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 2015, OmniTI Computer Consulting, Inc. All rights reserved.
  27  * Copyright (c) 2014, Joyent, Inc. All rights reserved.
  28  * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
  29  * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
  30  * Copyright (c) 2013 Steven Hartland. All rights reserved.
  31  * Copyright (c) 2014, Nexenta Systems, Inc. All rights reserved.
  32  */
  33 
  34 /*
  35  * ZFS ioctls.
  36  *
  37  * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
  38  * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
  39  *
  40  * There are two ways that we handle ioctls: the legacy way where almost
  41  * all of the logic is in the ioctl callback, and the new way where most
  42  * of the marshalling is handled in the common entry point, zfsdev_ioctl().
  43  *
  44  * Non-legacy ioctls should be registered by calling
  45  * zfs_ioctl_register() from zfs_ioctl_init().  The ioctl is invoked
  46  * from userland by lzc_ioctl().
  47  *
  48  * The registration arguments are as follows:
  49  *
  50  * const char *name
  51  *   The name of the ioctl.  This is used for history logging.  If the
  52  *   ioctl returns successfully (the callback returns 0), and allow_log
  53  *   is true, then a history log entry will be recorded with the input &
  54  *   output nvlists.  The log entry can be printed with "zpool history -i".
  55  *
  56  * zfs_ioc_t ioc
  57  *   The ioctl request number, which userland will pass to ioctl(2).
  58  *   The ioctl numbers can change from release to release, because
  59  *   the caller (libzfs) must be matched to the kernel.
  60  *
  61  * zfs_secpolicy_func_t *secpolicy
  62  *   This function will be called before the zfs_ioc_func_t, to
  63  *   determine if this operation is permitted.  It should return EPERM
  64  *   on failure, and 0 on success.  Checks include determining if the
  65  *   dataset is visible in this zone, and if the user has either all
  66  *   zfs privileges in the zone (SYS_MOUNT), or has been granted permission
  67  *   to do this operation on this dataset with "zfs allow".
  68  *
  69  * zfs_ioc_namecheck_t namecheck
  70  *   This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
  71  *   name, a dataset name, or nothing.  If the name is not well-formed,
  72  *   the ioctl will fail and the callback will not be called.
  73  *   Therefore, the callback can assume that the name is well-formed
  74  *   (e.g. is null-terminated, doesn't have more than one '@' character,
  75  *   doesn't have invalid characters).
  76  *
  77  * zfs_ioc_poolcheck_t pool_check
  78  *   This specifies requirements on the pool state.  If the pool does
  79  *   not meet them (is suspended or is readonly), the ioctl will fail
  80  *   and the callback will not be called.  If any checks are specified
  81  *   (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
  82  *   Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
  83  *   POOL_CHECK_READONLY).
  84  *
  85  * boolean_t smush_outnvlist
  86  *   If smush_outnvlist is true, then the output is presumed to be a
  87  *   list of errors, and it will be "smushed" down to fit into the
  88  *   caller's buffer, by removing some entries and replacing them with a
  89  *   single "N_MORE_ERRORS" entry indicating how many were removed.  See
  90  *   nvlist_smush() for details.  If smush_outnvlist is false, and the
  91  *   outnvlist does not fit into the userland-provided buffer, then the
  92  *   ioctl will fail with ENOMEM.
  93  *
  94  * zfs_ioc_func_t *func
  95  *   The callback function that will perform the operation.
  96  *
  97  *   The callback should return 0 on success, or an error number on
  98  *   failure.  If the function fails, the userland ioctl will return -1,
  99  *   and errno will be set to the callback's return value.  The callback
 100  *   will be called with the following arguments:
 101  *
 102  *   const char *name
 103  *     The name of the pool or dataset to operate on, from
 104  *     zfs_cmd_t:zc_name.  The 'namecheck' argument specifies the
 105  *     expected type (pool, dataset, or none).
 106  *
 107  *   nvlist_t *innvl
 108  *     The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src.  Or
 109  *     NULL if no input nvlist was provided.  Changes to this nvlist are
 110  *     ignored.  If the input nvlist could not be deserialized, the
 111  *     ioctl will fail and the callback will not be called.
 112  *
 113  *   nvlist_t *outnvl
 114  *     The output nvlist, initially empty.  The callback can fill it in,
 115  *     and it will be returned to userland by serializing it into
 116  *     zfs_cmd_t:zc_nvlist_dst.  If it is non-empty, and serialization
 117  *     fails (e.g. because the caller didn't supply a large enough
 118  *     buffer), then the overall ioctl will fail.  See the
 119  *     'smush_nvlist' argument above for additional behaviors.
 120  *
 121  *     There are two typical uses of the output nvlist:
 122  *       - To return state, e.g. property values.  In this case,
 123  *         smush_outnvlist should be false.  If the buffer was not large
 124  *         enough, the caller will reallocate a larger buffer and try
 125  *         the ioctl again.
 126  *
 127  *       - To return multiple errors from an ioctl which makes on-disk
 128  *         changes.  In this case, smush_outnvlist should be true.
 129  *         Ioctls which make on-disk modifications should generally not
 130  *         use the outnvl if they succeed, because the caller can not
 131  *         distinguish between the operation failing, and
 132  *         deserialization failing.
 133  */
 134 
 135 #include <sys/types.h>
 136 #include <sys/param.h>
 137 #include <sys/errno.h>
 138 #include <sys/uio.h>
 139 #include <sys/buf.h>
 140 #include <sys/modctl.h>
 141 #include <sys/open.h>
 142 #include <sys/file.h>
 143 #include <sys/kmem.h>
 144 #include <sys/conf.h>
 145 #include <sys/cmn_err.h>
 146 #include <sys/stat.h>
 147 #include <sys/zfs_ioctl.h>
 148 #include <sys/zfs_vfsops.h>
 149 #include <sys/zfs_znode.h>
 150 #include <sys/zap.h>
 151 #include <sys/spa.h>
 152 #include <sys/spa_impl.h>
 153 #include <sys/vdev.h>
 154 #include <sys/priv_impl.h>
 155 #include <sys/dmu.h>
 156 #include <sys/dsl_dir.h>
 157 #include <sys/dsl_dataset.h>
 158 #include <sys/dsl_prop.h>
 159 #include <sys/dsl_deleg.h>
 160 #include <sys/dmu_objset.h>
 161 #include <sys/dmu_impl.h>
 162 #include <sys/dmu_tx.h>
 163 #include <sys/ddi.h>
 164 #include <sys/sunddi.h>
 165 #include <sys/sunldi.h>
 166 #include <sys/policy.h>
 167 #include <sys/zone.h>
 168 #include <sys/nvpair.h>
 169 #include <sys/pathname.h>
 170 #include <sys/mount.h>
 171 #include <sys/sdt.h>
 172 #include <sys/fs/zfs.h>
 173 #include <sys/zfs_ctldir.h>
 174 #include <sys/zfs_dir.h>
 175 #include <sys/zfs_onexit.h>
 176 #include <sys/zvol.h>
 177 #include <sys/dsl_scan.h>
 178 #include <sharefs/share.h>
 179 #include <sys/dmu_objset.h>
 180 #include <sys/dmu_send.h>
 181 #include <sys/dsl_destroy.h>
 182 #include <sys/dsl_bookmark.h>
 183 #include <sys/dsl_userhold.h>
 184 #include <sys/zfeature.h>
 185 
 186 #include "zfs_namecheck.h"
 187 #include "zfs_prop.h"
 188 #include "zfs_deleg.h"
 189 #include "zfs_comutil.h"
 190 
 191 extern struct modlfs zfs_modlfs;
 192 
 193 extern void zfs_init(void);
 194 extern void zfs_fini(void);
 195 
 196 ldi_ident_t zfs_li = NULL;
 197 dev_info_t *zfs_dip;
 198 
 199 uint_t zfs_fsyncer_key;
 200 extern uint_t rrw_tsd_key;
 201 static uint_t zfs_allow_log_key;
 202 
 203 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
 204 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
 205 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
 206 
 207 typedef enum {
 208         NO_NAME,
 209         POOL_NAME,
 210         DATASET_NAME
 211 } zfs_ioc_namecheck_t;
 212 
 213 typedef enum {
 214         POOL_CHECK_NONE         = 1 << 0,
 215         POOL_CHECK_SUSPENDED    = 1 << 1,
 216         POOL_CHECK_READONLY     = 1 << 2,
 217 } zfs_ioc_poolcheck_t;
 218 
 219 typedef struct zfs_ioc_vec {
 220         zfs_ioc_legacy_func_t   *zvec_legacy_func;
 221         zfs_ioc_func_t          *zvec_func;
 222         zfs_secpolicy_func_t    *zvec_secpolicy;
 223         zfs_ioc_namecheck_t     zvec_namecheck;
 224         boolean_t               zvec_allow_log;
 225         zfs_ioc_poolcheck_t     zvec_pool_check;
 226         boolean_t               zvec_smush_outnvlist;
 227         const char              *zvec_name;
 228 } zfs_ioc_vec_t;
 229 
 230 /* This array is indexed by zfs_userquota_prop_t */
 231 static const char *userquota_perms[] = {
 232         ZFS_DELEG_PERM_USERUSED,
 233         ZFS_DELEG_PERM_USERQUOTA,
 234         ZFS_DELEG_PERM_GROUPUSED,
 235         ZFS_DELEG_PERM_GROUPQUOTA,
 236 };
 237 
 238 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
 239 static int zfs_check_settable(const char *name, nvpair_t *property,
 240     cred_t *cr);
 241 static int zfs_check_clearable(char *dataset, nvlist_t *props,
 242     nvlist_t **errors);
 243 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
 244     boolean_t *);
 245 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
 246 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
 247 
 248 static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature);
 249 
 250 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
 251 void
 252 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
 253 {
 254         const char *newfile;
 255         char buf[512];
 256         va_list adx;
 257 
 258         /*
 259          * Get rid of annoying "../common/" prefix to filename.
 260          */
 261         newfile = strrchr(file, '/');
 262         if (newfile != NULL) {
 263                 newfile = newfile + 1; /* Get rid of leading / */
 264         } else {
 265                 newfile = file;
 266         }
 267 
 268         va_start(adx, fmt);
 269         (void) vsnprintf(buf, sizeof (buf), fmt, adx);
 270         va_end(adx);
 271 
 272         /*
 273          * To get this data, use the zfs-dprintf probe as so:
 274          * dtrace -q -n 'zfs-dprintf \
 275          *      /stringof(arg0) == "dbuf.c"/ \
 276          *      {printf("%s: %s", stringof(arg1), stringof(arg3))}'
 277          * arg0 = file name
 278          * arg1 = function name
 279          * arg2 = line number
 280          * arg3 = message
 281          */
 282         DTRACE_PROBE4(zfs__dprintf,
 283             char *, newfile, char *, func, int, line, char *, buf);
 284 }
 285 
 286 static void
 287 history_str_free(char *buf)
 288 {
 289         kmem_free(buf, HIS_MAX_RECORD_LEN);
 290 }
 291 
 292 static char *
 293 history_str_get(zfs_cmd_t *zc)
 294 {
 295         char *buf;
 296 
 297         if (zc->zc_history == NULL)
 298                 return (NULL);
 299 
 300         buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
 301         if (copyinstr((void *)(uintptr_t)zc->zc_history,
 302             buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
 303                 history_str_free(buf);
 304                 return (NULL);
 305         }
 306 
 307         buf[HIS_MAX_RECORD_LEN -1] = '\0';
 308 
 309         return (buf);
 310 }
 311 
 312 /*
 313  * Check to see if the named dataset is currently defined as bootable
 314  */
 315 static boolean_t
 316 zfs_is_bootfs(const char *name)
 317 {
 318         objset_t *os;
 319 
 320         if (dmu_objset_hold(name, FTAG, &os) == 0) {
 321                 boolean_t ret;
 322                 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
 323                 dmu_objset_rele(os, FTAG);
 324                 return (ret);
 325         }
 326         return (B_FALSE);
 327 }
 328 
 329 /*
 330  * Return non-zero if the spa version is less than requested version.
 331  */
 332 static int
 333 zfs_earlier_version(const char *name, int version)
 334 {
 335         spa_t *spa;
 336 
 337         if (spa_open(name, &spa, FTAG) == 0) {
 338                 if (spa_version(spa) < version) {
 339                         spa_close(spa, FTAG);
 340                         return (1);
 341                 }
 342                 spa_close(spa, FTAG);
 343         }
 344         return (0);
 345 }
 346 
 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         case ZFS_PROP_FILESYSTEM_LIMIT:
 626         case ZFS_PROP_SNAPSHOT_LIMIT:
 627                 if (!INGLOBALZONE(curproc)) {
 628                         uint64_t zoned;
 629                         char setpoint[MAXNAMELEN];
 630                         /*
 631                          * Unprivileged users are allowed to modify the
 632                          * limit on things *under* (ie. contained by)
 633                          * the thing they own.
 634                          */
 635                         if (dsl_prop_get_integer(dsname, "zoned", &zoned,
 636                             setpoint))
 637                                 return (SET_ERROR(EPERM));
 638                         if (!zoned || strlen(dsname) <= strlen(setpoint))
 639                                 return (SET_ERROR(EPERM));
 640                 }
 641                 break;
 642 
 643         case ZFS_PROP_MLSLABEL:
 644                 if (!is_system_labeled())
 645                         return (SET_ERROR(EPERM));
 646 
 647                 if (nvpair_value_string(propval, &strval) == 0) {
 648                         int err;
 649 
 650                         err = zfs_set_slabel_policy(dsname, strval, CRED());
 651                         if (err != 0)
 652                                 return (err);
 653                 }
 654                 break;
 655         }
 656 
 657         return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
 658 }
 659 
 660 /* ARGSUSED */
 661 static int
 662 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 663 {
 664         int error;
 665 
 666         error = zfs_dozonecheck(zc->zc_name, cr);
 667         if (error != 0)
 668                 return (error);
 669 
 670         /*
 671          * permission to set permissions will be evaluated later in
 672          * dsl_deleg_can_allow()
 673          */
 674         return (0);
 675 }
 676 
 677 /* ARGSUSED */
 678 static int
 679 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 680 {
 681         return (zfs_secpolicy_write_perms(zc->zc_name,
 682             ZFS_DELEG_PERM_ROLLBACK, cr));
 683 }
 684 
 685 /* ARGSUSED */
 686 static int
 687 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 688 {
 689         dsl_pool_t *dp;
 690         dsl_dataset_t *ds;
 691         char *cp;
 692         int error;
 693 
 694         /*
 695          * Generate the current snapshot name from the given objsetid, then
 696          * use that name for the secpolicy/zone checks.
 697          */
 698         cp = strchr(zc->zc_name, '@');
 699         if (cp == NULL)
 700                 return (SET_ERROR(EINVAL));
 701         error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
 702         if (error != 0)
 703                 return (error);
 704 
 705         error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
 706         if (error != 0) {
 707                 dsl_pool_rele(dp, FTAG);
 708                 return (error);
 709         }
 710 
 711         dsl_dataset_name(ds, zc->zc_name);
 712 
 713         error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
 714             ZFS_DELEG_PERM_SEND, cr);
 715         dsl_dataset_rele(ds, FTAG);
 716         dsl_pool_rele(dp, FTAG);
 717 
 718         return (error);
 719 }
 720 
 721 /* ARGSUSED */
 722 static int
 723 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 724 {
 725         return (zfs_secpolicy_write_perms(zc->zc_name,
 726             ZFS_DELEG_PERM_SEND, cr));
 727 }
 728 
 729 /* ARGSUSED */
 730 static int
 731 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 732 {
 733         vnode_t *vp;
 734         int error;
 735 
 736         if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
 737             NO_FOLLOW, NULL, &vp)) != 0)
 738                 return (error);
 739 
 740         /* Now make sure mntpnt and dataset are ZFS */
 741 
 742         if (vp->v_vfsp->vfs_fstype != zfsfstype ||
 743             (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
 744             zc->zc_name) != 0)) {
 745                 VN_RELE(vp);
 746                 return (SET_ERROR(EPERM));
 747         }
 748 
 749         VN_RELE(vp);
 750         return (dsl_deleg_access(zc->zc_name,
 751             ZFS_DELEG_PERM_SHARE, cr));
 752 }
 753 
 754 int
 755 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 756 {
 757         if (!INGLOBALZONE(curproc))
 758                 return (SET_ERROR(EPERM));
 759 
 760         if (secpolicy_nfs(cr) == 0) {
 761                 return (0);
 762         } else {
 763                 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
 764         }
 765 }
 766 
 767 int
 768 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 769 {
 770         if (!INGLOBALZONE(curproc))
 771                 return (SET_ERROR(EPERM));
 772 
 773         if (secpolicy_smb(cr) == 0) {
 774                 return (0);
 775         } else {
 776                 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
 777         }
 778 }
 779 
 780 static int
 781 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
 782 {
 783         char *cp;
 784 
 785         /*
 786          * Remove the @bla or /bla from the end of the name to get the parent.
 787          */
 788         (void) strncpy(parent, datasetname, parentsize);
 789         cp = strrchr(parent, '@');
 790         if (cp != NULL) {
 791                 cp[0] = '\0';
 792         } else {
 793                 cp = strrchr(parent, '/');
 794                 if (cp == NULL)
 795                         return (SET_ERROR(ENOENT));
 796                 cp[0] = '\0';
 797         }
 798 
 799         return (0);
 800 }
 801 
 802 int
 803 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
 804 {
 805         int error;
 806 
 807         if ((error = zfs_secpolicy_write_perms(name,
 808             ZFS_DELEG_PERM_MOUNT, cr)) != 0)
 809                 return (error);
 810 
 811         return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
 812 }
 813 
 814 /* ARGSUSED */
 815 static int
 816 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 817 {
 818         return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
 819 }
 820 
 821 /*
 822  * Destroying snapshots with delegated permissions requires
 823  * descendant mount and destroy permissions.
 824  */
 825 /* ARGSUSED */
 826 static int
 827 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 828 {
 829         nvlist_t *snaps;
 830         nvpair_t *pair, *nextpair;
 831         int error = 0;
 832 
 833         if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
 834                 return (SET_ERROR(EINVAL));
 835         for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
 836             pair = nextpair) {
 837                 nextpair = nvlist_next_nvpair(snaps, pair);
 838                 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
 839                 if (error == ENOENT) {
 840                         /*
 841                          * Ignore any snapshots that don't exist (we consider
 842                          * them "already destroyed").  Remove the name from the
 843                          * nvl here in case the snapshot is created between
 844                          * now and when we try to destroy it (in which case
 845                          * we don't want to destroy it since we haven't
 846                          * checked for permission).
 847                          */
 848                         fnvlist_remove_nvpair(snaps, pair);
 849                         error = 0;
 850                 }
 851                 if (error != 0)
 852                         break;
 853         }
 854 
 855         return (error);
 856 }
 857 
 858 int
 859 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
 860 {
 861         char    parentname[MAXNAMELEN];
 862         int     error;
 863 
 864         if ((error = zfs_secpolicy_write_perms(from,
 865             ZFS_DELEG_PERM_RENAME, cr)) != 0)
 866                 return (error);
 867 
 868         if ((error = zfs_secpolicy_write_perms(from,
 869             ZFS_DELEG_PERM_MOUNT, cr)) != 0)
 870                 return (error);
 871 
 872         if ((error = zfs_get_parent(to, parentname,
 873             sizeof (parentname))) != 0)
 874                 return (error);
 875 
 876         if ((error = zfs_secpolicy_write_perms(parentname,
 877             ZFS_DELEG_PERM_CREATE, cr)) != 0)
 878                 return (error);
 879 
 880         if ((error = zfs_secpolicy_write_perms(parentname,
 881             ZFS_DELEG_PERM_MOUNT, cr)) != 0)
 882                 return (error);
 883 
 884         return (error);
 885 }
 886 
 887 /* ARGSUSED */
 888 static int
 889 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 890 {
 891         return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
 892 }
 893 
 894 /* ARGSUSED */
 895 static int
 896 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 897 {
 898         dsl_pool_t *dp;
 899         dsl_dataset_t *clone;
 900         int error;
 901 
 902         error = zfs_secpolicy_write_perms(zc->zc_name,
 903             ZFS_DELEG_PERM_PROMOTE, cr);
 904         if (error != 0)
 905                 return (error);
 906 
 907         error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
 908         if (error != 0)
 909                 return (error);
 910 
 911         error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
 912 
 913         if (error == 0) {
 914                 char parentname[MAXNAMELEN];
 915                 dsl_dataset_t *origin = NULL;
 916                 dsl_dir_t *dd;
 917                 dd = clone->ds_dir;
 918 
 919                 error = dsl_dataset_hold_obj(dd->dd_pool,
 920                     dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin);
 921                 if (error != 0) {
 922                         dsl_dataset_rele(clone, FTAG);
 923                         dsl_pool_rele(dp, FTAG);
 924                         return (error);
 925                 }
 926 
 927                 error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
 928                     ZFS_DELEG_PERM_MOUNT, cr);
 929 
 930                 dsl_dataset_name(origin, parentname);
 931                 if (error == 0) {
 932                         error = zfs_secpolicy_write_perms_ds(parentname, origin,
 933                             ZFS_DELEG_PERM_PROMOTE, cr);
 934                 }
 935                 dsl_dataset_rele(clone, FTAG);
 936                 dsl_dataset_rele(origin, FTAG);
 937         }
 938         dsl_pool_rele(dp, FTAG);
 939         return (error);
 940 }
 941 
 942 /* ARGSUSED */
 943 static int
 944 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 945 {
 946         int error;
 947 
 948         if ((error = zfs_secpolicy_write_perms(zc->zc_name,
 949             ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
 950                 return (error);
 951 
 952         if ((error = zfs_secpolicy_write_perms(zc->zc_name,
 953             ZFS_DELEG_PERM_MOUNT, cr)) != 0)
 954                 return (error);
 955 
 956         return (zfs_secpolicy_write_perms(zc->zc_name,
 957             ZFS_DELEG_PERM_CREATE, cr));
 958 }
 959 
 960 int
 961 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
 962 {
 963         return (zfs_secpolicy_write_perms(name,
 964             ZFS_DELEG_PERM_SNAPSHOT, cr));
 965 }
 966 
 967 /*
 968  * Check for permission to create each snapshot in the nvlist.
 969  */
 970 /* ARGSUSED */
 971 static int
 972 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 973 {
 974         nvlist_t *snaps;
 975         int error = 0;
 976         nvpair_t *pair;
 977 
 978         if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
 979                 return (SET_ERROR(EINVAL));
 980         for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
 981             pair = nvlist_next_nvpair(snaps, pair)) {
 982                 char *name = nvpair_name(pair);
 983                 char *atp = strchr(name, '@');
 984 
 985                 if (atp == NULL) {
 986                         error = SET_ERROR(EINVAL);
 987                         break;
 988                 }
 989                 *atp = '\0';
 990                 error = zfs_secpolicy_snapshot_perms(name, cr);
 991                 *atp = '@';
 992                 if (error != 0)
 993                         break;
 994         }
 995         return (error);
 996 }
 997 
 998 /*
 999  * Check for permission to create each snapshot in the nvlist.
1000  */
1001 /* ARGSUSED */
1002 static int
1003 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1004 {
1005         int error = 0;
1006 
1007         for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
1008             pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
1009                 char *name = nvpair_name(pair);
1010                 char *hashp = strchr(name, '#');
1011 
1012                 if (hashp == NULL) {
1013                         error = SET_ERROR(EINVAL);
1014                         break;
1015                 }
1016                 *hashp = '\0';
1017                 error = zfs_secpolicy_write_perms(name,
1018                     ZFS_DELEG_PERM_BOOKMARK, cr);
1019                 *hashp = '#';
1020                 if (error != 0)
1021                         break;
1022         }
1023         return (error);
1024 }
1025 
1026 /* ARGSUSED */
1027 static int
1028 zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1029 {
1030         nvpair_t *pair, *nextpair;
1031         int error = 0;
1032 
1033         for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1034             pair = nextpair) {
1035                 char *name = nvpair_name(pair);
1036                 char *hashp = strchr(name, '#');
1037                 nextpair = nvlist_next_nvpair(innvl, pair);
1038 
1039                 if (hashp == NULL) {
1040                         error = SET_ERROR(EINVAL);
1041                         break;
1042                 }
1043 
1044                 *hashp = '\0';
1045                 error = zfs_secpolicy_write_perms(name,
1046                     ZFS_DELEG_PERM_DESTROY, cr);
1047                 *hashp = '#';
1048                 if (error == ENOENT) {
1049                         /*
1050                          * Ignore any filesystems that don't exist (we consider
1051                          * their bookmarks "already destroyed").  Remove
1052                          * the name from the nvl here in case the filesystem
1053                          * is created between now and when we try to destroy
1054                          * the bookmark (in which case we don't want to
1055                          * destroy it since we haven't checked for permission).
1056                          */
1057                         fnvlist_remove_nvpair(innvl, pair);
1058                         error = 0;
1059                 }
1060                 if (error != 0)
1061                         break;
1062         }
1063 
1064         return (error);
1065 }
1066 
1067 /* ARGSUSED */
1068 static int
1069 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1070 {
1071         /*
1072          * Even root must have a proper TSD so that we know what pool
1073          * to log to.
1074          */
1075         if (tsd_get(zfs_allow_log_key) == NULL)
1076                 return (SET_ERROR(EPERM));
1077         return (0);
1078 }
1079 
1080 static int
1081 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1082 {
1083         char    parentname[MAXNAMELEN];
1084         int     error;
1085         char    *origin;
1086 
1087         if ((error = zfs_get_parent(zc->zc_name, parentname,
1088             sizeof (parentname))) != 0)
1089                 return (error);
1090 
1091         if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1092             (error = zfs_secpolicy_write_perms(origin,
1093             ZFS_DELEG_PERM_CLONE, cr)) != 0)
1094                 return (error);
1095 
1096         if ((error = zfs_secpolicy_write_perms(parentname,
1097             ZFS_DELEG_PERM_CREATE, cr)) != 0)
1098                 return (error);
1099 
1100         return (zfs_secpolicy_write_perms(parentname,
1101             ZFS_DELEG_PERM_MOUNT, cr));
1102 }
1103 
1104 /*
1105  * Policy for pool operations - create/destroy pools, add vdevs, etc.  Requires
1106  * SYS_CONFIG privilege, which is not available in a local zone.
1107  */
1108 /* ARGSUSED */
1109 static int
1110 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1111 {
1112         if (secpolicy_sys_config(cr, B_FALSE) != 0)
1113                 return (SET_ERROR(EPERM));
1114 
1115         return (0);
1116 }
1117 
1118 /*
1119  * Policy for object to name lookups.
1120  */
1121 /* ARGSUSED */
1122 static int
1123 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1124 {
1125         int error;
1126 
1127         if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1128                 return (0);
1129 
1130         error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1131         return (error);
1132 }
1133 
1134 /*
1135  * Policy for fault injection.  Requires all privileges.
1136  */
1137 /* ARGSUSED */
1138 static int
1139 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1140 {
1141         return (secpolicy_zinject(cr));
1142 }
1143 
1144 /* ARGSUSED */
1145 static int
1146 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1147 {
1148         zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1149 
1150         if (prop == ZPROP_INVAL) {
1151                 if (!zfs_prop_user(zc->zc_value))
1152                         return (SET_ERROR(EINVAL));
1153                 return (zfs_secpolicy_write_perms(zc->zc_name,
1154                     ZFS_DELEG_PERM_USERPROP, cr));
1155         } else {
1156                 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1157                     NULL, cr));
1158         }
1159 }
1160 
1161 static int
1162 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1163 {
1164         int err = zfs_secpolicy_read(zc, innvl, cr);
1165         if (err)
1166                 return (err);
1167 
1168         if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1169                 return (SET_ERROR(EINVAL));
1170 
1171         if (zc->zc_value[0] == 0) {
1172                 /*
1173                  * They are asking about a posix uid/gid.  If it's
1174                  * themself, allow it.
1175                  */
1176                 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1177                     zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
1178                         if (zc->zc_guid == crgetuid(cr))
1179                                 return (0);
1180                 } else {
1181                         if (groupmember(zc->zc_guid, cr))
1182                                 return (0);
1183                 }
1184         }
1185 
1186         return (zfs_secpolicy_write_perms(zc->zc_name,
1187             userquota_perms[zc->zc_objset_type], cr));
1188 }
1189 
1190 static int
1191 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1192 {
1193         int err = zfs_secpolicy_read(zc, innvl, cr);
1194         if (err)
1195                 return (err);
1196 
1197         if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1198                 return (SET_ERROR(EINVAL));
1199 
1200         return (zfs_secpolicy_write_perms(zc->zc_name,
1201             userquota_perms[zc->zc_objset_type], cr));
1202 }
1203 
1204 /* ARGSUSED */
1205 static int
1206 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1207 {
1208         return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1209             NULL, cr));
1210 }
1211 
1212 /* ARGSUSED */
1213 static int
1214 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1215 {
1216         nvpair_t *pair;
1217         nvlist_t *holds;
1218         int error;
1219 
1220         error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1221         if (error != 0)
1222                 return (SET_ERROR(EINVAL));
1223 
1224         for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1225             pair = nvlist_next_nvpair(holds, pair)) {
1226                 char fsname[MAXNAMELEN];
1227                 error = dmu_fsname(nvpair_name(pair), fsname);
1228                 if (error != 0)
1229                         return (error);
1230                 error = zfs_secpolicy_write_perms(fsname,
1231                     ZFS_DELEG_PERM_HOLD, cr);
1232                 if (error != 0)
1233                         return (error);
1234         }
1235         return (0);
1236 }
1237 
1238 /* ARGSUSED */
1239 static int
1240 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1241 {
1242         nvpair_t *pair;
1243         int error;
1244 
1245         for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1246             pair = nvlist_next_nvpair(innvl, pair)) {
1247                 char fsname[MAXNAMELEN];
1248                 error = dmu_fsname(nvpair_name(pair), fsname);
1249                 if (error != 0)
1250                         return (error);
1251                 error = zfs_secpolicy_write_perms(fsname,
1252                     ZFS_DELEG_PERM_RELEASE, cr);
1253                 if (error != 0)
1254                         return (error);
1255         }
1256         return (0);
1257 }
1258 
1259 /*
1260  * Policy for allowing temporary snapshots to be taken or released
1261  */
1262 static int
1263 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1264 {
1265         /*
1266          * A temporary snapshot is the same as a snapshot,
1267          * hold, destroy and release all rolled into one.
1268          * Delegated diff alone is sufficient that we allow this.
1269          */
1270         int error;
1271 
1272         if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1273             ZFS_DELEG_PERM_DIFF, cr)) == 0)
1274                 return (0);
1275 
1276         error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1277         if (error == 0)
1278                 error = zfs_secpolicy_hold(zc, innvl, cr);
1279         if (error == 0)
1280                 error = zfs_secpolicy_release(zc, innvl, cr);
1281         if (error == 0)
1282                 error = zfs_secpolicy_destroy(zc, innvl, cr);
1283         return (error);
1284 }
1285 
1286 /*
1287  * Returns the nvlist as specified by the user in the zfs_cmd_t.
1288  */
1289 static int
1290 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1291 {
1292         char *packed;
1293         int error;
1294         nvlist_t *list = NULL;
1295 
1296         /*
1297          * Read in and unpack the user-supplied nvlist.
1298          */
1299         if (size == 0)
1300                 return (SET_ERROR(EINVAL));
1301 
1302         packed = kmem_alloc(size, KM_SLEEP);
1303 
1304         if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1305             iflag)) != 0) {
1306                 kmem_free(packed, size);
1307                 return (error);
1308         }
1309 
1310         if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1311                 kmem_free(packed, size);
1312                 return (error);
1313         }
1314 
1315         kmem_free(packed, size);
1316 
1317         *nvp = list;
1318         return (0);
1319 }
1320 
1321 /*
1322  * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1323  * Entries will be removed from the end of the nvlist, and one int32 entry
1324  * named "N_MORE_ERRORS" will be added indicating how many entries were
1325  * removed.
1326  */
1327 static int
1328 nvlist_smush(nvlist_t *errors, size_t max)
1329 {
1330         size_t size;
1331 
1332         size = fnvlist_size(errors);
1333 
1334         if (size > max) {
1335                 nvpair_t *more_errors;
1336                 int n = 0;
1337 
1338                 if (max < 1024)
1339                         return (SET_ERROR(ENOMEM));
1340 
1341                 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1342                 more_errors = nvlist_prev_nvpair(errors, NULL);
1343 
1344                 do {
1345                         nvpair_t *pair = nvlist_prev_nvpair(errors,
1346                             more_errors);
1347                         fnvlist_remove_nvpair(errors, pair);
1348                         n++;
1349                         size = fnvlist_size(errors);
1350                 } while (size > max);
1351 
1352                 fnvlist_remove_nvpair(errors, more_errors);
1353                 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1354                 ASSERT3U(fnvlist_size(errors), <=, max);
1355         }
1356 
1357         return (0);
1358 }
1359 
1360 static int
1361 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1362 {
1363         char *packed = NULL;
1364         int error = 0;
1365         size_t size;
1366 
1367         size = fnvlist_size(nvl);
1368 
1369         if (size > zc->zc_nvlist_dst_size) {
1370                 error = SET_ERROR(ENOMEM);
1371         } else {
1372                 packed = fnvlist_pack(nvl, &size);
1373                 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1374                     size, zc->zc_iflags) != 0)
1375                         error = SET_ERROR(EFAULT);
1376                 fnvlist_pack_free(packed, size);
1377         }
1378 
1379         zc->zc_nvlist_dst_size = size;
1380         zc->zc_nvlist_dst_filled = B_TRUE;
1381         return (error);
1382 }
1383 
1384 static int
1385 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1386 {
1387         objset_t *os;
1388         int error;
1389 
1390         error = dmu_objset_hold(dsname, FTAG, &os);
1391         if (error != 0)
1392                 return (error);
1393         if (dmu_objset_type(os) != DMU_OST_ZFS) {
1394                 dmu_objset_rele(os, FTAG);
1395                 return (SET_ERROR(EINVAL));
1396         }
1397 
1398         mutex_enter(&os->os_user_ptr_lock);
1399         *zfvp = dmu_objset_get_user(os);
1400         if (*zfvp) {
1401                 VFS_HOLD((*zfvp)->z_vfs);
1402         } else {
1403                 error = SET_ERROR(ESRCH);
1404         }
1405         mutex_exit(&os->os_user_ptr_lock);
1406         dmu_objset_rele(os, FTAG);
1407         return (error);
1408 }
1409 
1410 /*
1411  * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1412  * case its z_vfs will be NULL, and it will be opened as the owner.
1413  * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1414  * which prevents all vnode ops from running.
1415  */
1416 static int
1417 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1418 {
1419         int error = 0;
1420 
1421         if (getzfsvfs(name, zfvp) != 0)
1422                 error = zfsvfs_create(name, zfvp);
1423         if (error == 0) {
1424                 rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1425                     RW_READER, tag);
1426                 if ((*zfvp)->z_unmounted) {
1427                         /*
1428                          * XXX we could probably try again, since the unmounting
1429                          * thread should be just about to disassociate the
1430                          * objset from the zfsvfs.
1431                          */
1432                         rrm_exit(&(*zfvp)->z_teardown_lock, tag);
1433                         return (SET_ERROR(EBUSY));
1434                 }
1435         }
1436         return (error);
1437 }
1438 
1439 static void
1440 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1441 {
1442         rrm_exit(&zfsvfs->z_teardown_lock, tag);
1443 
1444         if (zfsvfs->z_vfs) {
1445                 VFS_RELE(zfsvfs->z_vfs);
1446         } else {
1447                 dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1448                 zfsvfs_free(zfsvfs);
1449         }
1450 }
1451 
1452 static int
1453 zfs_ioc_pool_create(zfs_cmd_t *zc)
1454 {
1455         int error;
1456         nvlist_t *config, *props = NULL;
1457         nvlist_t *rootprops = NULL;
1458         nvlist_t *zplprops = NULL;
1459 
1460         if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1461             zc->zc_iflags, &config))
1462                 return (error);
1463 
1464         if (zc->zc_nvlist_src_size != 0 && (error =
1465             get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1466             zc->zc_iflags, &props))) {
1467                 nvlist_free(config);
1468                 return (error);
1469         }
1470 
1471         if (props) {
1472                 nvlist_t *nvl = NULL;
1473                 uint64_t version = SPA_VERSION;
1474 
1475                 (void) nvlist_lookup_uint64(props,
1476                     zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1477                 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1478                         error = SET_ERROR(EINVAL);
1479                         goto pool_props_bad;
1480                 }
1481                 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1482                 if (nvl) {
1483                         error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1484                         if (error != 0) {
1485                                 nvlist_free(config);
1486                                 nvlist_free(props);
1487                                 return (error);
1488                         }
1489                         (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1490                 }
1491                 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1492                 error = zfs_fill_zplprops_root(version, rootprops,
1493                     zplprops, NULL);
1494                 if (error != 0)
1495                         goto pool_props_bad;
1496         }
1497 
1498         error = spa_create(zc->zc_name, config, props, zplprops);
1499 
1500         /*
1501          * Set the remaining root properties
1502          */
1503         if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1504             ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1505                 (void) spa_destroy(zc->zc_name);
1506 
1507 pool_props_bad:
1508         nvlist_free(rootprops);
1509         nvlist_free(zplprops);
1510         nvlist_free(config);
1511         nvlist_free(props);
1512 
1513         return (error);
1514 }
1515 
1516 static int
1517 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1518 {
1519         int error;
1520         zfs_log_history(zc);
1521         error = spa_destroy(zc->zc_name);
1522         if (error == 0)
1523                 zvol_remove_minors(zc->zc_name);
1524         return (error);
1525 }
1526 
1527 static int
1528 zfs_ioc_pool_import(zfs_cmd_t *zc)
1529 {
1530         nvlist_t *config, *props = NULL;
1531         uint64_t guid;
1532         int error;
1533 
1534         if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1535             zc->zc_iflags, &config)) != 0)
1536                 return (error);
1537 
1538         if (zc->zc_nvlist_src_size != 0 && (error =
1539             get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1540             zc->zc_iflags, &props))) {
1541                 nvlist_free(config);
1542                 return (error);
1543         }
1544 
1545         if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1546             guid != zc->zc_guid)
1547                 error = SET_ERROR(EINVAL);
1548         else
1549                 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1550 
1551         if (zc->zc_nvlist_dst != 0) {
1552                 int err;
1553 
1554                 if ((err = put_nvlist(zc, config)) != 0)
1555                         error = err;
1556         }
1557 
1558         nvlist_free(config);
1559 
1560         if (props)
1561                 nvlist_free(props);
1562 
1563         return (error);
1564 }
1565 
1566 static int
1567 zfs_ioc_pool_export(zfs_cmd_t *zc)
1568 {
1569         int error;
1570         boolean_t force = (boolean_t)zc->zc_cookie;
1571         boolean_t hardforce = (boolean_t)zc->zc_guid;
1572 
1573         zfs_log_history(zc);
1574         error = spa_export(zc->zc_name, NULL, force, hardforce);
1575         if (error == 0)
1576                 zvol_remove_minors(zc->zc_name);
1577         return (error);
1578 }
1579 
1580 static int
1581 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1582 {
1583         nvlist_t *configs;
1584         int error;
1585 
1586         if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1587                 return (SET_ERROR(EEXIST));
1588 
1589         error = put_nvlist(zc, configs);
1590 
1591         nvlist_free(configs);
1592 
1593         return (error);
1594 }
1595 
1596 /*
1597  * inputs:
1598  * zc_name              name of the pool
1599  *
1600  * outputs:
1601  * zc_cookie            real errno
1602  * zc_nvlist_dst        config nvlist
1603  * zc_nvlist_dst_size   size of config nvlist
1604  */
1605 static int
1606 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1607 {
1608         nvlist_t *config;
1609         int error;
1610         int ret = 0;
1611 
1612         error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1613             sizeof (zc->zc_value));
1614 
1615         if (config != NULL) {
1616                 ret = put_nvlist(zc, config);
1617                 nvlist_free(config);
1618 
1619                 /*
1620                  * The config may be present even if 'error' is non-zero.
1621                  * In this case we return success, and preserve the real errno
1622                  * in 'zc_cookie'.
1623                  */
1624                 zc->zc_cookie = error;
1625         } else {
1626                 ret = error;
1627         }
1628 
1629         return (ret);
1630 }
1631 
1632 /*
1633  * Try to import the given pool, returning pool stats as appropriate so that
1634  * user land knows which devices are available and overall pool health.
1635  */
1636 static int
1637 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1638 {
1639         nvlist_t *tryconfig, *config;
1640         int error;
1641 
1642         if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1643             zc->zc_iflags, &tryconfig)) != 0)
1644                 return (error);
1645 
1646         config = spa_tryimport(tryconfig);
1647 
1648         nvlist_free(tryconfig);
1649 
1650         if (config == NULL)
1651                 return (SET_ERROR(EINVAL));
1652 
1653         error = put_nvlist(zc, config);
1654         nvlist_free(config);
1655 
1656         return (error);
1657 }
1658 
1659 /*
1660  * inputs:
1661  * zc_name              name of the pool
1662  * zc_cookie            scan func (pool_scan_func_t)
1663  */
1664 static int
1665 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1666 {
1667         spa_t *spa;
1668         int error;
1669 
1670         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1671                 return (error);
1672 
1673         if (zc->zc_cookie == POOL_SCAN_NONE)
1674                 error = spa_scan_stop(spa);
1675         else
1676                 error = spa_scan(spa, zc->zc_cookie);
1677 
1678         spa_close(spa, FTAG);
1679 
1680         return (error);
1681 }
1682 
1683 static int
1684 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1685 {
1686         spa_t *spa;
1687         int error;
1688 
1689         error = spa_open(zc->zc_name, &spa, FTAG);
1690         if (error == 0) {
1691                 spa_freeze(spa);
1692                 spa_close(spa, FTAG);
1693         }
1694         return (error);
1695 }
1696 
1697 static int
1698 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1699 {
1700         spa_t *spa;
1701         int error;
1702 
1703         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1704                 return (error);
1705 
1706         if (zc->zc_cookie < spa_version(spa) ||
1707             !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1708                 spa_close(spa, FTAG);
1709                 return (SET_ERROR(EINVAL));
1710         }
1711 
1712         spa_upgrade(spa, zc->zc_cookie);
1713         spa_close(spa, FTAG);
1714 
1715         return (error);
1716 }
1717 
1718 static int
1719 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1720 {
1721         spa_t *spa;
1722         char *hist_buf;
1723         uint64_t size;
1724         int error;
1725 
1726         if ((size = zc->zc_history_len) == 0)
1727                 return (SET_ERROR(EINVAL));
1728 
1729         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1730                 return (error);
1731 
1732         if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1733                 spa_close(spa, FTAG);
1734                 return (SET_ERROR(ENOTSUP));
1735         }
1736 
1737         hist_buf = kmem_alloc(size, KM_SLEEP);
1738         if ((error = spa_history_get(spa, &zc->zc_history_offset,
1739             &zc->zc_history_len, hist_buf)) == 0) {
1740                 error = ddi_copyout(hist_buf,
1741                     (void *)(uintptr_t)zc->zc_history,
1742                     zc->zc_history_len, zc->zc_iflags);
1743         }
1744 
1745         spa_close(spa, FTAG);
1746         kmem_free(hist_buf, size);
1747         return (error);
1748 }
1749 
1750 static int
1751 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1752 {
1753         spa_t *spa;
1754         int error;
1755 
1756         error = spa_open(zc->zc_name, &spa, FTAG);
1757         if (error == 0) {
1758                 error = spa_change_guid(spa);
1759                 spa_close(spa, FTAG);
1760         }
1761         return (error);
1762 }
1763 
1764 static int
1765 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1766 {
1767         return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1768 }
1769 
1770 /*
1771  * inputs:
1772  * zc_name              name of filesystem
1773  * zc_obj               object to find
1774  *
1775  * outputs:
1776  * zc_value             name of object
1777  */
1778 static int
1779 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1780 {
1781         objset_t *os;
1782         int error;
1783 
1784         /* XXX reading from objset not owned */
1785         if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1786                 return (error);
1787         if (dmu_objset_type(os) != DMU_OST_ZFS) {
1788                 dmu_objset_rele(os, FTAG);
1789                 return (SET_ERROR(EINVAL));
1790         }
1791         error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1792             sizeof (zc->zc_value));
1793         dmu_objset_rele(os, FTAG);
1794 
1795         return (error);
1796 }
1797 
1798 /*
1799  * inputs:
1800  * zc_name              name of filesystem
1801  * zc_obj               object to find
1802  *
1803  * outputs:
1804  * zc_stat              stats on object
1805  * zc_value             path to object
1806  */
1807 static int
1808 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1809 {
1810         objset_t *os;
1811         int error;
1812 
1813         /* XXX reading from objset not owned */
1814         if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1815                 return (error);
1816         if (dmu_objset_type(os) != DMU_OST_ZFS) {
1817                 dmu_objset_rele(os, FTAG);
1818                 return (SET_ERROR(EINVAL));
1819         }
1820         error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1821             sizeof (zc->zc_value));
1822         dmu_objset_rele(os, FTAG);
1823 
1824         return (error);
1825 }
1826 
1827 static int
1828 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1829 {
1830         spa_t *spa;
1831         int error;
1832         nvlist_t *config, **l2cache, **spares;
1833         uint_t nl2cache = 0, nspares = 0;
1834 
1835         error = spa_open(zc->zc_name, &spa, FTAG);
1836         if (error != 0)
1837                 return (error);
1838 
1839         error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1840             zc->zc_iflags, &config);
1841         (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1842             &l2cache, &nl2cache);
1843 
1844         (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1845             &spares, &nspares);
1846 
1847         /*
1848          * A root pool with concatenated devices is not supported.
1849          * Thus, can not add a device to a root pool.
1850          *
1851          * Intent log device can not be added to a rootpool because
1852          * during mountroot, zil is replayed, a seperated log device
1853          * can not be accessed during the mountroot time.
1854          *
1855          * l2cache and spare devices are ok to be added to a rootpool.
1856          */
1857         if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1858                 nvlist_free(config);
1859                 spa_close(spa, FTAG);
1860                 return (SET_ERROR(EDOM));
1861         }
1862 
1863         if (error == 0) {
1864                 error = spa_vdev_add(spa, config);
1865                 nvlist_free(config);
1866         }
1867         spa_close(spa, FTAG);
1868         return (error);
1869 }
1870 
1871 /*
1872  * inputs:
1873  * zc_name              name of the pool
1874  * zc_nvlist_conf       nvlist of devices to remove
1875  * zc_cookie            to stop the remove?
1876  */
1877 static int
1878 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1879 {
1880         spa_t *spa;
1881         int error;
1882 
1883         error = spa_open(zc->zc_name, &spa, FTAG);
1884         if (error != 0)
1885                 return (error);
1886         error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1887         spa_close(spa, FTAG);
1888         return (error);
1889 }
1890 
1891 static int
1892 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1893 {
1894         spa_t *spa;
1895         int error;
1896         vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1897 
1898         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1899                 return (error);
1900         switch (zc->zc_cookie) {
1901         case VDEV_STATE_ONLINE:
1902                 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1903                 break;
1904 
1905         case VDEV_STATE_OFFLINE:
1906                 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1907                 break;
1908 
1909         case VDEV_STATE_FAULTED:
1910                 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1911                     zc->zc_obj != VDEV_AUX_EXTERNAL)
1912                         zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1913 
1914                 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1915                 break;
1916 
1917         case VDEV_STATE_DEGRADED:
1918                 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1919                     zc->zc_obj != VDEV_AUX_EXTERNAL)
1920                         zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1921 
1922                 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1923                 break;
1924 
1925         default:
1926                 error = SET_ERROR(EINVAL);
1927         }
1928         zc->zc_cookie = newstate;
1929         spa_close(spa, FTAG);
1930         return (error);
1931 }
1932 
1933 static int
1934 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1935 {
1936         spa_t *spa;
1937         int replacing = zc->zc_cookie;
1938         nvlist_t *config;
1939         int error;
1940 
1941         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1942                 return (error);
1943 
1944         if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1945             zc->zc_iflags, &config)) == 0) {
1946                 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1947                 nvlist_free(config);
1948         }
1949 
1950         spa_close(spa, FTAG);
1951         return (error);
1952 }
1953 
1954 static int
1955 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1956 {
1957         spa_t *spa;
1958         int error;
1959 
1960         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1961                 return (error);
1962 
1963         error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1964 
1965         spa_close(spa, FTAG);
1966         return (error);
1967 }
1968 
1969 static int
1970 zfs_ioc_vdev_split(zfs_cmd_t *zc)
1971 {
1972         spa_t *spa;
1973         nvlist_t *config, *props = NULL;
1974         int error;
1975         boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
1976 
1977         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1978                 return (error);
1979 
1980         if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1981             zc->zc_iflags, &config)) {
1982                 spa_close(spa, FTAG);
1983                 return (error);
1984         }
1985 
1986         if (zc->zc_nvlist_src_size != 0 && (error =
1987             get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1988             zc->zc_iflags, &props))) {
1989                 spa_close(spa, FTAG);
1990                 nvlist_free(config);
1991                 return (error);
1992         }
1993 
1994         error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
1995 
1996         spa_close(spa, FTAG);
1997 
1998         nvlist_free(config);
1999         nvlist_free(props);
2000 
2001         return (error);
2002 }
2003 
2004 static int
2005 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
2006 {
2007         spa_t *spa;
2008         char *path = zc->zc_value;
2009         uint64_t guid = zc->zc_guid;
2010         int error;
2011 
2012         error = spa_open(zc->zc_name, &spa, FTAG);
2013         if (error != 0)
2014                 return (error);
2015 
2016         error = spa_vdev_setpath(spa, guid, path);
2017         spa_close(spa, FTAG);
2018         return (error);
2019 }
2020 
2021 static int
2022 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
2023 {
2024         spa_t *spa;
2025         char *fru = zc->zc_value;
2026         uint64_t guid = zc->zc_guid;
2027         int error;
2028 
2029         error = spa_open(zc->zc_name, &spa, FTAG);
2030         if (error != 0)
2031                 return (error);
2032 
2033         error = spa_vdev_setfru(spa, guid, fru);
2034         spa_close(spa, FTAG);
2035         return (error);
2036 }
2037 
2038 static int
2039 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
2040 {
2041         int error = 0;
2042         nvlist_t *nv;
2043 
2044         dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2045 
2046         if (zc->zc_nvlist_dst != 0 &&
2047             (error = dsl_prop_get_all(os, &nv)) == 0) {
2048                 dmu_objset_stats(os, nv);
2049                 /*
2050                  * NB: zvol_get_stats() will read the objset contents,
2051                  * which we aren't supposed to do with a
2052                  * DS_MODE_USER hold, because it could be
2053                  * inconsistent.  So this is a bit of a workaround...
2054                  * XXX reading with out owning
2055                  */
2056                 if (!zc->zc_objset_stats.dds_inconsistent &&
2057                     dmu_objset_type(os) == DMU_OST_ZVOL) {
2058                         error = zvol_get_stats(os, nv);
2059                         if (error == EIO)
2060                                 return (error);
2061                         VERIFY0(error);
2062                 }
2063                 error = put_nvlist(zc, nv);
2064                 nvlist_free(nv);
2065         }
2066 
2067         return (error);
2068 }
2069 
2070 /*
2071  * inputs:
2072  * zc_name              name of filesystem
2073  * zc_nvlist_dst_size   size of buffer for property nvlist
2074  *
2075  * outputs:
2076  * zc_objset_stats      stats
2077  * zc_nvlist_dst        property nvlist
2078  * zc_nvlist_dst_size   size of property nvlist
2079  */
2080 static int
2081 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2082 {
2083         objset_t *os;
2084         int error;
2085 
2086         error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2087         if (error == 0) {
2088                 error = zfs_ioc_objset_stats_impl(zc, os);
2089                 dmu_objset_rele(os, FTAG);
2090         }
2091 
2092         return (error);
2093 }
2094 
2095 /*
2096  * inputs:
2097  * zc_name              name of filesystem
2098  * zc_nvlist_dst_size   size of buffer for property nvlist
2099  *
2100  * outputs:
2101  * zc_nvlist_dst        received property nvlist
2102  * zc_nvlist_dst_size   size of received property nvlist
2103  *
2104  * Gets received properties (distinct from local properties on or after
2105  * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2106  * local property values.
2107  */
2108 static int
2109 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2110 {
2111         int error = 0;
2112         nvlist_t *nv;
2113 
2114         /*
2115          * Without this check, we would return local property values if the
2116          * caller has not already received properties on or after
2117          * SPA_VERSION_RECVD_PROPS.
2118          */
2119         if (!dsl_prop_get_hasrecvd(zc->zc_name))
2120                 return (SET_ERROR(ENOTSUP));
2121 
2122         if (zc->zc_nvlist_dst != 0 &&
2123             (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2124                 error = put_nvlist(zc, nv);
2125                 nvlist_free(nv);
2126         }
2127 
2128         return (error);
2129 }
2130 
2131 static int
2132 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2133 {
2134         uint64_t value;
2135         int error;
2136 
2137         /*
2138          * zfs_get_zplprop() will either find a value or give us
2139          * the default value (if there is one).
2140          */
2141         if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2142                 return (error);
2143         VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2144         return (0);
2145 }
2146 
2147 /*
2148  * inputs:
2149  * zc_name              name of filesystem
2150  * zc_nvlist_dst_size   size of buffer for zpl property nvlist
2151  *
2152  * outputs:
2153  * zc_nvlist_dst        zpl property nvlist
2154  * zc_nvlist_dst_size   size of zpl property nvlist
2155  */
2156 static int
2157 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2158 {
2159         objset_t *os;
2160         int err;
2161 
2162         /* XXX reading without owning */
2163         if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2164                 return (err);
2165 
2166         dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2167 
2168         /*
2169          * NB: nvl_add_zplprop() will read the objset contents,
2170          * which we aren't supposed to do with a DS_MODE_USER
2171          * hold, because it could be inconsistent.
2172          */
2173         if (zc->zc_nvlist_dst != NULL &&
2174             !zc->zc_objset_stats.dds_inconsistent &&
2175             dmu_objset_type(os) == DMU_OST_ZFS) {
2176                 nvlist_t *nv;
2177 
2178                 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2179                 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2180                     (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2181                     (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2182                     (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2183                         err = put_nvlist(zc, nv);
2184                 nvlist_free(nv);
2185         } else {
2186                 err = SET_ERROR(ENOENT);
2187         }
2188         dmu_objset_rele(os, FTAG);
2189         return (err);
2190 }
2191 
2192 static boolean_t
2193 dataset_name_hidden(const char *name)
2194 {
2195         /*
2196          * Skip over datasets that are not visible in this zone,
2197          * internal datasets (which have a $ in their name), and
2198          * temporary datasets (which have a % in their name).
2199          */
2200         if (strchr(name, '$') != NULL)
2201                 return (B_TRUE);
2202         if (strchr(name, '%') != NULL)
2203                 return (B_TRUE);
2204         if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
2205                 return (B_TRUE);
2206         return (B_FALSE);
2207 }
2208 
2209 /*
2210  * inputs:
2211  * zc_name              name of filesystem
2212  * zc_cookie            zap cursor
2213  * zc_nvlist_dst_size   size of buffer for property nvlist
2214  *
2215  * outputs:
2216  * zc_name              name of next filesystem
2217  * zc_cookie            zap cursor
2218  * zc_objset_stats      stats
2219  * zc_nvlist_dst        property nvlist
2220  * zc_nvlist_dst_size   size of property nvlist
2221  */
2222 static int
2223 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2224 {
2225         objset_t *os;
2226         int error;
2227         char *p;
2228         size_t orig_len = strlen(zc->zc_name);
2229 
2230 top:
2231         if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2232                 if (error == ENOENT)
2233                         error = SET_ERROR(ESRCH);
2234                 return (error);
2235         }
2236 
2237         p = strrchr(zc->zc_name, '/');
2238         if (p == NULL || p[1] != '\0')
2239                 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2240         p = zc->zc_name + strlen(zc->zc_name);
2241 
2242         do {
2243                 error = dmu_dir_list_next(os,
2244                     sizeof (zc->zc_name) - (p - zc->zc_name), p,
2245                     NULL, &zc->zc_cookie);
2246                 if (error == ENOENT)
2247                         error = SET_ERROR(ESRCH);
2248         } while (error == 0 && dataset_name_hidden(zc->zc_name));
2249         dmu_objset_rele(os, FTAG);
2250 
2251         /*
2252          * If it's an internal dataset (ie. with a '$' in its name),
2253          * don't try to get stats for it, otherwise we'll return ENOENT.
2254          */
2255         if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2256                 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2257                 if (error == ENOENT) {
2258                         /* We lost a race with destroy, get the next one. */
2259                         zc->zc_name[orig_len] = '\0';
2260                         goto top;
2261                 }
2262         }
2263         return (error);
2264 }
2265 
2266 /*
2267  * inputs:
2268  * zc_name              name of filesystem
2269  * zc_cookie            zap cursor
2270  * zc_nvlist_dst_size   size of buffer for property nvlist
2271  *
2272  * outputs:
2273  * zc_name              name of next snapshot
2274  * zc_objset_stats      stats
2275  * zc_nvlist_dst        property nvlist
2276  * zc_nvlist_dst_size   size of property nvlist
2277  */
2278 static int
2279 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2280 {
2281         objset_t *os;
2282         int error;
2283 
2284         error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2285         if (error != 0) {
2286                 return (error == ENOENT ? ESRCH : error);
2287         }
2288 
2289         /*
2290          * A dataset name of maximum length cannot have any snapshots,
2291          * so exit immediately.
2292          */
2293         if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2294                 dmu_objset_rele(os, FTAG);
2295                 return (SET_ERROR(ESRCH));
2296         }
2297 
2298         error = dmu_snapshot_list_next(os,
2299             sizeof (zc->zc_name) - strlen(zc->zc_name),
2300             zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2301             NULL);
2302 
2303         if (error == 0) {
2304                 dsl_dataset_t *ds;
2305                 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2306 
2307                 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2308                 if (error == 0) {
2309                         objset_t *ossnap;
2310 
2311                         error = dmu_objset_from_ds(ds, &ossnap);
2312                         if (error == 0)
2313                                 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2314                         dsl_dataset_rele(ds, FTAG);
2315                 }
2316         } else if (error == ENOENT) {
2317                 error = SET_ERROR(ESRCH);
2318         }
2319 
2320         dmu_objset_rele(os, FTAG);
2321         /* if we failed, undo the @ that we tacked on to zc_name */
2322         if (error != 0)
2323                 *strchr(zc->zc_name, '@') = '\0';
2324         return (error);
2325 }
2326 
2327 static int
2328 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2329 {
2330         const char *propname = nvpair_name(pair);
2331         uint64_t *valary;
2332         unsigned int vallen;
2333         const char *domain;
2334         char *dash;
2335         zfs_userquota_prop_t type;
2336         uint64_t rid;
2337         uint64_t quota;
2338         zfsvfs_t *zfsvfs;
2339         int err;
2340 
2341         if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2342                 nvlist_t *attrs;
2343                 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2344                 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2345                     &pair) != 0)
2346                         return (SET_ERROR(EINVAL));
2347         }
2348 
2349         /*
2350          * A correctly constructed propname is encoded as
2351          * userquota@<rid>-<domain>.
2352          */
2353         if ((dash = strchr(propname, '-')) == NULL ||
2354             nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2355             vallen != 3)
2356                 return (SET_ERROR(EINVAL));
2357 
2358         domain = dash + 1;
2359         type = valary[0];
2360         rid = valary[1];
2361         quota = valary[2];
2362 
2363         err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2364         if (err == 0) {
2365                 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2366                 zfsvfs_rele(zfsvfs, FTAG);
2367         }
2368 
2369         return (err);
2370 }
2371 
2372 /*
2373  * If the named property is one that has a special function to set its value,
2374  * return 0 on success and a positive error code on failure; otherwise if it is
2375  * not one of the special properties handled by this function, return -1.
2376  *
2377  * XXX: It would be better for callers of the property interface if we handled
2378  * these special cases in dsl_prop.c (in the dsl layer).
2379  */
2380 static int
2381 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2382     nvpair_t *pair)
2383 {
2384         const char *propname = nvpair_name(pair);
2385         zfs_prop_t prop = zfs_name_to_prop(propname);
2386         uint64_t intval;
2387         int err = -1;
2388 
2389         if (prop == ZPROP_INVAL) {
2390                 if (zfs_prop_userquota(propname))
2391                         return (zfs_prop_set_userquota(dsname, pair));
2392                 return (-1);
2393         }
2394 
2395         if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2396                 nvlist_t *attrs;
2397                 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2398                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2399                     &pair) == 0);
2400         }
2401 
2402         if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2403                 return (-1);
2404 
2405         VERIFY(0 == nvpair_value_uint64(pair, &intval));
2406 
2407         switch (prop) {
2408         case ZFS_PROP_QUOTA:
2409                 err = dsl_dir_set_quota(dsname, source, intval);
2410                 break;
2411         case ZFS_PROP_REFQUOTA:
2412                 err = dsl_dataset_set_refquota(dsname, source, intval);
2413                 break;
2414         case ZFS_PROP_FILESYSTEM_LIMIT:
2415         case ZFS_PROP_SNAPSHOT_LIMIT:
2416                 if (intval == UINT64_MAX) {
2417                         /* clearing the limit, just do it */
2418                         err = 0;
2419                 } else {
2420                         err = dsl_dir_activate_fs_ss_limit(dsname);
2421                 }
2422                 /*
2423                  * Set err to -1 to force the zfs_set_prop_nvlist code down the
2424                  * default path to set the value in the nvlist.
2425                  */
2426                 if (err == 0)
2427                         err = -1;
2428                 break;
2429         case ZFS_PROP_RESERVATION:
2430                 err = dsl_dir_set_reservation(dsname, source, intval);
2431                 break;
2432         case ZFS_PROP_REFRESERVATION:
2433                 err = dsl_dataset_set_refreservation(dsname, source, intval);
2434                 break;
2435         case ZFS_PROP_VOLSIZE:
2436                 err = zvol_set_volsize(dsname, intval);
2437                 break;
2438         case ZFS_PROP_VERSION:
2439         {
2440                 zfsvfs_t *zfsvfs;
2441 
2442                 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2443                         break;
2444 
2445                 err = zfs_set_version(zfsvfs, intval);
2446                 zfsvfs_rele(zfsvfs, FTAG);
2447 
2448                 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2449                         zfs_cmd_t *zc;
2450 
2451                         zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2452                         (void) strcpy(zc->zc_name, dsname);
2453                         (void) zfs_ioc_userspace_upgrade(zc);
2454                         kmem_free(zc, sizeof (zfs_cmd_t));
2455                 }
2456                 break;
2457         }
2458         default:
2459                 err = -1;
2460         }
2461 
2462         return (err);
2463 }
2464 
2465 /*
2466  * This function is best effort. If it fails to set any of the given properties,
2467  * it continues to set as many as it can and returns the last error
2468  * encountered. If the caller provides a non-NULL errlist, it will be filled in
2469  * with the list of names of all the properties that failed along with the
2470  * corresponding error numbers.
2471  *
2472  * If every property is set successfully, zero is returned and errlist is not
2473  * modified.
2474  */
2475 int
2476 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2477     nvlist_t *errlist)
2478 {
2479         nvpair_t *pair;
2480         nvpair_t *propval;
2481         int rv = 0;
2482         uint64_t intval;
2483         char *strval;
2484         nvlist_t *genericnvl = fnvlist_alloc();
2485         nvlist_t *retrynvl = fnvlist_alloc();
2486 
2487 retry:
2488         pair = NULL;
2489         while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2490                 const char *propname = nvpair_name(pair);
2491                 zfs_prop_t prop = zfs_name_to_prop(propname);
2492                 int err = 0;
2493 
2494                 /* decode the property value */
2495                 propval = pair;
2496                 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2497                         nvlist_t *attrs;
2498                         attrs = fnvpair_value_nvlist(pair);
2499                         if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2500                             &propval) != 0)
2501                                 err = SET_ERROR(EINVAL);
2502                 }
2503 
2504                 /* Validate value type */
2505                 if (err == 0 && prop == ZPROP_INVAL) {
2506                         if (zfs_prop_user(propname)) {
2507                                 if (nvpair_type(propval) != DATA_TYPE_STRING)
2508                                         err = SET_ERROR(EINVAL);
2509                         } else if (zfs_prop_userquota(propname)) {
2510                                 if (nvpair_type(propval) !=
2511                                     DATA_TYPE_UINT64_ARRAY)
2512                                         err = SET_ERROR(EINVAL);
2513                         } else {
2514                                 err = SET_ERROR(EINVAL);
2515                         }
2516                 } else if (err == 0) {
2517                         if (nvpair_type(propval) == DATA_TYPE_STRING) {
2518                                 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2519                                         err = SET_ERROR(EINVAL);
2520                         } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2521                                 const char *unused;
2522 
2523                                 intval = fnvpair_value_uint64(propval);
2524 
2525                                 switch (zfs_prop_get_type(prop)) {
2526                                 case PROP_TYPE_NUMBER:
2527                                         break;
2528                                 case PROP_TYPE_STRING:
2529                                         err = SET_ERROR(EINVAL);
2530                                         break;
2531                                 case PROP_TYPE_INDEX:
2532                                         if (zfs_prop_index_to_string(prop,
2533                                             intval, &unused) != 0)
2534                                                 err = SET_ERROR(EINVAL);
2535                                         break;
2536                                 default:
2537                                         cmn_err(CE_PANIC,
2538                                             "unknown property type");
2539                                 }
2540                         } else {
2541                                 err = SET_ERROR(EINVAL);
2542                         }
2543                 }
2544 
2545                 /* Validate permissions */
2546                 if (err == 0)
2547                         err = zfs_check_settable(dsname, pair, CRED());
2548 
2549                 if (err == 0) {
2550                         err = zfs_prop_set_special(dsname, source, pair);
2551                         if (err == -1) {
2552                                 /*
2553                                  * For better performance we build up a list of
2554                                  * properties to set in a single transaction.
2555                                  */
2556                                 err = nvlist_add_nvpair(genericnvl, pair);
2557                         } else if (err != 0 && nvl != retrynvl) {
2558                                 /*
2559                                  * This may be a spurious error caused by
2560                                  * receiving quota and reservation out of order.
2561                                  * Try again in a second pass.
2562                                  */
2563                                 err = nvlist_add_nvpair(retrynvl, pair);
2564                         }
2565                 }
2566 
2567                 if (err != 0) {
2568                         if (errlist != NULL)
2569                                 fnvlist_add_int32(errlist, propname, err);
2570                         rv = err;
2571                 }
2572         }
2573 
2574         if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2575                 nvl = retrynvl;
2576                 goto retry;
2577         }
2578 
2579         if (!nvlist_empty(genericnvl) &&
2580             dsl_props_set(dsname, source, genericnvl) != 0) {
2581                 /*
2582                  * If this fails, we still want to set as many properties as we
2583                  * can, so try setting them individually.
2584                  */
2585                 pair = NULL;
2586                 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2587                         const char *propname = nvpair_name(pair);
2588                         int err = 0;
2589 
2590                         propval = pair;
2591                         if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2592                                 nvlist_t *attrs;
2593                                 attrs = fnvpair_value_nvlist(pair);
2594                                 propval = fnvlist_lookup_nvpair(attrs,
2595                                     ZPROP_VALUE);
2596                         }
2597 
2598                         if (nvpair_type(propval) == DATA_TYPE_STRING) {
2599                                 strval = fnvpair_value_string(propval);
2600                                 err = dsl_prop_set_string(dsname, propname,
2601                                     source, strval);
2602                         } else {
2603                                 intval = fnvpair_value_uint64(propval);
2604                                 err = dsl_prop_set_int(dsname, propname, source,
2605                                     intval);
2606                         }
2607 
2608                         if (err != 0) {
2609                                 if (errlist != NULL) {
2610                                         fnvlist_add_int32(errlist, propname,
2611                                             err);
2612                                 }
2613                                 rv = err;
2614                         }
2615                 }
2616         }
2617         nvlist_free(genericnvl);
2618         nvlist_free(retrynvl);
2619 
2620         return (rv);
2621 }
2622 
2623 /*
2624  * Check that all the properties are valid user properties.
2625  */
2626 static int
2627 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2628 {
2629         nvpair_t *pair = NULL;
2630         int error = 0;
2631 
2632         while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2633                 const char *propname = nvpair_name(pair);
2634 
2635                 if (!zfs_prop_user(propname) ||
2636                     nvpair_type(pair) != DATA_TYPE_STRING)
2637                         return (SET_ERROR(EINVAL));
2638 
2639                 if (error = zfs_secpolicy_write_perms(fsname,
2640                     ZFS_DELEG_PERM_USERPROP, CRED()))
2641                         return (error);
2642 
2643                 if (strlen(propname) >= ZAP_MAXNAMELEN)
2644                         return (SET_ERROR(ENAMETOOLONG));
2645 
2646                 if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
2647                         return (E2BIG);
2648         }
2649         return (0);
2650 }
2651 
2652 static void
2653 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2654 {
2655         nvpair_t *pair;
2656 
2657         VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2658 
2659         pair = NULL;
2660         while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2661                 if (nvlist_exists(skipped, nvpair_name(pair)))
2662                         continue;
2663 
2664                 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2665         }
2666 }
2667 
2668 static int
2669 clear_received_props(const char *dsname, nvlist_t *props,
2670     nvlist_t *skipped)
2671 {
2672         int err = 0;
2673         nvlist_t *cleared_props = NULL;
2674         props_skip(props, skipped, &cleared_props);
2675         if (!nvlist_empty(cleared_props)) {
2676                 /*
2677                  * Acts on local properties until the dataset has received
2678                  * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2679                  */
2680                 zprop_source_t flags = (ZPROP_SRC_NONE |
2681                     (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2682                 err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2683         }
2684         nvlist_free(cleared_props);
2685         return (err);
2686 }
2687 
2688 /*
2689  * inputs:
2690  * zc_name              name of filesystem
2691  * zc_value             name of property to set
2692  * zc_nvlist_src{_size} nvlist of properties to apply
2693  * zc_cookie            received properties flag
2694  *
2695  * outputs:
2696  * zc_nvlist_dst{_size} error for each unapplied received property
2697  */
2698 static int
2699 zfs_ioc_set_prop(zfs_cmd_t *zc)
2700 {
2701         nvlist_t *nvl;
2702         boolean_t received = zc->zc_cookie;
2703         zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2704             ZPROP_SRC_LOCAL);
2705         nvlist_t *errors;
2706         int error;
2707 
2708         if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2709             zc->zc_iflags, &nvl)) != 0)
2710                 return (error);
2711 
2712         if (received) {
2713                 nvlist_t *origprops;
2714 
2715                 if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2716                         (void) clear_received_props(zc->zc_name,
2717                             origprops, nvl);
2718                         nvlist_free(origprops);
2719                 }
2720 
2721                 error = dsl_prop_set_hasrecvd(zc->zc_name);
2722         }
2723 
2724         errors = fnvlist_alloc();
2725         if (error == 0)
2726                 error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2727 
2728         if (zc->zc_nvlist_dst != NULL && errors != NULL) {
2729                 (void) put_nvlist(zc, errors);
2730         }
2731 
2732         nvlist_free(errors);
2733         nvlist_free(nvl);
2734         return (error);
2735 }
2736 
2737 /*
2738  * inputs:
2739  * zc_name              name of filesystem
2740  * zc_value             name of property to inherit
2741  * zc_cookie            revert to received value if TRUE
2742  *
2743  * outputs:             none
2744  */
2745 static int
2746 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2747 {
2748         const char *propname = zc->zc_value;
2749         zfs_prop_t prop = zfs_name_to_prop(propname);
2750         boolean_t received = zc->zc_cookie;
2751         zprop_source_t source = (received
2752             ? ZPROP_SRC_NONE            /* revert to received value, if any */
2753             : ZPROP_SRC_INHERITED);     /* explicitly inherit */
2754 
2755         if (received) {
2756                 nvlist_t *dummy;
2757                 nvpair_t *pair;
2758                 zprop_type_t type;
2759                 int err;
2760 
2761                 /*
2762                  * zfs_prop_set_special() expects properties in the form of an
2763                  * nvpair with type info.
2764                  */
2765                 if (prop == ZPROP_INVAL) {
2766                         if (!zfs_prop_user(propname))
2767                                 return (SET_ERROR(EINVAL));
2768 
2769                         type = PROP_TYPE_STRING;
2770                 } else if (prop == ZFS_PROP_VOLSIZE ||
2771                     prop == ZFS_PROP_VERSION) {
2772                         return (SET_ERROR(EINVAL));
2773                 } else {
2774                         type = zfs_prop_get_type(prop);
2775                 }
2776 
2777                 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2778 
2779                 switch (type) {
2780                 case PROP_TYPE_STRING:
2781                         VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2782                         break;
2783                 case PROP_TYPE_NUMBER:
2784                 case PROP_TYPE_INDEX:
2785                         VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2786                         break;
2787                 default:
2788                         nvlist_free(dummy);
2789                         return (SET_ERROR(EINVAL));
2790                 }
2791 
2792                 pair = nvlist_next_nvpair(dummy, NULL);
2793                 err = zfs_prop_set_special(zc->zc_name, source, pair);
2794                 nvlist_free(dummy);
2795                 if (err != -1)
2796                         return (err); /* special property already handled */
2797         } else {
2798                 /*
2799                  * Only check this in the non-received case. We want to allow
2800                  * 'inherit -S' to revert non-inheritable properties like quota
2801                  * and reservation to the received or default values even though
2802                  * they are not considered inheritable.
2803                  */
2804                 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2805                         return (SET_ERROR(EINVAL));
2806         }
2807 
2808         /* property name has been validated by zfs_secpolicy_inherit_prop() */
2809         return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2810 }
2811 
2812 static int
2813 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2814 {
2815         nvlist_t *props;
2816         spa_t *spa;
2817         int error;
2818         nvpair_t *pair;
2819 
2820         if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2821             zc->zc_iflags, &props))
2822                 return (error);
2823 
2824         /*
2825          * If the only property is the configfile, then just do a spa_lookup()
2826          * to handle the faulted case.
2827          */
2828         pair = nvlist_next_nvpair(props, NULL);
2829         if (pair != NULL && strcmp(nvpair_name(pair),
2830             zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2831             nvlist_next_nvpair(props, pair) == NULL) {
2832                 mutex_enter(&spa_namespace_lock);
2833                 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2834                         spa_configfile_set(spa, props, B_FALSE);
2835                         spa_config_sync(spa, B_FALSE, B_TRUE);
2836                 }
2837                 mutex_exit(&spa_namespace_lock);
2838                 if (spa != NULL) {
2839                         nvlist_free(props);
2840                         return (0);
2841                 }
2842         }
2843 
2844         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2845                 nvlist_free(props);
2846                 return (error);
2847         }
2848 
2849         error = spa_prop_set(spa, props);
2850 
2851         nvlist_free(props);
2852         spa_close(spa, FTAG);
2853 
2854         return (error);
2855 }
2856 
2857 static int
2858 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2859 {
2860         spa_t *spa;
2861         int error;
2862         nvlist_t *nvp = NULL;
2863 
2864         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2865                 /*
2866                  * If the pool is faulted, there may be properties we can still
2867                  * get (such as altroot and cachefile), so attempt to get them
2868                  * anyway.
2869                  */
2870                 mutex_enter(&spa_namespace_lock);
2871                 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2872                         error = spa_prop_get(spa, &nvp);
2873                 mutex_exit(&spa_namespace_lock);
2874         } else {
2875                 error = spa_prop_get(spa, &nvp);
2876                 spa_close(spa, FTAG);
2877         }
2878 
2879         if (error == 0 && zc->zc_nvlist_dst != NULL)
2880                 error = put_nvlist(zc, nvp);
2881         else
2882                 error = SET_ERROR(EFAULT);
2883 
2884         nvlist_free(nvp);
2885         return (error);
2886 }
2887 
2888 /*
2889  * inputs:
2890  * zc_name              name of filesystem
2891  * zc_nvlist_src{_size} nvlist of delegated permissions
2892  * zc_perm_action       allow/unallow flag
2893  *
2894  * outputs:             none
2895  */
2896 static int
2897 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2898 {
2899         int error;
2900         nvlist_t *fsaclnv = NULL;
2901 
2902         if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2903             zc->zc_iflags, &fsaclnv)) != 0)
2904                 return (error);
2905 
2906         /*
2907          * Verify nvlist is constructed correctly
2908          */
2909         if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2910                 nvlist_free(fsaclnv);
2911                 return (SET_ERROR(EINVAL));
2912         }
2913 
2914         /*
2915          * If we don't have PRIV_SYS_MOUNT, then validate
2916          * that user is allowed to hand out each permission in
2917          * the nvlist(s)
2918          */
2919 
2920         error = secpolicy_zfs(CRED());
2921         if (error != 0) {
2922                 if (zc->zc_perm_action == B_FALSE) {
2923                         error = dsl_deleg_can_allow(zc->zc_name,
2924                             fsaclnv, CRED());
2925                 } else {
2926                         error = dsl_deleg_can_unallow(zc->zc_name,
2927                             fsaclnv, CRED());
2928                 }
2929         }
2930 
2931         if (error == 0)
2932                 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2933 
2934         nvlist_free(fsaclnv);
2935         return (error);
2936 }
2937 
2938 /*
2939  * inputs:
2940  * zc_name              name of filesystem
2941  *
2942  * outputs:
2943  * zc_nvlist_src{_size} nvlist of delegated permissions
2944  */
2945 static int
2946 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2947 {
2948         nvlist_t *nvp;
2949         int error;
2950 
2951         if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2952                 error = put_nvlist(zc, nvp);
2953                 nvlist_free(nvp);
2954         }
2955 
2956         return (error);
2957 }
2958 
2959 /*
2960  * Search the vfs list for a specified resource.  Returns a pointer to it
2961  * or NULL if no suitable entry is found. The caller of this routine
2962  * is responsible for releasing the returned vfs pointer.
2963  */
2964 static vfs_t *
2965 zfs_get_vfs(const char *resource)
2966 {
2967         struct vfs *vfsp;
2968         struct vfs *vfs_found = NULL;
2969 
2970         vfs_list_read_lock();
2971         vfsp = rootvfs;
2972         do {
2973                 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
2974                         VFS_HOLD(vfsp);
2975                         vfs_found = vfsp;
2976                         break;
2977                 }
2978                 vfsp = vfsp->vfs_next;
2979         } while (vfsp != rootvfs);
2980         vfs_list_unlock();
2981         return (vfs_found);
2982 }
2983 
2984 /* ARGSUSED */
2985 static void
2986 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
2987 {
2988         zfs_creat_t *zct = arg;
2989 
2990         zfs_create_fs(os, cr, zct->zct_zplprops, tx);
2991 }
2992 
2993 #define ZFS_PROP_UNDEFINED      ((uint64_t)-1)
2994 
2995 /*
2996  * inputs:
2997  * os                   parent objset pointer (NULL if root fs)
2998  * fuids_ok             fuids allowed in this version of the spa?
2999  * sa_ok                SAs allowed in this version of the spa?
3000  * createprops          list of properties requested by creator
3001  *
3002  * outputs:
3003  * zplprops     values for the zplprops we attach to the master node object
3004  * is_ci        true if requested file system will be purely case-insensitive
3005  *
3006  * Determine the settings for utf8only, normalization and
3007  * casesensitivity.  Specific values may have been requested by the
3008  * creator and/or we can inherit values from the parent dataset.  If
3009  * the file system is of too early a vintage, a creator can not
3010  * request settings for these properties, even if the requested
3011  * setting is the default value.  We don't actually want to create dsl
3012  * properties for these, so remove them from the source nvlist after
3013  * processing.
3014  */
3015 static int
3016 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3017     boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3018     nvlist_t *zplprops, boolean_t *is_ci)
3019 {
3020         uint64_t sense = ZFS_PROP_UNDEFINED;
3021         uint64_t norm = ZFS_PROP_UNDEFINED;
3022         uint64_t u8 = ZFS_PROP_UNDEFINED;
3023 
3024         ASSERT(zplprops != NULL);
3025 
3026         /*
3027          * Pull out creator prop choices, if any.
3028          */
3029         if (createprops) {
3030                 (void) nvlist_lookup_uint64(createprops,
3031                     zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3032                 (void) nvlist_lookup_uint64(createprops,
3033                     zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3034                 (void) nvlist_remove_all(createprops,
3035                     zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3036                 (void) nvlist_lookup_uint64(createprops,
3037                     zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3038                 (void) nvlist_remove_all(createprops,
3039                     zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3040                 (void) nvlist_lookup_uint64(createprops,
3041                     zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3042                 (void) nvlist_remove_all(createprops,
3043                     zfs_prop_to_name(ZFS_PROP_CASE));
3044         }
3045 
3046         /*
3047          * If the zpl version requested is whacky or the file system
3048          * or pool is version is too "young" to support normalization
3049          * and the creator tried to set a value for one of the props,
3050          * error out.
3051          */
3052         if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3053             (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3054             (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3055             (zplver < ZPL_VERSION_NORMALIZATION &&
3056             (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3057             sense != ZFS_PROP_UNDEFINED)))
3058                 return (SET_ERROR(ENOTSUP));
3059 
3060         /*
3061          * Put the version in the zplprops
3062          */
3063         VERIFY(nvlist_add_uint64(zplprops,
3064             zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3065 
3066         if (norm == ZFS_PROP_UNDEFINED)
3067                 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3068         VERIFY(nvlist_add_uint64(zplprops,
3069             zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3070 
3071         /*
3072          * If we're normalizing, names must always be valid UTF-8 strings.
3073          */
3074         if (norm)
3075                 u8 = 1;
3076         if (u8 == ZFS_PROP_UNDEFINED)
3077                 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3078         VERIFY(nvlist_add_uint64(zplprops,
3079             zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3080 
3081         if (sense == ZFS_PROP_UNDEFINED)
3082                 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3083         VERIFY(nvlist_add_uint64(zplprops,
3084             zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3085 
3086         if (is_ci)
3087                 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3088 
3089         return (0);
3090 }
3091 
3092 static int
3093 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3094     nvlist_t *zplprops, boolean_t *is_ci)
3095 {
3096         boolean_t fuids_ok, sa_ok;
3097         uint64_t zplver = ZPL_VERSION;
3098         objset_t *os = NULL;
3099         char parentname[MAXNAMELEN];
3100         char *cp;
3101         spa_t *spa;
3102         uint64_t spa_vers;
3103         int error;
3104 
3105         (void) strlcpy(parentname, dataset, sizeof (parentname));
3106         cp = strrchr(parentname, '/');
3107         ASSERT(cp != NULL);
3108         cp[0] = '\0';
3109 
3110         if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3111                 return (error);
3112 
3113         spa_vers = spa_version(spa);
3114         spa_close(spa, FTAG);
3115 
3116         zplver = zfs_zpl_version_map(spa_vers);
3117         fuids_ok = (zplver >= ZPL_VERSION_FUID);
3118         sa_ok = (zplver >= ZPL_VERSION_SA);
3119 
3120         /*
3121          * Open parent object set so we can inherit zplprop values.
3122          */
3123         if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3124                 return (error);
3125 
3126         error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3127             zplprops, is_ci);
3128         dmu_objset_rele(os, FTAG);
3129         return (error);
3130 }
3131 
3132 static int
3133 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3134     nvlist_t *zplprops, boolean_t *is_ci)
3135 {
3136         boolean_t fuids_ok;
3137         boolean_t sa_ok;
3138         uint64_t zplver = ZPL_VERSION;
3139         int error;
3140 
3141         zplver = zfs_zpl_version_map(spa_vers);
3142         fuids_ok = (zplver >= ZPL_VERSION_FUID);
3143         sa_ok = (zplver >= ZPL_VERSION_SA);
3144 
3145         error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3146             createprops, zplprops, is_ci);
3147         return (error);
3148 }
3149 
3150 /*
3151  * innvl: {
3152  *     "type" -> dmu_objset_type_t (int32)
3153  *     (optional) "props" -> { prop -> value }
3154  * }
3155  *
3156  * outnvl: propname -> error code (int32)
3157  */
3158 static int
3159 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3160 {
3161         int error = 0;
3162         zfs_creat_t zct = { 0 };
3163         nvlist_t *nvprops = NULL;
3164         void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3165         int32_t type32;
3166         dmu_objset_type_t type;
3167         boolean_t is_insensitive = B_FALSE;
3168 
3169         if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3170                 return (SET_ERROR(EINVAL));
3171         type = type32;
3172         (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3173 
3174         switch (type) {
3175         case DMU_OST_ZFS:
3176                 cbfunc = zfs_create_cb;
3177                 break;
3178 
3179         case DMU_OST_ZVOL:
3180                 cbfunc = zvol_create_cb;
3181                 break;
3182 
3183         default:
3184                 cbfunc = NULL;
3185                 break;
3186         }
3187         if (strchr(fsname, '@') ||
3188             strchr(fsname, '%'))
3189                 return (SET_ERROR(EINVAL));
3190 
3191         zct.zct_props = nvprops;
3192 
3193         if (cbfunc == NULL)
3194                 return (SET_ERROR(EINVAL));
3195 
3196         if (type == DMU_OST_ZVOL) {
3197                 uint64_t volsize, volblocksize;
3198 
3199                 if (nvprops == NULL)
3200                         return (SET_ERROR(EINVAL));
3201                 if (nvlist_lookup_uint64(nvprops,
3202                     zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3203                         return (SET_ERROR(EINVAL));
3204 
3205                 if ((error = nvlist_lookup_uint64(nvprops,
3206                     zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3207                     &volblocksize)) != 0 && error != ENOENT)
3208                         return (SET_ERROR(EINVAL));
3209 
3210                 if (error != 0)
3211                         volblocksize = zfs_prop_default_numeric(
3212                             ZFS_PROP_VOLBLOCKSIZE);
3213 
3214                 if ((error = zvol_check_volblocksize(
3215                     volblocksize)) != 0 ||
3216                     (error = zvol_check_volsize(volsize,
3217                     volblocksize)) != 0)
3218                         return (error);
3219         } else if (type == DMU_OST_ZFS) {
3220                 int error;
3221 
3222                 /*
3223                  * We have to have normalization and
3224                  * case-folding flags correct when we do the
3225                  * file system creation, so go figure them out
3226                  * now.
3227                  */
3228                 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3229                     NV_UNIQUE_NAME, KM_SLEEP) == 0);
3230                 error = zfs_fill_zplprops(fsname, nvprops,
3231                     zct.zct_zplprops, &is_insensitive);
3232                 if (error != 0) {
3233                         nvlist_free(zct.zct_zplprops);
3234                         return (error);
3235                 }
3236         }
3237 
3238         error = dmu_objset_create(fsname, type,
3239             is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3240         nvlist_free(zct.zct_zplprops);
3241 
3242         /*
3243          * It would be nice to do this atomically.
3244          */
3245         if (error == 0) {
3246                 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3247                     nvprops, outnvl);
3248                 if (error != 0)
3249                         (void) dsl_destroy_head(fsname);
3250         }
3251         return (error);
3252 }
3253 
3254 /*
3255  * innvl: {
3256  *     "origin" -> name of origin snapshot
3257  *     (optional) "props" -> { prop -> value }
3258  * }
3259  *
3260  * outnvl: propname -> error code (int32)
3261  */
3262 static int
3263 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3264 {
3265         int error = 0;
3266         nvlist_t *nvprops = NULL;
3267         char *origin_name;
3268 
3269         if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3270                 return (SET_ERROR(EINVAL));
3271         (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3272 
3273         if (strchr(fsname, '@') ||
3274             strchr(fsname, '%'))
3275                 return (SET_ERROR(EINVAL));
3276 
3277         if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3278                 return (SET_ERROR(EINVAL));
3279         error = dmu_objset_clone(fsname, origin_name);
3280         if (error != 0)
3281                 return (error);
3282 
3283         /*
3284          * It would be nice to do this atomically.
3285          */
3286         if (error == 0) {
3287                 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3288                     nvprops, outnvl);
3289                 if (error != 0)
3290                         (void) dsl_destroy_head(fsname);
3291         }
3292         return (error);
3293 }
3294 
3295 /*
3296  * innvl: {
3297  *     "snaps" -> { snapshot1, snapshot2 }
3298  *     (optional) "props" -> { prop -> value (string) }
3299  * }
3300  *
3301  * outnvl: snapshot -> error code (int32)
3302  */
3303 static int
3304 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3305 {
3306         nvlist_t *snaps;
3307         nvlist_t *props = NULL;
3308         int error, poollen;
3309         nvpair_t *pair;
3310 
3311         (void) nvlist_lookup_nvlist(innvl, "props", &props);
3312         if ((error = zfs_check_userprops(poolname, props)) != 0)
3313                 return (error);
3314 
3315         if (!nvlist_empty(props) &&
3316             zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3317                 return (SET_ERROR(ENOTSUP));
3318 
3319         if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3320                 return (SET_ERROR(EINVAL));
3321         poollen = strlen(poolname);
3322         for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3323             pair = nvlist_next_nvpair(snaps, pair)) {
3324                 const char *name = nvpair_name(pair);
3325                 const char *cp = strchr(name, '@');
3326 
3327                 /*
3328                  * The snap name must contain an @, and the part after it must
3329                  * contain only valid characters.
3330                  */
3331                 if (cp == NULL ||
3332                     zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3333                         return (SET_ERROR(EINVAL));
3334 
3335                 /*
3336                  * The snap must be in the specified pool.
3337                  */
3338                 if (strncmp(name, poolname, poollen) != 0 ||
3339                     (name[poollen] != '/' && name[poollen] != '@'))
3340                         return (SET_ERROR(EXDEV));
3341 
3342                 /* This must be the only snap of this fs. */
3343                 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3344                     pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3345                         if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3346                             == 0) {
3347                                 return (SET_ERROR(EXDEV));
3348                         }
3349                 }
3350         }
3351 
3352         error = dsl_dataset_snapshot(snaps, props, outnvl);
3353         return (error);
3354 }
3355 
3356 /*
3357  * innvl: "message" -> string
3358  */
3359 /* ARGSUSED */
3360 static int
3361 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3362 {
3363         char *message;
3364         spa_t *spa;
3365         int error;
3366         char *poolname;
3367 
3368         /*
3369          * The poolname in the ioctl is not set, we get it from the TSD,
3370          * which was set at the end of the last successful ioctl that allows
3371          * logging.  The secpolicy func already checked that it is set.
3372          * Only one log ioctl is allowed after each successful ioctl, so
3373          * we clear the TSD here.
3374          */
3375         poolname = tsd_get(zfs_allow_log_key);
3376         (void) tsd_set(zfs_allow_log_key, NULL);
3377         error = spa_open(poolname, &spa, FTAG);
3378         strfree(poolname);
3379         if (error != 0)
3380                 return (error);
3381 
3382         if (nvlist_lookup_string(innvl, "message", &message) != 0)  {
3383                 spa_close(spa, FTAG);
3384                 return (SET_ERROR(EINVAL));
3385         }
3386 
3387         if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3388                 spa_close(spa, FTAG);
3389                 return (SET_ERROR(ENOTSUP));
3390         }
3391 
3392         error = spa_history_log(spa, message);
3393         spa_close(spa, FTAG);
3394         return (error);
3395 }
3396 
3397 /*
3398  * The dp_config_rwlock must not be held when calling this, because the
3399  * unmount may need to write out data.
3400  *
3401  * This function is best-effort.  Callers must deal gracefully if it
3402  * remains mounted (or is remounted after this call).
3403  *
3404  * Returns 0 if the argument is not a snapshot, or it is not currently a
3405  * filesystem, or we were able to unmount it.  Returns error code otherwise.
3406  */
3407 int
3408 zfs_unmount_snap(const char *snapname)
3409 {
3410         vfs_t *vfsp;
3411         zfsvfs_t *zfsvfs;
3412         int err;
3413 
3414         if (strchr(snapname, '@') == NULL)
3415                 return (0);
3416 
3417         vfsp = zfs_get_vfs(snapname);
3418         if (vfsp == NULL)
3419                 return (0);
3420 
3421         zfsvfs = vfsp->vfs_data;
3422         ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os)));
3423 
3424         err = vn_vfswlock(vfsp->vfs_vnodecovered);
3425         VFS_RELE(vfsp);
3426         if (err != 0)
3427                 return (SET_ERROR(err));
3428 
3429         /*
3430          * Always force the unmount for snapshots.
3431          */
3432         (void) dounmount(vfsp, MS_FORCE, kcred);
3433         return (0);
3434 }
3435 
3436 /* ARGSUSED */
3437 static int
3438 zfs_unmount_snap_cb(const char *snapname, void *arg)
3439 {
3440         return (zfs_unmount_snap(snapname));
3441 }
3442 
3443 /*
3444  * When a clone is destroyed, its origin may also need to be destroyed,
3445  * in which case it must be unmounted.  This routine will do that unmount
3446  * if necessary.
3447  */
3448 void
3449 zfs_destroy_unmount_origin(const char *fsname)
3450 {
3451         int error;
3452         objset_t *os;
3453         dsl_dataset_t *ds;
3454 
3455         error = dmu_objset_hold(fsname, FTAG, &os);
3456         if (error != 0)
3457                 return;
3458         ds = dmu_objset_ds(os);
3459         if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3460                 char originname[MAXNAMELEN];
3461                 dsl_dataset_name(ds->ds_prev, originname);
3462                 dmu_objset_rele(os, FTAG);
3463                 (void) zfs_unmount_snap(originname);
3464         } else {
3465                 dmu_objset_rele(os, FTAG);
3466         }
3467 }
3468 
3469 /*
3470  * innvl: {
3471  *     "snaps" -> { snapshot1, snapshot2 }
3472  *     (optional boolean) "defer"
3473  * }
3474  *
3475  * outnvl: snapshot -> error code (int32)
3476  *
3477  */
3478 /* ARGSUSED */
3479 static int
3480 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3481 {
3482         nvlist_t *snaps;
3483         nvpair_t *pair;
3484         boolean_t defer;
3485 
3486         if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3487                 return (SET_ERROR(EINVAL));
3488         defer = nvlist_exists(innvl, "defer");
3489 
3490         for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3491             pair = nvlist_next_nvpair(snaps, pair)) {
3492                 (void) zfs_unmount_snap(nvpair_name(pair));
3493         }
3494 
3495         return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3496 }
3497 
3498 /*
3499  * Create bookmarks.  Bookmark names are of the form <fs>#<bmark>.
3500  * All bookmarks must be in the same pool.
3501  *
3502  * innvl: {
3503  *     bookmark1 -> snapshot1, bookmark2 -> snapshot2
3504  * }
3505  *
3506  * outnvl: bookmark -> error code (int32)
3507  *
3508  */
3509 /* ARGSUSED */
3510 static int
3511 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3512 {
3513         for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3514             pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3515                 char *snap_name;
3516 
3517                 /*
3518                  * Verify the snapshot argument.
3519                  */
3520                 if (nvpair_value_string(pair, &snap_name) != 0)
3521                         return (SET_ERROR(EINVAL));
3522 
3523 
3524                 /* Verify that the keys (bookmarks) are unique */
3525                 for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair);
3526                     pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) {
3527                         if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0)
3528                                 return (SET_ERROR(EINVAL));
3529                 }
3530         }
3531 
3532         return (dsl_bookmark_create(innvl, outnvl));
3533 }
3534 
3535 /*
3536  * innvl: {
3537  *     property 1, property 2, ...
3538  * }
3539  *
3540  * outnvl: {
3541  *     bookmark name 1 -> { property 1, property 2, ... },
3542  *     bookmark name 2 -> { property 1, property 2, ... }
3543  * }
3544  *
3545  */
3546 static int
3547 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3548 {
3549         return (dsl_get_bookmarks(fsname, innvl, outnvl));
3550 }
3551 
3552 /*
3553  * innvl: {
3554  *     bookmark name 1, bookmark name 2
3555  * }
3556  *
3557  * outnvl: bookmark -> error code (int32)
3558  *
3559  */
3560 static int
3561 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
3562     nvlist_t *outnvl)
3563 {
3564         int error, poollen;
3565 
3566         poollen = strlen(poolname);
3567         for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3568             pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3569                 const char *name = nvpair_name(pair);
3570                 const char *cp = strchr(name, '#');
3571 
3572                 /*
3573                  * The bookmark name must contain an #, and the part after it
3574                  * must contain only valid characters.
3575                  */
3576                 if (cp == NULL ||
3577                     zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3578                         return (SET_ERROR(EINVAL));
3579 
3580                 /*
3581                  * The bookmark must be in the specified pool.
3582                  */
3583                 if (strncmp(name, poolname, poollen) != 0 ||
3584                     (name[poollen] != '/' && name[poollen] != '#'))
3585                         return (SET_ERROR(EXDEV));
3586         }
3587 
3588         error = dsl_bookmark_destroy(innvl, outnvl);
3589         return (error);
3590 }
3591 
3592 /*
3593  * inputs:
3594  * zc_name              name of dataset to destroy
3595  * zc_objset_type       type of objset
3596  * zc_defer_destroy     mark for deferred destroy
3597  *
3598  * outputs:             none
3599  */
3600 static int
3601 zfs_ioc_destroy(zfs_cmd_t *zc)
3602 {
3603         int err;
3604 
3605         if (zc->zc_objset_type == DMU_OST_ZFS) {
3606                 err = zfs_unmount_snap(zc->zc_name);
3607                 if (err != 0)
3608                         return (err);
3609         }
3610 
3611         if (strchr(zc->zc_name, '@'))
3612                 err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3613         else
3614                 err = dsl_destroy_head(zc->zc_name);
3615         if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3616                 (void) zvol_remove_minor(zc->zc_name);
3617         return (err);
3618 }
3619 
3620 /*
3621  * fsname is name of dataset to rollback (to most recent snapshot)
3622  *
3623  * innvl is not used.
3624  *
3625  * outnvl: "target" -> name of most recent snapshot
3626  * }
3627  */
3628 /* ARGSUSED */
3629 static int
3630 zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl)
3631 {
3632         zfsvfs_t *zfsvfs;
3633         int error;
3634 
3635         if (getzfsvfs(fsname, &zfsvfs) == 0) {
3636                 error = zfs_suspend_fs(zfsvfs);
3637                 if (error == 0) {
3638                         int resume_err;
3639 
3640                         error = dsl_dataset_rollback(fsname, zfsvfs, outnvl);
3641                         resume_err = zfs_resume_fs(zfsvfs, fsname);
3642                         error = error ? error : resume_err;
3643                 }
3644                 VFS_RELE(zfsvfs->z_vfs);
3645         } else {
3646                 error = dsl_dataset_rollback(fsname, NULL, outnvl);
3647         }
3648         return (error);
3649 }
3650 
3651 static int
3652 recursive_unmount(const char *fsname, void *arg)
3653 {
3654         const char *snapname = arg;
3655         char fullname[MAXNAMELEN];
3656 
3657         (void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname);
3658         return (zfs_unmount_snap(fullname));
3659 }
3660 
3661 /*
3662  * inputs:
3663  * zc_name      old name of dataset
3664  * zc_value     new name of dataset
3665  * zc_cookie    recursive flag (only valid for snapshots)
3666  *
3667  * outputs:     none
3668  */
3669 static int
3670 zfs_ioc_rename(zfs_cmd_t *zc)
3671 {
3672         boolean_t recursive = zc->zc_cookie & 1;
3673         char *at;
3674 
3675         zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3676         if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3677             strchr(zc->zc_value, '%'))
3678                 return (SET_ERROR(EINVAL));
3679 
3680         at = strchr(zc->zc_name, '@');
3681         if (at != NULL) {
3682                 /* snaps must be in same fs */
3683                 int error;
3684 
3685                 if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3686                         return (SET_ERROR(EXDEV));
3687                 *at = '\0';
3688                 if (zc->zc_objset_type == DMU_OST_ZFS) {
3689                         error = dmu_objset_find(zc->zc_name,
3690                             recursive_unmount, at + 1,
3691                             recursive ? DS_FIND_CHILDREN : 0);
3692                         if (error != 0) {
3693                                 *at = '@';
3694                                 return (error);
3695                         }
3696                 }
3697                 error = dsl_dataset_rename_snapshot(zc->zc_name,
3698                     at + 1, strchr(zc->zc_value, '@') + 1, recursive);
3699                 *at = '@';
3700 
3701                 return (error);
3702         } else {
3703                 if (zc->zc_objset_type == DMU_OST_ZVOL)
3704                         (void) zvol_remove_minor(zc->zc_name);
3705                 return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3706         }
3707 }
3708 
3709 static int
3710 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3711 {
3712         const char *propname = nvpair_name(pair);
3713         boolean_t issnap = (strchr(dsname, '@') != NULL);
3714         zfs_prop_t prop = zfs_name_to_prop(propname);
3715         uint64_t intval;
3716         int err;
3717 
3718         if (prop == ZPROP_INVAL) {
3719                 if (zfs_prop_user(propname)) {
3720                         if (err = zfs_secpolicy_write_perms(dsname,
3721                             ZFS_DELEG_PERM_USERPROP, cr))
3722                                 return (err);
3723                         return (0);
3724                 }
3725 
3726                 if (!issnap && zfs_prop_userquota(propname)) {
3727                         const char *perm = NULL;
3728                         const char *uq_prefix =
3729                             zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3730                         const char *gq_prefix =
3731                             zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3732 
3733                         if (strncmp(propname, uq_prefix,
3734                             strlen(uq_prefix)) == 0) {
3735                                 perm = ZFS_DELEG_PERM_USERQUOTA;
3736                         } else if (strncmp(propname, gq_prefix,
3737                             strlen(gq_prefix)) == 0) {
3738                                 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3739                         } else {
3740                                 /* USERUSED and GROUPUSED are read-only */
3741                                 return (SET_ERROR(EINVAL));
3742                         }
3743 
3744                         if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3745                                 return (err);
3746                         return (0);
3747                 }
3748 
3749                 return (SET_ERROR(EINVAL));
3750         }
3751 
3752         if (issnap)
3753                 return (SET_ERROR(EINVAL));
3754 
3755         if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3756                 /*
3757                  * dsl_prop_get_all_impl() returns properties in this
3758                  * format.
3759                  */
3760                 nvlist_t *attrs;
3761                 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3762                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3763                     &pair) == 0);
3764         }
3765 
3766         /*
3767          * Check that this value is valid for this pool version
3768          */
3769         switch (prop) {
3770         case ZFS_PROP_COMPRESSION:
3771                 /*
3772                  * If the user specified gzip compression, make sure
3773                  * the SPA supports it. We ignore any errors here since
3774                  * we'll catch them later.
3775                  */
3776                 if (nvpair_value_uint64(pair, &intval) == 0) {
3777                         if (intval >= ZIO_COMPRESS_GZIP_1 &&
3778                             intval <= ZIO_COMPRESS_GZIP_9 &&
3779                             zfs_earlier_version(dsname,
3780                             SPA_VERSION_GZIP_COMPRESSION)) {
3781                                 return (SET_ERROR(ENOTSUP));
3782                         }
3783 
3784                         if (intval == ZIO_COMPRESS_ZLE &&
3785                             zfs_earlier_version(dsname,
3786                             SPA_VERSION_ZLE_COMPRESSION))
3787                                 return (SET_ERROR(ENOTSUP));
3788 
3789                         if (intval == ZIO_COMPRESS_LZ4) {
3790                                 spa_t *spa;
3791 
3792                                 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3793                                         return (err);
3794 
3795                                 if (!spa_feature_is_enabled(spa,
3796                                     SPA_FEATURE_LZ4_COMPRESS)) {
3797                                         spa_close(spa, FTAG);
3798                                         return (SET_ERROR(ENOTSUP));
3799                                 }
3800                                 spa_close(spa, FTAG);
3801                         }
3802 
3803                         /*
3804                          * If this is a bootable dataset then
3805                          * verify that the compression algorithm
3806                          * is supported for booting. We must return
3807                          * something other than ENOTSUP since it
3808                          * implies a downrev pool version.
3809                          */
3810                         if (zfs_is_bootfs(dsname) &&
3811                             !BOOTFS_COMPRESS_VALID(intval)) {
3812                                 return (SET_ERROR(ERANGE));
3813                         }
3814                 }
3815                 break;
3816 
3817         case ZFS_PROP_COPIES:
3818                 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3819                         return (SET_ERROR(ENOTSUP));
3820                 break;
3821 
3822         case ZFS_PROP_DEDUP:
3823                 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3824                         return (SET_ERROR(ENOTSUP));
3825                 break;
3826 
3827         case ZFS_PROP_RECORDSIZE:
3828                 /* Record sizes above 128k need the feature to be enabled */
3829                 if (nvpair_value_uint64(pair, &intval) == 0 &&
3830                     intval > SPA_OLD_MAXBLOCKSIZE) {
3831                         spa_t *spa;
3832 
3833                         /*
3834                          * If this is a bootable dataset then
3835                          * the we don't allow large (>128K) blocks,
3836                          * because GRUB doesn't support them.
3837                          */
3838                         if (zfs_is_bootfs(dsname) &&
3839                             intval > SPA_OLD_MAXBLOCKSIZE) {
3840                                 return (SET_ERROR(EDOM));
3841                         }
3842 
3843                         /*
3844                          * We don't allow setting the property above 1MB,
3845                          * unless the tunable has been changed.
3846                          */
3847                         if (intval > zfs_max_recordsize ||
3848                             intval > SPA_MAXBLOCKSIZE)
3849                                 return (SET_ERROR(EDOM));
3850 
3851                         if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3852                                 return (err);
3853 
3854                         if (!spa_feature_is_enabled(spa,
3855                             SPA_FEATURE_LARGE_BLOCKS)) {
3856                                 spa_close(spa, FTAG);
3857                                 return (SET_ERROR(ENOTSUP));
3858                         }
3859                         spa_close(spa, FTAG);
3860                 }
3861                 break;
3862 
3863         case ZFS_PROP_SHARESMB:
3864                 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3865                         return (SET_ERROR(ENOTSUP));
3866                 break;
3867 
3868         case ZFS_PROP_ACLINHERIT:
3869                 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3870                     nvpair_value_uint64(pair, &intval) == 0) {
3871                         if (intval == ZFS_ACL_PASSTHROUGH_X &&
3872                             zfs_earlier_version(dsname,
3873                             SPA_VERSION_PASSTHROUGH_X))
3874                                 return (SET_ERROR(ENOTSUP));
3875                 }
3876                 break;
3877         }
3878 
3879         return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3880 }
3881 
3882 /*
3883  * Checks for a race condition to make sure we don't increment a feature flag
3884  * multiple times.
3885  */
3886 static int
3887 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
3888 {
3889         spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3890         spa_feature_t *featurep = arg;
3891 
3892         if (!spa_feature_is_active(spa, *featurep))
3893                 return (0);
3894         else
3895                 return (SET_ERROR(EBUSY));
3896 }
3897 
3898 /*
3899  * The callback invoked on feature activation in the sync task caused by
3900  * zfs_prop_activate_feature.
3901  */
3902 static void
3903 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
3904 {
3905         spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3906         spa_feature_t *featurep = arg;
3907 
3908         spa_feature_incr(spa, *featurep, tx);
3909 }
3910 
3911 /*
3912  * Activates a feature on a pool in response to a property setting. This
3913  * creates a new sync task which modifies the pool to reflect the feature
3914  * as being active.
3915  */
3916 static int
3917 zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature)
3918 {
3919         int err;
3920 
3921         /* EBUSY here indicates that the feature is already active */
3922         err = dsl_sync_task(spa_name(spa),
3923             zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
3924             &feature, 2, ZFS_SPACE_CHECK_RESERVED);
3925 
3926         if (err != 0 && err != EBUSY)
3927                 return (err);
3928         else
3929                 return (0);
3930 }
3931 
3932 /*
3933  * Removes properties from the given props list that fail permission checks
3934  * needed to clear them and to restore them in case of a receive error. For each
3935  * property, make sure we have both set and inherit permissions.
3936  *
3937  * Returns the first error encountered if any permission checks fail. If the
3938  * caller provides a non-NULL errlist, it also gives the complete list of names
3939  * of all the properties that failed a permission check along with the
3940  * corresponding error numbers. The caller is responsible for freeing the
3941  * returned errlist.
3942  *
3943  * If every property checks out successfully, zero is returned and the list
3944  * pointed at by errlist is NULL.
3945  */
3946 static int
3947 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
3948 {
3949         zfs_cmd_t *zc;
3950         nvpair_t *pair, *next_pair;
3951         nvlist_t *errors;
3952         int err, rv = 0;
3953 
3954         if (props == NULL)
3955                 return (0);
3956 
3957         VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3958 
3959         zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
3960         (void) strcpy(zc->zc_name, dataset);
3961         pair = nvlist_next_nvpair(props, NULL);
3962         while (pair != NULL) {
3963                 next_pair = nvlist_next_nvpair(props, pair);
3964 
3965                 (void) strcpy(zc->zc_value, nvpair_name(pair));
3966                 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
3967                     (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
3968                         VERIFY(nvlist_remove_nvpair(props, pair) == 0);
3969                         VERIFY(nvlist_add_int32(errors,
3970                             zc->zc_value, err) == 0);
3971                 }
3972                 pair = next_pair;
3973         }
3974         kmem_free(zc, sizeof (zfs_cmd_t));
3975 
3976         if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
3977                 nvlist_free(errors);
3978                 errors = NULL;
3979         } else {
3980                 VERIFY(nvpair_value_int32(pair, &rv) == 0);
3981         }
3982 
3983         if (errlist == NULL)
3984                 nvlist_free(errors);
3985         else
3986                 *errlist = errors;
3987 
3988         return (rv);
3989 }
3990 
3991 static boolean_t
3992 propval_equals(nvpair_t *p1, nvpair_t *p2)
3993 {
3994         if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
3995                 /* dsl_prop_get_all_impl() format */
3996                 nvlist_t *attrs;
3997                 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
3998                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3999                     &p1) == 0);
4000         }
4001 
4002         if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
4003                 nvlist_t *attrs;
4004                 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
4005                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4006                     &p2) == 0);
4007         }
4008 
4009         if (nvpair_type(p1) != nvpair_type(p2))
4010                 return (B_FALSE);
4011 
4012         if (nvpair_type(p1) == DATA_TYPE_STRING) {
4013                 char *valstr1, *valstr2;
4014 
4015                 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
4016                 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
4017                 return (strcmp(valstr1, valstr2) == 0);
4018         } else {
4019                 uint64_t intval1, intval2;
4020 
4021                 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4022                 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4023                 return (intval1 == intval2);
4024         }
4025 }
4026 
4027 /*
4028  * Remove properties from props if they are not going to change (as determined
4029  * by comparison with origprops). Remove them from origprops as well, since we
4030  * do not need to clear or restore properties that won't change.
4031  */
4032 static void
4033 props_reduce(nvlist_t *props, nvlist_t *origprops)
4034 {
4035         nvpair_t *pair, *next_pair;
4036 
4037         if (origprops == NULL)
4038                 return; /* all props need to be received */
4039 
4040         pair = nvlist_next_nvpair(props, NULL);
4041         while (pair != NULL) {
4042                 const char *propname = nvpair_name(pair);
4043                 nvpair_t *match;
4044 
4045                 next_pair = nvlist_next_nvpair(props, pair);
4046 
4047                 if ((nvlist_lookup_nvpair(origprops, propname,
4048                     &match) != 0) || !propval_equals(pair, match))
4049                         goto next; /* need to set received value */
4050 
4051                 /* don't clear the existing received value */
4052                 (void) nvlist_remove_nvpair(origprops, match);
4053                 /* don't bother receiving the property */
4054                 (void) nvlist_remove_nvpair(props, pair);
4055 next:
4056                 pair = next_pair;
4057         }
4058 }
4059 
4060 /*
4061  * Extract properties that cannot be set PRIOR to the receipt of a dataset.
4062  * For example, refquota cannot be set until after the receipt of a dataset,
4063  * because in replication streams, an older/earlier snapshot may exceed the
4064  * refquota.  We want to receive the older/earlier snapshot, but setting
4065  * refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent
4066  * the older/earlier snapshot from being received (with EDQUOT).
4067  *
4068  * The ZFS test "zfs_receive_011_pos" demonstrates such a scenario.
4069  *
4070  * libzfs will need to be judicious handling errors encountered by props
4071  * extracted by this function.
4072  */
4073 static nvlist_t *
4074 extract_delay_props(nvlist_t *props)
4075 {
4076         nvlist_t *delayprops;
4077         nvpair_t *nvp, *tmp;
4078         static const zfs_prop_t delayable[] = { ZFS_PROP_REFQUOTA, 0 };
4079         int i;
4080 
4081         VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4082 
4083         for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL;
4084             nvp = nvlist_next_nvpair(props, nvp)) {
4085                 /*
4086                  * strcmp() is safe because zfs_prop_to_name() always returns
4087                  * a bounded string.
4088                  */
4089                 for (i = 0; delayable[i] != 0; i++) {
4090                         if (strcmp(zfs_prop_to_name(delayable[i]),
4091                             nvpair_name(nvp)) == 0) {
4092                                 break;
4093                         }
4094                 }
4095                 if (delayable[i] != 0) {
4096                         tmp = nvlist_prev_nvpair(props, nvp);
4097                         VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0);
4098                         VERIFY(nvlist_remove_nvpair(props, nvp) == 0);
4099                         nvp = tmp;
4100                 }
4101         }
4102 
4103         if (nvlist_empty(delayprops)) {
4104                 nvlist_free(delayprops);
4105                 delayprops = NULL;
4106         }
4107         return (delayprops);
4108 }
4109 
4110 #ifdef  DEBUG
4111 static boolean_t zfs_ioc_recv_inject_err;
4112 #endif
4113 
4114 /*
4115  * inputs:
4116  * zc_name              name of containing filesystem
4117  * zc_nvlist_src{_size} nvlist of properties to apply
4118  * zc_value             name of snapshot to create
4119  * zc_string            name of clone origin (if DRR_FLAG_CLONE)
4120  * zc_cookie            file descriptor to recv from
4121  * zc_begin_record      the BEGIN record of the stream (not byteswapped)
4122  * zc_guid              force flag
4123  * zc_cleanup_fd        cleanup-on-exit file descriptor
4124  * zc_action_handle     handle for this guid/ds mapping (or zero on first call)
4125  *
4126  * outputs:
4127  * zc_cookie            number of bytes read
4128  * zc_nvlist_dst{_size} error for each unapplied received property
4129  * zc_obj               zprop_errflags_t
4130  * zc_action_handle     handle for this guid/ds mapping
4131  */
4132 static int
4133 zfs_ioc_recv(zfs_cmd_t *zc)
4134 {
4135         file_t *fp;
4136         dmu_recv_cookie_t drc;
4137         boolean_t force = (boolean_t)zc->zc_guid;
4138         int fd;
4139         int error = 0;
4140         int props_error = 0;
4141         nvlist_t *errors;
4142         offset_t off;
4143         nvlist_t *props = NULL; /* sent properties */
4144         nvlist_t *origprops = NULL; /* existing properties */
4145         nvlist_t *delayprops = NULL; /* sent properties applied post-receive */
4146         char *origin = NULL;
4147         char *tosnap;
4148         char tofs[ZFS_MAXNAMELEN];
4149         boolean_t first_recvd_props = B_FALSE;
4150 
4151         if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4152             strchr(zc->zc_value, '@') == NULL ||
4153             strchr(zc->zc_value, '%'))
4154                 return (SET_ERROR(EINVAL));
4155 
4156         (void) strcpy(tofs, zc->zc_value);
4157         tosnap = strchr(tofs, '@');
4158         *tosnap++ = '\0';
4159 
4160         if (zc->zc_nvlist_src != NULL &&
4161             (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
4162             zc->zc_iflags, &props)) != 0)
4163                 return (error);
4164 
4165         fd = zc->zc_cookie;
4166         fp = getf(fd);
4167         if (fp == NULL) {
4168                 nvlist_free(props);
4169                 return (SET_ERROR(EBADF));
4170         }
4171 
4172         VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4173 
4174         if (zc->zc_string[0])
4175                 origin = zc->zc_string;
4176 
4177         error = dmu_recv_begin(tofs, tosnap,
4178             &zc->zc_begin_record, force, origin, &drc);
4179         if (error != 0)
4180                 goto out;
4181 
4182         /*
4183          * Set properties before we receive the stream so that they are applied
4184          * to the new data. Note that we must call dmu_recv_stream() if
4185          * dmu_recv_begin() succeeds.
4186          */
4187         if (props != NULL && !drc.drc_newfs) {
4188                 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
4189                     SPA_VERSION_RECVD_PROPS &&
4190                     !dsl_prop_get_hasrecvd(tofs))
4191                         first_recvd_props = B_TRUE;
4192 
4193                 /*
4194                  * If new received properties are supplied, they are to
4195                  * completely replace the existing received properties, so stash
4196                  * away the existing ones.
4197                  */
4198                 if (dsl_prop_get_received(tofs, &origprops) == 0) {
4199                         nvlist_t *errlist = NULL;
4200                         /*
4201                          * Don't bother writing a property if its value won't
4202                          * change (and avoid the unnecessary security checks).
4203                          *
4204                          * The first receive after SPA_VERSION_RECVD_PROPS is a
4205                          * special case where we blow away all local properties
4206                          * regardless.
4207                          */
4208                         if (!first_recvd_props)
4209                                 props_reduce(props, origprops);
4210                         if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
4211                                 (void) nvlist_merge(errors, errlist, 0);
4212                         nvlist_free(errlist);
4213 
4214                         if (clear_received_props(tofs, origprops,
4215                             first_recvd_props ? NULL : props) != 0)
4216                                 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4217                 } else {
4218                         zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4219                 }
4220         }
4221 
4222         if (props != NULL) {
4223                 props_error = dsl_prop_set_hasrecvd(tofs);
4224 
4225                 if (props_error == 0) {
4226                         delayprops = extract_delay_props(props);
4227                         (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4228                             props, errors);
4229                 }
4230         }
4231 
4232         off = fp->f_offset;
4233         error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
4234             &zc->zc_action_handle);
4235 
4236         if (error == 0) {
4237                 zfsvfs_t *zfsvfs = NULL;
4238 
4239                 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4240                         /* online recv */
4241                         int end_err;
4242 
4243                         error = zfs_suspend_fs(zfsvfs);
4244                         /*
4245                          * If the suspend fails, then the recv_end will
4246                          * likely also fail, and clean up after itself.
4247                          */
4248                         end_err = dmu_recv_end(&drc, zfsvfs);
4249                         if (error == 0)
4250                                 error = zfs_resume_fs(zfsvfs, tofs);
4251                         error = error ? error : end_err;
4252                         VFS_RELE(zfsvfs->z_vfs);
4253                 } else {
4254                         error = dmu_recv_end(&drc, NULL);
4255                 }
4256 
4257                 /* Set delayed properties now, after we're done receiving. */
4258                 if (delayprops != NULL && error == 0) {
4259                         (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4260                             delayprops, errors);
4261                 }
4262         }
4263 
4264         if (delayprops != NULL) {
4265                 /*
4266                  * Merge delayed props back in with initial props, in case
4267                  * we're DEBUG and zfs_ioc_recv_inject_err is set (which means
4268                  * we have to make sure clear_received_props() includes
4269                  * the delayed properties).
4270                  *
4271                  * Since zfs_ioc_recv_inject_err is only in DEBUG kernels,
4272                  * using ASSERT() will be just like a VERIFY.
4273                  */
4274                 ASSERT(nvlist_merge(props, delayprops, 0) == 0);
4275                 nvlist_free(delayprops);
4276         }
4277 
4278         /*
4279          * Now that all props, initial and delayed, are set, report the prop
4280          * errors to the caller.
4281          */
4282         if (zc->zc_nvlist_dst_size != 0 &&
4283             (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4284             put_nvlist(zc, errors) != 0)) {
4285                 /*
4286                  * Caller made zc->zc_nvlist_dst less than the minimum expected
4287                  * size or supplied an invalid address.
4288                  */
4289                 props_error = SET_ERROR(EINVAL);
4290         }
4291 
4292         zc->zc_cookie = off - fp->f_offset;
4293         if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4294                 fp->f_offset = off;
4295 
4296 #ifdef  DEBUG
4297         if (zfs_ioc_recv_inject_err) {
4298                 zfs_ioc_recv_inject_err = B_FALSE;
4299                 error = 1;
4300         }
4301 #endif
4302         /*
4303          * On error, restore the original props.
4304          */
4305         if (error != 0 && props != NULL && !drc.drc_newfs) {
4306                 if (clear_received_props(tofs, props, NULL) != 0) {
4307                         /*
4308                          * We failed to clear the received properties.
4309                          * Since we may have left a $recvd value on the
4310                          * system, we can't clear the $hasrecvd flag.
4311                          */
4312                         zc->zc_obj |= ZPROP_ERR_NORESTORE;
4313                 } else if (first_recvd_props) {
4314                         dsl_prop_unset_hasrecvd(tofs);
4315                 }
4316 
4317                 if (origprops == NULL && !drc.drc_newfs) {
4318                         /* We failed to stash the original properties. */
4319                         zc->zc_obj |= ZPROP_ERR_NORESTORE;
4320                 }
4321 
4322                 /*
4323                  * dsl_props_set() will not convert RECEIVED to LOCAL on or
4324                  * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4325                  * explictly if we're restoring local properties cleared in the
4326                  * first new-style receive.
4327                  */
4328                 if (origprops != NULL &&
4329                     zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4330                     ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4331                     origprops, NULL) != 0) {
4332                         /*
4333                          * We stashed the original properties but failed to
4334                          * restore them.
4335                          */
4336                         zc->zc_obj |= ZPROP_ERR_NORESTORE;
4337                 }
4338         }
4339 out:
4340         nvlist_free(props);
4341         nvlist_free(origprops);
4342         nvlist_free(errors);
4343         releasef(fd);
4344 
4345         if (error == 0)
4346                 error = props_error;
4347 
4348         return (error);
4349 }
4350 
4351 /*
4352  * inputs:
4353  * zc_name      name of snapshot to send
4354  * zc_cookie    file descriptor to send stream to
4355  * zc_obj       fromorigin flag (mutually exclusive with zc_fromobj)
4356  * zc_sendobj   objsetid of snapshot to send
4357  * zc_fromobj   objsetid of incremental fromsnap (may be zero)
4358  * zc_guid      if set, estimate size of stream only.  zc_cookie is ignored.
4359  *              output size in zc_objset_type.
4360  * zc_flags     lzc_send_flags
4361  *
4362  * outputs:
4363  * zc_objset_type       estimated size, if zc_guid is set
4364  */
4365 static int
4366 zfs_ioc_send(zfs_cmd_t *zc)
4367 {
4368         int error;
4369         offset_t off;
4370         boolean_t estimate = (zc->zc_guid != 0);
4371         boolean_t embedok = (zc->zc_flags & 0x1);
4372         boolean_t large_block_ok = (zc->zc_flags & 0x2);
4373 
4374         if (zc->zc_obj != 0) {
4375                 dsl_pool_t *dp;
4376                 dsl_dataset_t *tosnap;
4377 
4378                 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4379                 if (error != 0)
4380                         return (error);
4381 
4382                 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4383                 if (error != 0) {
4384                         dsl_pool_rele(dp, FTAG);
4385                         return (error);
4386                 }
4387 
4388                 if (dsl_dir_is_clone(tosnap->ds_dir))
4389                         zc->zc_fromobj =
4390                             dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj;
4391                 dsl_dataset_rele(tosnap, FTAG);
4392                 dsl_pool_rele(dp, FTAG);
4393         }
4394 
4395         if (estimate) {
4396                 dsl_pool_t *dp;
4397                 dsl_dataset_t *tosnap;
4398                 dsl_dataset_t *fromsnap = NULL;
4399 
4400                 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4401                 if (error != 0)
4402                         return (error);
4403 
4404                 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4405                 if (error != 0) {
4406                         dsl_pool_rele(dp, FTAG);
4407                         return (error);
4408                 }
4409 
4410                 if (zc->zc_fromobj != 0) {
4411                         error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4412                             FTAG, &fromsnap);
4413                         if (error != 0) {
4414                                 dsl_dataset_rele(tosnap, FTAG);
4415                                 dsl_pool_rele(dp, FTAG);
4416                                 return (error);
4417                         }
4418                 }
4419 
4420                 error = dmu_send_estimate(tosnap, fromsnap,
4421                     &zc->zc_objset_type);
4422 
4423                 if (fromsnap != NULL)
4424                         dsl_dataset_rele(fromsnap, FTAG);
4425                 dsl_dataset_rele(tosnap, FTAG);
4426                 dsl_pool_rele(dp, FTAG);
4427         } else {
4428                 file_t *fp = getf(zc->zc_cookie);
4429                 if (fp == NULL)
4430                         return (SET_ERROR(EBADF));
4431 
4432                 off = fp->f_offset;
4433                 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4434                     zc->zc_fromobj, embedok, large_block_ok,
4435                     zc->zc_cookie, fp->f_vnode, &off);
4436 
4437                 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4438                         fp->f_offset = off;
4439                 releasef(zc->zc_cookie);
4440         }
4441         return (error);
4442 }
4443 
4444 /*
4445  * inputs:
4446  * zc_name      name of snapshot on which to report progress
4447  * zc_cookie    file descriptor of send stream
4448  *
4449  * outputs:
4450  * zc_cookie    number of bytes written in send stream thus far
4451  */
4452 static int
4453 zfs_ioc_send_progress(zfs_cmd_t *zc)
4454 {
4455         dsl_pool_t *dp;
4456         dsl_dataset_t *ds;
4457         dmu_sendarg_t *dsp = NULL;
4458         int error;
4459 
4460         error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4461         if (error != 0)
4462                 return (error);
4463 
4464         error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4465         if (error != 0) {
4466                 dsl_pool_rele(dp, FTAG);
4467                 return (error);
4468         }
4469 
4470         mutex_enter(&ds->ds_sendstream_lock);
4471 
4472         /*
4473          * Iterate over all the send streams currently active on this dataset.
4474          * If there's one which matches the specified file descriptor _and_ the
4475          * stream was started by the current process, return the progress of
4476          * that stream.
4477          */
4478         for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4479             dsp = list_next(&ds->ds_sendstreams, dsp)) {
4480                 if (dsp->dsa_outfd == zc->zc_cookie &&
4481                     dsp->dsa_proc == curproc)
4482                         break;
4483         }
4484 
4485         if (dsp != NULL)
4486                 zc->zc_cookie = *(dsp->dsa_off);
4487         else
4488                 error = SET_ERROR(ENOENT);
4489 
4490         mutex_exit(&ds->ds_sendstream_lock);
4491         dsl_dataset_rele(ds, FTAG);
4492         dsl_pool_rele(dp, FTAG);
4493         return (error);
4494 }
4495 
4496 static int
4497 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4498 {
4499         int id, error;
4500 
4501         error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4502             &zc->zc_inject_record);
4503 
4504         if (error == 0)
4505                 zc->zc_guid = (uint64_t)id;
4506 
4507         return (error);
4508 }
4509 
4510 static int
4511 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4512 {
4513         return (zio_clear_fault((int)zc->zc_guid));
4514 }
4515 
4516 static int
4517 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4518 {
4519         int id = (int)zc->zc_guid;
4520         int error;
4521 
4522         error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4523             &zc->zc_inject_record);
4524 
4525         zc->zc_guid = id;
4526 
4527         return (error);
4528 }
4529 
4530 static int
4531 zfs_ioc_error_log(zfs_cmd_t *zc)
4532 {
4533         spa_t *spa;
4534         int error;
4535         size_t count = (size_t)zc->zc_nvlist_dst_size;
4536 
4537         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4538                 return (error);
4539 
4540         error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4541             &count);
4542         if (error == 0)
4543                 zc->zc_nvlist_dst_size = count;
4544         else
4545                 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4546 
4547         spa_close(spa, FTAG);
4548 
4549         return (error);
4550 }
4551 
4552 static int
4553 zfs_ioc_clear(zfs_cmd_t *zc)
4554 {
4555         spa_t *spa;
4556         vdev_t *vd;
4557         int error;
4558 
4559         /*
4560          * On zpool clear we also fix up missing slogs
4561          */
4562         mutex_enter(&spa_namespace_lock);
4563         spa = spa_lookup(zc->zc_name);
4564         if (spa == NULL) {
4565                 mutex_exit(&spa_namespace_lock);
4566                 return (SET_ERROR(EIO));
4567         }
4568         if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4569                 /* we need to let spa_open/spa_load clear the chains */
4570                 spa_set_log_state(spa, SPA_LOG_CLEAR);
4571         }
4572         spa->spa_last_open_failed = 0;
4573         mutex_exit(&spa_namespace_lock);
4574 
4575         if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4576                 error = spa_open(zc->zc_name, &spa, FTAG);
4577         } else {
4578                 nvlist_t *policy;
4579                 nvlist_t *config = NULL;
4580 
4581                 if (zc->zc_nvlist_src == NULL)
4582                         return (SET_ERROR(EINVAL));
4583 
4584                 if ((error = get_nvlist(zc->zc_nvlist_src,
4585                     zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4586                         error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4587                             policy, &config);
4588                         if (config != NULL) {
4589                                 int err;
4590 
4591                                 if ((err = put_nvlist(zc, config)) != 0)
4592                                         error = err;
4593                                 nvlist_free(config);
4594                         }
4595                         nvlist_free(policy);
4596                 }
4597         }
4598 
4599         if (error != 0)
4600                 return (error);
4601 
4602         spa_vdev_state_enter(spa, SCL_NONE);
4603 
4604         if (zc->zc_guid == 0) {
4605                 vd = NULL;
4606         } else {
4607                 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4608                 if (vd == NULL) {
4609                         (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4610                         spa_close(spa, FTAG);
4611                         return (SET_ERROR(ENODEV));
4612                 }
4613         }
4614 
4615         vdev_clear(spa, vd);
4616 
4617         (void) spa_vdev_state_exit(spa, NULL, 0);
4618 
4619         /*
4620          * Resume any suspended I/Os.
4621          */
4622         if (zio_resume(spa) != 0)
4623                 error = SET_ERROR(EIO);
4624 
4625         spa_close(spa, FTAG);
4626 
4627         return (error);
4628 }
4629 
4630 static int
4631 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4632 {
4633         spa_t *spa;
4634         int error;
4635 
4636         error = spa_open(zc->zc_name, &spa, FTAG);
4637         if (error != 0)
4638                 return (error);
4639 
4640         spa_vdev_state_enter(spa, SCL_NONE);
4641 
4642         /*
4643          * If a resilver is already in progress then set the
4644          * spa_scrub_reopen flag to B_TRUE so that we don't restart
4645          * the scan as a side effect of the reopen. Otherwise, let
4646          * vdev_open() decided if a resilver is required.
4647          */
4648         spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4649         vdev_reopen(spa->spa_root_vdev);
4650         spa->spa_scrub_reopen = B_FALSE;
4651 
4652         (void) spa_vdev_state_exit(spa, NULL, 0);
4653         spa_close(spa, FTAG);
4654         return (0);
4655 }
4656 /*
4657  * inputs:
4658  * zc_name      name of filesystem
4659  * zc_value     name of origin snapshot
4660  *
4661  * outputs:
4662  * zc_string    name of conflicting snapshot, if there is one
4663  */
4664 static int
4665 zfs_ioc_promote(zfs_cmd_t *zc)
4666 {
4667         char *cp;
4668 
4669         /*
4670          * We don't need to unmount *all* the origin fs's snapshots, but
4671          * it's easier.
4672          */
4673         cp = strchr(zc->zc_value, '@');
4674         if (cp)
4675                 *cp = '\0';
4676         (void) dmu_objset_find(zc->zc_value,
4677             zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
4678         return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4679 }
4680 
4681 /*
4682  * Retrieve a single {user|group}{used|quota}@... property.
4683  *
4684  * inputs:
4685  * zc_name      name of filesystem
4686  * zc_objset_type zfs_userquota_prop_t
4687  * zc_value     domain name (eg. "S-1-234-567-89")
4688  * zc_guid      RID/UID/GID
4689  *
4690  * outputs:
4691  * zc_cookie    property value
4692  */
4693 static int
4694 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4695 {
4696         zfsvfs_t *zfsvfs;
4697         int error;
4698 
4699         if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4700                 return (SET_ERROR(EINVAL));
4701 
4702         error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4703         if (error != 0)
4704                 return (error);
4705 
4706         error = zfs_userspace_one(zfsvfs,
4707             zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4708         zfsvfs_rele(zfsvfs, FTAG);
4709 
4710         return (error);
4711 }
4712 
4713 /*
4714  * inputs:
4715  * zc_name              name of filesystem
4716  * zc_cookie            zap cursor
4717  * zc_objset_type       zfs_userquota_prop_t
4718  * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4719  *
4720  * outputs:
4721  * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4722  * zc_cookie    zap cursor
4723  */
4724 static int
4725 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4726 {
4727         zfsvfs_t *zfsvfs;
4728         int bufsize = zc->zc_nvlist_dst_size;
4729 
4730         if (bufsize <= 0)
4731                 return (SET_ERROR(ENOMEM));
4732 
4733         int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4734         if (error != 0)
4735                 return (error);
4736 
4737         void *buf = kmem_alloc(bufsize, KM_SLEEP);
4738 
4739         error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
4740             buf, &zc->zc_nvlist_dst_size);
4741 
4742         if (error == 0) {
4743                 error = xcopyout(buf,
4744                     (void *)(uintptr_t)zc->zc_nvlist_dst,
4745                     zc->zc_nvlist_dst_size);
4746         }
4747         kmem_free(buf, bufsize);
4748         zfsvfs_rele(zfsvfs, FTAG);
4749 
4750         return (error);
4751 }
4752 
4753 /*
4754  * inputs:
4755  * zc_name              name of filesystem
4756  *
4757  * outputs:
4758  * none
4759  */
4760 static int
4761 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4762 {
4763         objset_t *os;
4764         int error = 0;
4765         zfsvfs_t *zfsvfs;
4766 
4767         if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
4768                 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
4769                         /*
4770                          * If userused is not enabled, it may be because the
4771                          * objset needs to be closed & reopened (to grow the
4772                          * objset_phys_t).  Suspend/resume the fs will do that.
4773                          */
4774                         error = zfs_suspend_fs(zfsvfs);
4775                         if (error == 0) {
4776                                 dmu_objset_refresh_ownership(zfsvfs->z_os,
4777                                     zfsvfs);
4778                                 error = zfs_resume_fs(zfsvfs, zc->zc_name);
4779                         }
4780                 }
4781                 if (error == 0)
4782                         error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
4783                 VFS_RELE(zfsvfs->z_vfs);
4784         } else {
4785                 /* XXX kind of reading contents without owning */
4786                 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4787                 if (error != 0)
4788                         return (error);
4789 
4790                 error = dmu_objset_userspace_upgrade(os);
4791                 dmu_objset_rele(os, FTAG);
4792         }
4793 
4794         return (error);
4795 }
4796 
4797 /*
4798  * We don't want to have a hard dependency
4799  * against some special symbols in sharefs
4800  * nfs, and smbsrv.  Determine them if needed when
4801  * the first file system is shared.
4802  * Neither sharefs, nfs or smbsrv are unloadable modules.
4803  */
4804 int (*znfsexport_fs)(void *arg);
4805 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
4806 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
4807 
4808 int zfs_nfsshare_inited;
4809 int zfs_smbshare_inited;
4810 
4811 ddi_modhandle_t nfs_mod;
4812 ddi_modhandle_t sharefs_mod;
4813 ddi_modhandle_t smbsrv_mod;
4814 kmutex_t zfs_share_lock;
4815 
4816 static int
4817 zfs_init_sharefs()
4818 {
4819         int error;
4820 
4821         ASSERT(MUTEX_HELD(&zfs_share_lock));
4822         /* Both NFS and SMB shares also require sharetab support. */
4823         if (sharefs_mod == NULL && ((sharefs_mod =
4824             ddi_modopen("fs/sharefs",
4825             KRTLD_MODE_FIRST, &error)) == NULL)) {
4826                 return (SET_ERROR(ENOSYS));
4827         }
4828         if (zshare_fs == NULL && ((zshare_fs =
4829             (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
4830             ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
4831                 return (SET_ERROR(ENOSYS));
4832         }
4833         return (0);
4834 }
4835 
4836 static int
4837 zfs_ioc_share(zfs_cmd_t *zc)
4838 {
4839         int error;
4840         int opcode;
4841 
4842         switch (zc->zc_share.z_sharetype) {
4843         case ZFS_SHARE_NFS:
4844         case ZFS_UNSHARE_NFS:
4845                 if (zfs_nfsshare_inited == 0) {
4846                         mutex_enter(&zfs_share_lock);
4847                         if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
4848                             KRTLD_MODE_FIRST, &error)) == NULL)) {
4849                                 mutex_exit(&zfs_share_lock);
4850                                 return (SET_ERROR(ENOSYS));
4851                         }
4852                         if (znfsexport_fs == NULL &&
4853                             ((znfsexport_fs = (int (*)(void *))
4854                             ddi_modsym(nfs_mod,
4855                             "nfs_export", &error)) == NULL)) {
4856                                 mutex_exit(&zfs_share_lock);
4857                                 return (SET_ERROR(ENOSYS));
4858                         }
4859                         error = zfs_init_sharefs();
4860                         if (error != 0) {
4861                                 mutex_exit(&zfs_share_lock);
4862                                 return (SET_ERROR(ENOSYS));
4863                         }
4864                         zfs_nfsshare_inited = 1;
4865                         mutex_exit(&zfs_share_lock);
4866                 }
4867                 break;
4868         case ZFS_SHARE_SMB:
4869         case ZFS_UNSHARE_SMB:
4870                 if (zfs_smbshare_inited == 0) {
4871                         mutex_enter(&zfs_share_lock);
4872                         if (smbsrv_mod == NULL && ((smbsrv_mod =
4873                             ddi_modopen("drv/smbsrv",
4874                             KRTLD_MODE_FIRST, &error)) == NULL)) {
4875                                 mutex_exit(&zfs_share_lock);
4876                                 return (SET_ERROR(ENOSYS));
4877                         }
4878                         if (zsmbexport_fs == NULL && ((zsmbexport_fs =
4879                             (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
4880                             "smb_server_share", &error)) == NULL)) {
4881                                 mutex_exit(&zfs_share_lock);
4882                                 return (SET_ERROR(ENOSYS));
4883                         }
4884                         error = zfs_init_sharefs();
4885                         if (error != 0) {
4886                                 mutex_exit(&zfs_share_lock);
4887                                 return (SET_ERROR(ENOSYS));
4888                         }
4889                         zfs_smbshare_inited = 1;
4890                         mutex_exit(&zfs_share_lock);
4891                 }
4892                 break;
4893         default:
4894                 return (SET_ERROR(EINVAL));
4895         }
4896 
4897         switch (zc->zc_share.z_sharetype) {
4898         case ZFS_SHARE_NFS:
4899         case ZFS_UNSHARE_NFS:
4900                 if (error =
4901                     znfsexport_fs((void *)
4902                     (uintptr_t)zc->zc_share.z_exportdata))
4903                         return (error);
4904                 break;
4905         case ZFS_SHARE_SMB:
4906         case ZFS_UNSHARE_SMB:
4907                 if (error = zsmbexport_fs((void *)
4908                     (uintptr_t)zc->zc_share.z_exportdata,
4909                     zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
4910                     B_TRUE: B_FALSE)) {
4911                         return (error);
4912                 }
4913                 break;
4914         }
4915 
4916         opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
4917             zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
4918             SHAREFS_ADD : SHAREFS_REMOVE;
4919 
4920         /*
4921          * Add or remove share from sharetab
4922          */
4923         error = zshare_fs(opcode,
4924             (void *)(uintptr_t)zc->zc_share.z_sharedata,
4925             zc->zc_share.z_sharemax);
4926 
4927         return (error);
4928 
4929 }
4930 
4931 ace_t full_access[] = {
4932         {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4933 };
4934 
4935 /*
4936  * inputs:
4937  * zc_name              name of containing filesystem
4938  * zc_obj               object # beyond which we want next in-use object #
4939  *
4940  * outputs:
4941  * zc_obj               next in-use object #
4942  */
4943 static int
4944 zfs_ioc_next_obj(zfs_cmd_t *zc)
4945 {
4946         objset_t *os = NULL;
4947         int error;
4948 
4949         error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4950         if (error != 0)
4951                 return (error);
4952 
4953         error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4954             dsl_dataset_phys(os->os_dsl_dataset)->ds_prev_snap_txg);
4955 
4956         dmu_objset_rele(os, FTAG);
4957         return (error);
4958 }
4959 
4960 /*
4961  * inputs:
4962  * zc_name              name of filesystem
4963  * zc_value             prefix name for snapshot
4964  * zc_cleanup_fd        cleanup-on-exit file descriptor for calling process
4965  *
4966  * outputs:
4967  * zc_value             short name of new snapshot
4968  */
4969 static int
4970 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
4971 {
4972         char *snap_name;
4973         char *hold_name;
4974         int error;
4975         minor_t minor;
4976 
4977         error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
4978         if (error != 0)
4979                 return (error);
4980 
4981         snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
4982             (u_longlong_t)ddi_get_lbolt64());
4983         hold_name = kmem_asprintf("%%%s", zc->zc_value);
4984 
4985         error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
4986             hold_name);
4987         if (error == 0)
4988                 (void) strcpy(zc->zc_value, snap_name);
4989         strfree(snap_name);
4990         strfree(hold_name);
4991         zfs_onexit_fd_rele(zc->zc_cleanup_fd);
4992         return (error);
4993 }
4994 
4995 /*
4996  * inputs:
4997  * zc_name              name of "to" snapshot
4998  * zc_value             name of "from" snapshot
4999  * zc_cookie            file descriptor to write diff data on
5000  *
5001  * outputs:
5002  * dmu_diff_record_t's to the file descriptor
5003  */
5004 static int
5005 zfs_ioc_diff(zfs_cmd_t *zc)
5006 {
5007         file_t *fp;
5008         offset_t off;
5009         int error;
5010 
5011         fp = getf(zc->zc_cookie);
5012         if (fp == NULL)
5013                 return (SET_ERROR(EBADF));
5014 
5015         off = fp->f_offset;
5016 
5017         error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
5018 
5019         if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5020                 fp->f_offset = off;
5021         releasef(zc->zc_cookie);
5022 
5023         return (error);
5024 }
5025 
5026 /*
5027  * Remove all ACL files in shares dir
5028  */
5029 static int
5030 zfs_smb_acl_purge(znode_t *dzp)
5031 {
5032         zap_cursor_t    zc;
5033         zap_attribute_t zap;
5034         zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
5035         int error;
5036 
5037         for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
5038             (error = zap_cursor_retrieve(&zc, &zap)) == 0;
5039             zap_cursor_advance(&zc)) {
5040                 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
5041                     NULL, 0)) != 0)
5042                         break;
5043         }
5044         zap_cursor_fini(&zc);
5045         return (error);
5046 }
5047 
5048 static int
5049 zfs_ioc_smb_acl(zfs_cmd_t *zc)
5050 {
5051         vnode_t *vp;
5052         znode_t *dzp;
5053         vnode_t *resourcevp = NULL;
5054         znode_t *sharedir;
5055         zfsvfs_t *zfsvfs;
5056         nvlist_t *nvlist;
5057         char *src, *target;
5058         vattr_t vattr;
5059         vsecattr_t vsec;
5060         int error = 0;
5061 
5062         if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
5063             NO_FOLLOW, NULL, &vp)) != 0)
5064                 return (error);
5065 
5066         /* Now make sure mntpnt and dataset are ZFS */
5067 
5068         if (vp->v_vfsp->vfs_fstype != zfsfstype ||
5069             (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
5070             zc->zc_name) != 0)) {
5071                 VN_RELE(vp);
5072                 return (SET_ERROR(EINVAL));
5073         }
5074 
5075         dzp = VTOZ(vp);
5076         zfsvfs = dzp->z_zfsvfs;
5077         ZFS_ENTER(zfsvfs);
5078 
5079         /*
5080          * Create share dir if its missing.
5081          */
5082         mutex_enter(&zfsvfs->z_lock);
5083         if (zfsvfs->z_shares_dir == 0) {
5084                 dmu_tx_t *tx;
5085 
5086                 tx = dmu_tx_create(zfsvfs->z_os);
5087                 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
5088                     ZFS_SHARES_DIR);
5089                 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
5090                 error = dmu_tx_assign(tx, TXG_WAIT);
5091                 if (error != 0) {
5092                         dmu_tx_abort(tx);
5093                 } else {
5094                         error = zfs_create_share_dir(zfsvfs, tx);
5095                         dmu_tx_commit(tx);
5096                 }
5097                 if (error != 0) {
5098                         mutex_exit(&zfsvfs->z_lock);
5099                         VN_RELE(vp);
5100                         ZFS_EXIT(zfsvfs);
5101                         return (error);
5102                 }
5103         }
5104         mutex_exit(&zfsvfs->z_lock);
5105 
5106         ASSERT(zfsvfs->z_shares_dir);
5107         if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
5108                 VN_RELE(vp);
5109                 ZFS_EXIT(zfsvfs);
5110                 return (error);
5111         }
5112 
5113         switch (zc->zc_cookie) {
5114         case ZFS_SMB_ACL_ADD:
5115                 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
5116                 vattr.va_type = VREG;
5117                 vattr.va_mode = S_IFREG|0777;
5118                 vattr.va_uid = 0;
5119                 vattr.va_gid = 0;
5120 
5121                 vsec.vsa_mask = VSA_ACE;
5122                 vsec.vsa_aclentp = &full_access;
5123                 vsec.vsa_aclentsz = sizeof (full_access);
5124                 vsec.vsa_aclcnt = 1;
5125 
5126                 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
5127                     &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
5128                 if (resourcevp)
5129                         VN_RELE(resourcevp);
5130                 break;
5131 
5132         case ZFS_SMB_ACL_REMOVE:
5133                 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
5134                     NULL, 0);
5135                 break;
5136 
5137         case ZFS_SMB_ACL_RENAME:
5138                 if ((error = get_nvlist(zc->zc_nvlist_src,
5139                     zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
5140                         VN_RELE(vp);
5141                         VN_RELE(ZTOV(sharedir));
5142                         ZFS_EXIT(zfsvfs);
5143                         return (error);
5144                 }
5145                 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
5146                     nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
5147                     &target)) {
5148                         VN_RELE(vp);
5149                         VN_RELE(ZTOV(sharedir));
5150                         ZFS_EXIT(zfsvfs);
5151                         nvlist_free(nvlist);
5152                         return (error);
5153                 }
5154                 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
5155                     kcred, NULL, 0);
5156                 nvlist_free(nvlist);
5157                 break;
5158 
5159         case ZFS_SMB_ACL_PURGE:
5160                 error = zfs_smb_acl_purge(sharedir);
5161                 break;
5162 
5163         default:
5164                 error = SET_ERROR(EINVAL);
5165                 break;
5166         }
5167 
5168         VN_RELE(vp);
5169         VN_RELE(ZTOV(sharedir));
5170 
5171         ZFS_EXIT(zfsvfs);
5172 
5173         return (error);
5174 }
5175 
5176 /*
5177  * innvl: {
5178  *     "holds" -> { snapname -> holdname (string), ... }
5179  *     (optional) "cleanup_fd" -> fd (int32)
5180  * }
5181  *
5182  * outnvl: {
5183  *     snapname -> error value (int32)
5184  *     ...
5185  * }
5186  */
5187 /* ARGSUSED */
5188 static int
5189 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
5190 {
5191         nvlist_t *holds;
5192         int cleanup_fd = -1;
5193         int error;
5194         minor_t minor = 0;
5195 
5196         error = nvlist_lookup_nvlist(args, "holds", &holds);
5197         if (error != 0)
5198                 return (SET_ERROR(EINVAL));
5199 
5200         if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
5201                 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
5202                 if (error != 0)
5203                         return (error);
5204         }
5205 
5206         error = dsl_dataset_user_hold(holds, minor, errlist);
5207         if (minor != 0)
5208                 zfs_onexit_fd_rele(cleanup_fd);
5209         return (error);
5210 }
5211 
5212 /*
5213  * innvl is not used.
5214  *
5215  * outnvl: {
5216  *    holdname -> time added (uint64 seconds since epoch)
5217  *    ...
5218  * }
5219  */
5220 /* ARGSUSED */
5221 static int
5222 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
5223 {
5224         return (dsl_dataset_get_holds(snapname, outnvl));
5225 }
5226 
5227 /*
5228  * innvl: {
5229  *     snapname -> { holdname, ... }
5230  *     ...
5231  * }
5232  *
5233  * outnvl: {
5234  *     snapname -> error value (int32)
5235  *     ...
5236  * }
5237  */
5238 /* ARGSUSED */
5239 static int
5240 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
5241 {
5242         return (dsl_dataset_user_release(holds, errlist));
5243 }
5244 
5245 /*
5246  * inputs:
5247  * zc_name              name of new filesystem or snapshot
5248  * zc_value             full name of old snapshot
5249  *
5250  * outputs:
5251  * zc_cookie            space in bytes
5252  * zc_objset_type       compressed space in bytes
5253  * zc_perm_action       uncompressed space in bytes
5254  */
5255 static int
5256 zfs_ioc_space_written(zfs_cmd_t *zc)
5257 {
5258         int error;
5259         dsl_pool_t *dp;
5260         dsl_dataset_t *new, *old;
5261 
5262         error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5263         if (error != 0)
5264                 return (error);
5265         error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5266         if (error != 0) {
5267                 dsl_pool_rele(dp, FTAG);
5268                 return (error);
5269         }
5270         error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5271         if (error != 0) {
5272                 dsl_dataset_rele(new, FTAG);
5273                 dsl_pool_rele(dp, FTAG);
5274                 return (error);
5275         }
5276 
5277         error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5278             &zc->zc_objset_type, &zc->zc_perm_action);
5279         dsl_dataset_rele(old, FTAG);
5280         dsl_dataset_rele(new, FTAG);
5281         dsl_pool_rele(dp, FTAG);
5282         return (error);
5283 }
5284 
5285 /*
5286  * innvl: {
5287  *     "firstsnap" -> snapshot name
5288  * }
5289  *
5290  * outnvl: {
5291  *     "used" -> space in bytes
5292  *     "compressed" -> compressed space in bytes
5293  *     "uncompressed" -> uncompressed space in bytes
5294  * }
5295  */
5296 static int
5297 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5298 {
5299         int error;
5300         dsl_pool_t *dp;
5301         dsl_dataset_t *new, *old;
5302         char *firstsnap;
5303         uint64_t used, comp, uncomp;
5304 
5305         if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5306                 return (SET_ERROR(EINVAL));
5307 
5308         error = dsl_pool_hold(lastsnap, FTAG, &dp);
5309         if (error != 0)
5310                 return (error);
5311 
5312         error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5313         if (error != 0) {
5314                 dsl_pool_rele(dp, FTAG);
5315                 return (error);
5316         }
5317         error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5318         if (error != 0) {
5319                 dsl_dataset_rele(new, FTAG);
5320                 dsl_pool_rele(dp, FTAG);
5321                 return (error);
5322         }
5323 
5324         error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5325         dsl_dataset_rele(old, FTAG);
5326         dsl_dataset_rele(new, FTAG);
5327         dsl_pool_rele(dp, FTAG);
5328         fnvlist_add_uint64(outnvl, "used", used);
5329         fnvlist_add_uint64(outnvl, "compressed", comp);
5330         fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5331         return (error);
5332 }
5333 
5334 /*
5335  * innvl: {
5336  *     "fd" -> file descriptor to write stream to (int32)
5337  *     (optional) "fromsnap" -> full snap name to send an incremental from
5338  *     (optional) "largeblockok" -> (value ignored)
5339  *         indicates that blocks > 128KB are permitted
5340  *     (optional) "embedok" -> (value ignored)
5341  *         presence indicates DRR_WRITE_EMBEDDED records are permitted
5342  * }
5343  *
5344  * outnvl is unused
5345  */
5346 /* ARGSUSED */
5347 static int
5348 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5349 {
5350         int error;
5351         offset_t off;
5352         char *fromname = NULL;
5353         int fd;
5354         boolean_t largeblockok;
5355         boolean_t embedok;
5356 
5357         error = nvlist_lookup_int32(innvl, "fd", &fd);
5358         if (error != 0)
5359                 return (SET_ERROR(EINVAL));
5360 
5361         (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5362 
5363         largeblockok = nvlist_exists(innvl, "largeblockok");
5364         embedok = nvlist_exists(innvl, "embedok");
5365 
5366         file_t *fp = getf(fd);
5367         if (fp == NULL)
5368                 return (SET_ERROR(EBADF));
5369 
5370         off = fp->f_offset;
5371         error = dmu_send(snapname, fromname, embedok, largeblockok,
5372             fd, fp->f_vnode, &off);
5373 
5374         if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5375                 fp->f_offset = off;
5376         releasef(fd);
5377         return (error);
5378 }
5379 
5380 /*
5381  * Determine approximately how large a zfs send stream will be -- the number
5382  * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5383  *
5384  * innvl: {
5385  *     (optional) "from" -> full snap or bookmark name to send an incremental
5386  *                          from
5387  * }
5388  *
5389  * outnvl: {
5390  *     "space" -> bytes of space (uint64)
5391  * }
5392  */
5393 static int
5394 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5395 {
5396         dsl_pool_t *dp;
5397         dsl_dataset_t *tosnap;
5398         int error;
5399         char *fromname;
5400         uint64_t space;
5401 
5402         error = dsl_pool_hold(snapname, FTAG, &dp);
5403         if (error != 0)
5404                 return (error);
5405 
5406         error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5407         if (error != 0) {
5408                 dsl_pool_rele(dp, FTAG);
5409                 return (error);
5410         }
5411 
5412         error = nvlist_lookup_string(innvl, "from", &fromname);
5413         if (error == 0) {
5414                 if (strchr(fromname, '@') != NULL) {
5415                         /*
5416                          * If from is a snapshot, hold it and use the more
5417                          * efficient dmu_send_estimate to estimate send space
5418                          * size using deadlists.
5419                          */
5420                         dsl_dataset_t *fromsnap;
5421                         error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5422                         if (error != 0)
5423                                 goto out;
5424                         error = dmu_send_estimate(tosnap, fromsnap, &space);
5425                         dsl_dataset_rele(fromsnap, FTAG);
5426                 } else if (strchr(fromname, '#') != NULL) {
5427                         /*
5428                          * If from is a bookmark, fetch the creation TXG of the
5429                          * snapshot it was created from and use that to find
5430                          * blocks that were born after it.
5431                          */
5432                         zfs_bookmark_phys_t frombm;
5433 
5434                         error = dsl_bookmark_lookup(dp, fromname, tosnap,
5435                             &frombm);
5436                         if (error != 0)
5437                                 goto out;
5438                         error = dmu_send_estimate_from_txg(tosnap,
5439                             frombm.zbm_creation_txg, &space);
5440                 } else {
5441                         /*
5442                          * from is not properly formatted as a snapshot or
5443                          * bookmark
5444                          */
5445                         error = SET_ERROR(EINVAL);
5446                         goto out;
5447                 }
5448         } else {
5449                 // If estimating the size of a full send, use dmu_send_estimate
5450                 error = dmu_send_estimate(tosnap, NULL, &space);
5451         }
5452 
5453         fnvlist_add_uint64(outnvl, "space", space);
5454 
5455 out:
5456         dsl_dataset_rele(tosnap, FTAG);
5457         dsl_pool_rele(dp, FTAG);
5458         return (error);
5459 }
5460 
5461 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5462 
5463 static void
5464 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5465     zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5466     boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5467 {
5468         zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5469 
5470         ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5471         ASSERT3U(ioc, <, ZFS_IOC_LAST);
5472         ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5473         ASSERT3P(vec->zvec_func, ==, NULL);
5474 
5475         vec->zvec_legacy_func = func;
5476         vec->zvec_secpolicy = secpolicy;
5477         vec->zvec_namecheck = namecheck;
5478         vec->zvec_allow_log = log_history;
5479         vec->zvec_pool_check = pool_check;
5480 }
5481 
5482 /*
5483  * See the block comment at the beginning of this file for details on
5484  * each argument to this function.
5485  */
5486 static void
5487 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5488     zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5489     zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5490     boolean_t allow_log)
5491 {
5492         zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5493 
5494         ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5495         ASSERT3U(ioc, <, ZFS_IOC_LAST);
5496         ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5497         ASSERT3P(vec->zvec_func, ==, NULL);
5498 
5499         /* if we are logging, the name must be valid */
5500         ASSERT(!allow_log || namecheck != NO_NAME);
5501 
5502         vec->zvec_name = name;
5503         vec->zvec_func = func;
5504         vec->zvec_secpolicy = secpolicy;
5505         vec->zvec_namecheck = namecheck;
5506         vec->zvec_pool_check = pool_check;
5507         vec->zvec_smush_outnvlist = smush_outnvlist;
5508         vec->zvec_allow_log = allow_log;
5509 }
5510 
5511 static void
5512 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5513     zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5514     zfs_ioc_poolcheck_t pool_check)
5515 {
5516         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5517             POOL_NAME, log_history, pool_check);
5518 }
5519 
5520 static void
5521 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5522     zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5523 {
5524         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5525             DATASET_NAME, B_FALSE, pool_check);
5526 }
5527 
5528 static void
5529 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5530 {
5531         zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5532             POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5533 }
5534 
5535 static void
5536 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5537     zfs_secpolicy_func_t *secpolicy)
5538 {
5539         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5540             NO_NAME, B_FALSE, POOL_CHECK_NONE);
5541 }
5542 
5543 static void
5544 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5545     zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5546 {
5547         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5548             DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5549 }
5550 
5551 static void
5552 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5553 {
5554         zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5555             zfs_secpolicy_read);
5556 }
5557 
5558 static void
5559 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5560         zfs_secpolicy_func_t *secpolicy)
5561 {
5562         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5563             DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5564 }
5565 
5566 static void
5567 zfs_ioctl_init(void)
5568 {
5569         zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5570             zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5571             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5572 
5573         zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5574             zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5575             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5576 
5577         zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5578             zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5579             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5580 
5581         zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5582             zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5583             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5584 
5585         zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5586             zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5587             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5588 
5589         zfs_ioctl_register("create", ZFS_IOC_CREATE,
5590             zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5591             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5592 
5593         zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5594             zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5595             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5596 
5597         zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5598             zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5599             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5600 
5601         zfs_ioctl_register("hold", ZFS_IOC_HOLD,
5602             zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
5603             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5604         zfs_ioctl_register("release", ZFS_IOC_RELEASE,
5605             zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
5606             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5607 
5608         zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
5609             zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
5610             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5611 
5612         zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
5613             zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
5614             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
5615 
5616         zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
5617             zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
5618             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5619 
5620         zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
5621             zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
5622             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5623 
5624         zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
5625             zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
5626             POOL_NAME,
5627             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5628 
5629         /* IOCTLS that use the legacy function signature */
5630 
5631         zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5632             zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5633 
5634         zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5635             zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5636         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5637             zfs_ioc_pool_scan);
5638         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5639             zfs_ioc_pool_upgrade);
5640         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5641             zfs_ioc_vdev_add);
5642         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5643             zfs_ioc_vdev_remove);
5644         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5645             zfs_ioc_vdev_set_state);
5646         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5647             zfs_ioc_vdev_attach);
5648         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5649             zfs_ioc_vdev_detach);
5650         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5651             zfs_ioc_vdev_setpath);
5652         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5653             zfs_ioc_vdev_setfru);
5654         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5655             zfs_ioc_pool_set_props);
5656         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5657             zfs_ioc_vdev_split);
5658         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5659             zfs_ioc_pool_reguid);
5660 
5661         zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5662             zfs_ioc_pool_configs, zfs_secpolicy_none);
5663         zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5664             zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5665         zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5666             zfs_ioc_inject_fault, zfs_secpolicy_inject);
5667         zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5668             zfs_ioc_clear_fault, zfs_secpolicy_inject);
5669         zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
5670             zfs_ioc_inject_list_next, zfs_secpolicy_inject);
5671 
5672         /*
5673          * pool destroy, and export don't log the history as part of
5674          * zfsdev_ioctl, but rather zfs_ioc_pool_export
5675          * does the logging of those commands.
5676          */
5677         zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
5678             zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5679         zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
5680             zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5681 
5682         zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
5683             zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5684         zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
5685             zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5686 
5687         zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
5688             zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
5689         zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
5690             zfs_ioc_dsobj_to_dsname,
5691             zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
5692         zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
5693             zfs_ioc_pool_get_history,
5694             zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
5695 
5696         zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
5697             zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5698 
5699         zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
5700             zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5701         zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
5702             zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5703 
5704         zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
5705             zfs_ioc_space_written);
5706         zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
5707             zfs_ioc_objset_recvd_props);
5708         zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
5709             zfs_ioc_next_obj);
5710         zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
5711             zfs_ioc_get_fsacl);
5712         zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
5713             zfs_ioc_objset_stats);
5714         zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
5715             zfs_ioc_objset_zplprops);
5716         zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
5717             zfs_ioc_dataset_list_next);
5718         zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
5719             zfs_ioc_snapshot_list_next);
5720         zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
5721             zfs_ioc_send_progress);
5722 
5723         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
5724             zfs_ioc_diff, zfs_secpolicy_diff);
5725         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
5726             zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
5727         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
5728             zfs_ioc_obj_to_path, zfs_secpolicy_diff);
5729         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
5730             zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
5731         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
5732             zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
5733         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
5734             zfs_ioc_send, zfs_secpolicy_send);
5735 
5736         zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
5737             zfs_secpolicy_none);
5738         zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
5739             zfs_secpolicy_destroy);
5740         zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
5741             zfs_secpolicy_rename);
5742         zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
5743             zfs_secpolicy_recv);
5744         zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
5745             zfs_secpolicy_promote);
5746         zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
5747             zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
5748         zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
5749             zfs_secpolicy_set_fsacl);
5750 
5751         zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
5752             zfs_secpolicy_share, POOL_CHECK_NONE);
5753         zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
5754             zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
5755         zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
5756             zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
5757             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5758         zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
5759             zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
5760             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5761 }
5762 
5763 int
5764 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
5765     zfs_ioc_poolcheck_t check)
5766 {
5767         spa_t *spa;
5768         int error;
5769 
5770         ASSERT(type == POOL_NAME || type == DATASET_NAME);
5771 
5772         if (check & POOL_CHECK_NONE)
5773                 return (0);
5774 
5775         error = spa_open(name, &spa, FTAG);
5776         if (error == 0) {
5777                 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
5778                         error = SET_ERROR(EAGAIN);
5779                 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
5780                         error = SET_ERROR(EROFS);
5781                 spa_close(spa, FTAG);
5782         }
5783         return (error);
5784 }
5785 
5786 /*
5787  * Find a free minor number.
5788  */
5789 minor_t
5790 zfsdev_minor_alloc(void)
5791 {
5792         static minor_t last_minor;
5793         minor_t m;
5794 
5795         ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5796 
5797         for (m = last_minor + 1; m != last_minor; m++) {
5798                 if (m > ZFSDEV_MAX_MINOR)
5799                         m = 1;
5800                 if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
5801                         last_minor = m;
5802                         return (m);
5803                 }
5804         }
5805 
5806         return (0);
5807 }
5808 
5809 static int
5810 zfs_ctldev_init(dev_t *devp)
5811 {
5812         minor_t minor;
5813         zfs_soft_state_t *zs;
5814 
5815         ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5816         ASSERT(getminor(*devp) == 0);
5817 
5818         minor = zfsdev_minor_alloc();
5819         if (minor == 0)
5820                 return (SET_ERROR(ENXIO));
5821 
5822         if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
5823                 return (SET_ERROR(EAGAIN));
5824 
5825         *devp = makedevice(getemajor(*devp), minor);
5826 
5827         zs = ddi_get_soft_state(zfsdev_state, minor);
5828         zs->zss_type = ZSST_CTLDEV;
5829         zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
5830 
5831         return (0);
5832 }
5833 
5834 static void
5835 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
5836 {
5837         ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5838 
5839         zfs_onexit_destroy(zo);
5840         ddi_soft_state_free(zfsdev_state, minor);
5841 }
5842 
5843 void *
5844 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
5845 {
5846         zfs_soft_state_t *zp;
5847 
5848         zp = ddi_get_soft_state(zfsdev_state, minor);
5849         if (zp == NULL || zp->zss_type != which)
5850                 return (NULL);
5851 
5852         return (zp->zss_data);
5853 }
5854 
5855 static int
5856 zfsdev_open(dev_t *devp, int flag, int otyp, cred_t *cr)
5857 {
5858         int error = 0;
5859 
5860         if (getminor(*devp) != 0)
5861                 return (zvol_open(devp, flag, otyp, cr));
5862 
5863         /* This is the control device. Allocate a new minor if requested. */
5864         if (flag & FEXCL) {
5865                 mutex_enter(&zfsdev_state_lock);
5866                 error = zfs_ctldev_init(devp);
5867                 mutex_exit(&zfsdev_state_lock);
5868         }
5869 
5870         return (error);
5871 }
5872 
5873 static int
5874 zfsdev_close(dev_t dev, int flag, int otyp, cred_t *cr)
5875 {
5876         zfs_onexit_t *zo;
5877         minor_t minor = getminor(dev);
5878 
5879         if (minor == 0)
5880                 return (0);
5881 
5882         mutex_enter(&zfsdev_state_lock);
5883         zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
5884         if (zo == NULL) {
5885                 mutex_exit(&zfsdev_state_lock);
5886                 return (zvol_close(dev, flag, otyp, cr));
5887         }
5888         zfs_ctldev_destroy(zo, minor);
5889         mutex_exit(&zfsdev_state_lock);
5890 
5891         return (0);
5892 }
5893 
5894 static int
5895 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
5896 {
5897         zfs_cmd_t *zc;
5898         uint_t vecnum;
5899         int error, rc, len;
5900         minor_t minor = getminor(dev);
5901         const zfs_ioc_vec_t *vec;
5902         char *saved_poolname = NULL;
5903         nvlist_t *innvl = NULL;
5904 
5905         if (minor != 0 &&
5906             zfsdev_get_soft_state(minor, ZSST_CTLDEV) == NULL)
5907                 return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
5908 
5909         vecnum = cmd - ZFS_IOC_FIRST;
5910         ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
5911 
5912         if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
5913                 return (SET_ERROR(EINVAL));
5914         vec = &zfs_ioc_vec[vecnum];
5915 
5916         zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
5917 
5918         error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
5919         if (error != 0) {
5920                 error = SET_ERROR(EFAULT);
5921                 goto out;
5922         }
5923 
5924         zc->zc_iflags = flag & FKIOCTL;
5925         if (zc->zc_nvlist_src_size != 0) {
5926                 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
5927                     zc->zc_iflags, &innvl);
5928                 if (error != 0)
5929                         goto out;
5930         }
5931 
5932         /*
5933          * Ensure that all pool/dataset names are valid before we pass down to
5934          * the lower layers.
5935          */
5936         zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
5937         switch (vec->zvec_namecheck) {
5938         case POOL_NAME:
5939                 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
5940                         error = SET_ERROR(EINVAL);
5941                 else
5942                         error = pool_status_check(zc->zc_name,
5943                             vec->zvec_namecheck, vec->zvec_pool_check);
5944                 break;
5945 
5946         case DATASET_NAME:
5947                 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
5948                         error = SET_ERROR(EINVAL);
5949                 else
5950                         error = pool_status_check(zc->zc_name,
5951                             vec->zvec_namecheck, vec->zvec_pool_check);
5952                 break;
5953 
5954         case NO_NAME:
5955                 break;
5956         }
5957 
5958 
5959         if (error == 0 && !(flag & FKIOCTL))
5960                 error = vec->zvec_secpolicy(zc, innvl, cr);
5961 
5962         if (error != 0)
5963                 goto out;
5964 
5965         /* legacy ioctls can modify zc_name */
5966         len = strcspn(zc->zc_name, "/@#") + 1;
5967         saved_poolname = kmem_alloc(len, KM_SLEEP);
5968         (void) strlcpy(saved_poolname, zc->zc_name, len);
5969 
5970         if (vec->zvec_func != NULL) {
5971                 nvlist_t *outnvl;
5972                 int puterror = 0;
5973                 spa_t *spa;
5974                 nvlist_t *lognv = NULL;
5975 
5976                 ASSERT(vec->zvec_legacy_func == NULL);
5977 
5978                 /*
5979                  * Add the innvl to the lognv before calling the func,
5980                  * in case the func changes the innvl.
5981                  */
5982                 if (vec->zvec_allow_log) {
5983                         lognv = fnvlist_alloc();
5984                         fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
5985                             vec->zvec_name);
5986                         if (!nvlist_empty(innvl)) {
5987                                 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
5988                                     innvl);
5989                         }
5990                 }
5991 
5992                 outnvl = fnvlist_alloc();
5993                 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
5994 
5995                 if (error == 0 && vec->zvec_allow_log &&
5996                     spa_open(zc->zc_name, &spa, FTAG) == 0) {
5997                         if (!nvlist_empty(outnvl)) {
5998                                 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
5999                                     outnvl);
6000                         }
6001                         (void) spa_history_log_nvl(spa, lognv);
6002                         spa_close(spa, FTAG);
6003                 }
6004                 fnvlist_free(lognv);
6005 
6006                 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
6007                         int smusherror = 0;
6008                         if (vec->zvec_smush_outnvlist) {
6009                                 smusherror = nvlist_smush(outnvl,
6010                                     zc->zc_nvlist_dst_size);
6011                         }
6012                         if (smusherror == 0)
6013                                 puterror = put_nvlist(zc, outnvl);
6014                 }
6015 
6016                 if (puterror != 0)
6017                         error = puterror;
6018 
6019                 nvlist_free(outnvl);
6020         } else {
6021                 error = vec->zvec_legacy_func(zc);
6022         }
6023 
6024 out:
6025         nvlist_free(innvl);
6026         rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
6027         if (error == 0 && rc != 0)
6028                 error = SET_ERROR(EFAULT);
6029         if (error == 0 && vec->zvec_allow_log) {
6030                 char *s = tsd_get(zfs_allow_log_key);
6031                 if (s != NULL)
6032                         strfree(s);
6033                 (void) tsd_set(zfs_allow_log_key, saved_poolname);
6034         } else {
6035                 if (saved_poolname != NULL)
6036                         strfree(saved_poolname);
6037         }
6038 
6039         kmem_free(zc, sizeof (zfs_cmd_t));
6040         return (error);
6041 }
6042 
6043 static int
6044 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
6045 {
6046         if (cmd != DDI_ATTACH)
6047                 return (DDI_FAILURE);
6048 
6049         if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
6050             DDI_PSEUDO, 0) == DDI_FAILURE)
6051                 return (DDI_FAILURE);
6052 
6053         zfs_dip = dip;
6054 
6055         ddi_report_dev(dip);
6056 
6057         return (DDI_SUCCESS);
6058 }
6059 
6060 static int
6061 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
6062 {
6063         if (spa_busy() || zfs_busy() || zvol_busy())
6064                 return (DDI_FAILURE);
6065 
6066         if (cmd != DDI_DETACH)
6067                 return (DDI_FAILURE);
6068 
6069         zfs_dip = NULL;
6070 
6071         ddi_prop_remove_all(dip);
6072         ddi_remove_minor_node(dip, NULL);
6073 
6074         return (DDI_SUCCESS);
6075 }
6076 
6077 /*ARGSUSED*/
6078 static int
6079 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
6080 {
6081         switch (infocmd) {
6082         case DDI_INFO_DEVT2DEVINFO:
6083                 *result = zfs_dip;
6084                 return (DDI_SUCCESS);
6085 
6086         case DDI_INFO_DEVT2INSTANCE:
6087                 *result = (void *)0;
6088                 return (DDI_SUCCESS);
6089         }
6090 
6091         return (DDI_FAILURE);
6092 }
6093 
6094 /*
6095  * OK, so this is a little weird.
6096  *
6097  * /dev/zfs is the control node, i.e. minor 0.
6098  * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
6099  *
6100  * /dev/zfs has basically nothing to do except serve up ioctls,
6101  * so most of the standard driver entry points are in zvol.c.
6102  */
6103 static struct cb_ops zfs_cb_ops = {
6104         zfsdev_open,    /* open */
6105         zfsdev_close,   /* close */
6106         zvol_strategy,  /* strategy */
6107         nodev,          /* print */
6108         zvol_dump,      /* dump */
6109         zvol_read,      /* read */
6110         zvol_write,     /* write */
6111         zfsdev_ioctl,   /* ioctl */
6112         nodev,          /* devmap */
6113         nodev,          /* mmap */
6114         nodev,          /* segmap */
6115         nochpoll,       /* poll */
6116         ddi_prop_op,    /* prop_op */
6117         NULL,           /* streamtab */
6118         D_NEW | D_MP | D_64BIT,         /* Driver compatibility flag */
6119         CB_REV,         /* version */
6120         nodev,          /* async read */
6121         nodev,          /* async write */
6122 };
6123 
6124 static struct dev_ops zfs_dev_ops = {
6125         DEVO_REV,       /* version */
6126         0,              /* refcnt */
6127         zfs_info,       /* info */
6128         nulldev,        /* identify */
6129         nulldev,        /* probe */
6130         zfs_attach,     /* attach */
6131         zfs_detach,     /* detach */
6132         nodev,          /* reset */
6133         &zfs_cb_ops,        /* driver operations */
6134         NULL,           /* no bus operations */
6135         NULL,           /* power */
6136         ddi_quiesce_not_needed, /* quiesce */
6137 };
6138 
6139 static struct modldrv zfs_modldrv = {
6140         &mod_driverops,
6141         "ZFS storage pool",
6142         &zfs_dev_ops
6143 };
6144 
6145 static struct modlinkage modlinkage = {
6146         MODREV_1,
6147         (void *)&zfs_modlfs,
6148         (void *)&zfs_modldrv,
6149         NULL
6150 };
6151 
6152 static void
6153 zfs_allow_log_destroy(void *arg)
6154 {
6155         char *poolname = arg;
6156         strfree(poolname);
6157 }
6158 
6159 int
6160 _init(void)
6161 {
6162         int error;
6163 
6164         spa_init(FREAD | FWRITE);
6165         zfs_init();
6166         zvol_init();
6167         zfs_ioctl_init();
6168 
6169         if ((error = mod_install(&modlinkage)) != 0) {
6170                 zvol_fini();
6171                 zfs_fini();
6172                 spa_fini();
6173                 return (error);
6174         }
6175 
6176         tsd_create(&zfs_fsyncer_key, NULL);
6177         tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6178         tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6179 
6180         error = ldi_ident_from_mod(&modlinkage, &zfs_li);
6181         ASSERT(error == 0);
6182         mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6183 
6184         return (0);
6185 }
6186 
6187 int
6188 _fini(void)
6189 {
6190         int error;
6191 
6192         if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
6193                 return (SET_ERROR(EBUSY));
6194 
6195         if ((error = mod_remove(&modlinkage)) != 0)
6196                 return (error);
6197 
6198         zvol_fini();
6199         zfs_fini();
6200         spa_fini();
6201         if (zfs_nfsshare_inited)
6202                 (void) ddi_modclose(nfs_mod);
6203         if (zfs_smbshare_inited)
6204                 (void) ddi_modclose(smbsrv_mod);
6205         if (zfs_nfsshare_inited || zfs_smbshare_inited)
6206                 (void) ddi_modclose(sharefs_mod);
6207 
6208         tsd_destroy(&zfs_fsyncer_key);
6209         ldi_ident_release(zfs_li);
6210         zfs_li = NULL;
6211         mutex_destroy(&zfs_share_lock);
6212 
6213         return (error);
6214 }
6215 
6216 int
6217 _info(struct modinfo *modinfop)
6218 {
6219         return (mod_info(&modlinkage, modinfop));
6220 }