xref: /illumos-gate/usr/src/uts/common/fs/zfs/zfs_ioctl.c (revision 088f389458728c464569a5506b58070254fa4f7d)
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 2008 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  * Check to see if the named dataset is currently defined as bootable
159  */
160 static boolean_t
161 zfs_is_bootfs(const char *name)
162 {
163 	spa_t *spa;
164 	boolean_t ret = B_FALSE;
165 
166 	if (spa_open(name, &spa, FTAG) == 0) {
167 		if (spa->spa_bootfs) {
168 			objset_t *os;
169 
170 			if (dmu_objset_open(name, DMU_OST_ZFS,
171 			    DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
172 				ret = (dmu_objset_id(os) == spa->spa_bootfs);
173 				dmu_objset_close(os);
174 			}
175 		}
176 		spa_close(spa, FTAG);
177 	}
178 	return (ret);
179 }
180 
181 /*
182  * zfs_check_version
183  *
184  *	Return non-zero if the spa version is less than requested version.
185  */
186 static int
187 zfs_check_version(const char *name, int version)
188 {
189 
190 	spa_t *spa;
191 
192 	if (spa_open(name, &spa, FTAG) == 0) {
193 		if (spa_version(spa) < version) {
194 			spa_close(spa, FTAG);
195 			return (1);
196 		}
197 		spa_close(spa, FTAG);
198 	}
199 	return (0);
200 }
201 
202 /*
203  * zpl_earlier_version
204  *
205  * Return TRUE if the ZPL version is less than requested version.
206  */
207 static boolean_t
208 zpl_earlier_version(const char *name, int version)
209 {
210 	objset_t *os;
211 	boolean_t rc = B_TRUE;
212 
213 	if (dmu_objset_open(name, DMU_OST_ANY,
214 	    DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
215 		uint64_t zplversion;
216 
217 		if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
218 			rc = zplversion < version;
219 		dmu_objset_close(os);
220 	}
221 	return (rc);
222 }
223 
224 static void
225 zfs_log_history(zfs_cmd_t *zc)
226 {
227 	spa_t *spa;
228 	char *buf;
229 
230 	if ((buf = history_str_get(zc)) == NULL)
231 		return;
232 
233 	if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
234 		if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
235 			(void) spa_history_log(spa, buf, LOG_CMD_NORMAL);
236 		spa_close(spa, FTAG);
237 	}
238 	history_str_free(buf);
239 }
240 
241 /*
242  * Policy for top-level read operations (list pools).  Requires no privileges,
243  * and can be used in the local zone, as there is no associated dataset.
244  */
245 /* ARGSUSED */
246 static int
247 zfs_secpolicy_none(zfs_cmd_t *zc, cred_t *cr)
248 {
249 	return (0);
250 }
251 
252 /*
253  * Policy for dataset read operations (list children, get statistics).  Requires
254  * no privileges, but must be visible in the local zone.
255  */
256 /* ARGSUSED */
257 static int
258 zfs_secpolicy_read(zfs_cmd_t *zc, cred_t *cr)
259 {
260 	if (INGLOBALZONE(curproc) ||
261 	    zone_dataset_visible(zc->zc_name, NULL))
262 		return (0);
263 
264 	return (ENOENT);
265 }
266 
267 static int
268 zfs_dozonecheck(const char *dataset, cred_t *cr)
269 {
270 	uint64_t zoned;
271 	int writable = 1;
272 
273 	/*
274 	 * The dataset must be visible by this zone -- check this first
275 	 * so they don't see EPERM on something they shouldn't know about.
276 	 */
277 	if (!INGLOBALZONE(curproc) &&
278 	    !zone_dataset_visible(dataset, &writable))
279 		return (ENOENT);
280 
281 	if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
282 		return (ENOENT);
283 
284 	if (INGLOBALZONE(curproc)) {
285 		/*
286 		 * If the fs is zoned, only root can access it from the
287 		 * global zone.
288 		 */
289 		if (secpolicy_zfs(cr) && zoned)
290 			return (EPERM);
291 	} else {
292 		/*
293 		 * If we are in a local zone, the 'zoned' property must be set.
294 		 */
295 		if (!zoned)
296 			return (EPERM);
297 
298 		/* must be writable by this zone */
299 		if (!writable)
300 			return (EPERM);
301 	}
302 	return (0);
303 }
304 
305 int
306 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
307 {
308 	int error;
309 
310 	error = zfs_dozonecheck(name, cr);
311 	if (error == 0) {
312 		error = secpolicy_zfs(cr);
313 		if (error)
314 			error = dsl_deleg_access(name, perm, cr);
315 	}
316 	return (error);
317 }
318 
319 static int
320 zfs_secpolicy_setprop(const char *name, zfs_prop_t prop, cred_t *cr)
321 {
322 	/*
323 	 * Check permissions for special properties.
324 	 */
325 	switch (prop) {
326 	case ZFS_PROP_ZONED:
327 		/*
328 		 * Disallow setting of 'zoned' from within a local zone.
329 		 */
330 		if (!INGLOBALZONE(curproc))
331 			return (EPERM);
332 		break;
333 
334 	case ZFS_PROP_QUOTA:
335 		if (!INGLOBALZONE(curproc)) {
336 			uint64_t zoned;
337 			char setpoint[MAXNAMELEN];
338 			/*
339 			 * Unprivileged users are allowed to modify the
340 			 * quota on things *under* (ie. contained by)
341 			 * the thing they own.
342 			 */
343 			if (dsl_prop_get_integer(name, "zoned", &zoned,
344 			    setpoint))
345 				return (EPERM);
346 			if (!zoned || strlen(name) <= strlen(setpoint))
347 				return (EPERM);
348 		}
349 		break;
350 	}
351 
352 	return (zfs_secpolicy_write_perms(name, zfs_prop_to_name(prop), cr));
353 }
354 
355 int
356 zfs_secpolicy_fsacl(zfs_cmd_t *zc, cred_t *cr)
357 {
358 	int error;
359 
360 	error = zfs_dozonecheck(zc->zc_name, cr);
361 	if (error)
362 		return (error);
363 
364 	/*
365 	 * permission to set permissions will be evaluated later in
366 	 * dsl_deleg_can_allow()
367 	 */
368 	return (0);
369 }
370 
371 int
372 zfs_secpolicy_rollback(zfs_cmd_t *zc, cred_t *cr)
373 {
374 	int error;
375 	error = zfs_secpolicy_write_perms(zc->zc_name,
376 	    ZFS_DELEG_PERM_ROLLBACK, cr);
377 	if (error == 0)
378 		error = zfs_secpolicy_write_perms(zc->zc_name,
379 		    ZFS_DELEG_PERM_MOUNT, cr);
380 	return (error);
381 }
382 
383 int
384 zfs_secpolicy_send(zfs_cmd_t *zc, cred_t *cr)
385 {
386 	return (zfs_secpolicy_write_perms(zc->zc_name,
387 	    ZFS_DELEG_PERM_SEND, cr));
388 }
389 
390 int
391 zfs_secpolicy_share(zfs_cmd_t *zc, cred_t *cr)
392 {
393 	if (!INGLOBALZONE(curproc))
394 		return (EPERM);
395 
396 	if (secpolicy_nfs(cr) == 0) {
397 		return (0);
398 	} else {
399 		vnode_t *vp;
400 		int error;
401 
402 		if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
403 		    NO_FOLLOW, NULL, &vp)) != 0)
404 			return (error);
405 
406 		/* Now make sure mntpnt and dataset are ZFS */
407 
408 		if (vp->v_vfsp->vfs_fstype != zfsfstype ||
409 		    (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
410 		    zc->zc_name) != 0)) {
411 			VN_RELE(vp);
412 			return (EPERM);
413 		}
414 
415 		VN_RELE(vp);
416 		return (dsl_deleg_access(zc->zc_name,
417 		    ZFS_DELEG_PERM_SHARE, cr));
418 	}
419 }
420 
421 static int
422 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
423 {
424 	char *cp;
425 
426 	/*
427 	 * Remove the @bla or /bla from the end of the name to get the parent.
428 	 */
429 	(void) strncpy(parent, datasetname, parentsize);
430 	cp = strrchr(parent, '@');
431 	if (cp != NULL) {
432 		cp[0] = '\0';
433 	} else {
434 		cp = strrchr(parent, '/');
435 		if (cp == NULL)
436 			return (ENOENT);
437 		cp[0] = '\0';
438 	}
439 
440 	return (0);
441 }
442 
443 int
444 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
445 {
446 	int error;
447 
448 	if ((error = zfs_secpolicy_write_perms(name,
449 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
450 		return (error);
451 
452 	return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
453 }
454 
455 static int
456 zfs_secpolicy_destroy(zfs_cmd_t *zc, cred_t *cr)
457 {
458 	return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
459 }
460 
461 /*
462  * Must have sys_config privilege to check the iscsi permission
463  */
464 /* ARGSUSED */
465 static int
466 zfs_secpolicy_iscsi(zfs_cmd_t *zc, cred_t *cr)
467 {
468 	return (secpolicy_zfs(cr));
469 }
470 
471 int
472 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
473 {
474 	char 	parentname[MAXNAMELEN];
475 	int	error;
476 
477 	if ((error = zfs_secpolicy_write_perms(from,
478 	    ZFS_DELEG_PERM_RENAME, cr)) != 0)
479 		return (error);
480 
481 	if ((error = zfs_secpolicy_write_perms(from,
482 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
483 		return (error);
484 
485 	if ((error = zfs_get_parent(to, parentname,
486 	    sizeof (parentname))) != 0)
487 		return (error);
488 
489 	if ((error = zfs_secpolicy_write_perms(parentname,
490 	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
491 		return (error);
492 
493 	if ((error = zfs_secpolicy_write_perms(parentname,
494 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
495 		return (error);
496 
497 	return (error);
498 }
499 
500 static int
501 zfs_secpolicy_rename(zfs_cmd_t *zc, cred_t *cr)
502 {
503 	return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
504 }
505 
506 static int
507 zfs_secpolicy_promote(zfs_cmd_t *zc, cred_t *cr)
508 {
509 	char 	parentname[MAXNAMELEN];
510 	objset_t *clone;
511 	int error;
512 
513 	error = zfs_secpolicy_write_perms(zc->zc_name,
514 	    ZFS_DELEG_PERM_PROMOTE, cr);
515 	if (error)
516 		return (error);
517 
518 	error = dmu_objset_open(zc->zc_name, DMU_OST_ANY,
519 	    DS_MODE_USER | DS_MODE_READONLY, &clone);
520 
521 	if (error == 0) {
522 		dsl_dataset_t *pclone = NULL;
523 		dsl_dir_t *dd;
524 		dd = clone->os->os_dsl_dataset->ds_dir;
525 
526 		rw_enter(&dd->dd_pool->dp_config_rwlock, RW_READER);
527 		error = dsl_dataset_hold_obj(dd->dd_pool,
528 		    dd->dd_phys->dd_origin_obj, FTAG, &pclone);
529 		rw_exit(&dd->dd_pool->dp_config_rwlock);
530 		if (error) {
531 			dmu_objset_close(clone);
532 			return (error);
533 		}
534 
535 		error = zfs_secpolicy_write_perms(zc->zc_name,
536 		    ZFS_DELEG_PERM_MOUNT, cr);
537 
538 		dsl_dataset_name(pclone, parentname);
539 		dmu_objset_close(clone);
540 		dsl_dataset_rele(pclone, FTAG);
541 		if (error == 0)
542 			error = zfs_secpolicy_write_perms(parentname,
543 			    ZFS_DELEG_PERM_PROMOTE, cr);
544 	}
545 	return (error);
546 }
547 
548 static int
549 zfs_secpolicy_receive(zfs_cmd_t *zc, cred_t *cr)
550 {
551 	int error;
552 
553 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
554 	    ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
555 		return (error);
556 
557 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
558 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
559 		return (error);
560 
561 	return (zfs_secpolicy_write_perms(zc->zc_name,
562 	    ZFS_DELEG_PERM_CREATE, cr));
563 }
564 
565 int
566 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
567 {
568 	int error;
569 
570 	if ((error = zfs_secpolicy_write_perms(name,
571 	    ZFS_DELEG_PERM_SNAPSHOT, cr)) != 0)
572 		return (error);
573 
574 	error = zfs_secpolicy_write_perms(name,
575 	    ZFS_DELEG_PERM_MOUNT, cr);
576 
577 	return (error);
578 }
579 
580 static int
581 zfs_secpolicy_snapshot(zfs_cmd_t *zc, cred_t *cr)
582 {
583 
584 	return (zfs_secpolicy_snapshot_perms(zc->zc_name, cr));
585 }
586 
587 static int
588 zfs_secpolicy_create(zfs_cmd_t *zc, cred_t *cr)
589 {
590 	char 	parentname[MAXNAMELEN];
591 	int 	error;
592 
593 	if ((error = zfs_get_parent(zc->zc_name, parentname,
594 	    sizeof (parentname))) != 0)
595 		return (error);
596 
597 	if (zc->zc_value[0] != '\0') {
598 		if ((error = zfs_secpolicy_write_perms(zc->zc_value,
599 		    ZFS_DELEG_PERM_CLONE, cr)) != 0)
600 			return (error);
601 	}
602 
603 	if ((error = zfs_secpolicy_write_perms(parentname,
604 	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
605 		return (error);
606 
607 	error = zfs_secpolicy_write_perms(parentname,
608 	    ZFS_DELEG_PERM_MOUNT, cr);
609 
610 	return (error);
611 }
612 
613 static int
614 zfs_secpolicy_umount(zfs_cmd_t *zc, cred_t *cr)
615 {
616 	int error;
617 
618 	error = secpolicy_fs_unmount(cr, NULL);
619 	if (error) {
620 		error = dsl_deleg_access(zc->zc_name, ZFS_DELEG_PERM_MOUNT, cr);
621 	}
622 	return (error);
623 }
624 
625 /*
626  * Policy for pool operations - create/destroy pools, add vdevs, etc.  Requires
627  * SYS_CONFIG privilege, which is not available in a local zone.
628  */
629 /* ARGSUSED */
630 static int
631 zfs_secpolicy_config(zfs_cmd_t *zc, cred_t *cr)
632 {
633 	if (secpolicy_sys_config(cr, B_FALSE) != 0)
634 		return (EPERM);
635 
636 	return (0);
637 }
638 
639 /*
640  * Just like zfs_secpolicy_config, except that we will check for
641  * mount permission on the dataset for permission to create/remove
642  * the minor nodes.
643  */
644 static int
645 zfs_secpolicy_minor(zfs_cmd_t *zc, cred_t *cr)
646 {
647 	if (secpolicy_sys_config(cr, B_FALSE) != 0) {
648 		return (dsl_deleg_access(zc->zc_name,
649 		    ZFS_DELEG_PERM_MOUNT, cr));
650 	}
651 
652 	return (0);
653 }
654 
655 /*
656  * Policy for fault injection.  Requires all privileges.
657  */
658 /* ARGSUSED */
659 static int
660 zfs_secpolicy_inject(zfs_cmd_t *zc, cred_t *cr)
661 {
662 	return (secpolicy_zinject(cr));
663 }
664 
665 static int
666 zfs_secpolicy_inherit(zfs_cmd_t *zc, cred_t *cr)
667 {
668 	zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
669 
670 	if (prop == ZPROP_INVAL) {
671 		if (!zfs_prop_user(zc->zc_value))
672 			return (EINVAL);
673 		return (zfs_secpolicy_write_perms(zc->zc_name,
674 		    ZFS_DELEG_PERM_USERPROP, cr));
675 	} else {
676 		if (!zfs_prop_inheritable(prop))
677 			return (EINVAL);
678 		return (zfs_secpolicy_setprop(zc->zc_name, prop, cr));
679 	}
680 }
681 
682 /*
683  * Returns the nvlist as specified by the user in the zfs_cmd_t.
684  */
685 static int
686 get_nvlist(uint64_t nvl, uint64_t size, nvlist_t **nvp)
687 {
688 	char *packed;
689 	int error;
690 	nvlist_t *list = NULL;
691 
692 	/*
693 	 * Read in and unpack the user-supplied nvlist.
694 	 */
695 	if (size == 0)
696 		return (EINVAL);
697 
698 	packed = kmem_alloc(size, KM_SLEEP);
699 
700 	if ((error = xcopyin((void *)(uintptr_t)nvl, packed, size)) != 0) {
701 		kmem_free(packed, size);
702 		return (error);
703 	}
704 
705 	if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
706 		kmem_free(packed, size);
707 		return (error);
708 	}
709 
710 	kmem_free(packed, size);
711 
712 	*nvp = list;
713 	return (0);
714 }
715 
716 static int
717 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
718 {
719 	char *packed = NULL;
720 	size_t size;
721 	int error;
722 
723 	VERIFY(nvlist_size(nvl, &size, NV_ENCODE_NATIVE) == 0);
724 
725 	if (size > zc->zc_nvlist_dst_size) {
726 		error = ENOMEM;
727 	} else {
728 		packed = kmem_alloc(size, KM_SLEEP);
729 		VERIFY(nvlist_pack(nvl, &packed, &size, NV_ENCODE_NATIVE,
730 		    KM_SLEEP) == 0);
731 		error = xcopyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
732 		    size);
733 		kmem_free(packed, size);
734 	}
735 
736 	zc->zc_nvlist_dst_size = size;
737 	return (error);
738 }
739 
740 static int
741 zfs_ioc_pool_create(zfs_cmd_t *zc)
742 {
743 	int error;
744 	nvlist_t *config, *props = NULL;
745 	char *buf;
746 
747 	if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
748 	    &config))
749 		return (error);
750 
751 	if (zc->zc_nvlist_src_size != 0 && (error =
752 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, &props))) {
753 		nvlist_free(config);
754 		return (error);
755 	}
756 
757 	buf = history_str_get(zc);
758 
759 	error = spa_create(zc->zc_name, config, props, buf);
760 
761 	if (buf != NULL)
762 		history_str_free(buf);
763 
764 	nvlist_free(config);
765 
766 	if (props)
767 		nvlist_free(props);
768 
769 	return (error);
770 }
771 
772 static int
773 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
774 {
775 	int error;
776 	zfs_log_history(zc);
777 	error = spa_destroy(zc->zc_name);
778 	return (error);
779 }
780 
781 static int
782 zfs_ioc_pool_import(zfs_cmd_t *zc)
783 {
784 	int error;
785 	nvlist_t *config, *props = NULL;
786 	uint64_t guid;
787 
788 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
789 	    &config)) != 0)
790 		return (error);
791 
792 	if (zc->zc_nvlist_src_size != 0 && (error =
793 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, &props))) {
794 		nvlist_free(config);
795 		return (error);
796 	}
797 
798 	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
799 	    guid != zc->zc_guid)
800 		error = EINVAL;
801 	else if (zc->zc_cookie)
802 		error = spa_import_faulted(zc->zc_name, config,
803 		    props);
804 	else
805 		error = spa_import(zc->zc_name, config, props);
806 
807 	nvlist_free(config);
808 
809 	if (props)
810 		nvlist_free(props);
811 
812 	return (error);
813 }
814 
815 static int
816 zfs_ioc_pool_export(zfs_cmd_t *zc)
817 {
818 	int error;
819 	zfs_log_history(zc);
820 	error = spa_export(zc->zc_name, NULL);
821 	return (error);
822 }
823 
824 static int
825 zfs_ioc_pool_configs(zfs_cmd_t *zc)
826 {
827 	nvlist_t *configs;
828 	int error;
829 
830 	if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
831 		return (EEXIST);
832 
833 	error = put_nvlist(zc, configs);
834 
835 	nvlist_free(configs);
836 
837 	return (error);
838 }
839 
840 static int
841 zfs_ioc_pool_stats(zfs_cmd_t *zc)
842 {
843 	nvlist_t *config;
844 	int error;
845 	int ret = 0;
846 
847 	error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
848 	    sizeof (zc->zc_value));
849 
850 	if (config != NULL) {
851 		ret = put_nvlist(zc, config);
852 		nvlist_free(config);
853 
854 		/*
855 		 * The config may be present even if 'error' is non-zero.
856 		 * In this case we return success, and preserve the real errno
857 		 * in 'zc_cookie'.
858 		 */
859 		zc->zc_cookie = error;
860 	} else {
861 		ret = error;
862 	}
863 
864 	return (ret);
865 }
866 
867 /*
868  * Try to import the given pool, returning pool stats as appropriate so that
869  * user land knows which devices are available and overall pool health.
870  */
871 static int
872 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
873 {
874 	nvlist_t *tryconfig, *config;
875 	int error;
876 
877 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
878 	    &tryconfig)) != 0)
879 		return (error);
880 
881 	config = spa_tryimport(tryconfig);
882 
883 	nvlist_free(tryconfig);
884 
885 	if (config == NULL)
886 		return (EINVAL);
887 
888 	error = put_nvlist(zc, config);
889 	nvlist_free(config);
890 
891 	return (error);
892 }
893 
894 static int
895 zfs_ioc_pool_scrub(zfs_cmd_t *zc)
896 {
897 	spa_t *spa;
898 	int error;
899 
900 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
901 		return (error);
902 
903 	error = spa_scrub(spa, zc->zc_cookie);
904 
905 	spa_close(spa, FTAG);
906 
907 	return (error);
908 }
909 
910 static int
911 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
912 {
913 	spa_t *spa;
914 	int error;
915 
916 	error = spa_open(zc->zc_name, &spa, FTAG);
917 	if (error == 0) {
918 		spa_freeze(spa);
919 		spa_close(spa, FTAG);
920 	}
921 	return (error);
922 }
923 
924 static int
925 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
926 {
927 	spa_t *spa;
928 	int error;
929 
930 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
931 		return (error);
932 
933 	if (zc->zc_cookie < spa_version(spa) || zc->zc_cookie > SPA_VERSION) {
934 		spa_close(spa, FTAG);
935 		return (EINVAL);
936 	}
937 
938 	spa_upgrade(spa, zc->zc_cookie);
939 	spa_close(spa, FTAG);
940 
941 	return (error);
942 }
943 
944 static int
945 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
946 {
947 	spa_t *spa;
948 	char *hist_buf;
949 	uint64_t size;
950 	int error;
951 
952 	if ((size = zc->zc_history_len) == 0)
953 		return (EINVAL);
954 
955 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
956 		return (error);
957 
958 	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
959 		spa_close(spa, FTAG);
960 		return (ENOTSUP);
961 	}
962 
963 	hist_buf = kmem_alloc(size, KM_SLEEP);
964 	if ((error = spa_history_get(spa, &zc->zc_history_offset,
965 	    &zc->zc_history_len, hist_buf)) == 0) {
966 		error = xcopyout(hist_buf,
967 		    (char *)(uintptr_t)zc->zc_history,
968 		    zc->zc_history_len);
969 	}
970 
971 	spa_close(spa, FTAG);
972 	kmem_free(hist_buf, size);
973 	return (error);
974 }
975 
976 static int
977 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
978 {
979 	int error;
980 
981 	if (error = dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value))
982 		return (error);
983 
984 	return (0);
985 }
986 
987 static int
988 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
989 {
990 	objset_t *osp;
991 	int error;
992 
993 	if ((error = dmu_objset_open(zc->zc_name, DMU_OST_ZFS,
994 	    DS_MODE_USER | DS_MODE_READONLY, &osp)) != 0)
995 		return (error);
996 	error = zfs_obj_to_path(osp, zc->zc_obj, zc->zc_value,
997 	    sizeof (zc->zc_value));
998 	dmu_objset_close(osp);
999 
1000 	return (error);
1001 }
1002 
1003 static int
1004 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1005 {
1006 	spa_t *spa;
1007 	int error;
1008 	nvlist_t *config, **l2cache, **spares;
1009 	uint_t nl2cache = 0, nspares = 0;
1010 
1011 	error = spa_open(zc->zc_name, &spa, FTAG);
1012 	if (error != 0)
1013 		return (error);
1014 
1015 	error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1016 	    &config);
1017 	(void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1018 	    &l2cache, &nl2cache);
1019 
1020 	(void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1021 	    &spares, &nspares);
1022 
1023 	/*
1024 	 * A root pool with concatenated devices is not supported.
1025 	 * Thus, can not add a device to a root pool.
1026 	 *
1027 	 * Intent log device can not be added to a rootpool because
1028 	 * during mountroot, zil is replayed, a seperated log device
1029 	 * can not be accessed during the mountroot time.
1030 	 *
1031 	 * l2cache and spare devices are ok to be added to a rootpool.
1032 	 */
1033 	if (spa->spa_bootfs != 0 && nl2cache == 0 && nspares == 0) {
1034 		spa_close(spa, FTAG);
1035 		return (EDOM);
1036 	}
1037 
1038 	if (error == 0) {
1039 		error = spa_vdev_add(spa, config);
1040 		nvlist_free(config);
1041 	}
1042 	spa_close(spa, FTAG);
1043 	return (error);
1044 }
1045 
1046 static int
1047 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1048 {
1049 	spa_t *spa;
1050 	int error;
1051 
1052 	error = spa_open(zc->zc_name, &spa, FTAG);
1053 	if (error != 0)
1054 		return (error);
1055 	error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1056 	spa_close(spa, FTAG);
1057 	return (error);
1058 }
1059 
1060 static int
1061 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1062 {
1063 	spa_t *spa;
1064 	int error;
1065 	vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1066 
1067 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1068 		return (error);
1069 	switch (zc->zc_cookie) {
1070 	case VDEV_STATE_ONLINE:
1071 		error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1072 		break;
1073 
1074 	case VDEV_STATE_OFFLINE:
1075 		error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1076 		break;
1077 
1078 	case VDEV_STATE_FAULTED:
1079 		error = vdev_fault(spa, zc->zc_guid);
1080 		break;
1081 
1082 	case VDEV_STATE_DEGRADED:
1083 		error = vdev_degrade(spa, zc->zc_guid);
1084 		break;
1085 
1086 	default:
1087 		error = EINVAL;
1088 	}
1089 	zc->zc_cookie = newstate;
1090 	spa_close(spa, FTAG);
1091 	return (error);
1092 }
1093 
1094 static int
1095 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1096 {
1097 	spa_t *spa;
1098 	int replacing = zc->zc_cookie;
1099 	nvlist_t *config;
1100 	int error;
1101 
1102 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1103 		return (error);
1104 
1105 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1106 	    &config)) == 0) {
1107 		error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1108 		nvlist_free(config);
1109 	}
1110 
1111 	spa_close(spa, FTAG);
1112 	return (error);
1113 }
1114 
1115 static int
1116 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1117 {
1118 	spa_t *spa;
1119 	int error;
1120 
1121 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1122 		return (error);
1123 
1124 	error = spa_vdev_detach(spa, zc->zc_guid, B_FALSE);
1125 
1126 	spa_close(spa, FTAG);
1127 	return (error);
1128 }
1129 
1130 static int
1131 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
1132 {
1133 	spa_t *spa;
1134 	char *path = zc->zc_value;
1135 	uint64_t guid = zc->zc_guid;
1136 	int error;
1137 
1138 	error = spa_open(zc->zc_name, &spa, FTAG);
1139 	if (error != 0)
1140 		return (error);
1141 
1142 	error = spa_vdev_setpath(spa, guid, path);
1143 	spa_close(spa, FTAG);
1144 	return (error);
1145 }
1146 
1147 /*
1148  * inputs:
1149  * zc_name		name of filesystem
1150  * zc_nvlist_dst_size	size of buffer for property nvlist
1151  *
1152  * outputs:
1153  * zc_objset_stats	stats
1154  * zc_nvlist_dst	property nvlist
1155  * zc_nvlist_dst_size	size of property nvlist
1156  */
1157 static int
1158 zfs_ioc_objset_stats(zfs_cmd_t *zc)
1159 {
1160 	objset_t *os = NULL;
1161 	int error;
1162 	nvlist_t *nv;
1163 
1164 	if (error = dmu_objset_open(zc->zc_name,
1165 	    DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os))
1166 		return (error);
1167 
1168 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1169 
1170 	if (zc->zc_nvlist_dst != 0 &&
1171 	    (error = dsl_prop_get_all(os, &nv, FALSE)) == 0) {
1172 		dmu_objset_stats(os, nv);
1173 		/*
1174 		 * NB: zvol_get_stats() will read the objset contents,
1175 		 * which we aren't supposed to do with a
1176 		 * DS_MODE_USER hold, because it could be
1177 		 * inconsistent.  So this is a bit of a workaround...
1178 		 */
1179 		if (!zc->zc_objset_stats.dds_inconsistent) {
1180 			if (dmu_objset_type(os) == DMU_OST_ZVOL)
1181 				VERIFY(zvol_get_stats(os, nv) == 0);
1182 		}
1183 		error = put_nvlist(zc, nv);
1184 		nvlist_free(nv);
1185 	}
1186 
1187 	dmu_objset_close(os);
1188 	return (error);
1189 }
1190 
1191 static int
1192 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
1193 {
1194 	uint64_t value;
1195 	int error;
1196 
1197 	/*
1198 	 * zfs_get_zplprop() will either find a value or give us
1199 	 * the default value (if there is one).
1200 	 */
1201 	if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
1202 		return (error);
1203 	VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
1204 	return (0);
1205 }
1206 
1207 /*
1208  * inputs:
1209  * zc_name		name of filesystem
1210  * zc_nvlist_dst_size	size of buffer for zpl property nvlist
1211  *
1212  * outputs:
1213  * zc_nvlist_dst	zpl property nvlist
1214  * zc_nvlist_dst_size	size of zpl property nvlist
1215  */
1216 static int
1217 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
1218 {
1219 	objset_t *os;
1220 	int err;
1221 
1222 	if (err = dmu_objset_open(zc->zc_name,
1223 	    DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os))
1224 		return (err);
1225 
1226 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1227 
1228 	/*
1229 	 * NB: nvl_add_zplprop() will read the objset contents,
1230 	 * which we aren't supposed to do with a DS_MODE_USER
1231 	 * hold, because it could be inconsistent.
1232 	 */
1233 	if (zc->zc_nvlist_dst != NULL &&
1234 	    !zc->zc_objset_stats.dds_inconsistent &&
1235 	    dmu_objset_type(os) == DMU_OST_ZFS) {
1236 		nvlist_t *nv;
1237 
1238 		VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1239 		if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
1240 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
1241 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
1242 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
1243 			err = put_nvlist(zc, nv);
1244 		nvlist_free(nv);
1245 	} else {
1246 		err = ENOENT;
1247 	}
1248 	dmu_objset_close(os);
1249 	return (err);
1250 }
1251 
1252 /*
1253  * inputs:
1254  * zc_name		name of filesystem
1255  * zc_cookie		zap cursor
1256  * zc_nvlist_dst_size	size of buffer for property nvlist
1257  *
1258  * outputs:
1259  * zc_name		name of next filesystem
1260  * zc_objset_stats	stats
1261  * zc_nvlist_dst	property nvlist
1262  * zc_nvlist_dst_size	size of property nvlist
1263  */
1264 static int
1265 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
1266 {
1267 	objset_t *os;
1268 	int error;
1269 	char *p;
1270 
1271 	if (error = dmu_objset_open(zc->zc_name,
1272 	    DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os)) {
1273 		if (error == ENOENT)
1274 			error = ESRCH;
1275 		return (error);
1276 	}
1277 
1278 	p = strrchr(zc->zc_name, '/');
1279 	if (p == NULL || p[1] != '\0')
1280 		(void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
1281 	p = zc->zc_name + strlen(zc->zc_name);
1282 
1283 	do {
1284 		error = dmu_dir_list_next(os,
1285 		    sizeof (zc->zc_name) - (p - zc->zc_name), p,
1286 		    NULL, &zc->zc_cookie);
1287 		if (error == ENOENT)
1288 			error = ESRCH;
1289 	} while (error == 0 && !INGLOBALZONE(curproc) &&
1290 	    !zone_dataset_visible(zc->zc_name, NULL));
1291 	dmu_objset_close(os);
1292 
1293 	/*
1294 	 * If it's a hidden dataset (ie. with a '$' in its name), don't
1295 	 * try to get stats for it.  Userland will skip over it.
1296 	 */
1297 	if (error == 0 && strchr(zc->zc_name, '$') == NULL)
1298 		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
1299 
1300 	return (error);
1301 }
1302 
1303 /*
1304  * inputs:
1305  * zc_name		name of filesystem
1306  * zc_cookie		zap cursor
1307  * zc_nvlist_dst_size	size of buffer for property nvlist
1308  *
1309  * outputs:
1310  * zc_name		name of next snapshot
1311  * zc_objset_stats	stats
1312  * zc_nvlist_dst	property nvlist
1313  * zc_nvlist_dst_size	size of property nvlist
1314  */
1315 static int
1316 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
1317 {
1318 	objset_t *os;
1319 	int error;
1320 
1321 	error = dmu_objset_open(zc->zc_name,
1322 	    DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os);
1323 	if (error)
1324 		return (error == ENOENT ? ESRCH : error);
1325 
1326 	/*
1327 	 * A dataset name of maximum length cannot have any snapshots,
1328 	 * so exit immediately.
1329 	 */
1330 	if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
1331 		dmu_objset_close(os);
1332 		return (ESRCH);
1333 	}
1334 
1335 	error = dmu_snapshot_list_next(os,
1336 	    sizeof (zc->zc_name) - strlen(zc->zc_name),
1337 	    zc->zc_name + strlen(zc->zc_name), NULL, &zc->zc_cookie, NULL);
1338 	dmu_objset_close(os);
1339 	if (error == 0)
1340 		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
1341 	else if (error == ENOENT)
1342 		error = ESRCH;
1343 
1344 	/* if we failed, undo the @ that we tacked on to zc_name */
1345 	if (error)
1346 		*strchr(zc->zc_name, '@') = '\0';
1347 	return (error);
1348 }
1349 
1350 int
1351 zfs_set_prop_nvlist(const char *name, nvlist_t *nvl)
1352 {
1353 	nvpair_t *elem;
1354 	int error;
1355 	uint64_t intval;
1356 	char *strval;
1357 
1358 	/*
1359 	 * First validate permission to set all of the properties
1360 	 */
1361 	elem = NULL;
1362 	while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) {
1363 		const char *propname = nvpair_name(elem);
1364 		zfs_prop_t prop = zfs_name_to_prop(propname);
1365 
1366 		if (prop == ZPROP_INVAL) {
1367 			/*
1368 			 * If this is a user-defined property, it must be a
1369 			 * string, and there is no further validation to do.
1370 			 */
1371 			if (!zfs_prop_user(propname) ||
1372 			    nvpair_type(elem) != DATA_TYPE_STRING)
1373 				return (EINVAL);
1374 
1375 			if (error = zfs_secpolicy_write_perms(name,
1376 			    ZFS_DELEG_PERM_USERPROP, CRED()))
1377 				return (error);
1378 			continue;
1379 		}
1380 
1381 		if ((error = zfs_secpolicy_setprop(name, prop, CRED())) != 0)
1382 			return (error);
1383 
1384 		/*
1385 		 * Check that this value is valid for this pool version
1386 		 */
1387 		switch (prop) {
1388 		case ZFS_PROP_COMPRESSION:
1389 			/*
1390 			 * If the user specified gzip compression, make sure
1391 			 * the SPA supports it. We ignore any errors here since
1392 			 * we'll catch them later.
1393 			 */
1394 			if (nvpair_type(elem) == DATA_TYPE_UINT64 &&
1395 			    nvpair_value_uint64(elem, &intval) == 0) {
1396 				if (intval >= ZIO_COMPRESS_GZIP_1 &&
1397 				    intval <= ZIO_COMPRESS_GZIP_9 &&
1398 				    zfs_check_version(name,
1399 				    SPA_VERSION_GZIP_COMPRESSION))
1400 					return (ENOTSUP);
1401 
1402 				/*
1403 				 * If this is a bootable dataset then
1404 				 * verify that the compression algorithm
1405 				 * is supported for booting. We must return
1406 				 * something other than ENOTSUP since it
1407 				 * implies a downrev pool version.
1408 				 */
1409 				if (zfs_is_bootfs(name) &&
1410 				    !BOOTFS_COMPRESS_VALID(intval))
1411 					return (ERANGE);
1412 			}
1413 			break;
1414 
1415 		case ZFS_PROP_COPIES:
1416 			if (zfs_check_version(name, SPA_VERSION_DITTO_BLOCKS))
1417 				return (ENOTSUP);
1418 			break;
1419 
1420 		case ZFS_PROP_SHARESMB:
1421 			if (zpl_earlier_version(name, ZPL_VERSION_FUID))
1422 				return (ENOTSUP);
1423 			break;
1424 		}
1425 		if ((error = zfs_secpolicy_setprop(name, prop, CRED())) != 0)
1426 			return (error);
1427 	}
1428 
1429 	elem = NULL;
1430 	while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) {
1431 		const char *propname = nvpair_name(elem);
1432 		zfs_prop_t prop = zfs_name_to_prop(propname);
1433 
1434 		if (prop == ZPROP_INVAL) {
1435 			VERIFY(nvpair_value_string(elem, &strval) == 0);
1436 			error = dsl_prop_set(name, propname, 1,
1437 			    strlen(strval) + 1, strval);
1438 			if (error == 0)
1439 				continue;
1440 			else
1441 				return (error);
1442 		}
1443 
1444 		switch (prop) {
1445 		case ZFS_PROP_QUOTA:
1446 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1447 			    (error = dsl_dir_set_quota(name, intval)) != 0)
1448 				return (error);
1449 			break;
1450 
1451 		case ZFS_PROP_REFQUOTA:
1452 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1453 			    (error = dsl_dataset_set_quota(name, intval)) != 0)
1454 				return (error);
1455 			break;
1456 
1457 		case ZFS_PROP_RESERVATION:
1458 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1459 			    (error = dsl_dir_set_reservation(name,
1460 			    intval)) != 0)
1461 				return (error);
1462 			break;
1463 
1464 		case ZFS_PROP_REFRESERVATION:
1465 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1466 			    (error = dsl_dataset_set_reservation(name,
1467 			    intval)) != 0)
1468 				return (error);
1469 			break;
1470 
1471 		case ZFS_PROP_VOLSIZE:
1472 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1473 			    (error = zvol_set_volsize(name,
1474 			    ddi_driver_major(zfs_dip), intval)) != 0)
1475 				return (error);
1476 			break;
1477 
1478 		case ZFS_PROP_VOLBLOCKSIZE:
1479 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1480 			    (error = zvol_set_volblocksize(name, intval)) != 0)
1481 				return (error);
1482 			break;
1483 
1484 		case ZFS_PROP_VERSION:
1485 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1486 			    (error = zfs_set_version(name, intval)) != 0)
1487 				return (error);
1488 			break;
1489 
1490 		default:
1491 			if (nvpair_type(elem) == DATA_TYPE_STRING) {
1492 				if (zfs_prop_get_type(prop) !=
1493 				    PROP_TYPE_STRING)
1494 					return (EINVAL);
1495 				VERIFY(nvpair_value_string(elem, &strval) == 0);
1496 				if ((error = dsl_prop_set(name,
1497 				    nvpair_name(elem), 1, strlen(strval) + 1,
1498 				    strval)) != 0)
1499 					return (error);
1500 			} else if (nvpair_type(elem) == DATA_TYPE_UINT64) {
1501 				const char *unused;
1502 
1503 				VERIFY(nvpair_value_uint64(elem, &intval) == 0);
1504 
1505 				switch (zfs_prop_get_type(prop)) {
1506 				case PROP_TYPE_NUMBER:
1507 					break;
1508 				case PROP_TYPE_STRING:
1509 					return (EINVAL);
1510 				case PROP_TYPE_INDEX:
1511 					if (zfs_prop_index_to_string(prop,
1512 					    intval, &unused) != 0)
1513 						return (EINVAL);
1514 					break;
1515 				default:
1516 					cmn_err(CE_PANIC,
1517 					    "unknown property type");
1518 					break;
1519 				}
1520 
1521 				if ((error = dsl_prop_set(name, propname,
1522 				    8, 1, &intval)) != 0)
1523 					return (error);
1524 			} else {
1525 				return (EINVAL);
1526 			}
1527 			break;
1528 		}
1529 	}
1530 
1531 	return (0);
1532 }
1533 
1534 /*
1535  * inputs:
1536  * zc_name		name of filesystem
1537  * zc_value		name of property to inherit
1538  * zc_nvlist_src{_size}	nvlist of properties to apply
1539  *
1540  * outputs:		none
1541  */
1542 static int
1543 zfs_ioc_set_prop(zfs_cmd_t *zc)
1544 {
1545 	nvlist_t *nvl;
1546 	int error;
1547 
1548 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1549 	    &nvl)) != 0)
1550 		return (error);
1551 
1552 	error = zfs_set_prop_nvlist(zc->zc_name, nvl);
1553 
1554 	nvlist_free(nvl);
1555 	return (error);
1556 }
1557 
1558 /*
1559  * inputs:
1560  * zc_name		name of filesystem
1561  * zc_value		name of property to inherit
1562  *
1563  * outputs:		none
1564  */
1565 static int
1566 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
1567 {
1568 	/* the property name has been validated by zfs_secpolicy_inherit() */
1569 	return (dsl_prop_set(zc->zc_name, zc->zc_value, 0, 0, NULL));
1570 }
1571 
1572 static int
1573 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
1574 {
1575 	nvlist_t *props;
1576 	spa_t *spa;
1577 	int error;
1578 
1579 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1580 	    &props)))
1581 		return (error);
1582 
1583 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
1584 		nvlist_free(props);
1585 		return (error);
1586 	}
1587 
1588 	error = spa_prop_set(spa, props);
1589 
1590 	nvlist_free(props);
1591 	spa_close(spa, FTAG);
1592 
1593 	return (error);
1594 }
1595 
1596 static int
1597 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
1598 {
1599 	spa_t *spa;
1600 	int error;
1601 	nvlist_t *nvp = NULL;
1602 
1603 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1604 		return (error);
1605 
1606 	error = spa_prop_get(spa, &nvp);
1607 
1608 	if (error == 0 && zc->zc_nvlist_dst != NULL)
1609 		error = put_nvlist(zc, nvp);
1610 	else
1611 		error = EFAULT;
1612 
1613 	spa_close(spa, FTAG);
1614 
1615 	if (nvp)
1616 		nvlist_free(nvp);
1617 	return (error);
1618 }
1619 
1620 static int
1621 zfs_ioc_iscsi_perm_check(zfs_cmd_t *zc)
1622 {
1623 	nvlist_t *nvp;
1624 	int error;
1625 	uint32_t uid;
1626 	uint32_t gid;
1627 	uint32_t *groups;
1628 	uint_t group_cnt;
1629 	cred_t	*usercred;
1630 
1631 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1632 	    &nvp)) != 0) {
1633 		return (error);
1634 	}
1635 
1636 	if ((error = nvlist_lookup_uint32(nvp,
1637 	    ZFS_DELEG_PERM_UID, &uid)) != 0) {
1638 		nvlist_free(nvp);
1639 		return (EPERM);
1640 	}
1641 
1642 	if ((error = nvlist_lookup_uint32(nvp,
1643 	    ZFS_DELEG_PERM_GID, &gid)) != 0) {
1644 		nvlist_free(nvp);
1645 		return (EPERM);
1646 	}
1647 
1648 	if ((error = nvlist_lookup_uint32_array(nvp, ZFS_DELEG_PERM_GROUPS,
1649 	    &groups, &group_cnt)) != 0) {
1650 		nvlist_free(nvp);
1651 		return (EPERM);
1652 	}
1653 	usercred = cralloc();
1654 	if ((crsetugid(usercred, uid, gid) != 0) ||
1655 	    (crsetgroups(usercred, group_cnt, (gid_t *)groups) != 0)) {
1656 		nvlist_free(nvp);
1657 		crfree(usercred);
1658 		return (EPERM);
1659 	}
1660 	nvlist_free(nvp);
1661 	error = dsl_deleg_access(zc->zc_name,
1662 	    zfs_prop_to_name(ZFS_PROP_SHAREISCSI), usercred);
1663 	crfree(usercred);
1664 	return (error);
1665 }
1666 
1667 /*
1668  * inputs:
1669  * zc_name		name of filesystem
1670  * zc_nvlist_src{_size}	nvlist of delegated permissions
1671  * zc_perm_action	allow/unallow flag
1672  *
1673  * outputs:		none
1674  */
1675 static int
1676 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
1677 {
1678 	int error;
1679 	nvlist_t *fsaclnv = NULL;
1680 
1681 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1682 	    &fsaclnv)) != 0)
1683 		return (error);
1684 
1685 	/*
1686 	 * Verify nvlist is constructed correctly
1687 	 */
1688 	if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
1689 		nvlist_free(fsaclnv);
1690 		return (EINVAL);
1691 	}
1692 
1693 	/*
1694 	 * If we don't have PRIV_SYS_MOUNT, then validate
1695 	 * that user is allowed to hand out each permission in
1696 	 * the nvlist(s)
1697 	 */
1698 
1699 	error = secpolicy_zfs(CRED());
1700 	if (error) {
1701 		if (zc->zc_perm_action == B_FALSE) {
1702 			error = dsl_deleg_can_allow(zc->zc_name,
1703 			    fsaclnv, CRED());
1704 		} else {
1705 			error = dsl_deleg_can_unallow(zc->zc_name,
1706 			    fsaclnv, CRED());
1707 		}
1708 	}
1709 
1710 	if (error == 0)
1711 		error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
1712 
1713 	nvlist_free(fsaclnv);
1714 	return (error);
1715 }
1716 
1717 /*
1718  * inputs:
1719  * zc_name		name of filesystem
1720  *
1721  * outputs:
1722  * zc_nvlist_src{_size}	nvlist of delegated permissions
1723  */
1724 static int
1725 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
1726 {
1727 	nvlist_t *nvp;
1728 	int error;
1729 
1730 	if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
1731 		error = put_nvlist(zc, nvp);
1732 		nvlist_free(nvp);
1733 	}
1734 
1735 	return (error);
1736 }
1737 
1738 /*
1739  * inputs:
1740  * zc_name		name of volume
1741  *
1742  * outputs:		none
1743  */
1744 static int
1745 zfs_ioc_create_minor(zfs_cmd_t *zc)
1746 {
1747 	return (zvol_create_minor(zc->zc_name, ddi_driver_major(zfs_dip)));
1748 }
1749 
1750 /*
1751  * inputs:
1752  * zc_name		name of volume
1753  *
1754  * outputs:		none
1755  */
1756 static int
1757 zfs_ioc_remove_minor(zfs_cmd_t *zc)
1758 {
1759 	return (zvol_remove_minor(zc->zc_name));
1760 }
1761 
1762 /*
1763  * Search the vfs list for a specified resource.  Returns a pointer to it
1764  * or NULL if no suitable entry is found. The caller of this routine
1765  * is responsible for releasing the returned vfs pointer.
1766  */
1767 static vfs_t *
1768 zfs_get_vfs(const char *resource)
1769 {
1770 	struct vfs *vfsp;
1771 	struct vfs *vfs_found = NULL;
1772 
1773 	vfs_list_read_lock();
1774 	vfsp = rootvfs;
1775 	do {
1776 		if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
1777 			VFS_HOLD(vfsp);
1778 			vfs_found = vfsp;
1779 			break;
1780 		}
1781 		vfsp = vfsp->vfs_next;
1782 	} while (vfsp != rootvfs);
1783 	vfs_list_unlock();
1784 	return (vfs_found);
1785 }
1786 
1787 /* ARGSUSED */
1788 static void
1789 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
1790 {
1791 	zfs_creat_t *zct = arg;
1792 
1793 	zfs_create_fs(os, cr, zct->zct_zplprops, tx);
1794 }
1795 
1796 #define	ZFS_PROP_UNDEFINED	((uint64_t)-1)
1797 
1798 /*
1799  * inputs:
1800  * createprops	list of properties requested by creator
1801  * dataset	name of dataset we are creating
1802  *
1803  * outputs:
1804  * zplprops	values for the zplprops we attach to the master node object
1805  *
1806  * Determine the settings for utf8only, normalization and
1807  * casesensitivity.  Specific values may have been requested by the
1808  * creator and/or we can inherit values from the parent dataset.  If
1809  * the file system is of too early a vintage, a creator can not
1810  * request settings for these properties, even if the requested
1811  * setting is the default value.  We don't actually want to create dsl
1812  * properties for these, so remove them from the source nvlist after
1813  * processing.
1814  */
1815 static int
1816 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
1817     nvlist_t *zplprops, uint64_t zplver, boolean_t *is_ci)
1818 {
1819 	objset_t *os;
1820 	char parentname[MAXNAMELEN];
1821 	char *cp;
1822 	uint64_t sense = ZFS_PROP_UNDEFINED;
1823 	uint64_t norm = ZFS_PROP_UNDEFINED;
1824 	uint64_t u8 = ZFS_PROP_UNDEFINED;
1825 	int error = 0;
1826 
1827 	ASSERT(zplprops != NULL);
1828 
1829 	(void) strlcpy(parentname, dataset, sizeof (parentname));
1830 	cp = strrchr(parentname, '/');
1831 	ASSERT(cp != NULL);
1832 	cp[0] = '\0';
1833 
1834 	/*
1835 	 * Pull out creator prop choices, if any.
1836 	 */
1837 	if (createprops) {
1838 		(void) nvlist_lookup_uint64(createprops,
1839 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
1840 		(void) nvlist_remove_all(createprops,
1841 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE));
1842 		(void) nvlist_lookup_uint64(createprops,
1843 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
1844 		(void) nvlist_remove_all(createprops,
1845 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
1846 		(void) nvlist_lookup_uint64(createprops,
1847 		    zfs_prop_to_name(ZFS_PROP_CASE), &sense);
1848 		(void) nvlist_remove_all(createprops,
1849 		    zfs_prop_to_name(ZFS_PROP_CASE));
1850 	}
1851 
1852 	/*
1853 	 * If the file system or pool is version is too "young" to
1854 	 * support normalization and the creator tried to set a value
1855 	 * for one of the props, error out.  We only need check the
1856 	 * ZPL version because we've already checked by now that the
1857 	 * SPA version is compatible with the selected ZPL version.
1858 	 */
1859 	if (zplver < ZPL_VERSION_NORMALIZATION &&
1860 	    (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
1861 	    sense != ZFS_PROP_UNDEFINED))
1862 		return (ENOTSUP);
1863 
1864 	/*
1865 	 * Put the version in the zplprops
1866 	 */
1867 	VERIFY(nvlist_add_uint64(zplprops,
1868 	    zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
1869 
1870 	/*
1871 	 * Open parent object set so we can inherit zplprop values if
1872 	 * necessary.
1873 	 */
1874 	if (error = dmu_objset_open(parentname,
1875 	    DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os))
1876 		return (error);
1877 
1878 	if (norm == ZFS_PROP_UNDEFINED)
1879 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
1880 	VERIFY(nvlist_add_uint64(zplprops,
1881 	    zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
1882 
1883 	/*
1884 	 * If we're normalizing, names must always be valid UTF-8 strings.
1885 	 */
1886 	if (norm)
1887 		u8 = 1;
1888 	if (u8 == ZFS_PROP_UNDEFINED)
1889 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
1890 	VERIFY(nvlist_add_uint64(zplprops,
1891 	    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
1892 
1893 	if (sense == ZFS_PROP_UNDEFINED)
1894 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
1895 	VERIFY(nvlist_add_uint64(zplprops,
1896 	    zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
1897 
1898 	if (is_ci)
1899 		*is_ci = (sense == ZFS_CASE_INSENSITIVE);
1900 
1901 	dmu_objset_close(os);
1902 	return (0);
1903 }
1904 
1905 /*
1906  * inputs:
1907  * zc_objset_type	type of objset to create (fs vs zvol)
1908  * zc_name		name of new objset
1909  * zc_value		name of snapshot to clone from (may be empty)
1910  * zc_nvlist_src{_size}	nvlist of properties to apply
1911  *
1912  * outputs: none
1913  */
1914 static int
1915 zfs_ioc_create(zfs_cmd_t *zc)
1916 {
1917 	objset_t *clone;
1918 	int error = 0;
1919 	zfs_creat_t zct;
1920 	nvlist_t *nvprops = NULL;
1921 	void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
1922 	dmu_objset_type_t type = zc->zc_objset_type;
1923 
1924 	switch (type) {
1925 
1926 	case DMU_OST_ZFS:
1927 		cbfunc = zfs_create_cb;
1928 		break;
1929 
1930 	case DMU_OST_ZVOL:
1931 		cbfunc = zvol_create_cb;
1932 		break;
1933 
1934 	default:
1935 		cbfunc = NULL;
1936 		break;
1937 	}
1938 	if (strchr(zc->zc_name, '@') ||
1939 	    strchr(zc->zc_name, '%'))
1940 		return (EINVAL);
1941 
1942 	if (zc->zc_nvlist_src != NULL &&
1943 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1944 	    &nvprops)) != 0)
1945 		return (error);
1946 
1947 	zct.zct_zplprops = NULL;
1948 	zct.zct_props = nvprops;
1949 
1950 	if (zc->zc_value[0] != '\0') {
1951 		/*
1952 		 * We're creating a clone of an existing snapshot.
1953 		 */
1954 		zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
1955 		if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0) {
1956 			nvlist_free(nvprops);
1957 			return (EINVAL);
1958 		}
1959 
1960 		error = dmu_objset_open(zc->zc_value, type,
1961 		    DS_MODE_USER | DS_MODE_READONLY, &clone);
1962 		if (error) {
1963 			nvlist_free(nvprops);
1964 			return (error);
1965 		}
1966 
1967 		error = dmu_objset_create(zc->zc_name, type, clone, 0,
1968 		    NULL, NULL);
1969 		if (error) {
1970 			dmu_objset_close(clone);
1971 			nvlist_free(nvprops);
1972 			return (error);
1973 		}
1974 		dmu_objset_close(clone);
1975 	} else {
1976 		boolean_t is_insensitive = B_FALSE;
1977 
1978 		if (cbfunc == NULL) {
1979 			nvlist_free(nvprops);
1980 			return (EINVAL);
1981 		}
1982 
1983 		if (type == DMU_OST_ZVOL) {
1984 			uint64_t volsize, volblocksize;
1985 
1986 			if (nvprops == NULL ||
1987 			    nvlist_lookup_uint64(nvprops,
1988 			    zfs_prop_to_name(ZFS_PROP_VOLSIZE),
1989 			    &volsize) != 0) {
1990 				nvlist_free(nvprops);
1991 				return (EINVAL);
1992 			}
1993 
1994 			if ((error = nvlist_lookup_uint64(nvprops,
1995 			    zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
1996 			    &volblocksize)) != 0 && error != ENOENT) {
1997 				nvlist_free(nvprops);
1998 				return (EINVAL);
1999 			}
2000 
2001 			if (error != 0)
2002 				volblocksize = zfs_prop_default_numeric(
2003 				    ZFS_PROP_VOLBLOCKSIZE);
2004 
2005 			if ((error = zvol_check_volblocksize(
2006 			    volblocksize)) != 0 ||
2007 			    (error = zvol_check_volsize(volsize,
2008 			    volblocksize)) != 0) {
2009 				nvlist_free(nvprops);
2010 				return (error);
2011 			}
2012 		} else if (type == DMU_OST_ZFS) {
2013 			uint64_t version;
2014 			int error;
2015 
2016 			/*
2017 			 * Default ZPL version to non-FUID capable if the
2018 			 * pool is not upgraded to support FUIDs.
2019 			 */
2020 			if (zfs_check_version(zc->zc_name, SPA_VERSION_FUID))
2021 				version = ZPL_VERSION_FUID - 1;
2022 			else
2023 				version = ZPL_VERSION;
2024 
2025 			/*
2026 			 * Potentially override default ZPL version based
2027 			 * on creator's request.
2028 			 */
2029 			(void) nvlist_lookup_uint64(nvprops,
2030 			    zfs_prop_to_name(ZFS_PROP_VERSION), &version);
2031 
2032 			/*
2033 			 * Make sure version we ended up with is kosher
2034 			 */
2035 			if ((version < ZPL_VERSION_INITIAL ||
2036 			    version > ZPL_VERSION) ||
2037 			    (version >= ZPL_VERSION_FUID &&
2038 			    zfs_check_version(zc->zc_name, SPA_VERSION_FUID))) {
2039 				nvlist_free(nvprops);
2040 				return (ENOTSUP);
2041 			}
2042 
2043 			/*
2044 			 * We have to have normalization and
2045 			 * case-folding flags correct when we do the
2046 			 * file system creation, so go figure them out
2047 			 * now.
2048 			 */
2049 			VERIFY(nvlist_alloc(&zct.zct_zplprops,
2050 			    NV_UNIQUE_NAME, KM_SLEEP) == 0);
2051 			error = zfs_fill_zplprops(zc->zc_name, nvprops,
2052 			    zct.zct_zplprops, version, &is_insensitive);
2053 			if (error != 0) {
2054 				nvlist_free(nvprops);
2055 				nvlist_free(zct.zct_zplprops);
2056 				return (error);
2057 			}
2058 		}
2059 		error = dmu_objset_create(zc->zc_name, type, NULL,
2060 		    is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
2061 		nvlist_free(zct.zct_zplprops);
2062 	}
2063 
2064 	/*
2065 	 * It would be nice to do this atomically.
2066 	 */
2067 	if (error == 0) {
2068 		if ((error = zfs_set_prop_nvlist(zc->zc_name, nvprops)) != 0)
2069 			(void) dmu_objset_destroy(zc->zc_name);
2070 	}
2071 	nvlist_free(nvprops);
2072 	return (error);
2073 }
2074 
2075 /*
2076  * inputs:
2077  * zc_name	name of filesystem
2078  * zc_value	short name of snapshot
2079  * zc_cookie	recursive flag
2080  *
2081  * outputs:	none
2082  */
2083 static int
2084 zfs_ioc_snapshot(zfs_cmd_t *zc)
2085 {
2086 	if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
2087 		return (EINVAL);
2088 	return (dmu_objset_snapshot(zc->zc_name,
2089 	    zc->zc_value, zc->zc_cookie));
2090 }
2091 
2092 int
2093 zfs_unmount_snap(char *name, void *arg)
2094 {
2095 	vfs_t *vfsp = NULL;
2096 
2097 	if (arg) {
2098 		char *snapname = arg;
2099 		int len = strlen(name) + strlen(snapname) + 2;
2100 		char *buf = kmem_alloc(len, KM_SLEEP);
2101 
2102 		(void) strcpy(buf, name);
2103 		(void) strcat(buf, "@");
2104 		(void) strcat(buf, snapname);
2105 		vfsp = zfs_get_vfs(buf);
2106 		kmem_free(buf, len);
2107 	} else if (strchr(name, '@')) {
2108 		vfsp = zfs_get_vfs(name);
2109 	}
2110 
2111 	if (vfsp) {
2112 		/*
2113 		 * Always force the unmount for snapshots.
2114 		 */
2115 		int flag = MS_FORCE;
2116 		int err;
2117 
2118 		if ((err = vn_vfswlock(vfsp->vfs_vnodecovered)) != 0) {
2119 			VFS_RELE(vfsp);
2120 			return (err);
2121 		}
2122 		VFS_RELE(vfsp);
2123 		if ((err = dounmount(vfsp, flag, kcred)) != 0)
2124 			return (err);
2125 	}
2126 	return (0);
2127 }
2128 
2129 /*
2130  * inputs:
2131  * zc_name	name of filesystem
2132  * zc_value	short name of snapshot
2133  *
2134  * outputs:	none
2135  */
2136 static int
2137 zfs_ioc_destroy_snaps(zfs_cmd_t *zc)
2138 {
2139 	int err;
2140 
2141 	if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
2142 		return (EINVAL);
2143 	err = dmu_objset_find(zc->zc_name,
2144 	    zfs_unmount_snap, zc->zc_value, DS_FIND_CHILDREN);
2145 	if (err)
2146 		return (err);
2147 	return (dmu_snapshots_destroy(zc->zc_name, zc->zc_value));
2148 }
2149 
2150 /*
2151  * inputs:
2152  * zc_name		name of dataset to destroy
2153  * zc_objset_type	type of objset
2154  *
2155  * outputs:		none
2156  */
2157 static int
2158 zfs_ioc_destroy(zfs_cmd_t *zc)
2159 {
2160 	if (strchr(zc->zc_name, '@') && zc->zc_objset_type == DMU_OST_ZFS) {
2161 		int err = zfs_unmount_snap(zc->zc_name, NULL);
2162 		if (err)
2163 			return (err);
2164 	}
2165 
2166 	return (dmu_objset_destroy(zc->zc_name));
2167 }
2168 
2169 /*
2170  * inputs:
2171  * zc_name	name of dataset to rollback (to most recent snapshot)
2172  *
2173  * outputs:	none
2174  */
2175 static int
2176 zfs_ioc_rollback(zfs_cmd_t *zc)
2177 {
2178 	objset_t *os;
2179 	int error;
2180 	zfsvfs_t *zfsvfs = NULL;
2181 
2182 	/*
2183 	 * Get the zfsvfs for the receiving objset. There
2184 	 * won't be one if we're operating on a zvol, if the
2185 	 * objset doesn't exist yet, or is not mounted.
2186 	 */
2187 	error = dmu_objset_open(zc->zc_name, DMU_OST_ANY, DS_MODE_USER, &os);
2188 	if (error)
2189 		return (error);
2190 
2191 	if (dmu_objset_type(os) == DMU_OST_ZFS) {
2192 		mutex_enter(&os->os->os_user_ptr_lock);
2193 		zfsvfs = dmu_objset_get_user(os);
2194 		if (zfsvfs != NULL)
2195 			VFS_HOLD(zfsvfs->z_vfs);
2196 		mutex_exit(&os->os->os_user_ptr_lock);
2197 	}
2198 
2199 	if (zfsvfs != NULL) {
2200 		char osname[MAXNAMELEN];
2201 		int mode;
2202 
2203 		error = zfs_suspend_fs(zfsvfs, osname, &mode);
2204 		if (error == 0) {
2205 			int resume_err;
2206 
2207 			ASSERT(strcmp(osname, zc->zc_name) == 0);
2208 			error = dmu_objset_rollback(os);
2209 			resume_err = zfs_resume_fs(zfsvfs, osname, mode);
2210 			error = error ? error : resume_err;
2211 		} else {
2212 			dmu_objset_close(os);
2213 		}
2214 		VFS_RELE(zfsvfs->z_vfs);
2215 	} else {
2216 		error = dmu_objset_rollback(os);
2217 	}
2218 	/* Note, the dmu_objset_rollback() releases the objset for us. */
2219 
2220 	return (error);
2221 }
2222 
2223 /*
2224  * inputs:
2225  * zc_name	old name of dataset
2226  * zc_value	new name of dataset
2227  * zc_cookie	recursive flag (only valid for snapshots)
2228  *
2229  * outputs:	none
2230  */
2231 static int
2232 zfs_ioc_rename(zfs_cmd_t *zc)
2233 {
2234 	boolean_t recursive = zc->zc_cookie & 1;
2235 
2236 	zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
2237 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
2238 	    strchr(zc->zc_value, '%'))
2239 		return (EINVAL);
2240 
2241 	/*
2242 	 * Unmount snapshot unless we're doing a recursive rename,
2243 	 * in which case the dataset code figures out which snapshots
2244 	 * to unmount.
2245 	 */
2246 	if (!recursive && strchr(zc->zc_name, '@') != NULL &&
2247 	    zc->zc_objset_type == DMU_OST_ZFS) {
2248 		int err = zfs_unmount_snap(zc->zc_name, NULL);
2249 		if (err)
2250 			return (err);
2251 	}
2252 	return (dmu_objset_rename(zc->zc_name, zc->zc_value, recursive));
2253 }
2254 
2255 static void
2256 clear_props(char *dataset, nvlist_t *props)
2257 {
2258 	zfs_cmd_t *zc;
2259 	nvpair_t *prop;
2260 
2261 	if (props == NULL)
2262 		return;
2263 	zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
2264 	(void) strcpy(zc->zc_name, dataset);
2265 	for (prop = nvlist_next_nvpair(props, NULL); prop;
2266 	    prop = nvlist_next_nvpair(props, prop)) {
2267 		(void) strcpy(zc->zc_value, nvpair_name(prop));
2268 		if (zfs_secpolicy_inherit(zc, CRED()) == 0)
2269 			(void) zfs_ioc_inherit_prop(zc);
2270 	}
2271 	kmem_free(zc, sizeof (zfs_cmd_t));
2272 }
2273 
2274 /*
2275  * inputs:
2276  * zc_name		name of containing filesystem
2277  * zc_nvlist_src{_size}	nvlist of properties to apply
2278  * zc_value		name of snapshot to create
2279  * zc_string		name of clone origin (if DRR_FLAG_CLONE)
2280  * zc_cookie		file descriptor to recv from
2281  * zc_begin_record	the BEGIN record of the stream (not byteswapped)
2282  * zc_guid		force flag
2283  *
2284  * outputs:
2285  * zc_cookie		number of bytes read
2286  */
2287 static int
2288 zfs_ioc_recv(zfs_cmd_t *zc)
2289 {
2290 	file_t *fp;
2291 	objset_t *os;
2292 	dmu_recv_cookie_t drc;
2293 	zfsvfs_t *zfsvfs = NULL;
2294 	boolean_t force = (boolean_t)zc->zc_guid;
2295 	int error, fd;
2296 	offset_t off;
2297 	nvlist_t *props = NULL;
2298 	nvlist_t *origprops = NULL;
2299 	objset_t *origin = NULL;
2300 	char *tosnap;
2301 	char tofs[ZFS_MAXNAMELEN];
2302 
2303 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
2304 	    strchr(zc->zc_value, '@') == NULL ||
2305 	    strchr(zc->zc_value, '%'))
2306 		return (EINVAL);
2307 
2308 	(void) strcpy(tofs, zc->zc_value);
2309 	tosnap = strchr(tofs, '@');
2310 	*tosnap = '\0';
2311 	tosnap++;
2312 
2313 	if (zc->zc_nvlist_src != NULL &&
2314 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2315 	    &props)) != 0)
2316 		return (error);
2317 
2318 	fd = zc->zc_cookie;
2319 	fp = getf(fd);
2320 	if (fp == NULL) {
2321 		nvlist_free(props);
2322 		return (EBADF);
2323 	}
2324 
2325 	if (dmu_objset_open(tofs, DMU_OST_ANY,
2326 	    DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
2327 		/*
2328 		 * Try to get the zfsvfs for the receiving objset.
2329 		 * There won't be one if we're operating on a zvol,
2330 		 * if the objset doesn't exist yet, or is not mounted.
2331 		 */
2332 		mutex_enter(&os->os->os_user_ptr_lock);
2333 		if (zfsvfs = dmu_objset_get_user(os)) {
2334 			if (!mutex_tryenter(&zfsvfs->z_online_recv_lock)) {
2335 				mutex_exit(&os->os->os_user_ptr_lock);
2336 				dmu_objset_close(os);
2337 				zfsvfs = NULL;
2338 				error = EBUSY;
2339 				goto out;
2340 			}
2341 			VFS_HOLD(zfsvfs->z_vfs);
2342 		}
2343 		mutex_exit(&os->os->os_user_ptr_lock);
2344 
2345 		/*
2346 		 * If new properties are supplied, they are to completely
2347 		 * replace the existing ones, so stash away the existing ones.
2348 		 */
2349 		if (props)
2350 			(void) dsl_prop_get_all(os, &origprops, TRUE);
2351 
2352 		dmu_objset_close(os);
2353 	}
2354 
2355 	if (zc->zc_string[0]) {
2356 		error = dmu_objset_open(zc->zc_string, DMU_OST_ANY,
2357 		    DS_MODE_USER | DS_MODE_READONLY, &origin);
2358 		if (error)
2359 			goto out;
2360 	}
2361 
2362 	error = dmu_recv_begin(tofs, tosnap, &zc->zc_begin_record,
2363 	    force, origin, zfsvfs != NULL, &drc);
2364 	if (origin)
2365 		dmu_objset_close(origin);
2366 	if (error)
2367 		goto out;
2368 
2369 	/*
2370 	 * Reset properties.  We do this before we receive the stream
2371 	 * so that the properties are applied to the new data.
2372 	 */
2373 	if (props) {
2374 		clear_props(tofs, origprops);
2375 		/*
2376 		 * XXX - Note, this is all-or-nothing; should be best-effort.
2377 		 */
2378 		(void) zfs_set_prop_nvlist(tofs, props);
2379 	}
2380 
2381 	off = fp->f_offset;
2382 	error = dmu_recv_stream(&drc, fp->f_vnode, &off);
2383 
2384 	if (error == 0 && zfsvfs) {
2385 		char osname[MAXNAMELEN];
2386 		int mode;
2387 
2388 		/* online recv */
2389 		error = zfs_suspend_fs(zfsvfs, osname, &mode);
2390 		if (error == 0) {
2391 			int resume_err;
2392 
2393 			error = dmu_recv_end(&drc);
2394 			resume_err = zfs_resume_fs(zfsvfs, osname, mode);
2395 			error = error ? error : resume_err;
2396 		} else {
2397 			dmu_recv_abort_cleanup(&drc);
2398 		}
2399 	} else if (error == 0) {
2400 		error = dmu_recv_end(&drc);
2401 	}
2402 
2403 	zc->zc_cookie = off - fp->f_offset;
2404 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
2405 		fp->f_offset = off;
2406 
2407 	/*
2408 	 * On error, restore the original props.
2409 	 */
2410 	if (error && props) {
2411 		clear_props(tofs, props);
2412 		(void) zfs_set_prop_nvlist(tofs, origprops);
2413 	}
2414 out:
2415 	if (zfsvfs) {
2416 		mutex_exit(&zfsvfs->z_online_recv_lock);
2417 		VFS_RELE(zfsvfs->z_vfs);
2418 	}
2419 	nvlist_free(props);
2420 	nvlist_free(origprops);
2421 	releasef(fd);
2422 	return (error);
2423 }
2424 
2425 /*
2426  * inputs:
2427  * zc_name	name of snapshot to send
2428  * zc_value	short name of incremental fromsnap (may be empty)
2429  * zc_cookie	file descriptor to send stream to
2430  * zc_obj	fromorigin flag (mutually exclusive with zc_value)
2431  *
2432  * outputs: none
2433  */
2434 static int
2435 zfs_ioc_send(zfs_cmd_t *zc)
2436 {
2437 	objset_t *fromsnap = NULL;
2438 	objset_t *tosnap;
2439 	file_t *fp;
2440 	int error;
2441 	offset_t off;
2442 
2443 	error = dmu_objset_open(zc->zc_name, DMU_OST_ANY,
2444 	    DS_MODE_USER | DS_MODE_READONLY, &tosnap);
2445 	if (error)
2446 		return (error);
2447 
2448 	if (zc->zc_value[0] != '\0') {
2449 		char buf[MAXPATHLEN];
2450 		char *cp;
2451 
2452 		(void) strncpy(buf, zc->zc_name, sizeof (buf));
2453 		cp = strchr(buf, '@');
2454 		if (cp)
2455 			*(cp+1) = 0;
2456 		(void) strncat(buf, zc->zc_value, sizeof (buf));
2457 		error = dmu_objset_open(buf, DMU_OST_ANY,
2458 		    DS_MODE_USER | DS_MODE_READONLY, &fromsnap);
2459 		if (error) {
2460 			dmu_objset_close(tosnap);
2461 			return (error);
2462 		}
2463 	}
2464 
2465 	fp = getf(zc->zc_cookie);
2466 	if (fp == NULL) {
2467 		dmu_objset_close(tosnap);
2468 		if (fromsnap)
2469 			dmu_objset_close(fromsnap);
2470 		return (EBADF);
2471 	}
2472 
2473 	off = fp->f_offset;
2474 	error = dmu_sendbackup(tosnap, fromsnap, zc->zc_obj, fp->f_vnode, &off);
2475 
2476 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
2477 		fp->f_offset = off;
2478 	releasef(zc->zc_cookie);
2479 	if (fromsnap)
2480 		dmu_objset_close(fromsnap);
2481 	dmu_objset_close(tosnap);
2482 	return (error);
2483 }
2484 
2485 static int
2486 zfs_ioc_inject_fault(zfs_cmd_t *zc)
2487 {
2488 	int id, error;
2489 
2490 	error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
2491 	    &zc->zc_inject_record);
2492 
2493 	if (error == 0)
2494 		zc->zc_guid = (uint64_t)id;
2495 
2496 	return (error);
2497 }
2498 
2499 static int
2500 zfs_ioc_clear_fault(zfs_cmd_t *zc)
2501 {
2502 	return (zio_clear_fault((int)zc->zc_guid));
2503 }
2504 
2505 static int
2506 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
2507 {
2508 	int id = (int)zc->zc_guid;
2509 	int error;
2510 
2511 	error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
2512 	    &zc->zc_inject_record);
2513 
2514 	zc->zc_guid = id;
2515 
2516 	return (error);
2517 }
2518 
2519 static int
2520 zfs_ioc_error_log(zfs_cmd_t *zc)
2521 {
2522 	spa_t *spa;
2523 	int error;
2524 	size_t count = (size_t)zc->zc_nvlist_dst_size;
2525 
2526 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2527 		return (error);
2528 
2529 	error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
2530 	    &count);
2531 	if (error == 0)
2532 		zc->zc_nvlist_dst_size = count;
2533 	else
2534 		zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
2535 
2536 	spa_close(spa, FTAG);
2537 
2538 	return (error);
2539 }
2540 
2541 static int
2542 zfs_ioc_clear(zfs_cmd_t *zc)
2543 {
2544 	spa_t *spa;
2545 	vdev_t *vd;
2546 	uint64_t txg;
2547 	int error;
2548 
2549 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2550 		return (error);
2551 
2552 	/*
2553 	 * Try to resume any I/Os which may have been suspended
2554 	 * as a result of a complete pool failure.
2555 	 */
2556 	if (!list_is_empty(&spa->spa_zio_list)) {
2557 		if (zio_vdev_resume_io(spa) != 0) {
2558 			spa_close(spa, FTAG);
2559 			return (EIO);
2560 		}
2561 	}
2562 
2563 	txg = spa_vdev_enter(spa);
2564 
2565 	if (zc->zc_guid == 0) {
2566 		vd = NULL;
2567 	} else {
2568 		vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
2569 		if (vd == NULL) {
2570 			(void) spa_vdev_exit(spa, NULL, txg, ENODEV);
2571 			spa_close(spa, FTAG);
2572 			return (ENODEV);
2573 		}
2574 	}
2575 
2576 	vdev_clear(spa, vd, B_TRUE);
2577 
2578 	(void) spa_vdev_exit(spa, NULL, txg, 0);
2579 
2580 	spa_close(spa, FTAG);
2581 
2582 	return (0);
2583 }
2584 
2585 /*
2586  * inputs:
2587  * zc_name	name of filesystem
2588  * zc_value	name of origin snapshot
2589  *
2590  * outputs:	none
2591  */
2592 static int
2593 zfs_ioc_promote(zfs_cmd_t *zc)
2594 {
2595 	char *cp;
2596 
2597 	/*
2598 	 * We don't need to unmount *all* the origin fs's snapshots, but
2599 	 * it's easier.
2600 	 */
2601 	cp = strchr(zc->zc_value, '@');
2602 	if (cp)
2603 		*cp = '\0';
2604 	(void) dmu_objset_find(zc->zc_value,
2605 	    zfs_unmount_snap, NULL, DS_FIND_SNAPSHOTS);
2606 	return (dsl_dataset_promote(zc->zc_name));
2607 }
2608 
2609 /*
2610  * We don't want to have a hard dependency
2611  * against some special symbols in sharefs
2612  * nfs, and smbsrv.  Determine them if needed when
2613  * the first file system is shared.
2614  * Neither sharefs, nfs or smbsrv are unloadable modules.
2615  */
2616 int (*znfsexport_fs)(void *arg);
2617 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
2618 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
2619 
2620 int zfs_nfsshare_inited;
2621 int zfs_smbshare_inited;
2622 
2623 ddi_modhandle_t nfs_mod;
2624 ddi_modhandle_t sharefs_mod;
2625 ddi_modhandle_t smbsrv_mod;
2626 kmutex_t zfs_share_lock;
2627 
2628 static int
2629 zfs_init_sharefs()
2630 {
2631 	int error;
2632 
2633 	ASSERT(MUTEX_HELD(&zfs_share_lock));
2634 	/* Both NFS and SMB shares also require sharetab support. */
2635 	if (sharefs_mod == NULL && ((sharefs_mod =
2636 	    ddi_modopen("fs/sharefs",
2637 	    KRTLD_MODE_FIRST, &error)) == NULL)) {
2638 		return (ENOSYS);
2639 	}
2640 	if (zshare_fs == NULL && ((zshare_fs =
2641 	    (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
2642 	    ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
2643 		return (ENOSYS);
2644 	}
2645 	return (0);
2646 }
2647 
2648 static int
2649 zfs_ioc_share(zfs_cmd_t *zc)
2650 {
2651 	int error;
2652 	int opcode;
2653 
2654 	switch (zc->zc_share.z_sharetype) {
2655 	case ZFS_SHARE_NFS:
2656 	case ZFS_UNSHARE_NFS:
2657 		if (zfs_nfsshare_inited == 0) {
2658 			mutex_enter(&zfs_share_lock);
2659 			if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
2660 			    KRTLD_MODE_FIRST, &error)) == NULL)) {
2661 				mutex_exit(&zfs_share_lock);
2662 				return (ENOSYS);
2663 			}
2664 			if (znfsexport_fs == NULL &&
2665 			    ((znfsexport_fs = (int (*)(void *))
2666 			    ddi_modsym(nfs_mod,
2667 			    "nfs_export", &error)) == NULL)) {
2668 				mutex_exit(&zfs_share_lock);
2669 				return (ENOSYS);
2670 			}
2671 			error = zfs_init_sharefs();
2672 			if (error) {
2673 				mutex_exit(&zfs_share_lock);
2674 				return (ENOSYS);
2675 			}
2676 			zfs_nfsshare_inited = 1;
2677 			mutex_exit(&zfs_share_lock);
2678 		}
2679 		break;
2680 	case ZFS_SHARE_SMB:
2681 	case ZFS_UNSHARE_SMB:
2682 		if (zfs_smbshare_inited == 0) {
2683 			mutex_enter(&zfs_share_lock);
2684 			if (smbsrv_mod == NULL && ((smbsrv_mod =
2685 			    ddi_modopen("drv/smbsrv",
2686 			    KRTLD_MODE_FIRST, &error)) == NULL)) {
2687 				mutex_exit(&zfs_share_lock);
2688 				return (ENOSYS);
2689 			}
2690 			if (zsmbexport_fs == NULL && ((zsmbexport_fs =
2691 			    (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
2692 			    "smb_server_share", &error)) == NULL)) {
2693 				mutex_exit(&zfs_share_lock);
2694 				return (ENOSYS);
2695 			}
2696 			error = zfs_init_sharefs();
2697 			if (error) {
2698 				mutex_exit(&zfs_share_lock);
2699 				return (ENOSYS);
2700 			}
2701 			zfs_smbshare_inited = 1;
2702 			mutex_exit(&zfs_share_lock);
2703 		}
2704 		break;
2705 	default:
2706 		return (EINVAL);
2707 	}
2708 
2709 	switch (zc->zc_share.z_sharetype) {
2710 	case ZFS_SHARE_NFS:
2711 	case ZFS_UNSHARE_NFS:
2712 		if (error =
2713 		    znfsexport_fs((void *)
2714 		    (uintptr_t)zc->zc_share.z_exportdata))
2715 			return (error);
2716 		break;
2717 	case ZFS_SHARE_SMB:
2718 	case ZFS_UNSHARE_SMB:
2719 		if (error = zsmbexport_fs((void *)
2720 		    (uintptr_t)zc->zc_share.z_exportdata,
2721 		    zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
2722 		    B_TRUE : B_FALSE)) {
2723 			return (error);
2724 		}
2725 		break;
2726 	}
2727 
2728 	opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
2729 	    zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
2730 	    SHAREFS_ADD : SHAREFS_REMOVE;
2731 
2732 	/*
2733 	 * Add or remove share from sharetab
2734 	 */
2735 	error = zshare_fs(opcode,
2736 	    (void *)(uintptr_t)zc->zc_share.z_sharedata,
2737 	    zc->zc_share.z_sharemax);
2738 
2739 	return (error);
2740 
2741 }
2742 
2743 /*
2744  * pool create, destroy, and export don't log the history as part of
2745  * zfsdev_ioctl, but rather zfs_ioc_pool_create, and zfs_ioc_pool_export
2746  * do the logging of those commands.
2747  */
2748 static zfs_ioc_vec_t zfs_ioc_vec[] = {
2749 	{ zfs_ioc_pool_create, zfs_secpolicy_config, POOL_NAME, B_FALSE },
2750 	{ zfs_ioc_pool_destroy,	zfs_secpolicy_config, POOL_NAME, B_FALSE },
2751 	{ zfs_ioc_pool_import, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2752 	{ zfs_ioc_pool_export, zfs_secpolicy_config, POOL_NAME, B_FALSE },
2753 	{ zfs_ioc_pool_configs,	zfs_secpolicy_none, NO_NAME, B_FALSE },
2754 	{ zfs_ioc_pool_stats, zfs_secpolicy_read, POOL_NAME, B_FALSE },
2755 	{ zfs_ioc_pool_tryimport, zfs_secpolicy_config, NO_NAME, B_FALSE },
2756 	{ zfs_ioc_pool_scrub, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2757 	{ zfs_ioc_pool_freeze, zfs_secpolicy_config, NO_NAME, B_FALSE },
2758 	{ zfs_ioc_pool_upgrade,	zfs_secpolicy_config, POOL_NAME, B_TRUE },
2759 	{ zfs_ioc_pool_get_history, zfs_secpolicy_config, POOL_NAME, B_FALSE },
2760 	{ zfs_ioc_vdev_add, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2761 	{ zfs_ioc_vdev_remove, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2762 	{ zfs_ioc_vdev_set_state, zfs_secpolicy_config,	POOL_NAME, B_TRUE },
2763 	{ zfs_ioc_vdev_attach, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2764 	{ zfs_ioc_vdev_detach, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2765 	{ zfs_ioc_vdev_setpath,	zfs_secpolicy_config, POOL_NAME, B_FALSE },
2766 	{ zfs_ioc_objset_stats,	zfs_secpolicy_read, DATASET_NAME, B_FALSE },
2767 	{ zfs_ioc_objset_zplprops, zfs_secpolicy_read, DATASET_NAME, B_FALSE },
2768 	{ zfs_ioc_dataset_list_next, zfs_secpolicy_read,
2769 	    DATASET_NAME, B_FALSE },
2770 	{ zfs_ioc_snapshot_list_next, zfs_secpolicy_read,
2771 	    DATASET_NAME, B_FALSE },
2772 	{ zfs_ioc_set_prop, zfs_secpolicy_none, DATASET_NAME, B_TRUE },
2773 	{ zfs_ioc_create_minor,	zfs_secpolicy_minor, DATASET_NAME, B_FALSE },
2774 	{ zfs_ioc_remove_minor,	zfs_secpolicy_minor, DATASET_NAME, B_FALSE },
2775 	{ zfs_ioc_create, zfs_secpolicy_create, DATASET_NAME, B_TRUE },
2776 	{ zfs_ioc_destroy, zfs_secpolicy_destroy, DATASET_NAME, B_TRUE },
2777 	{ zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, B_TRUE },
2778 	{ zfs_ioc_rename, zfs_secpolicy_rename,	DATASET_NAME, B_TRUE },
2779 	{ zfs_ioc_recv, zfs_secpolicy_receive, DATASET_NAME, B_TRUE },
2780 	{ zfs_ioc_send, zfs_secpolicy_send, DATASET_NAME, B_TRUE },
2781 	{ zfs_ioc_inject_fault,	zfs_secpolicy_inject, NO_NAME, B_FALSE },
2782 	{ zfs_ioc_clear_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE },
2783 	{ zfs_ioc_inject_list_next, zfs_secpolicy_inject, NO_NAME, B_FALSE },
2784 	{ zfs_ioc_error_log, zfs_secpolicy_inject, POOL_NAME, B_FALSE },
2785 	{ zfs_ioc_clear, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2786 	{ zfs_ioc_promote, zfs_secpolicy_promote, DATASET_NAME, B_TRUE },
2787 	{ zfs_ioc_destroy_snaps, zfs_secpolicy_destroy,	DATASET_NAME, B_TRUE },
2788 	{ zfs_ioc_snapshot, zfs_secpolicy_snapshot, DATASET_NAME, B_TRUE },
2789 	{ zfs_ioc_dsobj_to_dsname, zfs_secpolicy_config, POOL_NAME, B_FALSE },
2790 	{ zfs_ioc_obj_to_path, zfs_secpolicy_config, NO_NAME, B_FALSE },
2791 	{ zfs_ioc_pool_set_props, zfs_secpolicy_config,	POOL_NAME, B_TRUE },
2792 	{ zfs_ioc_pool_get_props, zfs_secpolicy_read, POOL_NAME, B_FALSE },
2793 	{ zfs_ioc_set_fsacl, zfs_secpolicy_fsacl, DATASET_NAME, B_TRUE },
2794 	{ zfs_ioc_get_fsacl, zfs_secpolicy_read, DATASET_NAME, B_FALSE },
2795 	{ zfs_ioc_iscsi_perm_check, zfs_secpolicy_iscsi,
2796 	    DATASET_NAME, B_FALSE },
2797 	{ zfs_ioc_share, zfs_secpolicy_share, DATASET_NAME, B_FALSE },
2798 	{ zfs_ioc_inherit_prop, zfs_secpolicy_inherit, DATASET_NAME, B_TRUE },
2799 };
2800 
2801 static int
2802 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
2803 {
2804 	zfs_cmd_t *zc;
2805 	uint_t vec;
2806 	int error, rc;
2807 
2808 	if (getminor(dev) != 0)
2809 		return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
2810 
2811 	vec = cmd - ZFS_IOC;
2812 	ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
2813 
2814 	if (vec >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
2815 		return (EINVAL);
2816 
2817 	zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2818 
2819 	error = xcopyin((void *)arg, zc, sizeof (zfs_cmd_t));
2820 
2821 	if (error == 0)
2822 		error = zfs_ioc_vec[vec].zvec_secpolicy(zc, cr);
2823 
2824 	/*
2825 	 * Ensure that all pool/dataset names are valid before we pass down to
2826 	 * the lower layers.
2827 	 */
2828 	if (error == 0) {
2829 		zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
2830 		switch (zfs_ioc_vec[vec].zvec_namecheck) {
2831 		case POOL_NAME:
2832 			if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
2833 				error = EINVAL;
2834 			break;
2835 
2836 		case DATASET_NAME:
2837 			if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
2838 				error = EINVAL;
2839 			break;
2840 
2841 		case NO_NAME:
2842 			break;
2843 		}
2844 	}
2845 
2846 	if (error == 0)
2847 		error = zfs_ioc_vec[vec].zvec_func(zc);
2848 
2849 	rc = xcopyout(zc, (void *)arg, sizeof (zfs_cmd_t));
2850 	if (error == 0) {
2851 		error = rc;
2852 		if (zfs_ioc_vec[vec].zvec_his_log == B_TRUE)
2853 			zfs_log_history(zc);
2854 	}
2855 
2856 	kmem_free(zc, sizeof (zfs_cmd_t));
2857 	return (error);
2858 }
2859 
2860 static int
2861 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
2862 {
2863 	if (cmd != DDI_ATTACH)
2864 		return (DDI_FAILURE);
2865 
2866 	if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
2867 	    DDI_PSEUDO, 0) == DDI_FAILURE)
2868 		return (DDI_FAILURE);
2869 
2870 	zfs_dip = dip;
2871 
2872 	ddi_report_dev(dip);
2873 
2874 	return (DDI_SUCCESS);
2875 }
2876 
2877 static int
2878 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
2879 {
2880 	if (spa_busy() || zfs_busy() || zvol_busy())
2881 		return (DDI_FAILURE);
2882 
2883 	if (cmd != DDI_DETACH)
2884 		return (DDI_FAILURE);
2885 
2886 	zfs_dip = NULL;
2887 
2888 	ddi_prop_remove_all(dip);
2889 	ddi_remove_minor_node(dip, NULL);
2890 
2891 	return (DDI_SUCCESS);
2892 }
2893 
2894 /*ARGSUSED*/
2895 static int
2896 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
2897 {
2898 	switch (infocmd) {
2899 	case DDI_INFO_DEVT2DEVINFO:
2900 		*result = zfs_dip;
2901 		return (DDI_SUCCESS);
2902 
2903 	case DDI_INFO_DEVT2INSTANCE:
2904 		*result = (void *)0;
2905 		return (DDI_SUCCESS);
2906 	}
2907 
2908 	return (DDI_FAILURE);
2909 }
2910 
2911 /*
2912  * OK, so this is a little weird.
2913  *
2914  * /dev/zfs is the control node, i.e. minor 0.
2915  * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
2916  *
2917  * /dev/zfs has basically nothing to do except serve up ioctls,
2918  * so most of the standard driver entry points are in zvol.c.
2919  */
2920 static struct cb_ops zfs_cb_ops = {
2921 	zvol_open,	/* open */
2922 	zvol_close,	/* close */
2923 	zvol_strategy,	/* strategy */
2924 	nodev,		/* print */
2925 	zvol_dump,	/* dump */
2926 	zvol_read,	/* read */
2927 	zvol_write,	/* write */
2928 	zfsdev_ioctl,	/* ioctl */
2929 	nodev,		/* devmap */
2930 	nodev,		/* mmap */
2931 	nodev,		/* segmap */
2932 	nochpoll,	/* poll */
2933 	ddi_prop_op,	/* prop_op */
2934 	NULL,		/* streamtab */
2935 	D_NEW | D_MP | D_64BIT,		/* Driver compatibility flag */
2936 	CB_REV,		/* version */
2937 	nodev,		/* async read */
2938 	nodev,		/* async write */
2939 };
2940 
2941 static struct dev_ops zfs_dev_ops = {
2942 	DEVO_REV,	/* version */
2943 	0,		/* refcnt */
2944 	zfs_info,	/* info */
2945 	nulldev,	/* identify */
2946 	nulldev,	/* probe */
2947 	zfs_attach,	/* attach */
2948 	zfs_detach,	/* detach */
2949 	nodev,		/* reset */
2950 	&zfs_cb_ops,	/* driver operations */
2951 	NULL		/* no bus operations */
2952 };
2953 
2954 static struct modldrv zfs_modldrv = {
2955 	&mod_driverops, "ZFS storage pool version " SPA_VERSION_STRING,
2956 	    &zfs_dev_ops
2957 };
2958 
2959 static struct modlinkage modlinkage = {
2960 	MODREV_1,
2961 	(void *)&zfs_modlfs,
2962 	(void *)&zfs_modldrv,
2963 	NULL
2964 };
2965 
2966 
2967 uint_t zfs_fsyncer_key;
2968 extern uint_t rrw_tsd_key;
2969 
2970 int
2971 _init(void)
2972 {
2973 	int error;
2974 
2975 	spa_init(FREAD | FWRITE);
2976 	zfs_init();
2977 	zvol_init();
2978 
2979 	if ((error = mod_install(&modlinkage)) != 0) {
2980 		zvol_fini();
2981 		zfs_fini();
2982 		spa_fini();
2983 		return (error);
2984 	}
2985 
2986 	tsd_create(&zfs_fsyncer_key, NULL);
2987 	tsd_create(&rrw_tsd_key, NULL);
2988 
2989 	error = ldi_ident_from_mod(&modlinkage, &zfs_li);
2990 	ASSERT(error == 0);
2991 	mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
2992 
2993 	return (0);
2994 }
2995 
2996 int
2997 _fini(void)
2998 {
2999 	int error;
3000 
3001 	if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
3002 		return (EBUSY);
3003 
3004 	if ((error = mod_remove(&modlinkage)) != 0)
3005 		return (error);
3006 
3007 	zvol_fini();
3008 	zfs_fini();
3009 	spa_fini();
3010 	if (zfs_nfsshare_inited)
3011 		(void) ddi_modclose(nfs_mod);
3012 	if (zfs_smbshare_inited)
3013 		(void) ddi_modclose(smbsrv_mod);
3014 	if (zfs_nfsshare_inited || zfs_smbshare_inited)
3015 		(void) ddi_modclose(sharefs_mod);
3016 
3017 	tsd_destroy(&zfs_fsyncer_key);
3018 	ldi_ident_release(zfs_li);
3019 	zfs_li = NULL;
3020 	mutex_destroy(&zfs_share_lock);
3021 
3022 	return (error);
3023 }
3024 
3025 int
3026 _info(struct modinfo *modinfop)
3027 {
3028 	return (mod_info(&modlinkage, modinfop));
3029 }
3030