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