lib/libzfs/libzfs_pool.c

   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, 2018 by Delphix. All rights reserved.
  26  * Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>
  27  * Copyright (c) 2018 Datto Inc.
  28  * Copyright (c) 2017 Open-E, Inc. All Rights Reserved.
  29  * Copyright (c) 2017, Intel Corporation.
  30  * Copyright (c) 2018, loli10K <ezomori.nozomu@gmail.com>
  31  */
  32
  33 #include <errno.h>
  34 #include <libintl.h>
  35 #include <stdio.h>
  36 #include <stdlib.h>
  37 #include <strings.h>
  38 #include <unistd.h>
  39 #include <libgen.h>
  40 #include <zone.h>
  41 #include <sys/stat.h>
  42 #include <sys/efi_partition.h>
  43 #include <sys/systeminfo.h>
  44 #include <sys/zfs_ioctl.h>
  45 #include <sys/vdev_disk.h>
  46 #include <dlfcn.h>
  47 #include <libzutil.h>
  48
  49 #include "zfs_namecheck.h"
  50 #include "zfs_prop.h"
  51 #include "libzfs_impl.h"
  52 #include "zfs_comutil.h"
  53 #include "zfeature_common.h"
  54
  55 static boolean_t zpool_vdev_is_interior(const char *name);
  56
  57 typedef struct prop_flags {
  58         int create:1;   /* Validate property on creation */
  59         int import:1;   /* Validate property on import */
  60 } prop_flags_t;
  61
  62 /*
  63  * ====================================================================
  64  *   zpool property functions
  65  * ====================================================================
  66  */
  67
  68 static int
  69 zpool_get_all_props(zpool_handle_t *zhp)
  70 {
  71         zfs_cmd_t zc = {"\0"};
  72         libzfs_handle_t *hdl = zhp->zpool_hdl;
  73
  74         (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
  75
  76         if (zcmd_alloc_dst_nvlist(hdl, &zc, 0) != 0)
  77                 return (-1);
  78
  79         while (zfs_ioctl(hdl, ZFS_IOC_POOL_GET_PROPS, &zc) != 0) {
  80                 if (errno == ENOMEM) {
  81                         if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
  82                                 zcmd_free_nvlists(&zc);
  83                                 return (-1);
  84                         }
  85                 } else {
  86                         zcmd_free_nvlists(&zc);
  87                         return (-1);
  88                 }
  89         }
  90
  91         if (zcmd_read_dst_nvlist(hdl, &zc, &zhp->zpool_props) != 0) {
  92                 zcmd_free_nvlists(&zc);
  93                 return (-1);
  94         }
  95
  96         zcmd_free_nvlists(&zc);
  97
  98         return (0);
  99 }
 100
 101 int
 102 zpool_props_refresh(zpool_handle_t *zhp)
 103 {
 104         nvlist_t *old_props;
 105
 106         old_props = zhp->zpool_props;
 107
 108         if (zpool_get_all_props(zhp) != 0)
 109                 return (-1);
 110
 111         nvlist_free(old_props);
 112         return (0);
 113 }
 114
 115 static const char *
 116 zpool_get_prop_string(zpool_handle_t *zhp, zpool_prop_t prop,
 117     zprop_source_t *src)
 118 {
 119         nvlist_t *nv, *nvl;
 120         uint64_t ival;
 121         char *value;
 122         zprop_source_t source;
 123
 124         nvl = zhp->zpool_props;
 125         if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
 126                 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &ival) == 0);
 127                 source = ival;
 128                 verify(nvlist_lookup_string(nv, ZPROP_VALUE, &value) == 0);
 129         } else {
 130                 source = ZPROP_SRC_DEFAULT;
 131                 if ((value = (char *)zpool_prop_default_string(prop)) == NULL)
 132                         value = "-";
 133         }
 134
 135         if (src)
 136                 *src = source;
 137
 138         return (value);
 139 }
 140
 141 uint64_t
 142 zpool_get_prop_int(zpool_handle_t *zhp, zpool_prop_t prop, zprop_source_t *src)
 143 {
 144         nvlist_t *nv, *nvl;
 145         uint64_t value;
 146         zprop_source_t source;
 147
 148         if (zhp->zpool_props == NULL && zpool_get_all_props(zhp)) {
 149                 /*
 150                  * zpool_get_all_props() has most likely failed because
 151                  * the pool is faulted, but if all we need is the top level
 152                  * vdev's guid then get it from the zhp config nvlist.
 153                  */
 154                 if ((prop == ZPOOL_PROP_GUID) &&
 155                     (nvlist_lookup_nvlist(zhp->zpool_config,
 156                     ZPOOL_CONFIG_VDEV_TREE, &nv) == 0) &&
 157                     (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value)
 158                     == 0)) {
 159                         return (value);
 160                 }
 161                 return (zpool_prop_default_numeric(prop));
 162         }
 163
 164         nvl = zhp->zpool_props;
 165         if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
 166                 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &value) == 0);
 167                 source = value;
 168                 verify(nvlist_lookup_uint64(nv, ZPROP_VALUE, &value) == 0);
 169         } else {
 170                 source = ZPROP_SRC_DEFAULT;
 171                 value = zpool_prop_default_numeric(prop);
 172         }
 173
 174         if (src)
 175                 *src = source;
 176
 177         return (value);
 178 }
 179
 180 /*
 181  * Map VDEV STATE to printed strings.
 182  */
 183 const char *
 184 zpool_state_to_name(vdev_state_t state, vdev_aux_t aux)
 185 {
 186         switch (state) {
 187         case VDEV_STATE_CLOSED:
 188         case VDEV_STATE_OFFLINE:
 189                 return (gettext("OFFLINE"));
 190         case VDEV_STATE_REMOVED:
 191                 return (gettext("REMOVED"));
 192         case VDEV_STATE_CANT_OPEN:
 193                 if (aux == VDEV_AUX_CORRUPT_DATA || aux == VDEV_AUX_BAD_LOG)
 194                         return (gettext("FAULTED"));
 195                 else if (aux == VDEV_AUX_SPLIT_POOL)
 196                         return (gettext("SPLIT"));
 197                 else
 198                         return (gettext("UNAVAIL"));
 199         case VDEV_STATE_FAULTED:
 200                 return (gettext("FAULTED"));
 201         case VDEV_STATE_DEGRADED:
 202                 return (gettext("DEGRADED"));
 203         case VDEV_STATE_HEALTHY:
 204                 return (gettext("ONLINE"));
 205
 206         default:
 207                 break;
 208         }
 209
 210         return (gettext("UNKNOWN"));
 211 }
 212
 213 /*
 214  * Map POOL STATE to printed strings.
 215  */
 216 const char *
 217 zpool_pool_state_to_name(pool_state_t state)
 218 {
 219         switch (state) {
 220         default:
 221                 break;
 222         case POOL_STATE_ACTIVE:
 223                 return (gettext("ACTIVE"));
 224         case POOL_STATE_EXPORTED:
 225                 return (gettext("EXPORTED"));
 226         case POOL_STATE_DESTROYED:
 227                 return (gettext("DESTROYED"));
 228         case POOL_STATE_SPARE:
 229                 return (gettext("SPARE"));
 230         case POOL_STATE_L2CACHE:
 231                 return (gettext("L2CACHE"));
 232         case POOL_STATE_UNINITIALIZED:
 233                 return (gettext("UNINITIALIZED"));
 234         case POOL_STATE_UNAVAIL:
 235                 return (gettext("UNAVAIL"));
 236         case POOL_STATE_POTENTIALLY_ACTIVE:
 237                 return (gettext("POTENTIALLY_ACTIVE"));
 238         }
 239
 240         return (gettext("UNKNOWN"));
 241 }
 242
 243 /*
 244  * Given a pool handle, return the pool health string ("ONLINE", "DEGRADED",
 245  * "SUSPENDED", etc).
 246  */
 247 const char *
 248 zpool_get_state_str(zpool_handle_t *zhp)
 249 {
 250         zpool_errata_t errata;
 251         zpool_status_t status;
 252         nvlist_t *nvroot;
 253         vdev_stat_t *vs;
 254         uint_t vsc;
 255         const char *str;
 256
 257         status = zpool_get_status(zhp, NULL, &errata);
 258
 259         if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
 260                 str = gettext("FAULTED");
 261         } else if (status == ZPOOL_STATUS_IO_FAILURE_WAIT ||
 262             status == ZPOOL_STATUS_IO_FAILURE_MMP) {
 263                 str = gettext("SUSPENDED");
 264         } else {
 265                 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
 266                     ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
 267                 verify(nvlist_lookup_uint64_array(nvroot,
 268                     ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc)
 269                     == 0);
 270                 str = zpool_state_to_name(vs->vs_state, vs->vs_aux);
 271         }
 272         return (str);
 273 }
 274
 275 /*
 276  * Get a zpool property value for 'prop' and return the value in
 277  * a pre-allocated buffer.
 278  */
 279 int
 280 zpool_get_prop(zpool_handle_t *zhp, zpool_prop_t prop, char *buf,
 281     size_t len, zprop_source_t *srctype, boolean_t literal)
 282 {
 283         uint64_t intval;
 284         const char *strval;
 285         zprop_source_t src = ZPROP_SRC_NONE;
 286
 287         if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
 288                 switch (prop) {
 289                 case ZPOOL_PROP_NAME:
 290                         (void) strlcpy(buf, zpool_get_name(zhp), len);
 291                         break;
 292
 293                 case ZPOOL_PROP_HEALTH:
 294                         (void) strlcpy(buf, zpool_get_state_str(zhp), len);
 295                         break;
 296
 297                 case ZPOOL_PROP_GUID:
 298                         intval = zpool_get_prop_int(zhp, prop, &src);
 299                         (void) snprintf(buf, len, "%llu", (u_longlong_t)intval);
 300                         break;
 301
 302                 case ZPOOL_PROP_ALTROOT:
 303                 case ZPOOL_PROP_CACHEFILE:
 304                 case ZPOOL_PROP_COMMENT:
 305                         if (zhp->zpool_props != NULL ||
 306                             zpool_get_all_props(zhp) == 0) {
 307                                 (void) strlcpy(buf,
 308                                     zpool_get_prop_string(zhp, prop, &src),
 309                                     len);
 310                                 break;
 311                         }
 312                         /* FALLTHROUGH */
 313                 default:
 314                         (void) strlcpy(buf, "-", len);
 315                         break;
 316                 }
 317
 318                 if (srctype != NULL)
 319                         *srctype = src;
 320                 return (0);
 321         }
 322
 323         if (zhp->zpool_props == NULL && zpool_get_all_props(zhp) &&
 324             prop != ZPOOL_PROP_NAME)
 325                 return (-1);
 326
 327         switch (zpool_prop_get_type(prop)) {
 328         case PROP_TYPE_STRING:
 329                 (void) strlcpy(buf, zpool_get_prop_string(zhp, prop, &src),
 330                     len);
 331                 break;
 332
 333         case PROP_TYPE_NUMBER:
 334                 intval = zpool_get_prop_int(zhp, prop, &src);
 335
 336                 switch (prop) {
 337                 case ZPOOL_PROP_SIZE:
 338                 case ZPOOL_PROP_ALLOCATED:
 339                 case ZPOOL_PROP_FREE:
 340                 case ZPOOL_PROP_FREEING:
 341                 case ZPOOL_PROP_LEAKED:
 342                 case ZPOOL_PROP_ASHIFT:
 343                         if (literal)
 344                                 (void) snprintf(buf, len, "%llu",
 345                                     (u_longlong_t)intval);
 346                         else
 347                                 (void) zfs_nicenum(intval, buf, len);
 348                         break;
 349
 350                 case ZPOOL_PROP_EXPANDSZ:
 351                 case ZPOOL_PROP_CHECKPOINT:
 352                         if (intval == 0) {
 353                                 (void) strlcpy(buf, "-", len);
 354                         } else if (literal) {
 355                                 (void) snprintf(buf, len, "%llu",
 356                                     (u_longlong_t)intval);
 357                         } else {
 358                                 (void) zfs_nicebytes(intval, buf, len);
 359                         }
 360                         break;
 361
 362                 case ZPOOL_PROP_CAPACITY:
 363                         if (literal) {
 364                                 (void) snprintf(buf, len, "%llu",
 365                                     (u_longlong_t)intval);
 366                         } else {
 367                                 (void) snprintf(buf, len, "%llu%%",
 368                                     (u_longlong_t)intval);
 369                         }
 370                         break;
 371
 372                 case ZPOOL_PROP_FRAGMENTATION:
 373                         if (intval == UINT64_MAX) {
 374                                 (void) strlcpy(buf, "-", len);
 375                         } else if (literal) {
 376                                 (void) snprintf(buf, len, "%llu",
 377                                     (u_longlong_t)intval);
 378                         } else {
 379                                 (void) snprintf(buf, len, "%llu%%",
 380                                     (u_longlong_t)intval);
 381                         }
 382                         break;
 383
 384                 case ZPOOL_PROP_DEDUPRATIO:
 385                         if (literal)
 386                                 (void) snprintf(buf, len, "%llu.%02llu",
 387                                     (u_longlong_t)(intval / 100),
 388                                     (u_longlong_t)(intval % 100));
 389                         else
 390                                 (void) snprintf(buf, len, "%llu.%02llux",
 391                                     (u_longlong_t)(intval / 100),
 392                                     (u_longlong_t)(intval % 100));
 393                         break;
 394
 395                 case ZPOOL_PROP_HEALTH:
 396                         (void) strlcpy(buf, zpool_get_state_str(zhp), len);
 397                         break;
 398                 case ZPOOL_PROP_VERSION:
 399                         if (intval >= SPA_VERSION_FEATURES) {
 400                                 (void) snprintf(buf, len, "-");
 401                                 break;
 402                         }
 403                         /* FALLTHROUGH */
 404                 default:
 405                         (void) snprintf(buf, len, "%llu", (u_longlong_t)intval);
 406                 }
 407                 break;
 408
 409         case PROP_TYPE_INDEX:
 410                 intval = zpool_get_prop_int(zhp, prop, &src);
 411                 if (zpool_prop_index_to_string(prop, intval, &strval)
 412                     != 0)
 413                         return (-1);
 414                 (void) strlcpy(buf, strval, len);
 415                 break;
 416
 417         default:
 418                 abort();
 419         }
 420
 421         if (srctype)
 422                 *srctype = src;
 423
 424         return (0);
 425 }
 426
 427 /*
 428  * Check if the bootfs name has the same pool name as it is set to.
 429  * Assuming bootfs is a valid dataset name.
 430  */
 431 static boolean_t
 432 bootfs_name_valid(const char *pool, char *bootfs)
 433 {
 434         int len = strlen(pool);
 435         if (bootfs[0] == '\0')
 436                 return (B_TRUE);
 437
 438         if (!zfs_name_valid(bootfs, ZFS_TYPE_FILESYSTEM|ZFS_TYPE_SNAPSHOT))
 439                 return (B_FALSE);
 440
 441         if (strncmp(pool, bootfs, len) == 0 &&
 442             (bootfs[len] == '/' || bootfs[len] == '\0'))
 443                 return (B_TRUE);
 444
 445         return (B_FALSE);
 446 }
 447
 448 boolean_t
 449 zpool_is_bootable(zpool_handle_t *zhp)
 450 {
 451         char bootfs[ZFS_MAX_DATASET_NAME_LEN];
 452
 453         return (zpool_get_prop(zhp, ZPOOL_PROP_BOOTFS, bootfs,
 454             sizeof (bootfs), NULL, B_FALSE) == 0 && strncmp(bootfs, "-",
 455             sizeof (bootfs)) != 0);
 456 }
 457
 458
 459 /*
 460  * Given an nvlist of zpool properties to be set, validate that they are
 461  * correct, and parse any numeric properties (index, boolean, etc) if they are
 462  * specified as strings.
 463  */
 464 static nvlist_t *
 465 zpool_valid_proplist(libzfs_handle_t *hdl, const char *poolname,
 466     nvlist_t *props, uint64_t version, prop_flags_t flags, char *errbuf)
 467 {
 468         nvpair_t *elem;
 469         nvlist_t *retprops;
 470         zpool_prop_t prop;
 471         char *strval;
 472         uint64_t intval;
 473         char *slash, *check;
 474         struct stat64 statbuf;
 475         zpool_handle_t *zhp;
 476
 477         if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) {
 478                 (void) no_memory(hdl);
 479                 return (NULL);
 480         }
 481
 482         elem = NULL;
 483         while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
 484                 const char *propname = nvpair_name(elem);
 485
 486                 prop = zpool_name_to_prop(propname);
 487                 if (prop == ZPOOL_PROP_INVAL && zpool_prop_feature(propname)) {
 488                         int err;
 489                         char *fname = strchr(propname, '@') + 1;
 490
 491                         err = zfeature_lookup_name(fname, NULL);
 492                         if (err != 0) {
 493                                 ASSERT3U(err, ==, ENOENT);
 494                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 495                                     "invalid feature '%s'"), fname);
 496                                 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
 497                                 goto error;
 498                         }
 499
 500                         if (nvpair_type(elem) != DATA_TYPE_STRING) {
 501                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 502                                     "'%s' must be a string"), propname);
 503                                 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
 504                                 goto error;
 505                         }
 506
 507                         (void) nvpair_value_string(elem, &strval);
 508                         if (strcmp(strval, ZFS_FEATURE_ENABLED) != 0 &&
 509                             strcmp(strval, ZFS_FEATURE_DISABLED) != 0) {
 510                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 511                                     "property '%s' can only be set to "
 512                                     "'enabled' or 'disabled'"), propname);
 513                                 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
 514                                 goto error;
 515                         }
 516
 517                         if (!flags.create &&
 518                             strcmp(strval, ZFS_FEATURE_DISABLED) == 0) {
 519                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 520                                     "property '%s' can only be set to "
 521                                     "'disabled' at creation time"), propname);
 522                                 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
 523                                 goto error;
 524                         }
 525
 526                         if (nvlist_add_uint64(retprops, propname, 0) != 0) {
 527                                 (void) no_memory(hdl);
 528                                 goto error;
 529                         }
 530                         continue;
 531                 }
 532
 533                 /*
 534                  * Make sure this property is valid and applies to this type.
 535                  */
 536                 if (prop == ZPOOL_PROP_INVAL) {
 537                         zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 538                             "invalid property '%s'"), propname);
 539                         (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
 540                         goto error;
 541                 }
 542
 543                 if (zpool_prop_readonly(prop)) {
 544                         zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
 545                             "is readonly"), propname);
 546                         (void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf);
 547                         goto error;
 548                 }
 549
 550                 if (!flags.create && zpool_prop_setonce(prop)) {
 551                         zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 552                             "property '%s' can only be set at "
 553                             "creation time"), propname);
 554                         (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
 555                         goto error;
 556                 }
 557
 558                 if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_POOL, retprops,
 559                     &strval, &intval, errbuf) != 0)
 560                         goto error;
 561
 562                 /*
 563                  * Perform additional checking for specific properties.
 564                  */
 565                 switch (prop) {
 566                 case ZPOOL_PROP_VERSION:
 567                         if (intval < version ||
 568                             !SPA_VERSION_IS_SUPPORTED(intval)) {
 569                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 570                                     "property '%s' number %d is invalid."),
 571                                     propname, intval);
 572                                 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
 573                                 goto error;
 574                         }
 575                         break;
 576
 577                 case ZPOOL_PROP_ASHIFT:
 578                         if (intval != 0 &&
 579                             (intval < ASHIFT_MIN || intval > ASHIFT_MAX)) {
 580                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 581                                     "property '%s' number %d is invalid, only "
 582                                     "values between %" PRId32 " and "
 583                                     "%" PRId32 " are allowed."),
 584                                     propname, intval, ASHIFT_MIN, ASHIFT_MAX);
 585                                 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
 586                                 goto error;
 587                         }
 588                         break;
 589
 590                 case ZPOOL_PROP_BOOTFS:
 591                         if (flags.create || flags.import) {
 592                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 593                                     "property '%s' cannot be set at creation "
 594                                     "or import time"), propname);
 595                                 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
 596                                 goto error;
 597                         }
 598
 599                         if (version < SPA_VERSION_BOOTFS) {
 600                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 601                                     "pool must be upgraded to support "
 602                                     "'%s' property"), propname);
 603                                 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
 604                                 goto error;
 605                         }
 606
 607                         /*
 608                          * bootfs property value has to be a dataset name and
 609                          * the dataset has to be in the same pool as it sets to.
 610                          */
 611                         if (!bootfs_name_valid(poolname, strval)) {
 612                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
 613                                     "is an invalid name"), strval);
 614                                 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
 615                                 goto error;
 616                         }
 617
 618                         if ((zhp = zpool_open_canfail(hdl, poolname)) == NULL) {
 619                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 620                                     "could not open pool '%s'"), poolname);
 621                                 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
 622                                 goto error;
 623                         }
 624                         zpool_close(zhp);
 625                         break;
 626
 627                 case ZPOOL_PROP_ALTROOT:
 628                         if (!flags.create && !flags.import) {
 629                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 630                                     "property '%s' can only be set during pool "
 631                                     "creation or import"), propname);
 632                                 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
 633                                 goto error;
 634                         }
 635
 636                         if (strval[0] != '/') {
 637                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 638                                     "bad alternate root '%s'"), strval);
 639                                 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
 640                                 goto error;
 641                         }
 642                         break;
 643
 644                 case ZPOOL_PROP_CACHEFILE:
 645                         if (strval[0] == '\0')
 646                                 break;
 647
 648                         if (strcmp(strval, "none") == 0)
 649                                 break;
 650
 651                         if (strval[0] != '/') {
 652                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 653                                     "property '%s' must be empty, an "
 654                                     "absolute path, or 'none'"), propname);
 655                                 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
 656                                 goto error;
 657                         }
 658
 659                         slash = strrchr(strval, '/');
 660
 661                         if (slash[1] == '\0' || strcmp(slash, "/.") == 0 ||
 662                             strcmp(slash, "/..") == 0) {
 663                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 664                                     "'%s' is not a valid file"), strval);
 665                                 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
 666                                 goto error;
 667                         }
 668
 669                         *slash = '\0';
 670
 671                         if (strval[0] != '\0' &&
 672                             (stat64(strval, &statbuf) != 0 ||
 673                             !S_ISDIR(statbuf.st_mode))) {
 674                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 675                                     "'%s' is not a valid directory"),
 676                                     strval);
 677                                 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
 678                                 goto error;
 679                         }
 680
 681                         *slash = '/';
 682                         break;
 683
 684                 case ZPOOL_PROP_COMMENT:
 685                         for (check = strval; *check != '\0'; check++) {
 686                                 if (!isprint(*check)) {
 687                                         zfs_error_aux(hdl,
 688                                             dgettext(TEXT_DOMAIN,
 689                                             "comment may only have printable "
 690                                             "characters"));
 691                                         (void) zfs_error(hdl, EZFS_BADPROP,
 692                                             errbuf);
 693                                         goto error;
 694                                 }
 695                         }
 696                         if (strlen(strval) > ZPROP_MAX_COMMENT) {
 697                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 698                                     "comment must not exceed %d characters"),
 699                                     ZPROP_MAX_COMMENT);
 700                                 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
 701                                 goto error;
 702                         }
 703                         break;
 704                 case ZPOOL_PROP_READONLY:
 705                         if (!flags.import) {
 706                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 707                                     "property '%s' can only be set at "
 708                                     "import time"), propname);
 709                                 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
 710                                 goto error;
 711                         }
 712                         break;
 713                 case ZPOOL_PROP_MULTIHOST:
 714                         if (get_system_hostid() == 0) {
 715                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 716                                     "requires a non-zero system hostid"));
 717                                 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
 718                                 goto error;
 719                         }
 720                         break;
 721                 case ZPOOL_PROP_DEDUPDITTO:
 722                         printf("Note: property '%s' no longer has "
 723                             "any effect\n", propname);
 724                         break;
 725
 726                 default:
 727                         zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 728                             "property '%s'(%d) not defined"), propname, prop);
 729                         break;
 730                 }
 731         }
 732
 733         return (retprops);
 734 error:
 735         nvlist_free(retprops);
 736         return (NULL);
 737 }
 738
 739 /*
 740  * Set zpool property : propname=propval.
 741  */
 742 int
 743 zpool_set_prop(zpool_handle_t *zhp, const char *propname, const char *propval)
 744 {
 745         zfs_cmd_t zc = {"\0"};
 746         int ret = -1;
 747         char errbuf[1024];
 748         nvlist_t *nvl = NULL;
 749         nvlist_t *realprops;
 750         uint64_t version;
 751         prop_flags_t flags = { 0 };
 752
 753         (void) snprintf(errbuf, sizeof (errbuf),
 754             dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
 755             zhp->zpool_name);
 756
 757         if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
 758                 return (no_memory(zhp->zpool_hdl));
 759
 760         if (nvlist_add_string(nvl, propname, propval) != 0) {
 761                 nvlist_free(nvl);
 762                 return (no_memory(zhp->zpool_hdl));
 763         }
 764
 765         version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
 766         if ((realprops = zpool_valid_proplist(zhp->zpool_hdl,
 767             zhp->zpool_name, nvl, version, flags, errbuf)) == NULL) {
 768                 nvlist_free(nvl);
 769                 return (-1);
 770         }
 771
 772         nvlist_free(nvl);
 773         nvl = realprops;
 774
 775         /*
 776          * Execute the corresponding ioctl() to set this property.
 777          */
 778         (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
 779
 780         if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) {
 781                 nvlist_free(nvl);
 782                 return (-1);
 783         }
 784
 785         ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc);
 786
 787         zcmd_free_nvlists(&zc);
 788         nvlist_free(nvl);
 789
 790         if (ret)
 791                 (void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf);
 792         else
 793                 (void) zpool_props_refresh(zhp);
 794
 795         return (ret);
 796 }
 797
 798 int
 799 zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp)
 800 {
 801         libzfs_handle_t *hdl = zhp->zpool_hdl;
 802         zprop_list_t *entry;
 803         char buf[ZFS_MAXPROPLEN];
 804         nvlist_t *features = NULL;
 805         nvpair_t *nvp;
 806         zprop_list_t **last;
 807         boolean_t firstexpand = (NULL == *plp);
 808         int i;
 809
 810         if (zprop_expand_list(hdl, plp, ZFS_TYPE_POOL) != 0)
 811                 return (-1);
 812
 813         last = plp;
 814         while (*last != NULL)
 815                 last = &(*last)->pl_next;
 816
 817         if ((*plp)->pl_all)
 818                 features = zpool_get_features(zhp);
 819
 820         if ((*plp)->pl_all && firstexpand) {
 821                 for (i = 0; i < SPA_FEATURES; i++) {
 822                         zprop_list_t *entry = zfs_alloc(hdl,
 823                             sizeof (zprop_list_t));
 824                         entry->pl_prop = ZPROP_INVAL;
 825                         entry->pl_user_prop = zfs_asprintf(hdl, "feature@%s",
 826                             spa_feature_table[i].fi_uname);
 827                         entry->pl_width = strlen(entry->pl_user_prop);
 828                         entry->pl_all = B_TRUE;
 829
 830                         *last = entry;
 831                         last = &entry->pl_next;
 832                 }
 833         }
 834
 835         /* add any unsupported features */
 836         for (nvp = nvlist_next_nvpair(features, NULL);
 837             nvp != NULL; nvp = nvlist_next_nvpair(features, nvp)) {
 838                 char *propname;
 839                 boolean_t found;
 840                 zprop_list_t *entry;
 841
 842                 if (zfeature_is_supported(nvpair_name(nvp)))
 843                         continue;
 844
 845                 propname = zfs_asprintf(hdl, "unsupported@%s",
 846                     nvpair_name(nvp));
 847
 848                 /*
 849                  * Before adding the property to the list make sure that no
 850                  * other pool already added the same property.
 851                  */
 852                 found = B_FALSE;
 853                 entry = *plp;
 854                 while (entry != NULL) {
 855                         if (entry->pl_user_prop != NULL &&
 856                             strcmp(propname, entry->pl_user_prop) == 0) {
 857                                 found = B_TRUE;
 858                                 break;
 859                         }
 860                         entry = entry->pl_next;
 861                 }
 862                 if (found) {
 863                         free(propname);
 864                         continue;
 865                 }
 866
 867                 entry = zfs_alloc(hdl, sizeof (zprop_list_t));
 868                 entry->pl_prop = ZPROP_INVAL;
 869                 entry->pl_user_prop = propname;
 870                 entry->pl_width = strlen(entry->pl_user_prop);
 871                 entry->pl_all = B_TRUE;
 872
 873                 *last = entry;
 874                 last = &entry->pl_next;
 875         }
 876
 877         for (entry = *plp; entry != NULL; entry = entry->pl_next) {
 878
 879                 if (entry->pl_fixed)
 880                         continue;
 881
 882                 if (entry->pl_prop != ZPROP_INVAL &&
 883                     zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf),
 884                     NULL, B_FALSE) == 0) {
 885                         if (strlen(buf) > entry->pl_width)
 886                                 entry->pl_width = strlen(buf);
 887                 }
 888         }
 889
 890         return (0);
 891 }
 892
 893 /*
 894  * Get the state for the given feature on the given ZFS pool.
 895  */
 896 int
 897 zpool_prop_get_feature(zpool_handle_t *zhp, const char *propname, char *buf,
 898     size_t len)
 899 {
 900         uint64_t refcount;
 901         boolean_t found = B_FALSE;
 902         nvlist_t *features = zpool_get_features(zhp);
 903         boolean_t supported;
 904         const char *feature = strchr(propname, '@') + 1;
 905
 906         supported = zpool_prop_feature(propname);
 907         ASSERT(supported || zpool_prop_unsupported(propname));
 908
 909         /*
 910          * Convert from feature name to feature guid. This conversion is
 911          * unnecessary for unsupported@... properties because they already
 912          * use guids.
 913          */
 914         if (supported) {
 915                 int ret;
 916                 spa_feature_t fid;
 917
 918                 ret = zfeature_lookup_name(feature, &fid);
 919                 if (ret != 0) {
 920                         (void) strlcpy(buf, "-", len);
 921                         return (ENOTSUP);
 922                 }
 923                 feature = spa_feature_table[fid].fi_guid;
 924         }
 925
 926         if (nvlist_lookup_uint64(features, feature, &refcount) == 0)
 927                 found = B_TRUE;
 928
 929         if (supported) {
 930                 if (!found) {
 931                         (void) strlcpy(buf, ZFS_FEATURE_DISABLED, len);
 932                 } else  {
 933                         if (refcount == 0)
 934                                 (void) strlcpy(buf, ZFS_FEATURE_ENABLED, len);
 935                         else
 936                                 (void) strlcpy(buf, ZFS_FEATURE_ACTIVE, len);
 937                 }
 938         } else {
 939                 if (found) {
 940                         if (refcount == 0) {
 941                                 (void) strcpy(buf, ZFS_UNSUPPORTED_INACTIVE);
 942                         } else {
 943                                 (void) strcpy(buf, ZFS_UNSUPPORTED_READONLY);
 944                         }
 945                 } else {
 946                         (void) strlcpy(buf, "-", len);
 947                         return (ENOTSUP);
 948                 }
 949         }
 950
 951         return (0);
 952 }
 953
 954 /*
 955  * Validate the given pool name, optionally putting an extended error message in
 956  * 'buf'.
 957  */
 958 boolean_t
 959 zpool_name_valid(libzfs_handle_t *hdl, boolean_t isopen, const char *pool)
 960 {
 961         namecheck_err_t why;
 962         char what;
 963         int ret;
 964
 965         ret = pool_namecheck(pool, &why, &what);
 966
 967         /*
 968          * The rules for reserved pool names were extended at a later point.
 969          * But we need to support users with existing pools that may now be
 970          * invalid.  So we only check for this expanded set of names during a
 971          * create (or import), and only in userland.
 972          */
 973         if (ret == 0 && !isopen &&
 974             (strncmp(pool, "mirror", 6) == 0 ||
 975             strncmp(pool, "raidz", 5) == 0 ||
 976             strncmp(pool, "spare", 5) == 0 ||
 977             strcmp(pool, "log") == 0)) {
 978                 if (hdl != NULL)
 979                         zfs_error_aux(hdl,
 980                             dgettext(TEXT_DOMAIN, "name is reserved"));
 981                 return (B_FALSE);
 982         }
 983
 984
 985         if (ret != 0) {
 986                 if (hdl != NULL) {
 987                         switch (why) {
 988                         case NAME_ERR_TOOLONG:
 989                                 zfs_error_aux(hdl,
 990                                     dgettext(TEXT_DOMAIN, "name is too long"));
 991                                 break;
 992
 993                         case NAME_ERR_INVALCHAR:
 994                                 zfs_error_aux(hdl,
 995                                     dgettext(TEXT_DOMAIN, "invalid character "
 996                                     "'%c' in pool name"), what);
 997                                 break;
 998
 999                         case NAME_ERR_NOLETTER:
1000                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1001                                     "name must begin with a letter"));
1002                                 break;
1003
1004                         case NAME_ERR_RESERVED:
1005                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1006                                     "name is reserved"));
1007                                 break;
1008
1009                         case NAME_ERR_DISKLIKE:
1010                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1011                                     "pool name is reserved"));
1012                                 break;
1013
1014                         case NAME_ERR_LEADING_SLASH:
1015                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1016                                     "leading slash in name"));
1017                                 break;
1018
1019                         case NAME_ERR_EMPTY_COMPONENT:
1020                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1021                                     "empty component in name"));
1022                                 break;
1023
1024                         case NAME_ERR_TRAILING_SLASH:
1025                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1026                                     "trailing slash in name"));
1027                                 break;
1028
1029                         case NAME_ERR_MULTIPLE_DELIMITERS:
1030                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1031                                     "multiple '@' and/or '#' delimiters in "
1032                                     "name"));
1033                                 break;
1034
1035                         case NAME_ERR_NO_AT:
1036                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1037                                     "permission set is missing '@'"));
1038                                 break;
1039
1040                         default:
1041                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1042                                     "(%d) not defined"), why);
1043                                 break;
1044                         }
1045                 }
1046                 return (B_FALSE);
1047         }
1048
1049         return (B_TRUE);
1050 }
1051
1052 /*
1053  * Open a handle to the given pool, even if the pool is currently in the FAULTED
1054  * state.
1055  */
1056 zpool_handle_t *
1057 zpool_open_canfail(libzfs_handle_t *hdl, const char *pool)
1058 {
1059         zpool_handle_t *zhp;
1060         boolean_t missing;
1061
1062         /*
1063          * Make sure the pool name is valid.
1064          */
1065         if (!zpool_name_valid(hdl, B_TRUE, pool)) {
1066                 (void) zfs_error_fmt(hdl, EZFS_INVALIDNAME,
1067                     dgettext(TEXT_DOMAIN, "cannot open '%s'"),
1068                     pool);
1069                 return (NULL);
1070         }
1071
1072         if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
1073                 return (NULL);
1074
1075         zhp->zpool_hdl = hdl;
1076         (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
1077
1078         if (zpool_refresh_stats(zhp, &missing) != 0) {
1079                 zpool_close(zhp);
1080                 return (NULL);
1081         }
1082
1083         if (missing) {
1084                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool"));
1085                 (void) zfs_error_fmt(hdl, EZFS_NOENT,
1086                     dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool);
1087                 zpool_close(zhp);
1088                 return (NULL);
1089         }
1090
1091         return (zhp);
1092 }
1093
1094 /*
1095  * Like the above, but silent on error.  Used when iterating over pools (because
1096  * the configuration cache may be out of date).
1097  */
1098 int
1099 zpool_open_silent(libzfs_handle_t *hdl, const char *pool, zpool_handle_t **ret)
1100 {
1101         zpool_handle_t *zhp;
1102         boolean_t missing;
1103
1104         if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
1105                 return (-1);
1106
1107         zhp->zpool_hdl = hdl;
1108         (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
1109
1110         if (zpool_refresh_stats(zhp, &missing) != 0) {
1111                 zpool_close(zhp);
1112                 return (-1);
1113         }
1114
1115         if (missing) {
1116                 zpool_close(zhp);
1117                 *ret = NULL;
1118                 return (0);
1119         }
1120
1121         *ret = zhp;
1122         return (0);
1123 }
1124
1125 /*
1126  * Similar to zpool_open_canfail(), but refuses to open pools in the faulted
1127  * state.
1128  */
1129 zpool_handle_t *
1130 zpool_open(libzfs_handle_t *hdl, const char *pool)
1131 {
1132         zpool_handle_t *zhp;
1133
1134         if ((zhp = zpool_open_canfail(hdl, pool)) == NULL)
1135                 return (NULL);
1136
1137         if (zhp->zpool_state == POOL_STATE_UNAVAIL) {
1138                 (void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
1139                     dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name);
1140                 zpool_close(zhp);
1141                 return (NULL);
1142         }
1143
1144         return (zhp);
1145 }
1146
1147 /*
1148  * Close the handle.  Simply frees the memory associated with the handle.
1149  */
1150 void
1151 zpool_close(zpool_handle_t *zhp)
1152 {
1153         nvlist_free(zhp->zpool_config);
1154         nvlist_free(zhp->zpool_old_config);
1155         nvlist_free(zhp->zpool_props);
1156         free(zhp);
1157 }
1158
1159 /*
1160  * Return the name of the pool.
1161  */
1162 const char *
1163 zpool_get_name(zpool_handle_t *zhp)
1164 {
1165         return (zhp->zpool_name);
1166 }
1167
1168
1169 /*
1170  * Return the state of the pool (ACTIVE or UNAVAILABLE)
1171  */
1172 int
1173 zpool_get_state(zpool_handle_t *zhp)
1174 {
1175         return (zhp->zpool_state);
1176 }
1177
1178 /*
1179  * Check if vdev list contains a special vdev
1180  */
1181 static boolean_t
1182 zpool_has_special_vdev(nvlist_t *nvroot)
1183 {
1184         nvlist_t **child;
1185         uint_t children;
1186
1187         if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, &child,
1188             &children) == 0) {
1189                 for (uint_t c = 0; c < children; c++) {
1190                         char *bias;
1191
1192                         if (nvlist_lookup_string(child[c],
1193                             ZPOOL_CONFIG_ALLOCATION_BIAS, &bias) == 0 &&
1194                             strcmp(bias, VDEV_ALLOC_BIAS_SPECIAL) == 0) {
1195                                 return (B_TRUE);
1196                         }
1197                 }
1198         }
1199         return (B_FALSE);
1200 }
1201
1202 /*
1203  * Create the named pool, using the provided vdev list.  It is assumed
1204  * that the consumer has already validated the contents of the nvlist, so we
1205  * don't have to worry about error semantics.
1206  */
1207 int
1208 zpool_create(libzfs_handle_t *hdl, const char *pool, nvlist_t *nvroot,
1209     nvlist_t *props, nvlist_t *fsprops)
1210 {
1211         zfs_cmd_t zc = {"\0"};
1212         nvlist_t *zc_fsprops = NULL;
1213         nvlist_t *zc_props = NULL;
1214         nvlist_t *hidden_args = NULL;
1215         uint8_t *wkeydata = NULL;
1216         uint_t wkeylen = 0;
1217         char msg[1024];
1218         int ret = -1;
1219
1220         (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1221             "cannot create '%s'"), pool);
1222
1223         if (!zpool_name_valid(hdl, B_FALSE, pool))
1224                 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
1225
1226         if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
1227                 return (-1);
1228
1229         if (props) {
1230                 prop_flags_t flags = { .create = B_TRUE, .import = B_FALSE };
1231
1232                 if ((zc_props = zpool_valid_proplist(hdl, pool, props,
1233                     SPA_VERSION_1, flags, msg)) == NULL) {
1234                         goto create_failed;
1235                 }
1236         }
1237
1238         if (fsprops) {
1239                 uint64_t zoned;
1240                 char *zonestr;
1241
1242                 zoned = ((nvlist_lookup_string(fsprops,
1243                     zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) &&
1244                     strcmp(zonestr, "on") == 0);
1245
1246                 if ((zc_fsprops = zfs_valid_proplist(hdl, ZFS_TYPE_FILESYSTEM,
1247                     fsprops, zoned, NULL, NULL, B_TRUE, msg)) == NULL) {
1248                         goto create_failed;
1249                 }
1250
1251                 if (nvlist_exists(zc_fsprops,
1252                     zfs_prop_to_name(ZFS_PROP_SPECIAL_SMALL_BLOCKS)) &&
1253                     !zpool_has_special_vdev(nvroot)) {
1254                         zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1255                             "%s property requires a special vdev"),
1256                             zfs_prop_to_name(ZFS_PROP_SPECIAL_SMALL_BLOCKS));
1257                         (void) zfs_error(hdl, EZFS_BADPROP, msg);
1258                         goto create_failed;
1259                 }
1260
1261                 if (!zc_props &&
1262                     (nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) {
1263                         goto create_failed;
1264                 }
1265                 if (zfs_crypto_create(hdl, NULL, zc_fsprops, props, B_TRUE,
1266                     &wkeydata, &wkeylen) != 0) {
1267                         zfs_error(hdl, EZFS_CRYPTOFAILED, msg);
1268                         goto create_failed;
1269                 }
1270                 if (nvlist_add_nvlist(zc_props,
1271                     ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) {
1272                         goto create_failed;
1273                 }
1274                 if (wkeydata != NULL) {
1275                         if (nvlist_alloc(&hidden_args, NV_UNIQUE_NAME, 0) != 0)
1276                                 goto create_failed;
1277
1278                         if (nvlist_add_uint8_array(hidden_args, "wkeydata",
1279                             wkeydata, wkeylen) != 0)
1280                                 goto create_failed;
1281
1282                         if (nvlist_add_nvlist(zc_props, ZPOOL_HIDDEN_ARGS,
1283                             hidden_args) != 0)
1284                                 goto create_failed;
1285                 }
1286         }
1287
1288         if (zc_props && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
1289                 goto create_failed;
1290
1291         (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name));
1292
1293         if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) {
1294
1295                 zcmd_free_nvlists(&zc);
1296                 nvlist_free(zc_props);
1297                 nvlist_free(zc_fsprops);
1298                 nvlist_free(hidden_args);
1299                 if (wkeydata != NULL)
1300                         free(wkeydata);
1301
1302                 switch (errno) {
1303                 case EBUSY:
1304                         /*
1305                          * This can happen if the user has specified the same
1306                          * device multiple times.  We can't reliably detect this
1307                          * until we try to add it and see we already have a
1308                          * label.  This can also happen under if the device is
1309                          * part of an active md or lvm device.
1310                          */
1311                         zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1312                             "one or more vdevs refer to the same device, or "
1313                             "one of\nthe devices is part of an active md or "
1314                             "lvm device"));
1315                         return (zfs_error(hdl, EZFS_BADDEV, msg));
1316
1317                 case ERANGE:
1318                         /*
1319                          * This happens if the record size is smaller or larger
1320                          * than the allowed size range, or not a power of 2.
1321                          *
1322                          * NOTE: although zfs_valid_proplist is called earlier,
1323                          * this case may have slipped through since the
1324                          * pool does not exist yet and it is therefore
1325                          * impossible to read properties e.g. max blocksize
1326                          * from the pool.
1327                          */
1328                         zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1329                             "record size invalid"));
1330                         return (zfs_error(hdl, EZFS_BADPROP, msg));
1331
1332                 case EOVERFLOW:
1333                         /*
1334                          * This occurs when one of the devices is below
1335                          * SPA_MINDEVSIZE.  Unfortunately, we can't detect which
1336                          * device was the problem device since there's no
1337                          * reliable way to determine device size from userland.
1338                          */
1339                         {
1340                                 char buf[64];
1341
1342                                 zfs_nicebytes(SPA_MINDEVSIZE, buf,
1343                                     sizeof (buf));
1344
1345                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1346                                     "one or more devices is less than the "
1347                                     "minimum size (%s)"), buf);
1348                         }
1349                         return (zfs_error(hdl, EZFS_BADDEV, msg));
1350
1351                 case ENOSPC:
1352                         zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1353                             "one or more devices is out of space"));
1354                         return (zfs_error(hdl, EZFS_BADDEV, msg));
1355
1356                 case ENOTBLK:
1357                         zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1358                             "cache device must be a disk or disk slice"));
1359                         return (zfs_error(hdl, EZFS_BADDEV, msg));
1360
1361                 default:
1362                         return (zpool_standard_error(hdl, errno, msg));
1363                 }
1364         }
1365
1366 create_failed:
1367         zcmd_free_nvlists(&zc);
1368         nvlist_free(zc_props);
1369         nvlist_free(zc_fsprops);
1370         nvlist_free(hidden_args);
1371         if (wkeydata != NULL)
1372                 free(wkeydata);
1373         return (ret);
1374 }
1375
1376 /*
1377  * Destroy the given pool.  It is up to the caller to ensure that there are no
1378  * datasets left in the pool.
1379  */
1380 int
1381 zpool_destroy(zpool_handle_t *zhp, const char *log_str)
1382 {
1383         zfs_cmd_t zc = {"\0"};
1384         zfs_handle_t *zfp = NULL;
1385         libzfs_handle_t *hdl = zhp->zpool_hdl;
1386         char msg[1024];
1387
1388         if (zhp->zpool_state == POOL_STATE_ACTIVE &&
1389             (zfp = zfs_open(hdl, zhp->zpool_name, ZFS_TYPE_FILESYSTEM)) == NULL)
1390                 return (-1);
1391
1392         (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1393         zc.zc_history = (uint64_t)(uintptr_t)log_str;
1394
1395         if (zfs_ioctl(hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) {
1396                 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1397                     "cannot destroy '%s'"), zhp->zpool_name);
1398
1399                 if (errno == EROFS) {
1400                         zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1401                             "one or more devices is read only"));
1402                         (void) zfs_error(hdl, EZFS_BADDEV, msg);
1403                 } else {
1404                         (void) zpool_standard_error(hdl, errno, msg);
1405                 }
1406
1407                 if (zfp)
1408                         zfs_close(zfp);
1409                 return (-1);
1410         }
1411
1412         if (zfp) {
1413                 remove_mountpoint(zfp);
1414                 zfs_close(zfp);
1415         }
1416
1417         return (0);
1418 }
1419
1420 /*
1421  * Create a checkpoint in the given pool.
1422  */
1423 int
1424 zpool_checkpoint(zpool_handle_t *zhp)
1425 {
1426         libzfs_handle_t *hdl = zhp->zpool_hdl;
1427         char msg[1024];
1428         int error;
1429
1430         error = lzc_pool_checkpoint(zhp->zpool_name);
1431         if (error != 0) {
1432                 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1433                     "cannot checkpoint '%s'"), zhp->zpool_name);
1434                 (void) zpool_standard_error(hdl, error, msg);
1435                 return (-1);
1436         }
1437
1438         return (0);
1439 }
1440
1441 /*
1442  * Discard the checkpoint from the given pool.
1443  */
1444 int
1445 zpool_discard_checkpoint(zpool_handle_t *zhp)
1446 {
1447         libzfs_handle_t *hdl = zhp->zpool_hdl;
1448         char msg[1024];
1449         int error;
1450
1451         error = lzc_pool_checkpoint_discard(zhp->zpool_name);
1452         if (error != 0) {
1453                 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1454                     "cannot discard checkpoint in '%s'"), zhp->zpool_name);
1455                 (void) zpool_standard_error(hdl, error, msg);
1456                 return (-1);
1457         }
1458
1459         return (0);
1460 }
1461
1462 /*
1463  * Add the given vdevs to the pool.  The caller must have already performed the
1464  * necessary verification to ensure that the vdev specification is well-formed.
1465  */
1466 int
1467 zpool_add(zpool_handle_t *zhp, nvlist_t *nvroot)
1468 {
1469         zfs_cmd_t zc = {"\0"};
1470         int ret;
1471         libzfs_handle_t *hdl = zhp->zpool_hdl;
1472         char msg[1024];
1473         nvlist_t **spares, **l2cache;
1474         uint_t nspares, nl2cache;
1475
1476         (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1477             "cannot add to '%s'"), zhp->zpool_name);
1478
1479         if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1480             SPA_VERSION_SPARES &&
1481             nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
1482             &spares, &nspares) == 0) {
1483                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1484                     "upgraded to add hot spares"));
1485                 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1486         }
1487
1488         if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1489             SPA_VERSION_L2CACHE &&
1490             nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
1491             &l2cache, &nl2cache) == 0) {
1492                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1493                     "upgraded to add cache devices"));
1494                 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1495         }
1496
1497         if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
1498                 return (-1);
1499         (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1500
1501         if (zfs_ioctl(hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) {
1502                 switch (errno) {
1503                 case EBUSY:
1504                         /*
1505                          * This can happen if the user has specified the same
1506                          * device multiple times.  We can't reliably detect this
1507                          * until we try to add it and see we already have a
1508                          * label.
1509                          */
1510                         zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1511                             "one or more vdevs refer to the same device"));
1512                         (void) zfs_error(hdl, EZFS_BADDEV, msg);
1513                         break;
1514
1515                 case EINVAL:
1516                         zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1517                             "invalid config; a pool with removing/removed "
1518                             "vdevs does not support adding raidz vdevs"));
1519                         (void) zfs_error(hdl, EZFS_BADDEV, msg);
1520                         break;
1521
1522                 case EOVERFLOW:
1523                         /*
1524                          * This occurs when one of the devices is below
1525                          * SPA_MINDEVSIZE.  Unfortunately, we can't detect which
1526                          * device was the problem device since there's no
1527                          * reliable way to determine device size from userland.
1528                          */
1529                         {
1530                                 char buf[64];
1531
1532                                 zfs_nicebytes(SPA_MINDEVSIZE, buf,
1533                                     sizeof (buf));
1534
1535                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1536                                     "device is less than the minimum "
1537                                     "size (%s)"), buf);
1538                         }
1539                         (void) zfs_error(hdl, EZFS_BADDEV, msg);
1540                         break;
1541
1542                 case ENOTSUP:
1543                         zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1544                             "pool must be upgraded to add these vdevs"));
1545                         (void) zfs_error(hdl, EZFS_BADVERSION, msg);
1546                         break;
1547
1548                 case ENOTBLK:
1549                         zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1550                             "cache device must be a disk or disk slice"));
1551                         (void) zfs_error(hdl, EZFS_BADDEV, msg);
1552                         break;
1553
1554                 default:
1555                         (void) zpool_standard_error(hdl, errno, msg);
1556                 }
1557
1558                 ret = -1;
1559         } else {
1560                 ret = 0;
1561         }
1562
1563         zcmd_free_nvlists(&zc);
1564
1565         return (ret);
1566 }
1567
1568 /*
1569  * Exports the pool from the system.  The caller must ensure that there are no
1570  * mounted datasets in the pool.
1571  */
1572 static int
1573 zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce,
1574     const char *log_str)
1575 {
1576         zfs_cmd_t zc = {"\0"};
1577         char msg[1024];
1578
1579         (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1580             "cannot export '%s'"), zhp->zpool_name);
1581
1582         (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1583         zc.zc_cookie = force;
1584         zc.zc_guid = hardforce;
1585         zc.zc_history = (uint64_t)(uintptr_t)log_str;
1586
1587         if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) {
1588                 switch (errno) {
1589                 case EXDEV:
1590                         zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN,
1591                             "use '-f' to override the following errors:\n"
1592                             "'%s' has an active shared spare which could be"
1593                             " used by other pools once '%s' is exported."),
1594                             zhp->zpool_name, zhp->zpool_name);
1595                         return (zfs_error(zhp->zpool_hdl, EZFS_ACTIVE_SPARE,
1596                             msg));
1597                 default:
1598                         return (zpool_standard_error_fmt(zhp->zpool_hdl, errno,
1599                             msg));
1600                 }
1601         }
1602
1603         return (0);
1604 }
1605
1606 int
1607 zpool_export(zpool_handle_t *zhp, boolean_t force, const char *log_str)
1608 {
1609         return (zpool_export_common(zhp, force, B_FALSE, log_str));
1610 }
1611
1612 int
1613 zpool_export_force(zpool_handle_t *zhp, const char *log_str)
1614 {
1615         return (zpool_export_common(zhp, B_TRUE, B_TRUE, log_str));
1616 }
1617
1618 static void
1619 zpool_rewind_exclaim(libzfs_handle_t *hdl, const char *name, boolean_t dryrun,
1620     nvlist_t *config)
1621 {
1622         nvlist_t *nv = NULL;
1623         uint64_t rewindto;
1624         int64_t loss = -1;
1625         struct tm t;
1626         char timestr[128];
1627
1628         if (!hdl->libzfs_printerr || config == NULL)
1629                 return;
1630
1631         if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
1632             nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0) {
1633                 return;
1634         }
1635
1636         if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1637                 return;
1638         (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
1639
1640         if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1641             strftime(timestr, 128, "%c", &t) != 0) {
1642                 if (dryrun) {
1643                         (void) printf(dgettext(TEXT_DOMAIN,
1644                             "Would be able to return %s "
1645                             "to its state as of %s.\n"),
1646                             name, timestr);
1647                 } else {
1648                         (void) printf(dgettext(TEXT_DOMAIN,
1649                             "Pool %s returned to its state as of %s.\n"),
1650                             name, timestr);
1651                 }
1652                 if (loss > 120) {
1653                         (void) printf(dgettext(TEXT_DOMAIN,
1654                             "%s approximately %lld "),
1655                             dryrun ? "Would discard" : "Discarded",
1656                             ((longlong_t)loss + 30) / 60);
1657                         (void) printf(dgettext(TEXT_DOMAIN,
1658                             "minutes of transactions.\n"));
1659                 } else if (loss > 0) {
1660                         (void) printf(dgettext(TEXT_DOMAIN,
1661                             "%s approximately %lld "),
1662                             dryrun ? "Would discard" : "Discarded",
1663                             (longlong_t)loss);
1664                         (void) printf(dgettext(TEXT_DOMAIN,
1665                             "seconds of transactions.\n"));
1666                 }
1667         }
1668 }
1669
1670 void
1671 zpool_explain_recover(libzfs_handle_t *hdl, const char *name, int reason,
1672     nvlist_t *config)
1673 {
1674         nvlist_t *nv = NULL;
1675         int64_t loss = -1;
1676         uint64_t edata = UINT64_MAX;
1677         uint64_t rewindto;
1678         struct tm t;
1679         char timestr[128];
1680
1681         if (!hdl->libzfs_printerr)
1682                 return;
1683
1684         if (reason >= 0)
1685                 (void) printf(dgettext(TEXT_DOMAIN, "action: "));
1686         else
1687                 (void) printf(dgettext(TEXT_DOMAIN, "\t"));
1688
1689         /* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */
1690         if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
1691             nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0 ||
1692             nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1693                 goto no_info;
1694
1695         (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
1696         (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_DATA_ERRORS,
1697             &edata);
1698
1699         (void) printf(dgettext(TEXT_DOMAIN,
1700             "Recovery is possible, but will result in some data loss.\n"));
1701
1702         if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1703             strftime(timestr, 128, "%c", &t) != 0) {
1704                 (void) printf(dgettext(TEXT_DOMAIN,
1705                     "\tReturning the pool to its state as of %s\n"
1706                     "\tshould correct the problem.  "),
1707                     timestr);
1708         } else {
1709                 (void) printf(dgettext(TEXT_DOMAIN,
1710                     "\tReverting the pool to an earlier state "
1711                     "should correct the problem.\n\t"));
1712         }
1713
1714         if (loss > 120) {
1715                 (void) printf(dgettext(TEXT_DOMAIN,
1716                     "Approximately %lld minutes of data\n"
1717                     "\tmust be discarded, irreversibly.  "),
1718                     ((longlong_t)loss + 30) / 60);
1719         } else if (loss > 0) {
1720                 (void) printf(dgettext(TEXT_DOMAIN,
1721                     "Approximately %lld seconds of data\n"
1722                     "\tmust be discarded, irreversibly.  "),
1723                     (longlong_t)loss);
1724         }
1725         if (edata != 0 && edata != UINT64_MAX) {
1726                 if (edata == 1) {
1727                         (void) printf(dgettext(TEXT_DOMAIN,
1728                             "After rewind, at least\n"
1729                             "\tone persistent user-data error will remain.  "));
1730                 } else {
1731                         (void) printf(dgettext(TEXT_DOMAIN,
1732                             "After rewind, several\n"
1733                             "\tpersistent user-data errors will remain.  "));
1734                 }
1735         }
1736         (void) printf(dgettext(TEXT_DOMAIN,
1737             "Recovery can be attempted\n\tby executing 'zpool %s -F %s'.  "),
1738             reason >= 0 ? "clear" : "import", name);
1739
1740         (void) printf(dgettext(TEXT_DOMAIN,
1741             "A scrub of the pool\n"
1742             "\tis strongly recommended after recovery.\n"));
1743         return;
1744
1745 no_info:
1746         (void) printf(dgettext(TEXT_DOMAIN,
1747             "Destroy and re-create the pool from\n\ta backup source.\n"));
1748 }
1749
1750 /*
1751  * zpool_import() is a contracted interface. Should be kept the same
1752  * if possible.
1753  *
1754  * Applications should use zpool_import_props() to import a pool with
1755  * new properties value to be set.
1756  */
1757 int
1758 zpool_import(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1759     char *altroot)
1760 {
1761         nvlist_t *props = NULL;
1762         int ret;
1763
1764         if (altroot != NULL) {
1765                 if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) {
1766                         return (zfs_error_fmt(hdl, EZFS_NOMEM,
1767                             dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1768                             newname));
1769                 }
1770
1771                 if (nvlist_add_string(props,
1772                     zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 ||
1773                     nvlist_add_string(props,
1774                     zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) {
1775                         nvlist_free(props);
1776                         return (zfs_error_fmt(hdl, EZFS_NOMEM,
1777                             dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1778                             newname));
1779                 }
1780         }
1781
1782         ret = zpool_import_props(hdl, config, newname, props,
1783             ZFS_IMPORT_NORMAL);
1784         nvlist_free(props);
1785         return (ret);
1786 }
1787
1788 static void
1789 print_vdev_tree(libzfs_handle_t *hdl, const char *name, nvlist_t *nv,
1790     int indent)
1791 {
1792         nvlist_t **child;
1793         uint_t c, children;
1794         char *vname;
1795         uint64_t is_log = 0;
1796
1797         (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG,
1798             &is_log);
1799
1800         if (name != NULL)
1801                 (void) printf("\t%*s%s%s\n", indent, "", name,
1802                     is_log ? " [log]" : "");
1803
1804         if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1805             &child, &children) != 0)
1806                 return;
1807
1808         for (c = 0; c < children; c++) {
1809                 vname = zpool_vdev_name(hdl, NULL, child[c], VDEV_NAME_TYPE_ID);
1810                 print_vdev_tree(hdl, vname, child[c], indent + 2);
1811                 free(vname);
1812         }
1813 }
1814
1815 void
1816 zpool_print_unsup_feat(nvlist_t *config)
1817 {
1818         nvlist_t *nvinfo, *unsup_feat;
1819         nvpair_t *nvp;
1820
1821         verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nvinfo) ==
1822             0);
1823         verify(nvlist_lookup_nvlist(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT,
1824             &unsup_feat) == 0);
1825
1826         for (nvp = nvlist_next_nvpair(unsup_feat, NULL); nvp != NULL;
1827             nvp = nvlist_next_nvpair(unsup_feat, nvp)) {
1828                 char *desc;
1829
1830                 verify(nvpair_type(nvp) == DATA_TYPE_STRING);
1831                 verify(nvpair_value_string(nvp, &desc) == 0);
1832
1833                 if (strlen(desc) > 0)
1834                         (void) printf("\t%s (%s)\n", nvpair_name(nvp), desc);
1835                 else
1836                         (void) printf("\t%s\n", nvpair_name(nvp));
1837         }
1838 }
1839
1840 /*
1841  * Import the given pool using the known configuration and a list of
1842  * properties to be set. The configuration should have come from
1843  * zpool_find_import(). The 'newname' parameters control whether the pool
1844  * is imported with a different name.
1845  */
1846 int
1847 zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1848     nvlist_t *props, int flags)
1849 {
1850         zfs_cmd_t zc = {"\0"};
1851         zpool_load_policy_t policy;
1852         nvlist_t *nv = NULL;
1853         nvlist_t *nvinfo = NULL;
1854         nvlist_t *missing = NULL;
1855         char *thename;
1856         char *origname;
1857         int ret;
1858         int error = 0;
1859         char errbuf[1024];
1860
1861         verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
1862             &origname) == 0);
1863
1864         (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1865             "cannot import pool '%s'"), origname);
1866
1867         if (newname != NULL) {
1868                 if (!zpool_name_valid(hdl, B_FALSE, newname))
1869                         return (zfs_error_fmt(hdl, EZFS_INVALIDNAME,
1870                             dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1871                             newname));
1872                 thename = (char *)newname;
1873         } else {
1874                 thename = origname;
1875         }
1876
1877         if (props != NULL) {
1878                 uint64_t version;
1879                 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
1880
1881                 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
1882                     &version) == 0);
1883
1884                 if ((props = zpool_valid_proplist(hdl, origname,
1885                     props, version, flags, errbuf)) == NULL)
1886                         return (-1);
1887                 if (zcmd_write_src_nvlist(hdl, &zc, props) != 0) {
1888                         nvlist_free(props);
1889                         return (-1);
1890                 }
1891                 nvlist_free(props);
1892         }
1893
1894         (void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name));
1895
1896         verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
1897             &zc.zc_guid) == 0);
1898
1899         if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) {
1900                 zcmd_free_nvlists(&zc);
1901                 return (-1);
1902         }
1903         if (zcmd_alloc_dst_nvlist(hdl, &zc, zc.zc_nvlist_conf_size * 2) != 0) {
1904                 zcmd_free_nvlists(&zc);
1905                 return (-1);
1906         }
1907
1908         zc.zc_cookie = flags;
1909         while ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc)) != 0 &&
1910             errno == ENOMEM) {
1911                 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
1912                         zcmd_free_nvlists(&zc);
1913                         return (-1);
1914                 }
1915         }
1916         if (ret != 0)
1917                 error = errno;
1918
1919         (void) zcmd_read_dst_nvlist(hdl, &zc, &nv);
1920
1921         zcmd_free_nvlists(&zc);
1922
1923         zpool_get_load_policy(config, &policy);
1924
1925         if (error) {
1926                 char desc[1024];
1927                 char aux[256];
1928
1929                 /*
1930                  * Dry-run failed, but we print out what success
1931                  * looks like if we found a best txg
1932                  */
1933                 if (policy.zlp_rewind & ZPOOL_TRY_REWIND) {
1934                         zpool_rewind_exclaim(hdl, newname ? origname : thename,
1935                             B_TRUE, nv);
1936                         nvlist_free(nv);
1937                         return (-1);
1938                 }
1939
1940                 if (newname == NULL)
1941                         (void) snprintf(desc, sizeof (desc),
1942                             dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1943                             thename);
1944                 else
1945                         (void) snprintf(desc, sizeof (desc),
1946                             dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"),
1947                             origname, thename);
1948
1949                 switch (error) {
1950                 case ENOTSUP:
1951                         if (nv != NULL && nvlist_lookup_nvlist(nv,
1952                             ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
1953                             nvlist_exists(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT)) {
1954                                 (void) printf(dgettext(TEXT_DOMAIN, "This "
1955                                     "pool uses the following feature(s) not "
1956                                     "supported by this system:\n"));
1957                                 zpool_print_unsup_feat(nv);
1958                                 if (nvlist_exists(nvinfo,
1959                                     ZPOOL_CONFIG_CAN_RDONLY)) {
1960                                         (void) printf(dgettext(TEXT_DOMAIN,
1961                                             "All unsupported features are only "
1962                                             "required for writing to the pool."
1963                                             "\nThe pool can be imported using "
1964                                             "'-o readonly=on'.\n"));
1965                                 }
1966                         }
1967                         /*
1968                          * Unsupported version.
1969                          */
1970                         (void) zfs_error(hdl, EZFS_BADVERSION, desc);
1971                         break;
1972
1973                 case EREMOTEIO:
1974                         if (nv != NULL && nvlist_lookup_nvlist(nv,
1975                             ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0) {
1976                                 char *hostname = "<unknown>";
1977                                 uint64_t hostid = 0;
1978                                 mmp_state_t mmp_state;
1979
1980                                 mmp_state = fnvlist_lookup_uint64(nvinfo,
1981                                     ZPOOL_CONFIG_MMP_STATE);
1982
1983                                 if (nvlist_exists(nvinfo,
1984                                     ZPOOL_CONFIG_MMP_HOSTNAME))
1985                                         hostname = fnvlist_lookup_string(nvinfo,
1986                                             ZPOOL_CONFIG_MMP_HOSTNAME);
1987
1988                                 if (nvlist_exists(nvinfo,
1989                                     ZPOOL_CONFIG_MMP_HOSTID))
1990                                         hostid = fnvlist_lookup_uint64(nvinfo,
1991                                             ZPOOL_CONFIG_MMP_HOSTID);
1992
1993                                 if (mmp_state == MMP_STATE_ACTIVE) {
1994                                         (void) snprintf(aux, sizeof (aux),
1995                                             dgettext(TEXT_DOMAIN, "pool is imp"
1996                                             "orted on host '%s' (hostid=%lx).\n"
1997                                             "Export the pool on the other "
1998                                             "system, then run 'zpool import'."),
1999                                             hostname, (unsigned long) hostid);
2000                                 } else if (mmp_state == MMP_STATE_NO_HOSTID) {
2001                                         (void) snprintf(aux, sizeof (aux),
2002                                             dgettext(TEXT_DOMAIN, "pool has "
2003                                             "the multihost property on and "
2004                                             "the\nsystem's hostid is not set. "
2005                                             "Set a unique system hostid with "
2006                                             "the zgenhostid(8) command.\n"));
2007                                 }
2008
2009                                 (void) zfs_error_aux(hdl, aux);
2010                         }
2011                         (void) zfs_error(hdl, EZFS_ACTIVE_POOL, desc);
2012                         break;
2013
2014                 case EINVAL:
2015                         (void) zfs_error(hdl, EZFS_INVALCONFIG, desc);
2016                         break;
2017
2018                 case EROFS:
2019                         zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2020                             "one or more devices is read only"));
2021                         (void) zfs_error(hdl, EZFS_BADDEV, desc);
2022                         break;
2023
2024                 case ENXIO:
2025                         if (nv && nvlist_lookup_nvlist(nv,
2026                             ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
2027                             nvlist_lookup_nvlist(nvinfo,
2028                             ZPOOL_CONFIG_MISSING_DEVICES, &missing) == 0) {
2029                                 (void) printf(dgettext(TEXT_DOMAIN,
2030                                     "The devices below are missing or "
2031                                     "corrupted, use '-m' to import the pool "
2032                                     "anyway:\n"));
2033                                 print_vdev_tree(hdl, NULL, missing, 2);
2034                                 (void) printf("\n");
2035                         }
2036                         (void) zpool_standard_error(hdl, error, desc);
2037                         break;
2038
2039                 case EEXIST:
2040                         (void) zpool_standard_error(hdl, error, desc);
2041                         break;
2042
2043                 case EBUSY:
2044                         zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2045                             "one or more devices are already in use\n"));
2046                         (void) zfs_error(hdl, EZFS_BADDEV, desc);
2047                         break;
2048                 case ENAMETOOLONG:
2049                         zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2050                             "new name of at least one dataset is longer than "
2051                             "the maximum allowable length"));
2052                         (void) zfs_error(hdl, EZFS_NAMETOOLONG, desc);
2053                         break;
2054                 default:
2055                         (void) zpool_standard_error(hdl, error, desc);
2056                         zpool_explain_recover(hdl,
2057                             newname ? origname : thename, -error, nv);
2058                         break;
2059                 }
2060
2061                 nvlist_free(nv);
2062                 ret = -1;
2063         } else {
2064                 zpool_handle_t *zhp;
2065
2066                 /*
2067                  * This should never fail, but play it safe anyway.
2068                  */
2069                 if (zpool_open_silent(hdl, thename, &zhp) != 0)
2070                         ret = -1;
2071                 else if (zhp != NULL)
2072                         zpool_close(zhp);
2073                 if (policy.zlp_rewind &
2074                     (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
2075                         zpool_rewind_exclaim(hdl, newname ? origname : thename,
2076                             ((policy.zlp_rewind & ZPOOL_TRY_REWIND) != 0), nv);
2077                 }
2078                 nvlist_free(nv);
2079                 return (0);
2080         }
2081
2082         return (ret);
2083 }
2084
2085 /*
2086  * Translate vdev names to guids.  If a vdev_path is determined to be
2087  * unsuitable then a vd_errlist is allocated and the vdev path and errno
2088  * are added to it.
2089  */
2090 static int
2091 zpool_translate_vdev_guids(zpool_handle_t *zhp, nvlist_t *vds,
2092     nvlist_t *vdev_guids, nvlist_t *guids_to_paths, nvlist_t **vd_errlist)
2093 {
2094         nvlist_t *errlist = NULL;
2095         int error = 0;
2096
2097         for (nvpair_t *elem = nvlist_next_nvpair(vds, NULL); elem != NULL;
2098             elem = nvlist_next_nvpair(vds, elem)) {
2099                 boolean_t spare, cache;
2100
2101                 char *vd_path = nvpair_name(elem);
2102                 nvlist_t *tgt = zpool_find_vdev(zhp, vd_path, &spare, &cache,
2103                     NULL);
2104
2105                 if ((tgt == NULL) || cache || spare) {
2106                         if (errlist == NULL) {
2107                                 errlist = fnvlist_alloc();
2108                                 error = EINVAL;
2109                         }
2110
2111                         uint64_t err = (tgt == NULL) ? EZFS_NODEVICE :
2112                             (spare ? EZFS_ISSPARE : EZFS_ISL2CACHE);
2113                         fnvlist_add_int64(errlist, vd_path, err);
2114                         continue;
2115                 }
2116
2117                 uint64_t guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
2118                 fnvlist_add_uint64(vdev_guids, vd_path, guid);
2119
2120                 char msg[MAXNAMELEN];
2121                 (void) snprintf(msg, sizeof (msg), "%llu", (u_longlong_t)guid);
2122                 fnvlist_add_string(guids_to_paths, msg, vd_path);
2123         }
2124
2125         if (error != 0) {
2126                 verify(errlist != NULL);
2127                 if (vd_errlist != NULL)
2128                         *vd_errlist = errlist;
2129                 else
2130                         fnvlist_free(errlist);
2131         }
2132
2133         return (error);
2134 }
2135
2136 static int
2137 xlate_init_err(int err)
2138 {
2139         switch (err) {
2140         case ENODEV:
2141                 return (EZFS_NODEVICE);
2142         case EINVAL:
2143         case EROFS:
2144                 return (EZFS_BADDEV);
2145         case EBUSY:
2146                 return (EZFS_INITIALIZING);
2147         case ESRCH:
2148                 return (EZFS_NO_INITIALIZE);
2149         }
2150         return (err);
2151 }
2152
2153 /*
2154  * Begin, suspend, or cancel the initialization (initializing of all free
2155  * blocks) for the given vdevs in the given pool.
2156  */
2157 int
2158 zpool_initialize_impl(zpool_handle_t *zhp, pool_initialize_func_t cmd_type,
2159     nvlist_t *vds, boolean_t wait)
2160 {
2161         int err;
2162
2163         nvlist_t *vdev_guids = fnvlist_alloc();
2164         nvlist_t *guids_to_paths = fnvlist_alloc();
2165         nvlist_t *vd_errlist = NULL;
2166         nvlist_t *errlist;
2167         nvpair_t *elem;
2168
2169         err = zpool_translate_vdev_guids(zhp, vds, vdev_guids,
2170             guids_to_paths, &vd_errlist);
2171
2172         if (err != 0) {
2173                 verify(vd_errlist != NULL);
2174                 goto list_errors;
2175         }
2176
2177         err = lzc_initialize(zhp->zpool_name, cmd_type,
2178             vdev_guids, &errlist);
2179
2180         if (err != 0) {
2181                 if (errlist != NULL) {
2182                         vd_errlist = fnvlist_lookup_nvlist(errlist,
2183                             ZPOOL_INITIALIZE_VDEVS);
2184                         goto list_errors;
2185                 }
2186                 (void) zpool_standard_error(zhp->zpool_hdl, err,
2187                     dgettext(TEXT_DOMAIN, "operation failed"));
2188                 goto out;
2189         }
2190
2191         if (wait) {
2192                 for (elem = nvlist_next_nvpair(vdev_guids, NULL); elem != NULL;
2193                     elem = nvlist_next_nvpair(vdev_guids, elem)) {
2194
2195                         uint64_t guid = fnvpair_value_uint64(elem);
2196
2197                         err = lzc_wait_tag(zhp->zpool_name,
2198                             ZPOOL_WAIT_INITIALIZE, guid, NULL);
2199                         if (err != 0) {
2200                                 (void) zpool_standard_error_fmt(zhp->zpool_hdl,
2201                                     err, dgettext(TEXT_DOMAIN, "error "
2202                                     "waiting for '%s' to initialize"),
2203                                     nvpair_name(elem));
2204
2205                                 goto out;
2206                         }
2207                 }
2208         }
2209         goto out;
2210
2211 list_errors:
2212         for (elem = nvlist_next_nvpair(vd_errlist, NULL); elem != NULL;
2213             elem = nvlist_next_nvpair(vd_errlist, elem)) {
2214                 int64_t vd_error = xlate_init_err(fnvpair_value_int64(elem));
2215                 char *path;
2216
2217                 if (nvlist_lookup_string(guids_to_paths, nvpair_name(elem),
2218                     &path) != 0)
2219                         path = nvpair_name(elem);
2220
2221                 (void) zfs_error_fmt(zhp->zpool_hdl, vd_error,
2222                     "cannot initialize '%s'", path);
2223         }
2224
2225 out:
2226         fnvlist_free(vdev_guids);
2227         fnvlist_free(guids_to_paths);
2228
2229         if (vd_errlist != NULL)
2230                 fnvlist_free(vd_errlist);
2231
2232         return (err == 0 ? 0 : -1);
2233 }
2234
2235 int
2236 zpool_initialize(zpool_handle_t *zhp, pool_initialize_func_t cmd_type,
2237     nvlist_t *vds)
2238 {
2239         return (zpool_initialize_impl(zhp, cmd_type, vds, B_FALSE));
2240 }
2241
2242 int
2243 zpool_initialize_wait(zpool_handle_t *zhp, pool_initialize_func_t cmd_type,
2244     nvlist_t *vds)
2245 {
2246         return (zpool_initialize_impl(zhp, cmd_type, vds, B_TRUE));
2247 }
2248
2249 static int
2250 xlate_trim_err(int err)
2251 {
2252         switch (err) {
2253         case ENODEV:
2254                 return (EZFS_NODEVICE);
2255         case EINVAL:
2256         case EROFS:
2257                 return (EZFS_BADDEV);
2258         case EBUSY:
2259                 return (EZFS_TRIMMING);
2260         case ESRCH:
2261                 return (EZFS_NO_TRIM);
2262         case EOPNOTSUPP:
2263                 return (EZFS_TRIM_NOTSUP);
2264         }
2265         return (err);
2266 }
2267
2268 /*
2269  * Begin, suspend, or cancel the TRIM (discarding of all free blocks) for
2270  * the given vdevs in the given pool.
2271  */
2272 int
2273 zpool_trim(zpool_handle_t *zhp, pool_trim_func_t cmd_type, nvlist_t *vds,
2274     trimflags_t *trim_flags)
2275 {
2276         char msg[1024];
2277         int err;
2278
2279         nvlist_t *vdev_guids = fnvlist_alloc();
2280         nvlist_t *guids_to_paths = fnvlist_alloc();
2281         nvlist_t *vd_errlist = NULL;
2282         nvlist_t *errlist;
2283         nvpair_t *elem;
2284
2285         err = zpool_translate_vdev_guids(zhp, vds, vdev_guids,
2286             guids_to_paths, &vd_errlist);
2287         if (err == 0) {
2288                 err = lzc_trim(zhp->zpool_name, cmd_type, trim_flags->rate,
2289                     trim_flags->secure, vdev_guids, &errlist);
2290                 if (err == 0) {
2291                         fnvlist_free(vdev_guids);
2292                         fnvlist_free(guids_to_paths);
2293                         return (0);
2294                 }
2295
2296                 if (errlist != NULL) {
2297                         vd_errlist = fnvlist_lookup_nvlist(errlist,
2298                             ZPOOL_TRIM_VDEVS);
2299                 }
2300
2301                 (void) snprintf(msg, sizeof (msg),
2302                     dgettext(TEXT_DOMAIN, "operation failed"));
2303         } else {
2304                 verify(vd_errlist != NULL);
2305         }
2306
2307         for (elem = nvlist_next_nvpair(vd_errlist, NULL);
2308             elem != NULL; elem = nvlist_next_nvpair(vd_errlist, elem)) {
2309                 int64_t vd_error = xlate_trim_err(fnvpair_value_int64(elem));
2310                 char *path;
2311
2312                 /*
2313                  * If only the pool was specified, and it was not a secure
2314                  * trim then suppress warnings for individual vdevs which
2315                  * do not support trimming.
2316                  */
2317                 if (vd_error == EZFS_TRIM_NOTSUP &&
2318                     trim_flags->fullpool &&
2319                     !trim_flags->secure) {
2320                         continue;
2321                 }
2322
2323                 if (nvlist_lookup_string(guids_to_paths, nvpair_name(elem),
2324                     &path) != 0)
2325                         path = nvpair_name(elem);
2326
2327                 (void) zfs_error_fmt(zhp->zpool_hdl, vd_error,
2328                     "cannot trim '%s'", path);
2329         }
2330
2331         fnvlist_free(vdev_guids);
2332         fnvlist_free(guids_to_paths);
2333
2334         if (vd_errlist != NULL) {
2335                 fnvlist_free(vd_errlist);
2336                 return (-1);
2337         }
2338
2339         return (zpool_standard_error(zhp->zpool_hdl, err, msg));
2340 }
2341
2342 /*
2343  * Scan the pool.
2344  */
2345 int
2346 zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func, pool_scrub_cmd_t cmd)
2347 {
2348         zfs_cmd_t zc = {"\0"};
2349         char msg[1024];
2350         int err;
2351         libzfs_handle_t *hdl = zhp->zpool_hdl;
2352
2353         (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2354         zc.zc_cookie = func;
2355         zc.zc_flags = cmd;
2356
2357         if (zfs_ioctl(hdl, ZFS_IOC_POOL_SCAN, &zc) == 0)
2358                 return (0);
2359
2360         err = errno;
2361
2362         /* ECANCELED on a scrub means we resumed a paused scrub */
2363         if (err == ECANCELED && func == POOL_SCAN_SCRUB &&
2364             cmd == POOL_SCRUB_NORMAL)
2365                 return (0);
2366
2367         if (err == ENOENT && func != POOL_SCAN_NONE && cmd == POOL_SCRUB_NORMAL)
2368                 return (0);
2369
2370         if (func == POOL_SCAN_SCRUB) {
2371                 if (cmd == POOL_SCRUB_PAUSE) {
2372                         (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2373                             "cannot pause scrubbing %s"), zc.zc_name);
2374                 } else {
2375                         assert(cmd == POOL_SCRUB_NORMAL);
2376                         (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2377                             "cannot scrub %s"), zc.zc_name);
2378                 }
2379         } else if (func == POOL_SCAN_RESILVER) {
2380                 assert(cmd == POOL_SCRUB_NORMAL);
2381                 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2382                     "cannot restart resilver on %s"), zc.zc_name);
2383         } else if (func == POOL_SCAN_NONE) {
2384                 (void) snprintf(msg, sizeof (msg),
2385                     dgettext(TEXT_DOMAIN, "cannot cancel scrubbing %s"),
2386                     zc.zc_name);
2387         } else {
2388                 assert(!"unexpected result");
2389         }
2390
2391         if (err == EBUSY) {
2392                 nvlist_t *nvroot;
2393                 pool_scan_stat_t *ps = NULL;
2394                 uint_t psc;
2395
2396                 verify(nvlist_lookup_nvlist(zhp->zpool_config,
2397                     ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
2398                 (void) nvlist_lookup_uint64_array(nvroot,
2399                     ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &psc);
2400                 if (ps && ps->pss_func == POOL_SCAN_SCRUB) {
2401                         if (cmd == POOL_SCRUB_PAUSE)
2402                                 return (zfs_error(hdl, EZFS_SCRUB_PAUSED, msg));
2403                         else
2404                                 return (zfs_error(hdl, EZFS_SCRUBBING, msg));
2405                 } else {
2406                         return (zfs_error(hdl, EZFS_RESILVERING, msg));
2407                 }
2408         } else if (err == ENOENT) {
2409                 return (zfs_error(hdl, EZFS_NO_SCRUB, msg));
2410         } else if (err == ENOTSUP && func == POOL_SCAN_RESILVER) {
2411                 return (zfs_error(hdl, EZFS_NO_RESILVER_DEFER, msg));
2412         } else {
2413                 return (zpool_standard_error(hdl, err, msg));
2414         }
2415 }
2416
2417 /*
2418  * Find a vdev that matches the search criteria specified. We use the
2419  * the nvpair name to determine how we should look for the device.
2420  * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL
2421  * spare; but FALSE if its an INUSE spare.
2422  */
2423 static nvlist_t *
2424 vdev_to_nvlist_iter(nvlist_t *nv, nvlist_t *search, boolean_t *avail_spare,
2425     boolean_t *l2cache, boolean_t *log)
2426 {
2427         uint_t c, children;
2428         nvlist_t **child;
2429         nvlist_t *ret;
2430         uint64_t is_log;
2431         char *srchkey;
2432         nvpair_t *pair = nvlist_next_nvpair(search, NULL);
2433
2434         /* Nothing to look for */
2435         if (search == NULL || pair == NULL)
2436                 return (NULL);
2437
2438         /* Obtain the key we will use to search */
2439         srchkey = nvpair_name(pair);
2440
2441         switch (nvpair_type(pair)) {
2442         case DATA_TYPE_UINT64:
2443                 if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) {
2444                         uint64_t srchval, theguid;
2445
2446                         verify(nvpair_value_uint64(pair, &srchval) == 0);
2447                         verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
2448                             &theguid) == 0);
2449                         if (theguid == srchval)
2450                                 return (nv);
2451                 }
2452                 break;
2453
2454         case DATA_TYPE_STRING: {
2455                 char *srchval, *val;
2456
2457                 verify(nvpair_value_string(pair, &srchval) == 0);
2458                 if (nvlist_lookup_string(nv, srchkey, &val) != 0)
2459                         break;
2460
2461                 /*
2462                  * Search for the requested value. Special cases:
2463                  *
2464                  * - ZPOOL_CONFIG_PATH for whole disk entries.  These end in
2465                  *   "-part1", or "p1".  The suffix is hidden from the user,
2466                  *   but included in the string, so this matches around it.
2467                  * - ZPOOL_CONFIG_PATH for short names zfs_strcmp_shortname()
2468                  *   is used to check all possible expanded paths.
2469                  * - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE).
2470                  *
2471                  * Otherwise, all other searches are simple string compares.
2472                  */
2473                 if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0) {
2474                         uint64_t wholedisk = 0;
2475
2476                         (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
2477                             &wholedisk);
2478                         if (zfs_strcmp_pathname(srchval, val, wholedisk) == 0)
2479                                 return (nv);
2480
2481                 } else if (strcmp(srchkey, ZPOOL_CONFIG_TYPE) == 0 && val) {
2482                         char *type, *idx, *end, *p;
2483                         uint64_t id, vdev_id;
2484
2485                         /*
2486                          * Determine our vdev type, keeping in mind
2487                          * that the srchval is composed of a type and
2488                          * vdev id pair (i.e. mirror-4).
2489                          */
2490                         if ((type = strdup(srchval)) == NULL)
2491                                 return (NULL);
2492
2493                         if ((p = strrchr(type, '-')) == NULL) {
2494                                 free(type);
2495                                 break;
2496                         }
2497                         idx = p + 1;
2498                         *p = '\0';
2499
2500                         /*
2501                          * If the types don't match then keep looking.
2502                          */
2503                         if (strncmp(val, type, strlen(val)) != 0) {
2504                                 free(type);
2505                                 break;
2506                         }
2507
2508                         verify(zpool_vdev_is_interior(type));
2509                         verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
2510                             &id) == 0);
2511
2512                         errno = 0;
2513                         vdev_id = strtoull(idx, &end, 10);
2514
2515                         free(type);
2516                         if (errno != 0)
2517                                 return (NULL);
2518
2519                         /*
2520                          * Now verify that we have the correct vdev id.
2521                          */
2522                         if (vdev_id == id)
2523                                 return (nv);
2524                 }
2525
2526                 /*
2527                  * Common case
2528                  */
2529                 if (strcmp(srchval, val) == 0)
2530                         return (nv);
2531                 break;
2532         }
2533
2534         default:
2535                 break;
2536         }
2537
2538         if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
2539             &child, &children) != 0)
2540                 return (NULL);
2541
2542         for (c = 0; c < children; c++) {
2543                 if ((ret = vdev_to_nvlist_iter(child[c], search,
2544                     avail_spare, l2cache, NULL)) != NULL) {
2545                         /*
2546                          * The 'is_log' value is only set for the toplevel
2547                          * vdev, not the leaf vdevs.  So we always lookup the
2548                          * log device from the root of the vdev tree (where
2549                          * 'log' is non-NULL).
2550                          */
2551                         if (log != NULL &&
2552                             nvlist_lookup_uint64(child[c],
2553                             ZPOOL_CONFIG_IS_LOG, &is_log) == 0 &&
2554                             is_log) {
2555                                 *log = B_TRUE;
2556                         }
2557                         return (ret);
2558                 }
2559         }
2560
2561         if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
2562             &child, &children) == 0) {
2563                 for (c = 0; c < children; c++) {
2564                         if ((ret = vdev_to_nvlist_iter(child[c], search,
2565                             avail_spare, l2cache, NULL)) != NULL) {
2566                                 *avail_spare = B_TRUE;
2567                                 return (ret);
2568                         }
2569                 }
2570         }
2571
2572         if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
2573             &child, &children) == 0) {
2574                 for (c = 0; c < children; c++) {
2575                         if ((ret = vdev_to_nvlist_iter(child[c], search,
2576                             avail_spare, l2cache, NULL)) != NULL) {
2577                                 *l2cache = B_TRUE;
2578                                 return (ret);
2579                         }
2580                 }
2581         }
2582
2583         return (NULL);
2584 }
2585
2586 /*
2587  * Given a physical path or guid, find the associated vdev.
2588  */
2589 nvlist_t *
2590 zpool_find_vdev_by_physpath(zpool_handle_t *zhp, const char *ppath,
2591     boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log)
2592 {
2593         nvlist_t *search, *nvroot, *ret;
2594         uint64_t guid;
2595         char *end;
2596
2597         verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2598
2599         guid = strtoull(ppath, &end, 0);
2600         if (guid != 0 && *end == '\0') {
2601                 verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0);
2602         } else {
2603                 verify(nvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH,
2604                     ppath) == 0);
2605         }
2606
2607         verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
2608             &nvroot) == 0);
2609
2610         *avail_spare = B_FALSE;
2611         *l2cache = B_FALSE;
2612         if (log != NULL)
2613                 *log = B_FALSE;
2614         ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
2615         nvlist_free(search);
2616
2617         return (ret);
2618 }
2619
2620 /*
2621  * Determine if we have an "interior" top-level vdev (i.e mirror/raidz).
2622  */
2623 static boolean_t
2624 zpool_vdev_is_interior(const char *name)
2625 {
2626         if (strncmp(name, VDEV_TYPE_RAIDZ, strlen(VDEV_TYPE_RAIDZ)) == 0 ||
2627             strncmp(name, VDEV_TYPE_SPARE, strlen(VDEV_TYPE_SPARE)) == 0 ||
2628             strncmp(name,
2629             VDEV_TYPE_REPLACING, strlen(VDEV_TYPE_REPLACING)) == 0 ||
2630             strncmp(name, VDEV_TYPE_MIRROR, strlen(VDEV_TYPE_MIRROR)) == 0)
2631                 return (B_TRUE);
2632         return (B_FALSE);
2633 }
2634
2635 nvlist_t *
2636 zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare,
2637     boolean_t *l2cache, boolean_t *log)
2638 {
2639         char *end;
2640         nvlist_t *nvroot, *search, *ret;
2641         uint64_t guid;
2642
2643         verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2644
2645         guid = strtoull(path, &end, 0);
2646         if (guid != 0 && *end == '\0') {
2647                 verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0);
2648         } else if (zpool_vdev_is_interior(path)) {
2649                 verify(nvlist_add_string(search, ZPOOL_CONFIG_TYPE, path) == 0);
2650         } else {
2651                 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, path) == 0);
2652         }
2653
2654         verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
2655             &nvroot) == 0);
2656
2657         *avail_spare = B_FALSE;
2658         *l2cache = B_FALSE;
2659         if (log != NULL)
2660                 *log = B_FALSE;
2661         ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
2662         nvlist_free(search);
2663
2664         return (ret);
2665 }
2666
2667 static int
2668 vdev_is_online(nvlist_t *nv)
2669 {
2670         uint64_t ival;
2671
2672         if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE, &ival) == 0 ||
2673             nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED, &ival) == 0 ||
2674             nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED, &ival) == 0)
2675                 return (0);
2676
2677         return (1);
2678 }
2679
2680 /*
2681  * Helper function for zpool_get_physpaths().
2682  */
2683 static int
2684 vdev_get_one_physpath(nvlist_t *config, char *physpath, size_t physpath_size,
2685     size_t *bytes_written)
2686 {
2687         size_t bytes_left, pos, rsz;
2688         char *tmppath;
2689         const char *format;
2690
2691         if (nvlist_lookup_string(config, ZPOOL_CONFIG_PHYS_PATH,
2692             &tmppath) != 0)
2693                 return (EZFS_NODEVICE);
2694
2695         pos = *bytes_written;
2696         bytes_left = physpath_size - pos;
2697         format = (pos == 0) ? "%s" : " %s";
2698
2699         rsz = snprintf(physpath + pos, bytes_left, format, tmppath);
2700         *bytes_written += rsz;
2701
2702         if (rsz >= bytes_left) {
2703                 /* if physpath was not copied properly, clear it */
2704                 if (bytes_left != 0) {
2705                         physpath[pos] = 0;
2706                 }
2707                 return (EZFS_NOSPC);
2708         }
2709         return (0);
2710 }
2711
2712 static int
2713 vdev_get_physpaths(nvlist_t *nv, char *physpath, size_t phypath_size,
2714     size_t *rsz, boolean_t is_spare)
2715 {
2716         char *type;
2717         int ret;
2718
2719         if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0)
2720                 return (EZFS_INVALCONFIG);
2721
2722         if (strcmp(type, VDEV_TYPE_DISK) == 0) {
2723                 /*
2724                  * An active spare device has ZPOOL_CONFIG_IS_SPARE set.
2725                  * For a spare vdev, we only want to boot from the active
2726                  * spare device.
2727                  */
2728                 if (is_spare) {
2729                         uint64_t spare = 0;
2730                         (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE,
2731                             &spare);
2732                         if (!spare)
2733                                 return (EZFS_INVALCONFIG);
2734                 }
2735
2736                 if (vdev_is_online(nv)) {
2737                         if ((ret = vdev_get_one_physpath(nv, physpath,
2738                             phypath_size, rsz)) != 0)
2739                                 return (ret);
2740                 }
2741         } else if (strcmp(type, VDEV_TYPE_MIRROR) == 0 ||
2742             strcmp(type, VDEV_TYPE_RAIDZ) == 0 ||
2743             strcmp(type, VDEV_TYPE_REPLACING) == 0 ||
2744             (is_spare = (strcmp(type, VDEV_TYPE_SPARE) == 0))) {
2745                 nvlist_t **child;
2746                 uint_t count;
2747                 int i, ret;
2748
2749                 if (nvlist_lookup_nvlist_array(nv,
2750                     ZPOOL_CONFIG_CHILDREN, &child, &count) != 0)
2751                         return (EZFS_INVALCONFIG);
2752
2753                 for (i = 0; i < count; i++) {
2754                         ret = vdev_get_physpaths(child[i], physpath,
2755                             phypath_size, rsz, is_spare);
2756                         if (ret == EZFS_NOSPC)
2757                                 return (ret);
2758                 }
2759         }
2760
2761         return (EZFS_POOL_INVALARG);
2762 }
2763
2764 /*
2765  * Get phys_path for a root pool config.
2766  * Return 0 on success; non-zero on failure.
2767  */
2768 static int
2769 zpool_get_config_physpath(nvlist_t *config, char *physpath, size_t phypath_size)
2770 {
2771         size_t rsz;
2772         nvlist_t *vdev_root;
2773         nvlist_t **child;
2774         uint_t count;
2775         char *type;
2776
2777         rsz = 0;
2778
2779         if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
2780             &vdev_root) != 0)
2781                 return (EZFS_INVALCONFIG);
2782
2783         if (nvlist_lookup_string(vdev_root, ZPOOL_CONFIG_TYPE, &type) != 0 ||
2784             nvlist_lookup_nvlist_array(vdev_root, ZPOOL_CONFIG_CHILDREN,
2785             &child, &count) != 0)
2786                 return (EZFS_INVALCONFIG);
2787
2788         /*
2789          * root pool can only have a single top-level vdev.
2790          */
2791         if (strcmp(type, VDEV_TYPE_ROOT) != 0 || count != 1)
2792                 return (EZFS_POOL_INVALARG);
2793
2794         (void) vdev_get_physpaths(child[0], physpath, phypath_size, &rsz,
2795             B_FALSE);
2796
2797         /* No online devices */
2798         if (rsz == 0)
2799                 return (EZFS_NODEVICE);
2800
2801         return (0);
2802 }
2803
2804 /*
2805  * Get phys_path for a root pool
2806  * Return 0 on success; non-zero on failure.
2807  */
2808 int
2809 zpool_get_physpath(zpool_handle_t *zhp, char *physpath, size_t phypath_size)
2810 {
2811         return (zpool_get_config_physpath(zhp->zpool_config, physpath,
2812             phypath_size));
2813 }
2814
2815 /*
2816  * Convert a vdev path to a GUID.  Returns GUID or 0 on error.
2817  *
2818  * If is_spare, is_l2cache, or is_log is non-NULL, then store within it
2819  * if the VDEV is a spare, l2cache, or log device.  If they're NULL then
2820  * ignore them.
2821  */
2822 static uint64_t
2823 zpool_vdev_path_to_guid_impl(zpool_handle_t *zhp, const char *path,
2824     boolean_t *is_spare, boolean_t *is_l2cache, boolean_t *is_log)
2825 {
2826         uint64_t guid;
2827         boolean_t spare = B_FALSE, l2cache = B_FALSE, log = B_FALSE;
2828         nvlist_t *tgt;
2829
2830         if ((tgt = zpool_find_vdev(zhp, path, &spare, &l2cache,
2831             &log)) == NULL)
2832                 return (0);
2833
2834         verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &guid) == 0);
2835         if (is_spare != NULL)
2836                 *is_spare = spare;
2837         if (is_l2cache != NULL)
2838                 *is_l2cache = l2cache;
2839         if (is_log != NULL)
2840                 *is_log = log;
2841
2842         return (guid);
2843 }
2844
2845 /* Convert a vdev path to a GUID.  Returns GUID or 0 on error. */
2846 uint64_t
2847 zpool_vdev_path_to_guid(zpool_handle_t *zhp, const char *path)
2848 {
2849         return (zpool_vdev_path_to_guid_impl(zhp, path, NULL, NULL, NULL));
2850 }
2851
2852 /*
2853  * Bring the specified vdev online.   The 'flags' parameter is a set of the
2854  * ZFS_ONLINE_* flags.
2855  */
2856 int
2857 zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags,
2858     vdev_state_t *newstate)
2859 {
2860         zfs_cmd_t zc = {"\0"};
2861         char msg[1024];
2862         char *pathname;
2863         nvlist_t *tgt;
2864         boolean_t avail_spare, l2cache, islog;
2865         libzfs_handle_t *hdl = zhp->zpool_hdl;
2866         int error;
2867
2868         if (flags & ZFS_ONLINE_EXPAND) {
2869                 (void) snprintf(msg, sizeof (msg),
2870                     dgettext(TEXT_DOMAIN, "cannot expand %s"), path);
2871         } else {
2872                 (void) snprintf(msg, sizeof (msg),
2873                     dgettext(TEXT_DOMAIN, "cannot online %s"), path);
2874         }
2875
2876         (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2877         if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2878             &islog)) == NULL)
2879                 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2880
2881         verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2882
2883         if (avail_spare)
2884                 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2885
2886         if ((flags & ZFS_ONLINE_EXPAND ||
2887             zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) &&
2888             nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH, &pathname) == 0) {
2889                 uint64_t wholedisk = 0;
2890
2891                 (void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK,
2892                     &wholedisk);
2893
2894                 /*
2895                  * XXX - L2ARC 1.0 devices can't support expansion.
2896                  */
2897                 if (l2cache) {
2898                         zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2899                             "cannot expand cache devices"));
2900                         return (zfs_error(hdl, EZFS_VDEVNOTSUP, msg));
2901                 }
2902
2903                 if (wholedisk) {
2904                         const char *fullpath = path;
2905                         char buf[MAXPATHLEN];
2906
2907                         if (path[0] != '/') {
2908                                 error = zfs_resolve_shortname(path, buf,
2909                                     sizeof (buf));
2910                                 if (error != 0)
2911                                         return (zfs_error(hdl, EZFS_NODEVICE,
2912                                             msg));
2913
2914                                 fullpath = buf;
2915                         }
2916
2917                         error = zpool_relabel_disk(hdl, fullpath, msg);
2918                         if (error != 0)
2919                                 return (error);
2920                 }
2921         }
2922
2923         zc.zc_cookie = VDEV_STATE_ONLINE;
2924         zc.zc_obj = flags;
2925
2926         if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0) {
2927                 if (errno == EINVAL) {
2928                         zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "was split "
2929                             "from this pool into a new one.  Use '%s' "
2930                             "instead"), "zpool detach");
2931                         return (zfs_error(hdl, EZFS_POSTSPLIT_ONLINE, msg));
2932                 }
2933                 return (zpool_standard_error(hdl, errno, msg));
2934         }
2935
2936         *newstate = zc.zc_cookie;
2937         return (0);
2938 }
2939
2940 /*
2941  * Take the specified vdev offline
2942  */
2943 int
2944 zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp)
2945 {
2946         zfs_cmd_t zc = {"\0"};
2947         char msg[1024];
2948         nvlist_t *tgt;
2949         boolean_t avail_spare, l2cache;
2950         libzfs_handle_t *hdl = zhp->zpool_hdl;
2951
2952         (void) snprintf(msg, sizeof (msg),
2953             dgettext(TEXT_DOMAIN, "cannot offline %s"), path);
2954
2955         (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2956         if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2957             NULL)) == NULL)
2958                 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2959
2960         verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2961
2962         if (avail_spare)
2963                 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2964
2965         zc.zc_cookie = VDEV_STATE_OFFLINE;
2966         zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0;
2967
2968         if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2969                 return (0);
2970
2971         switch (errno) {
2972         case EBUSY:
2973
2974                 /*
2975                  * There are no other replicas of this device.
2976                  */
2977                 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2978
2979         case EEXIST:
2980                 /*
2981                  * The log device has unplayed logs
2982                  */
2983                 return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, msg));
2984
2985         default:
2986                 return (zpool_standard_error(hdl, errno, msg));
2987         }
2988 }
2989
2990 /*
2991  * Mark the given vdev faulted.
2992  */
2993 int
2994 zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
2995 {
2996         zfs_cmd_t zc = {"\0"};
2997         char msg[1024];
2998         libzfs_handle_t *hdl = zhp->zpool_hdl;
2999
3000         (void) snprintf(msg, sizeof (msg),
3001             dgettext(TEXT_DOMAIN, "cannot fault %llu"), (u_longlong_t)guid);
3002
3003         (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3004         zc.zc_guid = guid;
3005         zc.zc_cookie = VDEV_STATE_FAULTED;
3006         zc.zc_obj = aux;
3007
3008         if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
3009                 return (0);
3010
3011         switch (errno) {
3012         case EBUSY:
3013
3014                 /*
3015                  * There are no other replicas of this device.
3016                  */
3017                 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
3018
3019         default:
3020                 return (zpool_standard_error(hdl, errno, msg));
3021         }
3022
3023 }
3024
3025 /*
3026  * Mark the given vdev degraded.
3027  */
3028 int
3029 zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
3030 {
3031         zfs_cmd_t zc = {"\0"};
3032         char msg[1024];
3033         libzfs_handle_t *hdl = zhp->zpool_hdl;
3034
3035         (void) snprintf(msg, sizeof (msg),
3036             dgettext(TEXT_DOMAIN, "cannot degrade %llu"), (u_longlong_t)guid);
3037
3038         (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3039         zc.zc_guid = guid;
3040         zc.zc_cookie = VDEV_STATE_DEGRADED;
3041         zc.zc_obj = aux;
3042
3043         if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
3044                 return (0);
3045
3046         return (zpool_standard_error(hdl, errno, msg));
3047 }
3048
3049 /*
3050  * Returns TRUE if the given nvlist is a vdev that was originally swapped in as
3051  * a hot spare.
3052  */
3053 static boolean_t
3054 is_replacing_spare(nvlist_t *search, nvlist_t *tgt, int which)
3055 {
3056         nvlist_t **child;
3057         uint_t c, children;
3058         char *type;
3059
3060         if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child,
3061             &children) == 0) {
3062                 verify(nvlist_lookup_string(search, ZPOOL_CONFIG_TYPE,
3063                     &type) == 0);
3064
3065                 if (strcmp(type, VDEV_TYPE_SPARE) == 0 &&
3066                     children == 2 && child[which] == tgt)
3067                         return (B_TRUE);
3068
3069                 for (c = 0; c < children; c++)
3070                         if (is_replacing_spare(child[c], tgt, which))
3071                                 return (B_TRUE);
3072         }
3073
3074         return (B_FALSE);
3075 }
3076
3077 /*
3078  * Attach new_disk (fully described by nvroot) to old_disk.
3079  * If 'replacing' is specified, the new disk will replace the old one.
3080  */
3081 int
3082 zpool_vdev_attach(zpool_handle_t *zhp,
3083     const char *old_disk, const char *new_disk, nvlist_t *nvroot, int replacing)
3084 {
3085         zfs_cmd_t zc = {"\0"};
3086         char msg[1024];
3087         int ret;
3088         nvlist_t *tgt;
3089         boolean_t avail_spare, l2cache, islog;
3090         uint64_t val;
3091         char *newname;
3092         nvlist_t **child;
3093         uint_t children;
3094         nvlist_t *config_root;
3095         libzfs_handle_t *hdl = zhp->zpool_hdl;
3096         boolean_t rootpool = zpool_is_bootable(zhp);
3097
3098         if (replacing)
3099                 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
3100                     "cannot replace %s with %s"), old_disk, new_disk);
3101         else
3102                 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
3103                     "cannot attach %s to %s"), new_disk, old_disk);
3104
3105         (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3106         if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache,
3107             &islog)) == NULL)
3108                 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3109
3110         if (avail_spare)
3111                 return (zfs_error(hdl, EZFS_ISSPARE, msg));
3112
3113         if (l2cache)
3114                 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
3115
3116         verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
3117         zc.zc_cookie = replacing;
3118
3119         if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
3120             &child, &children) != 0 || children != 1) {
3121                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3122                     "new device must be a single disk"));
3123                 return (zfs_error(hdl, EZFS_INVALCONFIG, msg));
3124         }
3125
3126         verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
3127             ZPOOL_CONFIG_VDEV_TREE, &config_root) == 0);
3128
3129         if ((newname = zpool_vdev_name(NULL, NULL, child[0], 0)) == NULL)
3130                 return (-1);
3131
3132         /*
3133          * If the target is a hot spare that has been swapped in, we can only
3134          * replace it with another hot spare.
3135          */
3136         if (replacing &&
3137             nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 &&
3138             (zpool_find_vdev(zhp, newname, &avail_spare, &l2cache,
3139             NULL) == NULL || !avail_spare) &&
3140             is_replacing_spare(config_root, tgt, 1)) {
3141                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3142                     "can only be replaced by another hot spare"));
3143                 free(newname);
3144                 return (zfs_error(hdl, EZFS_BADTARGET, msg));
3145         }
3146
3147         free(newname);
3148
3149         if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
3150                 return (-1);
3151
3152         ret = zfs_ioctl(hdl, ZFS_IOC_VDEV_ATTACH, &zc);
3153
3154         zcmd_free_nvlists(&zc);
3155
3156         if (ret == 0) {
3157                 if (rootpool) {
3158                         /*
3159                          * XXX need a better way to prevent user from
3160                          * booting up a half-baked vdev.
3161                          */
3162                         (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Make "
3163                             "sure to wait until resilver is done "
3164                             "before rebooting.\n"));
3165                 }
3166                 return (0);
3167         }
3168
3169         switch (errno) {
3170         case ENOTSUP:
3171                 /*
3172                  * Can't attach to or replace this type of vdev.
3173                  */
3174                 if (replacing) {
3175                         uint64_t version = zpool_get_prop_int(zhp,
3176                             ZPOOL_PROP_VERSION, NULL);
3177
3178                         if (islog)
3179                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3180                                     "cannot replace a log with a spare"));
3181                         else if (version >= SPA_VERSION_MULTI_REPLACE)
3182                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3183                                     "already in replacing/spare config; wait "
3184                                     "for completion or use 'zpool detach'"));
3185                         else
3186                                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3187                                     "cannot replace a replacing device"));
3188                 } else {
3189                         zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3190                             "can only attach to mirrors and top-level "
3191                             "disks"));
3192                 }
3193                 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
3194                 break;
3195
3196         case EINVAL:
3197                 /*
3198                  * The new device must be a single disk.
3199                  */
3200                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3201                     "new device must be a single disk"));
3202                 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg);
3203                 break;
3204
3205         case EBUSY:
3206                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy, "
3207                     "or device removal is in progress"),
3208                     new_disk);
3209                 (void) zfs_error(hdl, EZFS_BADDEV, msg);
3210                 break;
3211
3212         case EOVERFLOW:
3213                 /*
3214                  * The new device is too small.
3215                  */
3216                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3217                     "device is too small"));
3218                 (void) zfs_error(hdl, EZFS_BADDEV, msg);
3219                 break;
3220
3221         case EDOM:
3222                 /*
3223                  * The new device has a different optimal sector size.
3224                  */
3225                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3226                     "new device has a different optimal sector size; use the "
3227                     "option '-o ashift=N' to override the optimal size"));
3228                 (void) zfs_error(hdl, EZFS_BADDEV, msg);
3229                 break;
3230
3231         case ENAMETOOLONG:
3232                 /*
3233                  * The resulting top-level vdev spec won't fit in the label.
3234                  */
3235                 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, msg);
3236                 break;
3237
3238         default:
3239                 (void) zpool_standard_error(hdl, errno, msg);
3240         }
3241
3242         return (-1);
3243 }
3244
3245 /*
3246  * Detach the specified device.
3247  */
3248 int
3249 zpool_vdev_detach(zpool_handle_t *zhp, const char *path)
3250 {
3251         zfs_cmd_t zc = {"\0"};
3252         char msg[1024];
3253         nvlist_t *tgt;
3254         boolean_t avail_spare, l2cache;
3255         libzfs_handle_t *hdl = zhp->zpool_hdl;
3256
3257         (void) snprintf(msg, sizeof (msg),
3258             dgettext(TEXT_DOMAIN, "cannot detach %s"), path);
3259
3260         (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3261         if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
3262             NULL)) == NULL)
3263                 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3264
3265         if (avail_spare)
3266                 return (zfs_error(hdl, EZFS_ISSPARE, msg));
3267
3268         if (l2cache)
3269                 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
3270
3271         verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
3272
3273         if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0)
3274                 return (0);
3275
3276         switch (errno) {
3277
3278         case ENOTSUP:
3279                 /*
3280                  * Can't detach from this type of vdev.
3281                  */
3282                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only "
3283                     "applicable to mirror and replacing vdevs"));
3284                 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
3285                 break;
3286
3287         case EBUSY:
3288                 /*
3289                  * There are no other replicas of this device.
3290                  */
3291                 (void) zfs_error(hdl, EZFS_NOREPLICAS, msg);
3292                 break;
3293
3294         default:
3295                 (void) zpool_standard_error(hdl, errno, msg);
3296         }
3297
3298         return (-1);
3299 }
3300
3301 /*
3302  * Find a mirror vdev in the source nvlist.
3303  *
3304  * The mchild array contains a list of disks in one of the top-level mirrors
3305  * of the source pool.  The schild array contains a list of disks that the
3306  * user specified on the command line.  We loop over the mchild array to
3307  * see if any entry in the schild array matches.
3308  *
3309  * If a disk in the mchild array is found in the schild array, we return
3310  * the index of that entry.  Otherwise we return -1.
3311  */
3312 static int
3313 find_vdev_entry(zpool_handle_t *zhp, nvlist_t **mchild, uint_t mchildren,
3314     nvlist_t **schild, uint_t schildren)
3315 {
3316         uint_t mc;
3317
3318         for (mc = 0; mc < mchildren; mc++) {
3319                 uint_t sc;
3320                 char *mpath = zpool_vdev_name(zhp->zpool_hdl, zhp,
3321                     mchild[mc], 0);
3322
3323                 for (sc = 0; sc < schildren; sc++) {
3324                         char *spath = zpool_vdev_name(zhp->zpool_hdl, zhp,
3325                             schild[sc], 0);
3326                         boolean_t result = (strcmp(mpath, spath) == 0);
3327
3328                         free(spath);
3329                         if (result) {
3330                                 free(mpath);
3331                                 return (mc);
3332                         }
3333                 }
3334
3335                 free(mpath);
3336         }
3337
3338         return (-1);
3339 }
3340
3341 /*
3342  * Split a mirror pool.  If newroot points to null, then a new nvlist
3343  * is generated and it is the responsibility of the caller to free it.
3344  */
3345 int
3346 zpool_vdev_split(zpool_handle_t *zhp, char *newname, nvlist_t **newroot,
3347     nvlist_t *props, splitflags_t flags)
3348 {
3349         zfs_cmd_t zc = {"\0"};
3350         char msg[1024];
3351         nvlist_t *tree, *config, **child, **newchild, *newconfig = NULL;
3352         nvlist_t **varray = NULL, *zc_props = NULL;
3353         uint_t c, children, newchildren, lastlog = 0, vcount, found = 0;
3354         libzfs_handle_t *hdl = zhp->zpool_hdl;
3355         uint64_t vers, readonly = B_FALSE;
3356         boolean_t freelist = B_FALSE, memory_err = B_TRUE;
3357         int retval = 0;
3358
3359         (void) snprintf(msg, sizeof (msg),
3360             dgettext(TEXT_DOMAIN, "Unable to split %s"), zhp->zpool_name);
3361
3362         if (!zpool_name_valid(hdl, B_FALSE, newname))
3363                 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
3364
3365         if ((config = zpool_get_config(zhp, NULL)) == NULL) {
3366                 (void) fprintf(stderr, gettext("Internal error: unable to "
3367                     "retrieve pool configuration\n"));
3368                 return (-1);
3369         }
3370
3371         verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree)
3372             == 0);
3373         verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &vers) == 0);
3374
3375         if (props) {
3376                 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
3377                 if ((zc_props = zpool_valid_proplist(hdl, zhp->zpool_name,
3378                     props, vers, flags, msg)) == NULL)
3379                         return (-1);
3380                 (void) nvlist_lookup_uint64(zc_props,
3381                     zpool_prop_to_name(ZPOOL_PROP_READONLY), &readonly);
3382                 if (readonly) {
3383                         zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3384                             "property %s can only be set at import time"),
3385                             zpool_prop_to_name(ZPOOL_PROP_READONLY));
3386                         return (-1);
3387                 }
3388         }
3389
3390         if (nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child,
3391             &children) != 0) {
3392                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3393                     "Source pool is missing vdev tree"));
3394                 nvlist_free(zc_props);
3395                 return (-1);
3396         }
3397
3398         varray = zfs_alloc(hdl, children * sizeof (nvlist_t *));
3399         vcount = 0;
3400
3401         if (*newroot == NULL ||
3402             nvlist_lookup_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN,
3403             &newchild, &newchildren) != 0)
3404                 newchildren = 0;
3405
3406         for (c = 0; c < children; c++) {
3407                 uint64_t is_log = B_FALSE, is_hole = B_FALSE;
3408                 char *type;
3409                 nvlist_t **mchild, *vdev;
3410                 uint_t mchildren;
3411                 int entry;
3412
3413                 /*
3414                  * Unlike cache & spares, slogs are stored in the
3415                  * ZPOOL_CONFIG_CHILDREN array.  We filter them out here.
3416                  */
3417                 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
3418                     &is_log);
3419                 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
3420                     &is_hole);
3421                 if (is_log || is_hole) {
3422                         /*
3423                          * Create a hole vdev and put it in the config.
3424                          */
3425                         if (nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) != 0)
3426                                 goto out;
3427                         if (nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE,
3428                             VDEV_TYPE_HOLE) != 0)
3429                                 goto out;
3430                         if (nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_HOLE,
3431                             1) != 0)
3432                                 goto out;
3433                         if (lastlog == 0)
3434                                 lastlog = vcount;
3435                         varray[vcount++] = vdev;
3436                         continue;
3437                 }
3438                 lastlog = 0;
3439                 verify(nvlist_lookup_string(child[c], ZPOOL_CONFIG_TYPE, &type)
3440                     == 0);
3441                 if (strcmp(type, VDEV_TYPE_MIRROR) != 0) {
3442                         zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3443                             "Source pool must be composed only of mirrors\n"));
3444                         retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
3445                         goto out;
3446                 }
3447
3448                 verify(nvlist_lookup_nvlist_array(child[c],
3449                     ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0);
3450
3451                 /* find or add an entry for this top-level vdev */
3452                 if (newchildren > 0 &&
3453                     (entry = find_vdev_entry(zhp, mchild, mchildren,
3454                     newchild, newchildren)) >= 0) {
3455                         /* We found a disk that the user specified. */
3456                         vdev = mchild[entry];
3457                         ++found;
3458                 } else {
3459                         /* User didn't specify a disk for this vdev. */
3460                         vdev = mchild[mchildren - 1];
3461                 }
3462
3463                 if (nvlist_dup(vdev, &varray[vcount++], 0) != 0)
3464                         goto out;
3465         }
3466
3467         /* did we find every disk the user specified? */
3468         if (found != newchildren) {
3469                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "Device list must "
3470                     "include at most one disk from each mirror"));
3471                 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
3472                 goto out;
3473         }
3474
3475         /* Prepare the nvlist for populating. */
3476         if (*newroot == NULL) {
3477                 if (nvlist_alloc(newroot, NV_UNIQUE_NAME, 0) != 0)
3478                         goto out;
3479                 freelist = B_TRUE;
3480                 if (nvlist_add_string(*newroot, ZPOOL_CONFIG_TYPE,
3481                     VDEV_TYPE_ROOT) != 0)
3482                         goto out;
3483         } else {
3484                 verify(nvlist_remove_all(*newroot, ZPOOL_CONFIG_CHILDREN) == 0);
3485         }
3486
3487         /* Add all the children we found */
3488         if (nvlist_add_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, varray,
3489             lastlog == 0 ? vcount : lastlog) != 0)
3490                 goto out;
3491
3492         /*
3493          * If we're just doing a dry run, exit now with success.
3494          */
3495         if (flags.dryrun) {
3496                 memory_err = B_FALSE;
3497                 freelist = B_FALSE;
3498                 goto out;
3499         }
3500
3501         /* now build up the config list & call the ioctl */
3502         if (nvlist_alloc(&newconfig, NV_UNIQUE_NAME, 0) != 0)
3503                 goto out;
3504
3505         if (nvlist_add_nvlist(newconfig,
3506             ZPOOL_CONFIG_VDEV_TREE, *newroot) != 0 ||
3507             nvlist_add_string(newconfig,
3508             ZPOOL_CONFIG_POOL_NAME, newname) != 0 ||
3509             nvlist_add_uint64(newconfig, ZPOOL_CONFIG_VERSION, vers) != 0)
3510                 goto out;
3511
3512         /*
3513          * The new pool is automatically part of the namespace unless we
3514          * explicitly export it.
3515          */
3516         if (!flags.import)
3517                 zc.zc_cookie = ZPOOL_EXPORT_AFTER_SPLIT;
3518         (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3519         (void) strlcpy(zc.zc_string, newname, sizeof (zc.zc_string));
3520         if (zcmd_write_conf_nvlist(hdl, &zc, newconfig) != 0)
3521                 goto out;
3522         if (zc_props != NULL && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
3523                 goto out;
3524
3525         if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SPLIT, &zc) != 0) {
3526                 retval = zpool_standard_error(hdl, errno, msg);
3527                 goto out;
3528         }
3529
3530         freelist = B_FALSE;
3531         memory_err = B_FALSE;
3532
3533 out:
3534         if (varray != NULL) {
3535                 int v;
3536
3537                 for (v = 0; v < vcount; v++)
3538                         nvlist_free(varray[v]);
3539                 free(varray);
3540         }
3541         zcmd_free_nvlists(&zc);
3542         nvlist_free(zc_props);
3543         nvlist_free(newconfig);
3544         if (freelist) {
3545                 nvlist_free(*newroot);
3546                 *newroot = NULL;
3547         }
3548
3549         if (retval != 0)
3550                 return (retval);
3551
3552         if (memory_err)
3553                 return (no_memory(hdl));
3554
3555         return (0);
3556 }
3557
3558 /*
3559  * Remove the given device.
3560  */
3561 int
3562 zpool_vdev_remove(zpool_handle_t *zhp, const char *path)
3563 {
3564         zfs_cmd_t zc = {"\0"};
3565         char msg[1024];
3566         nvlist_t *tgt;
3567         boolean_t avail_spare, l2cache, islog;
3568         libzfs_handle_t *hdl = zhp->zpool_hdl;
3569         uint64_t version;
3570
3571         (void) snprintf(msg, sizeof (msg),
3572             dgettext(TEXT_DOMAIN, "cannot remove %s"), path);
3573
3574         (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3575         if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
3576             &islog)) == NULL)
3577                 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3578
3579         version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
3580         if (islog && version < SPA_VERSION_HOLES) {
3581                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3582                     "pool must be upgraded to support log removal"));
3583                 return (zfs_error(hdl, EZFS_BADVERSION, msg));
3584         }
3585
3586         if (!islog && !avail_spare && !l2cache && zpool_is_bootable(zhp)) {
3587                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3588                     "root pool can not have removed devices, "
3589                     "because GRUB does not understand them"));
3590                 return (zfs_error(hdl, EINVAL, msg));
3591         }
3592
3593         zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
3594
3595         if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
3596                 return (0);
3597
3598         switch (errno) {
3599
3600         case EINVAL:
3601                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3602                     "invalid config; all top-level vdevs must "
3603                     "have the same sector size and not be raidz."));
3604                 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg);
3605                 break;
3606
3607         case EBUSY:
3608                 if (islog) {
3609                         zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3610                             "Mount encrypted datasets to replay logs."));
3611                 } else {
3612                         zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3613                             "Pool busy; removal may already be in progress"));
3614                 }
3615                 (void) zfs_error(hdl, EZFS_BUSY, msg);
3616                 break;
3617
3618         case EACCES:
3619                 if (islog) {
3620                         zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3621                             "Mount encrypted datasets to replay logs."));
3622                         (void) zfs_error(hdl, EZFS_BUSY, msg);
3623                 } else {
3624                         (void) zpool_standard_error(hdl, errno, msg);
3625                 }
3626                 break;
3627
3628         default:
3629                 (void) zpool_standard_error(hdl, errno, msg);
3630         }
3631         return (-1);
3632 }
3633
3634 int
3635 zpool_vdev_remove_cancel(zpool_handle_t *zhp)
3636 {
3637         zfs_cmd_t zc;
3638         char msg[1024];
3639         libzfs_handle_t *hdl = zhp->zpool_hdl;
3640
3641         (void) snprintf(msg, sizeof (msg),
3642             dgettext(TEXT_DOMAIN, "cannot cancel removal"));
3643
3644         bzero(&zc, sizeof (zc));
3645         (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3646         zc.zc_cookie = 1;
3647
3648         if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
3649                 return (0);
3650
3651         return (zpool_standard_error(hdl, errno, msg));
3652 }
3653
3654 int
3655 zpool_vdev_indirect_size(zpool_handle_t *zhp, const char *path,
3656     uint64_t *sizep)
3657 {
3658         char msg[1024];
3659         nvlist_t *tgt;
3660         boolean_t avail_spare, l2cache, islog;
3661         libzfs_handle_t *hdl = zhp->zpool_hdl;
3662
3663         (void) snprintf(msg, sizeof (msg),
3664             dgettext(TEXT_DOMAIN, "cannot determine indirect size of %s"),
3665             path);
3666
3667         if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
3668             &islog)) == NULL)
3669                 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3670
3671         if (avail_spare || l2cache || islog) {
3672                 *sizep = 0;
3673                 return (0);
3674         }
3675
3676         if (nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_INDIRECT_SIZE, sizep) != 0) {
3677                 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3678                     "indirect size not available"));
3679                 return (zfs_error(hdl, EINVAL, msg));
3680         }
3681         return (0);
3682 }
3683
3684 /*
3685  * Clear the errors for the pool, or the particular device if specified.
3686  */
3687 int
3688 zpool_clear(zpool_handle_t *zhp, const char *path, nvlist_t *rewindnvl)
3689 {
3690         zfs_cmd_t zc = {"\0"};
3691         char msg[1024];
3692         nvlist_t *tgt;
3693         zpool_load_policy_t policy;
3694         boolean_t avail_spare, l2cache;
3695         libzfs_handle_t *hdl = zhp->zpool_hdl;
3696         nvlist_t *nvi = NULL;
3697         int error;
3698
3699         if (path)
3700                 (void) snprintf(msg, sizeof (msg),
3701                     dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
3702                     path);
3703         else
3704                 (void) snprintf(msg, sizeof (msg),
3705                     dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
3706                     zhp->zpool_name);
3707
3708         (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3709         if (path) {
3710                 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare,
3711                     &l2cache, NULL)) == NULL)
3712                         return (zfs_error(hdl, EZFS_NODEVICE, msg));
3713
3714                 /*
3715                  * Don't allow error clearing for hot spares.  Do allow
3716                  * error clearing for l2cache devices.
3717                  */
3718                 if (avail_spare)
3719                         return (zfs_error(hdl, EZFS_ISSPARE, msg));
3720
3721                 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID,
3722                     &zc.zc_guid) == 0);
3723         }
3724
3725         zpool_get_load_policy(rewindnvl, &policy);
3726         zc.zc_cookie = policy.zlp_rewind;
3727
3728         if (zcmd_alloc_dst_nvlist(hdl, &zc, zhp->zpool_config_size * 2) != 0)
3729                 return (-1);
3730
3731         if (zcmd_write_src_nvlist(hdl, &zc, rewindnvl) != 0)
3732                 return (-1);
3733
3734         while ((error = zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc)) != 0 &&
3735             errno == ENOMEM) {
3736                 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
3737                         zcmd_free_nvlists(&zc);
3738                         return (-1);
3739                 }
3740         }
3741
3742         if (!error || ((policy.zlp_rewind & ZPOOL_TRY_REWIND) &&
3743             errno != EPERM && errno != EACCES)) {
3744                 if (policy.zlp_rewind &
3745                     (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
3746                         (void) zcmd_read_dst_nvlist(hdl, &zc, &nvi);
3747                         zpool_rewind_exclaim(hdl, zc.zc_name,
3748                             ((policy.zlp_rewind & ZPOOL_TRY_REWIND) != 0),
3749                             nvi);
3750                         nvlist_free(nvi);
3751                 }
3752                 zcmd_free_nvlists(&zc);
3753                 return (0);
3754         }
3755
3756         zcmd_free_nvlists(&zc);
3757         return (zpool_standard_error(hdl, errno, msg));
3758 }
3759
3760 /*
3761  * Similar to zpool_clear(), but takes a GUID (used by fmd).
3762  */
3763 int
3764 zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid)
3765 {
3766         zfs_cmd_t zc = {"\0"};
3767         char msg[1024];
3768         libzfs_handle_t *hdl = zhp->zpool_hdl;
3769
3770         (void) snprintf(msg, sizeof (msg),
3771             dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"),
3772             (u_longlong_t)guid);
3773
3774         (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3775         zc.zc_guid = guid;
3776         zc.zc_cookie = ZPOOL_NO_REWIND;
3777
3778         if (zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc) == 0)
3779                 return (0);
3780
3781         return (zpool_standard_error(hdl, errno, msg));
3782 }
3783
3784 /*
3785  * Change the GUID for a pool.
3786  */
3787 int
3788 zpool_reguid(zpool_handle_t *zhp)
3789 {
3790         char msg[1024];
3791         libzfs_handle_t *hdl = zhp->zpool_hdl;
3792         zfs_cmd_t zc = {"\0"};
3793
3794         (void) snprintf(msg, sizeof (msg),
3795             dgettext(TEXT_DOMAIN, "cannot reguid '%s'"), zhp->zpool_name);
3796
3797         (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3798         if (zfs_ioctl(hdl, ZFS_IOC_POOL_REGUID, &zc) == 0)
3799                 return (0);
3800
3801         return (zpool_standard_error(hdl, errno, msg));
3802 }
3803
3804 /*
3805  * Reopen the pool.
3806  */
3807 int
3808 zpool_reopen_one(zpool_handle_t *zhp, void *data)
3809 {
3810         libzfs_handle_t *hdl = zpool_get_handle(zhp);
3811         const char *pool_name = zpool_get_name(zhp);
3812         boolean_t *scrub_restart = data;
3813         int error;
3814
3815         error = lzc_reopen(pool_name, *scrub_restart);
3816         if (error) {
3817                 return (zpool_standard_error_fmt(hdl, error,
3818                     dgettext(TEXT_DOMAIN, "cannot reopen '%s'"), pool_name));
3819         }
3820
3821         return (0);
3822 }
3823
3824 /* call into libzfs_core to execute the sync IOCTL per pool */
3825 int
3826 zpool_sync_one(zpool_handle_t *zhp, void *data)
3827 {
3828         int ret;
3829         libzfs_handle_t *hdl = zpool_get_handle(zhp);
3830         const char *pool_name = zpool_get_name(zhp);
3831         boolean_t *force = data;
3832         nvlist_t *innvl = fnvlist_alloc();
3833
3834         fnvlist_add_boolean_value(innvl, "force", *force);
3835         if ((ret = lzc_sync(pool_name, innvl, NULL)) != 0) {
3836                 nvlist_free(innvl);
3837                 return (zpool_standard_error_fmt(hdl, ret,
3838                     dgettext(TEXT_DOMAIN, "sync '%s' failed"), pool_name));
3839         }
3840         nvlist_free(innvl);
3841
3842         return (0);
3843 }
3844
3845 #define PATH_BUF_LEN    64
3846
3847 /*
3848  * Given a vdev, return the name to display in iostat.  If the vdev has a path,
3849  * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
3850  * We also check if this is a whole disk, in which case we strip off the
3851  * trailing 's0' slice name.
3852  *
3853  * This routine is also responsible for identifying when disks have been
3854  * reconfigured in a new location.  The kernel will have opened the device by
3855  * devid, but the path will still refer to the old location.  To catch this, we
3856  * first do a path -> devid translation (which is fast for the common case).  If
3857  * the devid matches, we're done.  If not, we do a reverse devid -> path
3858  * translation and issue the appropriate ioctl() to update the path of the vdev.
3859  * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
3860  * of these checks.
3861  */
3862 char *
3863 zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv,
3864     int name_flags)
3865 {
3866         char *path, *type, *env;
3867         uint64_t value;
3868         char buf[PATH_BUF_LEN];
3869         char tmpbuf[PATH_BUF_LEN];
3870
3871         /*
3872          * vdev_name will be "root"/"root-0" for the root vdev, but it is the
3873          * zpool name that will be displayed to the user.
3874          */
3875         verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0);
3876         if (zhp != NULL && strcmp(type, "root") == 0)
3877                 return (zfs_strdup(hdl, zpool_get_name(zhp)));
3878
3879         env = getenv("ZPOOL_VDEV_NAME_PATH");
3880         if (env && (strtoul(env, NULL, 0) > 0 ||
3881             !strncasecmp(env, "YES", 3) || !strncasecmp(env, "ON", 2)))
3882                 name_flags |= VDEV_NAME_PATH;
3883
3884         env = getenv("ZPOOL_VDEV_NAME_GUID");
3885         if (env && (strtoul(env, NULL, 0) > 0 ||
3886             !strncasecmp(env, "YES", 3) || !strncasecmp(env, "ON", 2)))
3887                 name_flags |= VDEV_NAME_GUID;
3888
3889         env = getenv("ZPOOL_VDEV_NAME_FOLLOW_LINKS");
3890         if (env && (strtoul(env, NULL, 0) > 0 ||
3891             !strncasecmp(env, "YES", 3) || !strncasecmp(env, "ON", 2)))
3892                 name_flags |= VDEV_NAME_FOLLOW_LINKS;
3893
3894         if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, &value) == 0 ||
3895             name_flags & VDEV_NAME_GUID) {
3896                 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value);
3897                 (void) snprintf(buf, sizeof (buf), "%llu", (u_longlong_t)value);
3898                 path = buf;
3899         } else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) {
3900                 if (name_flags & VDEV_NAME_FOLLOW_LINKS) {
3901                         char *rp = realpath(path, NULL);
3902                         if (rp) {
3903                                 strlcpy(buf, rp, sizeof (buf));
3904                                 path = buf;
3905                                 free(rp);
3906                         }
3907                 }
3908
3909                 /*
3910                  * For a block device only use the name.
3911                  */
3912                 if ((strcmp(type, VDEV_TYPE_DISK) == 0) &&
3913                     !(name_flags & VDEV_NAME_PATH)) {
3914                         path = strrchr(path, '/');
3915                         path++;
3916                 }
3917
3918                 /*
3919                  * Remove the partition from the path it this is a whole disk.
3920                  */
3921                 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, &value)
3922                     == 0 && value && !(name_flags & VDEV_NAME_PATH)) {
3923                         return (zfs_strip_partition(path));
3924                 }
3925         } else {
3926                 path = type;
3927
3928                 /*
3929                  * If it's a raidz device, we need to stick in the parity level.
3930                  */
3931                 if (strcmp(path, VDEV_TYPE_RAIDZ) == 0) {
3932                         verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY,
3933                             &value) == 0);
3934                         (void) snprintf(buf, sizeof (buf), "%s%llu", path,
3935                             (u_longlong_t)value);
3936                         path = buf;
3937                 }
3938
3939                 /*
3940                  * We identify each top-level vdev by using a <type-id>
3941                  * naming convention.
3942                  */
3943                 if (name_flags & VDEV_NAME_TYPE_ID) {
3944                         uint64_t id;
3945                         verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
3946                             &id) == 0);
3947                         (void) snprintf(tmpbuf, sizeof (tmpbuf), "%s-%llu",
3948                             path, (u_longlong_t)id);
3949                         path = tmpbuf;
3950                 }
3951         }
3952
3953         return (zfs_strdup(hdl, path));
3954 }
3955
3956 static int
3957 zbookmark_mem_compare(const void *a, const void *b)
3958 {
3959         return (memcmp(a, b, sizeof (zbookmark_phys_t)));
3960 }
3961
3962 /*
3963  * Retrieve the persistent error log, uniquify the members, and return to the
3964  * caller.
3965  */
3966 int
3967 zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp)
3968 {
3969         zfs_cmd_t zc = {"\0"};
3970         libzfs_handle_t *hdl = zhp->zpool_hdl;
3971         uint64_t count;
3972         zbookmark_phys_t *zb = NULL;
3973         int i;
3974
3975         /*
3976          * Retrieve the raw error list from the kernel.  If the number of errors
3977          * has increased, allocate more space and continue until we get the
3978          * entire list.
3979          */
3980         verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_ERRCOUNT,
3981             &count) == 0);
3982         if (count == 0)
3983                 return (0);
3984         zc.zc_nvlist_dst = (uintptr_t)zfs_alloc(zhp->zpool_hdl,
3985             count * sizeof (zbookmark_phys_t));
3986         zc.zc_nvlist_dst_size = count;
3987         (void) strcpy(zc.zc_name, zhp->zpool_name);
3988         for (;;) {
3989                 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_ERROR_LOG,
3990                     &zc) != 0) {
3991                         free((void *)(uintptr_t)zc.zc_nvlist_dst);
3992                         if (errno == ENOMEM) {
3993                                 void *dst;
3994
3995                                 count = zc.zc_nvlist_dst_size;
3996                                 dst = zfs_alloc(zhp->zpool_hdl, count *
3997                                     sizeof (zbookmark_phys_t));
3998                                 zc.zc_nvlist_dst = (uintptr_t)dst;
3999                         } else {
4000                                 return (zpool_standard_error_fmt(hdl, errno,
4001                                     dgettext(TEXT_DOMAIN, "errors: List of "
4002                                     "errors unavailable")));
4003                         }
4004                 } else {
4005                         break;
4006                 }
4007         }
4008
4009         /*
4010          * Sort the resulting bookmarks.  This is a little confusing due to the
4011          * implementation of ZFS_IOC_ERROR_LOG.  The bookmarks are copied last
4012          * to first, and 'zc_nvlist_dst_size' indicates the number of bookmarks
4013          * _not_ copied as part of the process.  So we point the start of our
4014          * array appropriate and decrement the total number of elements.
4015          */
4016         zb = ((zbookmark_phys_t *)(uintptr_t)zc.zc_nvlist_dst) +
4017             zc.zc_nvlist_dst_size;
4018         count -= zc.zc_nvlist_dst_size;
4019
4020         qsort(zb, count, sizeof (zbookmark_phys_t), zbookmark_mem_compare);
4021
4022         verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0);
4023
4024         /*
4025          * Fill in the nverrlistp with nvlist's of dataset and object numbers.
4026          */
4027         for (i = 0; i < count; i++) {
4028                 nvlist_t *nv;
4029
4030                 /* ignoring zb_blkid and zb_level for now */
4031                 if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset &&
4032                     zb[i-1].zb_object == zb[i].zb_object)
4033                         continue;
4034
4035                 if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0)
4036                         goto nomem;
4037                 if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET,
4038                     zb[i].zb_objset) != 0) {
4039                         nvlist_free(nv);
4040                         goto nomem;
4041                 }
4042                 if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT,
4043                     zb[i].zb_object) != 0) {
4044                         nvlist_free(nv);
4045                         goto nomem;
4046                 }
4047                 if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) {
4048                         nvlist_free(nv);
4049                         goto nomem;
4050                 }
4051                 nvlist_free(nv);
4052         }
4053
4054         free((void *)(uintptr_t)zc.zc_nvlist_dst);
4055         return (0);
4056
4057 nomem:
4058         free((void *)(uintptr_t)zc.zc_nvlist_dst);
4059         return (no_memory(zhp->zpool_hdl));
4060 }
4061
4062 /*
4063  * Upgrade a ZFS pool to the latest on-disk version.
4064  */
4065 int
4066 zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version)
4067 {
4068         zfs_cmd_t zc = {"\0"};
4069         libzfs_handle_t *hdl = zhp->zpool_hdl;
4070
4071         (void) strcpy(zc.zc_name, zhp->zpool_name);
4072         zc.zc_cookie = new_version;
4073
4074         if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0)
4075                 return (zpool_standard_error_fmt(hdl, errno,
4076                     dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"),
4077                     zhp->zpool_name));
4078         return (0);
4079 }
4080
4081 void
4082 zfs_save_arguments(int argc, char **argv, char *string, int len)
4083 {
4084         int i;
4085
4086         (void) strlcpy(string, basename(argv[0]), len);
4087         for (i = 1; i < argc; i++) {
4088                 (void) strlcat(string, " ", len);
4089                 (void) strlcat(string, argv[i], len);
4090         }
4091 }
4092
4093 int
4094 zpool_log_history(libzfs_handle_t *hdl, const char *message)
4095 {
4096         zfs_cmd_t zc = {"\0"};
4097         nvlist_t *args;
4098         int err;
4099
4100         args = fnvlist_alloc();
4101         fnvlist_add_string(args, "message", message);
4102         err = zcmd_write_src_nvlist(hdl, &zc, args);
4103         if (err == 0)
4104                 err = zfs_ioctl(hdl, ZFS_IOC_LOG_HISTORY, &zc);
4105         nvlist_free(args);
4106         zcmd_free_nvlists(&zc);
4107         return (err);
4108 }
4109
4110 /*
4111  * Perform ioctl to get some command history of a pool.
4112  *
4113  * 'buf' is the buffer to fill up to 'len' bytes.  'off' is the
4114  * logical offset of the history buffer to start reading from.
4115  *
4116  * Upon return, 'off' is the next logical offset to read from and
4117  * 'len' is the actual amount of bytes read into 'buf'.
4118  */
4119 static int
4120 get_history(zpool_handle_t *zhp, char *buf, uint64_t *off, uint64_t *len)
4121 {
4122         zfs_cmd_t zc = {"\0"};
4123         libzfs_handle_t *hdl = zhp->zpool_hdl;
4124
4125         (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4126
4127         zc.zc_history = (uint64_t)(uintptr_t)buf;
4128         zc.zc_history_len = *len;
4129         zc.zc_history_offset = *off;
4130
4131         if (zfs_ioctl(hdl, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) {
4132                 switch (errno) {
4133                 case EPERM:
4134                         return (zfs_error_fmt(hdl, EZFS_PERM,
4135                             dgettext(TEXT_DOMAIN,
4136                             "cannot show history for pool '%s'"),
4137                             zhp->zpool_name));
4138                 case ENOENT:
4139                         return (zfs_error_fmt(hdl, EZFS_NOHISTORY,
4140                             dgettext(TEXT_DOMAIN, "cannot get history for pool "
4141                             "'%s'"), zhp->zpool_name));
4142                 case ENOTSUP:
4143                         return (zfs_error_fmt(hdl, EZFS_BADVERSION,
4144                             dgettext(TEXT_DOMAIN, "cannot get history for pool "
4145                             "'%s', pool must be upgraded"), zhp->zpool_name));
4146                 default:
4147                         return (zpool_standard_error_fmt(hdl, errno,
4148                             dgettext(TEXT_DOMAIN,
4149                             "cannot get history for '%s'"), zhp->zpool_name));
4150                 }
4151         }
4152
4153         *len = zc.zc_history_len;
4154         *off = zc.zc_history_offset;
4155
4156         return (0);
4157 }
4158
4159 /*
4160  * Retrieve the command history of a pool.
4161  */
4162 int
4163 zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp, uint64_t *off,
4164     boolean_t *eof)
4165 {
4166         char *buf;
4167         int buflen = 128 * 1024;
4168         nvlist_t **records = NULL;
4169         uint_t numrecords = 0;
4170         int err, i;
4171         uint64_t start = *off;
4172
4173         buf = malloc(buflen);
4174         if (buf == NULL)
4175                 return (ENOMEM);
4176         /* process about 1MB a time */
4177         while (*off - start < 1024 * 1024) {
4178                 uint64_t bytes_read = buflen;
4179                 uint64_t leftover;
4180
4181                 if ((err = get_history(zhp, buf, off, &bytes_read)) != 0)
4182                         break;
4183
4184                 /* if nothing else was read in, we're at EOF, just return */
4185                 if (!bytes_read) {
4186                         *eof = B_TRUE;
4187                         break;
4188                 }
4189
4190                 if ((err = zpool_history_unpack(buf, bytes_read,
4191                     &leftover, &records, &numrecords)) != 0)
4192                         break;
4193                 *off -= leftover;
4194                 if (leftover == bytes_read) {
4195                         /*
4196                          * no progress made, because buffer is not big enough
4197                          * to hold this record; resize and retry.
4198                          */
4199                         buflen *= 2;
4200                         free(buf);
4201                         buf = malloc(buflen);
4202                         if (buf == NULL)
4203                                 return (ENOMEM);
4204                 }
4205         }
4206
4207         free(buf);
4208
4209         if (!err) {
4210                 verify(nvlist_alloc(nvhisp, NV_UNIQUE_NAME, 0) == 0);
4211                 verify(nvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD,
4212                     records, numrecords) == 0);
4213         }
4214         for (i = 0; i < numrecords; i++)
4215                 nvlist_free(records[i]);
4216         free(records);
4217
4218         return (err);
4219 }
4220
4221 /*
4222  * Retrieve the next event given the passed 'zevent_fd' file descriptor.
4223  * If there is a new event available 'nvp' will contain a newly allocated
4224  * nvlist and 'dropped' will be set to the number of missed events since
4225  * the last call to this function.  When 'nvp' is set to NULL it indicates
4226  * no new events are available.  In either case the function returns 0 and
4227  * it is up to the caller to free 'nvp'.  In the case of a fatal error the
4228  * function will return a non-zero value.  When the function is called in
4229  * blocking mode (the default, unless the ZEVENT_NONBLOCK flag is passed),
4230  * it will not return until a new event is available.
4231  */
4232 int
4233 zpool_events_next(libzfs_handle_t *hdl, nvlist_t **nvp,
4234     int *dropped, unsigned flags, int zevent_fd)
4235 {
4236         zfs_cmd_t zc = {"\0"};
4237         int error = 0;
4238
4239         *nvp = NULL;
4240         *dropped = 0;
4241         zc.zc_cleanup_fd = zevent_fd;
4242
4243         if (flags & ZEVENT_NONBLOCK)
4244                 zc.zc_guid = ZEVENT_NONBLOCK;
4245
4246         if (zcmd_alloc_dst_nvlist(hdl, &zc, ZEVENT_SIZE) != 0)
4247                 return (-1);
4248
4249 retry:
4250         if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_NEXT, &zc) != 0) {
4251                 switch (errno) {
4252                 case ESHUTDOWN:
4253                         error = zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
4254                             dgettext(TEXT_DOMAIN, "zfs shutdown"));
4255                         goto out;
4256                 case ENOENT:
4257                         /* Blocking error case should not occur */
4258                         if (!(flags & ZEVENT_NONBLOCK))
4259                                 error = zpool_standard_error_fmt(hdl, errno,
4260                                     dgettext(TEXT_DOMAIN, "cannot get event"));
4261
4262                         goto out;
4263                 case ENOMEM:
4264                         if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
4265                                 error = zfs_error_fmt(hdl, EZFS_NOMEM,
4266                                     dgettext(TEXT_DOMAIN, "cannot get event"));
4267                                 goto out;
4268                         } else {
4269                                 goto retry;
4270                         }
4271                 default:
4272                         error = zpool_standard_error_fmt(hdl, errno,
4273                             dgettext(TEXT_DOMAIN, "cannot get event"));
4274                         goto out;
4275                 }
4276         }
4277
4278         error = zcmd_read_dst_nvlist(hdl, &zc, nvp);
4279         if (error != 0)
4280                 goto out;
4281
4282         *dropped = (int)zc.zc_cookie;
4283 out:
4284         zcmd_free_nvlists(&zc);
4285
4286         return (error);
4287 }
4288
4289 /*
4290  * Clear all events.
4291  */
4292 int
4293 zpool_events_clear(libzfs_handle_t *hdl, int *count)
4294 {
4295         zfs_cmd_t zc = {"\0"};
4296         char msg[1024];
4297
4298         (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
4299             "cannot clear events"));
4300
4301         if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_CLEAR, &zc) != 0)
4302                 return (zpool_standard_error_fmt(hdl, errno, msg));
4303
4304         if (count != NULL)
4305                 *count = (int)zc.zc_cookie; /* # of events cleared */
4306
4307         return (0);
4308 }
4309
4310 /*
4311  * Seek to a specific EID, ZEVENT_SEEK_START, or ZEVENT_SEEK_END for
4312  * the passed zevent_fd file handle.  On success zero is returned,
4313  * otherwise -1 is returned and hdl->libzfs_error is set to the errno.
4314  */
4315 int
4316 zpool_events_seek(libzfs_handle_t *hdl, uint64_t eid, int zevent_fd)
4317 {
4318         zfs_cmd_t zc = {"\0"};
4319         int error = 0;
4320
4321         zc.zc_guid = eid;
4322         zc.zc_cleanup_fd = zevent_fd;
4323
4324         if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_SEEK, &zc) != 0) {
4325                 switch (errno) {
4326                 case ENOENT:
4327                         error = zfs_error_fmt(hdl, EZFS_NOENT,
4328                             dgettext(TEXT_DOMAIN, "cannot get event"));
4329                         break;
4330
4331                 case ENOMEM:
4332                         error = zfs_error_fmt(hdl, EZFS_NOMEM,
4333                             dgettext(TEXT_DOMAIN, "cannot get event"));
4334                         break;
4335
4336                 default:
4337                         error = zpool_standard_error_fmt(hdl, errno,
4338                             dgettext(TEXT_DOMAIN, "cannot get event"));
4339                         break;
4340                 }
4341         }
4342
4343         return (error);
4344 }
4345
4346 void
4347 zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj,
4348     char *pathname, size_t len)
4349 {
4350         zfs_cmd_t zc = {"\0"};
4351         boolean_t mounted = B_FALSE;
4352         char *mntpnt = NULL;
4353         char dsname[ZFS_MAX_DATASET_NAME_LEN];
4354
4355         if (dsobj == 0) {
4356                 /* special case for the MOS */
4357                 (void) snprintf(pathname, len, "<metadata>:<0x%llx>",
4358                     (longlong_t)obj);
4359                 return;
4360         }
4361
4362         /* get the dataset's name */
4363         (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4364         zc.zc_obj = dsobj;
4365         if (zfs_ioctl(zhp->zpool_hdl,
4366             ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) {
4367                 /* just write out a path of two object numbers */
4368                 (void) snprintf(pathname, len, "<0x%llx>:<0x%llx>",
4369                     (longlong_t)dsobj, (longlong_t)obj);
4370                 return;
4371         }
4372         (void) strlcpy(dsname, zc.zc_value, sizeof (dsname));
4373
4374         /* find out if the dataset is mounted */
4375         mounted = is_mounted(zhp->zpool_hdl, dsname, &mntpnt);
4376
4377         /* get the corrupted object's path */
4378         (void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name));
4379         zc.zc_obj = obj;
4380         if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_OBJ_TO_PATH,
4381             &zc) == 0) {
4382                 if (mounted) {
4383                         (void) snprintf(pathname, len, "%s%s", mntpnt,
4384                             zc.zc_value);
4385                 } else {
4386                         (void) snprintf(pathname, len, "%s:%s",
4387                             dsname, zc.zc_value);
4388                 }
4389         } else {
4390                 (void) snprintf(pathname, len, "%s:<0x%llx>", dsname,
4391                     (longlong_t)obj);
4392         }
4393         free(mntpnt);
4394 }
4395
4396 /*
4397  * Wait while the specified activity is in progress in the pool.
4398  */
4399 int
4400 zpool_wait(zpool_handle_t *zhp, zpool_wait_activity_t activity)
4401 {
4402         boolean_t missing;
4403
4404         int error = zpool_wait_status(zhp, activity, &missing, NULL);
4405
4406         if (missing) {
4407                 (void) zpool_standard_error_fmt(zhp->zpool_hdl, ENOENT,
4408                     dgettext(TEXT_DOMAIN, "error waiting in pool '%s'"),
4409                     zhp->zpool_name);
4410                 return (ENOENT);
4411         } else {
4412                 return (error);
4413         }
4414 }
4415
4416 /*
4417  * Wait for the given activity and return the status of the wait (whether or not
4418  * any waiting was done) in the 'waited' parameter. Non-existent pools are
4419  * reported via the 'missing' parameter, rather than by printing an error
4420  * message. This is convenient when this function is called in a loop over a
4421  * long period of time (as it is, for example, by zpool's wait cmd). In that
4422  * scenario, a pool being exported or destroyed should be considered a normal
4423  * event, so we don't want to print an error when we find that the pool doesn't
4424  * exist.
4425  */
4426 int
4427 zpool_wait_status(zpool_handle_t *zhp, zpool_wait_activity_t activity,
4428     boolean_t *missing, boolean_t *waited)
4429 {
4430         int error = lzc_wait(zhp->zpool_name, activity, waited);
4431         *missing = (error == ENOENT);
4432         if (*missing)
4433                 return (0);
4434
4435         if (error != 0) {
4436                 (void) zpool_standard_error_fmt(zhp->zpool_hdl, error,
4437                     dgettext(TEXT_DOMAIN, "error waiting in pool '%s'"),
4438                     zhp->zpool_name);
4439         }
4440
4441         return (error);
4442 }