xref: /titanic_44/usr/src/uts/common/fs/zfs/zfs_ioctl.c (revision 193974072f41a843678abf5f61979c748687e66b)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 /*
22  * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 
27 #include <sys/types.h>
28 #include <sys/param.h>
29 #include <sys/errno.h>
30 #include <sys/uio.h>
31 #include <sys/buf.h>
32 #include <sys/modctl.h>
33 #include <sys/open.h>
34 #include <sys/file.h>
35 #include <sys/kmem.h>
36 #include <sys/conf.h>
37 #include <sys/cmn_err.h>
38 #include <sys/stat.h>
39 #include <sys/zfs_ioctl.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/vdev_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 struct zfs_ioc_vec {
84 	zfs_ioc_func_t		*zvec_func;
85 	zfs_secpolicy_func_t	*zvec_secpolicy;
86 	enum {
87 		NO_NAME,
88 		POOL_NAME,
89 		DATASET_NAME
90 	} zvec_namecheck;
91 	boolean_t		zvec_his_log;
92 } zfs_ioc_vec_t;
93 
94 static void clear_props(char *dataset, nvlist_t *props);
95 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
96     boolean_t *);
97 int zfs_set_prop_nvlist(const char *, nvlist_t *);
98 
99 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
100 void
101 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
102 {
103 	const char *newfile;
104 	char buf[256];
105 	va_list adx;
106 
107 	/*
108 	 * Get rid of annoying "../common/" prefix to filename.
109 	 */
110 	newfile = strrchr(file, '/');
111 	if (newfile != NULL) {
112 		newfile = newfile + 1; /* Get rid of leading / */
113 	} else {
114 		newfile = file;
115 	}
116 
117 	va_start(adx, fmt);
118 	(void) vsnprintf(buf, sizeof (buf), fmt, adx);
119 	va_end(adx);
120 
121 	/*
122 	 * To get this data, use the zfs-dprintf probe as so:
123 	 * dtrace -q -n 'zfs-dprintf \
124 	 *	/stringof(arg0) == "dbuf.c"/ \
125 	 *	{printf("%s: %s", stringof(arg1), stringof(arg3))}'
126 	 * arg0 = file name
127 	 * arg1 = function name
128 	 * arg2 = line number
129 	 * arg3 = message
130 	 */
131 	DTRACE_PROBE4(zfs__dprintf,
132 	    char *, newfile, char *, func, int, line, char *, buf);
133 }
134 
135 static void
136 history_str_free(char *buf)
137 {
138 	kmem_free(buf, HIS_MAX_RECORD_LEN);
139 }
140 
141 static char *
142 history_str_get(zfs_cmd_t *zc)
143 {
144 	char *buf;
145 
146 	if (zc->zc_history == NULL)
147 		return (NULL);
148 
149 	buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
150 	if (copyinstr((void *)(uintptr_t)zc->zc_history,
151 	    buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
152 		history_str_free(buf);
153 		return (NULL);
154 	}
155 
156 	buf[HIS_MAX_RECORD_LEN -1] = '\0';
157 
158 	return (buf);
159 }
160 
161 /*
162  * Check to see if the named dataset is currently defined as bootable
163  */
164 static boolean_t
165 zfs_is_bootfs(const char *name)
166 {
167 	spa_t *spa;
168 	boolean_t ret = B_FALSE;
169 
170 	if (spa_open(name, &spa, FTAG) == 0) {
171 		if (spa->spa_bootfs) {
172 			objset_t *os;
173 
174 			if (dmu_objset_open(name, DMU_OST_ZFS,
175 			    DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
176 				ret = (dmu_objset_id(os) == spa->spa_bootfs);
177 				dmu_objset_close(os);
178 			}
179 		}
180 		spa_close(spa, FTAG);
181 	}
182 	return (ret);
183 }
184 
185 /*
186  * zfs_earlier_version
187  *
188  *	Return non-zero if the spa version is less than requested version.
189  */
190 static int
191 zfs_earlier_version(const char *name, int version)
192 {
193 	spa_t *spa;
194 
195 	if (spa_open(name, &spa, FTAG) == 0) {
196 		if (spa_version(spa) < version) {
197 			spa_close(spa, FTAG);
198 			return (1);
199 		}
200 		spa_close(spa, FTAG);
201 	}
202 	return (0);
203 }
204 
205 /*
206  * zpl_earlier_version
207  *
208  * Return TRUE if the ZPL version is less than requested version.
209  */
210 static boolean_t
211 zpl_earlier_version(const char *name, int version)
212 {
213 	objset_t *os;
214 	boolean_t rc = B_TRUE;
215 
216 	if (dmu_objset_open(name, DMU_OST_ANY,
217 	    DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
218 		uint64_t zplversion;
219 
220 		if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
221 			rc = zplversion < version;
222 		dmu_objset_close(os);
223 	}
224 	return (rc);
225 }
226 
227 static void
228 zfs_log_history(zfs_cmd_t *zc)
229 {
230 	spa_t *spa;
231 	char *buf;
232 
233 	if ((buf = history_str_get(zc)) == NULL)
234 		return;
235 
236 	if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
237 		if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
238 			(void) spa_history_log(spa, buf, LOG_CMD_NORMAL);
239 		spa_close(spa, FTAG);
240 	}
241 	history_str_free(buf);
242 }
243 
244 /*
245  * Policy for top-level read operations (list pools).  Requires no privileges,
246  * and can be used in the local zone, as there is no associated dataset.
247  */
248 /* ARGSUSED */
249 static int
250 zfs_secpolicy_none(zfs_cmd_t *zc, cred_t *cr)
251 {
252 	return (0);
253 }
254 
255 /*
256  * Policy for dataset read operations (list children, get statistics).  Requires
257  * no privileges, but must be visible in the local zone.
258  */
259 /* ARGSUSED */
260 static int
261 zfs_secpolicy_read(zfs_cmd_t *zc, cred_t *cr)
262 {
263 	if (INGLOBALZONE(curproc) ||
264 	    zone_dataset_visible(zc->zc_name, NULL))
265 		return (0);
266 
267 	return (ENOENT);
268 }
269 
270 static int
271 zfs_dozonecheck(const char *dataset, cred_t *cr)
272 {
273 	uint64_t zoned;
274 	int writable = 1;
275 
276 	/*
277 	 * The dataset must be visible by this zone -- check this first
278 	 * so they don't see EPERM on something they shouldn't know about.
279 	 */
280 	if (!INGLOBALZONE(curproc) &&
281 	    !zone_dataset_visible(dataset, &writable))
282 		return (ENOENT);
283 
284 	if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
285 		return (ENOENT);
286 
287 	if (INGLOBALZONE(curproc)) {
288 		/*
289 		 * If the fs is zoned, only root can access it from the
290 		 * global zone.
291 		 */
292 		if (secpolicy_zfs(cr) && zoned)
293 			return (EPERM);
294 	} else {
295 		/*
296 		 * If we are in a local zone, the 'zoned' property must be set.
297 		 */
298 		if (!zoned)
299 			return (EPERM);
300 
301 		/* must be writable by this zone */
302 		if (!writable)
303 			return (EPERM);
304 	}
305 	return (0);
306 }
307 
308 int
309 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
310 {
311 	int error;
312 
313 	error = zfs_dozonecheck(name, cr);
314 	if (error == 0) {
315 		error = secpolicy_zfs(cr);
316 		if (error)
317 			error = dsl_deleg_access(name, perm, cr);
318 	}
319 	return (error);
320 }
321 
322 static int
323 zfs_secpolicy_setprop(const char *name, zfs_prop_t prop, cred_t *cr)
324 {
325 	/*
326 	 * Check permissions for special properties.
327 	 */
328 	switch (prop) {
329 	case ZFS_PROP_ZONED:
330 		/*
331 		 * Disallow setting of 'zoned' from within a local zone.
332 		 */
333 		if (!INGLOBALZONE(curproc))
334 			return (EPERM);
335 		break;
336 
337 	case ZFS_PROP_QUOTA:
338 		if (!INGLOBALZONE(curproc)) {
339 			uint64_t zoned;
340 			char setpoint[MAXNAMELEN];
341 			/*
342 			 * Unprivileged users are allowed to modify the
343 			 * quota on things *under* (ie. contained by)
344 			 * the thing they own.
345 			 */
346 			if (dsl_prop_get_integer(name, "zoned", &zoned,
347 			    setpoint))
348 				return (EPERM);
349 			if (!zoned || strlen(name) <= strlen(setpoint))
350 				return (EPERM);
351 		}
352 		break;
353 	}
354 
355 	return (zfs_secpolicy_write_perms(name, zfs_prop_to_name(prop), cr));
356 }
357 
358 int
359 zfs_secpolicy_fsacl(zfs_cmd_t *zc, cred_t *cr)
360 {
361 	int error;
362 
363 	error = zfs_dozonecheck(zc->zc_name, cr);
364 	if (error)
365 		return (error);
366 
367 	/*
368 	 * permission to set permissions will be evaluated later in
369 	 * dsl_deleg_can_allow()
370 	 */
371 	return (0);
372 }
373 
374 int
375 zfs_secpolicy_rollback(zfs_cmd_t *zc, cred_t *cr)
376 {
377 	int error;
378 	error = zfs_secpolicy_write_perms(zc->zc_name,
379 	    ZFS_DELEG_PERM_ROLLBACK, cr);
380 	if (error == 0)
381 		error = zfs_secpolicy_write_perms(zc->zc_name,
382 		    ZFS_DELEG_PERM_MOUNT, cr);
383 	return (error);
384 }
385 
386 int
387 zfs_secpolicy_send(zfs_cmd_t *zc, cred_t *cr)
388 {
389 	return (zfs_secpolicy_write_perms(zc->zc_name,
390 	    ZFS_DELEG_PERM_SEND, cr));
391 }
392 
393 int
394 zfs_secpolicy_share(zfs_cmd_t *zc, cred_t *cr)
395 {
396 	if (!INGLOBALZONE(curproc))
397 		return (EPERM);
398 
399 	if (secpolicy_nfs(cr) == 0) {
400 		return (0);
401 	} else {
402 		vnode_t *vp;
403 		int error;
404 
405 		if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
406 		    NO_FOLLOW, NULL, &vp)) != 0)
407 			return (error);
408 
409 		/* Now make sure mntpnt and dataset are ZFS */
410 
411 		if (vp->v_vfsp->vfs_fstype != zfsfstype ||
412 		    (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
413 		    zc->zc_name) != 0)) {
414 			VN_RELE(vp);
415 			return (EPERM);
416 		}
417 
418 		VN_RELE(vp);
419 		return (dsl_deleg_access(zc->zc_name,
420 		    ZFS_DELEG_PERM_SHARE, cr));
421 	}
422 }
423 
424 static int
425 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
426 {
427 	char *cp;
428 
429 	/*
430 	 * Remove the @bla or /bla from the end of the name to get the parent.
431 	 */
432 	(void) strncpy(parent, datasetname, parentsize);
433 	cp = strrchr(parent, '@');
434 	if (cp != NULL) {
435 		cp[0] = '\0';
436 	} else {
437 		cp = strrchr(parent, '/');
438 		if (cp == NULL)
439 			return (ENOENT);
440 		cp[0] = '\0';
441 	}
442 
443 	return (0);
444 }
445 
446 int
447 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
448 {
449 	int error;
450 
451 	if ((error = zfs_secpolicy_write_perms(name,
452 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
453 		return (error);
454 
455 	return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
456 }
457 
458 static int
459 zfs_secpolicy_destroy(zfs_cmd_t *zc, cred_t *cr)
460 {
461 	return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
462 }
463 
464 /*
465  * Must have sys_config privilege to check the iscsi permission
466  */
467 /* ARGSUSED */
468 static int
469 zfs_secpolicy_iscsi(zfs_cmd_t *zc, cred_t *cr)
470 {
471 	return (secpolicy_zfs(cr));
472 }
473 
474 int
475 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
476 {
477 	char 	parentname[MAXNAMELEN];
478 	int	error;
479 
480 	if ((error = zfs_secpolicy_write_perms(from,
481 	    ZFS_DELEG_PERM_RENAME, cr)) != 0)
482 		return (error);
483 
484 	if ((error = zfs_secpolicy_write_perms(from,
485 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
486 		return (error);
487 
488 	if ((error = zfs_get_parent(to, parentname,
489 	    sizeof (parentname))) != 0)
490 		return (error);
491 
492 	if ((error = zfs_secpolicy_write_perms(parentname,
493 	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
494 		return (error);
495 
496 	if ((error = zfs_secpolicy_write_perms(parentname,
497 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
498 		return (error);
499 
500 	return (error);
501 }
502 
503 static int
504 zfs_secpolicy_rename(zfs_cmd_t *zc, cred_t *cr)
505 {
506 	return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
507 }
508 
509 static int
510 zfs_secpolicy_promote(zfs_cmd_t *zc, cred_t *cr)
511 {
512 	char 	parentname[MAXNAMELEN];
513 	objset_t *clone;
514 	int error;
515 
516 	error = zfs_secpolicy_write_perms(zc->zc_name,
517 	    ZFS_DELEG_PERM_PROMOTE, cr);
518 	if (error)
519 		return (error);
520 
521 	error = dmu_objset_open(zc->zc_name, DMU_OST_ANY,
522 	    DS_MODE_USER | DS_MODE_READONLY, &clone);
523 
524 	if (error == 0) {
525 		dsl_dataset_t *pclone = NULL;
526 		dsl_dir_t *dd;
527 		dd = clone->os->os_dsl_dataset->ds_dir;
528 
529 		rw_enter(&dd->dd_pool->dp_config_rwlock, RW_READER);
530 		error = dsl_dataset_hold_obj(dd->dd_pool,
531 		    dd->dd_phys->dd_origin_obj, FTAG, &pclone);
532 		rw_exit(&dd->dd_pool->dp_config_rwlock);
533 		if (error) {
534 			dmu_objset_close(clone);
535 			return (error);
536 		}
537 
538 		error = zfs_secpolicy_write_perms(zc->zc_name,
539 		    ZFS_DELEG_PERM_MOUNT, cr);
540 
541 		dsl_dataset_name(pclone, parentname);
542 		dmu_objset_close(clone);
543 		dsl_dataset_rele(pclone, FTAG);
544 		if (error == 0)
545 			error = zfs_secpolicy_write_perms(parentname,
546 			    ZFS_DELEG_PERM_PROMOTE, cr);
547 	}
548 	return (error);
549 }
550 
551 static int
552 zfs_secpolicy_receive(zfs_cmd_t *zc, cred_t *cr)
553 {
554 	int error;
555 
556 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
557 	    ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
558 		return (error);
559 
560 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
561 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
562 		return (error);
563 
564 	return (zfs_secpolicy_write_perms(zc->zc_name,
565 	    ZFS_DELEG_PERM_CREATE, cr));
566 }
567 
568 int
569 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
570 {
571 	int error;
572 
573 	if ((error = zfs_secpolicy_write_perms(name,
574 	    ZFS_DELEG_PERM_SNAPSHOT, cr)) != 0)
575 		return (error);
576 
577 	error = zfs_secpolicy_write_perms(name,
578 	    ZFS_DELEG_PERM_MOUNT, cr);
579 
580 	return (error);
581 }
582 
583 static int
584 zfs_secpolicy_snapshot(zfs_cmd_t *zc, cred_t *cr)
585 {
586 
587 	return (zfs_secpolicy_snapshot_perms(zc->zc_name, cr));
588 }
589 
590 static int
591 zfs_secpolicy_create(zfs_cmd_t *zc, cred_t *cr)
592 {
593 	char 	parentname[MAXNAMELEN];
594 	int 	error;
595 
596 	if ((error = zfs_get_parent(zc->zc_name, parentname,
597 	    sizeof (parentname))) != 0)
598 		return (error);
599 
600 	if (zc->zc_value[0] != '\0') {
601 		if ((error = zfs_secpolicy_write_perms(zc->zc_value,
602 		    ZFS_DELEG_PERM_CLONE, cr)) != 0)
603 			return (error);
604 	}
605 
606 	if ((error = zfs_secpolicy_write_perms(parentname,
607 	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
608 		return (error);
609 
610 	error = zfs_secpolicy_write_perms(parentname,
611 	    ZFS_DELEG_PERM_MOUNT, cr);
612 
613 	return (error);
614 }
615 
616 static int
617 zfs_secpolicy_umount(zfs_cmd_t *zc, cred_t *cr)
618 {
619 	int error;
620 
621 	error = secpolicy_fs_unmount(cr, NULL);
622 	if (error) {
623 		error = dsl_deleg_access(zc->zc_name, ZFS_DELEG_PERM_MOUNT, cr);
624 	}
625 	return (error);
626 }
627 
628 /*
629  * Policy for pool operations - create/destroy pools, add vdevs, etc.  Requires
630  * SYS_CONFIG privilege, which is not available in a local zone.
631  */
632 /* ARGSUSED */
633 static int
634 zfs_secpolicy_config(zfs_cmd_t *zc, cred_t *cr)
635 {
636 	if (secpolicy_sys_config(cr, B_FALSE) != 0)
637 		return (EPERM);
638 
639 	return (0);
640 }
641 
642 /*
643  * Just like zfs_secpolicy_config, except that we will check for
644  * mount permission on the dataset for permission to create/remove
645  * the minor nodes.
646  */
647 static int
648 zfs_secpolicy_minor(zfs_cmd_t *zc, cred_t *cr)
649 {
650 	if (secpolicy_sys_config(cr, B_FALSE) != 0) {
651 		return (dsl_deleg_access(zc->zc_name,
652 		    ZFS_DELEG_PERM_MOUNT, cr));
653 	}
654 
655 	return (0);
656 }
657 
658 /*
659  * Policy for fault injection.  Requires all privileges.
660  */
661 /* ARGSUSED */
662 static int
663 zfs_secpolicy_inject(zfs_cmd_t *zc, cred_t *cr)
664 {
665 	return (secpolicy_zinject(cr));
666 }
667 
668 static int
669 zfs_secpolicy_inherit(zfs_cmd_t *zc, cred_t *cr)
670 {
671 	zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
672 
673 	if (prop == ZPROP_INVAL) {
674 		if (!zfs_prop_user(zc->zc_value))
675 			return (EINVAL);
676 		return (zfs_secpolicy_write_perms(zc->zc_name,
677 		    ZFS_DELEG_PERM_USERPROP, cr));
678 	} else {
679 		if (!zfs_prop_inheritable(prop))
680 			return (EINVAL);
681 		return (zfs_secpolicy_setprop(zc->zc_name, prop, cr));
682 	}
683 }
684 
685 /*
686  * Returns the nvlist as specified by the user in the zfs_cmd_t.
687  */
688 static int
689 get_nvlist(uint64_t nvl, uint64_t size, nvlist_t **nvp)
690 {
691 	char *packed;
692 	int error;
693 	nvlist_t *list = NULL;
694 
695 	/*
696 	 * Read in and unpack the user-supplied nvlist.
697 	 */
698 	if (size == 0)
699 		return (EINVAL);
700 
701 	packed = kmem_alloc(size, KM_SLEEP);
702 
703 	if ((error = xcopyin((void *)(uintptr_t)nvl, packed, size)) != 0) {
704 		kmem_free(packed, size);
705 		return (error);
706 	}
707 
708 	if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
709 		kmem_free(packed, size);
710 		return (error);
711 	}
712 
713 	kmem_free(packed, size);
714 
715 	*nvp = list;
716 	return (0);
717 }
718 
719 static int
720 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
721 {
722 	char *packed = NULL;
723 	size_t size;
724 	int error;
725 
726 	VERIFY(nvlist_size(nvl, &size, NV_ENCODE_NATIVE) == 0);
727 
728 	if (size > zc->zc_nvlist_dst_size) {
729 		error = ENOMEM;
730 	} else {
731 		packed = kmem_alloc(size, KM_SLEEP);
732 		VERIFY(nvlist_pack(nvl, &packed, &size, NV_ENCODE_NATIVE,
733 		    KM_SLEEP) == 0);
734 		error = xcopyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
735 		    size);
736 		kmem_free(packed, size);
737 	}
738 
739 	zc->zc_nvlist_dst_size = size;
740 	return (error);
741 }
742 
743 static int
744 zfs_ioc_pool_create(zfs_cmd_t *zc)
745 {
746 	int error;
747 	nvlist_t *config, *props = NULL;
748 	nvlist_t *rootprops = NULL;
749 	nvlist_t *zplprops = NULL;
750 	char *buf;
751 
752 	if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
753 	    &config))
754 		return (error);
755 
756 	if (zc->zc_nvlist_src_size != 0 && (error =
757 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, &props))) {
758 		nvlist_free(config);
759 		return (error);
760 	}
761 
762 	if (props) {
763 		nvlist_t *nvl = NULL;
764 		uint64_t version = SPA_VERSION;
765 
766 		(void) nvlist_lookup_uint64(props,
767 		    zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
768 		if (version < SPA_VERSION_INITIAL || version > SPA_VERSION) {
769 			error = EINVAL;
770 			goto pool_props_bad;
771 		}
772 		(void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
773 		if (nvl) {
774 			error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
775 			if (error != 0) {
776 				nvlist_free(config);
777 				nvlist_free(props);
778 				return (error);
779 			}
780 			(void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
781 		}
782 		VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
783 		error = zfs_fill_zplprops_root(version, rootprops,
784 		    zplprops, NULL);
785 		if (error)
786 			goto pool_props_bad;
787 	}
788 
789 	buf = history_str_get(zc);
790 
791 	error = spa_create(zc->zc_name, config, props, buf, zplprops);
792 
793 	/*
794 	 * Set the remaining root properties
795 	 */
796 	if (!error &&
797 	    (error = zfs_set_prop_nvlist(zc->zc_name, rootprops)) != 0)
798 		(void) spa_destroy(zc->zc_name);
799 
800 	if (buf != NULL)
801 		history_str_free(buf);
802 
803 pool_props_bad:
804 	nvlist_free(rootprops);
805 	nvlist_free(zplprops);
806 	nvlist_free(config);
807 	nvlist_free(props);
808 
809 	return (error);
810 }
811 
812 static int
813 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
814 {
815 	int error;
816 	zfs_log_history(zc);
817 	error = spa_destroy(zc->zc_name);
818 	return (error);
819 }
820 
821 static int
822 zfs_ioc_pool_import(zfs_cmd_t *zc)
823 {
824 	int error;
825 	nvlist_t *config, *props = NULL;
826 	uint64_t guid;
827 
828 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
829 	    &config)) != 0)
830 		return (error);
831 
832 	if (zc->zc_nvlist_src_size != 0 && (error =
833 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, &props))) {
834 		nvlist_free(config);
835 		return (error);
836 	}
837 
838 	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
839 	    guid != zc->zc_guid)
840 		error = EINVAL;
841 	else if (zc->zc_cookie)
842 		error = spa_import_faulted(zc->zc_name, config,
843 		    props);
844 	else
845 		error = spa_import(zc->zc_name, config, props);
846 
847 	nvlist_free(config);
848 
849 	if (props)
850 		nvlist_free(props);
851 
852 	return (error);
853 }
854 
855 static int
856 zfs_ioc_pool_export(zfs_cmd_t *zc)
857 {
858 	int error;
859 	boolean_t force = (boolean_t)zc->zc_cookie;
860 
861 	zfs_log_history(zc);
862 	error = spa_export(zc->zc_name, NULL, force);
863 	return (error);
864 }
865 
866 static int
867 zfs_ioc_pool_configs(zfs_cmd_t *zc)
868 {
869 	nvlist_t *configs;
870 	int error;
871 
872 	if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
873 		return (EEXIST);
874 
875 	error = put_nvlist(zc, configs);
876 
877 	nvlist_free(configs);
878 
879 	return (error);
880 }
881 
882 static int
883 zfs_ioc_pool_stats(zfs_cmd_t *zc)
884 {
885 	nvlist_t *config;
886 	int error;
887 	int ret = 0;
888 
889 	error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
890 	    sizeof (zc->zc_value));
891 
892 	if (config != NULL) {
893 		ret = put_nvlist(zc, config);
894 		nvlist_free(config);
895 
896 		/*
897 		 * The config may be present even if 'error' is non-zero.
898 		 * In this case we return success, and preserve the real errno
899 		 * in 'zc_cookie'.
900 		 */
901 		zc->zc_cookie = error;
902 	} else {
903 		ret = error;
904 	}
905 
906 	return (ret);
907 }
908 
909 /*
910  * Try to import the given pool, returning pool stats as appropriate so that
911  * user land knows which devices are available and overall pool health.
912  */
913 static int
914 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
915 {
916 	nvlist_t *tryconfig, *config;
917 	int error;
918 
919 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
920 	    &tryconfig)) != 0)
921 		return (error);
922 
923 	config = spa_tryimport(tryconfig);
924 
925 	nvlist_free(tryconfig);
926 
927 	if (config == NULL)
928 		return (EINVAL);
929 
930 	error = put_nvlist(zc, config);
931 	nvlist_free(config);
932 
933 	return (error);
934 }
935 
936 static int
937 zfs_ioc_pool_scrub(zfs_cmd_t *zc)
938 {
939 	spa_t *spa;
940 	int error;
941 
942 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
943 		return (error);
944 
945 	error = spa_scrub(spa, zc->zc_cookie);
946 
947 	spa_close(spa, FTAG);
948 
949 	return (error);
950 }
951 
952 static int
953 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
954 {
955 	spa_t *spa;
956 	int error;
957 
958 	error = spa_open(zc->zc_name, &spa, FTAG);
959 	if (error == 0) {
960 		spa_freeze(spa);
961 		spa_close(spa, FTAG);
962 	}
963 	return (error);
964 }
965 
966 static int
967 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
968 {
969 	spa_t *spa;
970 	int error;
971 
972 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
973 		return (error);
974 
975 	if (zc->zc_cookie < spa_version(spa) || zc->zc_cookie > SPA_VERSION) {
976 		spa_close(spa, FTAG);
977 		return (EINVAL);
978 	}
979 
980 	spa_upgrade(spa, zc->zc_cookie);
981 	spa_close(spa, FTAG);
982 
983 	return (error);
984 }
985 
986 static int
987 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
988 {
989 	spa_t *spa;
990 	char *hist_buf;
991 	uint64_t size;
992 	int error;
993 
994 	if ((size = zc->zc_history_len) == 0)
995 		return (EINVAL);
996 
997 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
998 		return (error);
999 
1000 	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1001 		spa_close(spa, FTAG);
1002 		return (ENOTSUP);
1003 	}
1004 
1005 	hist_buf = kmem_alloc(size, KM_SLEEP);
1006 	if ((error = spa_history_get(spa, &zc->zc_history_offset,
1007 	    &zc->zc_history_len, hist_buf)) == 0) {
1008 		error = xcopyout(hist_buf,
1009 		    (char *)(uintptr_t)zc->zc_history,
1010 		    zc->zc_history_len);
1011 	}
1012 
1013 	spa_close(spa, FTAG);
1014 	kmem_free(hist_buf, size);
1015 	return (error);
1016 }
1017 
1018 static int
1019 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1020 {
1021 	int error;
1022 
1023 	if (error = dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value))
1024 		return (error);
1025 
1026 	return (0);
1027 }
1028 
1029 static int
1030 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1031 {
1032 	objset_t *osp;
1033 	int error;
1034 
1035 	if ((error = dmu_objset_open(zc->zc_name, DMU_OST_ZFS,
1036 	    DS_MODE_USER | DS_MODE_READONLY, &osp)) != 0)
1037 		return (error);
1038 	error = zfs_obj_to_path(osp, zc->zc_obj, zc->zc_value,
1039 	    sizeof (zc->zc_value));
1040 	dmu_objset_close(osp);
1041 
1042 	return (error);
1043 }
1044 
1045 static int
1046 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1047 {
1048 	spa_t *spa;
1049 	int error;
1050 	nvlist_t *config, **l2cache, **spares;
1051 	uint_t nl2cache = 0, nspares = 0;
1052 
1053 	error = spa_open(zc->zc_name, &spa, FTAG);
1054 	if (error != 0)
1055 		return (error);
1056 
1057 	error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1058 	    &config);
1059 	(void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1060 	    &l2cache, &nl2cache);
1061 
1062 	(void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1063 	    &spares, &nspares);
1064 
1065 	/*
1066 	 * A root pool with concatenated devices is not supported.
1067 	 * Thus, can not add a device to a root pool.
1068 	 *
1069 	 * Intent log device can not be added to a rootpool because
1070 	 * during mountroot, zil is replayed, a seperated log device
1071 	 * can not be accessed during the mountroot time.
1072 	 *
1073 	 * l2cache and spare devices are ok to be added to a rootpool.
1074 	 */
1075 	if (spa->spa_bootfs != 0 && nl2cache == 0 && nspares == 0) {
1076 		spa_close(spa, FTAG);
1077 		return (EDOM);
1078 	}
1079 
1080 	if (error == 0) {
1081 		error = spa_vdev_add(spa, config);
1082 		nvlist_free(config);
1083 	}
1084 	spa_close(spa, FTAG);
1085 	return (error);
1086 }
1087 
1088 static int
1089 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1090 {
1091 	spa_t *spa;
1092 	int error;
1093 
1094 	error = spa_open(zc->zc_name, &spa, FTAG);
1095 	if (error != 0)
1096 		return (error);
1097 	error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1098 	spa_close(spa, FTAG);
1099 	return (error);
1100 }
1101 
1102 static int
1103 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1104 {
1105 	spa_t *spa;
1106 	int error;
1107 	vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1108 
1109 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1110 		return (error);
1111 	switch (zc->zc_cookie) {
1112 	case VDEV_STATE_ONLINE:
1113 		error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1114 		break;
1115 
1116 	case VDEV_STATE_OFFLINE:
1117 		error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1118 		break;
1119 
1120 	case VDEV_STATE_FAULTED:
1121 		error = vdev_fault(spa, zc->zc_guid);
1122 		break;
1123 
1124 	case VDEV_STATE_DEGRADED:
1125 		error = vdev_degrade(spa, zc->zc_guid);
1126 		break;
1127 
1128 	default:
1129 		error = EINVAL;
1130 	}
1131 	zc->zc_cookie = newstate;
1132 	spa_close(spa, FTAG);
1133 	return (error);
1134 }
1135 
1136 static int
1137 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1138 {
1139 	spa_t *spa;
1140 	int replacing = zc->zc_cookie;
1141 	nvlist_t *config;
1142 	int error;
1143 
1144 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1145 		return (error);
1146 
1147 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1148 	    &config)) == 0) {
1149 		error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1150 		nvlist_free(config);
1151 	}
1152 
1153 	spa_close(spa, FTAG);
1154 	return (error);
1155 }
1156 
1157 static int
1158 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1159 {
1160 	spa_t *spa;
1161 	int error;
1162 
1163 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1164 		return (error);
1165 
1166 	error = spa_vdev_detach(spa, zc->zc_guid, B_FALSE);
1167 
1168 	spa_close(spa, FTAG);
1169 	return (error);
1170 }
1171 
1172 static int
1173 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
1174 {
1175 	spa_t *spa;
1176 	char *path = zc->zc_value;
1177 	uint64_t guid = zc->zc_guid;
1178 	int error;
1179 
1180 	error = spa_open(zc->zc_name, &spa, FTAG);
1181 	if (error != 0)
1182 		return (error);
1183 
1184 	error = spa_vdev_setpath(spa, guid, path);
1185 	spa_close(spa, FTAG);
1186 	return (error);
1187 }
1188 
1189 /*
1190  * inputs:
1191  * zc_name		name of filesystem
1192  * zc_nvlist_dst_size	size of buffer for property nvlist
1193  *
1194  * outputs:
1195  * zc_objset_stats	stats
1196  * zc_nvlist_dst	property nvlist
1197  * zc_nvlist_dst_size	size of property nvlist
1198  */
1199 static int
1200 zfs_ioc_objset_stats(zfs_cmd_t *zc)
1201 {
1202 	objset_t *os = NULL;
1203 	int error;
1204 	nvlist_t *nv;
1205 
1206 	if (error = dmu_objset_open(zc->zc_name,
1207 	    DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os))
1208 		return (error);
1209 
1210 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1211 
1212 	if (zc->zc_nvlist_dst != 0 &&
1213 	    (error = dsl_prop_get_all(os, &nv, FALSE)) == 0) {
1214 		dmu_objset_stats(os, nv);
1215 		/*
1216 		 * NB: zvol_get_stats() will read the objset contents,
1217 		 * which we aren't supposed to do with a
1218 		 * DS_MODE_USER hold, because it could be
1219 		 * inconsistent.  So this is a bit of a workaround...
1220 		 */
1221 		if (!zc->zc_objset_stats.dds_inconsistent) {
1222 			if (dmu_objset_type(os) == DMU_OST_ZVOL)
1223 				VERIFY(zvol_get_stats(os, nv) == 0);
1224 		}
1225 		error = put_nvlist(zc, nv);
1226 		nvlist_free(nv);
1227 	}
1228 
1229 	dmu_objset_close(os);
1230 	return (error);
1231 }
1232 
1233 static int
1234 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
1235 {
1236 	uint64_t value;
1237 	int error;
1238 
1239 	/*
1240 	 * zfs_get_zplprop() will either find a value or give us
1241 	 * the default value (if there is one).
1242 	 */
1243 	if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
1244 		return (error);
1245 	VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
1246 	return (0);
1247 }
1248 
1249 /*
1250  * inputs:
1251  * zc_name		name of filesystem
1252  * zc_nvlist_dst_size	size of buffer for zpl property nvlist
1253  *
1254  * outputs:
1255  * zc_nvlist_dst	zpl property nvlist
1256  * zc_nvlist_dst_size	size of zpl property nvlist
1257  */
1258 static int
1259 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
1260 {
1261 	objset_t *os;
1262 	int err;
1263 
1264 	if (err = dmu_objset_open(zc->zc_name,
1265 	    DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os))
1266 		return (err);
1267 
1268 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1269 
1270 	/*
1271 	 * NB: nvl_add_zplprop() will read the objset contents,
1272 	 * which we aren't supposed to do with a DS_MODE_USER
1273 	 * hold, because it could be inconsistent.
1274 	 */
1275 	if (zc->zc_nvlist_dst != NULL &&
1276 	    !zc->zc_objset_stats.dds_inconsistent &&
1277 	    dmu_objset_type(os) == DMU_OST_ZFS) {
1278 		nvlist_t *nv;
1279 
1280 		VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1281 		if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
1282 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
1283 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
1284 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
1285 			err = put_nvlist(zc, nv);
1286 		nvlist_free(nv);
1287 	} else {
1288 		err = ENOENT;
1289 	}
1290 	dmu_objset_close(os);
1291 	return (err);
1292 }
1293 
1294 /*
1295  * inputs:
1296  * zc_name		name of filesystem
1297  * zc_cookie		zap cursor
1298  * zc_nvlist_dst_size	size of buffer for property nvlist
1299  *
1300  * outputs:
1301  * zc_name		name of next filesystem
1302  * zc_objset_stats	stats
1303  * zc_nvlist_dst	property nvlist
1304  * zc_nvlist_dst_size	size of property nvlist
1305  */
1306 static int
1307 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
1308 {
1309 	objset_t *os;
1310 	int error;
1311 	char *p;
1312 
1313 	if (error = dmu_objset_open(zc->zc_name,
1314 	    DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os)) {
1315 		if (error == ENOENT)
1316 			error = ESRCH;
1317 		return (error);
1318 	}
1319 
1320 	p = strrchr(zc->zc_name, '/');
1321 	if (p == NULL || p[1] != '\0')
1322 		(void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
1323 	p = zc->zc_name + strlen(zc->zc_name);
1324 
1325 	do {
1326 		error = dmu_dir_list_next(os,
1327 		    sizeof (zc->zc_name) - (p - zc->zc_name), p,
1328 		    NULL, &zc->zc_cookie);
1329 		if (error == ENOENT)
1330 			error = ESRCH;
1331 	} while (error == 0 && !INGLOBALZONE(curproc) &&
1332 	    !zone_dataset_visible(zc->zc_name, NULL));
1333 	dmu_objset_close(os);
1334 
1335 	/*
1336 	 * If it's a hidden dataset (ie. with a '$' in its name), don't
1337 	 * try to get stats for it.  Userland will skip over it.
1338 	 */
1339 	if (error == 0 && strchr(zc->zc_name, '$') == NULL)
1340 		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
1341 
1342 	return (error);
1343 }
1344 
1345 /*
1346  * inputs:
1347  * zc_name		name of filesystem
1348  * zc_cookie		zap cursor
1349  * zc_nvlist_dst_size	size of buffer for property nvlist
1350  *
1351  * outputs:
1352  * zc_name		name of next snapshot
1353  * zc_objset_stats	stats
1354  * zc_nvlist_dst	property nvlist
1355  * zc_nvlist_dst_size	size of property nvlist
1356  */
1357 static int
1358 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
1359 {
1360 	objset_t *os;
1361 	int error;
1362 
1363 	error = dmu_objset_open(zc->zc_name,
1364 	    DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os);
1365 	if (error)
1366 		return (error == ENOENT ? ESRCH : error);
1367 
1368 	/*
1369 	 * A dataset name of maximum length cannot have any snapshots,
1370 	 * so exit immediately.
1371 	 */
1372 	if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
1373 		dmu_objset_close(os);
1374 		return (ESRCH);
1375 	}
1376 
1377 	error = dmu_snapshot_list_next(os,
1378 	    sizeof (zc->zc_name) - strlen(zc->zc_name),
1379 	    zc->zc_name + strlen(zc->zc_name), NULL, &zc->zc_cookie, NULL);
1380 	dmu_objset_close(os);
1381 	if (error == 0)
1382 		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
1383 	else if (error == ENOENT)
1384 		error = ESRCH;
1385 
1386 	/* if we failed, undo the @ that we tacked on to zc_name */
1387 	if (error)
1388 		*strchr(zc->zc_name, '@') = '\0';
1389 	return (error);
1390 }
1391 
1392 int
1393 zfs_set_prop_nvlist(const char *name, nvlist_t *nvl)
1394 {
1395 	nvpair_t *elem;
1396 	int error;
1397 	uint64_t intval;
1398 	char *strval;
1399 
1400 	/*
1401 	 * First validate permission to set all of the properties
1402 	 */
1403 	elem = NULL;
1404 	while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) {
1405 		const char *propname = nvpair_name(elem);
1406 		zfs_prop_t prop = zfs_name_to_prop(propname);
1407 
1408 		if (prop == ZPROP_INVAL) {
1409 			/*
1410 			 * If this is a user-defined property, it must be a
1411 			 * string, and there is no further validation to do.
1412 			 */
1413 			if (!zfs_prop_user(propname) ||
1414 			    nvpair_type(elem) != DATA_TYPE_STRING)
1415 				return (EINVAL);
1416 
1417 			if (error = zfs_secpolicy_write_perms(name,
1418 			    ZFS_DELEG_PERM_USERPROP, CRED()))
1419 				return (error);
1420 			continue;
1421 		}
1422 
1423 		if ((error = zfs_secpolicy_setprop(name, prop, CRED())) != 0)
1424 			return (error);
1425 
1426 		/*
1427 		 * Check that this value is valid for this pool version
1428 		 */
1429 		switch (prop) {
1430 		case ZFS_PROP_COMPRESSION:
1431 			/*
1432 			 * If the user specified gzip compression, make sure
1433 			 * the SPA supports it. We ignore any errors here since
1434 			 * we'll catch them later.
1435 			 */
1436 			if (nvpair_type(elem) == DATA_TYPE_UINT64 &&
1437 			    nvpair_value_uint64(elem, &intval) == 0) {
1438 				if (intval >= ZIO_COMPRESS_GZIP_1 &&
1439 				    intval <= ZIO_COMPRESS_GZIP_9 &&
1440 				    zfs_earlier_version(name,
1441 				    SPA_VERSION_GZIP_COMPRESSION))
1442 					return (ENOTSUP);
1443 
1444 				/*
1445 				 * If this is a bootable dataset then
1446 				 * verify that the compression algorithm
1447 				 * is supported for booting. We must return
1448 				 * something other than ENOTSUP since it
1449 				 * implies a downrev pool version.
1450 				 */
1451 				if (zfs_is_bootfs(name) &&
1452 				    !BOOTFS_COMPRESS_VALID(intval))
1453 					return (ERANGE);
1454 			}
1455 			break;
1456 
1457 		case ZFS_PROP_COPIES:
1458 			if (zfs_earlier_version(name,
1459 			    SPA_VERSION_DITTO_BLOCKS))
1460 				return (ENOTSUP);
1461 			break;
1462 
1463 		case ZFS_PROP_SHARESMB:
1464 			if (zpl_earlier_version(name, ZPL_VERSION_FUID))
1465 				return (ENOTSUP);
1466 			break;
1467 		}
1468 	}
1469 
1470 	elem = NULL;
1471 	while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) {
1472 		const char *propname = nvpair_name(elem);
1473 		zfs_prop_t prop = zfs_name_to_prop(propname);
1474 
1475 		if (prop == ZPROP_INVAL) {
1476 			VERIFY(nvpair_value_string(elem, &strval) == 0);
1477 			error = dsl_prop_set(name, propname, 1,
1478 			    strlen(strval) + 1, strval);
1479 			if (error == 0)
1480 				continue;
1481 			else
1482 				return (error);
1483 		}
1484 
1485 		switch (prop) {
1486 		case ZFS_PROP_QUOTA:
1487 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1488 			    (error = dsl_dir_set_quota(name, intval)) != 0)
1489 				return (error);
1490 			break;
1491 
1492 		case ZFS_PROP_REFQUOTA:
1493 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1494 			    (error = dsl_dataset_set_quota(name, intval)) != 0)
1495 				return (error);
1496 			break;
1497 
1498 		case ZFS_PROP_RESERVATION:
1499 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1500 			    (error = dsl_dir_set_reservation(name,
1501 			    intval)) != 0)
1502 				return (error);
1503 			break;
1504 
1505 		case ZFS_PROP_REFRESERVATION:
1506 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1507 			    (error = dsl_dataset_set_reservation(name,
1508 			    intval)) != 0)
1509 				return (error);
1510 			break;
1511 
1512 		case ZFS_PROP_VOLSIZE:
1513 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1514 			    (error = zvol_set_volsize(name,
1515 			    ddi_driver_major(zfs_dip), intval)) != 0)
1516 				return (error);
1517 			break;
1518 
1519 		case ZFS_PROP_VOLBLOCKSIZE:
1520 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1521 			    (error = zvol_set_volblocksize(name, intval)) != 0)
1522 				return (error);
1523 			break;
1524 
1525 		case ZFS_PROP_VERSION:
1526 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1527 			    (error = zfs_set_version(name, intval)) != 0)
1528 				return (error);
1529 			break;
1530 
1531 		default:
1532 			if (nvpair_type(elem) == DATA_TYPE_STRING) {
1533 				if (zfs_prop_get_type(prop) !=
1534 				    PROP_TYPE_STRING)
1535 					return (EINVAL);
1536 				VERIFY(nvpair_value_string(elem, &strval) == 0);
1537 				if ((error = dsl_prop_set(name,
1538 				    nvpair_name(elem), 1, strlen(strval) + 1,
1539 				    strval)) != 0)
1540 					return (error);
1541 			} else if (nvpair_type(elem) == DATA_TYPE_UINT64) {
1542 				const char *unused;
1543 
1544 				VERIFY(nvpair_value_uint64(elem, &intval) == 0);
1545 
1546 				switch (zfs_prop_get_type(prop)) {
1547 				case PROP_TYPE_NUMBER:
1548 					break;
1549 				case PROP_TYPE_STRING:
1550 					return (EINVAL);
1551 				case PROP_TYPE_INDEX:
1552 					if (zfs_prop_index_to_string(prop,
1553 					    intval, &unused) != 0)
1554 						return (EINVAL);
1555 					break;
1556 				default:
1557 					cmn_err(CE_PANIC,
1558 					    "unknown property type");
1559 					break;
1560 				}
1561 
1562 				if ((error = dsl_prop_set(name, propname,
1563 				    8, 1, &intval)) != 0)
1564 					return (error);
1565 			} else {
1566 				return (EINVAL);
1567 			}
1568 			break;
1569 		}
1570 	}
1571 
1572 	return (0);
1573 }
1574 
1575 /*
1576  * inputs:
1577  * zc_name		name of filesystem
1578  * zc_value		name of property to inherit
1579  * zc_nvlist_src{_size}	nvlist of properties to apply
1580  * zc_cookie		clear existing local props?
1581  *
1582  * outputs:		none
1583  */
1584 static int
1585 zfs_ioc_set_prop(zfs_cmd_t *zc)
1586 {
1587 	nvlist_t *nvl;
1588 	int error;
1589 
1590 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1591 	    &nvl)) != 0)
1592 		return (error);
1593 
1594 	if (zc->zc_cookie) {
1595 		nvlist_t *origprops;
1596 		objset_t *os;
1597 
1598 		if (dmu_objset_open(zc->zc_name, DMU_OST_ANY,
1599 		    DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
1600 			if (dsl_prop_get_all(os, &origprops, TRUE) == 0) {
1601 				clear_props(zc->zc_name, origprops);
1602 				nvlist_free(origprops);
1603 			}
1604 			dmu_objset_close(os);
1605 		}
1606 
1607 	}
1608 
1609 	error = zfs_set_prop_nvlist(zc->zc_name, nvl);
1610 
1611 	nvlist_free(nvl);
1612 	return (error);
1613 }
1614 
1615 /*
1616  * inputs:
1617  * zc_name		name of filesystem
1618  * zc_value		name of property to inherit
1619  *
1620  * outputs:		none
1621  */
1622 static int
1623 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
1624 {
1625 	/* the property name has been validated by zfs_secpolicy_inherit() */
1626 	return (dsl_prop_set(zc->zc_name, zc->zc_value, 0, 0, NULL));
1627 }
1628 
1629 static int
1630 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
1631 {
1632 	nvlist_t *props;
1633 	spa_t *spa;
1634 	int error;
1635 
1636 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1637 	    &props)))
1638 		return (error);
1639 
1640 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
1641 		nvlist_free(props);
1642 		return (error);
1643 	}
1644 
1645 	error = spa_prop_set(spa, props);
1646 
1647 	nvlist_free(props);
1648 	spa_close(spa, FTAG);
1649 
1650 	return (error);
1651 }
1652 
1653 static int
1654 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
1655 {
1656 	spa_t *spa;
1657 	int error;
1658 	nvlist_t *nvp = NULL;
1659 
1660 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1661 		return (error);
1662 
1663 	error = spa_prop_get(spa, &nvp);
1664 
1665 	if (error == 0 && zc->zc_nvlist_dst != NULL)
1666 		error = put_nvlist(zc, nvp);
1667 	else
1668 		error = EFAULT;
1669 
1670 	spa_close(spa, FTAG);
1671 
1672 	if (nvp)
1673 		nvlist_free(nvp);
1674 	return (error);
1675 }
1676 
1677 static int
1678 zfs_ioc_iscsi_perm_check(zfs_cmd_t *zc)
1679 {
1680 	nvlist_t *nvp;
1681 	int error;
1682 	uint32_t uid;
1683 	uint32_t gid;
1684 	uint32_t *groups;
1685 	uint_t group_cnt;
1686 	cred_t	*usercred;
1687 
1688 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1689 	    &nvp)) != 0) {
1690 		return (error);
1691 	}
1692 
1693 	if ((error = nvlist_lookup_uint32(nvp,
1694 	    ZFS_DELEG_PERM_UID, &uid)) != 0) {
1695 		nvlist_free(nvp);
1696 		return (EPERM);
1697 	}
1698 
1699 	if ((error = nvlist_lookup_uint32(nvp,
1700 	    ZFS_DELEG_PERM_GID, &gid)) != 0) {
1701 		nvlist_free(nvp);
1702 		return (EPERM);
1703 	}
1704 
1705 	if ((error = nvlist_lookup_uint32_array(nvp, ZFS_DELEG_PERM_GROUPS,
1706 	    &groups, &group_cnt)) != 0) {
1707 		nvlist_free(nvp);
1708 		return (EPERM);
1709 	}
1710 	usercred = cralloc();
1711 	if ((crsetugid(usercred, uid, gid) != 0) ||
1712 	    (crsetgroups(usercred, group_cnt, (gid_t *)groups) != 0)) {
1713 		nvlist_free(nvp);
1714 		crfree(usercred);
1715 		return (EPERM);
1716 	}
1717 	nvlist_free(nvp);
1718 	error = dsl_deleg_access(zc->zc_name,
1719 	    zfs_prop_to_name(ZFS_PROP_SHAREISCSI), usercred);
1720 	crfree(usercred);
1721 	return (error);
1722 }
1723 
1724 /*
1725  * inputs:
1726  * zc_name		name of filesystem
1727  * zc_nvlist_src{_size}	nvlist of delegated permissions
1728  * zc_perm_action	allow/unallow flag
1729  *
1730  * outputs:		none
1731  */
1732 static int
1733 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
1734 {
1735 	int error;
1736 	nvlist_t *fsaclnv = NULL;
1737 
1738 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1739 	    &fsaclnv)) != 0)
1740 		return (error);
1741 
1742 	/*
1743 	 * Verify nvlist is constructed correctly
1744 	 */
1745 	if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
1746 		nvlist_free(fsaclnv);
1747 		return (EINVAL);
1748 	}
1749 
1750 	/*
1751 	 * If we don't have PRIV_SYS_MOUNT, then validate
1752 	 * that user is allowed to hand out each permission in
1753 	 * the nvlist(s)
1754 	 */
1755 
1756 	error = secpolicy_zfs(CRED());
1757 	if (error) {
1758 		if (zc->zc_perm_action == B_FALSE) {
1759 			error = dsl_deleg_can_allow(zc->zc_name,
1760 			    fsaclnv, CRED());
1761 		} else {
1762 			error = dsl_deleg_can_unallow(zc->zc_name,
1763 			    fsaclnv, CRED());
1764 		}
1765 	}
1766 
1767 	if (error == 0)
1768 		error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
1769 
1770 	nvlist_free(fsaclnv);
1771 	return (error);
1772 }
1773 
1774 /*
1775  * inputs:
1776  * zc_name		name of filesystem
1777  *
1778  * outputs:
1779  * zc_nvlist_src{_size}	nvlist of delegated permissions
1780  */
1781 static int
1782 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
1783 {
1784 	nvlist_t *nvp;
1785 	int error;
1786 
1787 	if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
1788 		error = put_nvlist(zc, nvp);
1789 		nvlist_free(nvp);
1790 	}
1791 
1792 	return (error);
1793 }
1794 
1795 /*
1796  * inputs:
1797  * zc_name		name of volume
1798  *
1799  * outputs:		none
1800  */
1801 static int
1802 zfs_ioc_create_minor(zfs_cmd_t *zc)
1803 {
1804 	return (zvol_create_minor(zc->zc_name, ddi_driver_major(zfs_dip)));
1805 }
1806 
1807 /*
1808  * inputs:
1809  * zc_name		name of volume
1810  *
1811  * outputs:		none
1812  */
1813 static int
1814 zfs_ioc_remove_minor(zfs_cmd_t *zc)
1815 {
1816 	return (zvol_remove_minor(zc->zc_name));
1817 }
1818 
1819 /*
1820  * Search the vfs list for a specified resource.  Returns a pointer to it
1821  * or NULL if no suitable entry is found. The caller of this routine
1822  * is responsible for releasing the returned vfs pointer.
1823  */
1824 static vfs_t *
1825 zfs_get_vfs(const char *resource)
1826 {
1827 	struct vfs *vfsp;
1828 	struct vfs *vfs_found = NULL;
1829 
1830 	vfs_list_read_lock();
1831 	vfsp = rootvfs;
1832 	do {
1833 		if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
1834 			VFS_HOLD(vfsp);
1835 			vfs_found = vfsp;
1836 			break;
1837 		}
1838 		vfsp = vfsp->vfs_next;
1839 	} while (vfsp != rootvfs);
1840 	vfs_list_unlock();
1841 	return (vfs_found);
1842 }
1843 
1844 /* ARGSUSED */
1845 static void
1846 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
1847 {
1848 	zfs_creat_t *zct = arg;
1849 
1850 	zfs_create_fs(os, cr, zct->zct_zplprops, tx);
1851 }
1852 
1853 #define	ZFS_PROP_UNDEFINED	((uint64_t)-1)
1854 
1855 /*
1856  * inputs:
1857  * createprops		list of properties requested by creator
1858  * default_zplver	zpl version to use if unspecified in createprops
1859  * fuids_ok		fuids allowed in this version of the spa?
1860  * os			parent objset pointer (NULL if root fs)
1861  *
1862  * outputs:
1863  * zplprops	values for the zplprops we attach to the master node object
1864  * is_ci	true if requested file system will be purely case-insensitive
1865  *
1866  * Determine the settings for utf8only, normalization and
1867  * casesensitivity.  Specific values may have been requested by the
1868  * creator and/or we can inherit values from the parent dataset.  If
1869  * the file system is of too early a vintage, a creator can not
1870  * request settings for these properties, even if the requested
1871  * setting is the default value.  We don't actually want to create dsl
1872  * properties for these, so remove them from the source nvlist after
1873  * processing.
1874  */
1875 static int
1876 zfs_fill_zplprops_impl(objset_t *os, uint64_t default_zplver,
1877     boolean_t fuids_ok, nvlist_t *createprops, nvlist_t *zplprops,
1878     boolean_t *is_ci)
1879 {
1880 	uint64_t zplver = default_zplver;
1881 	uint64_t sense = ZFS_PROP_UNDEFINED;
1882 	uint64_t norm = ZFS_PROP_UNDEFINED;
1883 	uint64_t u8 = ZFS_PROP_UNDEFINED;
1884 
1885 	ASSERT(zplprops != NULL);
1886 
1887 	/*
1888 	 * Pull out creator prop choices, if any.
1889 	 */
1890 	if (createprops) {
1891 		(void) nvlist_lookup_uint64(createprops,
1892 		    zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
1893 		(void) nvlist_lookup_uint64(createprops,
1894 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
1895 		(void) nvlist_remove_all(createprops,
1896 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE));
1897 		(void) nvlist_lookup_uint64(createprops,
1898 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
1899 		(void) nvlist_remove_all(createprops,
1900 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
1901 		(void) nvlist_lookup_uint64(createprops,
1902 		    zfs_prop_to_name(ZFS_PROP_CASE), &sense);
1903 		(void) nvlist_remove_all(createprops,
1904 		    zfs_prop_to_name(ZFS_PROP_CASE));
1905 	}
1906 
1907 	/*
1908 	 * If the zpl version requested is whacky or the file system
1909 	 * or pool is version is too "young" to support normalization
1910 	 * and the creator tried to set a value for one of the props,
1911 	 * error out.
1912 	 */
1913 	if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
1914 	    (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
1915 	    (zplver < ZPL_VERSION_NORMALIZATION &&
1916 	    (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
1917 	    sense != ZFS_PROP_UNDEFINED)))
1918 		return (ENOTSUP);
1919 
1920 	/*
1921 	 * Put the version in the zplprops
1922 	 */
1923 	VERIFY(nvlist_add_uint64(zplprops,
1924 	    zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
1925 
1926 	if (norm == ZFS_PROP_UNDEFINED)
1927 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
1928 	VERIFY(nvlist_add_uint64(zplprops,
1929 	    zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
1930 
1931 	/*
1932 	 * If we're normalizing, names must always be valid UTF-8 strings.
1933 	 */
1934 	if (norm)
1935 		u8 = 1;
1936 	if (u8 == ZFS_PROP_UNDEFINED)
1937 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
1938 	VERIFY(nvlist_add_uint64(zplprops,
1939 	    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
1940 
1941 	if (sense == ZFS_PROP_UNDEFINED)
1942 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
1943 	VERIFY(nvlist_add_uint64(zplprops,
1944 	    zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
1945 
1946 	if (is_ci)
1947 		*is_ci = (sense == ZFS_CASE_INSENSITIVE);
1948 
1949 	return (0);
1950 }
1951 
1952 static int
1953 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
1954     nvlist_t *zplprops, boolean_t *is_ci)
1955 {
1956 	boolean_t fuids_ok = B_TRUE;
1957 	uint64_t zplver = ZPL_VERSION;
1958 	objset_t *os = NULL;
1959 	char parentname[MAXNAMELEN];
1960 	char *cp;
1961 	int error;
1962 
1963 	(void) strlcpy(parentname, dataset, sizeof (parentname));
1964 	cp = strrchr(parentname, '/');
1965 	ASSERT(cp != NULL);
1966 	cp[0] = '\0';
1967 
1968 	if (zfs_earlier_version(dataset, SPA_VERSION_FUID)) {
1969 		zplver = ZPL_VERSION_FUID - 1;
1970 		fuids_ok = B_FALSE;
1971 	}
1972 
1973 	/*
1974 	 * Open parent object set so we can inherit zplprop values.
1975 	 */
1976 	if ((error = dmu_objset_open(parentname, DMU_OST_ANY,
1977 	    DS_MODE_USER | DS_MODE_READONLY, &os)) != 0)
1978 		return (error);
1979 
1980 	error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, createprops,
1981 	    zplprops, is_ci);
1982 	dmu_objset_close(os);
1983 	return (error);
1984 }
1985 
1986 static int
1987 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
1988     nvlist_t *zplprops, boolean_t *is_ci)
1989 {
1990 	boolean_t fuids_ok = B_TRUE;
1991 	uint64_t zplver = ZPL_VERSION;
1992 	int error;
1993 
1994 	if (spa_vers < SPA_VERSION_FUID) {
1995 		zplver = ZPL_VERSION_FUID - 1;
1996 		fuids_ok = B_FALSE;
1997 	}
1998 
1999 	error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, createprops,
2000 	    zplprops, is_ci);
2001 	return (error);
2002 }
2003 
2004 /*
2005  * inputs:
2006  * zc_objset_type	type of objset to create (fs vs zvol)
2007  * zc_name		name of new objset
2008  * zc_value		name of snapshot to clone from (may be empty)
2009  * zc_nvlist_src{_size}	nvlist of properties to apply
2010  *
2011  * outputs: none
2012  */
2013 static int
2014 zfs_ioc_create(zfs_cmd_t *zc)
2015 {
2016 	objset_t *clone;
2017 	int error = 0;
2018 	zfs_creat_t zct;
2019 	nvlist_t *nvprops = NULL;
2020 	void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
2021 	dmu_objset_type_t type = zc->zc_objset_type;
2022 
2023 	switch (type) {
2024 
2025 	case DMU_OST_ZFS:
2026 		cbfunc = zfs_create_cb;
2027 		break;
2028 
2029 	case DMU_OST_ZVOL:
2030 		cbfunc = zvol_create_cb;
2031 		break;
2032 
2033 	default:
2034 		cbfunc = NULL;
2035 		break;
2036 	}
2037 	if (strchr(zc->zc_name, '@') ||
2038 	    strchr(zc->zc_name, '%'))
2039 		return (EINVAL);
2040 
2041 	if (zc->zc_nvlist_src != NULL &&
2042 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2043 	    &nvprops)) != 0)
2044 		return (error);
2045 
2046 	zct.zct_zplprops = NULL;
2047 	zct.zct_props = nvprops;
2048 
2049 	if (zc->zc_value[0] != '\0') {
2050 		/*
2051 		 * We're creating a clone of an existing snapshot.
2052 		 */
2053 		zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
2054 		if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0) {
2055 			nvlist_free(nvprops);
2056 			return (EINVAL);
2057 		}
2058 
2059 		error = dmu_objset_open(zc->zc_value, type,
2060 		    DS_MODE_USER | DS_MODE_READONLY, &clone);
2061 		if (error) {
2062 			nvlist_free(nvprops);
2063 			return (error);
2064 		}
2065 
2066 		error = dmu_objset_create(zc->zc_name, type, clone, 0,
2067 		    NULL, NULL);
2068 		if (error) {
2069 			dmu_objset_close(clone);
2070 			nvlist_free(nvprops);
2071 			return (error);
2072 		}
2073 		dmu_objset_close(clone);
2074 	} else {
2075 		boolean_t is_insensitive = B_FALSE;
2076 
2077 		if (cbfunc == NULL) {
2078 			nvlist_free(nvprops);
2079 			return (EINVAL);
2080 		}
2081 
2082 		if (type == DMU_OST_ZVOL) {
2083 			uint64_t volsize, volblocksize;
2084 
2085 			if (nvprops == NULL ||
2086 			    nvlist_lookup_uint64(nvprops,
2087 			    zfs_prop_to_name(ZFS_PROP_VOLSIZE),
2088 			    &volsize) != 0) {
2089 				nvlist_free(nvprops);
2090 				return (EINVAL);
2091 			}
2092 
2093 			if ((error = nvlist_lookup_uint64(nvprops,
2094 			    zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
2095 			    &volblocksize)) != 0 && error != ENOENT) {
2096 				nvlist_free(nvprops);
2097 				return (EINVAL);
2098 			}
2099 
2100 			if (error != 0)
2101 				volblocksize = zfs_prop_default_numeric(
2102 				    ZFS_PROP_VOLBLOCKSIZE);
2103 
2104 			if ((error = zvol_check_volblocksize(
2105 			    volblocksize)) != 0 ||
2106 			    (error = zvol_check_volsize(volsize,
2107 			    volblocksize)) != 0) {
2108 				nvlist_free(nvprops);
2109 				return (error);
2110 			}
2111 		} else if (type == DMU_OST_ZFS) {
2112 			int error;
2113 
2114 			/*
2115 			 * We have to have normalization and
2116 			 * case-folding flags correct when we do the
2117 			 * file system creation, so go figure them out
2118 			 * now.
2119 			 */
2120 			VERIFY(nvlist_alloc(&zct.zct_zplprops,
2121 			    NV_UNIQUE_NAME, KM_SLEEP) == 0);
2122 			error = zfs_fill_zplprops(zc->zc_name, nvprops,
2123 			    zct.zct_zplprops, &is_insensitive);
2124 			if (error != 0) {
2125 				nvlist_free(nvprops);
2126 				nvlist_free(zct.zct_zplprops);
2127 				return (error);
2128 			}
2129 		}
2130 		error = dmu_objset_create(zc->zc_name, type, NULL,
2131 		    is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
2132 		nvlist_free(zct.zct_zplprops);
2133 	}
2134 
2135 	/*
2136 	 * It would be nice to do this atomically.
2137 	 */
2138 	if (error == 0) {
2139 		if ((error = zfs_set_prop_nvlist(zc->zc_name, nvprops)) != 0)
2140 			(void) dmu_objset_destroy(zc->zc_name);
2141 	}
2142 	nvlist_free(nvprops);
2143 	return (error);
2144 }
2145 
2146 struct snap_prop_arg {
2147 	nvlist_t *nvprops;
2148 	const char *snapname;
2149 };
2150 
2151 static int
2152 set_snap_props(char *name, void *arg)
2153 {
2154 	struct snap_prop_arg *snpa = arg;
2155 	int len = strlen(name) + strlen(snpa->snapname) + 2;
2156 	char *buf = kmem_alloc(len, KM_SLEEP);
2157 	int err;
2158 
2159 	(void) snprintf(buf, len, "%s@%s", name, snpa->snapname);
2160 	err = zfs_set_prop_nvlist(buf, snpa->nvprops);
2161 	if (err)
2162 		(void) dmu_objset_destroy(buf);
2163 	kmem_free(buf, len);
2164 	return (err);
2165 }
2166 
2167 /*
2168  * inputs:
2169  * zc_name	name of filesystem
2170  * zc_value	short name of snapshot
2171  * zc_cookie	recursive flag
2172  *
2173  * outputs:	none
2174  */
2175 static int
2176 zfs_ioc_snapshot(zfs_cmd_t *zc)
2177 {
2178 	nvlist_t *nvprops = NULL;
2179 	int error;
2180 	boolean_t recursive = zc->zc_cookie;
2181 
2182 	if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
2183 		return (EINVAL);
2184 
2185 	if (zc->zc_nvlist_src != NULL &&
2186 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2187 	    &nvprops)) != 0)
2188 		return (error);
2189 
2190 	error = dmu_objset_snapshot(zc->zc_name, zc->zc_value, recursive);
2191 
2192 	/*
2193 	 * It would be nice to do this atomically.
2194 	 */
2195 	if (error == 0) {
2196 		struct snap_prop_arg snpa;
2197 		snpa.nvprops = nvprops;
2198 		snpa.snapname = zc->zc_value;
2199 		if (recursive) {
2200 			error = dmu_objset_find(zc->zc_name,
2201 			    set_snap_props, &snpa, DS_FIND_CHILDREN);
2202 			if (error) {
2203 				(void) dmu_snapshots_destroy(zc->zc_name,
2204 				    zc->zc_value);
2205 			}
2206 		} else {
2207 			error = set_snap_props(zc->zc_name, &snpa);
2208 		}
2209 	}
2210 	nvlist_free(nvprops);
2211 	return (error);
2212 }
2213 
2214 int
2215 zfs_unmount_snap(char *name, void *arg)
2216 {
2217 	vfs_t *vfsp = NULL;
2218 
2219 	if (arg) {
2220 		char *snapname = arg;
2221 		int len = strlen(name) + strlen(snapname) + 2;
2222 		char *buf = kmem_alloc(len, KM_SLEEP);
2223 
2224 		(void) strcpy(buf, name);
2225 		(void) strcat(buf, "@");
2226 		(void) strcat(buf, snapname);
2227 		vfsp = zfs_get_vfs(buf);
2228 		kmem_free(buf, len);
2229 	} else if (strchr(name, '@')) {
2230 		vfsp = zfs_get_vfs(name);
2231 	}
2232 
2233 	if (vfsp) {
2234 		/*
2235 		 * Always force the unmount for snapshots.
2236 		 */
2237 		int flag = MS_FORCE;
2238 		int err;
2239 
2240 		if ((err = vn_vfswlock(vfsp->vfs_vnodecovered)) != 0) {
2241 			VFS_RELE(vfsp);
2242 			return (err);
2243 		}
2244 		VFS_RELE(vfsp);
2245 		if ((err = dounmount(vfsp, flag, kcred)) != 0)
2246 			return (err);
2247 	}
2248 	return (0);
2249 }
2250 
2251 /*
2252  * inputs:
2253  * zc_name	name of filesystem
2254  * zc_value	short name of snapshot
2255  *
2256  * outputs:	none
2257  */
2258 static int
2259 zfs_ioc_destroy_snaps(zfs_cmd_t *zc)
2260 {
2261 	int err;
2262 
2263 	if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
2264 		return (EINVAL);
2265 	err = dmu_objset_find(zc->zc_name,
2266 	    zfs_unmount_snap, zc->zc_value, DS_FIND_CHILDREN);
2267 	if (err)
2268 		return (err);
2269 	return (dmu_snapshots_destroy(zc->zc_name, zc->zc_value));
2270 }
2271 
2272 /*
2273  * inputs:
2274  * zc_name		name of dataset to destroy
2275  * zc_objset_type	type of objset
2276  *
2277  * outputs:		none
2278  */
2279 static int
2280 zfs_ioc_destroy(zfs_cmd_t *zc)
2281 {
2282 	if (strchr(zc->zc_name, '@') && zc->zc_objset_type == DMU_OST_ZFS) {
2283 		int err = zfs_unmount_snap(zc->zc_name, NULL);
2284 		if (err)
2285 			return (err);
2286 	}
2287 
2288 	return (dmu_objset_destroy(zc->zc_name));
2289 }
2290 
2291 /*
2292  * inputs:
2293  * zc_name	name of dataset to rollback (to most recent snapshot)
2294  *
2295  * outputs:	none
2296  */
2297 static int
2298 zfs_ioc_rollback(zfs_cmd_t *zc)
2299 {
2300 	objset_t *os;
2301 	int error;
2302 	zfsvfs_t *zfsvfs = NULL;
2303 
2304 	/*
2305 	 * Get the zfsvfs for the receiving objset. There
2306 	 * won't be one if we're operating on a zvol, if the
2307 	 * objset doesn't exist yet, or is not mounted.
2308 	 */
2309 	error = dmu_objset_open(zc->zc_name, DMU_OST_ANY, DS_MODE_USER, &os);
2310 	if (error)
2311 		return (error);
2312 
2313 	if (dmu_objset_type(os) == DMU_OST_ZFS) {
2314 		mutex_enter(&os->os->os_user_ptr_lock);
2315 		zfsvfs = dmu_objset_get_user(os);
2316 		if (zfsvfs != NULL)
2317 			VFS_HOLD(zfsvfs->z_vfs);
2318 		mutex_exit(&os->os->os_user_ptr_lock);
2319 	}
2320 
2321 	if (zfsvfs != NULL) {
2322 		char osname[MAXNAMELEN];
2323 		int mode;
2324 
2325 		error = zfs_suspend_fs(zfsvfs, osname, &mode);
2326 		if (error == 0) {
2327 			int resume_err;
2328 
2329 			ASSERT(strcmp(osname, zc->zc_name) == 0);
2330 			error = dmu_objset_rollback(os);
2331 			resume_err = zfs_resume_fs(zfsvfs, osname, mode);
2332 			error = error ? error : resume_err;
2333 		} else {
2334 			dmu_objset_close(os);
2335 		}
2336 		VFS_RELE(zfsvfs->z_vfs);
2337 	} else {
2338 		error = dmu_objset_rollback(os);
2339 	}
2340 	/* Note, the dmu_objset_rollback() releases the objset for us. */
2341 
2342 	return (error);
2343 }
2344 
2345 /*
2346  * inputs:
2347  * zc_name	old name of dataset
2348  * zc_value	new name of dataset
2349  * zc_cookie	recursive flag (only valid for snapshots)
2350  *
2351  * outputs:	none
2352  */
2353 static int
2354 zfs_ioc_rename(zfs_cmd_t *zc)
2355 {
2356 	boolean_t recursive = zc->zc_cookie & 1;
2357 
2358 	zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
2359 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
2360 	    strchr(zc->zc_value, '%'))
2361 		return (EINVAL);
2362 
2363 	/*
2364 	 * Unmount snapshot unless we're doing a recursive rename,
2365 	 * in which case the dataset code figures out which snapshots
2366 	 * to unmount.
2367 	 */
2368 	if (!recursive && strchr(zc->zc_name, '@') != NULL &&
2369 	    zc->zc_objset_type == DMU_OST_ZFS) {
2370 		int err = zfs_unmount_snap(zc->zc_name, NULL);
2371 		if (err)
2372 			return (err);
2373 	}
2374 	return (dmu_objset_rename(zc->zc_name, zc->zc_value, recursive));
2375 }
2376 
2377 static void
2378 clear_props(char *dataset, nvlist_t *props)
2379 {
2380 	zfs_cmd_t *zc;
2381 	nvpair_t *prop;
2382 
2383 	if (props == NULL)
2384 		return;
2385 	zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
2386 	(void) strcpy(zc->zc_name, dataset);
2387 	for (prop = nvlist_next_nvpair(props, NULL); prop;
2388 	    prop = nvlist_next_nvpair(props, prop)) {
2389 		(void) strcpy(zc->zc_value, nvpair_name(prop));
2390 		if (zfs_secpolicy_inherit(zc, CRED()) == 0)
2391 			(void) zfs_ioc_inherit_prop(zc);
2392 	}
2393 	kmem_free(zc, sizeof (zfs_cmd_t));
2394 }
2395 
2396 /*
2397  * inputs:
2398  * zc_name		name of containing filesystem
2399  * zc_nvlist_src{_size}	nvlist of properties to apply
2400  * zc_value		name of snapshot to create
2401  * zc_string		name of clone origin (if DRR_FLAG_CLONE)
2402  * zc_cookie		file descriptor to recv from
2403  * zc_begin_record	the BEGIN record of the stream (not byteswapped)
2404  * zc_guid		force flag
2405  *
2406  * outputs:
2407  * zc_cookie		number of bytes read
2408  */
2409 static int
2410 zfs_ioc_recv(zfs_cmd_t *zc)
2411 {
2412 	file_t *fp;
2413 	objset_t *os;
2414 	dmu_recv_cookie_t drc;
2415 	zfsvfs_t *zfsvfs = NULL;
2416 	boolean_t force = (boolean_t)zc->zc_guid;
2417 	int error, fd;
2418 	offset_t off;
2419 	nvlist_t *props = NULL;
2420 	nvlist_t *origprops = NULL;
2421 	objset_t *origin = NULL;
2422 	char *tosnap;
2423 	char tofs[ZFS_MAXNAMELEN];
2424 
2425 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
2426 	    strchr(zc->zc_value, '@') == NULL ||
2427 	    strchr(zc->zc_value, '%'))
2428 		return (EINVAL);
2429 
2430 	(void) strcpy(tofs, zc->zc_value);
2431 	tosnap = strchr(tofs, '@');
2432 	*tosnap = '\0';
2433 	tosnap++;
2434 
2435 	if (zc->zc_nvlist_src != NULL &&
2436 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2437 	    &props)) != 0)
2438 		return (error);
2439 
2440 	fd = zc->zc_cookie;
2441 	fp = getf(fd);
2442 	if (fp == NULL) {
2443 		nvlist_free(props);
2444 		return (EBADF);
2445 	}
2446 
2447 	if (dmu_objset_open(tofs, DMU_OST_ANY,
2448 	    DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
2449 		/*
2450 		 * Try to get the zfsvfs for the receiving objset.
2451 		 * There won't be one if we're operating on a zvol,
2452 		 * if the objset doesn't exist yet, or is not mounted.
2453 		 */
2454 		mutex_enter(&os->os->os_user_ptr_lock);
2455 		if (zfsvfs = dmu_objset_get_user(os)) {
2456 			if (!mutex_tryenter(&zfsvfs->z_online_recv_lock)) {
2457 				mutex_exit(&os->os->os_user_ptr_lock);
2458 				dmu_objset_close(os);
2459 				zfsvfs = NULL;
2460 				error = EBUSY;
2461 				goto out;
2462 			}
2463 			VFS_HOLD(zfsvfs->z_vfs);
2464 		}
2465 		mutex_exit(&os->os->os_user_ptr_lock);
2466 
2467 		/*
2468 		 * If new properties are supplied, they are to completely
2469 		 * replace the existing ones, so stash away the existing ones.
2470 		 */
2471 		if (props)
2472 			(void) dsl_prop_get_all(os, &origprops, TRUE);
2473 
2474 		dmu_objset_close(os);
2475 	}
2476 
2477 	if (zc->zc_string[0]) {
2478 		error = dmu_objset_open(zc->zc_string, DMU_OST_ANY,
2479 		    DS_MODE_USER | DS_MODE_READONLY, &origin);
2480 		if (error)
2481 			goto out;
2482 	}
2483 
2484 	error = dmu_recv_begin(tofs, tosnap, &zc->zc_begin_record,
2485 	    force, origin, zfsvfs != NULL, &drc);
2486 	if (origin)
2487 		dmu_objset_close(origin);
2488 	if (error)
2489 		goto out;
2490 
2491 	/*
2492 	 * Reset properties.  We do this before we receive the stream
2493 	 * so that the properties are applied to the new data.
2494 	 */
2495 	if (props) {
2496 		clear_props(tofs, origprops);
2497 		/*
2498 		 * XXX - Note, this is all-or-nothing; should be best-effort.
2499 		 */
2500 		(void) zfs_set_prop_nvlist(tofs, props);
2501 	}
2502 
2503 	off = fp->f_offset;
2504 	error = dmu_recv_stream(&drc, fp->f_vnode, &off);
2505 
2506 	if (error == 0 && zfsvfs) {
2507 		char osname[MAXNAMELEN];
2508 		int mode;
2509 
2510 		/* online recv */
2511 		error = zfs_suspend_fs(zfsvfs, osname, &mode);
2512 		if (error == 0) {
2513 			int resume_err;
2514 
2515 			error = dmu_recv_end(&drc);
2516 			resume_err = zfs_resume_fs(zfsvfs, osname, mode);
2517 			error = error ? error : resume_err;
2518 		} else {
2519 			dmu_recv_abort_cleanup(&drc);
2520 		}
2521 	} else if (error == 0) {
2522 		error = dmu_recv_end(&drc);
2523 	}
2524 
2525 	zc->zc_cookie = off - fp->f_offset;
2526 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
2527 		fp->f_offset = off;
2528 
2529 	/*
2530 	 * On error, restore the original props.
2531 	 */
2532 	if (error && props) {
2533 		clear_props(tofs, props);
2534 		(void) zfs_set_prop_nvlist(tofs, origprops);
2535 	}
2536 out:
2537 	if (zfsvfs) {
2538 		mutex_exit(&zfsvfs->z_online_recv_lock);
2539 		VFS_RELE(zfsvfs->z_vfs);
2540 	}
2541 	nvlist_free(props);
2542 	nvlist_free(origprops);
2543 	releasef(fd);
2544 	return (error);
2545 }
2546 
2547 /*
2548  * inputs:
2549  * zc_name	name of snapshot to send
2550  * zc_value	short name of incremental fromsnap (may be empty)
2551  * zc_cookie	file descriptor to send stream to
2552  * zc_obj	fromorigin flag (mutually exclusive with zc_value)
2553  *
2554  * outputs: none
2555  */
2556 static int
2557 zfs_ioc_send(zfs_cmd_t *zc)
2558 {
2559 	objset_t *fromsnap = NULL;
2560 	objset_t *tosnap;
2561 	file_t *fp;
2562 	int error;
2563 	offset_t off;
2564 
2565 	error = dmu_objset_open(zc->zc_name, DMU_OST_ANY,
2566 	    DS_MODE_USER | DS_MODE_READONLY, &tosnap);
2567 	if (error)
2568 		return (error);
2569 
2570 	if (zc->zc_value[0] != '\0') {
2571 		char buf[MAXPATHLEN];
2572 		char *cp;
2573 
2574 		(void) strncpy(buf, zc->zc_name, sizeof (buf));
2575 		cp = strchr(buf, '@');
2576 		if (cp)
2577 			*(cp+1) = 0;
2578 		(void) strncat(buf, zc->zc_value, sizeof (buf));
2579 		error = dmu_objset_open(buf, DMU_OST_ANY,
2580 		    DS_MODE_USER | DS_MODE_READONLY, &fromsnap);
2581 		if (error) {
2582 			dmu_objset_close(tosnap);
2583 			return (error);
2584 		}
2585 	}
2586 
2587 	fp = getf(zc->zc_cookie);
2588 	if (fp == NULL) {
2589 		dmu_objset_close(tosnap);
2590 		if (fromsnap)
2591 			dmu_objset_close(fromsnap);
2592 		return (EBADF);
2593 	}
2594 
2595 	off = fp->f_offset;
2596 	error = dmu_sendbackup(tosnap, fromsnap, zc->zc_obj, fp->f_vnode, &off);
2597 
2598 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
2599 		fp->f_offset = off;
2600 	releasef(zc->zc_cookie);
2601 	if (fromsnap)
2602 		dmu_objset_close(fromsnap);
2603 	dmu_objset_close(tosnap);
2604 	return (error);
2605 }
2606 
2607 static int
2608 zfs_ioc_inject_fault(zfs_cmd_t *zc)
2609 {
2610 	int id, error;
2611 
2612 	error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
2613 	    &zc->zc_inject_record);
2614 
2615 	if (error == 0)
2616 		zc->zc_guid = (uint64_t)id;
2617 
2618 	return (error);
2619 }
2620 
2621 static int
2622 zfs_ioc_clear_fault(zfs_cmd_t *zc)
2623 {
2624 	return (zio_clear_fault((int)zc->zc_guid));
2625 }
2626 
2627 static int
2628 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
2629 {
2630 	int id = (int)zc->zc_guid;
2631 	int error;
2632 
2633 	error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
2634 	    &zc->zc_inject_record);
2635 
2636 	zc->zc_guid = id;
2637 
2638 	return (error);
2639 }
2640 
2641 static int
2642 zfs_ioc_error_log(zfs_cmd_t *zc)
2643 {
2644 	spa_t *spa;
2645 	int error;
2646 	size_t count = (size_t)zc->zc_nvlist_dst_size;
2647 
2648 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2649 		return (error);
2650 
2651 	error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
2652 	    &count);
2653 	if (error == 0)
2654 		zc->zc_nvlist_dst_size = count;
2655 	else
2656 		zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
2657 
2658 	spa_close(spa, FTAG);
2659 
2660 	return (error);
2661 }
2662 
2663 static int
2664 zfs_ioc_clear(zfs_cmd_t *zc)
2665 {
2666 	spa_t *spa;
2667 	vdev_t *vd;
2668 	uint64_t txg;
2669 	int error;
2670 
2671 	/*
2672 	 * On zpool clear we also fix up missing slogs
2673 	 */
2674 	mutex_enter(&spa_namespace_lock);
2675 	spa = spa_lookup(zc->zc_name);
2676 	if (spa == NULL) {
2677 		mutex_exit(&spa_namespace_lock);
2678 		return (EIO);
2679 	}
2680 	if (spa->spa_log_state == SPA_LOG_MISSING) {
2681 		/* we need to let spa_open/spa_load clear the chains */
2682 		spa->spa_log_state = SPA_LOG_CLEAR;
2683 	}
2684 	mutex_exit(&spa_namespace_lock);
2685 
2686 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2687 		return (error);
2688 
2689 	/*
2690 	 * Try to resume any I/Os which may have been suspended
2691 	 * as a result of a complete pool failure.
2692 	 */
2693 	if (!list_is_empty(&spa->spa_zio_list)) {
2694 		if (zio_vdev_resume_io(spa) != 0) {
2695 			spa_close(spa, FTAG);
2696 			return (EIO);
2697 		}
2698 	}
2699 
2700 	txg = spa_vdev_enter(spa);
2701 
2702 	if (zc->zc_guid == 0) {
2703 		vd = NULL;
2704 	} else {
2705 		vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
2706 		if (vd == NULL) {
2707 			(void) spa_vdev_exit(spa, NULL, txg, ENODEV);
2708 			spa_close(spa, FTAG);
2709 			return (ENODEV);
2710 		}
2711 	}
2712 
2713 	vdev_clear(spa, vd, B_TRUE);
2714 
2715 	(void) spa_vdev_exit(spa, NULL, txg, 0);
2716 
2717 	spa_close(spa, FTAG);
2718 
2719 	return (0);
2720 }
2721 
2722 /*
2723  * inputs:
2724  * zc_name	name of filesystem
2725  * zc_value	name of origin snapshot
2726  *
2727  * outputs:	none
2728  */
2729 static int
2730 zfs_ioc_promote(zfs_cmd_t *zc)
2731 {
2732 	char *cp;
2733 
2734 	/*
2735 	 * We don't need to unmount *all* the origin fs's snapshots, but
2736 	 * it's easier.
2737 	 */
2738 	cp = strchr(zc->zc_value, '@');
2739 	if (cp)
2740 		*cp = '\0';
2741 	(void) dmu_objset_find(zc->zc_value,
2742 	    zfs_unmount_snap, NULL, DS_FIND_SNAPSHOTS);
2743 	return (dsl_dataset_promote(zc->zc_name));
2744 }
2745 
2746 /*
2747  * We don't want to have a hard dependency
2748  * against some special symbols in sharefs
2749  * nfs, and smbsrv.  Determine them if needed when
2750  * the first file system is shared.
2751  * Neither sharefs, nfs or smbsrv are unloadable modules.
2752  */
2753 int (*znfsexport_fs)(void *arg);
2754 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
2755 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
2756 
2757 int zfs_nfsshare_inited;
2758 int zfs_smbshare_inited;
2759 
2760 ddi_modhandle_t nfs_mod;
2761 ddi_modhandle_t sharefs_mod;
2762 ddi_modhandle_t smbsrv_mod;
2763 kmutex_t zfs_share_lock;
2764 
2765 static int
2766 zfs_init_sharefs()
2767 {
2768 	int error;
2769 
2770 	ASSERT(MUTEX_HELD(&zfs_share_lock));
2771 	/* Both NFS and SMB shares also require sharetab support. */
2772 	if (sharefs_mod == NULL && ((sharefs_mod =
2773 	    ddi_modopen("fs/sharefs",
2774 	    KRTLD_MODE_FIRST, &error)) == NULL)) {
2775 		return (ENOSYS);
2776 	}
2777 	if (zshare_fs == NULL && ((zshare_fs =
2778 	    (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
2779 	    ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
2780 		return (ENOSYS);
2781 	}
2782 	return (0);
2783 }
2784 
2785 static int
2786 zfs_ioc_share(zfs_cmd_t *zc)
2787 {
2788 	int error;
2789 	int opcode;
2790 
2791 	switch (zc->zc_share.z_sharetype) {
2792 	case ZFS_SHARE_NFS:
2793 	case ZFS_UNSHARE_NFS:
2794 		if (zfs_nfsshare_inited == 0) {
2795 			mutex_enter(&zfs_share_lock);
2796 			if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
2797 			    KRTLD_MODE_FIRST, &error)) == NULL)) {
2798 				mutex_exit(&zfs_share_lock);
2799 				return (ENOSYS);
2800 			}
2801 			if (znfsexport_fs == NULL &&
2802 			    ((znfsexport_fs = (int (*)(void *))
2803 			    ddi_modsym(nfs_mod,
2804 			    "nfs_export", &error)) == NULL)) {
2805 				mutex_exit(&zfs_share_lock);
2806 				return (ENOSYS);
2807 			}
2808 			error = zfs_init_sharefs();
2809 			if (error) {
2810 				mutex_exit(&zfs_share_lock);
2811 				return (ENOSYS);
2812 			}
2813 			zfs_nfsshare_inited = 1;
2814 			mutex_exit(&zfs_share_lock);
2815 		}
2816 		break;
2817 	case ZFS_SHARE_SMB:
2818 	case ZFS_UNSHARE_SMB:
2819 		if (zfs_smbshare_inited == 0) {
2820 			mutex_enter(&zfs_share_lock);
2821 			if (smbsrv_mod == NULL && ((smbsrv_mod =
2822 			    ddi_modopen("drv/smbsrv",
2823 			    KRTLD_MODE_FIRST, &error)) == NULL)) {
2824 				mutex_exit(&zfs_share_lock);
2825 				return (ENOSYS);
2826 			}
2827 			if (zsmbexport_fs == NULL && ((zsmbexport_fs =
2828 			    (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
2829 			    "smb_server_share", &error)) == NULL)) {
2830 				mutex_exit(&zfs_share_lock);
2831 				return (ENOSYS);
2832 			}
2833 			error = zfs_init_sharefs();
2834 			if (error) {
2835 				mutex_exit(&zfs_share_lock);
2836 				return (ENOSYS);
2837 			}
2838 			zfs_smbshare_inited = 1;
2839 			mutex_exit(&zfs_share_lock);
2840 		}
2841 		break;
2842 	default:
2843 		return (EINVAL);
2844 	}
2845 
2846 	switch (zc->zc_share.z_sharetype) {
2847 	case ZFS_SHARE_NFS:
2848 	case ZFS_UNSHARE_NFS:
2849 		if (error =
2850 		    znfsexport_fs((void *)
2851 		    (uintptr_t)zc->zc_share.z_exportdata))
2852 			return (error);
2853 		break;
2854 	case ZFS_SHARE_SMB:
2855 	case ZFS_UNSHARE_SMB:
2856 		if (error = zsmbexport_fs((void *)
2857 		    (uintptr_t)zc->zc_share.z_exportdata,
2858 		    zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
2859 		    B_TRUE : B_FALSE)) {
2860 			return (error);
2861 		}
2862 		break;
2863 	}
2864 
2865 	opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
2866 	    zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
2867 	    SHAREFS_ADD : SHAREFS_REMOVE;
2868 
2869 	/*
2870 	 * Add or remove share from sharetab
2871 	 */
2872 	error = zshare_fs(opcode,
2873 	    (void *)(uintptr_t)zc->zc_share.z_sharedata,
2874 	    zc->zc_share.z_sharemax);
2875 
2876 	return (error);
2877 
2878 }
2879 
2880 /*
2881  * pool create, destroy, and export don't log the history as part of
2882  * zfsdev_ioctl, but rather zfs_ioc_pool_create, and zfs_ioc_pool_export
2883  * do the logging of those commands.
2884  */
2885 static zfs_ioc_vec_t zfs_ioc_vec[] = {
2886 	{ zfs_ioc_pool_create, zfs_secpolicy_config, POOL_NAME, B_FALSE },
2887 	{ zfs_ioc_pool_destroy,	zfs_secpolicy_config, POOL_NAME, B_FALSE },
2888 	{ zfs_ioc_pool_import, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2889 	{ zfs_ioc_pool_export, zfs_secpolicy_config, POOL_NAME, B_FALSE },
2890 	{ zfs_ioc_pool_configs,	zfs_secpolicy_none, NO_NAME, B_FALSE },
2891 	{ zfs_ioc_pool_stats, zfs_secpolicy_read, POOL_NAME, B_FALSE },
2892 	{ zfs_ioc_pool_tryimport, zfs_secpolicy_config, NO_NAME, B_FALSE },
2893 	{ zfs_ioc_pool_scrub, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2894 	{ zfs_ioc_pool_freeze, zfs_secpolicy_config, NO_NAME, B_FALSE },
2895 	{ zfs_ioc_pool_upgrade,	zfs_secpolicy_config, POOL_NAME, B_TRUE },
2896 	{ zfs_ioc_pool_get_history, zfs_secpolicy_config, POOL_NAME, B_FALSE },
2897 	{ zfs_ioc_vdev_add, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2898 	{ zfs_ioc_vdev_remove, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2899 	{ zfs_ioc_vdev_set_state, zfs_secpolicy_config,	POOL_NAME, B_TRUE },
2900 	{ zfs_ioc_vdev_attach, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2901 	{ zfs_ioc_vdev_detach, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2902 	{ zfs_ioc_vdev_setpath,	zfs_secpolicy_config, POOL_NAME, B_FALSE },
2903 	{ zfs_ioc_objset_stats,	zfs_secpolicy_read, DATASET_NAME, B_FALSE },
2904 	{ zfs_ioc_objset_zplprops, zfs_secpolicy_read, DATASET_NAME, B_FALSE },
2905 	{ zfs_ioc_dataset_list_next, zfs_secpolicy_read,
2906 	    DATASET_NAME, B_FALSE },
2907 	{ zfs_ioc_snapshot_list_next, zfs_secpolicy_read,
2908 	    DATASET_NAME, B_FALSE },
2909 	{ zfs_ioc_set_prop, zfs_secpolicy_none, DATASET_NAME, B_TRUE },
2910 	{ zfs_ioc_create_minor,	zfs_secpolicy_minor, DATASET_NAME, B_FALSE },
2911 	{ zfs_ioc_remove_minor,	zfs_secpolicy_minor, DATASET_NAME, B_FALSE },
2912 	{ zfs_ioc_create, zfs_secpolicy_create, DATASET_NAME, B_TRUE },
2913 	{ zfs_ioc_destroy, zfs_secpolicy_destroy, DATASET_NAME, B_TRUE },
2914 	{ zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, B_TRUE },
2915 	{ zfs_ioc_rename, zfs_secpolicy_rename,	DATASET_NAME, B_TRUE },
2916 	{ zfs_ioc_recv, zfs_secpolicy_receive, DATASET_NAME, B_TRUE },
2917 	{ zfs_ioc_send, zfs_secpolicy_send, DATASET_NAME, B_TRUE },
2918 	{ zfs_ioc_inject_fault,	zfs_secpolicy_inject, NO_NAME, B_FALSE },
2919 	{ zfs_ioc_clear_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE },
2920 	{ zfs_ioc_inject_list_next, zfs_secpolicy_inject, NO_NAME, B_FALSE },
2921 	{ zfs_ioc_error_log, zfs_secpolicy_inject, POOL_NAME, B_FALSE },
2922 	{ zfs_ioc_clear, zfs_secpolicy_config, POOL_NAME, B_TRUE },
2923 	{ zfs_ioc_promote, zfs_secpolicy_promote, DATASET_NAME, B_TRUE },
2924 	{ zfs_ioc_destroy_snaps, zfs_secpolicy_destroy,	DATASET_NAME, B_TRUE },
2925 	{ zfs_ioc_snapshot, zfs_secpolicy_snapshot, DATASET_NAME, B_TRUE },
2926 	{ zfs_ioc_dsobj_to_dsname, zfs_secpolicy_config, POOL_NAME, B_FALSE },
2927 	{ zfs_ioc_obj_to_path, zfs_secpolicy_config, NO_NAME, B_FALSE },
2928 	{ zfs_ioc_pool_set_props, zfs_secpolicy_config,	POOL_NAME, B_TRUE },
2929 	{ zfs_ioc_pool_get_props, zfs_secpolicy_read, POOL_NAME, B_FALSE },
2930 	{ zfs_ioc_set_fsacl, zfs_secpolicy_fsacl, DATASET_NAME, B_TRUE },
2931 	{ zfs_ioc_get_fsacl, zfs_secpolicy_read, DATASET_NAME, B_FALSE },
2932 	{ zfs_ioc_iscsi_perm_check, zfs_secpolicy_iscsi,
2933 	    DATASET_NAME, B_FALSE },
2934 	{ zfs_ioc_share, zfs_secpolicy_share, DATASET_NAME, B_FALSE },
2935 	{ zfs_ioc_inherit_prop, zfs_secpolicy_inherit, DATASET_NAME, B_TRUE },
2936 };
2937 
2938 static int
2939 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
2940 {
2941 	zfs_cmd_t *zc;
2942 	uint_t vec;
2943 	int error, rc;
2944 
2945 	if (getminor(dev) != 0)
2946 		return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
2947 
2948 	vec = cmd - ZFS_IOC;
2949 	ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
2950 
2951 	if (vec >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
2952 		return (EINVAL);
2953 
2954 	zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2955 
2956 	error = xcopyin((void *)arg, zc, sizeof (zfs_cmd_t));
2957 
2958 	if (error == 0)
2959 		error = zfs_ioc_vec[vec].zvec_secpolicy(zc, cr);
2960 
2961 	/*
2962 	 * Ensure that all pool/dataset names are valid before we pass down to
2963 	 * the lower layers.
2964 	 */
2965 	if (error == 0) {
2966 		zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
2967 		switch (zfs_ioc_vec[vec].zvec_namecheck) {
2968 		case POOL_NAME:
2969 			if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
2970 				error = EINVAL;
2971 			break;
2972 
2973 		case DATASET_NAME:
2974 			if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
2975 				error = EINVAL;
2976 			break;
2977 
2978 		case NO_NAME:
2979 			break;
2980 		}
2981 	}
2982 
2983 	if (error == 0)
2984 		error = zfs_ioc_vec[vec].zvec_func(zc);
2985 
2986 	rc = xcopyout(zc, (void *)arg, sizeof (zfs_cmd_t));
2987 	if (error == 0) {
2988 		error = rc;
2989 		if (zfs_ioc_vec[vec].zvec_his_log == B_TRUE)
2990 			zfs_log_history(zc);
2991 	}
2992 
2993 	kmem_free(zc, sizeof (zfs_cmd_t));
2994 	return (error);
2995 }
2996 
2997 static int
2998 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
2999 {
3000 	if (cmd != DDI_ATTACH)
3001 		return (DDI_FAILURE);
3002 
3003 	if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
3004 	    DDI_PSEUDO, 0) == DDI_FAILURE)
3005 		return (DDI_FAILURE);
3006 
3007 	zfs_dip = dip;
3008 
3009 	ddi_report_dev(dip);
3010 
3011 	return (DDI_SUCCESS);
3012 }
3013 
3014 static int
3015 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
3016 {
3017 	if (spa_busy() || zfs_busy() || zvol_busy())
3018 		return (DDI_FAILURE);
3019 
3020 	if (cmd != DDI_DETACH)
3021 		return (DDI_FAILURE);
3022 
3023 	zfs_dip = NULL;
3024 
3025 	ddi_prop_remove_all(dip);
3026 	ddi_remove_minor_node(dip, NULL);
3027 
3028 	return (DDI_SUCCESS);
3029 }
3030 
3031 /*ARGSUSED*/
3032 static int
3033 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
3034 {
3035 	switch (infocmd) {
3036 	case DDI_INFO_DEVT2DEVINFO:
3037 		*result = zfs_dip;
3038 		return (DDI_SUCCESS);
3039 
3040 	case DDI_INFO_DEVT2INSTANCE:
3041 		*result = (void *)0;
3042 		return (DDI_SUCCESS);
3043 	}
3044 
3045 	return (DDI_FAILURE);
3046 }
3047 
3048 /*
3049  * OK, so this is a little weird.
3050  *
3051  * /dev/zfs is the control node, i.e. minor 0.
3052  * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
3053  *
3054  * /dev/zfs has basically nothing to do except serve up ioctls,
3055  * so most of the standard driver entry points are in zvol.c.
3056  */
3057 static struct cb_ops zfs_cb_ops = {
3058 	zvol_open,	/* open */
3059 	zvol_close,	/* close */
3060 	zvol_strategy,	/* strategy */
3061 	nodev,		/* print */
3062 	zvol_dump,	/* dump */
3063 	zvol_read,	/* read */
3064 	zvol_write,	/* write */
3065 	zfsdev_ioctl,	/* ioctl */
3066 	nodev,		/* devmap */
3067 	nodev,		/* mmap */
3068 	nodev,		/* segmap */
3069 	nochpoll,	/* poll */
3070 	ddi_prop_op,	/* prop_op */
3071 	NULL,		/* streamtab */
3072 	D_NEW | D_MP | D_64BIT,		/* Driver compatibility flag */
3073 	CB_REV,		/* version */
3074 	nodev,		/* async read */
3075 	nodev,		/* async write */
3076 };
3077 
3078 static struct dev_ops zfs_dev_ops = {
3079 	DEVO_REV,	/* version */
3080 	0,		/* refcnt */
3081 	zfs_info,	/* info */
3082 	nulldev,	/* identify */
3083 	nulldev,	/* probe */
3084 	zfs_attach,	/* attach */
3085 	zfs_detach,	/* detach */
3086 	nodev,		/* reset */
3087 	&zfs_cb_ops,	/* driver operations */
3088 	NULL,		/* no bus operations */
3089 	NULL,		/* power */
3090 	ddi_quiesce_not_needed,	/* quiesce */
3091 };
3092 
3093 static struct modldrv zfs_modldrv = {
3094 	&mod_driverops,
3095 	"ZFS storage pool",
3096 	&zfs_dev_ops
3097 };
3098 
3099 static struct modlinkage modlinkage = {
3100 	MODREV_1,
3101 	(void *)&zfs_modlfs,
3102 	(void *)&zfs_modldrv,
3103 	NULL
3104 };
3105 
3106 
3107 uint_t zfs_fsyncer_key;
3108 extern uint_t rrw_tsd_key;
3109 
3110 int
3111 _init(void)
3112 {
3113 	int error;
3114 
3115 	spa_init(FREAD | FWRITE);
3116 	zfs_init();
3117 	zvol_init();
3118 
3119 	if ((error = mod_install(&modlinkage)) != 0) {
3120 		zvol_fini();
3121 		zfs_fini();
3122 		spa_fini();
3123 		return (error);
3124 	}
3125 
3126 	tsd_create(&zfs_fsyncer_key, NULL);
3127 	tsd_create(&rrw_tsd_key, NULL);
3128 
3129 	error = ldi_ident_from_mod(&modlinkage, &zfs_li);
3130 	ASSERT(error == 0);
3131 	mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
3132 
3133 	return (0);
3134 }
3135 
3136 int
3137 _fini(void)
3138 {
3139 	int error;
3140 
3141 	if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
3142 		return (EBUSY);
3143 
3144 	if ((error = mod_remove(&modlinkage)) != 0)
3145 		return (error);
3146 
3147 	zvol_fini();
3148 	zfs_fini();
3149 	spa_fini();
3150 	if (zfs_nfsshare_inited)
3151 		(void) ddi_modclose(nfs_mod);
3152 	if (zfs_smbshare_inited)
3153 		(void) ddi_modclose(smbsrv_mod);
3154 	if (zfs_nfsshare_inited || zfs_smbshare_inited)
3155 		(void) ddi_modclose(sharefs_mod);
3156 
3157 	tsd_destroy(&zfs_fsyncer_key);
3158 	ldi_ident_release(zfs_li);
3159 	zfs_li = NULL;
3160 	mutex_destroy(&zfs_share_lock);
3161 
3162 	return (error);
3163 }
3164 
3165 int
3166 _info(struct modinfo *modinfop)
3167 {
3168 	return (mod_info(&modlinkage, modinfop));
3169 }
3170