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                         /*
2004                          * If the zvol's parent dataset was being destroyed
2005                          * when we called zvol_get_stats, then it's possible
2006                          * that the ZAP still existed but its blocks had
2007                          * been already been freed when we tried to read it.
2008                          * It would then appear that the ZAP had no entries.
2009                          */
2010                         if (error == ENOENT)
2011                                 return (error);
2012                         VERIFY0(error);
2013                 }
2014                 error = put_nvlist(zc, nv);
2015                 nvlist_free(nv);
2016         }
2017 
2018         return (error);
2019 }
2020 
2021 /*
2022  * inputs:
2023  * zc_name              name of filesystem
2024  * zc_nvlist_dst_size   size of buffer for property nvlist
2025  *
2026  * outputs:
2027  * zc_objset_stats      stats
2028  * zc_nvlist_dst        property nvlist
2029  * zc_nvlist_dst_size   size of property nvlist
2030  */
2031 static int
2032 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2033 {
2034         objset_t *os;
2035         int error;
2036 
2037         error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2038         if (error == 0) {
2039                 error = zfs_ioc_objset_stats_impl(zc, os);
2040                 dmu_objset_rele(os, FTAG);
2041         }
2042 
2043         return (error);
2044 }
2045 
2046 /*
2047  * inputs:
2048  * zc_name              name of filesystem
2049  * zc_nvlist_dst_size   size of buffer for property nvlist
2050  *
2051  * outputs:
2052  * zc_nvlist_dst        received property nvlist
2053  * zc_nvlist_dst_size   size of received property nvlist
2054  *
2055  * Gets received properties (distinct from local properties on or after
2056  * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2057  * local property values.
2058  */
2059 static int
2060 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2061 {
2062         int error = 0;
2063         nvlist_t *nv;
2064 
2065         /*
2066          * Without this check, we would return local property values if the
2067          * caller has not already received properties on or after
2068          * SPA_VERSION_RECVD_PROPS.
2069          */
2070         if (!dsl_prop_get_hasrecvd(zc->zc_name))
2071                 return (SET_ERROR(ENOTSUP));
2072 
2073         if (zc->zc_nvlist_dst != 0 &&
2074             (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2075                 error = put_nvlist(zc, nv);
2076                 nvlist_free(nv);
2077         }
2078 
2079         return (error);
2080 }
2081 
2082 static int
2083 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2084 {
2085         uint64_t value;
2086         int error;
2087 
2088         /*
2089          * zfs_get_zplprop() will either find a value or give us
2090          * the default value (if there is one).
2091          */
2092         if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2093                 return (error);
2094         VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2095         return (0);
2096 }
2097 
2098 /*
2099  * inputs:
2100  * zc_name              name of filesystem
2101  * zc_nvlist_dst_size   size of buffer for zpl property nvlist
2102  *
2103  * outputs:
2104  * zc_nvlist_dst        zpl property nvlist
2105  * zc_nvlist_dst_size   size of zpl property nvlist
2106  */
2107 static int
2108 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2109 {
2110         objset_t *os;
2111         int err;
2112 
2113         /* XXX reading without owning */
2114         if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2115                 return (err);
2116 
2117         dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2118 
2119         /*
2120          * NB: nvl_add_zplprop() will read the objset contents,
2121          * which we aren't supposed to do with a DS_MODE_USER
2122          * hold, because it could be inconsistent.
2123          */
2124         if (zc->zc_nvlist_dst != NULL &&
2125             !zc->zc_objset_stats.dds_inconsistent &&
2126             dmu_objset_type(os) == DMU_OST_ZFS) {
2127                 nvlist_t *nv;
2128 
2129                 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2130                 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2131                     (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2132                     (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2133                     (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2134                         err = put_nvlist(zc, nv);
2135                 nvlist_free(nv);
2136         } else {
2137                 err = SET_ERROR(ENOENT);
2138         }
2139         dmu_objset_rele(os, FTAG);
2140         return (err);
2141 }
2142 
2143 static boolean_t
2144 dataset_name_hidden(const char *name)
2145 {
2146         /*
2147          * Skip over datasets that are not visible in this zone,
2148          * internal datasets (which have a $ in their name), and
2149          * temporary datasets (which have a % in their name).
2150          */
2151         if (strchr(name, '$') != NULL)
2152                 return (B_TRUE);
2153         if (strchr(name, '%') != NULL)
2154                 return (B_TRUE);
2155         if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
2156                 return (B_TRUE);
2157         return (B_FALSE);
2158 }
2159 
2160 /*
2161  * inputs:
2162  * zc_name              name of filesystem
2163  * zc_cookie            zap cursor
2164  * zc_nvlist_dst_size   size of buffer for property nvlist
2165  *
2166  * outputs:
2167  * zc_name              name of next filesystem
2168  * zc_cookie            zap cursor
2169  * zc_objset_stats      stats
2170  * zc_nvlist_dst        property nvlist
2171  * zc_nvlist_dst_size   size of property nvlist
2172  */
2173 static int
2174 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2175 {
2176         objset_t *os;
2177         int error;
2178         char *p;
2179         size_t orig_len = strlen(zc->zc_name);
2180 
2181 top:
2182         if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2183                 if (error == ENOENT)
2184                         error = SET_ERROR(ESRCH);
2185                 return (error);
2186         }
2187 
2188         p = strrchr(zc->zc_name, '/');
2189         if (p == NULL || p[1] != '\0')
2190                 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2191         p = zc->zc_name + strlen(zc->zc_name);
2192 
2193         do {
2194                 error = dmu_dir_list_next(os,
2195                     sizeof (zc->zc_name) - (p - zc->zc_name), p,
2196                     NULL, &zc->zc_cookie);
2197                 if (error == ENOENT)
2198                         error = SET_ERROR(ESRCH);
2199         } while (error == 0 && dataset_name_hidden(zc->zc_name));
2200         dmu_objset_rele(os, FTAG);
2201 
2202         /*
2203          * If it's an internal dataset (ie. with a '$' in its name),
2204          * don't try to get stats for it, otherwise we'll return ENOENT.
2205          */
2206         if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2207                 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2208                 if (error == ENOENT) {
2209                         /* We lost a race with destroy, get the next one. */
2210                         zc->zc_name[orig_len] = '\0';
2211                         goto top;
2212                 }
2213         }
2214         return (error);
2215 }
2216 
2217 /*
2218  * inputs:
2219  * zc_name              name of filesystem
2220  * zc_cookie            zap cursor
2221  * zc_nvlist_dst_size   size of buffer for property nvlist
2222  *
2223  * outputs:
2224  * zc_name              name of next snapshot
2225  * zc_objset_stats      stats
2226  * zc_nvlist_dst        property nvlist
2227  * zc_nvlist_dst_size   size of property nvlist
2228  */
2229 static int
2230 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2231 {
2232         objset_t *os;
2233         int error;
2234 
2235         error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2236         if (error != 0) {
2237                 return (error == ENOENT ? ESRCH : error);
2238         }
2239 
2240         /*
2241          * A dataset name of maximum length cannot have any snapshots,
2242          * so exit immediately.
2243          */
2244         if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2245                 dmu_objset_rele(os, FTAG);
2246                 return (SET_ERROR(ESRCH));
2247         }
2248 
2249         error = dmu_snapshot_list_next(os,
2250             sizeof (zc->zc_name) - strlen(zc->zc_name),
2251             zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2252             NULL);
2253 
2254         if (error == 0) {
2255                 dsl_dataset_t *ds;
2256                 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2257 
2258                 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2259                 if (error == 0) {
2260                         objset_t *ossnap;
2261 
2262                         error = dmu_objset_from_ds(ds, &ossnap);
2263                         if (error == 0)
2264                                 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2265                         dsl_dataset_rele(ds, FTAG);
2266                 }
2267         } else if (error == ENOENT) {
2268                 error = SET_ERROR(ESRCH);
2269         }
2270 
2271         dmu_objset_rele(os, FTAG);
2272         /* if we failed, undo the @ that we tacked on to zc_name */
2273         if (error != 0)
2274                 *strchr(zc->zc_name, '@') = '\0';
2275         return (error);
2276 }
2277 
2278 static int
2279 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2280 {
2281         const char *propname = nvpair_name(pair);
2282         uint64_t *valary;
2283         unsigned int vallen;
2284         const char *domain;
2285         char *dash;
2286         zfs_userquota_prop_t type;
2287         uint64_t rid;
2288         uint64_t quota;
2289         zfsvfs_t *zfsvfs;
2290         int err;
2291 
2292         if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2293                 nvlist_t *attrs;
2294                 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2295                 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2296                     &pair) != 0)
2297                         return (SET_ERROR(EINVAL));
2298         }
2299 
2300         /*
2301          * A correctly constructed propname is encoded as
2302          * userquota@<rid>-<domain>.
2303          */
2304         if ((dash = strchr(propname, '-')) == NULL ||
2305             nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2306             vallen != 3)
2307                 return (SET_ERROR(EINVAL));
2308 
2309         domain = dash + 1;
2310         type = valary[0];
2311         rid = valary[1];
2312         quota = valary[2];
2313 
2314         err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2315         if (err == 0) {
2316                 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2317                 zfsvfs_rele(zfsvfs, FTAG);
2318         }
2319 
2320         return (err);
2321 }
2322 
2323 /*
2324  * If the named property is one that has a special function to set its value,
2325  * return 0 on success and a positive error code on failure; otherwise if it is
2326  * not one of the special properties handled by this function, return -1.
2327  *
2328  * XXX: It would be better for callers of the property interface if we handled
2329  * these special cases in dsl_prop.c (in the dsl layer).
2330  */
2331 static int
2332 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2333     nvpair_t *pair)
2334 {
2335         const char *propname = nvpair_name(pair);
2336         zfs_prop_t prop = zfs_name_to_prop(propname);
2337         uint64_t intval;
2338         int err;
2339 
2340         if (prop == ZPROP_INVAL) {
2341                 if (zfs_prop_userquota(propname))
2342                         return (zfs_prop_set_userquota(dsname, pair));
2343                 return (-1);
2344         }
2345 
2346         if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2347                 nvlist_t *attrs;
2348                 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2349                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2350                     &pair) == 0);
2351         }
2352 
2353         if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2354                 return (-1);
2355 
2356         VERIFY(0 == nvpair_value_uint64(pair, &intval));
2357 
2358         switch (prop) {
2359         case ZFS_PROP_QUOTA:
2360                 err = dsl_dir_set_quota(dsname, source, intval);
2361                 break;
2362         case ZFS_PROP_REFQUOTA:
2363                 err = dsl_dataset_set_refquota(dsname, source, intval);
2364                 break;
2365         case ZFS_PROP_RESERVATION:
2366                 err = dsl_dir_set_reservation(dsname, source, intval);
2367                 break;
2368         case ZFS_PROP_REFRESERVATION:
2369                 err = dsl_dataset_set_refreservation(dsname, source, intval);
2370                 break;
2371         case ZFS_PROP_VOLSIZE:
2372                 err = zvol_set_volsize(dsname, intval);
2373                 break;
2374         case ZFS_PROP_VERSION:
2375         {
2376                 zfsvfs_t *zfsvfs;
2377 
2378                 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2379                         break;
2380 
2381                 err = zfs_set_version(zfsvfs, intval);
2382                 zfsvfs_rele(zfsvfs, FTAG);
2383 
2384                 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2385                         zfs_cmd_t *zc;
2386 
2387                         zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2388                         (void) strcpy(zc->zc_name, dsname);
2389                         (void) zfs_ioc_userspace_upgrade(zc);
2390                         kmem_free(zc, sizeof (zfs_cmd_t));
2391                 }
2392                 break;
2393         }
2394         case ZFS_PROP_COMPRESSION:
2395         {
2396                 if (intval == ZIO_COMPRESS_LZ4) {
2397                         zfeature_info_t *feature =
2398                             &spa_feature_table[SPA_FEATURE_LZ4_COMPRESS];
2399                         spa_t *spa;
2400 
2401                         if ((err = spa_open(dsname, &spa, FTAG)) != 0)
2402                                 return (err);
2403 
2404                         /*
2405                          * Setting the LZ4 compression algorithm activates
2406                          * the feature.
2407                          */
2408                         if (!spa_feature_is_active(spa, feature)) {
2409                                 if ((err = zfs_prop_activate_feature(spa,
2410                                     feature)) != 0) {
2411                                         spa_close(spa, FTAG);
2412                                         return (err);
2413                                 }
2414                         }
2415 
2416                         spa_close(spa, FTAG);
2417                 }
2418                 /*
2419                  * We still want the default set action to be performed in the
2420                  * caller, we only performed zfeature settings here.
2421                  */
2422                 err = -1;
2423                 break;
2424         }
2425 
2426         default:
2427                 err = -1;
2428         }
2429 
2430         return (err);
2431 }
2432 
2433 /*
2434  * This function is best effort. If it fails to set any of the given properties,
2435  * it continues to set as many as it can and returns the last error
2436  * encountered. If the caller provides a non-NULL errlist, it will be filled in
2437  * with the list of names of all the properties that failed along with the
2438  * corresponding error numbers.
2439  *
2440  * If every property is set successfully, zero is returned and errlist is not
2441  * modified.
2442  */
2443 int
2444 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2445     nvlist_t *errlist)
2446 {
2447         nvpair_t *pair;
2448         nvpair_t *propval;
2449         int rv = 0;
2450         uint64_t intval;
2451         char *strval;
2452         nvlist_t *genericnvl = fnvlist_alloc();
2453         nvlist_t *retrynvl = fnvlist_alloc();
2454 
2455 retry:
2456         pair = NULL;
2457         while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2458                 const char *propname = nvpair_name(pair);
2459                 zfs_prop_t prop = zfs_name_to_prop(propname);
2460                 int err = 0;
2461 
2462                 /* decode the property value */
2463                 propval = pair;
2464                 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2465                         nvlist_t *attrs;
2466                         attrs = fnvpair_value_nvlist(pair);
2467                         if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2468                             &propval) != 0)
2469                                 err = SET_ERROR(EINVAL);
2470                 }
2471 
2472                 /* Validate value type */
2473                 if (err == 0 && prop == ZPROP_INVAL) {
2474                         if (zfs_prop_user(propname)) {
2475                                 if (nvpair_type(propval) != DATA_TYPE_STRING)
2476                                         err = SET_ERROR(EINVAL);
2477                         } else if (zfs_prop_userquota(propname)) {
2478                                 if (nvpair_type(propval) !=
2479                                     DATA_TYPE_UINT64_ARRAY)
2480                                         err = SET_ERROR(EINVAL);
2481                         } else {
2482                                 err = SET_ERROR(EINVAL);
2483                         }
2484                 } else if (err == 0) {
2485                         if (nvpair_type(propval) == DATA_TYPE_STRING) {
2486                                 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2487                                         err = SET_ERROR(EINVAL);
2488                         } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2489                                 const char *unused;
2490 
2491                                 intval = fnvpair_value_uint64(propval);
2492 
2493                                 switch (zfs_prop_get_type(prop)) {
2494                                 case PROP_TYPE_NUMBER:
2495                                         break;
2496                                 case PROP_TYPE_STRING:
2497                                         err = SET_ERROR(EINVAL);
2498                                         break;
2499                                 case PROP_TYPE_INDEX:
2500                                         if (zfs_prop_index_to_string(prop,
2501                                             intval, &unused) != 0)
2502                                                 err = SET_ERROR(EINVAL);
2503                                         break;
2504                                 default:
2505                                         cmn_err(CE_PANIC,
2506                                             "unknown property type");
2507                                 }
2508                         } else {
2509                                 err = SET_ERROR(EINVAL);
2510                         }
2511                 }
2512 
2513                 /* Validate permissions */
2514                 if (err == 0)
2515                         err = zfs_check_settable(dsname, pair, CRED());
2516 
2517                 if (err == 0) {
2518                         err = zfs_prop_set_special(dsname, source, pair);
2519                         if (err == -1) {
2520                                 /*
2521                                  * For better performance we build up a list of
2522                                  * properties to set in a single transaction.
2523                                  */
2524                                 err = nvlist_add_nvpair(genericnvl, pair);
2525                         } else if (err != 0 && nvl != retrynvl) {
2526                                 /*
2527                                  * This may be a spurious error caused by
2528                                  * receiving quota and reservation out of order.
2529                                  * Try again in a second pass.
2530                                  */
2531                                 err = nvlist_add_nvpair(retrynvl, pair);
2532                         }
2533                 }
2534 
2535                 if (err != 0) {
2536                         if (errlist != NULL)
2537                                 fnvlist_add_int32(errlist, propname, err);
2538                         rv = err;
2539                 }
2540         }
2541 
2542         if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2543                 nvl = retrynvl;
2544                 goto retry;
2545         }
2546 
2547         if (!nvlist_empty(genericnvl) &&
2548             dsl_props_set(dsname, source, genericnvl) != 0) {
2549                 /*
2550                  * If this fails, we still want to set as many properties as we
2551                  * can, so try setting them individually.
2552                  */
2553                 pair = NULL;
2554                 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2555                         const char *propname = nvpair_name(pair);
2556                         int err = 0;
2557 
2558                         propval = pair;
2559                         if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2560                                 nvlist_t *attrs;
2561                                 attrs = fnvpair_value_nvlist(pair);
2562                                 propval = fnvlist_lookup_nvpair(attrs,
2563                                     ZPROP_VALUE);
2564                         }
2565 
2566                         if (nvpair_type(propval) == DATA_TYPE_STRING) {
2567                                 strval = fnvpair_value_string(propval);
2568                                 err = dsl_prop_set_string(dsname, propname,
2569                                     source, strval);
2570                         } else {
2571                                 intval = fnvpair_value_uint64(propval);
2572                                 err = dsl_prop_set_int(dsname, propname, source,
2573                                     intval);
2574                         }
2575 
2576                         if (err != 0) {
2577                                 if (errlist != NULL) {
2578                                         fnvlist_add_int32(errlist, propname,
2579                                             err);
2580                                 }
2581                                 rv = err;
2582                         }
2583                 }
2584         }
2585         nvlist_free(genericnvl);
2586         nvlist_free(retrynvl);
2587 
2588         return (rv);
2589 }
2590 
2591 /*
2592  * Check that all the properties are valid user properties.
2593  */
2594 static int
2595 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2596 {
2597         nvpair_t *pair = NULL;
2598         int error = 0;
2599 
2600         while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2601                 const char *propname = nvpair_name(pair);
2602                 char *valstr;
2603 
2604                 if (!zfs_prop_user(propname) ||
2605                     nvpair_type(pair) != DATA_TYPE_STRING)
2606                         return (SET_ERROR(EINVAL));
2607 
2608                 if (error = zfs_secpolicy_write_perms(fsname,
2609                     ZFS_DELEG_PERM_USERPROP, CRED()))
2610                         return (error);
2611 
2612                 if (strlen(propname) >= ZAP_MAXNAMELEN)
2613                         return (SET_ERROR(ENAMETOOLONG));
2614 
2615                 VERIFY(nvpair_value_string(pair, &valstr) == 0);
2616                 if (strlen(valstr) >= ZAP_MAXVALUELEN)
2617                         return (E2BIG);
2618         }
2619         return (0);
2620 }
2621 
2622 static void
2623 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2624 {
2625         nvpair_t *pair;
2626 
2627         VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2628 
2629         pair = NULL;
2630         while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2631                 if (nvlist_exists(skipped, nvpair_name(pair)))
2632                         continue;
2633 
2634                 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2635         }
2636 }
2637 
2638 static int
2639 clear_received_props(const char *dsname, nvlist_t *props,
2640     nvlist_t *skipped)
2641 {
2642         int err = 0;
2643         nvlist_t *cleared_props = NULL;
2644         props_skip(props, skipped, &cleared_props);
2645         if (!nvlist_empty(cleared_props)) {
2646                 /*
2647                  * Acts on local properties until the dataset has received
2648                  * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2649                  */
2650                 zprop_source_t flags = (ZPROP_SRC_NONE |
2651                     (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2652                 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2653         }
2654         nvlist_free(cleared_props);
2655         return (err);
2656 }
2657 
2658 /*
2659  * inputs:
2660  * zc_name              name of filesystem
2661  * zc_value             name of property to set
2662  * zc_nvlist_src{_size} nvlist of properties to apply
2663  * zc_cookie            received properties flag
2664  *
2665  * outputs:
2666  * zc_nvlist_dst{_size} error for each unapplied received property
2667  */
2668 static int
2669 zfs_ioc_set_prop(zfs_cmd_t *zc)
2670 {
2671         nvlist_t *nvl;
2672         boolean_t received = zc->zc_cookie;
2673         zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2674             ZPROP_SRC_LOCAL);
2675         nvlist_t *errors;
2676         int error;
2677 
2678         if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2679             zc->zc_iflags, &nvl)) != 0)
2680                 return (error);
2681 
2682         if (received) {
2683                 nvlist_t *origprops;
2684 
2685                 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2686                         (void) clear_received_props(zc->zc_name,
2687                             origprops, nvl);
2688                         nvlist_free(origprops);
2689                 }
2690 
2691                 error = dsl_prop_set_hasrecvd(zc->zc_name);
2692         }
2693 
2694         errors = fnvlist_alloc();
2695         if (error == 0)
2696                 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2697 
2698         if (zc->zc_nvlist_dst != NULL && errors != NULL) {
2699                 (void) put_nvlist(zc, errors);
2700         }
2701 
2702         nvlist_free(errors);
2703         nvlist_free(nvl);
2704         return (error);
2705 }
2706 
2707 /*
2708  * inputs:
2709  * zc_name              name of filesystem
2710  * zc_value             name of property to inherit
2711  * zc_cookie            revert to received value if TRUE
2712  *
2713  * outputs:             none
2714  */
2715 static int
2716 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2717 {
2718         const char *propname = zc->zc_value;
2719         zfs_prop_t prop = zfs_name_to_prop(propname);
2720         boolean_t received = zc->zc_cookie;
2721         zprop_source_t source = (received
2722             ? ZPROP_SRC_NONE            /* revert to received value, if any */
2723             : ZPROP_SRC_INHERITED);     /* explicitly inherit */
2724 
2725         if (received) {
2726                 nvlist_t *dummy;
2727                 nvpair_t *pair;
2728                 zprop_type_t type;
2729                 int err;
2730 
2731                 /*
2732                  * zfs_prop_set_special() expects properties in the form of an
2733                  * nvpair with type info.
2734                  */
2735                 if (prop == ZPROP_INVAL) {
2736                         if (!zfs_prop_user(propname))
2737                                 return (SET_ERROR(EINVAL));
2738 
2739                         type = PROP_TYPE_STRING;
2740                 } else if (prop == ZFS_PROP_VOLSIZE ||
2741                     prop == ZFS_PROP_VERSION) {
2742                         return (SET_ERROR(EINVAL));
2743                 } else {
2744                         type = zfs_prop_get_type(prop);
2745                 }
2746 
2747                 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2748 
2749                 switch (type) {
2750                 case PROP_TYPE_STRING:
2751                         VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2752                         break;
2753                 case PROP_TYPE_NUMBER:
2754                 case PROP_TYPE_INDEX:
2755                         VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2756                         break;
2757                 default:
2758                         nvlist_free(dummy);
2759                         return (SET_ERROR(EINVAL));
2760                 }
2761 
2762                 pair = nvlist_next_nvpair(dummy, NULL);
2763                 err = zfs_prop_set_special(zc->zc_name, source, pair);
2764                 nvlist_free(dummy);
2765                 if (err != -1)
2766                         return (err); /* special property already handled */
2767         } else {
2768                 /*
2769                  * Only check this in the non-received case. We want to allow
2770                  * 'inherit -S' to revert non-inheritable properties like quota
2771                  * and reservation to the received or default values even though
2772                  * they are not considered inheritable.
2773                  */
2774                 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2775                         return (SET_ERROR(EINVAL));
2776         }
2777 
2778         /* property name has been validated by zfs_secpolicy_inherit_prop() */
2779         return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2780 }
2781 
2782 static int
2783 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2784 {
2785         nvlist_t *props;
2786         spa_t *spa;
2787         int error;
2788         nvpair_t *pair;
2789 
2790         if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2791             zc->zc_iflags, &props))
2792                 return (error);
2793 
2794         /*
2795          * If the only property is the configfile, then just do a spa_lookup()
2796          * to handle the faulted case.
2797          */
2798         pair = nvlist_next_nvpair(props, NULL);
2799         if (pair != NULL && strcmp(nvpair_name(pair),
2800             zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2801             nvlist_next_nvpair(props, pair) == NULL) {
2802                 mutex_enter(&spa_namespace_lock);
2803                 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2804                         spa_configfile_set(spa, props, B_FALSE);
2805                         spa_config_sync(spa, B_FALSE, B_TRUE);
2806                 }
2807                 mutex_exit(&spa_namespace_lock);
2808                 if (spa != NULL) {
2809                         nvlist_free(props);
2810                         return (0);
2811                 }
2812         }
2813 
2814         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2815                 nvlist_free(props);
2816                 return (error);
2817         }
2818 
2819         error = spa_prop_set(spa, props);
2820 
2821         nvlist_free(props);
2822         spa_close(spa, FTAG);
2823 
2824         return (error);
2825 }
2826 
2827 static int
2828 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2829 {
2830         spa_t *spa;
2831         int error;
2832         nvlist_t *nvp = NULL;
2833 
2834         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2835                 /*
2836                  * If the pool is faulted, there may be properties we can still
2837                  * get (such as altroot and cachefile), so attempt to get them
2838                  * anyway.
2839                  */
2840                 mutex_enter(&spa_namespace_lock);
2841                 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2842                         error = spa_prop_get(spa, &nvp);
2843                 mutex_exit(&spa_namespace_lock);
2844         } else {
2845                 error = spa_prop_get(spa, &nvp);
2846                 spa_close(spa, FTAG);
2847         }
2848 
2849         if (error == 0 && zc->zc_nvlist_dst != NULL)
2850                 error = put_nvlist(zc, nvp);
2851         else
2852                 error = SET_ERROR(EFAULT);
2853 
2854         nvlist_free(nvp);
2855         return (error);
2856 }
2857 
2858 /*
2859  * inputs:
2860  * zc_name              name of filesystem
2861  * zc_nvlist_src{_size} nvlist of delegated permissions
2862  * zc_perm_action       allow/unallow flag
2863  *
2864  * outputs:             none
2865  */
2866 static int
2867 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2868 {
2869         int error;
2870         nvlist_t *fsaclnv = NULL;
2871 
2872         if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2873             zc->zc_iflags, &fsaclnv)) != 0)
2874                 return (error);
2875 
2876         /*
2877          * Verify nvlist is constructed correctly
2878          */
2879         if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2880                 nvlist_free(fsaclnv);
2881                 return (SET_ERROR(EINVAL));
2882         }
2883 
2884         /*
2885          * If we don't have PRIV_SYS_MOUNT, then validate
2886          * that user is allowed to hand out each permission in
2887          * the nvlist(s)
2888          */
2889 
2890         error = secpolicy_zfs(CRED());
2891         if (error != 0) {
2892                 if (zc->zc_perm_action == B_FALSE) {
2893                         error = dsl_deleg_can_allow(zc->zc_name,
2894                             fsaclnv, CRED());
2895                 } else {
2896                         error = dsl_deleg_can_unallow(zc->zc_name,
2897                             fsaclnv, CRED());
2898                 }
2899         }
2900 
2901         if (error == 0)
2902                 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2903 
2904         nvlist_free(fsaclnv);
2905         return (error);
2906 }
2907 
2908 /*
2909  * inputs:
2910  * zc_name              name of filesystem
2911  *
2912  * outputs:
2913  * zc_nvlist_src{_size} nvlist of delegated permissions
2914  */
2915 static int
2916 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2917 {
2918         nvlist_t *nvp;
2919         int error;
2920 
2921         if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2922                 error = put_nvlist(zc, nvp);
2923                 nvlist_free(nvp);
2924         }
2925 
2926         return (error);
2927 }
2928 
2929 /*
2930  * Search the vfs list for a specified resource.  Returns a pointer to it
2931  * or NULL if no suitable entry is found. The caller of this routine
2932  * is responsible for releasing the returned vfs pointer.
2933  */
2934 static vfs_t *
2935 zfs_get_vfs(const char *resource)
2936 {
2937         struct vfs *vfsp;
2938         struct vfs *vfs_found = NULL;
2939 
2940         vfs_list_read_lock();
2941         vfsp = rootvfs;
2942         do {
2943                 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
2944                         VFS_HOLD(vfsp);
2945                         vfs_found = vfsp;
2946                         break;
2947                 }
2948                 vfsp = vfsp->vfs_next;
2949         } while (vfsp != rootvfs);
2950         vfs_list_unlock();
2951         return (vfs_found);
2952 }
2953 
2954 /* ARGSUSED */
2955 static void
2956 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
2957 {
2958         zfs_creat_t *zct = arg;
2959 
2960         zfs_create_fs(os, cr, zct->zct_zplprops, tx);
2961 }
2962 
2963 #define ZFS_PROP_UNDEFINED      ((uint64_t)-1)
2964 
2965 /*
2966  * inputs:
2967  * createprops          list of properties requested by creator
2968  * default_zplver       zpl version to use if unspecified in createprops
2969  * fuids_ok             fuids allowed in this version of the spa?
2970  * os                   parent objset pointer (NULL if root fs)
2971  *
2972  * outputs:
2973  * zplprops     values for the zplprops we attach to the master node object
2974  * is_ci        true if requested file system will be purely case-insensitive
2975  *
2976  * Determine the settings for utf8only, normalization and
2977  * casesensitivity.  Specific values may have been requested by the
2978  * creator and/or we can inherit values from the parent dataset.  If
2979  * the file system is of too early a vintage, a creator can not
2980  * request settings for these properties, even if the requested
2981  * setting is the default value.  We don't actually want to create dsl
2982  * properties for these, so remove them from the source nvlist after
2983  * processing.
2984  */
2985 static int
2986 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
2987     boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
2988     nvlist_t *zplprops, boolean_t *is_ci)
2989 {
2990         uint64_t sense = ZFS_PROP_UNDEFINED;
2991         uint64_t norm = ZFS_PROP_UNDEFINED;
2992         uint64_t u8 = ZFS_PROP_UNDEFINED;
2993 
2994         ASSERT(zplprops != NULL);
2995 
2996         /*
2997          * Pull out creator prop choices, if any.
2998          */
2999         if (createprops) {
3000                 (void) nvlist_lookup_uint64(createprops,
3001                     zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3002                 (void) nvlist_lookup_uint64(createprops,
3003                     zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3004                 (void) nvlist_remove_all(createprops,
3005                     zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3006                 (void) nvlist_lookup_uint64(createprops,
3007                     zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3008                 (void) nvlist_remove_all(createprops,
3009                     zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3010                 (void) nvlist_lookup_uint64(createprops,
3011                     zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3012                 (void) nvlist_remove_all(createprops,
3013                     zfs_prop_to_name(ZFS_PROP_CASE));
3014         }
3015 
3016         /*
3017          * If the zpl version requested is whacky or the file system
3018          * or pool is version is too "young" to support normalization
3019          * and the creator tried to set a value for one of the props,
3020          * error out.
3021          */
3022         if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3023             (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3024             (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3025             (zplver < ZPL_VERSION_NORMALIZATION &&
3026             (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3027             sense != ZFS_PROP_UNDEFINED)))
3028                 return (SET_ERROR(ENOTSUP));
3029 
3030         /*
3031          * Put the version in the zplprops
3032          */
3033         VERIFY(nvlist_add_uint64(zplprops,
3034             zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3035 
3036         if (norm == ZFS_PROP_UNDEFINED)
3037                 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3038         VERIFY(nvlist_add_uint64(zplprops,
3039             zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3040 
3041         /*
3042          * If we're normalizing, names must always be valid UTF-8 strings.
3043          */
3044         if (norm)
3045                 u8 = 1;
3046         if (u8 == ZFS_PROP_UNDEFINED)
3047                 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3048         VERIFY(nvlist_add_uint64(zplprops,
3049             zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3050 
3051         if (sense == ZFS_PROP_UNDEFINED)
3052                 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3053         VERIFY(nvlist_add_uint64(zplprops,
3054             zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3055 
3056         if (is_ci)
3057                 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3058 
3059         return (0);
3060 }
3061 
3062 static int
3063 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3064     nvlist_t *zplprops, boolean_t *is_ci)
3065 {
3066         boolean_t fuids_ok, sa_ok;
3067         uint64_t zplver = ZPL_VERSION;
3068         objset_t *os = NULL;
3069         char parentname[MAXNAMELEN];
3070         char *cp;
3071         spa_t *spa;
3072         uint64_t spa_vers;
3073         int error;
3074 
3075         (void) strlcpy(parentname, dataset, sizeof (parentname));
3076         cp = strrchr(parentname, '/');
3077         ASSERT(cp != NULL);
3078         cp[0] = '\0';
3079 
3080         if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3081                 return (error);
3082 
3083         spa_vers = spa_version(spa);
3084         spa_close(spa, FTAG);
3085 
3086         zplver = zfs_zpl_version_map(spa_vers);
3087         fuids_ok = (zplver >= ZPL_VERSION_FUID);
3088         sa_ok = (zplver >= ZPL_VERSION_SA);
3089 
3090         /*
3091          * Open parent object set so we can inherit zplprop values.
3092          */
3093         if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3094                 return (error);
3095 
3096         error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3097             zplprops, is_ci);
3098         dmu_objset_rele(os, FTAG);
3099         return (error);
3100 }
3101 
3102 static int
3103 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3104     nvlist_t *zplprops, boolean_t *is_ci)
3105 {
3106         boolean_t fuids_ok;
3107         boolean_t sa_ok;
3108         uint64_t zplver = ZPL_VERSION;
3109         int error;
3110 
3111         zplver = zfs_zpl_version_map(spa_vers);
3112         fuids_ok = (zplver >= ZPL_VERSION_FUID);
3113         sa_ok = (zplver >= ZPL_VERSION_SA);
3114 
3115         error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3116             createprops, zplprops, is_ci);
3117         return (error);
3118 }
3119 
3120 /*
3121  * innvl: {
3122  *     "type" -> dmu_objset_type_t (int32)
3123  *     (optional) "props" -> { prop -> value }
3124  * }
3125  *
3126  * outnvl: propname -> error code (int32)
3127  */
3128 static int
3129 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3130 {
3131         int error = 0;
3132         zfs_creat_t zct = { 0 };
3133         nvlist_t *nvprops = NULL;
3134         void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3135         int32_t type32;
3136         dmu_objset_type_t type;
3137         boolean_t is_insensitive = B_FALSE;
3138 
3139         if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3140                 return (SET_ERROR(EINVAL));
3141         type = type32;
3142         (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3143 
3144         switch (type) {
3145         case DMU_OST_ZFS:
3146                 cbfunc = zfs_create_cb;
3147                 break;
3148 
3149         case DMU_OST_ZVOL:
3150                 cbfunc = zvol_create_cb;
3151                 break;
3152 
3153         default:
3154                 cbfunc = NULL;
3155                 break;
3156         }
3157         if (strchr(fsname, '@') ||
3158             strchr(fsname, '%'))
3159                 return (SET_ERROR(EINVAL));
3160 
3161         zct.zct_props = nvprops;
3162 
3163         if (cbfunc == NULL)
3164                 return (SET_ERROR(EINVAL));
3165 
3166         if (type == DMU_OST_ZVOL) {
3167                 uint64_t volsize, volblocksize;
3168 
3169                 if (nvprops == NULL)
3170                         return (SET_ERROR(EINVAL));
3171                 if (nvlist_lookup_uint64(nvprops,
3172                     zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3173                         return (SET_ERROR(EINVAL));
3174 
3175                 if ((error = nvlist_lookup_uint64(nvprops,
3176                     zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3177                     &volblocksize)) != 0 && error != ENOENT)
3178                         return (SET_ERROR(EINVAL));
3179 
3180                 if (error != 0)
3181                         volblocksize = zfs_prop_default_numeric(
3182                             ZFS_PROP_VOLBLOCKSIZE);
3183 
3184                 if ((error = zvol_check_volblocksize(
3185                     volblocksize)) != 0 ||
3186                     (error = zvol_check_volsize(volsize,
3187                     volblocksize)) != 0)
3188                         return (error);
3189         } else if (type == DMU_OST_ZFS) {
3190                 int error;
3191 
3192                 /*
3193                  * We have to have normalization and
3194                  * case-folding flags correct when we do the
3195                  * file system creation, so go figure them out
3196                  * now.
3197                  */
3198                 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3199                     NV_UNIQUE_NAME, KM_SLEEP) == 0);
3200                 error = zfs_fill_zplprops(fsname, nvprops,
3201                     zct.zct_zplprops, &is_insensitive);
3202                 if (error != 0) {
3203                         nvlist_free(zct.zct_zplprops);
3204                         return (error);
3205                 }
3206         }
3207 
3208         error = dmu_objset_create(fsname, type,
3209             is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3210         nvlist_free(zct.zct_zplprops);
3211 
3212         /*
3213          * It would be nice to do this atomically.
3214          */
3215         if (error == 0)
3216                 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3217                     nvprops, outnvl);
3218         if (error != 0)
3219                 (void) dsl_destroy_head(fsname);
3220         return (error);
3221 }
3222 
3223 /*
3224  * innvl: {
3225  *     "origin" -> name of origin snapshot
3226  *     (optional) "props" -> { prop -> value }
3227  * }
3228  *
3229  * outnvl: propname -> error code (int32)
3230  */
3231 static int
3232 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3233 {
3234         int error = 0;
3235         nvlist_t *nvprops = NULL;
3236         char *origin_name;
3237 
3238         if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3239                 return (SET_ERROR(EINVAL));
3240         (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3241 
3242         if (strchr(fsname, '@') ||
3243             strchr(fsname, '%'))
3244                 return (SET_ERROR(EINVAL));
3245 
3246         if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3247                 return (SET_ERROR(EINVAL));
3248         error = dmu_objset_clone(fsname, origin_name);
3249         if (error != 0)
3250                 return (error);
3251 
3252         /*
3253          * It would be nice to do this atomically.
3254          */
3255         if (error == 0) {
3256                 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3257                     nvprops, outnvl);
3258                 if (error != 0)
3259                         (void) dsl_destroy_head(fsname);
3260         }
3261         return (error);
3262 }
3263 
3264 /*
3265  * innvl: {
3266  *     "snaps" -> { snapshot1, snapshot2 }
3267  *     (optional) "props" -> { prop -> value (string) }
3268  * }
3269  *
3270  * outnvl: snapshot -> error code (int32)
3271  */
3272 static int
3273 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3274 {
3275         nvlist_t *snaps;
3276         nvlist_t *props = NULL;
3277         int error, poollen;
3278         nvpair_t *pair;
3279 
3280         (void) nvlist_lookup_nvlist(innvl, "props", &props);
3281         if ((error = zfs_check_userprops(poolname, props)) != 0)
3282                 return (error);
3283 
3284         if (!nvlist_empty(props) &&
3285             zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3286                 return (SET_ERROR(ENOTSUP));
3287 
3288         if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3289                 return (SET_ERROR(EINVAL));
3290         poollen = strlen(poolname);
3291         for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3292             pair = nvlist_next_nvpair(snaps, pair)) {
3293                 const char *name = nvpair_name(pair);
3294                 const char *cp = strchr(name, '@');
3295 
3296                 /*
3297                  * The snap name must contain an @, and the part after it must
3298                  * contain only valid characters.
3299                  */
3300                 if (cp == NULL || snapshot_namecheck(cp + 1, NULL, NULL) != 0)
3301                         return (SET_ERROR(EINVAL));
3302 
3303                 /*
3304                  * The snap must be in the specified pool.
3305                  */
3306                 if (strncmp(name, poolname, poollen) != 0 ||
3307                     (name[poollen] != '/' && name[poollen] != '@'))
3308                         return (SET_ERROR(EXDEV));
3309 
3310                 /* This must be the only snap of this fs. */
3311                 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3312                     pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3313                         if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3314                             == 0) {
3315                                 return (SET_ERROR(EXDEV));
3316                         }
3317                 }
3318         }
3319 
3320         error = dsl_dataset_snapshot(snaps, props, outnvl);
3321         return (error);
3322 }
3323 
3324 /*
3325  * innvl: "message" -> string
3326  */
3327 /* ARGSUSED */
3328 static int
3329 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3330 {
3331         char *message;
3332         spa_t *spa;
3333         int error;
3334         char *poolname;
3335 
3336         /*
3337          * The poolname in the ioctl is not set, we get it from the TSD,
3338          * which was set at the end of the last successful ioctl that allows
3339          * logging.  The secpolicy func already checked that it is set.
3340          * Only one log ioctl is allowed after each successful ioctl, so
3341          * we clear the TSD here.
3342          */
3343         poolname = tsd_get(zfs_allow_log_key);
3344         (void) tsd_set(zfs_allow_log_key, NULL);
3345         error = spa_open(poolname, &spa, FTAG);
3346         strfree(poolname);
3347         if (error != 0)
3348                 return (error);
3349 
3350         if (nvlist_lookup_string(innvl, "message", &message) != 0)  {
3351                 spa_close(spa, FTAG);
3352                 return (SET_ERROR(EINVAL));
3353         }
3354 
3355         if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3356                 spa_close(spa, FTAG);
3357                 return (SET_ERROR(ENOTSUP));
3358         }
3359 
3360         error = spa_history_log(spa, message);
3361         spa_close(spa, FTAG);
3362         return (error);
3363 }
3364 
3365 /*
3366  * The dp_config_rwlock must not be held when calling this, because the
3367  * unmount may need to write out data.
3368  *
3369  * This function is best-effort.  Callers must deal gracefully if it
3370  * remains mounted (or is remounted after this call).
3371  *
3372  * Returns 0 if the argument is not a snapshot, or it is not currently a
3373  * filesystem, or we were able to unmount it.  Returns error code otherwise.
3374  */
3375 int
3376 zfs_unmount_snap(const char *snapname)
3377 {
3378         vfs_t *vfsp;
3379         zfsvfs_t *zfsvfs;
3380         int err;
3381 
3382         if (strchr(snapname, '@') == NULL)
3383                 return (0);
3384 
3385         vfsp = zfs_get_vfs(snapname);
3386         if (vfsp == NULL)
3387                 return (0);
3388 
3389         zfsvfs = vfsp->vfs_data;
3390         ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os)));
3391 
3392         err = vn_vfswlock(vfsp->vfs_vnodecovered);
3393         VFS_RELE(vfsp);
3394         if (err != 0)
3395                 return (err);
3396 
3397         /*
3398          * Always force the unmount for snapshots.
3399          */
3400         (void) dounmount(vfsp, MS_FORCE, kcred);
3401         return (0);
3402 }
3403 
3404 /* ARGSUSED */
3405 static int
3406 zfs_unmount_snap_cb(const char *snapname, void *arg)
3407 {
3408         return (zfs_unmount_snap(snapname));
3409 }
3410 
3411 /*
3412  * When a clone is destroyed, its origin may also need to be destroyed,
3413  * in which case it must be unmounted.  This routine will do that unmount
3414  * if necessary.
3415  */
3416 void
3417 zfs_destroy_unmount_origin(const char *fsname)
3418 {
3419         int error;
3420         objset_t *os;
3421         dsl_dataset_t *ds;
3422 
3423         error = dmu_objset_hold(fsname, FTAG, &os);
3424         if (error != 0)
3425                 return;
3426         ds = dmu_objset_ds(os);
3427         if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3428                 char originname[MAXNAMELEN];
3429                 dsl_dataset_name(ds->ds_prev, originname);
3430                 dmu_objset_rele(os, FTAG);
3431                 (void) zfs_unmount_snap(originname);
3432         } else {
3433                 dmu_objset_rele(os, FTAG);
3434         }
3435 }
3436 
3437 /*
3438  * innvl: {
3439  *     "snaps" -> { snapshot1, snapshot2 }
3440  *     (optional boolean) "defer"
3441  * }
3442  *
3443  * outnvl: snapshot -> error code (int32)
3444  *
3445  */
3446 static int
3447 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3448 {
3449         int error, poollen;
3450         nvlist_t *snaps;
3451         nvpair_t *pair;
3452         boolean_t defer;
3453 
3454         if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3455                 return (SET_ERROR(EINVAL));
3456         defer = nvlist_exists(innvl, "defer");
3457 
3458         poollen = strlen(poolname);
3459         for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3460             pair = nvlist_next_nvpair(snaps, pair)) {
3461                 const char *name = nvpair_name(pair);
3462 
3463                 /*
3464                  * The snap must be in the specified pool.
3465                  */
3466                 if (strncmp(name, poolname, poollen) != 0 ||
3467                     (name[poollen] != '/' && name[poollen] != '@'))
3468                         return (SET_ERROR(EXDEV));
3469 
3470                 error = zfs_unmount_snap(name);
3471                 if (error)
3472                         return (SET_ERROR(error));
3473         }
3474 
3475         return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3476 }
3477 
3478 /*
3479  * inputs:
3480  * zc_name              name of dataset to destroy
3481  * zc_objset_type       type of objset
3482  * zc_defer_destroy     mark for deferred destroy
3483  *
3484  * outputs:             none
3485  */
3486 static int
3487 zfs_ioc_destroy(zfs_cmd_t *zc)
3488 {
3489         int err;
3490 
3491         if (zc->zc_objset_type == DMU_OST_ZFS) {
3492                 err = zfs_unmount_snap(zc->zc_name);
3493                 if (err)
3494                         return (err);
3495         }
3496 
3497         if (strchr(zc->zc_name, '@'))
3498                 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3499         else
3500                 err = dsl_destroy_head(zc->zc_name);
3501         if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3502                 (void) zvol_remove_minor(zc->zc_name);
3503         return (err);
3504 }
3505 
3506 /*
3507  * inputs:
3508  * zc_name      name of dataset to rollback (to most recent snapshot)
3509  *
3510  * outputs:     none
3511  */
3512 static int
3513 zfs_ioc_rollback(zfs_cmd_t *zc)
3514 {
3515         zfsvfs_t *zfsvfs;
3516         int error;
3517 
3518         if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
3519                 error = zfs_suspend_fs(zfsvfs);
3520                 if (error == 0) {
3521                         int resume_err;
3522 
3523                         error = dsl_dataset_rollback(zc->zc_name);
3524                         resume_err = zfs_resume_fs(zfsvfs, zc->zc_name);
3525                         error = error ? error : resume_err;
3526                 }
3527                 VFS_RELE(zfsvfs->z_vfs);
3528         } else {
3529                 error = dsl_dataset_rollback(zc->zc_name);
3530         }
3531         return (error);
3532 }
3533 
3534 static int
3535 recursive_unmount(const char *fsname, void *arg)
3536 {
3537         const char *snapname = arg;
3538         char fullname[MAXNAMELEN];
3539 
3540         (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname);
3541         return (zfs_unmount_snap(fullname));
3542 }
3543 
3544 /*
3545  * inputs:
3546  * zc_name      old name of dataset
3547  * zc_value     new name of dataset
3548  * zc_cookie    recursive flag (only valid for snapshots)
3549  *
3550  * outputs:     none
3551  */
3552 static int
3553 zfs_ioc_rename(zfs_cmd_t *zc)
3554 {
3555         boolean_t recursive = zc->zc_cookie & 1;
3556         char *at;
3557 
3558         zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3559         if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3560             strchr(zc->zc_value, '%'))
3561                 return (SET_ERROR(EINVAL));
3562 
3563         at = strchr(zc->zc_name, '@');
3564         if (at != NULL) {
3565                 /* snaps must be in same fs */
3566                 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3567                         return (SET_ERROR(EXDEV));
3568                 *at = '\0';
3569                 if (zc->zc_objset_type == DMU_OST_ZFS) {
3570                         int error = dmu_objset_find(zc->zc_name,
3571                             recursive_unmount, at + 1,
3572                             recursive ? DS_FIND_CHILDREN : 0);
3573                         if (error != 0)
3574                                 return (error);
3575                 }
3576                 return (dsl_dataset_rename_snapshot(zc->zc_name,
3577                     at + 1, strchr(zc->zc_value, '@') + 1, recursive));
3578         } else {
3579                 if (zc->zc_objset_type == DMU_OST_ZVOL)
3580                         (void) zvol_remove_minor(zc->zc_name);
3581                 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3582         }
3583 }
3584 
3585 static int
3586 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3587 {
3588         const char *propname = nvpair_name(pair);
3589         boolean_t issnap = (strchr(dsname, '@') != NULL);
3590         zfs_prop_t prop = zfs_name_to_prop(propname);
3591         uint64_t intval;
3592         int err;
3593 
3594         if (prop == ZPROP_INVAL) {
3595                 if (zfs_prop_user(propname)) {
3596                         if (err = zfs_secpolicy_write_perms(dsname,
3597                             ZFS_DELEG_PERM_USERPROP, cr))
3598                                 return (err);
3599                         return (0);
3600                 }
3601 
3602                 if (!issnap && zfs_prop_userquota(propname)) {
3603                         const char *perm = NULL;
3604                         const char *uq_prefix =
3605                             zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3606                         const char *gq_prefix =
3607                             zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3608 
3609                         if (strncmp(propname, uq_prefix,
3610                             strlen(uq_prefix)) == 0) {
3611                                 perm = ZFS_DELEG_PERM_USERQUOTA;
3612                         } else if (strncmp(propname, gq_prefix,
3613                             strlen(gq_prefix)) == 0) {
3614                                 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3615                         } else {
3616                                 /* USERUSED and GROUPUSED are read-only */
3617                                 return (SET_ERROR(EINVAL));
3618                         }
3619 
3620                         if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3621                                 return (err);
3622                         return (0);
3623                 }
3624 
3625                 return (SET_ERROR(EINVAL));
3626         }
3627 
3628         if (issnap)
3629                 return (SET_ERROR(EINVAL));
3630 
3631         if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3632                 /*
3633                  * dsl_prop_get_all_impl() returns properties in this
3634                  * format.
3635                  */
3636                 nvlist_t *attrs;
3637                 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3638                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3639                     &pair) == 0);
3640         }
3641 
3642         /*
3643          * Check that this value is valid for this pool version
3644          */
3645         switch (prop) {
3646         case ZFS_PROP_COMPRESSION:
3647                 /*
3648                  * If the user specified gzip compression, make sure
3649                  * the SPA supports it. We ignore any errors here since
3650                  * we'll catch them later.
3651                  */
3652                 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3653                     nvpair_value_uint64(pair, &intval) == 0) {
3654                         if (intval >= ZIO_COMPRESS_GZIP_1 &&
3655                             intval <= ZIO_COMPRESS_GZIP_9 &&
3656                             zfs_earlier_version(dsname,
3657                             SPA_VERSION_GZIP_COMPRESSION)) {
3658                                 return (SET_ERROR(ENOTSUP));
3659                         }
3660 
3661                         if (intval == ZIO_COMPRESS_ZLE &&
3662                             zfs_earlier_version(dsname,
3663                             SPA_VERSION_ZLE_COMPRESSION))
3664                                 return (SET_ERROR(ENOTSUP));
3665 
3666                         if (intval == ZIO_COMPRESS_LZ4) {
3667                                 zfeature_info_t *feature =
3668                                     &spa_feature_table[
3669                                     SPA_FEATURE_LZ4_COMPRESS];
3670                                 spa_t *spa;
3671 
3672                                 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3673                                         return (err);
3674 
3675                                 if (!spa_feature_is_enabled(spa, feature)) {
3676                                         spa_close(spa, FTAG);
3677                                         return (SET_ERROR(ENOTSUP));
3678                                 }
3679                                 spa_close(spa, FTAG);
3680                         }
3681 
3682                         /*
3683                          * If this is a bootable dataset then
3684                          * verify that the compression algorithm
3685                          * is supported for booting. We must return
3686                          * something other than ENOTSUP since it
3687                          * implies a downrev pool version.
3688                          */
3689                         if (zfs_is_bootfs(dsname) &&
3690                             !BOOTFS_COMPRESS_VALID(intval)) {
3691                                 return (SET_ERROR(ERANGE));
3692                         }
3693                 }
3694                 break;
3695 
3696         case ZFS_PROP_COPIES:
3697                 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3698                         return (SET_ERROR(ENOTSUP));
3699                 break;
3700 
3701         case ZFS_PROP_DEDUP:
3702                 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3703                         return (SET_ERROR(ENOTSUP));
3704                 break;
3705 
3706         case ZFS_PROP_SHARESMB:
3707                 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3708                         return (SET_ERROR(ENOTSUP));
3709                 break;
3710 
3711         case ZFS_PROP_ACLINHERIT:
3712                 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3713                     nvpair_value_uint64(pair, &intval) == 0) {
3714                         if (intval == ZFS_ACL_PASSTHROUGH_X &&
3715                             zfs_earlier_version(dsname,
3716                             SPA_VERSION_PASSTHROUGH_X))
3717                                 return (SET_ERROR(ENOTSUP));
3718                 }
3719                 break;
3720         }
3721 
3722         return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3723 }
3724 
3725 /*
3726  * Checks for a race condition to make sure we don't increment a feature flag
3727  * multiple times.
3728  */
3729 static int
3730 zfs_prop_activate_feature_check(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         if (!spa_feature_is_active(spa, feature))
3736                 return (0);
3737         else
3738                 return (SET_ERROR(EBUSY));
3739 }
3740 
3741 /*
3742  * The callback invoked on feature activation in the sync task caused by
3743  * zfs_prop_activate_feature.
3744  */
3745 static void
3746 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
3747 {
3748         spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3749         zfeature_info_t *feature = arg;
3750 
3751         spa_feature_incr(spa, feature, tx);
3752 }
3753 
3754 /*
3755  * Activates a feature on a pool in response to a property setting. This
3756  * creates a new sync task which modifies the pool to reflect the feature
3757  * as being active.
3758  */
3759 static int
3760 zfs_prop_activate_feature(spa_t *spa, zfeature_info_t *feature)
3761 {
3762         int err;
3763 
3764         /* EBUSY here indicates that the feature is already active */
3765         err = dsl_sync_task(spa_name(spa),
3766             zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
3767             feature, 2);
3768 
3769         if (err != 0 && err != EBUSY)
3770                 return (err);
3771         else
3772                 return (0);
3773 }
3774 
3775 /*
3776  * Removes properties from the given props list that fail permission checks
3777  * needed to clear them and to restore them in case of a receive error. For each
3778  * property, make sure we have both set and inherit permissions.
3779  *
3780  * Returns the first error encountered if any permission checks fail. If the
3781  * caller provides a non-NULL errlist, it also gives the complete list of names
3782  * of all the properties that failed a permission check along with the
3783  * corresponding error numbers. The caller is responsible for freeing the
3784  * returned errlist.
3785  *
3786  * If every property checks out successfully, zero is returned and the list
3787  * pointed at by errlist is NULL.
3788  */
3789 static int
3790 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
3791 {
3792         zfs_cmd_t *zc;
3793         nvpair_t *pair, *next_pair;
3794         nvlist_t *errors;
3795         int err, rv = 0;
3796 
3797         if (props == NULL)
3798                 return (0);
3799 
3800         VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3801 
3802         zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
3803         (void) strcpy(zc->zc_name, dataset);
3804         pair = nvlist_next_nvpair(props, NULL);
3805         while (pair != NULL) {
3806                 next_pair = nvlist_next_nvpair(props, pair);
3807 
3808                 (void) strcpy(zc->zc_value, nvpair_name(pair));
3809                 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
3810                     (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
3811                         VERIFY(nvlist_remove_nvpair(props, pair) == 0);
3812                         VERIFY(nvlist_add_int32(errors,
3813                             zc->zc_value, err) == 0);
3814                 }
3815                 pair = next_pair;
3816         }
3817         kmem_free(zc, sizeof (zfs_cmd_t));
3818 
3819         if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
3820                 nvlist_free(errors);
3821                 errors = NULL;
3822         } else {
3823                 VERIFY(nvpair_value_int32(pair, &rv) == 0);
3824         }
3825 
3826         if (errlist == NULL)
3827                 nvlist_free(errors);
3828         else
3829                 *errlist = errors;
3830 
3831         return (rv);
3832 }
3833 
3834 static boolean_t
3835 propval_equals(nvpair_t *p1, nvpair_t *p2)
3836 {
3837         if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
3838                 /* dsl_prop_get_all_impl() format */
3839                 nvlist_t *attrs;
3840                 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
3841                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3842                     &p1) == 0);
3843         }
3844 
3845         if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
3846                 nvlist_t *attrs;
3847                 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
3848                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3849                     &p2) == 0);
3850         }
3851 
3852         if (nvpair_type(p1) != nvpair_type(p2))
3853                 return (B_FALSE);
3854 
3855         if (nvpair_type(p1) == DATA_TYPE_STRING) {
3856                 char *valstr1, *valstr2;
3857 
3858                 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
3859                 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
3860                 return (strcmp(valstr1, valstr2) == 0);
3861         } else {
3862                 uint64_t intval1, intval2;
3863 
3864                 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
3865                 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
3866                 return (intval1 == intval2);
3867         }
3868 }
3869 
3870 /*
3871  * Remove properties from props if they are not going to change (as determined
3872  * by comparison with origprops). Remove them from origprops as well, since we
3873  * do not need to clear or restore properties that won't change.
3874  */
3875 static void
3876 props_reduce(nvlist_t *props, nvlist_t *origprops)
3877 {
3878         nvpair_t *pair, *next_pair;
3879 
3880         if (origprops == NULL)
3881                 return; /* all props need to be received */
3882 
3883         pair = nvlist_next_nvpair(props, NULL);
3884         while (pair != NULL) {
3885                 const char *propname = nvpair_name(pair);
3886                 nvpair_t *match;
3887 
3888                 next_pair = nvlist_next_nvpair(props, pair);
3889 
3890                 if ((nvlist_lookup_nvpair(origprops, propname,
3891                     &match) != 0) || !propval_equals(pair, match))
3892                         goto next; /* need to set received value */
3893 
3894                 /* don't clear the existing received value */
3895                 (void) nvlist_remove_nvpair(origprops, match);
3896                 /* don't bother receiving the property */
3897                 (void) nvlist_remove_nvpair(props, pair);
3898 next:
3899                 pair = next_pair;
3900         }
3901 }
3902 
3903 #ifdef  DEBUG
3904 static boolean_t zfs_ioc_recv_inject_err;
3905 #endif
3906 
3907 /*
3908  * inputs:
3909  * zc_name              name of containing filesystem
3910  * zc_nvlist_src{_size} nvlist of properties to apply
3911  * zc_value             name of snapshot to create
3912  * zc_string            name of clone origin (if DRR_FLAG_CLONE)
3913  * zc_cookie            file descriptor to recv from
3914  * zc_begin_record      the BEGIN record of the stream (not byteswapped)
3915  * zc_guid              force flag
3916  * zc_cleanup_fd        cleanup-on-exit file descriptor
3917  * zc_action_handle     handle for this guid/ds mapping (or zero on first call)
3918  *
3919  * outputs:
3920  * zc_cookie            number of bytes read
3921  * zc_nvlist_dst{_size} error for each unapplied received property
3922  * zc_obj               zprop_errflags_t
3923  * zc_action_handle     handle for this guid/ds mapping
3924  */
3925 static int
3926 zfs_ioc_recv(zfs_cmd_t *zc)
3927 {
3928         file_t *fp;
3929         dmu_recv_cookie_t drc;
3930         boolean_t force = (boolean_t)zc->zc_guid;
3931         int fd;
3932         int error = 0;
3933         int props_error = 0;
3934         nvlist_t *errors;
3935         offset_t off;
3936         nvlist_t *props = NULL; /* sent properties */
3937         nvlist_t *origprops = NULL; /* existing properties */
3938         char *origin = NULL;
3939         char *tosnap;
3940         char tofs[ZFS_MAXNAMELEN];
3941         boolean_t first_recvd_props = B_FALSE;
3942 
3943         if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3944             strchr(zc->zc_value, '@') == NULL ||
3945             strchr(zc->zc_value, '%'))
3946                 return (SET_ERROR(EINVAL));
3947 
3948         (void) strcpy(tofs, zc->zc_value);
3949         tosnap = strchr(tofs, '@');
3950         *tosnap++ = '\0';
3951 
3952         if (zc->zc_nvlist_src != NULL &&
3953             (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3954             zc->zc_iflags, &props)) != 0)
3955                 return (error);
3956 
3957         fd = zc->zc_cookie;
3958         fp = getf(fd);
3959         if (fp == NULL) {
3960                 nvlist_free(props);
3961                 return (SET_ERROR(EBADF));
3962         }
3963 
3964         VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3965 
3966         if (zc->zc_string[0])
3967                 origin = zc->zc_string;
3968 
3969         error = dmu_recv_begin(tofs, tosnap,
3970             &zc->zc_begin_record, force, origin, &drc);
3971         if (error != 0)
3972                 goto out;
3973 
3974         /*
3975          * Set properties before we receive the stream so that they are applied
3976          * to the new data. Note that we must call dmu_recv_stream() if
3977          * dmu_recv_begin() succeeds.
3978          */
3979         if (props != NULL && !drc.drc_newfs) {
3980                 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
3981                     SPA_VERSION_RECVD_PROPS &&
3982                     !dsl_prop_get_hasrecvd(tofs))
3983                         first_recvd_props = B_TRUE;
3984 
3985                 /*
3986                  * If new received properties are supplied, they are to
3987                  * completely replace the existing received properties, so stash
3988                  * away the existing ones.
3989                  */
3990                 if (dsl_prop_get_received(tofs, &origprops) == 0) {
3991                         nvlist_t *errlist = NULL;
3992                         /*
3993                          * Don't bother writing a property if its value won't
3994                          * change (and avoid the unnecessary security checks).
3995                          *
3996                          * The first receive after SPA_VERSION_RECVD_PROPS is a
3997                          * special case where we blow away all local properties
3998                          * regardless.
3999                          */
4000                         if (!first_recvd_props)
4001                                 props_reduce(props, origprops);
4002                         if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
4003                                 (void) nvlist_merge(errors, errlist, 0);
4004                         nvlist_free(errlist);
4005 
4006                         if (clear_received_props(tofs, origprops,
4007                             first_recvd_props ? NULL : props) != 0)
4008                                 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4009                 } else {
4010                         zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4011                 }
4012         }
4013 
4014         if (props != NULL) {
4015                 props_error = dsl_prop_set_hasrecvd(tofs);
4016 
4017                 if (props_error == 0) {
4018                         (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4019                             props, errors);
4020                 }
4021         }
4022 
4023         if (zc->zc_nvlist_dst_size != 0 &&
4024             (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4025             put_nvlist(zc, errors) != 0)) {
4026                 /*
4027                  * Caller made zc->zc_nvlist_dst less than the minimum expected
4028                  * size or supplied an invalid address.
4029                  */
4030                 props_error = SET_ERROR(EINVAL);
4031         }
4032 
4033         off = fp->f_offset;
4034         error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
4035             &zc->zc_action_handle);
4036 
4037         if (error == 0) {
4038                 zfsvfs_t *zfsvfs = NULL;
4039 
4040                 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4041                         /* online recv */
4042                         int end_err;
4043 
4044                         error = zfs_suspend_fs(zfsvfs);
4045                         /*
4046                          * If the suspend fails, then the recv_end will
4047                          * likely also fail, and clean up after itself.
4048                          */
4049                         end_err = dmu_recv_end(&drc);
4050                         if (error == 0)
4051                                 error = zfs_resume_fs(zfsvfs, tofs);
4052                         error = error ? error : end_err;
4053                         VFS_RELE(zfsvfs->z_vfs);
4054                 } else {
4055                         error = dmu_recv_end(&drc);
4056                 }
4057         }
4058 
4059         zc->zc_cookie = off - fp->f_offset;
4060         if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4061                 fp->f_offset = off;
4062 
4063 #ifdef  DEBUG
4064         if (zfs_ioc_recv_inject_err) {
4065                 zfs_ioc_recv_inject_err = B_FALSE;
4066                 error = 1;
4067         }
4068 #endif
4069         /*
4070          * On error, restore the original props.
4071          */
4072         if (error != 0 && props != NULL && !drc.drc_newfs) {
4073                 if (clear_received_props(tofs, props, NULL) != 0) {
4074                         /*
4075                          * We failed to clear the received properties.
4076                          * Since we may have left a $recvd value on the
4077                          * system, we can't clear the $hasrecvd flag.
4078                          */
4079                         zc->zc_obj |= ZPROP_ERR_NORESTORE;
4080                 } else if (first_recvd_props) {
4081                         dsl_prop_unset_hasrecvd(tofs);
4082                 }
4083 
4084                 if (origprops == NULL && !drc.drc_newfs) {
4085                         /* We failed to stash the original properties. */
4086                         zc->zc_obj |= ZPROP_ERR_NORESTORE;
4087                 }
4088 
4089                 /*
4090                  * dsl_props_set() will not convert RECEIVED to LOCAL on or
4091                  * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4092                  * explictly if we're restoring local properties cleared in the
4093                  * first new-style receive.
4094                  */
4095                 if (origprops != NULL &&
4096                     zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4097                     ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4098                     origprops, NULL) != 0) {
4099                         /*
4100                          * We stashed the original properties but failed to
4101                          * restore them.
4102                          */
4103                         zc->zc_obj |= ZPROP_ERR_NORESTORE;
4104                 }
4105         }
4106 out:
4107         nvlist_free(props);
4108         nvlist_free(origprops);
4109         nvlist_free(errors);
4110         releasef(fd);
4111 
4112         if (error == 0)
4113                 error = props_error;
4114 
4115         return (error);
4116 }
4117 
4118 /*
4119  * inputs:
4120  * zc_name      name of snapshot to send
4121  * zc_cookie    file descriptor to send stream to
4122  * zc_obj       fromorigin flag (mutually exclusive with zc_fromobj)
4123  * zc_sendobj   objsetid of snapshot to send
4124  * zc_fromobj   objsetid of incremental fromsnap (may be zero)
4125  * zc_guid      if set, estimate size of stream only.  zc_cookie is ignored.
4126  *              output size in zc_objset_type.
4127  *
4128  * outputs: none
4129  */
4130 static int
4131 zfs_ioc_send(zfs_cmd_t *zc)
4132 {
4133         int error;
4134         offset_t off;
4135         boolean_t estimate = (zc->zc_guid != 0);
4136 
4137         if (zc->zc_obj != 0) {
4138                 dsl_pool_t *dp;
4139                 dsl_dataset_t *tosnap;
4140 
4141                 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4142                 if (error != 0)
4143                         return (error);
4144 
4145                 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4146                 if (error != 0) {
4147                         dsl_pool_rele(dp, FTAG);
4148                         return (error);
4149                 }
4150 
4151                 if (dsl_dir_is_clone(tosnap->ds_dir))
4152                         zc->zc_fromobj = tosnap->ds_dir->dd_phys->dd_origin_obj;
4153                 dsl_dataset_rele(tosnap, FTAG);
4154                 dsl_pool_rele(dp, FTAG);
4155         }
4156 
4157         if (estimate) {
4158                 dsl_pool_t *dp;
4159                 dsl_dataset_t *tosnap;
4160                 dsl_dataset_t *fromsnap = NULL;
4161 
4162                 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4163                 if (error != 0)
4164                         return (error);
4165 
4166                 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4167                 if (error != 0) {
4168                         dsl_pool_rele(dp, FTAG);
4169                         return (error);
4170                 }
4171 
4172                 if (zc->zc_fromobj != 0) {
4173                         error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4174                             FTAG, &fromsnap);
4175                         if (error != 0) {
4176                                 dsl_dataset_rele(tosnap, FTAG);
4177                                 dsl_pool_rele(dp, FTAG);
4178                                 return (error);
4179                         }
4180                 }
4181 
4182                 error = dmu_send_estimate(tosnap, fromsnap,
4183                     &zc->zc_objset_type);
4184 
4185                 if (fromsnap != NULL)
4186                         dsl_dataset_rele(fromsnap, FTAG);
4187                 dsl_dataset_rele(tosnap, FTAG);
4188                 dsl_pool_rele(dp, FTAG);
4189         } else {
4190                 file_t *fp = getf(zc->zc_cookie);
4191                 if (fp == NULL)
4192                         return (SET_ERROR(EBADF));
4193 
4194                 off = fp->f_offset;
4195                 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4196                     zc->zc_fromobj, zc->zc_cookie, fp->f_vnode, &off);
4197 
4198                 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4199                         fp->f_offset = off;
4200                 releasef(zc->zc_cookie);
4201         }
4202         return (error);
4203 }
4204 
4205 /*
4206  * inputs:
4207  * zc_name      name of snapshot on which to report progress
4208  * zc_cookie    file descriptor of send stream
4209  *
4210  * outputs:
4211  * zc_cookie    number of bytes written in send stream thus far
4212  */
4213 static int
4214 zfs_ioc_send_progress(zfs_cmd_t *zc)
4215 {
4216         dsl_pool_t *dp;
4217         dsl_dataset_t *ds;
4218         dmu_sendarg_t *dsp = NULL;
4219         int error;
4220 
4221         error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4222         if (error != 0)
4223                 return (error);
4224 
4225         error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4226         if (error != 0) {
4227                 dsl_pool_rele(dp, FTAG);
4228                 return (error);
4229         }
4230 
4231         mutex_enter(&ds->ds_sendstream_lock);
4232 
4233         /*
4234          * Iterate over all the send streams currently active on this dataset.
4235          * If there's one which matches the specified file descriptor _and_ the
4236          * stream was started by the current process, return the progress of
4237          * that stream.
4238          */
4239         for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4240             dsp = list_next(&ds->ds_sendstreams, dsp)) {
4241                 if (dsp->dsa_outfd == zc->zc_cookie &&
4242                     dsp->dsa_proc == curproc)
4243                         break;
4244         }
4245 
4246         if (dsp != NULL)
4247                 zc->zc_cookie = *(dsp->dsa_off);
4248         else
4249                 error = SET_ERROR(ENOENT);
4250 
4251         mutex_exit(&ds->ds_sendstream_lock);
4252         dsl_dataset_rele(ds, FTAG);
4253         dsl_pool_rele(dp, FTAG);
4254         return (error);
4255 }
4256 
4257 static int
4258 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4259 {
4260         int id, error;
4261 
4262         error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4263             &zc->zc_inject_record);
4264 
4265         if (error == 0)
4266                 zc->zc_guid = (uint64_t)id;
4267 
4268         return (error);
4269 }
4270 
4271 static int
4272 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4273 {
4274         return (zio_clear_fault((int)zc->zc_guid));
4275 }
4276 
4277 static int
4278 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4279 {
4280         int id = (int)zc->zc_guid;
4281         int error;
4282 
4283         error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4284             &zc->zc_inject_record);
4285 
4286         zc->zc_guid = id;
4287 
4288         return (error);
4289 }
4290 
4291 static int
4292 zfs_ioc_error_log(zfs_cmd_t *zc)
4293 {
4294         spa_t *spa;
4295         int error;
4296         size_t count = (size_t)zc->zc_nvlist_dst_size;
4297 
4298         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4299                 return (error);
4300 
4301         error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4302             &count);
4303         if (error == 0)
4304                 zc->zc_nvlist_dst_size = count;
4305         else
4306                 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4307 
4308         spa_close(spa, FTAG);
4309 
4310         return (error);
4311 }
4312 
4313 static int
4314 zfs_ioc_clear(zfs_cmd_t *zc)
4315 {
4316         spa_t *spa;
4317         vdev_t *vd;
4318         int error;
4319 
4320         /*
4321          * On zpool clear we also fix up missing slogs
4322          */
4323         mutex_enter(&spa_namespace_lock);
4324         spa = spa_lookup(zc->zc_name);
4325         if (spa == NULL) {
4326                 mutex_exit(&spa_namespace_lock);
4327                 return (SET_ERROR(EIO));
4328         }
4329         if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4330                 /* we need to let spa_open/spa_load clear the chains */
4331                 spa_set_log_state(spa, SPA_LOG_CLEAR);
4332         }
4333         spa->spa_last_open_failed = 0;
4334         mutex_exit(&spa_namespace_lock);
4335 
4336         if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4337                 error = spa_open(zc->zc_name, &spa, FTAG);
4338         } else {
4339                 nvlist_t *policy;
4340                 nvlist_t *config = NULL;
4341 
4342                 if (zc->zc_nvlist_src == NULL)
4343                         return (SET_ERROR(EINVAL));
4344 
4345                 if ((error = get_nvlist(zc->zc_nvlist_src,
4346                     zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4347                         error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4348                             policy, &config);
4349                         if (config != NULL) {
4350                                 int err;
4351 
4352                                 if ((err = put_nvlist(zc, config)) != 0)
4353                                         error = err;
4354                                 nvlist_free(config);
4355                         }
4356                         nvlist_free(policy);
4357                 }
4358         }
4359 
4360         if (error != 0)
4361                 return (error);
4362 
4363         spa_vdev_state_enter(spa, SCL_NONE);
4364 
4365         if (zc->zc_guid == 0) {
4366                 vd = NULL;
4367         } else {
4368                 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4369                 if (vd == NULL) {
4370                         (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4371                         spa_close(spa, FTAG);
4372                         return (SET_ERROR(ENODEV));
4373                 }
4374         }
4375 
4376         vdev_clear(spa, vd);
4377 
4378         (void) spa_vdev_state_exit(spa, NULL, 0);
4379 
4380         /*
4381          * Resume any suspended I/Os.
4382          */
4383         if (zio_resume(spa) != 0)
4384                 error = SET_ERROR(EIO);
4385 
4386         spa_close(spa, FTAG);
4387 
4388         return (error);
4389 }
4390 
4391 static int
4392 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4393 {
4394         spa_t *spa;
4395         int error;
4396 
4397         error = spa_open(zc->zc_name, &spa, FTAG);
4398         if (error != 0)
4399                 return (error);
4400 
4401         spa_vdev_state_enter(spa, SCL_NONE);
4402 
4403         /*
4404          * If a resilver is already in progress then set the
4405          * spa_scrub_reopen flag to B_TRUE so that we don't restart
4406          * the scan as a side effect of the reopen. Otherwise, let
4407          * vdev_open() decided if a resilver is required.
4408          */
4409         spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4410         vdev_reopen(spa->spa_root_vdev);
4411         spa->spa_scrub_reopen = B_FALSE;
4412 
4413         (void) spa_vdev_state_exit(spa, NULL, 0);
4414         spa_close(spa, FTAG);
4415         return (0);
4416 }
4417 /*
4418  * inputs:
4419  * zc_name      name of filesystem
4420  * zc_value     name of origin snapshot
4421  *
4422  * outputs:
4423  * zc_string    name of conflicting snapshot, if there is one
4424  */
4425 static int
4426 zfs_ioc_promote(zfs_cmd_t *zc)
4427 {
4428         char *cp;
4429 
4430         /*
4431          * We don't need to unmount *all* the origin fs's snapshots, but
4432          * it's easier.
4433          */
4434         cp = strchr(zc->zc_value, '@');
4435         if (cp)
4436                 *cp = '\0';
4437         (void) dmu_objset_find(zc->zc_value,
4438             zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
4439         return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4440 }
4441 
4442 /*
4443  * Retrieve a single {user|group}{used|quota}@... property.
4444  *
4445  * inputs:
4446  * zc_name      name of filesystem
4447  * zc_objset_type zfs_userquota_prop_t
4448  * zc_value     domain name (eg. "S-1-234-567-89")
4449  * zc_guid      RID/UID/GID
4450  *
4451  * outputs:
4452  * zc_cookie    property value
4453  */
4454 static int
4455 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4456 {
4457         zfsvfs_t *zfsvfs;
4458         int error;
4459 
4460         if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4461                 return (SET_ERROR(EINVAL));
4462 
4463         error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4464         if (error != 0)
4465                 return (error);
4466 
4467         error = zfs_userspace_one(zfsvfs,
4468             zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4469         zfsvfs_rele(zfsvfs, FTAG);
4470 
4471         return (error);
4472 }
4473 
4474 /*
4475  * inputs:
4476  * zc_name              name of filesystem
4477  * zc_cookie            zap cursor
4478  * zc_objset_type       zfs_userquota_prop_t
4479  * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4480  *
4481  * outputs:
4482  * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4483  * zc_cookie    zap cursor
4484  */
4485 static int
4486 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4487 {
4488         zfsvfs_t *zfsvfs;
4489         int bufsize = zc->zc_nvlist_dst_size;
4490 
4491         if (bufsize <= 0)
4492                 return (SET_ERROR(ENOMEM));
4493 
4494         int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4495         if (error != 0)
4496                 return (error);
4497 
4498         void *buf = kmem_alloc(bufsize, KM_SLEEP);
4499 
4500         error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
4501             buf, &zc->zc_nvlist_dst_size);
4502 
4503         if (error == 0) {
4504                 error = xcopyout(buf,
4505                     (void *)(uintptr_t)zc->zc_nvlist_dst,
4506                     zc->zc_nvlist_dst_size);
4507         }
4508         kmem_free(buf, bufsize);
4509         zfsvfs_rele(zfsvfs, FTAG);
4510 
4511         return (error);
4512 }
4513 
4514 /*
4515  * inputs:
4516  * zc_name              name of filesystem
4517  *
4518  * outputs:
4519  * none
4520  */
4521 static int
4522 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4523 {
4524         objset_t *os;
4525         int error = 0;
4526         zfsvfs_t *zfsvfs;
4527 
4528         if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
4529                 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
4530                         /*
4531                          * If userused is not enabled, it may be because the
4532                          * objset needs to be closed & reopened (to grow the
4533                          * objset_phys_t).  Suspend/resume the fs will do that.
4534                          */
4535                         error = zfs_suspend_fs(zfsvfs);
4536                         if (error == 0)
4537                                 error = zfs_resume_fs(zfsvfs, zc->zc_name);
4538                 }
4539                 if (error == 0)
4540                         error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
4541                 VFS_RELE(zfsvfs->z_vfs);
4542         } else {
4543                 /* XXX kind of reading contents without owning */
4544                 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4545                 if (error != 0)
4546                         return (error);
4547 
4548                 error = dmu_objset_userspace_upgrade(os);
4549                 dmu_objset_rele(os, FTAG);
4550         }
4551 
4552         return (error);
4553 }
4554 
4555 /*
4556  * We don't want to have a hard dependency
4557  * against some special symbols in sharefs
4558  * nfs, and smbsrv.  Determine them if needed when
4559  * the first file system is shared.
4560  * Neither sharefs, nfs or smbsrv are unloadable modules.
4561  */
4562 int (*znfsexport_fs)(void *arg);
4563 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
4564 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
4565 
4566 int zfs_nfsshare_inited;
4567 int zfs_smbshare_inited;
4568 
4569 ddi_modhandle_t nfs_mod;
4570 ddi_modhandle_t sharefs_mod;
4571 ddi_modhandle_t smbsrv_mod;
4572 kmutex_t zfs_share_lock;
4573 
4574 static int
4575 zfs_init_sharefs()
4576 {
4577         int error;
4578 
4579         ASSERT(MUTEX_HELD(&zfs_share_lock));
4580         /* Both NFS and SMB shares also require sharetab support. */
4581         if (sharefs_mod == NULL && ((sharefs_mod =
4582             ddi_modopen("fs/sharefs",
4583             KRTLD_MODE_FIRST, &error)) == NULL)) {
4584                 return (SET_ERROR(ENOSYS));
4585         }
4586         if (zshare_fs == NULL && ((zshare_fs =
4587             (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
4588             ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
4589                 return (SET_ERROR(ENOSYS));
4590         }
4591         return (0);
4592 }
4593 
4594 static int
4595 zfs_ioc_share(zfs_cmd_t *zc)
4596 {
4597         int error;
4598         int opcode;
4599 
4600         switch (zc->zc_share.z_sharetype) {
4601         case ZFS_SHARE_NFS:
4602         case ZFS_UNSHARE_NFS:
4603                 if (zfs_nfsshare_inited == 0) {
4604                         mutex_enter(&zfs_share_lock);
4605                         if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
4606                             KRTLD_MODE_FIRST, &error)) == NULL)) {
4607                                 mutex_exit(&zfs_share_lock);
4608                                 return (SET_ERROR(ENOSYS));
4609                         }
4610                         if (znfsexport_fs == NULL &&
4611                             ((znfsexport_fs = (int (*)(void *))
4612                             ddi_modsym(nfs_mod,
4613                             "nfs_export", &error)) == NULL)) {
4614                                 mutex_exit(&zfs_share_lock);
4615                                 return (SET_ERROR(ENOSYS));
4616                         }
4617                         error = zfs_init_sharefs();
4618                         if (error != 0) {
4619                                 mutex_exit(&zfs_share_lock);
4620                                 return (SET_ERROR(ENOSYS));
4621                         }
4622                         zfs_nfsshare_inited = 1;
4623                         mutex_exit(&zfs_share_lock);
4624                 }
4625                 break;
4626         case ZFS_SHARE_SMB:
4627         case ZFS_UNSHARE_SMB:
4628                 if (zfs_smbshare_inited == 0) {
4629                         mutex_enter(&zfs_share_lock);
4630                         if (smbsrv_mod == NULL && ((smbsrv_mod =
4631                             ddi_modopen("drv/smbsrv",
4632                             KRTLD_MODE_FIRST, &error)) == NULL)) {
4633                                 mutex_exit(&zfs_share_lock);
4634                                 return (SET_ERROR(ENOSYS));
4635                         }
4636                         if (zsmbexport_fs == NULL && ((zsmbexport_fs =
4637                             (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
4638                             "smb_server_share", &error)) == NULL)) {
4639                                 mutex_exit(&zfs_share_lock);
4640                                 return (SET_ERROR(ENOSYS));
4641                         }
4642                         error = zfs_init_sharefs();
4643                         if (error != 0) {
4644                                 mutex_exit(&zfs_share_lock);
4645                                 return (SET_ERROR(ENOSYS));
4646                         }
4647                         zfs_smbshare_inited = 1;
4648                         mutex_exit(&zfs_share_lock);
4649                 }
4650                 break;
4651         default:
4652                 return (SET_ERROR(EINVAL));
4653         }
4654 
4655         switch (zc->zc_share.z_sharetype) {
4656         case ZFS_SHARE_NFS:
4657         case ZFS_UNSHARE_NFS:
4658                 if (error =
4659                     znfsexport_fs((void *)
4660                     (uintptr_t)zc->zc_share.z_exportdata))
4661                         return (error);
4662                 break;
4663         case ZFS_SHARE_SMB:
4664         case ZFS_UNSHARE_SMB:
4665                 if (error = zsmbexport_fs((void *)
4666                     (uintptr_t)zc->zc_share.z_exportdata,
4667                     zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
4668                     B_TRUE: B_FALSE)) {
4669                         return (error);
4670                 }
4671                 break;
4672         }
4673 
4674         opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
4675             zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
4676             SHAREFS_ADD : SHAREFS_REMOVE;
4677 
4678         /*
4679          * Add or remove share from sharetab
4680          */
4681         error = zshare_fs(opcode,
4682             (void *)(uintptr_t)zc->zc_share.z_sharedata,
4683             zc->zc_share.z_sharemax);
4684 
4685         return (error);
4686 
4687 }
4688 
4689 ace_t full_access[] = {
4690         {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4691 };
4692 
4693 /*
4694  * inputs:
4695  * zc_name              name of containing filesystem
4696  * zc_obj               object # beyond which we want next in-use object #
4697  *
4698  * outputs:
4699  * zc_obj               next in-use object #
4700  */
4701 static int
4702 zfs_ioc_next_obj(zfs_cmd_t *zc)
4703 {
4704         objset_t *os = NULL;
4705         int error;
4706 
4707         error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4708         if (error != 0)
4709                 return (error);
4710 
4711         error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4712             os->os_dsl_dataset->ds_phys->ds_prev_snap_txg);
4713 
4714         dmu_objset_rele(os, FTAG);
4715         return (error);
4716 }
4717 
4718 /*
4719  * inputs:
4720  * zc_name              name of filesystem
4721  * zc_value             prefix name for snapshot
4722  * zc_cleanup_fd        cleanup-on-exit file descriptor for calling process
4723  *
4724  * outputs:
4725  * zc_value             short name of new snapshot
4726  */
4727 static int
4728 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
4729 {
4730         char *snap_name;
4731         char *hold_name;
4732         int error;
4733         minor_t minor;
4734 
4735         error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
4736         if (error != 0)
4737                 return (error);
4738 
4739         snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
4740             (u_longlong_t)ddi_get_lbolt64());
4741         hold_name = kmem_asprintf("%%%s", zc->zc_value);
4742 
4743         error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
4744             hold_name);
4745         if (error == 0)
4746                 (void) strcpy(zc->zc_value, snap_name);
4747         strfree(snap_name);
4748         strfree(hold_name);
4749         zfs_onexit_fd_rele(zc->zc_cleanup_fd);
4750         return (error);
4751 }
4752 
4753 /*
4754  * inputs:
4755  * zc_name              name of "to" snapshot
4756  * zc_value             name of "from" snapshot
4757  * zc_cookie            file descriptor to write diff data on
4758  *
4759  * outputs:
4760  * dmu_diff_record_t's to the file descriptor
4761  */
4762 static int
4763 zfs_ioc_diff(zfs_cmd_t *zc)
4764 {
4765         file_t *fp;
4766         offset_t off;
4767         int error;
4768 
4769         fp = getf(zc->zc_cookie);
4770         if (fp == NULL)
4771                 return (SET_ERROR(EBADF));
4772 
4773         off = fp->f_offset;
4774 
4775         error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
4776 
4777         if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4778                 fp->f_offset = off;
4779         releasef(zc->zc_cookie);
4780 
4781         return (error);
4782 }
4783 
4784 /*
4785  * Remove all ACL files in shares dir
4786  */
4787 static int
4788 zfs_smb_acl_purge(znode_t *dzp)
4789 {
4790         zap_cursor_t    zc;
4791         zap_attribute_t zap;
4792         zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
4793         int error;
4794 
4795         for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
4796             (error = zap_cursor_retrieve(&zc, &zap)) == 0;
4797             zap_cursor_advance(&zc)) {
4798                 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
4799                     NULL, 0)) != 0)
4800                         break;
4801         }
4802         zap_cursor_fini(&zc);
4803         return (error);
4804 }
4805 
4806 static int
4807 zfs_ioc_smb_acl(zfs_cmd_t *zc)
4808 {
4809         vnode_t *vp;
4810         znode_t *dzp;
4811         vnode_t *resourcevp = NULL;
4812         znode_t *sharedir;
4813         zfsvfs_t *zfsvfs;
4814         nvlist_t *nvlist;
4815         char *src, *target;
4816         vattr_t vattr;
4817         vsecattr_t vsec;
4818         int error = 0;
4819 
4820         if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
4821             NO_FOLLOW, NULL, &vp)) != 0)
4822                 return (error);
4823 
4824         /* Now make sure mntpnt and dataset are ZFS */
4825 
4826         if (vp->v_vfsp->vfs_fstype != zfsfstype ||
4827             (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
4828             zc->zc_name) != 0)) {
4829                 VN_RELE(vp);
4830                 return (SET_ERROR(EINVAL));
4831         }
4832 
4833         dzp = VTOZ(vp);
4834         zfsvfs = dzp->z_zfsvfs;
4835         ZFS_ENTER(zfsvfs);
4836 
4837         /*
4838          * Create share dir if its missing.
4839          */
4840         mutex_enter(&zfsvfs->z_lock);
4841         if (zfsvfs->z_shares_dir == 0) {
4842                 dmu_tx_t *tx;
4843 
4844                 tx = dmu_tx_create(zfsvfs->z_os);
4845                 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
4846                     ZFS_SHARES_DIR);
4847                 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
4848                 error = dmu_tx_assign(tx, TXG_WAIT);
4849                 if (error != 0) {
4850                         dmu_tx_abort(tx);
4851                 } else {
4852                         error = zfs_create_share_dir(zfsvfs, tx);
4853                         dmu_tx_commit(tx);
4854                 }
4855                 if (error != 0) {
4856                         mutex_exit(&zfsvfs->z_lock);
4857                         VN_RELE(vp);
4858                         ZFS_EXIT(zfsvfs);
4859                         return (error);
4860                 }
4861         }
4862         mutex_exit(&zfsvfs->z_lock);
4863 
4864         ASSERT(zfsvfs->z_shares_dir);
4865         if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
4866                 VN_RELE(vp);
4867                 ZFS_EXIT(zfsvfs);
4868                 return (error);
4869         }
4870 
4871         switch (zc->zc_cookie) {
4872         case ZFS_SMB_ACL_ADD:
4873                 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
4874                 vattr.va_type = VREG;
4875                 vattr.va_mode = S_IFREG|0777;
4876                 vattr.va_uid = 0;
4877                 vattr.va_gid = 0;
4878 
4879                 vsec.vsa_mask = VSA_ACE;
4880                 vsec.vsa_aclentp = &full_access;
4881                 vsec.vsa_aclentsz = sizeof (full_access);
4882                 vsec.vsa_aclcnt = 1;
4883 
4884                 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
4885                     &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
4886                 if (resourcevp)
4887                         VN_RELE(resourcevp);
4888                 break;
4889 
4890         case ZFS_SMB_ACL_REMOVE:
4891                 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
4892                     NULL, 0);
4893                 break;
4894 
4895         case ZFS_SMB_ACL_RENAME:
4896                 if ((error = get_nvlist(zc->zc_nvlist_src,
4897                     zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
4898                         VN_RELE(vp);
4899                         ZFS_EXIT(zfsvfs);
4900                         return (error);
4901                 }
4902                 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
4903                     nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
4904                     &target)) {
4905                         VN_RELE(vp);
4906                         VN_RELE(ZTOV(sharedir));
4907                         ZFS_EXIT(zfsvfs);
4908                         nvlist_free(nvlist);
4909                         return (error);
4910                 }
4911                 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
4912                     kcred, NULL, 0);
4913                 nvlist_free(nvlist);
4914                 break;
4915 
4916         case ZFS_SMB_ACL_PURGE:
4917                 error = zfs_smb_acl_purge(sharedir);
4918                 break;
4919 
4920         default:
4921                 error = SET_ERROR(EINVAL);
4922                 break;
4923         }
4924 
4925         VN_RELE(vp);
4926         VN_RELE(ZTOV(sharedir));
4927 
4928         ZFS_EXIT(zfsvfs);
4929 
4930         return (error);
4931 }
4932 
4933 /*
4934  * innvl: {
4935  *     "holds" -> { snapname -> holdname (string), ... }
4936  *     (optional) "cleanup_fd" -> fd (int32)
4937  * }
4938  *
4939  * outnvl: {
4940  *     snapname -> error value (int32)
4941  *     ...
4942  * }
4943  */
4944 /* ARGSUSED */
4945 static int
4946 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
4947 {
4948         nvlist_t *holds;
4949         int cleanup_fd = -1;
4950         int error;
4951         minor_t minor = 0;
4952 
4953         error = nvlist_lookup_nvlist(args, "holds", &holds);
4954         if (error != 0)
4955                 return (SET_ERROR(EINVAL));
4956 
4957         if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
4958                 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
4959                 if (error != 0)
4960                         return (error);
4961         }
4962 
4963         error = dsl_dataset_user_hold(holds, minor, errlist);
4964         if (minor != 0)
4965                 zfs_onexit_fd_rele(cleanup_fd);
4966         return (error);
4967 }
4968 
4969 /*
4970  * innvl is not used.
4971  *
4972  * outnvl: {
4973  *    holdname -> time added (uint64 seconds since epoch)
4974  *    ...
4975  * }
4976  */
4977 /* ARGSUSED */
4978 static int
4979 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
4980 {
4981         return (dsl_dataset_get_holds(snapname, outnvl));
4982 }
4983 
4984 /*
4985  * innvl: {
4986  *     snapname -> { holdname, ... }
4987  *     ...
4988  * }
4989  *
4990  * outnvl: {
4991  *     snapname -> error value (int32)
4992  *     ...
4993  * }
4994  */
4995 /* ARGSUSED */
4996 static int
4997 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
4998 {
4999         nvpair_t *pair;
5000         int err;
5001 
5002         /*
5003          * The release may cause the snapshot to be destroyed; make sure it
5004          * is not mounted.
5005          */
5006         for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
5007             pair = nvlist_next_nvpair(holds, pair)) {
5008                 err = zfs_unmount_snap(nvpair_name(pair));
5009                 if (err)
5010                         return (err);
5011         }
5012 
5013         return (dsl_dataset_user_release(holds, errlist));
5014 }
5015 
5016 /*
5017  * inputs:
5018  * zc_name              name of new filesystem or snapshot
5019  * zc_value             full name of old snapshot
5020  *
5021  * outputs:
5022  * zc_cookie            space in bytes
5023  * zc_objset_type       compressed space in bytes
5024  * zc_perm_action       uncompressed space in bytes
5025  */
5026 static int
5027 zfs_ioc_space_written(zfs_cmd_t *zc)
5028 {
5029         int error;
5030         dsl_pool_t *dp;
5031         dsl_dataset_t *new, *old;
5032 
5033         error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5034         if (error != 0)
5035                 return (error);
5036         error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5037         if (error != 0) {
5038                 dsl_pool_rele(dp, FTAG);
5039                 return (error);
5040         }
5041         error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5042         if (error != 0) {
5043                 dsl_dataset_rele(new, FTAG);
5044                 dsl_pool_rele(dp, FTAG);
5045                 return (error);
5046         }
5047 
5048         error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5049             &zc->zc_objset_type, &zc->zc_perm_action);
5050         dsl_dataset_rele(old, FTAG);
5051         dsl_dataset_rele(new, FTAG);
5052         dsl_pool_rele(dp, FTAG);
5053         return (error);
5054 }
5055 
5056 /*
5057  * innvl: {
5058  *     "firstsnap" -> snapshot name
5059  * }
5060  *
5061  * outnvl: {
5062  *     "used" -> space in bytes
5063  *     "compressed" -> compressed space in bytes
5064  *     "uncompressed" -> uncompressed space in bytes
5065  * }
5066  */
5067 static int
5068 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5069 {
5070         int error;
5071         dsl_pool_t *dp;
5072         dsl_dataset_t *new, *old;
5073         char *firstsnap;
5074         uint64_t used, comp, uncomp;
5075 
5076         if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5077                 return (SET_ERROR(EINVAL));
5078 
5079         error = dsl_pool_hold(lastsnap, FTAG, &dp);
5080         if (error != 0)
5081                 return (error);
5082 
5083         error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5084         if (error != 0) {
5085                 dsl_pool_rele(dp, FTAG);
5086                 return (error);
5087         }
5088         error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5089         if (error != 0) {
5090                 dsl_dataset_rele(new, FTAG);
5091                 dsl_pool_rele(dp, FTAG);
5092                 return (error);
5093         }
5094 
5095         error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5096         dsl_dataset_rele(old, FTAG);
5097         dsl_dataset_rele(new, FTAG);
5098         dsl_pool_rele(dp, FTAG);
5099         fnvlist_add_uint64(outnvl, "used", used);
5100         fnvlist_add_uint64(outnvl, "compressed", comp);
5101         fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5102         return (error);
5103 }
5104 
5105 /*
5106  * innvl: {
5107  *     "fd" -> file descriptor to write stream to (int32)
5108  *     (optional) "fromsnap" -> full snap name to send an incremental from
5109  * }
5110  *
5111  * outnvl is unused
5112  */
5113 /* ARGSUSED */
5114 static int
5115 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5116 {
5117         int error;
5118         offset_t off;
5119         char *fromname = NULL;
5120         int fd;
5121 
5122         error = nvlist_lookup_int32(innvl, "fd", &fd);
5123         if (error != 0)
5124                 return (SET_ERROR(EINVAL));
5125 
5126         (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5127 
5128         file_t *fp = getf(fd);
5129         if (fp == NULL)
5130                 return (SET_ERROR(EBADF));
5131 
5132         off = fp->f_offset;
5133         error = dmu_send(snapname, fromname, fd, fp->f_vnode, &off);
5134 
5135         if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5136                 fp->f_offset = off;
5137         releasef(fd);
5138         return (error);
5139 }
5140 
5141 /*
5142  * Determine approximately how large a zfs send stream will be -- the number
5143  * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5144  *
5145  * innvl: {
5146  *     (optional) "fromsnap" -> full snap name to send an incremental from
5147  * }
5148  *
5149  * outnvl: {
5150  *     "space" -> bytes of space (uint64)
5151  * }
5152  */
5153 static int
5154 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5155 {
5156         dsl_pool_t *dp;
5157         dsl_dataset_t *fromsnap = NULL;
5158         dsl_dataset_t *tosnap;
5159         int error;
5160         char *fromname;
5161         uint64_t space;
5162 
5163         error = dsl_pool_hold(snapname, FTAG, &dp);
5164         if (error != 0)
5165                 return (error);
5166 
5167         error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5168         if (error != 0) {
5169                 dsl_pool_rele(dp, FTAG);
5170                 return (error);
5171         }
5172 
5173         error = nvlist_lookup_string(innvl, "fromsnap", &fromname);
5174         if (error == 0) {
5175                 error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5176                 if (error != 0) {
5177                         dsl_dataset_rele(tosnap, FTAG);
5178                         dsl_pool_rele(dp, FTAG);
5179                         return (error);
5180                 }
5181         }
5182 
5183         error = dmu_send_estimate(tosnap, fromsnap, &space);
5184         fnvlist_add_uint64(outnvl, "space", space);
5185 
5186         if (fromsnap != NULL)
5187                 dsl_dataset_rele(fromsnap, FTAG);
5188         dsl_dataset_rele(tosnap, FTAG);
5189         dsl_pool_rele(dp, FTAG);
5190         return (error);
5191 }
5192 
5193 
5194 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5195 
5196 static void
5197 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5198     zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5199     boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5200 {
5201         zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5202 
5203         ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5204         ASSERT3U(ioc, <, ZFS_IOC_LAST);
5205         ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5206         ASSERT3P(vec->zvec_func, ==, NULL);
5207 
5208         vec->zvec_legacy_func = func;
5209         vec->zvec_secpolicy = secpolicy;
5210         vec->zvec_namecheck = namecheck;
5211         vec->zvec_allow_log = log_history;
5212         vec->zvec_pool_check = pool_check;
5213 }
5214 
5215 /*
5216  * See the block comment at the beginning of this file for details on
5217  * each argument to this function.
5218  */
5219 static void
5220 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5221     zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5222     zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5223     boolean_t allow_log)
5224 {
5225         zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5226 
5227         ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5228         ASSERT3U(ioc, <, ZFS_IOC_LAST);
5229         ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5230         ASSERT3P(vec->zvec_func, ==, NULL);
5231 
5232         /* if we are logging, the name must be valid */
5233         ASSERT(!allow_log || namecheck != NO_NAME);
5234 
5235         vec->zvec_name = name;
5236         vec->zvec_func = func;
5237         vec->zvec_secpolicy = secpolicy;
5238         vec->zvec_namecheck = namecheck;
5239         vec->zvec_pool_check = pool_check;
5240         vec->zvec_smush_outnvlist = smush_outnvlist;
5241         vec->zvec_allow_log = allow_log;
5242 }
5243 
5244 static void
5245 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5246     zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5247     zfs_ioc_poolcheck_t pool_check)
5248 {
5249         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5250             POOL_NAME, log_history, pool_check);
5251 }
5252 
5253 static void
5254 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5255     zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5256 {
5257         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5258             DATASET_NAME, B_FALSE, pool_check);
5259 }
5260 
5261 static void
5262 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5263 {
5264         zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5265             POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5266 }
5267 
5268 static void
5269 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5270     zfs_secpolicy_func_t *secpolicy)
5271 {
5272         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5273             NO_NAME, B_FALSE, POOL_CHECK_NONE);
5274 }
5275 
5276 static void
5277 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5278     zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5279 {
5280         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5281             DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5282 }
5283 
5284 static void
5285 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5286 {
5287         zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5288             zfs_secpolicy_read);
5289 }
5290 
5291 static void
5292 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5293         zfs_secpolicy_func_t *secpolicy)
5294 {
5295         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5296             DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5297 }
5298 
5299 static void
5300 zfs_ioctl_init(void)
5301 {
5302         zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5303             zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5304             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5305 
5306         zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5307             zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5308             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5309 
5310         zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5311             zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5312             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5313 
5314         zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5315             zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5316             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5317 
5318         zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5319             zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5320             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5321 
5322         zfs_ioctl_register("create", ZFS_IOC_CREATE,
5323             zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5324             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5325 
5326         zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5327             zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5328             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5329 
5330         zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5331             zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5332             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5333 
5334         zfs_ioctl_register("hold", ZFS_IOC_HOLD,
5335             zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
5336             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5337         zfs_ioctl_register("release", ZFS_IOC_RELEASE,
5338             zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
5339             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5340 
5341         zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
5342             zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
5343             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5344 
5345         /* IOCTLS that use the legacy function signature */
5346 
5347         zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5348             zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5349 
5350         zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5351             zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5352         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5353             zfs_ioc_pool_scan);
5354         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5355             zfs_ioc_pool_upgrade);
5356         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5357             zfs_ioc_vdev_add);
5358         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5359             zfs_ioc_vdev_remove);
5360         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5361             zfs_ioc_vdev_set_state);
5362         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5363             zfs_ioc_vdev_attach);
5364         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5365             zfs_ioc_vdev_detach);
5366         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5367             zfs_ioc_vdev_setpath);
5368         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5369             zfs_ioc_vdev_setfru);
5370         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5371             zfs_ioc_pool_set_props);
5372         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5373             zfs_ioc_vdev_split);
5374         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5375             zfs_ioc_pool_reguid);
5376 
5377         zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5378             zfs_ioc_pool_configs, zfs_secpolicy_none);
5379         zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5380             zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5381         zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5382             zfs_ioc_inject_fault, zfs_secpolicy_inject);
5383         zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5384             zfs_ioc_clear_fault, zfs_secpolicy_inject);
5385         zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
5386             zfs_ioc_inject_list_next, zfs_secpolicy_inject);
5387 
5388         /*
5389          * pool destroy, and export don't log the history as part of
5390          * zfsdev_ioctl, but rather zfs_ioc_pool_export
5391          * does the logging of those commands.
5392          */
5393         zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
5394             zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5395         zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
5396             zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5397 
5398         zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
5399             zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5400         zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
5401             zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5402 
5403         zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
5404             zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
5405         zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
5406             zfs_ioc_dsobj_to_dsname,
5407             zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
5408         zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
5409             zfs_ioc_pool_get_history,
5410             zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
5411 
5412         zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
5413             zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5414 
5415         zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
5416             zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5417         zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
5418             zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5419 
5420         zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
5421             zfs_ioc_space_written);
5422         zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
5423             zfs_ioc_objset_recvd_props);
5424         zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
5425             zfs_ioc_next_obj);
5426         zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
5427             zfs_ioc_get_fsacl);
5428         zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
5429             zfs_ioc_objset_stats);
5430         zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
5431             zfs_ioc_objset_zplprops);
5432         zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
5433             zfs_ioc_dataset_list_next);
5434         zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
5435             zfs_ioc_snapshot_list_next);
5436         zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
5437             zfs_ioc_send_progress);
5438 
5439         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
5440             zfs_ioc_diff, zfs_secpolicy_diff);
5441         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
5442             zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
5443         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
5444             zfs_ioc_obj_to_path, zfs_secpolicy_diff);
5445         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
5446             zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
5447         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
5448             zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
5449         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
5450             zfs_ioc_send, zfs_secpolicy_send);
5451 
5452         zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
5453             zfs_secpolicy_none);
5454         zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
5455             zfs_secpolicy_destroy);
5456         zfs_ioctl_register_dataset_modify(ZFS_IOC_ROLLBACK, zfs_ioc_rollback,
5457             zfs_secpolicy_rollback);
5458         zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
5459             zfs_secpolicy_rename);
5460         zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
5461             zfs_secpolicy_recv);
5462         zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
5463             zfs_secpolicy_promote);
5464         zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
5465             zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
5466         zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
5467             zfs_secpolicy_set_fsacl);
5468 
5469         zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
5470             zfs_secpolicy_share, POOL_CHECK_NONE);
5471         zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
5472             zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
5473         zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
5474             zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
5475             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5476         zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
5477             zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
5478             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5479 }
5480 
5481 int
5482 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
5483     zfs_ioc_poolcheck_t check)
5484 {
5485         spa_t *spa;
5486         int error;
5487 
5488         ASSERT(type == POOL_NAME || type == DATASET_NAME);
5489 
5490         if (check & POOL_CHECK_NONE)
5491                 return (0);
5492 
5493         error = spa_open(name, &spa, FTAG);
5494         if (error == 0) {
5495                 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
5496                         error = SET_ERROR(EAGAIN);
5497                 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
5498                         error = SET_ERROR(EROFS);
5499                 spa_close(spa, FTAG);
5500         }
5501         return (error);
5502 }
5503 
5504 /*
5505  * Find a free minor number.
5506  */
5507 minor_t
5508 zfsdev_minor_alloc(void)
5509 {
5510         static minor_t last_minor;
5511         minor_t m;
5512 
5513         ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5514 
5515         for (m = last_minor + 1; m != last_minor; m++) {
5516                 if (m > ZFSDEV_MAX_MINOR)
5517                         m = 1;
5518                 if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
5519                         last_minor = m;
5520                         return (m);
5521                 }
5522         }
5523 
5524         return (0);
5525 }
5526 
5527 static int
5528 zfs_ctldev_init(dev_t *devp)
5529 {
5530         minor_t minor;
5531         zfs_soft_state_t *zs;
5532 
5533         ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5534         ASSERT(getminor(*devp) == 0);
5535 
5536         minor = zfsdev_minor_alloc();
5537         if (minor == 0)
5538                 return (SET_ERROR(ENXIO));
5539 
5540         if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
5541                 return (SET_ERROR(EAGAIN));
5542 
5543         *devp = makedevice(getemajor(*devp), minor);
5544 
5545         zs = ddi_get_soft_state(zfsdev_state, minor);
5546         zs->zss_type = ZSST_CTLDEV;
5547         zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
5548 
5549         return (0);
5550 }
5551 
5552 static void
5553 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
5554 {
5555         ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5556 
5557         zfs_onexit_destroy(zo);
5558         ddi_soft_state_free(zfsdev_state, minor);
5559 }
5560 
5561 void *
5562 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
5563 {
5564         zfs_soft_state_t *zp;
5565 
5566         zp = ddi_get_soft_state(zfsdev_state, minor);
5567         if (zp == NULL || zp->zss_type != which)
5568                 return (NULL);
5569 
5570         return (zp->zss_data);
5571 }
5572 
5573 static int
5574 zfsdev_open(dev_t *devp, int flag, int otyp, cred_t *cr)
5575 {
5576         int error = 0;
5577 
5578         if (getminor(*devp) != 0)
5579                 return (zvol_open(devp, flag, otyp, cr));
5580 
5581         /* This is the control device. Allocate a new minor if requested. */
5582         if (flag & FEXCL) {
5583                 mutex_enter(&zfsdev_state_lock);
5584                 error = zfs_ctldev_init(devp);
5585                 mutex_exit(&zfsdev_state_lock);
5586         }
5587 
5588         return (error);
5589 }
5590 
5591 static int
5592 zfsdev_close(dev_t dev, int flag, int otyp, cred_t *cr)
5593 {
5594         zfs_onexit_t *zo;
5595         minor_t minor = getminor(dev);
5596 
5597         if (minor == 0)
5598                 return (0);
5599 
5600         mutex_enter(&zfsdev_state_lock);
5601         zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
5602         if (zo == NULL) {
5603                 mutex_exit(&zfsdev_state_lock);
5604                 return (zvol_close(dev, flag, otyp, cr));
5605         }
5606         zfs_ctldev_destroy(zo, minor);
5607         mutex_exit(&zfsdev_state_lock);
5608 
5609         return (0);
5610 }
5611 
5612 static int
5613 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
5614 {
5615         zfs_cmd_t *zc;
5616         uint_t vecnum;
5617         int error, rc, len;
5618         minor_t minor = getminor(dev);
5619         const zfs_ioc_vec_t *vec;
5620         char *saved_poolname = NULL;
5621         nvlist_t *innvl = NULL;
5622 
5623         if (minor != 0 &&
5624             zfsdev_get_soft_state(minor, ZSST_CTLDEV) == NULL)
5625                 return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
5626 
5627         vecnum = cmd - ZFS_IOC_FIRST;
5628         ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
5629 
5630         if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
5631                 return (SET_ERROR(EINVAL));
5632         vec = &zfs_ioc_vec[vecnum];
5633 
5634         zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
5635 
5636         error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
5637         if (error != 0) {
5638                 error = SET_ERROR(EFAULT);
5639                 goto out;
5640         }
5641 
5642         zc->zc_iflags = flag & FKIOCTL;
5643         if (zc->zc_nvlist_src_size != 0) {
5644                 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
5645                     zc->zc_iflags, &innvl);
5646                 if (error != 0)
5647                         goto out;
5648         }
5649 
5650         /*
5651          * Ensure that all pool/dataset names are valid before we pass down to
5652          * the lower layers.
5653          */
5654         zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
5655         switch (vec->zvec_namecheck) {
5656         case POOL_NAME:
5657                 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
5658                         error = SET_ERROR(EINVAL);
5659                 else
5660                         error = pool_status_check(zc->zc_name,
5661                             vec->zvec_namecheck, vec->zvec_pool_check);
5662                 break;
5663 
5664         case DATASET_NAME:
5665                 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
5666                         error = SET_ERROR(EINVAL);
5667                 else
5668                         error = pool_status_check(zc->zc_name,
5669                             vec->zvec_namecheck, vec->zvec_pool_check);
5670                 break;
5671 
5672         case NO_NAME:
5673                 break;
5674         }
5675 
5676 
5677         if (error == 0 && !(flag & FKIOCTL))
5678                 error = vec->zvec_secpolicy(zc, innvl, cr);
5679 
5680         if (error != 0)
5681                 goto out;
5682 
5683         /* legacy ioctls can modify zc_name */
5684         len = strcspn(zc->zc_name, "/@") + 1;
5685         saved_poolname = kmem_alloc(len, KM_SLEEP);
5686         (void) strlcpy(saved_poolname, zc->zc_name, len);
5687 
5688         if (vec->zvec_func != NULL) {
5689                 nvlist_t *outnvl;
5690                 int puterror = 0;
5691                 spa_t *spa;
5692                 nvlist_t *lognv = NULL;
5693 
5694                 ASSERT(vec->zvec_legacy_func == NULL);
5695 
5696                 /*
5697                  * Add the innvl to the lognv before calling the func,
5698                  * in case the func changes the innvl.
5699                  */
5700                 if (vec->zvec_allow_log) {
5701                         lognv = fnvlist_alloc();
5702                         fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
5703                             vec->zvec_name);
5704                         if (!nvlist_empty(innvl)) {
5705                                 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
5706                                     innvl);
5707                         }
5708                 }
5709 
5710                 outnvl = fnvlist_alloc();
5711                 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
5712 
5713                 if (error == 0 && vec->zvec_allow_log &&
5714                     spa_open(zc->zc_name, &spa, FTAG) == 0) {
5715                         if (!nvlist_empty(outnvl)) {
5716                                 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
5717                                     outnvl);
5718                         }
5719                         (void) spa_history_log_nvl(spa, lognv);
5720                         spa_close(spa, FTAG);
5721                 }
5722                 fnvlist_free(lognv);
5723 
5724                 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
5725                         int smusherror = 0;
5726                         if (vec->zvec_smush_outnvlist) {
5727                                 smusherror = nvlist_smush(outnvl,
5728                                     zc->zc_nvlist_dst_size);
5729                         }
5730                         if (smusherror == 0)
5731                                 puterror = put_nvlist(zc, outnvl);
5732                 }
5733 
5734                 if (puterror != 0)
5735                         error = puterror;
5736 
5737                 nvlist_free(outnvl);
5738         } else {
5739                 error = vec->zvec_legacy_func(zc);
5740         }
5741 
5742 out:
5743         nvlist_free(innvl);
5744         rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
5745         if (error == 0 && rc != 0)
5746                 error = SET_ERROR(EFAULT);
5747         if (error == 0 && vec->zvec_allow_log) {
5748                 char *s = tsd_get(zfs_allow_log_key);
5749                 if (s != NULL)
5750                         strfree(s);
5751                 (void) tsd_set(zfs_allow_log_key, saved_poolname);
5752         } else {
5753                 if (saved_poolname != NULL)
5754                         strfree(saved_poolname);
5755         }
5756 
5757         kmem_free(zc, sizeof (zfs_cmd_t));
5758         return (error);
5759 }
5760 
5761 static int
5762 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
5763 {
5764         if (cmd != DDI_ATTACH)
5765                 return (DDI_FAILURE);
5766 
5767         if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
5768             DDI_PSEUDO, 0) == DDI_FAILURE)
5769                 return (DDI_FAILURE);
5770 
5771         zfs_dip = dip;
5772 
5773         ddi_report_dev(dip);
5774 
5775         return (DDI_SUCCESS);
5776 }
5777 
5778 static int
5779 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
5780 {
5781         if (spa_busy() || zfs_busy() || zvol_busy())
5782                 return (DDI_FAILURE);
5783 
5784         if (cmd != DDI_DETACH)
5785                 return (DDI_FAILURE);
5786 
5787         zfs_dip = NULL;
5788 
5789         ddi_prop_remove_all(dip);
5790         ddi_remove_minor_node(dip, NULL);
5791 
5792         return (DDI_SUCCESS);
5793 }
5794 
5795 /*ARGSUSED*/
5796 static int
5797 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
5798 {
5799         switch (infocmd) {
5800         case DDI_INFO_DEVT2DEVINFO:
5801                 *result = zfs_dip;
5802                 return (DDI_SUCCESS);
5803 
5804         case DDI_INFO_DEVT2INSTANCE:
5805                 *result = (void *)0;
5806                 return (DDI_SUCCESS);
5807         }
5808 
5809         return (DDI_FAILURE);
5810 }
5811 
5812 /*
5813  * OK, so this is a little weird.
5814  *
5815  * /dev/zfs is the control node, i.e. minor 0.
5816  * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
5817  *
5818  * /dev/zfs has basically nothing to do except serve up ioctls,
5819  * so most of the standard driver entry points are in zvol.c.
5820  */
5821 static struct cb_ops zfs_cb_ops = {
5822         zfsdev_open,    /* open */
5823         zfsdev_close,   /* close */
5824         zvol_strategy,  /* strategy */
5825         nodev,          /* print */
5826         zvol_dump,      /* dump */
5827         zvol_read,      /* read */
5828         zvol_write,     /* write */
5829         zfsdev_ioctl,   /* ioctl */
5830         nodev,          /* devmap */
5831         nodev,          /* mmap */
5832         nodev,          /* segmap */
5833         nochpoll,       /* poll */
5834         ddi_prop_op,    /* prop_op */
5835         NULL,           /* streamtab */
5836         D_NEW | D_MP | D_64BIT,         /* Driver compatibility flag */
5837         CB_REV,         /* version */
5838         nodev,          /* async read */
5839         nodev,          /* async write */
5840 };
5841 
5842 static struct dev_ops zfs_dev_ops = {
5843         DEVO_REV,       /* version */
5844         0,              /* refcnt */
5845         zfs_info,       /* info */
5846         nulldev,        /* identify */
5847         nulldev,        /* probe */
5848         zfs_attach,     /* attach */
5849         zfs_detach,     /* detach */
5850         nodev,          /* reset */
5851         &zfs_cb_ops,        /* driver operations */
5852         NULL,           /* no bus operations */
5853         NULL,           /* power */
5854         ddi_quiesce_not_needed, /* quiesce */
5855 };
5856 
5857 static struct modldrv zfs_modldrv = {
5858         &mod_driverops,
5859         "ZFS storage pool",
5860         &zfs_dev_ops
5861 };
5862 
5863 static struct modlinkage modlinkage = {
5864         MODREV_1,
5865         (void *)&zfs_modlfs,
5866         (void *)&zfs_modldrv,
5867         NULL
5868 };
5869 
5870 static void
5871 zfs_allow_log_destroy(void *arg)
5872 {
5873         char *poolname = arg;
5874         strfree(poolname);
5875 }
5876 
5877 int
5878 _init(void)
5879 {
5880         int error;
5881 
5882         spa_init(FREAD | FWRITE);
5883         zfs_init();
5884         zvol_init();
5885         zfs_ioctl_init();
5886 
5887         if ((error = mod_install(&modlinkage)) != 0) {
5888                 zvol_fini();
5889                 zfs_fini();
5890                 spa_fini();
5891                 return (error);
5892         }
5893 
5894         tsd_create(&zfs_fsyncer_key, NULL);
5895         tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
5896         tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
5897 
5898         error = ldi_ident_from_mod(&modlinkage, &zfs_li);
5899         ASSERT(error == 0);
5900         mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
5901 
5902         return (0);
5903 }
5904 
5905 int
5906 _fini(void)
5907 {
5908         int error;
5909 
5910         if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
5911                 return (SET_ERROR(EBUSY));
5912 
5913         if ((error = mod_remove(&modlinkage)) != 0)
5914                 return (error);
5915 
5916         zvol_fini();
5917         zfs_fini();
5918         spa_fini();
5919         if (zfs_nfsshare_inited)
5920                 (void) ddi_modclose(nfs_mod);
5921         if (zfs_smbshare_inited)
5922                 (void) ddi_modclose(smbsrv_mod);
5923         if (zfs_nfsshare_inited || zfs_smbshare_inited)
5924                 (void) ddi_modclose(sharefs_mod);
5925 
5926         tsd_destroy(&zfs_fsyncer_key);
5927         ldi_ident_release(zfs_li);
5928         zfs_li = NULL;
5929         mutex_destroy(&zfs_share_lock);
5930 
5931         return (error);
5932 }
5933 
5934 int
5935 _info(struct modinfo *modinfop)
5936 {
5937         return (mod_info(&modlinkage, modinfop));
5938 }