xref: /titanic_51/usr/src/uts/common/fs/zfs/zfs_ioctl.c (revision 81d9f076db88c1f40c85831ce1ebb444a209c5a8)
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 2010 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_vfsops.h>
40 #include <sys/zfs_znode.h>
41 #include <sys/zap.h>
42 #include <sys/spa.h>
43 #include <sys/spa_impl.h>
44 #include <sys/vdev.h>
45 #include <sys/priv_impl.h>
46 #include <sys/dmu.h>
47 #include <sys/dsl_dir.h>
48 #include <sys/dsl_dataset.h>
49 #include <sys/dsl_prop.h>
50 #include <sys/dsl_deleg.h>
51 #include <sys/dmu_objset.h>
52 #include <sys/ddi.h>
53 #include <sys/sunddi.h>
54 #include <sys/sunldi.h>
55 #include <sys/policy.h>
56 #include <sys/zone.h>
57 #include <sys/nvpair.h>
58 #include <sys/pathname.h>
59 #include <sys/mount.h>
60 #include <sys/sdt.h>
61 #include <sys/fs/zfs.h>
62 #include <sys/zfs_ctldir.h>
63 #include <sys/zfs_dir.h>
64 #include <sys/zvol.h>
65 #include <sharefs/share.h>
66 #include <sys/dmu_objset.h>
67 
68 #include "zfs_namecheck.h"
69 #include "zfs_prop.h"
70 #include "zfs_deleg.h"
71 
72 extern struct modlfs zfs_modlfs;
73 
74 extern void zfs_init(void);
75 extern void zfs_fini(void);
76 
77 ldi_ident_t zfs_li = NULL;
78 dev_info_t *zfs_dip;
79 
80 typedef int zfs_ioc_func_t(zfs_cmd_t *);
81 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, cred_t *);
82 
83 typedef enum {
84 	NO_NAME,
85 	POOL_NAME,
86 	DATASET_NAME
87 } zfs_ioc_namecheck_t;
88 
89 typedef struct zfs_ioc_vec {
90 	zfs_ioc_func_t		*zvec_func;
91 	zfs_secpolicy_func_t	*zvec_secpolicy;
92 	zfs_ioc_namecheck_t	zvec_namecheck;
93 	boolean_t		zvec_his_log;
94 	boolean_t		zvec_pool_check;
95 } zfs_ioc_vec_t;
96 
97 /* This array is indexed by zfs_userquota_prop_t */
98 static const char *userquota_perms[] = {
99 	ZFS_DELEG_PERM_USERUSED,
100 	ZFS_DELEG_PERM_USERQUOTA,
101 	ZFS_DELEG_PERM_GROUPUSED,
102 	ZFS_DELEG_PERM_GROUPQUOTA,
103 };
104 
105 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
106 static int zfs_check_settable(const char *name, nvpair_t *property,
107     cred_t *cr);
108 static int zfs_check_clearable(char *dataset, nvlist_t *props,
109     nvlist_t **errors);
110 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
111     boolean_t *);
112 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t **);
113 
114 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
115 void
116 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
117 {
118 	const char *newfile;
119 	char buf[256];
120 	va_list adx;
121 
122 	/*
123 	 * Get rid of annoying "../common/" prefix to filename.
124 	 */
125 	newfile = strrchr(file, '/');
126 	if (newfile != NULL) {
127 		newfile = newfile + 1; /* Get rid of leading / */
128 	} else {
129 		newfile = file;
130 	}
131 
132 	va_start(adx, fmt);
133 	(void) vsnprintf(buf, sizeof (buf), fmt, adx);
134 	va_end(adx);
135 
136 	/*
137 	 * To get this data, use the zfs-dprintf probe as so:
138 	 * dtrace -q -n 'zfs-dprintf \
139 	 *	/stringof(arg0) == "dbuf.c"/ \
140 	 *	{printf("%s: %s", stringof(arg1), stringof(arg3))}'
141 	 * arg0 = file name
142 	 * arg1 = function name
143 	 * arg2 = line number
144 	 * arg3 = message
145 	 */
146 	DTRACE_PROBE4(zfs__dprintf,
147 	    char *, newfile, char *, func, int, line, char *, buf);
148 }
149 
150 static void
151 history_str_free(char *buf)
152 {
153 	kmem_free(buf, HIS_MAX_RECORD_LEN);
154 }
155 
156 static char *
157 history_str_get(zfs_cmd_t *zc)
158 {
159 	char *buf;
160 
161 	if (zc->zc_history == NULL)
162 		return (NULL);
163 
164 	buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
165 	if (copyinstr((void *)(uintptr_t)zc->zc_history,
166 	    buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
167 		history_str_free(buf);
168 		return (NULL);
169 	}
170 
171 	buf[HIS_MAX_RECORD_LEN -1] = '\0';
172 
173 	return (buf);
174 }
175 
176 /*
177  * Check to see if the named dataset is currently defined as bootable
178  */
179 static boolean_t
180 zfs_is_bootfs(const char *name)
181 {
182 	objset_t *os;
183 
184 	if (dmu_objset_hold(name, FTAG, &os) == 0) {
185 		boolean_t ret;
186 		ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
187 		dmu_objset_rele(os, FTAG);
188 		return (ret);
189 	}
190 	return (B_FALSE);
191 }
192 
193 /*
194  * zfs_earlier_version
195  *
196  *	Return non-zero if the spa version is less than requested version.
197  */
198 static int
199 zfs_earlier_version(const char *name, int version)
200 {
201 	spa_t *spa;
202 
203 	if (spa_open(name, &spa, FTAG) == 0) {
204 		if (spa_version(spa) < version) {
205 			spa_close(spa, FTAG);
206 			return (1);
207 		}
208 		spa_close(spa, FTAG);
209 	}
210 	return (0);
211 }
212 
213 /*
214  * zpl_earlier_version
215  *
216  * Return TRUE if the ZPL version is less than requested version.
217  */
218 static boolean_t
219 zpl_earlier_version(const char *name, int version)
220 {
221 	objset_t *os;
222 	boolean_t rc = B_TRUE;
223 
224 	if (dmu_objset_hold(name, FTAG, &os) == 0) {
225 		uint64_t zplversion;
226 
227 		if (dmu_objset_type(os) != DMU_OST_ZFS) {
228 			dmu_objset_rele(os, FTAG);
229 			return (B_TRUE);
230 		}
231 		/* XXX reading from non-owned objset */
232 		if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
233 			rc = zplversion < version;
234 		dmu_objset_rele(os, FTAG);
235 	}
236 	return (rc);
237 }
238 
239 static void
240 zfs_log_history(zfs_cmd_t *zc)
241 {
242 	spa_t *spa;
243 	char *buf;
244 
245 	if ((buf = history_str_get(zc)) == NULL)
246 		return;
247 
248 	if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
249 		if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
250 			(void) spa_history_log(spa, buf, LOG_CMD_NORMAL);
251 		spa_close(spa, FTAG);
252 	}
253 	history_str_free(buf);
254 }
255 
256 /*
257  * Policy for top-level read operations (list pools).  Requires no privileges,
258  * and can be used in the local zone, as there is no associated dataset.
259  */
260 /* ARGSUSED */
261 static int
262 zfs_secpolicy_none(zfs_cmd_t *zc, cred_t *cr)
263 {
264 	return (0);
265 }
266 
267 /*
268  * Policy for dataset read operations (list children, get statistics).  Requires
269  * no privileges, but must be visible in the local zone.
270  */
271 /* ARGSUSED */
272 static int
273 zfs_secpolicy_read(zfs_cmd_t *zc, cred_t *cr)
274 {
275 	if (INGLOBALZONE(curproc) ||
276 	    zone_dataset_visible(zc->zc_name, NULL))
277 		return (0);
278 
279 	return (ENOENT);
280 }
281 
282 static int
283 zfs_dozonecheck(const char *dataset, cred_t *cr)
284 {
285 	uint64_t zoned;
286 	int writable = 1;
287 
288 	/*
289 	 * The dataset must be visible by this zone -- check this first
290 	 * so they don't see EPERM on something they shouldn't know about.
291 	 */
292 	if (!INGLOBALZONE(curproc) &&
293 	    !zone_dataset_visible(dataset, &writable))
294 		return (ENOENT);
295 
296 	if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
297 		return (ENOENT);
298 
299 	if (INGLOBALZONE(curproc)) {
300 		/*
301 		 * If the fs is zoned, only root can access it from the
302 		 * global zone.
303 		 */
304 		if (secpolicy_zfs(cr) && zoned)
305 			return (EPERM);
306 	} else {
307 		/*
308 		 * If we are in a local zone, the 'zoned' property must be set.
309 		 */
310 		if (!zoned)
311 			return (EPERM);
312 
313 		/* must be writable by this zone */
314 		if (!writable)
315 			return (EPERM);
316 	}
317 	return (0);
318 }
319 
320 int
321 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
322 {
323 	int error;
324 
325 	error = zfs_dozonecheck(name, cr);
326 	if (error == 0) {
327 		error = secpolicy_zfs(cr);
328 		if (error)
329 			error = dsl_deleg_access(name, perm, cr);
330 	}
331 	return (error);
332 }
333 
334 /*
335  * Policy for setting the security label property.
336  *
337  * Returns 0 for success, non-zero for access and other errors.
338  */
339 static int
340 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
341 {
342 	char		ds_hexsl[MAXNAMELEN];
343 	bslabel_t	ds_sl, new_sl;
344 	boolean_t	new_default = FALSE;
345 	uint64_t	zoned;
346 	int		needed_priv = -1;
347 	int		error;
348 
349 	/* First get the existing dataset label. */
350 	error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
351 	    1, sizeof (ds_hexsl), &ds_hexsl, NULL);
352 	if (error)
353 		return (EPERM);
354 
355 	if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
356 		new_default = TRUE;
357 
358 	/* The label must be translatable */
359 	if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
360 		return (EINVAL);
361 
362 	/*
363 	 * In a non-global zone, disallow attempts to set a label that
364 	 * doesn't match that of the zone; otherwise no other checks
365 	 * are needed.
366 	 */
367 	if (!INGLOBALZONE(curproc)) {
368 		if (new_default || !blequal(&new_sl, CR_SL(CRED())))
369 			return (EPERM);
370 		return (0);
371 	}
372 
373 	/*
374 	 * For global-zone datasets (i.e., those whose zoned property is
375 	 * "off", verify that the specified new label is valid for the
376 	 * global zone.
377 	 */
378 	if (dsl_prop_get_integer(name,
379 	    zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
380 		return (EPERM);
381 	if (!zoned) {
382 		if (zfs_check_global_label(name, strval) != 0)
383 			return (EPERM);
384 	}
385 
386 	/*
387 	 * If the existing dataset label is nondefault, check if the
388 	 * dataset is mounted (label cannot be changed while mounted).
389 	 * Get the zfsvfs; if there isn't one, then the dataset isn't
390 	 * mounted (or isn't a dataset, doesn't exist, ...).
391 	 */
392 	if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
393 		objset_t *os;
394 		static char *setsl_tag = "setsl_tag";
395 
396 		/*
397 		 * Try to own the dataset; abort if there is any error,
398 		 * (e.g., already mounted, in use, or other error).
399 		 */
400 		error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE,
401 		    setsl_tag, &os);
402 		if (error)
403 			return (EPERM);
404 
405 		dmu_objset_disown(os, setsl_tag);
406 
407 		if (new_default) {
408 			needed_priv = PRIV_FILE_DOWNGRADE_SL;
409 			goto out_check;
410 		}
411 
412 		if (hexstr_to_label(strval, &new_sl) != 0)
413 			return (EPERM);
414 
415 		if (blstrictdom(&ds_sl, &new_sl))
416 			needed_priv = PRIV_FILE_DOWNGRADE_SL;
417 		else if (blstrictdom(&new_sl, &ds_sl))
418 			needed_priv = PRIV_FILE_UPGRADE_SL;
419 	} else {
420 		/* dataset currently has a default label */
421 		if (!new_default)
422 			needed_priv = PRIV_FILE_UPGRADE_SL;
423 	}
424 
425 out_check:
426 	if (needed_priv != -1)
427 		return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
428 	return (0);
429 }
430 
431 static int
432 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
433     cred_t *cr)
434 {
435 	char *strval;
436 
437 	/*
438 	 * Check permissions for special properties.
439 	 */
440 	switch (prop) {
441 	case ZFS_PROP_ZONED:
442 		/*
443 		 * Disallow setting of 'zoned' from within a local zone.
444 		 */
445 		if (!INGLOBALZONE(curproc))
446 			return (EPERM);
447 		break;
448 
449 	case ZFS_PROP_QUOTA:
450 		if (!INGLOBALZONE(curproc)) {
451 			uint64_t zoned;
452 			char setpoint[MAXNAMELEN];
453 			/*
454 			 * Unprivileged users are allowed to modify the
455 			 * quota on things *under* (ie. contained by)
456 			 * the thing they own.
457 			 */
458 			if (dsl_prop_get_integer(dsname, "zoned", &zoned,
459 			    setpoint))
460 				return (EPERM);
461 			if (!zoned || strlen(dsname) <= strlen(setpoint))
462 				return (EPERM);
463 		}
464 		break;
465 
466 	case ZFS_PROP_MLSLABEL:
467 		if (!is_system_labeled())
468 			return (EPERM);
469 
470 		if (nvpair_value_string(propval, &strval) == 0) {
471 			int err;
472 
473 			err = zfs_set_slabel_policy(dsname, strval, CRED());
474 			if (err != 0)
475 				return (err);
476 		}
477 		break;
478 	}
479 
480 	return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
481 }
482 
483 int
484 zfs_secpolicy_fsacl(zfs_cmd_t *zc, cred_t *cr)
485 {
486 	int error;
487 
488 	error = zfs_dozonecheck(zc->zc_name, cr);
489 	if (error)
490 		return (error);
491 
492 	/*
493 	 * permission to set permissions will be evaluated later in
494 	 * dsl_deleg_can_allow()
495 	 */
496 	return (0);
497 }
498 
499 int
500 zfs_secpolicy_rollback(zfs_cmd_t *zc, cred_t *cr)
501 {
502 	return (zfs_secpolicy_write_perms(zc->zc_name,
503 	    ZFS_DELEG_PERM_ROLLBACK, cr));
504 }
505 
506 int
507 zfs_secpolicy_send(zfs_cmd_t *zc, cred_t *cr)
508 {
509 	return (zfs_secpolicy_write_perms(zc->zc_name,
510 	    ZFS_DELEG_PERM_SEND, cr));
511 }
512 
513 static int
514 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, cred_t *cr)
515 {
516 	vnode_t *vp;
517 	int error;
518 
519 	if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
520 	    NO_FOLLOW, NULL, &vp)) != 0)
521 		return (error);
522 
523 	/* Now make sure mntpnt and dataset are ZFS */
524 
525 	if (vp->v_vfsp->vfs_fstype != zfsfstype ||
526 	    (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
527 	    zc->zc_name) != 0)) {
528 		VN_RELE(vp);
529 		return (EPERM);
530 	}
531 
532 	VN_RELE(vp);
533 	return (dsl_deleg_access(zc->zc_name,
534 	    ZFS_DELEG_PERM_SHARE, cr));
535 }
536 
537 int
538 zfs_secpolicy_share(zfs_cmd_t *zc, cred_t *cr)
539 {
540 	if (!INGLOBALZONE(curproc))
541 		return (EPERM);
542 
543 	if (secpolicy_nfs(cr) == 0) {
544 		return (0);
545 	} else {
546 		return (zfs_secpolicy_deleg_share(zc, cr));
547 	}
548 }
549 
550 int
551 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, cred_t *cr)
552 {
553 	if (!INGLOBALZONE(curproc))
554 		return (EPERM);
555 
556 	if (secpolicy_smb(cr) == 0) {
557 		return (0);
558 	} else {
559 		return (zfs_secpolicy_deleg_share(zc, cr));
560 	}
561 }
562 
563 static int
564 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
565 {
566 	char *cp;
567 
568 	/*
569 	 * Remove the @bla or /bla from the end of the name to get the parent.
570 	 */
571 	(void) strncpy(parent, datasetname, parentsize);
572 	cp = strrchr(parent, '@');
573 	if (cp != NULL) {
574 		cp[0] = '\0';
575 	} else {
576 		cp = strrchr(parent, '/');
577 		if (cp == NULL)
578 			return (ENOENT);
579 		cp[0] = '\0';
580 	}
581 
582 	return (0);
583 }
584 
585 int
586 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
587 {
588 	int error;
589 
590 	if ((error = zfs_secpolicy_write_perms(name,
591 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
592 		return (error);
593 
594 	return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
595 }
596 
597 static int
598 zfs_secpolicy_destroy(zfs_cmd_t *zc, cred_t *cr)
599 {
600 	return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
601 }
602 
603 /*
604  * Destroying snapshots with delegated permissions requires
605  * descendent mount and destroy permissions.
606  * Reassemble the full filesystem@snap name so dsl_deleg_access()
607  * can do the correct permission check.
608  *
609  * Since this routine is used when doing a recursive destroy of snapshots
610  * and destroying snapshots requires descendent permissions, a successfull
611  * check of the top level snapshot applies to snapshots of all descendent
612  * datasets as well.
613  */
614 static int
615 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, cred_t *cr)
616 {
617 	int error;
618 	char *dsname;
619 
620 	dsname = kmem_asprintf("%s@%s", zc->zc_name, zc->zc_value);
621 
622 	error = zfs_secpolicy_destroy_perms(dsname, cr);
623 
624 	strfree(dsname);
625 	return (error);
626 }
627 
628 /*
629  * Must have sys_config privilege to check the iscsi permission
630  */
631 /* ARGSUSED */
632 static int
633 zfs_secpolicy_iscsi(zfs_cmd_t *zc, cred_t *cr)
634 {
635 	return (secpolicy_zfs(cr));
636 }
637 
638 int
639 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
640 {
641 	char	parentname[MAXNAMELEN];
642 	int	error;
643 
644 	if ((error = zfs_secpolicy_write_perms(from,
645 	    ZFS_DELEG_PERM_RENAME, cr)) != 0)
646 		return (error);
647 
648 	if ((error = zfs_secpolicy_write_perms(from,
649 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
650 		return (error);
651 
652 	if ((error = zfs_get_parent(to, parentname,
653 	    sizeof (parentname))) != 0)
654 		return (error);
655 
656 	if ((error = zfs_secpolicy_write_perms(parentname,
657 	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
658 		return (error);
659 
660 	if ((error = zfs_secpolicy_write_perms(parentname,
661 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
662 		return (error);
663 
664 	return (error);
665 }
666 
667 static int
668 zfs_secpolicy_rename(zfs_cmd_t *zc, cred_t *cr)
669 {
670 	return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
671 }
672 
673 static int
674 zfs_secpolicy_promote(zfs_cmd_t *zc, cred_t *cr)
675 {
676 	char	parentname[MAXNAMELEN];
677 	objset_t *clone;
678 	int error;
679 
680 	error = zfs_secpolicy_write_perms(zc->zc_name,
681 	    ZFS_DELEG_PERM_PROMOTE, cr);
682 	if (error)
683 		return (error);
684 
685 	error = dmu_objset_hold(zc->zc_name, FTAG, &clone);
686 
687 	if (error == 0) {
688 		dsl_dataset_t *pclone = NULL;
689 		dsl_dir_t *dd;
690 		dd = clone->os_dsl_dataset->ds_dir;
691 
692 		rw_enter(&dd->dd_pool->dp_config_rwlock, RW_READER);
693 		error = dsl_dataset_hold_obj(dd->dd_pool,
694 		    dd->dd_phys->dd_origin_obj, FTAG, &pclone);
695 		rw_exit(&dd->dd_pool->dp_config_rwlock);
696 		if (error) {
697 			dmu_objset_rele(clone, FTAG);
698 			return (error);
699 		}
700 
701 		error = zfs_secpolicy_write_perms(zc->zc_name,
702 		    ZFS_DELEG_PERM_MOUNT, cr);
703 
704 		dsl_dataset_name(pclone, parentname);
705 		dmu_objset_rele(clone, FTAG);
706 		dsl_dataset_rele(pclone, FTAG);
707 		if (error == 0)
708 			error = zfs_secpolicy_write_perms(parentname,
709 			    ZFS_DELEG_PERM_PROMOTE, cr);
710 	}
711 	return (error);
712 }
713 
714 static int
715 zfs_secpolicy_receive(zfs_cmd_t *zc, cred_t *cr)
716 {
717 	int error;
718 
719 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
720 	    ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
721 		return (error);
722 
723 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
724 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
725 		return (error);
726 
727 	return (zfs_secpolicy_write_perms(zc->zc_name,
728 	    ZFS_DELEG_PERM_CREATE, cr));
729 }
730 
731 int
732 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
733 {
734 	return (zfs_secpolicy_write_perms(name,
735 	    ZFS_DELEG_PERM_SNAPSHOT, cr));
736 }
737 
738 static int
739 zfs_secpolicy_snapshot(zfs_cmd_t *zc, cred_t *cr)
740 {
741 
742 	return (zfs_secpolicy_snapshot_perms(zc->zc_name, cr));
743 }
744 
745 static int
746 zfs_secpolicy_create(zfs_cmd_t *zc, cred_t *cr)
747 {
748 	char	parentname[MAXNAMELEN];
749 	int	error;
750 
751 	if ((error = zfs_get_parent(zc->zc_name, parentname,
752 	    sizeof (parentname))) != 0)
753 		return (error);
754 
755 	if (zc->zc_value[0] != '\0') {
756 		if ((error = zfs_secpolicy_write_perms(zc->zc_value,
757 		    ZFS_DELEG_PERM_CLONE, cr)) != 0)
758 			return (error);
759 	}
760 
761 	if ((error = zfs_secpolicy_write_perms(parentname,
762 	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
763 		return (error);
764 
765 	error = zfs_secpolicy_write_perms(parentname,
766 	    ZFS_DELEG_PERM_MOUNT, cr);
767 
768 	return (error);
769 }
770 
771 static int
772 zfs_secpolicy_umount(zfs_cmd_t *zc, cred_t *cr)
773 {
774 	int error;
775 
776 	error = secpolicy_fs_unmount(cr, NULL);
777 	if (error) {
778 		error = dsl_deleg_access(zc->zc_name, ZFS_DELEG_PERM_MOUNT, cr);
779 	}
780 	return (error);
781 }
782 
783 /*
784  * Policy for pool operations - create/destroy pools, add vdevs, etc.  Requires
785  * SYS_CONFIG privilege, which is not available in a local zone.
786  */
787 /* ARGSUSED */
788 static int
789 zfs_secpolicy_config(zfs_cmd_t *zc, cred_t *cr)
790 {
791 	if (secpolicy_sys_config(cr, B_FALSE) != 0)
792 		return (EPERM);
793 
794 	return (0);
795 }
796 
797 /*
798  * Policy for fault injection.  Requires all privileges.
799  */
800 /* ARGSUSED */
801 static int
802 zfs_secpolicy_inject(zfs_cmd_t *zc, cred_t *cr)
803 {
804 	return (secpolicy_zinject(cr));
805 }
806 
807 static int
808 zfs_secpolicy_inherit(zfs_cmd_t *zc, cred_t *cr)
809 {
810 	zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
811 
812 	if (prop == ZPROP_INVAL) {
813 		if (!zfs_prop_user(zc->zc_value))
814 			return (EINVAL);
815 		return (zfs_secpolicy_write_perms(zc->zc_name,
816 		    ZFS_DELEG_PERM_USERPROP, cr));
817 	} else {
818 		return (zfs_secpolicy_setprop(zc->zc_name, prop,
819 		    NULL, cr));
820 	}
821 }
822 
823 static int
824 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, cred_t *cr)
825 {
826 	int err = zfs_secpolicy_read(zc, cr);
827 	if (err)
828 		return (err);
829 
830 	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
831 		return (EINVAL);
832 
833 	if (zc->zc_value[0] == 0) {
834 		/*
835 		 * They are asking about a posix uid/gid.  If it's
836 		 * themself, allow it.
837 		 */
838 		if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
839 		    zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
840 			if (zc->zc_guid == crgetuid(cr))
841 				return (0);
842 		} else {
843 			if (groupmember(zc->zc_guid, cr))
844 				return (0);
845 		}
846 	}
847 
848 	return (zfs_secpolicy_write_perms(zc->zc_name,
849 	    userquota_perms[zc->zc_objset_type], cr));
850 }
851 
852 static int
853 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, cred_t *cr)
854 {
855 	int err = zfs_secpolicy_read(zc, cr);
856 	if (err)
857 		return (err);
858 
859 	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
860 		return (EINVAL);
861 
862 	return (zfs_secpolicy_write_perms(zc->zc_name,
863 	    userquota_perms[zc->zc_objset_type], cr));
864 }
865 
866 static int
867 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, cred_t *cr)
868 {
869 	return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
870 	    NULL, cr));
871 }
872 
873 static int
874 zfs_secpolicy_hold(zfs_cmd_t *zc, cred_t *cr)
875 {
876 	return (zfs_secpolicy_write_perms(zc->zc_name,
877 	    ZFS_DELEG_PERM_HOLD, cr));
878 }
879 
880 static int
881 zfs_secpolicy_release(zfs_cmd_t *zc, cred_t *cr)
882 {
883 	return (zfs_secpolicy_write_perms(zc->zc_name,
884 	    ZFS_DELEG_PERM_RELEASE, cr));
885 }
886 
887 /*
888  * Returns the nvlist as specified by the user in the zfs_cmd_t.
889  */
890 static int
891 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
892 {
893 	char *packed;
894 	int error;
895 	nvlist_t *list = NULL;
896 
897 	/*
898 	 * Read in and unpack the user-supplied nvlist.
899 	 */
900 	if (size == 0)
901 		return (EINVAL);
902 
903 	packed = kmem_alloc(size, KM_SLEEP);
904 
905 	if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
906 	    iflag)) != 0) {
907 		kmem_free(packed, size);
908 		return (error);
909 	}
910 
911 	if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
912 		kmem_free(packed, size);
913 		return (error);
914 	}
915 
916 	kmem_free(packed, size);
917 
918 	*nvp = list;
919 	return (0);
920 }
921 
922 static int
923 fit_error_list(zfs_cmd_t *zc, nvlist_t **errors)
924 {
925 	size_t size;
926 
927 	VERIFY(nvlist_size(*errors, &size, NV_ENCODE_NATIVE) == 0);
928 
929 	if (size > zc->zc_nvlist_dst_size) {
930 		nvpair_t *more_errors;
931 		int n = 0;
932 
933 		if (zc->zc_nvlist_dst_size < 1024)
934 			return (ENOMEM);
935 
936 		VERIFY(nvlist_add_int32(*errors, ZPROP_N_MORE_ERRORS, 0) == 0);
937 		more_errors = nvlist_prev_nvpair(*errors, NULL);
938 
939 		do {
940 			nvpair_t *pair = nvlist_prev_nvpair(*errors,
941 			    more_errors);
942 			VERIFY(nvlist_remove_nvpair(*errors, pair) == 0);
943 			n++;
944 			VERIFY(nvlist_size(*errors, &size,
945 			    NV_ENCODE_NATIVE) == 0);
946 		} while (size > zc->zc_nvlist_dst_size);
947 
948 		VERIFY(nvlist_remove_nvpair(*errors, more_errors) == 0);
949 		VERIFY(nvlist_add_int32(*errors, ZPROP_N_MORE_ERRORS, n) == 0);
950 		ASSERT(nvlist_size(*errors, &size, NV_ENCODE_NATIVE) == 0);
951 		ASSERT(size <= zc->zc_nvlist_dst_size);
952 	}
953 
954 	return (0);
955 }
956 
957 static int
958 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
959 {
960 	char *packed = NULL;
961 	size_t size;
962 	int error;
963 
964 	VERIFY(nvlist_size(nvl, &size, NV_ENCODE_NATIVE) == 0);
965 
966 	if (size > zc->zc_nvlist_dst_size) {
967 		error = ENOMEM;
968 	} else {
969 		packed = kmem_alloc(size, KM_SLEEP);
970 		VERIFY(nvlist_pack(nvl, &packed, &size, NV_ENCODE_NATIVE,
971 		    KM_SLEEP) == 0);
972 		error = ddi_copyout(packed,
973 		    (void *)(uintptr_t)zc->zc_nvlist_dst, size, zc->zc_iflags);
974 		kmem_free(packed, size);
975 	}
976 
977 	zc->zc_nvlist_dst_size = size;
978 	return (error);
979 }
980 
981 static int
982 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
983 {
984 	objset_t *os;
985 	int error;
986 
987 	error = dmu_objset_hold(dsname, FTAG, &os);
988 	if (error)
989 		return (error);
990 	if (dmu_objset_type(os) != DMU_OST_ZFS) {
991 		dmu_objset_rele(os, FTAG);
992 		return (EINVAL);
993 	}
994 
995 	mutex_enter(&os->os_user_ptr_lock);
996 	*zfvp = dmu_objset_get_user(os);
997 	if (*zfvp) {
998 		VFS_HOLD((*zfvp)->z_vfs);
999 	} else {
1000 		error = ESRCH;
1001 	}
1002 	mutex_exit(&os->os_user_ptr_lock);
1003 	dmu_objset_rele(os, FTAG);
1004 	return (error);
1005 }
1006 
1007 /*
1008  * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1009  * case its z_vfs will be NULL, and it will be opened as the owner.
1010  */
1011 static int
1012 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp)
1013 {
1014 	int error = 0;
1015 
1016 	if (getzfsvfs(name, zfvp) != 0)
1017 		error = zfsvfs_create(name, zfvp);
1018 	if (error == 0) {
1019 		rrw_enter(&(*zfvp)->z_teardown_lock, RW_READER, tag);
1020 		if ((*zfvp)->z_unmounted) {
1021 			/*
1022 			 * XXX we could probably try again, since the unmounting
1023 			 * thread should be just about to disassociate the
1024 			 * objset from the zfsvfs.
1025 			 */
1026 			rrw_exit(&(*zfvp)->z_teardown_lock, tag);
1027 			return (EBUSY);
1028 		}
1029 	}
1030 	return (error);
1031 }
1032 
1033 static void
1034 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1035 {
1036 	rrw_exit(&zfsvfs->z_teardown_lock, tag);
1037 
1038 	if (zfsvfs->z_vfs) {
1039 		VFS_RELE(zfsvfs->z_vfs);
1040 	} else {
1041 		dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1042 		zfsvfs_free(zfsvfs);
1043 	}
1044 }
1045 
1046 static int
1047 zfs_ioc_pool_create(zfs_cmd_t *zc)
1048 {
1049 	int error;
1050 	nvlist_t *config, *props = NULL;
1051 	nvlist_t *rootprops = NULL;
1052 	nvlist_t *zplprops = NULL;
1053 	char *buf;
1054 
1055 	if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1056 	    zc->zc_iflags, &config))
1057 		return (error);
1058 
1059 	if (zc->zc_nvlist_src_size != 0 && (error =
1060 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1061 	    zc->zc_iflags, &props))) {
1062 		nvlist_free(config);
1063 		return (error);
1064 	}
1065 
1066 	if (props) {
1067 		nvlist_t *nvl = NULL;
1068 		uint64_t version = SPA_VERSION;
1069 
1070 		(void) nvlist_lookup_uint64(props,
1071 		    zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1072 		if (version < SPA_VERSION_INITIAL || version > SPA_VERSION) {
1073 			error = EINVAL;
1074 			goto pool_props_bad;
1075 		}
1076 		(void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1077 		if (nvl) {
1078 			error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1079 			if (error != 0) {
1080 				nvlist_free(config);
1081 				nvlist_free(props);
1082 				return (error);
1083 			}
1084 			(void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1085 		}
1086 		VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1087 		error = zfs_fill_zplprops_root(version, rootprops,
1088 		    zplprops, NULL);
1089 		if (error)
1090 			goto pool_props_bad;
1091 	}
1092 
1093 	buf = history_str_get(zc);
1094 
1095 	error = spa_create(zc->zc_name, config, props, buf, zplprops);
1096 
1097 	/*
1098 	 * Set the remaining root properties
1099 	 */
1100 	if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1101 	    ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1102 		(void) spa_destroy(zc->zc_name);
1103 
1104 	if (buf != NULL)
1105 		history_str_free(buf);
1106 
1107 pool_props_bad:
1108 	nvlist_free(rootprops);
1109 	nvlist_free(zplprops);
1110 	nvlist_free(config);
1111 	nvlist_free(props);
1112 
1113 	return (error);
1114 }
1115 
1116 static int
1117 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1118 {
1119 	int error;
1120 	zfs_log_history(zc);
1121 	error = spa_destroy(zc->zc_name);
1122 	if (error == 0)
1123 		zvol_remove_minors(zc->zc_name);
1124 	return (error);
1125 }
1126 
1127 static int
1128 zfs_ioc_pool_import(zfs_cmd_t *zc)
1129 {
1130 	nvlist_t *config, *props = NULL;
1131 	uint64_t guid;
1132 	int error;
1133 
1134 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1135 	    zc->zc_iflags, &config)) != 0)
1136 		return (error);
1137 
1138 	if (zc->zc_nvlist_src_size != 0 && (error =
1139 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1140 	    zc->zc_iflags, &props))) {
1141 		nvlist_free(config);
1142 		return (error);
1143 	}
1144 
1145 	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1146 	    guid != zc->zc_guid)
1147 		error = EINVAL;
1148 	else if (zc->zc_cookie)
1149 		error = spa_import_verbatim(zc->zc_name, config, props);
1150 	else
1151 		error = spa_import(zc->zc_name, config, props);
1152 
1153 	if (zc->zc_nvlist_dst != 0)
1154 		(void) put_nvlist(zc, config);
1155 
1156 	nvlist_free(config);
1157 
1158 	if (props)
1159 		nvlist_free(props);
1160 
1161 	return (error);
1162 }
1163 
1164 static int
1165 zfs_ioc_pool_export(zfs_cmd_t *zc)
1166 {
1167 	int error;
1168 	boolean_t force = (boolean_t)zc->zc_cookie;
1169 	boolean_t hardforce = (boolean_t)zc->zc_guid;
1170 
1171 	zfs_log_history(zc);
1172 	error = spa_export(zc->zc_name, NULL, force, hardforce);
1173 	if (error == 0)
1174 		zvol_remove_minors(zc->zc_name);
1175 	return (error);
1176 }
1177 
1178 static int
1179 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1180 {
1181 	nvlist_t *configs;
1182 	int error;
1183 
1184 	if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1185 		return (EEXIST);
1186 
1187 	error = put_nvlist(zc, configs);
1188 
1189 	nvlist_free(configs);
1190 
1191 	return (error);
1192 }
1193 
1194 static int
1195 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1196 {
1197 	nvlist_t *config;
1198 	int error;
1199 	int ret = 0;
1200 
1201 	error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1202 	    sizeof (zc->zc_value));
1203 
1204 	if (config != NULL) {
1205 		ret = put_nvlist(zc, config);
1206 		nvlist_free(config);
1207 
1208 		/*
1209 		 * The config may be present even if 'error' is non-zero.
1210 		 * In this case we return success, and preserve the real errno
1211 		 * in 'zc_cookie'.
1212 		 */
1213 		zc->zc_cookie = error;
1214 	} else {
1215 		ret = error;
1216 	}
1217 
1218 	return (ret);
1219 }
1220 
1221 /*
1222  * Try to import the given pool, returning pool stats as appropriate so that
1223  * user land knows which devices are available and overall pool health.
1224  */
1225 static int
1226 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1227 {
1228 	nvlist_t *tryconfig, *config;
1229 	int error;
1230 
1231 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1232 	    zc->zc_iflags, &tryconfig)) != 0)
1233 		return (error);
1234 
1235 	config = spa_tryimport(tryconfig);
1236 
1237 	nvlist_free(tryconfig);
1238 
1239 	if (config == NULL)
1240 		return (EINVAL);
1241 
1242 	error = put_nvlist(zc, config);
1243 	nvlist_free(config);
1244 
1245 	return (error);
1246 }
1247 
1248 static int
1249 zfs_ioc_pool_scrub(zfs_cmd_t *zc)
1250 {
1251 	spa_t *spa;
1252 	int error;
1253 
1254 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1255 		return (error);
1256 
1257 	error = spa_scrub(spa, zc->zc_cookie);
1258 
1259 	spa_close(spa, FTAG);
1260 
1261 	return (error);
1262 }
1263 
1264 static int
1265 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1266 {
1267 	spa_t *spa;
1268 	int error;
1269 
1270 	error = spa_open(zc->zc_name, &spa, FTAG);
1271 	if (error == 0) {
1272 		spa_freeze(spa);
1273 		spa_close(spa, FTAG);
1274 	}
1275 	return (error);
1276 }
1277 
1278 static int
1279 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1280 {
1281 	spa_t *spa;
1282 	int error;
1283 
1284 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1285 		return (error);
1286 
1287 	if (zc->zc_cookie < spa_version(spa) || zc->zc_cookie > SPA_VERSION) {
1288 		spa_close(spa, FTAG);
1289 		return (EINVAL);
1290 	}
1291 
1292 	spa_upgrade(spa, zc->zc_cookie);
1293 	spa_close(spa, FTAG);
1294 
1295 	return (error);
1296 }
1297 
1298 static int
1299 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1300 {
1301 	spa_t *spa;
1302 	char *hist_buf;
1303 	uint64_t size;
1304 	int error;
1305 
1306 	if ((size = zc->zc_history_len) == 0)
1307 		return (EINVAL);
1308 
1309 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1310 		return (error);
1311 
1312 	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1313 		spa_close(spa, FTAG);
1314 		return (ENOTSUP);
1315 	}
1316 
1317 	hist_buf = kmem_alloc(size, KM_SLEEP);
1318 	if ((error = spa_history_get(spa, &zc->zc_history_offset,
1319 	    &zc->zc_history_len, hist_buf)) == 0) {
1320 		error = ddi_copyout(hist_buf,
1321 		    (void *)(uintptr_t)zc->zc_history,
1322 		    zc->zc_history_len, zc->zc_iflags);
1323 	}
1324 
1325 	spa_close(spa, FTAG);
1326 	kmem_free(hist_buf, size);
1327 	return (error);
1328 }
1329 
1330 static int
1331 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1332 {
1333 	int error;
1334 
1335 	if (error = dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value))
1336 		return (error);
1337 
1338 	return (0);
1339 }
1340 
1341 /*
1342  * inputs:
1343  * zc_name		name of filesystem
1344  * zc_obj		object to find
1345  *
1346  * outputs:
1347  * zc_value		name of object
1348  */
1349 static int
1350 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1351 {
1352 	objset_t *os;
1353 	int error;
1354 
1355 	/* XXX reading from objset not owned */
1356 	if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1357 		return (error);
1358 	if (dmu_objset_type(os) != DMU_OST_ZFS) {
1359 		dmu_objset_rele(os, FTAG);
1360 		return (EINVAL);
1361 	}
1362 	error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1363 	    sizeof (zc->zc_value));
1364 	dmu_objset_rele(os, FTAG);
1365 
1366 	return (error);
1367 }
1368 
1369 static int
1370 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1371 {
1372 	spa_t *spa;
1373 	int error;
1374 	nvlist_t *config, **l2cache, **spares;
1375 	uint_t nl2cache = 0, nspares = 0;
1376 
1377 	error = spa_open(zc->zc_name, &spa, FTAG);
1378 	if (error != 0)
1379 		return (error);
1380 
1381 	error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1382 	    zc->zc_iflags, &config);
1383 	(void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1384 	    &l2cache, &nl2cache);
1385 
1386 	(void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1387 	    &spares, &nspares);
1388 
1389 	/*
1390 	 * A root pool with concatenated devices is not supported.
1391 	 * Thus, can not add a device to a root pool.
1392 	 *
1393 	 * Intent log device can not be added to a rootpool because
1394 	 * during mountroot, zil is replayed, a seperated log device
1395 	 * can not be accessed during the mountroot time.
1396 	 *
1397 	 * l2cache and spare devices are ok to be added to a rootpool.
1398 	 */
1399 	if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1400 		nvlist_free(config);
1401 		spa_close(spa, FTAG);
1402 		return (EDOM);
1403 	}
1404 
1405 	if (error == 0) {
1406 		error = spa_vdev_add(spa, config);
1407 		nvlist_free(config);
1408 	}
1409 	spa_close(spa, FTAG);
1410 	return (error);
1411 }
1412 
1413 static int
1414 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1415 {
1416 	spa_t *spa;
1417 	int error;
1418 
1419 	error = spa_open(zc->zc_name, &spa, FTAG);
1420 	if (error != 0)
1421 		return (error);
1422 	error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1423 	spa_close(spa, FTAG);
1424 	return (error);
1425 }
1426 
1427 static int
1428 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1429 {
1430 	spa_t *spa;
1431 	int error;
1432 	vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1433 
1434 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1435 		return (error);
1436 	switch (zc->zc_cookie) {
1437 	case VDEV_STATE_ONLINE:
1438 		error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1439 		break;
1440 
1441 	case VDEV_STATE_OFFLINE:
1442 		error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1443 		break;
1444 
1445 	case VDEV_STATE_FAULTED:
1446 		if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1447 		    zc->zc_obj != VDEV_AUX_EXTERNAL)
1448 			zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1449 
1450 		error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1451 		break;
1452 
1453 	case VDEV_STATE_DEGRADED:
1454 		if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1455 		    zc->zc_obj != VDEV_AUX_EXTERNAL)
1456 			zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1457 
1458 		error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1459 		break;
1460 
1461 	default:
1462 		error = EINVAL;
1463 	}
1464 	zc->zc_cookie = newstate;
1465 	spa_close(spa, FTAG);
1466 	return (error);
1467 }
1468 
1469 static int
1470 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1471 {
1472 	spa_t *spa;
1473 	int replacing = zc->zc_cookie;
1474 	nvlist_t *config;
1475 	int error;
1476 
1477 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1478 		return (error);
1479 
1480 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1481 	    zc->zc_iflags, &config)) == 0) {
1482 		error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1483 		nvlist_free(config);
1484 	}
1485 
1486 	spa_close(spa, FTAG);
1487 	return (error);
1488 }
1489 
1490 static int
1491 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1492 {
1493 	spa_t *spa;
1494 	int error;
1495 
1496 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1497 		return (error);
1498 
1499 	error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1500 
1501 	spa_close(spa, FTAG);
1502 	return (error);
1503 }
1504 
1505 static int
1506 zfs_ioc_vdev_split(zfs_cmd_t *zc)
1507 {
1508 	spa_t *spa;
1509 	nvlist_t *config, *props = NULL;
1510 	int error;
1511 	boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
1512 
1513 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1514 		return (error);
1515 
1516 	if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1517 	    zc->zc_iflags, &config)) {
1518 		spa_close(spa, FTAG);
1519 		return (error);
1520 	}
1521 
1522 	if (zc->zc_nvlist_src_size != 0 && (error =
1523 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1524 	    zc->zc_iflags, &props))) {
1525 		spa_close(spa, FTAG);
1526 		nvlist_free(config);
1527 		return (error);
1528 	}
1529 
1530 	error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
1531 
1532 	spa_close(spa, FTAG);
1533 
1534 	nvlist_free(config);
1535 	nvlist_free(props);
1536 
1537 	return (error);
1538 }
1539 
1540 static int
1541 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
1542 {
1543 	spa_t *spa;
1544 	char *path = zc->zc_value;
1545 	uint64_t guid = zc->zc_guid;
1546 	int error;
1547 
1548 	error = spa_open(zc->zc_name, &spa, FTAG);
1549 	if (error != 0)
1550 		return (error);
1551 
1552 	error = spa_vdev_setpath(spa, guid, path);
1553 	spa_close(spa, FTAG);
1554 	return (error);
1555 }
1556 
1557 static int
1558 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
1559 {
1560 	spa_t *spa;
1561 	char *fru = zc->zc_value;
1562 	uint64_t guid = zc->zc_guid;
1563 	int error;
1564 
1565 	error = spa_open(zc->zc_name, &spa, FTAG);
1566 	if (error != 0)
1567 		return (error);
1568 
1569 	error = spa_vdev_setfru(spa, guid, fru);
1570 	spa_close(spa, FTAG);
1571 	return (error);
1572 }
1573 
1574 /*
1575  * inputs:
1576  * zc_name		name of filesystem
1577  * zc_nvlist_dst_size	size of buffer for property nvlist
1578  *
1579  * outputs:
1580  * zc_objset_stats	stats
1581  * zc_nvlist_dst	property nvlist
1582  * zc_nvlist_dst_size	size of property nvlist
1583  */
1584 static int
1585 zfs_ioc_objset_stats(zfs_cmd_t *zc)
1586 {
1587 	objset_t *os = NULL;
1588 	int error;
1589 	nvlist_t *nv;
1590 
1591 	if (error = dmu_objset_hold(zc->zc_name, FTAG, &os))
1592 		return (error);
1593 
1594 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1595 
1596 	if (zc->zc_nvlist_dst != 0 &&
1597 	    (error = dsl_prop_get_all(os, &nv)) == 0) {
1598 		dmu_objset_stats(os, nv);
1599 		/*
1600 		 * NB: zvol_get_stats() will read the objset contents,
1601 		 * which we aren't supposed to do with a
1602 		 * DS_MODE_USER hold, because it could be
1603 		 * inconsistent.  So this is a bit of a workaround...
1604 		 * XXX reading with out owning
1605 		 */
1606 		if (!zc->zc_objset_stats.dds_inconsistent) {
1607 			if (dmu_objset_type(os) == DMU_OST_ZVOL)
1608 				VERIFY(zvol_get_stats(os, nv) == 0);
1609 		}
1610 		error = put_nvlist(zc, nv);
1611 		nvlist_free(nv);
1612 	}
1613 
1614 	dmu_objset_rele(os, FTAG);
1615 	return (error);
1616 }
1617 
1618 /*
1619  * inputs:
1620  * zc_name		name of filesystem
1621  * zc_nvlist_dst_size	size of buffer for property nvlist
1622  *
1623  * outputs:
1624  * zc_nvlist_dst	received property nvlist
1625  * zc_nvlist_dst_size	size of received property nvlist
1626  *
1627  * Gets received properties (distinct from local properties on or after
1628  * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
1629  * local property values.
1630  */
1631 static int
1632 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
1633 {
1634 	objset_t *os = NULL;
1635 	int error;
1636 	nvlist_t *nv;
1637 
1638 	if (error = dmu_objset_hold(zc->zc_name, FTAG, &os))
1639 		return (error);
1640 
1641 	/*
1642 	 * Without this check, we would return local property values if the
1643 	 * caller has not already received properties on or after
1644 	 * SPA_VERSION_RECVD_PROPS.
1645 	 */
1646 	if (!dsl_prop_get_hasrecvd(os)) {
1647 		dmu_objset_rele(os, FTAG);
1648 		return (ENOTSUP);
1649 	}
1650 
1651 	if (zc->zc_nvlist_dst != 0 &&
1652 	    (error = dsl_prop_get_received(os, &nv)) == 0) {
1653 		error = put_nvlist(zc, nv);
1654 		nvlist_free(nv);
1655 	}
1656 
1657 	dmu_objset_rele(os, FTAG);
1658 	return (error);
1659 }
1660 
1661 static int
1662 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
1663 {
1664 	uint64_t value;
1665 	int error;
1666 
1667 	/*
1668 	 * zfs_get_zplprop() will either find a value or give us
1669 	 * the default value (if there is one).
1670 	 */
1671 	if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
1672 		return (error);
1673 	VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
1674 	return (0);
1675 }
1676 
1677 /*
1678  * inputs:
1679  * zc_name		name of filesystem
1680  * zc_nvlist_dst_size	size of buffer for zpl property nvlist
1681  *
1682  * outputs:
1683  * zc_nvlist_dst	zpl property nvlist
1684  * zc_nvlist_dst_size	size of zpl property nvlist
1685  */
1686 static int
1687 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
1688 {
1689 	objset_t *os;
1690 	int err;
1691 
1692 	/* XXX reading without owning */
1693 	if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
1694 		return (err);
1695 
1696 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1697 
1698 	/*
1699 	 * NB: nvl_add_zplprop() will read the objset contents,
1700 	 * which we aren't supposed to do with a DS_MODE_USER
1701 	 * hold, because it could be inconsistent.
1702 	 */
1703 	if (zc->zc_nvlist_dst != NULL &&
1704 	    !zc->zc_objset_stats.dds_inconsistent &&
1705 	    dmu_objset_type(os) == DMU_OST_ZFS) {
1706 		nvlist_t *nv;
1707 
1708 		VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1709 		if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
1710 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
1711 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
1712 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
1713 			err = put_nvlist(zc, nv);
1714 		nvlist_free(nv);
1715 	} else {
1716 		err = ENOENT;
1717 	}
1718 	dmu_objset_rele(os, FTAG);
1719 	return (err);
1720 }
1721 
1722 static boolean_t
1723 dataset_name_hidden(const char *name)
1724 {
1725 	/*
1726 	 * Skip over datasets that are not visible in this zone,
1727 	 * internal datasets (which have a $ in their name), and
1728 	 * temporary datasets (which have a % in their name).
1729 	 */
1730 	if (strchr(name, '$') != NULL)
1731 		return (B_TRUE);
1732 	if (strchr(name, '%') != NULL)
1733 		return (B_TRUE);
1734 	if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
1735 		return (B_TRUE);
1736 	return (B_FALSE);
1737 }
1738 
1739 /*
1740  * inputs:
1741  * zc_name		name of filesystem
1742  * zc_cookie		zap cursor
1743  * zc_nvlist_dst_size	size of buffer for property nvlist
1744  *
1745  * outputs:
1746  * zc_name		name of next filesystem
1747  * zc_cookie		zap cursor
1748  * zc_objset_stats	stats
1749  * zc_nvlist_dst	property nvlist
1750  * zc_nvlist_dst_size	size of property nvlist
1751  */
1752 static int
1753 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
1754 {
1755 	objset_t *os;
1756 	int error;
1757 	char *p;
1758 	size_t orig_len = strlen(zc->zc_name);
1759 
1760 top:
1761 	if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
1762 		if (error == ENOENT)
1763 			error = ESRCH;
1764 		return (error);
1765 	}
1766 
1767 	p = strrchr(zc->zc_name, '/');
1768 	if (p == NULL || p[1] != '\0')
1769 		(void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
1770 	p = zc->zc_name + strlen(zc->zc_name);
1771 
1772 	/*
1773 	 * Pre-fetch the datasets.  dmu_objset_prefetch() always returns 0
1774 	 * but is not declared void because its called by dmu_objset_find().
1775 	 */
1776 	if (zc->zc_cookie == 0) {
1777 		uint64_t cookie = 0;
1778 		int len = sizeof (zc->zc_name) - (p - zc->zc_name);
1779 
1780 		while (dmu_dir_list_next(os, len, p, NULL, &cookie) == 0)
1781 			(void) dmu_objset_prefetch(p, NULL);
1782 	}
1783 
1784 	do {
1785 		error = dmu_dir_list_next(os,
1786 		    sizeof (zc->zc_name) - (p - zc->zc_name), p,
1787 		    NULL, &zc->zc_cookie);
1788 		if (error == ENOENT)
1789 			error = ESRCH;
1790 	} while (error == 0 && dataset_name_hidden(zc->zc_name) &&
1791 	    !(zc->zc_iflags & FKIOCTL));
1792 	dmu_objset_rele(os, FTAG);
1793 
1794 	/*
1795 	 * If it's an internal dataset (ie. with a '$' in its name),
1796 	 * don't try to get stats for it, otherwise we'll return ENOENT.
1797 	 */
1798 	if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
1799 		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
1800 		if (error == ENOENT) {
1801 			/* We lost a race with destroy, get the next one. */
1802 			zc->zc_name[orig_len] = '\0';
1803 			goto top;
1804 		}
1805 	}
1806 	return (error);
1807 }
1808 
1809 /*
1810  * inputs:
1811  * zc_name		name of filesystem
1812  * zc_cookie		zap cursor
1813  * zc_nvlist_dst_size	size of buffer for property nvlist
1814  *
1815  * outputs:
1816  * zc_name		name of next snapshot
1817  * zc_objset_stats	stats
1818  * zc_nvlist_dst	property nvlist
1819  * zc_nvlist_dst_size	size of property nvlist
1820  */
1821 static int
1822 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
1823 {
1824 	objset_t *os;
1825 	int error;
1826 
1827 top:
1828 	if (zc->zc_cookie == 0)
1829 		(void) dmu_objset_find(zc->zc_name, dmu_objset_prefetch,
1830 		    NULL, DS_FIND_SNAPSHOTS);
1831 
1832 	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
1833 	if (error)
1834 		return (error == ENOENT ? ESRCH : error);
1835 
1836 	/*
1837 	 * A dataset name of maximum length cannot have any snapshots,
1838 	 * so exit immediately.
1839 	 */
1840 	if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
1841 		dmu_objset_rele(os, FTAG);
1842 		return (ESRCH);
1843 	}
1844 
1845 	error = dmu_snapshot_list_next(os,
1846 	    sizeof (zc->zc_name) - strlen(zc->zc_name),
1847 	    zc->zc_name + strlen(zc->zc_name), NULL, &zc->zc_cookie, NULL);
1848 	dmu_objset_rele(os, FTAG);
1849 	if (error == 0) {
1850 		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
1851 		if (error == ENOENT)  {
1852 			/* We lost a race with destroy, get the next one. */
1853 			*strchr(zc->zc_name, '@') = '\0';
1854 			goto top;
1855 		}
1856 	} else if (error == ENOENT) {
1857 		error = ESRCH;
1858 	}
1859 
1860 	/* if we failed, undo the @ that we tacked on to zc_name */
1861 	if (error)
1862 		*strchr(zc->zc_name, '@') = '\0';
1863 	return (error);
1864 }
1865 
1866 static int
1867 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
1868 {
1869 	const char *propname = nvpair_name(pair);
1870 	uint64_t *valary;
1871 	unsigned int vallen;
1872 	const char *domain;
1873 	zfs_userquota_prop_t type;
1874 	uint64_t rid;
1875 	uint64_t quota;
1876 	zfsvfs_t *zfsvfs;
1877 	int err;
1878 
1879 	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
1880 		nvlist_t *attrs;
1881 		VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
1882 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
1883 		    &pair) == 0);
1884 	}
1885 
1886 	VERIFY(nvpair_value_uint64_array(pair, &valary, &vallen) == 0);
1887 	VERIFY(vallen == 3);
1888 	type = valary[0];
1889 	rid = valary[1];
1890 	quota = valary[2];
1891 	/*
1892 	 * The propname is encoded as
1893 	 * userquota@<rid>-<domain>.
1894 	 */
1895 	domain = strchr(propname, '-') + 1;
1896 
1897 	err = zfsvfs_hold(dsname, FTAG, &zfsvfs);
1898 	if (err == 0) {
1899 		err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
1900 		zfsvfs_rele(zfsvfs, FTAG);
1901 	}
1902 
1903 	return (err);
1904 }
1905 
1906 /*
1907  * If the named property is one that has a special function to set its value,
1908  * return 0 on success and a positive error code on failure; otherwise if it is
1909  * not one of the special properties handled by this function, return -1.
1910  *
1911  * XXX: It would be better for callers of the properety interface if we handled
1912  * these special cases in dsl_prop.c (in the dsl layer).
1913  */
1914 static int
1915 zfs_prop_set_special(const char *dsname, zprop_source_t source,
1916     nvpair_t *pair)
1917 {
1918 	const char *propname = nvpair_name(pair);
1919 	zfs_prop_t prop = zfs_name_to_prop(propname);
1920 	uint64_t intval;
1921 	int err;
1922 
1923 	if (prop == ZPROP_INVAL) {
1924 		if (zfs_prop_userquota(propname))
1925 			return (zfs_prop_set_userquota(dsname, pair));
1926 		return (-1);
1927 	}
1928 
1929 	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
1930 		nvlist_t *attrs;
1931 		VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
1932 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
1933 		    &pair) == 0);
1934 	}
1935 
1936 	if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
1937 		return (-1);
1938 
1939 	VERIFY(0 == nvpair_value_uint64(pair, &intval));
1940 
1941 	switch (prop) {
1942 	case ZFS_PROP_QUOTA:
1943 		err = dsl_dir_set_quota(dsname, source, intval);
1944 		break;
1945 	case ZFS_PROP_REFQUOTA:
1946 		err = dsl_dataset_set_quota(dsname, source, intval);
1947 		break;
1948 	case ZFS_PROP_RESERVATION:
1949 		err = dsl_dir_set_reservation(dsname, source, intval);
1950 		break;
1951 	case ZFS_PROP_REFRESERVATION:
1952 		err = dsl_dataset_set_reservation(dsname, source, intval);
1953 		break;
1954 	case ZFS_PROP_VOLSIZE:
1955 		err = zvol_set_volsize(dsname, ddi_driver_major(zfs_dip),
1956 		    intval);
1957 		break;
1958 	case ZFS_PROP_VERSION:
1959 	{
1960 		zfsvfs_t *zfsvfs;
1961 
1962 		if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs)) != 0)
1963 			break;
1964 
1965 		err = zfs_set_version(zfsvfs, intval);
1966 		zfsvfs_rele(zfsvfs, FTAG);
1967 
1968 		if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
1969 			zfs_cmd_t *zc;
1970 
1971 			zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
1972 			(void) strcpy(zc->zc_name, dsname);
1973 			(void) zfs_ioc_userspace_upgrade(zc);
1974 			kmem_free(zc, sizeof (zfs_cmd_t));
1975 		}
1976 		break;
1977 	}
1978 
1979 	default:
1980 		err = -1;
1981 	}
1982 
1983 	return (err);
1984 }
1985 
1986 /*
1987  * This function is best effort. If it fails to set any of the given properties,
1988  * it continues to set as many as it can and returns the first error
1989  * encountered. If the caller provides a non-NULL errlist, it also gives the
1990  * complete list of names of all the properties it failed to set along with the
1991  * corresponding error numbers. The caller is responsible for freeing the
1992  * returned errlist.
1993  *
1994  * If every property is set successfully, zero is returned and the list pointed
1995  * at by errlist is NULL.
1996  */
1997 int
1998 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
1999     nvlist_t **errlist)
2000 {
2001 	nvpair_t *pair;
2002 	nvpair_t *propval;
2003 	int rv = 0;
2004 	uint64_t intval;
2005 	char *strval;
2006 	nvlist_t *genericnvl;
2007 	nvlist_t *errors;
2008 	nvlist_t *retrynvl;
2009 
2010 	VERIFY(nvlist_alloc(&genericnvl, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2011 	VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2012 	VERIFY(nvlist_alloc(&retrynvl, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2013 
2014 retry:
2015 	pair = NULL;
2016 	while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2017 		const char *propname = nvpair_name(pair);
2018 		zfs_prop_t prop = zfs_name_to_prop(propname);
2019 		int err = 0;
2020 
2021 		/* decode the property value */
2022 		propval = pair;
2023 		if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2024 			nvlist_t *attrs;
2025 			VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2026 			VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2027 			    &propval) == 0);
2028 		}
2029 
2030 		/* Validate value type */
2031 		if (prop == ZPROP_INVAL) {
2032 			if (zfs_prop_user(propname)) {
2033 				if (nvpair_type(propval) != DATA_TYPE_STRING)
2034 					err = EINVAL;
2035 			} else if (zfs_prop_userquota(propname)) {
2036 				if (nvpair_type(propval) !=
2037 				    DATA_TYPE_UINT64_ARRAY)
2038 					err = EINVAL;
2039 			}
2040 		} else {
2041 			if (nvpair_type(propval) == DATA_TYPE_STRING) {
2042 				if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2043 					err = EINVAL;
2044 			} else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2045 				const char *unused;
2046 
2047 				VERIFY(nvpair_value_uint64(propval,
2048 				    &intval) == 0);
2049 
2050 				switch (zfs_prop_get_type(prop)) {
2051 				case PROP_TYPE_NUMBER:
2052 					break;
2053 				case PROP_TYPE_STRING:
2054 					err = EINVAL;
2055 					break;
2056 				case PROP_TYPE_INDEX:
2057 					if (zfs_prop_index_to_string(prop,
2058 					    intval, &unused) != 0)
2059 						err = EINVAL;
2060 					break;
2061 				default:
2062 					cmn_err(CE_PANIC,
2063 					    "unknown property type");
2064 				}
2065 			} else {
2066 				err = EINVAL;
2067 			}
2068 		}
2069 
2070 		/* Validate permissions */
2071 		if (err == 0)
2072 			err = zfs_check_settable(dsname, pair, CRED());
2073 
2074 		if (err == 0) {
2075 			err = zfs_prop_set_special(dsname, source, pair);
2076 			if (err == -1) {
2077 				/*
2078 				 * For better performance we build up a list of
2079 				 * properties to set in a single transaction.
2080 				 */
2081 				err = nvlist_add_nvpair(genericnvl, pair);
2082 			} else if (err != 0 && nvl != retrynvl) {
2083 				/*
2084 				 * This may be a spurious error caused by
2085 				 * receiving quota and reservation out of order.
2086 				 * Try again in a second pass.
2087 				 */
2088 				err = nvlist_add_nvpair(retrynvl, pair);
2089 			}
2090 		}
2091 
2092 		if (err != 0)
2093 			VERIFY(nvlist_add_int32(errors, propname, err) == 0);
2094 	}
2095 
2096 	if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2097 		nvl = retrynvl;
2098 		goto retry;
2099 	}
2100 
2101 	if (!nvlist_empty(genericnvl) &&
2102 	    dsl_props_set(dsname, source, genericnvl) != 0) {
2103 		/*
2104 		 * If this fails, we still want to set as many properties as we
2105 		 * can, so try setting them individually.
2106 		 */
2107 		pair = NULL;
2108 		while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2109 			const char *propname = nvpair_name(pair);
2110 			int err = 0;
2111 
2112 			propval = pair;
2113 			if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2114 				nvlist_t *attrs;
2115 				VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2116 				VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2117 				    &propval) == 0);
2118 			}
2119 
2120 			if (nvpair_type(propval) == DATA_TYPE_STRING) {
2121 				VERIFY(nvpair_value_string(propval,
2122 				    &strval) == 0);
2123 				err = dsl_prop_set(dsname, propname, source, 1,
2124 				    strlen(strval) + 1, strval);
2125 			} else {
2126 				VERIFY(nvpair_value_uint64(propval,
2127 				    &intval) == 0);
2128 				err = dsl_prop_set(dsname, propname, source, 8,
2129 				    1, &intval);
2130 			}
2131 
2132 			if (err != 0) {
2133 				VERIFY(nvlist_add_int32(errors, propname,
2134 				    err) == 0);
2135 			}
2136 		}
2137 	}
2138 	nvlist_free(genericnvl);
2139 	nvlist_free(retrynvl);
2140 
2141 	if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
2142 		nvlist_free(errors);
2143 		errors = NULL;
2144 	} else {
2145 		VERIFY(nvpair_value_int32(pair, &rv) == 0);
2146 	}
2147 
2148 	if (errlist == NULL)
2149 		nvlist_free(errors);
2150 	else
2151 		*errlist = errors;
2152 
2153 	return (rv);
2154 }
2155 
2156 /*
2157  * Check that all the properties are valid user properties.
2158  */
2159 static int
2160 zfs_check_userprops(char *fsname, nvlist_t *nvl)
2161 {
2162 	nvpair_t *pair = NULL;
2163 	int error = 0;
2164 
2165 	while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2166 		const char *propname = nvpair_name(pair);
2167 		char *valstr;
2168 
2169 		if (!zfs_prop_user(propname) ||
2170 		    nvpair_type(pair) != DATA_TYPE_STRING)
2171 			return (EINVAL);
2172 
2173 		if (error = zfs_secpolicy_write_perms(fsname,
2174 		    ZFS_DELEG_PERM_USERPROP, CRED()))
2175 			return (error);
2176 
2177 		if (strlen(propname) >= ZAP_MAXNAMELEN)
2178 			return (ENAMETOOLONG);
2179 
2180 		VERIFY(nvpair_value_string(pair, &valstr) == 0);
2181 		if (strlen(valstr) >= ZAP_MAXVALUELEN)
2182 			return (E2BIG);
2183 	}
2184 	return (0);
2185 }
2186 
2187 static void
2188 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2189 {
2190 	nvpair_t *pair;
2191 
2192 	VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2193 
2194 	pair = NULL;
2195 	while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2196 		if (nvlist_exists(skipped, nvpair_name(pair)))
2197 			continue;
2198 
2199 		VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2200 	}
2201 }
2202 
2203 static int
2204 clear_received_props(objset_t *os, const char *fs, nvlist_t *props,
2205     nvlist_t *skipped)
2206 {
2207 	int err = 0;
2208 	nvlist_t *cleared_props = NULL;
2209 	props_skip(props, skipped, &cleared_props);
2210 	if (!nvlist_empty(cleared_props)) {
2211 		/*
2212 		 * Acts on local properties until the dataset has received
2213 		 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2214 		 */
2215 		zprop_source_t flags = (ZPROP_SRC_NONE |
2216 		    (dsl_prop_get_hasrecvd(os) ? ZPROP_SRC_RECEIVED : 0));
2217 		err = zfs_set_prop_nvlist(fs, flags, cleared_props, NULL);
2218 	}
2219 	nvlist_free(cleared_props);
2220 	return (err);
2221 }
2222 
2223 /*
2224  * inputs:
2225  * zc_name		name of filesystem
2226  * zc_value		name of property to set
2227  * zc_nvlist_src{_size}	nvlist of properties to apply
2228  * zc_cookie		received properties flag
2229  *
2230  * outputs:
2231  * zc_nvlist_dst{_size} error for each unapplied received property
2232  */
2233 static int
2234 zfs_ioc_set_prop(zfs_cmd_t *zc)
2235 {
2236 	nvlist_t *nvl;
2237 	boolean_t received = zc->zc_cookie;
2238 	zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2239 	    ZPROP_SRC_LOCAL);
2240 	nvlist_t *errors = NULL;
2241 	int error;
2242 
2243 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2244 	    zc->zc_iflags, &nvl)) != 0)
2245 		return (error);
2246 
2247 	if (received) {
2248 		nvlist_t *origprops;
2249 		objset_t *os;
2250 
2251 		if (dmu_objset_hold(zc->zc_name, FTAG, &os) == 0) {
2252 			if (dsl_prop_get_received(os, &origprops) == 0) {
2253 				(void) clear_received_props(os,
2254 				    zc->zc_name, origprops, nvl);
2255 				nvlist_free(origprops);
2256 			}
2257 
2258 			dsl_prop_set_hasrecvd(os);
2259 			dmu_objset_rele(os, FTAG);
2260 		}
2261 	}
2262 
2263 	error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, &errors);
2264 
2265 	if (zc->zc_nvlist_dst != NULL && errors != NULL) {
2266 		(void) put_nvlist(zc, errors);
2267 	}
2268 
2269 	nvlist_free(errors);
2270 	nvlist_free(nvl);
2271 	return (error);
2272 }
2273 
2274 /*
2275  * inputs:
2276  * zc_name		name of filesystem
2277  * zc_value		name of property to inherit
2278  * zc_cookie		revert to received value if TRUE
2279  *
2280  * outputs:		none
2281  */
2282 static int
2283 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2284 {
2285 	const char *propname = zc->zc_value;
2286 	zfs_prop_t prop = zfs_name_to_prop(propname);
2287 	boolean_t received = zc->zc_cookie;
2288 	zprop_source_t source = (received
2289 	    ? ZPROP_SRC_NONE		/* revert to received value, if any */
2290 	    : ZPROP_SRC_INHERITED);	/* explicitly inherit */
2291 
2292 	if (received) {
2293 		nvlist_t *dummy;
2294 		nvpair_t *pair;
2295 		zprop_type_t type;
2296 		int err;
2297 
2298 		/*
2299 		 * zfs_prop_set_special() expects properties in the form of an
2300 		 * nvpair with type info.
2301 		 */
2302 		if (prop == ZPROP_INVAL) {
2303 			if (!zfs_prop_user(propname))
2304 				return (EINVAL);
2305 
2306 			type = PROP_TYPE_STRING;
2307 		} else if (prop == ZFS_PROP_VOLSIZE ||
2308 		    prop == ZFS_PROP_VERSION) {
2309 			return (EINVAL);
2310 		} else {
2311 			type = zfs_prop_get_type(prop);
2312 		}
2313 
2314 		VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2315 
2316 		switch (type) {
2317 		case PROP_TYPE_STRING:
2318 			VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2319 			break;
2320 		case PROP_TYPE_NUMBER:
2321 		case PROP_TYPE_INDEX:
2322 			VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2323 			break;
2324 		default:
2325 			nvlist_free(dummy);
2326 			return (EINVAL);
2327 		}
2328 
2329 		pair = nvlist_next_nvpair(dummy, NULL);
2330 		err = zfs_prop_set_special(zc->zc_name, source, pair);
2331 		nvlist_free(dummy);
2332 		if (err != -1)
2333 			return (err); /* special property already handled */
2334 	} else {
2335 		/*
2336 		 * Only check this in the non-received case. We want to allow
2337 		 * 'inherit -S' to revert non-inheritable properties like quota
2338 		 * and reservation to the received or default values even though
2339 		 * they are not considered inheritable.
2340 		 */
2341 		if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2342 			return (EINVAL);
2343 	}
2344 
2345 	/* the property name has been validated by zfs_secpolicy_inherit() */
2346 	return (dsl_prop_set(zc->zc_name, zc->zc_value, source, 0, 0, NULL));
2347 }
2348 
2349 static int
2350 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2351 {
2352 	nvlist_t *props;
2353 	spa_t *spa;
2354 	int error;
2355 	nvpair_t *pair;
2356 
2357 	if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2358 	    zc->zc_iflags, &props))
2359 		return (error);
2360 
2361 	/*
2362 	 * If the only property is the configfile, then just do a spa_lookup()
2363 	 * to handle the faulted case.
2364 	 */
2365 	pair = nvlist_next_nvpair(props, NULL);
2366 	if (pair != NULL && strcmp(nvpair_name(pair),
2367 	    zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2368 	    nvlist_next_nvpair(props, pair) == NULL) {
2369 		mutex_enter(&spa_namespace_lock);
2370 		if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2371 			spa_configfile_set(spa, props, B_FALSE);
2372 			spa_config_sync(spa, B_FALSE, B_TRUE);
2373 		}
2374 		mutex_exit(&spa_namespace_lock);
2375 		if (spa != NULL) {
2376 			nvlist_free(props);
2377 			return (0);
2378 		}
2379 	}
2380 
2381 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2382 		nvlist_free(props);
2383 		return (error);
2384 	}
2385 
2386 	error = spa_prop_set(spa, props);
2387 
2388 	nvlist_free(props);
2389 	spa_close(spa, FTAG);
2390 
2391 	return (error);
2392 }
2393 
2394 static int
2395 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2396 {
2397 	spa_t *spa;
2398 	int error;
2399 	nvlist_t *nvp = NULL;
2400 
2401 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2402 		/*
2403 		 * If the pool is faulted, there may be properties we can still
2404 		 * get (such as altroot and cachefile), so attempt to get them
2405 		 * anyway.
2406 		 */
2407 		mutex_enter(&spa_namespace_lock);
2408 		if ((spa = spa_lookup(zc->zc_name)) != NULL)
2409 			error = spa_prop_get(spa, &nvp);
2410 		mutex_exit(&spa_namespace_lock);
2411 	} else {
2412 		error = spa_prop_get(spa, &nvp);
2413 		spa_close(spa, FTAG);
2414 	}
2415 
2416 	if (error == 0 && zc->zc_nvlist_dst != NULL)
2417 		error = put_nvlist(zc, nvp);
2418 	else
2419 		error = EFAULT;
2420 
2421 	nvlist_free(nvp);
2422 	return (error);
2423 }
2424 
2425 static int
2426 zfs_ioc_iscsi_perm_check(zfs_cmd_t *zc)
2427 {
2428 	nvlist_t *nvp;
2429 	int error;
2430 	uint32_t uid;
2431 	uint32_t gid;
2432 	uint32_t *groups;
2433 	uint_t group_cnt;
2434 	cred_t	*usercred;
2435 
2436 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2437 	    zc->zc_iflags, &nvp)) != 0) {
2438 		return (error);
2439 	}
2440 
2441 	if ((error = nvlist_lookup_uint32(nvp,
2442 	    ZFS_DELEG_PERM_UID, &uid)) != 0) {
2443 		nvlist_free(nvp);
2444 		return (EPERM);
2445 	}
2446 
2447 	if ((error = nvlist_lookup_uint32(nvp,
2448 	    ZFS_DELEG_PERM_GID, &gid)) != 0) {
2449 		nvlist_free(nvp);
2450 		return (EPERM);
2451 	}
2452 
2453 	if ((error = nvlist_lookup_uint32_array(nvp, ZFS_DELEG_PERM_GROUPS,
2454 	    &groups, &group_cnt)) != 0) {
2455 		nvlist_free(nvp);
2456 		return (EPERM);
2457 	}
2458 	usercred = cralloc();
2459 	if ((crsetugid(usercred, uid, gid) != 0) ||
2460 	    (crsetgroups(usercred, group_cnt, (gid_t *)groups) != 0)) {
2461 		nvlist_free(nvp);
2462 		crfree(usercred);
2463 		return (EPERM);
2464 	}
2465 	nvlist_free(nvp);
2466 	error = dsl_deleg_access(zc->zc_name,
2467 	    zfs_prop_to_name(ZFS_PROP_SHAREISCSI), usercred);
2468 	crfree(usercred);
2469 	return (error);
2470 }
2471 
2472 /*
2473  * inputs:
2474  * zc_name		name of filesystem
2475  * zc_nvlist_src{_size}	nvlist of delegated permissions
2476  * zc_perm_action	allow/unallow flag
2477  *
2478  * outputs:		none
2479  */
2480 static int
2481 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2482 {
2483 	int error;
2484 	nvlist_t *fsaclnv = NULL;
2485 
2486 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2487 	    zc->zc_iflags, &fsaclnv)) != 0)
2488 		return (error);
2489 
2490 	/*
2491 	 * Verify nvlist is constructed correctly
2492 	 */
2493 	if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2494 		nvlist_free(fsaclnv);
2495 		return (EINVAL);
2496 	}
2497 
2498 	/*
2499 	 * If we don't have PRIV_SYS_MOUNT, then validate
2500 	 * that user is allowed to hand out each permission in
2501 	 * the nvlist(s)
2502 	 */
2503 
2504 	error = secpolicy_zfs(CRED());
2505 	if (error) {
2506 		if (zc->zc_perm_action == B_FALSE) {
2507 			error = dsl_deleg_can_allow(zc->zc_name,
2508 			    fsaclnv, CRED());
2509 		} else {
2510 			error = dsl_deleg_can_unallow(zc->zc_name,
2511 			    fsaclnv, CRED());
2512 		}
2513 	}
2514 
2515 	if (error == 0)
2516 		error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2517 
2518 	nvlist_free(fsaclnv);
2519 	return (error);
2520 }
2521 
2522 /*
2523  * inputs:
2524  * zc_name		name of filesystem
2525  *
2526  * outputs:
2527  * zc_nvlist_src{_size}	nvlist of delegated permissions
2528  */
2529 static int
2530 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2531 {
2532 	nvlist_t *nvp;
2533 	int error;
2534 
2535 	if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2536 		error = put_nvlist(zc, nvp);
2537 		nvlist_free(nvp);
2538 	}
2539 
2540 	return (error);
2541 }
2542 
2543 /*
2544  * Search the vfs list for a specified resource.  Returns a pointer to it
2545  * or NULL if no suitable entry is found. The caller of this routine
2546  * is responsible for releasing the returned vfs pointer.
2547  */
2548 static vfs_t *
2549 zfs_get_vfs(const char *resource)
2550 {
2551 	struct vfs *vfsp;
2552 	struct vfs *vfs_found = NULL;
2553 
2554 	vfs_list_read_lock();
2555 	vfsp = rootvfs;
2556 	do {
2557 		if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
2558 			VFS_HOLD(vfsp);
2559 			vfs_found = vfsp;
2560 			break;
2561 		}
2562 		vfsp = vfsp->vfs_next;
2563 	} while (vfsp != rootvfs);
2564 	vfs_list_unlock();
2565 	return (vfs_found);
2566 }
2567 
2568 /* ARGSUSED */
2569 static void
2570 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
2571 {
2572 	zfs_creat_t *zct = arg;
2573 
2574 	zfs_create_fs(os, cr, zct->zct_zplprops, tx);
2575 }
2576 
2577 #define	ZFS_PROP_UNDEFINED	((uint64_t)-1)
2578 
2579 /*
2580  * inputs:
2581  * createprops		list of properties requested by creator
2582  * default_zplver	zpl version to use if unspecified in createprops
2583  * fuids_ok		fuids allowed in this version of the spa?
2584  * os			parent objset pointer (NULL if root fs)
2585  *
2586  * outputs:
2587  * zplprops	values for the zplprops we attach to the master node object
2588  * is_ci	true if requested file system will be purely case-insensitive
2589  *
2590  * Determine the settings for utf8only, normalization and
2591  * casesensitivity.  Specific values may have been requested by the
2592  * creator and/or we can inherit values from the parent dataset.  If
2593  * the file system is of too early a vintage, a creator can not
2594  * request settings for these properties, even if the requested
2595  * setting is the default value.  We don't actually want to create dsl
2596  * properties for these, so remove them from the source nvlist after
2597  * processing.
2598  */
2599 static int
2600 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
2601     boolean_t fuids_ok, nvlist_t *createprops, nvlist_t *zplprops,
2602     boolean_t *is_ci)
2603 {
2604 	uint64_t sense = ZFS_PROP_UNDEFINED;
2605 	uint64_t norm = ZFS_PROP_UNDEFINED;
2606 	uint64_t u8 = ZFS_PROP_UNDEFINED;
2607 
2608 	ASSERT(zplprops != NULL);
2609 
2610 	/*
2611 	 * Pull out creator prop choices, if any.
2612 	 */
2613 	if (createprops) {
2614 		(void) nvlist_lookup_uint64(createprops,
2615 		    zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
2616 		(void) nvlist_lookup_uint64(createprops,
2617 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
2618 		(void) nvlist_remove_all(createprops,
2619 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE));
2620 		(void) nvlist_lookup_uint64(createprops,
2621 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
2622 		(void) nvlist_remove_all(createprops,
2623 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
2624 		(void) nvlist_lookup_uint64(createprops,
2625 		    zfs_prop_to_name(ZFS_PROP_CASE), &sense);
2626 		(void) nvlist_remove_all(createprops,
2627 		    zfs_prop_to_name(ZFS_PROP_CASE));
2628 	}
2629 
2630 	/*
2631 	 * If the zpl version requested is whacky or the file system
2632 	 * or pool is version is too "young" to support normalization
2633 	 * and the creator tried to set a value for one of the props,
2634 	 * error out.
2635 	 */
2636 	if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
2637 	    (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
2638 	    (zplver < ZPL_VERSION_NORMALIZATION &&
2639 	    (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
2640 	    sense != ZFS_PROP_UNDEFINED)))
2641 		return (ENOTSUP);
2642 
2643 	/*
2644 	 * Put the version in the zplprops
2645 	 */
2646 	VERIFY(nvlist_add_uint64(zplprops,
2647 	    zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
2648 
2649 	if (norm == ZFS_PROP_UNDEFINED)
2650 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
2651 	VERIFY(nvlist_add_uint64(zplprops,
2652 	    zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
2653 
2654 	/*
2655 	 * If we're normalizing, names must always be valid UTF-8 strings.
2656 	 */
2657 	if (norm)
2658 		u8 = 1;
2659 	if (u8 == ZFS_PROP_UNDEFINED)
2660 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
2661 	VERIFY(nvlist_add_uint64(zplprops,
2662 	    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
2663 
2664 	if (sense == ZFS_PROP_UNDEFINED)
2665 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
2666 	VERIFY(nvlist_add_uint64(zplprops,
2667 	    zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
2668 
2669 	if (is_ci)
2670 		*is_ci = (sense == ZFS_CASE_INSENSITIVE);
2671 
2672 	return (0);
2673 }
2674 
2675 static int
2676 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
2677     nvlist_t *zplprops, boolean_t *is_ci)
2678 {
2679 	boolean_t fuids_ok = B_TRUE;
2680 	uint64_t zplver = ZPL_VERSION;
2681 	objset_t *os = NULL;
2682 	char parentname[MAXNAMELEN];
2683 	char *cp;
2684 	int error;
2685 
2686 	(void) strlcpy(parentname, dataset, sizeof (parentname));
2687 	cp = strrchr(parentname, '/');
2688 	ASSERT(cp != NULL);
2689 	cp[0] = '\0';
2690 
2691 	if (zfs_earlier_version(dataset, SPA_VERSION_USERSPACE))
2692 		zplver = ZPL_VERSION_USERSPACE - 1;
2693 	if (zfs_earlier_version(dataset, SPA_VERSION_FUID)) {
2694 		zplver = ZPL_VERSION_FUID - 1;
2695 		fuids_ok = B_FALSE;
2696 	}
2697 
2698 	/*
2699 	 * Open parent object set so we can inherit zplprop values.
2700 	 */
2701 	if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
2702 		return (error);
2703 
2704 	error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, createprops,
2705 	    zplprops, is_ci);
2706 	dmu_objset_rele(os, FTAG);
2707 	return (error);
2708 }
2709 
2710 static int
2711 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
2712     nvlist_t *zplprops, boolean_t *is_ci)
2713 {
2714 	boolean_t fuids_ok = B_TRUE;
2715 	uint64_t zplver = ZPL_VERSION;
2716 	int error;
2717 
2718 	if (spa_vers < SPA_VERSION_FUID) {
2719 		zplver = ZPL_VERSION_FUID - 1;
2720 		fuids_ok = B_FALSE;
2721 	}
2722 
2723 	error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, createprops,
2724 	    zplprops, is_ci);
2725 	return (error);
2726 }
2727 
2728 /*
2729  * inputs:
2730  * zc_objset_type	type of objset to create (fs vs zvol)
2731  * zc_name		name of new objset
2732  * zc_value		name of snapshot to clone from (may be empty)
2733  * zc_nvlist_src{_size}	nvlist of properties to apply
2734  *
2735  * outputs: none
2736  */
2737 static int
2738 zfs_ioc_create(zfs_cmd_t *zc)
2739 {
2740 	objset_t *clone;
2741 	int error = 0;
2742 	zfs_creat_t zct;
2743 	nvlist_t *nvprops = NULL;
2744 	void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
2745 	dmu_objset_type_t type = zc->zc_objset_type;
2746 
2747 	switch (type) {
2748 
2749 	case DMU_OST_ZFS:
2750 		cbfunc = zfs_create_cb;
2751 		break;
2752 
2753 	case DMU_OST_ZVOL:
2754 		cbfunc = zvol_create_cb;
2755 		break;
2756 
2757 	default:
2758 		cbfunc = NULL;
2759 		break;
2760 	}
2761 	if (strchr(zc->zc_name, '@') ||
2762 	    strchr(zc->zc_name, '%'))
2763 		return (EINVAL);
2764 
2765 	if (zc->zc_nvlist_src != NULL &&
2766 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2767 	    zc->zc_iflags, &nvprops)) != 0)
2768 		return (error);
2769 
2770 	zct.zct_zplprops = NULL;
2771 	zct.zct_props = nvprops;
2772 
2773 	if (zc->zc_value[0] != '\0') {
2774 		/*
2775 		 * We're creating a clone of an existing snapshot.
2776 		 */
2777 		zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
2778 		if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0) {
2779 			nvlist_free(nvprops);
2780 			return (EINVAL);
2781 		}
2782 
2783 		error = dmu_objset_hold(zc->zc_value, FTAG, &clone);
2784 		if (error) {
2785 			nvlist_free(nvprops);
2786 			return (error);
2787 		}
2788 
2789 		error = dmu_objset_clone(zc->zc_name, dmu_objset_ds(clone), 0);
2790 		dmu_objset_rele(clone, FTAG);
2791 		if (error) {
2792 			nvlist_free(nvprops);
2793 			return (error);
2794 		}
2795 	} else {
2796 		boolean_t is_insensitive = B_FALSE;
2797 
2798 		if (cbfunc == NULL) {
2799 			nvlist_free(nvprops);
2800 			return (EINVAL);
2801 		}
2802 
2803 		if (type == DMU_OST_ZVOL) {
2804 			uint64_t volsize, volblocksize;
2805 
2806 			if (nvprops == NULL ||
2807 			    nvlist_lookup_uint64(nvprops,
2808 			    zfs_prop_to_name(ZFS_PROP_VOLSIZE),
2809 			    &volsize) != 0) {
2810 				nvlist_free(nvprops);
2811 				return (EINVAL);
2812 			}
2813 
2814 			if ((error = nvlist_lookup_uint64(nvprops,
2815 			    zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
2816 			    &volblocksize)) != 0 && error != ENOENT) {
2817 				nvlist_free(nvprops);
2818 				return (EINVAL);
2819 			}
2820 
2821 			if (error != 0)
2822 				volblocksize = zfs_prop_default_numeric(
2823 				    ZFS_PROP_VOLBLOCKSIZE);
2824 
2825 			if ((error = zvol_check_volblocksize(
2826 			    volblocksize)) != 0 ||
2827 			    (error = zvol_check_volsize(volsize,
2828 			    volblocksize)) != 0) {
2829 				nvlist_free(nvprops);
2830 				return (error);
2831 			}
2832 		} else if (type == DMU_OST_ZFS) {
2833 			int error;
2834 
2835 			/*
2836 			 * We have to have normalization and
2837 			 * case-folding flags correct when we do the
2838 			 * file system creation, so go figure them out
2839 			 * now.
2840 			 */
2841 			VERIFY(nvlist_alloc(&zct.zct_zplprops,
2842 			    NV_UNIQUE_NAME, KM_SLEEP) == 0);
2843 			error = zfs_fill_zplprops(zc->zc_name, nvprops,
2844 			    zct.zct_zplprops, &is_insensitive);
2845 			if (error != 0) {
2846 				nvlist_free(nvprops);
2847 				nvlist_free(zct.zct_zplprops);
2848 				return (error);
2849 			}
2850 		}
2851 		error = dmu_objset_create(zc->zc_name, type,
2852 		    is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
2853 		nvlist_free(zct.zct_zplprops);
2854 	}
2855 
2856 	/*
2857 	 * It would be nice to do this atomically.
2858 	 */
2859 	if (error == 0) {
2860 		error = zfs_set_prop_nvlist(zc->zc_name, ZPROP_SRC_LOCAL,
2861 		    nvprops, NULL);
2862 		if (error != 0)
2863 			(void) dmu_objset_destroy(zc->zc_name, B_FALSE);
2864 	}
2865 	nvlist_free(nvprops);
2866 	return (error);
2867 }
2868 
2869 /*
2870  * inputs:
2871  * zc_name	name of filesystem
2872  * zc_value	short name of snapshot
2873  * zc_cookie	recursive flag
2874  * zc_nvlist_src[_size] property list
2875  *
2876  * outputs:
2877  * zc_value	short snapname (i.e. part after the '@')
2878  */
2879 static int
2880 zfs_ioc_snapshot(zfs_cmd_t *zc)
2881 {
2882 	nvlist_t *nvprops = NULL;
2883 	int error;
2884 	boolean_t recursive = zc->zc_cookie;
2885 
2886 	if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
2887 		return (EINVAL);
2888 
2889 	if (zc->zc_nvlist_src != NULL &&
2890 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2891 	    zc->zc_iflags, &nvprops)) != 0)
2892 		return (error);
2893 
2894 	error = zfs_check_userprops(zc->zc_name, nvprops);
2895 	if (error)
2896 		goto out;
2897 
2898 	if (!nvlist_empty(nvprops) &&
2899 	    zfs_earlier_version(zc->zc_name, SPA_VERSION_SNAP_PROPS)) {
2900 		error = ENOTSUP;
2901 		goto out;
2902 	}
2903 
2904 	error = dmu_objset_snapshot(zc->zc_name, zc->zc_value,
2905 	    nvprops, recursive);
2906 
2907 out:
2908 	nvlist_free(nvprops);
2909 	return (error);
2910 }
2911 
2912 int
2913 zfs_unmount_snap(const char *name, void *arg)
2914 {
2915 	vfs_t *vfsp = NULL;
2916 
2917 	if (arg) {
2918 		char *snapname = arg;
2919 		char *fullname = kmem_asprintf("%s@%s", name, snapname);
2920 		vfsp = zfs_get_vfs(fullname);
2921 		strfree(fullname);
2922 	} else if (strchr(name, '@')) {
2923 		vfsp = zfs_get_vfs(name);
2924 	}
2925 
2926 	if (vfsp) {
2927 		/*
2928 		 * Always force the unmount for snapshots.
2929 		 */
2930 		int flag = MS_FORCE;
2931 		int err;
2932 
2933 		if ((err = vn_vfswlock(vfsp->vfs_vnodecovered)) != 0) {
2934 			VFS_RELE(vfsp);
2935 			return (err);
2936 		}
2937 		VFS_RELE(vfsp);
2938 		if ((err = dounmount(vfsp, flag, kcred)) != 0)
2939 			return (err);
2940 	}
2941 	return (0);
2942 }
2943 
2944 /*
2945  * inputs:
2946  * zc_name		name of filesystem
2947  * zc_value		short name of snapshot
2948  * zc_defer_destroy	mark for deferred destroy
2949  *
2950  * outputs:	none
2951  */
2952 static int
2953 zfs_ioc_destroy_snaps(zfs_cmd_t *zc)
2954 {
2955 	int err;
2956 
2957 	if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
2958 		return (EINVAL);
2959 	err = dmu_objset_find(zc->zc_name,
2960 	    zfs_unmount_snap, zc->zc_value, DS_FIND_CHILDREN);
2961 	if (err)
2962 		return (err);
2963 	return (dmu_snapshots_destroy(zc->zc_name, zc->zc_value,
2964 	    zc->zc_defer_destroy));
2965 }
2966 
2967 /*
2968  * inputs:
2969  * zc_name		name of dataset to destroy
2970  * zc_objset_type	type of objset
2971  * zc_defer_destroy	mark for deferred destroy
2972  *
2973  * outputs:		none
2974  */
2975 static int
2976 zfs_ioc_destroy(zfs_cmd_t *zc)
2977 {
2978 	int err;
2979 	if (strchr(zc->zc_name, '@') && zc->zc_objset_type == DMU_OST_ZFS) {
2980 		err = zfs_unmount_snap(zc->zc_name, NULL);
2981 		if (err)
2982 			return (err);
2983 	}
2984 
2985 	err = dmu_objset_destroy(zc->zc_name, zc->zc_defer_destroy);
2986 	if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
2987 		(void) zvol_remove_minor(zc->zc_name);
2988 	return (err);
2989 }
2990 
2991 /*
2992  * inputs:
2993  * zc_name	name of dataset to rollback (to most recent snapshot)
2994  *
2995  * outputs:	none
2996  */
2997 static int
2998 zfs_ioc_rollback(zfs_cmd_t *zc)
2999 {
3000 	dsl_dataset_t *ds, *clone;
3001 	int error;
3002 	zfsvfs_t *zfsvfs;
3003 	char *clone_name;
3004 
3005 	error = dsl_dataset_hold(zc->zc_name, FTAG, &ds);
3006 	if (error)
3007 		return (error);
3008 
3009 	/* must not be a snapshot */
3010 	if (dsl_dataset_is_snapshot(ds)) {
3011 		dsl_dataset_rele(ds, FTAG);
3012 		return (EINVAL);
3013 	}
3014 
3015 	/* must have a most recent snapshot */
3016 	if (ds->ds_phys->ds_prev_snap_txg < TXG_INITIAL) {
3017 		dsl_dataset_rele(ds, FTAG);
3018 		return (EINVAL);
3019 	}
3020 
3021 	/*
3022 	 * Create clone of most recent snapshot.
3023 	 */
3024 	clone_name = kmem_asprintf("%s/%%rollback", zc->zc_name);
3025 	error = dmu_objset_clone(clone_name, ds->ds_prev, DS_FLAG_INCONSISTENT);
3026 	if (error)
3027 		goto out;
3028 
3029 	error = dsl_dataset_own(clone_name, B_TRUE, FTAG, &clone);
3030 	if (error)
3031 		goto out;
3032 
3033 	/*
3034 	 * Do clone swap.
3035 	 */
3036 	if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
3037 		error = zfs_suspend_fs(zfsvfs);
3038 		if (error == 0) {
3039 			int resume_err;
3040 
3041 			if (dsl_dataset_tryown(ds, B_FALSE, FTAG)) {
3042 				error = dsl_dataset_clone_swap(clone, ds,
3043 				    B_TRUE);
3044 				dsl_dataset_disown(ds, FTAG);
3045 				ds = NULL;
3046 			} else {
3047 				error = EBUSY;
3048 			}
3049 			resume_err = zfs_resume_fs(zfsvfs, zc->zc_name);
3050 			error = error ? error : resume_err;
3051 		}
3052 		VFS_RELE(zfsvfs->z_vfs);
3053 	} else {
3054 		if (dsl_dataset_tryown(ds, B_FALSE, FTAG)) {
3055 			error = dsl_dataset_clone_swap(clone, ds, B_TRUE);
3056 			dsl_dataset_disown(ds, FTAG);
3057 			ds = NULL;
3058 		} else {
3059 			error = EBUSY;
3060 		}
3061 	}
3062 
3063 	/*
3064 	 * Destroy clone (which also closes it).
3065 	 */
3066 	(void) dsl_dataset_destroy(clone, FTAG, B_FALSE);
3067 
3068 out:
3069 	strfree(clone_name);
3070 	if (ds)
3071 		dsl_dataset_rele(ds, FTAG);
3072 	return (error);
3073 }
3074 
3075 /*
3076  * inputs:
3077  * zc_name	old name of dataset
3078  * zc_value	new name of dataset
3079  * zc_cookie	recursive flag (only valid for snapshots)
3080  *
3081  * outputs:	none
3082  */
3083 static int
3084 zfs_ioc_rename(zfs_cmd_t *zc)
3085 {
3086 	boolean_t recursive = zc->zc_cookie & 1;
3087 
3088 	zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3089 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3090 	    strchr(zc->zc_value, '%'))
3091 		return (EINVAL);
3092 
3093 	/*
3094 	 * Unmount snapshot unless we're doing a recursive rename,
3095 	 * in which case the dataset code figures out which snapshots
3096 	 * to unmount.
3097 	 */
3098 	if (!recursive && strchr(zc->zc_name, '@') != NULL &&
3099 	    zc->zc_objset_type == DMU_OST_ZFS) {
3100 		int err = zfs_unmount_snap(zc->zc_name, NULL);
3101 		if (err)
3102 			return (err);
3103 	}
3104 	if (zc->zc_objset_type == DMU_OST_ZVOL)
3105 		(void) zvol_remove_minor(zc->zc_name);
3106 	return (dmu_objset_rename(zc->zc_name, zc->zc_value, recursive));
3107 }
3108 
3109 static int
3110 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3111 {
3112 	const char *propname = nvpair_name(pair);
3113 	boolean_t issnap = (strchr(dsname, '@') != NULL);
3114 	zfs_prop_t prop = zfs_name_to_prop(propname);
3115 	uint64_t intval;
3116 	int err;
3117 
3118 	if (prop == ZPROP_INVAL) {
3119 		if (zfs_prop_user(propname)) {
3120 			if (err = zfs_secpolicy_write_perms(dsname,
3121 			    ZFS_DELEG_PERM_USERPROP, cr))
3122 				return (err);
3123 			return (0);
3124 		}
3125 
3126 		if (!issnap && zfs_prop_userquota(propname)) {
3127 			const char *perm = NULL;
3128 			const char *uq_prefix =
3129 			    zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3130 			const char *gq_prefix =
3131 			    zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3132 
3133 			if (strncmp(propname, uq_prefix,
3134 			    strlen(uq_prefix)) == 0) {
3135 				perm = ZFS_DELEG_PERM_USERQUOTA;
3136 			} else if (strncmp(propname, gq_prefix,
3137 			    strlen(gq_prefix)) == 0) {
3138 				perm = ZFS_DELEG_PERM_GROUPQUOTA;
3139 			} else {
3140 				/* USERUSED and GROUPUSED are read-only */
3141 				return (EINVAL);
3142 			}
3143 
3144 			if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3145 				return (err);
3146 			return (0);
3147 		}
3148 
3149 		return (EINVAL);
3150 	}
3151 
3152 	if (issnap)
3153 		return (EINVAL);
3154 
3155 	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3156 		/*
3157 		 * dsl_prop_get_all_impl() returns properties in this
3158 		 * format.
3159 		 */
3160 		nvlist_t *attrs;
3161 		VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3162 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3163 		    &pair) == 0);
3164 	}
3165 
3166 	/*
3167 	 * Check that this value is valid for this pool version
3168 	 */
3169 	switch (prop) {
3170 	case ZFS_PROP_COMPRESSION:
3171 		/*
3172 		 * If the user specified gzip compression, make sure
3173 		 * the SPA supports it. We ignore any errors here since
3174 		 * we'll catch them later.
3175 		 */
3176 		if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3177 		    nvpair_value_uint64(pair, &intval) == 0) {
3178 			if (intval >= ZIO_COMPRESS_GZIP_1 &&
3179 			    intval <= ZIO_COMPRESS_GZIP_9 &&
3180 			    zfs_earlier_version(dsname,
3181 			    SPA_VERSION_GZIP_COMPRESSION)) {
3182 				return (ENOTSUP);
3183 			}
3184 
3185 			if (intval == ZIO_COMPRESS_ZLE &&
3186 			    zfs_earlier_version(dsname,
3187 			    SPA_VERSION_ZLE_COMPRESSION))
3188 				return (ENOTSUP);
3189 
3190 			/*
3191 			 * If this is a bootable dataset then
3192 			 * verify that the compression algorithm
3193 			 * is supported for booting. We must return
3194 			 * something other than ENOTSUP since it
3195 			 * implies a downrev pool version.
3196 			 */
3197 			if (zfs_is_bootfs(dsname) &&
3198 			    !BOOTFS_COMPRESS_VALID(intval)) {
3199 				return (ERANGE);
3200 			}
3201 		}
3202 		break;
3203 
3204 	case ZFS_PROP_COPIES:
3205 		if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3206 			return (ENOTSUP);
3207 		break;
3208 
3209 	case ZFS_PROP_DEDUP:
3210 		if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3211 			return (ENOTSUP);
3212 		break;
3213 
3214 	case ZFS_PROP_SHARESMB:
3215 		if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3216 			return (ENOTSUP);
3217 		break;
3218 
3219 	case ZFS_PROP_ACLINHERIT:
3220 		if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3221 		    nvpair_value_uint64(pair, &intval) == 0) {
3222 			if (intval == ZFS_ACL_PASSTHROUGH_X &&
3223 			    zfs_earlier_version(dsname,
3224 			    SPA_VERSION_PASSTHROUGH_X))
3225 				return (ENOTSUP);
3226 		}
3227 		break;
3228 	}
3229 
3230 	return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3231 }
3232 
3233 /*
3234  * Removes properties from the given props list that fail permission checks
3235  * needed to clear them and to restore them in case of a receive error. For each
3236  * property, make sure we have both set and inherit permissions.
3237  *
3238  * Returns the first error encountered if any permission checks fail. If the
3239  * caller provides a non-NULL errlist, it also gives the complete list of names
3240  * of all the properties that failed a permission check along with the
3241  * corresponding error numbers. The caller is responsible for freeing the
3242  * returned errlist.
3243  *
3244  * If every property checks out successfully, zero is returned and the list
3245  * pointed at by errlist is NULL.
3246  */
3247 static int
3248 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
3249 {
3250 	zfs_cmd_t *zc;
3251 	nvpair_t *pair, *next_pair;
3252 	nvlist_t *errors;
3253 	int err, rv = 0;
3254 
3255 	if (props == NULL)
3256 		return (0);
3257 
3258 	VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3259 
3260 	zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
3261 	(void) strcpy(zc->zc_name, dataset);
3262 	pair = nvlist_next_nvpair(props, NULL);
3263 	while (pair != NULL) {
3264 		next_pair = nvlist_next_nvpair(props, pair);
3265 
3266 		(void) strcpy(zc->zc_value, nvpair_name(pair));
3267 		if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
3268 		    (err = zfs_secpolicy_inherit(zc, CRED())) != 0) {
3269 			VERIFY(nvlist_remove_nvpair(props, pair) == 0);
3270 			VERIFY(nvlist_add_int32(errors,
3271 			    zc->zc_value, err) == 0);
3272 		}
3273 		pair = next_pair;
3274 	}
3275 	kmem_free(zc, sizeof (zfs_cmd_t));
3276 
3277 	if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
3278 		nvlist_free(errors);
3279 		errors = NULL;
3280 	} else {
3281 		VERIFY(nvpair_value_int32(pair, &rv) == 0);
3282 	}
3283 
3284 	if (errlist == NULL)
3285 		nvlist_free(errors);
3286 	else
3287 		*errlist = errors;
3288 
3289 	return (rv);
3290 }
3291 
3292 static boolean_t
3293 propval_equals(nvpair_t *p1, nvpair_t *p2)
3294 {
3295 	if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
3296 		/* dsl_prop_get_all_impl() format */
3297 		nvlist_t *attrs;
3298 		VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
3299 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3300 		    &p1) == 0);
3301 	}
3302 
3303 	if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
3304 		nvlist_t *attrs;
3305 		VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
3306 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3307 		    &p2) == 0);
3308 	}
3309 
3310 	if (nvpair_type(p1) != nvpair_type(p2))
3311 		return (B_FALSE);
3312 
3313 	if (nvpair_type(p1) == DATA_TYPE_STRING) {
3314 		char *valstr1, *valstr2;
3315 
3316 		VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
3317 		VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
3318 		return (strcmp(valstr1, valstr2) == 0);
3319 	} else {
3320 		uint64_t intval1, intval2;
3321 
3322 		VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
3323 		VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
3324 		return (intval1 == intval2);
3325 	}
3326 }
3327 
3328 /*
3329  * Remove properties from props if they are not going to change (as determined
3330  * by comparison with origprops). Remove them from origprops as well, since we
3331  * do not need to clear or restore properties that won't change.
3332  */
3333 static void
3334 props_reduce(nvlist_t *props, nvlist_t *origprops)
3335 {
3336 	nvpair_t *pair, *next_pair;
3337 
3338 	if (origprops == NULL)
3339 		return; /* all props need to be received */
3340 
3341 	pair = nvlist_next_nvpair(props, NULL);
3342 	while (pair != NULL) {
3343 		const char *propname = nvpair_name(pair);
3344 		nvpair_t *match;
3345 
3346 		next_pair = nvlist_next_nvpair(props, pair);
3347 
3348 		if ((nvlist_lookup_nvpair(origprops, propname,
3349 		    &match) != 0) || !propval_equals(pair, match))
3350 			goto next; /* need to set received value */
3351 
3352 		/* don't clear the existing received value */
3353 		(void) nvlist_remove_nvpair(origprops, match);
3354 		/* don't bother receiving the property */
3355 		(void) nvlist_remove_nvpair(props, pair);
3356 next:
3357 		pair = next_pair;
3358 	}
3359 }
3360 
3361 #ifdef	DEBUG
3362 static boolean_t zfs_ioc_recv_inject_err;
3363 #endif
3364 
3365 /*
3366  * inputs:
3367  * zc_name		name of containing filesystem
3368  * zc_nvlist_src{_size}	nvlist of properties to apply
3369  * zc_value		name of snapshot to create
3370  * zc_string		name of clone origin (if DRR_FLAG_CLONE)
3371  * zc_cookie		file descriptor to recv from
3372  * zc_begin_record	the BEGIN record of the stream (not byteswapped)
3373  * zc_guid		force flag
3374  *
3375  * outputs:
3376  * zc_cookie		number of bytes read
3377  * zc_nvlist_dst{_size} error for each unapplied received property
3378  * zc_obj		zprop_errflags_t
3379  */
3380 static int
3381 zfs_ioc_recv(zfs_cmd_t *zc)
3382 {
3383 	file_t *fp;
3384 	objset_t *os;
3385 	dmu_recv_cookie_t drc;
3386 	boolean_t force = (boolean_t)zc->zc_guid;
3387 	int fd;
3388 	int error = 0;
3389 	int props_error = 0;
3390 	nvlist_t *errors;
3391 	offset_t off;
3392 	nvlist_t *props = NULL; /* sent properties */
3393 	nvlist_t *origprops = NULL; /* existing properties */
3394 	objset_t *origin = NULL;
3395 	char *tosnap;
3396 	char tofs[ZFS_MAXNAMELEN];
3397 	boolean_t first_recvd_props = B_FALSE;
3398 
3399 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3400 	    strchr(zc->zc_value, '@') == NULL ||
3401 	    strchr(zc->zc_value, '%'))
3402 		return (EINVAL);
3403 
3404 	(void) strcpy(tofs, zc->zc_value);
3405 	tosnap = strchr(tofs, '@');
3406 	*tosnap++ = '\0';
3407 
3408 	if (zc->zc_nvlist_src != NULL &&
3409 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3410 	    zc->zc_iflags, &props)) != 0)
3411 		return (error);
3412 
3413 	fd = zc->zc_cookie;
3414 	fp = getf(fd);
3415 	if (fp == NULL) {
3416 		nvlist_free(props);
3417 		return (EBADF);
3418 	}
3419 
3420 	VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3421 
3422 	if (props && dmu_objset_hold(tofs, FTAG, &os) == 0) {
3423 		if ((spa_version(os->os_spa) >= SPA_VERSION_RECVD_PROPS) &&
3424 		    !dsl_prop_get_hasrecvd(os)) {
3425 			first_recvd_props = B_TRUE;
3426 		}
3427 
3428 		/*
3429 		 * If new received properties are supplied, they are to
3430 		 * completely replace the existing received properties, so stash
3431 		 * away the existing ones.
3432 		 */
3433 		if (dsl_prop_get_received(os, &origprops) == 0) {
3434 			nvlist_t *errlist = NULL;
3435 			/*
3436 			 * Don't bother writing a property if its value won't
3437 			 * change (and avoid the unnecessary security checks).
3438 			 *
3439 			 * The first receive after SPA_VERSION_RECVD_PROPS is a
3440 			 * special case where we blow away all local properties
3441 			 * regardless.
3442 			 */
3443 			if (!first_recvd_props)
3444 				props_reduce(props, origprops);
3445 			if (zfs_check_clearable(tofs, origprops,
3446 			    &errlist) != 0)
3447 				(void) nvlist_merge(errors, errlist, 0);
3448 			nvlist_free(errlist);
3449 		}
3450 
3451 		dmu_objset_rele(os, FTAG);
3452 	}
3453 
3454 	if (zc->zc_string[0]) {
3455 		error = dmu_objset_hold(zc->zc_string, FTAG, &origin);
3456 		if (error)
3457 			goto out;
3458 	}
3459 
3460 	error = dmu_recv_begin(tofs, tosnap, zc->zc_top_ds,
3461 	    &zc->zc_begin_record, force, origin, &drc);
3462 	if (origin)
3463 		dmu_objset_rele(origin, FTAG);
3464 	if (error)
3465 		goto out;
3466 
3467 	/*
3468 	 * Set properties before we receive the stream so that they are applied
3469 	 * to the new data. Note that we must call dmu_recv_stream() if
3470 	 * dmu_recv_begin() succeeds.
3471 	 */
3472 	if (props) {
3473 		nvlist_t *errlist;
3474 
3475 		if (dmu_objset_from_ds(drc.drc_logical_ds, &os) == 0) {
3476 			if (drc.drc_newfs) {
3477 				if (spa_version(os->os_spa) >=
3478 				    SPA_VERSION_RECVD_PROPS)
3479 					first_recvd_props = B_TRUE;
3480 			} else if (origprops != NULL) {
3481 				if (clear_received_props(os, tofs, origprops,
3482 				    first_recvd_props ? NULL : props) != 0)
3483 					zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3484 			} else {
3485 				zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3486 			}
3487 			dsl_prop_set_hasrecvd(os);
3488 		} else if (!drc.drc_newfs) {
3489 			zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3490 		}
3491 
3492 		(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
3493 		    props, &errlist);
3494 		(void) nvlist_merge(errors, errlist, 0);
3495 		nvlist_free(errlist);
3496 	}
3497 
3498 	if (fit_error_list(zc, &errors) != 0 || put_nvlist(zc, errors) != 0) {
3499 		/*
3500 		 * Caller made zc->zc_nvlist_dst less than the minimum expected
3501 		 * size or supplied an invalid address.
3502 		 */
3503 		props_error = EINVAL;
3504 	}
3505 
3506 	off = fp->f_offset;
3507 	error = dmu_recv_stream(&drc, fp->f_vnode, &off);
3508 
3509 	if (error == 0) {
3510 		zfsvfs_t *zfsvfs = NULL;
3511 
3512 		if (getzfsvfs(tofs, &zfsvfs) == 0) {
3513 			/* online recv */
3514 			int end_err;
3515 
3516 			error = zfs_suspend_fs(zfsvfs);
3517 			/*
3518 			 * If the suspend fails, then the recv_end will
3519 			 * likely also fail, and clean up after itself.
3520 			 */
3521 			end_err = dmu_recv_end(&drc);
3522 			if (error == 0) {
3523 				int resume_err =
3524 				    zfs_resume_fs(zfsvfs, tofs);
3525 				error = error ? error : resume_err;
3526 			}
3527 			error = error ? error : end_err;
3528 			VFS_RELE(zfsvfs->z_vfs);
3529 		} else {
3530 			error = dmu_recv_end(&drc);
3531 		}
3532 	}
3533 
3534 	zc->zc_cookie = off - fp->f_offset;
3535 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
3536 		fp->f_offset = off;
3537 
3538 #ifdef	DEBUG
3539 	if (zfs_ioc_recv_inject_err) {
3540 		zfs_ioc_recv_inject_err = B_FALSE;
3541 		error = 1;
3542 	}
3543 #endif
3544 	/*
3545 	 * On error, restore the original props.
3546 	 */
3547 	if (error && props) {
3548 		if (dmu_objset_hold(tofs, FTAG, &os) == 0) {
3549 			if (clear_received_props(os, tofs, props, NULL) != 0) {
3550 				/*
3551 				 * We failed to clear the received properties.
3552 				 * Since we may have left a $recvd value on the
3553 				 * system, we can't clear the $hasrecvd flag.
3554 				 */
3555 				zc->zc_obj |= ZPROP_ERR_NORESTORE;
3556 			} else if (first_recvd_props) {
3557 				dsl_prop_unset_hasrecvd(os);
3558 			}
3559 			dmu_objset_rele(os, FTAG);
3560 		} else if (!drc.drc_newfs) {
3561 			/* We failed to clear the received properties. */
3562 			zc->zc_obj |= ZPROP_ERR_NORESTORE;
3563 		}
3564 
3565 		if (origprops == NULL && !drc.drc_newfs) {
3566 			/* We failed to stash the original properties. */
3567 			zc->zc_obj |= ZPROP_ERR_NORESTORE;
3568 		}
3569 
3570 		/*
3571 		 * dsl_props_set() will not convert RECEIVED to LOCAL on or
3572 		 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
3573 		 * explictly if we're restoring local properties cleared in the
3574 		 * first new-style receive.
3575 		 */
3576 		if (origprops != NULL &&
3577 		    zfs_set_prop_nvlist(tofs, (first_recvd_props ?
3578 		    ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
3579 		    origprops, NULL) != 0) {
3580 			/*
3581 			 * We stashed the original properties but failed to
3582 			 * restore them.
3583 			 */
3584 			zc->zc_obj |= ZPROP_ERR_NORESTORE;
3585 		}
3586 	}
3587 out:
3588 	nvlist_free(props);
3589 	nvlist_free(origprops);
3590 	nvlist_free(errors);
3591 	releasef(fd);
3592 
3593 	if (error == 0)
3594 		error = props_error;
3595 
3596 	return (error);
3597 }
3598 
3599 /*
3600  * inputs:
3601  * zc_name	name of snapshot to send
3602  * zc_value	short name of incremental fromsnap (may be empty)
3603  * zc_cookie	file descriptor to send stream to
3604  * zc_obj	fromorigin flag (mutually exclusive with zc_value)
3605  *
3606  * outputs: none
3607  */
3608 static int
3609 zfs_ioc_send(zfs_cmd_t *zc)
3610 {
3611 	objset_t *fromsnap = NULL;
3612 	objset_t *tosnap;
3613 	file_t *fp;
3614 	int error;
3615 	offset_t off;
3616 
3617 	error = dmu_objset_hold(zc->zc_name, FTAG, &tosnap);
3618 	if (error)
3619 		return (error);
3620 
3621 	if (zc->zc_value[0] != '\0') {
3622 		char *buf;
3623 		char *cp;
3624 
3625 		buf = kmem_alloc(MAXPATHLEN, KM_SLEEP);
3626 		(void) strncpy(buf, zc->zc_name, MAXPATHLEN);
3627 		cp = strchr(buf, '@');
3628 		if (cp)
3629 			*(cp+1) = 0;
3630 		(void) strncat(buf, zc->zc_value, MAXPATHLEN);
3631 		error = dmu_objset_hold(buf, FTAG, &fromsnap);
3632 		kmem_free(buf, MAXPATHLEN);
3633 		if (error) {
3634 			dmu_objset_rele(tosnap, FTAG);
3635 			return (error);
3636 		}
3637 	}
3638 
3639 	fp = getf(zc->zc_cookie);
3640 	if (fp == NULL) {
3641 		dmu_objset_rele(tosnap, FTAG);
3642 		if (fromsnap)
3643 			dmu_objset_rele(fromsnap, FTAG);
3644 		return (EBADF);
3645 	}
3646 
3647 	off = fp->f_offset;
3648 	error = dmu_sendbackup(tosnap, fromsnap, zc->zc_obj, fp->f_vnode, &off);
3649 
3650 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
3651 		fp->f_offset = off;
3652 	releasef(zc->zc_cookie);
3653 	if (fromsnap)
3654 		dmu_objset_rele(fromsnap, FTAG);
3655 	dmu_objset_rele(tosnap, FTAG);
3656 	return (error);
3657 }
3658 
3659 static int
3660 zfs_ioc_inject_fault(zfs_cmd_t *zc)
3661 {
3662 	int id, error;
3663 
3664 	error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
3665 	    &zc->zc_inject_record);
3666 
3667 	if (error == 0)
3668 		zc->zc_guid = (uint64_t)id;
3669 
3670 	return (error);
3671 }
3672 
3673 static int
3674 zfs_ioc_clear_fault(zfs_cmd_t *zc)
3675 {
3676 	return (zio_clear_fault((int)zc->zc_guid));
3677 }
3678 
3679 static int
3680 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
3681 {
3682 	int id = (int)zc->zc_guid;
3683 	int error;
3684 
3685 	error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
3686 	    &zc->zc_inject_record);
3687 
3688 	zc->zc_guid = id;
3689 
3690 	return (error);
3691 }
3692 
3693 static int
3694 zfs_ioc_error_log(zfs_cmd_t *zc)
3695 {
3696 	spa_t *spa;
3697 	int error;
3698 	size_t count = (size_t)zc->zc_nvlist_dst_size;
3699 
3700 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
3701 		return (error);
3702 
3703 	error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
3704 	    &count);
3705 	if (error == 0)
3706 		zc->zc_nvlist_dst_size = count;
3707 	else
3708 		zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
3709 
3710 	spa_close(spa, FTAG);
3711 
3712 	return (error);
3713 }
3714 
3715 static int
3716 zfs_ioc_clear(zfs_cmd_t *zc)
3717 {
3718 	spa_t *spa;
3719 	vdev_t *vd;
3720 	int error;
3721 
3722 	/*
3723 	 * On zpool clear we also fix up missing slogs
3724 	 */
3725 	mutex_enter(&spa_namespace_lock);
3726 	spa = spa_lookup(zc->zc_name);
3727 	if (spa == NULL) {
3728 		mutex_exit(&spa_namespace_lock);
3729 		return (EIO);
3730 	}
3731 	if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
3732 		/* we need to let spa_open/spa_load clear the chains */
3733 		spa_set_log_state(spa, SPA_LOG_CLEAR);
3734 	}
3735 	spa->spa_last_open_failed = 0;
3736 	mutex_exit(&spa_namespace_lock);
3737 
3738 	if (zc->zc_cookie & ZPOOL_NO_REWIND) {
3739 		error = spa_open(zc->zc_name, &spa, FTAG);
3740 	} else {
3741 		nvlist_t *policy;
3742 		nvlist_t *config = NULL;
3743 
3744 		if (zc->zc_nvlist_src == NULL)
3745 			return (EINVAL);
3746 
3747 		if ((error = get_nvlist(zc->zc_nvlist_src,
3748 		    zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
3749 			error = spa_open_rewind(zc->zc_name, &spa, FTAG,
3750 			    policy, &config);
3751 			if (config != NULL) {
3752 				(void) put_nvlist(zc, config);
3753 				nvlist_free(config);
3754 			}
3755 			nvlist_free(policy);
3756 		}
3757 	}
3758 
3759 	if (error)
3760 		return (error);
3761 
3762 	spa_vdev_state_enter(spa, SCL_NONE);
3763 
3764 	if (zc->zc_guid == 0) {
3765 		vd = NULL;
3766 	} else {
3767 		vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
3768 		if (vd == NULL) {
3769 			(void) spa_vdev_state_exit(spa, NULL, ENODEV);
3770 			spa_close(spa, FTAG);
3771 			return (ENODEV);
3772 		}
3773 	}
3774 
3775 	vdev_clear(spa, vd);
3776 
3777 	(void) spa_vdev_state_exit(spa, NULL, 0);
3778 
3779 	/*
3780 	 * Resume any suspended I/Os.
3781 	 */
3782 	if (zio_resume(spa) != 0)
3783 		error = EIO;
3784 
3785 	spa_close(spa, FTAG);
3786 
3787 	return (error);
3788 }
3789 
3790 /*
3791  * inputs:
3792  * zc_name	name of filesystem
3793  * zc_value	name of origin snapshot
3794  *
3795  * outputs:
3796  * zc_string	name of conflicting snapshot, if there is one
3797  */
3798 static int
3799 zfs_ioc_promote(zfs_cmd_t *zc)
3800 {
3801 	char *cp;
3802 
3803 	/*
3804 	 * We don't need to unmount *all* the origin fs's snapshots, but
3805 	 * it's easier.
3806 	 */
3807 	cp = strchr(zc->zc_value, '@');
3808 	if (cp)
3809 		*cp = '\0';
3810 	(void) dmu_objset_find(zc->zc_value,
3811 	    zfs_unmount_snap, NULL, DS_FIND_SNAPSHOTS);
3812 	return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
3813 }
3814 
3815 /*
3816  * Retrieve a single {user|group}{used|quota}@... property.
3817  *
3818  * inputs:
3819  * zc_name	name of filesystem
3820  * zc_objset_type zfs_userquota_prop_t
3821  * zc_value	domain name (eg. "S-1-234-567-89")
3822  * zc_guid	RID/UID/GID
3823  *
3824  * outputs:
3825  * zc_cookie	property value
3826  */
3827 static int
3828 zfs_ioc_userspace_one(zfs_cmd_t *zc)
3829 {
3830 	zfsvfs_t *zfsvfs;
3831 	int error;
3832 
3833 	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
3834 		return (EINVAL);
3835 
3836 	error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs);
3837 	if (error)
3838 		return (error);
3839 
3840 	error = zfs_userspace_one(zfsvfs,
3841 	    zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
3842 	zfsvfs_rele(zfsvfs, FTAG);
3843 
3844 	return (error);
3845 }
3846 
3847 /*
3848  * inputs:
3849  * zc_name		name of filesystem
3850  * zc_cookie		zap cursor
3851  * zc_objset_type	zfs_userquota_prop_t
3852  * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
3853  *
3854  * outputs:
3855  * zc_nvlist_dst[_size]	data buffer (array of zfs_useracct_t)
3856  * zc_cookie	zap cursor
3857  */
3858 static int
3859 zfs_ioc_userspace_many(zfs_cmd_t *zc)
3860 {
3861 	zfsvfs_t *zfsvfs;
3862 	int error;
3863 
3864 	error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs);
3865 	if (error)
3866 		return (error);
3867 
3868 	int bufsize = zc->zc_nvlist_dst_size;
3869 	void *buf = kmem_alloc(bufsize, KM_SLEEP);
3870 
3871 	error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
3872 	    buf, &zc->zc_nvlist_dst_size);
3873 
3874 	if (error == 0) {
3875 		error = xcopyout(buf,
3876 		    (void *)(uintptr_t)zc->zc_nvlist_dst,
3877 		    zc->zc_nvlist_dst_size);
3878 	}
3879 	kmem_free(buf, bufsize);
3880 	zfsvfs_rele(zfsvfs, FTAG);
3881 
3882 	return (error);
3883 }
3884 
3885 /*
3886  * inputs:
3887  * zc_name		name of filesystem
3888  *
3889  * outputs:
3890  * none
3891  */
3892 static int
3893 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
3894 {
3895 	objset_t *os;
3896 	int error = 0;
3897 	zfsvfs_t *zfsvfs;
3898 
3899 	if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
3900 		if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
3901 			/*
3902 			 * If userused is not enabled, it may be because the
3903 			 * objset needs to be closed & reopened (to grow the
3904 			 * objset_phys_t).  Suspend/resume the fs will do that.
3905 			 */
3906 			error = zfs_suspend_fs(zfsvfs);
3907 			if (error == 0)
3908 				error = zfs_resume_fs(zfsvfs, zc->zc_name);
3909 		}
3910 		if (error == 0)
3911 			error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
3912 		VFS_RELE(zfsvfs->z_vfs);
3913 	} else {
3914 		/* XXX kind of reading contents without owning */
3915 		error = dmu_objset_hold(zc->zc_name, FTAG, &os);
3916 		if (error)
3917 			return (error);
3918 
3919 		error = dmu_objset_userspace_upgrade(os);
3920 		dmu_objset_rele(os, FTAG);
3921 	}
3922 
3923 	return (error);
3924 }
3925 
3926 /*
3927  * We don't want to have a hard dependency
3928  * against some special symbols in sharefs
3929  * nfs, and smbsrv.  Determine them if needed when
3930  * the first file system is shared.
3931  * Neither sharefs, nfs or smbsrv are unloadable modules.
3932  */
3933 int (*znfsexport_fs)(void *arg);
3934 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
3935 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
3936 
3937 int zfs_nfsshare_inited;
3938 int zfs_smbshare_inited;
3939 
3940 ddi_modhandle_t nfs_mod;
3941 ddi_modhandle_t sharefs_mod;
3942 ddi_modhandle_t smbsrv_mod;
3943 kmutex_t zfs_share_lock;
3944 
3945 static int
3946 zfs_init_sharefs()
3947 {
3948 	int error;
3949 
3950 	ASSERT(MUTEX_HELD(&zfs_share_lock));
3951 	/* Both NFS and SMB shares also require sharetab support. */
3952 	if (sharefs_mod == NULL && ((sharefs_mod =
3953 	    ddi_modopen("fs/sharefs",
3954 	    KRTLD_MODE_FIRST, &error)) == NULL)) {
3955 		return (ENOSYS);
3956 	}
3957 	if (zshare_fs == NULL && ((zshare_fs =
3958 	    (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
3959 	    ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
3960 		return (ENOSYS);
3961 	}
3962 	return (0);
3963 }
3964 
3965 static int
3966 zfs_ioc_share(zfs_cmd_t *zc)
3967 {
3968 	int error;
3969 	int opcode;
3970 
3971 	switch (zc->zc_share.z_sharetype) {
3972 	case ZFS_SHARE_NFS:
3973 	case ZFS_UNSHARE_NFS:
3974 		if (zfs_nfsshare_inited == 0) {
3975 			mutex_enter(&zfs_share_lock);
3976 			if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
3977 			    KRTLD_MODE_FIRST, &error)) == NULL)) {
3978 				mutex_exit(&zfs_share_lock);
3979 				return (ENOSYS);
3980 			}
3981 			if (znfsexport_fs == NULL &&
3982 			    ((znfsexport_fs = (int (*)(void *))
3983 			    ddi_modsym(nfs_mod,
3984 			    "nfs_export", &error)) == NULL)) {
3985 				mutex_exit(&zfs_share_lock);
3986 				return (ENOSYS);
3987 			}
3988 			error = zfs_init_sharefs();
3989 			if (error) {
3990 				mutex_exit(&zfs_share_lock);
3991 				return (ENOSYS);
3992 			}
3993 			zfs_nfsshare_inited = 1;
3994 			mutex_exit(&zfs_share_lock);
3995 		}
3996 		break;
3997 	case ZFS_SHARE_SMB:
3998 	case ZFS_UNSHARE_SMB:
3999 		if (zfs_smbshare_inited == 0) {
4000 			mutex_enter(&zfs_share_lock);
4001 			if (smbsrv_mod == NULL && ((smbsrv_mod =
4002 			    ddi_modopen("drv/smbsrv",
4003 			    KRTLD_MODE_FIRST, &error)) == NULL)) {
4004 				mutex_exit(&zfs_share_lock);
4005 				return (ENOSYS);
4006 			}
4007 			if (zsmbexport_fs == NULL && ((zsmbexport_fs =
4008 			    (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
4009 			    "smb_server_share", &error)) == NULL)) {
4010 				mutex_exit(&zfs_share_lock);
4011 				return (ENOSYS);
4012 			}
4013 			error = zfs_init_sharefs();
4014 			if (error) {
4015 				mutex_exit(&zfs_share_lock);
4016 				return (ENOSYS);
4017 			}
4018 			zfs_smbshare_inited = 1;
4019 			mutex_exit(&zfs_share_lock);
4020 		}
4021 		break;
4022 	default:
4023 		return (EINVAL);
4024 	}
4025 
4026 	switch (zc->zc_share.z_sharetype) {
4027 	case ZFS_SHARE_NFS:
4028 	case ZFS_UNSHARE_NFS:
4029 		if (error =
4030 		    znfsexport_fs((void *)
4031 		    (uintptr_t)zc->zc_share.z_exportdata))
4032 			return (error);
4033 		break;
4034 	case ZFS_SHARE_SMB:
4035 	case ZFS_UNSHARE_SMB:
4036 		if (error = zsmbexport_fs((void *)
4037 		    (uintptr_t)zc->zc_share.z_exportdata,
4038 		    zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
4039 		    B_TRUE: B_FALSE)) {
4040 			return (error);
4041 		}
4042 		break;
4043 	}
4044 
4045 	opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
4046 	    zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
4047 	    SHAREFS_ADD : SHAREFS_REMOVE;
4048 
4049 	/*
4050 	 * Add or remove share from sharetab
4051 	 */
4052 	error = zshare_fs(opcode,
4053 	    (void *)(uintptr_t)zc->zc_share.z_sharedata,
4054 	    zc->zc_share.z_sharemax);
4055 
4056 	return (error);
4057 
4058 }
4059 
4060 ace_t full_access[] = {
4061 	{(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4062 };
4063 
4064 /*
4065  * Remove all ACL files in shares dir
4066  */
4067 static int
4068 zfs_smb_acl_purge(znode_t *dzp)
4069 {
4070 	zap_cursor_t	zc;
4071 	zap_attribute_t	zap;
4072 	zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
4073 	int error;
4074 
4075 	for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
4076 	    (error = zap_cursor_retrieve(&zc, &zap)) == 0;
4077 	    zap_cursor_advance(&zc)) {
4078 		if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
4079 		    NULL, 0)) != 0)
4080 			break;
4081 	}
4082 	zap_cursor_fini(&zc);
4083 	return (error);
4084 }
4085 
4086 static int
4087 zfs_ioc_smb_acl(zfs_cmd_t *zc)
4088 {
4089 	vnode_t *vp;
4090 	znode_t *dzp;
4091 	vnode_t *resourcevp = NULL;
4092 	znode_t *sharedir;
4093 	zfsvfs_t *zfsvfs;
4094 	nvlist_t *nvlist;
4095 	char *src, *target;
4096 	vattr_t vattr;
4097 	vsecattr_t vsec;
4098 	int error = 0;
4099 
4100 	if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
4101 	    NO_FOLLOW, NULL, &vp)) != 0)
4102 		return (error);
4103 
4104 	/* Now make sure mntpnt and dataset are ZFS */
4105 
4106 	if (vp->v_vfsp->vfs_fstype != zfsfstype ||
4107 	    (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
4108 	    zc->zc_name) != 0)) {
4109 		VN_RELE(vp);
4110 		return (EINVAL);
4111 	}
4112 
4113 	dzp = VTOZ(vp);
4114 	zfsvfs = dzp->z_zfsvfs;
4115 	ZFS_ENTER(zfsvfs);
4116 
4117 	/*
4118 	 * Create share dir if its missing.
4119 	 */
4120 	mutex_enter(&zfsvfs->z_lock);
4121 	if (zfsvfs->z_shares_dir == 0) {
4122 		dmu_tx_t *tx;
4123 
4124 		tx = dmu_tx_create(zfsvfs->z_os);
4125 		dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
4126 		    ZFS_SHARES_DIR);
4127 		dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
4128 		error = dmu_tx_assign(tx, TXG_WAIT);
4129 		if (error) {
4130 			dmu_tx_abort(tx);
4131 		} else {
4132 			error = zfs_create_share_dir(zfsvfs, tx);
4133 			dmu_tx_commit(tx);
4134 		}
4135 		if (error) {
4136 			mutex_exit(&zfsvfs->z_lock);
4137 			VN_RELE(vp);
4138 			ZFS_EXIT(zfsvfs);
4139 			return (error);
4140 		}
4141 	}
4142 	mutex_exit(&zfsvfs->z_lock);
4143 
4144 	ASSERT(zfsvfs->z_shares_dir);
4145 	if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
4146 		VN_RELE(vp);
4147 		ZFS_EXIT(zfsvfs);
4148 		return (error);
4149 	}
4150 
4151 	switch (zc->zc_cookie) {
4152 	case ZFS_SMB_ACL_ADD:
4153 		vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
4154 		vattr.va_type = VREG;
4155 		vattr.va_mode = S_IFREG|0777;
4156 		vattr.va_uid = 0;
4157 		vattr.va_gid = 0;
4158 
4159 		vsec.vsa_mask = VSA_ACE;
4160 		vsec.vsa_aclentp = &full_access;
4161 		vsec.vsa_aclentsz = sizeof (full_access);
4162 		vsec.vsa_aclcnt = 1;
4163 
4164 		error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
4165 		    &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
4166 		if (resourcevp)
4167 			VN_RELE(resourcevp);
4168 		break;
4169 
4170 	case ZFS_SMB_ACL_REMOVE:
4171 		error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
4172 		    NULL, 0);
4173 		break;
4174 
4175 	case ZFS_SMB_ACL_RENAME:
4176 		if ((error = get_nvlist(zc->zc_nvlist_src,
4177 		    zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
4178 			VN_RELE(vp);
4179 			ZFS_EXIT(zfsvfs);
4180 			return (error);
4181 		}
4182 		if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
4183 		    nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
4184 		    &target)) {
4185 			VN_RELE(vp);
4186 			VN_RELE(ZTOV(sharedir));
4187 			ZFS_EXIT(zfsvfs);
4188 			nvlist_free(nvlist);
4189 			return (error);
4190 		}
4191 		error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
4192 		    kcred, NULL, 0);
4193 		nvlist_free(nvlist);
4194 		break;
4195 
4196 	case ZFS_SMB_ACL_PURGE:
4197 		error = zfs_smb_acl_purge(sharedir);
4198 		break;
4199 
4200 	default:
4201 		error = EINVAL;
4202 		break;
4203 	}
4204 
4205 	VN_RELE(vp);
4206 	VN_RELE(ZTOV(sharedir));
4207 
4208 	ZFS_EXIT(zfsvfs);
4209 
4210 	return (error);
4211 }
4212 
4213 /*
4214  * inputs:
4215  * zc_name	name of filesystem
4216  * zc_value	short name of snap
4217  * zc_string	user-supplied tag for this reference
4218  * zc_cookie	recursive flag
4219  * zc_temphold	set if hold is temporary
4220  *
4221  * outputs:		none
4222  */
4223 static int
4224 zfs_ioc_hold(zfs_cmd_t *zc)
4225 {
4226 	boolean_t recursive = zc->zc_cookie;
4227 
4228 	if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
4229 		return (EINVAL);
4230 
4231 	return (dsl_dataset_user_hold(zc->zc_name, zc->zc_value,
4232 	    zc->zc_string, recursive, zc->zc_temphold));
4233 }
4234 
4235 /*
4236  * inputs:
4237  * zc_name	name of dataset from which we're releasing a user reference
4238  * zc_value	short name of snap
4239  * zc_string	user-supplied tag for this reference
4240  * zc_cookie	recursive flag
4241  *
4242  * outputs:		none
4243  */
4244 static int
4245 zfs_ioc_release(zfs_cmd_t *zc)
4246 {
4247 	boolean_t recursive = zc->zc_cookie;
4248 
4249 	if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
4250 		return (EINVAL);
4251 
4252 	return (dsl_dataset_user_release(zc->zc_name, zc->zc_value,
4253 	    zc->zc_string, recursive));
4254 }
4255 
4256 /*
4257  * inputs:
4258  * zc_name		name of filesystem
4259  *
4260  * outputs:
4261  * zc_nvlist_src{_size}	nvlist of snapshot holds
4262  */
4263 static int
4264 zfs_ioc_get_holds(zfs_cmd_t *zc)
4265 {
4266 	nvlist_t *nvp;
4267 	int error;
4268 
4269 	if ((error = dsl_dataset_get_holds(zc->zc_name, &nvp)) == 0) {
4270 		error = put_nvlist(zc, nvp);
4271 		nvlist_free(nvp);
4272 	}
4273 
4274 	return (error);
4275 }
4276 
4277 /*
4278  * pool create, destroy, and export don't log the history as part of
4279  * zfsdev_ioctl, but rather zfs_ioc_pool_create, and zfs_ioc_pool_export
4280  * do the logging of those commands.
4281  */
4282 static zfs_ioc_vec_t zfs_ioc_vec[] = {
4283 	{ zfs_ioc_pool_create, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4284 	    B_FALSE },
4285 	{ zfs_ioc_pool_destroy,	zfs_secpolicy_config, POOL_NAME, B_FALSE,
4286 	    B_FALSE },
4287 	{ zfs_ioc_pool_import, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4288 	    B_FALSE },
4289 	{ zfs_ioc_pool_export, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4290 	    B_FALSE },
4291 	{ zfs_ioc_pool_configs,	zfs_secpolicy_none, NO_NAME, B_FALSE,
4292 	    B_FALSE },
4293 	{ zfs_ioc_pool_stats, zfs_secpolicy_read, POOL_NAME, B_FALSE,
4294 	    B_FALSE },
4295 	{ zfs_ioc_pool_tryimport, zfs_secpolicy_config, NO_NAME, B_FALSE,
4296 	    B_FALSE },
4297 	{ zfs_ioc_pool_scrub, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4298 	    B_TRUE },
4299 	{ zfs_ioc_pool_freeze, zfs_secpolicy_config, NO_NAME, B_FALSE,
4300 	    B_FALSE },
4301 	{ zfs_ioc_pool_upgrade,	zfs_secpolicy_config, POOL_NAME, B_TRUE,
4302 	    B_TRUE },
4303 	{ zfs_ioc_pool_get_history, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4304 	    B_FALSE },
4305 	{ zfs_ioc_vdev_add, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4306 	    B_TRUE },
4307 	{ zfs_ioc_vdev_remove, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4308 	    B_TRUE },
4309 	{ zfs_ioc_vdev_set_state, zfs_secpolicy_config,	POOL_NAME, B_TRUE,
4310 	    B_FALSE },
4311 	{ zfs_ioc_vdev_attach, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4312 	    B_TRUE },
4313 	{ zfs_ioc_vdev_detach, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4314 	    B_TRUE },
4315 	{ zfs_ioc_vdev_setpath,	zfs_secpolicy_config, POOL_NAME, B_FALSE,
4316 	    B_TRUE },
4317 	{ zfs_ioc_vdev_setfru,	zfs_secpolicy_config, POOL_NAME, B_FALSE,
4318 	    B_TRUE },
4319 	{ zfs_ioc_objset_stats,	zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4320 	    B_TRUE },
4321 	{ zfs_ioc_objset_zplprops, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4322 	    B_FALSE },
4323 	{ zfs_ioc_dataset_list_next, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4324 	    B_TRUE },
4325 	{ zfs_ioc_snapshot_list_next, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4326 	    B_TRUE },
4327 	{ zfs_ioc_set_prop, zfs_secpolicy_none, DATASET_NAME, B_TRUE, B_TRUE },
4328 	{ zfs_ioc_create, zfs_secpolicy_create, DATASET_NAME, B_TRUE, B_TRUE },
4329 	{ zfs_ioc_destroy, zfs_secpolicy_destroy, DATASET_NAME, B_TRUE,
4330 	    B_TRUE},
4331 	{ zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, B_TRUE,
4332 	    B_TRUE },
4333 	{ zfs_ioc_rename, zfs_secpolicy_rename,	DATASET_NAME, B_TRUE, B_TRUE },
4334 	{ zfs_ioc_recv, zfs_secpolicy_receive, DATASET_NAME, B_TRUE, B_TRUE },
4335 	{ zfs_ioc_send, zfs_secpolicy_send, DATASET_NAME, B_TRUE, B_FALSE },
4336 	{ zfs_ioc_inject_fault,	zfs_secpolicy_inject, NO_NAME, B_FALSE,
4337 	    B_FALSE },
4338 	{ zfs_ioc_clear_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE,
4339 	    B_FALSE },
4340 	{ zfs_ioc_inject_list_next, zfs_secpolicy_inject, NO_NAME, B_FALSE,
4341 	    B_FALSE },
4342 	{ zfs_ioc_error_log, zfs_secpolicy_inject, POOL_NAME, B_FALSE,
4343 	    B_FALSE },
4344 	{ zfs_ioc_clear, zfs_secpolicy_config, POOL_NAME, B_TRUE, B_FALSE },
4345 	{ zfs_ioc_promote, zfs_secpolicy_promote, DATASET_NAME, B_TRUE,
4346 	    B_TRUE },
4347 	{ zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, DATASET_NAME,
4348 	    B_TRUE, B_TRUE },
4349 	{ zfs_ioc_snapshot, zfs_secpolicy_snapshot, DATASET_NAME, B_TRUE,
4350 	    B_TRUE },
4351 	{ zfs_ioc_dsobj_to_dsname, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4352 	    B_FALSE },
4353 	{ zfs_ioc_obj_to_path, zfs_secpolicy_config, DATASET_NAME, B_FALSE,
4354 	    B_TRUE },
4355 	{ zfs_ioc_pool_set_props, zfs_secpolicy_config,	POOL_NAME, B_TRUE,
4356 	    B_TRUE },
4357 	{ zfs_ioc_pool_get_props, zfs_secpolicy_read, POOL_NAME, B_FALSE,
4358 	    B_FALSE },
4359 	{ zfs_ioc_set_fsacl, zfs_secpolicy_fsacl, DATASET_NAME, B_TRUE,
4360 	    B_TRUE },
4361 	{ zfs_ioc_get_fsacl, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4362 	    B_FALSE },
4363 	{ zfs_ioc_iscsi_perm_check, zfs_secpolicy_iscsi, DATASET_NAME, B_FALSE,
4364 	    B_FALSE },
4365 	{ zfs_ioc_share, zfs_secpolicy_share, DATASET_NAME, B_FALSE, B_FALSE },
4366 	{ zfs_ioc_inherit_prop, zfs_secpolicy_inherit, DATASET_NAME, B_TRUE,
4367 	    B_TRUE },
4368 	{ zfs_ioc_smb_acl, zfs_secpolicy_smb_acl, DATASET_NAME, B_FALSE,
4369 	    B_FALSE },
4370 	{ zfs_ioc_userspace_one, zfs_secpolicy_userspace_one,
4371 	    DATASET_NAME, B_FALSE, B_FALSE },
4372 	{ zfs_ioc_userspace_many, zfs_secpolicy_userspace_many,
4373 	    DATASET_NAME, B_FALSE, B_FALSE },
4374 	{ zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
4375 	    DATASET_NAME, B_FALSE, B_TRUE },
4376 	{ zfs_ioc_hold, zfs_secpolicy_hold, DATASET_NAME, B_TRUE, B_TRUE },
4377 	{ zfs_ioc_release, zfs_secpolicy_release, DATASET_NAME, B_TRUE,
4378 	    B_TRUE },
4379 	{ zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4380 	    B_TRUE },
4381 	{ zfs_ioc_objset_recvd_props, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4382 	    B_FALSE },
4383 	{ zfs_ioc_vdev_split, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4384 	    B_TRUE }
4385 };
4386 
4387 int
4388 pool_status_check(const char *name, zfs_ioc_namecheck_t type)
4389 {
4390 	spa_t *spa;
4391 	int error;
4392 
4393 	ASSERT(type == POOL_NAME || type == DATASET_NAME);
4394 
4395 	error = spa_open(name, &spa, FTAG);
4396 	if (error == 0) {
4397 		if (spa_suspended(spa))
4398 			error = EAGAIN;
4399 		spa_close(spa, FTAG);
4400 	}
4401 	return (error);
4402 }
4403 
4404 static int
4405 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
4406 {
4407 	zfs_cmd_t *zc;
4408 	uint_t vec;
4409 	int error, rc;
4410 
4411 	if (getminor(dev) != 0)
4412 		return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
4413 
4414 	vec = cmd - ZFS_IOC;
4415 	ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
4416 
4417 	if (vec >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
4418 		return (EINVAL);
4419 
4420 	zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
4421 
4422 	error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
4423 
4424 	if ((error == 0) && !(flag & FKIOCTL))
4425 		error = zfs_ioc_vec[vec].zvec_secpolicy(zc, cr);
4426 
4427 	/*
4428 	 * Ensure that all pool/dataset names are valid before we pass down to
4429 	 * the lower layers.
4430 	 */
4431 	if (error == 0) {
4432 		zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
4433 		zc->zc_iflags = flag & FKIOCTL;
4434 		switch (zfs_ioc_vec[vec].zvec_namecheck) {
4435 		case POOL_NAME:
4436 			if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
4437 				error = EINVAL;
4438 			if (zfs_ioc_vec[vec].zvec_pool_check)
4439 				error = pool_status_check(zc->zc_name,
4440 				    zfs_ioc_vec[vec].zvec_namecheck);
4441 			break;
4442 
4443 		case DATASET_NAME:
4444 			if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
4445 				error = EINVAL;
4446 			if (zfs_ioc_vec[vec].zvec_pool_check)
4447 				error = pool_status_check(zc->zc_name,
4448 				    zfs_ioc_vec[vec].zvec_namecheck);
4449 			break;
4450 
4451 		case NO_NAME:
4452 			break;
4453 		}
4454 	}
4455 
4456 	if (error == 0)
4457 		error = zfs_ioc_vec[vec].zvec_func(zc);
4458 
4459 	rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
4460 	if (error == 0) {
4461 		error = rc;
4462 		if (zfs_ioc_vec[vec].zvec_his_log)
4463 			zfs_log_history(zc);
4464 	}
4465 
4466 	kmem_free(zc, sizeof (zfs_cmd_t));
4467 	return (error);
4468 }
4469 
4470 static int
4471 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
4472 {
4473 	if (cmd != DDI_ATTACH)
4474 		return (DDI_FAILURE);
4475 
4476 	if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
4477 	    DDI_PSEUDO, 0) == DDI_FAILURE)
4478 		return (DDI_FAILURE);
4479 
4480 	zfs_dip = dip;
4481 
4482 	ddi_report_dev(dip);
4483 
4484 	return (DDI_SUCCESS);
4485 }
4486 
4487 static int
4488 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
4489 {
4490 	if (spa_busy() || zfs_busy() || zvol_busy())
4491 		return (DDI_FAILURE);
4492 
4493 	if (cmd != DDI_DETACH)
4494 		return (DDI_FAILURE);
4495 
4496 	zfs_dip = NULL;
4497 
4498 	ddi_prop_remove_all(dip);
4499 	ddi_remove_minor_node(dip, NULL);
4500 
4501 	return (DDI_SUCCESS);
4502 }
4503 
4504 /*ARGSUSED*/
4505 static int
4506 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
4507 {
4508 	switch (infocmd) {
4509 	case DDI_INFO_DEVT2DEVINFO:
4510 		*result = zfs_dip;
4511 		return (DDI_SUCCESS);
4512 
4513 	case DDI_INFO_DEVT2INSTANCE:
4514 		*result = (void *)0;
4515 		return (DDI_SUCCESS);
4516 	}
4517 
4518 	return (DDI_FAILURE);
4519 }
4520 
4521 /*
4522  * OK, so this is a little weird.
4523  *
4524  * /dev/zfs is the control node, i.e. minor 0.
4525  * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
4526  *
4527  * /dev/zfs has basically nothing to do except serve up ioctls,
4528  * so most of the standard driver entry points are in zvol.c.
4529  */
4530 static struct cb_ops zfs_cb_ops = {
4531 	zvol_open,	/* open */
4532 	zvol_close,	/* close */
4533 	zvol_strategy,	/* strategy */
4534 	nodev,		/* print */
4535 	zvol_dump,	/* dump */
4536 	zvol_read,	/* read */
4537 	zvol_write,	/* write */
4538 	zfsdev_ioctl,	/* ioctl */
4539 	nodev,		/* devmap */
4540 	nodev,		/* mmap */
4541 	nodev,		/* segmap */
4542 	nochpoll,	/* poll */
4543 	ddi_prop_op,	/* prop_op */
4544 	NULL,		/* streamtab */
4545 	D_NEW | D_MP | D_64BIT,		/* Driver compatibility flag */
4546 	CB_REV,		/* version */
4547 	nodev,		/* async read */
4548 	nodev,		/* async write */
4549 };
4550 
4551 static struct dev_ops zfs_dev_ops = {
4552 	DEVO_REV,	/* version */
4553 	0,		/* refcnt */
4554 	zfs_info,	/* info */
4555 	nulldev,	/* identify */
4556 	nulldev,	/* probe */
4557 	zfs_attach,	/* attach */
4558 	zfs_detach,	/* detach */
4559 	nodev,		/* reset */
4560 	&zfs_cb_ops,	/* driver operations */
4561 	NULL,		/* no bus operations */
4562 	NULL,		/* power */
4563 	ddi_quiesce_not_needed,	/* quiesce */
4564 };
4565 
4566 static struct modldrv zfs_modldrv = {
4567 	&mod_driverops,
4568 	"ZFS storage pool",
4569 	&zfs_dev_ops
4570 };
4571 
4572 static struct modlinkage modlinkage = {
4573 	MODREV_1,
4574 	(void *)&zfs_modlfs,
4575 	(void *)&zfs_modldrv,
4576 	NULL
4577 };
4578 
4579 
4580 uint_t zfs_fsyncer_key;
4581 extern uint_t rrw_tsd_key;
4582 
4583 int
4584 _init(void)
4585 {
4586 	int error;
4587 
4588 	spa_init(FREAD | FWRITE);
4589 	zfs_init();
4590 	zvol_init();
4591 
4592 	if ((error = mod_install(&modlinkage)) != 0) {
4593 		zvol_fini();
4594 		zfs_fini();
4595 		spa_fini();
4596 		return (error);
4597 	}
4598 
4599 	tsd_create(&zfs_fsyncer_key, NULL);
4600 	tsd_create(&rrw_tsd_key, NULL);
4601 
4602 	error = ldi_ident_from_mod(&modlinkage, &zfs_li);
4603 	ASSERT(error == 0);
4604 	mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
4605 
4606 	return (0);
4607 }
4608 
4609 int
4610 _fini(void)
4611 {
4612 	int error;
4613 
4614 	if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
4615 		return (EBUSY);
4616 
4617 	if ((error = mod_remove(&modlinkage)) != 0)
4618 		return (error);
4619 
4620 	zvol_fini();
4621 	zfs_fini();
4622 	spa_fini();
4623 	if (zfs_nfsshare_inited)
4624 		(void) ddi_modclose(nfs_mod);
4625 	if (zfs_smbshare_inited)
4626 		(void) ddi_modclose(smbsrv_mod);
4627 	if (zfs_nfsshare_inited || zfs_smbshare_inited)
4628 		(void) ddi_modclose(sharefs_mod);
4629 
4630 	tsd_destroy(&zfs_fsyncer_key);
4631 	ldi_ident_release(zfs_li);
4632 	zfs_li = NULL;
4633 	mutex_destroy(&zfs_share_lock);
4634 
4635 	return (error);
4636 }
4637 
4638 int
4639 _info(struct modinfo *modinfop)
4640 {
4641 	return (mod_info(&modlinkage, modinfop));
4642 }
4643