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