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 2015, OmniTI Computer Consulting, Inc. All rights reserved.
  26  * Copyright 2015 Nexenta Systems, Inc.  All rights reserved.
  27  * Copyright (c) 2014, Joyent, Inc. All rights reserved.
  28  * Copyright (c) 2011, 2015 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  */
  32 
  33 /*
  34  * ZFS ioctls.
  35  *
  36  * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
  37  * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
  38  *
  39  * There are two ways that we handle ioctls: the legacy way where almost
  40  * all of the logic is in the ioctl callback, and the new way where most
  41  * of the marshalling is handled in the common entry point, zfsdev_ioctl().
  42  *
  43  * Non-legacy ioctls should be registered by calling
  44  * zfs_ioctl_register() from zfs_ioctl_init().  The ioctl is invoked
  45  * from userland by lzc_ioctl().
  46  *
  47  * The registration arguments are as follows:
  48  *
  49  * const char *name
  50  *   The name of the ioctl.  This is used for history logging.  If the
  51  *   ioctl returns successfully (the callback returns 0), and allow_log
  52  *   is true, then a history log entry will be recorded with the input &
  53  *   output nvlists.  The log entry can be printed with "zpool history -i".
  54  *
  55  * zfs_ioc_t ioc
  56  *   The ioctl request number, which userland will pass to ioctl(2).
  57  *   The ioctl numbers can change from release to release, because
  58  *   the caller (libzfs) must be matched to the kernel.
  59  *
  60  * zfs_secpolicy_func_t *secpolicy
  61  *   This function will be called before the zfs_ioc_func_t, to
  62  *   determine if this operation is permitted.  It should return EPERM
  63  *   on failure, and 0 on success.  Checks include determining if the
  64  *   dataset is visible in this zone, and if the user has either all
  65  *   zfs privileges in the zone (SYS_MOUNT), or has been granted permission
  66  *   to do this operation on this dataset with "zfs allow".
  67  *
  68  * zfs_ioc_namecheck_t namecheck
  69  *   This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
  70  *   name, a dataset name, or nothing.  If the name is not well-formed,
  71  *   the ioctl will fail and the callback will not be called.
  72  *   Therefore, the callback can assume that the name is well-formed
  73  *   (e.g. is null-terminated, doesn't have more than one '@' character,
  74  *   doesn't have invalid characters).
  75  *
  76  * zfs_ioc_poolcheck_t pool_check
  77  *   This specifies requirements on the pool state.  If the pool does
  78  *   not meet them (is suspended or is readonly), the ioctl will fail
  79  *   and the callback will not be called.  If any checks are specified
  80  *   (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
  81  *   Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
  82  *   POOL_CHECK_READONLY).
  83  *
  84  * boolean_t smush_outnvlist
  85  *   If smush_outnvlist is true, then the output is presumed to be a
  86  *   list of errors, and it will be "smushed" down to fit into the
  87  *   caller's buffer, by removing some entries and replacing them with a
  88  *   single "N_MORE_ERRORS" entry indicating how many were removed.  See
  89  *   nvlist_smush() for details.  If smush_outnvlist is false, and the
  90  *   outnvlist does not fit into the userland-provided buffer, then the
  91  *   ioctl will fail with ENOMEM.
  92  *
  93  * zfs_ioc_func_t *func
  94  *   The callback function that will perform the operation.
  95  *
  96  *   The callback should return 0 on success, or an error number on
  97  *   failure.  If the function fails, the userland ioctl will return -1,
  98  *   and errno will be set to the callback's return value.  The callback
  99  *   will be called with the following arguments:
 100  *
 101  *   const char *name
 102  *     The name of the pool or dataset to operate on, from
 103  *     zfs_cmd_t:zc_name.  The 'namecheck' argument specifies the
 104  *     expected type (pool, dataset, or none).
 105  *
 106  *   nvlist_t *innvl
 107  *     The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src.  Or
 108  *     NULL if no input nvlist was provided.  Changes to this nvlist are
 109  *     ignored.  If the input nvlist could not be deserialized, the
 110  *     ioctl will fail and the callback will not be called.
 111  *
 112  *   nvlist_t *outnvl
 113  *     The output nvlist, initially empty.  The callback can fill it in,
 114  *     and it will be returned to userland by serializing it into
 115  *     zfs_cmd_t:zc_nvlist_dst.  If it is non-empty, and serialization
 116  *     fails (e.g. because the caller didn't supply a large enough
 117  *     buffer), then the overall ioctl will fail.  See the
 118  *     'smush_nvlist' argument above for additional behaviors.
 119  *
 120  *     There are two typical uses of the output nvlist:
 121  *       - To return state, e.g. property values.  In this case,
 122  *         smush_outnvlist should be false.  If the buffer was not large
 123  *         enough, the caller will reallocate a larger buffer and try
 124  *         the ioctl again.
 125  *
 126  *       - To return multiple errors from an ioctl which makes on-disk
 127  *         changes.  In this case, smush_outnvlist should be true.
 128  *         Ioctls which make on-disk modifications should generally not
 129  *         use the outnvl if they succeed, because the caller can not
 130  *         distinguish between the operation failing, and
 131  *         deserialization failing.
 132  */
 133 
 134 #include <sys/types.h>
 135 #include <sys/param.h>
 136 #include <sys/errno.h>
 137 #include <sys/uio.h>
 138 #include <sys/buf.h>
 139 #include <sys/modctl.h>
 140 #include <sys/open.h>
 141 #include <sys/file.h>
 142 #include <sys/kmem.h>
 143 #include <sys/conf.h>
 144 #include <sys/cmn_err.h>
 145 #include <sys/stat.h>
 146 #include <sys/zfs_ioctl.h>
 147 #include <sys/zfs_vfsops.h>
 148 #include <sys/zfs_znode.h>
 149 #include <sys/zap.h>
 150 #include <sys/spa.h>
 151 #include <sys/spa_impl.h>
 152 #include <sys/vdev.h>
 153 #include <sys/priv_impl.h>
 154 #include <sys/dmu.h>
 155 #include <sys/dsl_dir.h>
 156 #include <sys/dsl_dataset.h>
 157 #include <sys/dsl_prop.h>
 158 #include <sys/dsl_deleg.h>
 159 #include <sys/dmu_objset.h>
 160 #include <sys/dmu_impl.h>
 161 #include <sys/dmu_tx.h>
 162 #include <sys/ddi.h>
 163 #include <sys/sunddi.h>
 164 #include <sys/sunldi.h>
 165 #include <sys/policy.h>
 166 #include <sys/zone.h>
 167 #include <sys/nvpair.h>
 168 #include <sys/pathname.h>
 169 #include <sys/mount.h>
 170 #include <sys/sdt.h>
 171 #include <sys/fs/zfs.h>
 172 #include <sys/zfs_ctldir.h>
 173 #include <sys/zfs_dir.h>
 174 #include <sys/zfs_onexit.h>
 175 #include <sys/zvol.h>
 176 #include <sys/dsl_scan.h>
 177 #include <sharefs/share.h>
 178 #include <sys/dmu_objset.h>
 179 #include <sys/dmu_send.h>
 180 #include <sys/dsl_destroy.h>
 181 #include <sys/dsl_bookmark.h>
 182 #include <sys/dsl_userhold.h>
 183 #include <sys/zfeature.h>
 184 #include <sys/zio_checksum.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 (SET_ERROR(EFAULT));
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_RECORDSIZE:
3823                 /* Record sizes above 128k need the feature to be enabled */
3824                 if (nvpair_value_uint64(pair, &intval) == 0 &&
3825                     intval > SPA_OLD_MAXBLOCKSIZE) {
3826                         spa_t *spa;
3827 
3828                         /*
3829                          * If this is a bootable dataset then
3830                          * the we don't allow large (>128K) blocks,
3831                          * because GRUB doesn't support them.
3832                          */
3833                         if (zfs_is_bootfs(dsname) &&
3834                             intval > SPA_OLD_MAXBLOCKSIZE) {
3835                                 return (SET_ERROR(ERANGE));
3836                         }
3837 
3838                         /*
3839                          * We don't allow setting the property above 1MB,
3840                          * unless the tunable has been changed.
3841                          */
3842                         if (intval > zfs_max_recordsize ||
3843                             intval > SPA_MAXBLOCKSIZE)
3844                                 return (SET_ERROR(ERANGE));
3845 
3846                         if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3847                                 return (err);
3848 
3849                         if (!spa_feature_is_enabled(spa,
3850                             SPA_FEATURE_LARGE_BLOCKS)) {
3851                                 spa_close(spa, FTAG);
3852                                 return (SET_ERROR(ENOTSUP));
3853                         }
3854                         spa_close(spa, FTAG);
3855                 }
3856                 break;
3857 
3858         case ZFS_PROP_SHARESMB:
3859                 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3860                         return (SET_ERROR(ENOTSUP));
3861                 break;
3862 
3863         case ZFS_PROP_ACLINHERIT:
3864                 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3865                     nvpair_value_uint64(pair, &intval) == 0) {
3866                         if (intval == ZFS_ACL_PASSTHROUGH_X &&
3867                             zfs_earlier_version(dsname,
3868                             SPA_VERSION_PASSTHROUGH_X))
3869                                 return (SET_ERROR(ENOTSUP));
3870                 }
3871                 break;
3872 
3873         case ZFS_PROP_CHECKSUM:
3874         case ZFS_PROP_DEDUP:
3875         {
3876                 spa_feature_t feature;
3877                 spa_t *spa;
3878 
3879                 /* dedup feature version checks */
3880                 if (prop == ZFS_PROP_DEDUP &&
3881                     zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3882                         return (SET_ERROR(ENOTSUP));
3883 
3884                 if (nvpair_value_uint64(pair, &intval) != 0)
3885                         return (SET_ERROR(EINVAL));
3886 
3887                 /* check prop value is enabled in features */
3888                 feature = zio_checksum_to_feature(intval);
3889                 if (feature == SPA_FEATURE_NONE)
3890                         break;
3891 
3892                 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3893                         return (err);
3894                 /*
3895                  * Salted checksums are not supported on root pools.
3896                  */
3897                 if (spa_bootfs(spa) != 0 &&
3898                     intval < ZIO_CHECKSUM_FUNCTIONS &&
3899                     (zio_checksum_table[intval].ci_flags &
3900                     ZCHECKSUM_FLAG_SALTED)) {
3901                         spa_close(spa, FTAG);
3902                         return (SET_ERROR(ERANGE));
3903                 }
3904                 if (!spa_feature_is_enabled(spa, feature)) {
3905                         spa_close(spa, FTAG);
3906                         return (SET_ERROR(ENOTSUP));
3907                 }
3908                 spa_close(spa, FTAG);
3909                 break;
3910         }
3911         }
3912 
3913         return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3914 }
3915 
3916 /*
3917  * Checks for a race condition to make sure we don't increment a feature flag
3918  * multiple times.
3919  */
3920 static int
3921 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
3922 {
3923         spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3924         spa_feature_t *featurep = arg;
3925 
3926         if (!spa_feature_is_active(spa, *featurep))
3927                 return (0);
3928         else
3929                 return (SET_ERROR(EBUSY));
3930 }
3931 
3932 /*
3933  * The callback invoked on feature activation in the sync task caused by
3934  * zfs_prop_activate_feature.
3935  */
3936 static void
3937 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
3938 {
3939         spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3940         spa_feature_t *featurep = arg;
3941 
3942         spa_feature_incr(spa, *featurep, tx);
3943 }
3944 
3945 /*
3946  * Activates a feature on a pool in response to a property setting. This
3947  * creates a new sync task which modifies the pool to reflect the feature
3948  * as being active.
3949  */
3950 static int
3951 zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature)
3952 {
3953         int err;
3954 
3955         /* EBUSY here indicates that the feature is already active */
3956         err = dsl_sync_task(spa_name(spa),
3957             zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
3958             &feature, 2, ZFS_SPACE_CHECK_RESERVED);
3959 
3960         if (err != 0 && err != EBUSY)
3961                 return (err);
3962         else
3963                 return (0);
3964 }
3965 
3966 /*
3967  * Removes properties from the given props list that fail permission checks
3968  * needed to clear them and to restore them in case of a receive error. For each
3969  * property, make sure we have both set and inherit permissions.
3970  *
3971  * Returns the first error encountered if any permission checks fail. If the
3972  * caller provides a non-NULL errlist, it also gives the complete list of names
3973  * of all the properties that failed a permission check along with the
3974  * corresponding error numbers. The caller is responsible for freeing the
3975  * returned errlist.
3976  *
3977  * If every property checks out successfully, zero is returned and the list
3978  * pointed at by errlist is NULL.
3979  */
3980 static int
3981 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
3982 {
3983         zfs_cmd_t *zc;
3984         nvpair_t *pair, *next_pair;
3985         nvlist_t *errors;
3986         int err, rv = 0;
3987 
3988         if (props == NULL)
3989                 return (0);
3990 
3991         VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3992 
3993         zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
3994         (void) strcpy(zc->zc_name, dataset);
3995         pair = nvlist_next_nvpair(props, NULL);
3996         while (pair != NULL) {
3997                 next_pair = nvlist_next_nvpair(props, pair);
3998 
3999                 (void) strcpy(zc->zc_value, nvpair_name(pair));
4000                 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
4001                     (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
4002                         VERIFY(nvlist_remove_nvpair(props, pair) == 0);
4003                         VERIFY(nvlist_add_int32(errors,
4004                             zc->zc_value, err) == 0);
4005                 }
4006                 pair = next_pair;
4007         }
4008         kmem_free(zc, sizeof (zfs_cmd_t));
4009 
4010         if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
4011                 nvlist_free(errors);
4012                 errors = NULL;
4013         } else {
4014                 VERIFY(nvpair_value_int32(pair, &rv) == 0);
4015         }
4016 
4017         if (errlist == NULL)
4018                 nvlist_free(errors);
4019         else
4020                 *errlist = errors;
4021 
4022         return (rv);
4023 }
4024 
4025 static boolean_t
4026 propval_equals(nvpair_t *p1, nvpair_t *p2)
4027 {
4028         if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
4029                 /* dsl_prop_get_all_impl() format */
4030                 nvlist_t *attrs;
4031                 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
4032                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4033                     &p1) == 0);
4034         }
4035 
4036         if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
4037                 nvlist_t *attrs;
4038                 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
4039                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4040                     &p2) == 0);
4041         }
4042 
4043         if (nvpair_type(p1) != nvpair_type(p2))
4044                 return (B_FALSE);
4045 
4046         if (nvpair_type(p1) == DATA_TYPE_STRING) {
4047                 char *valstr1, *valstr2;
4048 
4049                 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
4050                 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
4051                 return (strcmp(valstr1, valstr2) == 0);
4052         } else {
4053                 uint64_t intval1, intval2;
4054 
4055                 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4056                 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4057                 return (intval1 == intval2);
4058         }
4059 }
4060 
4061 /*
4062  * Remove properties from props if they are not going to change (as determined
4063  * by comparison with origprops). Remove them from origprops as well, since we
4064  * do not need to clear or restore properties that won't change.
4065  */
4066 static void
4067 props_reduce(nvlist_t *props, nvlist_t *origprops)
4068 {
4069         nvpair_t *pair, *next_pair;
4070 
4071         if (origprops == NULL)
4072                 return; /* all props need to be received */
4073 
4074         pair = nvlist_next_nvpair(props, NULL);
4075         while (pair != NULL) {
4076                 const char *propname = nvpair_name(pair);
4077                 nvpair_t *match;
4078 
4079                 next_pair = nvlist_next_nvpair(props, pair);
4080 
4081                 if ((nvlist_lookup_nvpair(origprops, propname,
4082                     &match) != 0) || !propval_equals(pair, match))
4083                         goto next; /* need to set received value */
4084 
4085                 /* don't clear the existing received value */
4086                 (void) nvlist_remove_nvpair(origprops, match);
4087                 /* don't bother receiving the property */
4088                 (void) nvlist_remove_nvpair(props, pair);
4089 next:
4090                 pair = next_pair;
4091         }
4092 }
4093 
4094 /*
4095  * Extract properties that cannot be set PRIOR to the receipt of a dataset.
4096  * For example, refquota cannot be set until after the receipt of a dataset,
4097  * because in replication streams, an older/earlier snapshot may exceed the
4098  * refquota.  We want to receive the older/earlier snapshot, but setting
4099  * refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent
4100  * the older/earlier snapshot from being received (with EDQUOT).
4101  *
4102  * The ZFS test "zfs_receive_011_pos" demonstrates such a scenario.
4103  *
4104  * libzfs will need to be judicious handling errors encountered by props
4105  * extracted by this function.
4106  */
4107 static nvlist_t *
4108 extract_delay_props(nvlist_t *props)
4109 {
4110         nvlist_t *delayprops;
4111         nvpair_t *nvp, *tmp;
4112         static const zfs_prop_t delayable[] = { ZFS_PROP_REFQUOTA, 0 };
4113         int i;
4114 
4115         VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4116 
4117         for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL;
4118             nvp = nvlist_next_nvpair(props, nvp)) {
4119                 /*
4120                  * strcmp() is safe because zfs_prop_to_name() always returns
4121                  * a bounded string.
4122                  */
4123                 for (i = 0; delayable[i] != 0; i++) {
4124                         if (strcmp(zfs_prop_to_name(delayable[i]),
4125                             nvpair_name(nvp)) == 0) {
4126                                 break;
4127                         }
4128                 }
4129                 if (delayable[i] != 0) {
4130                         tmp = nvlist_prev_nvpair(props, nvp);
4131                         VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0);
4132                         VERIFY(nvlist_remove_nvpair(props, nvp) == 0);
4133                         nvp = tmp;
4134                 }
4135         }
4136 
4137         if (nvlist_empty(delayprops)) {
4138                 nvlist_free(delayprops);
4139                 delayprops = NULL;
4140         }
4141         return (delayprops);
4142 }
4143 
4144 #ifdef  DEBUG
4145 static boolean_t zfs_ioc_recv_inject_err;
4146 #endif
4147 
4148 /*
4149  * inputs:
4150  * zc_name              name of containing filesystem
4151  * zc_nvlist_src{_size} nvlist of properties to apply
4152  * zc_value             name of snapshot to create
4153  * zc_string            name of clone origin (if DRR_FLAG_CLONE)
4154  * zc_cookie            file descriptor to recv from
4155  * zc_begin_record      the BEGIN record of the stream (not byteswapped)
4156  * zc_guid              force flag
4157  * zc_cleanup_fd        cleanup-on-exit file descriptor
4158  * zc_action_handle     handle for this guid/ds mapping (or zero on first call)
4159  * zc_resumable         if data is incomplete assume sender will resume
4160  *
4161  * outputs:
4162  * zc_cookie            number of bytes read
4163  * zc_nvlist_dst{_size} error for each unapplied received property
4164  * zc_obj               zprop_errflags_t
4165  * zc_action_handle     handle for this guid/ds mapping
4166  */
4167 static int
4168 zfs_ioc_recv(zfs_cmd_t *zc)
4169 {
4170         file_t *fp;
4171         dmu_recv_cookie_t drc;
4172         boolean_t force = (boolean_t)zc->zc_guid;
4173         int fd;
4174         int error = 0;
4175         int props_error = 0;
4176         nvlist_t *errors;
4177         offset_t off;
4178         nvlist_t *props = NULL; /* sent properties */
4179         nvlist_t *origprops = NULL; /* existing properties */
4180         nvlist_t *delayprops = NULL; /* sent properties applied post-receive */
4181         char *origin = NULL;
4182         char *tosnap;
4183         char tofs[ZFS_MAXNAMELEN];
4184         boolean_t first_recvd_props = B_FALSE;
4185 
4186         if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4187             strchr(zc->zc_value, '@') == NULL ||
4188             strchr(zc->zc_value, '%'))
4189                 return (SET_ERROR(EINVAL));
4190 
4191         (void) strcpy(tofs, zc->zc_value);
4192         tosnap = strchr(tofs, '@');
4193         *tosnap++ = '\0';
4194 
4195         if (zc->zc_nvlist_src != NULL &&
4196             (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
4197             zc->zc_iflags, &props)) != 0)
4198                 return (error);
4199 
4200         fd = zc->zc_cookie;
4201         fp = getf(fd);
4202         if (fp == NULL) {
4203                 nvlist_free(props);
4204                 return (SET_ERROR(EBADF));
4205         }
4206 
4207         errors = fnvlist_alloc();
4208 
4209         if (zc->zc_string[0])
4210                 origin = zc->zc_string;
4211 
4212         error = dmu_recv_begin(tofs, tosnap,
4213             &zc->zc_begin_record, force, zc->zc_resumable, origin, &drc);
4214         if (error != 0)
4215                 goto out;
4216 
4217         /*
4218          * Set properties before we receive the stream so that they are applied
4219          * to the new data. Note that we must call dmu_recv_stream() if
4220          * dmu_recv_begin() succeeds.
4221          */
4222         if (props != NULL && !drc.drc_newfs) {
4223                 if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
4224                     SPA_VERSION_RECVD_PROPS &&
4225                     !dsl_prop_get_hasrecvd(tofs))
4226                         first_recvd_props = B_TRUE;
4227 
4228                 /*
4229                  * If new received properties are supplied, they are to
4230                  * completely replace the existing received properties, so stash
4231                  * away the existing ones.
4232                  */
4233                 if (dsl_prop_get_received(tofs, &origprops) == 0) {
4234                         nvlist_t *errlist = NULL;
4235                         /*
4236                          * Don't bother writing a property if its value won't
4237                          * change (and avoid the unnecessary security checks).
4238                          *
4239                          * The first receive after SPA_VERSION_RECVD_PROPS is a
4240                          * special case where we blow away all local properties
4241                          * regardless.
4242                          */
4243                         if (!first_recvd_props)
4244                                 props_reduce(props, origprops);
4245                         if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
4246                                 (void) nvlist_merge(errors, errlist, 0);
4247                         nvlist_free(errlist);
4248 
4249                         if (clear_received_props(tofs, origprops,
4250                             first_recvd_props ? NULL : props) != 0)
4251                                 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4252                 } else {
4253                         zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4254                 }
4255         }
4256 
4257         if (props != NULL) {
4258                 props_error = dsl_prop_set_hasrecvd(tofs);
4259 
4260                 if (props_error == 0) {
4261                         delayprops = extract_delay_props(props);
4262                         (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4263                             props, errors);
4264                 }
4265         }
4266 
4267         off = fp->f_offset;
4268         error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
4269             &zc->zc_action_handle);
4270 
4271         if (error == 0) {
4272                 zfsvfs_t *zfsvfs = NULL;
4273 
4274                 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4275                         /* online recv */
4276                         int end_err;
4277 
4278                         error = zfs_suspend_fs(zfsvfs);
4279                         /*
4280                          * If the suspend fails, then the recv_end will
4281                          * likely also fail, and clean up after itself.
4282                          */
4283                         end_err = dmu_recv_end(&drc, zfsvfs);
4284                         if (error == 0)
4285                                 error = zfs_resume_fs(zfsvfs, tofs);
4286                         error = error ? error : end_err;
4287                         VFS_RELE(zfsvfs->z_vfs);
4288                 } else {
4289                         error = dmu_recv_end(&drc, NULL);
4290                 }
4291 
4292                 /* Set delayed properties now, after we're done receiving. */
4293                 if (delayprops != NULL && error == 0) {
4294                         (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4295                             delayprops, errors);
4296                 }
4297         }
4298 
4299         if (delayprops != NULL) {
4300                 /*
4301                  * Merge delayed props back in with initial props, in case
4302                  * we're DEBUG and zfs_ioc_recv_inject_err is set (which means
4303                  * we have to make sure clear_received_props() includes
4304                  * the delayed properties).
4305                  *
4306                  * Since zfs_ioc_recv_inject_err is only in DEBUG kernels,
4307                  * using ASSERT() will be just like a VERIFY.
4308                  */
4309                 ASSERT(nvlist_merge(props, delayprops, 0) == 0);
4310                 nvlist_free(delayprops);
4311         }
4312 
4313         /*
4314          * Now that all props, initial and delayed, are set, report the prop
4315          * errors to the caller.
4316          */
4317         if (zc->zc_nvlist_dst_size != 0 &&
4318             (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4319             put_nvlist(zc, errors) != 0)) {
4320                 /*
4321                  * Caller made zc->zc_nvlist_dst less than the minimum expected
4322                  * size or supplied an invalid address.
4323                  */
4324                 props_error = SET_ERROR(EINVAL);
4325         }
4326 
4327         zc->zc_cookie = off - fp->f_offset;
4328         if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4329                 fp->f_offset = off;
4330 
4331 #ifdef  DEBUG
4332         if (zfs_ioc_recv_inject_err) {
4333                 zfs_ioc_recv_inject_err = B_FALSE;
4334                 error = 1;
4335         }
4336 #endif
4337         /*
4338          * On error, restore the original props.
4339          */
4340         if (error != 0 && props != NULL && !drc.drc_newfs) {
4341                 if (clear_received_props(tofs, props, NULL) != 0) {
4342                         /*
4343                          * We failed to clear the received properties.
4344                          * Since we may have left a $recvd value on the
4345                          * system, we can't clear the $hasrecvd flag.
4346                          */
4347                         zc->zc_obj |= ZPROP_ERR_NORESTORE;
4348                 } else if (first_recvd_props) {
4349                         dsl_prop_unset_hasrecvd(tofs);
4350                 }
4351 
4352                 if (origprops == NULL && !drc.drc_newfs) {
4353                         /* We failed to stash the original properties. */
4354                         zc->zc_obj |= ZPROP_ERR_NORESTORE;
4355                 }
4356 
4357                 /*
4358                  * dsl_props_set() will not convert RECEIVED to LOCAL on or
4359                  * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4360                  * explictly if we're restoring local properties cleared in the
4361                  * first new-style receive.
4362                  */
4363                 if (origprops != NULL &&
4364                     zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4365                     ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4366                     origprops, NULL) != 0) {
4367                         /*
4368                          * We stashed the original properties but failed to
4369                          * restore them.
4370                          */
4371                         zc->zc_obj |= ZPROP_ERR_NORESTORE;
4372                 }
4373         }
4374 out:
4375         nvlist_free(props);
4376         nvlist_free(origprops);
4377         nvlist_free(errors);
4378         releasef(fd);
4379 
4380         if (error == 0)
4381                 error = props_error;
4382 
4383         return (error);
4384 }
4385 
4386 /*
4387  * inputs:
4388  * zc_name      name of snapshot to send
4389  * zc_cookie    file descriptor to send stream to
4390  * zc_obj       fromorigin flag (mutually exclusive with zc_fromobj)
4391  * zc_sendobj   objsetid of snapshot to send
4392  * zc_fromobj   objsetid of incremental fromsnap (may be zero)
4393  * zc_guid      if set, estimate size of stream only.  zc_cookie is ignored.
4394  *              output size in zc_objset_type.
4395  * zc_flags     lzc_send_flags
4396  *
4397  * outputs:
4398  * zc_objset_type       estimated size, if zc_guid is set
4399  */
4400 static int
4401 zfs_ioc_send(zfs_cmd_t *zc)
4402 {
4403         int error;
4404         offset_t off;
4405         boolean_t estimate = (zc->zc_guid != 0);
4406         boolean_t embedok = (zc->zc_flags & 0x1);
4407         boolean_t large_block_ok = (zc->zc_flags & 0x2);
4408 
4409         if (zc->zc_obj != 0) {
4410                 dsl_pool_t *dp;
4411                 dsl_dataset_t *tosnap;
4412 
4413                 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4414                 if (error != 0)
4415                         return (error);
4416 
4417                 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4418                 if (error != 0) {
4419                         dsl_pool_rele(dp, FTAG);
4420                         return (error);
4421                 }
4422 
4423                 if (dsl_dir_is_clone(tosnap->ds_dir))
4424                         zc->zc_fromobj =
4425                             dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj;
4426                 dsl_dataset_rele(tosnap, FTAG);
4427                 dsl_pool_rele(dp, FTAG);
4428         }
4429 
4430         if (estimate) {
4431                 dsl_pool_t *dp;
4432                 dsl_dataset_t *tosnap;
4433                 dsl_dataset_t *fromsnap = NULL;
4434 
4435                 error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4436                 if (error != 0)
4437                         return (error);
4438 
4439                 error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4440                 if (error != 0) {
4441                         dsl_pool_rele(dp, FTAG);
4442                         return (error);
4443                 }
4444 
4445                 if (zc->zc_fromobj != 0) {
4446                         error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4447                             FTAG, &fromsnap);
4448                         if (error != 0) {
4449                                 dsl_dataset_rele(tosnap, FTAG);
4450                                 dsl_pool_rele(dp, FTAG);
4451                                 return (error);
4452                         }
4453                 }
4454 
4455                 error = dmu_send_estimate(tosnap, fromsnap,
4456                     &zc->zc_objset_type);
4457 
4458                 if (fromsnap != NULL)
4459                         dsl_dataset_rele(fromsnap, FTAG);
4460                 dsl_dataset_rele(tosnap, FTAG);
4461                 dsl_pool_rele(dp, FTAG);
4462         } else {
4463                 file_t *fp = getf(zc->zc_cookie);
4464                 if (fp == NULL)
4465                         return (SET_ERROR(EBADF));
4466 
4467                 off = fp->f_offset;
4468                 error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4469                     zc->zc_fromobj, embedok, large_block_ok,
4470                     zc->zc_cookie, fp->f_vnode, &off);
4471 
4472                 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4473                         fp->f_offset = off;
4474                 releasef(zc->zc_cookie);
4475         }
4476         return (error);
4477 }
4478 
4479 /*
4480  * inputs:
4481  * zc_name      name of snapshot on which to report progress
4482  * zc_cookie    file descriptor of send stream
4483  *
4484  * outputs:
4485  * zc_cookie    number of bytes written in send stream thus far
4486  */
4487 static int
4488 zfs_ioc_send_progress(zfs_cmd_t *zc)
4489 {
4490         dsl_pool_t *dp;
4491         dsl_dataset_t *ds;
4492         dmu_sendarg_t *dsp = NULL;
4493         int error;
4494 
4495         error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4496         if (error != 0)
4497                 return (error);
4498 
4499         error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4500         if (error != 0) {
4501                 dsl_pool_rele(dp, FTAG);
4502                 return (error);
4503         }
4504 
4505         mutex_enter(&ds->ds_sendstream_lock);
4506 
4507         /*
4508          * Iterate over all the send streams currently active on this dataset.
4509          * If there's one which matches the specified file descriptor _and_ the
4510          * stream was started by the current process, return the progress of
4511          * that stream.
4512          */
4513         for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4514             dsp = list_next(&ds->ds_sendstreams, dsp)) {
4515                 if (dsp->dsa_outfd == zc->zc_cookie &&
4516                     dsp->dsa_proc == curproc)
4517                         break;
4518         }
4519 
4520         if (dsp != NULL)
4521                 zc->zc_cookie = *(dsp->dsa_off);
4522         else
4523                 error = SET_ERROR(ENOENT);
4524 
4525         mutex_exit(&ds->ds_sendstream_lock);
4526         dsl_dataset_rele(ds, FTAG);
4527         dsl_pool_rele(dp, FTAG);
4528         return (error);
4529 }
4530 
4531 static int
4532 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4533 {
4534         int id, error;
4535 
4536         error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4537             &zc->zc_inject_record);
4538 
4539         if (error == 0)
4540                 zc->zc_guid = (uint64_t)id;
4541 
4542         return (error);
4543 }
4544 
4545 static int
4546 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4547 {
4548         return (zio_clear_fault((int)zc->zc_guid));
4549 }
4550 
4551 static int
4552 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4553 {
4554         int id = (int)zc->zc_guid;
4555         int error;
4556 
4557         error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4558             &zc->zc_inject_record);
4559 
4560         zc->zc_guid = id;
4561 
4562         return (error);
4563 }
4564 
4565 static int
4566 zfs_ioc_error_log(zfs_cmd_t *zc)
4567 {
4568         spa_t *spa;
4569         int error;
4570         size_t count = (size_t)zc->zc_nvlist_dst_size;
4571 
4572         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4573                 return (error);
4574 
4575         error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4576             &count);
4577         if (error == 0)
4578                 zc->zc_nvlist_dst_size = count;
4579         else
4580                 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4581 
4582         spa_close(spa, FTAG);
4583 
4584         return (error);
4585 }
4586 
4587 static int
4588 zfs_ioc_clear(zfs_cmd_t *zc)
4589 {
4590         spa_t *spa;
4591         vdev_t *vd;
4592         int error;
4593 
4594         /*
4595          * On zpool clear we also fix up missing slogs
4596          */
4597         mutex_enter(&spa_namespace_lock);
4598         spa = spa_lookup(zc->zc_name);
4599         if (spa == NULL) {
4600                 mutex_exit(&spa_namespace_lock);
4601                 return (SET_ERROR(EIO));
4602         }
4603         if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4604                 /* we need to let spa_open/spa_load clear the chains */
4605                 spa_set_log_state(spa, SPA_LOG_CLEAR);
4606         }
4607         spa->spa_last_open_failed = 0;
4608         mutex_exit(&spa_namespace_lock);
4609 
4610         if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4611                 error = spa_open(zc->zc_name, &spa, FTAG);
4612         } else {
4613                 nvlist_t *policy;
4614                 nvlist_t *config = NULL;
4615 
4616                 if (zc->zc_nvlist_src == NULL)
4617                         return (SET_ERROR(EINVAL));
4618 
4619                 if ((error = get_nvlist(zc->zc_nvlist_src,
4620                     zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4621                         error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4622                             policy, &config);
4623                         if (config != NULL) {
4624                                 int err;
4625 
4626                                 if ((err = put_nvlist(zc, config)) != 0)
4627                                         error = err;
4628                                 nvlist_free(config);
4629                         }
4630                         nvlist_free(policy);
4631                 }
4632         }
4633 
4634         if (error != 0)
4635                 return (error);
4636 
4637         spa_vdev_state_enter(spa, SCL_NONE);
4638 
4639         if (zc->zc_guid == 0) {
4640                 vd = NULL;
4641         } else {
4642                 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4643                 if (vd == NULL) {
4644                         (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4645                         spa_close(spa, FTAG);
4646                         return (SET_ERROR(ENODEV));
4647                 }
4648         }
4649 
4650         vdev_clear(spa, vd);
4651 
4652         (void) spa_vdev_state_exit(spa, NULL, 0);
4653 
4654         /*
4655          * Resume any suspended I/Os.
4656          */
4657         if (zio_resume(spa) != 0)
4658                 error = SET_ERROR(EIO);
4659 
4660         spa_close(spa, FTAG);
4661 
4662         return (error);
4663 }
4664 
4665 static int
4666 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4667 {
4668         spa_t *spa;
4669         int error;
4670 
4671         error = spa_open(zc->zc_name, &spa, FTAG);
4672         if (error != 0)
4673                 return (error);
4674 
4675         spa_vdev_state_enter(spa, SCL_NONE);
4676 
4677         /*
4678          * If a resilver is already in progress then set the
4679          * spa_scrub_reopen flag to B_TRUE so that we don't restart
4680          * the scan as a side effect of the reopen. Otherwise, let
4681          * vdev_open() decided if a resilver is required.
4682          */
4683         spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4684         vdev_reopen(spa->spa_root_vdev);
4685         spa->spa_scrub_reopen = B_FALSE;
4686 
4687         (void) spa_vdev_state_exit(spa, NULL, 0);
4688         spa_close(spa, FTAG);
4689         return (0);
4690 }
4691 /*
4692  * inputs:
4693  * zc_name      name of filesystem
4694  * zc_value     name of origin snapshot
4695  *
4696  * outputs:
4697  * zc_string    name of conflicting snapshot, if there is one
4698  */
4699 static int
4700 zfs_ioc_promote(zfs_cmd_t *zc)
4701 {
4702         char *cp;
4703 
4704         /*
4705          * We don't need to unmount *all* the origin fs's snapshots, but
4706          * it's easier.
4707          */
4708         cp = strchr(zc->zc_value, '@');
4709         if (cp)
4710                 *cp = '\0';
4711         (void) dmu_objset_find(zc->zc_value,
4712             zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
4713         return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4714 }
4715 
4716 /*
4717  * Retrieve a single {user|group}{used|quota}@... property.
4718  *
4719  * inputs:
4720  * zc_name      name of filesystem
4721  * zc_objset_type zfs_userquota_prop_t
4722  * zc_value     domain name (eg. "S-1-234-567-89")
4723  * zc_guid      RID/UID/GID
4724  *
4725  * outputs:
4726  * zc_cookie    property value
4727  */
4728 static int
4729 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4730 {
4731         zfsvfs_t *zfsvfs;
4732         int error;
4733 
4734         if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4735                 return (SET_ERROR(EINVAL));
4736 
4737         error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4738         if (error != 0)
4739                 return (error);
4740 
4741         error = zfs_userspace_one(zfsvfs,
4742             zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4743         zfsvfs_rele(zfsvfs, FTAG);
4744 
4745         return (error);
4746 }
4747 
4748 /*
4749  * inputs:
4750  * zc_name              name of filesystem
4751  * zc_cookie            zap cursor
4752  * zc_objset_type       zfs_userquota_prop_t
4753  * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4754  *
4755  * outputs:
4756  * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4757  * zc_cookie    zap cursor
4758  */
4759 static int
4760 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4761 {
4762         zfsvfs_t *zfsvfs;
4763         int bufsize = zc->zc_nvlist_dst_size;
4764 
4765         if (bufsize <= 0)
4766                 return (SET_ERROR(ENOMEM));
4767 
4768         int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4769         if (error != 0)
4770                 return (error);
4771 
4772         void *buf = kmem_alloc(bufsize, KM_SLEEP);
4773 
4774         error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
4775             buf, &zc->zc_nvlist_dst_size);
4776 
4777         if (error == 0) {
4778                 error = xcopyout(buf,
4779                     (void *)(uintptr_t)zc->zc_nvlist_dst,
4780                     zc->zc_nvlist_dst_size);
4781         }
4782         kmem_free(buf, bufsize);
4783         zfsvfs_rele(zfsvfs, FTAG);
4784 
4785         return (error);
4786 }
4787 
4788 /*
4789  * inputs:
4790  * zc_name              name of filesystem
4791  *
4792  * outputs:
4793  * none
4794  */
4795 static int
4796 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4797 {
4798         objset_t *os;
4799         int error = 0;
4800         zfsvfs_t *zfsvfs;
4801 
4802         if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
4803                 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
4804                         /*
4805                          * If userused is not enabled, it may be because the
4806                          * objset needs to be closed & reopened (to grow the
4807                          * objset_phys_t).  Suspend/resume the fs will do that.
4808                          */
4809                         error = zfs_suspend_fs(zfsvfs);
4810                         if (error == 0) {
4811                                 dmu_objset_refresh_ownership(zfsvfs->z_os,
4812                                     zfsvfs);
4813                                 error = zfs_resume_fs(zfsvfs, zc->zc_name);
4814                         }
4815                 }
4816                 if (error == 0)
4817                         error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
4818                 VFS_RELE(zfsvfs->z_vfs);
4819         } else {
4820                 /* XXX kind of reading contents without owning */
4821                 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4822                 if (error != 0)
4823                         return (error);
4824 
4825                 error = dmu_objset_userspace_upgrade(os);
4826                 dmu_objset_rele(os, FTAG);
4827         }
4828 
4829         return (error);
4830 }
4831 
4832 /*
4833  * We don't want to have a hard dependency
4834  * against some special symbols in sharefs
4835  * nfs, and smbsrv.  Determine them if needed when
4836  * the first file system is shared.
4837  * Neither sharefs, nfs or smbsrv are unloadable modules.
4838  */
4839 int (*znfsexport_fs)(void *arg);
4840 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
4841 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
4842 
4843 int zfs_nfsshare_inited;
4844 int zfs_smbshare_inited;
4845 
4846 ddi_modhandle_t nfs_mod;
4847 ddi_modhandle_t sharefs_mod;
4848 ddi_modhandle_t smbsrv_mod;
4849 kmutex_t zfs_share_lock;
4850 
4851 static int
4852 zfs_init_sharefs()
4853 {
4854         int error;
4855 
4856         ASSERT(MUTEX_HELD(&zfs_share_lock));
4857         /* Both NFS and SMB shares also require sharetab support. */
4858         if (sharefs_mod == NULL && ((sharefs_mod =
4859             ddi_modopen("fs/sharefs",
4860             KRTLD_MODE_FIRST, &error)) == NULL)) {
4861                 return (SET_ERROR(ENOSYS));
4862         }
4863         if (zshare_fs == NULL && ((zshare_fs =
4864             (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
4865             ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
4866                 return (SET_ERROR(ENOSYS));
4867         }
4868         return (0);
4869 }
4870 
4871 static int
4872 zfs_ioc_share(zfs_cmd_t *zc)
4873 {
4874         int error;
4875         int opcode;
4876 
4877         switch (zc->zc_share.z_sharetype) {
4878         case ZFS_SHARE_NFS:
4879         case ZFS_UNSHARE_NFS:
4880                 if (zfs_nfsshare_inited == 0) {
4881                         mutex_enter(&zfs_share_lock);
4882                         if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
4883                             KRTLD_MODE_FIRST, &error)) == NULL)) {
4884                                 mutex_exit(&zfs_share_lock);
4885                                 return (SET_ERROR(ENOSYS));
4886                         }
4887                         if (znfsexport_fs == NULL &&
4888                             ((znfsexport_fs = (int (*)(void *))
4889                             ddi_modsym(nfs_mod,
4890                             "nfs_export", &error)) == NULL)) {
4891                                 mutex_exit(&zfs_share_lock);
4892                                 return (SET_ERROR(ENOSYS));
4893                         }
4894                         error = zfs_init_sharefs();
4895                         if (error != 0) {
4896                                 mutex_exit(&zfs_share_lock);
4897                                 return (SET_ERROR(ENOSYS));
4898                         }
4899                         zfs_nfsshare_inited = 1;
4900                         mutex_exit(&zfs_share_lock);
4901                 }
4902                 break;
4903         case ZFS_SHARE_SMB:
4904         case ZFS_UNSHARE_SMB:
4905                 if (zfs_smbshare_inited == 0) {
4906                         mutex_enter(&zfs_share_lock);
4907                         if (smbsrv_mod == NULL && ((smbsrv_mod =
4908                             ddi_modopen("drv/smbsrv",
4909                             KRTLD_MODE_FIRST, &error)) == NULL)) {
4910                                 mutex_exit(&zfs_share_lock);
4911                                 return (SET_ERROR(ENOSYS));
4912                         }
4913                         if (zsmbexport_fs == NULL && ((zsmbexport_fs =
4914                             (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
4915                             "smb_server_share", &error)) == NULL)) {
4916                                 mutex_exit(&zfs_share_lock);
4917                                 return (SET_ERROR(ENOSYS));
4918                         }
4919                         error = zfs_init_sharefs();
4920                         if (error != 0) {
4921                                 mutex_exit(&zfs_share_lock);
4922                                 return (SET_ERROR(ENOSYS));
4923                         }
4924                         zfs_smbshare_inited = 1;
4925                         mutex_exit(&zfs_share_lock);
4926                 }
4927                 break;
4928         default:
4929                 return (SET_ERROR(EINVAL));
4930         }
4931 
4932         switch (zc->zc_share.z_sharetype) {
4933         case ZFS_SHARE_NFS:
4934         case ZFS_UNSHARE_NFS:
4935                 if (error =
4936                     znfsexport_fs((void *)
4937                     (uintptr_t)zc->zc_share.z_exportdata))
4938                         return (error);
4939                 break;
4940         case ZFS_SHARE_SMB:
4941         case ZFS_UNSHARE_SMB:
4942                 if (error = zsmbexport_fs((void *)
4943                     (uintptr_t)zc->zc_share.z_exportdata,
4944                     zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
4945                     B_TRUE: B_FALSE)) {
4946                         return (error);
4947                 }
4948                 break;
4949         }
4950 
4951         opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
4952             zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
4953             SHAREFS_ADD : SHAREFS_REMOVE;
4954 
4955         /*
4956          * Add or remove share from sharetab
4957          */
4958         error = zshare_fs(opcode,
4959             (void *)(uintptr_t)zc->zc_share.z_sharedata,
4960             zc->zc_share.z_sharemax);
4961 
4962         return (error);
4963 
4964 }
4965 
4966 ace_t full_access[] = {
4967         {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4968 };
4969 
4970 /*
4971  * inputs:
4972  * zc_name              name of containing filesystem
4973  * zc_obj               object # beyond which we want next in-use object #
4974  *
4975  * outputs:
4976  * zc_obj               next in-use object #
4977  */
4978 static int
4979 zfs_ioc_next_obj(zfs_cmd_t *zc)
4980 {
4981         objset_t *os = NULL;
4982         int error;
4983 
4984         error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4985         if (error != 0)
4986                 return (error);
4987 
4988         error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4989             dsl_dataset_phys(os->os_dsl_dataset)->ds_prev_snap_txg);
4990 
4991         dmu_objset_rele(os, FTAG);
4992         return (error);
4993 }
4994 
4995 /*
4996  * inputs:
4997  * zc_name              name of filesystem
4998  * zc_value             prefix name for snapshot
4999  * zc_cleanup_fd        cleanup-on-exit file descriptor for calling process
5000  *
5001  * outputs:
5002  * zc_value             short name of new snapshot
5003  */
5004 static int
5005 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
5006 {
5007         char *snap_name;
5008         char *hold_name;
5009         int error;
5010         minor_t minor;
5011 
5012         error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
5013         if (error != 0)
5014                 return (error);
5015 
5016         snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
5017             (u_longlong_t)ddi_get_lbolt64());
5018         hold_name = kmem_asprintf("%%%s", zc->zc_value);
5019 
5020         error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
5021             hold_name);
5022         if (error == 0)
5023                 (void) strcpy(zc->zc_value, snap_name);
5024         strfree(snap_name);
5025         strfree(hold_name);
5026         zfs_onexit_fd_rele(zc->zc_cleanup_fd);
5027         return (error);
5028 }
5029 
5030 /*
5031  * inputs:
5032  * zc_name              name of "to" snapshot
5033  * zc_value             name of "from" snapshot
5034  * zc_cookie            file descriptor to write diff data on
5035  *
5036  * outputs:
5037  * dmu_diff_record_t's to the file descriptor
5038  */
5039 static int
5040 zfs_ioc_diff(zfs_cmd_t *zc)
5041 {
5042         file_t *fp;
5043         offset_t off;
5044         int error;
5045 
5046         fp = getf(zc->zc_cookie);
5047         if (fp == NULL)
5048                 return (SET_ERROR(EBADF));
5049 
5050         off = fp->f_offset;
5051 
5052         error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
5053 
5054         if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5055                 fp->f_offset = off;
5056         releasef(zc->zc_cookie);
5057 
5058         return (error);
5059 }
5060 
5061 /*
5062  * Remove all ACL files in shares dir
5063  */
5064 static int
5065 zfs_smb_acl_purge(znode_t *dzp)
5066 {
5067         zap_cursor_t    zc;
5068         zap_attribute_t zap;
5069         zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
5070         int error;
5071 
5072         for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
5073             (error = zap_cursor_retrieve(&zc, &zap)) == 0;
5074             zap_cursor_advance(&zc)) {
5075                 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
5076                     NULL, 0)) != 0)
5077                         break;
5078         }
5079         zap_cursor_fini(&zc);
5080         return (error);
5081 }
5082 
5083 static int
5084 zfs_ioc_smb_acl(zfs_cmd_t *zc)
5085 {
5086         vnode_t *vp;
5087         znode_t *dzp;
5088         vnode_t *resourcevp = NULL;
5089         znode_t *sharedir;
5090         zfsvfs_t *zfsvfs;
5091         nvlist_t *nvlist;
5092         char *src, *target;
5093         vattr_t vattr;
5094         vsecattr_t vsec;
5095         int error = 0;
5096 
5097         if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
5098             NO_FOLLOW, NULL, &vp)) != 0)
5099                 return (error);
5100 
5101         /* Now make sure mntpnt and dataset are ZFS */
5102 
5103         if (vp->v_vfsp->vfs_fstype != zfsfstype ||
5104             (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
5105             zc->zc_name) != 0)) {
5106                 VN_RELE(vp);
5107                 return (SET_ERROR(EINVAL));
5108         }
5109 
5110         dzp = VTOZ(vp);
5111         zfsvfs = dzp->z_zfsvfs;
5112         ZFS_ENTER(zfsvfs);
5113 
5114         /*
5115          * Create share dir if its missing.
5116          */
5117         mutex_enter(&zfsvfs->z_lock);
5118         if (zfsvfs->z_shares_dir == 0) {
5119                 dmu_tx_t *tx;
5120 
5121                 tx = dmu_tx_create(zfsvfs->z_os);
5122                 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
5123                     ZFS_SHARES_DIR);
5124                 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
5125                 error = dmu_tx_assign(tx, TXG_WAIT);
5126                 if (error != 0) {
5127                         dmu_tx_abort(tx);
5128                 } else {
5129                         error = zfs_create_share_dir(zfsvfs, tx);
5130                         dmu_tx_commit(tx);
5131                 }
5132                 if (error != 0) {
5133                         mutex_exit(&zfsvfs->z_lock);
5134                         VN_RELE(vp);
5135                         ZFS_EXIT(zfsvfs);
5136                         return (error);
5137                 }
5138         }
5139         mutex_exit(&zfsvfs->z_lock);
5140 
5141         ASSERT(zfsvfs->z_shares_dir);
5142         if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
5143                 VN_RELE(vp);
5144                 ZFS_EXIT(zfsvfs);
5145                 return (error);
5146         }
5147 
5148         switch (zc->zc_cookie) {
5149         case ZFS_SMB_ACL_ADD:
5150                 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
5151                 vattr.va_type = VREG;
5152                 vattr.va_mode = S_IFREG|0777;
5153                 vattr.va_uid = 0;
5154                 vattr.va_gid = 0;
5155 
5156                 vsec.vsa_mask = VSA_ACE;
5157                 vsec.vsa_aclentp = &full_access;
5158                 vsec.vsa_aclentsz = sizeof (full_access);
5159                 vsec.vsa_aclcnt = 1;
5160 
5161                 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
5162                     &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
5163                 if (resourcevp)
5164                         VN_RELE(resourcevp);
5165                 break;
5166 
5167         case ZFS_SMB_ACL_REMOVE:
5168                 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
5169                     NULL, 0);
5170                 break;
5171 
5172         case ZFS_SMB_ACL_RENAME:
5173                 if ((error = get_nvlist(zc->zc_nvlist_src,
5174                     zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
5175                         VN_RELE(vp);
5176                         VN_RELE(ZTOV(sharedir));
5177                         ZFS_EXIT(zfsvfs);
5178                         return (error);
5179                 }
5180                 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
5181                     nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
5182                     &target)) {
5183                         VN_RELE(vp);
5184                         VN_RELE(ZTOV(sharedir));
5185                         ZFS_EXIT(zfsvfs);
5186                         nvlist_free(nvlist);
5187                         return (error);
5188                 }
5189                 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
5190                     kcred, NULL, 0);
5191                 nvlist_free(nvlist);
5192                 break;
5193 
5194         case ZFS_SMB_ACL_PURGE:
5195                 error = zfs_smb_acl_purge(sharedir);
5196                 break;
5197 
5198         default:
5199                 error = SET_ERROR(EINVAL);
5200                 break;
5201         }
5202 
5203         VN_RELE(vp);
5204         VN_RELE(ZTOV(sharedir));
5205 
5206         ZFS_EXIT(zfsvfs);
5207 
5208         return (error);
5209 }
5210 
5211 /*
5212  * innvl: {
5213  *     "holds" -> { snapname -> holdname (string), ... }
5214  *     (optional) "cleanup_fd" -> fd (int32)
5215  * }
5216  *
5217  * outnvl: {
5218  *     snapname -> error value (int32)
5219  *     ...
5220  * }
5221  */
5222 /* ARGSUSED */
5223 static int
5224 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
5225 {
5226         nvpair_t *pair;
5227         nvlist_t *holds;
5228         int cleanup_fd = -1;
5229         int error;
5230         minor_t minor = 0;
5231 
5232         error = nvlist_lookup_nvlist(args, "holds", &holds);
5233         if (error != 0)
5234                 return (SET_ERROR(EINVAL));
5235 
5236         /* make sure the user didn't pass us any invalid (empty) tags */
5237         for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
5238             pair = nvlist_next_nvpair(holds, pair)) {
5239                 char *htag;
5240 
5241                 error = nvpair_value_string(pair, &htag);
5242                 if (error != 0)
5243                         return (SET_ERROR(error));
5244 
5245                 if (strlen(htag) == 0)
5246                         return (SET_ERROR(EINVAL));
5247         }
5248 
5249         if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
5250                 error = zfs_onexit_fd_hold(cleanup_fd, &minor);
5251                 if (error != 0)
5252                         return (error);
5253         }
5254 
5255         error = dsl_dataset_user_hold(holds, minor, errlist);
5256         if (minor != 0)
5257                 zfs_onexit_fd_rele(cleanup_fd);
5258         return (error);
5259 }
5260 
5261 /*
5262  * innvl is not used.
5263  *
5264  * outnvl: {
5265  *    holdname -> time added (uint64 seconds since epoch)
5266  *    ...
5267  * }
5268  */
5269 /* ARGSUSED */
5270 static int
5271 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
5272 {
5273         return (dsl_dataset_get_holds(snapname, outnvl));
5274 }
5275 
5276 /*
5277  * innvl: {
5278  *     snapname -> { holdname, ... }
5279  *     ...
5280  * }
5281  *
5282  * outnvl: {
5283  *     snapname -> error value (int32)
5284  *     ...
5285  * }
5286  */
5287 /* ARGSUSED */
5288 static int
5289 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
5290 {
5291         return (dsl_dataset_user_release(holds, errlist));
5292 }
5293 
5294 /*
5295  * inputs:
5296  * zc_name              name of new filesystem or snapshot
5297  * zc_value             full name of old snapshot
5298  *
5299  * outputs:
5300  * zc_cookie            space in bytes
5301  * zc_objset_type       compressed space in bytes
5302  * zc_perm_action       uncompressed space in bytes
5303  */
5304 static int
5305 zfs_ioc_space_written(zfs_cmd_t *zc)
5306 {
5307         int error;
5308         dsl_pool_t *dp;
5309         dsl_dataset_t *new, *old;
5310 
5311         error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5312         if (error != 0)
5313                 return (error);
5314         error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5315         if (error != 0) {
5316                 dsl_pool_rele(dp, FTAG);
5317                 return (error);
5318         }
5319         error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5320         if (error != 0) {
5321                 dsl_dataset_rele(new, FTAG);
5322                 dsl_pool_rele(dp, FTAG);
5323                 return (error);
5324         }
5325 
5326         error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5327             &zc->zc_objset_type, &zc->zc_perm_action);
5328         dsl_dataset_rele(old, FTAG);
5329         dsl_dataset_rele(new, FTAG);
5330         dsl_pool_rele(dp, FTAG);
5331         return (error);
5332 }
5333 
5334 /*
5335  * innvl: {
5336  *     "firstsnap" -> snapshot name
5337  * }
5338  *
5339  * outnvl: {
5340  *     "used" -> space in bytes
5341  *     "compressed" -> compressed space in bytes
5342  *     "uncompressed" -> uncompressed space in bytes
5343  * }
5344  */
5345 static int
5346 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5347 {
5348         int error;
5349         dsl_pool_t *dp;
5350         dsl_dataset_t *new, *old;
5351         char *firstsnap;
5352         uint64_t used, comp, uncomp;
5353 
5354         if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5355                 return (SET_ERROR(EINVAL));
5356 
5357         error = dsl_pool_hold(lastsnap, FTAG, &dp);
5358         if (error != 0)
5359                 return (error);
5360 
5361         error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5362         if (error == 0 && !new->ds_is_snapshot) {
5363                 dsl_dataset_rele(new, FTAG);
5364                 error = SET_ERROR(EINVAL);
5365         }
5366         if (error != 0) {
5367                 dsl_pool_rele(dp, FTAG);
5368                 return (error);
5369         }
5370         error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5371         if (error == 0 && !old->ds_is_snapshot) {
5372                 dsl_dataset_rele(old, FTAG);
5373                 error = SET_ERROR(EINVAL);
5374         }
5375         if (error != 0) {
5376                 dsl_dataset_rele(new, FTAG);
5377                 dsl_pool_rele(dp, FTAG);
5378                 return (error);
5379         }
5380 
5381         error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5382         dsl_dataset_rele(old, FTAG);
5383         dsl_dataset_rele(new, FTAG);
5384         dsl_pool_rele(dp, FTAG);
5385         fnvlist_add_uint64(outnvl, "used", used);
5386         fnvlist_add_uint64(outnvl, "compressed", comp);
5387         fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5388         return (error);
5389 }
5390 
5391 /*
5392  * innvl: {
5393  *     "fd" -> file descriptor to write stream to (int32)
5394  *     (optional) "fromsnap" -> full snap name to send an incremental from
5395  *     (optional) "largeblockok" -> (value ignored)
5396  *         indicates that blocks > 128KB are permitted
5397  *     (optional) "embedok" -> (value ignored)
5398  *         presence indicates DRR_WRITE_EMBEDDED records are permitted
5399  *     (optional) "resume_object" and "resume_offset" -> (uint64)
5400  *         if present, resume send stream from specified object and offset.
5401  * }
5402  *
5403  * outnvl is unused
5404  */
5405 /* ARGSUSED */
5406 static int
5407 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5408 {
5409         int error;
5410         offset_t off;
5411         char *fromname = NULL;
5412         int fd;
5413         boolean_t largeblockok;
5414         boolean_t embedok;
5415         uint64_t resumeobj = 0;
5416         uint64_t resumeoff = 0;
5417 
5418         error = nvlist_lookup_int32(innvl, "fd", &fd);
5419         if (error != 0)
5420                 return (SET_ERROR(EINVAL));
5421 
5422         (void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5423 
5424         largeblockok = nvlist_exists(innvl, "largeblockok");
5425         embedok = nvlist_exists(innvl, "embedok");
5426 
5427         (void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
5428         (void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);
5429 
5430         file_t *fp = getf(fd);
5431         if (fp == NULL)
5432                 return (SET_ERROR(EBADF));
5433 
5434         off = fp->f_offset;
5435         error = dmu_send(snapname, fromname, embedok, largeblockok, fd,
5436             resumeobj, resumeoff, fp->f_vnode, &off);
5437 
5438         if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5439                 fp->f_offset = off;
5440         releasef(fd);
5441         return (error);
5442 }
5443 
5444 /*
5445  * Determine approximately how large a zfs send stream will be -- the number
5446  * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5447  *
5448  * innvl: {
5449  *     (optional) "from" -> full snap or bookmark name to send an incremental
5450  *                          from
5451  * }
5452  *
5453  * outnvl: {
5454  *     "space" -> bytes of space (uint64)
5455  * }
5456  */
5457 static int
5458 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5459 {
5460         dsl_pool_t *dp;
5461         dsl_dataset_t *tosnap;
5462         int error;
5463         char *fromname;
5464         uint64_t space;
5465 
5466         error = dsl_pool_hold(snapname, FTAG, &dp);
5467         if (error != 0)
5468                 return (error);
5469 
5470         error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5471         if (error != 0) {
5472                 dsl_pool_rele(dp, FTAG);
5473                 return (error);
5474         }
5475 
5476         error = nvlist_lookup_string(innvl, "from", &fromname);
5477         if (error == 0) {
5478                 if (strchr(fromname, '@') != NULL) {
5479                         /*
5480                          * If from is a snapshot, hold it and use the more
5481                          * efficient dmu_send_estimate to estimate send space
5482                          * size using deadlists.
5483                          */
5484                         dsl_dataset_t *fromsnap;
5485                         error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5486                         if (error != 0)
5487                                 goto out;
5488                         error = dmu_send_estimate(tosnap, fromsnap, &space);
5489                         dsl_dataset_rele(fromsnap, FTAG);
5490                 } else if (strchr(fromname, '#') != NULL) {
5491                         /*
5492                          * If from is a bookmark, fetch the creation TXG of the
5493                          * snapshot it was created from and use that to find
5494                          * blocks that were born after it.
5495                          */
5496                         zfs_bookmark_phys_t frombm;
5497 
5498                         error = dsl_bookmark_lookup(dp, fromname, tosnap,
5499                             &frombm);
5500                         if (error != 0)
5501                                 goto out;
5502                         error = dmu_send_estimate_from_txg(tosnap,
5503                             frombm.zbm_creation_txg, &space);
5504                 } else {
5505                         /*
5506                          * from is not properly formatted as a snapshot or
5507                          * bookmark
5508                          */
5509                         error = SET_ERROR(EINVAL);
5510                         goto out;
5511                 }
5512         } else {
5513                 // If estimating the size of a full send, use dmu_send_estimate
5514                 error = dmu_send_estimate(tosnap, NULL, &space);
5515         }
5516 
5517         fnvlist_add_uint64(outnvl, "space", space);
5518 
5519 out:
5520         dsl_dataset_rele(tosnap, FTAG);
5521         dsl_pool_rele(dp, FTAG);
5522         return (error);
5523 }
5524 
5525 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5526 
5527 static void
5528 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5529     zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5530     boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5531 {
5532         zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5533 
5534         ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5535         ASSERT3U(ioc, <, ZFS_IOC_LAST);
5536         ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5537         ASSERT3P(vec->zvec_func, ==, NULL);
5538 
5539         vec->zvec_legacy_func = func;
5540         vec->zvec_secpolicy = secpolicy;
5541         vec->zvec_namecheck = namecheck;
5542         vec->zvec_allow_log = log_history;
5543         vec->zvec_pool_check = pool_check;
5544 }
5545 
5546 /*
5547  * See the block comment at the beginning of this file for details on
5548  * each argument to this function.
5549  */
5550 static void
5551 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5552     zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5553     zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5554     boolean_t allow_log)
5555 {
5556         zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5557 
5558         ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5559         ASSERT3U(ioc, <, ZFS_IOC_LAST);
5560         ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5561         ASSERT3P(vec->zvec_func, ==, NULL);
5562 
5563         /* if we are logging, the name must be valid */
5564         ASSERT(!allow_log || namecheck != NO_NAME);
5565 
5566         vec->zvec_name = name;
5567         vec->zvec_func = func;
5568         vec->zvec_secpolicy = secpolicy;
5569         vec->zvec_namecheck = namecheck;
5570         vec->zvec_pool_check = pool_check;
5571         vec->zvec_smush_outnvlist = smush_outnvlist;
5572         vec->zvec_allow_log = allow_log;
5573 }
5574 
5575 static void
5576 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5577     zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5578     zfs_ioc_poolcheck_t pool_check)
5579 {
5580         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5581             POOL_NAME, log_history, pool_check);
5582 }
5583 
5584 static void
5585 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5586     zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5587 {
5588         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5589             DATASET_NAME, B_FALSE, pool_check);
5590 }
5591 
5592 static void
5593 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5594 {
5595         zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5596             POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5597 }
5598 
5599 static void
5600 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5601     zfs_secpolicy_func_t *secpolicy)
5602 {
5603         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5604             NO_NAME, B_FALSE, POOL_CHECK_NONE);
5605 }
5606 
5607 static void
5608 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5609     zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5610 {
5611         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5612             DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5613 }
5614 
5615 static void
5616 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5617 {
5618         zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5619             zfs_secpolicy_read);
5620 }
5621 
5622 static void
5623 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5624     zfs_secpolicy_func_t *secpolicy)
5625 {
5626         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5627             DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5628 }
5629 
5630 static void
5631 zfs_ioctl_init(void)
5632 {
5633         zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5634             zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5635             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5636 
5637         zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5638             zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5639             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5640 
5641         zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5642             zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5643             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5644 
5645         zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5646             zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5647             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5648 
5649         zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5650             zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5651             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5652 
5653         zfs_ioctl_register("create", ZFS_IOC_CREATE,
5654             zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5655             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5656 
5657         zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5658             zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5659             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5660 
5661         zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5662             zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5663             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5664 
5665         zfs_ioctl_register("hold", ZFS_IOC_HOLD,
5666             zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
5667             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5668         zfs_ioctl_register("release", ZFS_IOC_RELEASE,
5669             zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
5670             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5671 
5672         zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
5673             zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
5674             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5675 
5676         zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
5677             zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
5678             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
5679 
5680         zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
5681             zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
5682             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5683 
5684         zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
5685             zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
5686             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5687 
5688         zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
5689             zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
5690             POOL_NAME,
5691             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5692 
5693         /* IOCTLS that use the legacy function signature */
5694 
5695         zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5696             zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5697 
5698         zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5699             zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5700         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5701             zfs_ioc_pool_scan);
5702         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5703             zfs_ioc_pool_upgrade);
5704         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5705             zfs_ioc_vdev_add);
5706         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5707             zfs_ioc_vdev_remove);
5708         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5709             zfs_ioc_vdev_set_state);
5710         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5711             zfs_ioc_vdev_attach);
5712         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5713             zfs_ioc_vdev_detach);
5714         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5715             zfs_ioc_vdev_setpath);
5716         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5717             zfs_ioc_vdev_setfru);
5718         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5719             zfs_ioc_pool_set_props);
5720         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5721             zfs_ioc_vdev_split);
5722         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5723             zfs_ioc_pool_reguid);
5724 
5725         zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5726             zfs_ioc_pool_configs, zfs_secpolicy_none);
5727         zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5728             zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5729         zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5730             zfs_ioc_inject_fault, zfs_secpolicy_inject);
5731         zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5732             zfs_ioc_clear_fault, zfs_secpolicy_inject);
5733         zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
5734             zfs_ioc_inject_list_next, zfs_secpolicy_inject);
5735 
5736         /*
5737          * pool destroy, and export don't log the history as part of
5738          * zfsdev_ioctl, but rather zfs_ioc_pool_export
5739          * does the logging of those commands.
5740          */
5741         zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
5742             zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5743         zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
5744             zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5745 
5746         zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
5747             zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5748         zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
5749             zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5750 
5751         zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
5752             zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
5753         zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
5754             zfs_ioc_dsobj_to_dsname,
5755             zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
5756         zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
5757             zfs_ioc_pool_get_history,
5758             zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
5759 
5760         zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
5761             zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5762 
5763         zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
5764             zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5765         zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
5766             zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5767 
5768         zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
5769             zfs_ioc_space_written);
5770         zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
5771             zfs_ioc_objset_recvd_props);
5772         zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
5773             zfs_ioc_next_obj);
5774         zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
5775             zfs_ioc_get_fsacl);
5776         zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
5777             zfs_ioc_objset_stats);
5778         zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
5779             zfs_ioc_objset_zplprops);
5780         zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
5781             zfs_ioc_dataset_list_next);
5782         zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
5783             zfs_ioc_snapshot_list_next);
5784         zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
5785             zfs_ioc_send_progress);
5786 
5787         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
5788             zfs_ioc_diff, zfs_secpolicy_diff);
5789         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
5790             zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
5791         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
5792             zfs_ioc_obj_to_path, zfs_secpolicy_diff);
5793         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
5794             zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
5795         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
5796             zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
5797         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
5798             zfs_ioc_send, zfs_secpolicy_send);
5799 
5800         zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
5801             zfs_secpolicy_none);
5802         zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
5803             zfs_secpolicy_destroy);
5804         zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
5805             zfs_secpolicy_rename);
5806         zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
5807             zfs_secpolicy_recv);
5808         zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
5809             zfs_secpolicy_promote);
5810         zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
5811             zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
5812         zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
5813             zfs_secpolicy_set_fsacl);
5814 
5815         zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
5816             zfs_secpolicy_share, POOL_CHECK_NONE);
5817         zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
5818             zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
5819         zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
5820             zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
5821             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5822         zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
5823             zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
5824             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5825 }
5826 
5827 int
5828 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
5829     zfs_ioc_poolcheck_t check)
5830 {
5831         spa_t *spa;
5832         int error;
5833 
5834         ASSERT(type == POOL_NAME || type == DATASET_NAME);
5835 
5836         if (check & POOL_CHECK_NONE)
5837                 return (0);
5838 
5839         error = spa_open(name, &spa, FTAG);
5840         if (error == 0) {
5841                 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
5842                         error = SET_ERROR(EAGAIN);
5843                 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
5844                         error = SET_ERROR(EROFS);
5845                 spa_close(spa, FTAG);
5846         }
5847         return (error);
5848 }
5849 
5850 /*
5851  * Find a free minor number.
5852  */
5853 minor_t
5854 zfsdev_minor_alloc(void)
5855 {
5856         static minor_t last_minor;
5857         minor_t m;
5858 
5859         ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5860 
5861         for (m = last_minor + 1; m != last_minor; m++) {
5862                 if (m > ZFSDEV_MAX_MINOR)
5863                         m = 1;
5864                 if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
5865                         last_minor = m;
5866                         return (m);
5867                 }
5868         }
5869 
5870         return (0);
5871 }
5872 
5873 static int
5874 zfs_ctldev_init(dev_t *devp)
5875 {
5876         minor_t minor;
5877         zfs_soft_state_t *zs;
5878 
5879         ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5880         ASSERT(getminor(*devp) == 0);
5881 
5882         minor = zfsdev_minor_alloc();
5883         if (minor == 0)
5884                 return (SET_ERROR(ENXIO));
5885 
5886         if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
5887                 return (SET_ERROR(EAGAIN));
5888 
5889         *devp = makedevice(getemajor(*devp), minor);
5890 
5891         zs = ddi_get_soft_state(zfsdev_state, minor);
5892         zs->zss_type = ZSST_CTLDEV;
5893         zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
5894 
5895         return (0);
5896 }
5897 
5898 static void
5899 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
5900 {
5901         ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5902 
5903         zfs_onexit_destroy(zo);
5904         ddi_soft_state_free(zfsdev_state, minor);
5905 }
5906 
5907 void *
5908 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
5909 {
5910         zfs_soft_state_t *zp;
5911 
5912         zp = ddi_get_soft_state(zfsdev_state, minor);
5913         if (zp == NULL || zp->zss_type != which)
5914                 return (NULL);
5915 
5916         return (zp->zss_data);
5917 }
5918 
5919 static int
5920 zfsdev_open(dev_t *devp, int flag, int otyp, cred_t *cr)
5921 {
5922         int error = 0;
5923 
5924         if (getminor(*devp) != 0)
5925                 return (zvol_open(devp, flag, otyp, cr));
5926 
5927         /* This is the control device. Allocate a new minor if requested. */
5928         if (flag & FEXCL) {
5929                 mutex_enter(&zfsdev_state_lock);
5930                 error = zfs_ctldev_init(devp);
5931                 mutex_exit(&zfsdev_state_lock);
5932         }
5933 
5934         return (error);
5935 }
5936 
5937 static int
5938 zfsdev_close(dev_t dev, int flag, int otyp, cred_t *cr)
5939 {
5940         zfs_onexit_t *zo;
5941         minor_t minor = getminor(dev);
5942 
5943         if (minor == 0)
5944                 return (0);
5945 
5946         mutex_enter(&zfsdev_state_lock);
5947         zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
5948         if (zo == NULL) {
5949                 mutex_exit(&zfsdev_state_lock);
5950                 return (zvol_close(dev, flag, otyp, cr));
5951         }
5952         zfs_ctldev_destroy(zo, minor);
5953         mutex_exit(&zfsdev_state_lock);
5954 
5955         return (0);
5956 }
5957 
5958 static int
5959 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
5960 {
5961         zfs_cmd_t *zc;
5962         uint_t vecnum;
5963         int error, rc, len;
5964         minor_t minor = getminor(dev);
5965         const zfs_ioc_vec_t *vec;
5966         char *saved_poolname = NULL;
5967         nvlist_t *innvl = NULL;
5968 
5969         if (minor != 0 &&
5970             zfsdev_get_soft_state(minor, ZSST_CTLDEV) == NULL)
5971                 return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
5972 
5973         vecnum = cmd - ZFS_IOC_FIRST;
5974         ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
5975 
5976         if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
5977                 return (SET_ERROR(EINVAL));
5978         vec = &zfs_ioc_vec[vecnum];
5979 
5980         zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
5981 
5982         error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
5983         if (error != 0) {
5984                 error = SET_ERROR(EFAULT);
5985                 goto out;
5986         }
5987 
5988         zc->zc_iflags = flag & FKIOCTL;
5989         if (zc->zc_nvlist_src_size != 0) {
5990                 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
5991                     zc->zc_iflags, &innvl);
5992                 if (error != 0)
5993                         goto out;
5994         }
5995 
5996         /*
5997          * Ensure that all pool/dataset names are valid before we pass down to
5998          * the lower layers.
5999          */
6000         zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
6001         switch (vec->zvec_namecheck) {
6002         case POOL_NAME:
6003                 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
6004                         error = SET_ERROR(EINVAL);
6005                 else
6006                         error = pool_status_check(zc->zc_name,
6007                             vec->zvec_namecheck, vec->zvec_pool_check);
6008                 break;
6009 
6010         case DATASET_NAME:
6011                 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
6012                         error = SET_ERROR(EINVAL);
6013                 else
6014                         error = pool_status_check(zc->zc_name,
6015                             vec->zvec_namecheck, vec->zvec_pool_check);
6016                 break;
6017 
6018         case NO_NAME:
6019                 break;
6020         }
6021 
6022 
6023         if (error == 0 && !(flag & FKIOCTL))
6024                 error = vec->zvec_secpolicy(zc, innvl, cr);
6025 
6026         if (error != 0)
6027                 goto out;
6028 
6029         /* legacy ioctls can modify zc_name */
6030         len = strcspn(zc->zc_name, "/@#") + 1;
6031         saved_poolname = kmem_alloc(len, KM_SLEEP);
6032         (void) strlcpy(saved_poolname, zc->zc_name, len);
6033 
6034         if (vec->zvec_func != NULL) {
6035                 nvlist_t *outnvl;
6036                 int puterror = 0;
6037                 spa_t *spa;
6038                 nvlist_t *lognv = NULL;
6039 
6040                 ASSERT(vec->zvec_legacy_func == NULL);
6041 
6042                 /*
6043                  * Add the innvl to the lognv before calling the func,
6044                  * in case the func changes the innvl.
6045                  */
6046                 if (vec->zvec_allow_log) {
6047                         lognv = fnvlist_alloc();
6048                         fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
6049                             vec->zvec_name);
6050                         if (!nvlist_empty(innvl)) {
6051                                 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
6052                                     innvl);
6053                         }
6054                 }
6055 
6056                 outnvl = fnvlist_alloc();
6057                 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
6058 
6059                 if (error == 0 && vec->zvec_allow_log &&
6060                     spa_open(zc->zc_name, &spa, FTAG) == 0) {
6061                         if (!nvlist_empty(outnvl)) {
6062                                 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
6063                                     outnvl);
6064                         }
6065                         (void) spa_history_log_nvl(spa, lognv);
6066                         spa_close(spa, FTAG);
6067                 }
6068                 fnvlist_free(lognv);
6069 
6070                 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
6071                         int smusherror = 0;
6072                         if (vec->zvec_smush_outnvlist) {
6073                                 smusherror = nvlist_smush(outnvl,
6074                                     zc->zc_nvlist_dst_size);
6075                         }
6076                         if (smusherror == 0)
6077                                 puterror = put_nvlist(zc, outnvl);
6078                 }
6079 
6080                 if (puterror != 0)
6081                         error = puterror;
6082 
6083                 nvlist_free(outnvl);
6084         } else {
6085                 error = vec->zvec_legacy_func(zc);
6086         }
6087 
6088 out:
6089         nvlist_free(innvl);
6090         rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
6091         if (error == 0 && rc != 0)
6092                 error = SET_ERROR(EFAULT);
6093         if (error == 0 && vec->zvec_allow_log) {
6094                 char *s = tsd_get(zfs_allow_log_key);
6095                 if (s != NULL)
6096                         strfree(s);
6097                 (void) tsd_set(zfs_allow_log_key, saved_poolname);
6098         } else {
6099                 if (saved_poolname != NULL)
6100                         strfree(saved_poolname);
6101         }
6102 
6103         kmem_free(zc, sizeof (zfs_cmd_t));
6104         return (error);
6105 }
6106 
6107 static int
6108 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
6109 {
6110         if (cmd != DDI_ATTACH)
6111                 return (DDI_FAILURE);
6112 
6113         if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
6114             DDI_PSEUDO, 0) == DDI_FAILURE)
6115                 return (DDI_FAILURE);
6116 
6117         zfs_dip = dip;
6118 
6119         ddi_report_dev(dip);
6120 
6121         return (DDI_SUCCESS);
6122 }
6123 
6124 static int
6125 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
6126 {
6127         if (spa_busy() || zfs_busy() || zvol_busy())
6128                 return (DDI_FAILURE);
6129 
6130         if (cmd != DDI_DETACH)
6131                 return (DDI_FAILURE);
6132 
6133         zfs_dip = NULL;
6134 
6135         ddi_prop_remove_all(dip);
6136         ddi_remove_minor_node(dip, NULL);
6137 
6138         return (DDI_SUCCESS);
6139 }
6140 
6141 /*ARGSUSED*/
6142 static int
6143 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
6144 {
6145         switch (infocmd) {
6146         case DDI_INFO_DEVT2DEVINFO:
6147                 *result = zfs_dip;
6148                 return (DDI_SUCCESS);
6149 
6150         case DDI_INFO_DEVT2INSTANCE:
6151                 *result = (void *)0;
6152                 return (DDI_SUCCESS);
6153         }
6154 
6155         return (DDI_FAILURE);
6156 }
6157 
6158 /*
6159  * OK, so this is a little weird.
6160  *
6161  * /dev/zfs is the control node, i.e. minor 0.
6162  * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
6163  *
6164  * /dev/zfs has basically nothing to do except serve up ioctls,
6165  * so most of the standard driver entry points are in zvol.c.
6166  */
6167 static struct cb_ops zfs_cb_ops = {
6168         zfsdev_open,    /* open */
6169         zfsdev_close,   /* close */
6170         zvol_strategy,  /* strategy */
6171         nodev,          /* print */
6172         zvol_dump,      /* dump */
6173         zvol_read,      /* read */
6174         zvol_write,     /* write */
6175         zfsdev_ioctl,   /* ioctl */
6176         nodev,          /* devmap */
6177         nodev,          /* mmap */
6178         nodev,          /* segmap */
6179         nochpoll,       /* poll */
6180         ddi_prop_op,    /* prop_op */
6181         NULL,           /* streamtab */
6182         D_NEW | D_MP | D_64BIT,         /* Driver compatibility flag */
6183         CB_REV,         /* version */
6184         nodev,          /* async read */
6185         nodev,          /* async write */
6186 };
6187 
6188 static struct dev_ops zfs_dev_ops = {
6189         DEVO_REV,       /* version */
6190         0,              /* refcnt */
6191         zfs_info,       /* info */
6192         nulldev,        /* identify */
6193         nulldev,        /* probe */
6194         zfs_attach,     /* attach */
6195         zfs_detach,     /* detach */
6196         nodev,          /* reset */
6197         &zfs_cb_ops,        /* driver operations */
6198         NULL,           /* no bus operations */
6199         NULL,           /* power */
6200         ddi_quiesce_not_needed, /* quiesce */
6201 };
6202 
6203 static struct modldrv zfs_modldrv = {
6204         &mod_driverops,
6205         "ZFS storage pool",
6206         &zfs_dev_ops
6207 };
6208 
6209 static struct modlinkage modlinkage = {
6210         MODREV_1,
6211         (void *)&zfs_modlfs,
6212         (void *)&zfs_modldrv,
6213         NULL
6214 };
6215 
6216 static void
6217 zfs_allow_log_destroy(void *arg)
6218 {
6219         char *poolname = arg;
6220         strfree(poolname);
6221 }
6222 
6223 int
6224 _init(void)
6225 {
6226         int error;
6227 
6228         spa_init(FREAD | FWRITE);
6229         zfs_init();
6230         zvol_init();
6231         zfs_ioctl_init();
6232 
6233         if ((error = mod_install(&modlinkage)) != 0) {
6234                 zvol_fini();
6235                 zfs_fini();
6236                 spa_fini();
6237                 return (error);
6238         }
6239 
6240         tsd_create(&zfs_fsyncer_key, NULL);
6241         tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6242         tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6243 
6244         error = ldi_ident_from_mod(&modlinkage, &zfs_li);
6245         ASSERT(error == 0);
6246         mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6247 
6248         return (0);
6249 }
6250 
6251 int
6252 _fini(void)
6253 {
6254         int error;
6255 
6256         if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
6257                 return (SET_ERROR(EBUSY));
6258 
6259         if ((error = mod_remove(&modlinkage)) != 0)
6260                 return (error);
6261 
6262         zvol_fini();
6263         zfs_fini();
6264         spa_fini();
6265         if (zfs_nfsshare_inited)
6266                 (void) ddi_modclose(nfs_mod);
6267         if (zfs_smbshare_inited)
6268                 (void) ddi_modclose(smbsrv_mod);
6269         if (zfs_nfsshare_inited || zfs_smbshare_inited)
6270                 (void) ddi_modclose(sharefs_mod);
6271 
6272         tsd_destroy(&zfs_fsyncer_key);
6273         ldi_ident_release(zfs_li);
6274         zfs_li = NULL;
6275         mutex_destroy(&zfs_share_lock);
6276 
6277         return (error);
6278 }
6279 
6280 int
6281 _info(struct modinfo *modinfop)
6282 {
6283         return (mod_info(&modlinkage, modinfop));
6284 }