1 /*
   2  * CDDL HEADER START
   3  *
   4  * The contents of this file are subject to the terms of the
   5  * Common Development and Distribution License (the "License").
   6  * You may not use this file except in compliance with the License.
   7  *
   8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
   9  * or http://www.opensolaris.org/os/licensing.
  10  * See the License for the specific language governing permissions
  11  * and limitations under the License.
  12  *
  13  * When distributing Covered Code, include this CDDL HEADER in each
  14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
  15  * If applicable, add the following below this CDDL HEADER, with the
  16  * fields enclosed by brackets "[]" replaced with your own identifying
  17  * information: Portions Copyright [yyyy] [name of copyright owner]
  18  *
  19  * CDDL HEADER END
  20  */
  21 
  22 /*
  23  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
  24  * Portions Copyright 2011 Martin Matuska
  25  * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
  26  * Copyright (c) 2012, Joyent, Inc. All rights reserved.
  27  * Copyright (c) 2013 by Delphix. All rights reserved.
  28  * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
  29  */
  30 
  31 /*
  32  * ZFS ioctls.
  33  *
  34  * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
  35  * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
  36  *
  37  * There are two ways that we handle ioctls: the legacy way where almost
  38  * all of the logic is in the ioctl callback, and the new way where most
  39  * of the marshalling is handled in the common entry point, zfsdev_ioctl().
  40  *
  41  * Non-legacy ioctls should be registered by calling
  42  * zfs_ioctl_register() from zfs_ioctl_init().  The ioctl is invoked
  43  * from userland by lzc_ioctl().
  44  *
  45  * The registration arguments are as follows:
  46  *
  47  * const char *name
  48  *   The name of the ioctl.  This is used for history logging.  If the
  49  *   ioctl returns successfully (the callback returns 0), and allow_log
  50  *   is true, then a history log entry will be recorded with the input &
  51  *   output nvlists.  The log entry can be printed with "zpool history -i".
  52  *
  53  * zfs_ioc_t ioc
  54  *   The ioctl request number, which userland will pass to ioctl(2).
  55  *   The ioctl numbers can change from release to release, because
  56  *   the caller (libzfs) must be matched to the kernel.
  57  *
  58  * zfs_secpolicy_func_t *secpolicy
  59  *   This function will be called before the zfs_ioc_func_t, to
  60  *   determine if this operation is permitted.  It should return EPERM
  61  *   on failure, and 0 on success.  Checks include determining if the
  62  *   dataset is visible in this zone, and if the user has either all
  63  *   zfs privileges in the zone (SYS_MOUNT), or has been granted permission
  64  *   to do this operation on this dataset with "zfs allow".
  65  *
  66  * zfs_ioc_namecheck_t namecheck
  67  *   This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
  68  *   name, a dataset name, or nothing.  If the name is not well-formed,
  69  *   the ioctl will fail and the callback will not be called.
  70  *   Therefore, the callback can assume that the name is well-formed
  71  *   (e.g. is null-terminated, doesn't have more than one '@' character,
  72  *   doesn't have invalid characters).
  73  *
  74  * zfs_ioc_poolcheck_t pool_check
  75  *   This specifies requirements on the pool state.  If the pool does
  76  *   not meet them (is suspended or is readonly), the ioctl will fail
  77  *   and the callback will not be called.  If any checks are specified
  78  *   (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
  79  *   Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
  80  *   POOL_CHECK_READONLY).
  81  *
  82  * boolean_t smush_outnvlist
  83  *   If smush_outnvlist is true, then the output is presumed to be a
  84  *   list of errors, and it will be "smushed" down to fit into the
  85  *   caller's buffer, by removing some entries and replacing them with a
  86  *   single "N_MORE_ERRORS" entry indicating how many were removed.  See
  87  *   nvlist_smush() for details.  If smush_outnvlist is false, and the
  88  *   outnvlist does not fit into the userland-provided buffer, then the
  89  *   ioctl will fail with ENOMEM.
  90  *
  91  * zfs_ioc_func_t *func
  92  *   The callback function that will perform the operation.
  93  *
  94  *   The callback should return 0 on success, or an error number on
  95  *   failure.  If the function fails, the userland ioctl will return -1,
  96  *   and errno will be set to the callback's return value.  The callback
  97  *   will be called with the following arguments:
  98  *
  99  *   const char *name
 100  *     The name of the pool or dataset to operate on, from
 101  *     zfs_cmd_t:zc_name.  The 'namecheck' argument specifies the
 102  *     expected type (pool, dataset, or none).
 103  *
 104  *   nvlist_t *innvl
 105  *     The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src.  Or
 106  *     NULL if no input nvlist was provided.  Changes to this nvlist are
 107  *     ignored.  If the input nvlist could not be deserialized, the
 108  *     ioctl will fail and the callback will not be called.
 109  *
 110  *   nvlist_t *outnvl
 111  *     The output nvlist, initially empty.  The callback can fill it in,
 112  *     and it will be returned to userland by serializing it into
 113  *     zfs_cmd_t:zc_nvlist_dst.  If it is non-empty, and serialization
 114  *     fails (e.g. because the caller didn't supply a large enough
 115  *     buffer), then the overall ioctl will fail.  See the
 116  *     'smush_nvlist' argument above for additional behaviors.
 117  *
 118  *     There are two typical uses of the output nvlist:
 119  *       - To return state, e.g. property values.  In this case,
 120  *         smush_outnvlist should be false.  If the buffer was not large
 121  *         enough, the caller will reallocate a larger buffer and try
 122  *         the ioctl again.
 123  *
 124  *       - To return multiple errors from an ioctl which makes on-disk
 125  *         changes.  In this case, smush_outnvlist should be true.
 126  *         Ioctls which make on-disk modifications should generally not
 127  *         use the outnvl if they succeed, because the caller can not
 128  *         distinguish between the operation failing, and
 129  *         deserialization failing.
 130  */
 131 
 132 #include <sys/types.h>
 133 #include <sys/param.h>
 134 #include <sys/errno.h>
 135 #include <sys/uio.h>
 136 #include <sys/buf.h>
 137 #include <sys/modctl.h>
 138 #include <sys/open.h>
 139 #include <sys/file.h>
 140 #include <sys/kmem.h>
 141 #include <sys/conf.h>
 142 #include <sys/cmn_err.h>
 143 #include <sys/stat.h>
 144 #include <sys/zfs_ioctl.h>
 145 #include <sys/zfs_vfsops.h>
 146 #include <sys/zfs_znode.h>
 147 #include <sys/zap.h>
 148 #include <sys/spa.h>
 149 #include <sys/spa_impl.h>
 150 #include <sys/vdev.h>
 151 #include <sys/priv_impl.h>
 152 #include <sys/dmu.h>
 153 #include <sys/dsl_dir.h>
 154 #include <sys/dsl_dataset.h>
 155 #include <sys/dsl_prop.h>
 156 #include <sys/dsl_deleg.h>
 157 #include <sys/dmu_objset.h>
 158 #include <sys/dmu_impl.h>
 159 #include <sys/dmu_tx.h>
 160 #include <sys/ddi.h>
 161 #include <sys/sunddi.h>
 162 #include <sys/sunldi.h>
 163 #include <sys/policy.h>
 164 #include <sys/zone.h>
 165 #include <sys/nvpair.h>
 166 #include <sys/pathname.h>
 167 #include <sys/mount.h>
 168 #include <sys/sdt.h>
 169 #include <sys/fs/zfs.h>
 170 #include <sys/zfs_ctldir.h>
 171 #include <sys/zfs_dir.h>
 172 #include <sys/zfs_onexit.h>
 173 #include <sys/zvol.h>
 174 #include <sys/dsl_scan.h>
 175 #include <sharefs/share.h>
 176 #include <sys/dmu_objset.h>
 177 #include <sys/dmu_send.h>
 178 #include <sys/dsl_destroy.h>
 179 #include <sys/dsl_userhold.h>
 180 #include <sys/zfeature.h>
 181 
 182 #include "zfs_namecheck.h"
 183 #include "zfs_prop.h"
 184 #include "zfs_deleg.h"
 185 #include "zfs_comutil.h"
 186 
 187 extern struct modlfs zfs_modlfs;
 188 
 189 extern void zfs_init(void);
 190 extern void zfs_fini(void);
 191 
 192 ldi_ident_t zfs_li = NULL;
 193 dev_info_t *zfs_dip;
 194 
 195 uint_t zfs_fsyncer_key;
 196 extern uint_t rrw_tsd_key;
 197 static uint_t zfs_allow_log_key;
 198 
 199 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
 200 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
 201 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
 202 
 203 typedef enum {
 204         NO_NAME,
 205         POOL_NAME,
 206         DATASET_NAME
 207 } zfs_ioc_namecheck_t;
 208 
 209 typedef enum {
 210         POOL_CHECK_NONE         = 1 << 0,
 211         POOL_CHECK_SUSPENDED    = 1 << 1,
 212         POOL_CHECK_READONLY     = 1 << 2,
 213 } zfs_ioc_poolcheck_t;
 214 
 215 typedef struct zfs_ioc_vec {
 216         zfs_ioc_legacy_func_t   *zvec_legacy_func;
 217         zfs_ioc_func_t          *zvec_func;
 218         zfs_secpolicy_func_t    *zvec_secpolicy;
 219         zfs_ioc_namecheck_t     zvec_namecheck;
 220         boolean_t               zvec_allow_log;
 221         zfs_ioc_poolcheck_t     zvec_pool_check;
 222         boolean_t               zvec_smush_outnvlist;
 223         const char              *zvec_name;
 224 } zfs_ioc_vec_t;
 225 
 226 /* This array is indexed by zfs_userquota_prop_t */
 227 static const char *userquota_perms[] = {
 228         ZFS_DELEG_PERM_USERUSED,
 229         ZFS_DELEG_PERM_USERQUOTA,
 230         ZFS_DELEG_PERM_GROUPUSED,
 231         ZFS_DELEG_PERM_GROUPQUOTA,
 232 };
 233 
 234 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
 235 static int zfs_check_settable(const char *name, nvpair_t *property,
 236     cred_t *cr);
 237 static int zfs_check_clearable(char *dataset, nvlist_t *props,
 238     nvlist_t **errors);
 239 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
 240     boolean_t *);
 241 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
 242 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
 243 
 244 static int zfs_prop_activate_feature(spa_t *spa, zfeature_info_t *feature);
 245 
 246 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
 247 void
 248 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
 249 {
 250         const char *newfile;
 251         char buf[512];
 252         va_list adx;
 253 
 254         /*
 255          * Get rid of annoying "../common/" prefix to filename.
 256          */
 257         newfile = strrchr(file, '/');
 258         if (newfile != NULL) {
 259                 newfile = newfile + 1; /* Get rid of leading / */
 260         } else {
 261                 newfile = file;
 262         }
 263 
 264         va_start(adx, fmt);
 265         (void) vsnprintf(buf, sizeof (buf), fmt, adx);
 266         va_end(adx);
 267 
 268         /*
 269          * To get this data, use the zfs-dprintf probe as so:
 270          * dtrace -q -n 'zfs-dprintf \
 271          *      /stringof(arg0) == "dbuf.c"/ \
 272          *      {printf("%s: %s", stringof(arg1), stringof(arg3))}'
 273          * arg0 = file name
 274          * arg1 = function name
 275          * arg2 = line number
 276          * arg3 = message
 277          */
 278         DTRACE_PROBE4(zfs__dprintf,
 279             char *, newfile, char *, func, int, line, char *, buf);
 280 }
 281 
 282 static void
 283 history_str_free(char *buf)
 284 {
 285         kmem_free(buf, HIS_MAX_RECORD_LEN);
 286 }
 287 
 288 static char *
 289 history_str_get(zfs_cmd_t *zc)
 290 {
 291         char *buf;
 292 
 293         if (zc->zc_history == NULL)
 294                 return (NULL);
 295 
 296         buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
 297         if (copyinstr((void *)(uintptr_t)zc->zc_history,
 298             buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
 299                 history_str_free(buf);
 300                 return (NULL);
 301         }
 302 
 303         buf[HIS_MAX_RECORD_LEN -1] = '\0';
 304 
 305         return (buf);
 306 }
 307 
 308 /*
 309  * Check to see if the named dataset is currently defined as bootable
 310  */
 311 static boolean_t
 312 zfs_is_bootfs(const char *name)
 313 {
 314         objset_t *os;
 315 
 316         if (dmu_objset_hold(name, FTAG, &os) == 0) {
 317                 boolean_t ret;
 318                 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
 319                 dmu_objset_rele(os, FTAG);
 320                 return (ret);
 321         }
 322         return (B_FALSE);
 323 }
 324 
 325 /*
 326  * zfs_earlier_version
 327  *
 328  *      Return non-zero if the spa version is less than requested version.
 329  */
 330 static int
 331 zfs_earlier_version(const char *name, int version)
 332 {
 333         spa_t *spa;
 334 
 335         if (spa_open(name, &spa, FTAG) == 0) {
 336                 if (spa_version(spa) < version) {
 337                         spa_close(spa, FTAG);
 338                         return (1);
 339                 }
 340                 spa_close(spa, FTAG);
 341         }
 342         return (0);
 343 }
 344 
 345 /*
 346  * zpl_earlier_version
 347  *
 348  * Return TRUE if the ZPL version is less than requested version.
 349  */
 350 static boolean_t
 351 zpl_earlier_version(const char *name, int version)
 352 {
 353         objset_t *os;
 354         boolean_t rc = B_TRUE;
 355 
 356         if (dmu_objset_hold(name, FTAG, &os) == 0) {
 357                 uint64_t zplversion;
 358 
 359                 if (dmu_objset_type(os) != DMU_OST_ZFS) {
 360                         dmu_objset_rele(os, FTAG);
 361                         return (B_TRUE);
 362                 }
 363                 /* XXX reading from non-owned objset */
 364                 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
 365                         rc = zplversion < version;
 366                 dmu_objset_rele(os, FTAG);
 367         }
 368         return (rc);
 369 }
 370 
 371 static void
 372 zfs_log_history(zfs_cmd_t *zc)
 373 {
 374         spa_t *spa;
 375         char *buf;
 376 
 377         if ((buf = history_str_get(zc)) == NULL)
 378                 return;
 379 
 380         if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
 381                 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
 382                         (void) spa_history_log(spa, buf);
 383                 spa_close(spa, FTAG);
 384         }
 385         history_str_free(buf);
 386 }
 387 
 388 /*
 389  * Policy for top-level read operations (list pools).  Requires no privileges,
 390  * and can be used in the local zone, as there is no associated dataset.
 391  */
 392 /* ARGSUSED */
 393 static int
 394 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 395 {
 396         return (0);
 397 }
 398 
 399 /*
 400  * Policy for dataset read operations (list children, get statistics).  Requires
 401  * no privileges, but must be visible in the local zone.
 402  */
 403 /* ARGSUSED */
 404 static int
 405 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 406 {
 407         if (INGLOBALZONE(curproc) ||
 408             zone_dataset_visible(zc->zc_name, NULL))
 409                 return (0);
 410 
 411         return (SET_ERROR(ENOENT));
 412 }
 413 
 414 static int
 415 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
 416 {
 417         int writable = 1;
 418 
 419         /*
 420          * The dataset must be visible by this zone -- check this first
 421          * so they don't see EPERM on something they shouldn't know about.
 422          */
 423         if (!INGLOBALZONE(curproc) &&
 424             !zone_dataset_visible(dataset, &writable))
 425                 return (SET_ERROR(ENOENT));
 426 
 427         if (INGLOBALZONE(curproc)) {
 428                 /*
 429                  * If the fs is zoned, only root can access it from the
 430                  * global zone.
 431                  */
 432                 if (secpolicy_zfs(cr) && zoned)
 433                         return (SET_ERROR(EPERM));
 434         } else {
 435                 /*
 436                  * If we are in a local zone, the 'zoned' property must be set.
 437                  */
 438                 if (!zoned)
 439                         return (SET_ERROR(EPERM));
 440 
 441                 /* must be writable by this zone */
 442                 if (!writable)
 443                         return (SET_ERROR(EPERM));
 444         }
 445         return (0);
 446 }
 447 
 448 static int
 449 zfs_dozonecheck(const char *dataset, cred_t *cr)
 450 {
 451         uint64_t zoned;
 452 
 453         if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
 454                 return (SET_ERROR(ENOENT));
 455 
 456         return (zfs_dozonecheck_impl(dataset, zoned, cr));
 457 }
 458 
 459 static int
 460 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
 461 {
 462         uint64_t zoned;
 463 
 464         if (dsl_prop_get_int_ds(ds, "zoned", &zoned))
 465                 return (SET_ERROR(ENOENT));
 466 
 467         return (zfs_dozonecheck_impl(dataset, zoned, cr));
 468 }
 469 
 470 static int
 471 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
 472     const char *perm, cred_t *cr)
 473 {
 474         int error;
 475 
 476         error = zfs_dozonecheck_ds(name, ds, cr);
 477         if (error == 0) {
 478                 error = secpolicy_zfs(cr);
 479                 if (error != 0)
 480                         error = dsl_deleg_access_impl(ds, perm, cr);
 481         }
 482         return (error);
 483 }
 484 
 485 static int
 486 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
 487 {
 488         int error;
 489         dsl_dataset_t *ds;
 490         dsl_pool_t *dp;
 491 
 492         error = dsl_pool_hold(name, FTAG, &dp);
 493         if (error != 0)
 494                 return (error);
 495 
 496         error = dsl_dataset_hold(dp, name, FTAG, &ds);
 497         if (error != 0) {
 498                 dsl_pool_rele(dp, FTAG);
 499                 return (error);
 500         }
 501 
 502         error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
 503 
 504         dsl_dataset_rele(ds, FTAG);
 505         dsl_pool_rele(dp, FTAG);
 506         return (error);
 507 }
 508 
 509 /*
 510  * Policy for setting the security label property.
 511  *
 512  * Returns 0 for success, non-zero for access and other errors.
 513  */
 514 static int
 515 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
 516 {
 517         char            ds_hexsl[MAXNAMELEN];
 518         bslabel_t       ds_sl, new_sl;
 519         boolean_t       new_default = FALSE;
 520         uint64_t        zoned;
 521         int             needed_priv = -1;
 522         int             error;
 523 
 524         /* First get the existing dataset label. */
 525         error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
 526             1, sizeof (ds_hexsl), &ds_hexsl, NULL);
 527         if (error != 0)
 528                 return (SET_ERROR(EPERM));
 529 
 530         if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
 531                 new_default = TRUE;
 532 
 533         /* The label must be translatable */
 534         if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
 535                 return (SET_ERROR(EINVAL));
 536 
 537         /*
 538          * In a non-global zone, disallow attempts to set a label that
 539          * doesn't match that of the zone; otherwise no other checks
 540          * are needed.
 541          */
 542         if (!INGLOBALZONE(curproc)) {
 543                 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
 544                         return (SET_ERROR(EPERM));
 545                 return (0);
 546         }
 547 
 548         /*
 549          * For global-zone datasets (i.e., those whose zoned property is
 550          * "off", verify that the specified new label is valid for the
 551          * global zone.
 552          */
 553         if (dsl_prop_get_integer(name,
 554             zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
 555                 return (SET_ERROR(EPERM));
 556         if (!zoned) {
 557                 if (zfs_check_global_label(name, strval) != 0)
 558                         return (SET_ERROR(EPERM));
 559         }
 560 
 561         /*
 562          * If the existing dataset label is nondefault, check if the
 563          * dataset is mounted (label cannot be changed while mounted).
 564          * Get the zfsvfs; if there isn't one, then the dataset isn't
 565          * mounted (or isn't a dataset, doesn't exist, ...).
 566          */
 567         if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
 568                 objset_t *os;
 569                 static char *setsl_tag = "setsl_tag";
 570 
 571                 /*
 572                  * Try to own the dataset; abort if there is any error,
 573                  * (e.g., already mounted, in use, or other error).
 574                  */
 575                 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
 576                     setsl_tag, &os);
 577                 if (error != 0)
 578                         return (SET_ERROR(EPERM));
 579 
 580                 dmu_objset_disown(os, setsl_tag);
 581 
 582                 if (new_default) {
 583                         needed_priv = PRIV_FILE_DOWNGRADE_SL;
 584                         goto out_check;
 585                 }
 586 
 587                 if (hexstr_to_label(strval, &new_sl) != 0)
 588                         return (SET_ERROR(EPERM));
 589 
 590                 if (blstrictdom(&ds_sl, &new_sl))
 591                         needed_priv = PRIV_FILE_DOWNGRADE_SL;
 592                 else if (blstrictdom(&new_sl, &ds_sl))
 593                         needed_priv = PRIV_FILE_UPGRADE_SL;
 594         } else {
 595                 /* dataset currently has a default label */
 596                 if (!new_default)
 597                         needed_priv = PRIV_FILE_UPGRADE_SL;
 598         }
 599 
 600 out_check:
 601         if (needed_priv != -1)
 602                 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
 603         return (0);
 604 }
 605 
 606 static int
 607 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
 608     cred_t *cr)
 609 {
 610         char *strval;
 611 
 612         /*
 613          * Check permissions for special properties.
 614          */
 615         switch (prop) {
 616         case ZFS_PROP_ZONED:
 617                 /*
 618                  * Disallow setting of 'zoned' from within a local zone.
 619                  */
 620                 if (!INGLOBALZONE(curproc))
 621                         return (SET_ERROR(EPERM));
 622                 break;
 623 
 624         case ZFS_PROP_QUOTA:
 625                 if (!INGLOBALZONE(curproc)) {
 626                         uint64_t zoned;
 627                         char setpoint[MAXNAMELEN];
 628                         /*
 629                          * Unprivileged users are allowed to modify the
 630                          * quota on things *under* (ie. contained by)
 631                          * the thing they own.
 632                          */
 633                         if (dsl_prop_get_integer(dsname, "zoned", &zoned,
 634                             setpoint))
 635                                 return (SET_ERROR(EPERM));
 636                         if (!zoned || strlen(dsname) <= strlen(setpoint))
 637                                 return (SET_ERROR(EPERM));
 638                 }
 639                 break;
 640 
 641         case ZFS_PROP_MLSLABEL:
 642                 if (!is_system_labeled())
 643                         return (SET_ERROR(EPERM));
 644 
 645                 if (nvpair_value_string(propval, &strval) == 0) {
 646                         int err;
 647 
 648                         err = zfs_set_slabel_policy(dsname, strval, CRED());
 649                         if (err != 0)
 650                                 return (err);
 651                 }
 652                 break;
 653         }
 654 
 655         return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
 656 }
 657 
 658 /* ARGSUSED */
 659 static int
 660 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 661 {
 662         int error;
 663 
 664         error = zfs_dozonecheck(zc->zc_name, cr);
 665         if (error != 0)
 666                 return (error);
 667 
 668         /*
 669          * permission to set permissions will be evaluated later in
 670          * dsl_deleg_can_allow()
 671          */
 672         return (0);
 673 }
 674 
 675 /* ARGSUSED */
 676 static int
 677 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 678 {
 679         return (zfs_secpolicy_write_perms(zc->zc_name,
 680             ZFS_DELEG_PERM_ROLLBACK, cr));
 681 }
 682 
 683 /* ARGSUSED */
 684 static int
 685 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 686 {
 687         dsl_pool_t *dp;
 688         dsl_dataset_t *ds;
 689         char *cp;
 690         int error;
 691 
 692         /*
 693          * Generate the current snapshot name from the given objsetid, then
 694          * use that name for the secpolicy/zone checks.
 695          */
 696         cp = strchr(zc->zc_name, '@');
 697         if (cp == NULL)
 698                 return (SET_ERROR(EINVAL));
 699         error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
 700         if (error != 0)
 701                 return (error);
 702 
 703         error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
 704         if (error != 0) {
 705                 dsl_pool_rele(dp, FTAG);
 706                 return (error);
 707         }
 708 
 709         dsl_dataset_name(ds, zc->zc_name);
 710 
 711         error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
 712             ZFS_DELEG_PERM_SEND, cr);
 713         dsl_dataset_rele(ds, FTAG);
 714         dsl_pool_rele(dp, FTAG);
 715 
 716         return (error);
 717 }
 718 
 719 /* ARGSUSED */
 720 static int
 721 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 722 {
 723         return (zfs_secpolicy_write_perms(zc->zc_name,
 724             ZFS_DELEG_PERM_SEND, cr));
 725 }
 726 
 727 /* ARGSUSED */
 728 static int
 729 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 730 {
 731         vnode_t *vp;
 732         int error;
 733 
 734         if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
 735             NO_FOLLOW, NULL, &vp)) != 0)
 736                 return (error);
 737 
 738         /* Now make sure mntpnt and dataset are ZFS */
 739 
 740         if (vp->v_vfsp->vfs_fstype != zfsfstype ||
 741             (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
 742             zc->zc_name) != 0)) {
 743                 VN_RELE(vp);
 744                 return (SET_ERROR(EPERM));
 745         }
 746 
 747         VN_RELE(vp);
 748         return (dsl_deleg_access(zc->zc_name,
 749             ZFS_DELEG_PERM_SHARE, cr));
 750 }
 751 
 752 int
 753 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 754 {
 755         if (!INGLOBALZONE(curproc))
 756                 return (SET_ERROR(EPERM));
 757 
 758         if (secpolicy_nfs(cr) == 0) {
 759                 return (0);
 760         } else {
 761                 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
 762         }
 763 }
 764 
 765 int
 766 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 767 {
 768         if (!INGLOBALZONE(curproc))
 769                 return (SET_ERROR(EPERM));
 770 
 771         if (secpolicy_smb(cr) == 0) {
 772                 return (0);
 773         } else {
 774                 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
 775         }
 776 }
 777 
 778 static int
 779 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
 780 {
 781         char *cp;
 782 
 783         /*
 784          * Remove the @bla or /bla from the end of the name to get the parent.
 785          */
 786         (void) strncpy(parent, datasetname, parentsize);
 787         cp = strrchr(parent, '@');
 788         if (cp != NULL) {
 789                 cp[0] = '\0';
 790         } else {
 791                 cp = strrchr(parent, '/');
 792                 if (cp == NULL)
 793                         return (SET_ERROR(ENOENT));
 794                 cp[0] = '\0';
 795         }
 796 
 797         return (0);
 798 }
 799 
 800 int
 801 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
 802 {
 803         int error;
 804 
 805         if ((error = zfs_secpolicy_write_perms(name,
 806             ZFS_DELEG_PERM_MOUNT, cr)) != 0)
 807                 return (error);
 808 
 809         return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
 810 }
 811 
 812 /* ARGSUSED */
 813 static int
 814 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 815 {
 816         return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
 817 }
 818 
 819 /*
 820  * Destroying snapshots with delegated permissions requires
 821  * descendant mount and destroy permissions.
 822  */
 823 /* ARGSUSED */
 824 static int
 825 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 826 {
 827         nvlist_t *snaps;
 828         nvpair_t *pair, *nextpair;
 829         int error = 0;
 830 
 831         if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
 832                 return (SET_ERROR(EINVAL));
 833         for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
 834             pair = nextpair) {
 835                 dsl_pool_t *dp;
 836                 dsl_dataset_t *ds;
 837 
 838                 error = dsl_pool_hold(nvpair_name(pair), FTAG, &dp);
 839                 if (error != 0)
 840                         break;
 841                 nextpair = nvlist_next_nvpair(snaps, pair);
 842                 error = dsl_dataset_hold(dp, nvpair_name(pair), FTAG, &ds);
 843                 if (error == 0)
 844                         dsl_dataset_rele(ds, FTAG);
 845                 dsl_pool_rele(dp, FTAG);
 846 
 847                 if (error == 0) {
 848                         error = zfs_secpolicy_destroy_perms(nvpair_name(pair),
 849                             cr);
 850                 } else if (error == ENOENT) {
 851                         /*
 852                          * Ignore any snapshots that don't exist (we consider
 853                          * them "already destroyed").  Remove the name from the
 854                          * nvl here in case the snapshot is created between
 855                          * now and when we try to destroy it (in which case
 856                          * we don't want to destroy it since we haven't
 857                          * checked for permission).
 858                          */
 859                         fnvlist_remove_nvpair(snaps, pair);
 860                         error = 0;
 861                 }
 862                 if (error != 0)
 863                         break;
 864         }
 865 
 866         return (error);
 867 }
 868 
 869 int
 870 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
 871 {
 872         char    parentname[MAXNAMELEN];
 873         int     error;
 874 
 875         if ((error = zfs_secpolicy_write_perms(from,
 876             ZFS_DELEG_PERM_RENAME, cr)) != 0)
 877                 return (error);
 878 
 879         if ((error = zfs_secpolicy_write_perms(from,
 880             ZFS_DELEG_PERM_MOUNT, cr)) != 0)
 881                 return (error);
 882 
 883         if ((error = zfs_get_parent(to, parentname,
 884             sizeof (parentname))) != 0)
 885                 return (error);
 886 
 887         if ((error = zfs_secpolicy_write_perms(parentname,
 888             ZFS_DELEG_PERM_CREATE, cr)) != 0)
 889                 return (error);
 890 
 891         if ((error = zfs_secpolicy_write_perms(parentname,
 892             ZFS_DELEG_PERM_MOUNT, cr)) != 0)
 893                 return (error);
 894 
 895         return (error);
 896 }
 897 
 898 /* ARGSUSED */
 899 static int
 900 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 901 {
 902         return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
 903 }
 904 
 905 /* ARGSUSED */
 906 static int
 907 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 908 {
 909         dsl_pool_t *dp;
 910         dsl_dataset_t *clone;
 911         int error;
 912 
 913         error = zfs_secpolicy_write_perms(zc->zc_name,
 914             ZFS_DELEG_PERM_PROMOTE, cr);
 915         if (error != 0)
 916                 return (error);
 917 
 918         error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
 919         if (error != 0)
 920                 return (error);
 921 
 922         error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
 923 
 924         if (error == 0) {
 925                 char parentname[MAXNAMELEN];
 926                 dsl_dataset_t *origin = NULL;
 927                 dsl_dir_t *dd;
 928                 dd = clone->ds_dir;
 929 
 930                 error = dsl_dataset_hold_obj(dd->dd_pool,
 931                     dd->dd_phys->dd_origin_obj, FTAG, &origin);
 932                 if (error != 0) {
 933                         dsl_dataset_rele(clone, FTAG);
 934                         dsl_pool_rele(dp, FTAG);
 935                         return (error);
 936                 }
 937 
 938                 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
 939                     ZFS_DELEG_PERM_MOUNT, cr);
 940 
 941                 dsl_dataset_name(origin, parentname);
 942                 if (error == 0) {
 943                         error = zfs_secpolicy_write_perms_ds(parentname, origin,
 944                             ZFS_DELEG_PERM_PROMOTE, cr);
 945                 }
 946                 dsl_dataset_rele(clone, FTAG);
 947                 dsl_dataset_rele(origin, FTAG);
 948         }
 949         dsl_pool_rele(dp, FTAG);
 950         return (error);
 951 }
 952 
 953 /* ARGSUSED */
 954 static int
 955 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 956 {
 957         int error;
 958 
 959         if ((error = zfs_secpolicy_write_perms(zc->zc_name,
 960             ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
 961                 return (error);
 962 
 963         if ((error = zfs_secpolicy_write_perms(zc->zc_name,
 964             ZFS_DELEG_PERM_MOUNT, cr)) != 0)
 965                 return (error);
 966 
 967         return (zfs_secpolicy_write_perms(zc->zc_name,
 968             ZFS_DELEG_PERM_CREATE, cr));
 969 }
 970 
 971 int
 972 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
 973 {
 974         return (zfs_secpolicy_write_perms(name,
 975             ZFS_DELEG_PERM_SNAPSHOT, cr));
 976 }
 977 
 978 /*
 979  * Check for permission to create each snapshot in the nvlist.
 980  */
 981 /* ARGSUSED */
 982 static int
 983 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 984 {
 985         nvlist_t *snaps;
 986         int error = 0;
 987         nvpair_t *pair;
 988 
 989         if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
 990                 return (SET_ERROR(EINVAL));
 991         for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
 992             pair = nvlist_next_nvpair(snaps, pair)) {
 993                 char *name = nvpair_name(pair);
 994                 char *atp = strchr(name, '@');
 995 
 996                 if (atp == NULL) {
 997                         error = SET_ERROR(EINVAL);
 998                         break;
 999                 }
1000                 *atp = '\0';
1001                 error = zfs_secpolicy_snapshot_perms(name, cr);
1002                 *atp = '@';
1003                 if (error != 0)
1004                         break;
1005         }
1006         return (error);
1007 }
1008 
1009 /* ARGSUSED */
1010 static int
1011 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1012 {
1013         /*
1014          * Even root must have a proper TSD so that we know what pool
1015          * to log to.
1016          */
1017         if (tsd_get(zfs_allow_log_key) == NULL)
1018                 return (SET_ERROR(EPERM));
1019         return (0);
1020 }
1021 
1022 static int
1023 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1024 {
1025         char    parentname[MAXNAMELEN];
1026         int     error;
1027         char    *origin;
1028 
1029         if ((error = zfs_get_parent(zc->zc_name, parentname,
1030             sizeof (parentname))) != 0)
1031                 return (error);
1032 
1033         if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1034             (error = zfs_secpolicy_write_perms(origin,
1035             ZFS_DELEG_PERM_CLONE, cr)) != 0)
1036                 return (error);
1037 
1038         if ((error = zfs_secpolicy_write_perms(parentname,
1039             ZFS_DELEG_PERM_CREATE, cr)) != 0)
1040                 return (error);
1041 
1042         return (zfs_secpolicy_write_perms(parentname,
1043             ZFS_DELEG_PERM_MOUNT, cr));
1044 }
1045 
1046 /*
1047  * Policy for pool operations - create/destroy pools, add vdevs, etc.  Requires
1048  * SYS_CONFIG privilege, which is not available in a local zone.
1049  */
1050 /* ARGSUSED */
1051 static int
1052 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1053 {
1054         if (secpolicy_sys_config(cr, B_FALSE) != 0)
1055                 return (SET_ERROR(EPERM));
1056 
1057         return (0);
1058 }
1059 
1060 /*
1061  * Policy for object to name lookups.
1062  */
1063 /* ARGSUSED */
1064 static int
1065 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1066 {
1067         int error;
1068 
1069         if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1070                 return (0);
1071 
1072         error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1073         return (error);
1074 }
1075 
1076 /*
1077  * Policy for fault injection.  Requires all privileges.
1078  */
1079 /* ARGSUSED */
1080 static int
1081 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1082 {
1083         return (secpolicy_zinject(cr));
1084 }
1085 
1086 /* ARGSUSED */
1087 static int
1088 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1089 {
1090         zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1091 
1092         if (prop == ZPROP_INVAL) {
1093                 if (!zfs_prop_user(zc->zc_value))
1094                         return (SET_ERROR(EINVAL));
1095                 return (zfs_secpolicy_write_perms(zc->zc_name,
1096                     ZFS_DELEG_PERM_USERPROP, cr));
1097         } else {
1098                 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1099                     NULL, cr));
1100         }
1101 }
1102 
1103 static int
1104 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1105 {
1106         int err = zfs_secpolicy_read(zc, innvl, cr);
1107         if (err)
1108                 return (err);
1109 
1110         if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1111                 return (SET_ERROR(EINVAL));
1112 
1113         if (zc->zc_value[0] == 0) {
1114                 /*
1115                  * They are asking about a posix uid/gid.  If it's
1116                  * themself, allow it.
1117                  */
1118                 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1119                     zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
1120                         if (zc->zc_guid == crgetuid(cr))
1121                                 return (0);
1122                 } else {
1123                         if (groupmember(zc->zc_guid, cr))
1124                                 return (0);
1125                 }
1126         }
1127 
1128         return (zfs_secpolicy_write_perms(zc->zc_name,
1129             userquota_perms[zc->zc_objset_type], cr));
1130 }
1131 
1132 static int
1133 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1134 {
1135         int err = zfs_secpolicy_read(zc, innvl, cr);
1136         if (err)
1137                 return (err);
1138 
1139         if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1140                 return (SET_ERROR(EINVAL));
1141 
1142         return (zfs_secpolicy_write_perms(zc->zc_name,
1143             userquota_perms[zc->zc_objset_type], cr));
1144 }
1145 
1146 /* ARGSUSED */
1147 static int
1148 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1149 {
1150         return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1151             NULL, cr));
1152 }
1153 
1154 /* ARGSUSED */
1155 static int
1156 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1157 {
1158         nvpair_t *pair;
1159         nvlist_t *holds;
1160         int error;
1161 
1162         error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1163         if (error != 0)
1164                 return (SET_ERROR(EINVAL));
1165 
1166         for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1167             pair = nvlist_next_nvpair(holds, pair)) {
1168                 char fsname[MAXNAMELEN];
1169                 error = dmu_fsname(nvpair_name(pair), fsname);
1170                 if (error != 0)
1171                         return (error);
1172                 error = zfs_secpolicy_write_perms(fsname,
1173                     ZFS_DELEG_PERM_HOLD, cr);
1174                 if (error != 0)
1175                         return (error);
1176         }
1177         return (0);
1178 }
1179 
1180 /* ARGSUSED */
1181 static int
1182 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1183 {
1184         nvpair_t *pair;
1185         int error;
1186 
1187         for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1188             pair = nvlist_next_nvpair(innvl, pair)) {
1189                 char fsname[MAXNAMELEN];
1190                 error = dmu_fsname(nvpair_name(pair), fsname);
1191                 if (error != 0)
1192                         return (error);
1193                 error = zfs_secpolicy_write_perms(fsname,
1194                     ZFS_DELEG_PERM_RELEASE, cr);
1195                 if (error != 0)
1196                         return (error);
1197         }
1198         return (0);
1199 }
1200 
1201 /*
1202  * Policy for allowing temporary snapshots to be taken or released
1203  */
1204 static int
1205 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1206 {
1207         /*
1208          * A temporary snapshot is the same as a snapshot,
1209          * hold, destroy and release all rolled into one.
1210          * Delegated diff alone is sufficient that we allow this.
1211          */
1212         int error;
1213 
1214         if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1215             ZFS_DELEG_PERM_DIFF, cr)) == 0)
1216                 return (0);
1217 
1218         error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1219         if (error == 0)
1220                 error = zfs_secpolicy_hold(zc, innvl, cr);
1221         if (error == 0)
1222                 error = zfs_secpolicy_release(zc, innvl, cr);
1223         if (error == 0)
1224                 error = zfs_secpolicy_destroy(zc, innvl, cr);
1225         return (error);
1226 }
1227 
1228 /*
1229  * Returns the nvlist as specified by the user in the zfs_cmd_t.
1230  */
1231 static int
1232 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1233 {
1234         char *packed;
1235         int error;
1236         nvlist_t *list = NULL;
1237 
1238         /*
1239          * Read in and unpack the user-supplied nvlist.
1240          */
1241         if (size == 0)
1242                 return (SET_ERROR(EINVAL));
1243 
1244         packed = kmem_alloc(size, KM_SLEEP);
1245 
1246         if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1247             iflag)) != 0) {
1248                 kmem_free(packed, size);
1249                 return (error);
1250         }
1251 
1252         if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1253                 kmem_free(packed, size);
1254                 return (error);
1255         }
1256 
1257         kmem_free(packed, size);
1258 
1259         *nvp = list;
1260         return (0);
1261 }
1262 
1263 /*
1264  * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1265  * Entries will be removed from the end of the nvlist, and one int32 entry
1266  * named "N_MORE_ERRORS" will be added indicating how many entries were
1267  * removed.
1268  */
1269 static int
1270 nvlist_smush(nvlist_t *errors, size_t max)
1271 {
1272         size_t size;
1273 
1274         size = fnvlist_size(errors);
1275 
1276         if (size > max) {
1277                 nvpair_t *more_errors;
1278                 int n = 0;
1279 
1280                 if (max < 1024)
1281                         return (SET_ERROR(ENOMEM));
1282 
1283                 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1284                 more_errors = nvlist_prev_nvpair(errors, NULL);
1285 
1286                 do {
1287                         nvpair_t *pair = nvlist_prev_nvpair(errors,
1288                             more_errors);
1289                         fnvlist_remove_nvpair(errors, pair);
1290                         n++;
1291                         size = fnvlist_size(errors);
1292                 } while (size > max);
1293 
1294                 fnvlist_remove_nvpair(errors, more_errors);
1295                 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1296                 ASSERT3U(fnvlist_size(errors), <=, max);
1297         }
1298 
1299         return (0);
1300 }
1301 
1302 static int
1303 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1304 {
1305         char *packed = NULL;
1306         int error = 0;
1307         size_t size;
1308 
1309         size = fnvlist_size(nvl);
1310 
1311         if (size > zc->zc_nvlist_dst_size) {
1312                 error = SET_ERROR(ENOMEM);
1313         } else {
1314                 packed = fnvlist_pack(nvl, &size);
1315                 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1316                     size, zc->zc_iflags) != 0)
1317                         error = SET_ERROR(EFAULT);
1318                 fnvlist_pack_free(packed, size);
1319         }
1320 
1321         zc->zc_nvlist_dst_size = size;
1322         zc->zc_nvlist_dst_filled = B_TRUE;
1323         return (error);
1324 }
1325 
1326 static int
1327 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1328 {
1329         objset_t *os;
1330         int error;
1331 
1332         error = dmu_objset_hold(dsname, FTAG, &os);
1333         if (error != 0)
1334                 return (error);
1335         if (dmu_objset_type(os) != DMU_OST_ZFS) {
1336                 dmu_objset_rele(os, FTAG);
1337                 return (SET_ERROR(EINVAL));
1338         }
1339 
1340         mutex_enter(&os->os_user_ptr_lock);
1341         *zfvp = dmu_objset_get_user(os);
1342         if (*zfvp) {
1343                 VFS_HOLD((*zfvp)->z_vfs);
1344         } else {
1345                 error = SET_ERROR(ESRCH);
1346         }
1347         mutex_exit(&os->os_user_ptr_lock);
1348         dmu_objset_rele(os, FTAG);
1349         return (error);
1350 }
1351 
1352 /*
1353  * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1354  * case its z_vfs will be NULL, and it will be opened as the owner.
1355  * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1356  * which prevents all vnode ops from running.
1357  */
1358 static int
1359 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1360 {
1361         int error = 0;
1362 
1363         if (getzfsvfs(name, zfvp) != 0)
1364                 error = zfsvfs_create(name, zfvp);
1365         if (error == 0) {
1366                 rrw_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1367                     RW_READER, tag);
1368                 if ((*zfvp)->z_unmounted) {
1369                         /*
1370                          * XXX we could probably try again, since the unmounting
1371                          * thread should be just about to disassociate the
1372                          * objset from the zfsvfs.
1373                          */
1374                         rrw_exit(&(*zfvp)->z_teardown_lock, tag);
1375                         return (SET_ERROR(EBUSY));
1376                 }
1377         }
1378         return (error);
1379 }
1380 
1381 static void
1382 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1383 {
1384         rrw_exit(&zfsvfs->z_teardown_lock, tag);
1385 
1386         if (zfsvfs->z_vfs) {
1387                 VFS_RELE(zfsvfs->z_vfs);
1388         } else {
1389                 dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1390                 zfsvfs_free(zfsvfs);
1391         }
1392 }
1393 
1394 static int
1395 zfs_ioc_pool_create(zfs_cmd_t *zc)
1396 {
1397         int error;
1398         nvlist_t *config, *props = NULL;
1399         nvlist_t *rootprops = NULL;
1400         nvlist_t *zplprops = NULL;
1401 
1402         if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1403             zc->zc_iflags, &config))
1404                 return (error);
1405 
1406         if (zc->zc_nvlist_src_size != 0 && (error =
1407             get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1408             zc->zc_iflags, &props))) {
1409                 nvlist_free(config);
1410                 return (error);
1411         }
1412 
1413         if (props) {
1414                 nvlist_t *nvl = NULL;
1415                 uint64_t version = SPA_VERSION;
1416 
1417                 (void) nvlist_lookup_uint64(props,
1418                     zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1419                 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1420                         error = SET_ERROR(EINVAL);
1421                         goto pool_props_bad;
1422                 }
1423                 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1424                 if (nvl) {
1425                         error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1426                         if (error != 0) {
1427                                 nvlist_free(config);
1428                                 nvlist_free(props);
1429                                 return (error);
1430                         }
1431                         (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1432                 }
1433                 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1434                 error = zfs_fill_zplprops_root(version, rootprops,
1435                     zplprops, NULL);
1436                 if (error != 0)
1437                         goto pool_props_bad;
1438         }
1439 
1440         error = spa_create(zc->zc_name, config, props, zplprops);
1441 
1442         /*
1443          * Set the remaining root properties
1444          */
1445         if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1446             ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1447                 (void) spa_destroy(zc->zc_name);
1448 
1449 pool_props_bad:
1450         nvlist_free(rootprops);
1451         nvlist_free(zplprops);
1452         nvlist_free(config);
1453         nvlist_free(props);
1454 
1455         return (error);
1456 }
1457 
1458 static int
1459 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1460 {
1461         int error;
1462         zfs_log_history(zc);
1463         error = spa_destroy(zc->zc_name);
1464         if (error == 0)
1465                 zvol_remove_minors(zc->zc_name);
1466         return (error);
1467 }
1468 
1469 static int
1470 zfs_ioc_pool_import(zfs_cmd_t *zc)
1471 {
1472         nvlist_t *config, *props = NULL;
1473         uint64_t guid;
1474         int error;
1475 
1476         if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1477             zc->zc_iflags, &config)) != 0)
1478                 return (error);
1479 
1480         if (zc->zc_nvlist_src_size != 0 && (error =
1481             get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1482             zc->zc_iflags, &props))) {
1483                 nvlist_free(config);
1484                 return (error);
1485         }
1486 
1487         if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1488             guid != zc->zc_guid)
1489                 error = SET_ERROR(EINVAL);
1490         else
1491                 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1492 
1493         if (zc->zc_nvlist_dst != 0) {
1494                 int err;
1495 
1496                 if ((err = put_nvlist(zc, config)) != 0)
1497                         error = err;
1498         }
1499 
1500         nvlist_free(config);
1501 
1502         if (props)
1503                 nvlist_free(props);
1504 
1505         return (error);
1506 }
1507 
1508 static int
1509 zfs_ioc_pool_export(zfs_cmd_t *zc)
1510 {
1511         int error;
1512         boolean_t force = (boolean_t)zc->zc_cookie;
1513         boolean_t hardforce = (boolean_t)zc->zc_guid;
1514 
1515         zfs_log_history(zc);
1516         error = spa_export(zc->zc_name, NULL, force, hardforce);
1517         if (error == 0)
1518                 zvol_remove_minors(zc->zc_name);
1519         return (error);
1520 }
1521 
1522 static int
1523 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1524 {
1525         nvlist_t *configs;
1526         int error;
1527 
1528         if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1529                 return (SET_ERROR(EEXIST));
1530 
1531         error = put_nvlist(zc, configs);
1532 
1533         nvlist_free(configs);
1534 
1535         return (error);
1536 }
1537 
1538 /*
1539  * inputs:
1540  * zc_name              name of the pool
1541  *
1542  * outputs:
1543  * zc_cookie            real errno
1544  * zc_nvlist_dst        config nvlist
1545  * zc_nvlist_dst_size   size of config nvlist
1546  */
1547 static int
1548 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1549 {
1550         nvlist_t *config;
1551         int error;
1552         int ret = 0;
1553 
1554         error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1555             sizeof (zc->zc_value));
1556 
1557         if (config != NULL) {
1558                 ret = put_nvlist(zc, config);
1559                 nvlist_free(config);
1560 
1561                 /*
1562                  * The config may be present even if 'error' is non-zero.
1563                  * In this case we return success, and preserve the real errno
1564                  * in 'zc_cookie'.
1565                  */
1566                 zc->zc_cookie = error;
1567         } else {
1568                 ret = error;
1569         }
1570 
1571         return (ret);
1572 }
1573 
1574 /*
1575  * Try to import the given pool, returning pool stats as appropriate so that
1576  * user land knows which devices are available and overall pool health.
1577  */
1578 static int
1579 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1580 {
1581         nvlist_t *tryconfig, *config;
1582         int error;
1583 
1584         if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1585             zc->zc_iflags, &tryconfig)) != 0)
1586                 return (error);
1587 
1588         config = spa_tryimport(tryconfig);
1589 
1590         nvlist_free(tryconfig);
1591 
1592         if (config == NULL)
1593                 return (SET_ERROR(EINVAL));
1594 
1595         error = put_nvlist(zc, config);
1596         nvlist_free(config);
1597 
1598         return (error);
1599 }
1600 
1601 /*
1602  * inputs:
1603  * zc_name              name of the pool
1604  * zc_cookie            scan func (pool_scan_func_t)
1605  */
1606 static int
1607 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1608 {
1609         spa_t *spa;
1610         int error;
1611 
1612         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1613                 return (error);
1614 
1615         if (zc->zc_cookie == POOL_SCAN_NONE)
1616                 error = spa_scan_stop(spa);
1617         else
1618                 error = spa_scan(spa, zc->zc_cookie);
1619 
1620         spa_close(spa, FTAG);
1621 
1622         return (error);
1623 }
1624 
1625 static int
1626 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1627 {
1628         spa_t *spa;
1629         int error;
1630 
1631         error = spa_open(zc->zc_name, &spa, FTAG);
1632         if (error == 0) {
1633                 spa_freeze(spa);
1634                 spa_close(spa, FTAG);
1635         }
1636         return (error);
1637 }
1638 
1639 static int
1640 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1641 {
1642         spa_t *spa;
1643         int error;
1644 
1645         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1646                 return (error);
1647 
1648         if (zc->zc_cookie < spa_version(spa) ||
1649             !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1650                 spa_close(spa, FTAG);
1651                 return (SET_ERROR(EINVAL));
1652         }
1653 
1654         spa_upgrade(spa, zc->zc_cookie);
1655         spa_close(spa, FTAG);
1656 
1657         return (error);
1658 }
1659 
1660 static int
1661 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1662 {
1663         spa_t *spa;
1664         char *hist_buf;
1665         uint64_t size;
1666         int error;
1667 
1668         if ((size = zc->zc_history_len) == 0)
1669                 return (SET_ERROR(EINVAL));
1670 
1671         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1672                 return (error);
1673 
1674         if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1675                 spa_close(spa, FTAG);
1676                 return (SET_ERROR(ENOTSUP));
1677         }
1678 
1679         hist_buf = kmem_alloc(size, KM_SLEEP);
1680         if ((error = spa_history_get(spa, &zc->zc_history_offset,
1681             &zc->zc_history_len, hist_buf)) == 0) {
1682                 error = ddi_copyout(hist_buf,
1683                     (void *)(uintptr_t)zc->zc_history,
1684                     zc->zc_history_len, zc->zc_iflags);
1685         }
1686 
1687         spa_close(spa, FTAG);
1688         kmem_free(hist_buf, size);
1689         return (error);
1690 }
1691 
1692 static int
1693 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1694 {
1695         spa_t *spa;
1696         int error;
1697 
1698         error = spa_open(zc->zc_name, &spa, FTAG);
1699         if (error == 0) {
1700                 error = spa_change_guid(spa);
1701                 spa_close(spa, FTAG);
1702         }
1703         return (error);
1704 }
1705 
1706 static int
1707 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1708 {
1709         return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1710 }
1711 
1712 /*
1713  * inputs:
1714  * zc_name              name of filesystem
1715  * zc_obj               object to find
1716  *
1717  * outputs:
1718  * zc_value             name of object
1719  */
1720 static int
1721 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1722 {
1723         objset_t *os;
1724         int error;
1725 
1726         /* XXX reading from objset not owned */
1727         if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1728                 return (error);
1729         if (dmu_objset_type(os) != DMU_OST_ZFS) {
1730                 dmu_objset_rele(os, FTAG);
1731                 return (SET_ERROR(EINVAL));
1732         }
1733         error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1734             sizeof (zc->zc_value));
1735         dmu_objset_rele(os, FTAG);
1736 
1737         return (error);
1738 }
1739 
1740 /*
1741  * inputs:
1742  * zc_name              name of filesystem
1743  * zc_obj               object to find
1744  *
1745  * outputs:
1746  * zc_stat              stats on object
1747  * zc_value             path to object
1748  */
1749 static int
1750 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1751 {
1752         objset_t *os;
1753         int error;
1754 
1755         /* XXX reading from objset not owned */
1756         if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1757                 return (error);
1758         if (dmu_objset_type(os) != DMU_OST_ZFS) {
1759                 dmu_objset_rele(os, FTAG);
1760                 return (SET_ERROR(EINVAL));
1761         }
1762         error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1763             sizeof (zc->zc_value));
1764         dmu_objset_rele(os, FTAG);
1765 
1766         return (error);
1767 }
1768 
1769 static int
1770 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1771 {
1772         spa_t *spa;
1773         int error;
1774         nvlist_t *config, **l2cache, **spares;
1775         uint_t nl2cache = 0, nspares = 0;
1776 
1777         error = spa_open(zc->zc_name, &spa, FTAG);
1778         if (error != 0)
1779                 return (error);
1780 
1781         error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1782             zc->zc_iflags, &config);
1783         (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1784             &l2cache, &nl2cache);
1785 
1786         (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1787             &spares, &nspares);
1788 
1789         /*
1790          * A root pool with concatenated devices is not supported.
1791          * Thus, can not add a device to a root pool.
1792          *
1793          * Intent log device can not be added to a rootpool because
1794          * during mountroot, zil is replayed, a seperated log device
1795          * can not be accessed during the mountroot time.
1796          *
1797          * l2cache and spare devices are ok to be added to a rootpool.
1798          */
1799         if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1800                 nvlist_free(config);
1801                 spa_close(spa, FTAG);
1802                 return (SET_ERROR(EDOM));
1803         }
1804 
1805         if (error == 0) {
1806                 error = spa_vdev_add(spa, config);
1807                 nvlist_free(config);
1808         }
1809         spa_close(spa, FTAG);
1810         return (error);
1811 }
1812 
1813 /*
1814  * inputs:
1815  * zc_name              name of the pool
1816  * zc_nvlist_conf       nvlist of devices to remove
1817  * zc_cookie            to stop the remove?
1818  */
1819 static int
1820 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1821 {
1822         spa_t *spa;
1823         int error;
1824 
1825         error = spa_open(zc->zc_name, &spa, FTAG);
1826         if (error != 0)
1827                 return (error);
1828         error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1829         spa_close(spa, FTAG);
1830         return (error);
1831 }
1832 
1833 static int
1834 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1835 {
1836         spa_t *spa;
1837         int error;
1838         vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1839 
1840         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1841                 return (error);
1842         switch (zc->zc_cookie) {
1843         case VDEV_STATE_ONLINE:
1844                 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1845                 break;
1846 
1847         case VDEV_STATE_OFFLINE:
1848                 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1849                 break;
1850 
1851         case VDEV_STATE_FAULTED:
1852                 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1853                     zc->zc_obj != VDEV_AUX_EXTERNAL)
1854                         zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1855 
1856                 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1857                 break;
1858 
1859         case VDEV_STATE_DEGRADED:
1860                 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1861                     zc->zc_obj != VDEV_AUX_EXTERNAL)
1862                         zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1863 
1864                 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1865                 break;
1866 
1867         default:
1868                 error = SET_ERROR(EINVAL);
1869         }
1870         zc->zc_cookie = newstate;
1871         spa_close(spa, FTAG);
1872         return (error);
1873 }
1874 
1875 static int
1876 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1877 {
1878         spa_t *spa;
1879         int replacing = zc->zc_cookie;
1880         nvlist_t *config;
1881         int error;
1882 
1883         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1884                 return (error);
1885 
1886         if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1887             zc->zc_iflags, &config)) == 0) {
1888                 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1889                 nvlist_free(config);
1890         }
1891 
1892         spa_close(spa, FTAG);
1893         return (error);
1894 }
1895 
1896 static int
1897 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1898 {
1899         spa_t *spa;
1900         int error;
1901 
1902         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1903                 return (error);
1904 
1905         error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1906 
1907         spa_close(spa, FTAG);
1908         return (error);
1909 }
1910 
1911 static int
1912 zfs_ioc_vdev_split(zfs_cmd_t *zc)
1913 {
1914         spa_t *spa;
1915         nvlist_t *config, *props = NULL;
1916         int error;
1917         boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
1918 
1919         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1920                 return (error);
1921 
1922         if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1923             zc->zc_iflags, &config)) {
1924                 spa_close(spa, FTAG);
1925                 return (error);
1926         }
1927 
1928         if (zc->zc_nvlist_src_size != 0 && (error =
1929             get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1930             zc->zc_iflags, &props))) {
1931                 spa_close(spa, FTAG);
1932                 nvlist_free(config);
1933                 return (error);
1934         }
1935 
1936         error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
1937 
1938         spa_close(spa, FTAG);
1939 
1940         nvlist_free(config);
1941         nvlist_free(props);
1942 
1943         return (error);
1944 }
1945 
1946 static int
1947 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
1948 {
1949         spa_t *spa;
1950         char *path = zc->zc_value;
1951         uint64_t guid = zc->zc_guid;
1952         int error;
1953 
1954         error = spa_open(zc->zc_name, &spa, FTAG);
1955         if (error != 0)
1956                 return (error);
1957 
1958         error = spa_vdev_setpath(spa, guid, path);
1959         spa_close(spa, FTAG);
1960         return (error);
1961 }
1962 
1963 static int
1964 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
1965 {
1966         spa_t *spa;
1967         char *fru = zc->zc_value;
1968         uint64_t guid = zc->zc_guid;
1969         int error;
1970 
1971         error = spa_open(zc->zc_name, &spa, FTAG);
1972         if (error != 0)
1973                 return (error);
1974 
1975         error = spa_vdev_setfru(spa, guid, fru);
1976         spa_close(spa, FTAG);
1977         return (error);
1978 }
1979 
1980 static int
1981 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
1982 {
1983         int error = 0;
1984         nvlist_t *nv;
1985 
1986         dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1987 
1988         if (zc->zc_nvlist_dst != 0 &&
1989             (error = dsl_prop_get_all(os, &nv)) == 0) {
1990                 dmu_objset_stats(os, nv);
1991                 /*
1992                  * NB: zvol_get_stats() will read the objset contents,
1993                  * which we aren't supposed to do with a
1994                  * DS_MODE_USER hold, because it could be
1995                  * inconsistent.  So this is a bit of a workaround...
1996                  * XXX reading with out owning
1997                  */
1998                 if (!zc->zc_objset_stats.dds_inconsistent &&
1999                     dmu_objset_type(os) == DMU_OST_ZVOL) {
2000                         error = zvol_get_stats(os, nv);
2001                         if (error == EIO)
2002                                 return (error);
2003                         VERIFY0(error);
2004                 }
2005                 error = put_nvlist(zc, nv);
2006                 nvlist_free(nv);
2007         }
2008 
2009         return (error);
2010 }
2011 
2012 /*
2013  * inputs:
2014  * zc_name              name of filesystem
2015  * zc_nvlist_dst_size   size of buffer for property nvlist
2016  *
2017  * outputs:
2018  * zc_objset_stats      stats
2019  * zc_nvlist_dst        property nvlist
2020  * zc_nvlist_dst_size   size of property nvlist
2021  */
2022 static int
2023 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2024 {
2025         objset_t *os;
2026         int error;
2027 
2028         error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2029         if (error == 0) {
2030                 error = zfs_ioc_objset_stats_impl(zc, os);
2031                 dmu_objset_rele(os, FTAG);
2032         }
2033 
2034         return (error);
2035 }
2036 
2037 /*
2038  * inputs:
2039  * zc_name              name of filesystem
2040  * zc_nvlist_dst_size   size of buffer for property nvlist
2041  *
2042  * outputs:
2043  * zc_nvlist_dst        received property nvlist
2044  * zc_nvlist_dst_size   size of received property nvlist
2045  *
2046  * Gets received properties (distinct from local properties on or after
2047  * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2048  * local property values.
2049  */
2050 static int
2051 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2052 {
2053         int error = 0;
2054         nvlist_t *nv;
2055 
2056         /*
2057          * Without this check, we would return local property values if the
2058          * caller has not already received properties on or after
2059          * SPA_VERSION_RECVD_PROPS.
2060          */
2061         if (!dsl_prop_get_hasrecvd(zc->zc_name))
2062                 return (SET_ERROR(ENOTSUP));
2063 
2064         if (zc->zc_nvlist_dst != 0 &&
2065             (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2066                 error = put_nvlist(zc, nv);
2067                 nvlist_free(nv);
2068         }
2069 
2070         return (error);
2071 }
2072 
2073 static int
2074 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2075 {
2076         uint64_t value;
2077         int error;
2078 
2079         /*
2080          * zfs_get_zplprop() will either find a value or give us
2081          * the default value (if there is one).
2082          */
2083         if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2084                 return (error);
2085         VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2086         return (0);
2087 }
2088 
2089 /*
2090  * inputs:
2091  * zc_name              name of filesystem
2092  * zc_nvlist_dst_size   size of buffer for zpl property nvlist
2093  *
2094  * outputs:
2095  * zc_nvlist_dst        zpl property nvlist
2096  * zc_nvlist_dst_size   size of zpl property nvlist
2097  */
2098 static int
2099 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2100 {
2101         objset_t *os;
2102         int err;
2103 
2104         /* XXX reading without owning */
2105         if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2106                 return (err);
2107 
2108         dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2109 
2110         /*
2111          * NB: nvl_add_zplprop() will read the objset contents,
2112          * which we aren't supposed to do with a DS_MODE_USER
2113          * hold, because it could be inconsistent.
2114          */
2115         if (zc->zc_nvlist_dst != NULL &&
2116             !zc->zc_objset_stats.dds_inconsistent &&
2117             dmu_objset_type(os) == DMU_OST_ZFS) {
2118                 nvlist_t *nv;
2119 
2120                 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2121                 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2122                     (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2123                     (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2124                     (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2125                         err = put_nvlist(zc, nv);
2126                 nvlist_free(nv);
2127         } else {
2128                 err = SET_ERROR(ENOENT);
2129         }
2130         dmu_objset_rele(os, FTAG);
2131         return (err);
2132 }
2133 
2134 static boolean_t
2135 dataset_name_hidden(const char *name)
2136 {
2137         /*
2138          * Skip over datasets that are not visible in this zone,
2139          * internal datasets (which have a $ in their name), and
2140          * temporary datasets (which have a % in their name).
2141          */
2142         if (strchr(name, '$') != NULL)
2143                 return (B_TRUE);
2144         if (strchr(name, '%') != NULL)
2145                 return (B_TRUE);
2146         if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
2147                 return (B_TRUE);
2148         return (B_FALSE);
2149 }
2150 
2151 /*
2152  * inputs:
2153  * zc_name              name of filesystem
2154  * zc_cookie            zap cursor
2155  * zc_nvlist_dst_size   size of buffer for property nvlist
2156  *
2157  * outputs:
2158  * zc_name              name of next filesystem
2159  * zc_cookie            zap cursor
2160  * zc_objset_stats      stats
2161  * zc_nvlist_dst        property nvlist
2162  * zc_nvlist_dst_size   size of property nvlist
2163  */
2164 static int
2165 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2166 {
2167         objset_t *os;
2168         int error;
2169         char *p;
2170         size_t orig_len = strlen(zc->zc_name);
2171 
2172 top:
2173         if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2174                 if (error == ENOENT)
2175                         error = SET_ERROR(ESRCH);
2176                 return (error);
2177         }
2178 
2179         p = strrchr(zc->zc_name, '/');
2180         if (p == NULL || p[1] != '\0')
2181                 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2182         p = zc->zc_name + strlen(zc->zc_name);
2183 
2184         do {
2185                 error = dmu_dir_list_next(os,
2186                     sizeof (zc->zc_name) - (p - zc->zc_name), p,
2187                     NULL, &zc->zc_cookie);
2188                 if (error == ENOENT)
2189                         error = SET_ERROR(ESRCH);
2190         } while (error == 0 && dataset_name_hidden(zc->zc_name));
2191         dmu_objset_rele(os, FTAG);
2192 
2193         /*
2194          * If it's an internal dataset (ie. with a '$' in its name),
2195          * don't try to get stats for it, otherwise we'll return ENOENT.
2196          */
2197         if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2198                 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2199                 if (error == ENOENT) {
2200                         /* We lost a race with destroy, get the next one. */
2201                         zc->zc_name[orig_len] = '\0';
2202                         goto top;
2203                 }
2204         }
2205         return (error);
2206 }
2207 
2208 /*
2209  * inputs:
2210  * zc_name              name of filesystem
2211  * zc_cookie            zap cursor
2212  * zc_nvlist_dst_size   size of buffer for property nvlist
2213  *
2214  * outputs:
2215  * zc_name              name of next snapshot
2216  * zc_objset_stats      stats
2217  * zc_nvlist_dst        property nvlist
2218  * zc_nvlist_dst_size   size of property nvlist
2219  */
2220 static int
2221 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2222 {
2223         objset_t *os;
2224         int error;
2225 
2226         error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2227         if (error != 0) {
2228                 return (error == ENOENT ? ESRCH : error);
2229         }
2230 
2231         /*
2232          * A dataset name of maximum length cannot have any snapshots,
2233          * so exit immediately.
2234          */
2235         if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2236                 dmu_objset_rele(os, FTAG);
2237                 return (SET_ERROR(ESRCH));
2238         }
2239 
2240         error = dmu_snapshot_list_next(os,
2241             sizeof (zc->zc_name) - strlen(zc->zc_name),
2242             zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2243             NULL);
2244 
2245         if (error == 0) {
2246                 dsl_dataset_t *ds;
2247                 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2248 
2249                 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2250                 if (error == 0) {
2251                         objset_t *ossnap;
2252 
2253                         error = dmu_objset_from_ds(ds, &ossnap);
2254                         if (error == 0)
2255                                 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2256                         dsl_dataset_rele(ds, FTAG);
2257                 }
2258         } else if (error == ENOENT) {
2259                 error = SET_ERROR(ESRCH);
2260         }
2261 
2262         dmu_objset_rele(os, FTAG);
2263         /* if we failed, undo the @ that we tacked on to zc_name */
2264         if (error != 0)
2265                 *strchr(zc->zc_name, '@') = '\0';
2266         return (error);
2267 }
2268 
2269 static int
2270 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2271 {
2272         const char *propname = nvpair_name(pair);
2273         uint64_t *valary;
2274         unsigned int vallen;
2275         const char *domain;
2276         char *dash;
2277         zfs_userquota_prop_t type;
2278         uint64_t rid;
2279         uint64_t quota;
2280         zfsvfs_t *zfsvfs;
2281         int err;
2282 
2283         if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2284                 nvlist_t *attrs;
2285                 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2286                 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2287                     &pair) != 0)
2288                         return (SET_ERROR(EINVAL));
2289         }
2290 
2291         /*
2292          * A correctly constructed propname is encoded as
2293          * userquota@<rid>-<domain>.
2294          */
2295         if ((dash = strchr(propname, '-')) == NULL ||
2296             nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2297             vallen != 3)
2298                 return (SET_ERROR(EINVAL));
2299 
2300         domain = dash + 1;
2301         type = valary[0];
2302         rid = valary[1];
2303         quota = valary[2];
2304 
2305         err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2306         if (err == 0) {
2307                 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2308                 zfsvfs_rele(zfsvfs, FTAG);
2309         }
2310 
2311         return (err);
2312 }
2313 
2314 /*
2315  * If the named property is one that has a special function to set its value,
2316  * return 0 on success and a positive error code on failure; otherwise if it is
2317  * not one of the special properties handled by this function, return -1.
2318  *
2319  * XXX: It would be better for callers of the property interface if we handled
2320  * these special cases in dsl_prop.c (in the dsl layer).
2321  */
2322 static int
2323 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2324     nvpair_t *pair)
2325 {
2326         const char *propname = nvpair_name(pair);
2327         zfs_prop_t prop = zfs_name_to_prop(propname);
2328         uint64_t intval;
2329         int err;
2330 
2331         if (prop == ZPROP_INVAL) {
2332                 if (zfs_prop_userquota(propname))
2333                         return (zfs_prop_set_userquota(dsname, pair));
2334                 return (-1);
2335         }
2336 
2337         if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2338                 nvlist_t *attrs;
2339                 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2340                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2341                     &pair) == 0);
2342         }
2343 
2344         if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2345                 return (-1);
2346 
2347         VERIFY(0 == nvpair_value_uint64(pair, &intval));
2348 
2349         switch (prop) {
2350         case ZFS_PROP_QUOTA:
2351                 err = dsl_dir_set_quota(dsname, source, intval);
2352                 break;
2353         case ZFS_PROP_REFQUOTA:
2354                 err = dsl_dataset_set_refquota(dsname, source, intval);
2355                 break;
2356         case ZFS_PROP_RESERVATION:
2357                 err = dsl_dir_set_reservation(dsname, source, intval);
2358                 break;
2359         case ZFS_PROP_REFRESERVATION:
2360                 err = dsl_dataset_set_refreservation(dsname, source, intval);
2361                 break;
2362         case ZFS_PROP_VOLSIZE:
2363                 err = zvol_set_volsize(dsname, intval);
2364                 break;
2365         case ZFS_PROP_VERSION:
2366         {
2367                 zfsvfs_t *zfsvfs;
2368 
2369                 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2370                         break;
2371 
2372                 err = zfs_set_version(zfsvfs, intval);
2373                 zfsvfs_rele(zfsvfs, FTAG);
2374 
2375                 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2376                         zfs_cmd_t *zc;
2377 
2378                         zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2379                         (void) strcpy(zc->zc_name, dsname);
2380                         (void) zfs_ioc_userspace_upgrade(zc);
2381                         kmem_free(zc, sizeof (zfs_cmd_t));
2382                 }
2383                 break;
2384         }
2385         case ZFS_PROP_COMPRESSION:
2386         {
2387                 if (intval == ZIO_COMPRESS_LZ4) {
2388                         zfeature_info_t *feature =
2389                             &spa_feature_table[SPA_FEATURE_LZ4_COMPRESS];
2390                         spa_t *spa;
2391 
2392                         if ((err = spa_open(dsname, &spa, FTAG)) != 0)
2393                                 return (err);
2394 
2395                         /*
2396                          * Setting the LZ4 compression algorithm activates
2397                          * the feature.
2398                          */
2399                         if (!spa_feature_is_active(spa, feature)) {
2400                                 if ((err = zfs_prop_activate_feature(spa,
2401                                     feature)) != 0) {
2402                                         spa_close(spa, FTAG);
2403                                         return (err);
2404                                 }
2405                         }
2406 
2407                         spa_close(spa, FTAG);
2408                 }
2409                 /*
2410                  * We still want the default set action to be performed in the
2411                  * caller, we only performed zfeature settings here.
2412                  */
2413                 err = -1;
2414                 break;
2415         }
2416 
2417         default:
2418                 err = -1;
2419         }
2420 
2421         return (err);
2422 }
2423 
2424 /*
2425  * This function is best effort. If it fails to set any of the given properties,
2426  * it continues to set as many as it can and returns the last error
2427  * encountered. If the caller provides a non-NULL errlist, it will be filled in
2428  * with the list of names of all the properties that failed along with the
2429  * corresponding error numbers.
2430  *
2431  * If every property is set successfully, zero is returned and errlist is not
2432  * modified.
2433  */
2434 int
2435 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2436     nvlist_t *errlist)
2437 {
2438         nvpair_t *pair;
2439         nvpair_t *propval;
2440         int rv = 0;
2441         uint64_t intval;
2442         char *strval;
2443         nvlist_t *genericnvl = fnvlist_alloc();
2444         nvlist_t *retrynvl = fnvlist_alloc();
2445 
2446 retry:
2447         pair = NULL;
2448         while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2449                 const char *propname = nvpair_name(pair);
2450                 zfs_prop_t prop = zfs_name_to_prop(propname);
2451                 int err = 0;
2452 
2453                 /* decode the property value */
2454                 propval = pair;
2455                 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2456                         nvlist_t *attrs;
2457                         attrs = fnvpair_value_nvlist(pair);
2458                         if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2459                             &propval) != 0)
2460                                 err = SET_ERROR(EINVAL);
2461                 }
2462 
2463                 /* Validate value type */
2464                 if (err == 0 && prop == ZPROP_INVAL) {
2465                         if (zfs_prop_user(propname)) {
2466                                 if (nvpair_type(propval) != DATA_TYPE_STRING)
2467                                         err = SET_ERROR(EINVAL);
2468                         } else if (zfs_prop_userquota(propname)) {
2469                                 if (nvpair_type(propval) !=
2470                                     DATA_TYPE_UINT64_ARRAY)
2471                                         err = SET_ERROR(EINVAL);
2472                         } else {
2473                                 err = SET_ERROR(EINVAL);
2474                         }
2475                 } else if (err == 0) {
2476                         if (nvpair_type(propval) == DATA_TYPE_STRING) {
2477                                 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2478                                         err = SET_ERROR(EINVAL);
2479                         } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2480                                 const char *unused;
2481 
2482                                 intval = fnvpair_value_uint64(propval);
2483 
2484                                 switch (zfs_prop_get_type(prop)) {
2485                                 case PROP_TYPE_NUMBER:
2486                                         break;
2487                                 case PROP_TYPE_STRING:
2488                                         err = SET_ERROR(EINVAL);
2489                                         break;
2490                                 case PROP_TYPE_INDEX:
2491                                         if (zfs_prop_index_to_string(prop,
2492                                             intval, &unused) != 0)
2493                                                 err = SET_ERROR(EINVAL);
2494                                         break;
2495                                 default:
2496                                         cmn_err(CE_PANIC,
2497                                             "unknown property type");
2498                                 }
2499                         } else {
2500                                 err = SET_ERROR(EINVAL);
2501                         }
2502                 }
2503 
2504                 /* Validate permissions */
2505                 if (err == 0)
2506                         err = zfs_check_settable(dsname, pair, CRED());
2507 
2508                 if (err == 0) {
2509                         err = zfs_prop_set_special(dsname, source, pair);
2510                         if (err == -1) {
2511                                 /*
2512                                  * For better performance we build up a list of
2513                                  * properties to set in a single transaction.
2514                                  */
2515                                 err = nvlist_add_nvpair(genericnvl, pair);
2516                         } else if (err != 0 && nvl != retrynvl) {
2517                                 /*
2518                                  * This may be a spurious error caused by
2519                                  * receiving quota and reservation out of order.
2520                                  * Try again in a second pass.
2521                                  */
2522                                 err = nvlist_add_nvpair(retrynvl, pair);
2523                         }
2524                 }
2525 
2526                 if (err != 0) {
2527                         if (errlist != NULL)
2528                                 fnvlist_add_int32(errlist, propname, err);
2529                         rv = err;
2530                 }
2531         }
2532 
2533         if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2534                 nvl = retrynvl;
2535                 goto retry;
2536         }
2537 
2538         if (!nvlist_empty(genericnvl) &&
2539             dsl_props_set(dsname, source, genericnvl) != 0) {
2540                 /*
2541                  * If this fails, we still want to set as many properties as we
2542                  * can, so try setting them individually.
2543                  */
2544                 pair = NULL;
2545                 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2546                         const char *propname = nvpair_name(pair);
2547                         int err = 0;
2548 
2549                         propval = pair;
2550                         if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2551                                 nvlist_t *attrs;
2552                                 attrs = fnvpair_value_nvlist(pair);
2553                                 propval = fnvlist_lookup_nvpair(attrs,
2554                                     ZPROP_VALUE);
2555                         }
2556 
2557                         if (nvpair_type(propval) == DATA_TYPE_STRING) {
2558                                 strval = fnvpair_value_string(propval);
2559                                 err = dsl_prop_set_string(dsname, propname,
2560                                     source, strval);
2561                         } else {
2562                                 intval = fnvpair_value_uint64(propval);
2563                                 err = dsl_prop_set_int(dsname, propname, source,
2564                                     intval);
2565                         }
2566 
2567                         if (err != 0) {
2568                                 if (errlist != NULL) {
2569                                         fnvlist_add_int32(errlist, propname,
2570                                             err);
2571                                 }
2572                                 rv = err;
2573                         }
2574                 }
2575         }
2576         nvlist_free(genericnvl);
2577         nvlist_free(retrynvl);
2578 
2579         return (rv);
2580 }
2581 
2582 /*
2583  * Check that all the properties are valid user properties.
2584  */
2585 static int
2586 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2587 {
2588         nvpair_t *pair = NULL;
2589         int error = 0;
2590 
2591         while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2592                 const char *propname = nvpair_name(pair);
2593                 char *valstr;
2594 
2595                 if (!zfs_prop_user(propname) ||
2596                     nvpair_type(pair) != DATA_TYPE_STRING)
2597                         return (SET_ERROR(EINVAL));
2598 
2599                 if (error = zfs_secpolicy_write_perms(fsname,
2600                     ZFS_DELEG_PERM_USERPROP, CRED()))
2601                         return (error);
2602 
2603                 if (strlen(propname) >= ZAP_MAXNAMELEN)
2604                         return (SET_ERROR(ENAMETOOLONG));
2605 
2606                 VERIFY(nvpair_value_string(pair, &valstr) == 0);
2607                 if (strlen(valstr) >= ZAP_MAXVALUELEN)
2608                         return (E2BIG);
2609         }
2610         return (0);
2611 }
2612 
2613 static void
2614 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2615 {
2616         nvpair_t *pair;
2617 
2618         VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2619 
2620         pair = NULL;
2621         while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2622                 if (nvlist_exists(skipped, nvpair_name(pair)))
2623                         continue;
2624 
2625                 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2626         }
2627 }
2628 
2629 static int
2630 clear_received_props(const char *dsname, nvlist_t *props,
2631     nvlist_t *skipped)
2632 {
2633         int err = 0;
2634         nvlist_t *cleared_props = NULL;
2635         props_skip(props, skipped, &cleared_props);
2636         if (!nvlist_empty(cleared_props)) {
2637                 /*
2638                  * Acts on local properties until the dataset has received
2639                  * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2640                  */
2641                 zprop_source_t flags = (ZPROP_SRC_NONE |
2642                     (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2643                 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2644         }
2645         nvlist_free(cleared_props);
2646         return (err);
2647 }
2648 
2649 /*
2650  * inputs:
2651  * zc_name              name of filesystem
2652  * zc_value             name of property to set
2653  * zc_nvlist_src{_size} nvlist of properties to apply
2654  * zc_cookie            received properties flag
2655  *
2656  * outputs:
2657  * zc_nvlist_dst{_size} error for each unapplied received property
2658  */
2659 static int
2660 zfs_ioc_set_prop(zfs_cmd_t *zc)
2661 {
2662         nvlist_t *nvl;
2663         boolean_t received = zc->zc_cookie;
2664         zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2665             ZPROP_SRC_LOCAL);
2666         nvlist_t *errors;
2667         int error;
2668 
2669         if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2670             zc->zc_iflags, &nvl)) != 0)
2671                 return (error);
2672 
2673         if (received) {
2674                 nvlist_t *origprops;
2675 
2676                 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2677                         (void) clear_received_props(zc->zc_name,
2678                             origprops, nvl);
2679                         nvlist_free(origprops);
2680                 }
2681 
2682                 error = dsl_prop_set_hasrecvd(zc->zc_name);
2683         }
2684 
2685         errors = fnvlist_alloc();
2686         if (error == 0)
2687                 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2688 
2689         if (zc->zc_nvlist_dst != NULL && errors != NULL) {
2690                 (void) put_nvlist(zc, errors);
2691         }
2692 
2693         nvlist_free(errors);
2694         nvlist_free(nvl);
2695         return (error);
2696 }
2697 
2698 /*
2699  * inputs:
2700  * zc_name              name of filesystem
2701  * zc_value             name of property to inherit
2702  * zc_cookie            revert to received value if TRUE
2703  *
2704  * outputs:             none
2705  */
2706 static int
2707 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2708 {
2709         const char *propname = zc->zc_value;
2710         zfs_prop_t prop = zfs_name_to_prop(propname);
2711         boolean_t received = zc->zc_cookie;
2712         zprop_source_t source = (received
2713             ? ZPROP_SRC_NONE            /* revert to received value, if any */
2714             : ZPROP_SRC_INHERITED);     /* explicitly inherit */
2715 
2716         if (received) {
2717                 nvlist_t *dummy;
2718                 nvpair_t *pair;
2719                 zprop_type_t type;
2720                 int err;
2721 
2722                 /*
2723                  * zfs_prop_set_special() expects properties in the form of an
2724                  * nvpair with type info.
2725                  */
2726                 if (prop == ZPROP_INVAL) {
2727                         if (!zfs_prop_user(propname))
2728                                 return (SET_ERROR(EINVAL));
2729 
2730                         type = PROP_TYPE_STRING;
2731                 } else if (prop == ZFS_PROP_VOLSIZE ||
2732                     prop == ZFS_PROP_VERSION) {
2733                         return (SET_ERROR(EINVAL));
2734                 } else {
2735                         type = zfs_prop_get_type(prop);
2736                 }
2737 
2738                 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2739 
2740                 switch (type) {
2741                 case PROP_TYPE_STRING:
2742                         VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2743                         break;
2744                 case PROP_TYPE_NUMBER:
2745                 case PROP_TYPE_INDEX:
2746                         VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2747                         break;
2748                 default:
2749                         nvlist_free(dummy);
2750                         return (SET_ERROR(EINVAL));
2751                 }
2752 
2753                 pair = nvlist_next_nvpair(dummy, NULL);
2754                 err = zfs_prop_set_special(zc->zc_name, source, pair);
2755                 nvlist_free(dummy);
2756                 if (err != -1)
2757                         return (err); /* special property already handled */
2758         } else {
2759                 /*
2760                  * Only check this in the non-received case. We want to allow
2761                  * 'inherit -S' to revert non-inheritable properties like quota
2762                  * and reservation to the received or default values even though
2763                  * they are not considered inheritable.
2764                  */
2765                 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2766                         return (SET_ERROR(EINVAL));
2767         }
2768 
2769         /* property name has been validated by zfs_secpolicy_inherit_prop() */
2770         return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2771 }
2772 
2773 static int
2774 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2775 {
2776         nvlist_t *props;
2777         spa_t *spa;
2778         int error;
2779         nvpair_t *pair;
2780 
2781         if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2782             zc->zc_iflags, &props))
2783                 return (error);
2784 
2785         /*
2786          * If the only property is the configfile, then just do a spa_lookup()
2787          * to handle the faulted case.
2788          */
2789         pair = nvlist_next_nvpair(props, NULL);
2790         if (pair != NULL && strcmp(nvpair_name(pair),
2791             zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2792             nvlist_next_nvpair(props, pair) == NULL) {
2793                 mutex_enter(&spa_namespace_lock);
2794                 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2795                         spa_configfile_set(spa, props, B_FALSE);
2796                         spa_config_sync(spa, B_FALSE, B_TRUE);
2797                 }
2798                 mutex_exit(&spa_namespace_lock);
2799                 if (spa != NULL) {
2800                         nvlist_free(props);
2801                         return (0);
2802                 }
2803         }
2804 
2805         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2806                 nvlist_free(props);
2807                 return (error);
2808         }
2809 
2810         error = spa_prop_set(spa, props);
2811 
2812         nvlist_free(props);
2813         spa_close(spa, FTAG);
2814 
2815         return (error);
2816 }
2817 
2818 static int
2819 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2820 {
2821         spa_t *spa;
2822         int error;
2823         nvlist_t *nvp = NULL;
2824 
2825         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2826                 /*
2827                  * If the pool is faulted, there may be properties we can still
2828                  * get (such as altroot and cachefile), so attempt to get them
2829                  * anyway.
2830                  */
2831                 mutex_enter(&spa_namespace_lock);
2832                 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2833                         error = spa_prop_get(spa, &nvp);
2834                 mutex_exit(&spa_namespace_lock);
2835         } else {
2836                 error = spa_prop_get(spa, &nvp);
2837                 spa_close(spa, FTAG);
2838         }
2839 
2840         if (error == 0 && zc->zc_nvlist_dst != NULL)
2841                 error = put_nvlist(zc, nvp);
2842         else
2843                 error = SET_ERROR(EFAULT);
2844 
2845         nvlist_free(nvp);
2846         return (error);
2847 }
2848 
2849 /*
2850  * inputs:
2851  * zc_name              name of filesystem
2852  * zc_nvlist_src{_size} nvlist of delegated permissions
2853  * zc_perm_action       allow/unallow flag
2854  *
2855  * outputs:             none
2856  */
2857 static int
2858 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2859 {
2860         int error;
2861         nvlist_t *fsaclnv = NULL;
2862 
2863         if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2864             zc->zc_iflags, &fsaclnv)) != 0)
2865                 return (error);
2866 
2867         /*
2868          * Verify nvlist is constructed correctly
2869          */
2870         if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2871                 nvlist_free(fsaclnv);
2872                 return (SET_ERROR(EINVAL));
2873         }
2874 
2875         /*
2876          * If we don't have PRIV_SYS_MOUNT, then validate
2877          * that user is allowed to hand out each permission in
2878          * the nvlist(s)
2879          */
2880 
2881         error = secpolicy_zfs(CRED());
2882         if (error != 0) {
2883                 if (zc->zc_perm_action == B_FALSE) {
2884                         error = dsl_deleg_can_allow(zc->zc_name,
2885                             fsaclnv, CRED());
2886                 } else {
2887                         error = dsl_deleg_can_unallow(zc->zc_name,
2888                             fsaclnv, CRED());
2889                 }
2890         }
2891 
2892         if (error == 0)
2893                 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2894 
2895         nvlist_free(fsaclnv);
2896         return (error);
2897 }
2898 
2899 /*
2900  * inputs:
2901  * zc_name              name of filesystem
2902  *
2903  * outputs:
2904  * zc_nvlist_src{_size} nvlist of delegated permissions
2905  */
2906 static int
2907 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2908 {
2909         nvlist_t *nvp;
2910         int error;
2911 
2912         if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2913                 error = put_nvlist(zc, nvp);
2914                 nvlist_free(nvp);
2915         }
2916 
2917         return (error);
2918 }
2919 
2920 /*
2921  * Search the vfs list for a specified resource.  Returns a pointer to it
2922  * or NULL if no suitable entry is found. The caller of this routine
2923  * is responsible for releasing the returned vfs pointer.
2924  */
2925 static vfs_t *
2926 zfs_get_vfs(const char *resource)
2927 {
2928         struct vfs *vfsp;
2929         struct vfs *vfs_found = NULL;
2930 
2931         vfs_list_read_lock();
2932         vfsp = rootvfs;
2933         do {
2934                 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
2935                         VFS_HOLD(vfsp);
2936                         vfs_found = vfsp;
2937                         break;
2938                 }
2939                 vfsp = vfsp->vfs_next;
2940         } while (vfsp != rootvfs);
2941         vfs_list_unlock();
2942         return (vfs_found);
2943 }
2944 
2945 /* ARGSUSED */
2946 static void
2947 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
2948 {
2949         zfs_creat_t *zct = arg;
2950 
2951         zfs_create_fs(os, cr, zct->zct_zplprops, tx);
2952 }
2953 
2954 #define ZFS_PROP_UNDEFINED      ((uint64_t)-1)
2955 
2956 /*
2957  * inputs:
2958  * createprops          list of properties requested by creator
2959  * default_zplver       zpl version to use if unspecified in createprops
2960  * fuids_ok             fuids allowed in this version of the spa?
2961  * os                   parent objset pointer (NULL if root fs)
2962  *
2963  * outputs:
2964  * zplprops     values for the zplprops we attach to the master node object
2965  * is_ci        true if requested file system will be purely case-insensitive
2966  *
2967  * Determine the settings for utf8only, normalization and
2968  * casesensitivity.  Specific values may have been requested by the
2969  * creator and/or we can inherit values from the parent dataset.  If
2970  * the file system is of too early a vintage, a creator can not
2971  * request settings for these properties, even if the requested
2972  * setting is the default value.  We don't actually want to create dsl
2973  * properties for these, so remove them from the source nvlist after
2974  * processing.
2975  */
2976 static int
2977 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
2978     boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
2979     nvlist_t *zplprops, boolean_t *is_ci)
2980 {
2981         uint64_t sense = ZFS_PROP_UNDEFINED;
2982         uint64_t norm = ZFS_PROP_UNDEFINED;
2983         uint64_t u8 = ZFS_PROP_UNDEFINED;
2984 
2985         ASSERT(zplprops != NULL);
2986 
2987         /*
2988          * Pull out creator prop choices, if any.
2989          */
2990         if (createprops) {
2991                 (void) nvlist_lookup_uint64(createprops,
2992                     zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
2993                 (void) nvlist_lookup_uint64(createprops,
2994                     zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
2995                 (void) nvlist_remove_all(createprops,
2996                     zfs_prop_to_name(ZFS_PROP_NORMALIZE));
2997                 (void) nvlist_lookup_uint64(createprops,
2998                     zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
2999                 (void) nvlist_remove_all(createprops,
3000                     zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3001                 (void) nvlist_lookup_uint64(createprops,
3002                     zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3003                 (void) nvlist_remove_all(createprops,
3004                     zfs_prop_to_name(ZFS_PROP_CASE));
3005         }
3006 
3007         /*
3008          * If the zpl version requested is whacky or the file system
3009          * or pool is version is too "young" to support normalization
3010          * and the creator tried to set a value for one of the props,
3011          * error out.
3012          */
3013         if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3014             (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3015             (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3016             (zplver < ZPL_VERSION_NORMALIZATION &&
3017             (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3018             sense != ZFS_PROP_UNDEFINED)))
3019                 return (SET_ERROR(ENOTSUP));
3020 
3021         /*
3022          * Put the version in the zplprops
3023          */
3024         VERIFY(nvlist_add_uint64(zplprops,
3025             zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3026 
3027         if (norm == ZFS_PROP_UNDEFINED)
3028                 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3029         VERIFY(nvlist_add_uint64(zplprops,
3030             zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3031 
3032         /*
3033          * If we're normalizing, names must always be valid UTF-8 strings.
3034          */
3035         if (norm)
3036                 u8 = 1;
3037         if (u8 == ZFS_PROP_UNDEFINED)
3038                 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3039         VERIFY(nvlist_add_uint64(zplprops,
3040             zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3041 
3042         if (sense == ZFS_PROP_UNDEFINED)
3043                 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3044         VERIFY(nvlist_add_uint64(zplprops,
3045             zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3046 
3047         if (is_ci)
3048                 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3049 
3050         return (0);
3051 }
3052 
3053 static int
3054 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3055     nvlist_t *zplprops, boolean_t *is_ci)
3056 {
3057         boolean_t fuids_ok, sa_ok;
3058         uint64_t zplver = ZPL_VERSION;
3059         objset_t *os = NULL;
3060         char parentname[MAXNAMELEN];
3061         char *cp;
3062         spa_t *spa;
3063         uint64_t spa_vers;
3064         int error;
3065 
3066         (void) strlcpy(parentname, dataset, sizeof (parentname));
3067         cp = strrchr(parentname, '/');
3068         ASSERT(cp != NULL);
3069         cp[0] = '\0';
3070 
3071         if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3072                 return (error);
3073 
3074         spa_vers = spa_version(spa);
3075         spa_close(spa, FTAG);
3076 
3077         zplver = zfs_zpl_version_map(spa_vers);
3078         fuids_ok = (zplver >= ZPL_VERSION_FUID);
3079         sa_ok = (zplver >= ZPL_VERSION_SA);
3080 
3081         /*
3082          * Open parent object set so we can inherit zplprop values.
3083          */
3084         if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3085                 return (error);
3086 
3087         error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3088             zplprops, is_ci);
3089         dmu_objset_rele(os, FTAG);
3090         return (error);
3091 }
3092 
3093 static int
3094 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3095     nvlist_t *zplprops, boolean_t *is_ci)
3096 {
3097         boolean_t fuids_ok;
3098         boolean_t sa_ok;
3099         uint64_t zplver = ZPL_VERSION;
3100         int error;
3101 
3102         zplver = zfs_zpl_version_map(spa_vers);
3103         fuids_ok = (zplver >= ZPL_VERSION_FUID);
3104         sa_ok = (zplver >= ZPL_VERSION_SA);
3105 
3106         error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3107             createprops, zplprops, is_ci);
3108         return (error);
3109 }
3110 
3111 /*
3112  * innvl: {
3113  *     "type" -> dmu_objset_type_t (int32)
3114  *     (optional) "props" -> { prop -> value }
3115  * }
3116  *
3117  * outnvl: propname -> error code (int32)
3118  */
3119 static int
3120 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3121 {
3122         int error = 0;
3123         zfs_creat_t zct = { 0 };
3124         nvlist_t *nvprops = NULL;
3125         void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3126         int32_t type32;
3127         dmu_objset_type_t type;
3128         boolean_t is_insensitive = B_FALSE;
3129 
3130         if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3131                 return (SET_ERROR(EINVAL));
3132         type = type32;
3133         (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3134 
3135         switch (type) {
3136         case DMU_OST_ZFS:
3137                 cbfunc = zfs_create_cb;
3138                 break;
3139 
3140         case DMU_OST_ZVOL:
3141                 cbfunc = zvol_create_cb;
3142                 break;
3143 
3144         default:
3145                 cbfunc = NULL;
3146                 break;
3147         }
3148         if (strchr(fsname, '@') ||
3149             strchr(fsname, '%'))
3150                 return (SET_ERROR(EINVAL));
3151 
3152         zct.zct_props = nvprops;
3153 
3154         if (cbfunc == NULL)
3155                 return (SET_ERROR(EINVAL));
3156 
3157         if (type == DMU_OST_ZVOL) {
3158                 uint64_t volsize, volblocksize;
3159 
3160                 if (nvprops == NULL)
3161                         return (SET_ERROR(EINVAL));
3162                 if (nvlist_lookup_uint64(nvprops,
3163                     zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3164                         return (SET_ERROR(EINVAL));
3165 
3166                 if ((error = nvlist_lookup_uint64(nvprops,
3167                     zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3168                     &volblocksize)) != 0 && error != ENOENT)
3169                         return (SET_ERROR(EINVAL));
3170 
3171                 if (error != 0)
3172                         volblocksize = zfs_prop_default_numeric(
3173                             ZFS_PROP_VOLBLOCKSIZE);
3174 
3175                 if ((error = zvol_check_volblocksize(
3176                     volblocksize)) != 0 ||
3177                     (error = zvol_check_volsize(volsize,
3178                     volblocksize)) != 0)
3179                         return (error);
3180         } else if (type == DMU_OST_ZFS) {
3181                 int error;
3182 
3183                 /*
3184                  * We have to have normalization and
3185                  * case-folding flags correct when we do the
3186                  * file system creation, so go figure them out
3187                  * now.
3188                  */
3189                 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3190                     NV_UNIQUE_NAME, KM_SLEEP) == 0);
3191                 error = zfs_fill_zplprops(fsname, nvprops,
3192                     zct.zct_zplprops, &is_insensitive);
3193                 if (error != 0) {
3194                         nvlist_free(zct.zct_zplprops);
3195                         return (error);
3196                 }
3197         }
3198 
3199         error = dmu_objset_create(fsname, type,
3200             is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3201         nvlist_free(zct.zct_zplprops);
3202 
3203         /*
3204          * It would be nice to do this atomically.
3205          */
3206         if (error == 0) {
3207                 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3208                     nvprops, outnvl);
3209                 if (error != 0)
3210                         (void) dsl_destroy_head(fsname);
3211         }
3212         return (error);
3213 }
3214 
3215 /*
3216  * innvl: {
3217  *     "origin" -> name of origin snapshot
3218  *     (optional) "props" -> { prop -> value }
3219  * }
3220  *
3221  * outnvl: propname -> error code (int32)
3222  */
3223 static int
3224 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3225 {
3226         int error = 0;
3227         nvlist_t *nvprops = NULL;
3228         char *origin_name;
3229 
3230         if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3231                 return (SET_ERROR(EINVAL));
3232         (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3233 
3234         if (strchr(fsname, '@') ||
3235             strchr(fsname, '%'))
3236                 return (SET_ERROR(EINVAL));
3237 
3238         if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3239                 return (SET_ERROR(EINVAL));
3240         error = dmu_objset_clone(fsname, origin_name);
3241         if (error != 0)
3242                 return (error);
3243 
3244         /*
3245          * It would be nice to do this atomically.
3246          */
3247         if (error == 0) {
3248                 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3249                     nvprops, outnvl);
3250                 if (error != 0)
3251                         (void) dsl_destroy_head(fsname);
3252         }
3253         return (error);
3254 }
3255 
3256 /*
3257  * innvl: {
3258  *     "snaps" -> { snapshot1, snapshot2 }
3259  *     (optional) "props" -> { prop -> value (string) }
3260  * }
3261  *
3262  * outnvl: snapshot -> error code (int32)
3263  */
3264 static int
3265 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3266 {
3267         nvlist_t *snaps;
3268         nvlist_t *props = NULL;
3269         int error, poollen;
3270         nvpair_t *pair;
3271 
3272         (void) nvlist_lookup_nvlist(innvl, "props", &props);
3273         if ((error = zfs_check_userprops(poolname, props)) != 0)
3274                 return (error);
3275 
3276         if (!nvlist_empty(props) &&
3277             zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3278                 return (SET_ERROR(ENOTSUP));
3279 
3280         if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3281                 return (SET_ERROR(EINVAL));
3282         poollen = strlen(poolname);
3283         for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3284             pair = nvlist_next_nvpair(snaps, pair)) {
3285                 const char *name = nvpair_name(pair);
3286                 const char *cp = strchr(name, '@');
3287 
3288                 /*
3289                  * The snap name must contain an @, and the part after it must
3290                  * contain only valid characters.
3291                  */
3292                 if (cp == NULL || snapshot_namecheck(cp + 1, NULL, NULL) != 0)
3293                         return (SET_ERROR(EINVAL));
3294 
3295                 /*
3296                  * The snap must be in the specified pool.
3297                  */
3298                 if (strncmp(name, poolname, poollen) != 0 ||
3299                     (name[poollen] != '/' && name[poollen] != '@'))
3300                         return (SET_ERROR(EXDEV));
3301 
3302                 /* This must be the only snap of this fs. */
3303                 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3304                     pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3305                         if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3306                             == 0) {
3307                                 return (SET_ERROR(EXDEV));
3308                         }
3309                 }
3310         }
3311 
3312         error = dsl_dataset_snapshot(snaps, props, outnvl);
3313         return (error);
3314 }
3315 
3316 /*
3317  * innvl: "message" -> string
3318  */
3319 /* ARGSUSED */
3320 static int
3321 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3322 {
3323         char *message;
3324         spa_t *spa;
3325         int error;
3326         char *poolname;
3327 
3328         /*
3329          * The poolname in the ioctl is not set, we get it from the TSD,
3330          * which was set at the end of the last successful ioctl that allows
3331          * logging.  The secpolicy func already checked that it is set.
3332          * Only one log ioctl is allowed after each successful ioctl, so
3333          * we clear the TSD here.
3334          */
3335         poolname = tsd_get(zfs_allow_log_key);
3336         (void) tsd_set(zfs_allow_log_key, NULL);
3337         error = spa_open(poolname, &spa, FTAG);
3338         strfree(poolname);
3339         if (error != 0)
3340                 return (error);
3341 
3342         if (nvlist_lookup_string(innvl, "message", &message) != 0)  {
3343                 spa_close(spa, FTAG);
3344                 return (SET_ERROR(EINVAL));
3345         }
3346 
3347         if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3348                 spa_close(spa, FTAG);
3349                 return (SET_ERROR(ENOTSUP));
3350         }
3351 
3352         error = spa_history_log(spa, message);
3353         spa_close(spa, FTAG);
3354         return (error);
3355 }
3356 
3357 /*
3358  * The dp_config_rwlock must not be held when calling this, because the
3359  * unmount may need to write out data.
3360  *
3361  * This function is best-effort.  Callers must deal gracefully if it
3362  * remains mounted (or is remounted after this call).
3363  */
3364 void
3365 zfs_unmount_snap(const char *snapname)
3366 {
3367         vfs_t *vfsp;
3368         zfsvfs_t *zfsvfs;
3369 
3370         if (strchr(snapname, '@') == NULL)
3371                 return;
3372 
3373         vfsp = zfs_get_vfs(snapname);
3374         if (vfsp == NULL)
3375                 return;
3376 
3377         zfsvfs = vfsp->vfs_data;
3378         ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os)));
3379 
3380         if (vn_vfswlock(vfsp->vfs_vnodecovered) != 0) {
3381                 VFS_RELE(vfsp);
3382                 return;
3383         }
3384         VFS_RELE(vfsp);
3385 
3386         /*
3387          * Always force the unmount for snapshots.
3388          */
3389         (void) dounmount(vfsp, MS_FORCE, kcred);
3390 }
3391 
3392 /* ARGSUSED */
3393 static int
3394 zfs_unmount_snap_cb(const char *snapname, void *arg)
3395 {
3396         zfs_unmount_snap(snapname);
3397         return (0);
3398 }
3399 
3400 /*
3401  * When a clone is destroyed, its origin may also need to be destroyed,
3402  * in which case it must be unmounted.  This routine will do that unmount
3403  * if necessary.
3404  */
3405 void
3406 zfs_destroy_unmount_origin(const char *fsname)
3407 {
3408         int error;
3409         objset_t *os;
3410         dsl_dataset_t *ds;
3411 
3412         error = dmu_objset_hold(fsname, FTAG, &os);
3413         if (error != 0)
3414                 return;
3415         ds = dmu_objset_ds(os);
3416         if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3417                 char originname[MAXNAMELEN];
3418                 dsl_dataset_name(ds->ds_prev, originname);
3419                 dmu_objset_rele(os, FTAG);
3420                 zfs_unmount_snap(originname);
3421         } else {
3422                 dmu_objset_rele(os, FTAG);
3423         }
3424 }
3425 
3426 /*
3427  * innvl: {
3428  *     "snaps" -> { snapshot1, snapshot2 }
3429  *     (optional boolean) "defer"
3430  * }
3431  *
3432  * outnvl: snapshot -> error code (int32)
3433  *
3434  */
3435 static int
3436 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3437 {
3438         int poollen;
3439         nvlist_t *snaps;
3440         nvpair_t *pair;
3441         boolean_t defer;
3442 
3443         if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3444                 return (SET_ERROR(EINVAL));
3445         defer = nvlist_exists(innvl, "defer");
3446 
3447         poollen = strlen(poolname);
3448         for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3449             pair = nvlist_next_nvpair(snaps, pair)) {
3450                 const char *name = nvpair_name(pair);
3451 
3452                 /*
3453                  * The snap must be in the specified pool.
3454                  */
3455                 if (strncmp(name, poolname, poollen) != 0 ||
3456                     (name[poollen] != '/' && name[poollen] != '@'))
3457                         return (SET_ERROR(EXDEV));
3458 
3459                 zfs_unmount_snap(name);
3460         }
3461 
3462         return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3463 }
3464 
3465 /*
3466  * inputs:
3467  * zc_name              name of dataset to destroy
3468  * zc_objset_type       type of objset
3469  * zc_defer_destroy     mark for deferred destroy
3470  *
3471  * outputs:             none
3472  */
3473 static int
3474 zfs_ioc_destroy(zfs_cmd_t *zc)
3475 {
3476         int err;
3477         if (strchr(zc->zc_name, '@') && zc->zc_objset_type == DMU_OST_ZFS)
3478                 zfs_unmount_snap(zc->zc_name);
3479 
3480         if (strchr(zc->zc_name, '@'))
3481                 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3482         else
3483                 err = dsl_destroy_head(zc->zc_name);
3484         if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3485                 (void) zvol_remove_minor(zc->zc_name);
3486         return (err);
3487 }
3488 
3489 /*
3490  * inputs:
3491  * zc_name      name of dataset to rollback (to most recent snapshot)
3492  *
3493  * outputs:     none
3494  */
3495 static int
3496 zfs_ioc_rollback(zfs_cmd_t *zc)
3497 {
3498         zfsvfs_t *zfsvfs;
3499         int error;
3500 
3501         if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
3502                 error = zfs_suspend_fs(zfsvfs);
3503                 if (error == 0) {
3504                         int resume_err;
3505 
3506                         error = dsl_dataset_rollback(zc->zc_name);
3507                         resume_err = zfs_resume_fs(zfsvfs, zc->zc_name);
3508                         error = error ? error : resume_err;
3509                 }
3510                 VFS_RELE(zfsvfs->z_vfs);
3511         } else {
3512                 error = dsl_dataset_rollback(zc->zc_name);
3513         }
3514         return (error);
3515 }
3516 
3517 static int
3518 recursive_unmount(const char *fsname, void *arg)
3519 {
3520         const char *snapname = arg;
3521         char fullname[MAXNAMELEN];
3522 
3523         (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname);
3524         zfs_unmount_snap(fullname);
3525         return (0);
3526 }
3527 
3528 /*
3529  * inputs:
3530  * zc_name      old name of dataset
3531  * zc_value     new name of dataset
3532  * zc_cookie    recursive flag (only valid for snapshots)
3533  *
3534  * outputs:     none
3535  */
3536 static int
3537 zfs_ioc_rename(zfs_cmd_t *zc)
3538 {
3539         boolean_t recursive = zc->zc_cookie & 1;
3540         char *at;
3541 
3542         zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3543         if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3544             strchr(zc->zc_value, '%'))
3545                 return (SET_ERROR(EINVAL));
3546 
3547         at = strchr(zc->zc_name, '@');
3548         if (at != NULL) {
3549                 /* snaps must be in same fs */
3550                 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3551                         return (SET_ERROR(EXDEV));
3552                 *at = '\0';
3553                 if (zc->zc_objset_type == DMU_OST_ZFS) {
3554                         int error = dmu_objset_find(zc->zc_name,
3555                             recursive_unmount, at + 1,
3556                             recursive ? DS_FIND_CHILDREN : 0);
3557                         if (error != 0)
3558                                 return (error);
3559                 }
3560                 return (dsl_dataset_rename_snapshot(zc->zc_name,
3561                     at + 1, strchr(zc->zc_value, '@') + 1, recursive));
3562         } else {
3563                 if (zc->zc_objset_type == DMU_OST_ZVOL)
3564                         (void) zvol_remove_minor(zc->zc_name);
3565                 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3566         }
3567 }
3568 
3569 static int
3570 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3571 {
3572         const char *propname = nvpair_name(pair);
3573         boolean_t issnap = (strchr(dsname, '@') != NULL);
3574         zfs_prop_t prop = zfs_name_to_prop(propname);
3575         uint64_t intval;
3576         int err;
3577 
3578         if (prop == ZPROP_INVAL) {
3579                 if (zfs_prop_user(propname)) {
3580                         if (err = zfs_secpolicy_write_perms(dsname,
3581                             ZFS_DELEG_PERM_USERPROP, cr))
3582                                 return (err);
3583                         return (0);
3584                 }
3585 
3586                 if (!issnap && zfs_prop_userquota(propname)) {
3587                         const char *perm = NULL;
3588                         const char *uq_prefix =
3589                             zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3590                         const char *gq_prefix =
3591                             zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3592 
3593                         if (strncmp(propname, uq_prefix,
3594                             strlen(uq_prefix)) == 0) {
3595                                 perm = ZFS_DELEG_PERM_USERQUOTA;
3596                         } else if (strncmp(propname, gq_prefix,
3597                             strlen(gq_prefix)) == 0) {
3598                                 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3599                         } else {
3600                                 /* USERUSED and GROUPUSED are read-only */
3601                                 return (SET_ERROR(EINVAL));
3602                         }
3603 
3604                         if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3605                                 return (err);
3606                         return (0);
3607                 }
3608 
3609                 return (SET_ERROR(EINVAL));
3610         }
3611 
3612         if (issnap)
3613                 return (SET_ERROR(EINVAL));
3614 
3615         if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3616                 /*
3617                  * dsl_prop_get_all_impl() returns properties in this
3618                  * format.
3619                  */
3620                 nvlist_t *attrs;
3621                 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3622                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3623                     &pair) == 0);
3624         }
3625 
3626         /*
3627          * Check that this value is valid for this pool version
3628          */
3629         switch (prop) {
3630         case ZFS_PROP_COMPRESSION:
3631                 /*
3632                  * If the user specified gzip compression, make sure
3633                  * the SPA supports it. We ignore any errors here since
3634                  * we'll catch them later.
3635                  */
3636                 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3637                     nvpair_value_uint64(pair, &intval) == 0) {
3638                         if (intval >= ZIO_COMPRESS_GZIP_1 &&
3639                             intval <= ZIO_COMPRESS_GZIP_9 &&
3640                             zfs_earlier_version(dsname,
3641                             SPA_VERSION_GZIP_COMPRESSION)) {
3642                                 return (SET_ERROR(ENOTSUP));
3643                         }
3644 
3645                         if (intval == ZIO_COMPRESS_ZLE &&
3646                             zfs_earlier_version(dsname,
3647                             SPA_VERSION_ZLE_COMPRESSION))
3648                                 return (SET_ERROR(ENOTSUP));
3649 
3650                         if (intval == ZIO_COMPRESS_LZ4) {
3651                                 zfeature_info_t *feature =
3652                                     &spa_feature_table[
3653                                     SPA_FEATURE_LZ4_COMPRESS];
3654                                 spa_t *spa;
3655 
3656                                 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3657                                         return (err);
3658 
3659                                 if (!spa_feature_is_enabled(spa, feature)) {
3660                                         spa_close(spa, FTAG);
3661                                         return (SET_ERROR(ENOTSUP));
3662                                 }
3663                                 spa_close(spa, FTAG);
3664                         }
3665 
3666                         /*
3667                          * If this is a bootable dataset then
3668                          * verify that the compression algorithm
3669                          * is supported for booting. We must return
3670                          * something other than ENOTSUP since it
3671                          * implies a downrev pool version.
3672                          */
3673                         if (zfs_is_bootfs(dsname) &&
3674                             !BOOTFS_COMPRESS_VALID(intval)) {
3675                                 return (SET_ERROR(ERANGE));
3676                         }
3677                 }
3678                 break;
3679 
3680         case ZFS_PROP_COPIES:
3681                 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3682                         return (SET_ERROR(ENOTSUP));
3683                 break;
3684 
3685         case ZFS_PROP_DEDUP:
3686                 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3687                         return (SET_ERROR(ENOTSUP));
3688                 break;
3689 
3690         case ZFS_PROP_SHARESMB:
3691                 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3692                         return (SET_ERROR(ENOTSUP));
3693                 break;
3694 
3695         case ZFS_PROP_ACLINHERIT:
3696                 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3697                     nvpair_value_uint64(pair, &intval) == 0) {
3698                         if (intval == ZFS_ACL_PASSTHROUGH_X &&
3699                             zfs_earlier_version(dsname,
3700                             SPA_VERSION_PASSTHROUGH_X))
3701                                 return (SET_ERROR(ENOTSUP));
3702                 }
3703                 break;
3704         }
3705 
3706         return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3707 }
3708 
3709 /*
3710  * Checks for a race condition to make sure we don't increment a feature flag
3711  * multiple times.
3712  */
3713 static int
3714 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
3715 {
3716         spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3717         zfeature_info_t *feature = arg;
3718 
3719         if (!spa_feature_is_active(spa, feature))
3720                 return (0);
3721         else
3722                 return (SET_ERROR(EBUSY));
3723 }
3724 
3725 /*
3726  * The callback invoked on feature activation in the sync task caused by
3727  * zfs_prop_activate_feature.
3728  */
3729 static void
3730 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
3731 {
3732         spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3733         zfeature_info_t *feature = arg;
3734 
3735         spa_feature_incr(spa, feature, tx);
3736 }
3737 
3738 /*
3739  * Activates a feature on a pool in response to a property setting. This
3740  * creates a new sync task which modifies the pool to reflect the feature
3741  * as being active.
3742  */
3743 static int
3744 zfs_prop_activate_feature(spa_t *spa, zfeature_info_t *feature)
3745 {
3746         int err;
3747 
3748         /* EBUSY here indicates that the feature is already active */
3749         err = dsl_sync_task(spa_name(spa),
3750             zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
3751             feature, 2);
3752 
3753         if (err != 0 && err != EBUSY)
3754                 return (err);
3755         else
3756                 return (0);
3757 }
3758 
3759 /*
3760  * Removes properties from the given props list that fail permission checks
3761  * needed to clear them and to restore them in case of a receive error. For each
3762  * property, make sure we have both set and inherit permissions.
3763  *
3764  * Returns the first error encountered if any permission checks fail. If the
3765  * caller provides a non-NULL errlist, it also gives the complete list of names
3766  * of all the properties that failed a permission check along with the
3767  * corresponding error numbers. The caller is responsible for freeing the
3768  * returned errlist.
3769  *
3770  * If every property checks out successfully, zero is returned and the list
3771  * pointed at by errlist is NULL.
3772  */
3773 static int
3774 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
3775 {
3776         zfs_cmd_t *zc;
3777         nvpair_t *pair, *next_pair;
3778         nvlist_t *errors;
3779         int err, rv = 0;
3780 
3781         if (props == NULL)
3782                 return (0);
3783 
3784         VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3785 
3786         zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
3787         (void) strcpy(zc->zc_name, dataset);
3788         pair = nvlist_next_nvpair(props, NULL);
3789         while (pair != NULL) {
3790                 next_pair = nvlist_next_nvpair(props, pair);
3791 
3792                 (void) strcpy(zc->zc_value, nvpair_name(pair));
3793                 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
3794                     (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
3795                         VERIFY(nvlist_remove_nvpair(props, pair) == 0);
3796                         VERIFY(nvlist_add_int32(errors,
3797                             zc->zc_value, err) == 0);
3798                 }
3799                 pair = next_pair;
3800         }
3801         kmem_free(zc, sizeof (zfs_cmd_t));
3802 
3803         if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
3804                 nvlist_free(errors);
3805                 errors = NULL;
3806         } else {
3807                 VERIFY(nvpair_value_int32(pair, &rv) == 0);
3808         }
3809 
3810         if (errlist == NULL)
3811                 nvlist_free(errors);
3812         else
3813                 *errlist = errors;
3814 
3815         return (rv);
3816 }
3817 
3818 static boolean_t
3819 propval_equals(nvpair_t *p1, nvpair_t *p2)
3820 {
3821         if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
3822                 /* dsl_prop_get_all_impl() format */
3823                 nvlist_t *attrs;
3824                 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
3825                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3826                     &p1) == 0);
3827         }
3828 
3829         if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
3830                 nvlist_t *attrs;
3831                 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
3832                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3833                     &p2) == 0);
3834         }
3835 
3836         if (nvpair_type(p1) != nvpair_type(p2))
3837                 return (B_FALSE);
3838 
3839         if (nvpair_type(p1) == DATA_TYPE_STRING) {
3840                 char *valstr1, *valstr2;
3841 
3842                 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
3843                 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
3844                 return (strcmp(valstr1, valstr2) == 0);
3845         } else {
3846                 uint64_t intval1, intval2;
3847 
3848                 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
3849                 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
3850                 return (intval1 == intval2);
3851         }
3852 }
3853 
3854 /*
3855  * Remove properties from props if they are not going to change (as determined
3856  * by comparison with origprops). Remove them from origprops as well, since we
3857  * do not need to clear or restore properties that won't change.
3858  */
3859 static void
3860 props_reduce(nvlist_t *props, nvlist_t *origprops)
3861 {
3862         nvpair_t *pair, *next_pair;
3863 
3864         if (origprops == NULL)
3865                 return; /* all props need to be received */
3866 
3867         pair = nvlist_next_nvpair(props, NULL);
3868         while (pair != NULL) {
3869                 const char *propname = nvpair_name(pair);
3870                 nvpair_t *match;
3871 
3872                 next_pair = nvlist_next_nvpair(props, pair);
3873 
3874                 if ((nvlist_lookup_nvpair(origprops, propname,
3875                     &match) != 0) || !propval_equals(pair, match))
3876                         goto next; /* need to set received value */
3877 
3878                 /* don't clear the existing received value */
3879                 (void) nvlist_remove_nvpair(origprops, match);
3880                 /* don't bother receiving the property */
3881                 (void) nvlist_remove_nvpair(props, pair);
3882 next:
3883                 pair = next_pair;
3884         }
3885 }
3886 
3887 #ifdef  DEBUG
3888 static boolean_t zfs_ioc_recv_inject_err;
3889 #endif
3890 
3891 /*
3892  * inputs:
3893  * zc_name              name of containing filesystem
3894  * zc_nvlist_src{_size} nvlist of properties to apply
3895  * zc_value             name of snapshot to create
3896  * zc_string            name of clone origin (if DRR_FLAG_CLONE)
3897  * zc_cookie            file descriptor to recv from
3898  * zc_begin_record      the BEGIN record of the stream (not byteswapped)
3899  * zc_guid              force flag
3900  * zc_cleanup_fd        cleanup-on-exit file descriptor
3901  * zc_action_handle     handle for this guid/ds mapping (or zero on first call)
3902  *
3903  * outputs:
3904  * zc_cookie            number of bytes read
3905  * zc_nvlist_dst{_size} error for each unapplied received property
3906  * zc_obj               zprop_errflags_t
3907  * zc_action_handle     handle for this guid/ds mapping
3908  */
3909 static int
3910 zfs_ioc_recv(zfs_cmd_t *zc)
3911 {
3912         file_t *fp;
3913         dmu_recv_cookie_t drc;
3914         boolean_t force = (boolean_t)zc->zc_guid;
3915         int fd;
3916         int error = 0;
3917         int props_error = 0;
3918         nvlist_t *errors;
3919         offset_t off;
3920         nvlist_t *props = NULL; /* sent properties */
3921         nvlist_t *origprops = NULL; /* existing properties */
3922         char *origin = NULL;
3923         char *tosnap;
3924         char tofs[ZFS_MAXNAMELEN];
3925         boolean_t first_recvd_props = B_FALSE;
3926 
3927         if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3928             strchr(zc->zc_value, '@') == NULL ||
3929             strchr(zc->zc_value, '%'))
3930                 return (SET_ERROR(EINVAL));
3931 
3932         (void) strcpy(tofs, zc->zc_value);
3933         tosnap = strchr(tofs, '@');
3934         *tosnap++ = '\0';
3935 
3936         if (zc->zc_nvlist_src != NULL &&
3937             (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3938             zc->zc_iflags, &props)) != 0)
3939                 return (error);
3940 
3941         fd = zc->zc_cookie;
3942         fp = getf(fd);
3943         if (fp == NULL) {
3944                 nvlist_free(props);
3945                 return (SET_ERROR(EBADF));
3946         }
3947 
3948         VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3949 
3950         if (zc->zc_string[0])
3951                 origin = zc->zc_string;
3952 
3953         error = dmu_recv_begin(tofs, tosnap,
3954             &zc->zc_begin_record, force, origin, &drc);
3955         if (error != 0)
3956                 goto out;
3957 
3958         /*
3959          * Set properties before we receive the stream so that they are applied
3960          * to the new data. Note that we must call dmu_recv_stream() if
3961          * dmu_recv_begin() succeeds.
3962          */
3963         if (props != NULL && !drc.drc_newfs) {
3964                 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
3965                     SPA_VERSION_RECVD_PROPS &&
3966                     !dsl_prop_get_hasrecvd(tofs))
3967                         first_recvd_props = B_TRUE;
3968 
3969                 /*
3970                  * If new received properties are supplied, they are to
3971                  * completely replace the existing received properties, so stash
3972                  * away the existing ones.
3973                  */
3974                 if (dsl_prop_get_received(tofs, &origprops) == 0) {
3975                         nvlist_t *errlist = NULL;
3976                         /*
3977                          * Don't bother writing a property if its value won't
3978                          * change (and avoid the unnecessary security checks).
3979                          *
3980                          * The first receive after SPA_VERSION_RECVD_PROPS is a
3981                          * special case where we blow away all local properties
3982                          * regardless.
3983                          */
3984                         if (!first_recvd_props)
3985                                 props_reduce(props, origprops);
3986                         if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
3987                                 (void) nvlist_merge(errors, errlist, 0);
3988                         nvlist_free(errlist);
3989 
3990                         if (clear_received_props(tofs, origprops,
3991                             first_recvd_props ? NULL : props) != 0)
3992                                 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3993                 } else {
3994                         zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3995                 }
3996         }
3997 
3998         if (props != NULL) {
3999                 props_error = dsl_prop_set_hasrecvd(tofs);
4000 
4001                 if (props_error == 0) {
4002                         (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4003                             props, errors);
4004                 }
4005         }
4006 
4007         if (zc->zc_nvlist_dst_size != 0 &&
4008             (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4009             put_nvlist(zc, errors) != 0)) {
4010                 /*
4011                  * Caller made zc->zc_nvlist_dst less than the minimum expected
4012                  * size or supplied an invalid address.
4013                  */
4014                 props_error = SET_ERROR(EINVAL);
4015         }
4016 
4017         off = fp->f_offset;
4018         error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
4019             &zc->zc_action_handle);
4020 
4021         if (error == 0) {
4022                 zfsvfs_t *zfsvfs = NULL;
4023 
4024                 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4025                         /* online recv */
4026                         int end_err;
4027 
4028                         error = zfs_suspend_fs(zfsvfs);
4029                         /*
4030                          * If the suspend fails, then the recv_end will
4031                          * likely also fail, and clean up after itself.
4032                          */
4033                         end_err = dmu_recv_end(&drc);
4034                         if (error == 0)
4035                                 error = zfs_resume_fs(zfsvfs, tofs);
4036                         error = error ? error : end_err;
4037                         VFS_RELE(zfsvfs->z_vfs);
4038                 } else {
4039                         error = dmu_recv_end(&drc);
4040                 }
4041         }
4042 
4043         zc->zc_cookie = off - fp->f_offset;
4044         if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4045                 fp->f_offset = off;
4046 
4047 #ifdef  DEBUG
4048         if (zfs_ioc_recv_inject_err) {
4049                 zfs_ioc_recv_inject_err = B_FALSE;
4050                 error = 1;
4051         }
4052 #endif
4053         /*
4054          * On error, restore the original props.
4055          */
4056         if (error != 0 && props != NULL && !drc.drc_newfs) {
4057                 if (clear_received_props(tofs, props, NULL) != 0) {
4058                         /*
4059                          * We failed to clear the received properties.
4060                          * Since we may have left a $recvd value on the
4061                          * system, we can't clear the $hasrecvd flag.
4062                          */
4063                         zc->zc_obj |= ZPROP_ERR_NORESTORE;
4064                 } else if (first_recvd_props) {
4065                         dsl_prop_unset_hasrecvd(tofs);
4066                 }
4067 
4068                 if (origprops == NULL && !drc.drc_newfs) {
4069                         /* We failed to stash the original properties. */
4070                         zc->zc_obj |= ZPROP_ERR_NORESTORE;
4071                 }
4072 
4073                 /*
4074                  * dsl_props_set() will not convert RECEIVED to LOCAL on or
4075                  * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4076                  * explictly if we're restoring local properties cleared in the
4077                  * first new-style receive.
4078                  */
4079                 if (origprops != NULL &&
4080                     zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4081                     ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4082                     origprops, NULL) != 0) {
4083                         /*
4084                          * We stashed the original properties but failed to
4085                          * restore them.
4086                          */
4087                         zc->zc_obj |= ZPROP_ERR_NORESTORE;
4088                 }
4089         }
4090 out:
4091         nvlist_free(props);
4092         nvlist_free(origprops);
4093         nvlist_free(errors);
4094         releasef(fd);
4095 
4096         if (error == 0)
4097                 error = props_error;
4098 
4099         return (error);
4100 }
4101 
4102 /*
4103  * inputs:
4104  * zc_name      name of snapshot to send
4105  * zc_cookie    file descriptor to send stream to
4106  * zc_obj       fromorigin flag (mutually exclusive with zc_fromobj)
4107  * zc_sendobj   objsetid of snapshot to send
4108  * zc_fromobj   objsetid of incremental fromsnap (may be zero)
4109  * zc_guid      if set, estimate size of stream only.  zc_cookie is ignored.
4110  *              output size in zc_objset_type.
4111  *
4112  * outputs: none
4113  */
4114 static int
4115 zfs_ioc_send(zfs_cmd_t *zc)
4116 {
4117         int error;
4118         offset_t off;
4119         boolean_t estimate = (zc->zc_guid != 0);
4120 
4121         if (zc->zc_obj != 0) {
4122                 dsl_pool_t *dp;
4123                 dsl_dataset_t *tosnap;
4124 
4125                 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4126                 if (error != 0)
4127                         return (error);
4128 
4129                 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4130                 if (error != 0) {
4131                         dsl_pool_rele(dp, FTAG);
4132                         return (error);
4133                 }
4134 
4135                 if (dsl_dir_is_clone(tosnap->ds_dir))
4136                         zc->zc_fromobj = tosnap->ds_dir->dd_phys->dd_origin_obj;
4137                 dsl_dataset_rele(tosnap, FTAG);
4138                 dsl_pool_rele(dp, FTAG);
4139         }
4140 
4141         if (estimate) {
4142                 dsl_pool_t *dp;
4143                 dsl_dataset_t *tosnap;
4144                 dsl_dataset_t *fromsnap = NULL;
4145 
4146                 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4147                 if (error != 0)
4148                         return (error);
4149 
4150                 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4151                 if (error != 0) {
4152                         dsl_pool_rele(dp, FTAG);
4153                         return (error);
4154                 }
4155 
4156                 if (zc->zc_fromobj != 0) {
4157                         error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4158                             FTAG, &fromsnap);
4159                         if (error != 0) {
4160                                 dsl_dataset_rele(tosnap, FTAG);
4161                                 dsl_pool_rele(dp, FTAG);
4162                                 return (error);
4163                         }
4164                 }
4165 
4166                 error = dmu_send_estimate(tosnap, fromsnap,
4167                     &zc->zc_objset_type);
4168 
4169                 if (fromsnap != NULL)
4170                         dsl_dataset_rele(fromsnap, FTAG);
4171                 dsl_dataset_rele(tosnap, FTAG);
4172                 dsl_pool_rele(dp, FTAG);
4173         } else {
4174                 file_t *fp = getf(zc->zc_cookie);
4175                 if (fp == NULL)
4176                         return (SET_ERROR(EBADF));
4177 
4178                 off = fp->f_offset;
4179                 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4180                     zc->zc_fromobj, zc->zc_cookie, fp->f_vnode, &off);
4181 
4182                 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4183                         fp->f_offset = off;
4184                 releasef(zc->zc_cookie);
4185         }
4186         return (error);
4187 }
4188 
4189 /*
4190  * inputs:
4191  * zc_name      name of snapshot on which to report progress
4192  * zc_cookie    file descriptor of send stream
4193  *
4194  * outputs:
4195  * zc_cookie    number of bytes written in send stream thus far
4196  */
4197 static int
4198 zfs_ioc_send_progress(zfs_cmd_t *zc)
4199 {
4200         dsl_pool_t *dp;
4201         dsl_dataset_t *ds;
4202         dmu_sendarg_t *dsp = NULL;
4203         int error;
4204 
4205         error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4206         if (error != 0)
4207                 return (error);
4208 
4209         error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4210         if (error != 0) {
4211                 dsl_pool_rele(dp, FTAG);
4212                 return (error);
4213         }
4214 
4215         mutex_enter(&ds->ds_sendstream_lock);
4216 
4217         /*
4218          * Iterate over all the send streams currently active on this dataset.
4219          * If there's one which matches the specified file descriptor _and_ the
4220          * stream was started by the current process, return the progress of
4221          * that stream.
4222          */
4223         for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4224             dsp = list_next(&ds->ds_sendstreams, dsp)) {
4225                 if (dsp->dsa_outfd == zc->zc_cookie &&
4226                     dsp->dsa_proc == curproc)
4227                         break;
4228         }
4229 
4230         if (dsp != NULL)
4231                 zc->zc_cookie = *(dsp->dsa_off);
4232         else
4233                 error = SET_ERROR(ENOENT);
4234 
4235         mutex_exit(&ds->ds_sendstream_lock);
4236         dsl_dataset_rele(ds, FTAG);
4237         dsl_pool_rele(dp, FTAG);
4238         return (error);
4239 }
4240 
4241 static int
4242 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4243 {
4244         int id, error;
4245 
4246         error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4247             &zc->zc_inject_record);
4248 
4249         if (error == 0)
4250                 zc->zc_guid = (uint64_t)id;
4251 
4252         return (error);
4253 }
4254 
4255 static int
4256 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4257 {
4258         return (zio_clear_fault((int)zc->zc_guid));
4259 }
4260 
4261 static int
4262 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4263 {
4264         int id = (int)zc->zc_guid;
4265         int error;
4266 
4267         error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4268             &zc->zc_inject_record);
4269 
4270         zc->zc_guid = id;
4271 
4272         return (error);
4273 }
4274 
4275 static int
4276 zfs_ioc_error_log(zfs_cmd_t *zc)
4277 {
4278         spa_t *spa;
4279         int error;
4280         size_t count = (size_t)zc->zc_nvlist_dst_size;
4281 
4282         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4283                 return (error);
4284 
4285         error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4286             &count);
4287         if (error == 0)
4288                 zc->zc_nvlist_dst_size = count;
4289         else
4290                 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4291 
4292         spa_close(spa, FTAG);
4293 
4294         return (error);
4295 }
4296 
4297 static int
4298 zfs_ioc_clear(zfs_cmd_t *zc)
4299 {
4300         spa_t *spa;
4301         vdev_t *vd;
4302         int error;
4303 
4304         /*
4305          * On zpool clear we also fix up missing slogs
4306          */
4307         mutex_enter(&spa_namespace_lock);
4308         spa = spa_lookup(zc->zc_name);
4309         if (spa == NULL) {
4310                 mutex_exit(&spa_namespace_lock);
4311                 return (SET_ERROR(EIO));
4312         }
4313         if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4314                 /* we need to let spa_open/spa_load clear the chains */
4315                 spa_set_log_state(spa, SPA_LOG_CLEAR);
4316         }
4317         spa->spa_last_open_failed = 0;
4318         mutex_exit(&spa_namespace_lock);
4319 
4320         if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4321                 error = spa_open(zc->zc_name, &spa, FTAG);
4322         } else {
4323                 nvlist_t *policy;
4324                 nvlist_t *config = NULL;
4325 
4326                 if (zc->zc_nvlist_src == NULL)
4327                         return (SET_ERROR(EINVAL));
4328 
4329                 if ((error = get_nvlist(zc->zc_nvlist_src,
4330                     zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4331                         error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4332                             policy, &config);
4333                         if (config != NULL) {
4334                                 int err;
4335 
4336                                 if ((err = put_nvlist(zc, config)) != 0)
4337                                         error = err;
4338                                 nvlist_free(config);
4339                         }
4340                         nvlist_free(policy);
4341                 }
4342         }
4343 
4344         if (error != 0)
4345                 return (error);
4346 
4347         spa_vdev_state_enter(spa, SCL_NONE);
4348 
4349         if (zc->zc_guid == 0) {
4350                 vd = NULL;
4351         } else {
4352                 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4353                 if (vd == NULL) {
4354                         (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4355                         spa_close(spa, FTAG);
4356                         return (SET_ERROR(ENODEV));
4357                 }
4358         }
4359 
4360         vdev_clear(spa, vd);
4361 
4362         (void) spa_vdev_state_exit(spa, NULL, 0);
4363 
4364         /*
4365          * Resume any suspended I/Os.
4366          */
4367         if (zio_resume(spa) != 0)
4368                 error = SET_ERROR(EIO);
4369 
4370         spa_close(spa, FTAG);
4371 
4372         return (error);
4373 }
4374 
4375 static int
4376 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4377 {
4378         spa_t *spa;
4379         int error;
4380 
4381         error = spa_open(zc->zc_name, &spa, FTAG);
4382         if (error != 0)
4383                 return (error);
4384 
4385         spa_vdev_state_enter(spa, SCL_NONE);
4386 
4387         /*
4388          * If a resilver is already in progress then set the
4389          * spa_scrub_reopen flag to B_TRUE so that we don't restart
4390          * the scan as a side effect of the reopen. Otherwise, let
4391          * vdev_open() decided if a resilver is required.
4392          */
4393         spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4394         vdev_reopen(spa->spa_root_vdev);
4395         spa->spa_scrub_reopen = B_FALSE;
4396 
4397         (void) spa_vdev_state_exit(spa, NULL, 0);
4398         spa_close(spa, FTAG);
4399         return (0);
4400 }
4401 /*
4402  * inputs:
4403  * zc_name      name of filesystem
4404  * zc_value     name of origin snapshot
4405  *
4406  * outputs:
4407  * zc_string    name of conflicting snapshot, if there is one
4408  */
4409 static int
4410 zfs_ioc_promote(zfs_cmd_t *zc)
4411 {
4412         char *cp;
4413 
4414         /*
4415          * We don't need to unmount *all* the origin fs's snapshots, but
4416          * it's easier.
4417          */
4418         cp = strchr(zc->zc_value, '@');
4419         if (cp)
4420                 *cp = '\0';
4421         (void) dmu_objset_find(zc->zc_value,
4422             zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
4423         return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4424 }
4425 
4426 /*
4427  * Retrieve a single {user|group}{used|quota}@... property.
4428  *
4429  * inputs:
4430  * zc_name      name of filesystem
4431  * zc_objset_type zfs_userquota_prop_t
4432  * zc_value     domain name (eg. "S-1-234-567-89")
4433  * zc_guid      RID/UID/GID
4434  *
4435  * outputs:
4436  * zc_cookie    property value
4437  */
4438 static int
4439 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4440 {
4441         zfsvfs_t *zfsvfs;
4442         int error;
4443 
4444         if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4445                 return (SET_ERROR(EINVAL));
4446 
4447         error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4448         if (error != 0)
4449                 return (error);
4450 
4451         error = zfs_userspace_one(zfsvfs,
4452             zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4453         zfsvfs_rele(zfsvfs, FTAG);
4454 
4455         return (error);
4456 }
4457 
4458 /*
4459  * inputs:
4460  * zc_name              name of filesystem
4461  * zc_cookie            zap cursor
4462  * zc_objset_type       zfs_userquota_prop_t
4463  * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4464  *
4465  * outputs:
4466  * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4467  * zc_cookie    zap cursor
4468  */
4469 static int
4470 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4471 {
4472         zfsvfs_t *zfsvfs;
4473         int bufsize = zc->zc_nvlist_dst_size;
4474 
4475         if (bufsize <= 0)
4476                 return (SET_ERROR(ENOMEM));
4477 
4478         int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4479         if (error != 0)
4480                 return (error);
4481 
4482         void *buf = kmem_alloc(bufsize, KM_SLEEP);
4483 
4484         error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
4485             buf, &zc->zc_nvlist_dst_size);
4486 
4487         if (error == 0) {
4488                 error = xcopyout(buf,
4489                     (void *)(uintptr_t)zc->zc_nvlist_dst,
4490                     zc->zc_nvlist_dst_size);
4491         }
4492         kmem_free(buf, bufsize);
4493         zfsvfs_rele(zfsvfs, FTAG);
4494 
4495         return (error);
4496 }
4497 
4498 /*
4499  * inputs:
4500  * zc_name              name of filesystem
4501  *
4502  * outputs:
4503  * none
4504  */
4505 static int
4506 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4507 {
4508         objset_t *os;
4509         int error = 0;
4510         zfsvfs_t *zfsvfs;
4511 
4512         if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
4513                 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
4514                         /*
4515                          * If userused is not enabled, it may be because the
4516                          * objset needs to be closed & reopened (to grow the
4517                          * objset_phys_t).  Suspend/resume the fs will do that.
4518                          */
4519                         error = zfs_suspend_fs(zfsvfs);
4520                         if (error == 0)
4521                                 error = zfs_resume_fs(zfsvfs, zc->zc_name);
4522                 }
4523                 if (error == 0)
4524                         error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
4525                 VFS_RELE(zfsvfs->z_vfs);
4526         } else {
4527                 /* XXX kind of reading contents without owning */
4528                 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4529                 if (error != 0)
4530                         return (error);
4531 
4532                 error = dmu_objset_userspace_upgrade(os);
4533                 dmu_objset_rele(os, FTAG);
4534         }
4535 
4536         return (error);
4537 }
4538 
4539 /*
4540  * We don't want to have a hard dependency
4541  * against some special symbols in sharefs
4542  * nfs, and smbsrv.  Determine them if needed when
4543  * the first file system is shared.
4544  * Neither sharefs, nfs or smbsrv are unloadable modules.
4545  */
4546 int (*znfsexport_fs)(void *arg);
4547 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
4548 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
4549 
4550 int zfs_nfsshare_inited;
4551 int zfs_smbshare_inited;
4552 
4553 ddi_modhandle_t nfs_mod;
4554 ddi_modhandle_t sharefs_mod;
4555 ddi_modhandle_t smbsrv_mod;
4556 kmutex_t zfs_share_lock;
4557 
4558 static int
4559 zfs_init_sharefs()
4560 {
4561         int error;
4562 
4563         ASSERT(MUTEX_HELD(&zfs_share_lock));
4564         /* Both NFS and SMB shares also require sharetab support. */
4565         if (sharefs_mod == NULL && ((sharefs_mod =
4566             ddi_modopen("fs/sharefs",
4567             KRTLD_MODE_FIRST, &error)) == NULL)) {
4568                 return (SET_ERROR(ENOSYS));
4569         }
4570         if (zshare_fs == NULL && ((zshare_fs =
4571             (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
4572             ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
4573                 return (SET_ERROR(ENOSYS));
4574         }
4575         return (0);
4576 }
4577 
4578 static int
4579 zfs_ioc_share(zfs_cmd_t *zc)
4580 {
4581         int error;
4582         int opcode;
4583 
4584         switch (zc->zc_share.z_sharetype) {
4585         case ZFS_SHARE_NFS:
4586         case ZFS_UNSHARE_NFS:
4587                 if (zfs_nfsshare_inited == 0) {
4588                         mutex_enter(&zfs_share_lock);
4589                         if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
4590                             KRTLD_MODE_FIRST, &error)) == NULL)) {
4591                                 mutex_exit(&zfs_share_lock);
4592                                 return (SET_ERROR(ENOSYS));
4593                         }
4594                         if (znfsexport_fs == NULL &&
4595                             ((znfsexport_fs = (int (*)(void *))
4596                             ddi_modsym(nfs_mod,
4597                             "nfs_export", &error)) == NULL)) {
4598                                 mutex_exit(&zfs_share_lock);
4599                                 return (SET_ERROR(ENOSYS));
4600                         }
4601                         error = zfs_init_sharefs();
4602                         if (error != 0) {
4603                                 mutex_exit(&zfs_share_lock);
4604                                 return (SET_ERROR(ENOSYS));
4605                         }
4606                         zfs_nfsshare_inited = 1;
4607                         mutex_exit(&zfs_share_lock);
4608                 }
4609                 break;
4610         case ZFS_SHARE_SMB:
4611         case ZFS_UNSHARE_SMB:
4612                 if (zfs_smbshare_inited == 0) {
4613                         mutex_enter(&zfs_share_lock);
4614                         if (smbsrv_mod == NULL && ((smbsrv_mod =
4615                             ddi_modopen("drv/smbsrv",
4616                             KRTLD_MODE_FIRST, &error)) == NULL)) {
4617                                 mutex_exit(&zfs_share_lock);
4618                                 return (SET_ERROR(ENOSYS));
4619                         }
4620                         if (zsmbexport_fs == NULL && ((zsmbexport_fs =
4621                             (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
4622                             "smb_server_share", &error)) == NULL)) {
4623                                 mutex_exit(&zfs_share_lock);
4624                                 return (SET_ERROR(ENOSYS));
4625                         }
4626                         error = zfs_init_sharefs();
4627                         if (error != 0) {
4628                                 mutex_exit(&zfs_share_lock);
4629                                 return (SET_ERROR(ENOSYS));
4630                         }
4631                         zfs_smbshare_inited = 1;
4632                         mutex_exit(&zfs_share_lock);
4633                 }
4634                 break;
4635         default:
4636                 return (SET_ERROR(EINVAL));
4637         }
4638 
4639         switch (zc->zc_share.z_sharetype) {
4640         case ZFS_SHARE_NFS:
4641         case ZFS_UNSHARE_NFS:
4642                 if (error =
4643                     znfsexport_fs((void *)
4644                     (uintptr_t)zc->zc_share.z_exportdata))
4645                         return (error);
4646                 break;
4647         case ZFS_SHARE_SMB:
4648         case ZFS_UNSHARE_SMB:
4649                 if (error = zsmbexport_fs((void *)
4650                     (uintptr_t)zc->zc_share.z_exportdata,
4651                     zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
4652                     B_TRUE: B_FALSE)) {
4653                         return (error);
4654                 }
4655                 break;
4656         }
4657 
4658         opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
4659             zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
4660             SHAREFS_ADD : SHAREFS_REMOVE;
4661 
4662         /*
4663          * Add or remove share from sharetab
4664          */
4665         error = zshare_fs(opcode,
4666             (void *)(uintptr_t)zc->zc_share.z_sharedata,
4667             zc->zc_share.z_sharemax);
4668 
4669         return (error);
4670 
4671 }
4672 
4673 ace_t full_access[] = {
4674         {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4675 };
4676 
4677 /*
4678  * inputs:
4679  * zc_name              name of containing filesystem
4680  * zc_obj               object # beyond which we want next in-use object #
4681  *
4682  * outputs:
4683  * zc_obj               next in-use object #
4684  */
4685 static int
4686 zfs_ioc_next_obj(zfs_cmd_t *zc)
4687 {
4688         objset_t *os = NULL;
4689         int error;
4690 
4691         error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4692         if (error != 0)
4693                 return (error);
4694 
4695         error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4696             os->os_dsl_dataset->ds_phys->ds_prev_snap_txg);
4697 
4698         dmu_objset_rele(os, FTAG);
4699         return (error);
4700 }
4701 
4702 /*
4703  * inputs:
4704  * zc_name              name of filesystem
4705  * zc_value             prefix name for snapshot
4706  * zc_cleanup_fd        cleanup-on-exit file descriptor for calling process
4707  *
4708  * outputs:
4709  * zc_value             short name of new snapshot
4710  */
4711 static int
4712 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
4713 {
4714         char *snap_name;
4715         char *hold_name;
4716         int error;
4717         minor_t minor;
4718 
4719         error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
4720         if (error != 0)
4721                 return (error);
4722 
4723         snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
4724             (u_longlong_t)ddi_get_lbolt64());
4725         hold_name = kmem_asprintf("%%%s", zc->zc_value);
4726 
4727         error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
4728             hold_name);
4729         if (error == 0)
4730                 (void) strcpy(zc->zc_value, snap_name);
4731         strfree(snap_name);
4732         strfree(hold_name);
4733         zfs_onexit_fd_rele(zc->zc_cleanup_fd);
4734         return (error);
4735 }
4736 
4737 /*
4738  * inputs:
4739  * zc_name              name of "to" snapshot
4740  * zc_value             name of "from" snapshot
4741  * zc_cookie            file descriptor to write diff data on
4742  *
4743  * outputs:
4744  * dmu_diff_record_t's to the file descriptor
4745  */
4746 static int
4747 zfs_ioc_diff(zfs_cmd_t *zc)
4748 {
4749         file_t *fp;
4750         offset_t off;
4751         int error;
4752 
4753         fp = getf(zc->zc_cookie);
4754         if (fp == NULL)
4755                 return (SET_ERROR(EBADF));
4756 
4757         off = fp->f_offset;
4758 
4759         error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
4760 
4761         if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4762                 fp->f_offset = off;
4763         releasef(zc->zc_cookie);
4764 
4765         return (error);
4766 }
4767 
4768 /*
4769  * Remove all ACL files in shares dir
4770  */
4771 static int
4772 zfs_smb_acl_purge(znode_t *dzp)
4773 {
4774         zap_cursor_t    zc;
4775         zap_attribute_t zap;
4776         zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
4777         int error;
4778 
4779         for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
4780             (error = zap_cursor_retrieve(&zc, &zap)) == 0;
4781             zap_cursor_advance(&zc)) {
4782                 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
4783                     NULL, 0)) != 0)
4784                         break;
4785         }
4786         zap_cursor_fini(&zc);
4787         return (error);
4788 }
4789 
4790 static int
4791 zfs_ioc_smb_acl(zfs_cmd_t *zc)
4792 {
4793         vnode_t *vp;
4794         znode_t *dzp;
4795         vnode_t *resourcevp = NULL;
4796         znode_t *sharedir;
4797         zfsvfs_t *zfsvfs;
4798         nvlist_t *nvlist;
4799         char *src, *target;
4800         vattr_t vattr;
4801         vsecattr_t vsec;
4802         int error = 0;
4803 
4804         if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
4805             NO_FOLLOW, NULL, &vp)) != 0)
4806                 return (error);
4807 
4808         /* Now make sure mntpnt and dataset are ZFS */
4809 
4810         if (vp->v_vfsp->vfs_fstype != zfsfstype ||
4811             (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
4812             zc->zc_name) != 0)) {
4813                 VN_RELE(vp);
4814                 return (SET_ERROR(EINVAL));
4815         }
4816 
4817         dzp = VTOZ(vp);
4818         zfsvfs = dzp->z_zfsvfs;
4819         ZFS_ENTER(zfsvfs);
4820 
4821         /*
4822          * Create share dir if its missing.
4823          */
4824         mutex_enter(&zfsvfs->z_lock);
4825         if (zfsvfs->z_shares_dir == 0) {
4826                 dmu_tx_t *tx;
4827 
4828                 tx = dmu_tx_create(zfsvfs->z_os);
4829                 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
4830                     ZFS_SHARES_DIR);
4831                 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
4832                 error = dmu_tx_assign(tx, TXG_WAIT);
4833                 if (error != 0) {
4834                         dmu_tx_abort(tx);
4835                 } else {
4836                         error = zfs_create_share_dir(zfsvfs, tx);
4837                         dmu_tx_commit(tx);
4838                 }
4839                 if (error != 0) {
4840                         mutex_exit(&zfsvfs->z_lock);
4841                         VN_RELE(vp);
4842                         ZFS_EXIT(zfsvfs);
4843                         return (error);
4844                 }
4845         }
4846         mutex_exit(&zfsvfs->z_lock);
4847 
4848         ASSERT(zfsvfs->z_shares_dir);
4849         if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
4850                 VN_RELE(vp);
4851                 ZFS_EXIT(zfsvfs);
4852                 return (error);
4853         }
4854 
4855         switch (zc->zc_cookie) {
4856         case ZFS_SMB_ACL_ADD:
4857                 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
4858                 vattr.va_type = VREG;
4859                 vattr.va_mode = S_IFREG|0777;
4860                 vattr.va_uid = 0;
4861                 vattr.va_gid = 0;
4862 
4863                 vsec.vsa_mask = VSA_ACE;
4864                 vsec.vsa_aclentp = &full_access;
4865                 vsec.vsa_aclentsz = sizeof (full_access);
4866                 vsec.vsa_aclcnt = 1;
4867 
4868                 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
4869                     &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
4870                 if (resourcevp)
4871                         VN_RELE(resourcevp);
4872                 break;
4873 
4874         case ZFS_SMB_ACL_REMOVE:
4875                 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
4876                     NULL, 0);
4877                 break;
4878 
4879         case ZFS_SMB_ACL_RENAME:
4880                 if ((error = get_nvlist(zc->zc_nvlist_src,
4881                     zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
4882                         VN_RELE(vp);
4883                         ZFS_EXIT(zfsvfs);
4884                         return (error);
4885                 }
4886                 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
4887                     nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
4888                     &target)) {
4889                         VN_RELE(vp);
4890                         VN_RELE(ZTOV(sharedir));
4891                         ZFS_EXIT(zfsvfs);
4892                         nvlist_free(nvlist);
4893                         return (error);
4894                 }
4895                 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
4896                     kcred, NULL, 0);
4897                 nvlist_free(nvlist);
4898                 break;
4899 
4900         case ZFS_SMB_ACL_PURGE:
4901                 error = zfs_smb_acl_purge(sharedir);
4902                 break;
4903 
4904         default:
4905                 error = SET_ERROR(EINVAL);
4906                 break;
4907         }
4908 
4909         VN_RELE(vp);
4910         VN_RELE(ZTOV(sharedir));
4911 
4912         ZFS_EXIT(zfsvfs);
4913 
4914         return (error);
4915 }
4916 
4917 /*
4918  * innvl: {
4919  *     "holds" -> { snapname -> holdname (string), ... }
4920  *     (optional) "cleanup_fd" -> fd (int32)
4921  * }
4922  *
4923  * outnvl: {
4924  *     snapname -> error value (int32)
4925  *     ...
4926  * }
4927  */
4928 /* ARGSUSED */
4929 static int
4930 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
4931 {
4932         nvlist_t *holds;
4933         int cleanup_fd = -1;
4934         int error;
4935         minor_t minor = 0;
4936 
4937         error = nvlist_lookup_nvlist(args, "holds", &holds);
4938         if (error != 0)
4939                 return (SET_ERROR(EINVAL));
4940 
4941         if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
4942                 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
4943                 if (error != 0)
4944                         return (error);
4945         }
4946 
4947         error = dsl_dataset_user_hold(holds, minor, errlist);
4948         if (minor != 0)
4949                 zfs_onexit_fd_rele(cleanup_fd);
4950         return (error);
4951 }
4952 
4953 /*
4954  * innvl is not used.
4955  *
4956  * outnvl: {
4957  *    holdname -> time added (uint64 seconds since epoch)
4958  *    ...
4959  * }
4960  */
4961 /* ARGSUSED */
4962 static int
4963 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
4964 {
4965         return (dsl_dataset_get_holds(snapname, outnvl));
4966 }
4967 
4968 /*
4969  * innvl: {
4970  *     snapname -> { holdname, ... }
4971  *     ...
4972  * }
4973  *
4974  * outnvl: {
4975  *     snapname -> error value (int32)
4976  *     ...
4977  * }
4978  */
4979 /* ARGSUSED */
4980 static int
4981 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
4982 {
4983         nvpair_t *pair;
4984 
4985         /*
4986          * The release may cause the snapshot to be destroyed; make sure it
4987          * is not mounted.
4988          */
4989         for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
4990             pair = nvlist_next_nvpair(holds, pair))
4991                 zfs_unmount_snap(nvpair_name(pair));
4992 
4993         return (dsl_dataset_user_release(holds, errlist));
4994 }
4995 
4996 /*
4997  * inputs:
4998  * zc_name              name of new filesystem or snapshot
4999  * zc_value             full name of old snapshot
5000  *
5001  * outputs:
5002  * zc_cookie            space in bytes
5003  * zc_objset_type       compressed space in bytes
5004  * zc_perm_action       uncompressed space in bytes
5005  */
5006 static int
5007 zfs_ioc_space_written(zfs_cmd_t *zc)
5008 {
5009         int error;
5010         dsl_pool_t *dp;
5011         dsl_dataset_t *new, *old;
5012 
5013         error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5014         if (error != 0)
5015                 return (error);
5016         error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5017         if (error != 0) {
5018                 dsl_pool_rele(dp, FTAG);
5019                 return (error);
5020         }
5021         error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5022         if (error != 0) {
5023                 dsl_dataset_rele(new, FTAG);
5024                 dsl_pool_rele(dp, FTAG);
5025                 return (error);
5026         }
5027 
5028         error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5029             &zc->zc_objset_type, &zc->zc_perm_action);
5030         dsl_dataset_rele(old, FTAG);
5031         dsl_dataset_rele(new, FTAG);
5032         dsl_pool_rele(dp, FTAG);
5033         return (error);
5034 }
5035 
5036 /*
5037  * innvl: {
5038  *     "firstsnap" -> snapshot name
5039  * }
5040  *
5041  * outnvl: {
5042  *     "used" -> space in bytes
5043  *     "compressed" -> compressed space in bytes
5044  *     "uncompressed" -> uncompressed space in bytes
5045  * }
5046  */
5047 static int
5048 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5049 {
5050         int error;
5051         dsl_pool_t *dp;
5052         dsl_dataset_t *new, *old;
5053         char *firstsnap;
5054         uint64_t used, comp, uncomp;
5055 
5056         if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5057                 return (SET_ERROR(EINVAL));
5058 
5059         error = dsl_pool_hold(lastsnap, FTAG, &dp);
5060         if (error != 0)
5061                 return (error);
5062 
5063         error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5064         if (error != 0) {
5065                 dsl_pool_rele(dp, FTAG);
5066                 return (error);
5067         }
5068         error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5069         if (error != 0) {
5070                 dsl_dataset_rele(new, FTAG);
5071                 dsl_pool_rele(dp, FTAG);
5072                 return (error);
5073         }
5074 
5075         error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5076         dsl_dataset_rele(old, FTAG);
5077         dsl_dataset_rele(new, FTAG);
5078         dsl_pool_rele(dp, FTAG);
5079         fnvlist_add_uint64(outnvl, "used", used);
5080         fnvlist_add_uint64(outnvl, "compressed", comp);
5081         fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5082         return (error);
5083 }
5084 
5085 /*
5086  * innvl: {
5087  *     "fd" -> file descriptor to write stream to (int32)
5088  *     (optional) "fromsnap" -> full snap name to send an incremental from
5089  * }
5090  *
5091  * outnvl is unused
5092  */
5093 /* ARGSUSED */
5094 static int
5095 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5096 {
5097         int error;
5098         offset_t off;
5099         char *fromname = NULL;
5100         int fd;
5101 
5102         error = nvlist_lookup_int32(innvl, "fd", &fd);
5103         if (error != 0)
5104                 return (SET_ERROR(EINVAL));
5105 
5106         (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5107 
5108         file_t *fp = getf(fd);
5109         if (fp == NULL)
5110                 return (SET_ERROR(EBADF));
5111 
5112         off = fp->f_offset;
5113         error = dmu_send(snapname, fromname, fd, fp->f_vnode, &off);
5114 
5115         if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5116                 fp->f_offset = off;
5117         releasef(fd);
5118         return (error);
5119 }
5120 
5121 /*
5122  * Determine approximately how large a zfs send stream will be -- the number
5123  * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5124  *
5125  * innvl: {
5126  *     (optional) "fromsnap" -> full snap name to send an incremental from
5127  * }
5128  *
5129  * outnvl: {
5130  *     "space" -> bytes of space (uint64)
5131  * }
5132  */
5133 static int
5134 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5135 {
5136         dsl_pool_t *dp;
5137         dsl_dataset_t *fromsnap = NULL;
5138         dsl_dataset_t *tosnap;
5139         int error;
5140         char *fromname;
5141         uint64_t space;
5142 
5143         error = dsl_pool_hold(snapname, FTAG, &dp);
5144         if (error != 0)
5145                 return (error);
5146 
5147         error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5148         if (error != 0) {
5149                 dsl_pool_rele(dp, FTAG);
5150                 return (error);
5151         }
5152 
5153         error = nvlist_lookup_string(innvl, "fromsnap", &fromname);
5154         if (error == 0) {
5155                 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5156                 if (error != 0) {
5157                         dsl_dataset_rele(tosnap, FTAG);
5158                         dsl_pool_rele(dp, FTAG);
5159                         return (error);
5160                 }
5161         }
5162 
5163         error = dmu_send_estimate(tosnap, fromsnap, &space);
5164         fnvlist_add_uint64(outnvl, "space", space);
5165 
5166         if (fromsnap != NULL)
5167                 dsl_dataset_rele(fromsnap, FTAG);
5168         dsl_dataset_rele(tosnap, FTAG);
5169         dsl_pool_rele(dp, FTAG);
5170         return (error);
5171 }
5172 
5173 
5174 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5175 
5176 static void
5177 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5178     zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5179     boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5180 {
5181         zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5182 
5183         ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5184         ASSERT3U(ioc, <, ZFS_IOC_LAST);
5185         ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5186         ASSERT3P(vec->zvec_func, ==, NULL);
5187 
5188         vec->zvec_legacy_func = func;
5189         vec->zvec_secpolicy = secpolicy;
5190         vec->zvec_namecheck = namecheck;
5191         vec->zvec_allow_log = log_history;
5192         vec->zvec_pool_check = pool_check;
5193 }
5194 
5195 /*
5196  * See the block comment at the beginning of this file for details on
5197  * each argument to this function.
5198  */
5199 static void
5200 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5201     zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5202     zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5203     boolean_t allow_log)
5204 {
5205         zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5206 
5207         ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5208         ASSERT3U(ioc, <, ZFS_IOC_LAST);
5209         ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5210         ASSERT3P(vec->zvec_func, ==, NULL);
5211 
5212         /* if we are logging, the name must be valid */
5213         ASSERT(!allow_log || namecheck != NO_NAME);
5214 
5215         vec->zvec_name = name;
5216         vec->zvec_func = func;
5217         vec->zvec_secpolicy = secpolicy;
5218         vec->zvec_namecheck = namecheck;
5219         vec->zvec_pool_check = pool_check;
5220         vec->zvec_smush_outnvlist = smush_outnvlist;
5221         vec->zvec_allow_log = allow_log;
5222 }
5223 
5224 static void
5225 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5226     zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5227     zfs_ioc_poolcheck_t pool_check)
5228 {
5229         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5230             POOL_NAME, log_history, pool_check);
5231 }
5232 
5233 static void
5234 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5235     zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5236 {
5237         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5238             DATASET_NAME, B_FALSE, pool_check);
5239 }
5240 
5241 static void
5242 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5243 {
5244         zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5245             POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5246 }
5247 
5248 static void
5249 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5250     zfs_secpolicy_func_t *secpolicy)
5251 {
5252         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5253             NO_NAME, B_FALSE, POOL_CHECK_NONE);
5254 }
5255 
5256 static void
5257 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5258     zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5259 {
5260         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5261             DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5262 }
5263 
5264 static void
5265 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5266 {
5267         zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5268             zfs_secpolicy_read);
5269 }
5270 
5271 static void
5272 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5273         zfs_secpolicy_func_t *secpolicy)
5274 {
5275         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5276             DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5277 }
5278 
5279 static void
5280 zfs_ioctl_init(void)
5281 {
5282         zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5283             zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5284             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5285 
5286         zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5287             zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5288             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5289 
5290         zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5291             zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5292             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5293 
5294         zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5295             zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5296             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5297 
5298         zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5299             zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5300             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5301 
5302         zfs_ioctl_register("create", ZFS_IOC_CREATE,
5303             zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5304             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5305 
5306         zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5307             zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5308             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5309 
5310         zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5311             zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5312             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5313 
5314         zfs_ioctl_register("hold", ZFS_IOC_HOLD,
5315             zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
5316             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5317         zfs_ioctl_register("release", ZFS_IOC_RELEASE,
5318             zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
5319             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5320 
5321         zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
5322             zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
5323             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5324 
5325         /* IOCTLS that use the legacy function signature */
5326 
5327         zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5328             zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5329 
5330         zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5331             zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5332         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5333             zfs_ioc_pool_scan);
5334         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5335             zfs_ioc_pool_upgrade);
5336         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5337             zfs_ioc_vdev_add);
5338         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5339             zfs_ioc_vdev_remove);
5340         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5341             zfs_ioc_vdev_set_state);
5342         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5343             zfs_ioc_vdev_attach);
5344         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5345             zfs_ioc_vdev_detach);
5346         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5347             zfs_ioc_vdev_setpath);
5348         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5349             zfs_ioc_vdev_setfru);
5350         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5351             zfs_ioc_pool_set_props);
5352         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5353             zfs_ioc_vdev_split);
5354         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5355             zfs_ioc_pool_reguid);
5356 
5357         zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5358             zfs_ioc_pool_configs, zfs_secpolicy_none);
5359         zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5360             zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5361         zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5362             zfs_ioc_inject_fault, zfs_secpolicy_inject);
5363         zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5364             zfs_ioc_clear_fault, zfs_secpolicy_inject);
5365         zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
5366             zfs_ioc_inject_list_next, zfs_secpolicy_inject);
5367 
5368         /*
5369          * pool destroy, and export don't log the history as part of
5370          * zfsdev_ioctl, but rather zfs_ioc_pool_export
5371          * does the logging of those commands.
5372          */
5373         zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
5374             zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5375         zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
5376             zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5377 
5378         zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
5379             zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5380         zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
5381             zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5382 
5383         zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
5384             zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
5385         zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
5386             zfs_ioc_dsobj_to_dsname,
5387             zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
5388         zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
5389             zfs_ioc_pool_get_history,
5390             zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
5391 
5392         zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
5393             zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5394 
5395         zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
5396             zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5397         zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
5398             zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5399 
5400         zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
5401             zfs_ioc_space_written);
5402         zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
5403             zfs_ioc_objset_recvd_props);
5404         zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
5405             zfs_ioc_next_obj);
5406         zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
5407             zfs_ioc_get_fsacl);
5408         zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
5409             zfs_ioc_objset_stats);
5410         zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
5411             zfs_ioc_objset_zplprops);
5412         zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
5413             zfs_ioc_dataset_list_next);
5414         zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
5415             zfs_ioc_snapshot_list_next);
5416         zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
5417             zfs_ioc_send_progress);
5418 
5419         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
5420             zfs_ioc_diff, zfs_secpolicy_diff);
5421         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
5422             zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
5423         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
5424             zfs_ioc_obj_to_path, zfs_secpolicy_diff);
5425         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
5426             zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
5427         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
5428             zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
5429         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
5430             zfs_ioc_send, zfs_secpolicy_send);
5431 
5432         zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
5433             zfs_secpolicy_none);
5434         zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
5435             zfs_secpolicy_destroy);
5436         zfs_ioctl_register_dataset_modify(ZFS_IOC_ROLLBACK, zfs_ioc_rollback,
5437             zfs_secpolicy_rollback);
5438         zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
5439             zfs_secpolicy_rename);
5440         zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
5441             zfs_secpolicy_recv);
5442         zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
5443             zfs_secpolicy_promote);
5444         zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
5445             zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
5446         zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
5447             zfs_secpolicy_set_fsacl);
5448 
5449         zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
5450             zfs_secpolicy_share, POOL_CHECK_NONE);
5451         zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
5452             zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
5453         zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
5454             zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
5455             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5456         zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
5457             zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
5458             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5459 }
5460 
5461 int
5462 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
5463     zfs_ioc_poolcheck_t check)
5464 {
5465         spa_t *spa;
5466         int error;
5467 
5468         ASSERT(type == POOL_NAME || type == DATASET_NAME);
5469 
5470         if (check & POOL_CHECK_NONE)
5471                 return (0);
5472 
5473         error = spa_open(name, &spa, FTAG);
5474         if (error == 0) {
5475                 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
5476                         error = SET_ERROR(EAGAIN);
5477                 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
5478                         error = SET_ERROR(EROFS);
5479                 spa_close(spa, FTAG);
5480         }
5481         return (error);
5482 }
5483 
5484 /*
5485  * Find a free minor number.
5486  */
5487 minor_t
5488 zfsdev_minor_alloc(void)
5489 {
5490         static minor_t last_minor;
5491         minor_t m;
5492 
5493         ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5494 
5495         for (m = last_minor + 1; m != last_minor; m++) {
5496                 if (m > ZFSDEV_MAX_MINOR)
5497                         m = 1;
5498                 if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
5499                         last_minor = m;
5500                         return (m);
5501                 }
5502         }
5503 
5504         return (0);
5505 }
5506 
5507 static int
5508 zfs_ctldev_init(dev_t *devp)
5509 {
5510         minor_t minor;
5511         zfs_soft_state_t *zs;
5512 
5513         ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5514         ASSERT(getminor(*devp) == 0);
5515 
5516         minor = zfsdev_minor_alloc();
5517         if (minor == 0)
5518                 return (SET_ERROR(ENXIO));
5519 
5520         if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
5521                 return (SET_ERROR(EAGAIN));
5522 
5523         *devp = makedevice(getemajor(*devp), minor);
5524 
5525         zs = ddi_get_soft_state(zfsdev_state, minor);
5526         zs->zss_type = ZSST_CTLDEV;
5527         zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
5528 
5529         return (0);
5530 }
5531 
5532 static void
5533 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
5534 {
5535         ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5536 
5537         zfs_onexit_destroy(zo);
5538         ddi_soft_state_free(zfsdev_state, minor);
5539 }
5540 
5541 void *
5542 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
5543 {
5544         zfs_soft_state_t *zp;
5545 
5546         zp = ddi_get_soft_state(zfsdev_state, minor);
5547         if (zp == NULL || zp->zss_type != which)
5548                 return (NULL);
5549 
5550         return (zp->zss_data);
5551 }
5552 
5553 static int
5554 zfsdev_open(dev_t *devp, int flag, int otyp, cred_t *cr)
5555 {
5556         int error = 0;
5557 
5558         if (getminor(*devp) != 0)
5559                 return (zvol_open(devp, flag, otyp, cr));
5560 
5561         /* This is the control device. Allocate a new minor if requested. */
5562         if (flag & FEXCL) {
5563                 mutex_enter(&zfsdev_state_lock);
5564                 error = zfs_ctldev_init(devp);
5565                 mutex_exit(&zfsdev_state_lock);
5566         }
5567 
5568         return (error);
5569 }
5570 
5571 static int
5572 zfsdev_close(dev_t dev, int flag, int otyp, cred_t *cr)
5573 {
5574         zfs_onexit_t *zo;
5575         minor_t minor = getminor(dev);
5576 
5577         if (minor == 0)
5578                 return (0);
5579 
5580         mutex_enter(&zfsdev_state_lock);
5581         zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
5582         if (zo == NULL) {
5583                 mutex_exit(&zfsdev_state_lock);
5584                 return (zvol_close(dev, flag, otyp, cr));
5585         }
5586         zfs_ctldev_destroy(zo, minor);
5587         mutex_exit(&zfsdev_state_lock);
5588 
5589         return (0);
5590 }
5591 
5592 static int
5593 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
5594 {
5595         zfs_cmd_t *zc;
5596         uint_t vecnum;
5597         int error, rc, len;
5598         minor_t minor = getminor(dev);
5599         const zfs_ioc_vec_t *vec;
5600         char *saved_poolname = NULL;
5601         nvlist_t *innvl = NULL;
5602 
5603         if (minor != 0 &&
5604             zfsdev_get_soft_state(minor, ZSST_CTLDEV) == NULL)
5605                 return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
5606 
5607         vecnum = cmd - ZFS_IOC_FIRST;
5608         ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
5609 
5610         if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
5611                 return (SET_ERROR(EINVAL));
5612         vec = &zfs_ioc_vec[vecnum];
5613 
5614         zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
5615 
5616         error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
5617         if (error != 0) {
5618                 error = SET_ERROR(EFAULT);
5619                 goto out;
5620         }
5621 
5622         zc->zc_iflags = flag & FKIOCTL;
5623         if (zc->zc_nvlist_src_size != 0) {
5624                 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
5625                     zc->zc_iflags, &innvl);
5626                 if (error != 0)
5627                         goto out;
5628         }
5629 
5630         /*
5631          * Ensure that all pool/dataset names are valid before we pass down to
5632          * the lower layers.
5633          */
5634         zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
5635         switch (vec->zvec_namecheck) {
5636         case POOL_NAME:
5637                 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
5638                         error = SET_ERROR(EINVAL);
5639                 else
5640                         error = pool_status_check(zc->zc_name,
5641                             vec->zvec_namecheck, vec->zvec_pool_check);
5642                 break;
5643 
5644         case DATASET_NAME:
5645                 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
5646                         error = SET_ERROR(EINVAL);
5647                 else
5648                         error = pool_status_check(zc->zc_name,
5649                             vec->zvec_namecheck, vec->zvec_pool_check);
5650                 break;
5651 
5652         case NO_NAME:
5653                 break;
5654         }
5655 
5656 
5657         if (error == 0 && !(flag & FKIOCTL))
5658                 error = vec->zvec_secpolicy(zc, innvl, cr);
5659 
5660         if (error != 0)
5661                 goto out;
5662 
5663         /* legacy ioctls can modify zc_name */
5664         len = strcspn(zc->zc_name, "/@") + 1;
5665         saved_poolname = kmem_alloc(len, KM_SLEEP);
5666         (void) strlcpy(saved_poolname, zc->zc_name, len);
5667 
5668         if (vec->zvec_func != NULL) {
5669                 nvlist_t *outnvl;
5670                 int puterror = 0;
5671                 spa_t *spa;
5672                 nvlist_t *lognv = NULL;
5673 
5674                 ASSERT(vec->zvec_legacy_func == NULL);
5675 
5676                 /*
5677                  * Add the innvl to the lognv before calling the func,
5678                  * in case the func changes the innvl.
5679                  */
5680                 if (vec->zvec_allow_log) {
5681                         lognv = fnvlist_alloc();
5682                         fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
5683                             vec->zvec_name);
5684                         if (!nvlist_empty(innvl)) {
5685                                 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
5686                                     innvl);
5687                         }
5688                 }
5689 
5690                 outnvl = fnvlist_alloc();
5691                 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
5692 
5693                 if (error == 0 && vec->zvec_allow_log &&
5694                     spa_open(zc->zc_name, &spa, FTAG) == 0) {
5695                         if (!nvlist_empty(outnvl)) {
5696                                 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
5697                                     outnvl);
5698                         }
5699                         (void) spa_history_log_nvl(spa, lognv);
5700                         spa_close(spa, FTAG);
5701                 }
5702                 fnvlist_free(lognv);
5703 
5704                 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
5705                         int smusherror = 0;
5706                         if (vec->zvec_smush_outnvlist) {
5707                                 smusherror = nvlist_smush(outnvl,
5708                                     zc->zc_nvlist_dst_size);
5709                         }
5710                         if (smusherror == 0)
5711                                 puterror = put_nvlist(zc, outnvl);
5712                 }
5713 
5714                 if (puterror != 0)
5715                         error = puterror;
5716 
5717                 nvlist_free(outnvl);
5718         } else {
5719                 error = vec->zvec_legacy_func(zc);
5720         }
5721 
5722 out:
5723         nvlist_free(innvl);
5724         rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
5725         if (error == 0 && rc != 0)
5726                 error = SET_ERROR(EFAULT);
5727         if (error == 0 && vec->zvec_allow_log) {
5728                 char *s = tsd_get(zfs_allow_log_key);
5729                 if (s != NULL)
5730                         strfree(s);
5731                 (void) tsd_set(zfs_allow_log_key, saved_poolname);
5732         } else {
5733                 if (saved_poolname != NULL)
5734                         strfree(saved_poolname);
5735         }
5736 
5737         kmem_free(zc, sizeof (zfs_cmd_t));
5738         return (error);
5739 }
5740 
5741 static int
5742 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
5743 {
5744         if (cmd != DDI_ATTACH)
5745                 return (DDI_FAILURE);
5746 
5747         if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
5748             DDI_PSEUDO, 0) == DDI_FAILURE)
5749                 return (DDI_FAILURE);
5750 
5751         zfs_dip = dip;
5752 
5753         ddi_report_dev(dip);
5754 
5755         return (DDI_SUCCESS);
5756 }
5757 
5758 static int
5759 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
5760 {
5761         if (spa_busy() || zfs_busy() || zvol_busy())
5762                 return (DDI_FAILURE);
5763 
5764         if (cmd != DDI_DETACH)
5765                 return (DDI_FAILURE);
5766 
5767         zfs_dip = NULL;
5768 
5769         ddi_prop_remove_all(dip);
5770         ddi_remove_minor_node(dip, NULL);
5771 
5772         return (DDI_SUCCESS);
5773 }
5774 
5775 /*ARGSUSED*/
5776 static int
5777 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
5778 {
5779         switch (infocmd) {
5780         case DDI_INFO_DEVT2DEVINFO:
5781                 *result = zfs_dip;
5782                 return (DDI_SUCCESS);
5783 
5784         case DDI_INFO_DEVT2INSTANCE:
5785                 *result = (void *)0;
5786                 return (DDI_SUCCESS);
5787         }
5788 
5789         return (DDI_FAILURE);
5790 }
5791 
5792 /*
5793  * OK, so this is a little weird.
5794  *
5795  * /dev/zfs is the control node, i.e. minor 0.
5796  * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
5797  *
5798  * /dev/zfs has basically nothing to do except serve up ioctls,
5799  * so most of the standard driver entry points are in zvol.c.
5800  */
5801 static struct cb_ops zfs_cb_ops = {
5802         zfsdev_open,    /* open */
5803         zfsdev_close,   /* close */
5804         zvol_strategy,  /* strategy */
5805         nodev,          /* print */
5806         zvol_dump,      /* dump */
5807         zvol_read,      /* read */
5808         zvol_write,     /* write */
5809         zfsdev_ioctl,   /* ioctl */
5810         nodev,          /* devmap */
5811         nodev,          /* mmap */
5812         nodev,          /* segmap */
5813         nochpoll,       /* poll */
5814         ddi_prop_op,    /* prop_op */
5815         NULL,           /* streamtab */
5816         D_NEW | D_MP | D_64BIT,         /* Driver compatibility flag */
5817         CB_REV,         /* version */
5818         nodev,          /* async read */
5819         nodev,          /* async write */
5820 };
5821 
5822 static struct dev_ops zfs_dev_ops = {
5823         DEVO_REV,       /* version */
5824         0,              /* refcnt */
5825         zfs_info,       /* info */
5826         nulldev,        /* identify */
5827         nulldev,        /* probe */
5828         zfs_attach,     /* attach */
5829         zfs_detach,     /* detach */
5830         nodev,          /* reset */
5831         &zfs_cb_ops,        /* driver operations */
5832         NULL,           /* no bus operations */
5833         NULL,           /* power */
5834         ddi_quiesce_not_needed, /* quiesce */
5835 };
5836 
5837 static struct modldrv zfs_modldrv = {
5838         &mod_driverops,
5839         "ZFS storage pool",
5840         &zfs_dev_ops
5841 };
5842 
5843 static struct modlinkage modlinkage = {
5844         MODREV_1,
5845         (void *)&zfs_modlfs,
5846         (void *)&zfs_modldrv,
5847         NULL
5848 };
5849 
5850 static void
5851 zfs_allow_log_destroy(void *arg)
5852 {
5853         char *poolname = arg;
5854         strfree(poolname);
5855 }
5856 
5857 int
5858 _init(void)
5859 {
5860         int error;
5861 
5862         spa_init(FREAD | FWRITE);
5863         zfs_init();
5864         zvol_init();
5865         zfs_ioctl_init();
5866 
5867         if ((error = mod_install(&modlinkage)) != 0) {
5868                 zvol_fini();
5869                 zfs_fini();
5870                 spa_fini();
5871                 return (error);
5872         }
5873 
5874         tsd_create(&zfs_fsyncer_key, NULL);
5875         tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
5876         tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
5877 
5878         error = ldi_ident_from_mod(&modlinkage, &zfs_li);
5879         ASSERT(error == 0);
5880         mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
5881 
5882         return (0);
5883 }
5884 
5885 int
5886 _fini(void)
5887 {
5888         int error;
5889 
5890         if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
5891                 return (SET_ERROR(EBUSY));
5892 
5893         if ((error = mod_remove(&modlinkage)) != 0)
5894                 return (error);
5895 
5896         zvol_fini();
5897         zfs_fini();
5898         spa_fini();
5899         if (zfs_nfsshare_inited)
5900                 (void) ddi_modclose(nfs_mod);
5901         if (zfs_smbshare_inited)
5902                 (void) ddi_modclose(smbsrv_mod);
5903         if (zfs_nfsshare_inited || zfs_smbshare_inited)
5904                 (void) ddi_modclose(sharefs_mod);
5905 
5906         tsd_destroy(&zfs_fsyncer_key);
5907         ldi_ident_release(zfs_li);
5908         zfs_li = NULL;
5909         mutex_destroy(&zfs_share_lock);
5910 
5911         return (error);
5912 }
5913 
5914 int
5915 _info(struct modinfo *modinfop)
5916 {
5917         return (mod_info(&modlinkage, modinfop));
5918 }