xref: /titanic_52/usr/src/uts/common/fs/zfs/zfs_ioctl.c (revision ab003da878e3fe36b164e1856f9e15a78384c9eb)
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 int
629 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
630 {
631 	char	parentname[MAXNAMELEN];
632 	int	error;
633 
634 	if ((error = zfs_secpolicy_write_perms(from,
635 	    ZFS_DELEG_PERM_RENAME, cr)) != 0)
636 		return (error);
637 
638 	if ((error = zfs_secpolicy_write_perms(from,
639 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
640 		return (error);
641 
642 	if ((error = zfs_get_parent(to, parentname,
643 	    sizeof (parentname))) != 0)
644 		return (error);
645 
646 	if ((error = zfs_secpolicy_write_perms(parentname,
647 	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
648 		return (error);
649 
650 	if ((error = zfs_secpolicy_write_perms(parentname,
651 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
652 		return (error);
653 
654 	return (error);
655 }
656 
657 static int
658 zfs_secpolicy_rename(zfs_cmd_t *zc, cred_t *cr)
659 {
660 	return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
661 }
662 
663 static int
664 zfs_secpolicy_promote(zfs_cmd_t *zc, cred_t *cr)
665 {
666 	char	parentname[MAXNAMELEN];
667 	objset_t *clone;
668 	int error;
669 
670 	error = zfs_secpolicy_write_perms(zc->zc_name,
671 	    ZFS_DELEG_PERM_PROMOTE, cr);
672 	if (error)
673 		return (error);
674 
675 	error = dmu_objset_hold(zc->zc_name, FTAG, &clone);
676 
677 	if (error == 0) {
678 		dsl_dataset_t *pclone = NULL;
679 		dsl_dir_t *dd;
680 		dd = clone->os_dsl_dataset->ds_dir;
681 
682 		rw_enter(&dd->dd_pool->dp_config_rwlock, RW_READER);
683 		error = dsl_dataset_hold_obj(dd->dd_pool,
684 		    dd->dd_phys->dd_origin_obj, FTAG, &pclone);
685 		rw_exit(&dd->dd_pool->dp_config_rwlock);
686 		if (error) {
687 			dmu_objset_rele(clone, FTAG);
688 			return (error);
689 		}
690 
691 		error = zfs_secpolicy_write_perms(zc->zc_name,
692 		    ZFS_DELEG_PERM_MOUNT, cr);
693 
694 		dsl_dataset_name(pclone, parentname);
695 		dmu_objset_rele(clone, FTAG);
696 		dsl_dataset_rele(pclone, FTAG);
697 		if (error == 0)
698 			error = zfs_secpolicy_write_perms(parentname,
699 			    ZFS_DELEG_PERM_PROMOTE, cr);
700 	}
701 	return (error);
702 }
703 
704 static int
705 zfs_secpolicy_receive(zfs_cmd_t *zc, cred_t *cr)
706 {
707 	int error;
708 
709 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
710 	    ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
711 		return (error);
712 
713 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
714 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
715 		return (error);
716 
717 	return (zfs_secpolicy_write_perms(zc->zc_name,
718 	    ZFS_DELEG_PERM_CREATE, cr));
719 }
720 
721 int
722 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
723 {
724 	return (zfs_secpolicy_write_perms(name,
725 	    ZFS_DELEG_PERM_SNAPSHOT, cr));
726 }
727 
728 static int
729 zfs_secpolicy_snapshot(zfs_cmd_t *zc, cred_t *cr)
730 {
731 
732 	return (zfs_secpolicy_snapshot_perms(zc->zc_name, cr));
733 }
734 
735 static int
736 zfs_secpolicy_create(zfs_cmd_t *zc, cred_t *cr)
737 {
738 	char	parentname[MAXNAMELEN];
739 	int	error;
740 
741 	if ((error = zfs_get_parent(zc->zc_name, parentname,
742 	    sizeof (parentname))) != 0)
743 		return (error);
744 
745 	if (zc->zc_value[0] != '\0') {
746 		if ((error = zfs_secpolicy_write_perms(zc->zc_value,
747 		    ZFS_DELEG_PERM_CLONE, cr)) != 0)
748 			return (error);
749 	}
750 
751 	if ((error = zfs_secpolicy_write_perms(parentname,
752 	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
753 		return (error);
754 
755 	error = zfs_secpolicy_write_perms(parentname,
756 	    ZFS_DELEG_PERM_MOUNT, cr);
757 
758 	return (error);
759 }
760 
761 static int
762 zfs_secpolicy_umount(zfs_cmd_t *zc, cred_t *cr)
763 {
764 	int error;
765 
766 	error = secpolicy_fs_unmount(cr, NULL);
767 	if (error) {
768 		error = dsl_deleg_access(zc->zc_name, ZFS_DELEG_PERM_MOUNT, cr);
769 	}
770 	return (error);
771 }
772 
773 /*
774  * Policy for pool operations - create/destroy pools, add vdevs, etc.  Requires
775  * SYS_CONFIG privilege, which is not available in a local zone.
776  */
777 /* ARGSUSED */
778 static int
779 zfs_secpolicy_config(zfs_cmd_t *zc, cred_t *cr)
780 {
781 	if (secpolicy_sys_config(cr, B_FALSE) != 0)
782 		return (EPERM);
783 
784 	return (0);
785 }
786 
787 /*
788  * Policy for fault injection.  Requires all privileges.
789  */
790 /* ARGSUSED */
791 static int
792 zfs_secpolicy_inject(zfs_cmd_t *zc, cred_t *cr)
793 {
794 	return (secpolicy_zinject(cr));
795 }
796 
797 static int
798 zfs_secpolicy_inherit(zfs_cmd_t *zc, cred_t *cr)
799 {
800 	zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
801 
802 	if (prop == ZPROP_INVAL) {
803 		if (!zfs_prop_user(zc->zc_value))
804 			return (EINVAL);
805 		return (zfs_secpolicy_write_perms(zc->zc_name,
806 		    ZFS_DELEG_PERM_USERPROP, cr));
807 	} else {
808 		return (zfs_secpolicy_setprop(zc->zc_name, prop,
809 		    NULL, cr));
810 	}
811 }
812 
813 static int
814 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, cred_t *cr)
815 {
816 	int err = zfs_secpolicy_read(zc, cr);
817 	if (err)
818 		return (err);
819 
820 	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
821 		return (EINVAL);
822 
823 	if (zc->zc_value[0] == 0) {
824 		/*
825 		 * They are asking about a posix uid/gid.  If it's
826 		 * themself, allow it.
827 		 */
828 		if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
829 		    zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
830 			if (zc->zc_guid == crgetuid(cr))
831 				return (0);
832 		} else {
833 			if (groupmember(zc->zc_guid, cr))
834 				return (0);
835 		}
836 	}
837 
838 	return (zfs_secpolicy_write_perms(zc->zc_name,
839 	    userquota_perms[zc->zc_objset_type], cr));
840 }
841 
842 static int
843 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, cred_t *cr)
844 {
845 	int err = zfs_secpolicy_read(zc, cr);
846 	if (err)
847 		return (err);
848 
849 	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
850 		return (EINVAL);
851 
852 	return (zfs_secpolicy_write_perms(zc->zc_name,
853 	    userquota_perms[zc->zc_objset_type], cr));
854 }
855 
856 static int
857 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, cred_t *cr)
858 {
859 	return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
860 	    NULL, cr));
861 }
862 
863 static int
864 zfs_secpolicy_hold(zfs_cmd_t *zc, cred_t *cr)
865 {
866 	return (zfs_secpolicy_write_perms(zc->zc_name,
867 	    ZFS_DELEG_PERM_HOLD, cr));
868 }
869 
870 static int
871 zfs_secpolicy_release(zfs_cmd_t *zc, cred_t *cr)
872 {
873 	return (zfs_secpolicy_write_perms(zc->zc_name,
874 	    ZFS_DELEG_PERM_RELEASE, cr));
875 }
876 
877 /*
878  * Returns the nvlist as specified by the user in the zfs_cmd_t.
879  */
880 static int
881 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
882 {
883 	char *packed;
884 	int error;
885 	nvlist_t *list = NULL;
886 
887 	/*
888 	 * Read in and unpack the user-supplied nvlist.
889 	 */
890 	if (size == 0)
891 		return (EINVAL);
892 
893 	packed = kmem_alloc(size, KM_SLEEP);
894 
895 	if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
896 	    iflag)) != 0) {
897 		kmem_free(packed, size);
898 		return (error);
899 	}
900 
901 	if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
902 		kmem_free(packed, size);
903 		return (error);
904 	}
905 
906 	kmem_free(packed, size);
907 
908 	*nvp = list;
909 	return (0);
910 }
911 
912 static int
913 fit_error_list(zfs_cmd_t *zc, nvlist_t **errors)
914 {
915 	size_t size;
916 
917 	VERIFY(nvlist_size(*errors, &size, NV_ENCODE_NATIVE) == 0);
918 
919 	if (size > zc->zc_nvlist_dst_size) {
920 		nvpair_t *more_errors;
921 		int n = 0;
922 
923 		if (zc->zc_nvlist_dst_size < 1024)
924 			return (ENOMEM);
925 
926 		VERIFY(nvlist_add_int32(*errors, ZPROP_N_MORE_ERRORS, 0) == 0);
927 		more_errors = nvlist_prev_nvpair(*errors, NULL);
928 
929 		do {
930 			nvpair_t *pair = nvlist_prev_nvpair(*errors,
931 			    more_errors);
932 			VERIFY(nvlist_remove_nvpair(*errors, pair) == 0);
933 			n++;
934 			VERIFY(nvlist_size(*errors, &size,
935 			    NV_ENCODE_NATIVE) == 0);
936 		} while (size > zc->zc_nvlist_dst_size);
937 
938 		VERIFY(nvlist_remove_nvpair(*errors, more_errors) == 0);
939 		VERIFY(nvlist_add_int32(*errors, ZPROP_N_MORE_ERRORS, n) == 0);
940 		ASSERT(nvlist_size(*errors, &size, NV_ENCODE_NATIVE) == 0);
941 		ASSERT(size <= zc->zc_nvlist_dst_size);
942 	}
943 
944 	return (0);
945 }
946 
947 static int
948 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
949 {
950 	char *packed = NULL;
951 	int error = 0;
952 	size_t size;
953 
954 	VERIFY(nvlist_size(nvl, &size, NV_ENCODE_NATIVE) == 0);
955 
956 	if (size > zc->zc_nvlist_dst_size) {
957 		error = ENOMEM;
958 	} else {
959 		packed = kmem_alloc(size, KM_SLEEP);
960 		VERIFY(nvlist_pack(nvl, &packed, &size, NV_ENCODE_NATIVE,
961 		    KM_SLEEP) == 0);
962 		if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
963 		    size, zc->zc_iflags) != 0)
964 			error = EFAULT;
965 		kmem_free(packed, size);
966 	}
967 
968 	zc->zc_nvlist_dst_size = size;
969 	return (error);
970 }
971 
972 static int
973 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
974 {
975 	objset_t *os;
976 	int error;
977 
978 	error = dmu_objset_hold(dsname, FTAG, &os);
979 	if (error)
980 		return (error);
981 	if (dmu_objset_type(os) != DMU_OST_ZFS) {
982 		dmu_objset_rele(os, FTAG);
983 		return (EINVAL);
984 	}
985 
986 	mutex_enter(&os->os_user_ptr_lock);
987 	*zfvp = dmu_objset_get_user(os);
988 	if (*zfvp) {
989 		VFS_HOLD((*zfvp)->z_vfs);
990 	} else {
991 		error = ESRCH;
992 	}
993 	mutex_exit(&os->os_user_ptr_lock);
994 	dmu_objset_rele(os, FTAG);
995 	return (error);
996 }
997 
998 /*
999  * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1000  * case its z_vfs will be NULL, and it will be opened as the owner.
1001  */
1002 static int
1003 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp)
1004 {
1005 	int error = 0;
1006 
1007 	if (getzfsvfs(name, zfvp) != 0)
1008 		error = zfsvfs_create(name, zfvp);
1009 	if (error == 0) {
1010 		rrw_enter(&(*zfvp)->z_teardown_lock, RW_READER, tag);
1011 		if ((*zfvp)->z_unmounted) {
1012 			/*
1013 			 * XXX we could probably try again, since the unmounting
1014 			 * thread should be just about to disassociate the
1015 			 * objset from the zfsvfs.
1016 			 */
1017 			rrw_exit(&(*zfvp)->z_teardown_lock, tag);
1018 			return (EBUSY);
1019 		}
1020 	}
1021 	return (error);
1022 }
1023 
1024 static void
1025 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1026 {
1027 	rrw_exit(&zfsvfs->z_teardown_lock, tag);
1028 
1029 	if (zfsvfs->z_vfs) {
1030 		VFS_RELE(zfsvfs->z_vfs);
1031 	} else {
1032 		dmu_objset_disown(zfsvfs->z_os, zfsvfs);
1033 		zfsvfs_free(zfsvfs);
1034 	}
1035 }
1036 
1037 static int
1038 zfs_ioc_pool_create(zfs_cmd_t *zc)
1039 {
1040 	int error;
1041 	nvlist_t *config, *props = NULL;
1042 	nvlist_t *rootprops = NULL;
1043 	nvlist_t *zplprops = NULL;
1044 	char *buf;
1045 
1046 	if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1047 	    zc->zc_iflags, &config))
1048 		return (error);
1049 
1050 	if (zc->zc_nvlist_src_size != 0 && (error =
1051 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1052 	    zc->zc_iflags, &props))) {
1053 		nvlist_free(config);
1054 		return (error);
1055 	}
1056 
1057 	if (props) {
1058 		nvlist_t *nvl = NULL;
1059 		uint64_t version = SPA_VERSION;
1060 
1061 		(void) nvlist_lookup_uint64(props,
1062 		    zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1063 		if (version < SPA_VERSION_INITIAL || version > SPA_VERSION) {
1064 			error = EINVAL;
1065 			goto pool_props_bad;
1066 		}
1067 		(void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1068 		if (nvl) {
1069 			error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1070 			if (error != 0) {
1071 				nvlist_free(config);
1072 				nvlist_free(props);
1073 				return (error);
1074 			}
1075 			(void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1076 		}
1077 		VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1078 		error = zfs_fill_zplprops_root(version, rootprops,
1079 		    zplprops, NULL);
1080 		if (error)
1081 			goto pool_props_bad;
1082 	}
1083 
1084 	buf = history_str_get(zc);
1085 
1086 	error = spa_create(zc->zc_name, config, props, buf, zplprops);
1087 
1088 	/*
1089 	 * Set the remaining root properties
1090 	 */
1091 	if (!error && (error = zfs_set_prop_nvlist(zc->zc_name,
1092 	    ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1093 		(void) spa_destroy(zc->zc_name);
1094 
1095 	if (buf != NULL)
1096 		history_str_free(buf);
1097 
1098 pool_props_bad:
1099 	nvlist_free(rootprops);
1100 	nvlist_free(zplprops);
1101 	nvlist_free(config);
1102 	nvlist_free(props);
1103 
1104 	return (error);
1105 }
1106 
1107 static int
1108 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1109 {
1110 	int error;
1111 	zfs_log_history(zc);
1112 	error = spa_destroy(zc->zc_name);
1113 	if (error == 0)
1114 		zvol_remove_minors(zc->zc_name);
1115 	return (error);
1116 }
1117 
1118 static int
1119 zfs_ioc_pool_import(zfs_cmd_t *zc)
1120 {
1121 	nvlist_t *config, *props = NULL;
1122 	uint64_t guid;
1123 	int error;
1124 
1125 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1126 	    zc->zc_iflags, &config)) != 0)
1127 		return (error);
1128 
1129 	if (zc->zc_nvlist_src_size != 0 && (error =
1130 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1131 	    zc->zc_iflags, &props))) {
1132 		nvlist_free(config);
1133 		return (error);
1134 	}
1135 
1136 	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1137 	    guid != zc->zc_guid)
1138 		error = EINVAL;
1139 	else if (zc->zc_cookie)
1140 		error = spa_import_verbatim(zc->zc_name, config, props);
1141 	else
1142 		error = spa_import(zc->zc_name, config, props);
1143 
1144 	if (zc->zc_nvlist_dst != 0)
1145 		(void) put_nvlist(zc, config);
1146 
1147 	nvlist_free(config);
1148 
1149 	if (props)
1150 		nvlist_free(props);
1151 
1152 	return (error);
1153 }
1154 
1155 static int
1156 zfs_ioc_pool_export(zfs_cmd_t *zc)
1157 {
1158 	int error;
1159 	boolean_t force = (boolean_t)zc->zc_cookie;
1160 	boolean_t hardforce = (boolean_t)zc->zc_guid;
1161 
1162 	zfs_log_history(zc);
1163 	error = spa_export(zc->zc_name, NULL, force, hardforce);
1164 	if (error == 0)
1165 		zvol_remove_minors(zc->zc_name);
1166 	return (error);
1167 }
1168 
1169 static int
1170 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1171 {
1172 	nvlist_t *configs;
1173 	int error;
1174 
1175 	if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1176 		return (EEXIST);
1177 
1178 	error = put_nvlist(zc, configs);
1179 
1180 	nvlist_free(configs);
1181 
1182 	return (error);
1183 }
1184 
1185 static int
1186 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1187 {
1188 	nvlist_t *config;
1189 	int error;
1190 	int ret = 0;
1191 
1192 	error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1193 	    sizeof (zc->zc_value));
1194 
1195 	if (config != NULL) {
1196 		ret = put_nvlist(zc, config);
1197 		nvlist_free(config);
1198 
1199 		/*
1200 		 * The config may be present even if 'error' is non-zero.
1201 		 * In this case we return success, and preserve the real errno
1202 		 * in 'zc_cookie'.
1203 		 */
1204 		zc->zc_cookie = error;
1205 	} else {
1206 		ret = error;
1207 	}
1208 
1209 	return (ret);
1210 }
1211 
1212 /*
1213  * Try to import the given pool, returning pool stats as appropriate so that
1214  * user land knows which devices are available and overall pool health.
1215  */
1216 static int
1217 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1218 {
1219 	nvlist_t *tryconfig, *config;
1220 	int error;
1221 
1222 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1223 	    zc->zc_iflags, &tryconfig)) != 0)
1224 		return (error);
1225 
1226 	config = spa_tryimport(tryconfig);
1227 
1228 	nvlist_free(tryconfig);
1229 
1230 	if (config == NULL)
1231 		return (EINVAL);
1232 
1233 	error = put_nvlist(zc, config);
1234 	nvlist_free(config);
1235 
1236 	return (error);
1237 }
1238 
1239 static int
1240 zfs_ioc_pool_scrub(zfs_cmd_t *zc)
1241 {
1242 	spa_t *spa;
1243 	int error;
1244 
1245 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1246 		return (error);
1247 
1248 	error = spa_scrub(spa, zc->zc_cookie);
1249 
1250 	spa_close(spa, FTAG);
1251 
1252 	return (error);
1253 }
1254 
1255 static int
1256 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1257 {
1258 	spa_t *spa;
1259 	int error;
1260 
1261 	error = spa_open(zc->zc_name, &spa, FTAG);
1262 	if (error == 0) {
1263 		spa_freeze(spa);
1264 		spa_close(spa, FTAG);
1265 	}
1266 	return (error);
1267 }
1268 
1269 static int
1270 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1271 {
1272 	spa_t *spa;
1273 	int error;
1274 
1275 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1276 		return (error);
1277 
1278 	if (zc->zc_cookie < spa_version(spa) || zc->zc_cookie > SPA_VERSION) {
1279 		spa_close(spa, FTAG);
1280 		return (EINVAL);
1281 	}
1282 
1283 	spa_upgrade(spa, zc->zc_cookie);
1284 	spa_close(spa, FTAG);
1285 
1286 	return (error);
1287 }
1288 
1289 static int
1290 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1291 {
1292 	spa_t *spa;
1293 	char *hist_buf;
1294 	uint64_t size;
1295 	int error;
1296 
1297 	if ((size = zc->zc_history_len) == 0)
1298 		return (EINVAL);
1299 
1300 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1301 		return (error);
1302 
1303 	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1304 		spa_close(spa, FTAG);
1305 		return (ENOTSUP);
1306 	}
1307 
1308 	hist_buf = kmem_alloc(size, KM_SLEEP);
1309 	if ((error = spa_history_get(spa, &zc->zc_history_offset,
1310 	    &zc->zc_history_len, hist_buf)) == 0) {
1311 		error = ddi_copyout(hist_buf,
1312 		    (void *)(uintptr_t)zc->zc_history,
1313 		    zc->zc_history_len, zc->zc_iflags);
1314 	}
1315 
1316 	spa_close(spa, FTAG);
1317 	kmem_free(hist_buf, size);
1318 	return (error);
1319 }
1320 
1321 static int
1322 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1323 {
1324 	int error;
1325 
1326 	if (error = dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value))
1327 		return (error);
1328 
1329 	return (0);
1330 }
1331 
1332 /*
1333  * inputs:
1334  * zc_name		name of filesystem
1335  * zc_obj		object to find
1336  *
1337  * outputs:
1338  * zc_value		name of object
1339  */
1340 static int
1341 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1342 {
1343 	objset_t *os;
1344 	int error;
1345 
1346 	/* XXX reading from objset not owned */
1347 	if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os)) != 0)
1348 		return (error);
1349 	if (dmu_objset_type(os) != DMU_OST_ZFS) {
1350 		dmu_objset_rele(os, FTAG);
1351 		return (EINVAL);
1352 	}
1353 	error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1354 	    sizeof (zc->zc_value));
1355 	dmu_objset_rele(os, FTAG);
1356 
1357 	return (error);
1358 }
1359 
1360 static int
1361 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1362 {
1363 	spa_t *spa;
1364 	int error;
1365 	nvlist_t *config, **l2cache, **spares;
1366 	uint_t nl2cache = 0, nspares = 0;
1367 
1368 	error = spa_open(zc->zc_name, &spa, FTAG);
1369 	if (error != 0)
1370 		return (error);
1371 
1372 	error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1373 	    zc->zc_iflags, &config);
1374 	(void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1375 	    &l2cache, &nl2cache);
1376 
1377 	(void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1378 	    &spares, &nspares);
1379 
1380 	/*
1381 	 * A root pool with concatenated devices is not supported.
1382 	 * Thus, can not add a device to a root pool.
1383 	 *
1384 	 * Intent log device can not be added to a rootpool because
1385 	 * during mountroot, zil is replayed, a seperated log device
1386 	 * can not be accessed during the mountroot time.
1387 	 *
1388 	 * l2cache and spare devices are ok to be added to a rootpool.
1389 	 */
1390 	if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1391 		nvlist_free(config);
1392 		spa_close(spa, FTAG);
1393 		return (EDOM);
1394 	}
1395 
1396 	if (error == 0) {
1397 		error = spa_vdev_add(spa, config);
1398 		nvlist_free(config);
1399 	}
1400 	spa_close(spa, FTAG);
1401 	return (error);
1402 }
1403 
1404 static int
1405 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1406 {
1407 	spa_t *spa;
1408 	int error;
1409 
1410 	error = spa_open(zc->zc_name, &spa, FTAG);
1411 	if (error != 0)
1412 		return (error);
1413 	error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1414 	spa_close(spa, FTAG);
1415 	return (error);
1416 }
1417 
1418 static int
1419 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1420 {
1421 	spa_t *spa;
1422 	int error;
1423 	vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1424 
1425 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1426 		return (error);
1427 	switch (zc->zc_cookie) {
1428 	case VDEV_STATE_ONLINE:
1429 		error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1430 		break;
1431 
1432 	case VDEV_STATE_OFFLINE:
1433 		error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1434 		break;
1435 
1436 	case VDEV_STATE_FAULTED:
1437 		if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1438 		    zc->zc_obj != VDEV_AUX_EXTERNAL)
1439 			zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1440 
1441 		error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
1442 		break;
1443 
1444 	case VDEV_STATE_DEGRADED:
1445 		if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
1446 		    zc->zc_obj != VDEV_AUX_EXTERNAL)
1447 			zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
1448 
1449 		error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
1450 		break;
1451 
1452 	default:
1453 		error = EINVAL;
1454 	}
1455 	zc->zc_cookie = newstate;
1456 	spa_close(spa, FTAG);
1457 	return (error);
1458 }
1459 
1460 static int
1461 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1462 {
1463 	spa_t *spa;
1464 	int replacing = zc->zc_cookie;
1465 	nvlist_t *config;
1466 	int error;
1467 
1468 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1469 		return (error);
1470 
1471 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1472 	    zc->zc_iflags, &config)) == 0) {
1473 		error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1474 		nvlist_free(config);
1475 	}
1476 
1477 	spa_close(spa, FTAG);
1478 	return (error);
1479 }
1480 
1481 static int
1482 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1483 {
1484 	spa_t *spa;
1485 	int error;
1486 
1487 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1488 		return (error);
1489 
1490 	error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1491 
1492 	spa_close(spa, FTAG);
1493 	return (error);
1494 }
1495 
1496 static int
1497 zfs_ioc_vdev_split(zfs_cmd_t *zc)
1498 {
1499 	spa_t *spa;
1500 	nvlist_t *config, *props = NULL;
1501 	int error;
1502 	boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
1503 
1504 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1505 		return (error);
1506 
1507 	if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1508 	    zc->zc_iflags, &config)) {
1509 		spa_close(spa, FTAG);
1510 		return (error);
1511 	}
1512 
1513 	if (zc->zc_nvlist_src_size != 0 && (error =
1514 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1515 	    zc->zc_iflags, &props))) {
1516 		spa_close(spa, FTAG);
1517 		nvlist_free(config);
1518 		return (error);
1519 	}
1520 
1521 	error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
1522 
1523 	spa_close(spa, FTAG);
1524 
1525 	nvlist_free(config);
1526 	nvlist_free(props);
1527 
1528 	return (error);
1529 }
1530 
1531 static int
1532 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
1533 {
1534 	spa_t *spa;
1535 	char *path = zc->zc_value;
1536 	uint64_t guid = zc->zc_guid;
1537 	int error;
1538 
1539 	error = spa_open(zc->zc_name, &spa, FTAG);
1540 	if (error != 0)
1541 		return (error);
1542 
1543 	error = spa_vdev_setpath(spa, guid, path);
1544 	spa_close(spa, FTAG);
1545 	return (error);
1546 }
1547 
1548 static int
1549 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
1550 {
1551 	spa_t *spa;
1552 	char *fru = zc->zc_value;
1553 	uint64_t guid = zc->zc_guid;
1554 	int error;
1555 
1556 	error = spa_open(zc->zc_name, &spa, FTAG);
1557 	if (error != 0)
1558 		return (error);
1559 
1560 	error = spa_vdev_setfru(spa, guid, fru);
1561 	spa_close(spa, FTAG);
1562 	return (error);
1563 }
1564 
1565 /*
1566  * inputs:
1567  * zc_name		name of filesystem
1568  * zc_nvlist_dst_size	size of buffer for property nvlist
1569  *
1570  * outputs:
1571  * zc_objset_stats	stats
1572  * zc_nvlist_dst	property nvlist
1573  * zc_nvlist_dst_size	size of property nvlist
1574  */
1575 static int
1576 zfs_ioc_objset_stats(zfs_cmd_t *zc)
1577 {
1578 	objset_t *os = NULL;
1579 	int error;
1580 	nvlist_t *nv;
1581 
1582 	if (error = dmu_objset_hold(zc->zc_name, FTAG, &os))
1583 		return (error);
1584 
1585 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1586 
1587 	if (zc->zc_nvlist_dst != 0 &&
1588 	    (error = dsl_prop_get_all(os, &nv)) == 0) {
1589 		dmu_objset_stats(os, nv);
1590 		/*
1591 		 * NB: zvol_get_stats() will read the objset contents,
1592 		 * which we aren't supposed to do with a
1593 		 * DS_MODE_USER hold, because it could be
1594 		 * inconsistent.  So this is a bit of a workaround...
1595 		 * XXX reading with out owning
1596 		 */
1597 		if (!zc->zc_objset_stats.dds_inconsistent) {
1598 			if (dmu_objset_type(os) == DMU_OST_ZVOL)
1599 				VERIFY(zvol_get_stats(os, nv) == 0);
1600 		}
1601 		error = put_nvlist(zc, nv);
1602 		nvlist_free(nv);
1603 	}
1604 
1605 	dmu_objset_rele(os, FTAG);
1606 	return (error);
1607 }
1608 
1609 /*
1610  * inputs:
1611  * zc_name		name of filesystem
1612  * zc_nvlist_dst_size	size of buffer for property nvlist
1613  *
1614  * outputs:
1615  * zc_nvlist_dst	received property nvlist
1616  * zc_nvlist_dst_size	size of received property nvlist
1617  *
1618  * Gets received properties (distinct from local properties on or after
1619  * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
1620  * local property values.
1621  */
1622 static int
1623 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
1624 {
1625 	objset_t *os = NULL;
1626 	int error;
1627 	nvlist_t *nv;
1628 
1629 	if (error = dmu_objset_hold(zc->zc_name, FTAG, &os))
1630 		return (error);
1631 
1632 	/*
1633 	 * Without this check, we would return local property values if the
1634 	 * caller has not already received properties on or after
1635 	 * SPA_VERSION_RECVD_PROPS.
1636 	 */
1637 	if (!dsl_prop_get_hasrecvd(os)) {
1638 		dmu_objset_rele(os, FTAG);
1639 		return (ENOTSUP);
1640 	}
1641 
1642 	if (zc->zc_nvlist_dst != 0 &&
1643 	    (error = dsl_prop_get_received(os, &nv)) == 0) {
1644 		error = put_nvlist(zc, nv);
1645 		nvlist_free(nv);
1646 	}
1647 
1648 	dmu_objset_rele(os, FTAG);
1649 	return (error);
1650 }
1651 
1652 static int
1653 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
1654 {
1655 	uint64_t value;
1656 	int error;
1657 
1658 	/*
1659 	 * zfs_get_zplprop() will either find a value or give us
1660 	 * the default value (if there is one).
1661 	 */
1662 	if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
1663 		return (error);
1664 	VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
1665 	return (0);
1666 }
1667 
1668 /*
1669  * inputs:
1670  * zc_name		name of filesystem
1671  * zc_nvlist_dst_size	size of buffer for zpl property nvlist
1672  *
1673  * outputs:
1674  * zc_nvlist_dst	zpl property nvlist
1675  * zc_nvlist_dst_size	size of zpl property nvlist
1676  */
1677 static int
1678 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
1679 {
1680 	objset_t *os;
1681 	int err;
1682 
1683 	/* XXX reading without owning */
1684 	if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
1685 		return (err);
1686 
1687 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1688 
1689 	/*
1690 	 * NB: nvl_add_zplprop() will read the objset contents,
1691 	 * which we aren't supposed to do with a DS_MODE_USER
1692 	 * hold, because it could be inconsistent.
1693 	 */
1694 	if (zc->zc_nvlist_dst != NULL &&
1695 	    !zc->zc_objset_stats.dds_inconsistent &&
1696 	    dmu_objset_type(os) == DMU_OST_ZFS) {
1697 		nvlist_t *nv;
1698 
1699 		VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1700 		if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
1701 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
1702 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
1703 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
1704 			err = put_nvlist(zc, nv);
1705 		nvlist_free(nv);
1706 	} else {
1707 		err = ENOENT;
1708 	}
1709 	dmu_objset_rele(os, FTAG);
1710 	return (err);
1711 }
1712 
1713 static boolean_t
1714 dataset_name_hidden(const char *name)
1715 {
1716 	/*
1717 	 * Skip over datasets that are not visible in this zone,
1718 	 * internal datasets (which have a $ in their name), and
1719 	 * temporary datasets (which have a % in their name).
1720 	 */
1721 	if (strchr(name, '$') != NULL)
1722 		return (B_TRUE);
1723 	if (strchr(name, '%') != NULL)
1724 		return (B_TRUE);
1725 	if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
1726 		return (B_TRUE);
1727 	return (B_FALSE);
1728 }
1729 
1730 /*
1731  * inputs:
1732  * zc_name		name of filesystem
1733  * zc_cookie		zap cursor
1734  * zc_nvlist_dst_size	size of buffer for property nvlist
1735  *
1736  * outputs:
1737  * zc_name		name of next filesystem
1738  * zc_cookie		zap cursor
1739  * zc_objset_stats	stats
1740  * zc_nvlist_dst	property nvlist
1741  * zc_nvlist_dst_size	size of property nvlist
1742  */
1743 static int
1744 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
1745 {
1746 	objset_t *os;
1747 	int error;
1748 	char *p;
1749 	size_t orig_len = strlen(zc->zc_name);
1750 
1751 top:
1752 	if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
1753 		if (error == ENOENT)
1754 			error = ESRCH;
1755 		return (error);
1756 	}
1757 
1758 	p = strrchr(zc->zc_name, '/');
1759 	if (p == NULL || p[1] != '\0')
1760 		(void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
1761 	p = zc->zc_name + strlen(zc->zc_name);
1762 
1763 	/*
1764 	 * Pre-fetch the datasets.  dmu_objset_prefetch() always returns 0
1765 	 * but is not declared void because its called by dmu_objset_find().
1766 	 */
1767 	if (zc->zc_cookie == 0) {
1768 		uint64_t cookie = 0;
1769 		int len = sizeof (zc->zc_name) - (p - zc->zc_name);
1770 
1771 		while (dmu_dir_list_next(os, len, p, NULL, &cookie) == 0)
1772 			(void) dmu_objset_prefetch(p, NULL);
1773 	}
1774 
1775 	do {
1776 		error = dmu_dir_list_next(os,
1777 		    sizeof (zc->zc_name) - (p - zc->zc_name), p,
1778 		    NULL, &zc->zc_cookie);
1779 		if (error == ENOENT)
1780 			error = ESRCH;
1781 	} while (error == 0 && dataset_name_hidden(zc->zc_name) &&
1782 	    !(zc->zc_iflags & FKIOCTL));
1783 	dmu_objset_rele(os, FTAG);
1784 
1785 	/*
1786 	 * If it's an internal dataset (ie. with a '$' in its name),
1787 	 * don't try to get stats for it, otherwise we'll return ENOENT.
1788 	 */
1789 	if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
1790 		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
1791 		if (error == ENOENT) {
1792 			/* We lost a race with destroy, get the next one. */
1793 			zc->zc_name[orig_len] = '\0';
1794 			goto top;
1795 		}
1796 	}
1797 	return (error);
1798 }
1799 
1800 /*
1801  * inputs:
1802  * zc_name		name of filesystem
1803  * zc_cookie		zap cursor
1804  * zc_nvlist_dst_size	size of buffer for property nvlist
1805  *
1806  * outputs:
1807  * zc_name		name of next snapshot
1808  * zc_objset_stats	stats
1809  * zc_nvlist_dst	property nvlist
1810  * zc_nvlist_dst_size	size of property nvlist
1811  */
1812 static int
1813 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
1814 {
1815 	objset_t *os;
1816 	int error;
1817 
1818 top:
1819 	if (zc->zc_cookie == 0)
1820 		(void) dmu_objset_find(zc->zc_name, dmu_objset_prefetch,
1821 		    NULL, DS_FIND_SNAPSHOTS);
1822 
1823 	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
1824 	if (error)
1825 		return (error == ENOENT ? ESRCH : error);
1826 
1827 	/*
1828 	 * A dataset name of maximum length cannot have any snapshots,
1829 	 * so exit immediately.
1830 	 */
1831 	if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
1832 		dmu_objset_rele(os, FTAG);
1833 		return (ESRCH);
1834 	}
1835 
1836 	error = dmu_snapshot_list_next(os,
1837 	    sizeof (zc->zc_name) - strlen(zc->zc_name),
1838 	    zc->zc_name + strlen(zc->zc_name), NULL, &zc->zc_cookie, NULL);
1839 	dmu_objset_rele(os, FTAG);
1840 	if (error == 0) {
1841 		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
1842 		if (error == ENOENT)  {
1843 			/* We lost a race with destroy, get the next one. */
1844 			*strchr(zc->zc_name, '@') = '\0';
1845 			goto top;
1846 		}
1847 	} else if (error == ENOENT) {
1848 		error = ESRCH;
1849 	}
1850 
1851 	/* if we failed, undo the @ that we tacked on to zc_name */
1852 	if (error)
1853 		*strchr(zc->zc_name, '@') = '\0';
1854 	return (error);
1855 }
1856 
1857 static int
1858 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
1859 {
1860 	const char *propname = nvpair_name(pair);
1861 	uint64_t *valary;
1862 	unsigned int vallen;
1863 	const char *domain;
1864 	zfs_userquota_prop_t type;
1865 	uint64_t rid;
1866 	uint64_t quota;
1867 	zfsvfs_t *zfsvfs;
1868 	int err;
1869 
1870 	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
1871 		nvlist_t *attrs;
1872 		VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
1873 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
1874 		    &pair) == 0);
1875 	}
1876 
1877 	VERIFY(nvpair_value_uint64_array(pair, &valary, &vallen) == 0);
1878 	VERIFY(vallen == 3);
1879 	type = valary[0];
1880 	rid = valary[1];
1881 	quota = valary[2];
1882 	/*
1883 	 * The propname is encoded as
1884 	 * userquota@<rid>-<domain>.
1885 	 */
1886 	domain = strchr(propname, '-') + 1;
1887 
1888 	err = zfsvfs_hold(dsname, FTAG, &zfsvfs);
1889 	if (err == 0) {
1890 		err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
1891 		zfsvfs_rele(zfsvfs, FTAG);
1892 	}
1893 
1894 	return (err);
1895 }
1896 
1897 /*
1898  * If the named property is one that has a special function to set its value,
1899  * return 0 on success and a positive error code on failure; otherwise if it is
1900  * not one of the special properties handled by this function, return -1.
1901  *
1902  * XXX: It would be better for callers of the properety interface if we handled
1903  * these special cases in dsl_prop.c (in the dsl layer).
1904  */
1905 static int
1906 zfs_prop_set_special(const char *dsname, zprop_source_t source,
1907     nvpair_t *pair)
1908 {
1909 	const char *propname = nvpair_name(pair);
1910 	zfs_prop_t prop = zfs_name_to_prop(propname);
1911 	uint64_t intval;
1912 	int err;
1913 
1914 	if (prop == ZPROP_INVAL) {
1915 		if (zfs_prop_userquota(propname))
1916 			return (zfs_prop_set_userquota(dsname, pair));
1917 		return (-1);
1918 	}
1919 
1920 	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
1921 		nvlist_t *attrs;
1922 		VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
1923 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
1924 		    &pair) == 0);
1925 	}
1926 
1927 	if (zfs_prop_get_type(prop) == PROP_TYPE_STRING)
1928 		return (-1);
1929 
1930 	VERIFY(0 == nvpair_value_uint64(pair, &intval));
1931 
1932 	switch (prop) {
1933 	case ZFS_PROP_QUOTA:
1934 		err = dsl_dir_set_quota(dsname, source, intval);
1935 		break;
1936 	case ZFS_PROP_REFQUOTA:
1937 		err = dsl_dataset_set_quota(dsname, source, intval);
1938 		break;
1939 	case ZFS_PROP_RESERVATION:
1940 		err = dsl_dir_set_reservation(dsname, source, intval);
1941 		break;
1942 	case ZFS_PROP_REFRESERVATION:
1943 		err = dsl_dataset_set_reservation(dsname, source, intval);
1944 		break;
1945 	case ZFS_PROP_VOLSIZE:
1946 		err = zvol_set_volsize(dsname, ddi_driver_major(zfs_dip),
1947 		    intval);
1948 		break;
1949 	case ZFS_PROP_VERSION:
1950 	{
1951 		zfsvfs_t *zfsvfs;
1952 
1953 		if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs)) != 0)
1954 			break;
1955 
1956 		err = zfs_set_version(zfsvfs, intval);
1957 		zfsvfs_rele(zfsvfs, FTAG);
1958 
1959 		if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
1960 			zfs_cmd_t *zc;
1961 
1962 			zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
1963 			(void) strcpy(zc->zc_name, dsname);
1964 			(void) zfs_ioc_userspace_upgrade(zc);
1965 			kmem_free(zc, sizeof (zfs_cmd_t));
1966 		}
1967 		break;
1968 	}
1969 
1970 	default:
1971 		err = -1;
1972 	}
1973 
1974 	return (err);
1975 }
1976 
1977 /*
1978  * This function is best effort. If it fails to set any of the given properties,
1979  * it continues to set as many as it can and returns the first error
1980  * encountered. If the caller provides a non-NULL errlist, it also gives the
1981  * complete list of names of all the properties it failed to set along with the
1982  * corresponding error numbers. The caller is responsible for freeing the
1983  * returned errlist.
1984  *
1985  * If every property is set successfully, zero is returned and the list pointed
1986  * at by errlist is NULL.
1987  */
1988 int
1989 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
1990     nvlist_t **errlist)
1991 {
1992 	nvpair_t *pair;
1993 	nvpair_t *propval;
1994 	int rv = 0;
1995 	uint64_t intval;
1996 	char *strval;
1997 	nvlist_t *genericnvl;
1998 	nvlist_t *errors;
1999 	nvlist_t *retrynvl;
2000 
2001 	VERIFY(nvlist_alloc(&genericnvl, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2002 	VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2003 	VERIFY(nvlist_alloc(&retrynvl, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2004 
2005 retry:
2006 	pair = NULL;
2007 	while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2008 		const char *propname = nvpair_name(pair);
2009 		zfs_prop_t prop = zfs_name_to_prop(propname);
2010 		int err = 0;
2011 
2012 		/* decode the property value */
2013 		propval = pair;
2014 		if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2015 			nvlist_t *attrs;
2016 			VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2017 			VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2018 			    &propval) == 0);
2019 		}
2020 
2021 		/* Validate value type */
2022 		if (prop == ZPROP_INVAL) {
2023 			if (zfs_prop_user(propname)) {
2024 				if (nvpair_type(propval) != DATA_TYPE_STRING)
2025 					err = EINVAL;
2026 			} else if (zfs_prop_userquota(propname)) {
2027 				if (nvpair_type(propval) !=
2028 				    DATA_TYPE_UINT64_ARRAY)
2029 					err = EINVAL;
2030 			}
2031 		} else {
2032 			if (nvpair_type(propval) == DATA_TYPE_STRING) {
2033 				if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2034 					err = EINVAL;
2035 			} else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2036 				const char *unused;
2037 
2038 				VERIFY(nvpair_value_uint64(propval,
2039 				    &intval) == 0);
2040 
2041 				switch (zfs_prop_get_type(prop)) {
2042 				case PROP_TYPE_NUMBER:
2043 					break;
2044 				case PROP_TYPE_STRING:
2045 					err = EINVAL;
2046 					break;
2047 				case PROP_TYPE_INDEX:
2048 					if (zfs_prop_index_to_string(prop,
2049 					    intval, &unused) != 0)
2050 						err = EINVAL;
2051 					break;
2052 				default:
2053 					cmn_err(CE_PANIC,
2054 					    "unknown property type");
2055 				}
2056 			} else {
2057 				err = EINVAL;
2058 			}
2059 		}
2060 
2061 		/* Validate permissions */
2062 		if (err == 0)
2063 			err = zfs_check_settable(dsname, pair, CRED());
2064 
2065 		if (err == 0) {
2066 			err = zfs_prop_set_special(dsname, source, pair);
2067 			if (err == -1) {
2068 				/*
2069 				 * For better performance we build up a list of
2070 				 * properties to set in a single transaction.
2071 				 */
2072 				err = nvlist_add_nvpair(genericnvl, pair);
2073 			} else if (err != 0 && nvl != retrynvl) {
2074 				/*
2075 				 * This may be a spurious error caused by
2076 				 * receiving quota and reservation out of order.
2077 				 * Try again in a second pass.
2078 				 */
2079 				err = nvlist_add_nvpair(retrynvl, pair);
2080 			}
2081 		}
2082 
2083 		if (err != 0)
2084 			VERIFY(nvlist_add_int32(errors, propname, err) == 0);
2085 	}
2086 
2087 	if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2088 		nvl = retrynvl;
2089 		goto retry;
2090 	}
2091 
2092 	if (!nvlist_empty(genericnvl) &&
2093 	    dsl_props_set(dsname, source, genericnvl) != 0) {
2094 		/*
2095 		 * If this fails, we still want to set as many properties as we
2096 		 * can, so try setting them individually.
2097 		 */
2098 		pair = NULL;
2099 		while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2100 			const char *propname = nvpair_name(pair);
2101 			int err = 0;
2102 
2103 			propval = pair;
2104 			if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2105 				nvlist_t *attrs;
2106 				VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2107 				VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2108 				    &propval) == 0);
2109 			}
2110 
2111 			if (nvpair_type(propval) == DATA_TYPE_STRING) {
2112 				VERIFY(nvpair_value_string(propval,
2113 				    &strval) == 0);
2114 				err = dsl_prop_set(dsname, propname, source, 1,
2115 				    strlen(strval) + 1, strval);
2116 			} else {
2117 				VERIFY(nvpair_value_uint64(propval,
2118 				    &intval) == 0);
2119 				err = dsl_prop_set(dsname, propname, source, 8,
2120 				    1, &intval);
2121 			}
2122 
2123 			if (err != 0) {
2124 				VERIFY(nvlist_add_int32(errors, propname,
2125 				    err) == 0);
2126 			}
2127 		}
2128 	}
2129 	nvlist_free(genericnvl);
2130 	nvlist_free(retrynvl);
2131 
2132 	if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
2133 		nvlist_free(errors);
2134 		errors = NULL;
2135 	} else {
2136 		VERIFY(nvpair_value_int32(pair, &rv) == 0);
2137 	}
2138 
2139 	if (errlist == NULL)
2140 		nvlist_free(errors);
2141 	else
2142 		*errlist = errors;
2143 
2144 	return (rv);
2145 }
2146 
2147 /*
2148  * Check that all the properties are valid user properties.
2149  */
2150 static int
2151 zfs_check_userprops(char *fsname, nvlist_t *nvl)
2152 {
2153 	nvpair_t *pair = NULL;
2154 	int error = 0;
2155 
2156 	while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2157 		const char *propname = nvpair_name(pair);
2158 		char *valstr;
2159 
2160 		if (!zfs_prop_user(propname) ||
2161 		    nvpair_type(pair) != DATA_TYPE_STRING)
2162 			return (EINVAL);
2163 
2164 		if (error = zfs_secpolicy_write_perms(fsname,
2165 		    ZFS_DELEG_PERM_USERPROP, CRED()))
2166 			return (error);
2167 
2168 		if (strlen(propname) >= ZAP_MAXNAMELEN)
2169 			return (ENAMETOOLONG);
2170 
2171 		VERIFY(nvpair_value_string(pair, &valstr) == 0);
2172 		if (strlen(valstr) >= ZAP_MAXVALUELEN)
2173 			return (E2BIG);
2174 	}
2175 	return (0);
2176 }
2177 
2178 static void
2179 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2180 {
2181 	nvpair_t *pair;
2182 
2183 	VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2184 
2185 	pair = NULL;
2186 	while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2187 		if (nvlist_exists(skipped, nvpair_name(pair)))
2188 			continue;
2189 
2190 		VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2191 	}
2192 }
2193 
2194 static int
2195 clear_received_props(objset_t *os, const char *fs, nvlist_t *props,
2196     nvlist_t *skipped)
2197 {
2198 	int err = 0;
2199 	nvlist_t *cleared_props = NULL;
2200 	props_skip(props, skipped, &cleared_props);
2201 	if (!nvlist_empty(cleared_props)) {
2202 		/*
2203 		 * Acts on local properties until the dataset has received
2204 		 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2205 		 */
2206 		zprop_source_t flags = (ZPROP_SRC_NONE |
2207 		    (dsl_prop_get_hasrecvd(os) ? ZPROP_SRC_RECEIVED : 0));
2208 		err = zfs_set_prop_nvlist(fs, flags, cleared_props, NULL);
2209 	}
2210 	nvlist_free(cleared_props);
2211 	return (err);
2212 }
2213 
2214 /*
2215  * inputs:
2216  * zc_name		name of filesystem
2217  * zc_value		name of property to set
2218  * zc_nvlist_src{_size}	nvlist of properties to apply
2219  * zc_cookie		received properties flag
2220  *
2221  * outputs:
2222  * zc_nvlist_dst{_size} error for each unapplied received property
2223  */
2224 static int
2225 zfs_ioc_set_prop(zfs_cmd_t *zc)
2226 {
2227 	nvlist_t *nvl;
2228 	boolean_t received = zc->zc_cookie;
2229 	zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2230 	    ZPROP_SRC_LOCAL);
2231 	nvlist_t *errors = NULL;
2232 	int error;
2233 
2234 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2235 	    zc->zc_iflags, &nvl)) != 0)
2236 		return (error);
2237 
2238 	if (received) {
2239 		nvlist_t *origprops;
2240 		objset_t *os;
2241 
2242 		if (dmu_objset_hold(zc->zc_name, FTAG, &os) == 0) {
2243 			if (dsl_prop_get_received(os, &origprops) == 0) {
2244 				(void) clear_received_props(os,
2245 				    zc->zc_name, origprops, nvl);
2246 				nvlist_free(origprops);
2247 			}
2248 
2249 			dsl_prop_set_hasrecvd(os);
2250 			dmu_objset_rele(os, FTAG);
2251 		}
2252 	}
2253 
2254 	error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, &errors);
2255 
2256 	if (zc->zc_nvlist_dst != NULL && errors != NULL) {
2257 		(void) put_nvlist(zc, errors);
2258 	}
2259 
2260 	nvlist_free(errors);
2261 	nvlist_free(nvl);
2262 	return (error);
2263 }
2264 
2265 /*
2266  * inputs:
2267  * zc_name		name of filesystem
2268  * zc_value		name of property to inherit
2269  * zc_cookie		revert to received value if TRUE
2270  *
2271  * outputs:		none
2272  */
2273 static int
2274 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2275 {
2276 	const char *propname = zc->zc_value;
2277 	zfs_prop_t prop = zfs_name_to_prop(propname);
2278 	boolean_t received = zc->zc_cookie;
2279 	zprop_source_t source = (received
2280 	    ? ZPROP_SRC_NONE		/* revert to received value, if any */
2281 	    : ZPROP_SRC_INHERITED);	/* explicitly inherit */
2282 
2283 	if (received) {
2284 		nvlist_t *dummy;
2285 		nvpair_t *pair;
2286 		zprop_type_t type;
2287 		int err;
2288 
2289 		/*
2290 		 * zfs_prop_set_special() expects properties in the form of an
2291 		 * nvpair with type info.
2292 		 */
2293 		if (prop == ZPROP_INVAL) {
2294 			if (!zfs_prop_user(propname))
2295 				return (EINVAL);
2296 
2297 			type = PROP_TYPE_STRING;
2298 		} else if (prop == ZFS_PROP_VOLSIZE ||
2299 		    prop == ZFS_PROP_VERSION) {
2300 			return (EINVAL);
2301 		} else {
2302 			type = zfs_prop_get_type(prop);
2303 		}
2304 
2305 		VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2306 
2307 		switch (type) {
2308 		case PROP_TYPE_STRING:
2309 			VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2310 			break;
2311 		case PROP_TYPE_NUMBER:
2312 		case PROP_TYPE_INDEX:
2313 			VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2314 			break;
2315 		default:
2316 			nvlist_free(dummy);
2317 			return (EINVAL);
2318 		}
2319 
2320 		pair = nvlist_next_nvpair(dummy, NULL);
2321 		err = zfs_prop_set_special(zc->zc_name, source, pair);
2322 		nvlist_free(dummy);
2323 		if (err != -1)
2324 			return (err); /* special property already handled */
2325 	} else {
2326 		/*
2327 		 * Only check this in the non-received case. We want to allow
2328 		 * 'inherit -S' to revert non-inheritable properties like quota
2329 		 * and reservation to the received or default values even though
2330 		 * they are not considered inheritable.
2331 		 */
2332 		if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2333 			return (EINVAL);
2334 	}
2335 
2336 	/* the property name has been validated by zfs_secpolicy_inherit() */
2337 	return (dsl_prop_set(zc->zc_name, zc->zc_value, source, 0, 0, NULL));
2338 }
2339 
2340 static int
2341 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2342 {
2343 	nvlist_t *props;
2344 	spa_t *spa;
2345 	int error;
2346 	nvpair_t *pair;
2347 
2348 	if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2349 	    zc->zc_iflags, &props))
2350 		return (error);
2351 
2352 	/*
2353 	 * If the only property is the configfile, then just do a spa_lookup()
2354 	 * to handle the faulted case.
2355 	 */
2356 	pair = nvlist_next_nvpair(props, NULL);
2357 	if (pair != NULL && strcmp(nvpair_name(pair),
2358 	    zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2359 	    nvlist_next_nvpair(props, pair) == NULL) {
2360 		mutex_enter(&spa_namespace_lock);
2361 		if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2362 			spa_configfile_set(spa, props, B_FALSE);
2363 			spa_config_sync(spa, B_FALSE, B_TRUE);
2364 		}
2365 		mutex_exit(&spa_namespace_lock);
2366 		if (spa != NULL) {
2367 			nvlist_free(props);
2368 			return (0);
2369 		}
2370 	}
2371 
2372 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2373 		nvlist_free(props);
2374 		return (error);
2375 	}
2376 
2377 	error = spa_prop_set(spa, props);
2378 
2379 	nvlist_free(props);
2380 	spa_close(spa, FTAG);
2381 
2382 	return (error);
2383 }
2384 
2385 static int
2386 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2387 {
2388 	spa_t *spa;
2389 	int error;
2390 	nvlist_t *nvp = NULL;
2391 
2392 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2393 		/*
2394 		 * If the pool is faulted, there may be properties we can still
2395 		 * get (such as altroot and cachefile), so attempt to get them
2396 		 * anyway.
2397 		 */
2398 		mutex_enter(&spa_namespace_lock);
2399 		if ((spa = spa_lookup(zc->zc_name)) != NULL)
2400 			error = spa_prop_get(spa, &nvp);
2401 		mutex_exit(&spa_namespace_lock);
2402 	} else {
2403 		error = spa_prop_get(spa, &nvp);
2404 		spa_close(spa, FTAG);
2405 	}
2406 
2407 	if (error == 0 && zc->zc_nvlist_dst != NULL)
2408 		error = put_nvlist(zc, nvp);
2409 	else
2410 		error = EFAULT;
2411 
2412 	nvlist_free(nvp);
2413 	return (error);
2414 }
2415 
2416 /*
2417  * inputs:
2418  * zc_name		name of filesystem
2419  * zc_nvlist_src{_size}	nvlist of delegated permissions
2420  * zc_perm_action	allow/unallow flag
2421  *
2422  * outputs:		none
2423  */
2424 static int
2425 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2426 {
2427 	int error;
2428 	nvlist_t *fsaclnv = NULL;
2429 
2430 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2431 	    zc->zc_iflags, &fsaclnv)) != 0)
2432 		return (error);
2433 
2434 	/*
2435 	 * Verify nvlist is constructed correctly
2436 	 */
2437 	if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2438 		nvlist_free(fsaclnv);
2439 		return (EINVAL);
2440 	}
2441 
2442 	/*
2443 	 * If we don't have PRIV_SYS_MOUNT, then validate
2444 	 * that user is allowed to hand out each permission in
2445 	 * the nvlist(s)
2446 	 */
2447 
2448 	error = secpolicy_zfs(CRED());
2449 	if (error) {
2450 		if (zc->zc_perm_action == B_FALSE) {
2451 			error = dsl_deleg_can_allow(zc->zc_name,
2452 			    fsaclnv, CRED());
2453 		} else {
2454 			error = dsl_deleg_can_unallow(zc->zc_name,
2455 			    fsaclnv, CRED());
2456 		}
2457 	}
2458 
2459 	if (error == 0)
2460 		error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2461 
2462 	nvlist_free(fsaclnv);
2463 	return (error);
2464 }
2465 
2466 /*
2467  * inputs:
2468  * zc_name		name of filesystem
2469  *
2470  * outputs:
2471  * zc_nvlist_src{_size}	nvlist of delegated permissions
2472  */
2473 static int
2474 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2475 {
2476 	nvlist_t *nvp;
2477 	int error;
2478 
2479 	if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2480 		error = put_nvlist(zc, nvp);
2481 		nvlist_free(nvp);
2482 	}
2483 
2484 	return (error);
2485 }
2486 
2487 /*
2488  * Search the vfs list for a specified resource.  Returns a pointer to it
2489  * or NULL if no suitable entry is found. The caller of this routine
2490  * is responsible for releasing the returned vfs pointer.
2491  */
2492 static vfs_t *
2493 zfs_get_vfs(const char *resource)
2494 {
2495 	struct vfs *vfsp;
2496 	struct vfs *vfs_found = NULL;
2497 
2498 	vfs_list_read_lock();
2499 	vfsp = rootvfs;
2500 	do {
2501 		if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
2502 			VFS_HOLD(vfsp);
2503 			vfs_found = vfsp;
2504 			break;
2505 		}
2506 		vfsp = vfsp->vfs_next;
2507 	} while (vfsp != rootvfs);
2508 	vfs_list_unlock();
2509 	return (vfs_found);
2510 }
2511 
2512 /* ARGSUSED */
2513 static void
2514 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
2515 {
2516 	zfs_creat_t *zct = arg;
2517 
2518 	zfs_create_fs(os, cr, zct->zct_zplprops, tx);
2519 }
2520 
2521 #define	ZFS_PROP_UNDEFINED	((uint64_t)-1)
2522 
2523 /*
2524  * inputs:
2525  * createprops		list of properties requested by creator
2526  * default_zplver	zpl version to use if unspecified in createprops
2527  * fuids_ok		fuids allowed in this version of the spa?
2528  * os			parent objset pointer (NULL if root fs)
2529  *
2530  * outputs:
2531  * zplprops	values for the zplprops we attach to the master node object
2532  * is_ci	true if requested file system will be purely case-insensitive
2533  *
2534  * Determine the settings for utf8only, normalization and
2535  * casesensitivity.  Specific values may have been requested by the
2536  * creator and/or we can inherit values from the parent dataset.  If
2537  * the file system is of too early a vintage, a creator can not
2538  * request settings for these properties, even if the requested
2539  * setting is the default value.  We don't actually want to create dsl
2540  * properties for these, so remove them from the source nvlist after
2541  * processing.
2542  */
2543 static int
2544 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
2545     boolean_t fuids_ok, nvlist_t *createprops, nvlist_t *zplprops,
2546     boolean_t *is_ci)
2547 {
2548 	uint64_t sense = ZFS_PROP_UNDEFINED;
2549 	uint64_t norm = ZFS_PROP_UNDEFINED;
2550 	uint64_t u8 = ZFS_PROP_UNDEFINED;
2551 
2552 	ASSERT(zplprops != NULL);
2553 
2554 	/*
2555 	 * Pull out creator prop choices, if any.
2556 	 */
2557 	if (createprops) {
2558 		(void) nvlist_lookup_uint64(createprops,
2559 		    zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
2560 		(void) nvlist_lookup_uint64(createprops,
2561 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
2562 		(void) nvlist_remove_all(createprops,
2563 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE));
2564 		(void) nvlist_lookup_uint64(createprops,
2565 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
2566 		(void) nvlist_remove_all(createprops,
2567 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
2568 		(void) nvlist_lookup_uint64(createprops,
2569 		    zfs_prop_to_name(ZFS_PROP_CASE), &sense);
2570 		(void) nvlist_remove_all(createprops,
2571 		    zfs_prop_to_name(ZFS_PROP_CASE));
2572 	}
2573 
2574 	/*
2575 	 * If the zpl version requested is whacky or the file system
2576 	 * or pool is version is too "young" to support normalization
2577 	 * and the creator tried to set a value for one of the props,
2578 	 * error out.
2579 	 */
2580 	if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
2581 	    (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
2582 	    (zplver < ZPL_VERSION_NORMALIZATION &&
2583 	    (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
2584 	    sense != ZFS_PROP_UNDEFINED)))
2585 		return (ENOTSUP);
2586 
2587 	/*
2588 	 * Put the version in the zplprops
2589 	 */
2590 	VERIFY(nvlist_add_uint64(zplprops,
2591 	    zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
2592 
2593 	if (norm == ZFS_PROP_UNDEFINED)
2594 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
2595 	VERIFY(nvlist_add_uint64(zplprops,
2596 	    zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
2597 
2598 	/*
2599 	 * If we're normalizing, names must always be valid UTF-8 strings.
2600 	 */
2601 	if (norm)
2602 		u8 = 1;
2603 	if (u8 == ZFS_PROP_UNDEFINED)
2604 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
2605 	VERIFY(nvlist_add_uint64(zplprops,
2606 	    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
2607 
2608 	if (sense == ZFS_PROP_UNDEFINED)
2609 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
2610 	VERIFY(nvlist_add_uint64(zplprops,
2611 	    zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
2612 
2613 	if (is_ci)
2614 		*is_ci = (sense == ZFS_CASE_INSENSITIVE);
2615 
2616 	return (0);
2617 }
2618 
2619 static int
2620 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
2621     nvlist_t *zplprops, boolean_t *is_ci)
2622 {
2623 	boolean_t fuids_ok = B_TRUE;
2624 	uint64_t zplver = ZPL_VERSION;
2625 	objset_t *os = NULL;
2626 	char parentname[MAXNAMELEN];
2627 	char *cp;
2628 	int error;
2629 
2630 	(void) strlcpy(parentname, dataset, sizeof (parentname));
2631 	cp = strrchr(parentname, '/');
2632 	ASSERT(cp != NULL);
2633 	cp[0] = '\0';
2634 
2635 	if (zfs_earlier_version(dataset, SPA_VERSION_USERSPACE))
2636 		zplver = ZPL_VERSION_USERSPACE - 1;
2637 	if (zfs_earlier_version(dataset, SPA_VERSION_FUID)) {
2638 		zplver = ZPL_VERSION_FUID - 1;
2639 		fuids_ok = B_FALSE;
2640 	}
2641 
2642 	/*
2643 	 * Open parent object set so we can inherit zplprop values.
2644 	 */
2645 	if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
2646 		return (error);
2647 
2648 	error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, createprops,
2649 	    zplprops, is_ci);
2650 	dmu_objset_rele(os, FTAG);
2651 	return (error);
2652 }
2653 
2654 static int
2655 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
2656     nvlist_t *zplprops, boolean_t *is_ci)
2657 {
2658 	boolean_t fuids_ok = B_TRUE;
2659 	uint64_t zplver = ZPL_VERSION;
2660 	int error;
2661 
2662 	if (spa_vers < SPA_VERSION_FUID) {
2663 		zplver = ZPL_VERSION_FUID - 1;
2664 		fuids_ok = B_FALSE;
2665 	}
2666 
2667 	error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, createprops,
2668 	    zplprops, is_ci);
2669 	return (error);
2670 }
2671 
2672 /*
2673  * inputs:
2674  * zc_objset_type	type of objset to create (fs vs zvol)
2675  * zc_name		name of new objset
2676  * zc_value		name of snapshot to clone from (may be empty)
2677  * zc_nvlist_src{_size}	nvlist of properties to apply
2678  *
2679  * outputs: none
2680  */
2681 static int
2682 zfs_ioc_create(zfs_cmd_t *zc)
2683 {
2684 	objset_t *clone;
2685 	int error = 0;
2686 	zfs_creat_t zct;
2687 	nvlist_t *nvprops = NULL;
2688 	void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
2689 	dmu_objset_type_t type = zc->zc_objset_type;
2690 
2691 	switch (type) {
2692 
2693 	case DMU_OST_ZFS:
2694 		cbfunc = zfs_create_cb;
2695 		break;
2696 
2697 	case DMU_OST_ZVOL:
2698 		cbfunc = zvol_create_cb;
2699 		break;
2700 
2701 	default:
2702 		cbfunc = NULL;
2703 		break;
2704 	}
2705 	if (strchr(zc->zc_name, '@') ||
2706 	    strchr(zc->zc_name, '%'))
2707 		return (EINVAL);
2708 
2709 	if (zc->zc_nvlist_src != NULL &&
2710 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2711 	    zc->zc_iflags, &nvprops)) != 0)
2712 		return (error);
2713 
2714 	zct.zct_zplprops = NULL;
2715 	zct.zct_props = nvprops;
2716 
2717 	if (zc->zc_value[0] != '\0') {
2718 		/*
2719 		 * We're creating a clone of an existing snapshot.
2720 		 */
2721 		zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
2722 		if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0) {
2723 			nvlist_free(nvprops);
2724 			return (EINVAL);
2725 		}
2726 
2727 		error = dmu_objset_hold(zc->zc_value, FTAG, &clone);
2728 		if (error) {
2729 			nvlist_free(nvprops);
2730 			return (error);
2731 		}
2732 
2733 		error = dmu_objset_clone(zc->zc_name, dmu_objset_ds(clone), 0);
2734 		dmu_objset_rele(clone, FTAG);
2735 		if (error) {
2736 			nvlist_free(nvprops);
2737 			return (error);
2738 		}
2739 	} else {
2740 		boolean_t is_insensitive = B_FALSE;
2741 
2742 		if (cbfunc == NULL) {
2743 			nvlist_free(nvprops);
2744 			return (EINVAL);
2745 		}
2746 
2747 		if (type == DMU_OST_ZVOL) {
2748 			uint64_t volsize, volblocksize;
2749 
2750 			if (nvprops == NULL ||
2751 			    nvlist_lookup_uint64(nvprops,
2752 			    zfs_prop_to_name(ZFS_PROP_VOLSIZE),
2753 			    &volsize) != 0) {
2754 				nvlist_free(nvprops);
2755 				return (EINVAL);
2756 			}
2757 
2758 			if ((error = nvlist_lookup_uint64(nvprops,
2759 			    zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
2760 			    &volblocksize)) != 0 && error != ENOENT) {
2761 				nvlist_free(nvprops);
2762 				return (EINVAL);
2763 			}
2764 
2765 			if (error != 0)
2766 				volblocksize = zfs_prop_default_numeric(
2767 				    ZFS_PROP_VOLBLOCKSIZE);
2768 
2769 			if ((error = zvol_check_volblocksize(
2770 			    volblocksize)) != 0 ||
2771 			    (error = zvol_check_volsize(volsize,
2772 			    volblocksize)) != 0) {
2773 				nvlist_free(nvprops);
2774 				return (error);
2775 			}
2776 		} else if (type == DMU_OST_ZFS) {
2777 			int error;
2778 
2779 			/*
2780 			 * We have to have normalization and
2781 			 * case-folding flags correct when we do the
2782 			 * file system creation, so go figure them out
2783 			 * now.
2784 			 */
2785 			VERIFY(nvlist_alloc(&zct.zct_zplprops,
2786 			    NV_UNIQUE_NAME, KM_SLEEP) == 0);
2787 			error = zfs_fill_zplprops(zc->zc_name, nvprops,
2788 			    zct.zct_zplprops, &is_insensitive);
2789 			if (error != 0) {
2790 				nvlist_free(nvprops);
2791 				nvlist_free(zct.zct_zplprops);
2792 				return (error);
2793 			}
2794 		}
2795 		error = dmu_objset_create(zc->zc_name, type,
2796 		    is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
2797 		nvlist_free(zct.zct_zplprops);
2798 	}
2799 
2800 	/*
2801 	 * It would be nice to do this atomically.
2802 	 */
2803 	if (error == 0) {
2804 		error = zfs_set_prop_nvlist(zc->zc_name, ZPROP_SRC_LOCAL,
2805 		    nvprops, NULL);
2806 		if (error != 0)
2807 			(void) dmu_objset_destroy(zc->zc_name, B_FALSE);
2808 	}
2809 	nvlist_free(nvprops);
2810 	return (error);
2811 }
2812 
2813 /*
2814  * inputs:
2815  * zc_name	name of filesystem
2816  * zc_value	short name of snapshot
2817  * zc_cookie	recursive flag
2818  * zc_nvlist_src[_size] property list
2819  *
2820  * outputs:
2821  * zc_value	short snapname (i.e. part after the '@')
2822  */
2823 static int
2824 zfs_ioc_snapshot(zfs_cmd_t *zc)
2825 {
2826 	nvlist_t *nvprops = NULL;
2827 	int error;
2828 	boolean_t recursive = zc->zc_cookie;
2829 
2830 	if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
2831 		return (EINVAL);
2832 
2833 	if (zc->zc_nvlist_src != NULL &&
2834 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2835 	    zc->zc_iflags, &nvprops)) != 0)
2836 		return (error);
2837 
2838 	error = zfs_check_userprops(zc->zc_name, nvprops);
2839 	if (error)
2840 		goto out;
2841 
2842 	if (!nvlist_empty(nvprops) &&
2843 	    zfs_earlier_version(zc->zc_name, SPA_VERSION_SNAP_PROPS)) {
2844 		error = ENOTSUP;
2845 		goto out;
2846 	}
2847 
2848 	error = dmu_objset_snapshot(zc->zc_name, zc->zc_value,
2849 	    nvprops, recursive);
2850 
2851 out:
2852 	nvlist_free(nvprops);
2853 	return (error);
2854 }
2855 
2856 int
2857 zfs_unmount_snap(const char *name, void *arg)
2858 {
2859 	vfs_t *vfsp = NULL;
2860 
2861 	if (arg) {
2862 		char *snapname = arg;
2863 		char *fullname = kmem_asprintf("%s@%s", name, snapname);
2864 		vfsp = zfs_get_vfs(fullname);
2865 		strfree(fullname);
2866 	} else if (strchr(name, '@')) {
2867 		vfsp = zfs_get_vfs(name);
2868 	}
2869 
2870 	if (vfsp) {
2871 		/*
2872 		 * Always force the unmount for snapshots.
2873 		 */
2874 		int flag = MS_FORCE;
2875 		int err;
2876 
2877 		if ((err = vn_vfswlock(vfsp->vfs_vnodecovered)) != 0) {
2878 			VFS_RELE(vfsp);
2879 			return (err);
2880 		}
2881 		VFS_RELE(vfsp);
2882 		if ((err = dounmount(vfsp, flag, kcred)) != 0)
2883 			return (err);
2884 	}
2885 	return (0);
2886 }
2887 
2888 /*
2889  * inputs:
2890  * zc_name		name of filesystem
2891  * zc_value		short name of snapshot
2892  * zc_defer_destroy	mark for deferred destroy
2893  *
2894  * outputs:	none
2895  */
2896 static int
2897 zfs_ioc_destroy_snaps(zfs_cmd_t *zc)
2898 {
2899 	int err;
2900 
2901 	if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
2902 		return (EINVAL);
2903 	err = dmu_objset_find(zc->zc_name,
2904 	    zfs_unmount_snap, zc->zc_value, DS_FIND_CHILDREN);
2905 	if (err)
2906 		return (err);
2907 	return (dmu_snapshots_destroy(zc->zc_name, zc->zc_value,
2908 	    zc->zc_defer_destroy));
2909 }
2910 
2911 /*
2912  * inputs:
2913  * zc_name		name of dataset to destroy
2914  * zc_objset_type	type of objset
2915  * zc_defer_destroy	mark for deferred destroy
2916  *
2917  * outputs:		none
2918  */
2919 static int
2920 zfs_ioc_destroy(zfs_cmd_t *zc)
2921 {
2922 	int err;
2923 	if (strchr(zc->zc_name, '@') && zc->zc_objset_type == DMU_OST_ZFS) {
2924 		err = zfs_unmount_snap(zc->zc_name, NULL);
2925 		if (err)
2926 			return (err);
2927 	}
2928 
2929 	err = dmu_objset_destroy(zc->zc_name, zc->zc_defer_destroy);
2930 	if (zc->zc_objset_type == DMU_OST_ZVOL && err == 0)
2931 		(void) zvol_remove_minor(zc->zc_name);
2932 	return (err);
2933 }
2934 
2935 /*
2936  * inputs:
2937  * zc_name	name of dataset to rollback (to most recent snapshot)
2938  *
2939  * outputs:	none
2940  */
2941 static int
2942 zfs_ioc_rollback(zfs_cmd_t *zc)
2943 {
2944 	dsl_dataset_t *ds, *clone;
2945 	int error;
2946 	zfsvfs_t *zfsvfs;
2947 	char *clone_name;
2948 
2949 	error = dsl_dataset_hold(zc->zc_name, FTAG, &ds);
2950 	if (error)
2951 		return (error);
2952 
2953 	/* must not be a snapshot */
2954 	if (dsl_dataset_is_snapshot(ds)) {
2955 		dsl_dataset_rele(ds, FTAG);
2956 		return (EINVAL);
2957 	}
2958 
2959 	/* must have a most recent snapshot */
2960 	if (ds->ds_phys->ds_prev_snap_txg < TXG_INITIAL) {
2961 		dsl_dataset_rele(ds, FTAG);
2962 		return (EINVAL);
2963 	}
2964 
2965 	/*
2966 	 * Create clone of most recent snapshot.
2967 	 */
2968 	clone_name = kmem_asprintf("%s/%%rollback", zc->zc_name);
2969 	error = dmu_objset_clone(clone_name, ds->ds_prev, DS_FLAG_INCONSISTENT);
2970 	if (error)
2971 		goto out;
2972 
2973 	error = dsl_dataset_own(clone_name, B_TRUE, FTAG, &clone);
2974 	if (error)
2975 		goto out;
2976 
2977 	/*
2978 	 * Do clone swap.
2979 	 */
2980 	if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
2981 		error = zfs_suspend_fs(zfsvfs);
2982 		if (error == 0) {
2983 			int resume_err;
2984 
2985 			if (dsl_dataset_tryown(ds, B_FALSE, FTAG)) {
2986 				error = dsl_dataset_clone_swap(clone, ds,
2987 				    B_TRUE);
2988 				dsl_dataset_disown(ds, FTAG);
2989 				ds = NULL;
2990 			} else {
2991 				error = EBUSY;
2992 			}
2993 			resume_err = zfs_resume_fs(zfsvfs, zc->zc_name);
2994 			error = error ? error : resume_err;
2995 		}
2996 		VFS_RELE(zfsvfs->z_vfs);
2997 	} else {
2998 		if (dsl_dataset_tryown(ds, B_FALSE, FTAG)) {
2999 			error = dsl_dataset_clone_swap(clone, ds, B_TRUE);
3000 			dsl_dataset_disown(ds, FTAG);
3001 			ds = NULL;
3002 		} else {
3003 			error = EBUSY;
3004 		}
3005 	}
3006 
3007 	/*
3008 	 * Destroy clone (which also closes it).
3009 	 */
3010 	(void) dsl_dataset_destroy(clone, FTAG, B_FALSE);
3011 
3012 out:
3013 	strfree(clone_name);
3014 	if (ds)
3015 		dsl_dataset_rele(ds, FTAG);
3016 	return (error);
3017 }
3018 
3019 /*
3020  * inputs:
3021  * zc_name	old name of dataset
3022  * zc_value	new name of dataset
3023  * zc_cookie	recursive flag (only valid for snapshots)
3024  *
3025  * outputs:	none
3026  */
3027 static int
3028 zfs_ioc_rename(zfs_cmd_t *zc)
3029 {
3030 	boolean_t recursive = zc->zc_cookie & 1;
3031 
3032 	zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
3033 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3034 	    strchr(zc->zc_value, '%'))
3035 		return (EINVAL);
3036 
3037 	/*
3038 	 * Unmount snapshot unless we're doing a recursive rename,
3039 	 * in which case the dataset code figures out which snapshots
3040 	 * to unmount.
3041 	 */
3042 	if (!recursive && strchr(zc->zc_name, '@') != NULL &&
3043 	    zc->zc_objset_type == DMU_OST_ZFS) {
3044 		int err = zfs_unmount_snap(zc->zc_name, NULL);
3045 		if (err)
3046 			return (err);
3047 	}
3048 	if (zc->zc_objset_type == DMU_OST_ZVOL)
3049 		(void) zvol_remove_minor(zc->zc_name);
3050 	return (dmu_objset_rename(zc->zc_name, zc->zc_value, recursive));
3051 }
3052 
3053 static int
3054 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
3055 {
3056 	const char *propname = nvpair_name(pair);
3057 	boolean_t issnap = (strchr(dsname, '@') != NULL);
3058 	zfs_prop_t prop = zfs_name_to_prop(propname);
3059 	uint64_t intval;
3060 	int err;
3061 
3062 	if (prop == ZPROP_INVAL) {
3063 		if (zfs_prop_user(propname)) {
3064 			if (err = zfs_secpolicy_write_perms(dsname,
3065 			    ZFS_DELEG_PERM_USERPROP, cr))
3066 				return (err);
3067 			return (0);
3068 		}
3069 
3070 		if (!issnap && zfs_prop_userquota(propname)) {
3071 			const char *perm = NULL;
3072 			const char *uq_prefix =
3073 			    zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
3074 			const char *gq_prefix =
3075 			    zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
3076 
3077 			if (strncmp(propname, uq_prefix,
3078 			    strlen(uq_prefix)) == 0) {
3079 				perm = ZFS_DELEG_PERM_USERQUOTA;
3080 			} else if (strncmp(propname, gq_prefix,
3081 			    strlen(gq_prefix)) == 0) {
3082 				perm = ZFS_DELEG_PERM_GROUPQUOTA;
3083 			} else {
3084 				/* USERUSED and GROUPUSED are read-only */
3085 				return (EINVAL);
3086 			}
3087 
3088 			if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
3089 				return (err);
3090 			return (0);
3091 		}
3092 
3093 		return (EINVAL);
3094 	}
3095 
3096 	if (issnap)
3097 		return (EINVAL);
3098 
3099 	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
3100 		/*
3101 		 * dsl_prop_get_all_impl() returns properties in this
3102 		 * format.
3103 		 */
3104 		nvlist_t *attrs;
3105 		VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
3106 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3107 		    &pair) == 0);
3108 	}
3109 
3110 	/*
3111 	 * Check that this value is valid for this pool version
3112 	 */
3113 	switch (prop) {
3114 	case ZFS_PROP_COMPRESSION:
3115 		/*
3116 		 * If the user specified gzip compression, make sure
3117 		 * the SPA supports it. We ignore any errors here since
3118 		 * we'll catch them later.
3119 		 */
3120 		if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3121 		    nvpair_value_uint64(pair, &intval) == 0) {
3122 			if (intval >= ZIO_COMPRESS_GZIP_1 &&
3123 			    intval <= ZIO_COMPRESS_GZIP_9 &&
3124 			    zfs_earlier_version(dsname,
3125 			    SPA_VERSION_GZIP_COMPRESSION)) {
3126 				return (ENOTSUP);
3127 			}
3128 
3129 			if (intval == ZIO_COMPRESS_ZLE &&
3130 			    zfs_earlier_version(dsname,
3131 			    SPA_VERSION_ZLE_COMPRESSION))
3132 				return (ENOTSUP);
3133 
3134 			/*
3135 			 * If this is a bootable dataset then
3136 			 * verify that the compression algorithm
3137 			 * is supported for booting. We must return
3138 			 * something other than ENOTSUP since it
3139 			 * implies a downrev pool version.
3140 			 */
3141 			if (zfs_is_bootfs(dsname) &&
3142 			    !BOOTFS_COMPRESS_VALID(intval)) {
3143 				return (ERANGE);
3144 			}
3145 		}
3146 		break;
3147 
3148 	case ZFS_PROP_COPIES:
3149 		if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
3150 			return (ENOTSUP);
3151 		break;
3152 
3153 	case ZFS_PROP_DEDUP:
3154 		if (zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
3155 			return (ENOTSUP);
3156 		break;
3157 
3158 	case ZFS_PROP_SHARESMB:
3159 		if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
3160 			return (ENOTSUP);
3161 		break;
3162 
3163 	case ZFS_PROP_ACLINHERIT:
3164 		if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
3165 		    nvpair_value_uint64(pair, &intval) == 0) {
3166 			if (intval == ZFS_ACL_PASSTHROUGH_X &&
3167 			    zfs_earlier_version(dsname,
3168 			    SPA_VERSION_PASSTHROUGH_X))
3169 				return (ENOTSUP);
3170 		}
3171 		break;
3172 	}
3173 
3174 	return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
3175 }
3176 
3177 /*
3178  * Removes properties from the given props list that fail permission checks
3179  * needed to clear them and to restore them in case of a receive error. For each
3180  * property, make sure we have both set and inherit permissions.
3181  *
3182  * Returns the first error encountered if any permission checks fail. If the
3183  * caller provides a non-NULL errlist, it also gives the complete list of names
3184  * of all the properties that failed a permission check along with the
3185  * corresponding error numbers. The caller is responsible for freeing the
3186  * returned errlist.
3187  *
3188  * If every property checks out successfully, zero is returned and the list
3189  * pointed at by errlist is NULL.
3190  */
3191 static int
3192 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
3193 {
3194 	zfs_cmd_t *zc;
3195 	nvpair_t *pair, *next_pair;
3196 	nvlist_t *errors;
3197 	int err, rv = 0;
3198 
3199 	if (props == NULL)
3200 		return (0);
3201 
3202 	VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3203 
3204 	zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
3205 	(void) strcpy(zc->zc_name, dataset);
3206 	pair = nvlist_next_nvpair(props, NULL);
3207 	while (pair != NULL) {
3208 		next_pair = nvlist_next_nvpair(props, pair);
3209 
3210 		(void) strcpy(zc->zc_value, nvpair_name(pair));
3211 		if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
3212 		    (err = zfs_secpolicy_inherit(zc, CRED())) != 0) {
3213 			VERIFY(nvlist_remove_nvpair(props, pair) == 0);
3214 			VERIFY(nvlist_add_int32(errors,
3215 			    zc->zc_value, err) == 0);
3216 		}
3217 		pair = next_pair;
3218 	}
3219 	kmem_free(zc, sizeof (zfs_cmd_t));
3220 
3221 	if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
3222 		nvlist_free(errors);
3223 		errors = NULL;
3224 	} else {
3225 		VERIFY(nvpair_value_int32(pair, &rv) == 0);
3226 	}
3227 
3228 	if (errlist == NULL)
3229 		nvlist_free(errors);
3230 	else
3231 		*errlist = errors;
3232 
3233 	return (rv);
3234 }
3235 
3236 static boolean_t
3237 propval_equals(nvpair_t *p1, nvpair_t *p2)
3238 {
3239 	if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
3240 		/* dsl_prop_get_all_impl() format */
3241 		nvlist_t *attrs;
3242 		VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
3243 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3244 		    &p1) == 0);
3245 	}
3246 
3247 	if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
3248 		nvlist_t *attrs;
3249 		VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
3250 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
3251 		    &p2) == 0);
3252 	}
3253 
3254 	if (nvpair_type(p1) != nvpair_type(p2))
3255 		return (B_FALSE);
3256 
3257 	if (nvpair_type(p1) == DATA_TYPE_STRING) {
3258 		char *valstr1, *valstr2;
3259 
3260 		VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
3261 		VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
3262 		return (strcmp(valstr1, valstr2) == 0);
3263 	} else {
3264 		uint64_t intval1, intval2;
3265 
3266 		VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
3267 		VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
3268 		return (intval1 == intval2);
3269 	}
3270 }
3271 
3272 /*
3273  * Remove properties from props if they are not going to change (as determined
3274  * by comparison with origprops). Remove them from origprops as well, since we
3275  * do not need to clear or restore properties that won't change.
3276  */
3277 static void
3278 props_reduce(nvlist_t *props, nvlist_t *origprops)
3279 {
3280 	nvpair_t *pair, *next_pair;
3281 
3282 	if (origprops == NULL)
3283 		return; /* all props need to be received */
3284 
3285 	pair = nvlist_next_nvpair(props, NULL);
3286 	while (pair != NULL) {
3287 		const char *propname = nvpair_name(pair);
3288 		nvpair_t *match;
3289 
3290 		next_pair = nvlist_next_nvpair(props, pair);
3291 
3292 		if ((nvlist_lookup_nvpair(origprops, propname,
3293 		    &match) != 0) || !propval_equals(pair, match))
3294 			goto next; /* need to set received value */
3295 
3296 		/* don't clear the existing received value */
3297 		(void) nvlist_remove_nvpair(origprops, match);
3298 		/* don't bother receiving the property */
3299 		(void) nvlist_remove_nvpair(props, pair);
3300 next:
3301 		pair = next_pair;
3302 	}
3303 }
3304 
3305 #ifdef	DEBUG
3306 static boolean_t zfs_ioc_recv_inject_err;
3307 #endif
3308 
3309 /*
3310  * inputs:
3311  * zc_name		name of containing filesystem
3312  * zc_nvlist_src{_size}	nvlist of properties to apply
3313  * zc_value		name of snapshot to create
3314  * zc_string		name of clone origin (if DRR_FLAG_CLONE)
3315  * zc_cookie		file descriptor to recv from
3316  * zc_begin_record	the BEGIN record of the stream (not byteswapped)
3317  * zc_guid		force flag
3318  *
3319  * outputs:
3320  * zc_cookie		number of bytes read
3321  * zc_nvlist_dst{_size} error for each unapplied received property
3322  * zc_obj		zprop_errflags_t
3323  */
3324 static int
3325 zfs_ioc_recv(zfs_cmd_t *zc)
3326 {
3327 	file_t *fp;
3328 	objset_t *os;
3329 	dmu_recv_cookie_t drc;
3330 	boolean_t force = (boolean_t)zc->zc_guid;
3331 	int fd;
3332 	int error = 0;
3333 	int props_error = 0;
3334 	nvlist_t *errors;
3335 	offset_t off;
3336 	nvlist_t *props = NULL; /* sent properties */
3337 	nvlist_t *origprops = NULL; /* existing properties */
3338 	objset_t *origin = NULL;
3339 	char *tosnap;
3340 	char tofs[ZFS_MAXNAMELEN];
3341 	boolean_t first_recvd_props = B_FALSE;
3342 
3343 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
3344 	    strchr(zc->zc_value, '@') == NULL ||
3345 	    strchr(zc->zc_value, '%'))
3346 		return (EINVAL);
3347 
3348 	(void) strcpy(tofs, zc->zc_value);
3349 	tosnap = strchr(tofs, '@');
3350 	*tosnap++ = '\0';
3351 
3352 	if (zc->zc_nvlist_src != NULL &&
3353 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3354 	    zc->zc_iflags, &props)) != 0)
3355 		return (error);
3356 
3357 	fd = zc->zc_cookie;
3358 	fp = getf(fd);
3359 	if (fp == NULL) {
3360 		nvlist_free(props);
3361 		return (EBADF);
3362 	}
3363 
3364 	VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
3365 
3366 	if (props && dmu_objset_hold(tofs, FTAG, &os) == 0) {
3367 		if ((spa_version(os->os_spa) >= SPA_VERSION_RECVD_PROPS) &&
3368 		    !dsl_prop_get_hasrecvd(os)) {
3369 			first_recvd_props = B_TRUE;
3370 		}
3371 
3372 		/*
3373 		 * If new received properties are supplied, they are to
3374 		 * completely replace the existing received properties, so stash
3375 		 * away the existing ones.
3376 		 */
3377 		if (dsl_prop_get_received(os, &origprops) == 0) {
3378 			nvlist_t *errlist = NULL;
3379 			/*
3380 			 * Don't bother writing a property if its value won't
3381 			 * change (and avoid the unnecessary security checks).
3382 			 *
3383 			 * The first receive after SPA_VERSION_RECVD_PROPS is a
3384 			 * special case where we blow away all local properties
3385 			 * regardless.
3386 			 */
3387 			if (!first_recvd_props)
3388 				props_reduce(props, origprops);
3389 			if (zfs_check_clearable(tofs, origprops,
3390 			    &errlist) != 0)
3391 				(void) nvlist_merge(errors, errlist, 0);
3392 			nvlist_free(errlist);
3393 		}
3394 
3395 		dmu_objset_rele(os, FTAG);
3396 	}
3397 
3398 	if (zc->zc_string[0]) {
3399 		error = dmu_objset_hold(zc->zc_string, FTAG, &origin);
3400 		if (error)
3401 			goto out;
3402 	}
3403 
3404 	error = dmu_recv_begin(tofs, tosnap, zc->zc_top_ds,
3405 	    &zc->zc_begin_record, force, origin, &drc);
3406 	if (origin)
3407 		dmu_objset_rele(origin, FTAG);
3408 	if (error)
3409 		goto out;
3410 
3411 	/*
3412 	 * Set properties before we receive the stream so that they are applied
3413 	 * to the new data. Note that we must call dmu_recv_stream() if
3414 	 * dmu_recv_begin() succeeds.
3415 	 */
3416 	if (props) {
3417 		nvlist_t *errlist;
3418 
3419 		if (dmu_objset_from_ds(drc.drc_logical_ds, &os) == 0) {
3420 			if (drc.drc_newfs) {
3421 				if (spa_version(os->os_spa) >=
3422 				    SPA_VERSION_RECVD_PROPS)
3423 					first_recvd_props = B_TRUE;
3424 			} else if (origprops != NULL) {
3425 				if (clear_received_props(os, tofs, origprops,
3426 				    first_recvd_props ? NULL : props) != 0)
3427 					zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3428 			} else {
3429 				zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3430 			}
3431 			dsl_prop_set_hasrecvd(os);
3432 		} else if (!drc.drc_newfs) {
3433 			zc->zc_obj |= ZPROP_ERR_NOCLEAR;
3434 		}
3435 
3436 		(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
3437 		    props, &errlist);
3438 		(void) nvlist_merge(errors, errlist, 0);
3439 		nvlist_free(errlist);
3440 	}
3441 
3442 	if (fit_error_list(zc, &errors) != 0 || put_nvlist(zc, errors) != 0) {
3443 		/*
3444 		 * Caller made zc->zc_nvlist_dst less than the minimum expected
3445 		 * size or supplied an invalid address.
3446 		 */
3447 		props_error = EINVAL;
3448 	}
3449 
3450 	off = fp->f_offset;
3451 	error = dmu_recv_stream(&drc, fp->f_vnode, &off);
3452 
3453 	if (error == 0) {
3454 		zfsvfs_t *zfsvfs = NULL;
3455 
3456 		if (getzfsvfs(tofs, &zfsvfs) == 0) {
3457 			/* online recv */
3458 			int end_err;
3459 
3460 			error = zfs_suspend_fs(zfsvfs);
3461 			/*
3462 			 * If the suspend fails, then the recv_end will
3463 			 * likely also fail, and clean up after itself.
3464 			 */
3465 			end_err = dmu_recv_end(&drc);
3466 			if (error == 0)
3467 				error = zfs_resume_fs(zfsvfs, tofs);
3468 			error = error ? error : end_err;
3469 			VFS_RELE(zfsvfs->z_vfs);
3470 		} else {
3471 			error = dmu_recv_end(&drc);
3472 		}
3473 	}
3474 
3475 	zc->zc_cookie = off - fp->f_offset;
3476 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
3477 		fp->f_offset = off;
3478 
3479 #ifdef	DEBUG
3480 	if (zfs_ioc_recv_inject_err) {
3481 		zfs_ioc_recv_inject_err = B_FALSE;
3482 		error = 1;
3483 	}
3484 #endif
3485 	/*
3486 	 * On error, restore the original props.
3487 	 */
3488 	if (error && props) {
3489 		if (dmu_objset_hold(tofs, FTAG, &os) == 0) {
3490 			if (clear_received_props(os, tofs, props, NULL) != 0) {
3491 				/*
3492 				 * We failed to clear the received properties.
3493 				 * Since we may have left a $recvd value on the
3494 				 * system, we can't clear the $hasrecvd flag.
3495 				 */
3496 				zc->zc_obj |= ZPROP_ERR_NORESTORE;
3497 			} else if (first_recvd_props) {
3498 				dsl_prop_unset_hasrecvd(os);
3499 			}
3500 			dmu_objset_rele(os, FTAG);
3501 		} else if (!drc.drc_newfs) {
3502 			/* We failed to clear the received properties. */
3503 			zc->zc_obj |= ZPROP_ERR_NORESTORE;
3504 		}
3505 
3506 		if (origprops == NULL && !drc.drc_newfs) {
3507 			/* We failed to stash the original properties. */
3508 			zc->zc_obj |= ZPROP_ERR_NORESTORE;
3509 		}
3510 
3511 		/*
3512 		 * dsl_props_set() will not convert RECEIVED to LOCAL on or
3513 		 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
3514 		 * explictly if we're restoring local properties cleared in the
3515 		 * first new-style receive.
3516 		 */
3517 		if (origprops != NULL &&
3518 		    zfs_set_prop_nvlist(tofs, (first_recvd_props ?
3519 		    ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
3520 		    origprops, NULL) != 0) {
3521 			/*
3522 			 * We stashed the original properties but failed to
3523 			 * restore them.
3524 			 */
3525 			zc->zc_obj |= ZPROP_ERR_NORESTORE;
3526 		}
3527 	}
3528 out:
3529 	nvlist_free(props);
3530 	nvlist_free(origprops);
3531 	nvlist_free(errors);
3532 	releasef(fd);
3533 
3534 	if (error == 0)
3535 		error = props_error;
3536 
3537 	return (error);
3538 }
3539 
3540 /*
3541  * inputs:
3542  * zc_name	name of snapshot to send
3543  * zc_value	short name of incremental fromsnap (may be empty)
3544  * zc_cookie	file descriptor to send stream to
3545  * zc_obj	fromorigin flag (mutually exclusive with zc_value)
3546  *
3547  * outputs: none
3548  */
3549 static int
3550 zfs_ioc_send(zfs_cmd_t *zc)
3551 {
3552 	objset_t *fromsnap = NULL;
3553 	objset_t *tosnap;
3554 	file_t *fp;
3555 	int error;
3556 	offset_t off;
3557 
3558 	error = dmu_objset_hold(zc->zc_name, FTAG, &tosnap);
3559 	if (error)
3560 		return (error);
3561 
3562 	if (zc->zc_value[0] != '\0') {
3563 		char *buf;
3564 		char *cp;
3565 
3566 		buf = kmem_alloc(MAXPATHLEN, KM_SLEEP);
3567 		(void) strncpy(buf, zc->zc_name, MAXPATHLEN);
3568 		cp = strchr(buf, '@');
3569 		if (cp)
3570 			*(cp+1) = 0;
3571 		(void) strncat(buf, zc->zc_value, MAXPATHLEN);
3572 		error = dmu_objset_hold(buf, FTAG, &fromsnap);
3573 		kmem_free(buf, MAXPATHLEN);
3574 		if (error) {
3575 			dmu_objset_rele(tosnap, FTAG);
3576 			return (error);
3577 		}
3578 	}
3579 
3580 	fp = getf(zc->zc_cookie);
3581 	if (fp == NULL) {
3582 		dmu_objset_rele(tosnap, FTAG);
3583 		if (fromsnap)
3584 			dmu_objset_rele(fromsnap, FTAG);
3585 		return (EBADF);
3586 	}
3587 
3588 	off = fp->f_offset;
3589 	error = dmu_sendbackup(tosnap, fromsnap, zc->zc_obj, fp->f_vnode, &off);
3590 
3591 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
3592 		fp->f_offset = off;
3593 	releasef(zc->zc_cookie);
3594 	if (fromsnap)
3595 		dmu_objset_rele(fromsnap, FTAG);
3596 	dmu_objset_rele(tosnap, FTAG);
3597 	return (error);
3598 }
3599 
3600 static int
3601 zfs_ioc_inject_fault(zfs_cmd_t *zc)
3602 {
3603 	int id, error;
3604 
3605 	error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
3606 	    &zc->zc_inject_record);
3607 
3608 	if (error == 0)
3609 		zc->zc_guid = (uint64_t)id;
3610 
3611 	return (error);
3612 }
3613 
3614 static int
3615 zfs_ioc_clear_fault(zfs_cmd_t *zc)
3616 {
3617 	return (zio_clear_fault((int)zc->zc_guid));
3618 }
3619 
3620 static int
3621 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
3622 {
3623 	int id = (int)zc->zc_guid;
3624 	int error;
3625 
3626 	error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
3627 	    &zc->zc_inject_record);
3628 
3629 	zc->zc_guid = id;
3630 
3631 	return (error);
3632 }
3633 
3634 static int
3635 zfs_ioc_error_log(zfs_cmd_t *zc)
3636 {
3637 	spa_t *spa;
3638 	int error;
3639 	size_t count = (size_t)zc->zc_nvlist_dst_size;
3640 
3641 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
3642 		return (error);
3643 
3644 	error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
3645 	    &count);
3646 	if (error == 0)
3647 		zc->zc_nvlist_dst_size = count;
3648 	else
3649 		zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
3650 
3651 	spa_close(spa, FTAG);
3652 
3653 	return (error);
3654 }
3655 
3656 static int
3657 zfs_ioc_clear(zfs_cmd_t *zc)
3658 {
3659 	spa_t *spa;
3660 	vdev_t *vd;
3661 	int error;
3662 
3663 	/*
3664 	 * On zpool clear we also fix up missing slogs
3665 	 */
3666 	mutex_enter(&spa_namespace_lock);
3667 	spa = spa_lookup(zc->zc_name);
3668 	if (spa == NULL) {
3669 		mutex_exit(&spa_namespace_lock);
3670 		return (EIO);
3671 	}
3672 	if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
3673 		/* we need to let spa_open/spa_load clear the chains */
3674 		spa_set_log_state(spa, SPA_LOG_CLEAR);
3675 	}
3676 	spa->spa_last_open_failed = 0;
3677 	mutex_exit(&spa_namespace_lock);
3678 
3679 	if (zc->zc_cookie & ZPOOL_NO_REWIND) {
3680 		error = spa_open(zc->zc_name, &spa, FTAG);
3681 	} else {
3682 		nvlist_t *policy;
3683 		nvlist_t *config = NULL;
3684 
3685 		if (zc->zc_nvlist_src == NULL)
3686 			return (EINVAL);
3687 
3688 		if ((error = get_nvlist(zc->zc_nvlist_src,
3689 		    zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
3690 			error = spa_open_rewind(zc->zc_name, &spa, FTAG,
3691 			    policy, &config);
3692 			if (config != NULL) {
3693 				(void) put_nvlist(zc, config);
3694 				nvlist_free(config);
3695 			}
3696 			nvlist_free(policy);
3697 		}
3698 	}
3699 
3700 	if (error)
3701 		return (error);
3702 
3703 	spa_vdev_state_enter(spa, SCL_NONE);
3704 
3705 	if (zc->zc_guid == 0) {
3706 		vd = NULL;
3707 	} else {
3708 		vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
3709 		if (vd == NULL) {
3710 			(void) spa_vdev_state_exit(spa, NULL, ENODEV);
3711 			spa_close(spa, FTAG);
3712 			return (ENODEV);
3713 		}
3714 	}
3715 
3716 	vdev_clear(spa, vd);
3717 
3718 	(void) spa_vdev_state_exit(spa, NULL, 0);
3719 
3720 	/*
3721 	 * Resume any suspended I/Os.
3722 	 */
3723 	if (zio_resume(spa) != 0)
3724 		error = EIO;
3725 
3726 	spa_close(spa, FTAG);
3727 
3728 	return (error);
3729 }
3730 
3731 /*
3732  * inputs:
3733  * zc_name	name of filesystem
3734  * zc_value	name of origin snapshot
3735  *
3736  * outputs:
3737  * zc_string	name of conflicting snapshot, if there is one
3738  */
3739 static int
3740 zfs_ioc_promote(zfs_cmd_t *zc)
3741 {
3742 	char *cp;
3743 
3744 	/*
3745 	 * We don't need to unmount *all* the origin fs's snapshots, but
3746 	 * it's easier.
3747 	 */
3748 	cp = strchr(zc->zc_value, '@');
3749 	if (cp)
3750 		*cp = '\0';
3751 	(void) dmu_objset_find(zc->zc_value,
3752 	    zfs_unmount_snap, NULL, DS_FIND_SNAPSHOTS);
3753 	return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
3754 }
3755 
3756 /*
3757  * Retrieve a single {user|group}{used|quota}@... property.
3758  *
3759  * inputs:
3760  * zc_name	name of filesystem
3761  * zc_objset_type zfs_userquota_prop_t
3762  * zc_value	domain name (eg. "S-1-234-567-89")
3763  * zc_guid	RID/UID/GID
3764  *
3765  * outputs:
3766  * zc_cookie	property value
3767  */
3768 static int
3769 zfs_ioc_userspace_one(zfs_cmd_t *zc)
3770 {
3771 	zfsvfs_t *zfsvfs;
3772 	int error;
3773 
3774 	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
3775 		return (EINVAL);
3776 
3777 	error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs);
3778 	if (error)
3779 		return (error);
3780 
3781 	error = zfs_userspace_one(zfsvfs,
3782 	    zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
3783 	zfsvfs_rele(zfsvfs, FTAG);
3784 
3785 	return (error);
3786 }
3787 
3788 /*
3789  * inputs:
3790  * zc_name		name of filesystem
3791  * zc_cookie		zap cursor
3792  * zc_objset_type	zfs_userquota_prop_t
3793  * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
3794  *
3795  * outputs:
3796  * zc_nvlist_dst[_size]	data buffer (array of zfs_useracct_t)
3797  * zc_cookie	zap cursor
3798  */
3799 static int
3800 zfs_ioc_userspace_many(zfs_cmd_t *zc)
3801 {
3802 	zfsvfs_t *zfsvfs;
3803 	int error;
3804 
3805 	error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs);
3806 	if (error)
3807 		return (error);
3808 
3809 	int bufsize = zc->zc_nvlist_dst_size;
3810 	void *buf = kmem_alloc(bufsize, KM_SLEEP);
3811 
3812 	error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
3813 	    buf, &zc->zc_nvlist_dst_size);
3814 
3815 	if (error == 0) {
3816 		error = xcopyout(buf,
3817 		    (void *)(uintptr_t)zc->zc_nvlist_dst,
3818 		    zc->zc_nvlist_dst_size);
3819 	}
3820 	kmem_free(buf, bufsize);
3821 	zfsvfs_rele(zfsvfs, FTAG);
3822 
3823 	return (error);
3824 }
3825 
3826 /*
3827  * inputs:
3828  * zc_name		name of filesystem
3829  *
3830  * outputs:
3831  * none
3832  */
3833 static int
3834 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
3835 {
3836 	objset_t *os;
3837 	int error = 0;
3838 	zfsvfs_t *zfsvfs;
3839 
3840 	if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
3841 		if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
3842 			/*
3843 			 * If userused is not enabled, it may be because the
3844 			 * objset needs to be closed & reopened (to grow the
3845 			 * objset_phys_t).  Suspend/resume the fs will do that.
3846 			 */
3847 			error = zfs_suspend_fs(zfsvfs);
3848 			if (error == 0)
3849 				error = zfs_resume_fs(zfsvfs, zc->zc_name);
3850 		}
3851 		if (error == 0)
3852 			error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
3853 		VFS_RELE(zfsvfs->z_vfs);
3854 	} else {
3855 		/* XXX kind of reading contents without owning */
3856 		error = dmu_objset_hold(zc->zc_name, FTAG, &os);
3857 		if (error)
3858 			return (error);
3859 
3860 		error = dmu_objset_userspace_upgrade(os);
3861 		dmu_objset_rele(os, FTAG);
3862 	}
3863 
3864 	return (error);
3865 }
3866 
3867 /*
3868  * We don't want to have a hard dependency
3869  * against some special symbols in sharefs
3870  * nfs, and smbsrv.  Determine them if needed when
3871  * the first file system is shared.
3872  * Neither sharefs, nfs or smbsrv are unloadable modules.
3873  */
3874 int (*znfsexport_fs)(void *arg);
3875 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
3876 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
3877 
3878 int zfs_nfsshare_inited;
3879 int zfs_smbshare_inited;
3880 
3881 ddi_modhandle_t nfs_mod;
3882 ddi_modhandle_t sharefs_mod;
3883 ddi_modhandle_t smbsrv_mod;
3884 kmutex_t zfs_share_lock;
3885 
3886 static int
3887 zfs_init_sharefs()
3888 {
3889 	int error;
3890 
3891 	ASSERT(MUTEX_HELD(&zfs_share_lock));
3892 	/* Both NFS and SMB shares also require sharetab support. */
3893 	if (sharefs_mod == NULL && ((sharefs_mod =
3894 	    ddi_modopen("fs/sharefs",
3895 	    KRTLD_MODE_FIRST, &error)) == NULL)) {
3896 		return (ENOSYS);
3897 	}
3898 	if (zshare_fs == NULL && ((zshare_fs =
3899 	    (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
3900 	    ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
3901 		return (ENOSYS);
3902 	}
3903 	return (0);
3904 }
3905 
3906 static int
3907 zfs_ioc_share(zfs_cmd_t *zc)
3908 {
3909 	int error;
3910 	int opcode;
3911 
3912 	switch (zc->zc_share.z_sharetype) {
3913 	case ZFS_SHARE_NFS:
3914 	case ZFS_UNSHARE_NFS:
3915 		if (zfs_nfsshare_inited == 0) {
3916 			mutex_enter(&zfs_share_lock);
3917 			if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
3918 			    KRTLD_MODE_FIRST, &error)) == NULL)) {
3919 				mutex_exit(&zfs_share_lock);
3920 				return (ENOSYS);
3921 			}
3922 			if (znfsexport_fs == NULL &&
3923 			    ((znfsexport_fs = (int (*)(void *))
3924 			    ddi_modsym(nfs_mod,
3925 			    "nfs_export", &error)) == NULL)) {
3926 				mutex_exit(&zfs_share_lock);
3927 				return (ENOSYS);
3928 			}
3929 			error = zfs_init_sharefs();
3930 			if (error) {
3931 				mutex_exit(&zfs_share_lock);
3932 				return (ENOSYS);
3933 			}
3934 			zfs_nfsshare_inited = 1;
3935 			mutex_exit(&zfs_share_lock);
3936 		}
3937 		break;
3938 	case ZFS_SHARE_SMB:
3939 	case ZFS_UNSHARE_SMB:
3940 		if (zfs_smbshare_inited == 0) {
3941 			mutex_enter(&zfs_share_lock);
3942 			if (smbsrv_mod == NULL && ((smbsrv_mod =
3943 			    ddi_modopen("drv/smbsrv",
3944 			    KRTLD_MODE_FIRST, &error)) == NULL)) {
3945 				mutex_exit(&zfs_share_lock);
3946 				return (ENOSYS);
3947 			}
3948 			if (zsmbexport_fs == NULL && ((zsmbexport_fs =
3949 			    (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
3950 			    "smb_server_share", &error)) == NULL)) {
3951 				mutex_exit(&zfs_share_lock);
3952 				return (ENOSYS);
3953 			}
3954 			error = zfs_init_sharefs();
3955 			if (error) {
3956 				mutex_exit(&zfs_share_lock);
3957 				return (ENOSYS);
3958 			}
3959 			zfs_smbshare_inited = 1;
3960 			mutex_exit(&zfs_share_lock);
3961 		}
3962 		break;
3963 	default:
3964 		return (EINVAL);
3965 	}
3966 
3967 	switch (zc->zc_share.z_sharetype) {
3968 	case ZFS_SHARE_NFS:
3969 	case ZFS_UNSHARE_NFS:
3970 		if (error =
3971 		    znfsexport_fs((void *)
3972 		    (uintptr_t)zc->zc_share.z_exportdata))
3973 			return (error);
3974 		break;
3975 	case ZFS_SHARE_SMB:
3976 	case ZFS_UNSHARE_SMB:
3977 		if (error = zsmbexport_fs((void *)
3978 		    (uintptr_t)zc->zc_share.z_exportdata,
3979 		    zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
3980 		    B_TRUE: B_FALSE)) {
3981 			return (error);
3982 		}
3983 		break;
3984 	}
3985 
3986 	opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
3987 	    zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
3988 	    SHAREFS_ADD : SHAREFS_REMOVE;
3989 
3990 	/*
3991 	 * Add or remove share from sharetab
3992 	 */
3993 	error = zshare_fs(opcode,
3994 	    (void *)(uintptr_t)zc->zc_share.z_sharedata,
3995 	    zc->zc_share.z_sharemax);
3996 
3997 	return (error);
3998 
3999 }
4000 
4001 ace_t full_access[] = {
4002 	{(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
4003 };
4004 
4005 /*
4006  * Remove all ACL files in shares dir
4007  */
4008 static int
4009 zfs_smb_acl_purge(znode_t *dzp)
4010 {
4011 	zap_cursor_t	zc;
4012 	zap_attribute_t	zap;
4013 	zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
4014 	int error;
4015 
4016 	for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
4017 	    (error = zap_cursor_retrieve(&zc, &zap)) == 0;
4018 	    zap_cursor_advance(&zc)) {
4019 		if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
4020 		    NULL, 0)) != 0)
4021 			break;
4022 	}
4023 	zap_cursor_fini(&zc);
4024 	return (error);
4025 }
4026 
4027 static int
4028 zfs_ioc_smb_acl(zfs_cmd_t *zc)
4029 {
4030 	vnode_t *vp;
4031 	znode_t *dzp;
4032 	vnode_t *resourcevp = NULL;
4033 	znode_t *sharedir;
4034 	zfsvfs_t *zfsvfs;
4035 	nvlist_t *nvlist;
4036 	char *src, *target;
4037 	vattr_t vattr;
4038 	vsecattr_t vsec;
4039 	int error = 0;
4040 
4041 	if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
4042 	    NO_FOLLOW, NULL, &vp)) != 0)
4043 		return (error);
4044 
4045 	/* Now make sure mntpnt and dataset are ZFS */
4046 
4047 	if (vp->v_vfsp->vfs_fstype != zfsfstype ||
4048 	    (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
4049 	    zc->zc_name) != 0)) {
4050 		VN_RELE(vp);
4051 		return (EINVAL);
4052 	}
4053 
4054 	dzp = VTOZ(vp);
4055 	zfsvfs = dzp->z_zfsvfs;
4056 	ZFS_ENTER(zfsvfs);
4057 
4058 	/*
4059 	 * Create share dir if its missing.
4060 	 */
4061 	mutex_enter(&zfsvfs->z_lock);
4062 	if (zfsvfs->z_shares_dir == 0) {
4063 		dmu_tx_t *tx;
4064 
4065 		tx = dmu_tx_create(zfsvfs->z_os);
4066 		dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
4067 		    ZFS_SHARES_DIR);
4068 		dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
4069 		error = dmu_tx_assign(tx, TXG_WAIT);
4070 		if (error) {
4071 			dmu_tx_abort(tx);
4072 		} else {
4073 			error = zfs_create_share_dir(zfsvfs, tx);
4074 			dmu_tx_commit(tx);
4075 		}
4076 		if (error) {
4077 			mutex_exit(&zfsvfs->z_lock);
4078 			VN_RELE(vp);
4079 			ZFS_EXIT(zfsvfs);
4080 			return (error);
4081 		}
4082 	}
4083 	mutex_exit(&zfsvfs->z_lock);
4084 
4085 	ASSERT(zfsvfs->z_shares_dir);
4086 	if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
4087 		VN_RELE(vp);
4088 		ZFS_EXIT(zfsvfs);
4089 		return (error);
4090 	}
4091 
4092 	switch (zc->zc_cookie) {
4093 	case ZFS_SMB_ACL_ADD:
4094 		vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
4095 		vattr.va_type = VREG;
4096 		vattr.va_mode = S_IFREG|0777;
4097 		vattr.va_uid = 0;
4098 		vattr.va_gid = 0;
4099 
4100 		vsec.vsa_mask = VSA_ACE;
4101 		vsec.vsa_aclentp = &full_access;
4102 		vsec.vsa_aclentsz = sizeof (full_access);
4103 		vsec.vsa_aclcnt = 1;
4104 
4105 		error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
4106 		    &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
4107 		if (resourcevp)
4108 			VN_RELE(resourcevp);
4109 		break;
4110 
4111 	case ZFS_SMB_ACL_REMOVE:
4112 		error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
4113 		    NULL, 0);
4114 		break;
4115 
4116 	case ZFS_SMB_ACL_RENAME:
4117 		if ((error = get_nvlist(zc->zc_nvlist_src,
4118 		    zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
4119 			VN_RELE(vp);
4120 			ZFS_EXIT(zfsvfs);
4121 			return (error);
4122 		}
4123 		if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
4124 		    nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
4125 		    &target)) {
4126 			VN_RELE(vp);
4127 			VN_RELE(ZTOV(sharedir));
4128 			ZFS_EXIT(zfsvfs);
4129 			nvlist_free(nvlist);
4130 			return (error);
4131 		}
4132 		error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
4133 		    kcred, NULL, 0);
4134 		nvlist_free(nvlist);
4135 		break;
4136 
4137 	case ZFS_SMB_ACL_PURGE:
4138 		error = zfs_smb_acl_purge(sharedir);
4139 		break;
4140 
4141 	default:
4142 		error = EINVAL;
4143 		break;
4144 	}
4145 
4146 	VN_RELE(vp);
4147 	VN_RELE(ZTOV(sharedir));
4148 
4149 	ZFS_EXIT(zfsvfs);
4150 
4151 	return (error);
4152 }
4153 
4154 /*
4155  * inputs:
4156  * zc_name	name of filesystem
4157  * zc_value	short name of snap
4158  * zc_string	user-supplied tag for this reference
4159  * zc_cookie	recursive flag
4160  * zc_temphold	set if hold is temporary
4161  *
4162  * outputs:		none
4163  */
4164 static int
4165 zfs_ioc_hold(zfs_cmd_t *zc)
4166 {
4167 	boolean_t recursive = zc->zc_cookie;
4168 
4169 	if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
4170 		return (EINVAL);
4171 
4172 	return (dsl_dataset_user_hold(zc->zc_name, zc->zc_value,
4173 	    zc->zc_string, recursive, zc->zc_temphold));
4174 }
4175 
4176 /*
4177  * inputs:
4178  * zc_name	name of dataset from which we're releasing a user reference
4179  * zc_value	short name of snap
4180  * zc_string	user-supplied tag for this reference
4181  * zc_cookie	recursive flag
4182  *
4183  * outputs:		none
4184  */
4185 static int
4186 zfs_ioc_release(zfs_cmd_t *zc)
4187 {
4188 	boolean_t recursive = zc->zc_cookie;
4189 
4190 	if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
4191 		return (EINVAL);
4192 
4193 	return (dsl_dataset_user_release(zc->zc_name, zc->zc_value,
4194 	    zc->zc_string, recursive));
4195 }
4196 
4197 /*
4198  * inputs:
4199  * zc_name		name of filesystem
4200  *
4201  * outputs:
4202  * zc_nvlist_src{_size}	nvlist of snapshot holds
4203  */
4204 static int
4205 zfs_ioc_get_holds(zfs_cmd_t *zc)
4206 {
4207 	nvlist_t *nvp;
4208 	int error;
4209 
4210 	if ((error = dsl_dataset_get_holds(zc->zc_name, &nvp)) == 0) {
4211 		error = put_nvlist(zc, nvp);
4212 		nvlist_free(nvp);
4213 	}
4214 
4215 	return (error);
4216 }
4217 
4218 /*
4219  * pool create, destroy, and export don't log the history as part of
4220  * zfsdev_ioctl, but rather zfs_ioc_pool_create, and zfs_ioc_pool_export
4221  * do the logging of those commands.
4222  */
4223 static zfs_ioc_vec_t zfs_ioc_vec[] = {
4224 	{ zfs_ioc_pool_create, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4225 	    B_FALSE },
4226 	{ zfs_ioc_pool_destroy,	zfs_secpolicy_config, POOL_NAME, B_FALSE,
4227 	    B_FALSE },
4228 	{ zfs_ioc_pool_import, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4229 	    B_FALSE },
4230 	{ zfs_ioc_pool_export, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4231 	    B_FALSE },
4232 	{ zfs_ioc_pool_configs,	zfs_secpolicy_none, NO_NAME, B_FALSE,
4233 	    B_FALSE },
4234 	{ zfs_ioc_pool_stats, zfs_secpolicy_read, POOL_NAME, B_FALSE,
4235 	    B_FALSE },
4236 	{ zfs_ioc_pool_tryimport, zfs_secpolicy_config, NO_NAME, B_FALSE,
4237 	    B_FALSE },
4238 	{ zfs_ioc_pool_scrub, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4239 	    B_TRUE },
4240 	{ zfs_ioc_pool_freeze, zfs_secpolicy_config, NO_NAME, B_FALSE,
4241 	    B_FALSE },
4242 	{ zfs_ioc_pool_upgrade,	zfs_secpolicy_config, POOL_NAME, B_TRUE,
4243 	    B_TRUE },
4244 	{ zfs_ioc_pool_get_history, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4245 	    B_FALSE },
4246 	{ zfs_ioc_vdev_add, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4247 	    B_TRUE },
4248 	{ zfs_ioc_vdev_remove, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4249 	    B_TRUE },
4250 	{ zfs_ioc_vdev_set_state, zfs_secpolicy_config,	POOL_NAME, B_TRUE,
4251 	    B_FALSE },
4252 	{ zfs_ioc_vdev_attach, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4253 	    B_TRUE },
4254 	{ zfs_ioc_vdev_detach, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4255 	    B_TRUE },
4256 	{ zfs_ioc_vdev_setpath,	zfs_secpolicy_config, POOL_NAME, B_FALSE,
4257 	    B_TRUE },
4258 	{ zfs_ioc_vdev_setfru,	zfs_secpolicy_config, POOL_NAME, B_FALSE,
4259 	    B_TRUE },
4260 	{ zfs_ioc_objset_stats,	zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4261 	    B_TRUE },
4262 	{ zfs_ioc_objset_zplprops, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4263 	    B_FALSE },
4264 	{ zfs_ioc_dataset_list_next, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4265 	    B_TRUE },
4266 	{ zfs_ioc_snapshot_list_next, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4267 	    B_TRUE },
4268 	{ zfs_ioc_set_prop, zfs_secpolicy_none, DATASET_NAME, B_TRUE, B_TRUE },
4269 	{ zfs_ioc_create, zfs_secpolicy_create, DATASET_NAME, B_TRUE, B_TRUE },
4270 	{ zfs_ioc_destroy, zfs_secpolicy_destroy, DATASET_NAME, B_TRUE,
4271 	    B_TRUE},
4272 	{ zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, B_TRUE,
4273 	    B_TRUE },
4274 	{ zfs_ioc_rename, zfs_secpolicy_rename,	DATASET_NAME, B_TRUE, B_TRUE },
4275 	{ zfs_ioc_recv, zfs_secpolicy_receive, DATASET_NAME, B_TRUE, B_TRUE },
4276 	{ zfs_ioc_send, zfs_secpolicy_send, DATASET_NAME, B_TRUE, B_FALSE },
4277 	{ zfs_ioc_inject_fault,	zfs_secpolicy_inject, NO_NAME, B_FALSE,
4278 	    B_FALSE },
4279 	{ zfs_ioc_clear_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE,
4280 	    B_FALSE },
4281 	{ zfs_ioc_inject_list_next, zfs_secpolicy_inject, NO_NAME, B_FALSE,
4282 	    B_FALSE },
4283 	{ zfs_ioc_error_log, zfs_secpolicy_inject, POOL_NAME, B_FALSE,
4284 	    B_FALSE },
4285 	{ zfs_ioc_clear, zfs_secpolicy_config, POOL_NAME, B_TRUE, B_FALSE },
4286 	{ zfs_ioc_promote, zfs_secpolicy_promote, DATASET_NAME, B_TRUE,
4287 	    B_TRUE },
4288 	{ zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, DATASET_NAME,
4289 	    B_TRUE, B_TRUE },
4290 	{ zfs_ioc_snapshot, zfs_secpolicy_snapshot, DATASET_NAME, B_TRUE,
4291 	    B_TRUE },
4292 	{ zfs_ioc_dsobj_to_dsname, zfs_secpolicy_config, POOL_NAME, B_FALSE,
4293 	    B_FALSE },
4294 	{ zfs_ioc_obj_to_path, zfs_secpolicy_config, DATASET_NAME, B_FALSE,
4295 	    B_TRUE },
4296 	{ zfs_ioc_pool_set_props, zfs_secpolicy_config,	POOL_NAME, B_TRUE,
4297 	    B_TRUE },
4298 	{ zfs_ioc_pool_get_props, zfs_secpolicy_read, POOL_NAME, B_FALSE,
4299 	    B_FALSE },
4300 	{ zfs_ioc_set_fsacl, zfs_secpolicy_fsacl, DATASET_NAME, B_TRUE,
4301 	    B_TRUE },
4302 	{ zfs_ioc_get_fsacl, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4303 	    B_FALSE },
4304 	{ zfs_ioc_share, zfs_secpolicy_share, DATASET_NAME, B_FALSE, B_FALSE },
4305 	{ zfs_ioc_inherit_prop, zfs_secpolicy_inherit, DATASET_NAME, B_TRUE,
4306 	    B_TRUE },
4307 	{ zfs_ioc_smb_acl, zfs_secpolicy_smb_acl, DATASET_NAME, B_FALSE,
4308 	    B_FALSE },
4309 	{ zfs_ioc_userspace_one, zfs_secpolicy_userspace_one,
4310 	    DATASET_NAME, B_FALSE, B_FALSE },
4311 	{ zfs_ioc_userspace_many, zfs_secpolicy_userspace_many,
4312 	    DATASET_NAME, B_FALSE, B_FALSE },
4313 	{ zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
4314 	    DATASET_NAME, B_FALSE, B_TRUE },
4315 	{ zfs_ioc_hold, zfs_secpolicy_hold, DATASET_NAME, B_TRUE, B_TRUE },
4316 	{ zfs_ioc_release, zfs_secpolicy_release, DATASET_NAME, B_TRUE,
4317 	    B_TRUE },
4318 	{ zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4319 	    B_TRUE },
4320 	{ zfs_ioc_objset_recvd_props, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
4321 	    B_FALSE },
4322 	{ zfs_ioc_vdev_split, zfs_secpolicy_config, POOL_NAME, B_TRUE,
4323 	    B_TRUE }
4324 };
4325 
4326 int
4327 pool_status_check(const char *name, zfs_ioc_namecheck_t type)
4328 {
4329 	spa_t *spa;
4330 	int error;
4331 
4332 	ASSERT(type == POOL_NAME || type == DATASET_NAME);
4333 
4334 	error = spa_open(name, &spa, FTAG);
4335 	if (error == 0) {
4336 		if (spa_suspended(spa))
4337 			error = EAGAIN;
4338 		spa_close(spa, FTAG);
4339 	}
4340 	return (error);
4341 }
4342 
4343 static int
4344 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
4345 {
4346 	zfs_cmd_t *zc;
4347 	uint_t vec;
4348 	int error, rc;
4349 
4350 	if (getminor(dev) != 0)
4351 		return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
4352 
4353 	vec = cmd - ZFS_IOC;
4354 	ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
4355 
4356 	if (vec >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
4357 		return (EINVAL);
4358 
4359 	zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
4360 
4361 	error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
4362 	if (error != 0)
4363 		error = EFAULT;
4364 
4365 	if ((error == 0) && !(flag & FKIOCTL))
4366 		error = zfs_ioc_vec[vec].zvec_secpolicy(zc, cr);
4367 
4368 	/*
4369 	 * Ensure that all pool/dataset names are valid before we pass down to
4370 	 * the lower layers.
4371 	 */
4372 	if (error == 0) {
4373 		zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
4374 		zc->zc_iflags = flag & FKIOCTL;
4375 		switch (zfs_ioc_vec[vec].zvec_namecheck) {
4376 		case POOL_NAME:
4377 			if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
4378 				error = EINVAL;
4379 			if (zfs_ioc_vec[vec].zvec_pool_check)
4380 				error = pool_status_check(zc->zc_name,
4381 				    zfs_ioc_vec[vec].zvec_namecheck);
4382 			break;
4383 
4384 		case DATASET_NAME:
4385 			if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
4386 				error = EINVAL;
4387 			if (zfs_ioc_vec[vec].zvec_pool_check)
4388 				error = pool_status_check(zc->zc_name,
4389 				    zfs_ioc_vec[vec].zvec_namecheck);
4390 			break;
4391 
4392 		case NO_NAME:
4393 			break;
4394 		}
4395 	}
4396 
4397 	if (error == 0)
4398 		error = zfs_ioc_vec[vec].zvec_func(zc);
4399 
4400 	rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
4401 	if (error == 0) {
4402 		if (rc != 0)
4403 			error = EFAULT;
4404 		if (zfs_ioc_vec[vec].zvec_his_log)
4405 			zfs_log_history(zc);
4406 	}
4407 
4408 	kmem_free(zc, sizeof (zfs_cmd_t));
4409 	return (error);
4410 }
4411 
4412 static int
4413 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
4414 {
4415 	if (cmd != DDI_ATTACH)
4416 		return (DDI_FAILURE);
4417 
4418 	if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
4419 	    DDI_PSEUDO, 0) == DDI_FAILURE)
4420 		return (DDI_FAILURE);
4421 
4422 	zfs_dip = dip;
4423 
4424 	ddi_report_dev(dip);
4425 
4426 	return (DDI_SUCCESS);
4427 }
4428 
4429 static int
4430 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
4431 {
4432 	if (spa_busy() || zfs_busy() || zvol_busy())
4433 		return (DDI_FAILURE);
4434 
4435 	if (cmd != DDI_DETACH)
4436 		return (DDI_FAILURE);
4437 
4438 	zfs_dip = NULL;
4439 
4440 	ddi_prop_remove_all(dip);
4441 	ddi_remove_minor_node(dip, NULL);
4442 
4443 	return (DDI_SUCCESS);
4444 }
4445 
4446 /*ARGSUSED*/
4447 static int
4448 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
4449 {
4450 	switch (infocmd) {
4451 	case DDI_INFO_DEVT2DEVINFO:
4452 		*result = zfs_dip;
4453 		return (DDI_SUCCESS);
4454 
4455 	case DDI_INFO_DEVT2INSTANCE:
4456 		*result = (void *)0;
4457 		return (DDI_SUCCESS);
4458 	}
4459 
4460 	return (DDI_FAILURE);
4461 }
4462 
4463 /*
4464  * OK, so this is a little weird.
4465  *
4466  * /dev/zfs is the control node, i.e. minor 0.
4467  * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
4468  *
4469  * /dev/zfs has basically nothing to do except serve up ioctls,
4470  * so most of the standard driver entry points are in zvol.c.
4471  */
4472 static struct cb_ops zfs_cb_ops = {
4473 	zvol_open,	/* open */
4474 	zvol_close,	/* close */
4475 	zvol_strategy,	/* strategy */
4476 	nodev,		/* print */
4477 	zvol_dump,	/* dump */
4478 	zvol_read,	/* read */
4479 	zvol_write,	/* write */
4480 	zfsdev_ioctl,	/* ioctl */
4481 	nodev,		/* devmap */
4482 	nodev,		/* mmap */
4483 	nodev,		/* segmap */
4484 	nochpoll,	/* poll */
4485 	ddi_prop_op,	/* prop_op */
4486 	NULL,		/* streamtab */
4487 	D_NEW | D_MP | D_64BIT,		/* Driver compatibility flag */
4488 	CB_REV,		/* version */
4489 	nodev,		/* async read */
4490 	nodev,		/* async write */
4491 };
4492 
4493 static struct dev_ops zfs_dev_ops = {
4494 	DEVO_REV,	/* version */
4495 	0,		/* refcnt */
4496 	zfs_info,	/* info */
4497 	nulldev,	/* identify */
4498 	nulldev,	/* probe */
4499 	zfs_attach,	/* attach */
4500 	zfs_detach,	/* detach */
4501 	nodev,		/* reset */
4502 	&zfs_cb_ops,	/* driver operations */
4503 	NULL,		/* no bus operations */
4504 	NULL,		/* power */
4505 	ddi_quiesce_not_needed,	/* quiesce */
4506 };
4507 
4508 static struct modldrv zfs_modldrv = {
4509 	&mod_driverops,
4510 	"ZFS storage pool",
4511 	&zfs_dev_ops
4512 };
4513 
4514 static struct modlinkage modlinkage = {
4515 	MODREV_1,
4516 	(void *)&zfs_modlfs,
4517 	(void *)&zfs_modldrv,
4518 	NULL
4519 };
4520 
4521 
4522 uint_t zfs_fsyncer_key;
4523 extern uint_t rrw_tsd_key;
4524 
4525 int
4526 _init(void)
4527 {
4528 	int error;
4529 
4530 	spa_init(FREAD | FWRITE);
4531 	zfs_init();
4532 	zvol_init();
4533 
4534 	if ((error = mod_install(&modlinkage)) != 0) {
4535 		zvol_fini();
4536 		zfs_fini();
4537 		spa_fini();
4538 		return (error);
4539 	}
4540 
4541 	tsd_create(&zfs_fsyncer_key, NULL);
4542 	tsd_create(&rrw_tsd_key, NULL);
4543 
4544 	error = ldi_ident_from_mod(&modlinkage, &zfs_li);
4545 	ASSERT(error == 0);
4546 	mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
4547 
4548 	return (0);
4549 }
4550 
4551 int
4552 _fini(void)
4553 {
4554 	int error;
4555 
4556 	if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
4557 		return (EBUSY);
4558 
4559 	if ((error = mod_remove(&modlinkage)) != 0)
4560 		return (error);
4561 
4562 	zvol_fini();
4563 	zfs_fini();
4564 	spa_fini();
4565 	if (zfs_nfsshare_inited)
4566 		(void) ddi_modclose(nfs_mod);
4567 	if (zfs_smbshare_inited)
4568 		(void) ddi_modclose(smbsrv_mod);
4569 	if (zfs_nfsshare_inited || zfs_smbshare_inited)
4570 		(void) ddi_modclose(sharefs_mod);
4571 
4572 	tsd_destroy(&zfs_fsyncer_key);
4573 	ldi_ident_release(zfs_li);
4574 	zfs_li = NULL;
4575 	mutex_destroy(&zfs_share_lock);
4576 
4577 	return (error);
4578 }
4579 
4580 int
4581 _info(struct modinfo *modinfop)
4582 {
4583 	return (mod_info(&modlinkage, modinfop));
4584 }
4585