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