1 /*
   2  * CDDL HEADER START
   3  *
   4  * The contents of this file are subject to the terms of the
   5  * Common Development and Distribution License (the "License").
   6  * You may not use this file except in compliance with the License.
   7  *
   8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
   9  * or http://www.opensolaris.org/os/licensing.
  10  * See the License for the specific language governing permissions
  11  * and limitations under the License.
  12  *
  13  * When distributing Covered Code, include this CDDL HEADER in each
  14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
  15  * If applicable, add the following below this CDDL HEADER, with the
  16  * fields enclosed by brackets "[]" replaced with your own identifying
  17  * information: Portions Copyright [yyyy] [name of copyright owner]
  18  *
  19  * CDDL HEADER END
  20  */
  21 
  22 /*
  23  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
  24  * Portions Copyright 2011 Martin Matuska
  25  * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
  26  * Copyright (c) 2012, Joyent, Inc. All rights reserved.
  27  * Copyright (c) 2012 by Delphix. All rights reserved.
  28  * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
  29  */
  30 
  31 /*
  32  * ZFS ioctls.
  33  *
  34  * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
  35  * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
  36  *
  37  * There are two ways that we handle ioctls: the legacy way where almost
  38  * all of the logic is in the ioctl callback, and the new way where most
  39  * of the marshalling is handled in the common entry point, zfsdev_ioctl().
  40  *
  41  * Non-legacy ioctls should be registered by calling
  42  * zfs_ioctl_register() from zfs_ioctl_init().  The ioctl is invoked
  43  * from userland by lzc_ioctl().
  44  *
  45  * The registration arguments are as follows:
  46  *
  47  * const char *name
  48  *   The name of the ioctl.  This is used for history logging.  If the
  49  *   ioctl returns successfully (the callback returns 0), and allow_log
  50  *   is true, then a history log entry will be recorded with the input &
  51  *   output nvlists.  The log entry can be printed with "zpool history -i".
  52  *
  53  * zfs_ioc_t ioc
  54  *   The ioctl request number, which userland will pass to ioctl(2).
  55  *   The ioctl numbers can change from release to release, because
  56  *   the caller (libzfs) must be matched to the kernel.
  57  *
  58  * zfs_secpolicy_func_t *secpolicy
  59  *   This function will be called before the zfs_ioc_func_t, to
  60  *   determine if this operation is permitted.  It should return EPERM
  61  *   on failure, and 0 on success.  Checks include determining if the
  62  *   dataset is visible in this zone, and if the user has either all
  63  *   zfs privileges in the zone (SYS_MOUNT), or has been granted permission
  64  *   to do this operation on this dataset with "zfs allow".
  65  *
  66  * zfs_ioc_namecheck_t namecheck
  67  *   This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
  68  *   name, a dataset name, or nothing.  If the name is not well-formed,
  69  *   the ioctl will fail and the callback will not be called.
  70  *   Therefore, the callback can assume that the name is well-formed
  71  *   (e.g. is null-terminated, doesn't have more than one '@' character,
  72  *   doesn't have invalid characters).
  73  *
  74  * zfs_ioc_poolcheck_t pool_check
  75  *   This specifies requirements on the pool state.  If the pool does
  76  *   not meet them (is suspended or is readonly), the ioctl will fail
  77  *   and the callback will not be called.  If any checks are specified
  78  *   (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
  79  *   Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
  80  *   POOL_CHECK_READONLY).
  81  *
  82  * boolean_t smush_outnvlist
  83  *   If smush_outnvlist is true, then the output is presumed to be a
  84  *   list of errors, and it will be "smushed" down to fit into the
  85  *   caller's buffer, by removing some entries and replacing them with a
  86  *   single "N_MORE_ERRORS" entry indicating how many were removed.  See
  87  *   nvlist_smush() for details.  If smush_outnvlist is false, and the
  88  *   outnvlist does not fit into the userland-provided buffer, then the
  89  *   ioctl will fail with ENOMEM.
  90  *
  91  * zfs_ioc_func_t *func
  92  *   The callback function that will perform the operation.
  93  *
  94  *   The callback should return 0 on success, or an error number on
  95  *   failure.  If the function fails, the userland ioctl will return -1,
  96  *   and errno will be set to the callback's return value.  The callback
  97  *   will be called with the following arguments:
  98  *
  99  *   const char *name
 100  *     The name of the pool or dataset to operate on, from
 101  *     zfs_cmd_t:zc_name.  The 'namecheck' argument specifies the
 102  *     expected type (pool, dataset, or none).
 103  *
 104  *   nvlist_t *innvl
 105  *     The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src.  Or
 106  *     NULL if no input nvlist was provided.  Changes to this nvlist are
 107  *     ignored.  If the input nvlist could not be deserialized, the
 108  *     ioctl will fail and the callback will not be called.
 109  *
 110  *   nvlist_t *outnvl
 111  *     The output nvlist, initially empty.  The callback can fill it in,
 112  *     and it will be returned to userland by serializing it into
 113  *     zfs_cmd_t:zc_nvlist_dst.  If it is non-empty, and serialization
 114  *     fails (e.g. because the caller didn't supply a large enough
 115  *     buffer), then the overall ioctl will fail.  See the
 116  *     'smush_nvlist' argument above for additional behaviors.
 117  *
 118  *     There are two typical uses of the output nvlist:
 119  *       - To return state, e.g. property values.  In this case,
 120  *         smush_outnvlist should be false.  If the buffer was not large
 121  *         enough, the caller will reallocate a larger buffer and try
 122  *         the ioctl again.
 123  *
 124  *       - To return multiple errors from an ioctl which makes on-disk
 125  *         changes.  In this case, smush_outnvlist should be true.
 126  *         Ioctls which make on-disk modifications should generally not
 127  *         use the outnvl if they succeed, because the caller can not
 128  *         distinguish between the operation failing, and
 129  *         deserialization failing.
 130  */
 131 
 132 #include <sys/types.h>
 133 #include <sys/param.h>
 134 #include <sys/errno.h>
 135 #include <sys/uio.h>
 136 #include <sys/buf.h>
 137 #include <sys/modctl.h>
 138 #include <sys/open.h>
 139 #include <sys/file.h>
 140 #include <sys/kmem.h>
 141 #include <sys/conf.h>
 142 #include <sys/cmn_err.h>
 143 #include <sys/stat.h>
 144 #include <sys/zfs_ioctl.h>
 145 #include <sys/zfs_vfsops.h>
 146 #include <sys/zfs_znode.h>
 147 #include <sys/zap.h>
 148 #include <sys/spa.h>
 149 #include <sys/spa_impl.h>
 150 #include <sys/vdev.h>
 151 #include <sys/priv_impl.h>
 152 #include <sys/dmu.h>
 153 #include <sys/dsl_dir.h>
 154 #include <sys/dsl_dataset.h>
 155 #include <sys/dsl_prop.h>
 156 #include <sys/dsl_deleg.h>
 157 #include <sys/dmu_objset.h>
 158 #include <sys/dmu_impl.h>
 159 #include <sys/ddi.h>
 160 #include <sys/sunddi.h>
 161 #include <sys/sunldi.h>
 162 #include <sys/policy.h>
 163 #include <sys/zone.h>
 164 #include <sys/nvpair.h>
 165 #include <sys/pathname.h>
 166 #include <sys/mount.h>
 167 #include <sys/sdt.h>
 168 #include <sys/fs/zfs.h>
 169 #include <sys/zfs_ctldir.h>
 170 #include <sys/zfs_dir.h>
 171 #include <sys/zfs_onexit.h>
 172 #include <sys/zvol.h>
 173 #include <sys/dsl_scan.h>
 174 #include <sharefs/share.h>
 175 #include <sys/dmu_objset.h>
 176 #include <sys/zfeature.h>
 177 
 178 #include "zfs_namecheck.h"
 179 #include "zfs_prop.h"
 180 #include "zfs_deleg.h"
 181 #include "zfs_comutil.h"
 182 
 183 extern struct modlfs zfs_modlfs;
 184 
 185 extern void zfs_init(void);
 186 extern void zfs_fini(void);
 187 
 188 ldi_ident_t zfs_li = NULL;
 189 dev_info_t *zfs_dip;
 190 
 191 uint_t zfs_fsyncer_key;
 192 extern uint_t rrw_tsd_key;
 193 static uint_t zfs_allow_log_key;
 194 
 195 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
 196 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
 197 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
 198 
 199 typedef enum {
 200         NO_NAME,
 201         POOL_NAME,
 202         DATASET_NAME
 203 } zfs_ioc_namecheck_t;
 204 
 205 typedef enum {
 206         POOL_CHECK_NONE         = 1 << 0,
 207         POOL_CHECK_SUSPENDED    = 1 << 1,
 208         POOL_CHECK_READONLY     = 1 << 2,
 209 } zfs_ioc_poolcheck_t;
 210 
 211 typedef struct zfs_ioc_vec {
 212         zfs_ioc_legacy_func_t   *zvec_legacy_func;
 213         zfs_ioc_func_t          *zvec_func;
 214         zfs_secpolicy_func_t    *zvec_secpolicy;
 215         zfs_ioc_namecheck_t     zvec_namecheck;
 216         boolean_t               zvec_allow_log;
 217         zfs_ioc_poolcheck_t     zvec_pool_check;
 218         boolean_t               zvec_smush_outnvlist;
 219         const char              *zvec_name;
 220 } zfs_ioc_vec_t;
 221 
 222 /* This array is indexed by zfs_userquota_prop_t */
 223 static const char *userquota_perms[] = {
 224         ZFS_DELEG_PERM_USERUSED,
 225         ZFS_DELEG_PERM_USERQUOTA,
 226         ZFS_DELEG_PERM_GROUPUSED,
 227         ZFS_DELEG_PERM_GROUPQUOTA,
 228 };
 229 
 230 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
 231 static int zfs_check_settable(const char *name, nvpair_t *property,
 232     cred_t *cr);
 233 static int zfs_check_clearable(char *dataset, nvlist_t *props,
 234     nvlist_t **errors);
 235 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
 236     boolean_t *);
 237 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
 238 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
 239 
 240 static int zfs_prop_activate_feature(dsl_pool_t *dp, zfeature_info_t *feature);
 241 static int zfs_prop_activate_feature_check(void *arg1, void *arg2,
 242     dmu_tx_t *tx);
 243 static void zfs_prop_activate_feature_sync(void *arg1, void *arg2,
 244     dmu_tx_t *tx);
 245 
 246 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
 247 void
 248 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
 249 {
 250         const char *newfile;
 251         char buf[512];
 252         va_list adx;
 253 
 254         /*
 255          * Get rid of annoying "../common/" prefix to filename.
 256          */
 257         newfile = strrchr(file, '/');
 258         if (newfile != NULL) {
 259                 newfile = newfile + 1; /* Get rid of leading / */
 260         } else {
 261                 newfile = file;
 262         }
 263 
 264         va_start(adx, fmt);
 265         (void) vsnprintf(buf, sizeof (buf), fmt, adx);
 266         va_end(adx);
 267 
 268         /*
 269          * To get this data, use the zfs-dprintf probe as so:
 270          * dtrace -q -n 'zfs-dprintf \
 271          *      /stringof(arg0) == "dbuf.c"/ \
 272          *      {printf("%s: %s", stringof(arg1), stringof(arg3))}'
 273          * arg0 = file name
 274          * arg1 = function name
 275          * arg2 = line number
 276          * arg3 = message
 277          */
 278         DTRACE_PROBE4(zfs__dprintf,
 279             char *, newfile, char *, func, int, line, char *, buf);
 280 }
 281 
 282 static void
 283 history_str_free(char *buf)
 284 {
 285         kmem_free(buf, HIS_MAX_RECORD_LEN);
 286 }
 287 
 288 static char *
 289 history_str_get(zfs_cmd_t *zc)
 290 {
 291         char *buf;
 292 
 293         if (zc->zc_history == NULL)
 294                 return (NULL);
 295 
 296         buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
 297         if (copyinstr((void *)(uintptr_t)zc->zc_history,
 298             buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
 299                 history_str_free(buf);
 300                 return (NULL);
 301         }
 302 
 303         buf[HIS_MAX_RECORD_LEN -1] = '\0';
 304 
 305         return (buf);
 306 }
 307 
 308 /*
 309  * Check to see if the named dataset is currently defined as bootable
 310  */
 311 static boolean_t
 312 zfs_is_bootfs(const char *name)
 313 {
 314         objset_t *os;
 315 
 316         if (dmu_objset_hold(name, FTAG, &os) == 0) {
 317                 boolean_t ret;
 318                 ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
 319                 dmu_objset_rele(os, FTAG);
 320                 return (ret);
 321         }
 322         return (B_FALSE);
 323 }
 324 
 325 /*
 326  * zfs_earlier_version
 327  *
 328  *      Return non-zero if the spa version is less than requested version.
 329  */
 330 static int
 331 zfs_earlier_version(const char *name, int version)
 332 {
 333         spa_t *spa;
 334 
 335         if (spa_open(name, &spa, FTAG) == 0) {
 336                 if (spa_version(spa) < version) {
 337                         spa_close(spa, FTAG);
 338                         return (1);
 339                 }
 340                 spa_close(spa, FTAG);
 341         }
 342         return (0);
 343 }
 344 
 345 /*
 346  * zpl_earlier_version
 347  *
 348  * Return TRUE if the ZPL version is less than requested version.
 349  */
 350 static boolean_t
 351 zpl_earlier_version(const char *name, int version)
 352 {
 353         objset_t *os;
 354         boolean_t rc = B_TRUE;
 355 
 356         if (dmu_objset_hold(name, FTAG, &os) == 0) {
 357                 uint64_t zplversion;
 358 
 359                 if (dmu_objset_type(os) != DMU_OST_ZFS) {
 360                         dmu_objset_rele(os, FTAG);
 361                         return (B_TRUE);
 362                 }
 363                 /* XXX reading from non-owned objset */
 364                 if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
 365                         rc = zplversion < version;
 366                 dmu_objset_rele(os, FTAG);
 367         }
 368         return (rc);
 369 }
 370 
 371 static void
 372 zfs_log_history(zfs_cmd_t *zc)
 373 {
 374         spa_t *spa;
 375         char *buf;
 376 
 377         if ((buf = history_str_get(zc)) == NULL)
 378                 return;
 379 
 380         if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
 381                 if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
 382                         (void) spa_history_log(spa, buf);
 383                 spa_close(spa, FTAG);
 384         }
 385         history_str_free(buf);
 386 }
 387 
 388 /*
 389  * Policy for top-level read operations (list pools).  Requires no privileges,
 390  * and can be used in the local zone, as there is no associated dataset.
 391  */
 392 /* ARGSUSED */
 393 static int
 394 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 395 {
 396         return (0);
 397 }
 398 
 399 /*
 400  * Policy for dataset read operations (list children, get statistics).  Requires
 401  * no privileges, but must be visible in the local zone.
 402  */
 403 /* ARGSUSED */
 404 static int
 405 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 406 {
 407         if (INGLOBALZONE(curproc) ||
 408             zone_dataset_visible(zc->zc_name, NULL))
 409                 return (0);
 410 
 411         return (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 (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 (EPERM);
 434         } else {
 435                 /*
 436                  * If we are in a local zone, the 'zoned' property must be set.
 437                  */
 438                 if (!zoned)
 439                         return (EPERM);
 440 
 441                 /* must be writable by this zone */
 442                 if (!writable)
 443                         return (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 (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         rw_enter(&ds->ds_dir->dd_pool->dp_config_rwlock, RW_READER);
 465         if (dsl_prop_get_ds(ds, "zoned", 8, 1, &zoned, NULL)) {
 466                 rw_exit(&ds->ds_dir->dd_pool->dp_config_rwlock);
 467                 return (ENOENT);
 468         }
 469         rw_exit(&ds->ds_dir->dd_pool->dp_config_rwlock);
 470 
 471         return (zfs_dozonecheck_impl(dataset, zoned, cr));
 472 }
 473 
 474 static int
 475 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
 476 {
 477         int error;
 478         dsl_dataset_t *ds;
 479 
 480         error = dsl_dataset_hold(name, FTAG, &ds);
 481         if (error != 0)
 482                 return (error);
 483 
 484         error = zfs_dozonecheck_ds(name, ds, cr);
 485         if (error == 0) {
 486                 error = secpolicy_zfs(cr);
 487                 if (error)
 488                         error = dsl_deleg_access_impl(ds, perm, cr);
 489         }
 490 
 491         dsl_dataset_rele(ds, FTAG);
 492         return (error);
 493 }
 494 
 495 static int
 496 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
 497     const char *perm, cred_t *cr)
 498 {
 499         int error;
 500 
 501         error = zfs_dozonecheck_ds(name, ds, cr);
 502         if (error == 0) {
 503                 error = secpolicy_zfs(cr);
 504                 if (error)
 505                         error = dsl_deleg_access_impl(ds, perm, cr);
 506         }
 507         return (error);
 508 }
 509 
 510 /*
 511  * Policy for setting the security label property.
 512  *
 513  * Returns 0 for success, non-zero for access and other errors.
 514  */
 515 static int
 516 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
 517 {
 518         char            ds_hexsl[MAXNAMELEN];
 519         bslabel_t       ds_sl, new_sl;
 520         boolean_t       new_default = FALSE;
 521         uint64_t        zoned;
 522         int             needed_priv = -1;
 523         int             error;
 524 
 525         /* First get the existing dataset label. */
 526         error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
 527             1, sizeof (ds_hexsl), &ds_hexsl, NULL);
 528         if (error)
 529                 return (EPERM);
 530 
 531         if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
 532                 new_default = TRUE;
 533 
 534         /* The label must be translatable */
 535         if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
 536                 return (EINVAL);
 537 
 538         /*
 539          * In a non-global zone, disallow attempts to set a label that
 540          * doesn't match that of the zone; otherwise no other checks
 541          * are needed.
 542          */
 543         if (!INGLOBALZONE(curproc)) {
 544                 if (new_default || !blequal(&new_sl, CR_SL(CRED())))
 545                         return (EPERM);
 546                 return (0);
 547         }
 548 
 549         /*
 550          * For global-zone datasets (i.e., those whose zoned property is
 551          * "off", verify that the specified new label is valid for the
 552          * global zone.
 553          */
 554         if (dsl_prop_get_integer(name,
 555             zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
 556                 return (EPERM);
 557         if (!zoned) {
 558                 if (zfs_check_global_label(name, strval) != 0)
 559                         return (EPERM);
 560         }
 561 
 562         /*
 563          * If the existing dataset label is nondefault, check if the
 564          * dataset is mounted (label cannot be changed while mounted).
 565          * Get the zfsvfs; if there isn't one, then the dataset isn't
 566          * mounted (or isn't a dataset, doesn't exist, ...).
 567          */
 568         if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
 569                 objset_t *os;
 570                 static char *setsl_tag = "setsl_tag";
 571 
 572                 /*
 573                  * Try to own the dataset; abort if there is any error,
 574                  * (e.g., already mounted, in use, or other error).
 575                  */
 576                 error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
 577                     setsl_tag, &os);
 578                 if (error)
 579                         return (EPERM);
 580 
 581                 dmu_objset_disown(os, setsl_tag);
 582 
 583                 if (new_default) {
 584                         needed_priv = PRIV_FILE_DOWNGRADE_SL;
 585                         goto out_check;
 586                 }
 587 
 588                 if (hexstr_to_label(strval, &new_sl) != 0)
 589                         return (EPERM);
 590 
 591                 if (blstrictdom(&ds_sl, &new_sl))
 592                         needed_priv = PRIV_FILE_DOWNGRADE_SL;
 593                 else if (blstrictdom(&new_sl, &ds_sl))
 594                         needed_priv = PRIV_FILE_UPGRADE_SL;
 595         } else {
 596                 /* dataset currently has a default label */
 597                 if (!new_default)
 598                         needed_priv = PRIV_FILE_UPGRADE_SL;
 599         }
 600 
 601 out_check:
 602         if (needed_priv != -1)
 603                 return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
 604         return (0);
 605 }
 606 
 607 static int
 608 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
 609     cred_t *cr)
 610 {
 611         char *strval;
 612 
 613         /*
 614          * Check permissions for special properties.
 615          */
 616         switch (prop) {
 617         case ZFS_PROP_ZONED:
 618                 /*
 619                  * Disallow setting of 'zoned' from within a local zone.
 620                  */
 621                 if (!INGLOBALZONE(curproc))
 622                         return (EPERM);
 623                 break;
 624 
 625         case ZFS_PROP_QUOTA:
 626                 if (!INGLOBALZONE(curproc)) {
 627                         uint64_t zoned;
 628                         char setpoint[MAXNAMELEN];
 629                         /*
 630                          * Unprivileged users are allowed to modify the
 631                          * quota on things *under* (ie. contained by)
 632                          * the thing they own.
 633                          */
 634                         if (dsl_prop_get_integer(dsname, "zoned", &zoned,
 635                             setpoint))
 636                                 return (EPERM);
 637                         if (!zoned || strlen(dsname) <= strlen(setpoint))
 638                                 return (EPERM);
 639                 }
 640                 break;
 641 
 642         case ZFS_PROP_MLSLABEL:
 643                 if (!is_system_labeled())
 644                         return (EPERM);
 645 
 646                 if (nvpair_value_string(propval, &strval) == 0) {
 647                         int err;
 648 
 649                         err = zfs_set_slabel_policy(dsname, strval, CRED());
 650                         if (err != 0)
 651                                 return (err);
 652                 }
 653                 break;
 654         }
 655 
 656         return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
 657 }
 658 
 659 /* ARGSUSED */
 660 static int
 661 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 662 {
 663         int error;
 664 
 665         error = zfs_dozonecheck(zc->zc_name, cr);
 666         if (error)
 667                 return (error);
 668 
 669         /*
 670          * permission to set permissions will be evaluated later in
 671          * dsl_deleg_can_allow()
 672          */
 673         return (0);
 674 }
 675 
 676 /* ARGSUSED */
 677 static int
 678 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 679 {
 680         return (zfs_secpolicy_write_perms(zc->zc_name,
 681             ZFS_DELEG_PERM_ROLLBACK, cr));
 682 }
 683 
 684 /* ARGSUSED */
 685 static int
 686 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 687 {
 688         spa_t *spa;
 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 (EINVAL);
 701         error = spa_open(zc->zc_name, &spa, FTAG);
 702         if (error)
 703                 return (error);
 704 
 705         dp = spa_get_dsl(spa);
 706         rw_enter(&dp->dp_config_rwlock, RW_READER);
 707         error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
 708         rw_exit(&dp->dp_config_rwlock);
 709         spa_close(spa, FTAG);
 710         if (error)
 711                 return (error);
 712 
 713         dsl_dataset_name(ds, zc->zc_name);
 714 
 715         error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
 716             ZFS_DELEG_PERM_SEND, cr);
 717         dsl_dataset_rele(ds, FTAG);
 718 
 719         return (error);
 720 }
 721 
 722 /* ARGSUSED */
 723 static int
 724 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 725 {
 726         return (zfs_secpolicy_write_perms(zc->zc_name,
 727             ZFS_DELEG_PERM_SEND, cr));
 728 }
 729 
 730 /* ARGSUSED */
 731 static int
 732 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 733 {
 734         vnode_t *vp;
 735         int error;
 736 
 737         if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
 738             NO_FOLLOW, NULL, &vp)) != 0)
 739                 return (error);
 740 
 741         /* Now make sure mntpnt and dataset are ZFS */
 742 
 743         if (vp->v_vfsp->vfs_fstype != zfsfstype ||
 744             (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
 745             zc->zc_name) != 0)) {
 746                 VN_RELE(vp);
 747                 return (EPERM);
 748         }
 749 
 750         VN_RELE(vp);
 751         return (dsl_deleg_access(zc->zc_name,
 752             ZFS_DELEG_PERM_SHARE, cr));
 753 }
 754 
 755 int
 756 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 757 {
 758         if (!INGLOBALZONE(curproc))
 759                 return (EPERM);
 760 
 761         if (secpolicy_nfs(cr) == 0) {
 762                 return (0);
 763         } else {
 764                 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
 765         }
 766 }
 767 
 768 int
 769 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 770 {
 771         if (!INGLOBALZONE(curproc))
 772                 return (EPERM);
 773 
 774         if (secpolicy_smb(cr) == 0) {
 775                 return (0);
 776         } else {
 777                 return (zfs_secpolicy_deleg_share(zc, innvl, cr));
 778         }
 779 }
 780 
 781 static int
 782 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
 783 {
 784         char *cp;
 785 
 786         /*
 787          * Remove the @bla or /bla from the end of the name to get the parent.
 788          */
 789         (void) strncpy(parent, datasetname, parentsize);
 790         cp = strrchr(parent, '@');
 791         if (cp != NULL) {
 792                 cp[0] = '\0';
 793         } else {
 794                 cp = strrchr(parent, '/');
 795                 if (cp == NULL)
 796                         return (ENOENT);
 797                 cp[0] = '\0';
 798         }
 799 
 800         return (0);
 801 }
 802 
 803 int
 804 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
 805 {
 806         int error;
 807 
 808         if ((error = zfs_secpolicy_write_perms(name,
 809             ZFS_DELEG_PERM_MOUNT, cr)) != 0)
 810                 return (error);
 811 
 812         return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
 813 }
 814 
 815 /* ARGSUSED */
 816 static int
 817 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 818 {
 819         return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
 820 }
 821 
 822 /*
 823  * Destroying snapshots with delegated permissions requires
 824  * descendant mount and destroy permissions.
 825  */
 826 /* ARGSUSED */
 827 static int
 828 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 829 {
 830         nvlist_t *snaps;
 831         nvpair_t *pair, *nextpair;
 832         int error = 0;
 833 
 834         if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
 835                 return (EINVAL);
 836         for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
 837             pair = nextpair) {
 838                 dsl_dataset_t *ds;
 839 
 840                 nextpair = nvlist_next_nvpair(snaps, pair);
 841                 error = dsl_dataset_hold(nvpair_name(pair), FTAG, &ds);
 842                 if (error == 0) {
 843                         dsl_dataset_rele(ds, FTAG);
 844                 } else if (error == ENOENT) {
 845                         /*
 846                          * Ignore any snapshots that don't exist (we consider
 847                          * them "already destroyed").  Remove the name from the
 848                          * nvl here in case the snapshot is created between
 849                          * now and when we try to destroy it (in which case
 850                          * we don't want to destroy it since we haven't
 851                          * checked for permission).
 852                          */
 853                         fnvlist_remove_nvpair(snaps, pair);
 854                         error = 0;
 855                         continue;
 856                 } else {
 857                         break;
 858                 }
 859                 error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
 860                 if (error != 0)
 861                         break;
 862         }
 863 
 864         return (error);
 865 }
 866 
 867 int
 868 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
 869 {
 870         char    parentname[MAXNAMELEN];
 871         int     error;
 872 
 873         if ((error = zfs_secpolicy_write_perms(from,
 874             ZFS_DELEG_PERM_RENAME, cr)) != 0)
 875                 return (error);
 876 
 877         if ((error = zfs_secpolicy_write_perms(from,
 878             ZFS_DELEG_PERM_MOUNT, cr)) != 0)
 879                 return (error);
 880 
 881         if ((error = zfs_get_parent(to, parentname,
 882             sizeof (parentname))) != 0)
 883                 return (error);
 884 
 885         if ((error = zfs_secpolicy_write_perms(parentname,
 886             ZFS_DELEG_PERM_CREATE, cr)) != 0)
 887                 return (error);
 888 
 889         if ((error = zfs_secpolicy_write_perms(parentname,
 890             ZFS_DELEG_PERM_MOUNT, cr)) != 0)
 891                 return (error);
 892 
 893         return (error);
 894 }
 895 
 896 /* ARGSUSED */
 897 static int
 898 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 899 {
 900         return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
 901 }
 902 
 903 /* ARGSUSED */
 904 static int
 905 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 906 {
 907         char    parentname[MAXNAMELEN];
 908         objset_t *clone;
 909         int error;
 910 
 911         error = zfs_secpolicy_write_perms(zc->zc_name,
 912             ZFS_DELEG_PERM_PROMOTE, cr);
 913         if (error)
 914                 return (error);
 915 
 916         error = dmu_objset_hold(zc->zc_name, FTAG, &clone);
 917 
 918         if (error == 0) {
 919                 dsl_dataset_t *pclone = NULL;
 920                 dsl_dir_t *dd;
 921                 dd = clone->os_dsl_dataset->ds_dir;
 922 
 923                 rw_enter(&dd->dd_pool->dp_config_rwlock, RW_READER);
 924                 error = dsl_dataset_hold_obj(dd->dd_pool,
 925                     dd->dd_phys->dd_origin_obj, FTAG, &pclone);
 926                 rw_exit(&dd->dd_pool->dp_config_rwlock);
 927                 if (error) {
 928                         dmu_objset_rele(clone, FTAG);
 929                         return (error);
 930                 }
 931 
 932                 error = zfs_secpolicy_write_perms(zc->zc_name,
 933                     ZFS_DELEG_PERM_MOUNT, cr);
 934 
 935                 dsl_dataset_name(pclone, parentname);
 936                 dmu_objset_rele(clone, FTAG);
 937                 dsl_dataset_rele(pclone, FTAG);
 938                 if (error == 0)
 939                         error = zfs_secpolicy_write_perms(parentname,
 940                             ZFS_DELEG_PERM_PROMOTE, cr);
 941         }
 942         return (error);
 943 }
 944 
 945 /* ARGSUSED */
 946 static int
 947 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 948 {
 949         int error;
 950 
 951         if ((error = zfs_secpolicy_write_perms(zc->zc_name,
 952             ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
 953                 return (error);
 954 
 955         if ((error = zfs_secpolicy_write_perms(zc->zc_name,
 956             ZFS_DELEG_PERM_MOUNT, cr)) != 0)
 957                 return (error);
 958 
 959         return (zfs_secpolicy_write_perms(zc->zc_name,
 960             ZFS_DELEG_PERM_CREATE, cr));
 961 }
 962 
 963 int
 964 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
 965 {
 966         return (zfs_secpolicy_write_perms(name,
 967             ZFS_DELEG_PERM_SNAPSHOT, cr));
 968 }
 969 
 970 /*
 971  * Check for permission to create each snapshot in the nvlist.
 972  */
 973 /* ARGSUSED */
 974 static int
 975 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
 976 {
 977         nvlist_t *snaps;
 978         int error;
 979         nvpair_t *pair;
 980 
 981         if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
 982                 return (EINVAL);
 983         for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
 984             pair = nvlist_next_nvpair(snaps, pair)) {
 985                 char *name = nvpair_name(pair);
 986                 char *atp = strchr(name, '@');
 987 
 988                 if (atp == NULL) {
 989                         error = EINVAL;
 990                         break;
 991                 }
 992                 *atp = '\0';
 993                 error = zfs_secpolicy_snapshot_perms(name, cr);
 994                 *atp = '@';
 995                 if (error != 0)
 996                         break;
 997         }
 998         return (error);
 999 }
1000 
1001 /* ARGSUSED */
1002 static int
1003 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1004 {
1005         /*
1006          * Even root must have a proper TSD so that we know what pool
1007          * to log to.
1008          */
1009         if (tsd_get(zfs_allow_log_key) == NULL)
1010                 return (EPERM);
1011         return (0);
1012 }
1013 
1014 static int
1015 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1016 {
1017         char    parentname[MAXNAMELEN];
1018         int     error;
1019         char    *origin;
1020 
1021         if ((error = zfs_get_parent(zc->zc_name, parentname,
1022             sizeof (parentname))) != 0)
1023                 return (error);
1024 
1025         if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1026             (error = zfs_secpolicy_write_perms(origin,
1027             ZFS_DELEG_PERM_CLONE, cr)) != 0)
1028                 return (error);
1029 
1030         if ((error = zfs_secpolicy_write_perms(parentname,
1031             ZFS_DELEG_PERM_CREATE, cr)) != 0)
1032                 return (error);
1033 
1034         return (zfs_secpolicy_write_perms(parentname,
1035             ZFS_DELEG_PERM_MOUNT, cr));
1036 }
1037 
1038 /*
1039  * Policy for pool operations - create/destroy pools, add vdevs, etc.  Requires
1040  * SYS_CONFIG privilege, which is not available in a local zone.
1041  */
1042 /* ARGSUSED */
1043 static int
1044 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1045 {
1046         if (secpolicy_sys_config(cr, B_FALSE) != 0)
1047                 return (EPERM);
1048 
1049         return (0);
1050 }
1051 
1052 /*
1053  * Policy for object to name lookups.
1054  */
1055 /* ARGSUSED */
1056 static int
1057 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1058 {
1059         int error;
1060 
1061         if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1062                 return (0);
1063 
1064         error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1065         return (error);
1066 }
1067 
1068 /*
1069  * Policy for fault injection.  Requires all privileges.
1070  */
1071 /* ARGSUSED */
1072 static int
1073 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1074 {
1075         return (secpolicy_zinject(cr));
1076 }
1077 
1078 /* ARGSUSED */
1079 static int
1080 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1081 {
1082         zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1083 
1084         if (prop == ZPROP_INVAL) {
1085                 if (!zfs_prop_user(zc->zc_value))
1086                         return (EINVAL);
1087                 return (zfs_secpolicy_write_perms(zc->zc_name,
1088                     ZFS_DELEG_PERM_USERPROP, cr));
1089         } else {
1090                 return (zfs_secpolicy_setprop(zc->zc_name, prop,
1091                     NULL, cr));
1092         }
1093 }
1094 
1095 static int
1096 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1097 {
1098         int err = zfs_secpolicy_read(zc, innvl, cr);
1099         if (err)
1100                 return (err);
1101 
1102         if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1103                 return (EINVAL);
1104 
1105         if (zc->zc_value[0] == 0) {
1106                 /*
1107                  * They are asking about a posix uid/gid.  If it's
1108                  * themself, allow it.
1109                  */
1110                 if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1111                     zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
1112                         if (zc->zc_guid == crgetuid(cr))
1113                                 return (0);
1114                 } else {
1115                         if (groupmember(zc->zc_guid, cr))
1116                                 return (0);
1117                 }
1118         }
1119 
1120         return (zfs_secpolicy_write_perms(zc->zc_name,
1121             userquota_perms[zc->zc_objset_type], cr));
1122 }
1123 
1124 static int
1125 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1126 {
1127         int err = zfs_secpolicy_read(zc, innvl, cr);
1128         if (err)
1129                 return (err);
1130 
1131         if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1132                 return (EINVAL);
1133 
1134         return (zfs_secpolicy_write_perms(zc->zc_name,
1135             userquota_perms[zc->zc_objset_type], cr));
1136 }
1137 
1138 /* ARGSUSED */
1139 static int
1140 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1141 {
1142         return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1143             NULL, cr));
1144 }
1145 
1146 /* ARGSUSED */
1147 static int
1148 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1149 {
1150         return (zfs_secpolicy_write_perms(zc->zc_name,
1151             ZFS_DELEG_PERM_HOLD, cr));
1152 }
1153 
1154 /* ARGSUSED */
1155 static int
1156 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1157 {
1158         return (zfs_secpolicy_write_perms(zc->zc_name,
1159             ZFS_DELEG_PERM_RELEASE, cr));
1160 }
1161 
1162 /*
1163  * Policy for allowing temporary snapshots to be taken or released
1164  */
1165 static int
1166 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1167 {
1168         /*
1169          * A temporary snapshot is the same as a snapshot,
1170          * hold, destroy and release all rolled into one.
1171          * Delegated diff alone is sufficient that we allow this.
1172          */
1173         int error;
1174 
1175         if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1176             ZFS_DELEG_PERM_DIFF, cr)) == 0)
1177                 return (0);
1178 
1179         error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1180         if (!error)
1181                 error = zfs_secpolicy_hold(zc, innvl, cr);
1182         if (!error)
1183                 error = zfs_secpolicy_release(zc, innvl, cr);
1184         if (!error)
1185                 error = zfs_secpolicy_destroy(zc, innvl, cr);
1186         return (error);
1187 }
1188 
1189 /*
1190  * Returns the nvlist as specified by the user in the zfs_cmd_t.
1191  */
1192 static int
1193 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1194 {
1195         char *packed;
1196         int error;
1197         nvlist_t *list = NULL;
1198 
1199         /*
1200          * Read in and unpack the user-supplied nvlist.
1201          */
1202         if (size == 0)
1203                 return (EINVAL);
1204 
1205         packed = kmem_alloc(size, KM_SLEEP);
1206 
1207         if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1208             iflag)) != 0) {
1209                 kmem_free(packed, size);
1210                 return (error);
1211         }
1212 
1213         if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1214                 kmem_free(packed, size);
1215                 return (error);
1216         }
1217 
1218         kmem_free(packed, size);
1219 
1220         *nvp = list;
1221         return (0);
1222 }
1223 
1224 /*
1225  * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1226  * Entries will be removed from the end of the nvlist, and one int32 entry
1227  * named "N_MORE_ERRORS" will be added indicating how many entries were
1228  * removed.
1229  */
1230 static int
1231 nvlist_smush(nvlist_t *errors, size_t max)
1232 {
1233         size_t size;
1234 
1235         size = fnvlist_size(errors);
1236 
1237         if (size > max) {
1238                 nvpair_t *more_errors;
1239                 int n = 0;
1240 
1241                 if (max < 1024)
1242                         return (ENOMEM);
1243 
1244                 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1245                 more_errors = nvlist_prev_nvpair(errors, NULL);
1246 
1247                 do {
1248                         nvpair_t *pair = nvlist_prev_nvpair(errors,
1249                             more_errors);
1250                         fnvlist_remove_nvpair(errors, pair);
1251                         n++;
1252                         size = fnvlist_size(errors);
1253                 } while (size > max);
1254 
1255                 fnvlist_remove_nvpair(errors, more_errors);
1256                 fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1257                 ASSERT3U(fnvlist_size(errors), <=, max);
1258         }
1259 
1260         return (0);
1261 }
1262 
1263 static int
1264 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1265 {
1266         char *packed = NULL;
1267         int error = 0;
1268         size_t size;
1269 
1270         size = fnvlist_size(nvl);
1271 
1272         if (size > zc->zc_nvlist_dst_size) {
1273                 error = ENOMEM;
1274         } else {
1275                 packed = fnvlist_pack(nvl, &size);
1276                 if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1277                     size, zc->zc_iflags) != 0)
1278                         error = EFAULT;
1279                 fnvlist_pack_free(packed, size);
1280         }
1281 
1282         zc->zc_nvlist_dst_size = size;
1283         zc->zc_nvlist_dst_filled = B_TRUE;
1284         return (error);
1285 }
1286 
1287 static int
1288 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1289 {
1290         objset_t *os;
1291         int error;
1292 
1293         error = dmu_objset_hold(dsname, FTAG, &os);
1294         if (error)
1295                 return (error);
1296         if (dmu_objset_type(os) != DMU_OST_ZFS) {
1297                 dmu_objset_rele(os, FTAG);
1298                 return (EINVAL);
1299         }
1300 
1301         mutex_enter(&os->os_user_ptr_lock);
1302         *zfvp = dmu_objset_get_user(os);
1303         if (*zfvp) {
1304                 VFS_HOLD((*zfvp)->z_vfs);
1305         } else {
1306                 error = ESRCH;
1307         }
1308         mutex_exit(&os->os_user_ptr_lock);
1309         dmu_objset_rele(os, FTAG);
1310         return (error);
1311 }
1312 
1313 /*
1314  * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1315  * case its z_vfs will be NULL, and it will be opened as the owner.
1316  * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1317  * which prevents all vnode ops from running.
1318  */
1319 static int
1320 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1321 {
1322         int error = 0;
1323 
1324         if (getzfsvfs(name, zfvp) != 0)
1325                 error = zfsvfs_create(name, zfvp);
1326         if (error == 0) {
1327                 rrw_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1328                     RW_READER, tag);
1329                 if ((*zfvp)->z_unmounted) {
1330                         /*
1331                          * XXX we could probably try again, since the unmounting
1332                          * thread should be just about to disassociate the
1333                          * objset from the zfsvfs.
1334                          */
1335                         rrw_exit(&(*zfvp)->z_teardown_lock, tag);
1336                         return (EBUSY);
1337                 }
1338         }
1339         return (error);
1340 }
1341 
1342 static void
1343 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1344 {
1345         rrw_exit(&zfsvfs->z_teardown_lock, tag);
1346 
1347         if (zfsvfs->z_vfs) {
1348                 VFS_RELE(zfsvfs->z_vfs);
1349         } else {
1350                 dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1351                 zfsvfs_free(zfsvfs);
1352         }
1353 }
1354 
1355 static int
1356 zfs_ioc_pool_create(zfs_cmd_t *zc)
1357 {
1358         int error;
1359         nvlist_t *config, *props = NULL;
1360         nvlist_t *rootprops = NULL;
1361         nvlist_t *zplprops = NULL;
1362 
1363         if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1364             zc->zc_iflags, &config))
1365                 return (error);
1366 
1367         if (zc->zc_nvlist_src_size != 0 && (error =
1368             get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1369             zc->zc_iflags, &props))) {
1370                 nvlist_free(config);
1371                 return (error);
1372         }
1373 
1374         if (props) {
1375                 nvlist_t *nvl = NULL;
1376                 uint64_t version = SPA_VERSION;
1377 
1378                 (void) nvlist_lookup_uint64(props,
1379                     zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1380                 if (!SPA_VERSION_IS_SUPPORTED(version)) {
1381                         error = EINVAL;
1382                         goto pool_props_bad;
1383                 }
1384                 (void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1385                 if (nvl) {
1386                         error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1387                         if (error != 0) {
1388                                 nvlist_free(config);
1389                                 nvlist_free(props);
1390                                 return (error);
1391                         }
1392                         (void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1393                 }
1394                 VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1395                 error = zfs_fill_zplprops_root(version, rootprops,
1396                     zplprops, NULL);
1397                 if (error)
1398                         goto pool_props_bad;
1399         }
1400 
1401         error = spa_create(zc->zc_name, config, props, zplprops);
1402 
1403         /*
1404          * Set the remaining root properties
1405          */
1406         if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1407             ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1408                 (void) spa_destroy(zc->zc_name);
1409 
1410 pool_props_bad:
1411         nvlist_free(rootprops);
1412         nvlist_free(zplprops);
1413         nvlist_free(config);
1414         nvlist_free(props);
1415 
1416         return (error);
1417 }
1418 
1419 static int
1420 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1421 {
1422         int error;
1423         zfs_log_history(zc);
1424         error = spa_destroy(zc->zc_name);
1425         if (error == 0)
1426                 zvol_remove_minors(zc->zc_name);
1427         return (error);
1428 }
1429 
1430 static int
1431 zfs_ioc_pool_import(zfs_cmd_t *zc)
1432 {
1433         nvlist_t *config, *props = NULL;
1434         uint64_t guid;
1435         int error;
1436 
1437         if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1438             zc->zc_iflags, &config)) != 0)
1439                 return (error);
1440 
1441         if (zc->zc_nvlist_src_size != 0 && (error =
1442             get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1443             zc->zc_iflags, &props))) {
1444                 nvlist_free(config);
1445                 return (error);
1446         }
1447 
1448         if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1449             guid != zc->zc_guid)
1450                 error = EINVAL;
1451         else
1452                 error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1453 
1454         if (zc->zc_nvlist_dst != 0) {
1455                 int err;
1456 
1457                 if ((err = put_nvlist(zc, config)) != 0)
1458                         error = err;
1459         }
1460 
1461         nvlist_free(config);
1462 
1463         if (props)
1464                 nvlist_free(props);
1465 
1466         return (error);
1467 }
1468 
1469 static int
1470 zfs_ioc_pool_export(zfs_cmd_t *zc)
1471 {
1472         int error;
1473         boolean_t force = (boolean_t)zc->zc_cookie;
1474         boolean_t hardforce = (boolean_t)zc->zc_guid;
1475 
1476         zfs_log_history(zc);
1477         error = spa_export(zc->zc_name, NULL, force, hardforce);
1478         if (error == 0)
1479                 zvol_remove_minors(zc->zc_name);
1480         return (error);
1481 }
1482 
1483 static int
1484 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1485 {
1486         nvlist_t *configs;
1487         int error;
1488 
1489         if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1490                 return (EEXIST);
1491 
1492         error = put_nvlist(zc, configs);
1493 
1494         nvlist_free(configs);
1495 
1496         return (error);
1497 }
1498 
1499 /*
1500  * inputs:
1501  * zc_name              name of the pool
1502  *
1503  * outputs:
1504  * zc_cookie            real errno
1505  * zc_nvlist_dst        config nvlist
1506  * zc_nvlist_dst_size   size of config nvlist
1507  */
1508 static int
1509 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1510 {
1511         nvlist_t *config;
1512         int error;
1513         int ret = 0;
1514 
1515         error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1516             sizeof (zc->zc_value));
1517 
1518         if (config != NULL) {
1519                 ret = put_nvlist(zc, config);
1520                 nvlist_free(config);
1521 
1522                 /*
1523                  * The config may be present even if 'error' is non-zero.
1524                  * In this case we return success, and preserve the real errno
1525                  * in 'zc_cookie'.
1526                  */
1527                 zc->zc_cookie = error;
1528         } else {
1529                 ret = error;
1530         }
1531 
1532         return (ret);
1533 }
1534 
1535 /*
1536  * Try to import the given pool, returning pool stats as appropriate so that
1537  * user land knows which devices are available and overall pool health.
1538  */
1539 static int
1540 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1541 {
1542         nvlist_t *tryconfig, *config;
1543         int error;
1544 
1545         if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1546             zc->zc_iflags, &tryconfig)) != 0)
1547                 return (error);
1548 
1549         config = spa_tryimport(tryconfig);
1550 
1551         nvlist_free(tryconfig);
1552 
1553         if (config == NULL)
1554                 return (EINVAL);
1555 
1556         error = put_nvlist(zc, config);
1557         nvlist_free(config);
1558 
1559         return (error);
1560 }
1561 
1562 /*
1563  * inputs:
1564  * zc_name              name of the pool
1565  * zc_cookie            scan func (pool_scan_func_t)
1566  */
1567 static int
1568 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1569 {
1570         spa_t *spa;
1571         int error;
1572 
1573         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1574                 return (error);
1575 
1576         if (zc->zc_cookie == POOL_SCAN_NONE)
1577                 error = spa_scan_stop(spa);
1578         else
1579                 error = spa_scan(spa, zc->zc_cookie);
1580 
1581         spa_close(spa, FTAG);
1582 
1583         return (error);
1584 }
1585 
1586 static int
1587 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1588 {
1589         spa_t *spa;
1590         int error;
1591 
1592         error = spa_open(zc->zc_name, &spa, FTAG);
1593         if (error == 0) {
1594                 spa_freeze(spa);
1595                 spa_close(spa, FTAG);
1596         }
1597         return (error);
1598 }
1599 
1600 static int
1601 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1602 {
1603         spa_t *spa;
1604         int error;
1605 
1606         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1607                 return (error);
1608 
1609         if (zc->zc_cookie < spa_version(spa) ||
1610             !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1611                 spa_close(spa, FTAG);
1612                 return (EINVAL);
1613         }
1614 
1615         spa_upgrade(spa, zc->zc_cookie);
1616         spa_close(spa, FTAG);
1617 
1618         return (error);
1619 }
1620 
1621 static int
1622 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1623 {
1624         spa_t *spa;
1625         char *hist_buf;
1626         uint64_t size;
1627         int error;
1628 
1629         if ((size = zc->zc_history_len) == 0)
1630                 return (EINVAL);
1631 
1632         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1633                 return (error);
1634 
1635         if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1636                 spa_close(spa, FTAG);
1637                 return (ENOTSUP);
1638         }
1639 
1640         hist_buf = kmem_alloc(size, KM_SLEEP);
1641         if ((error = spa_history_get(spa, &zc->zc_history_offset,
1642             &zc->zc_history_len, hist_buf)) == 0) {
1643                 error = ddi_copyout(hist_buf,
1644                     (void *)(uintptr_t)zc->zc_history,
1645                     zc->zc_history_len, zc->zc_iflags);
1646         }
1647 
1648         spa_close(spa, FTAG);
1649         kmem_free(hist_buf, size);
1650         return (error);
1651 }
1652 
1653 static int
1654 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1655 {
1656         spa_t *spa;
1657         int error;
1658 
1659         error = spa_open(zc->zc_name, &spa, FTAG);
1660         if (error == 0) {
1661                 error = spa_change_guid(spa);
1662                 spa_close(spa, FTAG);
1663         }
1664         return (error);
1665 }
1666 
1667 static int
1668 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1669 {
1670         int error;
1671 
1672         if (error = dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value))
1673                 return (error);
1674 
1675         return (0);
1676 }
1677 
1678 /*
1679  * inputs:
1680  * zc_name              name of filesystem
1681  * zc_obj               object to find
1682  *
1683  * outputs:
1684  * zc_value             name of object
1685  */
1686 static int
1687 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1688 {
1689         objset_t *os;
1690         int error;
1691 
1692         /* XXX reading from objset not owned */
1693         if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1694                 return (error);
1695         if (dmu_objset_type(os) != DMU_OST_ZFS) {
1696                 dmu_objset_rele(os, FTAG);
1697                 return (EINVAL);
1698         }
1699         error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1700             sizeof (zc->zc_value));
1701         dmu_objset_rele(os, FTAG);
1702 
1703         return (error);
1704 }
1705 
1706 /*
1707  * inputs:
1708  * zc_name              name of filesystem
1709  * zc_obj               object to find
1710  *
1711  * outputs:
1712  * zc_stat              stats on object
1713  * zc_value             path to object
1714  */
1715 static int
1716 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1717 {
1718         objset_t *os;
1719         int error;
1720 
1721         /* XXX reading from objset not owned */
1722         if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1723                 return (error);
1724         if (dmu_objset_type(os) != DMU_OST_ZFS) {
1725                 dmu_objset_rele(os, FTAG);
1726                 return (EINVAL);
1727         }
1728         error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1729             sizeof (zc->zc_value));
1730         dmu_objset_rele(os, FTAG);
1731 
1732         return (error);
1733 }
1734 
1735 static int
1736 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1737 {
1738         spa_t *spa;
1739         int error;
1740         nvlist_t *config, **l2cache, **spares;
1741         uint_t nl2cache = 0, nspares = 0;
1742 
1743         error = spa_open(zc->zc_name, &spa, FTAG);
1744         if (error != 0)
1745                 return (error);
1746 
1747         error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1748             zc->zc_iflags, &config);
1749         (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1750             &l2cache, &nl2cache);
1751 
1752         (void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1753             &spares, &nspares);
1754 
1755         /*
1756          * A root pool with concatenated devices is not supported.
1757          * Thus, can not add a device to a root pool.
1758          *
1759          * Intent log device can not be added to a rootpool because
1760          * during mountroot, zil is replayed, a seperated log device
1761          * can not be accessed during the mountroot time.
1762          *
1763          * l2cache and spare devices are ok to be added to a rootpool.
1764          */
1765         if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1766                 nvlist_free(config);
1767                 spa_close(spa, FTAG);
1768                 return (EDOM);
1769         }
1770 
1771         if (error == 0) {
1772                 error = spa_vdev_add(spa, config);
1773                 nvlist_free(config);
1774         }
1775         spa_close(spa, FTAG);
1776         return (error);
1777 }
1778 
1779 /*
1780  * inputs:
1781  * zc_name              name of the pool
1782  * zc_nvlist_conf       nvlist of devices to remove
1783  * zc_cookie            to stop the remove?
1784  */
1785 static int
1786 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1787 {
1788         spa_t *spa;
1789         int error;
1790 
1791         error = spa_open(zc->zc_name, &spa, FTAG);
1792         if (error != 0)
1793                 return (error);
1794         error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1795         spa_close(spa, FTAG);
1796         return (error);
1797 }
1798 
1799 static int
1800 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1801 {
1802         spa_t *spa;
1803         int error;
1804         vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1805 
1806         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1807                 return (error);
1808         switch (zc->zc_cookie) {
1809         case VDEV_STATE_ONLINE:
1810                 error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1811                 break;
1812 
1813         case VDEV_STATE_OFFLINE:
1814                 error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1815                 break;
1816 
1817         case VDEV_STATE_FAULTED:
1818                 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1819                     zc->zc_obj != VDEV_AUX_EXTERNAL)
1820                         zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1821 
1822                 error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1823                 break;
1824 
1825         case VDEV_STATE_DEGRADED:
1826                 if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1827                     zc->zc_obj != VDEV_AUX_EXTERNAL)
1828                         zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1829 
1830                 error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1831                 break;
1832 
1833         default:
1834                 error = EINVAL;
1835         }
1836         zc->zc_cookie = newstate;
1837         spa_close(spa, FTAG);
1838         return (error);
1839 }
1840 
1841 static int
1842 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1843 {
1844         spa_t *spa;
1845         int replacing = zc->zc_cookie;
1846         nvlist_t *config;
1847         int error;
1848 
1849         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1850                 return (error);
1851 
1852         if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1853             zc->zc_iflags, &config)) == 0) {
1854                 error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1855                 nvlist_free(config);
1856         }
1857 
1858         spa_close(spa, FTAG);
1859         return (error);
1860 }
1861 
1862 static int
1863 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1864 {
1865         spa_t *spa;
1866         int error;
1867 
1868         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1869                 return (error);
1870 
1871         error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1872 
1873         spa_close(spa, FTAG);
1874         return (error);
1875 }
1876 
1877 static int
1878 zfs_ioc_vdev_split(zfs_cmd_t *zc)
1879 {
1880         spa_t *spa;
1881         nvlist_t *config, *props = NULL;
1882         int error;
1883         boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
1884 
1885         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1886                 return (error);
1887 
1888         if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1889             zc->zc_iflags, &config)) {
1890                 spa_close(spa, FTAG);
1891                 return (error);
1892         }
1893 
1894         if (zc->zc_nvlist_src_size != 0 && (error =
1895             get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1896             zc->zc_iflags, &props))) {
1897                 spa_close(spa, FTAG);
1898                 nvlist_free(config);
1899                 return (error);
1900         }
1901 
1902         error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
1903 
1904         spa_close(spa, FTAG);
1905 
1906         nvlist_free(config);
1907         nvlist_free(props);
1908 
1909         return (error);
1910 }
1911 
1912 static int
1913 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
1914 {
1915         spa_t *spa;
1916         char *path = zc->zc_value;
1917         uint64_t guid = zc->zc_guid;
1918         int error;
1919 
1920         error = spa_open(zc->zc_name, &spa, FTAG);
1921         if (error != 0)
1922                 return (error);
1923 
1924         error = spa_vdev_setpath(spa, guid, path);
1925         spa_close(spa, FTAG);
1926         return (error);
1927 }
1928 
1929 static int
1930 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
1931 {
1932         spa_t *spa;
1933         char *fru = zc->zc_value;
1934         uint64_t guid = zc->zc_guid;
1935         int error;
1936 
1937         error = spa_open(zc->zc_name, &spa, FTAG);
1938         if (error != 0)
1939                 return (error);
1940 
1941         error = spa_vdev_setfru(spa, guid, fru);
1942         spa_close(spa, FTAG);
1943         return (error);
1944 }
1945 
1946 static int
1947 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
1948 {
1949         int error = 0;
1950         nvlist_t *nv;
1951 
1952         dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1953 
1954         if (zc->zc_nvlist_dst != 0 &&
1955             (error = dsl_prop_get_all(os, &nv)) == 0) {
1956                 dmu_objset_stats(os, nv);
1957                 /*
1958                  * NB: zvol_get_stats() will read the objset contents,
1959                  * which we aren't supposed to do with a
1960                  * DS_MODE_USER hold, because it could be
1961                  * inconsistent.  So this is a bit of a workaround...
1962                  * XXX reading with out owning
1963                  */
1964                 if (!zc->zc_objset_stats.dds_inconsistent &&
1965                     dmu_objset_type(os) == DMU_OST_ZVOL) {
1966                         error = zvol_get_stats(os, nv);
1967                         if (error == EIO)
1968                                 return (error);
1969                         VERIFY0(error);
1970                 }
1971                 error = put_nvlist(zc, nv);
1972                 nvlist_free(nv);
1973         }
1974 
1975         return (error);
1976 }
1977 
1978 /*
1979  * inputs:
1980  * zc_name              name of filesystem
1981  * zc_nvlist_dst_size   size of buffer for property nvlist
1982  *
1983  * outputs:
1984  * zc_objset_stats      stats
1985  * zc_nvlist_dst        property nvlist
1986  * zc_nvlist_dst_size   size of property nvlist
1987  */
1988 static int
1989 zfs_ioc_objset_stats(zfs_cmd_t *zc)
1990 {
1991         objset_t *os = NULL;
1992         int error;
1993 
1994         if (error = dmu_objset_hold(zc->zc_name, FTAG, &os))
1995                 return (error);
1996 
1997         error = zfs_ioc_objset_stats_impl(zc, os);
1998 
1999         dmu_objset_rele(os, FTAG);
2000 
2001         return (error);
2002 }
2003 
2004 /*
2005  * inputs:
2006  * zc_name              name of filesystem
2007  * zc_nvlist_dst_size   size of buffer for property nvlist
2008  *
2009  * outputs:
2010  * zc_nvlist_dst        received property nvlist
2011  * zc_nvlist_dst_size   size of received property nvlist
2012  *
2013  * Gets received properties (distinct from local properties on or after
2014  * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2015  * local property values.
2016  */
2017 static int
2018 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2019 {
2020         objset_t *os = NULL;
2021         int error;
2022         nvlist_t *nv;
2023 
2024         if (error = dmu_objset_hold(zc->zc_name, FTAG, &os))
2025                 return (error);
2026 
2027         /*
2028          * Without this check, we would return local property values if the
2029          * caller has not already received properties on or after
2030          * SPA_VERSION_RECVD_PROPS.
2031          */
2032         if (!dsl_prop_get_hasrecvd(os)) {
2033                 dmu_objset_rele(os, FTAG);
2034                 return (ENOTSUP);
2035         }
2036 
2037         if (zc->zc_nvlist_dst != 0 &&
2038             (error = dsl_prop_get_received(os, &nv)) == 0) {
2039                 error = put_nvlist(zc, nv);
2040                 nvlist_free(nv);
2041         }
2042 
2043         dmu_objset_rele(os, FTAG);
2044         return (error);
2045 }
2046 
2047 static int
2048 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2049 {
2050         uint64_t value;
2051         int error;
2052 
2053         /*
2054          * zfs_get_zplprop() will either find a value or give us
2055          * the default value (if there is one).
2056          */
2057         if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2058                 return (error);
2059         VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2060         return (0);
2061 }
2062 
2063 /*
2064  * inputs:
2065  * zc_name              name of filesystem
2066  * zc_nvlist_dst_size   size of buffer for zpl property nvlist
2067  *
2068  * outputs:
2069  * zc_nvlist_dst        zpl property nvlist
2070  * zc_nvlist_dst_size   size of zpl property nvlist
2071  */
2072 static int
2073 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2074 {
2075         objset_t *os;
2076         int err;
2077 
2078         /* XXX reading without owning */
2079         if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2080                 return (err);
2081 
2082         dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2083 
2084         /*
2085          * NB: nvl_add_zplprop() will read the objset contents,
2086          * which we aren't supposed to do with a DS_MODE_USER
2087          * hold, because it could be inconsistent.
2088          */
2089         if (zc->zc_nvlist_dst != NULL &&
2090             !zc->zc_objset_stats.dds_inconsistent &&
2091             dmu_objset_type(os) == DMU_OST_ZFS) {
2092                 nvlist_t *nv;
2093 
2094                 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2095                 if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2096                     (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2097                     (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2098                     (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2099                         err = put_nvlist(zc, nv);
2100                 nvlist_free(nv);
2101         } else {
2102                 err = ENOENT;
2103         }
2104         dmu_objset_rele(os, FTAG);
2105         return (err);
2106 }
2107 
2108 static boolean_t
2109 dataset_name_hidden(const char *name)
2110 {
2111         /*
2112          * Skip over datasets that are not visible in this zone,
2113          * internal datasets (which have a $ in their name), and
2114          * temporary datasets (which have a % in their name).
2115          */
2116         if (strchr(name, '$') != NULL)
2117                 return (B_TRUE);
2118         if (strchr(name, '%') != NULL)
2119                 return (B_TRUE);
2120         if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
2121                 return (B_TRUE);
2122         return (B_FALSE);
2123 }
2124 
2125 /*
2126  * inputs:
2127  * zc_name              name of filesystem
2128  * zc_cookie            zap cursor
2129  * zc_nvlist_dst_size   size of buffer for property nvlist
2130  *
2131  * outputs:
2132  * zc_name              name of next filesystem
2133  * zc_cookie            zap cursor
2134  * zc_objset_stats      stats
2135  * zc_nvlist_dst        property nvlist
2136  * zc_nvlist_dst_size   size of property nvlist
2137  */
2138 static int
2139 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2140 {
2141         objset_t *os;
2142         int error;
2143         char *p;
2144         size_t orig_len = strlen(zc->zc_name);
2145 
2146 top:
2147         if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2148                 if (error == ENOENT)
2149                         error = ESRCH;
2150                 return (error);
2151         }
2152 
2153         p = strrchr(zc->zc_name, '/');
2154         if (p == NULL || p[1] != '\0')
2155                 (void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2156         p = zc->zc_name + strlen(zc->zc_name);
2157 
2158         /*
2159          * Pre-fetch the datasets.  dmu_objset_prefetch() always returns 0
2160          * but is not declared void because its called by dmu_objset_find().
2161          */
2162         if (zc->zc_cookie == 0) {
2163                 uint64_t cookie = 0;
2164                 int len = sizeof (zc->zc_name) - (p - zc->zc_name);
2165 
2166                 while (dmu_dir_list_next(os, len, p, NULL, &cookie) == 0) {
2167                         if (!dataset_name_hidden(zc->zc_name))
2168                                 (void) dmu_objset_prefetch(zc->zc_name, NULL);
2169                 }
2170         }
2171 
2172         do {
2173                 error = dmu_dir_list_next(os,
2174                     sizeof (zc->zc_name) - (p - zc->zc_name), p,
2175                     NULL, &zc->zc_cookie);
2176                 if (error == ENOENT)
2177                         error = ESRCH;
2178         } while (error == 0 && dataset_name_hidden(zc->zc_name));
2179         dmu_objset_rele(os, FTAG);
2180 
2181         /*
2182          * If it's an internal dataset (ie. with a '$' in its name),
2183          * don't try to get stats for it, otherwise we'll return ENOENT.
2184          */
2185         if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2186                 error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2187                 if (error == ENOENT) {
2188                         /* We lost a race with destroy, get the next one. */
2189                         zc->zc_name[orig_len] = '\0';
2190                         goto top;
2191                 }
2192         }
2193         return (error);
2194 }
2195 
2196 /*
2197  * inputs:
2198  * zc_name              name of filesystem
2199  * zc_cookie            zap cursor
2200  * zc_nvlist_dst_size   size of buffer for property nvlist
2201  *
2202  * outputs:
2203  * zc_name              name of next snapshot
2204  * zc_objset_stats      stats
2205  * zc_nvlist_dst        property nvlist
2206  * zc_nvlist_dst_size   size of property nvlist
2207  */
2208 static int
2209 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2210 {
2211         objset_t *os;
2212         int error;
2213 
2214 top:
2215         if (zc->zc_cookie == 0)
2216                 (void) dmu_objset_find(zc->zc_name, dmu_objset_prefetch,
2217                     NULL, DS_FIND_SNAPSHOTS);
2218 
2219         error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2220         if (error)
2221                 return (error == ENOENT ? ESRCH : error);
2222 
2223         /*
2224          * A dataset name of maximum length cannot have any snapshots,
2225          * so exit immediately.
2226          */
2227         if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2228                 dmu_objset_rele(os, FTAG);
2229                 return (ESRCH);
2230         }
2231 
2232         error = dmu_snapshot_list_next(os,
2233             sizeof (zc->zc_name) - strlen(zc->zc_name),
2234             zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2235             NULL);
2236 
2237         if (error == 0) {
2238                 dsl_dataset_t *ds;
2239                 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2240 
2241                 /*
2242                  * Since we probably don't have a hold on this snapshot,
2243                  * it's possible that the objsetid could have been destroyed
2244                  * and reused for a new objset. It's OK if this happens during
2245                  * a zfs send operation, since the new createtxg will be
2246                  * beyond the range we're interested in.
2247                  */
2248                 rw_enter(&dp->dp_config_rwlock, RW_READER);
2249                 error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2250                 rw_exit(&dp->dp_config_rwlock);
2251                 if (error) {
2252                         if (error == ENOENT) {
2253                                 /* Racing with destroy, get the next one. */
2254                                 *strchr(zc->zc_name, '@') = '\0';
2255                                 dmu_objset_rele(os, FTAG);
2256                                 goto top;
2257                         }
2258                 } else {
2259                         objset_t *ossnap;
2260 
2261                         error = dmu_objset_from_ds(ds, &ossnap);
2262                         if (error == 0)
2263                                 error = zfs_ioc_objset_stats_impl(zc, ossnap);
2264                         dsl_dataset_rele(ds, FTAG);
2265                 }
2266         } else if (error == ENOENT) {
2267                 error = ESRCH;
2268         }
2269 
2270         dmu_objset_rele(os, FTAG);
2271         /* if we failed, undo the @ that we tacked on to zc_name */
2272         if (error)
2273                 *strchr(zc->zc_name, '@') = '\0';
2274         return (error);
2275 }
2276 
2277 static int
2278 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2279 {
2280         const char *propname = nvpair_name(pair);
2281         uint64_t *valary;
2282         unsigned int vallen;
2283         const char *domain;
2284         char *dash;
2285         zfs_userquota_prop_t type;
2286         uint64_t rid;
2287         uint64_t quota;
2288         zfsvfs_t *zfsvfs;
2289         int err;
2290 
2291         if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2292                 nvlist_t *attrs;
2293                 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2294                 if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2295                     &pair) != 0)
2296                         return (EINVAL);
2297         }
2298 
2299         /*
2300          * A correctly constructed propname is encoded as
2301          * userquota@<rid>-<domain>.
2302          */
2303         if ((dash = strchr(propname, '-')) == NULL ||
2304             nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2305             vallen != 3)
2306                 return (EINVAL);
2307 
2308         domain = dash + 1;
2309         type = valary[0];
2310         rid = valary[1];
2311         quota = valary[2];
2312 
2313         err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2314         if (err == 0) {
2315                 err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2316                 zfsvfs_rele(zfsvfs, FTAG);
2317         }
2318 
2319         return (err);
2320 }
2321 
2322 /*
2323  * If the named property is one that has a special function to set its value,
2324  * return 0 on success and a positive error code on failure; otherwise if it is
2325  * not one of the special properties handled by this function, return -1.
2326  *
2327  * XXX: It would be better for callers of the property interface if we handled
2328  * these special cases in dsl_prop.c (in the dsl layer).
2329  */
2330 static int
2331 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2332     nvpair_t *pair)
2333 {
2334         const char *propname = nvpair_name(pair);
2335         zfs_prop_t prop = zfs_name_to_prop(propname);
2336         uint64_t intval;
2337         int err;
2338 
2339         if (prop == ZPROP_INVAL) {
2340                 if (zfs_prop_userquota(propname))
2341                         return (zfs_prop_set_userquota(dsname, pair));
2342                 return (-1);
2343         }
2344 
2345         if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2346                 nvlist_t *attrs;
2347                 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2348                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2349                     &pair) == 0);
2350         }
2351 
2352         if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2353                 return (-1);
2354 
2355         VERIFY(0 == nvpair_value_uint64(pair, &intval));
2356 
2357         switch (prop) {
2358         case ZFS_PROP_QUOTA:
2359                 err = dsl_dir_set_quota(dsname, source, intval);
2360                 break;
2361         case ZFS_PROP_REFQUOTA:
2362                 err = dsl_dataset_set_quota(dsname, source, intval);
2363                 break;
2364         case ZFS_PROP_RESERVATION:
2365                 err = dsl_dir_set_reservation(dsname, source, intval);
2366                 break;
2367         case ZFS_PROP_REFRESERVATION:
2368                 err = dsl_dataset_set_reservation(dsname, source, intval);
2369                 break;
2370         case ZFS_PROP_VOLSIZE:
2371                 err = zvol_set_volsize(dsname, ddi_driver_major(zfs_dip),
2372                     intval);
2373                 break;
2374         case ZFS_PROP_VERSION:
2375         {
2376                 zfsvfs_t *zfsvfs;
2377 
2378                 if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2379                         break;
2380 
2381                 err = zfs_set_version(zfsvfs, intval);
2382                 zfsvfs_rele(zfsvfs, FTAG);
2383 
2384                 if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2385                         zfs_cmd_t *zc;
2386 
2387                         zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2388                         (void) strcpy(zc->zc_name, dsname);
2389                         (void) zfs_ioc_userspace_upgrade(zc);
2390                         kmem_free(zc, sizeof (zfs_cmd_t));
2391                 }
2392                 break;
2393         }
2394         case ZFS_PROP_COMPRESSION:
2395         {
2396                 if (intval == ZIO_COMPRESS_LZ4) {
2397                         zfeature_info_t *feature =
2398                             &spa_feature_table[SPA_FEATURE_LZ4_COMPRESS];
2399                         spa_t *spa;
2400                         dsl_pool_t *dp;
2401 
2402                         if ((err = spa_open(dsname, &spa, FTAG)) != 0)
2403                                 return (err);
2404 
2405                         dp = spa->spa_dsl_pool;
2406 
2407                         /*
2408                          * Setting the LZ4 compression algorithm activates
2409                          * the feature.
2410                          */
2411                         if (!spa_feature_is_active(spa, feature)) {
2412                                 if ((err = zfs_prop_activate_feature(dp,
2413                                     feature)) != 0) {
2414                                         spa_close(spa, FTAG);
2415                                         return (err);
2416                                 }
2417                         }
2418 
2419                         spa_close(spa, FTAG);
2420                 }
2421                 /*
2422                  * We still want the default set action to be performed in the
2423                  * caller, we only performed zfeature settings here.
2424                  */
2425                 err = -1;
2426                 break;
2427         }
2428 
2429         default:
2430                 err = -1;
2431         }
2432 
2433         return (err);
2434 }
2435 
2436 /*
2437  * This function is best effort. If it fails to set any of the given properties,
2438  * it continues to set as many as it can and returns the last error
2439  * encountered. If the caller provides a non-NULL errlist, it will be filled in
2440  * with the list of names of all the properties that failed along with the
2441  * corresponding error numbers.
2442  *
2443  * If every property is set successfully, zero is returned and errlist is not
2444  * modified.
2445  */
2446 int
2447 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2448     nvlist_t *errlist)
2449 {
2450         nvpair_t *pair;
2451         nvpair_t *propval;
2452         int rv = 0;
2453         uint64_t intval;
2454         char *strval;
2455         nvlist_t *genericnvl = fnvlist_alloc();
2456         nvlist_t *retrynvl = fnvlist_alloc();
2457 
2458 retry:
2459         pair = NULL;
2460         while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2461                 const char *propname = nvpair_name(pair);
2462                 zfs_prop_t prop = zfs_name_to_prop(propname);
2463                 int err = 0;
2464 
2465                 /* decode the property value */
2466                 propval = pair;
2467                 if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2468                         nvlist_t *attrs;
2469                         attrs = fnvpair_value_nvlist(pair);
2470                         if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2471                             &propval) != 0)
2472                                 err = EINVAL;
2473                 }
2474 
2475                 /* Validate value type */
2476                 if (err == 0 && prop == ZPROP_INVAL) {
2477                         if (zfs_prop_user(propname)) {
2478                                 if (nvpair_type(propval) != DATA_TYPE_STRING)
2479                                         err = EINVAL;
2480                         } else if (zfs_prop_userquota(propname)) {
2481                                 if (nvpair_type(propval) !=
2482                                     DATA_TYPE_UINT64_ARRAY)
2483                                         err = EINVAL;
2484                         } else {
2485                                 err = EINVAL;
2486                         }
2487                 } else if (err == 0) {
2488                         if (nvpair_type(propval) == DATA_TYPE_STRING) {
2489                                 if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2490                                         err = EINVAL;
2491                         } else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2492                                 const char *unused;
2493 
2494                                 intval = fnvpair_value_uint64(propval);
2495 
2496                                 switch (zfs_prop_get_type(prop)) {
2497                                 case PROP_TYPE_NUMBER:
2498                                         break;
2499                                 case PROP_TYPE_STRING:
2500                                         err = EINVAL;
2501                                         break;
2502                                 case PROP_TYPE_INDEX:
2503                                         if (zfs_prop_index_to_string(prop,
2504                                             intval, &unused) != 0)
2505                                                 err = EINVAL;
2506                                         break;
2507                                 default:
2508                                         cmn_err(CE_PANIC,
2509                                             "unknown property type");
2510                                 }
2511                         } else {
2512                                 err = EINVAL;
2513                         }
2514                 }
2515 
2516                 /* Validate permissions */
2517                 if (err == 0)
2518                         err = zfs_check_settable(dsname, pair, CRED());
2519 
2520                 if (err == 0) {
2521                         err = zfs_prop_set_special(dsname, source, pair);
2522                         if (err == -1) {
2523                                 /*
2524                                  * For better performance we build up a list of
2525                                  * properties to set in a single transaction.
2526                                  */
2527                                 err = nvlist_add_nvpair(genericnvl, pair);
2528                         } else if (err != 0 && nvl != retrynvl) {
2529                                 /*
2530                                  * This may be a spurious error caused by
2531                                  * receiving quota and reservation out of order.
2532                                  * Try again in a second pass.
2533                                  */
2534                                 err = nvlist_add_nvpair(retrynvl, pair);
2535                         }
2536                 }
2537 
2538                 if (err != 0) {
2539                         if (errlist != NULL)
2540                                 fnvlist_add_int32(errlist, propname, err);
2541                         rv = err;
2542                 }
2543         }
2544 
2545         if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2546                 nvl = retrynvl;
2547                 goto retry;
2548         }
2549 
2550         if (!nvlist_empty(genericnvl) &&
2551             dsl_props_set(dsname, source, genericnvl) != 0) {
2552                 /*
2553                  * If this fails, we still want to set as many properties as we
2554                  * can, so try setting them individually.
2555                  */
2556                 pair = NULL;
2557                 while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2558                         const char *propname = nvpair_name(pair);
2559                         int err = 0;
2560 
2561                         propval = pair;
2562                         if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2563                                 nvlist_t *attrs;
2564                                 attrs = fnvpair_value_nvlist(pair);
2565                                 propval = fnvlist_lookup_nvpair(attrs,
2566                                     ZPROP_VALUE);
2567                         }
2568 
2569                         if (nvpair_type(propval) == DATA_TYPE_STRING) {
2570                                 strval = fnvpair_value_string(propval);
2571                                 err = dsl_prop_set(dsname, propname, source, 1,
2572                                     strlen(strval) + 1, strval);
2573                         } else {
2574                                 intval = fnvpair_value_uint64(propval);
2575                                 err = dsl_prop_set(dsname, propname, source, 8,
2576                                     1, &intval);
2577                         }
2578 
2579                         if (err != 0) {
2580                                 if (errlist != NULL) {
2581                                         fnvlist_add_int32(errlist, propname,
2582                                             err);
2583                                 }
2584                                 rv = err;
2585                         }
2586                 }
2587         }
2588         nvlist_free(genericnvl);
2589         nvlist_free(retrynvl);
2590 
2591         return (rv);
2592 }
2593 
2594 /*
2595  * Check that all the properties are valid user properties.
2596  */
2597 static int
2598 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2599 {
2600         nvpair_t *pair = NULL;
2601         int error = 0;
2602 
2603         while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2604                 const char *propname = nvpair_name(pair);
2605                 char *valstr;
2606 
2607                 if (!zfs_prop_user(propname) ||
2608                     nvpair_type(pair) != DATA_TYPE_STRING)
2609                         return (EINVAL);
2610 
2611                 if (error = zfs_secpolicy_write_perms(fsname,
2612                     ZFS_DELEG_PERM_USERPROP, CRED()))
2613                         return (error);
2614 
2615                 if (strlen(propname) >= ZAP_MAXNAMELEN)
2616                         return (ENAMETOOLONG);
2617 
2618                 VERIFY(nvpair_value_string(pair, &valstr) == 0);
2619                 if (strlen(valstr) >= ZAP_MAXVALUELEN)
2620                         return (E2BIG);
2621         }
2622         return (0);
2623 }
2624 
2625 static void
2626 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2627 {
2628         nvpair_t *pair;
2629 
2630         VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2631 
2632         pair = NULL;
2633         while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2634                 if (nvlist_exists(skipped, nvpair_name(pair)))
2635                         continue;
2636 
2637                 VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2638         }
2639 }
2640 
2641 static int
2642 clear_received_props(objset_t *os, const char *fs, nvlist_t *props,
2643     nvlist_t *skipped)
2644 {
2645         int err = 0;
2646         nvlist_t *cleared_props = NULL;
2647         props_skip(props, skipped, &cleared_props);
2648         if (!nvlist_empty(cleared_props)) {
2649                 /*
2650                  * Acts on local properties until the dataset has received
2651                  * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2652                  */
2653                 zprop_source_t flags = (ZPROP_SRC_NONE |
2654                     (dsl_prop_get_hasrecvd(os) ? ZPROP_SRC_RECEIVED : 0));
2655                 err = zfs_set_prop_nvlist(fs, flags, cleared_props, NULL);
2656         }
2657         nvlist_free(cleared_props);
2658         return (err);
2659 }
2660 
2661 /*
2662  * inputs:
2663  * zc_name              name of filesystem
2664  * zc_value             name of property to set
2665  * zc_nvlist_src{_size} nvlist of properties to apply
2666  * zc_cookie            received properties flag
2667  *
2668  * outputs:
2669  * zc_nvlist_dst{_size} error for each unapplied received property
2670  */
2671 static int
2672 zfs_ioc_set_prop(zfs_cmd_t *zc)
2673 {
2674         nvlist_t *nvl;
2675         boolean_t received = zc->zc_cookie;
2676         zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2677             ZPROP_SRC_LOCAL);
2678         nvlist_t *errors;
2679         int error;
2680 
2681         if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2682             zc->zc_iflags, &nvl)) != 0)
2683                 return (error);
2684 
2685         if (received) {
2686                 nvlist_t *origprops;
2687                 objset_t *os;
2688 
2689                 if (dmu_objset_hold(zc->zc_name, FTAG, &os) == 0) {
2690                         if (dsl_prop_get_received(os, &origprops) == 0) {
2691                                 (void) clear_received_props(os,
2692                                     zc->zc_name, origprops, nvl);
2693                                 nvlist_free(origprops);
2694                         }
2695 
2696                         dsl_prop_set_hasrecvd(os);
2697                         dmu_objset_rele(os, FTAG);
2698                 }
2699         }
2700 
2701         errors = fnvlist_alloc();
2702         error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2703 
2704         if (zc->zc_nvlist_dst != NULL && errors != NULL) {
2705                 (void) put_nvlist(zc, errors);
2706         }
2707 
2708         nvlist_free(errors);
2709         nvlist_free(nvl);
2710         return (error);
2711 }
2712 
2713 /*
2714  * inputs:
2715  * zc_name              name of filesystem
2716  * zc_value             name of property to inherit
2717  * zc_cookie            revert to received value if TRUE
2718  *
2719  * outputs:             none
2720  */
2721 static int
2722 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2723 {
2724         const char *propname = zc->zc_value;
2725         zfs_prop_t prop = zfs_name_to_prop(propname);
2726         boolean_t received = zc->zc_cookie;
2727         zprop_source_t source = (received
2728             ? ZPROP_SRC_NONE            /* revert to received value, if any */
2729             : ZPROP_SRC_INHERITED);     /* explicitly inherit */
2730 
2731         if (received) {
2732                 nvlist_t *dummy;
2733                 nvpair_t *pair;
2734                 zprop_type_t type;
2735                 int err;
2736 
2737                 /*
2738                  * zfs_prop_set_special() expects properties in the form of an
2739                  * nvpair with type info.
2740                  */
2741                 if (prop == ZPROP_INVAL) {
2742                         if (!zfs_prop_user(propname))
2743                                 return (EINVAL);
2744 
2745                         type = PROP_TYPE_STRING;
2746                 } else if (prop == ZFS_PROP_VOLSIZE ||
2747                     prop == ZFS_PROP_VERSION) {
2748                         return (EINVAL);
2749                 } else {
2750                         type = zfs_prop_get_type(prop);
2751                 }
2752 
2753                 VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2754 
2755                 switch (type) {
2756                 case PROP_TYPE_STRING:
2757                         VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2758                         break;
2759                 case PROP_TYPE_NUMBER:
2760                 case PROP_TYPE_INDEX:
2761                         VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2762                         break;
2763                 default:
2764                         nvlist_free(dummy);
2765                         return (EINVAL);
2766                 }
2767 
2768                 pair = nvlist_next_nvpair(dummy, NULL);
2769                 err = zfs_prop_set_special(zc->zc_name, source, pair);
2770                 nvlist_free(dummy);
2771                 if (err != -1)
2772                         return (err); /* special property already handled */
2773         } else {
2774                 /*
2775                  * Only check this in the non-received case. We want to allow
2776                  * 'inherit -S' to revert non-inheritable properties like quota
2777                  * and reservation to the received or default values even though
2778                  * they are not considered inheritable.
2779                  */
2780                 if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2781                         return (EINVAL);
2782         }
2783 
2784         /* property name has been validated by zfs_secpolicy_inherit_prop() */
2785         return (dsl_prop_set(zc->zc_name, zc->zc_value, source, 0, 0, NULL));
2786 }
2787 
2788 static int
2789 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2790 {
2791         nvlist_t *props;
2792         spa_t *spa;
2793         int error;
2794         nvpair_t *pair;
2795 
2796         if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2797             zc->zc_iflags, &props))
2798                 return (error);
2799 
2800         /*
2801          * If the only property is the configfile, then just do a spa_lookup()
2802          * to handle the faulted case.
2803          */
2804         pair = nvlist_next_nvpair(props, NULL);
2805         if (pair != NULL && strcmp(nvpair_name(pair),
2806             zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2807             nvlist_next_nvpair(props, pair) == NULL) {
2808                 mutex_enter(&spa_namespace_lock);
2809                 if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2810                         spa_configfile_set(spa, props, B_FALSE);
2811                         spa_config_sync(spa, B_FALSE, B_TRUE);
2812                 }
2813                 mutex_exit(&spa_namespace_lock);
2814                 if (spa != NULL) {
2815                         nvlist_free(props);
2816                         return (0);
2817                 }
2818         }
2819 
2820         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2821                 nvlist_free(props);
2822                 return (error);
2823         }
2824 
2825         error = spa_prop_set(spa, props);
2826 
2827         nvlist_free(props);
2828         spa_close(spa, FTAG);
2829 
2830         return (error);
2831 }
2832 
2833 static int
2834 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2835 {
2836         spa_t *spa;
2837         int error;
2838         nvlist_t *nvp = NULL;
2839 
2840         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2841                 /*
2842                  * If the pool is faulted, there may be properties we can still
2843                  * get (such as altroot and cachefile), so attempt to get them
2844                  * anyway.
2845                  */
2846                 mutex_enter(&spa_namespace_lock);
2847                 if ((spa = spa_lookup(zc->zc_name)) != NULL)
2848                         error = spa_prop_get(spa, &nvp);
2849                 mutex_exit(&spa_namespace_lock);
2850         } else {
2851                 error = spa_prop_get(spa, &nvp);
2852                 spa_close(spa, FTAG);
2853         }
2854 
2855         if (error == 0 && zc->zc_nvlist_dst != NULL)
2856                 error = put_nvlist(zc, nvp);
2857         else
2858                 error = EFAULT;
2859 
2860         nvlist_free(nvp);
2861         return (error);
2862 }
2863 
2864 /*
2865  * inputs:
2866  * zc_name              name of filesystem
2867  * zc_nvlist_src{_size} nvlist of delegated permissions
2868  * zc_perm_action       allow/unallow flag
2869  *
2870  * outputs:             none
2871  */
2872 static int
2873 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2874 {
2875         int error;
2876         nvlist_t *fsaclnv = NULL;
2877 
2878         if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2879             zc->zc_iflags, &fsaclnv)) != 0)
2880                 return (error);
2881 
2882         /*
2883          * Verify nvlist is constructed correctly
2884          */
2885         if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2886                 nvlist_free(fsaclnv);
2887                 return (EINVAL);
2888         }
2889 
2890         /*
2891          * If we don't have PRIV_SYS_MOUNT, then validate
2892          * that user is allowed to hand out each permission in
2893          * the nvlist(s)
2894          */
2895 
2896         error = secpolicy_zfs(CRED());
2897         if (error) {
2898                 if (zc->zc_perm_action == B_FALSE) {
2899                         error = dsl_deleg_can_allow(zc->zc_name,
2900                             fsaclnv, CRED());
2901                 } else {
2902                         error = dsl_deleg_can_unallow(zc->zc_name,
2903                             fsaclnv, CRED());
2904                 }
2905         }
2906 
2907         if (error == 0)
2908                 error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2909 
2910         nvlist_free(fsaclnv);
2911         return (error);
2912 }
2913 
2914 /*
2915  * inputs:
2916  * zc_name              name of filesystem
2917  *
2918  * outputs:
2919  * zc_nvlist_src{_size} nvlist of delegated permissions
2920  */
2921 static int
2922 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2923 {
2924         nvlist_t *nvp;
2925         int error;
2926 
2927         if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2928                 error = put_nvlist(zc, nvp);
2929                 nvlist_free(nvp);
2930         }
2931 
2932         return (error);
2933 }
2934 
2935 /*
2936  * Search the vfs list for a specified resource.  Returns a pointer to it
2937  * or NULL if no suitable entry is found. The caller of this routine
2938  * is responsible for releasing the returned vfs pointer.
2939  */
2940 static vfs_t *
2941 zfs_get_vfs(const char *resource)
2942 {
2943         struct vfs *vfsp;
2944         struct vfs *vfs_found = NULL;
2945 
2946         vfs_list_read_lock();
2947         vfsp = rootvfs;
2948         do {
2949                 if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
2950                         VFS_HOLD(vfsp);
2951                         vfs_found = vfsp;
2952                         break;
2953                 }
2954                 vfsp = vfsp->vfs_next;
2955         } while (vfsp != rootvfs);
2956         vfs_list_unlock();
2957         return (vfs_found);
2958 }
2959 
2960 /* ARGSUSED */
2961 static void
2962 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
2963 {
2964         zfs_creat_t *zct = arg;
2965 
2966         zfs_create_fs(os, cr, zct->zct_zplprops, tx);
2967 }
2968 
2969 #define ZFS_PROP_UNDEFINED      ((uint64_t)-1)
2970 
2971 /*
2972  * inputs:
2973  * createprops          list of properties requested by creator
2974  * default_zplver       zpl version to use if unspecified in createprops
2975  * fuids_ok             fuids allowed in this version of the spa?
2976  * os                   parent objset pointer (NULL if root fs)
2977  *
2978  * outputs:
2979  * zplprops     values for the zplprops we attach to the master node object
2980  * is_ci        true if requested file system will be purely case-insensitive
2981  *
2982  * Determine the settings for utf8only, normalization and
2983  * casesensitivity.  Specific values may have been requested by the
2984  * creator and/or we can inherit values from the parent dataset.  If
2985  * the file system is of too early a vintage, a creator can not
2986  * request settings for these properties, even if the requested
2987  * setting is the default value.  We don't actually want to create dsl
2988  * properties for these, so remove them from the source nvlist after
2989  * processing.
2990  */
2991 static int
2992 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
2993     boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
2994     nvlist_t *zplprops, boolean_t *is_ci)
2995 {
2996         uint64_t sense = ZFS_PROP_UNDEFINED;
2997         uint64_t norm = ZFS_PROP_UNDEFINED;
2998         uint64_t u8 = ZFS_PROP_UNDEFINED;
2999 
3000         ASSERT(zplprops != NULL);
3001 
3002         /*
3003          * Pull out creator prop choices, if any.
3004          */
3005         if (createprops) {
3006                 (void) nvlist_lookup_uint64(createprops,
3007                     zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3008                 (void) nvlist_lookup_uint64(createprops,
3009                     zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3010                 (void) nvlist_remove_all(createprops,
3011                     zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3012                 (void) nvlist_lookup_uint64(createprops,
3013                     zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3014                 (void) nvlist_remove_all(createprops,
3015                     zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3016                 (void) nvlist_lookup_uint64(createprops,
3017                     zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3018                 (void) nvlist_remove_all(createprops,
3019                     zfs_prop_to_name(ZFS_PROP_CASE));
3020         }
3021 
3022         /*
3023          * If the zpl version requested is whacky or the file system
3024          * or pool is version is too "young" to support normalization
3025          * and the creator tried to set a value for one of the props,
3026          * error out.
3027          */
3028         if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3029             (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3030             (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3031             (zplver < ZPL_VERSION_NORMALIZATION &&
3032             (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3033             sense != ZFS_PROP_UNDEFINED)))
3034                 return (ENOTSUP);
3035 
3036         /*
3037          * Put the version in the zplprops
3038          */
3039         VERIFY(nvlist_add_uint64(zplprops,
3040             zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3041 
3042         if (norm == ZFS_PROP_UNDEFINED)
3043                 VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3044         VERIFY(nvlist_add_uint64(zplprops,
3045             zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3046 
3047         /*
3048          * If we're normalizing, names must always be valid UTF-8 strings.
3049          */
3050         if (norm)
3051                 u8 = 1;
3052         if (u8 == ZFS_PROP_UNDEFINED)
3053                 VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3054         VERIFY(nvlist_add_uint64(zplprops,
3055             zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3056 
3057         if (sense == ZFS_PROP_UNDEFINED)
3058                 VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3059         VERIFY(nvlist_add_uint64(zplprops,
3060             zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3061 
3062         if (is_ci)
3063                 *is_ci = (sense == ZFS_CASE_INSENSITIVE);
3064 
3065         return (0);
3066 }
3067 
3068 static int
3069 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3070     nvlist_t *zplprops, boolean_t *is_ci)
3071 {
3072         boolean_t fuids_ok, sa_ok;
3073         uint64_t zplver = ZPL_VERSION;
3074         objset_t *os = NULL;
3075         char parentname[MAXNAMELEN];
3076         char *cp;
3077         spa_t *spa;
3078         uint64_t spa_vers;
3079         int error;
3080 
3081         (void) strlcpy(parentname, dataset, sizeof (parentname));
3082         cp = strrchr(parentname, '/');
3083         ASSERT(cp != NULL);
3084         cp[0] = '\0';
3085 
3086         if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3087                 return (error);
3088 
3089         spa_vers = spa_version(spa);
3090         spa_close(spa, FTAG);
3091 
3092         zplver = zfs_zpl_version_map(spa_vers);
3093         fuids_ok = (zplver >= ZPL_VERSION_FUID);
3094         sa_ok = (zplver >= ZPL_VERSION_SA);
3095 
3096         /*
3097          * Open parent object set so we can inherit zplprop values.
3098          */
3099         if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3100                 return (error);
3101 
3102         error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3103             zplprops, is_ci);
3104         dmu_objset_rele(os, FTAG);
3105         return (error);
3106 }
3107 
3108 static int
3109 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3110     nvlist_t *zplprops, boolean_t *is_ci)
3111 {
3112         boolean_t fuids_ok;
3113         boolean_t sa_ok;
3114         uint64_t zplver = ZPL_VERSION;
3115         int error;
3116 
3117         zplver = zfs_zpl_version_map(spa_vers);
3118         fuids_ok = (zplver >= ZPL_VERSION_FUID);
3119         sa_ok = (zplver >= ZPL_VERSION_SA);
3120 
3121         error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3122             createprops, zplprops, is_ci);
3123         return (error);
3124 }
3125 
3126 /*
3127  * innvl: {
3128  *     "type" -> dmu_objset_type_t (int32)
3129  *     (optional) "props" -> { prop -> value }
3130  * }
3131  *
3132  * outnvl: propname -> error code (int32)
3133  */
3134 static int
3135 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3136 {
3137         int error = 0;
3138         zfs_creat_t zct = { 0 };
3139         nvlist_t *nvprops = NULL;
3140         void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3141         int32_t type32;
3142         dmu_objset_type_t type;
3143         boolean_t is_insensitive = B_FALSE;
3144 
3145         if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3146                 return (EINVAL);
3147         type = type32;
3148         (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3149 
3150         switch (type) {
3151         case DMU_OST_ZFS:
3152                 cbfunc = zfs_create_cb;
3153                 break;
3154 
3155         case DMU_OST_ZVOL:
3156                 cbfunc = zvol_create_cb;
3157                 break;
3158 
3159         default:
3160                 cbfunc = NULL;
3161                 break;
3162         }
3163         if (strchr(fsname, '@') ||
3164             strchr(fsname, '%'))
3165                 return (EINVAL);
3166 
3167         zct.zct_props = nvprops;
3168 
3169         if (cbfunc == NULL)
3170                 return (EINVAL);
3171 
3172         if (type == DMU_OST_ZVOL) {
3173                 uint64_t volsize, volblocksize;
3174 
3175                 if (nvprops == NULL)
3176                         return (EINVAL);
3177                 if (nvlist_lookup_uint64(nvprops,
3178                     zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3179                         return (EINVAL);
3180 
3181                 if ((error = nvlist_lookup_uint64(nvprops,
3182                     zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3183                     &volblocksize)) != 0 && error != ENOENT)
3184                         return (EINVAL);
3185 
3186                 if (error != 0)
3187                         volblocksize = zfs_prop_default_numeric(
3188                             ZFS_PROP_VOLBLOCKSIZE);
3189 
3190                 if ((error = zvol_check_volblocksize(
3191                     volblocksize)) != 0 ||
3192                     (error = zvol_check_volsize(volsize,
3193                     volblocksize)) != 0)
3194                         return (error);
3195         } else if (type == DMU_OST_ZFS) {
3196                 int error;
3197 
3198                 /*
3199                  * We have to have normalization and
3200                  * case-folding flags correct when we do the
3201                  * file system creation, so go figure them out
3202                  * now.
3203                  */
3204                 VERIFY(nvlist_alloc(&zct.zct_zplprops,
3205                     NV_UNIQUE_NAME, KM_SLEEP) == 0);
3206                 error = zfs_fill_zplprops(fsname, nvprops,
3207                     zct.zct_zplprops, &is_insensitive);
3208                 if (error != 0) {
3209                         nvlist_free(zct.zct_zplprops);
3210                         return (error);
3211                 }
3212         }
3213 
3214         error = dmu_objset_create(fsname, type,
3215             is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3216         nvlist_free(zct.zct_zplprops);
3217 
3218         /*
3219          * It would be nice to do this atomically.
3220          */
3221         if (error == 0) {
3222                 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3223                     nvprops, outnvl);
3224                 if (error != 0)
3225                         (void) dmu_objset_destroy(fsname, B_FALSE);
3226         }
3227         return (error);
3228 }
3229 
3230 /*
3231  * innvl: {
3232  *     "origin" -> name of origin snapshot
3233  *     (optional) "props" -> { prop -> value }
3234  * }
3235  *
3236  * outnvl: propname -> error code (int32)
3237  */
3238 static int
3239 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3240 {
3241         int error = 0;
3242         nvlist_t *nvprops = NULL;
3243         char *origin_name;
3244         dsl_dataset_t *origin;
3245 
3246         if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3247                 return (EINVAL);
3248         (void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3249 
3250         if (strchr(fsname, '@') ||
3251             strchr(fsname, '%'))
3252                 return (EINVAL);
3253 
3254         if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3255                 return (EINVAL);
3256 
3257         error = dsl_dataset_hold(origin_name, FTAG, &origin);
3258         if (error)
3259                 return (error);
3260 
3261         error = dmu_objset_clone(fsname, origin, 0);
3262         dsl_dataset_rele(origin, FTAG);
3263         if (error)
3264                 return (error);
3265 
3266         /*
3267          * It would be nice to do this atomically.
3268          */
3269         if (error == 0) {
3270                 error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3271                     nvprops, outnvl);
3272                 if (error != 0)
3273                         (void) dmu_objset_destroy(fsname, B_FALSE);
3274         }
3275         return (error);
3276 }
3277 
3278 /*
3279  * innvl: {
3280  *     "snaps" -> { snapshot1, snapshot2 }
3281  *     (optional) "props" -> { prop -> value (string) }
3282  * }
3283  *
3284  * outnvl: snapshot -> error code (int32)
3285  *
3286  */
3287 static int
3288 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3289 {
3290         nvlist_t *snaps;
3291         nvlist_t *props = NULL;
3292         int error, poollen;
3293         nvpair_t *pair;
3294 
3295         (void) nvlist_lookup_nvlist(innvl, "props", &props);
3296         if ((error = zfs_check_userprops(poolname, props)) != 0)
3297                 return (error);
3298 
3299         if (!nvlist_empty(props) &&
3300             zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3301                 return (ENOTSUP);
3302 
3303         if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3304                 return (EINVAL);
3305         poollen = strlen(poolname);
3306         for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3307             pair = nvlist_next_nvpair(snaps, pair)) {
3308                 const char *name = nvpair_name(pair);
3309                 const char *cp = strchr(name, '@');
3310 
3311                 /*
3312                  * The snap name must contain an @, and the part after it must
3313                  * contain only valid characters.
3314                  */
3315                 if (cp == NULL || snapshot_namecheck(cp + 1, NULL, NULL) != 0)
3316                         return (EINVAL);
3317 
3318                 /*
3319                  * The snap must be in the specified pool.
3320                  */
3321                 if (strncmp(name, poolname, poollen) != 0 ||
3322                     (name[poollen] != '/' && name[poollen] != '@'))
3323                         return (EXDEV);
3324 
3325                 /* This must be the only snap of this fs. */
3326                 for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3327                     pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3328                         if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3329                             == 0) {
3330                                 return (EXDEV);
3331                         }
3332                 }
3333         }
3334 
3335         error = dmu_objset_snapshot(snaps, props, outnvl);
3336         return (error);
3337 }
3338 
3339 /*
3340  * innvl: "message" -> string
3341  */
3342 /* ARGSUSED */
3343 static int
3344 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3345 {
3346         char *message;
3347         spa_t *spa;
3348         int error;
3349         char *poolname;
3350 
3351         /*
3352          * The poolname in the ioctl is not set, we get it from the TSD,
3353          * which was set at the end of the last successful ioctl that allows
3354          * logging.  The secpolicy func already checked that it is set.
3355          * Only one log ioctl is allowed after each successful ioctl, so
3356          * we clear the TSD here.
3357          */
3358         poolname = tsd_get(zfs_allow_log_key);
3359         (void) tsd_set(zfs_allow_log_key, NULL);
3360         error = spa_open(poolname, &spa, FTAG);
3361         strfree(poolname);
3362         if (error != 0)
3363                 return (error);
3364 
3365         if (nvlist_lookup_string(innvl, "message", &message) != 0)  {
3366                 spa_close(spa, FTAG);
3367                 return (EINVAL);
3368         }
3369 
3370         if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3371                 spa_close(spa, FTAG);
3372                 return (ENOTSUP);
3373         }
3374 
3375         error = spa_history_log(spa, message);
3376         spa_close(spa, FTAG);
3377         return (error);
3378 }
3379 
3380 /* ARGSUSED */
3381 int
3382 zfs_unmount_snap(const char *name, void *arg)
3383 {
3384         vfs_t *vfsp;
3385         int err;
3386 
3387         if (strchr(name, '@') == NULL)
3388                 return (0);
3389 
3390         vfsp = zfs_get_vfs(name);
3391         if (vfsp == NULL)
3392                 return (0);
3393 
3394         if ((err = vn_vfswlock(vfsp->vfs_vnodecovered)) != 0) {
3395                 VFS_RELE(vfsp);
3396                 return (err);
3397         }
3398         VFS_RELE(vfsp);
3399 
3400         /*
3401          * Always force the unmount for snapshots.
3402          */
3403         return (dounmount(vfsp, MS_FORCE, kcred));
3404 }
3405 
3406 /*
3407  * innvl: {
3408  *     "snaps" -> { snapshot1, snapshot2 }
3409  *     (optional boolean) "defer"
3410  * }
3411  *
3412  * outnvl: snapshot -> error code (int32)
3413  *
3414  */
3415 static int
3416 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3417 {
3418         int poollen;
3419         nvlist_t *snaps;
3420         nvpair_t *pair;
3421         boolean_t defer;
3422 
3423         if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3424                 return (EINVAL);
3425         defer = nvlist_exists(innvl, "defer");
3426 
3427         poollen = strlen(poolname);
3428         for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3429             pair = nvlist_next_nvpair(snaps, pair)) {
3430                 const char *name = nvpair_name(pair);
3431 
3432                 /*
3433                  * The snap must be in the specified pool.
3434                  */
3435                 if (strncmp(name, poolname, poollen) != 0 ||
3436                     (name[poollen] != '/' && name[poollen] != '@'))
3437                         return (EXDEV);
3438 
3439                 /*
3440                  * Ignore failures to unmount; dmu_snapshots_destroy_nvl()
3441                  * will deal with this gracefully (by filling in outnvl).
3442                  */
3443                 (void) zfs_unmount_snap(name, NULL);
3444         }
3445 
3446         return (dmu_snapshots_destroy_nvl(snaps, defer, outnvl));
3447 }
3448 
3449 /*
3450  * inputs:
3451  * zc_name              name of dataset to destroy
3452  * zc_objset_type       type of objset
3453  * zc_defer_destroy     mark for deferred destroy
3454  *
3455  * outputs:             none
3456  */
3457 static int
3458 zfs_ioc_destroy(zfs_cmd_t *zc)
3459 {
3460         int err;
3461         if (strchr(zc->zc_name, '@') && zc->zc_objset_type == DMU_OST_ZFS) {
3462                 err = zfs_unmount_snap(zc->zc_name, NULL);
3463                 if (err)
3464                         return (err);
3465         }
3466 
3467         err = dmu_objset_destroy(zc->zc_name, zc->zc_defer_destroy);
3468         if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3469                 (void) zvol_remove_minor(zc->zc_name);
3470         return (err);
3471 }
3472 
3473 /*
3474  * inputs:
3475  * zc_name      name of dataset to rollback (to most recent snapshot)
3476  *
3477  * outputs:     none
3478  */
3479 static int
3480 zfs_ioc_rollback(zfs_cmd_t *zc)
3481 {
3482         dsl_dataset_t *ds, *clone;
3483         int error;
3484         zfsvfs_t *zfsvfs;
3485         char *clone_name;
3486 
3487         error = dsl_dataset_hold(zc->zc_name, FTAG, &ds);
3488         if (error)
3489                 return (error);
3490 
3491         /* must not be a snapshot */
3492         if (dsl_dataset_is_snapshot(ds)) {
3493                 dsl_dataset_rele(ds, FTAG);
3494                 return (EINVAL);
3495         }
3496 
3497         /* must have a most recent snapshot */
3498         if (ds->ds_phys->ds_prev_snap_txg < TXG_INITIAL) {
3499                 dsl_dataset_rele(ds, FTAG);
3500                 return (EINVAL);
3501         }
3502 
3503         /*
3504          * Create clone of most recent snapshot.
3505          */
3506         clone_name = kmem_asprintf("%s/%%rollback", zc->zc_name);
3507         error = dmu_objset_clone(clone_name, ds->ds_prev, DS_FLAG_INCONSISTENT);
3508         if (error)
3509                 goto out;
3510 
3511         error = dsl_dataset_own(clone_name, B_TRUE, FTAG, &clone);
3512         if (error)
3513                 goto out;
3514 
3515         /*
3516          * Do clone swap.
3517          */
3518         if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
3519                 error = zfs_suspend_fs(zfsvfs);
3520                 if (error == 0) {
3521                         int resume_err;
3522 
3523                         if (dsl_dataset_tryown(ds, B_FALSE, FTAG)) {
3524                                 error = dsl_dataset_clone_swap(clone, ds,
3525                                     B_TRUE);
3526                                 dsl_dataset_disown(ds, FTAG);
3527                                 ds = NULL;
3528                         } else {
3529                                 error = EBUSY;
3530                         }
3531                         resume_err = zfs_resume_fs(zfsvfs, zc->zc_name);
3532                         error = error ? error : resume_err;
3533                 }
3534                 VFS_RELE(zfsvfs->z_vfs);
3535         } else {
3536                 if (dsl_dataset_tryown(ds, B_FALSE, FTAG)) {
3537                         error = dsl_dataset_clone_swap(clone, ds, B_TRUE);
3538                         dsl_dataset_disown(ds, FTAG);
3539                         ds = NULL;
3540                 } else {
3541                         error = EBUSY;
3542                 }
3543         }
3544 
3545         /*
3546          * Destroy clone (which also closes it).
3547          */
3548         (void) dsl_dataset_destroy(clone, FTAG, B_FALSE);
3549 
3550 out:
3551         strfree(clone_name);
3552         if (ds)
3553                 dsl_dataset_rele(ds, FTAG);
3554         return (error);
3555 }
3556 
3557 /*
3558  * inputs:
3559  * zc_name      old name of dataset
3560  * zc_value     new name of dataset
3561  * zc_cookie    recursive flag (only valid for snapshots)
3562  *
3563  * outputs:     none
3564  */
3565 static int
3566 zfs_ioc_rename(zfs_cmd_t *zc)
3567 {
3568         boolean_t recursive = zc->zc_cookie & 1;
3569 
3570         zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3571         if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3572             strchr(zc->zc_value, '%'))
3573                 return (EINVAL);
3574 
3575         /*
3576          * Unmount snapshot unless we're doing a recursive rename,
3577          * in which case the dataset code figures out which snapshots
3578          * to unmount.
3579          */
3580         if (!recursive && strchr(zc->zc_name, '@') != NULL &&
3581             zc->zc_objset_type == DMU_OST_ZFS) {
3582                 int err = zfs_unmount_snap(zc->zc_name, NULL);
3583                 if (err)
3584                         return (err);
3585         }
3586         if (zc->zc_objset_type == DMU_OST_ZVOL)
3587                 (void) zvol_remove_minor(zc->zc_name);
3588         return (dmu_objset_rename(zc->zc_name, zc->zc_value, recursive));
3589 }
3590 
3591 static int
3592 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3593 {
3594         const char *propname = nvpair_name(pair);
3595         boolean_t issnap = (strchr(dsname, '@') != NULL);
3596         zfs_prop_t prop = zfs_name_to_prop(propname);
3597         uint64_t intval;
3598         int err;
3599 
3600         if (prop == ZPROP_INVAL) {
3601                 if (zfs_prop_user(propname)) {
3602                         if (err = zfs_secpolicy_write_perms(dsname,
3603                             ZFS_DELEG_PERM_USERPROP, cr))
3604                                 return (err);
3605                         return (0);
3606                 }
3607 
3608                 if (!issnap && zfs_prop_userquota(propname)) {
3609                         const char *perm = NULL;
3610                         const char *uq_prefix =
3611                             zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3612                         const char *gq_prefix =
3613                             zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3614 
3615                         if (strncmp(propname, uq_prefix,
3616                             strlen(uq_prefix)) == 0) {
3617                                 perm = ZFS_DELEG_PERM_USERQUOTA;
3618                         } else if (strncmp(propname, gq_prefix,
3619                             strlen(gq_prefix)) == 0) {
3620                                 perm = ZFS_DELEG_PERM_GROUPQUOTA;
3621                         } else {
3622                                 /* USERUSED and GROUPUSED are read-only */
3623                                 return (EINVAL);
3624                         }
3625 
3626                         if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3627                                 return (err);
3628                         return (0);
3629                 }
3630 
3631                 return (EINVAL);
3632         }
3633 
3634         if (issnap)
3635                 return (EINVAL);
3636 
3637         if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3638                 /*
3639                  * dsl_prop_get_all_impl() returns properties in this
3640                  * format.
3641                  */
3642                 nvlist_t *attrs;
3643                 VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3644                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3645                     &pair) == 0);
3646         }
3647 
3648         /*
3649          * Check that this value is valid for this pool version
3650          */
3651         switch (prop) {
3652         case ZFS_PROP_COMPRESSION:
3653                 /*
3654                  * If the user specified gzip compression, make sure
3655                  * the SPA supports it. We ignore any errors here since
3656                  * we'll catch them later.
3657                  */
3658                 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3659                     nvpair_value_uint64(pair, &intval) == 0) {
3660                         if (intval >= ZIO_COMPRESS_GZIP_1 &&
3661                             intval <= ZIO_COMPRESS_GZIP_9 &&
3662                             zfs_earlier_version(dsname,
3663                             SPA_VERSION_GZIP_COMPRESSION)) {
3664                                 return (ENOTSUP);
3665                         }
3666 
3667                         if (intval == ZIO_COMPRESS_ZLE &&
3668                             zfs_earlier_version(dsname,
3669                             SPA_VERSION_ZLE_COMPRESSION))
3670                                 return (ENOTSUP);
3671 
3672                         if (intval == ZIO_COMPRESS_LZ4) {
3673                                 zfeature_info_t *feature =
3674                                     &spa_feature_table[
3675                                     SPA_FEATURE_LZ4_COMPRESS];
3676                                 spa_t *spa;
3677 
3678                                 if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3679                                         return (err);
3680 
3681                                 if (!spa_feature_is_enabled(spa, feature)) {
3682                                         spa_close(spa, FTAG);
3683                                         return (ENOTSUP);
3684                                 }
3685                                 spa_close(spa, FTAG);
3686                         }
3687 
3688                         /*
3689                          * If this is a bootable dataset then
3690                          * verify that the compression algorithm
3691                          * is supported for booting. We must return
3692                          * something other than ENOTSUP since it
3693                          * implies a downrev pool version.
3694                          */
3695                         if (zfs_is_bootfs(dsname) &&
3696                             !BOOTFS_COMPRESS_VALID(intval)) {
3697                                 return (ERANGE);
3698                         }
3699                 }
3700                 break;
3701 
3702         case ZFS_PROP_COPIES:
3703                 if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3704                         return (ENOTSUP);
3705                 break;
3706 
3707         case ZFS_PROP_DEDUP:
3708                 if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3709                         return (ENOTSUP);
3710                 break;
3711 
3712         case ZFS_PROP_SHARESMB:
3713                 if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3714                         return (ENOTSUP);
3715                 break;
3716 
3717         case ZFS_PROP_ACLINHERIT:
3718                 if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3719                     nvpair_value_uint64(pair, &intval) == 0) {
3720                         if (intval == ZFS_ACL_PASSTHROUGH_X &&
3721                             zfs_earlier_version(dsname,
3722                             SPA_VERSION_PASSTHROUGH_X))
3723                                 return (ENOTSUP);
3724                 }
3725                 break;
3726         }
3727 
3728         return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3729 }
3730 
3731 /*
3732  * Activates a feature on a pool in response to a property setting. This
3733  * creates a new sync task which modifies the pool to reflect the feature
3734  * as being active.
3735  */
3736 static int
3737 zfs_prop_activate_feature(dsl_pool_t *dp, zfeature_info_t *feature)
3738 {
3739         int err;
3740 
3741         /* EBUSY here indicates that the feature is already active */
3742         err = dsl_sync_task_do(dp, zfs_prop_activate_feature_check,
3743             zfs_prop_activate_feature_sync, dp->dp_spa, feature, 2);
3744 
3745         if (err != 0 && err != EBUSY)
3746                 return (err);
3747         else
3748                 return (0);
3749 }
3750 
3751 /*
3752  * Checks for a race condition to make sure we don't increment a feature flag
3753  * multiple times.
3754  */
3755 /*ARGSUSED*/
3756 static int
3757 zfs_prop_activate_feature_check(void *arg1, void *arg2, dmu_tx_t *tx)
3758 {
3759         spa_t *spa = arg1;
3760         zfeature_info_t *feature = arg2;
3761 
3762         if (!spa_feature_is_active(spa, feature))
3763                 return (0);
3764         else
3765                 return (EBUSY);
3766 }
3767 
3768 /*
3769  * The callback invoked on feature activation in the sync task caused by
3770  * zfs_prop_activate_feature.
3771  */
3772 static void
3773 zfs_prop_activate_feature_sync(void *arg1, void *arg2, dmu_tx_t *tx)
3774 {
3775         spa_t *spa = arg1;
3776         zfeature_info_t *feature = arg2;
3777 
3778         spa_feature_incr(spa, feature, tx);
3779 }
3780 
3781 /*
3782  * Removes properties from the given props list that fail permission checks
3783  * needed to clear them and to restore them in case of a receive error. For each
3784  * property, make sure we have both set and inherit permissions.
3785  *
3786  * Returns the first error encountered if any permission checks fail. If the
3787  * caller provides a non-NULL errlist, it also gives the complete list of names
3788  * of all the properties that failed a permission check along with the
3789  * corresponding error numbers. The caller is responsible for freeing the
3790  * returned errlist.
3791  *
3792  * If every property checks out successfully, zero is returned and the list
3793  * pointed at by errlist is NULL.
3794  */
3795 static int
3796 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
3797 {
3798         zfs_cmd_t *zc;
3799         nvpair_t *pair, *next_pair;
3800         nvlist_t *errors;
3801         int err, rv = 0;
3802 
3803         if (props == NULL)
3804                 return (0);
3805 
3806         VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3807 
3808         zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
3809         (void) strcpy(zc->zc_name, dataset);
3810         pair = nvlist_next_nvpair(props, NULL);
3811         while (pair != NULL) {
3812                 next_pair = nvlist_next_nvpair(props, pair);
3813 
3814                 (void) strcpy(zc->zc_value, nvpair_name(pair));
3815                 if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
3816                     (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
3817                         VERIFY(nvlist_remove_nvpair(props, pair) == 0);
3818                         VERIFY(nvlist_add_int32(errors,
3819                             zc->zc_value, err) == 0);
3820                 }
3821                 pair = next_pair;
3822         }
3823         kmem_free(zc, sizeof (zfs_cmd_t));
3824 
3825         if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
3826                 nvlist_free(errors);
3827                 errors = NULL;
3828         } else {
3829                 VERIFY(nvpair_value_int32(pair, &rv) == 0);
3830         }
3831 
3832         if (errlist == NULL)
3833                 nvlist_free(errors);
3834         else
3835                 *errlist = errors;
3836 
3837         return (rv);
3838 }
3839 
3840 static boolean_t
3841 propval_equals(nvpair_t *p1, nvpair_t *p2)
3842 {
3843         if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
3844                 /* dsl_prop_get_all_impl() format */
3845                 nvlist_t *attrs;
3846                 VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
3847                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3848                     &p1) == 0);
3849         }
3850 
3851         if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
3852                 nvlist_t *attrs;
3853                 VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
3854                 VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3855                     &p2) == 0);
3856         }
3857 
3858         if (nvpair_type(p1) != nvpair_type(p2))
3859                 return (B_FALSE);
3860 
3861         if (nvpair_type(p1) == DATA_TYPE_STRING) {
3862                 char *valstr1, *valstr2;
3863 
3864                 VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
3865                 VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
3866                 return (strcmp(valstr1, valstr2) == 0);
3867         } else {
3868                 uint64_t intval1, intval2;
3869 
3870                 VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
3871                 VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
3872                 return (intval1 == intval2);
3873         }
3874 }
3875 
3876 /*
3877  * Remove properties from props if they are not going to change (as determined
3878  * by comparison with origprops). Remove them from origprops as well, since we
3879  * do not need to clear or restore properties that won't change.
3880  */
3881 static void
3882 props_reduce(nvlist_t *props, nvlist_t *origprops)
3883 {
3884         nvpair_t *pair, *next_pair;
3885 
3886         if (origprops == NULL)
3887                 return; /* all props need to be received */
3888 
3889         pair = nvlist_next_nvpair(props, NULL);
3890         while (pair != NULL) {
3891                 const char *propname = nvpair_name(pair);
3892                 nvpair_t *match;
3893 
3894                 next_pair = nvlist_next_nvpair(props, pair);
3895 
3896                 if ((nvlist_lookup_nvpair(origprops, propname,
3897                     &match) != 0) || !propval_equals(pair, match))
3898                         goto next; /* need to set received value */
3899 
3900                 /* don't clear the existing received value */
3901                 (void) nvlist_remove_nvpair(origprops, match);
3902                 /* don't bother receiving the property */
3903                 (void) nvlist_remove_nvpair(props, pair);
3904 next:
3905                 pair = next_pair;
3906         }
3907 }
3908 
3909 #ifdef  DEBUG
3910 static boolean_t zfs_ioc_recv_inject_err;
3911 #endif
3912 
3913 /*
3914  * inputs:
3915  * zc_name              name of containing filesystem
3916  * zc_nvlist_src{_size} nvlist of properties to apply
3917  * zc_value             name of snapshot to create
3918  * zc_string            name of clone origin (if DRR_FLAG_CLONE)
3919  * zc_cookie            file descriptor to recv from
3920  * zc_begin_record      the BEGIN record of the stream (not byteswapped)
3921  * zc_guid              force flag
3922  * zc_cleanup_fd        cleanup-on-exit file descriptor
3923  * zc_action_handle     handle for this guid/ds mapping (or zero on first call)
3924  *
3925  * outputs:
3926  * zc_cookie            number of bytes read
3927  * zc_nvlist_dst{_size} error for each unapplied received property
3928  * zc_obj               zprop_errflags_t
3929  * zc_action_handle     handle for this guid/ds mapping
3930  */
3931 static int
3932 zfs_ioc_recv(zfs_cmd_t *zc)
3933 {
3934         file_t *fp;
3935         objset_t *os;
3936         dmu_recv_cookie_t drc;
3937         boolean_t force = (boolean_t)zc->zc_guid;
3938         int fd;
3939         int error = 0;
3940         int props_error = 0;
3941         nvlist_t *errors;
3942         offset_t off;
3943         nvlist_t *props = NULL; /* sent properties */
3944         nvlist_t *origprops = NULL; /* existing properties */
3945         objset_t *origin = NULL;
3946         char *tosnap;
3947         char tofs[ZFS_MAXNAMELEN];
3948         boolean_t first_recvd_props = B_FALSE;
3949 
3950         if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3951             strchr(zc->zc_value, '@') == NULL ||
3952             strchr(zc->zc_value, '%'))
3953                 return (EINVAL);
3954 
3955         (void) strcpy(tofs, zc->zc_value);
3956         tosnap = strchr(tofs, '@');
3957         *tosnap++ = '\0';
3958 
3959         if (zc->zc_nvlist_src != NULL &&
3960             (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3961             zc->zc_iflags, &props)) != 0)
3962                 return (error);
3963 
3964         fd = zc->zc_cookie;
3965         fp = getf(fd);
3966         if (fp == NULL) {
3967                 nvlist_free(props);
3968                 return (EBADF);
3969         }
3970 
3971         VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3972 
3973         if (props && dmu_objset_hold(tofs, FTAG, &os) == 0) {
3974                 if ((spa_version(os->os_spa) >= SPA_VERSION_RECVD_PROPS) &&
3975                     !dsl_prop_get_hasrecvd(os)) {
3976                         first_recvd_props = B_TRUE;
3977                 }
3978 
3979                 /*
3980                  * If new received properties are supplied, they are to
3981                  * completely replace the existing received properties, so stash
3982                  * away the existing ones.
3983                  */
3984                 if (dsl_prop_get_received(os, &origprops) == 0) {
3985                         nvlist_t *errlist = NULL;
3986                         /*
3987                          * Don't bother writing a property if its value won't
3988                          * change (and avoid the unnecessary security checks).
3989                          *
3990                          * The first receive after SPA_VERSION_RECVD_PROPS is a
3991                          * special case where we blow away all local properties
3992                          * regardless.
3993                          */
3994                         if (!first_recvd_props)
3995                                 props_reduce(props, origprops);
3996                         if (zfs_check_clearable(tofs, origprops,
3997                             &errlist) != 0)
3998                                 (void) nvlist_merge(errors, errlist, 0);
3999                         nvlist_free(errlist);
4000                 }
4001 
4002                 dmu_objset_rele(os, FTAG);
4003         }
4004 
4005         if (zc->zc_string[0]) {
4006                 error = dmu_objset_hold(zc->zc_string, FTAG, &origin);
4007                 if (error)
4008                         goto out;
4009         }
4010 
4011         error = dmu_recv_begin(tofs, tosnap, zc->zc_top_ds,
4012             &zc->zc_begin_record, force, origin, &drc);
4013         if (origin)
4014                 dmu_objset_rele(origin, FTAG);
4015         if (error)
4016                 goto out;
4017 
4018         /*
4019          * Set properties before we receive the stream so that they are applied
4020          * to the new data. Note that we must call dmu_recv_stream() if
4021          * dmu_recv_begin() succeeds.
4022          */
4023         if (props) {
4024                 if (dmu_objset_from_ds(drc.drc_logical_ds, &os) == 0) {
4025                         if (drc.drc_newfs) {
4026                                 if (spa_version(os->os_spa) >=
4027                                     SPA_VERSION_RECVD_PROPS)
4028                                         first_recvd_props = B_TRUE;
4029                         } else if (origprops != NULL) {
4030                                 if (clear_received_props(os, tofs, origprops,
4031                                     first_recvd_props ? NULL : props) != 0)
4032                                         zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4033                         } else {
4034                                 zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4035                         }
4036                         dsl_prop_set_hasrecvd(os);
4037                 } else if (!drc.drc_newfs) {
4038                         zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4039                 }
4040 
4041                 (void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4042                     props, errors);
4043         }
4044 
4045         if (zc->zc_nvlist_dst_size != 0 &&
4046             (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4047             put_nvlist(zc, errors) != 0)) {
4048                 /*
4049                  * Caller made zc->zc_nvlist_dst less than the minimum expected
4050                  * size or supplied an invalid address.
4051                  */
4052                 props_error = EINVAL;
4053         }
4054 
4055         off = fp->f_offset;
4056         error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
4057             &zc->zc_action_handle);
4058 
4059         if (error == 0) {
4060                 zfsvfs_t *zfsvfs = NULL;
4061 
4062                 if (getzfsvfs(tofs, &zfsvfs) == 0) {
4063                         /* online recv */
4064                         int end_err;
4065 
4066                         error = zfs_suspend_fs(zfsvfs);
4067                         /*
4068                          * If the suspend fails, then the recv_end will
4069                          * likely also fail, and clean up after itself.
4070                          */
4071                         end_err = dmu_recv_end(&drc);
4072                         if (error == 0)
4073                                 error = zfs_resume_fs(zfsvfs, tofs);
4074                         error = error ? error : end_err;
4075                         VFS_RELE(zfsvfs->z_vfs);
4076                 } else {
4077                         error = dmu_recv_end(&drc);
4078                 }
4079         }
4080 
4081         zc->zc_cookie = off - fp->f_offset;
4082         if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4083                 fp->f_offset = off;
4084 
4085 #ifdef  DEBUG
4086         if (zfs_ioc_recv_inject_err) {
4087                 zfs_ioc_recv_inject_err = B_FALSE;
4088                 error = 1;
4089         }
4090 #endif
4091         /*
4092          * On error, restore the original props.
4093          */
4094         if (error && props) {
4095                 if (dmu_objset_hold(tofs, FTAG, &os) == 0) {
4096                         if (clear_received_props(os, tofs, props, NULL) != 0) {
4097                                 /*
4098                                  * We failed to clear the received properties.
4099                                  * Since we may have left a $recvd value on the
4100                                  * system, we can't clear the $hasrecvd flag.
4101                                  */
4102                                 zc->zc_obj |= ZPROP_ERR_NORESTORE;
4103                         } else if (first_recvd_props) {
4104                                 dsl_prop_unset_hasrecvd(os);
4105                         }
4106                         dmu_objset_rele(os, FTAG);
4107                 } else if (!drc.drc_newfs) {
4108                         /* We failed to clear the received properties. */
4109                         zc->zc_obj |= ZPROP_ERR_NORESTORE;
4110                 }
4111 
4112                 if (origprops == NULL && !drc.drc_newfs) {
4113                         /* We failed to stash the original properties. */
4114                         zc->zc_obj |= ZPROP_ERR_NORESTORE;
4115                 }
4116 
4117                 /*
4118                  * dsl_props_set() will not convert RECEIVED to LOCAL on or
4119                  * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4120                  * explictly if we're restoring local properties cleared in the
4121                  * first new-style receive.
4122                  */
4123                 if (origprops != NULL &&
4124                     zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4125                     ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4126                     origprops, NULL) != 0) {
4127                         /*
4128                          * We stashed the original properties but failed to
4129                          * restore them.
4130                          */
4131                         zc->zc_obj |= ZPROP_ERR_NORESTORE;
4132                 }
4133         }
4134 out:
4135         nvlist_free(props);
4136         nvlist_free(origprops);
4137         nvlist_free(errors);
4138         releasef(fd);
4139 
4140         if (error == 0)
4141                 error = props_error;
4142 
4143         return (error);
4144 }
4145 
4146 /*
4147  * inputs:
4148  * zc_name      name of snapshot to send
4149  * zc_cookie    file descriptor to send stream to
4150  * zc_obj       fromorigin flag (mutually exclusive with zc_fromobj)
4151  * zc_sendobj   objsetid of snapshot to send
4152  * zc_fromobj   objsetid of incremental fromsnap (may be zero)
4153  * zc_guid      if set, estimate size of stream only.  zc_cookie is ignored.
4154  *              output size in zc_objset_type.
4155  *
4156  * outputs: none
4157  */
4158 static int
4159 zfs_ioc_send(zfs_cmd_t *zc)
4160 {
4161         objset_t *fromsnap = NULL;
4162         objset_t *tosnap;
4163         int error;
4164         offset_t off;
4165         dsl_dataset_t *ds;
4166         dsl_dataset_t *dsfrom = NULL;
4167         spa_t *spa;
4168         dsl_pool_t *dp;
4169         boolean_t estimate = (zc->zc_guid != 0);
4170 
4171         error = spa_open(zc->zc_name, &spa, FTAG);
4172         if (error)
4173                 return (error);
4174 
4175         dp = spa_get_dsl(spa);
4176         rw_enter(&dp->dp_config_rwlock, RW_READER);
4177         error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
4178         rw_exit(&dp->dp_config_rwlock);
4179         spa_close(spa, FTAG);
4180         if (error)
4181                 return (error);
4182 
4183         error = dmu_objset_from_ds(ds, &tosnap);
4184         if (error) {
4185                 dsl_dataset_rele(ds, FTAG);
4186                 return (error);
4187         }
4188 
4189         if (zc->zc_fromobj != 0) {
4190                 rw_enter(&dp->dp_config_rwlock, RW_READER);
4191                 error = dsl_dataset_hold_obj(dp, zc->zc_fromobj, FTAG, &dsfrom);
4192                 rw_exit(&dp->dp_config_rwlock);
4193                 if (error) {
4194                         dsl_dataset_rele(ds, FTAG);
4195                         return (error);
4196                 }
4197                 error = dmu_objset_from_ds(dsfrom, &fromsnap);
4198                 if (error) {
4199                         dsl_dataset_rele(dsfrom, FTAG);
4200                         dsl_dataset_rele(ds, FTAG);
4201                         return (error);
4202                 }
4203         }
4204 
4205         if (zc->zc_obj) {
4206                 dsl_pool_t *dp = ds->ds_dir->dd_pool;
4207 
4208                 if (fromsnap != NULL) {
4209                         dsl_dataset_rele(dsfrom, FTAG);
4210                         dsl_dataset_rele(ds, FTAG);
4211                         return (EINVAL);
4212                 }
4213 
4214                 if (dsl_dir_is_clone(ds->ds_dir)) {
4215                         rw_enter(&dp->dp_config_rwlock, RW_READER);
4216                         error = dsl_dataset_hold_obj(dp,
4217                             ds->ds_dir->dd_phys->dd_origin_obj, FTAG, &dsfrom);
4218                         rw_exit(&dp->dp_config_rwlock);
4219                         if (error) {
4220                                 dsl_dataset_rele(ds, FTAG);
4221                                 return (error);
4222                         }
4223                         error = dmu_objset_from_ds(dsfrom, &fromsnap);
4224                         if (error) {
4225                                 dsl_dataset_rele(dsfrom, FTAG);
4226                                 dsl_dataset_rele(ds, FTAG);
4227                                 return (error);
4228                         }
4229                 }
4230         }
4231 
4232         if (estimate) {
4233                 error = dmu_send_estimate(tosnap, fromsnap,
4234                     &zc->zc_objset_type);
4235         } else {
4236                 file_t *fp = getf(zc->zc_cookie);
4237                 if (fp == NULL) {
4238                         dsl_dataset_rele(ds, FTAG);
4239                         if (dsfrom)
4240                                 dsl_dataset_rele(dsfrom, FTAG);
4241                         return (EBADF);
4242                 }
4243 
4244                 off = fp->f_offset;
4245                 error = dmu_send(tosnap, fromsnap,
4246                     zc->zc_cookie, fp->f_vnode, &off);
4247 
4248                 if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4249                         fp->f_offset = off;
4250                 releasef(zc->zc_cookie);
4251         }
4252         if (dsfrom)
4253                 dsl_dataset_rele(dsfrom, FTAG);
4254         dsl_dataset_rele(ds, FTAG);
4255         return (error);
4256 }
4257 
4258 /*
4259  * inputs:
4260  * zc_name      name of snapshot on which to report progress
4261  * zc_cookie    file descriptor of send stream
4262  *
4263  * outputs:
4264  * zc_cookie    number of bytes written in send stream thus far
4265  */
4266 static int
4267 zfs_ioc_send_progress(zfs_cmd_t *zc)
4268 {
4269         dsl_dataset_t *ds;
4270         dmu_sendarg_t *dsp = NULL;
4271         int error;
4272 
4273         if ((error = dsl_dataset_hold(zc->zc_name, FTAG, &ds)) != 0)
4274                 return (error);
4275 
4276         mutex_enter(&ds->ds_sendstream_lock);
4277 
4278         /*
4279          * Iterate over all the send streams currently active on this dataset.
4280          * If there's one which matches the specified file descriptor _and_ the
4281          * stream was started by the current process, return the progress of
4282          * that stream.
4283          */
4284         for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4285             dsp = list_next(&ds->ds_sendstreams, dsp)) {
4286                 if (dsp->dsa_outfd == zc->zc_cookie &&
4287                     dsp->dsa_proc == curproc)
4288                         break;
4289         }
4290 
4291         if (dsp != NULL)
4292                 zc->zc_cookie = *(dsp->dsa_off);
4293         else
4294                 error = ENOENT;
4295 
4296         mutex_exit(&ds->ds_sendstream_lock);
4297         dsl_dataset_rele(ds, FTAG);
4298         return (error);
4299 }
4300 
4301 static int
4302 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4303 {
4304         int id, error;
4305 
4306         error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4307             &zc->zc_inject_record);
4308 
4309         if (error == 0)
4310                 zc->zc_guid = (uint64_t)id;
4311 
4312         return (error);
4313 }
4314 
4315 static int
4316 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4317 {
4318         return (zio_clear_fault((int)zc->zc_guid));
4319 }
4320 
4321 static int
4322 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4323 {
4324         int id = (int)zc->zc_guid;
4325         int error;
4326 
4327         error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4328             &zc->zc_inject_record);
4329 
4330         zc->zc_guid = id;
4331 
4332         return (error);
4333 }
4334 
4335 static int
4336 zfs_ioc_error_log(zfs_cmd_t *zc)
4337 {
4338         spa_t *spa;
4339         int error;
4340         size_t count = (size_t)zc->zc_nvlist_dst_size;
4341 
4342         if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4343                 return (error);
4344 
4345         error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4346             &count);
4347         if (error == 0)
4348                 zc->zc_nvlist_dst_size = count;
4349         else
4350                 zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4351 
4352         spa_close(spa, FTAG);
4353 
4354         return (error);
4355 }
4356 
4357 static int
4358 zfs_ioc_clear(zfs_cmd_t *zc)
4359 {
4360         spa_t *spa;
4361         vdev_t *vd;
4362         int error;
4363 
4364         /*
4365          * On zpool clear we also fix up missing slogs
4366          */
4367         mutex_enter(&spa_namespace_lock);
4368         spa = spa_lookup(zc->zc_name);
4369         if (spa == NULL) {
4370                 mutex_exit(&spa_namespace_lock);
4371                 return (EIO);
4372         }
4373         if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4374                 /* we need to let spa_open/spa_load clear the chains */
4375                 spa_set_log_state(spa, SPA_LOG_CLEAR);
4376         }
4377         spa->spa_last_open_failed = 0;
4378         mutex_exit(&spa_namespace_lock);
4379 
4380         if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4381                 error = spa_open(zc->zc_name, &spa, FTAG);
4382         } else {
4383                 nvlist_t *policy;
4384                 nvlist_t *config = NULL;
4385 
4386                 if (zc->zc_nvlist_src == NULL)
4387                         return (EINVAL);
4388 
4389                 if ((error = get_nvlist(zc->zc_nvlist_src,
4390                     zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4391                         error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4392                             policy, &config);
4393                         if (config != NULL) {
4394                                 int err;
4395 
4396                                 if ((err = put_nvlist(zc, config)) != 0)
4397                                         error = err;
4398                                 nvlist_free(config);
4399                         }
4400                         nvlist_free(policy);
4401                 }
4402         }
4403 
4404         if (error)
4405                 return (error);
4406 
4407         spa_vdev_state_enter(spa, SCL_NONE);
4408 
4409         if (zc->zc_guid == 0) {
4410                 vd = NULL;
4411         } else {
4412                 vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4413                 if (vd == NULL) {
4414                         (void) spa_vdev_state_exit(spa, NULL, ENODEV);
4415                         spa_close(spa, FTAG);
4416                         return (ENODEV);
4417                 }
4418         }
4419 
4420         vdev_clear(spa, vd);
4421 
4422         (void) spa_vdev_state_exit(spa, NULL, 0);
4423 
4424         /*
4425          * Resume any suspended I/Os.
4426          */
4427         if (zio_resume(spa) != 0)
4428                 error = EIO;
4429 
4430         spa_close(spa, FTAG);
4431 
4432         return (error);
4433 }
4434 
4435 static int
4436 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4437 {
4438         spa_t *spa;
4439         int error;
4440 
4441         error = spa_open(zc->zc_name, &spa, FTAG);
4442         if (error)
4443                 return (error);
4444 
4445         spa_vdev_state_enter(spa, SCL_NONE);
4446 
4447         /*
4448          * If a resilver is already in progress then set the
4449          * spa_scrub_reopen flag to B_TRUE so that we don't restart
4450          * the scan as a side effect of the reopen. Otherwise, let
4451          * vdev_open() decided if a resilver is required.
4452          */
4453         spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4454         vdev_reopen(spa->spa_root_vdev);
4455         spa->spa_scrub_reopen = B_FALSE;
4456 
4457         (void) spa_vdev_state_exit(spa, NULL, 0);
4458         spa_close(spa, FTAG);
4459         return (0);
4460 }
4461 /*
4462  * inputs:
4463  * zc_name      name of filesystem
4464  * zc_value     name of origin snapshot
4465  *
4466  * outputs:
4467  * zc_string    name of conflicting snapshot, if there is one
4468  */
4469 static int
4470 zfs_ioc_promote(zfs_cmd_t *zc)
4471 {
4472         char *cp;
4473 
4474         /*
4475          * We don't need to unmount *all* the origin fs's snapshots, but
4476          * it's easier.
4477          */
4478         cp = strchr(zc->zc_value, '@');
4479         if (cp)
4480                 *cp = '\0';
4481         (void) dmu_objset_find(zc->zc_value,
4482             zfs_unmount_snap, NULL, DS_FIND_SNAPSHOTS);
4483         return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4484 }
4485 
4486 /*
4487  * Retrieve a single {user|group}{used|quota}@... property.
4488  *
4489  * inputs:
4490  * zc_name      name of filesystem
4491  * zc_objset_type zfs_userquota_prop_t
4492  * zc_value     domain name (eg. "S-1-234-567-89")
4493  * zc_guid      RID/UID/GID
4494  *
4495  * outputs:
4496  * zc_cookie    property value
4497  */
4498 static int
4499 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4500 {
4501         zfsvfs_t *zfsvfs;
4502         int error;
4503 
4504         if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4505                 return (EINVAL);
4506 
4507         error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4508         if (error)
4509                 return (error);
4510 
4511         error = zfs_userspace_one(zfsvfs,
4512             zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4513         zfsvfs_rele(zfsvfs, FTAG);
4514 
4515         return (error);
4516 }
4517 
4518 /*
4519  * inputs:
4520  * zc_name              name of filesystem
4521  * zc_cookie            zap cursor
4522  * zc_objset_type       zfs_userquota_prop_t
4523  * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4524  *
4525  * outputs:
4526  * zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
4527  * zc_cookie    zap cursor
4528  */
4529 static int
4530 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4531 {
4532         zfsvfs_t *zfsvfs;
4533         int bufsize = zc->zc_nvlist_dst_size;
4534 
4535         if (bufsize <= 0)
4536                 return (ENOMEM);
4537 
4538         int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4539         if (error)
4540                 return (error);
4541 
4542         void *buf = kmem_alloc(bufsize, KM_SLEEP);
4543 
4544         error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
4545             buf, &zc->zc_nvlist_dst_size);
4546 
4547         if (error == 0) {
4548                 error = xcopyout(buf,
4549                     (void *)(uintptr_t)zc->zc_nvlist_dst,
4550                     zc->zc_nvlist_dst_size);
4551         }
4552         kmem_free(buf, bufsize);
4553         zfsvfs_rele(zfsvfs, FTAG);
4554 
4555         return (error);
4556 }
4557 
4558 /*
4559  * inputs:
4560  * zc_name              name of filesystem
4561  *
4562  * outputs:
4563  * none
4564  */
4565 static int
4566 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4567 {
4568         objset_t *os;
4569         int error = 0;
4570         zfsvfs_t *zfsvfs;
4571 
4572         if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
4573                 if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
4574                         /*
4575                          * If userused is not enabled, it may be because the
4576                          * objset needs to be closed & reopened (to grow the
4577                          * objset_phys_t).  Suspend/resume the fs will do that.
4578                          */
4579                         error = zfs_suspend_fs(zfsvfs);
4580                         if (error == 0)
4581                                 error = zfs_resume_fs(zfsvfs, zc->zc_name);
4582                 }
4583                 if (error == 0)
4584                         error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
4585                 VFS_RELE(zfsvfs->z_vfs);
4586         } else {
4587                 /* XXX kind of reading contents without owning */
4588                 error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4589                 if (error)
4590                         return (error);
4591 
4592                 error = dmu_objset_userspace_upgrade(os);
4593                 dmu_objset_rele(os, FTAG);
4594         }
4595 
4596         return (error);
4597 }
4598 
4599 /*
4600  * We don't want to have a hard dependency
4601  * against some special symbols in sharefs
4602  * nfs, and smbsrv.  Determine them if needed when
4603  * the first file system is shared.
4604  * Neither sharefs, nfs or smbsrv are unloadable modules.
4605  */
4606 int (*znfsexport_fs)(void *arg);
4607 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
4608 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
4609 
4610 int zfs_nfsshare_inited;
4611 int zfs_smbshare_inited;
4612 
4613 ddi_modhandle_t nfs_mod;
4614 ddi_modhandle_t sharefs_mod;
4615 ddi_modhandle_t smbsrv_mod;
4616 kmutex_t zfs_share_lock;
4617 
4618 static int
4619 zfs_init_sharefs()
4620 {
4621         int error;
4622 
4623         ASSERT(MUTEX_HELD(&zfs_share_lock));
4624         /* Both NFS and SMB shares also require sharetab support. */
4625         if (sharefs_mod == NULL && ((sharefs_mod =
4626             ddi_modopen("fs/sharefs",
4627             KRTLD_MODE_FIRST, &error)) == NULL)) {
4628                 return (ENOSYS);
4629         }
4630         if (zshare_fs == NULL && ((zshare_fs =
4631             (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
4632             ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
4633                 return (ENOSYS);
4634         }
4635         return (0);
4636 }
4637 
4638 static int
4639 zfs_ioc_share(zfs_cmd_t *zc)
4640 {
4641         int error;
4642         int opcode;
4643 
4644         switch (zc->zc_share.z_sharetype) {
4645         case ZFS_SHARE_NFS:
4646         case ZFS_UNSHARE_NFS:
4647                 if (zfs_nfsshare_inited == 0) {
4648                         mutex_enter(&zfs_share_lock);
4649                         if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
4650                             KRTLD_MODE_FIRST, &error)) == NULL)) {
4651                                 mutex_exit(&zfs_share_lock);
4652                                 return (ENOSYS);
4653                         }
4654                         if (znfsexport_fs == NULL &&
4655                             ((znfsexport_fs = (int (*)(void *))
4656                             ddi_modsym(nfs_mod,
4657                             "nfs_export", &error)) == NULL)) {
4658                                 mutex_exit(&zfs_share_lock);
4659                                 return (ENOSYS);
4660                         }
4661                         error = zfs_init_sharefs();
4662                         if (error) {
4663                                 mutex_exit(&zfs_share_lock);
4664                                 return (ENOSYS);
4665                         }
4666                         zfs_nfsshare_inited = 1;
4667                         mutex_exit(&zfs_share_lock);
4668                 }
4669                 break;
4670         case ZFS_SHARE_SMB:
4671         case ZFS_UNSHARE_SMB:
4672                 if (zfs_smbshare_inited == 0) {
4673                         mutex_enter(&zfs_share_lock);
4674                         if (smbsrv_mod == NULL && ((smbsrv_mod =
4675                             ddi_modopen("drv/smbsrv",
4676                             KRTLD_MODE_FIRST, &error)) == NULL)) {
4677                                 mutex_exit(&zfs_share_lock);
4678                                 return (ENOSYS);
4679                         }
4680                         if (zsmbexport_fs == NULL && ((zsmbexport_fs =
4681                             (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
4682                             "smb_server_share", &error)) == NULL)) {
4683                                 mutex_exit(&zfs_share_lock);
4684                                 return (ENOSYS);
4685                         }
4686                         error = zfs_init_sharefs();
4687                         if (error) {
4688                                 mutex_exit(&zfs_share_lock);
4689                                 return (ENOSYS);
4690                         }
4691                         zfs_smbshare_inited = 1;
4692                         mutex_exit(&zfs_share_lock);
4693                 }
4694                 break;
4695         default:
4696                 return (EINVAL);
4697         }
4698 
4699         switch (zc->zc_share.z_sharetype) {
4700         case ZFS_SHARE_NFS:
4701         case ZFS_UNSHARE_NFS:
4702                 if (error =
4703                     znfsexport_fs((void *)
4704                     (uintptr_t)zc->zc_share.z_exportdata))
4705                         return (error);
4706                 break;
4707         case ZFS_SHARE_SMB:
4708         case ZFS_UNSHARE_SMB:
4709                 if (error = zsmbexport_fs((void *)
4710                     (uintptr_t)zc->zc_share.z_exportdata,
4711                     zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
4712                     B_TRUE: B_FALSE)) {
4713                         return (error);
4714                 }
4715                 break;
4716         }
4717 
4718         opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
4719             zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
4720             SHAREFS_ADD : SHAREFS_REMOVE;
4721 
4722         /*
4723          * Add or remove share from sharetab
4724          */
4725         error = zshare_fs(opcode,
4726             (void *)(uintptr_t)zc->zc_share.z_sharedata,
4727             zc->zc_share.z_sharemax);
4728 
4729         return (error);
4730 
4731 }
4732 
4733 ace_t full_access[] = {
4734         {(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4735 };
4736 
4737 /*
4738  * inputs:
4739  * zc_name              name of containing filesystem
4740  * zc_obj               object # beyond which we want next in-use object #
4741  *
4742  * outputs:
4743  * zc_obj               next in-use object #
4744  */
4745 static int
4746 zfs_ioc_next_obj(zfs_cmd_t *zc)
4747 {
4748         objset_t *os = NULL;
4749         int error;
4750 
4751         error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4752         if (error)
4753                 return (error);
4754 
4755         error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4756             os->os_dsl_dataset->ds_phys->ds_prev_snap_txg);
4757 
4758         dmu_objset_rele(os, FTAG);
4759         return (error);
4760 }
4761 
4762 /*
4763  * inputs:
4764  * zc_name              name of filesystem
4765  * zc_value             prefix name for snapshot
4766  * zc_cleanup_fd        cleanup-on-exit file descriptor for calling process
4767  *
4768  * outputs:
4769  * zc_value             short name of new snapshot
4770  */
4771 static int
4772 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
4773 {
4774         char *snap_name;
4775         int error;
4776 
4777         snap_name = kmem_asprintf("%s@%s-%016llx", zc->zc_name, zc->zc_value,
4778             (u_longlong_t)ddi_get_lbolt64());
4779 
4780         if (strlen(snap_name) >= MAXPATHLEN) {
4781                 strfree(snap_name);
4782                 return (E2BIG);
4783         }
4784 
4785         error = dmu_objset_snapshot_tmp(snap_name, "%temp", zc->zc_cleanup_fd);
4786         if (error != 0) {
4787                 strfree(snap_name);
4788                 return (error);
4789         }
4790 
4791         (void) strcpy(zc->zc_value, strchr(snap_name, '@') + 1);
4792         strfree(snap_name);
4793         return (0);
4794 }
4795 
4796 /*
4797  * inputs:
4798  * zc_name              name of "to" snapshot
4799  * zc_value             name of "from" snapshot
4800  * zc_cookie            file descriptor to write diff data on
4801  *
4802  * outputs:
4803  * dmu_diff_record_t's to the file descriptor
4804  */
4805 static int
4806 zfs_ioc_diff(zfs_cmd_t *zc)
4807 {
4808         objset_t *fromsnap;
4809         objset_t *tosnap;
4810         file_t *fp;
4811         offset_t off;
4812         int error;
4813 
4814         error = dmu_objset_hold(zc->zc_name, FTAG, &tosnap);
4815         if (error)
4816                 return (error);
4817 
4818         error = dmu_objset_hold(zc->zc_value, FTAG, &fromsnap);
4819         if (error) {
4820                 dmu_objset_rele(tosnap, FTAG);
4821                 return (error);
4822         }
4823 
4824         fp = getf(zc->zc_cookie);
4825         if (fp == NULL) {
4826                 dmu_objset_rele(fromsnap, FTAG);
4827                 dmu_objset_rele(tosnap, FTAG);
4828                 return (EBADF);
4829         }
4830 
4831         off = fp->f_offset;
4832 
4833         error = dmu_diff(tosnap, fromsnap, fp->f_vnode, &off);
4834 
4835         if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4836                 fp->f_offset = off;
4837         releasef(zc->zc_cookie);
4838 
4839         dmu_objset_rele(fromsnap, FTAG);
4840         dmu_objset_rele(tosnap, FTAG);
4841         return (error);
4842 }
4843 
4844 /*
4845  * Remove all ACL files in shares dir
4846  */
4847 static int
4848 zfs_smb_acl_purge(znode_t *dzp)
4849 {
4850         zap_cursor_t    zc;
4851         zap_attribute_t zap;
4852         zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
4853         int error;
4854 
4855         for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
4856             (error = zap_cursor_retrieve(&zc, &zap)) == 0;
4857             zap_cursor_advance(&zc)) {
4858                 if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
4859                     NULL, 0)) != 0)
4860                         break;
4861         }
4862         zap_cursor_fini(&zc);
4863         return (error);
4864 }
4865 
4866 static int
4867 zfs_ioc_smb_acl(zfs_cmd_t *zc)
4868 {
4869         vnode_t *vp;
4870         znode_t *dzp;
4871         vnode_t *resourcevp = NULL;
4872         znode_t *sharedir;
4873         zfsvfs_t *zfsvfs;
4874         nvlist_t *nvlist;
4875         char *src, *target;
4876         vattr_t vattr;
4877         vsecattr_t vsec;
4878         int error = 0;
4879 
4880         if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
4881             NO_FOLLOW, NULL, &vp)) != 0)
4882                 return (error);
4883 
4884         /* Now make sure mntpnt and dataset are ZFS */
4885 
4886         if (vp->v_vfsp->vfs_fstype != zfsfstype ||
4887             (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
4888             zc->zc_name) != 0)) {
4889                 VN_RELE(vp);
4890                 return (EINVAL);
4891         }
4892 
4893         dzp = VTOZ(vp);
4894         zfsvfs = dzp->z_zfsvfs;
4895         ZFS_ENTER(zfsvfs);
4896 
4897         /*
4898          * Create share dir if its missing.
4899          */
4900         mutex_enter(&zfsvfs->z_lock);
4901         if (zfsvfs->z_shares_dir == 0) {
4902                 dmu_tx_t *tx;
4903 
4904                 tx = dmu_tx_create(zfsvfs->z_os);
4905                 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
4906                     ZFS_SHARES_DIR);
4907                 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
4908                 error = dmu_tx_assign(tx, TXG_WAIT);
4909                 if (error) {
4910                         dmu_tx_abort(tx);
4911                 } else {
4912                         error = zfs_create_share_dir(zfsvfs, tx);
4913                         dmu_tx_commit(tx);
4914                 }
4915                 if (error) {
4916                         mutex_exit(&zfsvfs->z_lock);
4917                         VN_RELE(vp);
4918                         ZFS_EXIT(zfsvfs);
4919                         return (error);
4920                 }
4921         }
4922         mutex_exit(&zfsvfs->z_lock);
4923 
4924         ASSERT(zfsvfs->z_shares_dir);
4925         if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
4926                 VN_RELE(vp);
4927                 ZFS_EXIT(zfsvfs);
4928                 return (error);
4929         }
4930 
4931         switch (zc->zc_cookie) {
4932         case ZFS_SMB_ACL_ADD:
4933                 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
4934                 vattr.va_type = VREG;
4935                 vattr.va_mode = S_IFREG|0777;
4936                 vattr.va_uid = 0;
4937                 vattr.va_gid = 0;
4938 
4939                 vsec.vsa_mask = VSA_ACE;
4940                 vsec.vsa_aclentp = &full_access;
4941                 vsec.vsa_aclentsz = sizeof (full_access);
4942                 vsec.vsa_aclcnt = 1;
4943 
4944                 error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
4945                     &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
4946                 if (resourcevp)
4947                         VN_RELE(resourcevp);
4948                 break;
4949 
4950         case ZFS_SMB_ACL_REMOVE:
4951                 error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
4952                     NULL, 0);
4953                 break;
4954 
4955         case ZFS_SMB_ACL_RENAME:
4956                 if ((error = get_nvlist(zc->zc_nvlist_src,
4957                     zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
4958                         VN_RELE(vp);
4959                         ZFS_EXIT(zfsvfs);
4960                         return (error);
4961                 }
4962                 if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
4963                     nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
4964                     &target)) {
4965                         VN_RELE(vp);
4966                         VN_RELE(ZTOV(sharedir));
4967                         ZFS_EXIT(zfsvfs);
4968                         nvlist_free(nvlist);
4969                         return (error);
4970                 }
4971                 error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
4972                     kcred, NULL, 0);
4973                 nvlist_free(nvlist);
4974                 break;
4975 
4976         case ZFS_SMB_ACL_PURGE:
4977                 error = zfs_smb_acl_purge(sharedir);
4978                 break;
4979 
4980         default:
4981                 error = EINVAL;
4982                 break;
4983         }
4984 
4985         VN_RELE(vp);
4986         VN_RELE(ZTOV(sharedir));
4987 
4988         ZFS_EXIT(zfsvfs);
4989 
4990         return (error);
4991 }
4992 
4993 /*
4994  * inputs:
4995  * zc_name              name of filesystem
4996  * zc_value             short name of snap
4997  * zc_string            user-supplied tag for this hold
4998  * zc_cookie            recursive flag
4999  * zc_temphold          set if hold is temporary
5000  * zc_cleanup_fd        cleanup-on-exit file descriptor for calling process
5001  * zc_sendobj           if non-zero, the objid for zc_name@zc_value
5002  * zc_createtxg         if zc_sendobj is non-zero, snap must have zc_createtxg
5003  *
5004  * outputs:             none
5005  */
5006 static int
5007 zfs_ioc_hold(zfs_cmd_t *zc)
5008 {
5009         boolean_t recursive = zc->zc_cookie;
5010         spa_t *spa;
5011         dsl_pool_t *dp;
5012         dsl_dataset_t *ds;
5013         int error;
5014         minor_t minor = 0;
5015 
5016         if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
5017                 return (EINVAL);
5018 
5019         if (zc->zc_sendobj == 0) {
5020                 return (dsl_dataset_user_hold(zc->zc_name, zc->zc_value,
5021                     zc->zc_string, recursive, zc->zc_temphold,
5022                     zc->zc_cleanup_fd));
5023         }
5024 
5025         if (recursive)
5026                 return (EINVAL);
5027 
5028         error = spa_open(zc->zc_name, &spa, FTAG);
5029         if (error)
5030                 return (error);
5031 
5032         dp = spa_get_dsl(spa);
5033         rw_enter(&dp->dp_config_rwlock, RW_READER);
5034         error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
5035         rw_exit(&dp->dp_config_rwlock);
5036         spa_close(spa, FTAG);
5037         if (error)
5038                 return (error);
5039 
5040         /*
5041          * Until we have a hold on this snapshot, it's possible that
5042          * zc_sendobj could've been destroyed and reused as part
5043          * of a later txg.  Make sure we're looking at the right object.
5044          */
5045         if (zc->zc_createtxg != ds->ds_phys->ds_creation_txg) {
5046                 dsl_dataset_rele(ds, FTAG);
5047                 return (ENOENT);
5048         }
5049 
5050         if (zc->zc_cleanup_fd != -1 && zc->zc_temphold) {
5051                 error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
5052                 if (error) {
5053                         dsl_dataset_rele(ds, FTAG);
5054                         return (error);
5055                 }
5056         }
5057 
5058         error = dsl_dataset_user_hold_for_send(ds, zc->zc_string,
5059             zc->zc_temphold);
5060         if (minor != 0) {
5061                 if (error == 0) {
5062                         dsl_register_onexit_hold_cleanup(ds, zc->zc_string,
5063                             minor);
5064                 }
5065                 zfs_onexit_fd_rele(zc->zc_cleanup_fd);
5066         }
5067         dsl_dataset_rele(ds, FTAG);
5068 
5069         return (error);
5070 }
5071 
5072 /*
5073  * inputs:
5074  * zc_name      name of dataset from which we're releasing a user hold
5075  * zc_value     short name of snap
5076  * zc_string    user-supplied tag for this hold
5077  * zc_cookie    recursive flag
5078  *
5079  * outputs:     none
5080  */
5081 static int
5082 zfs_ioc_release(zfs_cmd_t *zc)
5083 {
5084         boolean_t recursive = zc->zc_cookie;
5085 
5086         if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
5087                 return (EINVAL);
5088 
5089         return (dsl_dataset_user_release(zc->zc_name, zc->zc_value,
5090             zc->zc_string, recursive));
5091 }
5092 
5093 /*
5094  * inputs:
5095  * zc_name              name of filesystem
5096  *
5097  * outputs:
5098  * zc_nvlist_src{_size} nvlist of snapshot holds
5099  */
5100 static int
5101 zfs_ioc_get_holds(zfs_cmd_t *zc)
5102 {
5103         nvlist_t *nvp;
5104         int error;
5105 
5106         if ((error = dsl_dataset_get_holds(zc->zc_name, &nvp)) == 0) {
5107                 error = put_nvlist(zc, nvp);
5108                 nvlist_free(nvp);
5109         }
5110 
5111         return (error);
5112 }
5113 
5114 /*
5115  * inputs:
5116  * zc_name              name of new filesystem or snapshot
5117  * zc_value             full name of old snapshot
5118  *
5119  * outputs:
5120  * zc_cookie            space in bytes
5121  * zc_objset_type       compressed space in bytes
5122  * zc_perm_action       uncompressed space in bytes
5123  */
5124 static int
5125 zfs_ioc_space_written(zfs_cmd_t *zc)
5126 {
5127         int error;
5128         dsl_dataset_t *new, *old;
5129 
5130         error = dsl_dataset_hold(zc->zc_name, FTAG, &new);
5131         if (error != 0)
5132                 return (error);
5133         error = dsl_dataset_hold(zc->zc_value, FTAG, &old);
5134         if (error != 0) {
5135                 dsl_dataset_rele(new, FTAG);
5136                 return (error);
5137         }
5138 
5139         error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5140             &zc->zc_objset_type, &zc->zc_perm_action);
5141         dsl_dataset_rele(old, FTAG);
5142         dsl_dataset_rele(new, FTAG);
5143         return (error);
5144 }
5145 /*
5146  * innvl: {
5147  *     "firstsnap" -> snapshot name
5148  * }
5149  *
5150  * outnvl: {
5151  *     "used" -> space in bytes
5152  *     "compressed" -> compressed space in bytes
5153  *     "uncompressed" -> uncompressed space in bytes
5154  * }
5155  */
5156 static int
5157 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5158 {
5159         int error;
5160         dsl_dataset_t *new, *old;
5161         char *firstsnap;
5162         uint64_t used, comp, uncomp;
5163 
5164         if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5165                 return (EINVAL);
5166 
5167         error = dsl_dataset_hold(lastsnap, FTAG, &new);
5168         if (error != 0)
5169                 return (error);
5170         error = dsl_dataset_hold(firstsnap, FTAG, &old);
5171         if (error != 0) {
5172                 dsl_dataset_rele(new, FTAG);
5173                 return (error);
5174         }
5175 
5176         error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5177         dsl_dataset_rele(old, FTAG);
5178         dsl_dataset_rele(new, FTAG);
5179         fnvlist_add_uint64(outnvl, "used", used);
5180         fnvlist_add_uint64(outnvl, "compressed", comp);
5181         fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5182         return (error);
5183 }
5184 
5185 /*
5186  * innvl: {
5187  *     "fd" -> file descriptor to write stream to (int32)
5188  *     (optional) "fromsnap" -> full snap name to send an incremental from
5189  * }
5190  *
5191  * outnvl is unused
5192  */
5193 /* ARGSUSED */
5194 static int
5195 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5196 {
5197         objset_t *fromsnap = NULL;
5198         objset_t *tosnap;
5199         int error;
5200         offset_t off;
5201         char *fromname;
5202         int fd;
5203 
5204         error = nvlist_lookup_int32(innvl, "fd", &fd);
5205         if (error != 0)
5206                 return (EINVAL);
5207 
5208         error = dmu_objset_hold(snapname, FTAG, &tosnap);
5209         if (error)
5210                 return (error);
5211 
5212         error = nvlist_lookup_string(innvl, "fromsnap", &fromname);
5213         if (error == 0) {
5214                 error = dmu_objset_hold(fromname, FTAG, &fromsnap);
5215                 if (error) {
5216                         dmu_objset_rele(tosnap, FTAG);
5217                         return (error);
5218                 }
5219         }
5220 
5221         file_t *fp = getf(fd);
5222         if (fp == NULL) {
5223                 dmu_objset_rele(tosnap, FTAG);
5224                 if (fromsnap != NULL)
5225                         dmu_objset_rele(fromsnap, FTAG);
5226                 return (EBADF);
5227         }
5228 
5229         off = fp->f_offset;
5230         error = dmu_send(tosnap, fromsnap, fd, fp->f_vnode, &off);
5231 
5232         if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5233                 fp->f_offset = off;
5234         releasef(fd);
5235         if (fromsnap != NULL)
5236                 dmu_objset_rele(fromsnap, FTAG);
5237         dmu_objset_rele(tosnap, FTAG);
5238         return (error);
5239 }
5240 
5241 /*
5242  * Determine approximately how large a zfs send stream will be -- the number
5243  * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5244  *
5245  * innvl: {
5246  *     (optional) "fromsnap" -> full snap name to send an incremental from
5247  * }
5248  *
5249  * outnvl: {
5250  *     "space" -> bytes of space (uint64)
5251  * }
5252  */
5253 static int
5254 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5255 {
5256         objset_t *fromsnap = NULL;
5257         objset_t *tosnap;
5258         int error;
5259         char *fromname;
5260         uint64_t space;
5261 
5262         error = dmu_objset_hold(snapname, FTAG, &tosnap);
5263         if (error)
5264                 return (error);
5265 
5266         error = nvlist_lookup_string(innvl, "fromsnap", &fromname);
5267         if (error == 0) {
5268                 error = dmu_objset_hold(fromname, FTAG, &fromsnap);
5269                 if (error) {
5270                         dmu_objset_rele(tosnap, FTAG);
5271                         return (error);
5272                 }
5273         }
5274 
5275         error = dmu_send_estimate(tosnap, fromsnap, &space);
5276         fnvlist_add_uint64(outnvl, "space", space);
5277 
5278         if (fromsnap != NULL)
5279                 dmu_objset_rele(fromsnap, FTAG);
5280         dmu_objset_rele(tosnap, FTAG);
5281         return (error);
5282 }
5283 
5284 
5285 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5286 
5287 static void
5288 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5289     zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5290     boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5291 {
5292         zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5293 
5294         ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5295         ASSERT3U(ioc, <, ZFS_IOC_LAST);
5296         ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5297         ASSERT3P(vec->zvec_func, ==, NULL);
5298 
5299         vec->zvec_legacy_func = func;
5300         vec->zvec_secpolicy = secpolicy;
5301         vec->zvec_namecheck = namecheck;
5302         vec->zvec_allow_log = log_history;
5303         vec->zvec_pool_check = pool_check;
5304 }
5305 
5306 /*
5307  * See the block comment at the beginning of this file for details on
5308  * each argument to this function.
5309  */
5310 static void
5311 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5312     zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5313     zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5314     boolean_t allow_log)
5315 {
5316         zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5317 
5318         ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5319         ASSERT3U(ioc, <, ZFS_IOC_LAST);
5320         ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5321         ASSERT3P(vec->zvec_func, ==, NULL);
5322 
5323         /* if we are logging, the name must be valid */
5324         ASSERT(!allow_log || namecheck != NO_NAME);
5325 
5326         vec->zvec_name = name;
5327         vec->zvec_func = func;
5328         vec->zvec_secpolicy = secpolicy;
5329         vec->zvec_namecheck = namecheck;
5330         vec->zvec_pool_check = pool_check;
5331         vec->zvec_smush_outnvlist = smush_outnvlist;
5332         vec->zvec_allow_log = allow_log;
5333 }
5334 
5335 static void
5336 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5337     zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5338     zfs_ioc_poolcheck_t pool_check)
5339 {
5340         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5341             POOL_NAME, log_history, pool_check);
5342 }
5343 
5344 static void
5345 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5346     zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5347 {
5348         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5349             DATASET_NAME, B_FALSE, pool_check);
5350 }
5351 
5352 static void
5353 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5354 {
5355         zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5356             POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5357 }
5358 
5359 static void
5360 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5361     zfs_secpolicy_func_t *secpolicy)
5362 {
5363         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5364             NO_NAME, B_FALSE, POOL_CHECK_NONE);
5365 }
5366 
5367 static void
5368 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5369     zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5370 {
5371         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5372             DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5373 }
5374 
5375 static void
5376 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5377 {
5378         zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5379             zfs_secpolicy_read);
5380 }
5381 
5382 static void
5383 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5384         zfs_secpolicy_func_t *secpolicy)
5385 {
5386         zfs_ioctl_register_legacy(ioc, func, secpolicy,
5387             DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5388 }
5389 
5390 static void
5391 zfs_ioctl_init(void)
5392 {
5393         zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5394             zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5395             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5396 
5397         zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5398             zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5399             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5400 
5401         zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5402             zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5403             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5404 
5405         zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5406             zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5407             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5408 
5409         zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5410             zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5411             POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5412 
5413         zfs_ioctl_register("create", ZFS_IOC_CREATE,
5414             zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5415             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5416 
5417         zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5418             zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5419             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5420 
5421         zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5422             zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5423             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5424 
5425         /* IOCTLS that use the legacy function signature */
5426 
5427         zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5428             zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5429 
5430         zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5431             zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5432         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5433             zfs_ioc_pool_scan);
5434         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5435             zfs_ioc_pool_upgrade);
5436         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5437             zfs_ioc_vdev_add);
5438         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5439             zfs_ioc_vdev_remove);
5440         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5441             zfs_ioc_vdev_set_state);
5442         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5443             zfs_ioc_vdev_attach);
5444         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5445             zfs_ioc_vdev_detach);
5446         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5447             zfs_ioc_vdev_setpath);
5448         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5449             zfs_ioc_vdev_setfru);
5450         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5451             zfs_ioc_pool_set_props);
5452         zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5453             zfs_ioc_vdev_split);
5454         zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5455             zfs_ioc_pool_reguid);
5456 
5457         zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5458             zfs_ioc_pool_configs, zfs_secpolicy_none);
5459         zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5460             zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5461         zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5462             zfs_ioc_inject_fault, zfs_secpolicy_inject);
5463         zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5464             zfs_ioc_clear_fault, zfs_secpolicy_inject);
5465         zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
5466             zfs_ioc_inject_list_next, zfs_secpolicy_inject);
5467 
5468         /*
5469          * pool destroy, and export don't log the history as part of
5470          * zfsdev_ioctl, but rather zfs_ioc_pool_export
5471          * does the logging of those commands.
5472          */
5473         zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
5474             zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5475         zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
5476             zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5477 
5478         zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
5479             zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5480         zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
5481             zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5482 
5483         zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
5484             zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
5485         zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
5486             zfs_ioc_dsobj_to_dsname,
5487             zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
5488         zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
5489             zfs_ioc_pool_get_history,
5490             zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
5491 
5492         zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
5493             zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5494 
5495         zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
5496             zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5497         zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
5498             zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5499 
5500         zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
5501             zfs_ioc_space_written);
5502         zfs_ioctl_register_dataset_read(ZFS_IOC_GET_HOLDS,
5503             zfs_ioc_get_holds);
5504         zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
5505             zfs_ioc_objset_recvd_props);
5506         zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
5507             zfs_ioc_next_obj);
5508         zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
5509             zfs_ioc_get_fsacl);
5510         zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
5511             zfs_ioc_objset_stats);
5512         zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
5513             zfs_ioc_objset_zplprops);
5514         zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
5515             zfs_ioc_dataset_list_next);
5516         zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
5517             zfs_ioc_snapshot_list_next);
5518         zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
5519             zfs_ioc_send_progress);
5520 
5521         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
5522             zfs_ioc_diff, zfs_secpolicy_diff);
5523         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
5524             zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
5525         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
5526             zfs_ioc_obj_to_path, zfs_secpolicy_diff);
5527         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
5528             zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
5529         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
5530             zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
5531         zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
5532             zfs_ioc_send, zfs_secpolicy_send);
5533 
5534         zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
5535             zfs_secpolicy_none);
5536         zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
5537             zfs_secpolicy_destroy);
5538         zfs_ioctl_register_dataset_modify(ZFS_IOC_ROLLBACK, zfs_ioc_rollback,
5539             zfs_secpolicy_rollback);
5540         zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
5541             zfs_secpolicy_rename);
5542         zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
5543             zfs_secpolicy_recv);
5544         zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
5545             zfs_secpolicy_promote);
5546         zfs_ioctl_register_dataset_modify(ZFS_IOC_HOLD, zfs_ioc_hold,
5547             zfs_secpolicy_hold);
5548         zfs_ioctl_register_dataset_modify(ZFS_IOC_RELEASE, zfs_ioc_release,
5549             zfs_secpolicy_release);
5550         zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
5551             zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
5552         zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
5553             zfs_secpolicy_set_fsacl);
5554 
5555         zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
5556             zfs_secpolicy_share, POOL_CHECK_NONE);
5557         zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
5558             zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
5559         zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
5560             zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
5561             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5562         zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
5563             zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
5564             POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5565 }
5566 
5567 int
5568 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
5569     zfs_ioc_poolcheck_t check)
5570 {
5571         spa_t *spa;
5572         int error;
5573 
5574         ASSERT(type == POOL_NAME || type == DATASET_NAME);
5575 
5576         if (check & POOL_CHECK_NONE)
5577                 return (0);
5578 
5579         error = spa_open(name, &spa, FTAG);
5580         if (error == 0) {
5581                 if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
5582                         error = EAGAIN;
5583                 else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
5584                         error = EROFS;
5585                 spa_close(spa, FTAG);
5586         }
5587         return (error);
5588 }
5589 
5590 /*
5591  * Find a free minor number.
5592  */
5593 minor_t
5594 zfsdev_minor_alloc(void)
5595 {
5596         static minor_t last_minor;
5597         minor_t m;
5598 
5599         ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5600 
5601         for (m = last_minor + 1; m != last_minor; m++) {
5602                 if (m > ZFSDEV_MAX_MINOR)
5603                         m = 1;
5604                 if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
5605                         last_minor = m;
5606                         return (m);
5607                 }
5608         }
5609 
5610         return (0);
5611 }
5612 
5613 static int
5614 zfs_ctldev_init(dev_t *devp)
5615 {
5616         minor_t minor;
5617         zfs_soft_state_t *zs;
5618 
5619         ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5620         ASSERT(getminor(*devp) == 0);
5621 
5622         minor = zfsdev_minor_alloc();
5623         if (minor == 0)
5624                 return (ENXIO);
5625 
5626         if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
5627                 return (EAGAIN);
5628 
5629         *devp = makedevice(getemajor(*devp), minor);
5630 
5631         zs = ddi_get_soft_state(zfsdev_state, minor);
5632         zs->zss_type = ZSST_CTLDEV;
5633         zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
5634 
5635         return (0);
5636 }
5637 
5638 static void
5639 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
5640 {
5641         ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5642 
5643         zfs_onexit_destroy(zo);
5644         ddi_soft_state_free(zfsdev_state, minor);
5645 }
5646 
5647 void *
5648 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
5649 {
5650         zfs_soft_state_t *zp;
5651 
5652         zp = ddi_get_soft_state(zfsdev_state, minor);
5653         if (zp == NULL || zp->zss_type != which)
5654                 return (NULL);
5655 
5656         return (zp->zss_data);
5657 }
5658 
5659 static int
5660 zfsdev_open(dev_t *devp, int flag, int otyp, cred_t *cr)
5661 {
5662         int error = 0;
5663 
5664         if (getminor(*devp) != 0)
5665                 return (zvol_open(devp, flag, otyp, cr));
5666 
5667         /* This is the control device. Allocate a new minor if requested. */
5668         if (flag & FEXCL) {
5669                 mutex_enter(&zfsdev_state_lock);
5670                 error = zfs_ctldev_init(devp);
5671                 mutex_exit(&zfsdev_state_lock);
5672         }
5673 
5674         return (error);
5675 }
5676 
5677 static int
5678 zfsdev_close(dev_t dev, int flag, int otyp, cred_t *cr)
5679 {
5680         zfs_onexit_t *zo;
5681         minor_t minor = getminor(dev);
5682 
5683         if (minor == 0)
5684                 return (0);
5685 
5686         mutex_enter(&zfsdev_state_lock);
5687         zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
5688         if (zo == NULL) {
5689                 mutex_exit(&zfsdev_state_lock);
5690                 return (zvol_close(dev, flag, otyp, cr));
5691         }
5692         zfs_ctldev_destroy(zo, minor);
5693         mutex_exit(&zfsdev_state_lock);
5694 
5695         return (0);
5696 }
5697 
5698 static int
5699 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
5700 {
5701         zfs_cmd_t *zc;
5702         uint_t vecnum;
5703         int error, rc, len;
5704         minor_t minor = getminor(dev);
5705         const zfs_ioc_vec_t *vec;
5706         char *saved_poolname = NULL;
5707         nvlist_t *innvl = NULL;
5708 
5709         if (minor != 0 &&
5710             zfsdev_get_soft_state(minor, ZSST_CTLDEV) == NULL)
5711                 return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
5712 
5713         vecnum = cmd - ZFS_IOC_FIRST;
5714         ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
5715 
5716         if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
5717                 return (EINVAL);
5718         vec = &zfs_ioc_vec[vecnum];
5719 
5720         zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
5721 
5722         error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
5723         if (error != 0) {
5724                 error = EFAULT;
5725                 goto out;
5726         }
5727 
5728         zc->zc_iflags = flag & FKIOCTL;
5729         if (zc->zc_nvlist_src_size != 0) {
5730                 error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
5731                     zc->zc_iflags, &innvl);
5732                 if (error != 0)
5733                         goto out;
5734         }
5735 
5736         /*
5737          * Ensure that all pool/dataset names are valid before we pass down to
5738          * the lower layers.
5739          */
5740         zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
5741         switch (vec->zvec_namecheck) {
5742         case POOL_NAME:
5743                 if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
5744                         error = EINVAL;
5745                 else
5746                         error = pool_status_check(zc->zc_name,
5747                             vec->zvec_namecheck, vec->zvec_pool_check);
5748                 break;
5749 
5750         case DATASET_NAME:
5751                 if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
5752                         error = EINVAL;
5753                 else
5754                         error = pool_status_check(zc->zc_name,
5755                             vec->zvec_namecheck, vec->zvec_pool_check);
5756                 break;
5757 
5758         case NO_NAME:
5759                 break;
5760         }
5761 
5762 
5763         if (error == 0 && !(flag & FKIOCTL))
5764                 error = vec->zvec_secpolicy(zc, innvl, cr);
5765 
5766         if (error != 0)
5767                 goto out;
5768 
5769         /* legacy ioctls can modify zc_name */
5770         len = strcspn(zc->zc_name, "/@") + 1;
5771         saved_poolname = kmem_alloc(len, KM_SLEEP);
5772         (void) strlcpy(saved_poolname, zc->zc_name, len);
5773 
5774         if (vec->zvec_func != NULL) {
5775                 nvlist_t *outnvl;
5776                 int puterror = 0;
5777                 spa_t *spa;
5778                 nvlist_t *lognv = NULL;
5779 
5780                 ASSERT(vec->zvec_legacy_func == NULL);
5781 
5782                 /*
5783                  * Add the innvl to the lognv before calling the func,
5784                  * in case the func changes the innvl.
5785                  */
5786                 if (vec->zvec_allow_log) {
5787                         lognv = fnvlist_alloc();
5788                         fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
5789                             vec->zvec_name);
5790                         if (!nvlist_empty(innvl)) {
5791                                 fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
5792                                     innvl);
5793                         }
5794                 }
5795 
5796                 outnvl = fnvlist_alloc();
5797                 error = vec->zvec_func(zc->zc_name, innvl, outnvl);
5798 
5799                 if (error == 0 && vec->zvec_allow_log &&
5800                     spa_open(zc->zc_name, &spa, FTAG) == 0) {
5801                         if (!nvlist_empty(outnvl)) {
5802                                 fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
5803                                     outnvl);
5804                         }
5805                         (void) spa_history_log_nvl(spa, lognv);
5806                         spa_close(spa, FTAG);
5807                 }
5808                 fnvlist_free(lognv);
5809 
5810                 if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
5811                         int smusherror = 0;
5812                         if (vec->zvec_smush_outnvlist) {
5813                                 smusherror = nvlist_smush(outnvl,
5814                                     zc->zc_nvlist_dst_size);
5815                         }
5816                         if (smusherror == 0)
5817                                 puterror = put_nvlist(zc, outnvl);
5818                 }
5819 
5820                 if (puterror != 0)
5821                         error = puterror;
5822 
5823                 nvlist_free(outnvl);
5824         } else {
5825                 error = vec->zvec_legacy_func(zc);
5826         }
5827 
5828 out:
5829         nvlist_free(innvl);
5830         rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
5831         if (error == 0 && rc != 0)
5832                 error = EFAULT;
5833         if (error == 0 && vec->zvec_allow_log) {
5834                 char *s = tsd_get(zfs_allow_log_key);
5835                 if (s != NULL)
5836                         strfree(s);
5837                 (void) tsd_set(zfs_allow_log_key, saved_poolname);
5838         } else {
5839                 if (saved_poolname != NULL)
5840                         strfree(saved_poolname);
5841         }
5842 
5843         kmem_free(zc, sizeof (zfs_cmd_t));
5844         return (error);
5845 }
5846 
5847 static int
5848 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
5849 {
5850         if (cmd != DDI_ATTACH)
5851                 return (DDI_FAILURE);
5852 
5853         if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
5854             DDI_PSEUDO, 0) == DDI_FAILURE)
5855                 return (DDI_FAILURE);
5856 
5857         zfs_dip = dip;
5858 
5859         ddi_report_dev(dip);
5860 
5861         return (DDI_SUCCESS);
5862 }
5863 
5864 static int
5865 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
5866 {
5867         if (spa_busy() || zfs_busy() || zvol_busy())
5868                 return (DDI_FAILURE);
5869 
5870         if (cmd != DDI_DETACH)
5871                 return (DDI_FAILURE);
5872 
5873         zfs_dip = NULL;
5874 
5875         ddi_prop_remove_all(dip);
5876         ddi_remove_minor_node(dip, NULL);
5877 
5878         return (DDI_SUCCESS);
5879 }
5880 
5881 /*ARGSUSED*/
5882 static int
5883 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
5884 {
5885         switch (infocmd) {
5886         case DDI_INFO_DEVT2DEVINFO:
5887                 *result = zfs_dip;
5888                 return (DDI_SUCCESS);
5889 
5890         case DDI_INFO_DEVT2INSTANCE:
5891                 *result = (void *)0;
5892                 return (DDI_SUCCESS);
5893         }
5894 
5895         return (DDI_FAILURE);
5896 }
5897 
5898 /*
5899  * OK, so this is a little weird.
5900  *
5901  * /dev/zfs is the control node, i.e. minor 0.
5902  * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
5903  *
5904  * /dev/zfs has basically nothing to do except serve up ioctls,
5905  * so most of the standard driver entry points are in zvol.c.
5906  */
5907 static struct cb_ops zfs_cb_ops = {
5908         zfsdev_open,    /* open */
5909         zfsdev_close,   /* close */
5910         zvol_strategy,  /* strategy */
5911         nodev,          /* print */
5912         zvol_dump,      /* dump */
5913         zvol_read,      /* read */
5914         zvol_write,     /* write */
5915         zfsdev_ioctl,   /* ioctl */
5916         nodev,          /* devmap */
5917         nodev,          /* mmap */
5918         nodev,          /* segmap */
5919         nochpoll,       /* poll */
5920         ddi_prop_op,    /* prop_op */
5921         NULL,           /* streamtab */
5922         D_NEW | D_MP | D_64BIT,         /* Driver compatibility flag */
5923         CB_REV,         /* version */
5924         nodev,          /* async read */
5925         nodev,          /* async write */
5926 };
5927 
5928 static struct dev_ops zfs_dev_ops = {
5929         DEVO_REV,       /* version */
5930         0,              /* refcnt */
5931         zfs_info,       /* info */
5932         nulldev,        /* identify */
5933         nulldev,        /* probe */
5934         zfs_attach,     /* attach */
5935         zfs_detach,     /* detach */
5936         nodev,          /* reset */
5937         &zfs_cb_ops,        /* driver operations */
5938         NULL,           /* no bus operations */
5939         NULL,           /* power */
5940         ddi_quiesce_not_needed, /* quiesce */
5941 };
5942 
5943 static struct modldrv zfs_modldrv = {
5944         &mod_driverops,
5945         "ZFS storage pool",
5946         &zfs_dev_ops
5947 };
5948 
5949 static struct modlinkage modlinkage = {
5950         MODREV_1,
5951         (void *)&zfs_modlfs,
5952         (void *)&zfs_modldrv,
5953         NULL
5954 };
5955 
5956 static void
5957 zfs_allow_log_destroy(void *arg)
5958 {
5959         char *poolname = arg;
5960         strfree(poolname);
5961 }
5962 
5963 int
5964 _init(void)
5965 {
5966         int error;
5967 
5968         spa_init(FREAD | FWRITE);
5969         zfs_init();
5970         zvol_init();
5971         zfs_ioctl_init();
5972 
5973         if ((error = mod_install(&modlinkage)) != 0) {
5974                 zvol_fini();
5975                 zfs_fini();
5976                 spa_fini();
5977                 return (error);
5978         }
5979 
5980         tsd_create(&zfs_fsyncer_key, NULL);
5981         tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
5982         tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
5983 
5984         error = ldi_ident_from_mod(&modlinkage, &zfs_li);
5985         ASSERT(error == 0);
5986         mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
5987 
5988         return (0);
5989 }
5990 
5991 int
5992 _fini(void)
5993 {
5994         int error;
5995 
5996         if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
5997                 return (EBUSY);
5998 
5999         if ((error = mod_remove(&modlinkage)) != 0)
6000                 return (error);
6001 
6002         zvol_fini();
6003         zfs_fini();
6004         spa_fini();
6005         if (zfs_nfsshare_inited)
6006                 (void) ddi_modclose(nfs_mod);
6007         if (zfs_smbshare_inited)
6008                 (void) ddi_modclose(smbsrv_mod);
6009         if (zfs_nfsshare_inited || zfs_smbshare_inited)
6010                 (void) ddi_modclose(sharefs_mod);
6011 
6012         tsd_destroy(&zfs_fsyncer_key);
6013         ldi_ident_release(zfs_li);
6014         zfs_li = NULL;
6015         mutex_destroy(&zfs_share_lock);
6016 
6017         return (error);
6018 }
6019 
6020 int
6021 _info(struct modinfo *modinfop)
6022 {
6023         return (mod_info(&modlinkage, modinfop));
6024 }