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