xref: /illumos-gate/usr/src/uts/common/fs/zfs/zfs_ioctl.c (revision b38f097029665f4ece801ca129913d36f757b49c)
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  * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 #pragma ident	"%Z%%M%	%I%	%E% SMI"
27 
28 #include <sys/types.h>
29 #include <sys/param.h>
30 #include <sys/errno.h>
31 #include <sys/uio.h>
32 #include <sys/buf.h>
33 #include <sys/modctl.h>
34 #include <sys/open.h>
35 #include <sys/file.h>
36 #include <sys/kmem.h>
37 #include <sys/conf.h>
38 #include <sys/cmn_err.h>
39 #include <sys/stat.h>
40 #include <sys/zfs_ioctl.h>
41 #include <sys/zfs_znode.h>
42 #include <sys/zap.h>
43 #include <sys/spa.h>
44 #include <sys/spa_impl.h>
45 #include <sys/vdev.h>
46 #include <sys/vdev_impl.h>
47 #include <sys/dmu.h>
48 #include <sys/dsl_dir.h>
49 #include <sys/dsl_dataset.h>
50 #include <sys/dsl_prop.h>
51 #include <sys/dsl_deleg.h>
52 #include <sys/dmu_objset.h>
53 #include <sys/ddi.h>
54 #include <sys/sunddi.h>
55 #include <sys/sunldi.h>
56 #include <sys/policy.h>
57 #include <sys/zone.h>
58 #include <sys/nvpair.h>
59 #include <sys/pathname.h>
60 #include <sys/mount.h>
61 #include <sys/sdt.h>
62 #include <sys/fs/zfs.h>
63 #include <sys/zfs_ctldir.h>
64 #include <sys/zfs_dir.h>
65 #include <sys/zvol.h>
66 #include <sharefs/share.h>
67 #include <sys/dmu_objset.h>
68 
69 #include "zfs_namecheck.h"
70 #include "zfs_prop.h"
71 #include "zfs_deleg.h"
72 
73 extern struct modlfs zfs_modlfs;
74 
75 extern void zfs_init(void);
76 extern void zfs_fini(void);
77 
78 ldi_ident_t zfs_li = NULL;
79 dev_info_t *zfs_dip;
80 
81 typedef int zfs_ioc_func_t(zfs_cmd_t *);
82 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, cred_t *);
83 
84 typedef struct zfs_ioc_vec {
85 	zfs_ioc_func_t		*zvec_func;
86 	zfs_secpolicy_func_t	*zvec_secpolicy;
87 	enum {
88 		NO_NAME,
89 		POOL_NAME,
90 		DATASET_NAME
91 	} zvec_namecheck;
92 	boolean_t		zvec_his_log;
93 } zfs_ioc_vec_t;
94 
95 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
96 void
97 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
98 {
99 	const char *newfile;
100 	char buf[256];
101 	va_list adx;
102 
103 	/*
104 	 * Get rid of annoying "../common/" prefix to filename.
105 	 */
106 	newfile = strrchr(file, '/');
107 	if (newfile != NULL) {
108 		newfile = newfile + 1; /* Get rid of leading / */
109 	} else {
110 		newfile = file;
111 	}
112 
113 	va_start(adx, fmt);
114 	(void) vsnprintf(buf, sizeof (buf), fmt, adx);
115 	va_end(adx);
116 
117 	/*
118 	 * To get this data, use the zfs-dprintf probe as so:
119 	 * dtrace -q -n 'zfs-dprintf \
120 	 *	/stringof(arg0) == "dbuf.c"/ \
121 	 *	{printf("%s: %s", stringof(arg1), stringof(arg3))}'
122 	 * arg0 = file name
123 	 * arg1 = function name
124 	 * arg2 = line number
125 	 * arg3 = message
126 	 */
127 	DTRACE_PROBE4(zfs__dprintf,
128 	    char *, newfile, char *, func, int, line, char *, buf);
129 }
130 
131 static void
132 history_str_free(char *buf)
133 {
134 	kmem_free(buf, HIS_MAX_RECORD_LEN);
135 }
136 
137 static char *
138 history_str_get(zfs_cmd_t *zc)
139 {
140 	char *buf;
141 
142 	if (zc->zc_history == NULL)
143 		return (NULL);
144 
145 	buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
146 	if (copyinstr((void *)(uintptr_t)zc->zc_history,
147 	    buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
148 		history_str_free(buf);
149 		return (NULL);
150 	}
151 
152 	buf[HIS_MAX_RECORD_LEN -1] = '\0';
153 
154 	return (buf);
155 }
156 
157 /*
158  * zfs_check_version
159  *
160  *	Return non-zero if the spa version is less than requested version.
161  */
162 static int
163 zfs_check_version(const char *name, int version)
164 {
165 
166 	spa_t *spa;
167 
168 	if (spa_open(name, &spa, FTAG) == 0) {
169 		if (spa_version(spa) < version) {
170 			spa_close(spa, FTAG);
171 			return (1);
172 		}
173 		spa_close(spa, FTAG);
174 	}
175 	return (0);
176 }
177 
178 static void
179 zfs_log_history(zfs_cmd_t *zc)
180 {
181 	spa_t *spa;
182 	char *buf;
183 
184 	if ((buf = history_str_get(zc)) == NULL)
185 		return;
186 
187 	if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
188 		if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
189 			(void) spa_history_log(spa, buf, LOG_CMD_NORMAL);
190 		spa_close(spa, FTAG);
191 	}
192 	history_str_free(buf);
193 }
194 
195 /*
196  * Policy for top-level read operations (list pools).  Requires no privileges,
197  * and can be used in the local zone, as there is no associated dataset.
198  */
199 /* ARGSUSED */
200 static int
201 zfs_secpolicy_none(zfs_cmd_t *zc, cred_t *cr)
202 {
203 	return (0);
204 }
205 
206 /*
207  * Policy for dataset read operations (list children, get statistics).  Requires
208  * no privileges, but must be visible in the local zone.
209  */
210 /* ARGSUSED */
211 static int
212 zfs_secpolicy_read(zfs_cmd_t *zc, cred_t *cr)
213 {
214 	if (INGLOBALZONE(curproc) ||
215 	    zone_dataset_visible(zc->zc_name, NULL))
216 		return (0);
217 
218 	return (ENOENT);
219 }
220 
221 static int
222 zfs_dozonecheck(const char *dataset, cred_t *cr)
223 {
224 	uint64_t zoned;
225 	int writable = 1;
226 
227 	/*
228 	 * The dataset must be visible by this zone -- check this first
229 	 * so they don't see EPERM on something they shouldn't know about.
230 	 */
231 	if (!INGLOBALZONE(curproc) &&
232 	    !zone_dataset_visible(dataset, &writable))
233 		return (ENOENT);
234 
235 	if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
236 		return (ENOENT);
237 
238 	if (INGLOBALZONE(curproc)) {
239 		/*
240 		 * If the fs is zoned, only root can access it from the
241 		 * global zone.
242 		 */
243 		if (secpolicy_zfs(cr) && zoned)
244 			return (EPERM);
245 	} else {
246 		/*
247 		 * If we are in a local zone, the 'zoned' property must be set.
248 		 */
249 		if (!zoned)
250 			return (EPERM);
251 
252 		/* must be writable by this zone */
253 		if (!writable)
254 			return (EPERM);
255 	}
256 	return (0);
257 }
258 
259 int
260 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
261 {
262 	int error;
263 
264 	error = zfs_dozonecheck(name, cr);
265 	if (error == 0) {
266 		error = secpolicy_zfs(cr);
267 		if (error)
268 			error = dsl_deleg_access(name, perm, cr);
269 	}
270 	return (error);
271 }
272 
273 static int
274 zfs_secpolicy_setprop(const char *name, zfs_prop_t prop, cred_t *cr)
275 {
276 	/*
277 	 * Check permissions for special properties.
278 	 */
279 	switch (prop) {
280 	case ZFS_PROP_ZONED:
281 		/*
282 		 * Disallow setting of 'zoned' from within a local zone.
283 		 */
284 		if (!INGLOBALZONE(curproc))
285 			return (EPERM);
286 		break;
287 
288 	case ZFS_PROP_QUOTA:
289 		if (!INGLOBALZONE(curproc)) {
290 			uint64_t zoned;
291 			char setpoint[MAXNAMELEN];
292 			/*
293 			 * Unprivileged users are allowed to modify the
294 			 * quota on things *under* (ie. contained by)
295 			 * the thing they own.
296 			 */
297 			if (dsl_prop_get_integer(name, "zoned", &zoned,
298 			    setpoint))
299 				return (EPERM);
300 			if (!zoned || strlen(name) <= strlen(setpoint))
301 				return (EPERM);
302 		}
303 		break;
304 	}
305 
306 	return (zfs_secpolicy_write_perms(name, zfs_prop_to_name(prop), cr));
307 }
308 
309 int
310 zfs_secpolicy_fsacl(zfs_cmd_t *zc, cred_t *cr)
311 {
312 	int error;
313 
314 	error = zfs_dozonecheck(zc->zc_name, cr);
315 	if (error)
316 		return (error);
317 
318 	/*
319 	 * permission to set permissions will be evaluated later in
320 	 * dsl_deleg_can_allow()
321 	 */
322 	return (0);
323 }
324 
325 int
326 zfs_secpolicy_rollback(zfs_cmd_t *zc, cred_t *cr)
327 {
328 	int error;
329 	error = zfs_secpolicy_write_perms(zc->zc_name,
330 	    ZFS_DELEG_PERM_ROLLBACK, cr);
331 	if (error == 0)
332 		error = zfs_secpolicy_write_perms(zc->zc_name,
333 		    ZFS_DELEG_PERM_MOUNT, cr);
334 	return (error);
335 }
336 
337 int
338 zfs_secpolicy_send(zfs_cmd_t *zc, cred_t *cr)
339 {
340 	return (zfs_secpolicy_write_perms(zc->zc_name,
341 	    ZFS_DELEG_PERM_SEND, cr));
342 }
343 
344 int
345 zfs_secpolicy_share(zfs_cmd_t *zc, cred_t *cr)
346 {
347 	if (!INGLOBALZONE(curproc))
348 		return (EPERM);
349 
350 	if (secpolicy_nfs(cr) == 0) {
351 		return (0);
352 	} else {
353 		vnode_t *vp;
354 		int error;
355 
356 		if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
357 		    NO_FOLLOW, NULL, &vp)) != 0)
358 			return (error);
359 
360 		/* Now make sure mntpnt and dataset are ZFS */
361 
362 		if (vp->v_vfsp->vfs_fstype != zfsfstype ||
363 		    (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
364 		    zc->zc_name) != 0)) {
365 			VN_RELE(vp);
366 			return (EPERM);
367 		}
368 
369 		VN_RELE(vp);
370 		return (dsl_deleg_access(zc->zc_name,
371 		    ZFS_DELEG_PERM_SHARE, cr));
372 	}
373 }
374 
375 static int
376 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
377 {
378 	char *cp;
379 
380 	/*
381 	 * Remove the @bla or /bla from the end of the name to get the parent.
382 	 */
383 	(void) strncpy(parent, datasetname, parentsize);
384 	cp = strrchr(parent, '@');
385 	if (cp != NULL) {
386 		cp[0] = '\0';
387 	} else {
388 		cp = strrchr(parent, '/');
389 		if (cp == NULL)
390 			return (ENOENT);
391 		cp[0] = '\0';
392 	}
393 
394 	return (0);
395 }
396 
397 int
398 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
399 {
400 	int error;
401 
402 	if ((error = zfs_secpolicy_write_perms(name,
403 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
404 		return (error);
405 
406 	return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
407 }
408 
409 static int
410 zfs_secpolicy_destroy(zfs_cmd_t *zc, cred_t *cr)
411 {
412 	return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
413 }
414 
415 /*
416  * Must have sys_config privilege to check the iscsi permission
417  */
418 /* ARGSUSED */
419 static int
420 zfs_secpolicy_iscsi(zfs_cmd_t *zc, cred_t *cr)
421 {
422 	return (secpolicy_zfs(cr));
423 }
424 
425 int
426 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
427 {
428 	char 	parentname[MAXNAMELEN];
429 	int	error;
430 
431 	if ((error = zfs_secpolicy_write_perms(from,
432 	    ZFS_DELEG_PERM_RENAME, cr)) != 0)
433 		return (error);
434 
435 	if ((error = zfs_secpolicy_write_perms(from,
436 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
437 		return (error);
438 
439 	if ((error = zfs_get_parent(to, parentname,
440 	    sizeof (parentname))) != 0)
441 		return (error);
442 
443 	if ((error = zfs_secpolicy_write_perms(parentname,
444 	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
445 		return (error);
446 
447 	if ((error = zfs_secpolicy_write_perms(parentname,
448 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
449 		return (error);
450 
451 	return (error);
452 }
453 
454 static int
455 zfs_secpolicy_rename(zfs_cmd_t *zc, cred_t *cr)
456 {
457 	return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
458 }
459 
460 static int
461 zfs_secpolicy_promote(zfs_cmd_t *zc, cred_t *cr)
462 {
463 	char 	parentname[MAXNAMELEN];
464 	objset_t *clone;
465 	int error;
466 
467 	error = zfs_secpolicy_write_perms(zc->zc_name,
468 	    ZFS_DELEG_PERM_PROMOTE, cr);
469 	if (error)
470 		return (error);
471 
472 	error = dmu_objset_open(zc->zc_name, DMU_OST_ANY,
473 	    DS_MODE_STANDARD | DS_MODE_READONLY, &clone);
474 
475 	if (error == 0) {
476 		dsl_dataset_t *pclone = NULL;
477 		dsl_dir_t *dd;
478 		dd = clone->os->os_dsl_dataset->ds_dir;
479 
480 		rw_enter(&dd->dd_pool->dp_config_rwlock, RW_READER);
481 		error = dsl_dataset_open_obj(dd->dd_pool,
482 		    dd->dd_phys->dd_origin_obj, NULL,
483 		    DS_MODE_NONE, FTAG, &pclone);
484 		rw_exit(&dd->dd_pool->dp_config_rwlock);
485 		if (error) {
486 			dmu_objset_close(clone);
487 			return (error);
488 		}
489 
490 		error = zfs_secpolicy_write_perms(zc->zc_name,
491 		    ZFS_DELEG_PERM_MOUNT, cr);
492 
493 		dsl_dataset_name(pclone, parentname);
494 		dmu_objset_close(clone);
495 		dsl_dataset_close(pclone, DS_MODE_NONE, FTAG);
496 		if (error == 0)
497 			error = zfs_secpolicy_write_perms(parentname,
498 			    ZFS_DELEG_PERM_PROMOTE, cr);
499 	}
500 	return (error);
501 }
502 
503 static int
504 zfs_secpolicy_receive(zfs_cmd_t *zc, cred_t *cr)
505 {
506 	int error;
507 
508 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
509 	    ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
510 		return (error);
511 
512 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
513 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
514 		return (error);
515 
516 	return (zfs_secpolicy_write_perms(zc->zc_name,
517 	    ZFS_DELEG_PERM_CREATE, cr));
518 }
519 
520 int
521 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
522 {
523 	int error;
524 
525 	if ((error = zfs_secpolicy_write_perms(name,
526 	    ZFS_DELEG_PERM_SNAPSHOT, cr)) != 0)
527 		return (error);
528 
529 	error = zfs_secpolicy_write_perms(name,
530 	    ZFS_DELEG_PERM_MOUNT, cr);
531 
532 	return (error);
533 }
534 
535 static int
536 zfs_secpolicy_snapshot(zfs_cmd_t *zc, cred_t *cr)
537 {
538 
539 	return (zfs_secpolicy_snapshot_perms(zc->zc_name, cr));
540 }
541 
542 static int
543 zfs_secpolicy_create(zfs_cmd_t *zc, cred_t *cr)
544 {
545 	char 	parentname[MAXNAMELEN];
546 	int 	error;
547 
548 	if ((error = zfs_get_parent(zc->zc_name, parentname,
549 	    sizeof (parentname))) != 0)
550 		return (error);
551 
552 	if (zc->zc_value[0] != '\0') {
553 		if ((error = zfs_secpolicy_write_perms(zc->zc_value,
554 		    ZFS_DELEG_PERM_CLONE, cr)) != 0)
555 			return (error);
556 	}
557 
558 	if ((error = zfs_secpolicy_write_perms(parentname,
559 	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
560 		return (error);
561 
562 	error = zfs_secpolicy_write_perms(parentname,
563 	    ZFS_DELEG_PERM_MOUNT, cr);
564 
565 	return (error);
566 }
567 
568 static int
569 zfs_secpolicy_umount(zfs_cmd_t *zc, cred_t *cr)
570 {
571 	int error;
572 
573 	error = secpolicy_fs_unmount(cr, NULL);
574 	if (error) {
575 		error = dsl_deleg_access(zc->zc_name, ZFS_DELEG_PERM_MOUNT, cr);
576 	}
577 	return (error);
578 }
579 
580 /*
581  * Policy for pool operations - create/destroy pools, add vdevs, etc.  Requires
582  * SYS_CONFIG privilege, which is not available in a local zone.
583  */
584 /* ARGSUSED */
585 static int
586 zfs_secpolicy_config(zfs_cmd_t *zc, cred_t *cr)
587 {
588 	if (secpolicy_sys_config(cr, B_FALSE) != 0)
589 		return (EPERM);
590 
591 	return (0);
592 }
593 
594 /*
595  * Just like zfs_secpolicy_config, except that we will check for
596  * mount permission on the dataset for permission to create/remove
597  * the minor nodes.
598  */
599 static int
600 zfs_secpolicy_minor(zfs_cmd_t *zc, cred_t *cr)
601 {
602 	if (secpolicy_sys_config(cr, B_FALSE) != 0) {
603 		return (dsl_deleg_access(zc->zc_name,
604 		    ZFS_DELEG_PERM_MOUNT, cr));
605 	}
606 
607 	return (0);
608 }
609 
610 /*
611  * Policy for fault injection.  Requires all privileges.
612  */
613 /* ARGSUSED */
614 static int
615 zfs_secpolicy_inject(zfs_cmd_t *zc, cred_t *cr)
616 {
617 	return (secpolicy_zinject(cr));
618 }
619 
620 static int
621 zfs_secpolicy_inherit(zfs_cmd_t *zc, cred_t *cr)
622 {
623 	zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
624 
625 	if (prop == ZPROP_INVAL) {
626 		if (!zfs_prop_user(zc->zc_value))
627 			return (EINVAL);
628 		return (zfs_secpolicy_write_perms(zc->zc_name,
629 		    ZFS_DELEG_PERM_USERPROP, cr));
630 	} else {
631 		if (!zfs_prop_inheritable(prop))
632 			return (EINVAL);
633 		return (zfs_secpolicy_setprop(zc->zc_name, prop, cr));
634 	}
635 }
636 
637 /*
638  * Returns the nvlist as specified by the user in the zfs_cmd_t.
639  */
640 static int
641 get_nvlist(uint64_t nvl, uint64_t size, nvlist_t **nvp)
642 {
643 	char *packed;
644 	int error;
645 	nvlist_t *list = NULL;
646 
647 	/*
648 	 * Read in and unpack the user-supplied nvlist.
649 	 */
650 	if (size == 0)
651 		return (EINVAL);
652 
653 	packed = kmem_alloc(size, KM_SLEEP);
654 
655 	if ((error = xcopyin((void *)(uintptr_t)nvl, packed, size)) != 0) {
656 		kmem_free(packed, size);
657 		return (error);
658 	}
659 
660 	if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
661 		kmem_free(packed, size);
662 		return (error);
663 	}
664 
665 	kmem_free(packed, size);
666 
667 	*nvp = list;
668 	return (0);
669 }
670 
671 static int
672 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
673 {
674 	char *packed = NULL;
675 	size_t size;
676 	int error;
677 
678 	VERIFY(nvlist_size(nvl, &size, NV_ENCODE_NATIVE) == 0);
679 
680 	if (size > zc->zc_nvlist_dst_size) {
681 		error = ENOMEM;
682 	} else {
683 		packed = kmem_alloc(size, KM_SLEEP);
684 		VERIFY(nvlist_pack(nvl, &packed, &size, NV_ENCODE_NATIVE,
685 		    KM_SLEEP) == 0);
686 		error = xcopyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
687 		    size);
688 		kmem_free(packed, size);
689 	}
690 
691 	zc->zc_nvlist_dst_size = size;
692 	return (error);
693 }
694 
695 static int
696 zfs_ioc_pool_create(zfs_cmd_t *zc)
697 {
698 	int error;
699 	nvlist_t *config, *props = NULL;
700 	char *buf;
701 
702 	if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
703 	    &config))
704 		return (error);
705 
706 	if (zc->zc_nvlist_src_size != 0 && (error =
707 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, &props))) {
708 		nvlist_free(config);
709 		return (error);
710 	}
711 
712 	buf = history_str_get(zc);
713 
714 	error = spa_create(zc->zc_name, config, props, buf);
715 
716 	if (buf != NULL)
717 		history_str_free(buf);
718 
719 	nvlist_free(config);
720 
721 	if (props)
722 		nvlist_free(props);
723 
724 	return (error);
725 }
726 
727 static int
728 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
729 {
730 	int error;
731 	zfs_log_history(zc);
732 	error = spa_destroy(zc->zc_name);
733 	return (error);
734 }
735 
736 static int
737 zfs_ioc_pool_import(zfs_cmd_t *zc)
738 {
739 	int error;
740 	nvlist_t *config, *props = NULL;
741 	uint64_t guid;
742 
743 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
744 	    &config)) != 0)
745 		return (error);
746 
747 	if (zc->zc_nvlist_src_size != 0 && (error =
748 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, &props))) {
749 		nvlist_free(config);
750 		return (error);
751 	}
752 
753 	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
754 	    guid != zc->zc_guid)
755 		error = EINVAL;
756 	else
757 		error = spa_import(zc->zc_name, config, props);
758 
759 	nvlist_free(config);
760 
761 	if (props)
762 		nvlist_free(props);
763 
764 	return (error);
765 }
766 
767 static int
768 zfs_ioc_pool_export(zfs_cmd_t *zc)
769 {
770 	int error;
771 	zfs_log_history(zc);
772 	error = spa_export(zc->zc_name, NULL);
773 	return (error);
774 }
775 
776 static int
777 zfs_ioc_pool_configs(zfs_cmd_t *zc)
778 {
779 	nvlist_t *configs;
780 	int error;
781 
782 	if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
783 		return (EEXIST);
784 
785 	error = put_nvlist(zc, configs);
786 
787 	nvlist_free(configs);
788 
789 	return (error);
790 }
791 
792 static int
793 zfs_ioc_pool_stats(zfs_cmd_t *zc)
794 {
795 	nvlist_t *config;
796 	int error;
797 	int ret = 0;
798 
799 	error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
800 	    sizeof (zc->zc_value));
801 
802 	if (config != NULL) {
803 		ret = put_nvlist(zc, config);
804 		nvlist_free(config);
805 
806 		/*
807 		 * The config may be present even if 'error' is non-zero.
808 		 * In this case we return success, and preserve the real errno
809 		 * in 'zc_cookie'.
810 		 */
811 		zc->zc_cookie = error;
812 	} else {
813 		ret = error;
814 	}
815 
816 	return (ret);
817 }
818 
819 /*
820  * Try to import the given pool, returning pool stats as appropriate so that
821  * user land knows which devices are available and overall pool health.
822  */
823 static int
824 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
825 {
826 	nvlist_t *tryconfig, *config;
827 	int error;
828 
829 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
830 	    &tryconfig)) != 0)
831 		return (error);
832 
833 	config = spa_tryimport(tryconfig);
834 
835 	nvlist_free(tryconfig);
836 
837 	if (config == NULL)
838 		return (EINVAL);
839 
840 	error = put_nvlist(zc, config);
841 	nvlist_free(config);
842 
843 	return (error);
844 }
845 
846 static int
847 zfs_ioc_pool_scrub(zfs_cmd_t *zc)
848 {
849 	spa_t *spa;
850 	int error;
851 
852 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
853 		return (error);
854 
855 	mutex_enter(&spa_namespace_lock);
856 	error = spa_scrub(spa, zc->zc_cookie, B_FALSE);
857 	mutex_exit(&spa_namespace_lock);
858 
859 	spa_close(spa, FTAG);
860 
861 	return (error);
862 }
863 
864 static int
865 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
866 {
867 	spa_t *spa;
868 	int error;
869 
870 	error = spa_open(zc->zc_name, &spa, FTAG);
871 	if (error == 0) {
872 		spa_freeze(spa);
873 		spa_close(spa, FTAG);
874 	}
875 	return (error);
876 }
877 
878 static int
879 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
880 {
881 	spa_t *spa;
882 	int error;
883 
884 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
885 		return (error);
886 
887 	if (zc->zc_cookie < spa_version(spa) || zc->zc_cookie > SPA_VERSION) {
888 		spa_close(spa, FTAG);
889 		return (EINVAL);
890 	}
891 
892 	spa_upgrade(spa, zc->zc_cookie);
893 	spa_close(spa, FTAG);
894 
895 	return (error);
896 }
897 
898 static int
899 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
900 {
901 	spa_t *spa;
902 	char *hist_buf;
903 	uint64_t size;
904 	int error;
905 
906 	if ((size = zc->zc_history_len) == 0)
907 		return (EINVAL);
908 
909 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
910 		return (error);
911 
912 	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
913 		spa_close(spa, FTAG);
914 		return (ENOTSUP);
915 	}
916 
917 	hist_buf = kmem_alloc(size, KM_SLEEP);
918 	if ((error = spa_history_get(spa, &zc->zc_history_offset,
919 	    &zc->zc_history_len, hist_buf)) == 0) {
920 		error = xcopyout(hist_buf,
921 		    (char *)(uintptr_t)zc->zc_history,
922 		    zc->zc_history_len);
923 	}
924 
925 	spa_close(spa, FTAG);
926 	kmem_free(hist_buf, size);
927 	return (error);
928 }
929 
930 static int
931 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
932 {
933 	int error;
934 
935 	if (error = dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value))
936 		return (error);
937 
938 	return (0);
939 }
940 
941 static int
942 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
943 {
944 	objset_t *osp;
945 	int error;
946 
947 	if ((error = dmu_objset_open(zc->zc_name, DMU_OST_ZFS,
948 	    DS_MODE_NONE | DS_MODE_READONLY, &osp)) != 0)
949 		return (error);
950 
951 	error = zfs_obj_to_path(osp, zc->zc_obj, zc->zc_value,
952 	    sizeof (zc->zc_value));
953 	dmu_objset_close(osp);
954 
955 	return (error);
956 }
957 
958 static int
959 zfs_ioc_vdev_add(zfs_cmd_t *zc)
960 {
961 	spa_t *spa;
962 	int error;
963 	nvlist_t *config, **l2cache;
964 	uint_t nl2cache;
965 
966 	error = spa_open(zc->zc_name, &spa, FTAG);
967 	if (error != 0)
968 		return (error);
969 
970 	error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
971 	    &config);
972 	(void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
973 	    &l2cache, &nl2cache);
974 
975 	/*
976 	 * A root pool with concatenated devices is not supported.
977 	 * Thus, can not add a device to a root pool with one device.
978 	 * Allow for l2cache devices to be added.
979 	 */
980 	if (spa->spa_root_vdev->vdev_children == 1 && spa->spa_bootfs != 0 &&
981 	    nl2cache == 0) {
982 		spa_close(spa, FTAG);
983 		return (EDOM);
984 	}
985 
986 	if (error == 0) {
987 		error = spa_vdev_add(spa, config);
988 		nvlist_free(config);
989 	}
990 	spa_close(spa, FTAG);
991 	return (error);
992 }
993 
994 static int
995 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
996 {
997 	spa_t *spa;
998 	int error;
999 
1000 	error = spa_open(zc->zc_name, &spa, FTAG);
1001 	if (error != 0)
1002 		return (error);
1003 	error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1004 	spa_close(spa, FTAG);
1005 	return (error);
1006 }
1007 
1008 static int
1009 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1010 {
1011 	spa_t *spa;
1012 	int error;
1013 	vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1014 
1015 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1016 		return (error);
1017 	switch (zc->zc_cookie) {
1018 	case VDEV_STATE_ONLINE:
1019 		error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1020 		break;
1021 
1022 	case VDEV_STATE_OFFLINE:
1023 		error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1024 		break;
1025 
1026 	case VDEV_STATE_FAULTED:
1027 		error = vdev_fault(spa, zc->zc_guid);
1028 		break;
1029 
1030 	case VDEV_STATE_DEGRADED:
1031 		error = vdev_degrade(spa, zc->zc_guid);
1032 		break;
1033 
1034 	default:
1035 		error = EINVAL;
1036 	}
1037 	zc->zc_cookie = newstate;
1038 	spa_close(spa, FTAG);
1039 	return (error);
1040 }
1041 
1042 static int
1043 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1044 {
1045 	spa_t *spa;
1046 	int replacing = zc->zc_cookie;
1047 	nvlist_t *config;
1048 	int error;
1049 
1050 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1051 		return (error);
1052 
1053 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1054 	    &config)) == 0) {
1055 		error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1056 		nvlist_free(config);
1057 	}
1058 
1059 	spa_close(spa, FTAG);
1060 	return (error);
1061 }
1062 
1063 static int
1064 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1065 {
1066 	spa_t *spa;
1067 	int error;
1068 
1069 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1070 		return (error);
1071 
1072 	error = spa_vdev_detach(spa, zc->zc_guid, B_FALSE);
1073 
1074 	spa_close(spa, FTAG);
1075 	return (error);
1076 }
1077 
1078 static int
1079 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
1080 {
1081 	spa_t *spa;
1082 	char *path = zc->zc_value;
1083 	uint64_t guid = zc->zc_guid;
1084 	int error;
1085 
1086 	error = spa_open(zc->zc_name, &spa, FTAG);
1087 	if (error != 0)
1088 		return (error);
1089 
1090 	error = spa_vdev_setpath(spa, guid, path);
1091 	spa_close(spa, FTAG);
1092 	return (error);
1093 }
1094 
1095 static int
1096 zfs_os_open_retry(char *name, objset_t **os)
1097 {
1098 	int error;
1099 
1100 retry:
1101 	error = dmu_objset_open(name, DMU_OST_ANY,
1102 	    DS_MODE_STANDARD | DS_MODE_READONLY, os);
1103 	if (error != 0) {
1104 		/*
1105 		 * This is ugly: dmu_objset_open() can return EBUSY if
1106 		 * the objset is held exclusively. Fortunately this hold is
1107 		 * only for a short while, so we retry here.
1108 		 * This avoids user code having to handle EBUSY,
1109 		 * for example for a "zfs list".
1110 		 */
1111 		if (error == EBUSY) {
1112 			delay(1);
1113 			goto retry;
1114 		}
1115 	}
1116 	return (error);
1117 }
1118 
1119 /*
1120  * inputs:
1121  * zc_name		name of filesystem
1122  * zc_nvlist_dst_size	size of buffer for property nvlist
1123  *
1124  * outputs:
1125  * zc_objset_stats	stats
1126  * zc_nvlist_dst	property nvlist
1127  * zc_nvlist_dst_size	size of property nvlist
1128  * zc_value		alternate root
1129  */
1130 static int
1131 zfs_ioc_objset_stats(zfs_cmd_t *zc)
1132 {
1133 	objset_t *os = NULL;
1134 	int error;
1135 	nvlist_t *nv;
1136 
1137 	if ((error = zfs_os_open_retry(zc->zc_name, &os)) != 0)
1138 		return (error);
1139 
1140 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1141 
1142 	if (zc->zc_nvlist_dst != 0 &&
1143 	    (error = dsl_prop_get_all(os, &nv)) == 0) {
1144 		dmu_objset_stats(os, nv);
1145 		/*
1146 		 * NB: zvol_get_stats() will read the objset contents,
1147 		 * which we aren't supposed to do with a
1148 		 * DS_MODE_STANDARD open, because it could be
1149 		 * inconsistent.  So this is a bit of a workaround...
1150 		 */
1151 		if (!zc->zc_objset_stats.dds_inconsistent) {
1152 			if (dmu_objset_type(os) == DMU_OST_ZVOL)
1153 				VERIFY(zvol_get_stats(os, nv) == 0);
1154 		}
1155 		error = put_nvlist(zc, nv);
1156 		nvlist_free(nv);
1157 	}
1158 
1159 	spa_altroot(dmu_objset_spa(os), zc->zc_value, sizeof (zc->zc_value));
1160 
1161 	dmu_objset_close(os);
1162 	return (error);
1163 }
1164 
1165 static int
1166 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
1167 {
1168 	uint64_t value;
1169 	int error;
1170 
1171 	/*
1172 	 * zfs_get_zplprop() will either find a value or give us
1173 	 * the default value (if there is one).
1174 	 */
1175 	if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
1176 		return (error);
1177 	VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
1178 	return (0);
1179 }
1180 
1181 /*
1182  * inputs:
1183  * zc_name		name of filesystem
1184  * zc_nvlist_dst_size	size of buffer for zpl property nvlist
1185  *
1186  * outputs:
1187  * zc_nvlist_dst	zpl property nvlist
1188  * zc_nvlist_dst_size	size of zpl property nvlist
1189  */
1190 static int
1191 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
1192 {
1193 	objset_t *os;
1194 	int err;
1195 
1196 	if ((err = zfs_os_open_retry(zc->zc_name, &os)) != 0)
1197 		return (err);
1198 
1199 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1200 
1201 	/*
1202 	 * NB: nvl_add_zplprop() will read the objset contents,
1203 	 * which we aren't supposed to do with a DS_MODE_STANDARD
1204 	 * open, because it could be inconsistent.
1205 	 */
1206 	if (zc->zc_nvlist_dst != NULL &&
1207 	    !zc->zc_objset_stats.dds_inconsistent &&
1208 	    dmu_objset_type(os) == DMU_OST_ZFS) {
1209 		nvlist_t *nv;
1210 
1211 		VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1212 		if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
1213 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
1214 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
1215 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
1216 			err = put_nvlist(zc, nv);
1217 		nvlist_free(nv);
1218 	} else {
1219 		err = ENOENT;
1220 	}
1221 	dmu_objset_close(os);
1222 	return (err);
1223 }
1224 
1225 /*
1226  * inputs:
1227  * zc_name		name of filesystem
1228  * zc_cookie		zap cursor
1229  * zc_nvlist_dst_size	size of buffer for property nvlist
1230  *
1231  * outputs:
1232  * zc_name		name of next filesystem
1233  * zc_objset_stats	stats
1234  * zc_nvlist_dst	property nvlist
1235  * zc_nvlist_dst_size	size of property nvlist
1236  * zc_value		alternate root
1237  */
1238 static int
1239 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
1240 {
1241 	objset_t *os;
1242 	int error;
1243 	char *p;
1244 
1245 	if ((error = zfs_os_open_retry(zc->zc_name, &os)) != 0) {
1246 		if (error == ENOENT)
1247 			error = ESRCH;
1248 		return (error);
1249 	}
1250 
1251 	p = strrchr(zc->zc_name, '/');
1252 	if (p == NULL || p[1] != '\0')
1253 		(void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
1254 	p = zc->zc_name + strlen(zc->zc_name);
1255 
1256 	do {
1257 		error = dmu_dir_list_next(os,
1258 		    sizeof (zc->zc_name) - (p - zc->zc_name), p,
1259 		    NULL, &zc->zc_cookie);
1260 		if (error == ENOENT)
1261 			error = ESRCH;
1262 	} while (error == 0 && !INGLOBALZONE(curproc) &&
1263 	    !zone_dataset_visible(zc->zc_name, NULL));
1264 
1265 	/*
1266 	 * If it's a hidden dataset (ie. with a '$' in its name), don't
1267 	 * try to get stats for it.  Userland will skip over it.
1268 	 */
1269 	if (error == 0 && strchr(zc->zc_name, '$') == NULL)
1270 		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
1271 
1272 	dmu_objset_close(os);
1273 	return (error);
1274 }
1275 
1276 /*
1277  * inputs:
1278  * zc_name		name of filesystem
1279  * zc_cookie		zap cursor
1280  * zc_nvlist_dst_size	size of buffer for property nvlist
1281  *
1282  * outputs:
1283  * zc_name		name of next snapshot
1284  * zc_objset_stats	stats
1285  * zc_nvlist_dst	property nvlist
1286  * zc_nvlist_dst_size	size of property nvlist
1287  * zc_value		alternate root
1288  */
1289 static int
1290 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
1291 {
1292 	objset_t *os;
1293 	int error;
1294 
1295 	if ((error = zfs_os_open_retry(zc->zc_name, &os)) != 0) {
1296 		if (error == ENOENT)
1297 			error = ESRCH;
1298 		return (error);
1299 	}
1300 
1301 	/*
1302 	 * A dataset name of maximum length cannot have any snapshots,
1303 	 * so exit immediately.
1304 	 */
1305 	if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
1306 		dmu_objset_close(os);
1307 		return (ESRCH);
1308 	}
1309 
1310 	error = dmu_snapshot_list_next(os,
1311 	    sizeof (zc->zc_name) - strlen(zc->zc_name),
1312 	    zc->zc_name + strlen(zc->zc_name), NULL, &zc->zc_cookie, NULL);
1313 	if (error == ENOENT)
1314 		error = ESRCH;
1315 
1316 	if (error == 0)
1317 		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
1318 
1319 	/* if we failed, undo the @ that we tacked on to zc_name */
1320 	if (error != 0)
1321 		*strchr(zc->zc_name, '@') = '\0';
1322 
1323 	dmu_objset_close(os);
1324 	return (error);
1325 }
1326 
1327 static int
1328 zfs_set_prop_nvlist(const char *name, nvlist_t *nvl)
1329 {
1330 	nvpair_t *elem;
1331 	int error;
1332 	uint64_t intval;
1333 	char *strval;
1334 
1335 	/*
1336 	 * First validate permission to set all of the properties
1337 	 */
1338 	elem = NULL;
1339 	while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) {
1340 		const char *propname = nvpair_name(elem);
1341 		zfs_prop_t prop = zfs_name_to_prop(propname);
1342 
1343 		if (prop == ZPROP_INVAL) {
1344 			/*
1345 			 * If this is a user-defined property, it must be a
1346 			 * string, and there is no further validation to do.
1347 			 */
1348 			if (!zfs_prop_user(propname) ||
1349 			    nvpair_type(elem) != DATA_TYPE_STRING)
1350 				return (EINVAL);
1351 
1352 			if (error = zfs_secpolicy_write_perms(name,
1353 			    ZFS_DELEG_PERM_USERPROP, CRED()))
1354 				return (error);
1355 			continue;
1356 		}
1357 
1358 		if ((error = zfs_secpolicy_setprop(name, prop, CRED())) != 0)
1359 			return (error);
1360 
1361 		/*
1362 		 * Check that this value is valid for this pool version
1363 		 */
1364 		switch (prop) {
1365 		case ZFS_PROP_COMPRESSION:
1366 			/*
1367 			 * If the user specified gzip compression, make sure
1368 			 * the SPA supports it. We ignore any errors here since
1369 			 * we'll catch them later.
1370 			 */
1371 			if (nvpair_type(elem) == DATA_TYPE_UINT64 &&
1372 			    nvpair_value_uint64(elem, &intval) == 0 &&
1373 			    intval >= ZIO_COMPRESS_GZIP_1 &&
1374 			    intval <= ZIO_COMPRESS_GZIP_9) {
1375 				if (zfs_check_version(name,
1376 				    SPA_VERSION_GZIP_COMPRESSION))
1377 					return (ENOTSUP);
1378 			}
1379 			break;
1380 
1381 		case ZFS_PROP_COPIES:
1382 			if (zfs_check_version(name, SPA_VERSION_DITTO_BLOCKS))
1383 				return (ENOTSUP);
1384 			break;
1385 		}
1386 		if ((error = zfs_secpolicy_setprop(name, prop, CRED())) != 0)
1387 			return (error);
1388 	}
1389 
1390 	elem = NULL;
1391 	while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) {
1392 		const char *propname = nvpair_name(elem);
1393 		zfs_prop_t prop = zfs_name_to_prop(propname);
1394 
1395 		if (prop == ZPROP_INVAL) {
1396 			VERIFY(nvpair_value_string(elem, &strval) == 0);
1397 			error = dsl_prop_set(name, propname, 1,
1398 			    strlen(strval) + 1, strval);
1399 			if (error == 0)
1400 				continue;
1401 			else
1402 				return (error);
1403 		}
1404 
1405 		switch (prop) {
1406 		case ZFS_PROP_QUOTA:
1407 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1408 			    (error = dsl_dir_set_quota(name, intval)) != 0)
1409 				return (error);
1410 			break;
1411 
1412 		case ZFS_PROP_REFQUOTA:
1413 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1414 			    (error = dsl_dataset_set_quota(name, intval)) != 0)
1415 				return (error);
1416 			break;
1417 
1418 		case ZFS_PROP_RESERVATION:
1419 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1420 			    (error = dsl_dir_set_reservation(name,
1421 			    intval)) != 0)
1422 				return (error);
1423 			break;
1424 
1425 		case ZFS_PROP_REFRESERVATION:
1426 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1427 			    (error = dsl_dataset_set_reservation(name,
1428 			    intval)) != 0)
1429 				return (error);
1430 			break;
1431 
1432 		case ZFS_PROP_VOLSIZE:
1433 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1434 			    (error = zvol_set_volsize(name,
1435 			    ddi_driver_major(zfs_dip), intval)) != 0)
1436 				return (error);
1437 			break;
1438 
1439 		case ZFS_PROP_VOLBLOCKSIZE:
1440 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1441 			    (error = zvol_set_volblocksize(name, intval)) != 0)
1442 				return (error);
1443 			break;
1444 
1445 		case ZFS_PROP_VERSION:
1446 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1447 			    (error = zfs_set_version(name, intval)) != 0)
1448 				return (error);
1449 			break;
1450 
1451 		default:
1452 			if (nvpair_type(elem) == DATA_TYPE_STRING) {
1453 				if (zfs_prop_get_type(prop) !=
1454 				    PROP_TYPE_STRING)
1455 					return (EINVAL);
1456 				VERIFY(nvpair_value_string(elem, &strval) == 0);
1457 				if ((error = dsl_prop_set(name,
1458 				    nvpair_name(elem), 1, strlen(strval) + 1,
1459 				    strval)) != 0)
1460 					return (error);
1461 			} else if (nvpair_type(elem) == DATA_TYPE_UINT64) {
1462 				const char *unused;
1463 
1464 				VERIFY(nvpair_value_uint64(elem, &intval) == 0);
1465 
1466 				switch (zfs_prop_get_type(prop)) {
1467 				case PROP_TYPE_NUMBER:
1468 					break;
1469 				case PROP_TYPE_STRING:
1470 					return (EINVAL);
1471 				case PROP_TYPE_INDEX:
1472 					if (zfs_prop_index_to_string(prop,
1473 					    intval, &unused) != 0)
1474 						return (EINVAL);
1475 					break;
1476 				default:
1477 					cmn_err(CE_PANIC,
1478 					    "unknown property type");
1479 					break;
1480 				}
1481 
1482 				if ((error = dsl_prop_set(name, propname,
1483 				    8, 1, &intval)) != 0)
1484 					return (error);
1485 			} else {
1486 				return (EINVAL);
1487 			}
1488 			break;
1489 		}
1490 	}
1491 
1492 	return (0);
1493 }
1494 
1495 /*
1496  * inputs:
1497  * zc_name		name of filesystem
1498  * zc_value		name of property to inherit
1499  * zc_nvlist_src{_size}	nvlist of properties to apply
1500  *
1501  * outputs:		none
1502  */
1503 static int
1504 zfs_ioc_set_prop(zfs_cmd_t *zc)
1505 {
1506 	nvlist_t *nvl;
1507 	int error;
1508 
1509 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1510 	    &nvl)) != 0)
1511 		return (error);
1512 
1513 	error = zfs_set_prop_nvlist(zc->zc_name, nvl);
1514 
1515 	nvlist_free(nvl);
1516 	return (error);
1517 }
1518 
1519 /*
1520  * inputs:
1521  * zc_name		name of filesystem
1522  * zc_value		name of property to inherit
1523  *
1524  * outputs:		none
1525  */
1526 static int
1527 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
1528 {
1529 	/* the property name has been validated by zfs_secpolicy_inherit() */
1530 	return (dsl_prop_set(zc->zc_name, zc->zc_value, 0, 0, NULL));
1531 }
1532 
1533 static int
1534 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
1535 {
1536 	nvlist_t *props;
1537 	spa_t *spa;
1538 	int error;
1539 
1540 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1541 	    &props)))
1542 		return (error);
1543 
1544 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
1545 		nvlist_free(props);
1546 		return (error);
1547 	}
1548 
1549 	error = spa_prop_set(spa, props);
1550 
1551 	nvlist_free(props);
1552 	spa_close(spa, FTAG);
1553 
1554 	return (error);
1555 }
1556 
1557 static int
1558 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
1559 {
1560 	spa_t *spa;
1561 	int error;
1562 	nvlist_t *nvp = NULL;
1563 
1564 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1565 		return (error);
1566 
1567 	error = spa_prop_get(spa, &nvp);
1568 
1569 	if (error == 0 && zc->zc_nvlist_dst != NULL)
1570 		error = put_nvlist(zc, nvp);
1571 	else
1572 		error = EFAULT;
1573 
1574 	spa_close(spa, FTAG);
1575 
1576 	if (nvp)
1577 		nvlist_free(nvp);
1578 	return (error);
1579 }
1580 
1581 static int
1582 zfs_ioc_iscsi_perm_check(zfs_cmd_t *zc)
1583 {
1584 	nvlist_t *nvp;
1585 	int error;
1586 	uint32_t uid;
1587 	uint32_t gid;
1588 	uint32_t *groups;
1589 	uint_t group_cnt;
1590 	cred_t	*usercred;
1591 
1592 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1593 	    &nvp)) != 0) {
1594 		return (error);
1595 	}
1596 
1597 	if ((error = nvlist_lookup_uint32(nvp,
1598 	    ZFS_DELEG_PERM_UID, &uid)) != 0) {
1599 		nvlist_free(nvp);
1600 		return (EPERM);
1601 	}
1602 
1603 	if ((error = nvlist_lookup_uint32(nvp,
1604 	    ZFS_DELEG_PERM_GID, &gid)) != 0) {
1605 		nvlist_free(nvp);
1606 		return (EPERM);
1607 	}
1608 
1609 	if ((error = nvlist_lookup_uint32_array(nvp, ZFS_DELEG_PERM_GROUPS,
1610 	    &groups, &group_cnt)) != 0) {
1611 		nvlist_free(nvp);
1612 		return (EPERM);
1613 	}
1614 	usercred = cralloc();
1615 	if ((crsetugid(usercred, uid, gid) != 0) ||
1616 	    (crsetgroups(usercred, group_cnt, (gid_t *)groups) != 0)) {
1617 		nvlist_free(nvp);
1618 		crfree(usercred);
1619 		return (EPERM);
1620 	}
1621 	nvlist_free(nvp);
1622 	error = dsl_deleg_access(zc->zc_name,
1623 	    zfs_prop_to_name(ZFS_PROP_SHAREISCSI), usercred);
1624 	crfree(usercred);
1625 	return (error);
1626 }
1627 
1628 /*
1629  * inputs:
1630  * zc_name		name of filesystem
1631  * zc_nvlist_src{_size}	nvlist of delegated permissions
1632  * zc_perm_action	allow/unallow flag
1633  *
1634  * outputs:		none
1635  */
1636 static int
1637 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
1638 {
1639 	int error;
1640 	nvlist_t *fsaclnv = NULL;
1641 
1642 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1643 	    &fsaclnv)) != 0)
1644 		return (error);
1645 
1646 	/*
1647 	 * Verify nvlist is constructed correctly
1648 	 */
1649 	if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
1650 		nvlist_free(fsaclnv);
1651 		return (EINVAL);
1652 	}
1653 
1654 	/*
1655 	 * If we don't have PRIV_SYS_MOUNT, then validate
1656 	 * that user is allowed to hand out each permission in
1657 	 * the nvlist(s)
1658 	 */
1659 
1660 	error = secpolicy_zfs(CRED());
1661 	if (error) {
1662 		if (zc->zc_perm_action == B_FALSE) {
1663 			error = dsl_deleg_can_allow(zc->zc_name,
1664 			    fsaclnv, CRED());
1665 		} else {
1666 			error = dsl_deleg_can_unallow(zc->zc_name,
1667 			    fsaclnv, CRED());
1668 		}
1669 	}
1670 
1671 	if (error == 0)
1672 		error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
1673 
1674 	nvlist_free(fsaclnv);
1675 	return (error);
1676 }
1677 
1678 /*
1679  * inputs:
1680  * zc_name		name of filesystem
1681  *
1682  * outputs:
1683  * zc_nvlist_src{_size}	nvlist of delegated permissions
1684  */
1685 static int
1686 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
1687 {
1688 	nvlist_t *nvp;
1689 	int error;
1690 
1691 	if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
1692 		error = put_nvlist(zc, nvp);
1693 		nvlist_free(nvp);
1694 	}
1695 
1696 	return (error);
1697 }
1698 
1699 /*
1700  * inputs:
1701  * zc_name		name of volume
1702  *
1703  * outputs:		none
1704  */
1705 static int
1706 zfs_ioc_create_minor(zfs_cmd_t *zc)
1707 {
1708 	return (zvol_create_minor(zc->zc_name, ddi_driver_major(zfs_dip)));
1709 }
1710 
1711 /*
1712  * inputs:
1713  * zc_name		name of volume
1714  *
1715  * outputs:		none
1716  */
1717 static int
1718 zfs_ioc_remove_minor(zfs_cmd_t *zc)
1719 {
1720 	return (zvol_remove_minor(zc->zc_name));
1721 }
1722 
1723 /*
1724  * Search the vfs list for a specified resource.  Returns a pointer to it
1725  * or NULL if no suitable entry is found. The caller of this routine
1726  * is responsible for releasing the returned vfs pointer.
1727  */
1728 static vfs_t *
1729 zfs_get_vfs(const char *resource)
1730 {
1731 	struct vfs *vfsp;
1732 	struct vfs *vfs_found = NULL;
1733 
1734 	vfs_list_read_lock();
1735 	vfsp = rootvfs;
1736 	do {
1737 		if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
1738 			VFS_HOLD(vfsp);
1739 			vfs_found = vfsp;
1740 			break;
1741 		}
1742 		vfsp = vfsp->vfs_next;
1743 	} while (vfsp != rootvfs);
1744 	vfs_list_unlock();
1745 	return (vfs_found);
1746 }
1747 
1748 /* ARGSUSED */
1749 static void
1750 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
1751 {
1752 	zfs_creat_t *zct = arg;
1753 
1754 	zfs_create_fs(os, cr, zct->zct_zplprops, tx);
1755 }
1756 
1757 #define	ZFS_PROP_UNDEFINED	((uint64_t)-1)
1758 
1759 /*
1760  * inputs:
1761  * createprops	list of properties requested by creator
1762  * dataset	name of dataset we are creating
1763  *
1764  * outputs:
1765  * zplprops	values for the zplprops we attach to the master node object
1766  *
1767  * Determine the settings for utf8only, normalization and
1768  * casesensitivity.  Specific values may have been requested by the
1769  * creator and/or we can inherit values from the parent dataset.  If
1770  * the file system is of too early a vintage, a creator can not
1771  * request settings for these properties, even if the requested
1772  * setting is the default value.  We don't actually want to create dsl
1773  * properties for these, so remove them from the source nvlist after
1774  * processing.
1775  */
1776 static int
1777 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
1778     nvlist_t *zplprops, uint64_t zplver)
1779 {
1780 	objset_t *os;
1781 	char parentname[MAXNAMELEN];
1782 	char *cp;
1783 	uint64_t sense = ZFS_PROP_UNDEFINED;
1784 	uint64_t norm = ZFS_PROP_UNDEFINED;
1785 	uint64_t u8 = ZFS_PROP_UNDEFINED;
1786 	int error = 0;
1787 
1788 	ASSERT(zplprops != NULL);
1789 
1790 	(void) strlcpy(parentname, dataset, sizeof (parentname));
1791 	cp = strrchr(parentname, '/');
1792 	ASSERT(cp != NULL);
1793 	cp[0] = '\0';
1794 
1795 	/*
1796 	 * Pull out creator prop choices, if any.
1797 	 */
1798 	if (createprops) {
1799 		(void) nvlist_lookup_uint64(createprops,
1800 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
1801 		(void) nvlist_remove_all(createprops,
1802 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE));
1803 		(void) nvlist_lookup_uint64(createprops,
1804 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
1805 		(void) nvlist_remove_all(createprops,
1806 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
1807 		(void) nvlist_lookup_uint64(createprops,
1808 		    zfs_prop_to_name(ZFS_PROP_CASE), &sense);
1809 		(void) nvlist_remove_all(createprops,
1810 		    zfs_prop_to_name(ZFS_PROP_CASE));
1811 	}
1812 
1813 	/*
1814 	 * If the file system or pool is version is too "young" to
1815 	 * support normalization and the creator tried to set a value
1816 	 * for one of the props, error out.  We only need check the
1817 	 * ZPL version because we've already checked by now that the
1818 	 * SPA version is compatible with the selected ZPL version.
1819 	 */
1820 	if (zplver < ZPL_VERSION_NORMALIZATION &&
1821 	    (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
1822 	    sense != ZFS_PROP_UNDEFINED))
1823 		return (ENOTSUP);
1824 
1825 	/*
1826 	 * Put the version in the zplprops
1827 	 */
1828 	VERIFY(nvlist_add_uint64(zplprops,
1829 	    zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
1830 
1831 	/*
1832 	 * Open parent object set so we can inherit zplprop values if
1833 	 * necessary.
1834 	 */
1835 	if ((error = zfs_os_open_retry(parentname, &os)) != 0)
1836 		return (error);
1837 
1838 	if (norm == ZFS_PROP_UNDEFINED)
1839 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
1840 	VERIFY(nvlist_add_uint64(zplprops,
1841 	    zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
1842 
1843 	/*
1844 	 * If we're normalizing, names must always be valid UTF-8 strings.
1845 	 */
1846 	if (norm)
1847 		u8 = 1;
1848 	if (u8 == ZFS_PROP_UNDEFINED)
1849 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
1850 	VERIFY(nvlist_add_uint64(zplprops,
1851 	    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
1852 
1853 	if (sense == ZFS_PROP_UNDEFINED)
1854 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
1855 	VERIFY(nvlist_add_uint64(zplprops,
1856 	    zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
1857 
1858 	dmu_objset_close(os);
1859 	return (0);
1860 }
1861 
1862 /*
1863  * inputs:
1864  * zc_objset_type	type of objset to create (fs vs zvol)
1865  * zc_name		name of new objset
1866  * zc_value		name of snapshot to clone from (may be empty)
1867  * zc_nvlist_src{_size}	nvlist of properties to apply
1868  *
1869  * outputs: none
1870  */
1871 static int
1872 zfs_ioc_create(zfs_cmd_t *zc)
1873 {
1874 	objset_t *clone;
1875 	int error = 0;
1876 	zfs_creat_t zct;
1877 	nvlist_t *nvprops = NULL;
1878 	void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
1879 	dmu_objset_type_t type = zc->zc_objset_type;
1880 
1881 	switch (type) {
1882 
1883 	case DMU_OST_ZFS:
1884 		cbfunc = zfs_create_cb;
1885 		break;
1886 
1887 	case DMU_OST_ZVOL:
1888 		cbfunc = zvol_create_cb;
1889 		break;
1890 
1891 	default:
1892 		cbfunc = NULL;
1893 	}
1894 	if (strchr(zc->zc_name, '@') ||
1895 	    strchr(zc->zc_name, '%'))
1896 		return (EINVAL);
1897 
1898 	if (zc->zc_nvlist_src != NULL &&
1899 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1900 	    &nvprops)) != 0)
1901 		return (error);
1902 
1903 	zct.zct_zplprops = NULL;
1904 	zct.zct_props = nvprops;
1905 
1906 	if (zc->zc_value[0] != '\0') {
1907 		/*
1908 		 * We're creating a clone of an existing snapshot.
1909 		 */
1910 		zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
1911 		if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0) {
1912 			nvlist_free(nvprops);
1913 			return (EINVAL);
1914 		}
1915 
1916 		error = dmu_objset_open(zc->zc_value, type,
1917 		    DS_MODE_STANDARD | DS_MODE_READONLY, &clone);
1918 		if (error) {
1919 			nvlist_free(nvprops);
1920 			return (error);
1921 		}
1922 		error = dmu_objset_create(zc->zc_name, type, clone, NULL, NULL);
1923 		if (error) {
1924 			dmu_objset_close(clone);
1925 			nvlist_free(nvprops);
1926 			return (error);
1927 		}
1928 		dmu_objset_close(clone);
1929 	} else {
1930 		if (cbfunc == NULL) {
1931 			nvlist_free(nvprops);
1932 			return (EINVAL);
1933 		}
1934 
1935 		if (type == DMU_OST_ZVOL) {
1936 			uint64_t volsize, volblocksize;
1937 
1938 			if (nvprops == NULL ||
1939 			    nvlist_lookup_uint64(nvprops,
1940 			    zfs_prop_to_name(ZFS_PROP_VOLSIZE),
1941 			    &volsize) != 0) {
1942 				nvlist_free(nvprops);
1943 				return (EINVAL);
1944 			}
1945 
1946 			if ((error = nvlist_lookup_uint64(nvprops,
1947 			    zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
1948 			    &volblocksize)) != 0 && error != ENOENT) {
1949 				nvlist_free(nvprops);
1950 				return (EINVAL);
1951 			}
1952 
1953 			if (error != 0)
1954 				volblocksize = zfs_prop_default_numeric(
1955 				    ZFS_PROP_VOLBLOCKSIZE);
1956 
1957 			if ((error = zvol_check_volblocksize(
1958 			    volblocksize)) != 0 ||
1959 			    (error = zvol_check_volsize(volsize,
1960 			    volblocksize)) != 0) {
1961 				nvlist_free(nvprops);
1962 				return (error);
1963 			}
1964 		} else if (type == DMU_OST_ZFS) {
1965 			uint64_t version;
1966 			int error;
1967 
1968 			/*
1969 			 * Default ZPL version to non-FUID capable if the
1970 			 * pool is not upgraded to support FUIDs.
1971 			 */
1972 			if (zfs_check_version(zc->zc_name, SPA_VERSION_FUID))
1973 				version = ZPL_VERSION_FUID - 1;
1974 			else
1975 				version = ZPL_VERSION;
1976 
1977 			/*
1978 			 * Potentially override default ZPL version based
1979 			 * on creator's request.
1980 			 */
1981 			(void) nvlist_lookup_uint64(nvprops,
1982 			    zfs_prop_to_name(ZFS_PROP_VERSION), &version);
1983 
1984 			/*
1985 			 * Make sure version we ended up with is kosher
1986 			 */
1987 			if ((version < ZPL_VERSION_INITIAL ||
1988 			    version > ZPL_VERSION) ||
1989 			    (version >= ZPL_VERSION_FUID &&
1990 			    zfs_check_version(zc->zc_name, SPA_VERSION_FUID))) {
1991 				nvlist_free(nvprops);
1992 				return (ENOTSUP);
1993 			}
1994 
1995 			/*
1996 			 * We have to have normalization and
1997 			 * case-folding flags correct when we do the
1998 			 * file system creation, so go figure them out
1999 			 * now.
2000 			 */
2001 			VERIFY(nvlist_alloc(&zct.zct_zplprops,
2002 			    NV_UNIQUE_NAME, KM_SLEEP) == 0);
2003 			error = zfs_fill_zplprops(zc->zc_name, nvprops,
2004 			    zct.zct_zplprops, version);
2005 			if (error != 0) {
2006 				nvlist_free(nvprops);
2007 				nvlist_free(zct.zct_zplprops);
2008 				return (error);
2009 			}
2010 		}
2011 		error = dmu_objset_create(zc->zc_name, type, NULL, cbfunc,
2012 		    &zct);
2013 		nvlist_free(zct.zct_zplprops);
2014 	}
2015 
2016 	/*
2017 	 * It would be nice to do this atomically.
2018 	 */
2019 	if (error == 0) {
2020 		if ((error = zfs_set_prop_nvlist(zc->zc_name, nvprops)) != 0)
2021 			(void) dmu_objset_destroy(zc->zc_name);
2022 	}
2023 
2024 	nvlist_free(nvprops);
2025 	return (error);
2026 }
2027 
2028 /*
2029  * inputs:
2030  * zc_name	name of filesystem
2031  * zc_value	short name of snapshot
2032  * zc_cookie	recursive flag
2033  *
2034  * outputs:	none
2035  */
2036 static int
2037 zfs_ioc_snapshot(zfs_cmd_t *zc)
2038 {
2039 	if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
2040 		return (EINVAL);
2041 	return (dmu_objset_snapshot(zc->zc_name,
2042 	    zc->zc_value, zc->zc_cookie));
2043 }
2044 
2045 int
2046 zfs_unmount_snap(char *name, void *arg)
2047 {
2048 	char *snapname = arg;
2049 	char *cp;
2050 	vfs_t *vfsp = NULL;
2051 
2052 	/*
2053 	 * Snapshots (which are under .zfs control) must be unmounted
2054 	 * before they can be destroyed.
2055 	 */
2056 
2057 	if (snapname) {
2058 		(void) strcat(name, "@");
2059 		(void) strcat(name, snapname);
2060 		vfsp = zfs_get_vfs(name);
2061 		cp = strchr(name, '@');
2062 		*cp = '\0';
2063 	} else if (strchr(name, '@')) {
2064 		vfsp = zfs_get_vfs(name);
2065 	}
2066 
2067 	if (vfsp) {
2068 		/*
2069 		 * Always force the unmount for snapshots.
2070 		 */
2071 		int flag = MS_FORCE;
2072 		int err;
2073 
2074 		if ((err = vn_vfswlock(vfsp->vfs_vnodecovered)) != 0) {
2075 			VFS_RELE(vfsp);
2076 			return (err);
2077 		}
2078 		VFS_RELE(vfsp);
2079 		if ((err = dounmount(vfsp, flag, kcred)) != 0)
2080 			return (err);
2081 	}
2082 	return (0);
2083 }
2084 
2085 /*
2086  * inputs:
2087  * zc_name	name of filesystem
2088  * zc_value	short name of snapshot
2089  *
2090  * outputs:	none
2091  */
2092 static int
2093 zfs_ioc_destroy_snaps(zfs_cmd_t *zc)
2094 {
2095 	int err;
2096 
2097 	if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
2098 		return (EINVAL);
2099 	err = dmu_objset_find(zc->zc_name,
2100 	    zfs_unmount_snap, zc->zc_value, DS_FIND_CHILDREN);
2101 	if (err)
2102 		return (err);
2103 	return (dmu_snapshots_destroy(zc->zc_name, zc->zc_value));
2104 }
2105 
2106 /*
2107  * inputs:
2108  * zc_name		name of dataset to destroy
2109  * zc_objset_type	type of objset
2110  *
2111  * outputs:		none
2112  */
2113 static int
2114 zfs_ioc_destroy(zfs_cmd_t *zc)
2115 {
2116 	if (strchr(zc->zc_name, '@') && zc->zc_objset_type == DMU_OST_ZFS) {
2117 		int err = zfs_unmount_snap(zc->zc_name, NULL);
2118 		if (err)
2119 			return (err);
2120 	}
2121 
2122 	return (dmu_objset_destroy(zc->zc_name));
2123 }
2124 
2125 /*
2126  * inputs:
2127  * zc_name	name of dataset to rollback (to most recent snapshot)
2128  *
2129  * outputs:	none
2130  */
2131 static int
2132 zfs_ioc_rollback(zfs_cmd_t *zc)
2133 {
2134 	objset_t *os;
2135 	int error;
2136 	zfsvfs_t *zfsvfs = NULL;
2137 
2138 	/*
2139 	 * Get the zfsvfs for the receiving objset. There
2140 	 * won't be one if we're operating on a zvol, if the
2141 	 * objset doesn't exist yet, or is not mounted.
2142 	 */
2143 	error = dmu_objset_open(zc->zc_name, DMU_OST_ANY,
2144 	    DS_MODE_STANDARD, &os);
2145 	if (error)
2146 		return (error);
2147 
2148 	if (dmu_objset_type(os) == DMU_OST_ZFS) {
2149 		mutex_enter(&os->os->os_user_ptr_lock);
2150 		zfsvfs = dmu_objset_get_user(os);
2151 		if (zfsvfs != NULL)
2152 			VFS_HOLD(zfsvfs->z_vfs);
2153 		mutex_exit(&os->os->os_user_ptr_lock);
2154 	}
2155 
2156 	if (zfsvfs != NULL) {
2157 		char osname[MAXNAMELEN];
2158 		int mode;
2159 
2160 		VERIFY3U(0, ==, zfs_suspend_fs(zfsvfs, osname, &mode));
2161 		ASSERT(strcmp(osname, zc->zc_name) == 0);
2162 		error = dmu_objset_rollback(os);
2163 		VERIFY3U(0, ==, zfs_resume_fs(zfsvfs, osname, mode));
2164 
2165 		VFS_RELE(zfsvfs->z_vfs);
2166 	} else {
2167 		error = dmu_objset_rollback(os);
2168 	}
2169 	/* Note, the dmu_objset_rollback() closes the objset for us. */
2170 
2171 	return (error);
2172 }
2173 
2174 /*
2175  * inputs:
2176  * zc_name	old name of dataset
2177  * zc_value	new name of dataset
2178  * zc_cookie	recursive flag (only valid for snapshots)
2179  *
2180  * outputs:	none
2181  */
2182 static int
2183 zfs_ioc_rename(zfs_cmd_t *zc)
2184 {
2185 	boolean_t recursive = zc->zc_cookie & 1;
2186 
2187 	zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
2188 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
2189 	    strchr(zc->zc_value, '%'))
2190 		return (EINVAL);
2191 
2192 	/*
2193 	 * Unmount snapshot unless we're doing a recursive rename,
2194 	 * in which case the dataset code figures out which snapshots
2195 	 * to unmount.
2196 	 */
2197 	if (!recursive && strchr(zc->zc_name, '@') != NULL &&
2198 	    zc->zc_objset_type == DMU_OST_ZFS) {
2199 		int err = zfs_unmount_snap(zc->zc_name, NULL);
2200 		if (err)
2201 			return (err);
2202 	}
2203 
2204 	return (dmu_objset_rename(zc->zc_name, zc->zc_value, recursive));
2205 }
2206 
2207 /*
2208  * inputs:
2209  * zc_name		name of containing filesystem
2210  * zc_nvlist_src{_size}	nvlist of properties to apply
2211  * zc_value		name of snapshot to create
2212  * zc_string		name of clone origin (if DRR_FLAG_CLONE)
2213  * zc_cookie		file descriptor to recv from
2214  * zc_begin_record	the BEGIN record of the stream (not byteswapped)
2215  * zc_guid		force flag
2216  *
2217  * outputs:
2218  * zc_cookie		number of bytes read
2219  */
2220 static int
2221 zfs_ioc_recv(zfs_cmd_t *zc)
2222 {
2223 	file_t *fp;
2224 	objset_t *os;
2225 	dmu_recv_cookie_t drc;
2226 	zfsvfs_t *zfsvfs = NULL;
2227 	boolean_t force = (boolean_t)zc->zc_guid;
2228 	int error, fd;
2229 	offset_t off;
2230 	nvlist_t *props = NULL;
2231 	objset_t *origin = NULL;
2232 	char *tosnap;
2233 	char tofs[ZFS_MAXNAMELEN];
2234 
2235 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
2236 	    strchr(zc->zc_value, '@') == NULL ||
2237 	    strchr(zc->zc_value, '%'))
2238 		return (EINVAL);
2239 
2240 	(void) strcpy(tofs, zc->zc_value);
2241 	tosnap = strchr(tofs, '@');
2242 	*tosnap = '\0';
2243 	tosnap++;
2244 
2245 	if (zc->zc_nvlist_src != NULL &&
2246 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2247 	    &props)) != 0)
2248 		return (error);
2249 
2250 	fd = zc->zc_cookie;
2251 	fp = getf(fd);
2252 	if (fp == NULL) {
2253 		nvlist_free(props);
2254 		return (EBADF);
2255 	}
2256 
2257 	/*
2258 	 * Get the zfsvfs for the receiving objset. There
2259 	 * won't be one if we're operating on a zvol, if the
2260 	 * objset doesn't exist yet, or is not mounted.
2261 	 */
2262 
2263 	error = dmu_objset_open(tofs, DMU_OST_ZFS,
2264 	    DS_MODE_STANDARD | DS_MODE_READONLY, &os);
2265 	if (!error) {
2266 		mutex_enter(&os->os->os_user_ptr_lock);
2267 		zfsvfs = dmu_objset_get_user(os);
2268 		if (zfsvfs != NULL)
2269 			VFS_HOLD(zfsvfs->z_vfs);
2270 		mutex_exit(&os->os->os_user_ptr_lock);
2271 		dmu_objset_close(os);
2272 	}
2273 
2274 	if (zc->zc_string[0]) {
2275 		error = dmu_objset_open(zc->zc_string, DMU_OST_ANY,
2276 		    DS_MODE_STANDARD | DS_MODE_READONLY, &origin);
2277 		if (error) {
2278 			if (zfsvfs != NULL)
2279 				VFS_RELE(zfsvfs->z_vfs);
2280 			nvlist_free(props);
2281 			releasef(fd);
2282 			return (error);
2283 		}
2284 	}
2285 
2286 	error = dmu_recv_begin(tofs, tosnap, &zc->zc_begin_record,
2287 	    force, origin, zfsvfs != NULL, &drc);
2288 	if (origin)
2289 		dmu_objset_close(origin);
2290 	if (error) {
2291 		if (zfsvfs != NULL)
2292 			VFS_RELE(zfsvfs->z_vfs);
2293 		nvlist_free(props);
2294 		releasef(fd);
2295 		return (error);
2296 	}
2297 
2298 	/*
2299 	 * If properties are supplied, they are to completely replace
2300 	 * the existing ones; "inherit" any existing properties.
2301 	 */
2302 	if (props) {
2303 		objset_t *os;
2304 		nvlist_t *nv = NULL;
2305 
2306 		error = dmu_objset_open(tofs, DMU_OST_ANY,
2307 		    DS_MODE_STANDARD | DS_MODE_READONLY | DS_MODE_INCONSISTENT,
2308 		    &os);
2309 		if (error == 0) {
2310 			error = dsl_prop_get_all(os, &nv);
2311 			dmu_objset_close(os);
2312 		}
2313 		if (error == 0) {
2314 			nvpair_t *elem;
2315 			zfs_cmd_t *zc2;
2316 			zc2 = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
2317 
2318 			(void) strcpy(zc2->zc_name, tofs);
2319 			for (elem = nvlist_next_nvpair(nv, NULL); elem;
2320 			    elem = nvlist_next_nvpair(nv, elem)) {
2321 				(void) strcpy(zc2->zc_value, nvpair_name(elem));
2322 				if (zfs_secpolicy_inherit(zc2, CRED()) == 0)
2323 					(void) zfs_ioc_inherit_prop(zc2);
2324 			}
2325 			kmem_free(zc2, sizeof (zfs_cmd_t));
2326 		}
2327 		if (nv)
2328 			nvlist_free(nv);
2329 	}
2330 
2331 	/*
2332 	 * Set properties.  Note, we ignore errors.  Would be better to
2333 	 * do best-effort in zfs_set_prop_nvlist, too.
2334 	 */
2335 	(void) zfs_set_prop_nvlist(tofs, props);
2336 	nvlist_free(props);
2337 
2338 	off = fp->f_offset;
2339 	error = dmu_recv_stream(&drc, fp->f_vnode, &off);
2340 
2341 	if (error == 0) {
2342 		if (zfsvfs != NULL) {
2343 			char osname[MAXNAMELEN];
2344 			int mode;
2345 
2346 			(void) zfs_suspend_fs(zfsvfs, osname, &mode);
2347 			error = dmu_recv_end(&drc);
2348 			error |= zfs_resume_fs(zfsvfs, osname, mode);
2349 		} else {
2350 			error = dmu_recv_end(&drc);
2351 		}
2352 	}
2353 	if (zfsvfs != NULL)
2354 		VFS_RELE(zfsvfs->z_vfs);
2355 
2356 	zc->zc_cookie = off - fp->f_offset;
2357 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
2358 		fp->f_offset = off;
2359 
2360 	releasef(fd);
2361 	return (error);
2362 }
2363 
2364 /*
2365  * inputs:
2366  * zc_name	name of snapshot to send
2367  * zc_value	short name of incremental fromsnap (may be empty)
2368  * zc_cookie	file descriptor to send stream to
2369  * zc_obj	fromorigin flag (mutually exclusive with zc_value)
2370  *
2371  * outputs: none
2372  */
2373 static int
2374 zfs_ioc_send(zfs_cmd_t *zc)
2375 {
2376 	objset_t *fromsnap = NULL;
2377 	objset_t *tosnap;
2378 	file_t *fp;
2379 	int error;
2380 	offset_t off;
2381 
2382 	error = dmu_objset_open(zc->zc_name, DMU_OST_ANY,
2383 	    DS_MODE_STANDARD | DS_MODE_READONLY, &tosnap);
2384 	if (error)
2385 		return (error);
2386 
2387 	if (zc->zc_value[0] != '\0') {
2388 		char buf[MAXPATHLEN];
2389 		char *cp;
2390 
2391 		(void) strncpy(buf, zc->zc_name, sizeof (buf));
2392 		cp = strchr(buf, '@');
2393 		if (cp)
2394 			*(cp+1) = 0;
2395 		(void) strncat(buf, zc->zc_value, sizeof (buf));
2396 		error = dmu_objset_open(buf, DMU_OST_ANY,
2397 		    DS_MODE_STANDARD | DS_MODE_READONLY, &fromsnap);
2398 		if (error) {
2399 			dmu_objset_close(tosnap);
2400 			return (error);
2401 		}
2402 	}
2403 
2404 	fp = getf(zc->zc_cookie);
2405 	if (fp == NULL) {
2406 		dmu_objset_close(tosnap);
2407 		if (fromsnap)
2408 			dmu_objset_close(fromsnap);
2409 		return (EBADF);
2410 	}
2411 
2412 	off = fp->f_offset;
2413 	error = dmu_sendbackup(tosnap, fromsnap, zc->zc_obj, fp->f_vnode, &off);
2414 
2415 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
2416 		fp->f_offset = off;
2417 	releasef(zc->zc_cookie);
2418 	if (fromsnap)
2419 		dmu_objset_close(fromsnap);
2420 	dmu_objset_close(tosnap);
2421 	return (error);
2422 }
2423 
2424 static int
2425 zfs_ioc_inject_fault(zfs_cmd_t *zc)
2426 {
2427 	int id, error;
2428 
2429 	error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
2430 	    &zc->zc_inject_record);
2431 
2432 	if (error == 0)
2433 		zc->zc_guid = (uint64_t)id;
2434 
2435 	return (error);
2436 }
2437 
2438 static int
2439 zfs_ioc_clear_fault(zfs_cmd_t *zc)
2440 {
2441 	return (zio_clear_fault((int)zc->zc_guid));
2442 }
2443 
2444 static int
2445 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
2446 {
2447 	int id = (int)zc->zc_guid;
2448 	int error;
2449 
2450 	error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
2451 	    &zc->zc_inject_record);
2452 
2453 	zc->zc_guid = id;
2454 
2455 	return (error);
2456 }
2457 
2458 static int
2459 zfs_ioc_error_log(zfs_cmd_t *zc)
2460 {
2461 	spa_t *spa;
2462 	int error;
2463 	size_t count = (size_t)zc->zc_nvlist_dst_size;
2464 
2465 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2466 		return (error);
2467 
2468 	error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
2469 	    &count);
2470 	if (error == 0)
2471 		zc->zc_nvlist_dst_size = count;
2472 	else
2473 		zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
2474 
2475 	spa_close(spa, FTAG);
2476 
2477 	return (error);
2478 }
2479 
2480 static int
2481 zfs_ioc_clear(zfs_cmd_t *zc)
2482 {
2483 	spa_t *spa;
2484 	vdev_t *vd;
2485 	uint64_t txg;
2486 	int error;
2487 
2488 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2489 		return (error);
2490 
2491 	/*
2492 	 * Try to resume any I/Os which may have been suspended
2493 	 * as a result of a complete pool failure.
2494 	 */
2495 	if (!list_is_empty(&spa->spa_zio_list)) {
2496 		if (zio_vdev_resume_io(spa) != 0) {
2497 			spa_close(spa, FTAG);
2498 			return (EIO);
2499 		}
2500 	}
2501 
2502 	txg = spa_vdev_enter(spa);
2503 
2504 	if (zc->zc_guid == 0) {
2505 		vd = NULL;
2506 	} else if ((vd = spa_lookup_by_guid(spa, zc->zc_guid)) == NULL) {
2507 		spa_aux_vdev_t *sav;
2508 		int i;
2509 
2510 		/*
2511 		 * Check if this is an l2cache device.
2512 		 */
2513 		ASSERT(spa != NULL);
2514 		sav = &spa->spa_l2cache;
2515 		for (i = 0; i < sav->sav_count; i++) {
2516 			if (sav->sav_vdevs[i]->vdev_guid == zc->zc_guid) {
2517 				vd = sav->sav_vdevs[i];
2518 				break;
2519 			}
2520 		}
2521 
2522 		if (vd == NULL) {
2523 			(void) spa_vdev_exit(spa, NULL, txg, ENODEV);
2524 			spa_close(spa, FTAG);
2525 			return (ENODEV);
2526 		}
2527 	}
2528 
2529 	vdev_clear(spa, vd, B_TRUE);
2530 
2531 	(void) spa_vdev_exit(spa, NULL, txg, 0);
2532 
2533 	spa_close(spa, FTAG);
2534 
2535 	return (0);
2536 }
2537 
2538 /*
2539  * inputs:
2540  * zc_name	name of filesystem
2541  * zc_value	name of origin snapshot
2542  *
2543  * outputs:	none
2544  */
2545 static int
2546 zfs_ioc_promote(zfs_cmd_t *zc)
2547 {
2548 	char *cp;
2549 
2550 	/*
2551 	 * We don't need to unmount *all* the origin fs's snapshots, but
2552 	 * it's easier.
2553 	 */
2554 	cp = strchr(zc->zc_value, '@');
2555 	if (cp)
2556 		*cp = '\0';
2557 	(void) dmu_objset_find(zc->zc_value,
2558 	    zfs_unmount_snap, NULL, DS_FIND_SNAPSHOTS);
2559 	return (dsl_dataset_promote(zc->zc_name));
2560 }
2561 
2562 /*
2563  * We don't want to have a hard dependency
2564  * against some special symbols in sharefs
2565  * nfs, and smbsrv.  Determine them if needed when
2566  * the first file system is shared.
2567  * Neither sharefs, nfs or smbsrv are unloadable modules.
2568  */
2569 int (*znfsexport_fs)(void *arg);
2570 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
2571 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
2572 
2573 int zfs_nfsshare_inited;
2574 int zfs_smbshare_inited;
2575 
2576 ddi_modhandle_t nfs_mod;
2577 ddi_modhandle_t sharefs_mod;
2578 ddi_modhandle_t smbsrv_mod;
2579 kmutex_t zfs_share_lock;
2580 
2581 static int
2582 zfs_init_sharefs()
2583 {
2584 	int error;
2585 
2586 	ASSERT(MUTEX_HELD(&zfs_share_lock));
2587 	/* Both NFS and SMB shares also require sharetab support. */
2588 	if (sharefs_mod == NULL && ((sharefs_mod =
2589 	    ddi_modopen("fs/sharefs",
2590 	    KRTLD_MODE_FIRST, &error)) == NULL)) {
2591 		return (ENOSYS);
2592 	}
2593 	if (zshare_fs == NULL && ((zshare_fs =
2594 	    (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
2595 	    ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
2596 		return (ENOSYS);
2597 	}
2598 	return (0);
2599 }
2600 
2601 static int
2602 zfs_ioc_share(zfs_cmd_t *zc)
2603 {
2604 	int error;
2605 	int opcode;
2606 
2607 	switch (zc->zc_share.z_sharetype) {
2608 	case ZFS_SHARE_NFS:
2609 	case ZFS_UNSHARE_NFS:
2610 		if (zfs_nfsshare_inited == 0) {
2611 			mutex_enter(&zfs_share_lock);
2612 			if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
2613 			    KRTLD_MODE_FIRST, &error)) == NULL)) {
2614 				mutex_exit(&zfs_share_lock);
2615 				return (ENOSYS);
2616 			}
2617 			if (znfsexport_fs == NULL &&
2618 			    ((znfsexport_fs = (int (*)(void *))
2619 			    ddi_modsym(nfs_mod,
2620 			    "nfs_export", &error)) == NULL)) {
2621 				mutex_exit(&zfs_share_lock);
2622 				return (ENOSYS);
2623 			}
2624 			error = zfs_init_sharefs();
2625 			if (error) {
2626 				mutex_exit(&zfs_share_lock);
2627 				return (ENOSYS);
2628 			}
2629 			zfs_nfsshare_inited = 1;
2630 			mutex_exit(&zfs_share_lock);
2631 		}
2632 		break;
2633 	case ZFS_SHARE_SMB:
2634 	case ZFS_UNSHARE_SMB:
2635 		if (zfs_smbshare_inited == 0) {
2636 			mutex_enter(&zfs_share_lock);
2637 			if (smbsrv_mod == NULL && ((smbsrv_mod =
2638 			    ddi_modopen("drv/smbsrv",
2639 			    KRTLD_MODE_FIRST, &error)) == NULL)) {
2640 				mutex_exit(&zfs_share_lock);
2641 				return (ENOSYS);
2642 			}
2643 			if (zsmbexport_fs == NULL && ((zsmbexport_fs =
2644 			    (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
2645 			    "lmshrd_share_upcall", &error)) == NULL)) {
2646 				mutex_exit(&zfs_share_lock);
2647 				return (ENOSYS);
2648 			}
2649 			error = zfs_init_sharefs();
2650 			if (error) {
2651 				mutex_exit(&zfs_share_lock);
2652 				return (ENOSYS);
2653 			}
2654 			zfs_smbshare_inited = 1;
2655 			mutex_exit(&zfs_share_lock);
2656 		}
2657 		break;
2658 	default:
2659 		return (EINVAL);
2660 	}
2661 
2662 	switch (zc->zc_share.z_sharetype) {
2663 	case ZFS_SHARE_NFS:
2664 	case ZFS_UNSHARE_NFS:
2665 		if (error =
2666 		    znfsexport_fs((void *)
2667 		    (uintptr_t)zc->zc_share.z_exportdata))
2668 			return (error);
2669 		break;
2670 	case ZFS_SHARE_SMB:
2671 	case ZFS_UNSHARE_SMB:
2672 		if (error = zsmbexport_fs((void *)
2673 		    (uintptr_t)zc->zc_share.z_exportdata,
2674 		    zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
2675 		    B_TRUE : B_FALSE)) {
2676 			return (error);
2677 		}
2678 		break;
2679 	}
2680 
2681 	opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
2682 	    zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
2683 	    SHAREFS_ADD : SHAREFS_REMOVE;
2684 
2685 	/*
2686 	 * Add or remove share from sharetab
2687 	 */
2688 	error = zshare_fs(opcode,
2689 	    (void *)(uintptr_t)zc->zc_share.z_sharedata,
2690 	    zc->zc_share.z_sharemax);
2691 
2692 	return (error);
2693 
2694 }
2695 
2696 /*
2697  * pool create, destroy, and export don't log the history as part of
2698  * zfsdev_ioctl, but rather zfs_ioc_pool_create, and zfs_ioc_pool_export
2699  * do the logging of those commands.
2700  */
2701 static zfs_ioc_vec_t zfs_ioc_vec[] = {
2702 	{ zfs_ioc_pool_create, zfs_secpolicy_config, POOL_NAME, B_FALSE },
2703 	{ zfs_ioc_pool_destroy,	zfs_secpolicy_config, POOL_NAME, B_FALSE },
2704 	{ zfs_ioc_pool_import, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2705 	{ zfs_ioc_pool_export, zfs_secpolicy_config, POOL_NAME, B_FALSE },
2706 	{ zfs_ioc_pool_configs,	zfs_secpolicy_none, NO_NAME, B_FALSE },
2707 	{ zfs_ioc_pool_stats, zfs_secpolicy_read, POOL_NAME, B_FALSE },
2708 	{ zfs_ioc_pool_tryimport, zfs_secpolicy_config, NO_NAME, B_FALSE },
2709 	{ zfs_ioc_pool_scrub, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2710 	{ zfs_ioc_pool_freeze, zfs_secpolicy_config, NO_NAME, B_FALSE },
2711 	{ zfs_ioc_pool_upgrade,	zfs_secpolicy_config, POOL_NAME, B_TRUE },
2712 	{ zfs_ioc_pool_get_history, zfs_secpolicy_config, POOL_NAME, B_FALSE },
2713 	{ zfs_ioc_vdev_add, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2714 	{ zfs_ioc_vdev_remove, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2715 	{ zfs_ioc_vdev_set_state, zfs_secpolicy_config,	POOL_NAME, B_TRUE },
2716 	{ zfs_ioc_vdev_attach, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2717 	{ zfs_ioc_vdev_detach, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2718 	{ zfs_ioc_vdev_setpath,	zfs_secpolicy_config, POOL_NAME, B_FALSE },
2719 	{ zfs_ioc_objset_stats,	zfs_secpolicy_read, DATASET_NAME, B_FALSE },
2720 	{ zfs_ioc_objset_zplprops, zfs_secpolicy_read, DATASET_NAME, B_FALSE },
2721 	{ zfs_ioc_dataset_list_next, zfs_secpolicy_read,
2722 	    DATASET_NAME, B_FALSE },
2723 	{ zfs_ioc_snapshot_list_next, zfs_secpolicy_read,
2724 	    DATASET_NAME, B_FALSE },
2725 	{ zfs_ioc_set_prop, zfs_secpolicy_none, DATASET_NAME, B_TRUE },
2726 	{ zfs_ioc_create_minor,	zfs_secpolicy_minor, DATASET_NAME, B_FALSE },
2727 	{ zfs_ioc_remove_minor,	zfs_secpolicy_minor, DATASET_NAME, B_FALSE },
2728 	{ zfs_ioc_create, zfs_secpolicy_create, DATASET_NAME, B_TRUE },
2729 	{ zfs_ioc_destroy, zfs_secpolicy_destroy, DATASET_NAME, B_TRUE },
2730 	{ zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, B_TRUE },
2731 	{ zfs_ioc_rename, zfs_secpolicy_rename,	DATASET_NAME, B_TRUE },
2732 	{ zfs_ioc_recv, zfs_secpolicy_receive, DATASET_NAME, B_TRUE },
2733 	{ zfs_ioc_send, zfs_secpolicy_send, DATASET_NAME, B_TRUE },
2734 	{ zfs_ioc_inject_fault,	zfs_secpolicy_inject, NO_NAME, B_FALSE },
2735 	{ zfs_ioc_clear_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE },
2736 	{ zfs_ioc_inject_list_next, zfs_secpolicy_inject, NO_NAME, B_FALSE },
2737 	{ zfs_ioc_error_log, zfs_secpolicy_inject, POOL_NAME, B_FALSE },
2738 	{ zfs_ioc_clear, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2739 	{ zfs_ioc_promote, zfs_secpolicy_promote, DATASET_NAME, B_TRUE },
2740 	{ zfs_ioc_destroy_snaps, zfs_secpolicy_destroy,	DATASET_NAME, B_TRUE },
2741 	{ zfs_ioc_snapshot, zfs_secpolicy_snapshot, DATASET_NAME, B_TRUE },
2742 	{ zfs_ioc_dsobj_to_dsname, zfs_secpolicy_config, POOL_NAME, B_FALSE },
2743 	{ zfs_ioc_obj_to_path, zfs_secpolicy_config, NO_NAME, B_FALSE },
2744 	{ zfs_ioc_pool_set_props, zfs_secpolicy_config,	POOL_NAME, B_TRUE },
2745 	{ zfs_ioc_pool_get_props, zfs_secpolicy_read, POOL_NAME, B_FALSE },
2746 	{ zfs_ioc_set_fsacl, zfs_secpolicy_fsacl, DATASET_NAME, B_TRUE },
2747 	{ zfs_ioc_get_fsacl, zfs_secpolicy_read, DATASET_NAME, B_FALSE },
2748 	{ zfs_ioc_iscsi_perm_check, zfs_secpolicy_iscsi,
2749 	    DATASET_NAME, B_FALSE },
2750 	{ zfs_ioc_share, zfs_secpolicy_share, DATASET_NAME, B_FALSE },
2751 	{ zfs_ioc_inherit_prop, zfs_secpolicy_inherit, DATASET_NAME, B_TRUE },
2752 };
2753 
2754 static int
2755 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
2756 {
2757 	zfs_cmd_t *zc;
2758 	uint_t vec;
2759 	int error, rc;
2760 
2761 	if (getminor(dev) != 0)
2762 		return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
2763 
2764 	vec = cmd - ZFS_IOC;
2765 	ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
2766 
2767 	if (vec >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
2768 		return (EINVAL);
2769 
2770 	zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2771 
2772 	error = xcopyin((void *)arg, zc, sizeof (zfs_cmd_t));
2773 
2774 	if (error == 0)
2775 		error = zfs_ioc_vec[vec].zvec_secpolicy(zc, cr);
2776 
2777 	/*
2778 	 * Ensure that all pool/dataset names are valid before we pass down to
2779 	 * the lower layers.
2780 	 */
2781 	if (error == 0) {
2782 		zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
2783 		switch (zfs_ioc_vec[vec].zvec_namecheck) {
2784 		case POOL_NAME:
2785 			if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
2786 				error = EINVAL;
2787 			break;
2788 
2789 		case DATASET_NAME:
2790 			if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
2791 				error = EINVAL;
2792 			break;
2793 
2794 		case NO_NAME:
2795 			break;
2796 		}
2797 	}
2798 
2799 	if (error == 0)
2800 		error = zfs_ioc_vec[vec].zvec_func(zc);
2801 
2802 	rc = xcopyout(zc, (void *)arg, sizeof (zfs_cmd_t));
2803 	if (error == 0) {
2804 		error = rc;
2805 		if (zfs_ioc_vec[vec].zvec_his_log == B_TRUE)
2806 			zfs_log_history(zc);
2807 	}
2808 
2809 	kmem_free(zc, sizeof (zfs_cmd_t));
2810 	return (error);
2811 }
2812 
2813 static int
2814 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
2815 {
2816 	if (cmd != DDI_ATTACH)
2817 		return (DDI_FAILURE);
2818 
2819 	if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
2820 	    DDI_PSEUDO, 0) == DDI_FAILURE)
2821 		return (DDI_FAILURE);
2822 
2823 	zfs_dip = dip;
2824 
2825 	ddi_report_dev(dip);
2826 
2827 	return (DDI_SUCCESS);
2828 }
2829 
2830 static int
2831 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
2832 {
2833 	if (spa_busy() || zfs_busy() || zvol_busy())
2834 		return (DDI_FAILURE);
2835 
2836 	if (cmd != DDI_DETACH)
2837 		return (DDI_FAILURE);
2838 
2839 	zfs_dip = NULL;
2840 
2841 	ddi_prop_remove_all(dip);
2842 	ddi_remove_minor_node(dip, NULL);
2843 
2844 	return (DDI_SUCCESS);
2845 }
2846 
2847 /*ARGSUSED*/
2848 static int
2849 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
2850 {
2851 	switch (infocmd) {
2852 	case DDI_INFO_DEVT2DEVINFO:
2853 		*result = zfs_dip;
2854 		return (DDI_SUCCESS);
2855 
2856 	case DDI_INFO_DEVT2INSTANCE:
2857 		*result = (void *)0;
2858 		return (DDI_SUCCESS);
2859 	}
2860 
2861 	return (DDI_FAILURE);
2862 }
2863 
2864 /*
2865  * OK, so this is a little weird.
2866  *
2867  * /dev/zfs is the control node, i.e. minor 0.
2868  * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
2869  *
2870  * /dev/zfs has basically nothing to do except serve up ioctls,
2871  * so most of the standard driver entry points are in zvol.c.
2872  */
2873 static struct cb_ops zfs_cb_ops = {
2874 	zvol_open,	/* open */
2875 	zvol_close,	/* close */
2876 	zvol_strategy,	/* strategy */
2877 	nodev,		/* print */
2878 	nodev,		/* dump */
2879 	zvol_read,	/* read */
2880 	zvol_write,	/* write */
2881 	zfsdev_ioctl,	/* ioctl */
2882 	nodev,		/* devmap */
2883 	nodev,		/* mmap */
2884 	nodev,		/* segmap */
2885 	nochpoll,	/* poll */
2886 	ddi_prop_op,	/* prop_op */
2887 	NULL,		/* streamtab */
2888 	D_NEW | D_MP | D_64BIT,		/* Driver compatibility flag */
2889 	CB_REV,		/* version */
2890 	nodev,		/* async read */
2891 	nodev,		/* async write */
2892 };
2893 
2894 static struct dev_ops zfs_dev_ops = {
2895 	DEVO_REV,	/* version */
2896 	0,		/* refcnt */
2897 	zfs_info,	/* info */
2898 	nulldev,	/* identify */
2899 	nulldev,	/* probe */
2900 	zfs_attach,	/* attach */
2901 	zfs_detach,	/* detach */
2902 	nodev,		/* reset */
2903 	&zfs_cb_ops,	/* driver operations */
2904 	NULL		/* no bus operations */
2905 };
2906 
2907 static struct modldrv zfs_modldrv = {
2908 	&mod_driverops, "ZFS storage pool version " SPA_VERSION_STRING,
2909 	    &zfs_dev_ops
2910 };
2911 
2912 static struct modlinkage modlinkage = {
2913 	MODREV_1,
2914 	(void *)&zfs_modlfs,
2915 	(void *)&zfs_modldrv,
2916 	NULL
2917 };
2918 
2919 
2920 uint_t zfs_fsyncer_key;
2921 extern uint_t rrw_tsd_key;
2922 
2923 int
2924 _init(void)
2925 {
2926 	int error;
2927 
2928 	spa_init(FREAD | FWRITE);
2929 	zfs_init();
2930 	zvol_init();
2931 
2932 	if ((error = mod_install(&modlinkage)) != 0) {
2933 		zvol_fini();
2934 		zfs_fini();
2935 		spa_fini();
2936 		return (error);
2937 	}
2938 
2939 	tsd_create(&zfs_fsyncer_key, NULL);
2940 	tsd_create(&rrw_tsd_key, NULL);
2941 
2942 	error = ldi_ident_from_mod(&modlinkage, &zfs_li);
2943 	ASSERT(error == 0);
2944 	mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
2945 
2946 	return (0);
2947 }
2948 
2949 int
2950 _fini(void)
2951 {
2952 	int error;
2953 
2954 	if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
2955 		return (EBUSY);
2956 
2957 	if ((error = mod_remove(&modlinkage)) != 0)
2958 		return (error);
2959 
2960 	zvol_fini();
2961 	zfs_fini();
2962 	spa_fini();
2963 	if (zfs_nfsshare_inited)
2964 		(void) ddi_modclose(nfs_mod);
2965 	if (zfs_smbshare_inited)
2966 		(void) ddi_modclose(smbsrv_mod);
2967 	if (zfs_nfsshare_inited || zfs_smbshare_inited)
2968 		(void) ddi_modclose(sharefs_mod);
2969 
2970 	tsd_destroy(&zfs_fsyncer_key);
2971 	ldi_ident_release(zfs_li);
2972 	zfs_li = NULL;
2973 	mutex_destroy(&zfs_share_lock);
2974 
2975 	return (error);
2976 }
2977 
2978 int
2979 _info(struct modinfo *modinfop)
2980 {
2981 	return (mod_info(&modlinkage, modinfop));
2982 }
2983