xref: /titanic_44/usr/src/uts/common/fs/zfs/zfs_ioctl.c (revision 48a4016cae8aa2b8b3d8b258eb22e0c781912bed)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 
22 /*
23  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24  * Portions Copyright 2011 Martin Matuska
25  * Copyright 2015 Nexenta Systems, Inc.  All rights reserved.
26  * Copyright (c) 2014, Joyent, Inc. All rights reserved.
27  * Copyright (c) 2011, 2015 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  */
31 
32 /*
33  * ZFS ioctls.
34  *
35  * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
36  * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
37  *
38  * There are two ways that we handle ioctls: the legacy way where almost
39  * all of the logic is in the ioctl callback, and the new way where most
40  * of the marshalling is handled in the common entry point, zfsdev_ioctl().
41  *
42  * Non-legacy ioctls should be registered by calling
43  * zfs_ioctl_register() from zfs_ioctl_init().  The ioctl is invoked
44  * from userland by lzc_ioctl().
45  *
46  * The registration arguments are as follows:
47  *
48  * const char *name
49  *   The name of the ioctl.  This is used for history logging.  If the
50  *   ioctl returns successfully (the callback returns 0), and allow_log
51  *   is true, then a history log entry will be recorded with the input &
52  *   output nvlists.  The log entry can be printed with "zpool history -i".
53  *
54  * zfs_ioc_t ioc
55  *   The ioctl request number, which userland will pass to ioctl(2).
56  *   The ioctl numbers can change from release to release, because
57  *   the caller (libzfs) must be matched to the kernel.
58  *
59  * zfs_secpolicy_func_t *secpolicy
60  *   This function will be called before the zfs_ioc_func_t, to
61  *   determine if this operation is permitted.  It should return EPERM
62  *   on failure, and 0 on success.  Checks include determining if the
63  *   dataset is visible in this zone, and if the user has either all
64  *   zfs privileges in the zone (SYS_MOUNT), or has been granted permission
65  *   to do this operation on this dataset with "zfs allow".
66  *
67  * zfs_ioc_namecheck_t namecheck
68  *   This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
69  *   name, a dataset name, or nothing.  If the name is not well-formed,
70  *   the ioctl will fail and the callback will not be called.
71  *   Therefore, the callback can assume that the name is well-formed
72  *   (e.g. is null-terminated, doesn't have more than one '@' character,
73  *   doesn't have invalid characters).
74  *
75  * zfs_ioc_poolcheck_t pool_check
76  *   This specifies requirements on the pool state.  If the pool does
77  *   not meet them (is suspended or is readonly), the ioctl will fail
78  *   and the callback will not be called.  If any checks are specified
79  *   (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
80  *   Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
81  *   POOL_CHECK_READONLY).
82  *
83  * boolean_t smush_outnvlist
84  *   If smush_outnvlist is true, then the output is presumed to be a
85  *   list of errors, and it will be "smushed" down to fit into the
86  *   caller's buffer, by removing some entries and replacing them with a
87  *   single "N_MORE_ERRORS" entry indicating how many were removed.  See
88  *   nvlist_smush() for details.  If smush_outnvlist is false, and the
89  *   outnvlist does not fit into the userland-provided buffer, then the
90  *   ioctl will fail with ENOMEM.
91  *
92  * zfs_ioc_func_t *func
93  *   The callback function that will perform the operation.
94  *
95  *   The callback should return 0 on success, or an error number on
96  *   failure.  If the function fails, the userland ioctl will return -1,
97  *   and errno will be set to the callback's return value.  The callback
98  *   will be called with the following arguments:
99  *
100  *   const char *name
101  *     The name of the pool or dataset to operate on, from
102  *     zfs_cmd_t:zc_name.  The 'namecheck' argument specifies the
103  *     expected type (pool, dataset, or none).
104  *
105  *   nvlist_t *innvl
106  *     The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src.  Or
107  *     NULL if no input nvlist was provided.  Changes to this nvlist are
108  *     ignored.  If the input nvlist could not be deserialized, the
109  *     ioctl will fail and the callback will not be called.
110  *
111  *   nvlist_t *outnvl
112  *     The output nvlist, initially empty.  The callback can fill it in,
113  *     and it will be returned to userland by serializing it into
114  *     zfs_cmd_t:zc_nvlist_dst.  If it is non-empty, and serialization
115  *     fails (e.g. because the caller didn't supply a large enough
116  *     buffer), then the overall ioctl will fail.  See the
117  *     'smush_nvlist' argument above for additional behaviors.
118  *
119  *     There are two typical uses of the output nvlist:
120  *       - To return state, e.g. property values.  In this case,
121  *         smush_outnvlist should be false.  If the buffer was not large
122  *         enough, the caller will reallocate a larger buffer and try
123  *         the ioctl again.
124  *
125  *       - To return multiple errors from an ioctl which makes on-disk
126  *         changes.  In this case, smush_outnvlist should be true.
127  *         Ioctls which make on-disk modifications should generally not
128  *         use the outnvl if they succeed, because the caller can not
129  *         distinguish between the operation failing, and
130  *         deserialization failing.
131  */
132 
133 #include <sys/types.h>
134 #include <sys/param.h>
135 #include <sys/errno.h>
136 #include <sys/uio.h>
137 #include <sys/buf.h>
138 #include <sys/modctl.h>
139 #include <sys/open.h>
140 #include <sys/file.h>
141 #include <sys/kmem.h>
142 #include <sys/conf.h>
143 #include <sys/cmn_err.h>
144 #include <sys/stat.h>
145 #include <sys/zfs_ioctl.h>
146 #include <sys/zfs_vfsops.h>
147 #include <sys/zfs_znode.h>
148 #include <sys/zap.h>
149 #include <sys/spa.h>
150 #include <sys/spa_impl.h>
151 #include <sys/vdev.h>
152 #include <sys/priv_impl.h>
153 #include <sys/dmu.h>
154 #include <sys/dsl_dir.h>
155 #include <sys/dsl_dataset.h>
156 #include <sys/dsl_prop.h>
157 #include <sys/dsl_deleg.h>
158 #include <sys/dmu_objset.h>
159 #include <sys/dmu_impl.h>
160 #include <sys/dmu_tx.h>
161 #include <sys/ddi.h>
162 #include <sys/sunddi.h>
163 #include <sys/sunldi.h>
164 #include <sys/policy.h>
165 #include <sys/zone.h>
166 #include <sys/nvpair.h>
167 #include <sys/pathname.h>
168 #include <sys/mount.h>
169 #include <sys/sdt.h>
170 #include <sys/fs/zfs.h>
171 #include <sys/zfs_ctldir.h>
172 #include <sys/zfs_dir.h>
173 #include <sys/zfs_onexit.h>
174 #include <sys/zvol.h>
175 #include <sys/dsl_scan.h>
176 #include <sharefs/share.h>
177 #include <sys/dmu_objset.h>
178 #include <sys/dmu_send.h>
179 #include <sys/dsl_destroy.h>
180 #include <sys/dsl_bookmark.h>
181 #include <sys/dsl_userhold.h>
182 #include <sys/zfeature.h>
183 #include <sys/zfs_events.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[ZFS_MAX_DATASET_NAME_LEN];
629 			/*
630 			 * Unprivileged users are allowed to modify the
631 			 * limit on things *under* (ie. contained by)
632 			 * the thing they own.
633 			 */
634 			if (dsl_prop_get_integer(dsname, "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[ZFS_MAX_DATASET_NAME_LEN];
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[ZFS_MAX_DATASET_NAME_LEN];
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[ZFS_MAX_DATASET_NAME_LEN];
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[ZFS_MAX_DATASET_NAME_LEN];
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[ZFS_MAX_DATASET_NAME_LEN];
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  * Policy for allowing setting the zev callback list.
1287  */
1288 static int
1289 zfs_secpolicy_set_zev_callbacks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1290 {
1291 	/* must be called from kernel context */
1292 	if (!(zc->zc_iflags & FKIOCTL))
1293 		return (SET_ERROR(EPERM));
1294 	/* callback pointer must be unset (set to default value) */
1295 	rw_enter(&rz_zev_rwlock, RW_READER);
1296 	if (rz_zev_callbacks != rz_zev_default_callbacks) {
1297 		rw_exit(&rz_zev_rwlock);
1298 		return (SET_ERROR(EBUSY));
1299 	}
1300 	rw_exit(&rz_zev_rwlock);
1301 	return (0);
1302 }
1303 
1304 /*
1305  * Policy for allowing unsetting/resetting the zev callback list.
1306  */
1307 static int
1308 zfs_secpolicy_unset_zev_callbacks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1309 {
1310 	/* must be called from kernel context */
1311 	if (!(zc->zc_iflags & FKIOCTL))
1312 		return (SET_ERROR(EPERM));
1313 	return (0);
1314 }
1315 
1316 /*
1317  * Returns the nvlist as specified by the user in the zfs_cmd_t.
1318  */
1319 static int
1320 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1321 {
1322 	char *packed;
1323 	int error;
1324 	nvlist_t *list = NULL;
1325 
1326 	/*
1327 	 * Read in and unpack the user-supplied nvlist.
1328 	 */
1329 	if (size == 0)
1330 		return (SET_ERROR(EINVAL));
1331 
1332 	packed = kmem_alloc(size, KM_SLEEP);
1333 
1334 	if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1335 	    iflag)) != 0) {
1336 		kmem_free(packed, size);
1337 		return (error);
1338 	}
1339 
1340 	if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1341 		kmem_free(packed, size);
1342 		return (error);
1343 	}
1344 
1345 	kmem_free(packed, size);
1346 
1347 	*nvp = list;
1348 	return (0);
1349 }
1350 
1351 /*
1352  * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1353  * Entries will be removed from the end of the nvlist, and one int32 entry
1354  * named "N_MORE_ERRORS" will be added indicating how many entries were
1355  * removed.
1356  */
1357 static int
1358 nvlist_smush(nvlist_t *errors, size_t max)
1359 {
1360 	size_t size;
1361 
1362 	size = fnvlist_size(errors);
1363 
1364 	if (size > max) {
1365 		nvpair_t *more_errors;
1366 		int n = 0;
1367 
1368 		if (max < 1024)
1369 			return (SET_ERROR(ENOMEM));
1370 
1371 		fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1372 		more_errors = nvlist_prev_nvpair(errors, NULL);
1373 
1374 		do {
1375 			nvpair_t *pair = nvlist_prev_nvpair(errors,
1376 			    more_errors);
1377 			fnvlist_remove_nvpair(errors, pair);
1378 			n++;
1379 			size = fnvlist_size(errors);
1380 		} while (size > max);
1381 
1382 		fnvlist_remove_nvpair(errors, more_errors);
1383 		fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1384 		ASSERT3U(fnvlist_size(errors), <=, max);
1385 	}
1386 
1387 	return (0);
1388 }
1389 
1390 static int
1391 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1392 {
1393 	char *packed = NULL;
1394 	int error = 0;
1395 	size_t size;
1396 
1397 	size = fnvlist_size(nvl);
1398 
1399 	if (size > zc->zc_nvlist_dst_size) {
1400 		error = SET_ERROR(ENOMEM);
1401 	} else {
1402 		packed = fnvlist_pack(nvl, &size);
1403 		if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1404 		    size, zc->zc_iflags) != 0)
1405 			error = SET_ERROR(EFAULT);
1406 		fnvlist_pack_free(packed, size);
1407 	}
1408 
1409 	zc->zc_nvlist_dst_size = size;
1410 	zc->zc_nvlist_dst_filled = B_TRUE;
1411 	return (error);
1412 }
1413 
1414 static int
1415 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1416 {
1417 	objset_t *os;
1418 	int error;
1419 
1420 	error = dmu_objset_hold(dsname, FTAG, &os);
1421 	if (error != 0)
1422 		return (error);
1423 	if (dmu_objset_type(os) != DMU_OST_ZFS) {
1424 		dmu_objset_rele(os, FTAG);
1425 		return (SET_ERROR(EINVAL));
1426 	}
1427 
1428 	mutex_enter(&os->os_user_ptr_lock);
1429 	*zfvp = dmu_objset_get_user(os);
1430 	if (*zfvp) {
1431 		VFS_HOLD((*zfvp)->z_vfs);
1432 	} else {
1433 		error = SET_ERROR(ESRCH);
1434 	}
1435 	mutex_exit(&os->os_user_ptr_lock);
1436 	dmu_objset_rele(os, FTAG);
1437 	return (error);
1438 }
1439 
1440 /*
1441  * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1442  * case its z_vfs will be NULL, and it will be opened as the owner.
1443  * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1444  * which prevents all vnode ops from running.
1445  */
1446 static int
1447 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1448 {
1449 	int error = 0;
1450 
1451 	if (getzfsvfs(name, zfvp) != 0)
1452 		error = zfsvfs_create(name, zfvp);
1453 	if (error == 0) {
1454 		rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1455 		    RW_READER, tag);
1456 		if ((*zfvp)->z_unmounted) {
1457 			/*
1458 			 * XXX we could probably try again, since the unmounting
1459 			 * thread should be just about to disassociate the
1460 			 * objset from the zfsvfs.
1461 			 */
1462 			rrm_exit(&(*zfvp)->z_teardown_lock, tag);
1463 			return (SET_ERROR(EBUSY));
1464 		}
1465 	}
1466 	return (error);
1467 }
1468 
1469 static void
1470 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1471 {
1472 	rrm_exit(&zfsvfs->z_teardown_lock, tag);
1473 
1474 	if (zfsvfs->z_vfs) {
1475 		VFS_RELE(zfsvfs->z_vfs);
1476 	} else {
1477 		dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1478 		zfsvfs_free(zfsvfs);
1479 	}
1480 }
1481 
1482 static int
1483 zfs_ioc_pool_create(zfs_cmd_t *zc)
1484 {
1485 	int error;
1486 	nvlist_t *config, *props = NULL;
1487 	nvlist_t *rootprops = NULL;
1488 	nvlist_t *zplprops = NULL;
1489 
1490 	if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1491 	    zc->zc_iflags, &config))
1492 		return (error);
1493 
1494 	if (zc->zc_nvlist_src_size != 0 && (error =
1495 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1496 	    zc->zc_iflags, &props))) {
1497 		nvlist_free(config);
1498 		return (error);
1499 	}
1500 
1501 	if (props) {
1502 		nvlist_t *nvl = NULL;
1503 		uint64_t version = SPA_VERSION;
1504 
1505 		(void) nvlist_lookup_uint64(props,
1506 		    zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1507 		if (!SPA_VERSION_IS_SUPPORTED(version)) {
1508 			error = SET_ERROR(EINVAL);
1509 			goto pool_props_bad;
1510 		}
1511 		(void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1512 		if (nvl) {
1513 			error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1514 			if (error != 0) {
1515 				nvlist_free(config);
1516 				nvlist_free(props);
1517 				return (error);
1518 			}
1519 			(void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1520 		}
1521 		VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1522 		error = zfs_fill_zplprops_root(version, rootprops,
1523 		    zplprops, NULL);
1524 		if (error != 0)
1525 			goto pool_props_bad;
1526 	}
1527 
1528 	error = spa_create(zc->zc_name, config, props, zplprops);
1529 
1530 	/*
1531 	 * Set the remaining root properties
1532 	 */
1533 	if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1534 	    ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1535 		(void) spa_destroy(zc->zc_name);
1536 
1537 pool_props_bad:
1538 	nvlist_free(rootprops);
1539 	nvlist_free(zplprops);
1540 	nvlist_free(config);
1541 	nvlist_free(props);
1542 
1543 	return (error);
1544 }
1545 
1546 static int
1547 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1548 {
1549 	int error;
1550 	zfs_log_history(zc);
1551 	error = spa_destroy(zc->zc_name);
1552 	if (error == 0)
1553 		zvol_remove_minors(zc->zc_name);
1554 	return (error);
1555 }
1556 
1557 static int
1558 zfs_ioc_pool_import(zfs_cmd_t *zc)
1559 {
1560 	nvlist_t *config, *props = NULL;
1561 	uint64_t guid;
1562 	int error;
1563 
1564 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1565 	    zc->zc_iflags, &config)) != 0)
1566 		return (error);
1567 
1568 	if (zc->zc_nvlist_src_size != 0 && (error =
1569 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1570 	    zc->zc_iflags, &props))) {
1571 		nvlist_free(config);
1572 		return (error);
1573 	}
1574 
1575 	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1576 	    guid != zc->zc_guid)
1577 		error = SET_ERROR(EINVAL);
1578 	else
1579 		error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1580 
1581 	if (zc->zc_nvlist_dst != 0) {
1582 		int err;
1583 
1584 		if ((err = put_nvlist(zc, config)) != 0)
1585 			error = err;
1586 	}
1587 
1588 	nvlist_free(config);
1589 
1590 	if (props)
1591 		nvlist_free(props);
1592 
1593 	return (error);
1594 }
1595 
1596 static int
1597 zfs_ioc_pool_export(zfs_cmd_t *zc)
1598 {
1599 	int error;
1600 	boolean_t force = (boolean_t)zc->zc_cookie;
1601 	boolean_t hardforce = (boolean_t)zc->zc_guid;
1602 
1603 	zfs_log_history(zc);
1604 	error = spa_export(zc->zc_name, NULL, force, hardforce);
1605 	if (error == 0)
1606 		zvol_remove_minors(zc->zc_name);
1607 	return (error);
1608 }
1609 
1610 static int
1611 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1612 {
1613 	nvlist_t *configs;
1614 	int error;
1615 
1616 	if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1617 		return (SET_ERROR(EEXIST));
1618 
1619 	error = put_nvlist(zc, configs);
1620 
1621 	nvlist_free(configs);
1622 
1623 	return (error);
1624 }
1625 
1626 /*
1627  * inputs:
1628  * zc_name		name of the pool
1629  *
1630  * outputs:
1631  * zc_cookie		real errno
1632  * zc_nvlist_dst	config nvlist
1633  * zc_nvlist_dst_size	size of config nvlist
1634  */
1635 static int
1636 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1637 {
1638 	nvlist_t *config;
1639 	int error;
1640 	int ret = 0;
1641 
1642 	error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1643 	    sizeof (zc->zc_value));
1644 
1645 	if (config != NULL) {
1646 		ret = put_nvlist(zc, config);
1647 		nvlist_free(config);
1648 
1649 		/*
1650 		 * The config may be present even if 'error' is non-zero.
1651 		 * In this case we return success, and preserve the real errno
1652 		 * in 'zc_cookie'.
1653 		 */
1654 		zc->zc_cookie = error;
1655 	} else {
1656 		ret = error;
1657 	}
1658 
1659 	return (ret);
1660 }
1661 
1662 /*
1663  * Try to import the given pool, returning pool stats as appropriate so that
1664  * user land knows which devices are available and overall pool health.
1665  */
1666 static int
1667 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1668 {
1669 	nvlist_t *tryconfig, *config;
1670 	int error;
1671 
1672 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1673 	    zc->zc_iflags, &tryconfig)) != 0)
1674 		return (error);
1675 
1676 	config = spa_tryimport(tryconfig);
1677 
1678 	nvlist_free(tryconfig);
1679 
1680 	if (config == NULL)
1681 		return (SET_ERROR(EINVAL));
1682 
1683 	error = put_nvlist(zc, config);
1684 	nvlist_free(config);
1685 
1686 	return (error);
1687 }
1688 
1689 /*
1690  * inputs:
1691  * zc_name              name of the pool
1692  * zc_cookie            scan func (pool_scan_func_t)
1693  */
1694 static int
1695 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1696 {
1697 	spa_t *spa;
1698 	int error;
1699 
1700 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1701 		return (error);
1702 
1703 	if (zc->zc_cookie == POOL_SCAN_NONE)
1704 		error = spa_scan_stop(spa);
1705 	else
1706 		error = spa_scan(spa, zc->zc_cookie);
1707 
1708 	spa_close(spa, FTAG);
1709 
1710 	return (error);
1711 }
1712 
1713 static int
1714 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1715 {
1716 	spa_t *spa;
1717 	int error;
1718 
1719 	error = spa_open(zc->zc_name, &spa, FTAG);
1720 	if (error == 0) {
1721 		spa_freeze(spa);
1722 		spa_close(spa, FTAG);
1723 	}
1724 	return (error);
1725 }
1726 
1727 static int
1728 zfs_ioc_arc_info(zfs_cmd_t *zc)
1729 {
1730 	int ret;
1731 	void *buf;
1732 	size_t sz = zc->zc_nvlist_dst_size;
1733 	size_t returned_bytes;
1734 
1735 	if (zc->zc_nvlist_dst == 0)
1736 		return (SET_ERROR(EINVAL));
1737 
1738 	buf = kmem_alloc(sz, KM_NOSLEEP);
1739 	if (buf == NULL)
1740 		return (SET_ERROR(ENOMEM));
1741 
1742 	ret = arc_dump(zc->zc_obj, buf, sz, &returned_bytes);
1743 	if (ret != 0) {
1744 		kmem_free(buf, sz);
1745 		return (SET_ERROR(ret));
1746 	}
1747 
1748 	zc->zc_nvlist_dst_filled = 1;
1749 	ret = ddi_copyout(buf, (void *)(uintptr_t)zc->zc_nvlist_dst,
1750 	    returned_bytes, zc->zc_iflags);
1751 	kmem_free(buf, sz);
1752 	if (ret != 0)
1753 		ret = SET_ERROR(EFAULT);
1754 
1755 	return (ret);
1756 }
1757 
1758 static int
1759 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1760 {
1761 	spa_t *spa;
1762 	int error;
1763 
1764 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1765 		return (error);
1766 
1767 	if (zc->zc_cookie < spa_version(spa) ||
1768 	    !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1769 		spa_close(spa, FTAG);
1770 		return (SET_ERROR(EINVAL));
1771 	}
1772 
1773 	spa_upgrade(spa, zc->zc_cookie);
1774 	spa_close(spa, FTAG);
1775 
1776 	return (error);
1777 }
1778 
1779 static int
1780 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1781 {
1782 	spa_t *spa;
1783 	char *hist_buf;
1784 	uint64_t size;
1785 	int error;
1786 
1787 	if ((size = zc->zc_history_len) == 0)
1788 		return (SET_ERROR(EINVAL));
1789 
1790 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1791 		return (error);
1792 
1793 	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1794 		spa_close(spa, FTAG);
1795 		return (SET_ERROR(ENOTSUP));
1796 	}
1797 
1798 	hist_buf = kmem_alloc(size, KM_SLEEP);
1799 	if ((error = spa_history_get(spa, &zc->zc_history_offset,
1800 	    &zc->zc_history_len, hist_buf)) == 0) {
1801 		error = ddi_copyout(hist_buf,
1802 		    (void *)(uintptr_t)zc->zc_history,
1803 		    zc->zc_history_len, zc->zc_iflags);
1804 	}
1805 
1806 	spa_close(spa, FTAG);
1807 	kmem_free(hist_buf, size);
1808 	return (error);
1809 }
1810 
1811 static int
1812 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1813 {
1814 	spa_t *spa;
1815 	int error;
1816 
1817 	error = spa_open(zc->zc_name, &spa, FTAG);
1818 	if (error == 0) {
1819 		error = spa_change_guid(spa);
1820 		spa_close(spa, FTAG);
1821 	}
1822 	return (error);
1823 }
1824 
1825 static int
1826 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1827 {
1828 	return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1829 }
1830 
1831 /*
1832  * inputs:
1833  * zc_name		name of filesystem
1834  * zc_obj		object to find
1835  *
1836  * outputs:
1837  * zc_value		name of object
1838  */
1839 static int
1840 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1841 {
1842 	objset_t *os;
1843 	int error;
1844 
1845 	/* XXX reading from objset not owned */
1846 	if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1847 		return (error);
1848 	if (dmu_objset_type(os) != DMU_OST_ZFS) {
1849 		dmu_objset_rele(os, FTAG);
1850 		return (SET_ERROR(EINVAL));
1851 	}
1852 	error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1853 	    sizeof (zc->zc_value));
1854 	dmu_objset_rele(os, FTAG);
1855 
1856 	return (error);
1857 }
1858 
1859 /*
1860  * inputs:
1861  * zc_name		name of filesystem
1862  * zc_obj		object to find
1863  *
1864  * outputs:
1865  * zc_stat		stats on object
1866  * zc_value		path to object
1867  */
1868 static int
1869 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1870 {
1871 	objset_t *os;
1872 	int error;
1873 
1874 	/* XXX reading from objset not owned */
1875 	if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1876 		return (error);
1877 	if (dmu_objset_type(os) != DMU_OST_ZFS) {
1878 		dmu_objset_rele(os, FTAG);
1879 		return (SET_ERROR(EINVAL));
1880 	}
1881 	error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1882 	    sizeof (zc->zc_value));
1883 	dmu_objset_rele(os, FTAG);
1884 
1885 	return (error);
1886 }
1887 
1888 static int
1889 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1890 {
1891 	spa_t *spa;
1892 	int error;
1893 	nvlist_t *config, **l2cache, **spares;
1894 	uint_t nl2cache = 0, nspares = 0;
1895 
1896 	error = spa_open(zc->zc_name, &spa, FTAG);
1897 	if (error != 0)
1898 		return (error);
1899 
1900 	error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1901 	    zc->zc_iflags, &config);
1902 	(void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1903 	    &l2cache, &nl2cache);
1904 
1905 	(void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1906 	    &spares, &nspares);
1907 
1908 	/*
1909 	 * A root pool with concatenated devices is not supported.
1910 	 * Thus, can not add a device to a root pool.
1911 	 *
1912 	 * Intent log device can not be added to a rootpool because
1913 	 * during mountroot, zil is replayed, a seperated log device
1914 	 * can not be accessed during the mountroot time.
1915 	 *
1916 	 * l2cache and spare devices are ok to be added to a rootpool.
1917 	 */
1918 	if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1919 		nvlist_free(config);
1920 		spa_close(spa, FTAG);
1921 		return (SET_ERROR(EDOM));
1922 	}
1923 
1924 	if (error == 0) {
1925 		error = spa_vdev_add(spa, config);
1926 		nvlist_free(config);
1927 	}
1928 	spa_close(spa, FTAG);
1929 	return (error);
1930 }
1931 
1932 /*
1933  * inputs:
1934  * zc_name		name of the pool
1935  * zc_nvlist_conf	nvlist of devices to remove
1936  * zc_cookie		to stop the remove?
1937  */
1938 static int
1939 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1940 {
1941 	spa_t *spa;
1942 	int error;
1943 
1944 	error = spa_open(zc->zc_name, &spa, FTAG);
1945 	if (error != 0)
1946 		return (error);
1947 	error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1948 	spa_close(spa, FTAG);
1949 	return (error);
1950 }
1951 
1952 static int
1953 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1954 {
1955 	spa_t *spa;
1956 	int error;
1957 	vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1958 
1959 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1960 		return (error);
1961 	switch (zc->zc_cookie) {
1962 	case VDEV_STATE_ONLINE:
1963 		error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1964 		break;
1965 
1966 	case VDEV_STATE_OFFLINE:
1967 		error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1968 		break;
1969 
1970 	case VDEV_STATE_FAULTED:
1971 		if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1972 		    zc->zc_obj != VDEV_AUX_EXTERNAL)
1973 			zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1974 
1975 		error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1976 		break;
1977 
1978 	case VDEV_STATE_DEGRADED:
1979 		if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1980 		    zc->zc_obj != VDEV_AUX_EXTERNAL)
1981 			zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1982 
1983 		error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1984 		break;
1985 
1986 	default:
1987 		error = SET_ERROR(EINVAL);
1988 	}
1989 	zc->zc_cookie = newstate;
1990 	spa_close(spa, FTAG);
1991 	return (error);
1992 }
1993 
1994 static int
1995 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1996 {
1997 	spa_t *spa;
1998 	int replacing = zc->zc_cookie;
1999 	nvlist_t *config;
2000 	int error;
2001 
2002 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2003 		return (error);
2004 
2005 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
2006 	    zc->zc_iflags, &config)) == 0) {
2007 		error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
2008 		nvlist_free(config);
2009 	}
2010 
2011 	spa_close(spa, FTAG);
2012 	return (error);
2013 }
2014 
2015 static int
2016 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
2017 {
2018 	spa_t *spa;
2019 	int error;
2020 
2021 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2022 		return (error);
2023 
2024 	error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
2025 
2026 	spa_close(spa, FTAG);
2027 	return (error);
2028 }
2029 
2030 static int
2031 zfs_ioc_vdev_split(zfs_cmd_t *zc)
2032 {
2033 	spa_t *spa;
2034 	nvlist_t *config, *props = NULL;
2035 	int error;
2036 	boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
2037 
2038 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2039 		return (error);
2040 
2041 	if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
2042 	    zc->zc_iflags, &config)) {
2043 		spa_close(spa, FTAG);
2044 		return (error);
2045 	}
2046 
2047 	if (zc->zc_nvlist_src_size != 0 && (error =
2048 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2049 	    zc->zc_iflags, &props))) {
2050 		spa_close(spa, FTAG);
2051 		nvlist_free(config);
2052 		return (error);
2053 	}
2054 
2055 	error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
2056 
2057 	spa_close(spa, FTAG);
2058 
2059 	nvlist_free(config);
2060 	nvlist_free(props);
2061 
2062 	return (error);
2063 }
2064 
2065 static int
2066 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
2067 {
2068 	spa_t *spa;
2069 	char *path = zc->zc_value;
2070 	uint64_t guid = zc->zc_guid;
2071 	int error;
2072 
2073 	error = spa_open(zc->zc_name, &spa, FTAG);
2074 	if (error != 0)
2075 		return (error);
2076 
2077 	error = spa_vdev_setpath(spa, guid, path);
2078 	spa_close(spa, FTAG);
2079 	return (error);
2080 }
2081 
2082 static int
2083 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
2084 {
2085 	spa_t *spa;
2086 	char *fru = zc->zc_value;
2087 	uint64_t guid = zc->zc_guid;
2088 	int error;
2089 
2090 	error = spa_open(zc->zc_name, &spa, FTAG);
2091 	if (error != 0)
2092 		return (error);
2093 
2094 	error = spa_vdev_setfru(spa, guid, fru);
2095 	spa_close(spa, FTAG);
2096 	return (error);
2097 }
2098 
2099 static int
2100 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
2101 {
2102 	int error = 0;
2103 	nvlist_t *nv;
2104 
2105 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2106 
2107 	if (zc->zc_nvlist_dst != 0 &&
2108 	    (error = dsl_prop_get_all(os, &nv)) == 0) {
2109 		dmu_objset_stats(os, nv);
2110 		/*
2111 		 * NB: zvol_get_stats() will read the objset contents,
2112 		 * which we aren't supposed to do with a
2113 		 * DS_MODE_USER hold, because it could be
2114 		 * inconsistent.  So this is a bit of a workaround...
2115 		 * XXX reading with out owning
2116 		 */
2117 		if (!zc->zc_objset_stats.dds_inconsistent &&
2118 		    dmu_objset_type(os) == DMU_OST_ZVOL) {
2119 			error = zvol_get_stats(os, nv);
2120 			if (error == EIO)
2121 				return (error);
2122 			VERIFY0(error);
2123 		}
2124 		error = put_nvlist(zc, nv);
2125 		nvlist_free(nv);
2126 	}
2127 
2128 	return (error);
2129 }
2130 
2131 /*
2132  * inputs:
2133  * zc_name		name of filesystem
2134  * zc_nvlist_dst_size	size of buffer for property nvlist
2135  *
2136  * outputs:
2137  * zc_objset_stats	stats
2138  * zc_nvlist_dst	property nvlist
2139  * zc_nvlist_dst_size	size of property nvlist
2140  */
2141 static int
2142 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2143 {
2144 	objset_t *os;
2145 	int error;
2146 
2147 	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2148 	if (error == 0) {
2149 		error = zfs_ioc_objset_stats_impl(zc, os);
2150 		dmu_objset_rele(os, FTAG);
2151 	}
2152 
2153 	return (error);
2154 }
2155 
2156 /*
2157  * inputs:
2158  * zc_name		name of filesystem
2159  * zc_nvlist_dst_size	size of buffer for property nvlist
2160  *
2161  * outputs:
2162  * zc_nvlist_dst	received property nvlist
2163  * zc_nvlist_dst_size	size of received property nvlist
2164  *
2165  * Gets received properties (distinct from local properties on or after
2166  * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2167  * local property values.
2168  */
2169 static int
2170 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2171 {
2172 	int error = 0;
2173 	nvlist_t *nv;
2174 
2175 	/*
2176 	 * Without this check, we would return local property values if the
2177 	 * caller has not already received properties on or after
2178 	 * SPA_VERSION_RECVD_PROPS.
2179 	 */
2180 	if (!dsl_prop_get_hasrecvd(zc->zc_name))
2181 		return (SET_ERROR(ENOTSUP));
2182 
2183 	if (zc->zc_nvlist_dst != 0 &&
2184 	    (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2185 		error = put_nvlist(zc, nv);
2186 		nvlist_free(nv);
2187 	}
2188 
2189 	return (error);
2190 }
2191 
2192 static int
2193 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2194 {
2195 	uint64_t value;
2196 	int error;
2197 
2198 	/*
2199 	 * zfs_get_zplprop() will either find a value or give us
2200 	 * the default value (if there is one).
2201 	 */
2202 	if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2203 		return (error);
2204 	VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2205 	return (0);
2206 }
2207 
2208 /*
2209  * inputs:
2210  * zc_name		name of filesystem
2211  * zc_nvlist_dst_size	size of buffer for zpl property nvlist
2212  *
2213  * outputs:
2214  * zc_nvlist_dst	zpl property nvlist
2215  * zc_nvlist_dst_size	size of zpl property nvlist
2216  */
2217 static int
2218 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2219 {
2220 	objset_t *os;
2221 	int err;
2222 
2223 	/* XXX reading without owning */
2224 	if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2225 		return (err);
2226 
2227 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2228 
2229 	/*
2230 	 * NB: nvl_add_zplprop() will read the objset contents,
2231 	 * which we aren't supposed to do with a DS_MODE_USER
2232 	 * hold, because it could be inconsistent.
2233 	 */
2234 	if (zc->zc_nvlist_dst != NULL &&
2235 	    !zc->zc_objset_stats.dds_inconsistent &&
2236 	    dmu_objset_type(os) == DMU_OST_ZFS) {
2237 		nvlist_t *nv;
2238 
2239 		VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2240 		if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2241 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2242 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2243 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2244 			err = put_nvlist(zc, nv);
2245 		nvlist_free(nv);
2246 	} else {
2247 		err = SET_ERROR(ENOENT);
2248 	}
2249 	dmu_objset_rele(os, FTAG);
2250 	return (err);
2251 }
2252 
2253 static boolean_t
2254 dataset_name_hidden(const char *name)
2255 {
2256 	/*
2257 	 * Skip over datasets that are not visible in this zone,
2258 	 * internal datasets (which have a $ in their name), and
2259 	 * temporary datasets (which have a % in their name).
2260 	 */
2261 	if (strchr(name, '$') != NULL)
2262 		return (B_TRUE);
2263 	if (strchr(name, '%') != NULL)
2264 		return (B_TRUE);
2265 	if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
2266 		return (B_TRUE);
2267 	return (B_FALSE);
2268 }
2269 
2270 /*
2271  * inputs:
2272  * zc_name		name of filesystem
2273  * zc_cookie		zap cursor
2274  * zc_nvlist_dst_size	size of buffer for property nvlist
2275  *
2276  * outputs:
2277  * zc_name		name of next filesystem
2278  * zc_cookie		zap cursor
2279  * zc_objset_stats	stats
2280  * zc_nvlist_dst	property nvlist
2281  * zc_nvlist_dst_size	size of property nvlist
2282  */
2283 static int
2284 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2285 {
2286 	objset_t *os;
2287 	int error;
2288 	char *p;
2289 	size_t orig_len = strlen(zc->zc_name);
2290 
2291 top:
2292 	if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2293 		if (error == ENOENT)
2294 			error = SET_ERROR(ESRCH);
2295 		return (error);
2296 	}
2297 
2298 	p = strrchr(zc->zc_name, '/');
2299 	if (p == NULL || p[1] != '\0')
2300 		(void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2301 	p = zc->zc_name + strlen(zc->zc_name);
2302 
2303 	do {
2304 		error = dmu_dir_list_next(os,
2305 		    sizeof (zc->zc_name) - (p - zc->zc_name), p,
2306 		    NULL, &zc->zc_cookie);
2307 		if (error == ENOENT)
2308 			error = SET_ERROR(ESRCH);
2309 	} while (error == 0 && dataset_name_hidden(zc->zc_name));
2310 	dmu_objset_rele(os, FTAG);
2311 
2312 	/*
2313 	 * If it's an internal dataset (ie. with a '$' in its name),
2314 	 * don't try to get stats for it, otherwise we'll return ENOENT.
2315 	 */
2316 	if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2317 		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2318 		if (error == ENOENT) {
2319 			/* We lost a race with destroy, get the next one. */
2320 			zc->zc_name[orig_len] = '\0';
2321 			goto top;
2322 		}
2323 	}
2324 	return (error);
2325 }
2326 
2327 /*
2328  * inputs:
2329  * zc_name		name of filesystem
2330  * zc_cookie		zap cursor
2331  * zc_nvlist_dst_size	size of buffer for property nvlist
2332  *
2333  * outputs:
2334  * zc_name		name of next snapshot
2335  * zc_objset_stats	stats
2336  * zc_nvlist_dst	property nvlist
2337  * zc_nvlist_dst_size	size of property nvlist
2338  */
2339 static int
2340 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2341 {
2342 	objset_t *os;
2343 	int error;
2344 
2345 	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2346 	if (error != 0) {
2347 		return (error == ENOENT ? ESRCH : error);
2348 	}
2349 
2350 	/*
2351 	 * A dataset name of maximum length cannot have any snapshots,
2352 	 * so exit immediately.
2353 	 */
2354 	if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >=
2355 	    ZFS_MAX_DATASET_NAME_LEN) {
2356 		dmu_objset_rele(os, FTAG);
2357 		return (SET_ERROR(ESRCH));
2358 	}
2359 
2360 	error = dmu_snapshot_list_next(os,
2361 	    sizeof (zc->zc_name) - strlen(zc->zc_name),
2362 	    zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2363 	    NULL);
2364 
2365 	if (error == 0) {
2366 		dsl_dataset_t *ds;
2367 		dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2368 
2369 		error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2370 		if (error == 0) {
2371 			objset_t *ossnap;
2372 
2373 			error = dmu_objset_from_ds(ds, &ossnap);
2374 			if (error == 0)
2375 				error = zfs_ioc_objset_stats_impl(zc, ossnap);
2376 			dsl_dataset_rele(ds, FTAG);
2377 		}
2378 	} else if (error == ENOENT) {
2379 		error = SET_ERROR(ESRCH);
2380 	}
2381 
2382 	dmu_objset_rele(os, FTAG);
2383 	/* if we failed, undo the @ that we tacked on to zc_name */
2384 	if (error != 0)
2385 		*strchr(zc->zc_name, '@') = '\0';
2386 	return (error);
2387 }
2388 
2389 static int
2390 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2391 {
2392 	const char *propname = nvpair_name(pair);
2393 	uint64_t *valary;
2394 	unsigned int vallen;
2395 	const char *domain;
2396 	char *dash;
2397 	zfs_userquota_prop_t type;
2398 	uint64_t rid;
2399 	uint64_t quota;
2400 	zfsvfs_t *zfsvfs;
2401 	int err;
2402 
2403 	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2404 		nvlist_t *attrs;
2405 		VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2406 		if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2407 		    &pair) != 0)
2408 			return (SET_ERROR(EINVAL));
2409 	}
2410 
2411 	/*
2412 	 * A correctly constructed propname is encoded as
2413 	 * userquota@<rid>-<domain>.
2414 	 */
2415 	if ((dash = strchr(propname, '-')) == NULL ||
2416 	    nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2417 	    vallen != 3)
2418 		return (SET_ERROR(EINVAL));
2419 
2420 	domain = dash + 1;
2421 	type = valary[0];
2422 	rid = valary[1];
2423 	quota = valary[2];
2424 
2425 	err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2426 	if (err == 0) {
2427 		err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2428 		zfsvfs_rele(zfsvfs, FTAG);
2429 	}
2430 
2431 	return (err);
2432 }
2433 
2434 /*
2435  * If the named property is one that has a special function to set its value,
2436  * return 0 on success and a positive error code on failure; otherwise if it is
2437  * not one of the special properties handled by this function, return -1.
2438  *
2439  * XXX: It would be better for callers of the property interface if we handled
2440  * these special cases in dsl_prop.c (in the dsl layer).
2441  */
2442 static int
2443 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2444     nvpair_t *pair)
2445 {
2446 	const char *propname = nvpair_name(pair);
2447 	zfs_prop_t prop = zfs_name_to_prop(propname);
2448 	uint64_t intval;
2449 	int err = -1;
2450 
2451 	if (prop == ZPROP_INVAL) {
2452 		if (zfs_prop_userquota(propname))
2453 			return (zfs_prop_set_userquota(dsname, pair));
2454 		return (-1);
2455 	}
2456 
2457 	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2458 		nvlist_t *attrs;
2459 		VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2460 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2461 		    &pair) == 0);
2462 	}
2463 
2464 	if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
2465 		return (-1);
2466 
2467 	VERIFY(0 == nvpair_value_uint64(pair, &intval));
2468 
2469 	switch (prop) {
2470 	case ZFS_PROP_QUOTA:
2471 		err = dsl_dir_set_quota(dsname, source, intval);
2472 		break;
2473 	case ZFS_PROP_REFQUOTA:
2474 		err = dsl_dataset_set_refquota(dsname, source, intval);
2475 		break;
2476 	case ZFS_PROP_FILESYSTEM_LIMIT:
2477 	case ZFS_PROP_SNAPSHOT_LIMIT:
2478 		if (intval == UINT64_MAX) {
2479 			/* clearing the limit, just do it */
2480 			err = 0;
2481 		} else {
2482 			err = dsl_dir_activate_fs_ss_limit(dsname);
2483 		}
2484 		/*
2485 		 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2486 		 * default path to set the value in the nvlist.
2487 		 */
2488 		if (err == 0)
2489 			err = -1;
2490 		break;
2491 	case ZFS_PROP_RESERVATION:
2492 		err = dsl_dir_set_reservation(dsname, source, intval);
2493 		break;
2494 	case ZFS_PROP_REFRESERVATION:
2495 		err = dsl_dataset_set_refreservation(dsname, source, intval);
2496 		break;
2497 	case ZFS_PROP_VOLSIZE:
2498 		err = zvol_set_volsize(dsname, intval);
2499 		break;
2500 	case ZFS_PROP_VERSION:
2501 	{
2502 		zfsvfs_t *zfsvfs;
2503 
2504 		if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2505 			break;
2506 
2507 		err = zfs_set_version(zfsvfs, intval);
2508 		zfsvfs_rele(zfsvfs, FTAG);
2509 
2510 		if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2511 			zfs_cmd_t *zc;
2512 
2513 			zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2514 			(void) strcpy(zc->zc_name, dsname);
2515 			(void) zfs_ioc_userspace_upgrade(zc);
2516 			kmem_free(zc, sizeof (zfs_cmd_t));
2517 		}
2518 		break;
2519 	}
2520 	default:
2521 		err = -1;
2522 	}
2523 
2524 	return (err);
2525 }
2526 
2527 /*
2528  * This function is best effort. If it fails to set any of the given properties,
2529  * it continues to set as many as it can and returns the last error
2530  * encountered. If the caller provides a non-NULL errlist, it will be filled in
2531  * with the list of names of all the properties that failed along with the
2532  * corresponding error numbers.
2533  *
2534  * If every property is set successfully, zero is returned and errlist is not
2535  * modified.
2536  */
2537 int
2538 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2539     nvlist_t *errlist)
2540 {
2541 	nvpair_t *pair;
2542 	nvpair_t *propval;
2543 	int rv = 0;
2544 	uint64_t intval;
2545 	char *strval;
2546 	nvlist_t *genericnvl = fnvlist_alloc();
2547 	nvlist_t *retrynvl = fnvlist_alloc();
2548 
2549 retry:
2550 	pair = NULL;
2551 	while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2552 		const char *propname = nvpair_name(pair);
2553 		zfs_prop_t prop = zfs_name_to_prop(propname);
2554 		int err = 0;
2555 
2556 		/* decode the property value */
2557 		propval = pair;
2558 		if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2559 			nvlist_t *attrs;
2560 			attrs = fnvpair_value_nvlist(pair);
2561 			if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2562 			    &propval) != 0)
2563 				err = SET_ERROR(EINVAL);
2564 		}
2565 
2566 		/* Validate value type */
2567 		if (err == 0 && prop == ZPROP_INVAL) {
2568 			if (zfs_prop_user(propname)) {
2569 				if (nvpair_type(propval) != DATA_TYPE_STRING)
2570 					err = SET_ERROR(EINVAL);
2571 			} else if (zfs_prop_userquota(propname)) {
2572 				if (nvpair_type(propval) !=
2573 				    DATA_TYPE_UINT64_ARRAY)
2574 					err = SET_ERROR(EINVAL);
2575 			} else {
2576 				err = SET_ERROR(EINVAL);
2577 			}
2578 		} else if (err == 0) {
2579 			if (nvpair_type(propval) == DATA_TYPE_STRING) {
2580 				if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2581 					err = SET_ERROR(EINVAL);
2582 			} else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2583 				const char *unused;
2584 
2585 				intval = fnvpair_value_uint64(propval);
2586 
2587 				switch (zfs_prop_get_type(prop)) {
2588 				case PROP_TYPE_NUMBER:
2589 					break;
2590 				case PROP_TYPE_STRING:
2591 					err = SET_ERROR(EINVAL);
2592 					break;
2593 				case PROP_TYPE_INDEX:
2594 					if (zfs_prop_index_to_string(prop,
2595 					    intval, &unused) != 0)
2596 						err = SET_ERROR(EINVAL);
2597 					break;
2598 				default:
2599 					cmn_err(CE_PANIC,
2600 					    "unknown property type");
2601 				}
2602 			} else {
2603 				err = SET_ERROR(EINVAL);
2604 			}
2605 		}
2606 
2607 		/* Validate permissions */
2608 		if (err == 0)
2609 			err = zfs_check_settable(dsname, pair, CRED());
2610 
2611 		if (err == 0) {
2612 			err = zfs_prop_set_special(dsname, source, pair);
2613 			if (err == -1) {
2614 				/*
2615 				 * For better performance we build up a list of
2616 				 * properties to set in a single transaction.
2617 				 */
2618 				err = nvlist_add_nvpair(genericnvl, pair);
2619 			} else if (err != 0 && nvl != retrynvl) {
2620 				/*
2621 				 * This may be a spurious error caused by
2622 				 * receiving quota and reservation out of order.
2623 				 * Try again in a second pass.
2624 				 */
2625 				err = nvlist_add_nvpair(retrynvl, pair);
2626 			}
2627 		}
2628 
2629 		if (err != 0) {
2630 			if (errlist != NULL)
2631 				fnvlist_add_int32(errlist, propname, err);
2632 			rv = err;
2633 		}
2634 	}
2635 
2636 	if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2637 		nvl = retrynvl;
2638 		goto retry;
2639 	}
2640 
2641 	if (!nvlist_empty(genericnvl) &&
2642 	    dsl_props_set(dsname, source, genericnvl) != 0) {
2643 		/*
2644 		 * If this fails, we still want to set as many properties as we
2645 		 * can, so try setting them individually.
2646 		 */
2647 		pair = NULL;
2648 		while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2649 			const char *propname = nvpair_name(pair);
2650 			int err = 0;
2651 
2652 			propval = pair;
2653 			if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2654 				nvlist_t *attrs;
2655 				attrs = fnvpair_value_nvlist(pair);
2656 				propval = fnvlist_lookup_nvpair(attrs,
2657 				    ZPROP_VALUE);
2658 			}
2659 
2660 			if (nvpair_type(propval) == DATA_TYPE_STRING) {
2661 				strval = fnvpair_value_string(propval);
2662 				err = dsl_prop_set_string(dsname, propname,
2663 				    source, strval);
2664 			} else {
2665 				intval = fnvpair_value_uint64(propval);
2666 				err = dsl_prop_set_int(dsname, propname, source,
2667 				    intval);
2668 			}
2669 
2670 			if (err != 0) {
2671 				if (errlist != NULL) {
2672 					fnvlist_add_int32(errlist, propname,
2673 					    err);
2674 				}
2675 				rv = err;
2676 			}
2677 		}
2678 	}
2679 	nvlist_free(genericnvl);
2680 	nvlist_free(retrynvl);
2681 
2682 	return (rv);
2683 }
2684 
2685 /*
2686  * Check that all the properties are valid user properties.
2687  */
2688 static int
2689 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2690 {
2691 	nvpair_t *pair = NULL;
2692 	int error = 0;
2693 
2694 	while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2695 		const char *propname = nvpair_name(pair);
2696 
2697 		if (!zfs_prop_user(propname) ||
2698 		    nvpair_type(pair) != DATA_TYPE_STRING)
2699 			return (SET_ERROR(EINVAL));
2700 
2701 		if (error = zfs_secpolicy_write_perms(fsname,
2702 		    ZFS_DELEG_PERM_USERPROP, CRED()))
2703 			return (error);
2704 
2705 		if (strlen(propname) >= ZAP_MAXNAMELEN)
2706 			return (SET_ERROR(ENAMETOOLONG));
2707 
2708 		if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
2709 			return (E2BIG);
2710 	}
2711 	return (0);
2712 }
2713 
2714 static void
2715 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2716 {
2717 	nvpair_t *pair;
2718 
2719 	VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2720 
2721 	pair = NULL;
2722 	while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2723 		if (nvlist_exists(skipped, nvpair_name(pair)))
2724 			continue;
2725 
2726 		VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2727 	}
2728 }
2729 
2730 static int
2731 clear_received_props(const char *dsname, nvlist_t *props,
2732     nvlist_t *skipped)
2733 {
2734 	int err = 0;
2735 	nvlist_t *cleared_props = NULL;
2736 	props_skip(props, skipped, &cleared_props);
2737 	if (!nvlist_empty(cleared_props)) {
2738 		/*
2739 		 * Acts on local properties until the dataset has received
2740 		 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2741 		 */
2742 		zprop_source_t flags = (ZPROP_SRC_NONE |
2743 		    (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2744 		err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2745 	}
2746 	nvlist_free(cleared_props);
2747 	return (err);
2748 }
2749 
2750 /*
2751  * inputs:
2752  * zc_name		name of filesystem
2753  * zc_value		name of property to set
2754  * zc_nvlist_src{_size}	nvlist of properties to apply
2755  * zc_cookie		received properties flag
2756  *
2757  * outputs:
2758  * zc_nvlist_dst{_size} error for each unapplied received property
2759  */
2760 static int
2761 zfs_ioc_set_prop(zfs_cmd_t *zc)
2762 {
2763 	nvlist_t *nvl;
2764 	boolean_t received = zc->zc_cookie;
2765 	zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2766 	    ZPROP_SRC_LOCAL);
2767 	nvlist_t *errors;
2768 	int error;
2769 
2770 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2771 	    zc->zc_iflags, &nvl)) != 0)
2772 		return (error);
2773 
2774 	if (received) {
2775 		nvlist_t *origprops;
2776 
2777 		if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2778 			(void) clear_received_props(zc->zc_name,
2779 			    origprops, nvl);
2780 			nvlist_free(origprops);
2781 		}
2782 
2783 		error = dsl_prop_set_hasrecvd(zc->zc_name);
2784 	}
2785 
2786 	errors = fnvlist_alloc();
2787 	if (error == 0)
2788 		error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2789 
2790 	if (zc->zc_nvlist_dst != NULL && errors != NULL) {
2791 		(void) put_nvlist(zc, errors);
2792 	}
2793 
2794 	nvlist_free(errors);
2795 	nvlist_free(nvl);
2796 	return (error);
2797 }
2798 
2799 /*
2800  * inputs:
2801  * zc_name		name of filesystem
2802  * zc_value		name of property to inherit
2803  * zc_cookie		revert to received value if TRUE
2804  *
2805  * outputs:		none
2806  */
2807 static int
2808 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2809 {
2810 	const char *propname = zc->zc_value;
2811 	zfs_prop_t prop = zfs_name_to_prop(propname);
2812 	boolean_t received = zc->zc_cookie;
2813 	zprop_source_t source = (received
2814 	    ? ZPROP_SRC_NONE		/* revert to received value, if any */
2815 	    : ZPROP_SRC_INHERITED);	/* explicitly inherit */
2816 
2817 	if (received) {
2818 		nvlist_t *dummy;
2819 		nvpair_t *pair;
2820 		zprop_type_t type;
2821 		int err;
2822 
2823 		/*
2824 		 * zfs_prop_set_special() expects properties in the form of an
2825 		 * nvpair with type info.
2826 		 */
2827 		if (prop == ZPROP_INVAL) {
2828 			if (!zfs_prop_user(propname))
2829 				return (SET_ERROR(EINVAL));
2830 
2831 			type = PROP_TYPE_STRING;
2832 		} else if (prop == ZFS_PROP_VOLSIZE ||
2833 		    prop == ZFS_PROP_VERSION) {
2834 			return (SET_ERROR(EINVAL));
2835 		} else {
2836 			type = zfs_prop_get_type(prop);
2837 		}
2838 
2839 		VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2840 
2841 		switch (type) {
2842 		case PROP_TYPE_STRING:
2843 			VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2844 			break;
2845 		case PROP_TYPE_NUMBER:
2846 		case PROP_TYPE_INDEX:
2847 			VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2848 			break;
2849 		default:
2850 			nvlist_free(dummy);
2851 			return (SET_ERROR(EINVAL));
2852 		}
2853 
2854 		pair = nvlist_next_nvpair(dummy, NULL);
2855 		err = zfs_prop_set_special(zc->zc_name, source, pair);
2856 		nvlist_free(dummy);
2857 		if (err != -1)
2858 			return (err); /* special property already handled */
2859 	} else {
2860 		/*
2861 		 * Only check this in the non-received case. We want to allow
2862 		 * 'inherit -S' to revert non-inheritable properties like quota
2863 		 * and reservation to the received or default values even though
2864 		 * they are not considered inheritable.
2865 		 */
2866 		if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2867 			return (SET_ERROR(EINVAL));
2868 	}
2869 
2870 	/* property name has been validated by zfs_secpolicy_inherit_prop() */
2871 	return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2872 }
2873 
2874 static int
2875 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2876 {
2877 	nvlist_t *props;
2878 	spa_t *spa;
2879 	int error;
2880 	nvpair_t *pair;
2881 
2882 	if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2883 	    zc->zc_iflags, &props))
2884 		return (error);
2885 
2886 	/*
2887 	 * If the only property is the configfile, then just do a spa_lookup()
2888 	 * to handle the faulted case.
2889 	 */
2890 	pair = nvlist_next_nvpair(props, NULL);
2891 	if (pair != NULL && strcmp(nvpair_name(pair),
2892 	    zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2893 	    nvlist_next_nvpair(props, pair) == NULL) {
2894 		mutex_enter(&spa_namespace_lock);
2895 		if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2896 			spa_configfile_set(spa, props, B_FALSE);
2897 			spa_config_sync(spa, B_FALSE, B_TRUE);
2898 		}
2899 		mutex_exit(&spa_namespace_lock);
2900 		if (spa != NULL) {
2901 			nvlist_free(props);
2902 			return (0);
2903 		}
2904 	}
2905 
2906 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2907 		nvlist_free(props);
2908 		return (error);
2909 	}
2910 
2911 	error = spa_prop_set(spa, props);
2912 
2913 	nvlist_free(props);
2914 	spa_close(spa, FTAG);
2915 
2916 	return (error);
2917 }
2918 
2919 static int
2920 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2921 {
2922 	spa_t *spa;
2923 	int error;
2924 	nvlist_t *nvp = NULL;
2925 
2926 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2927 		/*
2928 		 * If the pool is faulted, there may be properties we can still
2929 		 * get (such as altroot and cachefile), so attempt to get them
2930 		 * anyway.
2931 		 */
2932 		mutex_enter(&spa_namespace_lock);
2933 		if ((spa = spa_lookup(zc->zc_name)) != NULL)
2934 			error = spa_prop_get(spa, &nvp);
2935 		mutex_exit(&spa_namespace_lock);
2936 	} else {
2937 		error = spa_prop_get(spa, &nvp);
2938 		spa_close(spa, FTAG);
2939 	}
2940 
2941 	if (error == 0 && zc->zc_nvlist_dst != NULL)
2942 		error = put_nvlist(zc, nvp);
2943 	else
2944 		error = SET_ERROR(EFAULT);
2945 
2946 	nvlist_free(nvp);
2947 	return (error);
2948 }
2949 
2950 /*
2951  * inputs:
2952  * zc_name		name of filesystem
2953  * zc_nvlist_src{_size}	nvlist of delegated permissions
2954  * zc_perm_action	allow/unallow flag
2955  *
2956  * outputs:		none
2957  */
2958 static int
2959 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2960 {
2961 	int error;
2962 	nvlist_t *fsaclnv = NULL;
2963 
2964 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2965 	    zc->zc_iflags, &fsaclnv)) != 0)
2966 		return (error);
2967 
2968 	/*
2969 	 * Verify nvlist is constructed correctly
2970 	 */
2971 	if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2972 		nvlist_free(fsaclnv);
2973 		return (SET_ERROR(EINVAL));
2974 	}
2975 
2976 	/*
2977 	 * If we don't have PRIV_SYS_MOUNT, then validate
2978 	 * that user is allowed to hand out each permission in
2979 	 * the nvlist(s)
2980 	 */
2981 
2982 	error = secpolicy_zfs(CRED());
2983 	if (error != 0) {
2984 		if (zc->zc_perm_action == B_FALSE) {
2985 			error = dsl_deleg_can_allow(zc->zc_name,
2986 			    fsaclnv, CRED());
2987 		} else {
2988 			error = dsl_deleg_can_unallow(zc->zc_name,
2989 			    fsaclnv, CRED());
2990 		}
2991 	}
2992 
2993 	if (error == 0)
2994 		error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2995 
2996 	nvlist_free(fsaclnv);
2997 	return (error);
2998 }
2999 
3000 /*
3001  * inputs:
3002  * zc_name		name of filesystem
3003  *
3004  * outputs:
3005  * zc_nvlist_src{_size}	nvlist of delegated permissions
3006  */
3007 static int
3008 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
3009 {
3010 	nvlist_t *nvp;
3011 	int error;
3012 
3013 	if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
3014 		error = put_nvlist(zc, nvp);
3015 		nvlist_free(nvp);
3016 	}
3017 
3018 	return (error);
3019 }
3020 
3021 /*
3022  * Search the vfs list for a specified resource.  Returns a pointer to it
3023  * or NULL if no suitable entry is found. The caller of this routine
3024  * is responsible for releasing the returned vfs pointer.
3025  */
3026 static vfs_t *
3027 zfs_get_vfs(const char *resource)
3028 {
3029 	struct vfs *vfsp;
3030 	struct vfs *vfs_found = NULL;
3031 
3032 	vfs_list_read_lock();
3033 	vfsp = rootvfs;
3034 	do {
3035 		if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
3036 			VFS_HOLD(vfsp);
3037 			vfs_found = vfsp;
3038 			break;
3039 		}
3040 		vfsp = vfsp->vfs_next;
3041 	} while (vfsp != rootvfs);
3042 	vfs_list_unlock();
3043 	return (vfs_found);
3044 }
3045 
3046 /* ARGSUSED */
3047 static void
3048 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
3049 {
3050 	zfs_creat_t *zct = arg;
3051 
3052 	zfs_create_fs(os, cr, zct->zct_zplprops, tx);
3053 }
3054 
3055 #define	ZFS_PROP_UNDEFINED	((uint64_t)-1)
3056 
3057 /*
3058  * inputs:
3059  * os			parent objset pointer (NULL if root fs)
3060  * fuids_ok		fuids allowed in this version of the spa?
3061  * sa_ok		SAs allowed in this version of the spa?
3062  * createprops		list of properties requested by creator
3063  *
3064  * outputs:
3065  * zplprops	values for the zplprops we attach to the master node object
3066  * is_ci	true if requested file system will be purely case-insensitive
3067  *
3068  * Determine the settings for utf8only, normalization and
3069  * casesensitivity.  Specific values may have been requested by the
3070  * creator and/or we can inherit values from the parent dataset.  If
3071  * the file system is of too early a vintage, a creator can not
3072  * request settings for these properties, even if the requested
3073  * setting is the default value.  We don't actually want to create dsl
3074  * properties for these, so remove them from the source nvlist after
3075  * processing.
3076  */
3077 static int
3078 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3079     boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3080     nvlist_t *zplprops, boolean_t *is_ci)
3081 {
3082 	uint64_t sense = ZFS_PROP_UNDEFINED;
3083 	uint64_t norm = ZFS_PROP_UNDEFINED;
3084 	uint64_t u8 = ZFS_PROP_UNDEFINED;
3085 
3086 	ASSERT(zplprops != NULL);
3087 
3088 	/*
3089 	 * Pull out creator prop choices, if any.
3090 	 */
3091 	if (createprops) {
3092 		(void) nvlist_lookup_uint64(createprops,
3093 		    zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3094 		(void) nvlist_lookup_uint64(createprops,
3095 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3096 		(void) nvlist_remove_all(createprops,
3097 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3098 		(void) nvlist_lookup_uint64(createprops,
3099 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3100 		(void) nvlist_remove_all(createprops,
3101 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3102 		(void) nvlist_lookup_uint64(createprops,
3103 		    zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3104 		(void) nvlist_remove_all(createprops,
3105 		    zfs_prop_to_name(ZFS_PROP_CASE));
3106 	}
3107 
3108 	/*
3109 	 * If the zpl version requested is whacky or the file system
3110 	 * or pool is version is too "young" to support normalization
3111 	 * and the creator tried to set a value for one of the props,
3112 	 * error out.
3113 	 */
3114 	if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3115 	    (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3116 	    (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3117 	    (zplver < ZPL_VERSION_NORMALIZATION &&
3118 	    (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3119 	    sense != ZFS_PROP_UNDEFINED)))
3120 		return (SET_ERROR(ENOTSUP));
3121 
3122 	/*
3123 	 * Put the version in the zplprops
3124 	 */
3125 	VERIFY(nvlist_add_uint64(zplprops,
3126 	    zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3127 
3128 	if (norm == ZFS_PROP_UNDEFINED)
3129 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3130 	VERIFY(nvlist_add_uint64(zplprops,
3131 	    zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3132 
3133 	/*
3134 	 * If we're normalizing, names must always be valid UTF-8 strings.
3135 	 */
3136 	if (norm)
3137 		u8 = 1;
3138 	if (u8 == ZFS_PROP_UNDEFINED)
3139 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3140 	VERIFY(nvlist_add_uint64(zplprops,
3141 	    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3142 
3143 	if (sense == ZFS_PROP_UNDEFINED)
3144 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3145 	VERIFY(nvlist_add_uint64(zplprops,
3146 	    zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3147 
3148 	if (is_ci)
3149 		*is_ci = (sense == ZFS_CASE_INSENSITIVE);
3150 
3151 	return (0);
3152 }
3153 
3154 static int
3155 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3156     nvlist_t *zplprops, boolean_t *is_ci)
3157 {
3158 	boolean_t fuids_ok, sa_ok;
3159 	uint64_t zplver = ZPL_VERSION;
3160 	objset_t *os = NULL;
3161 	char parentname[ZFS_MAX_DATASET_NAME_LEN];
3162 	char *cp;
3163 	spa_t *spa;
3164 	uint64_t spa_vers;
3165 	int error;
3166 
3167 	(void) strlcpy(parentname, dataset, sizeof (parentname));
3168 	cp = strrchr(parentname, '/');
3169 	ASSERT(cp != NULL);
3170 	cp[0] = '\0';
3171 
3172 	if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3173 		return (error);
3174 
3175 	spa_vers = spa_version(spa);
3176 	spa_close(spa, FTAG);
3177 
3178 	zplver = zfs_zpl_version_map(spa_vers);
3179 	fuids_ok = (zplver >= ZPL_VERSION_FUID);
3180 	sa_ok = (zplver >= ZPL_VERSION_SA);
3181 
3182 	/*
3183 	 * Open parent object set so we can inherit zplprop values.
3184 	 */
3185 	if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3186 		return (error);
3187 
3188 	error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3189 	    zplprops, is_ci);
3190 	dmu_objset_rele(os, FTAG);
3191 	return (error);
3192 }
3193 
3194 static int
3195 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3196     nvlist_t *zplprops, boolean_t *is_ci)
3197 {
3198 	boolean_t fuids_ok;
3199 	boolean_t sa_ok;
3200 	uint64_t zplver = ZPL_VERSION;
3201 	int error;
3202 
3203 	zplver = zfs_zpl_version_map(spa_vers);
3204 	fuids_ok = (zplver >= ZPL_VERSION_FUID);
3205 	sa_ok = (zplver >= ZPL_VERSION_SA);
3206 
3207 	error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3208 	    createprops, zplprops, is_ci);
3209 	return (error);
3210 }
3211 
3212 /*
3213  * innvl: {
3214  *     "type" -> dmu_objset_type_t (int32)
3215  *     (optional) "props" -> { prop -> value }
3216  * }
3217  *
3218  * outnvl: propname -> error code (int32)
3219  */
3220 static int
3221 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3222 {
3223 	int error = 0;
3224 	zfs_creat_t zct = { 0 };
3225 	nvlist_t *nvprops = NULL;
3226 	void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3227 	int32_t type32;
3228 	dmu_objset_type_t type;
3229 	boolean_t is_insensitive = B_FALSE;
3230 
3231 	if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3232 		return (SET_ERROR(EINVAL));
3233 	type = type32;
3234 	(void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3235 
3236 	switch (type) {
3237 	case DMU_OST_ZFS:
3238 		cbfunc = zfs_create_cb;
3239 		break;
3240 
3241 	case DMU_OST_ZVOL:
3242 		cbfunc = zvol_create_cb;
3243 		break;
3244 
3245 	default:
3246 		cbfunc = NULL;
3247 		break;
3248 	}
3249 	if (strchr(fsname, '@') ||
3250 	    strchr(fsname, '%'))
3251 		return (SET_ERROR(EINVAL));
3252 
3253 	zct.zct_props = nvprops;
3254 
3255 	if (cbfunc == NULL)
3256 		return (SET_ERROR(EINVAL));
3257 
3258 	if (type == DMU_OST_ZVOL) {
3259 		uint64_t volsize, volblocksize;
3260 
3261 		if (nvprops == NULL)
3262 			return (SET_ERROR(EINVAL));
3263 		if (nvlist_lookup_uint64(nvprops,
3264 		    zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3265 			return (SET_ERROR(EINVAL));
3266 
3267 		if ((error = nvlist_lookup_uint64(nvprops,
3268 		    zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3269 		    &volblocksize)) != 0 && error != ENOENT)
3270 			return (SET_ERROR(EINVAL));
3271 
3272 		if (error != 0)
3273 			volblocksize = zfs_prop_default_numeric(
3274 			    ZFS_PROP_VOLBLOCKSIZE);
3275 
3276 		if ((error = zvol_check_volblocksize(
3277 		    volblocksize)) != 0 ||
3278 		    (error = zvol_check_volsize(volsize,
3279 		    volblocksize)) != 0)
3280 			return (error);
3281 	} else if (type == DMU_OST_ZFS) {
3282 		int error;
3283 
3284 		/*
3285 		 * We have to have normalization and
3286 		 * case-folding flags correct when we do the
3287 		 * file system creation, so go figure them out
3288 		 * now.
3289 		 */
3290 		VERIFY(nvlist_alloc(&zct.zct_zplprops,
3291 		    NV_UNIQUE_NAME, KM_SLEEP) == 0);
3292 		error = zfs_fill_zplprops(fsname, nvprops,
3293 		    zct.zct_zplprops, &is_insensitive);
3294 		if (error != 0) {
3295 			nvlist_free(zct.zct_zplprops);
3296 			return (error);
3297 		}
3298 	}
3299 
3300 	error = dmu_objset_create(fsname, type,
3301 	    is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
3302 	nvlist_free(zct.zct_zplprops);
3303 
3304 	/*
3305 	 * It would be nice to do this atomically.
3306 	 */
3307 	if (error == 0) {
3308 		error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3309 		    nvprops, outnvl);
3310 		if (error != 0)
3311 			(void) dsl_destroy_head(fsname);
3312 	}
3313 	return (error);
3314 }
3315 
3316 /*
3317  * innvl: {
3318  *     "origin" -> name of origin snapshot
3319  *     (optional) "props" -> { prop -> value }
3320  * }
3321  *
3322  * outnvl: propname -> error code (int32)
3323  */
3324 static int
3325 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3326 {
3327 	int error = 0;
3328 	nvlist_t *nvprops = NULL;
3329 	char *origin_name;
3330 
3331 	if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3332 		return (SET_ERROR(EINVAL));
3333 	(void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3334 
3335 	if (strchr(fsname, '@') ||
3336 	    strchr(fsname, '%'))
3337 		return (SET_ERROR(EINVAL));
3338 
3339 	if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3340 		return (SET_ERROR(EINVAL));
3341 	error = dmu_objset_clone(fsname, origin_name);
3342 	if (error != 0)
3343 		return (error);
3344 
3345 	/*
3346 	 * It would be nice to do this atomically.
3347 	 */
3348 	if (error == 0) {
3349 		error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3350 		    nvprops, outnvl);
3351 		if (error != 0)
3352 			(void) dsl_destroy_head(fsname);
3353 	}
3354 	return (error);
3355 }
3356 
3357 /*
3358  * innvl: {
3359  *     "snaps" -> { snapshot1, snapshot2 }
3360  *     (optional) "props" -> { prop -> value (string) }
3361  * }
3362  *
3363  * outnvl: snapshot -> error code (int32)
3364  */
3365 static int
3366 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3367 {
3368 	nvlist_t *snaps;
3369 	nvlist_t *props = NULL;
3370 	int error, poollen;
3371 	nvpair_t *pair;
3372 
3373 	(void) nvlist_lookup_nvlist(innvl, "props", &props);
3374 	if ((error = zfs_check_userprops(poolname, props)) != 0)
3375 		return (error);
3376 
3377 	if (!nvlist_empty(props) &&
3378 	    zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3379 		return (SET_ERROR(ENOTSUP));
3380 
3381 	if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3382 		return (SET_ERROR(EINVAL));
3383 	poollen = strlen(poolname);
3384 	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3385 	    pair = nvlist_next_nvpair(snaps, pair)) {
3386 		const char *name = nvpair_name(pair);
3387 		const char *cp = strchr(name, '@');
3388 
3389 		/*
3390 		 * The snap name must contain an @, and the part after it must
3391 		 * contain only valid characters.
3392 		 */
3393 		if (cp == NULL ||
3394 		    zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3395 			return (SET_ERROR(EINVAL));
3396 
3397 		/*
3398 		 * The snap must be in the specified pool.
3399 		 */
3400 		if (strncmp(name, poolname, poollen) != 0 ||
3401 		    (name[poollen] != '/' && name[poollen] != '@'))
3402 			return (SET_ERROR(EXDEV));
3403 
3404 		/* This must be the only snap of this fs. */
3405 		for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3406 		    pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3407 			if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3408 			    == 0) {
3409 				return (SET_ERROR(EXDEV));
3410 			}
3411 		}
3412 	}
3413 
3414 	error = dsl_dataset_snapshot(snaps, props, outnvl);
3415 	return (error);
3416 }
3417 
3418 /*
3419  * innvl: "message" -> string
3420  */
3421 /* ARGSUSED */
3422 static int
3423 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3424 {
3425 	char *message;
3426 	spa_t *spa;
3427 	int error;
3428 	char *poolname;
3429 
3430 	/*
3431 	 * The poolname in the ioctl is not set, we get it from the TSD,
3432 	 * which was set at the end of the last successful ioctl that allows
3433 	 * logging.  The secpolicy func already checked that it is set.
3434 	 * Only one log ioctl is allowed after each successful ioctl, so
3435 	 * we clear the TSD here.
3436 	 */
3437 	poolname = tsd_get(zfs_allow_log_key);
3438 	(void) tsd_set(zfs_allow_log_key, NULL);
3439 	error = spa_open(poolname, &spa, FTAG);
3440 	strfree(poolname);
3441 	if (error != 0)
3442 		return (error);
3443 
3444 	if (nvlist_lookup_string(innvl, "message", &message) != 0)  {
3445 		spa_close(spa, FTAG);
3446 		return (SET_ERROR(EINVAL));
3447 	}
3448 
3449 	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3450 		spa_close(spa, FTAG);
3451 		return (SET_ERROR(ENOTSUP));
3452 	}
3453 
3454 	error = spa_history_log(spa, message);
3455 	spa_close(spa, FTAG);
3456 	return (error);
3457 }
3458 
3459 /*
3460  * The dp_config_rwlock must not be held when calling this, because the
3461  * unmount may need to write out data.
3462  *
3463  * This function is best-effort.  Callers must deal gracefully if it
3464  * remains mounted (or is remounted after this call).
3465  *
3466  * Returns 0 if the argument is not a snapshot, or it is not currently a
3467  * filesystem, or we were able to unmount it.  Returns error code otherwise.
3468  */
3469 int
3470 zfs_unmount_snap(const char *snapname)
3471 {
3472 	vfs_t *vfsp;
3473 	zfsvfs_t *zfsvfs;
3474 	int err;
3475 
3476 	if (strchr(snapname, '@') == NULL)
3477 		return (0);
3478 
3479 	vfsp = zfs_get_vfs(snapname);
3480 	if (vfsp == NULL)
3481 		return (0);
3482 
3483 	zfsvfs = vfsp->vfs_data;
3484 	ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os)));
3485 
3486 	err = vn_vfswlock(vfsp->vfs_vnodecovered);
3487 	VFS_RELE(vfsp);
3488 	if (err != 0)
3489 		return (SET_ERROR(err));
3490 
3491 	/*
3492 	 * Always force the unmount for snapshots.
3493 	 */
3494 	(void) dounmount(vfsp, MS_FORCE, kcred);
3495 	return (0);
3496 }
3497 
3498 /* ARGSUSED */
3499 static int
3500 zfs_unmount_snap_cb(const char *snapname, void *arg)
3501 {
3502 	return (zfs_unmount_snap(snapname));
3503 }
3504 
3505 /*
3506  * When a clone is destroyed, its origin may also need to be destroyed,
3507  * in which case it must be unmounted.  This routine will do that unmount
3508  * if necessary.
3509  */
3510 void
3511 zfs_destroy_unmount_origin(const char *fsname)
3512 {
3513 	int error;
3514 	objset_t *os;
3515 	dsl_dataset_t *ds;
3516 
3517 	error = dmu_objset_hold(fsname, FTAG, &os);
3518 	if (error != 0)
3519 		return;
3520 	ds = dmu_objset_ds(os);
3521 	if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3522 		char originname[ZFS_MAX_DATASET_NAME_LEN];
3523 		dsl_dataset_name(ds->ds_prev, originname);
3524 		dmu_objset_rele(os, FTAG);
3525 		(void) zfs_unmount_snap(originname);
3526 	} else {
3527 		dmu_objset_rele(os, FTAG);
3528 	}
3529 }
3530 
3531 /*
3532  * innvl: {
3533  *     "snaps" -> { snapshot1, snapshot2 }
3534  *     (optional boolean) "defer"
3535  * }
3536  *
3537  * outnvl: snapshot -> error code (int32)
3538  *
3539  */
3540 /* ARGSUSED */
3541 static int
3542 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3543 {
3544 	nvlist_t *snaps;
3545 	nvpair_t *pair;
3546 	boolean_t defer;
3547 
3548 	if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3549 		return (SET_ERROR(EINVAL));
3550 	defer = nvlist_exists(innvl, "defer");
3551 
3552 	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3553 	    pair = nvlist_next_nvpair(snaps, pair)) {
3554 		(void) zfs_unmount_snap(nvpair_name(pair));
3555 	}
3556 
3557 	return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3558 }
3559 
3560 /*
3561  * Create bookmarks.  Bookmark names are of the form <fs>#<bmark>.
3562  * All bookmarks must be in the same pool.
3563  *
3564  * innvl: {
3565  *     bookmark1 -> snapshot1, bookmark2 -> snapshot2
3566  * }
3567  *
3568  * outnvl: bookmark -> error code (int32)
3569  *
3570  */
3571 /* ARGSUSED */
3572 static int
3573 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3574 {
3575 	for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3576 	    pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3577 		char *snap_name;
3578 
3579 		/*
3580 		 * Verify the snapshot argument.
3581 		 */
3582 		if (nvpair_value_string(pair, &snap_name) != 0)
3583 			return (SET_ERROR(EINVAL));
3584 
3585 
3586 		/* Verify that the keys (bookmarks) are unique */
3587 		for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair);
3588 		    pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) {
3589 			if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0)
3590 				return (SET_ERROR(EINVAL));
3591 		}
3592 	}
3593 
3594 	return (dsl_bookmark_create(innvl, outnvl));
3595 }
3596 
3597 /*
3598  * innvl: {
3599  *     property 1, property 2, ...
3600  * }
3601  *
3602  * outnvl: {
3603  *     bookmark name 1 -> { property 1, property 2, ... },
3604  *     bookmark name 2 -> { property 1, property 2, ... }
3605  * }
3606  *
3607  */
3608 static int
3609 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3610 {
3611 	return (dsl_get_bookmarks(fsname, innvl, outnvl));
3612 }
3613 
3614 /*
3615  * innvl: {
3616  *     bookmark name 1, bookmark name 2
3617  * }
3618  *
3619  * outnvl: bookmark -> error code (int32)
3620  *
3621  */
3622 static int
3623 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
3624     nvlist_t *outnvl)
3625 {
3626 	int error, poollen;
3627 
3628 	poollen = strlen(poolname);
3629 	for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3630 	    pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3631 		const char *name = nvpair_name(pair);
3632 		const char *cp = strchr(name, '#');
3633 
3634 		/*
3635 		 * The bookmark name must contain an #, and the part after it
3636 		 * must contain only valid characters.
3637 		 */
3638 		if (cp == NULL ||
3639 		    zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3640 			return (SET_ERROR(EINVAL));
3641 
3642 		/*
3643 		 * The bookmark must be in the specified pool.
3644 		 */
3645 		if (strncmp(name, poolname, poollen) != 0 ||
3646 		    (name[poollen] != '/' && name[poollen] != '#'))
3647 			return (SET_ERROR(EXDEV));
3648 	}
3649 
3650 	error = dsl_bookmark_destroy(innvl, outnvl);
3651 	return (error);
3652 }
3653 
3654 /*
3655  * inputs:
3656  * zc_name		name of dataset to destroy
3657  * zc_objset_type	type of objset
3658  * zc_defer_destroy	mark for deferred destroy
3659  *
3660  * outputs:		none
3661  */
3662 static int
3663 zfs_ioc_destroy(zfs_cmd_t *zc)
3664 {
3665 	int err;
3666 
3667 	if (zc->zc_objset_type == DMU_OST_ZFS) {
3668 		err = zfs_unmount_snap(zc->zc_name);
3669 		if (err != 0)
3670 			return (err);
3671 	}
3672 
3673 	if (strchr(zc->zc_name, '@'))
3674 		err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3675 	else
3676 		err = dsl_destroy_head(zc->zc_name);
3677 	if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
3678 		(void) zvol_remove_minor(zc->zc_name);
3679 	return (err);
3680 }
3681 
3682 /*
3683  * fsname is name of dataset to rollback (to most recent snapshot)
3684  *
3685  * innvl is not used.
3686  *
3687  * outnvl: "target" -> name of most recent snapshot
3688  * }
3689  */
3690 /* ARGSUSED */
3691 static int
3692 zfs_ioc_rollback(const char *fsname, nvlist_t *args, nvlist_t *outnvl)
3693 {
3694 	zfsvfs_t *zfsvfs;
3695 	int error;
3696 
3697 	if (getzfsvfs(fsname, &zfsvfs) == 0) {
3698 		error = zfs_suspend_fs(zfsvfs);
3699 		if (error == 0) {
3700 			int resume_err;
3701 
3702 			error = dsl_dataset_rollback(fsname, zfsvfs, outnvl);
3703 			resume_err = zfs_resume_fs(zfsvfs, fsname);
3704 			error = error ? error : resume_err;
3705 		}
3706 		VFS_RELE(zfsvfs->z_vfs);
3707 	} else {
3708 		error = dsl_dataset_rollback(fsname, NULL, outnvl);
3709 	}
3710 	return (error);
3711 }
3712 
3713 static int
3714 recursive_unmount(const char *fsname, void *arg)
3715 {
3716 	const char *snapname = arg;
3717 	char fullname[ZFS_MAX_DATASET_NAME_LEN];
3718 
3719 	(void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname);
3720 	return (zfs_unmount_snap(fullname));
3721 }
3722 
3723 /*
3724  * inputs:
3725  * zc_name	old name of dataset
3726  * zc_value	new name of dataset
3727  * zc_cookie	recursive flag (only valid for snapshots)
3728  *
3729  * outputs:	none
3730  */
3731 static int
3732 zfs_ioc_rename(zfs_cmd_t *zc)
3733 {
3734 	boolean_t recursive = zc->zc_cookie & 1;
3735 	char *at;
3736 
3737 	zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3738 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3739 	    strchr(zc->zc_value, '%'))
3740 		return (SET_ERROR(EINVAL));
3741 
3742 	at = strchr(zc->zc_name, '@');
3743 	if (at != NULL) {
3744 		/* snaps must be in same fs */
3745 		int error;
3746 
3747 		if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
3748 			return (SET_ERROR(EXDEV));
3749 		*at = '\0';
3750 		if (zc->zc_objset_type == DMU_OST_ZFS) {
3751 			error = dmu_objset_find(zc->zc_name,
3752 			    recursive_unmount, at + 1,
3753 			    recursive ? DS_FIND_CHILDREN : 0);
3754 			if (error != 0) {
3755 				*at = '@';
3756 				return (error);
3757 			}
3758 		}
3759 		error = dsl_dataset_rename_snapshot(zc->zc_name,
3760 		    at + 1, strchr(zc->zc_value, '@') + 1, recursive);
3761 		*at = '@';
3762 
3763 		return (error);
3764 	} else {
3765 		if (zc->zc_objset_type == DMU_OST_ZVOL)
3766 			(void) zvol_remove_minor(zc->zc_name);
3767 		return (dsl_dir_rename(zc->zc_name, zc->zc_value));
3768 	}
3769 }
3770 
3771 static int
3772 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3773 {
3774 	const char *propname = nvpair_name(pair);
3775 	boolean_t issnap = (strchr(dsname, '@') != NULL);
3776 	zfs_prop_t prop = zfs_name_to_prop(propname);
3777 	uint64_t intval;
3778 	int err;
3779 
3780 	if (prop == ZPROP_INVAL) {
3781 		if (zfs_prop_user(propname)) {
3782 			if (err = zfs_secpolicy_write_perms(dsname,
3783 			    ZFS_DELEG_PERM_USERPROP, cr))
3784 				return (err);
3785 			return (0);
3786 		}
3787 
3788 		if (!issnap && zfs_prop_userquota(propname)) {
3789 			const char *perm = NULL;
3790 			const char *uq_prefix =
3791 			    zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3792 			const char *gq_prefix =
3793 			    zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3794 
3795 			if (strncmp(propname, uq_prefix,
3796 			    strlen(uq_prefix)) == 0) {
3797 				perm = ZFS_DELEG_PERM_USERQUOTA;
3798 			} else if (strncmp(propname, gq_prefix,
3799 			    strlen(gq_prefix)) == 0) {
3800 				perm = ZFS_DELEG_PERM_GROUPQUOTA;
3801 			} else {
3802 				/* USERUSED and GROUPUSED are read-only */
3803 				return (SET_ERROR(EINVAL));
3804 			}
3805 
3806 			if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3807 				return (err);
3808 			return (0);
3809 		}
3810 
3811 		return (SET_ERROR(EINVAL));
3812 	}
3813 
3814 	if (issnap)
3815 		return (SET_ERROR(EINVAL));
3816 
3817 	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3818 		/*
3819 		 * dsl_prop_get_all_impl() returns properties in this
3820 		 * format.
3821 		 */
3822 		nvlist_t *attrs;
3823 		VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3824 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3825 		    &pair) == 0);
3826 	}
3827 
3828 	/*
3829 	 * Check that this value is valid for this pool version
3830 	 */
3831 	switch (prop) {
3832 	case ZFS_PROP_COMPRESSION:
3833 		/*
3834 		 * If the user specified gzip compression, make sure
3835 		 * the SPA supports it. We ignore any errors here since
3836 		 * we'll catch them later.
3837 		 */
3838 		if (nvpair_value_uint64(pair, &intval) == 0) {
3839 			if (intval >= ZIO_COMPRESS_GZIP_1 &&
3840 			    intval <= ZIO_COMPRESS_GZIP_9 &&
3841 			    zfs_earlier_version(dsname,
3842 			    SPA_VERSION_GZIP_COMPRESSION)) {
3843 				return (SET_ERROR(ENOTSUP));
3844 			}
3845 
3846 			if (intval == ZIO_COMPRESS_ZLE &&
3847 			    zfs_earlier_version(dsname,
3848 			    SPA_VERSION_ZLE_COMPRESSION))
3849 				return (SET_ERROR(ENOTSUP));
3850 
3851 			if (intval == ZIO_COMPRESS_LZ4) {
3852 				spa_t *spa;
3853 
3854 				if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3855 					return (err);
3856 
3857 				if (!spa_feature_is_enabled(spa,
3858 				    SPA_FEATURE_LZ4_COMPRESS)) {
3859 					spa_close(spa, FTAG);
3860 					return (SET_ERROR(ENOTSUP));
3861 				}
3862 				spa_close(spa, FTAG);
3863 			}
3864 
3865 			/*
3866 			 * If this is a bootable dataset then
3867 			 * verify that the compression algorithm
3868 			 * is supported for booting. We must return
3869 			 * something other than ENOTSUP since it
3870 			 * implies a downrev pool version.
3871 			 */
3872 			if (zfs_is_bootfs(dsname) &&
3873 			    !BOOTFS_COMPRESS_VALID(intval)) {
3874 				return (SET_ERROR(ERANGE));
3875 			}
3876 		}
3877 		break;
3878 
3879 	case ZFS_PROP_COPIES:
3880 		if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3881 			return (SET_ERROR(ENOTSUP));
3882 		break;
3883 
3884 	case ZFS_PROP_DEDUP:
3885 		if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3886 			return (SET_ERROR(ENOTSUP));
3887 		break;
3888 
3889 	case ZFS_PROP_RECORDSIZE:
3890 		/* Record sizes above 128k need the feature to be enabled */
3891 		if (nvpair_value_uint64(pair, &intval) == 0 &&
3892 		    intval > SPA_OLD_MAXBLOCKSIZE) {
3893 			spa_t *spa;
3894 
3895 			/*
3896 			 * If this is a bootable dataset then
3897 			 * the we don't allow large (>128K) blocks,
3898 			 * because GRUB doesn't support them.
3899 			 */
3900 			if (zfs_is_bootfs(dsname) &&
3901 			    intval > SPA_OLD_MAXBLOCKSIZE) {
3902 				return (SET_ERROR(EDOM));
3903 			}
3904 
3905 			/*
3906 			 * We don't allow setting the property above 1MB,
3907 			 * unless the tunable has been changed.
3908 			 */
3909 			if (intval > zfs_max_recordsize ||
3910 			    intval > SPA_MAXBLOCKSIZE)
3911 				return (SET_ERROR(EDOM));
3912 
3913 			if ((err = spa_open(dsname, &spa, FTAG)) != 0)
3914 				return (err);
3915 
3916 			if (!spa_feature_is_enabled(spa,
3917 			    SPA_FEATURE_LARGE_BLOCKS)) {
3918 				spa_close(spa, FTAG);
3919 				return (SET_ERROR(ENOTSUP));
3920 			}
3921 			spa_close(spa, FTAG);
3922 		}
3923 		break;
3924 
3925 	case ZFS_PROP_SHARESMB:
3926 		if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3927 			return (SET_ERROR(ENOTSUP));
3928 		break;
3929 
3930 	case ZFS_PROP_ACLINHERIT:
3931 		if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3932 		    nvpair_value_uint64(pair, &intval) == 0) {
3933 			if (intval == ZFS_ACL_PASSTHROUGH_X &&
3934 			    zfs_earlier_version(dsname,
3935 			    SPA_VERSION_PASSTHROUGH_X))
3936 				return (SET_ERROR(ENOTSUP));
3937 		}
3938 		break;
3939 	}
3940 
3941 	return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3942 }
3943 
3944 /*
3945  * Checks for a race condition to make sure we don't increment a feature flag
3946  * multiple times.
3947  */
3948 static int
3949 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
3950 {
3951 	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3952 	spa_feature_t *featurep = arg;
3953 
3954 	if (!spa_feature_is_active(spa, *featurep))
3955 		return (0);
3956 	else
3957 		return (SET_ERROR(EBUSY));
3958 }
3959 
3960 /*
3961  * The callback invoked on feature activation in the sync task caused by
3962  * zfs_prop_activate_feature.
3963  */
3964 static void
3965 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
3966 {
3967 	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
3968 	spa_feature_t *featurep = arg;
3969 
3970 	spa_feature_incr(spa, *featurep, tx);
3971 }
3972 
3973 /*
3974  * Activates a feature on a pool in response to a property setting. This
3975  * creates a new sync task which modifies the pool to reflect the feature
3976  * as being active.
3977  */
3978 static int
3979 zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature)
3980 {
3981 	int err;
3982 
3983 	/* EBUSY here indicates that the feature is already active */
3984 	err = dsl_sync_task(spa_name(spa),
3985 	    zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
3986 	    &feature, 2, ZFS_SPACE_CHECK_RESERVED);
3987 
3988 	if (err != 0 && err != EBUSY)
3989 		return (err);
3990 	else
3991 		return (0);
3992 }
3993 
3994 /*
3995  * Removes properties from the given props list that fail permission checks
3996  * needed to clear them and to restore them in case of a receive error. For each
3997  * property, make sure we have both set and inherit permissions.
3998  *
3999  * Returns the first error encountered if any permission checks fail. If the
4000  * caller provides a non-NULL errlist, it also gives the complete list of names
4001  * of all the properties that failed a permission check along with the
4002  * corresponding error numbers. The caller is responsible for freeing the
4003  * returned errlist.
4004  *
4005  * If every property checks out successfully, zero is returned and the list
4006  * pointed at by errlist is NULL.
4007  */
4008 static int
4009 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
4010 {
4011 	zfs_cmd_t *zc;
4012 	nvpair_t *pair, *next_pair;
4013 	nvlist_t *errors;
4014 	int err, rv = 0;
4015 
4016 	if (props == NULL)
4017 		return (0);
4018 
4019 	VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4020 
4021 	zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
4022 	(void) strcpy(zc->zc_name, dataset);
4023 	pair = nvlist_next_nvpair(props, NULL);
4024 	while (pair != NULL) {
4025 		next_pair = nvlist_next_nvpair(props, pair);
4026 
4027 		(void) strcpy(zc->zc_value, nvpair_name(pair));
4028 		if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
4029 		    (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
4030 			VERIFY(nvlist_remove_nvpair(props, pair) == 0);
4031 			VERIFY(nvlist_add_int32(errors,
4032 			    zc->zc_value, err) == 0);
4033 		}
4034 		pair = next_pair;
4035 	}
4036 	kmem_free(zc, sizeof (zfs_cmd_t));
4037 
4038 	if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
4039 		nvlist_free(errors);
4040 		errors = NULL;
4041 	} else {
4042 		VERIFY(nvpair_value_int32(pair, &rv) == 0);
4043 	}
4044 
4045 	if (errlist == NULL)
4046 		nvlist_free(errors);
4047 	else
4048 		*errlist = errors;
4049 
4050 	return (rv);
4051 }
4052 
4053 static boolean_t
4054 propval_equals(nvpair_t *p1, nvpair_t *p2)
4055 {
4056 	if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
4057 		/* dsl_prop_get_all_impl() format */
4058 		nvlist_t *attrs;
4059 		VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
4060 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4061 		    &p1) == 0);
4062 	}
4063 
4064 	if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
4065 		nvlist_t *attrs;
4066 		VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
4067 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4068 		    &p2) == 0);
4069 	}
4070 
4071 	if (nvpair_type(p1) != nvpair_type(p2))
4072 		return (B_FALSE);
4073 
4074 	if (nvpair_type(p1) == DATA_TYPE_STRING) {
4075 		char *valstr1, *valstr2;
4076 
4077 		VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
4078 		VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
4079 		return (strcmp(valstr1, valstr2) == 0);
4080 	} else {
4081 		uint64_t intval1, intval2;
4082 
4083 		VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4084 		VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4085 		return (intval1 == intval2);
4086 	}
4087 }
4088 
4089 /*
4090  * Remove properties from props if they are not going to change (as determined
4091  * by comparison with origprops). Remove them from origprops as well, since we
4092  * do not need to clear or restore properties that won't change.
4093  */
4094 static void
4095 props_reduce(nvlist_t *props, nvlist_t *origprops)
4096 {
4097 	nvpair_t *pair, *next_pair;
4098 
4099 	if (origprops == NULL)
4100 		return; /* all props need to be received */
4101 
4102 	pair = nvlist_next_nvpair(props, NULL);
4103 	while (pair != NULL) {
4104 		const char *propname = nvpair_name(pair);
4105 		nvpair_t *match;
4106 
4107 		next_pair = nvlist_next_nvpair(props, pair);
4108 
4109 		if ((nvlist_lookup_nvpair(origprops, propname,
4110 		    &match) != 0) || !propval_equals(pair, match))
4111 			goto next; /* need to set received value */
4112 
4113 		/* don't clear the existing received value */
4114 		(void) nvlist_remove_nvpair(origprops, match);
4115 		/* don't bother receiving the property */
4116 		(void) nvlist_remove_nvpair(props, pair);
4117 next:
4118 		pair = next_pair;
4119 	}
4120 }
4121 
4122 #ifdef	DEBUG
4123 static boolean_t zfs_ioc_recv_inject_err;
4124 #endif
4125 
4126 /*
4127  * inputs:
4128  * zc_name		name of containing filesystem
4129  * zc_nvlist_src{_size}	nvlist of properties to apply
4130  * zc_value		name of snapshot to create
4131  * zc_string		name of clone origin (if DRR_FLAG_CLONE)
4132  * zc_cookie		file descriptor to recv from
4133  * zc_begin_record	the BEGIN record of the stream (not byteswapped)
4134  * zc_guid		force flag
4135  * zc_cleanup_fd	cleanup-on-exit file descriptor
4136  * zc_action_handle	handle for this guid/ds mapping (or zero on first call)
4137  * zc_resumable		if data is incomplete assume sender will resume
4138  *
4139  * outputs:
4140  * zc_cookie		number of bytes read
4141  * zc_nvlist_dst{_size} error for each unapplied received property
4142  * zc_obj		zprop_errflags_t
4143  * zc_action_handle	handle for this guid/ds mapping
4144  */
4145 static int
4146 zfs_ioc_recv(zfs_cmd_t *zc)
4147 {
4148 	file_t *fp;
4149 	dmu_recv_cookie_t drc;
4150 	boolean_t force = (boolean_t)zc->zc_guid;
4151 	int fd;
4152 	int error = 0;
4153 	int props_error = 0;
4154 	nvlist_t *errors;
4155 	offset_t off;
4156 	nvlist_t *props = NULL; /* sent properties */
4157 	nvlist_t *origprops = NULL; /* existing properties */
4158 	char *origin = NULL;
4159 	char *tosnap;
4160 	char tofs[ZFS_MAX_DATASET_NAME_LEN];
4161 	boolean_t first_recvd_props = B_FALSE;
4162 
4163 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4164 	    strchr(zc->zc_value, '@') == NULL ||
4165 	    strchr(zc->zc_value, '%'))
4166 		return (SET_ERROR(EINVAL));
4167 
4168 	(void) strcpy(tofs, zc->zc_value);
4169 	tosnap = strchr(tofs, '@');
4170 	*tosnap++ = '\0';
4171 
4172 	if (zc->zc_nvlist_src != NULL &&
4173 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
4174 	    zc->zc_iflags, &props)) != 0)
4175 		return (error);
4176 
4177 	fd = zc->zc_cookie;
4178 	fp = getf(fd);
4179 	if (fp == NULL) {
4180 		nvlist_free(props);
4181 		return (SET_ERROR(EBADF));
4182 	}
4183 
4184 	errors = fnvlist_alloc();
4185 
4186 	if (zc->zc_string[0])
4187 		origin = zc->zc_string;
4188 
4189 	error = dmu_recv_begin(tofs, tosnap,
4190 	    &zc->zc_begin_record, force, zc->zc_resumable, origin, &drc);
4191 	if (error != 0)
4192 		goto out;
4193 
4194 	/*
4195 	 * Set properties before we receive the stream so that they are applied
4196 	 * to the new data. Note that we must call dmu_recv_stream() if
4197 	 * dmu_recv_begin() succeeds.
4198 	 */
4199 	if (props != NULL && !drc.drc_newfs) {
4200 		if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
4201 		    SPA_VERSION_RECVD_PROPS &&
4202 		    !dsl_prop_get_hasrecvd(tofs))
4203 			first_recvd_props = B_TRUE;
4204 
4205 		/*
4206 		 * If new received properties are supplied, they are to
4207 		 * completely replace the existing received properties, so stash
4208 		 * away the existing ones.
4209 		 */
4210 		if (dsl_prop_get_received(tofs, &origprops) == 0) {
4211 			nvlist_t *errlist = NULL;
4212 			/*
4213 			 * Don't bother writing a property if its value won't
4214 			 * change (and avoid the unnecessary security checks).
4215 			 *
4216 			 * The first receive after SPA_VERSION_RECVD_PROPS is a
4217 			 * special case where we blow away all local properties
4218 			 * regardless.
4219 			 */
4220 			if (!first_recvd_props)
4221 				props_reduce(props, origprops);
4222 			if (zfs_check_clearable(tofs, origprops, &errlist) != 0)
4223 				(void) nvlist_merge(errors, errlist, 0);
4224 			nvlist_free(errlist);
4225 
4226 			if (clear_received_props(tofs, origprops,
4227 			    first_recvd_props ? NULL : props) != 0)
4228 				zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4229 		} else {
4230 			zc->zc_obj |= ZPROP_ERR_NOCLEAR;
4231 		}
4232 	}
4233 
4234 	if (props != NULL) {
4235 		props_error = dsl_prop_set_hasrecvd(tofs);
4236 
4237 		if (props_error == 0) {
4238 			(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4239 			    props, errors);
4240 		}
4241 	}
4242 
4243 	if (zc->zc_nvlist_dst_size != 0 &&
4244 	    (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4245 	    put_nvlist(zc, errors) != 0)) {
4246 		/*
4247 		 * Caller made zc->zc_nvlist_dst less than the minimum expected
4248 		 * size or supplied an invalid address.
4249 		 */
4250 		props_error = SET_ERROR(EINVAL);
4251 	}
4252 
4253 	off = fp->f_offset;
4254 	error = dmu_recv_stream(&drc, fp->f_vnode, &off, zc->zc_cleanup_fd,
4255 	    &zc->zc_action_handle);
4256 
4257 	if (error == 0) {
4258 		zfsvfs_t *zfsvfs = NULL;
4259 
4260 		if (getzfsvfs(tofs, &zfsvfs) == 0) {
4261 			/* online recv */
4262 			int end_err;
4263 
4264 			error = zfs_suspend_fs(zfsvfs);
4265 			/*
4266 			 * If the suspend fails, then the recv_end will
4267 			 * likely also fail, and clean up after itself.
4268 			 */
4269 			end_err = dmu_recv_end(&drc, zfsvfs);
4270 			if (error == 0)
4271 				error = zfs_resume_fs(zfsvfs, tofs);
4272 			error = error ? error : end_err;
4273 			VFS_RELE(zfsvfs->z_vfs);
4274 		} else {
4275 			error = dmu_recv_end(&drc, NULL);
4276 		}
4277 	}
4278 
4279 	zc->zc_cookie = off - fp->f_offset;
4280 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4281 		fp->f_offset = off;
4282 
4283 #ifdef	DEBUG
4284 	if (zfs_ioc_recv_inject_err) {
4285 		zfs_ioc_recv_inject_err = B_FALSE;
4286 		error = 1;
4287 	}
4288 #endif
4289 	/*
4290 	 * On error, restore the original props.
4291 	 */
4292 	if (error != 0 && props != NULL && !drc.drc_newfs) {
4293 		if (clear_received_props(tofs, props, NULL) != 0) {
4294 			/*
4295 			 * We failed to clear the received properties.
4296 			 * Since we may have left a $recvd value on the
4297 			 * system, we can't clear the $hasrecvd flag.
4298 			 */
4299 			zc->zc_obj |= ZPROP_ERR_NORESTORE;
4300 		} else if (first_recvd_props) {
4301 			dsl_prop_unset_hasrecvd(tofs);
4302 		}
4303 
4304 		if (origprops == NULL && !drc.drc_newfs) {
4305 			/* We failed to stash the original properties. */
4306 			zc->zc_obj |= ZPROP_ERR_NORESTORE;
4307 		}
4308 
4309 		/*
4310 		 * dsl_props_set() will not convert RECEIVED to LOCAL on or
4311 		 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4312 		 * explictly if we're restoring local properties cleared in the
4313 		 * first new-style receive.
4314 		 */
4315 		if (origprops != NULL &&
4316 		    zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4317 		    ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4318 		    origprops, NULL) != 0) {
4319 			/*
4320 			 * We stashed the original properties but failed to
4321 			 * restore them.
4322 			 */
4323 			zc->zc_obj |= ZPROP_ERR_NORESTORE;
4324 		}
4325 	}
4326 out:
4327 	nvlist_free(props);
4328 	nvlist_free(origprops);
4329 	nvlist_free(errors);
4330 	releasef(fd);
4331 
4332 	if (error == 0)
4333 		error = props_error;
4334 
4335 	return (error);
4336 }
4337 
4338 /*
4339  * inputs:
4340  * zc_name	name of snapshot to send
4341  * zc_cookie	file descriptor to send stream to
4342  * zc_obj	fromorigin flag (mutually exclusive with zc_fromobj)
4343  * zc_sendobj	objsetid of snapshot to send
4344  * zc_fromobj	objsetid of incremental fromsnap (may be zero)
4345  * zc_guid	if set, estimate size of stream only.  zc_cookie is ignored.
4346  *		output size in zc_objset_type.
4347  * zc_flags	lzc_send_flags
4348  *
4349  * outputs:
4350  * zc_objset_type	estimated size, if zc_guid is set
4351  */
4352 static int
4353 zfs_ioc_send(zfs_cmd_t *zc)
4354 {
4355 	int error;
4356 	offset_t off;
4357 	boolean_t estimate = (zc->zc_guid != 0);
4358 	boolean_t embedok = (zc->zc_flags & 0x1);
4359 	boolean_t large_block_ok = (zc->zc_flags & 0x2);
4360 
4361 	if (zc->zc_obj != 0) {
4362 		dsl_pool_t *dp;
4363 		dsl_dataset_t *tosnap;
4364 
4365 		error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4366 		if (error != 0)
4367 			return (error);
4368 
4369 		error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4370 		if (error != 0) {
4371 			dsl_pool_rele(dp, FTAG);
4372 			return (error);
4373 		}
4374 
4375 		if (dsl_dir_is_clone(tosnap->ds_dir))
4376 			zc->zc_fromobj =
4377 			    dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj;
4378 		dsl_dataset_rele(tosnap, FTAG);
4379 		dsl_pool_rele(dp, FTAG);
4380 	}
4381 
4382 	if (estimate) {
4383 		dsl_pool_t *dp;
4384 		dsl_dataset_t *tosnap;
4385 		dsl_dataset_t *fromsnap = NULL;
4386 
4387 		error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4388 		if (error != 0)
4389 			return (error);
4390 
4391 		error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
4392 		if (error != 0) {
4393 			dsl_pool_rele(dp, FTAG);
4394 			return (error);
4395 		}
4396 
4397 		if (zc->zc_fromobj != 0) {
4398 			error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
4399 			    FTAG, &fromsnap);
4400 			if (error != 0) {
4401 				dsl_dataset_rele(tosnap, FTAG);
4402 				dsl_pool_rele(dp, FTAG);
4403 				return (error);
4404 			}
4405 		}
4406 
4407 		error = dmu_send_estimate(tosnap, fromsnap,
4408 		    &zc->zc_objset_type);
4409 
4410 		if (fromsnap != NULL)
4411 			dsl_dataset_rele(fromsnap, FTAG);
4412 		dsl_dataset_rele(tosnap, FTAG);
4413 		dsl_pool_rele(dp, FTAG);
4414 	} else {
4415 		file_t *fp = getf(zc->zc_cookie);
4416 		if (fp == NULL)
4417 			return (SET_ERROR(EBADF));
4418 
4419 		off = fp->f_offset;
4420 		error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
4421 		    zc->zc_fromobj, embedok, large_block_ok,
4422 		    zc->zc_cookie, fp->f_vnode, &off);
4423 
4424 		if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
4425 			fp->f_offset = off;
4426 		releasef(zc->zc_cookie);
4427 	}
4428 	return (error);
4429 }
4430 
4431 /*
4432  * inputs:
4433  * zc_name	name of snapshot on which to report progress
4434  * zc_cookie	file descriptor of send stream
4435  *
4436  * outputs:
4437  * zc_cookie	number of bytes written in send stream thus far
4438  */
4439 static int
4440 zfs_ioc_send_progress(zfs_cmd_t *zc)
4441 {
4442 	dsl_pool_t *dp;
4443 	dsl_dataset_t *ds;
4444 	dmu_sendarg_t *dsp = NULL;
4445 	int error;
4446 
4447 	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
4448 	if (error != 0)
4449 		return (error);
4450 
4451 	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
4452 	if (error != 0) {
4453 		dsl_pool_rele(dp, FTAG);
4454 		return (error);
4455 	}
4456 
4457 	mutex_enter(&ds->ds_sendstream_lock);
4458 
4459 	/*
4460 	 * Iterate over all the send streams currently active on this dataset.
4461 	 * If there's one which matches the specified file descriptor _and_ the
4462 	 * stream was started by the current process, return the progress of
4463 	 * that stream.
4464 	 */
4465 	for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
4466 	    dsp = list_next(&ds->ds_sendstreams, dsp)) {
4467 		if (dsp->dsa_outfd == zc->zc_cookie &&
4468 		    dsp->dsa_proc == curproc)
4469 			break;
4470 	}
4471 
4472 	if (dsp != NULL)
4473 		zc->zc_cookie = *(dsp->dsa_off);
4474 	else
4475 		error = SET_ERROR(ENOENT);
4476 
4477 	mutex_exit(&ds->ds_sendstream_lock);
4478 	dsl_dataset_rele(ds, FTAG);
4479 	dsl_pool_rele(dp, FTAG);
4480 	return (error);
4481 }
4482 
4483 static int
4484 zfs_ioc_inject_fault(zfs_cmd_t *zc)
4485 {
4486 	int id, error;
4487 
4488 	error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
4489 	    &zc->zc_inject_record);
4490 
4491 	if (error == 0)
4492 		zc->zc_guid = (uint64_t)id;
4493 
4494 	return (error);
4495 }
4496 
4497 static int
4498 zfs_ioc_clear_fault(zfs_cmd_t *zc)
4499 {
4500 	return (zio_clear_fault((int)zc->zc_guid));
4501 }
4502 
4503 static int
4504 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
4505 {
4506 	int id = (int)zc->zc_guid;
4507 	int error;
4508 
4509 	error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
4510 	    &zc->zc_inject_record);
4511 
4512 	zc->zc_guid = id;
4513 
4514 	return (error);
4515 }
4516 
4517 static int
4518 zfs_ioc_error_log(zfs_cmd_t *zc)
4519 {
4520 	spa_t *spa;
4521 	int error;
4522 	size_t count = (size_t)zc->zc_nvlist_dst_size;
4523 
4524 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
4525 		return (error);
4526 
4527 	error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
4528 	    &count);
4529 	if (error == 0)
4530 		zc->zc_nvlist_dst_size = count;
4531 	else
4532 		zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
4533 
4534 	spa_close(spa, FTAG);
4535 
4536 	return (error);
4537 }
4538 
4539 static int
4540 zfs_ioc_clear(zfs_cmd_t *zc)
4541 {
4542 	spa_t *spa;
4543 	vdev_t *vd;
4544 	int error;
4545 
4546 	/*
4547 	 * On zpool clear we also fix up missing slogs
4548 	 */
4549 	mutex_enter(&spa_namespace_lock);
4550 	spa = spa_lookup(zc->zc_name);
4551 	if (spa == NULL) {
4552 		mutex_exit(&spa_namespace_lock);
4553 		return (SET_ERROR(EIO));
4554 	}
4555 	if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
4556 		/* we need to let spa_open/spa_load clear the chains */
4557 		spa_set_log_state(spa, SPA_LOG_CLEAR);
4558 	}
4559 	spa->spa_last_open_failed = 0;
4560 	mutex_exit(&spa_namespace_lock);
4561 
4562 	if (zc->zc_cookie & ZPOOL_NO_REWIND) {
4563 		error = spa_open(zc->zc_name, &spa, FTAG);
4564 	} else {
4565 		nvlist_t *policy;
4566 		nvlist_t *config = NULL;
4567 
4568 		if (zc->zc_nvlist_src == NULL)
4569 			return (SET_ERROR(EINVAL));
4570 
4571 		if ((error = get_nvlist(zc->zc_nvlist_src,
4572 		    zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
4573 			error = spa_open_rewind(zc->zc_name, &spa, FTAG,
4574 			    policy, &config);
4575 			if (config != NULL) {
4576 				int err;
4577 
4578 				if ((err = put_nvlist(zc, config)) != 0)
4579 					error = err;
4580 				nvlist_free(config);
4581 			}
4582 			nvlist_free(policy);
4583 		}
4584 	}
4585 
4586 	if (error != 0)
4587 		return (error);
4588 
4589 	spa_vdev_state_enter(spa, SCL_NONE);
4590 
4591 	if (zc->zc_guid == 0) {
4592 		vd = NULL;
4593 	} else {
4594 		vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
4595 		if (vd == NULL) {
4596 			(void) spa_vdev_state_exit(spa, NULL, ENODEV);
4597 			spa_close(spa, FTAG);
4598 			return (SET_ERROR(ENODEV));
4599 		}
4600 	}
4601 
4602 	vdev_clear(spa, vd);
4603 
4604 	(void) spa_vdev_state_exit(spa, NULL, 0);
4605 
4606 	/*
4607 	 * Resume any suspended I/Os.
4608 	 */
4609 	if (zio_resume(spa) != 0)
4610 		error = SET_ERROR(EIO);
4611 
4612 	spa_close(spa, FTAG);
4613 
4614 	return (error);
4615 }
4616 
4617 static int
4618 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
4619 {
4620 	spa_t *spa;
4621 	int error;
4622 
4623 	error = spa_open(zc->zc_name, &spa, FTAG);
4624 	if (error != 0)
4625 		return (error);
4626 
4627 	spa_vdev_state_enter(spa, SCL_NONE);
4628 
4629 	/*
4630 	 * If a resilver is already in progress then set the
4631 	 * spa_scrub_reopen flag to B_TRUE so that we don't restart
4632 	 * the scan as a side effect of the reopen. Otherwise, let
4633 	 * vdev_open() decided if a resilver is required.
4634 	 */
4635 	spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
4636 	vdev_reopen(spa->spa_root_vdev);
4637 	spa->spa_scrub_reopen = B_FALSE;
4638 
4639 	(void) spa_vdev_state_exit(spa, NULL, 0);
4640 	spa_close(spa, FTAG);
4641 	return (0);
4642 }
4643 /*
4644  * inputs:
4645  * zc_name	name of filesystem
4646  * zc_value	name of origin snapshot
4647  *
4648  * outputs:
4649  * zc_string	name of conflicting snapshot, if there is one
4650  */
4651 static int
4652 zfs_ioc_promote(zfs_cmd_t *zc)
4653 {
4654 	char *cp;
4655 
4656 	/*
4657 	 * We don't need to unmount *all* the origin fs's snapshots, but
4658 	 * it's easier.
4659 	 */
4660 	cp = strchr(zc->zc_value, '@');
4661 	if (cp)
4662 		*cp = '\0';
4663 	(void) dmu_objset_find(zc->zc_value,
4664 	    zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
4665 	return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
4666 }
4667 
4668 /*
4669  * Retrieve a single {user|group}{used|quota}@... property.
4670  *
4671  * inputs:
4672  * zc_name	name of filesystem
4673  * zc_objset_type zfs_userquota_prop_t
4674  * zc_value	domain name (eg. "S-1-234-567-89")
4675  * zc_guid	RID/UID/GID
4676  *
4677  * outputs:
4678  * zc_cookie	property value
4679  */
4680 static int
4681 zfs_ioc_userspace_one(zfs_cmd_t *zc)
4682 {
4683 	zfsvfs_t *zfsvfs;
4684 	int error;
4685 
4686 	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
4687 		return (SET_ERROR(EINVAL));
4688 
4689 	error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4690 	if (error != 0)
4691 		return (error);
4692 
4693 	error = zfs_userspace_one(zfsvfs,
4694 	    zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
4695 	zfsvfs_rele(zfsvfs, FTAG);
4696 
4697 	return (error);
4698 }
4699 
4700 /*
4701  * inputs:
4702  * zc_name		name of filesystem
4703  * zc_cookie		zap cursor
4704  * zc_objset_type	zfs_userquota_prop_t
4705  * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
4706  *
4707  * outputs:
4708  * zc_nvlist_dst[_size]	data buffer (array of zfs_useracct_t)
4709  * zc_cookie	zap cursor
4710  */
4711 static int
4712 zfs_ioc_userspace_many(zfs_cmd_t *zc)
4713 {
4714 	zfsvfs_t *zfsvfs;
4715 	int bufsize = zc->zc_nvlist_dst_size;
4716 
4717 	if (bufsize <= 0)
4718 		return (SET_ERROR(ENOMEM));
4719 
4720 	int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
4721 	if (error != 0)
4722 		return (error);
4723 
4724 	void *buf = kmem_alloc(bufsize, KM_SLEEP);
4725 
4726 	error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
4727 	    buf, &zc->zc_nvlist_dst_size);
4728 
4729 	if (error == 0) {
4730 		error = xcopyout(buf,
4731 		    (void *)(uintptr_t)zc->zc_nvlist_dst,
4732 		    zc->zc_nvlist_dst_size);
4733 	}
4734 	kmem_free(buf, bufsize);
4735 	zfsvfs_rele(zfsvfs, FTAG);
4736 
4737 	return (error);
4738 }
4739 
4740 /*
4741  * inputs:
4742  * zc_name		name of filesystem
4743  *
4744  * outputs:
4745  * none
4746  */
4747 static int
4748 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
4749 {
4750 	objset_t *os;
4751 	int error = 0;
4752 	zfsvfs_t *zfsvfs;
4753 
4754 	if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
4755 		if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
4756 			/*
4757 			 * If userused is not enabled, it may be because the
4758 			 * objset needs to be closed & reopened (to grow the
4759 			 * objset_phys_t).  Suspend/resume the fs will do that.
4760 			 */
4761 			error = zfs_suspend_fs(zfsvfs);
4762 			if (error == 0) {
4763 				dmu_objset_refresh_ownership(zfsvfs->z_os,
4764 				    zfsvfs);
4765 				error = zfs_resume_fs(zfsvfs, zc->zc_name);
4766 			}
4767 		}
4768 		if (error == 0)
4769 			error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
4770 		VFS_RELE(zfsvfs->z_vfs);
4771 	} else {
4772 		/* XXX kind of reading contents without owning */
4773 		error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4774 		if (error != 0)
4775 			return (error);
4776 
4777 		error = dmu_objset_userspace_upgrade(os);
4778 		dmu_objset_rele(os, FTAG);
4779 	}
4780 
4781 	return (error);
4782 }
4783 
4784 /*
4785  * We don't want to have a hard dependency
4786  * against some special symbols in sharefs
4787  * nfs, and smbsrv.  Determine them if needed when
4788  * the first file system is shared.
4789  * Neither sharefs, nfs or smbsrv are unloadable modules.
4790  */
4791 int (*znfsexport_fs)(void *arg);
4792 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
4793 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
4794 
4795 int zfs_nfsshare_inited;
4796 int zfs_smbshare_inited;
4797 
4798 ddi_modhandle_t nfs_mod;
4799 ddi_modhandle_t sharefs_mod;
4800 ddi_modhandle_t smbsrv_mod;
4801 kmutex_t zfs_share_lock;
4802 
4803 static int
4804 zfs_init_sharefs()
4805 {
4806 	int error;
4807 
4808 	ASSERT(MUTEX_HELD(&zfs_share_lock));
4809 	/* Both NFS and SMB shares also require sharetab support. */
4810 	if (sharefs_mod == NULL && ((sharefs_mod =
4811 	    ddi_modopen("fs/sharefs",
4812 	    KRTLD_MODE_FIRST, &error)) == NULL)) {
4813 		return (SET_ERROR(ENOSYS));
4814 	}
4815 	if (zshare_fs == NULL && ((zshare_fs =
4816 	    (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
4817 	    ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
4818 		return (SET_ERROR(ENOSYS));
4819 	}
4820 	return (0);
4821 }
4822 
4823 static int
4824 zfs_ioc_share(zfs_cmd_t *zc)
4825 {
4826 	int error;
4827 	int opcode;
4828 
4829 	switch (zc->zc_share.z_sharetype) {
4830 	case ZFS_SHARE_NFS:
4831 	case ZFS_UNSHARE_NFS:
4832 		if (zfs_nfsshare_inited == 0) {
4833 			mutex_enter(&zfs_share_lock);
4834 			if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
4835 			    KRTLD_MODE_FIRST, &error)) == NULL)) {
4836 				mutex_exit(&zfs_share_lock);
4837 				return (SET_ERROR(ENOSYS));
4838 			}
4839 			if (znfsexport_fs == NULL &&
4840 			    ((znfsexport_fs = (int (*)(void *))
4841 			    ddi_modsym(nfs_mod,
4842 			    "nfs_export", &error)) == NULL)) {
4843 				mutex_exit(&zfs_share_lock);
4844 				return (SET_ERROR(ENOSYS));
4845 			}
4846 			error = zfs_init_sharefs();
4847 			if (error != 0) {
4848 				mutex_exit(&zfs_share_lock);
4849 				return (SET_ERROR(ENOSYS));
4850 			}
4851 			zfs_nfsshare_inited = 1;
4852 			mutex_exit(&zfs_share_lock);
4853 		}
4854 		break;
4855 	case ZFS_SHARE_SMB:
4856 	case ZFS_UNSHARE_SMB:
4857 		if (zfs_smbshare_inited == 0) {
4858 			mutex_enter(&zfs_share_lock);
4859 			if (smbsrv_mod == NULL && ((smbsrv_mod =
4860 			    ddi_modopen("drv/smbsrv",
4861 			    KRTLD_MODE_FIRST, &error)) == NULL)) {
4862 				mutex_exit(&zfs_share_lock);
4863 				return (SET_ERROR(ENOSYS));
4864 			}
4865 			if (zsmbexport_fs == NULL && ((zsmbexport_fs =
4866 			    (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
4867 			    "smb_server_share", &error)) == NULL)) {
4868 				mutex_exit(&zfs_share_lock);
4869 				return (SET_ERROR(ENOSYS));
4870 			}
4871 			error = zfs_init_sharefs();
4872 			if (error != 0) {
4873 				mutex_exit(&zfs_share_lock);
4874 				return (SET_ERROR(ENOSYS));
4875 			}
4876 			zfs_smbshare_inited = 1;
4877 			mutex_exit(&zfs_share_lock);
4878 		}
4879 		break;
4880 	default:
4881 		return (SET_ERROR(EINVAL));
4882 	}
4883 
4884 	switch (zc->zc_share.z_sharetype) {
4885 	case ZFS_SHARE_NFS:
4886 	case ZFS_UNSHARE_NFS:
4887 		if (error =
4888 		    znfsexport_fs((void *)
4889 		    (uintptr_t)zc->zc_share.z_exportdata))
4890 			return (error);
4891 		break;
4892 	case ZFS_SHARE_SMB:
4893 	case ZFS_UNSHARE_SMB:
4894 		if (error = zsmbexport_fs((void *)
4895 		    (uintptr_t)zc->zc_share.z_exportdata,
4896 		    zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
4897 		    B_TRUE: B_FALSE)) {
4898 			return (error);
4899 		}
4900 		break;
4901 	}
4902 
4903 	opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
4904 	    zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
4905 	    SHAREFS_ADD : SHAREFS_REMOVE;
4906 
4907 	/*
4908 	 * Add or remove share from sharetab
4909 	 */
4910 	error = zshare_fs(opcode,
4911 	    (void *)(uintptr_t)zc->zc_share.z_sharedata,
4912 	    zc->zc_share.z_sharemax);
4913 
4914 	return (error);
4915 
4916 }
4917 
4918 ace_t full_access[] = {
4919 	{(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4920 };
4921 
4922 /*
4923  * inputs:
4924  * zc_name		name of containing filesystem
4925  * zc_obj		object # beyond which we want next in-use object #
4926  *
4927  * outputs:
4928  * zc_obj		next in-use object #
4929  */
4930 static int
4931 zfs_ioc_next_obj(zfs_cmd_t *zc)
4932 {
4933 	objset_t *os = NULL;
4934 	int error;
4935 
4936 	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
4937 	if (error != 0)
4938 		return (error);
4939 
4940 	error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
4941 	    dsl_dataset_phys(os->os_dsl_dataset)->ds_prev_snap_txg);
4942 
4943 	dmu_objset_rele(os, FTAG);
4944 	return (error);
4945 }
4946 
4947 /*
4948  * inputs:
4949  * zc_name		name of filesystem
4950  * zc_value		prefix name for snapshot
4951  * zc_cleanup_fd	cleanup-on-exit file descriptor for calling process
4952  *
4953  * outputs:
4954  * zc_value		short name of new snapshot
4955  */
4956 static int
4957 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
4958 {
4959 	char *snap_name;
4960 	char *hold_name;
4961 	int error;
4962 	minor_t minor;
4963 
4964 	error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
4965 	if (error != 0)
4966 		return (error);
4967 
4968 	snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
4969 	    (u_longlong_t)ddi_get_lbolt64());
4970 	hold_name = kmem_asprintf("%%%s", zc->zc_value);
4971 
4972 	error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
4973 	    hold_name);
4974 	if (error == 0)
4975 		(void) strcpy(zc->zc_value, snap_name);
4976 	strfree(snap_name);
4977 	strfree(hold_name);
4978 	zfs_onexit_fd_rele(zc->zc_cleanup_fd);
4979 	return (error);
4980 }
4981 
4982 /*
4983  * inputs:
4984  * zc_name		name of "to" snapshot
4985  * zc_value		name of "from" snapshot
4986  * zc_cookie		file descriptor to write diff data on
4987  *
4988  * outputs:
4989  * dmu_diff_record_t's to the file descriptor
4990  */
4991 static int
4992 zfs_ioc_diff(zfs_cmd_t *zc)
4993 {
4994 	file_t *fp;
4995 	offset_t off;
4996 	int error;
4997 
4998 	fp = getf(zc->zc_cookie);
4999 	if (fp == NULL)
5000 		return (SET_ERROR(EBADF));
5001 
5002 	off = fp->f_offset;
5003 
5004 	error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
5005 
5006 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5007 		fp->f_offset = off;
5008 	releasef(zc->zc_cookie);
5009 
5010 	return (error);
5011 }
5012 
5013 /*
5014  * Remove all ACL files in shares dir
5015  */
5016 static int
5017 zfs_smb_acl_purge(znode_t *dzp)
5018 {
5019 	zap_cursor_t	zc;
5020 	zap_attribute_t	zap;
5021 	zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
5022 	int error;
5023 
5024 	for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
5025 	    (error = zap_cursor_retrieve(&zc, &zap)) == 0;
5026 	    zap_cursor_advance(&zc)) {
5027 		if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
5028 		    NULL, 0)) != 0)
5029 			break;
5030 	}
5031 	zap_cursor_fini(&zc);
5032 	return (error);
5033 }
5034 
5035 static int
5036 zfs_ioc_smb_acl(zfs_cmd_t *zc)
5037 {
5038 	vnode_t *vp;
5039 	znode_t *dzp;
5040 	vnode_t *resourcevp = NULL;
5041 	znode_t *sharedir;
5042 	zfsvfs_t *zfsvfs;
5043 	nvlist_t *nvlist;
5044 	char *src, *target;
5045 	vattr_t vattr;
5046 	vsecattr_t vsec;
5047 	int error = 0;
5048 
5049 	if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
5050 	    NO_FOLLOW, NULL, &vp)) != 0)
5051 		return (error);
5052 
5053 	/* Now make sure mntpnt and dataset are ZFS */
5054 
5055 	if (vp->v_vfsp->vfs_fstype != zfsfstype ||
5056 	    (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
5057 	    zc->zc_name) != 0)) {
5058 		VN_RELE(vp);
5059 		return (SET_ERROR(EINVAL));
5060 	}
5061 
5062 	dzp = VTOZ(vp);
5063 	zfsvfs = dzp->z_zfsvfs;
5064 	ZFS_ENTER(zfsvfs);
5065 
5066 	/*
5067 	 * Create share dir if its missing.
5068 	 */
5069 	mutex_enter(&zfsvfs->z_lock);
5070 	if (zfsvfs->z_shares_dir == 0) {
5071 		dmu_tx_t *tx;
5072 
5073 		tx = dmu_tx_create(zfsvfs->z_os);
5074 		dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
5075 		    ZFS_SHARES_DIR);
5076 		dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
5077 		error = dmu_tx_assign(tx, TXG_WAIT);
5078 		if (error != 0) {
5079 			dmu_tx_abort(tx);
5080 		} else {
5081 			error = zfs_create_share_dir(zfsvfs, tx);
5082 			dmu_tx_commit(tx);
5083 		}
5084 		if (error != 0) {
5085 			mutex_exit(&zfsvfs->z_lock);
5086 			VN_RELE(vp);
5087 			ZFS_EXIT(zfsvfs);
5088 			return (error);
5089 		}
5090 	}
5091 	mutex_exit(&zfsvfs->z_lock);
5092 
5093 	ASSERT(zfsvfs->z_shares_dir);
5094 	if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
5095 		VN_RELE(vp);
5096 		ZFS_EXIT(zfsvfs);
5097 		return (error);
5098 	}
5099 
5100 	switch (zc->zc_cookie) {
5101 	case ZFS_SMB_ACL_ADD:
5102 		vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
5103 		vattr.va_type = VREG;
5104 		vattr.va_mode = S_IFREG|0777;
5105 		vattr.va_uid = 0;
5106 		vattr.va_gid = 0;
5107 
5108 		vsec.vsa_mask = VSA_ACE;
5109 		vsec.vsa_aclentp = &full_access;
5110 		vsec.vsa_aclentsz = sizeof (full_access);
5111 		vsec.vsa_aclcnt = 1;
5112 
5113 		error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
5114 		    &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
5115 		if (resourcevp)
5116 			VN_RELE(resourcevp);
5117 		break;
5118 
5119 	case ZFS_SMB_ACL_REMOVE:
5120 		error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
5121 		    NULL, 0);
5122 		break;
5123 
5124 	case ZFS_SMB_ACL_RENAME:
5125 		if ((error = get_nvlist(zc->zc_nvlist_src,
5126 		    zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
5127 			VN_RELE(vp);
5128 			ZFS_EXIT(zfsvfs);
5129 			return (error);
5130 		}
5131 		if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
5132 		    nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
5133 		    &target)) {
5134 			VN_RELE(vp);
5135 			VN_RELE(ZTOV(sharedir));
5136 			ZFS_EXIT(zfsvfs);
5137 			nvlist_free(nvlist);
5138 			return (error);
5139 		}
5140 		error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
5141 		    kcred, NULL, 0);
5142 		nvlist_free(nvlist);
5143 		break;
5144 
5145 	case ZFS_SMB_ACL_PURGE:
5146 		error = zfs_smb_acl_purge(sharedir);
5147 		break;
5148 
5149 	default:
5150 		error = SET_ERROR(EINVAL);
5151 		break;
5152 	}
5153 
5154 	VN_RELE(vp);
5155 	VN_RELE(ZTOV(sharedir));
5156 
5157 	ZFS_EXIT(zfsvfs);
5158 
5159 	return (error);
5160 }
5161 
5162 /*
5163  * innvl: {
5164  *     "holds" -> { snapname -> holdname (string), ... }
5165  *     (optional) "cleanup_fd" -> fd (int32)
5166  * }
5167  *
5168  * outnvl: {
5169  *     snapname -> error value (int32)
5170  *     ...
5171  * }
5172  */
5173 /* ARGSUSED */
5174 static int
5175 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
5176 {
5177 	nvpair_t *pair;
5178 	nvlist_t *holds;
5179 	int cleanup_fd = -1;
5180 	int error;
5181 	minor_t minor = 0;
5182 
5183 	error = nvlist_lookup_nvlist(args, "holds", &holds);
5184 	if (error != 0)
5185 		return (SET_ERROR(EINVAL));
5186 
5187 	/* make sure the user didn't pass us any invalid (empty) tags */
5188 	for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
5189 	    pair = nvlist_next_nvpair(holds, pair)) {
5190 		char *htag;
5191 
5192 		error = nvpair_value_string(pair, &htag);
5193 		if (error != 0)
5194 			return (SET_ERROR(error));
5195 
5196 		if (strlen(htag) == 0)
5197 			return (SET_ERROR(EINVAL));
5198 	}
5199 
5200 	if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
5201 		error = zfs_onexit_fd_hold(cleanup_fd, &minor);
5202 		if (error != 0)
5203 			return (error);
5204 	}
5205 
5206 	error = dsl_dataset_user_hold(holds, minor, errlist);
5207 	if (minor != 0)
5208 		zfs_onexit_fd_rele(cleanup_fd);
5209 	return (error);
5210 }
5211 
5212 /*
5213  * innvl is not used.
5214  *
5215  * outnvl: {
5216  *    holdname -> time added (uint64 seconds since epoch)
5217  *    ...
5218  * }
5219  */
5220 /* ARGSUSED */
5221 static int
5222 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
5223 {
5224 	return (dsl_dataset_get_holds(snapname, outnvl));
5225 }
5226 
5227 /*
5228  * innvl: {
5229  *     snapname -> { holdname, ... }
5230  *     ...
5231  * }
5232  *
5233  * outnvl: {
5234  *     snapname -> error value (int32)
5235  *     ...
5236  * }
5237  */
5238 /* ARGSUSED */
5239 static int
5240 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
5241 {
5242 	return (dsl_dataset_user_release(holds, errlist));
5243 }
5244 
5245 /*
5246  * inputs:
5247  * zc_name		name of new filesystem or snapshot
5248  * zc_value		full name of old snapshot
5249  *
5250  * outputs:
5251  * zc_cookie		space in bytes
5252  * zc_objset_type	compressed space in bytes
5253  * zc_perm_action	uncompressed space in bytes
5254  */
5255 static int
5256 zfs_ioc_space_written(zfs_cmd_t *zc)
5257 {
5258 	int error;
5259 	dsl_pool_t *dp;
5260 	dsl_dataset_t *new, *old;
5261 
5262 	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5263 	if (error != 0)
5264 		return (error);
5265 	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
5266 	if (error != 0) {
5267 		dsl_pool_rele(dp, FTAG);
5268 		return (error);
5269 	}
5270 	error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
5271 	if (error != 0) {
5272 		dsl_dataset_rele(new, FTAG);
5273 		dsl_pool_rele(dp, FTAG);
5274 		return (error);
5275 	}
5276 
5277 	error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
5278 	    &zc->zc_objset_type, &zc->zc_perm_action);
5279 	dsl_dataset_rele(old, FTAG);
5280 	dsl_dataset_rele(new, FTAG);
5281 	dsl_pool_rele(dp, FTAG);
5282 	return (error);
5283 }
5284 
5285 /*
5286  * innvl: {
5287  *     "firstsnap" -> snapshot name
5288  * }
5289  *
5290  * outnvl: {
5291  *     "used" -> space in bytes
5292  *     "compressed" -> compressed space in bytes
5293  *     "uncompressed" -> uncompressed space in bytes
5294  * }
5295  */
5296 static int
5297 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
5298 {
5299 	int error;
5300 	dsl_pool_t *dp;
5301 	dsl_dataset_t *new, *old;
5302 	char *firstsnap;
5303 	uint64_t used, comp, uncomp;
5304 
5305 	if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
5306 		return (SET_ERROR(EINVAL));
5307 
5308 	error = dsl_pool_hold(lastsnap, FTAG, &dp);
5309 	if (error != 0)
5310 		return (error);
5311 
5312 	error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
5313 	if (error == 0 && !new->ds_is_snapshot) {
5314 		dsl_dataset_rele(new, FTAG);
5315 		error = SET_ERROR(EINVAL);
5316 	}
5317 	if (error != 0) {
5318 		dsl_pool_rele(dp, FTAG);
5319 		return (error);
5320 	}
5321 	error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
5322 	if (error == 0 && !old->ds_is_snapshot) {
5323 		dsl_dataset_rele(old, FTAG);
5324 		error = SET_ERROR(EINVAL);
5325 	}
5326 	if (error != 0) {
5327 		dsl_dataset_rele(new, FTAG);
5328 		dsl_pool_rele(dp, FTAG);
5329 		return (error);
5330 	}
5331 
5332 	error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
5333 	dsl_dataset_rele(old, FTAG);
5334 	dsl_dataset_rele(new, FTAG);
5335 	dsl_pool_rele(dp, FTAG);
5336 	fnvlist_add_uint64(outnvl, "used", used);
5337 	fnvlist_add_uint64(outnvl, "compressed", comp);
5338 	fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
5339 	return (error);
5340 }
5341 
5342 /*
5343  * innvl: {
5344  *     "fd" -> file descriptor to write stream to (int32)
5345  *     (optional) "fromsnap" -> full snap name to send an incremental from
5346  *     (optional) "largeblockok" -> (value ignored)
5347  *         indicates that blocks > 128KB are permitted
5348  *     (optional) "embedok" -> (value ignored)
5349  *         presence indicates DRR_WRITE_EMBEDDED records are permitted
5350  *     (optional) "resume_object" and "resume_offset" -> (uint64)
5351  *         if present, resume send stream from specified object and offset.
5352  * }
5353  *
5354  * outnvl is unused
5355  */
5356 /* ARGSUSED */
5357 static int
5358 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5359 {
5360 	int error;
5361 	offset_t off;
5362 	char *fromname = NULL;
5363 	int fd;
5364 	boolean_t largeblockok;
5365 	boolean_t embedok;
5366 	uint64_t resumeobj = 0;
5367 	uint64_t resumeoff = 0;
5368 
5369 	error = nvlist_lookup_int32(innvl, "fd", &fd);
5370 	if (error != 0)
5371 		return (SET_ERROR(EINVAL));
5372 
5373 	(void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
5374 
5375 	largeblockok = nvlist_exists(innvl, "largeblockok");
5376 	embedok = nvlist_exists(innvl, "embedok");
5377 
5378 	(void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
5379 	(void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);
5380 
5381 	file_t *fp = getf(fd);
5382 	if (fp == NULL)
5383 		return (SET_ERROR(EBADF));
5384 
5385 	off = fp->f_offset;
5386 	error = dmu_send(snapname, fromname, embedok, largeblockok, fd,
5387 	    resumeobj, resumeoff, fp->f_vnode, &off);
5388 
5389 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5390 		fp->f_offset = off;
5391 	releasef(fd);
5392 	return (error);
5393 }
5394 
5395 /*
5396  * Determine approximately how large a zfs send stream will be -- the number
5397  * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
5398  *
5399  * innvl: {
5400  *     (optional) "from" -> full snap or bookmark name to send an incremental
5401  *                          from
5402  * }
5403  *
5404  * outnvl: {
5405  *     "space" -> bytes of space (uint64)
5406  * }
5407  */
5408 static int
5409 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
5410 {
5411 	dsl_pool_t *dp;
5412 	dsl_dataset_t *tosnap;
5413 	int error;
5414 	char *fromname;
5415 	uint64_t space;
5416 
5417 	error = dsl_pool_hold(snapname, FTAG, &dp);
5418 	if (error != 0)
5419 		return (error);
5420 
5421 	error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
5422 	if (error != 0) {
5423 		dsl_pool_rele(dp, FTAG);
5424 		return (error);
5425 	}
5426 
5427 	error = nvlist_lookup_string(innvl, "from", &fromname);
5428 	if (error == 0) {
5429 		if (strchr(fromname, '@') != NULL) {
5430 			/*
5431 			 * If from is a snapshot, hold it and use the more
5432 			 * efficient dmu_send_estimate to estimate send space
5433 			 * size using deadlists.
5434 			 */
5435 			dsl_dataset_t *fromsnap;
5436 			error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
5437 			if (error != 0)
5438 				goto out;
5439 			error = dmu_send_estimate(tosnap, fromsnap, &space);
5440 			dsl_dataset_rele(fromsnap, FTAG);
5441 		} else if (strchr(fromname, '#') != NULL) {
5442 			/*
5443 			 * If from is a bookmark, fetch the creation TXG of the
5444 			 * snapshot it was created from and use that to find
5445 			 * blocks that were born after it.
5446 			 */
5447 			zfs_bookmark_phys_t frombm;
5448 
5449 			error = dsl_bookmark_lookup(dp, fromname, tosnap,
5450 			    &frombm);
5451 			if (error != 0)
5452 				goto out;
5453 			error = dmu_send_estimate_from_txg(tosnap,
5454 			    frombm.zbm_creation_txg, &space);
5455 		} else {
5456 			/*
5457 			 * from is not properly formatted as a snapshot or
5458 			 * bookmark
5459 			 */
5460 			error = SET_ERROR(EINVAL);
5461 			goto out;
5462 		}
5463 	} else {
5464 		// If estimating the size of a full send, use dmu_send_estimate
5465 		error = dmu_send_estimate(tosnap, NULL, &space);
5466 	}
5467 
5468 	fnvlist_add_uint64(outnvl, "space", space);
5469 
5470 out:
5471 	dsl_dataset_rele(tosnap, FTAG);
5472 	dsl_pool_rele(dp, FTAG);
5473 	return (error);
5474 }
5475 
5476 static int
5477 zfs_ioc_set_zev_callbacks(const char *unused, nvlist_t *innvl,
5478     nvlist_t *outnvl)
5479 {
5480 	int error;
5481 	uint64_t cb_addr;
5482 	/*
5483 	 * Our secpolicy for this op makes sure it's called in
5484 	 * kernel context, and that no other callbacks have
5485 	 * been registered, yet.
5486 	 */
5487 	error = nvlist_lookup_uint64(innvl, "callbacks", &cb_addr);
5488 	if (error != 0) {
5489 		cmn_err(CE_WARN, "set_zev_callbacks nvlist lookup failed (%d)",
5490 		    error);
5491 		return (error);
5492 	}
5493 	/* cb_addr is always a kernel memory address */
5494 	rw_enter(&rz_zev_rwlock, RW_WRITER);
5495 	if (rz_zev_callbacks != rz_zev_default_callbacks) {
5496 		rw_exit(&rz_zev_rwlock);
5497 		return (EBUSY);
5498 	}
5499 	rz_zev_callbacks = (void *)(uintptr_t)cb_addr;
5500 	rw_exit(&rz_zev_rwlock);
5501 	return (0);
5502 }
5503 
5504 static int
5505 zfs_ioc_unset_zev_callbacks(const char *unused, nvlist_t *innvl,
5506     nvlist_t *outnvl)
5507 {
5508 	/*
5509 	 * Our secpolicy for this op makes sure it's called in
5510 	 * kernel context.
5511 	 */
5512 	rw_enter(&rz_zev_rwlock, RW_WRITER);
5513 	rz_zev_callbacks = rz_zev_default_callbacks;
5514 	rw_exit(&rz_zev_rwlock);
5515 	/* after mutex release, no thread is using the old table anymore. */
5516 	return (0);
5517 }
5518 
5519 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
5520 
5521 static void
5522 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5523     zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5524     boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
5525 {
5526 	zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5527 
5528 	ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5529 	ASSERT3U(ioc, <, ZFS_IOC_LAST);
5530 	ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5531 	ASSERT3P(vec->zvec_func, ==, NULL);
5532 
5533 	vec->zvec_legacy_func = func;
5534 	vec->zvec_secpolicy = secpolicy;
5535 	vec->zvec_namecheck = namecheck;
5536 	vec->zvec_allow_log = log_history;
5537 	vec->zvec_pool_check = pool_check;
5538 }
5539 
5540 /*
5541  * See the block comment at the beginning of this file for details on
5542  * each argument to this function.
5543  */
5544 static void
5545 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
5546     zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
5547     zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
5548     boolean_t allow_log)
5549 {
5550 	zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
5551 
5552 	ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
5553 	ASSERT3U(ioc, <, ZFS_IOC_LAST);
5554 	ASSERT3P(vec->zvec_legacy_func, ==, NULL);
5555 	ASSERT3P(vec->zvec_func, ==, NULL);
5556 
5557 	/* if we are logging, the name must be valid */
5558 	ASSERT(!allow_log || namecheck != NO_NAME);
5559 
5560 	vec->zvec_name = name;
5561 	vec->zvec_func = func;
5562 	vec->zvec_secpolicy = secpolicy;
5563 	vec->zvec_namecheck = namecheck;
5564 	vec->zvec_pool_check = pool_check;
5565 	vec->zvec_smush_outnvlist = smush_outnvlist;
5566 	vec->zvec_allow_log = allow_log;
5567 }
5568 
5569 static void
5570 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5571     zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
5572     zfs_ioc_poolcheck_t pool_check)
5573 {
5574 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
5575 	    POOL_NAME, log_history, pool_check);
5576 }
5577 
5578 static void
5579 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5580     zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
5581 {
5582 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
5583 	    DATASET_NAME, B_FALSE, pool_check);
5584 }
5585 
5586 static void
5587 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5588 {
5589 	zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
5590 	    POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5591 }
5592 
5593 static void
5594 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5595     zfs_secpolicy_func_t *secpolicy)
5596 {
5597 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
5598 	    NO_NAME, B_FALSE, POOL_CHECK_NONE);
5599 }
5600 
5601 static void
5602 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
5603     zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
5604 {
5605 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
5606 	    DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
5607 }
5608 
5609 static void
5610 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
5611 {
5612 	zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
5613 	    zfs_secpolicy_read);
5614 }
5615 
5616 static void
5617 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
5618 	zfs_secpolicy_func_t *secpolicy)
5619 {
5620 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
5621 	    DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5622 }
5623 
5624 static void
5625 zfs_ioctl_init(void)
5626 {
5627 	zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
5628 	    zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
5629 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5630 
5631 	zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
5632 	    zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
5633 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
5634 
5635 	zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
5636 	    zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
5637 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5638 
5639 	zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
5640 	    zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
5641 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5642 
5643 	zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
5644 	    zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
5645 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5646 
5647 	zfs_ioctl_register("create", ZFS_IOC_CREATE,
5648 	    zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
5649 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5650 
5651 	zfs_ioctl_register("clone", ZFS_IOC_CLONE,
5652 	    zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
5653 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5654 
5655 	zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
5656 	    zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
5657 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5658 
5659 	zfs_ioctl_register("hold", ZFS_IOC_HOLD,
5660 	    zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
5661 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5662 	zfs_ioctl_register("release", ZFS_IOC_RELEASE,
5663 	    zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
5664 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5665 
5666 	zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
5667 	    zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
5668 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5669 
5670 	zfs_ioctl_register("set_zev_callbacks", ZFS_IOC_SET_ZEV_CALLBACKS,
5671 	    zfs_ioc_set_zev_callbacks, zfs_secpolicy_set_zev_callbacks, NO_NAME,
5672 	    POOL_CHECK_NONE, B_TRUE, B_FALSE);
5673 
5674 	zfs_ioctl_register("unset_zev_callbacks", ZFS_IOC_UNSET_ZEV_CALLBACKS,
5675 	    zfs_ioc_unset_zev_callbacks, zfs_secpolicy_unset_zev_callbacks,
5676 	    NO_NAME, POOL_CHECK_NONE, B_TRUE, B_FALSE);
5677 	zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
5678 	    zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
5679 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
5680 
5681 	zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
5682 	    zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
5683 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5684 
5685 	zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
5686 	    zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
5687 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
5688 
5689 	zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
5690 	    zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
5691 	    POOL_NAME,
5692 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
5693 
5694 	/* IOCTLS that use the legacy function signature */
5695 
5696 	zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
5697 	    zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
5698 
5699 	zfs_ioctl_register_legacy(ZFS_IOC_ARC_INFO, zfs_ioc_arc_info,
5700 	    zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
5701 
5702 	zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
5703 	    zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5704 	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
5705 	    zfs_ioc_pool_scan);
5706 	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
5707 	    zfs_ioc_pool_upgrade);
5708 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
5709 	    zfs_ioc_vdev_add);
5710 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
5711 	    zfs_ioc_vdev_remove);
5712 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
5713 	    zfs_ioc_vdev_set_state);
5714 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
5715 	    zfs_ioc_vdev_attach);
5716 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
5717 	    zfs_ioc_vdev_detach);
5718 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
5719 	    zfs_ioc_vdev_setpath);
5720 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
5721 	    zfs_ioc_vdev_setfru);
5722 	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
5723 	    zfs_ioc_pool_set_props);
5724 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
5725 	    zfs_ioc_vdev_split);
5726 	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
5727 	    zfs_ioc_pool_reguid);
5728 
5729 	zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
5730 	    zfs_ioc_pool_configs, zfs_secpolicy_none);
5731 	zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
5732 	    zfs_ioc_pool_tryimport, zfs_secpolicy_config);
5733 	zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
5734 	    zfs_ioc_inject_fault, zfs_secpolicy_inject);
5735 	zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
5736 	    zfs_ioc_clear_fault, zfs_secpolicy_inject);
5737 	zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
5738 	    zfs_ioc_inject_list_next, zfs_secpolicy_inject);
5739 
5740 	/*
5741 	 * pool destroy, and export don't log the history as part of
5742 	 * zfsdev_ioctl, but rather zfs_ioc_pool_export
5743 	 * does the logging of those commands.
5744 	 */
5745 	zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
5746 	    zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5747 	zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
5748 	    zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
5749 
5750 	zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
5751 	    zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5752 	zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
5753 	    zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
5754 
5755 	zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
5756 	    zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
5757 	zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
5758 	    zfs_ioc_dsobj_to_dsname,
5759 	    zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
5760 	zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
5761 	    zfs_ioc_pool_get_history,
5762 	    zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
5763 
5764 	zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
5765 	    zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5766 
5767 	zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
5768 	    zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
5769 	zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
5770 	    zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
5771 
5772 	zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
5773 	    zfs_ioc_space_written);
5774 	zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
5775 	    zfs_ioc_objset_recvd_props);
5776 	zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
5777 	    zfs_ioc_next_obj);
5778 	zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
5779 	    zfs_ioc_get_fsacl);
5780 	zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
5781 	    zfs_ioc_objset_stats);
5782 	zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
5783 	    zfs_ioc_objset_zplprops);
5784 	zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
5785 	    zfs_ioc_dataset_list_next);
5786 	zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
5787 	    zfs_ioc_snapshot_list_next);
5788 	zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
5789 	    zfs_ioc_send_progress);
5790 
5791 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
5792 	    zfs_ioc_diff, zfs_secpolicy_diff);
5793 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
5794 	    zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
5795 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
5796 	    zfs_ioc_obj_to_path, zfs_secpolicy_diff);
5797 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
5798 	    zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
5799 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
5800 	    zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
5801 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
5802 	    zfs_ioc_send, zfs_secpolicy_send);
5803 
5804 	zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
5805 	    zfs_secpolicy_none);
5806 	zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
5807 	    zfs_secpolicy_destroy);
5808 	zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
5809 	    zfs_secpolicy_rename);
5810 	zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
5811 	    zfs_secpolicy_recv);
5812 	zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
5813 	    zfs_secpolicy_promote);
5814 	zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
5815 	    zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
5816 	zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
5817 	    zfs_secpolicy_set_fsacl);
5818 
5819 	zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
5820 	    zfs_secpolicy_share, POOL_CHECK_NONE);
5821 	zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
5822 	    zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
5823 	zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
5824 	    zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
5825 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5826 	zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
5827 	    zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
5828 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
5829 }
5830 
5831 int
5832 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
5833     zfs_ioc_poolcheck_t check)
5834 {
5835 	spa_t *spa;
5836 	int error;
5837 
5838 	ASSERT(type == POOL_NAME || type == DATASET_NAME);
5839 
5840 	if (check & POOL_CHECK_NONE)
5841 		return (0);
5842 
5843 	error = spa_open(name, &spa, FTAG);
5844 	if (error == 0) {
5845 		if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
5846 			error = SET_ERROR(EAGAIN);
5847 		else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
5848 			error = SET_ERROR(EROFS);
5849 		spa_close(spa, FTAG);
5850 	}
5851 	return (error);
5852 }
5853 
5854 /*
5855  * Find a free minor number.
5856  */
5857 minor_t
5858 zfsdev_minor_alloc(void)
5859 {
5860 	static minor_t last_minor;
5861 	minor_t m;
5862 
5863 	ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5864 
5865 	for (m = last_minor + 1; m != last_minor; m++) {
5866 		if (m > ZFSDEV_MAX_MINOR)
5867 			m = 1;
5868 		if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
5869 			last_minor = m;
5870 			return (m);
5871 		}
5872 	}
5873 
5874 	return (0);
5875 }
5876 
5877 static int
5878 zfs_ctldev_init(dev_t *devp)
5879 {
5880 	minor_t minor;
5881 	zfs_soft_state_t *zs;
5882 
5883 	ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5884 	ASSERT(getminor(*devp) == 0);
5885 
5886 	minor = zfsdev_minor_alloc();
5887 	if (minor == 0)
5888 		return (SET_ERROR(ENXIO));
5889 
5890 	if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
5891 		return (SET_ERROR(EAGAIN));
5892 
5893 	*devp = makedevice(getemajor(*devp), minor);
5894 
5895 	zs = ddi_get_soft_state(zfsdev_state, minor);
5896 	zs->zss_type = ZSST_CTLDEV;
5897 	zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
5898 
5899 	return (0);
5900 }
5901 
5902 static void
5903 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
5904 {
5905 	ASSERT(MUTEX_HELD(&zfsdev_state_lock));
5906 
5907 	zfs_onexit_destroy(zo);
5908 	ddi_soft_state_free(zfsdev_state, minor);
5909 }
5910 
5911 void *
5912 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
5913 {
5914 	zfs_soft_state_t *zp;
5915 
5916 	zp = ddi_get_soft_state(zfsdev_state, minor);
5917 	if (zp == NULL || zp->zss_type != which)
5918 		return (NULL);
5919 
5920 	return (zp->zss_data);
5921 }
5922 
5923 static int
5924 zfsdev_open(dev_t *devp, int flag, int otyp, cred_t *cr)
5925 {
5926 	int error = 0;
5927 
5928 	if (getminor(*devp) != 0)
5929 		return (zvol_open(devp, flag, otyp, cr));
5930 
5931 	/* This is the control device. Allocate a new minor if requested. */
5932 	if (flag & FEXCL) {
5933 		mutex_enter(&zfsdev_state_lock);
5934 		error = zfs_ctldev_init(devp);
5935 		mutex_exit(&zfsdev_state_lock);
5936 	}
5937 
5938 	return (error);
5939 }
5940 
5941 static int
5942 zfsdev_close(dev_t dev, int flag, int otyp, cred_t *cr)
5943 {
5944 	zfs_onexit_t *zo;
5945 	minor_t minor = getminor(dev);
5946 
5947 	if (minor == 0)
5948 		return (0);
5949 
5950 	mutex_enter(&zfsdev_state_lock);
5951 	zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
5952 	if (zo == NULL) {
5953 		mutex_exit(&zfsdev_state_lock);
5954 		return (zvol_close(dev, flag, otyp, cr));
5955 	}
5956 	zfs_ctldev_destroy(zo, minor);
5957 	mutex_exit(&zfsdev_state_lock);
5958 
5959 	return (0);
5960 }
5961 
5962 static int
5963 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
5964 {
5965 	zfs_cmd_t *zc;
5966 	uint_t vecnum;
5967 	int error, rc, len;
5968 	minor_t minor = getminor(dev);
5969 	const zfs_ioc_vec_t *vec;
5970 	char *saved_poolname = NULL;
5971 	nvlist_t *innvl = NULL;
5972 
5973 	if (minor != 0 &&
5974 	    zfsdev_get_soft_state(minor, ZSST_CTLDEV) == NULL)
5975 		return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
5976 
5977 	vecnum = cmd - ZFS_IOC_FIRST;
5978 	ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
5979 
5980 	if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
5981 		return (SET_ERROR(EINVAL));
5982 	vec = &zfs_ioc_vec[vecnum];
5983 
5984 	zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
5985 
5986 	error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
5987 	if (error != 0) {
5988 		error = SET_ERROR(EFAULT);
5989 		goto out;
5990 	}
5991 
5992 	zc->zc_iflags = flag & FKIOCTL;
5993 	if (zc->zc_nvlist_src_size != 0) {
5994 		error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
5995 		    zc->zc_iflags, &innvl);
5996 		if (error != 0)
5997 			goto out;
5998 	}
5999 
6000 	/*
6001 	 * Ensure that all pool/dataset names are valid before we pass down to
6002 	 * the lower layers.
6003 	 */
6004 	zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
6005 	switch (vec->zvec_namecheck) {
6006 	case POOL_NAME:
6007 		if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
6008 			error = SET_ERROR(EINVAL);
6009 		else
6010 			error = pool_status_check(zc->zc_name,
6011 			    vec->zvec_namecheck, vec->zvec_pool_check);
6012 		break;
6013 
6014 	case DATASET_NAME:
6015 		if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
6016 			error = SET_ERROR(EINVAL);
6017 		else
6018 			error = pool_status_check(zc->zc_name,
6019 			    vec->zvec_namecheck, vec->zvec_pool_check);
6020 		break;
6021 
6022 	case NO_NAME:
6023 		break;
6024 	}
6025 
6026 
6027 	if (error == 0 && !(flag & FKIOCTL))
6028 		error = vec->zvec_secpolicy(zc, innvl, cr);
6029 
6030 	if (error != 0)
6031 		goto out;
6032 
6033 	/* legacy ioctls can modify zc_name */
6034 	len = strcspn(zc->zc_name, "/@#") + 1;
6035 	saved_poolname = kmem_alloc(len, KM_SLEEP);
6036 	(void) strlcpy(saved_poolname, zc->zc_name, len);
6037 
6038 	if (vec->zvec_func != NULL) {
6039 		nvlist_t *outnvl;
6040 		int puterror = 0;
6041 		spa_t *spa;
6042 		nvlist_t *lognv = NULL;
6043 
6044 		ASSERT(vec->zvec_legacy_func == NULL);
6045 
6046 		/*
6047 		 * Add the innvl to the lognv before calling the func,
6048 		 * in case the func changes the innvl.
6049 		 */
6050 		if (vec->zvec_allow_log) {
6051 			lognv = fnvlist_alloc();
6052 			fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
6053 			    vec->zvec_name);
6054 			if (!nvlist_empty(innvl)) {
6055 				fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
6056 				    innvl);
6057 			}
6058 		}
6059 
6060 		outnvl = fnvlist_alloc();
6061 		error = vec->zvec_func(zc->zc_name, innvl, outnvl);
6062 
6063 		if (error == 0 && vec->zvec_allow_log &&
6064 		    spa_open(zc->zc_name, &spa, FTAG) == 0) {
6065 			if (!nvlist_empty(outnvl)) {
6066 				fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
6067 				    outnvl);
6068 			}
6069 			(void) spa_history_log_nvl(spa, lognv);
6070 			spa_close(spa, FTAG);
6071 		}
6072 		fnvlist_free(lognv);
6073 
6074 		if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
6075 			int smusherror = 0;
6076 			if (vec->zvec_smush_outnvlist) {
6077 				smusherror = nvlist_smush(outnvl,
6078 				    zc->zc_nvlist_dst_size);
6079 			}
6080 			if (smusherror == 0)
6081 				puterror = put_nvlist(zc, outnvl);
6082 		}
6083 
6084 		if (puterror != 0)
6085 			error = puterror;
6086 
6087 		nvlist_free(outnvl);
6088 	} else {
6089 		error = vec->zvec_legacy_func(zc);
6090 	}
6091 
6092 out:
6093 	nvlist_free(innvl);
6094 	rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
6095 	if (error == 0 && rc != 0)
6096 		error = SET_ERROR(EFAULT);
6097 	if (error == 0 && vec->zvec_allow_log) {
6098 		char *s = tsd_get(zfs_allow_log_key);
6099 		if (s != NULL)
6100 			strfree(s);
6101 		(void) tsd_set(zfs_allow_log_key, saved_poolname);
6102 	} else {
6103 		if (saved_poolname != NULL)
6104 			strfree(saved_poolname);
6105 	}
6106 
6107 	kmem_free(zc, sizeof (zfs_cmd_t));
6108 	return (error);
6109 }
6110 
6111 static int
6112 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
6113 {
6114 	if (cmd != DDI_ATTACH)
6115 		return (DDI_FAILURE);
6116 
6117 	if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
6118 	    DDI_PSEUDO, 0) == DDI_FAILURE)
6119 		return (DDI_FAILURE);
6120 
6121 	zfs_dip = dip;
6122 
6123 	ddi_report_dev(dip);
6124 
6125 	return (DDI_SUCCESS);
6126 }
6127 
6128 static int
6129 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
6130 {
6131 	if (spa_busy() || zfs_busy() || zvol_busy())
6132 		return (DDI_FAILURE);
6133 
6134 	if (cmd != DDI_DETACH)
6135 		return (DDI_FAILURE);
6136 
6137 	zfs_dip = NULL;
6138 
6139 	ddi_prop_remove_all(dip);
6140 	ddi_remove_minor_node(dip, NULL);
6141 
6142 	return (DDI_SUCCESS);
6143 }
6144 
6145 /*ARGSUSED*/
6146 static int
6147 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
6148 {
6149 	switch (infocmd) {
6150 	case DDI_INFO_DEVT2DEVINFO:
6151 		*result = zfs_dip;
6152 		return (DDI_SUCCESS);
6153 
6154 	case DDI_INFO_DEVT2INSTANCE:
6155 		*result = (void *)0;
6156 		return (DDI_SUCCESS);
6157 	}
6158 
6159 	return (DDI_FAILURE);
6160 }
6161 
6162 /*
6163  * OK, so this is a little weird.
6164  *
6165  * /dev/zfs is the control node, i.e. minor 0.
6166  * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
6167  *
6168  * /dev/zfs has basically nothing to do except serve up ioctls,
6169  * so most of the standard driver entry points are in zvol.c.
6170  */
6171 static struct cb_ops zfs_cb_ops = {
6172 	zfsdev_open,	/* open */
6173 	zfsdev_close,	/* close */
6174 	zvol_strategy,	/* strategy */
6175 	nodev,		/* print */
6176 	zvol_dump,	/* dump */
6177 	zvol_read,	/* read */
6178 	zvol_write,	/* write */
6179 	zfsdev_ioctl,	/* ioctl */
6180 	nodev,		/* devmap */
6181 	nodev,		/* mmap */
6182 	nodev,		/* segmap */
6183 	nochpoll,	/* poll */
6184 	ddi_prop_op,	/* prop_op */
6185 	NULL,		/* streamtab */
6186 	D_NEW | D_MP | D_64BIT,		/* Driver compatibility flag */
6187 	CB_REV,		/* version */
6188 	nodev,		/* async read */
6189 	nodev,		/* async write */
6190 };
6191 
6192 static struct dev_ops zfs_dev_ops = {
6193 	DEVO_REV,	/* version */
6194 	0,		/* refcnt */
6195 	zfs_info,	/* info */
6196 	nulldev,	/* identify */
6197 	nulldev,	/* probe */
6198 	zfs_attach,	/* attach */
6199 	zfs_detach,	/* detach */
6200 	nodev,		/* reset */
6201 	&zfs_cb_ops,	/* driver operations */
6202 	NULL,		/* no bus operations */
6203 	NULL,		/* power */
6204 	ddi_quiesce_not_needed,	/* quiesce */
6205 };
6206 
6207 static struct modldrv zfs_modldrv = {
6208 	&mod_driverops,
6209 	"ZFS storage pool",
6210 	&zfs_dev_ops
6211 };
6212 
6213 static struct modlinkage modlinkage = {
6214 	MODREV_1,
6215 	(void *)&zfs_modlfs,
6216 	(void *)&zfs_modldrv,
6217 	NULL
6218 };
6219 
6220 static void
6221 zfs_allow_log_destroy(void *arg)
6222 {
6223 	char *poolname = arg;
6224 	strfree(poolname);
6225 }
6226 
6227 int
6228 _init(void)
6229 {
6230 	int error;
6231 
6232 	spa_init(FREAD | FWRITE);
6233 	zfs_init();
6234 	zvol_init();
6235 	zfs_ioctl_init();
6236 	rz_zev_init();
6237 
6238 	if ((error = mod_install(&modlinkage)) != 0) {
6239 		zvol_fini();
6240 		zfs_fini();
6241 		spa_fini();
6242 		return (error);
6243 	}
6244 
6245 	tsd_create(&zfs_fsyncer_key, NULL);
6246 	tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
6247 	tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
6248 
6249 	error = ldi_ident_from_mod(&modlinkage, &zfs_li);
6250 	ASSERT(error == 0);
6251 	mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
6252 
6253 	return (0);
6254 }
6255 
6256 int
6257 _fini(void)
6258 {
6259 	int error;
6260 
6261 	if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
6262 		return (SET_ERROR(EBUSY));
6263 
6264 	if ((error = mod_remove(&modlinkage)) != 0)
6265 		return (error);
6266 
6267 	rz_zev_fini();
6268 	zvol_fini();
6269 	zfs_fini();
6270 	spa_fini();
6271 	if (zfs_nfsshare_inited)
6272 		(void) ddi_modclose(nfs_mod);
6273 	if (zfs_smbshare_inited)
6274 		(void) ddi_modclose(smbsrv_mod);
6275 	if (zfs_nfsshare_inited || zfs_smbshare_inited)
6276 		(void) ddi_modclose(sharefs_mod);
6277 
6278 	tsd_destroy(&zfs_fsyncer_key);
6279 	ldi_ident_release(zfs_li);
6280 	zfs_li = NULL;
6281 	mutex_destroy(&zfs_share_lock);
6282 
6283 	return (error);
6284 }
6285 
6286 int
6287 _info(struct modinfo *modinfop)
6288 {
6289 	return (mod_info(&modlinkage, modinfop));
6290 }
6291