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