xref: /titanic_50/usr/src/uts/common/fs/zfs/zfs_ioctl.c (revision 9d12795f87b63c2e39e87bff369182edd34677d3)
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 
185 #include "zfs_namecheck.h"
186 #include "zfs_prop.h"
187 #include "zfs_deleg.h"
188 #include "zfs_comutil.h"
189 
190 extern struct modlfs zfs_modlfs;
191 
192 extern void zfs_init(void);
193 extern void zfs_fini(void);
194 
195 ldi_ident_t zfs_li = NULL;
196 dev_info_t *zfs_dip;
197 
198 uint_t zfs_fsyncer_key;
199 extern uint_t rrw_tsd_key;
200 static uint_t zfs_allow_log_key;
201 
202 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
203 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
204 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
205 
206 typedef enum {
207 	NO_NAME,
208 	POOL_NAME,
209 	DATASET_NAME
210 } zfs_ioc_namecheck_t;
211 
212 typedef enum {
213 	POOL_CHECK_NONE		= 1 << 0,
214 	POOL_CHECK_SUSPENDED	= 1 << 1,
215 	POOL_CHECK_READONLY	= 1 << 2,
216 } zfs_ioc_poolcheck_t;
217 
218 typedef struct zfs_ioc_vec {
219 	zfs_ioc_legacy_func_t	*zvec_legacy_func;
220 	zfs_ioc_func_t		*zvec_func;
221 	zfs_secpolicy_func_t	*zvec_secpolicy;
222 	zfs_ioc_namecheck_t	zvec_namecheck;
223 	boolean_t		zvec_allow_log;
224 	zfs_ioc_poolcheck_t	zvec_pool_check;
225 	boolean_t		zvec_smush_outnvlist;
226 	const char		*zvec_name;
227 } zfs_ioc_vec_t;
228 
229 /* This array is indexed by zfs_userquota_prop_t */
230 static const char *userquota_perms[] = {
231 	ZFS_DELEG_PERM_USERUSED,
232 	ZFS_DELEG_PERM_USERQUOTA,
233 	ZFS_DELEG_PERM_GROUPUSED,
234 	ZFS_DELEG_PERM_GROUPQUOTA,
235 };
236 
237 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
238 static int zfs_check_settable(const char *name, nvpair_t *property,
239     cred_t *cr);
240 static int zfs_check_clearable(char *dataset, nvlist_t *props,
241     nvlist_t **errors);
242 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
243     boolean_t *);
244 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
245 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
246 
247 static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature);
248 
249 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
250 void
251 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
252 {
253 	const char *newfile;
254 	char buf[512];
255 	va_list adx;
256 
257 	/*
258 	 * Get rid of annoying "../common/" prefix to filename.
259 	 */
260 	newfile = strrchr(file, '/');
261 	if (newfile != NULL) {
262 		newfile = newfile + 1; /* Get rid of leading / */
263 	} else {
264 		newfile = file;
265 	}
266 
267 	va_start(adx, fmt);
268 	(void) vsnprintf(buf, sizeof (buf), fmt, adx);
269 	va_end(adx);
270 
271 	/*
272 	 * To get this data, use the zfs-dprintf probe as so:
273 	 * dtrace -q -n 'zfs-dprintf \
274 	 *	/stringof(arg0) == "dbuf.c"/ \
275 	 *	{printf("%s: %s", stringof(arg1), stringof(arg3))}'
276 	 * arg0 = file name
277 	 * arg1 = function name
278 	 * arg2 = line number
279 	 * arg3 = message
280 	 */
281 	DTRACE_PROBE4(zfs__dprintf,
282 	    char *, newfile, char *, func, int, line, char *, buf);
283 }
284 
285 static void
286 history_str_free(char *buf)
287 {
288 	kmem_free(buf, HIS_MAX_RECORD_LEN);
289 }
290 
291 static char *
292 history_str_get(zfs_cmd_t *zc)
293 {
294 	char *buf;
295 
296 	if (zc->zc_history == NULL)
297 		return (NULL);
298 
299 	buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
300 	if (copyinstr((void *)(uintptr_t)zc->zc_history,
301 	    buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
302 		history_str_free(buf);
303 		return (NULL);
304 	}
305 
306 	buf[HIS_MAX_RECORD_LEN -1] = '\0';
307 
308 	return (buf);
309 }
310 
311 /*
312  * Check to see if the named dataset is currently defined as bootable
313  */
314 static boolean_t
315 zfs_is_bootfs(const char *name)
316 {
317 	objset_t *os;
318 
319 	if (dmu_objset_hold(name, FTAG, &os) == 0) {
320 		boolean_t ret;
321 		ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
322 		dmu_objset_rele(os, FTAG);
323 		return (ret);
324 	}
325 	return (B_FALSE);
326 }
327 
328 /*
329  * Return non-zero if the spa version is less than requested version.
330  */
331 static int
332 zfs_earlier_version(const char *name, int version)
333 {
334 	spa_t *spa;
335 
336 	if (spa_open(name, &spa, FTAG) == 0) {
337 		if (spa_version(spa) < version) {
338 			spa_close(spa, FTAG);
339 			return (1);
340 		}
341 		spa_close(spa, FTAG);
342 	}
343 	return (0);
344 }
345 
346 /*
347  * Return TRUE if the ZPL version is less than requested version.
348  */
349 static boolean_t
350 zpl_earlier_version(const char *name, int version)
351 {
352 	objset_t *os;
353 	boolean_t rc = B_TRUE;
354 
355 	if (dmu_objset_hold(name, FTAG, &os) == 0) {
356 		uint64_t zplversion;
357 
358 		if (dmu_objset_type(os) != DMU_OST_ZFS) {
359 			dmu_objset_rele(os, FTAG);
360 			return (B_TRUE);
361 		}
362 		/* XXX reading from non-owned objset */
363 		if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
364 			rc = zplversion < version;
365 		dmu_objset_rele(os, FTAG);
366 	}
367 	return (rc);
368 }
369 
370 static void
371 zfs_log_history(zfs_cmd_t *zc)
372 {
373 	spa_t *spa;
374 	char *buf;
375 
376 	if ((buf = history_str_get(zc)) == NULL)
377 		return;
378 
379 	if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
380 		if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
381 			(void) spa_history_log(spa, buf);
382 		spa_close(spa, FTAG);
383 	}
384 	history_str_free(buf);
385 }
386 
387 /*
388  * Policy for top-level read operations (list pools).  Requires no privileges,
389  * and can be used in the local zone, as there is no associated dataset.
390  */
391 /* ARGSUSED */
392 static int
393 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
394 {
395 	return (0);
396 }
397 
398 /*
399  * Policy for dataset read operations (list children, get statistics).  Requires
400  * no privileges, but must be visible in the local zone.
401  */
402 /* ARGSUSED */
403 static int
404 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
405 {
406 	if (INGLOBALZONE(curproc) ||
407 	    zone_dataset_visible(zc->zc_name, NULL))
408 		return (0);
409 
410 	return (SET_ERROR(ENOENT));
411 }
412 
413 static int
414 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
415 {
416 	int writable = 1;
417 
418 	/*
419 	 * The dataset must be visible by this zone -- check this first
420 	 * so they don't see EPERM on something they shouldn't know about.
421 	 */
422 	if (!INGLOBALZONE(curproc) &&
423 	    !zone_dataset_visible(dataset, &writable))
424 		return (SET_ERROR(ENOENT));
425 
426 	if (INGLOBALZONE(curproc)) {
427 		/*
428 		 * If the fs is zoned, only root can access it from the
429 		 * global zone.
430 		 */
431 		if (secpolicy_zfs(cr) && zoned)
432 			return (SET_ERROR(EPERM));
433 	} else {
434 		/*
435 		 * If we are in a local zone, the 'zoned' property must be set.
436 		 */
437 		if (!zoned)
438 			return (SET_ERROR(EPERM));
439 
440 		/* must be writable by this zone */
441 		if (!writable)
442 			return (SET_ERROR(EPERM));
443 	}
444 	return (0);
445 }
446 
447 static int
448 zfs_dozonecheck(const char *dataset, cred_t *cr)
449 {
450 	uint64_t zoned;
451 
452 	if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
453 		return (SET_ERROR(ENOENT));
454 
455 	return (zfs_dozonecheck_impl(dataset, zoned, cr));
456 }
457 
458 static int
459 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
460 {
461 	uint64_t zoned;
462 
463 	if (dsl_prop_get_int_ds(ds, "zoned", &zoned))
464 		return (SET_ERROR(ENOENT));
465 
466 	return (zfs_dozonecheck_impl(dataset, zoned, cr));
467 }
468 
469 static int
470 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
471     const char *perm, cred_t *cr)
472 {
473 	int error;
474 
475 	error = zfs_dozonecheck_ds(name, ds, cr);
476 	if (error == 0) {
477 		error = secpolicy_zfs(cr);
478 		if (error != 0)
479 			error = dsl_deleg_access_impl(ds, perm, cr);
480 	}
481 	return (error);
482 }
483 
484 static int
485 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
486 {
487 	int error;
488 	dsl_dataset_t *ds;
489 	dsl_pool_t *dp;
490 
491 	error = dsl_pool_hold(name, FTAG, &dp);
492 	if (error != 0)
493 		return (error);
494 
495 	error = dsl_dataset_hold(dp, name, FTAG, &ds);
496 	if (error != 0) {
497 		dsl_pool_rele(dp, FTAG);
498 		return (error);
499 	}
500 
501 	error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
502 
503 	dsl_dataset_rele(ds, FTAG);
504 	dsl_pool_rele(dp, FTAG);
505 	return (error);
506 }
507 
508 /*
509  * Policy for setting the security label property.
510  *
511  * Returns 0 for success, non-zero for access and other errors.
512  */
513 static int
514 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
515 {
516 	char		ds_hexsl[MAXNAMELEN];
517 	bslabel_t	ds_sl, new_sl;
518 	boolean_t	new_default = FALSE;
519 	uint64_t	zoned;
520 	int		needed_priv = -1;
521 	int		error;
522 
523 	/* First get the existing dataset label. */
524 	error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
525 	    1, sizeof (ds_hexsl), &ds_hexsl, NULL);
526 	if (error != 0)
527 		return (SET_ERROR(EPERM));
528 
529 	if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
530 		new_default = TRUE;
531 
532 	/* The label must be translatable */
533 	if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
534 		return (SET_ERROR(EINVAL));
535 
536 	/*
537 	 * In a non-global zone, disallow attempts to set a label that
538 	 * doesn't match that of the zone; otherwise no other checks
539 	 * are needed.
540 	 */
541 	if (!INGLOBALZONE(curproc)) {
542 		if (new_default || !blequal(&new_sl, CR_SL(CRED())))
543 			return (SET_ERROR(EPERM));
544 		return (0);
545 	}
546 
547 	/*
548 	 * For global-zone datasets (i.e., those whose zoned property is
549 	 * "off", verify that the specified new label is valid for the
550 	 * global zone.
551 	 */
552 	if (dsl_prop_get_integer(name,
553 	    zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
554 		return (SET_ERROR(EPERM));
555 	if (!zoned) {
556 		if (zfs_check_global_label(name, strval) != 0)
557 			return (SET_ERROR(EPERM));
558 	}
559 
560 	/*
561 	 * If the existing dataset label is nondefault, check if the
562 	 * dataset is mounted (label cannot be changed while mounted).
563 	 * Get the zfsvfs; if there isn't one, then the dataset isn't
564 	 * mounted (or isn't a dataset, doesn't exist, ...).
565 	 */
566 	if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
567 		objset_t *os;
568 		static char *setsl_tag = "setsl_tag";
569 
570 		/*
571 		 * Try to own the dataset; abort if there is any error,
572 		 * (e.g., already mounted, in use, or other error).
573 		 */
574 		error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
575 		    setsl_tag, &os);
576 		if (error != 0)
577 			return (SET_ERROR(EPERM));
578 
579 		dmu_objset_disown(os, setsl_tag);
580 
581 		if (new_default) {
582 			needed_priv = PRIV_FILE_DOWNGRADE_SL;
583 			goto out_check;
584 		}
585 
586 		if (hexstr_to_label(strval, &new_sl) != 0)
587 			return (SET_ERROR(EPERM));
588 
589 		if (blstrictdom(&ds_sl, &new_sl))
590 			needed_priv = PRIV_FILE_DOWNGRADE_SL;
591 		else if (blstrictdom(&new_sl, &ds_sl))
592 			needed_priv = PRIV_FILE_UPGRADE_SL;
593 	} else {
594 		/* dataset currently has a default label */
595 		if (!new_default)
596 			needed_priv = PRIV_FILE_UPGRADE_SL;
597 	}
598 
599 out_check:
600 	if (needed_priv != -1)
601 		return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
602 	return (0);
603 }
604 
605 static int
606 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
607     cred_t *cr)
608 {
609 	char *strval;
610 
611 	/*
612 	 * Check permissions for special properties.
613 	 */
614 	switch (prop) {
615 	case ZFS_PROP_ZONED:
616 		/*
617 		 * Disallow setting of 'zoned' from within a local zone.
618 		 */
619 		if (!INGLOBALZONE(curproc))
620 			return (SET_ERROR(EPERM));
621 		break;
622 
623 	case ZFS_PROP_QUOTA:
624 	case ZFS_PROP_FILESYSTEM_LIMIT:
625 	case ZFS_PROP_SNAPSHOT_LIMIT:
626 		if (!INGLOBALZONE(curproc)) {
627 			uint64_t zoned;
628 			char setpoint[MAXNAMELEN];
629 			/*
630 			 * Unprivileged users are allowed to modify the
631 			 * limit on things *under* (ie. contained by)
632 			 * the thing they own.
633 			 */
634 			if (dsl_prop_get_integer(dsname, "zoned", &zoned,
635 			    setpoint))
636 				return (SET_ERROR(EPERM));
637 			if (!zoned || strlen(dsname) <= strlen(setpoint))
638 				return (SET_ERROR(EPERM));
639 		}
640 		break;
641 
642 	case ZFS_PROP_MLSLABEL:
643 		if (!is_system_labeled())
644 			return (SET_ERROR(EPERM));
645 
646 		if (nvpair_value_string(propval, &strval) == 0) {
647 			int err;
648 
649 			err = zfs_set_slabel_policy(dsname, strval, CRED());
650 			if (err != 0)
651 				return (err);
652 		}
653 		break;
654 	}
655 
656 	return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
657 }
658 
659 /* ARGSUSED */
660 static int
661 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
662 {
663 	int error;
664 
665 	error = zfs_dozonecheck(zc->zc_name, cr);
666 	if (error != 0)
667 		return (error);
668 
669 	/*
670 	 * permission to set permissions will be evaluated later in
671 	 * dsl_deleg_can_allow()
672 	 */
673 	return (0);
674 }
675 
676 /* ARGSUSED */
677 static int
678 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
679 {
680 	return (zfs_secpolicy_write_perms(zc->zc_name,
681 	    ZFS_DELEG_PERM_ROLLBACK, cr));
682 }
683 
684 /* ARGSUSED */
685 static int
686 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
687 {
688 	dsl_pool_t *dp;
689 	dsl_dataset_t *ds;
690 	char *cp;
691 	int error;
692 
693 	/*
694 	 * Generate the current snapshot name from the given objsetid, then
695 	 * use that name for the secpolicy/zone checks.
696 	 */
697 	cp = strchr(zc->zc_name, '@');
698 	if (cp == NULL)
699 		return (SET_ERROR(EINVAL));
700 	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
701 	if (error != 0)
702 		return (error);
703 
704 	error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
705 	if (error != 0) {
706 		dsl_pool_rele(dp, FTAG);
707 		return (error);
708 	}
709 
710 	dsl_dataset_name(ds, zc->zc_name);
711 
712 	error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
713 	    ZFS_DELEG_PERM_SEND, cr);
714 	dsl_dataset_rele(ds, FTAG);
715 	dsl_pool_rele(dp, FTAG);
716 
717 	return (error);
718 }
719 
720 /* ARGSUSED */
721 static int
722 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
723 {
724 	return (zfs_secpolicy_write_perms(zc->zc_name,
725 	    ZFS_DELEG_PERM_SEND, cr));
726 }
727 
728 /* ARGSUSED */
729 static int
730 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
731 {
732 	vnode_t *vp;
733 	int error;
734 
735 	if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
736 	    NO_FOLLOW, NULL, &vp)) != 0)
737 		return (error);
738 
739 	/* Now make sure mntpnt and dataset are ZFS */
740 
741 	if (vp->v_vfsp->vfs_fstype != zfsfstype ||
742 	    (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
743 	    zc->zc_name) != 0)) {
744 		VN_RELE(vp);
745 		return (SET_ERROR(EPERM));
746 	}
747 
748 	VN_RELE(vp);
749 	return (dsl_deleg_access(zc->zc_name,
750 	    ZFS_DELEG_PERM_SHARE, cr));
751 }
752 
753 int
754 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
755 {
756 	if (!INGLOBALZONE(curproc))
757 		return (SET_ERROR(EPERM));
758 
759 	if (secpolicy_nfs(cr) == 0) {
760 		return (0);
761 	} else {
762 		return (zfs_secpolicy_deleg_share(zc, innvl, cr));
763 	}
764 }
765 
766 int
767 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
768 {
769 	if (!INGLOBALZONE(curproc))
770 		return (SET_ERROR(EPERM));
771 
772 	if (secpolicy_smb(cr) == 0) {
773 		return (0);
774 	} else {
775 		return (zfs_secpolicy_deleg_share(zc, innvl, cr));
776 	}
777 }
778 
779 static int
780 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
781 {
782 	char *cp;
783 
784 	/*
785 	 * Remove the @bla or /bla from the end of the name to get the parent.
786 	 */
787 	(void) strncpy(parent, datasetname, parentsize);
788 	cp = strrchr(parent, '@');
789 	if (cp != NULL) {
790 		cp[0] = '\0';
791 	} else {
792 		cp = strrchr(parent, '/');
793 		if (cp == NULL)
794 			return (SET_ERROR(ENOENT));
795 		cp[0] = '\0';
796 	}
797 
798 	return (0);
799 }
800 
801 int
802 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
803 {
804 	int error;
805 
806 	if ((error = zfs_secpolicy_write_perms(name,
807 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
808 		return (error);
809 
810 	return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
811 }
812 
813 /* ARGSUSED */
814 static int
815 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
816 {
817 	return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
818 }
819 
820 /*
821  * Destroying snapshots with delegated permissions requires
822  * descendant mount and destroy permissions.
823  */
824 /* ARGSUSED */
825 static int
826 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
827 {
828 	nvlist_t *snaps;
829 	nvpair_t *pair, *nextpair;
830 	int error = 0;
831 
832 	if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
833 		return (SET_ERROR(EINVAL));
834 	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
835 	    pair = nextpair) {
836 		nextpair = nvlist_next_nvpair(snaps, pair);
837 		error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
838 		if (error == ENOENT) {
839 			/*
840 			 * Ignore any snapshots that don't exist (we consider
841 			 * them "already destroyed").  Remove the name from the
842 			 * nvl here in case the snapshot is created between
843 			 * now and when we try to destroy it (in which case
844 			 * we don't want to destroy it since we haven't
845 			 * checked for permission).
846 			 */
847 			fnvlist_remove_nvpair(snaps, pair);
848 			error = 0;
849 		}
850 		if (error != 0)
851 			break;
852 	}
853 
854 	return (error);
855 }
856 
857 int
858 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
859 {
860 	char	parentname[MAXNAMELEN];
861 	int	error;
862 
863 	if ((error = zfs_secpolicy_write_perms(from,
864 	    ZFS_DELEG_PERM_RENAME, cr)) != 0)
865 		return (error);
866 
867 	if ((error = zfs_secpolicy_write_perms(from,
868 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
869 		return (error);
870 
871 	if ((error = zfs_get_parent(to, parentname,
872 	    sizeof (parentname))) != 0)
873 		return (error);
874 
875 	if ((error = zfs_secpolicy_write_perms(parentname,
876 	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
877 		return (error);
878 
879 	if ((error = zfs_secpolicy_write_perms(parentname,
880 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
881 		return (error);
882 
883 	return (error);
884 }
885 
886 /* ARGSUSED */
887 static int
888 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
889 {
890 	return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
891 }
892 
893 /* ARGSUSED */
894 static int
895 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
896 {
897 	dsl_pool_t *dp;
898 	dsl_dataset_t *clone;
899 	int error;
900 
901 	error = zfs_secpolicy_write_perms(zc->zc_name,
902 	    ZFS_DELEG_PERM_PROMOTE, cr);
903 	if (error != 0)
904 		return (error);
905 
906 	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
907 	if (error != 0)
908 		return (error);
909 
910 	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
911 
912 	if (error == 0) {
913 		char parentname[MAXNAMELEN];
914 		dsl_dataset_t *origin = NULL;
915 		dsl_dir_t *dd;
916 		dd = clone->ds_dir;
917 
918 		error = dsl_dataset_hold_obj(dd->dd_pool,
919 		    dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin);
920 		if (error != 0) {
921 			dsl_dataset_rele(clone, FTAG);
922 			dsl_pool_rele(dp, FTAG);
923 			return (error);
924 		}
925 
926 		error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
927 		    ZFS_DELEG_PERM_MOUNT, cr);
928 
929 		dsl_dataset_name(origin, parentname);
930 		if (error == 0) {
931 			error = zfs_secpolicy_write_perms_ds(parentname, origin,
932 			    ZFS_DELEG_PERM_PROMOTE, cr);
933 		}
934 		dsl_dataset_rele(clone, FTAG);
935 		dsl_dataset_rele(origin, FTAG);
936 	}
937 	dsl_pool_rele(dp, FTAG);
938 	return (error);
939 }
940 
941 /* ARGSUSED */
942 static int
943 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
944 {
945 	int error;
946 
947 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
948 	    ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
949 		return (error);
950 
951 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
952 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
953 		return (error);
954 
955 	return (zfs_secpolicy_write_perms(zc->zc_name,
956 	    ZFS_DELEG_PERM_CREATE, cr));
957 }
958 
959 int
960 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
961 {
962 	return (zfs_secpolicy_write_perms(name,
963 	    ZFS_DELEG_PERM_SNAPSHOT, cr));
964 }
965 
966 /*
967  * Check for permission to create each snapshot in the nvlist.
968  */
969 /* ARGSUSED */
970 static int
971 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
972 {
973 	nvlist_t *snaps;
974 	int error = 0;
975 	nvpair_t *pair;
976 
977 	if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
978 		return (SET_ERROR(EINVAL));
979 	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
980 	    pair = nvlist_next_nvpair(snaps, pair)) {
981 		char *name = nvpair_name(pair);
982 		char *atp = strchr(name, '@');
983 
984 		if (atp == NULL) {
985 			error = SET_ERROR(EINVAL);
986 			break;
987 		}
988 		*atp = '\0';
989 		error = zfs_secpolicy_snapshot_perms(name, cr);
990 		*atp = '@';
991 		if (error != 0)
992 			break;
993 	}
994 	return (error);
995 }
996 
997 /*
998  * Check for permission to create each snapshot in the nvlist.
999  */
1000 /* ARGSUSED */
1001 static int
1002 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1003 {
1004 	int error = 0;
1005 
1006 	for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
1007 	    pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
1008 		char *name = nvpair_name(pair);
1009 		char *hashp = strchr(name, '#');
1010 
1011 		if (hashp == NULL) {
1012 			error = SET_ERROR(EINVAL);
1013 			break;
1014 		}
1015 		*hashp = '\0';
1016 		error = zfs_secpolicy_write_perms(name,
1017 		    ZFS_DELEG_PERM_BOOKMARK, cr);
1018 		*hashp = '#';
1019 		if (error != 0)
1020 			break;
1021 	}
1022 	return (error);
1023 }
1024 
1025 /* ARGSUSED */
1026 static int
1027 zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1028 {
1029 	nvpair_t *pair, *nextpair;
1030 	int error = 0;
1031 
1032 	for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1033 	    pair = nextpair) {
1034 		char *name = nvpair_name(pair);
1035 		char *hashp = strchr(name, '#');
1036 		nextpair = nvlist_next_nvpair(innvl, pair);
1037 
1038 		if (hashp == NULL) {
1039 			error = SET_ERROR(EINVAL);
1040 			break;
1041 		}
1042 
1043 		*hashp = '\0';
1044 		error = zfs_secpolicy_write_perms(name,
1045 		    ZFS_DELEG_PERM_DESTROY, cr);
1046 		*hashp = '#';
1047 		if (error == ENOENT) {
1048 			/*
1049 			 * Ignore any filesystems that don't exist (we consider
1050 			 * their bookmarks "already destroyed").  Remove
1051 			 * the name from the nvl here in case the filesystem
1052 			 * is created between now and when we try to destroy
1053 			 * the bookmark (in which case we don't want to
1054 			 * destroy it since we haven't checked for permission).
1055 			 */
1056 			fnvlist_remove_nvpair(innvl, pair);
1057 			error = 0;
1058 		}
1059 		if (error != 0)
1060 			break;
1061 	}
1062 
1063 	return (error);
1064 }
1065 
1066 /* ARGSUSED */
1067 static int
1068 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1069 {
1070 	/*
1071 	 * Even root must have a proper TSD so that we know what pool
1072 	 * to log to.
1073 	 */
1074 	if (tsd_get(zfs_allow_log_key) == NULL)
1075 		return (SET_ERROR(EPERM));
1076 	return (0);
1077 }
1078 
1079 static int
1080 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1081 {
1082 	char	parentname[MAXNAMELEN];
1083 	int	error;
1084 	char	*origin;
1085 
1086 	if ((error = zfs_get_parent(zc->zc_name, parentname,
1087 	    sizeof (parentname))) != 0)
1088 		return (error);
1089 
1090 	if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1091 	    (error = zfs_secpolicy_write_perms(origin,
1092 	    ZFS_DELEG_PERM_CLONE, cr)) != 0)
1093 		return (error);
1094 
1095 	if ((error = zfs_secpolicy_write_perms(parentname,
1096 	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
1097 		return (error);
1098 
1099 	return (zfs_secpolicy_write_perms(parentname,
1100 	    ZFS_DELEG_PERM_MOUNT, cr));
1101 }
1102 
1103 /*
1104  * Policy for pool operations - create/destroy pools, add vdevs, etc.  Requires
1105  * SYS_CONFIG privilege, which is not available in a local zone.
1106  */
1107 /* ARGSUSED */
1108 static int
1109 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1110 {
1111 	if (secpolicy_sys_config(cr, B_FALSE) != 0)
1112 		return (SET_ERROR(EPERM));
1113 
1114 	return (0);
1115 }
1116 
1117 /*
1118  * Policy for object to name lookups.
1119  */
1120 /* ARGSUSED */
1121 static int
1122 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1123 {
1124 	int error;
1125 
1126 	if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1127 		return (0);
1128 
1129 	error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1130 	return (error);
1131 }
1132 
1133 /*
1134  * Policy for fault injection.  Requires all privileges.
1135  */
1136 /* ARGSUSED */
1137 static int
1138 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1139 {
1140 	return (secpolicy_zinject(cr));
1141 }
1142 
1143 /* ARGSUSED */
1144 static int
1145 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1146 {
1147 	zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1148 
1149 	if (prop == ZPROP_INVAL) {
1150 		if (!zfs_prop_user(zc->zc_value))
1151 			return (SET_ERROR(EINVAL));
1152 		return (zfs_secpolicy_write_perms(zc->zc_name,
1153 		    ZFS_DELEG_PERM_USERPROP, cr));
1154 	} else {
1155 		return (zfs_secpolicy_setprop(zc->zc_name, prop,
1156 		    NULL, cr));
1157 	}
1158 }
1159 
1160 static int
1161 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1162 {
1163 	int err = zfs_secpolicy_read(zc, innvl, cr);
1164 	if (err)
1165 		return (err);
1166 
1167 	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1168 		return (SET_ERROR(EINVAL));
1169 
1170 	if (zc->zc_value[0] == 0) {
1171 		/*
1172 		 * They are asking about a posix uid/gid.  If it's
1173 		 * themself, allow it.
1174 		 */
1175 		if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1176 		    zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
1177 			if (zc->zc_guid == crgetuid(cr))
1178 				return (0);
1179 		} else {
1180 			if (groupmember(zc->zc_guid, cr))
1181 				return (0);
1182 		}
1183 	}
1184 
1185 	return (zfs_secpolicy_write_perms(zc->zc_name,
1186 	    userquota_perms[zc->zc_objset_type], cr));
1187 }
1188 
1189 static int
1190 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1191 {
1192 	int err = zfs_secpolicy_read(zc, innvl, cr);
1193 	if (err)
1194 		return (err);
1195 
1196 	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1197 		return (SET_ERROR(EINVAL));
1198 
1199 	return (zfs_secpolicy_write_perms(zc->zc_name,
1200 	    userquota_perms[zc->zc_objset_type], cr));
1201 }
1202 
1203 /* ARGSUSED */
1204 static int
1205 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1206 {
1207 	return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1208 	    NULL, cr));
1209 }
1210 
1211 /* ARGSUSED */
1212 static int
1213 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1214 {
1215 	nvpair_t *pair;
1216 	nvlist_t *holds;
1217 	int error;
1218 
1219 	error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1220 	if (error != 0)
1221 		return (SET_ERROR(EINVAL));
1222 
1223 	for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1224 	    pair = nvlist_next_nvpair(holds, pair)) {
1225 		char fsname[MAXNAMELEN];
1226 		error = dmu_fsname(nvpair_name(pair), fsname);
1227 		if (error != 0)
1228 			return (error);
1229 		error = zfs_secpolicy_write_perms(fsname,
1230 		    ZFS_DELEG_PERM_HOLD, cr);
1231 		if (error != 0)
1232 			return (error);
1233 	}
1234 	return (0);
1235 }
1236 
1237 /* ARGSUSED */
1238 static int
1239 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1240 {
1241 	nvpair_t *pair;
1242 	int error;
1243 
1244 	for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1245 	    pair = nvlist_next_nvpair(innvl, pair)) {
1246 		char fsname[MAXNAMELEN];
1247 		error = dmu_fsname(nvpair_name(pair), fsname);
1248 		if (error != 0)
1249 			return (error);
1250 		error = zfs_secpolicy_write_perms(fsname,
1251 		    ZFS_DELEG_PERM_RELEASE, cr);
1252 		if (error != 0)
1253 			return (error);
1254 	}
1255 	return (0);
1256 }
1257 
1258 /*
1259  * Policy for allowing temporary snapshots to be taken or released
1260  */
1261 static int
1262 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1263 {
1264 	/*
1265 	 * A temporary snapshot is the same as a snapshot,
1266 	 * hold, destroy and release all rolled into one.
1267 	 * Delegated diff alone is sufficient that we allow this.
1268 	 */
1269 	int error;
1270 
1271 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1272 	    ZFS_DELEG_PERM_DIFF, cr)) == 0)
1273 		return (0);
1274 
1275 	error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1276 	if (error == 0)
1277 		error = zfs_secpolicy_hold(zc, innvl, cr);
1278 	if (error == 0)
1279 		error = zfs_secpolicy_release(zc, innvl, cr);
1280 	if (error == 0)
1281 		error = zfs_secpolicy_destroy(zc, innvl, cr);
1282 	return (error);
1283 }
1284 
1285 /*
1286  * Returns the nvlist as specified by the user in the zfs_cmd_t.
1287  */
1288 static int
1289 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1290 {
1291 	char *packed;
1292 	int error;
1293 	nvlist_t *list = NULL;
1294 
1295 	/*
1296 	 * Read in and unpack the user-supplied nvlist.
1297 	 */
1298 	if (size == 0)
1299 		return (SET_ERROR(EINVAL));
1300 
1301 	packed = kmem_alloc(size, KM_SLEEP);
1302 
1303 	if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1304 	    iflag)) != 0) {
1305 		kmem_free(packed, size);
1306 		return (error);
1307 	}
1308 
1309 	if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1310 		kmem_free(packed, size);
1311 		return (error);
1312 	}
1313 
1314 	kmem_free(packed, size);
1315 
1316 	*nvp = list;
1317 	return (0);
1318 }
1319 
1320 /*
1321  * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1322  * Entries will be removed from the end of the nvlist, and one int32 entry
1323  * named "N_MORE_ERRORS" will be added indicating how many entries were
1324  * removed.
1325  */
1326 static int
1327 nvlist_smush(nvlist_t *errors, size_t max)
1328 {
1329 	size_t size;
1330 
1331 	size = fnvlist_size(errors);
1332 
1333 	if (size > max) {
1334 		nvpair_t *more_errors;
1335 		int n = 0;
1336 
1337 		if (max < 1024)
1338 			return (SET_ERROR(ENOMEM));
1339 
1340 		fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1341 		more_errors = nvlist_prev_nvpair(errors, NULL);
1342 
1343 		do {
1344 			nvpair_t *pair = nvlist_prev_nvpair(errors,
1345 			    more_errors);
1346 			fnvlist_remove_nvpair(errors, pair);
1347 			n++;
1348 			size = fnvlist_size(errors);
1349 		} while (size > max);
1350 
1351 		fnvlist_remove_nvpair(errors, more_errors);
1352 		fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1353 		ASSERT3U(fnvlist_size(errors), <=, max);
1354 	}
1355 
1356 	return (0);
1357 }
1358 
1359 static int
1360 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1361 {
1362 	char *packed = NULL;
1363 	int error = 0;
1364 	size_t size;
1365 
1366 	size = fnvlist_size(nvl);
1367 
1368 	if (size > zc->zc_nvlist_dst_size) {
1369 		error = SET_ERROR(ENOMEM);
1370 	} else {
1371 		packed = fnvlist_pack(nvl, &size);
1372 		if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1373 		    size, zc->zc_iflags) != 0)
1374 			error = SET_ERROR(EFAULT);
1375 		fnvlist_pack_free(packed, size);
1376 	}
1377 
1378 	zc->zc_nvlist_dst_size = size;
1379 	zc->zc_nvlist_dst_filled = B_TRUE;
1380 	return (error);
1381 }
1382 
1383 static int
1384 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1385 {
1386 	objset_t *os;
1387 	int error;
1388 
1389 	error = dmu_objset_hold(dsname, FTAG, &os);
1390 	if (error != 0)
1391 		return (error);
1392 	if (dmu_objset_type(os) != DMU_OST_ZFS) {
1393 		dmu_objset_rele(os, FTAG);
1394 		return (SET_ERROR(EINVAL));
1395 	}
1396 
1397 	mutex_enter(&os->os_user_ptr_lock);
1398 	*zfvp = dmu_objset_get_user(os);
1399 	if (*zfvp) {
1400 		VFS_HOLD((*zfvp)->z_vfs);
1401 	} else {
1402 		error = SET_ERROR(ESRCH);
1403 	}
1404 	mutex_exit(&os->os_user_ptr_lock);
1405 	dmu_objset_rele(os, FTAG);
1406 	return (error);
1407 }
1408 
1409 /*
1410  * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1411  * case its z_vfs will be NULL, and it will be opened as the owner.
1412  * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1413  * which prevents all vnode ops from running.
1414  */
1415 static int
1416 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1417 {
1418 	int error = 0;
1419 
1420 	if (getzfsvfs(name, zfvp) != 0)
1421 		error = zfsvfs_create(name, zfvp);
1422 	if (error == 0) {
1423 		rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1424 		    RW_READER, tag);
1425 		if ((*zfvp)->z_unmounted) {
1426 			/*
1427 			 * XXX we could probably try again, since the unmounting
1428 			 * thread should be just about to disassociate the
1429 			 * objset from the zfsvfs.
1430 			 */
1431 			rrm_exit(&(*zfvp)->z_teardown_lock, tag);
1432 			return (SET_ERROR(EBUSY));
1433 		}
1434 	}
1435 	return (error);
1436 }
1437 
1438 static void
1439 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1440 {
1441 	rrm_exit(&zfsvfs->z_teardown_lock, tag);
1442 
1443 	if (zfsvfs->z_vfs) {
1444 		VFS_RELE(zfsvfs->z_vfs);
1445 	} else {
1446 		dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1447 		zfsvfs_free(zfsvfs);
1448 	}
1449 }
1450 
1451 static int
1452 zfs_ioc_pool_create(zfs_cmd_t *zc)
1453 {
1454 	int error;
1455 	nvlist_t *config, *props = NULL;
1456 	nvlist_t *rootprops = NULL;
1457 	nvlist_t *zplprops = NULL;
1458 
1459 	if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1460 	    zc->zc_iflags, &config))
1461 		return (error);
1462 
1463 	if (zc->zc_nvlist_src_size != 0 && (error =
1464 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1465 	    zc->zc_iflags, &props))) {
1466 		nvlist_free(config);
1467 		return (error);
1468 	}
1469 
1470 	if (props) {
1471 		nvlist_t *nvl = NULL;
1472 		uint64_t version = SPA_VERSION;
1473 
1474 		(void) nvlist_lookup_uint64(props,
1475 		    zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1476 		if (!SPA_VERSION_IS_SUPPORTED(version)) {
1477 			error = SET_ERROR(EINVAL);
1478 			goto pool_props_bad;
1479 		}
1480 		(void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1481 		if (nvl) {
1482 			error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1483 			if (error != 0) {
1484 				nvlist_free(config);
1485 				nvlist_free(props);
1486 				return (error);
1487 			}
1488 			(void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1489 		}
1490 		VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1491 		error = zfs_fill_zplprops_root(version, rootprops,
1492 		    zplprops, NULL);
1493 		if (error != 0)
1494 			goto pool_props_bad;
1495 	}
1496 
1497 	error = spa_create(zc->zc_name, config, props, zplprops);
1498 
1499 	/*
1500 	 * Set the remaining root properties
1501 	 */
1502 	if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1503 	    ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1504 		(void) spa_destroy(zc->zc_name);
1505 
1506 pool_props_bad:
1507 	nvlist_free(rootprops);
1508 	nvlist_free(zplprops);
1509 	nvlist_free(config);
1510 	nvlist_free(props);
1511 
1512 	return (error);
1513 }
1514 
1515 static int
1516 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1517 {
1518 	int error;
1519 	zfs_log_history(zc);
1520 	error = spa_destroy(zc->zc_name);
1521 	if (error == 0)
1522 		zvol_remove_minors(zc->zc_name);
1523 	return (error);
1524 }
1525 
1526 static int
1527 zfs_ioc_pool_import(zfs_cmd_t *zc)
1528 {
1529 	nvlist_t *config, *props = NULL;
1530 	uint64_t guid;
1531 	int error;
1532 
1533 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1534 	    zc->zc_iflags, &config)) != 0)
1535 		return (error);
1536 
1537 	if (zc->zc_nvlist_src_size != 0 && (error =
1538 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1539 	    zc->zc_iflags, &props))) {
1540 		nvlist_free(config);
1541 		return (error);
1542 	}
1543 
1544 	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1545 	    guid != zc->zc_guid)
1546 		error = SET_ERROR(EINVAL);
1547 	else
1548 		error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1549 
1550 	if (zc->zc_nvlist_dst != 0) {
1551 		int err;
1552 
1553 		if ((err = put_nvlist(zc, config)) != 0)
1554 			error = err;
1555 	}
1556 
1557 	nvlist_free(config);
1558 
1559 	if (props)
1560 		nvlist_free(props);
1561 
1562 	return (error);
1563 }
1564 
1565 static int
1566 zfs_ioc_pool_export(zfs_cmd_t *zc)
1567 {
1568 	int error;
1569 	boolean_t force = (boolean_t)zc->zc_cookie;
1570 	boolean_t hardforce = (boolean_t)zc->zc_guid;
1571 
1572 	zfs_log_history(zc);
1573 	error = spa_export(zc->zc_name, NULL, force, hardforce);
1574 	if (error == 0)
1575 		zvol_remove_minors(zc->zc_name);
1576 	return (error);
1577 }
1578 
1579 static int
1580 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1581 {
1582 	nvlist_t *configs;
1583 	int error;
1584 
1585 	if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1586 		return (SET_ERROR(EEXIST));
1587 
1588 	error = put_nvlist(zc, configs);
1589 
1590 	nvlist_free(configs);
1591 
1592 	return (error);
1593 }
1594 
1595 /*
1596  * inputs:
1597  * zc_name		name of the pool
1598  *
1599  * outputs:
1600  * zc_cookie		real errno
1601  * zc_nvlist_dst	config nvlist
1602  * zc_nvlist_dst_size	size of config nvlist
1603  */
1604 static int
1605 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1606 {
1607 	nvlist_t *config;
1608 	int error;
1609 	int ret = 0;
1610 
1611 	error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1612 	    sizeof (zc->zc_value));
1613 
1614 	if (config != NULL) {
1615 		ret = put_nvlist(zc, config);
1616 		nvlist_free(config);
1617 
1618 		/*
1619 		 * The config may be present even if 'error' is non-zero.
1620 		 * In this case we return success, and preserve the real errno
1621 		 * in 'zc_cookie'.
1622 		 */
1623 		zc->zc_cookie = error;
1624 	} else {
1625 		ret = error;
1626 	}
1627 
1628 	return (ret);
1629 }
1630 
1631 /*
1632  * Try to import the given pool, returning pool stats as appropriate so that
1633  * user land knows which devices are available and overall pool health.
1634  */
1635 static int
1636 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1637 {
1638 	nvlist_t *tryconfig, *config;
1639 	int error;
1640 
1641 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1642 	    zc->zc_iflags, &tryconfig)) != 0)
1643 		return (error);
1644 
1645 	config = spa_tryimport(tryconfig);
1646 
1647 	nvlist_free(tryconfig);
1648 
1649 	if (config == NULL)
1650 		return (SET_ERROR(EINVAL));
1651 
1652 	error = put_nvlist(zc, config);
1653 	nvlist_free(config);
1654 
1655 	return (error);
1656 }
1657 
1658 /*
1659  * inputs:
1660  * zc_name              name of the pool
1661  * zc_cookie            scan func (pool_scan_func_t)
1662  */
1663 static int
1664 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1665 {
1666 	spa_t *spa;
1667 	int error;
1668 
1669 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1670 		return (error);
1671 
1672 	if (zc->zc_cookie == POOL_SCAN_NONE)
1673 		error = spa_scan_stop(spa);
1674 	else
1675 		error = spa_scan(spa, zc->zc_cookie);
1676 
1677 	spa_close(spa, FTAG);
1678 
1679 	return (error);
1680 }
1681 
1682 static int
1683 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1684 {
1685 	spa_t *spa;
1686 	int error;
1687 
1688 	error = spa_open(zc->zc_name, &spa, FTAG);
1689 	if (error == 0) {
1690 		spa_freeze(spa);
1691 		spa_close(spa, FTAG);
1692 	}
1693 	return (error);
1694 }
1695 
1696 static int
1697 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1698 {
1699 	spa_t *spa;
1700 	int error;
1701 
1702 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1703 		return (error);
1704 
1705 	if (zc->zc_cookie < spa_version(spa) ||
1706 	    !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1707 		spa_close(spa, FTAG);
1708 		return (SET_ERROR(EINVAL));
1709 	}
1710 
1711 	spa_upgrade(spa, zc->zc_cookie);
1712 	spa_close(spa, FTAG);
1713 
1714 	return (error);
1715 }
1716 
1717 static int
1718 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1719 {
1720 	spa_t *spa;
1721 	char *hist_buf;
1722 	uint64_t size;
1723 	int error;
1724 
1725 	if ((size = zc->zc_history_len) == 0)
1726 		return (SET_ERROR(EINVAL));
1727 
1728 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1729 		return (error);
1730 
1731 	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1732 		spa_close(spa, FTAG);
1733 		return (SET_ERROR(ENOTSUP));
1734 	}
1735 
1736 	hist_buf = kmem_alloc(size, KM_SLEEP);
1737 	if ((error = spa_history_get(spa, &zc->zc_history_offset,
1738 	    &zc->zc_history_len, hist_buf)) == 0) {
1739 		error = ddi_copyout(hist_buf,
1740 		    (void *)(uintptr_t)zc->zc_history,
1741 		    zc->zc_history_len, zc->zc_iflags);
1742 	}
1743 
1744 	spa_close(spa, FTAG);
1745 	kmem_free(hist_buf, size);
1746 	return (error);
1747 }
1748 
1749 static int
1750 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1751 {
1752 	spa_t *spa;
1753 	int error;
1754 
1755 	error = spa_open(zc->zc_name, &spa, FTAG);
1756 	if (error == 0) {
1757 		error = spa_change_guid(spa);
1758 		spa_close(spa, FTAG);
1759 	}
1760 	return (error);
1761 }
1762 
1763 static int
1764 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1765 {
1766 	return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1767 }
1768 
1769 /*
1770  * inputs:
1771  * zc_name		name of filesystem
1772  * zc_obj		object to find
1773  *
1774  * outputs:
1775  * zc_value		name of object
1776  */
1777 static int
1778 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1779 {
1780 	objset_t *os;
1781 	int error;
1782 
1783 	/* XXX reading from objset not owned */
1784 	if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1785 		return (error);
1786 	if (dmu_objset_type(os) != DMU_OST_ZFS) {
1787 		dmu_objset_rele(os, FTAG);
1788 		return (SET_ERROR(EINVAL));
1789 	}
1790 	error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1791 	    sizeof (zc->zc_value));
1792 	dmu_objset_rele(os, FTAG);
1793 
1794 	return (error);
1795 }
1796 
1797 /*
1798  * inputs:
1799  * zc_name		name of filesystem
1800  * zc_obj		object to find
1801  *
1802  * outputs:
1803  * zc_stat		stats on object
1804  * zc_value		path to object
1805  */
1806 static int
1807 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1808 {
1809 	objset_t *os;
1810 	int error;
1811 
1812 	/* XXX reading from objset not owned */
1813 	if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1814 		return (error);
1815 	if (dmu_objset_type(os) != DMU_OST_ZFS) {
1816 		dmu_objset_rele(os, FTAG);
1817 		return (SET_ERROR(EINVAL));
1818 	}
1819 	error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1820 	    sizeof (zc->zc_value));
1821 	dmu_objset_rele(os, FTAG);
1822 
1823 	return (error);
1824 }
1825 
1826 static int
1827 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1828 {
1829 	spa_t *spa;
1830 	int error;
1831 	nvlist_t *config, **l2cache, **spares;
1832 	uint_t nl2cache = 0, nspares = 0;
1833 
1834 	error = spa_open(zc->zc_name, &spa, FTAG);
1835 	if (error != 0)
1836 		return (error);
1837 
1838 	error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1839 	    zc->zc_iflags, &config);
1840 	(void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1841 	    &l2cache, &nl2cache);
1842 
1843 	(void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1844 	    &spares, &nspares);
1845 
1846 	/*
1847 	 * A root pool with concatenated devices is not supported.
1848 	 * Thus, can not add a device to a root pool.
1849 	 *
1850 	 * Intent log device can not be added to a rootpool because
1851 	 * during mountroot, zil is replayed, a seperated log device
1852 	 * can not be accessed during the mountroot time.
1853 	 *
1854 	 * l2cache and spare devices are ok to be added to a rootpool.
1855 	 */
1856 	if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1857 		nvlist_free(config);
1858 		spa_close(spa, FTAG);
1859 		return (SET_ERROR(EDOM));
1860 	}
1861 
1862 	if (error == 0) {
1863 		error = spa_vdev_add(spa, config);
1864 		nvlist_free(config);
1865 	}
1866 	spa_close(spa, FTAG);
1867 	return (error);
1868 }
1869 
1870 /*
1871  * inputs:
1872  * zc_name		name of the pool
1873  * zc_nvlist_conf	nvlist of devices to remove
1874  * zc_cookie		to stop the remove?
1875  */
1876 static int
1877 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1878 {
1879 	spa_t *spa;
1880 	int error;
1881 
1882 	error = spa_open(zc->zc_name, &spa, FTAG);
1883 	if (error != 0)
1884 		return (error);
1885 	error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1886 	spa_close(spa, FTAG);
1887 	return (error);
1888 }
1889 
1890 static int
1891 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1892 {
1893 	spa_t *spa;
1894 	int error;
1895 	vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1896 
1897 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1898 		return (error);
1899 	switch (zc->zc_cookie) {
1900 	case VDEV_STATE_ONLINE:
1901 		error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1902 		break;
1903 
1904 	case VDEV_STATE_OFFLINE:
1905 		error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1906 		break;
1907 
1908 	case VDEV_STATE_FAULTED:
1909 		if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1910 		    zc->zc_obj != VDEV_AUX_EXTERNAL)
1911 			zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1912 
1913 		error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1914 		break;
1915 
1916 	case VDEV_STATE_DEGRADED:
1917 		if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1918 		    zc->zc_obj != VDEV_AUX_EXTERNAL)
1919 			zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1920 
1921 		error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1922 		break;
1923 
1924 	default:
1925 		error = SET_ERROR(EINVAL);
1926 	}
1927 	zc->zc_cookie = newstate;
1928 	spa_close(spa, FTAG);
1929 	return (error);
1930 }
1931 
1932 static int
1933 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1934 {
1935 	spa_t *spa;
1936 	int replacing = zc->zc_cookie;
1937 	nvlist_t *config;
1938 	int error;
1939 
1940 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1941 		return (error);
1942 
1943 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1944 	    zc->zc_iflags, &config)) == 0) {
1945 		error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1946 		nvlist_free(config);
1947 	}
1948 
1949 	spa_close(spa, FTAG);
1950 	return (error);
1951 }
1952 
1953 static int
1954 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1955 {
1956 	spa_t *spa;
1957 	int error;
1958 
1959 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1960 		return (error);
1961 
1962 	error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1963 
1964 	spa_close(spa, FTAG);
1965 	return (error);
1966 }
1967 
1968 static int
1969 zfs_ioc_vdev_split(zfs_cmd_t *zc)
1970 {
1971 	spa_t *spa;
1972 	nvlist_t *config, *props = NULL;
1973 	int error;
1974 	boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
1975 
1976 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1977 		return (error);
1978 
1979 	if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1980 	    zc->zc_iflags, &config)) {
1981 		spa_close(spa, FTAG);
1982 		return (error);
1983 	}
1984 
1985 	if (zc->zc_nvlist_src_size != 0 && (error =
1986 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1987 	    zc->zc_iflags, &props))) {
1988 		spa_close(spa, FTAG);
1989 		nvlist_free(config);
1990 		return (error);
1991 	}
1992 
1993 	error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
1994 
1995 	spa_close(spa, FTAG);
1996 
1997 	nvlist_free(config);
1998 	nvlist_free(props);
1999 
2000 	return (error);
2001 }
2002 
2003 static int
2004 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
2005 {
2006 	spa_t *spa;
2007 	char *path = zc->zc_value;
2008 	uint64_t guid = zc->zc_guid;
2009 	int error;
2010 
2011 	error = spa_open(zc->zc_name, &spa, FTAG);
2012 	if (error != 0)
2013 		return (error);
2014 
2015 	error = spa_vdev_setpath(spa, guid, path);
2016 	spa_close(spa, FTAG);
2017 	return (error);
2018 }
2019 
2020 static int
2021 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
2022 {
2023 	spa_t *spa;
2024 	char *fru = zc->zc_value;
2025 	uint64_t guid = zc->zc_guid;
2026 	int error;
2027 
2028 	error = spa_open(zc->zc_name, &spa, FTAG);
2029 	if (error != 0)
2030 		return (error);
2031 
2032 	error = spa_vdev_setfru(spa, guid, fru);
2033 	spa_close(spa, FTAG);
2034 	return (error);
2035 }
2036 
2037 static int
2038 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
2039 {
2040 	int error = 0;
2041 	nvlist_t *nv;
2042 
2043 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2044 
2045 	if (zc->zc_nvlist_dst != 0 &&
2046 	    (error = dsl_prop_get_all(os, &nv)) == 0) {
2047 		dmu_objset_stats(os, nv);
2048 		/*
2049 		 * NB: zvol_get_stats() will read the objset contents,
2050 		 * which we aren't supposed to do with a
2051 		 * DS_MODE_USER hold, because it could be
2052 		 * inconsistent.  So this is a bit of a workaround...
2053 		 * XXX reading with out owning
2054 		 */
2055 		if (!zc->zc_objset_stats.dds_inconsistent &&
2056 		    dmu_objset_type(os) == DMU_OST_ZVOL) {
2057 			error = zvol_get_stats(os, nv);
2058 			if (error == EIO)
2059 				return (error);
2060 			VERIFY0(error);
2061 		}
2062 		error = put_nvlist(zc, nv);
2063 		nvlist_free(nv);
2064 	}
2065 
2066 	return (error);
2067 }
2068 
2069 /*
2070  * inputs:
2071  * zc_name		name of filesystem
2072  * zc_nvlist_dst_size	size of buffer for property nvlist
2073  *
2074  * outputs:
2075  * zc_objset_stats	stats
2076  * zc_nvlist_dst	property nvlist
2077  * zc_nvlist_dst_size	size of property nvlist
2078  */
2079 static int
2080 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2081 {
2082 	objset_t *os;
2083 	int error;
2084 
2085 	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2086 	if (error == 0) {
2087 		error = zfs_ioc_objset_stats_impl(zc, os);
2088 		dmu_objset_rele(os, FTAG);
2089 	}
2090 
2091 	return (error);
2092 }
2093 
2094 /*
2095  * inputs:
2096  * zc_name		name of filesystem
2097  * zc_nvlist_dst_size	size of buffer for property nvlist
2098  *
2099  * outputs:
2100  * zc_nvlist_dst	received property nvlist
2101  * zc_nvlist_dst_size	size of received property nvlist
2102  *
2103  * Gets received properties (distinct from local properties on or after
2104  * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2105  * local property values.
2106  */
2107 static int
2108 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2109 {
2110 	int error = 0;
2111 	nvlist_t *nv;
2112 
2113 	/*
2114 	 * Without this check, we would return local property values if the
2115 	 * caller has not already received properties on or after
2116 	 * SPA_VERSION_RECVD_PROPS.
2117 	 */
2118 	if (!dsl_prop_get_hasrecvd(zc->zc_name))
2119 		return (SET_ERROR(ENOTSUP));
2120 
2121 	if (zc->zc_nvlist_dst != 0 &&
2122 	    (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2123 		error = put_nvlist(zc, nv);
2124 		nvlist_free(nv);
2125 	}
2126 
2127 	return (error);
2128 }
2129 
2130 static int
2131 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2132 {
2133 	uint64_t value;
2134 	int error;
2135 
2136 	/*
2137 	 * zfs_get_zplprop() will either find a value or give us
2138 	 * the default value (if there is one).
2139 	 */
2140 	if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2141 		return (error);
2142 	VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2143 	return (0);
2144 }
2145 
2146 /*
2147  * inputs:
2148  * zc_name		name of filesystem
2149  * zc_nvlist_dst_size	size of buffer for zpl property nvlist
2150  *
2151  * outputs:
2152  * zc_nvlist_dst	zpl property nvlist
2153  * zc_nvlist_dst_size	size of zpl property nvlist
2154  */
2155 static int
2156 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2157 {
2158 	objset_t *os;
2159 	int err;
2160 
2161 	/* XXX reading without owning */
2162 	if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2163 		return (err);
2164 
2165 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2166 
2167 	/*
2168 	 * NB: nvl_add_zplprop() will read the objset contents,
2169 	 * which we aren't supposed to do with a DS_MODE_USER
2170 	 * hold, because it could be inconsistent.
2171 	 */
2172 	if (zc->zc_nvlist_dst != NULL &&
2173 	    !zc->zc_objset_stats.dds_inconsistent &&
2174 	    dmu_objset_type(os) == DMU_OST_ZFS) {
2175 		nvlist_t *nv;
2176 
2177 		VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2178 		if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2179 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2180 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2181 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2182 			err = put_nvlist(zc, nv);
2183 		nvlist_free(nv);
2184 	} else {
2185 		err = SET_ERROR(ENOENT);
2186 	}
2187 	dmu_objset_rele(os, FTAG);
2188 	return (err);
2189 }
2190 
2191 static boolean_t
2192 dataset_name_hidden(const char *name)
2193 {
2194 	/*
2195 	 * Skip over datasets that are not visible in this zone,
2196 	 * internal datasets (which have a $ in their name), and
2197 	 * temporary datasets (which have a % in their name).
2198 	 */
2199 	if (strchr(name, '$') != NULL)
2200 		return (B_TRUE);
2201 	if (strchr(name, '%') != NULL)
2202 		return (B_TRUE);
2203 	if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
2204 		return (B_TRUE);
2205 	return (B_FALSE);
2206 }
2207 
2208 /*
2209  * inputs:
2210  * zc_name		name of filesystem
2211  * zc_cookie		zap cursor
2212  * zc_nvlist_dst_size	size of buffer for property nvlist
2213  *
2214  * outputs:
2215  * zc_name		name of next filesystem
2216  * zc_cookie		zap cursor
2217  * zc_objset_stats	stats
2218  * zc_nvlist_dst	property nvlist
2219  * zc_nvlist_dst_size	size of property nvlist
2220  */
2221 static int
2222 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2223 {
2224 	objset_t *os;
2225 	int error;
2226 	char *p;
2227 	size_t orig_len = strlen(zc->zc_name);
2228 
2229 top:
2230 	if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2231 		if (error == ENOENT)
2232 			error = SET_ERROR(ESRCH);
2233 		return (error);
2234 	}
2235 
2236 	p = strrchr(zc->zc_name, '/');
2237 	if (p == NULL || p[1] != '\0')
2238 		(void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2239 	p = zc->zc_name + strlen(zc->zc_name);
2240 
2241 	do {
2242 		error = dmu_dir_list_next(os,
2243 		    sizeof (zc->zc_name) - (p - zc->zc_name), p,
2244 		    NULL, &zc->zc_cookie);
2245 		if (error == ENOENT)
2246 			error = SET_ERROR(ESRCH);
2247 	} while (error == 0 && dataset_name_hidden(zc->zc_name));
2248 	dmu_objset_rele(os, FTAG);
2249 
2250 	/*
2251 	 * If it's an internal dataset (ie. with a '$' in its name),
2252 	 * don't try to get stats for it, otherwise we'll return ENOENT.
2253 	 */
2254 	if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2255 		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2256 		if (error == ENOENT) {
2257 			/* We lost a race with destroy, get the next one. */
2258 			zc->zc_name[orig_len] = '\0';
2259 			goto top;
2260 		}
2261 	}
2262 	return (error);
2263 }
2264 
2265 /*
2266  * inputs:
2267  * zc_name		name of filesystem
2268  * zc_cookie		zap cursor
2269  * zc_nvlist_dst_size	size of buffer for property nvlist
2270  *
2271  * outputs:
2272  * zc_name		name of next snapshot
2273  * zc_objset_stats	stats
2274  * zc_nvlist_dst	property nvlist
2275  * zc_nvlist_dst_size	size of property nvlist
2276  */
2277 static int
2278 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2279 {
2280 	objset_t *os;
2281 	int error;
2282 
2283 	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2284 	if (error != 0) {
2285 		return (error == ENOENT ? ESRCH : error);
2286 	}
2287 
2288 	/*
2289 	 * A dataset name of maximum length cannot have any snapshots,
2290 	 * so exit immediately.
2291 	 */
2292 	if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
2293 		dmu_objset_rele(os, FTAG);
2294 		return (SET_ERROR(ESRCH));
2295 	}
2296 
2297 	error = dmu_snapshot_list_next(os,
2298 	    sizeof (zc->zc_name) - strlen(zc->zc_name),
2299 	    zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2300 	    NULL);
2301 
2302 	if (error == 0) {
2303 		dsl_dataset_t *ds;
2304 		dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2305 
2306 		error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2307 		if (error == 0) {
2308 			objset_t *ossnap;
2309 
2310 			error = dmu_objset_from_ds(ds, &ossnap);
2311 			if (error == 0)
2312 				error = zfs_ioc_objset_stats_impl(zc, ossnap);
2313 			dsl_dataset_rele(ds, FTAG);
2314 		}
2315 	} else if (error == ENOENT) {
2316 		error = SET_ERROR(ESRCH);
2317 	}
2318 
2319 	dmu_objset_rele(os, FTAG);
2320 	/* if we failed, undo the @ that we tacked on to zc_name */
2321 	if (error != 0)
2322 		*strchr(zc->zc_name, '@') = '\0';
2323 	return (error);
2324 }
2325 
2326 static int
2327 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2328 {
2329 	const char *propname = nvpair_name(pair);
2330 	uint64_t *valary;
2331 	unsigned int vallen;
2332 	const char *domain;
2333 	char *dash;
2334 	zfs_userquota_prop_t type;
2335 	uint64_t rid;
2336 	uint64_t quota;
2337 	zfsvfs_t *zfsvfs;
2338 	int err;
2339 
2340 	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2341 		nvlist_t *attrs;
2342 		VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2343 		if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2344 		    &pair) != 0)
2345 			return (SET_ERROR(EINVAL));
2346 	}
2347 
2348 	/*
2349 	 * A correctly constructed propname is encoded as
2350 	 * userquota@<rid>-<domain>.
2351 	 */
2352 	if ((dash = strchr(propname, '-')) == NULL ||
2353 	    nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2354 	    vallen != 3)
2355 		return (SET_ERROR(EINVAL));
2356 
2357 	domain = dash + 1;
2358 	type = valary[0];
2359 	rid = valary[1];
2360 	quota = valary[2];
2361 
2362 	err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2363 	if (err == 0) {
2364 		err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2365 		zfsvfs_rele(zfsvfs, FTAG);
2366 	}
2367 
2368 	return (err);
2369 }
2370 
2371 /*
2372  * If the named property is one that has a special function to set its value,
2373  * return 0 on success and a positive error code on failure; otherwise if it is
2374  * not one of the special properties handled by this function, return -1.
2375  *
2376  * XXX: It would be better for callers of the property interface if we handled
2377  * these special cases in dsl_prop.c (in the dsl layer).
2378  */
2379 static int
2380 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2381     nvpair_t *pair)
2382 {
2383 	const char *propname = nvpair_name(pair);
2384 	zfs_prop_t prop = zfs_name_to_prop(propname);
2385 	uint64_t intval;
2386 	int err = -1;
2387 
2388 	if (prop == ZPROP_INVAL) {
2389 		if (zfs_prop_userquota(propname))
2390 			return (zfs_prop_set_userquota(dsname, pair));
2391 		return (-1);
2392 	}
2393 
2394 	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2395 		nvlist_t *attrs;
2396 		VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2397 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2398 		    &pair) == 0);
2399 	}
2400 
2401 	if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2402 		return (-1);
2403 
2404 	VERIFY(0 == nvpair_value_uint64(pair, &intval));
2405 
2406 	switch (prop) {
2407 	case ZFS_PROP_QUOTA:
2408 		err = dsl_dir_set_quota(dsname, source, intval);
2409 		break;
2410 	case ZFS_PROP_REFQUOTA:
2411 		err = dsl_dataset_set_refquota(dsname, source, intval);
2412 		break;
2413 	case ZFS_PROP_FILESYSTEM_LIMIT:
2414 	case ZFS_PROP_SNAPSHOT_LIMIT:
2415 		if (intval == UINT64_MAX) {
2416 			/* clearing the limit, just do it */
2417 			err = 0;
2418 		} else {
2419 			err = dsl_dir_activate_fs_ss_limit(dsname);
2420 		}
2421 		/*
2422 		 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2423 		 * default path to set the value in the nvlist.
2424 		 */
2425 		if (err == 0)
2426 			err = -1;
2427 		break;
2428 	case ZFS_PROP_RESERVATION:
2429 		err = dsl_dir_set_reservation(dsname, source, intval);
2430 		break;
2431 	case ZFS_PROP_REFRESERVATION:
2432 		err = dsl_dataset_set_refreservation(dsname, source, intval);
2433 		break;
2434 	case ZFS_PROP_VOLSIZE:
2435 		err = zvol_set_volsize(dsname, intval);
2436 		break;
2437 	case ZFS_PROP_VERSION:
2438 	{
2439 		zfsvfs_t *zfsvfs;
2440 
2441 		if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2442 			break;
2443 
2444 		err = zfs_set_version(zfsvfs, intval);
2445 		zfsvfs_rele(zfsvfs, FTAG);
2446 
2447 		if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2448 			zfs_cmd_t *zc;
2449 
2450 			zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2451 			(void) strcpy(zc->zc_name, dsname);
2452 			(void) zfs_ioc_userspace_upgrade(zc);
2453 			kmem_free(zc, sizeof (zfs_cmd_t));
2454 		}
2455 		break;
2456 	}
2457 	default:
2458 		err = -1;
2459 	}
2460 
2461 	return (err);
2462 }
2463 
2464 /*
2465  * This function is best effort. If it fails to set any of the given properties,
2466  * it continues to set as many as it can and returns the last error
2467  * encountered. If the caller provides a non-NULL errlist, it will be filled in
2468  * with the list of names of all the properties that failed along with the
2469  * corresponding error numbers.
2470  *
2471  * If every property is set successfully, zero is returned and errlist is not
2472  * modified.
2473  */
2474 int
2475 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2476     nvlist_t *errlist)
2477 {
2478 	nvpair_t *pair;
2479 	nvpair_t *propval;
2480 	int rv = 0;
2481 	uint64_t intval;
2482 	char *strval;
2483 	nvlist_t *genericnvl = fnvlist_alloc();
2484 	nvlist_t *retrynvl = fnvlist_alloc();
2485 
2486 retry:
2487 	pair = NULL;
2488 	while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2489 		const char *propname = nvpair_name(pair);
2490 		zfs_prop_t prop = zfs_name_to_prop(propname);
2491 		int err = 0;
2492 
2493 		/* decode the property value */
2494 		propval = pair;
2495 		if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2496 			nvlist_t *attrs;
2497 			attrs = fnvpair_value_nvlist(pair);
2498 			if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2499 			    &propval) != 0)
2500 				err = SET_ERROR(EINVAL);
2501 		}
2502 
2503 		/* Validate value type */
2504 		if (err == 0 && prop == ZPROP_INVAL) {
2505 			if (zfs_prop_user(propname)) {
2506 				if (nvpair_type(propval) != DATA_TYPE_STRING)
2507 					err = SET_ERROR(EINVAL);
2508 			} else if (zfs_prop_userquota(propname)) {
2509 				if (nvpair_type(propval) !=
2510 				    DATA_TYPE_UINT64_ARRAY)
2511 					err = SET_ERROR(EINVAL);
2512 			} else {
2513 				err = SET_ERROR(EINVAL);
2514 			}
2515 		} else if (err == 0) {
2516 			if (nvpair_type(propval) == DATA_TYPE_STRING) {
2517 				if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2518 					err = SET_ERROR(EINVAL);
2519 			} else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2520 				const char *unused;
2521 
2522 				intval = fnvpair_value_uint64(propval);
2523 
2524 				switch (zfs_prop_get_type(prop)) {
2525 				case PROP_TYPE_NUMBER:
2526 					break;
2527 				case PROP_TYPE_STRING:
2528 					err = SET_ERROR(EINVAL);
2529 					break;
2530 				case PROP_TYPE_INDEX:
2531 					if (zfs_prop_index_to_string(prop,
2532 					    intval, &unused) != 0)
2533 						err = SET_ERROR(EINVAL);
2534 					break;
2535 				default:
2536 					cmn_err(CE_PANIC,
2537 					    "unknown property type");
2538 				}
2539 			} else {
2540 				err = SET_ERROR(EINVAL);
2541 			}
2542 		}
2543 
2544 		/* Validate permissions */
2545 		if (err == 0)
2546 			err = zfs_check_settable(dsname, pair, CRED());
2547 
2548 		if (err == 0) {
2549 			err = zfs_prop_set_special(dsname, source, pair);
2550 			if (err == -1) {
2551 				/*
2552 				 * For better performance we build up a list of
2553 				 * properties to set in a single transaction.
2554 				 */
2555 				err = nvlist_add_nvpair(genericnvl, pair);
2556 			} else if (err != 0 && nvl != retrynvl) {
2557 				/*
2558 				 * This may be a spurious error caused by
2559 				 * receiving quota and reservation out of order.
2560 				 * Try again in a second pass.
2561 				 */
2562 				err = nvlist_add_nvpair(retrynvl, pair);
2563 			}
2564 		}
2565 
2566 		if (err != 0) {
2567 			if (errlist != NULL)
2568 				fnvlist_add_int32(errlist, propname, err);
2569 			rv = err;
2570 		}
2571 	}
2572 
2573 	if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2574 		nvl = retrynvl;
2575 		goto retry;
2576 	}
2577 
2578 	if (!nvlist_empty(genericnvl) &&
2579 	    dsl_props_set(dsname, source, genericnvl) != 0) {
2580 		/*
2581 		 * If this fails, we still want to set as many properties as we
2582 		 * can, so try setting them individually.
2583 		 */
2584 		pair = NULL;
2585 		while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2586 			const char *propname = nvpair_name(pair);
2587 			int err = 0;
2588 
2589 			propval = pair;
2590 			if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2591 				nvlist_t *attrs;
2592 				attrs = fnvpair_value_nvlist(pair);
2593 				propval = fnvlist_lookup_nvpair(attrs,
2594 				    ZPROP_VALUE);
2595 			}
2596 
2597 			if (nvpair_type(propval) == DATA_TYPE_STRING) {
2598 				strval = fnvpair_value_string(propval);
2599 				err = dsl_prop_set_string(dsname, propname,
2600 				    source, strval);
2601 			} else {
2602 				intval = fnvpair_value_uint64(propval);
2603 				err = dsl_prop_set_int(dsname, propname, source,
2604 				    intval);
2605 			}
2606 
2607 			if (err != 0) {
2608 				if (errlist != NULL) {
2609 					fnvlist_add_int32(errlist, propname,
2610 					    err);
2611 				}
2612 				rv = err;
2613 			}
2614 		}
2615 	}
2616 	nvlist_free(genericnvl);
2617 	nvlist_free(retrynvl);
2618 
2619 	return (rv);
2620 }
2621 
2622 /*
2623  * Check that all the properties are valid user properties.
2624  */
2625 static int
2626 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2627 {
2628 	nvpair_t *pair = NULL;
2629 	int error = 0;
2630 
2631 	while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2632 		const char *propname = nvpair_name(pair);
2633 
2634 		if (!zfs_prop_user(propname) ||
2635 		    nvpair_type(pair) != DATA_TYPE_STRING)
2636 			return (SET_ERROR(EINVAL));
2637 
2638 		if (error = zfs_secpolicy_write_perms(fsname,
2639 		    ZFS_DELEG_PERM_USERPROP, CRED()))
2640 			return (error);
2641 
2642 		if (strlen(propname) >= ZAP_MAXNAMELEN)
2643 			return (SET_ERROR(ENAMETOOLONG));
2644 
2645 		if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
2646 			return (E2BIG);
2647 	}
2648 	return (0);
2649 }
2650 
2651 static void
2652 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2653 {
2654 	nvpair_t *pair;
2655 
2656 	VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2657 
2658 	pair = NULL;
2659 	while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2660 		if (nvlist_exists(skipped, nvpair_name(pair)))
2661 			continue;
2662 
2663 		VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2664 	}
2665 }
2666 
2667 static int
2668 clear_received_props(const char *dsname, nvlist_t *props,
2669     nvlist_t *skipped)
2670 {
2671 	int err = 0;
2672 	nvlist_t *cleared_props = NULL;
2673 	props_skip(props, skipped, &cleared_props);
2674 	if (!nvlist_empty(cleared_props)) {
2675 		/*
2676 		 * Acts on local properties until the dataset has received
2677 		 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2678 		 */
2679 		zprop_source_t flags = (ZPROP_SRC_NONE |
2680 		    (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2681 		err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2682 	}
2683 	nvlist_free(cleared_props);
2684 	return (err);
2685 }
2686 
2687 /*
2688  * inputs:
2689  * zc_name		name of filesystem
2690  * zc_value		name of property to set
2691  * zc_nvlist_src{_size}	nvlist of properties to apply
2692  * zc_cookie		received properties flag
2693  *
2694  * outputs:
2695  * zc_nvlist_dst{_size} error for each unapplied received property
2696  */
2697 static int
2698 zfs_ioc_set_prop(zfs_cmd_t *zc)
2699 {
2700 	nvlist_t *nvl;
2701 	boolean_t received = zc->zc_cookie;
2702 	zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2703 	    ZPROP_SRC_LOCAL);
2704 	nvlist_t *errors;
2705 	int error;
2706 
2707 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2708 	    zc->zc_iflags, &nvl)) != 0)
2709 		return (error);
2710 
2711 	if (received) {
2712 		nvlist_t *origprops;
2713 
2714 		if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2715 			(void) clear_received_props(zc->zc_name,
2716 			    origprops, nvl);
2717 			nvlist_free(origprops);
2718 		}
2719 
2720 		error = dsl_prop_set_hasrecvd(zc->zc_name);
2721 	}
2722 
2723 	errors = fnvlist_alloc();
2724 	if (error == 0)
2725 		error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2726 
2727 	if (zc->zc_nvlist_dst != NULL && errors != NULL) {
2728 		(void) put_nvlist(zc, errors);
2729 	}
2730 
2731 	nvlist_free(errors);
2732 	nvlist_free(nvl);
2733 	return (error);
2734 }
2735 
2736 /*
2737  * inputs:
2738  * zc_name		name of filesystem
2739  * zc_value		name of property to inherit
2740  * zc_cookie		revert to received value if TRUE
2741  *
2742  * outputs:		none
2743  */
2744 static int
2745 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2746 {
2747 	const char *propname = zc->zc_value;
2748 	zfs_prop_t prop = zfs_name_to_prop(propname);
2749 	boolean_t received = zc->zc_cookie;
2750 	zprop_source_t source = (received
2751 	    ? ZPROP_SRC_NONE		/* revert to received value, if any */
2752 	    : ZPROP_SRC_INHERITED);	/* explicitly inherit */
2753 
2754 	if (received) {
2755 		nvlist_t *dummy;
2756 		nvpair_t *pair;
2757 		zprop_type_t type;
2758 		int err;
2759 
2760 		/*
2761 		 * zfs_prop_set_special() expects properties in the form of an
2762 		 * nvpair with type info.
2763 		 */
2764 		if (prop == ZPROP_INVAL) {
2765 			if (!zfs_prop_user(propname))
2766 				return (SET_ERROR(EINVAL));
2767 
2768 			type = PROP_TYPE_STRING;
2769 		} else if (prop == ZFS_PROP_VOLSIZE ||
2770 		    prop == ZFS_PROP_VERSION) {
2771 			return (SET_ERROR(EINVAL));
2772 		} else {
2773 			type = zfs_prop_get_type(prop);
2774 		}
2775 
2776 		VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2777 
2778 		switch (type) {
2779 		case PROP_TYPE_STRING:
2780 			VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2781 			break;
2782 		case PROP_TYPE_NUMBER:
2783 		case PROP_TYPE_INDEX:
2784 			VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2785 			break;
2786 		default:
2787 			nvlist_free(dummy);
2788 			return (SET_ERROR(EINVAL));
2789 		}
2790 
2791 		pair = nvlist_next_nvpair(dummy, NULL);
2792 		err = zfs_prop_set_special(zc->zc_name, source, pair);
2793 		nvlist_free(dummy);
2794 		if (err != -1)
2795 			return (err); /* special property already handled */
2796 	} else {
2797 		/*
2798 		 * Only check this in the non-received case. We want to allow
2799 		 * 'inherit -S' to revert non-inheritable properties like quota
2800 		 * and reservation to the received or default values even though
2801 		 * they are not considered inheritable.
2802 		 */
2803 		if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2804 			return (SET_ERROR(EINVAL));
2805 	}
2806 
2807 	/* property name has been validated by zfs_secpolicy_inherit_prop() */
2808 	return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2809 }
2810 
2811 static int
2812 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2813 {
2814 	nvlist_t *props;
2815 	spa_t *spa;
2816 	int error;
2817 	nvpair_t *pair;
2818 
2819 	if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2820 	    zc->zc_iflags, &props))
2821 		return (error);
2822 
2823 	/*
2824 	 * If the only property is the configfile, then just do a spa_lookup()
2825 	 * to handle the faulted case.
2826 	 */
2827 	pair = nvlist_next_nvpair(props, NULL);
2828 	if (pair != NULL && strcmp(nvpair_name(pair),
2829 	    zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2830 	    nvlist_next_nvpair(props, pair) == NULL) {
2831 		mutex_enter(&spa_namespace_lock);
2832 		if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2833 			spa_configfile_set(spa, props, B_FALSE);
2834 			spa_config_sync(spa, B_FALSE, B_TRUE);
2835 		}
2836 		mutex_exit(&spa_namespace_lock);
2837 		if (spa != NULL) {
2838 			nvlist_free(props);
2839 			return (0);
2840 		}
2841 	}
2842 
2843 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2844 		nvlist_free(props);
2845 		return (error);
2846 	}
2847 
2848 	error = spa_prop_set(spa, props);
2849 
2850 	nvlist_free(props);
2851 	spa_close(spa, FTAG);
2852 
2853 	return (error);
2854 }
2855 
2856 static int
2857 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2858 {
2859 	spa_t *spa;
2860 	int error;
2861 	nvlist_t *nvp = NULL;
2862 
2863 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2864 		/*
2865 		 * If the pool is faulted, there may be properties we can still
2866 		 * get (such as altroot and cachefile), so attempt to get them
2867 		 * anyway.
2868 		 */
2869 		mutex_enter(&spa_namespace_lock);
2870 		if ((spa = spa_lookup(zc->zc_name)) != NULL)
2871 			error = spa_prop_get(spa, &nvp);
2872 		mutex_exit(&spa_namespace_lock);
2873 	} else {
2874 		error = spa_prop_get(spa, &nvp);
2875 		spa_close(spa, FTAG);
2876 	}
2877 
2878 	if (error == 0 && zc->zc_nvlist_dst != NULL)
2879 		error = put_nvlist(zc, nvp);
2880 	else
2881 		error = SET_ERROR(EFAULT);
2882 
2883 	nvlist_free(nvp);
2884 	return (error);
2885 }
2886 
2887 /*
2888  * inputs:
2889  * zc_name		name of filesystem
2890  * zc_nvlist_src{_size}	nvlist of delegated permissions
2891  * zc_perm_action	allow/unallow flag
2892  *
2893  * outputs:		none
2894  */
2895 static int
2896 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2897 {
2898 	int error;
2899 	nvlist_t *fsaclnv = NULL;
2900 
2901 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2902 	    zc->zc_iflags, &fsaclnv)) != 0)
2903 		return (error);
2904 
2905 	/*
2906 	 * Verify nvlist is constructed correctly
2907 	 */
2908 	if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2909 		nvlist_free(fsaclnv);
2910 		return (SET_ERROR(EINVAL));
2911 	}
2912 
2913 	/*
2914 	 * If we don't have PRIV_SYS_MOUNT, then validate
2915 	 * that user is allowed to hand out each permission in
2916 	 * the nvlist(s)
2917 	 */
2918 
2919 	error = secpolicy_zfs(CRED());
2920 	if (error != 0) {
2921 		if (zc->zc_perm_action == B_FALSE) {
2922 			error = dsl_deleg_can_allow(zc->zc_name,
2923 			    fsaclnv, CRED());
2924 		} else {
2925 			error = dsl_deleg_can_unallow(zc->zc_name,
2926 			    fsaclnv, CRED());
2927 		}
2928 	}
2929 
2930 	if (error == 0)
2931 		error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2932 
2933 	nvlist_free(fsaclnv);
2934 	return (error);
2935 }
2936 
2937 /*
2938  * inputs:
2939  * zc_name		name of filesystem
2940  *
2941  * outputs:
2942  * zc_nvlist_src{_size}	nvlist of delegated permissions
2943  */
2944 static int
2945 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2946 {
2947 	nvlist_t *nvp;
2948 	int error;
2949 
2950 	if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2951 		error = put_nvlist(zc, nvp);
2952 		nvlist_free(nvp);
2953 	}
2954 
2955 	return (error);
2956 }
2957 
2958 /*
2959  * Search the vfs list for a specified resource.  Returns a pointer to it
2960  * or NULL if no suitable entry is found. The caller of this routine
2961  * is responsible for releasing the returned vfs pointer.
2962  */
2963 static vfs_t *
2964 zfs_get_vfs(const char *resource)
2965 {
2966 	struct vfs *vfsp;
2967 	struct vfs *vfs_found = NULL;
2968 
2969 	vfs_list_read_lock();
2970 	vfsp = rootvfs;
2971 	do {
2972 		if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
2973 			VFS_HOLD(vfsp);
2974 			vfs_found = vfsp;
2975 			break;
2976 		}
2977 		vfsp = vfsp->vfs_next;
2978 	} while (vfsp != rootvfs);
2979 	vfs_list_unlock();
2980 	return (vfs_found);
2981 }
2982 
2983 /* ARGSUSED */
2984 static void
2985 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
2986 {
2987 	zfs_creat_t *zct = arg;
2988 
2989 	zfs_create_fs(os, cr, zct->zct_zplprops, tx);
2990 }
2991 
2992 #define	ZFS_PROP_UNDEFINED	((uint64_t)-1)
2993 
2994 /*
2995  * inputs:
2996  * os			parent objset pointer (NULL if root fs)
2997  * fuids_ok		fuids allowed in this version of the spa?
2998  * sa_ok		SAs allowed in this version of the spa?
2999  * createprops		list of properties requested by creator
3000  *
3001  * outputs:
3002  * zplprops	values for the zplprops we attach to the master node object
3003  * is_ci	true if requested file system will be purely case-insensitive
3004  *
3005  * Determine the settings for utf8only, normalization and
3006  * casesensitivity.  Specific values may have been requested by the
3007  * creator and/or we can inherit values from the parent dataset.  If
3008  * the file system is of too early a vintage, a creator can not
3009  * request settings for these properties, even if the requested
3010  * setting is the default value.  We don't actually want to create dsl
3011  * properties for these, so remove them from the source nvlist after
3012  * processing.
3013  */
3014 static int
3015 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3016     boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3017     nvlist_t *zplprops, boolean_t *is_ci)
3018 {
3019 	uint64_t sense = ZFS_PROP_UNDEFINED;
3020 	uint64_t norm = ZFS_PROP_UNDEFINED;
3021 	uint64_t u8 = ZFS_PROP_UNDEFINED;
3022 
3023 	ASSERT(zplprops != NULL);
3024 
3025 	/*
3026 	 * Pull out creator prop choices, if any.
3027 	 */
3028 	if (createprops) {
3029 		(void) nvlist_lookup_uint64(createprops,
3030 		    zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3031 		(void) nvlist_lookup_uint64(createprops,
3032 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3033 		(void) nvlist_remove_all(createprops,
3034 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3035 		(void) nvlist_lookup_uint64(createprops,
3036 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3037 		(void) nvlist_remove_all(createprops,
3038 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3039 		(void) nvlist_lookup_uint64(createprops,
3040 		    zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3041 		(void) nvlist_remove_all(createprops,
3042 		    zfs_prop_to_name(ZFS_PROP_CASE));
3043 	}
3044 
3045 	/*
3046 	 * If the zpl version requested is whacky or the file system
3047 	 * or pool is version is too "young" to support normalization
3048 	 * and the creator tried to set a value for one of the props,
3049 	 * error out.
3050 	 */
3051 	if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3052 	    (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3053 	    (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3054 	    (zplver < ZPL_VERSION_NORMALIZATION &&
3055 	    (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3056 	    sense != ZFS_PROP_UNDEFINED)))
3057 		return (SET_ERROR(ENOTSUP));
3058 
3059 	/*
3060 	 * Put the version in the zplprops
3061 	 */
3062 	VERIFY(nvlist_add_uint64(zplprops,
3063 	    zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3064 
3065 	if (norm == ZFS_PROP_UNDEFINED)
3066 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3067 	VERIFY(nvlist_add_uint64(zplprops,
3068 	    zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3069 
3070 	/*
3071 	 * If we're normalizing, names must always be valid UTF-8 strings.
3072 	 */
3073 	if (norm)
3074 		u8 = 1;
3075 	if (u8 == ZFS_PROP_UNDEFINED)
3076 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3077 	VERIFY(nvlist_add_uint64(zplprops,
3078 	    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3079 
3080 	if (sense == ZFS_PROP_UNDEFINED)
3081 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3082 	VERIFY(nvlist_add_uint64(zplprops,
3083 	    zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3084 
3085 	if (is_ci)
3086 		*is_ci = (sense == ZFS_CASE_INSENSITIVE);
3087 
3088 	return (0);
3089 }
3090 
3091 static int
3092 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3093     nvlist_t *zplprops, boolean_t *is_ci)
3094 {
3095 	boolean_t fuids_ok, sa_ok;
3096 	uint64_t zplver = ZPL_VERSION;
3097 	objset_t *os = NULL;
3098 	char parentname[MAXNAMELEN];
3099 	char *cp;
3100 	spa_t *spa;
3101 	uint64_t spa_vers;
3102 	int error;
3103 
3104 	(void) strlcpy(parentname, dataset, sizeof (parentname));
3105 	cp = strrchr(parentname, '/');
3106 	ASSERT(cp != NULL);
3107 	cp[0] = '\0';
3108 
3109 	if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3110 		return (error);
3111 
3112 	spa_vers = spa_version(spa);
3113 	spa_close(spa, FTAG);
3114 
3115 	zplver = zfs_zpl_version_map(spa_vers);
3116 	fuids_ok = (zplver >= ZPL_VERSION_FUID);
3117 	sa_ok = (zplver >= ZPL_VERSION_SA);
3118 
3119 	/*
3120 	 * Open parent object set so we can inherit zplprop values.
3121 	 */
3122 	if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3123 		return (error);
3124 
3125 	error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3126 	    zplprops, is_ci);
3127 	dmu_objset_rele(os, FTAG);
3128 	return (error);
3129 }
3130 
3131 static int
3132 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3133     nvlist_t *zplprops, boolean_t *is_ci)
3134 {
3135 	boolean_t fuids_ok;
3136 	boolean_t sa_ok;
3137 	uint64_t zplver = ZPL_VERSION;
3138 	int error;
3139 
3140 	zplver = zfs_zpl_version_map(spa_vers);
3141 	fuids_ok = (zplver >= ZPL_VERSION_FUID);
3142 	sa_ok = (zplver >= ZPL_VERSION_SA);
3143 
3144 	error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3145 	    createprops, zplprops, is_ci);
3146 	return (error);
3147 }
3148 
3149 /*
3150  * innvl: {
3151  *     "type" -> dmu_objset_type_t (int32)
3152  *     (optional) "props" -> { prop -> value }
3153  * }
3154  *
3155  * outnvl: propname -> error code (int32)
3156  */
3157 static int
3158 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3159 {
3160 	int error = 0;
3161 	zfs_creat_t zct = { 0 };
3162 	nvlist_t *nvprops = NULL;
3163 	void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3164 	int32_t type32;
3165 	dmu_objset_type_t type;
3166 	boolean_t is_insensitive = B_FALSE;
3167 
3168 	if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3169 		return (SET_ERROR(EINVAL));
3170 	type = type32;
3171 	(void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3172 
3173 	switch (type) {
3174 	case DMU_OST_ZFS:
3175 		cbfunc = zfs_create_cb;
3176 		break;
3177 
3178 	case DMU_OST_ZVOL:
3179 		cbfunc = zvol_create_cb;
3180 		break;
3181 
3182 	default:
3183 		cbfunc = NULL;
3184 		break;
3185 	}
3186 	if (strchr(fsname, '@') ||
3187 	    strchr(fsname, '%'))
3188 		return (SET_ERROR(EINVAL));
3189 
3190 	zct.zct_props = nvprops;
3191 
3192 	if (cbfunc == NULL)
3193 		return (SET_ERROR(EINVAL));
3194 
3195 	if (type == DMU_OST_ZVOL) {
3196 		uint64_t volsize, volblocksize;
3197 
3198 		if (nvprops == NULL)
3199 			return (SET_ERROR(EINVAL));
3200 		if (nvlist_lookup_uint64(nvprops,
3201 		    zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3202 			return (SET_ERROR(EINVAL));
3203 
3204 		if ((error = nvlist_lookup_uint64(nvprops,
3205 		    zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3206 		    &volblocksize)) != 0 && error != ENOENT)
3207 			return (SET_ERROR(EINVAL));
3208 
3209 		if (error != 0)
3210 			volblocksize = zfs_prop_default_numeric(
3211 			    ZFS_PROP_VOLBLOCKSIZE);
3212 
3213 		if ((error = zvol_check_volblocksize(
3214 		    volblocksize)) != 0 ||
3215 		    (error = zvol_check_volsize(volsize,
3216 		    volblocksize)) != 0)
3217 			return (error);
3218 	} else if (type == DMU_OST_ZFS) {
3219 		int error;
3220 
3221 		/*
3222 		 * We have to have normalization and
3223 		 * case-folding flags correct when we do the
3224 		 * file system creation, so go figure them out
3225 		 * now.
3226 		 */
3227 		VERIFY(nvlist_alloc(&zct.zct_zplprops,
3228 		    NV_UNIQUE_NAME, KM_SLEEP) == 0);
3229 		error = zfs_fill_zplprops(fsname, nvprops,
3230 		    zct.zct_zplprops, &is_insensitive);
3231 		if (error != 0) {
3232 			nvlist_free(zct.zct_zplprops);
3233 			return (error);
3234 		}
3235 	}
3236 
3237 	error = dmu_objset_create(fsname, type,
3238 	    is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3239 	nvlist_free(zct.zct_zplprops);
3240 
3241 	/*
3242 	 * It would be nice to do this atomically.
3243 	 */
3244 	if (error == 0) {
3245 		error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3246 		    nvprops, outnvl);
3247 		if (error != 0)
3248 			(void) dsl_destroy_head(fsname);
3249 	}
3250 	return (error);
3251 }
3252 
3253 /*
3254  * innvl: {
3255  *     "origin" -> name of origin snapshot
3256  *     (optional) "props" -> { prop -> value }
3257  * }
3258  *
3259  * outnvl: propname -> error code (int32)
3260  */
3261 static int
3262 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3263 {
3264 	int error = 0;
3265 	nvlist_t *nvprops = NULL;
3266 	char *origin_name;
3267 
3268 	if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3269 		return (SET_ERROR(EINVAL));
3270 	(void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3271 
3272 	if (strchr(fsname, '@') ||
3273 	    strchr(fsname, '%'))
3274 		return (SET_ERROR(EINVAL));
3275 
3276 	if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3277 		return (SET_ERROR(EINVAL));
3278 	error = dmu_objset_clone(fsname, origin_name);
3279 	if (error != 0)
3280 		return (error);
3281 
3282 	/*
3283 	 * It would be nice to do this atomically.
3284 	 */
3285 	if (error == 0) {
3286 		error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3287 		    nvprops, outnvl);
3288 		if (error != 0)
3289 			(void) dsl_destroy_head(fsname);
3290 	}
3291 	return (error);
3292 }
3293 
3294 /*
3295  * innvl: {
3296  *     "snaps" -> { snapshot1, snapshot2 }
3297  *     (optional) "props" -> { prop -> value (string) }
3298  * }
3299  *
3300  * outnvl: snapshot -> error code (int32)
3301  */
3302 static int
3303 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3304 {
3305 	nvlist_t *snaps;
3306 	nvlist_t *props = NULL;
3307 	int error, poollen;
3308 	nvpair_t *pair;
3309 
3310 	(void) nvlist_lookup_nvlist(innvl, "props", &props);
3311 	if ((error = zfs_check_userprops(poolname, props)) != 0)
3312 		return (error);
3313 
3314 	if (!nvlist_empty(props) &&
3315 	    zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3316 		return (SET_ERROR(ENOTSUP));
3317 
3318 	if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3319 		return (SET_ERROR(EINVAL));
3320 	poollen = strlen(poolname);
3321 	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3322 	    pair = nvlist_next_nvpair(snaps, pair)) {
3323 		const char *name = nvpair_name(pair);
3324 		const char *cp = strchr(name, '@');
3325 
3326 		/*
3327 		 * The snap name must contain an @, and the part after it must
3328 		 * contain only valid characters.
3329 		 */
3330 		if (cp == NULL ||
3331 		    zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3332 			return (SET_ERROR(EINVAL));
3333 
3334 		/*
3335 		 * The snap must be in the specified pool.
3336 		 */
3337 		if (strncmp(name, poolname, poollen) != 0 ||
3338 		    (name[poollen] != '/' && name[poollen] != '@'))
3339 			return (SET_ERROR(EXDEV));
3340 
3341 		/* This must be the only snap of this fs. */
3342 		for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3343 		    pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3344 			if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3345 			    == 0) {
3346 				return (SET_ERROR(EXDEV));
3347 			}
3348 		}
3349 	}
3350 
3351 	error = dsl_dataset_snapshot(snaps, props, outnvl);
3352 	return (error);
3353 }
3354 
3355 /*
3356  * innvl: "message" -> string
3357  */
3358 /* ARGSUSED */
3359 static int
3360 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3361 {
3362 	char *message;
3363 	spa_t *spa;
3364 	int error;
3365 	char *poolname;
3366 
3367 	/*
3368 	 * The poolname in the ioctl is not set, we get it from the TSD,
3369 	 * which was set at the end of the last successful ioctl that allows
3370 	 * logging.  The secpolicy func already checked that it is set.
3371 	 * Only one log ioctl is allowed after each successful ioctl, so
3372 	 * we clear the TSD here.
3373 	 */
3374 	poolname = tsd_get(zfs_allow_log_key);
3375 	(void) tsd_set(zfs_allow_log_key, NULL);
3376 	error = spa_open(poolname, &spa, FTAG);
3377 	strfree(poolname);
3378 	if (error != 0)
3379 		return (error);
3380 
3381 	if (nvlist_lookup_string(innvl, "message", &message) != 0)  {
3382 		spa_close(spa, FTAG);
3383 		return (SET_ERROR(EINVAL));
3384 	}
3385 
3386 	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3387 		spa_close(spa, FTAG);
3388 		return (SET_ERROR(ENOTSUP));
3389 	}
3390 
3391 	error = spa_history_log(spa, message);
3392 	spa_close(spa, FTAG);
3393 	return (error);
3394 }
3395 
3396 /*
3397  * The dp_config_rwlock must not be held when calling this, because the
3398  * unmount may need to write out data.
3399  *
3400  * This function is best-effort.  Callers must deal gracefully if it
3401  * remains mounted (or is remounted after this call).
3402  *
3403  * Returns 0 if the argument is not a snapshot, or it is not currently a
3404  * filesystem, or we were able to unmount it.  Returns error code otherwise.
3405  */
3406 int
3407 zfs_unmount_snap(const char *snapname)
3408 {
3409 	vfs_t *vfsp;
3410 	zfsvfs_t *zfsvfs;
3411 	int err;
3412 
3413 	if (strchr(snapname, '@') == NULL)
3414 		return (0);
3415 
3416 	vfsp = zfs_get_vfs(snapname);
3417 	if (vfsp == NULL)
3418 		return (0);
3419 
3420 	zfsvfs = vfsp->vfs_data;
3421 	ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os)));
3422 
3423 	err = vn_vfswlock(vfsp->vfs_vnodecovered);
3424 	VFS_RELE(vfsp);
3425 	if (err != 0)
3426 		return (SET_ERROR(err));
3427 
3428 	/*
3429 	 * Always force the unmount for snapshots.
3430 	 */
3431 	(void) dounmount(vfsp, MS_FORCE, kcred);
3432 	return (0);
3433 }
3434 
3435 /* ARGSUSED */
3436 static int
3437 zfs_unmount_snap_cb(const char *snapname, void *arg)
3438 {
3439 	return (zfs_unmount_snap(snapname));
3440 }
3441 
3442 /*
3443  * When a clone is destroyed, its origin may also need to be destroyed,
3444  * in which case it must be unmounted.  This routine will do that unmount
3445  * if necessary.
3446  */
3447 void
3448 zfs_destroy_unmount_origin(const char *fsname)
3449 {
3450 	int error;
3451 	objset_t *os;
3452 	dsl_dataset_t *ds;
3453 
3454 	error = dmu_objset_hold(fsname, FTAG, &os);
3455 	if (error != 0)
3456 		return;
3457 	ds = dmu_objset_ds(os);
3458 	if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3459 		char originname[MAXNAMELEN];
3460 		dsl_dataset_name(ds->ds_prev, originname);
3461 		dmu_objset_rele(os, FTAG);
3462 		(void) zfs_unmount_snap(originname);
3463 	} else {
3464 		dmu_objset_rele(os, FTAG);
3465 	}
3466 }
3467 
3468 /*
3469  * innvl: {
3470  *     "snaps" -> { snapshot1, snapshot2 }
3471  *     (optional boolean) "defer"
3472  * }
3473  *
3474  * outnvl: snapshot -> error code (int32)
3475  *
3476  */
3477 /* ARGSUSED */
3478 static int
3479 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3480 {
3481 	nvlist_t *snaps;
3482 	nvpair_t *pair;
3483 	boolean_t defer;
3484 
3485 	if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3486 		return (SET_ERROR(EINVAL));
3487 	defer = nvlist_exists(innvl, "defer");
3488 
3489 	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3490 	    pair = nvlist_next_nvpair(snaps, pair)) {
3491 		(void) zfs_unmount_snap(nvpair_name(pair));
3492 	}
3493 
3494 	return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3495 }
3496 
3497 /*
3498  * Create bookmarks.  Bookmark names are of the form <fs>#<bmark>.
3499  * All bookmarks must be in the same pool.
3500  *
3501  * innvl: {
3502  *     bookmark1 -> snapshot1, bookmark2 -> snapshot2
3503  * }
3504  *
3505  * outnvl: bookmark -> error code (int32)
3506  *
3507  */
3508 /* ARGSUSED */
3509 static int
3510 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3511 {
3512 	for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3513 	    pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3514 		char *snap_name;
3515 
3516 		/*
3517 		 * Verify the snapshot argument.
3518 		 */
3519 		if (nvpair_value_string(pair, &snap_name) != 0)
3520 			return (SET_ERROR(EINVAL));
3521 
3522 
3523 		/* Verify that the keys (bookmarks) are unique */
3524 		for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair);
3525 		    pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) {
3526 			if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0)
3527 				return (SET_ERROR(EINVAL));
3528 		}
3529 	}
3530 
3531 	return (dsl_bookmark_create(innvl, outnvl));
3532 }
3533 
3534 /*
3535  * innvl: {
3536  *     property 1, property 2, ...
3537  * }
3538  *
3539  * outnvl: {
3540  *     bookmark name 1 -> { property 1, property 2, ... },
3541  *     bookmark name 2 -> { property 1, property 2, ... }
3542  * }
3543  *
3544  */
3545 static int
3546 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3547 {
3548 	return (dsl_get_bookmarks(fsname, innvl, outnvl));
3549 }
3550 
3551 /*
3552  * innvl: {
3553  *     bookmark name 1, bookmark name 2
3554  * }
3555  *
3556  * outnvl: bookmark -> error code (int32)
3557  *
3558  */
3559 static int
3560 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
3561     nvlist_t *outnvl)
3562 {
3563 	int error, poollen;
3564 
3565 	poollen = strlen(poolname);
3566 	for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3567 	    pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3568 		const char *name = nvpair_name(pair);
3569 		const char *cp = strchr(name, '#');
3570 
3571 		/*
3572 		 * The bookmark name must contain an #, and the part after it
3573 		 * must contain only valid characters.
3574 		 */
3575 		if (cp == NULL ||
3576 		    zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3577 			return (SET_ERROR(EINVAL));
3578 
3579 		/*
3580 		 * The bookmark must be in the specified pool.
3581 		 */
3582 		if (strncmp(name, poolname, poollen) != 0 ||
3583 		    (name[poollen] != '/' && name[poollen] != '#'))
3584 			return (SET_ERROR(EXDEV));
3585 	}
3586 
3587 	error = dsl_bookmark_destroy(innvl, outnvl);
3588 	return (error);
3589 }
3590 
3591 /*
3592  * inputs:
3593  * zc_name		name of dataset to destroy
3594  * zc_objset_type	type of objset
3595  * zc_defer_destroy	mark for deferred destroy
3596  *
3597  * outputs:		none
3598  */
3599 static int
3600 zfs_ioc_destroy(zfs_cmd_t *zc)
3601 {
3602 	int err;
3603 
3604 	if (zc->zc_objset_type == DMU_OST_ZFS) {
3605 		err = zfs_unmount_snap(zc->zc_name);
3606 		if (err != 0)
3607 			return (err);
3608 	}
3609 
3610 	if (strchr(zc->zc_name, '@'))
3611 		err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3612 	else
3613 		err = dsl_destroy_head(zc->zc_name);
3614 	if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3615 		(void) zvol_remove_minor(zc->zc_name);
3616 	return (err);
3617 }
3618 
3619 /*
3620  * fsname is name of dataset to rollback (to most recent snapshot)
3621  *
3622  * innvl is not used.
3623  *
3624  * outnvl: "target" -> name of most recent snapshot
3625  * }
3626  */
3627 /* ARGSUSED */
3628 static int
3629 zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl)
3630 {
3631 	zfsvfs_t *zfsvfs;
3632 	int error;
3633 
3634 	if (getzfsvfs(fsname, &zfsvfs) == 0) {
3635 		error = zfs_suspend_fs(zfsvfs);
3636 		if (error == 0) {
3637 			int resume_err;
3638 
3639 			error = dsl_dataset_rollback(fsname, zfsvfs, outnvl);
3640 			resume_err = zfs_resume_fs(zfsvfs, fsname);
3641 			error = error ? error : resume_err;
3642 		}
3643 		VFS_RELE(zfsvfs->z_vfs);
3644 	} else {
3645 		error = dsl_dataset_rollback(fsname, NULL, outnvl);
3646 	}
3647 	return (error);
3648 }
3649 
3650 static int
3651 recursive_unmount(const char *fsname, void *arg)
3652 {
3653 	const char *snapname = arg;
3654 	char fullname[MAXNAMELEN];
3655 
3656 	(void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname);
3657 	return (zfs_unmount_snap(fullname));
3658 }
3659 
3660 /*
3661  * inputs:
3662  * zc_name	old name of dataset
3663  * zc_value	new name of dataset
3664  * zc_cookie	recursive flag (only valid for snapshots)
3665  *
3666  * outputs:	none
3667  */
3668 static int
3669 zfs_ioc_rename(zfs_cmd_t *zc)
3670 {
3671 	boolean_t recursive = zc->zc_cookie & 1;
3672 	char *at;
3673 
3674 	zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3675 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3676 	    strchr(zc->zc_value, '%'))
3677 		return (SET_ERROR(EINVAL));
3678 
3679 	at = strchr(zc->zc_name, '@');
3680 	if (at != NULL) {
3681 		/* snaps must be in same fs */
3682 		int error;
3683 
3684 		if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3685 			return (SET_ERROR(EXDEV));
3686 		*at = '\0';
3687 		if (zc->zc_objset_type == DMU_OST_ZFS) {
3688 			error = dmu_objset_find(zc->zc_name,
3689 			    recursive_unmount, at + 1,
3690 			    recursive ? DS_FIND_CHILDREN : 0);
3691 			if (error != 0) {
3692 				*at = '@';
3693 				return (error);
3694 			}
3695 		}
3696 		error = dsl_dataset_rename_snapshot(zc->zc_name,
3697 		    at + 1, strchr(zc->zc_value, '@') + 1, recursive);
3698 		*at = '@';
3699 
3700 		return (error);
3701 	} else {
3702 		if (zc->zc_objset_type == DMU_OST_ZVOL)
3703 			(void) zvol_remove_minor(zc->zc_name);
3704 		return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3705 	}
3706 }
3707 
3708 static int
3709 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3710 {
3711 	const char *propname = nvpair_name(pair);
3712 	boolean_t issnap = (strchr(dsname, '@') != NULL);
3713 	zfs_prop_t prop = zfs_name_to_prop(propname);
3714 	uint64_t intval;
3715 	int err;
3716 
3717 	if (prop == ZPROP_INVAL) {
3718 		if (zfs_prop_user(propname)) {
3719 			if (err = zfs_secpolicy_write_perms(dsname,
3720 			    ZFS_DELEG_PERM_USERPROP, cr))
3721 				return (err);
3722 			return (0);
3723 		}
3724 
3725 		if (!issnap && zfs_prop_userquota(propname)) {
3726 			const char *perm = NULL;
3727 			const char *uq_prefix =
3728 			    zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3729 			const char *gq_prefix =
3730 			    zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3731 
3732 			if (strncmp(propname, uq_prefix,
3733 			    strlen(uq_prefix)) == 0) {
3734 				perm = ZFS_DELEG_PERM_USERQUOTA;
3735 			} else if (strncmp(propname, gq_prefix,
3736 			    strlen(gq_prefix)) == 0) {
3737 				perm = ZFS_DELEG_PERM_GROUPQUOTA;
3738 			} else {
3739 				/* USERUSED and GROUPUSED are read-only */
3740 				return (SET_ERROR(EINVAL));
3741 			}
3742 
3743 			if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3744 				return (err);
3745 			return (0);
3746 		}
3747 
3748 		return (SET_ERROR(EINVAL));
3749 	}
3750 
3751 	if (issnap)
3752 		return (SET_ERROR(EINVAL));
3753 
3754 	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3755 		/*
3756 		 * dsl_prop_get_all_impl() returns properties in this
3757 		 * format.
3758 		 */
3759 		nvlist_t *attrs;
3760 		VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3761 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3762 		    &pair) == 0);
3763 	}
3764 
3765 	/*
3766 	 * Check that this value is valid for this pool version
3767 	 */
3768 	switch (prop) {
3769 	case ZFS_PROP_COMPRESSION:
3770 		/*
3771 		 * If the user specified gzip compression, make sure
3772 		 * the SPA supports it. We ignore any errors here since
3773 		 * we'll catch them later.
3774 		 */
3775 		if (nvpair_value_uint64(pair, &intval) == 0) {
3776 			if (intval >= ZIO_COMPRESS_GZIP_1 &&
3777 			    intval <= ZIO_COMPRESS_GZIP_9 &&
3778 			    zfs_earlier_version(dsname,
3779 			    SPA_VERSION_GZIP_COMPRESSION)) {
3780 				return (SET_ERROR(ENOTSUP));
3781 			}
3782 
3783 			if (intval == ZIO_COMPRESS_ZLE &&
3784 			    zfs_earlier_version(dsname,
3785 			    SPA_VERSION_ZLE_COMPRESSION))
3786 				return (SET_ERROR(ENOTSUP));
3787 
3788 			if (intval == ZIO_COMPRESS_LZ4) {
3789 				spa_t *spa;
3790 
3791 				if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3792 					return (err);
3793 
3794 				if (!spa_feature_is_enabled(spa,
3795 				    SPA_FEATURE_LZ4_COMPRESS)) {
3796 					spa_close(spa, FTAG);
3797 					return (SET_ERROR(ENOTSUP));
3798 				}
3799 				spa_close(spa, FTAG);
3800 			}
3801 
3802 			/*
3803 			 * If this is a bootable dataset then
3804 			 * verify that the compression algorithm
3805 			 * is supported for booting. We must return
3806 			 * something other than ENOTSUP since it
3807 			 * implies a downrev pool version.
3808 			 */
3809 			if (zfs_is_bootfs(dsname) &&
3810 			    !BOOTFS_COMPRESS_VALID(intval)) {
3811 				return (SET_ERROR(ERANGE));
3812 			}
3813 		}
3814 		break;
3815 
3816 	case ZFS_PROP_COPIES:
3817 		if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3818 			return (SET_ERROR(ENOTSUP));
3819 		break;
3820 
3821 	case ZFS_PROP_DEDUP:
3822 		if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3823 			return (SET_ERROR(ENOTSUP));
3824 		break;
3825 
3826 	case ZFS_PROP_RECORDSIZE:
3827 		/* Record sizes above 128k need the feature to be enabled */
3828 		if (nvpair_value_uint64(pair, &intval) == 0 &&
3829 		    intval > SPA_OLD_MAXBLOCKSIZE) {
3830 			spa_t *spa;
3831 
3832 			/*
3833 			 * If this is a bootable dataset then
3834 			 * the we don't allow large (>128K) blocks,
3835 			 * because GRUB doesn't support them.
3836 			 */
3837 			if (zfs_is_bootfs(dsname) &&
3838 			    intval > SPA_OLD_MAXBLOCKSIZE) {
3839 				return (SET_ERROR(EDOM));
3840 			}
3841 
3842 			/*
3843 			 * We don't allow setting the property above 1MB,
3844 			 * unless the tunable has been changed.
3845 			 */
3846 			if (intval > zfs_max_recordsize ||
3847 			    intval > SPA_MAXBLOCKSIZE)
3848 				return (SET_ERROR(EDOM));
3849 
3850 			if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3851 				return (err);
3852 
3853 			if (!spa_feature_is_enabled(spa,
3854 			    SPA_FEATURE_LARGE_BLOCKS)) {
3855 				spa_close(spa, FTAG);
3856 				return (SET_ERROR(ENOTSUP));
3857 			}
3858 			spa_close(spa, FTAG);
3859 		}
3860 		break;
3861 
3862 	case ZFS_PROP_SHARESMB:
3863 		if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3864 			return (SET_ERROR(ENOTSUP));
3865 		break;
3866 
3867 	case ZFS_PROP_ACLINHERIT:
3868 		if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3869 		    nvpair_value_uint64(pair, &intval) == 0) {
3870 			if (intval == ZFS_ACL_PASSTHROUGH_X &&
3871 			    zfs_earlier_version(dsname,
3872 			    SPA_VERSION_PASSTHROUGH_X))
3873 				return (SET_ERROR(ENOTSUP));
3874 		}
3875 		break;
3876 	}
3877 
3878 	return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3879 }
3880 
3881 /*
3882  * Checks for a race condition to make sure we don't increment a feature flag
3883  * multiple times.
3884  */
3885 static int
3886 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
3887 {
3888 	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3889 	spa_feature_t *featurep = arg;
3890 
3891 	if (!spa_feature_is_active(spa, *featurep))
3892 		return (0);
3893 	else
3894 		return (SET_ERROR(EBUSY));
3895 }
3896 
3897 /*
3898  * The callback invoked on feature activation in the sync task caused by
3899  * zfs_prop_activate_feature.
3900  */
3901 static void
3902 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
3903 {
3904 	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3905 	spa_feature_t *featurep = arg;
3906 
3907 	spa_feature_incr(spa, *featurep, tx);
3908 }
3909 
3910 /*
3911  * Activates a feature on a pool in response to a property setting. This
3912  * creates a new sync task which modifies the pool to reflect the feature
3913  * as being active.
3914  */
3915 static int
3916 zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature)
3917 {
3918 	int err;
3919 
3920 	/* EBUSY here indicates that the feature is already active */
3921 	err = dsl_sync_task(spa_name(spa),
3922 	    zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
3923 	    &feature, 2, ZFS_SPACE_CHECK_RESERVED);
3924 
3925 	if (err != 0 && err != EBUSY)
3926 		return (err);
3927 	else
3928 		return (0);
3929 }
3930 
3931 /*
3932  * Removes properties from the given props list that fail permission checks
3933  * needed to clear them and to restore them in case of a receive error. For each
3934  * property, make sure we have both set and inherit permissions.
3935  *
3936  * Returns the first error encountered if any permission checks fail. If the
3937  * caller provides a non-NULL errlist, it also gives the complete list of names
3938  * of all the properties that failed a permission check along with the
3939  * corresponding error numbers. The caller is responsible for freeing the
3940  * returned errlist.
3941  *
3942  * If every property checks out successfully, zero is returned and the list
3943  * pointed at by errlist is NULL.
3944  */
3945 static int
3946 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
3947 {
3948 	zfs_cmd_t *zc;
3949 	nvpair_t *pair, *next_pair;
3950 	nvlist_t *errors;
3951 	int err, rv = 0;
3952 
3953 	if (props == NULL)
3954 		return (0);
3955 
3956 	VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3957 
3958 	zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
3959 	(void) strcpy(zc->zc_name, dataset);
3960 	pair = nvlist_next_nvpair(props, NULL);
3961 	while (pair != NULL) {
3962 		next_pair = nvlist_next_nvpair(props, pair);
3963 
3964 		(void) strcpy(zc->zc_value, nvpair_name(pair));
3965 		if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
3966 		    (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
3967 			VERIFY(nvlist_remove_nvpair(props, pair) == 0);
3968 			VERIFY(nvlist_add_int32(errors,
3969 			    zc->zc_value, err) == 0);
3970 		}
3971 		pair = next_pair;
3972 	}
3973 	kmem_free(zc, sizeof (zfs_cmd_t));
3974 
3975 	if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
3976 		nvlist_free(errors);
3977 		errors = NULL;
3978 	} else {
3979 		VERIFY(nvpair_value_int32(pair, &rv) == 0);
3980 	}
3981 
3982 	if (errlist == NULL)
3983 		nvlist_free(errors);
3984 	else
3985 		*errlist = errors;
3986 
3987 	return (rv);
3988 }
3989 
3990 static boolean_t
3991 propval_equals(nvpair_t *p1, nvpair_t *p2)
3992 {
3993 	if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
3994 		/* dsl_prop_get_all_impl() format */
3995 		nvlist_t *attrs;
3996 		VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
3997 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3998 		    &p1) == 0);
3999 	}
4000 
4001 	if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
4002 		nvlist_t *attrs;
4003 		VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
4004 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4005 		    &p2) == 0);
4006 	}
4007 
4008 	if (nvpair_type(p1) != nvpair_type(p2))
4009 		return (B_FALSE);
4010 
4011 	if (nvpair_type(p1) == DATA_TYPE_STRING) {
4012 		char *valstr1, *valstr2;
4013 
4014 		VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
4015 		VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
4016 		return (strcmp(valstr1, valstr2) == 0);
4017 	} else {
4018 		uint64_t intval1, intval2;
4019 
4020 		VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4021 		VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4022 		return (intval1 == intval2);
4023 	}
4024 }
4025 
4026 /*
4027  * Remove properties from props if they are not going to change (as determined
4028  * by comparison with origprops). Remove them from origprops as well, since we
4029  * do not need to clear or restore properties that won't change.
4030  */
4031 static void
4032 props_reduce(nvlist_t *props, nvlist_t *origprops)
4033 {
4034 	nvpair_t *pair, *next_pair;
4035 
4036 	if (origprops == NULL)
4037 		return; /* all props need to be received */
4038 
4039 	pair = nvlist_next_nvpair(props, NULL);
4040 	while (pair != NULL) {
4041 		const char *propname = nvpair_name(pair);
4042 		nvpair_t *match;
4043 
4044 		next_pair = nvlist_next_nvpair(props, pair);
4045 
4046 		if ((nvlist_lookup_nvpair(origprops, propname,
4047 		    &match) != 0) || !propval_equals(pair, match))
4048 			goto next; /* need to set received value */
4049 
4050 		/* don't clear the existing received value */
4051 		(void) nvlist_remove_nvpair(origprops, match);
4052 		/* don't bother receiving the property */
4053 		(void) nvlist_remove_nvpair(props, pair);
4054 next:
4055 		pair = next_pair;
4056 	}
4057 }
4058 
4059 #ifdef	DEBUG
4060 static boolean_t zfs_ioc_recv_inject_err;
4061 #endif
4062 
4063 /*
4064  * inputs:
4065  * zc_name		name of containing filesystem
4066  * zc_nvlist_src{_size}	nvlist of properties to apply
4067  * zc_value		name of snapshot to create
4068  * zc_string		name of clone origin (if DRR_FLAG_CLONE)
4069  * zc_cookie		file descriptor to recv from
4070  * zc_begin_record	the BEGIN record of the stream (not byteswapped)
4071  * zc_guid		force flag
4072  * zc_cleanup_fd	cleanup-on-exit file descriptor
4073  * zc_action_handle	handle for this guid/ds mapping (or zero on first call)
4074  *
4075  * outputs:
4076  * zc_cookie		number of bytes read
4077  * zc_nvlist_dst{_size} error for each unapplied received property
4078  * zc_obj		zprop_errflags_t
4079  * zc_action_handle	handle for this guid/ds mapping
4080  */
4081 static int
4082 zfs_ioc_recv(zfs_cmd_t *zc)
4083 {
4084 	file_t *fp;
4085 	dmu_recv_cookie_t drc;
4086 	boolean_t force = (boolean_t)zc->zc_guid;
4087 	int fd;
4088 	int error = 0;
4089 	int props_error = 0;
4090 	nvlist_t *errors;
4091 	offset_t off;
4092 	nvlist_t *props = NULL; /* sent properties */
4093 	nvlist_t *origprops = NULL; /* existing properties */
4094 	char *origin = NULL;
4095 	char *tosnap;
4096 	char tofs[ZFS_MAXNAMELEN];
4097 	boolean_t first_recvd_props = B_FALSE;
4098 
4099 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4100 	    strchr(zc->zc_value, '@') == NULL ||
4101 	    strchr(zc->zc_value, '%'))
4102 		return (SET_ERROR(EINVAL));
4103 
4104 	(void) strcpy(tofs, zc->zc_value);
4105 	tosnap = strchr(tofs, '@');
4106 	*tosnap++ = '\0';
4107 
4108 	if (zc->zc_nvlist_src != NULL &&
4109 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
4110 	    zc->zc_iflags, &props)) != 0)
4111 		return (error);
4112 
4113 	fd = zc->zc_cookie;
4114 	fp = getf(fd);
4115 	if (fp == NULL) {
4116 		nvlist_free(props);
4117 		return (SET_ERROR(EBADF));
4118 	}
4119 
4120 	VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4121 
4122 	if (zc->zc_string[0])
4123 		origin = zc->zc_string;
4124 
4125 	error = dmu_recv_begin(tofs, tosnap,
4126 	    &zc->zc_begin_record, force, origin, &drc);
4127 	if (error != 0)
4128 		goto out;
4129 
4130 	/*
4131 	 * Set properties before we receive the stream so that they are applied
4132 	 * to the new data. Note that we must call dmu_recv_stream() if
4133 	 * dmu_recv_begin() succeeds.
4134 	 */
4135 	if (props != NULL && !drc.drc_newfs) {
4136 		if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
4137 		    SPA_VERSION_RECVD_PROPS &&
4138 		    !dsl_prop_get_hasrecvd(tofs))
4139 			first_recvd_props = B_TRUE;
4140 
4141 		/*
4142 		 * If new received properties are supplied, they are to
4143 		 * completely replace the existing received properties, so stash
4144 		 * away the existing ones.
4145 		 */
4146 		if (dsl_prop_get_received(tofs, &origprops) == 0) {
4147 			nvlist_t *errlist = NULL;
4148 			/*
4149 			 * Don't bother writing a property if its value won't
4150 			 * change (and avoid the unnecessary security checks).
4151 			 *
4152 			 * The first receive after SPA_VERSION_RECVD_PROPS is a
4153 			 * special case where we blow away all local properties
4154 			 * regardless.
4155 			 */
4156 			if (!first_recvd_props)
4157 				props_reduce(props, origprops);
4158 			if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
4159 				(void) nvlist_merge(errors, errlist, 0);
4160 			nvlist_free(errlist);
4161 
4162 			if (clear_received_props(tofs, origprops,
4163 			    first_recvd_props ? NULL : props) != 0)
4164 				zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4165 		} else {
4166 			zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4167 		}
4168 	}
4169 
4170 	if (props != NULL) {
4171 		props_error = dsl_prop_set_hasrecvd(tofs);
4172 
4173 		if (props_error == 0) {
4174 			(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4175 			    props, errors);
4176 		}
4177 	}
4178 
4179 	if (zc->zc_nvlist_dst_size != 0 &&
4180 	    (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4181 	    put_nvlist(zc, errors) != 0)) {
4182 		/*
4183 		 * Caller made zc->zc_nvlist_dst less than the minimum expected
4184 		 * size or supplied an invalid address.
4185 		 */
4186 		props_error = SET_ERROR(EINVAL);
4187 	}
4188 
4189 	off = fp->f_offset;
4190 	error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
4191 	    &zc->zc_action_handle);
4192 
4193 	if (error == 0) {
4194 		zfsvfs_t *zfsvfs = NULL;
4195 
4196 		if (getzfsvfs(tofs, &zfsvfs) == 0) {
4197 			/* online recv */
4198 			int end_err;
4199 
4200 			error = zfs_suspend_fs(zfsvfs);
4201 			/*
4202 			 * If the suspend fails, then the recv_end will
4203 			 * likely also fail, and clean up after itself.
4204 			 */
4205 			end_err = dmu_recv_end(&drc, zfsvfs);
4206 			if (error == 0)
4207 				error = zfs_resume_fs(zfsvfs, tofs);
4208 			error = error ? error : end_err;
4209 			VFS_RELE(zfsvfs->z_vfs);
4210 		} else {
4211 			error = dmu_recv_end(&drc, NULL);
4212 		}
4213 	}
4214 
4215 	zc->zc_cookie = off - fp->f_offset;
4216 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4217 		fp->f_offset = off;
4218 
4219 #ifdef	DEBUG
4220 	if (zfs_ioc_recv_inject_err) {
4221 		zfs_ioc_recv_inject_err = B_FALSE;
4222 		error = 1;
4223 	}
4224 #endif
4225 	/*
4226 	 * On error, restore the original props.
4227 	 */
4228 	if (error != 0 && props != NULL && !drc.drc_newfs) {
4229 		if (clear_received_props(tofs, props, NULL) != 0) {
4230 			/*
4231 			 * We failed to clear the received properties.
4232 			 * Since we may have left a $recvd value on the
4233 			 * system, we can't clear the $hasrecvd flag.
4234 			 */
4235 			zc->zc_obj |= ZPROP_ERR_NORESTORE;
4236 		} else if (first_recvd_props) {
4237 			dsl_prop_unset_hasrecvd(tofs);
4238 		}
4239 
4240 		if (origprops == NULL && !drc.drc_newfs) {
4241 			/* We failed to stash the original properties. */
4242 			zc->zc_obj |= ZPROP_ERR_NORESTORE;
4243 		}
4244 
4245 		/*
4246 		 * dsl_props_set() will not convert RECEIVED to LOCAL on or
4247 		 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4248 		 * explictly if we're restoring local properties cleared in the
4249 		 * first new-style receive.
4250 		 */
4251 		if (origprops != NULL &&
4252 		    zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4253 		    ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4254 		    origprops, NULL) != 0) {
4255 			/*
4256 			 * We stashed the original properties but failed to
4257 			 * restore them.
4258 			 */
4259 			zc->zc_obj |= ZPROP_ERR_NORESTORE;
4260 		}
4261 	}
4262 out:
4263 	nvlist_free(props);
4264 	nvlist_free(origprops);
4265 	nvlist_free(errors);
4266 	releasef(fd);
4267 
4268 	if (error == 0)
4269 		error = props_error;
4270 
4271 	return (error);
4272 }
4273 
4274 /*
4275  * inputs:
4276  * zc_name	name of snapshot to send
4277  * zc_cookie	file descriptor to send stream to
4278  * zc_obj	fromorigin flag (mutually exclusive with zc_fromobj)
4279  * zc_sendobj	objsetid of snapshot to send
4280  * zc_fromobj	objsetid of incremental fromsnap (may be zero)
4281  * zc_guid	if set, estimate size of stream only.  zc_cookie is ignored.
4282  *		output size in zc_objset_type.
4283  * zc_flags	lzc_send_flags
4284  *
4285  * outputs:
4286  * zc_objset_type	estimated size, if zc_guid is set
4287  */
4288 static int
4289 zfs_ioc_send(zfs_cmd_t *zc)
4290 {
4291 	int error;
4292 	offset_t off;
4293 	boolean_t estimate = (zc->zc_guid != 0);
4294 	boolean_t embedok = (zc->zc_flags & 0x1);
4295 	boolean_t large_block_ok = (zc->zc_flags & 0x2);
4296 
4297 	if (zc->zc_obj != 0) {
4298 		dsl_pool_t *dp;
4299 		dsl_dataset_t *tosnap;
4300 
4301 		error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4302 		if (error != 0)
4303 			return (error);
4304 
4305 		error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4306 		if (error != 0) {
4307 			dsl_pool_rele(dp, FTAG);
4308 			return (error);
4309 		}
4310 
4311 		if (dsl_dir_is_clone(tosnap->ds_dir))
4312 			zc->zc_fromobj =
4313 			    dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj;
4314 		dsl_dataset_rele(tosnap, FTAG);
4315 		dsl_pool_rele(dp, FTAG);
4316 	}
4317 
4318 	if (estimate) {
4319 		dsl_pool_t *dp;
4320 		dsl_dataset_t *tosnap;
4321 		dsl_dataset_t *fromsnap = NULL;
4322 
4323 		error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4324 		if (error != 0)
4325 			return (error);
4326 
4327 		error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4328 		if (error != 0) {
4329 			dsl_pool_rele(dp, FTAG);
4330 			return (error);
4331 		}
4332 
4333 		if (zc->zc_fromobj != 0) {
4334 			error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4335 			    FTAG, &fromsnap);
4336 			if (error != 0) {
4337 				dsl_dataset_rele(tosnap, FTAG);
4338 				dsl_pool_rele(dp, FTAG);
4339 				return (error);
4340 			}
4341 		}
4342 
4343 		error = dmu_send_estimate(tosnap, fromsnap,
4344 		    &zc->zc_objset_type);
4345 
4346 		if (fromsnap != NULL)
4347 			dsl_dataset_rele(fromsnap, FTAG);
4348 		dsl_dataset_rele(tosnap, FTAG);
4349 		dsl_pool_rele(dp, FTAG);
4350 	} else {
4351 		file_t *fp = getf(zc->zc_cookie);
4352 		if (fp == NULL)
4353 			return (SET_ERROR(EBADF));
4354 
4355 		off = fp->f_offset;
4356 		error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4357 		    zc->zc_fromobj, embedok, large_block_ok,
4358 		    zc->zc_cookie, fp->f_vnode, &off);
4359 
4360 		if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4361 			fp->f_offset = off;
4362 		releasef(zc->zc_cookie);
4363 	}
4364 	return (error);
4365 }
4366 
4367 /*
4368  * inputs:
4369  * zc_name	name of snapshot on which to report progress
4370  * zc_cookie	file descriptor of send stream
4371  *
4372  * outputs:
4373  * zc_cookie	number of bytes written in send stream thus far
4374  */
4375 static int
4376 zfs_ioc_send_progress(zfs_cmd_t *zc)
4377 {
4378 	dsl_pool_t *dp;
4379 	dsl_dataset_t *ds;
4380 	dmu_sendarg_t *dsp = NULL;
4381 	int error;
4382 
4383 	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4384 	if (error != 0)
4385 		return (error);
4386 
4387 	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4388 	if (error != 0) {
4389 		dsl_pool_rele(dp, FTAG);
4390 		return (error);
4391 	}
4392 
4393 	mutex_enter(&ds->ds_sendstream_lock);
4394 
4395 	/*
4396 	 * Iterate over all the send streams currently active on this dataset.
4397 	 * If there's one which matches the specified file descriptor _and_ the
4398 	 * stream was started by the current process, return the progress of
4399 	 * that stream.
4400 	 */
4401 	for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4402 	    dsp = list_next(&ds->ds_sendstreams, dsp)) {
4403 		if (dsp->dsa_outfd == zc->zc_cookie &&
4404 		    dsp->dsa_proc == curproc)
4405 			break;
4406 	}
4407 
4408 	if (dsp != NULL)
4409 		zc->zc_cookie = *(dsp->dsa_off);
4410 	else
4411 		error = SET_ERROR(ENOENT);
4412 
4413 	mutex_exit(&ds->ds_sendstream_lock);
4414 	dsl_dataset_rele(ds, FTAG);
4415 	dsl_pool_rele(dp, FTAG);
4416 	return (error);
4417 }
4418 
4419 static int
4420 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4421 {
4422 	int id, error;
4423 
4424 	error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4425 	    &zc->zc_inject_record);
4426 
4427 	if (error == 0)
4428 		zc->zc_guid = (uint64_t)id;
4429 
4430 	return (error);
4431 }
4432 
4433 static int
4434 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4435 {
4436 	return (zio_clear_fault((int)zc->zc_guid));
4437 }
4438 
4439 static int
4440 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4441 {
4442 	int id = (int)zc->zc_guid;
4443 	int error;
4444 
4445 	error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4446 	    &zc->zc_inject_record);
4447 
4448 	zc->zc_guid = id;
4449 
4450 	return (error);
4451 }
4452 
4453 static int
4454 zfs_ioc_error_log(zfs_cmd_t *zc)
4455 {
4456 	spa_t *spa;
4457 	int error;
4458 	size_t count = (size_t)zc->zc_nvlist_dst_size;
4459 
4460 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4461 		return (error);
4462 
4463 	error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4464 	    &count);
4465 	if (error == 0)
4466 		zc->zc_nvlist_dst_size = count;
4467 	else
4468 		zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4469 
4470 	spa_close(spa, FTAG);
4471 
4472 	return (error);
4473 }
4474 
4475 static int
4476 zfs_ioc_clear(zfs_cmd_t *zc)
4477 {
4478 	spa_t *spa;
4479 	vdev_t *vd;
4480 	int error;
4481 
4482 	/*
4483 	 * On zpool clear we also fix up missing slogs
4484 	 */
4485 	mutex_enter(&spa_namespace_lock);
4486 	spa = spa_lookup(zc->zc_name);
4487 	if (spa == NULL) {
4488 		mutex_exit(&spa_namespace_lock);
4489 		return (SET_ERROR(EIO));
4490 	}
4491 	if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4492 		/* we need to let spa_open/spa_load clear the chains */
4493 		spa_set_log_state(spa, SPA_LOG_CLEAR);
4494 	}
4495 	spa->spa_last_open_failed = 0;
4496 	mutex_exit(&spa_namespace_lock);
4497 
4498 	if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4499 		error = spa_open(zc->zc_name, &spa, FTAG);
4500 	} else {
4501 		nvlist_t *policy;
4502 		nvlist_t *config = NULL;
4503 
4504 		if (zc->zc_nvlist_src == NULL)
4505 			return (SET_ERROR(EINVAL));
4506 
4507 		if ((error = get_nvlist(zc->zc_nvlist_src,
4508 		    zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4509 			error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4510 			    policy, &config);
4511 			if (config != NULL) {
4512 				int err;
4513 
4514 				if ((err = put_nvlist(zc, config)) != 0)
4515 					error = err;
4516 				nvlist_free(config);
4517 			}
4518 			nvlist_free(policy);
4519 		}
4520 	}
4521 
4522 	if (error != 0)
4523 		return (error);
4524 
4525 	spa_vdev_state_enter(spa, SCL_NONE);
4526 
4527 	if (zc->zc_guid == 0) {
4528 		vd = NULL;
4529 	} else {
4530 		vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4531 		if (vd == NULL) {
4532 			(void) spa_vdev_state_exit(spa, NULL, ENODEV);
4533 			spa_close(spa, FTAG);
4534 			return (SET_ERROR(ENODEV));
4535 		}
4536 	}
4537 
4538 	vdev_clear(spa, vd);
4539 
4540 	(void) spa_vdev_state_exit(spa, NULL, 0);
4541 
4542 	/*
4543 	 * Resume any suspended I/Os.
4544 	 */
4545 	if (zio_resume(spa) != 0)
4546 		error = SET_ERROR(EIO);
4547 
4548 	spa_close(spa, FTAG);
4549 
4550 	return (error);
4551 }
4552 
4553 static int
4554 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4555 {
4556 	spa_t *spa;
4557 	int error;
4558 
4559 	error = spa_open(zc->zc_name, &spa, FTAG);
4560 	if (error != 0)
4561 		return (error);
4562 
4563 	spa_vdev_state_enter(spa, SCL_NONE);
4564 
4565 	/*
4566 	 * If a resilver is already in progress then set the
4567 	 * spa_scrub_reopen flag to B_TRUE so that we don't restart
4568 	 * the scan as a side effect of the reopen. Otherwise, let
4569 	 * vdev_open() decided if a resilver is required.
4570 	 */
4571 	spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4572 	vdev_reopen(spa->spa_root_vdev);
4573 	spa->spa_scrub_reopen = B_FALSE;
4574 
4575 	(void) spa_vdev_state_exit(spa, NULL, 0);
4576 	spa_close(spa, FTAG);
4577 	return (0);
4578 }
4579 /*
4580  * inputs:
4581  * zc_name	name of filesystem
4582  * zc_value	name of origin snapshot
4583  *
4584  * outputs:
4585  * zc_string	name of conflicting snapshot, if there is one
4586  */
4587 static int
4588 zfs_ioc_promote(zfs_cmd_t *zc)
4589 {
4590 	char *cp;
4591 
4592 	/*
4593 	 * We don't need to unmount *all* the origin fs's snapshots, but
4594 	 * it's easier.
4595 	 */
4596 	cp = strchr(zc->zc_value, '@');
4597 	if (cp)
4598 		*cp = '\0';
4599 	(void) dmu_objset_find(zc->zc_value,
4600 	    zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
4601 	return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4602 }
4603 
4604 /*
4605  * Retrieve a single {user|group}{used|quota}@... property.
4606  *
4607  * inputs:
4608  * zc_name	name of filesystem
4609  * zc_objset_type zfs_userquota_prop_t
4610  * zc_value	domain name (eg. "S-1-234-567-89")
4611  * zc_guid	RID/UID/GID
4612  *
4613  * outputs:
4614  * zc_cookie	property value
4615  */
4616 static int
4617 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4618 {
4619 	zfsvfs_t *zfsvfs;
4620 	int error;
4621 
4622 	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4623 		return (SET_ERROR(EINVAL));
4624 
4625 	error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4626 	if (error != 0)
4627 		return (error);
4628 
4629 	error = zfs_userspace_one(zfsvfs,
4630 	    zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4631 	zfsvfs_rele(zfsvfs, FTAG);
4632 
4633 	return (error);
4634 }
4635 
4636 /*
4637  * inputs:
4638  * zc_name		name of filesystem
4639  * zc_cookie		zap cursor
4640  * zc_objset_type	zfs_userquota_prop_t
4641  * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4642  *
4643  * outputs:
4644  * zc_nvlist_dst[_size]	data buffer (array of zfs_useracct_t)
4645  * zc_cookie	zap cursor
4646  */
4647 static int
4648 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4649 {
4650 	zfsvfs_t *zfsvfs;
4651 	int bufsize = zc->zc_nvlist_dst_size;
4652 
4653 	if (bufsize <= 0)
4654 		return (SET_ERROR(ENOMEM));
4655 
4656 	int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4657 	if (error != 0)
4658 		return (error);
4659 
4660 	void *buf = kmem_alloc(bufsize, KM_SLEEP);
4661 
4662 	error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
4663 	    buf, &zc->zc_nvlist_dst_size);
4664 
4665 	if (error == 0) {
4666 		error = xcopyout(buf,
4667 		    (void *)(uintptr_t)zc->zc_nvlist_dst,
4668 		    zc->zc_nvlist_dst_size);
4669 	}
4670 	kmem_free(buf, bufsize);
4671 	zfsvfs_rele(zfsvfs, FTAG);
4672 
4673 	return (error);
4674 }
4675 
4676 /*
4677  * inputs:
4678  * zc_name		name of filesystem
4679  *
4680  * outputs:
4681  * none
4682  */
4683 static int
4684 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4685 {
4686 	objset_t *os;
4687 	int error = 0;
4688 	zfsvfs_t *zfsvfs;
4689 
4690 	if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
4691 		if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
4692 			/*
4693 			 * If userused is not enabled, it may be because the
4694 			 * objset needs to be closed & reopened (to grow the
4695 			 * objset_phys_t).  Suspend/resume the fs will do that.
4696 			 */
4697 			error = zfs_suspend_fs(zfsvfs);
4698 			if (error == 0) {
4699 				dmu_objset_refresh_ownership(zfsvfs->z_os,
4700 				    zfsvfs);
4701 				error = zfs_resume_fs(zfsvfs, zc->zc_name);
4702 			}
4703 		}
4704 		if (error == 0)
4705 			error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
4706 		VFS_RELE(zfsvfs->z_vfs);
4707 	} else {
4708 		/* XXX kind of reading contents without owning */
4709 		error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4710 		if (error != 0)
4711 			return (error);
4712 
4713 		error = dmu_objset_userspace_upgrade(os);
4714 		dmu_objset_rele(os, FTAG);
4715 	}
4716 
4717 	return (error);
4718 }
4719 
4720 /*
4721  * We don't want to have a hard dependency
4722  * against some special symbols in sharefs
4723  * nfs, and smbsrv.  Determine them if needed when
4724  * the first file system is shared.
4725  * Neither sharefs, nfs or smbsrv are unloadable modules.
4726  */
4727 int (*znfsexport_fs)(void *arg);
4728 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
4729 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
4730 
4731 int zfs_nfsshare_inited;
4732 int zfs_smbshare_inited;
4733 
4734 ddi_modhandle_t nfs_mod;
4735 ddi_modhandle_t sharefs_mod;
4736 ddi_modhandle_t smbsrv_mod;
4737 kmutex_t zfs_share_lock;
4738 
4739 static int
4740 zfs_init_sharefs()
4741 {
4742 	int error;
4743 
4744 	ASSERT(MUTEX_HELD(&zfs_share_lock));
4745 	/* Both NFS and SMB shares also require sharetab support. */
4746 	if (sharefs_mod == NULL && ((sharefs_mod =
4747 	    ddi_modopen("fs/sharefs",
4748 	    KRTLD_MODE_FIRST, &error)) == NULL)) {
4749 		return (SET_ERROR(ENOSYS));
4750 	}
4751 	if (zshare_fs == NULL && ((zshare_fs =
4752 	    (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
4753 	    ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
4754 		return (SET_ERROR(ENOSYS));
4755 	}
4756 	return (0);
4757 }
4758 
4759 static int
4760 zfs_ioc_share(zfs_cmd_t *zc)
4761 {
4762 	int error;
4763 	int opcode;
4764 
4765 	switch (zc->zc_share.z_sharetype) {
4766 	case ZFS_SHARE_NFS:
4767 	case ZFS_UNSHARE_NFS:
4768 		if (zfs_nfsshare_inited == 0) {
4769 			mutex_enter(&zfs_share_lock);
4770 			if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
4771 			    KRTLD_MODE_FIRST, &error)) == NULL)) {
4772 				mutex_exit(&zfs_share_lock);
4773 				return (SET_ERROR(ENOSYS));
4774 			}
4775 			if (znfsexport_fs == NULL &&
4776 			    ((znfsexport_fs = (int (*)(void *))
4777 			    ddi_modsym(nfs_mod,
4778 			    "nfs_export", &error)) == NULL)) {
4779 				mutex_exit(&zfs_share_lock);
4780 				return (SET_ERROR(ENOSYS));
4781 			}
4782 			error = zfs_init_sharefs();
4783 			if (error != 0) {
4784 				mutex_exit(&zfs_share_lock);
4785 				return (SET_ERROR(ENOSYS));
4786 			}
4787 			zfs_nfsshare_inited = 1;
4788 			mutex_exit(&zfs_share_lock);
4789 		}
4790 		break;
4791 	case ZFS_SHARE_SMB:
4792 	case ZFS_UNSHARE_SMB:
4793 		if (zfs_smbshare_inited == 0) {
4794 			mutex_enter(&zfs_share_lock);
4795 			if (smbsrv_mod == NULL && ((smbsrv_mod =
4796 			    ddi_modopen("drv/smbsrv",
4797 			    KRTLD_MODE_FIRST, &error)) == NULL)) {
4798 				mutex_exit(&zfs_share_lock);
4799 				return (SET_ERROR(ENOSYS));
4800 			}
4801 			if (zsmbexport_fs == NULL && ((zsmbexport_fs =
4802 			    (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
4803 			    "smb_server_share", &error)) == NULL)) {
4804 				mutex_exit(&zfs_share_lock);
4805 				return (SET_ERROR(ENOSYS));
4806 			}
4807 			error = zfs_init_sharefs();
4808 			if (error != 0) {
4809 				mutex_exit(&zfs_share_lock);
4810 				return (SET_ERROR(ENOSYS));
4811 			}
4812 			zfs_smbshare_inited = 1;
4813 			mutex_exit(&zfs_share_lock);
4814 		}
4815 		break;
4816 	default:
4817 		return (SET_ERROR(EINVAL));
4818 	}
4819 
4820 	switch (zc->zc_share.z_sharetype) {
4821 	case ZFS_SHARE_NFS:
4822 	case ZFS_UNSHARE_NFS:
4823 		if (error =
4824 		    znfsexport_fs((void *)
4825 		    (uintptr_t)zc->zc_share.z_exportdata))
4826 			return (error);
4827 		break;
4828 	case ZFS_SHARE_SMB:
4829 	case ZFS_UNSHARE_SMB:
4830 		if (error = zsmbexport_fs((void *)
4831 		    (uintptr_t)zc->zc_share.z_exportdata,
4832 		    zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
4833 		    B_TRUE: B_FALSE)) {
4834 			return (error);
4835 		}
4836 		break;
4837 	}
4838 
4839 	opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
4840 	    zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
4841 	    SHAREFS_ADD : SHAREFS_REMOVE;
4842 
4843 	/*
4844 	 * Add or remove share from sharetab
4845 	 */
4846 	error = zshare_fs(opcode,
4847 	    (void *)(uintptr_t)zc->zc_share.z_sharedata,
4848 	    zc->zc_share.z_sharemax);
4849 
4850 	return (error);
4851 
4852 }
4853 
4854 ace_t full_access[] = {
4855 	{(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4856 };
4857 
4858 /*
4859  * inputs:
4860  * zc_name		name of containing filesystem
4861  * zc_obj		object # beyond which we want next in-use object #
4862  *
4863  * outputs:
4864  * zc_obj		next in-use object #
4865  */
4866 static int
4867 zfs_ioc_next_obj(zfs_cmd_t *zc)
4868 {
4869 	objset_t *os = NULL;
4870 	int error;
4871 
4872 	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4873 	if (error != 0)
4874 		return (error);
4875 
4876 	error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4877 	    dsl_dataset_phys(os->os_dsl_dataset)->ds_prev_snap_txg);
4878 
4879 	dmu_objset_rele(os, FTAG);
4880 	return (error);
4881 }
4882 
4883 /*
4884  * inputs:
4885  * zc_name		name of filesystem
4886  * zc_value		prefix name for snapshot
4887  * zc_cleanup_fd	cleanup-on-exit file descriptor for calling process
4888  *
4889  * outputs:
4890  * zc_value		short name of new snapshot
4891  */
4892 static int
4893 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
4894 {
4895 	char *snap_name;
4896 	char *hold_name;
4897 	int error;
4898 	minor_t minor;
4899 
4900 	error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
4901 	if (error != 0)
4902 		return (error);
4903 
4904 	snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
4905 	    (u_longlong_t)ddi_get_lbolt64());
4906 	hold_name = kmem_asprintf("%%%s", zc->zc_value);
4907 
4908 	error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
4909 	    hold_name);
4910 	if (error == 0)
4911 		(void) strcpy(zc->zc_value, snap_name);
4912 	strfree(snap_name);
4913 	strfree(hold_name);
4914 	zfs_onexit_fd_rele(zc->zc_cleanup_fd);
4915 	return (error);
4916 }
4917 
4918 /*
4919  * inputs:
4920  * zc_name		name of "to" snapshot
4921  * zc_value		name of "from" snapshot
4922  * zc_cookie		file descriptor to write diff data on
4923  *
4924  * outputs:
4925  * dmu_diff_record_t's to the file descriptor
4926  */
4927 static int
4928 zfs_ioc_diff(zfs_cmd_t *zc)
4929 {
4930 	file_t *fp;
4931 	offset_t off;
4932 	int error;
4933 
4934 	fp = getf(zc->zc_cookie);
4935 	if (fp == NULL)
4936 		return (SET_ERROR(EBADF));
4937 
4938 	off = fp->f_offset;
4939 
4940 	error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
4941 
4942 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4943 		fp->f_offset = off;
4944 	releasef(zc->zc_cookie);
4945 
4946 	return (error);
4947 }
4948 
4949 /*
4950  * Remove all ACL files in shares dir
4951  */
4952 static int
4953 zfs_smb_acl_purge(znode_t *dzp)
4954 {
4955 	zap_cursor_t	zc;
4956 	zap_attribute_t	zap;
4957 	zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
4958 	int error;
4959 
4960 	for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
4961 	    (error = zap_cursor_retrieve(&zc, &zap)) == 0;
4962 	    zap_cursor_advance(&zc)) {
4963 		if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
4964 		    NULL, 0)) != 0)
4965 			break;
4966 	}
4967 	zap_cursor_fini(&zc);
4968 	return (error);
4969 }
4970 
4971 static int
4972 zfs_ioc_smb_acl(zfs_cmd_t *zc)
4973 {
4974 	vnode_t *vp;
4975 	znode_t *dzp;
4976 	vnode_t *resourcevp = NULL;
4977 	znode_t *sharedir;
4978 	zfsvfs_t *zfsvfs;
4979 	nvlist_t *nvlist;
4980 	char *src, *target;
4981 	vattr_t vattr;
4982 	vsecattr_t vsec;
4983 	int error = 0;
4984 
4985 	if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
4986 	    NO_FOLLOW, NULL, &vp)) != 0)
4987 		return (error);
4988 
4989 	/* Now make sure mntpnt and dataset are ZFS */
4990 
4991 	if (vp->v_vfsp->vfs_fstype != zfsfstype ||
4992 	    (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
4993 	    zc->zc_name) != 0)) {
4994 		VN_RELE(vp);
4995 		return (SET_ERROR(EINVAL));
4996 	}
4997 
4998 	dzp = VTOZ(vp);
4999 	zfsvfs = dzp->z_zfsvfs;
5000 	ZFS_ENTER(zfsvfs);
5001 
5002 	/*
5003 	 * Create share dir if its missing.
5004 	 */
5005 	mutex_enter(&zfsvfs->z_lock);
5006 	if (zfsvfs->z_shares_dir == 0) {
5007 		dmu_tx_t *tx;
5008 
5009 		tx = dmu_tx_create(zfsvfs->z_os);
5010 		dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
5011 		    ZFS_SHARES_DIR);
5012 		dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
5013 		error = dmu_tx_assign(tx, TXG_WAIT);
5014 		if (error != 0) {
5015 			dmu_tx_abort(tx);
5016 		} else {
5017 			error = zfs_create_share_dir(zfsvfs, tx);
5018 			dmu_tx_commit(tx);
5019 		}
5020 		if (error != 0) {
5021 			mutex_exit(&zfsvfs->z_lock);
5022 			VN_RELE(vp);
5023 			ZFS_EXIT(zfsvfs);
5024 			return (error);
5025 		}
5026 	}
5027 	mutex_exit(&zfsvfs->z_lock);
5028 
5029 	ASSERT(zfsvfs->z_shares_dir);
5030 	if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
5031 		VN_RELE(vp);
5032 		ZFS_EXIT(zfsvfs);
5033 		return (error);
5034 	}
5035 
5036 	switch (zc->zc_cookie) {
5037 	case ZFS_SMB_ACL_ADD:
5038 		vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
5039 		vattr.va_type = VREG;
5040 		vattr.va_mode = S_IFREG|0777;
5041 		vattr.va_uid = 0;
5042 		vattr.va_gid = 0;
5043 
5044 		vsec.vsa_mask = VSA_ACE;
5045 		vsec.vsa_aclentp = &full_access;
5046 		vsec.vsa_aclentsz = sizeof (full_access);
5047 		vsec.vsa_aclcnt = 1;
5048 
5049 		error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
5050 		    &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
5051 		if (resourcevp)
5052 			VN_RELE(resourcevp);
5053 		break;
5054 
5055 	case ZFS_SMB_ACL_REMOVE:
5056 		error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
5057 		    NULL, 0);
5058 		break;
5059 
5060 	case ZFS_SMB_ACL_RENAME:
5061 		if ((error = get_nvlist(zc->zc_nvlist_src,
5062 		    zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
5063 			VN_RELE(vp);
5064 			ZFS_EXIT(zfsvfs);
5065 			return (error);
5066 		}
5067 		if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
5068 		    nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
5069 		    &target)) {
5070 			VN_RELE(vp);
5071 			VN_RELE(ZTOV(sharedir));
5072 			ZFS_EXIT(zfsvfs);
5073 			nvlist_free(nvlist);
5074 			return (error);
5075 		}
5076 		error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
5077 		    kcred, NULL, 0);
5078 		nvlist_free(nvlist);
5079 		break;
5080 
5081 	case ZFS_SMB_ACL_PURGE:
5082 		error = zfs_smb_acl_purge(sharedir);
5083 		break;
5084 
5085 	default:
5086 		error = SET_ERROR(EINVAL);
5087 		break;
5088 	}
5089 
5090 	VN_RELE(vp);
5091 	VN_RELE(ZTOV(sharedir));
5092 
5093 	ZFS_EXIT(zfsvfs);
5094 
5095 	return (error);
5096 }
5097 
5098 /*
5099  * innvl: {
5100  *     "holds" -> { snapname -> holdname (string), ... }
5101  *     (optional) "cleanup_fd" -> fd (int32)
5102  * }
5103  *
5104  * outnvl: {
5105  *     snapname -> error value (int32)
5106  *     ...
5107  * }
5108  */
5109 /* ARGSUSED */
5110 static int
5111 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
5112 {
5113 	nvlist_t *holds;
5114 	int cleanup_fd = -1;
5115 	int error;
5116 	minor_t minor = 0;
5117 
5118 	error = nvlist_lookup_nvlist(args, "holds", &holds);
5119 	if (error != 0)
5120 		return (SET_ERROR(EINVAL));
5121 
5122 	if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
5123 		error = zfs_onexit_fd_hold(cleanup_fd, &minor);
5124 		if (error != 0)
5125 			return (error);
5126 	}
5127 
5128 	error = dsl_dataset_user_hold(holds, minor, errlist);
5129 	if (minor != 0)
5130 		zfs_onexit_fd_rele(cleanup_fd);
5131 	return (error);
5132 }
5133 
5134 /*
5135  * innvl is not used.
5136  *
5137  * outnvl: {
5138  *    holdname -> time added (uint64 seconds since epoch)
5139  *    ...
5140  * }
5141  */
5142 /* ARGSUSED */
5143 static int
5144 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
5145 {
5146 	return (dsl_dataset_get_holds(snapname, outnvl));
5147 }
5148 
5149 /*
5150  * innvl: {
5151  *     snapname -> { holdname, ... }
5152  *     ...
5153  * }
5154  *
5155  * outnvl: {
5156  *     snapname -> error value (int32)
5157  *     ...
5158  * }
5159  */
5160 /* ARGSUSED */
5161 static int
5162 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
5163 {
5164 	return (dsl_dataset_user_release(holds, errlist));
5165 }
5166 
5167 /*
5168  * inputs:
5169  * zc_name		name of new filesystem or snapshot
5170  * zc_value		full name of old snapshot
5171  *
5172  * outputs:
5173  * zc_cookie		space in bytes
5174  * zc_objset_type	compressed space in bytes
5175  * zc_perm_action	uncompressed space in bytes
5176  */
5177 static int
5178 zfs_ioc_space_written(zfs_cmd_t *zc)
5179 {
5180 	int error;
5181 	dsl_pool_t *dp;
5182 	dsl_dataset_t *new, *old;
5183 
5184 	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5185 	if (error != 0)
5186 		return (error);
5187 	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5188 	if (error != 0) {
5189 		dsl_pool_rele(dp, FTAG);
5190 		return (error);
5191 	}
5192 	error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5193 	if (error != 0) {
5194 		dsl_dataset_rele(new, FTAG);
5195 		dsl_pool_rele(dp, FTAG);
5196 		return (error);
5197 	}
5198 
5199 	error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5200 	    &zc->zc_objset_type, &zc->zc_perm_action);
5201 	dsl_dataset_rele(old, FTAG);
5202 	dsl_dataset_rele(new, FTAG);
5203 	dsl_pool_rele(dp, FTAG);
5204 	return (error);
5205 }
5206 
5207 /*
5208  * innvl: {
5209  *     "firstsnap" -> snapshot name
5210  * }
5211  *
5212  * outnvl: {
5213  *     "used" -> space in bytes
5214  *     "compressed" -> compressed space in bytes
5215  *     "uncompressed" -> uncompressed space in bytes
5216  * }
5217  */
5218 static int
5219 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5220 {
5221 	int error;
5222 	dsl_pool_t *dp;
5223 	dsl_dataset_t *new, *old;
5224 	char *firstsnap;
5225 	uint64_t used, comp, uncomp;
5226 
5227 	if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5228 		return (SET_ERROR(EINVAL));
5229 
5230 	error = dsl_pool_hold(lastsnap, FTAG, &dp);
5231 	if (error != 0)
5232 		return (error);
5233 
5234 	error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5235 	if (error != 0) {
5236 		dsl_pool_rele(dp, FTAG);
5237 		return (error);
5238 	}
5239 	error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5240 	if (error != 0) {
5241 		dsl_dataset_rele(new, FTAG);
5242 		dsl_pool_rele(dp, FTAG);
5243 		return (error);
5244 	}
5245 
5246 	error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5247 	dsl_dataset_rele(old, FTAG);
5248 	dsl_dataset_rele(new, FTAG);
5249 	dsl_pool_rele(dp, FTAG);
5250 	fnvlist_add_uint64(outnvl, "used", used);
5251 	fnvlist_add_uint64(outnvl, "compressed", comp);
5252 	fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5253 	return (error);
5254 }
5255 
5256 /*
5257  * innvl: {
5258  *     "fd" -> file descriptor to write stream to (int32)
5259  *     (optional) "fromsnap" -> full snap name to send an incremental from
5260  *     (optional) "largeblockok" -> (value ignored)
5261  *         indicates that blocks > 128KB are permitted
5262  *     (optional) "embedok" -> (value ignored)
5263  *         presence indicates DRR_WRITE_EMBEDDED records are permitted
5264  * }
5265  *
5266  * outnvl is unused
5267  */
5268 /* ARGSUSED */
5269 static int
5270 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5271 {
5272 	int error;
5273 	offset_t off;
5274 	char *fromname = NULL;
5275 	int fd;
5276 	boolean_t largeblockok;
5277 	boolean_t embedok;
5278 
5279 	error = nvlist_lookup_int32(innvl, "fd", &fd);
5280 	if (error != 0)
5281 		return (SET_ERROR(EINVAL));
5282 
5283 	(void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5284 
5285 	largeblockok = nvlist_exists(innvl, "largeblockok");
5286 	embedok = nvlist_exists(innvl, "embedok");
5287 
5288 	file_t *fp = getf(fd);
5289 	if (fp == NULL)
5290 		return (SET_ERROR(EBADF));
5291 
5292 	off = fp->f_offset;
5293 	error = dmu_send(snapname, fromname, embedok, largeblockok,
5294 	    fd, fp->f_vnode, &off);
5295 
5296 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5297 		fp->f_offset = off;
5298 	releasef(fd);
5299 	return (error);
5300 }
5301 
5302 /*
5303  * Determine approximately how large a zfs send stream will be -- the number
5304  * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5305  *
5306  * innvl: {
5307  *     (optional) "from" -> full snap or bookmark name to send an incremental
5308  *                          from
5309  * }
5310  *
5311  * outnvl: {
5312  *     "space" -> bytes of space (uint64)
5313  * }
5314  */
5315 static int
5316 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5317 {
5318 	dsl_pool_t *dp;
5319 	dsl_dataset_t *tosnap;
5320 	int error;
5321 	char *fromname;
5322 	uint64_t space;
5323 
5324 	error = dsl_pool_hold(snapname, FTAG, &dp);
5325 	if (error != 0)
5326 		return (error);
5327 
5328 	error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5329 	if (error != 0) {
5330 		dsl_pool_rele(dp, FTAG);
5331 		return (error);
5332 	}
5333 
5334 	error = nvlist_lookup_string(innvl, "from", &fromname);
5335 	if (error == 0) {
5336 		if (strchr(fromname, '@') != NULL) {
5337 			/*
5338 			 * If from is a snapshot, hold it and use the more
5339 			 * efficient dmu_send_estimate to estimate send space
5340 			 * size using deadlists.
5341 			 */
5342 			dsl_dataset_t *fromsnap;
5343 			error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5344 			if (error != 0)
5345 				goto out;
5346 			error = dmu_send_estimate(tosnap, fromsnap, &space);
5347 			dsl_dataset_rele(fromsnap, FTAG);
5348 		} else if (strchr(fromname, '#') != NULL) {
5349 			/*
5350 			 * If from is a bookmark, fetch the creation TXG of the
5351 			 * snapshot it was created from and use that to find
5352 			 * blocks that were born after it.
5353 			 */
5354 			zfs_bookmark_phys_t frombm;
5355 
5356 			error = dsl_bookmark_lookup(dp, fromname, tosnap,
5357 			    &frombm);
5358 			if (error != 0)
5359 				goto out;
5360 			error = dmu_send_estimate_from_txg(tosnap,
5361 			    frombm.zbm_creation_txg, &space);
5362 		} else {
5363 			/*
5364 			 * from is not properly formatted as a snapshot or
5365 			 * bookmark
5366 			 */
5367 			error = SET_ERROR(EINVAL);
5368 			goto out;
5369 		}
5370 	} else {
5371 		// If estimating the size of a full send, use dmu_send_estimate
5372 		error = dmu_send_estimate(tosnap, NULL, &space);
5373 	}
5374 
5375 	fnvlist_add_uint64(outnvl, "space", space);
5376 
5377 out:
5378 	dsl_dataset_rele(tosnap, FTAG);
5379 	dsl_pool_rele(dp, FTAG);
5380 	return (error);
5381 }
5382 
5383 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5384 
5385 static void
5386 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5387     zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5388     boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5389 {
5390 	zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5391 
5392 	ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5393 	ASSERT3U(ioc, <, ZFS_IOC_LAST);
5394 	ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5395 	ASSERT3P(vec->zvec_func, ==, NULL);
5396 
5397 	vec->zvec_legacy_func = func;
5398 	vec->zvec_secpolicy = secpolicy;
5399 	vec->zvec_namecheck = namecheck;
5400 	vec->zvec_allow_log = log_history;
5401 	vec->zvec_pool_check = pool_check;
5402 }
5403 
5404 /*
5405  * See the block comment at the beginning of this file for details on
5406  * each argument to this function.
5407  */
5408 static void
5409 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5410     zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5411     zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5412     boolean_t allow_log)
5413 {
5414 	zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5415 
5416 	ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5417 	ASSERT3U(ioc, <, ZFS_IOC_LAST);
5418 	ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5419 	ASSERT3P(vec->zvec_func, ==, NULL);
5420 
5421 	/* if we are logging, the name must be valid */
5422 	ASSERT(!allow_log || namecheck != NO_NAME);
5423 
5424 	vec->zvec_name = name;
5425 	vec->zvec_func = func;
5426 	vec->zvec_secpolicy = secpolicy;
5427 	vec->zvec_namecheck = namecheck;
5428 	vec->zvec_pool_check = pool_check;
5429 	vec->zvec_smush_outnvlist = smush_outnvlist;
5430 	vec->zvec_allow_log = allow_log;
5431 }
5432 
5433 static void
5434 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5435     zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5436     zfs_ioc_poolcheck_t pool_check)
5437 {
5438 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
5439 	    POOL_NAME, log_history, pool_check);
5440 }
5441 
5442 static void
5443 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5444     zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5445 {
5446 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
5447 	    DATASET_NAME, B_FALSE, pool_check);
5448 }
5449 
5450 static void
5451 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5452 {
5453 	zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5454 	    POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5455 }
5456 
5457 static void
5458 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5459     zfs_secpolicy_func_t *secpolicy)
5460 {
5461 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
5462 	    NO_NAME, B_FALSE, POOL_CHECK_NONE);
5463 }
5464 
5465 static void
5466 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5467     zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5468 {
5469 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
5470 	    DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5471 }
5472 
5473 static void
5474 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5475 {
5476 	zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5477 	    zfs_secpolicy_read);
5478 }
5479 
5480 static void
5481 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5482 	zfs_secpolicy_func_t *secpolicy)
5483 {
5484 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
5485 	    DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5486 }
5487 
5488 static void
5489 zfs_ioctl_init(void)
5490 {
5491 	zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5492 	    zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5493 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5494 
5495 	zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5496 	    zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5497 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5498 
5499 	zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5500 	    zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5501 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5502 
5503 	zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5504 	    zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5505 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5506 
5507 	zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5508 	    zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5509 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5510 
5511 	zfs_ioctl_register("create", ZFS_IOC_CREATE,
5512 	    zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5513 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5514 
5515 	zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5516 	    zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5517 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5518 
5519 	zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5520 	    zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5521 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5522 
5523 	zfs_ioctl_register("hold", ZFS_IOC_HOLD,
5524 	    zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
5525 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5526 	zfs_ioctl_register("release", ZFS_IOC_RELEASE,
5527 	    zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
5528 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5529 
5530 	zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
5531 	    zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
5532 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5533 
5534 	zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
5535 	    zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
5536 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
5537 
5538 	zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
5539 	    zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
5540 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5541 
5542 	zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
5543 	    zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
5544 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5545 
5546 	zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
5547 	    zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
5548 	    POOL_NAME,
5549 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5550 
5551 	/* IOCTLS that use the legacy function signature */
5552 
5553 	zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5554 	    zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5555 
5556 	zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5557 	    zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5558 	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5559 	    zfs_ioc_pool_scan);
5560 	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5561 	    zfs_ioc_pool_upgrade);
5562 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5563 	    zfs_ioc_vdev_add);
5564 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5565 	    zfs_ioc_vdev_remove);
5566 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5567 	    zfs_ioc_vdev_set_state);
5568 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5569 	    zfs_ioc_vdev_attach);
5570 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5571 	    zfs_ioc_vdev_detach);
5572 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5573 	    zfs_ioc_vdev_setpath);
5574 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5575 	    zfs_ioc_vdev_setfru);
5576 	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5577 	    zfs_ioc_pool_set_props);
5578 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5579 	    zfs_ioc_vdev_split);
5580 	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5581 	    zfs_ioc_pool_reguid);
5582 
5583 	zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5584 	    zfs_ioc_pool_configs, zfs_secpolicy_none);
5585 	zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5586 	    zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5587 	zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5588 	    zfs_ioc_inject_fault, zfs_secpolicy_inject);
5589 	zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5590 	    zfs_ioc_clear_fault, zfs_secpolicy_inject);
5591 	zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
5592 	    zfs_ioc_inject_list_next, zfs_secpolicy_inject);
5593 
5594 	/*
5595 	 * pool destroy, and export don't log the history as part of
5596 	 * zfsdev_ioctl, but rather zfs_ioc_pool_export
5597 	 * does the logging of those commands.
5598 	 */
5599 	zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
5600 	    zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5601 	zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
5602 	    zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5603 
5604 	zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
5605 	    zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5606 	zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
5607 	    zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5608 
5609 	zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
5610 	    zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
5611 	zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
5612 	    zfs_ioc_dsobj_to_dsname,
5613 	    zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
5614 	zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
5615 	    zfs_ioc_pool_get_history,
5616 	    zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
5617 
5618 	zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
5619 	    zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5620 
5621 	zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
5622 	    zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5623 	zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
5624 	    zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5625 
5626 	zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
5627 	    zfs_ioc_space_written);
5628 	zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
5629 	    zfs_ioc_objset_recvd_props);
5630 	zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
5631 	    zfs_ioc_next_obj);
5632 	zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
5633 	    zfs_ioc_get_fsacl);
5634 	zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
5635 	    zfs_ioc_objset_stats);
5636 	zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
5637 	    zfs_ioc_objset_zplprops);
5638 	zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
5639 	    zfs_ioc_dataset_list_next);
5640 	zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
5641 	    zfs_ioc_snapshot_list_next);
5642 	zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
5643 	    zfs_ioc_send_progress);
5644 
5645 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
5646 	    zfs_ioc_diff, zfs_secpolicy_diff);
5647 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
5648 	    zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
5649 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
5650 	    zfs_ioc_obj_to_path, zfs_secpolicy_diff);
5651 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
5652 	    zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
5653 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
5654 	    zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
5655 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
5656 	    zfs_ioc_send, zfs_secpolicy_send);
5657 
5658 	zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
5659 	    zfs_secpolicy_none);
5660 	zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
5661 	    zfs_secpolicy_destroy);
5662 	zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
5663 	    zfs_secpolicy_rename);
5664 	zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
5665 	    zfs_secpolicy_recv);
5666 	zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
5667 	    zfs_secpolicy_promote);
5668 	zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
5669 	    zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
5670 	zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
5671 	    zfs_secpolicy_set_fsacl);
5672 
5673 	zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
5674 	    zfs_secpolicy_share, POOL_CHECK_NONE);
5675 	zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
5676 	    zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
5677 	zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
5678 	    zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
5679 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5680 	zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
5681 	    zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
5682 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5683 }
5684 
5685 int
5686 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
5687     zfs_ioc_poolcheck_t check)
5688 {
5689 	spa_t *spa;
5690 	int error;
5691 
5692 	ASSERT(type == POOL_NAME || type == DATASET_NAME);
5693 
5694 	if (check & POOL_CHECK_NONE)
5695 		return (0);
5696 
5697 	error = spa_open(name, &spa, FTAG);
5698 	if (error == 0) {
5699 		if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
5700 			error = SET_ERROR(EAGAIN);
5701 		else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
5702 			error = SET_ERROR(EROFS);
5703 		spa_close(spa, FTAG);
5704 	}
5705 	return (error);
5706 }
5707 
5708 /*
5709  * Find a free minor number.
5710  */
5711 minor_t
5712 zfsdev_minor_alloc(void)
5713 {
5714 	static minor_t last_minor;
5715 	minor_t m;
5716 
5717 	ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5718 
5719 	for (m = last_minor + 1; m != last_minor; m++) {
5720 		if (m > ZFSDEV_MAX_MINOR)
5721 			m = 1;
5722 		if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
5723 			last_minor = m;
5724 			return (m);
5725 		}
5726 	}
5727 
5728 	return (0);
5729 }
5730 
5731 static int
5732 zfs_ctldev_init(dev_t *devp)
5733 {
5734 	minor_t minor;
5735 	zfs_soft_state_t *zs;
5736 
5737 	ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5738 	ASSERT(getminor(*devp) == 0);
5739 
5740 	minor = zfsdev_minor_alloc();
5741 	if (minor == 0)
5742 		return (SET_ERROR(ENXIO));
5743 
5744 	if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
5745 		return (SET_ERROR(EAGAIN));
5746 
5747 	*devp = makedevice(getemajor(*devp), minor);
5748 
5749 	zs = ddi_get_soft_state(zfsdev_state, minor);
5750 	zs->zss_type = ZSST_CTLDEV;
5751 	zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
5752 
5753 	return (0);
5754 }
5755 
5756 static void
5757 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
5758 {
5759 	ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5760 
5761 	zfs_onexit_destroy(zo);
5762 	ddi_soft_state_free(zfsdev_state, minor);
5763 }
5764 
5765 void *
5766 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
5767 {
5768 	zfs_soft_state_t *zp;
5769 
5770 	zp = ddi_get_soft_state(zfsdev_state, minor);
5771 	if (zp == NULL || zp->zss_type != which)
5772 		return (NULL);
5773 
5774 	return (zp->zss_data);
5775 }
5776 
5777 static int
5778 zfsdev_open(dev_t *devp, int flag, int otyp, cred_t *cr)
5779 {
5780 	int error = 0;
5781 
5782 	if (getminor(*devp) != 0)
5783 		return (zvol_open(devp, flag, otyp, cr));
5784 
5785 	/* This is the control device. Allocate a new minor if requested. */
5786 	if (flag & FEXCL) {
5787 		mutex_enter(&zfsdev_state_lock);
5788 		error = zfs_ctldev_init(devp);
5789 		mutex_exit(&zfsdev_state_lock);
5790 	}
5791 
5792 	return (error);
5793 }
5794 
5795 static int
5796 zfsdev_close(dev_t dev, int flag, int otyp, cred_t *cr)
5797 {
5798 	zfs_onexit_t *zo;
5799 	minor_t minor = getminor(dev);
5800 
5801 	if (minor == 0)
5802 		return (0);
5803 
5804 	mutex_enter(&zfsdev_state_lock);
5805 	zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
5806 	if (zo == NULL) {
5807 		mutex_exit(&zfsdev_state_lock);
5808 		return (zvol_close(dev, flag, otyp, cr));
5809 	}
5810 	zfs_ctldev_destroy(zo, minor);
5811 	mutex_exit(&zfsdev_state_lock);
5812 
5813 	return (0);
5814 }
5815 
5816 static int
5817 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
5818 {
5819 	zfs_cmd_t *zc;
5820 	uint_t vecnum;
5821 	int error, rc, len;
5822 	minor_t minor = getminor(dev);
5823 	const zfs_ioc_vec_t *vec;
5824 	char *saved_poolname = NULL;
5825 	nvlist_t *innvl = NULL;
5826 
5827 	if (minor != 0 &&
5828 	    zfsdev_get_soft_state(minor, ZSST_CTLDEV) == NULL)
5829 		return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
5830 
5831 	vecnum = cmd - ZFS_IOC_FIRST;
5832 	ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
5833 
5834 	if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
5835 		return (SET_ERROR(EINVAL));
5836 	vec = &zfs_ioc_vec[vecnum];
5837 
5838 	zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
5839 
5840 	error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
5841 	if (error != 0) {
5842 		error = SET_ERROR(EFAULT);
5843 		goto out;
5844 	}
5845 
5846 	zc->zc_iflags = flag & FKIOCTL;
5847 	if (zc->zc_nvlist_src_size != 0) {
5848 		error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
5849 		    zc->zc_iflags, &innvl);
5850 		if (error != 0)
5851 			goto out;
5852 	}
5853 
5854 	/*
5855 	 * Ensure that all pool/dataset names are valid before we pass down to
5856 	 * the lower layers.
5857 	 */
5858 	zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
5859 	switch (vec->zvec_namecheck) {
5860 	case POOL_NAME:
5861 		if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
5862 			error = SET_ERROR(EINVAL);
5863 		else
5864 			error = pool_status_check(zc->zc_name,
5865 			    vec->zvec_namecheck, vec->zvec_pool_check);
5866 		break;
5867 
5868 	case DATASET_NAME:
5869 		if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
5870 			error = SET_ERROR(EINVAL);
5871 		else
5872 			error = pool_status_check(zc->zc_name,
5873 			    vec->zvec_namecheck, vec->zvec_pool_check);
5874 		break;
5875 
5876 	case NO_NAME:
5877 		break;
5878 	}
5879 
5880 
5881 	if (error == 0 && !(flag & FKIOCTL))
5882 		error = vec->zvec_secpolicy(zc, innvl, cr);
5883 
5884 	if (error != 0)
5885 		goto out;
5886 
5887 	/* legacy ioctls can modify zc_name */
5888 	len = strcspn(zc->zc_name, "/@#") + 1;
5889 	saved_poolname = kmem_alloc(len, KM_SLEEP);
5890 	(void) strlcpy(saved_poolname, zc->zc_name, len);
5891 
5892 	if (vec->zvec_func != NULL) {
5893 		nvlist_t *outnvl;
5894 		int puterror = 0;
5895 		spa_t *spa;
5896 		nvlist_t *lognv = NULL;
5897 
5898 		ASSERT(vec->zvec_legacy_func == NULL);
5899 
5900 		/*
5901 		 * Add the innvl to the lognv before calling the func,
5902 		 * in case the func changes the innvl.
5903 		 */
5904 		if (vec->zvec_allow_log) {
5905 			lognv = fnvlist_alloc();
5906 			fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
5907 			    vec->zvec_name);
5908 			if (!nvlist_empty(innvl)) {
5909 				fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
5910 				    innvl);
5911 			}
5912 		}
5913 
5914 		outnvl = fnvlist_alloc();
5915 		error = vec->zvec_func(zc->zc_name, innvl, outnvl);
5916 
5917 		if (error == 0 && vec->zvec_allow_log &&
5918 		    spa_open(zc->zc_name, &spa, FTAG) == 0) {
5919 			if (!nvlist_empty(outnvl)) {
5920 				fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
5921 				    outnvl);
5922 			}
5923 			(void) spa_history_log_nvl(spa, lognv);
5924 			spa_close(spa, FTAG);
5925 		}
5926 		fnvlist_free(lognv);
5927 
5928 		if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
5929 			int smusherror = 0;
5930 			if (vec->zvec_smush_outnvlist) {
5931 				smusherror = nvlist_smush(outnvl,
5932 				    zc->zc_nvlist_dst_size);
5933 			}
5934 			if (smusherror == 0)
5935 				puterror = put_nvlist(zc, outnvl);
5936 		}
5937 
5938 		if (puterror != 0)
5939 			error = puterror;
5940 
5941 		nvlist_free(outnvl);
5942 	} else {
5943 		error = vec->zvec_legacy_func(zc);
5944 	}
5945 
5946 out:
5947 	nvlist_free(innvl);
5948 	rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
5949 	if (error == 0 && rc != 0)
5950 		error = SET_ERROR(EFAULT);
5951 	if (error == 0 && vec->zvec_allow_log) {
5952 		char *s = tsd_get(zfs_allow_log_key);
5953 		if (s != NULL)
5954 			strfree(s);
5955 		(void) tsd_set(zfs_allow_log_key, saved_poolname);
5956 	} else {
5957 		if (saved_poolname != NULL)
5958 			strfree(saved_poolname);
5959 	}
5960 
5961 	kmem_free(zc, sizeof (zfs_cmd_t));
5962 	return (error);
5963 }
5964 
5965 static int
5966 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
5967 {
5968 	if (cmd != DDI_ATTACH)
5969 		return (DDI_FAILURE);
5970 
5971 	if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
5972 	    DDI_PSEUDO, 0) == DDI_FAILURE)
5973 		return (DDI_FAILURE);
5974 
5975 	zfs_dip = dip;
5976 
5977 	ddi_report_dev(dip);
5978 
5979 	return (DDI_SUCCESS);
5980 }
5981 
5982 static int
5983 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
5984 {
5985 	if (spa_busy() || zfs_busy() || zvol_busy())
5986 		return (DDI_FAILURE);
5987 
5988 	if (cmd != DDI_DETACH)
5989 		return (DDI_FAILURE);
5990 
5991 	zfs_dip = NULL;
5992 
5993 	ddi_prop_remove_all(dip);
5994 	ddi_remove_minor_node(dip, NULL);
5995 
5996 	return (DDI_SUCCESS);
5997 }
5998 
5999 /*ARGSUSED*/
6000 static int
6001 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
6002 {
6003 	switch (infocmd) {
6004 	case DDI_INFO_DEVT2DEVINFO:
6005 		*result = zfs_dip;
6006 		return (DDI_SUCCESS);
6007 
6008 	case DDI_INFO_DEVT2INSTANCE:
6009 		*result = (void *)0;
6010 		return (DDI_SUCCESS);
6011 	}
6012 
6013 	return (DDI_FAILURE);
6014 }
6015 
6016 /*
6017  * OK, so this is a little weird.
6018  *
6019  * /dev/zfs is the control node, i.e. minor 0.
6020  * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
6021  *
6022  * /dev/zfs has basically nothing to do except serve up ioctls,
6023  * so most of the standard driver entry points are in zvol.c.
6024  */
6025 static struct cb_ops zfs_cb_ops = {
6026 	zfsdev_open,	/* open */
6027 	zfsdev_close,	/* close */
6028 	zvol_strategy,	/* strategy */
6029 	nodev,		/* print */
6030 	zvol_dump,	/* dump */
6031 	zvol_read,	/* read */
6032 	zvol_write,	/* write */
6033 	zfsdev_ioctl,	/* ioctl */
6034 	nodev,		/* devmap */
6035 	nodev,		/* mmap */
6036 	nodev,		/* segmap */
6037 	nochpoll,	/* poll */
6038 	ddi_prop_op,	/* prop_op */
6039 	NULL,		/* streamtab */
6040 	D_NEW | D_MP | D_64BIT,		/* Driver compatibility flag */
6041 	CB_REV,		/* version */
6042 	nodev,		/* async read */
6043 	nodev,		/* async write */
6044 };
6045 
6046 static struct dev_ops zfs_dev_ops = {
6047 	DEVO_REV,	/* version */
6048 	0,		/* refcnt */
6049 	zfs_info,	/* info */
6050 	nulldev,	/* identify */
6051 	nulldev,	/* probe */
6052 	zfs_attach,	/* attach */
6053 	zfs_detach,	/* detach */
6054 	nodev,		/* reset */
6055 	&zfs_cb_ops,	/* driver operations */
6056 	NULL,		/* no bus operations */
6057 	NULL,		/* power */
6058 	ddi_quiesce_not_needed,	/* quiesce */
6059 };
6060 
6061 static struct modldrv zfs_modldrv = {
6062 	&mod_driverops,
6063 	"ZFS storage pool",
6064 	&zfs_dev_ops
6065 };
6066 
6067 static struct modlinkage modlinkage = {
6068 	MODREV_1,
6069 	(void *)&zfs_modlfs,
6070 	(void *)&zfs_modldrv,
6071 	NULL
6072 };
6073 
6074 static void
6075 zfs_allow_log_destroy(void *arg)
6076 {
6077 	char *poolname = arg;
6078 	strfree(poolname);
6079 }
6080 
6081 int
6082 _init(void)
6083 {
6084 	int error;
6085 
6086 	spa_init(FREAD | FWRITE);
6087 	zfs_init();
6088 	zvol_init();
6089 	zfs_ioctl_init();
6090 
6091 	if ((error = mod_install(&modlinkage)) != 0) {
6092 		zvol_fini();
6093 		zfs_fini();
6094 		spa_fini();
6095 		return (error);
6096 	}
6097 
6098 	tsd_create(&zfs_fsyncer_key, NULL);
6099 	tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6100 	tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6101 
6102 	error = ldi_ident_from_mod(&modlinkage, &zfs_li);
6103 	ASSERT(error == 0);
6104 	mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6105 
6106 	return (0);
6107 }
6108 
6109 int
6110 _fini(void)
6111 {
6112 	int error;
6113 
6114 	if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
6115 		return (SET_ERROR(EBUSY));
6116 
6117 	if ((error = mod_remove(&modlinkage)) != 0)
6118 		return (error);
6119 
6120 	zvol_fini();
6121 	zfs_fini();
6122 	spa_fini();
6123 	if (zfs_nfsshare_inited)
6124 		(void) ddi_modclose(nfs_mod);
6125 	if (zfs_smbshare_inited)
6126 		(void) ddi_modclose(smbsrv_mod);
6127 	if (zfs_nfsshare_inited || zfs_smbshare_inited)
6128 		(void) ddi_modclose(sharefs_mod);
6129 
6130 	tsd_destroy(&zfs_fsyncer_key);
6131 	ldi_ident_release(zfs_li);
6132 	zfs_li = NULL;
6133 	mutex_destroy(&zfs_share_lock);
6134 
6135 	return (error);
6136 }
6137 
6138 int
6139 _info(struct modinfo *modinfop)
6140 {
6141 	return (mod_info(&modlinkage, modinfop));
6142 }
6143