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