xref: /titanic_41/usr/src/uts/common/fs/zfs/zfs_ioctl.c (revision 299d957d28501350bcae7fa4333ff183c062c427)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 /*
22  * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 #include <sys/types.h>
27 #include <sys/param.h>
28 #include <sys/errno.h>
29 #include <sys/uio.h>
30 #include <sys/buf.h>
31 #include <sys/modctl.h>
32 #include <sys/open.h>
33 #include <sys/file.h>
34 #include <sys/kmem.h>
35 #include <sys/conf.h>
36 #include <sys/cmn_err.h>
37 #include <sys/stat.h>
38 #include <sys/zfs_ioctl.h>
39 #include <sys/zfs_znode.h>
40 #include <sys/zap.h>
41 #include <sys/spa.h>
42 #include <sys/spa_impl.h>
43 #include <sys/vdev.h>
44 #include <sys/vdev_impl.h>
45 #include <sys/dmu.h>
46 #include <sys/dsl_dir.h>
47 #include <sys/dsl_dataset.h>
48 #include <sys/dsl_prop.h>
49 #include <sys/dsl_deleg.h>
50 #include <sys/dmu_objset.h>
51 #include <sys/ddi.h>
52 #include <sys/sunddi.h>
53 #include <sys/sunldi.h>
54 #include <sys/policy.h>
55 #include <sys/zone.h>
56 #include <sys/nvpair.h>
57 #include <sys/pathname.h>
58 #include <sys/mount.h>
59 #include <sys/sdt.h>
60 #include <sys/fs/zfs.h>
61 #include <sys/zfs_ctldir.h>
62 #include <sys/zfs_dir.h>
63 #include <sys/zvol.h>
64 #include <sharefs/share.h>
65 #include <sys/dmu_objset.h>
66 
67 #include "zfs_namecheck.h"
68 #include "zfs_prop.h"
69 #include "zfs_deleg.h"
70 
71 extern struct modlfs zfs_modlfs;
72 
73 extern void zfs_init(void);
74 extern void zfs_fini(void);
75 
76 ldi_ident_t zfs_li = NULL;
77 dev_info_t *zfs_dip;
78 
79 typedef int zfs_ioc_func_t(zfs_cmd_t *);
80 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, cred_t *);
81 
82 typedef struct zfs_ioc_vec {
83 	zfs_ioc_func_t		*zvec_func;
84 	zfs_secpolicy_func_t	*zvec_secpolicy;
85 	enum {
86 		NO_NAME,
87 		POOL_NAME,
88 		DATASET_NAME
89 	} zvec_namecheck;
90 	boolean_t		zvec_his_log;
91 } zfs_ioc_vec_t;
92 
93 static void clear_props(char *dataset, nvlist_t *props, nvlist_t *newprops);
94 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
95     boolean_t *);
96 int zfs_set_prop_nvlist(const char *, nvlist_t *);
97 
98 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
99 void
100 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
101 {
102 	const char *newfile;
103 	char buf[256];
104 	va_list adx;
105 
106 	/*
107 	 * Get rid of annoying "../common/" prefix to filename.
108 	 */
109 	newfile = strrchr(file, '/');
110 	if (newfile != NULL) {
111 		newfile = newfile + 1; /* Get rid of leading / */
112 	} else {
113 		newfile = file;
114 	}
115 
116 	va_start(adx, fmt);
117 	(void) vsnprintf(buf, sizeof (buf), fmt, adx);
118 	va_end(adx);
119 
120 	/*
121 	 * To get this data, use the zfs-dprintf probe as so:
122 	 * dtrace -q -n 'zfs-dprintf \
123 	 *	/stringof(arg0) == "dbuf.c"/ \
124 	 *	{printf("%s: %s", stringof(arg1), stringof(arg3))}'
125 	 * arg0 = file name
126 	 * arg1 = function name
127 	 * arg2 = line number
128 	 * arg3 = message
129 	 */
130 	DTRACE_PROBE4(zfs__dprintf,
131 	    char *, newfile, char *, func, int, line, char *, buf);
132 }
133 
134 static void
135 history_str_free(char *buf)
136 {
137 	kmem_free(buf, HIS_MAX_RECORD_LEN);
138 }
139 
140 static char *
141 history_str_get(zfs_cmd_t *zc)
142 {
143 	char *buf;
144 
145 	if (zc->zc_history == NULL)
146 		return (NULL);
147 
148 	buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
149 	if (copyinstr((void *)(uintptr_t)zc->zc_history,
150 	    buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
151 		history_str_free(buf);
152 		return (NULL);
153 	}
154 
155 	buf[HIS_MAX_RECORD_LEN -1] = '\0';
156 
157 	return (buf);
158 }
159 
160 /*
161  * Check to see if the named dataset is currently defined as bootable
162  */
163 static boolean_t
164 zfs_is_bootfs(const char *name)
165 {
166 	spa_t *spa;
167 	boolean_t ret = B_FALSE;
168 
169 	if (spa_open(name, &spa, FTAG) == 0) {
170 		if (spa->spa_bootfs) {
171 			objset_t *os;
172 
173 			if (dmu_objset_open(name, DMU_OST_ZFS,
174 			    DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
175 				ret = (dmu_objset_id(os) == spa->spa_bootfs);
176 				dmu_objset_close(os);
177 			}
178 		}
179 		spa_close(spa, FTAG);
180 	}
181 	return (ret);
182 }
183 
184 /*
185  * zfs_earlier_version
186  *
187  *	Return non-zero if the spa version is less than requested version.
188  */
189 static int
190 zfs_earlier_version(const char *name, int version)
191 {
192 	spa_t *spa;
193 
194 	if (spa_open(name, &spa, FTAG) == 0) {
195 		if (spa_version(spa) < version) {
196 			spa_close(spa, FTAG);
197 			return (1);
198 		}
199 		spa_close(spa, FTAG);
200 	}
201 	return (0);
202 }
203 
204 /*
205  * zpl_earlier_version
206  *
207  * Return TRUE if the ZPL version is less than requested version.
208  */
209 static boolean_t
210 zpl_earlier_version(const char *name, int version)
211 {
212 	objset_t *os;
213 	boolean_t rc = B_TRUE;
214 
215 	if (dmu_objset_open(name, DMU_OST_ANY,
216 	    DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
217 		uint64_t zplversion;
218 
219 		if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
220 			rc = zplversion < version;
221 		dmu_objset_close(os);
222 	}
223 	return (rc);
224 }
225 
226 static void
227 zfs_log_history(zfs_cmd_t *zc)
228 {
229 	spa_t *spa;
230 	char *buf;
231 
232 	if ((buf = history_str_get(zc)) == NULL)
233 		return;
234 
235 	if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
236 		if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
237 			(void) spa_history_log(spa, buf, LOG_CMD_NORMAL);
238 		spa_close(spa, FTAG);
239 	}
240 	history_str_free(buf);
241 }
242 
243 /*
244  * Policy for top-level read operations (list pools).  Requires no privileges,
245  * and can be used in the local zone, as there is no associated dataset.
246  */
247 /* ARGSUSED */
248 static int
249 zfs_secpolicy_none(zfs_cmd_t *zc, cred_t *cr)
250 {
251 	return (0);
252 }
253 
254 /*
255  * Policy for dataset read operations (list children, get statistics).  Requires
256  * no privileges, but must be visible in the local zone.
257  */
258 /* ARGSUSED */
259 static int
260 zfs_secpolicy_read(zfs_cmd_t *zc, cred_t *cr)
261 {
262 	if (INGLOBALZONE(curproc) ||
263 	    zone_dataset_visible(zc->zc_name, NULL))
264 		return (0);
265 
266 	return (ENOENT);
267 }
268 
269 static int
270 zfs_dozonecheck(const char *dataset, cred_t *cr)
271 {
272 	uint64_t zoned;
273 	int writable = 1;
274 
275 	/*
276 	 * The dataset must be visible by this zone -- check this first
277 	 * so they don't see EPERM on something they shouldn't know about.
278 	 */
279 	if (!INGLOBALZONE(curproc) &&
280 	    !zone_dataset_visible(dataset, &writable))
281 		return (ENOENT);
282 
283 	if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
284 		return (ENOENT);
285 
286 	if (INGLOBALZONE(curproc)) {
287 		/*
288 		 * If the fs is zoned, only root can access it from the
289 		 * global zone.
290 		 */
291 		if (secpolicy_zfs(cr) && zoned)
292 			return (EPERM);
293 	} else {
294 		/*
295 		 * If we are in a local zone, the 'zoned' property must be set.
296 		 */
297 		if (!zoned)
298 			return (EPERM);
299 
300 		/* must be writable by this zone */
301 		if (!writable)
302 			return (EPERM);
303 	}
304 	return (0);
305 }
306 
307 int
308 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
309 {
310 	int error;
311 
312 	error = zfs_dozonecheck(name, cr);
313 	if (error == 0) {
314 		error = secpolicy_zfs(cr);
315 		if (error)
316 			error = dsl_deleg_access(name, perm, cr);
317 	}
318 	return (error);
319 }
320 
321 static int
322 zfs_secpolicy_setprop(const char *name, zfs_prop_t prop, cred_t *cr)
323 {
324 	/*
325 	 * Check permissions for special properties.
326 	 */
327 	switch (prop) {
328 	case ZFS_PROP_ZONED:
329 		/*
330 		 * Disallow setting of 'zoned' from within a local zone.
331 		 */
332 		if (!INGLOBALZONE(curproc))
333 			return (EPERM);
334 		break;
335 
336 	case ZFS_PROP_QUOTA:
337 		if (!INGLOBALZONE(curproc)) {
338 			uint64_t zoned;
339 			char setpoint[MAXNAMELEN];
340 			/*
341 			 * Unprivileged users are allowed to modify the
342 			 * quota on things *under* (ie. contained by)
343 			 * the thing they own.
344 			 */
345 			if (dsl_prop_get_integer(name, "zoned", &zoned,
346 			    setpoint))
347 				return (EPERM);
348 			if (!zoned || strlen(name) <= strlen(setpoint))
349 				return (EPERM);
350 		}
351 		break;
352 	}
353 
354 	return (zfs_secpolicy_write_perms(name, zfs_prop_to_name(prop), cr));
355 }
356 
357 int
358 zfs_secpolicy_fsacl(zfs_cmd_t *zc, cred_t *cr)
359 {
360 	int error;
361 
362 	error = zfs_dozonecheck(zc->zc_name, cr);
363 	if (error)
364 		return (error);
365 
366 	/*
367 	 * permission to set permissions will be evaluated later in
368 	 * dsl_deleg_can_allow()
369 	 */
370 	return (0);
371 }
372 
373 int
374 zfs_secpolicy_rollback(zfs_cmd_t *zc, cred_t *cr)
375 {
376 	int error;
377 	error = zfs_secpolicy_write_perms(zc->zc_name,
378 	    ZFS_DELEG_PERM_ROLLBACK, cr);
379 	if (error == 0)
380 		error = zfs_secpolicy_write_perms(zc->zc_name,
381 		    ZFS_DELEG_PERM_MOUNT, cr);
382 	return (error);
383 }
384 
385 int
386 zfs_secpolicy_send(zfs_cmd_t *zc, cred_t *cr)
387 {
388 	return (zfs_secpolicy_write_perms(zc->zc_name,
389 	    ZFS_DELEG_PERM_SEND, cr));
390 }
391 
392 int
393 zfs_secpolicy_share(zfs_cmd_t *zc, cred_t *cr)
394 {
395 	if (!INGLOBALZONE(curproc))
396 		return (EPERM);
397 
398 	if (secpolicy_nfs(cr) == 0) {
399 		return (0);
400 	} else {
401 		vnode_t *vp;
402 		int error;
403 
404 		if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
405 		    NO_FOLLOW, NULL, &vp)) != 0)
406 			return (error);
407 
408 		/* Now make sure mntpnt and dataset are ZFS */
409 
410 		if (vp->v_vfsp->vfs_fstype != zfsfstype ||
411 		    (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
412 		    zc->zc_name) != 0)) {
413 			VN_RELE(vp);
414 			return (EPERM);
415 		}
416 
417 		VN_RELE(vp);
418 		return (dsl_deleg_access(zc->zc_name,
419 		    ZFS_DELEG_PERM_SHARE, cr));
420 	}
421 }
422 
423 static int
424 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
425 {
426 	char *cp;
427 
428 	/*
429 	 * Remove the @bla or /bla from the end of the name to get the parent.
430 	 */
431 	(void) strncpy(parent, datasetname, parentsize);
432 	cp = strrchr(parent, '@');
433 	if (cp != NULL) {
434 		cp[0] = '\0';
435 	} else {
436 		cp = strrchr(parent, '/');
437 		if (cp == NULL)
438 			return (ENOENT);
439 		cp[0] = '\0';
440 	}
441 
442 	return (0);
443 }
444 
445 int
446 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
447 {
448 	int error;
449 
450 	if ((error = zfs_secpolicy_write_perms(name,
451 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
452 		return (error);
453 
454 	return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
455 }
456 
457 static int
458 zfs_secpolicy_destroy(zfs_cmd_t *zc, cred_t *cr)
459 {
460 	return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
461 }
462 
463 /*
464  * Must have sys_config privilege to check the iscsi permission
465  */
466 /* ARGSUSED */
467 static int
468 zfs_secpolicy_iscsi(zfs_cmd_t *zc, cred_t *cr)
469 {
470 	return (secpolicy_zfs(cr));
471 }
472 
473 int
474 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
475 {
476 	char 	parentname[MAXNAMELEN];
477 	int	error;
478 
479 	if ((error = zfs_secpolicy_write_perms(from,
480 	    ZFS_DELEG_PERM_RENAME, cr)) != 0)
481 		return (error);
482 
483 	if ((error = zfs_secpolicy_write_perms(from,
484 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
485 		return (error);
486 
487 	if ((error = zfs_get_parent(to, parentname,
488 	    sizeof (parentname))) != 0)
489 		return (error);
490 
491 	if ((error = zfs_secpolicy_write_perms(parentname,
492 	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
493 		return (error);
494 
495 	if ((error = zfs_secpolicy_write_perms(parentname,
496 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
497 		return (error);
498 
499 	return (error);
500 }
501 
502 static int
503 zfs_secpolicy_rename(zfs_cmd_t *zc, cred_t *cr)
504 {
505 	return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
506 }
507 
508 static int
509 zfs_secpolicy_promote(zfs_cmd_t *zc, cred_t *cr)
510 {
511 	char 	parentname[MAXNAMELEN];
512 	objset_t *clone;
513 	int error;
514 
515 	error = zfs_secpolicy_write_perms(zc->zc_name,
516 	    ZFS_DELEG_PERM_PROMOTE, cr);
517 	if (error)
518 		return (error);
519 
520 	error = dmu_objset_open(zc->zc_name, DMU_OST_ANY,
521 	    DS_MODE_USER | DS_MODE_READONLY, &clone);
522 
523 	if (error == 0) {
524 		dsl_dataset_t *pclone = NULL;
525 		dsl_dir_t *dd;
526 		dd = clone->os->os_dsl_dataset->ds_dir;
527 
528 		rw_enter(&dd->dd_pool->dp_config_rwlock, RW_READER);
529 		error = dsl_dataset_hold_obj(dd->dd_pool,
530 		    dd->dd_phys->dd_origin_obj, FTAG, &pclone);
531 		rw_exit(&dd->dd_pool->dp_config_rwlock);
532 		if (error) {
533 			dmu_objset_close(clone);
534 			return (error);
535 		}
536 
537 		error = zfs_secpolicy_write_perms(zc->zc_name,
538 		    ZFS_DELEG_PERM_MOUNT, cr);
539 
540 		dsl_dataset_name(pclone, parentname);
541 		dmu_objset_close(clone);
542 		dsl_dataset_rele(pclone, FTAG);
543 		if (error == 0)
544 			error = zfs_secpolicy_write_perms(parentname,
545 			    ZFS_DELEG_PERM_PROMOTE, cr);
546 	}
547 	return (error);
548 }
549 
550 static int
551 zfs_secpolicy_receive(zfs_cmd_t *zc, cred_t *cr)
552 {
553 	int error;
554 
555 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
556 	    ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
557 		return (error);
558 
559 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
560 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
561 		return (error);
562 
563 	return (zfs_secpolicy_write_perms(zc->zc_name,
564 	    ZFS_DELEG_PERM_CREATE, cr));
565 }
566 
567 int
568 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
569 {
570 	int error;
571 
572 	if ((error = zfs_secpolicy_write_perms(name,
573 	    ZFS_DELEG_PERM_SNAPSHOT, cr)) != 0)
574 		return (error);
575 
576 	error = zfs_secpolicy_write_perms(name,
577 	    ZFS_DELEG_PERM_MOUNT, cr);
578 
579 	return (error);
580 }
581 
582 static int
583 zfs_secpolicy_snapshot(zfs_cmd_t *zc, cred_t *cr)
584 {
585 
586 	return (zfs_secpolicy_snapshot_perms(zc->zc_name, cr));
587 }
588 
589 static int
590 zfs_secpolicy_create(zfs_cmd_t *zc, cred_t *cr)
591 {
592 	char 	parentname[MAXNAMELEN];
593 	int 	error;
594 
595 	if ((error = zfs_get_parent(zc->zc_name, parentname,
596 	    sizeof (parentname))) != 0)
597 		return (error);
598 
599 	if (zc->zc_value[0] != '\0') {
600 		if ((error = zfs_secpolicy_write_perms(zc->zc_value,
601 		    ZFS_DELEG_PERM_CLONE, cr)) != 0)
602 			return (error);
603 	}
604 
605 	if ((error = zfs_secpolicy_write_perms(parentname,
606 	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
607 		return (error);
608 
609 	error = zfs_secpolicy_write_perms(parentname,
610 	    ZFS_DELEG_PERM_MOUNT, cr);
611 
612 	return (error);
613 }
614 
615 static int
616 zfs_secpolicy_umount(zfs_cmd_t *zc, cred_t *cr)
617 {
618 	int error;
619 
620 	error = secpolicy_fs_unmount(cr, NULL);
621 	if (error) {
622 		error = dsl_deleg_access(zc->zc_name, ZFS_DELEG_PERM_MOUNT, cr);
623 	}
624 	return (error);
625 }
626 
627 /*
628  * Policy for pool operations - create/destroy pools, add vdevs, etc.  Requires
629  * SYS_CONFIG privilege, which is not available in a local zone.
630  */
631 /* ARGSUSED */
632 static int
633 zfs_secpolicy_config(zfs_cmd_t *zc, cred_t *cr)
634 {
635 	if (secpolicy_sys_config(cr, B_FALSE) != 0)
636 		return (EPERM);
637 
638 	return (0);
639 }
640 
641 /*
642  * Just like zfs_secpolicy_config, except that we will check for
643  * mount permission on the dataset for permission to create/remove
644  * the minor nodes.
645  */
646 static int
647 zfs_secpolicy_minor(zfs_cmd_t *zc, cred_t *cr)
648 {
649 	if (secpolicy_sys_config(cr, B_FALSE) != 0) {
650 		return (dsl_deleg_access(zc->zc_name,
651 		    ZFS_DELEG_PERM_MOUNT, cr));
652 	}
653 
654 	return (0);
655 }
656 
657 /*
658  * Policy for fault injection.  Requires all privileges.
659  */
660 /* ARGSUSED */
661 static int
662 zfs_secpolicy_inject(zfs_cmd_t *zc, cred_t *cr)
663 {
664 	return (secpolicy_zinject(cr));
665 }
666 
667 static int
668 zfs_secpolicy_inherit(zfs_cmd_t *zc, cred_t *cr)
669 {
670 	zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
671 
672 	if (prop == ZPROP_INVAL) {
673 		if (!zfs_prop_user(zc->zc_value))
674 			return (EINVAL);
675 		return (zfs_secpolicy_write_perms(zc->zc_name,
676 		    ZFS_DELEG_PERM_USERPROP, cr));
677 	} else {
678 		if (!zfs_prop_inheritable(prop))
679 			return (EINVAL);
680 		return (zfs_secpolicy_setprop(zc->zc_name, prop, cr));
681 	}
682 }
683 
684 /*
685  * Returns the nvlist as specified by the user in the zfs_cmd_t.
686  */
687 static int
688 get_nvlist(uint64_t nvl, uint64_t size, nvlist_t **nvp)
689 {
690 	char *packed;
691 	int error;
692 	nvlist_t *list = NULL;
693 
694 	/*
695 	 * Read in and unpack the user-supplied nvlist.
696 	 */
697 	if (size == 0)
698 		return (EINVAL);
699 
700 	packed = kmem_alloc(size, KM_SLEEP);
701 
702 	if ((error = xcopyin((void *)(uintptr_t)nvl, packed, size)) != 0) {
703 		kmem_free(packed, size);
704 		return (error);
705 	}
706 
707 	if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
708 		kmem_free(packed, size);
709 		return (error);
710 	}
711 
712 	kmem_free(packed, size);
713 
714 	*nvp = list;
715 	return (0);
716 }
717 
718 static int
719 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
720 {
721 	char *packed = NULL;
722 	size_t size;
723 	int error;
724 
725 	VERIFY(nvlist_size(nvl, &size, NV_ENCODE_NATIVE) == 0);
726 
727 	if (size > zc->zc_nvlist_dst_size) {
728 		error = ENOMEM;
729 	} else {
730 		packed = kmem_alloc(size, KM_SLEEP);
731 		VERIFY(nvlist_pack(nvl, &packed, &size, NV_ENCODE_NATIVE,
732 		    KM_SLEEP) == 0);
733 		error = xcopyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
734 		    size);
735 		kmem_free(packed, size);
736 	}
737 
738 	zc->zc_nvlist_dst_size = size;
739 	return (error);
740 }
741 
742 static int
743 zfs_ioc_pool_create(zfs_cmd_t *zc)
744 {
745 	int error;
746 	nvlist_t *config, *props = NULL;
747 	nvlist_t *rootprops = NULL;
748 	nvlist_t *zplprops = NULL;
749 	char *buf;
750 
751 	if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
752 	    &config))
753 		return (error);
754 
755 	if (zc->zc_nvlist_src_size != 0 && (error =
756 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, &props))) {
757 		nvlist_free(config);
758 		return (error);
759 	}
760 
761 	if (props) {
762 		nvlist_t *nvl = NULL;
763 		uint64_t version = SPA_VERSION;
764 
765 		(void) nvlist_lookup_uint64(props,
766 		    zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
767 		if (version < SPA_VERSION_INITIAL || version > SPA_VERSION) {
768 			error = EINVAL;
769 			goto pool_props_bad;
770 		}
771 		(void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
772 		if (nvl) {
773 			error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
774 			if (error != 0) {
775 				nvlist_free(config);
776 				nvlist_free(props);
777 				return (error);
778 			}
779 			(void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
780 		}
781 		VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
782 		error = zfs_fill_zplprops_root(version, rootprops,
783 		    zplprops, NULL);
784 		if (error)
785 			goto pool_props_bad;
786 	}
787 
788 	buf = history_str_get(zc);
789 
790 	error = spa_create(zc->zc_name, config, props, buf, zplprops);
791 
792 	/*
793 	 * Set the remaining root properties
794 	 */
795 	if (!error &&
796 	    (error = zfs_set_prop_nvlist(zc->zc_name, rootprops)) != 0)
797 		(void) spa_destroy(zc->zc_name);
798 
799 	if (buf != NULL)
800 		history_str_free(buf);
801 
802 pool_props_bad:
803 	nvlist_free(rootprops);
804 	nvlist_free(zplprops);
805 	nvlist_free(config);
806 	nvlist_free(props);
807 
808 	return (error);
809 }
810 
811 static int
812 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
813 {
814 	int error;
815 	zfs_log_history(zc);
816 	error = spa_destroy(zc->zc_name);
817 	return (error);
818 }
819 
820 static int
821 zfs_ioc_pool_import(zfs_cmd_t *zc)
822 {
823 	int error;
824 	nvlist_t *config, *props = NULL;
825 	uint64_t guid;
826 
827 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
828 	    &config)) != 0)
829 		return (error);
830 
831 	if (zc->zc_nvlist_src_size != 0 && (error =
832 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, &props))) {
833 		nvlist_free(config);
834 		return (error);
835 	}
836 
837 	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
838 	    guid != zc->zc_guid)
839 		error = EINVAL;
840 	else if (zc->zc_cookie)
841 		error = spa_import_faulted(zc->zc_name, config,
842 		    props);
843 	else
844 		error = spa_import(zc->zc_name, config, props);
845 
846 	nvlist_free(config);
847 
848 	if (props)
849 		nvlist_free(props);
850 
851 	return (error);
852 }
853 
854 static int
855 zfs_ioc_pool_export(zfs_cmd_t *zc)
856 {
857 	int error;
858 	boolean_t force = (boolean_t)zc->zc_cookie;
859 	boolean_t hardforce = (boolean_t)zc->zc_guid;
860 
861 	zfs_log_history(zc);
862 	error = spa_export(zc->zc_name, NULL, force, hardforce);
863 	return (error);
864 }
865 
866 static int
867 zfs_ioc_pool_configs(zfs_cmd_t *zc)
868 {
869 	nvlist_t *configs;
870 	int error;
871 
872 	if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
873 		return (EEXIST);
874 
875 	error = put_nvlist(zc, configs);
876 
877 	nvlist_free(configs);
878 
879 	return (error);
880 }
881 
882 static int
883 zfs_ioc_pool_stats(zfs_cmd_t *zc)
884 {
885 	nvlist_t *config;
886 	int error;
887 	int ret = 0;
888 
889 	error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
890 	    sizeof (zc->zc_value));
891 
892 	if (config != NULL) {
893 		ret = put_nvlist(zc, config);
894 		nvlist_free(config);
895 
896 		/*
897 		 * The config may be present even if 'error' is non-zero.
898 		 * In this case we return success, and preserve the real errno
899 		 * in 'zc_cookie'.
900 		 */
901 		zc->zc_cookie = error;
902 	} else {
903 		ret = error;
904 	}
905 
906 	return (ret);
907 }
908 
909 /*
910  * Try to import the given pool, returning pool stats as appropriate so that
911  * user land knows which devices are available and overall pool health.
912  */
913 static int
914 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
915 {
916 	nvlist_t *tryconfig, *config;
917 	int error;
918 
919 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
920 	    &tryconfig)) != 0)
921 		return (error);
922 
923 	config = spa_tryimport(tryconfig);
924 
925 	nvlist_free(tryconfig);
926 
927 	if (config == NULL)
928 		return (EINVAL);
929 
930 	error = put_nvlist(zc, config);
931 	nvlist_free(config);
932 
933 	return (error);
934 }
935 
936 static int
937 zfs_ioc_pool_scrub(zfs_cmd_t *zc)
938 {
939 	spa_t *spa;
940 	int error;
941 
942 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
943 		return (error);
944 
945 	error = spa_scrub(spa, zc->zc_cookie);
946 
947 	spa_close(spa, FTAG);
948 
949 	return (error);
950 }
951 
952 static int
953 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
954 {
955 	spa_t *spa;
956 	int error;
957 
958 	error = spa_open(zc->zc_name, &spa, FTAG);
959 	if (error == 0) {
960 		spa_freeze(spa);
961 		spa_close(spa, FTAG);
962 	}
963 	return (error);
964 }
965 
966 static int
967 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
968 {
969 	spa_t *spa;
970 	int error;
971 
972 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
973 		return (error);
974 
975 	if (zc->zc_cookie < spa_version(spa) || zc->zc_cookie > SPA_VERSION) {
976 		spa_close(spa, FTAG);
977 		return (EINVAL);
978 	}
979 
980 	spa_upgrade(spa, zc->zc_cookie);
981 	spa_close(spa, FTAG);
982 
983 	return (error);
984 }
985 
986 static int
987 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
988 {
989 	spa_t *spa;
990 	char *hist_buf;
991 	uint64_t size;
992 	int error;
993 
994 	if ((size = zc->zc_history_len) == 0)
995 		return (EINVAL);
996 
997 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
998 		return (error);
999 
1000 	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1001 		spa_close(spa, FTAG);
1002 		return (ENOTSUP);
1003 	}
1004 
1005 	hist_buf = kmem_alloc(size, KM_SLEEP);
1006 	if ((error = spa_history_get(spa, &zc->zc_history_offset,
1007 	    &zc->zc_history_len, hist_buf)) == 0) {
1008 		error = xcopyout(hist_buf,
1009 		    (char *)(uintptr_t)zc->zc_history,
1010 		    zc->zc_history_len);
1011 	}
1012 
1013 	spa_close(spa, FTAG);
1014 	kmem_free(hist_buf, size);
1015 	return (error);
1016 }
1017 
1018 static int
1019 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1020 {
1021 	int error;
1022 
1023 	if (error = dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value))
1024 		return (error);
1025 
1026 	return (0);
1027 }
1028 
1029 static int
1030 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1031 {
1032 	objset_t *osp;
1033 	int error;
1034 
1035 	if ((error = dmu_objset_open(zc->zc_name, DMU_OST_ZFS,
1036 	    DS_MODE_USER | DS_MODE_READONLY, &osp)) != 0)
1037 		return (error);
1038 	error = zfs_obj_to_path(osp, zc->zc_obj, zc->zc_value,
1039 	    sizeof (zc->zc_value));
1040 	dmu_objset_close(osp);
1041 
1042 	return (error);
1043 }
1044 
1045 static int
1046 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1047 {
1048 	spa_t *spa;
1049 	int error;
1050 	nvlist_t *config, **l2cache, **spares;
1051 	uint_t nl2cache = 0, nspares = 0;
1052 
1053 	error = spa_open(zc->zc_name, &spa, FTAG);
1054 	if (error != 0)
1055 		return (error);
1056 
1057 	error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1058 	    &config);
1059 	(void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1060 	    &l2cache, &nl2cache);
1061 
1062 	(void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1063 	    &spares, &nspares);
1064 
1065 	/*
1066 	 * A root pool with concatenated devices is not supported.
1067 	 * Thus, can not add a device to a root pool.
1068 	 *
1069 	 * Intent log device can not be added to a rootpool because
1070 	 * during mountroot, zil is replayed, a seperated log device
1071 	 * can not be accessed during the mountroot time.
1072 	 *
1073 	 * l2cache and spare devices are ok to be added to a rootpool.
1074 	 */
1075 	if (spa->spa_bootfs != 0 && nl2cache == 0 && nspares == 0) {
1076 		spa_close(spa, FTAG);
1077 		return (EDOM);
1078 	}
1079 
1080 	if (error == 0) {
1081 		error = spa_vdev_add(spa, config);
1082 		nvlist_free(config);
1083 	}
1084 	spa_close(spa, FTAG);
1085 	return (error);
1086 }
1087 
1088 static int
1089 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1090 {
1091 	spa_t *spa;
1092 	int error;
1093 
1094 	error = spa_open(zc->zc_name, &spa, FTAG);
1095 	if (error != 0)
1096 		return (error);
1097 	error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1098 	spa_close(spa, FTAG);
1099 	return (error);
1100 }
1101 
1102 static int
1103 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1104 {
1105 	spa_t *spa;
1106 	int error;
1107 	vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1108 
1109 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1110 		return (error);
1111 	switch (zc->zc_cookie) {
1112 	case VDEV_STATE_ONLINE:
1113 		error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1114 		break;
1115 
1116 	case VDEV_STATE_OFFLINE:
1117 		error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1118 		break;
1119 
1120 	case VDEV_STATE_FAULTED:
1121 		error = vdev_fault(spa, zc->zc_guid);
1122 		break;
1123 
1124 	case VDEV_STATE_DEGRADED:
1125 		error = vdev_degrade(spa, zc->zc_guid);
1126 		break;
1127 
1128 	default:
1129 		error = EINVAL;
1130 	}
1131 	zc->zc_cookie = newstate;
1132 	spa_close(spa, FTAG);
1133 	return (error);
1134 }
1135 
1136 static int
1137 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1138 {
1139 	spa_t *spa;
1140 	int replacing = zc->zc_cookie;
1141 	nvlist_t *config;
1142 	int error;
1143 
1144 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1145 		return (error);
1146 
1147 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1148 	    &config)) == 0) {
1149 		error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1150 		nvlist_free(config);
1151 	}
1152 
1153 	spa_close(spa, FTAG);
1154 	return (error);
1155 }
1156 
1157 static int
1158 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1159 {
1160 	spa_t *spa;
1161 	int error;
1162 
1163 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1164 		return (error);
1165 
1166 	error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1167 
1168 	spa_close(spa, FTAG);
1169 	return (error);
1170 }
1171 
1172 static int
1173 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
1174 {
1175 	spa_t *spa;
1176 	char *path = zc->zc_value;
1177 	uint64_t guid = zc->zc_guid;
1178 	int error;
1179 
1180 	error = spa_open(zc->zc_name, &spa, FTAG);
1181 	if (error != 0)
1182 		return (error);
1183 
1184 	error = spa_vdev_setpath(spa, guid, path);
1185 	spa_close(spa, FTAG);
1186 	return (error);
1187 }
1188 
1189 /*
1190  * inputs:
1191  * zc_name		name of filesystem
1192  * zc_nvlist_dst_size	size of buffer for property nvlist
1193  *
1194  * outputs:
1195  * zc_objset_stats	stats
1196  * zc_nvlist_dst	property nvlist
1197  * zc_nvlist_dst_size	size of property nvlist
1198  */
1199 static int
1200 zfs_ioc_objset_stats(zfs_cmd_t *zc)
1201 {
1202 	objset_t *os = NULL;
1203 	int error;
1204 	nvlist_t *nv;
1205 
1206 	if (error = dmu_objset_open(zc->zc_name,
1207 	    DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os))
1208 		return (error);
1209 
1210 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1211 
1212 	if (zc->zc_nvlist_dst != 0 &&
1213 	    (error = dsl_prop_get_all(os, &nv, FALSE)) == 0) {
1214 		dmu_objset_stats(os, nv);
1215 		/*
1216 		 * NB: zvol_get_stats() will read the objset contents,
1217 		 * which we aren't supposed to do with a
1218 		 * DS_MODE_USER hold, because it could be
1219 		 * inconsistent.  So this is a bit of a workaround...
1220 		 */
1221 		if (!zc->zc_objset_stats.dds_inconsistent) {
1222 			if (dmu_objset_type(os) == DMU_OST_ZVOL)
1223 				VERIFY(zvol_get_stats(os, nv) == 0);
1224 		}
1225 		error = put_nvlist(zc, nv);
1226 		nvlist_free(nv);
1227 	}
1228 
1229 	dmu_objset_close(os);
1230 	return (error);
1231 }
1232 
1233 static int
1234 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
1235 {
1236 	uint64_t value;
1237 	int error;
1238 
1239 	/*
1240 	 * zfs_get_zplprop() will either find a value or give us
1241 	 * the default value (if there is one).
1242 	 */
1243 	if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
1244 		return (error);
1245 	VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
1246 	return (0);
1247 }
1248 
1249 /*
1250  * inputs:
1251  * zc_name		name of filesystem
1252  * zc_nvlist_dst_size	size of buffer for zpl property nvlist
1253  *
1254  * outputs:
1255  * zc_nvlist_dst	zpl property nvlist
1256  * zc_nvlist_dst_size	size of zpl property nvlist
1257  */
1258 static int
1259 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
1260 {
1261 	objset_t *os;
1262 	int err;
1263 
1264 	if (err = dmu_objset_open(zc->zc_name,
1265 	    DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os))
1266 		return (err);
1267 
1268 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1269 
1270 	/*
1271 	 * NB: nvl_add_zplprop() will read the objset contents,
1272 	 * which we aren't supposed to do with a DS_MODE_USER
1273 	 * hold, because it could be inconsistent.
1274 	 */
1275 	if (zc->zc_nvlist_dst != NULL &&
1276 	    !zc->zc_objset_stats.dds_inconsistent &&
1277 	    dmu_objset_type(os) == DMU_OST_ZFS) {
1278 		nvlist_t *nv;
1279 
1280 		VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1281 		if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
1282 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
1283 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
1284 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
1285 			err = put_nvlist(zc, nv);
1286 		nvlist_free(nv);
1287 	} else {
1288 		err = ENOENT;
1289 	}
1290 	dmu_objset_close(os);
1291 	return (err);
1292 }
1293 
1294 /*
1295  * inputs:
1296  * zc_name		name of filesystem
1297  * zc_cookie		zap cursor
1298  * zc_nvlist_dst_size	size of buffer for property nvlist
1299  *
1300  * outputs:
1301  * zc_name		name of next filesystem
1302  * zc_objset_stats	stats
1303  * zc_nvlist_dst	property nvlist
1304  * zc_nvlist_dst_size	size of property nvlist
1305  */
1306 static int
1307 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
1308 {
1309 	objset_t *os;
1310 	int error;
1311 	char *p;
1312 
1313 	if (error = dmu_objset_open(zc->zc_name,
1314 	    DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os)) {
1315 		if (error == ENOENT)
1316 			error = ESRCH;
1317 		return (error);
1318 	}
1319 
1320 	p = strrchr(zc->zc_name, '/');
1321 	if (p == NULL || p[1] != '\0')
1322 		(void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
1323 	p = zc->zc_name + strlen(zc->zc_name);
1324 
1325 	/*
1326 	 * Pre-fetch the datasets.  dmu_objset_prefetch() always returns 0
1327 	 * but is not declared void because its called by dmu_objset_find().
1328 	 */
1329 	if (zc->zc_cookie == 0) {
1330 		uint64_t cookie = 0;
1331 		int len = sizeof (zc->zc_name) - (p - zc->zc_name);
1332 
1333 		while (dmu_dir_list_next(os, len, p, NULL, &cookie) == 0)
1334 			(void) dmu_objset_prefetch(p, NULL);
1335 	}
1336 
1337 	do {
1338 		error = dmu_dir_list_next(os,
1339 		    sizeof (zc->zc_name) - (p - zc->zc_name), p,
1340 		    NULL, &zc->zc_cookie);
1341 		if (error == ENOENT)
1342 			error = ESRCH;
1343 	} while (error == 0 && !INGLOBALZONE(curproc) &&
1344 	    !zone_dataset_visible(zc->zc_name, NULL));
1345 	dmu_objset_close(os);
1346 
1347 	/*
1348 	 * If it's a hidden dataset (ie. with a '$' in its name), don't
1349 	 * try to get stats for it.  Userland will skip over it.
1350 	 */
1351 	if (error == 0 && strchr(zc->zc_name, '$') == NULL)
1352 		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
1353 
1354 	return (error);
1355 }
1356 
1357 /*
1358  * inputs:
1359  * zc_name		name of filesystem
1360  * zc_cookie		zap cursor
1361  * zc_nvlist_dst_size	size of buffer for property nvlist
1362  *
1363  * outputs:
1364  * zc_name		name of next snapshot
1365  * zc_objset_stats	stats
1366  * zc_nvlist_dst	property nvlist
1367  * zc_nvlist_dst_size	size of property nvlist
1368  */
1369 static int
1370 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
1371 {
1372 	objset_t *os;
1373 	int error;
1374 
1375 	error = dmu_objset_open(zc->zc_name,
1376 	    DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os);
1377 	if (error)
1378 		return (error == ENOENT ? ESRCH : error);
1379 
1380 	if (zc->zc_cookie == 0)
1381 		(void) dmu_objset_find(zc->zc_name, dmu_objset_prefetch,
1382 		    NULL, DS_FIND_SNAPSHOTS);
1383 	/*
1384 	 * A dataset name of maximum length cannot have any snapshots,
1385 	 * so exit immediately.
1386 	 */
1387 	if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
1388 		dmu_objset_close(os);
1389 		return (ESRCH);
1390 	}
1391 
1392 	error = dmu_snapshot_list_next(os,
1393 	    sizeof (zc->zc_name) - strlen(zc->zc_name),
1394 	    zc->zc_name + strlen(zc->zc_name), NULL, &zc->zc_cookie, NULL);
1395 	dmu_objset_close(os);
1396 	if (error == 0)
1397 		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
1398 	else if (error == ENOENT)
1399 		error = ESRCH;
1400 
1401 	/* if we failed, undo the @ that we tacked on to zc_name */
1402 	if (error)
1403 		*strchr(zc->zc_name, '@') = '\0';
1404 	return (error);
1405 }
1406 
1407 int
1408 zfs_set_prop_nvlist(const char *name, nvlist_t *nvl)
1409 {
1410 	nvpair_t *elem;
1411 	int error = 0;
1412 	uint64_t intval;
1413 	char *strval;
1414 	nvlist_t *genericnvl;
1415 
1416 	/*
1417 	 * First validate permission to set all of the properties
1418 	 */
1419 	elem = NULL;
1420 	while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) {
1421 		const char *propname = nvpair_name(elem);
1422 		zfs_prop_t prop = zfs_name_to_prop(propname);
1423 
1424 		if (prop == ZPROP_INVAL) {
1425 			/*
1426 			 * If this is a user-defined property, it must be a
1427 			 * string, and there is no further validation to do.
1428 			 */
1429 			if (!zfs_prop_user(propname) ||
1430 			    nvpair_type(elem) != DATA_TYPE_STRING)
1431 				return (EINVAL);
1432 
1433 			if (error = zfs_secpolicy_write_perms(name,
1434 			    ZFS_DELEG_PERM_USERPROP, CRED()))
1435 				return (error);
1436 			continue;
1437 		}
1438 
1439 		if ((error = zfs_secpolicy_setprop(name, prop, CRED())) != 0)
1440 			return (error);
1441 
1442 		/*
1443 		 * Check that this value is valid for this pool version
1444 		 */
1445 		switch (prop) {
1446 		case ZFS_PROP_COMPRESSION:
1447 			/*
1448 			 * If the user specified gzip compression, make sure
1449 			 * the SPA supports it. We ignore any errors here since
1450 			 * we'll catch them later.
1451 			 */
1452 			if (nvpair_type(elem) == DATA_TYPE_UINT64 &&
1453 			    nvpair_value_uint64(elem, &intval) == 0) {
1454 				if (intval >= ZIO_COMPRESS_GZIP_1 &&
1455 				    intval <= ZIO_COMPRESS_GZIP_9 &&
1456 				    zfs_earlier_version(name,
1457 				    SPA_VERSION_GZIP_COMPRESSION))
1458 					return (ENOTSUP);
1459 
1460 				/*
1461 				 * If this is a bootable dataset then
1462 				 * verify that the compression algorithm
1463 				 * is supported for booting. We must return
1464 				 * something other than ENOTSUP since it
1465 				 * implies a downrev pool version.
1466 				 */
1467 				if (zfs_is_bootfs(name) &&
1468 				    !BOOTFS_COMPRESS_VALID(intval))
1469 					return (ERANGE);
1470 			}
1471 			break;
1472 
1473 		case ZFS_PROP_COPIES:
1474 			if (zfs_earlier_version(name,
1475 			    SPA_VERSION_DITTO_BLOCKS))
1476 				return (ENOTSUP);
1477 			break;
1478 
1479 		case ZFS_PROP_SHARESMB:
1480 			if (zpl_earlier_version(name, ZPL_VERSION_FUID))
1481 				return (ENOTSUP);
1482 			break;
1483 
1484 		case ZFS_PROP_ACLINHERIT:
1485 			if (nvpair_type(elem) == DATA_TYPE_UINT64 &&
1486 			    nvpair_value_uint64(elem, &intval) == 0)
1487 				if (intval == ZFS_ACL_PASSTHROUGH_X &&
1488 				    zfs_earlier_version(name,
1489 				    SPA_VERSION_PASSTHROUGH_X))
1490 					return (ENOTSUP);
1491 		}
1492 	}
1493 
1494 	VERIFY(nvlist_alloc(&genericnvl, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1495 	elem = NULL;
1496 	while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) {
1497 		const char *propname = nvpair_name(elem);
1498 		zfs_prop_t prop = zfs_name_to_prop(propname);
1499 
1500 		if (prop == ZPROP_INVAL) {
1501 			VERIFY(nvpair_value_string(elem, &strval) == 0);
1502 			error = dsl_prop_set(name, propname, 1,
1503 			    strlen(strval) + 1, strval);
1504 			if (error == 0)
1505 				continue;
1506 			else
1507 				goto out;
1508 		}
1509 
1510 		switch (prop) {
1511 		case ZFS_PROP_QUOTA:
1512 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1513 			    (error = dsl_dir_set_quota(name, intval)) != 0)
1514 				goto out;
1515 			break;
1516 
1517 		case ZFS_PROP_REFQUOTA:
1518 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1519 			    (error = dsl_dataset_set_quota(name, intval)) != 0)
1520 				goto out;
1521 			break;
1522 
1523 		case ZFS_PROP_RESERVATION:
1524 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1525 			    (error = dsl_dir_set_reservation(name,
1526 			    intval)) != 0)
1527 				goto out;
1528 			break;
1529 
1530 		case ZFS_PROP_REFRESERVATION:
1531 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1532 			    (error = dsl_dataset_set_reservation(name,
1533 			    intval)) != 0)
1534 				goto out;
1535 			break;
1536 
1537 		case ZFS_PROP_VOLSIZE:
1538 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1539 			    (error = zvol_set_volsize(name,
1540 			    ddi_driver_major(zfs_dip), intval)) != 0)
1541 				goto out;
1542 			break;
1543 
1544 		case ZFS_PROP_VOLBLOCKSIZE:
1545 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1546 			    (error = zvol_set_volblocksize(name, intval)) != 0)
1547 				goto out;
1548 			break;
1549 
1550 		case ZFS_PROP_VERSION:
1551 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1552 			    (error = zfs_set_version(name, intval)) != 0)
1553 				goto out;
1554 			break;
1555 
1556 		default:
1557 			if (nvpair_type(elem) == DATA_TYPE_STRING) {
1558 				if (zfs_prop_get_type(prop) !=
1559 				    PROP_TYPE_STRING) {
1560 					error = EINVAL;
1561 					goto out;
1562 				}
1563 			} else if (nvpair_type(elem) == DATA_TYPE_UINT64) {
1564 				const char *unused;
1565 
1566 				VERIFY(nvpair_value_uint64(elem, &intval) == 0);
1567 
1568 				switch (zfs_prop_get_type(prop)) {
1569 				case PROP_TYPE_NUMBER:
1570 					break;
1571 				case PROP_TYPE_STRING:
1572 					error = EINVAL;
1573 					goto out;
1574 				case PROP_TYPE_INDEX:
1575 					if (zfs_prop_index_to_string(prop,
1576 					    intval, &unused) != 0) {
1577 						error = EINVAL;
1578 						goto out;
1579 					}
1580 					break;
1581 				default:
1582 					cmn_err(CE_PANIC,
1583 					    "unknown property type");
1584 					break;
1585 				}
1586 			} else {
1587 				error = EINVAL;
1588 				goto out;
1589 			}
1590 			if ((error = nvlist_add_nvpair(genericnvl, elem)) != 0)
1591 				goto out;
1592 		}
1593 	}
1594 
1595 	if (nvlist_next_nvpair(genericnvl, NULL) != NULL) {
1596 		error = dsl_props_set(name, genericnvl);
1597 	}
1598 out:
1599 	nvlist_free(genericnvl);
1600 	return (error);
1601 }
1602 
1603 /*
1604  * inputs:
1605  * zc_name		name of filesystem
1606  * zc_value		name of property to set
1607  * zc_nvlist_src{_size}	nvlist of properties to apply
1608  * zc_cookie		clear existing local props?
1609  *
1610  * outputs:		none
1611  */
1612 static int
1613 zfs_ioc_set_prop(zfs_cmd_t *zc)
1614 {
1615 	nvlist_t *nvl;
1616 	int error;
1617 
1618 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1619 	    &nvl)) != 0)
1620 		return (error);
1621 
1622 	if (zc->zc_cookie) {
1623 		nvlist_t *origprops;
1624 		objset_t *os;
1625 
1626 		if (dmu_objset_open(zc->zc_name, DMU_OST_ANY,
1627 		    DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
1628 			if (dsl_prop_get_all(os, &origprops, TRUE) == 0) {
1629 				clear_props(zc->zc_name, origprops, nvl);
1630 				nvlist_free(origprops);
1631 			}
1632 			dmu_objset_close(os);
1633 		}
1634 
1635 	}
1636 
1637 	error = zfs_set_prop_nvlist(zc->zc_name, nvl);
1638 
1639 	nvlist_free(nvl);
1640 	return (error);
1641 }
1642 
1643 /*
1644  * inputs:
1645  * zc_name		name of filesystem
1646  * zc_value		name of property to inherit
1647  *
1648  * outputs:		none
1649  */
1650 static int
1651 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
1652 {
1653 	/* the property name has been validated by zfs_secpolicy_inherit() */
1654 	return (dsl_prop_set(zc->zc_name, zc->zc_value, 0, 0, NULL));
1655 }
1656 
1657 static int
1658 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
1659 {
1660 	nvlist_t *props;
1661 	spa_t *spa;
1662 	int error;
1663 	nvpair_t *elem;
1664 
1665 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1666 	    &props)))
1667 		return (error);
1668 
1669 	/*
1670 	 * If the only property is the configfile, then just do a spa_lookup()
1671 	 * to handle the faulted case.
1672 	 */
1673 	elem = nvlist_next_nvpair(props, NULL);
1674 	if (elem != NULL && strcmp(nvpair_name(elem),
1675 	    zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
1676 	    nvlist_next_nvpair(props, elem) == NULL) {
1677 		mutex_enter(&spa_namespace_lock);
1678 		if ((spa = spa_lookup(zc->zc_name)) != NULL) {
1679 			spa_configfile_set(spa, props, B_FALSE);
1680 			spa_config_sync(spa, B_FALSE, B_TRUE);
1681 		}
1682 		mutex_exit(&spa_namespace_lock);
1683 		if (spa != NULL)
1684 			return (0);
1685 	}
1686 
1687 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
1688 		nvlist_free(props);
1689 		return (error);
1690 	}
1691 
1692 	error = spa_prop_set(spa, props);
1693 
1694 	nvlist_free(props);
1695 	spa_close(spa, FTAG);
1696 
1697 	return (error);
1698 }
1699 
1700 static int
1701 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
1702 {
1703 	spa_t *spa;
1704 	int error;
1705 	nvlist_t *nvp = NULL;
1706 
1707 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
1708 		/*
1709 		 * If the pool is faulted, there may be properties we can still
1710 		 * get (such as altroot and cachefile), so attempt to get them
1711 		 * anyway.
1712 		 */
1713 		mutex_enter(&spa_namespace_lock);
1714 		if ((spa = spa_lookup(zc->zc_name)) != NULL)
1715 			error = spa_prop_get(spa, &nvp);
1716 		mutex_exit(&spa_namespace_lock);
1717 	} else {
1718 		error = spa_prop_get(spa, &nvp);
1719 		spa_close(spa, FTAG);
1720 	}
1721 
1722 	if (error == 0 && zc->zc_nvlist_dst != NULL)
1723 		error = put_nvlist(zc, nvp);
1724 	else
1725 		error = EFAULT;
1726 
1727 	nvlist_free(nvp);
1728 	return (error);
1729 }
1730 
1731 static int
1732 zfs_ioc_iscsi_perm_check(zfs_cmd_t *zc)
1733 {
1734 	nvlist_t *nvp;
1735 	int error;
1736 	uint32_t uid;
1737 	uint32_t gid;
1738 	uint32_t *groups;
1739 	uint_t group_cnt;
1740 	cred_t	*usercred;
1741 
1742 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1743 	    &nvp)) != 0) {
1744 		return (error);
1745 	}
1746 
1747 	if ((error = nvlist_lookup_uint32(nvp,
1748 	    ZFS_DELEG_PERM_UID, &uid)) != 0) {
1749 		nvlist_free(nvp);
1750 		return (EPERM);
1751 	}
1752 
1753 	if ((error = nvlist_lookup_uint32(nvp,
1754 	    ZFS_DELEG_PERM_GID, &gid)) != 0) {
1755 		nvlist_free(nvp);
1756 		return (EPERM);
1757 	}
1758 
1759 	if ((error = nvlist_lookup_uint32_array(nvp, ZFS_DELEG_PERM_GROUPS,
1760 	    &groups, &group_cnt)) != 0) {
1761 		nvlist_free(nvp);
1762 		return (EPERM);
1763 	}
1764 	usercred = cralloc();
1765 	if ((crsetugid(usercred, uid, gid) != 0) ||
1766 	    (crsetgroups(usercred, group_cnt, (gid_t *)groups) != 0)) {
1767 		nvlist_free(nvp);
1768 		crfree(usercred);
1769 		return (EPERM);
1770 	}
1771 	nvlist_free(nvp);
1772 	error = dsl_deleg_access(zc->zc_name,
1773 	    zfs_prop_to_name(ZFS_PROP_SHAREISCSI), usercred);
1774 	crfree(usercred);
1775 	return (error);
1776 }
1777 
1778 /*
1779  * inputs:
1780  * zc_name		name of filesystem
1781  * zc_nvlist_src{_size}	nvlist of delegated permissions
1782  * zc_perm_action	allow/unallow flag
1783  *
1784  * outputs:		none
1785  */
1786 static int
1787 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
1788 {
1789 	int error;
1790 	nvlist_t *fsaclnv = NULL;
1791 
1792 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1793 	    &fsaclnv)) != 0)
1794 		return (error);
1795 
1796 	/*
1797 	 * Verify nvlist is constructed correctly
1798 	 */
1799 	if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
1800 		nvlist_free(fsaclnv);
1801 		return (EINVAL);
1802 	}
1803 
1804 	/*
1805 	 * If we don't have PRIV_SYS_MOUNT, then validate
1806 	 * that user is allowed to hand out each permission in
1807 	 * the nvlist(s)
1808 	 */
1809 
1810 	error = secpolicy_zfs(CRED());
1811 	if (error) {
1812 		if (zc->zc_perm_action == B_FALSE) {
1813 			error = dsl_deleg_can_allow(zc->zc_name,
1814 			    fsaclnv, CRED());
1815 		} else {
1816 			error = dsl_deleg_can_unallow(zc->zc_name,
1817 			    fsaclnv, CRED());
1818 		}
1819 	}
1820 
1821 	if (error == 0)
1822 		error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
1823 
1824 	nvlist_free(fsaclnv);
1825 	return (error);
1826 }
1827 
1828 /*
1829  * inputs:
1830  * zc_name		name of filesystem
1831  *
1832  * outputs:
1833  * zc_nvlist_src{_size}	nvlist of delegated permissions
1834  */
1835 static int
1836 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
1837 {
1838 	nvlist_t *nvp;
1839 	int error;
1840 
1841 	if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
1842 		error = put_nvlist(zc, nvp);
1843 		nvlist_free(nvp);
1844 	}
1845 
1846 	return (error);
1847 }
1848 
1849 /*
1850  * inputs:
1851  * zc_name		name of volume
1852  *
1853  * outputs:		none
1854  */
1855 static int
1856 zfs_ioc_create_minor(zfs_cmd_t *zc)
1857 {
1858 	return (zvol_create_minor(zc->zc_name, ddi_driver_major(zfs_dip)));
1859 }
1860 
1861 /*
1862  * inputs:
1863  * zc_name		name of volume
1864  *
1865  * outputs:		none
1866  */
1867 static int
1868 zfs_ioc_remove_minor(zfs_cmd_t *zc)
1869 {
1870 	return (zvol_remove_minor(zc->zc_name));
1871 }
1872 
1873 /*
1874  * Search the vfs list for a specified resource.  Returns a pointer to it
1875  * or NULL if no suitable entry is found. The caller of this routine
1876  * is responsible for releasing the returned vfs pointer.
1877  */
1878 static vfs_t *
1879 zfs_get_vfs(const char *resource)
1880 {
1881 	struct vfs *vfsp;
1882 	struct vfs *vfs_found = NULL;
1883 
1884 	vfs_list_read_lock();
1885 	vfsp = rootvfs;
1886 	do {
1887 		if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
1888 			VFS_HOLD(vfsp);
1889 			vfs_found = vfsp;
1890 			break;
1891 		}
1892 		vfsp = vfsp->vfs_next;
1893 	} while (vfsp != rootvfs);
1894 	vfs_list_unlock();
1895 	return (vfs_found);
1896 }
1897 
1898 /* ARGSUSED */
1899 static void
1900 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
1901 {
1902 	zfs_creat_t *zct = arg;
1903 
1904 	zfs_create_fs(os, cr, zct->zct_zplprops, tx);
1905 }
1906 
1907 #define	ZFS_PROP_UNDEFINED	((uint64_t)-1)
1908 
1909 /*
1910  * inputs:
1911  * createprops		list of properties requested by creator
1912  * default_zplver	zpl version to use if unspecified in createprops
1913  * fuids_ok		fuids allowed in this version of the spa?
1914  * os			parent objset pointer (NULL if root fs)
1915  *
1916  * outputs:
1917  * zplprops	values for the zplprops we attach to the master node object
1918  * is_ci	true if requested file system will be purely case-insensitive
1919  *
1920  * Determine the settings for utf8only, normalization and
1921  * casesensitivity.  Specific values may have been requested by the
1922  * creator and/or we can inherit values from the parent dataset.  If
1923  * the file system is of too early a vintage, a creator can not
1924  * request settings for these properties, even if the requested
1925  * setting is the default value.  We don't actually want to create dsl
1926  * properties for these, so remove them from the source nvlist after
1927  * processing.
1928  */
1929 static int
1930 zfs_fill_zplprops_impl(objset_t *os, uint64_t default_zplver,
1931     boolean_t fuids_ok, nvlist_t *createprops, nvlist_t *zplprops,
1932     boolean_t *is_ci)
1933 {
1934 	uint64_t zplver = default_zplver;
1935 	uint64_t sense = ZFS_PROP_UNDEFINED;
1936 	uint64_t norm = ZFS_PROP_UNDEFINED;
1937 	uint64_t u8 = ZFS_PROP_UNDEFINED;
1938 
1939 	ASSERT(zplprops != NULL);
1940 
1941 	/*
1942 	 * Pull out creator prop choices, if any.
1943 	 */
1944 	if (createprops) {
1945 		(void) nvlist_lookup_uint64(createprops,
1946 		    zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
1947 		(void) nvlist_lookup_uint64(createprops,
1948 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
1949 		(void) nvlist_remove_all(createprops,
1950 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE));
1951 		(void) nvlist_lookup_uint64(createprops,
1952 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
1953 		(void) nvlist_remove_all(createprops,
1954 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
1955 		(void) nvlist_lookup_uint64(createprops,
1956 		    zfs_prop_to_name(ZFS_PROP_CASE), &sense);
1957 		(void) nvlist_remove_all(createprops,
1958 		    zfs_prop_to_name(ZFS_PROP_CASE));
1959 	}
1960 
1961 	/*
1962 	 * If the zpl version requested is whacky or the file system
1963 	 * or pool is version is too "young" to support normalization
1964 	 * and the creator tried to set a value for one of the props,
1965 	 * error out.
1966 	 */
1967 	if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
1968 	    (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
1969 	    (zplver < ZPL_VERSION_NORMALIZATION &&
1970 	    (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
1971 	    sense != ZFS_PROP_UNDEFINED)))
1972 		return (ENOTSUP);
1973 
1974 	/*
1975 	 * Put the version in the zplprops
1976 	 */
1977 	VERIFY(nvlist_add_uint64(zplprops,
1978 	    zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
1979 
1980 	if (norm == ZFS_PROP_UNDEFINED)
1981 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
1982 	VERIFY(nvlist_add_uint64(zplprops,
1983 	    zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
1984 
1985 	/*
1986 	 * If we're normalizing, names must always be valid UTF-8 strings.
1987 	 */
1988 	if (norm)
1989 		u8 = 1;
1990 	if (u8 == ZFS_PROP_UNDEFINED)
1991 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
1992 	VERIFY(nvlist_add_uint64(zplprops,
1993 	    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
1994 
1995 	if (sense == ZFS_PROP_UNDEFINED)
1996 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
1997 	VERIFY(nvlist_add_uint64(zplprops,
1998 	    zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
1999 
2000 	if (is_ci)
2001 		*is_ci = (sense == ZFS_CASE_INSENSITIVE);
2002 
2003 	return (0);
2004 }
2005 
2006 static int
2007 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
2008     nvlist_t *zplprops, boolean_t *is_ci)
2009 {
2010 	boolean_t fuids_ok = B_TRUE;
2011 	uint64_t zplver = ZPL_VERSION;
2012 	objset_t *os = NULL;
2013 	char parentname[MAXNAMELEN];
2014 	char *cp;
2015 	int error;
2016 
2017 	(void) strlcpy(parentname, dataset, sizeof (parentname));
2018 	cp = strrchr(parentname, '/');
2019 	ASSERT(cp != NULL);
2020 	cp[0] = '\0';
2021 
2022 	if (zfs_earlier_version(dataset, SPA_VERSION_FUID)) {
2023 		zplver = ZPL_VERSION_FUID - 1;
2024 		fuids_ok = B_FALSE;
2025 	}
2026 
2027 	/*
2028 	 * Open parent object set so we can inherit zplprop values.
2029 	 */
2030 	if ((error = dmu_objset_open(parentname, DMU_OST_ANY,
2031 	    DS_MODE_USER | DS_MODE_READONLY, &os)) != 0)
2032 		return (error);
2033 
2034 	error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, createprops,
2035 	    zplprops, is_ci);
2036 	dmu_objset_close(os);
2037 	return (error);
2038 }
2039 
2040 static int
2041 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
2042     nvlist_t *zplprops, boolean_t *is_ci)
2043 {
2044 	boolean_t fuids_ok = B_TRUE;
2045 	uint64_t zplver = ZPL_VERSION;
2046 	int error;
2047 
2048 	if (spa_vers < SPA_VERSION_FUID) {
2049 		zplver = ZPL_VERSION_FUID - 1;
2050 		fuids_ok = B_FALSE;
2051 	}
2052 
2053 	error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, createprops,
2054 	    zplprops, is_ci);
2055 	return (error);
2056 }
2057 
2058 /*
2059  * inputs:
2060  * zc_objset_type	type of objset to create (fs vs zvol)
2061  * zc_name		name of new objset
2062  * zc_value		name of snapshot to clone from (may be empty)
2063  * zc_nvlist_src{_size}	nvlist of properties to apply
2064  *
2065  * outputs: none
2066  */
2067 static int
2068 zfs_ioc_create(zfs_cmd_t *zc)
2069 {
2070 	objset_t *clone;
2071 	int error = 0;
2072 	zfs_creat_t zct;
2073 	nvlist_t *nvprops = NULL;
2074 	void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
2075 	dmu_objset_type_t type = zc->zc_objset_type;
2076 
2077 	switch (type) {
2078 
2079 	case DMU_OST_ZFS:
2080 		cbfunc = zfs_create_cb;
2081 		break;
2082 
2083 	case DMU_OST_ZVOL:
2084 		cbfunc = zvol_create_cb;
2085 		break;
2086 
2087 	default:
2088 		cbfunc = NULL;
2089 		break;
2090 	}
2091 	if (strchr(zc->zc_name, '@') ||
2092 	    strchr(zc->zc_name, '%'))
2093 		return (EINVAL);
2094 
2095 	if (zc->zc_nvlist_src != NULL &&
2096 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2097 	    &nvprops)) != 0)
2098 		return (error);
2099 
2100 	zct.zct_zplprops = NULL;
2101 	zct.zct_props = nvprops;
2102 
2103 	if (zc->zc_value[0] != '\0') {
2104 		/*
2105 		 * We're creating a clone of an existing snapshot.
2106 		 */
2107 		zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
2108 		if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0) {
2109 			nvlist_free(nvprops);
2110 			return (EINVAL);
2111 		}
2112 
2113 		error = dmu_objset_open(zc->zc_value, type,
2114 		    DS_MODE_USER | DS_MODE_READONLY, &clone);
2115 		if (error) {
2116 			nvlist_free(nvprops);
2117 			return (error);
2118 		}
2119 
2120 		error = dmu_objset_create(zc->zc_name, type, clone, 0,
2121 		    NULL, NULL);
2122 		if (error) {
2123 			dmu_objset_close(clone);
2124 			nvlist_free(nvprops);
2125 			return (error);
2126 		}
2127 		dmu_objset_close(clone);
2128 	} else {
2129 		boolean_t is_insensitive = B_FALSE;
2130 
2131 		if (cbfunc == NULL) {
2132 			nvlist_free(nvprops);
2133 			return (EINVAL);
2134 		}
2135 
2136 		if (type == DMU_OST_ZVOL) {
2137 			uint64_t volsize, volblocksize;
2138 
2139 			if (nvprops == NULL ||
2140 			    nvlist_lookup_uint64(nvprops,
2141 			    zfs_prop_to_name(ZFS_PROP_VOLSIZE),
2142 			    &volsize) != 0) {
2143 				nvlist_free(nvprops);
2144 				return (EINVAL);
2145 			}
2146 
2147 			if ((error = nvlist_lookup_uint64(nvprops,
2148 			    zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
2149 			    &volblocksize)) != 0 && error != ENOENT) {
2150 				nvlist_free(nvprops);
2151 				return (EINVAL);
2152 			}
2153 
2154 			if (error != 0)
2155 				volblocksize = zfs_prop_default_numeric(
2156 				    ZFS_PROP_VOLBLOCKSIZE);
2157 
2158 			if ((error = zvol_check_volblocksize(
2159 			    volblocksize)) != 0 ||
2160 			    (error = zvol_check_volsize(volsize,
2161 			    volblocksize)) != 0) {
2162 				nvlist_free(nvprops);
2163 				return (error);
2164 			}
2165 		} else if (type == DMU_OST_ZFS) {
2166 			int error;
2167 
2168 			/*
2169 			 * We have to have normalization and
2170 			 * case-folding flags correct when we do the
2171 			 * file system creation, so go figure them out
2172 			 * now.
2173 			 */
2174 			VERIFY(nvlist_alloc(&zct.zct_zplprops,
2175 			    NV_UNIQUE_NAME, KM_SLEEP) == 0);
2176 			error = zfs_fill_zplprops(zc->zc_name, nvprops,
2177 			    zct.zct_zplprops, &is_insensitive);
2178 			if (error != 0) {
2179 				nvlist_free(nvprops);
2180 				nvlist_free(zct.zct_zplprops);
2181 				return (error);
2182 			}
2183 		}
2184 		error = dmu_objset_create(zc->zc_name, type, NULL,
2185 		    is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
2186 		nvlist_free(zct.zct_zplprops);
2187 	}
2188 
2189 	/*
2190 	 * It would be nice to do this atomically.
2191 	 */
2192 	if (error == 0) {
2193 		if ((error = zfs_set_prop_nvlist(zc->zc_name, nvprops)) != 0)
2194 			(void) dmu_objset_destroy(zc->zc_name);
2195 	}
2196 	nvlist_free(nvprops);
2197 	return (error);
2198 }
2199 
2200 struct snap_prop_arg {
2201 	nvlist_t *nvprops;
2202 	const char *snapname;
2203 };
2204 
2205 static int
2206 set_snap_props(char *name, void *arg)
2207 {
2208 	struct snap_prop_arg *snpa = arg;
2209 	int len = strlen(name) + strlen(snpa->snapname) + 2;
2210 	char *buf = kmem_alloc(len, KM_SLEEP);
2211 	int err;
2212 
2213 	(void) snprintf(buf, len, "%s@%s", name, snpa->snapname);
2214 	err = zfs_set_prop_nvlist(buf, snpa->nvprops);
2215 	if (err)
2216 		(void) dmu_objset_destroy(buf);
2217 	kmem_free(buf, len);
2218 	return (err);
2219 }
2220 
2221 /*
2222  * inputs:
2223  * zc_name	name of filesystem
2224  * zc_value	short name of snapshot
2225  * zc_cookie	recursive flag
2226  *
2227  * outputs:	none
2228  */
2229 static int
2230 zfs_ioc_snapshot(zfs_cmd_t *zc)
2231 {
2232 	nvlist_t *nvprops = NULL;
2233 	int error;
2234 	boolean_t recursive = zc->zc_cookie;
2235 
2236 	if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
2237 		return (EINVAL);
2238 
2239 	if (zc->zc_nvlist_src != NULL &&
2240 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2241 	    &nvprops)) != 0)
2242 		return (error);
2243 
2244 	error = dmu_objset_snapshot(zc->zc_name, zc->zc_value, recursive);
2245 
2246 	/*
2247 	 * It would be nice to do this atomically.
2248 	 */
2249 	if (error == 0) {
2250 		struct snap_prop_arg snpa;
2251 		snpa.nvprops = nvprops;
2252 		snpa.snapname = zc->zc_value;
2253 		if (recursive) {
2254 			error = dmu_objset_find(zc->zc_name,
2255 			    set_snap_props, &snpa, DS_FIND_CHILDREN);
2256 			if (error) {
2257 				(void) dmu_snapshots_destroy(zc->zc_name,
2258 				    zc->zc_value);
2259 			}
2260 		} else {
2261 			error = set_snap_props(zc->zc_name, &snpa);
2262 		}
2263 	}
2264 	nvlist_free(nvprops);
2265 	return (error);
2266 }
2267 
2268 int
2269 zfs_unmount_snap(char *name, void *arg)
2270 {
2271 	vfs_t *vfsp = NULL;
2272 
2273 	if (arg) {
2274 		char *snapname = arg;
2275 		int len = strlen(name) + strlen(snapname) + 2;
2276 		char *buf = kmem_alloc(len, KM_SLEEP);
2277 
2278 		(void) strcpy(buf, name);
2279 		(void) strcat(buf, "@");
2280 		(void) strcat(buf, snapname);
2281 		vfsp = zfs_get_vfs(buf);
2282 		kmem_free(buf, len);
2283 	} else if (strchr(name, '@')) {
2284 		vfsp = zfs_get_vfs(name);
2285 	}
2286 
2287 	if (vfsp) {
2288 		/*
2289 		 * Always force the unmount for snapshots.
2290 		 */
2291 		int flag = MS_FORCE;
2292 		int err;
2293 
2294 		if ((err = vn_vfswlock(vfsp->vfs_vnodecovered)) != 0) {
2295 			VFS_RELE(vfsp);
2296 			return (err);
2297 		}
2298 		VFS_RELE(vfsp);
2299 		if ((err = dounmount(vfsp, flag, kcred)) != 0)
2300 			return (err);
2301 	}
2302 	return (0);
2303 }
2304 
2305 /*
2306  * inputs:
2307  * zc_name	name of filesystem
2308  * zc_value	short name of snapshot
2309  *
2310  * outputs:	none
2311  */
2312 static int
2313 zfs_ioc_destroy_snaps(zfs_cmd_t *zc)
2314 {
2315 	int err;
2316 
2317 	if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
2318 		return (EINVAL);
2319 	err = dmu_objset_find(zc->zc_name,
2320 	    zfs_unmount_snap, zc->zc_value, DS_FIND_CHILDREN);
2321 	if (err)
2322 		return (err);
2323 	return (dmu_snapshots_destroy(zc->zc_name, zc->zc_value));
2324 }
2325 
2326 /*
2327  * inputs:
2328  * zc_name		name of dataset to destroy
2329  * zc_objset_type	type of objset
2330  *
2331  * outputs:		none
2332  */
2333 static int
2334 zfs_ioc_destroy(zfs_cmd_t *zc)
2335 {
2336 	if (strchr(zc->zc_name, '@') && zc->zc_objset_type == DMU_OST_ZFS) {
2337 		int err = zfs_unmount_snap(zc->zc_name, NULL);
2338 		if (err)
2339 			return (err);
2340 	}
2341 
2342 	return (dmu_objset_destroy(zc->zc_name));
2343 }
2344 
2345 /*
2346  * inputs:
2347  * zc_name	name of dataset to rollback (to most recent snapshot)
2348  *
2349  * outputs:	none
2350  */
2351 static int
2352 zfs_ioc_rollback(zfs_cmd_t *zc)
2353 {
2354 	objset_t *os;
2355 	int error;
2356 	zfsvfs_t *zfsvfs = NULL;
2357 
2358 	/*
2359 	 * Get the zfsvfs for the receiving objset. There
2360 	 * won't be one if we're operating on a zvol, if the
2361 	 * objset doesn't exist yet, or is not mounted.
2362 	 */
2363 	error = dmu_objset_open(zc->zc_name, DMU_OST_ANY, DS_MODE_USER, &os);
2364 	if (error)
2365 		return (error);
2366 
2367 	if (dmu_objset_type(os) == DMU_OST_ZFS) {
2368 		mutex_enter(&os->os->os_user_ptr_lock);
2369 		zfsvfs = dmu_objset_get_user(os);
2370 		if (zfsvfs != NULL)
2371 			VFS_HOLD(zfsvfs->z_vfs);
2372 		mutex_exit(&os->os->os_user_ptr_lock);
2373 	}
2374 
2375 	if (zfsvfs != NULL) {
2376 		char *osname;
2377 		int mode;
2378 
2379 		osname = kmem_alloc(MAXNAMELEN, KM_SLEEP);
2380 		error = zfs_suspend_fs(zfsvfs, osname, &mode);
2381 		if (error == 0) {
2382 			int resume_err;
2383 
2384 			ASSERT(strcmp(osname, zc->zc_name) == 0);
2385 			error = dmu_objset_rollback(os);
2386 			resume_err = zfs_resume_fs(zfsvfs, osname, mode);
2387 			error = error ? error : resume_err;
2388 		} else {
2389 			dmu_objset_close(os);
2390 		}
2391 		kmem_free(osname, MAXNAMELEN);
2392 		VFS_RELE(zfsvfs->z_vfs);
2393 	} else {
2394 		error = dmu_objset_rollback(os);
2395 	}
2396 	/* Note, the dmu_objset_rollback() releases the objset for us. */
2397 
2398 	return (error);
2399 }
2400 
2401 /*
2402  * inputs:
2403  * zc_name	old name of dataset
2404  * zc_value	new name of dataset
2405  * zc_cookie	recursive flag (only valid for snapshots)
2406  *
2407  * outputs:	none
2408  */
2409 static int
2410 zfs_ioc_rename(zfs_cmd_t *zc)
2411 {
2412 	boolean_t recursive = zc->zc_cookie & 1;
2413 
2414 	zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
2415 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
2416 	    strchr(zc->zc_value, '%'))
2417 		return (EINVAL);
2418 
2419 	/*
2420 	 * Unmount snapshot unless we're doing a recursive rename,
2421 	 * in which case the dataset code figures out which snapshots
2422 	 * to unmount.
2423 	 */
2424 	if (!recursive && strchr(zc->zc_name, '@') != NULL &&
2425 	    zc->zc_objset_type == DMU_OST_ZFS) {
2426 		int err = zfs_unmount_snap(zc->zc_name, NULL);
2427 		if (err)
2428 			return (err);
2429 	}
2430 	return (dmu_objset_rename(zc->zc_name, zc->zc_value, recursive));
2431 }
2432 
2433 static void
2434 clear_props(char *dataset, nvlist_t *props, nvlist_t *newprops)
2435 {
2436 	zfs_cmd_t *zc;
2437 	nvpair_t *prop;
2438 
2439 	if (props == NULL)
2440 		return;
2441 	zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
2442 	(void) strcpy(zc->zc_name, dataset);
2443 	for (prop = nvlist_next_nvpair(props, NULL); prop;
2444 	    prop = nvlist_next_nvpair(props, prop)) {
2445 		if (newprops != NULL &&
2446 		    nvlist_exists(newprops, nvpair_name(prop)))
2447 			continue;
2448 		(void) strcpy(zc->zc_value, nvpair_name(prop));
2449 		if (zfs_secpolicy_inherit(zc, CRED()) == 0)
2450 			(void) zfs_ioc_inherit_prop(zc);
2451 	}
2452 	kmem_free(zc, sizeof (zfs_cmd_t));
2453 }
2454 
2455 /*
2456  * inputs:
2457  * zc_name		name of containing filesystem
2458  * zc_nvlist_src{_size}	nvlist of properties to apply
2459  * zc_value		name of snapshot to create
2460  * zc_string		name of clone origin (if DRR_FLAG_CLONE)
2461  * zc_cookie		file descriptor to recv from
2462  * zc_begin_record	the BEGIN record of the stream (not byteswapped)
2463  * zc_guid		force flag
2464  *
2465  * outputs:
2466  * zc_cookie		number of bytes read
2467  */
2468 static int
2469 zfs_ioc_recv(zfs_cmd_t *zc)
2470 {
2471 	file_t *fp;
2472 	objset_t *os;
2473 	dmu_recv_cookie_t drc;
2474 	zfsvfs_t *zfsvfs = NULL;
2475 	boolean_t force = (boolean_t)zc->zc_guid;
2476 	int error, fd;
2477 	offset_t off;
2478 	nvlist_t *props = NULL;
2479 	nvlist_t *origprops = NULL;
2480 	objset_t *origin = NULL;
2481 	char *tosnap;
2482 	char tofs[ZFS_MAXNAMELEN];
2483 
2484 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
2485 	    strchr(zc->zc_value, '@') == NULL ||
2486 	    strchr(zc->zc_value, '%'))
2487 		return (EINVAL);
2488 
2489 	(void) strcpy(tofs, zc->zc_value);
2490 	tosnap = strchr(tofs, '@');
2491 	*tosnap = '\0';
2492 	tosnap++;
2493 
2494 	if (zc->zc_nvlist_src != NULL &&
2495 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2496 	    &props)) != 0)
2497 		return (error);
2498 
2499 	fd = zc->zc_cookie;
2500 	fp = getf(fd);
2501 	if (fp == NULL) {
2502 		nvlist_free(props);
2503 		return (EBADF);
2504 	}
2505 
2506 	if (dmu_objset_open(tofs, DMU_OST_ANY,
2507 	    DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
2508 		/*
2509 		 * Try to get the zfsvfs for the receiving objset.
2510 		 * There won't be one if we're operating on a zvol,
2511 		 * if the objset doesn't exist yet, or is not mounted.
2512 		 */
2513 		mutex_enter(&os->os->os_user_ptr_lock);
2514 		if (zfsvfs = dmu_objset_get_user(os)) {
2515 			if (!mutex_tryenter(&zfsvfs->z_online_recv_lock)) {
2516 				mutex_exit(&os->os->os_user_ptr_lock);
2517 				dmu_objset_close(os);
2518 				zfsvfs = NULL;
2519 				error = EBUSY;
2520 				goto out;
2521 			}
2522 			VFS_HOLD(zfsvfs->z_vfs);
2523 		}
2524 		mutex_exit(&os->os->os_user_ptr_lock);
2525 
2526 		/*
2527 		 * If new properties are supplied, they are to completely
2528 		 * replace the existing ones, so stash away the existing ones.
2529 		 */
2530 		if (props)
2531 			(void) dsl_prop_get_all(os, &origprops, TRUE);
2532 
2533 		dmu_objset_close(os);
2534 	}
2535 
2536 	if (zc->zc_string[0]) {
2537 		error = dmu_objset_open(zc->zc_string, DMU_OST_ANY,
2538 		    DS_MODE_USER | DS_MODE_READONLY, &origin);
2539 		if (error)
2540 			goto out;
2541 	}
2542 
2543 	error = dmu_recv_begin(tofs, tosnap, &zc->zc_begin_record,
2544 	    force, origin, zfsvfs != NULL, &drc);
2545 	if (origin)
2546 		dmu_objset_close(origin);
2547 	if (error)
2548 		goto out;
2549 
2550 	/*
2551 	 * Reset properties.  We do this before we receive the stream
2552 	 * so that the properties are applied to the new data.
2553 	 */
2554 	if (props) {
2555 		clear_props(tofs, origprops, props);
2556 		/*
2557 		 * XXX - Note, this is all-or-nothing; should be best-effort.
2558 		 */
2559 		(void) zfs_set_prop_nvlist(tofs, props);
2560 	}
2561 
2562 	off = fp->f_offset;
2563 	error = dmu_recv_stream(&drc, fp->f_vnode, &off);
2564 
2565 	if (error == 0 && zfsvfs) {
2566 		char *osname;
2567 		int mode;
2568 
2569 		/* online recv */
2570 		osname = kmem_alloc(MAXNAMELEN, KM_SLEEP);
2571 		error = zfs_suspend_fs(zfsvfs, osname, &mode);
2572 		if (error == 0) {
2573 			int resume_err;
2574 
2575 			error = dmu_recv_end(&drc);
2576 			resume_err = zfs_resume_fs(zfsvfs, osname, mode);
2577 			error = error ? error : resume_err;
2578 		} else {
2579 			dmu_recv_abort_cleanup(&drc);
2580 		}
2581 		kmem_free(osname, MAXNAMELEN);
2582 	} else if (error == 0) {
2583 		error = dmu_recv_end(&drc);
2584 	}
2585 
2586 	zc->zc_cookie = off - fp->f_offset;
2587 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
2588 		fp->f_offset = off;
2589 
2590 	/*
2591 	 * On error, restore the original props.
2592 	 */
2593 	if (error && props) {
2594 		clear_props(tofs, props, NULL);
2595 		(void) zfs_set_prop_nvlist(tofs, origprops);
2596 	}
2597 out:
2598 	if (zfsvfs) {
2599 		mutex_exit(&zfsvfs->z_online_recv_lock);
2600 		VFS_RELE(zfsvfs->z_vfs);
2601 	}
2602 	nvlist_free(props);
2603 	nvlist_free(origprops);
2604 	releasef(fd);
2605 	return (error);
2606 }
2607 
2608 /*
2609  * inputs:
2610  * zc_name	name of snapshot to send
2611  * zc_value	short name of incremental fromsnap (may be empty)
2612  * zc_cookie	file descriptor to send stream to
2613  * zc_obj	fromorigin flag (mutually exclusive with zc_value)
2614  *
2615  * outputs: none
2616  */
2617 static int
2618 zfs_ioc_send(zfs_cmd_t *zc)
2619 {
2620 	objset_t *fromsnap = NULL;
2621 	objset_t *tosnap;
2622 	file_t *fp;
2623 	int error;
2624 	offset_t off;
2625 
2626 	error = dmu_objset_open(zc->zc_name, DMU_OST_ANY,
2627 	    DS_MODE_USER | DS_MODE_READONLY, &tosnap);
2628 	if (error)
2629 		return (error);
2630 
2631 	if (zc->zc_value[0] != '\0') {
2632 		char *buf;
2633 		char *cp;
2634 
2635 		buf = kmem_alloc(MAXPATHLEN, KM_SLEEP);
2636 		(void) strncpy(buf, zc->zc_name, MAXPATHLEN);
2637 		cp = strchr(buf, '@');
2638 		if (cp)
2639 			*(cp+1) = 0;
2640 		(void) strncat(buf, zc->zc_value, MAXPATHLEN);
2641 		error = dmu_objset_open(buf, DMU_OST_ANY,
2642 		    DS_MODE_USER | DS_MODE_READONLY, &fromsnap);
2643 		kmem_free(buf, MAXPATHLEN);
2644 		if (error) {
2645 			dmu_objset_close(tosnap);
2646 			return (error);
2647 		}
2648 	}
2649 
2650 	fp = getf(zc->zc_cookie);
2651 	if (fp == NULL) {
2652 		dmu_objset_close(tosnap);
2653 		if (fromsnap)
2654 			dmu_objset_close(fromsnap);
2655 		return (EBADF);
2656 	}
2657 
2658 	off = fp->f_offset;
2659 	error = dmu_sendbackup(tosnap, fromsnap, zc->zc_obj, fp->f_vnode, &off);
2660 
2661 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
2662 		fp->f_offset = off;
2663 	releasef(zc->zc_cookie);
2664 	if (fromsnap)
2665 		dmu_objset_close(fromsnap);
2666 	dmu_objset_close(tosnap);
2667 	return (error);
2668 }
2669 
2670 static int
2671 zfs_ioc_inject_fault(zfs_cmd_t *zc)
2672 {
2673 	int id, error;
2674 
2675 	error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
2676 	    &zc->zc_inject_record);
2677 
2678 	if (error == 0)
2679 		zc->zc_guid = (uint64_t)id;
2680 
2681 	return (error);
2682 }
2683 
2684 static int
2685 zfs_ioc_clear_fault(zfs_cmd_t *zc)
2686 {
2687 	return (zio_clear_fault((int)zc->zc_guid));
2688 }
2689 
2690 static int
2691 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
2692 {
2693 	int id = (int)zc->zc_guid;
2694 	int error;
2695 
2696 	error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
2697 	    &zc->zc_inject_record);
2698 
2699 	zc->zc_guid = id;
2700 
2701 	return (error);
2702 }
2703 
2704 static int
2705 zfs_ioc_error_log(zfs_cmd_t *zc)
2706 {
2707 	spa_t *spa;
2708 	int error;
2709 	size_t count = (size_t)zc->zc_nvlist_dst_size;
2710 
2711 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2712 		return (error);
2713 
2714 	error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
2715 	    &count);
2716 	if (error == 0)
2717 		zc->zc_nvlist_dst_size = count;
2718 	else
2719 		zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
2720 
2721 	spa_close(spa, FTAG);
2722 
2723 	return (error);
2724 }
2725 
2726 static int
2727 zfs_ioc_clear(zfs_cmd_t *zc)
2728 {
2729 	spa_t *spa;
2730 	vdev_t *vd;
2731 	int error;
2732 
2733 	/*
2734 	 * On zpool clear we also fix up missing slogs
2735 	 */
2736 	mutex_enter(&spa_namespace_lock);
2737 	spa = spa_lookup(zc->zc_name);
2738 	if (spa == NULL) {
2739 		mutex_exit(&spa_namespace_lock);
2740 		return (EIO);
2741 	}
2742 	if (spa->spa_log_state == SPA_LOG_MISSING) {
2743 		/* we need to let spa_open/spa_load clear the chains */
2744 		spa->spa_log_state = SPA_LOG_CLEAR;
2745 	}
2746 	mutex_exit(&spa_namespace_lock);
2747 
2748 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2749 		return (error);
2750 
2751 	spa_vdev_state_enter(spa);
2752 
2753 	if (zc->zc_guid == 0) {
2754 		vd = NULL;
2755 	} else {
2756 		vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
2757 		if (vd == NULL) {
2758 			(void) spa_vdev_state_exit(spa, NULL, ENODEV);
2759 			spa_close(spa, FTAG);
2760 			return (ENODEV);
2761 		}
2762 	}
2763 
2764 	vdev_clear(spa, vd);
2765 
2766 	(void) spa_vdev_state_exit(spa, NULL, 0);
2767 
2768 	/*
2769 	 * Resume any suspended I/Os.
2770 	 */
2771 	zio_resume(spa);
2772 
2773 	spa_close(spa, FTAG);
2774 
2775 	return (0);
2776 }
2777 
2778 /*
2779  * inputs:
2780  * zc_name	name of filesystem
2781  * zc_value	name of origin snapshot
2782  *
2783  * outputs:	none
2784  */
2785 static int
2786 zfs_ioc_promote(zfs_cmd_t *zc)
2787 {
2788 	char *cp;
2789 
2790 	/*
2791 	 * We don't need to unmount *all* the origin fs's snapshots, but
2792 	 * it's easier.
2793 	 */
2794 	cp = strchr(zc->zc_value, '@');
2795 	if (cp)
2796 		*cp = '\0';
2797 	(void) dmu_objset_find(zc->zc_value,
2798 	    zfs_unmount_snap, NULL, DS_FIND_SNAPSHOTS);
2799 	return (dsl_dataset_promote(zc->zc_name));
2800 }
2801 
2802 /*
2803  * We don't want to have a hard dependency
2804  * against some special symbols in sharefs
2805  * nfs, and smbsrv.  Determine them if needed when
2806  * the first file system is shared.
2807  * Neither sharefs, nfs or smbsrv are unloadable modules.
2808  */
2809 int (*znfsexport_fs)(void *arg);
2810 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
2811 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
2812 
2813 int zfs_nfsshare_inited;
2814 int zfs_smbshare_inited;
2815 
2816 ddi_modhandle_t nfs_mod;
2817 ddi_modhandle_t sharefs_mod;
2818 ddi_modhandle_t smbsrv_mod;
2819 kmutex_t zfs_share_lock;
2820 
2821 static int
2822 zfs_init_sharefs()
2823 {
2824 	int error;
2825 
2826 	ASSERT(MUTEX_HELD(&zfs_share_lock));
2827 	/* Both NFS and SMB shares also require sharetab support. */
2828 	if (sharefs_mod == NULL && ((sharefs_mod =
2829 	    ddi_modopen("fs/sharefs",
2830 	    KRTLD_MODE_FIRST, &error)) == NULL)) {
2831 		return (ENOSYS);
2832 	}
2833 	if (zshare_fs == NULL && ((zshare_fs =
2834 	    (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
2835 	    ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
2836 		return (ENOSYS);
2837 	}
2838 	return (0);
2839 }
2840 
2841 static int
2842 zfs_ioc_share(zfs_cmd_t *zc)
2843 {
2844 	int error;
2845 	int opcode;
2846 
2847 	switch (zc->zc_share.z_sharetype) {
2848 	case ZFS_SHARE_NFS:
2849 	case ZFS_UNSHARE_NFS:
2850 		if (zfs_nfsshare_inited == 0) {
2851 			mutex_enter(&zfs_share_lock);
2852 			if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
2853 			    KRTLD_MODE_FIRST, &error)) == NULL)) {
2854 				mutex_exit(&zfs_share_lock);
2855 				return (ENOSYS);
2856 			}
2857 			if (znfsexport_fs == NULL &&
2858 			    ((znfsexport_fs = (int (*)(void *))
2859 			    ddi_modsym(nfs_mod,
2860 			    "nfs_export", &error)) == NULL)) {
2861 				mutex_exit(&zfs_share_lock);
2862 				return (ENOSYS);
2863 			}
2864 			error = zfs_init_sharefs();
2865 			if (error) {
2866 				mutex_exit(&zfs_share_lock);
2867 				return (ENOSYS);
2868 			}
2869 			zfs_nfsshare_inited = 1;
2870 			mutex_exit(&zfs_share_lock);
2871 		}
2872 		break;
2873 	case ZFS_SHARE_SMB:
2874 	case ZFS_UNSHARE_SMB:
2875 		if (zfs_smbshare_inited == 0) {
2876 			mutex_enter(&zfs_share_lock);
2877 			if (smbsrv_mod == NULL && ((smbsrv_mod =
2878 			    ddi_modopen("drv/smbsrv",
2879 			    KRTLD_MODE_FIRST, &error)) == NULL)) {
2880 				mutex_exit(&zfs_share_lock);
2881 				return (ENOSYS);
2882 			}
2883 			if (zsmbexport_fs == NULL && ((zsmbexport_fs =
2884 			    (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
2885 			    "smb_server_share", &error)) == NULL)) {
2886 				mutex_exit(&zfs_share_lock);
2887 				return (ENOSYS);
2888 			}
2889 			error = zfs_init_sharefs();
2890 			if (error) {
2891 				mutex_exit(&zfs_share_lock);
2892 				return (ENOSYS);
2893 			}
2894 			zfs_smbshare_inited = 1;
2895 			mutex_exit(&zfs_share_lock);
2896 		}
2897 		break;
2898 	default:
2899 		return (EINVAL);
2900 	}
2901 
2902 	switch (zc->zc_share.z_sharetype) {
2903 	case ZFS_SHARE_NFS:
2904 	case ZFS_UNSHARE_NFS:
2905 		if (error =
2906 		    znfsexport_fs((void *)
2907 		    (uintptr_t)zc->zc_share.z_exportdata))
2908 			return (error);
2909 		break;
2910 	case ZFS_SHARE_SMB:
2911 	case ZFS_UNSHARE_SMB:
2912 		if (error = zsmbexport_fs((void *)
2913 		    (uintptr_t)zc->zc_share.z_exportdata,
2914 		    zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
2915 		    B_TRUE : B_FALSE)) {
2916 			return (error);
2917 		}
2918 		break;
2919 	}
2920 
2921 	opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
2922 	    zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
2923 	    SHAREFS_ADD : SHAREFS_REMOVE;
2924 
2925 	/*
2926 	 * Add or remove share from sharetab
2927 	 */
2928 	error = zshare_fs(opcode,
2929 	    (void *)(uintptr_t)zc->zc_share.z_sharedata,
2930 	    zc->zc_share.z_sharemax);
2931 
2932 	return (error);
2933 
2934 }
2935 
2936 /*
2937  * pool create, destroy, and export don't log the history as part of
2938  * zfsdev_ioctl, but rather zfs_ioc_pool_create, and zfs_ioc_pool_export
2939  * do the logging of those commands.
2940  */
2941 static zfs_ioc_vec_t zfs_ioc_vec[] = {
2942 	{ zfs_ioc_pool_create, zfs_secpolicy_config, POOL_NAME, B_FALSE },
2943 	{ zfs_ioc_pool_destroy,	zfs_secpolicy_config, POOL_NAME, B_FALSE },
2944 	{ zfs_ioc_pool_import, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2945 	{ zfs_ioc_pool_export, zfs_secpolicy_config, POOL_NAME, B_FALSE },
2946 	{ zfs_ioc_pool_configs,	zfs_secpolicy_none, NO_NAME, B_FALSE },
2947 	{ zfs_ioc_pool_stats, zfs_secpolicy_read, POOL_NAME, B_FALSE },
2948 	{ zfs_ioc_pool_tryimport, zfs_secpolicy_config, NO_NAME, B_FALSE },
2949 	{ zfs_ioc_pool_scrub, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2950 	{ zfs_ioc_pool_freeze, zfs_secpolicy_config, NO_NAME, B_FALSE },
2951 	{ zfs_ioc_pool_upgrade,	zfs_secpolicy_config, POOL_NAME, B_TRUE },
2952 	{ zfs_ioc_pool_get_history, zfs_secpolicy_config, POOL_NAME, B_FALSE },
2953 	{ zfs_ioc_vdev_add, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2954 	{ zfs_ioc_vdev_remove, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2955 	{ zfs_ioc_vdev_set_state, zfs_secpolicy_config,	POOL_NAME, B_TRUE },
2956 	{ zfs_ioc_vdev_attach, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2957 	{ zfs_ioc_vdev_detach, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2958 	{ zfs_ioc_vdev_setpath,	zfs_secpolicy_config, POOL_NAME, B_FALSE },
2959 	{ zfs_ioc_objset_stats,	zfs_secpolicy_read, DATASET_NAME, B_FALSE },
2960 	{ zfs_ioc_objset_zplprops, zfs_secpolicy_read, DATASET_NAME, B_FALSE },
2961 	{ zfs_ioc_dataset_list_next, zfs_secpolicy_read,
2962 	    DATASET_NAME, B_FALSE },
2963 	{ zfs_ioc_snapshot_list_next, zfs_secpolicy_read,
2964 	    DATASET_NAME, B_FALSE },
2965 	{ zfs_ioc_set_prop, zfs_secpolicy_none, DATASET_NAME, B_TRUE },
2966 	{ zfs_ioc_create_minor,	zfs_secpolicy_minor, DATASET_NAME, B_FALSE },
2967 	{ zfs_ioc_remove_minor,	zfs_secpolicy_minor, DATASET_NAME, B_FALSE },
2968 	{ zfs_ioc_create, zfs_secpolicy_create, DATASET_NAME, B_TRUE },
2969 	{ zfs_ioc_destroy, zfs_secpolicy_destroy, DATASET_NAME, B_TRUE },
2970 	{ zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, B_TRUE },
2971 	{ zfs_ioc_rename, zfs_secpolicy_rename,	DATASET_NAME, B_TRUE },
2972 	{ zfs_ioc_recv, zfs_secpolicy_receive, DATASET_NAME, B_TRUE },
2973 	{ zfs_ioc_send, zfs_secpolicy_send, DATASET_NAME, B_TRUE },
2974 	{ zfs_ioc_inject_fault,	zfs_secpolicy_inject, NO_NAME, B_FALSE },
2975 	{ zfs_ioc_clear_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE },
2976 	{ zfs_ioc_inject_list_next, zfs_secpolicy_inject, NO_NAME, B_FALSE },
2977 	{ zfs_ioc_error_log, zfs_secpolicy_inject, POOL_NAME, B_FALSE },
2978 	{ zfs_ioc_clear, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2979 	{ zfs_ioc_promote, zfs_secpolicy_promote, DATASET_NAME, B_TRUE },
2980 	{ zfs_ioc_destroy_snaps, zfs_secpolicy_destroy,	DATASET_NAME, B_TRUE },
2981 	{ zfs_ioc_snapshot, zfs_secpolicy_snapshot, DATASET_NAME, B_TRUE },
2982 	{ zfs_ioc_dsobj_to_dsname, zfs_secpolicy_config, POOL_NAME, B_FALSE },
2983 	{ zfs_ioc_obj_to_path, zfs_secpolicy_config, NO_NAME, B_FALSE },
2984 	{ zfs_ioc_pool_set_props, zfs_secpolicy_config,	POOL_NAME, B_TRUE },
2985 	{ zfs_ioc_pool_get_props, zfs_secpolicy_read, POOL_NAME, B_FALSE },
2986 	{ zfs_ioc_set_fsacl, zfs_secpolicy_fsacl, DATASET_NAME, B_TRUE },
2987 	{ zfs_ioc_get_fsacl, zfs_secpolicy_read, DATASET_NAME, B_FALSE },
2988 	{ zfs_ioc_iscsi_perm_check, zfs_secpolicy_iscsi,
2989 	    DATASET_NAME, B_FALSE },
2990 	{ zfs_ioc_share, zfs_secpolicy_share, DATASET_NAME, B_FALSE },
2991 	{ zfs_ioc_inherit_prop, zfs_secpolicy_inherit, DATASET_NAME, B_TRUE },
2992 };
2993 
2994 static int
2995 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
2996 {
2997 	zfs_cmd_t *zc;
2998 	uint_t vec;
2999 	int error, rc;
3000 
3001 	if (getminor(dev) != 0)
3002 		return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
3003 
3004 	vec = cmd - ZFS_IOC;
3005 	ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
3006 
3007 	if (vec >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
3008 		return (EINVAL);
3009 
3010 	zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
3011 
3012 	error = xcopyin((void *)arg, zc, sizeof (zfs_cmd_t));
3013 
3014 	if (error == 0)
3015 		error = zfs_ioc_vec[vec].zvec_secpolicy(zc, cr);
3016 
3017 	/*
3018 	 * Ensure that all pool/dataset names are valid before we pass down to
3019 	 * the lower layers.
3020 	 */
3021 	if (error == 0) {
3022 		zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
3023 		switch (zfs_ioc_vec[vec].zvec_namecheck) {
3024 		case POOL_NAME:
3025 			if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
3026 				error = EINVAL;
3027 			break;
3028 
3029 		case DATASET_NAME:
3030 			if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
3031 				error = EINVAL;
3032 			break;
3033 
3034 		case NO_NAME:
3035 			break;
3036 		}
3037 	}
3038 
3039 	if (error == 0)
3040 		error = zfs_ioc_vec[vec].zvec_func(zc);
3041 
3042 	rc = xcopyout(zc, (void *)arg, sizeof (zfs_cmd_t));
3043 	if (error == 0) {
3044 		error = rc;
3045 		if (zfs_ioc_vec[vec].zvec_his_log == B_TRUE)
3046 			zfs_log_history(zc);
3047 	}
3048 
3049 	kmem_free(zc, sizeof (zfs_cmd_t));
3050 	return (error);
3051 }
3052 
3053 static int
3054 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
3055 {
3056 	if (cmd != DDI_ATTACH)
3057 		return (DDI_FAILURE);
3058 
3059 	if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
3060 	    DDI_PSEUDO, 0) == DDI_FAILURE)
3061 		return (DDI_FAILURE);
3062 
3063 	zfs_dip = dip;
3064 
3065 	ddi_report_dev(dip);
3066 
3067 	return (DDI_SUCCESS);
3068 }
3069 
3070 static int
3071 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
3072 {
3073 	if (spa_busy() || zfs_busy() || zvol_busy())
3074 		return (DDI_FAILURE);
3075 
3076 	if (cmd != DDI_DETACH)
3077 		return (DDI_FAILURE);
3078 
3079 	zfs_dip = NULL;
3080 
3081 	ddi_prop_remove_all(dip);
3082 	ddi_remove_minor_node(dip, NULL);
3083 
3084 	return (DDI_SUCCESS);
3085 }
3086 
3087 /*ARGSUSED*/
3088 static int
3089 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
3090 {
3091 	switch (infocmd) {
3092 	case DDI_INFO_DEVT2DEVINFO:
3093 		*result = zfs_dip;
3094 		return (DDI_SUCCESS);
3095 
3096 	case DDI_INFO_DEVT2INSTANCE:
3097 		*result = (void *)0;
3098 		return (DDI_SUCCESS);
3099 	}
3100 
3101 	return (DDI_FAILURE);
3102 }
3103 
3104 /*
3105  * OK, so this is a little weird.
3106  *
3107  * /dev/zfs is the control node, i.e. minor 0.
3108  * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
3109  *
3110  * /dev/zfs has basically nothing to do except serve up ioctls,
3111  * so most of the standard driver entry points are in zvol.c.
3112  */
3113 static struct cb_ops zfs_cb_ops = {
3114 	zvol_open,	/* open */
3115 	zvol_close,	/* close */
3116 	zvol_strategy,	/* strategy */
3117 	nodev,		/* print */
3118 	zvol_dump,	/* dump */
3119 	zvol_read,	/* read */
3120 	zvol_write,	/* write */
3121 	zfsdev_ioctl,	/* ioctl */
3122 	nodev,		/* devmap */
3123 	nodev,		/* mmap */
3124 	nodev,		/* segmap */
3125 	nochpoll,	/* poll */
3126 	ddi_prop_op,	/* prop_op */
3127 	NULL,		/* streamtab */
3128 	D_NEW | D_MP | D_64BIT,		/* Driver compatibility flag */
3129 	CB_REV,		/* version */
3130 	nodev,		/* async read */
3131 	nodev,		/* async write */
3132 };
3133 
3134 static struct dev_ops zfs_dev_ops = {
3135 	DEVO_REV,	/* version */
3136 	0,		/* refcnt */
3137 	zfs_info,	/* info */
3138 	nulldev,	/* identify */
3139 	nulldev,	/* probe */
3140 	zfs_attach,	/* attach */
3141 	zfs_detach,	/* detach */
3142 	nodev,		/* reset */
3143 	&zfs_cb_ops,	/* driver operations */
3144 	NULL,		/* no bus operations */
3145 	NULL,		/* power */
3146 	ddi_quiesce_not_needed,	/* quiesce */
3147 };
3148 
3149 static struct modldrv zfs_modldrv = {
3150 	&mod_driverops,
3151 	"ZFS storage pool",
3152 	&zfs_dev_ops
3153 };
3154 
3155 static struct modlinkage modlinkage = {
3156 	MODREV_1,
3157 	(void *)&zfs_modlfs,
3158 	(void *)&zfs_modldrv,
3159 	NULL
3160 };
3161 
3162 
3163 uint_t zfs_fsyncer_key;
3164 extern uint_t rrw_tsd_key;
3165 
3166 int
3167 _init(void)
3168 {
3169 	int error;
3170 
3171 	spa_init(FREAD | FWRITE);
3172 	zfs_init();
3173 	zvol_init();
3174 
3175 	if ((error = mod_install(&modlinkage)) != 0) {
3176 		zvol_fini();
3177 		zfs_fini();
3178 		spa_fini();
3179 		return (error);
3180 	}
3181 
3182 	tsd_create(&zfs_fsyncer_key, NULL);
3183 	tsd_create(&rrw_tsd_key, NULL);
3184 
3185 	error = ldi_ident_from_mod(&modlinkage, &zfs_li);
3186 	ASSERT(error == 0);
3187 	mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
3188 
3189 	return (0);
3190 }
3191 
3192 int
3193 _fini(void)
3194 {
3195 	int error;
3196 
3197 	if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
3198 		return (EBUSY);
3199 
3200 	if ((error = mod_remove(&modlinkage)) != 0)
3201 		return (error);
3202 
3203 	zvol_fini();
3204 	zfs_fini();
3205 	spa_fini();
3206 	if (zfs_nfsshare_inited)
3207 		(void) ddi_modclose(nfs_mod);
3208 	if (zfs_smbshare_inited)
3209 		(void) ddi_modclose(smbsrv_mod);
3210 	if (zfs_nfsshare_inited || zfs_smbshare_inited)
3211 		(void) ddi_modclose(sharefs_mod);
3212 
3213 	tsd_destroy(&zfs_fsyncer_key);
3214 	ldi_ident_release(zfs_li);
3215 	zfs_li = NULL;
3216 	mutex_destroy(&zfs_share_lock);
3217 
3218 	return (error);
3219 }
3220 
3221 int
3222 _info(struct modinfo *modinfop)
3223 {
3224 	return (mod_info(&modlinkage, modinfop));
3225 }
3226