libzfs_pool.c revision 332525
1/* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 22/* 23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 24 * Copyright (c) 2011, 2017 by Delphix. All rights reserved. 25 * Copyright (c) 2013, Joyent, Inc. All rights reserved. 26 * Copyright 2016 Nexenta Systems, Inc. 27 * Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com> 28 * Copyright (c) 2017 Datto Inc. 29 */ 30 31#include <sys/types.h> 32#include <sys/stat.h> 33#include <ctype.h> 34#include <errno.h> 35#include <devid.h> 36#include <fcntl.h> 37#include <libintl.h> 38#include <stdio.h> 39#include <stdlib.h> 40#include <strings.h> 41#include <unistd.h> 42#include <libgen.h> 43#include <sys/zfs_ioctl.h> 44#include <dlfcn.h> 45 46#include "zfs_namecheck.h" 47#include "zfs_prop.h" 48#include "libzfs_impl.h" 49#include "zfs_comutil.h" 50#include "zfeature_common.h" 51 52static int read_efi_label(nvlist_t *, diskaddr_t *, boolean_t *); 53static boolean_t zpool_vdev_is_interior(const char *name); 54 55#define BACKUP_SLICE "s2" 56 57typedef struct prop_flags { 58 int create:1; /* Validate property on creation */ 59 int import:1; /* Validate property on import */ 60} prop_flags_t; 61 62/* 63 * ==================================================================== 64 * zpool property functions 65 * ==================================================================== 66 */ 67 68static int 69zpool_get_all_props(zpool_handle_t *zhp) 70{ 71 zfs_cmd_t zc = { 0 }; 72 libzfs_handle_t *hdl = zhp->zpool_hdl; 73 74 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 75 76 if (zcmd_alloc_dst_nvlist(hdl, &zc, 0) != 0) 77 return (-1); 78 79 while (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_PROPS, &zc) != 0) { 80 if (errno == ENOMEM) { 81 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) { 82 zcmd_free_nvlists(&zc); 83 return (-1); 84 } 85 } else { 86 zcmd_free_nvlists(&zc); 87 return (-1); 88 } 89 } 90 91 if (zcmd_read_dst_nvlist(hdl, &zc, &zhp->zpool_props) != 0) { 92 zcmd_free_nvlists(&zc); 93 return (-1); 94 } 95 96 zcmd_free_nvlists(&zc); 97 98 return (0); 99} 100 101static int 102zpool_props_refresh(zpool_handle_t *zhp) 103{ 104 nvlist_t *old_props; 105 106 old_props = zhp->zpool_props; 107 108 if (zpool_get_all_props(zhp) != 0) 109 return (-1); 110 111 nvlist_free(old_props); 112 return (0); 113} 114 115static char * 116zpool_get_prop_string(zpool_handle_t *zhp, zpool_prop_t prop, 117 zprop_source_t *src) 118{ 119 nvlist_t *nv, *nvl; 120 uint64_t ival; 121 char *value; 122 zprop_source_t source; 123 124 nvl = zhp->zpool_props; 125 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) { 126 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &ival) == 0); 127 source = ival; 128 verify(nvlist_lookup_string(nv, ZPROP_VALUE, &value) == 0); 129 } else { 130 source = ZPROP_SRC_DEFAULT; 131 if ((value = (char *)zpool_prop_default_string(prop)) == NULL) 132 value = "-"; 133 } 134 135 if (src) 136 *src = source; 137 138 return (value); 139} 140 141uint64_t 142zpool_get_prop_int(zpool_handle_t *zhp, zpool_prop_t prop, zprop_source_t *src) 143{ 144 nvlist_t *nv, *nvl; 145 uint64_t value; 146 zprop_source_t source; 147 148 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp)) { 149 /* 150 * zpool_get_all_props() has most likely failed because 151 * the pool is faulted, but if all we need is the top level 152 * vdev's guid then get it from the zhp config nvlist. 153 */ 154 if ((prop == ZPOOL_PROP_GUID) && 155 (nvlist_lookup_nvlist(zhp->zpool_config, 156 ZPOOL_CONFIG_VDEV_TREE, &nv) == 0) && 157 (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value) 158 == 0)) { 159 return (value); 160 } 161 return (zpool_prop_default_numeric(prop)); 162 } 163 164 nvl = zhp->zpool_props; 165 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) { 166 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &value) == 0); 167 source = value; 168 verify(nvlist_lookup_uint64(nv, ZPROP_VALUE, &value) == 0); 169 } else { 170 source = ZPROP_SRC_DEFAULT; 171 value = zpool_prop_default_numeric(prop); 172 } 173 174 if (src) 175 *src = source; 176 177 return (value); 178} 179 180/* 181 * Map VDEV STATE to printed strings. 182 */ 183const char * 184zpool_state_to_name(vdev_state_t state, vdev_aux_t aux) 185{ 186 switch (state) { 187 case VDEV_STATE_CLOSED: 188 case VDEV_STATE_OFFLINE: 189 return (gettext("OFFLINE")); 190 case VDEV_STATE_REMOVED: 191 return (gettext("REMOVED")); 192 case VDEV_STATE_CANT_OPEN: 193 if (aux == VDEV_AUX_CORRUPT_DATA || aux == VDEV_AUX_BAD_LOG) 194 return (gettext("FAULTED")); 195 else if (aux == VDEV_AUX_SPLIT_POOL) 196 return (gettext("SPLIT")); 197 else 198 return (gettext("UNAVAIL")); 199 case VDEV_STATE_FAULTED: 200 return (gettext("FAULTED")); 201 case VDEV_STATE_DEGRADED: 202 return (gettext("DEGRADED")); 203 case VDEV_STATE_HEALTHY: 204 return (gettext("ONLINE")); 205 206 default: 207 break; 208 } 209 210 return (gettext("UNKNOWN")); 211} 212 213/* 214 * Map POOL STATE to printed strings. 215 */ 216const char * 217zpool_pool_state_to_name(pool_state_t state) 218{ 219 switch (state) { 220 case POOL_STATE_ACTIVE: 221 return (gettext("ACTIVE")); 222 case POOL_STATE_EXPORTED: 223 return (gettext("EXPORTED")); 224 case POOL_STATE_DESTROYED: 225 return (gettext("DESTROYED")); 226 case POOL_STATE_SPARE: 227 return (gettext("SPARE")); 228 case POOL_STATE_L2CACHE: 229 return (gettext("L2CACHE")); 230 case POOL_STATE_UNINITIALIZED: 231 return (gettext("UNINITIALIZED")); 232 case POOL_STATE_UNAVAIL: 233 return (gettext("UNAVAIL")); 234 case POOL_STATE_POTENTIALLY_ACTIVE: 235 return (gettext("POTENTIALLY_ACTIVE")); 236 } 237 238 return (gettext("UNKNOWN")); 239} 240 241/* 242 * Get a zpool property value for 'prop' and return the value in 243 * a pre-allocated buffer. 244 */ 245int 246zpool_get_prop(zpool_handle_t *zhp, zpool_prop_t prop, char *buf, size_t len, 247 zprop_source_t *srctype, boolean_t literal) 248{ 249 uint64_t intval; 250 const char *strval; 251 zprop_source_t src = ZPROP_SRC_NONE; 252 nvlist_t *nvroot; 253 vdev_stat_t *vs; 254 uint_t vsc; 255 256 if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) { 257 switch (prop) { 258 case ZPOOL_PROP_NAME: 259 (void) strlcpy(buf, zpool_get_name(zhp), len); 260 break; 261 262 case ZPOOL_PROP_HEALTH: 263 (void) strlcpy(buf, 264 zpool_pool_state_to_name(POOL_STATE_UNAVAIL), len); 265 break; 266 267 case ZPOOL_PROP_GUID: 268 intval = zpool_get_prop_int(zhp, prop, &src); 269 (void) snprintf(buf, len, "%llu", intval); 270 break; 271 272 case ZPOOL_PROP_ALTROOT: 273 case ZPOOL_PROP_CACHEFILE: 274 case ZPOOL_PROP_COMMENT: 275 if (zhp->zpool_props != NULL || 276 zpool_get_all_props(zhp) == 0) { 277 (void) strlcpy(buf, 278 zpool_get_prop_string(zhp, prop, &src), 279 len); 280 break; 281 } 282 /* FALLTHROUGH */ 283 default: 284 (void) strlcpy(buf, "-", len); 285 break; 286 } 287 288 if (srctype != NULL) 289 *srctype = src; 290 return (0); 291 } 292 293 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp) && 294 prop != ZPOOL_PROP_NAME) 295 return (-1); 296 297 switch (zpool_prop_get_type(prop)) { 298 case PROP_TYPE_STRING: 299 (void) strlcpy(buf, zpool_get_prop_string(zhp, prop, &src), 300 len); 301 break; 302 303 case PROP_TYPE_NUMBER: 304 intval = zpool_get_prop_int(zhp, prop, &src); 305 306 switch (prop) { 307 case ZPOOL_PROP_SIZE: 308 case ZPOOL_PROP_ALLOCATED: 309 case ZPOOL_PROP_FREE: 310 case ZPOOL_PROP_FREEING: 311 case ZPOOL_PROP_LEAKED: 312 if (literal) { 313 (void) snprintf(buf, len, "%llu", 314 (u_longlong_t)intval); 315 } else { 316 (void) zfs_nicenum(intval, buf, len); 317 } 318 break; 319 case ZPOOL_PROP_BOOTSIZE: 320 case ZPOOL_PROP_EXPANDSZ: 321 if (intval == 0) { 322 (void) strlcpy(buf, "-", len); 323 } else if (literal) { 324 (void) snprintf(buf, len, "%llu", 325 (u_longlong_t)intval); 326 } else { 327 (void) zfs_nicenum(intval, buf, len); 328 } 329 break; 330 case ZPOOL_PROP_CAPACITY: 331 if (literal) { 332 (void) snprintf(buf, len, "%llu", 333 (u_longlong_t)intval); 334 } else { 335 (void) snprintf(buf, len, "%llu%%", 336 (u_longlong_t)intval); 337 } 338 break; 339 case ZPOOL_PROP_FRAGMENTATION: 340 if (intval == UINT64_MAX) { 341 (void) strlcpy(buf, "-", len); 342 } else { 343 (void) snprintf(buf, len, "%llu%%", 344 (u_longlong_t)intval); 345 } 346 break; 347 case ZPOOL_PROP_DEDUPRATIO: 348 (void) snprintf(buf, len, "%llu.%02llux", 349 (u_longlong_t)(intval / 100), 350 (u_longlong_t)(intval % 100)); 351 break; 352 case ZPOOL_PROP_HEALTH: 353 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL), 354 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 355 verify(nvlist_lookup_uint64_array(nvroot, 356 ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc) 357 == 0); 358 359 (void) strlcpy(buf, zpool_state_to_name(intval, 360 vs->vs_aux), len); 361 break; 362 case ZPOOL_PROP_VERSION: 363 if (intval >= SPA_VERSION_FEATURES) { 364 (void) snprintf(buf, len, "-"); 365 break; 366 } 367 /* FALLTHROUGH */ 368 default: 369 (void) snprintf(buf, len, "%llu", intval); 370 } 371 break; 372 373 case PROP_TYPE_INDEX: 374 intval = zpool_get_prop_int(zhp, prop, &src); 375 if (zpool_prop_index_to_string(prop, intval, &strval) 376 != 0) 377 return (-1); 378 (void) strlcpy(buf, strval, len); 379 break; 380 381 default: 382 abort(); 383 } 384 385 if (srctype) 386 *srctype = src; 387 388 return (0); 389} 390 391/* 392 * Check if the bootfs name has the same pool name as it is set to. 393 * Assuming bootfs is a valid dataset name. 394 */ 395static boolean_t 396bootfs_name_valid(const char *pool, char *bootfs) 397{ 398 int len = strlen(pool); 399 400 if (!zfs_name_valid(bootfs, ZFS_TYPE_FILESYSTEM|ZFS_TYPE_SNAPSHOT)) 401 return (B_FALSE); 402 403 if (strncmp(pool, bootfs, len) == 0 && 404 (bootfs[len] == '/' || bootfs[len] == '\0')) 405 return (B_TRUE); 406 407 return (B_FALSE); 408} 409 410boolean_t 411zpool_is_bootable(zpool_handle_t *zhp) 412{ 413 char bootfs[ZFS_MAX_DATASET_NAME_LEN]; 414 415 return (zpool_get_prop(zhp, ZPOOL_PROP_BOOTFS, bootfs, 416 sizeof (bootfs), NULL, B_FALSE) == 0 && strncmp(bootfs, "-", 417 sizeof (bootfs)) != 0); 418} 419 420 421/* 422 * Given an nvlist of zpool properties to be set, validate that they are 423 * correct, and parse any numeric properties (index, boolean, etc) if they are 424 * specified as strings. 425 */ 426static nvlist_t * 427zpool_valid_proplist(libzfs_handle_t *hdl, const char *poolname, 428 nvlist_t *props, uint64_t version, prop_flags_t flags, char *errbuf) 429{ 430 nvpair_t *elem; 431 nvlist_t *retprops; 432 zpool_prop_t prop; 433 char *strval; 434 uint64_t intval; 435 char *slash, *check; 436 struct stat64 statbuf; 437 zpool_handle_t *zhp; 438 439 if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) { 440 (void) no_memory(hdl); 441 return (NULL); 442 } 443 444 elem = NULL; 445 while ((elem = nvlist_next_nvpair(props, elem)) != NULL) { 446 const char *propname = nvpair_name(elem); 447 448 prop = zpool_name_to_prop(propname); 449 if (prop == ZPOOL_PROP_INVAL && zpool_prop_feature(propname)) { 450 int err; 451 char *fname = strchr(propname, '@') + 1; 452 453 err = zfeature_lookup_name(fname, NULL); 454 if (err != 0) { 455 ASSERT3U(err, ==, ENOENT); 456 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 457 "invalid feature '%s'"), fname); 458 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 459 goto error; 460 } 461 462 if (nvpair_type(elem) != DATA_TYPE_STRING) { 463 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 464 "'%s' must be a string"), propname); 465 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 466 goto error; 467 } 468 469 (void) nvpair_value_string(elem, &strval); 470 if (strcmp(strval, ZFS_FEATURE_ENABLED) != 0) { 471 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 472 "property '%s' can only be set to " 473 "'enabled'"), propname); 474 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 475 goto error; 476 } 477 478 if (nvlist_add_uint64(retprops, propname, 0) != 0) { 479 (void) no_memory(hdl); 480 goto error; 481 } 482 continue; 483 } 484 485 /* 486 * Make sure this property is valid and applies to this type. 487 */ 488 if (prop == ZPOOL_PROP_INVAL) { 489 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 490 "invalid property '%s'"), propname); 491 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 492 goto error; 493 } 494 495 if (zpool_prop_readonly(prop)) { 496 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' " 497 "is readonly"), propname); 498 (void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf); 499 goto error; 500 } 501 502 if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_POOL, retprops, 503 &strval, &intval, errbuf) != 0) 504 goto error; 505 506 /* 507 * Perform additional checking for specific properties. 508 */ 509 switch (prop) { 510 case ZPOOL_PROP_VERSION: 511 if (intval < version || 512 !SPA_VERSION_IS_SUPPORTED(intval)) { 513 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 514 "property '%s' number %d is invalid."), 515 propname, intval); 516 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf); 517 goto error; 518 } 519 break; 520 521 case ZPOOL_PROP_BOOTSIZE: 522 if (!flags.create) { 523 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 524 "property '%s' can only be set during pool " 525 "creation"), propname); 526 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 527 goto error; 528 } 529 break; 530 531 case ZPOOL_PROP_BOOTFS: 532 if (flags.create || flags.import) { 533 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 534 "property '%s' cannot be set at creation " 535 "or import time"), propname); 536 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 537 goto error; 538 } 539 540 if (version < SPA_VERSION_BOOTFS) { 541 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 542 "pool must be upgraded to support " 543 "'%s' property"), propname); 544 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf); 545 goto error; 546 } 547 548 /* 549 * bootfs property value has to be a dataset name and 550 * the dataset has to be in the same pool as it sets to. 551 */ 552 if (strval[0] != '\0' && !bootfs_name_valid(poolname, 553 strval)) { 554 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' " 555 "is an invalid name"), strval); 556 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf); 557 goto error; 558 } 559 560 if ((zhp = zpool_open_canfail(hdl, poolname)) == NULL) { 561 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 562 "could not open pool '%s'"), poolname); 563 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf); 564 goto error; 565 } 566 zpool_close(zhp); 567 break; 568 569 case ZPOOL_PROP_ALTROOT: 570 if (!flags.create && !flags.import) { 571 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 572 "property '%s' can only be set during pool " 573 "creation or import"), propname); 574 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 575 goto error; 576 } 577 578 if (strval[0] != '/') { 579 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 580 "bad alternate root '%s'"), strval); 581 (void) zfs_error(hdl, EZFS_BADPATH, errbuf); 582 goto error; 583 } 584 break; 585 586 case ZPOOL_PROP_CACHEFILE: 587 if (strval[0] == '\0') 588 break; 589 590 if (strcmp(strval, "none") == 0) 591 break; 592 593 if (strval[0] != '/') { 594 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 595 "property '%s' must be empty, an " 596 "absolute path, or 'none'"), propname); 597 (void) zfs_error(hdl, EZFS_BADPATH, errbuf); 598 goto error; 599 } 600 601 slash = strrchr(strval, '/'); 602 603 if (slash[1] == '\0' || strcmp(slash, "/.") == 0 || 604 strcmp(slash, "/..") == 0) { 605 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 606 "'%s' is not a valid file"), strval); 607 (void) zfs_error(hdl, EZFS_BADPATH, errbuf); 608 goto error; 609 } 610 611 *slash = '\0'; 612 613 if (strval[0] != '\0' && 614 (stat64(strval, &statbuf) != 0 || 615 !S_ISDIR(statbuf.st_mode))) { 616 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 617 "'%s' is not a valid directory"), 618 strval); 619 (void) zfs_error(hdl, EZFS_BADPATH, errbuf); 620 goto error; 621 } 622 623 *slash = '/'; 624 break; 625 626 case ZPOOL_PROP_COMMENT: 627 for (check = strval; *check != '\0'; check++) { 628 if (!isprint(*check)) { 629 zfs_error_aux(hdl, 630 dgettext(TEXT_DOMAIN, 631 "comment may only have printable " 632 "characters")); 633 (void) zfs_error(hdl, EZFS_BADPROP, 634 errbuf); 635 goto error; 636 } 637 } 638 if (strlen(strval) > ZPROP_MAX_COMMENT) { 639 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 640 "comment must not exceed %d characters"), 641 ZPROP_MAX_COMMENT); 642 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 643 goto error; 644 } 645 break; 646 case ZPOOL_PROP_READONLY: 647 if (!flags.import) { 648 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 649 "property '%s' can only be set at " 650 "import time"), propname); 651 (void) zfs_error(hdl, EZFS_BADPROP, errbuf); 652 goto error; 653 } 654 break; 655 656 default: 657 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 658 "property '%s'(%d) not defined"), propname, prop); 659 break; 660 } 661 } 662 663 return (retprops); 664error: 665 nvlist_free(retprops); 666 return (NULL); 667} 668 669/* 670 * Set zpool property : propname=propval. 671 */ 672int 673zpool_set_prop(zpool_handle_t *zhp, const char *propname, const char *propval) 674{ 675 zfs_cmd_t zc = { 0 }; 676 int ret = -1; 677 char errbuf[1024]; 678 nvlist_t *nvl = NULL; 679 nvlist_t *realprops; 680 uint64_t version; 681 prop_flags_t flags = { 0 }; 682 683 (void) snprintf(errbuf, sizeof (errbuf), 684 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"), 685 zhp->zpool_name); 686 687 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0) 688 return (no_memory(zhp->zpool_hdl)); 689 690 if (nvlist_add_string(nvl, propname, propval) != 0) { 691 nvlist_free(nvl); 692 return (no_memory(zhp->zpool_hdl)); 693 } 694 695 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL); 696 if ((realprops = zpool_valid_proplist(zhp->zpool_hdl, 697 zhp->zpool_name, nvl, version, flags, errbuf)) == NULL) { 698 nvlist_free(nvl); 699 return (-1); 700 } 701 702 nvlist_free(nvl); 703 nvl = realprops; 704 705 /* 706 * Execute the corresponding ioctl() to set this property. 707 */ 708 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 709 710 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) { 711 nvlist_free(nvl); 712 return (-1); 713 } 714 715 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc); 716 717 zcmd_free_nvlists(&zc); 718 nvlist_free(nvl); 719 720 if (ret) 721 (void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf); 722 else 723 (void) zpool_props_refresh(zhp); 724 725 return (ret); 726} 727 728int 729zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp) 730{ 731 libzfs_handle_t *hdl = zhp->zpool_hdl; 732 zprop_list_t *entry; 733 char buf[ZFS_MAXPROPLEN]; 734 nvlist_t *features = NULL; 735 zprop_list_t **last; 736 boolean_t firstexpand = (NULL == *plp); 737 738 if (zprop_expand_list(hdl, plp, ZFS_TYPE_POOL) != 0) 739 return (-1); 740 741 last = plp; 742 while (*last != NULL) 743 last = &(*last)->pl_next; 744 745 if ((*plp)->pl_all) 746 features = zpool_get_features(zhp); 747 748 if ((*plp)->pl_all && firstexpand) { 749 for (int i = 0; i < SPA_FEATURES; i++) { 750 zprop_list_t *entry = zfs_alloc(hdl, 751 sizeof (zprop_list_t)); 752 entry->pl_prop = ZPROP_INVAL; 753 entry->pl_user_prop = zfs_asprintf(hdl, "feature@%s", 754 spa_feature_table[i].fi_uname); 755 entry->pl_width = strlen(entry->pl_user_prop); 756 entry->pl_all = B_TRUE; 757 758 *last = entry; 759 last = &entry->pl_next; 760 } 761 } 762 763 /* add any unsupported features */ 764 for (nvpair_t *nvp = nvlist_next_nvpair(features, NULL); 765 nvp != NULL; nvp = nvlist_next_nvpair(features, nvp)) { 766 char *propname; 767 boolean_t found; 768 zprop_list_t *entry; 769 770 if (zfeature_is_supported(nvpair_name(nvp))) 771 continue; 772 773 propname = zfs_asprintf(hdl, "unsupported@%s", 774 nvpair_name(nvp)); 775 776 /* 777 * Before adding the property to the list make sure that no 778 * other pool already added the same property. 779 */ 780 found = B_FALSE; 781 entry = *plp; 782 while (entry != NULL) { 783 if (entry->pl_user_prop != NULL && 784 strcmp(propname, entry->pl_user_prop) == 0) { 785 found = B_TRUE; 786 break; 787 } 788 entry = entry->pl_next; 789 } 790 if (found) { 791 free(propname); 792 continue; 793 } 794 795 entry = zfs_alloc(hdl, sizeof (zprop_list_t)); 796 entry->pl_prop = ZPROP_INVAL; 797 entry->pl_user_prop = propname; 798 entry->pl_width = strlen(entry->pl_user_prop); 799 entry->pl_all = B_TRUE; 800 801 *last = entry; 802 last = &entry->pl_next; 803 } 804 805 for (entry = *plp; entry != NULL; entry = entry->pl_next) { 806 807 if (entry->pl_fixed) 808 continue; 809 810 if (entry->pl_prop != ZPROP_INVAL && 811 zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf), 812 NULL, B_FALSE) == 0) { 813 if (strlen(buf) > entry->pl_width) 814 entry->pl_width = strlen(buf); 815 } 816 } 817 818 return (0); 819} 820 821/* 822 * Get the state for the given feature on the given ZFS pool. 823 */ 824int 825zpool_prop_get_feature(zpool_handle_t *zhp, const char *propname, char *buf, 826 size_t len) 827{ 828 uint64_t refcount; 829 boolean_t found = B_FALSE; 830 nvlist_t *features = zpool_get_features(zhp); 831 boolean_t supported; 832 const char *feature = strchr(propname, '@') + 1; 833 834 supported = zpool_prop_feature(propname); 835 ASSERT(supported || zpool_prop_unsupported(propname)); 836 837 /* 838 * Convert from feature name to feature guid. This conversion is 839 * unecessary for unsupported@... properties because they already 840 * use guids. 841 */ 842 if (supported) { 843 int ret; 844 spa_feature_t fid; 845 846 ret = zfeature_lookup_name(feature, &fid); 847 if (ret != 0) { 848 (void) strlcpy(buf, "-", len); 849 return (ENOTSUP); 850 } 851 feature = spa_feature_table[fid].fi_guid; 852 } 853 854 if (nvlist_lookup_uint64(features, feature, &refcount) == 0) 855 found = B_TRUE; 856 857 if (supported) { 858 if (!found) { 859 (void) strlcpy(buf, ZFS_FEATURE_DISABLED, len); 860 } else { 861 if (refcount == 0) 862 (void) strlcpy(buf, ZFS_FEATURE_ENABLED, len); 863 else 864 (void) strlcpy(buf, ZFS_FEATURE_ACTIVE, len); 865 } 866 } else { 867 if (found) { 868 if (refcount == 0) { 869 (void) strcpy(buf, ZFS_UNSUPPORTED_INACTIVE); 870 } else { 871 (void) strcpy(buf, ZFS_UNSUPPORTED_READONLY); 872 } 873 } else { 874 (void) strlcpy(buf, "-", len); 875 return (ENOTSUP); 876 } 877 } 878 879 return (0); 880} 881 882/* 883 * Don't start the slice at the default block of 34; many storage 884 * devices will use a stripe width of 128k, so start there instead. 885 */ 886#define NEW_START_BLOCK 256 887 888/* 889 * Validate the given pool name, optionally putting an extended error message in 890 * 'buf'. 891 */ 892boolean_t 893zpool_name_valid(libzfs_handle_t *hdl, boolean_t isopen, const char *pool) 894{ 895 namecheck_err_t why; 896 char what; 897 int ret; 898 899 ret = pool_namecheck(pool, &why, &what); 900 901 /* 902 * The rules for reserved pool names were extended at a later point. 903 * But we need to support users with existing pools that may now be 904 * invalid. So we only check for this expanded set of names during a 905 * create (or import), and only in userland. 906 */ 907 if (ret == 0 && !isopen && 908 (strncmp(pool, "mirror", 6) == 0 || 909 strncmp(pool, "raidz", 5) == 0 || 910 strncmp(pool, "spare", 5) == 0 || 911 strcmp(pool, "log") == 0)) { 912 if (hdl != NULL) 913 zfs_error_aux(hdl, 914 dgettext(TEXT_DOMAIN, "name is reserved")); 915 return (B_FALSE); 916 } 917 918 919 if (ret != 0) { 920 if (hdl != NULL) { 921 switch (why) { 922 case NAME_ERR_TOOLONG: 923 zfs_error_aux(hdl, 924 dgettext(TEXT_DOMAIN, "name is too long")); 925 break; 926 927 case NAME_ERR_INVALCHAR: 928 zfs_error_aux(hdl, 929 dgettext(TEXT_DOMAIN, "invalid character " 930 "'%c' in pool name"), what); 931 break; 932 933 case NAME_ERR_NOLETTER: 934 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 935 "name must begin with a letter")); 936 break; 937 938 case NAME_ERR_RESERVED: 939 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 940 "name is reserved")); 941 break; 942 943 case NAME_ERR_DISKLIKE: 944 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 945 "pool name is reserved")); 946 break; 947 948 case NAME_ERR_LEADING_SLASH: 949 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 950 "leading slash in name")); 951 break; 952 953 case NAME_ERR_EMPTY_COMPONENT: 954 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 955 "empty component in name")); 956 break; 957 958 case NAME_ERR_TRAILING_SLASH: 959 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 960 "trailing slash in name")); 961 break; 962 963 case NAME_ERR_MULTIPLE_DELIMITERS: 964 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 965 "multiple '@' and/or '#' delimiters in " 966 "name")); 967 break; 968 969 default: 970 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 971 "(%d) not defined"), why); 972 break; 973 } 974 } 975 return (B_FALSE); 976 } 977 978 return (B_TRUE); 979} 980 981/* 982 * Open a handle to the given pool, even if the pool is currently in the FAULTED 983 * state. 984 */ 985zpool_handle_t * 986zpool_open_canfail(libzfs_handle_t *hdl, const char *pool) 987{ 988 zpool_handle_t *zhp; 989 boolean_t missing; 990 991 /* 992 * Make sure the pool name is valid. 993 */ 994 if (!zpool_name_valid(hdl, B_TRUE, pool)) { 995 (void) zfs_error_fmt(hdl, EZFS_INVALIDNAME, 996 dgettext(TEXT_DOMAIN, "cannot open '%s'"), 997 pool); 998 return (NULL); 999 } 1000 1001 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL) 1002 return (NULL); 1003 1004 zhp->zpool_hdl = hdl; 1005 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name)); 1006 1007 if (zpool_refresh_stats(zhp, &missing) != 0) { 1008 zpool_close(zhp); 1009 return (NULL); 1010 } 1011 1012 if (missing) { 1013 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool")); 1014 (void) zfs_error_fmt(hdl, EZFS_NOENT, 1015 dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool); 1016 zpool_close(zhp); 1017 return (NULL); 1018 } 1019 1020 return (zhp); 1021} 1022 1023/* 1024 * Like the above, but silent on error. Used when iterating over pools (because 1025 * the configuration cache may be out of date). 1026 */ 1027int 1028zpool_open_silent(libzfs_handle_t *hdl, const char *pool, zpool_handle_t **ret) 1029{ 1030 zpool_handle_t *zhp; 1031 boolean_t missing; 1032 1033 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL) 1034 return (-1); 1035 1036 zhp->zpool_hdl = hdl; 1037 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name)); 1038 1039 if (zpool_refresh_stats(zhp, &missing) != 0) { 1040 zpool_close(zhp); 1041 return (-1); 1042 } 1043 1044 if (missing) { 1045 zpool_close(zhp); 1046 *ret = NULL; 1047 return (0); 1048 } 1049 1050 *ret = zhp; 1051 return (0); 1052} 1053 1054/* 1055 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted 1056 * state. 1057 */ 1058zpool_handle_t * 1059zpool_open(libzfs_handle_t *hdl, const char *pool) 1060{ 1061 zpool_handle_t *zhp; 1062 1063 if ((zhp = zpool_open_canfail(hdl, pool)) == NULL) 1064 return (NULL); 1065 1066 if (zhp->zpool_state == POOL_STATE_UNAVAIL) { 1067 (void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL, 1068 dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name); 1069 zpool_close(zhp); 1070 return (NULL); 1071 } 1072 1073 return (zhp); 1074} 1075 1076/* 1077 * Close the handle. Simply frees the memory associated with the handle. 1078 */ 1079void 1080zpool_close(zpool_handle_t *zhp) 1081{ 1082 nvlist_free(zhp->zpool_config); 1083 nvlist_free(zhp->zpool_old_config); 1084 nvlist_free(zhp->zpool_props); 1085 free(zhp); 1086} 1087 1088/* 1089 * Return the name of the pool. 1090 */ 1091const char * 1092zpool_get_name(zpool_handle_t *zhp) 1093{ 1094 return (zhp->zpool_name); 1095} 1096 1097 1098/* 1099 * Return the state of the pool (ACTIVE or UNAVAILABLE) 1100 */ 1101int 1102zpool_get_state(zpool_handle_t *zhp) 1103{ 1104 return (zhp->zpool_state); 1105} 1106 1107/* 1108 * Create the named pool, using the provided vdev list. It is assumed 1109 * that the consumer has already validated the contents of the nvlist, so we 1110 * don't have to worry about error semantics. 1111 */ 1112int 1113zpool_create(libzfs_handle_t *hdl, const char *pool, nvlist_t *nvroot, 1114 nvlist_t *props, nvlist_t *fsprops) 1115{ 1116 zfs_cmd_t zc = { 0 }; 1117 nvlist_t *zc_fsprops = NULL; 1118 nvlist_t *zc_props = NULL; 1119 char msg[1024]; 1120 int ret = -1; 1121 1122 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1123 "cannot create '%s'"), pool); 1124 1125 if (!zpool_name_valid(hdl, B_FALSE, pool)) 1126 return (zfs_error(hdl, EZFS_INVALIDNAME, msg)); 1127 1128 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0) 1129 return (-1); 1130 1131 if (props) { 1132 prop_flags_t flags = { .create = B_TRUE, .import = B_FALSE }; 1133 1134 if ((zc_props = zpool_valid_proplist(hdl, pool, props, 1135 SPA_VERSION_1, flags, msg)) == NULL) { 1136 goto create_failed; 1137 } 1138 } 1139 1140 if (fsprops) { 1141 uint64_t zoned; 1142 char *zonestr; 1143 1144 zoned = ((nvlist_lookup_string(fsprops, 1145 zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) && 1146 strcmp(zonestr, "on") == 0); 1147 1148 if ((zc_fsprops = zfs_valid_proplist(hdl, ZFS_TYPE_FILESYSTEM, 1149 fsprops, zoned, NULL, NULL, msg)) == NULL) { 1150 goto create_failed; 1151 } 1152 if (!zc_props && 1153 (nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) { 1154 goto create_failed; 1155 } 1156 if (nvlist_add_nvlist(zc_props, 1157 ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) { 1158 goto create_failed; 1159 } 1160 } 1161 1162 if (zc_props && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0) 1163 goto create_failed; 1164 1165 (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name)); 1166 1167 if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) { 1168 1169 zcmd_free_nvlists(&zc); 1170 nvlist_free(zc_props); 1171 nvlist_free(zc_fsprops); 1172 1173 switch (errno) { 1174 case EBUSY: 1175 /* 1176 * This can happen if the user has specified the same 1177 * device multiple times. We can't reliably detect this 1178 * until we try to add it and see we already have a 1179 * label. 1180 */ 1181 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1182 "one or more vdevs refer to the same device")); 1183 return (zfs_error(hdl, EZFS_BADDEV, msg)); 1184 1185 case ERANGE: 1186 /* 1187 * This happens if the record size is smaller or larger 1188 * than the allowed size range, or not a power of 2. 1189 * 1190 * NOTE: although zfs_valid_proplist is called earlier, 1191 * this case may have slipped through since the 1192 * pool does not exist yet and it is therefore 1193 * impossible to read properties e.g. max blocksize 1194 * from the pool. 1195 */ 1196 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1197 "record size invalid")); 1198 return (zfs_error(hdl, EZFS_BADPROP, msg)); 1199 1200 case EOVERFLOW: 1201 /* 1202 * This occurs when one of the devices is below 1203 * SPA_MINDEVSIZE. Unfortunately, we can't detect which 1204 * device was the problem device since there's no 1205 * reliable way to determine device size from userland. 1206 */ 1207 { 1208 char buf[64]; 1209 1210 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf)); 1211 1212 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1213 "one or more devices is less than the " 1214 "minimum size (%s)"), buf); 1215 } 1216 return (zfs_error(hdl, EZFS_BADDEV, msg)); 1217 1218 case ENOSPC: 1219 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1220 "one or more devices is out of space")); 1221 return (zfs_error(hdl, EZFS_BADDEV, msg)); 1222 1223 case ENOTBLK: 1224 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1225 "cache device must be a disk or disk slice")); 1226 return (zfs_error(hdl, EZFS_BADDEV, msg)); 1227 1228 default: 1229 return (zpool_standard_error(hdl, errno, msg)); 1230 } 1231 } 1232 1233create_failed: 1234 zcmd_free_nvlists(&zc); 1235 nvlist_free(zc_props); 1236 nvlist_free(zc_fsprops); 1237 return (ret); 1238} 1239 1240/* 1241 * Destroy the given pool. It is up to the caller to ensure that there are no 1242 * datasets left in the pool. 1243 */ 1244int 1245zpool_destroy(zpool_handle_t *zhp, const char *log_str) 1246{ 1247 zfs_cmd_t zc = { 0 }; 1248 zfs_handle_t *zfp = NULL; 1249 libzfs_handle_t *hdl = zhp->zpool_hdl; 1250 char msg[1024]; 1251 1252 if (zhp->zpool_state == POOL_STATE_ACTIVE && 1253 (zfp = zfs_open(hdl, zhp->zpool_name, ZFS_TYPE_FILESYSTEM)) == NULL) 1254 return (-1); 1255 1256 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1257 zc.zc_history = (uint64_t)(uintptr_t)log_str; 1258 1259 if (zfs_ioctl(hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) { 1260 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1261 "cannot destroy '%s'"), zhp->zpool_name); 1262 1263 if (errno == EROFS) { 1264 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1265 "one or more devices is read only")); 1266 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1267 } else { 1268 (void) zpool_standard_error(hdl, errno, msg); 1269 } 1270 1271 if (zfp) 1272 zfs_close(zfp); 1273 return (-1); 1274 } 1275 1276 if (zfp) { 1277 remove_mountpoint(zfp); 1278 zfs_close(zfp); 1279 } 1280 1281 return (0); 1282} 1283 1284/* 1285 * Add the given vdevs to the pool. The caller must have already performed the 1286 * necessary verification to ensure that the vdev specification is well-formed. 1287 */ 1288int 1289zpool_add(zpool_handle_t *zhp, nvlist_t *nvroot) 1290{ 1291 zfs_cmd_t zc = { 0 }; 1292 int ret; 1293 libzfs_handle_t *hdl = zhp->zpool_hdl; 1294 char msg[1024]; 1295 nvlist_t **spares, **l2cache; 1296 uint_t nspares, nl2cache; 1297 1298 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1299 "cannot add to '%s'"), zhp->zpool_name); 1300 1301 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) < 1302 SPA_VERSION_SPARES && 1303 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, 1304 &spares, &nspares) == 0) { 1305 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be " 1306 "upgraded to add hot spares")); 1307 return (zfs_error(hdl, EZFS_BADVERSION, msg)); 1308 } 1309 1310 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) < 1311 SPA_VERSION_L2CACHE && 1312 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, 1313 &l2cache, &nl2cache) == 0) { 1314 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be " 1315 "upgraded to add cache devices")); 1316 return (zfs_error(hdl, EZFS_BADVERSION, msg)); 1317 } 1318 1319 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0) 1320 return (-1); 1321 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1322 1323 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) { 1324 switch (errno) { 1325 case EBUSY: 1326 /* 1327 * This can happen if the user has specified the same 1328 * device multiple times. We can't reliably detect this 1329 * until we try to add it and see we already have a 1330 * label. 1331 */ 1332 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1333 "one or more vdevs refer to the same device")); 1334 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1335 break; 1336 1337 case EINVAL: 1338 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1339 "invalid config; a pool with removing/removed " 1340 "vdevs does not support adding raidz vdevs")); 1341 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1342 break; 1343 1344 case EOVERFLOW: 1345 /* 1346 * This occurrs when one of the devices is below 1347 * SPA_MINDEVSIZE. Unfortunately, we can't detect which 1348 * device was the problem device since there's no 1349 * reliable way to determine device size from userland. 1350 */ 1351 { 1352 char buf[64]; 1353 1354 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf)); 1355 1356 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1357 "device is less than the minimum " 1358 "size (%s)"), buf); 1359 } 1360 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1361 break; 1362 1363 case ENOTSUP: 1364 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1365 "pool must be upgraded to add these vdevs")); 1366 (void) zfs_error(hdl, EZFS_BADVERSION, msg); 1367 break; 1368 1369 case EDOM: 1370 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1371 "root pool can not have multiple vdevs" 1372 " or separate logs")); 1373 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, msg); 1374 break; 1375 1376 case ENOTBLK: 1377 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1378 "cache device must be a disk or disk slice")); 1379 (void) zfs_error(hdl, EZFS_BADDEV, msg); 1380 break; 1381 1382 default: 1383 (void) zpool_standard_error(hdl, errno, msg); 1384 } 1385 1386 ret = -1; 1387 } else { 1388 ret = 0; 1389 } 1390 1391 zcmd_free_nvlists(&zc); 1392 1393 return (ret); 1394} 1395 1396/* 1397 * Exports the pool from the system. The caller must ensure that there are no 1398 * mounted datasets in the pool. 1399 */ 1400static int 1401zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce, 1402 const char *log_str) 1403{ 1404 zfs_cmd_t zc = { 0 }; 1405 char msg[1024]; 1406 1407 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1408 "cannot export '%s'"), zhp->zpool_name); 1409 1410 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1411 zc.zc_cookie = force; 1412 zc.zc_guid = hardforce; 1413 zc.zc_history = (uint64_t)(uintptr_t)log_str; 1414 1415 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) { 1416 switch (errno) { 1417 case EXDEV: 1418 zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN, 1419 "use '-f' to override the following errors:\n" 1420 "'%s' has an active shared spare which could be" 1421 " used by other pools once '%s' is exported."), 1422 zhp->zpool_name, zhp->zpool_name); 1423 return (zfs_error(zhp->zpool_hdl, EZFS_ACTIVE_SPARE, 1424 msg)); 1425 default: 1426 return (zpool_standard_error_fmt(zhp->zpool_hdl, errno, 1427 msg)); 1428 } 1429 } 1430 1431 return (0); 1432} 1433 1434int 1435zpool_export(zpool_handle_t *zhp, boolean_t force, const char *log_str) 1436{ 1437 return (zpool_export_common(zhp, force, B_FALSE, log_str)); 1438} 1439 1440int 1441zpool_export_force(zpool_handle_t *zhp, const char *log_str) 1442{ 1443 return (zpool_export_common(zhp, B_TRUE, B_TRUE, log_str)); 1444} 1445 1446static void 1447zpool_rewind_exclaim(libzfs_handle_t *hdl, const char *name, boolean_t dryrun, 1448 nvlist_t *config) 1449{ 1450 nvlist_t *nv = NULL; 1451 uint64_t rewindto; 1452 int64_t loss = -1; 1453 struct tm t; 1454 char timestr[128]; 1455 1456 if (!hdl->libzfs_printerr || config == NULL) 1457 return; 1458 1459 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 || 1460 nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0) { 1461 return; 1462 } 1463 1464 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0) 1465 return; 1466 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss); 1467 1468 if (localtime_r((time_t *)&rewindto, &t) != NULL && 1469 strftime(timestr, 128, 0, &t) != 0) { 1470 if (dryrun) { 1471 (void) printf(dgettext(TEXT_DOMAIN, 1472 "Would be able to return %s " 1473 "to its state as of %s.\n"), 1474 name, timestr); 1475 } else { 1476 (void) printf(dgettext(TEXT_DOMAIN, 1477 "Pool %s returned to its state as of %s.\n"), 1478 name, timestr); 1479 } 1480 if (loss > 120) { 1481 (void) printf(dgettext(TEXT_DOMAIN, 1482 "%s approximately %lld "), 1483 dryrun ? "Would discard" : "Discarded", 1484 (loss + 30) / 60); 1485 (void) printf(dgettext(TEXT_DOMAIN, 1486 "minutes of transactions.\n")); 1487 } else if (loss > 0) { 1488 (void) printf(dgettext(TEXT_DOMAIN, 1489 "%s approximately %lld "), 1490 dryrun ? "Would discard" : "Discarded", loss); 1491 (void) printf(dgettext(TEXT_DOMAIN, 1492 "seconds of transactions.\n")); 1493 } 1494 } 1495} 1496 1497void 1498zpool_explain_recover(libzfs_handle_t *hdl, const char *name, int reason, 1499 nvlist_t *config) 1500{ 1501 nvlist_t *nv = NULL; 1502 int64_t loss = -1; 1503 uint64_t edata = UINT64_MAX; 1504 uint64_t rewindto; 1505 struct tm t; 1506 char timestr[128]; 1507 1508 if (!hdl->libzfs_printerr) 1509 return; 1510 1511 if (reason >= 0) 1512 (void) printf(dgettext(TEXT_DOMAIN, "action: ")); 1513 else 1514 (void) printf(dgettext(TEXT_DOMAIN, "\t")); 1515 1516 /* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */ 1517 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 || 1518 nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0 || 1519 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0) 1520 goto no_info; 1521 1522 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss); 1523 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_DATA_ERRORS, 1524 &edata); 1525 1526 (void) printf(dgettext(TEXT_DOMAIN, 1527 "Recovery is possible, but will result in some data loss.\n")); 1528 1529 if (localtime_r((time_t *)&rewindto, &t) != NULL && 1530 strftime(timestr, 128, 0, &t) != 0) { 1531 (void) printf(dgettext(TEXT_DOMAIN, 1532 "\tReturning the pool to its state as of %s\n" 1533 "\tshould correct the problem. "), 1534 timestr); 1535 } else { 1536 (void) printf(dgettext(TEXT_DOMAIN, 1537 "\tReverting the pool to an earlier state " 1538 "should correct the problem.\n\t")); 1539 } 1540 1541 if (loss > 120) { 1542 (void) printf(dgettext(TEXT_DOMAIN, 1543 "Approximately %lld minutes of data\n" 1544 "\tmust be discarded, irreversibly. "), (loss + 30) / 60); 1545 } else if (loss > 0) { 1546 (void) printf(dgettext(TEXT_DOMAIN, 1547 "Approximately %lld seconds of data\n" 1548 "\tmust be discarded, irreversibly. "), loss); 1549 } 1550 if (edata != 0 && edata != UINT64_MAX) { 1551 if (edata == 1) { 1552 (void) printf(dgettext(TEXT_DOMAIN, 1553 "After rewind, at least\n" 1554 "\tone persistent user-data error will remain. ")); 1555 } else { 1556 (void) printf(dgettext(TEXT_DOMAIN, 1557 "After rewind, several\n" 1558 "\tpersistent user-data errors will remain. ")); 1559 } 1560 } 1561 (void) printf(dgettext(TEXT_DOMAIN, 1562 "Recovery can be attempted\n\tby executing 'zpool %s -F %s'. "), 1563 reason >= 0 ? "clear" : "import", name); 1564 1565 (void) printf(dgettext(TEXT_DOMAIN, 1566 "A scrub of the pool\n" 1567 "\tis strongly recommended after recovery.\n")); 1568 return; 1569 1570no_info: 1571 (void) printf(dgettext(TEXT_DOMAIN, 1572 "Destroy and re-create the pool from\n\ta backup source.\n")); 1573} 1574 1575/* 1576 * zpool_import() is a contracted interface. Should be kept the same 1577 * if possible. 1578 * 1579 * Applications should use zpool_import_props() to import a pool with 1580 * new properties value to be set. 1581 */ 1582int 1583zpool_import(libzfs_handle_t *hdl, nvlist_t *config, const char *newname, 1584 char *altroot) 1585{ 1586 nvlist_t *props = NULL; 1587 int ret; 1588 1589 if (altroot != NULL) { 1590 if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) { 1591 return (zfs_error_fmt(hdl, EZFS_NOMEM, 1592 dgettext(TEXT_DOMAIN, "cannot import '%s'"), 1593 newname)); 1594 } 1595 1596 if (nvlist_add_string(props, 1597 zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 || 1598 nvlist_add_string(props, 1599 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) { 1600 nvlist_free(props); 1601 return (zfs_error_fmt(hdl, EZFS_NOMEM, 1602 dgettext(TEXT_DOMAIN, "cannot import '%s'"), 1603 newname)); 1604 } 1605 } 1606 1607 ret = zpool_import_props(hdl, config, newname, props, 1608 ZFS_IMPORT_NORMAL); 1609 nvlist_free(props); 1610 return (ret); 1611} 1612 1613static void 1614print_vdev_tree(libzfs_handle_t *hdl, const char *name, nvlist_t *nv, 1615 int indent) 1616{ 1617 nvlist_t **child; 1618 uint_t c, children; 1619 char *vname; 1620 uint64_t is_log = 0; 1621 1622 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG, 1623 &is_log); 1624 1625 if (name != NULL) 1626 (void) printf("\t%*s%s%s\n", indent, "", name, 1627 is_log ? " [log]" : ""); 1628 1629 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, 1630 &child, &children) != 0) 1631 return; 1632 1633 for (c = 0; c < children; c++) { 1634 vname = zpool_vdev_name(hdl, NULL, child[c], B_TRUE); 1635 print_vdev_tree(hdl, vname, child[c], indent + 2); 1636 free(vname); 1637 } 1638} 1639 1640void 1641zpool_print_unsup_feat(nvlist_t *config) 1642{ 1643 nvlist_t *nvinfo, *unsup_feat; 1644 1645 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 1646 0); 1647 verify(nvlist_lookup_nvlist(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT, 1648 &unsup_feat) == 0); 1649 1650 for (nvpair_t *nvp = nvlist_next_nvpair(unsup_feat, NULL); nvp != NULL; 1651 nvp = nvlist_next_nvpair(unsup_feat, nvp)) { 1652 char *desc; 1653 1654 verify(nvpair_type(nvp) == DATA_TYPE_STRING); 1655 verify(nvpair_value_string(nvp, &desc) == 0); 1656 1657 if (strlen(desc) > 0) 1658 (void) printf("\t%s (%s)\n", nvpair_name(nvp), desc); 1659 else 1660 (void) printf("\t%s\n", nvpair_name(nvp)); 1661 } 1662} 1663 1664/* 1665 * Import the given pool using the known configuration and a list of 1666 * properties to be set. The configuration should have come from 1667 * zpool_find_import(). The 'newname' parameters control whether the pool 1668 * is imported with a different name. 1669 */ 1670int 1671zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname, 1672 nvlist_t *props, int flags) 1673{ 1674 zfs_cmd_t zc = { 0 }; 1675 zpool_rewind_policy_t policy; 1676 nvlist_t *nv = NULL; 1677 nvlist_t *nvinfo = NULL; 1678 nvlist_t *missing = NULL; 1679 char *thename; 1680 char *origname; 1681 int ret; 1682 int error = 0; 1683 char errbuf[1024]; 1684 1685 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, 1686 &origname) == 0); 1687 1688 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, 1689 "cannot import pool '%s'"), origname); 1690 1691 if (newname != NULL) { 1692 if (!zpool_name_valid(hdl, B_FALSE, newname)) 1693 return (zfs_error_fmt(hdl, EZFS_INVALIDNAME, 1694 dgettext(TEXT_DOMAIN, "cannot import '%s'"), 1695 newname)); 1696 thename = (char *)newname; 1697 } else { 1698 thename = origname; 1699 } 1700 1701 if (props != NULL) { 1702 uint64_t version; 1703 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE }; 1704 1705 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, 1706 &version) == 0); 1707 1708 if ((props = zpool_valid_proplist(hdl, origname, 1709 props, version, flags, errbuf)) == NULL) 1710 return (-1); 1711 if (zcmd_write_src_nvlist(hdl, &zc, props) != 0) { 1712 nvlist_free(props); 1713 return (-1); 1714 } 1715 nvlist_free(props); 1716 } 1717 1718 (void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name)); 1719 1720 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, 1721 &zc.zc_guid) == 0); 1722 1723 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) { 1724 zcmd_free_nvlists(&zc); 1725 return (-1); 1726 } 1727 if (zcmd_alloc_dst_nvlist(hdl, &zc, zc.zc_nvlist_conf_size * 2) != 0) { 1728 zcmd_free_nvlists(&zc); 1729 return (-1); 1730 } 1731 1732 zc.zc_cookie = flags; 1733 while ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc)) != 0 && 1734 errno == ENOMEM) { 1735 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) { 1736 zcmd_free_nvlists(&zc); 1737 return (-1); 1738 } 1739 } 1740 if (ret != 0) 1741 error = errno; 1742 1743 (void) zcmd_read_dst_nvlist(hdl, &zc, &nv); 1744 1745 zcmd_free_nvlists(&zc); 1746 1747 zpool_get_rewind_policy(config, &policy); 1748 1749 if (error) { 1750 char desc[1024]; 1751 1752 /* 1753 * Dry-run failed, but we print out what success 1754 * looks like if we found a best txg 1755 */ 1756 if (policy.zrp_request & ZPOOL_TRY_REWIND) { 1757 zpool_rewind_exclaim(hdl, newname ? origname : thename, 1758 B_TRUE, nv); 1759 nvlist_free(nv); 1760 return (-1); 1761 } 1762 1763 if (newname == NULL) 1764 (void) snprintf(desc, sizeof (desc), 1765 dgettext(TEXT_DOMAIN, "cannot import '%s'"), 1766 thename); 1767 else 1768 (void) snprintf(desc, sizeof (desc), 1769 dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"), 1770 origname, thename); 1771 1772 switch (error) { 1773 case ENOTSUP: 1774 if (nv != NULL && nvlist_lookup_nvlist(nv, 1775 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 && 1776 nvlist_exists(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT)) { 1777 (void) printf(dgettext(TEXT_DOMAIN, "This " 1778 "pool uses the following feature(s) not " 1779 "supported by this system:\n")); 1780 zpool_print_unsup_feat(nv); 1781 if (nvlist_exists(nvinfo, 1782 ZPOOL_CONFIG_CAN_RDONLY)) { 1783 (void) printf(dgettext(TEXT_DOMAIN, 1784 "All unsupported features are only " 1785 "required for writing to the pool." 1786 "\nThe pool can be imported using " 1787 "'-o readonly=on'.\n")); 1788 } 1789 } 1790 /* 1791 * Unsupported version. 1792 */ 1793 (void) zfs_error(hdl, EZFS_BADVERSION, desc); 1794 break; 1795 1796 case EINVAL: 1797 (void) zfs_error(hdl, EZFS_INVALCONFIG, desc); 1798 break; 1799 1800 case EROFS: 1801 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1802 "one or more devices is read only")); 1803 (void) zfs_error(hdl, EZFS_BADDEV, desc); 1804 break; 1805 1806 case ENXIO: 1807 if (nv && nvlist_lookup_nvlist(nv, 1808 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 && 1809 nvlist_lookup_nvlist(nvinfo, 1810 ZPOOL_CONFIG_MISSING_DEVICES, &missing) == 0) { 1811 (void) printf(dgettext(TEXT_DOMAIN, 1812 "The devices below are missing, use " 1813 "'-m' to import the pool anyway:\n")); 1814 print_vdev_tree(hdl, NULL, missing, 2); 1815 (void) printf("\n"); 1816 } 1817 (void) zpool_standard_error(hdl, error, desc); 1818 break; 1819 1820 case EEXIST: 1821 (void) zpool_standard_error(hdl, error, desc); 1822 break; 1823 case ENAMETOOLONG: 1824 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 1825 "new name of at least one dataset is longer than " 1826 "the maximum allowable length")); 1827 (void) zfs_error(hdl, EZFS_NAMETOOLONG, desc); 1828 break; 1829 default: 1830 (void) zpool_standard_error(hdl, error, desc); 1831 zpool_explain_recover(hdl, 1832 newname ? origname : thename, -error, nv); 1833 break; 1834 } 1835 1836 nvlist_free(nv); 1837 ret = -1; 1838 } else { 1839 zpool_handle_t *zhp; 1840 1841 /* 1842 * This should never fail, but play it safe anyway. 1843 */ 1844 if (zpool_open_silent(hdl, thename, &zhp) != 0) 1845 ret = -1; 1846 else if (zhp != NULL) 1847 zpool_close(zhp); 1848 if (policy.zrp_request & 1849 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) { 1850 zpool_rewind_exclaim(hdl, newname ? origname : thename, 1851 ((policy.zrp_request & ZPOOL_TRY_REWIND) != 0), nv); 1852 } 1853 nvlist_free(nv); 1854 return (0); 1855 } 1856 1857 return (ret); 1858} 1859 1860/* 1861 * Scan the pool. 1862 */ 1863int 1864zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func, pool_scrub_cmd_t cmd) 1865{ 1866 zfs_cmd_t zc = { 0 }; 1867 char msg[1024]; 1868 int err; 1869 libzfs_handle_t *hdl = zhp->zpool_hdl; 1870 1871 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 1872 zc.zc_cookie = func; 1873 zc.zc_flags = cmd; 1874 1875 if (zfs_ioctl(hdl, ZFS_IOC_POOL_SCAN, &zc) == 0) 1876 return (0); 1877 1878 err = errno; 1879 1880 /* ECANCELED on a scrub means we resumed a paused scrub */ 1881 if (err == ECANCELED && func == POOL_SCAN_SCRUB && 1882 cmd == POOL_SCRUB_NORMAL) 1883 return (0); 1884 1885 if (err == ENOENT && func != POOL_SCAN_NONE && cmd == POOL_SCRUB_NORMAL) 1886 return (0); 1887 1888 if (func == POOL_SCAN_SCRUB) { 1889 if (cmd == POOL_SCRUB_PAUSE) { 1890 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1891 "cannot pause scrubbing %s"), zc.zc_name); 1892 } else { 1893 assert(cmd == POOL_SCRUB_NORMAL); 1894 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 1895 "cannot scrub %s"), zc.zc_name); 1896 } 1897 } else if (func == POOL_SCAN_NONE) { 1898 (void) snprintf(msg, sizeof (msg), 1899 dgettext(TEXT_DOMAIN, "cannot cancel scrubbing %s"), 1900 zc.zc_name); 1901 } else { 1902 assert(!"unexpected result"); 1903 } 1904 1905 if (err == EBUSY) { 1906 nvlist_t *nvroot; 1907 pool_scan_stat_t *ps = NULL; 1908 uint_t psc; 1909 1910 verify(nvlist_lookup_nvlist(zhp->zpool_config, 1911 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 1912 (void) nvlist_lookup_uint64_array(nvroot, 1913 ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &psc); 1914 if (ps && ps->pss_func == POOL_SCAN_SCRUB) { 1915 if (cmd == POOL_SCRUB_PAUSE) 1916 return (zfs_error(hdl, EZFS_SCRUB_PAUSED, msg)); 1917 else 1918 return (zfs_error(hdl, EZFS_SCRUBBING, msg)); 1919 } else { 1920 return (zfs_error(hdl, EZFS_RESILVERING, msg)); 1921 } 1922 } else if (err == ENOENT) { 1923 return (zfs_error(hdl, EZFS_NO_SCRUB, msg)); 1924 } else { 1925 return (zpool_standard_error(hdl, err, msg)); 1926 } 1927} 1928 1929#ifdef illumos 1930/* 1931 * This provides a very minimal check whether a given string is likely a 1932 * c#t#d# style string. Users of this are expected to do their own 1933 * verification of the s# part. 1934 */ 1935#define CTD_CHECK(str) (str && str[0] == 'c' && isdigit(str[1])) 1936 1937/* 1938 * More elaborate version for ones which may start with "/dev/dsk/" 1939 * and the like. 1940 */ 1941static int 1942ctd_check_path(char *str) 1943{ 1944 /* 1945 * If it starts with a slash, check the last component. 1946 */ 1947 if (str && str[0] == '/') { 1948 char *tmp = strrchr(str, '/'); 1949 1950 /* 1951 * If it ends in "/old", check the second-to-last 1952 * component of the string instead. 1953 */ 1954 if (tmp != str && strcmp(tmp, "/old") == 0) { 1955 for (tmp--; *tmp != '/'; tmp--) 1956 ; 1957 } 1958 str = tmp + 1; 1959 } 1960 return (CTD_CHECK(str)); 1961} 1962#endif 1963 1964/* 1965 * Find a vdev that matches the search criteria specified. We use the 1966 * the nvpair name to determine how we should look for the device. 1967 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL 1968 * spare; but FALSE if its an INUSE spare. 1969 */ 1970static nvlist_t * 1971vdev_to_nvlist_iter(nvlist_t *nv, nvlist_t *search, boolean_t *avail_spare, 1972 boolean_t *l2cache, boolean_t *log) 1973{ 1974 uint_t c, children; 1975 nvlist_t **child; 1976 nvlist_t *ret; 1977 uint64_t is_log; 1978 char *srchkey; 1979 nvpair_t *pair = nvlist_next_nvpair(search, NULL); 1980 1981 /* Nothing to look for */ 1982 if (search == NULL || pair == NULL) 1983 return (NULL); 1984 1985 /* Obtain the key we will use to search */ 1986 srchkey = nvpair_name(pair); 1987 1988 switch (nvpair_type(pair)) { 1989 case DATA_TYPE_UINT64: 1990 if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) { 1991 uint64_t srchval, theguid; 1992 1993 verify(nvpair_value_uint64(pair, &srchval) == 0); 1994 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, 1995 &theguid) == 0); 1996 if (theguid == srchval) 1997 return (nv); 1998 } 1999 break; 2000 2001 case DATA_TYPE_STRING: { 2002 char *srchval, *val; 2003 2004 verify(nvpair_value_string(pair, &srchval) == 0); 2005 if (nvlist_lookup_string(nv, srchkey, &val) != 0) 2006 break; 2007 2008 /* 2009 * Search for the requested value. Special cases: 2010 * 2011 * - ZPOOL_CONFIG_PATH for whole disk entries. To support 2012 * UEFI boot, these end in "s0" or "s0/old" or "s1" or 2013 * "s1/old". The "s0" or "s1" part is hidden from the user, 2014 * but included in the string, so this matches around it. 2015 * - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE). 2016 * 2017 * Otherwise, all other searches are simple string compares. 2018 */ 2019#ifdef illumos 2020 if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0 && 2021 ctd_check_path(val)) { 2022 uint64_t wholedisk = 0; 2023 2024 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, 2025 &wholedisk); 2026 if (wholedisk) { 2027 int slen = strlen(srchval); 2028 int vlen = strlen(val); 2029 2030 if (slen != vlen - 2) 2031 break; 2032 2033 /* 2034 * make_leaf_vdev() should only set 2035 * wholedisk for ZPOOL_CONFIG_PATHs which 2036 * will include "/dev/dsk/", giving plenty of 2037 * room for the indices used next. 2038 */ 2039 ASSERT(vlen >= 6); 2040 2041 /* 2042 * strings identical except trailing "s0" 2043 */ 2044 if ((strcmp(&val[vlen - 2], "s0") == 0 || 2045 strcmp(&val[vlen - 2], "s1") == 0) && 2046 strncmp(srchval, val, slen) == 0) 2047 return (nv); 2048 2049 /* 2050 * strings identical except trailing "s0/old" 2051 */ 2052 if ((strcmp(&val[vlen - 6], "s0/old") == 0 || 2053 strcmp(&val[vlen - 6], "s1/old") == 0) && 2054 strcmp(&srchval[slen - 4], "/old") == 0 && 2055 strncmp(srchval, val, slen - 4) == 0) 2056 return (nv); 2057 2058 break; 2059 } 2060 } else if (strcmp(srchkey, ZPOOL_CONFIG_TYPE) == 0 && val) { 2061#else 2062 if (strcmp(srchkey, ZPOOL_CONFIG_TYPE) == 0 && val) { 2063#endif 2064 char *type, *idx, *end, *p; 2065 uint64_t id, vdev_id; 2066 2067 /* 2068 * Determine our vdev type, keeping in mind 2069 * that the srchval is composed of a type and 2070 * vdev id pair (i.e. mirror-4). 2071 */ 2072 if ((type = strdup(srchval)) == NULL) 2073 return (NULL); 2074 2075 if ((p = strrchr(type, '-')) == NULL) { 2076 free(type); 2077 break; 2078 } 2079 idx = p + 1; 2080 *p = '\0'; 2081 2082 /* 2083 * If the types don't match then keep looking. 2084 */ 2085 if (strncmp(val, type, strlen(val)) != 0) { 2086 free(type); 2087 break; 2088 } 2089 2090 verify(zpool_vdev_is_interior(type)); 2091 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID, 2092 &id) == 0); 2093 2094 errno = 0; 2095 vdev_id = strtoull(idx, &end, 10); 2096 2097 free(type); 2098 if (errno != 0) 2099 return (NULL); 2100 2101 /* 2102 * Now verify that we have the correct vdev id. 2103 */ 2104 if (vdev_id == id) 2105 return (nv); 2106 } 2107 2108 /* 2109 * Common case 2110 */ 2111 if (strcmp(srchval, val) == 0) 2112 return (nv); 2113 break; 2114 } 2115 2116 default: 2117 break; 2118 } 2119 2120 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, 2121 &child, &children) != 0) 2122 return (NULL); 2123 2124 for (c = 0; c < children; c++) { 2125 if ((ret = vdev_to_nvlist_iter(child[c], search, 2126 avail_spare, l2cache, NULL)) != NULL) { 2127 /* 2128 * The 'is_log' value is only set for the toplevel 2129 * vdev, not the leaf vdevs. So we always lookup the 2130 * log device from the root of the vdev tree (where 2131 * 'log' is non-NULL). 2132 */ 2133 if (log != NULL && 2134 nvlist_lookup_uint64(child[c], 2135 ZPOOL_CONFIG_IS_LOG, &is_log) == 0 && 2136 is_log) { 2137 *log = B_TRUE; 2138 } 2139 return (ret); 2140 } 2141 } 2142 2143 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES, 2144 &child, &children) == 0) { 2145 for (c = 0; c < children; c++) { 2146 if ((ret = vdev_to_nvlist_iter(child[c], search, 2147 avail_spare, l2cache, NULL)) != NULL) { 2148 *avail_spare = B_TRUE; 2149 return (ret); 2150 } 2151 } 2152 } 2153 2154 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE, 2155 &child, &children) == 0) { 2156 for (c = 0; c < children; c++) { 2157 if ((ret = vdev_to_nvlist_iter(child[c], search, 2158 avail_spare, l2cache, NULL)) != NULL) { 2159 *l2cache = B_TRUE; 2160 return (ret); 2161 } 2162 } 2163 } 2164 2165 return (NULL); 2166} 2167 2168/* 2169 * Given a physical path (minus the "/devices" prefix), find the 2170 * associated vdev. 2171 */ 2172nvlist_t * 2173zpool_find_vdev_by_physpath(zpool_handle_t *zhp, const char *ppath, 2174 boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log) 2175{ 2176 nvlist_t *search, *nvroot, *ret; 2177 2178 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2179 verify(nvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH, ppath) == 0); 2180 2181 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE, 2182 &nvroot) == 0); 2183 2184 *avail_spare = B_FALSE; 2185 *l2cache = B_FALSE; 2186 if (log != NULL) 2187 *log = B_FALSE; 2188 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log); 2189 nvlist_free(search); 2190 2191 return (ret); 2192} 2193 2194/* 2195 * Determine if we have an "interior" top-level vdev (i.e mirror/raidz). 2196 */ 2197static boolean_t 2198zpool_vdev_is_interior(const char *name) 2199{ 2200 if (strncmp(name, VDEV_TYPE_RAIDZ, strlen(VDEV_TYPE_RAIDZ)) == 0 || 2201 strncmp(name, VDEV_TYPE_SPARE, strlen(VDEV_TYPE_SPARE)) == 0 || 2202 strncmp(name, 2203 VDEV_TYPE_REPLACING, strlen(VDEV_TYPE_REPLACING)) == 0 || 2204 strncmp(name, VDEV_TYPE_MIRROR, strlen(VDEV_TYPE_MIRROR)) == 0) 2205 return (B_TRUE); 2206 return (B_FALSE); 2207} 2208 2209nvlist_t * 2210zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare, 2211 boolean_t *l2cache, boolean_t *log) 2212{ 2213 char buf[MAXPATHLEN]; 2214 char *end; 2215 nvlist_t *nvroot, *search, *ret; 2216 uint64_t guid; 2217 2218 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0); 2219 2220 guid = strtoull(path, &end, 10); 2221 if (guid != 0 && *end == '\0') { 2222 verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0); 2223 } else if (zpool_vdev_is_interior(path)) { 2224 verify(nvlist_add_string(search, ZPOOL_CONFIG_TYPE, path) == 0); 2225 } else if (path[0] != '/') { 2226 (void) snprintf(buf, sizeof (buf), "%s%s", _PATH_DEV, path); 2227 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, buf) == 0); 2228 } else { 2229 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, path) == 0); 2230 } 2231 2232 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE, 2233 &nvroot) == 0); 2234 2235 *avail_spare = B_FALSE; 2236 *l2cache = B_FALSE; 2237 if (log != NULL) 2238 *log = B_FALSE; 2239 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log); 2240 nvlist_free(search); 2241 2242 return (ret); 2243} 2244 2245static int 2246vdev_online(nvlist_t *nv) 2247{ 2248 uint64_t ival; 2249 2250 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE, &ival) == 0 || 2251 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED, &ival) == 0 || 2252 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED, &ival) == 0) 2253 return (0); 2254 2255 return (1); 2256} 2257 2258/* 2259 * Helper function for zpool_get_physpaths(). 2260 */ 2261static int 2262vdev_get_one_physpath(nvlist_t *config, char *physpath, size_t physpath_size, 2263 size_t *bytes_written) 2264{ 2265 size_t bytes_left, pos, rsz; 2266 char *tmppath; 2267 const char *format; 2268 2269 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PHYS_PATH, 2270 &tmppath) != 0) 2271 return (EZFS_NODEVICE); 2272 2273 pos = *bytes_written; 2274 bytes_left = physpath_size - pos; 2275 format = (pos == 0) ? "%s" : " %s"; 2276 2277 rsz = snprintf(physpath + pos, bytes_left, format, tmppath); 2278 *bytes_written += rsz; 2279 2280 if (rsz >= bytes_left) { 2281 /* if physpath was not copied properly, clear it */ 2282 if (bytes_left != 0) { 2283 physpath[pos] = 0; 2284 } 2285 return (EZFS_NOSPC); 2286 } 2287 return (0); 2288} 2289 2290static int 2291vdev_get_physpaths(nvlist_t *nv, char *physpath, size_t phypath_size, 2292 size_t *rsz, boolean_t is_spare) 2293{ 2294 char *type; 2295 int ret; 2296 2297 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0) 2298 return (EZFS_INVALCONFIG); 2299 2300 if (strcmp(type, VDEV_TYPE_DISK) == 0) { 2301 /* 2302 * An active spare device has ZPOOL_CONFIG_IS_SPARE set. 2303 * For a spare vdev, we only want to boot from the active 2304 * spare device. 2305 */ 2306 if (is_spare) { 2307 uint64_t spare = 0; 2308 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE, 2309 &spare); 2310 if (!spare) 2311 return (EZFS_INVALCONFIG); 2312 } 2313 2314 if (vdev_online(nv)) { 2315 if ((ret = vdev_get_one_physpath(nv, physpath, 2316 phypath_size, rsz)) != 0) 2317 return (ret); 2318 } 2319 } else if (strcmp(type, VDEV_TYPE_MIRROR) == 0 || 2320 strcmp(type, VDEV_TYPE_RAIDZ) == 0 || 2321 strcmp(type, VDEV_TYPE_REPLACING) == 0 || 2322 (is_spare = (strcmp(type, VDEV_TYPE_SPARE) == 0))) { 2323 nvlist_t **child; 2324 uint_t count; 2325 int i, ret; 2326 2327 if (nvlist_lookup_nvlist_array(nv, 2328 ZPOOL_CONFIG_CHILDREN, &child, &count) != 0) 2329 return (EZFS_INVALCONFIG); 2330 2331 for (i = 0; i < count; i++) { 2332 ret = vdev_get_physpaths(child[i], physpath, 2333 phypath_size, rsz, is_spare); 2334 if (ret == EZFS_NOSPC) 2335 return (ret); 2336 } 2337 } 2338 2339 return (EZFS_POOL_INVALARG); 2340} 2341 2342/* 2343 * Get phys_path for a root pool config. 2344 * Return 0 on success; non-zero on failure. 2345 */ 2346static int 2347zpool_get_config_physpath(nvlist_t *config, char *physpath, size_t phypath_size) 2348{ 2349 size_t rsz; 2350 nvlist_t *vdev_root; 2351 nvlist_t **child; 2352 uint_t count; 2353 char *type; 2354 2355 rsz = 0; 2356 2357 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 2358 &vdev_root) != 0) 2359 return (EZFS_INVALCONFIG); 2360 2361 if (nvlist_lookup_string(vdev_root, ZPOOL_CONFIG_TYPE, &type) != 0 || 2362 nvlist_lookup_nvlist_array(vdev_root, ZPOOL_CONFIG_CHILDREN, 2363 &child, &count) != 0) 2364 return (EZFS_INVALCONFIG); 2365 2366 /* 2367 * root pool can only have a single top-level vdev. 2368 */ 2369 if (strcmp(type, VDEV_TYPE_ROOT) != 0 || count != 1) 2370 return (EZFS_POOL_INVALARG); 2371 2372 (void) vdev_get_physpaths(child[0], physpath, phypath_size, &rsz, 2373 B_FALSE); 2374 2375 /* No online devices */ 2376 if (rsz == 0) 2377 return (EZFS_NODEVICE); 2378 2379 return (0); 2380} 2381 2382/* 2383 * Get phys_path for a root pool 2384 * Return 0 on success; non-zero on failure. 2385 */ 2386int 2387zpool_get_physpath(zpool_handle_t *zhp, char *physpath, size_t phypath_size) 2388{ 2389 return (zpool_get_config_physpath(zhp->zpool_config, physpath, 2390 phypath_size)); 2391} 2392 2393/* 2394 * If the device has being dynamically expanded then we need to relabel 2395 * the disk to use the new unallocated space. 2396 */ 2397static int 2398zpool_relabel_disk(libzfs_handle_t *hdl, const char *name) 2399{ 2400#ifdef illumos 2401 char path[MAXPATHLEN]; 2402 char errbuf[1024]; 2403 int fd, error; 2404 int (*_efi_use_whole_disk)(int); 2405 2406 if ((_efi_use_whole_disk = (int (*)(int))dlsym(RTLD_DEFAULT, 2407 "efi_use_whole_disk")) == NULL) 2408 return (-1); 2409 2410 (void) snprintf(path, sizeof (path), "%s/%s", ZFS_RDISK_ROOT, name); 2411 2412 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) { 2413 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot " 2414 "relabel '%s': unable to open device"), name); 2415 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf)); 2416 } 2417 2418 /* 2419 * It's possible that we might encounter an error if the device 2420 * does not have any unallocated space left. If so, we simply 2421 * ignore that error and continue on. 2422 */ 2423 error = _efi_use_whole_disk(fd); 2424 (void) close(fd); 2425 if (error && error != VT_ENOSPC) { 2426 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot " 2427 "relabel '%s': unable to read disk capacity"), name); 2428 return (zfs_error(hdl, EZFS_NOCAP, errbuf)); 2429 } 2430#endif /* illumos */ 2431 return (0); 2432} 2433 2434/* 2435 * Bring the specified vdev online. The 'flags' parameter is a set of the 2436 * ZFS_ONLINE_* flags. 2437 */ 2438int 2439zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags, 2440 vdev_state_t *newstate) 2441{ 2442 zfs_cmd_t zc = { 0 }; 2443 char msg[1024]; 2444 char *pathname; 2445 nvlist_t *tgt; 2446 boolean_t avail_spare, l2cache, islog; 2447 libzfs_handle_t *hdl = zhp->zpool_hdl; 2448 2449 if (flags & ZFS_ONLINE_EXPAND) { 2450 (void) snprintf(msg, sizeof (msg), 2451 dgettext(TEXT_DOMAIN, "cannot expand %s"), path); 2452 } else { 2453 (void) snprintf(msg, sizeof (msg), 2454 dgettext(TEXT_DOMAIN, "cannot online %s"), path); 2455 } 2456 2457 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2458 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 2459 &islog)) == NULL) 2460 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 2461 2462 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 2463 2464 if (avail_spare) 2465 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 2466 2467 if ((flags & ZFS_ONLINE_EXPAND || 2468 zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) && 2469 nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH, &pathname) == 0) { 2470 uint64_t wholedisk = 0; 2471 2472 (void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK, 2473 &wholedisk); 2474 2475 /* 2476 * XXX - L2ARC 1.0 devices can't support expansion. 2477 */ 2478 if (l2cache) { 2479 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2480 "cannot expand cache devices")); 2481 return (zfs_error(hdl, EZFS_VDEVNOTSUP, msg)); 2482 } 2483 2484 if (wholedisk) { 2485 pathname += strlen(ZFS_DISK_ROOT) + 1; 2486 (void) zpool_relabel_disk(hdl, pathname); 2487 } 2488 } 2489 2490 zc.zc_cookie = VDEV_STATE_ONLINE; 2491 zc.zc_obj = flags; 2492 2493 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0) { 2494 if (errno == EINVAL) { 2495 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "was split " 2496 "from this pool into a new one. Use '%s' " 2497 "instead"), "zpool detach"); 2498 return (zfs_error(hdl, EZFS_POSTSPLIT_ONLINE, msg)); 2499 } 2500 return (zpool_standard_error(hdl, errno, msg)); 2501 } 2502 2503 *newstate = zc.zc_cookie; 2504 return (0); 2505} 2506 2507/* 2508 * Take the specified vdev offline 2509 */ 2510int 2511zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp) 2512{ 2513 zfs_cmd_t zc = { 0 }; 2514 char msg[1024]; 2515 nvlist_t *tgt; 2516 boolean_t avail_spare, l2cache; 2517 libzfs_handle_t *hdl = zhp->zpool_hdl; 2518 2519 (void) snprintf(msg, sizeof (msg), 2520 dgettext(TEXT_DOMAIN, "cannot offline %s"), path); 2521 2522 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2523 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 2524 NULL)) == NULL) 2525 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 2526 2527 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 2528 2529 if (avail_spare) 2530 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 2531 2532 zc.zc_cookie = VDEV_STATE_OFFLINE; 2533 zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0; 2534 2535 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0) 2536 return (0); 2537 2538 switch (errno) { 2539 case EBUSY: 2540 2541 /* 2542 * There are no other replicas of this device. 2543 */ 2544 return (zfs_error(hdl, EZFS_NOREPLICAS, msg)); 2545 2546 case EEXIST: 2547 /* 2548 * The log device has unplayed logs 2549 */ 2550 return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, msg)); 2551 2552 default: 2553 return (zpool_standard_error(hdl, errno, msg)); 2554 } 2555} 2556 2557/* 2558 * Mark the given vdev faulted. 2559 */ 2560int 2561zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux) 2562{ 2563 zfs_cmd_t zc = { 0 }; 2564 char msg[1024]; 2565 libzfs_handle_t *hdl = zhp->zpool_hdl; 2566 2567 (void) snprintf(msg, sizeof (msg), 2568 dgettext(TEXT_DOMAIN, "cannot fault %llu"), guid); 2569 2570 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2571 zc.zc_guid = guid; 2572 zc.zc_cookie = VDEV_STATE_FAULTED; 2573 zc.zc_obj = aux; 2574 2575 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0) 2576 return (0); 2577 2578 switch (errno) { 2579 case EBUSY: 2580 2581 /* 2582 * There are no other replicas of this device. 2583 */ 2584 return (zfs_error(hdl, EZFS_NOREPLICAS, msg)); 2585 2586 default: 2587 return (zpool_standard_error(hdl, errno, msg)); 2588 } 2589 2590} 2591 2592/* 2593 * Mark the given vdev degraded. 2594 */ 2595int 2596zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux) 2597{ 2598 zfs_cmd_t zc = { 0 }; 2599 char msg[1024]; 2600 libzfs_handle_t *hdl = zhp->zpool_hdl; 2601 2602 (void) snprintf(msg, sizeof (msg), 2603 dgettext(TEXT_DOMAIN, "cannot degrade %llu"), guid); 2604 2605 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2606 zc.zc_guid = guid; 2607 zc.zc_cookie = VDEV_STATE_DEGRADED; 2608 zc.zc_obj = aux; 2609 2610 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0) 2611 return (0); 2612 2613 return (zpool_standard_error(hdl, errno, msg)); 2614} 2615 2616/* 2617 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as 2618 * a hot spare. 2619 */ 2620static boolean_t 2621is_replacing_spare(nvlist_t *search, nvlist_t *tgt, int which) 2622{ 2623 nvlist_t **child; 2624 uint_t c, children; 2625 char *type; 2626 2627 if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child, 2628 &children) == 0) { 2629 verify(nvlist_lookup_string(search, ZPOOL_CONFIG_TYPE, 2630 &type) == 0); 2631 2632 if (strcmp(type, VDEV_TYPE_SPARE) == 0 && 2633 children == 2 && child[which] == tgt) 2634 return (B_TRUE); 2635 2636 for (c = 0; c < children; c++) 2637 if (is_replacing_spare(child[c], tgt, which)) 2638 return (B_TRUE); 2639 } 2640 2641 return (B_FALSE); 2642} 2643 2644/* 2645 * Attach new_disk (fully described by nvroot) to old_disk. 2646 * If 'replacing' is specified, the new disk will replace the old one. 2647 */ 2648int 2649zpool_vdev_attach(zpool_handle_t *zhp, 2650 const char *old_disk, const char *new_disk, nvlist_t *nvroot, int replacing) 2651{ 2652 zfs_cmd_t zc = { 0 }; 2653 char msg[1024]; 2654 int ret; 2655 nvlist_t *tgt; 2656 boolean_t avail_spare, l2cache, islog; 2657 uint64_t val; 2658 char *newname; 2659 nvlist_t **child; 2660 uint_t children; 2661 nvlist_t *config_root; 2662 libzfs_handle_t *hdl = zhp->zpool_hdl; 2663 boolean_t rootpool = zpool_is_bootable(zhp); 2664 2665 if (replacing) 2666 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 2667 "cannot replace %s with %s"), old_disk, new_disk); 2668 else 2669 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, 2670 "cannot attach %s to %s"), new_disk, old_disk); 2671 2672 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2673 if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache, 2674 &islog)) == NULL) 2675 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 2676 2677 if (avail_spare) 2678 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 2679 2680 if (l2cache) 2681 return (zfs_error(hdl, EZFS_ISL2CACHE, msg)); 2682 2683 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 2684 zc.zc_cookie = replacing; 2685 2686 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, 2687 &child, &children) != 0 || children != 1) { 2688 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2689 "new device must be a single disk")); 2690 return (zfs_error(hdl, EZFS_INVALCONFIG, msg)); 2691 } 2692 2693 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL), 2694 ZPOOL_CONFIG_VDEV_TREE, &config_root) == 0); 2695 2696 if ((newname = zpool_vdev_name(NULL, NULL, child[0], B_FALSE)) == NULL) 2697 return (-1); 2698 2699 /* 2700 * If the target is a hot spare that has been swapped in, we can only 2701 * replace it with another hot spare. 2702 */ 2703 if (replacing && 2704 nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 && 2705 (zpool_find_vdev(zhp, newname, &avail_spare, &l2cache, 2706 NULL) == NULL || !avail_spare) && 2707 is_replacing_spare(config_root, tgt, 1)) { 2708 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2709 "can only be replaced by another hot spare")); 2710 free(newname); 2711 return (zfs_error(hdl, EZFS_BADTARGET, msg)); 2712 } 2713 2714 free(newname); 2715 2716 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0) 2717 return (-1); 2718 2719 ret = zfs_ioctl(hdl, ZFS_IOC_VDEV_ATTACH, &zc); 2720 2721 zcmd_free_nvlists(&zc); 2722 2723 if (ret == 0) { 2724 if (rootpool) { 2725 /* 2726 * XXX need a better way to prevent user from 2727 * booting up a half-baked vdev. 2728 */ 2729 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Make " 2730 "sure to wait until resilver is done " 2731 "before rebooting.\n")); 2732 (void) fprintf(stderr, "\n"); 2733 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "If " 2734 "you boot from pool '%s', you may need to update\n" 2735 "boot code on newly attached disk '%s'.\n\n" 2736 "Assuming you use GPT partitioning and 'da0' is " 2737 "your new boot disk\n" 2738 "you may use the following command:\n\n" 2739 "\tgpart bootcode -b /boot/pmbr -p " 2740 "/boot/gptzfsboot -i 1 da0\n\n"), 2741 zhp->zpool_name, new_disk); 2742 } 2743 return (0); 2744 } 2745 2746 switch (errno) { 2747 case ENOTSUP: 2748 /* 2749 * Can't attach to or replace this type of vdev. 2750 */ 2751 if (replacing) { 2752 uint64_t version = zpool_get_prop_int(zhp, 2753 ZPOOL_PROP_VERSION, NULL); 2754 2755 if (islog) 2756 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2757 "cannot replace a log with a spare")); 2758 else if (version >= SPA_VERSION_MULTI_REPLACE) 2759 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2760 "already in replacing/spare config; wait " 2761 "for completion or use 'zpool detach'")); 2762 else 2763 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2764 "cannot replace a replacing device")); 2765 } else { 2766 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2767 "can only attach to mirrors and top-level " 2768 "disks")); 2769 } 2770 (void) zfs_error(hdl, EZFS_BADTARGET, msg); 2771 break; 2772 2773 case EINVAL: 2774 /* 2775 * The new device must be a single disk. 2776 */ 2777 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2778 "new device must be a single disk")); 2779 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg); 2780 break; 2781 2782 case EBUSY: 2783 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy, " 2784 "or pool has removing/removed vdevs"), 2785 new_disk); 2786 (void) zfs_error(hdl, EZFS_BADDEV, msg); 2787 break; 2788 2789 case EOVERFLOW: 2790 /* 2791 * The new device is too small. 2792 */ 2793 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2794 "device is too small")); 2795 (void) zfs_error(hdl, EZFS_BADDEV, msg); 2796 break; 2797 2798 case EDOM: 2799 /* 2800 * The new device has a different alignment requirement. 2801 */ 2802 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2803 "devices have different sector alignment")); 2804 (void) zfs_error(hdl, EZFS_BADDEV, msg); 2805 break; 2806 2807 case ENAMETOOLONG: 2808 /* 2809 * The resulting top-level vdev spec won't fit in the label. 2810 */ 2811 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, msg); 2812 break; 2813 2814 default: 2815 (void) zpool_standard_error(hdl, errno, msg); 2816 } 2817 2818 return (-1); 2819} 2820 2821/* 2822 * Detach the specified device. 2823 */ 2824int 2825zpool_vdev_detach(zpool_handle_t *zhp, const char *path) 2826{ 2827 zfs_cmd_t zc = { 0 }; 2828 char msg[1024]; 2829 nvlist_t *tgt; 2830 boolean_t avail_spare, l2cache; 2831 libzfs_handle_t *hdl = zhp->zpool_hdl; 2832 2833 (void) snprintf(msg, sizeof (msg), 2834 dgettext(TEXT_DOMAIN, "cannot detach %s"), path); 2835 2836 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 2837 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 2838 NULL)) == NULL) 2839 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 2840 2841 if (avail_spare) 2842 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 2843 2844 if (l2cache) 2845 return (zfs_error(hdl, EZFS_ISL2CACHE, msg)); 2846 2847 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); 2848 2849 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0) 2850 return (0); 2851 2852 switch (errno) { 2853 2854 case ENOTSUP: 2855 /* 2856 * Can't detach from this type of vdev. 2857 */ 2858 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only " 2859 "applicable to mirror and replacing vdevs")); 2860 (void) zfs_error(hdl, EZFS_BADTARGET, msg); 2861 break; 2862 2863 case EBUSY: 2864 /* 2865 * There are no other replicas of this device. 2866 */ 2867 (void) zfs_error(hdl, EZFS_NOREPLICAS, msg); 2868 break; 2869 2870 default: 2871 (void) zpool_standard_error(hdl, errno, msg); 2872 } 2873 2874 return (-1); 2875} 2876 2877/* 2878 * Find a mirror vdev in the source nvlist. 2879 * 2880 * The mchild array contains a list of disks in one of the top-level mirrors 2881 * of the source pool. The schild array contains a list of disks that the 2882 * user specified on the command line. We loop over the mchild array to 2883 * see if any entry in the schild array matches. 2884 * 2885 * If a disk in the mchild array is found in the schild array, we return 2886 * the index of that entry. Otherwise we return -1. 2887 */ 2888static int 2889find_vdev_entry(zpool_handle_t *zhp, nvlist_t **mchild, uint_t mchildren, 2890 nvlist_t **schild, uint_t schildren) 2891{ 2892 uint_t mc; 2893 2894 for (mc = 0; mc < mchildren; mc++) { 2895 uint_t sc; 2896 char *mpath = zpool_vdev_name(zhp->zpool_hdl, zhp, 2897 mchild[mc], B_FALSE); 2898 2899 for (sc = 0; sc < schildren; sc++) { 2900 char *spath = zpool_vdev_name(zhp->zpool_hdl, zhp, 2901 schild[sc], B_FALSE); 2902 boolean_t result = (strcmp(mpath, spath) == 0); 2903 2904 free(spath); 2905 if (result) { 2906 free(mpath); 2907 return (mc); 2908 } 2909 } 2910 2911 free(mpath); 2912 } 2913 2914 return (-1); 2915} 2916 2917/* 2918 * Split a mirror pool. If newroot points to null, then a new nvlist 2919 * is generated and it is the responsibility of the caller to free it. 2920 */ 2921int 2922zpool_vdev_split(zpool_handle_t *zhp, char *newname, nvlist_t **newroot, 2923 nvlist_t *props, splitflags_t flags) 2924{ 2925 zfs_cmd_t zc = { 0 }; 2926 char msg[1024]; 2927 nvlist_t *tree, *config, **child, **newchild, *newconfig = NULL; 2928 nvlist_t **varray = NULL, *zc_props = NULL; 2929 uint_t c, children, newchildren, lastlog = 0, vcount, found = 0; 2930 libzfs_handle_t *hdl = zhp->zpool_hdl; 2931 uint64_t vers; 2932 boolean_t freelist = B_FALSE, memory_err = B_TRUE; 2933 int retval = 0; 2934 2935 (void) snprintf(msg, sizeof (msg), 2936 dgettext(TEXT_DOMAIN, "Unable to split %s"), zhp->zpool_name); 2937 2938 if (!zpool_name_valid(hdl, B_FALSE, newname)) 2939 return (zfs_error(hdl, EZFS_INVALIDNAME, msg)); 2940 2941 if ((config = zpool_get_config(zhp, NULL)) == NULL) { 2942 (void) fprintf(stderr, gettext("Internal error: unable to " 2943 "retrieve pool configuration\n")); 2944 return (-1); 2945 } 2946 2947 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree) 2948 == 0); 2949 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &vers) == 0); 2950 2951 if (props) { 2952 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE }; 2953 if ((zc_props = zpool_valid_proplist(hdl, zhp->zpool_name, 2954 props, vers, flags, msg)) == NULL) 2955 return (-1); 2956 } 2957 2958 if (nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child, 2959 &children) != 0) { 2960 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 2961 "Source pool is missing vdev tree")); 2962 nvlist_free(zc_props); 2963 return (-1); 2964 } 2965 2966 varray = zfs_alloc(hdl, children * sizeof (nvlist_t *)); 2967 vcount = 0; 2968 2969 if (*newroot == NULL || 2970 nvlist_lookup_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, 2971 &newchild, &newchildren) != 0) 2972 newchildren = 0; 2973 2974 for (c = 0; c < children; c++) { 2975 uint64_t is_log = B_FALSE, is_hole = B_FALSE; 2976 char *type; 2977 nvlist_t **mchild, *vdev; 2978 uint_t mchildren; 2979 int entry; 2980 2981 /* 2982 * Unlike cache & spares, slogs are stored in the 2983 * ZPOOL_CONFIG_CHILDREN array. We filter them out here. 2984 */ 2985 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG, 2986 &is_log); 2987 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE, 2988 &is_hole); 2989 if (is_log || is_hole) { 2990 /* 2991 * Create a hole vdev and put it in the config. 2992 */ 2993 if (nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) != 0) 2994 goto out; 2995 if (nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE, 2996 VDEV_TYPE_HOLE) != 0) 2997 goto out; 2998 if (nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_HOLE, 2999 1) != 0) 3000 goto out; 3001 if (lastlog == 0) 3002 lastlog = vcount; 3003 varray[vcount++] = vdev; 3004 continue; 3005 } 3006 lastlog = 0; 3007 verify(nvlist_lookup_string(child[c], ZPOOL_CONFIG_TYPE, &type) 3008 == 0); 3009 if (strcmp(type, VDEV_TYPE_MIRROR) != 0) { 3010 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3011 "Source pool must be composed only of mirrors\n")); 3012 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg); 3013 goto out; 3014 } 3015 3016 verify(nvlist_lookup_nvlist_array(child[c], 3017 ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0); 3018 3019 /* find or add an entry for this top-level vdev */ 3020 if (newchildren > 0 && 3021 (entry = find_vdev_entry(zhp, mchild, mchildren, 3022 newchild, newchildren)) >= 0) { 3023 /* We found a disk that the user specified. */ 3024 vdev = mchild[entry]; 3025 ++found; 3026 } else { 3027 /* User didn't specify a disk for this vdev. */ 3028 vdev = mchild[mchildren - 1]; 3029 } 3030 3031 if (nvlist_dup(vdev, &varray[vcount++], 0) != 0) 3032 goto out; 3033 } 3034 3035 /* did we find every disk the user specified? */ 3036 if (found != newchildren) { 3037 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "Device list must " 3038 "include at most one disk from each mirror")); 3039 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg); 3040 goto out; 3041 } 3042 3043 /* Prepare the nvlist for populating. */ 3044 if (*newroot == NULL) { 3045 if (nvlist_alloc(newroot, NV_UNIQUE_NAME, 0) != 0) 3046 goto out; 3047 freelist = B_TRUE; 3048 if (nvlist_add_string(*newroot, ZPOOL_CONFIG_TYPE, 3049 VDEV_TYPE_ROOT) != 0) 3050 goto out; 3051 } else { 3052 verify(nvlist_remove_all(*newroot, ZPOOL_CONFIG_CHILDREN) == 0); 3053 } 3054 3055 /* Add all the children we found */ 3056 if (nvlist_add_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, varray, 3057 lastlog == 0 ? vcount : lastlog) != 0) 3058 goto out; 3059 3060 /* 3061 * If we're just doing a dry run, exit now with success. 3062 */ 3063 if (flags.dryrun) { 3064 memory_err = B_FALSE; 3065 freelist = B_FALSE; 3066 goto out; 3067 } 3068 3069 /* now build up the config list & call the ioctl */ 3070 if (nvlist_alloc(&newconfig, NV_UNIQUE_NAME, 0) != 0) 3071 goto out; 3072 3073 if (nvlist_add_nvlist(newconfig, 3074 ZPOOL_CONFIG_VDEV_TREE, *newroot) != 0 || 3075 nvlist_add_string(newconfig, 3076 ZPOOL_CONFIG_POOL_NAME, newname) != 0 || 3077 nvlist_add_uint64(newconfig, ZPOOL_CONFIG_VERSION, vers) != 0) 3078 goto out; 3079 3080 /* 3081 * The new pool is automatically part of the namespace unless we 3082 * explicitly export it. 3083 */ 3084 if (!flags.import) 3085 zc.zc_cookie = ZPOOL_EXPORT_AFTER_SPLIT; 3086 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3087 (void) strlcpy(zc.zc_string, newname, sizeof (zc.zc_string)); 3088 if (zcmd_write_conf_nvlist(hdl, &zc, newconfig) != 0) 3089 goto out; 3090 if (zc_props != NULL && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0) 3091 goto out; 3092 3093 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SPLIT, &zc) != 0) { 3094 retval = zpool_standard_error(hdl, errno, msg); 3095 goto out; 3096 } 3097 3098 freelist = B_FALSE; 3099 memory_err = B_FALSE; 3100 3101out: 3102 if (varray != NULL) { 3103 int v; 3104 3105 for (v = 0; v < vcount; v++) 3106 nvlist_free(varray[v]); 3107 free(varray); 3108 } 3109 zcmd_free_nvlists(&zc); 3110 nvlist_free(zc_props); 3111 nvlist_free(newconfig); 3112 if (freelist) { 3113 nvlist_free(*newroot); 3114 *newroot = NULL; 3115 } 3116 3117 if (retval != 0) 3118 return (retval); 3119 3120 if (memory_err) 3121 return (no_memory(hdl)); 3122 3123 return (0); 3124} 3125 3126/* 3127 * Remove the given device. 3128 */ 3129int 3130zpool_vdev_remove(zpool_handle_t *zhp, const char *path) 3131{ 3132 zfs_cmd_t zc = { 0 }; 3133 char msg[1024]; 3134 nvlist_t *tgt; 3135 boolean_t avail_spare, l2cache, islog; 3136 libzfs_handle_t *hdl = zhp->zpool_hdl; 3137 uint64_t version; 3138 3139 (void) snprintf(msg, sizeof (msg), 3140 dgettext(TEXT_DOMAIN, "cannot remove %s"), path); 3141 3142 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3143 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 3144 &islog)) == NULL) 3145 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 3146 3147 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL); 3148 if (islog && version < SPA_VERSION_HOLES) { 3149 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3150 "pool must be upgraded to support log removal")); 3151 return (zfs_error(hdl, EZFS_BADVERSION, msg)); 3152 } 3153 3154 if (!islog && !avail_spare && !l2cache && zpool_is_bootable(zhp)) { 3155 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3156 "root pool can not have removed devices, " 3157 "because GRUB does not understand them")); 3158 return (zfs_error(hdl, EINVAL, msg)); 3159 } 3160 3161 zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID); 3162 3163 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0) 3164 return (0); 3165 3166 switch (errno) { 3167 3168 case EINVAL: 3169 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3170 "invalid config; all top-level vdevs must " 3171 "have the same sector size and not be raidz.")); 3172 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg); 3173 break; 3174 3175 case EBUSY: 3176 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3177 "Pool busy; removal may already be in progress")); 3178 (void) zfs_error(hdl, EZFS_BUSY, msg); 3179 break; 3180 3181 default: 3182 (void) zpool_standard_error(hdl, errno, msg); 3183 } 3184 return (-1); 3185} 3186 3187int 3188zpool_vdev_remove_cancel(zpool_handle_t *zhp) 3189{ 3190 zfs_cmd_t zc = { 0 }; 3191 char msg[1024]; 3192 libzfs_handle_t *hdl = zhp->zpool_hdl; 3193 3194 (void) snprintf(msg, sizeof (msg), 3195 dgettext(TEXT_DOMAIN, "cannot cancel removal")); 3196 3197 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3198 zc.zc_cookie = 1; 3199 3200 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0) 3201 return (0); 3202 3203 return (zpool_standard_error(hdl, errno, msg)); 3204} 3205 3206int 3207zpool_vdev_indirect_size(zpool_handle_t *zhp, const char *path, 3208 uint64_t *sizep) 3209{ 3210 char msg[1024]; 3211 nvlist_t *tgt; 3212 boolean_t avail_spare, l2cache, islog; 3213 libzfs_handle_t *hdl = zhp->zpool_hdl; 3214 3215 (void) snprintf(msg, sizeof (msg), 3216 dgettext(TEXT_DOMAIN, "cannot determine indirect size of %s"), 3217 path); 3218 3219 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, 3220 &islog)) == NULL) 3221 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 3222 3223 if (avail_spare || l2cache || islog) { 3224 *sizep = 0; 3225 return (0); 3226 } 3227 3228 if (nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_INDIRECT_SIZE, sizep) != 0) { 3229 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 3230 "indirect size not available")); 3231 return (zfs_error(hdl, EINVAL, msg)); 3232 } 3233 return (0); 3234} 3235 3236/* 3237 * Clear the errors for the pool, or the particular device if specified. 3238 */ 3239int 3240zpool_clear(zpool_handle_t *zhp, const char *path, nvlist_t *rewindnvl) 3241{ 3242 zfs_cmd_t zc = { 0 }; 3243 char msg[1024]; 3244 nvlist_t *tgt; 3245 zpool_rewind_policy_t policy; 3246 boolean_t avail_spare, l2cache; 3247 libzfs_handle_t *hdl = zhp->zpool_hdl; 3248 nvlist_t *nvi = NULL; 3249 int error; 3250 3251 if (path) 3252 (void) snprintf(msg, sizeof (msg), 3253 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"), 3254 path); 3255 else 3256 (void) snprintf(msg, sizeof (msg), 3257 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"), 3258 zhp->zpool_name); 3259 3260 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3261 if (path) { 3262 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, 3263 &l2cache, NULL)) == NULL) 3264 return (zfs_error(hdl, EZFS_NODEVICE, msg)); 3265 3266 /* 3267 * Don't allow error clearing for hot spares. Do allow 3268 * error clearing for l2cache devices. 3269 */ 3270 if (avail_spare) 3271 return (zfs_error(hdl, EZFS_ISSPARE, msg)); 3272 3273 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, 3274 &zc.zc_guid) == 0); 3275 } 3276 3277 zpool_get_rewind_policy(rewindnvl, &policy); 3278 zc.zc_cookie = policy.zrp_request; 3279 3280 if (zcmd_alloc_dst_nvlist(hdl, &zc, zhp->zpool_config_size * 2) != 0) 3281 return (-1); 3282 3283 if (zcmd_write_src_nvlist(hdl, &zc, rewindnvl) != 0) 3284 return (-1); 3285 3286 while ((error = zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc)) != 0 && 3287 errno == ENOMEM) { 3288 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) { 3289 zcmd_free_nvlists(&zc); 3290 return (-1); 3291 } 3292 } 3293 3294 if (!error || ((policy.zrp_request & ZPOOL_TRY_REWIND) && 3295 errno != EPERM && errno != EACCES)) { 3296 if (policy.zrp_request & 3297 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) { 3298 (void) zcmd_read_dst_nvlist(hdl, &zc, &nvi); 3299 zpool_rewind_exclaim(hdl, zc.zc_name, 3300 ((policy.zrp_request & ZPOOL_TRY_REWIND) != 0), 3301 nvi); 3302 nvlist_free(nvi); 3303 } 3304 zcmd_free_nvlists(&zc); 3305 return (0); 3306 } 3307 3308 zcmd_free_nvlists(&zc); 3309 return (zpool_standard_error(hdl, errno, msg)); 3310} 3311 3312/* 3313 * Similar to zpool_clear(), but takes a GUID (used by fmd). 3314 */ 3315int 3316zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid) 3317{ 3318 zfs_cmd_t zc = { 0 }; 3319 char msg[1024]; 3320 libzfs_handle_t *hdl = zhp->zpool_hdl; 3321 3322 (void) snprintf(msg, sizeof (msg), 3323 dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"), 3324 guid); 3325 3326 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3327 zc.zc_guid = guid; 3328 zc.zc_cookie = ZPOOL_NO_REWIND; 3329 3330 if (ioctl(hdl->libzfs_fd, ZFS_IOC_CLEAR, &zc) == 0) 3331 return (0); 3332 3333 return (zpool_standard_error(hdl, errno, msg)); 3334} 3335 3336/* 3337 * Change the GUID for a pool. 3338 */ 3339int 3340zpool_reguid(zpool_handle_t *zhp) 3341{ 3342 char msg[1024]; 3343 libzfs_handle_t *hdl = zhp->zpool_hdl; 3344 zfs_cmd_t zc = { 0 }; 3345 3346 (void) snprintf(msg, sizeof (msg), 3347 dgettext(TEXT_DOMAIN, "cannot reguid '%s'"), zhp->zpool_name); 3348 3349 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3350 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REGUID, &zc) == 0) 3351 return (0); 3352 3353 return (zpool_standard_error(hdl, errno, msg)); 3354} 3355 3356/* 3357 * Reopen the pool. 3358 */ 3359int 3360zpool_reopen(zpool_handle_t *zhp) 3361{ 3362 zfs_cmd_t zc = { 0 }; 3363 char msg[1024]; 3364 libzfs_handle_t *hdl = zhp->zpool_hdl; 3365 3366 (void) snprintf(msg, sizeof (msg), 3367 dgettext(TEXT_DOMAIN, "cannot reopen '%s'"), 3368 zhp->zpool_name); 3369 3370 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3371 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REOPEN, &zc) == 0) 3372 return (0); 3373 return (zpool_standard_error(hdl, errno, msg)); 3374} 3375 3376/* 3377 * Convert from a devid string to a path. 3378 */ 3379static char * 3380devid_to_path(char *devid_str) 3381{ 3382 ddi_devid_t devid; 3383 char *minor; 3384 char *path; 3385 devid_nmlist_t *list = NULL; 3386 int ret; 3387 3388 if (devid_str_decode(devid_str, &devid, &minor) != 0) 3389 return (NULL); 3390 3391 ret = devid_deviceid_to_nmlist("/dev", devid, minor, &list); 3392 3393 devid_str_free(minor); 3394 devid_free(devid); 3395 3396 if (ret != 0) 3397 return (NULL); 3398 3399 /* 3400 * In a case the strdup() fails, we will just return NULL below. 3401 */ 3402 path = strdup(list[0].devname); 3403 3404 devid_free_nmlist(list); 3405 3406 return (path); 3407} 3408 3409/* 3410 * Convert from a path to a devid string. 3411 */ 3412static char * 3413path_to_devid(const char *path) 3414{ 3415#ifdef have_devid 3416 int fd; 3417 ddi_devid_t devid; 3418 char *minor, *ret; 3419 3420 if ((fd = open(path, O_RDONLY)) < 0) 3421 return (NULL); 3422 3423 minor = NULL; 3424 ret = NULL; 3425 if (devid_get(fd, &devid) == 0) { 3426 if (devid_get_minor_name(fd, &minor) == 0) 3427 ret = devid_str_encode(devid, minor); 3428 if (minor != NULL) 3429 devid_str_free(minor); 3430 devid_free(devid); 3431 } 3432 (void) close(fd); 3433 3434 return (ret); 3435#else 3436 return (NULL); 3437#endif 3438} 3439 3440/* 3441 * Issue the necessary ioctl() to update the stored path value for the vdev. We 3442 * ignore any failure here, since a common case is for an unprivileged user to 3443 * type 'zpool status', and we'll display the correct information anyway. 3444 */ 3445static void 3446set_path(zpool_handle_t *zhp, nvlist_t *nv, const char *path) 3447{ 3448 zfs_cmd_t zc = { 0 }; 3449 3450 (void) strncpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3451 (void) strncpy(zc.zc_value, path, sizeof (zc.zc_value)); 3452 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, 3453 &zc.zc_guid) == 0); 3454 3455 (void) ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SETPATH, &zc); 3456} 3457 3458/* 3459 * Given a vdev, return the name to display in iostat. If the vdev has a path, 3460 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type. 3461 * We also check if this is a whole disk, in which case we strip off the 3462 * trailing 's0' slice name. 3463 * 3464 * This routine is also responsible for identifying when disks have been 3465 * reconfigured in a new location. The kernel will have opened the device by 3466 * devid, but the path will still refer to the old location. To catch this, we 3467 * first do a path -> devid translation (which is fast for the common case). If 3468 * the devid matches, we're done. If not, we do a reverse devid -> path 3469 * translation and issue the appropriate ioctl() to update the path of the vdev. 3470 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any 3471 * of these checks. 3472 */ 3473char * 3474zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv, 3475 boolean_t verbose) 3476{ 3477 char *path, *devid; 3478 uint64_t value; 3479 char buf[64]; 3480 vdev_stat_t *vs; 3481 uint_t vsc; 3482 int have_stats; 3483 int have_path; 3484 3485 have_stats = nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS, 3486 (uint64_t **)&vs, &vsc) == 0; 3487 have_path = nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0; 3488 3489 /* 3490 * If the device is not currently present, assume it will not 3491 * come back at the same device path. Display the device by GUID. 3492 */ 3493 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, &value) == 0 || 3494 have_path && have_stats && vs->vs_state <= VDEV_STATE_CANT_OPEN) { 3495 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, 3496 &value) == 0); 3497 (void) snprintf(buf, sizeof (buf), "%llu", 3498 (u_longlong_t)value); 3499 path = buf; 3500 } else if (have_path) { 3501 3502 /* 3503 * If the device is dead (faulted, offline, etc) then don't 3504 * bother opening it. Otherwise we may be forcing the user to 3505 * open a misbehaving device, which can have undesirable 3506 * effects. 3507 */ 3508 if ((have_stats == 0 || 3509 vs->vs_state >= VDEV_STATE_DEGRADED) && 3510 zhp != NULL && 3511 nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &devid) == 0) { 3512 /* 3513 * Determine if the current path is correct. 3514 */ 3515 char *newdevid = path_to_devid(path); 3516 3517 if (newdevid == NULL || 3518 strcmp(devid, newdevid) != 0) { 3519 char *newpath; 3520 3521 if ((newpath = devid_to_path(devid)) != NULL) { 3522 /* 3523 * Update the path appropriately. 3524 */ 3525 set_path(zhp, nv, newpath); 3526 if (nvlist_add_string(nv, 3527 ZPOOL_CONFIG_PATH, newpath) == 0) 3528 verify(nvlist_lookup_string(nv, 3529 ZPOOL_CONFIG_PATH, 3530 &path) == 0); 3531 free(newpath); 3532 } 3533 } 3534 3535 if (newdevid) 3536 devid_str_free(newdevid); 3537 } 3538 3539#ifdef illumos 3540 if (strncmp(path, ZFS_DISK_ROOTD, strlen(ZFS_DISK_ROOTD)) == 0) 3541 path += strlen(ZFS_DISK_ROOTD); 3542 3543 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, 3544 &value) == 0 && value) { 3545 int pathlen = strlen(path); 3546 char *tmp = zfs_strdup(hdl, path); 3547 3548 /* 3549 * If it starts with c#, and ends with "s0" or "s1", 3550 * chop the slice off, or if it ends with "s0/old" or 3551 * "s1/old", remove the slice from the middle. 3552 */ 3553 if (CTD_CHECK(tmp)) { 3554 if (strcmp(&tmp[pathlen - 2], "s0") == 0 || 3555 strcmp(&tmp[pathlen - 2], "s1") == 0) { 3556 tmp[pathlen - 2] = '\0'; 3557 } else if (pathlen > 6 && 3558 (strcmp(&tmp[pathlen - 6], "s0/old") == 0 || 3559 strcmp(&tmp[pathlen - 6], "s1/old") == 0)) { 3560 (void) strcpy(&tmp[pathlen - 6], 3561 "/old"); 3562 } 3563 } 3564 return (tmp); 3565 } 3566#else /* !illumos */ 3567 if (strncmp(path, _PATH_DEV, sizeof(_PATH_DEV) - 1) == 0) 3568 path += sizeof(_PATH_DEV) - 1; 3569#endif /* illumos */ 3570 } else { 3571 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &path) == 0); 3572 3573 /* 3574 * If it's a raidz device, we need to stick in the parity level. 3575 */ 3576 if (strcmp(path, VDEV_TYPE_RAIDZ) == 0) { 3577 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY, 3578 &value) == 0); 3579 (void) snprintf(buf, sizeof (buf), "%s%llu", path, 3580 (u_longlong_t)value); 3581 path = buf; 3582 } 3583 3584 /* 3585 * We identify each top-level vdev by using a <type-id> 3586 * naming convention. 3587 */ 3588 if (verbose) { 3589 uint64_t id; 3590 3591 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID, 3592 &id) == 0); 3593 (void) snprintf(buf, sizeof (buf), "%s-%llu", path, 3594 (u_longlong_t)id); 3595 path = buf; 3596 } 3597 } 3598 3599 return (zfs_strdup(hdl, path)); 3600} 3601 3602static int 3603zbookmark_mem_compare(const void *a, const void *b) 3604{ 3605 return (memcmp(a, b, sizeof (zbookmark_phys_t))); 3606} 3607 3608/* 3609 * Retrieve the persistent error log, uniquify the members, and return to the 3610 * caller. 3611 */ 3612int 3613zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp) 3614{ 3615 zfs_cmd_t zc = { 0 }; 3616 uint64_t count; 3617 zbookmark_phys_t *zb = NULL; 3618 int i; 3619 3620 /* 3621 * Retrieve the raw error list from the kernel. If the number of errors 3622 * has increased, allocate more space and continue until we get the 3623 * entire list. 3624 */ 3625 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_ERRCOUNT, 3626 &count) == 0); 3627 if (count == 0) 3628 return (0); 3629 if ((zc.zc_nvlist_dst = (uintptr_t)zfs_alloc(zhp->zpool_hdl, 3630 count * sizeof (zbookmark_phys_t))) == (uintptr_t)NULL) 3631 return (-1); 3632 zc.zc_nvlist_dst_size = count; 3633 (void) strcpy(zc.zc_name, zhp->zpool_name); 3634 for (;;) { 3635 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_ERROR_LOG, 3636 &zc) != 0) { 3637 free((void *)(uintptr_t)zc.zc_nvlist_dst); 3638 if (errno == ENOMEM) { 3639 void *dst; 3640 3641 count = zc.zc_nvlist_dst_size; 3642 dst = zfs_alloc(zhp->zpool_hdl, count * 3643 sizeof (zbookmark_phys_t)); 3644 if (dst == NULL) 3645 return (-1); 3646 zc.zc_nvlist_dst = (uintptr_t)dst; 3647 } else { 3648 return (-1); 3649 } 3650 } else { 3651 break; 3652 } 3653 } 3654 3655 /* 3656 * Sort the resulting bookmarks. This is a little confusing due to the 3657 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last 3658 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks 3659 * _not_ copied as part of the process. So we point the start of our 3660 * array appropriate and decrement the total number of elements. 3661 */ 3662 zb = ((zbookmark_phys_t *)(uintptr_t)zc.zc_nvlist_dst) + 3663 zc.zc_nvlist_dst_size; 3664 count -= zc.zc_nvlist_dst_size; 3665 3666 qsort(zb, count, sizeof (zbookmark_phys_t), zbookmark_mem_compare); 3667 3668 verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0); 3669 3670 /* 3671 * Fill in the nverrlistp with nvlist's of dataset and object numbers. 3672 */ 3673 for (i = 0; i < count; i++) { 3674 nvlist_t *nv; 3675 3676 /* ignoring zb_blkid and zb_level for now */ 3677 if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset && 3678 zb[i-1].zb_object == zb[i].zb_object) 3679 continue; 3680 3681 if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0) 3682 goto nomem; 3683 if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET, 3684 zb[i].zb_objset) != 0) { 3685 nvlist_free(nv); 3686 goto nomem; 3687 } 3688 if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT, 3689 zb[i].zb_object) != 0) { 3690 nvlist_free(nv); 3691 goto nomem; 3692 } 3693 if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) { 3694 nvlist_free(nv); 3695 goto nomem; 3696 } 3697 nvlist_free(nv); 3698 } 3699 3700 free((void *)(uintptr_t)zc.zc_nvlist_dst); 3701 return (0); 3702 3703nomem: 3704 free((void *)(uintptr_t)zc.zc_nvlist_dst); 3705 return (no_memory(zhp->zpool_hdl)); 3706} 3707 3708/* 3709 * Upgrade a ZFS pool to the latest on-disk version. 3710 */ 3711int 3712zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version) 3713{ 3714 zfs_cmd_t zc = { 0 }; 3715 libzfs_handle_t *hdl = zhp->zpool_hdl; 3716 3717 (void) strcpy(zc.zc_name, zhp->zpool_name); 3718 zc.zc_cookie = new_version; 3719 3720 if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0) 3721 return (zpool_standard_error_fmt(hdl, errno, 3722 dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"), 3723 zhp->zpool_name)); 3724 return (0); 3725} 3726 3727void 3728zfs_save_arguments(int argc, char **argv, char *string, int len) 3729{ 3730 (void) strlcpy(string, basename(argv[0]), len); 3731 for (int i = 1; i < argc; i++) { 3732 (void) strlcat(string, " ", len); 3733 (void) strlcat(string, argv[i], len); 3734 } 3735} 3736 3737int 3738zpool_log_history(libzfs_handle_t *hdl, const char *message) 3739{ 3740 zfs_cmd_t zc = { 0 }; 3741 nvlist_t *args; 3742 int err; 3743 3744 args = fnvlist_alloc(); 3745 fnvlist_add_string(args, "message", message); 3746 err = zcmd_write_src_nvlist(hdl, &zc, args); 3747 if (err == 0) 3748 err = ioctl(hdl->libzfs_fd, ZFS_IOC_LOG_HISTORY, &zc); 3749 nvlist_free(args); 3750 zcmd_free_nvlists(&zc); 3751 return (err); 3752} 3753 3754/* 3755 * Perform ioctl to get some command history of a pool. 3756 * 3757 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the 3758 * logical offset of the history buffer to start reading from. 3759 * 3760 * Upon return, 'off' is the next logical offset to read from and 3761 * 'len' is the actual amount of bytes read into 'buf'. 3762 */ 3763static int 3764get_history(zpool_handle_t *zhp, char *buf, uint64_t *off, uint64_t *len) 3765{ 3766 zfs_cmd_t zc = { 0 }; 3767 libzfs_handle_t *hdl = zhp->zpool_hdl; 3768 3769 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3770 3771 zc.zc_history = (uint64_t)(uintptr_t)buf; 3772 zc.zc_history_len = *len; 3773 zc.zc_history_offset = *off; 3774 3775 if (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) { 3776 switch (errno) { 3777 case EPERM: 3778 return (zfs_error_fmt(hdl, EZFS_PERM, 3779 dgettext(TEXT_DOMAIN, 3780 "cannot show history for pool '%s'"), 3781 zhp->zpool_name)); 3782 case ENOENT: 3783 return (zfs_error_fmt(hdl, EZFS_NOHISTORY, 3784 dgettext(TEXT_DOMAIN, "cannot get history for pool " 3785 "'%s'"), zhp->zpool_name)); 3786 case ENOTSUP: 3787 return (zfs_error_fmt(hdl, EZFS_BADVERSION, 3788 dgettext(TEXT_DOMAIN, "cannot get history for pool " 3789 "'%s', pool must be upgraded"), zhp->zpool_name)); 3790 default: 3791 return (zpool_standard_error_fmt(hdl, errno, 3792 dgettext(TEXT_DOMAIN, 3793 "cannot get history for '%s'"), zhp->zpool_name)); 3794 } 3795 } 3796 3797 *len = zc.zc_history_len; 3798 *off = zc.zc_history_offset; 3799 3800 return (0); 3801} 3802 3803/* 3804 * Process the buffer of nvlists, unpacking and storing each nvlist record 3805 * into 'records'. 'leftover' is set to the number of bytes that weren't 3806 * processed as there wasn't a complete record. 3807 */ 3808int 3809zpool_history_unpack(char *buf, uint64_t bytes_read, uint64_t *leftover, 3810 nvlist_t ***records, uint_t *numrecords) 3811{ 3812 uint64_t reclen; 3813 nvlist_t *nv; 3814 int i; 3815 3816 while (bytes_read > sizeof (reclen)) { 3817 3818 /* get length of packed record (stored as little endian) */ 3819 for (i = 0, reclen = 0; i < sizeof (reclen); i++) 3820 reclen += (uint64_t)(((uchar_t *)buf)[i]) << (8*i); 3821 3822 if (bytes_read < sizeof (reclen) + reclen) 3823 break; 3824 3825 /* unpack record */ 3826 if (nvlist_unpack(buf + sizeof (reclen), reclen, &nv, 0) != 0) 3827 return (ENOMEM); 3828 bytes_read -= sizeof (reclen) + reclen; 3829 buf += sizeof (reclen) + reclen; 3830 3831 /* add record to nvlist array */ 3832 (*numrecords)++; 3833 if (ISP2(*numrecords + 1)) { 3834 *records = realloc(*records, 3835 *numrecords * 2 * sizeof (nvlist_t *)); 3836 } 3837 (*records)[*numrecords - 1] = nv; 3838 } 3839 3840 *leftover = bytes_read; 3841 return (0); 3842} 3843 3844/* from spa_history.c: spa_history_create_obj() */ 3845#define HIS_BUF_LEN_DEF (128 << 10) 3846#define HIS_BUF_LEN_MAX (1 << 30) 3847 3848/* 3849 * Retrieve the command history of a pool. 3850 */ 3851int 3852zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp) 3853{ 3854 char *buf; 3855 uint64_t buflen = HIS_BUF_LEN_DEF; 3856 uint64_t off = 0; 3857 nvlist_t **records = NULL; 3858 uint_t numrecords = 0; 3859 int err, i; 3860 3861 buf = malloc(buflen); 3862 if (buf == NULL) 3863 return (ENOMEM); 3864 do { 3865 uint64_t bytes_read = buflen; 3866 uint64_t leftover; 3867 3868 if ((err = get_history(zhp, buf, &off, &bytes_read)) != 0) 3869 break; 3870 3871 /* if nothing else was read in, we're at EOF, just return */ 3872 if (bytes_read == 0) 3873 break; 3874 3875 if ((err = zpool_history_unpack(buf, bytes_read, 3876 &leftover, &records, &numrecords)) != 0) 3877 break; 3878 off -= leftover; 3879 if (leftover == bytes_read) { 3880 /* 3881 * no progress made, because buffer is not big enough 3882 * to hold this record; resize and retry. 3883 */ 3884 buflen *= 2; 3885 free(buf); 3886 buf = NULL; 3887 if ((buflen >= HIS_BUF_LEN_MAX) || 3888 ((buf = malloc(buflen)) == NULL)) { 3889 err = ENOMEM; 3890 break; 3891 } 3892 } 3893 3894 /* CONSTCOND */ 3895 } while (1); 3896 3897 free(buf); 3898 3899 if (!err) { 3900 verify(nvlist_alloc(nvhisp, NV_UNIQUE_NAME, 0) == 0); 3901 verify(nvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD, 3902 records, numrecords) == 0); 3903 } 3904 for (i = 0; i < numrecords; i++) 3905 nvlist_free(records[i]); 3906 free(records); 3907 3908 return (err); 3909} 3910 3911void 3912zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj, 3913 char *pathname, size_t len) 3914{ 3915 zfs_cmd_t zc = { 0 }; 3916 boolean_t mounted = B_FALSE; 3917 char *mntpnt = NULL; 3918 char dsname[ZFS_MAX_DATASET_NAME_LEN]; 3919 3920 if (dsobj == 0) { 3921 /* special case for the MOS */ 3922 (void) snprintf(pathname, len, "<metadata>:<0x%llx>", obj); 3923 return; 3924 } 3925 3926 /* get the dataset's name */ 3927 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); 3928 zc.zc_obj = dsobj; 3929 if (ioctl(zhp->zpool_hdl->libzfs_fd, 3930 ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) { 3931 /* just write out a path of two object numbers */ 3932 (void) snprintf(pathname, len, "<0x%llx>:<0x%llx>", 3933 dsobj, obj); 3934 return; 3935 } 3936 (void) strlcpy(dsname, zc.zc_value, sizeof (dsname)); 3937 3938 /* find out if the dataset is mounted */ 3939 mounted = is_mounted(zhp->zpool_hdl, dsname, &mntpnt); 3940 3941 /* get the corrupted object's path */ 3942 (void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name)); 3943 zc.zc_obj = obj; 3944 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_OBJ_TO_PATH, 3945 &zc) == 0) { 3946 if (mounted) { 3947 (void) snprintf(pathname, len, "%s%s", mntpnt, 3948 zc.zc_value); 3949 } else { 3950 (void) snprintf(pathname, len, "%s:%s", 3951 dsname, zc.zc_value); 3952 } 3953 } else { 3954 (void) snprintf(pathname, len, "%s:<0x%llx>", dsname, obj); 3955 } 3956 free(mntpnt); 3957} 3958 3959#ifdef illumos 3960/* 3961 * Read the EFI label from the config, if a label does not exist then 3962 * pass back the error to the caller. If the caller has passed a non-NULL 3963 * diskaddr argument then we set it to the starting address of the EFI 3964 * partition. If the caller has passed a non-NULL boolean argument, then 3965 * we set it to indicate if the disk does have efi system partition. 3966 */ 3967static int 3968read_efi_label(nvlist_t *config, diskaddr_t *sb, boolean_t *system) 3969{ 3970 char *path; 3971 int fd; 3972 char diskname[MAXPATHLEN]; 3973 boolean_t boot = B_FALSE; 3974 int err = -1; 3975 int slice; 3976 3977 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PATH, &path) != 0) 3978 return (err); 3979 3980 (void) snprintf(diskname, sizeof (diskname), "%s%s", ZFS_RDISK_ROOT, 3981 strrchr(path, '/')); 3982 if ((fd = open(diskname, O_RDONLY|O_NDELAY)) >= 0) { 3983 struct dk_gpt *vtoc; 3984 3985 if ((err = efi_alloc_and_read(fd, &vtoc)) >= 0) { 3986 for (slice = 0; slice < vtoc->efi_nparts; slice++) { 3987 if (vtoc->efi_parts[slice].p_tag == V_SYSTEM) 3988 boot = B_TRUE; 3989 if (vtoc->efi_parts[slice].p_tag == V_USR) 3990 break; 3991 } 3992 if (sb != NULL && vtoc->efi_parts[slice].p_tag == V_USR) 3993 *sb = vtoc->efi_parts[slice].p_start; 3994 if (system != NULL) 3995 *system = boot; 3996 efi_free(vtoc); 3997 } 3998 (void) close(fd); 3999 } 4000 return (err); 4001} 4002 4003/* 4004 * determine where a partition starts on a disk in the current 4005 * configuration 4006 */ 4007static diskaddr_t 4008find_start_block(nvlist_t *config) 4009{ 4010 nvlist_t **child; 4011 uint_t c, children; 4012 diskaddr_t sb = MAXOFFSET_T; 4013 uint64_t wholedisk; 4014 4015 if (nvlist_lookup_nvlist_array(config, 4016 ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) { 4017 if (nvlist_lookup_uint64(config, 4018 ZPOOL_CONFIG_WHOLE_DISK, 4019 &wholedisk) != 0 || !wholedisk) { 4020 return (MAXOFFSET_T); 4021 } 4022 if (read_efi_label(config, &sb, NULL) < 0) 4023 sb = MAXOFFSET_T; 4024 return (sb); 4025 } 4026 4027 for (c = 0; c < children; c++) { 4028 sb = find_start_block(child[c]); 4029 if (sb != MAXOFFSET_T) { 4030 return (sb); 4031 } 4032 } 4033 return (MAXOFFSET_T); 4034} 4035#endif /* illumos */ 4036 4037/* 4038 * Label an individual disk. The name provided is the short name, 4039 * stripped of any leading /dev path. 4040 */ 4041int 4042zpool_label_disk(libzfs_handle_t *hdl, zpool_handle_t *zhp, const char *name, 4043 zpool_boot_label_t boot_type, uint64_t boot_size, int *slice) 4044{ 4045#ifdef illumos 4046 char path[MAXPATHLEN]; 4047 struct dk_gpt *vtoc; 4048 int fd; 4049 size_t resv = EFI_MIN_RESV_SIZE; 4050 uint64_t slice_size; 4051 diskaddr_t start_block; 4052 char errbuf[1024]; 4053 4054 /* prepare an error message just in case */ 4055 (void) snprintf(errbuf, sizeof (errbuf), 4056 dgettext(TEXT_DOMAIN, "cannot label '%s'"), name); 4057 4058 if (zhp) { 4059 nvlist_t *nvroot; 4060 4061 verify(nvlist_lookup_nvlist(zhp->zpool_config, 4062 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 4063 4064 if (zhp->zpool_start_block == 0) 4065 start_block = find_start_block(nvroot); 4066 else 4067 start_block = zhp->zpool_start_block; 4068 zhp->zpool_start_block = start_block; 4069 } else { 4070 /* new pool */ 4071 start_block = NEW_START_BLOCK; 4072 } 4073 4074 (void) snprintf(path, sizeof (path), "%s/%s%s", ZFS_RDISK_ROOT, name, 4075 BACKUP_SLICE); 4076 4077 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) { 4078 /* 4079 * This shouldn't happen. We've long since verified that this 4080 * is a valid device. 4081 */ 4082 zfs_error_aux(hdl, 4083 dgettext(TEXT_DOMAIN, "unable to open device")); 4084 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf)); 4085 } 4086 4087 if (efi_alloc_and_init(fd, EFI_NUMPAR, &vtoc) != 0) { 4088 /* 4089 * The only way this can fail is if we run out of memory, or we 4090 * were unable to read the disk's capacity 4091 */ 4092 if (errno == ENOMEM) 4093 (void) no_memory(hdl); 4094 4095 (void) close(fd); 4096 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4097 "unable to read disk capacity"), name); 4098 4099 return (zfs_error(hdl, EZFS_NOCAP, errbuf)); 4100 } 4101 4102 /* 4103 * Why we use V_USR: V_BACKUP confuses users, and is considered 4104 * disposable by some EFI utilities (since EFI doesn't have a backup 4105 * slice). V_UNASSIGNED is supposed to be used only for zero size 4106 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT, 4107 * etc. were all pretty specific. V_USR is as close to reality as we 4108 * can get, in the absence of V_OTHER. 4109 */ 4110 /* first fix the partition start block */ 4111 if (start_block == MAXOFFSET_T) 4112 start_block = NEW_START_BLOCK; 4113 4114 /* 4115 * EFI System partition is using slice 0. 4116 * ZFS is on slice 1 and slice 8 is reserved. 4117 * We assume the GPT partition table without system 4118 * partition has zfs p_start == NEW_START_BLOCK. 4119 * If start_block != NEW_START_BLOCK, it means we have 4120 * system partition. Correct solution would be to query/cache vtoc 4121 * from existing vdev member. 4122 */ 4123 if (boot_type == ZPOOL_CREATE_BOOT_LABEL) { 4124 if (boot_size % vtoc->efi_lbasize != 0) { 4125 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4126 "boot partition size must be a multiple of %d"), 4127 vtoc->efi_lbasize); 4128 (void) close(fd); 4129 efi_free(vtoc); 4130 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf)); 4131 } 4132 /* 4133 * System partition size checks. 4134 * Note the 1MB is quite arbitrary value, since we 4135 * are creating dedicated pool, it should be enough 4136 * to hold fat + efi bootloader. May need to be 4137 * adjusted if the bootloader size will grow. 4138 */ 4139 if (boot_size < 1024 * 1024) { 4140 char buf[64]; 4141 zfs_nicenum(boot_size, buf, sizeof (buf)); 4142 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4143 "Specified size %s for EFI System partition is too " 4144 "small, the minimum size is 1MB."), buf); 4145 (void) close(fd); 4146 efi_free(vtoc); 4147 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf)); 4148 } 4149 /* 33MB is tested with mkfs -F pcfs */ 4150 if (hdl->libzfs_printerr && 4151 ((vtoc->efi_lbasize == 512 && 4152 boot_size < 33 * 1024 * 1024) || 4153 (vtoc->efi_lbasize == 4096 && 4154 boot_size < 256 * 1024 * 1024))) { 4155 char buf[64]; 4156 zfs_nicenum(boot_size, buf, sizeof (buf)); 4157 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, 4158 "Warning: EFI System partition size %s is " 4159 "not allowing to create FAT32 file\nsystem, which " 4160 "may result in unbootable system.\n"), buf); 4161 } 4162 /* Adjust zfs partition start by size of system partition. */ 4163 start_block += boot_size / vtoc->efi_lbasize; 4164 } 4165 4166 if (start_block == NEW_START_BLOCK) { 4167 /* 4168 * Use default layout. 4169 * ZFS is on slice 0 and slice 8 is reserved. 4170 */ 4171 slice_size = vtoc->efi_last_u_lba + 1; 4172 slice_size -= EFI_MIN_RESV_SIZE; 4173 slice_size -= start_block; 4174 if (slice != NULL) 4175 *slice = 0; 4176 4177 vtoc->efi_parts[0].p_start = start_block; 4178 vtoc->efi_parts[0].p_size = slice_size; 4179 4180 vtoc->efi_parts[0].p_tag = V_USR; 4181 (void) strcpy(vtoc->efi_parts[0].p_name, "zfs"); 4182 4183 vtoc->efi_parts[8].p_start = slice_size + start_block; 4184 vtoc->efi_parts[8].p_size = resv; 4185 vtoc->efi_parts[8].p_tag = V_RESERVED; 4186 } else { 4187 slice_size = start_block - NEW_START_BLOCK; 4188 vtoc->efi_parts[0].p_start = NEW_START_BLOCK; 4189 vtoc->efi_parts[0].p_size = slice_size; 4190 vtoc->efi_parts[0].p_tag = V_SYSTEM; 4191 (void) strcpy(vtoc->efi_parts[0].p_name, "loader"); 4192 if (slice != NULL) 4193 *slice = 1; 4194 /* prepare slice 1 */ 4195 slice_size = vtoc->efi_last_u_lba + 1 - slice_size; 4196 slice_size -= resv; 4197 slice_size -= NEW_START_BLOCK; 4198 vtoc->efi_parts[1].p_start = start_block; 4199 vtoc->efi_parts[1].p_size = slice_size; 4200 vtoc->efi_parts[1].p_tag = V_USR; 4201 (void) strcpy(vtoc->efi_parts[1].p_name, "zfs"); 4202 4203 vtoc->efi_parts[8].p_start = slice_size + start_block; 4204 vtoc->efi_parts[8].p_size = resv; 4205 vtoc->efi_parts[8].p_tag = V_RESERVED; 4206 } 4207 4208 if (efi_write(fd, vtoc) != 0) { 4209 /* 4210 * Some block drivers (like pcata) may not support EFI 4211 * GPT labels. Print out a helpful error message dir- 4212 * ecting the user to manually label the disk and give 4213 * a specific slice. 4214 */ 4215 (void) close(fd); 4216 efi_free(vtoc); 4217 4218 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4219 "try using fdisk(1M) and then provide a specific slice")); 4220 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf)); 4221 } 4222 4223 (void) close(fd); 4224 efi_free(vtoc); 4225#endif /* illumos */ 4226 return (0); 4227} 4228 4229static boolean_t 4230supported_dump_vdev_type(libzfs_handle_t *hdl, nvlist_t *config, char *errbuf) 4231{ 4232 char *type; 4233 nvlist_t **child; 4234 uint_t children, c; 4235 4236 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_TYPE, &type) == 0); 4237 if (strcmp(type, VDEV_TYPE_FILE) == 0 || 4238 strcmp(type, VDEV_TYPE_HOLE) == 0 || 4239 strcmp(type, VDEV_TYPE_MISSING) == 0) { 4240 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4241 "vdev type '%s' is not supported"), type); 4242 (void) zfs_error(hdl, EZFS_VDEVNOTSUP, errbuf); 4243 return (B_FALSE); 4244 } 4245 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN, 4246 &child, &children) == 0) { 4247 for (c = 0; c < children; c++) { 4248 if (!supported_dump_vdev_type(hdl, child[c], errbuf)) 4249 return (B_FALSE); 4250 } 4251 } 4252 return (B_TRUE); 4253} 4254 4255/* 4256 * Check if this zvol is allowable for use as a dump device; zero if 4257 * it is, > 0 if it isn't, < 0 if it isn't a zvol. 4258 * 4259 * Allowable storage configurations include mirrors, all raidz variants, and 4260 * pools with log, cache, and spare devices. Pools which are backed by files or 4261 * have missing/hole vdevs are not suitable. 4262 */ 4263int 4264zvol_check_dump_config(char *arg) 4265{ 4266 zpool_handle_t *zhp = NULL; 4267 nvlist_t *config, *nvroot; 4268 char *p, *volname; 4269 nvlist_t **top; 4270 uint_t toplevels; 4271 libzfs_handle_t *hdl; 4272 char errbuf[1024]; 4273 char poolname[ZFS_MAX_DATASET_NAME_LEN]; 4274 int pathlen = strlen(ZVOL_FULL_DEV_DIR); 4275 int ret = 1; 4276 4277 if (strncmp(arg, ZVOL_FULL_DEV_DIR, pathlen)) { 4278 return (-1); 4279 } 4280 4281 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, 4282 "dump is not supported on device '%s'"), arg); 4283 4284 if ((hdl = libzfs_init()) == NULL) 4285 return (1); 4286 libzfs_print_on_error(hdl, B_TRUE); 4287 4288 volname = arg + pathlen; 4289 4290 /* check the configuration of the pool */ 4291 if ((p = strchr(volname, '/')) == NULL) { 4292 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4293 "malformed dataset name")); 4294 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf); 4295 return (1); 4296 } else if (p - volname >= ZFS_MAX_DATASET_NAME_LEN) { 4297 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4298 "dataset name is too long")); 4299 (void) zfs_error(hdl, EZFS_NAMETOOLONG, errbuf); 4300 return (1); 4301 } else { 4302 (void) strncpy(poolname, volname, p - volname); 4303 poolname[p - volname] = '\0'; 4304 } 4305 4306 if ((zhp = zpool_open(hdl, poolname)) == NULL) { 4307 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4308 "could not open pool '%s'"), poolname); 4309 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf); 4310 goto out; 4311 } 4312 config = zpool_get_config(zhp, NULL); 4313 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 4314 &nvroot) != 0) { 4315 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, 4316 "could not obtain vdev configuration for '%s'"), poolname); 4317 (void) zfs_error(hdl, EZFS_INVALCONFIG, errbuf); 4318 goto out; 4319 } 4320 4321 verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, 4322 &top, &toplevels) == 0); 4323 4324 if (!supported_dump_vdev_type(hdl, top[0], errbuf)) { 4325 goto out; 4326 } 4327 ret = 0; 4328 4329out: 4330 if (zhp) 4331 zpool_close(zhp); 4332 libzfs_fini(hdl); 4333 return (ret); 4334} 4335 4336int 4337zpool_nextboot(libzfs_handle_t *hdl, uint64_t pool_guid, uint64_t dev_guid, 4338 const char *command) 4339{ 4340 zfs_cmd_t zc = { 0 }; 4341 nvlist_t *args; 4342 char *packed; 4343 size_t size; 4344 int error; 4345 4346 args = fnvlist_alloc(); 4347 fnvlist_add_uint64(args, ZPOOL_CONFIG_POOL_GUID, pool_guid); 4348 fnvlist_add_uint64(args, ZPOOL_CONFIG_GUID, dev_guid); 4349 fnvlist_add_string(args, "command", command); 4350 error = zcmd_write_src_nvlist(hdl, &zc, args); 4351 if (error == 0) 4352 error = ioctl(hdl->libzfs_fd, ZFS_IOC_NEXTBOOT, &zc); 4353 zcmd_free_nvlists(&zc); 4354 nvlist_free(args); 4355 return (error); 4356} 4357