xref: /freebsd/sys/contrib/openzfs/module/zfs/zfs_ioctl.c (revision 963f5dc7a30624e95d72fb7f87b8892651164e46)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 
22 /*
23  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24  * Portions Copyright 2011 Martin Matuska
25  * Copyright 2015, OmniTI Computer Consulting, Inc. All rights reserved.
26  * Portions Copyright 2012 Pawel Jakub Dawidek <pawel@dawidek.net>
27  * Copyright (c) 2014, 2016 Joyent, Inc. All rights reserved.
28  * Copyright 2016 Nexenta Systems, Inc.  All rights reserved.
29  * Copyright (c) 2014, Joyent, Inc. All rights reserved.
30  * Copyright (c) 2011, 2020 by Delphix. All rights reserved.
31  * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
32  * Copyright (c) 2013 Steven Hartland. All rights reserved.
33  * Copyright (c) 2014 Integros [integros.com]
34  * Copyright 2016 Toomas Soome <tsoome@me.com>
35  * Copyright (c) 2016 Actifio, Inc. All rights reserved.
36  * Copyright (c) 2018, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
37  * Copyright 2017 RackTop Systems.
38  * Copyright (c) 2017 Open-E, Inc. All Rights Reserved.
39  * Copyright (c) 2019 Datto Inc.
40  * Copyright (c) 2019, 2020 by Christian Schwarz. All rights reserved.
41  * Copyright (c) 2019, 2021, Klara Inc.
42  * Copyright (c) 2019, Allan Jude
43  */
44 
45 /*
46  * ZFS ioctls.
47  *
48  * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
49  * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
50  *
51  * There are two ways that we handle ioctls: the legacy way where almost
52  * all of the logic is in the ioctl callback, and the new way where most
53  * of the marshalling is handled in the common entry point, zfsdev_ioctl().
54  *
55  * Non-legacy ioctls should be registered by calling
56  * zfs_ioctl_register() from zfs_ioctl_init().  The ioctl is invoked
57  * from userland by lzc_ioctl().
58  *
59  * The registration arguments are as follows:
60  *
61  * const char *name
62  *   The name of the ioctl.  This is used for history logging.  If the
63  *   ioctl returns successfully (the callback returns 0), and allow_log
64  *   is true, then a history log entry will be recorded with the input &
65  *   output nvlists.  The log entry can be printed with "zpool history -i".
66  *
67  * zfs_ioc_t ioc
68  *   The ioctl request number, which userland will pass to ioctl(2).
69  *   We want newer versions of libzfs and libzfs_core to run against
70  *   existing zfs kernel modules (i.e. a deferred reboot after an update).
71  *   Therefore the ioctl numbers cannot change from release to release.
72  *
73  * zfs_secpolicy_func_t *secpolicy
74  *   This function will be called before the zfs_ioc_func_t, to
75  *   determine if this operation is permitted.  It should return EPERM
76  *   on failure, and 0 on success.  Checks include determining if the
77  *   dataset is visible in this zone, and if the user has either all
78  *   zfs privileges in the zone (SYS_MOUNT), or has been granted permission
79  *   to do this operation on this dataset with "zfs allow".
80  *
81  * zfs_ioc_namecheck_t namecheck
82  *   This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
83  *   name, a dataset name, or nothing.  If the name is not well-formed,
84  *   the ioctl will fail and the callback will not be called.
85  *   Therefore, the callback can assume that the name is well-formed
86  *   (e.g. is null-terminated, doesn't have more than one '@' character,
87  *   doesn't have invalid characters).
88  *
89  * zfs_ioc_poolcheck_t pool_check
90  *   This specifies requirements on the pool state.  If the pool does
91  *   not meet them (is suspended or is readonly), the ioctl will fail
92  *   and the callback will not be called.  If any checks are specified
93  *   (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
94  *   Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
95  *   POOL_CHECK_READONLY).
96  *
97  * zfs_ioc_key_t *nvl_keys
98  *  The list of expected/allowable innvl input keys. This list is used
99  *  to validate the nvlist input to the ioctl.
100  *
101  * boolean_t smush_outnvlist
102  *   If smush_outnvlist is true, then the output is presumed to be a
103  *   list of errors, and it will be "smushed" down to fit into the
104  *   caller's buffer, by removing some entries and replacing them with a
105  *   single "N_MORE_ERRORS" entry indicating how many were removed.  See
106  *   nvlist_smush() for details.  If smush_outnvlist is false, and the
107  *   outnvlist does not fit into the userland-provided buffer, then the
108  *   ioctl will fail with ENOMEM.
109  *
110  * zfs_ioc_func_t *func
111  *   The callback function that will perform the operation.
112  *
113  *   The callback should return 0 on success, or an error number on
114  *   failure.  If the function fails, the userland ioctl will return -1,
115  *   and errno will be set to the callback's return value.  The callback
116  *   will be called with the following arguments:
117  *
118  *   const char *name
119  *     The name of the pool or dataset to operate on, from
120  *     zfs_cmd_t:zc_name.  The 'namecheck' argument specifies the
121  *     expected type (pool, dataset, or none).
122  *
123  *   nvlist_t *innvl
124  *     The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src.  Or
125  *     NULL if no input nvlist was provided.  Changes to this nvlist are
126  *     ignored.  If the input nvlist could not be deserialized, the
127  *     ioctl will fail and the callback will not be called.
128  *
129  *   nvlist_t *outnvl
130  *     The output nvlist, initially empty.  The callback can fill it in,
131  *     and it will be returned to userland by serializing it into
132  *     zfs_cmd_t:zc_nvlist_dst.  If it is non-empty, and serialization
133  *     fails (e.g. because the caller didn't supply a large enough
134  *     buffer), then the overall ioctl will fail.  See the
135  *     'smush_nvlist' argument above for additional behaviors.
136  *
137  *     There are two typical uses of the output nvlist:
138  *       - To return state, e.g. property values.  In this case,
139  *         smush_outnvlist should be false.  If the buffer was not large
140  *         enough, the caller will reallocate a larger buffer and try
141  *         the ioctl again.
142  *
143  *       - To return multiple errors from an ioctl which makes on-disk
144  *         changes.  In this case, smush_outnvlist should be true.
145  *         Ioctls which make on-disk modifications should generally not
146  *         use the outnvl if they succeed, because the caller can not
147  *         distinguish between the operation failing, and
148  *         deserialization failing.
149  *
150  * IOCTL Interface Errors
151  *
152  * The following ioctl input errors can be returned:
153  *   ZFS_ERR_IOC_CMD_UNAVAIL	the ioctl number is not supported by kernel
154  *   ZFS_ERR_IOC_ARG_UNAVAIL	an input argument is not supported by kernel
155  *   ZFS_ERR_IOC_ARG_REQUIRED	a required input argument is missing
156  *   ZFS_ERR_IOC_ARG_BADTYPE	an input argument has an invalid type
157  */
158 
159 #include <sys/types.h>
160 #include <sys/param.h>
161 #include <sys/errno.h>
162 #include <sys/uio_impl.h>
163 #include <sys/file.h>
164 #include <sys/kmem.h>
165 #include <sys/cmn_err.h>
166 #include <sys/stat.h>
167 #include <sys/zfs_ioctl.h>
168 #include <sys/zfs_quota.h>
169 #include <sys/zfs_vfsops.h>
170 #include <sys/zfs_znode.h>
171 #include <sys/zap.h>
172 #include <sys/spa.h>
173 #include <sys/spa_impl.h>
174 #include <sys/vdev.h>
175 #include <sys/vdev_impl.h>
176 #include <sys/dmu.h>
177 #include <sys/dsl_dir.h>
178 #include <sys/dsl_dataset.h>
179 #include <sys/dsl_prop.h>
180 #include <sys/dsl_deleg.h>
181 #include <sys/dmu_objset.h>
182 #include <sys/dmu_impl.h>
183 #include <sys/dmu_redact.h>
184 #include <sys/dmu_tx.h>
185 #include <sys/sunddi.h>
186 #include <sys/policy.h>
187 #include <sys/zone.h>
188 #include <sys/nvpair.h>
189 #include <sys/pathname.h>
190 #include <sys/fs/zfs.h>
191 #include <sys/zfs_ctldir.h>
192 #include <sys/zfs_dir.h>
193 #include <sys/zfs_onexit.h>
194 #include <sys/zvol.h>
195 #include <sys/dsl_scan.h>
196 #include <sys/fm/util.h>
197 #include <sys/dsl_crypt.h>
198 #include <sys/rrwlock.h>
199 #include <sys/zfs_file.h>
200 
201 #include <sys/dmu_recv.h>
202 #include <sys/dmu_send.h>
203 #include <sys/dmu_recv.h>
204 #include <sys/dsl_destroy.h>
205 #include <sys/dsl_bookmark.h>
206 #include <sys/dsl_userhold.h>
207 #include <sys/zfeature.h>
208 #include <sys/zcp.h>
209 #include <sys/zio_checksum.h>
210 #include <sys/vdev_removal.h>
211 #include <sys/vdev_impl.h>
212 #include <sys/vdev_initialize.h>
213 #include <sys/vdev_trim.h>
214 
215 #include "zfs_namecheck.h"
216 #include "zfs_prop.h"
217 #include "zfs_deleg.h"
218 #include "zfs_comutil.h"
219 
220 #include <sys/lua/lua.h>
221 #include <sys/lua/lauxlib.h>
222 #include <sys/zfs_ioctl_impl.h>
223 
224 kmutex_t zfsdev_state_lock;
225 zfsdev_state_t *zfsdev_state_list;
226 
227 /*
228  * Limit maximum nvlist size.  We don't want users passing in insane values
229  * for zc->zc_nvlist_src_size, since we will need to allocate that much memory.
230  * Defaults to 0=auto which is handled by platform code.
231  */
232 unsigned long zfs_max_nvlist_src_size = 0;
233 
234 /*
235  * When logging the output nvlist of an ioctl in the on-disk history, limit
236  * the logged size to this many bytes.  This must be less than DMU_MAX_ACCESS.
237  * This applies primarily to zfs_ioc_channel_program().
238  */
239 unsigned long zfs_history_output_max = 1024 * 1024;
240 
241 uint_t zfs_fsyncer_key;
242 uint_t zfs_allow_log_key;
243 
244 /* DATA_TYPE_ANY is used when zkey_type can vary. */
245 #define	DATA_TYPE_ANY	DATA_TYPE_UNKNOWN
246 
247 typedef struct zfs_ioc_vec {
248 	zfs_ioc_legacy_func_t	*zvec_legacy_func;
249 	zfs_ioc_func_t		*zvec_func;
250 	zfs_secpolicy_func_t	*zvec_secpolicy;
251 	zfs_ioc_namecheck_t	zvec_namecheck;
252 	boolean_t		zvec_allow_log;
253 	zfs_ioc_poolcheck_t	zvec_pool_check;
254 	boolean_t		zvec_smush_outnvlist;
255 	const char		*zvec_name;
256 	const zfs_ioc_key_t	*zvec_nvl_keys;
257 	size_t			zvec_nvl_key_count;
258 } zfs_ioc_vec_t;
259 
260 /* This array is indexed by zfs_userquota_prop_t */
261 static const char *userquota_perms[] = {
262 	ZFS_DELEG_PERM_USERUSED,
263 	ZFS_DELEG_PERM_USERQUOTA,
264 	ZFS_DELEG_PERM_GROUPUSED,
265 	ZFS_DELEG_PERM_GROUPQUOTA,
266 	ZFS_DELEG_PERM_USEROBJUSED,
267 	ZFS_DELEG_PERM_USEROBJQUOTA,
268 	ZFS_DELEG_PERM_GROUPOBJUSED,
269 	ZFS_DELEG_PERM_GROUPOBJQUOTA,
270 	ZFS_DELEG_PERM_PROJECTUSED,
271 	ZFS_DELEG_PERM_PROJECTQUOTA,
272 	ZFS_DELEG_PERM_PROJECTOBJUSED,
273 	ZFS_DELEG_PERM_PROJECTOBJQUOTA,
274 };
275 
276 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
277 static int zfs_ioc_id_quota_upgrade(zfs_cmd_t *zc);
278 static int zfs_check_settable(const char *name, nvpair_t *property,
279     cred_t *cr);
280 static int zfs_check_clearable(const char *dataset, nvlist_t *props,
281     nvlist_t **errors);
282 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
283     boolean_t *);
284 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
285 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
286 
287 static void
288 history_str_free(char *buf)
289 {
290 	kmem_free(buf, HIS_MAX_RECORD_LEN);
291 }
292 
293 static char *
294 history_str_get(zfs_cmd_t *zc)
295 {
296 	char *buf;
297 
298 	if (zc->zc_history == 0)
299 		return (NULL);
300 
301 	buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
302 	if (copyinstr((void *)(uintptr_t)zc->zc_history,
303 	    buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
304 		history_str_free(buf);
305 		return (NULL);
306 	}
307 
308 	buf[HIS_MAX_RECORD_LEN -1] = '\0';
309 
310 	return (buf);
311 }
312 
313 /*
314  * Return non-zero if the spa version is less than requested version.
315  */
316 static int
317 zfs_earlier_version(const char *name, int version)
318 {
319 	spa_t *spa;
320 
321 	if (spa_open(name, &spa, FTAG) == 0) {
322 		if (spa_version(spa) < version) {
323 			spa_close(spa, FTAG);
324 			return (1);
325 		}
326 		spa_close(spa, FTAG);
327 	}
328 	return (0);
329 }
330 
331 /*
332  * Return TRUE if the ZPL version is less than requested version.
333  */
334 static boolean_t
335 zpl_earlier_version(const char *name, int version)
336 {
337 	objset_t *os;
338 	boolean_t rc = B_TRUE;
339 
340 	if (dmu_objset_hold(name, FTAG, &os) == 0) {
341 		uint64_t zplversion;
342 
343 		if (dmu_objset_type(os) != DMU_OST_ZFS) {
344 			dmu_objset_rele(os, FTAG);
345 			return (B_TRUE);
346 		}
347 		/* XXX reading from non-owned objset */
348 		if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
349 			rc = zplversion < version;
350 		dmu_objset_rele(os, FTAG);
351 	}
352 	return (rc);
353 }
354 
355 static void
356 zfs_log_history(zfs_cmd_t *zc)
357 {
358 	spa_t *spa;
359 	char *buf;
360 
361 	if ((buf = history_str_get(zc)) == NULL)
362 		return;
363 
364 	if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
365 		if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
366 			(void) spa_history_log(spa, buf);
367 		spa_close(spa, FTAG);
368 	}
369 	history_str_free(buf);
370 }
371 
372 /*
373  * Policy for top-level read operations (list pools).  Requires no privileges,
374  * and can be used in the local zone, as there is no associated dataset.
375  */
376 /* ARGSUSED */
377 static int
378 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
379 {
380 	return (0);
381 }
382 
383 /*
384  * Policy for dataset read operations (list children, get statistics).  Requires
385  * no privileges, but must be visible in the local zone.
386  */
387 /* ARGSUSED */
388 static int
389 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
390 {
391 	if (INGLOBALZONE(curproc) ||
392 	    zone_dataset_visible(zc->zc_name, NULL))
393 		return (0);
394 
395 	return (SET_ERROR(ENOENT));
396 }
397 
398 static int
399 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
400 {
401 	int writable = 1;
402 
403 	/*
404 	 * The dataset must be visible by this zone -- check this first
405 	 * so they don't see EPERM on something they shouldn't know about.
406 	 */
407 	if (!INGLOBALZONE(curproc) &&
408 	    !zone_dataset_visible(dataset, &writable))
409 		return (SET_ERROR(ENOENT));
410 
411 	if (INGLOBALZONE(curproc)) {
412 		/*
413 		 * If the fs is zoned, only root can access it from the
414 		 * global zone.
415 		 */
416 		if (secpolicy_zfs(cr) && zoned)
417 			return (SET_ERROR(EPERM));
418 	} else {
419 		/*
420 		 * If we are in a local zone, the 'zoned' property must be set.
421 		 */
422 		if (!zoned)
423 			return (SET_ERROR(EPERM));
424 
425 		/* must be writable by this zone */
426 		if (!writable)
427 			return (SET_ERROR(EPERM));
428 	}
429 	return (0);
430 }
431 
432 static int
433 zfs_dozonecheck(const char *dataset, cred_t *cr)
434 {
435 	uint64_t zoned;
436 
437 	if (dsl_prop_get_integer(dataset, zfs_prop_to_name(ZFS_PROP_ZONED),
438 	    &zoned, NULL))
439 		return (SET_ERROR(ENOENT));
440 
441 	return (zfs_dozonecheck_impl(dataset, zoned, cr));
442 }
443 
444 static int
445 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
446 {
447 	uint64_t zoned;
448 
449 	if (dsl_prop_get_int_ds(ds, zfs_prop_to_name(ZFS_PROP_ZONED), &zoned))
450 		return (SET_ERROR(ENOENT));
451 
452 	return (zfs_dozonecheck_impl(dataset, zoned, cr));
453 }
454 
455 static int
456 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
457     const char *perm, cred_t *cr)
458 {
459 	int error;
460 
461 	error = zfs_dozonecheck_ds(name, ds, cr);
462 	if (error == 0) {
463 		error = secpolicy_zfs(cr);
464 		if (error != 0)
465 			error = dsl_deleg_access_impl(ds, perm, cr);
466 	}
467 	return (error);
468 }
469 
470 static int
471 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
472 {
473 	int error;
474 	dsl_dataset_t *ds;
475 	dsl_pool_t *dp;
476 
477 	/*
478 	 * First do a quick check for root in the global zone, which
479 	 * is allowed to do all write_perms.  This ensures that zfs_ioc_*
480 	 * will get to handle nonexistent datasets.
481 	 */
482 	if (INGLOBALZONE(curproc) && secpolicy_zfs(cr) == 0)
483 		return (0);
484 
485 	error = dsl_pool_hold(name, FTAG, &dp);
486 	if (error != 0)
487 		return (error);
488 
489 	error = dsl_dataset_hold(dp, name, FTAG, &ds);
490 	if (error != 0) {
491 		dsl_pool_rele(dp, FTAG);
492 		return (error);
493 	}
494 
495 	error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
496 
497 	dsl_dataset_rele(ds, FTAG);
498 	dsl_pool_rele(dp, FTAG);
499 	return (error);
500 }
501 
502 /*
503  * Policy for setting the security label property.
504  *
505  * Returns 0 for success, non-zero for access and other errors.
506  */
507 static int
508 zfs_set_slabel_policy(const char *name, const char *strval, cred_t *cr)
509 {
510 #ifdef HAVE_MLSLABEL
511 	char		ds_hexsl[MAXNAMELEN];
512 	bslabel_t	ds_sl, new_sl;
513 	boolean_t	new_default = FALSE;
514 	uint64_t	zoned;
515 	int		needed_priv = -1;
516 	int		error;
517 
518 	/* First get the existing dataset label. */
519 	error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
520 	    1, sizeof (ds_hexsl), &ds_hexsl, NULL);
521 	if (error != 0)
522 		return (SET_ERROR(EPERM));
523 
524 	if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
525 		new_default = TRUE;
526 
527 	/* The label must be translatable */
528 	if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
529 		return (SET_ERROR(EINVAL));
530 
531 	/*
532 	 * In a non-global zone, disallow attempts to set a label that
533 	 * doesn't match that of the zone; otherwise no other checks
534 	 * are needed.
535 	 */
536 	if (!INGLOBALZONE(curproc)) {
537 		if (new_default || !blequal(&new_sl, CR_SL(CRED())))
538 			return (SET_ERROR(EPERM));
539 		return (0);
540 	}
541 
542 	/*
543 	 * For global-zone datasets (i.e., those whose zoned property is
544 	 * "off", verify that the specified new label is valid for the
545 	 * global zone.
546 	 */
547 	if (dsl_prop_get_integer(name,
548 	    zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
549 		return (SET_ERROR(EPERM));
550 	if (!zoned) {
551 		if (zfs_check_global_label(name, strval) != 0)
552 			return (SET_ERROR(EPERM));
553 	}
554 
555 	/*
556 	 * If the existing dataset label is nondefault, check if the
557 	 * dataset is mounted (label cannot be changed while mounted).
558 	 * Get the zfsvfs_t; if there isn't one, then the dataset isn't
559 	 * mounted (or isn't a dataset, doesn't exist, ...).
560 	 */
561 	if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
562 		objset_t *os;
563 		static const char *setsl_tag = "setsl_tag";
564 
565 		/*
566 		 * Try to own the dataset; abort if there is any error,
567 		 * (e.g., already mounted, in use, or other error).
568 		 */
569 		error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE, B_TRUE,
570 		    setsl_tag, &os);
571 		if (error != 0)
572 			return (SET_ERROR(EPERM));
573 
574 		dmu_objset_disown(os, B_TRUE, setsl_tag);
575 
576 		if (new_default) {
577 			needed_priv = PRIV_FILE_DOWNGRADE_SL;
578 			goto out_check;
579 		}
580 
581 		if (hexstr_to_label(strval, &new_sl) != 0)
582 			return (SET_ERROR(EPERM));
583 
584 		if (blstrictdom(&ds_sl, &new_sl))
585 			needed_priv = PRIV_FILE_DOWNGRADE_SL;
586 		else if (blstrictdom(&new_sl, &ds_sl))
587 			needed_priv = PRIV_FILE_UPGRADE_SL;
588 	} else {
589 		/* dataset currently has a default label */
590 		if (!new_default)
591 			needed_priv = PRIV_FILE_UPGRADE_SL;
592 	}
593 
594 out_check:
595 	if (needed_priv != -1)
596 		return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
597 	return (0);
598 #else
599 	return (SET_ERROR(ENOTSUP));
600 #endif /* HAVE_MLSLABEL */
601 }
602 
603 static int
604 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
605     cred_t *cr)
606 {
607 	char *strval;
608 
609 	/*
610 	 * Check permissions for special properties.
611 	 */
612 	switch (prop) {
613 	default:
614 		break;
615 	case ZFS_PROP_ZONED:
616 		/*
617 		 * Disallow setting of 'zoned' from within a local zone.
618 		 */
619 		if (!INGLOBALZONE(curproc))
620 			return (SET_ERROR(EPERM));
621 		break;
622 
623 	case ZFS_PROP_QUOTA:
624 	case ZFS_PROP_FILESYSTEM_LIMIT:
625 	case ZFS_PROP_SNAPSHOT_LIMIT:
626 		if (!INGLOBALZONE(curproc)) {
627 			uint64_t zoned;
628 			char setpoint[ZFS_MAX_DATASET_NAME_LEN];
629 			/*
630 			 * Unprivileged users are allowed to modify the
631 			 * limit on things *under* (ie. contained by)
632 			 * the thing they own.
633 			 */
634 			if (dsl_prop_get_integer(dsname,
635 			    zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, setpoint))
636 				return (SET_ERROR(EPERM));
637 			if (!zoned || strlen(dsname) <= strlen(setpoint))
638 				return (SET_ERROR(EPERM));
639 		}
640 		break;
641 
642 	case ZFS_PROP_MLSLABEL:
643 		if (!is_system_labeled())
644 			return (SET_ERROR(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 != 0)
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 	dsl_pool_t *dp;
689 	dsl_dataset_t *ds;
690 	const char *cp;
691 	int error;
692 
693 	/*
694 	 * Generate the current snapshot name from the given objsetid, then
695 	 * use that name for the secpolicy/zone checks.
696 	 */
697 	cp = strchr(zc->zc_name, '@');
698 	if (cp == NULL)
699 		return (SET_ERROR(EINVAL));
700 	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
701 	if (error != 0)
702 		return (error);
703 
704 	error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
705 	if (error != 0) {
706 		dsl_pool_rele(dp, FTAG);
707 		return (error);
708 	}
709 
710 	dsl_dataset_name(ds, zc->zc_name);
711 
712 	error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
713 	    ZFS_DELEG_PERM_SEND, cr);
714 	dsl_dataset_rele(ds, FTAG);
715 	dsl_pool_rele(dp, FTAG);
716 
717 	return (error);
718 }
719 
720 /* ARGSUSED */
721 static int
722 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
723 {
724 	return (zfs_secpolicy_write_perms(zc->zc_name,
725 	    ZFS_DELEG_PERM_SEND, cr));
726 }
727 
728 static int
729 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
730 {
731 	return (SET_ERROR(ENOTSUP));
732 }
733 
734 static int
735 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
736 {
737 	return (SET_ERROR(ENOTSUP));
738 }
739 
740 static int
741 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
742 {
743 	char *cp;
744 
745 	/*
746 	 * Remove the @bla or /bla from the end of the name to get the parent.
747 	 */
748 	(void) strncpy(parent, datasetname, parentsize);
749 	cp = strrchr(parent, '@');
750 	if (cp != NULL) {
751 		cp[0] = '\0';
752 	} else {
753 		cp = strrchr(parent, '/');
754 		if (cp == NULL)
755 			return (SET_ERROR(ENOENT));
756 		cp[0] = '\0';
757 	}
758 
759 	return (0);
760 }
761 
762 int
763 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
764 {
765 	int error;
766 
767 	if ((error = zfs_secpolicy_write_perms(name,
768 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
769 		return (error);
770 
771 	return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
772 }
773 
774 /* ARGSUSED */
775 static int
776 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
777 {
778 	return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
779 }
780 
781 /*
782  * Destroying snapshots with delegated permissions requires
783  * descendant mount and destroy permissions.
784  */
785 /* ARGSUSED */
786 static int
787 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
788 {
789 	nvlist_t *snaps;
790 	nvpair_t *pair, *nextpair;
791 	int error = 0;
792 
793 	snaps = fnvlist_lookup_nvlist(innvl, "snaps");
794 
795 	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
796 	    pair = nextpair) {
797 		nextpair = nvlist_next_nvpair(snaps, pair);
798 		error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
799 		if (error == ENOENT) {
800 			/*
801 			 * Ignore any snapshots that don't exist (we consider
802 			 * them "already destroyed").  Remove the name from the
803 			 * nvl here in case the snapshot is created between
804 			 * now and when we try to destroy it (in which case
805 			 * we don't want to destroy it since we haven't
806 			 * checked for permission).
807 			 */
808 			fnvlist_remove_nvpair(snaps, pair);
809 			error = 0;
810 		}
811 		if (error != 0)
812 			break;
813 	}
814 
815 	return (error);
816 }
817 
818 int
819 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
820 {
821 	char	parentname[ZFS_MAX_DATASET_NAME_LEN];
822 	int	error;
823 
824 	if ((error = zfs_secpolicy_write_perms(from,
825 	    ZFS_DELEG_PERM_RENAME, cr)) != 0)
826 		return (error);
827 
828 	if ((error = zfs_secpolicy_write_perms(from,
829 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
830 		return (error);
831 
832 	if ((error = zfs_get_parent(to, parentname,
833 	    sizeof (parentname))) != 0)
834 		return (error);
835 
836 	if ((error = zfs_secpolicy_write_perms(parentname,
837 	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
838 		return (error);
839 
840 	if ((error = zfs_secpolicy_write_perms(parentname,
841 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
842 		return (error);
843 
844 	return (error);
845 }
846 
847 /* ARGSUSED */
848 static int
849 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
850 {
851 	return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
852 }
853 
854 /* ARGSUSED */
855 static int
856 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
857 {
858 	dsl_pool_t *dp;
859 	dsl_dataset_t *clone;
860 	int error;
861 
862 	error = zfs_secpolicy_write_perms(zc->zc_name,
863 	    ZFS_DELEG_PERM_PROMOTE, cr);
864 	if (error != 0)
865 		return (error);
866 
867 	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
868 	if (error != 0)
869 		return (error);
870 
871 	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
872 
873 	if (error == 0) {
874 		char parentname[ZFS_MAX_DATASET_NAME_LEN];
875 		dsl_dataset_t *origin = NULL;
876 		dsl_dir_t *dd;
877 		dd = clone->ds_dir;
878 
879 		error = dsl_dataset_hold_obj(dd->dd_pool,
880 		    dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin);
881 		if (error != 0) {
882 			dsl_dataset_rele(clone, FTAG);
883 			dsl_pool_rele(dp, FTAG);
884 			return (error);
885 		}
886 
887 		error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
888 		    ZFS_DELEG_PERM_MOUNT, cr);
889 
890 		dsl_dataset_name(origin, parentname);
891 		if (error == 0) {
892 			error = zfs_secpolicy_write_perms_ds(parentname, origin,
893 			    ZFS_DELEG_PERM_PROMOTE, cr);
894 		}
895 		dsl_dataset_rele(clone, FTAG);
896 		dsl_dataset_rele(origin, FTAG);
897 	}
898 	dsl_pool_rele(dp, FTAG);
899 	return (error);
900 }
901 
902 /* ARGSUSED */
903 static int
904 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
905 {
906 	int error;
907 
908 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
909 	    ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
910 		return (error);
911 
912 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
913 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
914 		return (error);
915 
916 	return (zfs_secpolicy_write_perms(zc->zc_name,
917 	    ZFS_DELEG_PERM_CREATE, cr));
918 }
919 
920 /* ARGSUSED */
921 static int
922 zfs_secpolicy_recv_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
923 {
924 	return (zfs_secpolicy_recv(zc, innvl, cr));
925 }
926 
927 int
928 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
929 {
930 	return (zfs_secpolicy_write_perms(name,
931 	    ZFS_DELEG_PERM_SNAPSHOT, cr));
932 }
933 
934 /*
935  * Check for permission to create each snapshot in the nvlist.
936  */
937 /* ARGSUSED */
938 static int
939 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
940 {
941 	nvlist_t *snaps;
942 	int error = 0;
943 	nvpair_t *pair;
944 
945 	snaps = fnvlist_lookup_nvlist(innvl, "snaps");
946 
947 	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
948 	    pair = nvlist_next_nvpair(snaps, pair)) {
949 		char *name = nvpair_name(pair);
950 		char *atp = strchr(name, '@');
951 
952 		if (atp == NULL) {
953 			error = SET_ERROR(EINVAL);
954 			break;
955 		}
956 		*atp = '\0';
957 		error = zfs_secpolicy_snapshot_perms(name, cr);
958 		*atp = '@';
959 		if (error != 0)
960 			break;
961 	}
962 	return (error);
963 }
964 
965 /*
966  * Check for permission to create each bookmark in the nvlist.
967  */
968 /* ARGSUSED */
969 static int
970 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
971 {
972 	int error = 0;
973 
974 	for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
975 	    pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
976 		char *name = nvpair_name(pair);
977 		char *hashp = strchr(name, '#');
978 
979 		if (hashp == NULL) {
980 			error = SET_ERROR(EINVAL);
981 			break;
982 		}
983 		*hashp = '\0';
984 		error = zfs_secpolicy_write_perms(name,
985 		    ZFS_DELEG_PERM_BOOKMARK, cr);
986 		*hashp = '#';
987 		if (error != 0)
988 			break;
989 	}
990 	return (error);
991 }
992 
993 /* ARGSUSED */
994 static int
995 zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
996 {
997 	nvpair_t *pair, *nextpair;
998 	int error = 0;
999 
1000 	for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1001 	    pair = nextpair) {
1002 		char *name = nvpair_name(pair);
1003 		char *hashp = strchr(name, '#');
1004 		nextpair = nvlist_next_nvpair(innvl, pair);
1005 
1006 		if (hashp == NULL) {
1007 			error = SET_ERROR(EINVAL);
1008 			break;
1009 		}
1010 
1011 		*hashp = '\0';
1012 		error = zfs_secpolicy_write_perms(name,
1013 		    ZFS_DELEG_PERM_DESTROY, cr);
1014 		*hashp = '#';
1015 		if (error == ENOENT) {
1016 			/*
1017 			 * Ignore any filesystems that don't exist (we consider
1018 			 * their bookmarks "already destroyed").  Remove
1019 			 * the name from the nvl here in case the filesystem
1020 			 * is created between now and when we try to destroy
1021 			 * the bookmark (in which case we don't want to
1022 			 * destroy it since we haven't checked for permission).
1023 			 */
1024 			fnvlist_remove_nvpair(innvl, pair);
1025 			error = 0;
1026 		}
1027 		if (error != 0)
1028 			break;
1029 	}
1030 
1031 	return (error);
1032 }
1033 
1034 /* ARGSUSED */
1035 static int
1036 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1037 {
1038 	/*
1039 	 * Even root must have a proper TSD so that we know what pool
1040 	 * to log to.
1041 	 */
1042 	if (tsd_get(zfs_allow_log_key) == NULL)
1043 		return (SET_ERROR(EPERM));
1044 	return (0);
1045 }
1046 
1047 static int
1048 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1049 {
1050 	char	parentname[ZFS_MAX_DATASET_NAME_LEN];
1051 	int	error;
1052 	char	*origin;
1053 
1054 	if ((error = zfs_get_parent(zc->zc_name, parentname,
1055 	    sizeof (parentname))) != 0)
1056 		return (error);
1057 
1058 	if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1059 	    (error = zfs_secpolicy_write_perms(origin,
1060 	    ZFS_DELEG_PERM_CLONE, cr)) != 0)
1061 		return (error);
1062 
1063 	if ((error = zfs_secpolicy_write_perms(parentname,
1064 	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
1065 		return (error);
1066 
1067 	return (zfs_secpolicy_write_perms(parentname,
1068 	    ZFS_DELEG_PERM_MOUNT, cr));
1069 }
1070 
1071 /*
1072  * Policy for pool operations - create/destroy pools, add vdevs, etc.  Requires
1073  * SYS_CONFIG privilege, which is not available in a local zone.
1074  */
1075 /* ARGSUSED */
1076 int
1077 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1078 {
1079 	if (secpolicy_sys_config(cr, B_FALSE) != 0)
1080 		return (SET_ERROR(EPERM));
1081 
1082 	return (0);
1083 }
1084 
1085 /*
1086  * Policy for object to name lookups.
1087  */
1088 /* ARGSUSED */
1089 static int
1090 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1091 {
1092 	int error;
1093 
1094 	if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1095 		return (0);
1096 
1097 	error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1098 	return (error);
1099 }
1100 
1101 /*
1102  * Policy for fault injection.  Requires all privileges.
1103  */
1104 /* ARGSUSED */
1105 static int
1106 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1107 {
1108 	return (secpolicy_zinject(cr));
1109 }
1110 
1111 /* ARGSUSED */
1112 static int
1113 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1114 {
1115 	zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1116 
1117 	if (prop == ZPROP_INVAL) {
1118 		if (!zfs_prop_user(zc->zc_value))
1119 			return (SET_ERROR(EINVAL));
1120 		return (zfs_secpolicy_write_perms(zc->zc_name,
1121 		    ZFS_DELEG_PERM_USERPROP, cr));
1122 	} else {
1123 		return (zfs_secpolicy_setprop(zc->zc_name, prop,
1124 		    NULL, cr));
1125 	}
1126 }
1127 
1128 static int
1129 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1130 {
1131 	int err = zfs_secpolicy_read(zc, innvl, cr);
1132 	if (err)
1133 		return (err);
1134 
1135 	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1136 		return (SET_ERROR(EINVAL));
1137 
1138 	if (zc->zc_value[0] == 0) {
1139 		/*
1140 		 * They are asking about a posix uid/gid.  If it's
1141 		 * themself, allow it.
1142 		 */
1143 		if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1144 		    zc->zc_objset_type == ZFS_PROP_USERQUOTA ||
1145 		    zc->zc_objset_type == ZFS_PROP_USEROBJUSED ||
1146 		    zc->zc_objset_type == ZFS_PROP_USEROBJQUOTA) {
1147 			if (zc->zc_guid == crgetuid(cr))
1148 				return (0);
1149 		} else if (zc->zc_objset_type == ZFS_PROP_GROUPUSED ||
1150 		    zc->zc_objset_type == ZFS_PROP_GROUPQUOTA ||
1151 		    zc->zc_objset_type == ZFS_PROP_GROUPOBJUSED ||
1152 		    zc->zc_objset_type == ZFS_PROP_GROUPOBJQUOTA) {
1153 			if (groupmember(zc->zc_guid, cr))
1154 				return (0);
1155 		}
1156 		/* else is for project quota/used */
1157 	}
1158 
1159 	return (zfs_secpolicy_write_perms(zc->zc_name,
1160 	    userquota_perms[zc->zc_objset_type], cr));
1161 }
1162 
1163 static int
1164 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1165 {
1166 	int err = zfs_secpolicy_read(zc, innvl, cr);
1167 	if (err)
1168 		return (err);
1169 
1170 	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1171 		return (SET_ERROR(EINVAL));
1172 
1173 	return (zfs_secpolicy_write_perms(zc->zc_name,
1174 	    userquota_perms[zc->zc_objset_type], cr));
1175 }
1176 
1177 /* ARGSUSED */
1178 static int
1179 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1180 {
1181 	return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1182 	    NULL, cr));
1183 }
1184 
1185 /* ARGSUSED */
1186 static int
1187 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1188 {
1189 	nvpair_t *pair;
1190 	nvlist_t *holds;
1191 	int error;
1192 
1193 	holds = fnvlist_lookup_nvlist(innvl, "holds");
1194 
1195 	for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1196 	    pair = nvlist_next_nvpair(holds, pair)) {
1197 		char fsname[ZFS_MAX_DATASET_NAME_LEN];
1198 		error = dmu_fsname(nvpair_name(pair), fsname);
1199 		if (error != 0)
1200 			return (error);
1201 		error = zfs_secpolicy_write_perms(fsname,
1202 		    ZFS_DELEG_PERM_HOLD, cr);
1203 		if (error != 0)
1204 			return (error);
1205 	}
1206 	return (0);
1207 }
1208 
1209 /* ARGSUSED */
1210 static int
1211 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1212 {
1213 	nvpair_t *pair;
1214 	int error;
1215 
1216 	for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1217 	    pair = nvlist_next_nvpair(innvl, pair)) {
1218 		char fsname[ZFS_MAX_DATASET_NAME_LEN];
1219 		error = dmu_fsname(nvpair_name(pair), fsname);
1220 		if (error != 0)
1221 			return (error);
1222 		error = zfs_secpolicy_write_perms(fsname,
1223 		    ZFS_DELEG_PERM_RELEASE, cr);
1224 		if (error != 0)
1225 			return (error);
1226 	}
1227 	return (0);
1228 }
1229 
1230 /*
1231  * Policy for allowing temporary snapshots to be taken or released
1232  */
1233 static int
1234 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1235 {
1236 	/*
1237 	 * A temporary snapshot is the same as a snapshot,
1238 	 * hold, destroy and release all rolled into one.
1239 	 * Delegated diff alone is sufficient that we allow this.
1240 	 */
1241 	int error;
1242 
1243 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1244 	    ZFS_DELEG_PERM_DIFF, cr)) == 0)
1245 		return (0);
1246 
1247 	error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1248 
1249 	if (innvl != NULL) {
1250 		if (error == 0)
1251 			error = zfs_secpolicy_hold(zc, innvl, cr);
1252 		if (error == 0)
1253 			error = zfs_secpolicy_release(zc, innvl, cr);
1254 		if (error == 0)
1255 			error = zfs_secpolicy_destroy(zc, innvl, cr);
1256 	}
1257 	return (error);
1258 }
1259 
1260 static int
1261 zfs_secpolicy_load_key(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1262 {
1263 	return (zfs_secpolicy_write_perms(zc->zc_name,
1264 	    ZFS_DELEG_PERM_LOAD_KEY, cr));
1265 }
1266 
1267 static int
1268 zfs_secpolicy_change_key(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1269 {
1270 	return (zfs_secpolicy_write_perms(zc->zc_name,
1271 	    ZFS_DELEG_PERM_CHANGE_KEY, cr));
1272 }
1273 
1274 /*
1275  * Returns the nvlist as specified by the user in the zfs_cmd_t.
1276  */
1277 static int
1278 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1279 {
1280 	char *packed;
1281 	int error;
1282 	nvlist_t *list = NULL;
1283 
1284 	/*
1285 	 * Read in and unpack the user-supplied nvlist.
1286 	 */
1287 	if (size == 0)
1288 		return (SET_ERROR(EINVAL));
1289 
1290 	packed = vmem_alloc(size, KM_SLEEP);
1291 
1292 	if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1293 	    iflag)) != 0) {
1294 		vmem_free(packed, size);
1295 		return (SET_ERROR(EFAULT));
1296 	}
1297 
1298 	if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1299 		vmem_free(packed, size);
1300 		return (error);
1301 	}
1302 
1303 	vmem_free(packed, size);
1304 
1305 	*nvp = list;
1306 	return (0);
1307 }
1308 
1309 /*
1310  * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1311  * Entries will be removed from the end of the nvlist, and one int32 entry
1312  * named "N_MORE_ERRORS" will be added indicating how many entries were
1313  * removed.
1314  */
1315 static int
1316 nvlist_smush(nvlist_t *errors, size_t max)
1317 {
1318 	size_t size;
1319 
1320 	size = fnvlist_size(errors);
1321 
1322 	if (size > max) {
1323 		nvpair_t *more_errors;
1324 		int n = 0;
1325 
1326 		if (max < 1024)
1327 			return (SET_ERROR(ENOMEM));
1328 
1329 		fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1330 		more_errors = nvlist_prev_nvpair(errors, NULL);
1331 
1332 		do {
1333 			nvpair_t *pair = nvlist_prev_nvpair(errors,
1334 			    more_errors);
1335 			fnvlist_remove_nvpair(errors, pair);
1336 			n++;
1337 			size = fnvlist_size(errors);
1338 		} while (size > max);
1339 
1340 		fnvlist_remove_nvpair(errors, more_errors);
1341 		fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1342 		ASSERT3U(fnvlist_size(errors), <=, max);
1343 	}
1344 
1345 	return (0);
1346 }
1347 
1348 static int
1349 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1350 {
1351 	char *packed = NULL;
1352 	int error = 0;
1353 	size_t size;
1354 
1355 	size = fnvlist_size(nvl);
1356 
1357 	if (size > zc->zc_nvlist_dst_size) {
1358 		error = SET_ERROR(ENOMEM);
1359 	} else {
1360 		packed = fnvlist_pack(nvl, &size);
1361 		if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1362 		    size, zc->zc_iflags) != 0)
1363 			error = SET_ERROR(EFAULT);
1364 		fnvlist_pack_free(packed, size);
1365 	}
1366 
1367 	zc->zc_nvlist_dst_size = size;
1368 	zc->zc_nvlist_dst_filled = B_TRUE;
1369 	return (error);
1370 }
1371 
1372 int
1373 getzfsvfs_impl(objset_t *os, zfsvfs_t **zfvp)
1374 {
1375 	int error = 0;
1376 	if (dmu_objset_type(os) != DMU_OST_ZFS) {
1377 		return (SET_ERROR(EINVAL));
1378 	}
1379 
1380 	mutex_enter(&os->os_user_ptr_lock);
1381 	*zfvp = dmu_objset_get_user(os);
1382 	/* bump s_active only when non-zero to prevent umount race */
1383 	error = zfs_vfs_ref(zfvp);
1384 	mutex_exit(&os->os_user_ptr_lock);
1385 	return (error);
1386 }
1387 
1388 int
1389 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1390 {
1391 	objset_t *os;
1392 	int error;
1393 
1394 	error = dmu_objset_hold(dsname, FTAG, &os);
1395 	if (error != 0)
1396 		return (error);
1397 
1398 	error = getzfsvfs_impl(os, zfvp);
1399 	dmu_objset_rele(os, FTAG);
1400 	return (error);
1401 }
1402 
1403 /*
1404  * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1405  * case its z_sb will be NULL, and it will be opened as the owner.
1406  * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1407  * which prevents all inode ops from running.
1408  */
1409 static int
1410 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1411 {
1412 	int error = 0;
1413 
1414 	if (getzfsvfs(name, zfvp) != 0)
1415 		error = zfsvfs_create(name, B_FALSE, zfvp);
1416 	if (error == 0) {
1417 		if (writer)
1418 			ZFS_TEARDOWN_ENTER_WRITE(*zfvp, tag);
1419 		else
1420 			ZFS_TEARDOWN_ENTER_READ(*zfvp, tag);
1421 		if ((*zfvp)->z_unmounted) {
1422 			/*
1423 			 * XXX we could probably try again, since the unmounting
1424 			 * thread should be just about to disassociate the
1425 			 * objset from the zfsvfs.
1426 			 */
1427 			ZFS_TEARDOWN_EXIT(*zfvp, tag);
1428 			return (SET_ERROR(EBUSY));
1429 		}
1430 	}
1431 	return (error);
1432 }
1433 
1434 static void
1435 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1436 {
1437 	ZFS_TEARDOWN_EXIT(zfsvfs, tag);
1438 
1439 	if (zfs_vfs_held(zfsvfs)) {
1440 		zfs_vfs_rele(zfsvfs);
1441 	} else {
1442 		dmu_objset_disown(zfsvfs->z_os, B_TRUE, zfsvfs);
1443 		zfsvfs_free(zfsvfs);
1444 	}
1445 }
1446 
1447 static int
1448 zfs_ioc_pool_create(zfs_cmd_t *zc)
1449 {
1450 	int error;
1451 	nvlist_t *config, *props = NULL;
1452 	nvlist_t *rootprops = NULL;
1453 	nvlist_t *zplprops = NULL;
1454 	dsl_crypto_params_t *dcp = NULL;
1455 	const char *spa_name = zc->zc_name;
1456 	boolean_t unload_wkey = B_TRUE;
1457 
1458 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1459 	    zc->zc_iflags, &config)))
1460 		return (error);
1461 
1462 	if (zc->zc_nvlist_src_size != 0 && (error =
1463 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1464 	    zc->zc_iflags, &props))) {
1465 		nvlist_free(config);
1466 		return (error);
1467 	}
1468 
1469 	if (props) {
1470 		nvlist_t *nvl = NULL;
1471 		nvlist_t *hidden_args = NULL;
1472 		uint64_t version = SPA_VERSION;
1473 		char *tname;
1474 
1475 		(void) nvlist_lookup_uint64(props,
1476 		    zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1477 		if (!SPA_VERSION_IS_SUPPORTED(version)) {
1478 			error = SET_ERROR(EINVAL);
1479 			goto pool_props_bad;
1480 		}
1481 		(void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1482 		if (nvl) {
1483 			error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1484 			if (error != 0)
1485 				goto pool_props_bad;
1486 			(void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1487 		}
1488 
1489 		(void) nvlist_lookup_nvlist(props, ZPOOL_HIDDEN_ARGS,
1490 		    &hidden_args);
1491 		error = dsl_crypto_params_create_nvlist(DCP_CMD_NONE,
1492 		    rootprops, hidden_args, &dcp);
1493 		if (error != 0)
1494 			goto pool_props_bad;
1495 		(void) nvlist_remove_all(props, ZPOOL_HIDDEN_ARGS);
1496 
1497 		VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1498 		error = zfs_fill_zplprops_root(version, rootprops,
1499 		    zplprops, NULL);
1500 		if (error != 0)
1501 			goto pool_props_bad;
1502 
1503 		if (nvlist_lookup_string(props,
1504 		    zpool_prop_to_name(ZPOOL_PROP_TNAME), &tname) == 0)
1505 			spa_name = tname;
1506 	}
1507 
1508 	error = spa_create(zc->zc_name, config, props, zplprops, dcp);
1509 
1510 	/*
1511 	 * Set the remaining root properties
1512 	 */
1513 	if (!error && (error = zfs_set_prop_nvlist(spa_name,
1514 	    ZPROP_SRC_LOCAL, rootprops, NULL)) != 0) {
1515 		(void) spa_destroy(spa_name);
1516 		unload_wkey = B_FALSE; /* spa_destroy() unloads wrapping keys */
1517 	}
1518 
1519 pool_props_bad:
1520 	nvlist_free(rootprops);
1521 	nvlist_free(zplprops);
1522 	nvlist_free(config);
1523 	nvlist_free(props);
1524 	dsl_crypto_params_free(dcp, unload_wkey && !!error);
1525 
1526 	return (error);
1527 }
1528 
1529 static int
1530 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1531 {
1532 	int error;
1533 	zfs_log_history(zc);
1534 	error = spa_destroy(zc->zc_name);
1535 
1536 	return (error);
1537 }
1538 
1539 static int
1540 zfs_ioc_pool_import(zfs_cmd_t *zc)
1541 {
1542 	nvlist_t *config, *props = NULL;
1543 	uint64_t guid;
1544 	int error;
1545 
1546 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1547 	    zc->zc_iflags, &config)) != 0)
1548 		return (error);
1549 
1550 	if (zc->zc_nvlist_src_size != 0 && (error =
1551 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1552 	    zc->zc_iflags, &props))) {
1553 		nvlist_free(config);
1554 		return (error);
1555 	}
1556 
1557 	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1558 	    guid != zc->zc_guid)
1559 		error = SET_ERROR(EINVAL);
1560 	else
1561 		error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1562 
1563 	if (zc->zc_nvlist_dst != 0) {
1564 		int err;
1565 
1566 		if ((err = put_nvlist(zc, config)) != 0)
1567 			error = err;
1568 	}
1569 
1570 	nvlist_free(config);
1571 	nvlist_free(props);
1572 
1573 	return (error);
1574 }
1575 
1576 static int
1577 zfs_ioc_pool_export(zfs_cmd_t *zc)
1578 {
1579 	int error;
1580 	boolean_t force = (boolean_t)zc->zc_cookie;
1581 	boolean_t hardforce = (boolean_t)zc->zc_guid;
1582 
1583 	zfs_log_history(zc);
1584 	error = spa_export(zc->zc_name, NULL, force, hardforce);
1585 
1586 	return (error);
1587 }
1588 
1589 static int
1590 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1591 {
1592 	nvlist_t *configs;
1593 	int error;
1594 
1595 	if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1596 		return (SET_ERROR(EEXIST));
1597 
1598 	error = put_nvlist(zc, configs);
1599 
1600 	nvlist_free(configs);
1601 
1602 	return (error);
1603 }
1604 
1605 /*
1606  * inputs:
1607  * zc_name		name of the pool
1608  *
1609  * outputs:
1610  * zc_cookie		real errno
1611  * zc_nvlist_dst	config nvlist
1612  * zc_nvlist_dst_size	size of config nvlist
1613  */
1614 static int
1615 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1616 {
1617 	nvlist_t *config;
1618 	int error;
1619 	int ret = 0;
1620 
1621 	error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1622 	    sizeof (zc->zc_value));
1623 
1624 	if (config != NULL) {
1625 		ret = put_nvlist(zc, config);
1626 		nvlist_free(config);
1627 
1628 		/*
1629 		 * The config may be present even if 'error' is non-zero.
1630 		 * In this case we return success, and preserve the real errno
1631 		 * in 'zc_cookie'.
1632 		 */
1633 		zc->zc_cookie = error;
1634 	} else {
1635 		ret = error;
1636 	}
1637 
1638 	return (ret);
1639 }
1640 
1641 /*
1642  * Try to import the given pool, returning pool stats as appropriate so that
1643  * user land knows which devices are available and overall pool health.
1644  */
1645 static int
1646 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1647 {
1648 	nvlist_t *tryconfig, *config = NULL;
1649 	int error;
1650 
1651 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1652 	    zc->zc_iflags, &tryconfig)) != 0)
1653 		return (error);
1654 
1655 	config = spa_tryimport(tryconfig);
1656 
1657 	nvlist_free(tryconfig);
1658 
1659 	if (config == NULL)
1660 		return (SET_ERROR(EINVAL));
1661 
1662 	error = put_nvlist(zc, config);
1663 	nvlist_free(config);
1664 
1665 	return (error);
1666 }
1667 
1668 /*
1669  * inputs:
1670  * zc_name              name of the pool
1671  * zc_cookie            scan func (pool_scan_func_t)
1672  * zc_flags             scrub pause/resume flag (pool_scrub_cmd_t)
1673  */
1674 static int
1675 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1676 {
1677 	spa_t *spa;
1678 	int error;
1679 
1680 	if (zc->zc_flags >= POOL_SCRUB_FLAGS_END)
1681 		return (SET_ERROR(EINVAL));
1682 
1683 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1684 		return (error);
1685 
1686 	if (zc->zc_flags == POOL_SCRUB_PAUSE)
1687 		error = spa_scrub_pause_resume(spa, POOL_SCRUB_PAUSE);
1688 	else if (zc->zc_cookie == POOL_SCAN_NONE)
1689 		error = spa_scan_stop(spa);
1690 	else
1691 		error = spa_scan(spa, zc->zc_cookie);
1692 
1693 	spa_close(spa, FTAG);
1694 
1695 	return (error);
1696 }
1697 
1698 static int
1699 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1700 {
1701 	spa_t *spa;
1702 	int error;
1703 
1704 	error = spa_open(zc->zc_name, &spa, FTAG);
1705 	if (error == 0) {
1706 		spa_freeze(spa);
1707 		spa_close(spa, FTAG);
1708 	}
1709 	return (error);
1710 }
1711 
1712 static int
1713 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1714 {
1715 	spa_t *spa;
1716 	int error;
1717 
1718 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1719 		return (error);
1720 
1721 	if (zc->zc_cookie < spa_version(spa) ||
1722 	    !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1723 		spa_close(spa, FTAG);
1724 		return (SET_ERROR(EINVAL));
1725 	}
1726 
1727 	spa_upgrade(spa, zc->zc_cookie);
1728 	spa_close(spa, FTAG);
1729 
1730 	return (error);
1731 }
1732 
1733 static int
1734 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1735 {
1736 	spa_t *spa;
1737 	char *hist_buf;
1738 	uint64_t size;
1739 	int error;
1740 
1741 	if ((size = zc->zc_history_len) == 0)
1742 		return (SET_ERROR(EINVAL));
1743 
1744 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1745 		return (error);
1746 
1747 	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1748 		spa_close(spa, FTAG);
1749 		return (SET_ERROR(ENOTSUP));
1750 	}
1751 
1752 	hist_buf = vmem_alloc(size, KM_SLEEP);
1753 	if ((error = spa_history_get(spa, &zc->zc_history_offset,
1754 	    &zc->zc_history_len, hist_buf)) == 0) {
1755 		error = ddi_copyout(hist_buf,
1756 		    (void *)(uintptr_t)zc->zc_history,
1757 		    zc->zc_history_len, zc->zc_iflags);
1758 	}
1759 
1760 	spa_close(spa, FTAG);
1761 	vmem_free(hist_buf, size);
1762 	return (error);
1763 }
1764 
1765 static int
1766 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1767 {
1768 	spa_t *spa;
1769 	int error;
1770 
1771 	error = spa_open(zc->zc_name, &spa, FTAG);
1772 	if (error == 0) {
1773 		error = spa_change_guid(spa);
1774 		spa_close(spa, FTAG);
1775 	}
1776 	return (error);
1777 }
1778 
1779 static int
1780 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1781 {
1782 	return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1783 }
1784 
1785 /*
1786  * inputs:
1787  * zc_name		name of filesystem
1788  * zc_obj		object to find
1789  *
1790  * outputs:
1791  * zc_value		name of object
1792  */
1793 static int
1794 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1795 {
1796 	objset_t *os;
1797 	int error;
1798 
1799 	/* XXX reading from objset not owned */
1800 	if ((error = dmu_objset_hold_flags(zc->zc_name, B_TRUE,
1801 	    FTAG, &os)) != 0)
1802 		return (error);
1803 	if (dmu_objset_type(os) != DMU_OST_ZFS) {
1804 		dmu_objset_rele_flags(os, B_TRUE, FTAG);
1805 		return (SET_ERROR(EINVAL));
1806 	}
1807 	error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1808 	    sizeof (zc->zc_value));
1809 	dmu_objset_rele_flags(os, B_TRUE, FTAG);
1810 
1811 	return (error);
1812 }
1813 
1814 /*
1815  * inputs:
1816  * zc_name		name of filesystem
1817  * zc_obj		object to find
1818  *
1819  * outputs:
1820  * zc_stat		stats on object
1821  * zc_value		path to object
1822  */
1823 static int
1824 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1825 {
1826 	objset_t *os;
1827 	int error;
1828 
1829 	/* XXX reading from objset not owned */
1830 	if ((error = dmu_objset_hold_flags(zc->zc_name, B_TRUE,
1831 	    FTAG, &os)) != 0)
1832 		return (error);
1833 	if (dmu_objset_type(os) != DMU_OST_ZFS) {
1834 		dmu_objset_rele_flags(os, B_TRUE, FTAG);
1835 		return (SET_ERROR(EINVAL));
1836 	}
1837 	error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1838 	    sizeof (zc->zc_value));
1839 	dmu_objset_rele_flags(os, B_TRUE, FTAG);
1840 
1841 	return (error);
1842 }
1843 
1844 static int
1845 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1846 {
1847 	spa_t *spa;
1848 	int error;
1849 	nvlist_t *config;
1850 
1851 	error = spa_open(zc->zc_name, &spa, FTAG);
1852 	if (error != 0)
1853 		return (error);
1854 
1855 	error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1856 	    zc->zc_iflags, &config);
1857 	if (error == 0) {
1858 		error = spa_vdev_add(spa, config);
1859 		nvlist_free(config);
1860 	}
1861 	spa_close(spa, FTAG);
1862 	return (error);
1863 }
1864 
1865 /*
1866  * inputs:
1867  * zc_name		name of the pool
1868  * zc_guid		guid of vdev to remove
1869  * zc_cookie		cancel removal
1870  */
1871 static int
1872 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1873 {
1874 	spa_t *spa;
1875 	int error;
1876 
1877 	error = spa_open(zc->zc_name, &spa, FTAG);
1878 	if (error != 0)
1879 		return (error);
1880 	if (zc->zc_cookie != 0) {
1881 		error = spa_vdev_remove_cancel(spa);
1882 	} else {
1883 		error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1884 	}
1885 	spa_close(spa, FTAG);
1886 	return (error);
1887 }
1888 
1889 static int
1890 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1891 {
1892 	spa_t *spa;
1893 	int error;
1894 	vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1895 
1896 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1897 		return (error);
1898 	switch (zc->zc_cookie) {
1899 	case VDEV_STATE_ONLINE:
1900 		error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1901 		break;
1902 
1903 	case VDEV_STATE_OFFLINE:
1904 		error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1905 		break;
1906 
1907 	case VDEV_STATE_FAULTED:
1908 		if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1909 		    zc->zc_obj != VDEV_AUX_EXTERNAL &&
1910 		    zc->zc_obj != VDEV_AUX_EXTERNAL_PERSIST)
1911 			zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1912 
1913 		error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1914 		break;
1915 
1916 	case VDEV_STATE_DEGRADED:
1917 		if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1918 		    zc->zc_obj != VDEV_AUX_EXTERNAL)
1919 			zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1920 
1921 		error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1922 		break;
1923 
1924 	default:
1925 		error = SET_ERROR(EINVAL);
1926 	}
1927 	zc->zc_cookie = newstate;
1928 	spa_close(spa, FTAG);
1929 	return (error);
1930 }
1931 
1932 static int
1933 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1934 {
1935 	spa_t *spa;
1936 	nvlist_t *config;
1937 	int replacing = zc->zc_cookie;
1938 	int rebuild = zc->zc_simple;
1939 	int error;
1940 
1941 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1942 		return (error);
1943 
1944 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1945 	    zc->zc_iflags, &config)) == 0) {
1946 		error = spa_vdev_attach(spa, zc->zc_guid, config, replacing,
1947 		    rebuild);
1948 		nvlist_free(config);
1949 	}
1950 
1951 	spa_close(spa, FTAG);
1952 	return (error);
1953 }
1954 
1955 static int
1956 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1957 {
1958 	spa_t *spa;
1959 	int error;
1960 
1961 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1962 		return (error);
1963 
1964 	error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1965 
1966 	spa_close(spa, FTAG);
1967 	return (error);
1968 }
1969 
1970 static int
1971 zfs_ioc_vdev_split(zfs_cmd_t *zc)
1972 {
1973 	spa_t *spa;
1974 	nvlist_t *config, *props = NULL;
1975 	int error;
1976 	boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
1977 
1978 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1979 		return (error);
1980 
1981 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1982 	    zc->zc_iflags, &config))) {
1983 		spa_close(spa, FTAG);
1984 		return (error);
1985 	}
1986 
1987 	if (zc->zc_nvlist_src_size != 0 && (error =
1988 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1989 	    zc->zc_iflags, &props))) {
1990 		spa_close(spa, FTAG);
1991 		nvlist_free(config);
1992 		return (error);
1993 	}
1994 
1995 	error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
1996 
1997 	spa_close(spa, FTAG);
1998 
1999 	nvlist_free(config);
2000 	nvlist_free(props);
2001 
2002 	return (error);
2003 }
2004 
2005 static int
2006 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
2007 {
2008 	spa_t *spa;
2009 	const char *path = zc->zc_value;
2010 	uint64_t guid = zc->zc_guid;
2011 	int error;
2012 
2013 	error = spa_open(zc->zc_name, &spa, FTAG);
2014 	if (error != 0)
2015 		return (error);
2016 
2017 	error = spa_vdev_setpath(spa, guid, path);
2018 	spa_close(spa, FTAG);
2019 	return (error);
2020 }
2021 
2022 static int
2023 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
2024 {
2025 	spa_t *spa;
2026 	const char *fru = zc->zc_value;
2027 	uint64_t guid = zc->zc_guid;
2028 	int error;
2029 
2030 	error = spa_open(zc->zc_name, &spa, FTAG);
2031 	if (error != 0)
2032 		return (error);
2033 
2034 	error = spa_vdev_setfru(spa, guid, fru);
2035 	spa_close(spa, FTAG);
2036 	return (error);
2037 }
2038 
2039 static int
2040 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
2041 {
2042 	int error = 0;
2043 	nvlist_t *nv;
2044 
2045 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2046 
2047 	if (zc->zc_nvlist_dst != 0 &&
2048 	    (error = dsl_prop_get_all(os, &nv)) == 0) {
2049 		dmu_objset_stats(os, nv);
2050 		/*
2051 		 * NB: zvol_get_stats() will read the objset contents,
2052 		 * which we aren't supposed to do with a
2053 		 * DS_MODE_USER hold, because it could be
2054 		 * inconsistent.  So this is a bit of a workaround...
2055 		 * XXX reading without owning
2056 		 */
2057 		if (!zc->zc_objset_stats.dds_inconsistent &&
2058 		    dmu_objset_type(os) == DMU_OST_ZVOL) {
2059 			error = zvol_get_stats(os, nv);
2060 			if (error == EIO) {
2061 				nvlist_free(nv);
2062 				return (error);
2063 			}
2064 			VERIFY0(error);
2065 		}
2066 		if (error == 0)
2067 			error = put_nvlist(zc, nv);
2068 		nvlist_free(nv);
2069 	}
2070 
2071 	return (error);
2072 }
2073 
2074 /*
2075  * inputs:
2076  * zc_name		name of filesystem
2077  * zc_nvlist_dst_size	size of buffer for property nvlist
2078  *
2079  * outputs:
2080  * zc_objset_stats	stats
2081  * zc_nvlist_dst	property nvlist
2082  * zc_nvlist_dst_size	size of property nvlist
2083  */
2084 static int
2085 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2086 {
2087 	objset_t *os;
2088 	int error;
2089 
2090 	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2091 	if (error == 0) {
2092 		error = zfs_ioc_objset_stats_impl(zc, os);
2093 		dmu_objset_rele(os, FTAG);
2094 	}
2095 
2096 	return (error);
2097 }
2098 
2099 /*
2100  * inputs:
2101  * zc_name		name of filesystem
2102  * zc_nvlist_dst_size	size of buffer for property nvlist
2103  *
2104  * outputs:
2105  * zc_nvlist_dst	received property nvlist
2106  * zc_nvlist_dst_size	size of received property nvlist
2107  *
2108  * Gets received properties (distinct from local properties on or after
2109  * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2110  * local property values.
2111  */
2112 static int
2113 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2114 {
2115 	int error = 0;
2116 	nvlist_t *nv;
2117 
2118 	/*
2119 	 * Without this check, we would return local property values if the
2120 	 * caller has not already received properties on or after
2121 	 * SPA_VERSION_RECVD_PROPS.
2122 	 */
2123 	if (!dsl_prop_get_hasrecvd(zc->zc_name))
2124 		return (SET_ERROR(ENOTSUP));
2125 
2126 	if (zc->zc_nvlist_dst != 0 &&
2127 	    (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2128 		error = put_nvlist(zc, nv);
2129 		nvlist_free(nv);
2130 	}
2131 
2132 	return (error);
2133 }
2134 
2135 static int
2136 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2137 {
2138 	uint64_t value;
2139 	int error;
2140 
2141 	/*
2142 	 * zfs_get_zplprop() will either find a value or give us
2143 	 * the default value (if there is one).
2144 	 */
2145 	if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2146 		return (error);
2147 	VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2148 	return (0);
2149 }
2150 
2151 /*
2152  * inputs:
2153  * zc_name		name of filesystem
2154  * zc_nvlist_dst_size	size of buffer for zpl property nvlist
2155  *
2156  * outputs:
2157  * zc_nvlist_dst	zpl property nvlist
2158  * zc_nvlist_dst_size	size of zpl property nvlist
2159  */
2160 static int
2161 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2162 {
2163 	objset_t *os;
2164 	int err;
2165 
2166 	/* XXX reading without owning */
2167 	if ((err = dmu_objset_hold(zc->zc_name, FTAG, &os)))
2168 		return (err);
2169 
2170 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2171 
2172 	/*
2173 	 * NB: nvl_add_zplprop() will read the objset contents,
2174 	 * which we aren't supposed to do with a DS_MODE_USER
2175 	 * hold, because it could be inconsistent.
2176 	 */
2177 	if (zc->zc_nvlist_dst != 0 &&
2178 	    !zc->zc_objset_stats.dds_inconsistent &&
2179 	    dmu_objset_type(os) == DMU_OST_ZFS) {
2180 		nvlist_t *nv;
2181 
2182 		VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2183 		if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2184 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2185 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2186 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2187 			err = put_nvlist(zc, nv);
2188 		nvlist_free(nv);
2189 	} else {
2190 		err = SET_ERROR(ENOENT);
2191 	}
2192 	dmu_objset_rele(os, FTAG);
2193 	return (err);
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 filesystem
2204  * zc_cookie		zap cursor
2205  * zc_objset_stats	stats
2206  * zc_nvlist_dst	property nvlist
2207  * zc_nvlist_dst_size	size of property nvlist
2208  */
2209 static int
2210 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2211 {
2212 	objset_t *os;
2213 	int error;
2214 	char *p;
2215 	size_t orig_len = strlen(zc->zc_name);
2216 
2217 top:
2218 	if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os))) {
2219 		if (error == ENOENT)
2220 			error = SET_ERROR(ESRCH);
2221 		return (error);
2222 	}
2223 
2224 	p = strrchr(zc->zc_name, '/');
2225 	if (p == NULL || p[1] != '\0')
2226 		(void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2227 	p = zc->zc_name + strlen(zc->zc_name);
2228 
2229 	do {
2230 		error = dmu_dir_list_next(os,
2231 		    sizeof (zc->zc_name) - (p - zc->zc_name), p,
2232 		    NULL, &zc->zc_cookie);
2233 		if (error == ENOENT)
2234 			error = SET_ERROR(ESRCH);
2235 	} while (error == 0 && zfs_dataset_name_hidden(zc->zc_name));
2236 	dmu_objset_rele(os, FTAG);
2237 
2238 	/*
2239 	 * If it's an internal dataset (ie. with a '$' in its name),
2240 	 * don't try to get stats for it, otherwise we'll return ENOENT.
2241 	 */
2242 	if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2243 		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2244 		if (error == ENOENT) {
2245 			/* We lost a race with destroy, get the next one. */
2246 			zc->zc_name[orig_len] = '\0';
2247 			goto top;
2248 		}
2249 	}
2250 	return (error);
2251 }
2252 
2253 /*
2254  * inputs:
2255  * zc_name		name of filesystem
2256  * zc_cookie		zap cursor
2257  * zc_nvlist_src	iteration range nvlist
2258  * zc_nvlist_src_size	size of iteration range nvlist
2259  *
2260  * outputs:
2261  * zc_name		name of next snapshot
2262  * zc_objset_stats	stats
2263  * zc_nvlist_dst	property nvlist
2264  * zc_nvlist_dst_size	size of property nvlist
2265  */
2266 static int
2267 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2268 {
2269 	int error;
2270 	objset_t *os, *ossnap;
2271 	dsl_dataset_t *ds;
2272 	uint64_t min_txg = 0, max_txg = 0;
2273 
2274 	if (zc->zc_nvlist_src_size != 0) {
2275 		nvlist_t *props = NULL;
2276 		error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2277 		    zc->zc_iflags, &props);
2278 		if (error != 0)
2279 			return (error);
2280 		(void) nvlist_lookup_uint64(props, SNAP_ITER_MIN_TXG,
2281 		    &min_txg);
2282 		(void) nvlist_lookup_uint64(props, SNAP_ITER_MAX_TXG,
2283 		    &max_txg);
2284 		nvlist_free(props);
2285 	}
2286 
2287 	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2288 	if (error != 0) {
2289 		return (error == ENOENT ? SET_ERROR(ESRCH) : error);
2290 	}
2291 
2292 	/*
2293 	 * A dataset name of maximum length cannot have any snapshots,
2294 	 * so exit immediately.
2295 	 */
2296 	if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >=
2297 	    ZFS_MAX_DATASET_NAME_LEN) {
2298 		dmu_objset_rele(os, FTAG);
2299 		return (SET_ERROR(ESRCH));
2300 	}
2301 
2302 	while (error == 0) {
2303 		if (issig(JUSTLOOKING) && issig(FORREAL)) {
2304 			error = SET_ERROR(EINTR);
2305 			break;
2306 		}
2307 
2308 		error = dmu_snapshot_list_next(os,
2309 		    sizeof (zc->zc_name) - strlen(zc->zc_name),
2310 		    zc->zc_name + strlen(zc->zc_name), &zc->zc_obj,
2311 		    &zc->zc_cookie, NULL);
2312 		if (error == ENOENT) {
2313 			error = SET_ERROR(ESRCH);
2314 			break;
2315 		} else if (error != 0) {
2316 			break;
2317 		}
2318 
2319 		error = dsl_dataset_hold_obj(dmu_objset_pool(os), zc->zc_obj,
2320 		    FTAG, &ds);
2321 		if (error != 0)
2322 			break;
2323 
2324 		if ((min_txg != 0 && dsl_get_creationtxg(ds) < min_txg) ||
2325 		    (max_txg != 0 && dsl_get_creationtxg(ds) > max_txg)) {
2326 			dsl_dataset_rele(ds, FTAG);
2327 			/* undo snapshot name append */
2328 			*(strchr(zc->zc_name, '@') + 1) = '\0';
2329 			/* skip snapshot */
2330 			continue;
2331 		}
2332 
2333 		if (zc->zc_simple) {
2334 			dsl_dataset_rele(ds, FTAG);
2335 			break;
2336 		}
2337 
2338 		if ((error = dmu_objset_from_ds(ds, &ossnap)) != 0) {
2339 			dsl_dataset_rele(ds, FTAG);
2340 			break;
2341 		}
2342 		if ((error = zfs_ioc_objset_stats_impl(zc, ossnap)) != 0) {
2343 			dsl_dataset_rele(ds, FTAG);
2344 			break;
2345 		}
2346 		dsl_dataset_rele(ds, FTAG);
2347 		break;
2348 	}
2349 
2350 	dmu_objset_rele(os, FTAG);
2351 	/* if we failed, undo the @ that we tacked on to zc_name */
2352 	if (error != 0)
2353 		*strchr(zc->zc_name, '@') = '\0';
2354 	return (error);
2355 }
2356 
2357 static int
2358 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2359 {
2360 	const char *propname = nvpair_name(pair);
2361 	uint64_t *valary;
2362 	unsigned int vallen;
2363 	const char *dash, *domain;
2364 	zfs_userquota_prop_t type;
2365 	uint64_t rid;
2366 	uint64_t quota;
2367 	zfsvfs_t *zfsvfs;
2368 	int err;
2369 
2370 	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2371 		nvlist_t *attrs;
2372 		VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2373 		if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2374 		    &pair) != 0)
2375 			return (SET_ERROR(EINVAL));
2376 	}
2377 
2378 	/*
2379 	 * A correctly constructed propname is encoded as
2380 	 * userquota@<rid>-<domain>.
2381 	 */
2382 	if ((dash = strchr(propname, '-')) == NULL ||
2383 	    nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2384 	    vallen != 3)
2385 		return (SET_ERROR(EINVAL));
2386 
2387 	domain = dash + 1;
2388 	type = valary[0];
2389 	rid = valary[1];
2390 	quota = valary[2];
2391 
2392 	err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2393 	if (err == 0) {
2394 		err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2395 		zfsvfs_rele(zfsvfs, FTAG);
2396 	}
2397 
2398 	return (err);
2399 }
2400 
2401 /*
2402  * If the named property is one that has a special function to set its value,
2403  * return 0 on success and a positive error code on failure; otherwise if it is
2404  * not one of the special properties handled by this function, return -1.
2405  *
2406  * XXX: It would be better for callers of the property interface if we handled
2407  * these special cases in dsl_prop.c (in the dsl layer).
2408  */
2409 static int
2410 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2411     nvpair_t *pair)
2412 {
2413 	const char *propname = nvpair_name(pair);
2414 	zfs_prop_t prop = zfs_name_to_prop(propname);
2415 	uint64_t intval = 0;
2416 	const char *strval = NULL;
2417 	int err = -1;
2418 
2419 	if (prop == ZPROP_INVAL) {
2420 		if (zfs_prop_userquota(propname))
2421 			return (zfs_prop_set_userquota(dsname, pair));
2422 		return (-1);
2423 	}
2424 
2425 	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2426 		nvlist_t *attrs;
2427 		VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2428 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2429 		    &pair) == 0);
2430 	}
2431 
2432 	/* all special properties are numeric except for keylocation */
2433 	if (zfs_prop_get_type(prop) == PROP_TYPE_STRING) {
2434 		strval = fnvpair_value_string(pair);
2435 	} else {
2436 		intval = fnvpair_value_uint64(pair);
2437 	}
2438 
2439 	switch (prop) {
2440 	case ZFS_PROP_QUOTA:
2441 		err = dsl_dir_set_quota(dsname, source, intval);
2442 		break;
2443 	case ZFS_PROP_REFQUOTA:
2444 		err = dsl_dataset_set_refquota(dsname, source, intval);
2445 		break;
2446 	case ZFS_PROP_FILESYSTEM_LIMIT:
2447 	case ZFS_PROP_SNAPSHOT_LIMIT:
2448 		if (intval == UINT64_MAX) {
2449 			/* clearing the limit, just do it */
2450 			err = 0;
2451 		} else {
2452 			err = dsl_dir_activate_fs_ss_limit(dsname);
2453 		}
2454 		/*
2455 		 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2456 		 * default path to set the value in the nvlist.
2457 		 */
2458 		if (err == 0)
2459 			err = -1;
2460 		break;
2461 	case ZFS_PROP_KEYLOCATION:
2462 		err = dsl_crypto_can_set_keylocation(dsname, strval);
2463 
2464 		/*
2465 		 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2466 		 * default path to set the value in the nvlist.
2467 		 */
2468 		if (err == 0)
2469 			err = -1;
2470 		break;
2471 	case ZFS_PROP_RESERVATION:
2472 		err = dsl_dir_set_reservation(dsname, source, intval);
2473 		break;
2474 	case ZFS_PROP_REFRESERVATION:
2475 		err = dsl_dataset_set_refreservation(dsname, source, intval);
2476 		break;
2477 	case ZFS_PROP_COMPRESSION:
2478 		err = dsl_dataset_set_compression(dsname, source, intval);
2479 		/*
2480 		 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2481 		 * default path to set the value in the nvlist.
2482 		 */
2483 		if (err == 0)
2484 			err = -1;
2485 		break;
2486 	case ZFS_PROP_VOLSIZE:
2487 		err = zvol_set_volsize(dsname, intval);
2488 		break;
2489 	case ZFS_PROP_SNAPDEV:
2490 		err = zvol_set_snapdev(dsname, source, intval);
2491 		break;
2492 	case ZFS_PROP_VOLMODE:
2493 		err = zvol_set_volmode(dsname, source, intval);
2494 		break;
2495 	case ZFS_PROP_VERSION:
2496 	{
2497 		zfsvfs_t *zfsvfs;
2498 
2499 		if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2500 			break;
2501 
2502 		err = zfs_set_version(zfsvfs, intval);
2503 		zfsvfs_rele(zfsvfs, FTAG);
2504 
2505 		if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2506 			zfs_cmd_t *zc;
2507 
2508 			zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2509 			(void) strlcpy(zc->zc_name, dsname,
2510 			    sizeof (zc->zc_name));
2511 			(void) zfs_ioc_userspace_upgrade(zc);
2512 			(void) zfs_ioc_id_quota_upgrade(zc);
2513 			kmem_free(zc, sizeof (zfs_cmd_t));
2514 		}
2515 		break;
2516 	}
2517 	default:
2518 		err = -1;
2519 	}
2520 
2521 	return (err);
2522 }
2523 
2524 static boolean_t
2525 zfs_is_namespace_prop(zfs_prop_t prop)
2526 {
2527 	switch (prop) {
2528 
2529 	case ZFS_PROP_ATIME:
2530 	case ZFS_PROP_RELATIME:
2531 	case ZFS_PROP_DEVICES:
2532 	case ZFS_PROP_EXEC:
2533 	case ZFS_PROP_SETUID:
2534 	case ZFS_PROP_READONLY:
2535 	case ZFS_PROP_XATTR:
2536 	case ZFS_PROP_NBMAND:
2537 		return (B_TRUE);
2538 
2539 	default:
2540 		return (B_FALSE);
2541 	}
2542 }
2543 
2544 /*
2545  * This function is best effort. If it fails to set any of the given properties,
2546  * it continues to set as many as it can and returns the last error
2547  * encountered. If the caller provides a non-NULL errlist, it will be filled in
2548  * with the list of names of all the properties that failed along with the
2549  * corresponding error numbers.
2550  *
2551  * If every property is set successfully, zero is returned and errlist is not
2552  * modified.
2553  */
2554 int
2555 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2556     nvlist_t *errlist)
2557 {
2558 	nvpair_t *pair;
2559 	nvpair_t *propval;
2560 	int rv = 0;
2561 	uint64_t intval;
2562 	const char *strval;
2563 	boolean_t should_update_mount_cache = B_FALSE;
2564 
2565 	nvlist_t *genericnvl = fnvlist_alloc();
2566 	nvlist_t *retrynvl = fnvlist_alloc();
2567 retry:
2568 	pair = NULL;
2569 	while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2570 		const char *propname = nvpair_name(pair);
2571 		zfs_prop_t prop = zfs_name_to_prop(propname);
2572 		int err = 0;
2573 
2574 		/* decode the property value */
2575 		propval = pair;
2576 		if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2577 			nvlist_t *attrs;
2578 			attrs = fnvpair_value_nvlist(pair);
2579 			if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2580 			    &propval) != 0)
2581 				err = SET_ERROR(EINVAL);
2582 		}
2583 
2584 		/* Validate value type */
2585 		if (err == 0 && source == ZPROP_SRC_INHERITED) {
2586 			/* inherited properties are expected to be booleans */
2587 			if (nvpair_type(propval) != DATA_TYPE_BOOLEAN)
2588 				err = SET_ERROR(EINVAL);
2589 		} else if (err == 0 && prop == ZPROP_INVAL) {
2590 			if (zfs_prop_user(propname)) {
2591 				if (nvpair_type(propval) != DATA_TYPE_STRING)
2592 					err = SET_ERROR(EINVAL);
2593 			} else if (zfs_prop_userquota(propname)) {
2594 				if (nvpair_type(propval) !=
2595 				    DATA_TYPE_UINT64_ARRAY)
2596 					err = SET_ERROR(EINVAL);
2597 			} else {
2598 				err = SET_ERROR(EINVAL);
2599 			}
2600 		} else if (err == 0) {
2601 			if (nvpair_type(propval) == DATA_TYPE_STRING) {
2602 				if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2603 					err = SET_ERROR(EINVAL);
2604 			} else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2605 				const char *unused;
2606 
2607 				intval = fnvpair_value_uint64(propval);
2608 
2609 				switch (zfs_prop_get_type(prop)) {
2610 				case PROP_TYPE_NUMBER:
2611 					break;
2612 				case PROP_TYPE_STRING:
2613 					err = SET_ERROR(EINVAL);
2614 					break;
2615 				case PROP_TYPE_INDEX:
2616 					if (zfs_prop_index_to_string(prop,
2617 					    intval, &unused) != 0)
2618 						err =
2619 						    SET_ERROR(ZFS_ERR_BADPROP);
2620 					break;
2621 				default:
2622 					cmn_err(CE_PANIC,
2623 					    "unknown property type");
2624 				}
2625 			} else {
2626 				err = SET_ERROR(EINVAL);
2627 			}
2628 		}
2629 
2630 		/* Validate permissions */
2631 		if (err == 0)
2632 			err = zfs_check_settable(dsname, pair, CRED());
2633 
2634 		if (err == 0) {
2635 			if (source == ZPROP_SRC_INHERITED)
2636 				err = -1; /* does not need special handling */
2637 			else
2638 				err = zfs_prop_set_special(dsname, source,
2639 				    pair);
2640 			if (err == -1) {
2641 				/*
2642 				 * For better performance we build up a list of
2643 				 * properties to set in a single transaction.
2644 				 */
2645 				err = nvlist_add_nvpair(genericnvl, pair);
2646 			} else if (err != 0 && nvl != retrynvl) {
2647 				/*
2648 				 * This may be a spurious error caused by
2649 				 * receiving quota and reservation out of order.
2650 				 * Try again in a second pass.
2651 				 */
2652 				err = nvlist_add_nvpair(retrynvl, pair);
2653 			}
2654 		}
2655 
2656 		if (err != 0) {
2657 			if (errlist != NULL)
2658 				fnvlist_add_int32(errlist, propname, err);
2659 			rv = err;
2660 		}
2661 
2662 		if (zfs_is_namespace_prop(prop))
2663 			should_update_mount_cache = B_TRUE;
2664 	}
2665 
2666 	if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2667 		nvl = retrynvl;
2668 		goto retry;
2669 	}
2670 
2671 	if (!nvlist_empty(genericnvl) &&
2672 	    dsl_props_set(dsname, source, genericnvl) != 0) {
2673 		/*
2674 		 * If this fails, we still want to set as many properties as we
2675 		 * can, so try setting them individually.
2676 		 */
2677 		pair = NULL;
2678 		while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2679 			const char *propname = nvpair_name(pair);
2680 			int err = 0;
2681 
2682 			propval = pair;
2683 			if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2684 				nvlist_t *attrs;
2685 				attrs = fnvpair_value_nvlist(pair);
2686 				propval = fnvlist_lookup_nvpair(attrs,
2687 				    ZPROP_VALUE);
2688 			}
2689 
2690 			if (nvpair_type(propval) == DATA_TYPE_STRING) {
2691 				strval = fnvpair_value_string(propval);
2692 				err = dsl_prop_set_string(dsname, propname,
2693 				    source, strval);
2694 			} else if (nvpair_type(propval) == DATA_TYPE_BOOLEAN) {
2695 				err = dsl_prop_inherit(dsname, propname,
2696 				    source);
2697 			} else {
2698 				intval = fnvpair_value_uint64(propval);
2699 				err = dsl_prop_set_int(dsname, propname, source,
2700 				    intval);
2701 			}
2702 
2703 			if (err != 0) {
2704 				if (errlist != NULL) {
2705 					fnvlist_add_int32(errlist, propname,
2706 					    err);
2707 				}
2708 				rv = err;
2709 			}
2710 		}
2711 	}
2712 	if (should_update_mount_cache)
2713 		zfs_ioctl_update_mount_cache(dsname);
2714 
2715 	nvlist_free(genericnvl);
2716 	nvlist_free(retrynvl);
2717 
2718 	return (rv);
2719 }
2720 
2721 /*
2722  * Check that all the properties are valid user properties.
2723  */
2724 static int
2725 zfs_check_userprops(nvlist_t *nvl)
2726 {
2727 	nvpair_t *pair = NULL;
2728 
2729 	while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2730 		const char *propname = nvpair_name(pair);
2731 
2732 		if (!zfs_prop_user(propname) ||
2733 		    nvpair_type(pair) != DATA_TYPE_STRING)
2734 			return (SET_ERROR(EINVAL));
2735 
2736 		if (strlen(propname) >= ZAP_MAXNAMELEN)
2737 			return (SET_ERROR(ENAMETOOLONG));
2738 
2739 		if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
2740 			return (SET_ERROR(E2BIG));
2741 	}
2742 	return (0);
2743 }
2744 
2745 static void
2746 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2747 {
2748 	nvpair_t *pair;
2749 
2750 	VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2751 
2752 	pair = NULL;
2753 	while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2754 		if (nvlist_exists(skipped, nvpair_name(pair)))
2755 			continue;
2756 
2757 		VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2758 	}
2759 }
2760 
2761 static int
2762 clear_received_props(const char *dsname, nvlist_t *props,
2763     nvlist_t *skipped)
2764 {
2765 	int err = 0;
2766 	nvlist_t *cleared_props = NULL;
2767 	props_skip(props, skipped, &cleared_props);
2768 	if (!nvlist_empty(cleared_props)) {
2769 		/*
2770 		 * Acts on local properties until the dataset has received
2771 		 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2772 		 */
2773 		zprop_source_t flags = (ZPROP_SRC_NONE |
2774 		    (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2775 		err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2776 	}
2777 	nvlist_free(cleared_props);
2778 	return (err);
2779 }
2780 
2781 /*
2782  * inputs:
2783  * zc_name		name of filesystem
2784  * zc_value		name of property to set
2785  * zc_nvlist_src{_size}	nvlist of properties to apply
2786  * zc_cookie		received properties flag
2787  *
2788  * outputs:
2789  * zc_nvlist_dst{_size} error for each unapplied received property
2790  */
2791 static int
2792 zfs_ioc_set_prop(zfs_cmd_t *zc)
2793 {
2794 	nvlist_t *nvl;
2795 	boolean_t received = zc->zc_cookie;
2796 	zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2797 	    ZPROP_SRC_LOCAL);
2798 	nvlist_t *errors;
2799 	int error;
2800 
2801 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2802 	    zc->zc_iflags, &nvl)) != 0)
2803 		return (error);
2804 
2805 	if (received) {
2806 		nvlist_t *origprops;
2807 
2808 		if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2809 			(void) clear_received_props(zc->zc_name,
2810 			    origprops, nvl);
2811 			nvlist_free(origprops);
2812 		}
2813 
2814 		error = dsl_prop_set_hasrecvd(zc->zc_name);
2815 	}
2816 
2817 	errors = fnvlist_alloc();
2818 	if (error == 0)
2819 		error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2820 
2821 	if (zc->zc_nvlist_dst != 0 && errors != NULL) {
2822 		(void) put_nvlist(zc, errors);
2823 	}
2824 
2825 	nvlist_free(errors);
2826 	nvlist_free(nvl);
2827 	return (error);
2828 }
2829 
2830 /*
2831  * inputs:
2832  * zc_name		name of filesystem
2833  * zc_value		name of property to inherit
2834  * zc_cookie		revert to received value if TRUE
2835  *
2836  * outputs:		none
2837  */
2838 static int
2839 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2840 {
2841 	const char *propname = zc->zc_value;
2842 	zfs_prop_t prop = zfs_name_to_prop(propname);
2843 	boolean_t received = zc->zc_cookie;
2844 	zprop_source_t source = (received
2845 	    ? ZPROP_SRC_NONE		/* revert to received value, if any */
2846 	    : ZPROP_SRC_INHERITED);	/* explicitly inherit */
2847 	nvlist_t *dummy;
2848 	nvpair_t *pair;
2849 	zprop_type_t type;
2850 	int err;
2851 
2852 	if (!received) {
2853 		/*
2854 		 * Only check this in the non-received case. We want to allow
2855 		 * 'inherit -S' to revert non-inheritable properties like quota
2856 		 * and reservation to the received or default values even though
2857 		 * they are not considered inheritable.
2858 		 */
2859 		if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2860 			return (SET_ERROR(EINVAL));
2861 	}
2862 
2863 	if (prop == ZPROP_INVAL) {
2864 		if (!zfs_prop_user(propname))
2865 			return (SET_ERROR(EINVAL));
2866 
2867 		type = PROP_TYPE_STRING;
2868 	} else if (prop == ZFS_PROP_VOLSIZE || prop == ZFS_PROP_VERSION) {
2869 		return (SET_ERROR(EINVAL));
2870 	} else {
2871 		type = zfs_prop_get_type(prop);
2872 	}
2873 
2874 	/*
2875 	 * zfs_prop_set_special() expects properties in the form of an
2876 	 * nvpair with type info.
2877 	 */
2878 	dummy = fnvlist_alloc();
2879 
2880 	switch (type) {
2881 	case PROP_TYPE_STRING:
2882 		VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2883 		break;
2884 	case PROP_TYPE_NUMBER:
2885 	case PROP_TYPE_INDEX:
2886 		VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2887 		break;
2888 	default:
2889 		err = SET_ERROR(EINVAL);
2890 		goto errout;
2891 	}
2892 
2893 	pair = nvlist_next_nvpair(dummy, NULL);
2894 	if (pair == NULL) {
2895 		err = SET_ERROR(EINVAL);
2896 	} else {
2897 		err = zfs_prop_set_special(zc->zc_name, source, pair);
2898 		if (err == -1) /* property is not "special", needs handling */
2899 			err = dsl_prop_inherit(zc->zc_name, zc->zc_value,
2900 			    source);
2901 	}
2902 
2903 errout:
2904 	nvlist_free(dummy);
2905 	return (err);
2906 }
2907 
2908 static int
2909 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2910 {
2911 	nvlist_t *props;
2912 	spa_t *spa;
2913 	int error;
2914 	nvpair_t *pair;
2915 
2916 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2917 	    zc->zc_iflags, &props)))
2918 		return (error);
2919 
2920 	/*
2921 	 * If the only property is the configfile, then just do a spa_lookup()
2922 	 * to handle the faulted case.
2923 	 */
2924 	pair = nvlist_next_nvpair(props, NULL);
2925 	if (pair != NULL && strcmp(nvpair_name(pair),
2926 	    zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2927 	    nvlist_next_nvpair(props, pair) == NULL) {
2928 		mutex_enter(&spa_namespace_lock);
2929 		if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2930 			spa_configfile_set(spa, props, B_FALSE);
2931 			spa_write_cachefile(spa, B_FALSE, B_TRUE);
2932 		}
2933 		mutex_exit(&spa_namespace_lock);
2934 		if (spa != NULL) {
2935 			nvlist_free(props);
2936 			return (0);
2937 		}
2938 	}
2939 
2940 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2941 		nvlist_free(props);
2942 		return (error);
2943 	}
2944 
2945 	error = spa_prop_set(spa, props);
2946 
2947 	nvlist_free(props);
2948 	spa_close(spa, FTAG);
2949 
2950 	return (error);
2951 }
2952 
2953 static int
2954 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2955 {
2956 	spa_t *spa;
2957 	int error;
2958 	nvlist_t *nvp = NULL;
2959 
2960 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2961 		/*
2962 		 * If the pool is faulted, there may be properties we can still
2963 		 * get (such as altroot and cachefile), so attempt to get them
2964 		 * anyway.
2965 		 */
2966 		mutex_enter(&spa_namespace_lock);
2967 		if ((spa = spa_lookup(zc->zc_name)) != NULL)
2968 			error = spa_prop_get(spa, &nvp);
2969 		mutex_exit(&spa_namespace_lock);
2970 	} else {
2971 		error = spa_prop_get(spa, &nvp);
2972 		spa_close(spa, FTAG);
2973 	}
2974 
2975 	if (error == 0 && zc->zc_nvlist_dst != 0)
2976 		error = put_nvlist(zc, nvp);
2977 	else
2978 		error = SET_ERROR(EFAULT);
2979 
2980 	nvlist_free(nvp);
2981 	return (error);
2982 }
2983 
2984 /*
2985  * innvl: {
2986  *     "vdevprops_set_vdev" -> guid
2987  *     "vdevprops_set_props" -> { prop -> value }
2988  * }
2989  *
2990  * outnvl: propname -> error code (int32)
2991  */
2992 static const zfs_ioc_key_t zfs_keys_vdev_set_props[] = {
2993 	{ZPOOL_VDEV_PROPS_SET_VDEV,	DATA_TYPE_UINT64,	0},
2994 	{ZPOOL_VDEV_PROPS_SET_PROPS,	DATA_TYPE_NVLIST,	0}
2995 };
2996 
2997 static int
2998 zfs_ioc_vdev_set_props(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
2999 {
3000 	spa_t *spa;
3001 	int error;
3002 	vdev_t *vd;
3003 	uint64_t vdev_guid;
3004 
3005 	/* Early validation */
3006 	if (nvlist_lookup_uint64(innvl, ZPOOL_VDEV_PROPS_SET_VDEV,
3007 	    &vdev_guid) != 0)
3008 		return (SET_ERROR(EINVAL));
3009 
3010 	if (outnvl == NULL)
3011 		return (SET_ERROR(EINVAL));
3012 
3013 	if ((error = spa_open(poolname, &spa, FTAG)) != 0)
3014 		return (error);
3015 
3016 	ASSERT(spa_writeable(spa));
3017 
3018 	if ((vd = spa_lookup_by_guid(spa, vdev_guid, B_TRUE)) == NULL) {
3019 		spa_close(spa, FTAG);
3020 		return (SET_ERROR(ENOENT));
3021 	}
3022 
3023 	error = vdev_prop_set(vd, innvl, outnvl);
3024 
3025 	spa_close(spa, FTAG);
3026 
3027 	return (error);
3028 }
3029 
3030 /*
3031  * innvl: {
3032  *     "vdevprops_get_vdev" -> guid
3033  *     (optional) "vdevprops_get_props" -> { propname -> propid }
3034  * }
3035  *
3036  * outnvl: propname -> value
3037  */
3038 static const zfs_ioc_key_t zfs_keys_vdev_get_props[] = {
3039 	{ZPOOL_VDEV_PROPS_GET_VDEV,	DATA_TYPE_UINT64,	0},
3040 	{ZPOOL_VDEV_PROPS_GET_PROPS,	DATA_TYPE_NVLIST,	ZK_OPTIONAL}
3041 };
3042 
3043 static int
3044 zfs_ioc_vdev_get_props(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3045 {
3046 	spa_t *spa;
3047 	int error;
3048 	vdev_t *vd;
3049 	uint64_t vdev_guid;
3050 
3051 	/* Early validation */
3052 	if (nvlist_lookup_uint64(innvl, ZPOOL_VDEV_PROPS_GET_VDEV,
3053 	    &vdev_guid) != 0)
3054 		return (SET_ERROR(EINVAL));
3055 
3056 	if (outnvl == NULL)
3057 		return (SET_ERROR(EINVAL));
3058 
3059 	if ((error = spa_open(poolname, &spa, FTAG)) != 0)
3060 		return (error);
3061 
3062 	if ((vd = spa_lookup_by_guid(spa, vdev_guid, B_TRUE)) == NULL) {
3063 		spa_close(spa, FTAG);
3064 		return (SET_ERROR(ENOENT));
3065 	}
3066 
3067 	error = vdev_prop_get(vd, innvl, outnvl);
3068 
3069 	spa_close(spa, FTAG);
3070 
3071 	return (error);
3072 }
3073 
3074 /*
3075  * inputs:
3076  * zc_name		name of filesystem
3077  * zc_nvlist_src{_size}	nvlist of delegated permissions
3078  * zc_perm_action	allow/unallow flag
3079  *
3080  * outputs:		none
3081  */
3082 static int
3083 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
3084 {
3085 	int error;
3086 	nvlist_t *fsaclnv = NULL;
3087 
3088 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3089 	    zc->zc_iflags, &fsaclnv)) != 0)
3090 		return (error);
3091 
3092 	/*
3093 	 * Verify nvlist is constructed correctly
3094 	 */
3095 	if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
3096 		nvlist_free(fsaclnv);
3097 		return (SET_ERROR(EINVAL));
3098 	}
3099 
3100 	/*
3101 	 * If we don't have PRIV_SYS_MOUNT, then validate
3102 	 * that user is allowed to hand out each permission in
3103 	 * the nvlist(s)
3104 	 */
3105 
3106 	error = secpolicy_zfs(CRED());
3107 	if (error != 0) {
3108 		if (zc->zc_perm_action == B_FALSE) {
3109 			error = dsl_deleg_can_allow(zc->zc_name,
3110 			    fsaclnv, CRED());
3111 		} else {
3112 			error = dsl_deleg_can_unallow(zc->zc_name,
3113 			    fsaclnv, CRED());
3114 		}
3115 	}
3116 
3117 	if (error == 0)
3118 		error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
3119 
3120 	nvlist_free(fsaclnv);
3121 	return (error);
3122 }
3123 
3124 /*
3125  * inputs:
3126  * zc_name		name of filesystem
3127  *
3128  * outputs:
3129  * zc_nvlist_src{_size}	nvlist of delegated permissions
3130  */
3131 static int
3132 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
3133 {
3134 	nvlist_t *nvp;
3135 	int error;
3136 
3137 	if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
3138 		error = put_nvlist(zc, nvp);
3139 		nvlist_free(nvp);
3140 	}
3141 
3142 	return (error);
3143 }
3144 
3145 /* ARGSUSED */
3146 static void
3147 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
3148 {
3149 	zfs_creat_t *zct = arg;
3150 
3151 	zfs_create_fs(os, cr, zct->zct_zplprops, tx);
3152 }
3153 
3154 #define	ZFS_PROP_UNDEFINED	((uint64_t)-1)
3155 
3156 /*
3157  * inputs:
3158  * os			parent objset pointer (NULL if root fs)
3159  * fuids_ok		fuids allowed in this version of the spa?
3160  * sa_ok		SAs allowed in this version of the spa?
3161  * createprops		list of properties requested by creator
3162  *
3163  * outputs:
3164  * zplprops	values for the zplprops we attach to the master node object
3165  * is_ci	true if requested file system will be purely case-insensitive
3166  *
3167  * Determine the settings for utf8only, normalization and
3168  * casesensitivity.  Specific values may have been requested by the
3169  * creator and/or we can inherit values from the parent dataset.  If
3170  * the file system is of too early a vintage, a creator can not
3171  * request settings for these properties, even if the requested
3172  * setting is the default value.  We don't actually want to create dsl
3173  * properties for these, so remove them from the source nvlist after
3174  * processing.
3175  */
3176 static int
3177 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3178     boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3179     nvlist_t *zplprops, boolean_t *is_ci)
3180 {
3181 	uint64_t sense = ZFS_PROP_UNDEFINED;
3182 	uint64_t norm = ZFS_PROP_UNDEFINED;
3183 	uint64_t u8 = ZFS_PROP_UNDEFINED;
3184 	int error;
3185 
3186 	ASSERT(zplprops != NULL);
3187 
3188 	/* parent dataset must be a filesystem */
3189 	if (os != NULL && os->os_phys->os_type != DMU_OST_ZFS)
3190 		return (SET_ERROR(ZFS_ERR_WRONG_PARENT));
3191 
3192 	/*
3193 	 * Pull out creator prop choices, if any.
3194 	 */
3195 	if (createprops) {
3196 		(void) nvlist_lookup_uint64(createprops,
3197 		    zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3198 		(void) nvlist_lookup_uint64(createprops,
3199 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3200 		(void) nvlist_remove_all(createprops,
3201 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3202 		(void) nvlist_lookup_uint64(createprops,
3203 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3204 		(void) nvlist_remove_all(createprops,
3205 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3206 		(void) nvlist_lookup_uint64(createprops,
3207 		    zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3208 		(void) nvlist_remove_all(createprops,
3209 		    zfs_prop_to_name(ZFS_PROP_CASE));
3210 	}
3211 
3212 	/*
3213 	 * If the zpl version requested is whacky or the file system
3214 	 * or pool is version is too "young" to support normalization
3215 	 * and the creator tried to set a value for one of the props,
3216 	 * error out.
3217 	 */
3218 	if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3219 	    (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3220 	    (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3221 	    (zplver < ZPL_VERSION_NORMALIZATION &&
3222 	    (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3223 	    sense != ZFS_PROP_UNDEFINED)))
3224 		return (SET_ERROR(ENOTSUP));
3225 
3226 	/*
3227 	 * Put the version in the zplprops
3228 	 */
3229 	VERIFY(nvlist_add_uint64(zplprops,
3230 	    zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3231 
3232 	if (norm == ZFS_PROP_UNDEFINED &&
3233 	    (error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm)) != 0)
3234 		return (error);
3235 	VERIFY(nvlist_add_uint64(zplprops,
3236 	    zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3237 
3238 	/*
3239 	 * If we're normalizing, names must always be valid UTF-8 strings.
3240 	 */
3241 	if (norm)
3242 		u8 = 1;
3243 	if (u8 == ZFS_PROP_UNDEFINED &&
3244 	    (error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8)) != 0)
3245 		return (error);
3246 	VERIFY(nvlist_add_uint64(zplprops,
3247 	    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3248 
3249 	if (sense == ZFS_PROP_UNDEFINED &&
3250 	    (error = zfs_get_zplprop(os, ZFS_PROP_CASE, &sense)) != 0)
3251 		return (error);
3252 	VERIFY(nvlist_add_uint64(zplprops,
3253 	    zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3254 
3255 	if (is_ci)
3256 		*is_ci = (sense == ZFS_CASE_INSENSITIVE);
3257 
3258 	return (0);
3259 }
3260 
3261 static int
3262 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3263     nvlist_t *zplprops, boolean_t *is_ci)
3264 {
3265 	boolean_t fuids_ok, sa_ok;
3266 	uint64_t zplver = ZPL_VERSION;
3267 	objset_t *os = NULL;
3268 	char parentname[ZFS_MAX_DATASET_NAME_LEN];
3269 	spa_t *spa;
3270 	uint64_t spa_vers;
3271 	int error;
3272 
3273 	zfs_get_parent(dataset, parentname, sizeof (parentname));
3274 
3275 	if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3276 		return (error);
3277 
3278 	spa_vers = spa_version(spa);
3279 	spa_close(spa, FTAG);
3280 
3281 	zplver = zfs_zpl_version_map(spa_vers);
3282 	fuids_ok = (zplver >= ZPL_VERSION_FUID);
3283 	sa_ok = (zplver >= ZPL_VERSION_SA);
3284 
3285 	/*
3286 	 * Open parent object set so we can inherit zplprop values.
3287 	 */
3288 	if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3289 		return (error);
3290 
3291 	error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3292 	    zplprops, is_ci);
3293 	dmu_objset_rele(os, FTAG);
3294 	return (error);
3295 }
3296 
3297 static int
3298 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3299     nvlist_t *zplprops, boolean_t *is_ci)
3300 {
3301 	boolean_t fuids_ok;
3302 	boolean_t sa_ok;
3303 	uint64_t zplver = ZPL_VERSION;
3304 	int error;
3305 
3306 	zplver = zfs_zpl_version_map(spa_vers);
3307 	fuids_ok = (zplver >= ZPL_VERSION_FUID);
3308 	sa_ok = (zplver >= ZPL_VERSION_SA);
3309 
3310 	error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3311 	    createprops, zplprops, is_ci);
3312 	return (error);
3313 }
3314 
3315 /*
3316  * innvl: {
3317  *     "type" -> dmu_objset_type_t (int32)
3318  *     (optional) "props" -> { prop -> value }
3319  *     (optional) "hidden_args" -> { "wkeydata" -> value }
3320  *         raw uint8_t array of encryption wrapping key data (32 bytes)
3321  * }
3322  *
3323  * outnvl: propname -> error code (int32)
3324  */
3325 
3326 static const zfs_ioc_key_t zfs_keys_create[] = {
3327 	{"type",	DATA_TYPE_INT32,	0},
3328 	{"props",	DATA_TYPE_NVLIST,	ZK_OPTIONAL},
3329 	{"hidden_args",	DATA_TYPE_NVLIST,	ZK_OPTIONAL},
3330 };
3331 
3332 static int
3333 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3334 {
3335 	int error = 0;
3336 	zfs_creat_t zct = { 0 };
3337 	nvlist_t *nvprops = NULL;
3338 	nvlist_t *hidden_args = NULL;
3339 	void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3340 	dmu_objset_type_t type;
3341 	boolean_t is_insensitive = B_FALSE;
3342 	dsl_crypto_params_t *dcp = NULL;
3343 
3344 	type = (dmu_objset_type_t)fnvlist_lookup_int32(innvl, "type");
3345 	(void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3346 	(void) nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);
3347 
3348 	switch (type) {
3349 	case DMU_OST_ZFS:
3350 		cbfunc = zfs_create_cb;
3351 		break;
3352 
3353 	case DMU_OST_ZVOL:
3354 		cbfunc = zvol_create_cb;
3355 		break;
3356 
3357 	default:
3358 		cbfunc = NULL;
3359 		break;
3360 	}
3361 	if (strchr(fsname, '@') ||
3362 	    strchr(fsname, '%'))
3363 		return (SET_ERROR(EINVAL));
3364 
3365 	zct.zct_props = nvprops;
3366 
3367 	if (cbfunc == NULL)
3368 		return (SET_ERROR(EINVAL));
3369 
3370 	if (type == DMU_OST_ZVOL) {
3371 		uint64_t volsize, volblocksize;
3372 
3373 		if (nvprops == NULL)
3374 			return (SET_ERROR(EINVAL));
3375 		if (nvlist_lookup_uint64(nvprops,
3376 		    zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3377 			return (SET_ERROR(EINVAL));
3378 
3379 		if ((error = nvlist_lookup_uint64(nvprops,
3380 		    zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3381 		    &volblocksize)) != 0 && error != ENOENT)
3382 			return (SET_ERROR(EINVAL));
3383 
3384 		if (error != 0)
3385 			volblocksize = zfs_prop_default_numeric(
3386 			    ZFS_PROP_VOLBLOCKSIZE);
3387 
3388 		if ((error = zvol_check_volblocksize(fsname,
3389 		    volblocksize)) != 0 ||
3390 		    (error = zvol_check_volsize(volsize,
3391 		    volblocksize)) != 0)
3392 			return (error);
3393 	} else if (type == DMU_OST_ZFS) {
3394 		int error;
3395 
3396 		/*
3397 		 * We have to have normalization and
3398 		 * case-folding flags correct when we do the
3399 		 * file system creation, so go figure them out
3400 		 * now.
3401 		 */
3402 		VERIFY(nvlist_alloc(&zct.zct_zplprops,
3403 		    NV_UNIQUE_NAME, KM_SLEEP) == 0);
3404 		error = zfs_fill_zplprops(fsname, nvprops,
3405 		    zct.zct_zplprops, &is_insensitive);
3406 		if (error != 0) {
3407 			nvlist_free(zct.zct_zplprops);
3408 			return (error);
3409 		}
3410 	}
3411 
3412 	error = dsl_crypto_params_create_nvlist(DCP_CMD_NONE, nvprops,
3413 	    hidden_args, &dcp);
3414 	if (error != 0) {
3415 		nvlist_free(zct.zct_zplprops);
3416 		return (error);
3417 	}
3418 
3419 	error = dmu_objset_create(fsname, type,
3420 	    is_insensitive ? DS_FLAG_CI_DATASET : 0, dcp, cbfunc, &zct);
3421 
3422 	nvlist_free(zct.zct_zplprops);
3423 	dsl_crypto_params_free(dcp, !!error);
3424 
3425 	/*
3426 	 * It would be nice to do this atomically.
3427 	 */
3428 	if (error == 0) {
3429 		error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3430 		    nvprops, outnvl);
3431 		if (error != 0) {
3432 			spa_t *spa;
3433 			int error2;
3434 
3435 			/*
3436 			 * Volumes will return EBUSY and cannot be destroyed
3437 			 * until all asynchronous minor handling (e.g. from
3438 			 * setting the volmode property) has completed. Wait for
3439 			 * the spa_zvol_taskq to drain then retry.
3440 			 */
3441 			error2 = dsl_destroy_head(fsname);
3442 			while ((error2 == EBUSY) && (type == DMU_OST_ZVOL)) {
3443 				error2 = spa_open(fsname, &spa, FTAG);
3444 				if (error2 == 0) {
3445 					taskq_wait(spa->spa_zvol_taskq);
3446 					spa_close(spa, FTAG);
3447 				}
3448 				error2 = dsl_destroy_head(fsname);
3449 			}
3450 		}
3451 	}
3452 	return (error);
3453 }
3454 
3455 /*
3456  * innvl: {
3457  *     "origin" -> name of origin snapshot
3458  *     (optional) "props" -> { prop -> value }
3459  *     (optional) "hidden_args" -> { "wkeydata" -> value }
3460  *         raw uint8_t array of encryption wrapping key data (32 bytes)
3461  * }
3462  *
3463  * outputs:
3464  * outnvl: propname -> error code (int32)
3465  */
3466 static const zfs_ioc_key_t zfs_keys_clone[] = {
3467 	{"origin",	DATA_TYPE_STRING,	0},
3468 	{"props",	DATA_TYPE_NVLIST,	ZK_OPTIONAL},
3469 	{"hidden_args",	DATA_TYPE_NVLIST,	ZK_OPTIONAL},
3470 };
3471 
3472 static int
3473 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3474 {
3475 	int error = 0;
3476 	nvlist_t *nvprops = NULL;
3477 	const char *origin_name;
3478 
3479 	origin_name = fnvlist_lookup_string(innvl, "origin");
3480 	(void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3481 
3482 	if (strchr(fsname, '@') ||
3483 	    strchr(fsname, '%'))
3484 		return (SET_ERROR(EINVAL));
3485 
3486 	if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3487 		return (SET_ERROR(EINVAL));
3488 
3489 	error = dmu_objset_clone(fsname, origin_name);
3490 
3491 	/*
3492 	 * It would be nice to do this atomically.
3493 	 */
3494 	if (error == 0) {
3495 		error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3496 		    nvprops, outnvl);
3497 		if (error != 0)
3498 			(void) dsl_destroy_head(fsname);
3499 	}
3500 	return (error);
3501 }
3502 
3503 static const zfs_ioc_key_t zfs_keys_remap[] = {
3504 	/* no nvl keys */
3505 };
3506 
3507 /* ARGSUSED */
3508 static int
3509 zfs_ioc_remap(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3510 {
3511 	/* This IOCTL is no longer supported. */
3512 	return (0);
3513 }
3514 
3515 /*
3516  * innvl: {
3517  *     "snaps" -> { snapshot1, snapshot2 }
3518  *     (optional) "props" -> { prop -> value (string) }
3519  * }
3520  *
3521  * outnvl: snapshot -> error code (int32)
3522  */
3523 static const zfs_ioc_key_t zfs_keys_snapshot[] = {
3524 	{"snaps",	DATA_TYPE_NVLIST,	0},
3525 	{"props",	DATA_TYPE_NVLIST,	ZK_OPTIONAL},
3526 };
3527 
3528 static int
3529 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3530 {
3531 	nvlist_t *snaps;
3532 	nvlist_t *props = NULL;
3533 	int error, poollen;
3534 	nvpair_t *pair;
3535 
3536 	(void) nvlist_lookup_nvlist(innvl, "props", &props);
3537 	if (!nvlist_empty(props) &&
3538 	    zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3539 		return (SET_ERROR(ENOTSUP));
3540 	if ((error = zfs_check_userprops(props)) != 0)
3541 		return (error);
3542 
3543 	snaps = fnvlist_lookup_nvlist(innvl, "snaps");
3544 	poollen = strlen(poolname);
3545 	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3546 	    pair = nvlist_next_nvpair(snaps, pair)) {
3547 		const char *name = nvpair_name(pair);
3548 		char *cp = strchr(name, '@');
3549 
3550 		/*
3551 		 * The snap name must contain an @, and the part after it must
3552 		 * contain only valid characters.
3553 		 */
3554 		if (cp == NULL ||
3555 		    zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3556 			return (SET_ERROR(EINVAL));
3557 
3558 		/*
3559 		 * The snap must be in the specified pool.
3560 		 */
3561 		if (strncmp(name, poolname, poollen) != 0 ||
3562 		    (name[poollen] != '/' && name[poollen] != '@'))
3563 			return (SET_ERROR(EXDEV));
3564 
3565 		/*
3566 		 * Check for permission to set the properties on the fs.
3567 		 */
3568 		if (!nvlist_empty(props)) {
3569 			*cp = '\0';
3570 			error = zfs_secpolicy_write_perms(name,
3571 			    ZFS_DELEG_PERM_USERPROP, CRED());
3572 			*cp = '@';
3573 			if (error != 0)
3574 				return (error);
3575 		}
3576 
3577 		/* This must be the only snap of this fs. */
3578 		for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3579 		    pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3580 			if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3581 			    == 0) {
3582 				return (SET_ERROR(EXDEV));
3583 			}
3584 		}
3585 	}
3586 
3587 	error = dsl_dataset_snapshot(snaps, props, outnvl);
3588 
3589 	return (error);
3590 }
3591 
3592 /*
3593  * innvl: "message" -> string
3594  */
3595 static const zfs_ioc_key_t zfs_keys_log_history[] = {
3596 	{"message",	DATA_TYPE_STRING,	0},
3597 };
3598 
3599 /* ARGSUSED */
3600 static int
3601 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3602 {
3603 	const char *message;
3604 	char *poolname;
3605 	spa_t *spa;
3606 	int error;
3607 
3608 	/*
3609 	 * The poolname in the ioctl is not set, we get it from the TSD,
3610 	 * which was set at the end of the last successful ioctl that allows
3611 	 * logging.  The secpolicy func already checked that it is set.
3612 	 * Only one log ioctl is allowed after each successful ioctl, so
3613 	 * we clear the TSD here.
3614 	 */
3615 	poolname = tsd_get(zfs_allow_log_key);
3616 	if (poolname == NULL)
3617 		return (SET_ERROR(EINVAL));
3618 	(void) tsd_set(zfs_allow_log_key, NULL);
3619 	error = spa_open(poolname, &spa, FTAG);
3620 	kmem_strfree(poolname);
3621 	if (error != 0)
3622 		return (error);
3623 
3624 	message = fnvlist_lookup_string(innvl, "message");
3625 
3626 	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3627 		spa_close(spa, FTAG);
3628 		return (SET_ERROR(ENOTSUP));
3629 	}
3630 
3631 	error = spa_history_log(spa, message);
3632 	spa_close(spa, FTAG);
3633 	return (error);
3634 }
3635 
3636 /*
3637  * This ioctl is used to set the bootenv configuration on the current
3638  * pool. This configuration is stored in the second padding area of the label,
3639  * and it is used by the bootloader(s) to store the bootloader and/or system
3640  * specific data.
3641  * The data is stored as nvlist data stream, and is protected by
3642  * an embedded checksum.
3643  * The version can have two possible values:
3644  * VB_RAW: nvlist should have key GRUB_ENVMAP, value DATA_TYPE_STRING.
3645  * VB_NVLIST: nvlist with arbitrary <key, value> pairs.
3646  */
3647 static const zfs_ioc_key_t zfs_keys_set_bootenv[] = {
3648 	{"version",	DATA_TYPE_UINT64,	0},
3649 	{"<keys>",	DATA_TYPE_ANY, ZK_OPTIONAL | ZK_WILDCARDLIST},
3650 };
3651 
3652 static int
3653 zfs_ioc_set_bootenv(const char *name, nvlist_t *innvl, nvlist_t *outnvl)
3654 {
3655 	int error;
3656 	spa_t *spa;
3657 
3658 	if ((error = spa_open(name, &spa, FTAG)) != 0)
3659 		return (error);
3660 	spa_vdev_state_enter(spa, SCL_ALL);
3661 	error = vdev_label_write_bootenv(spa->spa_root_vdev, innvl);
3662 	(void) spa_vdev_state_exit(spa, NULL, 0);
3663 	spa_close(spa, FTAG);
3664 	return (error);
3665 }
3666 
3667 static const zfs_ioc_key_t zfs_keys_get_bootenv[] = {
3668 	/* no nvl keys */
3669 };
3670 
3671 static int
3672 zfs_ioc_get_bootenv(const char *name, nvlist_t *innvl, nvlist_t *outnvl)
3673 {
3674 	spa_t *spa;
3675 	int error;
3676 
3677 	if ((error = spa_open(name, &spa, FTAG)) != 0)
3678 		return (error);
3679 	spa_vdev_state_enter(spa, SCL_ALL);
3680 	error = vdev_label_read_bootenv(spa->spa_root_vdev, outnvl);
3681 	(void) spa_vdev_state_exit(spa, NULL, 0);
3682 	spa_close(spa, FTAG);
3683 	return (error);
3684 }
3685 
3686 /*
3687  * The dp_config_rwlock must not be held when calling this, because the
3688  * unmount may need to write out data.
3689  *
3690  * This function is best-effort.  Callers must deal gracefully if it
3691  * remains mounted (or is remounted after this call).
3692  *
3693  * Returns 0 if the argument is not a snapshot, or it is not currently a
3694  * filesystem, or we were able to unmount it.  Returns error code otherwise.
3695  */
3696 void
3697 zfs_unmount_snap(const char *snapname)
3698 {
3699 	if (strchr(snapname, '@') == NULL)
3700 		return;
3701 
3702 	(void) zfsctl_snapshot_unmount(snapname, MNT_FORCE);
3703 }
3704 
3705 /* ARGSUSED */
3706 static int
3707 zfs_unmount_snap_cb(const char *snapname, void *arg)
3708 {
3709 	zfs_unmount_snap(snapname);
3710 	return (0);
3711 }
3712 
3713 /*
3714  * When a clone is destroyed, its origin may also need to be destroyed,
3715  * in which case it must be unmounted.  This routine will do that unmount
3716  * if necessary.
3717  */
3718 void
3719 zfs_destroy_unmount_origin(const char *fsname)
3720 {
3721 	int error;
3722 	objset_t *os;
3723 	dsl_dataset_t *ds;
3724 
3725 	error = dmu_objset_hold(fsname, FTAG, &os);
3726 	if (error != 0)
3727 		return;
3728 	ds = dmu_objset_ds(os);
3729 	if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3730 		char originname[ZFS_MAX_DATASET_NAME_LEN];
3731 		dsl_dataset_name(ds->ds_prev, originname);
3732 		dmu_objset_rele(os, FTAG);
3733 		zfs_unmount_snap(originname);
3734 	} else {
3735 		dmu_objset_rele(os, FTAG);
3736 	}
3737 }
3738 
3739 /*
3740  * innvl: {
3741  *     "snaps" -> { snapshot1, snapshot2 }
3742  *     (optional boolean) "defer"
3743  * }
3744  *
3745  * outnvl: snapshot -> error code (int32)
3746  */
3747 static const zfs_ioc_key_t zfs_keys_destroy_snaps[] = {
3748 	{"snaps",	DATA_TYPE_NVLIST,	0},
3749 	{"defer",	DATA_TYPE_BOOLEAN,	ZK_OPTIONAL},
3750 };
3751 
3752 /* ARGSUSED */
3753 static int
3754 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3755 {
3756 	int poollen;
3757 	nvlist_t *snaps;
3758 	nvpair_t *pair;
3759 	boolean_t defer;
3760 	spa_t *spa;
3761 
3762 	snaps = fnvlist_lookup_nvlist(innvl, "snaps");
3763 	defer = nvlist_exists(innvl, "defer");
3764 
3765 	poollen = strlen(poolname);
3766 	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3767 	    pair = nvlist_next_nvpair(snaps, pair)) {
3768 		const char *name = nvpair_name(pair);
3769 
3770 		/*
3771 		 * The snap must be in the specified pool to prevent the
3772 		 * invalid removal of zvol minors below.
3773 		 */
3774 		if (strncmp(name, poolname, poollen) != 0 ||
3775 		    (name[poollen] != '/' && name[poollen] != '@'))
3776 			return (SET_ERROR(EXDEV));
3777 
3778 		zfs_unmount_snap(nvpair_name(pair));
3779 		if (spa_open(name, &spa, FTAG) == 0) {
3780 			zvol_remove_minors(spa, name, B_TRUE);
3781 			spa_close(spa, FTAG);
3782 		}
3783 	}
3784 
3785 	return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3786 }
3787 
3788 /*
3789  * Create bookmarks. The bookmark names are of the form <fs>#<bmark>.
3790  * All bookmarks and snapshots must be in the same pool.
3791  * dsl_bookmark_create_nvl_validate describes the nvlist schema in more detail.
3792  *
3793  * innvl: {
3794  *     new_bookmark1 -> existing_snapshot,
3795  *     new_bookmark2 -> existing_bookmark,
3796  * }
3797  *
3798  * outnvl: bookmark -> error code (int32)
3799  *
3800  */
3801 static const zfs_ioc_key_t zfs_keys_bookmark[] = {
3802 	{"<bookmark>...",	DATA_TYPE_STRING,	ZK_WILDCARDLIST},
3803 };
3804 
3805 /* ARGSUSED */
3806 static int
3807 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3808 {
3809 	return (dsl_bookmark_create(innvl, outnvl));
3810 }
3811 
3812 /*
3813  * innvl: {
3814  *     property 1, property 2, ...
3815  * }
3816  *
3817  * outnvl: {
3818  *     bookmark name 1 -> { property 1, property 2, ... },
3819  *     bookmark name 2 -> { property 1, property 2, ... }
3820  * }
3821  *
3822  */
3823 static const zfs_ioc_key_t zfs_keys_get_bookmarks[] = {
3824 	{"<property>...", DATA_TYPE_BOOLEAN, ZK_WILDCARDLIST | ZK_OPTIONAL},
3825 };
3826 
3827 static int
3828 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3829 {
3830 	return (dsl_get_bookmarks(fsname, innvl, outnvl));
3831 }
3832 
3833 /*
3834  * innvl is not used.
3835  *
3836  * outnvl: {
3837  *     property 1, property 2, ...
3838  * }
3839  *
3840  */
3841 static const zfs_ioc_key_t zfs_keys_get_bookmark_props[] = {
3842 	/* no nvl keys */
3843 };
3844 
3845 /* ARGSUSED */
3846 static int
3847 zfs_ioc_get_bookmark_props(const char *bookmark, nvlist_t *innvl,
3848     nvlist_t *outnvl)
3849 {
3850 	char fsname[ZFS_MAX_DATASET_NAME_LEN];
3851 	char *bmname;
3852 
3853 	bmname = strchr(bookmark, '#');
3854 	if (bmname == NULL)
3855 		return (SET_ERROR(EINVAL));
3856 	bmname++;
3857 
3858 	(void) strlcpy(fsname, bookmark, sizeof (fsname));
3859 	*(strchr(fsname, '#')) = '\0';
3860 
3861 	return (dsl_get_bookmark_props(fsname, bmname, outnvl));
3862 }
3863 
3864 /*
3865  * innvl: {
3866  *     bookmark name 1, bookmark name 2
3867  * }
3868  *
3869  * outnvl: bookmark -> error code (int32)
3870  *
3871  */
3872 static const zfs_ioc_key_t zfs_keys_destroy_bookmarks[] = {
3873 	{"<bookmark>...",	DATA_TYPE_BOOLEAN,	ZK_WILDCARDLIST},
3874 };
3875 
3876 static int
3877 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
3878     nvlist_t *outnvl)
3879 {
3880 	int error, poollen;
3881 
3882 	poollen = strlen(poolname);
3883 	for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3884 	    pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3885 		const char *name = nvpair_name(pair);
3886 		const char *cp = strchr(name, '#');
3887 
3888 		/*
3889 		 * The bookmark name must contain an #, and the part after it
3890 		 * must contain only valid characters.
3891 		 */
3892 		if (cp == NULL ||
3893 		    zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3894 			return (SET_ERROR(EINVAL));
3895 
3896 		/*
3897 		 * The bookmark must be in the specified pool.
3898 		 */
3899 		if (strncmp(name, poolname, poollen) != 0 ||
3900 		    (name[poollen] != '/' && name[poollen] != '#'))
3901 			return (SET_ERROR(EXDEV));
3902 	}
3903 
3904 	error = dsl_bookmark_destroy(innvl, outnvl);
3905 	return (error);
3906 }
3907 
3908 static const zfs_ioc_key_t zfs_keys_channel_program[] = {
3909 	{"program",	DATA_TYPE_STRING,		0},
3910 	{"arg",		DATA_TYPE_ANY,			0},
3911 	{"sync",	DATA_TYPE_BOOLEAN_VALUE,	ZK_OPTIONAL},
3912 	{"instrlimit",	DATA_TYPE_UINT64,		ZK_OPTIONAL},
3913 	{"memlimit",	DATA_TYPE_UINT64,		ZK_OPTIONAL},
3914 };
3915 
3916 static int
3917 zfs_ioc_channel_program(const char *poolname, nvlist_t *innvl,
3918     nvlist_t *outnvl)
3919 {
3920 	char *program;
3921 	uint64_t instrlimit, memlimit;
3922 	boolean_t sync_flag;
3923 	nvpair_t *nvarg = NULL;
3924 
3925 	program = fnvlist_lookup_string(innvl, ZCP_ARG_PROGRAM);
3926 	if (0 != nvlist_lookup_boolean_value(innvl, ZCP_ARG_SYNC, &sync_flag)) {
3927 		sync_flag = B_TRUE;
3928 	}
3929 	if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_INSTRLIMIT, &instrlimit)) {
3930 		instrlimit = ZCP_DEFAULT_INSTRLIMIT;
3931 	}
3932 	if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_MEMLIMIT, &memlimit)) {
3933 		memlimit = ZCP_DEFAULT_MEMLIMIT;
3934 	}
3935 	nvarg = fnvlist_lookup_nvpair(innvl, ZCP_ARG_ARGLIST);
3936 
3937 	if (instrlimit == 0 || instrlimit > zfs_lua_max_instrlimit)
3938 		return (SET_ERROR(EINVAL));
3939 	if (memlimit == 0 || memlimit > zfs_lua_max_memlimit)
3940 		return (SET_ERROR(EINVAL));
3941 
3942 	return (zcp_eval(poolname, program, sync_flag, instrlimit, memlimit,
3943 	    nvarg, outnvl));
3944 }
3945 
3946 /*
3947  * innvl: unused
3948  * outnvl: empty
3949  */
3950 static const zfs_ioc_key_t zfs_keys_pool_checkpoint[] = {
3951 	/* no nvl keys */
3952 };
3953 
3954 /* ARGSUSED */
3955 static int
3956 zfs_ioc_pool_checkpoint(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3957 {
3958 	return (spa_checkpoint(poolname));
3959 }
3960 
3961 /*
3962  * innvl: unused
3963  * outnvl: empty
3964  */
3965 static const zfs_ioc_key_t zfs_keys_pool_discard_checkpoint[] = {
3966 	/* no nvl keys */
3967 };
3968 
3969 /* ARGSUSED */
3970 static int
3971 zfs_ioc_pool_discard_checkpoint(const char *poolname, nvlist_t *innvl,
3972     nvlist_t *outnvl)
3973 {
3974 	return (spa_checkpoint_discard(poolname));
3975 }
3976 
3977 /*
3978  * inputs:
3979  * zc_name		name of dataset to destroy
3980  * zc_defer_destroy	mark for deferred destroy
3981  *
3982  * outputs:		none
3983  */
3984 static int
3985 zfs_ioc_destroy(zfs_cmd_t *zc)
3986 {
3987 	objset_t *os;
3988 	dmu_objset_type_t ost;
3989 	int err;
3990 
3991 	err = dmu_objset_hold(zc->zc_name, FTAG, &os);
3992 	if (err != 0)
3993 		return (err);
3994 	ost = dmu_objset_type(os);
3995 	dmu_objset_rele(os, FTAG);
3996 
3997 	if (ost == DMU_OST_ZFS)
3998 		zfs_unmount_snap(zc->zc_name);
3999 
4000 	if (strchr(zc->zc_name, '@')) {
4001 		err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
4002 	} else {
4003 		err = dsl_destroy_head(zc->zc_name);
4004 		if (err == EEXIST) {
4005 			/*
4006 			 * It is possible that the given DS may have
4007 			 * hidden child (%recv) datasets - "leftovers"
4008 			 * resulting from the previously interrupted
4009 			 * 'zfs receive'.
4010 			 *
4011 			 * 6 extra bytes for /%recv
4012 			 */
4013 			char namebuf[ZFS_MAX_DATASET_NAME_LEN + 6];
4014 
4015 			if (snprintf(namebuf, sizeof (namebuf), "%s/%s",
4016 			    zc->zc_name, recv_clone_name) >=
4017 			    sizeof (namebuf))
4018 				return (SET_ERROR(EINVAL));
4019 
4020 			/*
4021 			 * Try to remove the hidden child (%recv) and after
4022 			 * that try to remove the target dataset.
4023 			 * If the hidden child (%recv) does not exist
4024 			 * the original error (EEXIST) will be returned
4025 			 */
4026 			err = dsl_destroy_head(namebuf);
4027 			if (err == 0)
4028 				err = dsl_destroy_head(zc->zc_name);
4029 			else if (err == ENOENT)
4030 				err = SET_ERROR(EEXIST);
4031 		}
4032 	}
4033 
4034 	return (err);
4035 }
4036 
4037 /*
4038  * innvl: {
4039  *     "initialize_command" -> POOL_INITIALIZE_{CANCEL|START|SUSPEND} (uint64)
4040  *     "initialize_vdevs": { -> guids to initialize (nvlist)
4041  *         "vdev_path_1": vdev_guid_1, (uint64),
4042  *         "vdev_path_2": vdev_guid_2, (uint64),
4043  *         ...
4044  *     },
4045  * }
4046  *
4047  * outnvl: {
4048  *     "initialize_vdevs": { -> initialization errors (nvlist)
4049  *         "vdev_path_1": errno, see function body for possible errnos (uint64)
4050  *         "vdev_path_2": errno, ... (uint64)
4051  *         ...
4052  *     }
4053  * }
4054  *
4055  * EINVAL is returned for an unknown commands or if any of the provided vdev
4056  * guids have be specified with a type other than uint64.
4057  */
4058 static const zfs_ioc_key_t zfs_keys_pool_initialize[] = {
4059 	{ZPOOL_INITIALIZE_COMMAND,	DATA_TYPE_UINT64,	0},
4060 	{ZPOOL_INITIALIZE_VDEVS,	DATA_TYPE_NVLIST,	0}
4061 };
4062 
4063 static int
4064 zfs_ioc_pool_initialize(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
4065 {
4066 	uint64_t cmd_type;
4067 	if (nvlist_lookup_uint64(innvl, ZPOOL_INITIALIZE_COMMAND,
4068 	    &cmd_type) != 0) {
4069 		return (SET_ERROR(EINVAL));
4070 	}
4071 
4072 	if (!(cmd_type == POOL_INITIALIZE_CANCEL ||
4073 	    cmd_type == POOL_INITIALIZE_START ||
4074 	    cmd_type == POOL_INITIALIZE_SUSPEND)) {
4075 		return (SET_ERROR(EINVAL));
4076 	}
4077 
4078 	nvlist_t *vdev_guids;
4079 	if (nvlist_lookup_nvlist(innvl, ZPOOL_INITIALIZE_VDEVS,
4080 	    &vdev_guids) != 0) {
4081 		return (SET_ERROR(EINVAL));
4082 	}
4083 
4084 	for (nvpair_t *pair = nvlist_next_nvpair(vdev_guids, NULL);
4085 	    pair != NULL; pair = nvlist_next_nvpair(vdev_guids, pair)) {
4086 		uint64_t vdev_guid;
4087 		if (nvpair_value_uint64(pair, &vdev_guid) != 0) {
4088 			return (SET_ERROR(EINVAL));
4089 		}
4090 	}
4091 
4092 	spa_t *spa;
4093 	int error = spa_open(poolname, &spa, FTAG);
4094 	if (error != 0)
4095 		return (error);
4096 
4097 	nvlist_t *vdev_errlist = fnvlist_alloc();
4098 	int total_errors = spa_vdev_initialize(spa, vdev_guids, cmd_type,
4099 	    vdev_errlist);
4100 
4101 	if (fnvlist_size(vdev_errlist) > 0) {
4102 		fnvlist_add_nvlist(outnvl, ZPOOL_INITIALIZE_VDEVS,
4103 		    vdev_errlist);
4104 	}
4105 	fnvlist_free(vdev_errlist);
4106 
4107 	spa_close(spa, FTAG);
4108 	return (total_errors > 0 ? SET_ERROR(EINVAL) : 0);
4109 }
4110 
4111 /*
4112  * innvl: {
4113  *     "trim_command" -> POOL_TRIM_{CANCEL|START|SUSPEND} (uint64)
4114  *     "trim_vdevs": { -> guids to TRIM (nvlist)
4115  *         "vdev_path_1": vdev_guid_1, (uint64),
4116  *         "vdev_path_2": vdev_guid_2, (uint64),
4117  *         ...
4118  *     },
4119  *     "trim_rate" -> Target TRIM rate in bytes/sec.
4120  *     "trim_secure" -> Set to request a secure TRIM.
4121  * }
4122  *
4123  * outnvl: {
4124  *     "trim_vdevs": { -> TRIM errors (nvlist)
4125  *         "vdev_path_1": errno, see function body for possible errnos (uint64)
4126  *         "vdev_path_2": errno, ... (uint64)
4127  *         ...
4128  *     }
4129  * }
4130  *
4131  * EINVAL is returned for an unknown commands or if any of the provided vdev
4132  * guids have be specified with a type other than uint64.
4133  */
4134 static const zfs_ioc_key_t zfs_keys_pool_trim[] = {
4135 	{ZPOOL_TRIM_COMMAND,	DATA_TYPE_UINT64,		0},
4136 	{ZPOOL_TRIM_VDEVS,	DATA_TYPE_NVLIST,		0},
4137 	{ZPOOL_TRIM_RATE,	DATA_TYPE_UINT64,		ZK_OPTIONAL},
4138 	{ZPOOL_TRIM_SECURE,	DATA_TYPE_BOOLEAN_VALUE,	ZK_OPTIONAL},
4139 };
4140 
4141 static int
4142 zfs_ioc_pool_trim(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
4143 {
4144 	uint64_t cmd_type;
4145 	if (nvlist_lookup_uint64(innvl, ZPOOL_TRIM_COMMAND, &cmd_type) != 0)
4146 		return (SET_ERROR(EINVAL));
4147 
4148 	if (!(cmd_type == POOL_TRIM_CANCEL ||
4149 	    cmd_type == POOL_TRIM_START ||
4150 	    cmd_type == POOL_TRIM_SUSPEND)) {
4151 		return (SET_ERROR(EINVAL));
4152 	}
4153 
4154 	nvlist_t *vdev_guids;
4155 	if (nvlist_lookup_nvlist(innvl, ZPOOL_TRIM_VDEVS, &vdev_guids) != 0)
4156 		return (SET_ERROR(EINVAL));
4157 
4158 	for (nvpair_t *pair = nvlist_next_nvpair(vdev_guids, NULL);
4159 	    pair != NULL; pair = nvlist_next_nvpair(vdev_guids, pair)) {
4160 		uint64_t vdev_guid;
4161 		if (nvpair_value_uint64(pair, &vdev_guid) != 0) {
4162 			return (SET_ERROR(EINVAL));
4163 		}
4164 	}
4165 
4166 	/* Optional, defaults to maximum rate when not provided */
4167 	uint64_t rate;
4168 	if (nvlist_lookup_uint64(innvl, ZPOOL_TRIM_RATE, &rate) != 0)
4169 		rate = 0;
4170 
4171 	/* Optional, defaults to standard TRIM when not provided */
4172 	boolean_t secure;
4173 	if (nvlist_lookup_boolean_value(innvl, ZPOOL_TRIM_SECURE,
4174 	    &secure) != 0) {
4175 		secure = B_FALSE;
4176 	}
4177 
4178 	spa_t *spa;
4179 	int error = spa_open(poolname, &spa, FTAG);
4180 	if (error != 0)
4181 		return (error);
4182 
4183 	nvlist_t *vdev_errlist = fnvlist_alloc();
4184 	int total_errors = spa_vdev_trim(spa, vdev_guids, cmd_type,
4185 	    rate, !!zfs_trim_metaslab_skip, secure, vdev_errlist);
4186 
4187 	if (fnvlist_size(vdev_errlist) > 0)
4188 		fnvlist_add_nvlist(outnvl, ZPOOL_TRIM_VDEVS, vdev_errlist);
4189 
4190 	fnvlist_free(vdev_errlist);
4191 
4192 	spa_close(spa, FTAG);
4193 	return (total_errors > 0 ? SET_ERROR(EINVAL) : 0);
4194 }
4195 
4196 /*
4197  * This ioctl waits for activity of a particular type to complete. If there is
4198  * no activity of that type in progress, it returns immediately, and the
4199  * returned value "waited" is false. If there is activity in progress, and no
4200  * tag is passed in, the ioctl blocks until all activity of that type is
4201  * complete, and then returns with "waited" set to true.
4202  *
4203  * If a tag is provided, it identifies a particular instance of an activity to
4204  * wait for. Currently, this is only valid for use with 'initialize', because
4205  * that is the only activity for which there can be multiple instances running
4206  * concurrently. In the case of 'initialize', the tag corresponds to the guid of
4207  * the vdev on which to wait.
4208  *
4209  * If a thread waiting in the ioctl receives a signal, the call will return
4210  * immediately, and the return value will be EINTR.
4211  *
4212  * innvl: {
4213  *     "wait_activity" -> int32_t
4214  *     (optional) "wait_tag" -> uint64_t
4215  * }
4216  *
4217  * outnvl: "waited" -> boolean_t
4218  */
4219 static const zfs_ioc_key_t zfs_keys_pool_wait[] = {
4220 	{ZPOOL_WAIT_ACTIVITY,	DATA_TYPE_INT32,		0},
4221 	{ZPOOL_WAIT_TAG,	DATA_TYPE_UINT64,		ZK_OPTIONAL},
4222 };
4223 
4224 static int
4225 zfs_ioc_wait(const char *name, nvlist_t *innvl, nvlist_t *outnvl)
4226 {
4227 	int32_t activity;
4228 	uint64_t tag;
4229 	boolean_t waited;
4230 	int error;
4231 
4232 	if (nvlist_lookup_int32(innvl, ZPOOL_WAIT_ACTIVITY, &activity) != 0)
4233 		return (EINVAL);
4234 
4235 	if (nvlist_lookup_uint64(innvl, ZPOOL_WAIT_TAG, &tag) == 0)
4236 		error = spa_wait_tag(name, activity, tag, &waited);
4237 	else
4238 		error = spa_wait(name, activity, &waited);
4239 
4240 	if (error == 0)
4241 		fnvlist_add_boolean_value(outnvl, ZPOOL_WAIT_WAITED, waited);
4242 
4243 	return (error);
4244 }
4245 
4246 /*
4247  * This ioctl waits for activity of a particular type to complete. If there is
4248  * no activity of that type in progress, it returns immediately, and the
4249  * returned value "waited" is false. If there is activity in progress, and no
4250  * tag is passed in, the ioctl blocks until all activity of that type is
4251  * complete, and then returns with "waited" set to true.
4252  *
4253  * If a thread waiting in the ioctl receives a signal, the call will return
4254  * immediately, and the return value will be EINTR.
4255  *
4256  * innvl: {
4257  *     "wait_activity" -> int32_t
4258  * }
4259  *
4260  * outnvl: "waited" -> boolean_t
4261  */
4262 static const zfs_ioc_key_t zfs_keys_fs_wait[] = {
4263 	{ZFS_WAIT_ACTIVITY,	DATA_TYPE_INT32,		0},
4264 };
4265 
4266 static int
4267 zfs_ioc_wait_fs(const char *name, nvlist_t *innvl, nvlist_t *outnvl)
4268 {
4269 	int32_t activity;
4270 	boolean_t waited = B_FALSE;
4271 	int error;
4272 	dsl_pool_t *dp;
4273 	dsl_dir_t *dd;
4274 	dsl_dataset_t *ds;
4275 
4276 	if (nvlist_lookup_int32(innvl, ZFS_WAIT_ACTIVITY, &activity) != 0)
4277 		return (SET_ERROR(EINVAL));
4278 
4279 	if (activity >= ZFS_WAIT_NUM_ACTIVITIES || activity < 0)
4280 		return (SET_ERROR(EINVAL));
4281 
4282 	if ((error = dsl_pool_hold(name, FTAG, &dp)) != 0)
4283 		return (error);
4284 
4285 	if ((error = dsl_dataset_hold(dp, name, FTAG, &ds)) != 0) {
4286 		dsl_pool_rele(dp, FTAG);
4287 		return (error);
4288 	}
4289 
4290 	dd = ds->ds_dir;
4291 	mutex_enter(&dd->dd_activity_lock);
4292 	dd->dd_activity_waiters++;
4293 
4294 	/*
4295 	 * We get a long-hold here so that the dsl_dataset_t and dsl_dir_t
4296 	 * aren't evicted while we're waiting. Normally this is prevented by
4297 	 * holding the pool, but we can't do that while we're waiting since
4298 	 * that would prevent TXGs from syncing out. Some of the functionality
4299 	 * of long-holds (e.g. preventing deletion) is unnecessary for this
4300 	 * case, since we would cancel the waiters before proceeding with a
4301 	 * deletion. An alternative mechanism for keeping the dataset around
4302 	 * could be developed but this is simpler.
4303 	 */
4304 	dsl_dataset_long_hold(ds, FTAG);
4305 	dsl_pool_rele(dp, FTAG);
4306 
4307 	error = dsl_dir_wait(dd, ds, activity, &waited);
4308 
4309 	dsl_dataset_long_rele(ds, FTAG);
4310 	dd->dd_activity_waiters--;
4311 	if (dd->dd_activity_waiters == 0)
4312 		cv_signal(&dd->dd_activity_cv);
4313 	mutex_exit(&dd->dd_activity_lock);
4314 
4315 	dsl_dataset_rele(ds, FTAG);
4316 
4317 	if (error == 0)
4318 		fnvlist_add_boolean_value(outnvl, ZFS_WAIT_WAITED, waited);
4319 
4320 	return (error);
4321 }
4322 
4323 /*
4324  * fsname is name of dataset to rollback (to most recent snapshot)
4325  *
4326  * innvl may contain name of expected target snapshot
4327  *
4328  * outnvl: "target" -> name of most recent snapshot
4329  * }
4330  */
4331 static const zfs_ioc_key_t zfs_keys_rollback[] = {
4332 	{"target",	DATA_TYPE_STRING,	ZK_OPTIONAL},
4333 };
4334 
4335 /* ARGSUSED */
4336 static int
4337 zfs_ioc_rollback(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
4338 {
4339 	zfsvfs_t *zfsvfs;
4340 	zvol_state_handle_t *zv;
4341 	char *target = NULL;
4342 	int error;
4343 
4344 	(void) nvlist_lookup_string(innvl, "target", &target);
4345 	if (target != NULL) {
4346 		const char *cp = strchr(target, '@');
4347 
4348 		/*
4349 		 * The snap name must contain an @, and the part after it must
4350 		 * contain only valid characters.
4351 		 */
4352 		if (cp == NULL ||
4353 		    zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
4354 			return (SET_ERROR(EINVAL));
4355 	}
4356 
4357 	if (getzfsvfs(fsname, &zfsvfs) == 0) {
4358 		dsl_dataset_t *ds;
4359 
4360 		ds = dmu_objset_ds(zfsvfs->z_os);
4361 		error = zfs_suspend_fs(zfsvfs);
4362 		if (error == 0) {
4363 			int resume_err;
4364 
4365 			error = dsl_dataset_rollback(fsname, target, zfsvfs,
4366 			    outnvl);
4367 			resume_err = zfs_resume_fs(zfsvfs, ds);
4368 			error = error ? error : resume_err;
4369 		}
4370 		zfs_vfs_rele(zfsvfs);
4371 	} else if ((zv = zvol_suspend(fsname)) != NULL) {
4372 		error = dsl_dataset_rollback(fsname, target, zvol_tag(zv),
4373 		    outnvl);
4374 		zvol_resume(zv);
4375 	} else {
4376 		error = dsl_dataset_rollback(fsname, target, NULL, outnvl);
4377 	}
4378 	return (error);
4379 }
4380 
4381 static int
4382 recursive_unmount(const char *fsname, void *arg)
4383 {
4384 	const char *snapname = arg;
4385 	char *fullname;
4386 
4387 	fullname = kmem_asprintf("%s@%s", fsname, snapname);
4388 	zfs_unmount_snap(fullname);
4389 	kmem_strfree(fullname);
4390 
4391 	return (0);
4392 }
4393 
4394 /*
4395  *
4396  * snapname is the snapshot to redact.
4397  * innvl: {
4398  *     "bookname" -> (string)
4399  *         shortname of the redaction bookmark to generate
4400  *     "snapnv" -> (nvlist, values ignored)
4401  *         snapshots to redact snapname with respect to
4402  * }
4403  *
4404  * outnvl is unused
4405  */
4406 
4407 /* ARGSUSED */
4408 static const zfs_ioc_key_t zfs_keys_redact[] = {
4409 	{"bookname",		DATA_TYPE_STRING,	0},
4410 	{"snapnv",		DATA_TYPE_NVLIST,	0},
4411 };
4412 static int
4413 zfs_ioc_redact(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
4414 {
4415 	nvlist_t *redactnvl = NULL;
4416 	char *redactbook = NULL;
4417 
4418 	if (nvlist_lookup_nvlist(innvl, "snapnv", &redactnvl) != 0)
4419 		return (SET_ERROR(EINVAL));
4420 	if (fnvlist_num_pairs(redactnvl) == 0)
4421 		return (SET_ERROR(ENXIO));
4422 	if (nvlist_lookup_string(innvl, "bookname", &redactbook) != 0)
4423 		return (SET_ERROR(EINVAL));
4424 
4425 	return (dmu_redact_snap(snapname, redactnvl, redactbook));
4426 }
4427 
4428 /*
4429  * inputs:
4430  * zc_name	old name of dataset
4431  * zc_value	new name of dataset
4432  * zc_cookie	recursive flag (only valid for snapshots)
4433  *
4434  * outputs:	none
4435  */
4436 static int
4437 zfs_ioc_rename(zfs_cmd_t *zc)
4438 {
4439 	objset_t *os;
4440 	dmu_objset_type_t ost;
4441 	boolean_t recursive = zc->zc_cookie & 1;
4442 	boolean_t nounmount = !!(zc->zc_cookie & 2);
4443 	char *at;
4444 	int err;
4445 
4446 	/* "zfs rename" from and to ...%recv datasets should both fail */
4447 	zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
4448 	zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
4449 	if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0 ||
4450 	    dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4451 	    strchr(zc->zc_name, '%') || strchr(zc->zc_value, '%'))
4452 		return (SET_ERROR(EINVAL));
4453 
4454 	err = dmu_objset_hold(zc->zc_name, FTAG, &os);
4455 	if (err != 0)
4456 		return (err);
4457 	ost = dmu_objset_type(os);
4458 	dmu_objset_rele(os, FTAG);
4459 
4460 	at = strchr(zc->zc_name, '@');
4461 	if (at != NULL) {
4462 		/* snaps must be in same fs */
4463 		int error;
4464 
4465 		if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
4466 			return (SET_ERROR(EXDEV));
4467 		*at = '\0';
4468 		if (ost == DMU_OST_ZFS && !nounmount) {
4469 			error = dmu_objset_find(zc->zc_name,
4470 			    recursive_unmount, at + 1,
4471 			    recursive ? DS_FIND_CHILDREN : 0);
4472 			if (error != 0) {
4473 				*at = '@';
4474 				return (error);
4475 			}
4476 		}
4477 		error = dsl_dataset_rename_snapshot(zc->zc_name,
4478 		    at + 1, strchr(zc->zc_value, '@') + 1, recursive);
4479 		*at = '@';
4480 
4481 		return (error);
4482 	} else {
4483 		return (dsl_dir_rename(zc->zc_name, zc->zc_value));
4484 	}
4485 }
4486 
4487 static int
4488 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
4489 {
4490 	const char *propname = nvpair_name(pair);
4491 	boolean_t issnap = (strchr(dsname, '@') != NULL);
4492 	zfs_prop_t prop = zfs_name_to_prop(propname);
4493 	uint64_t intval, compval;
4494 	int err;
4495 
4496 	if (prop == ZPROP_INVAL) {
4497 		if (zfs_prop_user(propname)) {
4498 			if ((err = zfs_secpolicy_write_perms(dsname,
4499 			    ZFS_DELEG_PERM_USERPROP, cr)))
4500 				return (err);
4501 			return (0);
4502 		}
4503 
4504 		if (!issnap && zfs_prop_userquota(propname)) {
4505 			const char *perm = NULL;
4506 			const char *uq_prefix =
4507 			    zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
4508 			const char *gq_prefix =
4509 			    zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
4510 			const char *uiq_prefix =
4511 			    zfs_userquota_prop_prefixes[ZFS_PROP_USEROBJQUOTA];
4512 			const char *giq_prefix =
4513 			    zfs_userquota_prop_prefixes[ZFS_PROP_GROUPOBJQUOTA];
4514 			const char *pq_prefix =
4515 			    zfs_userquota_prop_prefixes[ZFS_PROP_PROJECTQUOTA];
4516 			const char *piq_prefix = zfs_userquota_prop_prefixes[\
4517 			    ZFS_PROP_PROJECTOBJQUOTA];
4518 
4519 			if (strncmp(propname, uq_prefix,
4520 			    strlen(uq_prefix)) == 0) {
4521 				perm = ZFS_DELEG_PERM_USERQUOTA;
4522 			} else if (strncmp(propname, uiq_prefix,
4523 			    strlen(uiq_prefix)) == 0) {
4524 				perm = ZFS_DELEG_PERM_USEROBJQUOTA;
4525 			} else if (strncmp(propname, gq_prefix,
4526 			    strlen(gq_prefix)) == 0) {
4527 				perm = ZFS_DELEG_PERM_GROUPQUOTA;
4528 			} else if (strncmp(propname, giq_prefix,
4529 			    strlen(giq_prefix)) == 0) {
4530 				perm = ZFS_DELEG_PERM_GROUPOBJQUOTA;
4531 			} else if (strncmp(propname, pq_prefix,
4532 			    strlen(pq_prefix)) == 0) {
4533 				perm = ZFS_DELEG_PERM_PROJECTQUOTA;
4534 			} else if (strncmp(propname, piq_prefix,
4535 			    strlen(piq_prefix)) == 0) {
4536 				perm = ZFS_DELEG_PERM_PROJECTOBJQUOTA;
4537 			} else {
4538 				/* {USER|GROUP|PROJECT}USED are read-only */
4539 				return (SET_ERROR(EINVAL));
4540 			}
4541 
4542 			if ((err = zfs_secpolicy_write_perms(dsname, perm, cr)))
4543 				return (err);
4544 			return (0);
4545 		}
4546 
4547 		return (SET_ERROR(EINVAL));
4548 	}
4549 
4550 	if (issnap)
4551 		return (SET_ERROR(EINVAL));
4552 
4553 	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
4554 		/*
4555 		 * dsl_prop_get_all_impl() returns properties in this
4556 		 * format.
4557 		 */
4558 		nvlist_t *attrs;
4559 		VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
4560 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4561 		    &pair) == 0);
4562 	}
4563 
4564 	/*
4565 	 * Check that this value is valid for this pool version
4566 	 */
4567 	switch (prop) {
4568 	case ZFS_PROP_COMPRESSION:
4569 		/*
4570 		 * If the user specified gzip compression, make sure
4571 		 * the SPA supports it. We ignore any errors here since
4572 		 * we'll catch them later.
4573 		 */
4574 		if (nvpair_value_uint64(pair, &intval) == 0) {
4575 			compval = ZIO_COMPRESS_ALGO(intval);
4576 			if (compval >= ZIO_COMPRESS_GZIP_1 &&
4577 			    compval <= ZIO_COMPRESS_GZIP_9 &&
4578 			    zfs_earlier_version(dsname,
4579 			    SPA_VERSION_GZIP_COMPRESSION)) {
4580 				return (SET_ERROR(ENOTSUP));
4581 			}
4582 
4583 			if (compval == ZIO_COMPRESS_ZLE &&
4584 			    zfs_earlier_version(dsname,
4585 			    SPA_VERSION_ZLE_COMPRESSION))
4586 				return (SET_ERROR(ENOTSUP));
4587 
4588 			if (compval == ZIO_COMPRESS_LZ4) {
4589 				spa_t *spa;
4590 
4591 				if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4592 					return (err);
4593 
4594 				if (!spa_feature_is_enabled(spa,
4595 				    SPA_FEATURE_LZ4_COMPRESS)) {
4596 					spa_close(spa, FTAG);
4597 					return (SET_ERROR(ENOTSUP));
4598 				}
4599 				spa_close(spa, FTAG);
4600 			}
4601 
4602 			if (compval == ZIO_COMPRESS_ZSTD) {
4603 				spa_t *spa;
4604 
4605 				if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4606 					return (err);
4607 
4608 				if (!spa_feature_is_enabled(spa,
4609 				    SPA_FEATURE_ZSTD_COMPRESS)) {
4610 					spa_close(spa, FTAG);
4611 					return (SET_ERROR(ENOTSUP));
4612 				}
4613 				spa_close(spa, FTAG);
4614 			}
4615 		}
4616 		break;
4617 
4618 	case ZFS_PROP_COPIES:
4619 		if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
4620 			return (SET_ERROR(ENOTSUP));
4621 		break;
4622 
4623 	case ZFS_PROP_VOLBLOCKSIZE:
4624 	case ZFS_PROP_RECORDSIZE:
4625 		/* Record sizes above 128k need the feature to be enabled */
4626 		if (nvpair_value_uint64(pair, &intval) == 0 &&
4627 		    intval > SPA_OLD_MAXBLOCKSIZE) {
4628 			spa_t *spa;
4629 
4630 			/*
4631 			 * We don't allow setting the property above 1MB,
4632 			 * unless the tunable has been changed.
4633 			 */
4634 			if (intval > zfs_max_recordsize ||
4635 			    intval > SPA_MAXBLOCKSIZE)
4636 				return (SET_ERROR(ERANGE));
4637 
4638 			if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4639 				return (err);
4640 
4641 			if (!spa_feature_is_enabled(spa,
4642 			    SPA_FEATURE_LARGE_BLOCKS)) {
4643 				spa_close(spa, FTAG);
4644 				return (SET_ERROR(ENOTSUP));
4645 			}
4646 			spa_close(spa, FTAG);
4647 		}
4648 		break;
4649 
4650 	case ZFS_PROP_DNODESIZE:
4651 		/* Dnode sizes above 512 need the feature to be enabled */
4652 		if (nvpair_value_uint64(pair, &intval) == 0 &&
4653 		    intval != ZFS_DNSIZE_LEGACY) {
4654 			spa_t *spa;
4655 
4656 			if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4657 				return (err);
4658 
4659 			if (!spa_feature_is_enabled(spa,
4660 			    SPA_FEATURE_LARGE_DNODE)) {
4661 				spa_close(spa, FTAG);
4662 				return (SET_ERROR(ENOTSUP));
4663 			}
4664 			spa_close(spa, FTAG);
4665 		}
4666 		break;
4667 
4668 	case ZFS_PROP_SPECIAL_SMALL_BLOCKS:
4669 		/*
4670 		 * This property could require the allocation classes
4671 		 * feature to be active for setting, however we allow
4672 		 * it so that tests of settable properties succeed.
4673 		 * The CLI will issue a warning in this case.
4674 		 */
4675 		break;
4676 
4677 	case ZFS_PROP_SHARESMB:
4678 		if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
4679 			return (SET_ERROR(ENOTSUP));
4680 		break;
4681 
4682 	case ZFS_PROP_ACLINHERIT:
4683 		if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
4684 		    nvpair_value_uint64(pair, &intval) == 0) {
4685 			if (intval == ZFS_ACL_PASSTHROUGH_X &&
4686 			    zfs_earlier_version(dsname,
4687 			    SPA_VERSION_PASSTHROUGH_X))
4688 				return (SET_ERROR(ENOTSUP));
4689 		}
4690 		break;
4691 	case ZFS_PROP_CHECKSUM:
4692 	case ZFS_PROP_DEDUP:
4693 	{
4694 		spa_feature_t feature;
4695 		spa_t *spa;
4696 		int err;
4697 
4698 		/* dedup feature version checks */
4699 		if (prop == ZFS_PROP_DEDUP &&
4700 		    zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
4701 			return (SET_ERROR(ENOTSUP));
4702 
4703 		if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
4704 		    nvpair_value_uint64(pair, &intval) == 0) {
4705 			/* check prop value is enabled in features */
4706 			feature = zio_checksum_to_feature(
4707 			    intval & ZIO_CHECKSUM_MASK);
4708 			if (feature == SPA_FEATURE_NONE)
4709 				break;
4710 
4711 			if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4712 				return (err);
4713 
4714 			if (!spa_feature_is_enabled(spa, feature)) {
4715 				spa_close(spa, FTAG);
4716 				return (SET_ERROR(ENOTSUP));
4717 			}
4718 			spa_close(spa, FTAG);
4719 		}
4720 		break;
4721 	}
4722 
4723 	default:
4724 		break;
4725 	}
4726 
4727 	return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
4728 }
4729 
4730 /*
4731  * Removes properties from the given props list that fail permission checks
4732  * needed to clear them and to restore them in case of a receive error. For each
4733  * property, make sure we have both set and inherit permissions.
4734  *
4735  * Returns the first error encountered if any permission checks fail. If the
4736  * caller provides a non-NULL errlist, it also gives the complete list of names
4737  * of all the properties that failed a permission check along with the
4738  * corresponding error numbers. The caller is responsible for freeing the
4739  * returned errlist.
4740  *
4741  * If every property checks out successfully, zero is returned and the list
4742  * pointed at by errlist is NULL.
4743  */
4744 static int
4745 zfs_check_clearable(const char *dataset, nvlist_t *props, nvlist_t **errlist)
4746 {
4747 	zfs_cmd_t *zc;
4748 	nvpair_t *pair, *next_pair;
4749 	nvlist_t *errors;
4750 	int err, rv = 0;
4751 
4752 	if (props == NULL)
4753 		return (0);
4754 
4755 	VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4756 
4757 	zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
4758 	(void) strlcpy(zc->zc_name, dataset, sizeof (zc->zc_name));
4759 	pair = nvlist_next_nvpair(props, NULL);
4760 	while (pair != NULL) {
4761 		next_pair = nvlist_next_nvpair(props, pair);
4762 
4763 		(void) strlcpy(zc->zc_value, nvpair_name(pair),
4764 		    sizeof (zc->zc_value));
4765 		if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
4766 		    (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
4767 			VERIFY(nvlist_remove_nvpair(props, pair) == 0);
4768 			VERIFY(nvlist_add_int32(errors,
4769 			    zc->zc_value, err) == 0);
4770 		}
4771 		pair = next_pair;
4772 	}
4773 	kmem_free(zc, sizeof (zfs_cmd_t));
4774 
4775 	if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
4776 		nvlist_free(errors);
4777 		errors = NULL;
4778 	} else {
4779 		VERIFY(nvpair_value_int32(pair, &rv) == 0);
4780 	}
4781 
4782 	if (errlist == NULL)
4783 		nvlist_free(errors);
4784 	else
4785 		*errlist = errors;
4786 
4787 	return (rv);
4788 }
4789 
4790 static boolean_t
4791 propval_equals(nvpair_t *p1, nvpair_t *p2)
4792 {
4793 	if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
4794 		/* dsl_prop_get_all_impl() format */
4795 		nvlist_t *attrs;
4796 		VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
4797 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4798 		    &p1) == 0);
4799 	}
4800 
4801 	if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
4802 		nvlist_t *attrs;
4803 		VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
4804 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4805 		    &p2) == 0);
4806 	}
4807 
4808 	if (nvpair_type(p1) != nvpair_type(p2))
4809 		return (B_FALSE);
4810 
4811 	if (nvpair_type(p1) == DATA_TYPE_STRING) {
4812 		char *valstr1, *valstr2;
4813 
4814 		VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
4815 		VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
4816 		return (strcmp(valstr1, valstr2) == 0);
4817 	} else {
4818 		uint64_t intval1, intval2;
4819 
4820 		VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4821 		VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4822 		return (intval1 == intval2);
4823 	}
4824 }
4825 
4826 /*
4827  * Remove properties from props if they are not going to change (as determined
4828  * by comparison with origprops). Remove them from origprops as well, since we
4829  * do not need to clear or restore properties that won't change.
4830  */
4831 static void
4832 props_reduce(nvlist_t *props, nvlist_t *origprops)
4833 {
4834 	nvpair_t *pair, *next_pair;
4835 
4836 	if (origprops == NULL)
4837 		return; /* all props need to be received */
4838 
4839 	pair = nvlist_next_nvpair(props, NULL);
4840 	while (pair != NULL) {
4841 		const char *propname = nvpair_name(pair);
4842 		nvpair_t *match;
4843 
4844 		next_pair = nvlist_next_nvpair(props, pair);
4845 
4846 		if ((nvlist_lookup_nvpair(origprops, propname,
4847 		    &match) != 0) || !propval_equals(pair, match))
4848 			goto next; /* need to set received value */
4849 
4850 		/* don't clear the existing received value */
4851 		(void) nvlist_remove_nvpair(origprops, match);
4852 		/* don't bother receiving the property */
4853 		(void) nvlist_remove_nvpair(props, pair);
4854 next:
4855 		pair = next_pair;
4856 	}
4857 }
4858 
4859 /*
4860  * Extract properties that cannot be set PRIOR to the receipt of a dataset.
4861  * For example, refquota cannot be set until after the receipt of a dataset,
4862  * because in replication streams, an older/earlier snapshot may exceed the
4863  * refquota.  We want to receive the older/earlier snapshot, but setting
4864  * refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent
4865  * the older/earlier snapshot from being received (with EDQUOT).
4866  *
4867  * The ZFS test "zfs_receive_011_pos" demonstrates such a scenario.
4868  *
4869  * libzfs will need to be judicious handling errors encountered by props
4870  * extracted by this function.
4871  */
4872 static nvlist_t *
4873 extract_delay_props(nvlist_t *props)
4874 {
4875 	nvlist_t *delayprops;
4876 	nvpair_t *nvp, *tmp;
4877 	static const zfs_prop_t delayable[] = {
4878 		ZFS_PROP_REFQUOTA,
4879 		ZFS_PROP_KEYLOCATION,
4880 		0
4881 	};
4882 	int i;
4883 
4884 	VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4885 
4886 	for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL;
4887 	    nvp = nvlist_next_nvpair(props, nvp)) {
4888 		/*
4889 		 * strcmp() is safe because zfs_prop_to_name() always returns
4890 		 * a bounded string.
4891 		 */
4892 		for (i = 0; delayable[i] != 0; i++) {
4893 			if (strcmp(zfs_prop_to_name(delayable[i]),
4894 			    nvpair_name(nvp)) == 0) {
4895 				break;
4896 			}
4897 		}
4898 		if (delayable[i] != 0) {
4899 			tmp = nvlist_prev_nvpair(props, nvp);
4900 			VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0);
4901 			VERIFY(nvlist_remove_nvpair(props, nvp) == 0);
4902 			nvp = tmp;
4903 		}
4904 	}
4905 
4906 	if (nvlist_empty(delayprops)) {
4907 		nvlist_free(delayprops);
4908 		delayprops = NULL;
4909 	}
4910 	return (delayprops);
4911 }
4912 
4913 static void
4914 zfs_allow_log_destroy(void *arg)
4915 {
4916 	char *poolname = arg;
4917 
4918 	if (poolname != NULL)
4919 		kmem_strfree(poolname);
4920 }
4921 
4922 #ifdef	ZFS_DEBUG
4923 static boolean_t zfs_ioc_recv_inject_err;
4924 #endif
4925 
4926 /*
4927  * nvlist 'errors' is always allocated. It will contain descriptions of
4928  * encountered errors, if any. It's the callers responsibility to free.
4929  */
4930 static int
4931 zfs_ioc_recv_impl(char *tofs, char *tosnap, char *origin, nvlist_t *recvprops,
4932     nvlist_t *localprops, nvlist_t *hidden_args, boolean_t force,
4933     boolean_t resumable, int input_fd,
4934     dmu_replay_record_t *begin_record, uint64_t *read_bytes,
4935     uint64_t *errflags, nvlist_t **errors)
4936 {
4937 	dmu_recv_cookie_t drc;
4938 	int error = 0;
4939 	int props_error = 0;
4940 	offset_t off, noff;
4941 	nvlist_t *local_delayprops = NULL;
4942 	nvlist_t *recv_delayprops = NULL;
4943 	nvlist_t *origprops = NULL; /* existing properties */
4944 	nvlist_t *origrecvd = NULL; /* existing received properties */
4945 	boolean_t first_recvd_props = B_FALSE;
4946 	boolean_t tofs_was_redacted;
4947 	zfs_file_t *input_fp;
4948 
4949 	*read_bytes = 0;
4950 	*errflags = 0;
4951 	*errors = fnvlist_alloc();
4952 	off = 0;
4953 
4954 	if ((input_fp = zfs_file_get(input_fd)) == NULL)
4955 		return (SET_ERROR(EBADF));
4956 
4957 	noff = off = zfs_file_off(input_fp);
4958 	error = dmu_recv_begin(tofs, tosnap, begin_record, force,
4959 	    resumable, localprops, hidden_args, origin, &drc, input_fp,
4960 	    &off);
4961 	if (error != 0)
4962 		goto out;
4963 	tofs_was_redacted = dsl_get_redacted(drc.drc_ds);
4964 
4965 	/*
4966 	 * Set properties before we receive the stream so that they are applied
4967 	 * to the new data. Note that we must call dmu_recv_stream() if
4968 	 * dmu_recv_begin() succeeds.
4969 	 */
4970 	if (recvprops != NULL && !drc.drc_newfs) {
4971 		if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
4972 		    SPA_VERSION_RECVD_PROPS &&
4973 		    !dsl_prop_get_hasrecvd(tofs))
4974 			first_recvd_props = B_TRUE;
4975 
4976 		/*
4977 		 * If new received properties are supplied, they are to
4978 		 * completely replace the existing received properties,
4979 		 * so stash away the existing ones.
4980 		 */
4981 		if (dsl_prop_get_received(tofs, &origrecvd) == 0) {
4982 			nvlist_t *errlist = NULL;
4983 			/*
4984 			 * Don't bother writing a property if its value won't
4985 			 * change (and avoid the unnecessary security checks).
4986 			 *
4987 			 * The first receive after SPA_VERSION_RECVD_PROPS is a
4988 			 * special case where we blow away all local properties
4989 			 * regardless.
4990 			 */
4991 			if (!first_recvd_props)
4992 				props_reduce(recvprops, origrecvd);
4993 			if (zfs_check_clearable(tofs, origrecvd, &errlist) != 0)
4994 				(void) nvlist_merge(*errors, errlist, 0);
4995 			nvlist_free(errlist);
4996 
4997 			if (clear_received_props(tofs, origrecvd,
4998 			    first_recvd_props ? NULL : recvprops) != 0)
4999 				*errflags |= ZPROP_ERR_NOCLEAR;
5000 		} else {
5001 			*errflags |= ZPROP_ERR_NOCLEAR;
5002 		}
5003 	}
5004 
5005 	/*
5006 	 * Stash away existing properties so we can restore them on error unless
5007 	 * we're doing the first receive after SPA_VERSION_RECVD_PROPS, in which
5008 	 * case "origrecvd" will take care of that.
5009 	 */
5010 	if (localprops != NULL && !drc.drc_newfs && !first_recvd_props) {
5011 		objset_t *os;
5012 		if (dmu_objset_hold(tofs, FTAG, &os) == 0) {
5013 			if (dsl_prop_get_all(os, &origprops) != 0) {
5014 				*errflags |= ZPROP_ERR_NOCLEAR;
5015 			}
5016 			dmu_objset_rele(os, FTAG);
5017 		} else {
5018 			*errflags |= ZPROP_ERR_NOCLEAR;
5019 		}
5020 	}
5021 
5022 	if (recvprops != NULL) {
5023 		props_error = dsl_prop_set_hasrecvd(tofs);
5024 
5025 		if (props_error == 0) {
5026 			recv_delayprops = extract_delay_props(recvprops);
5027 			(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
5028 			    recvprops, *errors);
5029 		}
5030 	}
5031 
5032 	if (localprops != NULL) {
5033 		nvlist_t *oprops = fnvlist_alloc();
5034 		nvlist_t *xprops = fnvlist_alloc();
5035 		nvpair_t *nvp = NULL;
5036 
5037 		while ((nvp = nvlist_next_nvpair(localprops, nvp)) != NULL) {
5038 			if (nvpair_type(nvp) == DATA_TYPE_BOOLEAN) {
5039 				/* -x property */
5040 				const char *name = nvpair_name(nvp);
5041 				zfs_prop_t prop = zfs_name_to_prop(name);
5042 				if (prop != ZPROP_INVAL) {
5043 					if (!zfs_prop_inheritable(prop))
5044 						continue;
5045 				} else if (!zfs_prop_user(name))
5046 					continue;
5047 				fnvlist_add_boolean(xprops, name);
5048 			} else {
5049 				/* -o property=value */
5050 				fnvlist_add_nvpair(oprops, nvp);
5051 			}
5052 		}
5053 
5054 		local_delayprops = extract_delay_props(oprops);
5055 		(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL,
5056 		    oprops, *errors);
5057 		(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_INHERITED,
5058 		    xprops, *errors);
5059 
5060 		nvlist_free(oprops);
5061 		nvlist_free(xprops);
5062 	}
5063 
5064 	error = dmu_recv_stream(&drc, &off);
5065 
5066 	if (error == 0) {
5067 		zfsvfs_t *zfsvfs = NULL;
5068 		zvol_state_handle_t *zv = NULL;
5069 
5070 		if (getzfsvfs(tofs, &zfsvfs) == 0) {
5071 			/* online recv */
5072 			dsl_dataset_t *ds;
5073 			int end_err;
5074 			boolean_t stream_is_redacted = DMU_GET_FEATUREFLAGS(
5075 			    begin_record->drr_u.drr_begin.
5076 			    drr_versioninfo) & DMU_BACKUP_FEATURE_REDACTED;
5077 
5078 			ds = dmu_objset_ds(zfsvfs->z_os);
5079 			error = zfs_suspend_fs(zfsvfs);
5080 			/*
5081 			 * If the suspend fails, then the recv_end will
5082 			 * likely also fail, and clean up after itself.
5083 			 */
5084 			end_err = dmu_recv_end(&drc, zfsvfs);
5085 			/*
5086 			 * If the dataset was not redacted, but we received a
5087 			 * redacted stream onto it, we need to unmount the
5088 			 * dataset.  Otherwise, resume the filesystem.
5089 			 */
5090 			if (error == 0 && !drc.drc_newfs &&
5091 			    stream_is_redacted && !tofs_was_redacted) {
5092 				error = zfs_end_fs(zfsvfs, ds);
5093 			} else if (error == 0) {
5094 				error = zfs_resume_fs(zfsvfs, ds);
5095 			}
5096 			error = error ? error : end_err;
5097 			zfs_vfs_rele(zfsvfs);
5098 		} else if ((zv = zvol_suspend(tofs)) != NULL) {
5099 			error = dmu_recv_end(&drc, zvol_tag(zv));
5100 			zvol_resume(zv);
5101 		} else {
5102 			error = dmu_recv_end(&drc, NULL);
5103 		}
5104 
5105 		/* Set delayed properties now, after we're done receiving. */
5106 		if (recv_delayprops != NULL && error == 0) {
5107 			(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
5108 			    recv_delayprops, *errors);
5109 		}
5110 		if (local_delayprops != NULL && error == 0) {
5111 			(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL,
5112 			    local_delayprops, *errors);
5113 		}
5114 	}
5115 
5116 	/*
5117 	 * Merge delayed props back in with initial props, in case
5118 	 * we're DEBUG and zfs_ioc_recv_inject_err is set (which means
5119 	 * we have to make sure clear_received_props() includes
5120 	 * the delayed properties).
5121 	 *
5122 	 * Since zfs_ioc_recv_inject_err is only in DEBUG kernels,
5123 	 * using ASSERT() will be just like a VERIFY.
5124 	 */
5125 	if (recv_delayprops != NULL) {
5126 		ASSERT(nvlist_merge(recvprops, recv_delayprops, 0) == 0);
5127 		nvlist_free(recv_delayprops);
5128 	}
5129 	if (local_delayprops != NULL) {
5130 		ASSERT(nvlist_merge(localprops, local_delayprops, 0) == 0);
5131 		nvlist_free(local_delayprops);
5132 	}
5133 	*read_bytes = off - noff;
5134 
5135 #ifdef	ZFS_DEBUG
5136 	if (zfs_ioc_recv_inject_err) {
5137 		zfs_ioc_recv_inject_err = B_FALSE;
5138 		error = 1;
5139 	}
5140 #endif
5141 
5142 	/*
5143 	 * On error, restore the original props.
5144 	 */
5145 	if (error != 0 && recvprops != NULL && !drc.drc_newfs) {
5146 		if (clear_received_props(tofs, recvprops, NULL) != 0) {
5147 			/*
5148 			 * We failed to clear the received properties.
5149 			 * Since we may have left a $recvd value on the
5150 			 * system, we can't clear the $hasrecvd flag.
5151 			 */
5152 			*errflags |= ZPROP_ERR_NORESTORE;
5153 		} else if (first_recvd_props) {
5154 			dsl_prop_unset_hasrecvd(tofs);
5155 		}
5156 
5157 		if (origrecvd == NULL && !drc.drc_newfs) {
5158 			/* We failed to stash the original properties. */
5159 			*errflags |= ZPROP_ERR_NORESTORE;
5160 		}
5161 
5162 		/*
5163 		 * dsl_props_set() will not convert RECEIVED to LOCAL on or
5164 		 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
5165 		 * explicitly if we're restoring local properties cleared in the
5166 		 * first new-style receive.
5167 		 */
5168 		if (origrecvd != NULL &&
5169 		    zfs_set_prop_nvlist(tofs, (first_recvd_props ?
5170 		    ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
5171 		    origrecvd, NULL) != 0) {
5172 			/*
5173 			 * We stashed the original properties but failed to
5174 			 * restore them.
5175 			 */
5176 			*errflags |= ZPROP_ERR_NORESTORE;
5177 		}
5178 	}
5179 	if (error != 0 && localprops != NULL && !drc.drc_newfs &&
5180 	    !first_recvd_props) {
5181 		nvlist_t *setprops;
5182 		nvlist_t *inheritprops;
5183 		nvpair_t *nvp;
5184 
5185 		if (origprops == NULL) {
5186 			/* We failed to stash the original properties. */
5187 			*errflags |= ZPROP_ERR_NORESTORE;
5188 			goto out;
5189 		}
5190 
5191 		/* Restore original props */
5192 		setprops = fnvlist_alloc();
5193 		inheritprops = fnvlist_alloc();
5194 		nvp = NULL;
5195 		while ((nvp = nvlist_next_nvpair(localprops, nvp)) != NULL) {
5196 			const char *name = nvpair_name(nvp);
5197 			const char *source;
5198 			nvlist_t *attrs;
5199 
5200 			if (!nvlist_exists(origprops, name)) {
5201 				/*
5202 				 * Property was not present or was explicitly
5203 				 * inherited before the receive, restore this.
5204 				 */
5205 				fnvlist_add_boolean(inheritprops, name);
5206 				continue;
5207 			}
5208 			attrs = fnvlist_lookup_nvlist(origprops, name);
5209 			source = fnvlist_lookup_string(attrs, ZPROP_SOURCE);
5210 
5211 			/* Skip received properties */
5212 			if (strcmp(source, ZPROP_SOURCE_VAL_RECVD) == 0)
5213 				continue;
5214 
5215 			if (strcmp(source, tofs) == 0) {
5216 				/* Property was locally set */
5217 				fnvlist_add_nvlist(setprops, name, attrs);
5218 			} else {
5219 				/* Property was implicitly inherited */
5220 				fnvlist_add_boolean(inheritprops, name);
5221 			}
5222 		}
5223 
5224 		if (zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL, setprops,
5225 		    NULL) != 0)
5226 			*errflags |= ZPROP_ERR_NORESTORE;
5227 		if (zfs_set_prop_nvlist(tofs, ZPROP_SRC_INHERITED, inheritprops,
5228 		    NULL) != 0)
5229 			*errflags |= ZPROP_ERR_NORESTORE;
5230 
5231 		nvlist_free(setprops);
5232 		nvlist_free(inheritprops);
5233 	}
5234 out:
5235 	zfs_file_put(input_fp);
5236 	nvlist_free(origrecvd);
5237 	nvlist_free(origprops);
5238 
5239 	if (error == 0)
5240 		error = props_error;
5241 
5242 	return (error);
5243 }
5244 
5245 /*
5246  * inputs:
5247  * zc_name		name of containing filesystem (unused)
5248  * zc_nvlist_src{_size}	nvlist of properties to apply
5249  * zc_nvlist_conf{_size}	nvlist of properties to exclude
5250  *			(DATA_TYPE_BOOLEAN) and override (everything else)
5251  * zc_value		name of snapshot to create
5252  * zc_string		name of clone origin (if DRR_FLAG_CLONE)
5253  * zc_cookie		file descriptor to recv from
5254  * zc_begin_record	the BEGIN record of the stream (not byteswapped)
5255  * zc_guid		force flag
5256  *
5257  * outputs:
5258  * zc_cookie		number of bytes read
5259  * zc_obj		zprop_errflags_t
5260  * zc_nvlist_dst{_size} error for each unapplied received property
5261  */
5262 static int
5263 zfs_ioc_recv(zfs_cmd_t *zc)
5264 {
5265 	dmu_replay_record_t begin_record;
5266 	nvlist_t *errors = NULL;
5267 	nvlist_t *recvdprops = NULL;
5268 	nvlist_t *localprops = NULL;
5269 	char *origin = NULL;
5270 	char *tosnap;
5271 	char tofs[ZFS_MAX_DATASET_NAME_LEN];
5272 	int error = 0;
5273 
5274 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
5275 	    strchr(zc->zc_value, '@') == NULL ||
5276 	    strchr(zc->zc_value, '%'))
5277 		return (SET_ERROR(EINVAL));
5278 
5279 	(void) strlcpy(tofs, zc->zc_value, sizeof (tofs));
5280 	tosnap = strchr(tofs, '@');
5281 	*tosnap++ = '\0';
5282 
5283 	if (zc->zc_nvlist_src != 0 &&
5284 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
5285 	    zc->zc_iflags, &recvdprops)) != 0)
5286 		return (error);
5287 
5288 	if (zc->zc_nvlist_conf != 0 &&
5289 	    (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
5290 	    zc->zc_iflags, &localprops)) != 0)
5291 		return (error);
5292 
5293 	if (zc->zc_string[0])
5294 		origin = zc->zc_string;
5295 
5296 	begin_record.drr_type = DRR_BEGIN;
5297 	begin_record.drr_payloadlen = 0;
5298 	begin_record.drr_u.drr_begin = zc->zc_begin_record;
5299 
5300 	error = zfs_ioc_recv_impl(tofs, tosnap, origin, recvdprops, localprops,
5301 	    NULL, zc->zc_guid, B_FALSE, zc->zc_cookie, &begin_record,
5302 	    &zc->zc_cookie, &zc->zc_obj, &errors);
5303 	nvlist_free(recvdprops);
5304 	nvlist_free(localprops);
5305 
5306 	/*
5307 	 * Now that all props, initial and delayed, are set, report the prop
5308 	 * errors to the caller.
5309 	 */
5310 	if (zc->zc_nvlist_dst_size != 0 && errors != NULL &&
5311 	    (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
5312 	    put_nvlist(zc, errors) != 0)) {
5313 		/*
5314 		 * Caller made zc->zc_nvlist_dst less than the minimum expected
5315 		 * size or supplied an invalid address.
5316 		 */
5317 		error = SET_ERROR(EINVAL);
5318 	}
5319 
5320 	nvlist_free(errors);
5321 
5322 	return (error);
5323 }
5324 
5325 /*
5326  * innvl: {
5327  *     "snapname" -> full name of the snapshot to create
5328  *     (optional) "props" -> received properties to set (nvlist)
5329  *     (optional) "localprops" -> override and exclude properties (nvlist)
5330  *     (optional) "origin" -> name of clone origin (DRR_FLAG_CLONE)
5331  *     "begin_record" -> non-byteswapped dmu_replay_record_t
5332  *     "input_fd" -> file descriptor to read stream from (int32)
5333  *     (optional) "force" -> force flag (value ignored)
5334  *     (optional) "resumable" -> resumable flag (value ignored)
5335  *     (optional) "cleanup_fd" -> unused
5336  *     (optional) "action_handle" -> unused
5337  *     (optional) "hidden_args" -> { "wkeydata" -> value }
5338  * }
5339  *
5340  * outnvl: {
5341  *     "read_bytes" -> number of bytes read
5342  *     "error_flags" -> zprop_errflags_t
5343  *     "errors" -> error for each unapplied received property (nvlist)
5344  * }
5345  */
5346 static const zfs_ioc_key_t zfs_keys_recv_new[] = {
5347 	{"snapname",		DATA_TYPE_STRING,	0},
5348 	{"props",		DATA_TYPE_NVLIST,	ZK_OPTIONAL},
5349 	{"localprops",		DATA_TYPE_NVLIST,	ZK_OPTIONAL},
5350 	{"origin",		DATA_TYPE_STRING,	ZK_OPTIONAL},
5351 	{"begin_record",	DATA_TYPE_BYTE_ARRAY,	0},
5352 	{"input_fd",		DATA_TYPE_INT32,	0},
5353 	{"force",		DATA_TYPE_BOOLEAN,	ZK_OPTIONAL},
5354 	{"resumable",		DATA_TYPE_BOOLEAN,	ZK_OPTIONAL},
5355 	{"cleanup_fd",		DATA_TYPE_INT32,	ZK_OPTIONAL},
5356 	{"action_handle",	DATA_TYPE_UINT64,	ZK_OPTIONAL},
5357 	{"hidden_args",		DATA_TYPE_NVLIST,	ZK_OPTIONAL},
5358 };
5359 
5360 static int
5361 zfs_ioc_recv_new(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
5362 {
5363 	dmu_replay_record_t *begin_record;
5364 	uint_t begin_record_size;
5365 	nvlist_t *errors = NULL;
5366 	nvlist_t *recvprops = NULL;
5367 	nvlist_t *localprops = NULL;
5368 	nvlist_t *hidden_args = NULL;
5369 	char *snapname;
5370 	char *origin = NULL;
5371 	char *tosnap;
5372 	char tofs[ZFS_MAX_DATASET_NAME_LEN];
5373 	boolean_t force;
5374 	boolean_t resumable;
5375 	uint64_t read_bytes = 0;
5376 	uint64_t errflags = 0;
5377 	int input_fd = -1;
5378 	int error;
5379 
5380 	snapname = fnvlist_lookup_string(innvl, "snapname");
5381 
5382 	if (dataset_namecheck(snapname, NULL, NULL) != 0 ||
5383 	    strchr(snapname, '@') == NULL ||
5384 	    strchr(snapname, '%'))
5385 		return (SET_ERROR(EINVAL));
5386 
5387 	(void) strlcpy(tofs, snapname, sizeof (tofs));
5388 	tosnap = strchr(tofs, '@');
5389 	*tosnap++ = '\0';
5390 
5391 	error = nvlist_lookup_string(innvl, "origin", &origin);
5392 	if (error && error != ENOENT)
5393 		return (error);
5394 
5395 	error = nvlist_lookup_byte_array(innvl, "begin_record",
5396 	    (uchar_t **)&begin_record, &begin_record_size);
5397 	if (error != 0 || begin_record_size != sizeof (*begin_record))
5398 		return (SET_ERROR(EINVAL));
5399 
5400 	input_fd = fnvlist_lookup_int32(innvl, "input_fd");
5401 
5402 	force = nvlist_exists(innvl, "force");
5403 	resumable = nvlist_exists(innvl, "resumable");
5404 
5405 	/* we still use "props" here for backwards compatibility */
5406 	error = nvlist_lookup_nvlist(innvl, "props", &recvprops);
5407 	if (error && error != ENOENT)
5408 		return (error);
5409 
5410 	error = nvlist_lookup_nvlist(innvl, "localprops", &localprops);
5411 	if (error && error != ENOENT)
5412 		return (error);
5413 
5414 	error = nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);
5415 	if (error && error != ENOENT)
5416 		return (error);
5417 
5418 	error = zfs_ioc_recv_impl(tofs, tosnap, origin, recvprops, localprops,
5419 	    hidden_args, force, resumable, input_fd, begin_record,
5420 	    &read_bytes, &errflags, &errors);
5421 
5422 	fnvlist_add_uint64(outnvl, "read_bytes", read_bytes);
5423 	fnvlist_add_uint64(outnvl, "error_flags", errflags);
5424 	fnvlist_add_nvlist(outnvl, "errors", errors);
5425 
5426 	nvlist_free(errors);
5427 	nvlist_free(recvprops);
5428 	nvlist_free(localprops);
5429 
5430 	return (error);
5431 }
5432 
5433 typedef struct dump_bytes_io {
5434 	zfs_file_t	*dbi_fp;
5435 	caddr_t		dbi_buf;
5436 	int		dbi_len;
5437 	int		dbi_err;
5438 } dump_bytes_io_t;
5439 
5440 static void
5441 dump_bytes_cb(void *arg)
5442 {
5443 	dump_bytes_io_t *dbi = (dump_bytes_io_t *)arg;
5444 	zfs_file_t *fp;
5445 	caddr_t buf;
5446 
5447 	fp = dbi->dbi_fp;
5448 	buf = dbi->dbi_buf;
5449 
5450 	dbi->dbi_err = zfs_file_write(fp, buf, dbi->dbi_len, NULL);
5451 }
5452 
5453 static int
5454 dump_bytes(objset_t *os, void *buf, int len, void *arg)
5455 {
5456 	dump_bytes_io_t dbi;
5457 
5458 	dbi.dbi_fp = arg;
5459 	dbi.dbi_buf = buf;
5460 	dbi.dbi_len = len;
5461 
5462 #if defined(HAVE_LARGE_STACKS)
5463 	dump_bytes_cb(&dbi);
5464 #else
5465 	/*
5466 	 * The vn_rdwr() call is performed in a taskq to ensure that there is
5467 	 * always enough stack space to write safely to the target filesystem.
5468 	 * The ZIO_TYPE_FREE threads are used because there can be a lot of
5469 	 * them and they are used in vdev_file.c for a similar purpose.
5470 	 */
5471 	spa_taskq_dispatch_sync(dmu_objset_spa(os), ZIO_TYPE_FREE,
5472 	    ZIO_TASKQ_ISSUE, dump_bytes_cb, &dbi, TQ_SLEEP);
5473 #endif /* HAVE_LARGE_STACKS */
5474 
5475 	return (dbi.dbi_err);
5476 }
5477 
5478 /*
5479  * inputs:
5480  * zc_name	name of snapshot to send
5481  * zc_cookie	file descriptor to send stream to
5482  * zc_obj	fromorigin flag (mutually exclusive with zc_fromobj)
5483  * zc_sendobj	objsetid of snapshot to send
5484  * zc_fromobj	objsetid of incremental fromsnap (may be zero)
5485  * zc_guid	if set, estimate size of stream only.  zc_cookie is ignored.
5486  *		output size in zc_objset_type.
5487  * zc_flags	lzc_send_flags
5488  *
5489  * outputs:
5490  * zc_objset_type	estimated size, if zc_guid is set
5491  *
5492  * NOTE: This is no longer the preferred interface, any new functionality
5493  *	  should be added to zfs_ioc_send_new() instead.
5494  */
5495 static int
5496 zfs_ioc_send(zfs_cmd_t *zc)
5497 {
5498 	int error;
5499 	offset_t off;
5500 	boolean_t estimate = (zc->zc_guid != 0);
5501 	boolean_t embedok = (zc->zc_flags & 0x1);
5502 	boolean_t large_block_ok = (zc->zc_flags & 0x2);
5503 	boolean_t compressok = (zc->zc_flags & 0x4);
5504 	boolean_t rawok = (zc->zc_flags & 0x8);
5505 	boolean_t savedok = (zc->zc_flags & 0x10);
5506 
5507 	if (zc->zc_obj != 0) {
5508 		dsl_pool_t *dp;
5509 		dsl_dataset_t *tosnap;
5510 
5511 		error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5512 		if (error != 0)
5513 			return (error);
5514 
5515 		error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
5516 		if (error != 0) {
5517 			dsl_pool_rele(dp, FTAG);
5518 			return (error);
5519 		}
5520 
5521 		if (dsl_dir_is_clone(tosnap->ds_dir))
5522 			zc->zc_fromobj =
5523 			    dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj;
5524 		dsl_dataset_rele(tosnap, FTAG);
5525 		dsl_pool_rele(dp, FTAG);
5526 	}
5527 
5528 	if (estimate) {
5529 		dsl_pool_t *dp;
5530 		dsl_dataset_t *tosnap;
5531 		dsl_dataset_t *fromsnap = NULL;
5532 
5533 		error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5534 		if (error != 0)
5535 			return (error);
5536 
5537 		error = dsl_dataset_hold_obj(dp, zc->zc_sendobj,
5538 		    FTAG, &tosnap);
5539 		if (error != 0) {
5540 			dsl_pool_rele(dp, FTAG);
5541 			return (error);
5542 		}
5543 
5544 		if (zc->zc_fromobj != 0) {
5545 			error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
5546 			    FTAG, &fromsnap);
5547 			if (error != 0) {
5548 				dsl_dataset_rele(tosnap, FTAG);
5549 				dsl_pool_rele(dp, FTAG);
5550 				return (error);
5551 			}
5552 		}
5553 
5554 		error = dmu_send_estimate_fast(tosnap, fromsnap, NULL,
5555 		    compressok || rawok, savedok, &zc->zc_objset_type);
5556 
5557 		if (fromsnap != NULL)
5558 			dsl_dataset_rele(fromsnap, FTAG);
5559 		dsl_dataset_rele(tosnap, FTAG);
5560 		dsl_pool_rele(dp, FTAG);
5561 	} else {
5562 		zfs_file_t *fp;
5563 		dmu_send_outparams_t out = {0};
5564 
5565 		if ((fp = zfs_file_get(zc->zc_cookie)) == NULL)
5566 			return (SET_ERROR(EBADF));
5567 
5568 		off = zfs_file_off(fp);
5569 		out.dso_outfunc = dump_bytes;
5570 		out.dso_arg = fp;
5571 		out.dso_dryrun = B_FALSE;
5572 		error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
5573 		    zc->zc_fromobj, embedok, large_block_ok, compressok,
5574 		    rawok, savedok, zc->zc_cookie, &off, &out);
5575 
5576 		zfs_file_put(fp);
5577 	}
5578 	return (error);
5579 }
5580 
5581 /*
5582  * inputs:
5583  * zc_name		name of snapshot on which to report progress
5584  * zc_cookie		file descriptor of send stream
5585  *
5586  * outputs:
5587  * zc_cookie		number of bytes written in send stream thus far
5588  * zc_objset_type	logical size of data traversed by send thus far
5589  */
5590 static int
5591 zfs_ioc_send_progress(zfs_cmd_t *zc)
5592 {
5593 	dsl_pool_t *dp;
5594 	dsl_dataset_t *ds;
5595 	dmu_sendstatus_t *dsp = NULL;
5596 	int error;
5597 
5598 	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5599 	if (error != 0)
5600 		return (error);
5601 
5602 	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
5603 	if (error != 0) {
5604 		dsl_pool_rele(dp, FTAG);
5605 		return (error);
5606 	}
5607 
5608 	mutex_enter(&ds->ds_sendstream_lock);
5609 
5610 	/*
5611 	 * Iterate over all the send streams currently active on this dataset.
5612 	 * If there's one which matches the specified file descriptor _and_ the
5613 	 * stream was started by the current process, return the progress of
5614 	 * that stream.
5615 	 */
5616 
5617 	for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
5618 	    dsp = list_next(&ds->ds_sendstreams, dsp)) {
5619 		if (dsp->dss_outfd == zc->zc_cookie &&
5620 		    zfs_proc_is_caller(dsp->dss_proc))
5621 			break;
5622 	}
5623 
5624 	if (dsp != NULL) {
5625 		zc->zc_cookie = atomic_cas_64((volatile uint64_t *)dsp->dss_off,
5626 		    0, 0);
5627 		/* This is the closest thing we have to atomic_read_64. */
5628 		zc->zc_objset_type = atomic_cas_64(&dsp->dss_blocks, 0, 0);
5629 	} else {
5630 		error = SET_ERROR(ENOENT);
5631 	}
5632 
5633 	mutex_exit(&ds->ds_sendstream_lock);
5634 	dsl_dataset_rele(ds, FTAG);
5635 	dsl_pool_rele(dp, FTAG);
5636 	return (error);
5637 }
5638 
5639 static int
5640 zfs_ioc_inject_fault(zfs_cmd_t *zc)
5641 {
5642 	int id, error;
5643 
5644 	error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
5645 	    &zc->zc_inject_record);
5646 
5647 	if (error == 0)
5648 		zc->zc_guid = (uint64_t)id;
5649 
5650 	return (error);
5651 }
5652 
5653 static int
5654 zfs_ioc_clear_fault(zfs_cmd_t *zc)
5655 {
5656 	return (zio_clear_fault((int)zc->zc_guid));
5657 }
5658 
5659 static int
5660 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
5661 {
5662 	int id = (int)zc->zc_guid;
5663 	int error;
5664 
5665 	error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
5666 	    &zc->zc_inject_record);
5667 
5668 	zc->zc_guid = id;
5669 
5670 	return (error);
5671 }
5672 
5673 static int
5674 zfs_ioc_error_log(zfs_cmd_t *zc)
5675 {
5676 	spa_t *spa;
5677 	int error;
5678 	size_t count = (size_t)zc->zc_nvlist_dst_size;
5679 
5680 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
5681 		return (error);
5682 
5683 	error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
5684 	    &count);
5685 	if (error == 0)
5686 		zc->zc_nvlist_dst_size = count;
5687 	else
5688 		zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
5689 
5690 	spa_close(spa, FTAG);
5691 
5692 	return (error);
5693 }
5694 
5695 static int
5696 zfs_ioc_clear(zfs_cmd_t *zc)
5697 {
5698 	spa_t *spa;
5699 	vdev_t *vd;
5700 	int error;
5701 
5702 	/*
5703 	 * On zpool clear we also fix up missing slogs
5704 	 */
5705 	mutex_enter(&spa_namespace_lock);
5706 	spa = spa_lookup(zc->zc_name);
5707 	if (spa == NULL) {
5708 		mutex_exit(&spa_namespace_lock);
5709 		return (SET_ERROR(EIO));
5710 	}
5711 	if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
5712 		/* we need to let spa_open/spa_load clear the chains */
5713 		spa_set_log_state(spa, SPA_LOG_CLEAR);
5714 	}
5715 	spa->spa_last_open_failed = 0;
5716 	mutex_exit(&spa_namespace_lock);
5717 
5718 	if (zc->zc_cookie & ZPOOL_NO_REWIND) {
5719 		error = spa_open(zc->zc_name, &spa, FTAG);
5720 	} else {
5721 		nvlist_t *policy;
5722 		nvlist_t *config = NULL;
5723 
5724 		if (zc->zc_nvlist_src == 0)
5725 			return (SET_ERROR(EINVAL));
5726 
5727 		if ((error = get_nvlist(zc->zc_nvlist_src,
5728 		    zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
5729 			error = spa_open_rewind(zc->zc_name, &spa, FTAG,
5730 			    policy, &config);
5731 			if (config != NULL) {
5732 				int err;
5733 
5734 				if ((err = put_nvlist(zc, config)) != 0)
5735 					error = err;
5736 				nvlist_free(config);
5737 			}
5738 			nvlist_free(policy);
5739 		}
5740 	}
5741 
5742 	if (error != 0)
5743 		return (error);
5744 
5745 	/*
5746 	 * If multihost is enabled, resuming I/O is unsafe as another
5747 	 * host may have imported the pool.
5748 	 */
5749 	if (spa_multihost(spa) && spa_suspended(spa))
5750 		return (SET_ERROR(EINVAL));
5751 
5752 	spa_vdev_state_enter(spa, SCL_NONE);
5753 
5754 	if (zc->zc_guid == 0) {
5755 		vd = NULL;
5756 	} else {
5757 		vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
5758 		if (vd == NULL) {
5759 			error = SET_ERROR(ENODEV);
5760 			(void) spa_vdev_state_exit(spa, NULL, error);
5761 			spa_close(spa, FTAG);
5762 			return (error);
5763 		}
5764 	}
5765 
5766 	vdev_clear(spa, vd);
5767 
5768 	(void) spa_vdev_state_exit(spa, spa_suspended(spa) ?
5769 	    NULL : spa->spa_root_vdev, 0);
5770 
5771 	/*
5772 	 * Resume any suspended I/Os.
5773 	 */
5774 	if (zio_resume(spa) != 0)
5775 		error = SET_ERROR(EIO);
5776 
5777 	spa_close(spa, FTAG);
5778 
5779 	return (error);
5780 }
5781 
5782 /*
5783  * Reopen all the vdevs associated with the pool.
5784  *
5785  * innvl: {
5786  *  "scrub_restart" -> when true and scrub is running, allow to restart
5787  *              scrub as the side effect of the reopen (boolean).
5788  * }
5789  *
5790  * outnvl is unused
5791  */
5792 static const zfs_ioc_key_t zfs_keys_pool_reopen[] = {
5793 	{"scrub_restart",	DATA_TYPE_BOOLEAN_VALUE,	ZK_OPTIONAL},
5794 };
5795 
5796 /* ARGSUSED */
5797 static int
5798 zfs_ioc_pool_reopen(const char *pool, nvlist_t *innvl, nvlist_t *outnvl)
5799 {
5800 	spa_t *spa;
5801 	int error;
5802 	boolean_t rc, scrub_restart = B_TRUE;
5803 
5804 	if (innvl) {
5805 		error = nvlist_lookup_boolean_value(innvl,
5806 		    "scrub_restart", &rc);
5807 		if (error == 0)
5808 			scrub_restart = rc;
5809 	}
5810 
5811 	error = spa_open(pool, &spa, FTAG);
5812 	if (error != 0)
5813 		return (error);
5814 
5815 	spa_vdev_state_enter(spa, SCL_NONE);
5816 
5817 	/*
5818 	 * If the scrub_restart flag is B_FALSE and a scrub is already
5819 	 * in progress then set spa_scrub_reopen flag to B_TRUE so that
5820 	 * we don't restart the scrub as a side effect of the reopen.
5821 	 * Otherwise, let vdev_open() decided if a resilver is required.
5822 	 */
5823 
5824 	spa->spa_scrub_reopen = (!scrub_restart &&
5825 	    dsl_scan_scrubbing(spa->spa_dsl_pool));
5826 	vdev_reopen(spa->spa_root_vdev);
5827 	spa->spa_scrub_reopen = B_FALSE;
5828 
5829 	(void) spa_vdev_state_exit(spa, NULL, 0);
5830 	spa_close(spa, FTAG);
5831 	return (0);
5832 }
5833 
5834 /*
5835  * inputs:
5836  * zc_name	name of filesystem
5837  *
5838  * outputs:
5839  * zc_string	name of conflicting snapshot, if there is one
5840  */
5841 static int
5842 zfs_ioc_promote(zfs_cmd_t *zc)
5843 {
5844 	dsl_pool_t *dp;
5845 	dsl_dataset_t *ds, *ods;
5846 	char origin[ZFS_MAX_DATASET_NAME_LEN];
5847 	char *cp;
5848 	int error;
5849 
5850 	zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
5851 	if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0 ||
5852 	    strchr(zc->zc_name, '%'))
5853 		return (SET_ERROR(EINVAL));
5854 
5855 	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5856 	if (error != 0)
5857 		return (error);
5858 
5859 	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
5860 	if (error != 0) {
5861 		dsl_pool_rele(dp, FTAG);
5862 		return (error);
5863 	}
5864 
5865 	if (!dsl_dir_is_clone(ds->ds_dir)) {
5866 		dsl_dataset_rele(ds, FTAG);
5867 		dsl_pool_rele(dp, FTAG);
5868 		return (SET_ERROR(EINVAL));
5869 	}
5870 
5871 	error = dsl_dataset_hold_obj(dp,
5872 	    dsl_dir_phys(ds->ds_dir)->dd_origin_obj, FTAG, &ods);
5873 	if (error != 0) {
5874 		dsl_dataset_rele(ds, FTAG);
5875 		dsl_pool_rele(dp, FTAG);
5876 		return (error);
5877 	}
5878 
5879 	dsl_dataset_name(ods, origin);
5880 	dsl_dataset_rele(ods, FTAG);
5881 	dsl_dataset_rele(ds, FTAG);
5882 	dsl_pool_rele(dp, FTAG);
5883 
5884 	/*
5885 	 * We don't need to unmount *all* the origin fs's snapshots, but
5886 	 * it's easier.
5887 	 */
5888 	cp = strchr(origin, '@');
5889 	if (cp)
5890 		*cp = '\0';
5891 	(void) dmu_objset_find(origin,
5892 	    zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
5893 	return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
5894 }
5895 
5896 /*
5897  * Retrieve a single {user|group|project}{used|quota}@... property.
5898  *
5899  * inputs:
5900  * zc_name	name of filesystem
5901  * zc_objset_type zfs_userquota_prop_t
5902  * zc_value	domain name (eg. "S-1-234-567-89")
5903  * zc_guid	RID/UID/GID
5904  *
5905  * outputs:
5906  * zc_cookie	property value
5907  */
5908 static int
5909 zfs_ioc_userspace_one(zfs_cmd_t *zc)
5910 {
5911 	zfsvfs_t *zfsvfs;
5912 	int error;
5913 
5914 	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
5915 		return (SET_ERROR(EINVAL));
5916 
5917 	error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
5918 	if (error != 0)
5919 		return (error);
5920 
5921 	error = zfs_userspace_one(zfsvfs,
5922 	    zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
5923 	zfsvfs_rele(zfsvfs, FTAG);
5924 
5925 	return (error);
5926 }
5927 
5928 /*
5929  * inputs:
5930  * zc_name		name of filesystem
5931  * zc_cookie		zap cursor
5932  * zc_objset_type	zfs_userquota_prop_t
5933  * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
5934  *
5935  * outputs:
5936  * zc_nvlist_dst[_size]	data buffer (array of zfs_useracct_t)
5937  * zc_cookie	zap cursor
5938  */
5939 static int
5940 zfs_ioc_userspace_many(zfs_cmd_t *zc)
5941 {
5942 	zfsvfs_t *zfsvfs;
5943 	int bufsize = zc->zc_nvlist_dst_size;
5944 
5945 	if (bufsize <= 0)
5946 		return (SET_ERROR(ENOMEM));
5947 
5948 	int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
5949 	if (error != 0)
5950 		return (error);
5951 
5952 	void *buf = vmem_alloc(bufsize, KM_SLEEP);
5953 
5954 	error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
5955 	    buf, &zc->zc_nvlist_dst_size);
5956 
5957 	if (error == 0) {
5958 		error = xcopyout(buf,
5959 		    (void *)(uintptr_t)zc->zc_nvlist_dst,
5960 		    zc->zc_nvlist_dst_size);
5961 	}
5962 	vmem_free(buf, bufsize);
5963 	zfsvfs_rele(zfsvfs, FTAG);
5964 
5965 	return (error);
5966 }
5967 
5968 /*
5969  * inputs:
5970  * zc_name		name of filesystem
5971  *
5972  * outputs:
5973  * none
5974  */
5975 static int
5976 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
5977 {
5978 	int error = 0;
5979 	zfsvfs_t *zfsvfs;
5980 
5981 	if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
5982 		if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
5983 			/*
5984 			 * If userused is not enabled, it may be because the
5985 			 * objset needs to be closed & reopened (to grow the
5986 			 * objset_phys_t).  Suspend/resume the fs will do that.
5987 			 */
5988 			dsl_dataset_t *ds, *newds;
5989 
5990 			ds = dmu_objset_ds(zfsvfs->z_os);
5991 			error = zfs_suspend_fs(zfsvfs);
5992 			if (error == 0) {
5993 				dmu_objset_refresh_ownership(ds, &newds,
5994 				    B_TRUE, zfsvfs);
5995 				error = zfs_resume_fs(zfsvfs, newds);
5996 			}
5997 		}
5998 		if (error == 0) {
5999 			mutex_enter(&zfsvfs->z_os->os_upgrade_lock);
6000 			if (zfsvfs->z_os->os_upgrade_id == 0) {
6001 				/* clear potential error code and retry */
6002 				zfsvfs->z_os->os_upgrade_status = 0;
6003 				mutex_exit(&zfsvfs->z_os->os_upgrade_lock);
6004 
6005 				dsl_pool_config_enter(
6006 				    dmu_objset_pool(zfsvfs->z_os), FTAG);
6007 				dmu_objset_userspace_upgrade(zfsvfs->z_os);
6008 				dsl_pool_config_exit(
6009 				    dmu_objset_pool(zfsvfs->z_os), FTAG);
6010 			} else {
6011 				mutex_exit(&zfsvfs->z_os->os_upgrade_lock);
6012 			}
6013 
6014 			taskq_wait_id(zfsvfs->z_os->os_spa->spa_upgrade_taskq,
6015 			    zfsvfs->z_os->os_upgrade_id);
6016 			error = zfsvfs->z_os->os_upgrade_status;
6017 		}
6018 		zfs_vfs_rele(zfsvfs);
6019 	} else {
6020 		objset_t *os;
6021 
6022 		/* XXX kind of reading contents without owning */
6023 		error = dmu_objset_hold_flags(zc->zc_name, B_TRUE, FTAG, &os);
6024 		if (error != 0)
6025 			return (error);
6026 
6027 		mutex_enter(&os->os_upgrade_lock);
6028 		if (os->os_upgrade_id == 0) {
6029 			/* clear potential error code and retry */
6030 			os->os_upgrade_status = 0;
6031 			mutex_exit(&os->os_upgrade_lock);
6032 
6033 			dmu_objset_userspace_upgrade(os);
6034 		} else {
6035 			mutex_exit(&os->os_upgrade_lock);
6036 		}
6037 
6038 		dsl_pool_rele(dmu_objset_pool(os), FTAG);
6039 
6040 		taskq_wait_id(os->os_spa->spa_upgrade_taskq, os->os_upgrade_id);
6041 		error = os->os_upgrade_status;
6042 
6043 		dsl_dataset_rele_flags(dmu_objset_ds(os), DS_HOLD_FLAG_DECRYPT,
6044 		    FTAG);
6045 	}
6046 	return (error);
6047 }
6048 
6049 /*
6050  * inputs:
6051  * zc_name		name of filesystem
6052  *
6053  * outputs:
6054  * none
6055  */
6056 static int
6057 zfs_ioc_id_quota_upgrade(zfs_cmd_t *zc)
6058 {
6059 	objset_t *os;
6060 	int error;
6061 
6062 	error = dmu_objset_hold_flags(zc->zc_name, B_TRUE, FTAG, &os);
6063 	if (error != 0)
6064 		return (error);
6065 
6066 	if (dmu_objset_userobjspace_upgradable(os) ||
6067 	    dmu_objset_projectquota_upgradable(os)) {
6068 		mutex_enter(&os->os_upgrade_lock);
6069 		if (os->os_upgrade_id == 0) {
6070 			/* clear potential error code and retry */
6071 			os->os_upgrade_status = 0;
6072 			mutex_exit(&os->os_upgrade_lock);
6073 
6074 			dmu_objset_id_quota_upgrade(os);
6075 		} else {
6076 			mutex_exit(&os->os_upgrade_lock);
6077 		}
6078 
6079 		dsl_pool_rele(dmu_objset_pool(os), FTAG);
6080 
6081 		taskq_wait_id(os->os_spa->spa_upgrade_taskq, os->os_upgrade_id);
6082 		error = os->os_upgrade_status;
6083 	} else {
6084 		dsl_pool_rele(dmu_objset_pool(os), FTAG);
6085 	}
6086 
6087 	dsl_dataset_rele_flags(dmu_objset_ds(os), DS_HOLD_FLAG_DECRYPT, FTAG);
6088 
6089 	return (error);
6090 }
6091 
6092 static int
6093 zfs_ioc_share(zfs_cmd_t *zc)
6094 {
6095 	return (SET_ERROR(ENOSYS));
6096 }
6097 
6098 ace_t full_access[] = {
6099 	{(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
6100 };
6101 
6102 /*
6103  * inputs:
6104  * zc_name		name of containing filesystem
6105  * zc_obj		object # beyond which we want next in-use object #
6106  *
6107  * outputs:
6108  * zc_obj		next in-use object #
6109  */
6110 static int
6111 zfs_ioc_next_obj(zfs_cmd_t *zc)
6112 {
6113 	objset_t *os = NULL;
6114 	int error;
6115 
6116 	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
6117 	if (error != 0)
6118 		return (error);
6119 
6120 	error = dmu_object_next(os, &zc->zc_obj, B_FALSE, 0);
6121 
6122 	dmu_objset_rele(os, FTAG);
6123 	return (error);
6124 }
6125 
6126 /*
6127  * inputs:
6128  * zc_name		name of filesystem
6129  * zc_value		prefix name for snapshot
6130  * zc_cleanup_fd	cleanup-on-exit file descriptor for calling process
6131  *
6132  * outputs:
6133  * zc_value		short name of new snapshot
6134  */
6135 static int
6136 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
6137 {
6138 	char *snap_name;
6139 	char *hold_name;
6140 	minor_t minor;
6141 
6142 	zfs_file_t *fp = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
6143 	if (fp == NULL)
6144 		return (SET_ERROR(EBADF));
6145 
6146 	snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
6147 	    (u_longlong_t)ddi_get_lbolt64());
6148 	hold_name = kmem_asprintf("%%%s", zc->zc_value);
6149 
6150 	int error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
6151 	    hold_name);
6152 	if (error == 0)
6153 		(void) strlcpy(zc->zc_value, snap_name,
6154 		    sizeof (zc->zc_value));
6155 	kmem_strfree(snap_name);
6156 	kmem_strfree(hold_name);
6157 	zfs_onexit_fd_rele(fp);
6158 	return (error);
6159 }
6160 
6161 /*
6162  * inputs:
6163  * zc_name		name of "to" snapshot
6164  * zc_value		name of "from" snapshot
6165  * zc_cookie		file descriptor to write diff data on
6166  *
6167  * outputs:
6168  * dmu_diff_record_t's to the file descriptor
6169  */
6170 static int
6171 zfs_ioc_diff(zfs_cmd_t *zc)
6172 {
6173 	zfs_file_t *fp;
6174 	offset_t off;
6175 	int error;
6176 
6177 	if ((fp = zfs_file_get(zc->zc_cookie)) == NULL)
6178 		return (SET_ERROR(EBADF));
6179 
6180 	off = zfs_file_off(fp);
6181 	error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off);
6182 
6183 	zfs_file_put(fp);
6184 
6185 	return (error);
6186 }
6187 
6188 static int
6189 zfs_ioc_smb_acl(zfs_cmd_t *zc)
6190 {
6191 	return (SET_ERROR(ENOTSUP));
6192 }
6193 
6194 /*
6195  * innvl: {
6196  *     "holds" -> { snapname -> holdname (string), ... }
6197  *     (optional) "cleanup_fd" -> fd (int32)
6198  * }
6199  *
6200  * outnvl: {
6201  *     snapname -> error value (int32)
6202  *     ...
6203  * }
6204  */
6205 static const zfs_ioc_key_t zfs_keys_hold[] = {
6206 	{"holds",		DATA_TYPE_NVLIST,	0},
6207 	{"cleanup_fd",		DATA_TYPE_INT32,	ZK_OPTIONAL},
6208 };
6209 
6210 /* ARGSUSED */
6211 static int
6212 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
6213 {
6214 	nvpair_t *pair;
6215 	nvlist_t *holds;
6216 	int cleanup_fd = -1;
6217 	int error;
6218 	minor_t minor = 0;
6219 	zfs_file_t *fp = NULL;
6220 
6221 	holds = fnvlist_lookup_nvlist(args, "holds");
6222 
6223 	/* make sure the user didn't pass us any invalid (empty) tags */
6224 	for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
6225 	    pair = nvlist_next_nvpair(holds, pair)) {
6226 		char *htag;
6227 
6228 		error = nvpair_value_string(pair, &htag);
6229 		if (error != 0)
6230 			return (SET_ERROR(error));
6231 
6232 		if (strlen(htag) == 0)
6233 			return (SET_ERROR(EINVAL));
6234 	}
6235 
6236 	if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
6237 		fp = zfs_onexit_fd_hold(cleanup_fd, &minor);
6238 		if (fp == NULL)
6239 			return (SET_ERROR(EBADF));
6240 	}
6241 
6242 	error = dsl_dataset_user_hold(holds, minor, errlist);
6243 	if (fp != NULL) {
6244 		ASSERT3U(minor, !=, 0);
6245 		zfs_onexit_fd_rele(fp);
6246 	}
6247 	return (SET_ERROR(error));
6248 }
6249 
6250 /*
6251  * innvl is not used.
6252  *
6253  * outnvl: {
6254  *    holdname -> time added (uint64 seconds since epoch)
6255  *    ...
6256  * }
6257  */
6258 static const zfs_ioc_key_t zfs_keys_get_holds[] = {
6259 	/* no nvl keys */
6260 };
6261 
6262 /* ARGSUSED */
6263 static int
6264 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
6265 {
6266 	return (dsl_dataset_get_holds(snapname, outnvl));
6267 }
6268 
6269 /*
6270  * innvl: {
6271  *     snapname -> { holdname, ... }
6272  *     ...
6273  * }
6274  *
6275  * outnvl: {
6276  *     snapname -> error value (int32)
6277  *     ...
6278  * }
6279  */
6280 static const zfs_ioc_key_t zfs_keys_release[] = {
6281 	{"<snapname>...",	DATA_TYPE_NVLIST,	ZK_WILDCARDLIST},
6282 };
6283 
6284 /* ARGSUSED */
6285 static int
6286 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
6287 {
6288 	return (dsl_dataset_user_release(holds, errlist));
6289 }
6290 
6291 /*
6292  * inputs:
6293  * zc_guid		flags (ZEVENT_NONBLOCK)
6294  * zc_cleanup_fd	zevent file descriptor
6295  *
6296  * outputs:
6297  * zc_nvlist_dst	next nvlist event
6298  * zc_cookie		dropped events since last get
6299  */
6300 static int
6301 zfs_ioc_events_next(zfs_cmd_t *zc)
6302 {
6303 	zfs_zevent_t *ze;
6304 	nvlist_t *event = NULL;
6305 	minor_t minor;
6306 	uint64_t dropped = 0;
6307 	int error;
6308 
6309 	zfs_file_t *fp = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
6310 	if (fp == NULL)
6311 		return (SET_ERROR(EBADF));
6312 
6313 	do {
6314 		error = zfs_zevent_next(ze, &event,
6315 		    &zc->zc_nvlist_dst_size, &dropped);
6316 		if (event != NULL) {
6317 			zc->zc_cookie = dropped;
6318 			error = put_nvlist(zc, event);
6319 			nvlist_free(event);
6320 		}
6321 
6322 		if (zc->zc_guid & ZEVENT_NONBLOCK)
6323 			break;
6324 
6325 		if ((error == 0) || (error != ENOENT))
6326 			break;
6327 
6328 		error = zfs_zevent_wait(ze);
6329 		if (error != 0)
6330 			break;
6331 	} while (1);
6332 
6333 	zfs_zevent_fd_rele(fp);
6334 
6335 	return (error);
6336 }
6337 
6338 /*
6339  * outputs:
6340  * zc_cookie		cleared events count
6341  */
6342 static int
6343 zfs_ioc_events_clear(zfs_cmd_t *zc)
6344 {
6345 	int count;
6346 
6347 	zfs_zevent_drain_all(&count);
6348 	zc->zc_cookie = count;
6349 
6350 	return (0);
6351 }
6352 
6353 /*
6354  * inputs:
6355  * zc_guid		eid | ZEVENT_SEEK_START | ZEVENT_SEEK_END
6356  * zc_cleanup		zevent file descriptor
6357  */
6358 static int
6359 zfs_ioc_events_seek(zfs_cmd_t *zc)
6360 {
6361 	zfs_zevent_t *ze;
6362 	minor_t minor;
6363 	int error;
6364 
6365 	zfs_file_t *fp = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
6366 	if (fp == NULL)
6367 		return (SET_ERROR(EBADF));
6368 
6369 	error = zfs_zevent_seek(ze, zc->zc_guid);
6370 	zfs_zevent_fd_rele(fp);
6371 
6372 	return (error);
6373 }
6374 
6375 /*
6376  * inputs:
6377  * zc_name		name of later filesystem or snapshot
6378  * zc_value		full name of old snapshot or bookmark
6379  *
6380  * outputs:
6381  * zc_cookie		space in bytes
6382  * zc_objset_type	compressed space in bytes
6383  * zc_perm_action	uncompressed space in bytes
6384  */
6385 static int
6386 zfs_ioc_space_written(zfs_cmd_t *zc)
6387 {
6388 	int error;
6389 	dsl_pool_t *dp;
6390 	dsl_dataset_t *new;
6391 
6392 	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
6393 	if (error != 0)
6394 		return (error);
6395 	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
6396 	if (error != 0) {
6397 		dsl_pool_rele(dp, FTAG);
6398 		return (error);
6399 	}
6400 	if (strchr(zc->zc_value, '#') != NULL) {
6401 		zfs_bookmark_phys_t bmp;
6402 		error = dsl_bookmark_lookup(dp, zc->zc_value,
6403 		    new, &bmp);
6404 		if (error == 0) {
6405 			error = dsl_dataset_space_written_bookmark(&bmp, new,
6406 			    &zc->zc_cookie,
6407 			    &zc->zc_objset_type, &zc->zc_perm_action);
6408 		}
6409 	} else {
6410 		dsl_dataset_t *old;
6411 		error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
6412 
6413 		if (error == 0) {
6414 			error = dsl_dataset_space_written(old, new,
6415 			    &zc->zc_cookie,
6416 			    &zc->zc_objset_type, &zc->zc_perm_action);
6417 			dsl_dataset_rele(old, FTAG);
6418 		}
6419 	}
6420 	dsl_dataset_rele(new, FTAG);
6421 	dsl_pool_rele(dp, FTAG);
6422 	return (error);
6423 }
6424 
6425 /*
6426  * innvl: {
6427  *     "firstsnap" -> snapshot name
6428  * }
6429  *
6430  * outnvl: {
6431  *     "used" -> space in bytes
6432  *     "compressed" -> compressed space in bytes
6433  *     "uncompressed" -> uncompressed space in bytes
6434  * }
6435  */
6436 static const zfs_ioc_key_t zfs_keys_space_snaps[] = {
6437 	{"firstsnap",	DATA_TYPE_STRING,	0},
6438 };
6439 
6440 static int
6441 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
6442 {
6443 	int error;
6444 	dsl_pool_t *dp;
6445 	dsl_dataset_t *new, *old;
6446 	char *firstsnap;
6447 	uint64_t used, comp, uncomp;
6448 
6449 	firstsnap = fnvlist_lookup_string(innvl, "firstsnap");
6450 
6451 	error = dsl_pool_hold(lastsnap, FTAG, &dp);
6452 	if (error != 0)
6453 		return (error);
6454 
6455 	error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
6456 	if (error == 0 && !new->ds_is_snapshot) {
6457 		dsl_dataset_rele(new, FTAG);
6458 		error = SET_ERROR(EINVAL);
6459 	}
6460 	if (error != 0) {
6461 		dsl_pool_rele(dp, FTAG);
6462 		return (error);
6463 	}
6464 	error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
6465 	if (error == 0 && !old->ds_is_snapshot) {
6466 		dsl_dataset_rele(old, FTAG);
6467 		error = SET_ERROR(EINVAL);
6468 	}
6469 	if (error != 0) {
6470 		dsl_dataset_rele(new, FTAG);
6471 		dsl_pool_rele(dp, FTAG);
6472 		return (error);
6473 	}
6474 
6475 	error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
6476 	dsl_dataset_rele(old, FTAG);
6477 	dsl_dataset_rele(new, FTAG);
6478 	dsl_pool_rele(dp, FTAG);
6479 	fnvlist_add_uint64(outnvl, "used", used);
6480 	fnvlist_add_uint64(outnvl, "compressed", comp);
6481 	fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
6482 	return (error);
6483 }
6484 
6485 /*
6486  * innvl: {
6487  *     "fd" -> file descriptor to write stream to (int32)
6488  *     (optional) "fromsnap" -> full snap name to send an incremental from
6489  *     (optional) "largeblockok" -> (value ignored)
6490  *         indicates that blocks > 128KB are permitted
6491  *     (optional) "embedok" -> (value ignored)
6492  *         presence indicates DRR_WRITE_EMBEDDED records are permitted
6493  *     (optional) "compressok" -> (value ignored)
6494  *         presence indicates compressed DRR_WRITE records are permitted
6495  *     (optional) "rawok" -> (value ignored)
6496  *         presence indicates raw encrypted records should be used.
6497  *     (optional) "savedok" -> (value ignored)
6498  *         presence indicates we should send a partially received snapshot
6499  *     (optional) "resume_object" and "resume_offset" -> (uint64)
6500  *         if present, resume send stream from specified object and offset.
6501  *     (optional) "redactbook" -> (string)
6502  *         if present, use this bookmark's redaction list to generate a redacted
6503  *         send stream
6504  * }
6505  *
6506  * outnvl is unused
6507  */
6508 static const zfs_ioc_key_t zfs_keys_send_new[] = {
6509 	{"fd",			DATA_TYPE_INT32,	0},
6510 	{"fromsnap",		DATA_TYPE_STRING,	ZK_OPTIONAL},
6511 	{"largeblockok",	DATA_TYPE_BOOLEAN,	ZK_OPTIONAL},
6512 	{"embedok",		DATA_TYPE_BOOLEAN,	ZK_OPTIONAL},
6513 	{"compressok",		DATA_TYPE_BOOLEAN,	ZK_OPTIONAL},
6514 	{"rawok",		DATA_TYPE_BOOLEAN,	ZK_OPTIONAL},
6515 	{"savedok",		DATA_TYPE_BOOLEAN,	ZK_OPTIONAL},
6516 	{"resume_object",	DATA_TYPE_UINT64,	ZK_OPTIONAL},
6517 	{"resume_offset",	DATA_TYPE_UINT64,	ZK_OPTIONAL},
6518 	{"redactbook",		DATA_TYPE_STRING,	ZK_OPTIONAL},
6519 };
6520 
6521 /* ARGSUSED */
6522 static int
6523 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
6524 {
6525 	int error;
6526 	offset_t off;
6527 	char *fromname = NULL;
6528 	int fd;
6529 	zfs_file_t *fp;
6530 	boolean_t largeblockok;
6531 	boolean_t embedok;
6532 	boolean_t compressok;
6533 	boolean_t rawok;
6534 	boolean_t savedok;
6535 	uint64_t resumeobj = 0;
6536 	uint64_t resumeoff = 0;
6537 	char *redactbook = NULL;
6538 
6539 	fd = fnvlist_lookup_int32(innvl, "fd");
6540 
6541 	(void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
6542 
6543 	largeblockok = nvlist_exists(innvl, "largeblockok");
6544 	embedok = nvlist_exists(innvl, "embedok");
6545 	compressok = nvlist_exists(innvl, "compressok");
6546 	rawok = nvlist_exists(innvl, "rawok");
6547 	savedok = nvlist_exists(innvl, "savedok");
6548 
6549 	(void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
6550 	(void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);
6551 
6552 	(void) nvlist_lookup_string(innvl, "redactbook", &redactbook);
6553 
6554 	if ((fp = zfs_file_get(fd)) == NULL)
6555 		return (SET_ERROR(EBADF));
6556 
6557 	off = zfs_file_off(fp);
6558 
6559 	dmu_send_outparams_t out = {0};
6560 	out.dso_outfunc = dump_bytes;
6561 	out.dso_arg = fp;
6562 	out.dso_dryrun = B_FALSE;
6563 	error = dmu_send(snapname, fromname, embedok, largeblockok,
6564 	    compressok, rawok, savedok, resumeobj, resumeoff,
6565 	    redactbook, fd, &off, &out);
6566 
6567 	zfs_file_put(fp);
6568 	return (error);
6569 }
6570 
6571 /* ARGSUSED */
6572 static int
6573 send_space_sum(objset_t *os, void *buf, int len, void *arg)
6574 {
6575 	uint64_t *size = arg;
6576 	*size += len;
6577 	return (0);
6578 }
6579 
6580 /*
6581  * Determine approximately how large a zfs send stream will be -- the number
6582  * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
6583  *
6584  * innvl: {
6585  *     (optional) "from" -> full snap or bookmark name to send an incremental
6586  *                          from
6587  *     (optional) "largeblockok" -> (value ignored)
6588  *         indicates that blocks > 128KB are permitted
6589  *     (optional) "embedok" -> (value ignored)
6590  *         presence indicates DRR_WRITE_EMBEDDED records are permitted
6591  *     (optional) "compressok" -> (value ignored)
6592  *         presence indicates compressed DRR_WRITE records are permitted
6593  *     (optional) "rawok" -> (value ignored)
6594  *         presence indicates raw encrypted records should be used.
6595  *     (optional) "resume_object" and "resume_offset" -> (uint64)
6596  *         if present, resume send stream from specified object and offset.
6597  *     (optional) "fd" -> file descriptor to use as a cookie for progress
6598  *         tracking (int32)
6599  * }
6600  *
6601  * outnvl: {
6602  *     "space" -> bytes of space (uint64)
6603  * }
6604  */
6605 static const zfs_ioc_key_t zfs_keys_send_space[] = {
6606 	{"from",		DATA_TYPE_STRING,	ZK_OPTIONAL},
6607 	{"fromsnap",		DATA_TYPE_STRING,	ZK_OPTIONAL},
6608 	{"largeblockok",	DATA_TYPE_BOOLEAN,	ZK_OPTIONAL},
6609 	{"embedok",		DATA_TYPE_BOOLEAN,	ZK_OPTIONAL},
6610 	{"compressok",		DATA_TYPE_BOOLEAN,	ZK_OPTIONAL},
6611 	{"rawok",		DATA_TYPE_BOOLEAN,	ZK_OPTIONAL},
6612 	{"fd",			DATA_TYPE_INT32,	ZK_OPTIONAL},
6613 	{"redactbook",		DATA_TYPE_STRING,	ZK_OPTIONAL},
6614 	{"resume_object",	DATA_TYPE_UINT64,	ZK_OPTIONAL},
6615 	{"resume_offset",	DATA_TYPE_UINT64,	ZK_OPTIONAL},
6616 	{"bytes",		DATA_TYPE_UINT64,	ZK_OPTIONAL},
6617 };
6618 
6619 static int
6620 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
6621 {
6622 	dsl_pool_t *dp;
6623 	dsl_dataset_t *tosnap;
6624 	dsl_dataset_t *fromsnap = NULL;
6625 	int error;
6626 	char *fromname = NULL;
6627 	char *redactlist_book = NULL;
6628 	boolean_t largeblockok;
6629 	boolean_t embedok;
6630 	boolean_t compressok;
6631 	boolean_t rawok;
6632 	boolean_t savedok;
6633 	uint64_t space = 0;
6634 	boolean_t full_estimate = B_FALSE;
6635 	uint64_t resumeobj = 0;
6636 	uint64_t resumeoff = 0;
6637 	uint64_t resume_bytes = 0;
6638 	int32_t fd = -1;
6639 	zfs_bookmark_phys_t zbm = {0};
6640 
6641 	error = dsl_pool_hold(snapname, FTAG, &dp);
6642 	if (error != 0)
6643 		return (error);
6644 
6645 	error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
6646 	if (error != 0) {
6647 		dsl_pool_rele(dp, FTAG);
6648 		return (error);
6649 	}
6650 	(void) nvlist_lookup_int32(innvl, "fd", &fd);
6651 
6652 	largeblockok = nvlist_exists(innvl, "largeblockok");
6653 	embedok = nvlist_exists(innvl, "embedok");
6654 	compressok = nvlist_exists(innvl, "compressok");
6655 	rawok = nvlist_exists(innvl, "rawok");
6656 	savedok = nvlist_exists(innvl, "savedok");
6657 	boolean_t from = (nvlist_lookup_string(innvl, "from", &fromname) == 0);
6658 	boolean_t altbook = (nvlist_lookup_string(innvl, "redactbook",
6659 	    &redactlist_book) == 0);
6660 
6661 	(void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
6662 	(void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);
6663 	(void) nvlist_lookup_uint64(innvl, "bytes", &resume_bytes);
6664 
6665 	if (altbook) {
6666 		full_estimate = B_TRUE;
6667 	} else if (from) {
6668 		if (strchr(fromname, '#')) {
6669 			error = dsl_bookmark_lookup(dp, fromname, tosnap, &zbm);
6670 
6671 			/*
6672 			 * dsl_bookmark_lookup() will fail with EXDEV if
6673 			 * the from-bookmark and tosnap are at the same txg.
6674 			 * However, it's valid to do a send (and therefore,
6675 			 * a send estimate) from and to the same time point,
6676 			 * if the bookmark is redacted (the incremental send
6677 			 * can change what's redacted on the target).  In
6678 			 * this case, dsl_bookmark_lookup() fills in zbm
6679 			 * but returns EXDEV.  Ignore this error.
6680 			 */
6681 			if (error == EXDEV && zbm.zbm_redaction_obj != 0 &&
6682 			    zbm.zbm_guid ==
6683 			    dsl_dataset_phys(tosnap)->ds_guid)
6684 				error = 0;
6685 
6686 			if (error != 0) {
6687 				dsl_dataset_rele(tosnap, FTAG);
6688 				dsl_pool_rele(dp, FTAG);
6689 				return (error);
6690 			}
6691 			if (zbm.zbm_redaction_obj != 0 || !(zbm.zbm_flags &
6692 			    ZBM_FLAG_HAS_FBN)) {
6693 				full_estimate = B_TRUE;
6694 			}
6695 		} else if (strchr(fromname, '@')) {
6696 			error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
6697 			if (error != 0) {
6698 				dsl_dataset_rele(tosnap, FTAG);
6699 				dsl_pool_rele(dp, FTAG);
6700 				return (error);
6701 			}
6702 
6703 			if (!dsl_dataset_is_before(tosnap, fromsnap, 0)) {
6704 				full_estimate = B_TRUE;
6705 				dsl_dataset_rele(fromsnap, FTAG);
6706 			}
6707 		} else {
6708 			/*
6709 			 * from is not properly formatted as a snapshot or
6710 			 * bookmark
6711 			 */
6712 			dsl_dataset_rele(tosnap, FTAG);
6713 			dsl_pool_rele(dp, FTAG);
6714 			return (SET_ERROR(EINVAL));
6715 		}
6716 	}
6717 
6718 	if (full_estimate) {
6719 		dmu_send_outparams_t out = {0};
6720 		offset_t off = 0;
6721 		out.dso_outfunc = send_space_sum;
6722 		out.dso_arg = &space;
6723 		out.dso_dryrun = B_TRUE;
6724 		/*
6725 		 * We have to release these holds so dmu_send can take them.  It
6726 		 * will do all the error checking we need.
6727 		 */
6728 		dsl_dataset_rele(tosnap, FTAG);
6729 		dsl_pool_rele(dp, FTAG);
6730 		error = dmu_send(snapname, fromname, embedok, largeblockok,
6731 		    compressok, rawok, savedok, resumeobj, resumeoff,
6732 		    redactlist_book, fd, &off, &out);
6733 	} else {
6734 		error = dmu_send_estimate_fast(tosnap, fromsnap,
6735 		    (from && strchr(fromname, '#') != NULL ? &zbm : NULL),
6736 		    compressok || rawok, savedok, &space);
6737 		space -= resume_bytes;
6738 		if (fromsnap != NULL)
6739 			dsl_dataset_rele(fromsnap, FTAG);
6740 		dsl_dataset_rele(tosnap, FTAG);
6741 		dsl_pool_rele(dp, FTAG);
6742 	}
6743 
6744 	fnvlist_add_uint64(outnvl, "space", space);
6745 
6746 	return (error);
6747 }
6748 
6749 /*
6750  * Sync the currently open TXG to disk for the specified pool.
6751  * This is somewhat similar to 'zfs_sync()'.
6752  * For cases that do not result in error this ioctl will wait for
6753  * the currently open TXG to commit before returning back to the caller.
6754  *
6755  * innvl: {
6756  *  "force" -> when true, force uberblock update even if there is no dirty data.
6757  *             In addition this will cause the vdev configuration to be written
6758  *             out including updating the zpool cache file. (boolean_t)
6759  * }
6760  *
6761  * onvl is unused
6762  */
6763 static const zfs_ioc_key_t zfs_keys_pool_sync[] = {
6764 	{"force",	DATA_TYPE_BOOLEAN_VALUE,	0},
6765 };
6766 
6767 /* ARGSUSED */
6768 static int
6769 zfs_ioc_pool_sync(const char *pool, nvlist_t *innvl, nvlist_t *onvl)
6770 {
6771 	int err;
6772 	boolean_t rc, force = B_FALSE;
6773 	spa_t *spa;
6774 
6775 	if ((err = spa_open(pool, &spa, FTAG)) != 0)
6776 		return (err);
6777 
6778 	if (innvl) {
6779 		err = nvlist_lookup_boolean_value(innvl, "force", &rc);
6780 		if (err == 0)
6781 			force = rc;
6782 	}
6783 
6784 	if (force) {
6785 		spa_config_enter(spa, SCL_CONFIG, FTAG, RW_WRITER);
6786 		vdev_config_dirty(spa->spa_root_vdev);
6787 		spa_config_exit(spa, SCL_CONFIG, FTAG);
6788 	}
6789 	txg_wait_synced(spa_get_dsl(spa), 0);
6790 
6791 	spa_close(spa, FTAG);
6792 
6793 	return (0);
6794 }
6795 
6796 /*
6797  * Load a user's wrapping key into the kernel.
6798  * innvl: {
6799  *     "hidden_args" -> { "wkeydata" -> value }
6800  *         raw uint8_t array of encryption wrapping key data (32 bytes)
6801  *     (optional) "noop" -> (value ignored)
6802  *         presence indicated key should only be verified, not loaded
6803  * }
6804  */
6805 static const zfs_ioc_key_t zfs_keys_load_key[] = {
6806 	{"hidden_args",	DATA_TYPE_NVLIST,	0},
6807 	{"noop",	DATA_TYPE_BOOLEAN,	ZK_OPTIONAL},
6808 };
6809 
6810 /* ARGSUSED */
6811 static int
6812 zfs_ioc_load_key(const char *dsname, nvlist_t *innvl, nvlist_t *outnvl)
6813 {
6814 	int ret;
6815 	dsl_crypto_params_t *dcp = NULL;
6816 	nvlist_t *hidden_args;
6817 	boolean_t noop = nvlist_exists(innvl, "noop");
6818 
6819 	if (strchr(dsname, '@') != NULL || strchr(dsname, '%') != NULL) {
6820 		ret = SET_ERROR(EINVAL);
6821 		goto error;
6822 	}
6823 
6824 	hidden_args = fnvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS);
6825 
6826 	ret = dsl_crypto_params_create_nvlist(DCP_CMD_NONE, NULL,
6827 	    hidden_args, &dcp);
6828 	if (ret != 0)
6829 		goto error;
6830 
6831 	ret = spa_keystore_load_wkey(dsname, dcp, noop);
6832 	if (ret != 0)
6833 		goto error;
6834 
6835 	dsl_crypto_params_free(dcp, noop);
6836 
6837 	return (0);
6838 
6839 error:
6840 	dsl_crypto_params_free(dcp, B_TRUE);
6841 	return (ret);
6842 }
6843 
6844 /*
6845  * Unload a user's wrapping key from the kernel.
6846  * Both innvl and outnvl are unused.
6847  */
6848 static const zfs_ioc_key_t zfs_keys_unload_key[] = {
6849 	/* no nvl keys */
6850 };
6851 
6852 /* ARGSUSED */
6853 static int
6854 zfs_ioc_unload_key(const char *dsname, nvlist_t *innvl, nvlist_t *outnvl)
6855 {
6856 	int ret = 0;
6857 
6858 	if (strchr(dsname, '@') != NULL || strchr(dsname, '%') != NULL) {
6859 		ret = (SET_ERROR(EINVAL));
6860 		goto out;
6861 	}
6862 
6863 	ret = spa_keystore_unload_wkey(dsname);
6864 	if (ret != 0)
6865 		goto out;
6866 
6867 out:
6868 	return (ret);
6869 }
6870 
6871 /*
6872  * Changes a user's wrapping key used to decrypt a dataset. The keyformat,
6873  * keylocation, pbkdf2salt, and  pbkdf2iters properties can also be specified
6874  * here to change how the key is derived in userspace.
6875  *
6876  * innvl: {
6877  *    "hidden_args" (optional) -> { "wkeydata" -> value }
6878  *         raw uint8_t array of new encryption wrapping key data (32 bytes)
6879  *    "props" (optional) -> { prop -> value }
6880  * }
6881  *
6882  * outnvl is unused
6883  */
6884 static const zfs_ioc_key_t zfs_keys_change_key[] = {
6885 	{"crypt_cmd",	DATA_TYPE_UINT64,	ZK_OPTIONAL},
6886 	{"hidden_args",	DATA_TYPE_NVLIST,	ZK_OPTIONAL},
6887 	{"props",	DATA_TYPE_NVLIST,	ZK_OPTIONAL},
6888 };
6889 
6890 /* ARGSUSED */
6891 static int
6892 zfs_ioc_change_key(const char *dsname, nvlist_t *innvl, nvlist_t *outnvl)
6893 {
6894 	int ret;
6895 	uint64_t cmd = DCP_CMD_NONE;
6896 	dsl_crypto_params_t *dcp = NULL;
6897 	nvlist_t *args = NULL, *hidden_args = NULL;
6898 
6899 	if (strchr(dsname, '@') != NULL || strchr(dsname, '%') != NULL) {
6900 		ret = (SET_ERROR(EINVAL));
6901 		goto error;
6902 	}
6903 
6904 	(void) nvlist_lookup_uint64(innvl, "crypt_cmd", &cmd);
6905 	(void) nvlist_lookup_nvlist(innvl, "props", &args);
6906 	(void) nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);
6907 
6908 	ret = dsl_crypto_params_create_nvlist(cmd, args, hidden_args, &dcp);
6909 	if (ret != 0)
6910 		goto error;
6911 
6912 	ret = spa_keystore_change_key(dsname, dcp);
6913 	if (ret != 0)
6914 		goto error;
6915 
6916 	dsl_crypto_params_free(dcp, B_FALSE);
6917 
6918 	return (0);
6919 
6920 error:
6921 	dsl_crypto_params_free(dcp, B_TRUE);
6922 	return (ret);
6923 }
6924 
6925 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
6926 
6927 static void
6928 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6929     zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
6930     boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
6931 {
6932 	zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
6933 
6934 	ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
6935 	ASSERT3U(ioc, <, ZFS_IOC_LAST);
6936 	ASSERT3P(vec->zvec_legacy_func, ==, NULL);
6937 	ASSERT3P(vec->zvec_func, ==, NULL);
6938 
6939 	vec->zvec_legacy_func = func;
6940 	vec->zvec_secpolicy = secpolicy;
6941 	vec->zvec_namecheck = namecheck;
6942 	vec->zvec_allow_log = log_history;
6943 	vec->zvec_pool_check = pool_check;
6944 }
6945 
6946 /*
6947  * See the block comment at the beginning of this file for details on
6948  * each argument to this function.
6949  */
6950 void
6951 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
6952     zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
6953     zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
6954     boolean_t allow_log, const zfs_ioc_key_t *nvl_keys, size_t num_keys)
6955 {
6956 	zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
6957 
6958 	ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
6959 	ASSERT3U(ioc, <, ZFS_IOC_LAST);
6960 	ASSERT3P(vec->zvec_legacy_func, ==, NULL);
6961 	ASSERT3P(vec->zvec_func, ==, NULL);
6962 
6963 	/* if we are logging, the name must be valid */
6964 	ASSERT(!allow_log || namecheck != NO_NAME);
6965 
6966 	vec->zvec_name = name;
6967 	vec->zvec_func = func;
6968 	vec->zvec_secpolicy = secpolicy;
6969 	vec->zvec_namecheck = namecheck;
6970 	vec->zvec_pool_check = pool_check;
6971 	vec->zvec_smush_outnvlist = smush_outnvlist;
6972 	vec->zvec_allow_log = allow_log;
6973 	vec->zvec_nvl_keys = nvl_keys;
6974 	vec->zvec_nvl_key_count = num_keys;
6975 }
6976 
6977 static void
6978 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6979     zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
6980     zfs_ioc_poolcheck_t pool_check)
6981 {
6982 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
6983 	    POOL_NAME, log_history, pool_check);
6984 }
6985 
6986 void
6987 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6988     zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
6989 {
6990 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
6991 	    DATASET_NAME, B_FALSE, pool_check);
6992 }
6993 
6994 static void
6995 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
6996 {
6997 	zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
6998 	    POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
6999 }
7000 
7001 static void
7002 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
7003     zfs_secpolicy_func_t *secpolicy)
7004 {
7005 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
7006 	    NO_NAME, B_FALSE, POOL_CHECK_NONE);
7007 }
7008 
7009 static void
7010 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
7011     zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
7012 {
7013 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
7014 	    DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
7015 }
7016 
7017 static void
7018 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
7019 {
7020 	zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
7021 	    zfs_secpolicy_read);
7022 }
7023 
7024 static void
7025 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
7026     zfs_secpolicy_func_t *secpolicy)
7027 {
7028 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
7029 	    DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
7030 }
7031 
7032 static void
7033 zfs_ioctl_init(void)
7034 {
7035 	zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
7036 	    zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
7037 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7038 	    zfs_keys_snapshot, ARRAY_SIZE(zfs_keys_snapshot));
7039 
7040 	zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
7041 	    zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
7042 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE,
7043 	    zfs_keys_log_history, ARRAY_SIZE(zfs_keys_log_history));
7044 
7045 	zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
7046 	    zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
7047 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
7048 	    zfs_keys_space_snaps, ARRAY_SIZE(zfs_keys_space_snaps));
7049 
7050 	zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
7051 	    zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
7052 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
7053 	    zfs_keys_send_new, ARRAY_SIZE(zfs_keys_send_new));
7054 
7055 	zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
7056 	    zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
7057 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
7058 	    zfs_keys_send_space, ARRAY_SIZE(zfs_keys_send_space));
7059 
7060 	zfs_ioctl_register("create", ZFS_IOC_CREATE,
7061 	    zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
7062 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7063 	    zfs_keys_create, ARRAY_SIZE(zfs_keys_create));
7064 
7065 	zfs_ioctl_register("clone", ZFS_IOC_CLONE,
7066 	    zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
7067 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7068 	    zfs_keys_clone, ARRAY_SIZE(zfs_keys_clone));
7069 
7070 	zfs_ioctl_register("remap", ZFS_IOC_REMAP,
7071 	    zfs_ioc_remap, zfs_secpolicy_none, DATASET_NAME,
7072 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE,
7073 	    zfs_keys_remap, ARRAY_SIZE(zfs_keys_remap));
7074 
7075 	zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
7076 	    zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
7077 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7078 	    zfs_keys_destroy_snaps, ARRAY_SIZE(zfs_keys_destroy_snaps));
7079 
7080 	zfs_ioctl_register("hold", ZFS_IOC_HOLD,
7081 	    zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
7082 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7083 	    zfs_keys_hold, ARRAY_SIZE(zfs_keys_hold));
7084 	zfs_ioctl_register("release", ZFS_IOC_RELEASE,
7085 	    zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
7086 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7087 	    zfs_keys_release, ARRAY_SIZE(zfs_keys_release));
7088 
7089 	zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
7090 	    zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
7091 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
7092 	    zfs_keys_get_holds, ARRAY_SIZE(zfs_keys_get_holds));
7093 
7094 	zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
7095 	    zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
7096 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE,
7097 	    zfs_keys_rollback, ARRAY_SIZE(zfs_keys_rollback));
7098 
7099 	zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
7100 	    zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
7101 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7102 	    zfs_keys_bookmark, ARRAY_SIZE(zfs_keys_bookmark));
7103 
7104 	zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
7105 	    zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
7106 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
7107 	    zfs_keys_get_bookmarks, ARRAY_SIZE(zfs_keys_get_bookmarks));
7108 
7109 	zfs_ioctl_register("get_bookmark_props", ZFS_IOC_GET_BOOKMARK_PROPS,
7110 	    zfs_ioc_get_bookmark_props, zfs_secpolicy_read, ENTITY_NAME,
7111 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE, zfs_keys_get_bookmark_props,
7112 	    ARRAY_SIZE(zfs_keys_get_bookmark_props));
7113 
7114 	zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
7115 	    zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
7116 	    POOL_NAME,
7117 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7118 	    zfs_keys_destroy_bookmarks,
7119 	    ARRAY_SIZE(zfs_keys_destroy_bookmarks));
7120 
7121 	zfs_ioctl_register("receive", ZFS_IOC_RECV_NEW,
7122 	    zfs_ioc_recv_new, zfs_secpolicy_recv_new, DATASET_NAME,
7123 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7124 	    zfs_keys_recv_new, ARRAY_SIZE(zfs_keys_recv_new));
7125 	zfs_ioctl_register("load-key", ZFS_IOC_LOAD_KEY,
7126 	    zfs_ioc_load_key, zfs_secpolicy_load_key,
7127 	    DATASET_NAME, POOL_CHECK_SUSPENDED, B_TRUE, B_TRUE,
7128 	    zfs_keys_load_key, ARRAY_SIZE(zfs_keys_load_key));
7129 	zfs_ioctl_register("unload-key", ZFS_IOC_UNLOAD_KEY,
7130 	    zfs_ioc_unload_key, zfs_secpolicy_load_key,
7131 	    DATASET_NAME, POOL_CHECK_SUSPENDED, B_TRUE, B_TRUE,
7132 	    zfs_keys_unload_key, ARRAY_SIZE(zfs_keys_unload_key));
7133 	zfs_ioctl_register("change-key", ZFS_IOC_CHANGE_KEY,
7134 	    zfs_ioc_change_key, zfs_secpolicy_change_key,
7135 	    DATASET_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY,
7136 	    B_TRUE, B_TRUE, zfs_keys_change_key,
7137 	    ARRAY_SIZE(zfs_keys_change_key));
7138 
7139 	zfs_ioctl_register("sync", ZFS_IOC_POOL_SYNC,
7140 	    zfs_ioc_pool_sync, zfs_secpolicy_none, POOL_NAME,
7141 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE,
7142 	    zfs_keys_pool_sync, ARRAY_SIZE(zfs_keys_pool_sync));
7143 	zfs_ioctl_register("reopen", ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
7144 	    zfs_secpolicy_config, POOL_NAME, POOL_CHECK_SUSPENDED, B_TRUE,
7145 	    B_TRUE, zfs_keys_pool_reopen, ARRAY_SIZE(zfs_keys_pool_reopen));
7146 
7147 	zfs_ioctl_register("channel_program", ZFS_IOC_CHANNEL_PROGRAM,
7148 	    zfs_ioc_channel_program, zfs_secpolicy_config,
7149 	    POOL_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE,
7150 	    B_TRUE, zfs_keys_channel_program,
7151 	    ARRAY_SIZE(zfs_keys_channel_program));
7152 
7153 	zfs_ioctl_register("redact", ZFS_IOC_REDACT,
7154 	    zfs_ioc_redact, zfs_secpolicy_config, DATASET_NAME,
7155 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7156 	    zfs_keys_redact, ARRAY_SIZE(zfs_keys_redact));
7157 
7158 	zfs_ioctl_register("zpool_checkpoint", ZFS_IOC_POOL_CHECKPOINT,
7159 	    zfs_ioc_pool_checkpoint, zfs_secpolicy_config, POOL_NAME,
7160 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7161 	    zfs_keys_pool_checkpoint, ARRAY_SIZE(zfs_keys_pool_checkpoint));
7162 
7163 	zfs_ioctl_register("zpool_discard_checkpoint",
7164 	    ZFS_IOC_POOL_DISCARD_CHECKPOINT, zfs_ioc_pool_discard_checkpoint,
7165 	    zfs_secpolicy_config, POOL_NAME,
7166 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7167 	    zfs_keys_pool_discard_checkpoint,
7168 	    ARRAY_SIZE(zfs_keys_pool_discard_checkpoint));
7169 
7170 	zfs_ioctl_register("initialize", ZFS_IOC_POOL_INITIALIZE,
7171 	    zfs_ioc_pool_initialize, zfs_secpolicy_config, POOL_NAME,
7172 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7173 	    zfs_keys_pool_initialize, ARRAY_SIZE(zfs_keys_pool_initialize));
7174 
7175 	zfs_ioctl_register("trim", ZFS_IOC_POOL_TRIM,
7176 	    zfs_ioc_pool_trim, zfs_secpolicy_config, POOL_NAME,
7177 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7178 	    zfs_keys_pool_trim, ARRAY_SIZE(zfs_keys_pool_trim));
7179 
7180 	zfs_ioctl_register("wait", ZFS_IOC_WAIT,
7181 	    zfs_ioc_wait, zfs_secpolicy_none, POOL_NAME,
7182 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE,
7183 	    zfs_keys_pool_wait, ARRAY_SIZE(zfs_keys_pool_wait));
7184 
7185 	zfs_ioctl_register("wait_fs", ZFS_IOC_WAIT_FS,
7186 	    zfs_ioc_wait_fs, zfs_secpolicy_none, DATASET_NAME,
7187 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE,
7188 	    zfs_keys_fs_wait, ARRAY_SIZE(zfs_keys_fs_wait));
7189 
7190 	zfs_ioctl_register("set_bootenv", ZFS_IOC_SET_BOOTENV,
7191 	    zfs_ioc_set_bootenv, zfs_secpolicy_config, POOL_NAME,
7192 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE,
7193 	    zfs_keys_set_bootenv, ARRAY_SIZE(zfs_keys_set_bootenv));
7194 
7195 	zfs_ioctl_register("get_bootenv", ZFS_IOC_GET_BOOTENV,
7196 	    zfs_ioc_get_bootenv, zfs_secpolicy_none, POOL_NAME,
7197 	    POOL_CHECK_SUSPENDED, B_FALSE, B_TRUE,
7198 	    zfs_keys_get_bootenv, ARRAY_SIZE(zfs_keys_get_bootenv));
7199 
7200 	zfs_ioctl_register("zpool_vdev_get_props", ZFS_IOC_VDEV_GET_PROPS,
7201 	    zfs_ioc_vdev_get_props, zfs_secpolicy_read, POOL_NAME,
7202 	    POOL_CHECK_NONE, B_FALSE, B_FALSE, zfs_keys_vdev_get_props,
7203 	    ARRAY_SIZE(zfs_keys_vdev_get_props));
7204 
7205 	zfs_ioctl_register("zpool_vdev_set_props", ZFS_IOC_VDEV_SET_PROPS,
7206 	    zfs_ioc_vdev_set_props, zfs_secpolicy_config, POOL_NAME,
7207 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE,
7208 	    zfs_keys_vdev_set_props, ARRAY_SIZE(zfs_keys_vdev_set_props));
7209 
7210 	/* IOCTLS that use the legacy function signature */
7211 
7212 	zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
7213 	    zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
7214 
7215 	zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
7216 	    zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
7217 	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
7218 	    zfs_ioc_pool_scan);
7219 	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
7220 	    zfs_ioc_pool_upgrade);
7221 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
7222 	    zfs_ioc_vdev_add);
7223 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
7224 	    zfs_ioc_vdev_remove);
7225 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
7226 	    zfs_ioc_vdev_set_state);
7227 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
7228 	    zfs_ioc_vdev_attach);
7229 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
7230 	    zfs_ioc_vdev_detach);
7231 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
7232 	    zfs_ioc_vdev_setpath);
7233 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
7234 	    zfs_ioc_vdev_setfru);
7235 	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
7236 	    zfs_ioc_pool_set_props);
7237 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
7238 	    zfs_ioc_vdev_split);
7239 	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
7240 	    zfs_ioc_pool_reguid);
7241 
7242 	zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
7243 	    zfs_ioc_pool_configs, zfs_secpolicy_none);
7244 	zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
7245 	    zfs_ioc_pool_tryimport, zfs_secpolicy_config);
7246 	zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
7247 	    zfs_ioc_inject_fault, zfs_secpolicy_inject);
7248 	zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
7249 	    zfs_ioc_clear_fault, zfs_secpolicy_inject);
7250 	zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
7251 	    zfs_ioc_inject_list_next, zfs_secpolicy_inject);
7252 
7253 	/*
7254 	 * pool destroy, and export don't log the history as part of
7255 	 * zfsdev_ioctl, but rather zfs_ioc_pool_export
7256 	 * does the logging of those commands.
7257 	 */
7258 	zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
7259 	    zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
7260 	zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
7261 	    zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
7262 
7263 	zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
7264 	    zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
7265 	zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
7266 	    zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
7267 
7268 	zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
7269 	    zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
7270 	zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
7271 	    zfs_ioc_dsobj_to_dsname,
7272 	    zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
7273 	zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
7274 	    zfs_ioc_pool_get_history,
7275 	    zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
7276 
7277 	zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
7278 	    zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
7279 
7280 	zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
7281 	    zfs_secpolicy_config, B_TRUE, POOL_CHECK_READONLY);
7282 
7283 	zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
7284 	    zfs_ioc_space_written);
7285 	zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
7286 	    zfs_ioc_objset_recvd_props);
7287 	zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
7288 	    zfs_ioc_next_obj);
7289 	zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
7290 	    zfs_ioc_get_fsacl);
7291 	zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
7292 	    zfs_ioc_objset_stats);
7293 	zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
7294 	    zfs_ioc_objset_zplprops);
7295 	zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
7296 	    zfs_ioc_dataset_list_next);
7297 	zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
7298 	    zfs_ioc_snapshot_list_next);
7299 	zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
7300 	    zfs_ioc_send_progress);
7301 
7302 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
7303 	    zfs_ioc_diff, zfs_secpolicy_diff);
7304 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
7305 	    zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
7306 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
7307 	    zfs_ioc_obj_to_path, zfs_secpolicy_diff);
7308 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
7309 	    zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
7310 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
7311 	    zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
7312 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
7313 	    zfs_ioc_send, zfs_secpolicy_send);
7314 
7315 	zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
7316 	    zfs_secpolicy_none);
7317 	zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
7318 	    zfs_secpolicy_destroy);
7319 	zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
7320 	    zfs_secpolicy_rename);
7321 	zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
7322 	    zfs_secpolicy_recv);
7323 	zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
7324 	    zfs_secpolicy_promote);
7325 	zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
7326 	    zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
7327 	zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
7328 	    zfs_secpolicy_set_fsacl);
7329 
7330 	zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
7331 	    zfs_secpolicy_share, POOL_CHECK_NONE);
7332 	zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
7333 	    zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
7334 	zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
7335 	    zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
7336 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
7337 	zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
7338 	    zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
7339 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
7340 
7341 	zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_NEXT, zfs_ioc_events_next,
7342 	    zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
7343 	zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_CLEAR, zfs_ioc_events_clear,
7344 	    zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
7345 	zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_SEEK, zfs_ioc_events_seek,
7346 	    zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
7347 
7348 	zfs_ioctl_init_os();
7349 }
7350 
7351 /*
7352  * Verify that for non-legacy ioctls the input nvlist
7353  * pairs match against the expected input.
7354  *
7355  * Possible errors are:
7356  * ZFS_ERR_IOC_ARG_UNAVAIL	An unrecognized nvpair was encountered
7357  * ZFS_ERR_IOC_ARG_REQUIRED	A required nvpair is missing
7358  * ZFS_ERR_IOC_ARG_BADTYPE	Invalid type for nvpair
7359  */
7360 static int
7361 zfs_check_input_nvpairs(nvlist_t *innvl, const zfs_ioc_vec_t *vec)
7362 {
7363 	const zfs_ioc_key_t *nvl_keys = vec->zvec_nvl_keys;
7364 	boolean_t required_keys_found = B_FALSE;
7365 
7366 	/*
7367 	 * examine each input pair
7368 	 */
7369 	for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
7370 	    pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
7371 		char *name = nvpair_name(pair);
7372 		data_type_t type = nvpair_type(pair);
7373 		boolean_t identified = B_FALSE;
7374 
7375 		/*
7376 		 * check pair against the documented names and type
7377 		 */
7378 		for (int k = 0; k < vec->zvec_nvl_key_count; k++) {
7379 			/* if not a wild card name, check for an exact match */
7380 			if ((nvl_keys[k].zkey_flags & ZK_WILDCARDLIST) == 0 &&
7381 			    strcmp(nvl_keys[k].zkey_name, name) != 0)
7382 				continue;
7383 
7384 			identified = B_TRUE;
7385 
7386 			if (nvl_keys[k].zkey_type != DATA_TYPE_ANY &&
7387 			    nvl_keys[k].zkey_type != type) {
7388 				return (SET_ERROR(ZFS_ERR_IOC_ARG_BADTYPE));
7389 			}
7390 
7391 			if (nvl_keys[k].zkey_flags & ZK_OPTIONAL)
7392 				continue;
7393 
7394 			required_keys_found = B_TRUE;
7395 			break;
7396 		}
7397 
7398 		/* allow an 'optional' key, everything else is invalid */
7399 		if (!identified &&
7400 		    (strcmp(name, "optional") != 0 ||
7401 		    type != DATA_TYPE_NVLIST)) {
7402 			return (SET_ERROR(ZFS_ERR_IOC_ARG_UNAVAIL));
7403 		}
7404 	}
7405 
7406 	/* verify that all required keys were found */
7407 	for (int k = 0; k < vec->zvec_nvl_key_count; k++) {
7408 		if (nvl_keys[k].zkey_flags & ZK_OPTIONAL)
7409 			continue;
7410 
7411 		if (nvl_keys[k].zkey_flags & ZK_WILDCARDLIST) {
7412 			/* at least one non-optional key is expected here */
7413 			if (!required_keys_found)
7414 				return (SET_ERROR(ZFS_ERR_IOC_ARG_REQUIRED));
7415 			continue;
7416 		}
7417 
7418 		if (!nvlist_exists(innvl, nvl_keys[k].zkey_name))
7419 			return (SET_ERROR(ZFS_ERR_IOC_ARG_REQUIRED));
7420 	}
7421 
7422 	return (0);
7423 }
7424 
7425 static int
7426 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
7427     zfs_ioc_poolcheck_t check)
7428 {
7429 	spa_t *spa;
7430 	int error;
7431 
7432 	ASSERT(type == POOL_NAME || type == DATASET_NAME ||
7433 	    type == ENTITY_NAME);
7434 
7435 	if (check & POOL_CHECK_NONE)
7436 		return (0);
7437 
7438 	error = spa_open(name, &spa, FTAG);
7439 	if (error == 0) {
7440 		if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
7441 			error = SET_ERROR(EAGAIN);
7442 		else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
7443 			error = SET_ERROR(EROFS);
7444 		spa_close(spa, FTAG);
7445 	}
7446 	return (error);
7447 }
7448 
7449 int
7450 zfsdev_getminor(zfs_file_t *fp, minor_t *minorp)
7451 {
7452 	zfsdev_state_t *zs, *fpd;
7453 
7454 	ASSERT(!MUTEX_HELD(&zfsdev_state_lock));
7455 
7456 	fpd = zfs_file_private(fp);
7457 	if (fpd == NULL)
7458 		return (SET_ERROR(EBADF));
7459 
7460 	mutex_enter(&zfsdev_state_lock);
7461 
7462 	for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
7463 
7464 		if (zs->zs_minor == -1)
7465 			continue;
7466 
7467 		if (fpd == zs) {
7468 			*minorp = fpd->zs_minor;
7469 			mutex_exit(&zfsdev_state_lock);
7470 			return (0);
7471 		}
7472 	}
7473 
7474 	mutex_exit(&zfsdev_state_lock);
7475 
7476 	return (SET_ERROR(EBADF));
7477 }
7478 
7479 void *
7480 zfsdev_get_state(minor_t minor, enum zfsdev_state_type which)
7481 {
7482 	zfsdev_state_t *zs;
7483 
7484 	for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
7485 		if (zs->zs_minor == minor) {
7486 			smp_rmb();
7487 			switch (which) {
7488 			case ZST_ONEXIT:
7489 				return (zs->zs_onexit);
7490 			case ZST_ZEVENT:
7491 				return (zs->zs_zevent);
7492 			case ZST_ALL:
7493 				return (zs);
7494 			}
7495 		}
7496 	}
7497 
7498 	return (NULL);
7499 }
7500 
7501 /*
7502  * Find a free minor number.  The zfsdev_state_list is expected to
7503  * be short since it is only a list of currently open file handles.
7504  */
7505 static minor_t
7506 zfsdev_minor_alloc(void)
7507 {
7508 	static minor_t last_minor = 0;
7509 	minor_t m;
7510 
7511 	ASSERT(MUTEX_HELD(&zfsdev_state_lock));
7512 
7513 	for (m = last_minor + 1; m != last_minor; m++) {
7514 		if (m > ZFSDEV_MAX_MINOR)
7515 			m = 1;
7516 		if (zfsdev_get_state(m, ZST_ALL) == NULL) {
7517 			last_minor = m;
7518 			return (m);
7519 		}
7520 	}
7521 
7522 	return (0);
7523 }
7524 
7525 int
7526 zfsdev_state_init(void *priv)
7527 {
7528 	zfsdev_state_t *zs, *zsprev = NULL;
7529 	minor_t minor;
7530 	boolean_t newzs = B_FALSE;
7531 
7532 	ASSERT(MUTEX_HELD(&zfsdev_state_lock));
7533 
7534 	minor = zfsdev_minor_alloc();
7535 	if (minor == 0)
7536 		return (SET_ERROR(ENXIO));
7537 
7538 	for (zs = zfsdev_state_list; zs != NULL; zs = zs->zs_next) {
7539 		if (zs->zs_minor == -1)
7540 			break;
7541 		zsprev = zs;
7542 	}
7543 
7544 	if (!zs) {
7545 		zs = kmem_zalloc(sizeof (zfsdev_state_t), KM_SLEEP);
7546 		newzs = B_TRUE;
7547 	}
7548 
7549 	zfsdev_private_set_state(priv, zs);
7550 
7551 	zfs_onexit_init((zfs_onexit_t **)&zs->zs_onexit);
7552 	zfs_zevent_init((zfs_zevent_t **)&zs->zs_zevent);
7553 
7554 	/*
7555 	 * In order to provide for lock-free concurrent read access
7556 	 * to the minor list in zfsdev_get_state(), new entries
7557 	 * must be completely written before linking them into the
7558 	 * list whereas existing entries are already linked; the last
7559 	 * operation must be updating zs_minor (from -1 to the new
7560 	 * value).
7561 	 */
7562 	if (newzs) {
7563 		zs->zs_minor = minor;
7564 		membar_producer();
7565 		zsprev->zs_next = zs;
7566 	} else {
7567 		membar_producer();
7568 		zs->zs_minor = minor;
7569 	}
7570 
7571 	return (0);
7572 }
7573 
7574 void
7575 zfsdev_state_destroy(void *priv)
7576 {
7577 	zfsdev_state_t *zs = zfsdev_private_get_state(priv);
7578 
7579 	ASSERT(zs != NULL);
7580 	ASSERT3S(zs->zs_minor, >, 0);
7581 
7582 	/*
7583 	 * The last reference to this zfsdev file descriptor is being dropped.
7584 	 * We don't have to worry about lookup grabbing this state object, and
7585 	 * zfsdev_state_init() will not try to reuse this object until it is
7586 	 * invalidated by setting zs_minor to -1.  Invalidation must be done
7587 	 * last, with a memory barrier to ensure ordering.  This lets us avoid
7588 	 * taking the global zfsdev state lock around destruction.
7589 	 */
7590 	zfs_onexit_destroy(zs->zs_onexit);
7591 	zfs_zevent_destroy(zs->zs_zevent);
7592 	zs->zs_onexit = NULL;
7593 	zs->zs_zevent = NULL;
7594 	membar_producer();
7595 	zs->zs_minor = -1;
7596 }
7597 
7598 long
7599 zfsdev_ioctl_common(uint_t vecnum, zfs_cmd_t *zc, int flag)
7600 {
7601 	int error, cmd;
7602 	const zfs_ioc_vec_t *vec;
7603 	char *saved_poolname = NULL;
7604 	uint64_t max_nvlist_src_size;
7605 	size_t saved_poolname_len = 0;
7606 	nvlist_t *innvl = NULL;
7607 	fstrans_cookie_t cookie;
7608 	hrtime_t start_time = gethrtime();
7609 
7610 	cmd = vecnum;
7611 	error = 0;
7612 	if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
7613 		return (SET_ERROR(ZFS_ERR_IOC_CMD_UNAVAIL));
7614 
7615 	vec = &zfs_ioc_vec[vecnum];
7616 
7617 	/*
7618 	 * The registered ioctl list may be sparse, verify that either
7619 	 * a normal or legacy handler are registered.
7620 	 */
7621 	if (vec->zvec_func == NULL && vec->zvec_legacy_func == NULL)
7622 		return (SET_ERROR(ZFS_ERR_IOC_CMD_UNAVAIL));
7623 
7624 	zc->zc_iflags = flag & FKIOCTL;
7625 	max_nvlist_src_size = zfs_max_nvlist_src_size_os();
7626 	if (zc->zc_nvlist_src_size > max_nvlist_src_size) {
7627 		/*
7628 		 * Make sure the user doesn't pass in an insane value for
7629 		 * zc_nvlist_src_size.  We have to check, since we will end
7630 		 * up allocating that much memory inside of get_nvlist().  This
7631 		 * prevents a nefarious user from allocating tons of kernel
7632 		 * memory.
7633 		 *
7634 		 * Also, we return EINVAL instead of ENOMEM here.  The reason
7635 		 * being that returning ENOMEM from an ioctl() has a special
7636 		 * connotation; that the user's size value is too small and
7637 		 * needs to be expanded to hold the nvlist.  See
7638 		 * zcmd_expand_dst_nvlist() for details.
7639 		 */
7640 		error = SET_ERROR(EINVAL);	/* User's size too big */
7641 
7642 	} else if (zc->zc_nvlist_src_size != 0) {
7643 		error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
7644 		    zc->zc_iflags, &innvl);
7645 		if (error != 0)
7646 			goto out;
7647 	}
7648 
7649 	/*
7650 	 * Ensure that all pool/dataset names are valid before we pass down to
7651 	 * the lower layers.
7652 	 */
7653 	zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
7654 	switch (vec->zvec_namecheck) {
7655 	case POOL_NAME:
7656 		if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
7657 			error = SET_ERROR(EINVAL);
7658 		else
7659 			error = pool_status_check(zc->zc_name,
7660 			    vec->zvec_namecheck, vec->zvec_pool_check);
7661 		break;
7662 
7663 	case DATASET_NAME:
7664 		if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
7665 			error = SET_ERROR(EINVAL);
7666 		else
7667 			error = pool_status_check(zc->zc_name,
7668 			    vec->zvec_namecheck, vec->zvec_pool_check);
7669 		break;
7670 
7671 	case ENTITY_NAME:
7672 		if (entity_namecheck(zc->zc_name, NULL, NULL) != 0) {
7673 			error = SET_ERROR(EINVAL);
7674 		} else {
7675 			error = pool_status_check(zc->zc_name,
7676 			    vec->zvec_namecheck, vec->zvec_pool_check);
7677 		}
7678 		break;
7679 
7680 	case NO_NAME:
7681 		break;
7682 	}
7683 	/*
7684 	 * Ensure that all input pairs are valid before we pass them down
7685 	 * to the lower layers.
7686 	 *
7687 	 * The vectored functions can use fnvlist_lookup_{type} for any
7688 	 * required pairs since zfs_check_input_nvpairs() confirmed that
7689 	 * they exist and are of the correct type.
7690 	 */
7691 	if (error == 0 && vec->zvec_func != NULL) {
7692 		error = zfs_check_input_nvpairs(innvl, vec);
7693 		if (error != 0)
7694 			goto out;
7695 	}
7696 
7697 	if (error == 0) {
7698 		cookie = spl_fstrans_mark();
7699 		error = vec->zvec_secpolicy(zc, innvl, CRED());
7700 		spl_fstrans_unmark(cookie);
7701 	}
7702 
7703 	if (error != 0)
7704 		goto out;
7705 
7706 	/* legacy ioctls can modify zc_name */
7707 	/*
7708 	 * Can't use kmem_strdup() as we might truncate the string and
7709 	 * kmem_strfree() would then free with incorrect size.
7710 	 */
7711 	saved_poolname_len = strlen(zc->zc_name) + 1;
7712 	saved_poolname = kmem_alloc(saved_poolname_len, KM_SLEEP);
7713 
7714 	strlcpy(saved_poolname, zc->zc_name, saved_poolname_len);
7715 	saved_poolname[strcspn(saved_poolname, "/@#")] = '\0';
7716 
7717 	if (vec->zvec_func != NULL) {
7718 		nvlist_t *outnvl;
7719 		int puterror = 0;
7720 		spa_t *spa;
7721 		nvlist_t *lognv = NULL;
7722 
7723 		ASSERT(vec->zvec_legacy_func == NULL);
7724 
7725 		/*
7726 		 * Add the innvl to the lognv before calling the func,
7727 		 * in case the func changes the innvl.
7728 		 */
7729 		if (vec->zvec_allow_log) {
7730 			lognv = fnvlist_alloc();
7731 			fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
7732 			    vec->zvec_name);
7733 			if (!nvlist_empty(innvl)) {
7734 				fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
7735 				    innvl);
7736 			}
7737 		}
7738 
7739 		outnvl = fnvlist_alloc();
7740 		cookie = spl_fstrans_mark();
7741 		error = vec->zvec_func(zc->zc_name, innvl, outnvl);
7742 		spl_fstrans_unmark(cookie);
7743 
7744 		/*
7745 		 * Some commands can partially execute, modify state, and still
7746 		 * return an error.  In these cases, attempt to record what
7747 		 * was modified.
7748 		 */
7749 		if ((error == 0 ||
7750 		    (cmd == ZFS_IOC_CHANNEL_PROGRAM && error != EINVAL)) &&
7751 		    vec->zvec_allow_log &&
7752 		    spa_open(zc->zc_name, &spa, FTAG) == 0) {
7753 			if (!nvlist_empty(outnvl)) {
7754 				size_t out_size = fnvlist_size(outnvl);
7755 				if (out_size > zfs_history_output_max) {
7756 					fnvlist_add_int64(lognv,
7757 					    ZPOOL_HIST_OUTPUT_SIZE, out_size);
7758 				} else {
7759 					fnvlist_add_nvlist(lognv,
7760 					    ZPOOL_HIST_OUTPUT_NVL, outnvl);
7761 				}
7762 			}
7763 			if (error != 0) {
7764 				fnvlist_add_int64(lognv, ZPOOL_HIST_ERRNO,
7765 				    error);
7766 			}
7767 			fnvlist_add_int64(lognv, ZPOOL_HIST_ELAPSED_NS,
7768 			    gethrtime() - start_time);
7769 			(void) spa_history_log_nvl(spa, lognv);
7770 			spa_close(spa, FTAG);
7771 		}
7772 		fnvlist_free(lognv);
7773 
7774 		if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
7775 			int smusherror = 0;
7776 			if (vec->zvec_smush_outnvlist) {
7777 				smusherror = nvlist_smush(outnvl,
7778 				    zc->zc_nvlist_dst_size);
7779 			}
7780 			if (smusherror == 0)
7781 				puterror = put_nvlist(zc, outnvl);
7782 		}
7783 
7784 		if (puterror != 0)
7785 			error = puterror;
7786 
7787 		nvlist_free(outnvl);
7788 	} else {
7789 		cookie = spl_fstrans_mark();
7790 		error = vec->zvec_legacy_func(zc);
7791 		spl_fstrans_unmark(cookie);
7792 	}
7793 
7794 out:
7795 	nvlist_free(innvl);
7796 	if (error == 0 && vec->zvec_allow_log) {
7797 		char *s = tsd_get(zfs_allow_log_key);
7798 		if (s != NULL)
7799 			kmem_strfree(s);
7800 		(void) tsd_set(zfs_allow_log_key, kmem_strdup(saved_poolname));
7801 	}
7802 	if (saved_poolname != NULL)
7803 		kmem_free(saved_poolname, saved_poolname_len);
7804 
7805 	return (error);
7806 }
7807 
7808 int
7809 zfs_kmod_init(void)
7810 {
7811 	int error;
7812 
7813 	if ((error = zvol_init()) != 0)
7814 		return (error);
7815 
7816 	spa_init(SPA_MODE_READ | SPA_MODE_WRITE);
7817 	zfs_init();
7818 
7819 	zfs_ioctl_init();
7820 
7821 	mutex_init(&zfsdev_state_lock, NULL, MUTEX_DEFAULT, NULL);
7822 	zfsdev_state_list = kmem_zalloc(sizeof (zfsdev_state_t), KM_SLEEP);
7823 	zfsdev_state_list->zs_minor = -1;
7824 
7825 	if ((error = zfsdev_attach()) != 0)
7826 		goto out;
7827 
7828 	tsd_create(&zfs_fsyncer_key, NULL);
7829 	tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
7830 	tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
7831 
7832 	return (0);
7833 out:
7834 	zfs_fini();
7835 	spa_fini();
7836 	zvol_fini();
7837 
7838 	return (error);
7839 }
7840 
7841 void
7842 zfs_kmod_fini(void)
7843 {
7844 	zfsdev_state_t *zs, *zsnext = NULL;
7845 
7846 	zfsdev_detach();
7847 
7848 	mutex_destroy(&zfsdev_state_lock);
7849 
7850 	for (zs = zfsdev_state_list; zs != NULL; zs = zsnext) {
7851 		zsnext = zs->zs_next;
7852 		if (zs->zs_onexit)
7853 			zfs_onexit_destroy(zs->zs_onexit);
7854 		if (zs->zs_zevent)
7855 			zfs_zevent_destroy(zs->zs_zevent);
7856 		kmem_free(zs, sizeof (zfsdev_state_t));
7857 	}
7858 
7859 	zfs_ereport_taskq_fini();	/* run before zfs_fini() on Linux */
7860 	zfs_fini();
7861 	spa_fini();
7862 	zvol_fini();
7863 
7864 	tsd_destroy(&zfs_fsyncer_key);
7865 	tsd_destroy(&rrw_tsd_key);
7866 	tsd_destroy(&zfs_allow_log_key);
7867 }
7868 
7869 /* BEGIN CSTYLED */
7870 ZFS_MODULE_PARAM(zfs, zfs_, max_nvlist_src_size, ULONG, ZMOD_RW,
7871     "Maximum size in bytes allowed for src nvlist passed with ZFS ioctls");
7872 
7873 ZFS_MODULE_PARAM(zfs, zfs_, history_output_max, ULONG, ZMOD_RW,
7874     "Maximum size in bytes of ZFS ioctl output that will be logged");
7875 /* END CSTYLED */
7876