xref: /titanic_41/usr/src/uts/common/fs/zfs/zfs_ioctl.c (revision a9b5625ad449e28716053a1fd299e7e3271f750d)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 
22 /*
23  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24  * Portions Copyright 2011 Martin Matuska
25  * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
26  * Copyright (c) 2014, Joyent, Inc. All rights reserved.
27  * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
28  * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
29  * Copyright (c) 2013 Steven Hartland. All rights reserved.
30  * Copyright (c) 2014, Nexenta Systems, Inc. All rights reserved.
31  */
32 
33 /*
34  * ZFS ioctls.
35  *
36  * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
37  * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
38  *
39  * There are two ways that we handle ioctls: the legacy way where almost
40  * all of the logic is in the ioctl callback, and the new way where most
41  * of the marshalling is handled in the common entry point, zfsdev_ioctl().
42  *
43  * Non-legacy ioctls should be registered by calling
44  * zfs_ioctl_register() from zfs_ioctl_init().  The ioctl is invoked
45  * from userland by lzc_ioctl().
46  *
47  * The registration arguments are as follows:
48  *
49  * const char *name
50  *   The name of the ioctl.  This is used for history logging.  If the
51  *   ioctl returns successfully (the callback returns 0), and allow_log
52  *   is true, then a history log entry will be recorded with the input &
53  *   output nvlists.  The log entry can be printed with "zpool history -i".
54  *
55  * zfs_ioc_t ioc
56  *   The ioctl request number, which userland will pass to ioctl(2).
57  *   The ioctl numbers can change from release to release, because
58  *   the caller (libzfs) must be matched to the kernel.
59  *
60  * zfs_secpolicy_func_t *secpolicy
61  *   This function will be called before the zfs_ioc_func_t, to
62  *   determine if this operation is permitted.  It should return EPERM
63  *   on failure, and 0 on success.  Checks include determining if the
64  *   dataset is visible in this zone, and if the user has either all
65  *   zfs privileges in the zone (SYS_MOUNT), or has been granted permission
66  *   to do this operation on this dataset with "zfs allow".
67  *
68  * zfs_ioc_namecheck_t namecheck
69  *   This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
70  *   name, a dataset name, or nothing.  If the name is not well-formed,
71  *   the ioctl will fail and the callback will not be called.
72  *   Therefore, the callback can assume that the name is well-formed
73  *   (e.g. is null-terminated, doesn't have more than one '@' character,
74  *   doesn't have invalid characters).
75  *
76  * zfs_ioc_poolcheck_t pool_check
77  *   This specifies requirements on the pool state.  If the pool does
78  *   not meet them (is suspended or is readonly), the ioctl will fail
79  *   and the callback will not be called.  If any checks are specified
80  *   (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
81  *   Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
82  *   POOL_CHECK_READONLY).
83  *
84  * boolean_t smush_outnvlist
85  *   If smush_outnvlist is true, then the output is presumed to be a
86  *   list of errors, and it will be "smushed" down to fit into the
87  *   caller's buffer, by removing some entries and replacing them with a
88  *   single "N_MORE_ERRORS" entry indicating how many were removed.  See
89  *   nvlist_smush() for details.  If smush_outnvlist is false, and the
90  *   outnvlist does not fit into the userland-provided buffer, then the
91  *   ioctl will fail with ENOMEM.
92  *
93  * zfs_ioc_func_t *func
94  *   The callback function that will perform the operation.
95  *
96  *   The callback should return 0 on success, or an error number on
97  *   failure.  If the function fails, the userland ioctl will return -1,
98  *   and errno will be set to the callback's return value.  The callback
99  *   will be called with the following arguments:
100  *
101  *   const char *name
102  *     The name of the pool or dataset to operate on, from
103  *     zfs_cmd_t:zc_name.  The 'namecheck' argument specifies the
104  *     expected type (pool, dataset, or none).
105  *
106  *   nvlist_t *innvl
107  *     The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src.  Or
108  *     NULL if no input nvlist was provided.  Changes to this nvlist are
109  *     ignored.  If the input nvlist could not be deserialized, the
110  *     ioctl will fail and the callback will not be called.
111  *
112  *   nvlist_t *outnvl
113  *     The output nvlist, initially empty.  The callback can fill it in,
114  *     and it will be returned to userland by serializing it into
115  *     zfs_cmd_t:zc_nvlist_dst.  If it is non-empty, and serialization
116  *     fails (e.g. because the caller didn't supply a large enough
117  *     buffer), then the overall ioctl will fail.  See the
118  *     'smush_nvlist' argument above for additional behaviors.
119  *
120  *     There are two typical uses of the output nvlist:
121  *       - To return state, e.g. property values.  In this case,
122  *         smush_outnvlist should be false.  If the buffer was not large
123  *         enough, the caller will reallocate a larger buffer and try
124  *         the ioctl again.
125  *
126  *       - To return multiple errors from an ioctl which makes on-disk
127  *         changes.  In this case, smush_outnvlist should be true.
128  *         Ioctls which make on-disk modifications should generally not
129  *         use the outnvl if they succeed, because the caller can not
130  *         distinguish between the operation failing, and
131  *         deserialization failing.
132  */
133 
134 #include <sys/types.h>
135 #include <sys/param.h>
136 #include <sys/errno.h>
137 #include <sys/uio.h>
138 #include <sys/buf.h>
139 #include <sys/modctl.h>
140 #include <sys/open.h>
141 #include <sys/file.h>
142 #include <sys/kmem.h>
143 #include <sys/conf.h>
144 #include <sys/cmn_err.h>
145 #include <sys/stat.h>
146 #include <sys/zfs_ioctl.h>
147 #include <sys/zfs_vfsops.h>
148 #include <sys/zfs_znode.h>
149 #include <sys/zap.h>
150 #include <sys/spa.h>
151 #include <sys/spa_impl.h>
152 #include <sys/vdev.h>
153 #include <sys/priv_impl.h>
154 #include <sys/dmu.h>
155 #include <sys/dsl_dir.h>
156 #include <sys/dsl_dataset.h>
157 #include <sys/dsl_prop.h>
158 #include <sys/dsl_deleg.h>
159 #include <sys/dmu_objset.h>
160 #include <sys/dmu_impl.h>
161 #include <sys/dmu_tx.h>
162 #include <sys/ddi.h>
163 #include <sys/sunddi.h>
164 #include <sys/sunldi.h>
165 #include <sys/policy.h>
166 #include <sys/zone.h>
167 #include <sys/nvpair.h>
168 #include <sys/pathname.h>
169 #include <sys/mount.h>
170 #include <sys/sdt.h>
171 #include <sys/fs/zfs.h>
172 #include <sys/zfs_ctldir.h>
173 #include <sys/zfs_dir.h>
174 #include <sys/zfs_onexit.h>
175 #include <sys/zvol.h>
176 #include <sys/dsl_scan.h>
177 #include <sharefs/share.h>
178 #include <sys/dmu_objset.h>
179 #include <sys/dmu_send.h>
180 #include <sys/dsl_destroy.h>
181 #include <sys/dsl_bookmark.h>
182 #include <sys/dsl_userhold.h>
183 #include <sys/zfeature.h>
184 #include <sys/zfs_events.h>
185 
186 #include "zfs_namecheck.h"
187 #include "zfs_prop.h"
188 #include "zfs_deleg.h"
189 #include "zfs_comutil.h"
190 
191 extern struct modlfs zfs_modlfs;
192 
193 extern void zfs_init(void);
194 extern void zfs_fini(void);
195 
196 ldi_ident_t zfs_li = NULL;
197 dev_info_t *zfs_dip;
198 
199 uint_t zfs_fsyncer_key;
200 extern uint_t rrw_tsd_key;
201 static uint_t zfs_allow_log_key;
202 
203 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
204 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
205 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
206 
207 typedef enum {
208 	NO_NAME,
209 	POOL_NAME,
210 	DATASET_NAME
211 } zfs_ioc_namecheck_t;
212 
213 typedef enum {
214 	POOL_CHECK_NONE		= 1 << 0,
215 	POOL_CHECK_SUSPENDED	= 1 << 1,
216 	POOL_CHECK_READONLY	= 1 << 2,
217 } zfs_ioc_poolcheck_t;
218 
219 typedef struct zfs_ioc_vec {
220 	zfs_ioc_legacy_func_t	*zvec_legacy_func;
221 	zfs_ioc_func_t		*zvec_func;
222 	zfs_secpolicy_func_t	*zvec_secpolicy;
223 	zfs_ioc_namecheck_t	zvec_namecheck;
224 	boolean_t		zvec_allow_log;
225 	zfs_ioc_poolcheck_t	zvec_pool_check;
226 	boolean_t		zvec_smush_outnvlist;
227 	const char		*zvec_name;
228 } zfs_ioc_vec_t;
229 
230 /* This array is indexed by zfs_userquota_prop_t */
231 static const char *userquota_perms[] = {
232 	ZFS_DELEG_PERM_USERUSED,
233 	ZFS_DELEG_PERM_USERQUOTA,
234 	ZFS_DELEG_PERM_GROUPUSED,
235 	ZFS_DELEG_PERM_GROUPQUOTA,
236 };
237 
238 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
239 static int zfs_check_settable(const char *name, nvpair_t *property,
240     cred_t *cr);
241 static int zfs_check_clearable(char *dataset, nvlist_t *props,
242     nvlist_t **errors);
243 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
244     boolean_t *);
245 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
246 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
247 
248 static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature);
249 
250 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
251 void
252 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
253 {
254 	const char *newfile;
255 	char buf[512];
256 	va_list adx;
257 
258 	/*
259 	 * Get rid of annoying "../common/" prefix to filename.
260 	 */
261 	newfile = strrchr(file, '/');
262 	if (newfile != NULL) {
263 		newfile = newfile + 1; /* Get rid of leading / */
264 	} else {
265 		newfile = file;
266 	}
267 
268 	va_start(adx, fmt);
269 	(void) vsnprintf(buf, sizeof (buf), fmt, adx);
270 	va_end(adx);
271 
272 	/*
273 	 * To get this data, use the zfs-dprintf probe as so:
274 	 * dtrace -q -n 'zfs-dprintf \
275 	 *	/stringof(arg0) == "dbuf.c"/ \
276 	 *	{printf("%s: %s", stringof(arg1), stringof(arg3))}'
277 	 * arg0 = file name
278 	 * arg1 = function name
279 	 * arg2 = line number
280 	 * arg3 = message
281 	 */
282 	DTRACE_PROBE4(zfs__dprintf,
283 	    char *, newfile, char *, func, int, line, char *, buf);
284 }
285 
286 static void
287 history_str_free(char *buf)
288 {
289 	kmem_free(buf, HIS_MAX_RECORD_LEN);
290 }
291 
292 static char *
293 history_str_get(zfs_cmd_t *zc)
294 {
295 	char *buf;
296 
297 	if (zc->zc_history == NULL)
298 		return (NULL);
299 
300 	buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
301 	if (copyinstr((void *)(uintptr_t)zc->zc_history,
302 	    buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
303 		history_str_free(buf);
304 		return (NULL);
305 	}
306 
307 	buf[HIS_MAX_RECORD_LEN -1] = '\0';
308 
309 	return (buf);
310 }
311 
312 /*
313  * Check to see if the named dataset is currently defined as bootable
314  */
315 static boolean_t
316 zfs_is_bootfs(const char *name)
317 {
318 	objset_t *os;
319 
320 	if (dmu_objset_hold(name, FTAG, &os) == 0) {
321 		boolean_t ret;
322 		ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
323 		dmu_objset_rele(os, FTAG);
324 		return (ret);
325 	}
326 	return (B_FALSE);
327 }
328 
329 /*
330  * Return non-zero if the spa version is less than requested version.
331  */
332 static int
333 zfs_earlier_version(const char *name, int version)
334 {
335 	spa_t *spa;
336 
337 	if (spa_open(name, &spa, FTAG) == 0) {
338 		if (spa_version(spa) < version) {
339 			spa_close(spa, FTAG);
340 			return (1);
341 		}
342 		spa_close(spa, FTAG);
343 	}
344 	return (0);
345 }
346 
347 /*
348  * Return TRUE if the ZPL version is less than requested version.
349  */
350 static boolean_t
351 zpl_earlier_version(const char *name, int version)
352 {
353 	objset_t *os;
354 	boolean_t rc = B_TRUE;
355 
356 	if (dmu_objset_hold(name, FTAG, &os) == 0) {
357 		uint64_t zplversion;
358 
359 		if (dmu_objset_type(os) != DMU_OST_ZFS) {
360 			dmu_objset_rele(os, FTAG);
361 			return (B_TRUE);
362 		}
363 		/* XXX reading from non-owned objset */
364 		if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
365 			rc = zplversion < version;
366 		dmu_objset_rele(os, FTAG);
367 	}
368 	return (rc);
369 }
370 
371 static void
372 zfs_log_history(zfs_cmd_t *zc)
373 {
374 	spa_t *spa;
375 	char *buf;
376 
377 	if ((buf = history_str_get(zc)) == NULL)
378 		return;
379 
380 	if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
381 		if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
382 			(void) spa_history_log(spa, buf);
383 		spa_close(spa, FTAG);
384 	}
385 	history_str_free(buf);
386 }
387 
388 /*
389  * Policy for top-level read operations (list pools).  Requires no privileges,
390  * and can be used in the local zone, as there is no associated dataset.
391  */
392 /* ARGSUSED */
393 static int
394 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
395 {
396 	return (0);
397 }
398 
399 /*
400  * Policy for dataset read operations (list children, get statistics).  Requires
401  * no privileges, but must be visible in the local zone.
402  */
403 /* ARGSUSED */
404 static int
405 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
406 {
407 	if (INGLOBALZONE(curproc) ||
408 	    zone_dataset_visible(zc->zc_name, NULL))
409 		return (0);
410 
411 	return (SET_ERROR(ENOENT));
412 }
413 
414 static int
415 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
416 {
417 	int writable = 1;
418 
419 	/*
420 	 * The dataset must be visible by this zone -- check this first
421 	 * so they don't see EPERM on something they shouldn't know about.
422 	 */
423 	if (!INGLOBALZONE(curproc) &&
424 	    !zone_dataset_visible(dataset, &writable))
425 		return (SET_ERROR(ENOENT));
426 
427 	if (INGLOBALZONE(curproc)) {
428 		/*
429 		 * If the fs is zoned, only root can access it from the
430 		 * global zone.
431 		 */
432 		if (secpolicy_zfs(cr) && zoned)
433 			return (SET_ERROR(EPERM));
434 	} else {
435 		/*
436 		 * If we are in a local zone, the 'zoned' property must be set.
437 		 */
438 		if (!zoned)
439 			return (SET_ERROR(EPERM));
440 
441 		/* must be writable by this zone */
442 		if (!writable)
443 			return (SET_ERROR(EPERM));
444 	}
445 	return (0);
446 }
447 
448 static int
449 zfs_dozonecheck(const char *dataset, cred_t *cr)
450 {
451 	uint64_t zoned;
452 
453 	if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
454 		return (SET_ERROR(ENOENT));
455 
456 	return (zfs_dozonecheck_impl(dataset, zoned, cr));
457 }
458 
459 static int
460 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
461 {
462 	uint64_t zoned;
463 
464 	if (dsl_prop_get_int_ds(ds, "zoned", &zoned))
465 		return (SET_ERROR(ENOENT));
466 
467 	return (zfs_dozonecheck_impl(dataset, zoned, cr));
468 }
469 
470 static int
471 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
472     const char *perm, cred_t *cr)
473 {
474 	int error;
475 
476 	error = zfs_dozonecheck_ds(name, ds, cr);
477 	if (error == 0) {
478 		error = secpolicy_zfs(cr);
479 		if (error != 0)
480 			error = dsl_deleg_access_impl(ds, perm, cr);
481 	}
482 	return (error);
483 }
484 
485 static int
486 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
487 {
488 	int error;
489 	dsl_dataset_t *ds;
490 	dsl_pool_t *dp;
491 
492 	error = dsl_pool_hold(name, FTAG, &dp);
493 	if (error != 0)
494 		return (error);
495 
496 	error = dsl_dataset_hold(dp, name, FTAG, &ds);
497 	if (error != 0) {
498 		dsl_pool_rele(dp, FTAG);
499 		return (error);
500 	}
501 
502 	error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
503 
504 	dsl_dataset_rele(ds, FTAG);
505 	dsl_pool_rele(dp, FTAG);
506 	return (error);
507 }
508 
509 /*
510  * Policy for setting the security label property.
511  *
512  * Returns 0 for success, non-zero for access and other errors.
513  */
514 static int
515 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
516 {
517 	char		ds_hexsl[MAXNAMELEN];
518 	bslabel_t	ds_sl, new_sl;
519 	boolean_t	new_default = FALSE;
520 	uint64_t	zoned;
521 	int		needed_priv = -1;
522 	int		error;
523 
524 	/* First get the existing dataset label. */
525 	error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
526 	    1, sizeof (ds_hexsl), &ds_hexsl, NULL);
527 	if (error != 0)
528 		return (SET_ERROR(EPERM));
529 
530 	if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
531 		new_default = TRUE;
532 
533 	/* The label must be translatable */
534 	if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
535 		return (SET_ERROR(EINVAL));
536 
537 	/*
538 	 * In a non-global zone, disallow attempts to set a label that
539 	 * doesn't match that of the zone; otherwise no other checks
540 	 * are needed.
541 	 */
542 	if (!INGLOBALZONE(curproc)) {
543 		if (new_default || !blequal(&new_sl, CR_SL(CRED())))
544 			return (SET_ERROR(EPERM));
545 		return (0);
546 	}
547 
548 	/*
549 	 * For global-zone datasets (i.e., those whose zoned property is
550 	 * "off", verify that the specified new label is valid for the
551 	 * global zone.
552 	 */
553 	if (dsl_prop_get_integer(name,
554 	    zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
555 		return (SET_ERROR(EPERM));
556 	if (!zoned) {
557 		if (zfs_check_global_label(name, strval) != 0)
558 			return (SET_ERROR(EPERM));
559 	}
560 
561 	/*
562 	 * If the existing dataset label is nondefault, check if the
563 	 * dataset is mounted (label cannot be changed while mounted).
564 	 * Get the zfsvfs; if there isn't one, then the dataset isn't
565 	 * mounted (or isn't a dataset, doesn't exist, ...).
566 	 */
567 	if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
568 		objset_t *os;
569 		static char *setsl_tag = "setsl_tag";
570 
571 		/*
572 		 * Try to own the dataset; abort if there is any error,
573 		 * (e.g., already mounted, in use, or other error).
574 		 */
575 		error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
576 		    setsl_tag, &os);
577 		if (error != 0)
578 			return (SET_ERROR(EPERM));
579 
580 		dmu_objset_disown(os, setsl_tag);
581 
582 		if (new_default) {
583 			needed_priv = PRIV_FILE_DOWNGRADE_SL;
584 			goto out_check;
585 		}
586 
587 		if (hexstr_to_label(strval, &new_sl) != 0)
588 			return (SET_ERROR(EPERM));
589 
590 		if (blstrictdom(&ds_sl, &new_sl))
591 			needed_priv = PRIV_FILE_DOWNGRADE_SL;
592 		else if (blstrictdom(&new_sl, &ds_sl))
593 			needed_priv = PRIV_FILE_UPGRADE_SL;
594 	} else {
595 		/* dataset currently has a default label */
596 		if (!new_default)
597 			needed_priv = PRIV_FILE_UPGRADE_SL;
598 	}
599 
600 out_check:
601 	if (needed_priv != -1)
602 		return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
603 	return (0);
604 }
605 
606 static int
607 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
608     cred_t *cr)
609 {
610 	char *strval;
611 
612 	/*
613 	 * Check permissions for special properties.
614 	 */
615 	switch (prop) {
616 	case ZFS_PROP_ZONED:
617 		/*
618 		 * Disallow setting of 'zoned' from within a local zone.
619 		 */
620 		if (!INGLOBALZONE(curproc))
621 			return (SET_ERROR(EPERM));
622 		break;
623 
624 	case ZFS_PROP_QUOTA:
625 	case ZFS_PROP_FILESYSTEM_LIMIT:
626 	case ZFS_PROP_SNAPSHOT_LIMIT:
627 		if (!INGLOBALZONE(curproc)) {
628 			uint64_t zoned;
629 			char setpoint[MAXNAMELEN];
630 			/*
631 			 * Unprivileged users are allowed to modify the
632 			 * limit on things *under* (ie. contained by)
633 			 * the thing they own.
634 			 */
635 			if (dsl_prop_get_integer(dsname, "zoned", &zoned,
636 			    setpoint))
637 				return (SET_ERROR(EPERM));
638 			if (!zoned || strlen(dsname) <= strlen(setpoint))
639 				return (SET_ERROR(EPERM));
640 		}
641 		break;
642 
643 	case ZFS_PROP_MLSLABEL:
644 		if (!is_system_labeled())
645 			return (SET_ERROR(EPERM));
646 
647 		if (nvpair_value_string(propval, &strval) == 0) {
648 			int err;
649 
650 			err = zfs_set_slabel_policy(dsname, strval, CRED());
651 			if (err != 0)
652 				return (err);
653 		}
654 		break;
655 	}
656 
657 	return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
658 }
659 
660 /* ARGSUSED */
661 static int
662 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
663 {
664 	int error;
665 
666 	error = zfs_dozonecheck(zc->zc_name, cr);
667 	if (error != 0)
668 		return (error);
669 
670 	/*
671 	 * permission to set permissions will be evaluated later in
672 	 * dsl_deleg_can_allow()
673 	 */
674 	return (0);
675 }
676 
677 /* ARGSUSED */
678 static int
679 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
680 {
681 	return (zfs_secpolicy_write_perms(zc->zc_name,
682 	    ZFS_DELEG_PERM_ROLLBACK, cr));
683 }
684 
685 /* ARGSUSED */
686 static int
687 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
688 {
689 	dsl_pool_t *dp;
690 	dsl_dataset_t *ds;
691 	char *cp;
692 	int error;
693 
694 	/*
695 	 * Generate the current snapshot name from the given objsetid, then
696 	 * use that name for the secpolicy/zone checks.
697 	 */
698 	cp = strchr(zc->zc_name, '@');
699 	if (cp == NULL)
700 		return (SET_ERROR(EINVAL));
701 	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
702 	if (error != 0)
703 		return (error);
704 
705 	error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
706 	if (error != 0) {
707 		dsl_pool_rele(dp, FTAG);
708 		return (error);
709 	}
710 
711 	dsl_dataset_name(ds, zc->zc_name);
712 
713 	error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
714 	    ZFS_DELEG_PERM_SEND, cr);
715 	dsl_dataset_rele(ds, FTAG);
716 	dsl_pool_rele(dp, FTAG);
717 
718 	return (error);
719 }
720 
721 /* ARGSUSED */
722 static int
723 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
724 {
725 	return (zfs_secpolicy_write_perms(zc->zc_name,
726 	    ZFS_DELEG_PERM_SEND, cr));
727 }
728 
729 /* ARGSUSED */
730 static int
731 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
732 {
733 	vnode_t *vp;
734 	int error;
735 
736 	if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
737 	    NO_FOLLOW, NULL, &vp)) != 0)
738 		return (error);
739 
740 	/* Now make sure mntpnt and dataset are ZFS */
741 
742 	if (vp->v_vfsp->vfs_fstype != zfsfstype ||
743 	    (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
744 	    zc->zc_name) != 0)) {
745 		VN_RELE(vp);
746 		return (SET_ERROR(EPERM));
747 	}
748 
749 	VN_RELE(vp);
750 	return (dsl_deleg_access(zc->zc_name,
751 	    ZFS_DELEG_PERM_SHARE, cr));
752 }
753 
754 int
755 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
756 {
757 	if (!INGLOBALZONE(curproc))
758 		return (SET_ERROR(EPERM));
759 
760 	if (secpolicy_nfs(cr) == 0) {
761 		return (0);
762 	} else {
763 		return (zfs_secpolicy_deleg_share(zc, innvl, cr));
764 	}
765 }
766 
767 int
768 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
769 {
770 	if (!INGLOBALZONE(curproc))
771 		return (SET_ERROR(EPERM));
772 
773 	if (secpolicy_smb(cr) == 0) {
774 		return (0);
775 	} else {
776 		return (zfs_secpolicy_deleg_share(zc, innvl, cr));
777 	}
778 }
779 
780 static int
781 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
782 {
783 	char *cp;
784 
785 	/*
786 	 * Remove the @bla or /bla from the end of the name to get the parent.
787 	 */
788 	(void) strncpy(parent, datasetname, parentsize);
789 	cp = strrchr(parent, '@');
790 	if (cp != NULL) {
791 		cp[0] = '\0';
792 	} else {
793 		cp = strrchr(parent, '/');
794 		if (cp == NULL)
795 			return (SET_ERROR(ENOENT));
796 		cp[0] = '\0';
797 	}
798 
799 	return (0);
800 }
801 
802 int
803 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
804 {
805 	int error;
806 
807 	if ((error = zfs_secpolicy_write_perms(name,
808 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
809 		return (error);
810 
811 	return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
812 }
813 
814 /* ARGSUSED */
815 static int
816 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
817 {
818 	return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
819 }
820 
821 /*
822  * Destroying snapshots with delegated permissions requires
823  * descendant mount and destroy permissions.
824  */
825 /* ARGSUSED */
826 static int
827 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
828 {
829 	nvlist_t *snaps;
830 	nvpair_t *pair, *nextpair;
831 	int error = 0;
832 
833 	if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
834 		return (SET_ERROR(EINVAL));
835 	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
836 	    pair = nextpair) {
837 		nextpair = nvlist_next_nvpair(snaps, pair);
838 		error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
839 		if (error == ENOENT) {
840 			/*
841 			 * Ignore any snapshots that don't exist (we consider
842 			 * them "already destroyed").  Remove the name from the
843 			 * nvl here in case the snapshot is created between
844 			 * now and when we try to destroy it (in which case
845 			 * we don't want to destroy it since we haven't
846 			 * checked for permission).
847 			 */
848 			fnvlist_remove_nvpair(snaps, pair);
849 			error = 0;
850 		}
851 		if (error != 0)
852 			break;
853 	}
854 
855 	return (error);
856 }
857 
858 int
859 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
860 {
861 	char	parentname[MAXNAMELEN];
862 	int	error;
863 
864 	if ((error = zfs_secpolicy_write_perms(from,
865 	    ZFS_DELEG_PERM_RENAME, cr)) != 0)
866 		return (error);
867 
868 	if ((error = zfs_secpolicy_write_perms(from,
869 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
870 		return (error);
871 
872 	if ((error = zfs_get_parent(to, parentname,
873 	    sizeof (parentname))) != 0)
874 		return (error);
875 
876 	if ((error = zfs_secpolicy_write_perms(parentname,
877 	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
878 		return (error);
879 
880 	if ((error = zfs_secpolicy_write_perms(parentname,
881 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
882 		return (error);
883 
884 	return (error);
885 }
886 
887 /* ARGSUSED */
888 static int
889 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
890 {
891 	return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
892 }
893 
894 /* ARGSUSED */
895 static int
896 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
897 {
898 	dsl_pool_t *dp;
899 	dsl_dataset_t *clone;
900 	int error;
901 
902 	error = zfs_secpolicy_write_perms(zc->zc_name,
903 	    ZFS_DELEG_PERM_PROMOTE, cr);
904 	if (error != 0)
905 		return (error);
906 
907 	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
908 	if (error != 0)
909 		return (error);
910 
911 	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
912 
913 	if (error == 0) {
914 		char parentname[MAXNAMELEN];
915 		dsl_dataset_t *origin = NULL;
916 		dsl_dir_t *dd;
917 		dd = clone->ds_dir;
918 
919 		error = dsl_dataset_hold_obj(dd->dd_pool,
920 		    dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin);
921 		if (error != 0) {
922 			dsl_dataset_rele(clone, FTAG);
923 			dsl_pool_rele(dp, FTAG);
924 			return (error);
925 		}
926 
927 		error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
928 		    ZFS_DELEG_PERM_MOUNT, cr);
929 
930 		dsl_dataset_name(origin, parentname);
931 		if (error == 0) {
932 			error = zfs_secpolicy_write_perms_ds(parentname, origin,
933 			    ZFS_DELEG_PERM_PROMOTE, cr);
934 		}
935 		dsl_dataset_rele(clone, FTAG);
936 		dsl_dataset_rele(origin, FTAG);
937 	}
938 	dsl_pool_rele(dp, FTAG);
939 	return (error);
940 }
941 
942 /* ARGSUSED */
943 static int
944 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
945 {
946 	int error;
947 
948 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
949 	    ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
950 		return (error);
951 
952 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
953 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
954 		return (error);
955 
956 	return (zfs_secpolicy_write_perms(zc->zc_name,
957 	    ZFS_DELEG_PERM_CREATE, cr));
958 }
959 
960 int
961 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
962 {
963 	return (zfs_secpolicy_write_perms(name,
964 	    ZFS_DELEG_PERM_SNAPSHOT, cr));
965 }
966 
967 /*
968  * Check for permission to create each snapshot in the nvlist.
969  */
970 /* ARGSUSED */
971 static int
972 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
973 {
974 	nvlist_t *snaps;
975 	int error = 0;
976 	nvpair_t *pair;
977 
978 	if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
979 		return (SET_ERROR(EINVAL));
980 	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
981 	    pair = nvlist_next_nvpair(snaps, pair)) {
982 		char *name = nvpair_name(pair);
983 		char *atp = strchr(name, '@');
984 
985 		if (atp == NULL) {
986 			error = SET_ERROR(EINVAL);
987 			break;
988 		}
989 		*atp = '\0';
990 		error = zfs_secpolicy_snapshot_perms(name, cr);
991 		*atp = '@';
992 		if (error != 0)
993 			break;
994 	}
995 	return (error);
996 }
997 
998 /*
999  * Check for permission to create each snapshot in the nvlist.
1000  */
1001 /* ARGSUSED */
1002 static int
1003 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1004 {
1005 	int error = 0;
1006 
1007 	for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
1008 	    pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
1009 		char *name = nvpair_name(pair);
1010 		char *hashp = strchr(name, '#');
1011 
1012 		if (hashp == NULL) {
1013 			error = SET_ERROR(EINVAL);
1014 			break;
1015 		}
1016 		*hashp = '\0';
1017 		error = zfs_secpolicy_write_perms(name,
1018 		    ZFS_DELEG_PERM_BOOKMARK, cr);
1019 		*hashp = '#';
1020 		if (error != 0)
1021 			break;
1022 	}
1023 	return (error);
1024 }
1025 
1026 /* ARGSUSED */
1027 static int
1028 zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1029 {
1030 	nvpair_t *pair, *nextpair;
1031 	int error = 0;
1032 
1033 	for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1034 	    pair = nextpair) {
1035 		char *name = nvpair_name(pair);
1036 		char *hashp = strchr(name, '#');
1037 		nextpair = nvlist_next_nvpair(innvl, pair);
1038 
1039 		if (hashp == NULL) {
1040 			error = SET_ERROR(EINVAL);
1041 			break;
1042 		}
1043 
1044 		*hashp = '\0';
1045 		error = zfs_secpolicy_write_perms(name,
1046 		    ZFS_DELEG_PERM_DESTROY, cr);
1047 		*hashp = '#';
1048 		if (error == ENOENT) {
1049 			/*
1050 			 * Ignore any filesystems that don't exist (we consider
1051 			 * their bookmarks "already destroyed").  Remove
1052 			 * the name from the nvl here in case the filesystem
1053 			 * is created between now and when we try to destroy
1054 			 * the bookmark (in which case we don't want to
1055 			 * destroy it since we haven't checked for permission).
1056 			 */
1057 			fnvlist_remove_nvpair(innvl, pair);
1058 			error = 0;
1059 		}
1060 		if (error != 0)
1061 			break;
1062 	}
1063 
1064 	return (error);
1065 }
1066 
1067 /* ARGSUSED */
1068 static int
1069 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1070 {
1071 	/*
1072 	 * Even root must have a proper TSD so that we know what pool
1073 	 * to log to.
1074 	 */
1075 	if (tsd_get(zfs_allow_log_key) == NULL)
1076 		return (SET_ERROR(EPERM));
1077 	return (0);
1078 }
1079 
1080 static int
1081 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1082 {
1083 	char	parentname[MAXNAMELEN];
1084 	int	error;
1085 	char	*origin;
1086 
1087 	if ((error = zfs_get_parent(zc->zc_name, parentname,
1088 	    sizeof (parentname))) != 0)
1089 		return (error);
1090 
1091 	if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1092 	    (error = zfs_secpolicy_write_perms(origin,
1093 	    ZFS_DELEG_PERM_CLONE, cr)) != 0)
1094 		return (error);
1095 
1096 	if ((error = zfs_secpolicy_write_perms(parentname,
1097 	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
1098 		return (error);
1099 
1100 	return (zfs_secpolicy_write_perms(parentname,
1101 	    ZFS_DELEG_PERM_MOUNT, cr));
1102 }
1103 
1104 /*
1105  * Policy for pool operations - create/destroy pools, add vdevs, etc.  Requires
1106  * SYS_CONFIG privilege, which is not available in a local zone.
1107  */
1108 /* ARGSUSED */
1109 static int
1110 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1111 {
1112 	if (secpolicy_sys_config(cr, B_FALSE) != 0)
1113 		return (SET_ERROR(EPERM));
1114 
1115 	return (0);
1116 }
1117 
1118 /*
1119  * Policy for object to name lookups.
1120  */
1121 /* ARGSUSED */
1122 static int
1123 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1124 {
1125 	int error;
1126 
1127 	if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1128 		return (0);
1129 
1130 	error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1131 	return (error);
1132 }
1133 
1134 /*
1135  * Policy for fault injection.  Requires all privileges.
1136  */
1137 /* ARGSUSED */
1138 static int
1139 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1140 {
1141 	return (secpolicy_zinject(cr));
1142 }
1143 
1144 /* ARGSUSED */
1145 static int
1146 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1147 {
1148 	zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1149 
1150 	if (prop == ZPROP_INVAL) {
1151 		if (!zfs_prop_user(zc->zc_value))
1152 			return (SET_ERROR(EINVAL));
1153 		return (zfs_secpolicy_write_perms(zc->zc_name,
1154 		    ZFS_DELEG_PERM_USERPROP, cr));
1155 	} else {
1156 		return (zfs_secpolicy_setprop(zc->zc_name, prop,
1157 		    NULL, cr));
1158 	}
1159 }
1160 
1161 static int
1162 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1163 {
1164 	int err = zfs_secpolicy_read(zc, innvl, cr);
1165 	if (err)
1166 		return (err);
1167 
1168 	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1169 		return (SET_ERROR(EINVAL));
1170 
1171 	if (zc->zc_value[0] == 0) {
1172 		/*
1173 		 * They are asking about a posix uid/gid.  If it's
1174 		 * themself, allow it.
1175 		 */
1176 		if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1177 		    zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
1178 			if (zc->zc_guid == crgetuid(cr))
1179 				return (0);
1180 		} else {
1181 			if (groupmember(zc->zc_guid, cr))
1182 				return (0);
1183 		}
1184 	}
1185 
1186 	return (zfs_secpolicy_write_perms(zc->zc_name,
1187 	    userquota_perms[zc->zc_objset_type], cr));
1188 }
1189 
1190 static int
1191 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1192 {
1193 	int err = zfs_secpolicy_read(zc, innvl, cr);
1194 	if (err)
1195 		return (err);
1196 
1197 	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1198 		return (SET_ERROR(EINVAL));
1199 
1200 	return (zfs_secpolicy_write_perms(zc->zc_name,
1201 	    userquota_perms[zc->zc_objset_type], cr));
1202 }
1203 
1204 /* ARGSUSED */
1205 static int
1206 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1207 {
1208 	return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1209 	    NULL, cr));
1210 }
1211 
1212 /* ARGSUSED */
1213 static int
1214 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1215 {
1216 	nvpair_t *pair;
1217 	nvlist_t *holds;
1218 	int error;
1219 
1220 	error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1221 	if (error != 0)
1222 		return (SET_ERROR(EINVAL));
1223 
1224 	for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1225 	    pair = nvlist_next_nvpair(holds, pair)) {
1226 		char fsname[MAXNAMELEN];
1227 		error = dmu_fsname(nvpair_name(pair), fsname);
1228 		if (error != 0)
1229 			return (error);
1230 		error = zfs_secpolicy_write_perms(fsname,
1231 		    ZFS_DELEG_PERM_HOLD, cr);
1232 		if (error != 0)
1233 			return (error);
1234 	}
1235 	return (0);
1236 }
1237 
1238 /* ARGSUSED */
1239 static int
1240 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1241 {
1242 	nvpair_t *pair;
1243 	int error;
1244 
1245 	for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1246 	    pair = nvlist_next_nvpair(innvl, pair)) {
1247 		char fsname[MAXNAMELEN];
1248 		error = dmu_fsname(nvpair_name(pair), fsname);
1249 		if (error != 0)
1250 			return (error);
1251 		error = zfs_secpolicy_write_perms(fsname,
1252 		    ZFS_DELEG_PERM_RELEASE, cr);
1253 		if (error != 0)
1254 			return (error);
1255 	}
1256 	return (0);
1257 }
1258 
1259 /*
1260  * Policy for allowing temporary snapshots to be taken or released
1261  */
1262 static int
1263 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1264 {
1265 	/*
1266 	 * A temporary snapshot is the same as a snapshot,
1267 	 * hold, destroy and release all rolled into one.
1268 	 * Delegated diff alone is sufficient that we allow this.
1269 	 */
1270 	int error;
1271 
1272 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1273 	    ZFS_DELEG_PERM_DIFF, cr)) == 0)
1274 		return (0);
1275 
1276 	error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1277 	if (error == 0)
1278 		error = zfs_secpolicy_hold(zc, innvl, cr);
1279 	if (error == 0)
1280 		error = zfs_secpolicy_release(zc, innvl, cr);
1281 	if (error == 0)
1282 		error = zfs_secpolicy_destroy(zc, innvl, cr);
1283 	return (error);
1284 }
1285 
1286 /*
1287  * Policy for allowing setting the zev callback list.
1288  */
1289 static int
1290 zfs_secpolicy_set_zev_callbacks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1291 {
1292 	/* must be called from kernel context */
1293 	if (!(zc->zc_iflags & FKIOCTL))
1294 		return (SET_ERROR(EPERM));
1295 	/* callback pointer must be unset (set to default value) */
1296 	rw_enter(&rz_zev_rwlock, RW_READER);
1297 	if (rz_zev_callbacks != rz_zev_default_callbacks) {
1298 		rw_exit(&rz_zev_rwlock);
1299 		return (SET_ERROR(EBUSY));
1300 	}
1301 	rw_exit(&rz_zev_rwlock);
1302 	return (0);
1303 }
1304 
1305 /*
1306  * Policy for allowing unsetting/resetting the zev callback list.
1307  */
1308 static int
1309 zfs_secpolicy_unset_zev_callbacks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1310 {
1311 	/* must be called from kernel context */
1312 	if (!(zc->zc_iflags & FKIOCTL))
1313 		return (SET_ERROR(EPERM));
1314 	return (0);
1315 }
1316 
1317 /*
1318  * Returns the nvlist as specified by the user in the zfs_cmd_t.
1319  */
1320 static int
1321 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1322 {
1323 	char *packed;
1324 	int error;
1325 	nvlist_t *list = NULL;
1326 
1327 	/*
1328 	 * Read in and unpack the user-supplied nvlist.
1329 	 */
1330 	if (size == 0)
1331 		return (SET_ERROR(EINVAL));
1332 
1333 	packed = kmem_alloc(size, KM_SLEEP);
1334 
1335 	if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1336 	    iflag)) != 0) {
1337 		kmem_free(packed, size);
1338 		return (error);
1339 	}
1340 
1341 	if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1342 		kmem_free(packed, size);
1343 		return (error);
1344 	}
1345 
1346 	kmem_free(packed, size);
1347 
1348 	*nvp = list;
1349 	return (0);
1350 }
1351 
1352 /*
1353  * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1354  * Entries will be removed from the end of the nvlist, and one int32 entry
1355  * named "N_MORE_ERRORS" will be added indicating how many entries were
1356  * removed.
1357  */
1358 static int
1359 nvlist_smush(nvlist_t *errors, size_t max)
1360 {
1361 	size_t size;
1362 
1363 	size = fnvlist_size(errors);
1364 
1365 	if (size > max) {
1366 		nvpair_t *more_errors;
1367 		int n = 0;
1368 
1369 		if (max < 1024)
1370 			return (SET_ERROR(ENOMEM));
1371 
1372 		fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1373 		more_errors = nvlist_prev_nvpair(errors, NULL);
1374 
1375 		do {
1376 			nvpair_t *pair = nvlist_prev_nvpair(errors,
1377 			    more_errors);
1378 			fnvlist_remove_nvpair(errors, pair);
1379 			n++;
1380 			size = fnvlist_size(errors);
1381 		} while (size > max);
1382 
1383 		fnvlist_remove_nvpair(errors, more_errors);
1384 		fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1385 		ASSERT3U(fnvlist_size(errors), <=, max);
1386 	}
1387 
1388 	return (0);
1389 }
1390 
1391 static int
1392 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1393 {
1394 	char *packed = NULL;
1395 	int error = 0;
1396 	size_t size;
1397 
1398 	size = fnvlist_size(nvl);
1399 
1400 	if (size > zc->zc_nvlist_dst_size) {
1401 		error = SET_ERROR(ENOMEM);
1402 	} else {
1403 		packed = fnvlist_pack(nvl, &size);
1404 		if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1405 		    size, zc->zc_iflags) != 0)
1406 			error = SET_ERROR(EFAULT);
1407 		fnvlist_pack_free(packed, size);
1408 	}
1409 
1410 	zc->zc_nvlist_dst_size = size;
1411 	zc->zc_nvlist_dst_filled = B_TRUE;
1412 	return (error);
1413 }
1414 
1415 static int
1416 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1417 {
1418 	objset_t *os;
1419 	int error;
1420 
1421 	error = dmu_objset_hold(dsname, FTAG, &os);
1422 	if (error != 0)
1423 		return (error);
1424 	if (dmu_objset_type(os) != DMU_OST_ZFS) {
1425 		dmu_objset_rele(os, FTAG);
1426 		return (SET_ERROR(EINVAL));
1427 	}
1428 
1429 	mutex_enter(&os->os_user_ptr_lock);
1430 	*zfvp = dmu_objset_get_user(os);
1431 	if (*zfvp) {
1432 		VFS_HOLD((*zfvp)->z_vfs);
1433 	} else {
1434 		error = SET_ERROR(ESRCH);
1435 	}
1436 	mutex_exit(&os->os_user_ptr_lock);
1437 	dmu_objset_rele(os, FTAG);
1438 	return (error);
1439 }
1440 
1441 /*
1442  * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1443  * case its z_vfs will be NULL, and it will be opened as the owner.
1444  * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1445  * which prevents all vnode ops from running.
1446  */
1447 static int
1448 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1449 {
1450 	int error = 0;
1451 
1452 	if (getzfsvfs(name, zfvp) != 0)
1453 		error = zfsvfs_create(name, zfvp);
1454 	if (error == 0) {
1455 		rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1456 		    RW_READER, tag);
1457 		if ((*zfvp)->z_unmounted) {
1458 			/*
1459 			 * XXX we could probably try again, since the unmounting
1460 			 * thread should be just about to disassociate the
1461 			 * objset from the zfsvfs.
1462 			 */
1463 			rrm_exit(&(*zfvp)->z_teardown_lock, tag);
1464 			return (SET_ERROR(EBUSY));
1465 		}
1466 	}
1467 	return (error);
1468 }
1469 
1470 static void
1471 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1472 {
1473 	rrm_exit(&zfsvfs->z_teardown_lock, tag);
1474 
1475 	if (zfsvfs->z_vfs) {
1476 		VFS_RELE(zfsvfs->z_vfs);
1477 	} else {
1478 		dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1479 		zfsvfs_free(zfsvfs);
1480 	}
1481 }
1482 
1483 static int
1484 zfs_ioc_pool_create(zfs_cmd_t *zc)
1485 {
1486 	int error;
1487 	nvlist_t *config, *props = NULL;
1488 	nvlist_t *rootprops = NULL;
1489 	nvlist_t *zplprops = NULL;
1490 
1491 	if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1492 	    zc->zc_iflags, &config))
1493 		return (error);
1494 
1495 	if (zc->zc_nvlist_src_size != 0 && (error =
1496 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1497 	    zc->zc_iflags, &props))) {
1498 		nvlist_free(config);
1499 		return (error);
1500 	}
1501 
1502 	if (props) {
1503 		nvlist_t *nvl = NULL;
1504 		uint64_t version = SPA_VERSION;
1505 
1506 		(void) nvlist_lookup_uint64(props,
1507 		    zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1508 		if (!SPA_VERSION_IS_SUPPORTED(version)) {
1509 			error = SET_ERROR(EINVAL);
1510 			goto pool_props_bad;
1511 		}
1512 		(void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1513 		if (nvl) {
1514 			error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1515 			if (error != 0) {
1516 				nvlist_free(config);
1517 				nvlist_free(props);
1518 				return (error);
1519 			}
1520 			(void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1521 		}
1522 		VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1523 		error = zfs_fill_zplprops_root(version, rootprops,
1524 		    zplprops, NULL);
1525 		if (error != 0)
1526 			goto pool_props_bad;
1527 	}
1528 
1529 	error = spa_create(zc->zc_name, config, props, zplprops);
1530 
1531 	/*
1532 	 * Set the remaining root properties
1533 	 */
1534 	if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1535 	    ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1536 		(void) spa_destroy(zc->zc_name);
1537 
1538 pool_props_bad:
1539 	nvlist_free(rootprops);
1540 	nvlist_free(zplprops);
1541 	nvlist_free(config);
1542 	nvlist_free(props);
1543 
1544 	return (error);
1545 }
1546 
1547 static int
1548 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1549 {
1550 	int error;
1551 	zfs_log_history(zc);
1552 	error = spa_destroy(zc->zc_name);
1553 	if (error == 0)
1554 		zvol_remove_minors(zc->zc_name);
1555 	return (error);
1556 }
1557 
1558 static int
1559 zfs_ioc_pool_import(zfs_cmd_t *zc)
1560 {
1561 	nvlist_t *config, *props = NULL;
1562 	uint64_t guid;
1563 	int error;
1564 
1565 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1566 	    zc->zc_iflags, &config)) != 0)
1567 		return (error);
1568 
1569 	if (zc->zc_nvlist_src_size != 0 && (error =
1570 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1571 	    zc->zc_iflags, &props))) {
1572 		nvlist_free(config);
1573 		return (error);
1574 	}
1575 
1576 	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1577 	    guid != zc->zc_guid)
1578 		error = SET_ERROR(EINVAL);
1579 	else
1580 		error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1581 
1582 	if (zc->zc_nvlist_dst != 0) {
1583 		int err;
1584 
1585 		if ((err = put_nvlist(zc, config)) != 0)
1586 			error = err;
1587 	}
1588 
1589 	nvlist_free(config);
1590 
1591 	if (props)
1592 		nvlist_free(props);
1593 
1594 	return (error);
1595 }
1596 
1597 static int
1598 zfs_ioc_pool_export(zfs_cmd_t *zc)
1599 {
1600 	int error;
1601 	boolean_t force = (boolean_t)zc->zc_cookie;
1602 	boolean_t hardforce = (boolean_t)zc->zc_guid;
1603 
1604 	zfs_log_history(zc);
1605 	error = spa_export(zc->zc_name, NULL, force, hardforce);
1606 	if (error == 0)
1607 		zvol_remove_minors(zc->zc_name);
1608 	return (error);
1609 }
1610 
1611 static int
1612 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1613 {
1614 	nvlist_t *configs;
1615 	int error;
1616 
1617 	if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1618 		return (SET_ERROR(EEXIST));
1619 
1620 	error = put_nvlist(zc, configs);
1621 
1622 	nvlist_free(configs);
1623 
1624 	return (error);
1625 }
1626 
1627 /*
1628  * inputs:
1629  * zc_name		name of the pool
1630  *
1631  * outputs:
1632  * zc_cookie		real errno
1633  * zc_nvlist_dst	config nvlist
1634  * zc_nvlist_dst_size	size of config nvlist
1635  */
1636 static int
1637 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1638 {
1639 	nvlist_t *config;
1640 	int error;
1641 	int ret = 0;
1642 
1643 	error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1644 	    sizeof (zc->zc_value));
1645 
1646 	if (config != NULL) {
1647 		ret = put_nvlist(zc, config);
1648 		nvlist_free(config);
1649 
1650 		/*
1651 		 * The config may be present even if 'error' is non-zero.
1652 		 * In this case we return success, and preserve the real errno
1653 		 * in 'zc_cookie'.
1654 		 */
1655 		zc->zc_cookie = error;
1656 	} else {
1657 		ret = error;
1658 	}
1659 
1660 	return (ret);
1661 }
1662 
1663 /*
1664  * Try to import the given pool, returning pool stats as appropriate so that
1665  * user land knows which devices are available and overall pool health.
1666  */
1667 static int
1668 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1669 {
1670 	nvlist_t *tryconfig, *config;
1671 	int error;
1672 
1673 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1674 	    zc->zc_iflags, &tryconfig)) != 0)
1675 		return (error);
1676 
1677 	config = spa_tryimport(tryconfig);
1678 
1679 	nvlist_free(tryconfig);
1680 
1681 	if (config == NULL)
1682 		return (SET_ERROR(EINVAL));
1683 
1684 	error = put_nvlist(zc, config);
1685 	nvlist_free(config);
1686 
1687 	return (error);
1688 }
1689 
1690 /*
1691  * inputs:
1692  * zc_name              name of the pool
1693  * zc_cookie            scan func (pool_scan_func_t)
1694  */
1695 static int
1696 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1697 {
1698 	spa_t *spa;
1699 	int error;
1700 
1701 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1702 		return (error);
1703 
1704 	if (zc->zc_cookie == POOL_SCAN_NONE)
1705 		error = spa_scan_stop(spa);
1706 	else
1707 		error = spa_scan(spa, zc->zc_cookie);
1708 
1709 	spa_close(spa, FTAG);
1710 
1711 	return (error);
1712 }
1713 
1714 static int
1715 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1716 {
1717 	spa_t *spa;
1718 	int error;
1719 
1720 	error = spa_open(zc->zc_name, &spa, FTAG);
1721 	if (error == 0) {
1722 		spa_freeze(spa);
1723 		spa_close(spa, FTAG);
1724 	}
1725 	return (error);
1726 }
1727 
1728 static int
1729 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1730 {
1731 	spa_t *spa;
1732 	int error;
1733 
1734 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1735 		return (error);
1736 
1737 	if (zc->zc_cookie < spa_version(spa) ||
1738 	    !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1739 		spa_close(spa, FTAG);
1740 		return (SET_ERROR(EINVAL));
1741 	}
1742 
1743 	spa_upgrade(spa, zc->zc_cookie);
1744 	spa_close(spa, FTAG);
1745 
1746 	return (error);
1747 }
1748 
1749 static int
1750 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1751 {
1752 	spa_t *spa;
1753 	char *hist_buf;
1754 	uint64_t size;
1755 	int error;
1756 
1757 	if ((size = zc->zc_history_len) == 0)
1758 		return (SET_ERROR(EINVAL));
1759 
1760 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1761 		return (error);
1762 
1763 	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1764 		spa_close(spa, FTAG);
1765 		return (SET_ERROR(ENOTSUP));
1766 	}
1767 
1768 	hist_buf = kmem_alloc(size, KM_SLEEP);
1769 	if ((error = spa_history_get(spa, &zc->zc_history_offset,
1770 	    &zc->zc_history_len, hist_buf)) == 0) {
1771 		error = ddi_copyout(hist_buf,
1772 		    (void *)(uintptr_t)zc->zc_history,
1773 		    zc->zc_history_len, zc->zc_iflags);
1774 	}
1775 
1776 	spa_close(spa, FTAG);
1777 	kmem_free(hist_buf, size);
1778 	return (error);
1779 }
1780 
1781 static int
1782 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1783 {
1784 	spa_t *spa;
1785 	int error;
1786 
1787 	error = spa_open(zc->zc_name, &spa, FTAG);
1788 	if (error == 0) {
1789 		error = spa_change_guid(spa);
1790 		spa_close(spa, FTAG);
1791 	}
1792 	return (error);
1793 }
1794 
1795 static int
1796 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1797 {
1798 	return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1799 }
1800 
1801 /*
1802  * inputs:
1803  * zc_name		name of filesystem
1804  * zc_obj		object to find
1805  *
1806  * outputs:
1807  * zc_value		name of object
1808  */
1809 static int
1810 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1811 {
1812 	objset_t *os;
1813 	int error;
1814 
1815 	/* XXX reading from objset not owned */
1816 	if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1817 		return (error);
1818 	if (dmu_objset_type(os) != DMU_OST_ZFS) {
1819 		dmu_objset_rele(os, FTAG);
1820 		return (SET_ERROR(EINVAL));
1821 	}
1822 	error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1823 	    sizeof (zc->zc_value));
1824 	dmu_objset_rele(os, FTAG);
1825 
1826 	return (error);
1827 }
1828 
1829 /*
1830  * inputs:
1831  * zc_name		name of filesystem
1832  * zc_obj		object to find
1833  *
1834  * outputs:
1835  * zc_stat		stats on object
1836  * zc_value		path to object
1837  */
1838 static int
1839 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1840 {
1841 	objset_t *os;
1842 	int error;
1843 
1844 	/* XXX reading from objset not owned */
1845 	if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1846 		return (error);
1847 	if (dmu_objset_type(os) != DMU_OST_ZFS) {
1848 		dmu_objset_rele(os, FTAG);
1849 		return (SET_ERROR(EINVAL));
1850 	}
1851 	error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1852 	    sizeof (zc->zc_value));
1853 	dmu_objset_rele(os, FTAG);
1854 
1855 	return (error);
1856 }
1857 
1858 static int
1859 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1860 {
1861 	spa_t *spa;
1862 	int error;
1863 	nvlist_t *config, **l2cache, **spares;
1864 	uint_t nl2cache = 0, nspares = 0;
1865 
1866 	error = spa_open(zc->zc_name, &spa, FTAG);
1867 	if (error != 0)
1868 		return (error);
1869 
1870 	error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1871 	    zc->zc_iflags, &config);
1872 	(void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1873 	    &l2cache, &nl2cache);
1874 
1875 	(void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1876 	    &spares, &nspares);
1877 
1878 	/*
1879 	 * A root pool with concatenated devices is not supported.
1880 	 * Thus, can not add a device to a root pool.
1881 	 *
1882 	 * Intent log device can not be added to a rootpool because
1883 	 * during mountroot, zil is replayed, a seperated log device
1884 	 * can not be accessed during the mountroot time.
1885 	 *
1886 	 * l2cache and spare devices are ok to be added to a rootpool.
1887 	 */
1888 	if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1889 		nvlist_free(config);
1890 		spa_close(spa, FTAG);
1891 		return (SET_ERROR(EDOM));
1892 	}
1893 
1894 	if (error == 0) {
1895 		error = spa_vdev_add(spa, config);
1896 		nvlist_free(config);
1897 	}
1898 	spa_close(spa, FTAG);
1899 	return (error);
1900 }
1901 
1902 /*
1903  * inputs:
1904  * zc_name		name of the pool
1905  * zc_nvlist_conf	nvlist of devices to remove
1906  * zc_cookie		to stop the remove?
1907  */
1908 static int
1909 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1910 {
1911 	spa_t *spa;
1912 	int error;
1913 
1914 	error = spa_open(zc->zc_name, &spa, FTAG);
1915 	if (error != 0)
1916 		return (error);
1917 	error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1918 	spa_close(spa, FTAG);
1919 	return (error);
1920 }
1921 
1922 static int
1923 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1924 {
1925 	spa_t *spa;
1926 	int error;
1927 	vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1928 
1929 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1930 		return (error);
1931 	switch (zc->zc_cookie) {
1932 	case VDEV_STATE_ONLINE:
1933 		error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1934 		break;
1935 
1936 	case VDEV_STATE_OFFLINE:
1937 		error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1938 		break;
1939 
1940 	case VDEV_STATE_FAULTED:
1941 		if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1942 		    zc->zc_obj != VDEV_AUX_EXTERNAL)
1943 			zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1944 
1945 		error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1946 		break;
1947 
1948 	case VDEV_STATE_DEGRADED:
1949 		if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1950 		    zc->zc_obj != VDEV_AUX_EXTERNAL)
1951 			zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1952 
1953 		error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1954 		break;
1955 
1956 	default:
1957 		error = SET_ERROR(EINVAL);
1958 	}
1959 	zc->zc_cookie = newstate;
1960 	spa_close(spa, FTAG);
1961 	return (error);
1962 }
1963 
1964 static int
1965 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1966 {
1967 	spa_t *spa;
1968 	int replacing = zc->zc_cookie;
1969 	nvlist_t *config;
1970 	int error;
1971 
1972 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1973 		return (error);
1974 
1975 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1976 	    zc->zc_iflags, &config)) == 0) {
1977 		error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1978 		nvlist_free(config);
1979 	}
1980 
1981 	spa_close(spa, FTAG);
1982 	return (error);
1983 }
1984 
1985 static int
1986 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1987 {
1988 	spa_t *spa;
1989 	int error;
1990 
1991 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1992 		return (error);
1993 
1994 	error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1995 
1996 	spa_close(spa, FTAG);
1997 	return (error);
1998 }
1999 
2000 static int
2001 zfs_ioc_vdev_split(zfs_cmd_t *zc)
2002 {
2003 	spa_t *spa;
2004 	nvlist_t *config, *props = NULL;
2005 	int error;
2006 	boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
2007 
2008 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2009 		return (error);
2010 
2011 	if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
2012 	    zc->zc_iflags, &config)) {
2013 		spa_close(spa, FTAG);
2014 		return (error);
2015 	}
2016 
2017 	if (zc->zc_nvlist_src_size != 0 && (error =
2018 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2019 	    zc->zc_iflags, &props))) {
2020 		spa_close(spa, FTAG);
2021 		nvlist_free(config);
2022 		return (error);
2023 	}
2024 
2025 	error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
2026 
2027 	spa_close(spa, FTAG);
2028 
2029 	nvlist_free(config);
2030 	nvlist_free(props);
2031 
2032 	return (error);
2033 }
2034 
2035 static int
2036 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
2037 {
2038 	spa_t *spa;
2039 	char *path = zc->zc_value;
2040 	uint64_t guid = zc->zc_guid;
2041 	int error;
2042 
2043 	error = spa_open(zc->zc_name, &spa, FTAG);
2044 	if (error != 0)
2045 		return (error);
2046 
2047 	error = spa_vdev_setpath(spa, guid, path);
2048 	spa_close(spa, FTAG);
2049 	return (error);
2050 }
2051 
2052 static int
2053 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
2054 {
2055 	spa_t *spa;
2056 	char *fru = zc->zc_value;
2057 	uint64_t guid = zc->zc_guid;
2058 	int error;
2059 
2060 	error = spa_open(zc->zc_name, &spa, FTAG);
2061 	if (error != 0)
2062 		return (error);
2063 
2064 	error = spa_vdev_setfru(spa, guid, fru);
2065 	spa_close(spa, FTAG);
2066 	return (error);
2067 }
2068 
2069 static int
2070 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
2071 {
2072 	int error = 0;
2073 	nvlist_t *nv;
2074 
2075 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2076 
2077 	if (zc->zc_nvlist_dst != 0 &&
2078 	    (error = dsl_prop_get_all(os, &nv)) == 0) {
2079 		dmu_objset_stats(os, nv);
2080 		/*
2081 		 * NB: zvol_get_stats() will read the objset contents,
2082 		 * which we aren't supposed to do with a
2083 		 * DS_MODE_USER hold, because it could be
2084 		 * inconsistent.  So this is a bit of a workaround...
2085 		 * XXX reading with out owning
2086 		 */
2087 		if (!zc->zc_objset_stats.dds_inconsistent &&
2088 		    dmu_objset_type(os) == DMU_OST_ZVOL) {
2089 			error = zvol_get_stats(os, nv);
2090 			if (error == EIO)
2091 				return (error);
2092 			VERIFY0(error);
2093 		}
2094 		error = put_nvlist(zc, nv);
2095 		nvlist_free(nv);
2096 	}
2097 
2098 	return (error);
2099 }
2100 
2101 /*
2102  * inputs:
2103  * zc_name		name of filesystem
2104  * zc_nvlist_dst_size	size of buffer for property nvlist
2105  *
2106  * outputs:
2107  * zc_objset_stats	stats
2108  * zc_nvlist_dst	property nvlist
2109  * zc_nvlist_dst_size	size of property nvlist
2110  */
2111 static int
2112 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2113 {
2114 	objset_t *os;
2115 	int error;
2116 
2117 	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2118 	if (error == 0) {
2119 		error = zfs_ioc_objset_stats_impl(zc, os);
2120 		dmu_objset_rele(os, FTAG);
2121 	}
2122 
2123 	return (error);
2124 }
2125 
2126 /*
2127  * inputs:
2128  * zc_name		name of filesystem
2129  * zc_nvlist_dst_size	size of buffer for property nvlist
2130  *
2131  * outputs:
2132  * zc_nvlist_dst	received property nvlist
2133  * zc_nvlist_dst_size	size of received property nvlist
2134  *
2135  * Gets received properties (distinct from local properties on or after
2136  * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2137  * local property values.
2138  */
2139 static int
2140 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2141 {
2142 	int error = 0;
2143 	nvlist_t *nv;
2144 
2145 	/*
2146 	 * Without this check, we would return local property values if the
2147 	 * caller has not already received properties on or after
2148 	 * SPA_VERSION_RECVD_PROPS.
2149 	 */
2150 	if (!dsl_prop_get_hasrecvd(zc->zc_name))
2151 		return (SET_ERROR(ENOTSUP));
2152 
2153 	if (zc->zc_nvlist_dst != 0 &&
2154 	    (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2155 		error = put_nvlist(zc, nv);
2156 		nvlist_free(nv);
2157 	}
2158 
2159 	return (error);
2160 }
2161 
2162 static int
2163 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2164 {
2165 	uint64_t value;
2166 	int error;
2167 
2168 	/*
2169 	 * zfs_get_zplprop() will either find a value or give us
2170 	 * the default value (if there is one).
2171 	 */
2172 	if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2173 		return (error);
2174 	VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2175 	return (0);
2176 }
2177 
2178 /*
2179  * inputs:
2180  * zc_name		name of filesystem
2181  * zc_nvlist_dst_size	size of buffer for zpl property nvlist
2182  *
2183  * outputs:
2184  * zc_nvlist_dst	zpl property nvlist
2185  * zc_nvlist_dst_size	size of zpl property nvlist
2186  */
2187 static int
2188 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2189 {
2190 	objset_t *os;
2191 	int err;
2192 
2193 	/* XXX reading without owning */
2194 	if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2195 		return (err);
2196 
2197 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2198 
2199 	/*
2200 	 * NB: nvl_add_zplprop() will read the objset contents,
2201 	 * which we aren't supposed to do with a DS_MODE_USER
2202 	 * hold, because it could be inconsistent.
2203 	 */
2204 	if (zc->zc_nvlist_dst != NULL &&
2205 	    !zc->zc_objset_stats.dds_inconsistent &&
2206 	    dmu_objset_type(os) == DMU_OST_ZFS) {
2207 		nvlist_t *nv;
2208 
2209 		VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2210 		if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2211 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2212 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2213 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2214 			err = put_nvlist(zc, nv);
2215 		nvlist_free(nv);
2216 	} else {
2217 		err = SET_ERROR(ENOENT);
2218 	}
2219 	dmu_objset_rele(os, FTAG);
2220 	return (err);
2221 }
2222 
2223 static boolean_t
2224 dataset_name_hidden(const char *name)
2225 {
2226 	/*
2227 	 * Skip over datasets that are not visible in this zone,
2228 	 * internal datasets (which have a $ in their name), and
2229 	 * temporary datasets (which have a % in their name).
2230 	 */
2231 	if (strchr(name, '$') != NULL)
2232 		return (B_TRUE);
2233 	if (strchr(name, '%') != NULL)
2234 		return (B_TRUE);
2235 	if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
2236 		return (B_TRUE);
2237 	return (B_FALSE);
2238 }
2239 
2240 /*
2241  * inputs:
2242  * zc_name		name of filesystem
2243  * zc_cookie		zap cursor
2244  * zc_nvlist_dst_size	size of buffer for property nvlist
2245  *
2246  * outputs:
2247  * zc_name		name of next filesystem
2248  * zc_cookie		zap cursor
2249  * zc_objset_stats	stats
2250  * zc_nvlist_dst	property nvlist
2251  * zc_nvlist_dst_size	size of property nvlist
2252  */
2253 static int
2254 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2255 {
2256 	objset_t *os;
2257 	int error;
2258 	char *p;
2259 	size_t orig_len = strlen(zc->zc_name);
2260 
2261 top:
2262 	if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2263 		if (error == ENOENT)
2264 			error = SET_ERROR(ESRCH);
2265 		return (error);
2266 	}
2267 
2268 	p = strrchr(zc->zc_name, '/');
2269 	if (p == NULL || p[1] != '\0')
2270 		(void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2271 	p = zc->zc_name + strlen(zc->zc_name);
2272 
2273 	do {
2274 		error = dmu_dir_list_next(os,
2275 		    sizeof (zc->zc_name) - (p - zc->zc_name), p,
2276 		    NULL, &zc->zc_cookie);
2277 		if (error == ENOENT)
2278 			error = SET_ERROR(ESRCH);
2279 	} while (error == 0 && dataset_name_hidden(zc->zc_name));
2280 	dmu_objset_rele(os, FTAG);
2281 
2282 	/*
2283 	 * If it's an internal dataset (ie. with a '$' in its name),
2284 	 * don't try to get stats for it, otherwise we'll return ENOENT.
2285 	 */
2286 	if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2287 		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2288 		if (error == ENOENT) {
2289 			/* We lost a race with destroy, get the next one. */
2290 			zc->zc_name[orig_len] = '\0';
2291 			goto top;
2292 		}
2293 	}
2294 	return (error);
2295 }
2296 
2297 /*
2298  * inputs:
2299  * zc_name		name of filesystem
2300  * zc_cookie		zap cursor
2301  * zc_nvlist_dst_size	size of buffer for property nvlist
2302  *
2303  * outputs:
2304  * zc_name		name of next snapshot
2305  * zc_objset_stats	stats
2306  * zc_nvlist_dst	property nvlist
2307  * zc_nvlist_dst_size	size of property nvlist
2308  */
2309 static int
2310 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2311 {
2312 	objset_t *os;
2313 	int error;
2314 
2315 	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2316 	if (error != 0) {
2317 		return (error == ENOENT ? ESRCH : error);
2318 	}
2319 
2320 	/*
2321 	 * A dataset name of maximum length cannot have any snapshots,
2322 	 * so exit immediately.
2323 	 */
2324 	if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2325 		dmu_objset_rele(os, FTAG);
2326 		return (SET_ERROR(ESRCH));
2327 	}
2328 
2329 	error = dmu_snapshot_list_next(os,
2330 	    sizeof (zc->zc_name) - strlen(zc->zc_name),
2331 	    zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2332 	    NULL);
2333 
2334 	if (error == 0) {
2335 		dsl_dataset_t *ds;
2336 		dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2337 
2338 		error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2339 		if (error == 0) {
2340 			objset_t *ossnap;
2341 
2342 			error = dmu_objset_from_ds(ds, &ossnap);
2343 			if (error == 0)
2344 				error = zfs_ioc_objset_stats_impl(zc, ossnap);
2345 			dsl_dataset_rele(ds, FTAG);
2346 		}
2347 	} else if (error == ENOENT) {
2348 		error = SET_ERROR(ESRCH);
2349 	}
2350 
2351 	dmu_objset_rele(os, FTAG);
2352 	/* if we failed, undo the @ that we tacked on to zc_name */
2353 	if (error != 0)
2354 		*strchr(zc->zc_name, '@') = '\0';
2355 	return (error);
2356 }
2357 
2358 static int
2359 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2360 {
2361 	const char *propname = nvpair_name(pair);
2362 	uint64_t *valary;
2363 	unsigned int vallen;
2364 	const char *domain;
2365 	char *dash;
2366 	zfs_userquota_prop_t type;
2367 	uint64_t rid;
2368 	uint64_t quota;
2369 	zfsvfs_t *zfsvfs;
2370 	int err;
2371 
2372 	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2373 		nvlist_t *attrs;
2374 		VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2375 		if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2376 		    &pair) != 0)
2377 			return (SET_ERROR(EINVAL));
2378 	}
2379 
2380 	/*
2381 	 * A correctly constructed propname is encoded as
2382 	 * userquota@<rid>-<domain>.
2383 	 */
2384 	if ((dash = strchr(propname, '-')) == NULL ||
2385 	    nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2386 	    vallen != 3)
2387 		return (SET_ERROR(EINVAL));
2388 
2389 	domain = dash + 1;
2390 	type = valary[0];
2391 	rid = valary[1];
2392 	quota = valary[2];
2393 
2394 	err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2395 	if (err == 0) {
2396 		err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2397 		zfsvfs_rele(zfsvfs, FTAG);
2398 	}
2399 
2400 	return (err);
2401 }
2402 
2403 /*
2404  * If the named property is one that has a special function to set its value,
2405  * return 0 on success and a positive error code on failure; otherwise if it is
2406  * not one of the special properties handled by this function, return -1.
2407  *
2408  * XXX: It would be better for callers of the property interface if we handled
2409  * these special cases in dsl_prop.c (in the dsl layer).
2410  */
2411 static int
2412 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2413     nvpair_t *pair)
2414 {
2415 	const char *propname = nvpair_name(pair);
2416 	zfs_prop_t prop = zfs_name_to_prop(propname);
2417 	uint64_t intval;
2418 	int err = -1;
2419 
2420 	if (prop == ZPROP_INVAL) {
2421 		if (zfs_prop_userquota(propname))
2422 			return (zfs_prop_set_userquota(dsname, pair));
2423 		return (-1);
2424 	}
2425 
2426 	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2427 		nvlist_t *attrs;
2428 		VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2429 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2430 		    &pair) == 0);
2431 	}
2432 
2433 	if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2434 		return (-1);
2435 
2436 	VERIFY(0 == nvpair_value_uint64(pair, &intval));
2437 
2438 	switch (prop) {
2439 	case ZFS_PROP_QUOTA:
2440 		err = dsl_dir_set_quota(dsname, source, intval);
2441 		break;
2442 	case ZFS_PROP_REFQUOTA:
2443 		err = dsl_dataset_set_refquota(dsname, source, intval);
2444 		break;
2445 	case ZFS_PROP_FILESYSTEM_LIMIT:
2446 	case ZFS_PROP_SNAPSHOT_LIMIT:
2447 		if (intval == UINT64_MAX) {
2448 			/* clearing the limit, just do it */
2449 			err = 0;
2450 		} else {
2451 			err = dsl_dir_activate_fs_ss_limit(dsname);
2452 		}
2453 		/*
2454 		 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2455 		 * default path to set the value in the nvlist.
2456 		 */
2457 		if (err == 0)
2458 			err = -1;
2459 		break;
2460 	case ZFS_PROP_RESERVATION:
2461 		err = dsl_dir_set_reservation(dsname, source, intval);
2462 		break;
2463 	case ZFS_PROP_REFRESERVATION:
2464 		err = dsl_dataset_set_refreservation(dsname, source, intval);
2465 		break;
2466 	case ZFS_PROP_VOLSIZE:
2467 		err = zvol_set_volsize(dsname, intval);
2468 		break;
2469 	case ZFS_PROP_VERSION:
2470 	{
2471 		zfsvfs_t *zfsvfs;
2472 
2473 		if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2474 			break;
2475 
2476 		err = zfs_set_version(zfsvfs, intval);
2477 		zfsvfs_rele(zfsvfs, FTAG);
2478 
2479 		if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2480 			zfs_cmd_t *zc;
2481 
2482 			zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2483 			(void) strcpy(zc->zc_name, dsname);
2484 			(void) zfs_ioc_userspace_upgrade(zc);
2485 			kmem_free(zc, sizeof (zfs_cmd_t));
2486 		}
2487 		break;
2488 	}
2489 	default:
2490 		err = -1;
2491 	}
2492 
2493 	return (err);
2494 }
2495 
2496 /*
2497  * This function is best effort. If it fails to set any of the given properties,
2498  * it continues to set as many as it can and returns the last error
2499  * encountered. If the caller provides a non-NULL errlist, it will be filled in
2500  * with the list of names of all the properties that failed along with the
2501  * corresponding error numbers.
2502  *
2503  * If every property is set successfully, zero is returned and errlist is not
2504  * modified.
2505  */
2506 int
2507 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2508     nvlist_t *errlist)
2509 {
2510 	nvpair_t *pair;
2511 	nvpair_t *propval;
2512 	int rv = 0;
2513 	uint64_t intval;
2514 	char *strval;
2515 	nvlist_t *genericnvl = fnvlist_alloc();
2516 	nvlist_t *retrynvl = fnvlist_alloc();
2517 
2518 retry:
2519 	pair = NULL;
2520 	while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2521 		const char *propname = nvpair_name(pair);
2522 		zfs_prop_t prop = zfs_name_to_prop(propname);
2523 		int err = 0;
2524 
2525 		/* decode the property value */
2526 		propval = pair;
2527 		if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2528 			nvlist_t *attrs;
2529 			attrs = fnvpair_value_nvlist(pair);
2530 			if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2531 			    &propval) != 0)
2532 				err = SET_ERROR(EINVAL);
2533 		}
2534 
2535 		/* Validate value type */
2536 		if (err == 0 && prop == ZPROP_INVAL) {
2537 			if (zfs_prop_user(propname)) {
2538 				if (nvpair_type(propval) != DATA_TYPE_STRING)
2539 					err = SET_ERROR(EINVAL);
2540 			} else if (zfs_prop_userquota(propname)) {
2541 				if (nvpair_type(propval) !=
2542 				    DATA_TYPE_UINT64_ARRAY)
2543 					err = SET_ERROR(EINVAL);
2544 			} else {
2545 				err = SET_ERROR(EINVAL);
2546 			}
2547 		} else if (err == 0) {
2548 			if (nvpair_type(propval) == DATA_TYPE_STRING) {
2549 				if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2550 					err = SET_ERROR(EINVAL);
2551 			} else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2552 				const char *unused;
2553 
2554 				intval = fnvpair_value_uint64(propval);
2555 
2556 				switch (zfs_prop_get_type(prop)) {
2557 				case PROP_TYPE_NUMBER:
2558 					break;
2559 				case PROP_TYPE_STRING:
2560 					err = SET_ERROR(EINVAL);
2561 					break;
2562 				case PROP_TYPE_INDEX:
2563 					if (zfs_prop_index_to_string(prop,
2564 					    intval, &unused) != 0)
2565 						err = SET_ERROR(EINVAL);
2566 					break;
2567 				default:
2568 					cmn_err(CE_PANIC,
2569 					    "unknown property type");
2570 				}
2571 			} else {
2572 				err = SET_ERROR(EINVAL);
2573 			}
2574 		}
2575 
2576 		/* Validate permissions */
2577 		if (err == 0)
2578 			err = zfs_check_settable(dsname, pair, CRED());
2579 
2580 		if (err == 0) {
2581 			err = zfs_prop_set_special(dsname, source, pair);
2582 			if (err == -1) {
2583 				/*
2584 				 * For better performance we build up a list of
2585 				 * properties to set in a single transaction.
2586 				 */
2587 				err = nvlist_add_nvpair(genericnvl, pair);
2588 			} else if (err != 0 && nvl != retrynvl) {
2589 				/*
2590 				 * This may be a spurious error caused by
2591 				 * receiving quota and reservation out of order.
2592 				 * Try again in a second pass.
2593 				 */
2594 				err = nvlist_add_nvpair(retrynvl, pair);
2595 			}
2596 		}
2597 
2598 		if (err != 0) {
2599 			if (errlist != NULL)
2600 				fnvlist_add_int32(errlist, propname, err);
2601 			rv = err;
2602 		}
2603 	}
2604 
2605 	if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2606 		nvl = retrynvl;
2607 		goto retry;
2608 	}
2609 
2610 	if (!nvlist_empty(genericnvl) &&
2611 	    dsl_props_set(dsname, source, genericnvl) != 0) {
2612 		/*
2613 		 * If this fails, we still want to set as many properties as we
2614 		 * can, so try setting them individually.
2615 		 */
2616 		pair = NULL;
2617 		while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2618 			const char *propname = nvpair_name(pair);
2619 			int err = 0;
2620 
2621 			propval = pair;
2622 			if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2623 				nvlist_t *attrs;
2624 				attrs = fnvpair_value_nvlist(pair);
2625 				propval = fnvlist_lookup_nvpair(attrs,
2626 				    ZPROP_VALUE);
2627 			}
2628 
2629 			if (nvpair_type(propval) == DATA_TYPE_STRING) {
2630 				strval = fnvpair_value_string(propval);
2631 				err = dsl_prop_set_string(dsname, propname,
2632 				    source, strval);
2633 			} else {
2634 				intval = fnvpair_value_uint64(propval);
2635 				err = dsl_prop_set_int(dsname, propname, source,
2636 				    intval);
2637 			}
2638 
2639 			if (err != 0) {
2640 				if (errlist != NULL) {
2641 					fnvlist_add_int32(errlist, propname,
2642 					    err);
2643 				}
2644 				rv = err;
2645 			}
2646 		}
2647 	}
2648 	nvlist_free(genericnvl);
2649 	nvlist_free(retrynvl);
2650 
2651 	return (rv);
2652 }
2653 
2654 /*
2655  * Check that all the properties are valid user properties.
2656  */
2657 static int
2658 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2659 {
2660 	nvpair_t *pair = NULL;
2661 	int error = 0;
2662 
2663 	while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2664 		const char *propname = nvpair_name(pair);
2665 
2666 		if (!zfs_prop_user(propname) ||
2667 		    nvpair_type(pair) != DATA_TYPE_STRING)
2668 			return (SET_ERROR(EINVAL));
2669 
2670 		if (error = zfs_secpolicy_write_perms(fsname,
2671 		    ZFS_DELEG_PERM_USERPROP, CRED()))
2672 			return (error);
2673 
2674 		if (strlen(propname) >= ZAP_MAXNAMELEN)
2675 			return (SET_ERROR(ENAMETOOLONG));
2676 
2677 		if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
2678 			return (E2BIG);
2679 	}
2680 	return (0);
2681 }
2682 
2683 static void
2684 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2685 {
2686 	nvpair_t *pair;
2687 
2688 	VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2689 
2690 	pair = NULL;
2691 	while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2692 		if (nvlist_exists(skipped, nvpair_name(pair)))
2693 			continue;
2694 
2695 		VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2696 	}
2697 }
2698 
2699 static int
2700 clear_received_props(const char *dsname, nvlist_t *props,
2701     nvlist_t *skipped)
2702 {
2703 	int err = 0;
2704 	nvlist_t *cleared_props = NULL;
2705 	props_skip(props, skipped, &cleared_props);
2706 	if (!nvlist_empty(cleared_props)) {
2707 		/*
2708 		 * Acts on local properties until the dataset has received
2709 		 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2710 		 */
2711 		zprop_source_t flags = (ZPROP_SRC_NONE |
2712 		    (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2713 		err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2714 	}
2715 	nvlist_free(cleared_props);
2716 	return (err);
2717 }
2718 
2719 /*
2720  * inputs:
2721  * zc_name		name of filesystem
2722  * zc_value		name of property to set
2723  * zc_nvlist_src{_size}	nvlist of properties to apply
2724  * zc_cookie		received properties flag
2725  *
2726  * outputs:
2727  * zc_nvlist_dst{_size} error for each unapplied received property
2728  */
2729 static int
2730 zfs_ioc_set_prop(zfs_cmd_t *zc)
2731 {
2732 	nvlist_t *nvl;
2733 	boolean_t received = zc->zc_cookie;
2734 	zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2735 	    ZPROP_SRC_LOCAL);
2736 	nvlist_t *errors;
2737 	int error;
2738 
2739 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2740 	    zc->zc_iflags, &nvl)) != 0)
2741 		return (error);
2742 
2743 	if (received) {
2744 		nvlist_t *origprops;
2745 
2746 		if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2747 			(void) clear_received_props(zc->zc_name,
2748 			    origprops, nvl);
2749 			nvlist_free(origprops);
2750 		}
2751 
2752 		error = dsl_prop_set_hasrecvd(zc->zc_name);
2753 	}
2754 
2755 	errors = fnvlist_alloc();
2756 	if (error == 0)
2757 		error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2758 
2759 	if (zc->zc_nvlist_dst != NULL && errors != NULL) {
2760 		(void) put_nvlist(zc, errors);
2761 	}
2762 
2763 	nvlist_free(errors);
2764 	nvlist_free(nvl);
2765 	return (error);
2766 }
2767 
2768 /*
2769  * inputs:
2770  * zc_name		name of filesystem
2771  * zc_value		name of property to inherit
2772  * zc_cookie		revert to received value if TRUE
2773  *
2774  * outputs:		none
2775  */
2776 static int
2777 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2778 {
2779 	const char *propname = zc->zc_value;
2780 	zfs_prop_t prop = zfs_name_to_prop(propname);
2781 	boolean_t received = zc->zc_cookie;
2782 	zprop_source_t source = (received
2783 	    ? ZPROP_SRC_NONE		/* revert to received value, if any */
2784 	    : ZPROP_SRC_INHERITED);	/* explicitly inherit */
2785 
2786 	if (received) {
2787 		nvlist_t *dummy;
2788 		nvpair_t *pair;
2789 		zprop_type_t type;
2790 		int err;
2791 
2792 		/*
2793 		 * zfs_prop_set_special() expects properties in the form of an
2794 		 * nvpair with type info.
2795 		 */
2796 		if (prop == ZPROP_INVAL) {
2797 			if (!zfs_prop_user(propname))
2798 				return (SET_ERROR(EINVAL));
2799 
2800 			type = PROP_TYPE_STRING;
2801 		} else if (prop == ZFS_PROP_VOLSIZE ||
2802 		    prop == ZFS_PROP_VERSION) {
2803 			return (SET_ERROR(EINVAL));
2804 		} else {
2805 			type = zfs_prop_get_type(prop);
2806 		}
2807 
2808 		VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2809 
2810 		switch (type) {
2811 		case PROP_TYPE_STRING:
2812 			VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2813 			break;
2814 		case PROP_TYPE_NUMBER:
2815 		case PROP_TYPE_INDEX:
2816 			VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2817 			break;
2818 		default:
2819 			nvlist_free(dummy);
2820 			return (SET_ERROR(EINVAL));
2821 		}
2822 
2823 		pair = nvlist_next_nvpair(dummy, NULL);
2824 		err = zfs_prop_set_special(zc->zc_name, source, pair);
2825 		nvlist_free(dummy);
2826 		if (err != -1)
2827 			return (err); /* special property already handled */
2828 	} else {
2829 		/*
2830 		 * Only check this in the non-received case. We want to allow
2831 		 * 'inherit -S' to revert non-inheritable properties like quota
2832 		 * and reservation to the received or default values even though
2833 		 * they are not considered inheritable.
2834 		 */
2835 		if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2836 			return (SET_ERROR(EINVAL));
2837 	}
2838 
2839 	/* property name has been validated by zfs_secpolicy_inherit_prop() */
2840 	return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2841 }
2842 
2843 static int
2844 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2845 {
2846 	nvlist_t *props;
2847 	spa_t *spa;
2848 	int error;
2849 	nvpair_t *pair;
2850 
2851 	if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2852 	    zc->zc_iflags, &props))
2853 		return (error);
2854 
2855 	/*
2856 	 * If the only property is the configfile, then just do a spa_lookup()
2857 	 * to handle the faulted case.
2858 	 */
2859 	pair = nvlist_next_nvpair(props, NULL);
2860 	if (pair != NULL && strcmp(nvpair_name(pair),
2861 	    zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2862 	    nvlist_next_nvpair(props, pair) == NULL) {
2863 		mutex_enter(&spa_namespace_lock);
2864 		if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2865 			spa_configfile_set(spa, props, B_FALSE);
2866 			spa_config_sync(spa, B_FALSE, B_TRUE);
2867 		}
2868 		mutex_exit(&spa_namespace_lock);
2869 		if (spa != NULL) {
2870 			nvlist_free(props);
2871 			return (0);
2872 		}
2873 	}
2874 
2875 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2876 		nvlist_free(props);
2877 		return (error);
2878 	}
2879 
2880 	error = spa_prop_set(spa, props);
2881 
2882 	nvlist_free(props);
2883 	spa_close(spa, FTAG);
2884 
2885 	return (error);
2886 }
2887 
2888 static int
2889 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2890 {
2891 	spa_t *spa;
2892 	int error;
2893 	nvlist_t *nvp = NULL;
2894 
2895 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2896 		/*
2897 		 * If the pool is faulted, there may be properties we can still
2898 		 * get (such as altroot and cachefile), so attempt to get them
2899 		 * anyway.
2900 		 */
2901 		mutex_enter(&spa_namespace_lock);
2902 		if ((spa = spa_lookup(zc->zc_name)) != NULL)
2903 			error = spa_prop_get(spa, &nvp);
2904 		mutex_exit(&spa_namespace_lock);
2905 	} else {
2906 		error = spa_prop_get(spa, &nvp);
2907 		spa_close(spa, FTAG);
2908 	}
2909 
2910 	if (error == 0 && zc->zc_nvlist_dst != NULL)
2911 		error = put_nvlist(zc, nvp);
2912 	else
2913 		error = SET_ERROR(EFAULT);
2914 
2915 	nvlist_free(nvp);
2916 	return (error);
2917 }
2918 
2919 /*
2920  * inputs:
2921  * zc_name		name of filesystem
2922  * zc_nvlist_src{_size}	nvlist of delegated permissions
2923  * zc_perm_action	allow/unallow flag
2924  *
2925  * outputs:		none
2926  */
2927 static int
2928 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2929 {
2930 	int error;
2931 	nvlist_t *fsaclnv = NULL;
2932 
2933 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2934 	    zc->zc_iflags, &fsaclnv)) != 0)
2935 		return (error);
2936 
2937 	/*
2938 	 * Verify nvlist is constructed correctly
2939 	 */
2940 	if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2941 		nvlist_free(fsaclnv);
2942 		return (SET_ERROR(EINVAL));
2943 	}
2944 
2945 	/*
2946 	 * If we don't have PRIV_SYS_MOUNT, then validate
2947 	 * that user is allowed to hand out each permission in
2948 	 * the nvlist(s)
2949 	 */
2950 
2951 	error = secpolicy_zfs(CRED());
2952 	if (error != 0) {
2953 		if (zc->zc_perm_action == B_FALSE) {
2954 			error = dsl_deleg_can_allow(zc->zc_name,
2955 			    fsaclnv, CRED());
2956 		} else {
2957 			error = dsl_deleg_can_unallow(zc->zc_name,
2958 			    fsaclnv, CRED());
2959 		}
2960 	}
2961 
2962 	if (error == 0)
2963 		error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2964 
2965 	nvlist_free(fsaclnv);
2966 	return (error);
2967 }
2968 
2969 /*
2970  * inputs:
2971  * zc_name		name of filesystem
2972  *
2973  * outputs:
2974  * zc_nvlist_src{_size}	nvlist of delegated permissions
2975  */
2976 static int
2977 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2978 {
2979 	nvlist_t *nvp;
2980 	int error;
2981 
2982 	if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2983 		error = put_nvlist(zc, nvp);
2984 		nvlist_free(nvp);
2985 	}
2986 
2987 	return (error);
2988 }
2989 
2990 /*
2991  * Search the vfs list for a specified resource.  Returns a pointer to it
2992  * or NULL if no suitable entry is found. The caller of this routine
2993  * is responsible for releasing the returned vfs pointer.
2994  */
2995 static vfs_t *
2996 zfs_get_vfs(const char *resource)
2997 {
2998 	struct vfs *vfsp;
2999 	struct vfs *vfs_found = NULL;
3000 
3001 	vfs_list_read_lock();
3002 	vfsp = rootvfs;
3003 	do {
3004 		if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
3005 			VFS_HOLD(vfsp);
3006 			vfs_found = vfsp;
3007 			break;
3008 		}
3009 		vfsp = vfsp->vfs_next;
3010 	} while (vfsp != rootvfs);
3011 	vfs_list_unlock();
3012 	return (vfs_found);
3013 }
3014 
3015 /* ARGSUSED */
3016 static void
3017 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
3018 {
3019 	zfs_creat_t *zct = arg;
3020 
3021 	zfs_create_fs(os, cr, zct->zct_zplprops, tx);
3022 }
3023 
3024 #define	ZFS_PROP_UNDEFINED	((uint64_t)-1)
3025 
3026 /*
3027  * inputs:
3028  * os			parent objset pointer (NULL if root fs)
3029  * fuids_ok		fuids allowed in this version of the spa?
3030  * sa_ok		SAs allowed in this version of the spa?
3031  * createprops		list of properties requested by creator
3032  *
3033  * outputs:
3034  * zplprops	values for the zplprops we attach to the master node object
3035  * is_ci	true if requested file system will be purely case-insensitive
3036  *
3037  * Determine the settings for utf8only, normalization and
3038  * casesensitivity.  Specific values may have been requested by the
3039  * creator and/or we can inherit values from the parent dataset.  If
3040  * the file system is of too early a vintage, a creator can not
3041  * request settings for these properties, even if the requested
3042  * setting is the default value.  We don't actually want to create dsl
3043  * properties for these, so remove them from the source nvlist after
3044  * processing.
3045  */
3046 static int
3047 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3048     boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3049     nvlist_t *zplprops, boolean_t *is_ci)
3050 {
3051 	uint64_t sense = ZFS_PROP_UNDEFINED;
3052 	uint64_t norm = ZFS_PROP_UNDEFINED;
3053 	uint64_t u8 = ZFS_PROP_UNDEFINED;
3054 
3055 	ASSERT(zplprops != NULL);
3056 
3057 	/*
3058 	 * Pull out creator prop choices, if any.
3059 	 */
3060 	if (createprops) {
3061 		(void) nvlist_lookup_uint64(createprops,
3062 		    zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3063 		(void) nvlist_lookup_uint64(createprops,
3064 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3065 		(void) nvlist_remove_all(createprops,
3066 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3067 		(void) nvlist_lookup_uint64(createprops,
3068 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3069 		(void) nvlist_remove_all(createprops,
3070 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3071 		(void) nvlist_lookup_uint64(createprops,
3072 		    zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3073 		(void) nvlist_remove_all(createprops,
3074 		    zfs_prop_to_name(ZFS_PROP_CASE));
3075 	}
3076 
3077 	/*
3078 	 * If the zpl version requested is whacky or the file system
3079 	 * or pool is version is too "young" to support normalization
3080 	 * and the creator tried to set a value for one of the props,
3081 	 * error out.
3082 	 */
3083 	if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3084 	    (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3085 	    (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3086 	    (zplver < ZPL_VERSION_NORMALIZATION &&
3087 	    (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3088 	    sense != ZFS_PROP_UNDEFINED)))
3089 		return (SET_ERROR(ENOTSUP));
3090 
3091 	/*
3092 	 * Put the version in the zplprops
3093 	 */
3094 	VERIFY(nvlist_add_uint64(zplprops,
3095 	    zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3096 
3097 	if (norm == ZFS_PROP_UNDEFINED)
3098 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3099 	VERIFY(nvlist_add_uint64(zplprops,
3100 	    zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3101 
3102 	/*
3103 	 * If we're normalizing, names must always be valid UTF-8 strings.
3104 	 */
3105 	if (norm)
3106 		u8 = 1;
3107 	if (u8 == ZFS_PROP_UNDEFINED)
3108 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3109 	VERIFY(nvlist_add_uint64(zplprops,
3110 	    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3111 
3112 	if (sense == ZFS_PROP_UNDEFINED)
3113 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3114 	VERIFY(nvlist_add_uint64(zplprops,
3115 	    zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3116 
3117 	if (is_ci)
3118 		*is_ci = (sense == ZFS_CASE_INSENSITIVE);
3119 
3120 	return (0);
3121 }
3122 
3123 static int
3124 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3125     nvlist_t *zplprops, boolean_t *is_ci)
3126 {
3127 	boolean_t fuids_ok, sa_ok;
3128 	uint64_t zplver = ZPL_VERSION;
3129 	objset_t *os = NULL;
3130 	char parentname[MAXNAMELEN];
3131 	char *cp;
3132 	spa_t *spa;
3133 	uint64_t spa_vers;
3134 	int error;
3135 
3136 	(void) strlcpy(parentname, dataset, sizeof (parentname));
3137 	cp = strrchr(parentname, '/');
3138 	ASSERT(cp != NULL);
3139 	cp[0] = '\0';
3140 
3141 	if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3142 		return (error);
3143 
3144 	spa_vers = spa_version(spa);
3145 	spa_close(spa, FTAG);
3146 
3147 	zplver = zfs_zpl_version_map(spa_vers);
3148 	fuids_ok = (zplver >= ZPL_VERSION_FUID);
3149 	sa_ok = (zplver >= ZPL_VERSION_SA);
3150 
3151 	/*
3152 	 * Open parent object set so we can inherit zplprop values.
3153 	 */
3154 	if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3155 		return (error);
3156 
3157 	error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3158 	    zplprops, is_ci);
3159 	dmu_objset_rele(os, FTAG);
3160 	return (error);
3161 }
3162 
3163 static int
3164 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3165     nvlist_t *zplprops, boolean_t *is_ci)
3166 {
3167 	boolean_t fuids_ok;
3168 	boolean_t sa_ok;
3169 	uint64_t zplver = ZPL_VERSION;
3170 	int error;
3171 
3172 	zplver = zfs_zpl_version_map(spa_vers);
3173 	fuids_ok = (zplver >= ZPL_VERSION_FUID);
3174 	sa_ok = (zplver >= ZPL_VERSION_SA);
3175 
3176 	error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3177 	    createprops, zplprops, is_ci);
3178 	return (error);
3179 }
3180 
3181 /*
3182  * innvl: {
3183  *     "type" -> dmu_objset_type_t (int32)
3184  *     (optional) "props" -> { prop -> value }
3185  * }
3186  *
3187  * outnvl: propname -> error code (int32)
3188  */
3189 static int
3190 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3191 {
3192 	int error = 0;
3193 	zfs_creat_t zct = { 0 };
3194 	nvlist_t *nvprops = NULL;
3195 	void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3196 	int32_t type32;
3197 	dmu_objset_type_t type;
3198 	boolean_t is_insensitive = B_FALSE;
3199 
3200 	if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3201 		return (SET_ERROR(EINVAL));
3202 	type = type32;
3203 	(void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3204 
3205 	switch (type) {
3206 	case DMU_OST_ZFS:
3207 		cbfunc = zfs_create_cb;
3208 		break;
3209 
3210 	case DMU_OST_ZVOL:
3211 		cbfunc = zvol_create_cb;
3212 		break;
3213 
3214 	default:
3215 		cbfunc = NULL;
3216 		break;
3217 	}
3218 	if (strchr(fsname, '@') ||
3219 	    strchr(fsname, '%'))
3220 		return (SET_ERROR(EINVAL));
3221 
3222 	zct.zct_props = nvprops;
3223 
3224 	if (cbfunc == NULL)
3225 		return (SET_ERROR(EINVAL));
3226 
3227 	if (type == DMU_OST_ZVOL) {
3228 		uint64_t volsize, volblocksize;
3229 
3230 		if (nvprops == NULL)
3231 			return (SET_ERROR(EINVAL));
3232 		if (nvlist_lookup_uint64(nvprops,
3233 		    zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3234 			return (SET_ERROR(EINVAL));
3235 
3236 		if ((error = nvlist_lookup_uint64(nvprops,
3237 		    zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3238 		    &volblocksize)) != 0 && error != ENOENT)
3239 			return (SET_ERROR(EINVAL));
3240 
3241 		if (error != 0)
3242 			volblocksize = zfs_prop_default_numeric(
3243 			    ZFS_PROP_VOLBLOCKSIZE);
3244 
3245 		if ((error = zvol_check_volblocksize(
3246 		    volblocksize)) != 0 ||
3247 		    (error = zvol_check_volsize(volsize,
3248 		    volblocksize)) != 0)
3249 			return (error);
3250 	} else if (type == DMU_OST_ZFS) {
3251 		int error;
3252 
3253 		/*
3254 		 * We have to have normalization and
3255 		 * case-folding flags correct when we do the
3256 		 * file system creation, so go figure them out
3257 		 * now.
3258 		 */
3259 		VERIFY(nvlist_alloc(&zct.zct_zplprops,
3260 		    NV_UNIQUE_NAME, KM_SLEEP) == 0);
3261 		error = zfs_fill_zplprops(fsname, nvprops,
3262 		    zct.zct_zplprops, &is_insensitive);
3263 		if (error != 0) {
3264 			nvlist_free(zct.zct_zplprops);
3265 			return (error);
3266 		}
3267 	}
3268 
3269 	error = dmu_objset_create(fsname, type,
3270 	    is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3271 	nvlist_free(zct.zct_zplprops);
3272 
3273 	/*
3274 	 * It would be nice to do this atomically.
3275 	 */
3276 	if (error == 0) {
3277 		error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3278 		    nvprops, outnvl);
3279 		if (error != 0)
3280 			(void) dsl_destroy_head(fsname);
3281 	}
3282 	return (error);
3283 }
3284 
3285 /*
3286  * innvl: {
3287  *     "origin" -> name of origin snapshot
3288  *     (optional) "props" -> { prop -> value }
3289  * }
3290  *
3291  * outnvl: propname -> error code (int32)
3292  */
3293 static int
3294 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3295 {
3296 	int error = 0;
3297 	nvlist_t *nvprops = NULL;
3298 	char *origin_name;
3299 
3300 	if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3301 		return (SET_ERROR(EINVAL));
3302 	(void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3303 
3304 	if (strchr(fsname, '@') ||
3305 	    strchr(fsname, '%'))
3306 		return (SET_ERROR(EINVAL));
3307 
3308 	if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3309 		return (SET_ERROR(EINVAL));
3310 	error = dmu_objset_clone(fsname, origin_name);
3311 	if (error != 0)
3312 		return (error);
3313 
3314 	/*
3315 	 * It would be nice to do this atomically.
3316 	 */
3317 	if (error == 0) {
3318 		error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3319 		    nvprops, outnvl);
3320 		if (error != 0)
3321 			(void) dsl_destroy_head(fsname);
3322 	}
3323 	return (error);
3324 }
3325 
3326 /*
3327  * innvl: {
3328  *     "snaps" -> { snapshot1, snapshot2 }
3329  *     (optional) "props" -> { prop -> value (string) }
3330  * }
3331  *
3332  * outnvl: snapshot -> error code (int32)
3333  */
3334 static int
3335 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3336 {
3337 	nvlist_t *snaps;
3338 	nvlist_t *props = NULL;
3339 	int error, poollen;
3340 	nvpair_t *pair;
3341 
3342 	(void) nvlist_lookup_nvlist(innvl, "props", &props);
3343 	if ((error = zfs_check_userprops(poolname, props)) != 0)
3344 		return (error);
3345 
3346 	if (!nvlist_empty(props) &&
3347 	    zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3348 		return (SET_ERROR(ENOTSUP));
3349 
3350 	if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3351 		return (SET_ERROR(EINVAL));
3352 	poollen = strlen(poolname);
3353 	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3354 	    pair = nvlist_next_nvpair(snaps, pair)) {
3355 		const char *name = nvpair_name(pair);
3356 		const char *cp = strchr(name, '@');
3357 
3358 		/*
3359 		 * The snap name must contain an @, and the part after it must
3360 		 * contain only valid characters.
3361 		 */
3362 		if (cp == NULL ||
3363 		    zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3364 			return (SET_ERROR(EINVAL));
3365 
3366 		/*
3367 		 * The snap must be in the specified pool.
3368 		 */
3369 		if (strncmp(name, poolname, poollen) != 0 ||
3370 		    (name[poollen] != '/' && name[poollen] != '@'))
3371 			return (SET_ERROR(EXDEV));
3372 
3373 		/* This must be the only snap of this fs. */
3374 		for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3375 		    pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3376 			if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3377 			    == 0) {
3378 				return (SET_ERROR(EXDEV));
3379 			}
3380 		}
3381 	}
3382 
3383 	error = dsl_dataset_snapshot(snaps, props, outnvl);
3384 	return (error);
3385 }
3386 
3387 /*
3388  * innvl: "message" -> string
3389  */
3390 /* ARGSUSED */
3391 static int
3392 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3393 {
3394 	char *message;
3395 	spa_t *spa;
3396 	int error;
3397 	char *poolname;
3398 
3399 	/*
3400 	 * The poolname in the ioctl is not set, we get it from the TSD,
3401 	 * which was set at the end of the last successful ioctl that allows
3402 	 * logging.  The secpolicy func already checked that it is set.
3403 	 * Only one log ioctl is allowed after each successful ioctl, so
3404 	 * we clear the TSD here.
3405 	 */
3406 	poolname = tsd_get(zfs_allow_log_key);
3407 	(void) tsd_set(zfs_allow_log_key, NULL);
3408 	error = spa_open(poolname, &spa, FTAG);
3409 	strfree(poolname);
3410 	if (error != 0)
3411 		return (error);
3412 
3413 	if (nvlist_lookup_string(innvl, "message", &message) != 0)  {
3414 		spa_close(spa, FTAG);
3415 		return (SET_ERROR(EINVAL));
3416 	}
3417 
3418 	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3419 		spa_close(spa, FTAG);
3420 		return (SET_ERROR(ENOTSUP));
3421 	}
3422 
3423 	error = spa_history_log(spa, message);
3424 	spa_close(spa, FTAG);
3425 	return (error);
3426 }
3427 
3428 /*
3429  * The dp_config_rwlock must not be held when calling this, because the
3430  * unmount may need to write out data.
3431  *
3432  * This function is best-effort.  Callers must deal gracefully if it
3433  * remains mounted (or is remounted after this call).
3434  *
3435  * Returns 0 if the argument is not a snapshot, or it is not currently a
3436  * filesystem, or we were able to unmount it.  Returns error code otherwise.
3437  */
3438 int
3439 zfs_unmount_snap(const char *snapname)
3440 {
3441 	vfs_t *vfsp;
3442 	zfsvfs_t *zfsvfs;
3443 	int err;
3444 
3445 	if (strchr(snapname, '@') == NULL)
3446 		return (0);
3447 
3448 	vfsp = zfs_get_vfs(snapname);
3449 	if (vfsp == NULL)
3450 		return (0);
3451 
3452 	zfsvfs = vfsp->vfs_data;
3453 	ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os)));
3454 
3455 	err = vn_vfswlock(vfsp->vfs_vnodecovered);
3456 	VFS_RELE(vfsp);
3457 	if (err != 0)
3458 		return (SET_ERROR(err));
3459 
3460 	/*
3461 	 * Always force the unmount for snapshots.
3462 	 */
3463 	(void) dounmount(vfsp, MS_FORCE, kcred);
3464 	return (0);
3465 }
3466 
3467 /* ARGSUSED */
3468 static int
3469 zfs_unmount_snap_cb(const char *snapname, void *arg)
3470 {
3471 	return (zfs_unmount_snap(snapname));
3472 }
3473 
3474 /*
3475  * When a clone is destroyed, its origin may also need to be destroyed,
3476  * in which case it must be unmounted.  This routine will do that unmount
3477  * if necessary.
3478  */
3479 void
3480 zfs_destroy_unmount_origin(const char *fsname)
3481 {
3482 	int error;
3483 	objset_t *os;
3484 	dsl_dataset_t *ds;
3485 
3486 	error = dmu_objset_hold(fsname, FTAG, &os);
3487 	if (error != 0)
3488 		return;
3489 	ds = dmu_objset_ds(os);
3490 	if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3491 		char originname[MAXNAMELEN];
3492 		dsl_dataset_name(ds->ds_prev, originname);
3493 		dmu_objset_rele(os, FTAG);
3494 		(void) zfs_unmount_snap(originname);
3495 	} else {
3496 		dmu_objset_rele(os, FTAG);
3497 	}
3498 }
3499 
3500 /*
3501  * innvl: {
3502  *     "snaps" -> { snapshot1, snapshot2 }
3503  *     (optional boolean) "defer"
3504  * }
3505  *
3506  * outnvl: snapshot -> error code (int32)
3507  *
3508  */
3509 /* ARGSUSED */
3510 static int
3511 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3512 {
3513 	nvlist_t *snaps;
3514 	nvpair_t *pair;
3515 	boolean_t defer;
3516 
3517 	if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3518 		return (SET_ERROR(EINVAL));
3519 	defer = nvlist_exists(innvl, "defer");
3520 
3521 	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3522 	    pair = nvlist_next_nvpair(snaps, pair)) {
3523 		(void) zfs_unmount_snap(nvpair_name(pair));
3524 	}
3525 
3526 	return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3527 }
3528 
3529 /*
3530  * Create bookmarks.  Bookmark names are of the form <fs>#<bmark>.
3531  * All bookmarks must be in the same pool.
3532  *
3533  * innvl: {
3534  *     bookmark1 -> snapshot1, bookmark2 -> snapshot2
3535  * }
3536  *
3537  * outnvl: bookmark -> error code (int32)
3538  *
3539  */
3540 /* ARGSUSED */
3541 static int
3542 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3543 {
3544 	for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3545 	    pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3546 		char *snap_name;
3547 
3548 		/*
3549 		 * Verify the snapshot argument.
3550 		 */
3551 		if (nvpair_value_string(pair, &snap_name) != 0)
3552 			return (SET_ERROR(EINVAL));
3553 
3554 
3555 		/* Verify that the keys (bookmarks) are unique */
3556 		for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair);
3557 		    pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) {
3558 			if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0)
3559 				return (SET_ERROR(EINVAL));
3560 		}
3561 	}
3562 
3563 	return (dsl_bookmark_create(innvl, outnvl));
3564 }
3565 
3566 /*
3567  * innvl: {
3568  *     property 1, property 2, ...
3569  * }
3570  *
3571  * outnvl: {
3572  *     bookmark name 1 -> { property 1, property 2, ... },
3573  *     bookmark name 2 -> { property 1, property 2, ... }
3574  * }
3575  *
3576  */
3577 static int
3578 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3579 {
3580 	return (dsl_get_bookmarks(fsname, innvl, outnvl));
3581 }
3582 
3583 /*
3584  * innvl: {
3585  *     bookmark name 1, bookmark name 2
3586  * }
3587  *
3588  * outnvl: bookmark -> error code (int32)
3589  *
3590  */
3591 static int
3592 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
3593     nvlist_t *outnvl)
3594 {
3595 	int error, poollen;
3596 
3597 	poollen = strlen(poolname);
3598 	for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3599 	    pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3600 		const char *name = nvpair_name(pair);
3601 		const char *cp = strchr(name, '#');
3602 
3603 		/*
3604 		 * The bookmark name must contain an #, and the part after it
3605 		 * must contain only valid characters.
3606 		 */
3607 		if (cp == NULL ||
3608 		    zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3609 			return (SET_ERROR(EINVAL));
3610 
3611 		/*
3612 		 * The bookmark must be in the specified pool.
3613 		 */
3614 		if (strncmp(name, poolname, poollen) != 0 ||
3615 		    (name[poollen] != '/' && name[poollen] != '#'))
3616 			return (SET_ERROR(EXDEV));
3617 	}
3618 
3619 	error = dsl_bookmark_destroy(innvl, outnvl);
3620 	return (error);
3621 }
3622 
3623 /*
3624  * inputs:
3625  * zc_name		name of dataset to destroy
3626  * zc_objset_type	type of objset
3627  * zc_defer_destroy	mark for deferred destroy
3628  *
3629  * outputs:		none
3630  */
3631 static int
3632 zfs_ioc_destroy(zfs_cmd_t *zc)
3633 {
3634 	int err;
3635 
3636 	if (zc->zc_objset_type == DMU_OST_ZFS) {
3637 		err = zfs_unmount_snap(zc->zc_name);
3638 		if (err != 0)
3639 			return (err);
3640 	}
3641 
3642 	if (strchr(zc->zc_name, '@'))
3643 		err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3644 	else
3645 		err = dsl_destroy_head(zc->zc_name);
3646 	if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3647 		(void) zvol_remove_minor(zc->zc_name);
3648 	return (err);
3649 }
3650 
3651 /*
3652  * fsname is name of dataset to rollback (to most recent snapshot)
3653  *
3654  * innvl is not used.
3655  *
3656  * outnvl: "target" -> name of most recent snapshot
3657  * }
3658  */
3659 /* ARGSUSED */
3660 static int
3661 zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl)
3662 {
3663 	zfsvfs_t *zfsvfs;
3664 	int error;
3665 
3666 	if (getzfsvfs(fsname, &zfsvfs) == 0) {
3667 		error = zfs_suspend_fs(zfsvfs);
3668 		if (error == 0) {
3669 			int resume_err;
3670 
3671 			error = dsl_dataset_rollback(fsname, zfsvfs, outnvl);
3672 			resume_err = zfs_resume_fs(zfsvfs, fsname);
3673 			error = error ? error : resume_err;
3674 		}
3675 		VFS_RELE(zfsvfs->z_vfs);
3676 	} else {
3677 		error = dsl_dataset_rollback(fsname, NULL, outnvl);
3678 	}
3679 	return (error);
3680 }
3681 
3682 static int
3683 recursive_unmount(const char *fsname, void *arg)
3684 {
3685 	const char *snapname = arg;
3686 	char fullname[MAXNAMELEN];
3687 
3688 	(void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname);
3689 	return (zfs_unmount_snap(fullname));
3690 }
3691 
3692 /*
3693  * inputs:
3694  * zc_name	old name of dataset
3695  * zc_value	new name of dataset
3696  * zc_cookie	recursive flag (only valid for snapshots)
3697  *
3698  * outputs:	none
3699  */
3700 static int
3701 zfs_ioc_rename(zfs_cmd_t *zc)
3702 {
3703 	boolean_t recursive = zc->zc_cookie & 1;
3704 	char *at;
3705 
3706 	zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3707 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3708 	    strchr(zc->zc_value, '%'))
3709 		return (SET_ERROR(EINVAL));
3710 
3711 	at = strchr(zc->zc_name, '@');
3712 	if (at != NULL) {
3713 		/* snaps must be in same fs */
3714 		int error;
3715 
3716 		if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3717 			return (SET_ERROR(EXDEV));
3718 		*at = '\0';
3719 		if (zc->zc_objset_type == DMU_OST_ZFS) {
3720 			error = dmu_objset_find(zc->zc_name,
3721 			    recursive_unmount, at + 1,
3722 			    recursive ? DS_FIND_CHILDREN : 0);
3723 			if (error != 0) {
3724 				*at = '@';
3725 				return (error);
3726 			}
3727 		}
3728 		error = dsl_dataset_rename_snapshot(zc->zc_name,
3729 		    at + 1, strchr(zc->zc_value, '@') + 1, recursive);
3730 		*at = '@';
3731 
3732 		return (error);
3733 	} else {
3734 		if (zc->zc_objset_type == DMU_OST_ZVOL)
3735 			(void) zvol_remove_minor(zc->zc_name);
3736 		return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3737 	}
3738 }
3739 
3740 static int
3741 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3742 {
3743 	const char *propname = nvpair_name(pair);
3744 	boolean_t issnap = (strchr(dsname, '@') != NULL);
3745 	zfs_prop_t prop = zfs_name_to_prop(propname);
3746 	uint64_t intval;
3747 	int err;
3748 
3749 	if (prop == ZPROP_INVAL) {
3750 		if (zfs_prop_user(propname)) {
3751 			if (err = zfs_secpolicy_write_perms(dsname,
3752 			    ZFS_DELEG_PERM_USERPROP, cr))
3753 				return (err);
3754 			return (0);
3755 		}
3756 
3757 		if (!issnap && zfs_prop_userquota(propname)) {
3758 			const char *perm = NULL;
3759 			const char *uq_prefix =
3760 			    zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3761 			const char *gq_prefix =
3762 			    zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3763 
3764 			if (strncmp(propname, uq_prefix,
3765 			    strlen(uq_prefix)) == 0) {
3766 				perm = ZFS_DELEG_PERM_USERQUOTA;
3767 			} else if (strncmp(propname, gq_prefix,
3768 			    strlen(gq_prefix)) == 0) {
3769 				perm = ZFS_DELEG_PERM_GROUPQUOTA;
3770 			} else {
3771 				/* USERUSED and GROUPUSED are read-only */
3772 				return (SET_ERROR(EINVAL));
3773 			}
3774 
3775 			if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3776 				return (err);
3777 			return (0);
3778 		}
3779 
3780 		return (SET_ERROR(EINVAL));
3781 	}
3782 
3783 	if (issnap)
3784 		return (SET_ERROR(EINVAL));
3785 
3786 	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3787 		/*
3788 		 * dsl_prop_get_all_impl() returns properties in this
3789 		 * format.
3790 		 */
3791 		nvlist_t *attrs;
3792 		VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3793 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3794 		    &pair) == 0);
3795 	}
3796 
3797 	/*
3798 	 * Check that this value is valid for this pool version
3799 	 */
3800 	switch (prop) {
3801 	case ZFS_PROP_COMPRESSION:
3802 		/*
3803 		 * If the user specified gzip compression, make sure
3804 		 * the SPA supports it. We ignore any errors here since
3805 		 * we'll catch them later.
3806 		 */
3807 		if (nvpair_value_uint64(pair, &intval) == 0) {
3808 			if (intval >= ZIO_COMPRESS_GZIP_1 &&
3809 			    intval <= ZIO_COMPRESS_GZIP_9 &&
3810 			    zfs_earlier_version(dsname,
3811 			    SPA_VERSION_GZIP_COMPRESSION)) {
3812 				return (SET_ERROR(ENOTSUP));
3813 			}
3814 
3815 			if (intval == ZIO_COMPRESS_ZLE &&
3816 			    zfs_earlier_version(dsname,
3817 			    SPA_VERSION_ZLE_COMPRESSION))
3818 				return (SET_ERROR(ENOTSUP));
3819 
3820 			if (intval == ZIO_COMPRESS_LZ4) {
3821 				spa_t *spa;
3822 
3823 				if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3824 					return (err);
3825 
3826 				if (!spa_feature_is_enabled(spa,
3827 				    SPA_FEATURE_LZ4_COMPRESS)) {
3828 					spa_close(spa, FTAG);
3829 					return (SET_ERROR(ENOTSUP));
3830 				}
3831 				spa_close(spa, FTAG);
3832 			}
3833 
3834 			/*
3835 			 * If this is a bootable dataset then
3836 			 * verify that the compression algorithm
3837 			 * is supported for booting. We must return
3838 			 * something other than ENOTSUP since it
3839 			 * implies a downrev pool version.
3840 			 */
3841 			if (zfs_is_bootfs(dsname) &&
3842 			    !BOOTFS_COMPRESS_VALID(intval)) {
3843 				return (SET_ERROR(ERANGE));
3844 			}
3845 		}
3846 		break;
3847 
3848 	case ZFS_PROP_COPIES:
3849 		if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3850 			return (SET_ERROR(ENOTSUP));
3851 		break;
3852 
3853 	case ZFS_PROP_DEDUP:
3854 		if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3855 			return (SET_ERROR(ENOTSUP));
3856 		break;
3857 
3858 	case ZFS_PROP_RECORDSIZE:
3859 		/* Record sizes above 128k need the feature to be enabled */
3860 		if (nvpair_value_uint64(pair, &intval) == 0 &&
3861 		    intval > SPA_OLD_MAXBLOCKSIZE) {
3862 			spa_t *spa;
3863 
3864 			/*
3865 			 * If this is a bootable dataset then
3866 			 * the we don't allow large (>128K) blocks,
3867 			 * because GRUB doesn't support them.
3868 			 */
3869 			if (zfs_is_bootfs(dsname) &&
3870 			    intval > SPA_OLD_MAXBLOCKSIZE) {
3871 				return (SET_ERROR(EDOM));
3872 			}
3873 
3874 			/*
3875 			 * We don't allow setting the property above 1MB,
3876 			 * unless the tunable has been changed.
3877 			 */
3878 			if (intval > zfs_max_recordsize ||
3879 			    intval > SPA_MAXBLOCKSIZE)
3880 				return (SET_ERROR(EDOM));
3881 
3882 			if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3883 				return (err);
3884 
3885 			if (!spa_feature_is_enabled(spa,
3886 			    SPA_FEATURE_LARGE_BLOCKS)) {
3887 				spa_close(spa, FTAG);
3888 				return (SET_ERROR(ENOTSUP));
3889 			}
3890 			spa_close(spa, FTAG);
3891 		}
3892 		break;
3893 
3894 	case ZFS_PROP_SHARESMB:
3895 		if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3896 			return (SET_ERROR(ENOTSUP));
3897 		break;
3898 
3899 	case ZFS_PROP_ACLINHERIT:
3900 		if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3901 		    nvpair_value_uint64(pair, &intval) == 0) {
3902 			if (intval == ZFS_ACL_PASSTHROUGH_X &&
3903 			    zfs_earlier_version(dsname,
3904 			    SPA_VERSION_PASSTHROUGH_X))
3905 				return (SET_ERROR(ENOTSUP));
3906 		}
3907 		break;
3908 	}
3909 
3910 	return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3911 }
3912 
3913 /*
3914  * Checks for a race condition to make sure we don't increment a feature flag
3915  * multiple times.
3916  */
3917 static int
3918 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
3919 {
3920 	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3921 	spa_feature_t *featurep = arg;
3922 
3923 	if (!spa_feature_is_active(spa, *featurep))
3924 		return (0);
3925 	else
3926 		return (SET_ERROR(EBUSY));
3927 }
3928 
3929 /*
3930  * The callback invoked on feature activation in the sync task caused by
3931  * zfs_prop_activate_feature.
3932  */
3933 static void
3934 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
3935 {
3936 	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3937 	spa_feature_t *featurep = arg;
3938 
3939 	spa_feature_incr(spa, *featurep, tx);
3940 }
3941 
3942 /*
3943  * Activates a feature on a pool in response to a property setting. This
3944  * creates a new sync task which modifies the pool to reflect the feature
3945  * as being active.
3946  */
3947 static int
3948 zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature)
3949 {
3950 	int err;
3951 
3952 	/* EBUSY here indicates that the feature is already active */
3953 	err = dsl_sync_task(spa_name(spa),
3954 	    zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
3955 	    &feature, 2, ZFS_SPACE_CHECK_RESERVED);
3956 
3957 	if (err != 0 && err != EBUSY)
3958 		return (err);
3959 	else
3960 		return (0);
3961 }
3962 
3963 /*
3964  * Removes properties from the given props list that fail permission checks
3965  * needed to clear them and to restore them in case of a receive error. For each
3966  * property, make sure we have both set and inherit permissions.
3967  *
3968  * Returns the first error encountered if any permission checks fail. If the
3969  * caller provides a non-NULL errlist, it also gives the complete list of names
3970  * of all the properties that failed a permission check along with the
3971  * corresponding error numbers. The caller is responsible for freeing the
3972  * returned errlist.
3973  *
3974  * If every property checks out successfully, zero is returned and the list
3975  * pointed at by errlist is NULL.
3976  */
3977 static int
3978 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
3979 {
3980 	zfs_cmd_t *zc;
3981 	nvpair_t *pair, *next_pair;
3982 	nvlist_t *errors;
3983 	int err, rv = 0;
3984 
3985 	if (props == NULL)
3986 		return (0);
3987 
3988 	VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3989 
3990 	zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
3991 	(void) strcpy(zc->zc_name, dataset);
3992 	pair = nvlist_next_nvpair(props, NULL);
3993 	while (pair != NULL) {
3994 		next_pair = nvlist_next_nvpair(props, pair);
3995 
3996 		(void) strcpy(zc->zc_value, nvpair_name(pair));
3997 		if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
3998 		    (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
3999 			VERIFY(nvlist_remove_nvpair(props, pair) == 0);
4000 			VERIFY(nvlist_add_int32(errors,
4001 			    zc->zc_value, err) == 0);
4002 		}
4003 		pair = next_pair;
4004 	}
4005 	kmem_free(zc, sizeof (zfs_cmd_t));
4006 
4007 	if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
4008 		nvlist_free(errors);
4009 		errors = NULL;
4010 	} else {
4011 		VERIFY(nvpair_value_int32(pair, &rv) == 0);
4012 	}
4013 
4014 	if (errlist == NULL)
4015 		nvlist_free(errors);
4016 	else
4017 		*errlist = errors;
4018 
4019 	return (rv);
4020 }
4021 
4022 static boolean_t
4023 propval_equals(nvpair_t *p1, nvpair_t *p2)
4024 {
4025 	if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
4026 		/* dsl_prop_get_all_impl() format */
4027 		nvlist_t *attrs;
4028 		VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
4029 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4030 		    &p1) == 0);
4031 	}
4032 
4033 	if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
4034 		nvlist_t *attrs;
4035 		VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
4036 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4037 		    &p2) == 0);
4038 	}
4039 
4040 	if (nvpair_type(p1) != nvpair_type(p2))
4041 		return (B_FALSE);
4042 
4043 	if (nvpair_type(p1) == DATA_TYPE_STRING) {
4044 		char *valstr1, *valstr2;
4045 
4046 		VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
4047 		VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
4048 		return (strcmp(valstr1, valstr2) == 0);
4049 	} else {
4050 		uint64_t intval1, intval2;
4051 
4052 		VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4053 		VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4054 		return (intval1 == intval2);
4055 	}
4056 }
4057 
4058 /*
4059  * Remove properties from props if they are not going to change (as determined
4060  * by comparison with origprops). Remove them from origprops as well, since we
4061  * do not need to clear or restore properties that won't change.
4062  */
4063 static void
4064 props_reduce(nvlist_t *props, nvlist_t *origprops)
4065 {
4066 	nvpair_t *pair, *next_pair;
4067 
4068 	if (origprops == NULL)
4069 		return; /* all props need to be received */
4070 
4071 	pair = nvlist_next_nvpair(props, NULL);
4072 	while (pair != NULL) {
4073 		const char *propname = nvpair_name(pair);
4074 		nvpair_t *match;
4075 
4076 		next_pair = nvlist_next_nvpair(props, pair);
4077 
4078 		if ((nvlist_lookup_nvpair(origprops, propname,
4079 		    &match) != 0) || !propval_equals(pair, match))
4080 			goto next; /* need to set received value */
4081 
4082 		/* don't clear the existing received value */
4083 		(void) nvlist_remove_nvpair(origprops, match);
4084 		/* don't bother receiving the property */
4085 		(void) nvlist_remove_nvpair(props, pair);
4086 next:
4087 		pair = next_pair;
4088 	}
4089 }
4090 
4091 #ifdef	DEBUG
4092 static boolean_t zfs_ioc_recv_inject_err;
4093 #endif
4094 
4095 /*
4096  * inputs:
4097  * zc_name		name of containing filesystem
4098  * zc_nvlist_src{_size}	nvlist of properties to apply
4099  * zc_value		name of snapshot to create
4100  * zc_string		name of clone origin (if DRR_FLAG_CLONE)
4101  * zc_cookie		file descriptor to recv from
4102  * zc_begin_record	the BEGIN record of the stream (not byteswapped)
4103  * zc_guid		force flag
4104  * zc_cleanup_fd	cleanup-on-exit file descriptor
4105  * zc_action_handle	handle for this guid/ds mapping (or zero on first call)
4106  *
4107  * outputs:
4108  * zc_cookie		number of bytes read
4109  * zc_nvlist_dst{_size} error for each unapplied received property
4110  * zc_obj		zprop_errflags_t
4111  * zc_action_handle	handle for this guid/ds mapping
4112  */
4113 static int
4114 zfs_ioc_recv(zfs_cmd_t *zc)
4115 {
4116 	file_t *fp;
4117 	dmu_recv_cookie_t drc;
4118 	boolean_t force = (boolean_t)zc->zc_guid;
4119 	int fd;
4120 	int error = 0;
4121 	int props_error = 0;
4122 	nvlist_t *errors;
4123 	offset_t off;
4124 	nvlist_t *props = NULL; /* sent properties */
4125 	nvlist_t *origprops = NULL; /* existing properties */
4126 	char *origin = NULL;
4127 	char *tosnap;
4128 	char tofs[ZFS_MAXNAMELEN];
4129 	boolean_t first_recvd_props = B_FALSE;
4130 
4131 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4132 	    strchr(zc->zc_value, '@') == NULL ||
4133 	    strchr(zc->zc_value, '%'))
4134 		return (SET_ERROR(EINVAL));
4135 
4136 	(void) strcpy(tofs, zc->zc_value);
4137 	tosnap = strchr(tofs, '@');
4138 	*tosnap++ = '\0';
4139 
4140 	if (zc->zc_nvlist_src != NULL &&
4141 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
4142 	    zc->zc_iflags, &props)) != 0)
4143 		return (error);
4144 
4145 	fd = zc->zc_cookie;
4146 	fp = getf(fd);
4147 	if (fp == NULL) {
4148 		nvlist_free(props);
4149 		return (SET_ERROR(EBADF));
4150 	}
4151 
4152 	VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4153 
4154 	if (zc->zc_string[0])
4155 		origin = zc->zc_string;
4156 
4157 	error = dmu_recv_begin(tofs, tosnap,
4158 	    &zc->zc_begin_record, force, origin, &drc);
4159 	if (error != 0)
4160 		goto out;
4161 
4162 	/*
4163 	 * Set properties before we receive the stream so that they are applied
4164 	 * to the new data. Note that we must call dmu_recv_stream() if
4165 	 * dmu_recv_begin() succeeds.
4166 	 */
4167 	if (props != NULL && !drc.drc_newfs) {
4168 		if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
4169 		    SPA_VERSION_RECVD_PROPS &&
4170 		    !dsl_prop_get_hasrecvd(tofs))
4171 			first_recvd_props = B_TRUE;
4172 
4173 		/*
4174 		 * If new received properties are supplied, they are to
4175 		 * completely replace the existing received properties, so stash
4176 		 * away the existing ones.
4177 		 */
4178 		if (dsl_prop_get_received(tofs, &origprops) == 0) {
4179 			nvlist_t *errlist = NULL;
4180 			/*
4181 			 * Don't bother writing a property if its value won't
4182 			 * change (and avoid the unnecessary security checks).
4183 			 *
4184 			 * The first receive after SPA_VERSION_RECVD_PROPS is a
4185 			 * special case where we blow away all local properties
4186 			 * regardless.
4187 			 */
4188 			if (!first_recvd_props)
4189 				props_reduce(props, origprops);
4190 			if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
4191 				(void) nvlist_merge(errors, errlist, 0);
4192 			nvlist_free(errlist);
4193 
4194 			if (clear_received_props(tofs, origprops,
4195 			    first_recvd_props ? NULL : props) != 0)
4196 				zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4197 		} else {
4198 			zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4199 		}
4200 	}
4201 
4202 	if (props != NULL) {
4203 		props_error = dsl_prop_set_hasrecvd(tofs);
4204 
4205 		if (props_error == 0) {
4206 			(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4207 			    props, errors);
4208 		}
4209 	}
4210 
4211 	if (zc->zc_nvlist_dst_size != 0 &&
4212 	    (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4213 	    put_nvlist(zc, errors) != 0)) {
4214 		/*
4215 		 * Caller made zc->zc_nvlist_dst less than the minimum expected
4216 		 * size or supplied an invalid address.
4217 		 */
4218 		props_error = SET_ERROR(EINVAL);
4219 	}
4220 
4221 	off = fp->f_offset;
4222 	error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
4223 	    &zc->zc_action_handle);
4224 
4225 	if (error == 0) {
4226 		zfsvfs_t *zfsvfs = NULL;
4227 
4228 		if (getzfsvfs(tofs, &zfsvfs) == 0) {
4229 			/* online recv */
4230 			int end_err;
4231 
4232 			error = zfs_suspend_fs(zfsvfs);
4233 			/*
4234 			 * If the suspend fails, then the recv_end will
4235 			 * likely also fail, and clean up after itself.
4236 			 */
4237 			end_err = dmu_recv_end(&drc, zfsvfs);
4238 			if (error == 0)
4239 				error = zfs_resume_fs(zfsvfs, tofs);
4240 			error = error ? error : end_err;
4241 			VFS_RELE(zfsvfs->z_vfs);
4242 		} else {
4243 			error = dmu_recv_end(&drc, NULL);
4244 		}
4245 	}
4246 
4247 	zc->zc_cookie = off - fp->f_offset;
4248 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4249 		fp->f_offset = off;
4250 
4251 #ifdef	DEBUG
4252 	if (zfs_ioc_recv_inject_err) {
4253 		zfs_ioc_recv_inject_err = B_FALSE;
4254 		error = 1;
4255 	}
4256 #endif
4257 	/*
4258 	 * On error, restore the original props.
4259 	 */
4260 	if (error != 0 && props != NULL && !drc.drc_newfs) {
4261 		if (clear_received_props(tofs, props, NULL) != 0) {
4262 			/*
4263 			 * We failed to clear the received properties.
4264 			 * Since we may have left a $recvd value on the
4265 			 * system, we can't clear the $hasrecvd flag.
4266 			 */
4267 			zc->zc_obj |= ZPROP_ERR_NORESTORE;
4268 		} else if (first_recvd_props) {
4269 			dsl_prop_unset_hasrecvd(tofs);
4270 		}
4271 
4272 		if (origprops == NULL && !drc.drc_newfs) {
4273 			/* We failed to stash the original properties. */
4274 			zc->zc_obj |= ZPROP_ERR_NORESTORE;
4275 		}
4276 
4277 		/*
4278 		 * dsl_props_set() will not convert RECEIVED to LOCAL on or
4279 		 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4280 		 * explictly if we're restoring local properties cleared in the
4281 		 * first new-style receive.
4282 		 */
4283 		if (origprops != NULL &&
4284 		    zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4285 		    ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4286 		    origprops, NULL) != 0) {
4287 			/*
4288 			 * We stashed the original properties but failed to
4289 			 * restore them.
4290 			 */
4291 			zc->zc_obj |= ZPROP_ERR_NORESTORE;
4292 		}
4293 	}
4294 out:
4295 	nvlist_free(props);
4296 	nvlist_free(origprops);
4297 	nvlist_free(errors);
4298 	releasef(fd);
4299 
4300 	if (error == 0)
4301 		error = props_error;
4302 
4303 	return (error);
4304 }
4305 
4306 /*
4307  * inputs:
4308  * zc_name	name of snapshot to send
4309  * zc_cookie	file descriptor to send stream to
4310  * zc_obj	fromorigin flag (mutually exclusive with zc_fromobj)
4311  * zc_sendobj	objsetid of snapshot to send
4312  * zc_fromobj	objsetid of incremental fromsnap (may be zero)
4313  * zc_guid	if set, estimate size of stream only.  zc_cookie is ignored.
4314  *		output size in zc_objset_type.
4315  * zc_flags	lzc_send_flags
4316  *
4317  * outputs:
4318  * zc_objset_type	estimated size, if zc_guid is set
4319  */
4320 static int
4321 zfs_ioc_send(zfs_cmd_t *zc)
4322 {
4323 	int error;
4324 	offset_t off;
4325 	boolean_t estimate = (zc->zc_guid != 0);
4326 	boolean_t embedok = (zc->zc_flags & 0x1);
4327 	boolean_t large_block_ok = (zc->zc_flags & 0x2);
4328 
4329 	if (zc->zc_obj != 0) {
4330 		dsl_pool_t *dp;
4331 		dsl_dataset_t *tosnap;
4332 
4333 		error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4334 		if (error != 0)
4335 			return (error);
4336 
4337 		error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4338 		if (error != 0) {
4339 			dsl_pool_rele(dp, FTAG);
4340 			return (error);
4341 		}
4342 
4343 		if (dsl_dir_is_clone(tosnap->ds_dir))
4344 			zc->zc_fromobj =
4345 			    dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj;
4346 		dsl_dataset_rele(tosnap, FTAG);
4347 		dsl_pool_rele(dp, FTAG);
4348 	}
4349 
4350 	if (estimate) {
4351 		dsl_pool_t *dp;
4352 		dsl_dataset_t *tosnap;
4353 		dsl_dataset_t *fromsnap = NULL;
4354 
4355 		error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4356 		if (error != 0)
4357 			return (error);
4358 
4359 		error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4360 		if (error != 0) {
4361 			dsl_pool_rele(dp, FTAG);
4362 			return (error);
4363 		}
4364 
4365 		if (zc->zc_fromobj != 0) {
4366 			error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4367 			    FTAG, &fromsnap);
4368 			if (error != 0) {
4369 				dsl_dataset_rele(tosnap, FTAG);
4370 				dsl_pool_rele(dp, FTAG);
4371 				return (error);
4372 			}
4373 		}
4374 
4375 		error = dmu_send_estimate(tosnap, fromsnap,
4376 		    &zc->zc_objset_type);
4377 
4378 		if (fromsnap != NULL)
4379 			dsl_dataset_rele(fromsnap, FTAG);
4380 		dsl_dataset_rele(tosnap, FTAG);
4381 		dsl_pool_rele(dp, FTAG);
4382 	} else {
4383 		file_t *fp = getf(zc->zc_cookie);
4384 		if (fp == NULL)
4385 			return (SET_ERROR(EBADF));
4386 
4387 		off = fp->f_offset;
4388 		error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4389 		    zc->zc_fromobj, embedok, large_block_ok,
4390 		    zc->zc_cookie, fp->f_vnode, &off);
4391 
4392 		if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4393 			fp->f_offset = off;
4394 		releasef(zc->zc_cookie);
4395 	}
4396 	return (error);
4397 }
4398 
4399 /*
4400  * inputs:
4401  * zc_name	name of snapshot on which to report progress
4402  * zc_cookie	file descriptor of send stream
4403  *
4404  * outputs:
4405  * zc_cookie	number of bytes written in send stream thus far
4406  */
4407 static int
4408 zfs_ioc_send_progress(zfs_cmd_t *zc)
4409 {
4410 	dsl_pool_t *dp;
4411 	dsl_dataset_t *ds;
4412 	dmu_sendarg_t *dsp = NULL;
4413 	int error;
4414 
4415 	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4416 	if (error != 0)
4417 		return (error);
4418 
4419 	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4420 	if (error != 0) {
4421 		dsl_pool_rele(dp, FTAG);
4422 		return (error);
4423 	}
4424 
4425 	mutex_enter(&ds->ds_sendstream_lock);
4426 
4427 	/*
4428 	 * Iterate over all the send streams currently active on this dataset.
4429 	 * If there's one which matches the specified file descriptor _and_ the
4430 	 * stream was started by the current process, return the progress of
4431 	 * that stream.
4432 	 */
4433 	for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4434 	    dsp = list_next(&ds->ds_sendstreams, dsp)) {
4435 		if (dsp->dsa_outfd == zc->zc_cookie &&
4436 		    dsp->dsa_proc == curproc)
4437 			break;
4438 	}
4439 
4440 	if (dsp != NULL)
4441 		zc->zc_cookie = *(dsp->dsa_off);
4442 	else
4443 		error = SET_ERROR(ENOENT);
4444 
4445 	mutex_exit(&ds->ds_sendstream_lock);
4446 	dsl_dataset_rele(ds, FTAG);
4447 	dsl_pool_rele(dp, FTAG);
4448 	return (error);
4449 }
4450 
4451 static int
4452 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4453 {
4454 	int id, error;
4455 
4456 	error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4457 	    &zc->zc_inject_record);
4458 
4459 	if (error == 0)
4460 		zc->zc_guid = (uint64_t)id;
4461 
4462 	return (error);
4463 }
4464 
4465 static int
4466 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4467 {
4468 	return (zio_clear_fault((int)zc->zc_guid));
4469 }
4470 
4471 static int
4472 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4473 {
4474 	int id = (int)zc->zc_guid;
4475 	int error;
4476 
4477 	error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4478 	    &zc->zc_inject_record);
4479 
4480 	zc->zc_guid = id;
4481 
4482 	return (error);
4483 }
4484 
4485 static int
4486 zfs_ioc_error_log(zfs_cmd_t *zc)
4487 {
4488 	spa_t *spa;
4489 	int error;
4490 	size_t count = (size_t)zc->zc_nvlist_dst_size;
4491 
4492 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4493 		return (error);
4494 
4495 	error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4496 	    &count);
4497 	if (error == 0)
4498 		zc->zc_nvlist_dst_size = count;
4499 	else
4500 		zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4501 
4502 	spa_close(spa, FTAG);
4503 
4504 	return (error);
4505 }
4506 
4507 static int
4508 zfs_ioc_clear(zfs_cmd_t *zc)
4509 {
4510 	spa_t *spa;
4511 	vdev_t *vd;
4512 	int error;
4513 
4514 	/*
4515 	 * On zpool clear we also fix up missing slogs
4516 	 */
4517 	mutex_enter(&spa_namespace_lock);
4518 	spa = spa_lookup(zc->zc_name);
4519 	if (spa == NULL) {
4520 		mutex_exit(&spa_namespace_lock);
4521 		return (SET_ERROR(EIO));
4522 	}
4523 	if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4524 		/* we need to let spa_open/spa_load clear the chains */
4525 		spa_set_log_state(spa, SPA_LOG_CLEAR);
4526 	}
4527 	spa->spa_last_open_failed = 0;
4528 	mutex_exit(&spa_namespace_lock);
4529 
4530 	if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4531 		error = spa_open(zc->zc_name, &spa, FTAG);
4532 	} else {
4533 		nvlist_t *policy;
4534 		nvlist_t *config = NULL;
4535 
4536 		if (zc->zc_nvlist_src == NULL)
4537 			return (SET_ERROR(EINVAL));
4538 
4539 		if ((error = get_nvlist(zc->zc_nvlist_src,
4540 		    zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4541 			error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4542 			    policy, &config);
4543 			if (config != NULL) {
4544 				int err;
4545 
4546 				if ((err = put_nvlist(zc, config)) != 0)
4547 					error = err;
4548 				nvlist_free(config);
4549 			}
4550 			nvlist_free(policy);
4551 		}
4552 	}
4553 
4554 	if (error != 0)
4555 		return (error);
4556 
4557 	spa_vdev_state_enter(spa, SCL_NONE);
4558 
4559 	if (zc->zc_guid == 0) {
4560 		vd = NULL;
4561 	} else {
4562 		vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4563 		if (vd == NULL) {
4564 			(void) spa_vdev_state_exit(spa, NULL, ENODEV);
4565 			spa_close(spa, FTAG);
4566 			return (SET_ERROR(ENODEV));
4567 		}
4568 	}
4569 
4570 	vdev_clear(spa, vd);
4571 
4572 	(void) spa_vdev_state_exit(spa, NULL, 0);
4573 
4574 	/*
4575 	 * Resume any suspended I/Os.
4576 	 */
4577 	if (zio_resume(spa) != 0)
4578 		error = SET_ERROR(EIO);
4579 
4580 	spa_close(spa, FTAG);
4581 
4582 	return (error);
4583 }
4584 
4585 static int
4586 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4587 {
4588 	spa_t *spa;
4589 	int error;
4590 
4591 	error = spa_open(zc->zc_name, &spa, FTAG);
4592 	if (error != 0)
4593 		return (error);
4594 
4595 	spa_vdev_state_enter(spa, SCL_NONE);
4596 
4597 	/*
4598 	 * If a resilver is already in progress then set the
4599 	 * spa_scrub_reopen flag to B_TRUE so that we don't restart
4600 	 * the scan as a side effect of the reopen. Otherwise, let
4601 	 * vdev_open() decided if a resilver is required.
4602 	 */
4603 	spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4604 	vdev_reopen(spa->spa_root_vdev);
4605 	spa->spa_scrub_reopen = B_FALSE;
4606 
4607 	(void) spa_vdev_state_exit(spa, NULL, 0);
4608 	spa_close(spa, FTAG);
4609 	return (0);
4610 }
4611 /*
4612  * inputs:
4613  * zc_name	name of filesystem
4614  * zc_value	name of origin snapshot
4615  *
4616  * outputs:
4617  * zc_string	name of conflicting snapshot, if there is one
4618  */
4619 static int
4620 zfs_ioc_promote(zfs_cmd_t *zc)
4621 {
4622 	char *cp;
4623 
4624 	/*
4625 	 * We don't need to unmount *all* the origin fs's snapshots, but
4626 	 * it's easier.
4627 	 */
4628 	cp = strchr(zc->zc_value, '@');
4629 	if (cp)
4630 		*cp = '\0';
4631 	(void) dmu_objset_find(zc->zc_value,
4632 	    zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
4633 	return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4634 }
4635 
4636 /*
4637  * Retrieve a single {user|group}{used|quota}@... property.
4638  *
4639  * inputs:
4640  * zc_name	name of filesystem
4641  * zc_objset_type zfs_userquota_prop_t
4642  * zc_value	domain name (eg. "S-1-234-567-89")
4643  * zc_guid	RID/UID/GID
4644  *
4645  * outputs:
4646  * zc_cookie	property value
4647  */
4648 static int
4649 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4650 {
4651 	zfsvfs_t *zfsvfs;
4652 	int error;
4653 
4654 	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4655 		return (SET_ERROR(EINVAL));
4656 
4657 	error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4658 	if (error != 0)
4659 		return (error);
4660 
4661 	error = zfs_userspace_one(zfsvfs,
4662 	    zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4663 	zfsvfs_rele(zfsvfs, FTAG);
4664 
4665 	return (error);
4666 }
4667 
4668 /*
4669  * inputs:
4670  * zc_name		name of filesystem
4671  * zc_cookie		zap cursor
4672  * zc_objset_type	zfs_userquota_prop_t
4673  * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4674  *
4675  * outputs:
4676  * zc_nvlist_dst[_size]	data buffer (array of zfs_useracct_t)
4677  * zc_cookie	zap cursor
4678  */
4679 static int
4680 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4681 {
4682 	zfsvfs_t *zfsvfs;
4683 	int bufsize = zc->zc_nvlist_dst_size;
4684 
4685 	if (bufsize <= 0)
4686 		return (SET_ERROR(ENOMEM));
4687 
4688 	int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4689 	if (error != 0)
4690 		return (error);
4691 
4692 	void *buf = kmem_alloc(bufsize, KM_SLEEP);
4693 
4694 	error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
4695 	    buf, &zc->zc_nvlist_dst_size);
4696 
4697 	if (error == 0) {
4698 		error = xcopyout(buf,
4699 		    (void *)(uintptr_t)zc->zc_nvlist_dst,
4700 		    zc->zc_nvlist_dst_size);
4701 	}
4702 	kmem_free(buf, bufsize);
4703 	zfsvfs_rele(zfsvfs, FTAG);
4704 
4705 	return (error);
4706 }
4707 
4708 /*
4709  * inputs:
4710  * zc_name		name of filesystem
4711  *
4712  * outputs:
4713  * none
4714  */
4715 static int
4716 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4717 {
4718 	objset_t *os;
4719 	int error = 0;
4720 	zfsvfs_t *zfsvfs;
4721 
4722 	if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
4723 		if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
4724 			/*
4725 			 * If userused is not enabled, it may be because the
4726 			 * objset needs to be closed & reopened (to grow the
4727 			 * objset_phys_t).  Suspend/resume the fs will do that.
4728 			 */
4729 			error = zfs_suspend_fs(zfsvfs);
4730 			if (error == 0) {
4731 				dmu_objset_refresh_ownership(zfsvfs->z_os,
4732 				    zfsvfs);
4733 				error = zfs_resume_fs(zfsvfs, zc->zc_name);
4734 			}
4735 		}
4736 		if (error == 0)
4737 			error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
4738 		VFS_RELE(zfsvfs->z_vfs);
4739 	} else {
4740 		/* XXX kind of reading contents without owning */
4741 		error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4742 		if (error != 0)
4743 			return (error);
4744 
4745 		error = dmu_objset_userspace_upgrade(os);
4746 		dmu_objset_rele(os, FTAG);
4747 	}
4748 
4749 	return (error);
4750 }
4751 
4752 /*
4753  * We don't want to have a hard dependency
4754  * against some special symbols in sharefs
4755  * nfs, and smbsrv.  Determine them if needed when
4756  * the first file system is shared.
4757  * Neither sharefs, nfs or smbsrv are unloadable modules.
4758  */
4759 int (*znfsexport_fs)(void *arg);
4760 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
4761 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
4762 
4763 int zfs_nfsshare_inited;
4764 int zfs_smbshare_inited;
4765 
4766 ddi_modhandle_t nfs_mod;
4767 ddi_modhandle_t sharefs_mod;
4768 ddi_modhandle_t smbsrv_mod;
4769 kmutex_t zfs_share_lock;
4770 
4771 static int
4772 zfs_init_sharefs()
4773 {
4774 	int error;
4775 
4776 	ASSERT(MUTEX_HELD(&zfs_share_lock));
4777 	/* Both NFS and SMB shares also require sharetab support. */
4778 	if (sharefs_mod == NULL && ((sharefs_mod =
4779 	    ddi_modopen("fs/sharefs",
4780 	    KRTLD_MODE_FIRST, &error)) == NULL)) {
4781 		return (SET_ERROR(ENOSYS));
4782 	}
4783 	if (zshare_fs == NULL && ((zshare_fs =
4784 	    (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
4785 	    ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
4786 		return (SET_ERROR(ENOSYS));
4787 	}
4788 	return (0);
4789 }
4790 
4791 static int
4792 zfs_ioc_share(zfs_cmd_t *zc)
4793 {
4794 	int error;
4795 	int opcode;
4796 
4797 	switch (zc->zc_share.z_sharetype) {
4798 	case ZFS_SHARE_NFS:
4799 	case ZFS_UNSHARE_NFS:
4800 		if (zfs_nfsshare_inited == 0) {
4801 			mutex_enter(&zfs_share_lock);
4802 			if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
4803 			    KRTLD_MODE_FIRST, &error)) == NULL)) {
4804 				mutex_exit(&zfs_share_lock);
4805 				return (SET_ERROR(ENOSYS));
4806 			}
4807 			if (znfsexport_fs == NULL &&
4808 			    ((znfsexport_fs = (int (*)(void *))
4809 			    ddi_modsym(nfs_mod,
4810 			    "nfs_export", &error)) == NULL)) {
4811 				mutex_exit(&zfs_share_lock);
4812 				return (SET_ERROR(ENOSYS));
4813 			}
4814 			error = zfs_init_sharefs();
4815 			if (error != 0) {
4816 				mutex_exit(&zfs_share_lock);
4817 				return (SET_ERROR(ENOSYS));
4818 			}
4819 			zfs_nfsshare_inited = 1;
4820 			mutex_exit(&zfs_share_lock);
4821 		}
4822 		break;
4823 	case ZFS_SHARE_SMB:
4824 	case ZFS_UNSHARE_SMB:
4825 		if (zfs_smbshare_inited == 0) {
4826 			mutex_enter(&zfs_share_lock);
4827 			if (smbsrv_mod == NULL && ((smbsrv_mod =
4828 			    ddi_modopen("drv/smbsrv",
4829 			    KRTLD_MODE_FIRST, &error)) == NULL)) {
4830 				mutex_exit(&zfs_share_lock);
4831 				return (SET_ERROR(ENOSYS));
4832 			}
4833 			if (zsmbexport_fs == NULL && ((zsmbexport_fs =
4834 			    (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
4835 			    "smb_server_share", &error)) == NULL)) {
4836 				mutex_exit(&zfs_share_lock);
4837 				return (SET_ERROR(ENOSYS));
4838 			}
4839 			error = zfs_init_sharefs();
4840 			if (error != 0) {
4841 				mutex_exit(&zfs_share_lock);
4842 				return (SET_ERROR(ENOSYS));
4843 			}
4844 			zfs_smbshare_inited = 1;
4845 			mutex_exit(&zfs_share_lock);
4846 		}
4847 		break;
4848 	default:
4849 		return (SET_ERROR(EINVAL));
4850 	}
4851 
4852 	switch (zc->zc_share.z_sharetype) {
4853 	case ZFS_SHARE_NFS:
4854 	case ZFS_UNSHARE_NFS:
4855 		if (error =
4856 		    znfsexport_fs((void *)
4857 		    (uintptr_t)zc->zc_share.z_exportdata))
4858 			return (error);
4859 		break;
4860 	case ZFS_SHARE_SMB:
4861 	case ZFS_UNSHARE_SMB:
4862 		if (error = zsmbexport_fs((void *)
4863 		    (uintptr_t)zc->zc_share.z_exportdata,
4864 		    zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
4865 		    B_TRUE: B_FALSE)) {
4866 			return (error);
4867 		}
4868 		break;
4869 	}
4870 
4871 	opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
4872 	    zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
4873 	    SHAREFS_ADD : SHAREFS_REMOVE;
4874 
4875 	/*
4876 	 * Add or remove share from sharetab
4877 	 */
4878 	error = zshare_fs(opcode,
4879 	    (void *)(uintptr_t)zc->zc_share.z_sharedata,
4880 	    zc->zc_share.z_sharemax);
4881 
4882 	return (error);
4883 
4884 }
4885 
4886 ace_t full_access[] = {
4887 	{(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4888 };
4889 
4890 /*
4891  * inputs:
4892  * zc_name		name of containing filesystem
4893  * zc_obj		object # beyond which we want next in-use object #
4894  *
4895  * outputs:
4896  * zc_obj		next in-use object #
4897  */
4898 static int
4899 zfs_ioc_next_obj(zfs_cmd_t *zc)
4900 {
4901 	objset_t *os = NULL;
4902 	int error;
4903 
4904 	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4905 	if (error != 0)
4906 		return (error);
4907 
4908 	error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4909 	    dsl_dataset_phys(os->os_dsl_dataset)->ds_prev_snap_txg);
4910 
4911 	dmu_objset_rele(os, FTAG);
4912 	return (error);
4913 }
4914 
4915 /*
4916  * inputs:
4917  * zc_name		name of filesystem
4918  * zc_value		prefix name for snapshot
4919  * zc_cleanup_fd	cleanup-on-exit file descriptor for calling process
4920  *
4921  * outputs:
4922  * zc_value		short name of new snapshot
4923  */
4924 static int
4925 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
4926 {
4927 	char *snap_name;
4928 	char *hold_name;
4929 	int error;
4930 	minor_t minor;
4931 
4932 	error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
4933 	if (error != 0)
4934 		return (error);
4935 
4936 	snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
4937 	    (u_longlong_t)ddi_get_lbolt64());
4938 	hold_name = kmem_asprintf("%%%s", zc->zc_value);
4939 
4940 	error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
4941 	    hold_name);
4942 	if (error == 0)
4943 		(void) strcpy(zc->zc_value, snap_name);
4944 	strfree(snap_name);
4945 	strfree(hold_name);
4946 	zfs_onexit_fd_rele(zc->zc_cleanup_fd);
4947 	return (error);
4948 }
4949 
4950 /*
4951  * inputs:
4952  * zc_name		name of "to" snapshot
4953  * zc_value		name of "from" snapshot
4954  * zc_cookie		file descriptor to write diff data on
4955  *
4956  * outputs:
4957  * dmu_diff_record_t's to the file descriptor
4958  */
4959 static int
4960 zfs_ioc_diff(zfs_cmd_t *zc)
4961 {
4962 	file_t *fp;
4963 	offset_t off;
4964 	int error;
4965 
4966 	fp = getf(zc->zc_cookie);
4967 	if (fp == NULL)
4968 		return (SET_ERROR(EBADF));
4969 
4970 	off = fp->f_offset;
4971 
4972 	error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
4973 
4974 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4975 		fp->f_offset = off;
4976 	releasef(zc->zc_cookie);
4977 
4978 	return (error);
4979 }
4980 
4981 /*
4982  * Remove all ACL files in shares dir
4983  */
4984 static int
4985 zfs_smb_acl_purge(znode_t *dzp)
4986 {
4987 	zap_cursor_t	zc;
4988 	zap_attribute_t	zap;
4989 	zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
4990 	int error;
4991 
4992 	for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
4993 	    (error = zap_cursor_retrieve(&zc, &zap)) == 0;
4994 	    zap_cursor_advance(&zc)) {
4995 		if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
4996 		    NULL, 0)) != 0)
4997 			break;
4998 	}
4999 	zap_cursor_fini(&zc);
5000 	return (error);
5001 }
5002 
5003 static int
5004 zfs_ioc_smb_acl(zfs_cmd_t *zc)
5005 {
5006 	vnode_t *vp;
5007 	znode_t *dzp;
5008 	vnode_t *resourcevp = NULL;
5009 	znode_t *sharedir;
5010 	zfsvfs_t *zfsvfs;
5011 	nvlist_t *nvlist;
5012 	char *src, *target;
5013 	vattr_t vattr;
5014 	vsecattr_t vsec;
5015 	int error = 0;
5016 
5017 	if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
5018 	    NO_FOLLOW, NULL, &vp)) != 0)
5019 		return (error);
5020 
5021 	/* Now make sure mntpnt and dataset are ZFS */
5022 
5023 	if (vp->v_vfsp->vfs_fstype != zfsfstype ||
5024 	    (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
5025 	    zc->zc_name) != 0)) {
5026 		VN_RELE(vp);
5027 		return (SET_ERROR(EINVAL));
5028 	}
5029 
5030 	dzp = VTOZ(vp);
5031 	zfsvfs = dzp->z_zfsvfs;
5032 	ZFS_ENTER(zfsvfs);
5033 
5034 	/*
5035 	 * Create share dir if its missing.
5036 	 */
5037 	mutex_enter(&zfsvfs->z_lock);
5038 	if (zfsvfs->z_shares_dir == 0) {
5039 		dmu_tx_t *tx;
5040 
5041 		tx = dmu_tx_create(zfsvfs->z_os);
5042 		dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
5043 		    ZFS_SHARES_DIR);
5044 		dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
5045 		error = dmu_tx_assign(tx, TXG_WAIT);
5046 		if (error != 0) {
5047 			dmu_tx_abort(tx);
5048 		} else {
5049 			error = zfs_create_share_dir(zfsvfs, tx);
5050 			dmu_tx_commit(tx);
5051 		}
5052 		if (error != 0) {
5053 			mutex_exit(&zfsvfs->z_lock);
5054 			VN_RELE(vp);
5055 			ZFS_EXIT(zfsvfs);
5056 			return (error);
5057 		}
5058 	}
5059 	mutex_exit(&zfsvfs->z_lock);
5060 
5061 	ASSERT(zfsvfs->z_shares_dir);
5062 	if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
5063 		VN_RELE(vp);
5064 		ZFS_EXIT(zfsvfs);
5065 		return (error);
5066 	}
5067 
5068 	switch (zc->zc_cookie) {
5069 	case ZFS_SMB_ACL_ADD:
5070 		vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
5071 		vattr.va_type = VREG;
5072 		vattr.va_mode = S_IFREG|0777;
5073 		vattr.va_uid = 0;
5074 		vattr.va_gid = 0;
5075 
5076 		vsec.vsa_mask = VSA_ACE;
5077 		vsec.vsa_aclentp = &full_access;
5078 		vsec.vsa_aclentsz = sizeof (full_access);
5079 		vsec.vsa_aclcnt = 1;
5080 
5081 		error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
5082 		    &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
5083 		if (resourcevp)
5084 			VN_RELE(resourcevp);
5085 		break;
5086 
5087 	case ZFS_SMB_ACL_REMOVE:
5088 		error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
5089 		    NULL, 0);
5090 		break;
5091 
5092 	case ZFS_SMB_ACL_RENAME:
5093 		if ((error = get_nvlist(zc->zc_nvlist_src,
5094 		    zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
5095 			VN_RELE(vp);
5096 			ZFS_EXIT(zfsvfs);
5097 			return (error);
5098 		}
5099 		if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
5100 		    nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
5101 		    &target)) {
5102 			VN_RELE(vp);
5103 			VN_RELE(ZTOV(sharedir));
5104 			ZFS_EXIT(zfsvfs);
5105 			nvlist_free(nvlist);
5106 			return (error);
5107 		}
5108 		error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
5109 		    kcred, NULL, 0);
5110 		nvlist_free(nvlist);
5111 		break;
5112 
5113 	case ZFS_SMB_ACL_PURGE:
5114 		error = zfs_smb_acl_purge(sharedir);
5115 		break;
5116 
5117 	default:
5118 		error = SET_ERROR(EINVAL);
5119 		break;
5120 	}
5121 
5122 	VN_RELE(vp);
5123 	VN_RELE(ZTOV(sharedir));
5124 
5125 	ZFS_EXIT(zfsvfs);
5126 
5127 	return (error);
5128 }
5129 
5130 /*
5131  * innvl: {
5132  *     "holds" -> { snapname -> holdname (string), ... }
5133  *     (optional) "cleanup_fd" -> fd (int32)
5134  * }
5135  *
5136  * outnvl: {
5137  *     snapname -> error value (int32)
5138  *     ...
5139  * }
5140  */
5141 /* ARGSUSED */
5142 static int
5143 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
5144 {
5145 	nvlist_t *holds;
5146 	int cleanup_fd = -1;
5147 	int error;
5148 	minor_t minor = 0;
5149 
5150 	error = nvlist_lookup_nvlist(args, "holds", &holds);
5151 	if (error != 0)
5152 		return (SET_ERROR(EINVAL));
5153 
5154 	if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
5155 		error = zfs_onexit_fd_hold(cleanup_fd, &minor);
5156 		if (error != 0)
5157 			return (error);
5158 	}
5159 
5160 	error = dsl_dataset_user_hold(holds, minor, errlist);
5161 	if (minor != 0)
5162 		zfs_onexit_fd_rele(cleanup_fd);
5163 	return (error);
5164 }
5165 
5166 /*
5167  * innvl is not used.
5168  *
5169  * outnvl: {
5170  *    holdname -> time added (uint64 seconds since epoch)
5171  *    ...
5172  * }
5173  */
5174 /* ARGSUSED */
5175 static int
5176 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
5177 {
5178 	return (dsl_dataset_get_holds(snapname, outnvl));
5179 }
5180 
5181 /*
5182  * innvl: {
5183  *     snapname -> { holdname, ... }
5184  *     ...
5185  * }
5186  *
5187  * outnvl: {
5188  *     snapname -> error value (int32)
5189  *     ...
5190  * }
5191  */
5192 /* ARGSUSED */
5193 static int
5194 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
5195 {
5196 	return (dsl_dataset_user_release(holds, errlist));
5197 }
5198 
5199 /*
5200  * inputs:
5201  * zc_name		name of new filesystem or snapshot
5202  * zc_value		full name of old snapshot
5203  *
5204  * outputs:
5205  * zc_cookie		space in bytes
5206  * zc_objset_type	compressed space in bytes
5207  * zc_perm_action	uncompressed space in bytes
5208  */
5209 static int
5210 zfs_ioc_space_written(zfs_cmd_t *zc)
5211 {
5212 	int error;
5213 	dsl_pool_t *dp;
5214 	dsl_dataset_t *new, *old;
5215 
5216 	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5217 	if (error != 0)
5218 		return (error);
5219 	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5220 	if (error != 0) {
5221 		dsl_pool_rele(dp, FTAG);
5222 		return (error);
5223 	}
5224 	error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5225 	if (error != 0) {
5226 		dsl_dataset_rele(new, FTAG);
5227 		dsl_pool_rele(dp, FTAG);
5228 		return (error);
5229 	}
5230 
5231 	error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5232 	    &zc->zc_objset_type, &zc->zc_perm_action);
5233 	dsl_dataset_rele(old, FTAG);
5234 	dsl_dataset_rele(new, FTAG);
5235 	dsl_pool_rele(dp, FTAG);
5236 	return (error);
5237 }
5238 
5239 /*
5240  * innvl: {
5241  *     "firstsnap" -> snapshot name
5242  * }
5243  *
5244  * outnvl: {
5245  *     "used" -> space in bytes
5246  *     "compressed" -> compressed space in bytes
5247  *     "uncompressed" -> uncompressed space in bytes
5248  * }
5249  */
5250 static int
5251 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5252 {
5253 	int error;
5254 	dsl_pool_t *dp;
5255 	dsl_dataset_t *new, *old;
5256 	char *firstsnap;
5257 	uint64_t used, comp, uncomp;
5258 
5259 	if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5260 		return (SET_ERROR(EINVAL));
5261 
5262 	error = dsl_pool_hold(lastsnap, FTAG, &dp);
5263 	if (error != 0)
5264 		return (error);
5265 
5266 	error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5267 	if (error != 0) {
5268 		dsl_pool_rele(dp, FTAG);
5269 		return (error);
5270 	}
5271 	error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5272 	if (error != 0) {
5273 		dsl_dataset_rele(new, FTAG);
5274 		dsl_pool_rele(dp, FTAG);
5275 		return (error);
5276 	}
5277 
5278 	error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5279 	dsl_dataset_rele(old, FTAG);
5280 	dsl_dataset_rele(new, FTAG);
5281 	dsl_pool_rele(dp, FTAG);
5282 	fnvlist_add_uint64(outnvl, "used", used);
5283 	fnvlist_add_uint64(outnvl, "compressed", comp);
5284 	fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5285 	return (error);
5286 }
5287 
5288 /*
5289  * innvl: {
5290  *     "fd" -> file descriptor to write stream to (int32)
5291  *     (optional) "fromsnap" -> full snap name to send an incremental from
5292  *     (optional) "largeblockok" -> (value ignored)
5293  *         indicates that blocks > 128KB are permitted
5294  *     (optional) "embedok" -> (value ignored)
5295  *         presence indicates DRR_WRITE_EMBEDDED records are permitted
5296  * }
5297  *
5298  * outnvl is unused
5299  */
5300 /* ARGSUSED */
5301 static int
5302 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5303 {
5304 	int error;
5305 	offset_t off;
5306 	char *fromname = NULL;
5307 	int fd;
5308 	boolean_t largeblockok;
5309 	boolean_t embedok;
5310 
5311 	error = nvlist_lookup_int32(innvl, "fd", &fd);
5312 	if (error != 0)
5313 		return (SET_ERROR(EINVAL));
5314 
5315 	(void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5316 
5317 	largeblockok = nvlist_exists(innvl, "largeblockok");
5318 	embedok = nvlist_exists(innvl, "embedok");
5319 
5320 	file_t *fp = getf(fd);
5321 	if (fp == NULL)
5322 		return (SET_ERROR(EBADF));
5323 
5324 	off = fp->f_offset;
5325 	error = dmu_send(snapname, fromname, embedok, largeblockok,
5326 	    fd, fp->f_vnode, &off);
5327 
5328 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5329 		fp->f_offset = off;
5330 	releasef(fd);
5331 	return (error);
5332 }
5333 
5334 /*
5335  * Determine approximately how large a zfs send stream will be -- the number
5336  * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5337  *
5338  * innvl: {
5339  *     (optional) "fromsnap" -> full snap name to send an incremental from
5340  * }
5341  *
5342  * outnvl: {
5343  *     "space" -> bytes of space (uint64)
5344  * }
5345  */
5346 static int
5347 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5348 {
5349 	dsl_pool_t *dp;
5350 	dsl_dataset_t *fromsnap = NULL;
5351 	dsl_dataset_t *tosnap;
5352 	int error;
5353 	char *fromname;
5354 	uint64_t space;
5355 
5356 	error = dsl_pool_hold(snapname, FTAG, &dp);
5357 	if (error != 0)
5358 		return (error);
5359 
5360 	error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5361 	if (error != 0) {
5362 		dsl_pool_rele(dp, FTAG);
5363 		return (error);
5364 	}
5365 
5366 	error = nvlist_lookup_string(innvl, "fromsnap", &fromname);
5367 	if (error == 0) {
5368 		error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5369 		if (error != 0) {
5370 			dsl_dataset_rele(tosnap, FTAG);
5371 			dsl_pool_rele(dp, FTAG);
5372 			return (error);
5373 		}
5374 	}
5375 
5376 	error = dmu_send_estimate(tosnap, fromsnap, &space);
5377 	fnvlist_add_uint64(outnvl, "space", space);
5378 
5379 	if (fromsnap != NULL)
5380 		dsl_dataset_rele(fromsnap, FTAG);
5381 	dsl_dataset_rele(tosnap, FTAG);
5382 	dsl_pool_rele(dp, FTAG);
5383 	return (error);
5384 }
5385 
5386 static int
5387 zfs_ioc_set_zev_callbacks(const char *unused, nvlist_t *innvl,
5388     nvlist_t *outnvl)
5389 {
5390 	int error;
5391 	uint64_t cb_addr;
5392 	/*
5393 	 * Our secpolicy for this op makes sure it's called in
5394 	 * kernel context, and that no other callbacks have
5395 	 * been registered, yet.
5396 	 */
5397 	error = nvlist_lookup_uint64(innvl, "callbacks", &cb_addr);
5398 	if (error != 0) {
5399 		cmn_err(CE_WARN, "set_zev_callbacks nvlist lookup failed (%d)",
5400 		    error);
5401 		return (error);
5402 	}
5403 	/* cb_addr is always a kernel memory address */
5404 	rw_enter(&rz_zev_rwlock, RW_WRITER);
5405 	if (rz_zev_callbacks != rz_zev_default_callbacks) {
5406 		rw_exit(&rz_zev_rwlock);
5407 		return (EBUSY);
5408 	}
5409 	rz_zev_callbacks = (void *)(uintptr_t)cb_addr;
5410 	rw_exit(&rz_zev_rwlock);
5411 	return (0);
5412 }
5413 
5414 static int
5415 zfs_ioc_unset_zev_callbacks(const char *unused, nvlist_t *innvl,
5416     nvlist_t *outnvl)
5417 {
5418 	/*
5419 	 * Our secpolicy for this op makes sure it's called in
5420 	 * kernel context.
5421 	 */
5422 	rw_enter(&rz_zev_rwlock, RW_WRITER);
5423 	rz_zev_callbacks = rz_zev_default_callbacks;
5424 	rw_exit(&rz_zev_rwlock);
5425 	/* after mutex release, no thread is using the old table anymore. */
5426 	return (0);
5427 }
5428 
5429 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5430 
5431 static void
5432 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5433     zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5434     boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5435 {
5436 	zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5437 
5438 	ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5439 	ASSERT3U(ioc, <, ZFS_IOC_LAST);
5440 	ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5441 	ASSERT3P(vec->zvec_func, ==, NULL);
5442 
5443 	vec->zvec_legacy_func = func;
5444 	vec->zvec_secpolicy = secpolicy;
5445 	vec->zvec_namecheck = namecheck;
5446 	vec->zvec_allow_log = log_history;
5447 	vec->zvec_pool_check = pool_check;
5448 }
5449 
5450 /*
5451  * See the block comment at the beginning of this file for details on
5452  * each argument to this function.
5453  */
5454 static void
5455 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5456     zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5457     zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5458     boolean_t allow_log)
5459 {
5460 	zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5461 
5462 	ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5463 	ASSERT3U(ioc, <, ZFS_IOC_LAST);
5464 	ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5465 	ASSERT3P(vec->zvec_func, ==, NULL);
5466 
5467 	/* if we are logging, the name must be valid */
5468 	ASSERT(!allow_log || namecheck != NO_NAME);
5469 
5470 	vec->zvec_name = name;
5471 	vec->zvec_func = func;
5472 	vec->zvec_secpolicy = secpolicy;
5473 	vec->zvec_namecheck = namecheck;
5474 	vec->zvec_pool_check = pool_check;
5475 	vec->zvec_smush_outnvlist = smush_outnvlist;
5476 	vec->zvec_allow_log = allow_log;
5477 }
5478 
5479 static void
5480 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5481     zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5482     zfs_ioc_poolcheck_t pool_check)
5483 {
5484 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
5485 	    POOL_NAME, log_history, pool_check);
5486 }
5487 
5488 static void
5489 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5490     zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5491 {
5492 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
5493 	    DATASET_NAME, B_FALSE, pool_check);
5494 }
5495 
5496 static void
5497 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5498 {
5499 	zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5500 	    POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5501 }
5502 
5503 static void
5504 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5505     zfs_secpolicy_func_t *secpolicy)
5506 {
5507 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
5508 	    NO_NAME, B_FALSE, POOL_CHECK_NONE);
5509 }
5510 
5511 static void
5512 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5513     zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5514 {
5515 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
5516 	    DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5517 }
5518 
5519 static void
5520 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5521 {
5522 	zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5523 	    zfs_secpolicy_read);
5524 }
5525 
5526 static void
5527 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5528 	zfs_secpolicy_func_t *secpolicy)
5529 {
5530 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
5531 	    DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5532 }
5533 
5534 static void
5535 zfs_ioctl_init(void)
5536 {
5537 	zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5538 	    zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5539 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5540 
5541 	zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5542 	    zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5543 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5544 
5545 	zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5546 	    zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5547 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5548 
5549 	zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5550 	    zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5551 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5552 
5553 	zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5554 	    zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5555 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5556 
5557 	zfs_ioctl_register("create", ZFS_IOC_CREATE,
5558 	    zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5559 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5560 
5561 	zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5562 	    zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5563 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5564 
5565 	zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5566 	    zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5567 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5568 
5569 	zfs_ioctl_register("hold", ZFS_IOC_HOLD,
5570 	    zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
5571 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5572 	zfs_ioctl_register("release", ZFS_IOC_RELEASE,
5573 	    zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
5574 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5575 
5576 	zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
5577 	    zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
5578 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5579 
5580 	zfs_ioctl_register("set_zev_callbacks", ZFS_IOC_SET_ZEV_CALLBACKS,
5581 	    zfs_ioc_set_zev_callbacks, zfs_secpolicy_set_zev_callbacks, NO_NAME,
5582 	    POOL_CHECK_NONE, B_TRUE, B_FALSE);
5583 
5584 	zfs_ioctl_register("unset_zev_callbacks", ZFS_IOC_UNSET_ZEV_CALLBACKS,
5585 	    zfs_ioc_unset_zev_callbacks, zfs_secpolicy_unset_zev_callbacks,
5586 	    NO_NAME, POOL_CHECK_NONE, B_TRUE, B_FALSE);
5587 	zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
5588 	    zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
5589 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
5590 
5591 	zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
5592 	    zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
5593 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5594 
5595 	zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
5596 	    zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
5597 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5598 
5599 	zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
5600 	    zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
5601 	    POOL_NAME,
5602 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5603 
5604 	/* IOCTLS that use the legacy function signature */
5605 
5606 	zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5607 	    zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5608 
5609 	zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5610 	    zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5611 	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5612 	    zfs_ioc_pool_scan);
5613 	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5614 	    zfs_ioc_pool_upgrade);
5615 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5616 	    zfs_ioc_vdev_add);
5617 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5618 	    zfs_ioc_vdev_remove);
5619 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5620 	    zfs_ioc_vdev_set_state);
5621 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5622 	    zfs_ioc_vdev_attach);
5623 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5624 	    zfs_ioc_vdev_detach);
5625 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5626 	    zfs_ioc_vdev_setpath);
5627 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5628 	    zfs_ioc_vdev_setfru);
5629 	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5630 	    zfs_ioc_pool_set_props);
5631 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5632 	    zfs_ioc_vdev_split);
5633 	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5634 	    zfs_ioc_pool_reguid);
5635 
5636 	zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5637 	    zfs_ioc_pool_configs, zfs_secpolicy_none);
5638 	zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5639 	    zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5640 	zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5641 	    zfs_ioc_inject_fault, zfs_secpolicy_inject);
5642 	zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5643 	    zfs_ioc_clear_fault, zfs_secpolicy_inject);
5644 	zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
5645 	    zfs_ioc_inject_list_next, zfs_secpolicy_inject);
5646 
5647 	/*
5648 	 * pool destroy, and export don't log the history as part of
5649 	 * zfsdev_ioctl, but rather zfs_ioc_pool_export
5650 	 * does the logging of those commands.
5651 	 */
5652 	zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
5653 	    zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5654 	zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
5655 	    zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5656 
5657 	zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
5658 	    zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5659 	zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
5660 	    zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5661 
5662 	zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
5663 	    zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
5664 	zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
5665 	    zfs_ioc_dsobj_to_dsname,
5666 	    zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
5667 	zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
5668 	    zfs_ioc_pool_get_history,
5669 	    zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
5670 
5671 	zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
5672 	    zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5673 
5674 	zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
5675 	    zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5676 	zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
5677 	    zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5678 
5679 	zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
5680 	    zfs_ioc_space_written);
5681 	zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
5682 	    zfs_ioc_objset_recvd_props);
5683 	zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
5684 	    zfs_ioc_next_obj);
5685 	zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
5686 	    zfs_ioc_get_fsacl);
5687 	zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
5688 	    zfs_ioc_objset_stats);
5689 	zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
5690 	    zfs_ioc_objset_zplprops);
5691 	zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
5692 	    zfs_ioc_dataset_list_next);
5693 	zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
5694 	    zfs_ioc_snapshot_list_next);
5695 	zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
5696 	    zfs_ioc_send_progress);
5697 
5698 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
5699 	    zfs_ioc_diff, zfs_secpolicy_diff);
5700 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
5701 	    zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
5702 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
5703 	    zfs_ioc_obj_to_path, zfs_secpolicy_diff);
5704 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
5705 	    zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
5706 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
5707 	    zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
5708 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
5709 	    zfs_ioc_send, zfs_secpolicy_send);
5710 
5711 	zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
5712 	    zfs_secpolicy_none);
5713 	zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
5714 	    zfs_secpolicy_destroy);
5715 	zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
5716 	    zfs_secpolicy_rename);
5717 	zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
5718 	    zfs_secpolicy_recv);
5719 	zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
5720 	    zfs_secpolicy_promote);
5721 	zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
5722 	    zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
5723 	zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
5724 	    zfs_secpolicy_set_fsacl);
5725 
5726 	zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
5727 	    zfs_secpolicy_share, POOL_CHECK_NONE);
5728 	zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
5729 	    zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
5730 	zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
5731 	    zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
5732 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5733 	zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
5734 	    zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
5735 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5736 }
5737 
5738 int
5739 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
5740     zfs_ioc_poolcheck_t check)
5741 {
5742 	spa_t *spa;
5743 	int error;
5744 
5745 	ASSERT(type == POOL_NAME || type == DATASET_NAME);
5746 
5747 	if (check & POOL_CHECK_NONE)
5748 		return (0);
5749 
5750 	error = spa_open(name, &spa, FTAG);
5751 	if (error == 0) {
5752 		if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
5753 			error = SET_ERROR(EAGAIN);
5754 		else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
5755 			error = SET_ERROR(EROFS);
5756 		spa_close(spa, FTAG);
5757 	}
5758 	return (error);
5759 }
5760 
5761 /*
5762  * Find a free minor number.
5763  */
5764 minor_t
5765 zfsdev_minor_alloc(void)
5766 {
5767 	static minor_t last_minor;
5768 	minor_t m;
5769 
5770 	ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5771 
5772 	for (m = last_minor + 1; m != last_minor; m++) {
5773 		if (m > ZFSDEV_MAX_MINOR)
5774 			m = 1;
5775 		if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
5776 			last_minor = m;
5777 			return (m);
5778 		}
5779 	}
5780 
5781 	return (0);
5782 }
5783 
5784 static int
5785 zfs_ctldev_init(dev_t *devp)
5786 {
5787 	minor_t minor;
5788 	zfs_soft_state_t *zs;
5789 
5790 	ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5791 	ASSERT(getminor(*devp) == 0);
5792 
5793 	minor = zfsdev_minor_alloc();
5794 	if (minor == 0)
5795 		return (SET_ERROR(ENXIO));
5796 
5797 	if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
5798 		return (SET_ERROR(EAGAIN));
5799 
5800 	*devp = makedevice(getemajor(*devp), minor);
5801 
5802 	zs = ddi_get_soft_state(zfsdev_state, minor);
5803 	zs->zss_type = ZSST_CTLDEV;
5804 	zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
5805 
5806 	return (0);
5807 }
5808 
5809 static void
5810 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
5811 {
5812 	ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5813 
5814 	zfs_onexit_destroy(zo);
5815 	ddi_soft_state_free(zfsdev_state, minor);
5816 }
5817 
5818 void *
5819 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
5820 {
5821 	zfs_soft_state_t *zp;
5822 
5823 	zp = ddi_get_soft_state(zfsdev_state, minor);
5824 	if (zp == NULL || zp->zss_type != which)
5825 		return (NULL);
5826 
5827 	return (zp->zss_data);
5828 }
5829 
5830 static int
5831 zfsdev_open(dev_t *devp, int flag, int otyp, cred_t *cr)
5832 {
5833 	int error = 0;
5834 
5835 	if (getminor(*devp) != 0)
5836 		return (zvol_open(devp, flag, otyp, cr));
5837 
5838 	/* This is the control device. Allocate a new minor if requested. */
5839 	if (flag & FEXCL) {
5840 		mutex_enter(&zfsdev_state_lock);
5841 		error = zfs_ctldev_init(devp);
5842 		mutex_exit(&zfsdev_state_lock);
5843 	}
5844 
5845 	return (error);
5846 }
5847 
5848 static int
5849 zfsdev_close(dev_t dev, int flag, int otyp, cred_t *cr)
5850 {
5851 	zfs_onexit_t *zo;
5852 	minor_t minor = getminor(dev);
5853 
5854 	if (minor == 0)
5855 		return (0);
5856 
5857 	mutex_enter(&zfsdev_state_lock);
5858 	zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
5859 	if (zo == NULL) {
5860 		mutex_exit(&zfsdev_state_lock);
5861 		return (zvol_close(dev, flag, otyp, cr));
5862 	}
5863 	zfs_ctldev_destroy(zo, minor);
5864 	mutex_exit(&zfsdev_state_lock);
5865 
5866 	return (0);
5867 }
5868 
5869 static int
5870 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
5871 {
5872 	zfs_cmd_t *zc;
5873 	uint_t vecnum;
5874 	int error, rc, len;
5875 	minor_t minor = getminor(dev);
5876 	const zfs_ioc_vec_t *vec;
5877 	char *saved_poolname = NULL;
5878 	nvlist_t *innvl = NULL;
5879 
5880 	if (minor != 0 &&
5881 	    zfsdev_get_soft_state(minor, ZSST_CTLDEV) == NULL)
5882 		return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
5883 
5884 	vecnum = cmd - ZFS_IOC_FIRST;
5885 	ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
5886 
5887 	if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
5888 		return (SET_ERROR(EINVAL));
5889 	vec = &zfs_ioc_vec[vecnum];
5890 
5891 	zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
5892 
5893 	error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
5894 	if (error != 0) {
5895 		error = SET_ERROR(EFAULT);
5896 		goto out;
5897 	}
5898 
5899 	zc->zc_iflags = flag & FKIOCTL;
5900 	if (zc->zc_nvlist_src_size != 0) {
5901 		error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
5902 		    zc->zc_iflags, &innvl);
5903 		if (error != 0)
5904 			goto out;
5905 	}
5906 
5907 	/*
5908 	 * Ensure that all pool/dataset names are valid before we pass down to
5909 	 * the lower layers.
5910 	 */
5911 	zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
5912 	switch (vec->zvec_namecheck) {
5913 	case POOL_NAME:
5914 		if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
5915 			error = SET_ERROR(EINVAL);
5916 		else
5917 			error = pool_status_check(zc->zc_name,
5918 			    vec->zvec_namecheck, vec->zvec_pool_check);
5919 		break;
5920 
5921 	case DATASET_NAME:
5922 		if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
5923 			error = SET_ERROR(EINVAL);
5924 		else
5925 			error = pool_status_check(zc->zc_name,
5926 			    vec->zvec_namecheck, vec->zvec_pool_check);
5927 		break;
5928 
5929 	case NO_NAME:
5930 		break;
5931 	}
5932 
5933 
5934 	if (error == 0 && !(flag & FKIOCTL))
5935 		error = vec->zvec_secpolicy(zc, innvl, cr);
5936 
5937 	if (error != 0)
5938 		goto out;
5939 
5940 	/* legacy ioctls can modify zc_name */
5941 	len = strcspn(zc->zc_name, "/@#") + 1;
5942 	saved_poolname = kmem_alloc(len, KM_SLEEP);
5943 	(void) strlcpy(saved_poolname, zc->zc_name, len);
5944 
5945 	if (vec->zvec_func != NULL) {
5946 		nvlist_t *outnvl;
5947 		int puterror = 0;
5948 		spa_t *spa;
5949 		nvlist_t *lognv = NULL;
5950 
5951 		ASSERT(vec->zvec_legacy_func == NULL);
5952 
5953 		/*
5954 		 * Add the innvl to the lognv before calling the func,
5955 		 * in case the func changes the innvl.
5956 		 */
5957 		if (vec->zvec_allow_log) {
5958 			lognv = fnvlist_alloc();
5959 			fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
5960 			    vec->zvec_name);
5961 			if (!nvlist_empty(innvl)) {
5962 				fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
5963 				    innvl);
5964 			}
5965 		}
5966 
5967 		outnvl = fnvlist_alloc();
5968 		error = vec->zvec_func(zc->zc_name, innvl, outnvl);
5969 
5970 		if (error == 0 && vec->zvec_allow_log &&
5971 		    spa_open(zc->zc_name, &spa, FTAG) == 0) {
5972 			if (!nvlist_empty(outnvl)) {
5973 				fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
5974 				    outnvl);
5975 			}
5976 			(void) spa_history_log_nvl(spa, lognv);
5977 			spa_close(spa, FTAG);
5978 		}
5979 		fnvlist_free(lognv);
5980 
5981 		if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
5982 			int smusherror = 0;
5983 			if (vec->zvec_smush_outnvlist) {
5984 				smusherror = nvlist_smush(outnvl,
5985 				    zc->zc_nvlist_dst_size);
5986 			}
5987 			if (smusherror == 0)
5988 				puterror = put_nvlist(zc, outnvl);
5989 		}
5990 
5991 		if (puterror != 0)
5992 			error = puterror;
5993 
5994 		nvlist_free(outnvl);
5995 	} else {
5996 		error = vec->zvec_legacy_func(zc);
5997 	}
5998 
5999 out:
6000 	nvlist_free(innvl);
6001 	rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
6002 	if (error == 0 && rc != 0)
6003 		error = SET_ERROR(EFAULT);
6004 	if (error == 0 && vec->zvec_allow_log) {
6005 		char *s = tsd_get(zfs_allow_log_key);
6006 		if (s != NULL)
6007 			strfree(s);
6008 		(void) tsd_set(zfs_allow_log_key, saved_poolname);
6009 	} else {
6010 		if (saved_poolname != NULL)
6011 			strfree(saved_poolname);
6012 	}
6013 
6014 	kmem_free(zc, sizeof (zfs_cmd_t));
6015 	return (error);
6016 }
6017 
6018 static int
6019 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
6020 {
6021 	if (cmd != DDI_ATTACH)
6022 		return (DDI_FAILURE);
6023 
6024 	if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
6025 	    DDI_PSEUDO, 0) == DDI_FAILURE)
6026 		return (DDI_FAILURE);
6027 
6028 	zfs_dip = dip;
6029 
6030 	ddi_report_dev(dip);
6031 
6032 	return (DDI_SUCCESS);
6033 }
6034 
6035 static int
6036 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
6037 {
6038 	if (spa_busy() || zfs_busy() || zvol_busy())
6039 		return (DDI_FAILURE);
6040 
6041 	if (cmd != DDI_DETACH)
6042 		return (DDI_FAILURE);
6043 
6044 	zfs_dip = NULL;
6045 
6046 	ddi_prop_remove_all(dip);
6047 	ddi_remove_minor_node(dip, NULL);
6048 
6049 	return (DDI_SUCCESS);
6050 }
6051 
6052 /*ARGSUSED*/
6053 static int
6054 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
6055 {
6056 	switch (infocmd) {
6057 	case DDI_INFO_DEVT2DEVINFO:
6058 		*result = zfs_dip;
6059 		return (DDI_SUCCESS);
6060 
6061 	case DDI_INFO_DEVT2INSTANCE:
6062 		*result = (void *)0;
6063 		return (DDI_SUCCESS);
6064 	}
6065 
6066 	return (DDI_FAILURE);
6067 }
6068 
6069 /*
6070  * OK, so this is a little weird.
6071  *
6072  * /dev/zfs is the control node, i.e. minor 0.
6073  * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
6074  *
6075  * /dev/zfs has basically nothing to do except serve up ioctls,
6076  * so most of the standard driver entry points are in zvol.c.
6077  */
6078 static struct cb_ops zfs_cb_ops = {
6079 	zfsdev_open,	/* open */
6080 	zfsdev_close,	/* close */
6081 	zvol_strategy,	/* strategy */
6082 	nodev,		/* print */
6083 	zvol_dump,	/* dump */
6084 	zvol_read,	/* read */
6085 	zvol_write,	/* write */
6086 	zfsdev_ioctl,	/* ioctl */
6087 	nodev,		/* devmap */
6088 	nodev,		/* mmap */
6089 	nodev,		/* segmap */
6090 	nochpoll,	/* poll */
6091 	ddi_prop_op,	/* prop_op */
6092 	NULL,		/* streamtab */
6093 	D_NEW | D_MP | D_64BIT,		/* Driver compatibility flag */
6094 	CB_REV,		/* version */
6095 	nodev,		/* async read */
6096 	nodev,		/* async write */
6097 };
6098 
6099 static struct dev_ops zfs_dev_ops = {
6100 	DEVO_REV,	/* version */
6101 	0,		/* refcnt */
6102 	zfs_info,	/* info */
6103 	nulldev,	/* identify */
6104 	nulldev,	/* probe */
6105 	zfs_attach,	/* attach */
6106 	zfs_detach,	/* detach */
6107 	nodev,		/* reset */
6108 	&zfs_cb_ops,	/* driver operations */
6109 	NULL,		/* no bus operations */
6110 	NULL,		/* power */
6111 	ddi_quiesce_not_needed,	/* quiesce */
6112 };
6113 
6114 static struct modldrv zfs_modldrv = {
6115 	&mod_driverops,
6116 	"ZFS storage pool",
6117 	&zfs_dev_ops
6118 };
6119 
6120 static struct modlinkage modlinkage = {
6121 	MODREV_1,
6122 	(void *)&zfs_modlfs,
6123 	(void *)&zfs_modldrv,
6124 	NULL
6125 };
6126 
6127 static void
6128 zfs_allow_log_destroy(void *arg)
6129 {
6130 	char *poolname = arg;
6131 	strfree(poolname);
6132 }
6133 
6134 int
6135 _init(void)
6136 {
6137 	int error;
6138 
6139 	spa_init(FREAD | FWRITE);
6140 	zfs_init();
6141 	zvol_init();
6142 	zfs_ioctl_init();
6143 	rz_zev_init();
6144 
6145 	if ((error = mod_install(&modlinkage)) != 0) {
6146 		zvol_fini();
6147 		zfs_fini();
6148 		spa_fini();
6149 		return (error);
6150 	}
6151 
6152 	tsd_create(&zfs_fsyncer_key, NULL);
6153 	tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6154 	tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6155 
6156 	error = ldi_ident_from_mod(&modlinkage, &zfs_li);
6157 	ASSERT(error == 0);
6158 	mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6159 
6160 	return (0);
6161 }
6162 
6163 int
6164 _fini(void)
6165 {
6166 	int error;
6167 
6168 	if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
6169 		return (SET_ERROR(EBUSY));
6170 
6171 	if ((error = mod_remove(&modlinkage)) != 0)
6172 		return (error);
6173 
6174 	rz_zev_fini();
6175 	zvol_fini();
6176 	zfs_fini();
6177 	spa_fini();
6178 	if (zfs_nfsshare_inited)
6179 		(void) ddi_modclose(nfs_mod);
6180 	if (zfs_smbshare_inited)
6181 		(void) ddi_modclose(smbsrv_mod);
6182 	if (zfs_nfsshare_inited || zfs_smbshare_inited)
6183 		(void) ddi_modclose(sharefs_mod);
6184 
6185 	tsd_destroy(&zfs_fsyncer_key);
6186 	ldi_ident_release(zfs_li);
6187 	zfs_li = NULL;
6188 	mutex_destroy(&zfs_share_lock);
6189 
6190 	return (error);
6191 }
6192 
6193 int
6194 _info(struct modinfo *modinfop)
6195 {
6196 	return (mod_info(&modlinkage, modinfop));
6197 }
6198