xref: /titanic_51/usr/src/uts/common/fs/zfs/zfs_ioctl.c (revision f06271be56df67ca3faa4ca4bc51457dad15c3b5)
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 #include <sys/types.h>
27 #include <sys/param.h>
28 #include <sys/errno.h>
29 #include <sys/uio.h>
30 #include <sys/buf.h>
31 #include <sys/modctl.h>
32 #include <sys/open.h>
33 #include <sys/file.h>
34 #include <sys/kmem.h>
35 #include <sys/conf.h>
36 #include <sys/cmn_err.h>
37 #include <sys/stat.h>
38 #include <sys/zfs_ioctl.h>
39 #include <sys/zfs_znode.h>
40 #include <sys/zap.h>
41 #include <sys/spa.h>
42 #include <sys/spa_impl.h>
43 #include <sys/vdev.h>
44 #include <sys/vdev_impl.h>
45 #include <sys/dmu.h>
46 #include <sys/dsl_dir.h>
47 #include <sys/dsl_dataset.h>
48 #include <sys/dsl_prop.h>
49 #include <sys/dsl_deleg.h>
50 #include <sys/dmu_objset.h>
51 #include <sys/ddi.h>
52 #include <sys/sunddi.h>
53 #include <sys/sunldi.h>
54 #include <sys/policy.h>
55 #include <sys/zone.h>
56 #include <sys/nvpair.h>
57 #include <sys/pathname.h>
58 #include <sys/mount.h>
59 #include <sys/sdt.h>
60 #include <sys/fs/zfs.h>
61 #include <sys/zfs_ctldir.h>
62 #include <sys/zfs_dir.h>
63 #include <sys/zvol.h>
64 #include <sharefs/share.h>
65 #include <sys/dmu_objset.h>
66 
67 #include "zfs_namecheck.h"
68 #include "zfs_prop.h"
69 #include "zfs_deleg.h"
70 
71 extern struct modlfs zfs_modlfs;
72 
73 extern void zfs_init(void);
74 extern void zfs_fini(void);
75 
76 ldi_ident_t zfs_li = NULL;
77 dev_info_t *zfs_dip;
78 
79 typedef int zfs_ioc_func_t(zfs_cmd_t *);
80 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, cred_t *);
81 
82 typedef struct zfs_ioc_vec {
83 	zfs_ioc_func_t		*zvec_func;
84 	zfs_secpolicy_func_t	*zvec_secpolicy;
85 	enum {
86 		NO_NAME,
87 		POOL_NAME,
88 		DATASET_NAME
89 	} zvec_namecheck;
90 	boolean_t		zvec_his_log;
91 } zfs_ioc_vec_t;
92 
93 static void clear_props(char *dataset, nvlist_t *props);
94 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
95     boolean_t *);
96 int zfs_set_prop_nvlist(const char *, nvlist_t *);
97 
98 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
99 void
100 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
101 {
102 	const char *newfile;
103 	char buf[256];
104 	va_list adx;
105 
106 	/*
107 	 * Get rid of annoying "../common/" prefix to filename.
108 	 */
109 	newfile = strrchr(file, '/');
110 	if (newfile != NULL) {
111 		newfile = newfile + 1; /* Get rid of leading / */
112 	} else {
113 		newfile = file;
114 	}
115 
116 	va_start(adx, fmt);
117 	(void) vsnprintf(buf, sizeof (buf), fmt, adx);
118 	va_end(adx);
119 
120 	/*
121 	 * To get this data, use the zfs-dprintf probe as so:
122 	 * dtrace -q -n 'zfs-dprintf \
123 	 *	/stringof(arg0) == "dbuf.c"/ \
124 	 *	{printf("%s: %s", stringof(arg1), stringof(arg3))}'
125 	 * arg0 = file name
126 	 * arg1 = function name
127 	 * arg2 = line number
128 	 * arg3 = message
129 	 */
130 	DTRACE_PROBE4(zfs__dprintf,
131 	    char *, newfile, char *, func, int, line, char *, buf);
132 }
133 
134 static void
135 history_str_free(char *buf)
136 {
137 	kmem_free(buf, HIS_MAX_RECORD_LEN);
138 }
139 
140 static char *
141 history_str_get(zfs_cmd_t *zc)
142 {
143 	char *buf;
144 
145 	if (zc->zc_history == NULL)
146 		return (NULL);
147 
148 	buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
149 	if (copyinstr((void *)(uintptr_t)zc->zc_history,
150 	    buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
151 		history_str_free(buf);
152 		return (NULL);
153 	}
154 
155 	buf[HIS_MAX_RECORD_LEN -1] = '\0';
156 
157 	return (buf);
158 }
159 
160 /*
161  * Check to see if the named dataset is currently defined as bootable
162  */
163 static boolean_t
164 zfs_is_bootfs(const char *name)
165 {
166 	spa_t *spa;
167 	boolean_t ret = B_FALSE;
168 
169 	if (spa_open(name, &spa, FTAG) == 0) {
170 		if (spa->spa_bootfs) {
171 			objset_t *os;
172 
173 			if (dmu_objset_open(name, DMU_OST_ZFS,
174 			    DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
175 				ret = (dmu_objset_id(os) == spa->spa_bootfs);
176 				dmu_objset_close(os);
177 			}
178 		}
179 		spa_close(spa, FTAG);
180 	}
181 	return (ret);
182 }
183 
184 /*
185  * zfs_earlier_version
186  *
187  *	Return non-zero if the spa version is less than requested version.
188  */
189 static int
190 zfs_earlier_version(const char *name, int version)
191 {
192 	spa_t *spa;
193 
194 	if (spa_open(name, &spa, FTAG) == 0) {
195 		if (spa_version(spa) < version) {
196 			spa_close(spa, FTAG);
197 			return (1);
198 		}
199 		spa_close(spa, FTAG);
200 	}
201 	return (0);
202 }
203 
204 /*
205  * zpl_earlier_version
206  *
207  * Return TRUE if the ZPL version is less than requested version.
208  */
209 static boolean_t
210 zpl_earlier_version(const char *name, int version)
211 {
212 	objset_t *os;
213 	boolean_t rc = B_TRUE;
214 
215 	if (dmu_objset_open(name, DMU_OST_ANY,
216 	    DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
217 		uint64_t zplversion;
218 
219 		if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
220 			rc = zplversion < version;
221 		dmu_objset_close(os);
222 	}
223 	return (rc);
224 }
225 
226 static void
227 zfs_log_history(zfs_cmd_t *zc)
228 {
229 	spa_t *spa;
230 	char *buf;
231 
232 	if ((buf = history_str_get(zc)) == NULL)
233 		return;
234 
235 	if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
236 		if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
237 			(void) spa_history_log(spa, buf, LOG_CMD_NORMAL);
238 		spa_close(spa, FTAG);
239 	}
240 	history_str_free(buf);
241 }
242 
243 /*
244  * Policy for top-level read operations (list pools).  Requires no privileges,
245  * and can be used in the local zone, as there is no associated dataset.
246  */
247 /* ARGSUSED */
248 static int
249 zfs_secpolicy_none(zfs_cmd_t *zc, cred_t *cr)
250 {
251 	return (0);
252 }
253 
254 /*
255  * Policy for dataset read operations (list children, get statistics).  Requires
256  * no privileges, but must be visible in the local zone.
257  */
258 /* ARGSUSED */
259 static int
260 zfs_secpolicy_read(zfs_cmd_t *zc, cred_t *cr)
261 {
262 	if (INGLOBALZONE(curproc) ||
263 	    zone_dataset_visible(zc->zc_name, NULL))
264 		return (0);
265 
266 	return (ENOENT);
267 }
268 
269 static int
270 zfs_dozonecheck(const char *dataset, cred_t *cr)
271 {
272 	uint64_t zoned;
273 	int writable = 1;
274 
275 	/*
276 	 * The dataset must be visible by this zone -- check this first
277 	 * so they don't see EPERM on something they shouldn't know about.
278 	 */
279 	if (!INGLOBALZONE(curproc) &&
280 	    !zone_dataset_visible(dataset, &writable))
281 		return (ENOENT);
282 
283 	if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
284 		return (ENOENT);
285 
286 	if (INGLOBALZONE(curproc)) {
287 		/*
288 		 * If the fs is zoned, only root can access it from the
289 		 * global zone.
290 		 */
291 		if (secpolicy_zfs(cr) && zoned)
292 			return (EPERM);
293 	} else {
294 		/*
295 		 * If we are in a local zone, the 'zoned' property must be set.
296 		 */
297 		if (!zoned)
298 			return (EPERM);
299 
300 		/* must be writable by this zone */
301 		if (!writable)
302 			return (EPERM);
303 	}
304 	return (0);
305 }
306 
307 int
308 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
309 {
310 	int error;
311 
312 	error = zfs_dozonecheck(name, cr);
313 	if (error == 0) {
314 		error = secpolicy_zfs(cr);
315 		if (error)
316 			error = dsl_deleg_access(name, perm, cr);
317 	}
318 	return (error);
319 }
320 
321 static int
322 zfs_secpolicy_setprop(const char *name, zfs_prop_t prop, cred_t *cr)
323 {
324 	/*
325 	 * Check permissions for special properties.
326 	 */
327 	switch (prop) {
328 	case ZFS_PROP_ZONED:
329 		/*
330 		 * Disallow setting of 'zoned' from within a local zone.
331 		 */
332 		if (!INGLOBALZONE(curproc))
333 			return (EPERM);
334 		break;
335 
336 	case ZFS_PROP_QUOTA:
337 		if (!INGLOBALZONE(curproc)) {
338 			uint64_t zoned;
339 			char setpoint[MAXNAMELEN];
340 			/*
341 			 * Unprivileged users are allowed to modify the
342 			 * quota on things *under* (ie. contained by)
343 			 * the thing they own.
344 			 */
345 			if (dsl_prop_get_integer(name, "zoned", &zoned,
346 			    setpoint))
347 				return (EPERM);
348 			if (!zoned || strlen(name) <= strlen(setpoint))
349 				return (EPERM);
350 		}
351 		break;
352 	}
353 
354 	return (zfs_secpolicy_write_perms(name, zfs_prop_to_name(prop), cr));
355 }
356 
357 int
358 zfs_secpolicy_fsacl(zfs_cmd_t *zc, cred_t *cr)
359 {
360 	int error;
361 
362 	error = zfs_dozonecheck(zc->zc_name, cr);
363 	if (error)
364 		return (error);
365 
366 	/*
367 	 * permission to set permissions will be evaluated later in
368 	 * dsl_deleg_can_allow()
369 	 */
370 	return (0);
371 }
372 
373 int
374 zfs_secpolicy_rollback(zfs_cmd_t *zc, cred_t *cr)
375 {
376 	int error;
377 	error = zfs_secpolicy_write_perms(zc->zc_name,
378 	    ZFS_DELEG_PERM_ROLLBACK, cr);
379 	if (error == 0)
380 		error = zfs_secpolicy_write_perms(zc->zc_name,
381 		    ZFS_DELEG_PERM_MOUNT, cr);
382 	return (error);
383 }
384 
385 int
386 zfs_secpolicy_send(zfs_cmd_t *zc, cred_t *cr)
387 {
388 	return (zfs_secpolicy_write_perms(zc->zc_name,
389 	    ZFS_DELEG_PERM_SEND, cr));
390 }
391 
392 int
393 zfs_secpolicy_share(zfs_cmd_t *zc, cred_t *cr)
394 {
395 	if (!INGLOBALZONE(curproc))
396 		return (EPERM);
397 
398 	if (secpolicy_nfs(cr) == 0) {
399 		return (0);
400 	} else {
401 		vnode_t *vp;
402 		int error;
403 
404 		if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
405 		    NO_FOLLOW, NULL, &vp)) != 0)
406 			return (error);
407 
408 		/* Now make sure mntpnt and dataset are ZFS */
409 
410 		if (vp->v_vfsp->vfs_fstype != zfsfstype ||
411 		    (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
412 		    zc->zc_name) != 0)) {
413 			VN_RELE(vp);
414 			return (EPERM);
415 		}
416 
417 		VN_RELE(vp);
418 		return (dsl_deleg_access(zc->zc_name,
419 		    ZFS_DELEG_PERM_SHARE, cr));
420 	}
421 }
422 
423 static int
424 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
425 {
426 	char *cp;
427 
428 	/*
429 	 * Remove the @bla or /bla from the end of the name to get the parent.
430 	 */
431 	(void) strncpy(parent, datasetname, parentsize);
432 	cp = strrchr(parent, '@');
433 	if (cp != NULL) {
434 		cp[0] = '\0';
435 	} else {
436 		cp = strrchr(parent, '/');
437 		if (cp == NULL)
438 			return (ENOENT);
439 		cp[0] = '\0';
440 	}
441 
442 	return (0);
443 }
444 
445 int
446 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
447 {
448 	int error;
449 
450 	if ((error = zfs_secpolicy_write_perms(name,
451 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
452 		return (error);
453 
454 	return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
455 }
456 
457 static int
458 zfs_secpolicy_destroy(zfs_cmd_t *zc, cred_t *cr)
459 {
460 	return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
461 }
462 
463 /*
464  * Must have sys_config privilege to check the iscsi permission
465  */
466 /* ARGSUSED */
467 static int
468 zfs_secpolicy_iscsi(zfs_cmd_t *zc, cred_t *cr)
469 {
470 	return (secpolicy_zfs(cr));
471 }
472 
473 int
474 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
475 {
476 	char 	parentname[MAXNAMELEN];
477 	int	error;
478 
479 	if ((error = zfs_secpolicy_write_perms(from,
480 	    ZFS_DELEG_PERM_RENAME, cr)) != 0)
481 		return (error);
482 
483 	if ((error = zfs_secpolicy_write_perms(from,
484 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
485 		return (error);
486 
487 	if ((error = zfs_get_parent(to, parentname,
488 	    sizeof (parentname))) != 0)
489 		return (error);
490 
491 	if ((error = zfs_secpolicy_write_perms(parentname,
492 	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
493 		return (error);
494 
495 	if ((error = zfs_secpolicy_write_perms(parentname,
496 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
497 		return (error);
498 
499 	return (error);
500 }
501 
502 static int
503 zfs_secpolicy_rename(zfs_cmd_t *zc, cred_t *cr)
504 {
505 	return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
506 }
507 
508 static int
509 zfs_secpolicy_promote(zfs_cmd_t *zc, cred_t *cr)
510 {
511 	char 	parentname[MAXNAMELEN];
512 	objset_t *clone;
513 	int error;
514 
515 	error = zfs_secpolicy_write_perms(zc->zc_name,
516 	    ZFS_DELEG_PERM_PROMOTE, cr);
517 	if (error)
518 		return (error);
519 
520 	error = dmu_objset_open(zc->zc_name, DMU_OST_ANY,
521 	    DS_MODE_USER | DS_MODE_READONLY, &clone);
522 
523 	if (error == 0) {
524 		dsl_dataset_t *pclone = NULL;
525 		dsl_dir_t *dd;
526 		dd = clone->os->os_dsl_dataset->ds_dir;
527 
528 		rw_enter(&dd->dd_pool->dp_config_rwlock, RW_READER);
529 		error = dsl_dataset_hold_obj(dd->dd_pool,
530 		    dd->dd_phys->dd_origin_obj, FTAG, &pclone);
531 		rw_exit(&dd->dd_pool->dp_config_rwlock);
532 		if (error) {
533 			dmu_objset_close(clone);
534 			return (error);
535 		}
536 
537 		error = zfs_secpolicy_write_perms(zc->zc_name,
538 		    ZFS_DELEG_PERM_MOUNT, cr);
539 
540 		dsl_dataset_name(pclone, parentname);
541 		dmu_objset_close(clone);
542 		dsl_dataset_rele(pclone, FTAG);
543 		if (error == 0)
544 			error = zfs_secpolicy_write_perms(parentname,
545 			    ZFS_DELEG_PERM_PROMOTE, cr);
546 	}
547 	return (error);
548 }
549 
550 static int
551 zfs_secpolicy_receive(zfs_cmd_t *zc, cred_t *cr)
552 {
553 	int error;
554 
555 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
556 	    ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
557 		return (error);
558 
559 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
560 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
561 		return (error);
562 
563 	return (zfs_secpolicy_write_perms(zc->zc_name,
564 	    ZFS_DELEG_PERM_CREATE, cr));
565 }
566 
567 int
568 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
569 {
570 	int error;
571 
572 	if ((error = zfs_secpolicy_write_perms(name,
573 	    ZFS_DELEG_PERM_SNAPSHOT, cr)) != 0)
574 		return (error);
575 
576 	error = zfs_secpolicy_write_perms(name,
577 	    ZFS_DELEG_PERM_MOUNT, cr);
578 
579 	return (error);
580 }
581 
582 static int
583 zfs_secpolicy_snapshot(zfs_cmd_t *zc, cred_t *cr)
584 {
585 
586 	return (zfs_secpolicy_snapshot_perms(zc->zc_name, cr));
587 }
588 
589 static int
590 zfs_secpolicy_create(zfs_cmd_t *zc, cred_t *cr)
591 {
592 	char 	parentname[MAXNAMELEN];
593 	int 	error;
594 
595 	if ((error = zfs_get_parent(zc->zc_name, parentname,
596 	    sizeof (parentname))) != 0)
597 		return (error);
598 
599 	if (zc->zc_value[0] != '\0') {
600 		if ((error = zfs_secpolicy_write_perms(zc->zc_value,
601 		    ZFS_DELEG_PERM_CLONE, cr)) != 0)
602 			return (error);
603 	}
604 
605 	if ((error = zfs_secpolicy_write_perms(parentname,
606 	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
607 		return (error);
608 
609 	error = zfs_secpolicy_write_perms(parentname,
610 	    ZFS_DELEG_PERM_MOUNT, cr);
611 
612 	return (error);
613 }
614 
615 static int
616 zfs_secpolicy_umount(zfs_cmd_t *zc, cred_t *cr)
617 {
618 	int error;
619 
620 	error = secpolicy_fs_unmount(cr, NULL);
621 	if (error) {
622 		error = dsl_deleg_access(zc->zc_name, ZFS_DELEG_PERM_MOUNT, cr);
623 	}
624 	return (error);
625 }
626 
627 /*
628  * Policy for pool operations - create/destroy pools, add vdevs, etc.  Requires
629  * SYS_CONFIG privilege, which is not available in a local zone.
630  */
631 /* ARGSUSED */
632 static int
633 zfs_secpolicy_config(zfs_cmd_t *zc, cred_t *cr)
634 {
635 	if (secpolicy_sys_config(cr, B_FALSE) != 0)
636 		return (EPERM);
637 
638 	return (0);
639 }
640 
641 /*
642  * Just like zfs_secpolicy_config, except that we will check for
643  * mount permission on the dataset for permission to create/remove
644  * the minor nodes.
645  */
646 static int
647 zfs_secpolicy_minor(zfs_cmd_t *zc, cred_t *cr)
648 {
649 	if (secpolicy_sys_config(cr, B_FALSE) != 0) {
650 		return (dsl_deleg_access(zc->zc_name,
651 		    ZFS_DELEG_PERM_MOUNT, cr));
652 	}
653 
654 	return (0);
655 }
656 
657 /*
658  * Policy for fault injection.  Requires all privileges.
659  */
660 /* ARGSUSED */
661 static int
662 zfs_secpolicy_inject(zfs_cmd_t *zc, cred_t *cr)
663 {
664 	return (secpolicy_zinject(cr));
665 }
666 
667 static int
668 zfs_secpolicy_inherit(zfs_cmd_t *zc, cred_t *cr)
669 {
670 	zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
671 
672 	if (prop == ZPROP_INVAL) {
673 		if (!zfs_prop_user(zc->zc_value))
674 			return (EINVAL);
675 		return (zfs_secpolicy_write_perms(zc->zc_name,
676 		    ZFS_DELEG_PERM_USERPROP, cr));
677 	} else {
678 		if (!zfs_prop_inheritable(prop))
679 			return (EINVAL);
680 		return (zfs_secpolicy_setprop(zc->zc_name, prop, cr));
681 	}
682 }
683 
684 /*
685  * Returns the nvlist as specified by the user in the zfs_cmd_t.
686  */
687 static int
688 get_nvlist(uint64_t nvl, uint64_t size, nvlist_t **nvp)
689 {
690 	char *packed;
691 	int error;
692 	nvlist_t *list = NULL;
693 
694 	/*
695 	 * Read in and unpack the user-supplied nvlist.
696 	 */
697 	if (size == 0)
698 		return (EINVAL);
699 
700 	packed = kmem_alloc(size, KM_SLEEP);
701 
702 	if ((error = xcopyin((void *)(uintptr_t)nvl, packed, size)) != 0) {
703 		kmem_free(packed, size);
704 		return (error);
705 	}
706 
707 	if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
708 		kmem_free(packed, size);
709 		return (error);
710 	}
711 
712 	kmem_free(packed, size);
713 
714 	*nvp = list;
715 	return (0);
716 }
717 
718 static int
719 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
720 {
721 	char *packed = NULL;
722 	size_t size;
723 	int error;
724 
725 	VERIFY(nvlist_size(nvl, &size, NV_ENCODE_NATIVE) == 0);
726 
727 	if (size > zc->zc_nvlist_dst_size) {
728 		error = ENOMEM;
729 	} else {
730 		packed = kmem_alloc(size, KM_SLEEP);
731 		VERIFY(nvlist_pack(nvl, &packed, &size, NV_ENCODE_NATIVE,
732 		    KM_SLEEP) == 0);
733 		error = xcopyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
734 		    size);
735 		kmem_free(packed, size);
736 	}
737 
738 	zc->zc_nvlist_dst_size = size;
739 	return (error);
740 }
741 
742 static int
743 zfs_ioc_pool_create(zfs_cmd_t *zc)
744 {
745 	int error;
746 	nvlist_t *config, *props = NULL;
747 	nvlist_t *rootprops = NULL;
748 	nvlist_t *zplprops = NULL;
749 	char *buf;
750 
751 	if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
752 	    &config))
753 		return (error);
754 
755 	if (zc->zc_nvlist_src_size != 0 && (error =
756 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, &props))) {
757 		nvlist_free(config);
758 		return (error);
759 	}
760 
761 	if (props) {
762 		nvlist_t *nvl = NULL;
763 		uint64_t version = SPA_VERSION;
764 
765 		(void) nvlist_lookup_uint64(props,
766 		    zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
767 		if (version < SPA_VERSION_INITIAL || version > SPA_VERSION) {
768 			error = EINVAL;
769 			goto pool_props_bad;
770 		}
771 		(void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
772 		if (nvl) {
773 			error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
774 			if (error != 0) {
775 				nvlist_free(config);
776 				nvlist_free(props);
777 				return (error);
778 			}
779 			(void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
780 		}
781 		VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
782 		error = zfs_fill_zplprops_root(version, rootprops,
783 		    zplprops, NULL);
784 		if (error)
785 			goto pool_props_bad;
786 	}
787 
788 	buf = history_str_get(zc);
789 
790 	error = spa_create(zc->zc_name, config, props, buf, zplprops);
791 
792 	/*
793 	 * Set the remaining root properties
794 	 */
795 	if (!error &&
796 	    (error = zfs_set_prop_nvlist(zc->zc_name, rootprops)) != 0)
797 		(void) spa_destroy(zc->zc_name);
798 
799 	if (buf != NULL)
800 		history_str_free(buf);
801 
802 pool_props_bad:
803 	nvlist_free(rootprops);
804 	nvlist_free(zplprops);
805 	nvlist_free(config);
806 	nvlist_free(props);
807 
808 	return (error);
809 }
810 
811 static int
812 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
813 {
814 	int error;
815 	zfs_log_history(zc);
816 	error = spa_destroy(zc->zc_name);
817 	return (error);
818 }
819 
820 static int
821 zfs_ioc_pool_import(zfs_cmd_t *zc)
822 {
823 	int error;
824 	nvlist_t *config, *props = NULL;
825 	uint64_t guid;
826 
827 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
828 	    &config)) != 0)
829 		return (error);
830 
831 	if (zc->zc_nvlist_src_size != 0 && (error =
832 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, &props))) {
833 		nvlist_free(config);
834 		return (error);
835 	}
836 
837 	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
838 	    guid != zc->zc_guid)
839 		error = EINVAL;
840 	else if (zc->zc_cookie)
841 		error = spa_import_faulted(zc->zc_name, config,
842 		    props);
843 	else
844 		error = spa_import(zc->zc_name, config, props);
845 
846 	nvlist_free(config);
847 
848 	if (props)
849 		nvlist_free(props);
850 
851 	return (error);
852 }
853 
854 static int
855 zfs_ioc_pool_export(zfs_cmd_t *zc)
856 {
857 	int error;
858 	boolean_t force = (boolean_t)zc->zc_cookie;
859 	boolean_t hardforce = (boolean_t)zc->zc_guid;
860 
861 	zfs_log_history(zc);
862 	error = spa_export(zc->zc_name, NULL, force, hardforce);
863 	return (error);
864 }
865 
866 static int
867 zfs_ioc_pool_configs(zfs_cmd_t *zc)
868 {
869 	nvlist_t *configs;
870 	int error;
871 
872 	if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
873 		return (EEXIST);
874 
875 	error = put_nvlist(zc, configs);
876 
877 	nvlist_free(configs);
878 
879 	return (error);
880 }
881 
882 static int
883 zfs_ioc_pool_stats(zfs_cmd_t *zc)
884 {
885 	nvlist_t *config;
886 	int error;
887 	int ret = 0;
888 
889 	error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
890 	    sizeof (zc->zc_value));
891 
892 	if (config != NULL) {
893 		ret = put_nvlist(zc, config);
894 		nvlist_free(config);
895 
896 		/*
897 		 * The config may be present even if 'error' is non-zero.
898 		 * In this case we return success, and preserve the real errno
899 		 * in 'zc_cookie'.
900 		 */
901 		zc->zc_cookie = error;
902 	} else {
903 		ret = error;
904 	}
905 
906 	return (ret);
907 }
908 
909 /*
910  * Try to import the given pool, returning pool stats as appropriate so that
911  * user land knows which devices are available and overall pool health.
912  */
913 static int
914 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
915 {
916 	nvlist_t *tryconfig, *config;
917 	int error;
918 
919 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
920 	    &tryconfig)) != 0)
921 		return (error);
922 
923 	config = spa_tryimport(tryconfig);
924 
925 	nvlist_free(tryconfig);
926 
927 	if (config == NULL)
928 		return (EINVAL);
929 
930 	error = put_nvlist(zc, config);
931 	nvlist_free(config);
932 
933 	return (error);
934 }
935 
936 static int
937 zfs_ioc_pool_scrub(zfs_cmd_t *zc)
938 {
939 	spa_t *spa;
940 	int error;
941 
942 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
943 		return (error);
944 
945 	error = spa_scrub(spa, zc->zc_cookie);
946 
947 	spa_close(spa, FTAG);
948 
949 	return (error);
950 }
951 
952 static int
953 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
954 {
955 	spa_t *spa;
956 	int error;
957 
958 	error = spa_open(zc->zc_name, &spa, FTAG);
959 	if (error == 0) {
960 		spa_freeze(spa);
961 		spa_close(spa, FTAG);
962 	}
963 	return (error);
964 }
965 
966 static int
967 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
968 {
969 	spa_t *spa;
970 	int error;
971 
972 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
973 		return (error);
974 
975 	if (zc->zc_cookie < spa_version(spa) || zc->zc_cookie > SPA_VERSION) {
976 		spa_close(spa, FTAG);
977 		return (EINVAL);
978 	}
979 
980 	spa_upgrade(spa, zc->zc_cookie);
981 	spa_close(spa, FTAG);
982 
983 	return (error);
984 }
985 
986 static int
987 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
988 {
989 	spa_t *spa;
990 	char *hist_buf;
991 	uint64_t size;
992 	int error;
993 
994 	if ((size = zc->zc_history_len) == 0)
995 		return (EINVAL);
996 
997 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
998 		return (error);
999 
1000 	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1001 		spa_close(spa, FTAG);
1002 		return (ENOTSUP);
1003 	}
1004 
1005 	hist_buf = kmem_alloc(size, KM_SLEEP);
1006 	if ((error = spa_history_get(spa, &zc->zc_history_offset,
1007 	    &zc->zc_history_len, hist_buf)) == 0) {
1008 		error = xcopyout(hist_buf,
1009 		    (char *)(uintptr_t)zc->zc_history,
1010 		    zc->zc_history_len);
1011 	}
1012 
1013 	spa_close(spa, FTAG);
1014 	kmem_free(hist_buf, size);
1015 	return (error);
1016 }
1017 
1018 static int
1019 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1020 {
1021 	int error;
1022 
1023 	if (error = dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value))
1024 		return (error);
1025 
1026 	return (0);
1027 }
1028 
1029 static int
1030 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1031 {
1032 	objset_t *osp;
1033 	int error;
1034 
1035 	if ((error = dmu_objset_open(zc->zc_name, DMU_OST_ZFS,
1036 	    DS_MODE_USER | DS_MODE_READONLY, &osp)) != 0)
1037 		return (error);
1038 	error = zfs_obj_to_path(osp, zc->zc_obj, zc->zc_value,
1039 	    sizeof (zc->zc_value));
1040 	dmu_objset_close(osp);
1041 
1042 	return (error);
1043 }
1044 
1045 static int
1046 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1047 {
1048 	spa_t *spa;
1049 	int error;
1050 	nvlist_t *config, **l2cache, **spares;
1051 	uint_t nl2cache = 0, nspares = 0;
1052 
1053 	error = spa_open(zc->zc_name, &spa, FTAG);
1054 	if (error != 0)
1055 		return (error);
1056 
1057 	error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1058 	    &config);
1059 	(void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1060 	    &l2cache, &nl2cache);
1061 
1062 	(void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1063 	    &spares, &nspares);
1064 
1065 	/*
1066 	 * A root pool with concatenated devices is not supported.
1067 	 * Thus, can not add a device to a root pool.
1068 	 *
1069 	 * Intent log device can not be added to a rootpool because
1070 	 * during mountroot, zil is replayed, a seperated log device
1071 	 * can not be accessed during the mountroot time.
1072 	 *
1073 	 * l2cache and spare devices are ok to be added to a rootpool.
1074 	 */
1075 	if (spa->spa_bootfs != 0 && nl2cache == 0 && nspares == 0) {
1076 		spa_close(spa, FTAG);
1077 		return (EDOM);
1078 	}
1079 
1080 	if (error == 0) {
1081 		error = spa_vdev_add(spa, config);
1082 		nvlist_free(config);
1083 	}
1084 	spa_close(spa, FTAG);
1085 	return (error);
1086 }
1087 
1088 static int
1089 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1090 {
1091 	spa_t *spa;
1092 	int error;
1093 
1094 	error = spa_open(zc->zc_name, &spa, FTAG);
1095 	if (error != 0)
1096 		return (error);
1097 	error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1098 	spa_close(spa, FTAG);
1099 	return (error);
1100 }
1101 
1102 static int
1103 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1104 {
1105 	spa_t *spa;
1106 	int error;
1107 	vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1108 
1109 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1110 		return (error);
1111 	switch (zc->zc_cookie) {
1112 	case VDEV_STATE_ONLINE:
1113 		error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1114 		break;
1115 
1116 	case VDEV_STATE_OFFLINE:
1117 		error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1118 		break;
1119 
1120 	case VDEV_STATE_FAULTED:
1121 		error = vdev_fault(spa, zc->zc_guid);
1122 		break;
1123 
1124 	case VDEV_STATE_DEGRADED:
1125 		error = vdev_degrade(spa, zc->zc_guid);
1126 		break;
1127 
1128 	default:
1129 		error = EINVAL;
1130 	}
1131 	zc->zc_cookie = newstate;
1132 	spa_close(spa, FTAG);
1133 	return (error);
1134 }
1135 
1136 static int
1137 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1138 {
1139 	spa_t *spa;
1140 	int replacing = zc->zc_cookie;
1141 	nvlist_t *config;
1142 	int error;
1143 
1144 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1145 		return (error);
1146 
1147 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1148 	    &config)) == 0) {
1149 		error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1150 		nvlist_free(config);
1151 	}
1152 
1153 	spa_close(spa, FTAG);
1154 	return (error);
1155 }
1156 
1157 static int
1158 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1159 {
1160 	spa_t *spa;
1161 	int error;
1162 
1163 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1164 		return (error);
1165 
1166 	error = spa_vdev_detach(spa, zc->zc_guid, B_FALSE);
1167 
1168 	spa_close(spa, FTAG);
1169 	return (error);
1170 }
1171 
1172 static int
1173 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
1174 {
1175 	spa_t *spa;
1176 	char *path = zc->zc_value;
1177 	uint64_t guid = zc->zc_guid;
1178 	int error;
1179 
1180 	error = spa_open(zc->zc_name, &spa, FTAG);
1181 	if (error != 0)
1182 		return (error);
1183 
1184 	error = spa_vdev_setpath(spa, guid, path);
1185 	spa_close(spa, FTAG);
1186 	return (error);
1187 }
1188 
1189 /*
1190  * inputs:
1191  * zc_name		name of filesystem
1192  * zc_nvlist_dst_size	size of buffer for property nvlist
1193  *
1194  * outputs:
1195  * zc_objset_stats	stats
1196  * zc_nvlist_dst	property nvlist
1197  * zc_nvlist_dst_size	size of property nvlist
1198  */
1199 static int
1200 zfs_ioc_objset_stats(zfs_cmd_t *zc)
1201 {
1202 	objset_t *os = NULL;
1203 	int error;
1204 	nvlist_t *nv;
1205 
1206 	if (error = dmu_objset_open(zc->zc_name,
1207 	    DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os))
1208 		return (error);
1209 
1210 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1211 
1212 	if (zc->zc_nvlist_dst != 0 &&
1213 	    (error = dsl_prop_get_all(os, &nv, FALSE)) == 0) {
1214 		dmu_objset_stats(os, nv);
1215 		/*
1216 		 * NB: zvol_get_stats() will read the objset contents,
1217 		 * which we aren't supposed to do with a
1218 		 * DS_MODE_USER hold, because it could be
1219 		 * inconsistent.  So this is a bit of a workaround...
1220 		 */
1221 		if (!zc->zc_objset_stats.dds_inconsistent) {
1222 			if (dmu_objset_type(os) == DMU_OST_ZVOL)
1223 				VERIFY(zvol_get_stats(os, nv) == 0);
1224 		}
1225 		error = put_nvlist(zc, nv);
1226 		nvlist_free(nv);
1227 	}
1228 
1229 	dmu_objset_close(os);
1230 	return (error);
1231 }
1232 
1233 static int
1234 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
1235 {
1236 	uint64_t value;
1237 	int error;
1238 
1239 	/*
1240 	 * zfs_get_zplprop() will either find a value or give us
1241 	 * the default value (if there is one).
1242 	 */
1243 	if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
1244 		return (error);
1245 	VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
1246 	return (0);
1247 }
1248 
1249 /*
1250  * inputs:
1251  * zc_name		name of filesystem
1252  * zc_nvlist_dst_size	size of buffer for zpl property nvlist
1253  *
1254  * outputs:
1255  * zc_nvlist_dst	zpl property nvlist
1256  * zc_nvlist_dst_size	size of zpl property nvlist
1257  */
1258 static int
1259 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
1260 {
1261 	objset_t *os;
1262 	int err;
1263 
1264 	if (err = dmu_objset_open(zc->zc_name,
1265 	    DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os))
1266 		return (err);
1267 
1268 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1269 
1270 	/*
1271 	 * NB: nvl_add_zplprop() will read the objset contents,
1272 	 * which we aren't supposed to do with a DS_MODE_USER
1273 	 * hold, because it could be inconsistent.
1274 	 */
1275 	if (zc->zc_nvlist_dst != NULL &&
1276 	    !zc->zc_objset_stats.dds_inconsistent &&
1277 	    dmu_objset_type(os) == DMU_OST_ZFS) {
1278 		nvlist_t *nv;
1279 
1280 		VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1281 		if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
1282 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
1283 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
1284 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
1285 			err = put_nvlist(zc, nv);
1286 		nvlist_free(nv);
1287 	} else {
1288 		err = ENOENT;
1289 	}
1290 	dmu_objset_close(os);
1291 	return (err);
1292 }
1293 
1294 /*
1295  * inputs:
1296  * zc_name		name of filesystem
1297  * zc_cookie		zap cursor
1298  * zc_nvlist_dst_size	size of buffer for property nvlist
1299  *
1300  * outputs:
1301  * zc_name		name of next filesystem
1302  * zc_objset_stats	stats
1303  * zc_nvlist_dst	property nvlist
1304  * zc_nvlist_dst_size	size of property nvlist
1305  */
1306 static int
1307 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
1308 {
1309 	objset_t *os;
1310 	int error;
1311 	char *p;
1312 
1313 	if (error = dmu_objset_open(zc->zc_name,
1314 	    DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os)) {
1315 		if (error == ENOENT)
1316 			error = ESRCH;
1317 		return (error);
1318 	}
1319 
1320 	p = strrchr(zc->zc_name, '/');
1321 	if (p == NULL || p[1] != '\0')
1322 		(void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
1323 	p = zc->zc_name + strlen(zc->zc_name);
1324 
1325 	do {
1326 		error = dmu_dir_list_next(os,
1327 		    sizeof (zc->zc_name) - (p - zc->zc_name), p,
1328 		    NULL, &zc->zc_cookie);
1329 		if (error == ENOENT)
1330 			error = ESRCH;
1331 	} while (error == 0 && !INGLOBALZONE(curproc) &&
1332 	    !zone_dataset_visible(zc->zc_name, NULL));
1333 	dmu_objset_close(os);
1334 
1335 	/*
1336 	 * If it's a hidden dataset (ie. with a '$' in its name), don't
1337 	 * try to get stats for it.  Userland will skip over it.
1338 	 */
1339 	if (error == 0 && strchr(zc->zc_name, '$') == NULL)
1340 		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
1341 
1342 	return (error);
1343 }
1344 
1345 /*
1346  * inputs:
1347  * zc_name		name of filesystem
1348  * zc_cookie		zap cursor
1349  * zc_nvlist_dst_size	size of buffer for property nvlist
1350  *
1351  * outputs:
1352  * zc_name		name of next snapshot
1353  * zc_objset_stats	stats
1354  * zc_nvlist_dst	property nvlist
1355  * zc_nvlist_dst_size	size of property nvlist
1356  */
1357 static int
1358 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
1359 {
1360 	objset_t *os;
1361 	int error;
1362 
1363 	error = dmu_objset_open(zc->zc_name,
1364 	    DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os);
1365 	if (error)
1366 		return (error == ENOENT ? ESRCH : error);
1367 
1368 	/*
1369 	 * A dataset name of maximum length cannot have any snapshots,
1370 	 * so exit immediately.
1371 	 */
1372 	if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
1373 		dmu_objset_close(os);
1374 		return (ESRCH);
1375 	}
1376 
1377 	error = dmu_snapshot_list_next(os,
1378 	    sizeof (zc->zc_name) - strlen(zc->zc_name),
1379 	    zc->zc_name + strlen(zc->zc_name), NULL, &zc->zc_cookie, NULL);
1380 	dmu_objset_close(os);
1381 	if (error == 0)
1382 		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
1383 	else if (error == ENOENT)
1384 		error = ESRCH;
1385 
1386 	/* if we failed, undo the @ that we tacked on to zc_name */
1387 	if (error)
1388 		*strchr(zc->zc_name, '@') = '\0';
1389 	return (error);
1390 }
1391 
1392 int
1393 zfs_set_prop_nvlist(const char *name, nvlist_t *nvl)
1394 {
1395 	nvpair_t *elem;
1396 	int error;
1397 	uint64_t intval;
1398 	char *strval;
1399 
1400 	/*
1401 	 * First validate permission to set all of the properties
1402 	 */
1403 	elem = NULL;
1404 	while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) {
1405 		const char *propname = nvpair_name(elem);
1406 		zfs_prop_t prop = zfs_name_to_prop(propname);
1407 
1408 		if (prop == ZPROP_INVAL) {
1409 			/*
1410 			 * If this is a user-defined property, it must be a
1411 			 * string, and there is no further validation to do.
1412 			 */
1413 			if (!zfs_prop_user(propname) ||
1414 			    nvpair_type(elem) != DATA_TYPE_STRING)
1415 				return (EINVAL);
1416 
1417 			if (error = zfs_secpolicy_write_perms(name,
1418 			    ZFS_DELEG_PERM_USERPROP, CRED()))
1419 				return (error);
1420 			continue;
1421 		}
1422 
1423 		if ((error = zfs_secpolicy_setprop(name, prop, CRED())) != 0)
1424 			return (error);
1425 
1426 		/*
1427 		 * Check that this value is valid for this pool version
1428 		 */
1429 		switch (prop) {
1430 		case ZFS_PROP_COMPRESSION:
1431 			/*
1432 			 * If the user specified gzip compression, make sure
1433 			 * the SPA supports it. We ignore any errors here since
1434 			 * we'll catch them later.
1435 			 */
1436 			if (nvpair_type(elem) == DATA_TYPE_UINT64 &&
1437 			    nvpair_value_uint64(elem, &intval) == 0) {
1438 				if (intval >= ZIO_COMPRESS_GZIP_1 &&
1439 				    intval <= ZIO_COMPRESS_GZIP_9 &&
1440 				    zfs_earlier_version(name,
1441 				    SPA_VERSION_GZIP_COMPRESSION))
1442 					return (ENOTSUP);
1443 
1444 				/*
1445 				 * If this is a bootable dataset then
1446 				 * verify that the compression algorithm
1447 				 * is supported for booting. We must return
1448 				 * something other than ENOTSUP since it
1449 				 * implies a downrev pool version.
1450 				 */
1451 				if (zfs_is_bootfs(name) &&
1452 				    !BOOTFS_COMPRESS_VALID(intval))
1453 					return (ERANGE);
1454 			}
1455 			break;
1456 
1457 		case ZFS_PROP_COPIES:
1458 			if (zfs_earlier_version(name,
1459 			    SPA_VERSION_DITTO_BLOCKS))
1460 				return (ENOTSUP);
1461 			break;
1462 
1463 		case ZFS_PROP_SHARESMB:
1464 			if (zpl_earlier_version(name, ZPL_VERSION_FUID))
1465 				return (ENOTSUP);
1466 			break;
1467 
1468 		case ZFS_PROP_ACLINHERIT:
1469 			if (nvpair_type(elem) == DATA_TYPE_UINT64 &&
1470 			    nvpair_value_uint64(elem, &intval) == 0)
1471 				if (intval == ZFS_ACL_PASSTHROUGH_X &&
1472 				    zfs_earlier_version(name,
1473 				    SPA_VERSION_PASSTHROUGH_X))
1474 					return (ENOTSUP);
1475 		}
1476 	}
1477 
1478 	elem = NULL;
1479 	while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) {
1480 		const char *propname = nvpair_name(elem);
1481 		zfs_prop_t prop = zfs_name_to_prop(propname);
1482 
1483 		if (prop == ZPROP_INVAL) {
1484 			VERIFY(nvpair_value_string(elem, &strval) == 0);
1485 			error = dsl_prop_set(name, propname, 1,
1486 			    strlen(strval) + 1, strval);
1487 			if (error == 0)
1488 				continue;
1489 			else
1490 				return (error);
1491 		}
1492 
1493 		switch (prop) {
1494 		case ZFS_PROP_QUOTA:
1495 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1496 			    (error = dsl_dir_set_quota(name, intval)) != 0)
1497 				return (error);
1498 			break;
1499 
1500 		case ZFS_PROP_REFQUOTA:
1501 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1502 			    (error = dsl_dataset_set_quota(name, intval)) != 0)
1503 				return (error);
1504 			break;
1505 
1506 		case ZFS_PROP_RESERVATION:
1507 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1508 			    (error = dsl_dir_set_reservation(name,
1509 			    intval)) != 0)
1510 				return (error);
1511 			break;
1512 
1513 		case ZFS_PROP_REFRESERVATION:
1514 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1515 			    (error = dsl_dataset_set_reservation(name,
1516 			    intval)) != 0)
1517 				return (error);
1518 			break;
1519 
1520 		case ZFS_PROP_VOLSIZE:
1521 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1522 			    (error = zvol_set_volsize(name,
1523 			    ddi_driver_major(zfs_dip), intval)) != 0)
1524 				return (error);
1525 			break;
1526 
1527 		case ZFS_PROP_VOLBLOCKSIZE:
1528 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1529 			    (error = zvol_set_volblocksize(name, intval)) != 0)
1530 				return (error);
1531 			break;
1532 
1533 		case ZFS_PROP_VERSION:
1534 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1535 			    (error = zfs_set_version(name, intval)) != 0)
1536 				return (error);
1537 			break;
1538 
1539 		default:
1540 			if (nvpair_type(elem) == DATA_TYPE_STRING) {
1541 				if (zfs_prop_get_type(prop) !=
1542 				    PROP_TYPE_STRING)
1543 					return (EINVAL);
1544 				VERIFY(nvpair_value_string(elem, &strval) == 0);
1545 				if ((error = dsl_prop_set(name,
1546 				    nvpair_name(elem), 1, strlen(strval) + 1,
1547 				    strval)) != 0)
1548 					return (error);
1549 			} else if (nvpair_type(elem) == DATA_TYPE_UINT64) {
1550 				const char *unused;
1551 
1552 				VERIFY(nvpair_value_uint64(elem, &intval) == 0);
1553 
1554 				switch (zfs_prop_get_type(prop)) {
1555 				case PROP_TYPE_NUMBER:
1556 					break;
1557 				case PROP_TYPE_STRING:
1558 					return (EINVAL);
1559 				case PROP_TYPE_INDEX:
1560 					if (zfs_prop_index_to_string(prop,
1561 					    intval, &unused) != 0)
1562 						return (EINVAL);
1563 					break;
1564 				default:
1565 					cmn_err(CE_PANIC,
1566 					    "unknown property type");
1567 					break;
1568 				}
1569 
1570 				if ((error = dsl_prop_set(name, propname,
1571 				    8, 1, &intval)) != 0)
1572 					return (error);
1573 			} else {
1574 				return (EINVAL);
1575 			}
1576 			break;
1577 		}
1578 	}
1579 
1580 	return (0);
1581 }
1582 
1583 /*
1584  * inputs:
1585  * zc_name		name of filesystem
1586  * zc_value		name of property to inherit
1587  * zc_nvlist_src{_size}	nvlist of properties to apply
1588  * zc_cookie		clear existing local props?
1589  *
1590  * outputs:		none
1591  */
1592 static int
1593 zfs_ioc_set_prop(zfs_cmd_t *zc)
1594 {
1595 	nvlist_t *nvl;
1596 	int error;
1597 
1598 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1599 	    &nvl)) != 0)
1600 		return (error);
1601 
1602 	if (zc->zc_cookie) {
1603 		nvlist_t *origprops;
1604 		objset_t *os;
1605 
1606 		if (dmu_objset_open(zc->zc_name, DMU_OST_ANY,
1607 		    DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
1608 			if (dsl_prop_get_all(os, &origprops, TRUE) == 0) {
1609 				clear_props(zc->zc_name, origprops);
1610 				nvlist_free(origprops);
1611 			}
1612 			dmu_objset_close(os);
1613 		}
1614 
1615 	}
1616 
1617 	error = zfs_set_prop_nvlist(zc->zc_name, nvl);
1618 
1619 	nvlist_free(nvl);
1620 	return (error);
1621 }
1622 
1623 /*
1624  * inputs:
1625  * zc_name		name of filesystem
1626  * zc_value		name of property to inherit
1627  *
1628  * outputs:		none
1629  */
1630 static int
1631 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
1632 {
1633 	/* the property name has been validated by zfs_secpolicy_inherit() */
1634 	return (dsl_prop_set(zc->zc_name, zc->zc_value, 0, 0, NULL));
1635 }
1636 
1637 static int
1638 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
1639 {
1640 	nvlist_t *props;
1641 	spa_t *spa;
1642 	int error;
1643 
1644 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1645 	    &props)))
1646 		return (error);
1647 
1648 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
1649 		nvlist_free(props);
1650 		return (error);
1651 	}
1652 
1653 	error = spa_prop_set(spa, props);
1654 
1655 	nvlist_free(props);
1656 	spa_close(spa, FTAG);
1657 
1658 	return (error);
1659 }
1660 
1661 static int
1662 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
1663 {
1664 	spa_t *spa;
1665 	int error;
1666 	nvlist_t *nvp = NULL;
1667 
1668 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1669 		return (error);
1670 
1671 	error = spa_prop_get(spa, &nvp);
1672 
1673 	if (error == 0 && zc->zc_nvlist_dst != NULL)
1674 		error = put_nvlist(zc, nvp);
1675 	else
1676 		error = EFAULT;
1677 
1678 	spa_close(spa, FTAG);
1679 
1680 	if (nvp)
1681 		nvlist_free(nvp);
1682 	return (error);
1683 }
1684 
1685 static int
1686 zfs_ioc_iscsi_perm_check(zfs_cmd_t *zc)
1687 {
1688 	nvlist_t *nvp;
1689 	int error;
1690 	uint32_t uid;
1691 	uint32_t gid;
1692 	uint32_t *groups;
1693 	uint_t group_cnt;
1694 	cred_t	*usercred;
1695 
1696 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1697 	    &nvp)) != 0) {
1698 		return (error);
1699 	}
1700 
1701 	if ((error = nvlist_lookup_uint32(nvp,
1702 	    ZFS_DELEG_PERM_UID, &uid)) != 0) {
1703 		nvlist_free(nvp);
1704 		return (EPERM);
1705 	}
1706 
1707 	if ((error = nvlist_lookup_uint32(nvp,
1708 	    ZFS_DELEG_PERM_GID, &gid)) != 0) {
1709 		nvlist_free(nvp);
1710 		return (EPERM);
1711 	}
1712 
1713 	if ((error = nvlist_lookup_uint32_array(nvp, ZFS_DELEG_PERM_GROUPS,
1714 	    &groups, &group_cnt)) != 0) {
1715 		nvlist_free(nvp);
1716 		return (EPERM);
1717 	}
1718 	usercred = cralloc();
1719 	if ((crsetugid(usercred, uid, gid) != 0) ||
1720 	    (crsetgroups(usercred, group_cnt, (gid_t *)groups) != 0)) {
1721 		nvlist_free(nvp);
1722 		crfree(usercred);
1723 		return (EPERM);
1724 	}
1725 	nvlist_free(nvp);
1726 	error = dsl_deleg_access(zc->zc_name,
1727 	    zfs_prop_to_name(ZFS_PROP_SHAREISCSI), usercred);
1728 	crfree(usercred);
1729 	return (error);
1730 }
1731 
1732 /*
1733  * inputs:
1734  * zc_name		name of filesystem
1735  * zc_nvlist_src{_size}	nvlist of delegated permissions
1736  * zc_perm_action	allow/unallow flag
1737  *
1738  * outputs:		none
1739  */
1740 static int
1741 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
1742 {
1743 	int error;
1744 	nvlist_t *fsaclnv = NULL;
1745 
1746 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1747 	    &fsaclnv)) != 0)
1748 		return (error);
1749 
1750 	/*
1751 	 * Verify nvlist is constructed correctly
1752 	 */
1753 	if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
1754 		nvlist_free(fsaclnv);
1755 		return (EINVAL);
1756 	}
1757 
1758 	/*
1759 	 * If we don't have PRIV_SYS_MOUNT, then validate
1760 	 * that user is allowed to hand out each permission in
1761 	 * the nvlist(s)
1762 	 */
1763 
1764 	error = secpolicy_zfs(CRED());
1765 	if (error) {
1766 		if (zc->zc_perm_action == B_FALSE) {
1767 			error = dsl_deleg_can_allow(zc->zc_name,
1768 			    fsaclnv, CRED());
1769 		} else {
1770 			error = dsl_deleg_can_unallow(zc->zc_name,
1771 			    fsaclnv, CRED());
1772 		}
1773 	}
1774 
1775 	if (error == 0)
1776 		error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
1777 
1778 	nvlist_free(fsaclnv);
1779 	return (error);
1780 }
1781 
1782 /*
1783  * inputs:
1784  * zc_name		name of filesystem
1785  *
1786  * outputs:
1787  * zc_nvlist_src{_size}	nvlist of delegated permissions
1788  */
1789 static int
1790 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
1791 {
1792 	nvlist_t *nvp;
1793 	int error;
1794 
1795 	if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
1796 		error = put_nvlist(zc, nvp);
1797 		nvlist_free(nvp);
1798 	}
1799 
1800 	return (error);
1801 }
1802 
1803 /*
1804  * inputs:
1805  * zc_name		name of volume
1806  *
1807  * outputs:		none
1808  */
1809 static int
1810 zfs_ioc_create_minor(zfs_cmd_t *zc)
1811 {
1812 	return (zvol_create_minor(zc->zc_name, ddi_driver_major(zfs_dip)));
1813 }
1814 
1815 /*
1816  * inputs:
1817  * zc_name		name of volume
1818  *
1819  * outputs:		none
1820  */
1821 static int
1822 zfs_ioc_remove_minor(zfs_cmd_t *zc)
1823 {
1824 	return (zvol_remove_minor(zc->zc_name));
1825 }
1826 
1827 /*
1828  * Search the vfs list for a specified resource.  Returns a pointer to it
1829  * or NULL if no suitable entry is found. The caller of this routine
1830  * is responsible for releasing the returned vfs pointer.
1831  */
1832 static vfs_t *
1833 zfs_get_vfs(const char *resource)
1834 {
1835 	struct vfs *vfsp;
1836 	struct vfs *vfs_found = NULL;
1837 
1838 	vfs_list_read_lock();
1839 	vfsp = rootvfs;
1840 	do {
1841 		if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
1842 			VFS_HOLD(vfsp);
1843 			vfs_found = vfsp;
1844 			break;
1845 		}
1846 		vfsp = vfsp->vfs_next;
1847 	} while (vfsp != rootvfs);
1848 	vfs_list_unlock();
1849 	return (vfs_found);
1850 }
1851 
1852 /* ARGSUSED */
1853 static void
1854 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
1855 {
1856 	zfs_creat_t *zct = arg;
1857 
1858 	zfs_create_fs(os, cr, zct->zct_zplprops, tx);
1859 }
1860 
1861 #define	ZFS_PROP_UNDEFINED	((uint64_t)-1)
1862 
1863 /*
1864  * inputs:
1865  * createprops		list of properties requested by creator
1866  * default_zplver	zpl version to use if unspecified in createprops
1867  * fuids_ok		fuids allowed in this version of the spa?
1868  * os			parent objset pointer (NULL if root fs)
1869  *
1870  * outputs:
1871  * zplprops	values for the zplprops we attach to the master node object
1872  * is_ci	true if requested file system will be purely case-insensitive
1873  *
1874  * Determine the settings for utf8only, normalization and
1875  * casesensitivity.  Specific values may have been requested by the
1876  * creator and/or we can inherit values from the parent dataset.  If
1877  * the file system is of too early a vintage, a creator can not
1878  * request settings for these properties, even if the requested
1879  * setting is the default value.  We don't actually want to create dsl
1880  * properties for these, so remove them from the source nvlist after
1881  * processing.
1882  */
1883 static int
1884 zfs_fill_zplprops_impl(objset_t *os, uint64_t default_zplver,
1885     boolean_t fuids_ok, nvlist_t *createprops, nvlist_t *zplprops,
1886     boolean_t *is_ci)
1887 {
1888 	uint64_t zplver = default_zplver;
1889 	uint64_t sense = ZFS_PROP_UNDEFINED;
1890 	uint64_t norm = ZFS_PROP_UNDEFINED;
1891 	uint64_t u8 = ZFS_PROP_UNDEFINED;
1892 
1893 	ASSERT(zplprops != NULL);
1894 
1895 	/*
1896 	 * Pull out creator prop choices, if any.
1897 	 */
1898 	if (createprops) {
1899 		(void) nvlist_lookup_uint64(createprops,
1900 		    zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
1901 		(void) nvlist_lookup_uint64(createprops,
1902 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
1903 		(void) nvlist_remove_all(createprops,
1904 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE));
1905 		(void) nvlist_lookup_uint64(createprops,
1906 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
1907 		(void) nvlist_remove_all(createprops,
1908 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
1909 		(void) nvlist_lookup_uint64(createprops,
1910 		    zfs_prop_to_name(ZFS_PROP_CASE), &sense);
1911 		(void) nvlist_remove_all(createprops,
1912 		    zfs_prop_to_name(ZFS_PROP_CASE));
1913 	}
1914 
1915 	/*
1916 	 * If the zpl version requested is whacky or the file system
1917 	 * or pool is version is too "young" to support normalization
1918 	 * and the creator tried to set a value for one of the props,
1919 	 * error out.
1920 	 */
1921 	if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
1922 	    (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
1923 	    (zplver < ZPL_VERSION_NORMALIZATION &&
1924 	    (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
1925 	    sense != ZFS_PROP_UNDEFINED)))
1926 		return (ENOTSUP);
1927 
1928 	/*
1929 	 * Put the version in the zplprops
1930 	 */
1931 	VERIFY(nvlist_add_uint64(zplprops,
1932 	    zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
1933 
1934 	if (norm == ZFS_PROP_UNDEFINED)
1935 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
1936 	VERIFY(nvlist_add_uint64(zplprops,
1937 	    zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
1938 
1939 	/*
1940 	 * If we're normalizing, names must always be valid UTF-8 strings.
1941 	 */
1942 	if (norm)
1943 		u8 = 1;
1944 	if (u8 == ZFS_PROP_UNDEFINED)
1945 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
1946 	VERIFY(nvlist_add_uint64(zplprops,
1947 	    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
1948 
1949 	if (sense == ZFS_PROP_UNDEFINED)
1950 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
1951 	VERIFY(nvlist_add_uint64(zplprops,
1952 	    zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
1953 
1954 	if (is_ci)
1955 		*is_ci = (sense == ZFS_CASE_INSENSITIVE);
1956 
1957 	return (0);
1958 }
1959 
1960 static int
1961 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
1962     nvlist_t *zplprops, boolean_t *is_ci)
1963 {
1964 	boolean_t fuids_ok = B_TRUE;
1965 	uint64_t zplver = ZPL_VERSION;
1966 	objset_t *os = NULL;
1967 	char parentname[MAXNAMELEN];
1968 	char *cp;
1969 	int error;
1970 
1971 	(void) strlcpy(parentname, dataset, sizeof (parentname));
1972 	cp = strrchr(parentname, '/');
1973 	ASSERT(cp != NULL);
1974 	cp[0] = '\0';
1975 
1976 	if (zfs_earlier_version(dataset, SPA_VERSION_FUID)) {
1977 		zplver = ZPL_VERSION_FUID - 1;
1978 		fuids_ok = B_FALSE;
1979 	}
1980 
1981 	/*
1982 	 * Open parent object set so we can inherit zplprop values.
1983 	 */
1984 	if ((error = dmu_objset_open(parentname, DMU_OST_ANY,
1985 	    DS_MODE_USER | DS_MODE_READONLY, &os)) != 0)
1986 		return (error);
1987 
1988 	error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, createprops,
1989 	    zplprops, is_ci);
1990 	dmu_objset_close(os);
1991 	return (error);
1992 }
1993 
1994 static int
1995 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
1996     nvlist_t *zplprops, boolean_t *is_ci)
1997 {
1998 	boolean_t fuids_ok = B_TRUE;
1999 	uint64_t zplver = ZPL_VERSION;
2000 	int error;
2001 
2002 	if (spa_vers < SPA_VERSION_FUID) {
2003 		zplver = ZPL_VERSION_FUID - 1;
2004 		fuids_ok = B_FALSE;
2005 	}
2006 
2007 	error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, createprops,
2008 	    zplprops, is_ci);
2009 	return (error);
2010 }
2011 
2012 /*
2013  * inputs:
2014  * zc_objset_type	type of objset to create (fs vs zvol)
2015  * zc_name		name of new objset
2016  * zc_value		name of snapshot to clone from (may be empty)
2017  * zc_nvlist_src{_size}	nvlist of properties to apply
2018  *
2019  * outputs: none
2020  */
2021 static int
2022 zfs_ioc_create(zfs_cmd_t *zc)
2023 {
2024 	objset_t *clone;
2025 	int error = 0;
2026 	zfs_creat_t zct;
2027 	nvlist_t *nvprops = NULL;
2028 	void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
2029 	dmu_objset_type_t type = zc->zc_objset_type;
2030 
2031 	switch (type) {
2032 
2033 	case DMU_OST_ZFS:
2034 		cbfunc = zfs_create_cb;
2035 		break;
2036 
2037 	case DMU_OST_ZVOL:
2038 		cbfunc = zvol_create_cb;
2039 		break;
2040 
2041 	default:
2042 		cbfunc = NULL;
2043 		break;
2044 	}
2045 	if (strchr(zc->zc_name, '@') ||
2046 	    strchr(zc->zc_name, '%'))
2047 		return (EINVAL);
2048 
2049 	if (zc->zc_nvlist_src != NULL &&
2050 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2051 	    &nvprops)) != 0)
2052 		return (error);
2053 
2054 	zct.zct_zplprops = NULL;
2055 	zct.zct_props = nvprops;
2056 
2057 	if (zc->zc_value[0] != '\0') {
2058 		/*
2059 		 * We're creating a clone of an existing snapshot.
2060 		 */
2061 		zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
2062 		if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0) {
2063 			nvlist_free(nvprops);
2064 			return (EINVAL);
2065 		}
2066 
2067 		error = dmu_objset_open(zc->zc_value, type,
2068 		    DS_MODE_USER | DS_MODE_READONLY, &clone);
2069 		if (error) {
2070 			nvlist_free(nvprops);
2071 			return (error);
2072 		}
2073 
2074 		error = dmu_objset_create(zc->zc_name, type, clone, 0,
2075 		    NULL, NULL);
2076 		if (error) {
2077 			dmu_objset_close(clone);
2078 			nvlist_free(nvprops);
2079 			return (error);
2080 		}
2081 		dmu_objset_close(clone);
2082 	} else {
2083 		boolean_t is_insensitive = B_FALSE;
2084 
2085 		if (cbfunc == NULL) {
2086 			nvlist_free(nvprops);
2087 			return (EINVAL);
2088 		}
2089 
2090 		if (type == DMU_OST_ZVOL) {
2091 			uint64_t volsize, volblocksize;
2092 
2093 			if (nvprops == NULL ||
2094 			    nvlist_lookup_uint64(nvprops,
2095 			    zfs_prop_to_name(ZFS_PROP_VOLSIZE),
2096 			    &volsize) != 0) {
2097 				nvlist_free(nvprops);
2098 				return (EINVAL);
2099 			}
2100 
2101 			if ((error = nvlist_lookup_uint64(nvprops,
2102 			    zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
2103 			    &volblocksize)) != 0 && error != ENOENT) {
2104 				nvlist_free(nvprops);
2105 				return (EINVAL);
2106 			}
2107 
2108 			if (error != 0)
2109 				volblocksize = zfs_prop_default_numeric(
2110 				    ZFS_PROP_VOLBLOCKSIZE);
2111 
2112 			if ((error = zvol_check_volblocksize(
2113 			    volblocksize)) != 0 ||
2114 			    (error = zvol_check_volsize(volsize,
2115 			    volblocksize)) != 0) {
2116 				nvlist_free(nvprops);
2117 				return (error);
2118 			}
2119 		} else if (type == DMU_OST_ZFS) {
2120 			int error;
2121 
2122 			/*
2123 			 * We have to have normalization and
2124 			 * case-folding flags correct when we do the
2125 			 * file system creation, so go figure them out
2126 			 * now.
2127 			 */
2128 			VERIFY(nvlist_alloc(&zct.zct_zplprops,
2129 			    NV_UNIQUE_NAME, KM_SLEEP) == 0);
2130 			error = zfs_fill_zplprops(zc->zc_name, nvprops,
2131 			    zct.zct_zplprops, &is_insensitive);
2132 			if (error != 0) {
2133 				nvlist_free(nvprops);
2134 				nvlist_free(zct.zct_zplprops);
2135 				return (error);
2136 			}
2137 		}
2138 		error = dmu_objset_create(zc->zc_name, type, NULL,
2139 		    is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
2140 		nvlist_free(zct.zct_zplprops);
2141 	}
2142 
2143 	/*
2144 	 * It would be nice to do this atomically.
2145 	 */
2146 	if (error == 0) {
2147 		if ((error = zfs_set_prop_nvlist(zc->zc_name, nvprops)) != 0)
2148 			(void) dmu_objset_destroy(zc->zc_name);
2149 	}
2150 	nvlist_free(nvprops);
2151 	return (error);
2152 }
2153 
2154 struct snap_prop_arg {
2155 	nvlist_t *nvprops;
2156 	const char *snapname;
2157 };
2158 
2159 static int
2160 set_snap_props(char *name, void *arg)
2161 {
2162 	struct snap_prop_arg *snpa = arg;
2163 	int len = strlen(name) + strlen(snpa->snapname) + 2;
2164 	char *buf = kmem_alloc(len, KM_SLEEP);
2165 	int err;
2166 
2167 	(void) snprintf(buf, len, "%s@%s", name, snpa->snapname);
2168 	err = zfs_set_prop_nvlist(buf, snpa->nvprops);
2169 	if (err)
2170 		(void) dmu_objset_destroy(buf);
2171 	kmem_free(buf, len);
2172 	return (err);
2173 }
2174 
2175 /*
2176  * inputs:
2177  * zc_name	name of filesystem
2178  * zc_value	short name of snapshot
2179  * zc_cookie	recursive flag
2180  *
2181  * outputs:	none
2182  */
2183 static int
2184 zfs_ioc_snapshot(zfs_cmd_t *zc)
2185 {
2186 	nvlist_t *nvprops = NULL;
2187 	int error;
2188 	boolean_t recursive = zc->zc_cookie;
2189 
2190 	if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
2191 		return (EINVAL);
2192 
2193 	if (zc->zc_nvlist_src != NULL &&
2194 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2195 	    &nvprops)) != 0)
2196 		return (error);
2197 
2198 	error = dmu_objset_snapshot(zc->zc_name, zc->zc_value, recursive);
2199 
2200 	/*
2201 	 * It would be nice to do this atomically.
2202 	 */
2203 	if (error == 0) {
2204 		struct snap_prop_arg snpa;
2205 		snpa.nvprops = nvprops;
2206 		snpa.snapname = zc->zc_value;
2207 		if (recursive) {
2208 			error = dmu_objset_find(zc->zc_name,
2209 			    set_snap_props, &snpa, DS_FIND_CHILDREN);
2210 			if (error) {
2211 				(void) dmu_snapshots_destroy(zc->zc_name,
2212 				    zc->zc_value);
2213 			}
2214 		} else {
2215 			error = set_snap_props(zc->zc_name, &snpa);
2216 		}
2217 	}
2218 	nvlist_free(nvprops);
2219 	return (error);
2220 }
2221 
2222 int
2223 zfs_unmount_snap(char *name, void *arg)
2224 {
2225 	vfs_t *vfsp = NULL;
2226 
2227 	if (arg) {
2228 		char *snapname = arg;
2229 		int len = strlen(name) + strlen(snapname) + 2;
2230 		char *buf = kmem_alloc(len, KM_SLEEP);
2231 
2232 		(void) strcpy(buf, name);
2233 		(void) strcat(buf, "@");
2234 		(void) strcat(buf, snapname);
2235 		vfsp = zfs_get_vfs(buf);
2236 		kmem_free(buf, len);
2237 	} else if (strchr(name, '@')) {
2238 		vfsp = zfs_get_vfs(name);
2239 	}
2240 
2241 	if (vfsp) {
2242 		/*
2243 		 * Always force the unmount for snapshots.
2244 		 */
2245 		int flag = MS_FORCE;
2246 		int err;
2247 
2248 		if ((err = vn_vfswlock(vfsp->vfs_vnodecovered)) != 0) {
2249 			VFS_RELE(vfsp);
2250 			return (err);
2251 		}
2252 		VFS_RELE(vfsp);
2253 		if ((err = dounmount(vfsp, flag, kcred)) != 0)
2254 			return (err);
2255 	}
2256 	return (0);
2257 }
2258 
2259 /*
2260  * inputs:
2261  * zc_name	name of filesystem
2262  * zc_value	short name of snapshot
2263  *
2264  * outputs:	none
2265  */
2266 static int
2267 zfs_ioc_destroy_snaps(zfs_cmd_t *zc)
2268 {
2269 	int err;
2270 
2271 	if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
2272 		return (EINVAL);
2273 	err = dmu_objset_find(zc->zc_name,
2274 	    zfs_unmount_snap, zc->zc_value, DS_FIND_CHILDREN);
2275 	if (err)
2276 		return (err);
2277 	return (dmu_snapshots_destroy(zc->zc_name, zc->zc_value));
2278 }
2279 
2280 /*
2281  * inputs:
2282  * zc_name		name of dataset to destroy
2283  * zc_objset_type	type of objset
2284  *
2285  * outputs:		none
2286  */
2287 static int
2288 zfs_ioc_destroy(zfs_cmd_t *zc)
2289 {
2290 	if (strchr(zc->zc_name, '@') && zc->zc_objset_type == DMU_OST_ZFS) {
2291 		int err = zfs_unmount_snap(zc->zc_name, NULL);
2292 		if (err)
2293 			return (err);
2294 	}
2295 
2296 	return (dmu_objset_destroy(zc->zc_name));
2297 }
2298 
2299 /*
2300  * inputs:
2301  * zc_name	name of dataset to rollback (to most recent snapshot)
2302  *
2303  * outputs:	none
2304  */
2305 static int
2306 zfs_ioc_rollback(zfs_cmd_t *zc)
2307 {
2308 	objset_t *os;
2309 	int error;
2310 	zfsvfs_t *zfsvfs = NULL;
2311 
2312 	/*
2313 	 * Get the zfsvfs for the receiving objset. There
2314 	 * won't be one if we're operating on a zvol, if the
2315 	 * objset doesn't exist yet, or is not mounted.
2316 	 */
2317 	error = dmu_objset_open(zc->zc_name, DMU_OST_ANY, DS_MODE_USER, &os);
2318 	if (error)
2319 		return (error);
2320 
2321 	if (dmu_objset_type(os) == DMU_OST_ZFS) {
2322 		mutex_enter(&os->os->os_user_ptr_lock);
2323 		zfsvfs = dmu_objset_get_user(os);
2324 		if (zfsvfs != NULL)
2325 			VFS_HOLD(zfsvfs->z_vfs);
2326 		mutex_exit(&os->os->os_user_ptr_lock);
2327 	}
2328 
2329 	if (zfsvfs != NULL) {
2330 		char *osname;
2331 		int mode;
2332 
2333 		osname = kmem_alloc(MAXNAMELEN, KM_SLEEP);
2334 		error = zfs_suspend_fs(zfsvfs, osname, &mode);
2335 		if (error == 0) {
2336 			int resume_err;
2337 
2338 			ASSERT(strcmp(osname, zc->zc_name) == 0);
2339 			error = dmu_objset_rollback(os);
2340 			resume_err = zfs_resume_fs(zfsvfs, osname, mode);
2341 			error = error ? error : resume_err;
2342 		} else {
2343 			dmu_objset_close(os);
2344 		}
2345 		kmem_free(osname, MAXNAMELEN);
2346 		VFS_RELE(zfsvfs->z_vfs);
2347 	} else {
2348 		error = dmu_objset_rollback(os);
2349 	}
2350 	/* Note, the dmu_objset_rollback() releases the objset for us. */
2351 
2352 	return (error);
2353 }
2354 
2355 /*
2356  * inputs:
2357  * zc_name	old name of dataset
2358  * zc_value	new name of dataset
2359  * zc_cookie	recursive flag (only valid for snapshots)
2360  *
2361  * outputs:	none
2362  */
2363 static int
2364 zfs_ioc_rename(zfs_cmd_t *zc)
2365 {
2366 	boolean_t recursive = zc->zc_cookie & 1;
2367 
2368 	zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
2369 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
2370 	    strchr(zc->zc_value, '%'))
2371 		return (EINVAL);
2372 
2373 	/*
2374 	 * Unmount snapshot unless we're doing a recursive rename,
2375 	 * in which case the dataset code figures out which snapshots
2376 	 * to unmount.
2377 	 */
2378 	if (!recursive && strchr(zc->zc_name, '@') != NULL &&
2379 	    zc->zc_objset_type == DMU_OST_ZFS) {
2380 		int err = zfs_unmount_snap(zc->zc_name, NULL);
2381 		if (err)
2382 			return (err);
2383 	}
2384 	return (dmu_objset_rename(zc->zc_name, zc->zc_value, recursive));
2385 }
2386 
2387 static void
2388 clear_props(char *dataset, nvlist_t *props)
2389 {
2390 	zfs_cmd_t *zc;
2391 	nvpair_t *prop;
2392 
2393 	if (props == NULL)
2394 		return;
2395 	zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
2396 	(void) strcpy(zc->zc_name, dataset);
2397 	for (prop = nvlist_next_nvpair(props, NULL); prop;
2398 	    prop = nvlist_next_nvpair(props, prop)) {
2399 		(void) strcpy(zc->zc_value, nvpair_name(prop));
2400 		if (zfs_secpolicy_inherit(zc, CRED()) == 0)
2401 			(void) zfs_ioc_inherit_prop(zc);
2402 	}
2403 	kmem_free(zc, sizeof (zfs_cmd_t));
2404 }
2405 
2406 /*
2407  * inputs:
2408  * zc_name		name of containing filesystem
2409  * zc_nvlist_src{_size}	nvlist of properties to apply
2410  * zc_value		name of snapshot to create
2411  * zc_string		name of clone origin (if DRR_FLAG_CLONE)
2412  * zc_cookie		file descriptor to recv from
2413  * zc_begin_record	the BEGIN record of the stream (not byteswapped)
2414  * zc_guid		force flag
2415  *
2416  * outputs:
2417  * zc_cookie		number of bytes read
2418  */
2419 static int
2420 zfs_ioc_recv(zfs_cmd_t *zc)
2421 {
2422 	file_t *fp;
2423 	objset_t *os;
2424 	dmu_recv_cookie_t drc;
2425 	zfsvfs_t *zfsvfs = NULL;
2426 	boolean_t force = (boolean_t)zc->zc_guid;
2427 	int error, fd;
2428 	offset_t off;
2429 	nvlist_t *props = NULL;
2430 	nvlist_t *origprops = NULL;
2431 	objset_t *origin = NULL;
2432 	char *tosnap;
2433 	char tofs[ZFS_MAXNAMELEN];
2434 
2435 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
2436 	    strchr(zc->zc_value, '@') == NULL ||
2437 	    strchr(zc->zc_value, '%'))
2438 		return (EINVAL);
2439 
2440 	(void) strcpy(tofs, zc->zc_value);
2441 	tosnap = strchr(tofs, '@');
2442 	*tosnap = '\0';
2443 	tosnap++;
2444 
2445 	if (zc->zc_nvlist_src != NULL &&
2446 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2447 	    &props)) != 0)
2448 		return (error);
2449 
2450 	fd = zc->zc_cookie;
2451 	fp = getf(fd);
2452 	if (fp == NULL) {
2453 		nvlist_free(props);
2454 		return (EBADF);
2455 	}
2456 
2457 	if (dmu_objset_open(tofs, DMU_OST_ANY,
2458 	    DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
2459 		/*
2460 		 * Try to get the zfsvfs for the receiving objset.
2461 		 * There won't be one if we're operating on a zvol,
2462 		 * if the objset doesn't exist yet, or is not mounted.
2463 		 */
2464 		mutex_enter(&os->os->os_user_ptr_lock);
2465 		if (zfsvfs = dmu_objset_get_user(os)) {
2466 			if (!mutex_tryenter(&zfsvfs->z_online_recv_lock)) {
2467 				mutex_exit(&os->os->os_user_ptr_lock);
2468 				dmu_objset_close(os);
2469 				zfsvfs = NULL;
2470 				error = EBUSY;
2471 				goto out;
2472 			}
2473 			VFS_HOLD(zfsvfs->z_vfs);
2474 		}
2475 		mutex_exit(&os->os->os_user_ptr_lock);
2476 
2477 		/*
2478 		 * If new properties are supplied, they are to completely
2479 		 * replace the existing ones, so stash away the existing ones.
2480 		 */
2481 		if (props)
2482 			(void) dsl_prop_get_all(os, &origprops, TRUE);
2483 
2484 		dmu_objset_close(os);
2485 	}
2486 
2487 	if (zc->zc_string[0]) {
2488 		error = dmu_objset_open(zc->zc_string, DMU_OST_ANY,
2489 		    DS_MODE_USER | DS_MODE_READONLY, &origin);
2490 		if (error)
2491 			goto out;
2492 	}
2493 
2494 	error = dmu_recv_begin(tofs, tosnap, &zc->zc_begin_record,
2495 	    force, origin, zfsvfs != NULL, &drc);
2496 	if (origin)
2497 		dmu_objset_close(origin);
2498 	if (error)
2499 		goto out;
2500 
2501 	/*
2502 	 * Reset properties.  We do this before we receive the stream
2503 	 * so that the properties are applied to the new data.
2504 	 */
2505 	if (props) {
2506 		clear_props(tofs, origprops);
2507 		/*
2508 		 * XXX - Note, this is all-or-nothing; should be best-effort.
2509 		 */
2510 		(void) zfs_set_prop_nvlist(tofs, props);
2511 	}
2512 
2513 	off = fp->f_offset;
2514 	error = dmu_recv_stream(&drc, fp->f_vnode, &off);
2515 
2516 	if (error == 0 && zfsvfs) {
2517 		char *osname;
2518 		int mode;
2519 
2520 		/* online recv */
2521 		osname = kmem_alloc(MAXNAMELEN, KM_SLEEP);
2522 		error = zfs_suspend_fs(zfsvfs, osname, &mode);
2523 		if (error == 0) {
2524 			int resume_err;
2525 
2526 			error = dmu_recv_end(&drc);
2527 			resume_err = zfs_resume_fs(zfsvfs, osname, mode);
2528 			error = error ? error : resume_err;
2529 		} else {
2530 			dmu_recv_abort_cleanup(&drc);
2531 		}
2532 		kmem_free(osname, MAXNAMELEN);
2533 	} else if (error == 0) {
2534 		error = dmu_recv_end(&drc);
2535 	}
2536 
2537 	zc->zc_cookie = off - fp->f_offset;
2538 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
2539 		fp->f_offset = off;
2540 
2541 	/*
2542 	 * On error, restore the original props.
2543 	 */
2544 	if (error && props) {
2545 		clear_props(tofs, props);
2546 		(void) zfs_set_prop_nvlist(tofs, origprops);
2547 	}
2548 out:
2549 	if (zfsvfs) {
2550 		mutex_exit(&zfsvfs->z_online_recv_lock);
2551 		VFS_RELE(zfsvfs->z_vfs);
2552 	}
2553 	nvlist_free(props);
2554 	nvlist_free(origprops);
2555 	releasef(fd);
2556 	return (error);
2557 }
2558 
2559 /*
2560  * inputs:
2561  * zc_name	name of snapshot to send
2562  * zc_value	short name of incremental fromsnap (may be empty)
2563  * zc_cookie	file descriptor to send stream to
2564  * zc_obj	fromorigin flag (mutually exclusive with zc_value)
2565  *
2566  * outputs: none
2567  */
2568 static int
2569 zfs_ioc_send(zfs_cmd_t *zc)
2570 {
2571 	objset_t *fromsnap = NULL;
2572 	objset_t *tosnap;
2573 	file_t *fp;
2574 	int error;
2575 	offset_t off;
2576 
2577 	error = dmu_objset_open(zc->zc_name, DMU_OST_ANY,
2578 	    DS_MODE_USER | DS_MODE_READONLY, &tosnap);
2579 	if (error)
2580 		return (error);
2581 
2582 	if (zc->zc_value[0] != '\0') {
2583 		char *buf;
2584 		char *cp;
2585 
2586 		buf = kmem_alloc(MAXPATHLEN, KM_SLEEP);
2587 		(void) strncpy(buf, zc->zc_name, MAXPATHLEN);
2588 		cp = strchr(buf, '@');
2589 		if (cp)
2590 			*(cp+1) = 0;
2591 		(void) strncat(buf, zc->zc_value, MAXPATHLEN);
2592 		error = dmu_objset_open(buf, DMU_OST_ANY,
2593 		    DS_MODE_USER | DS_MODE_READONLY, &fromsnap);
2594 		kmem_free(buf, MAXPATHLEN);
2595 		if (error) {
2596 			dmu_objset_close(tosnap);
2597 			return (error);
2598 		}
2599 	}
2600 
2601 	fp = getf(zc->zc_cookie);
2602 	if (fp == NULL) {
2603 		dmu_objset_close(tosnap);
2604 		if (fromsnap)
2605 			dmu_objset_close(fromsnap);
2606 		return (EBADF);
2607 	}
2608 
2609 	off = fp->f_offset;
2610 	error = dmu_sendbackup(tosnap, fromsnap, zc->zc_obj, fp->f_vnode, &off);
2611 
2612 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
2613 		fp->f_offset = off;
2614 	releasef(zc->zc_cookie);
2615 	if (fromsnap)
2616 		dmu_objset_close(fromsnap);
2617 	dmu_objset_close(tosnap);
2618 	return (error);
2619 }
2620 
2621 static int
2622 zfs_ioc_inject_fault(zfs_cmd_t *zc)
2623 {
2624 	int id, error;
2625 
2626 	error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
2627 	    &zc->zc_inject_record);
2628 
2629 	if (error == 0)
2630 		zc->zc_guid = (uint64_t)id;
2631 
2632 	return (error);
2633 }
2634 
2635 static int
2636 zfs_ioc_clear_fault(zfs_cmd_t *zc)
2637 {
2638 	return (zio_clear_fault((int)zc->zc_guid));
2639 }
2640 
2641 static int
2642 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
2643 {
2644 	int id = (int)zc->zc_guid;
2645 	int error;
2646 
2647 	error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
2648 	    &zc->zc_inject_record);
2649 
2650 	zc->zc_guid = id;
2651 
2652 	return (error);
2653 }
2654 
2655 static int
2656 zfs_ioc_error_log(zfs_cmd_t *zc)
2657 {
2658 	spa_t *spa;
2659 	int error;
2660 	size_t count = (size_t)zc->zc_nvlist_dst_size;
2661 
2662 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2663 		return (error);
2664 
2665 	error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
2666 	    &count);
2667 	if (error == 0)
2668 		zc->zc_nvlist_dst_size = count;
2669 	else
2670 		zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
2671 
2672 	spa_close(spa, FTAG);
2673 
2674 	return (error);
2675 }
2676 
2677 static int
2678 zfs_ioc_clear(zfs_cmd_t *zc)
2679 {
2680 	spa_t *spa;
2681 	vdev_t *vd;
2682 	int error;
2683 
2684 	/*
2685 	 * On zpool clear we also fix up missing slogs
2686 	 */
2687 	mutex_enter(&spa_namespace_lock);
2688 	spa = spa_lookup(zc->zc_name);
2689 	if (spa == NULL) {
2690 		mutex_exit(&spa_namespace_lock);
2691 		return (EIO);
2692 	}
2693 	if (spa->spa_log_state == SPA_LOG_MISSING) {
2694 		/* we need to let spa_open/spa_load clear the chains */
2695 		spa->spa_log_state = SPA_LOG_CLEAR;
2696 	}
2697 	mutex_exit(&spa_namespace_lock);
2698 
2699 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2700 		return (error);
2701 
2702 	spa_vdev_state_enter(spa);
2703 
2704 	if (zc->zc_guid == 0) {
2705 		vd = NULL;
2706 	} else {
2707 		vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
2708 		if (vd == NULL) {
2709 			(void) spa_vdev_state_exit(spa, NULL, ENODEV);
2710 			spa_close(spa, FTAG);
2711 			return (ENODEV);
2712 		}
2713 	}
2714 
2715 	vdev_clear(spa, vd);
2716 
2717 	(void) spa_vdev_state_exit(spa, NULL, 0);
2718 
2719 	/*
2720 	 * Resume any suspended I/Os.
2721 	 */
2722 	zio_resume(spa);
2723 
2724 	spa_close(spa, FTAG);
2725 
2726 	return (0);
2727 }
2728 
2729 /*
2730  * inputs:
2731  * zc_name	name of filesystem
2732  * zc_value	name of origin snapshot
2733  *
2734  * outputs:	none
2735  */
2736 static int
2737 zfs_ioc_promote(zfs_cmd_t *zc)
2738 {
2739 	char *cp;
2740 
2741 	/*
2742 	 * We don't need to unmount *all* the origin fs's snapshots, but
2743 	 * it's easier.
2744 	 */
2745 	cp = strchr(zc->zc_value, '@');
2746 	if (cp)
2747 		*cp = '\0';
2748 	(void) dmu_objset_find(zc->zc_value,
2749 	    zfs_unmount_snap, NULL, DS_FIND_SNAPSHOTS);
2750 	return (dsl_dataset_promote(zc->zc_name));
2751 }
2752 
2753 /*
2754  * We don't want to have a hard dependency
2755  * against some special symbols in sharefs
2756  * nfs, and smbsrv.  Determine them if needed when
2757  * the first file system is shared.
2758  * Neither sharefs, nfs or smbsrv are unloadable modules.
2759  */
2760 int (*znfsexport_fs)(void *arg);
2761 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
2762 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
2763 
2764 int zfs_nfsshare_inited;
2765 int zfs_smbshare_inited;
2766 
2767 ddi_modhandle_t nfs_mod;
2768 ddi_modhandle_t sharefs_mod;
2769 ddi_modhandle_t smbsrv_mod;
2770 kmutex_t zfs_share_lock;
2771 
2772 static int
2773 zfs_init_sharefs()
2774 {
2775 	int error;
2776 
2777 	ASSERT(MUTEX_HELD(&zfs_share_lock));
2778 	/* Both NFS and SMB shares also require sharetab support. */
2779 	if (sharefs_mod == NULL && ((sharefs_mod =
2780 	    ddi_modopen("fs/sharefs",
2781 	    KRTLD_MODE_FIRST, &error)) == NULL)) {
2782 		return (ENOSYS);
2783 	}
2784 	if (zshare_fs == NULL && ((zshare_fs =
2785 	    (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
2786 	    ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
2787 		return (ENOSYS);
2788 	}
2789 	return (0);
2790 }
2791 
2792 static int
2793 zfs_ioc_share(zfs_cmd_t *zc)
2794 {
2795 	int error;
2796 	int opcode;
2797 
2798 	switch (zc->zc_share.z_sharetype) {
2799 	case ZFS_SHARE_NFS:
2800 	case ZFS_UNSHARE_NFS:
2801 		if (zfs_nfsshare_inited == 0) {
2802 			mutex_enter(&zfs_share_lock);
2803 			if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
2804 			    KRTLD_MODE_FIRST, &error)) == NULL)) {
2805 				mutex_exit(&zfs_share_lock);
2806 				return (ENOSYS);
2807 			}
2808 			if (znfsexport_fs == NULL &&
2809 			    ((znfsexport_fs = (int (*)(void *))
2810 			    ddi_modsym(nfs_mod,
2811 			    "nfs_export", &error)) == NULL)) {
2812 				mutex_exit(&zfs_share_lock);
2813 				return (ENOSYS);
2814 			}
2815 			error = zfs_init_sharefs();
2816 			if (error) {
2817 				mutex_exit(&zfs_share_lock);
2818 				return (ENOSYS);
2819 			}
2820 			zfs_nfsshare_inited = 1;
2821 			mutex_exit(&zfs_share_lock);
2822 		}
2823 		break;
2824 	case ZFS_SHARE_SMB:
2825 	case ZFS_UNSHARE_SMB:
2826 		if (zfs_smbshare_inited == 0) {
2827 			mutex_enter(&zfs_share_lock);
2828 			if (smbsrv_mod == NULL && ((smbsrv_mod =
2829 			    ddi_modopen("drv/smbsrv",
2830 			    KRTLD_MODE_FIRST, &error)) == NULL)) {
2831 				mutex_exit(&zfs_share_lock);
2832 				return (ENOSYS);
2833 			}
2834 			if (zsmbexport_fs == NULL && ((zsmbexport_fs =
2835 			    (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
2836 			    "smb_server_share", &error)) == NULL)) {
2837 				mutex_exit(&zfs_share_lock);
2838 				return (ENOSYS);
2839 			}
2840 			error = zfs_init_sharefs();
2841 			if (error) {
2842 				mutex_exit(&zfs_share_lock);
2843 				return (ENOSYS);
2844 			}
2845 			zfs_smbshare_inited = 1;
2846 			mutex_exit(&zfs_share_lock);
2847 		}
2848 		break;
2849 	default:
2850 		return (EINVAL);
2851 	}
2852 
2853 	switch (zc->zc_share.z_sharetype) {
2854 	case ZFS_SHARE_NFS:
2855 	case ZFS_UNSHARE_NFS:
2856 		if (error =
2857 		    znfsexport_fs((void *)
2858 		    (uintptr_t)zc->zc_share.z_exportdata))
2859 			return (error);
2860 		break;
2861 	case ZFS_SHARE_SMB:
2862 	case ZFS_UNSHARE_SMB:
2863 		if (error = zsmbexport_fs((void *)
2864 		    (uintptr_t)zc->zc_share.z_exportdata,
2865 		    zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
2866 		    B_TRUE : B_FALSE)) {
2867 			return (error);
2868 		}
2869 		break;
2870 	}
2871 
2872 	opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
2873 	    zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
2874 	    SHAREFS_ADD : SHAREFS_REMOVE;
2875 
2876 	/*
2877 	 * Add or remove share from sharetab
2878 	 */
2879 	error = zshare_fs(opcode,
2880 	    (void *)(uintptr_t)zc->zc_share.z_sharedata,
2881 	    zc->zc_share.z_sharemax);
2882 
2883 	return (error);
2884 
2885 }
2886 
2887 /*
2888  * pool create, destroy, and export don't log the history as part of
2889  * zfsdev_ioctl, but rather zfs_ioc_pool_create, and zfs_ioc_pool_export
2890  * do the logging of those commands.
2891  */
2892 static zfs_ioc_vec_t zfs_ioc_vec[] = {
2893 	{ zfs_ioc_pool_create, zfs_secpolicy_config, POOL_NAME, B_FALSE },
2894 	{ zfs_ioc_pool_destroy,	zfs_secpolicy_config, POOL_NAME, B_FALSE },
2895 	{ zfs_ioc_pool_import, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2896 	{ zfs_ioc_pool_export, zfs_secpolicy_config, POOL_NAME, B_FALSE },
2897 	{ zfs_ioc_pool_configs,	zfs_secpolicy_none, NO_NAME, B_FALSE },
2898 	{ zfs_ioc_pool_stats, zfs_secpolicy_read, POOL_NAME, B_FALSE },
2899 	{ zfs_ioc_pool_tryimport, zfs_secpolicy_config, NO_NAME, B_FALSE },
2900 	{ zfs_ioc_pool_scrub, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2901 	{ zfs_ioc_pool_freeze, zfs_secpolicy_config, NO_NAME, B_FALSE },
2902 	{ zfs_ioc_pool_upgrade,	zfs_secpolicy_config, POOL_NAME, B_TRUE },
2903 	{ zfs_ioc_pool_get_history, zfs_secpolicy_config, POOL_NAME, B_FALSE },
2904 	{ zfs_ioc_vdev_add, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2905 	{ zfs_ioc_vdev_remove, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2906 	{ zfs_ioc_vdev_set_state, zfs_secpolicy_config,	POOL_NAME, B_TRUE },
2907 	{ zfs_ioc_vdev_attach, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2908 	{ zfs_ioc_vdev_detach, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2909 	{ zfs_ioc_vdev_setpath,	zfs_secpolicy_config, POOL_NAME, B_FALSE },
2910 	{ zfs_ioc_objset_stats,	zfs_secpolicy_read, DATASET_NAME, B_FALSE },
2911 	{ zfs_ioc_objset_zplprops, zfs_secpolicy_read, DATASET_NAME, B_FALSE },
2912 	{ zfs_ioc_dataset_list_next, zfs_secpolicy_read,
2913 	    DATASET_NAME, B_FALSE },
2914 	{ zfs_ioc_snapshot_list_next, zfs_secpolicy_read,
2915 	    DATASET_NAME, B_FALSE },
2916 	{ zfs_ioc_set_prop, zfs_secpolicy_none, DATASET_NAME, B_TRUE },
2917 	{ zfs_ioc_create_minor,	zfs_secpolicy_minor, DATASET_NAME, B_FALSE },
2918 	{ zfs_ioc_remove_minor,	zfs_secpolicy_minor, DATASET_NAME, B_FALSE },
2919 	{ zfs_ioc_create, zfs_secpolicy_create, DATASET_NAME, B_TRUE },
2920 	{ zfs_ioc_destroy, zfs_secpolicy_destroy, DATASET_NAME, B_TRUE },
2921 	{ zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, B_TRUE },
2922 	{ zfs_ioc_rename, zfs_secpolicy_rename,	DATASET_NAME, B_TRUE },
2923 	{ zfs_ioc_recv, zfs_secpolicy_receive, DATASET_NAME, B_TRUE },
2924 	{ zfs_ioc_send, zfs_secpolicy_send, DATASET_NAME, B_TRUE },
2925 	{ zfs_ioc_inject_fault,	zfs_secpolicy_inject, NO_NAME, B_FALSE },
2926 	{ zfs_ioc_clear_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE },
2927 	{ zfs_ioc_inject_list_next, zfs_secpolicy_inject, NO_NAME, B_FALSE },
2928 	{ zfs_ioc_error_log, zfs_secpolicy_inject, POOL_NAME, B_FALSE },
2929 	{ zfs_ioc_clear, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2930 	{ zfs_ioc_promote, zfs_secpolicy_promote, DATASET_NAME, B_TRUE },
2931 	{ zfs_ioc_destroy_snaps, zfs_secpolicy_destroy,	DATASET_NAME, B_TRUE },
2932 	{ zfs_ioc_snapshot, zfs_secpolicy_snapshot, DATASET_NAME, B_TRUE },
2933 	{ zfs_ioc_dsobj_to_dsname, zfs_secpolicy_config, POOL_NAME, B_FALSE },
2934 	{ zfs_ioc_obj_to_path, zfs_secpolicy_config, NO_NAME, B_FALSE },
2935 	{ zfs_ioc_pool_set_props, zfs_secpolicy_config,	POOL_NAME, B_TRUE },
2936 	{ zfs_ioc_pool_get_props, zfs_secpolicy_read, POOL_NAME, B_FALSE },
2937 	{ zfs_ioc_set_fsacl, zfs_secpolicy_fsacl, DATASET_NAME, B_TRUE },
2938 	{ zfs_ioc_get_fsacl, zfs_secpolicy_read, DATASET_NAME, B_FALSE },
2939 	{ zfs_ioc_iscsi_perm_check, zfs_secpolicy_iscsi,
2940 	    DATASET_NAME, B_FALSE },
2941 	{ zfs_ioc_share, zfs_secpolicy_share, DATASET_NAME, B_FALSE },
2942 	{ zfs_ioc_inherit_prop, zfs_secpolicy_inherit, DATASET_NAME, B_TRUE },
2943 };
2944 
2945 static int
2946 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
2947 {
2948 	zfs_cmd_t *zc;
2949 	uint_t vec;
2950 	int error, rc;
2951 
2952 	if (getminor(dev) != 0)
2953 		return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
2954 
2955 	vec = cmd - ZFS_IOC;
2956 	ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
2957 
2958 	if (vec >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
2959 		return (EINVAL);
2960 
2961 	zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2962 
2963 	error = xcopyin((void *)arg, zc, sizeof (zfs_cmd_t));
2964 
2965 	if (error == 0)
2966 		error = zfs_ioc_vec[vec].zvec_secpolicy(zc, cr);
2967 
2968 	/*
2969 	 * Ensure that all pool/dataset names are valid before we pass down to
2970 	 * the lower layers.
2971 	 */
2972 	if (error == 0) {
2973 		zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
2974 		switch (zfs_ioc_vec[vec].zvec_namecheck) {
2975 		case POOL_NAME:
2976 			if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
2977 				error = EINVAL;
2978 			break;
2979 
2980 		case DATASET_NAME:
2981 			if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
2982 				error = EINVAL;
2983 			break;
2984 
2985 		case NO_NAME:
2986 			break;
2987 		}
2988 	}
2989 
2990 	if (error == 0)
2991 		error = zfs_ioc_vec[vec].zvec_func(zc);
2992 
2993 	rc = xcopyout(zc, (void *)arg, sizeof (zfs_cmd_t));
2994 	if (error == 0) {
2995 		error = rc;
2996 		if (zfs_ioc_vec[vec].zvec_his_log == B_TRUE)
2997 			zfs_log_history(zc);
2998 	}
2999 
3000 	kmem_free(zc, sizeof (zfs_cmd_t));
3001 	return (error);
3002 }
3003 
3004 static int
3005 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
3006 {
3007 	if (cmd != DDI_ATTACH)
3008 		return (DDI_FAILURE);
3009 
3010 	if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
3011 	    DDI_PSEUDO, 0) == DDI_FAILURE)
3012 		return (DDI_FAILURE);
3013 
3014 	zfs_dip = dip;
3015 
3016 	ddi_report_dev(dip);
3017 
3018 	return (DDI_SUCCESS);
3019 }
3020 
3021 static int
3022 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
3023 {
3024 	if (spa_busy() || zfs_busy() || zvol_busy())
3025 		return (DDI_FAILURE);
3026 
3027 	if (cmd != DDI_DETACH)
3028 		return (DDI_FAILURE);
3029 
3030 	zfs_dip = NULL;
3031 
3032 	ddi_prop_remove_all(dip);
3033 	ddi_remove_minor_node(dip, NULL);
3034 
3035 	return (DDI_SUCCESS);
3036 }
3037 
3038 /*ARGSUSED*/
3039 static int
3040 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
3041 {
3042 	switch (infocmd) {
3043 	case DDI_INFO_DEVT2DEVINFO:
3044 		*result = zfs_dip;
3045 		return (DDI_SUCCESS);
3046 
3047 	case DDI_INFO_DEVT2INSTANCE:
3048 		*result = (void *)0;
3049 		return (DDI_SUCCESS);
3050 	}
3051 
3052 	return (DDI_FAILURE);
3053 }
3054 
3055 /*
3056  * OK, so this is a little weird.
3057  *
3058  * /dev/zfs is the control node, i.e. minor 0.
3059  * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
3060  *
3061  * /dev/zfs has basically nothing to do except serve up ioctls,
3062  * so most of the standard driver entry points are in zvol.c.
3063  */
3064 static struct cb_ops zfs_cb_ops = {
3065 	zvol_open,	/* open */
3066 	zvol_close,	/* close */
3067 	zvol_strategy,	/* strategy */
3068 	nodev,		/* print */
3069 	zvol_dump,	/* dump */
3070 	zvol_read,	/* read */
3071 	zvol_write,	/* write */
3072 	zfsdev_ioctl,	/* ioctl */
3073 	nodev,		/* devmap */
3074 	nodev,		/* mmap */
3075 	nodev,		/* segmap */
3076 	nochpoll,	/* poll */
3077 	ddi_prop_op,	/* prop_op */
3078 	NULL,		/* streamtab */
3079 	D_NEW | D_MP | D_64BIT,		/* Driver compatibility flag */
3080 	CB_REV,		/* version */
3081 	nodev,		/* async read */
3082 	nodev,		/* async write */
3083 };
3084 
3085 static struct dev_ops zfs_dev_ops = {
3086 	DEVO_REV,	/* version */
3087 	0,		/* refcnt */
3088 	zfs_info,	/* info */
3089 	nulldev,	/* identify */
3090 	nulldev,	/* probe */
3091 	zfs_attach,	/* attach */
3092 	zfs_detach,	/* detach */
3093 	nodev,		/* reset */
3094 	&zfs_cb_ops,	/* driver operations */
3095 	NULL,		/* no bus operations */
3096 	NULL,		/* power */
3097 	ddi_quiesce_not_needed,	/* quiesce */
3098 };
3099 
3100 static struct modldrv zfs_modldrv = {
3101 	&mod_driverops,
3102 	"ZFS storage pool",
3103 	&zfs_dev_ops
3104 };
3105 
3106 static struct modlinkage modlinkage = {
3107 	MODREV_1,
3108 	(void *)&zfs_modlfs,
3109 	(void *)&zfs_modldrv,
3110 	NULL
3111 };
3112 
3113 
3114 uint_t zfs_fsyncer_key;
3115 extern uint_t rrw_tsd_key;
3116 
3117 int
3118 _init(void)
3119 {
3120 	int error;
3121 
3122 	spa_init(FREAD | FWRITE);
3123 	zfs_init();
3124 	zvol_init();
3125 
3126 	if ((error = mod_install(&modlinkage)) != 0) {
3127 		zvol_fini();
3128 		zfs_fini();
3129 		spa_fini();
3130 		return (error);
3131 	}
3132 
3133 	tsd_create(&zfs_fsyncer_key, NULL);
3134 	tsd_create(&rrw_tsd_key, NULL);
3135 
3136 	error = ldi_ident_from_mod(&modlinkage, &zfs_li);
3137 	ASSERT(error == 0);
3138 	mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
3139 
3140 	return (0);
3141 }
3142 
3143 int
3144 _fini(void)
3145 {
3146 	int error;
3147 
3148 	if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
3149 		return (EBUSY);
3150 
3151 	if ((error = mod_remove(&modlinkage)) != 0)
3152 		return (error);
3153 
3154 	zvol_fini();
3155 	zfs_fini();
3156 	spa_fini();
3157 	if (zfs_nfsshare_inited)
3158 		(void) ddi_modclose(nfs_mod);
3159 	if (zfs_smbshare_inited)
3160 		(void) ddi_modclose(smbsrv_mod);
3161 	if (zfs_nfsshare_inited || zfs_smbshare_inited)
3162 		(void) ddi_modclose(sharefs_mod);
3163 
3164 	tsd_destroy(&zfs_fsyncer_key);
3165 	ldi_ident_release(zfs_li);
3166 	zfs_li = NULL;
3167 	mutex_destroy(&zfs_share_lock);
3168 
3169 	return (error);
3170 }
3171 
3172 int
3173 _info(struct modinfo *modinfop)
3174 {
3175 	return (mod_info(&modlinkage, modinfop));
3176 }
3177