1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 22 /* 23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 24 * Copyright (c) 2011, 2020 by Delphix. All rights reserved. 25 * Copyright 2019 Joyent, Inc. 26 * Copyright 2016 Nexenta Systems, Inc. 27 * Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com> 28 * Copyright (c) 2017 Datto Inc. 29 * Copyright (c) 2017, Intel Corporation. 30 * Copyright 2022 OmniOS Community Edition (OmniOSce) Association. 31 * Copyright 2022 Oxide Computer Company 32 */ 33 34 #include <ctype.h> 35 #include <errno.h> 36 #include <devid.h> 37 #include <fcntl.h> 38 #include <libintl.h> 39 #include <stdio.h> 40 #include <stdlib.h> 41 #include <strings.h> 42 #include <unistd.h> 43 #include <libgen.h> 44 #include <sys/dkio.h> 45 #include <sys/efi_partition.h> 46 #include <sys/vtoc.h> 47 #include <sys/zfs_ioctl.h> 48 #include <sys/modctl.h> 49 #include <sys/mkdev.h> 50 #include <dlfcn.h> 51 #include <libzutil.h> 52 53 #include "zfs_namecheck.h" 54 #include "zfs_prop.h" 55 #include "libzfs_impl.h" 56 #include "zfs_comutil.h" 57 #include "zfeature_common.h" 58 59 static int read_efi_label(nvlist_t *, diskaddr_t *, boolean_t *); 60 static boolean_t zpool_vdev_is_interior(const char *name); 61 62 #define BACKUP_SLICE "s2" 63 64 typedef struct prop_flags { 65 int create:1; /* Validate property on creation */ 66 int import:1; /* Validate property on import */ 67 } prop_flags_t; 68 69 /* 70 * ==================================================================== 71 * zpool property functions 72 * ==================================================================== 73 */ 74 75 static int 76 zpool_get_all_props(zpool_handle_t *zhp) 77 { 78 zfs_cmd_t zc = { 0 }; 79 libzfs_handle_t *hdl = zhp->zpool_hdl; 80 81 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 82 83 if (zcmd_alloc_dst_nvlist(hdl, &zc, 0) != 0) 84 return (-1); 85 86 while (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_PROPS, &zc) != 0) { 87 if (errno == ENOMEM) { 88 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) { 89 zcmd_free_nvlists(&zc); 90 return (-1); 91 } 92 } else { 93 zcmd_free_nvlists(&zc); 94 return (-1); 95 } 96 } 97 98 if (zcmd_read_dst_nvlist(hdl, &zc, &zhp->zpool_props) != 0) { 99 zcmd_free_nvlists(&zc); 100 return (-1); 101 } 102 103 zcmd_free_nvlists(&zc); 104 105 return (0); 106 } 107 108 static int 109 zpool_props_refresh(zpool_handle_t *zhp) 110 { 111 nvlist_t *old_props; 112 113 old_props = zhp->zpool_props; 114 115 if (zpool_get_all_props(zhp) != 0) 116 return (-1); 117 118 nvlist_free(old_props); 119 return (0); 120 } 121 122 static char * 123 zpool_get_prop_string(zpool_handle_t *zhp, zpool_prop_t prop, 124 zprop_source_t *src) 125 { 126 nvlist_t *nv, *nvl; 127 uint64_t ival; 128 char *value; 129 zprop_source_t source; 130 131 nvl = zhp->zpool_props; 132 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) { 133 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &ival) == 0); 134 source = ival; 135 verify(nvlist_lookup_string(nv, ZPROP_VALUE, &value) == 0); 136 } else { 137 source = ZPROP_SRC_DEFAULT; 138 if ((value = (char *)zpool_prop_default_string(prop)) == NULL) 139 value = "-"; 140 } 141 142 if (src) 143 *src = source; 144 145 return (value); 146 } 147 148 uint64_t 149 zpool_get_prop_int(zpool_handle_t *zhp, zpool_prop_t prop, zprop_source_t *src) 150 { 151 nvlist_t *nv, *nvl; 152 uint64_t value; 153 zprop_source_t source; 154 155 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp)) { 156 /* 157 * zpool_get_all_props() has most likely failed because 158 * the pool is faulted, but if all we need is the top level 159 * vdev's guid then get it from the zhp config nvlist. 160 */ 161 if ((prop == ZPOOL_PROP_GUID) && 162 (nvlist_lookup_nvlist(zhp->zpool_config, 163 ZPOOL_CONFIG_VDEV_TREE, &nv) == 0) && 164 (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value) 165 == 0)) { 166 return (value); 167 } 168 return (zpool_prop_default_numeric(prop)); 169 } 170 171 nvl = zhp->zpool_props; 172 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) { 173 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &value) == 0); 174 source = value; 175 verify(nvlist_lookup_uint64(nv, ZPROP_VALUE, &value) == 0); 176 } else { 177 source = ZPROP_SRC_DEFAULT; 178 value = zpool_prop_default_numeric(prop); 179 } 180 181 if (src) 182 *src = source; 183 184 return (value); 185 } 186 187 /* 188 * Map VDEV STATE to printed strings. 189 */ 190 const char * 191 zpool_state_to_name(vdev_state_t state, vdev_aux_t aux) 192 { 193 switch (state) { 194 case VDEV_STATE_CLOSED: 195 case VDEV_STATE_OFFLINE: 196 return (gettext("OFFLINE")); 197 case VDEV_STATE_REMOVED: 198 return (gettext("REMOVED")); 199 case VDEV_STATE_CANT_OPEN: 200 if (aux == VDEV_AUX_CORRUPT_DATA || aux == VDEV_AUX_BAD_LOG) 201 return (gettext("FAULTED")); 202 else if (aux == VDEV_AUX_SPLIT_POOL) 203 return (gettext("SPLIT")); 204 else 205 return (gettext("UNAVAIL")); 206 case VDEV_STATE_FAULTED: 207 return (gettext("FAULTED")); 208 case VDEV_STATE_DEGRADED: 209 return (gettext("DEGRADED")); 210 case VDEV_STATE_HEALTHY: 211 return (gettext("ONLINE")); 212 213 default: 214 break; 215 } 216 217 return (gettext("UNKNOWN")); 218 } 219 220 /* 221 * Map POOL STATE to printed strings. 222 */ 223 const char * 224 zpool_pool_state_to_name(pool_state_t state) 225 { 226 switch (state) { 227 case POOL_STATE_ACTIVE: 228 return (gettext("ACTIVE")); 229 case POOL_STATE_EXPORTED: 230 return (gettext("EXPORTED")); 231 case POOL_STATE_DESTROYED: 232 return (gettext("DESTROYED")); 233 case POOL_STATE_SPARE: 234 return (gettext("SPARE")); 235 case POOL_STATE_L2CACHE: 236 return (gettext("L2CACHE")); 237 case POOL_STATE_UNINITIALIZED: 238 return (gettext("UNINITIALIZED")); 239 case POOL_STATE_UNAVAIL: 240 return (gettext("UNAVAIL")); 241 case POOL_STATE_POTENTIALLY_ACTIVE: 242 return (gettext("POTENTIALLY_ACTIVE")); 243 } 244 245 return (gettext("UNKNOWN")); 246 } 247 248 /* 249 * Get a zpool property value for 'prop' and return the value in 250 * a pre-allocated buffer. 251 */ 252 int 253 zpool_get_prop(zpool_handle_t *zhp, zpool_prop_t prop, char *buf, size_t len, 254 zprop_source_t *srctype, boolean_t literal) 255 { 256 uint64_t intval; 257 const char *strval; 258 zprop_source_t src = ZPROP_SRC_NONE; 259 nvlist_t *nvroot; 260 vdev_stat_t *vs; 261 uint_t vsc; 262 263 if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) { 264 switch (prop) { 265 case ZPOOL_PROP_NAME: 266 (void) strlcpy(buf, zpool_get_name(zhp), len); 267 break; 268 269 case ZPOOL_PROP_HEALTH: 270 (void) strlcpy(buf, "FAULTED", len); 271 break; 272 273 case ZPOOL_PROP_GUID: 274 intval = zpool_get_prop_int(zhp, prop, &src); 275 (void) snprintf(buf, len, "%llu", intval); 276 break; 277 278 case ZPOOL_PROP_ALTROOT: 279 case ZPOOL_PROP_CACHEFILE: 280 case ZPOOL_PROP_COMMENT: 281 if (zhp->zpool_props != NULL || 282 zpool_get_all_props(zhp) == 0) { 283 (void) strlcpy(buf, 284 zpool_get_prop_string(zhp, prop, &src), 285 len); 286 break; 287 } 288 /* FALLTHROUGH */ 289 default: 290 (void) strlcpy(buf, "-", len); 291 break; 292 } 293 294 if (srctype != NULL) 295 *srctype = src; 296 return (0); 297 } 298 299 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp) && 300 prop != ZPOOL_PROP_NAME) 301 return (-1); 302 303 switch (zpool_prop_get_type(prop)) { 304 case PROP_TYPE_STRING: 305 (void) strlcpy(buf, zpool_get_prop_string(zhp, prop, &src), 306 len); 307 break; 308 309 case PROP_TYPE_NUMBER: 310 intval = zpool_get_prop_int(zhp, prop, &src); 311 312 switch (prop) { 313 case ZPOOL_PROP_SIZE: 314 case ZPOOL_PROP_ALLOCATED: 315 case ZPOOL_PROP_FREE: 316 case ZPOOL_PROP_FREEING: 317 case ZPOOL_PROP_LEAKED: 318 case ZPOOL_PROP_ASHIFT: 319 if (literal) { 320 (void) snprintf(buf, len, "%llu", 321 (u_longlong_t)intval); 322 } else { 323 (void) zfs_nicenum(intval, buf, len); 324 } 325 break; 326 case ZPOOL_PROP_BOOTSIZE: 327 case ZPOOL_PROP_EXPANDSZ: 328 case ZPOOL_PROP_CHECKPOINT: 329 if (intval == 0) { 330 (void) strlcpy(buf, "-", len); 331 } else if (literal) { 332 (void) snprintf(buf, len, "%llu", 333 (u_longlong_t)intval); 334 } else { 335 (void) zfs_nicenum(intval, buf, len); 336 } 337 break; 338 case ZPOOL_PROP_CAPACITY: 339 if (literal) { 340 (void) snprintf(buf, len, "%llu", 341 (u_longlong_t)intval); 342 } else { 343 (void) snprintf(buf, len, "%llu%%", 344 (u_longlong_t)intval); 345 } 346 break; 347 case ZPOOL_PROP_FRAGMENTATION: 348 if (intval == UINT64_MAX) { 349 (void) strlcpy(buf, "-", len); 350 } else if (literal) { 351 (void) snprintf(buf, len, "%llu", 352 (u_longlong_t)intval); 353 } else { 354 (void) snprintf(buf, len, "%llu%%", 355 (u_longlong_t)intval); 356 } 357 break; 358 case ZPOOL_PROP_DEDUPRATIO: 359 if (literal) 360 (void) snprintf(buf, len, "%llu.%02llu", 361 (u_longlong_t)(intval / 100), 362 (u_longlong_t)(intval % 100)); 363 else 364 (void) snprintf(buf, len, "%llu.%02llux", 365 (u_longlong_t)(intval / 100), 366 (u_longlong_t)(intval % 100)); 367 break; 368 case ZPOOL_PROP_HEALTH: 369 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL), 370 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 371 verify(nvlist_lookup_uint64_array(nvroot, 372 ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc) 373 == 0); 374 375 (void) strlcpy(buf, zpool_state_to_name(intval, 376 vs->vs_aux), len); 377 break; 378 case ZPOOL_PROP_VERSION: 379 if (intval >= SPA_VERSION_FEATURES) { 380 (void) snprintf(buf, len, "-"); 381 break; 382 } 383 /* FALLTHROUGH */ 384 default: 385 (void) snprintf(buf, len, "%llu", intval); 386 } 387 break; 388 389 case PROP_TYPE_INDEX: 390 intval = zpool_get_prop_int(zhp, prop, &src); 391 if (zpool_prop_index_to_string(prop, intval, &strval) 392 != 0) 393 return (-1); 394 (void) strlcpy(buf, strval, len); 395 break; 396 397 default: 398 abort(); 399 } 400 401 if (srctype) 402 *srctype = src; 403 404 return (0); 405 } 406 407 /* 408 * Check if the bootfs name has the same pool name as it is set to. 409 * Assuming bootfs is a valid dataset name. 410 */ 411 static boolean_t 412 bootfs_name_valid(const char *pool, const char *bootfs) 413 { 414 int len = strlen(pool); 415 if (bootfs[0] == '\0') 416 return (B_TRUE); 417 418 if (!zfs_name_valid(bootfs, ZFS_TYPE_FILESYSTEM|ZFS_TYPE_SNAPSHOT)) 419 return (B_FALSE); 420 421 if (strncmp(pool, bootfs, len) == 0 && 422 (bootfs[len] == '/' || bootfs[len] == '\0')) 423 return (B_TRUE); 424 425 return (B_FALSE); 426 } 427 428 boolean_t 429 zpool_is_bootable(zpool_handle_t *zhp) 430 { 431 char bootfs[ZFS_MAX_DATASET_NAME_LEN]; 432 433 return (zpool_get_prop(zhp, ZPOOL_PROP_BOOTFS, bootfs, 434 sizeof (bootfs), NULL, B_FALSE) == 0 && strncmp(bootfs, "-", 435 sizeof (bootfs)) != 0); 436 } 437 438 439 /* 440 * Given an nvlist of zpool properties to be set, validate that they are 441 * correct, and parse any numeric properties (index, boolean, etc) if they are 442 * specified as strings. 443 */ 444 static nvlist_t * 445 zpool_valid_proplist(libzfs_handle_t *hdl, const char *poolname, 446 nvlist_t *props, uint64_t version, prop_flags_t flags, char *errbuf) 447 { 448 nvpair_t *elem; 449 nvlist_t *retprops; 450 zpool_prop_t prop; 451 char *strval; 452 uint64_t intval; 453 char *slash, *check; 454 struct stat64 statbuf; 455 zpool_handle_t *zhp; 456 457 if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) { 458 (void) no_memory(hdl); 459 return (NULL); 460 } 461 462 elem = NULL; 463 while ((elem = nvlist_next_nvpair(props, elem)) != NULL) { 464 const char *propname = nvpair_name(elem); 465 466 prop = zpool_name_to_prop(propname); 467 if (prop == ZPOOL_PROP_INVAL && zpool_prop_feature(propname)) { 468 int err; 469 char *fname = strchr(propname, '@') + 1; 470 471 err = zfeature_lookup_name(fname, NULL); 472 if (err != 0) { 473 ASSERT3U(err, ==, ENOENT); 474 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 475 "invalid feature '%s', '%s'"), fname, 476 propname); 477 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 478 goto error; 479 } 480 481 if (nvpair_type(elem) != DATA_TYPE_STRING) { 482 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 483 "'%s' must be a string"), propname); 484 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 485 goto error; 486 } 487 488 (void) nvpair_value_string(elem, &strval); 489 if (strcmp(strval, ZFS_FEATURE_ENABLED) != 0 && 490 strcmp(strval, ZFS_FEATURE_DISABLED) != 0) { 491 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 492 "property '%s' can only be set to " 493 "'enabled' or 'disabled'"), propname); 494 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 495 goto error; 496 } 497 498 if (nvlist_add_uint64(retprops, propname, 0) != 0) { 499 (void) no_memory(hdl); 500 goto error; 501 } 502 continue; 503 } 504 505 /* 506 * Make sure this property is valid and applies to this type. 507 */ 508 if (prop == ZPOOL_PROP_INVAL) { 509 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 510 "invalid property '%s'"), propname); 511 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 512 goto error; 513 } 514 515 if (zpool_prop_readonly(prop)) { 516 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' " 517 "is readonly"), propname); 518 (void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf); 519 goto error; 520 } 521 522 if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_POOL, retprops, 523 &strval, &intval, errbuf) != 0) 524 goto error; 525 526 /* 527 * Perform additional checking for specific properties. 528 */ 529 switch (prop) { 530 case ZPOOL_PROP_VERSION: 531 if (intval < version || 532 !SPA_VERSION_IS_SUPPORTED(intval)) { 533 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 534 "property '%s' number %d is invalid."), 535 propname, intval); 536 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf); 537 goto error; 538 } 539 break; 540 541 case ZPOOL_PROP_BOOTSIZE: 542 if (!flags.create) { 543 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 544 "property '%s' can only be set during pool " 545 "creation"), propname); 546 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 547 goto error; 548 } 549 break; 550 551 case ZPOOL_PROP_ASHIFT: 552 if (intval != 0 && 553 (intval < ASHIFT_MIN || intval > ASHIFT_MAX)) { 554 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 555 "invalid '%s=%d' property: only values " 556 "between %" PRId32 " and %" PRId32 " " 557 "are allowed.\n"), 558 propname, intval, ASHIFT_MIN, ASHIFT_MAX); 559 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 560 goto error; 561 } 562 break; 563 564 case ZPOOL_PROP_BOOTFS: 565 if (flags.create || flags.import) { 566 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 567 "property '%s' cannot be set at creation " 568 "or import time"), propname); 569 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 570 goto error; 571 } 572 573 if (version < SPA_VERSION_BOOTFS) { 574 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 575 "pool must be upgraded to support " 576 "'%s' property"), propname); 577 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf); 578 goto error; 579 } 580 581 /* 582 * bootfs property value has to be a dataset name and 583 * the dataset has to be in the same pool as it sets to. 584 */ 585 if (!bootfs_name_valid(poolname, strval)) { 586 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' " 587 "is an invalid name"), strval); 588 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf); 589 goto error; 590 } 591 592 if ((zhp = zpool_open_canfail(hdl, poolname)) == NULL) { 593 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 594 "could not open pool '%s'"), poolname); 595 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf); 596 goto error; 597 } 598 zpool_close(zhp); 599 break; 600 601 case ZPOOL_PROP_ALTROOT: 602 if (!flags.create && !flags.import) { 603 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 604 "property '%s' can only be set during pool " 605 "creation or import"), propname); 606 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 607 goto error; 608 } 609 610 if (strval[0] != '/') { 611 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 612 "bad alternate root '%s'"), strval); 613 (void) zfs_error(hdl, EZFS_BADPATH, errbuf); 614 goto error; 615 } 616 break; 617 618 case ZPOOL_PROP_CACHEFILE: 619 if (strval[0] == '\0') 620 break; 621 622 if (strcmp(strval, "none") == 0) 623 break; 624 625 if (strval[0] != '/') { 626 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 627 "property '%s' must be empty, an " 628 "absolute path, or 'none'"), propname); 629 (void) zfs_error(hdl, EZFS_BADPATH, errbuf); 630 goto error; 631 } 632 633 slash = strrchr(strval, '/'); 634 635 if (slash[1] == '\0' || strcmp(slash, "/.") == 0 || 636 strcmp(slash, "/..") == 0) { 637 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 638 "'%s' is not a valid file"), strval); 639 (void) zfs_error(hdl, EZFS_BADPATH, errbuf); 640 goto error; 641 } 642 643 *slash = '\0'; 644 645 if (strval[0] != '\0' && 646 (stat64(strval, &statbuf) != 0 || 647 !S_ISDIR(statbuf.st_mode))) { 648 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 649 "'%s' is not a valid directory"), 650 strval); 651 (void) zfs_error(hdl, EZFS_BADPATH, errbuf); 652 goto error; 653 } 654 655 *slash = '/'; 656 break; 657 658 case ZPOOL_PROP_COMMENT: 659 for (check = strval; *check != '\0'; check++) { 660 if (!isprint(*check)) { 661 zfs_error_aux(hdl, 662 dgettext(TEXT_DOMAIN, 663 "comment may only have printable " 664 "characters")); 665 (void) zfs_error(hdl, EZFS_BADPROP, 666 errbuf); 667 goto error; 668 } 669 } 670 if (strlen(strval) > ZPROP_MAX_COMMENT) { 671 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 672 "comment must not exceed %d characters"), 673 ZPROP_MAX_COMMENT); 674 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 675 goto error; 676 } 677 break; 678 679 case ZPOOL_PROP_READONLY: 680 if (!flags.import) { 681 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 682 "property '%s' can only be set at " 683 "import time"), propname); 684 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 685 goto error; 686 } 687 break; 688 689 case ZPOOL_PROP_TNAME: 690 if (!flags.create) { 691 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 692 "property '%s' can only be set at " 693 "creation time"), propname); 694 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 695 goto error; 696 } 697 break; 698 699 case ZPOOL_PROP_MULTIHOST: 700 if (get_system_hostid() == 0) { 701 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 702 "requires a non-zero system hostid")); 703 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 704 goto error; 705 } 706 break; 707 708 default: 709 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 710 "property '%s'(%d) not defined"), propname, prop); 711 break; 712 } 713 } 714 715 return (retprops); 716 error: 717 nvlist_free(retprops); 718 return (NULL); 719 } 720 721 /* 722 * Set zpool property : propname=propval. 723 */ 724 int 725 zpool_set_prop(zpool_handle_t *zhp, const char *propname, const char *propval) 726 { 727 zfs_cmd_t zc = { 0 }; 728 int ret = -1; 729 char errbuf[1024]; 730 nvlist_t *nvl = NULL; 731 nvlist_t *realprops; 732 uint64_t version; 733 prop_flags_t flags = { 0 }; 734 735 (void) snprintf(errbuf, sizeof (errbuf), 736 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"), 737 zhp->zpool_name); 738 739 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0) 740 return (no_memory(zhp->zpool_hdl)); 741 742 if (nvlist_add_string(nvl, propname, propval) != 0) { 743 nvlist_free(nvl); 744 return (no_memory(zhp->zpool_hdl)); 745 } 746 747 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL); 748 if ((realprops = zpool_valid_proplist(zhp->zpool_hdl, 749 zhp->zpool_name, nvl, version, flags, errbuf)) == NULL) { 750 nvlist_free(nvl); 751 return (-1); 752 } 753 754 nvlist_free(nvl); 755 nvl = realprops; 756 757 /* 758 * Execute the corresponding ioctl() to set this property. 759 */ 760 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 761 762 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) { 763 nvlist_free(nvl); 764 return (-1); 765 } 766 767 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc); 768 769 zcmd_free_nvlists(&zc); 770 nvlist_free(nvl); 771 772 if (ret) 773 (void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf); 774 else 775 (void) zpool_props_refresh(zhp); 776 777 return (ret); 778 } 779 780 int 781 zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp) 782 { 783 libzfs_handle_t *hdl = zhp->zpool_hdl; 784 zprop_list_t *entry; 785 char buf[ZFS_MAXPROPLEN]; 786 nvlist_t *features = NULL; 787 zprop_list_t **last; 788 boolean_t firstexpand = (NULL == *plp); 789 790 if (zprop_expand_list(hdl, plp, ZFS_TYPE_POOL) != 0) 791 return (-1); 792 793 last = plp; 794 while (*last != NULL) 795 last = &(*last)->pl_next; 796 797 if ((*plp)->pl_all) 798 features = zpool_get_features(zhp); 799 800 if ((*plp)->pl_all && firstexpand) { 801 for (int i = 0; i < SPA_FEATURES; i++) { 802 zprop_list_t *entry = zfs_alloc(hdl, 803 sizeof (zprop_list_t)); 804 entry->pl_prop = ZPROP_INVAL; 805 entry->pl_user_prop = zfs_asprintf(hdl, "feature@%s", 806 spa_feature_table[i].fi_uname); 807 entry->pl_width = strlen(entry->pl_user_prop); 808 entry->pl_all = B_TRUE; 809 810 *last = entry; 811 last = &entry->pl_next; 812 } 813 } 814 815 /* add any unsupported features */ 816 for (nvpair_t *nvp = nvlist_next_nvpair(features, NULL); 817 nvp != NULL; nvp = nvlist_next_nvpair(features, nvp)) { 818 char *propname; 819 boolean_t found; 820 zprop_list_t *entry; 821 822 if (zfeature_is_supported(nvpair_name(nvp))) 823 continue; 824 825 propname = zfs_asprintf(hdl, "unsupported@%s", 826 nvpair_name(nvp)); 827 828 /* 829 * Before adding the property to the list make sure that no 830 * other pool already added the same property. 831 */ 832 found = B_FALSE; 833 entry = *plp; 834 while (entry != NULL) { 835 if (entry->pl_user_prop != NULL && 836 strcmp(propname, entry->pl_user_prop) == 0) { 837 found = B_TRUE; 838 break; 839 } 840 entry = entry->pl_next; 841 } 842 if (found) { 843 free(propname); 844 continue; 845 } 846 847 entry = zfs_alloc(hdl, sizeof (zprop_list_t)); 848 entry->pl_prop = ZPROP_INVAL; 849 entry->pl_user_prop = propname; 850 entry->pl_width = strlen(entry->pl_user_prop); 851 entry->pl_all = B_TRUE; 852 853 *last = entry; 854 last = &entry->pl_next; 855 } 856 857 for (entry = *plp; entry != NULL; entry = entry->pl_next) { 858 859 if (entry->pl_fixed) 860 continue; 861 862 if (entry->pl_prop != ZPROP_INVAL && 863 zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf), 864 NULL, B_FALSE) == 0) { 865 if (strlen(buf) > entry->pl_width) 866 entry->pl_width = strlen(buf); 867 } 868 } 869 870 return (0); 871 } 872 873 /* 874 * Get the state for the given feature on the given ZFS pool. 875 */ 876 int 877 zpool_prop_get_feature(zpool_handle_t *zhp, const char *propname, char *buf, 878 size_t len) 879 { 880 uint64_t refcount; 881 boolean_t found = B_FALSE; 882 nvlist_t *features = zpool_get_features(zhp); 883 boolean_t supported; 884 const char *feature = strchr(propname, '@') + 1; 885 886 supported = zpool_prop_feature(propname); 887 ASSERT(supported || zpool_prop_unsupported(propname)); 888 889 /* 890 * Convert from feature name to feature guid. This conversion is 891 * unecessary for unsupported@... properties because they already 892 * use guids. 893 */ 894 if (supported) { 895 int ret; 896 spa_feature_t fid; 897 898 ret = zfeature_lookup_name(feature, &fid); 899 if (ret != 0) { 900 (void) strlcpy(buf, "-", len); 901 return (ENOTSUP); 902 } 903 feature = spa_feature_table[fid].fi_guid; 904 } 905 906 if (nvlist_lookup_uint64(features, feature, &refcount) == 0) 907 found = B_TRUE; 908 909 if (supported) { 910 if (!found) { 911 (void) strlcpy(buf, ZFS_FEATURE_DISABLED, len); 912 } else { 913 if (refcount == 0) 914 (void) strlcpy(buf, ZFS_FEATURE_ENABLED, len); 915 else 916 (void) strlcpy(buf, ZFS_FEATURE_ACTIVE, len); 917 } 918 } else { 919 if (found) { 920 if (refcount == 0) { 921 (void) strcpy(buf, ZFS_UNSUPPORTED_INACTIVE); 922 } else { 923 (void) strcpy(buf, ZFS_UNSUPPORTED_READONLY); 924 } 925 } else { 926 (void) strlcpy(buf, "-", len); 927 return (ENOTSUP); 928 } 929 } 930 931 return (0); 932 } 933 934 /* 935 * Don't start the slice at the default block of 34; many storage 936 * devices will use a stripe width of 128k, so start there instead. 937 */ 938 #define NEW_START_BLOCK 256 939 940 /* 941 * Validate the given pool name, optionally putting an extended error message in 942 * 'buf'. 943 */ 944 boolean_t 945 zpool_name_valid(libzfs_handle_t *hdl, boolean_t isopen, const char *pool) 946 { 947 namecheck_err_t why; 948 char what; 949 int ret; 950 951 ret = pool_namecheck(pool, &why, &what); 952 953 /* 954 * The rules for reserved pool names were extended at a later point. 955 * But we need to support users with existing pools that may now be 956 * invalid. So we only check for this expanded set of names during a 957 * create (or import), and only in userland. 958 */ 959 if (ret == 0 && !isopen && 960 (strncmp(pool, "mirror", 6) == 0 || 961 strncmp(pool, "raidz", 5) == 0 || 962 strncmp(pool, "spare", 5) == 0 || 963 strcmp(pool, "log") == 0)) { 964 if (hdl != NULL) 965 zfs_error_aux(hdl, 966 dgettext(TEXT_DOMAIN, "name is reserved")); 967 return (B_FALSE); 968 } 969 970 971 if (ret != 0) { 972 if (hdl != NULL) { 973 switch (why) { 974 case NAME_ERR_TOOLONG: 975 zfs_error_aux(hdl, 976 dgettext(TEXT_DOMAIN, "name is too long")); 977 break; 978 979 case NAME_ERR_INVALCHAR: 980 zfs_error_aux(hdl, 981 dgettext(TEXT_DOMAIN, "invalid character " 982 "'%c' in pool name"), what); 983 break; 984 985 case NAME_ERR_NOLETTER: 986 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 987 "name must begin with a letter")); 988 break; 989 990 case NAME_ERR_RESERVED: 991 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 992 "name is reserved")); 993 break; 994 995 case NAME_ERR_DISKLIKE: 996 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 997 "pool name is reserved")); 998 break; 999 1000 case NAME_ERR_LEADING_SLASH: 1001 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1002 "leading slash in name")); 1003 break; 1004 1005 case NAME_ERR_EMPTY_COMPONENT: 1006 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1007 "empty component in name")); 1008 break; 1009 1010 case NAME_ERR_TRAILING_SLASH: 1011 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1012 "trailing slash in name")); 1013 break; 1014 1015 case NAME_ERR_MULTIPLE_DELIMITERS: 1016 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1017 "multiple '@' and/or '#' delimiters in " 1018 "name")); 1019 break; 1020 1021 default: 1022 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1023 "(%d) not defined"), why); 1024 break; 1025 } 1026 } 1027 return (B_FALSE); 1028 } 1029 1030 return (B_TRUE); 1031 } 1032 1033 /* 1034 * Open a handle to the given pool, even if the pool is currently in the FAULTED 1035 * state. 1036 */ 1037 zpool_handle_t * 1038 zpool_open_canfail(libzfs_handle_t *hdl, const char *pool) 1039 { 1040 zpool_handle_t *zhp; 1041 boolean_t missing; 1042 1043 /* 1044 * Make sure the pool name is valid. 1045 */ 1046 if (!zpool_name_valid(hdl, B_TRUE, pool)) { 1047 (void) zfs_error_fmt(hdl, EZFS_INVALIDNAME, 1048 dgettext(TEXT_DOMAIN, "cannot open '%s'"), 1049 pool); 1050 return (NULL); 1051 } 1052 1053 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL) 1054 return (NULL); 1055 1056 zhp->zpool_hdl = hdl; 1057 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name)); 1058 1059 if (zpool_refresh_stats(zhp, &missing) != 0) { 1060 zpool_close(zhp); 1061 return (NULL); 1062 } 1063 1064 if (missing) { 1065 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool")); 1066 (void) zfs_error_fmt(hdl, EZFS_NOENT, 1067 dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool); 1068 zpool_close(zhp); 1069 return (NULL); 1070 } 1071 1072 return (zhp); 1073 } 1074 1075 /* 1076 * Like the above, but silent on error. Used when iterating over pools (because 1077 * the configuration cache may be out of date). 1078 */ 1079 int 1080 zpool_open_silent(libzfs_handle_t *hdl, const char *pool, zpool_handle_t **ret) 1081 { 1082 zpool_handle_t *zhp; 1083 boolean_t missing; 1084 1085 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL) 1086 return (-1); 1087 1088 zhp->zpool_hdl = hdl; 1089 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name)); 1090 1091 if (zpool_refresh_stats(zhp, &missing) != 0) { 1092 zpool_close(zhp); 1093 return (-1); 1094 } 1095 1096 if (missing) { 1097 zpool_close(zhp); 1098 *ret = NULL; 1099 return (0); 1100 } 1101 1102 *ret = zhp; 1103 return (0); 1104 } 1105 1106 /* 1107 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted 1108 * state. 1109 */ 1110 zpool_handle_t * 1111 zpool_open(libzfs_handle_t *hdl, const char *pool) 1112 { 1113 zpool_handle_t *zhp; 1114 1115 if ((zhp = zpool_open_canfail(hdl, pool)) == NULL) 1116 return (NULL); 1117 1118 if (zhp->zpool_state == POOL_STATE_UNAVAIL) { 1119 (void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL, 1120 dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name); 1121 zpool_close(zhp); 1122 return (NULL); 1123 } 1124 1125 return (zhp); 1126 } 1127 1128 /* 1129 * Close the handle. Simply frees the memory associated with the handle. 1130 */ 1131 void 1132 zpool_close(zpool_handle_t *zhp) 1133 { 1134 nvlist_free(zhp->zpool_config); 1135 nvlist_free(zhp->zpool_old_config); 1136 nvlist_free(zhp->zpool_props); 1137 free(zhp); 1138 } 1139 1140 /* 1141 * Return the name of the pool. 1142 */ 1143 const char * 1144 zpool_get_name(zpool_handle_t *zhp) 1145 { 1146 return (zhp->zpool_name); 1147 } 1148 1149 1150 /* 1151 * Return the state of the pool (ACTIVE or UNAVAILABLE) 1152 */ 1153 int 1154 zpool_get_state(zpool_handle_t *zhp) 1155 { 1156 return (zhp->zpool_state); 1157 } 1158 1159 /* 1160 * Check if vdev list contains a special vdev 1161 */ 1162 static boolean_t 1163 zpool_has_special_vdev(nvlist_t *nvroot) 1164 { 1165 nvlist_t **child; 1166 uint_t children; 1167 1168 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, &child, 1169 &children) == 0) { 1170 for (uint_t c = 0; c < children; c++) { 1171 char *bias; 1172 1173 if (nvlist_lookup_string(child[c], 1174 ZPOOL_CONFIG_ALLOCATION_BIAS, &bias) == 0 && 1175 strcmp(bias, VDEV_ALLOC_BIAS_SPECIAL) == 0) { 1176 return (B_TRUE); 1177 } 1178 } 1179 } 1180 return (B_FALSE); 1181 } 1182 1183 /* 1184 * Create the named pool, using the provided vdev list. It is assumed 1185 * that the consumer has already validated the contents of the nvlist, so we 1186 * don't have to worry about error semantics. 1187 */ 1188 int 1189 zpool_create(libzfs_handle_t *hdl, const char *pool, nvlist_t *nvroot, 1190 nvlist_t *props, nvlist_t *fsprops) 1191 { 1192 zfs_cmd_t zc = { 0 }; 1193 nvlist_t *zc_fsprops = NULL; 1194 nvlist_t *zc_props = NULL; 1195 nvlist_t *hidden_args = NULL; 1196 uint8_t *wkeydata = NULL; 1197 uint_t wkeylen = 0; 1198 char msg[1024]; 1199 int ret = -1; 1200 1201 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1202 "cannot create '%s'"), pool); 1203 1204 if (!zpool_name_valid(hdl, B_FALSE, pool)) 1205 return (zfs_error(hdl, EZFS_INVALIDNAME, msg)); 1206 1207 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0) 1208 return (-1); 1209 1210 if (props) { 1211 prop_flags_t flags = { .create = B_TRUE, .import = B_FALSE }; 1212 1213 if ((zc_props = zpool_valid_proplist(hdl, pool, props, 1214 SPA_VERSION_1, flags, msg)) == NULL) { 1215 goto create_failed; 1216 } 1217 } 1218 1219 if (fsprops) { 1220 uint64_t zoned; 1221 char *zonestr; 1222 1223 zoned = ((nvlist_lookup_string(fsprops, 1224 zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) && 1225 strcmp(zonestr, "on") == 0); 1226 1227 if ((zc_fsprops = zfs_valid_proplist(hdl, ZFS_TYPE_FILESYSTEM, 1228 fsprops, zoned, NULL, NULL, B_TRUE, msg)) == NULL) { 1229 goto create_failed; 1230 } 1231 1232 if (nvlist_exists(zc_fsprops, 1233 zfs_prop_to_name(ZFS_PROP_SPECIAL_SMALL_BLOCKS)) && 1234 !zpool_has_special_vdev(nvroot)) { 1235 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1236 "%s property requires a special vdev"), 1237 zfs_prop_to_name(ZFS_PROP_SPECIAL_SMALL_BLOCKS)); 1238 (void) zfs_error(hdl, EZFS_BADPROP, msg); 1239 goto create_failed; 1240 } 1241 1242 if (!zc_props && 1243 (nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) { 1244 goto create_failed; 1245 } 1246 if (zfs_crypto_create(hdl, NULL, zc_fsprops, props, B_TRUE, 1247 &wkeydata, &wkeylen) != 0) { 1248 (void) zfs_error(hdl, EZFS_CRYPTOFAILED, msg); 1249 goto create_failed; 1250 } 1251 if (nvlist_add_nvlist(zc_props, 1252 ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) { 1253 goto create_failed; 1254 } 1255 if (wkeydata != NULL) { 1256 if (nvlist_alloc(&hidden_args, NV_UNIQUE_NAME, 0) != 0) 1257 goto create_failed; 1258 1259 if (nvlist_add_uint8_array(hidden_args, "wkeydata", 1260 wkeydata, wkeylen) != 0) 1261 goto create_failed; 1262 1263 if (nvlist_add_nvlist(zc_props, ZPOOL_HIDDEN_ARGS, 1264 hidden_args) != 0) 1265 goto create_failed; 1266 } 1267 } 1268 1269 if (zc_props && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0) 1270 goto create_failed; 1271 1272 (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name)); 1273 1274 if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) { 1275 1276 zcmd_free_nvlists(&zc); 1277 nvlist_free(zc_props); 1278 nvlist_free(zc_fsprops); 1279 nvlist_free(hidden_args); 1280 if (wkeydata != NULL) 1281 free(wkeydata); 1282 1283 switch (errno) { 1284 case EBUSY: 1285 /* 1286 * This can happen if the user has specified the same 1287 * device multiple times. We can't reliably detect this 1288 * until we try to add it and see we already have a 1289 * label. 1290 */ 1291 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1292 "one or more vdevs refer to the same device")); 1293 return (zfs_error(hdl, EZFS_BADDEV, msg)); 1294 1295 case ERANGE: 1296 /* 1297 * This happens if the record size is smaller or larger 1298 * than the allowed size range, or not a power of 2. 1299 * 1300 * NOTE: although zfs_valid_proplist is called earlier, 1301 * this case may have slipped through since the 1302 * pool does not exist yet and it is therefore 1303 * impossible to read properties e.g. max blocksize 1304 * from the pool. 1305 */ 1306 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1307 "record size invalid")); 1308 return (zfs_error(hdl, EZFS_BADPROP, msg)); 1309 1310 case EOVERFLOW: 1311 /* 1312 * This occurs when one of the devices is below 1313 * SPA_MINDEVSIZE. Unfortunately, we can't detect which 1314 * device was the problem device since there's no 1315 * reliable way to determine device size from userland. 1316 */ 1317 { 1318 char buf[64]; 1319 1320 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf)); 1321 1322 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1323 "one or more devices is less than the " 1324 "minimum size (%s)"), buf); 1325 } 1326 return (zfs_error(hdl, EZFS_BADDEV, msg)); 1327 1328 case ENOSPC: 1329 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1330 "one or more devices is out of space")); 1331 return (zfs_error(hdl, EZFS_BADDEV, msg)); 1332 1333 default: 1334 return (zpool_standard_error(hdl, errno, msg)); 1335 } 1336 } 1337 1338 create_failed: 1339 zcmd_free_nvlists(&zc); 1340 nvlist_free(zc_props); 1341 nvlist_free(zc_fsprops); 1342 nvlist_free(hidden_args); 1343 if (wkeydata != NULL) 1344 free(wkeydata); 1345 return (ret); 1346 } 1347 1348 /* 1349 * Destroy the given pool. It is up to the caller to ensure that there are no 1350 * datasets left in the pool. 1351 */ 1352 int 1353 zpool_destroy(zpool_handle_t *zhp, const char *log_str) 1354 { 1355 zfs_cmd_t zc = { 0 }; 1356 zfs_handle_t *zfp = NULL; 1357 libzfs_handle_t *hdl = zhp->zpool_hdl; 1358 char msg[1024]; 1359 1360 if (zhp->zpool_state == POOL_STATE_ACTIVE && 1361 (zfp = zfs_open(hdl, zhp->zpool_name, ZFS_TYPE_FILESYSTEM)) == NULL) 1362 return (-1); 1363 1364 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1365 zc.zc_history = (uint64_t)(uintptr_t)log_str; 1366 1367 if (zfs_ioctl(hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) { 1368 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1369 "cannot destroy '%s'"), zhp->zpool_name); 1370 1371 if (errno == EROFS) { 1372 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1373 "one or more devices is read only")); 1374 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1375 } else { 1376 (void) zpool_standard_error(hdl, errno, msg); 1377 } 1378 1379 if (zfp) 1380 zfs_close(zfp); 1381 return (-1); 1382 } 1383 1384 if (zfp) { 1385 remove_mountpoint(zfp); 1386 zfs_close(zfp); 1387 } 1388 1389 return (0); 1390 } 1391 1392 /* 1393 * Create a checkpoint in the given pool. 1394 */ 1395 int 1396 zpool_checkpoint(zpool_handle_t *zhp) 1397 { 1398 libzfs_handle_t *hdl = zhp->zpool_hdl; 1399 char msg[1024]; 1400 int error; 1401 1402 error = lzc_pool_checkpoint(zhp->zpool_name); 1403 if (error != 0) { 1404 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1405 "cannot checkpoint '%s'"), zhp->zpool_name); 1406 (void) zpool_standard_error(hdl, error, msg); 1407 return (-1); 1408 } 1409 1410 return (0); 1411 } 1412 1413 /* 1414 * Discard the checkpoint from the given pool. 1415 */ 1416 int 1417 zpool_discard_checkpoint(zpool_handle_t *zhp) 1418 { 1419 libzfs_handle_t *hdl = zhp->zpool_hdl; 1420 char msg[1024]; 1421 int error; 1422 1423 error = lzc_pool_checkpoint_discard(zhp->zpool_name); 1424 if (error != 0) { 1425 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1426 "cannot discard checkpoint in '%s'"), zhp->zpool_name); 1427 (void) zpool_standard_error(hdl, error, msg); 1428 return (-1); 1429 } 1430 1431 return (0); 1432 } 1433 1434 /* 1435 * Add the given vdevs to the pool. The caller must have already performed the 1436 * necessary verification to ensure that the vdev specification is well-formed. 1437 */ 1438 int 1439 zpool_add(zpool_handle_t *zhp, nvlist_t *nvroot) 1440 { 1441 zfs_cmd_t zc = { 0 }; 1442 int ret; 1443 libzfs_handle_t *hdl = zhp->zpool_hdl; 1444 char msg[1024]; 1445 nvlist_t **spares, **l2cache; 1446 uint_t nspares, nl2cache; 1447 1448 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1449 "cannot add to '%s'"), zhp->zpool_name); 1450 1451 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) < 1452 SPA_VERSION_SPARES && 1453 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, 1454 &spares, &nspares) == 0) { 1455 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be " 1456 "upgraded to add hot spares")); 1457 return (zfs_error(hdl, EZFS_BADVERSION, msg)); 1458 } 1459 1460 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) < 1461 SPA_VERSION_L2CACHE && 1462 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, 1463 &l2cache, &nl2cache) == 0) { 1464 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be " 1465 "upgraded to add cache devices")); 1466 return (zfs_error(hdl, EZFS_BADVERSION, msg)); 1467 } 1468 1469 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0) 1470 return (-1); 1471 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1472 1473 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) { 1474 switch (errno) { 1475 case EBUSY: 1476 /* 1477 * This can happen if the user has specified the same 1478 * device multiple times. We can't reliably detect this 1479 * until we try to add it and see we already have a 1480 * label. 1481 */ 1482 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1483 "one or more vdevs refer to the same device")); 1484 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1485 break; 1486 1487 case EINVAL: 1488 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1489 "invalid config; a pool with removing/removed " 1490 "vdevs does not support adding raidz vdevs")); 1491 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1492 break; 1493 1494 case EOVERFLOW: 1495 /* 1496 * This occurrs when one of the devices is below 1497 * SPA_MINDEVSIZE. Unfortunately, we can't detect which 1498 * device was the problem device since there's no 1499 * reliable way to determine device size from userland. 1500 */ 1501 { 1502 char buf[64]; 1503 1504 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf)); 1505 1506 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1507 "device is less than the minimum " 1508 "size (%s)"), buf); 1509 } 1510 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1511 break; 1512 1513 case ENOTSUP: 1514 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1515 "pool must be upgraded to add these vdevs")); 1516 (void) zfs_error(hdl, EZFS_BADVERSION, msg); 1517 break; 1518 1519 default: 1520 (void) zpool_standard_error(hdl, errno, msg); 1521 } 1522 1523 ret = -1; 1524 } else { 1525 ret = 0; 1526 } 1527 1528 zcmd_free_nvlists(&zc); 1529 1530 return (ret); 1531 } 1532 1533 /* 1534 * Exports the pool from the system. The caller must ensure that there are no 1535 * mounted datasets in the pool. 1536 */ 1537 static int 1538 zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce, 1539 const char *log_str) 1540 { 1541 zfs_cmd_t zc = { 0 }; 1542 char msg[1024]; 1543 1544 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1545 "cannot export '%s'"), zhp->zpool_name); 1546 1547 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1548 zc.zc_cookie = force; 1549 zc.zc_guid = hardforce; 1550 zc.zc_history = (uint64_t)(uintptr_t)log_str; 1551 1552 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) { 1553 switch (errno) { 1554 case EXDEV: 1555 zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN, 1556 "use '-f' to override the following errors:\n" 1557 "'%s' has an active shared spare which could be" 1558 " used by other pools once '%s' is exported."), 1559 zhp->zpool_name, zhp->zpool_name); 1560 return (zfs_error(zhp->zpool_hdl, EZFS_ACTIVE_SPARE, 1561 msg)); 1562 default: 1563 return (zpool_standard_error_fmt(zhp->zpool_hdl, errno, 1564 msg)); 1565 } 1566 } 1567 1568 return (0); 1569 } 1570 1571 int 1572 zpool_export(zpool_handle_t *zhp, boolean_t force, const char *log_str) 1573 { 1574 return (zpool_export_common(zhp, force, B_FALSE, log_str)); 1575 } 1576 1577 int 1578 zpool_export_force(zpool_handle_t *zhp, const char *log_str) 1579 { 1580 return (zpool_export_common(zhp, B_TRUE, B_TRUE, log_str)); 1581 } 1582 1583 static void 1584 zpool_rewind_exclaim(libzfs_handle_t *hdl, const char *name, boolean_t dryrun, 1585 nvlist_t *config) 1586 { 1587 nvlist_t *nv = NULL; 1588 uint64_t rewindto; 1589 int64_t loss = -1; 1590 struct tm t; 1591 char timestr[128]; 1592 1593 if (!hdl->libzfs_printerr || config == NULL) 1594 return; 1595 1596 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 || 1597 nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0) { 1598 return; 1599 } 1600 1601 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0) 1602 return; 1603 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss); 1604 1605 if (localtime_r((time_t *)&rewindto, &t) != NULL && 1606 strftime(timestr, 128, 0, &t) != 0) { 1607 if (dryrun) { 1608 (void) printf(dgettext(TEXT_DOMAIN, 1609 "Would be able to return %s " 1610 "to its state as of %s.\n"), 1611 name, timestr); 1612 } else { 1613 (void) printf(dgettext(TEXT_DOMAIN, 1614 "Pool %s returned to its state as of %s.\n"), 1615 name, timestr); 1616 } 1617 if (loss > 120) { 1618 (void) printf(dgettext(TEXT_DOMAIN, 1619 "%s approximately %lld "), 1620 dryrun ? "Would discard" : "Discarded", 1621 (loss + 30) / 60); 1622 (void) printf(dgettext(TEXT_DOMAIN, 1623 "minutes of transactions.\n")); 1624 } else if (loss > 0) { 1625 (void) printf(dgettext(TEXT_DOMAIN, 1626 "%s approximately %lld "), 1627 dryrun ? "Would discard" : "Discarded", loss); 1628 (void) printf(dgettext(TEXT_DOMAIN, 1629 "seconds of transactions.\n")); 1630 } 1631 } 1632 } 1633 1634 void 1635 zpool_explain_recover(libzfs_handle_t *hdl, const char *name, int reason, 1636 nvlist_t *config) 1637 { 1638 nvlist_t *nv = NULL; 1639 int64_t loss = -1; 1640 uint64_t edata = UINT64_MAX; 1641 uint64_t rewindto; 1642 struct tm t; 1643 char timestr[128]; 1644 1645 if (!hdl->libzfs_printerr) 1646 return; 1647 1648 if (reason >= 0) 1649 (void) printf(dgettext(TEXT_DOMAIN, "action: ")); 1650 else 1651 (void) printf(dgettext(TEXT_DOMAIN, "\t")); 1652 1653 /* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */ 1654 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 || 1655 nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0 || 1656 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0) 1657 goto no_info; 1658 1659 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss); 1660 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_DATA_ERRORS, 1661 &edata); 1662 1663 (void) printf(dgettext(TEXT_DOMAIN, 1664 "Recovery is possible, but will result in some data loss.\n")); 1665 1666 if (localtime_r((time_t *)&rewindto, &t) != NULL && 1667 strftime(timestr, 128, 0, &t) != 0) { 1668 (void) printf(dgettext(TEXT_DOMAIN, 1669 "\tReturning the pool to its state as of %s\n" 1670 "\tshould correct the problem. "), 1671 timestr); 1672 } else { 1673 (void) printf(dgettext(TEXT_DOMAIN, 1674 "\tReverting the pool to an earlier state " 1675 "should correct the problem.\n\t")); 1676 } 1677 1678 if (loss > 120) { 1679 (void) printf(dgettext(TEXT_DOMAIN, 1680 "Approximately %lld minutes of data\n" 1681 "\tmust be discarded, irreversibly. "), (loss + 30) / 60); 1682 } else if (loss > 0) { 1683 (void) printf(dgettext(TEXT_DOMAIN, 1684 "Approximately %lld seconds of data\n" 1685 "\tmust be discarded, irreversibly. "), loss); 1686 } 1687 if (edata != 0 && edata != UINT64_MAX) { 1688 if (edata == 1) { 1689 (void) printf(dgettext(TEXT_DOMAIN, 1690 "After rewind, at least\n" 1691 "\tone persistent user-data error will remain. ")); 1692 } else { 1693 (void) printf(dgettext(TEXT_DOMAIN, 1694 "After rewind, several\n" 1695 "\tpersistent user-data errors will remain. ")); 1696 } 1697 } 1698 (void) printf(dgettext(TEXT_DOMAIN, 1699 "Recovery can be attempted\n\tby executing 'zpool %s -F %s'. "), 1700 reason >= 0 ? "clear" : "import", name); 1701 1702 (void) printf(dgettext(TEXT_DOMAIN, 1703 "A scrub of the pool\n" 1704 "\tis strongly recommended after recovery.\n")); 1705 return; 1706 1707 no_info: 1708 (void) printf(dgettext(TEXT_DOMAIN, 1709 "Destroy and re-create the pool from\n\ta backup source.\n")); 1710 } 1711 1712 /* 1713 * zpool_import() is a contracted interface. Should be kept the same 1714 * if possible. 1715 * 1716 * Applications should use zpool_import_props() to import a pool with 1717 * new properties value to be set. 1718 */ 1719 int 1720 zpool_import(libzfs_handle_t *hdl, nvlist_t *config, const char *newname, 1721 char *altroot) 1722 { 1723 nvlist_t *props = NULL; 1724 int ret; 1725 1726 if (altroot != NULL) { 1727 if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) { 1728 return (zfs_error_fmt(hdl, EZFS_NOMEM, 1729 dgettext(TEXT_DOMAIN, "cannot import '%s'"), 1730 newname)); 1731 } 1732 1733 if (nvlist_add_string(props, 1734 zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 || 1735 nvlist_add_string(props, 1736 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) { 1737 nvlist_free(props); 1738 return (zfs_error_fmt(hdl, EZFS_NOMEM, 1739 dgettext(TEXT_DOMAIN, "cannot import '%s'"), 1740 newname)); 1741 } 1742 } 1743 1744 ret = zpool_import_props(hdl, config, newname, props, 1745 ZFS_IMPORT_NORMAL); 1746 nvlist_free(props); 1747 return (ret); 1748 } 1749 1750 static void 1751 print_vdev_tree(libzfs_handle_t *hdl, const char *name, nvlist_t *nv, 1752 int indent) 1753 { 1754 nvlist_t **child; 1755 uint_t c, children; 1756 char *vname; 1757 uint64_t is_log = 0; 1758 1759 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG, 1760 &is_log); 1761 1762 if (name != NULL) 1763 (void) printf("\t%*s%s%s\n", indent, "", name, 1764 is_log ? " [log]" : ""); 1765 1766 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, 1767 &child, &children) != 0) 1768 return; 1769 1770 for (c = 0; c < children; c++) { 1771 vname = zpool_vdev_name(hdl, NULL, child[c], VDEV_NAME_TYPE_ID); 1772 print_vdev_tree(hdl, vname, child[c], indent + 2); 1773 free(vname); 1774 } 1775 } 1776 1777 void 1778 zpool_print_unsup_feat(nvlist_t *config) 1779 { 1780 nvlist_t *nvinfo, *unsup_feat; 1781 1782 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 1783 0); 1784 verify(nvlist_lookup_nvlist(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT, 1785 &unsup_feat) == 0); 1786 1787 for (nvpair_t *nvp = nvlist_next_nvpair(unsup_feat, NULL); nvp != NULL; 1788 nvp = nvlist_next_nvpair(unsup_feat, nvp)) { 1789 char *desc; 1790 1791 verify(nvpair_type(nvp) == DATA_TYPE_STRING); 1792 verify(nvpair_value_string(nvp, &desc) == 0); 1793 1794 if (strlen(desc) > 0) 1795 (void) printf("\t%s (%s)\n", nvpair_name(nvp), desc); 1796 else 1797 (void) printf("\t%s\n", nvpair_name(nvp)); 1798 } 1799 } 1800 1801 /* 1802 * Import the given pool using the known configuration and a list of 1803 * properties to be set. The configuration should have come from 1804 * zpool_find_import(). The 'newname' parameters control whether the pool 1805 * is imported with a different name. 1806 */ 1807 int 1808 zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname, 1809 nvlist_t *props, int flags) 1810 { 1811 zfs_cmd_t zc = { 0 }; 1812 zpool_load_policy_t policy; 1813 nvlist_t *nv = NULL; 1814 nvlist_t *nvinfo = NULL; 1815 nvlist_t *missing = NULL; 1816 char *thename; 1817 char *origname; 1818 int ret; 1819 int error = 0; 1820 char errbuf[1024]; 1821 1822 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, 1823 &origname) == 0); 1824 1825 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, 1826 "cannot import pool '%s'"), origname); 1827 1828 if (newname != NULL) { 1829 if (!zpool_name_valid(hdl, B_FALSE, newname)) 1830 return (zfs_error_fmt(hdl, EZFS_INVALIDNAME, 1831 dgettext(TEXT_DOMAIN, "cannot import '%s'"), 1832 newname)); 1833 thename = (char *)newname; 1834 } else { 1835 thename = origname; 1836 } 1837 1838 if (props != NULL) { 1839 uint64_t version; 1840 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE }; 1841 1842 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, 1843 &version) == 0); 1844 1845 if ((props = zpool_valid_proplist(hdl, origname, 1846 props, version, flags, errbuf)) == NULL) 1847 return (-1); 1848 if (zcmd_write_src_nvlist(hdl, &zc, props) != 0) { 1849 nvlist_free(props); 1850 return (-1); 1851 } 1852 nvlist_free(props); 1853 } 1854 1855 (void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name)); 1856 1857 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, 1858 &zc.zc_guid) == 0); 1859 1860 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) { 1861 zcmd_free_nvlists(&zc); 1862 return (-1); 1863 } 1864 if (zcmd_alloc_dst_nvlist(hdl, &zc, zc.zc_nvlist_conf_size * 2) != 0) { 1865 zcmd_free_nvlists(&zc); 1866 return (-1); 1867 } 1868 1869 zc.zc_cookie = flags; 1870 while ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc)) != 0 && 1871 errno == ENOMEM) { 1872 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) { 1873 zcmd_free_nvlists(&zc); 1874 return (-1); 1875 } 1876 } 1877 if (ret != 0) 1878 error = errno; 1879 1880 (void) zcmd_read_dst_nvlist(hdl, &zc, &nv); 1881 1882 zcmd_free_nvlists(&zc); 1883 1884 zpool_get_load_policy(config, &policy); 1885 1886 if (error) { 1887 char desc[1024]; 1888 char aux[256]; 1889 1890 /* 1891 * Dry-run failed, but we print out what success 1892 * looks like if we found a best txg 1893 */ 1894 if (policy.zlp_rewind & ZPOOL_TRY_REWIND) { 1895 zpool_rewind_exclaim(hdl, newname ? origname : thename, 1896 B_TRUE, nv); 1897 nvlist_free(nv); 1898 return (-1); 1899 } 1900 1901 if (newname == NULL) 1902 (void) snprintf(desc, sizeof (desc), 1903 dgettext(TEXT_DOMAIN, "cannot import '%s'"), 1904 thename); 1905 else 1906 (void) snprintf(desc, sizeof (desc), 1907 dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"), 1908 origname, thename); 1909 1910 switch (error) { 1911 case ENOTSUP: 1912 if (nv != NULL && nvlist_lookup_nvlist(nv, 1913 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 && 1914 nvlist_exists(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT)) { 1915 (void) printf(dgettext(TEXT_DOMAIN, "This " 1916 "pool uses the following feature(s) not " 1917 "supported by this system:\n")); 1918 zpool_print_unsup_feat(nv); 1919 if (nvlist_exists(nvinfo, 1920 ZPOOL_CONFIG_CAN_RDONLY)) { 1921 (void) printf(dgettext(TEXT_DOMAIN, 1922 "All unsupported features are only " 1923 "required for writing to the pool." 1924 "\nThe pool can be imported using " 1925 "'-o readonly=on'.\n")); 1926 } 1927 } 1928 /* 1929 * Unsupported version. 1930 */ 1931 (void) zfs_error(hdl, EZFS_BADVERSION, desc); 1932 break; 1933 1934 case EREMOTEIO: 1935 if (nv != NULL && nvlist_lookup_nvlist(nv, 1936 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0) { 1937 char *hostname = "<unknown>"; 1938 uint64_t hostid = 0; 1939 mmp_state_t mmp_state; 1940 1941 mmp_state = fnvlist_lookup_uint64(nvinfo, 1942 ZPOOL_CONFIG_MMP_STATE); 1943 1944 if (nvlist_exists(nvinfo, 1945 ZPOOL_CONFIG_MMP_HOSTNAME)) 1946 hostname = fnvlist_lookup_string(nvinfo, 1947 ZPOOL_CONFIG_MMP_HOSTNAME); 1948 1949 if (nvlist_exists(nvinfo, 1950 ZPOOL_CONFIG_MMP_HOSTID)) 1951 hostid = fnvlist_lookup_uint64(nvinfo, 1952 ZPOOL_CONFIG_MMP_HOSTID); 1953 1954 if (mmp_state == MMP_STATE_ACTIVE) { 1955 (void) snprintf(aux, sizeof (aux), 1956 dgettext(TEXT_DOMAIN, "pool is imp" 1957 "orted on host '%s' (hostid=%lx).\n" 1958 "Export the pool on the other " 1959 "system, then run 'zpool import'."), 1960 hostname, (unsigned long) hostid); 1961 } else if (mmp_state == MMP_STATE_NO_HOSTID) { 1962 (void) snprintf(aux, sizeof (aux), 1963 dgettext(TEXT_DOMAIN, "pool has " 1964 "the multihost property on and " 1965 "the\nsystem's hostid is not " 1966 "set.\n")); 1967 } 1968 1969 (void) zfs_error_aux(hdl, aux); 1970 } 1971 (void) zfs_error(hdl, EZFS_ACTIVE_POOL, desc); 1972 break; 1973 1974 case EINVAL: 1975 (void) zfs_error(hdl, EZFS_INVALCONFIG, desc); 1976 break; 1977 1978 case EROFS: 1979 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1980 "one or more devices is read only")); 1981 (void) zfs_error(hdl, EZFS_BADDEV, desc); 1982 break; 1983 1984 case ENXIO: 1985 if (nv && nvlist_lookup_nvlist(nv, 1986 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 && 1987 nvlist_lookup_nvlist(nvinfo, 1988 ZPOOL_CONFIG_MISSING_DEVICES, &missing) == 0) { 1989 (void) printf(dgettext(TEXT_DOMAIN, 1990 "The devices below are missing or " 1991 "corrupted, use '-m' to import the pool " 1992 "anyway:\n")); 1993 print_vdev_tree(hdl, NULL, missing, 2); 1994 (void) printf("\n"); 1995 } 1996 (void) zpool_standard_error(hdl, error, desc); 1997 break; 1998 1999 case EEXIST: 2000 (void) zpool_standard_error(hdl, error, desc); 2001 break; 2002 case ENAMETOOLONG: 2003 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2004 "new name of at least one dataset is longer than " 2005 "the maximum allowable length")); 2006 (void) zfs_error(hdl, EZFS_NAMETOOLONG, desc); 2007 break; 2008 default: 2009 (void) zpool_standard_error(hdl, error, desc); 2010 zpool_explain_recover(hdl, 2011 newname ? origname : thename, -error, nv); 2012 break; 2013 } 2014 2015 nvlist_free(nv); 2016 ret = -1; 2017 } else { 2018 zpool_handle_t *zhp; 2019 2020 /* 2021 * This should never fail, but play it safe anyway. 2022 */ 2023 if (zpool_open_silent(hdl, thename, &zhp) != 0) 2024 ret = -1; 2025 else if (zhp != NULL) 2026 zpool_close(zhp); 2027 if (policy.zlp_rewind & 2028 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) { 2029 zpool_rewind_exclaim(hdl, newname ? origname : thename, 2030 ((policy.zlp_rewind & ZPOOL_TRY_REWIND) != 0), nv); 2031 } 2032 nvlist_free(nv); 2033 return (0); 2034 } 2035 2036 return (ret); 2037 } 2038 2039 /* 2040 * Translate vdev names to guids. If a vdev_path is determined to be 2041 * unsuitable then a vd_errlist is allocated and the vdev path and errno 2042 * are added to it. 2043 */ 2044 static int 2045 zpool_translate_vdev_guids(zpool_handle_t *zhp, nvlist_t *vds, 2046 nvlist_t *vdev_guids, nvlist_t *guids_to_paths, nvlist_t **vd_errlist) 2047 { 2048 nvlist_t *errlist = NULL; 2049 int error = 0; 2050 2051 for (nvpair_t *elem = nvlist_next_nvpair(vds, NULL); elem != NULL; 2052 elem = nvlist_next_nvpair(vds, elem)) { 2053 boolean_t spare, cache; 2054 2055 char *vd_path = nvpair_name(elem); 2056 nvlist_t *tgt = zpool_find_vdev(zhp, vd_path, &spare, &cache, 2057 NULL); 2058 2059 if ((tgt == NULL) || cache || spare) { 2060 if (errlist == NULL) { 2061 errlist = fnvlist_alloc(); 2062 error = EINVAL; 2063 } 2064 2065 uint64_t err = (tgt == NULL) ? EZFS_NODEVICE : 2066 (spare ? EZFS_ISSPARE : EZFS_ISL2CACHE); 2067 fnvlist_add_int64(errlist, vd_path, err); 2068 continue; 2069 } 2070 2071 uint64_t guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID); 2072 fnvlist_add_uint64(vdev_guids, vd_path, guid); 2073 2074 char msg[MAXNAMELEN]; 2075 (void) snprintf(msg, sizeof (msg), "%llu", (u_longlong_t)guid); 2076 fnvlist_add_string(guids_to_paths, msg, vd_path); 2077 } 2078 2079 if (error != 0) { 2080 verify(errlist != NULL); 2081 if (vd_errlist != NULL) 2082 *vd_errlist = errlist; 2083 else 2084 fnvlist_free(errlist); 2085 } 2086 2087 return (error); 2088 } 2089 2090 /* 2091 * Scan the pool. 2092 */ 2093 int 2094 zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func, pool_scrub_cmd_t cmd) 2095 { 2096 zfs_cmd_t zc = { 0 }; 2097 char msg[1024]; 2098 int err; 2099 libzfs_handle_t *hdl = zhp->zpool_hdl; 2100 2101 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2102 zc.zc_cookie = func; 2103 zc.zc_flags = cmd; 2104 2105 if (zfs_ioctl(hdl, ZFS_IOC_POOL_SCAN, &zc) == 0) 2106 return (0); 2107 2108 err = errno; 2109 2110 /* ECANCELED on a scrub means we resumed a paused scrub */ 2111 if (err == ECANCELED && func == POOL_SCAN_SCRUB && 2112 cmd == POOL_SCRUB_NORMAL) 2113 return (0); 2114 2115 if (err == ENOENT && func != POOL_SCAN_NONE && cmd == POOL_SCRUB_NORMAL) 2116 return (0); 2117 2118 if (func == POOL_SCAN_SCRUB) { 2119 if (cmd == POOL_SCRUB_PAUSE) { 2120 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 2121 "cannot pause scrubbing %s"), zc.zc_name); 2122 } else { 2123 assert(cmd == POOL_SCRUB_NORMAL); 2124 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 2125 "cannot scrub %s"), zc.zc_name); 2126 } 2127 } else if (func == POOL_SCAN_RESILVER) { 2128 assert(cmd == POOL_SCRUB_NORMAL); 2129 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 2130 "cannot restart resilver on %s"), zc.zc_name); 2131 } else if (func == POOL_SCAN_NONE) { 2132 (void) snprintf(msg, sizeof (msg), 2133 dgettext(TEXT_DOMAIN, "cannot cancel scrubbing %s"), 2134 zc.zc_name); 2135 } else { 2136 assert(!"unexpected result"); 2137 } 2138 2139 if (err == EBUSY) { 2140 nvlist_t *nvroot; 2141 pool_scan_stat_t *ps = NULL; 2142 uint_t psc; 2143 2144 verify(nvlist_lookup_nvlist(zhp->zpool_config, 2145 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 2146 (void) nvlist_lookup_uint64_array(nvroot, 2147 ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &psc); 2148 if (ps && ps->pss_func == POOL_SCAN_SCRUB) { 2149 if (cmd == POOL_SCRUB_PAUSE) 2150 return (zfs_error(hdl, EZFS_SCRUB_PAUSED, msg)); 2151 else 2152 return (zfs_error(hdl, EZFS_SCRUBBING, msg)); 2153 } else { 2154 return (zfs_error(hdl, EZFS_RESILVERING, msg)); 2155 } 2156 } else if (err == ENOENT) { 2157 return (zfs_error(hdl, EZFS_NO_SCRUB, msg)); 2158 } else if (err == ENOTSUP && func == POOL_SCAN_RESILVER) { 2159 return (zfs_error(hdl, EZFS_NO_RESILVER_DEFER, msg)); 2160 } else { 2161 return (zpool_standard_error(hdl, err, msg)); 2162 } 2163 } 2164 2165 static int 2166 xlate_init_err(int err) 2167 { 2168 switch (err) { 2169 case ENODEV: 2170 return (EZFS_NODEVICE); 2171 case EINVAL: 2172 case EROFS: 2173 return (EZFS_BADDEV); 2174 case EBUSY: 2175 return (EZFS_INITIALIZING); 2176 case ESRCH: 2177 return (EZFS_NO_INITIALIZE); 2178 } 2179 return (err); 2180 } 2181 2182 /* 2183 * Begin, suspend, or cancel the initialization (initializing of all free 2184 * blocks) for the given vdevs in the given pool. 2185 */ 2186 int 2187 zpool_initialize(zpool_handle_t *zhp, pool_initialize_func_t cmd_type, 2188 nvlist_t *vds) 2189 { 2190 char msg[1024]; 2191 int err; 2192 2193 nvlist_t *vdev_guids = fnvlist_alloc(); 2194 nvlist_t *guids_to_paths = fnvlist_alloc(); 2195 nvlist_t *vd_errlist = NULL; 2196 nvlist_t *errlist; 2197 nvpair_t *elem; 2198 2199 err = zpool_translate_vdev_guids(zhp, vds, vdev_guids, 2200 guids_to_paths, &vd_errlist); 2201 2202 if (err == 0) { 2203 err = lzc_initialize(zhp->zpool_name, cmd_type, 2204 vdev_guids, &errlist); 2205 if (err == 0) { 2206 fnvlist_free(vdev_guids); 2207 fnvlist_free(guids_to_paths); 2208 return (0); 2209 } 2210 2211 if (errlist != NULL) { 2212 vd_errlist = fnvlist_lookup_nvlist(errlist, 2213 ZPOOL_INITIALIZE_VDEVS); 2214 } 2215 2216 (void) snprintf(msg, sizeof (msg), 2217 dgettext(TEXT_DOMAIN, "operation failed")); 2218 } else { 2219 verify(vd_errlist != NULL); 2220 } 2221 2222 for (elem = nvlist_next_nvpair(vd_errlist, NULL); elem != NULL; 2223 elem = nvlist_next_nvpair(vd_errlist, elem)) { 2224 int64_t vd_error = xlate_init_err(fnvpair_value_int64(elem)); 2225 char *path; 2226 2227 if (nvlist_lookup_string(guids_to_paths, nvpair_name(elem), 2228 &path) != 0) 2229 path = nvpair_name(elem); 2230 2231 (void) zfs_error_fmt(zhp->zpool_hdl, vd_error, 2232 "cannot initialize '%s'", path); 2233 } 2234 2235 fnvlist_free(vdev_guids); 2236 fnvlist_free(guids_to_paths); 2237 2238 if (vd_errlist != NULL) { 2239 fnvlist_free(vd_errlist); 2240 return (-1); 2241 } 2242 2243 return (zpool_standard_error(zhp->zpool_hdl, err, msg)); 2244 } 2245 2246 static int 2247 xlate_trim_err(int err) 2248 { 2249 switch (err) { 2250 case ENODEV: 2251 return (EZFS_NODEVICE); 2252 case EINVAL: 2253 case EROFS: 2254 return (EZFS_BADDEV); 2255 case EBUSY: 2256 return (EZFS_TRIMMING); 2257 case ESRCH: 2258 return (EZFS_NO_TRIM); 2259 case EOPNOTSUPP: 2260 return (EZFS_TRIM_NOTSUP); 2261 } 2262 return (err); 2263 } 2264 2265 /* 2266 * Begin, suspend, or cancel the TRIM (discarding of all free blocks) for 2267 * the given vdevs in the given pool. 2268 */ 2269 int 2270 zpool_trim(zpool_handle_t *zhp, pool_trim_func_t cmd_type, nvlist_t *vds, 2271 trimflags_t *trim_flags) 2272 { 2273 char msg[1024]; 2274 int err; 2275 2276 nvlist_t *vdev_guids = fnvlist_alloc(); 2277 nvlist_t *guids_to_paths = fnvlist_alloc(); 2278 nvlist_t *vd_errlist = NULL; 2279 nvlist_t *errlist; 2280 nvpair_t *elem; 2281 2282 err = zpool_translate_vdev_guids(zhp, vds, vdev_guids, 2283 guids_to_paths, &vd_errlist); 2284 if (err == 0) { 2285 err = lzc_trim(zhp->zpool_name, cmd_type, trim_flags->rate, 2286 trim_flags->secure, vdev_guids, &errlist); 2287 if (err == 0) { 2288 fnvlist_free(vdev_guids); 2289 fnvlist_free(guids_to_paths); 2290 return (0); 2291 } 2292 2293 if (errlist != NULL) { 2294 vd_errlist = fnvlist_lookup_nvlist(errlist, 2295 ZPOOL_TRIM_VDEVS); 2296 } 2297 2298 (void) snprintf(msg, sizeof (msg), 2299 dgettext(TEXT_DOMAIN, "operation failed")); 2300 } else { 2301 verify(vd_errlist != NULL); 2302 } 2303 2304 for (elem = nvlist_next_nvpair(vd_errlist, NULL); 2305 elem != NULL; elem = nvlist_next_nvpair(vd_errlist, elem)) { 2306 int64_t vd_error = xlate_trim_err(fnvpair_value_int64(elem)); 2307 char *path; 2308 /* 2309 * If only the pool was specified, and it was not a secure 2310 * trim then suppress warnings for individual vdevs which 2311 * do not support trimming. 2312 */ 2313 if (vd_error == EZFS_TRIM_NOTSUP && 2314 trim_flags->fullpool && 2315 !trim_flags->secure) { 2316 continue; 2317 } 2318 2319 if (nvlist_lookup_string(guids_to_paths, nvpair_name(elem), 2320 &path) != 0) 2321 path = nvpair_name(elem); 2322 2323 (void) zfs_error_fmt(zhp->zpool_hdl, vd_error, 2324 "cannot trim '%s'", path); 2325 } 2326 2327 fnvlist_free(vdev_guids); 2328 fnvlist_free(guids_to_paths); 2329 2330 if (vd_errlist != NULL) { 2331 fnvlist_free(vd_errlist); 2332 return (-1); 2333 } 2334 2335 return (zpool_standard_error(zhp->zpool_hdl, err, msg)); 2336 } 2337 2338 /* 2339 * This provides a very minimal check whether a given string is likely a 2340 * c#t#d# style string. Users of this are expected to do their own 2341 * verification of the s# part. 2342 */ 2343 #define CTD_CHECK(str) (str && str[0] == 'c' && isdigit(str[1])) 2344 2345 /* 2346 * More elaborate version for ones which may start with "/dev/dsk/" 2347 * and the like. 2348 */ 2349 static int 2350 ctd_check_path(char *str) 2351 { 2352 /* 2353 * If it starts with a slash, check the last component. 2354 */ 2355 if (str && str[0] == '/') { 2356 char *tmp = strrchr(str, '/'); 2357 2358 /* 2359 * If it ends in "/old", check the second-to-last 2360 * component of the string instead. 2361 */ 2362 if (tmp != str && strcmp(tmp, "/old") == 0) { 2363 for (tmp--; *tmp != '/'; tmp--) 2364 ; 2365 } 2366 str = tmp + 1; 2367 } 2368 return (CTD_CHECK(str)); 2369 } 2370 2371 /* 2372 * Find a vdev that matches the search criteria specified. We use the 2373 * the nvpair name to determine how we should look for the device. 2374 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL 2375 * spare; but FALSE if its an INUSE spare. 2376 */ 2377 static nvlist_t * 2378 vdev_to_nvlist_iter(nvlist_t *nv, nvlist_t *search, boolean_t *avail_spare, 2379 boolean_t *l2cache, boolean_t *log) 2380 { 2381 uint_t c, children; 2382 nvlist_t **child; 2383 nvlist_t *ret; 2384 uint64_t is_log; 2385 char *srchkey; 2386 nvpair_t *pair = nvlist_next_nvpair(search, NULL); 2387 2388 /* Nothing to look for */ 2389 if (search == NULL || pair == NULL) 2390 return (NULL); 2391 2392 /* Obtain the key we will use to search */ 2393 srchkey = nvpair_name(pair); 2394 2395 switch (nvpair_type(pair)) { 2396 case DATA_TYPE_UINT64: 2397 if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) { 2398 uint64_t srchval, theguid; 2399 2400 verify(nvpair_value_uint64(pair, &srchval) == 0); 2401 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, 2402 &theguid) == 0); 2403 if (theguid == srchval) 2404 return (nv); 2405 } 2406 break; 2407 2408 case DATA_TYPE_STRING: { 2409 char *srchval, *val; 2410 2411 verify(nvpair_value_string(pair, &srchval) == 0); 2412 if (nvlist_lookup_string(nv, srchkey, &val) != 0) 2413 break; 2414 2415 /* 2416 * Search for the requested value. Special cases: 2417 * 2418 * - ZPOOL_CONFIG_PATH for whole disk entries. To support 2419 * UEFI boot, these end in "s0" or "s0/old" or "s1" or 2420 * "s1/old". The "s0" or "s1" part is hidden from the user, 2421 * but included in the string, so this matches around it. 2422 * - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE). 2423 * 2424 * Otherwise, all other searches are simple string compares. 2425 */ 2426 if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0 && 2427 ctd_check_path(val)) { 2428 uint64_t wholedisk = 0; 2429 2430 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, 2431 &wholedisk); 2432 if (wholedisk) { 2433 int slen = strlen(srchval); 2434 int vlen = strlen(val); 2435 2436 if (slen != vlen - 2) 2437 break; 2438 2439 /* 2440 * make_leaf_vdev() should only set 2441 * wholedisk for ZPOOL_CONFIG_PATHs which 2442 * will include "/dev/dsk/", giving plenty of 2443 * room for the indices used next. 2444 */ 2445 ASSERT(vlen >= 6); 2446 2447 /* 2448 * strings identical except trailing "s0" 2449 */ 2450 if ((strcmp(&val[vlen - 2], "s0") == 0 || 2451 strcmp(&val[vlen - 2], "s1") == 0) && 2452 strncmp(srchval, val, slen) == 0) 2453 return (nv); 2454 2455 /* 2456 * strings identical except trailing "s0/old" 2457 */ 2458 if ((strcmp(&val[vlen - 6], "s0/old") == 0 || 2459 strcmp(&val[vlen - 6], "s1/old") == 0) && 2460 strcmp(&srchval[slen - 4], "/old") == 0 && 2461 strncmp(srchval, val, slen - 4) == 0) 2462 return (nv); 2463 2464 break; 2465 } 2466 } else if (strcmp(srchkey, ZPOOL_CONFIG_TYPE) == 0 && val) { 2467 char *type, *idx, *end, *p; 2468 uint64_t id, vdev_id; 2469 2470 /* 2471 * Determine our vdev type, keeping in mind 2472 * that the srchval is composed of a type and 2473 * vdev id pair (i.e. mirror-4). 2474 */ 2475 if ((type = strdup(srchval)) == NULL) 2476 return (NULL); 2477 2478 if ((p = strrchr(type, '-')) == NULL) { 2479 free(type); 2480 break; 2481 } 2482 idx = p + 1; 2483 *p = '\0'; 2484 2485 /* 2486 * If the types don't match then keep looking. 2487 */ 2488 if (strncmp(val, type, strlen(val)) != 0) { 2489 free(type); 2490 break; 2491 } 2492 2493 verify(zpool_vdev_is_interior(type)); 2494 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID, 2495 &id) == 0); 2496 2497 errno = 0; 2498 vdev_id = strtoull(idx, &end, 10); 2499 2500 free(type); 2501 if (errno != 0) 2502 return (NULL); 2503 2504 /* 2505 * Now verify that we have the correct vdev id. 2506 */ 2507 if (vdev_id == id) 2508 return (nv); 2509 } 2510 2511 /* 2512 * Common case 2513 */ 2514 if (strcmp(srchval, val) == 0) 2515 return (nv); 2516 break; 2517 } 2518 2519 default: 2520 break; 2521 } 2522 2523 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, 2524 &child, &children) != 0) 2525 return (NULL); 2526 2527 for (c = 0; c < children; c++) { 2528 if ((ret = vdev_to_nvlist_iter(child[c], search, 2529 avail_spare, l2cache, NULL)) != NULL) { 2530 /* 2531 * The 'is_log' value is only set for the toplevel 2532 * vdev, not the leaf vdevs. So we always lookup the 2533 * log device from the root of the vdev tree (where 2534 * 'log' is non-NULL). 2535 */ 2536 if (log != NULL && 2537 nvlist_lookup_uint64(child[c], 2538 ZPOOL_CONFIG_IS_LOG, &is_log) == 0 && 2539 is_log) { 2540 *log = B_TRUE; 2541 } 2542 return (ret); 2543 } 2544 } 2545 2546 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES, 2547 &child, &children) == 0) { 2548 for (c = 0; c < children; c++) { 2549 if ((ret = vdev_to_nvlist_iter(child[c], search, 2550 avail_spare, l2cache, NULL)) != NULL) { 2551 *avail_spare = B_TRUE; 2552 return (ret); 2553 } 2554 } 2555 } 2556 2557 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE, 2558 &child, &children) == 0) { 2559 for (c = 0; c < children; c++) { 2560 if ((ret = vdev_to_nvlist_iter(child[c], search, 2561 avail_spare, l2cache, NULL)) != NULL) { 2562 *l2cache = B_TRUE; 2563 return (ret); 2564 } 2565 } 2566 } 2567 2568 return (NULL); 2569 } 2570 2571 /* 2572 * Given a physical path (minus the "/devices" prefix), find the 2573 * associated vdev. 2574 */ 2575 nvlist_t * 2576 zpool_find_vdev_by_physpath(zpool_handle_t *zhp, const char *ppath, 2577 boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log) 2578 { 2579 nvlist_t *search, *nvroot, *ret; 2580 2581 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2582 verify(nvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH, ppath) == 0); 2583 2584 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE, 2585 &nvroot) == 0); 2586 2587 *avail_spare = B_FALSE; 2588 *l2cache = B_FALSE; 2589 if (log != NULL) 2590 *log = B_FALSE; 2591 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log); 2592 nvlist_free(search); 2593 2594 return (ret); 2595 } 2596 2597 /* 2598 * Determine if we have an "interior" top-level vdev (i.e mirror/raidz). 2599 */ 2600 static boolean_t 2601 zpool_vdev_is_interior(const char *name) 2602 { 2603 if (strncmp(name, VDEV_TYPE_RAIDZ, strlen(VDEV_TYPE_RAIDZ)) == 0 || 2604 strncmp(name, VDEV_TYPE_SPARE, strlen(VDEV_TYPE_SPARE)) == 0 || 2605 strncmp(name, 2606 VDEV_TYPE_REPLACING, strlen(VDEV_TYPE_REPLACING)) == 0 || 2607 strncmp(name, VDEV_TYPE_MIRROR, strlen(VDEV_TYPE_MIRROR)) == 0) 2608 return (B_TRUE); 2609 return (B_FALSE); 2610 } 2611 2612 nvlist_t * 2613 zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare, 2614 boolean_t *l2cache, boolean_t *log) 2615 { 2616 char buf[MAXPATHLEN]; 2617 char *end; 2618 nvlist_t *nvroot, *search, *ret; 2619 uint64_t guid; 2620 2621 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2622 2623 guid = strtoull(path, &end, 10); 2624 if (guid != 0 && *end == '\0') { 2625 verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0); 2626 } else if (zpool_vdev_is_interior(path)) { 2627 verify(nvlist_add_string(search, ZPOOL_CONFIG_TYPE, path) == 0); 2628 } else if (path[0] != '/') { 2629 (void) snprintf(buf, sizeof (buf), "%s/%s", ZFS_DISK_ROOT, 2630 path); 2631 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, buf) == 0); 2632 } else { 2633 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, path) == 0); 2634 } 2635 2636 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE, 2637 &nvroot) == 0); 2638 2639 *avail_spare = B_FALSE; 2640 *l2cache = B_FALSE; 2641 if (log != NULL) 2642 *log = B_FALSE; 2643 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log); 2644 nvlist_free(search); 2645 2646 return (ret); 2647 } 2648 2649 static int 2650 vdev_is_online(nvlist_t *nv) 2651 { 2652 uint64_t ival; 2653 2654 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE, &ival) == 0 || 2655 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED, &ival) == 0 || 2656 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED, &ival) == 0) 2657 return (0); 2658 2659 return (1); 2660 } 2661 2662 /* 2663 * Helper function for zpool_get_physpaths(). 2664 */ 2665 static int 2666 vdev_get_one_physpath(nvlist_t *config, char *physpath, size_t physpath_size, 2667 size_t *bytes_written) 2668 { 2669 size_t bytes_left, pos, rsz; 2670 char *tmppath; 2671 const char *format; 2672 2673 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PHYS_PATH, 2674 &tmppath) != 0) 2675 return (EZFS_NODEVICE); 2676 2677 pos = *bytes_written; 2678 bytes_left = physpath_size - pos; 2679 format = (pos == 0) ? "%s" : " %s"; 2680 2681 rsz = snprintf(physpath + pos, bytes_left, format, tmppath); 2682 *bytes_written += rsz; 2683 2684 if (rsz >= bytes_left) { 2685 /* if physpath was not copied properly, clear it */ 2686 if (bytes_left != 0) { 2687 physpath[pos] = 0; 2688 } 2689 return (EZFS_NOSPC); 2690 } 2691 return (0); 2692 } 2693 2694 static int 2695 vdev_get_physpaths(nvlist_t *nv, char *physpath, size_t phypath_size, 2696 size_t *rsz, boolean_t is_spare) 2697 { 2698 char *type; 2699 int ret; 2700 2701 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0) 2702 return (EZFS_INVALCONFIG); 2703 2704 if (strcmp(type, VDEV_TYPE_DISK) == 0) { 2705 /* 2706 * An active spare device has ZPOOL_CONFIG_IS_SPARE set. 2707 * For a spare vdev, we only want to boot from the active 2708 * spare device. 2709 */ 2710 if (is_spare) { 2711 uint64_t spare = 0; 2712 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE, 2713 &spare); 2714 if (!spare) 2715 return (EZFS_INVALCONFIG); 2716 } 2717 2718 if (vdev_is_online(nv)) { 2719 if ((ret = vdev_get_one_physpath(nv, physpath, 2720 phypath_size, rsz)) != 0) 2721 return (ret); 2722 } 2723 } else if (strcmp(type, VDEV_TYPE_MIRROR) == 0 || 2724 strcmp(type, VDEV_TYPE_RAIDZ) == 0 || 2725 strcmp(type, VDEV_TYPE_REPLACING) == 0 || 2726 (is_spare = (strcmp(type, VDEV_TYPE_SPARE) == 0))) { 2727 nvlist_t **child; 2728 uint_t count; 2729 int i, ret; 2730 2731 if (nvlist_lookup_nvlist_array(nv, 2732 ZPOOL_CONFIG_CHILDREN, &child, &count) != 0) 2733 return (EZFS_INVALCONFIG); 2734 2735 for (i = 0; i < count; i++) { 2736 ret = vdev_get_physpaths(child[i], physpath, 2737 phypath_size, rsz, is_spare); 2738 if (ret == EZFS_NOSPC) 2739 return (ret); 2740 } 2741 } 2742 2743 return (EZFS_POOL_INVALARG); 2744 } 2745 2746 /* 2747 * Get phys_path for a root pool config. 2748 * Return 0 on success; non-zero on failure. 2749 */ 2750 static int 2751 zpool_get_config_physpath(nvlist_t *config, char *physpath, size_t phypath_size) 2752 { 2753 size_t rsz; 2754 nvlist_t *vdev_root; 2755 nvlist_t **child; 2756 uint_t count; 2757 char *type; 2758 2759 rsz = 0; 2760 2761 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 2762 &vdev_root) != 0) 2763 return (EZFS_INVALCONFIG); 2764 2765 if (nvlist_lookup_string(vdev_root, ZPOOL_CONFIG_TYPE, &type) != 0 || 2766 nvlist_lookup_nvlist_array(vdev_root, ZPOOL_CONFIG_CHILDREN, 2767 &child, &count) != 0) 2768 return (EZFS_INVALCONFIG); 2769 2770 /* 2771 * root pool can only have a single top-level vdev. 2772 */ 2773 if (strcmp(type, VDEV_TYPE_ROOT) != 0 || count != 1) 2774 return (EZFS_POOL_INVALARG); 2775 2776 (void) vdev_get_physpaths(child[0], physpath, phypath_size, &rsz, 2777 B_FALSE); 2778 2779 /* No online devices */ 2780 if (rsz == 0) 2781 return (EZFS_NODEVICE); 2782 2783 return (0); 2784 } 2785 2786 /* 2787 * Get phys_path for a root pool 2788 * Return 0 on success; non-zero on failure. 2789 */ 2790 int 2791 zpool_get_physpath(zpool_handle_t *zhp, char *physpath, size_t phypath_size) 2792 { 2793 return (zpool_get_config_physpath(zhp->zpool_config, physpath, 2794 phypath_size)); 2795 } 2796 2797 /* 2798 * If the device has being dynamically expanded then we need to relabel 2799 * the disk to use the new unallocated space. 2800 */ 2801 static int 2802 zpool_relabel_disk(libzfs_handle_t *hdl, const char *name, const char *msg) 2803 { 2804 char path[MAXPATHLEN]; 2805 int fd, error; 2806 int (*_efi_use_whole_disk)(int); 2807 char drv[MODMAXNAMELEN]; 2808 major_t maj; 2809 struct stat st; 2810 2811 if ((_efi_use_whole_disk = (int (*)(int))dlsym(RTLD_DEFAULT, 2812 "efi_use_whole_disk")) == NULL) 2813 return (-1); 2814 2815 (void) snprintf(path, sizeof (path), "%s/%s", ZFS_RDISK_ROOT, name); 2816 2817 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) { 2818 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot " 2819 "relabel '%s': unable to open device"), name); 2820 return (zfs_error(hdl, EZFS_OPENFAILED, msg)); 2821 } 2822 2823 /* 2824 * It's possible that we might encounter an error if the device 2825 * does not have any unallocated space left. If so, we simply 2826 * ignore that error and continue on. 2827 */ 2828 error = _efi_use_whole_disk(fd); 2829 if (error && error != VT_ENOSPC) { 2830 (void) close(fd); 2831 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot " 2832 "relabel '%s': unable to read disk capacity"), name); 2833 return (zfs_error(hdl, EZFS_NOCAP, msg)); 2834 } 2835 2836 /* 2837 * Writing a new EFI partition table to the disk will have marked 2838 * the geometry as needing re-validation. Before returning, force 2839 * it to be checked by querying the device state, otherwise the 2840 * subsequent vdev_reopen() will very likely fail to read the device 2841 * size, faulting the pool. 2842 * 2843 * The dkio(4I) ioctls are implemented by the disk driver rather than 2844 * some generic framework, so we limit its use here to drivers with 2845 * which it has been tested. 2846 */ 2847 if (fstat(fd, &st) == 0 && 2848 (maj = major(st.st_rdev)) != (major_t)NODEV && 2849 modctl(MODGETNAME, drv, sizeof (drv), &maj) == 0 && 2850 (strcmp(drv, "blkdev") == 0 || strcmp(drv, "sd") == 0)) { 2851 enum dkio_state dkst = DKIO_NONE; 2852 (void) ioctl(fd, DKIOCSTATE, &dkst); 2853 } 2854 2855 (void) close(fd); 2856 2857 return (0); 2858 } 2859 2860 /* 2861 * Bring the specified vdev online. The 'flags' parameter is a set of the 2862 * ZFS_ONLINE_* flags. 2863 */ 2864 int 2865 zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags, 2866 vdev_state_t *newstate) 2867 { 2868 zfs_cmd_t zc = { 0 }; 2869 char msg[1024]; 2870 char *pathname; 2871 nvlist_t *tgt; 2872 boolean_t avail_spare, l2cache, islog; 2873 libzfs_handle_t *hdl = zhp->zpool_hdl; 2874 int error; 2875 2876 if (flags & ZFS_ONLINE_EXPAND) { 2877 (void) snprintf(msg, sizeof (msg), 2878 dgettext(TEXT_DOMAIN, "cannot expand %s"), path); 2879 } else { 2880 (void) snprintf(msg, sizeof (msg), 2881 dgettext(TEXT_DOMAIN, "cannot online %s"), path); 2882 } 2883 2884 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2885 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 2886 &islog)) == NULL) 2887 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 2888 2889 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 2890 2891 if (avail_spare) 2892 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 2893 2894 if ((flags & ZFS_ONLINE_EXPAND || 2895 zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) && 2896 nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH, &pathname) == 0) { 2897 uint64_t wholedisk = 0; 2898 2899 (void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK, 2900 &wholedisk); 2901 2902 /* 2903 * XXX - L2ARC 1.0 devices can't support expansion. 2904 */ 2905 if (l2cache) { 2906 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2907 "cannot expand cache devices")); 2908 return (zfs_error(hdl, EZFS_VDEVNOTSUP, msg)); 2909 } 2910 2911 if (wholedisk) { 2912 pathname += strlen(ZFS_DISK_ROOT) + 1; 2913 error = zpool_relabel_disk(hdl, pathname, msg); 2914 if (error != 0) 2915 return (error); 2916 } 2917 } 2918 2919 zc.zc_cookie = VDEV_STATE_ONLINE; 2920 zc.zc_obj = flags; 2921 2922 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0) { 2923 if (errno == EINVAL) { 2924 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "was split " 2925 "from this pool into a new one. Use '%s' " 2926 "instead"), "zpool detach"); 2927 return (zfs_error(hdl, EZFS_POSTSPLIT_ONLINE, msg)); 2928 } 2929 return (zpool_standard_error(hdl, errno, msg)); 2930 } 2931 2932 *newstate = zc.zc_cookie; 2933 return (0); 2934 } 2935 2936 /* 2937 * Take the specified vdev offline 2938 */ 2939 int 2940 zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp) 2941 { 2942 zfs_cmd_t zc = { 0 }; 2943 char msg[1024]; 2944 nvlist_t *tgt; 2945 boolean_t avail_spare, l2cache; 2946 libzfs_handle_t *hdl = zhp->zpool_hdl; 2947 2948 (void) snprintf(msg, sizeof (msg), 2949 dgettext(TEXT_DOMAIN, "cannot offline %s"), path); 2950 2951 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2952 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 2953 NULL)) == NULL) 2954 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 2955 2956 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 2957 2958 if (avail_spare) 2959 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 2960 2961 zc.zc_cookie = VDEV_STATE_OFFLINE; 2962 zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0; 2963 2964 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0) 2965 return (0); 2966 2967 switch (errno) { 2968 case EBUSY: 2969 2970 /* 2971 * There are no other replicas of this device. 2972 */ 2973 return (zfs_error(hdl, EZFS_NOREPLICAS, msg)); 2974 2975 case EEXIST: 2976 /* 2977 * The log device has unplayed logs 2978 */ 2979 return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, msg)); 2980 2981 default: 2982 return (zpool_standard_error(hdl, errno, msg)); 2983 } 2984 } 2985 2986 /* 2987 * Mark the given vdev faulted. 2988 */ 2989 int 2990 zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux) 2991 { 2992 zfs_cmd_t zc = { 0 }; 2993 char msg[1024]; 2994 libzfs_handle_t *hdl = zhp->zpool_hdl; 2995 2996 (void) snprintf(msg, sizeof (msg), 2997 dgettext(TEXT_DOMAIN, "cannot fault %llu"), guid); 2998 2999 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3000 zc.zc_guid = guid; 3001 zc.zc_cookie = VDEV_STATE_FAULTED; 3002 zc.zc_obj = aux; 3003 3004 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0) 3005 return (0); 3006 3007 switch (errno) { 3008 case EBUSY: 3009 3010 /* 3011 * There are no other replicas of this device. 3012 */ 3013 return (zfs_error(hdl, EZFS_NOREPLICAS, msg)); 3014 3015 default: 3016 return (zpool_standard_error(hdl, errno, msg)); 3017 } 3018 3019 } 3020 3021 /* 3022 * Mark the given vdev degraded. 3023 */ 3024 int 3025 zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux) 3026 { 3027 zfs_cmd_t zc = { 0 }; 3028 char msg[1024]; 3029 libzfs_handle_t *hdl = zhp->zpool_hdl; 3030 3031 (void) snprintf(msg, sizeof (msg), 3032 dgettext(TEXT_DOMAIN, "cannot degrade %llu"), guid); 3033 3034 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3035 zc.zc_guid = guid; 3036 zc.zc_cookie = VDEV_STATE_DEGRADED; 3037 zc.zc_obj = aux; 3038 3039 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0) 3040 return (0); 3041 3042 return (zpool_standard_error(hdl, errno, msg)); 3043 } 3044 3045 /* 3046 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as 3047 * a hot spare. 3048 */ 3049 static boolean_t 3050 is_replacing_spare(nvlist_t *search, nvlist_t *tgt, int which) 3051 { 3052 nvlist_t **child; 3053 uint_t c, children; 3054 char *type; 3055 3056 if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child, 3057 &children) == 0) { 3058 verify(nvlist_lookup_string(search, ZPOOL_CONFIG_TYPE, 3059 &type) == 0); 3060 3061 if (strcmp(type, VDEV_TYPE_SPARE) == 0 && 3062 children == 2 && child[which] == tgt) 3063 return (B_TRUE); 3064 3065 for (c = 0; c < children; c++) 3066 if (is_replacing_spare(child[c], tgt, which)) 3067 return (B_TRUE); 3068 } 3069 3070 return (B_FALSE); 3071 } 3072 3073 /* 3074 * Attach new_disk (fully described by nvroot) to old_disk. 3075 * If 'replacing' is specified, the new disk will replace the old one. 3076 */ 3077 int 3078 zpool_vdev_attach(zpool_handle_t *zhp, 3079 const char *old_disk, const char *new_disk, nvlist_t *nvroot, int replacing) 3080 { 3081 zfs_cmd_t zc = { 0 }; 3082 char msg[1024]; 3083 int ret; 3084 nvlist_t *tgt, *newvd; 3085 boolean_t avail_spare, l2cache, islog; 3086 uint64_t val; 3087 char *newname; 3088 nvlist_t **child; 3089 uint_t children; 3090 nvlist_t *config_root; 3091 libzfs_handle_t *hdl = zhp->zpool_hdl; 3092 boolean_t rootpool = zpool_is_bootable(zhp); 3093 3094 if (replacing) 3095 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 3096 "cannot replace %s with %s"), old_disk, new_disk); 3097 else 3098 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 3099 "cannot attach %s to %s"), new_disk, old_disk); 3100 3101 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3102 if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache, 3103 &islog)) == NULL) 3104 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 3105 3106 if (avail_spare) 3107 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 3108 3109 if (l2cache) 3110 return (zfs_error(hdl, EZFS_ISL2CACHE, msg)); 3111 3112 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 3113 zc.zc_cookie = replacing; 3114 3115 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, 3116 &child, &children) != 0 || children != 1) { 3117 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3118 "new device must be a single disk")); 3119 return (zfs_error(hdl, EZFS_INVALCONFIG, msg)); 3120 } 3121 3122 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL), 3123 ZPOOL_CONFIG_VDEV_TREE, &config_root) == 0); 3124 3125 if ((newname = zpool_vdev_name(NULL, NULL, child[0], 0)) == NULL) 3126 return (-1); 3127 3128 newvd = zpool_find_vdev(zhp, newname, &avail_spare, &l2cache, NULL); 3129 /* 3130 * If the target is a hot spare that has been swapped in, we can only 3131 * replace it with another hot spare. 3132 */ 3133 if (replacing && 3134 nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 && 3135 (newvd == NULL || !avail_spare) && 3136 is_replacing_spare(config_root, tgt, 1)) { 3137 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3138 "can only be replaced by another hot spare")); 3139 free(newname); 3140 return (zfs_error(hdl, EZFS_BADTARGET, msg)); 3141 } 3142 3143 free(newname); 3144 3145 if (replacing && avail_spare && !vdev_is_online(newvd)) { 3146 (void) zpool_standard_error(hdl, ENXIO, msg); 3147 return (-1); 3148 } 3149 3150 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0) 3151 return (-1); 3152 3153 ret = zfs_ioctl(hdl, ZFS_IOC_VDEV_ATTACH, &zc); 3154 3155 zcmd_free_nvlists(&zc); 3156 3157 if (ret == 0) { 3158 if (rootpool) { 3159 /* 3160 * XXX need a better way to prevent user from 3161 * booting up a half-baked vdev. 3162 */ 3163 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Make " 3164 "sure to wait until resilver is done " 3165 "before rebooting.\n")); 3166 } 3167 return (0); 3168 } 3169 3170 switch (errno) { 3171 case ENOTSUP: 3172 /* 3173 * Can't attach to or replace this type of vdev. 3174 */ 3175 if (replacing) { 3176 uint64_t version = zpool_get_prop_int(zhp, 3177 ZPOOL_PROP_VERSION, NULL); 3178 3179 if (islog) 3180 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3181 "cannot replace a log with a spare")); 3182 else if (version >= SPA_VERSION_MULTI_REPLACE) 3183 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3184 "already in replacing/spare config; wait " 3185 "for completion or use 'zpool detach'")); 3186 else 3187 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3188 "cannot replace a replacing device")); 3189 } else { 3190 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3191 "can only attach to mirrors and top-level " 3192 "disks")); 3193 } 3194 (void) zfs_error(hdl, EZFS_BADTARGET, msg); 3195 break; 3196 3197 case EINVAL: 3198 /* 3199 * The new device must be a single disk. 3200 */ 3201 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3202 "new device must be a single disk")); 3203 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg); 3204 break; 3205 3206 case EBUSY: 3207 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy, " 3208 "or device removal is in progress"), 3209 new_disk); 3210 (void) zfs_error(hdl, EZFS_BADDEV, msg); 3211 break; 3212 3213 case EOVERFLOW: 3214 /* 3215 * The new device is too small. 3216 */ 3217 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3218 "device is too small")); 3219 (void) zfs_error(hdl, EZFS_BADDEV, msg); 3220 break; 3221 3222 case EDOM: 3223 /* 3224 * The new device has a different optimal sector size. 3225 */ 3226 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3227 "new device has a different optimal sector size; use the " 3228 "option '-o ashift=N' to override the optimal size")); 3229 (void) zfs_error(hdl, EZFS_BADDEV, msg); 3230 break; 3231 3232 case ENAMETOOLONG: 3233 /* 3234 * The resulting top-level vdev spec won't fit in the label. 3235 */ 3236 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, msg); 3237 break; 3238 3239 default: 3240 (void) zpool_standard_error(hdl, errno, msg); 3241 } 3242 3243 return (-1); 3244 } 3245 3246 /* 3247 * Detach the specified device. 3248 */ 3249 int 3250 zpool_vdev_detach(zpool_handle_t *zhp, const char *path) 3251 { 3252 zfs_cmd_t zc = { 0 }; 3253 char msg[1024]; 3254 nvlist_t *tgt; 3255 boolean_t avail_spare, l2cache; 3256 libzfs_handle_t *hdl = zhp->zpool_hdl; 3257 3258 (void) snprintf(msg, sizeof (msg), 3259 dgettext(TEXT_DOMAIN, "cannot detach %s"), path); 3260 3261 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3262 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 3263 NULL)) == NULL) 3264 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 3265 3266 if (avail_spare) 3267 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 3268 3269 if (l2cache) 3270 return (zfs_error(hdl, EZFS_ISL2CACHE, msg)); 3271 3272 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 3273 3274 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0) 3275 return (0); 3276 3277 switch (errno) { 3278 3279 case ENOTSUP: 3280 /* 3281 * Can't detach from this type of vdev. 3282 */ 3283 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only " 3284 "applicable to mirror and replacing vdevs")); 3285 (void) zfs_error(hdl, EZFS_BADTARGET, msg); 3286 break; 3287 3288 case EBUSY: 3289 /* 3290 * There are no other replicas of this device. 3291 */ 3292 (void) zfs_error(hdl, EZFS_NOREPLICAS, msg); 3293 break; 3294 3295 default: 3296 (void) zpool_standard_error(hdl, errno, msg); 3297 } 3298 3299 return (-1); 3300 } 3301 3302 /* 3303 * Find a mirror vdev in the source nvlist. 3304 * 3305 * The mchild array contains a list of disks in one of the top-level mirrors 3306 * of the source pool. The schild array contains a list of disks that the 3307 * user specified on the command line. We loop over the mchild array to 3308 * see if any entry in the schild array matches. 3309 * 3310 * If a disk in the mchild array is found in the schild array, we return 3311 * the index of that entry. Otherwise we return -1. 3312 */ 3313 static int 3314 find_vdev_entry(zpool_handle_t *zhp, nvlist_t **mchild, uint_t mchildren, 3315 nvlist_t **schild, uint_t schildren) 3316 { 3317 uint_t mc; 3318 3319 for (mc = 0; mc < mchildren; mc++) { 3320 uint_t sc; 3321 char *mpath = zpool_vdev_name(zhp->zpool_hdl, zhp, 3322 mchild[mc], 0); 3323 3324 for (sc = 0; sc < schildren; sc++) { 3325 char *spath = zpool_vdev_name(zhp->zpool_hdl, zhp, 3326 schild[sc], 0); 3327 boolean_t result = (strcmp(mpath, spath) == 0); 3328 3329 free(spath); 3330 if (result) { 3331 free(mpath); 3332 return (mc); 3333 } 3334 } 3335 3336 free(mpath); 3337 } 3338 3339 return (-1); 3340 } 3341 3342 /* 3343 * Split a mirror pool. If newroot points to null, then a new nvlist 3344 * is generated and it is the responsibility of the caller to free it. 3345 */ 3346 int 3347 zpool_vdev_split(zpool_handle_t *zhp, char *newname, nvlist_t **newroot, 3348 nvlist_t *props, splitflags_t flags) 3349 { 3350 zfs_cmd_t zc = { 0 }; 3351 char msg[1024]; 3352 nvlist_t *tree, *config, **child, **newchild, *newconfig = NULL; 3353 nvlist_t **varray = NULL, *zc_props = NULL; 3354 uint_t c, children, newchildren, lastlog = 0, vcount, found = 0; 3355 libzfs_handle_t *hdl = zhp->zpool_hdl; 3356 uint64_t vers; 3357 boolean_t freelist = B_FALSE, memory_err = B_TRUE; 3358 int retval = 0; 3359 3360 (void) snprintf(msg, sizeof (msg), 3361 dgettext(TEXT_DOMAIN, "Unable to split %s"), zhp->zpool_name); 3362 3363 if (!zpool_name_valid(hdl, B_FALSE, newname)) 3364 return (zfs_error(hdl, EZFS_INVALIDNAME, msg)); 3365 3366 if ((config = zpool_get_config(zhp, NULL)) == NULL) { 3367 (void) fprintf(stderr, gettext("Internal error: unable to " 3368 "retrieve pool configuration\n")); 3369 return (-1); 3370 } 3371 3372 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree) 3373 == 0); 3374 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &vers) == 0); 3375 3376 if (props) { 3377 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE }; 3378 if ((zc_props = zpool_valid_proplist(hdl, zhp->zpool_name, 3379 props, vers, flags, msg)) == NULL) 3380 return (-1); 3381 } 3382 3383 if (nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child, 3384 &children) != 0) { 3385 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3386 "Source pool is missing vdev tree")); 3387 nvlist_free(zc_props); 3388 return (-1); 3389 } 3390 3391 varray = zfs_alloc(hdl, children * sizeof (nvlist_t *)); 3392 vcount = 0; 3393 3394 if (*newroot == NULL || 3395 nvlist_lookup_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, 3396 &newchild, &newchildren) != 0) 3397 newchildren = 0; 3398 3399 for (c = 0; c < children; c++) { 3400 uint64_t is_log = B_FALSE, is_hole = B_FALSE; 3401 char *type; 3402 nvlist_t **mchild, *vdev; 3403 uint_t mchildren; 3404 int entry; 3405 3406 /* 3407 * Unlike cache & spares, slogs are stored in the 3408 * ZPOOL_CONFIG_CHILDREN array. We filter them out here. 3409 */ 3410 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG, 3411 &is_log); 3412 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE, 3413 &is_hole); 3414 if (is_log || is_hole) { 3415 /* 3416 * Create a hole vdev and put it in the config. 3417 */ 3418 if (nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) != 0) 3419 goto out; 3420 if (nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE, 3421 VDEV_TYPE_HOLE) != 0) 3422 goto out; 3423 if (nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_HOLE, 3424 1) != 0) 3425 goto out; 3426 if (lastlog == 0) 3427 lastlog = vcount; 3428 varray[vcount++] = vdev; 3429 continue; 3430 } 3431 lastlog = 0; 3432 verify(nvlist_lookup_string(child[c], ZPOOL_CONFIG_TYPE, &type) 3433 == 0); 3434 if (strcmp(type, VDEV_TYPE_MIRROR) != 0) { 3435 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3436 "Source pool must be composed only of mirrors\n")); 3437 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg); 3438 goto out; 3439 } 3440 3441 verify(nvlist_lookup_nvlist_array(child[c], 3442 ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0); 3443 3444 /* find or add an entry for this top-level vdev */ 3445 if (newchildren > 0 && 3446 (entry = find_vdev_entry(zhp, mchild, mchildren, 3447 newchild, newchildren)) >= 0) { 3448 /* We found a disk that the user specified. */ 3449 vdev = mchild[entry]; 3450 ++found; 3451 } else { 3452 /* User didn't specify a disk for this vdev. */ 3453 vdev = mchild[mchildren - 1]; 3454 } 3455 3456 if (nvlist_dup(vdev, &varray[vcount++], 0) != 0) 3457 goto out; 3458 } 3459 3460 /* did we find every disk the user specified? */ 3461 if (found != newchildren) { 3462 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "Device list must " 3463 "include at most one disk from each mirror")); 3464 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg); 3465 goto out; 3466 } 3467 3468 /* Prepare the nvlist for populating. */ 3469 if (*newroot == NULL) { 3470 if (nvlist_alloc(newroot, NV_UNIQUE_NAME, 0) != 0) 3471 goto out; 3472 freelist = B_TRUE; 3473 if (nvlist_add_string(*newroot, ZPOOL_CONFIG_TYPE, 3474 VDEV_TYPE_ROOT) != 0) 3475 goto out; 3476 } else { 3477 verify(nvlist_remove_all(*newroot, ZPOOL_CONFIG_CHILDREN) == 0); 3478 } 3479 3480 /* Add all the children we found */ 3481 if (nvlist_add_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, varray, 3482 lastlog == 0 ? vcount : lastlog) != 0) 3483 goto out; 3484 3485 /* 3486 * If we're just doing a dry run, exit now with success. 3487 */ 3488 if (flags.dryrun) { 3489 memory_err = B_FALSE; 3490 freelist = B_FALSE; 3491 goto out; 3492 } 3493 3494 /* now build up the config list & call the ioctl */ 3495 if (nvlist_alloc(&newconfig, NV_UNIQUE_NAME, 0) != 0) 3496 goto out; 3497 3498 if (nvlist_add_nvlist(newconfig, 3499 ZPOOL_CONFIG_VDEV_TREE, *newroot) != 0 || 3500 nvlist_add_string(newconfig, 3501 ZPOOL_CONFIG_POOL_NAME, newname) != 0 || 3502 nvlist_add_uint64(newconfig, ZPOOL_CONFIG_VERSION, vers) != 0) 3503 goto out; 3504 3505 /* 3506 * The new pool is automatically part of the namespace unless we 3507 * explicitly export it. 3508 */ 3509 if (!flags.import) 3510 zc.zc_cookie = ZPOOL_EXPORT_AFTER_SPLIT; 3511 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3512 (void) strlcpy(zc.zc_string, newname, sizeof (zc.zc_string)); 3513 if (zcmd_write_conf_nvlist(hdl, &zc, newconfig) != 0) 3514 goto out; 3515 if (zc_props != NULL && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0) 3516 goto out; 3517 3518 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SPLIT, &zc) != 0) { 3519 retval = zpool_standard_error(hdl, errno, msg); 3520 goto out; 3521 } 3522 3523 freelist = B_FALSE; 3524 memory_err = B_FALSE; 3525 3526 out: 3527 if (varray != NULL) { 3528 int v; 3529 3530 for (v = 0; v < vcount; v++) 3531 nvlist_free(varray[v]); 3532 free(varray); 3533 } 3534 zcmd_free_nvlists(&zc); 3535 nvlist_free(zc_props); 3536 nvlist_free(newconfig); 3537 if (freelist) { 3538 nvlist_free(*newroot); 3539 *newroot = NULL; 3540 } 3541 3542 if (retval != 0) 3543 return (retval); 3544 3545 if (memory_err) 3546 return (no_memory(hdl)); 3547 3548 return (0); 3549 } 3550 3551 /* 3552 * Remove the given device. 3553 */ 3554 int 3555 zpool_vdev_remove(zpool_handle_t *zhp, const char *path) 3556 { 3557 zfs_cmd_t zc = { 0 }; 3558 char msg[1024]; 3559 nvlist_t *tgt; 3560 boolean_t avail_spare, l2cache, islog; 3561 libzfs_handle_t *hdl = zhp->zpool_hdl; 3562 uint64_t version; 3563 3564 (void) snprintf(msg, sizeof (msg), 3565 dgettext(TEXT_DOMAIN, "cannot remove %s"), path); 3566 3567 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3568 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 3569 &islog)) == NULL) 3570 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 3571 3572 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL); 3573 if (islog && version < SPA_VERSION_HOLES) { 3574 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3575 "pool must be upgraded to support log removal")); 3576 return (zfs_error(hdl, EZFS_BADVERSION, msg)); 3577 } 3578 3579 zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID); 3580 3581 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0) 3582 return (0); 3583 3584 switch (errno) { 3585 3586 case EINVAL: 3587 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3588 "invalid config; all top-level vdevs must " 3589 "have the same sector size and not be raidz.")); 3590 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg); 3591 break; 3592 3593 case EBUSY: 3594 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3595 "Pool busy; removal may already be in progress")); 3596 (void) zfs_error(hdl, EZFS_BUSY, msg); 3597 break; 3598 3599 default: 3600 (void) zpool_standard_error(hdl, errno, msg); 3601 } 3602 return (-1); 3603 } 3604 3605 int 3606 zpool_vdev_remove_cancel(zpool_handle_t *zhp) 3607 { 3608 zfs_cmd_t zc = { 0 }; 3609 char msg[1024]; 3610 libzfs_handle_t *hdl = zhp->zpool_hdl; 3611 3612 (void) snprintf(msg, sizeof (msg), 3613 dgettext(TEXT_DOMAIN, "cannot cancel removal")); 3614 3615 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3616 zc.zc_cookie = 1; 3617 3618 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0) 3619 return (0); 3620 3621 return (zpool_standard_error(hdl, errno, msg)); 3622 } 3623 3624 int 3625 zpool_vdev_indirect_size(zpool_handle_t *zhp, const char *path, 3626 uint64_t *sizep) 3627 { 3628 char msg[1024]; 3629 nvlist_t *tgt; 3630 boolean_t avail_spare, l2cache, islog; 3631 libzfs_handle_t *hdl = zhp->zpool_hdl; 3632 3633 (void) snprintf(msg, sizeof (msg), 3634 dgettext(TEXT_DOMAIN, "cannot determine indirect size of %s"), 3635 path); 3636 3637 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 3638 &islog)) == NULL) 3639 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 3640 3641 if (avail_spare || l2cache || islog) { 3642 *sizep = 0; 3643 return (0); 3644 } 3645 3646 if (nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_INDIRECT_SIZE, sizep) != 0) { 3647 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3648 "indirect size not available")); 3649 return (zfs_error(hdl, EINVAL, msg)); 3650 } 3651 return (0); 3652 } 3653 3654 /* 3655 * Clear the errors for the pool, or the particular device if specified. 3656 */ 3657 int 3658 zpool_clear(zpool_handle_t *zhp, const char *path, nvlist_t *rewindnvl) 3659 { 3660 zfs_cmd_t zc = { 0 }; 3661 char msg[1024]; 3662 nvlist_t *tgt; 3663 zpool_load_policy_t policy; 3664 boolean_t avail_spare, l2cache; 3665 libzfs_handle_t *hdl = zhp->zpool_hdl; 3666 nvlist_t *nvi = NULL; 3667 int error; 3668 3669 if (path) 3670 (void) snprintf(msg, sizeof (msg), 3671 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"), 3672 path); 3673 else 3674 (void) snprintf(msg, sizeof (msg), 3675 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"), 3676 zhp->zpool_name); 3677 3678 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3679 if (path) { 3680 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, 3681 &l2cache, NULL)) == NULL) 3682 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 3683 3684 /* 3685 * Don't allow error clearing for hot spares. Do allow 3686 * error clearing for l2cache devices. 3687 */ 3688 if (avail_spare) 3689 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 3690 3691 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, 3692 &zc.zc_guid) == 0); 3693 } 3694 3695 zpool_get_load_policy(rewindnvl, &policy); 3696 zc.zc_cookie = policy.zlp_rewind; 3697 3698 if (zcmd_alloc_dst_nvlist(hdl, &zc, zhp->zpool_config_size * 2) != 0) 3699 return (-1); 3700 3701 if (zcmd_write_src_nvlist(hdl, &zc, rewindnvl) != 0) 3702 return (-1); 3703 3704 while ((error = zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc)) != 0 && 3705 errno == ENOMEM) { 3706 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) { 3707 zcmd_free_nvlists(&zc); 3708 return (-1); 3709 } 3710 } 3711 3712 if (!error || ((policy.zlp_rewind & ZPOOL_TRY_REWIND) && 3713 errno != EPERM && errno != EACCES)) { 3714 if (policy.zlp_rewind & 3715 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) { 3716 (void) zcmd_read_dst_nvlist(hdl, &zc, &nvi); 3717 zpool_rewind_exclaim(hdl, zc.zc_name, 3718 ((policy.zlp_rewind & ZPOOL_TRY_REWIND) != 0), 3719 nvi); 3720 nvlist_free(nvi); 3721 } 3722 zcmd_free_nvlists(&zc); 3723 return (0); 3724 } 3725 3726 zcmd_free_nvlists(&zc); 3727 return (zpool_standard_error(hdl, errno, msg)); 3728 } 3729 3730 /* 3731 * Similar to zpool_clear(), but takes a GUID (used by fmd). 3732 */ 3733 int 3734 zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid) 3735 { 3736 zfs_cmd_t zc = { 0 }; 3737 char msg[1024]; 3738 libzfs_handle_t *hdl = zhp->zpool_hdl; 3739 3740 (void) snprintf(msg, sizeof (msg), 3741 dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"), 3742 guid); 3743 3744 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3745 zc.zc_guid = guid; 3746 zc.zc_cookie = ZPOOL_NO_REWIND; 3747 3748 if (ioctl(hdl->libzfs_fd, ZFS_IOC_CLEAR, &zc) == 0) 3749 return (0); 3750 3751 return (zpool_standard_error(hdl, errno, msg)); 3752 } 3753 3754 /* 3755 * Change the GUID for a pool. 3756 */ 3757 int 3758 zpool_reguid(zpool_handle_t *zhp) 3759 { 3760 char msg[1024]; 3761 libzfs_handle_t *hdl = zhp->zpool_hdl; 3762 zfs_cmd_t zc = { 0 }; 3763 3764 (void) snprintf(msg, sizeof (msg), 3765 dgettext(TEXT_DOMAIN, "cannot reguid '%s'"), zhp->zpool_name); 3766 3767 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3768 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REGUID, &zc) == 0) 3769 return (0); 3770 3771 return (zpool_standard_error(hdl, errno, msg)); 3772 } 3773 3774 /* 3775 * Reopen the pool. 3776 */ 3777 int 3778 zpool_reopen(zpool_handle_t *zhp) 3779 { 3780 zfs_cmd_t zc = { 0 }; 3781 char msg[1024]; 3782 libzfs_handle_t *hdl = zhp->zpool_hdl; 3783 3784 (void) snprintf(msg, sizeof (msg), 3785 dgettext(TEXT_DOMAIN, "cannot reopen '%s'"), 3786 zhp->zpool_name); 3787 3788 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3789 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REOPEN, &zc) == 0) 3790 return (0); 3791 return (zpool_standard_error(hdl, errno, msg)); 3792 } 3793 3794 /* call into libzfs_core to execute the sync IOCTL per pool */ 3795 int 3796 zpool_sync_one(zpool_handle_t *zhp, void *data) 3797 { 3798 int ret; 3799 libzfs_handle_t *hdl = zpool_get_handle(zhp); 3800 const char *pool_name = zpool_get_name(zhp); 3801 boolean_t *force = data; 3802 nvlist_t *innvl = fnvlist_alloc(); 3803 3804 fnvlist_add_boolean_value(innvl, "force", *force); 3805 if ((ret = lzc_sync(pool_name, innvl, NULL)) != 0) { 3806 nvlist_free(innvl); 3807 return (zpool_standard_error_fmt(hdl, ret, 3808 dgettext(TEXT_DOMAIN, "sync '%s' failed"), pool_name)); 3809 } 3810 nvlist_free(innvl); 3811 3812 return (0); 3813 } 3814 3815 /* 3816 * Convert from a devid string to a path. 3817 */ 3818 static char * 3819 devid_to_path(char *devid_str) 3820 { 3821 ddi_devid_t devid; 3822 char *minor; 3823 char *path; 3824 devid_nmlist_t *list = NULL; 3825 int ret; 3826 3827 if (devid_str_decode(devid_str, &devid, &minor) != 0) 3828 return (NULL); 3829 3830 ret = devid_deviceid_to_nmlist("/dev", devid, minor, &list); 3831 3832 devid_str_free(minor); 3833 devid_free(devid); 3834 3835 if (ret != 0) 3836 return (NULL); 3837 3838 /* 3839 * In a case the strdup() fails, we will just return NULL below. 3840 */ 3841 path = strdup(list[0].devname); 3842 3843 devid_free_nmlist(list); 3844 3845 return (path); 3846 } 3847 3848 /* 3849 * Convert from a path to a devid string. 3850 */ 3851 static char * 3852 path_to_devid(const char *path) 3853 { 3854 int fd; 3855 ddi_devid_t devid; 3856 char *minor, *ret; 3857 3858 if ((fd = open(path, O_RDONLY)) < 0) 3859 return (NULL); 3860 3861 minor = NULL; 3862 ret = NULL; 3863 if (devid_get(fd, &devid) == 0) { 3864 if (devid_get_minor_name(fd, &minor) == 0) 3865 ret = devid_str_encode(devid, minor); 3866 if (minor != NULL) 3867 devid_str_free(minor); 3868 devid_free(devid); 3869 } 3870 (void) close(fd); 3871 3872 return (ret); 3873 } 3874 3875 struct path_from_physpath_walker_args { 3876 char *pfpwa_path; 3877 }; 3878 3879 /* 3880 * Walker for use with di_devlink_walk(). Stores the "/dev" path of the first 3881 * primary devlink (i.e., the first devlink which refers to our "/devices" 3882 * node) and stops walking. 3883 */ 3884 static int 3885 path_from_physpath_walker(di_devlink_t devlink, void *arg) 3886 { 3887 struct path_from_physpath_walker_args *pfpwa = arg; 3888 3889 if (di_devlink_type(devlink) != DI_PRIMARY_LINK) { 3890 return (DI_WALK_CONTINUE); 3891 } 3892 3893 verify(pfpwa->pfpwa_path == NULL); 3894 if ((pfpwa->pfpwa_path = strdup(di_devlink_path(devlink))) != NULL) { 3895 return (DI_WALK_TERMINATE); 3896 } 3897 3898 return (DI_WALK_CONTINUE); 3899 } 3900 3901 /* 3902 * Search for a "/dev" path that refers to our physical path. Returns the new 3903 * path if one is found and it does not match the existing "path" value. If 3904 * the value is unchanged, or one could not be found, returns NULL. 3905 */ 3906 static char * 3907 path_from_physpath(libzfs_handle_t *hdl, const char *path, 3908 const char *physpath) 3909 { 3910 struct path_from_physpath_walker_args pfpwa; 3911 3912 if (physpath == NULL) { 3913 return (NULL); 3914 } 3915 3916 if (hdl->libzfs_devlink == NULL) { 3917 if ((hdl->libzfs_devlink = di_devlink_init(NULL, 0)) == 3918 DI_LINK_NIL) { 3919 /* 3920 * We may not be able to open a handle if this process 3921 * is insufficiently privileged, or we are too early in 3922 * boot for devfsadm to be ready. Ignore this error 3923 * and defer the path check to a subsequent run. 3924 */ 3925 return (NULL); 3926 } 3927 } 3928 3929 pfpwa.pfpwa_path = NULL; 3930 (void) di_devlink_walk(hdl->libzfs_devlink, NULL, physpath, 3931 DI_PRIMARY_LINK, &pfpwa, path_from_physpath_walker); 3932 3933 if (path != NULL && pfpwa.pfpwa_path != NULL && 3934 strcmp(path, pfpwa.pfpwa_path) == 0) { 3935 /* 3936 * If the path is already correct, no change is required. 3937 */ 3938 free(pfpwa.pfpwa_path); 3939 return (NULL); 3940 } 3941 3942 return (pfpwa.pfpwa_path); 3943 } 3944 3945 /* 3946 * Issue the necessary ioctl() to update the stored path value for the vdev. We 3947 * ignore any failure here, since a common case is for an unprivileged user to 3948 * type 'zpool status', and we'll display the correct information anyway. 3949 */ 3950 static void 3951 set_path(zpool_handle_t *zhp, nvlist_t *nv, const char *path) 3952 { 3953 zfs_cmd_t zc = { 0 }; 3954 3955 (void) strncpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3956 (void) strncpy(zc.zc_value, path, sizeof (zc.zc_value)); 3957 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, 3958 &zc.zc_guid) == 0); 3959 3960 (void) ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SETPATH, &zc); 3961 } 3962 3963 /* 3964 * This routine is responsible for identifying when disks have been 3965 * reconfigured in a new location. The kernel will have opened the device by 3966 * devid, but the path will still refer to the old location. To catch this, we 3967 * first do a path -> devid translation (which is fast for the common case). 3968 * If the devid matches, we're done. If not, we do a reverse devid -> path 3969 * translation and issue the appropriate ioctl() to update the path of the 3970 * vdev. 3971 */ 3972 void 3973 zpool_vdev_refresh_path(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv) 3974 { 3975 char *path = NULL; 3976 char *newpath = NULL; 3977 char *physpath = NULL; 3978 char *devid = NULL; 3979 3980 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) != 0) { 3981 return; 3982 } 3983 3984 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &devid) == 0) { 3985 /* 3986 * This vdev has a devid. We can use it to check the current 3987 * path. 3988 */ 3989 char *newdevid = path_to_devid(path); 3990 3991 if (newdevid == NULL || strcmp(devid, newdevid) != 0) { 3992 newpath = devid_to_path(devid); 3993 } 3994 3995 if (newdevid != NULL) { 3996 devid_str_free(newdevid); 3997 } 3998 3999 } else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PHYS_PATH, 4000 &physpath) == 0) { 4001 /* 4002 * This vdev does not have a devid, but it does have a physical 4003 * path. Attempt to translate this to a /dev path. 4004 */ 4005 newpath = path_from_physpath(hdl, path, physpath); 4006 } 4007 4008 if (newpath == NULL) { 4009 /* 4010 * No path update is required. 4011 */ 4012 return; 4013 } 4014 4015 set_path(zhp, nv, newpath); 4016 fnvlist_add_string(nv, ZPOOL_CONFIG_PATH, newpath); 4017 4018 free(newpath); 4019 } 4020 4021 /* 4022 * Given a vdev, return the name to display in iostat. If the vdev has a path, 4023 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type. 4024 * We will confirm that the path and name of the vdev are current, and update 4025 * them if not. We also check if this is a whole disk, in which case we strip 4026 * off the trailing 's0' slice name. 4027 * 4028 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any 4029 * of these checks. 4030 */ 4031 char * 4032 zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv, 4033 int name_flags) 4034 { 4035 char *path, *type, *env; 4036 uint64_t value; 4037 4038 /* 4039 * vdev_name will be "root"/"root-0" for the root vdev, but it is the 4040 * zpool name that will be displayed to the user. 4041 */ 4042 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0); 4043 if (zhp != NULL && strcmp(type, "root") == 0) 4044 return (zfs_strdup(hdl, zpool_get_name(zhp))); 4045 4046 env = getenv("ZPOOL_VDEV_NAME_PATH"); 4047 if (env && (strtoul(env, NULL, 0) > 0 || 4048 !strncasecmp(env, "YES", 3) || !strncasecmp(env, "ON", 2))) 4049 name_flags |= VDEV_NAME_PATH; 4050 4051 env = getenv("ZPOOL_VDEV_NAME_GUID"); 4052 if (env && (strtoul(env, NULL, 0) > 0 || 4053 !strncasecmp(env, "YES", 3) || !strncasecmp(env, "ON", 2))) 4054 name_flags |= VDEV_NAME_GUID; 4055 4056 env = getenv("ZPOOL_VDEV_NAME_FOLLOW_LINKS"); 4057 if (env && (strtoul(env, NULL, 0) > 0 || 4058 !strncasecmp(env, "YES", 3) || !strncasecmp(env, "ON", 2))) 4059 name_flags |= VDEV_NAME_FOLLOW_LINKS; 4060 4061 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, &value) == 0 || 4062 name_flags & VDEV_NAME_GUID) { 4063 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value); 4064 path = zfs_asprintf(hdl, "%llu", (u_longlong_t)value); 4065 } else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) { 4066 vdev_stat_t *vs; 4067 uint_t vsc; 4068 4069 /* 4070 * If the device is dead (faulted, offline, etc) then don't 4071 * bother opening it. Otherwise we may be forcing the user to 4072 * open a misbehaving device, which can have undesirable 4073 * effects. 4074 */ 4075 if (nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS, 4076 (uint64_t **)&vs, &vsc) != 0 || 4077 vs->vs_state < VDEV_STATE_DEGRADED || 4078 zhp == NULL) { 4079 path = zfs_strdup(hdl, path); 4080 goto after_open; 4081 } 4082 4083 /* 4084 * Refresh the /dev path for this vdev if required, then ensure 4085 * we're using the latest path value: 4086 */ 4087 zpool_vdev_refresh_path(hdl, zhp, nv); 4088 path = fnvlist_lookup_string(nv, ZPOOL_CONFIG_PATH); 4089 4090 if (name_flags & VDEV_NAME_FOLLOW_LINKS) { 4091 char *rp = realpath(path, NULL); 4092 if (rp == NULL) 4093 no_memory(hdl); 4094 path = rp; 4095 } else { 4096 path = zfs_strdup(hdl, path); 4097 } 4098 4099 after_open: 4100 if (strncmp(path, ZFS_DISK_ROOTD, 4101 sizeof (ZFS_DISK_ROOTD) - 1) == 0) { 4102 const char *p2 = path + sizeof (ZFS_DISK_ROOTD) - 1; 4103 4104 memmove(path, p2, strlen(p2) + 1); 4105 } 4106 4107 /* 4108 * Remove the partition from the path it this is a whole disk. 4109 */ 4110 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, &value) 4111 == 0 && value && !(name_flags & VDEV_NAME_PATH)) { 4112 int pathlen = strlen(path); 4113 4114 /* 4115 * If it starts with c#, and ends with "s0" or "s1", 4116 * chop the slice off, or if it ends with "s0/old" or 4117 * "s1/old", remove the slice from the middle. 4118 */ 4119 if (CTD_CHECK(path)) { 4120 if (strcmp(&path[pathlen - 2], "s0") == 0 || 4121 strcmp(&path[pathlen - 2], "s1") == 0) { 4122 path[pathlen - 2] = '\0'; 4123 } else if (pathlen > 6 && 4124 (strcmp(&path[pathlen - 6], 4125 "s0/old") == 0 || 4126 strcmp(&path[pathlen - 6], 4127 "s1/old") == 0)) { 4128 (void) strcpy(&path[pathlen - 6], 4129 "/old"); 4130 } 4131 } 4132 return (path); 4133 } 4134 } else { 4135 /* 4136 * If it's a raidz device, we need to stick in the parity level. 4137 */ 4138 if (strcmp(type, VDEV_TYPE_RAIDZ) == 0) { 4139 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY, 4140 &value) == 0); 4141 path = zfs_asprintf(hdl, "%s%llu", type, 4142 (u_longlong_t)value); 4143 } else { 4144 path = zfs_strdup(hdl, type); 4145 } 4146 4147 /* 4148 * We identify each top-level vdev by using a <type-id> 4149 * naming convention. 4150 */ 4151 if (name_flags & VDEV_NAME_TYPE_ID) { 4152 uint64_t id; 4153 char *tmp; 4154 4155 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID, 4156 &id) == 0); 4157 tmp = zfs_asprintf(hdl, "%s-%llu", path, 4158 (u_longlong_t)id); 4159 free(path); 4160 path = tmp; 4161 } 4162 } 4163 4164 return (path); 4165 } 4166 4167 static int 4168 zbookmark_mem_compare(const void *a, const void *b) 4169 { 4170 return (memcmp(a, b, sizeof (zbookmark_phys_t))); 4171 } 4172 4173 /* 4174 * Retrieve the persistent error log, uniquify the members, and return to the 4175 * caller. 4176 */ 4177 int 4178 zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp) 4179 { 4180 zfs_cmd_t zc = { 0 }; 4181 uint64_t count; 4182 zbookmark_phys_t *zb = NULL; 4183 int i; 4184 4185 /* 4186 * Retrieve the raw error list from the kernel. If the number of errors 4187 * has increased, allocate more space and continue until we get the 4188 * entire list. 4189 */ 4190 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_ERRCOUNT, 4191 &count) == 0); 4192 if (count == 0) 4193 return (0); 4194 if ((zc.zc_nvlist_dst = (uintptr_t)zfs_alloc(zhp->zpool_hdl, 4195 count * sizeof (zbookmark_phys_t))) == (uintptr_t)NULL) 4196 return (-1); 4197 zc.zc_nvlist_dst_size = count; 4198 (void) strcpy(zc.zc_name, zhp->zpool_name); 4199 for (;;) { 4200 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_ERROR_LOG, 4201 &zc) != 0) { 4202 free((void *)(uintptr_t)zc.zc_nvlist_dst); 4203 if (errno == ENOMEM) { 4204 void *dst; 4205 4206 count = zc.zc_nvlist_dst_size; 4207 dst = zfs_alloc(zhp->zpool_hdl, count * 4208 sizeof (zbookmark_phys_t)); 4209 if (dst == NULL) 4210 return (-1); 4211 zc.zc_nvlist_dst = (uintptr_t)dst; 4212 } else { 4213 return (-1); 4214 } 4215 } else { 4216 break; 4217 } 4218 } 4219 4220 /* 4221 * Sort the resulting bookmarks. This is a little confusing due to the 4222 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last 4223 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks 4224 * _not_ copied as part of the process. So we point the start of our 4225 * array appropriate and decrement the total number of elements. 4226 */ 4227 zb = ((zbookmark_phys_t *)(uintptr_t)zc.zc_nvlist_dst) + 4228 zc.zc_nvlist_dst_size; 4229 count -= zc.zc_nvlist_dst_size; 4230 4231 qsort(zb, count, sizeof (zbookmark_phys_t), zbookmark_mem_compare); 4232 4233 verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0); 4234 4235 /* 4236 * Fill in the nverrlistp with nvlist's of dataset and object numbers. 4237 */ 4238 for (i = 0; i < count; i++) { 4239 nvlist_t *nv; 4240 4241 /* ignoring zb_blkid and zb_level for now */ 4242 if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset && 4243 zb[i-1].zb_object == zb[i].zb_object) 4244 continue; 4245 4246 if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0) 4247 goto nomem; 4248 if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET, 4249 zb[i].zb_objset) != 0) { 4250 nvlist_free(nv); 4251 goto nomem; 4252 } 4253 if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT, 4254 zb[i].zb_object) != 0) { 4255 nvlist_free(nv); 4256 goto nomem; 4257 } 4258 if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) { 4259 nvlist_free(nv); 4260 goto nomem; 4261 } 4262 nvlist_free(nv); 4263 } 4264 4265 free((void *)(uintptr_t)zc.zc_nvlist_dst); 4266 return (0); 4267 4268 nomem: 4269 free((void *)(uintptr_t)zc.zc_nvlist_dst); 4270 return (no_memory(zhp->zpool_hdl)); 4271 } 4272 4273 /* 4274 * Upgrade a ZFS pool to the latest on-disk version. 4275 */ 4276 int 4277 zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version) 4278 { 4279 zfs_cmd_t zc = { 0 }; 4280 libzfs_handle_t *hdl = zhp->zpool_hdl; 4281 4282 (void) strcpy(zc.zc_name, zhp->zpool_name); 4283 zc.zc_cookie = new_version; 4284 4285 if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0) 4286 return (zpool_standard_error_fmt(hdl, errno, 4287 dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"), 4288 zhp->zpool_name)); 4289 return (0); 4290 } 4291 4292 void 4293 zfs_save_arguments(int argc, char **argv, char *string, int len) 4294 { 4295 (void) strlcpy(string, basename(argv[0]), len); 4296 for (int i = 1; i < argc; i++) { 4297 (void) strlcat(string, " ", len); 4298 (void) strlcat(string, argv[i], len); 4299 } 4300 } 4301 4302 int 4303 zpool_log_history(libzfs_handle_t *hdl, const char *message) 4304 { 4305 zfs_cmd_t zc = { 0 }; 4306 nvlist_t *args; 4307 int err; 4308 4309 args = fnvlist_alloc(); 4310 fnvlist_add_string(args, "message", message); 4311 err = zcmd_write_src_nvlist(hdl, &zc, args); 4312 if (err == 0) 4313 err = ioctl(hdl->libzfs_fd, ZFS_IOC_LOG_HISTORY, &zc); 4314 nvlist_free(args); 4315 zcmd_free_nvlists(&zc); 4316 return (err); 4317 } 4318 4319 /* 4320 * Perform ioctl to get some command history of a pool. 4321 * 4322 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the 4323 * logical offset of the history buffer to start reading from. 4324 * 4325 * Upon return, 'off' is the next logical offset to read from and 4326 * 'len' is the actual amount of bytes read into 'buf'. 4327 */ 4328 static int 4329 get_history(zpool_handle_t *zhp, char *buf, uint64_t *off, uint64_t *len) 4330 { 4331 zfs_cmd_t zc = { 0 }; 4332 libzfs_handle_t *hdl = zhp->zpool_hdl; 4333 4334 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 4335 4336 zc.zc_history = (uint64_t)(uintptr_t)buf; 4337 zc.zc_history_len = *len; 4338 zc.zc_history_offset = *off; 4339 4340 if (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) { 4341 switch (errno) { 4342 case EPERM: 4343 return (zfs_error_fmt(hdl, EZFS_PERM, 4344 dgettext(TEXT_DOMAIN, 4345 "cannot show history for pool '%s'"), 4346 zhp->zpool_name)); 4347 case ENOENT: 4348 return (zfs_error_fmt(hdl, EZFS_NOHISTORY, 4349 dgettext(TEXT_DOMAIN, "cannot get history for pool " 4350 "'%s'"), zhp->zpool_name)); 4351 case ENOTSUP: 4352 return (zfs_error_fmt(hdl, EZFS_BADVERSION, 4353 dgettext(TEXT_DOMAIN, "cannot get history for pool " 4354 "'%s', pool must be upgraded"), zhp->zpool_name)); 4355 default: 4356 return (zpool_standard_error_fmt(hdl, errno, 4357 dgettext(TEXT_DOMAIN, 4358 "cannot get history for '%s'"), zhp->zpool_name)); 4359 } 4360 } 4361 4362 *len = zc.zc_history_len; 4363 *off = zc.zc_history_offset; 4364 4365 return (0); 4366 } 4367 4368 /* 4369 * Retrieve the command history of a pool. 4370 */ 4371 int 4372 zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp, uint64_t *off, 4373 boolean_t *eof) 4374 { 4375 char *buf; 4376 int buflen = 128 * 1024; 4377 nvlist_t **records = NULL; 4378 uint_t numrecords = 0; 4379 int err = 0, i; 4380 uint64_t start = *off; 4381 4382 buf = malloc(buflen); 4383 if (buf == NULL) 4384 return (ENOMEM); 4385 /* process about 1MB a time */ 4386 while (*off - start < 1024 * 1024) { 4387 uint64_t bytes_read = buflen; 4388 uint64_t leftover; 4389 4390 if ((err = get_history(zhp, buf, off, &bytes_read)) != 0) 4391 break; 4392 4393 /* if nothing else was read in, we're at EOF, just return */ 4394 if (!bytes_read) { 4395 *eof = B_TRUE; 4396 break; 4397 } 4398 4399 if ((err = zpool_history_unpack(buf, bytes_read, 4400 &leftover, &records, &numrecords)) != 0) 4401 break; 4402 *off -= leftover; 4403 if (leftover == bytes_read) { 4404 /* 4405 * no progress made, because buffer is not big enough 4406 * to hold this record; resize and retry. 4407 */ 4408 buflen *= 2; 4409 free(buf); 4410 buf = malloc(buflen); 4411 if (buf == NULL) 4412 return (ENOMEM); 4413 } 4414 } 4415 4416 free(buf); 4417 4418 if (!err) { 4419 verify(nvlist_alloc(nvhisp, NV_UNIQUE_NAME, 0) == 0); 4420 verify(nvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD, 4421 records, numrecords) == 0); 4422 } 4423 for (i = 0; i < numrecords; i++) 4424 nvlist_free(records[i]); 4425 free(records); 4426 4427 return (err); 4428 } 4429 4430 void 4431 zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj, 4432 char *pathname, size_t len) 4433 { 4434 zfs_cmd_t zc = { 0 }; 4435 boolean_t mounted = B_FALSE; 4436 char *mntpnt = NULL; 4437 char dsname[ZFS_MAX_DATASET_NAME_LEN]; 4438 4439 if (dsobj == 0) { 4440 /* special case for the MOS */ 4441 (void) snprintf(pathname, len, "<metadata>:<0x%llx>", obj); 4442 return; 4443 } 4444 4445 /* get the dataset's name */ 4446 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 4447 zc.zc_obj = dsobj; 4448 if (ioctl(zhp->zpool_hdl->libzfs_fd, 4449 ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) { 4450 /* just write out a path of two object numbers */ 4451 (void) snprintf(pathname, len, "<0x%llx>:<0x%llx>", 4452 dsobj, obj); 4453 return; 4454 } 4455 (void) strlcpy(dsname, zc.zc_value, sizeof (dsname)); 4456 4457 /* find out if the dataset is mounted */ 4458 mounted = is_mounted(zhp->zpool_hdl, dsname, &mntpnt); 4459 4460 /* get the corrupted object's path */ 4461 (void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name)); 4462 zc.zc_obj = obj; 4463 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_OBJ_TO_PATH, 4464 &zc) == 0) { 4465 if (mounted) { 4466 (void) snprintf(pathname, len, "%s%s", mntpnt, 4467 zc.zc_value); 4468 } else { 4469 (void) snprintf(pathname, len, "%s:%s", 4470 dsname, zc.zc_value); 4471 } 4472 } else { 4473 (void) snprintf(pathname, len, "%s:<0x%llx>", dsname, obj); 4474 } 4475 free(mntpnt); 4476 } 4477 4478 int 4479 zpool_set_bootenv(zpool_handle_t *zhp, const nvlist_t *envmap) 4480 { 4481 int error = lzc_set_bootenv(zhp->zpool_name, envmap); 4482 if (error != 0) { 4483 (void) zpool_standard_error_fmt(zhp->zpool_hdl, error, 4484 dgettext(TEXT_DOMAIN, 4485 "error setting bootenv in pool '%s'"), zhp->zpool_name); 4486 } 4487 4488 return (error); 4489 } 4490 4491 int 4492 zpool_get_bootenv(zpool_handle_t *zhp, nvlist_t **nvlp) 4493 { 4494 nvlist_t *nvl; 4495 int error; 4496 4497 nvl = NULL; 4498 error = lzc_get_bootenv(zhp->zpool_name, &nvl); 4499 if (error != 0) { 4500 (void) zpool_standard_error_fmt(zhp->zpool_hdl, error, 4501 dgettext(TEXT_DOMAIN, 4502 "error getting bootenv in pool '%s'"), zhp->zpool_name); 4503 } else { 4504 *nvlp = nvl; 4505 } 4506 4507 return (error); 4508 } 4509 4510 /* 4511 * Read the EFI label from the config, if a label does not exist then 4512 * pass back the error to the caller. If the caller has passed a non-NULL 4513 * diskaddr argument then we set it to the starting address of the EFI 4514 * partition. If the caller has passed a non-NULL boolean argument, then 4515 * we set it to indicate if the disk does have efi system partition. 4516 */ 4517 static int 4518 read_efi_label(nvlist_t *config, diskaddr_t *sb, boolean_t *system) 4519 { 4520 char *path; 4521 int fd; 4522 char diskname[MAXPATHLEN]; 4523 boolean_t boot = B_FALSE; 4524 int err = -1; 4525 int slice; 4526 4527 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PATH, &path) != 0) 4528 return (err); 4529 4530 (void) snprintf(diskname, sizeof (diskname), "%s%s", ZFS_RDISK_ROOT, 4531 strrchr(path, '/')); 4532 if ((fd = open(diskname, O_RDONLY|O_NDELAY)) >= 0) { 4533 struct dk_gpt *vtoc; 4534 4535 if ((err = efi_alloc_and_read(fd, &vtoc)) >= 0) { 4536 for (slice = 0; slice < vtoc->efi_nparts; slice++) { 4537 if (vtoc->efi_parts[slice].p_tag == V_SYSTEM) 4538 boot = B_TRUE; 4539 if (vtoc->efi_parts[slice].p_tag == V_USR) 4540 break; 4541 } 4542 if (sb != NULL && vtoc->efi_parts[slice].p_tag == V_USR) 4543 *sb = vtoc->efi_parts[slice].p_start; 4544 if (system != NULL) 4545 *system = boot; 4546 efi_free(vtoc); 4547 } 4548 (void) close(fd); 4549 } 4550 return (err); 4551 } 4552 4553 /* 4554 * determine where a partition starts on a disk in the current 4555 * configuration 4556 */ 4557 static diskaddr_t 4558 find_start_block(nvlist_t *config) 4559 { 4560 nvlist_t **child; 4561 uint_t c, children; 4562 diskaddr_t sb = MAXOFFSET_T; 4563 uint64_t wholedisk; 4564 4565 if (nvlist_lookup_nvlist_array(config, 4566 ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) { 4567 if (nvlist_lookup_uint64(config, 4568 ZPOOL_CONFIG_WHOLE_DISK, 4569 &wholedisk) != 0 || !wholedisk) { 4570 return (MAXOFFSET_T); 4571 } 4572 if (read_efi_label(config, &sb, NULL) < 0) 4573 sb = MAXOFFSET_T; 4574 return (sb); 4575 } 4576 4577 for (c = 0; c < children; c++) { 4578 sb = find_start_block(child[c]); 4579 if (sb != MAXOFFSET_T) { 4580 return (sb); 4581 } 4582 } 4583 return (MAXOFFSET_T); 4584 } 4585 4586 /* 4587 * Label an individual disk. The name provided is the short name, 4588 * stripped of any leading /dev path. 4589 */ 4590 int 4591 zpool_label_disk(libzfs_handle_t *hdl, zpool_handle_t *zhp, const char *name, 4592 zpool_boot_label_t boot_type, uint64_t boot_size, int *slice) 4593 { 4594 char path[MAXPATHLEN]; 4595 struct dk_gpt *vtoc; 4596 int fd; 4597 size_t resv; 4598 uint64_t slice_size; 4599 diskaddr_t start_block; 4600 char errbuf[1024]; 4601 4602 /* prepare an error message just in case */ 4603 (void) snprintf(errbuf, sizeof (errbuf), 4604 dgettext(TEXT_DOMAIN, "cannot label '%s'"), name); 4605 4606 if (zhp) { 4607 nvlist_t *nvroot; 4608 4609 verify(nvlist_lookup_nvlist(zhp->zpool_config, 4610 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 4611 4612 if (zhp->zpool_start_block == 0) 4613 start_block = find_start_block(nvroot); 4614 else 4615 start_block = zhp->zpool_start_block; 4616 zhp->zpool_start_block = start_block; 4617 } else { 4618 /* new pool */ 4619 start_block = NEW_START_BLOCK; 4620 } 4621 4622 (void) snprintf(path, sizeof (path), "%s/%s%s", ZFS_RDISK_ROOT, name, 4623 BACKUP_SLICE); 4624 4625 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) { 4626 /* 4627 * This shouldn't happen. We've long since verified that this 4628 * is a valid device. 4629 */ 4630 zfs_error_aux(hdl, 4631 dgettext(TEXT_DOMAIN, "unable to open device")); 4632 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf)); 4633 } 4634 4635 if (efi_alloc_and_init(fd, EFI_NUMPAR, &vtoc) != 0) { 4636 /* 4637 * The only way this can fail is if we run out of memory, or we 4638 * were unable to read the disk's capacity 4639 */ 4640 if (errno == ENOMEM) 4641 (void) no_memory(hdl); 4642 4643 (void) close(fd); 4644 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4645 "unable to read disk capacity"), name); 4646 4647 return (zfs_error(hdl, EZFS_NOCAP, errbuf)); 4648 } 4649 resv = efi_reserved_sectors(vtoc); 4650 4651 /* 4652 * Why we use V_USR: V_BACKUP confuses users, and is considered 4653 * disposable by some EFI utilities (since EFI doesn't have a backup 4654 * slice). V_UNASSIGNED is supposed to be used only for zero size 4655 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT, 4656 * etc. were all pretty specific. V_USR is as close to reality as we 4657 * can get, in the absence of V_OTHER. 4658 */ 4659 /* first fix the partition start block */ 4660 if (start_block == MAXOFFSET_T) 4661 start_block = NEW_START_BLOCK; 4662 4663 /* 4664 * EFI System partition is using slice 0. 4665 * ZFS is on slice 1 and slice 8 is reserved. 4666 * We assume the GPT partition table without system 4667 * partition has zfs p_start == NEW_START_BLOCK. 4668 * If start_block != NEW_START_BLOCK, it means we have 4669 * system partition. Correct solution would be to query/cache vtoc 4670 * from existing vdev member. 4671 */ 4672 if (boot_type == ZPOOL_CREATE_BOOT_LABEL) { 4673 if (boot_size % vtoc->efi_lbasize != 0) { 4674 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4675 "boot partition size must be a multiple of %d"), 4676 vtoc->efi_lbasize); 4677 (void) close(fd); 4678 efi_free(vtoc); 4679 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf)); 4680 } 4681 /* 4682 * System partition size checks. 4683 * Note the 1MB is quite arbitrary value, since we 4684 * are creating dedicated pool, it should be enough 4685 * to hold fat + efi bootloader. May need to be 4686 * adjusted if the bootloader size will grow. 4687 */ 4688 if (boot_size < 1024 * 1024) { 4689 char buf[64]; 4690 zfs_nicenum(boot_size, buf, sizeof (buf)); 4691 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4692 "Specified size %s for EFI System partition is too " 4693 "small, the minimum size is 1MB."), buf); 4694 (void) close(fd); 4695 efi_free(vtoc); 4696 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf)); 4697 } 4698 /* 33MB is tested with mkfs -F pcfs */ 4699 if (hdl->libzfs_printerr && 4700 ((vtoc->efi_lbasize == 512 && 4701 boot_size < 33 * 1024 * 1024) || 4702 (vtoc->efi_lbasize == 4096 && 4703 boot_size < 256 * 1024 * 1024))) { 4704 char buf[64]; 4705 zfs_nicenum(boot_size, buf, sizeof (buf)); 4706 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, 4707 "Warning: EFI System partition size %s is " 4708 "not allowing to create FAT32 file\nsystem, which " 4709 "may result in unbootable system.\n"), buf); 4710 } 4711 /* Adjust zfs partition start by size of system partition. */ 4712 start_block += boot_size / vtoc->efi_lbasize; 4713 } 4714 4715 if (start_block == NEW_START_BLOCK) { 4716 /* 4717 * Use default layout. 4718 * ZFS is on slice 0 and slice 8 is reserved. 4719 */ 4720 slice_size = vtoc->efi_last_u_lba + 1; 4721 slice_size -= resv; 4722 slice_size -= start_block; 4723 if (slice != NULL) 4724 *slice = 0; 4725 4726 vtoc->efi_parts[0].p_start = start_block; 4727 vtoc->efi_parts[0].p_size = slice_size; 4728 4729 vtoc->efi_parts[0].p_tag = V_USR; 4730 (void) strcpy(vtoc->efi_parts[0].p_name, "zfs"); 4731 4732 vtoc->efi_parts[8].p_start = slice_size + start_block; 4733 vtoc->efi_parts[8].p_size = resv; 4734 vtoc->efi_parts[8].p_tag = V_RESERVED; 4735 } else { 4736 slice_size = start_block - NEW_START_BLOCK; 4737 vtoc->efi_parts[0].p_start = NEW_START_BLOCK; 4738 vtoc->efi_parts[0].p_size = slice_size; 4739 vtoc->efi_parts[0].p_tag = V_SYSTEM; 4740 (void) strcpy(vtoc->efi_parts[0].p_name, "loader"); 4741 if (slice != NULL) 4742 *slice = 1; 4743 /* prepare slice 1 */ 4744 slice_size = vtoc->efi_last_u_lba + 1 - slice_size; 4745 slice_size -= resv; 4746 slice_size -= NEW_START_BLOCK; 4747 vtoc->efi_parts[1].p_start = start_block; 4748 vtoc->efi_parts[1].p_size = slice_size; 4749 vtoc->efi_parts[1].p_tag = V_USR; 4750 (void) strcpy(vtoc->efi_parts[1].p_name, "zfs"); 4751 4752 vtoc->efi_parts[8].p_start = slice_size + start_block; 4753 vtoc->efi_parts[8].p_size = resv; 4754 vtoc->efi_parts[8].p_tag = V_RESERVED; 4755 } 4756 4757 if (efi_write(fd, vtoc) != 0) { 4758 /* 4759 * Some block drivers (like pcata) may not support EFI 4760 * GPT labels. Print out a helpful error message dir- 4761 * ecting the user to manually label the disk and give 4762 * a specific slice. 4763 */ 4764 (void) close(fd); 4765 efi_free(vtoc); 4766 4767 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4768 "try using fdisk(8) and then provide a specific slice")); 4769 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf)); 4770 } 4771 4772 (void) close(fd); 4773 efi_free(vtoc); 4774 return (0); 4775 } 4776 4777 static boolean_t 4778 supported_dump_vdev_type(libzfs_handle_t *hdl, nvlist_t *config, char *errbuf) 4779 { 4780 char *type; 4781 nvlist_t **child; 4782 uint_t children, c; 4783 4784 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_TYPE, &type) == 0); 4785 if (strcmp(type, VDEV_TYPE_FILE) == 0 || 4786 strcmp(type, VDEV_TYPE_HOLE) == 0 || 4787 strcmp(type, VDEV_TYPE_MISSING) == 0) { 4788 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4789 "vdev type '%s' is not supported"), type); 4790 (void) zfs_error(hdl, EZFS_VDEVNOTSUP, errbuf); 4791 return (B_FALSE); 4792 } 4793 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN, 4794 &child, &children) == 0) { 4795 for (c = 0; c < children; c++) { 4796 if (!supported_dump_vdev_type(hdl, child[c], errbuf)) 4797 return (B_FALSE); 4798 } 4799 } 4800 return (B_TRUE); 4801 } 4802 4803 /* 4804 * Check if this zvol is allowable for use as a dump device; zero if 4805 * it is, > 0 if it isn't, < 0 if it isn't a zvol. 4806 * 4807 * Allowable storage configurations include mirrors, all raidz variants, and 4808 * pools with log, cache, and spare devices. Pools which are backed by files or 4809 * have missing/hole vdevs are not suitable. 4810 */ 4811 int 4812 zvol_check_dump_config(char *arg) 4813 { 4814 zpool_handle_t *zhp = NULL; 4815 nvlist_t *config, *nvroot; 4816 char *p, *volname; 4817 nvlist_t **top; 4818 uint_t toplevels; 4819 libzfs_handle_t *hdl; 4820 char errbuf[1024]; 4821 char poolname[ZFS_MAX_DATASET_NAME_LEN]; 4822 int pathlen = strlen(ZVOL_FULL_DEV_DIR); 4823 int ret = 1; 4824 4825 if (strncmp(arg, ZVOL_FULL_DEV_DIR, pathlen)) { 4826 return (-1); 4827 } 4828 4829 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, 4830 "dump is not supported on device '%s'"), arg); 4831 4832 if ((hdl = libzfs_init()) == NULL) 4833 return (1); 4834 libzfs_print_on_error(hdl, B_TRUE); 4835 4836 volname = arg + pathlen; 4837 4838 /* check the configuration of the pool */ 4839 if ((p = strchr(volname, '/')) == NULL) { 4840 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4841 "malformed dataset name")); 4842 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf); 4843 return (1); 4844 } else if (p - volname >= ZFS_MAX_DATASET_NAME_LEN) { 4845 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4846 "dataset name is too long")); 4847 (void) zfs_error(hdl, EZFS_NAMETOOLONG, errbuf); 4848 return (1); 4849 } else { 4850 (void) strncpy(poolname, volname, p - volname); 4851 poolname[p - volname] = '\0'; 4852 } 4853 4854 if ((zhp = zpool_open(hdl, poolname)) == NULL) { 4855 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4856 "could not open pool '%s'"), poolname); 4857 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf); 4858 goto out; 4859 } 4860 config = zpool_get_config(zhp, NULL); 4861 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 4862 &nvroot) != 0) { 4863 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4864 "could not obtain vdev configuration for '%s'"), poolname); 4865 (void) zfs_error(hdl, EZFS_INVALCONFIG, errbuf); 4866 goto out; 4867 } 4868 4869 verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, 4870 &top, &toplevels) == 0); 4871 4872 if (!supported_dump_vdev_type(hdl, top[0], errbuf)) { 4873 goto out; 4874 } 4875 ret = 0; 4876 4877 out: 4878 if (zhp) 4879 zpool_close(zhp); 4880 libzfs_fini(hdl); 4881 return (ret); 4882 } 4883