xref: /illumos-gate/usr/src/uts/common/fs/zfs/zfs_ioctl.c (revision 379c004d1f26b343f034bba8a350290691d00d38)
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 2009 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, 0, 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 	if (zc->zc_cookie == 0) {
1326 		uint64_t cookie = 0;
1327 		int len = sizeof (zc->zc_name) - (p - zc->zc_name);
1328 
1329 		while (dmu_dir_list_next(os, len, p, NULL, &cookie) == 0)
1330 			dmu_objset_prefetch(p, NULL);
1331 	}
1332 
1333 	do {
1334 		error = dmu_dir_list_next(os,
1335 		    sizeof (zc->zc_name) - (p - zc->zc_name), p,
1336 		    NULL, &zc->zc_cookie);
1337 		if (error == ENOENT)
1338 			error = ESRCH;
1339 	} while (error == 0 && !INGLOBALZONE(curproc) &&
1340 	    !zone_dataset_visible(zc->zc_name, NULL));
1341 	dmu_objset_close(os);
1342 
1343 	/*
1344 	 * If it's a hidden dataset (ie. with a '$' in its name), don't
1345 	 * try to get stats for it.  Userland will skip over it.
1346 	 */
1347 	if (error == 0 && strchr(zc->zc_name, '$') == NULL)
1348 		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
1349 
1350 	return (error);
1351 }
1352 
1353 /*
1354  * inputs:
1355  * zc_name		name of filesystem
1356  * zc_cookie		zap cursor
1357  * zc_nvlist_dst_size	size of buffer for property nvlist
1358  *
1359  * outputs:
1360  * zc_name		name of next snapshot
1361  * zc_objset_stats	stats
1362  * zc_nvlist_dst	property nvlist
1363  * zc_nvlist_dst_size	size of property nvlist
1364  */
1365 static int
1366 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
1367 {
1368 	objset_t *os;
1369 	int error;
1370 
1371 	error = dmu_objset_open(zc->zc_name,
1372 	    DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os);
1373 	if (error)
1374 		return (error == ENOENT ? ESRCH : error);
1375 
1376 	if (zc->zc_cookie == 0)
1377 		dmu_objset_find(zc->zc_name, dmu_objset_prefetch,
1378 		    NULL, DS_FIND_SNAPSHOTS);
1379 	/*
1380 	 * A dataset name of maximum length cannot have any snapshots,
1381 	 * so exit immediately.
1382 	 */
1383 	if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
1384 		dmu_objset_close(os);
1385 		return (ESRCH);
1386 	}
1387 
1388 	error = dmu_snapshot_list_next(os,
1389 	    sizeof (zc->zc_name) - strlen(zc->zc_name),
1390 	    zc->zc_name + strlen(zc->zc_name), NULL, &zc->zc_cookie, NULL);
1391 	dmu_objset_close(os);
1392 	if (error == 0)
1393 		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
1394 	else if (error == ENOENT)
1395 		error = ESRCH;
1396 
1397 	/* if we failed, undo the @ that we tacked on to zc_name */
1398 	if (error)
1399 		*strchr(zc->zc_name, '@') = '\0';
1400 	return (error);
1401 }
1402 
1403 int
1404 zfs_set_prop_nvlist(const char *name, nvlist_t *nvl)
1405 {
1406 	nvpair_t *elem;
1407 	int error;
1408 	uint64_t intval;
1409 	char *strval;
1410 
1411 	/*
1412 	 * First validate permission to set all of the properties
1413 	 */
1414 	elem = NULL;
1415 	while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) {
1416 		const char *propname = nvpair_name(elem);
1417 		zfs_prop_t prop = zfs_name_to_prop(propname);
1418 
1419 		if (prop == ZPROP_INVAL) {
1420 			/*
1421 			 * If this is a user-defined property, it must be a
1422 			 * string, and there is no further validation to do.
1423 			 */
1424 			if (!zfs_prop_user(propname) ||
1425 			    nvpair_type(elem) != DATA_TYPE_STRING)
1426 				return (EINVAL);
1427 
1428 			if (error = zfs_secpolicy_write_perms(name,
1429 			    ZFS_DELEG_PERM_USERPROP, CRED()))
1430 				return (error);
1431 			continue;
1432 		}
1433 
1434 		if ((error = zfs_secpolicy_setprop(name, prop, CRED())) != 0)
1435 			return (error);
1436 
1437 		/*
1438 		 * Check that this value is valid for this pool version
1439 		 */
1440 		switch (prop) {
1441 		case ZFS_PROP_COMPRESSION:
1442 			/*
1443 			 * If the user specified gzip compression, make sure
1444 			 * the SPA supports it. We ignore any errors here since
1445 			 * we'll catch them later.
1446 			 */
1447 			if (nvpair_type(elem) == DATA_TYPE_UINT64 &&
1448 			    nvpair_value_uint64(elem, &intval) == 0) {
1449 				if (intval >= ZIO_COMPRESS_GZIP_1 &&
1450 				    intval <= ZIO_COMPRESS_GZIP_9 &&
1451 				    zfs_earlier_version(name,
1452 				    SPA_VERSION_GZIP_COMPRESSION))
1453 					return (ENOTSUP);
1454 
1455 				/*
1456 				 * If this is a bootable dataset then
1457 				 * verify that the compression algorithm
1458 				 * is supported for booting. We must return
1459 				 * something other than ENOTSUP since it
1460 				 * implies a downrev pool version.
1461 				 */
1462 				if (zfs_is_bootfs(name) &&
1463 				    !BOOTFS_COMPRESS_VALID(intval))
1464 					return (ERANGE);
1465 			}
1466 			break;
1467 
1468 		case ZFS_PROP_COPIES:
1469 			if (zfs_earlier_version(name,
1470 			    SPA_VERSION_DITTO_BLOCKS))
1471 				return (ENOTSUP);
1472 			break;
1473 
1474 		case ZFS_PROP_SHARESMB:
1475 			if (zpl_earlier_version(name, ZPL_VERSION_FUID))
1476 				return (ENOTSUP);
1477 			break;
1478 
1479 		case ZFS_PROP_ACLINHERIT:
1480 			if (nvpair_type(elem) == DATA_TYPE_UINT64 &&
1481 			    nvpair_value_uint64(elem, &intval) == 0)
1482 				if (intval == ZFS_ACL_PASSTHROUGH_X &&
1483 				    zfs_earlier_version(name,
1484 				    SPA_VERSION_PASSTHROUGH_X))
1485 					return (ENOTSUP);
1486 		}
1487 	}
1488 
1489 	elem = NULL;
1490 	while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) {
1491 		const char *propname = nvpair_name(elem);
1492 		zfs_prop_t prop = zfs_name_to_prop(propname);
1493 
1494 		if (prop == ZPROP_INVAL) {
1495 			VERIFY(nvpair_value_string(elem, &strval) == 0);
1496 			error = dsl_prop_set(name, propname, 1,
1497 			    strlen(strval) + 1, strval);
1498 			if (error == 0)
1499 				continue;
1500 			else
1501 				return (error);
1502 		}
1503 
1504 		switch (prop) {
1505 		case ZFS_PROP_QUOTA:
1506 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1507 			    (error = dsl_dir_set_quota(name, intval)) != 0)
1508 				return (error);
1509 			break;
1510 
1511 		case ZFS_PROP_REFQUOTA:
1512 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1513 			    (error = dsl_dataset_set_quota(name, intval)) != 0)
1514 				return (error);
1515 			break;
1516 
1517 		case ZFS_PROP_RESERVATION:
1518 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1519 			    (error = dsl_dir_set_reservation(name,
1520 			    intval)) != 0)
1521 				return (error);
1522 			break;
1523 
1524 		case ZFS_PROP_REFRESERVATION:
1525 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1526 			    (error = dsl_dataset_set_reservation(name,
1527 			    intval)) != 0)
1528 				return (error);
1529 			break;
1530 
1531 		case ZFS_PROP_VOLSIZE:
1532 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1533 			    (error = zvol_set_volsize(name,
1534 			    ddi_driver_major(zfs_dip), intval)) != 0)
1535 				return (error);
1536 			break;
1537 
1538 		case ZFS_PROP_VOLBLOCKSIZE:
1539 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1540 			    (error = zvol_set_volblocksize(name, intval)) != 0)
1541 				return (error);
1542 			break;
1543 
1544 		case ZFS_PROP_VERSION:
1545 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1546 			    (error = zfs_set_version(name, intval)) != 0)
1547 				return (error);
1548 			break;
1549 
1550 		default:
1551 			if (nvpair_type(elem) == DATA_TYPE_STRING) {
1552 				if (zfs_prop_get_type(prop) !=
1553 				    PROP_TYPE_STRING)
1554 					return (EINVAL);
1555 				VERIFY(nvpair_value_string(elem, &strval) == 0);
1556 				if ((error = dsl_prop_set(name,
1557 				    nvpair_name(elem), 1, strlen(strval) + 1,
1558 				    strval)) != 0)
1559 					return (error);
1560 			} else if (nvpair_type(elem) == DATA_TYPE_UINT64) {
1561 				const char *unused;
1562 
1563 				VERIFY(nvpair_value_uint64(elem, &intval) == 0);
1564 
1565 				switch (zfs_prop_get_type(prop)) {
1566 				case PROP_TYPE_NUMBER:
1567 					break;
1568 				case PROP_TYPE_STRING:
1569 					return (EINVAL);
1570 				case PROP_TYPE_INDEX:
1571 					if (zfs_prop_index_to_string(prop,
1572 					    intval, &unused) != 0)
1573 						return (EINVAL);
1574 					break;
1575 				default:
1576 					cmn_err(CE_PANIC,
1577 					    "unknown property type");
1578 					break;
1579 				}
1580 
1581 				if ((error = dsl_prop_set(name, propname,
1582 				    8, 1, &intval)) != 0)
1583 					return (error);
1584 			} else {
1585 				return (EINVAL);
1586 			}
1587 			break;
1588 		}
1589 	}
1590 
1591 	return (0);
1592 }
1593 
1594 /*
1595  * inputs:
1596  * zc_name		name of filesystem
1597  * zc_value		name of property to inherit
1598  * zc_nvlist_src{_size}	nvlist of properties to apply
1599  * zc_cookie		clear existing local props?
1600  *
1601  * outputs:		none
1602  */
1603 static int
1604 zfs_ioc_set_prop(zfs_cmd_t *zc)
1605 {
1606 	nvlist_t *nvl;
1607 	int error;
1608 
1609 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1610 	    &nvl)) != 0)
1611 		return (error);
1612 
1613 	if (zc->zc_cookie) {
1614 		nvlist_t *origprops;
1615 		objset_t *os;
1616 
1617 		if (dmu_objset_open(zc->zc_name, DMU_OST_ANY,
1618 		    DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
1619 			if (dsl_prop_get_all(os, &origprops, TRUE) == 0) {
1620 				clear_props(zc->zc_name, origprops);
1621 				nvlist_free(origprops);
1622 			}
1623 			dmu_objset_close(os);
1624 		}
1625 
1626 	}
1627 
1628 	error = zfs_set_prop_nvlist(zc->zc_name, nvl);
1629 
1630 	nvlist_free(nvl);
1631 	return (error);
1632 }
1633 
1634 /*
1635  * inputs:
1636  * zc_name		name of filesystem
1637  * zc_value		name of property to inherit
1638  *
1639  * outputs:		none
1640  */
1641 static int
1642 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
1643 {
1644 	/* the property name has been validated by zfs_secpolicy_inherit() */
1645 	return (dsl_prop_set(zc->zc_name, zc->zc_value, 0, 0, NULL));
1646 }
1647 
1648 static int
1649 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
1650 {
1651 	nvlist_t *props;
1652 	spa_t *spa;
1653 	int error;
1654 	nvpair_t *elem;
1655 
1656 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1657 	    &props)))
1658 		return (error);
1659 
1660 	/*
1661 	 * If the only property is the configfile, then just do a spa_lookup()
1662 	 * to handle the faulted case.
1663 	 */
1664 	elem = nvlist_next_nvpair(props, NULL);
1665 	if (elem != NULL && strcmp(nvpair_name(elem),
1666 	    zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
1667 	    nvlist_next_nvpair(props, elem) == NULL) {
1668 		mutex_enter(&spa_namespace_lock);
1669 		if ((spa = spa_lookup(zc->zc_name)) != NULL) {
1670 			spa_configfile_set(spa, props, B_FALSE);
1671 			spa_config_sync(spa, B_FALSE, B_TRUE);
1672 		}
1673 		mutex_exit(&spa_namespace_lock);
1674 		if (spa != NULL)
1675 			return (0);
1676 	}
1677 
1678 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
1679 		nvlist_free(props);
1680 		return (error);
1681 	}
1682 
1683 	error = spa_prop_set(spa, props);
1684 
1685 	nvlist_free(props);
1686 	spa_close(spa, FTAG);
1687 
1688 	return (error);
1689 }
1690 
1691 static int
1692 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
1693 {
1694 	spa_t *spa;
1695 	int error;
1696 	nvlist_t *nvp = NULL;
1697 
1698 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
1699 		/*
1700 		 * If the pool is faulted, there may be properties we can still
1701 		 * get (such as altroot and cachefile), so attempt to get them
1702 		 * anyway.
1703 		 */
1704 		mutex_enter(&spa_namespace_lock);
1705 		if ((spa = spa_lookup(zc->zc_name)) != NULL)
1706 			error = spa_prop_get(spa, &nvp);
1707 		mutex_exit(&spa_namespace_lock);
1708 	} else {
1709 		error = spa_prop_get(spa, &nvp);
1710 		spa_close(spa, FTAG);
1711 	}
1712 
1713 	if (error == 0 && zc->zc_nvlist_dst != NULL)
1714 		error = put_nvlist(zc, nvp);
1715 	else
1716 		error = EFAULT;
1717 
1718 	nvlist_free(nvp);
1719 	return (error);
1720 }
1721 
1722 static int
1723 zfs_ioc_iscsi_perm_check(zfs_cmd_t *zc)
1724 {
1725 	nvlist_t *nvp;
1726 	int error;
1727 	uint32_t uid;
1728 	uint32_t gid;
1729 	uint32_t *groups;
1730 	uint_t group_cnt;
1731 	cred_t	*usercred;
1732 
1733 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1734 	    &nvp)) != 0) {
1735 		return (error);
1736 	}
1737 
1738 	if ((error = nvlist_lookup_uint32(nvp,
1739 	    ZFS_DELEG_PERM_UID, &uid)) != 0) {
1740 		nvlist_free(nvp);
1741 		return (EPERM);
1742 	}
1743 
1744 	if ((error = nvlist_lookup_uint32(nvp,
1745 	    ZFS_DELEG_PERM_GID, &gid)) != 0) {
1746 		nvlist_free(nvp);
1747 		return (EPERM);
1748 	}
1749 
1750 	if ((error = nvlist_lookup_uint32_array(nvp, ZFS_DELEG_PERM_GROUPS,
1751 	    &groups, &group_cnt)) != 0) {
1752 		nvlist_free(nvp);
1753 		return (EPERM);
1754 	}
1755 	usercred = cralloc();
1756 	if ((crsetugid(usercred, uid, gid) != 0) ||
1757 	    (crsetgroups(usercred, group_cnt, (gid_t *)groups) != 0)) {
1758 		nvlist_free(nvp);
1759 		crfree(usercred);
1760 		return (EPERM);
1761 	}
1762 	nvlist_free(nvp);
1763 	error = dsl_deleg_access(zc->zc_name,
1764 	    zfs_prop_to_name(ZFS_PROP_SHAREISCSI), usercred);
1765 	crfree(usercred);
1766 	return (error);
1767 }
1768 
1769 /*
1770  * inputs:
1771  * zc_name		name of filesystem
1772  * zc_nvlist_src{_size}	nvlist of delegated permissions
1773  * zc_perm_action	allow/unallow flag
1774  *
1775  * outputs:		none
1776  */
1777 static int
1778 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
1779 {
1780 	int error;
1781 	nvlist_t *fsaclnv = NULL;
1782 
1783 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1784 	    &fsaclnv)) != 0)
1785 		return (error);
1786 
1787 	/*
1788 	 * Verify nvlist is constructed correctly
1789 	 */
1790 	if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
1791 		nvlist_free(fsaclnv);
1792 		return (EINVAL);
1793 	}
1794 
1795 	/*
1796 	 * If we don't have PRIV_SYS_MOUNT, then validate
1797 	 * that user is allowed to hand out each permission in
1798 	 * the nvlist(s)
1799 	 */
1800 
1801 	error = secpolicy_zfs(CRED());
1802 	if (error) {
1803 		if (zc->zc_perm_action == B_FALSE) {
1804 			error = dsl_deleg_can_allow(zc->zc_name,
1805 			    fsaclnv, CRED());
1806 		} else {
1807 			error = dsl_deleg_can_unallow(zc->zc_name,
1808 			    fsaclnv, CRED());
1809 		}
1810 	}
1811 
1812 	if (error == 0)
1813 		error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
1814 
1815 	nvlist_free(fsaclnv);
1816 	return (error);
1817 }
1818 
1819 /*
1820  * inputs:
1821  * zc_name		name of filesystem
1822  *
1823  * outputs:
1824  * zc_nvlist_src{_size}	nvlist of delegated permissions
1825  */
1826 static int
1827 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
1828 {
1829 	nvlist_t *nvp;
1830 	int error;
1831 
1832 	if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
1833 		error = put_nvlist(zc, nvp);
1834 		nvlist_free(nvp);
1835 	}
1836 
1837 	return (error);
1838 }
1839 
1840 /*
1841  * inputs:
1842  * zc_name		name of volume
1843  *
1844  * outputs:		none
1845  */
1846 static int
1847 zfs_ioc_create_minor(zfs_cmd_t *zc)
1848 {
1849 	return (zvol_create_minor(zc->zc_name, ddi_driver_major(zfs_dip)));
1850 }
1851 
1852 /*
1853  * inputs:
1854  * zc_name		name of volume
1855  *
1856  * outputs:		none
1857  */
1858 static int
1859 zfs_ioc_remove_minor(zfs_cmd_t *zc)
1860 {
1861 	return (zvol_remove_minor(zc->zc_name));
1862 }
1863 
1864 /*
1865  * Search the vfs list for a specified resource.  Returns a pointer to it
1866  * or NULL if no suitable entry is found. The caller of this routine
1867  * is responsible for releasing the returned vfs pointer.
1868  */
1869 static vfs_t *
1870 zfs_get_vfs(const char *resource)
1871 {
1872 	struct vfs *vfsp;
1873 	struct vfs *vfs_found = NULL;
1874 
1875 	vfs_list_read_lock();
1876 	vfsp = rootvfs;
1877 	do {
1878 		if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
1879 			VFS_HOLD(vfsp);
1880 			vfs_found = vfsp;
1881 			break;
1882 		}
1883 		vfsp = vfsp->vfs_next;
1884 	} while (vfsp != rootvfs);
1885 	vfs_list_unlock();
1886 	return (vfs_found);
1887 }
1888 
1889 /* ARGSUSED */
1890 static void
1891 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
1892 {
1893 	zfs_creat_t *zct = arg;
1894 
1895 	zfs_create_fs(os, cr, zct->zct_zplprops, tx);
1896 }
1897 
1898 #define	ZFS_PROP_UNDEFINED	((uint64_t)-1)
1899 
1900 /*
1901  * inputs:
1902  * createprops		list of properties requested by creator
1903  * default_zplver	zpl version to use if unspecified in createprops
1904  * fuids_ok		fuids allowed in this version of the spa?
1905  * os			parent objset pointer (NULL if root fs)
1906  *
1907  * outputs:
1908  * zplprops	values for the zplprops we attach to the master node object
1909  * is_ci	true if requested file system will be purely case-insensitive
1910  *
1911  * Determine the settings for utf8only, normalization and
1912  * casesensitivity.  Specific values may have been requested by the
1913  * creator and/or we can inherit values from the parent dataset.  If
1914  * the file system is of too early a vintage, a creator can not
1915  * request settings for these properties, even if the requested
1916  * setting is the default value.  We don't actually want to create dsl
1917  * properties for these, so remove them from the source nvlist after
1918  * processing.
1919  */
1920 static int
1921 zfs_fill_zplprops_impl(objset_t *os, uint64_t default_zplver,
1922     boolean_t fuids_ok, nvlist_t *createprops, nvlist_t *zplprops,
1923     boolean_t *is_ci)
1924 {
1925 	uint64_t zplver = default_zplver;
1926 	uint64_t sense = ZFS_PROP_UNDEFINED;
1927 	uint64_t norm = ZFS_PROP_UNDEFINED;
1928 	uint64_t u8 = ZFS_PROP_UNDEFINED;
1929 
1930 	ASSERT(zplprops != NULL);
1931 
1932 	/*
1933 	 * Pull out creator prop choices, if any.
1934 	 */
1935 	if (createprops) {
1936 		(void) nvlist_lookup_uint64(createprops,
1937 		    zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
1938 		(void) nvlist_lookup_uint64(createprops,
1939 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
1940 		(void) nvlist_remove_all(createprops,
1941 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE));
1942 		(void) nvlist_lookup_uint64(createprops,
1943 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
1944 		(void) nvlist_remove_all(createprops,
1945 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
1946 		(void) nvlist_lookup_uint64(createprops,
1947 		    zfs_prop_to_name(ZFS_PROP_CASE), &sense);
1948 		(void) nvlist_remove_all(createprops,
1949 		    zfs_prop_to_name(ZFS_PROP_CASE));
1950 	}
1951 
1952 	/*
1953 	 * If the zpl version requested is whacky or the file system
1954 	 * or pool is version is too "young" to support normalization
1955 	 * and the creator tried to set a value for one of the props,
1956 	 * error out.
1957 	 */
1958 	if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
1959 	    (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
1960 	    (zplver < ZPL_VERSION_NORMALIZATION &&
1961 	    (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
1962 	    sense != ZFS_PROP_UNDEFINED)))
1963 		return (ENOTSUP);
1964 
1965 	/*
1966 	 * Put the version in the zplprops
1967 	 */
1968 	VERIFY(nvlist_add_uint64(zplprops,
1969 	    zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
1970 
1971 	if (norm == ZFS_PROP_UNDEFINED)
1972 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
1973 	VERIFY(nvlist_add_uint64(zplprops,
1974 	    zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
1975 
1976 	/*
1977 	 * If we're normalizing, names must always be valid UTF-8 strings.
1978 	 */
1979 	if (norm)
1980 		u8 = 1;
1981 	if (u8 == ZFS_PROP_UNDEFINED)
1982 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
1983 	VERIFY(nvlist_add_uint64(zplprops,
1984 	    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
1985 
1986 	if (sense == ZFS_PROP_UNDEFINED)
1987 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
1988 	VERIFY(nvlist_add_uint64(zplprops,
1989 	    zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
1990 
1991 	if (is_ci)
1992 		*is_ci = (sense == ZFS_CASE_INSENSITIVE);
1993 
1994 	return (0);
1995 }
1996 
1997 static int
1998 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
1999     nvlist_t *zplprops, boolean_t *is_ci)
2000 {
2001 	boolean_t fuids_ok = B_TRUE;
2002 	uint64_t zplver = ZPL_VERSION;
2003 	objset_t *os = NULL;
2004 	char parentname[MAXNAMELEN];
2005 	char *cp;
2006 	int error;
2007 
2008 	(void) strlcpy(parentname, dataset, sizeof (parentname));
2009 	cp = strrchr(parentname, '/');
2010 	ASSERT(cp != NULL);
2011 	cp[0] = '\0';
2012 
2013 	if (zfs_earlier_version(dataset, SPA_VERSION_FUID)) {
2014 		zplver = ZPL_VERSION_FUID - 1;
2015 		fuids_ok = B_FALSE;
2016 	}
2017 
2018 	/*
2019 	 * Open parent object set so we can inherit zplprop values.
2020 	 */
2021 	if ((error = dmu_objset_open(parentname, DMU_OST_ANY,
2022 	    DS_MODE_USER | DS_MODE_READONLY, &os)) != 0)
2023 		return (error);
2024 
2025 	error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, createprops,
2026 	    zplprops, is_ci);
2027 	dmu_objset_close(os);
2028 	return (error);
2029 }
2030 
2031 static int
2032 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
2033     nvlist_t *zplprops, boolean_t *is_ci)
2034 {
2035 	boolean_t fuids_ok = B_TRUE;
2036 	uint64_t zplver = ZPL_VERSION;
2037 	int error;
2038 
2039 	if (spa_vers < SPA_VERSION_FUID) {
2040 		zplver = ZPL_VERSION_FUID - 1;
2041 		fuids_ok = B_FALSE;
2042 	}
2043 
2044 	error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, createprops,
2045 	    zplprops, is_ci);
2046 	return (error);
2047 }
2048 
2049 /*
2050  * inputs:
2051  * zc_objset_type	type of objset to create (fs vs zvol)
2052  * zc_name		name of new objset
2053  * zc_value		name of snapshot to clone from (may be empty)
2054  * zc_nvlist_src{_size}	nvlist of properties to apply
2055  *
2056  * outputs: none
2057  */
2058 static int
2059 zfs_ioc_create(zfs_cmd_t *zc)
2060 {
2061 	objset_t *clone;
2062 	int error = 0;
2063 	zfs_creat_t zct;
2064 	nvlist_t *nvprops = NULL;
2065 	void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
2066 	dmu_objset_type_t type = zc->zc_objset_type;
2067 
2068 	switch (type) {
2069 
2070 	case DMU_OST_ZFS:
2071 		cbfunc = zfs_create_cb;
2072 		break;
2073 
2074 	case DMU_OST_ZVOL:
2075 		cbfunc = zvol_create_cb;
2076 		break;
2077 
2078 	default:
2079 		cbfunc = NULL;
2080 		break;
2081 	}
2082 	if (strchr(zc->zc_name, '@') ||
2083 	    strchr(zc->zc_name, '%'))
2084 		return (EINVAL);
2085 
2086 	if (zc->zc_nvlist_src != NULL &&
2087 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2088 	    &nvprops)) != 0)
2089 		return (error);
2090 
2091 	zct.zct_zplprops = NULL;
2092 	zct.zct_props = nvprops;
2093 
2094 	if (zc->zc_value[0] != '\0') {
2095 		/*
2096 		 * We're creating a clone of an existing snapshot.
2097 		 */
2098 		zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
2099 		if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0) {
2100 			nvlist_free(nvprops);
2101 			return (EINVAL);
2102 		}
2103 
2104 		error = dmu_objset_open(zc->zc_value, type,
2105 		    DS_MODE_USER | DS_MODE_READONLY, &clone);
2106 		if (error) {
2107 			nvlist_free(nvprops);
2108 			return (error);
2109 		}
2110 
2111 		error = dmu_objset_create(zc->zc_name, type, clone, 0,
2112 		    NULL, NULL);
2113 		if (error) {
2114 			dmu_objset_close(clone);
2115 			nvlist_free(nvprops);
2116 			return (error);
2117 		}
2118 		dmu_objset_close(clone);
2119 	} else {
2120 		boolean_t is_insensitive = B_FALSE;
2121 
2122 		if (cbfunc == NULL) {
2123 			nvlist_free(nvprops);
2124 			return (EINVAL);
2125 		}
2126 
2127 		if (type == DMU_OST_ZVOL) {
2128 			uint64_t volsize, volblocksize;
2129 
2130 			if (nvprops == NULL ||
2131 			    nvlist_lookup_uint64(nvprops,
2132 			    zfs_prop_to_name(ZFS_PROP_VOLSIZE),
2133 			    &volsize) != 0) {
2134 				nvlist_free(nvprops);
2135 				return (EINVAL);
2136 			}
2137 
2138 			if ((error = nvlist_lookup_uint64(nvprops,
2139 			    zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
2140 			    &volblocksize)) != 0 && error != ENOENT) {
2141 				nvlist_free(nvprops);
2142 				return (EINVAL);
2143 			}
2144 
2145 			if (error != 0)
2146 				volblocksize = zfs_prop_default_numeric(
2147 				    ZFS_PROP_VOLBLOCKSIZE);
2148 
2149 			if ((error = zvol_check_volblocksize(
2150 			    volblocksize)) != 0 ||
2151 			    (error = zvol_check_volsize(volsize,
2152 			    volblocksize)) != 0) {
2153 				nvlist_free(nvprops);
2154 				return (error);
2155 			}
2156 		} else if (type == DMU_OST_ZFS) {
2157 			int error;
2158 
2159 			/*
2160 			 * We have to have normalization and
2161 			 * case-folding flags correct when we do the
2162 			 * file system creation, so go figure them out
2163 			 * now.
2164 			 */
2165 			VERIFY(nvlist_alloc(&zct.zct_zplprops,
2166 			    NV_UNIQUE_NAME, KM_SLEEP) == 0);
2167 			error = zfs_fill_zplprops(zc->zc_name, nvprops,
2168 			    zct.zct_zplprops, &is_insensitive);
2169 			if (error != 0) {
2170 				nvlist_free(nvprops);
2171 				nvlist_free(zct.zct_zplprops);
2172 				return (error);
2173 			}
2174 		}
2175 		error = dmu_objset_create(zc->zc_name, type, NULL,
2176 		    is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
2177 		nvlist_free(zct.zct_zplprops);
2178 	}
2179 
2180 	/*
2181 	 * It would be nice to do this atomically.
2182 	 */
2183 	if (error == 0) {
2184 		if ((error = zfs_set_prop_nvlist(zc->zc_name, nvprops)) != 0)
2185 			(void) dmu_objset_destroy(zc->zc_name);
2186 	}
2187 	nvlist_free(nvprops);
2188 	return (error);
2189 }
2190 
2191 struct snap_prop_arg {
2192 	nvlist_t *nvprops;
2193 	const char *snapname;
2194 };
2195 
2196 static int
2197 set_snap_props(char *name, void *arg)
2198 {
2199 	struct snap_prop_arg *snpa = arg;
2200 	int len = strlen(name) + strlen(snpa->snapname) + 2;
2201 	char *buf = kmem_alloc(len, KM_SLEEP);
2202 	int err;
2203 
2204 	(void) snprintf(buf, len, "%s@%s", name, snpa->snapname);
2205 	err = zfs_set_prop_nvlist(buf, snpa->nvprops);
2206 	if (err)
2207 		(void) dmu_objset_destroy(buf);
2208 	kmem_free(buf, len);
2209 	return (err);
2210 }
2211 
2212 /*
2213  * inputs:
2214  * zc_name	name of filesystem
2215  * zc_value	short name of snapshot
2216  * zc_cookie	recursive flag
2217  *
2218  * outputs:	none
2219  */
2220 static int
2221 zfs_ioc_snapshot(zfs_cmd_t *zc)
2222 {
2223 	nvlist_t *nvprops = NULL;
2224 	int error;
2225 	boolean_t recursive = zc->zc_cookie;
2226 
2227 	if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
2228 		return (EINVAL);
2229 
2230 	if (zc->zc_nvlist_src != NULL &&
2231 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2232 	    &nvprops)) != 0)
2233 		return (error);
2234 
2235 	error = dmu_objset_snapshot(zc->zc_name, zc->zc_value, recursive);
2236 
2237 	/*
2238 	 * It would be nice to do this atomically.
2239 	 */
2240 	if (error == 0) {
2241 		struct snap_prop_arg snpa;
2242 		snpa.nvprops = nvprops;
2243 		snpa.snapname = zc->zc_value;
2244 		if (recursive) {
2245 			error = dmu_objset_find(zc->zc_name,
2246 			    set_snap_props, &snpa, DS_FIND_CHILDREN);
2247 			if (error) {
2248 				(void) dmu_snapshots_destroy(zc->zc_name,
2249 				    zc->zc_value);
2250 			}
2251 		} else {
2252 			error = set_snap_props(zc->zc_name, &snpa);
2253 		}
2254 	}
2255 	nvlist_free(nvprops);
2256 	return (error);
2257 }
2258 
2259 int
2260 zfs_unmount_snap(char *name, void *arg)
2261 {
2262 	vfs_t *vfsp = NULL;
2263 
2264 	if (arg) {
2265 		char *snapname = arg;
2266 		int len = strlen(name) + strlen(snapname) + 2;
2267 		char *buf = kmem_alloc(len, KM_SLEEP);
2268 
2269 		(void) strcpy(buf, name);
2270 		(void) strcat(buf, "@");
2271 		(void) strcat(buf, snapname);
2272 		vfsp = zfs_get_vfs(buf);
2273 		kmem_free(buf, len);
2274 	} else if (strchr(name, '@')) {
2275 		vfsp = zfs_get_vfs(name);
2276 	}
2277 
2278 	if (vfsp) {
2279 		/*
2280 		 * Always force the unmount for snapshots.
2281 		 */
2282 		int flag = MS_FORCE;
2283 		int err;
2284 
2285 		if ((err = vn_vfswlock(vfsp->vfs_vnodecovered)) != 0) {
2286 			VFS_RELE(vfsp);
2287 			return (err);
2288 		}
2289 		VFS_RELE(vfsp);
2290 		if ((err = dounmount(vfsp, flag, kcred)) != 0)
2291 			return (err);
2292 	}
2293 	return (0);
2294 }
2295 
2296 /*
2297  * inputs:
2298  * zc_name	name of filesystem
2299  * zc_value	short name of snapshot
2300  *
2301  * outputs:	none
2302  */
2303 static int
2304 zfs_ioc_destroy_snaps(zfs_cmd_t *zc)
2305 {
2306 	int err;
2307 
2308 	if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
2309 		return (EINVAL);
2310 	err = dmu_objset_find(zc->zc_name,
2311 	    zfs_unmount_snap, zc->zc_value, DS_FIND_CHILDREN);
2312 	if (err)
2313 		return (err);
2314 	return (dmu_snapshots_destroy(zc->zc_name, zc->zc_value));
2315 }
2316 
2317 /*
2318  * inputs:
2319  * zc_name		name of dataset to destroy
2320  * zc_objset_type	type of objset
2321  *
2322  * outputs:		none
2323  */
2324 static int
2325 zfs_ioc_destroy(zfs_cmd_t *zc)
2326 {
2327 	if (strchr(zc->zc_name, '@') && zc->zc_objset_type == DMU_OST_ZFS) {
2328 		int err = zfs_unmount_snap(zc->zc_name, NULL);
2329 		if (err)
2330 			return (err);
2331 	}
2332 
2333 	return (dmu_objset_destroy(zc->zc_name));
2334 }
2335 
2336 /*
2337  * inputs:
2338  * zc_name	name of dataset to rollback (to most recent snapshot)
2339  *
2340  * outputs:	none
2341  */
2342 static int
2343 zfs_ioc_rollback(zfs_cmd_t *zc)
2344 {
2345 	objset_t *os;
2346 	int error;
2347 	zfsvfs_t *zfsvfs = NULL;
2348 
2349 	/*
2350 	 * Get the zfsvfs for the receiving objset. There
2351 	 * won't be one if we're operating on a zvol, if the
2352 	 * objset doesn't exist yet, or is not mounted.
2353 	 */
2354 	error = dmu_objset_open(zc->zc_name, DMU_OST_ANY, DS_MODE_USER, &os);
2355 	if (error)
2356 		return (error);
2357 
2358 	if (dmu_objset_type(os) == DMU_OST_ZFS) {
2359 		mutex_enter(&os->os->os_user_ptr_lock);
2360 		zfsvfs = dmu_objset_get_user(os);
2361 		if (zfsvfs != NULL)
2362 			VFS_HOLD(zfsvfs->z_vfs);
2363 		mutex_exit(&os->os->os_user_ptr_lock);
2364 	}
2365 
2366 	if (zfsvfs != NULL) {
2367 		char *osname;
2368 		int mode;
2369 
2370 		osname = kmem_alloc(MAXNAMELEN, KM_SLEEP);
2371 		error = zfs_suspend_fs(zfsvfs, osname, &mode);
2372 		if (error == 0) {
2373 			int resume_err;
2374 
2375 			ASSERT(strcmp(osname, zc->zc_name) == 0);
2376 			error = dmu_objset_rollback(os);
2377 			resume_err = zfs_resume_fs(zfsvfs, osname, mode);
2378 			error = error ? error : resume_err;
2379 		} else {
2380 			dmu_objset_close(os);
2381 		}
2382 		kmem_free(osname, MAXNAMELEN);
2383 		VFS_RELE(zfsvfs->z_vfs);
2384 	} else {
2385 		error = dmu_objset_rollback(os);
2386 	}
2387 	/* Note, the dmu_objset_rollback() releases the objset for us. */
2388 
2389 	return (error);
2390 }
2391 
2392 /*
2393  * inputs:
2394  * zc_name	old name of dataset
2395  * zc_value	new name of dataset
2396  * zc_cookie	recursive flag (only valid for snapshots)
2397  *
2398  * outputs:	none
2399  */
2400 static int
2401 zfs_ioc_rename(zfs_cmd_t *zc)
2402 {
2403 	boolean_t recursive = zc->zc_cookie & 1;
2404 
2405 	zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
2406 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
2407 	    strchr(zc->zc_value, '%'))
2408 		return (EINVAL);
2409 
2410 	/*
2411 	 * Unmount snapshot unless we're doing a recursive rename,
2412 	 * in which case the dataset code figures out which snapshots
2413 	 * to unmount.
2414 	 */
2415 	if (!recursive && strchr(zc->zc_name, '@') != NULL &&
2416 	    zc->zc_objset_type == DMU_OST_ZFS) {
2417 		int err = zfs_unmount_snap(zc->zc_name, NULL);
2418 		if (err)
2419 			return (err);
2420 	}
2421 	return (dmu_objset_rename(zc->zc_name, zc->zc_value, recursive));
2422 }
2423 
2424 static void
2425 clear_props(char *dataset, nvlist_t *props)
2426 {
2427 	zfs_cmd_t *zc;
2428 	nvpair_t *prop;
2429 
2430 	if (props == NULL)
2431 		return;
2432 	zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
2433 	(void) strcpy(zc->zc_name, dataset);
2434 	for (prop = nvlist_next_nvpair(props, NULL); prop;
2435 	    prop = nvlist_next_nvpair(props, prop)) {
2436 		(void) strcpy(zc->zc_value, nvpair_name(prop));
2437 		if (zfs_secpolicy_inherit(zc, CRED()) == 0)
2438 			(void) zfs_ioc_inherit_prop(zc);
2439 	}
2440 	kmem_free(zc, sizeof (zfs_cmd_t));
2441 }
2442 
2443 /*
2444  * inputs:
2445  * zc_name		name of containing filesystem
2446  * zc_nvlist_src{_size}	nvlist of properties to apply
2447  * zc_value		name of snapshot to create
2448  * zc_string		name of clone origin (if DRR_FLAG_CLONE)
2449  * zc_cookie		file descriptor to recv from
2450  * zc_begin_record	the BEGIN record of the stream (not byteswapped)
2451  * zc_guid		force flag
2452  *
2453  * outputs:
2454  * zc_cookie		number of bytes read
2455  */
2456 static int
2457 zfs_ioc_recv(zfs_cmd_t *zc)
2458 {
2459 	file_t *fp;
2460 	objset_t *os;
2461 	dmu_recv_cookie_t drc;
2462 	zfsvfs_t *zfsvfs = NULL;
2463 	boolean_t force = (boolean_t)zc->zc_guid;
2464 	int error, fd;
2465 	offset_t off;
2466 	nvlist_t *props = NULL;
2467 	nvlist_t *origprops = NULL;
2468 	objset_t *origin = NULL;
2469 	char *tosnap;
2470 	char tofs[ZFS_MAXNAMELEN];
2471 
2472 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
2473 	    strchr(zc->zc_value, '@') == NULL ||
2474 	    strchr(zc->zc_value, '%'))
2475 		return (EINVAL);
2476 
2477 	(void) strcpy(tofs, zc->zc_value);
2478 	tosnap = strchr(tofs, '@');
2479 	*tosnap = '\0';
2480 	tosnap++;
2481 
2482 	if (zc->zc_nvlist_src != NULL &&
2483 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2484 	    &props)) != 0)
2485 		return (error);
2486 
2487 	fd = zc->zc_cookie;
2488 	fp = getf(fd);
2489 	if (fp == NULL) {
2490 		nvlist_free(props);
2491 		return (EBADF);
2492 	}
2493 
2494 	if (dmu_objset_open(tofs, DMU_OST_ANY,
2495 	    DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
2496 		/*
2497 		 * Try to get the zfsvfs for the receiving objset.
2498 		 * There won't be one if we're operating on a zvol,
2499 		 * if the objset doesn't exist yet, or is not mounted.
2500 		 */
2501 		mutex_enter(&os->os->os_user_ptr_lock);
2502 		if (zfsvfs = dmu_objset_get_user(os)) {
2503 			if (!mutex_tryenter(&zfsvfs->z_online_recv_lock)) {
2504 				mutex_exit(&os->os->os_user_ptr_lock);
2505 				dmu_objset_close(os);
2506 				zfsvfs = NULL;
2507 				error = EBUSY;
2508 				goto out;
2509 			}
2510 			VFS_HOLD(zfsvfs->z_vfs);
2511 		}
2512 		mutex_exit(&os->os->os_user_ptr_lock);
2513 
2514 		/*
2515 		 * If new properties are supplied, they are to completely
2516 		 * replace the existing ones, so stash away the existing ones.
2517 		 */
2518 		if (props)
2519 			(void) dsl_prop_get_all(os, &origprops, TRUE);
2520 
2521 		dmu_objset_close(os);
2522 	}
2523 
2524 	if (zc->zc_string[0]) {
2525 		error = dmu_objset_open(zc->zc_string, DMU_OST_ANY,
2526 		    DS_MODE_USER | DS_MODE_READONLY, &origin);
2527 		if (error)
2528 			goto out;
2529 	}
2530 
2531 	error = dmu_recv_begin(tofs, tosnap, &zc->zc_begin_record,
2532 	    force, origin, zfsvfs != NULL, &drc);
2533 	if (origin)
2534 		dmu_objset_close(origin);
2535 	if (error)
2536 		goto out;
2537 
2538 	/*
2539 	 * Reset properties.  We do this before we receive the stream
2540 	 * so that the properties are applied to the new data.
2541 	 */
2542 	if (props) {
2543 		clear_props(tofs, origprops);
2544 		/*
2545 		 * XXX - Note, this is all-or-nothing; should be best-effort.
2546 		 */
2547 		(void) zfs_set_prop_nvlist(tofs, props);
2548 	}
2549 
2550 	off = fp->f_offset;
2551 	error = dmu_recv_stream(&drc, fp->f_vnode, &off);
2552 
2553 	if (error == 0 && zfsvfs) {
2554 		char *osname;
2555 		int mode;
2556 
2557 		/* online recv */
2558 		osname = kmem_alloc(MAXNAMELEN, KM_SLEEP);
2559 		error = zfs_suspend_fs(zfsvfs, osname, &mode);
2560 		if (error == 0) {
2561 			int resume_err;
2562 
2563 			error = dmu_recv_end(&drc);
2564 			resume_err = zfs_resume_fs(zfsvfs, osname, mode);
2565 			error = error ? error : resume_err;
2566 		} else {
2567 			dmu_recv_abort_cleanup(&drc);
2568 		}
2569 		kmem_free(osname, MAXNAMELEN);
2570 	} else if (error == 0) {
2571 		error = dmu_recv_end(&drc);
2572 	}
2573 
2574 	zc->zc_cookie = off - fp->f_offset;
2575 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
2576 		fp->f_offset = off;
2577 
2578 	/*
2579 	 * On error, restore the original props.
2580 	 */
2581 	if (error && props) {
2582 		clear_props(tofs, props);
2583 		(void) zfs_set_prop_nvlist(tofs, origprops);
2584 	}
2585 out:
2586 	if (zfsvfs) {
2587 		mutex_exit(&zfsvfs->z_online_recv_lock);
2588 		VFS_RELE(zfsvfs->z_vfs);
2589 	}
2590 	nvlist_free(props);
2591 	nvlist_free(origprops);
2592 	releasef(fd);
2593 	return (error);
2594 }
2595 
2596 /*
2597  * inputs:
2598  * zc_name	name of snapshot to send
2599  * zc_value	short name of incremental fromsnap (may be empty)
2600  * zc_cookie	file descriptor to send stream to
2601  * zc_obj	fromorigin flag (mutually exclusive with zc_value)
2602  *
2603  * outputs: none
2604  */
2605 static int
2606 zfs_ioc_send(zfs_cmd_t *zc)
2607 {
2608 	objset_t *fromsnap = NULL;
2609 	objset_t *tosnap;
2610 	file_t *fp;
2611 	int error;
2612 	offset_t off;
2613 
2614 	error = dmu_objset_open(zc->zc_name, DMU_OST_ANY,
2615 	    DS_MODE_USER | DS_MODE_READONLY, &tosnap);
2616 	if (error)
2617 		return (error);
2618 
2619 	if (zc->zc_value[0] != '\0') {
2620 		char *buf;
2621 		char *cp;
2622 
2623 		buf = kmem_alloc(MAXPATHLEN, KM_SLEEP);
2624 		(void) strncpy(buf, zc->zc_name, MAXPATHLEN);
2625 		cp = strchr(buf, '@');
2626 		if (cp)
2627 			*(cp+1) = 0;
2628 		(void) strncat(buf, zc->zc_value, MAXPATHLEN);
2629 		error = dmu_objset_open(buf, DMU_OST_ANY,
2630 		    DS_MODE_USER | DS_MODE_READONLY, &fromsnap);
2631 		kmem_free(buf, MAXPATHLEN);
2632 		if (error) {
2633 			dmu_objset_close(tosnap);
2634 			return (error);
2635 		}
2636 	}
2637 
2638 	fp = getf(zc->zc_cookie);
2639 	if (fp == NULL) {
2640 		dmu_objset_close(tosnap);
2641 		if (fromsnap)
2642 			dmu_objset_close(fromsnap);
2643 		return (EBADF);
2644 	}
2645 
2646 	off = fp->f_offset;
2647 	error = dmu_sendbackup(tosnap, fromsnap, zc->zc_obj, fp->f_vnode, &off);
2648 
2649 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
2650 		fp->f_offset = off;
2651 	releasef(zc->zc_cookie);
2652 	if (fromsnap)
2653 		dmu_objset_close(fromsnap);
2654 	dmu_objset_close(tosnap);
2655 	return (error);
2656 }
2657 
2658 static int
2659 zfs_ioc_inject_fault(zfs_cmd_t *zc)
2660 {
2661 	int id, error;
2662 
2663 	error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
2664 	    &zc->zc_inject_record);
2665 
2666 	if (error == 0)
2667 		zc->zc_guid = (uint64_t)id;
2668 
2669 	return (error);
2670 }
2671 
2672 static int
2673 zfs_ioc_clear_fault(zfs_cmd_t *zc)
2674 {
2675 	return (zio_clear_fault((int)zc->zc_guid));
2676 }
2677 
2678 static int
2679 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
2680 {
2681 	int id = (int)zc->zc_guid;
2682 	int error;
2683 
2684 	error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
2685 	    &zc->zc_inject_record);
2686 
2687 	zc->zc_guid = id;
2688 
2689 	return (error);
2690 }
2691 
2692 static int
2693 zfs_ioc_error_log(zfs_cmd_t *zc)
2694 {
2695 	spa_t *spa;
2696 	int error;
2697 	size_t count = (size_t)zc->zc_nvlist_dst_size;
2698 
2699 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2700 		return (error);
2701 
2702 	error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
2703 	    &count);
2704 	if (error == 0)
2705 		zc->zc_nvlist_dst_size = count;
2706 	else
2707 		zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
2708 
2709 	spa_close(spa, FTAG);
2710 
2711 	return (error);
2712 }
2713 
2714 static int
2715 zfs_ioc_clear(zfs_cmd_t *zc)
2716 {
2717 	spa_t *spa;
2718 	vdev_t *vd;
2719 	int error;
2720 
2721 	/*
2722 	 * On zpool clear we also fix up missing slogs
2723 	 */
2724 	mutex_enter(&spa_namespace_lock);
2725 	spa = spa_lookup(zc->zc_name);
2726 	if (spa == NULL) {
2727 		mutex_exit(&spa_namespace_lock);
2728 		return (EIO);
2729 	}
2730 	if (spa->spa_log_state == SPA_LOG_MISSING) {
2731 		/* we need to let spa_open/spa_load clear the chains */
2732 		spa->spa_log_state = SPA_LOG_CLEAR;
2733 	}
2734 	mutex_exit(&spa_namespace_lock);
2735 
2736 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2737 		return (error);
2738 
2739 	spa_vdev_state_enter(spa);
2740 
2741 	if (zc->zc_guid == 0) {
2742 		vd = NULL;
2743 	} else {
2744 		vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
2745 		if (vd == NULL) {
2746 			(void) spa_vdev_state_exit(spa, NULL, ENODEV);
2747 			spa_close(spa, FTAG);
2748 			return (ENODEV);
2749 		}
2750 	}
2751 
2752 	vdev_clear(spa, vd);
2753 
2754 	(void) spa_vdev_state_exit(spa, NULL, 0);
2755 
2756 	/*
2757 	 * Resume any suspended I/Os.
2758 	 */
2759 	zio_resume(spa);
2760 
2761 	spa_close(spa, FTAG);
2762 
2763 	return (0);
2764 }
2765 
2766 /*
2767  * inputs:
2768  * zc_name	name of filesystem
2769  * zc_value	name of origin snapshot
2770  *
2771  * outputs:	none
2772  */
2773 static int
2774 zfs_ioc_promote(zfs_cmd_t *zc)
2775 {
2776 	char *cp;
2777 
2778 	/*
2779 	 * We don't need to unmount *all* the origin fs's snapshots, but
2780 	 * it's easier.
2781 	 */
2782 	cp = strchr(zc->zc_value, '@');
2783 	if (cp)
2784 		*cp = '\0';
2785 	(void) dmu_objset_find(zc->zc_value,
2786 	    zfs_unmount_snap, NULL, DS_FIND_SNAPSHOTS);
2787 	return (dsl_dataset_promote(zc->zc_name));
2788 }
2789 
2790 /*
2791  * We don't want to have a hard dependency
2792  * against some special symbols in sharefs
2793  * nfs, and smbsrv.  Determine them if needed when
2794  * the first file system is shared.
2795  * Neither sharefs, nfs or smbsrv are unloadable modules.
2796  */
2797 int (*znfsexport_fs)(void *arg);
2798 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
2799 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
2800 
2801 int zfs_nfsshare_inited;
2802 int zfs_smbshare_inited;
2803 
2804 ddi_modhandle_t nfs_mod;
2805 ddi_modhandle_t sharefs_mod;
2806 ddi_modhandle_t smbsrv_mod;
2807 kmutex_t zfs_share_lock;
2808 
2809 static int
2810 zfs_init_sharefs()
2811 {
2812 	int error;
2813 
2814 	ASSERT(MUTEX_HELD(&zfs_share_lock));
2815 	/* Both NFS and SMB shares also require sharetab support. */
2816 	if (sharefs_mod == NULL && ((sharefs_mod =
2817 	    ddi_modopen("fs/sharefs",
2818 	    KRTLD_MODE_FIRST, &error)) == NULL)) {
2819 		return (ENOSYS);
2820 	}
2821 	if (zshare_fs == NULL && ((zshare_fs =
2822 	    (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
2823 	    ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
2824 		return (ENOSYS);
2825 	}
2826 	return (0);
2827 }
2828 
2829 static int
2830 zfs_ioc_share(zfs_cmd_t *zc)
2831 {
2832 	int error;
2833 	int opcode;
2834 
2835 	switch (zc->zc_share.z_sharetype) {
2836 	case ZFS_SHARE_NFS:
2837 	case ZFS_UNSHARE_NFS:
2838 		if (zfs_nfsshare_inited == 0) {
2839 			mutex_enter(&zfs_share_lock);
2840 			if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
2841 			    KRTLD_MODE_FIRST, &error)) == NULL)) {
2842 				mutex_exit(&zfs_share_lock);
2843 				return (ENOSYS);
2844 			}
2845 			if (znfsexport_fs == NULL &&
2846 			    ((znfsexport_fs = (int (*)(void *))
2847 			    ddi_modsym(nfs_mod,
2848 			    "nfs_export", &error)) == NULL)) {
2849 				mutex_exit(&zfs_share_lock);
2850 				return (ENOSYS);
2851 			}
2852 			error = zfs_init_sharefs();
2853 			if (error) {
2854 				mutex_exit(&zfs_share_lock);
2855 				return (ENOSYS);
2856 			}
2857 			zfs_nfsshare_inited = 1;
2858 			mutex_exit(&zfs_share_lock);
2859 		}
2860 		break;
2861 	case ZFS_SHARE_SMB:
2862 	case ZFS_UNSHARE_SMB:
2863 		if (zfs_smbshare_inited == 0) {
2864 			mutex_enter(&zfs_share_lock);
2865 			if (smbsrv_mod == NULL && ((smbsrv_mod =
2866 			    ddi_modopen("drv/smbsrv",
2867 			    KRTLD_MODE_FIRST, &error)) == NULL)) {
2868 				mutex_exit(&zfs_share_lock);
2869 				return (ENOSYS);
2870 			}
2871 			if (zsmbexport_fs == NULL && ((zsmbexport_fs =
2872 			    (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
2873 			    "smb_server_share", &error)) == NULL)) {
2874 				mutex_exit(&zfs_share_lock);
2875 				return (ENOSYS);
2876 			}
2877 			error = zfs_init_sharefs();
2878 			if (error) {
2879 				mutex_exit(&zfs_share_lock);
2880 				return (ENOSYS);
2881 			}
2882 			zfs_smbshare_inited = 1;
2883 			mutex_exit(&zfs_share_lock);
2884 		}
2885 		break;
2886 	default:
2887 		return (EINVAL);
2888 	}
2889 
2890 	switch (zc->zc_share.z_sharetype) {
2891 	case ZFS_SHARE_NFS:
2892 	case ZFS_UNSHARE_NFS:
2893 		if (error =
2894 		    znfsexport_fs((void *)
2895 		    (uintptr_t)zc->zc_share.z_exportdata))
2896 			return (error);
2897 		break;
2898 	case ZFS_SHARE_SMB:
2899 	case ZFS_UNSHARE_SMB:
2900 		if (error = zsmbexport_fs((void *)
2901 		    (uintptr_t)zc->zc_share.z_exportdata,
2902 		    zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
2903 		    B_TRUE : B_FALSE)) {
2904 			return (error);
2905 		}
2906 		break;
2907 	}
2908 
2909 	opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
2910 	    zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
2911 	    SHAREFS_ADD : SHAREFS_REMOVE;
2912 
2913 	/*
2914 	 * Add or remove share from sharetab
2915 	 */
2916 	error = zshare_fs(opcode,
2917 	    (void *)(uintptr_t)zc->zc_share.z_sharedata,
2918 	    zc->zc_share.z_sharemax);
2919 
2920 	return (error);
2921 
2922 }
2923 
2924 /*
2925  * pool create, destroy, and export don't log the history as part of
2926  * zfsdev_ioctl, but rather zfs_ioc_pool_create, and zfs_ioc_pool_export
2927  * do the logging of those commands.
2928  */
2929 static zfs_ioc_vec_t zfs_ioc_vec[] = {
2930 	{ zfs_ioc_pool_create, zfs_secpolicy_config, POOL_NAME, B_FALSE },
2931 	{ zfs_ioc_pool_destroy,	zfs_secpolicy_config, POOL_NAME, B_FALSE },
2932 	{ zfs_ioc_pool_import, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2933 	{ zfs_ioc_pool_export, zfs_secpolicy_config, POOL_NAME, B_FALSE },
2934 	{ zfs_ioc_pool_configs,	zfs_secpolicy_none, NO_NAME, B_FALSE },
2935 	{ zfs_ioc_pool_stats, zfs_secpolicy_read, POOL_NAME, B_FALSE },
2936 	{ zfs_ioc_pool_tryimport, zfs_secpolicy_config, NO_NAME, B_FALSE },
2937 	{ zfs_ioc_pool_scrub, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2938 	{ zfs_ioc_pool_freeze, zfs_secpolicy_config, NO_NAME, B_FALSE },
2939 	{ zfs_ioc_pool_upgrade,	zfs_secpolicy_config, POOL_NAME, B_TRUE },
2940 	{ zfs_ioc_pool_get_history, zfs_secpolicy_config, POOL_NAME, B_FALSE },
2941 	{ zfs_ioc_vdev_add, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2942 	{ zfs_ioc_vdev_remove, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2943 	{ zfs_ioc_vdev_set_state, zfs_secpolicy_config,	POOL_NAME, B_TRUE },
2944 	{ zfs_ioc_vdev_attach, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2945 	{ zfs_ioc_vdev_detach, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2946 	{ zfs_ioc_vdev_setpath,	zfs_secpolicy_config, POOL_NAME, B_FALSE },
2947 	{ zfs_ioc_objset_stats,	zfs_secpolicy_read, DATASET_NAME, B_FALSE },
2948 	{ zfs_ioc_objset_zplprops, zfs_secpolicy_read, DATASET_NAME, B_FALSE },
2949 	{ zfs_ioc_dataset_list_next, zfs_secpolicy_read,
2950 	    DATASET_NAME, B_FALSE },
2951 	{ zfs_ioc_snapshot_list_next, zfs_secpolicy_read,
2952 	    DATASET_NAME, B_FALSE },
2953 	{ zfs_ioc_set_prop, zfs_secpolicy_none, DATASET_NAME, B_TRUE },
2954 	{ zfs_ioc_create_minor,	zfs_secpolicy_minor, DATASET_NAME, B_FALSE },
2955 	{ zfs_ioc_remove_minor,	zfs_secpolicy_minor, DATASET_NAME, B_FALSE },
2956 	{ zfs_ioc_create, zfs_secpolicy_create, DATASET_NAME, B_TRUE },
2957 	{ zfs_ioc_destroy, zfs_secpolicy_destroy, DATASET_NAME, B_TRUE },
2958 	{ zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, B_TRUE },
2959 	{ zfs_ioc_rename, zfs_secpolicy_rename,	DATASET_NAME, B_TRUE },
2960 	{ zfs_ioc_recv, zfs_secpolicy_receive, DATASET_NAME, B_TRUE },
2961 	{ zfs_ioc_send, zfs_secpolicy_send, DATASET_NAME, B_TRUE },
2962 	{ zfs_ioc_inject_fault,	zfs_secpolicy_inject, NO_NAME, B_FALSE },
2963 	{ zfs_ioc_clear_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE },
2964 	{ zfs_ioc_inject_list_next, zfs_secpolicy_inject, NO_NAME, B_FALSE },
2965 	{ zfs_ioc_error_log, zfs_secpolicy_inject, POOL_NAME, B_FALSE },
2966 	{ zfs_ioc_clear, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2967 	{ zfs_ioc_promote, zfs_secpolicy_promote, DATASET_NAME, B_TRUE },
2968 	{ zfs_ioc_destroy_snaps, zfs_secpolicy_destroy,	DATASET_NAME, B_TRUE },
2969 	{ zfs_ioc_snapshot, zfs_secpolicy_snapshot, DATASET_NAME, B_TRUE },
2970 	{ zfs_ioc_dsobj_to_dsname, zfs_secpolicy_config, POOL_NAME, B_FALSE },
2971 	{ zfs_ioc_obj_to_path, zfs_secpolicy_config, NO_NAME, B_FALSE },
2972 	{ zfs_ioc_pool_set_props, zfs_secpolicy_config,	POOL_NAME, B_TRUE },
2973 	{ zfs_ioc_pool_get_props, zfs_secpolicy_read, POOL_NAME, B_FALSE },
2974 	{ zfs_ioc_set_fsacl, zfs_secpolicy_fsacl, DATASET_NAME, B_TRUE },
2975 	{ zfs_ioc_get_fsacl, zfs_secpolicy_read, DATASET_NAME, B_FALSE },
2976 	{ zfs_ioc_iscsi_perm_check, zfs_secpolicy_iscsi,
2977 	    DATASET_NAME, B_FALSE },
2978 	{ zfs_ioc_share, zfs_secpolicy_share, DATASET_NAME, B_FALSE },
2979 	{ zfs_ioc_inherit_prop, zfs_secpolicy_inherit, DATASET_NAME, B_TRUE },
2980 };
2981 
2982 static int
2983 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
2984 {
2985 	zfs_cmd_t *zc;
2986 	uint_t vec;
2987 	int error, rc;
2988 
2989 	if (getminor(dev) != 0)
2990 		return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
2991 
2992 	vec = cmd - ZFS_IOC;
2993 	ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
2994 
2995 	if (vec >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
2996 		return (EINVAL);
2997 
2998 	zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2999 
3000 	error = xcopyin((void *)arg, zc, sizeof (zfs_cmd_t));
3001 
3002 	if (error == 0)
3003 		error = zfs_ioc_vec[vec].zvec_secpolicy(zc, cr);
3004 
3005 	/*
3006 	 * Ensure that all pool/dataset names are valid before we pass down to
3007 	 * the lower layers.
3008 	 */
3009 	if (error == 0) {
3010 		zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
3011 		switch (zfs_ioc_vec[vec].zvec_namecheck) {
3012 		case POOL_NAME:
3013 			if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
3014 				error = EINVAL;
3015 			break;
3016 
3017 		case DATASET_NAME:
3018 			if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
3019 				error = EINVAL;
3020 			break;
3021 
3022 		case NO_NAME:
3023 			break;
3024 		}
3025 	}
3026 
3027 	if (error == 0)
3028 		error = zfs_ioc_vec[vec].zvec_func(zc);
3029 
3030 	rc = xcopyout(zc, (void *)arg, sizeof (zfs_cmd_t));
3031 	if (error == 0) {
3032 		error = rc;
3033 		if (zfs_ioc_vec[vec].zvec_his_log == B_TRUE)
3034 			zfs_log_history(zc);
3035 	}
3036 
3037 	kmem_free(zc, sizeof (zfs_cmd_t));
3038 	return (error);
3039 }
3040 
3041 static int
3042 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
3043 {
3044 	if (cmd != DDI_ATTACH)
3045 		return (DDI_FAILURE);
3046 
3047 	if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
3048 	    DDI_PSEUDO, 0) == DDI_FAILURE)
3049 		return (DDI_FAILURE);
3050 
3051 	zfs_dip = dip;
3052 
3053 	ddi_report_dev(dip);
3054 
3055 	return (DDI_SUCCESS);
3056 }
3057 
3058 static int
3059 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
3060 {
3061 	if (spa_busy() || zfs_busy() || zvol_busy())
3062 		return (DDI_FAILURE);
3063 
3064 	if (cmd != DDI_DETACH)
3065 		return (DDI_FAILURE);
3066 
3067 	zfs_dip = NULL;
3068 
3069 	ddi_prop_remove_all(dip);
3070 	ddi_remove_minor_node(dip, NULL);
3071 
3072 	return (DDI_SUCCESS);
3073 }
3074 
3075 /*ARGSUSED*/
3076 static int
3077 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
3078 {
3079 	switch (infocmd) {
3080 	case DDI_INFO_DEVT2DEVINFO:
3081 		*result = zfs_dip;
3082 		return (DDI_SUCCESS);
3083 
3084 	case DDI_INFO_DEVT2INSTANCE:
3085 		*result = (void *)0;
3086 		return (DDI_SUCCESS);
3087 	}
3088 
3089 	return (DDI_FAILURE);
3090 }
3091 
3092 /*
3093  * OK, so this is a little weird.
3094  *
3095  * /dev/zfs is the control node, i.e. minor 0.
3096  * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
3097  *
3098  * /dev/zfs has basically nothing to do except serve up ioctls,
3099  * so most of the standard driver entry points are in zvol.c.
3100  */
3101 static struct cb_ops zfs_cb_ops = {
3102 	zvol_open,	/* open */
3103 	zvol_close,	/* close */
3104 	zvol_strategy,	/* strategy */
3105 	nodev,		/* print */
3106 	zvol_dump,	/* dump */
3107 	zvol_read,	/* read */
3108 	zvol_write,	/* write */
3109 	zfsdev_ioctl,	/* ioctl */
3110 	nodev,		/* devmap */
3111 	nodev,		/* mmap */
3112 	nodev,		/* segmap */
3113 	nochpoll,	/* poll */
3114 	ddi_prop_op,	/* prop_op */
3115 	NULL,		/* streamtab */
3116 	D_NEW | D_MP | D_64BIT,		/* Driver compatibility flag */
3117 	CB_REV,		/* version */
3118 	nodev,		/* async read */
3119 	nodev,		/* async write */
3120 };
3121 
3122 static struct dev_ops zfs_dev_ops = {
3123 	DEVO_REV,	/* version */
3124 	0,		/* refcnt */
3125 	zfs_info,	/* info */
3126 	nulldev,	/* identify */
3127 	nulldev,	/* probe */
3128 	zfs_attach,	/* attach */
3129 	zfs_detach,	/* detach */
3130 	nodev,		/* reset */
3131 	&zfs_cb_ops,	/* driver operations */
3132 	NULL,		/* no bus operations */
3133 	NULL,		/* power */
3134 	ddi_quiesce_not_needed,	/* quiesce */
3135 };
3136 
3137 static struct modldrv zfs_modldrv = {
3138 	&mod_driverops,
3139 	"ZFS storage pool",
3140 	&zfs_dev_ops
3141 };
3142 
3143 static struct modlinkage modlinkage = {
3144 	MODREV_1,
3145 	(void *)&zfs_modlfs,
3146 	(void *)&zfs_modldrv,
3147 	NULL
3148 };
3149 
3150 
3151 uint_t zfs_fsyncer_key;
3152 extern uint_t rrw_tsd_key;
3153 
3154 int
3155 _init(void)
3156 {
3157 	int error;
3158 
3159 	spa_init(FREAD | FWRITE);
3160 	zfs_init();
3161 	zvol_init();
3162 
3163 	if ((error = mod_install(&modlinkage)) != 0) {
3164 		zvol_fini();
3165 		zfs_fini();
3166 		spa_fini();
3167 		return (error);
3168 	}
3169 
3170 	tsd_create(&zfs_fsyncer_key, NULL);
3171 	tsd_create(&rrw_tsd_key, NULL);
3172 
3173 	error = ldi_ident_from_mod(&modlinkage, &zfs_li);
3174 	ASSERT(error == 0);
3175 	mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
3176 
3177 	return (0);
3178 }
3179 
3180 int
3181 _fini(void)
3182 {
3183 	int error;
3184 
3185 	if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
3186 		return (EBUSY);
3187 
3188 	if ((error = mod_remove(&modlinkage)) != 0)
3189 		return (error);
3190 
3191 	zvol_fini();
3192 	zfs_fini();
3193 	spa_fini();
3194 	if (zfs_nfsshare_inited)
3195 		(void) ddi_modclose(nfs_mod);
3196 	if (zfs_smbshare_inited)
3197 		(void) ddi_modclose(smbsrv_mod);
3198 	if (zfs_nfsshare_inited || zfs_smbshare_inited)
3199 		(void) ddi_modclose(sharefs_mod);
3200 
3201 	tsd_destroy(&zfs_fsyncer_key);
3202 	ldi_ident_release(zfs_li);
3203 	zfs_li = NULL;
3204 	mutex_destroy(&zfs_share_lock);
3205 
3206 	return (error);
3207 }
3208 
3209 int
3210 _info(struct modinfo *modinfop)
3211 {
3212 	return (mod_info(&modlinkage, modinfop));
3213 }
3214