xref: /freebsd/sys/contrib/openzfs/module/zfs/zfs_ioctl.c (revision a2f733abcff64628b7771a47089628b7327a88bd)
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 https://opensource.org/licenses/CDDL-1.0.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 
22 /*
23  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24  * Portions Copyright 2011 Martin Matuska
25  * Copyright 2015, OmniTI Computer Consulting, Inc. All rights reserved.
26  * Copyright (c) 2012 Pawel Jakub Dawidek
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 static zfsdev_state_t zfsdev_state_listhead;
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 uint64_t 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 static uint64_t zfs_history_output_max = 1024 * 1024;
240 
241 uint_t zfs_allow_log_key;
242 
243 /* DATA_TYPE_ANY is used when zkey_type can vary. */
244 #define	DATA_TYPE_ANY	DATA_TYPE_UNKNOWN
245 
246 typedef struct zfs_ioc_vec {
247 	zfs_ioc_legacy_func_t	*zvec_legacy_func;
248 	zfs_ioc_func_t		*zvec_func;
249 	zfs_secpolicy_func_t	*zvec_secpolicy;
250 	zfs_ioc_namecheck_t	zvec_namecheck;
251 	boolean_t		zvec_allow_log;
252 	zfs_ioc_poolcheck_t	zvec_pool_check;
253 	boolean_t		zvec_smush_outnvlist;
254 	const char		*zvec_name;
255 	const zfs_ioc_key_t	*zvec_nvl_keys;
256 	size_t			zvec_nvl_key_count;
257 } zfs_ioc_vec_t;
258 
259 /* This array is indexed by zfs_userquota_prop_t */
260 static const char *userquota_perms[] = {
261 	ZFS_DELEG_PERM_USERUSED,
262 	ZFS_DELEG_PERM_USERQUOTA,
263 	ZFS_DELEG_PERM_GROUPUSED,
264 	ZFS_DELEG_PERM_GROUPQUOTA,
265 	ZFS_DELEG_PERM_USEROBJUSED,
266 	ZFS_DELEG_PERM_USEROBJQUOTA,
267 	ZFS_DELEG_PERM_GROUPOBJUSED,
268 	ZFS_DELEG_PERM_GROUPOBJQUOTA,
269 	ZFS_DELEG_PERM_PROJECTUSED,
270 	ZFS_DELEG_PERM_PROJECTQUOTA,
271 	ZFS_DELEG_PERM_PROJECTOBJUSED,
272 	ZFS_DELEG_PERM_PROJECTOBJQUOTA,
273 };
274 
275 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
276 static int zfs_ioc_id_quota_upgrade(zfs_cmd_t *zc);
277 static int zfs_check_settable(const char *name, nvpair_t *property,
278     cred_t *cr);
279 static int zfs_check_clearable(const char *dataset, nvlist_t *props,
280     nvlist_t **errors);
281 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
282     boolean_t *);
283 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
284 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
285 
286 static void
287 history_str_free(char *buf)
288 {
289 	kmem_free(buf, HIS_MAX_RECORD_LEN);
290 }
291 
292 static char *
293 history_str_get(zfs_cmd_t *zc)
294 {
295 	char *buf;
296 
297 	if (zc->zc_history == 0)
298 		return (NULL);
299 
300 	buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
301 	if (copyinstr((void *)(uintptr_t)zc->zc_history,
302 	    buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
303 		history_str_free(buf);
304 		return (NULL);
305 	}
306 
307 	buf[HIS_MAX_RECORD_LEN -1] = '\0';
308 
309 	return (buf);
310 }
311 
312 /*
313  * Return non-zero if the spa version is less than requested version.
314  */
315 static int
316 zfs_earlier_version(const char *name, int version)
317 {
318 	spa_t *spa;
319 
320 	if (spa_open(name, &spa, FTAG) == 0) {
321 		if (spa_version(spa) < version) {
322 			spa_close(spa, FTAG);
323 			return (1);
324 		}
325 		spa_close(spa, FTAG);
326 	}
327 	return (0);
328 }
329 
330 /*
331  * Return TRUE if the ZPL version is less than requested version.
332  */
333 static boolean_t
334 zpl_earlier_version(const char *name, int version)
335 {
336 	objset_t *os;
337 	boolean_t rc = B_TRUE;
338 
339 	if (dmu_objset_hold(name, FTAG, &os) == 0) {
340 		uint64_t zplversion;
341 
342 		if (dmu_objset_type(os) != DMU_OST_ZFS) {
343 			dmu_objset_rele(os, FTAG);
344 			return (B_TRUE);
345 		}
346 		/* XXX reading from non-owned objset */
347 		if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
348 			rc = zplversion < version;
349 		dmu_objset_rele(os, FTAG);
350 	}
351 	return (rc);
352 }
353 
354 static void
355 zfs_log_history(zfs_cmd_t *zc)
356 {
357 	spa_t *spa;
358 	char *buf;
359 
360 	if ((buf = history_str_get(zc)) == NULL)
361 		return;
362 
363 	if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
364 		if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
365 			(void) spa_history_log(spa, buf);
366 		spa_close(spa, FTAG);
367 	}
368 	history_str_free(buf);
369 }
370 
371 /*
372  * Policy for top-level read operations (list pools).  Requires no privileges,
373  * and can be used in the local zone, as there is no associated dataset.
374  */
375 static int
376 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
377 {
378 	(void) zc, (void) innvl, (void) cr;
379 	return (0);
380 }
381 
382 /*
383  * Policy for dataset read operations (list children, get statistics).  Requires
384  * no privileges, but must be visible in the local zone.
385  */
386 static int
387 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
388 {
389 	(void) innvl, (void) cr;
390 	if (INGLOBALZONE(curproc) ||
391 	    zone_dataset_visible(zc->zc_name, NULL))
392 		return (0);
393 
394 	return (SET_ERROR(ENOENT));
395 }
396 
397 static int
398 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
399 {
400 	int writable = 1;
401 
402 	/*
403 	 * The dataset must be visible by this zone -- check this first
404 	 * so they don't see EPERM on something they shouldn't know about.
405 	 */
406 	if (!INGLOBALZONE(curproc) &&
407 	    !zone_dataset_visible(dataset, &writable))
408 		return (SET_ERROR(ENOENT));
409 
410 	if (INGLOBALZONE(curproc)) {
411 		/*
412 		 * If the fs is zoned, only root can access it from the
413 		 * global zone.
414 		 */
415 		if (secpolicy_zfs(cr) && zoned)
416 			return (SET_ERROR(EPERM));
417 	} else {
418 		/*
419 		 * If we are in a local zone, the 'zoned' property must be set.
420 		 */
421 		if (!zoned)
422 			return (SET_ERROR(EPERM));
423 
424 		/* must be writable by this zone */
425 		if (!writable)
426 			return (SET_ERROR(EPERM));
427 	}
428 	return (0);
429 }
430 
431 static int
432 zfs_dozonecheck(const char *dataset, cred_t *cr)
433 {
434 	uint64_t zoned;
435 
436 	if (dsl_prop_get_integer(dataset, zfs_prop_to_name(ZFS_PROP_ZONED),
437 	    &zoned, NULL))
438 		return (SET_ERROR(ENOENT));
439 
440 	return (zfs_dozonecheck_impl(dataset, zoned, cr));
441 }
442 
443 static int
444 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
445 {
446 	uint64_t zoned;
447 
448 	if (dsl_prop_get_int_ds(ds, zfs_prop_to_name(ZFS_PROP_ZONED), &zoned))
449 		return (SET_ERROR(ENOENT));
450 
451 	return (zfs_dozonecheck_impl(dataset, zoned, cr));
452 }
453 
454 static int
455 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
456     const char *perm, cred_t *cr)
457 {
458 	int error;
459 
460 	error = zfs_dozonecheck_ds(name, ds, cr);
461 	if (error == 0) {
462 		error = secpolicy_zfs(cr);
463 		if (error != 0)
464 			error = dsl_deleg_access_impl(ds, perm, cr);
465 	}
466 	return (error);
467 }
468 
469 static int
470 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
471 {
472 	int error;
473 	dsl_dataset_t *ds;
474 	dsl_pool_t *dp;
475 
476 	/*
477 	 * First do a quick check for root in the global zone, which
478 	 * is allowed to do all write_perms.  This ensures that zfs_ioc_*
479 	 * will get to handle nonexistent datasets.
480 	 */
481 	if (INGLOBALZONE(curproc) && secpolicy_zfs(cr) == 0)
482 		return (0);
483 
484 	error = dsl_pool_hold(name, FTAG, &dp);
485 	if (error != 0)
486 		return (error);
487 
488 	error = dsl_dataset_hold(dp, name, FTAG, &ds);
489 	if (error != 0) {
490 		dsl_pool_rele(dp, FTAG);
491 		return (error);
492 	}
493 
494 	error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
495 
496 	dsl_dataset_rele(ds, FTAG);
497 	dsl_pool_rele(dp, FTAG);
498 	return (error);
499 }
500 
501 /*
502  * Policy for setting the security label property.
503  *
504  * Returns 0 for success, non-zero for access and other errors.
505  */
506 static int
507 zfs_set_slabel_policy(const char *name, const char *strval, cred_t *cr)
508 {
509 #ifdef HAVE_MLSLABEL
510 	char		ds_hexsl[MAXNAMELEN];
511 	bslabel_t	ds_sl, new_sl;
512 	boolean_t	new_default = FALSE;
513 	uint64_t	zoned;
514 	int		needed_priv = -1;
515 	int		error;
516 
517 	/* First get the existing dataset label. */
518 	error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
519 	    1, sizeof (ds_hexsl), &ds_hexsl, NULL);
520 	if (error != 0)
521 		return (SET_ERROR(EPERM));
522 
523 	if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
524 		new_default = TRUE;
525 
526 	/* The label must be translatable */
527 	if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
528 		return (SET_ERROR(EINVAL));
529 
530 	/*
531 	 * In a non-global zone, disallow attempts to set a label that
532 	 * doesn't match that of the zone; otherwise no other checks
533 	 * are needed.
534 	 */
535 	if (!INGLOBALZONE(curproc)) {
536 		if (new_default || !blequal(&new_sl, CR_SL(CRED())))
537 			return (SET_ERROR(EPERM));
538 		return (0);
539 	}
540 
541 	/*
542 	 * For global-zone datasets (i.e., those whose zoned property is
543 	 * "off", verify that the specified new label is valid for the
544 	 * global zone.
545 	 */
546 	if (dsl_prop_get_integer(name,
547 	    zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
548 		return (SET_ERROR(EPERM));
549 	if (!zoned) {
550 		if (zfs_check_global_label(name, strval) != 0)
551 			return (SET_ERROR(EPERM));
552 	}
553 
554 	/*
555 	 * If the existing dataset label is nondefault, check if the
556 	 * dataset is mounted (label cannot be changed while mounted).
557 	 * Get the zfsvfs_t; if there isn't one, then the dataset isn't
558 	 * mounted (or isn't a dataset, doesn't exist, ...).
559 	 */
560 	if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
561 		objset_t *os;
562 		static const char *setsl_tag = "setsl_tag";
563 
564 		/*
565 		 * Try to own the dataset; abort if there is any error,
566 		 * (e.g., already mounted, in use, or other error).
567 		 */
568 		error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE, B_TRUE,
569 		    setsl_tag, &os);
570 		if (error != 0)
571 			return (SET_ERROR(EPERM));
572 
573 		dmu_objset_disown(os, B_TRUE, setsl_tag);
574 
575 		if (new_default) {
576 			needed_priv = PRIV_FILE_DOWNGRADE_SL;
577 			goto out_check;
578 		}
579 
580 		if (hexstr_to_label(strval, &new_sl) != 0)
581 			return (SET_ERROR(EPERM));
582 
583 		if (blstrictdom(&ds_sl, &new_sl))
584 			needed_priv = PRIV_FILE_DOWNGRADE_SL;
585 		else if (blstrictdom(&new_sl, &ds_sl))
586 			needed_priv = PRIV_FILE_UPGRADE_SL;
587 	} else {
588 		/* dataset currently has a default label */
589 		if (!new_default)
590 			needed_priv = PRIV_FILE_UPGRADE_SL;
591 	}
592 
593 out_check:
594 	if (needed_priv != -1)
595 		return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
596 	return (0);
597 #else
598 	return (SET_ERROR(ENOTSUP));
599 #endif /* HAVE_MLSLABEL */
600 }
601 
602 static int
603 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
604     cred_t *cr)
605 {
606 	const char *strval;
607 
608 	/*
609 	 * Check permissions for special properties.
610 	 */
611 	switch (prop) {
612 	default:
613 		break;
614 	case ZFS_PROP_ZONED:
615 		/*
616 		 * Disallow setting of 'zoned' from within a local zone.
617 		 */
618 		if (!INGLOBALZONE(curproc))
619 			return (SET_ERROR(EPERM));
620 		break;
621 
622 	case ZFS_PROP_QUOTA:
623 	case ZFS_PROP_FILESYSTEM_LIMIT:
624 	case ZFS_PROP_SNAPSHOT_LIMIT:
625 		if (!INGLOBALZONE(curproc)) {
626 			uint64_t zoned;
627 			char setpoint[ZFS_MAX_DATASET_NAME_LEN];
628 			/*
629 			 * Unprivileged users are allowed to modify the
630 			 * limit on things *under* (ie. contained by)
631 			 * the thing they own.
632 			 */
633 			if (dsl_prop_get_integer(dsname,
634 			    zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, setpoint))
635 				return (SET_ERROR(EPERM));
636 			if (!zoned || strlen(dsname) <= strlen(setpoint))
637 				return (SET_ERROR(EPERM));
638 		}
639 		break;
640 
641 	case ZFS_PROP_MLSLABEL:
642 		if (!is_system_labeled())
643 			return (SET_ERROR(EPERM));
644 
645 		if (nvpair_value_string(propval, &strval) == 0) {
646 			int err;
647 
648 			err = zfs_set_slabel_policy(dsname, strval, CRED());
649 			if (err != 0)
650 				return (err);
651 		}
652 		break;
653 	}
654 
655 	return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
656 }
657 
658 static int
659 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
660 {
661 	/*
662 	 * permission to set permissions will be evaluated later in
663 	 * dsl_deleg_can_allow()
664 	 */
665 	(void) innvl;
666 	return (zfs_dozonecheck(zc->zc_name, cr));
667 }
668 
669 static int
670 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
671 {
672 	(void) innvl;
673 	return (zfs_secpolicy_write_perms(zc->zc_name,
674 	    ZFS_DELEG_PERM_ROLLBACK, cr));
675 }
676 
677 static int
678 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
679 {
680 	(void) innvl;
681 	dsl_pool_t *dp;
682 	dsl_dataset_t *ds;
683 	const char *cp;
684 	int error;
685 
686 	/*
687 	 * Generate the current snapshot name from the given objsetid, then
688 	 * use that name for the secpolicy/zone checks.
689 	 */
690 	cp = strchr(zc->zc_name, '@');
691 	if (cp == NULL)
692 		return (SET_ERROR(EINVAL));
693 	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
694 	if (error != 0)
695 		return (error);
696 
697 	error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
698 	if (error != 0) {
699 		dsl_pool_rele(dp, FTAG);
700 		return (error);
701 	}
702 
703 	dsl_dataset_name(ds, zc->zc_name);
704 
705 	error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
706 	    ZFS_DELEG_PERM_SEND, cr);
707 	dsl_dataset_rele(ds, FTAG);
708 	dsl_pool_rele(dp, FTAG);
709 
710 	return (error);
711 }
712 
713 static int
714 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
715 {
716 	(void) innvl;
717 	return (zfs_secpolicy_write_perms(zc->zc_name,
718 	    ZFS_DELEG_PERM_SEND, cr));
719 }
720 
721 static int
722 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
723 {
724 	(void) zc, (void) innvl, (void) cr;
725 	return (SET_ERROR(ENOTSUP));
726 }
727 
728 static int
729 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
730 {
731 	(void) zc, (void) innvl, (void) cr;
732 	return (SET_ERROR(ENOTSUP));
733 }
734 
735 static int
736 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
737 {
738 	char *cp;
739 
740 	/*
741 	 * Remove the @bla or /bla from the end of the name to get the parent.
742 	 */
743 	(void) strlcpy(parent, datasetname, parentsize);
744 	cp = strrchr(parent, '@');
745 	if (cp != NULL) {
746 		cp[0] = '\0';
747 	} else {
748 		cp = strrchr(parent, '/');
749 		if (cp == NULL)
750 			return (SET_ERROR(ENOENT));
751 		cp[0] = '\0';
752 	}
753 
754 	return (0);
755 }
756 
757 int
758 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
759 {
760 	int error;
761 
762 	if ((error = zfs_secpolicy_write_perms(name,
763 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
764 		return (error);
765 
766 	return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
767 }
768 
769 static int
770 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
771 {
772 	(void) innvl;
773 	return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
774 }
775 
776 /*
777  * Destroying snapshots with delegated permissions requires
778  * descendant mount and destroy permissions.
779  */
780 static int
781 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
782 {
783 	(void) zc;
784 	nvlist_t *snaps;
785 	nvpair_t *pair, *nextpair;
786 	int error = 0;
787 
788 	snaps = fnvlist_lookup_nvlist(innvl, "snaps");
789 
790 	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
791 	    pair = nextpair) {
792 		nextpair = nvlist_next_nvpair(snaps, pair);
793 		error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
794 		if (error == ENOENT) {
795 			/*
796 			 * Ignore any snapshots that don't exist (we consider
797 			 * them "already destroyed").  Remove the name from the
798 			 * nvl here in case the snapshot is created between
799 			 * now and when we try to destroy it (in which case
800 			 * we don't want to destroy it since we haven't
801 			 * checked for permission).
802 			 */
803 			fnvlist_remove_nvpair(snaps, pair);
804 			error = 0;
805 		}
806 		if (error != 0)
807 			break;
808 	}
809 
810 	return (error);
811 }
812 
813 int
814 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
815 {
816 	char	parentname[ZFS_MAX_DATASET_NAME_LEN];
817 	int	error;
818 
819 	if ((error = zfs_secpolicy_write_perms(from,
820 	    ZFS_DELEG_PERM_RENAME, cr)) != 0)
821 		return (error);
822 
823 	if ((error = zfs_secpolicy_write_perms(from,
824 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
825 		return (error);
826 
827 	if ((error = zfs_get_parent(to, parentname,
828 	    sizeof (parentname))) != 0)
829 		return (error);
830 
831 	if ((error = zfs_secpolicy_write_perms(parentname,
832 	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
833 		return (error);
834 
835 	if ((error = zfs_secpolicy_write_perms(parentname,
836 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
837 		return (error);
838 
839 	return (error);
840 }
841 
842 static int
843 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
844 {
845 	(void) innvl;
846 	return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
847 }
848 
849 static int
850 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
851 {
852 	(void) innvl;
853 	dsl_pool_t *dp;
854 	dsl_dataset_t *clone;
855 	int error;
856 
857 	error = zfs_secpolicy_write_perms(zc->zc_name,
858 	    ZFS_DELEG_PERM_PROMOTE, cr);
859 	if (error != 0)
860 		return (error);
861 
862 	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
863 	if (error != 0)
864 		return (error);
865 
866 	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
867 
868 	if (error == 0) {
869 		char parentname[ZFS_MAX_DATASET_NAME_LEN];
870 		dsl_dataset_t *origin = NULL;
871 		dsl_dir_t *dd;
872 		dd = clone->ds_dir;
873 
874 		error = dsl_dataset_hold_obj(dd->dd_pool,
875 		    dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin);
876 		if (error != 0) {
877 			dsl_dataset_rele(clone, FTAG);
878 			dsl_pool_rele(dp, FTAG);
879 			return (error);
880 		}
881 
882 		error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
883 		    ZFS_DELEG_PERM_MOUNT, cr);
884 
885 		dsl_dataset_name(origin, parentname);
886 		if (error == 0) {
887 			error = zfs_secpolicy_write_perms_ds(parentname, origin,
888 			    ZFS_DELEG_PERM_PROMOTE, cr);
889 		}
890 		dsl_dataset_rele(clone, FTAG);
891 		dsl_dataset_rele(origin, FTAG);
892 	}
893 	dsl_pool_rele(dp, FTAG);
894 	return (error);
895 }
896 
897 static int
898 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
899 {
900 	(void) innvl;
901 	int error;
902 
903 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
904 	    ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
905 		return (error);
906 
907 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
908 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
909 		return (error);
910 
911 	return (zfs_secpolicy_write_perms(zc->zc_name,
912 	    ZFS_DELEG_PERM_CREATE, cr));
913 }
914 
915 int
916 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
917 {
918 	return (zfs_secpolicy_write_perms(name,
919 	    ZFS_DELEG_PERM_SNAPSHOT, cr));
920 }
921 
922 /*
923  * Check for permission to create each snapshot in the nvlist.
924  */
925 static int
926 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
927 {
928 	(void) zc;
929 	nvlist_t *snaps;
930 	int error = 0;
931 	nvpair_t *pair;
932 
933 	snaps = fnvlist_lookup_nvlist(innvl, "snaps");
934 
935 	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
936 	    pair = nvlist_next_nvpair(snaps, pair)) {
937 		char *name = (char *)nvpair_name(pair);
938 		char *atp = strchr(name, '@');
939 
940 		if (atp == NULL) {
941 			error = SET_ERROR(EINVAL);
942 			break;
943 		}
944 		*atp = '\0';
945 		error = zfs_secpolicy_snapshot_perms(name, cr);
946 		*atp = '@';
947 		if (error != 0)
948 			break;
949 	}
950 	return (error);
951 }
952 
953 /*
954  * Check for permission to create each bookmark in the nvlist.
955  */
956 static int
957 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
958 {
959 	(void) zc;
960 	int error = 0;
961 
962 	for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
963 	    pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
964 		char *name = (char *)nvpair_name(pair);
965 		char *hashp = strchr(name, '#');
966 
967 		if (hashp == NULL) {
968 			error = SET_ERROR(EINVAL);
969 			break;
970 		}
971 		*hashp = '\0';
972 		error = zfs_secpolicy_write_perms(name,
973 		    ZFS_DELEG_PERM_BOOKMARK, cr);
974 		*hashp = '#';
975 		if (error != 0)
976 			break;
977 	}
978 	return (error);
979 }
980 
981 static int
982 zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
983 {
984 	(void) zc;
985 	nvpair_t *pair, *nextpair;
986 	int error = 0;
987 
988 	for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
989 	    pair = nextpair) {
990 		char *name = (char *)nvpair_name(pair);
991 		char *hashp = strchr(name, '#');
992 		nextpair = nvlist_next_nvpair(innvl, pair);
993 
994 		if (hashp == NULL) {
995 			error = SET_ERROR(EINVAL);
996 			break;
997 		}
998 
999 		*hashp = '\0';
1000 		error = zfs_secpolicy_write_perms(name,
1001 		    ZFS_DELEG_PERM_DESTROY, cr);
1002 		*hashp = '#';
1003 		if (error == ENOENT) {
1004 			/*
1005 			 * Ignore any filesystems that don't exist (we consider
1006 			 * their bookmarks "already destroyed").  Remove
1007 			 * the name from the nvl here in case the filesystem
1008 			 * is created between now and when we try to destroy
1009 			 * the bookmark (in which case we don't want to
1010 			 * destroy it since we haven't checked for permission).
1011 			 */
1012 			fnvlist_remove_nvpair(innvl, pair);
1013 			error = 0;
1014 		}
1015 		if (error != 0)
1016 			break;
1017 	}
1018 
1019 	return (error);
1020 }
1021 
1022 static int
1023 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1024 {
1025 	(void) zc, (void) innvl, (void) cr;
1026 	/*
1027 	 * Even root must have a proper TSD so that we know what pool
1028 	 * to log to.
1029 	 */
1030 	if (tsd_get(zfs_allow_log_key) == NULL)
1031 		return (SET_ERROR(EPERM));
1032 	return (0);
1033 }
1034 
1035 static int
1036 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1037 {
1038 	char		parentname[ZFS_MAX_DATASET_NAME_LEN];
1039 	int		error;
1040 	const char	*origin;
1041 
1042 	if ((error = zfs_get_parent(zc->zc_name, parentname,
1043 	    sizeof (parentname))) != 0)
1044 		return (error);
1045 
1046 	if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1047 	    (error = zfs_secpolicy_write_perms(origin,
1048 	    ZFS_DELEG_PERM_CLONE, cr)) != 0)
1049 		return (error);
1050 
1051 	if ((error = zfs_secpolicy_write_perms(parentname,
1052 	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
1053 		return (error);
1054 
1055 	return (zfs_secpolicy_write_perms(parentname,
1056 	    ZFS_DELEG_PERM_MOUNT, cr));
1057 }
1058 
1059 /*
1060  * Policy for pool operations - create/destroy pools, add vdevs, etc.  Requires
1061  * SYS_CONFIG privilege, which is not available in a local zone.
1062  */
1063 int
1064 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1065 {
1066 	(void) zc, (void) innvl;
1067 
1068 	if (secpolicy_sys_config(cr, B_FALSE) != 0)
1069 		return (SET_ERROR(EPERM));
1070 
1071 	return (0);
1072 }
1073 
1074 /*
1075  * Policy for object to name lookups.
1076  */
1077 static int
1078 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1079 {
1080 	(void) innvl;
1081 	int error;
1082 
1083 	if (secpolicy_sys_config(cr, B_FALSE) == 0)
1084 		return (0);
1085 
1086 	error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1087 	return (error);
1088 }
1089 
1090 /*
1091  * Policy for fault injection.  Requires all privileges.
1092  */
1093 static int
1094 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1095 {
1096 	(void) zc, (void) innvl;
1097 	return (secpolicy_zinject(cr));
1098 }
1099 
1100 static int
1101 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1102 {
1103 	(void) innvl;
1104 	zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1105 
1106 	if (prop == ZPROP_USERPROP) {
1107 		if (!zfs_prop_user(zc->zc_value))
1108 			return (SET_ERROR(EINVAL));
1109 		return (zfs_secpolicy_write_perms(zc->zc_name,
1110 		    ZFS_DELEG_PERM_USERPROP, cr));
1111 	} else {
1112 		return (zfs_secpolicy_setprop(zc->zc_name, prop,
1113 		    NULL, cr));
1114 	}
1115 }
1116 
1117 static int
1118 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1119 {
1120 	int err = zfs_secpolicy_read(zc, innvl, cr);
1121 	if (err)
1122 		return (err);
1123 
1124 	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1125 		return (SET_ERROR(EINVAL));
1126 
1127 	if (zc->zc_value[0] == 0) {
1128 		/*
1129 		 * They are asking about a posix uid/gid.  If it's
1130 		 * themself, allow it.
1131 		 */
1132 		if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1133 		    zc->zc_objset_type == ZFS_PROP_USERQUOTA ||
1134 		    zc->zc_objset_type == ZFS_PROP_USEROBJUSED ||
1135 		    zc->zc_objset_type == ZFS_PROP_USEROBJQUOTA) {
1136 			if (zc->zc_guid == crgetuid(cr))
1137 				return (0);
1138 		} else if (zc->zc_objset_type == ZFS_PROP_GROUPUSED ||
1139 		    zc->zc_objset_type == ZFS_PROP_GROUPQUOTA ||
1140 		    zc->zc_objset_type == ZFS_PROP_GROUPOBJUSED ||
1141 		    zc->zc_objset_type == ZFS_PROP_GROUPOBJQUOTA) {
1142 			if (groupmember(zc->zc_guid, cr))
1143 				return (0);
1144 		}
1145 		/* else is for project quota/used */
1146 	}
1147 
1148 	return (zfs_secpolicy_write_perms(zc->zc_name,
1149 	    userquota_perms[zc->zc_objset_type], cr));
1150 }
1151 
1152 static int
1153 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1154 {
1155 	int err = zfs_secpolicy_read(zc, innvl, cr);
1156 	if (err)
1157 		return (err);
1158 
1159 	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1160 		return (SET_ERROR(EINVAL));
1161 
1162 	return (zfs_secpolicy_write_perms(zc->zc_name,
1163 	    userquota_perms[zc->zc_objset_type], cr));
1164 }
1165 
1166 static int
1167 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1168 {
1169 	(void) innvl;
1170 	return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1171 	    NULL, cr));
1172 }
1173 
1174 static int
1175 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1176 {
1177 	(void) zc;
1178 	nvpair_t *pair;
1179 	nvlist_t *holds;
1180 	int error;
1181 
1182 	holds = fnvlist_lookup_nvlist(innvl, "holds");
1183 
1184 	for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1185 	    pair = nvlist_next_nvpair(holds, pair)) {
1186 		char fsname[ZFS_MAX_DATASET_NAME_LEN];
1187 		error = dmu_fsname(nvpair_name(pair), fsname);
1188 		if (error != 0)
1189 			return (error);
1190 		error = zfs_secpolicy_write_perms(fsname,
1191 		    ZFS_DELEG_PERM_HOLD, cr);
1192 		if (error != 0)
1193 			return (error);
1194 	}
1195 	return (0);
1196 }
1197 
1198 static int
1199 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1200 {
1201 	(void) zc;
1202 	nvpair_t *pair;
1203 	int error;
1204 
1205 	for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1206 	    pair = nvlist_next_nvpair(innvl, pair)) {
1207 		char fsname[ZFS_MAX_DATASET_NAME_LEN];
1208 		error = dmu_fsname(nvpair_name(pair), fsname);
1209 		if (error != 0)
1210 			return (error);
1211 		error = zfs_secpolicy_write_perms(fsname,
1212 		    ZFS_DELEG_PERM_RELEASE, cr);
1213 		if (error != 0)
1214 			return (error);
1215 	}
1216 	return (0);
1217 }
1218 
1219 /*
1220  * Policy for allowing temporary snapshots to be taken or released
1221  */
1222 static int
1223 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1224 {
1225 	/*
1226 	 * A temporary snapshot is the same as a snapshot,
1227 	 * hold, destroy and release all rolled into one.
1228 	 * Delegated diff alone is sufficient that we allow this.
1229 	 */
1230 	int error;
1231 
1232 	if (zfs_secpolicy_write_perms(zc->zc_name,
1233 	    ZFS_DELEG_PERM_DIFF, cr) == 0)
1234 		return (0);
1235 
1236 	error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1237 
1238 	if (innvl != NULL) {
1239 		if (error == 0)
1240 			error = zfs_secpolicy_hold(zc, innvl, cr);
1241 		if (error == 0)
1242 			error = zfs_secpolicy_release(zc, innvl, cr);
1243 		if (error == 0)
1244 			error = zfs_secpolicy_destroy(zc, innvl, cr);
1245 	}
1246 	return (error);
1247 }
1248 
1249 static int
1250 zfs_secpolicy_load_key(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1251 {
1252 	return (zfs_secpolicy_write_perms(zc->zc_name,
1253 	    ZFS_DELEG_PERM_LOAD_KEY, cr));
1254 }
1255 
1256 static int
1257 zfs_secpolicy_change_key(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1258 {
1259 	return (zfs_secpolicy_write_perms(zc->zc_name,
1260 	    ZFS_DELEG_PERM_CHANGE_KEY, cr));
1261 }
1262 
1263 /*
1264  * Returns the nvlist as specified by the user in the zfs_cmd_t.
1265  */
1266 static int
1267 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1268 {
1269 	char *packed;
1270 	int error;
1271 	nvlist_t *list = NULL;
1272 
1273 	/*
1274 	 * Read in and unpack the user-supplied nvlist.
1275 	 */
1276 	if (size == 0)
1277 		return (SET_ERROR(EINVAL));
1278 
1279 	packed = vmem_alloc(size, KM_SLEEP);
1280 
1281 	if (ddi_copyin((void *)(uintptr_t)nvl, packed, size, iflag) != 0) {
1282 		vmem_free(packed, size);
1283 		return (SET_ERROR(EFAULT));
1284 	}
1285 
1286 	if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1287 		vmem_free(packed, size);
1288 		return (error);
1289 	}
1290 
1291 	vmem_free(packed, size);
1292 
1293 	*nvp = list;
1294 	return (0);
1295 }
1296 
1297 /*
1298  * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1299  * Entries will be removed from the end of the nvlist, and one int32 entry
1300  * named "N_MORE_ERRORS" will be added indicating how many entries were
1301  * removed.
1302  */
1303 static int
1304 nvlist_smush(nvlist_t *errors, size_t max)
1305 {
1306 	size_t size;
1307 
1308 	size = fnvlist_size(errors);
1309 
1310 	if (size > max) {
1311 		nvpair_t *more_errors;
1312 		int n = 0;
1313 
1314 		if (max < 1024)
1315 			return (SET_ERROR(ENOMEM));
1316 
1317 		fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1318 		more_errors = nvlist_prev_nvpair(errors, NULL);
1319 
1320 		do {
1321 			nvpair_t *pair = nvlist_prev_nvpair(errors,
1322 			    more_errors);
1323 			fnvlist_remove_nvpair(errors, pair);
1324 			n++;
1325 			size = fnvlist_size(errors);
1326 		} while (size > max);
1327 
1328 		fnvlist_remove_nvpair(errors, more_errors);
1329 		fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1330 		ASSERT3U(fnvlist_size(errors), <=, max);
1331 	}
1332 
1333 	return (0);
1334 }
1335 
1336 static int
1337 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1338 {
1339 	char *packed = NULL;
1340 	int error = 0;
1341 	size_t size;
1342 
1343 	size = fnvlist_size(nvl);
1344 
1345 	if (size > zc->zc_nvlist_dst_size) {
1346 		error = SET_ERROR(ENOMEM);
1347 	} else {
1348 		packed = fnvlist_pack(nvl, &size);
1349 		if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1350 		    size, zc->zc_iflags) != 0)
1351 			error = SET_ERROR(EFAULT);
1352 		fnvlist_pack_free(packed, size);
1353 	}
1354 
1355 	zc->zc_nvlist_dst_size = size;
1356 	zc->zc_nvlist_dst_filled = B_TRUE;
1357 	return (error);
1358 }
1359 
1360 int
1361 getzfsvfs_impl(objset_t *os, zfsvfs_t **zfvp)
1362 {
1363 	int error = 0;
1364 	if (dmu_objset_type(os) != DMU_OST_ZFS) {
1365 		return (SET_ERROR(EINVAL));
1366 	}
1367 
1368 	mutex_enter(&os->os_user_ptr_lock);
1369 	*zfvp = dmu_objset_get_user(os);
1370 	/* bump s_active only when non-zero to prevent umount race */
1371 	error = zfs_vfs_ref(zfvp);
1372 	mutex_exit(&os->os_user_ptr_lock);
1373 	return (error);
1374 }
1375 
1376 int
1377 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1378 {
1379 	objset_t *os;
1380 	int error;
1381 
1382 	error = dmu_objset_hold(dsname, FTAG, &os);
1383 	if (error != 0)
1384 		return (error);
1385 
1386 	error = getzfsvfs_impl(os, zfvp);
1387 	dmu_objset_rele(os, FTAG);
1388 	return (error);
1389 }
1390 
1391 /*
1392  * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1393  * case its z_sb will be NULL, and it will be opened as the owner.
1394  * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1395  * which prevents all inode ops from running.
1396  */
1397 static int
1398 zfsvfs_hold(const char *name, const void *tag, zfsvfs_t **zfvp,
1399     boolean_t writer)
1400 {
1401 	int error = 0;
1402 
1403 	if (getzfsvfs(name, zfvp) != 0)
1404 		error = zfsvfs_create(name, B_FALSE, zfvp);
1405 	if (error == 0) {
1406 		if (writer)
1407 			ZFS_TEARDOWN_ENTER_WRITE(*zfvp, tag);
1408 		else
1409 			ZFS_TEARDOWN_ENTER_READ(*zfvp, tag);
1410 		if ((*zfvp)->z_unmounted) {
1411 			/*
1412 			 * XXX we could probably try again, since the unmounting
1413 			 * thread should be just about to disassociate the
1414 			 * objset from the zfsvfs.
1415 			 */
1416 			ZFS_TEARDOWN_EXIT(*zfvp, tag);
1417 			return (SET_ERROR(EBUSY));
1418 		}
1419 	}
1420 	return (error);
1421 }
1422 
1423 static void
1424 zfsvfs_rele(zfsvfs_t *zfsvfs, const void *tag)
1425 {
1426 	ZFS_TEARDOWN_EXIT(zfsvfs, tag);
1427 
1428 	if (zfs_vfs_held(zfsvfs)) {
1429 		zfs_vfs_rele(zfsvfs);
1430 	} else {
1431 		dmu_objset_disown(zfsvfs->z_os, B_TRUE, zfsvfs);
1432 		zfsvfs_free(zfsvfs);
1433 	}
1434 }
1435 
1436 static int
1437 zfs_ioc_pool_create(zfs_cmd_t *zc)
1438 {
1439 	int error;
1440 	nvlist_t *config, *props = NULL;
1441 	nvlist_t *rootprops = NULL;
1442 	nvlist_t *zplprops = NULL;
1443 	dsl_crypto_params_t *dcp = NULL;
1444 	const char *spa_name = zc->zc_name;
1445 	boolean_t unload_wkey = B_TRUE;
1446 
1447 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1448 	    zc->zc_iflags, &config)))
1449 		return (error);
1450 
1451 	if (zc->zc_nvlist_src_size != 0 && (error =
1452 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1453 	    zc->zc_iflags, &props))) {
1454 		nvlist_free(config);
1455 		return (error);
1456 	}
1457 
1458 	if (props) {
1459 		nvlist_t *nvl = NULL;
1460 		nvlist_t *hidden_args = NULL;
1461 		uint64_t version = SPA_VERSION;
1462 		const char *tname;
1463 
1464 		(void) nvlist_lookup_uint64(props,
1465 		    zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1466 		if (!SPA_VERSION_IS_SUPPORTED(version)) {
1467 			error = SET_ERROR(EINVAL);
1468 			goto pool_props_bad;
1469 		}
1470 		(void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1471 		if (nvl) {
1472 			error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1473 			if (error != 0)
1474 				goto pool_props_bad;
1475 			(void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1476 		}
1477 
1478 		(void) nvlist_lookup_nvlist(props, ZPOOL_HIDDEN_ARGS,
1479 		    &hidden_args);
1480 		error = dsl_crypto_params_create_nvlist(DCP_CMD_NONE,
1481 		    rootprops, hidden_args, &dcp);
1482 		if (error != 0)
1483 			goto pool_props_bad;
1484 		(void) nvlist_remove_all(props, ZPOOL_HIDDEN_ARGS);
1485 
1486 		VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1487 		error = zfs_fill_zplprops_root(version, rootprops,
1488 		    zplprops, NULL);
1489 		if (error != 0)
1490 			goto pool_props_bad;
1491 
1492 		if (nvlist_lookup_string(props,
1493 		    zpool_prop_to_name(ZPOOL_PROP_TNAME), &tname) == 0)
1494 			spa_name = tname;
1495 	}
1496 
1497 	error = spa_create(zc->zc_name, config, props, zplprops, dcp);
1498 
1499 	/*
1500 	 * Set the remaining root properties
1501 	 */
1502 	if (!error && (error = zfs_set_prop_nvlist(spa_name,
1503 	    ZPROP_SRC_LOCAL, rootprops, NULL)) != 0) {
1504 		(void) spa_destroy(spa_name);
1505 		unload_wkey = B_FALSE; /* spa_destroy() unloads wrapping keys */
1506 	}
1507 
1508 pool_props_bad:
1509 	nvlist_free(rootprops);
1510 	nvlist_free(zplprops);
1511 	nvlist_free(config);
1512 	nvlist_free(props);
1513 	dsl_crypto_params_free(dcp, unload_wkey && !!error);
1514 
1515 	return (error);
1516 }
1517 
1518 static int
1519 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1520 {
1521 	int error;
1522 	zfs_log_history(zc);
1523 	error = spa_destroy(zc->zc_name);
1524 
1525 	return (error);
1526 }
1527 
1528 static int
1529 zfs_ioc_pool_import(zfs_cmd_t *zc)
1530 {
1531 	nvlist_t *config, *props = NULL;
1532 	uint64_t guid;
1533 	int error;
1534 
1535 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1536 	    zc->zc_iflags, &config)) != 0)
1537 		return (error);
1538 
1539 	if (zc->zc_nvlist_src_size != 0 && (error =
1540 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1541 	    zc->zc_iflags, &props))) {
1542 		nvlist_free(config);
1543 		return (error);
1544 	}
1545 
1546 	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1547 	    guid != zc->zc_guid)
1548 		error = SET_ERROR(EINVAL);
1549 	else
1550 		error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1551 
1552 	if (zc->zc_nvlist_dst != 0) {
1553 		int err;
1554 
1555 		if ((err = put_nvlist(zc, config)) != 0)
1556 			error = err;
1557 	}
1558 
1559 	nvlist_free(config);
1560 	nvlist_free(props);
1561 
1562 	return (error);
1563 }
1564 
1565 static int
1566 zfs_ioc_pool_export(zfs_cmd_t *zc)
1567 {
1568 	int error;
1569 	boolean_t force = (boolean_t)zc->zc_cookie;
1570 	boolean_t hardforce = (boolean_t)zc->zc_guid;
1571 
1572 	zfs_log_history(zc);
1573 	error = spa_export(zc->zc_name, NULL, force, hardforce);
1574 
1575 	return (error);
1576 }
1577 
1578 static int
1579 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1580 {
1581 	nvlist_t *configs;
1582 	int error;
1583 
1584 	error = spa_all_configs(&zc->zc_cookie, &configs);
1585 	if (error)
1586 		return (error);
1587 
1588 	error = put_nvlist(zc, configs);
1589 
1590 	nvlist_free(configs);
1591 
1592 	return (error);
1593 }
1594 
1595 /*
1596  * inputs:
1597  * zc_name		name of the pool
1598  *
1599  * outputs:
1600  * zc_cookie		real errno
1601  * zc_nvlist_dst	config nvlist
1602  * zc_nvlist_dst_size	size of config nvlist
1603  */
1604 static int
1605 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1606 {
1607 	nvlist_t *config;
1608 	int error;
1609 	int ret = 0;
1610 
1611 	error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1612 	    sizeof (zc->zc_value));
1613 
1614 	if (config != NULL) {
1615 		ret = put_nvlist(zc, config);
1616 		nvlist_free(config);
1617 
1618 		/*
1619 		 * The config may be present even if 'error' is non-zero.
1620 		 * In this case we return success, and preserve the real errno
1621 		 * in 'zc_cookie'.
1622 		 */
1623 		zc->zc_cookie = error;
1624 	} else {
1625 		ret = error;
1626 	}
1627 
1628 	return (ret);
1629 }
1630 
1631 /*
1632  * Try to import the given pool, returning pool stats as appropriate so that
1633  * user land knows which devices are available and overall pool health.
1634  */
1635 static int
1636 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1637 {
1638 	nvlist_t *tryconfig, *config = NULL;
1639 	int error;
1640 
1641 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1642 	    zc->zc_iflags, &tryconfig)) != 0)
1643 		return (error);
1644 
1645 	config = spa_tryimport(tryconfig);
1646 
1647 	nvlist_free(tryconfig);
1648 
1649 	if (config == NULL)
1650 		return (SET_ERROR(EINVAL));
1651 
1652 	error = put_nvlist(zc, config);
1653 	nvlist_free(config);
1654 
1655 	return (error);
1656 }
1657 
1658 /*
1659  * inputs:
1660  * zc_name              name of the pool
1661  * zc_cookie            scan func (pool_scan_func_t)
1662  * zc_flags             scrub pause/resume flag (pool_scrub_cmd_t)
1663  */
1664 static int
1665 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1666 {
1667 	spa_t *spa;
1668 	int error;
1669 
1670 	if (zc->zc_flags >= POOL_SCRUB_FLAGS_END)
1671 		return (SET_ERROR(EINVAL));
1672 
1673 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1674 		return (error);
1675 
1676 	if (zc->zc_flags == POOL_SCRUB_PAUSE)
1677 		error = spa_scrub_pause_resume(spa, POOL_SCRUB_PAUSE);
1678 	else if (zc->zc_cookie == POOL_SCAN_NONE)
1679 		error = spa_scan_stop(spa);
1680 	else
1681 		error = spa_scan(spa, zc->zc_cookie);
1682 
1683 	spa_close(spa, FTAG);
1684 
1685 	return (error);
1686 }
1687 
1688 /*
1689  * inputs:
1690  * poolname             name of the pool
1691  * scan_type            scan func (pool_scan_func_t)
1692  * scan_command         scrub pause/resume flag (pool_scrub_cmd_t)
1693  */
1694 static const zfs_ioc_key_t zfs_keys_pool_scrub[] = {
1695 	{"scan_type",		DATA_TYPE_UINT64,	0},
1696 	{"scan_command",	DATA_TYPE_UINT64,	0},
1697 };
1698 
1699 static int
1700 zfs_ioc_pool_scrub(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
1701 {
1702 	spa_t *spa;
1703 	int error;
1704 	uint64_t scan_type, scan_cmd;
1705 
1706 	if (nvlist_lookup_uint64(innvl, "scan_type", &scan_type) != 0)
1707 		return (SET_ERROR(EINVAL));
1708 	if (nvlist_lookup_uint64(innvl, "scan_command", &scan_cmd) != 0)
1709 		return (SET_ERROR(EINVAL));
1710 
1711 	if (scan_cmd >= POOL_SCRUB_FLAGS_END)
1712 		return (SET_ERROR(EINVAL));
1713 
1714 	if ((error = spa_open(poolname, &spa, FTAG)) != 0)
1715 		return (error);
1716 
1717 	if (scan_cmd == POOL_SCRUB_PAUSE) {
1718 		error = spa_scrub_pause_resume(spa, POOL_SCRUB_PAUSE);
1719 	} else if (scan_type == POOL_SCAN_NONE) {
1720 		error = spa_scan_stop(spa);
1721 	} else {
1722 		error = spa_scan(spa, scan_type);
1723 	}
1724 
1725 	spa_close(spa, FTAG);
1726 	return (error);
1727 }
1728 
1729 static int
1730 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1731 {
1732 	spa_t *spa;
1733 	int error;
1734 
1735 	error = spa_open(zc->zc_name, &spa, FTAG);
1736 	if (error == 0) {
1737 		spa_freeze(spa);
1738 		spa_close(spa, FTAG);
1739 	}
1740 	return (error);
1741 }
1742 
1743 static int
1744 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1745 {
1746 	spa_t *spa;
1747 	int error;
1748 
1749 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1750 		return (error);
1751 
1752 	if (zc->zc_cookie < spa_version(spa) ||
1753 	    !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1754 		spa_close(spa, FTAG);
1755 		return (SET_ERROR(EINVAL));
1756 	}
1757 
1758 	spa_upgrade(spa, zc->zc_cookie);
1759 	spa_close(spa, FTAG);
1760 
1761 	return (error);
1762 }
1763 
1764 static int
1765 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1766 {
1767 	spa_t *spa;
1768 	char *hist_buf;
1769 	uint64_t size;
1770 	int error;
1771 
1772 	if ((size = zc->zc_history_len) == 0)
1773 		return (SET_ERROR(EINVAL));
1774 
1775 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1776 		return (error);
1777 
1778 	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1779 		spa_close(spa, FTAG);
1780 		return (SET_ERROR(ENOTSUP));
1781 	}
1782 
1783 	hist_buf = vmem_alloc(size, KM_SLEEP);
1784 	if ((error = spa_history_get(spa, &zc->zc_history_offset,
1785 	    &zc->zc_history_len, hist_buf)) == 0) {
1786 		error = ddi_copyout(hist_buf,
1787 		    (void *)(uintptr_t)zc->zc_history,
1788 		    zc->zc_history_len, zc->zc_iflags);
1789 	}
1790 
1791 	spa_close(spa, FTAG);
1792 	vmem_free(hist_buf, size);
1793 	return (error);
1794 }
1795 
1796 static int
1797 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1798 {
1799 	spa_t *spa;
1800 	int error;
1801 
1802 	error = spa_open(zc->zc_name, &spa, FTAG);
1803 	if (error == 0) {
1804 		error = spa_change_guid(spa);
1805 		spa_close(spa, FTAG);
1806 	}
1807 	return (error);
1808 }
1809 
1810 static int
1811 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1812 {
1813 	return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1814 }
1815 
1816 /*
1817  * inputs:
1818  * zc_name		name of filesystem
1819  * zc_obj		object to find
1820  *
1821  * outputs:
1822  * zc_value		name of object
1823  */
1824 static int
1825 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1826 {
1827 	objset_t *os;
1828 	int error;
1829 
1830 	/* XXX reading from objset not owned */
1831 	if ((error = dmu_objset_hold_flags(zc->zc_name, B_TRUE,
1832 	    FTAG, &os)) != 0)
1833 		return (error);
1834 	if (dmu_objset_type(os) != DMU_OST_ZFS) {
1835 		dmu_objset_rele_flags(os, B_TRUE, FTAG);
1836 		return (SET_ERROR(EINVAL));
1837 	}
1838 	error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1839 	    sizeof (zc->zc_value));
1840 	dmu_objset_rele_flags(os, B_TRUE, FTAG);
1841 
1842 	return (error);
1843 }
1844 
1845 /*
1846  * inputs:
1847  * zc_name		name of filesystem
1848  * zc_obj		object to find
1849  *
1850  * outputs:
1851  * zc_stat		stats on object
1852  * zc_value		path to object
1853  */
1854 static int
1855 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1856 {
1857 	objset_t *os;
1858 	int error;
1859 
1860 	/* XXX reading from objset not owned */
1861 	if ((error = dmu_objset_hold_flags(zc->zc_name, B_TRUE,
1862 	    FTAG, &os)) != 0)
1863 		return (error);
1864 	if (dmu_objset_type(os) != DMU_OST_ZFS) {
1865 		dmu_objset_rele_flags(os, B_TRUE, FTAG);
1866 		return (SET_ERROR(EINVAL));
1867 	}
1868 	error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1869 	    sizeof (zc->zc_value));
1870 	dmu_objset_rele_flags(os, B_TRUE, FTAG);
1871 
1872 	return (error);
1873 }
1874 
1875 static int
1876 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1877 {
1878 	spa_t *spa;
1879 	int error;
1880 	nvlist_t *config;
1881 
1882 	error = spa_open(zc->zc_name, &spa, FTAG);
1883 	if (error != 0)
1884 		return (error);
1885 
1886 	error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1887 	    zc->zc_iflags, &config);
1888 	if (error == 0) {
1889 		error = spa_vdev_add(spa, config);
1890 		nvlist_free(config);
1891 	}
1892 	spa_close(spa, FTAG);
1893 	return (error);
1894 }
1895 
1896 /*
1897  * inputs:
1898  * zc_name		name of the pool
1899  * zc_guid		guid of vdev to remove
1900  * zc_cookie		cancel removal
1901  */
1902 static int
1903 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1904 {
1905 	spa_t *spa;
1906 	int error;
1907 
1908 	error = spa_open(zc->zc_name, &spa, FTAG);
1909 	if (error != 0)
1910 		return (error);
1911 	if (zc->zc_cookie != 0) {
1912 		error = spa_vdev_remove_cancel(spa);
1913 	} else {
1914 		error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1915 	}
1916 	spa_close(spa, FTAG);
1917 	return (error);
1918 }
1919 
1920 static int
1921 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1922 {
1923 	spa_t *spa;
1924 	int error;
1925 	vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1926 
1927 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1928 		return (error);
1929 	switch (zc->zc_cookie) {
1930 	case VDEV_STATE_ONLINE:
1931 		error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1932 		break;
1933 
1934 	case VDEV_STATE_OFFLINE:
1935 		error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1936 		break;
1937 
1938 	case VDEV_STATE_FAULTED:
1939 		if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1940 		    zc->zc_obj != VDEV_AUX_EXTERNAL &&
1941 		    zc->zc_obj != VDEV_AUX_EXTERNAL_PERSIST)
1942 			zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1943 
1944 		error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1945 		break;
1946 
1947 	case VDEV_STATE_DEGRADED:
1948 		if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1949 		    zc->zc_obj != VDEV_AUX_EXTERNAL)
1950 			zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1951 
1952 		error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1953 		break;
1954 
1955 	case VDEV_STATE_REMOVED:
1956 		error = vdev_remove_wanted(spa, zc->zc_guid);
1957 		break;
1958 
1959 	default:
1960 		error = SET_ERROR(EINVAL);
1961 	}
1962 	zc->zc_cookie = newstate;
1963 	spa_close(spa, FTAG);
1964 	return (error);
1965 }
1966 
1967 static int
1968 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1969 {
1970 	spa_t *spa;
1971 	nvlist_t *config;
1972 	int replacing = zc->zc_cookie;
1973 	int rebuild = zc->zc_simple;
1974 	int error;
1975 
1976 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1977 		return (error);
1978 
1979 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1980 	    zc->zc_iflags, &config)) == 0) {
1981 		error = spa_vdev_attach(spa, zc->zc_guid, config, replacing,
1982 		    rebuild);
1983 		nvlist_free(config);
1984 	}
1985 
1986 	spa_close(spa, FTAG);
1987 	return (error);
1988 }
1989 
1990 static int
1991 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1992 {
1993 	spa_t *spa;
1994 	int error;
1995 
1996 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1997 		return (error);
1998 
1999 	error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
2000 
2001 	spa_close(spa, FTAG);
2002 	return (error);
2003 }
2004 
2005 static int
2006 zfs_ioc_vdev_split(zfs_cmd_t *zc)
2007 {
2008 	spa_t *spa;
2009 	nvlist_t *config, *props = NULL;
2010 	int error;
2011 	boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
2012 
2013 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2014 		return (error);
2015 
2016 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
2017 	    zc->zc_iflags, &config))) {
2018 		spa_close(spa, FTAG);
2019 		return (error);
2020 	}
2021 
2022 	if (zc->zc_nvlist_src_size != 0 && (error =
2023 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2024 	    zc->zc_iflags, &props))) {
2025 		spa_close(spa, FTAG);
2026 		nvlist_free(config);
2027 		return (error);
2028 	}
2029 
2030 	error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
2031 
2032 	spa_close(spa, FTAG);
2033 
2034 	nvlist_free(config);
2035 	nvlist_free(props);
2036 
2037 	return (error);
2038 }
2039 
2040 static int
2041 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
2042 {
2043 	spa_t *spa;
2044 	const char *path = zc->zc_value;
2045 	uint64_t guid = zc->zc_guid;
2046 	int error;
2047 
2048 	error = spa_open(zc->zc_name, &spa, FTAG);
2049 	if (error != 0)
2050 		return (error);
2051 
2052 	error = spa_vdev_setpath(spa, guid, path);
2053 	spa_close(spa, FTAG);
2054 	return (error);
2055 }
2056 
2057 static int
2058 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
2059 {
2060 	spa_t *spa;
2061 	const char *fru = zc->zc_value;
2062 	uint64_t guid = zc->zc_guid;
2063 	int error;
2064 
2065 	error = spa_open(zc->zc_name, &spa, FTAG);
2066 	if (error != 0)
2067 		return (error);
2068 
2069 	error = spa_vdev_setfru(spa, guid, fru);
2070 	spa_close(spa, FTAG);
2071 	return (error);
2072 }
2073 
2074 static int
2075 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
2076 {
2077 	int error = 0;
2078 	nvlist_t *nv;
2079 
2080 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2081 
2082 	if (!zc->zc_simple && zc->zc_nvlist_dst != 0 &&
2083 	    (error = dsl_prop_get_all(os, &nv)) == 0) {
2084 		dmu_objset_stats(os, nv);
2085 		/*
2086 		 * NB: zvol_get_stats() will read the objset contents,
2087 		 * which we aren't supposed to do with a
2088 		 * DS_MODE_USER hold, because it could be
2089 		 * inconsistent.  So this is a bit of a workaround...
2090 		 * XXX reading without owning
2091 		 */
2092 		if (!zc->zc_objset_stats.dds_inconsistent &&
2093 		    dmu_objset_type(os) == DMU_OST_ZVOL) {
2094 			error = zvol_get_stats(os, nv);
2095 			if (error == EIO) {
2096 				nvlist_free(nv);
2097 				return (error);
2098 			}
2099 			VERIFY0(error);
2100 		}
2101 		if (error == 0)
2102 			error = put_nvlist(zc, nv);
2103 		nvlist_free(nv);
2104 	}
2105 
2106 	return (error);
2107 }
2108 
2109 /*
2110  * inputs:
2111  * zc_name		name of filesystem
2112  * zc_nvlist_dst_size	size of buffer for property nvlist
2113  *
2114  * outputs:
2115  * zc_objset_stats	stats
2116  * zc_nvlist_dst	property nvlist
2117  * zc_nvlist_dst_size	size of property nvlist
2118  */
2119 static int
2120 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2121 {
2122 	objset_t *os;
2123 	int error;
2124 
2125 	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2126 	if (error == 0) {
2127 		error = zfs_ioc_objset_stats_impl(zc, os);
2128 		dmu_objset_rele(os, FTAG);
2129 	}
2130 
2131 	return (error);
2132 }
2133 
2134 /*
2135  * inputs:
2136  * zc_name		name of filesystem
2137  * zc_nvlist_dst_size	size of buffer for property nvlist
2138  *
2139  * outputs:
2140  * zc_nvlist_dst	received property nvlist
2141  * zc_nvlist_dst_size	size of received property nvlist
2142  *
2143  * Gets received properties (distinct from local properties on or after
2144  * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2145  * local property values.
2146  */
2147 static int
2148 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2149 {
2150 	int error = 0;
2151 	nvlist_t *nv;
2152 
2153 	/*
2154 	 * Without this check, we would return local property values if the
2155 	 * caller has not already received properties on or after
2156 	 * SPA_VERSION_RECVD_PROPS.
2157 	 */
2158 	if (!dsl_prop_get_hasrecvd(zc->zc_name))
2159 		return (SET_ERROR(ENOTSUP));
2160 
2161 	if (zc->zc_nvlist_dst != 0 &&
2162 	    (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2163 		error = put_nvlist(zc, nv);
2164 		nvlist_free(nv);
2165 	}
2166 
2167 	return (error);
2168 }
2169 
2170 static int
2171 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2172 {
2173 	uint64_t value;
2174 	int error;
2175 
2176 	/*
2177 	 * zfs_get_zplprop() will either find a value or give us
2178 	 * the default value (if there is one).
2179 	 */
2180 	if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2181 		return (error);
2182 	VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2183 	return (0);
2184 }
2185 
2186 /*
2187  * inputs:
2188  * zc_name		name of filesystem
2189  * zc_nvlist_dst_size	size of buffer for zpl property nvlist
2190  *
2191  * outputs:
2192  * zc_nvlist_dst	zpl property nvlist
2193  * zc_nvlist_dst_size	size of zpl property nvlist
2194  */
2195 static int
2196 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2197 {
2198 	objset_t *os;
2199 	int err;
2200 
2201 	/* XXX reading without owning */
2202 	if ((err = dmu_objset_hold(zc->zc_name, FTAG, &os)))
2203 		return (err);
2204 
2205 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2206 
2207 	/*
2208 	 * NB: nvl_add_zplprop() will read the objset contents,
2209 	 * which we aren't supposed to do with a DS_MODE_USER
2210 	 * hold, because it could be inconsistent.
2211 	 */
2212 	if (zc->zc_nvlist_dst != 0 &&
2213 	    !zc->zc_objset_stats.dds_inconsistent &&
2214 	    dmu_objset_type(os) == DMU_OST_ZFS) {
2215 		nvlist_t *nv;
2216 
2217 		VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2218 		if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2219 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2220 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2221 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2222 			err = put_nvlist(zc, nv);
2223 		nvlist_free(nv);
2224 	} else {
2225 		err = SET_ERROR(ENOENT);
2226 	}
2227 	dmu_objset_rele(os, FTAG);
2228 	return (err);
2229 }
2230 
2231 /*
2232  * inputs:
2233  * zc_name		name of filesystem
2234  * zc_cookie		zap cursor
2235  * zc_nvlist_dst_size	size of buffer for property nvlist
2236  *
2237  * outputs:
2238  * zc_name		name of next filesystem
2239  * zc_cookie		zap cursor
2240  * zc_objset_stats	stats
2241  * zc_nvlist_dst	property nvlist
2242  * zc_nvlist_dst_size	size of property nvlist
2243  */
2244 static int
2245 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2246 {
2247 	objset_t *os;
2248 	int error;
2249 	char *p;
2250 	size_t orig_len = strlen(zc->zc_name);
2251 
2252 top:
2253 	if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os))) {
2254 		if (error == ENOENT)
2255 			error = SET_ERROR(ESRCH);
2256 		return (error);
2257 	}
2258 
2259 	p = strrchr(zc->zc_name, '/');
2260 	if (p == NULL || p[1] != '\0')
2261 		(void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2262 	p = zc->zc_name + strlen(zc->zc_name);
2263 
2264 	do {
2265 		error = dmu_dir_list_next(os,
2266 		    sizeof (zc->zc_name) - (p - zc->zc_name), p,
2267 		    NULL, &zc->zc_cookie);
2268 		if (error == ENOENT)
2269 			error = SET_ERROR(ESRCH);
2270 	} while (error == 0 && zfs_dataset_name_hidden(zc->zc_name));
2271 	dmu_objset_rele(os, FTAG);
2272 
2273 	/*
2274 	 * If it's an internal dataset (ie. with a '$' in its name),
2275 	 * don't try to get stats for it, otherwise we'll return ENOENT.
2276 	 */
2277 	if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2278 		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2279 		if (error == ENOENT) {
2280 			/* We lost a race with destroy, get the next one. */
2281 			zc->zc_name[orig_len] = '\0';
2282 			goto top;
2283 		}
2284 	}
2285 	return (error);
2286 }
2287 
2288 /*
2289  * inputs:
2290  * zc_name		name of filesystem
2291  * zc_cookie		zap cursor
2292  * zc_nvlist_src	iteration range nvlist
2293  * zc_nvlist_src_size	size of iteration range nvlist
2294  *
2295  * outputs:
2296  * zc_name		name of next snapshot
2297  * zc_objset_stats	stats
2298  * zc_nvlist_dst	property nvlist
2299  * zc_nvlist_dst_size	size of property nvlist
2300  */
2301 static int
2302 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2303 {
2304 	int error;
2305 	objset_t *os, *ossnap;
2306 	dsl_dataset_t *ds;
2307 	uint64_t min_txg = 0, max_txg = 0;
2308 
2309 	if (zc->zc_nvlist_src_size != 0) {
2310 		nvlist_t *props = NULL;
2311 		error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2312 		    zc->zc_iflags, &props);
2313 		if (error != 0)
2314 			return (error);
2315 		(void) nvlist_lookup_uint64(props, SNAP_ITER_MIN_TXG,
2316 		    &min_txg);
2317 		(void) nvlist_lookup_uint64(props, SNAP_ITER_MAX_TXG,
2318 		    &max_txg);
2319 		nvlist_free(props);
2320 	}
2321 
2322 	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2323 	if (error != 0) {
2324 		return (error == ENOENT ? SET_ERROR(ESRCH) : error);
2325 	}
2326 
2327 	/*
2328 	 * A dataset name of maximum length cannot have any snapshots,
2329 	 * so exit immediately.
2330 	 */
2331 	if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >=
2332 	    ZFS_MAX_DATASET_NAME_LEN) {
2333 		dmu_objset_rele(os, FTAG);
2334 		return (SET_ERROR(ESRCH));
2335 	}
2336 
2337 	while (error == 0) {
2338 		if (issig(JUSTLOOKING) && issig(FORREAL)) {
2339 			error = SET_ERROR(EINTR);
2340 			break;
2341 		}
2342 
2343 		error = dmu_snapshot_list_next(os,
2344 		    sizeof (zc->zc_name) - strlen(zc->zc_name),
2345 		    zc->zc_name + strlen(zc->zc_name), &zc->zc_obj,
2346 		    &zc->zc_cookie, NULL);
2347 		if (error == ENOENT) {
2348 			error = SET_ERROR(ESRCH);
2349 			break;
2350 		} else if (error != 0) {
2351 			break;
2352 		}
2353 
2354 		error = dsl_dataset_hold_obj(dmu_objset_pool(os), zc->zc_obj,
2355 		    FTAG, &ds);
2356 		if (error != 0)
2357 			break;
2358 
2359 		if ((min_txg != 0 && dsl_get_creationtxg(ds) < min_txg) ||
2360 		    (max_txg != 0 && dsl_get_creationtxg(ds) > max_txg)) {
2361 			dsl_dataset_rele(ds, FTAG);
2362 			/* undo snapshot name append */
2363 			*(strchr(zc->zc_name, '@') + 1) = '\0';
2364 			/* skip snapshot */
2365 			continue;
2366 		}
2367 
2368 		if (zc->zc_simple) {
2369 			dsl_dataset_fast_stat(ds, &zc->zc_objset_stats);
2370 			dsl_dataset_rele(ds, FTAG);
2371 			break;
2372 		}
2373 
2374 		if ((error = dmu_objset_from_ds(ds, &ossnap)) != 0) {
2375 			dsl_dataset_rele(ds, FTAG);
2376 			break;
2377 		}
2378 		if ((error = zfs_ioc_objset_stats_impl(zc, ossnap)) != 0) {
2379 			dsl_dataset_rele(ds, FTAG);
2380 			break;
2381 		}
2382 		dsl_dataset_rele(ds, FTAG);
2383 		break;
2384 	}
2385 
2386 	dmu_objset_rele(os, FTAG);
2387 	/* if we failed, undo the @ that we tacked on to zc_name */
2388 	if (error != 0)
2389 		*strchr(zc->zc_name, '@') = '\0';
2390 	return (error);
2391 }
2392 
2393 static int
2394 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2395 {
2396 	const char *propname = nvpair_name(pair);
2397 	uint64_t *valary;
2398 	unsigned int vallen;
2399 	const char *dash, *domain;
2400 	zfs_userquota_prop_t type;
2401 	uint64_t rid;
2402 	uint64_t quota;
2403 	zfsvfs_t *zfsvfs;
2404 	int err;
2405 
2406 	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2407 		nvlist_t *attrs;
2408 		VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2409 		if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2410 		    &pair) != 0)
2411 			return (SET_ERROR(EINVAL));
2412 	}
2413 
2414 	/*
2415 	 * A correctly constructed propname is encoded as
2416 	 * userquota@<rid>-<domain>.
2417 	 */
2418 	if ((dash = strchr(propname, '-')) == NULL ||
2419 	    nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2420 	    vallen != 3)
2421 		return (SET_ERROR(EINVAL));
2422 
2423 	domain = dash + 1;
2424 	type = valary[0];
2425 	rid = valary[1];
2426 	quota = valary[2];
2427 
2428 	err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2429 	if (err == 0) {
2430 		err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2431 		zfsvfs_rele(zfsvfs, FTAG);
2432 	}
2433 
2434 	return (err);
2435 }
2436 
2437 /*
2438  * If the named property is one that has a special function to set its value,
2439  * return 0 on success and a positive error code on failure; otherwise if it is
2440  * not one of the special properties handled by this function, return -1.
2441  *
2442  * XXX: It would be better for callers of the property interface if we handled
2443  * these special cases in dsl_prop.c (in the dsl layer).
2444  */
2445 static int
2446 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2447     nvpair_t *pair)
2448 {
2449 	const char *propname = nvpair_name(pair);
2450 	zfs_prop_t prop = zfs_name_to_prop(propname);
2451 	uint64_t intval = 0;
2452 	const char *strval = NULL;
2453 	int err = -1;
2454 
2455 	if (prop == ZPROP_USERPROP) {
2456 		if (zfs_prop_userquota(propname))
2457 			return (zfs_prop_set_userquota(dsname, pair));
2458 		return (-1);
2459 	}
2460 
2461 	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2462 		nvlist_t *attrs;
2463 		VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2464 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2465 		    &pair) == 0);
2466 	}
2467 
2468 	/* all special properties are numeric except for keylocation */
2469 	if (zfs_prop_get_type(prop) == PROP_TYPE_STRING) {
2470 		strval = fnvpair_value_string(pair);
2471 	} else {
2472 		intval = fnvpair_value_uint64(pair);
2473 	}
2474 
2475 	switch (prop) {
2476 	case ZFS_PROP_QUOTA:
2477 		err = dsl_dir_set_quota(dsname, source, intval);
2478 		break;
2479 	case ZFS_PROP_REFQUOTA:
2480 		err = dsl_dataset_set_refquota(dsname, source, intval);
2481 		break;
2482 	case ZFS_PROP_FILESYSTEM_LIMIT:
2483 	case ZFS_PROP_SNAPSHOT_LIMIT:
2484 		if (intval == UINT64_MAX) {
2485 			/* clearing the limit, just do it */
2486 			err = 0;
2487 		} else {
2488 			err = dsl_dir_activate_fs_ss_limit(dsname);
2489 		}
2490 		/*
2491 		 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2492 		 * default path to set the value in the nvlist.
2493 		 */
2494 		if (err == 0)
2495 			err = -1;
2496 		break;
2497 	case ZFS_PROP_KEYLOCATION:
2498 		err = dsl_crypto_can_set_keylocation(dsname, strval);
2499 
2500 		/*
2501 		 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2502 		 * default path to set the value in the nvlist.
2503 		 */
2504 		if (err == 0)
2505 			err = -1;
2506 		break;
2507 	case ZFS_PROP_RESERVATION:
2508 		err = dsl_dir_set_reservation(dsname, source, intval);
2509 		break;
2510 	case ZFS_PROP_REFRESERVATION:
2511 		err = dsl_dataset_set_refreservation(dsname, source, intval);
2512 		break;
2513 	case ZFS_PROP_COMPRESSION:
2514 		err = dsl_dataset_set_compression(dsname, source, intval);
2515 		/*
2516 		 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2517 		 * default path to set the value in the nvlist.
2518 		 */
2519 		if (err == 0)
2520 			err = -1;
2521 		break;
2522 	case ZFS_PROP_VOLSIZE:
2523 		err = zvol_set_volsize(dsname, intval);
2524 		break;
2525 	case ZFS_PROP_VOLTHREADING:
2526 		err = zvol_set_volthreading(dsname, intval);
2527 		/*
2528 		 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2529 		 * default path to set the value in the nvlist.
2530 		 */
2531 		if (err == 0)
2532 			err = -1;
2533 		break;
2534 	case ZFS_PROP_SNAPDEV:
2535 	case ZFS_PROP_VOLMODE:
2536 		err = zvol_set_common(dsname, prop, source, intval);
2537 		break;
2538 	case ZFS_PROP_READONLY:
2539 		err = zvol_set_ro(dsname, intval);
2540 		/*
2541 		 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2542 		 * default path to set the value in the nvlist.
2543 		 */
2544 		if (err == 0)
2545 			err = -1;
2546 		break;
2547 	case ZFS_PROP_VERSION:
2548 	{
2549 		zfsvfs_t *zfsvfs;
2550 
2551 		if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2552 			break;
2553 
2554 		err = zfs_set_version(zfsvfs, intval);
2555 		zfsvfs_rele(zfsvfs, FTAG);
2556 
2557 		if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2558 			zfs_cmd_t *zc;
2559 
2560 			zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2561 			(void) strlcpy(zc->zc_name, dsname,
2562 			    sizeof (zc->zc_name));
2563 			(void) zfs_ioc_userspace_upgrade(zc);
2564 			(void) zfs_ioc_id_quota_upgrade(zc);
2565 			kmem_free(zc, sizeof (zfs_cmd_t));
2566 		}
2567 		break;
2568 	}
2569 	default:
2570 		err = -1;
2571 	}
2572 
2573 	return (err);
2574 }
2575 
2576 static boolean_t
2577 zfs_is_namespace_prop(zfs_prop_t prop)
2578 {
2579 	switch (prop) {
2580 
2581 	case ZFS_PROP_ATIME:
2582 	case ZFS_PROP_RELATIME:
2583 	case ZFS_PROP_DEVICES:
2584 	case ZFS_PROP_EXEC:
2585 	case ZFS_PROP_SETUID:
2586 	case ZFS_PROP_READONLY:
2587 	case ZFS_PROP_XATTR:
2588 	case ZFS_PROP_NBMAND:
2589 		return (B_TRUE);
2590 
2591 	default:
2592 		return (B_FALSE);
2593 	}
2594 }
2595 
2596 /*
2597  * This function is best effort. If it fails to set any of the given properties,
2598  * it continues to set as many as it can and returns the last error
2599  * encountered. If the caller provides a non-NULL errlist, it will be filled in
2600  * with the list of names of all the properties that failed along with the
2601  * corresponding error numbers.
2602  *
2603  * If every property is set successfully, zero is returned and errlist is not
2604  * modified.
2605  */
2606 int
2607 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2608     nvlist_t *errlist)
2609 {
2610 	nvpair_t *pair;
2611 	nvpair_t *propval;
2612 	int rv = 0;
2613 	int err;
2614 	uint64_t intval;
2615 	const char *strval;
2616 	boolean_t should_update_mount_cache = B_FALSE;
2617 
2618 	nvlist_t *genericnvl = fnvlist_alloc();
2619 	nvlist_t *retrynvl = fnvlist_alloc();
2620 retry:
2621 	pair = NULL;
2622 	while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2623 		const char *propname = nvpair_name(pair);
2624 		zfs_prop_t prop = zfs_name_to_prop(propname);
2625 		err = 0;
2626 
2627 		/* decode the property value */
2628 		propval = pair;
2629 		if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2630 			nvlist_t *attrs;
2631 			attrs = fnvpair_value_nvlist(pair);
2632 			if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2633 			    &propval) != 0)
2634 				err = SET_ERROR(EINVAL);
2635 		}
2636 
2637 		/* Validate value type */
2638 		if (err == 0 && source == ZPROP_SRC_INHERITED) {
2639 			/* inherited properties are expected to be booleans */
2640 			if (nvpair_type(propval) != DATA_TYPE_BOOLEAN)
2641 				err = SET_ERROR(EINVAL);
2642 		} else if (err == 0 && prop == ZPROP_USERPROP) {
2643 			if (zfs_prop_user(propname)) {
2644 				if (nvpair_type(propval) != DATA_TYPE_STRING)
2645 					err = SET_ERROR(EINVAL);
2646 			} else if (zfs_prop_userquota(propname)) {
2647 				if (nvpair_type(propval) !=
2648 				    DATA_TYPE_UINT64_ARRAY)
2649 					err = SET_ERROR(EINVAL);
2650 			} else {
2651 				err = SET_ERROR(EINVAL);
2652 			}
2653 		} else if (err == 0) {
2654 			if (nvpair_type(propval) == DATA_TYPE_STRING) {
2655 				if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2656 					err = SET_ERROR(EINVAL);
2657 			} else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2658 				const char *unused;
2659 
2660 				intval = fnvpair_value_uint64(propval);
2661 
2662 				switch (zfs_prop_get_type(prop)) {
2663 				case PROP_TYPE_NUMBER:
2664 					break;
2665 				case PROP_TYPE_STRING:
2666 					err = SET_ERROR(EINVAL);
2667 					break;
2668 				case PROP_TYPE_INDEX:
2669 					if (zfs_prop_index_to_string(prop,
2670 					    intval, &unused) != 0)
2671 						err =
2672 						    SET_ERROR(ZFS_ERR_BADPROP);
2673 					break;
2674 				default:
2675 					cmn_err(CE_PANIC,
2676 					    "unknown property type");
2677 				}
2678 			} else {
2679 				err = SET_ERROR(EINVAL);
2680 			}
2681 		}
2682 
2683 		/* Validate permissions */
2684 		if (err == 0)
2685 			err = zfs_check_settable(dsname, pair, CRED());
2686 
2687 		if (err == 0) {
2688 			if (source == ZPROP_SRC_INHERITED)
2689 				err = -1; /* does not need special handling */
2690 			else
2691 				err = zfs_prop_set_special(dsname, source,
2692 				    pair);
2693 			if (err == -1) {
2694 				/*
2695 				 * For better performance we build up a list of
2696 				 * properties to set in a single transaction.
2697 				 */
2698 				err = nvlist_add_nvpair(genericnvl, pair);
2699 			} else if (err != 0 && nvl != retrynvl) {
2700 				/*
2701 				 * This may be a spurious error caused by
2702 				 * receiving quota and reservation out of order.
2703 				 * Try again in a second pass.
2704 				 */
2705 				err = nvlist_add_nvpair(retrynvl, pair);
2706 			}
2707 		}
2708 
2709 		if (err != 0) {
2710 			if (errlist != NULL)
2711 				fnvlist_add_int32(errlist, propname, err);
2712 			rv = err;
2713 		}
2714 
2715 		if (zfs_is_namespace_prop(prop))
2716 			should_update_mount_cache = B_TRUE;
2717 	}
2718 
2719 	if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2720 		nvl = retrynvl;
2721 		goto retry;
2722 	}
2723 
2724 	if (nvlist_empty(genericnvl))
2725 		goto out;
2726 
2727 	/*
2728 	 * Try to set them all in one batch.
2729 	 */
2730 	err = dsl_props_set(dsname, source, genericnvl);
2731 	if (err == 0)
2732 		goto out;
2733 
2734 	/*
2735 	 * If batching fails, we still want to set as many properties as we
2736 	 * can, so try setting them individually.
2737 	 */
2738 	pair = NULL;
2739 	while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2740 		const char *propname = nvpair_name(pair);
2741 
2742 		propval = pair;
2743 		if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2744 			nvlist_t *attrs;
2745 			attrs = fnvpair_value_nvlist(pair);
2746 			propval = fnvlist_lookup_nvpair(attrs, ZPROP_VALUE);
2747 		}
2748 
2749 		if (nvpair_type(propval) == DATA_TYPE_STRING) {
2750 			strval = fnvpair_value_string(propval);
2751 			err = dsl_prop_set_string(dsname, propname,
2752 			    source, strval);
2753 		} else if (nvpair_type(propval) == DATA_TYPE_BOOLEAN) {
2754 			err = dsl_prop_inherit(dsname, propname, source);
2755 		} else {
2756 			intval = fnvpair_value_uint64(propval);
2757 			err = dsl_prop_set_int(dsname, propname, source,
2758 			    intval);
2759 		}
2760 
2761 		if (err != 0) {
2762 			if (errlist != NULL) {
2763 				fnvlist_add_int32(errlist, propname, err);
2764 			}
2765 			rv = err;
2766 		}
2767 	}
2768 
2769 out:
2770 	if (should_update_mount_cache)
2771 		zfs_ioctl_update_mount_cache(dsname);
2772 
2773 	nvlist_free(genericnvl);
2774 	nvlist_free(retrynvl);
2775 
2776 	return (rv);
2777 }
2778 
2779 /*
2780  * Check that all the properties are valid user properties.
2781  */
2782 static int
2783 zfs_check_userprops(nvlist_t *nvl)
2784 {
2785 	nvpair_t *pair = NULL;
2786 
2787 	while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2788 		const char *propname = nvpair_name(pair);
2789 
2790 		if (!zfs_prop_user(propname) ||
2791 		    nvpair_type(pair) != DATA_TYPE_STRING)
2792 			return (SET_ERROR(EINVAL));
2793 
2794 		if (strlen(propname) >= ZAP_MAXNAMELEN)
2795 			return (SET_ERROR(ENAMETOOLONG));
2796 
2797 		if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
2798 			return (SET_ERROR(E2BIG));
2799 	}
2800 	return (0);
2801 }
2802 
2803 static void
2804 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2805 {
2806 	nvpair_t *pair;
2807 
2808 	VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2809 
2810 	pair = NULL;
2811 	while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2812 		if (nvlist_exists(skipped, nvpair_name(pair)))
2813 			continue;
2814 
2815 		VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2816 	}
2817 }
2818 
2819 static int
2820 clear_received_props(const char *dsname, nvlist_t *props,
2821     nvlist_t *skipped)
2822 {
2823 	int err = 0;
2824 	nvlist_t *cleared_props = NULL;
2825 	props_skip(props, skipped, &cleared_props);
2826 	if (!nvlist_empty(cleared_props)) {
2827 		/*
2828 		 * Acts on local properties until the dataset has received
2829 		 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2830 		 */
2831 		zprop_source_t flags = (ZPROP_SRC_NONE |
2832 		    (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2833 		err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2834 	}
2835 	nvlist_free(cleared_props);
2836 	return (err);
2837 }
2838 
2839 /*
2840  * inputs:
2841  * zc_name		name of filesystem
2842  * zc_value		name of property to set
2843  * zc_nvlist_src{_size}	nvlist of properties to apply
2844  * zc_cookie		received properties flag
2845  *
2846  * outputs:
2847  * zc_nvlist_dst{_size} error for each unapplied received property
2848  */
2849 static int
2850 zfs_ioc_set_prop(zfs_cmd_t *zc)
2851 {
2852 	nvlist_t *nvl;
2853 	boolean_t received = zc->zc_cookie;
2854 	zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2855 	    ZPROP_SRC_LOCAL);
2856 	nvlist_t *errors;
2857 	int error;
2858 
2859 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2860 	    zc->zc_iflags, &nvl)) != 0)
2861 		return (error);
2862 
2863 	if (received) {
2864 		nvlist_t *origprops;
2865 
2866 		if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2867 			(void) clear_received_props(zc->zc_name,
2868 			    origprops, nvl);
2869 			nvlist_free(origprops);
2870 		}
2871 
2872 		error = dsl_prop_set_hasrecvd(zc->zc_name);
2873 	}
2874 
2875 	errors = fnvlist_alloc();
2876 	if (error == 0)
2877 		error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2878 
2879 	if (zc->zc_nvlist_dst != 0 && errors != NULL) {
2880 		(void) put_nvlist(zc, errors);
2881 	}
2882 
2883 	nvlist_free(errors);
2884 	nvlist_free(nvl);
2885 	return (error);
2886 }
2887 
2888 /*
2889  * inputs:
2890  * zc_name		name of filesystem
2891  * zc_value		name of property to inherit
2892  * zc_cookie		revert to received value if TRUE
2893  *
2894  * outputs:		none
2895  */
2896 static int
2897 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2898 {
2899 	const char *propname = zc->zc_value;
2900 	zfs_prop_t prop = zfs_name_to_prop(propname);
2901 	boolean_t received = zc->zc_cookie;
2902 	zprop_source_t source = (received
2903 	    ? ZPROP_SRC_NONE		/* revert to received value, if any */
2904 	    : ZPROP_SRC_INHERITED);	/* explicitly inherit */
2905 	nvlist_t *dummy;
2906 	nvpair_t *pair;
2907 	zprop_type_t type;
2908 	int err;
2909 
2910 	if (!received) {
2911 		/*
2912 		 * Only check this in the non-received case. We want to allow
2913 		 * 'inherit -S' to revert non-inheritable properties like quota
2914 		 * and reservation to the received or default values even though
2915 		 * they are not considered inheritable.
2916 		 */
2917 		if (prop != ZPROP_USERPROP && !zfs_prop_inheritable(prop))
2918 			return (SET_ERROR(EINVAL));
2919 	}
2920 
2921 	if (prop == ZPROP_USERPROP) {
2922 		if (!zfs_prop_user(propname))
2923 			return (SET_ERROR(EINVAL));
2924 
2925 		type = PROP_TYPE_STRING;
2926 	} else if (prop == ZFS_PROP_VOLSIZE || prop == ZFS_PROP_VERSION) {
2927 		return (SET_ERROR(EINVAL));
2928 	} else {
2929 		type = zfs_prop_get_type(prop);
2930 	}
2931 
2932 	/*
2933 	 * zfs_prop_set_special() expects properties in the form of an
2934 	 * nvpair with type info.
2935 	 */
2936 	dummy = fnvlist_alloc();
2937 
2938 	switch (type) {
2939 	case PROP_TYPE_STRING:
2940 		VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2941 		break;
2942 	case PROP_TYPE_NUMBER:
2943 	case PROP_TYPE_INDEX:
2944 		VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2945 		break;
2946 	default:
2947 		err = SET_ERROR(EINVAL);
2948 		goto errout;
2949 	}
2950 
2951 	pair = nvlist_next_nvpair(dummy, NULL);
2952 	if (pair == NULL) {
2953 		err = SET_ERROR(EINVAL);
2954 	} else {
2955 		err = zfs_prop_set_special(zc->zc_name, source, pair);
2956 		if (err == -1) /* property is not "special", needs handling */
2957 			err = dsl_prop_inherit(zc->zc_name, zc->zc_value,
2958 			    source);
2959 	}
2960 
2961 errout:
2962 	nvlist_free(dummy);
2963 	return (err);
2964 }
2965 
2966 static int
2967 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2968 {
2969 	nvlist_t *props;
2970 	spa_t *spa;
2971 	int error;
2972 	nvpair_t *pair;
2973 
2974 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2975 	    zc->zc_iflags, &props)))
2976 		return (error);
2977 
2978 	/*
2979 	 * If the only property is the configfile, then just do a spa_lookup()
2980 	 * to handle the faulted case.
2981 	 */
2982 	pair = nvlist_next_nvpair(props, NULL);
2983 	if (pair != NULL && strcmp(nvpair_name(pair),
2984 	    zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2985 	    nvlist_next_nvpair(props, pair) == NULL) {
2986 		mutex_enter(&spa_namespace_lock);
2987 		if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2988 			spa_configfile_set(spa, props, B_FALSE);
2989 			spa_write_cachefile(spa, B_FALSE, B_TRUE, B_FALSE);
2990 		}
2991 		mutex_exit(&spa_namespace_lock);
2992 		if (spa != NULL) {
2993 			nvlist_free(props);
2994 			return (0);
2995 		}
2996 	}
2997 
2998 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2999 		nvlist_free(props);
3000 		return (error);
3001 	}
3002 
3003 	error = spa_prop_set(spa, props);
3004 
3005 	nvlist_free(props);
3006 	spa_close(spa, FTAG);
3007 
3008 	return (error);
3009 }
3010 
3011 static int
3012 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
3013 {
3014 	spa_t *spa;
3015 	int error;
3016 	nvlist_t *nvp = NULL;
3017 
3018 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
3019 		/*
3020 		 * If the pool is faulted, there may be properties we can still
3021 		 * get (such as altroot and cachefile), so attempt to get them
3022 		 * anyway.
3023 		 */
3024 		mutex_enter(&spa_namespace_lock);
3025 		if ((spa = spa_lookup(zc->zc_name)) != NULL)
3026 			error = spa_prop_get(spa, &nvp);
3027 		mutex_exit(&spa_namespace_lock);
3028 	} else {
3029 		error = spa_prop_get(spa, &nvp);
3030 		spa_close(spa, FTAG);
3031 	}
3032 
3033 	if (error == 0 && zc->zc_nvlist_dst != 0)
3034 		error = put_nvlist(zc, nvp);
3035 	else
3036 		error = SET_ERROR(EFAULT);
3037 
3038 	nvlist_free(nvp);
3039 	return (error);
3040 }
3041 
3042 /*
3043  * innvl: {
3044  *     "vdevprops_set_vdev" -> guid
3045  *     "vdevprops_set_props" -> { prop -> value }
3046  * }
3047  *
3048  * outnvl: propname -> error code (int32)
3049  */
3050 static const zfs_ioc_key_t zfs_keys_vdev_set_props[] = {
3051 	{ZPOOL_VDEV_PROPS_SET_VDEV,	DATA_TYPE_UINT64,	0},
3052 	{ZPOOL_VDEV_PROPS_SET_PROPS,	DATA_TYPE_NVLIST,	0}
3053 };
3054 
3055 static int
3056 zfs_ioc_vdev_set_props(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3057 {
3058 	spa_t *spa;
3059 	int error;
3060 	vdev_t *vd;
3061 	uint64_t vdev_guid;
3062 
3063 	/* Early validation */
3064 	if (nvlist_lookup_uint64(innvl, ZPOOL_VDEV_PROPS_SET_VDEV,
3065 	    &vdev_guid) != 0)
3066 		return (SET_ERROR(EINVAL));
3067 
3068 	if (outnvl == NULL)
3069 		return (SET_ERROR(EINVAL));
3070 
3071 	if ((error = spa_open(poolname, &spa, FTAG)) != 0)
3072 		return (error);
3073 
3074 	ASSERT(spa_writeable(spa));
3075 
3076 	if ((vd = spa_lookup_by_guid(spa, vdev_guid, B_TRUE)) == NULL) {
3077 		spa_close(spa, FTAG);
3078 		return (SET_ERROR(ENOENT));
3079 	}
3080 
3081 	error = vdev_prop_set(vd, innvl, outnvl);
3082 
3083 	spa_close(spa, FTAG);
3084 
3085 	return (error);
3086 }
3087 
3088 /*
3089  * innvl: {
3090  *     "vdevprops_get_vdev" -> guid
3091  *     (optional) "vdevprops_get_props" -> { propname -> propid }
3092  * }
3093  *
3094  * outnvl: propname -> value
3095  */
3096 static const zfs_ioc_key_t zfs_keys_vdev_get_props[] = {
3097 	{ZPOOL_VDEV_PROPS_GET_VDEV,	DATA_TYPE_UINT64,	0},
3098 	{ZPOOL_VDEV_PROPS_GET_PROPS,	DATA_TYPE_NVLIST,	ZK_OPTIONAL}
3099 };
3100 
3101 static int
3102 zfs_ioc_vdev_get_props(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3103 {
3104 	spa_t *spa;
3105 	int error;
3106 	vdev_t *vd;
3107 	uint64_t vdev_guid;
3108 
3109 	/* Early validation */
3110 	if (nvlist_lookup_uint64(innvl, ZPOOL_VDEV_PROPS_GET_VDEV,
3111 	    &vdev_guid) != 0)
3112 		return (SET_ERROR(EINVAL));
3113 
3114 	if (outnvl == NULL)
3115 		return (SET_ERROR(EINVAL));
3116 
3117 	if ((error = spa_open(poolname, &spa, FTAG)) != 0)
3118 		return (error);
3119 
3120 	if ((vd = spa_lookup_by_guid(spa, vdev_guid, B_TRUE)) == NULL) {
3121 		spa_close(spa, FTAG);
3122 		return (SET_ERROR(ENOENT));
3123 	}
3124 
3125 	error = vdev_prop_get(vd, innvl, outnvl);
3126 
3127 	spa_close(spa, FTAG);
3128 
3129 	return (error);
3130 }
3131 
3132 /*
3133  * inputs:
3134  * zc_name		name of filesystem
3135  * zc_nvlist_src{_size}	nvlist of delegated permissions
3136  * zc_perm_action	allow/unallow flag
3137  *
3138  * outputs:		none
3139  */
3140 static int
3141 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
3142 {
3143 	int error;
3144 	nvlist_t *fsaclnv = NULL;
3145 
3146 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3147 	    zc->zc_iflags, &fsaclnv)) != 0)
3148 		return (error);
3149 
3150 	/*
3151 	 * Verify nvlist is constructed correctly
3152 	 */
3153 	if (zfs_deleg_verify_nvlist(fsaclnv) != 0) {
3154 		nvlist_free(fsaclnv);
3155 		return (SET_ERROR(EINVAL));
3156 	}
3157 
3158 	/*
3159 	 * If we don't have PRIV_SYS_MOUNT, then validate
3160 	 * that user is allowed to hand out each permission in
3161 	 * the nvlist(s)
3162 	 */
3163 
3164 	error = secpolicy_zfs(CRED());
3165 	if (error != 0) {
3166 		if (zc->zc_perm_action == B_FALSE) {
3167 			error = dsl_deleg_can_allow(zc->zc_name,
3168 			    fsaclnv, CRED());
3169 		} else {
3170 			error = dsl_deleg_can_unallow(zc->zc_name,
3171 			    fsaclnv, CRED());
3172 		}
3173 	}
3174 
3175 	if (error == 0)
3176 		error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
3177 
3178 	nvlist_free(fsaclnv);
3179 	return (error);
3180 }
3181 
3182 /*
3183  * inputs:
3184  * zc_name		name of filesystem
3185  *
3186  * outputs:
3187  * zc_nvlist_src{_size}	nvlist of delegated permissions
3188  */
3189 static int
3190 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
3191 {
3192 	nvlist_t *nvp;
3193 	int error;
3194 
3195 	if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
3196 		error = put_nvlist(zc, nvp);
3197 		nvlist_free(nvp);
3198 	}
3199 
3200 	return (error);
3201 }
3202 
3203 static void
3204 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
3205 {
3206 	zfs_creat_t *zct = arg;
3207 
3208 	zfs_create_fs(os, cr, zct->zct_zplprops, tx);
3209 }
3210 
3211 #define	ZFS_PROP_UNDEFINED	((uint64_t)-1)
3212 
3213 /*
3214  * inputs:
3215  * os			parent objset pointer (NULL if root fs)
3216  * fuids_ok		fuids allowed in this version of the spa?
3217  * sa_ok		SAs allowed in this version of the spa?
3218  * createprops		list of properties requested by creator
3219  *
3220  * outputs:
3221  * zplprops	values for the zplprops we attach to the master node object
3222  * is_ci	true if requested file system will be purely case-insensitive
3223  *
3224  * Determine the settings for utf8only, normalization and
3225  * casesensitivity.  Specific values may have been requested by the
3226  * creator and/or we can inherit values from the parent dataset.  If
3227  * the file system is of too early a vintage, a creator can not
3228  * request settings for these properties, even if the requested
3229  * setting is the default value.  We don't actually want to create dsl
3230  * properties for these, so remove them from the source nvlist after
3231  * processing.
3232  */
3233 static int
3234 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3235     boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3236     nvlist_t *zplprops, boolean_t *is_ci)
3237 {
3238 	uint64_t sense = ZFS_PROP_UNDEFINED;
3239 	uint64_t norm = ZFS_PROP_UNDEFINED;
3240 	uint64_t u8 = ZFS_PROP_UNDEFINED;
3241 	int error;
3242 
3243 	ASSERT(zplprops != NULL);
3244 
3245 	/* parent dataset must be a filesystem */
3246 	if (os != NULL && os->os_phys->os_type != DMU_OST_ZFS)
3247 		return (SET_ERROR(ZFS_ERR_WRONG_PARENT));
3248 
3249 	/*
3250 	 * Pull out creator prop choices, if any.
3251 	 */
3252 	if (createprops) {
3253 		(void) nvlist_lookup_uint64(createprops,
3254 		    zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3255 		(void) nvlist_lookup_uint64(createprops,
3256 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3257 		(void) nvlist_remove_all(createprops,
3258 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3259 		(void) nvlist_lookup_uint64(createprops,
3260 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3261 		(void) nvlist_remove_all(createprops,
3262 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3263 		(void) nvlist_lookup_uint64(createprops,
3264 		    zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3265 		(void) nvlist_remove_all(createprops,
3266 		    zfs_prop_to_name(ZFS_PROP_CASE));
3267 	}
3268 
3269 	/*
3270 	 * If the zpl version requested is whacky or the file system
3271 	 * or pool is version is too "young" to support normalization
3272 	 * and the creator tried to set a value for one of the props,
3273 	 * error out.
3274 	 */
3275 	if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3276 	    (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3277 	    (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3278 	    (zplver < ZPL_VERSION_NORMALIZATION &&
3279 	    (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3280 	    sense != ZFS_PROP_UNDEFINED)))
3281 		return (SET_ERROR(ENOTSUP));
3282 
3283 	/*
3284 	 * Put the version in the zplprops
3285 	 */
3286 	VERIFY(nvlist_add_uint64(zplprops,
3287 	    zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3288 
3289 	if (norm == ZFS_PROP_UNDEFINED &&
3290 	    (error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm)) != 0)
3291 		return (error);
3292 	VERIFY(nvlist_add_uint64(zplprops,
3293 	    zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3294 
3295 	/*
3296 	 * If we're normalizing, names must always be valid UTF-8 strings.
3297 	 */
3298 	if (norm)
3299 		u8 = 1;
3300 	if (u8 == ZFS_PROP_UNDEFINED &&
3301 	    (error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8)) != 0)
3302 		return (error);
3303 	VERIFY(nvlist_add_uint64(zplprops,
3304 	    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3305 
3306 	if (sense == ZFS_PROP_UNDEFINED &&
3307 	    (error = zfs_get_zplprop(os, ZFS_PROP_CASE, &sense)) != 0)
3308 		return (error);
3309 	VERIFY(nvlist_add_uint64(zplprops,
3310 	    zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3311 
3312 	if (is_ci)
3313 		*is_ci = (sense == ZFS_CASE_INSENSITIVE);
3314 
3315 	return (0);
3316 }
3317 
3318 static int
3319 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3320     nvlist_t *zplprops, boolean_t *is_ci)
3321 {
3322 	boolean_t fuids_ok, sa_ok;
3323 	uint64_t zplver = ZPL_VERSION;
3324 	objset_t *os = NULL;
3325 	char parentname[ZFS_MAX_DATASET_NAME_LEN];
3326 	spa_t *spa;
3327 	uint64_t spa_vers;
3328 	int error;
3329 
3330 	zfs_get_parent(dataset, parentname, sizeof (parentname));
3331 
3332 	if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3333 		return (error);
3334 
3335 	spa_vers = spa_version(spa);
3336 	spa_close(spa, FTAG);
3337 
3338 	zplver = zfs_zpl_version_map(spa_vers);
3339 	fuids_ok = (zplver >= ZPL_VERSION_FUID);
3340 	sa_ok = (zplver >= ZPL_VERSION_SA);
3341 
3342 	/*
3343 	 * Open parent object set so we can inherit zplprop values.
3344 	 */
3345 	if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3346 		return (error);
3347 
3348 	error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3349 	    zplprops, is_ci);
3350 	dmu_objset_rele(os, FTAG);
3351 	return (error);
3352 }
3353 
3354 static int
3355 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3356     nvlist_t *zplprops, boolean_t *is_ci)
3357 {
3358 	boolean_t fuids_ok;
3359 	boolean_t sa_ok;
3360 	uint64_t zplver = ZPL_VERSION;
3361 	int error;
3362 
3363 	zplver = zfs_zpl_version_map(spa_vers);
3364 	fuids_ok = (zplver >= ZPL_VERSION_FUID);
3365 	sa_ok = (zplver >= ZPL_VERSION_SA);
3366 
3367 	error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3368 	    createprops, zplprops, is_ci);
3369 	return (error);
3370 }
3371 
3372 /*
3373  * innvl: {
3374  *     "type" -> dmu_objset_type_t (int32)
3375  *     (optional) "props" -> { prop -> value }
3376  *     (optional) "hidden_args" -> { "wkeydata" -> value }
3377  *         raw uint8_t array of encryption wrapping key data (32 bytes)
3378  * }
3379  *
3380  * outnvl: propname -> error code (int32)
3381  */
3382 
3383 static const zfs_ioc_key_t zfs_keys_create[] = {
3384 	{"type",	DATA_TYPE_INT32,	0},
3385 	{"props",	DATA_TYPE_NVLIST,	ZK_OPTIONAL},
3386 	{"hidden_args",	DATA_TYPE_NVLIST,	ZK_OPTIONAL},
3387 };
3388 
3389 static int
3390 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3391 {
3392 	int error = 0;
3393 	zfs_creat_t zct = { 0 };
3394 	nvlist_t *nvprops = NULL;
3395 	nvlist_t *hidden_args = NULL;
3396 	void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3397 	dmu_objset_type_t type;
3398 	boolean_t is_insensitive = B_FALSE;
3399 	dsl_crypto_params_t *dcp = NULL;
3400 
3401 	type = (dmu_objset_type_t)fnvlist_lookup_int32(innvl, "type");
3402 	(void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3403 	(void) nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);
3404 
3405 	switch (type) {
3406 	case DMU_OST_ZFS:
3407 		cbfunc = zfs_create_cb;
3408 		break;
3409 
3410 	case DMU_OST_ZVOL:
3411 		cbfunc = zvol_create_cb;
3412 		break;
3413 
3414 	default:
3415 		cbfunc = NULL;
3416 		break;
3417 	}
3418 	if (strchr(fsname, '@') ||
3419 	    strchr(fsname, '%'))
3420 		return (SET_ERROR(EINVAL));
3421 
3422 	zct.zct_props = nvprops;
3423 
3424 	if (cbfunc == NULL)
3425 		return (SET_ERROR(EINVAL));
3426 
3427 	if (type == DMU_OST_ZVOL) {
3428 		uint64_t volsize, volblocksize;
3429 
3430 		if (nvprops == NULL)
3431 			return (SET_ERROR(EINVAL));
3432 		if (nvlist_lookup_uint64(nvprops,
3433 		    zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3434 			return (SET_ERROR(EINVAL));
3435 
3436 		if ((error = nvlist_lookup_uint64(nvprops,
3437 		    zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3438 		    &volblocksize)) != 0 && error != ENOENT)
3439 			return (SET_ERROR(EINVAL));
3440 
3441 		if (error != 0)
3442 			volblocksize = zfs_prop_default_numeric(
3443 			    ZFS_PROP_VOLBLOCKSIZE);
3444 
3445 		if ((error = zvol_check_volblocksize(fsname,
3446 		    volblocksize)) != 0 ||
3447 		    (error = zvol_check_volsize(volsize,
3448 		    volblocksize)) != 0)
3449 			return (error);
3450 	} else if (type == DMU_OST_ZFS) {
3451 		int error;
3452 
3453 		/*
3454 		 * We have to have normalization and
3455 		 * case-folding flags correct when we do the
3456 		 * file system creation, so go figure them out
3457 		 * now.
3458 		 */
3459 		VERIFY(nvlist_alloc(&zct.zct_zplprops,
3460 		    NV_UNIQUE_NAME, KM_SLEEP) == 0);
3461 		error = zfs_fill_zplprops(fsname, nvprops,
3462 		    zct.zct_zplprops, &is_insensitive);
3463 		if (error != 0) {
3464 			nvlist_free(zct.zct_zplprops);
3465 			return (error);
3466 		}
3467 	}
3468 
3469 	error = dsl_crypto_params_create_nvlist(DCP_CMD_NONE, nvprops,
3470 	    hidden_args, &dcp);
3471 	if (error != 0) {
3472 		nvlist_free(zct.zct_zplprops);
3473 		return (error);
3474 	}
3475 
3476 	error = dmu_objset_create(fsname, type,
3477 	    is_insensitive ? DS_FLAG_CI_DATASET : 0, dcp, cbfunc, &zct);
3478 
3479 	nvlist_free(zct.zct_zplprops);
3480 	dsl_crypto_params_free(dcp, !!error);
3481 
3482 	/*
3483 	 * It would be nice to do this atomically.
3484 	 */
3485 	if (error == 0) {
3486 		error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3487 		    nvprops, outnvl);
3488 		if (error != 0) {
3489 			spa_t *spa;
3490 			int error2;
3491 
3492 			/*
3493 			 * Volumes will return EBUSY and cannot be destroyed
3494 			 * until all asynchronous minor handling (e.g. from
3495 			 * setting the volmode property) has completed. Wait for
3496 			 * the spa_zvol_taskq to drain then retry.
3497 			 */
3498 			error2 = dsl_destroy_head(fsname);
3499 			while ((error2 == EBUSY) && (type == DMU_OST_ZVOL)) {
3500 				error2 = spa_open(fsname, &spa, FTAG);
3501 				if (error2 == 0) {
3502 					taskq_wait(spa->spa_zvol_taskq);
3503 					spa_close(spa, FTAG);
3504 				}
3505 				error2 = dsl_destroy_head(fsname);
3506 			}
3507 		}
3508 	}
3509 	return (error);
3510 }
3511 
3512 /*
3513  * innvl: {
3514  *     "origin" -> name of origin snapshot
3515  *     (optional) "props" -> { prop -> value }
3516  *     (optional) "hidden_args" -> { "wkeydata" -> value }
3517  *         raw uint8_t array of encryption wrapping key data (32 bytes)
3518  * }
3519  *
3520  * outputs:
3521  * outnvl: propname -> error code (int32)
3522  */
3523 static const zfs_ioc_key_t zfs_keys_clone[] = {
3524 	{"origin",	DATA_TYPE_STRING,	0},
3525 	{"props",	DATA_TYPE_NVLIST,	ZK_OPTIONAL},
3526 	{"hidden_args",	DATA_TYPE_NVLIST,	ZK_OPTIONAL},
3527 };
3528 
3529 static int
3530 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3531 {
3532 	int error = 0;
3533 	nvlist_t *nvprops = NULL;
3534 	const char *origin_name;
3535 
3536 	origin_name = fnvlist_lookup_string(innvl, "origin");
3537 	(void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3538 
3539 	if (strchr(fsname, '@') ||
3540 	    strchr(fsname, '%'))
3541 		return (SET_ERROR(EINVAL));
3542 
3543 	if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3544 		return (SET_ERROR(EINVAL));
3545 
3546 	error = dmu_objset_clone(fsname, origin_name);
3547 
3548 	/*
3549 	 * It would be nice to do this atomically.
3550 	 */
3551 	if (error == 0) {
3552 		error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3553 		    nvprops, outnvl);
3554 		if (error != 0)
3555 			(void) dsl_destroy_head(fsname);
3556 	}
3557 	return (error);
3558 }
3559 
3560 static const zfs_ioc_key_t zfs_keys_remap[] = {
3561 	/* no nvl keys */
3562 };
3563 
3564 static int
3565 zfs_ioc_remap(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3566 {
3567 	/* This IOCTL is no longer supported. */
3568 	(void) fsname, (void) innvl, (void) outnvl;
3569 	return (0);
3570 }
3571 
3572 /*
3573  * innvl: {
3574  *     "snaps" -> { snapshot1, snapshot2 }
3575  *     (optional) "props" -> { prop -> value (string) }
3576  * }
3577  *
3578  * outnvl: snapshot -> error code (int32)
3579  */
3580 static const zfs_ioc_key_t zfs_keys_snapshot[] = {
3581 	{"snaps",	DATA_TYPE_NVLIST,	0},
3582 	{"props",	DATA_TYPE_NVLIST,	ZK_OPTIONAL},
3583 };
3584 
3585 static int
3586 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3587 {
3588 	nvlist_t *snaps;
3589 	nvlist_t *props = NULL;
3590 	int error, poollen;
3591 	nvpair_t *pair;
3592 
3593 	(void) nvlist_lookup_nvlist(innvl, "props", &props);
3594 	if (!nvlist_empty(props) &&
3595 	    zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3596 		return (SET_ERROR(ENOTSUP));
3597 	if ((error = zfs_check_userprops(props)) != 0)
3598 		return (error);
3599 
3600 	snaps = fnvlist_lookup_nvlist(innvl, "snaps");
3601 	poollen = strlen(poolname);
3602 	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3603 	    pair = nvlist_next_nvpair(snaps, pair)) {
3604 		const char *name = nvpair_name(pair);
3605 		char *cp = strchr(name, '@');
3606 
3607 		/*
3608 		 * The snap name must contain an @, and the part after it must
3609 		 * contain only valid characters.
3610 		 */
3611 		if (cp == NULL ||
3612 		    zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3613 			return (SET_ERROR(EINVAL));
3614 
3615 		/*
3616 		 * The snap must be in the specified pool.
3617 		 */
3618 		if (strncmp(name, poolname, poollen) != 0 ||
3619 		    (name[poollen] != '/' && name[poollen] != '@'))
3620 			return (SET_ERROR(EXDEV));
3621 
3622 		/*
3623 		 * Check for permission to set the properties on the fs.
3624 		 */
3625 		if (!nvlist_empty(props)) {
3626 			*cp = '\0';
3627 			error = zfs_secpolicy_write_perms(name,
3628 			    ZFS_DELEG_PERM_USERPROP, CRED());
3629 			*cp = '@';
3630 			if (error != 0)
3631 				return (error);
3632 		}
3633 
3634 		/* This must be the only snap of this fs. */
3635 		for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3636 		    pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3637 			if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3638 			    == 0) {
3639 				return (SET_ERROR(EXDEV));
3640 			}
3641 		}
3642 	}
3643 
3644 	error = dsl_dataset_snapshot(snaps, props, outnvl);
3645 
3646 	return (error);
3647 }
3648 
3649 /*
3650  * innvl: "message" -> string
3651  */
3652 static const zfs_ioc_key_t zfs_keys_log_history[] = {
3653 	{"message",	DATA_TYPE_STRING,	0},
3654 };
3655 
3656 static int
3657 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3658 {
3659 	(void) unused, (void) outnvl;
3660 	const char *message;
3661 	char *poolname;
3662 	spa_t *spa;
3663 	int error;
3664 
3665 	/*
3666 	 * The poolname in the ioctl is not set, we get it from the TSD,
3667 	 * which was set at the end of the last successful ioctl that allows
3668 	 * logging.  The secpolicy func already checked that it is set.
3669 	 * Only one log ioctl is allowed after each successful ioctl, so
3670 	 * we clear the TSD here.
3671 	 */
3672 	poolname = tsd_get(zfs_allow_log_key);
3673 	if (poolname == NULL)
3674 		return (SET_ERROR(EINVAL));
3675 	(void) tsd_set(zfs_allow_log_key, NULL);
3676 	error = spa_open(poolname, &spa, FTAG);
3677 	kmem_strfree(poolname);
3678 	if (error != 0)
3679 		return (error);
3680 
3681 	message = fnvlist_lookup_string(innvl, "message");
3682 
3683 	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3684 		spa_close(spa, FTAG);
3685 		return (SET_ERROR(ENOTSUP));
3686 	}
3687 
3688 	error = spa_history_log(spa, message);
3689 	spa_close(spa, FTAG);
3690 	return (error);
3691 }
3692 
3693 /*
3694  * This ioctl is used to set the bootenv configuration on the current
3695  * pool. This configuration is stored in the second padding area of the label,
3696  * and it is used by the bootloader(s) to store the bootloader and/or system
3697  * specific data.
3698  * The data is stored as nvlist data stream, and is protected by
3699  * an embedded checksum.
3700  * The version can have two possible values:
3701  * VB_RAW: nvlist should have key GRUB_ENVMAP, value DATA_TYPE_STRING.
3702  * VB_NVLIST: nvlist with arbitrary <key, value> pairs.
3703  */
3704 static const zfs_ioc_key_t zfs_keys_set_bootenv[] = {
3705 	{"version",	DATA_TYPE_UINT64,	0},
3706 	{"<keys>",	DATA_TYPE_ANY, ZK_OPTIONAL | ZK_WILDCARDLIST},
3707 };
3708 
3709 static int
3710 zfs_ioc_set_bootenv(const char *name, nvlist_t *innvl, nvlist_t *outnvl)
3711 {
3712 	int error;
3713 	spa_t *spa;
3714 
3715 	if ((error = spa_open(name, &spa, FTAG)) != 0)
3716 		return (error);
3717 	spa_vdev_state_enter(spa, SCL_ALL);
3718 	error = vdev_label_write_bootenv(spa->spa_root_vdev, innvl);
3719 	(void) spa_vdev_state_exit(spa, NULL, 0);
3720 	spa_close(spa, FTAG);
3721 	return (error);
3722 }
3723 
3724 static const zfs_ioc_key_t zfs_keys_get_bootenv[] = {
3725 	/* no nvl keys */
3726 };
3727 
3728 static int
3729 zfs_ioc_get_bootenv(const char *name, nvlist_t *innvl, nvlist_t *outnvl)
3730 {
3731 	spa_t *spa;
3732 	int error;
3733 
3734 	if ((error = spa_open(name, &spa, FTAG)) != 0)
3735 		return (error);
3736 	spa_vdev_state_enter(spa, SCL_ALL);
3737 	error = vdev_label_read_bootenv(spa->spa_root_vdev, outnvl);
3738 	(void) spa_vdev_state_exit(spa, NULL, 0);
3739 	spa_close(spa, FTAG);
3740 	return (error);
3741 }
3742 
3743 /*
3744  * The dp_config_rwlock must not be held when calling this, because the
3745  * unmount may need to write out data.
3746  *
3747  * This function is best-effort.  Callers must deal gracefully if it
3748  * remains mounted (or is remounted after this call).
3749  *
3750  * Returns 0 if the argument is not a snapshot, or it is not currently a
3751  * filesystem, or we were able to unmount it.  Returns error code otherwise.
3752  */
3753 void
3754 zfs_unmount_snap(const char *snapname)
3755 {
3756 	if (strchr(snapname, '@') == NULL)
3757 		return;
3758 
3759 	(void) zfsctl_snapshot_unmount(snapname, MNT_FORCE);
3760 }
3761 
3762 static int
3763 zfs_unmount_snap_cb(const char *snapname, void *arg)
3764 {
3765 	(void) arg;
3766 	zfs_unmount_snap(snapname);
3767 	return (0);
3768 }
3769 
3770 /*
3771  * When a clone is destroyed, its origin may also need to be destroyed,
3772  * in which case it must be unmounted.  This routine will do that unmount
3773  * if necessary.
3774  */
3775 void
3776 zfs_destroy_unmount_origin(const char *fsname)
3777 {
3778 	int error;
3779 	objset_t *os;
3780 	dsl_dataset_t *ds;
3781 
3782 	error = dmu_objset_hold(fsname, FTAG, &os);
3783 	if (error != 0)
3784 		return;
3785 	ds = dmu_objset_ds(os);
3786 	if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3787 		char originname[ZFS_MAX_DATASET_NAME_LEN];
3788 		dsl_dataset_name(ds->ds_prev, originname);
3789 		dmu_objset_rele(os, FTAG);
3790 		zfs_unmount_snap(originname);
3791 	} else {
3792 		dmu_objset_rele(os, FTAG);
3793 	}
3794 }
3795 
3796 /*
3797  * innvl: {
3798  *     "snaps" -> { snapshot1, snapshot2 }
3799  *     (optional boolean) "defer"
3800  * }
3801  *
3802  * outnvl: snapshot -> error code (int32)
3803  */
3804 static const zfs_ioc_key_t zfs_keys_destroy_snaps[] = {
3805 	{"snaps",	DATA_TYPE_NVLIST,	0},
3806 	{"defer",	DATA_TYPE_BOOLEAN,	ZK_OPTIONAL},
3807 };
3808 
3809 static int
3810 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3811 {
3812 	int poollen;
3813 	nvlist_t *snaps;
3814 	nvpair_t *pair;
3815 	boolean_t defer;
3816 	spa_t *spa;
3817 
3818 	snaps = fnvlist_lookup_nvlist(innvl, "snaps");
3819 	defer = nvlist_exists(innvl, "defer");
3820 
3821 	poollen = strlen(poolname);
3822 	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3823 	    pair = nvlist_next_nvpair(snaps, pair)) {
3824 		const char *name = nvpair_name(pair);
3825 
3826 		/*
3827 		 * The snap must be in the specified pool to prevent the
3828 		 * invalid removal of zvol minors below.
3829 		 */
3830 		if (strncmp(name, poolname, poollen) != 0 ||
3831 		    (name[poollen] != '/' && name[poollen] != '@'))
3832 			return (SET_ERROR(EXDEV));
3833 
3834 		zfs_unmount_snap(nvpair_name(pair));
3835 		if (spa_open(name, &spa, FTAG) == 0) {
3836 			zvol_remove_minors(spa, name, B_TRUE);
3837 			spa_close(spa, FTAG);
3838 		}
3839 	}
3840 
3841 	return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3842 }
3843 
3844 /*
3845  * Create bookmarks. The bookmark names are of the form <fs>#<bmark>.
3846  * All bookmarks and snapshots must be in the same pool.
3847  * dsl_bookmark_create_nvl_validate describes the nvlist schema in more detail.
3848  *
3849  * innvl: {
3850  *     new_bookmark1 -> existing_snapshot,
3851  *     new_bookmark2 -> existing_bookmark,
3852  * }
3853  *
3854  * outnvl: bookmark -> error code (int32)
3855  *
3856  */
3857 static const zfs_ioc_key_t zfs_keys_bookmark[] = {
3858 	{"<bookmark>...",	DATA_TYPE_STRING,	ZK_WILDCARDLIST},
3859 };
3860 
3861 static int
3862 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3863 {
3864 	(void) poolname;
3865 	return (dsl_bookmark_create(innvl, outnvl));
3866 }
3867 
3868 /*
3869  * innvl: {
3870  *     property 1, property 2, ...
3871  * }
3872  *
3873  * outnvl: {
3874  *     bookmark name 1 -> { property 1, property 2, ... },
3875  *     bookmark name 2 -> { property 1, property 2, ... }
3876  * }
3877  *
3878  */
3879 static const zfs_ioc_key_t zfs_keys_get_bookmarks[] = {
3880 	{"<property>...", DATA_TYPE_BOOLEAN, ZK_WILDCARDLIST | ZK_OPTIONAL},
3881 };
3882 
3883 static int
3884 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3885 {
3886 	return (dsl_get_bookmarks(fsname, innvl, outnvl));
3887 }
3888 
3889 /*
3890  * innvl is not used.
3891  *
3892  * outnvl: {
3893  *     property 1, property 2, ...
3894  * }
3895  *
3896  */
3897 static const zfs_ioc_key_t zfs_keys_get_bookmark_props[] = {
3898 	/* no nvl keys */
3899 };
3900 
3901 static int
3902 zfs_ioc_get_bookmark_props(const char *bookmark, nvlist_t *innvl,
3903     nvlist_t *outnvl)
3904 {
3905 	(void) innvl;
3906 	char fsname[ZFS_MAX_DATASET_NAME_LEN];
3907 	char *bmname;
3908 
3909 	bmname = strchr(bookmark, '#');
3910 	if (bmname == NULL)
3911 		return (SET_ERROR(EINVAL));
3912 	bmname++;
3913 
3914 	(void) strlcpy(fsname, bookmark, sizeof (fsname));
3915 	*(strchr(fsname, '#')) = '\0';
3916 
3917 	return (dsl_get_bookmark_props(fsname, bmname, outnvl));
3918 }
3919 
3920 /*
3921  * innvl: {
3922  *     bookmark name 1, bookmark name 2
3923  * }
3924  *
3925  * outnvl: bookmark -> error code (int32)
3926  *
3927  */
3928 static const zfs_ioc_key_t zfs_keys_destroy_bookmarks[] = {
3929 	{"<bookmark>...",	DATA_TYPE_BOOLEAN,	ZK_WILDCARDLIST},
3930 };
3931 
3932 static int
3933 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
3934     nvlist_t *outnvl)
3935 {
3936 	int error, poollen;
3937 
3938 	poollen = strlen(poolname);
3939 	for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3940 	    pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3941 		const char *name = nvpair_name(pair);
3942 		const char *cp = strchr(name, '#');
3943 
3944 		/*
3945 		 * The bookmark name must contain an #, and the part after it
3946 		 * must contain only valid characters.
3947 		 */
3948 		if (cp == NULL ||
3949 		    zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3950 			return (SET_ERROR(EINVAL));
3951 
3952 		/*
3953 		 * The bookmark must be in the specified pool.
3954 		 */
3955 		if (strncmp(name, poolname, poollen) != 0 ||
3956 		    (name[poollen] != '/' && name[poollen] != '#'))
3957 			return (SET_ERROR(EXDEV));
3958 	}
3959 
3960 	error = dsl_bookmark_destroy(innvl, outnvl);
3961 	return (error);
3962 }
3963 
3964 static const zfs_ioc_key_t zfs_keys_channel_program[] = {
3965 	{"program",	DATA_TYPE_STRING,		0},
3966 	{"arg",		DATA_TYPE_ANY,			0},
3967 	{"sync",	DATA_TYPE_BOOLEAN_VALUE,	ZK_OPTIONAL},
3968 	{"instrlimit",	DATA_TYPE_UINT64,		ZK_OPTIONAL},
3969 	{"memlimit",	DATA_TYPE_UINT64,		ZK_OPTIONAL},
3970 };
3971 
3972 static int
3973 zfs_ioc_channel_program(const char *poolname, nvlist_t *innvl,
3974     nvlist_t *outnvl)
3975 {
3976 	const char *program;
3977 	uint64_t instrlimit, memlimit;
3978 	boolean_t sync_flag;
3979 	nvpair_t *nvarg = NULL;
3980 
3981 	program = fnvlist_lookup_string(innvl, ZCP_ARG_PROGRAM);
3982 	if (0 != nvlist_lookup_boolean_value(innvl, ZCP_ARG_SYNC, &sync_flag)) {
3983 		sync_flag = B_TRUE;
3984 	}
3985 	if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_INSTRLIMIT, &instrlimit)) {
3986 		instrlimit = ZCP_DEFAULT_INSTRLIMIT;
3987 	}
3988 	if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_MEMLIMIT, &memlimit)) {
3989 		memlimit = ZCP_DEFAULT_MEMLIMIT;
3990 	}
3991 	nvarg = fnvlist_lookup_nvpair(innvl, ZCP_ARG_ARGLIST);
3992 
3993 	if (instrlimit == 0 || instrlimit > zfs_lua_max_instrlimit)
3994 		return (SET_ERROR(EINVAL));
3995 	if (memlimit == 0 || memlimit > zfs_lua_max_memlimit)
3996 		return (SET_ERROR(EINVAL));
3997 
3998 	return (zcp_eval(poolname, program, sync_flag, instrlimit, memlimit,
3999 	    nvarg, outnvl));
4000 }
4001 
4002 /*
4003  * innvl: unused
4004  * outnvl: empty
4005  */
4006 static const zfs_ioc_key_t zfs_keys_pool_checkpoint[] = {
4007 	/* no nvl keys */
4008 };
4009 
4010 static int
4011 zfs_ioc_pool_checkpoint(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
4012 {
4013 	(void) innvl, (void) outnvl;
4014 	return (spa_checkpoint(poolname));
4015 }
4016 
4017 /*
4018  * innvl: unused
4019  * outnvl: empty
4020  */
4021 static const zfs_ioc_key_t zfs_keys_pool_discard_checkpoint[] = {
4022 	/* no nvl keys */
4023 };
4024 
4025 static int
4026 zfs_ioc_pool_discard_checkpoint(const char *poolname, nvlist_t *innvl,
4027     nvlist_t *outnvl)
4028 {
4029 	(void) innvl, (void) outnvl;
4030 	return (spa_checkpoint_discard(poolname));
4031 }
4032 
4033 /*
4034  * inputs:
4035  * zc_name		name of dataset to destroy
4036  * zc_defer_destroy	mark for deferred destroy
4037  *
4038  * outputs:		none
4039  */
4040 static int
4041 zfs_ioc_destroy(zfs_cmd_t *zc)
4042 {
4043 	objset_t *os;
4044 	dmu_objset_type_t ost;
4045 	int err;
4046 
4047 	err = dmu_objset_hold(zc->zc_name, FTAG, &os);
4048 	if (err != 0)
4049 		return (err);
4050 	ost = dmu_objset_type(os);
4051 	dmu_objset_rele(os, FTAG);
4052 
4053 	if (ost == DMU_OST_ZFS)
4054 		zfs_unmount_snap(zc->zc_name);
4055 
4056 	if (strchr(zc->zc_name, '@')) {
4057 		err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
4058 	} else {
4059 		err = dsl_destroy_head(zc->zc_name);
4060 		if (err == EEXIST) {
4061 			/*
4062 			 * It is possible that the given DS may have
4063 			 * hidden child (%recv) datasets - "leftovers"
4064 			 * resulting from the previously interrupted
4065 			 * 'zfs receive'.
4066 			 *
4067 			 * 6 extra bytes for /%recv
4068 			 */
4069 			char namebuf[ZFS_MAX_DATASET_NAME_LEN + 6];
4070 
4071 			if (snprintf(namebuf, sizeof (namebuf), "%s/%s",
4072 			    zc->zc_name, recv_clone_name) >=
4073 			    sizeof (namebuf))
4074 				return (SET_ERROR(EINVAL));
4075 
4076 			/*
4077 			 * Try to remove the hidden child (%recv) and after
4078 			 * that try to remove the target dataset.
4079 			 * If the hidden child (%recv) does not exist
4080 			 * the original error (EEXIST) will be returned
4081 			 */
4082 			err = dsl_destroy_head(namebuf);
4083 			if (err == 0)
4084 				err = dsl_destroy_head(zc->zc_name);
4085 			else if (err == ENOENT)
4086 				err = SET_ERROR(EEXIST);
4087 		}
4088 	}
4089 
4090 	return (err);
4091 }
4092 
4093 /*
4094  * innvl: {
4095  *     "initialize_command" -> POOL_INITIALIZE_{CANCEL|START|SUSPEND} (uint64)
4096  *     "initialize_vdevs": { -> guids to initialize (nvlist)
4097  *         "vdev_path_1": vdev_guid_1, (uint64),
4098  *         "vdev_path_2": vdev_guid_2, (uint64),
4099  *         ...
4100  *     },
4101  * }
4102  *
4103  * outnvl: {
4104  *     "initialize_vdevs": { -> initialization errors (nvlist)
4105  *         "vdev_path_1": errno, see function body for possible errnos (uint64)
4106  *         "vdev_path_2": errno, ... (uint64)
4107  *         ...
4108  *     }
4109  * }
4110  *
4111  * EINVAL is returned for an unknown commands or if any of the provided vdev
4112  * guids have be specified with a type other than uint64.
4113  */
4114 static const zfs_ioc_key_t zfs_keys_pool_initialize[] = {
4115 	{ZPOOL_INITIALIZE_COMMAND,	DATA_TYPE_UINT64,	0},
4116 	{ZPOOL_INITIALIZE_VDEVS,	DATA_TYPE_NVLIST,	0}
4117 };
4118 
4119 static int
4120 zfs_ioc_pool_initialize(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
4121 {
4122 	uint64_t cmd_type;
4123 	if (nvlist_lookup_uint64(innvl, ZPOOL_INITIALIZE_COMMAND,
4124 	    &cmd_type) != 0) {
4125 		return (SET_ERROR(EINVAL));
4126 	}
4127 
4128 	if (!(cmd_type == POOL_INITIALIZE_CANCEL ||
4129 	    cmd_type == POOL_INITIALIZE_START ||
4130 	    cmd_type == POOL_INITIALIZE_SUSPEND ||
4131 	    cmd_type == POOL_INITIALIZE_UNINIT)) {
4132 		return (SET_ERROR(EINVAL));
4133 	}
4134 
4135 	nvlist_t *vdev_guids;
4136 	if (nvlist_lookup_nvlist(innvl, ZPOOL_INITIALIZE_VDEVS,
4137 	    &vdev_guids) != 0) {
4138 		return (SET_ERROR(EINVAL));
4139 	}
4140 
4141 	for (nvpair_t *pair = nvlist_next_nvpair(vdev_guids, NULL);
4142 	    pair != NULL; pair = nvlist_next_nvpair(vdev_guids, pair)) {
4143 		uint64_t vdev_guid;
4144 		if (nvpair_value_uint64(pair, &vdev_guid) != 0) {
4145 			return (SET_ERROR(EINVAL));
4146 		}
4147 	}
4148 
4149 	spa_t *spa;
4150 	int error = spa_open(poolname, &spa, FTAG);
4151 	if (error != 0)
4152 		return (error);
4153 
4154 	nvlist_t *vdev_errlist = fnvlist_alloc();
4155 	int total_errors = spa_vdev_initialize(spa, vdev_guids, cmd_type,
4156 	    vdev_errlist);
4157 
4158 	if (fnvlist_size(vdev_errlist) > 0) {
4159 		fnvlist_add_nvlist(outnvl, ZPOOL_INITIALIZE_VDEVS,
4160 		    vdev_errlist);
4161 	}
4162 	fnvlist_free(vdev_errlist);
4163 
4164 	spa_close(spa, FTAG);
4165 	return (total_errors > 0 ? SET_ERROR(EINVAL) : 0);
4166 }
4167 
4168 /*
4169  * innvl: {
4170  *     "trim_command" -> POOL_TRIM_{CANCEL|START|SUSPEND} (uint64)
4171  *     "trim_vdevs": { -> guids to TRIM (nvlist)
4172  *         "vdev_path_1": vdev_guid_1, (uint64),
4173  *         "vdev_path_2": vdev_guid_2, (uint64),
4174  *         ...
4175  *     },
4176  *     "trim_rate" -> Target TRIM rate in bytes/sec.
4177  *     "trim_secure" -> Set to request a secure TRIM.
4178  * }
4179  *
4180  * outnvl: {
4181  *     "trim_vdevs": { -> TRIM errors (nvlist)
4182  *         "vdev_path_1": errno, see function body for possible errnos (uint64)
4183  *         "vdev_path_2": errno, ... (uint64)
4184  *         ...
4185  *     }
4186  * }
4187  *
4188  * EINVAL is returned for an unknown commands or if any of the provided vdev
4189  * guids have be specified with a type other than uint64.
4190  */
4191 static const zfs_ioc_key_t zfs_keys_pool_trim[] = {
4192 	{ZPOOL_TRIM_COMMAND,	DATA_TYPE_UINT64,		0},
4193 	{ZPOOL_TRIM_VDEVS,	DATA_TYPE_NVLIST,		0},
4194 	{ZPOOL_TRIM_RATE,	DATA_TYPE_UINT64,		ZK_OPTIONAL},
4195 	{ZPOOL_TRIM_SECURE,	DATA_TYPE_BOOLEAN_VALUE,	ZK_OPTIONAL},
4196 };
4197 
4198 static int
4199 zfs_ioc_pool_trim(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
4200 {
4201 	uint64_t cmd_type;
4202 	if (nvlist_lookup_uint64(innvl, ZPOOL_TRIM_COMMAND, &cmd_type) != 0)
4203 		return (SET_ERROR(EINVAL));
4204 
4205 	if (!(cmd_type == POOL_TRIM_CANCEL ||
4206 	    cmd_type == POOL_TRIM_START ||
4207 	    cmd_type == POOL_TRIM_SUSPEND)) {
4208 		return (SET_ERROR(EINVAL));
4209 	}
4210 
4211 	nvlist_t *vdev_guids;
4212 	if (nvlist_lookup_nvlist(innvl, ZPOOL_TRIM_VDEVS, &vdev_guids) != 0)
4213 		return (SET_ERROR(EINVAL));
4214 
4215 	for (nvpair_t *pair = nvlist_next_nvpair(vdev_guids, NULL);
4216 	    pair != NULL; pair = nvlist_next_nvpair(vdev_guids, pair)) {
4217 		uint64_t vdev_guid;
4218 		if (nvpair_value_uint64(pair, &vdev_guid) != 0) {
4219 			return (SET_ERROR(EINVAL));
4220 		}
4221 	}
4222 
4223 	/* Optional, defaults to maximum rate when not provided */
4224 	uint64_t rate;
4225 	if (nvlist_lookup_uint64(innvl, ZPOOL_TRIM_RATE, &rate) != 0)
4226 		rate = 0;
4227 
4228 	/* Optional, defaults to standard TRIM when not provided */
4229 	boolean_t secure;
4230 	if (nvlist_lookup_boolean_value(innvl, ZPOOL_TRIM_SECURE,
4231 	    &secure) != 0) {
4232 		secure = B_FALSE;
4233 	}
4234 
4235 	spa_t *spa;
4236 	int error = spa_open(poolname, &spa, FTAG);
4237 	if (error != 0)
4238 		return (error);
4239 
4240 	nvlist_t *vdev_errlist = fnvlist_alloc();
4241 	int total_errors = spa_vdev_trim(spa, vdev_guids, cmd_type,
4242 	    rate, !!zfs_trim_metaslab_skip, secure, vdev_errlist);
4243 
4244 	if (fnvlist_size(vdev_errlist) > 0)
4245 		fnvlist_add_nvlist(outnvl, ZPOOL_TRIM_VDEVS, vdev_errlist);
4246 
4247 	fnvlist_free(vdev_errlist);
4248 
4249 	spa_close(spa, FTAG);
4250 	return (total_errors > 0 ? SET_ERROR(EINVAL) : 0);
4251 }
4252 
4253 /*
4254  * This ioctl waits for activity of a particular type to complete. If there is
4255  * no activity of that type in progress, it returns immediately, and the
4256  * returned value "waited" is false. If there is activity in progress, and no
4257  * tag is passed in, the ioctl blocks until all activity of that type is
4258  * complete, and then returns with "waited" set to true.
4259  *
4260  * If a tag is provided, it identifies a particular instance of an activity to
4261  * wait for. Currently, this is only valid for use with 'initialize', because
4262  * that is the only activity for which there can be multiple instances running
4263  * concurrently. In the case of 'initialize', the tag corresponds to the guid of
4264  * the vdev on which to wait.
4265  *
4266  * If a thread waiting in the ioctl receives a signal, the call will return
4267  * immediately, and the return value will be EINTR.
4268  *
4269  * innvl: {
4270  *     "wait_activity" -> int32_t
4271  *     (optional) "wait_tag" -> uint64_t
4272  * }
4273  *
4274  * outnvl: "waited" -> boolean_t
4275  */
4276 static const zfs_ioc_key_t zfs_keys_pool_wait[] = {
4277 	{ZPOOL_WAIT_ACTIVITY,	DATA_TYPE_INT32,		0},
4278 	{ZPOOL_WAIT_TAG,	DATA_TYPE_UINT64,		ZK_OPTIONAL},
4279 };
4280 
4281 static int
4282 zfs_ioc_wait(const char *name, nvlist_t *innvl, nvlist_t *outnvl)
4283 {
4284 	int32_t activity;
4285 	uint64_t tag;
4286 	boolean_t waited;
4287 	int error;
4288 
4289 	if (nvlist_lookup_int32(innvl, ZPOOL_WAIT_ACTIVITY, &activity) != 0)
4290 		return (EINVAL);
4291 
4292 	if (nvlist_lookup_uint64(innvl, ZPOOL_WAIT_TAG, &tag) == 0)
4293 		error = spa_wait_tag(name, activity, tag, &waited);
4294 	else
4295 		error = spa_wait(name, activity, &waited);
4296 
4297 	if (error == 0)
4298 		fnvlist_add_boolean_value(outnvl, ZPOOL_WAIT_WAITED, waited);
4299 
4300 	return (error);
4301 }
4302 
4303 /*
4304  * This ioctl waits for activity of a particular type to complete. If there is
4305  * no activity of that type in progress, it returns immediately, and the
4306  * returned value "waited" is false. If there is activity in progress, and no
4307  * tag is passed in, the ioctl blocks until all activity of that type is
4308  * complete, and then returns with "waited" set to true.
4309  *
4310  * If a thread waiting in the ioctl receives a signal, the call will return
4311  * immediately, and the return value will be EINTR.
4312  *
4313  * innvl: {
4314  *     "wait_activity" -> int32_t
4315  * }
4316  *
4317  * outnvl: "waited" -> boolean_t
4318  */
4319 static const zfs_ioc_key_t zfs_keys_fs_wait[] = {
4320 	{ZFS_WAIT_ACTIVITY,	DATA_TYPE_INT32,		0},
4321 };
4322 
4323 static int
4324 zfs_ioc_wait_fs(const char *name, nvlist_t *innvl, nvlist_t *outnvl)
4325 {
4326 	int32_t activity;
4327 	boolean_t waited = B_FALSE;
4328 	int error;
4329 	dsl_pool_t *dp;
4330 	dsl_dir_t *dd;
4331 	dsl_dataset_t *ds;
4332 
4333 	if (nvlist_lookup_int32(innvl, ZFS_WAIT_ACTIVITY, &activity) != 0)
4334 		return (SET_ERROR(EINVAL));
4335 
4336 	if (activity >= ZFS_WAIT_NUM_ACTIVITIES || activity < 0)
4337 		return (SET_ERROR(EINVAL));
4338 
4339 	if ((error = dsl_pool_hold(name, FTAG, &dp)) != 0)
4340 		return (error);
4341 
4342 	if ((error = dsl_dataset_hold(dp, name, FTAG, &ds)) != 0) {
4343 		dsl_pool_rele(dp, FTAG);
4344 		return (error);
4345 	}
4346 
4347 	dd = ds->ds_dir;
4348 	mutex_enter(&dd->dd_activity_lock);
4349 	dd->dd_activity_waiters++;
4350 
4351 	/*
4352 	 * We get a long-hold here so that the dsl_dataset_t and dsl_dir_t
4353 	 * aren't evicted while we're waiting. Normally this is prevented by
4354 	 * holding the pool, but we can't do that while we're waiting since
4355 	 * that would prevent TXGs from syncing out. Some of the functionality
4356 	 * of long-holds (e.g. preventing deletion) is unnecessary for this
4357 	 * case, since we would cancel the waiters before proceeding with a
4358 	 * deletion. An alternative mechanism for keeping the dataset around
4359 	 * could be developed but this is simpler.
4360 	 */
4361 	dsl_dataset_long_hold(ds, FTAG);
4362 	dsl_pool_rele(dp, FTAG);
4363 
4364 	error = dsl_dir_wait(dd, ds, activity, &waited);
4365 
4366 	dsl_dataset_long_rele(ds, FTAG);
4367 	dd->dd_activity_waiters--;
4368 	if (dd->dd_activity_waiters == 0)
4369 		cv_signal(&dd->dd_activity_cv);
4370 	mutex_exit(&dd->dd_activity_lock);
4371 
4372 	dsl_dataset_rele(ds, FTAG);
4373 
4374 	if (error == 0)
4375 		fnvlist_add_boolean_value(outnvl, ZFS_WAIT_WAITED, waited);
4376 
4377 	return (error);
4378 }
4379 
4380 /*
4381  * fsname is name of dataset to rollback (to most recent snapshot)
4382  *
4383  * innvl may contain name of expected target snapshot
4384  *
4385  * outnvl: "target" -> name of most recent snapshot
4386  * }
4387  */
4388 static const zfs_ioc_key_t zfs_keys_rollback[] = {
4389 	{"target",	DATA_TYPE_STRING,	ZK_OPTIONAL},
4390 };
4391 
4392 static int
4393 zfs_ioc_rollback(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
4394 {
4395 	zfsvfs_t *zfsvfs;
4396 	zvol_state_handle_t *zv;
4397 	const char *target = NULL;
4398 	int error;
4399 
4400 	(void) nvlist_lookup_string(innvl, "target", &target);
4401 	if (target != NULL) {
4402 		const char *cp = strchr(target, '@');
4403 
4404 		/*
4405 		 * The snap name must contain an @, and the part after it must
4406 		 * contain only valid characters.
4407 		 */
4408 		if (cp == NULL ||
4409 		    zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
4410 			return (SET_ERROR(EINVAL));
4411 	}
4412 
4413 	if (getzfsvfs(fsname, &zfsvfs) == 0) {
4414 		dsl_dataset_t *ds;
4415 
4416 		ds = dmu_objset_ds(zfsvfs->z_os);
4417 		error = zfs_suspend_fs(zfsvfs);
4418 		if (error == 0) {
4419 			int resume_err;
4420 
4421 			error = dsl_dataset_rollback(fsname, target, zfsvfs,
4422 			    outnvl);
4423 			resume_err = zfs_resume_fs(zfsvfs, ds);
4424 			error = error ? error : resume_err;
4425 		}
4426 		zfs_vfs_rele(zfsvfs);
4427 	} else if ((zv = zvol_suspend(fsname)) != NULL) {
4428 		error = dsl_dataset_rollback(fsname, target, zvol_tag(zv),
4429 		    outnvl);
4430 		zvol_resume(zv);
4431 	} else {
4432 		error = dsl_dataset_rollback(fsname, target, NULL, outnvl);
4433 	}
4434 	return (error);
4435 }
4436 
4437 static int
4438 recursive_unmount(const char *fsname, void *arg)
4439 {
4440 	const char *snapname = arg;
4441 	char *fullname;
4442 
4443 	fullname = kmem_asprintf("%s@%s", fsname, snapname);
4444 	zfs_unmount_snap(fullname);
4445 	kmem_strfree(fullname);
4446 
4447 	return (0);
4448 }
4449 
4450 /*
4451  *
4452  * snapname is the snapshot to redact.
4453  * innvl: {
4454  *     "bookname" -> (string)
4455  *         shortname of the redaction bookmark to generate
4456  *     "snapnv" -> (nvlist, values ignored)
4457  *         snapshots to redact snapname with respect to
4458  * }
4459  *
4460  * outnvl is unused
4461  */
4462 
4463 static const zfs_ioc_key_t zfs_keys_redact[] = {
4464 	{"bookname",		DATA_TYPE_STRING,	0},
4465 	{"snapnv",		DATA_TYPE_NVLIST,	0},
4466 };
4467 
4468 static int
4469 zfs_ioc_redact(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
4470 {
4471 	(void) outnvl;
4472 	nvlist_t *redactnvl = NULL;
4473 	const char *redactbook = NULL;
4474 
4475 	if (nvlist_lookup_nvlist(innvl, "snapnv", &redactnvl) != 0)
4476 		return (SET_ERROR(EINVAL));
4477 	if (fnvlist_num_pairs(redactnvl) == 0)
4478 		return (SET_ERROR(ENXIO));
4479 	if (nvlist_lookup_string(innvl, "bookname", &redactbook) != 0)
4480 		return (SET_ERROR(EINVAL));
4481 
4482 	return (dmu_redact_snap(snapname, redactnvl, redactbook));
4483 }
4484 
4485 /*
4486  * inputs:
4487  * zc_name	old name of dataset
4488  * zc_value	new name of dataset
4489  * zc_cookie	recursive flag (only valid for snapshots)
4490  *
4491  * outputs:	none
4492  */
4493 static int
4494 zfs_ioc_rename(zfs_cmd_t *zc)
4495 {
4496 	objset_t *os;
4497 	dmu_objset_type_t ost;
4498 	boolean_t recursive = zc->zc_cookie & 1;
4499 	boolean_t nounmount = !!(zc->zc_cookie & 2);
4500 	char *at;
4501 	int err;
4502 
4503 	/* "zfs rename" from and to ...%recv datasets should both fail */
4504 	zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
4505 	zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
4506 	if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0 ||
4507 	    dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4508 	    strchr(zc->zc_name, '%') || strchr(zc->zc_value, '%'))
4509 		return (SET_ERROR(EINVAL));
4510 
4511 	err = dmu_objset_hold(zc->zc_name, FTAG, &os);
4512 	if (err != 0)
4513 		return (err);
4514 	ost = dmu_objset_type(os);
4515 	dmu_objset_rele(os, FTAG);
4516 
4517 	at = strchr(zc->zc_name, '@');
4518 	if (at != NULL) {
4519 		/* snaps must be in same fs */
4520 		int error;
4521 
4522 		if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
4523 			return (SET_ERROR(EXDEV));
4524 		*at = '\0';
4525 		if (ost == DMU_OST_ZFS && !nounmount) {
4526 			error = dmu_objset_find(zc->zc_name,
4527 			    recursive_unmount, at + 1,
4528 			    recursive ? DS_FIND_CHILDREN : 0);
4529 			if (error != 0) {
4530 				*at = '@';
4531 				return (error);
4532 			}
4533 		}
4534 		error = dsl_dataset_rename_snapshot(zc->zc_name,
4535 		    at + 1, strchr(zc->zc_value, '@') + 1, recursive);
4536 		*at = '@';
4537 
4538 		return (error);
4539 	} else {
4540 		return (dsl_dir_rename(zc->zc_name, zc->zc_value));
4541 	}
4542 }
4543 
4544 static int
4545 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
4546 {
4547 	const char *propname = nvpair_name(pair);
4548 	boolean_t issnap = (strchr(dsname, '@') != NULL);
4549 	zfs_prop_t prop = zfs_name_to_prop(propname);
4550 	uint64_t intval, compval;
4551 	int err;
4552 
4553 	if (prop == ZPROP_USERPROP) {
4554 		if (zfs_prop_user(propname)) {
4555 			if ((err = zfs_secpolicy_write_perms(dsname,
4556 			    ZFS_DELEG_PERM_USERPROP, cr)))
4557 				return (err);
4558 			return (0);
4559 		}
4560 
4561 		if (!issnap && zfs_prop_userquota(propname)) {
4562 			const char *perm = NULL;
4563 			const char *uq_prefix =
4564 			    zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
4565 			const char *gq_prefix =
4566 			    zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
4567 			const char *uiq_prefix =
4568 			    zfs_userquota_prop_prefixes[ZFS_PROP_USEROBJQUOTA];
4569 			const char *giq_prefix =
4570 			    zfs_userquota_prop_prefixes[ZFS_PROP_GROUPOBJQUOTA];
4571 			const char *pq_prefix =
4572 			    zfs_userquota_prop_prefixes[ZFS_PROP_PROJECTQUOTA];
4573 			const char *piq_prefix = zfs_userquota_prop_prefixes[\
4574 			    ZFS_PROP_PROJECTOBJQUOTA];
4575 
4576 			if (strncmp(propname, uq_prefix,
4577 			    strlen(uq_prefix)) == 0) {
4578 				perm = ZFS_DELEG_PERM_USERQUOTA;
4579 			} else if (strncmp(propname, uiq_prefix,
4580 			    strlen(uiq_prefix)) == 0) {
4581 				perm = ZFS_DELEG_PERM_USEROBJQUOTA;
4582 			} else if (strncmp(propname, gq_prefix,
4583 			    strlen(gq_prefix)) == 0) {
4584 				perm = ZFS_DELEG_PERM_GROUPQUOTA;
4585 			} else if (strncmp(propname, giq_prefix,
4586 			    strlen(giq_prefix)) == 0) {
4587 				perm = ZFS_DELEG_PERM_GROUPOBJQUOTA;
4588 			} else if (strncmp(propname, pq_prefix,
4589 			    strlen(pq_prefix)) == 0) {
4590 				perm = ZFS_DELEG_PERM_PROJECTQUOTA;
4591 			} else if (strncmp(propname, piq_prefix,
4592 			    strlen(piq_prefix)) == 0) {
4593 				perm = ZFS_DELEG_PERM_PROJECTOBJQUOTA;
4594 			} else {
4595 				/* {USER|GROUP|PROJECT}USED are read-only */
4596 				return (SET_ERROR(EINVAL));
4597 			}
4598 
4599 			if ((err = zfs_secpolicy_write_perms(dsname, perm, cr)))
4600 				return (err);
4601 			return (0);
4602 		}
4603 
4604 		return (SET_ERROR(EINVAL));
4605 	}
4606 
4607 	if (issnap)
4608 		return (SET_ERROR(EINVAL));
4609 
4610 	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
4611 		/*
4612 		 * dsl_prop_get_all_impl() returns properties in this
4613 		 * format.
4614 		 */
4615 		nvlist_t *attrs;
4616 		VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
4617 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4618 		    &pair) == 0);
4619 	}
4620 
4621 	/*
4622 	 * Check that this value is valid for this pool version
4623 	 */
4624 	switch (prop) {
4625 	case ZFS_PROP_COMPRESSION:
4626 		/*
4627 		 * If the user specified gzip compression, make sure
4628 		 * the SPA supports it. We ignore any errors here since
4629 		 * we'll catch them later.
4630 		 */
4631 		if (nvpair_value_uint64(pair, &intval) == 0) {
4632 			compval = ZIO_COMPRESS_ALGO(intval);
4633 			if (compval >= ZIO_COMPRESS_GZIP_1 &&
4634 			    compval <= ZIO_COMPRESS_GZIP_9 &&
4635 			    zfs_earlier_version(dsname,
4636 			    SPA_VERSION_GZIP_COMPRESSION)) {
4637 				return (SET_ERROR(ENOTSUP));
4638 			}
4639 
4640 			if (compval == ZIO_COMPRESS_ZLE &&
4641 			    zfs_earlier_version(dsname,
4642 			    SPA_VERSION_ZLE_COMPRESSION))
4643 				return (SET_ERROR(ENOTSUP));
4644 
4645 			if (compval == ZIO_COMPRESS_LZ4) {
4646 				spa_t *spa;
4647 
4648 				if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4649 					return (err);
4650 
4651 				if (!spa_feature_is_enabled(spa,
4652 				    SPA_FEATURE_LZ4_COMPRESS)) {
4653 					spa_close(spa, FTAG);
4654 					return (SET_ERROR(ENOTSUP));
4655 				}
4656 				spa_close(spa, FTAG);
4657 			}
4658 
4659 			if (compval == ZIO_COMPRESS_ZSTD) {
4660 				spa_t *spa;
4661 
4662 				if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4663 					return (err);
4664 
4665 				if (!spa_feature_is_enabled(spa,
4666 				    SPA_FEATURE_ZSTD_COMPRESS)) {
4667 					spa_close(spa, FTAG);
4668 					return (SET_ERROR(ENOTSUP));
4669 				}
4670 				spa_close(spa, FTAG);
4671 			}
4672 		}
4673 		break;
4674 
4675 	case ZFS_PROP_COPIES:
4676 		if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
4677 			return (SET_ERROR(ENOTSUP));
4678 		break;
4679 
4680 	case ZFS_PROP_VOLBLOCKSIZE:
4681 	case ZFS_PROP_RECORDSIZE:
4682 		/* Record sizes above 128k need the feature to be enabled */
4683 		if (nvpair_value_uint64(pair, &intval) == 0 &&
4684 		    intval > SPA_OLD_MAXBLOCKSIZE) {
4685 			spa_t *spa;
4686 
4687 			/*
4688 			 * We don't allow setting the property above 1MB,
4689 			 * unless the tunable has been changed.
4690 			 */
4691 			if (intval > zfs_max_recordsize ||
4692 			    intval > SPA_MAXBLOCKSIZE)
4693 				return (SET_ERROR(ERANGE));
4694 
4695 			if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4696 				return (err);
4697 
4698 			if (!spa_feature_is_enabled(spa,
4699 			    SPA_FEATURE_LARGE_BLOCKS)) {
4700 				spa_close(spa, FTAG);
4701 				return (SET_ERROR(ENOTSUP));
4702 			}
4703 			spa_close(spa, FTAG);
4704 		}
4705 		break;
4706 
4707 	case ZFS_PROP_DNODESIZE:
4708 		/* Dnode sizes above 512 need the feature to be enabled */
4709 		if (nvpair_value_uint64(pair, &intval) == 0 &&
4710 		    intval != ZFS_DNSIZE_LEGACY) {
4711 			spa_t *spa;
4712 
4713 			if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4714 				return (err);
4715 
4716 			if (!spa_feature_is_enabled(spa,
4717 			    SPA_FEATURE_LARGE_DNODE)) {
4718 				spa_close(spa, FTAG);
4719 				return (SET_ERROR(ENOTSUP));
4720 			}
4721 			spa_close(spa, FTAG);
4722 		}
4723 		break;
4724 
4725 	case ZFS_PROP_SPECIAL_SMALL_BLOCKS:
4726 		/*
4727 		 * This property could require the allocation classes
4728 		 * feature to be active for setting, however we allow
4729 		 * it so that tests of settable properties succeed.
4730 		 * The CLI will issue a warning in this case.
4731 		 */
4732 		break;
4733 
4734 	case ZFS_PROP_SHARESMB:
4735 		if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
4736 			return (SET_ERROR(ENOTSUP));
4737 		break;
4738 
4739 	case ZFS_PROP_ACLINHERIT:
4740 		if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
4741 		    nvpair_value_uint64(pair, &intval) == 0) {
4742 			if (intval == ZFS_ACL_PASSTHROUGH_X &&
4743 			    zfs_earlier_version(dsname,
4744 			    SPA_VERSION_PASSTHROUGH_X))
4745 				return (SET_ERROR(ENOTSUP));
4746 		}
4747 		break;
4748 	case ZFS_PROP_CHECKSUM:
4749 	case ZFS_PROP_DEDUP:
4750 	{
4751 		spa_feature_t feature;
4752 		spa_t *spa;
4753 		int err;
4754 
4755 		/* dedup feature version checks */
4756 		if (prop == ZFS_PROP_DEDUP &&
4757 		    zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
4758 			return (SET_ERROR(ENOTSUP));
4759 
4760 		if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
4761 		    nvpair_value_uint64(pair, &intval) == 0) {
4762 			/* check prop value is enabled in features */
4763 			feature = zio_checksum_to_feature(
4764 			    intval & ZIO_CHECKSUM_MASK);
4765 			if (feature == SPA_FEATURE_NONE)
4766 				break;
4767 
4768 			if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4769 				return (err);
4770 
4771 			if (!spa_feature_is_enabled(spa, feature)) {
4772 				spa_close(spa, FTAG);
4773 				return (SET_ERROR(ENOTSUP));
4774 			}
4775 			spa_close(spa, FTAG);
4776 		}
4777 		break;
4778 	}
4779 
4780 	default:
4781 		break;
4782 	}
4783 
4784 	return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
4785 }
4786 
4787 /*
4788  * Removes properties from the given props list that fail permission checks
4789  * needed to clear them and to restore them in case of a receive error. For each
4790  * property, make sure we have both set and inherit permissions.
4791  *
4792  * Returns the first error encountered if any permission checks fail. If the
4793  * caller provides a non-NULL errlist, it also gives the complete list of names
4794  * of all the properties that failed a permission check along with the
4795  * corresponding error numbers. The caller is responsible for freeing the
4796  * returned errlist.
4797  *
4798  * If every property checks out successfully, zero is returned and the list
4799  * pointed at by errlist is NULL.
4800  */
4801 static int
4802 zfs_check_clearable(const char *dataset, nvlist_t *props, nvlist_t **errlist)
4803 {
4804 	zfs_cmd_t *zc;
4805 	nvpair_t *pair, *next_pair;
4806 	nvlist_t *errors;
4807 	int err, rv = 0;
4808 
4809 	if (props == NULL)
4810 		return (0);
4811 
4812 	VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4813 
4814 	zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
4815 	(void) strlcpy(zc->zc_name, dataset, sizeof (zc->zc_name));
4816 	pair = nvlist_next_nvpair(props, NULL);
4817 	while (pair != NULL) {
4818 		next_pair = nvlist_next_nvpair(props, pair);
4819 
4820 		(void) strlcpy(zc->zc_value, nvpair_name(pair),
4821 		    sizeof (zc->zc_value));
4822 		if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
4823 		    (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
4824 			VERIFY(nvlist_remove_nvpair(props, pair) == 0);
4825 			VERIFY(nvlist_add_int32(errors,
4826 			    zc->zc_value, err) == 0);
4827 		}
4828 		pair = next_pair;
4829 	}
4830 	kmem_free(zc, sizeof (zfs_cmd_t));
4831 
4832 	if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
4833 		nvlist_free(errors);
4834 		errors = NULL;
4835 	} else {
4836 		VERIFY(nvpair_value_int32(pair, &rv) == 0);
4837 	}
4838 
4839 	if (errlist == NULL)
4840 		nvlist_free(errors);
4841 	else
4842 		*errlist = errors;
4843 
4844 	return (rv);
4845 }
4846 
4847 static boolean_t
4848 propval_equals(nvpair_t *p1, nvpair_t *p2)
4849 {
4850 	if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
4851 		/* dsl_prop_get_all_impl() format */
4852 		nvlist_t *attrs;
4853 		VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
4854 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4855 		    &p1) == 0);
4856 	}
4857 
4858 	if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
4859 		nvlist_t *attrs;
4860 		VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
4861 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4862 		    &p2) == 0);
4863 	}
4864 
4865 	if (nvpair_type(p1) != nvpair_type(p2))
4866 		return (B_FALSE);
4867 
4868 	if (nvpair_type(p1) == DATA_TYPE_STRING) {
4869 		const char *valstr1, *valstr2;
4870 
4871 		VERIFY(nvpair_value_string(p1, &valstr1) == 0);
4872 		VERIFY(nvpair_value_string(p2, &valstr2) == 0);
4873 		return (strcmp(valstr1, valstr2) == 0);
4874 	} else {
4875 		uint64_t intval1, intval2;
4876 
4877 		VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4878 		VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4879 		return (intval1 == intval2);
4880 	}
4881 }
4882 
4883 /*
4884  * Remove properties from props if they are not going to change (as determined
4885  * by comparison with origprops). Remove them from origprops as well, since we
4886  * do not need to clear or restore properties that won't change.
4887  */
4888 static void
4889 props_reduce(nvlist_t *props, nvlist_t *origprops)
4890 {
4891 	nvpair_t *pair, *next_pair;
4892 
4893 	if (origprops == NULL)
4894 		return; /* all props need to be received */
4895 
4896 	pair = nvlist_next_nvpair(props, NULL);
4897 	while (pair != NULL) {
4898 		const char *propname = nvpair_name(pair);
4899 		nvpair_t *match;
4900 
4901 		next_pair = nvlist_next_nvpair(props, pair);
4902 
4903 		if ((nvlist_lookup_nvpair(origprops, propname,
4904 		    &match) != 0) || !propval_equals(pair, match))
4905 			goto next; /* need to set received value */
4906 
4907 		/* don't clear the existing received value */
4908 		(void) nvlist_remove_nvpair(origprops, match);
4909 		/* don't bother receiving the property */
4910 		(void) nvlist_remove_nvpair(props, pair);
4911 next:
4912 		pair = next_pair;
4913 	}
4914 }
4915 
4916 /*
4917  * Extract properties that cannot be set PRIOR to the receipt of a dataset.
4918  * For example, refquota cannot be set until after the receipt of a dataset,
4919  * because in replication streams, an older/earlier snapshot may exceed the
4920  * refquota.  We want to receive the older/earlier snapshot, but setting
4921  * refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent
4922  * the older/earlier snapshot from being received (with EDQUOT).
4923  *
4924  * The ZFS test "zfs_receive_011_pos" demonstrates such a scenario.
4925  *
4926  * libzfs will need to be judicious handling errors encountered by props
4927  * extracted by this function.
4928  */
4929 static nvlist_t *
4930 extract_delay_props(nvlist_t *props)
4931 {
4932 	nvlist_t *delayprops;
4933 	nvpair_t *nvp, *tmp;
4934 	static const zfs_prop_t delayable[] = {
4935 		ZFS_PROP_REFQUOTA,
4936 		ZFS_PROP_KEYLOCATION,
4937 		/*
4938 		 * Setting ZFS_PROP_SHARESMB requires the objset type to be
4939 		 * known, which is not possible prior to receipt of raw sends.
4940 		 */
4941 		ZFS_PROP_SHARESMB,
4942 		0
4943 	};
4944 	int i;
4945 
4946 	VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4947 
4948 	for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL;
4949 	    nvp = nvlist_next_nvpair(props, nvp)) {
4950 		/*
4951 		 * strcmp() is safe because zfs_prop_to_name() always returns
4952 		 * a bounded string.
4953 		 */
4954 		for (i = 0; delayable[i] != 0; i++) {
4955 			if (strcmp(zfs_prop_to_name(delayable[i]),
4956 			    nvpair_name(nvp)) == 0) {
4957 				break;
4958 			}
4959 		}
4960 		if (delayable[i] != 0) {
4961 			tmp = nvlist_prev_nvpair(props, nvp);
4962 			VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0);
4963 			VERIFY(nvlist_remove_nvpair(props, nvp) == 0);
4964 			nvp = tmp;
4965 		}
4966 	}
4967 
4968 	if (nvlist_empty(delayprops)) {
4969 		nvlist_free(delayprops);
4970 		delayprops = NULL;
4971 	}
4972 	return (delayprops);
4973 }
4974 
4975 static void
4976 zfs_allow_log_destroy(void *arg)
4977 {
4978 	char *poolname = arg;
4979 
4980 	if (poolname != NULL)
4981 		kmem_strfree(poolname);
4982 }
4983 
4984 #ifdef	ZFS_DEBUG
4985 static boolean_t zfs_ioc_recv_inject_err;
4986 #endif
4987 
4988 /*
4989  * nvlist 'errors' is always allocated. It will contain descriptions of
4990  * encountered errors, if any. It's the callers responsibility to free.
4991  */
4992 static int
4993 zfs_ioc_recv_impl(char *tofs, char *tosnap, const char *origin,
4994     nvlist_t *recvprops, nvlist_t *localprops, nvlist_t *hidden_args,
4995     boolean_t force, boolean_t heal, boolean_t resumable, int input_fd,
4996     dmu_replay_record_t *begin_record, uint64_t *read_bytes,
4997     uint64_t *errflags, nvlist_t **errors)
4998 {
4999 	dmu_recv_cookie_t drc;
5000 	int error = 0;
5001 	int props_error = 0;
5002 	offset_t off, noff;
5003 	nvlist_t *local_delayprops = NULL;
5004 	nvlist_t *recv_delayprops = NULL;
5005 	nvlist_t *inherited_delayprops = NULL;
5006 	nvlist_t *origprops = NULL; /* existing properties */
5007 	nvlist_t *origrecvd = NULL; /* existing received properties */
5008 	boolean_t first_recvd_props = B_FALSE;
5009 	boolean_t tofs_was_redacted;
5010 	zfs_file_t *input_fp;
5011 
5012 	*read_bytes = 0;
5013 	*errflags = 0;
5014 	*errors = fnvlist_alloc();
5015 	off = 0;
5016 
5017 	if ((input_fp = zfs_file_get(input_fd)) == NULL)
5018 		return (SET_ERROR(EBADF));
5019 
5020 	noff = off = zfs_file_off(input_fp);
5021 	error = dmu_recv_begin(tofs, tosnap, begin_record, force, heal,
5022 	    resumable, localprops, hidden_args, origin, &drc, input_fp,
5023 	    &off);
5024 	if (error != 0)
5025 		goto out;
5026 	tofs_was_redacted = dsl_get_redacted(drc.drc_ds);
5027 
5028 	/*
5029 	 * Set properties before we receive the stream so that they are applied
5030 	 * to the new data. Note that we must call dmu_recv_stream() if
5031 	 * dmu_recv_begin() succeeds.
5032 	 */
5033 	if (recvprops != NULL && !drc.drc_newfs) {
5034 		if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
5035 		    SPA_VERSION_RECVD_PROPS &&
5036 		    !dsl_prop_get_hasrecvd(tofs))
5037 			first_recvd_props = B_TRUE;
5038 
5039 		/*
5040 		 * If new received properties are supplied, they are to
5041 		 * completely replace the existing received properties,
5042 		 * so stash away the existing ones.
5043 		 */
5044 		if (dsl_prop_get_received(tofs, &origrecvd) == 0) {
5045 			nvlist_t *errlist = NULL;
5046 			/*
5047 			 * Don't bother writing a property if its value won't
5048 			 * change (and avoid the unnecessary security checks).
5049 			 *
5050 			 * The first receive after SPA_VERSION_RECVD_PROPS is a
5051 			 * special case where we blow away all local properties
5052 			 * regardless.
5053 			 */
5054 			if (!first_recvd_props)
5055 				props_reduce(recvprops, origrecvd);
5056 			if (zfs_check_clearable(tofs, origrecvd, &errlist) != 0)
5057 				(void) nvlist_merge(*errors, errlist, 0);
5058 			nvlist_free(errlist);
5059 
5060 			if (clear_received_props(tofs, origrecvd,
5061 			    first_recvd_props ? NULL : recvprops) != 0)
5062 				*errflags |= ZPROP_ERR_NOCLEAR;
5063 		} else {
5064 			*errflags |= ZPROP_ERR_NOCLEAR;
5065 		}
5066 	}
5067 
5068 	/*
5069 	 * Stash away existing properties so we can restore them on error unless
5070 	 * we're doing the first receive after SPA_VERSION_RECVD_PROPS, in which
5071 	 * case "origrecvd" will take care of that.
5072 	 */
5073 	if (localprops != NULL && !drc.drc_newfs && !first_recvd_props) {
5074 		objset_t *os;
5075 		if (dmu_objset_hold(tofs, FTAG, &os) == 0) {
5076 			if (dsl_prop_get_all(os, &origprops) != 0) {
5077 				*errflags |= ZPROP_ERR_NOCLEAR;
5078 			}
5079 			dmu_objset_rele(os, FTAG);
5080 		} else {
5081 			*errflags |= ZPROP_ERR_NOCLEAR;
5082 		}
5083 	}
5084 
5085 	if (recvprops != NULL) {
5086 		props_error = dsl_prop_set_hasrecvd(tofs);
5087 
5088 		if (props_error == 0) {
5089 			recv_delayprops = extract_delay_props(recvprops);
5090 			(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
5091 			    recvprops, *errors);
5092 		}
5093 	}
5094 
5095 	if (localprops != NULL) {
5096 		nvlist_t *oprops = fnvlist_alloc();
5097 		nvlist_t *xprops = fnvlist_alloc();
5098 		nvpair_t *nvp = NULL;
5099 
5100 		while ((nvp = nvlist_next_nvpair(localprops, nvp)) != NULL) {
5101 			if (nvpair_type(nvp) == DATA_TYPE_BOOLEAN) {
5102 				/* -x property */
5103 				const char *name = nvpair_name(nvp);
5104 				zfs_prop_t prop = zfs_name_to_prop(name);
5105 				if (prop != ZPROP_USERPROP) {
5106 					if (!zfs_prop_inheritable(prop))
5107 						continue;
5108 				} else if (!zfs_prop_user(name))
5109 					continue;
5110 				fnvlist_add_boolean(xprops, name);
5111 			} else {
5112 				/* -o property=value */
5113 				fnvlist_add_nvpair(oprops, nvp);
5114 			}
5115 		}
5116 
5117 		local_delayprops = extract_delay_props(oprops);
5118 		(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL,
5119 		    oprops, *errors);
5120 		inherited_delayprops = extract_delay_props(xprops);
5121 		(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_INHERITED,
5122 		    xprops, *errors);
5123 
5124 		nvlist_free(oprops);
5125 		nvlist_free(xprops);
5126 	}
5127 
5128 	error = dmu_recv_stream(&drc, &off);
5129 
5130 	if (error == 0) {
5131 		zfsvfs_t *zfsvfs = NULL;
5132 		zvol_state_handle_t *zv = NULL;
5133 
5134 		if (getzfsvfs(tofs, &zfsvfs) == 0) {
5135 			/* online recv */
5136 			dsl_dataset_t *ds;
5137 			int end_err;
5138 			boolean_t stream_is_redacted = DMU_GET_FEATUREFLAGS(
5139 			    begin_record->drr_u.drr_begin.
5140 			    drr_versioninfo) & DMU_BACKUP_FEATURE_REDACTED;
5141 
5142 			ds = dmu_objset_ds(zfsvfs->z_os);
5143 			error = zfs_suspend_fs(zfsvfs);
5144 			/*
5145 			 * If the suspend fails, then the recv_end will
5146 			 * likely also fail, and clean up after itself.
5147 			 */
5148 			end_err = dmu_recv_end(&drc, zfsvfs);
5149 			/*
5150 			 * If the dataset was not redacted, but we received a
5151 			 * redacted stream onto it, we need to unmount the
5152 			 * dataset.  Otherwise, resume the filesystem.
5153 			 */
5154 			if (error == 0 && !drc.drc_newfs &&
5155 			    stream_is_redacted && !tofs_was_redacted) {
5156 				error = zfs_end_fs(zfsvfs, ds);
5157 			} else if (error == 0) {
5158 				error = zfs_resume_fs(zfsvfs, ds);
5159 			}
5160 			error = error ? error : end_err;
5161 			zfs_vfs_rele(zfsvfs);
5162 		} else if ((zv = zvol_suspend(tofs)) != NULL) {
5163 			error = dmu_recv_end(&drc, zvol_tag(zv));
5164 			zvol_resume(zv);
5165 		} else {
5166 			error = dmu_recv_end(&drc, NULL);
5167 		}
5168 
5169 		/* Set delayed properties now, after we're done receiving. */
5170 		if (recv_delayprops != NULL && error == 0) {
5171 			(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
5172 			    recv_delayprops, *errors);
5173 		}
5174 		if (local_delayprops != NULL && error == 0) {
5175 			(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL,
5176 			    local_delayprops, *errors);
5177 		}
5178 		if (inherited_delayprops != NULL && error == 0) {
5179 			(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_INHERITED,
5180 			    inherited_delayprops, *errors);
5181 		}
5182 	}
5183 
5184 	/*
5185 	 * Merge delayed props back in with initial props, in case
5186 	 * we're DEBUG and zfs_ioc_recv_inject_err is set (which means
5187 	 * we have to make sure clear_received_props() includes
5188 	 * the delayed properties).
5189 	 *
5190 	 * Since zfs_ioc_recv_inject_err is only in DEBUG kernels,
5191 	 * using ASSERT() will be just like a VERIFY.
5192 	 */
5193 	if (recv_delayprops != NULL) {
5194 		ASSERT(nvlist_merge(recvprops, recv_delayprops, 0) == 0);
5195 		nvlist_free(recv_delayprops);
5196 	}
5197 	if (local_delayprops != NULL) {
5198 		ASSERT(nvlist_merge(localprops, local_delayprops, 0) == 0);
5199 		nvlist_free(local_delayprops);
5200 	}
5201 	if (inherited_delayprops != NULL) {
5202 		ASSERT(nvlist_merge(localprops, inherited_delayprops, 0) == 0);
5203 		nvlist_free(inherited_delayprops);
5204 	}
5205 	*read_bytes = off - noff;
5206 
5207 #ifdef	ZFS_DEBUG
5208 	if (zfs_ioc_recv_inject_err) {
5209 		zfs_ioc_recv_inject_err = B_FALSE;
5210 		error = 1;
5211 	}
5212 #endif
5213 
5214 	/*
5215 	 * On error, restore the original props.
5216 	 */
5217 	if (error != 0 && recvprops != NULL && !drc.drc_newfs) {
5218 		if (clear_received_props(tofs, recvprops, NULL) != 0) {
5219 			/*
5220 			 * We failed to clear the received properties.
5221 			 * Since we may have left a $recvd value on the
5222 			 * system, we can't clear the $hasrecvd flag.
5223 			 */
5224 			*errflags |= ZPROP_ERR_NORESTORE;
5225 		} else if (first_recvd_props) {
5226 			dsl_prop_unset_hasrecvd(tofs);
5227 		}
5228 
5229 		if (origrecvd == NULL && !drc.drc_newfs) {
5230 			/* We failed to stash the original properties. */
5231 			*errflags |= ZPROP_ERR_NORESTORE;
5232 		}
5233 
5234 		/*
5235 		 * dsl_props_set() will not convert RECEIVED to LOCAL on or
5236 		 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
5237 		 * explicitly if we're restoring local properties cleared in the
5238 		 * first new-style receive.
5239 		 */
5240 		if (origrecvd != NULL &&
5241 		    zfs_set_prop_nvlist(tofs, (first_recvd_props ?
5242 		    ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
5243 		    origrecvd, NULL) != 0) {
5244 			/*
5245 			 * We stashed the original properties but failed to
5246 			 * restore them.
5247 			 */
5248 			*errflags |= ZPROP_ERR_NORESTORE;
5249 		}
5250 	}
5251 	if (error != 0 && localprops != NULL && !drc.drc_newfs &&
5252 	    !first_recvd_props) {
5253 		nvlist_t *setprops;
5254 		nvlist_t *inheritprops;
5255 		nvpair_t *nvp;
5256 
5257 		if (origprops == NULL) {
5258 			/* We failed to stash the original properties. */
5259 			*errflags |= ZPROP_ERR_NORESTORE;
5260 			goto out;
5261 		}
5262 
5263 		/* Restore original props */
5264 		setprops = fnvlist_alloc();
5265 		inheritprops = fnvlist_alloc();
5266 		nvp = NULL;
5267 		while ((nvp = nvlist_next_nvpair(localprops, nvp)) != NULL) {
5268 			const char *name = nvpair_name(nvp);
5269 			const char *source;
5270 			nvlist_t *attrs;
5271 
5272 			if (!nvlist_exists(origprops, name)) {
5273 				/*
5274 				 * Property was not present or was explicitly
5275 				 * inherited before the receive, restore this.
5276 				 */
5277 				fnvlist_add_boolean(inheritprops, name);
5278 				continue;
5279 			}
5280 			attrs = fnvlist_lookup_nvlist(origprops, name);
5281 			source = fnvlist_lookup_string(attrs, ZPROP_SOURCE);
5282 
5283 			/* Skip received properties */
5284 			if (strcmp(source, ZPROP_SOURCE_VAL_RECVD) == 0)
5285 				continue;
5286 
5287 			if (strcmp(source, tofs) == 0) {
5288 				/* Property was locally set */
5289 				fnvlist_add_nvlist(setprops, name, attrs);
5290 			} else {
5291 				/* Property was implicitly inherited */
5292 				fnvlist_add_boolean(inheritprops, name);
5293 			}
5294 		}
5295 
5296 		if (zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL, setprops,
5297 		    NULL) != 0)
5298 			*errflags |= ZPROP_ERR_NORESTORE;
5299 		if (zfs_set_prop_nvlist(tofs, ZPROP_SRC_INHERITED, inheritprops,
5300 		    NULL) != 0)
5301 			*errflags |= ZPROP_ERR_NORESTORE;
5302 
5303 		nvlist_free(setprops);
5304 		nvlist_free(inheritprops);
5305 	}
5306 out:
5307 	zfs_file_put(input_fp);
5308 	nvlist_free(origrecvd);
5309 	nvlist_free(origprops);
5310 
5311 	if (error == 0)
5312 		error = props_error;
5313 
5314 	return (error);
5315 }
5316 
5317 /*
5318  * inputs:
5319  * zc_name		name of containing filesystem (unused)
5320  * zc_nvlist_src{_size}	nvlist of properties to apply
5321  * zc_nvlist_conf{_size}	nvlist of properties to exclude
5322  *			(DATA_TYPE_BOOLEAN) and override (everything else)
5323  * zc_value		name of snapshot to create
5324  * zc_string		name of clone origin (if DRR_FLAG_CLONE)
5325  * zc_cookie		file descriptor to recv from
5326  * zc_begin_record	the BEGIN record of the stream (not byteswapped)
5327  * zc_guid		force flag
5328  *
5329  * outputs:
5330  * zc_cookie		number of bytes read
5331  * zc_obj		zprop_errflags_t
5332  * zc_nvlist_dst{_size} error for each unapplied received property
5333  */
5334 static int
5335 zfs_ioc_recv(zfs_cmd_t *zc)
5336 {
5337 	dmu_replay_record_t begin_record;
5338 	nvlist_t *errors = NULL;
5339 	nvlist_t *recvdprops = NULL;
5340 	nvlist_t *localprops = NULL;
5341 	const char *origin = NULL;
5342 	char *tosnap;
5343 	char tofs[ZFS_MAX_DATASET_NAME_LEN];
5344 	int error = 0;
5345 
5346 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
5347 	    strchr(zc->zc_value, '@') == NULL ||
5348 	    strchr(zc->zc_value, '%'))
5349 		return (SET_ERROR(EINVAL));
5350 
5351 	(void) strlcpy(tofs, zc->zc_value, sizeof (tofs));
5352 	tosnap = strchr(tofs, '@');
5353 	*tosnap++ = '\0';
5354 
5355 	if (zc->zc_nvlist_src != 0 &&
5356 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
5357 	    zc->zc_iflags, &recvdprops)) != 0)
5358 		return (error);
5359 
5360 	if (zc->zc_nvlist_conf != 0 &&
5361 	    (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
5362 	    zc->zc_iflags, &localprops)) != 0)
5363 		return (error);
5364 
5365 	if (zc->zc_string[0])
5366 		origin = zc->zc_string;
5367 
5368 	begin_record.drr_type = DRR_BEGIN;
5369 	begin_record.drr_payloadlen = 0;
5370 	begin_record.drr_u.drr_begin = zc->zc_begin_record;
5371 
5372 	error = zfs_ioc_recv_impl(tofs, tosnap, origin, recvdprops, localprops,
5373 	    NULL, zc->zc_guid, B_FALSE, B_FALSE, zc->zc_cookie, &begin_record,
5374 	    &zc->zc_cookie, &zc->zc_obj, &errors);
5375 	nvlist_free(recvdprops);
5376 	nvlist_free(localprops);
5377 
5378 	/*
5379 	 * Now that all props, initial and delayed, are set, report the prop
5380 	 * errors to the caller.
5381 	 */
5382 	if (zc->zc_nvlist_dst_size != 0 && errors != NULL &&
5383 	    (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
5384 	    put_nvlist(zc, errors) != 0)) {
5385 		/*
5386 		 * Caller made zc->zc_nvlist_dst less than the minimum expected
5387 		 * size or supplied an invalid address.
5388 		 */
5389 		error = SET_ERROR(EINVAL);
5390 	}
5391 
5392 	nvlist_free(errors);
5393 
5394 	return (error);
5395 }
5396 
5397 /*
5398  * innvl: {
5399  *     "snapname" -> full name of the snapshot to create
5400  *     (optional) "props" -> received properties to set (nvlist)
5401  *     (optional) "localprops" -> override and exclude properties (nvlist)
5402  *     (optional) "origin" -> name of clone origin (DRR_FLAG_CLONE)
5403  *     "begin_record" -> non-byteswapped dmu_replay_record_t
5404  *     "input_fd" -> file descriptor to read stream from (int32)
5405  *     (optional) "force" -> force flag (value ignored)
5406  *     (optional) "heal" -> use send stream to heal data corruption
5407  *     (optional) "resumable" -> resumable flag (value ignored)
5408  *     (optional) "cleanup_fd" -> unused
5409  *     (optional) "action_handle" -> unused
5410  *     (optional) "hidden_args" -> { "wkeydata" -> value }
5411  * }
5412  *
5413  * outnvl: {
5414  *     "read_bytes" -> number of bytes read
5415  *     "error_flags" -> zprop_errflags_t
5416  *     "errors" -> error for each unapplied received property (nvlist)
5417  * }
5418  */
5419 static const zfs_ioc_key_t zfs_keys_recv_new[] = {
5420 	{"snapname",		DATA_TYPE_STRING,	0},
5421 	{"props",		DATA_TYPE_NVLIST,	ZK_OPTIONAL},
5422 	{"localprops",		DATA_TYPE_NVLIST,	ZK_OPTIONAL},
5423 	{"origin",		DATA_TYPE_STRING,	ZK_OPTIONAL},
5424 	{"begin_record",	DATA_TYPE_BYTE_ARRAY,	0},
5425 	{"input_fd",		DATA_TYPE_INT32,	0},
5426 	{"force",		DATA_TYPE_BOOLEAN,	ZK_OPTIONAL},
5427 	{"heal",		DATA_TYPE_BOOLEAN,	ZK_OPTIONAL},
5428 	{"resumable",		DATA_TYPE_BOOLEAN,	ZK_OPTIONAL},
5429 	{"cleanup_fd",		DATA_TYPE_INT32,	ZK_OPTIONAL},
5430 	{"action_handle",	DATA_TYPE_UINT64,	ZK_OPTIONAL},
5431 	{"hidden_args",		DATA_TYPE_NVLIST,	ZK_OPTIONAL},
5432 };
5433 
5434 static int
5435 zfs_ioc_recv_new(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
5436 {
5437 	dmu_replay_record_t *begin_record;
5438 	uint_t begin_record_size;
5439 	nvlist_t *errors = NULL;
5440 	nvlist_t *recvprops = NULL;
5441 	nvlist_t *localprops = NULL;
5442 	nvlist_t *hidden_args = NULL;
5443 	const char *snapname;
5444 	const char *origin = NULL;
5445 	char *tosnap;
5446 	char tofs[ZFS_MAX_DATASET_NAME_LEN];
5447 	boolean_t force;
5448 	boolean_t heal;
5449 	boolean_t resumable;
5450 	uint64_t read_bytes = 0;
5451 	uint64_t errflags = 0;
5452 	int input_fd = -1;
5453 	int error;
5454 
5455 	snapname = fnvlist_lookup_string(innvl, "snapname");
5456 
5457 	if (dataset_namecheck(snapname, NULL, NULL) != 0 ||
5458 	    strchr(snapname, '@') == NULL ||
5459 	    strchr(snapname, '%'))
5460 		return (SET_ERROR(EINVAL));
5461 
5462 	(void) strlcpy(tofs, snapname, sizeof (tofs));
5463 	tosnap = strchr(tofs, '@');
5464 	*tosnap++ = '\0';
5465 
5466 	error = nvlist_lookup_string(innvl, "origin", &origin);
5467 	if (error && error != ENOENT)
5468 		return (error);
5469 
5470 	error = nvlist_lookup_byte_array(innvl, "begin_record",
5471 	    (uchar_t **)&begin_record, &begin_record_size);
5472 	if (error != 0 || begin_record_size != sizeof (*begin_record))
5473 		return (SET_ERROR(EINVAL));
5474 
5475 	input_fd = fnvlist_lookup_int32(innvl, "input_fd");
5476 
5477 	force = nvlist_exists(innvl, "force");
5478 	heal = nvlist_exists(innvl, "heal");
5479 	resumable = nvlist_exists(innvl, "resumable");
5480 
5481 	/* we still use "props" here for backwards compatibility */
5482 	error = nvlist_lookup_nvlist(innvl, "props", &recvprops);
5483 	if (error && error != ENOENT)
5484 		return (error);
5485 
5486 	error = nvlist_lookup_nvlist(innvl, "localprops", &localprops);
5487 	if (error && error != ENOENT)
5488 		return (error);
5489 
5490 	error = nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);
5491 	if (error && error != ENOENT)
5492 		return (error);
5493 
5494 	error = zfs_ioc_recv_impl(tofs, tosnap, origin, recvprops, localprops,
5495 	    hidden_args, force, heal, resumable, input_fd, begin_record,
5496 	    &read_bytes, &errflags, &errors);
5497 
5498 	fnvlist_add_uint64(outnvl, "read_bytes", read_bytes);
5499 	fnvlist_add_uint64(outnvl, "error_flags", errflags);
5500 	fnvlist_add_nvlist(outnvl, "errors", errors);
5501 
5502 	nvlist_free(errors);
5503 	nvlist_free(recvprops);
5504 	nvlist_free(localprops);
5505 
5506 	return (error);
5507 }
5508 
5509 typedef struct dump_bytes_io {
5510 	zfs_file_t	*dbi_fp;
5511 	caddr_t		dbi_buf;
5512 	int		dbi_len;
5513 	int		dbi_err;
5514 } dump_bytes_io_t;
5515 
5516 static void
5517 dump_bytes_cb(void *arg)
5518 {
5519 	dump_bytes_io_t *dbi = (dump_bytes_io_t *)arg;
5520 	zfs_file_t *fp;
5521 	caddr_t buf;
5522 
5523 	fp = dbi->dbi_fp;
5524 	buf = dbi->dbi_buf;
5525 
5526 	dbi->dbi_err = zfs_file_write(fp, buf, dbi->dbi_len, NULL);
5527 }
5528 
5529 static int
5530 dump_bytes(objset_t *os, void *buf, int len, void *arg)
5531 {
5532 	dump_bytes_io_t dbi;
5533 
5534 	dbi.dbi_fp = arg;
5535 	dbi.dbi_buf = buf;
5536 	dbi.dbi_len = len;
5537 
5538 #if defined(HAVE_LARGE_STACKS)
5539 	dump_bytes_cb(&dbi);
5540 #else
5541 	/*
5542 	 * The vn_rdwr() call is performed in a taskq to ensure that there is
5543 	 * always enough stack space to write safely to the target filesystem.
5544 	 * The ZIO_TYPE_FREE threads are used because there can be a lot of
5545 	 * them and they are used in vdev_file.c for a similar purpose.
5546 	 */
5547 	spa_taskq_dispatch_sync(dmu_objset_spa(os), ZIO_TYPE_FREE,
5548 	    ZIO_TASKQ_ISSUE, dump_bytes_cb, &dbi, TQ_SLEEP);
5549 #endif /* HAVE_LARGE_STACKS */
5550 
5551 	return (dbi.dbi_err);
5552 }
5553 
5554 /*
5555  * inputs:
5556  * zc_name	name of snapshot to send
5557  * zc_cookie	file descriptor to send stream to
5558  * zc_obj	fromorigin flag (mutually exclusive with zc_fromobj)
5559  * zc_sendobj	objsetid of snapshot to send
5560  * zc_fromobj	objsetid of incremental fromsnap (may be zero)
5561  * zc_guid	if set, estimate size of stream only.  zc_cookie is ignored.
5562  *		output size in zc_objset_type.
5563  * zc_flags	lzc_send_flags
5564  *
5565  * outputs:
5566  * zc_objset_type	estimated size, if zc_guid is set
5567  *
5568  * NOTE: This is no longer the preferred interface, any new functionality
5569  *	  should be added to zfs_ioc_send_new() instead.
5570  */
5571 static int
5572 zfs_ioc_send(zfs_cmd_t *zc)
5573 {
5574 	int error;
5575 	offset_t off;
5576 	boolean_t estimate = (zc->zc_guid != 0);
5577 	boolean_t embedok = (zc->zc_flags & 0x1);
5578 	boolean_t large_block_ok = (zc->zc_flags & 0x2);
5579 	boolean_t compressok = (zc->zc_flags & 0x4);
5580 	boolean_t rawok = (zc->zc_flags & 0x8);
5581 	boolean_t savedok = (zc->zc_flags & 0x10);
5582 
5583 	if (zc->zc_obj != 0) {
5584 		dsl_pool_t *dp;
5585 		dsl_dataset_t *tosnap;
5586 
5587 		error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5588 		if (error != 0)
5589 			return (error);
5590 
5591 		error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
5592 		if (error != 0) {
5593 			dsl_pool_rele(dp, FTAG);
5594 			return (error);
5595 		}
5596 
5597 		if (dsl_dir_is_clone(tosnap->ds_dir))
5598 			zc->zc_fromobj =
5599 			    dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj;
5600 		dsl_dataset_rele(tosnap, FTAG);
5601 		dsl_pool_rele(dp, FTAG);
5602 	}
5603 
5604 	if (estimate) {
5605 		dsl_pool_t *dp;
5606 		dsl_dataset_t *tosnap;
5607 		dsl_dataset_t *fromsnap = NULL;
5608 
5609 		error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5610 		if (error != 0)
5611 			return (error);
5612 
5613 		error = dsl_dataset_hold_obj(dp, zc->zc_sendobj,
5614 		    FTAG, &tosnap);
5615 		if (error != 0) {
5616 			dsl_pool_rele(dp, FTAG);
5617 			return (error);
5618 		}
5619 
5620 		if (zc->zc_fromobj != 0) {
5621 			error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
5622 			    FTAG, &fromsnap);
5623 			if (error != 0) {
5624 				dsl_dataset_rele(tosnap, FTAG);
5625 				dsl_pool_rele(dp, FTAG);
5626 				return (error);
5627 			}
5628 		}
5629 
5630 		error = dmu_send_estimate_fast(tosnap, fromsnap, NULL,
5631 		    compressok || rawok, savedok, &zc->zc_objset_type);
5632 
5633 		if (fromsnap != NULL)
5634 			dsl_dataset_rele(fromsnap, FTAG);
5635 		dsl_dataset_rele(tosnap, FTAG);
5636 		dsl_pool_rele(dp, FTAG);
5637 	} else {
5638 		zfs_file_t *fp;
5639 		dmu_send_outparams_t out = {0};
5640 
5641 		if ((fp = zfs_file_get(zc->zc_cookie)) == NULL)
5642 			return (SET_ERROR(EBADF));
5643 
5644 		off = zfs_file_off(fp);
5645 		out.dso_outfunc = dump_bytes;
5646 		out.dso_arg = fp;
5647 		out.dso_dryrun = B_FALSE;
5648 		error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
5649 		    zc->zc_fromobj, embedok, large_block_ok, compressok,
5650 		    rawok, savedok, zc->zc_cookie, &off, &out);
5651 
5652 		zfs_file_put(fp);
5653 	}
5654 	return (error);
5655 }
5656 
5657 /*
5658  * inputs:
5659  * zc_name		name of snapshot on which to report progress
5660  * zc_cookie		file descriptor of send stream
5661  *
5662  * outputs:
5663  * zc_cookie		number of bytes written in send stream thus far
5664  * zc_objset_type	logical size of data traversed by send thus far
5665  */
5666 static int
5667 zfs_ioc_send_progress(zfs_cmd_t *zc)
5668 {
5669 	dsl_pool_t *dp;
5670 	dsl_dataset_t *ds;
5671 	dmu_sendstatus_t *dsp = NULL;
5672 	int error;
5673 
5674 	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5675 	if (error != 0)
5676 		return (error);
5677 
5678 	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
5679 	if (error != 0) {
5680 		dsl_pool_rele(dp, FTAG);
5681 		return (error);
5682 	}
5683 
5684 	mutex_enter(&ds->ds_sendstream_lock);
5685 
5686 	/*
5687 	 * Iterate over all the send streams currently active on this dataset.
5688 	 * If there's one which matches the specified file descriptor _and_ the
5689 	 * stream was started by the current process, return the progress of
5690 	 * that stream.
5691 	 */
5692 
5693 	for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
5694 	    dsp = list_next(&ds->ds_sendstreams, dsp)) {
5695 		if (dsp->dss_outfd == zc->zc_cookie &&
5696 		    zfs_proc_is_caller(dsp->dss_proc))
5697 			break;
5698 	}
5699 
5700 	if (dsp != NULL) {
5701 		zc->zc_cookie = atomic_cas_64((volatile uint64_t *)dsp->dss_off,
5702 		    0, 0);
5703 		/* This is the closest thing we have to atomic_read_64. */
5704 		zc->zc_objset_type = atomic_cas_64(&dsp->dss_blocks, 0, 0);
5705 	} else {
5706 		error = SET_ERROR(ENOENT);
5707 	}
5708 
5709 	mutex_exit(&ds->ds_sendstream_lock);
5710 	dsl_dataset_rele(ds, FTAG);
5711 	dsl_pool_rele(dp, FTAG);
5712 	return (error);
5713 }
5714 
5715 static int
5716 zfs_ioc_inject_fault(zfs_cmd_t *zc)
5717 {
5718 	int id, error;
5719 
5720 	error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
5721 	    &zc->zc_inject_record);
5722 
5723 	if (error == 0)
5724 		zc->zc_guid = (uint64_t)id;
5725 
5726 	return (error);
5727 }
5728 
5729 static int
5730 zfs_ioc_clear_fault(zfs_cmd_t *zc)
5731 {
5732 	return (zio_clear_fault((int)zc->zc_guid));
5733 }
5734 
5735 static int
5736 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
5737 {
5738 	int id = (int)zc->zc_guid;
5739 	int error;
5740 
5741 	error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
5742 	    &zc->zc_inject_record);
5743 
5744 	zc->zc_guid = id;
5745 
5746 	return (error);
5747 }
5748 
5749 static int
5750 zfs_ioc_error_log(zfs_cmd_t *zc)
5751 {
5752 	spa_t *spa;
5753 	int error;
5754 
5755 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
5756 		return (error);
5757 
5758 	error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
5759 	    &zc->zc_nvlist_dst_size);
5760 
5761 	spa_close(spa, FTAG);
5762 
5763 	return (error);
5764 }
5765 
5766 static int
5767 zfs_ioc_clear(zfs_cmd_t *zc)
5768 {
5769 	spa_t *spa;
5770 	vdev_t *vd;
5771 	int error;
5772 
5773 	/*
5774 	 * On zpool clear we also fix up missing slogs
5775 	 */
5776 	mutex_enter(&spa_namespace_lock);
5777 	spa = spa_lookup(zc->zc_name);
5778 	if (spa == NULL) {
5779 		mutex_exit(&spa_namespace_lock);
5780 		return (SET_ERROR(EIO));
5781 	}
5782 	if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
5783 		/* we need to let spa_open/spa_load clear the chains */
5784 		spa_set_log_state(spa, SPA_LOG_CLEAR);
5785 	}
5786 	spa->spa_last_open_failed = 0;
5787 	mutex_exit(&spa_namespace_lock);
5788 
5789 	if (zc->zc_cookie & ZPOOL_NO_REWIND) {
5790 		error = spa_open(zc->zc_name, &spa, FTAG);
5791 	} else {
5792 		nvlist_t *policy;
5793 		nvlist_t *config = NULL;
5794 
5795 		if (zc->zc_nvlist_src == 0)
5796 			return (SET_ERROR(EINVAL));
5797 
5798 		if ((error = get_nvlist(zc->zc_nvlist_src,
5799 		    zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
5800 			error = spa_open_rewind(zc->zc_name, &spa, FTAG,
5801 			    policy, &config);
5802 			if (config != NULL) {
5803 				int err;
5804 
5805 				if ((err = put_nvlist(zc, config)) != 0)
5806 					error = err;
5807 				nvlist_free(config);
5808 			}
5809 			nvlist_free(policy);
5810 		}
5811 	}
5812 
5813 	if (error != 0)
5814 		return (error);
5815 
5816 	/*
5817 	 * If multihost is enabled, resuming I/O is unsafe as another
5818 	 * host may have imported the pool.
5819 	 */
5820 	if (spa_multihost(spa) && spa_suspended(spa))
5821 		return (SET_ERROR(EINVAL));
5822 
5823 	spa_vdev_state_enter(spa, SCL_NONE);
5824 
5825 	if (zc->zc_guid == 0) {
5826 		vd = NULL;
5827 	} else {
5828 		vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
5829 		if (vd == NULL) {
5830 			error = SET_ERROR(ENODEV);
5831 			(void) spa_vdev_state_exit(spa, NULL, error);
5832 			spa_close(spa, FTAG);
5833 			return (error);
5834 		}
5835 	}
5836 
5837 	vdev_clear(spa, vd);
5838 
5839 	(void) spa_vdev_state_exit(spa, spa_suspended(spa) ?
5840 	    NULL : spa->spa_root_vdev, 0);
5841 
5842 	/*
5843 	 * Resume any suspended I/Os.
5844 	 */
5845 	if (zio_resume(spa) != 0)
5846 		error = SET_ERROR(EIO);
5847 
5848 	spa_close(spa, FTAG);
5849 
5850 	return (error);
5851 }
5852 
5853 /*
5854  * Reopen all the vdevs associated with the pool.
5855  *
5856  * innvl: {
5857  *  "scrub_restart" -> when true and scrub is running, allow to restart
5858  *              scrub as the side effect of the reopen (boolean).
5859  * }
5860  *
5861  * outnvl is unused
5862  */
5863 static const zfs_ioc_key_t zfs_keys_pool_reopen[] = {
5864 	{"scrub_restart",	DATA_TYPE_BOOLEAN_VALUE,	ZK_OPTIONAL},
5865 };
5866 
5867 static int
5868 zfs_ioc_pool_reopen(const char *pool, nvlist_t *innvl, nvlist_t *outnvl)
5869 {
5870 	(void) outnvl;
5871 	spa_t *spa;
5872 	int error;
5873 	boolean_t rc, scrub_restart = B_TRUE;
5874 
5875 	if (innvl) {
5876 		error = nvlist_lookup_boolean_value(innvl,
5877 		    "scrub_restart", &rc);
5878 		if (error == 0)
5879 			scrub_restart = rc;
5880 	}
5881 
5882 	error = spa_open(pool, &spa, FTAG);
5883 	if (error != 0)
5884 		return (error);
5885 
5886 	spa_vdev_state_enter(spa, SCL_NONE);
5887 
5888 	/*
5889 	 * If the scrub_restart flag is B_FALSE and a scrub is already
5890 	 * in progress then set spa_scrub_reopen flag to B_TRUE so that
5891 	 * we don't restart the scrub as a side effect of the reopen.
5892 	 * Otherwise, let vdev_open() decided if a resilver is required.
5893 	 */
5894 
5895 	spa->spa_scrub_reopen = (!scrub_restart &&
5896 	    dsl_scan_scrubbing(spa->spa_dsl_pool));
5897 	vdev_reopen(spa->spa_root_vdev);
5898 	spa->spa_scrub_reopen = B_FALSE;
5899 
5900 	(void) spa_vdev_state_exit(spa, NULL, 0);
5901 	spa_close(spa, FTAG);
5902 	return (0);
5903 }
5904 
5905 /*
5906  * inputs:
5907  * zc_name	name of filesystem
5908  *
5909  * outputs:
5910  * zc_string	name of conflicting snapshot, if there is one
5911  */
5912 static int
5913 zfs_ioc_promote(zfs_cmd_t *zc)
5914 {
5915 	dsl_pool_t *dp;
5916 	dsl_dataset_t *ds, *ods;
5917 	char origin[ZFS_MAX_DATASET_NAME_LEN];
5918 	char *cp;
5919 	int error;
5920 
5921 	zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
5922 	if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0 ||
5923 	    strchr(zc->zc_name, '%'))
5924 		return (SET_ERROR(EINVAL));
5925 
5926 	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5927 	if (error != 0)
5928 		return (error);
5929 
5930 	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
5931 	if (error != 0) {
5932 		dsl_pool_rele(dp, FTAG);
5933 		return (error);
5934 	}
5935 
5936 	if (!dsl_dir_is_clone(ds->ds_dir)) {
5937 		dsl_dataset_rele(ds, FTAG);
5938 		dsl_pool_rele(dp, FTAG);
5939 		return (SET_ERROR(EINVAL));
5940 	}
5941 
5942 	error = dsl_dataset_hold_obj(dp,
5943 	    dsl_dir_phys(ds->ds_dir)->dd_origin_obj, FTAG, &ods);
5944 	if (error != 0) {
5945 		dsl_dataset_rele(ds, FTAG);
5946 		dsl_pool_rele(dp, FTAG);
5947 		return (error);
5948 	}
5949 
5950 	dsl_dataset_name(ods, origin);
5951 	dsl_dataset_rele(ods, FTAG);
5952 	dsl_dataset_rele(ds, FTAG);
5953 	dsl_pool_rele(dp, FTAG);
5954 
5955 	/*
5956 	 * We don't need to unmount *all* the origin fs's snapshots, but
5957 	 * it's easier.
5958 	 */
5959 	cp = strchr(origin, '@');
5960 	if (cp)
5961 		*cp = '\0';
5962 	(void) dmu_objset_find(origin,
5963 	    zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
5964 	return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
5965 }
5966 
5967 /*
5968  * Retrieve a single {user|group|project}{used|quota}@... property.
5969  *
5970  * inputs:
5971  * zc_name	name of filesystem
5972  * zc_objset_type zfs_userquota_prop_t
5973  * zc_value	domain name (eg. "S-1-234-567-89")
5974  * zc_guid	RID/UID/GID
5975  *
5976  * outputs:
5977  * zc_cookie	property value
5978  */
5979 static int
5980 zfs_ioc_userspace_one(zfs_cmd_t *zc)
5981 {
5982 	zfsvfs_t *zfsvfs;
5983 	int error;
5984 
5985 	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
5986 		return (SET_ERROR(EINVAL));
5987 
5988 	error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
5989 	if (error != 0)
5990 		return (error);
5991 
5992 	error = zfs_userspace_one(zfsvfs,
5993 	    zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
5994 	zfsvfs_rele(zfsvfs, FTAG);
5995 
5996 	return (error);
5997 }
5998 
5999 /*
6000  * inputs:
6001  * zc_name		name of filesystem
6002  * zc_cookie		zap cursor
6003  * zc_objset_type	zfs_userquota_prop_t
6004  * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
6005  *
6006  * outputs:
6007  * zc_nvlist_dst[_size]	data buffer (array of zfs_useracct_t)
6008  * zc_cookie	zap cursor
6009  */
6010 static int
6011 zfs_ioc_userspace_many(zfs_cmd_t *zc)
6012 {
6013 	zfsvfs_t *zfsvfs;
6014 	int bufsize = zc->zc_nvlist_dst_size;
6015 
6016 	if (bufsize <= 0)
6017 		return (SET_ERROR(ENOMEM));
6018 
6019 	int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
6020 	if (error != 0)
6021 		return (error);
6022 
6023 	void *buf = vmem_alloc(bufsize, KM_SLEEP);
6024 
6025 	error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
6026 	    buf, &zc->zc_nvlist_dst_size);
6027 
6028 	if (error == 0) {
6029 		error = xcopyout(buf,
6030 		    (void *)(uintptr_t)zc->zc_nvlist_dst,
6031 		    zc->zc_nvlist_dst_size);
6032 	}
6033 	vmem_free(buf, bufsize);
6034 	zfsvfs_rele(zfsvfs, FTAG);
6035 
6036 	return (error);
6037 }
6038 
6039 /*
6040  * inputs:
6041  * zc_name		name of filesystem
6042  *
6043  * outputs:
6044  * none
6045  */
6046 static int
6047 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
6048 {
6049 	int error = 0;
6050 	zfsvfs_t *zfsvfs;
6051 
6052 	if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
6053 		if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
6054 			/*
6055 			 * If userused is not enabled, it may be because the
6056 			 * objset needs to be closed & reopened (to grow the
6057 			 * objset_phys_t).  Suspend/resume the fs will do that.
6058 			 */
6059 			dsl_dataset_t *ds, *newds;
6060 
6061 			ds = dmu_objset_ds(zfsvfs->z_os);
6062 			error = zfs_suspend_fs(zfsvfs);
6063 			if (error == 0) {
6064 				dmu_objset_refresh_ownership(ds, &newds,
6065 				    B_TRUE, zfsvfs);
6066 				error = zfs_resume_fs(zfsvfs, newds);
6067 			}
6068 		}
6069 		if (error == 0) {
6070 			mutex_enter(&zfsvfs->z_os->os_upgrade_lock);
6071 			if (zfsvfs->z_os->os_upgrade_id == 0) {
6072 				/* clear potential error code and retry */
6073 				zfsvfs->z_os->os_upgrade_status = 0;
6074 				mutex_exit(&zfsvfs->z_os->os_upgrade_lock);
6075 
6076 				dsl_pool_config_enter(
6077 				    dmu_objset_pool(zfsvfs->z_os), FTAG);
6078 				dmu_objset_userspace_upgrade(zfsvfs->z_os);
6079 				dsl_pool_config_exit(
6080 				    dmu_objset_pool(zfsvfs->z_os), FTAG);
6081 			} else {
6082 				mutex_exit(&zfsvfs->z_os->os_upgrade_lock);
6083 			}
6084 
6085 			taskq_wait_id(zfsvfs->z_os->os_spa->spa_upgrade_taskq,
6086 			    zfsvfs->z_os->os_upgrade_id);
6087 			error = zfsvfs->z_os->os_upgrade_status;
6088 		}
6089 		zfs_vfs_rele(zfsvfs);
6090 	} else {
6091 		objset_t *os;
6092 
6093 		/* XXX kind of reading contents without owning */
6094 		error = dmu_objset_hold_flags(zc->zc_name, B_TRUE, FTAG, &os);
6095 		if (error != 0)
6096 			return (error);
6097 
6098 		mutex_enter(&os->os_upgrade_lock);
6099 		if (os->os_upgrade_id == 0) {
6100 			/* clear potential error code and retry */
6101 			os->os_upgrade_status = 0;
6102 			mutex_exit(&os->os_upgrade_lock);
6103 
6104 			dmu_objset_userspace_upgrade(os);
6105 		} else {
6106 			mutex_exit(&os->os_upgrade_lock);
6107 		}
6108 
6109 		dsl_pool_rele(dmu_objset_pool(os), FTAG);
6110 
6111 		taskq_wait_id(os->os_spa->spa_upgrade_taskq, os->os_upgrade_id);
6112 		error = os->os_upgrade_status;
6113 
6114 		dsl_dataset_rele_flags(dmu_objset_ds(os), DS_HOLD_FLAG_DECRYPT,
6115 		    FTAG);
6116 	}
6117 	return (error);
6118 }
6119 
6120 /*
6121  * inputs:
6122  * zc_name		name of filesystem
6123  *
6124  * outputs:
6125  * none
6126  */
6127 static int
6128 zfs_ioc_id_quota_upgrade(zfs_cmd_t *zc)
6129 {
6130 	objset_t *os;
6131 	int error;
6132 
6133 	error = dmu_objset_hold_flags(zc->zc_name, B_TRUE, FTAG, &os);
6134 	if (error != 0)
6135 		return (error);
6136 
6137 	if (dmu_objset_userobjspace_upgradable(os) ||
6138 	    dmu_objset_projectquota_upgradable(os)) {
6139 		mutex_enter(&os->os_upgrade_lock);
6140 		if (os->os_upgrade_id == 0) {
6141 			/* clear potential error code and retry */
6142 			os->os_upgrade_status = 0;
6143 			mutex_exit(&os->os_upgrade_lock);
6144 
6145 			dmu_objset_id_quota_upgrade(os);
6146 		} else {
6147 			mutex_exit(&os->os_upgrade_lock);
6148 		}
6149 
6150 		dsl_pool_rele(dmu_objset_pool(os), FTAG);
6151 
6152 		taskq_wait_id(os->os_spa->spa_upgrade_taskq, os->os_upgrade_id);
6153 		error = os->os_upgrade_status;
6154 	} else {
6155 		dsl_pool_rele(dmu_objset_pool(os), FTAG);
6156 	}
6157 
6158 	dsl_dataset_rele_flags(dmu_objset_ds(os), DS_HOLD_FLAG_DECRYPT, FTAG);
6159 
6160 	return (error);
6161 }
6162 
6163 static int
6164 zfs_ioc_share(zfs_cmd_t *zc)
6165 {
6166 	return (SET_ERROR(ENOSYS));
6167 }
6168 
6169 /*
6170  * inputs:
6171  * zc_name		name of containing filesystem
6172  * zc_obj		object # beyond which we want next in-use object #
6173  *
6174  * outputs:
6175  * zc_obj		next in-use object #
6176  */
6177 static int
6178 zfs_ioc_next_obj(zfs_cmd_t *zc)
6179 {
6180 	objset_t *os = NULL;
6181 	int error;
6182 
6183 	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
6184 	if (error != 0)
6185 		return (error);
6186 
6187 	error = dmu_object_next(os, &zc->zc_obj, B_FALSE, 0);
6188 
6189 	dmu_objset_rele(os, FTAG);
6190 	return (error);
6191 }
6192 
6193 /*
6194  * inputs:
6195  * zc_name		name of filesystem
6196  * zc_value		prefix name for snapshot
6197  * zc_cleanup_fd	cleanup-on-exit file descriptor for calling process
6198  *
6199  * outputs:
6200  * zc_value		short name of new snapshot
6201  */
6202 static int
6203 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
6204 {
6205 	char *snap_name;
6206 	char *hold_name;
6207 	minor_t minor;
6208 
6209 	zfs_file_t *fp = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
6210 	if (fp == NULL)
6211 		return (SET_ERROR(EBADF));
6212 
6213 	snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
6214 	    (u_longlong_t)ddi_get_lbolt64());
6215 	hold_name = kmem_asprintf("%%%s", zc->zc_value);
6216 
6217 	int error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
6218 	    hold_name);
6219 	if (error == 0)
6220 		(void) strlcpy(zc->zc_value, snap_name,
6221 		    sizeof (zc->zc_value));
6222 	kmem_strfree(snap_name);
6223 	kmem_strfree(hold_name);
6224 	zfs_onexit_fd_rele(fp);
6225 	return (error);
6226 }
6227 
6228 /*
6229  * inputs:
6230  * zc_name		name of "to" snapshot
6231  * zc_value		name of "from" snapshot
6232  * zc_cookie		file descriptor to write diff data on
6233  *
6234  * outputs:
6235  * dmu_diff_record_t's to the file descriptor
6236  */
6237 static int
6238 zfs_ioc_diff(zfs_cmd_t *zc)
6239 {
6240 	zfs_file_t *fp;
6241 	offset_t off;
6242 	int error;
6243 
6244 	if ((fp = zfs_file_get(zc->zc_cookie)) == NULL)
6245 		return (SET_ERROR(EBADF));
6246 
6247 	off = zfs_file_off(fp);
6248 	error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off);
6249 
6250 	zfs_file_put(fp);
6251 
6252 	return (error);
6253 }
6254 
6255 static int
6256 zfs_ioc_smb_acl(zfs_cmd_t *zc)
6257 {
6258 	return (SET_ERROR(ENOTSUP));
6259 }
6260 
6261 /*
6262  * innvl: {
6263  *     "holds" -> { snapname -> holdname (string), ... }
6264  *     (optional) "cleanup_fd" -> fd (int32)
6265  * }
6266  *
6267  * outnvl: {
6268  *     snapname -> error value (int32)
6269  *     ...
6270  * }
6271  */
6272 static const zfs_ioc_key_t zfs_keys_hold[] = {
6273 	{"holds",		DATA_TYPE_NVLIST,	0},
6274 	{"cleanup_fd",		DATA_TYPE_INT32,	ZK_OPTIONAL},
6275 };
6276 
6277 static int
6278 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
6279 {
6280 	(void) pool;
6281 	nvpair_t *pair;
6282 	nvlist_t *holds;
6283 	int cleanup_fd = -1;
6284 	int error;
6285 	minor_t minor = 0;
6286 	zfs_file_t *fp = NULL;
6287 
6288 	holds = fnvlist_lookup_nvlist(args, "holds");
6289 
6290 	/* make sure the user didn't pass us any invalid (empty) tags */
6291 	for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
6292 	    pair = nvlist_next_nvpair(holds, pair)) {
6293 		const char *htag;
6294 
6295 		error = nvpair_value_string(pair, &htag);
6296 		if (error != 0)
6297 			return (SET_ERROR(error));
6298 
6299 		if (strlen(htag) == 0)
6300 			return (SET_ERROR(EINVAL));
6301 	}
6302 
6303 	if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
6304 		fp = zfs_onexit_fd_hold(cleanup_fd, &minor);
6305 		if (fp == NULL)
6306 			return (SET_ERROR(EBADF));
6307 	}
6308 
6309 	error = dsl_dataset_user_hold(holds, minor, errlist);
6310 	if (fp != NULL) {
6311 		ASSERT3U(minor, !=, 0);
6312 		zfs_onexit_fd_rele(fp);
6313 	}
6314 	return (SET_ERROR(error));
6315 }
6316 
6317 /*
6318  * innvl is not used.
6319  *
6320  * outnvl: {
6321  *    holdname -> time added (uint64 seconds since epoch)
6322  *    ...
6323  * }
6324  */
6325 static const zfs_ioc_key_t zfs_keys_get_holds[] = {
6326 	/* no nvl keys */
6327 };
6328 
6329 static int
6330 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
6331 {
6332 	(void) args;
6333 	return (dsl_dataset_get_holds(snapname, outnvl));
6334 }
6335 
6336 /*
6337  * innvl: {
6338  *     snapname -> { holdname, ... }
6339  *     ...
6340  * }
6341  *
6342  * outnvl: {
6343  *     snapname -> error value (int32)
6344  *     ...
6345  * }
6346  */
6347 static const zfs_ioc_key_t zfs_keys_release[] = {
6348 	{"<snapname>...",	DATA_TYPE_NVLIST,	ZK_WILDCARDLIST},
6349 };
6350 
6351 static int
6352 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
6353 {
6354 	(void) pool;
6355 	return (dsl_dataset_user_release(holds, errlist));
6356 }
6357 
6358 /*
6359  * inputs:
6360  * zc_guid		flags (ZEVENT_NONBLOCK)
6361  * zc_cleanup_fd	zevent file descriptor
6362  *
6363  * outputs:
6364  * zc_nvlist_dst	next nvlist event
6365  * zc_cookie		dropped events since last get
6366  */
6367 static int
6368 zfs_ioc_events_next(zfs_cmd_t *zc)
6369 {
6370 	zfs_zevent_t *ze;
6371 	nvlist_t *event = NULL;
6372 	minor_t minor;
6373 	uint64_t dropped = 0;
6374 	int error;
6375 
6376 	zfs_file_t *fp = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
6377 	if (fp == NULL)
6378 		return (SET_ERROR(EBADF));
6379 
6380 	do {
6381 		error = zfs_zevent_next(ze, &event,
6382 		    &zc->zc_nvlist_dst_size, &dropped);
6383 		if (event != NULL) {
6384 			zc->zc_cookie = dropped;
6385 			error = put_nvlist(zc, event);
6386 			nvlist_free(event);
6387 		}
6388 
6389 		if (zc->zc_guid & ZEVENT_NONBLOCK)
6390 			break;
6391 
6392 		if ((error == 0) || (error != ENOENT))
6393 			break;
6394 
6395 		error = zfs_zevent_wait(ze);
6396 		if (error != 0)
6397 			break;
6398 	} while (1);
6399 
6400 	zfs_zevent_fd_rele(fp);
6401 
6402 	return (error);
6403 }
6404 
6405 /*
6406  * outputs:
6407  * zc_cookie		cleared events count
6408  */
6409 static int
6410 zfs_ioc_events_clear(zfs_cmd_t *zc)
6411 {
6412 	uint_t count;
6413 
6414 	zfs_zevent_drain_all(&count);
6415 	zc->zc_cookie = count;
6416 
6417 	return (0);
6418 }
6419 
6420 /*
6421  * inputs:
6422  * zc_guid		eid | ZEVENT_SEEK_START | ZEVENT_SEEK_END
6423  * zc_cleanup		zevent file descriptor
6424  */
6425 static int
6426 zfs_ioc_events_seek(zfs_cmd_t *zc)
6427 {
6428 	zfs_zevent_t *ze;
6429 	minor_t minor;
6430 	int error;
6431 
6432 	zfs_file_t *fp = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
6433 	if (fp == NULL)
6434 		return (SET_ERROR(EBADF));
6435 
6436 	error = zfs_zevent_seek(ze, zc->zc_guid);
6437 	zfs_zevent_fd_rele(fp);
6438 
6439 	return (error);
6440 }
6441 
6442 /*
6443  * inputs:
6444  * zc_name		name of later filesystem or snapshot
6445  * zc_value		full name of old snapshot or bookmark
6446  *
6447  * outputs:
6448  * zc_cookie		space in bytes
6449  * zc_objset_type	compressed space in bytes
6450  * zc_perm_action	uncompressed space in bytes
6451  */
6452 static int
6453 zfs_ioc_space_written(zfs_cmd_t *zc)
6454 {
6455 	int error;
6456 	dsl_pool_t *dp;
6457 	dsl_dataset_t *new;
6458 
6459 	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
6460 	if (error != 0)
6461 		return (error);
6462 	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
6463 	if (error != 0) {
6464 		dsl_pool_rele(dp, FTAG);
6465 		return (error);
6466 	}
6467 	if (strchr(zc->zc_value, '#') != NULL) {
6468 		zfs_bookmark_phys_t bmp;
6469 		error = dsl_bookmark_lookup(dp, zc->zc_value,
6470 		    new, &bmp);
6471 		if (error == 0) {
6472 			error = dsl_dataset_space_written_bookmark(&bmp, new,
6473 			    &zc->zc_cookie,
6474 			    &zc->zc_objset_type, &zc->zc_perm_action);
6475 		}
6476 	} else {
6477 		dsl_dataset_t *old;
6478 		error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
6479 
6480 		if (error == 0) {
6481 			error = dsl_dataset_space_written(old, new,
6482 			    &zc->zc_cookie,
6483 			    &zc->zc_objset_type, &zc->zc_perm_action);
6484 			dsl_dataset_rele(old, FTAG);
6485 		}
6486 	}
6487 	dsl_dataset_rele(new, FTAG);
6488 	dsl_pool_rele(dp, FTAG);
6489 	return (error);
6490 }
6491 
6492 /*
6493  * innvl: {
6494  *     "firstsnap" -> snapshot name
6495  * }
6496  *
6497  * outnvl: {
6498  *     "used" -> space in bytes
6499  *     "compressed" -> compressed space in bytes
6500  *     "uncompressed" -> uncompressed space in bytes
6501  * }
6502  */
6503 static const zfs_ioc_key_t zfs_keys_space_snaps[] = {
6504 	{"firstsnap",	DATA_TYPE_STRING,	0},
6505 };
6506 
6507 static int
6508 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
6509 {
6510 	int error;
6511 	dsl_pool_t *dp;
6512 	dsl_dataset_t *new, *old;
6513 	const char *firstsnap;
6514 	uint64_t used, comp, uncomp;
6515 
6516 	firstsnap = fnvlist_lookup_string(innvl, "firstsnap");
6517 
6518 	error = dsl_pool_hold(lastsnap, FTAG, &dp);
6519 	if (error != 0)
6520 		return (error);
6521 
6522 	error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
6523 	if (error == 0 && !new->ds_is_snapshot) {
6524 		dsl_dataset_rele(new, FTAG);
6525 		error = SET_ERROR(EINVAL);
6526 	}
6527 	if (error != 0) {
6528 		dsl_pool_rele(dp, FTAG);
6529 		return (error);
6530 	}
6531 	error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
6532 	if (error == 0 && !old->ds_is_snapshot) {
6533 		dsl_dataset_rele(old, FTAG);
6534 		error = SET_ERROR(EINVAL);
6535 	}
6536 	if (error != 0) {
6537 		dsl_dataset_rele(new, FTAG);
6538 		dsl_pool_rele(dp, FTAG);
6539 		return (error);
6540 	}
6541 
6542 	error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
6543 	dsl_dataset_rele(old, FTAG);
6544 	dsl_dataset_rele(new, FTAG);
6545 	dsl_pool_rele(dp, FTAG);
6546 	fnvlist_add_uint64(outnvl, "used", used);
6547 	fnvlist_add_uint64(outnvl, "compressed", comp);
6548 	fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
6549 	return (error);
6550 }
6551 
6552 /*
6553  * innvl: {
6554  *     "fd" -> file descriptor to write stream to (int32)
6555  *     (optional) "fromsnap" -> full snap name to send an incremental from
6556  *     (optional) "largeblockok" -> (value ignored)
6557  *         indicates that blocks > 128KB are permitted
6558  *     (optional) "embedok" -> (value ignored)
6559  *         presence indicates DRR_WRITE_EMBEDDED records are permitted
6560  *     (optional) "compressok" -> (value ignored)
6561  *         presence indicates compressed DRR_WRITE records are permitted
6562  *     (optional) "rawok" -> (value ignored)
6563  *         presence indicates raw encrypted records should be used.
6564  *     (optional) "savedok" -> (value ignored)
6565  *         presence indicates we should send a partially received snapshot
6566  *     (optional) "resume_object" and "resume_offset" -> (uint64)
6567  *         if present, resume send stream from specified object and offset.
6568  *     (optional) "redactbook" -> (string)
6569  *         if present, use this bookmark's redaction list to generate a redacted
6570  *         send stream
6571  * }
6572  *
6573  * outnvl is unused
6574  */
6575 static const zfs_ioc_key_t zfs_keys_send_new[] = {
6576 	{"fd",			DATA_TYPE_INT32,	0},
6577 	{"fromsnap",		DATA_TYPE_STRING,	ZK_OPTIONAL},
6578 	{"largeblockok",	DATA_TYPE_BOOLEAN,	ZK_OPTIONAL},
6579 	{"embedok",		DATA_TYPE_BOOLEAN,	ZK_OPTIONAL},
6580 	{"compressok",		DATA_TYPE_BOOLEAN,	ZK_OPTIONAL},
6581 	{"rawok",		DATA_TYPE_BOOLEAN,	ZK_OPTIONAL},
6582 	{"savedok",		DATA_TYPE_BOOLEAN,	ZK_OPTIONAL},
6583 	{"resume_object",	DATA_TYPE_UINT64,	ZK_OPTIONAL},
6584 	{"resume_offset",	DATA_TYPE_UINT64,	ZK_OPTIONAL},
6585 	{"redactbook",		DATA_TYPE_STRING,	ZK_OPTIONAL},
6586 };
6587 
6588 static int
6589 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
6590 {
6591 	(void) outnvl;
6592 	int error;
6593 	offset_t off;
6594 	const char *fromname = NULL;
6595 	int fd;
6596 	zfs_file_t *fp;
6597 	boolean_t largeblockok;
6598 	boolean_t embedok;
6599 	boolean_t compressok;
6600 	boolean_t rawok;
6601 	boolean_t savedok;
6602 	uint64_t resumeobj = 0;
6603 	uint64_t resumeoff = 0;
6604 	const char *redactbook = NULL;
6605 
6606 	fd = fnvlist_lookup_int32(innvl, "fd");
6607 
6608 	(void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
6609 
6610 	largeblockok = nvlist_exists(innvl, "largeblockok");
6611 	embedok = nvlist_exists(innvl, "embedok");
6612 	compressok = nvlist_exists(innvl, "compressok");
6613 	rawok = nvlist_exists(innvl, "rawok");
6614 	savedok = nvlist_exists(innvl, "savedok");
6615 
6616 	(void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
6617 	(void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);
6618 
6619 	(void) nvlist_lookup_string(innvl, "redactbook", &redactbook);
6620 
6621 	if ((fp = zfs_file_get(fd)) == NULL)
6622 		return (SET_ERROR(EBADF));
6623 
6624 	off = zfs_file_off(fp);
6625 
6626 	dmu_send_outparams_t out = {0};
6627 	out.dso_outfunc = dump_bytes;
6628 	out.dso_arg = fp;
6629 	out.dso_dryrun = B_FALSE;
6630 	error = dmu_send(snapname, fromname, embedok, largeblockok,
6631 	    compressok, rawok, savedok, resumeobj, resumeoff,
6632 	    redactbook, fd, &off, &out);
6633 
6634 	zfs_file_put(fp);
6635 	return (error);
6636 }
6637 
6638 static int
6639 send_space_sum(objset_t *os, void *buf, int len, void *arg)
6640 {
6641 	(void) os, (void) buf;
6642 	uint64_t *size = arg;
6643 
6644 	*size += len;
6645 	return (0);
6646 }
6647 
6648 /*
6649  * Determine approximately how large a zfs send stream will be -- the number
6650  * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
6651  *
6652  * innvl: {
6653  *     (optional) "from" -> full snap or bookmark name to send an incremental
6654  *                          from
6655  *     (optional) "largeblockok" -> (value ignored)
6656  *         indicates that blocks > 128KB are permitted
6657  *     (optional) "embedok" -> (value ignored)
6658  *         presence indicates DRR_WRITE_EMBEDDED records are permitted
6659  *     (optional) "compressok" -> (value ignored)
6660  *         presence indicates compressed DRR_WRITE records are permitted
6661  *     (optional) "rawok" -> (value ignored)
6662  *         presence indicates raw encrypted records should be used.
6663  *     (optional) "resume_object" and "resume_offset" -> (uint64)
6664  *         if present, resume send stream from specified object and offset.
6665  *     (optional) "fd" -> file descriptor to use as a cookie for progress
6666  *         tracking (int32)
6667  * }
6668  *
6669  * outnvl: {
6670  *     "space" -> bytes of space (uint64)
6671  * }
6672  */
6673 static const zfs_ioc_key_t zfs_keys_send_space[] = {
6674 	{"from",		DATA_TYPE_STRING,	ZK_OPTIONAL},
6675 	{"fromsnap",		DATA_TYPE_STRING,	ZK_OPTIONAL},
6676 	{"largeblockok",	DATA_TYPE_BOOLEAN,	ZK_OPTIONAL},
6677 	{"embedok",		DATA_TYPE_BOOLEAN,	ZK_OPTIONAL},
6678 	{"compressok",		DATA_TYPE_BOOLEAN,	ZK_OPTIONAL},
6679 	{"rawok",		DATA_TYPE_BOOLEAN,	ZK_OPTIONAL},
6680 	{"fd",			DATA_TYPE_INT32,	ZK_OPTIONAL},
6681 	{"redactbook",		DATA_TYPE_STRING,	ZK_OPTIONAL},
6682 	{"resume_object",	DATA_TYPE_UINT64,	ZK_OPTIONAL},
6683 	{"resume_offset",	DATA_TYPE_UINT64,	ZK_OPTIONAL},
6684 	{"bytes",		DATA_TYPE_UINT64,	ZK_OPTIONAL},
6685 };
6686 
6687 static int
6688 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
6689 {
6690 	dsl_pool_t *dp;
6691 	dsl_dataset_t *tosnap;
6692 	dsl_dataset_t *fromsnap = NULL;
6693 	int error;
6694 	const char *fromname = NULL;
6695 	const char *redactlist_book = NULL;
6696 	boolean_t largeblockok;
6697 	boolean_t embedok;
6698 	boolean_t compressok;
6699 	boolean_t rawok;
6700 	boolean_t savedok;
6701 	uint64_t space = 0;
6702 	boolean_t full_estimate = B_FALSE;
6703 	uint64_t resumeobj = 0;
6704 	uint64_t resumeoff = 0;
6705 	uint64_t resume_bytes = 0;
6706 	int32_t fd = -1;
6707 	zfs_bookmark_phys_t zbm = {0};
6708 
6709 	error = dsl_pool_hold(snapname, FTAG, &dp);
6710 	if (error != 0)
6711 		return (error);
6712 
6713 	error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
6714 	if (error != 0) {
6715 		dsl_pool_rele(dp, FTAG);
6716 		return (error);
6717 	}
6718 	(void) nvlist_lookup_int32(innvl, "fd", &fd);
6719 
6720 	largeblockok = nvlist_exists(innvl, "largeblockok");
6721 	embedok = nvlist_exists(innvl, "embedok");
6722 	compressok = nvlist_exists(innvl, "compressok");
6723 	rawok = nvlist_exists(innvl, "rawok");
6724 	savedok = nvlist_exists(innvl, "savedok");
6725 	boolean_t from = (nvlist_lookup_string(innvl, "from", &fromname) == 0);
6726 	boolean_t altbook = (nvlist_lookup_string(innvl, "redactbook",
6727 	    &redactlist_book) == 0);
6728 
6729 	(void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
6730 	(void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);
6731 	(void) nvlist_lookup_uint64(innvl, "bytes", &resume_bytes);
6732 
6733 	if (altbook) {
6734 		full_estimate = B_TRUE;
6735 	} else if (from) {
6736 		if (strchr(fromname, '#')) {
6737 			error = dsl_bookmark_lookup(dp, fromname, tosnap, &zbm);
6738 
6739 			/*
6740 			 * dsl_bookmark_lookup() will fail with EXDEV if
6741 			 * the from-bookmark and tosnap are at the same txg.
6742 			 * However, it's valid to do a send (and therefore,
6743 			 * a send estimate) from and to the same time point,
6744 			 * if the bookmark is redacted (the incremental send
6745 			 * can change what's redacted on the target).  In
6746 			 * this case, dsl_bookmark_lookup() fills in zbm
6747 			 * but returns EXDEV.  Ignore this error.
6748 			 */
6749 			if (error == EXDEV && zbm.zbm_redaction_obj != 0 &&
6750 			    zbm.zbm_guid ==
6751 			    dsl_dataset_phys(tosnap)->ds_guid)
6752 				error = 0;
6753 
6754 			if (error != 0) {
6755 				dsl_dataset_rele(tosnap, FTAG);
6756 				dsl_pool_rele(dp, FTAG);
6757 				return (error);
6758 			}
6759 			if (zbm.zbm_redaction_obj != 0 || !(zbm.zbm_flags &
6760 			    ZBM_FLAG_HAS_FBN)) {
6761 				full_estimate = B_TRUE;
6762 			}
6763 		} else if (strchr(fromname, '@')) {
6764 			error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
6765 			if (error != 0) {
6766 				dsl_dataset_rele(tosnap, FTAG);
6767 				dsl_pool_rele(dp, FTAG);
6768 				return (error);
6769 			}
6770 
6771 			if (!dsl_dataset_is_before(tosnap, fromsnap, 0)) {
6772 				full_estimate = B_TRUE;
6773 				dsl_dataset_rele(fromsnap, FTAG);
6774 			}
6775 		} else {
6776 			/*
6777 			 * from is not properly formatted as a snapshot or
6778 			 * bookmark
6779 			 */
6780 			dsl_dataset_rele(tosnap, FTAG);
6781 			dsl_pool_rele(dp, FTAG);
6782 			return (SET_ERROR(EINVAL));
6783 		}
6784 	}
6785 
6786 	if (full_estimate) {
6787 		dmu_send_outparams_t out = {0};
6788 		offset_t off = 0;
6789 		out.dso_outfunc = send_space_sum;
6790 		out.dso_arg = &space;
6791 		out.dso_dryrun = B_TRUE;
6792 		/*
6793 		 * We have to release these holds so dmu_send can take them.  It
6794 		 * will do all the error checking we need.
6795 		 */
6796 		dsl_dataset_rele(tosnap, FTAG);
6797 		dsl_pool_rele(dp, FTAG);
6798 		error = dmu_send(snapname, fromname, embedok, largeblockok,
6799 		    compressok, rawok, savedok, resumeobj, resumeoff,
6800 		    redactlist_book, fd, &off, &out);
6801 	} else {
6802 		error = dmu_send_estimate_fast(tosnap, fromsnap,
6803 		    (from && strchr(fromname, '#') != NULL ? &zbm : NULL),
6804 		    compressok || rawok, savedok, &space);
6805 		space -= resume_bytes;
6806 		if (fromsnap != NULL)
6807 			dsl_dataset_rele(fromsnap, FTAG);
6808 		dsl_dataset_rele(tosnap, FTAG);
6809 		dsl_pool_rele(dp, FTAG);
6810 	}
6811 
6812 	fnvlist_add_uint64(outnvl, "space", space);
6813 
6814 	return (error);
6815 }
6816 
6817 /*
6818  * Sync the currently open TXG to disk for the specified pool.
6819  * This is somewhat similar to 'zfs_sync()'.
6820  * For cases that do not result in error this ioctl will wait for
6821  * the currently open TXG to commit before returning back to the caller.
6822  *
6823  * innvl: {
6824  *  "force" -> when true, force uberblock update even if there is no dirty data.
6825  *             In addition this will cause the vdev configuration to be written
6826  *             out including updating the zpool cache file. (boolean_t)
6827  * }
6828  *
6829  * onvl is unused
6830  */
6831 static const zfs_ioc_key_t zfs_keys_pool_sync[] = {
6832 	{"force",	DATA_TYPE_BOOLEAN_VALUE,	0},
6833 };
6834 
6835 static int
6836 zfs_ioc_pool_sync(const char *pool, nvlist_t *innvl, nvlist_t *onvl)
6837 {
6838 	(void) onvl;
6839 	int err;
6840 	boolean_t rc, force = B_FALSE;
6841 	spa_t *spa;
6842 
6843 	if ((err = spa_open(pool, &spa, FTAG)) != 0)
6844 		return (err);
6845 
6846 	if (innvl) {
6847 		err = nvlist_lookup_boolean_value(innvl, "force", &rc);
6848 		if (err == 0)
6849 			force = rc;
6850 	}
6851 
6852 	if (force) {
6853 		spa_config_enter(spa, SCL_CONFIG, FTAG, RW_WRITER);
6854 		vdev_config_dirty(spa->spa_root_vdev);
6855 		spa_config_exit(spa, SCL_CONFIG, FTAG);
6856 	}
6857 	txg_wait_synced(spa_get_dsl(spa), 0);
6858 
6859 	spa_close(spa, FTAG);
6860 
6861 	return (0);
6862 }
6863 
6864 /*
6865  * Load a user's wrapping key into the kernel.
6866  * innvl: {
6867  *     "hidden_args" -> { "wkeydata" -> value }
6868  *         raw uint8_t array of encryption wrapping key data (32 bytes)
6869  *     (optional) "noop" -> (value ignored)
6870  *         presence indicated key should only be verified, not loaded
6871  * }
6872  */
6873 static const zfs_ioc_key_t zfs_keys_load_key[] = {
6874 	{"hidden_args",	DATA_TYPE_NVLIST,	0},
6875 	{"noop",	DATA_TYPE_BOOLEAN,	ZK_OPTIONAL},
6876 };
6877 
6878 static int
6879 zfs_ioc_load_key(const char *dsname, nvlist_t *innvl, nvlist_t *outnvl)
6880 {
6881 	(void) outnvl;
6882 	int ret;
6883 	dsl_crypto_params_t *dcp = NULL;
6884 	nvlist_t *hidden_args;
6885 	boolean_t noop = nvlist_exists(innvl, "noop");
6886 
6887 	if (strchr(dsname, '@') != NULL || strchr(dsname, '%') != NULL) {
6888 		ret = SET_ERROR(EINVAL);
6889 		goto error;
6890 	}
6891 
6892 	hidden_args = fnvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS);
6893 
6894 	ret = dsl_crypto_params_create_nvlist(DCP_CMD_NONE, NULL,
6895 	    hidden_args, &dcp);
6896 	if (ret != 0)
6897 		goto error;
6898 
6899 	ret = spa_keystore_load_wkey(dsname, dcp, noop);
6900 	if (ret != 0)
6901 		goto error;
6902 
6903 	dsl_crypto_params_free(dcp, noop);
6904 
6905 	return (0);
6906 
6907 error:
6908 	dsl_crypto_params_free(dcp, B_TRUE);
6909 	return (ret);
6910 }
6911 
6912 /*
6913  * Unload a user's wrapping key from the kernel.
6914  * Both innvl and outnvl are unused.
6915  */
6916 static const zfs_ioc_key_t zfs_keys_unload_key[] = {
6917 	/* no nvl keys */
6918 };
6919 
6920 static int
6921 zfs_ioc_unload_key(const char *dsname, nvlist_t *innvl, nvlist_t *outnvl)
6922 {
6923 	(void) innvl, (void) outnvl;
6924 	int ret = 0;
6925 
6926 	if (strchr(dsname, '@') != NULL || strchr(dsname, '%') != NULL) {
6927 		ret = (SET_ERROR(EINVAL));
6928 		goto out;
6929 	}
6930 
6931 	ret = spa_keystore_unload_wkey(dsname);
6932 	if (ret != 0)
6933 		goto out;
6934 
6935 out:
6936 	return (ret);
6937 }
6938 
6939 /*
6940  * Changes a user's wrapping key used to decrypt a dataset. The keyformat,
6941  * keylocation, pbkdf2salt, and pbkdf2iters properties can also be specified
6942  * here to change how the key is derived in userspace.
6943  *
6944  * innvl: {
6945  *    "hidden_args" (optional) -> { "wkeydata" -> value }
6946  *         raw uint8_t array of new encryption wrapping key data (32 bytes)
6947  *    "props" (optional) -> { prop -> value }
6948  * }
6949  *
6950  * outnvl is unused
6951  */
6952 static const zfs_ioc_key_t zfs_keys_change_key[] = {
6953 	{"crypt_cmd",	DATA_TYPE_UINT64,	ZK_OPTIONAL},
6954 	{"hidden_args",	DATA_TYPE_NVLIST,	ZK_OPTIONAL},
6955 	{"props",	DATA_TYPE_NVLIST,	ZK_OPTIONAL},
6956 };
6957 
6958 static int
6959 zfs_ioc_change_key(const char *dsname, nvlist_t *innvl, nvlist_t *outnvl)
6960 {
6961 	(void) outnvl;
6962 	int ret;
6963 	uint64_t cmd = DCP_CMD_NONE;
6964 	dsl_crypto_params_t *dcp = NULL;
6965 	nvlist_t *args = NULL, *hidden_args = NULL;
6966 
6967 	if (strchr(dsname, '@') != NULL || strchr(dsname, '%') != NULL) {
6968 		ret = (SET_ERROR(EINVAL));
6969 		goto error;
6970 	}
6971 
6972 	(void) nvlist_lookup_uint64(innvl, "crypt_cmd", &cmd);
6973 	(void) nvlist_lookup_nvlist(innvl, "props", &args);
6974 	(void) nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);
6975 
6976 	ret = dsl_crypto_params_create_nvlist(cmd, args, hidden_args, &dcp);
6977 	if (ret != 0)
6978 		goto error;
6979 
6980 	ret = spa_keystore_change_key(dsname, dcp);
6981 	if (ret != 0)
6982 		goto error;
6983 
6984 	dsl_crypto_params_free(dcp, B_FALSE);
6985 
6986 	return (0);
6987 
6988 error:
6989 	dsl_crypto_params_free(dcp, B_TRUE);
6990 	return (ret);
6991 }
6992 
6993 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
6994 
6995 static void
6996 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6997     zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
6998     boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
6999 {
7000 	zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
7001 
7002 	ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
7003 	ASSERT3U(ioc, <, ZFS_IOC_LAST);
7004 	ASSERT3P(vec->zvec_legacy_func, ==, NULL);
7005 	ASSERT3P(vec->zvec_func, ==, NULL);
7006 
7007 	vec->zvec_legacy_func = func;
7008 	vec->zvec_secpolicy = secpolicy;
7009 	vec->zvec_namecheck = namecheck;
7010 	vec->zvec_allow_log = log_history;
7011 	vec->zvec_pool_check = pool_check;
7012 }
7013 
7014 /*
7015  * See the block comment at the beginning of this file for details on
7016  * each argument to this function.
7017  */
7018 void
7019 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
7020     zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
7021     zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
7022     boolean_t allow_log, const zfs_ioc_key_t *nvl_keys, size_t num_keys)
7023 {
7024 	zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
7025 
7026 	ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
7027 	ASSERT3U(ioc, <, ZFS_IOC_LAST);
7028 	ASSERT3P(vec->zvec_legacy_func, ==, NULL);
7029 	ASSERT3P(vec->zvec_func, ==, NULL);
7030 
7031 	/* if we are logging, the name must be valid */
7032 	ASSERT(!allow_log || namecheck != NO_NAME);
7033 
7034 	vec->zvec_name = name;
7035 	vec->zvec_func = func;
7036 	vec->zvec_secpolicy = secpolicy;
7037 	vec->zvec_namecheck = namecheck;
7038 	vec->zvec_pool_check = pool_check;
7039 	vec->zvec_smush_outnvlist = smush_outnvlist;
7040 	vec->zvec_allow_log = allow_log;
7041 	vec->zvec_nvl_keys = nvl_keys;
7042 	vec->zvec_nvl_key_count = num_keys;
7043 }
7044 
7045 static void
7046 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
7047     zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
7048     zfs_ioc_poolcheck_t pool_check)
7049 {
7050 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
7051 	    POOL_NAME, log_history, pool_check);
7052 }
7053 
7054 void
7055 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
7056     zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
7057 {
7058 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
7059 	    DATASET_NAME, B_FALSE, pool_check);
7060 }
7061 
7062 static void
7063 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
7064 {
7065 	zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
7066 	    POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
7067 }
7068 
7069 static void
7070 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
7071     zfs_secpolicy_func_t *secpolicy)
7072 {
7073 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
7074 	    NO_NAME, B_FALSE, POOL_CHECK_NONE);
7075 }
7076 
7077 static void
7078 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
7079     zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
7080 {
7081 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
7082 	    DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
7083 }
7084 
7085 static void
7086 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
7087 {
7088 	zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
7089 	    zfs_secpolicy_read);
7090 }
7091 
7092 static void
7093 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
7094     zfs_secpolicy_func_t *secpolicy)
7095 {
7096 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
7097 	    DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
7098 }
7099 
7100 static void
7101 zfs_ioctl_init(void)
7102 {
7103 	zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
7104 	    zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
7105 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7106 	    zfs_keys_snapshot, ARRAY_SIZE(zfs_keys_snapshot));
7107 
7108 	zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
7109 	    zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
7110 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE,
7111 	    zfs_keys_log_history, ARRAY_SIZE(zfs_keys_log_history));
7112 
7113 	zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
7114 	    zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
7115 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
7116 	    zfs_keys_space_snaps, ARRAY_SIZE(zfs_keys_space_snaps));
7117 
7118 	zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
7119 	    zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
7120 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
7121 	    zfs_keys_send_new, ARRAY_SIZE(zfs_keys_send_new));
7122 
7123 	zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
7124 	    zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
7125 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
7126 	    zfs_keys_send_space, ARRAY_SIZE(zfs_keys_send_space));
7127 
7128 	zfs_ioctl_register("create", ZFS_IOC_CREATE,
7129 	    zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
7130 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7131 	    zfs_keys_create, ARRAY_SIZE(zfs_keys_create));
7132 
7133 	zfs_ioctl_register("clone", ZFS_IOC_CLONE,
7134 	    zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
7135 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7136 	    zfs_keys_clone, ARRAY_SIZE(zfs_keys_clone));
7137 
7138 	zfs_ioctl_register("remap", ZFS_IOC_REMAP,
7139 	    zfs_ioc_remap, zfs_secpolicy_none, DATASET_NAME,
7140 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE,
7141 	    zfs_keys_remap, ARRAY_SIZE(zfs_keys_remap));
7142 
7143 	zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
7144 	    zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
7145 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7146 	    zfs_keys_destroy_snaps, ARRAY_SIZE(zfs_keys_destroy_snaps));
7147 
7148 	zfs_ioctl_register("hold", ZFS_IOC_HOLD,
7149 	    zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
7150 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7151 	    zfs_keys_hold, ARRAY_SIZE(zfs_keys_hold));
7152 	zfs_ioctl_register("release", ZFS_IOC_RELEASE,
7153 	    zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
7154 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7155 	    zfs_keys_release, ARRAY_SIZE(zfs_keys_release));
7156 
7157 	zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
7158 	    zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
7159 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
7160 	    zfs_keys_get_holds, ARRAY_SIZE(zfs_keys_get_holds));
7161 
7162 	zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
7163 	    zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
7164 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE,
7165 	    zfs_keys_rollback, ARRAY_SIZE(zfs_keys_rollback));
7166 
7167 	zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
7168 	    zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
7169 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7170 	    zfs_keys_bookmark, ARRAY_SIZE(zfs_keys_bookmark));
7171 
7172 	zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
7173 	    zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
7174 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
7175 	    zfs_keys_get_bookmarks, ARRAY_SIZE(zfs_keys_get_bookmarks));
7176 
7177 	zfs_ioctl_register("get_bookmark_props", ZFS_IOC_GET_BOOKMARK_PROPS,
7178 	    zfs_ioc_get_bookmark_props, zfs_secpolicy_read, ENTITY_NAME,
7179 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE, zfs_keys_get_bookmark_props,
7180 	    ARRAY_SIZE(zfs_keys_get_bookmark_props));
7181 
7182 	zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
7183 	    zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
7184 	    POOL_NAME,
7185 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7186 	    zfs_keys_destroy_bookmarks,
7187 	    ARRAY_SIZE(zfs_keys_destroy_bookmarks));
7188 
7189 	zfs_ioctl_register("receive", ZFS_IOC_RECV_NEW,
7190 	    zfs_ioc_recv_new, zfs_secpolicy_recv, DATASET_NAME,
7191 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7192 	    zfs_keys_recv_new, ARRAY_SIZE(zfs_keys_recv_new));
7193 	zfs_ioctl_register("load-key", ZFS_IOC_LOAD_KEY,
7194 	    zfs_ioc_load_key, zfs_secpolicy_load_key,
7195 	    DATASET_NAME, POOL_CHECK_SUSPENDED, B_TRUE, B_TRUE,
7196 	    zfs_keys_load_key, ARRAY_SIZE(zfs_keys_load_key));
7197 	zfs_ioctl_register("unload-key", ZFS_IOC_UNLOAD_KEY,
7198 	    zfs_ioc_unload_key, zfs_secpolicy_load_key,
7199 	    DATASET_NAME, POOL_CHECK_SUSPENDED, B_TRUE, B_TRUE,
7200 	    zfs_keys_unload_key, ARRAY_SIZE(zfs_keys_unload_key));
7201 	zfs_ioctl_register("change-key", ZFS_IOC_CHANGE_KEY,
7202 	    zfs_ioc_change_key, zfs_secpolicy_change_key,
7203 	    DATASET_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY,
7204 	    B_TRUE, B_TRUE, zfs_keys_change_key,
7205 	    ARRAY_SIZE(zfs_keys_change_key));
7206 
7207 	zfs_ioctl_register("sync", ZFS_IOC_POOL_SYNC,
7208 	    zfs_ioc_pool_sync, zfs_secpolicy_none, POOL_NAME,
7209 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE,
7210 	    zfs_keys_pool_sync, ARRAY_SIZE(zfs_keys_pool_sync));
7211 	zfs_ioctl_register("reopen", ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
7212 	    zfs_secpolicy_config, POOL_NAME, POOL_CHECK_SUSPENDED, B_TRUE,
7213 	    B_TRUE, zfs_keys_pool_reopen, ARRAY_SIZE(zfs_keys_pool_reopen));
7214 
7215 	zfs_ioctl_register("channel_program", ZFS_IOC_CHANNEL_PROGRAM,
7216 	    zfs_ioc_channel_program, zfs_secpolicy_config,
7217 	    POOL_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE,
7218 	    B_TRUE, zfs_keys_channel_program,
7219 	    ARRAY_SIZE(zfs_keys_channel_program));
7220 
7221 	zfs_ioctl_register("redact", ZFS_IOC_REDACT,
7222 	    zfs_ioc_redact, zfs_secpolicy_config, DATASET_NAME,
7223 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7224 	    zfs_keys_redact, ARRAY_SIZE(zfs_keys_redact));
7225 
7226 	zfs_ioctl_register("zpool_checkpoint", ZFS_IOC_POOL_CHECKPOINT,
7227 	    zfs_ioc_pool_checkpoint, zfs_secpolicy_config, POOL_NAME,
7228 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7229 	    zfs_keys_pool_checkpoint, ARRAY_SIZE(zfs_keys_pool_checkpoint));
7230 
7231 	zfs_ioctl_register("zpool_discard_checkpoint",
7232 	    ZFS_IOC_POOL_DISCARD_CHECKPOINT, zfs_ioc_pool_discard_checkpoint,
7233 	    zfs_secpolicy_config, POOL_NAME,
7234 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7235 	    zfs_keys_pool_discard_checkpoint,
7236 	    ARRAY_SIZE(zfs_keys_pool_discard_checkpoint));
7237 
7238 	zfs_ioctl_register("initialize", ZFS_IOC_POOL_INITIALIZE,
7239 	    zfs_ioc_pool_initialize, zfs_secpolicy_config, POOL_NAME,
7240 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7241 	    zfs_keys_pool_initialize, ARRAY_SIZE(zfs_keys_pool_initialize));
7242 
7243 	zfs_ioctl_register("trim", ZFS_IOC_POOL_TRIM,
7244 	    zfs_ioc_pool_trim, zfs_secpolicy_config, POOL_NAME,
7245 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE,
7246 	    zfs_keys_pool_trim, ARRAY_SIZE(zfs_keys_pool_trim));
7247 
7248 	zfs_ioctl_register("wait", ZFS_IOC_WAIT,
7249 	    zfs_ioc_wait, zfs_secpolicy_none, POOL_NAME,
7250 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE,
7251 	    zfs_keys_pool_wait, ARRAY_SIZE(zfs_keys_pool_wait));
7252 
7253 	zfs_ioctl_register("wait_fs", ZFS_IOC_WAIT_FS,
7254 	    zfs_ioc_wait_fs, zfs_secpolicy_none, DATASET_NAME,
7255 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE,
7256 	    zfs_keys_fs_wait, ARRAY_SIZE(zfs_keys_fs_wait));
7257 
7258 	zfs_ioctl_register("set_bootenv", ZFS_IOC_SET_BOOTENV,
7259 	    zfs_ioc_set_bootenv, zfs_secpolicy_config, POOL_NAME,
7260 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE,
7261 	    zfs_keys_set_bootenv, ARRAY_SIZE(zfs_keys_set_bootenv));
7262 
7263 	zfs_ioctl_register("get_bootenv", ZFS_IOC_GET_BOOTENV,
7264 	    zfs_ioc_get_bootenv, zfs_secpolicy_none, POOL_NAME,
7265 	    POOL_CHECK_SUSPENDED, B_FALSE, B_TRUE,
7266 	    zfs_keys_get_bootenv, ARRAY_SIZE(zfs_keys_get_bootenv));
7267 
7268 	zfs_ioctl_register("zpool_vdev_get_props", ZFS_IOC_VDEV_GET_PROPS,
7269 	    zfs_ioc_vdev_get_props, zfs_secpolicy_read, POOL_NAME,
7270 	    POOL_CHECK_NONE, B_FALSE, B_FALSE, zfs_keys_vdev_get_props,
7271 	    ARRAY_SIZE(zfs_keys_vdev_get_props));
7272 
7273 	zfs_ioctl_register("zpool_vdev_set_props", ZFS_IOC_VDEV_SET_PROPS,
7274 	    zfs_ioc_vdev_set_props, zfs_secpolicy_config, POOL_NAME,
7275 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE,
7276 	    zfs_keys_vdev_set_props, ARRAY_SIZE(zfs_keys_vdev_set_props));
7277 
7278 	zfs_ioctl_register("scrub", ZFS_IOC_POOL_SCRUB,
7279 	    zfs_ioc_pool_scrub, zfs_secpolicy_config, POOL_NAME,
7280 	    POOL_CHECK_NONE, B_TRUE, B_TRUE,
7281 	    zfs_keys_pool_scrub, ARRAY_SIZE(zfs_keys_pool_scrub));
7282 
7283 	/* IOCTLS that use the legacy function signature */
7284 
7285 	zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
7286 	    zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
7287 
7288 	zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
7289 	    zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
7290 	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
7291 	    zfs_ioc_pool_scan);
7292 	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
7293 	    zfs_ioc_pool_upgrade);
7294 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
7295 	    zfs_ioc_vdev_add);
7296 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
7297 	    zfs_ioc_vdev_remove);
7298 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
7299 	    zfs_ioc_vdev_set_state);
7300 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
7301 	    zfs_ioc_vdev_attach);
7302 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
7303 	    zfs_ioc_vdev_detach);
7304 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
7305 	    zfs_ioc_vdev_setpath);
7306 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
7307 	    zfs_ioc_vdev_setfru);
7308 	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
7309 	    zfs_ioc_pool_set_props);
7310 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
7311 	    zfs_ioc_vdev_split);
7312 	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
7313 	    zfs_ioc_pool_reguid);
7314 
7315 	zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
7316 	    zfs_ioc_pool_configs, zfs_secpolicy_none);
7317 	zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
7318 	    zfs_ioc_pool_tryimport, zfs_secpolicy_config);
7319 	zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
7320 	    zfs_ioc_inject_fault, zfs_secpolicy_inject);
7321 	zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
7322 	    zfs_ioc_clear_fault, zfs_secpolicy_inject);
7323 	zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
7324 	    zfs_ioc_inject_list_next, zfs_secpolicy_inject);
7325 
7326 	/*
7327 	 * pool destroy, and export don't log the history as part of
7328 	 * zfsdev_ioctl, but rather zfs_ioc_pool_export
7329 	 * does the logging of those commands.
7330 	 */
7331 	zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
7332 	    zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
7333 	zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
7334 	    zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
7335 
7336 	zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
7337 	    zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
7338 	zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
7339 	    zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
7340 
7341 	zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
7342 	    zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
7343 	zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
7344 	    zfs_ioc_dsobj_to_dsname,
7345 	    zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
7346 	zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
7347 	    zfs_ioc_pool_get_history,
7348 	    zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
7349 
7350 	zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
7351 	    zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
7352 
7353 	zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
7354 	    zfs_secpolicy_config, B_TRUE, POOL_CHECK_READONLY);
7355 
7356 	zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
7357 	    zfs_ioc_space_written);
7358 	zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
7359 	    zfs_ioc_objset_recvd_props);
7360 	zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
7361 	    zfs_ioc_next_obj);
7362 	zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
7363 	    zfs_ioc_get_fsacl);
7364 	zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
7365 	    zfs_ioc_objset_stats);
7366 	zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
7367 	    zfs_ioc_objset_zplprops);
7368 	zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
7369 	    zfs_ioc_dataset_list_next);
7370 	zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
7371 	    zfs_ioc_snapshot_list_next);
7372 	zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
7373 	    zfs_ioc_send_progress);
7374 
7375 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
7376 	    zfs_ioc_diff, zfs_secpolicy_diff);
7377 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
7378 	    zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
7379 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
7380 	    zfs_ioc_obj_to_path, zfs_secpolicy_diff);
7381 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
7382 	    zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
7383 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
7384 	    zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
7385 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
7386 	    zfs_ioc_send, zfs_secpolicy_send);
7387 
7388 	zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
7389 	    zfs_secpolicy_none);
7390 	zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
7391 	    zfs_secpolicy_destroy);
7392 	zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
7393 	    zfs_secpolicy_rename);
7394 	zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
7395 	    zfs_secpolicy_recv);
7396 	zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
7397 	    zfs_secpolicy_promote);
7398 	zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
7399 	    zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
7400 	zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
7401 	    zfs_secpolicy_set_fsacl);
7402 
7403 	zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
7404 	    zfs_secpolicy_share, POOL_CHECK_NONE);
7405 	zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
7406 	    zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
7407 	zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
7408 	    zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
7409 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
7410 	zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
7411 	    zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
7412 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
7413 
7414 	zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_NEXT, zfs_ioc_events_next,
7415 	    zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
7416 	zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_CLEAR, zfs_ioc_events_clear,
7417 	    zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
7418 	zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_SEEK, zfs_ioc_events_seek,
7419 	    zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
7420 
7421 	zfs_ioctl_init_os();
7422 }
7423 
7424 /*
7425  * Verify that for non-legacy ioctls the input nvlist
7426  * pairs match against the expected input.
7427  *
7428  * Possible errors are:
7429  * ZFS_ERR_IOC_ARG_UNAVAIL	An unrecognized nvpair was encountered
7430  * ZFS_ERR_IOC_ARG_REQUIRED	A required nvpair is missing
7431  * ZFS_ERR_IOC_ARG_BADTYPE	Invalid type for nvpair
7432  */
7433 static int
7434 zfs_check_input_nvpairs(nvlist_t *innvl, const zfs_ioc_vec_t *vec)
7435 {
7436 	const zfs_ioc_key_t *nvl_keys = vec->zvec_nvl_keys;
7437 	boolean_t required_keys_found = B_FALSE;
7438 
7439 	/*
7440 	 * examine each input pair
7441 	 */
7442 	for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
7443 	    pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
7444 		const char *name = nvpair_name(pair);
7445 		data_type_t type = nvpair_type(pair);
7446 		boolean_t identified = B_FALSE;
7447 
7448 		/*
7449 		 * check pair against the documented names and type
7450 		 */
7451 		for (int k = 0; k < vec->zvec_nvl_key_count; k++) {
7452 			/* if not a wild card name, check for an exact match */
7453 			if ((nvl_keys[k].zkey_flags & ZK_WILDCARDLIST) == 0 &&
7454 			    strcmp(nvl_keys[k].zkey_name, name) != 0)
7455 				continue;
7456 
7457 			identified = B_TRUE;
7458 
7459 			if (nvl_keys[k].zkey_type != DATA_TYPE_ANY &&
7460 			    nvl_keys[k].zkey_type != type) {
7461 				return (SET_ERROR(ZFS_ERR_IOC_ARG_BADTYPE));
7462 			}
7463 
7464 			if (nvl_keys[k].zkey_flags & ZK_OPTIONAL)
7465 				continue;
7466 
7467 			required_keys_found = B_TRUE;
7468 			break;
7469 		}
7470 
7471 		/* allow an 'optional' key, everything else is invalid */
7472 		if (!identified &&
7473 		    (strcmp(name, "optional") != 0 ||
7474 		    type != DATA_TYPE_NVLIST)) {
7475 			return (SET_ERROR(ZFS_ERR_IOC_ARG_UNAVAIL));
7476 		}
7477 	}
7478 
7479 	/* verify that all required keys were found */
7480 	for (int k = 0; k < vec->zvec_nvl_key_count; k++) {
7481 		if (nvl_keys[k].zkey_flags & ZK_OPTIONAL)
7482 			continue;
7483 
7484 		if (nvl_keys[k].zkey_flags & ZK_WILDCARDLIST) {
7485 			/* at least one non-optional key is expected here */
7486 			if (!required_keys_found)
7487 				return (SET_ERROR(ZFS_ERR_IOC_ARG_REQUIRED));
7488 			continue;
7489 		}
7490 
7491 		if (!nvlist_exists(innvl, nvl_keys[k].zkey_name))
7492 			return (SET_ERROR(ZFS_ERR_IOC_ARG_REQUIRED));
7493 	}
7494 
7495 	return (0);
7496 }
7497 
7498 static int
7499 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
7500     zfs_ioc_poolcheck_t check)
7501 {
7502 	spa_t *spa;
7503 	int error;
7504 
7505 	ASSERT(type == POOL_NAME || type == DATASET_NAME ||
7506 	    type == ENTITY_NAME);
7507 
7508 	if (check & POOL_CHECK_NONE)
7509 		return (0);
7510 
7511 	error = spa_open(name, &spa, FTAG);
7512 	if (error == 0) {
7513 		if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
7514 			error = SET_ERROR(EAGAIN);
7515 		else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
7516 			error = SET_ERROR(EROFS);
7517 		spa_close(spa, FTAG);
7518 	}
7519 	return (error);
7520 }
7521 
7522 int
7523 zfsdev_getminor(zfs_file_t *fp, minor_t *minorp)
7524 {
7525 	zfsdev_state_t *zs, *fpd;
7526 
7527 	ASSERT(!MUTEX_HELD(&zfsdev_state_lock));
7528 
7529 	fpd = zfs_file_private(fp);
7530 	if (fpd == NULL)
7531 		return (SET_ERROR(EBADF));
7532 
7533 	mutex_enter(&zfsdev_state_lock);
7534 
7535 	for (zs = &zfsdev_state_listhead; zs != NULL; zs = zs->zs_next) {
7536 
7537 		if (zs->zs_minor == -1)
7538 			continue;
7539 
7540 		if (fpd == zs) {
7541 			*minorp = fpd->zs_minor;
7542 			mutex_exit(&zfsdev_state_lock);
7543 			return (0);
7544 		}
7545 	}
7546 
7547 	mutex_exit(&zfsdev_state_lock);
7548 
7549 	return (SET_ERROR(EBADF));
7550 }
7551 
7552 void *
7553 zfsdev_get_state(minor_t minor, enum zfsdev_state_type which)
7554 {
7555 	zfsdev_state_t *zs;
7556 
7557 	for (zs = &zfsdev_state_listhead; zs != NULL; zs = zs->zs_next) {
7558 		if (zs->zs_minor == minor) {
7559 			membar_consumer();
7560 			switch (which) {
7561 			case ZST_ONEXIT:
7562 				return (zs->zs_onexit);
7563 			case ZST_ZEVENT:
7564 				return (zs->zs_zevent);
7565 			case ZST_ALL:
7566 				return (zs);
7567 			}
7568 		}
7569 	}
7570 
7571 	return (NULL);
7572 }
7573 
7574 /*
7575  * Find a free minor number.  The zfsdev_state_list is expected to
7576  * be short since it is only a list of currently open file handles.
7577  */
7578 static minor_t
7579 zfsdev_minor_alloc(void)
7580 {
7581 	static minor_t last_minor = 0;
7582 	minor_t m;
7583 
7584 	ASSERT(MUTEX_HELD(&zfsdev_state_lock));
7585 
7586 	for (m = last_minor + 1; m != last_minor; m++) {
7587 		if (m > ZFSDEV_MAX_MINOR)
7588 			m = 1;
7589 		if (zfsdev_get_state(m, ZST_ALL) == NULL) {
7590 			last_minor = m;
7591 			return (m);
7592 		}
7593 	}
7594 
7595 	return (0);
7596 }
7597 
7598 int
7599 zfsdev_state_init(void *priv)
7600 {
7601 	zfsdev_state_t *zs, *zsprev = NULL;
7602 	minor_t minor;
7603 	boolean_t newzs = B_FALSE;
7604 
7605 	ASSERT(MUTEX_HELD(&zfsdev_state_lock));
7606 
7607 	minor = zfsdev_minor_alloc();
7608 	if (minor == 0)
7609 		return (SET_ERROR(ENXIO));
7610 
7611 	for (zs = &zfsdev_state_listhead; zs != NULL; zs = zs->zs_next) {
7612 		if (zs->zs_minor == -1)
7613 			break;
7614 		zsprev = zs;
7615 	}
7616 
7617 	if (!zs) {
7618 		zs = kmem_zalloc(sizeof (zfsdev_state_t), KM_SLEEP);
7619 		newzs = B_TRUE;
7620 	}
7621 
7622 	zfsdev_private_set_state(priv, zs);
7623 
7624 	zfs_onexit_init((zfs_onexit_t **)&zs->zs_onexit);
7625 	zfs_zevent_init((zfs_zevent_t **)&zs->zs_zevent);
7626 
7627 	/*
7628 	 * In order to provide for lock-free concurrent read access
7629 	 * to the minor list in zfsdev_get_state(), new entries
7630 	 * must be completely written before linking them into the
7631 	 * list whereas existing entries are already linked; the last
7632 	 * operation must be updating zs_minor (from -1 to the new
7633 	 * value).
7634 	 */
7635 	if (newzs) {
7636 		zs->zs_minor = minor;
7637 		membar_producer();
7638 		zsprev->zs_next = zs;
7639 	} else {
7640 		membar_producer();
7641 		zs->zs_minor = minor;
7642 	}
7643 
7644 	return (0);
7645 }
7646 
7647 void
7648 zfsdev_state_destroy(void *priv)
7649 {
7650 	zfsdev_state_t *zs = zfsdev_private_get_state(priv);
7651 
7652 	ASSERT(zs != NULL);
7653 	ASSERT3S(zs->zs_minor, >, 0);
7654 
7655 	/*
7656 	 * The last reference to this zfsdev file descriptor is being dropped.
7657 	 * We don't have to worry about lookup grabbing this state object, and
7658 	 * zfsdev_state_init() will not try to reuse this object until it is
7659 	 * invalidated by setting zs_minor to -1.  Invalidation must be done
7660 	 * last, with a memory barrier to ensure ordering.  This lets us avoid
7661 	 * taking the global zfsdev state lock around destruction.
7662 	 */
7663 	zfs_onexit_destroy(zs->zs_onexit);
7664 	zfs_zevent_destroy(zs->zs_zevent);
7665 	zs->zs_onexit = NULL;
7666 	zs->zs_zevent = NULL;
7667 	membar_producer();
7668 	zs->zs_minor = -1;
7669 }
7670 
7671 long
7672 zfsdev_ioctl_common(uint_t vecnum, zfs_cmd_t *zc, int flag)
7673 {
7674 	int error, cmd;
7675 	const zfs_ioc_vec_t *vec;
7676 	char *saved_poolname = NULL;
7677 	uint64_t max_nvlist_src_size;
7678 	size_t saved_poolname_len = 0;
7679 	nvlist_t *innvl = NULL;
7680 	fstrans_cookie_t cookie;
7681 	hrtime_t start_time = gethrtime();
7682 
7683 	cmd = vecnum;
7684 	error = 0;
7685 	if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
7686 		return (SET_ERROR(ZFS_ERR_IOC_CMD_UNAVAIL));
7687 
7688 	vec = &zfs_ioc_vec[vecnum];
7689 
7690 	/*
7691 	 * The registered ioctl list may be sparse, verify that either
7692 	 * a normal or legacy handler are registered.
7693 	 */
7694 	if (vec->zvec_func == NULL && vec->zvec_legacy_func == NULL)
7695 		return (SET_ERROR(ZFS_ERR_IOC_CMD_UNAVAIL));
7696 
7697 	zc->zc_iflags = flag & FKIOCTL;
7698 	max_nvlist_src_size = zfs_max_nvlist_src_size_os();
7699 	if (zc->zc_nvlist_src_size > max_nvlist_src_size) {
7700 		/*
7701 		 * Make sure the user doesn't pass in an insane value for
7702 		 * zc_nvlist_src_size.  We have to check, since we will end
7703 		 * up allocating that much memory inside of get_nvlist().  This
7704 		 * prevents a nefarious user from allocating tons of kernel
7705 		 * memory.
7706 		 *
7707 		 * Also, we return EINVAL instead of ENOMEM here.  The reason
7708 		 * being that returning ENOMEM from an ioctl() has a special
7709 		 * connotation; that the user's size value is too small and
7710 		 * needs to be expanded to hold the nvlist.  See
7711 		 * zcmd_expand_dst_nvlist() for details.
7712 		 */
7713 		error = SET_ERROR(EINVAL);	/* User's size too big */
7714 
7715 	} else if (zc->zc_nvlist_src_size != 0) {
7716 		error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
7717 		    zc->zc_iflags, &innvl);
7718 		if (error != 0)
7719 			goto out;
7720 	}
7721 
7722 	/*
7723 	 * Ensure that all pool/dataset names are valid before we pass down to
7724 	 * the lower layers.
7725 	 */
7726 	zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
7727 	switch (vec->zvec_namecheck) {
7728 	case POOL_NAME:
7729 		if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
7730 			error = SET_ERROR(EINVAL);
7731 		else
7732 			error = pool_status_check(zc->zc_name,
7733 			    vec->zvec_namecheck, vec->zvec_pool_check);
7734 		break;
7735 
7736 	case DATASET_NAME:
7737 		if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
7738 			error = SET_ERROR(EINVAL);
7739 		else
7740 			error = pool_status_check(zc->zc_name,
7741 			    vec->zvec_namecheck, vec->zvec_pool_check);
7742 		break;
7743 
7744 	case ENTITY_NAME:
7745 		if (entity_namecheck(zc->zc_name, NULL, NULL) != 0) {
7746 			error = SET_ERROR(EINVAL);
7747 		} else {
7748 			error = pool_status_check(zc->zc_name,
7749 			    vec->zvec_namecheck, vec->zvec_pool_check);
7750 		}
7751 		break;
7752 
7753 	case NO_NAME:
7754 		break;
7755 	}
7756 	/*
7757 	 * Ensure that all input pairs are valid before we pass them down
7758 	 * to the lower layers.
7759 	 *
7760 	 * The vectored functions can use fnvlist_lookup_{type} for any
7761 	 * required pairs since zfs_check_input_nvpairs() confirmed that
7762 	 * they exist and are of the correct type.
7763 	 */
7764 	if (error == 0 && vec->zvec_func != NULL) {
7765 		error = zfs_check_input_nvpairs(innvl, vec);
7766 		if (error != 0)
7767 			goto out;
7768 	}
7769 
7770 	if (error == 0) {
7771 		cookie = spl_fstrans_mark();
7772 		error = vec->zvec_secpolicy(zc, innvl, CRED());
7773 		spl_fstrans_unmark(cookie);
7774 	}
7775 
7776 	if (error != 0)
7777 		goto out;
7778 
7779 	/* legacy ioctls can modify zc_name */
7780 	/*
7781 	 * Can't use kmem_strdup() as we might truncate the string and
7782 	 * kmem_strfree() would then free with incorrect size.
7783 	 */
7784 	saved_poolname_len = strlen(zc->zc_name) + 1;
7785 	saved_poolname = kmem_alloc(saved_poolname_len, KM_SLEEP);
7786 
7787 	strlcpy(saved_poolname, zc->zc_name, saved_poolname_len);
7788 	saved_poolname[strcspn(saved_poolname, "/@#")] = '\0';
7789 
7790 	if (vec->zvec_func != NULL) {
7791 		nvlist_t *outnvl;
7792 		int puterror = 0;
7793 		spa_t *spa;
7794 		nvlist_t *lognv = NULL;
7795 
7796 		ASSERT(vec->zvec_legacy_func == NULL);
7797 
7798 		/*
7799 		 * Add the innvl to the lognv before calling the func,
7800 		 * in case the func changes the innvl.
7801 		 */
7802 		if (vec->zvec_allow_log) {
7803 			lognv = fnvlist_alloc();
7804 			fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
7805 			    vec->zvec_name);
7806 			if (!nvlist_empty(innvl)) {
7807 				fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
7808 				    innvl);
7809 			}
7810 		}
7811 
7812 		outnvl = fnvlist_alloc();
7813 		cookie = spl_fstrans_mark();
7814 		error = vec->zvec_func(zc->zc_name, innvl, outnvl);
7815 		spl_fstrans_unmark(cookie);
7816 
7817 		/*
7818 		 * Some commands can partially execute, modify state, and still
7819 		 * return an error.  In these cases, attempt to record what
7820 		 * was modified.
7821 		 */
7822 		if ((error == 0 ||
7823 		    (cmd == ZFS_IOC_CHANNEL_PROGRAM && error != EINVAL)) &&
7824 		    vec->zvec_allow_log &&
7825 		    spa_open(zc->zc_name, &spa, FTAG) == 0) {
7826 			if (!nvlist_empty(outnvl)) {
7827 				size_t out_size = fnvlist_size(outnvl);
7828 				if (out_size > zfs_history_output_max) {
7829 					fnvlist_add_int64(lognv,
7830 					    ZPOOL_HIST_OUTPUT_SIZE, out_size);
7831 				} else {
7832 					fnvlist_add_nvlist(lognv,
7833 					    ZPOOL_HIST_OUTPUT_NVL, outnvl);
7834 				}
7835 			}
7836 			if (error != 0) {
7837 				fnvlist_add_int64(lognv, ZPOOL_HIST_ERRNO,
7838 				    error);
7839 			}
7840 			fnvlist_add_int64(lognv, ZPOOL_HIST_ELAPSED_NS,
7841 			    gethrtime() - start_time);
7842 			(void) spa_history_log_nvl(spa, lognv);
7843 			spa_close(spa, FTAG);
7844 		}
7845 		fnvlist_free(lognv);
7846 
7847 		if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
7848 			int smusherror = 0;
7849 			if (vec->zvec_smush_outnvlist) {
7850 				smusherror = nvlist_smush(outnvl,
7851 				    zc->zc_nvlist_dst_size);
7852 			}
7853 			if (smusherror == 0)
7854 				puterror = put_nvlist(zc, outnvl);
7855 		}
7856 
7857 		if (puterror != 0)
7858 			error = puterror;
7859 
7860 		nvlist_free(outnvl);
7861 	} else {
7862 		cookie = spl_fstrans_mark();
7863 		error = vec->zvec_legacy_func(zc);
7864 		spl_fstrans_unmark(cookie);
7865 	}
7866 
7867 out:
7868 	nvlist_free(innvl);
7869 	if (error == 0 && vec->zvec_allow_log) {
7870 		char *s = tsd_get(zfs_allow_log_key);
7871 		if (s != NULL)
7872 			kmem_strfree(s);
7873 		(void) tsd_set(zfs_allow_log_key, kmem_strdup(saved_poolname));
7874 	}
7875 	if (saved_poolname != NULL)
7876 		kmem_free(saved_poolname, saved_poolname_len);
7877 
7878 	return (error);
7879 }
7880 
7881 int
7882 zfs_kmod_init(void)
7883 {
7884 	int error;
7885 
7886 	if ((error = zvol_init()) != 0)
7887 		return (error);
7888 
7889 	spa_init(SPA_MODE_READ | SPA_MODE_WRITE);
7890 	zfs_init();
7891 
7892 	zfs_ioctl_init();
7893 
7894 	mutex_init(&zfsdev_state_lock, NULL, MUTEX_DEFAULT, NULL);
7895 	zfsdev_state_listhead.zs_minor = -1;
7896 
7897 	if ((error = zfsdev_attach()) != 0)
7898 		goto out;
7899 
7900 	tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
7901 	tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
7902 
7903 	return (0);
7904 out:
7905 	zfs_fini();
7906 	spa_fini();
7907 	zvol_fini();
7908 
7909 	return (error);
7910 }
7911 
7912 void
7913 zfs_kmod_fini(void)
7914 {
7915 	zfsdev_state_t *zs, *zsnext = NULL;
7916 
7917 	zfsdev_detach();
7918 
7919 	mutex_destroy(&zfsdev_state_lock);
7920 
7921 	for (zs = &zfsdev_state_listhead; zs != NULL; zs = zsnext) {
7922 		zsnext = zs->zs_next;
7923 		if (zs->zs_onexit)
7924 			zfs_onexit_destroy(zs->zs_onexit);
7925 		if (zs->zs_zevent)
7926 			zfs_zevent_destroy(zs->zs_zevent);
7927 		if (zs != &zfsdev_state_listhead)
7928 			kmem_free(zs, sizeof (zfsdev_state_t));
7929 	}
7930 
7931 	zfs_ereport_taskq_fini();	/* run before zfs_fini() on Linux */
7932 	zfs_fini();
7933 	spa_fini();
7934 	zvol_fini();
7935 
7936 	tsd_destroy(&rrw_tsd_key);
7937 	tsd_destroy(&zfs_allow_log_key);
7938 }
7939 
7940 ZFS_MODULE_PARAM(zfs, zfs_, max_nvlist_src_size, U64, ZMOD_RW,
7941 	"Maximum size in bytes allowed for src nvlist passed with ZFS ioctls");
7942 
7943 ZFS_MODULE_PARAM(zfs, zfs_, history_output_max, U64, ZMOD_RW,
7944 	"Maximum size in bytes of ZFS ioctl output that will be logged");
7945