FALSE; 3354 memory_err = B_FALSE; 3355 3356 out: 3357 if (varray != NULL) { 3358 int v; 3359 3360 for (v = 0; v < vcount; v++) 3361 nvlist_free(varray[v]); 3362 free(varray); 3363 } 3364 zcmd_free_nvlists(&zc); 3365 nvlist_free(zc_props); 3366 nvlist_free(newconfig); 3367 if (freelist) { 3368 nvlist_free(*newroot); 3369 *newroot = NULL; 3370 } 3371 3372 if (retval != 0) 3373 return (retval); 3374 3375 if (memory_err) 3376 return (no_memory(hdl)); 3377 3378 return (0); 3379 } 3380 3381 /* 3382 * Remove the given device. 3383 */ 3384 int 3385 zpool_vdev_remove(zpool_handle_t *zhp, const char *path) 3386 { 3387 zfs_cmd_t zc = { 0 }; 3388 char msg[1024]; 3389 nvlist_t *tgt; 3390 boolean_t avail_spare, l2cache, islog; 3391 libzfs_handle_t *hdl = zhp->zpool_hdl; 3392 uint64_t version; 3393 3394 (void) snprintf(msg, sizeof (msg), 3395 dgettext(TEXT_DOMAIN, "cannot remove %s"), path); 3396 3397 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3398 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 3399 &islog)) == NULL) 3400 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 3401 3402 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL); 3403 if (islog && version < SPA_VERSION_HOLES) { 3404 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3405 "pool must be upgraded to support log removal")); 3406 return (zfs_error(hdl, EZFS_BADVERSION, msg)); 3407 } 3408 3409 zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID); 3410 3411 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0) 3412 return (0); 3413 3414 switch (errno) { 3415 3416 case EINVAL: 3417 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3418 "invalid config; all top-level vdevs must " 3419 "have the same sector size and not be raidz.")); 3420 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg); 3421 break; 3422 3423 case EBUSY: 3424 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3425 "Pool busy; removal may already be in progress")); 3426 (void) zfs_error(hdl, EZFS_BUSY, msg); 3427 break; 3428 3429 default: 3430 (void) zpool_standard_error(hdl, errno, msg); 3431 } 3432 return (-1); 3433 } 3434 3435 int 3436 zpool_vdev_remove_cancel(zpool_handle_t *zhp) 3437 { 3438 zfs_cmd_t zc = { 0 }; 3439 char msg[1024]; 3440 libzfs_handle_t *hdl = zhp->zpool_hdl; 3441 3442 (void) snprintf(msg, sizeof (msg), 3443 dgettext(TEXT_DOMAIN, "cannot cancel removal")); 3444 3445 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3446 zc.zc_cookie = 1; 3447 3448 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0) 3449 return (0); 3450 3451 return (zpool_standard_error(hdl, errno, msg)); 3452 } 3453 3454 int 3455 zpool_vdev_indirect_size(zpool_handle_t *zhp, const char *path, 3456 uint64_t *sizep) 3457 { 3458 char msg[1024]; 3459 nvlist_t *tgt; 3460 boolean_t avail_spare, l2cache, islog; 3461 libzfs_handle_t *hdl = zhp->zpool_hdl; 3462 3463 (void) snprintf(msg, sizeof (msg), 3464 dgettext(TEXT_DOMAIN, "cannot determine indirect size of %s"), 3465 path); 3466 3467 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 3468 &islog)) == NULL) 3469 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 3470 3471 if (avail_spare || l2cache || islog) { 3472 *sizep = 0; 3473 return (0); 3474 } 3475 3476 if (nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_INDIRECT_SIZE, sizep) != 0) { 3477 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3478 "indirect size not available")); 3479 return (zfs_error(hdl, EINVAL, msg)); 3480 } 3481 return (0); 3482 } 3483 3484 /* 3485 * Clear the errors for the pool, or the particular device if specified. 3486 */ 3487 int 3488 zpool_clear(zpool_handle_t *zhp, const char *path, nvlist_t *rewindnvl) 3489 { 3490 zfs_cmd_t zc = { 0 }; 3491 char msg[1024]; 3492 nvlist_t *tgt; 3493 zpool_load_policy_t policy; 3494 boolean_t avail_spare, l2cache; 3495 libzfs_handle_t *hdl = zhp->zpool_hdl; 3496 nvlist_t *nvi = NULL; 3497 int error; 3498 3499 if (path) 3500 (void) snprintf(msg, sizeof (msg), 3501 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"), 3502 path); 3503 else 3504 (void) snprintf(msg, sizeof (msg), 3505 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"), 3506 zhp->zpool_name); 3507 3508 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3509 if (path) { 3510 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, 3511 &l2cache, NULL)) == NULL) 3512 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 3513 3514 /* 3515 * Don't allow error clearing for hot spares. Do allow 3516 * error clearing for l2cache devices. 3517 */ 3518 if (avail_spare) 3519 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 3520 3521 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, 3522 &zc.zc_guid) == 0); 3523 } 3524 3525 zpool_get_load_policy(rewindnvl, &policy); 3526 zc.zc_cookie = policy.zlp_rewind; 3527 3528 if (zcmd_alloc_dst_nvlist(hdl, &zc, zhp->zpool_config_size * 2) != 0) 3529 return (-1); 3530 3531 if (zcmd_write_src_nvlist(hdl, &zc, rewindnvl) != 0) 3532 return (-1); 3533 3534 while ((error = zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc)) != 0 && 3535 errno == ENOMEM) { 3536 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) { 3537 zcmd_free_nvlists(&zc); 3538 return (-1); 3539 } 3540 } 3541 3542 if (!error || ((policy.zlp_rewind & ZPOOL_TRY_REWIND) && 3543 errno != EPERM && errno != EACCES)) { 3544 if (policy.zlp_rewind & 3545 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) { 3546 (void) zcmd_read_dst_nvlist(hdl, &zc, &nvi); 3547 zpool_rewind_exclaim(hdl, zc.zc_name, 3548 ((policy.zlp_rewind & ZPOOL_TRY_REWIND) != 0), 3549 nvi); 3550 nvlist_free(nvi); 3551 } 3552 zcmd_free_nvlists(&zc); 3553 return (0); 3554 } 3555 3556 zcmd_free_nvlists(&zc); 3557 return (zpool_standard_error(hdl, errno, msg)); 3558 } 3559 3560 /* 3561 * Similar to zpool_clear(), but takes a GUID (used by fmd). 3562 */ 3563 int 3564 zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid) 3565 { 3566 zfs_cmd_t zc = { 0 }; 3567 char msg[1024]; 3568 libzfs_handle_t *hdl = zhp->zpool_hdl; 3569 3570 (void) snprintf(msg, sizeof (msg), 3571 dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"), 3572 guid); 3573 3574 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3575 zc.zc_guid = guid; 3576 zc.zc_cookie = ZPOOL_NO_REWIND; 3577 3578 if (ioctl(hdl->libzfs_fd, ZFS_IOC_CLEAR, &zc) == 0) 3579 return (0); 3580 3581 return (zpool_standard_error(hdl, errno, msg)); 3582 } 3583 3584 /* 3585 * Change the GUID for a pool. 3586 */ 3587 int 3588 zpool_reguid(zpool_handle_t *zhp) 3589 { 3590 char msg[1024]; 3591 libzfs_handle_t *hdl = zhp->zpool_hdl; 3592 zfs_cmd_t zc = { 0 }; 3593 3594 (void) snprintf(msg, sizeof (msg), 3595 dgettext(TEXT_DOMAIN, "cannot reguid '%s'"), zhp->zpool_name); 3596 3597 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3598 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REGUID, &zc) == 0) 3599 return (0); 3600 3601 return (zpool_standard_error(hdl, errno, msg)); 3602 } 3603 3604 /* 3605 * Reopen the pool. 3606 */ 3607 int 3608 zpool_reopen(zpool_handle_t *zhp) 3609 { 3610 zfs_cmd_t zc = { 0 }; 3611 char msg[1024]; 3612 libzfs_handle_t *hdl = zhp->zpool_hdl; 3613 3614 (void) snprintf(msg, sizeof (msg), 3615 dgettext(TEXT_DOMAIN, "cannot reopen '%s'"), 3616 zhp->zpool_name); 3617 3618 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3619 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REOPEN, &zc) == 0) 3620 return (0); 3621 return (zpool_standard_error(hdl, errno, msg)); 3622 } 3623 3624 /* call into libzfs_core to execute the sync IOCTL per pool */ 3625 int 3626 zpool_sync_one(zpool_handle_t *zhp, void *data) 3627 { 3628 int ret; 3629 libzfs_handle_t *hdl = zpool_get_handle(zhp); 3630 const char *pool_name = zpool_get_name(zhp); 3631 boolean_t *force = data; 3632 nvlist_t *innvl = fnvlist_alloc(); 3633 3634 fnvlist_add_boolean_value(innvl, "force", *force); 3635 if ((ret = lzc_sync(pool_name, innvl, NULL)) != 0) { 3636 nvlist_free(innvl); 3637 return (zpool_standard_error_fmt(hdl, ret, 3638 dgettext(TEXT_DOMAIN, "sync '%s' failed"), pool_name)); 3639 } 3640 nvlist_free(innvl); 3641 3642 return (0); 3643 } 3644 3645 /* 3646 * Convert from a devid string to a path. 3647 */ 3648 static char * 3649 devid_to_path(char *devid_str) 3650 { 3651 ddi_devid_t devid; 3652 char *minor; 3653 char *path; 3654 devid_nmlist_t *list = NULL; 3655 int ret; 3656 3657 if (devid_str_decode(devid_str, &devid, &minor) != 0) 3658 return (NULL); 3659 3660 ret = devid_deviceid_to_nmlist("/dev", devid, minor, &list); 3661 3662 devid_str_free(minor); 3663 devid_free(devid); 3664 3665 if (ret != 0) 3666 return (NULL); 3667 3668 /* 3669 * In a case the strdup() fails, we will just return NULL below. 3670 */ 3671 path = strdup(list[0].devname); 3672 3673 devid_free_nmlist(list); 3674 3675 return (path); 3676 } 3677 3678 /* 3679 * Convert from a path to a devid string. 3680 */ 3681 static char * 3682 path_to_devid(const char *path) 3683 { 3684 int fd; 3685 ddi_devid_t devid; 3686 char *minor, *ret; 3687 3688 if ((fd = open(path, O_RDONLY)) < 0) 3689 return (NULL); 3690 3691 minor = NULL; 3692 ret = NULL; 3693 if (devid_get(fd, &devid) == 0) { 3694 if (devid_get_minor_name(fd, &minor) == 0) 3695 ret = devid_str_encode(devid, minor); 3696 if (minor != NULL) 3697 devid_str_free(minor); 3698 devid_free(devid); 3699 } 3700 (void) close(fd); 3701 3702 return (ret); 3703 } 3704 3705 struct path_from_physpath_walker_args { 3706 char *pfpwa_path; 3707 }; 3708 3709 /* 3710 * Walker for use with di_devlink_walk(). Stores the "/dev" path of the first 3711 * primary devlink (i.e., the first devlink which refers to our "/devices" 3712 * node) and stops walking. 3713 */ 3714 static int 3715 path_from_physpath_walker(di_devlink_t devlink, void *arg) 3716 { 3717 struct path_from_physpath_walker_args *pfpwa = arg; 3718 3719 if (di_devlink_type(devlink) != DI_PRIMARY_LINK) { 3720 return (DI_WALK_CONTINUE); 3721 } 3722 3723 verify(pfpwa->pfpwa_path == NULL); 3724 if ((pfpwa->pfpwa_path = strdup(di_devlink_path(devlink))) != NULL) { 3725 return (DI_WALK_TERMINATE); 3726 } 3727 3728 return (DI_WALK_CONTINUE); 3729 } 3730 3731 /* 3732 * Search for a "/dev" path that refers to our physical path. Returns the new 3733 * path if one is found and it does not match the existing "path" value. If 3734 * the value is unchanged, or one could not be found, returns NULL. 3735 */ 3736 static char * 3737 path_from_physpath(libzfs_handle_t *hdl, const char *path, 3738 const char *physpath) 3739 { 3740 struct path_from_physpath_walker_args pfpwa; 3741 3742 if (physpath == NULL) { 3743 return (NULL); 3744 } 3745 3746 if (hdl->libzfs_devlink == NULL) { 3747 if ((hdl->libzfs_devlink = di_devlink_init(NULL, 0)) == 3748 DI_LINK_NIL) { 3749 /* 3750 * We may not be able to open a handle if this process 3751 * is insufficiently privileged, or we are too early in 3752 * boot for devfsadm to be ready. Ignore this error 3753 * and defer the path check to a subsequent run. 3754 */ 3755 return (NULL); 3756 } 3757 } 3758 3759 pfpwa.pfpwa_path = NULL; 3760 (void) di_devlink_walk(hdl->libzfs_devlink, NULL, physpath, 3761 DI_PRIMARY_LINK, &pfpwa, path_from_physpath_walker); 3762 3763 if (path != NULL && pfpwa.pfpwa_path != NULL && 3764 strcmp(path, pfpwa.pfpwa_path) == 0) { 3765 /* 3766 * If the path is already correct, no change is required. 3767 */ 3768 free(pfpwa.pfpwa_path); 3769 return (NULL); 3770 } 3771 3772 return (pfpwa.pfpwa_path); 3773 } 3774 3775 /* 3776 * Issue the necessary ioctl() to update the stored path value for the vdev. We 3777 * ignore any failure here, since a common case is for an unprivileged user to 3778 * type 'zpool status', and we'll display the correct information anyway. 3779 */ 3780 static void 3781 set_path(zpool_handle_t *zhp, nvlist_t *nv, const char *path) 3782 { 3783 zfs_cmd_t zc = { 0 }; 3784 3785 (void) strncpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3786 (void) strncpy(zc.zc_value, path, sizeof (zc.zc_value)); 3787 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, 3788 &zc.zc_guid) == 0); 3789 3790 (void) ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SETPATH, &zc); 3791 } 3792 3793 /* 3794 * Given a vdev, return the name to display in iostat. If the vdev has a path, 3795 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type. 3796 * We also check if this is a whole disk, in which case we strip off the 3797 * trailing 's0' slice name. 3798 * 3799 * This routine is also responsible for identifying when disks have been 3800 * reconfigured in a new location. The kernel will have opened the device by 3801 * devid, but the path will still refer to the old location. To catch this, we 3802 * first do a path -> devid translation (which is fast for the common case). If 3803 * the devid matches, we're done. If not, we do a reverse devid -> path 3804 * translation and issue the appropriate ioctl() to update the path of the vdev. 3805 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any 3806 * of these checks. 3807 */ 3808 char * 3809 zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv, 3810 int name_flags) 3811 { 3812 char *path, *env; 3813 uint64_t value; 3814 char buf[64]; 3815 3816 env = getenv("ZPOOL_VDEV_NAME_PATH"); 3817 if (env && (strtoul(env, NULL, 0) > 0 || 3818 !strncasecmp(env, "YES", 3) || !strncasecmp(env, "ON", 2))) 3819 name_flags |= VDEV_NAME_PATH; 3820 3821 env = getenv("ZPOOL_VDEV_NAME_GUID"); 3822 if (env && (strtoul(env, NULL, 0) > 0 || 3823 !strncasecmp(env, "YES", 3) || !strncasecmp(env, "ON", 2))) 3824 name_flags |= VDEV_NAME_GUID; 3825 3826 env = getenv("ZPOOL_VDEV_NAME_FOLLOW_LINKS"); 3827 if (env && (strtoul(env, NULL, 0) > 0 || 3828 !strncasecmp(env, "YES", 3) || !strncasecmp(env, "ON", 2))) 3829 name_flags |= VDEV_NAME_FOLLOW_LINKS; 3830 3831 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, &value) == 0 || 3832 name_flags & VDEV_NAME_GUID) { 3833 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value); 3834 (void) snprintf(buf, sizeof (buf), "%llu", (u_longlong_t)value); 3835 path = buf; 3836 } else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) { 3837 vdev_stat_t *vs; 3838 uint_t vsc; 3839 char *newpath = NULL; 3840 char *physpath = NULL; 3841 char *devid = NULL; 3842 3843 /* 3844 * If the device is dead (faulted, offline, etc) then don't 3845 * bother opening it. Otherwise we may be forcing the user to 3846 * open a misbehaving device, which can have undesirable 3847 * effects. 3848 */ 3849 if (nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS, 3850 (uint64_t **)&vs, &vsc) != 0 || 3851 vs->vs_state < VDEV_STATE_DEGRADED || 3852 zhp == NULL) { 3853 goto after_open; 3854 } 3855 3856 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &devid) == 0) { 3857 /* 3858 * This vdev has a devid. We can use it to check the 3859 * current path. 3860 */ 3861 char *newdevid = path_to_devid(path); 3862 3863 if (newdevid == NULL || strcmp(devid, newdevid) != 0) { 3864 newpath = devid_to_path(devid); 3865 } 3866 3867 if (newdevid != NULL) 3868 devid_str_free(newdevid); 3869 3870 } else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PHYS_PATH, 3871 &physpath) == 0) { 3872 /* 3873 * This vdev does not have a devid, but it does have a 3874 * physical path. Attempt to translate this to a /dev 3875 * path. 3876 */ 3877 newpath = path_from_physpath(hdl, path, physpath); 3878 } 3879 3880 if (newpath != NULL) { 3881 /* 3882 * Update the path appropriately. 3883 */ 3884 set_path(zhp, nv, newpath); 3885 if (nvlist_add_string(nv, ZPOOL_CONFIG_PATH, 3886 newpath) == 0) { 3887 verify(nvlist_lookup_string(nv, 3888 ZPOOL_CONFIG_PATH, &path) == 0); 3889 } 3890 free(newpath); 3891 } 3892 3893 if (name_flags & VDEV_NAME_FOLLOW_LINKS) { 3894 char *rp = realpath(path, NULL); 3895 if (rp) { 3896 strlcpy(buf, rp, sizeof (buf)); 3897 path = buf; 3898 free(rp); 3899 } 3900 } 3901 3902 after_open: 3903 if (strncmp(path, ZFS_DISK_ROOTD, strlen(ZFS_DISK_ROOTD)) == 0) 3904 path += strlen(ZFS_DISK_ROOTD); 3905 3906 /* 3907 * Remove the partition from the path it this is a whole disk. 3908 */ 3909 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, &value) 3910 == 0 && value && !(name_flags & VDEV_NAME_PATH)) { 3911 int pathlen = strlen(path); 3912 char *tmp = zfs_strdup(hdl, path); 3913 3914 /* 3915 * If it starts with c#, and ends with "s0" or "s1", 3916 * chop the slice off, or if it ends with "s0/old" or 3917 * "s1/old", remove the slice from the middle. 3918 */ 3919 if (CTD_CHECK(tmp)) { 3920 if (strcmp(&tmp[pathlen - 2], "s0") == 0 || 3921 strcmp(&tmp[pathlen - 2], "s1") == 0) { 3922 tmp[pathlen - 2] = '\0'; 3923 } else if (pathlen > 6 && 3924 (strcmp(&tmp[pathlen - 6], "s0/old") == 0 || 3925 strcmp(&tmp[pathlen - 6], "s1/old") == 0)) { 3926 (void) strcpy(&tmp[pathlen - 6], 3927 "/old"); 3928 } 3929 } 3930 return (tmp); 3931 } 3932 } else { 3933 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &path) == 0); 3934 3935 /* 3936 * If it's a raidz device, we need to stick in the parity level. 3937 */ 3938 if (strcmp(path, VDEV_TYPE_RAIDZ) == 0) { 3939 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY, 3940 &value) == 0); 3941 (void) snprintf(buf, sizeof (buf), "%s%llu", path, 3942 (u_longlong_t)value); 3943 path = buf; 3944 } 3945 3946 /* 3947 * We identify each top-level vdev by using a <type-id> 3948 * naming convention. 3949 */ 3950 if (name_flags & VDEV_NAME_TYPE_ID) { 3951 uint64_t id; 3952 3953 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID, 3954 &id) == 0); 3955 (void) snprintf(buf, sizeof (buf), "%s-%llu", path, 3956 (u_longlong_t)id); 3957 path = buf; 3958 } 3959 } 3960 3961 return (zfs_strdup(hdl, path)); 3962 } 3963 3964 static int 3965 zbookmark_mem_compare(const void *a, const void *b) 3966 { 3967 return (memcmp(a, b, sizeof (zbookmark_phys_t))); 3968 } 3969 3970 /* 3971 * Retrieve the persistent error log, uniquify the members, and return to the 3972 * caller. 3973 */ 3974 int 3975 zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp) 3976 { 3977 zfs_cmd_t zc = { 0 }; 3978 uint64_t count; 3979 zbookmark_phys_t *zb = NULL; 3980 int i; 3981 3982 /* 3983 * Retrieve the raw error list from the kernel. If the number of errors 3984 * has increased, allocate more space and continue until we get the 3985 * entire list. 3986 */ 3987 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_ERRCOUNT, 3988 &count) == 0); 3989 if (count == 0) 3990 return (0); 3991 if ((zc.zc_nvlist_dst = (uintptr_t)zfs_alloc(zhp->zpool_hdl, 3992 count * sizeof (zbookmark_phys_t))) == (uintptr_t)NULL) 3993 return (-1); 3994 zc.zc_nvlist_dst_size = count; 3995 (void) strcpy(zc.zc_name, zhp->zpool_name); 3996 for (;;) { 3997 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_ERROR_LOG, 3998 &zc) != 0) { 3999 free((void *)(uintptr_t)zc.zc_nvlist_dst); 4000 if (errno == ENOMEM) { 4001 void *dst; 4002 4003 count = zc.zc_nvlist_dst_size; 4004 dst = zfs_alloc(zhp->zpool_hdl, count * 4005 sizeof (zbookmark_phys_t)); 4006 if (dst == NULL) 4007 return (-1); 4008 zc.zc_nvlist_dst = (uintptr_t)dst; 4009 } else { 4010 return (-1); 4011 } 4012 } else { 4013 break; 4014 } 4015 } 4016 4017 /* 4018 * Sort the resulting bookmarks. This is a little confusing due to the 4019 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last 4020 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks 4021 * _not_ copied as part of the process. So we point the start of our 4022 * array appropriate and decrement the total number of elements. 4023 */ 4024 zb = ((zbookmark_phys_t *)(uintptr_t)zc.zc_nvlist_dst) + 4025 zc.zc_nvlist_dst_size; 4026 count -= zc.zc_nvlist_dst_size; 4027 4028 qsort(zb, count, sizeof (zbookmark_phys_t), zbookmark_mem_compare); 4029 4030 verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0); 4031 4032 /* 4033 * Fill in the nverrlistp with nvlist's of dataset and object numbers. 4034 */ 4035 for (i = 0; i < count; i++) { 4036 nvlist_t *nv; 4037 4038 /* ignoring zb_blkid and zb_level for now */ 4039 if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset && 4040 zb[i-1].zb_object == zb[i].zb_object) 4041 continue; 4042 4043 if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0) 4044 goto nomem; 4045 if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET, 4046 zb[i].zb_objset) != 0) { 4047 nvlist_free(nv); 4048 goto nomem; 4049 } 4050 if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT, 4051 zb[i].zb_object) != 0) { 4052 nvlist_free(nv); 4053 goto nomem; 4054 } 4055 if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) { 4056 nvlist_free(nv); 4057 goto nomem; 4058 } 4059 nvlist_free(nv); 4060 } 4061 4062 free((void *)(uintptr_t)zc.zc_nvlist_dst); 4063 return (0); 4064 4065 nomem: 4066 free((void *)(uintptr_t)zc.zc_nvlist_dst); 4067 return (no_memory(zhp->zpool_hdl)); 4068 } 4069 4070 /* 4071 * Upgrade a ZFS pool to the latest on-disk version. 4072 */ 4073 int 4074 zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version) 4075 { 4076 zfs_cmd_t zc = { 0 }; 4077 libzfs_handle_t *hdl = zhp->zpool_hdl; 4078 4079 (void) strcpy(zc.zc_name, zhp->zpool_name); 4080 zc.zc_cookie = new_version; 4081 4082 if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0) 4083 return (zpool_standard_error_fmt(hdl, errno, 4084 dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"), 4085 zhp->zpool_name)); 4086 return (0); 4087 } 4088 4089 void 4090 zfs_save_arguments(int argc, char **argv, char *string, int len) 4091 { 4092 (void) strlcpy(string, basename(argv[0]), len); 4093 for (int i = 1; i < argc; i++) { 4094 (void) strlcat(string, " ", len); 4095 (void) strlcat(string, argv[i], len); 4096 } 4097 } 4098 4099 int 4100 zpool_log_history(libzfs_handle_t *hdl, const char *message) 4101 { 4102 zfs_cmd_t zc = { 0 }; 4103 nvlist_t *args; 4104 int err; 4105 4106 args = fnvlist_alloc(); 4107 fnvlist_add_string(args, "message", message); 4108 err = zcmd_write_src_nvlist(hdl, &zc, args); 4109 if (err == 0) 4110 err = ioctl(hdl->libzfs_fd, ZFS_IOC_LOG_HISTORY, &zc); 4111 nvlist_free(args); 4112 zcmd_free_nvlists(&zc); 4113 return (err); 4114 } 4115 4116 /* 4117 * Perform ioctl to get some command history of a pool. 4118 * 4119 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the 4120 * logical offset of the history buffer to start reading from. 4121 * 4122 * Upon return, 'off' is the next logical offset to read from and 4123 * 'len' is the actual amount of bytes read into 'buf'. 4124 */ 4125 static int 4126 get_history(zpool_handle_t *zhp, char *buf, uint64_t *off, uint64_t *len) 4127 { 4128 zfs_cmd_t zc = { 0 }; 4129 libzfs_handle_t *hdl = zhp->zpool_hdl; 4130 4131 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 4132 4133 zc.zc_history = (uint64_t)(uintptr_t)buf; 4134 zc.zc_history_len = *len; 4135 zc.zc_history_offset = *off; 4136 4137 if (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) { 4138 switch (errno) { 4139 case EPERM: 4140 return (zfs_error_fmt(hdl, EZFS_PERM, 4141 dgettext(TEXT_DOMAIN, 4142 "cannot show history for pool '%s'"), 4143 zhp->zpool_name)); 4144 case ENOENT: 4145 return (zfs_error_fmt(hdl, EZFS_NOHISTORY, 4146 dgettext(TEXT_DOMAIN, "cannot get history for pool " 4147 "'%s'"), zhp->zpool_name)); 4148 case ENOTSUP: 4149 return (zfs_error_fmt(hdl, EZFS_BADVERSION, 4150 dgettext(TEXT_DOMAIN, "cannot get history for pool " 4151 "'%s', pool must be upgraded"), zhp->zpool_name)); 4152 default: 4153 return (zpool_standard_error_fmt(hdl, errno, 4154 dgettext(TEXT_DOMAIN, 4155 "cannot get history for '%s'"), zhp->zpool_name)); 4156 } 4157 } 4158 4159 *len = zc.zc_history_len; 4160 *off = zc.zc_history_offset; 4161 4162 return (0); 4163 } 4164 4165 /* 4166 * Process the buffer of nvlists, unpacking and storing each nvlist record 4167 * into 'records'. 'leftover' is set to the number of bytes that weren't 4168 * processed as there wasn't a complete record. 4169 */ 4170 int 4171 zpool_history_unpack(char *buf, uint64_t bytes_read, uint64_t *leftover, 4172 nvlist_t ***records, uint_t *numrecords) 4173 { 4174 uint64_t reclen; 4175 nvlist_t *nv; 4176 int i; 4177 4178 while (bytes_read > sizeof (reclen)) { 4179 4180 /* get length of packed record (stored as little endian) */ 4181 for (i = 0, reclen = 0; i < sizeof (reclen); i++) 4182 reclen += (uint64_t)(((uchar_t *)buf)[i]) << (8*i); 4183 4184 if (bytes_read < sizeof (reclen) + reclen) 4185 break; 4186 4187 /* unpack record */ 4188 if (nvlist_unpack(buf + sizeof (reclen), reclen, &nv, 0) != 0) 4189 return (ENOMEM); 4190 bytes_read -= sizeof (reclen) + reclen; 4191 buf += sizeof (reclen) + reclen; 4192 4193 /* add record to nvlist array */ 4194 (*numrecords)++; 4195 if (ISP2(*numrecords + 1)) { 4196 *records = realloc(*records, 4197 *numrecords * 2 * sizeof (nvlist_t *)); 4198 } 4199 (*records)[*numrecords - 1] = nv; 4200 } 4201 4202 *leftover = bytes_read; 4203 return (0); 4204 } 4205 4206 /* 4207 * Retrieve the command history of a pool. 4208 */ 4209 int 4210 zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp) 4211 { 4212 char *buf; 4213 int buflen = 128 * 1024; 4214 uint64_t off = 0; 4215 nvlist_t **records = NULL; 4216 uint_t numrecords = 0; 4217 int err, i; 4218 4219 buf = malloc(buflen); 4220 if (buf == NULL) 4221 return (ENOMEM); 4222 do { 4223 uint64_t bytes_read = buflen; 4224 uint64_t leftover; 4225 4226 if ((err = get_history(zhp, buf, &off, &bytes_read)) != 0) 4227 break; 4228 4229 /* if nothing else was read in, we're at EOF, just return */ 4230 if (!bytes_read) 4231 break; 4232 4233 if ((err = zpool_history_unpack(buf, bytes_read, 4234 &leftover, &records, &numrecords)) != 0) 4235 break; 4236 off -= leftover; 4237 if (leftover == bytes_read) { 4238 /* 4239 * no progress made, because buffer is not big enough 4240 * to hold this record; resize and retry. 4241 */ 4242 buflen *= 2; 4243 free(buf); 4244 buf = malloc(buflen); 4245 if (buf == NULL) 4246 return (ENOMEM); 4247 } 4248 4249 /* CONSTCOND */ 4250 } while (1); 4251 4252 free(buf); 4253 4254 if (!err) { 4255 verify(nvlist_alloc(nvhisp, NV_UNIQUE_NAME, 0) == 0); 4256 verify(nvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD, 4257 records, numrecords) == 0); 4258 } 4259 for (i = 0; i < numrecords; i++) 4260 nvlist_free(records[i]); 4261 free(records); 4262 4263 return (err); 4264 } 4265 4266 void 4267 zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj, 4268 char *pathname, size_t len) 4269 { 4270 zfs_cmd_t zc = { 0 }; 4271 boolean_t mounted = B_FALSE; 4272 char *mntpnt = NULL; 4273 char dsname[ZFS_MAX_DATASET_NAME_LEN]; 4274 4275 if (dsobj == 0) { 4276 /* special case for the MOS */ 4277 (void) snprintf(pathname, len, "<metadata>:<0x%llx>", obj); 4278 return; 4279 } 4280 4281 /* get the dataset's name */ 4282 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 4283 zc.zc_obj = dsobj; 4284 if (ioctl(zhp->zpool_hdl->libzfs_fd, 4285 ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) { 4286 /* just write out a path of two object numbers */ 4287 (void) snprintf(pathname, len, "<0x%llx>:<0x%llx>", 4288 dsobj, obj); 4289 return; 4290 } 4291 (void) strlcpy(dsname, zc.zc_value, sizeof (dsname)); 4292 4293 /* find out if the dataset is mounted */ 4294 mounted = is_mounted(zhp->zpool_hdl, dsname, &mntpnt); 4295 4296 /* get the corrupted object's path */ 4297 (void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name)); 4298 zc.zc_obj = obj; 4299 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_OBJ_TO_PATH, 4300 &zc) == 0) { 4301 if (mounted) { 4302 (void) snprintf(pathname, len, "%s%s", mntpnt, 4303 zc.zc_value); 4304 } else { 4305 (void) snprintf(pathname, len, "%s:%s", 4306 dsname, zc.zc_value); 4307 } 4308 } else { 4309 (void) snprintf(pathname, len, "%s:<0x%llx>", dsname, obj); 4310 } 4311 free(mntpnt); 4312 } 4313 4314 /* 4315 * Read the EFI label from the config, if a label does not exist then 4316 * pass back the error to the caller. If the caller has passed a non-NULL 4317 * diskaddr argument then we set it to the starting address of the EFI 4318 * partition. If the caller has passed a non-NULL boolean argument, then 4319 * we set it to indicate if the disk does have efi system partition. 4320 */ 4321 static int 4322 read_efi_label(nvlist_t *config, diskaddr_t *sb, boolean_t *system) 4323 { 4324 char *path; 4325 int fd; 4326 char diskname[MAXPATHLEN]; 4327 boolean_t boot = B_FALSE; 4328 int err = -1; 4329 int slice; 4330 4331 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PATH, &path) != 0) 4332 return (err); 4333 4334 (void) snprintf(diskname, sizeof (diskname), "%s%s", ZFS_RDISK_ROOT, 4335 strrchr(path, '/')); 4336 if ((fd = open(diskname, O_RDONLY|O_NDELAY)) >= 0) { 4337 struct dk_gpt *vtoc; 4338 4339 if ((err = efi_alloc_and_read(fd, &vtoc)) >= 0) { 4340 for (slice = 0; slice < vtoc->efi_nparts; slice++) { 4341 if (vtoc->efi_parts[slice].p_tag == V_SYSTEM) 4342 boot = B_TRUE; 4343 if (vtoc->efi_parts[slice].p_tag == V_USR) 4344 break; 4345 } 4346 if (sb != NULL && vtoc->efi_parts[slice].p_tag == V_USR) 4347 *sb = vtoc->efi_parts[slice].p_start; 4348 if (system != NULL) 4349 *system = boot; 4350 efi_free(vtoc); 4351 } 4352 (void) close(fd); 4353 } 4354 return (err); 4355 } 4356 4357 /* 4358 * determine where a partition starts on a disk in the current 4359 * configuration 4360 */ 4361 static diskaddr_t 4362 find_start_block(nvlist_t *config) 4363 { 4364 nvlist_t **child; 4365 uint_t c, children; 4366 diskaddr_t sb = MAXOFFSET_T; 4367 uint64_t wholedisk; 4368 4369 if (nvlist_lookup_nvlist_array(config, 4370 ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) { 4371 if (nvlist_lookup_uint64(config, 4372 ZPOOL_CONFIG_WHOLE_DISK, 4373 &wholedisk) != 0 || !wholedisk) { 4374 return (MAXOFFSET_T); 4375 } 4376 if (read_efi_label(config, &sb, NULL) < 0) 4377 sb = MAXOFFSET_T; 4378 return (sb); 4379 } 4380 4381 for (c = 0; c < children; c++) { 4382 sb = find_start_block(child[c]); 4383 if (sb != MAXOFFSET_T) { 4384 return (sb); 4385 } 4386 } 4387 return (MAXOFFSET_T); 4388 } 4389 4390 /* 4391 * Label an individual disk. The name provided is the short name, 4392 * stripped of any leading /dev path. 4393 */ 4394 int 4395 zpool_label_disk(libzfs_handle_t *hdl, zpool_handle_t *zhp, const char *name, 4396 zpool_boot_label_t boot_type, uint64_t boot_size, int *slice) 4397 { 4398 char path[MAXPATHLEN]; 4399 struct dk_gpt *vtoc; 4400 int fd; 4401 size_t resv = EFI_MIN_RESV_SIZE; 4402 uint64_t slice_size; 4403 diskaddr_t start_block; 4404 char errbuf[1024]; 4405 4406 /* prepare an error message just in case */ 4407 (void) snprintf(errbuf, sizeof (errbuf), 4408 dgettext(TEXT_DOMAIN, "cannot label '%s'"), name); 4409 4410 if (zhp) { 4411 nvlist_t *nvroot; 4412 4413 verify(nvlist_lookup_nvlist(zhp->zpool_config, 4414 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 4415 4416 if (zhp->zpool_start_block == 0) 4417 start_block = find_start_block(nvroot); 4418 else 4419 start_block = zhp->zpool_start_block; 4420 zhp->zpool_start_block = start_block; 4421 } else { 4422 /* new pool */ 4423 start_block = NEW_START_BLOCK; 4424 } 4425 4426 (void) snprintf(path, sizeof (path), "%s/%s%s", ZFS_RDISK_ROOT, name, 4427 BACKUP_SLICE); 4428 4429 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) { 4430 /* 4431 * This shouldn't happen. We've long since verified that this 4432 * is a valid device. 4433 */ 4434 zfs_error_aux(hdl, 4435 dgettext(TEXT_DOMAIN, "unable to open device")); 4436 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf)); 4437 } 4438 4439 if (efi_alloc_and_init(fd, EFI_NUMPAR, &vtoc) != 0) { 4440 /* 4441 * The only way this can fail is if we run out of memory, or we 4442 * were unable to read the disk's capacity 4443 */ 4444 if (errno == ENOMEM) 4445 (void) no_memory(hdl); 4446 4447 (void) close(fd); 4448 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4449 "unable to read disk capacity"), name); 4450 4451 return (zfs_error(hdl, EZFS_NOCAP, errbuf)); 4452 } 4453 4454 /* 4455 * Why we use V_USR: V_BACKUP confuses users, and is considered 4456 * disposable by some EFI utilities (since EFI doesn't have a backup 4457 * slice). V_UNASSIGNED is supposed to be used only for zero size 4458 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT, 4459 * etc. were all pretty specific. V_USR is as close to reality as we 4460 * can get, in the absence of V_OTHER. 4461 */ 4462 /* first fix the partition start block */ 4463 if (start_block == MAXOFFSET_T) 4464 start_block = NEW_START_BLOCK; 4465 4466 /* 4467 * EFI System partition is using slice 0. 4468 * ZFS is on slice 1 and slice 8 is reserved. 4469 * We assume the GPT partition table without system 4470 * partition has zfs p_start == NEW_START_BLOCK. 4471 * If start_block != NEW_START_BLOCK, it means we have 4472 * system partition. Correct solution would be to query/cache vtoc 4473 * from existing vdev member. 4474 */ 4475 if (boot_type == ZPOOL_CREATE_BOOT_LABEL) { 4476 if (boot_size % vtoc->efi_lbasize != 0) { 4477 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4478 "boot partition size must be a multiple of %d"), 4479 vtoc->efi_lbasize); 4480 (void) close(fd); 4481 efi_free(vtoc); 4482 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf)); 4483 } 4484 /* 4485 * System partition size checks. 4486 * Note the 1MB is quite arbitrary value, since we 4487 * are creating dedicated pool, it should be enough 4488 * to hold fat + efi bootloader. May need to be 4489 * adjusted if the bootloader size will grow. 4490 */ 4491 if (boot_size < 1024 * 1024) { 4492 char buf[64]; 4493 zfs_nicenum(boot_size, buf, sizeof (buf)); 4494 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4495 "Specified size %s for EFI System partition is too " 4496 "small, the minimum size is 1MB."), buf); 4497 (void) close(fd); 4498 efi_free(vtoc); 4499 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf)); 4500 } 4501 /* 33MB is tested with mkfs -F pcfs */ 4502 if (hdl->libzfs_printerr && 4503 ((vtoc->efi_lbasize == 512 && 4504 boot_size < 33 * 1024 * 1024) || 4505 (vtoc->efi_lbasize == 4096 && 4506 boot_size < 256 * 1024 * 1024))) { 4507 char buf[64]; 4508 zfs_nicenum(boot_size, buf, sizeof (buf)); 4509 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, 4510 "Warning: EFI System partition size %s is " 4511 "not allowing to create FAT32 file\nsystem, which " 4512 "may result in unbootable system.\n"), buf); 4513 } 4514 /* Adjust zfs partition start by size of system partition. */ 4515 start_block += boot_size / vtoc->efi_lbasize; 4516 } 4517 4518 if (start_block == NEW_START_BLOCK) { 4519 /* 4520 * Use default layout. 4521 * ZFS is on slice 0 and slice 8 is reserved. 4522 */ 4523 slice_size = vtoc->efi_last_u_lba + 1; 4524 slice_size -= EFI_MIN_RESV_SIZE; 4525 slice_size -= start_block; 4526 if (slice != NULL) 4527 *slice = 0; 4528 4529 vtoc->efi_parts[0].p_start = start_block; 4530 vtoc->efi_parts[0].p_size = slice_size; 4531 4532 vtoc->efi_parts[0].p_tag = V_USR; 4533 (void) strcpy(vtoc->efi_parts[0].p_name, "zfs"); 4534 4535 vtoc->efi_parts[8].p_start = slice_size + start_block; 4536 vtoc->efi_parts[8].p_size = resv; 4537 vtoc->efi_parts[8].p_tag = V_RESERVED; 4538 } else { 4539 slice_size = start_block - NEW_START_BLOCK; 4540 vtoc->efi_parts[0].p_start = NEW_START_BLOCK; 4541 vtoc->efi_parts[0].p_size = slice_size; 4542 vtoc->efi_parts[0].p_tag = V_SYSTEM; 4543 (void) strcpy(vtoc->efi_parts[0].p_name, "loader"); 4544 if (slice != NULL) 4545 *slice = 1; 4546 /* prepare slice 1 */ 4547 slice_size = vtoc->efi_last_u_lba + 1 - slice_size; 4548 slice_size -= resv; 4549 slice_size -= NEW_START_BLOCK; 4550 vtoc->efi_parts[1].p_start = start_block; 4551 vtoc->efi_parts[1].p_size = slice_size; 4552 vtoc->efi_parts[1].p_tag = V_USR; 4553 (void) strcpy(vtoc->efi_parts[1].p_name, "zfs"); 4554 4555 vtoc->efi_parts[8].p_start = slice_size + start_block; 4556 vtoc->efi_parts[8].p_size = resv; 4557 vtoc->efi_parts[8].p_tag = V_RESERVED; 4558 } 4559 4560 if (efi_write(fd, vtoc) != 0) { 4561 /* 4562 * Some block drivers (like pcata) may not support EFI 4563 * GPT labels. Print out a helpful error message dir- 4564 * ecting the user to manually label the disk and give 4565 * a specific slice. 4566 */ 4567 (void) close(fd); 4568 efi_free(vtoc); 4569 4570 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4571 "try using fdisk(1M) and then provide a specific slice")); 4572 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf)); 4573 } 4574 4575 (void) close(fd); 4576 efi_free(vtoc); 4577 return (0); 4578 } 4579 4580 static boolean_t 4581 supported_dump_vdev_type(libzfs_handle_t *hdl, nvlist_t *config, char *errbuf) 4582 { 4583 char *type; 4584 nvlist_t **child; 4585 uint_t children, c; 4586 4587 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_TYPE, &type) == 0); 4588 if (strcmp(type, VDEV_TYPE_FILE) == 0 || 4589 strcmp(type, VDEV_TYPE_HOLE) == 0 || 4590 strcmp(type, VDEV_TYPE_MISSING) == 0) { 4591 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4592 "vdev type '%s' is not supported"), type); 4593 (void) zfs_error(hdl, EZFS_VDEVNOTSUP, errbuf); 4594 return (B_FALSE); 4595 } 4596 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN, 4597 &child, &children) == 0) { 4598 for (c = 0; c < children; c++) { 4599 if (!supported_dump_vdev_type(hdl, child[c], errbuf)) 4600 return (B_FALSE); 4601 } 4602 } 4603 return (B_TRUE); 4604 } 4605 4606 /* 4607 * Check if this zvol is allowable for use as a dump device; zero if 4608 * it is, > 0 if it isn't, < 0 if it isn't a zvol. 4609 * 4610 * Allowable storage configurations include mirrors, all raidz variants, and 4611 * pools with log, cache, and spare devices. Pools which are backed by files or 4612 * have missing/hole vdevs are not suitable. 4613 */ 4614 int 4615 zvol_check_dump_config(char *arg) 4616 { 4617 zpool_handle_t *zhp = NULL; 4618 nvlist_t *config, *nvroot; 4619 char *p, *volname; 4620 nvlist_t **top; 4621 uint_t toplevels; 4622 libzfs_handle_t *hdl; 4623 char errbuf[1024]; 4624 char poolname[ZFS_MAX_DATASET_NAME_LEN]; 4625 int pathlen = strlen(ZVOL_FULL_DEV_DIR); 4626 int ret = 1; 4627 4628 if (strncmp(arg, ZVOL_FULL_DEV_DIR, pathlen)) { 4629 return (-1); 4630 } 4631 4632 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, 4633 "dump is not supported on device '%s'"), arg); 4634 4635 if ((hdl = libzfs_init()) == NULL) 4636 return (1); 4637 libzfs_print_on_error(hdl, B_TRUE); 4638 4639 volname = arg + pathlen; 4640 4641 /* check the configuration of the pool */ 4642 if ((p = strchr(volname, '/')) == NULL) { 4643 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4644 "malformed dataset name")); 4645 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf); 4646 return (1); 4647 } else if (p - volname >= ZFS_MAX_DATASET_NAME_LEN) { 4648 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4649 "dataset name is too long")); 4650 (void) zfs_error(hdl, EZFS_NAMETOOLONG, errbuf); 4651 return (1); 4652 } else { 4653 (void) strncpy(poolname, volname, p - volname); 4654 poolname[p - volname] = '\0'; 4655 } 4656 4657 if ((zhp = zpool_open(hdl, poolname)) == NULL) { 4658 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4659 "could not open pool '%s'"), poolname); 4660 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf); 4661 goto out; 4662 } 4663 config = zpool_get_config(zhp, NULL); 4664 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 4665 &nvroot) != 0) { 4666 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4667 "could not obtain vdev configuration for '%s'"), poolname); 4668 (void) zfs_error(hdl, EZFS_INVALCONFIG, errbuf); 4669 goto out; 4670 } 4671 4672 verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, 4673 &top, &toplevels) == 0); 4674 4675 if (!supported_dump_vdev_type(hdl, top[0], errbuf)) { 4676 goto out; 4677 } 4678 ret = 0; 4679 4680 out: 4681 if (zhp) 4682 zpool_close(zhp); 4683 libzfs_fini(hdl); 4684 return (ret); 4685 } 4686