xref: /titanic_41/usr/src/uts/common/fs/zfs/zfs_ioctl.c (revision 2b637af7da499c08e98bb81902f8992f624e8d86)
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 enum {
83 	NO_NAME,
84 	POOL_NAME,
85 	DATASET_NAME
86 } zfs_ioc_namecheck_t;
87 
88 typedef struct zfs_ioc_vec {
89 	zfs_ioc_func_t		*zvec_func;
90 	zfs_secpolicy_func_t	*zvec_secpolicy;
91 	zfs_ioc_namecheck_t	zvec_namecheck;
92 	boolean_t		zvec_his_log;
93 	boolean_t		zvec_pool_check;
94 } zfs_ioc_vec_t;
95 
96 /* This array is indexed by zfs_userquota_prop_t */
97 static const char *userquota_perms[] = {
98 	ZFS_DELEG_PERM_USERUSED,
99 	ZFS_DELEG_PERM_USERQUOTA,
100 	ZFS_DELEG_PERM_GROUPUSED,
101 	ZFS_DELEG_PERM_GROUPQUOTA,
102 };
103 
104 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
105 static void clear_props(char *dataset, nvlist_t *props, nvlist_t *newprops);
106 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
107     boolean_t *);
108 int zfs_set_prop_nvlist(const char *, nvlist_t *);
109 
110 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
111 void
112 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
113 {
114 	const char *newfile;
115 	char buf[256];
116 	va_list adx;
117 
118 	/*
119 	 * Get rid of annoying "../common/" prefix to filename.
120 	 */
121 	newfile = strrchr(file, '/');
122 	if (newfile != NULL) {
123 		newfile = newfile + 1; /* Get rid of leading / */
124 	} else {
125 		newfile = file;
126 	}
127 
128 	va_start(adx, fmt);
129 	(void) vsnprintf(buf, sizeof (buf), fmt, adx);
130 	va_end(adx);
131 
132 	/*
133 	 * To get this data, use the zfs-dprintf probe as so:
134 	 * dtrace -q -n 'zfs-dprintf \
135 	 *	/stringof(arg0) == "dbuf.c"/ \
136 	 *	{printf("%s: %s", stringof(arg1), stringof(arg3))}'
137 	 * arg0 = file name
138 	 * arg1 = function name
139 	 * arg2 = line number
140 	 * arg3 = message
141 	 */
142 	DTRACE_PROBE4(zfs__dprintf,
143 	    char *, newfile, char *, func, int, line, char *, buf);
144 }
145 
146 static void
147 history_str_free(char *buf)
148 {
149 	kmem_free(buf, HIS_MAX_RECORD_LEN);
150 }
151 
152 static char *
153 history_str_get(zfs_cmd_t *zc)
154 {
155 	char *buf;
156 
157 	if (zc->zc_history == NULL)
158 		return (NULL);
159 
160 	buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
161 	if (copyinstr((void *)(uintptr_t)zc->zc_history,
162 	    buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
163 		history_str_free(buf);
164 		return (NULL);
165 	}
166 
167 	buf[HIS_MAX_RECORD_LEN -1] = '\0';
168 
169 	return (buf);
170 }
171 
172 /*
173  * Check to see if the named dataset is currently defined as bootable
174  */
175 static boolean_t
176 zfs_is_bootfs(const char *name)
177 {
178 	spa_t *spa;
179 	boolean_t ret = B_FALSE;
180 
181 	if (spa_open(name, &spa, FTAG) == 0) {
182 		if (spa->spa_bootfs) {
183 			objset_t *os;
184 
185 			if (dmu_objset_open(name, DMU_OST_ZFS,
186 			    DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
187 				ret = (dmu_objset_id(os) == spa->spa_bootfs);
188 				dmu_objset_close(os);
189 			}
190 		}
191 		spa_close(spa, FTAG);
192 	}
193 	return (ret);
194 }
195 
196 /*
197  * zfs_earlier_version
198  *
199  *	Return non-zero if the spa version is less than requested version.
200  */
201 static int
202 zfs_earlier_version(const char *name, int version)
203 {
204 	spa_t *spa;
205 
206 	if (spa_open(name, &spa, FTAG) == 0) {
207 		if (spa_version(spa) < version) {
208 			spa_close(spa, FTAG);
209 			return (1);
210 		}
211 		spa_close(spa, FTAG);
212 	}
213 	return (0);
214 }
215 
216 /*
217  * zpl_earlier_version
218  *
219  * Return TRUE if the ZPL version is less than requested version.
220  */
221 static boolean_t
222 zpl_earlier_version(const char *name, int version)
223 {
224 	objset_t *os;
225 	boolean_t rc = B_TRUE;
226 
227 	if (dmu_objset_open(name, DMU_OST_ANY,
228 	    DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
229 		uint64_t zplversion;
230 
231 		if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
232 			rc = zplversion < version;
233 		dmu_objset_close(os);
234 	}
235 	return (rc);
236 }
237 
238 static void
239 zfs_log_history(zfs_cmd_t *zc)
240 {
241 	spa_t *spa;
242 	char *buf;
243 
244 	if ((buf = history_str_get(zc)) == NULL)
245 		return;
246 
247 	if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
248 		if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
249 			(void) spa_history_log(spa, buf, LOG_CMD_NORMAL);
250 		spa_close(spa, FTAG);
251 	}
252 	history_str_free(buf);
253 }
254 
255 /*
256  * Policy for top-level read operations (list pools).  Requires no privileges,
257  * and can be used in the local zone, as there is no associated dataset.
258  */
259 /* ARGSUSED */
260 static int
261 zfs_secpolicy_none(zfs_cmd_t *zc, cred_t *cr)
262 {
263 	return (0);
264 }
265 
266 /*
267  * Policy for dataset read operations (list children, get statistics).  Requires
268  * no privileges, but must be visible in the local zone.
269  */
270 /* ARGSUSED */
271 static int
272 zfs_secpolicy_read(zfs_cmd_t *zc, cred_t *cr)
273 {
274 	if (INGLOBALZONE(curproc) ||
275 	    zone_dataset_visible(zc->zc_name, NULL))
276 		return (0);
277 
278 	return (ENOENT);
279 }
280 
281 static int
282 zfs_dozonecheck(const char *dataset, cred_t *cr)
283 {
284 	uint64_t zoned;
285 	int writable = 1;
286 
287 	/*
288 	 * The dataset must be visible by this zone -- check this first
289 	 * so they don't see EPERM on something they shouldn't know about.
290 	 */
291 	if (!INGLOBALZONE(curproc) &&
292 	    !zone_dataset_visible(dataset, &writable))
293 		return (ENOENT);
294 
295 	if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
296 		return (ENOENT);
297 
298 	if (INGLOBALZONE(curproc)) {
299 		/*
300 		 * If the fs is zoned, only root can access it from the
301 		 * global zone.
302 		 */
303 		if (secpolicy_zfs(cr) && zoned)
304 			return (EPERM);
305 	} else {
306 		/*
307 		 * If we are in a local zone, the 'zoned' property must be set.
308 		 */
309 		if (!zoned)
310 			return (EPERM);
311 
312 		/* must be writable by this zone */
313 		if (!writable)
314 			return (EPERM);
315 	}
316 	return (0);
317 }
318 
319 int
320 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
321 {
322 	int error;
323 
324 	error = zfs_dozonecheck(name, cr);
325 	if (error == 0) {
326 		error = secpolicy_zfs(cr);
327 		if (error)
328 			error = dsl_deleg_access(name, perm, cr);
329 	}
330 	return (error);
331 }
332 
333 static int
334 zfs_secpolicy_setprop(const char *name, zfs_prop_t prop, cred_t *cr)
335 {
336 	/*
337 	 * Check permissions for special properties.
338 	 */
339 	switch (prop) {
340 	case ZFS_PROP_ZONED:
341 		/*
342 		 * Disallow setting of 'zoned' from within a local zone.
343 		 */
344 		if (!INGLOBALZONE(curproc))
345 			return (EPERM);
346 		break;
347 
348 	case ZFS_PROP_QUOTA:
349 		if (!INGLOBALZONE(curproc)) {
350 			uint64_t zoned;
351 			char setpoint[MAXNAMELEN];
352 			/*
353 			 * Unprivileged users are allowed to modify the
354 			 * quota on things *under* (ie. contained by)
355 			 * the thing they own.
356 			 */
357 			if (dsl_prop_get_integer(name, "zoned", &zoned,
358 			    setpoint))
359 				return (EPERM);
360 			if (!zoned || strlen(name) <= strlen(setpoint))
361 				return (EPERM);
362 		}
363 		break;
364 	}
365 
366 	return (zfs_secpolicy_write_perms(name, zfs_prop_to_name(prop), cr));
367 }
368 
369 int
370 zfs_secpolicy_fsacl(zfs_cmd_t *zc, cred_t *cr)
371 {
372 	int error;
373 
374 	error = zfs_dozonecheck(zc->zc_name, cr);
375 	if (error)
376 		return (error);
377 
378 	/*
379 	 * permission to set permissions will be evaluated later in
380 	 * dsl_deleg_can_allow()
381 	 */
382 	return (0);
383 }
384 
385 int
386 zfs_secpolicy_rollback(zfs_cmd_t *zc, cred_t *cr)
387 {
388 	int error;
389 	error = zfs_secpolicy_write_perms(zc->zc_name,
390 	    ZFS_DELEG_PERM_ROLLBACK, cr);
391 	if (error == 0)
392 		error = zfs_secpolicy_write_perms(zc->zc_name,
393 		    ZFS_DELEG_PERM_MOUNT, cr);
394 	return (error);
395 }
396 
397 int
398 zfs_secpolicy_send(zfs_cmd_t *zc, cred_t *cr)
399 {
400 	return (zfs_secpolicy_write_perms(zc->zc_name,
401 	    ZFS_DELEG_PERM_SEND, cr));
402 }
403 
404 static int
405 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, cred_t *cr)
406 {
407 	vnode_t *vp;
408 	int error;
409 
410 	if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
411 	    NO_FOLLOW, NULL, &vp)) != 0)
412 		return (error);
413 
414 	/* Now make sure mntpnt and dataset are ZFS */
415 
416 	if (vp->v_vfsp->vfs_fstype != zfsfstype ||
417 	    (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
418 	    zc->zc_name) != 0)) {
419 		VN_RELE(vp);
420 		return (EPERM);
421 	}
422 
423 	VN_RELE(vp);
424 	return (dsl_deleg_access(zc->zc_name,
425 	    ZFS_DELEG_PERM_SHARE, cr));
426 }
427 
428 int
429 zfs_secpolicy_share(zfs_cmd_t *zc, cred_t *cr)
430 {
431 	if (!INGLOBALZONE(curproc))
432 		return (EPERM);
433 
434 	if (secpolicy_nfs(cr) == 0) {
435 		return (0);
436 	} else {
437 		return (zfs_secpolicy_deleg_share(zc, cr));
438 	}
439 }
440 
441 int
442 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, cred_t *cr)
443 {
444 	if (!INGLOBALZONE(curproc))
445 		return (EPERM);
446 
447 	if (secpolicy_smb(cr) == 0) {
448 		return (0);
449 	} else {
450 		return (zfs_secpolicy_deleg_share(zc, cr));
451 	}
452 }
453 
454 static int
455 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
456 {
457 	char *cp;
458 
459 	/*
460 	 * Remove the @bla or /bla from the end of the name to get the parent.
461 	 */
462 	(void) strncpy(parent, datasetname, parentsize);
463 	cp = strrchr(parent, '@');
464 	if (cp != NULL) {
465 		cp[0] = '\0';
466 	} else {
467 		cp = strrchr(parent, '/');
468 		if (cp == NULL)
469 			return (ENOENT);
470 		cp[0] = '\0';
471 	}
472 
473 	return (0);
474 }
475 
476 int
477 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
478 {
479 	int error;
480 
481 	if ((error = zfs_secpolicy_write_perms(name,
482 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
483 		return (error);
484 
485 	return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
486 }
487 
488 static int
489 zfs_secpolicy_destroy(zfs_cmd_t *zc, cred_t *cr)
490 {
491 	return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
492 }
493 
494 /*
495  * Must have sys_config privilege to check the iscsi permission
496  */
497 /* ARGSUSED */
498 static int
499 zfs_secpolicy_iscsi(zfs_cmd_t *zc, cred_t *cr)
500 {
501 	return (secpolicy_zfs(cr));
502 }
503 
504 int
505 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
506 {
507 	char 	parentname[MAXNAMELEN];
508 	int	error;
509 
510 	if ((error = zfs_secpolicy_write_perms(from,
511 	    ZFS_DELEG_PERM_RENAME, cr)) != 0)
512 		return (error);
513 
514 	if ((error = zfs_secpolicy_write_perms(from,
515 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
516 		return (error);
517 
518 	if ((error = zfs_get_parent(to, parentname,
519 	    sizeof (parentname))) != 0)
520 		return (error);
521 
522 	if ((error = zfs_secpolicy_write_perms(parentname,
523 	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
524 		return (error);
525 
526 	if ((error = zfs_secpolicy_write_perms(parentname,
527 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
528 		return (error);
529 
530 	return (error);
531 }
532 
533 static int
534 zfs_secpolicy_rename(zfs_cmd_t *zc, cred_t *cr)
535 {
536 	return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
537 }
538 
539 static int
540 zfs_secpolicy_promote(zfs_cmd_t *zc, cred_t *cr)
541 {
542 	char 	parentname[MAXNAMELEN];
543 	objset_t *clone;
544 	int error;
545 
546 	error = zfs_secpolicy_write_perms(zc->zc_name,
547 	    ZFS_DELEG_PERM_PROMOTE, cr);
548 	if (error)
549 		return (error);
550 
551 	error = dmu_objset_open(zc->zc_name, DMU_OST_ANY,
552 	    DS_MODE_USER | DS_MODE_READONLY, &clone);
553 
554 	if (error == 0) {
555 		dsl_dataset_t *pclone = NULL;
556 		dsl_dir_t *dd;
557 		dd = clone->os->os_dsl_dataset->ds_dir;
558 
559 		rw_enter(&dd->dd_pool->dp_config_rwlock, RW_READER);
560 		error = dsl_dataset_hold_obj(dd->dd_pool,
561 		    dd->dd_phys->dd_origin_obj, FTAG, &pclone);
562 		rw_exit(&dd->dd_pool->dp_config_rwlock);
563 		if (error) {
564 			dmu_objset_close(clone);
565 			return (error);
566 		}
567 
568 		error = zfs_secpolicy_write_perms(zc->zc_name,
569 		    ZFS_DELEG_PERM_MOUNT, cr);
570 
571 		dsl_dataset_name(pclone, parentname);
572 		dmu_objset_close(clone);
573 		dsl_dataset_rele(pclone, FTAG);
574 		if (error == 0)
575 			error = zfs_secpolicy_write_perms(parentname,
576 			    ZFS_DELEG_PERM_PROMOTE, cr);
577 	}
578 	return (error);
579 }
580 
581 static int
582 zfs_secpolicy_receive(zfs_cmd_t *zc, cred_t *cr)
583 {
584 	int error;
585 
586 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
587 	    ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
588 		return (error);
589 
590 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
591 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
592 		return (error);
593 
594 	return (zfs_secpolicy_write_perms(zc->zc_name,
595 	    ZFS_DELEG_PERM_CREATE, cr));
596 }
597 
598 int
599 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
600 {
601 	int error;
602 
603 	if ((error = zfs_secpolicy_write_perms(name,
604 	    ZFS_DELEG_PERM_SNAPSHOT, cr)) != 0)
605 		return (error);
606 
607 	error = zfs_secpolicy_write_perms(name,
608 	    ZFS_DELEG_PERM_MOUNT, cr);
609 
610 	return (error);
611 }
612 
613 static int
614 zfs_secpolicy_snapshot(zfs_cmd_t *zc, cred_t *cr)
615 {
616 
617 	return (zfs_secpolicy_snapshot_perms(zc->zc_name, cr));
618 }
619 
620 static int
621 zfs_secpolicy_create(zfs_cmd_t *zc, cred_t *cr)
622 {
623 	char 	parentname[MAXNAMELEN];
624 	int 	error;
625 
626 	if ((error = zfs_get_parent(zc->zc_name, parentname,
627 	    sizeof (parentname))) != 0)
628 		return (error);
629 
630 	if (zc->zc_value[0] != '\0') {
631 		if ((error = zfs_secpolicy_write_perms(zc->zc_value,
632 		    ZFS_DELEG_PERM_CLONE, cr)) != 0)
633 			return (error);
634 	}
635 
636 	if ((error = zfs_secpolicy_write_perms(parentname,
637 	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
638 		return (error);
639 
640 	error = zfs_secpolicy_write_perms(parentname,
641 	    ZFS_DELEG_PERM_MOUNT, cr);
642 
643 	return (error);
644 }
645 
646 static int
647 zfs_secpolicy_umount(zfs_cmd_t *zc, cred_t *cr)
648 {
649 	int error;
650 
651 	error = secpolicy_fs_unmount(cr, NULL);
652 	if (error) {
653 		error = dsl_deleg_access(zc->zc_name, ZFS_DELEG_PERM_MOUNT, cr);
654 	}
655 	return (error);
656 }
657 
658 /*
659  * Policy for pool operations - create/destroy pools, add vdevs, etc.  Requires
660  * SYS_CONFIG privilege, which is not available in a local zone.
661  */
662 /* ARGSUSED */
663 static int
664 zfs_secpolicy_config(zfs_cmd_t *zc, cred_t *cr)
665 {
666 	if (secpolicy_sys_config(cr, B_FALSE) != 0)
667 		return (EPERM);
668 
669 	return (0);
670 }
671 
672 /*
673  * Just like zfs_secpolicy_config, except that we will check for
674  * mount permission on the dataset for permission to create/remove
675  * the minor nodes.
676  */
677 static int
678 zfs_secpolicy_minor(zfs_cmd_t *zc, cred_t *cr)
679 {
680 	if (secpolicy_sys_config(cr, B_FALSE) != 0) {
681 		return (dsl_deleg_access(zc->zc_name,
682 		    ZFS_DELEG_PERM_MOUNT, cr));
683 	}
684 
685 	return (0);
686 }
687 
688 /*
689  * Policy for fault injection.  Requires all privileges.
690  */
691 /* ARGSUSED */
692 static int
693 zfs_secpolicy_inject(zfs_cmd_t *zc, cred_t *cr)
694 {
695 	return (secpolicy_zinject(cr));
696 }
697 
698 static int
699 zfs_secpolicy_inherit(zfs_cmd_t *zc, cred_t *cr)
700 {
701 	zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
702 
703 	if (prop == ZPROP_INVAL) {
704 		if (!zfs_prop_user(zc->zc_value))
705 			return (EINVAL);
706 		return (zfs_secpolicy_write_perms(zc->zc_name,
707 		    ZFS_DELEG_PERM_USERPROP, cr));
708 	} else {
709 		if (!zfs_prop_inheritable(prop))
710 			return (EINVAL);
711 		return (zfs_secpolicy_setprop(zc->zc_name, prop, cr));
712 	}
713 }
714 
715 static int
716 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, cred_t *cr)
717 {
718 	int err = zfs_secpolicy_read(zc, cr);
719 	if (err)
720 		return (err);
721 
722 	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
723 		return (EINVAL);
724 
725 	if (zc->zc_value[0] == 0) {
726 		/*
727 		 * They are asking about a posix uid/gid.  If it's
728 		 * themself, allow it.
729 		 */
730 		if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
731 		    zc->zc_objset_type == ZFS_PROP_USERQUOTA) {
732 			if (zc->zc_guid == crgetuid(cr))
733 				return (0);
734 		} else {
735 			if (groupmember(zc->zc_guid, cr))
736 				return (0);
737 		}
738 	}
739 
740 	return (zfs_secpolicy_write_perms(zc->zc_name,
741 	    userquota_perms[zc->zc_objset_type], cr));
742 }
743 
744 static int
745 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, cred_t *cr)
746 {
747 	int err = zfs_secpolicy_read(zc, cr);
748 	if (err)
749 		return (err);
750 
751 	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
752 		return (EINVAL);
753 
754 	return (zfs_secpolicy_write_perms(zc->zc_name,
755 	    userquota_perms[zc->zc_objset_type], cr));
756 }
757 
758 static int
759 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, cred_t *cr)
760 {
761 	return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION, cr));
762 }
763 
764 /*
765  * Returns the nvlist as specified by the user in the zfs_cmd_t.
766  */
767 static int
768 get_nvlist(uint64_t nvl, uint64_t size, nvlist_t **nvp)
769 {
770 	char *packed;
771 	int error;
772 	nvlist_t *list = NULL;
773 
774 	/*
775 	 * Read in and unpack the user-supplied nvlist.
776 	 */
777 	if (size == 0)
778 		return (EINVAL);
779 
780 	packed = kmem_alloc(size, KM_SLEEP);
781 
782 	if ((error = xcopyin((void *)(uintptr_t)nvl, packed, size)) != 0) {
783 		kmem_free(packed, size);
784 		return (error);
785 	}
786 
787 	if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
788 		kmem_free(packed, size);
789 		return (error);
790 	}
791 
792 	kmem_free(packed, size);
793 
794 	*nvp = list;
795 	return (0);
796 }
797 
798 static int
799 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
800 {
801 	char *packed = NULL;
802 	size_t size;
803 	int error;
804 
805 	VERIFY(nvlist_size(nvl, &size, NV_ENCODE_NATIVE) == 0);
806 
807 	if (size > zc->zc_nvlist_dst_size) {
808 		error = ENOMEM;
809 	} else {
810 		packed = kmem_alloc(size, KM_SLEEP);
811 		VERIFY(nvlist_pack(nvl, &packed, &size, NV_ENCODE_NATIVE,
812 		    KM_SLEEP) == 0);
813 		error = xcopyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
814 		    size);
815 		kmem_free(packed, size);
816 	}
817 
818 	zc->zc_nvlist_dst_size = size;
819 	return (error);
820 }
821 
822 static int
823 getzfsvfs(const char *dsname, zfsvfs_t **zvp)
824 {
825 	objset_t *os;
826 	int error;
827 
828 	error = dmu_objset_open(dsname, DMU_OST_ZFS,
829 	    DS_MODE_USER | DS_MODE_READONLY, &os);
830 	if (error)
831 		return (error);
832 
833 	mutex_enter(&os->os->os_user_ptr_lock);
834 	*zvp = dmu_objset_get_user(os);
835 	if (*zvp) {
836 		VFS_HOLD((*zvp)->z_vfs);
837 	} else {
838 		error = ESRCH;
839 	}
840 	mutex_exit(&os->os->os_user_ptr_lock);
841 	dmu_objset_close(os);
842 	return (error);
843 }
844 
845 /*
846  * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
847  * case its z_vfs will be NULL, and it will be opened as the owner.
848  */
849 static int
850 zfsvfs_hold(const char *name, boolean_t readonly, void *tag, zfsvfs_t **zvp)
851 {
852 	int error = 0;
853 	int mode = DS_MODE_OWNER | (readonly ? DS_MODE_READONLY : 0);
854 
855 	if (getzfsvfs(name, zvp) != 0)
856 		error = zfsvfs_create(name, mode, zvp);
857 	if (error == 0) {
858 		rrw_enter(&(*zvp)->z_teardown_lock, RW_READER, tag);
859 		if ((*zvp)->z_unmounted) {
860 			/*
861 			 * XXX we could probably try again, since the unmounting
862 			 * thread should be just about to disassociate the
863 			 * objset from the zfsvfs.
864 			 */
865 			rrw_exit(&(*zvp)->z_teardown_lock, tag);
866 			return (EBUSY);
867 		}
868 	}
869 	return (error);
870 }
871 
872 static void
873 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
874 {
875 	rrw_exit(&zfsvfs->z_teardown_lock, tag);
876 
877 	if (zfsvfs->z_vfs) {
878 		VFS_RELE(zfsvfs->z_vfs);
879 	} else {
880 		dmu_objset_close(zfsvfs->z_os);
881 		zfsvfs_free(zfsvfs);
882 	}
883 }
884 
885 static int
886 zfs_ioc_pool_create(zfs_cmd_t *zc)
887 {
888 	int error;
889 	nvlist_t *config, *props = NULL;
890 	nvlist_t *rootprops = NULL;
891 	nvlist_t *zplprops = NULL;
892 	char *buf;
893 
894 	if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
895 	    &config))
896 		return (error);
897 
898 	if (zc->zc_nvlist_src_size != 0 && (error =
899 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, &props))) {
900 		nvlist_free(config);
901 		return (error);
902 	}
903 
904 	if (props) {
905 		nvlist_t *nvl = NULL;
906 		uint64_t version = SPA_VERSION;
907 
908 		(void) nvlist_lookup_uint64(props,
909 		    zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
910 		if (version < SPA_VERSION_INITIAL || version > SPA_VERSION) {
911 			error = EINVAL;
912 			goto pool_props_bad;
913 		}
914 		(void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
915 		if (nvl) {
916 			error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
917 			if (error != 0) {
918 				nvlist_free(config);
919 				nvlist_free(props);
920 				return (error);
921 			}
922 			(void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
923 		}
924 		VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
925 		error = zfs_fill_zplprops_root(version, rootprops,
926 		    zplprops, NULL);
927 		if (error)
928 			goto pool_props_bad;
929 	}
930 
931 	buf = history_str_get(zc);
932 
933 	error = spa_create(zc->zc_name, config, props, buf, zplprops);
934 
935 	/*
936 	 * Set the remaining root properties
937 	 */
938 	if (!error &&
939 	    (error = zfs_set_prop_nvlist(zc->zc_name, rootprops)) != 0)
940 		(void) spa_destroy(zc->zc_name);
941 
942 	if (buf != NULL)
943 		history_str_free(buf);
944 
945 pool_props_bad:
946 	nvlist_free(rootprops);
947 	nvlist_free(zplprops);
948 	nvlist_free(config);
949 	nvlist_free(props);
950 
951 	return (error);
952 }
953 
954 static int
955 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
956 {
957 	int error;
958 	zfs_log_history(zc);
959 	error = spa_destroy(zc->zc_name);
960 	return (error);
961 }
962 
963 static int
964 zfs_ioc_pool_import(zfs_cmd_t *zc)
965 {
966 	int error;
967 	nvlist_t *config, *props = NULL;
968 	uint64_t guid;
969 
970 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
971 	    &config)) != 0)
972 		return (error);
973 
974 	if (zc->zc_nvlist_src_size != 0 && (error =
975 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, &props))) {
976 		nvlist_free(config);
977 		return (error);
978 	}
979 
980 	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
981 	    guid != zc->zc_guid)
982 		error = EINVAL;
983 	else if (zc->zc_cookie)
984 		error = spa_import_verbatim(zc->zc_name, config,
985 		    props);
986 	else
987 		error = spa_import(zc->zc_name, config, props);
988 
989 	nvlist_free(config);
990 
991 	if (props)
992 		nvlist_free(props);
993 
994 	return (error);
995 }
996 
997 static int
998 zfs_ioc_pool_export(zfs_cmd_t *zc)
999 {
1000 	int error;
1001 	boolean_t force = (boolean_t)zc->zc_cookie;
1002 	boolean_t hardforce = (boolean_t)zc->zc_guid;
1003 
1004 	zfs_log_history(zc);
1005 	error = spa_export(zc->zc_name, NULL, force, hardforce);
1006 	return (error);
1007 }
1008 
1009 static int
1010 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1011 {
1012 	nvlist_t *configs;
1013 	int error;
1014 
1015 	if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1016 		return (EEXIST);
1017 
1018 	error = put_nvlist(zc, configs);
1019 
1020 	nvlist_free(configs);
1021 
1022 	return (error);
1023 }
1024 
1025 static int
1026 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1027 {
1028 	nvlist_t *config;
1029 	int error;
1030 	int ret = 0;
1031 
1032 	error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1033 	    sizeof (zc->zc_value));
1034 
1035 	if (config != NULL) {
1036 		ret = put_nvlist(zc, config);
1037 		nvlist_free(config);
1038 
1039 		/*
1040 		 * The config may be present even if 'error' is non-zero.
1041 		 * In this case we return success, and preserve the real errno
1042 		 * in 'zc_cookie'.
1043 		 */
1044 		zc->zc_cookie = error;
1045 	} else {
1046 		ret = error;
1047 	}
1048 
1049 	return (ret);
1050 }
1051 
1052 /*
1053  * Try to import the given pool, returning pool stats as appropriate so that
1054  * user land knows which devices are available and overall pool health.
1055  */
1056 static int
1057 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1058 {
1059 	nvlist_t *tryconfig, *config;
1060 	int error;
1061 
1062 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1063 	    &tryconfig)) != 0)
1064 		return (error);
1065 
1066 	config = spa_tryimport(tryconfig);
1067 
1068 	nvlist_free(tryconfig);
1069 
1070 	if (config == NULL)
1071 		return (EINVAL);
1072 
1073 	error = put_nvlist(zc, config);
1074 	nvlist_free(config);
1075 
1076 	return (error);
1077 }
1078 
1079 static int
1080 zfs_ioc_pool_scrub(zfs_cmd_t *zc)
1081 {
1082 	spa_t *spa;
1083 	int error;
1084 
1085 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1086 		return (error);
1087 
1088 	error = spa_scrub(spa, zc->zc_cookie);
1089 
1090 	spa_close(spa, FTAG);
1091 
1092 	return (error);
1093 }
1094 
1095 static int
1096 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1097 {
1098 	spa_t *spa;
1099 	int error;
1100 
1101 	error = spa_open(zc->zc_name, &spa, FTAG);
1102 	if (error == 0) {
1103 		spa_freeze(spa);
1104 		spa_close(spa, FTAG);
1105 	}
1106 	return (error);
1107 }
1108 
1109 static int
1110 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1111 {
1112 	spa_t *spa;
1113 	int error;
1114 
1115 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1116 		return (error);
1117 
1118 	if (zc->zc_cookie < spa_version(spa) || zc->zc_cookie > SPA_VERSION) {
1119 		spa_close(spa, FTAG);
1120 		return (EINVAL);
1121 	}
1122 
1123 	spa_upgrade(spa, zc->zc_cookie);
1124 	spa_close(spa, FTAG);
1125 
1126 	return (error);
1127 }
1128 
1129 static int
1130 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1131 {
1132 	spa_t *spa;
1133 	char *hist_buf;
1134 	uint64_t size;
1135 	int error;
1136 
1137 	if ((size = zc->zc_history_len) == 0)
1138 		return (EINVAL);
1139 
1140 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1141 		return (error);
1142 
1143 	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1144 		spa_close(spa, FTAG);
1145 		return (ENOTSUP);
1146 	}
1147 
1148 	hist_buf = kmem_alloc(size, KM_SLEEP);
1149 	if ((error = spa_history_get(spa, &zc->zc_history_offset,
1150 	    &zc->zc_history_len, hist_buf)) == 0) {
1151 		error = xcopyout(hist_buf,
1152 		    (char *)(uintptr_t)zc->zc_history,
1153 		    zc->zc_history_len);
1154 	}
1155 
1156 	spa_close(spa, FTAG);
1157 	kmem_free(hist_buf, size);
1158 	return (error);
1159 }
1160 
1161 static int
1162 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1163 {
1164 	int error;
1165 
1166 	if (error = dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value))
1167 		return (error);
1168 
1169 	return (0);
1170 }
1171 
1172 static int
1173 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1174 {
1175 	objset_t *osp;
1176 	int error;
1177 
1178 	if ((error = dmu_objset_open(zc->zc_name, DMU_OST_ZFS,
1179 	    DS_MODE_USER | DS_MODE_READONLY, &osp)) != 0)
1180 		return (error);
1181 	error = zfs_obj_to_path(osp, zc->zc_obj, zc->zc_value,
1182 	    sizeof (zc->zc_value));
1183 	dmu_objset_close(osp);
1184 
1185 	return (error);
1186 }
1187 
1188 static int
1189 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1190 {
1191 	spa_t *spa;
1192 	int error;
1193 	nvlist_t *config, **l2cache, **spares;
1194 	uint_t nl2cache = 0, nspares = 0;
1195 
1196 	error = spa_open(zc->zc_name, &spa, FTAG);
1197 	if (error != 0)
1198 		return (error);
1199 
1200 	error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1201 	    &config);
1202 	(void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1203 	    &l2cache, &nl2cache);
1204 
1205 	(void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1206 	    &spares, &nspares);
1207 
1208 	/*
1209 	 * A root pool with concatenated devices is not supported.
1210 	 * Thus, can not add a device to a root pool.
1211 	 *
1212 	 * Intent log device can not be added to a rootpool because
1213 	 * during mountroot, zil is replayed, a seperated log device
1214 	 * can not be accessed during the mountroot time.
1215 	 *
1216 	 * l2cache and spare devices are ok to be added to a rootpool.
1217 	 */
1218 	if (spa->spa_bootfs != 0 && nl2cache == 0 && nspares == 0) {
1219 		spa_close(spa, FTAG);
1220 		return (EDOM);
1221 	}
1222 
1223 	if (error == 0) {
1224 		error = spa_vdev_add(spa, config);
1225 		nvlist_free(config);
1226 	}
1227 	spa_close(spa, FTAG);
1228 	return (error);
1229 }
1230 
1231 static int
1232 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1233 {
1234 	spa_t *spa;
1235 	int error;
1236 
1237 	error = spa_open(zc->zc_name, &spa, FTAG);
1238 	if (error != 0)
1239 		return (error);
1240 	error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1241 	spa_close(spa, FTAG);
1242 	return (error);
1243 }
1244 
1245 static int
1246 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1247 {
1248 	spa_t *spa;
1249 	int error;
1250 	vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1251 
1252 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1253 		return (error);
1254 	switch (zc->zc_cookie) {
1255 	case VDEV_STATE_ONLINE:
1256 		error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1257 		break;
1258 
1259 	case VDEV_STATE_OFFLINE:
1260 		error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1261 		break;
1262 
1263 	case VDEV_STATE_FAULTED:
1264 		error = vdev_fault(spa, zc->zc_guid);
1265 		break;
1266 
1267 	case VDEV_STATE_DEGRADED:
1268 		error = vdev_degrade(spa, zc->zc_guid);
1269 		break;
1270 
1271 	default:
1272 		error = EINVAL;
1273 	}
1274 	zc->zc_cookie = newstate;
1275 	spa_close(spa, FTAG);
1276 	return (error);
1277 }
1278 
1279 static int
1280 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1281 {
1282 	spa_t *spa;
1283 	int replacing = zc->zc_cookie;
1284 	nvlist_t *config;
1285 	int error;
1286 
1287 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1288 		return (error);
1289 
1290 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1291 	    &config)) == 0) {
1292 		error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1293 		nvlist_free(config);
1294 	}
1295 
1296 	spa_close(spa, FTAG);
1297 	return (error);
1298 }
1299 
1300 static int
1301 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1302 {
1303 	spa_t *spa;
1304 	int error;
1305 
1306 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1307 		return (error);
1308 
1309 	error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1310 
1311 	spa_close(spa, FTAG);
1312 	return (error);
1313 }
1314 
1315 static int
1316 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
1317 {
1318 	spa_t *spa;
1319 	char *path = zc->zc_value;
1320 	uint64_t guid = zc->zc_guid;
1321 	int error;
1322 
1323 	error = spa_open(zc->zc_name, &spa, FTAG);
1324 	if (error != 0)
1325 		return (error);
1326 
1327 	error = spa_vdev_setpath(spa, guid, path);
1328 	spa_close(spa, FTAG);
1329 	return (error);
1330 }
1331 
1332 static int
1333 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
1334 {
1335 	spa_t *spa;
1336 	char *fru = zc->zc_value;
1337 	uint64_t guid = zc->zc_guid;
1338 	int error;
1339 
1340 	error = spa_open(zc->zc_name, &spa, FTAG);
1341 	if (error != 0)
1342 		return (error);
1343 
1344 	error = spa_vdev_setfru(spa, guid, fru);
1345 	spa_close(spa, FTAG);
1346 	return (error);
1347 }
1348 
1349 /*
1350  * inputs:
1351  * zc_name		name of filesystem
1352  * zc_nvlist_dst_size	size of buffer for property nvlist
1353  *
1354  * outputs:
1355  * zc_objset_stats	stats
1356  * zc_nvlist_dst	property nvlist
1357  * zc_nvlist_dst_size	size of property nvlist
1358  */
1359 static int
1360 zfs_ioc_objset_stats(zfs_cmd_t *zc)
1361 {
1362 	objset_t *os = NULL;
1363 	int error;
1364 	nvlist_t *nv;
1365 
1366 	if (error = dmu_objset_open(zc->zc_name,
1367 	    DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os))
1368 		return (error);
1369 
1370 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1371 
1372 	if (zc->zc_nvlist_dst != 0 &&
1373 	    (error = dsl_prop_get_all(os, &nv, FALSE)) == 0) {
1374 		dmu_objset_stats(os, nv);
1375 		/*
1376 		 * NB: zvol_get_stats() will read the objset contents,
1377 		 * which we aren't supposed to do with a
1378 		 * DS_MODE_USER hold, because it could be
1379 		 * inconsistent.  So this is a bit of a workaround...
1380 		 */
1381 		if (!zc->zc_objset_stats.dds_inconsistent) {
1382 			if (dmu_objset_type(os) == DMU_OST_ZVOL)
1383 				VERIFY(zvol_get_stats(os, nv) == 0);
1384 		}
1385 		error = put_nvlist(zc, nv);
1386 		nvlist_free(nv);
1387 	}
1388 
1389 	dmu_objset_close(os);
1390 	return (error);
1391 }
1392 
1393 static int
1394 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
1395 {
1396 	uint64_t value;
1397 	int error;
1398 
1399 	/*
1400 	 * zfs_get_zplprop() will either find a value or give us
1401 	 * the default value (if there is one).
1402 	 */
1403 	if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
1404 		return (error);
1405 	VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
1406 	return (0);
1407 }
1408 
1409 /*
1410  * inputs:
1411  * zc_name		name of filesystem
1412  * zc_nvlist_dst_size	size of buffer for zpl property nvlist
1413  *
1414  * outputs:
1415  * zc_nvlist_dst	zpl property nvlist
1416  * zc_nvlist_dst_size	size of zpl property nvlist
1417  */
1418 static int
1419 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
1420 {
1421 	objset_t *os;
1422 	int err;
1423 
1424 	if (err = dmu_objset_open(zc->zc_name,
1425 	    DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os))
1426 		return (err);
1427 
1428 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1429 
1430 	/*
1431 	 * NB: nvl_add_zplprop() will read the objset contents,
1432 	 * which we aren't supposed to do with a DS_MODE_USER
1433 	 * hold, because it could be inconsistent.
1434 	 */
1435 	if (zc->zc_nvlist_dst != NULL &&
1436 	    !zc->zc_objset_stats.dds_inconsistent &&
1437 	    dmu_objset_type(os) == DMU_OST_ZFS) {
1438 		nvlist_t *nv;
1439 
1440 		VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1441 		if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
1442 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
1443 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
1444 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
1445 			err = put_nvlist(zc, nv);
1446 		nvlist_free(nv);
1447 	} else {
1448 		err = ENOENT;
1449 	}
1450 	dmu_objset_close(os);
1451 	return (err);
1452 }
1453 
1454 static boolean_t
1455 dataset_name_hidden(const char *name)
1456 {
1457 	/*
1458 	 * Skip over datasets that are not visible in this zone,
1459 	 * internal datasets (which have a $ in their name), and
1460 	 * temporary datasets (which have a % in their name).
1461 	 */
1462 	if (strchr(name, '$') != NULL)
1463 		return (B_TRUE);
1464 	if (strchr(name, '%') != NULL)
1465 		return (B_TRUE);
1466 	if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
1467 		return (B_TRUE);
1468 	return (B_FALSE);
1469 }
1470 
1471 /*
1472  * inputs:
1473  * zc_name		name of filesystem
1474  * zc_cookie		zap cursor
1475  * zc_nvlist_dst_size	size of buffer for property nvlist
1476  *
1477  * outputs:
1478  * zc_name		name of next filesystem
1479  * zc_cookie		zap cursor
1480  * zc_objset_stats	stats
1481  * zc_nvlist_dst	property nvlist
1482  * zc_nvlist_dst_size	size of property nvlist
1483  */
1484 static int
1485 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
1486 {
1487 	objset_t *os;
1488 	int error;
1489 	char *p;
1490 
1491 	if (error = dmu_objset_open(zc->zc_name,
1492 	    DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os)) {
1493 		if (error == ENOENT)
1494 			error = ESRCH;
1495 		return (error);
1496 	}
1497 
1498 	p = strrchr(zc->zc_name, '/');
1499 	if (p == NULL || p[1] != '\0')
1500 		(void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
1501 	p = zc->zc_name + strlen(zc->zc_name);
1502 
1503 	/*
1504 	 * Pre-fetch the datasets.  dmu_objset_prefetch() always returns 0
1505 	 * but is not declared void because its called by dmu_objset_find().
1506 	 */
1507 	if (zc->zc_cookie == 0) {
1508 		uint64_t cookie = 0;
1509 		int len = sizeof (zc->zc_name) - (p - zc->zc_name);
1510 
1511 		while (dmu_dir_list_next(os, len, p, NULL, &cookie) == 0)
1512 			(void) dmu_objset_prefetch(p, NULL);
1513 	}
1514 
1515 	do {
1516 		error = dmu_dir_list_next(os,
1517 		    sizeof (zc->zc_name) - (p - zc->zc_name), p,
1518 		    NULL, &zc->zc_cookie);
1519 		if (error == ENOENT)
1520 			error = ESRCH;
1521 	} while (error == 0 && dataset_name_hidden(zc->zc_name));
1522 	dmu_objset_close(os);
1523 
1524 	if (error == 0)
1525 		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
1526 
1527 	return (error);
1528 }
1529 
1530 /*
1531  * inputs:
1532  * zc_name		name of filesystem
1533  * zc_cookie		zap cursor
1534  * zc_nvlist_dst_size	size of buffer for property nvlist
1535  *
1536  * outputs:
1537  * zc_name		name of next snapshot
1538  * zc_objset_stats	stats
1539  * zc_nvlist_dst	property nvlist
1540  * zc_nvlist_dst_size	size of property nvlist
1541  */
1542 static int
1543 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
1544 {
1545 	objset_t *os;
1546 	int error;
1547 
1548 	error = dmu_objset_open(zc->zc_name,
1549 	    DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os);
1550 	if (error)
1551 		return (error == ENOENT ? ESRCH : error);
1552 
1553 	if (zc->zc_cookie == 0) {
1554 		(void) dmu_objset_find(zc->zc_name, dmu_objset_prefetch,
1555 		    NULL, DS_FIND_SNAPSHOTS);
1556 	}
1557 	/*
1558 	 * A dataset name of maximum length cannot have any snapshots,
1559 	 * so exit immediately.
1560 	 */
1561 	if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
1562 		dmu_objset_close(os);
1563 		return (ESRCH);
1564 	}
1565 
1566 	error = dmu_snapshot_list_next(os,
1567 	    sizeof (zc->zc_name) - strlen(zc->zc_name),
1568 	    zc->zc_name + strlen(zc->zc_name), NULL, &zc->zc_cookie, NULL);
1569 	dmu_objset_close(os);
1570 	if (error == 0)
1571 		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
1572 	else if (error == ENOENT)
1573 		error = ESRCH;
1574 
1575 	/* if we failed, undo the @ that we tacked on to zc_name */
1576 	if (error)
1577 		*strchr(zc->zc_name, '@') = '\0';
1578 	return (error);
1579 }
1580 
1581 int
1582 zfs_set_prop_nvlist(const char *name, nvlist_t *nvl)
1583 {
1584 	nvpair_t *elem;
1585 	int error = 0;
1586 	uint64_t intval;
1587 	char *strval;
1588 	nvlist_t *genericnvl;
1589 	boolean_t issnap = (strchr(name, '@') != NULL);
1590 
1591 	/*
1592 	 * First validate permission to set all of the properties
1593 	 */
1594 	elem = NULL;
1595 	while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) {
1596 		const char *propname = nvpair_name(elem);
1597 		zfs_prop_t prop = zfs_name_to_prop(propname);
1598 
1599 		if (prop == ZPROP_INVAL) {
1600 			/*
1601 			 * If this is a user-defined property, it must be a
1602 			 * string, and there is no further validation to do.
1603 			 */
1604 			if (zfs_prop_user(propname) &&
1605 			    nvpair_type(elem) == DATA_TYPE_STRING) {
1606 				if (error = zfs_secpolicy_write_perms(name,
1607 				    ZFS_DELEG_PERM_USERPROP, CRED()))
1608 					return (error);
1609 				continue;
1610 			}
1611 
1612 			if (!issnap && zfs_prop_userquota(propname) &&
1613 			    nvpair_type(elem) == DATA_TYPE_UINT64_ARRAY) {
1614 				const char *perm;
1615 				const char *up = zfs_userquota_prop_prefixes
1616 				    [ZFS_PROP_USERQUOTA];
1617 				if (strncmp(propname, up, strlen(up)) == 0)
1618 					perm = ZFS_DELEG_PERM_USERQUOTA;
1619 				else
1620 					perm = ZFS_DELEG_PERM_GROUPQUOTA;
1621 				if (error = zfs_secpolicy_write_perms(name,
1622 				    perm, CRED()))
1623 					return (error);
1624 				continue;
1625 			}
1626 
1627 			return (EINVAL);
1628 		}
1629 
1630 		if (issnap)
1631 			return (EINVAL);
1632 
1633 		if ((error = zfs_secpolicy_setprop(name, prop, CRED())) != 0)
1634 			return (error);
1635 
1636 		/*
1637 		 * Check that this value is valid for this pool version
1638 		 */
1639 		switch (prop) {
1640 		case ZFS_PROP_COMPRESSION:
1641 			/*
1642 			 * If the user specified gzip compression, make sure
1643 			 * the SPA supports it. We ignore any errors here since
1644 			 * we'll catch them later.
1645 			 */
1646 			if (nvpair_type(elem) == DATA_TYPE_UINT64 &&
1647 			    nvpair_value_uint64(elem, &intval) == 0) {
1648 				if (intval >= ZIO_COMPRESS_GZIP_1 &&
1649 				    intval <= ZIO_COMPRESS_GZIP_9 &&
1650 				    zfs_earlier_version(name,
1651 				    SPA_VERSION_GZIP_COMPRESSION))
1652 					return (ENOTSUP);
1653 
1654 				/*
1655 				 * If this is a bootable dataset then
1656 				 * verify that the compression algorithm
1657 				 * is supported for booting. We must return
1658 				 * something other than ENOTSUP since it
1659 				 * implies a downrev pool version.
1660 				 */
1661 				if (zfs_is_bootfs(name) &&
1662 				    !BOOTFS_COMPRESS_VALID(intval))
1663 					return (ERANGE);
1664 			}
1665 			break;
1666 
1667 		case ZFS_PROP_COPIES:
1668 			if (zfs_earlier_version(name, SPA_VERSION_DITTO_BLOCKS))
1669 				return (ENOTSUP);
1670 			break;
1671 
1672 		case ZFS_PROP_SHARESMB:
1673 			if (zpl_earlier_version(name, ZPL_VERSION_FUID))
1674 				return (ENOTSUP);
1675 			break;
1676 
1677 		case ZFS_PROP_ACLINHERIT:
1678 			if (nvpair_type(elem) == DATA_TYPE_UINT64 &&
1679 			    nvpair_value_uint64(elem, &intval) == 0)
1680 				if (intval == ZFS_ACL_PASSTHROUGH_X &&
1681 				    zfs_earlier_version(name,
1682 				    SPA_VERSION_PASSTHROUGH_X))
1683 					return (ENOTSUP);
1684 		}
1685 	}
1686 
1687 	VERIFY(nvlist_alloc(&genericnvl, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1688 	elem = NULL;
1689 	while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) {
1690 		const char *propname = nvpair_name(elem);
1691 		zfs_prop_t prop = zfs_name_to_prop(propname);
1692 
1693 		if (prop == ZPROP_INVAL) {
1694 			if (zfs_prop_userquota(propname)) {
1695 				uint64_t *valary;
1696 				unsigned int vallen;
1697 				const char *domain;
1698 				zfs_userquota_prop_t type;
1699 				uint64_t rid;
1700 				uint64_t quota;
1701 				zfsvfs_t *zfsvfs;
1702 
1703 				VERIFY(nvpair_value_uint64_array(elem,
1704 				    &valary, &vallen) == 0);
1705 				VERIFY(vallen == 3);
1706 				type = valary[0];
1707 				rid = valary[1];
1708 				quota = valary[2];
1709 				domain = propname +
1710 				    strlen(zfs_userquota_prop_prefixes[type]);
1711 
1712 				error = zfsvfs_hold(name, B_FALSE, FTAG,
1713 				    &zfsvfs);
1714 				if (error == 0) {
1715 					error = zfs_set_userquota(zfsvfs,
1716 					    type, domain, rid, quota);
1717 					zfsvfs_rele(zfsvfs, FTAG);
1718 				}
1719 				if (error == 0)
1720 					continue;
1721 				else
1722 					goto out;
1723 			} else if (zfs_prop_user(propname)) {
1724 				VERIFY(nvpair_value_string(elem, &strval) == 0);
1725 				error = dsl_prop_set(name, propname, 1,
1726 				    strlen(strval) + 1, strval);
1727 				if (error == 0)
1728 					continue;
1729 				else
1730 					goto out;
1731 			}
1732 		}
1733 
1734 		switch (prop) {
1735 		case ZFS_PROP_QUOTA:
1736 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1737 			    (error = dsl_dir_set_quota(name, intval)) != 0)
1738 				goto out;
1739 			break;
1740 
1741 		case ZFS_PROP_REFQUOTA:
1742 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1743 			    (error = dsl_dataset_set_quota(name, intval)) != 0)
1744 				goto out;
1745 			break;
1746 
1747 		case ZFS_PROP_RESERVATION:
1748 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1749 			    (error = dsl_dir_set_reservation(name,
1750 			    intval)) != 0)
1751 				goto out;
1752 			break;
1753 
1754 		case ZFS_PROP_REFRESERVATION:
1755 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1756 			    (error = dsl_dataset_set_reservation(name,
1757 			    intval)) != 0)
1758 				goto out;
1759 			break;
1760 
1761 		case ZFS_PROP_VOLSIZE:
1762 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1763 			    (error = zvol_set_volsize(name,
1764 			    ddi_driver_major(zfs_dip), intval)) != 0)
1765 				goto out;
1766 			break;
1767 
1768 		case ZFS_PROP_VOLBLOCKSIZE:
1769 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1770 			    (error = zvol_set_volblocksize(name, intval)) != 0)
1771 				goto out;
1772 			break;
1773 
1774 		case ZFS_PROP_VERSION:
1775 		{
1776 			zfsvfs_t *zfsvfs;
1777 
1778 			if ((error = nvpair_value_uint64(elem, &intval)) != 0)
1779 				goto out;
1780 			if ((error = zfsvfs_hold(name, B_FALSE, FTAG,
1781 			    &zfsvfs)) != 0)
1782 				goto out;
1783 			error = zfs_set_version(zfsvfs, intval);
1784 			zfsvfs_rele(zfsvfs, FTAG);
1785 
1786 			if (error == 0 && intval >= ZPL_VERSION_USERSPACE) {
1787 				zfs_cmd_t zc = { 0 };
1788 				(void) strcpy(zc.zc_name, name);
1789 				(void) zfs_ioc_userspace_upgrade(&zc);
1790 			}
1791 			if (error)
1792 				goto out;
1793 			break;
1794 		}
1795 
1796 		default:
1797 			if (nvpair_type(elem) == DATA_TYPE_STRING) {
1798 				if (zfs_prop_get_type(prop) !=
1799 				    PROP_TYPE_STRING) {
1800 					error = EINVAL;
1801 					goto out;
1802 				}
1803 			} else if (nvpair_type(elem) == DATA_TYPE_UINT64) {
1804 				const char *unused;
1805 
1806 				VERIFY(nvpair_value_uint64(elem, &intval) == 0);
1807 
1808 				switch (zfs_prop_get_type(prop)) {
1809 				case PROP_TYPE_NUMBER:
1810 					break;
1811 				case PROP_TYPE_STRING:
1812 					error = EINVAL;
1813 					goto out;
1814 				case PROP_TYPE_INDEX:
1815 					if (zfs_prop_index_to_string(prop,
1816 					    intval, &unused) != 0) {
1817 						error = EINVAL;
1818 						goto out;
1819 					}
1820 					break;
1821 				default:
1822 					cmn_err(CE_PANIC,
1823 					    "unknown property type");
1824 					break;
1825 				}
1826 			} else {
1827 				error = EINVAL;
1828 				goto out;
1829 			}
1830 			if ((error = nvlist_add_nvpair(genericnvl, elem)) != 0)
1831 				goto out;
1832 		}
1833 	}
1834 
1835 	if (nvlist_next_nvpair(genericnvl, NULL) != NULL) {
1836 		error = dsl_props_set(name, genericnvl);
1837 	}
1838 out:
1839 	nvlist_free(genericnvl);
1840 	return (error);
1841 }
1842 
1843 /*
1844  * Check that all the properties are valid user properties.
1845  */
1846 static int
1847 zfs_check_userprops(char *fsname, nvlist_t *nvl)
1848 {
1849 	nvpair_t *elem = NULL;
1850 	int error = 0;
1851 
1852 	while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) {
1853 		const char *propname = nvpair_name(elem);
1854 		char *valstr;
1855 
1856 		if (!zfs_prop_user(propname) ||
1857 		    nvpair_type(elem) != DATA_TYPE_STRING)
1858 			return (EINVAL);
1859 
1860 		if (error = zfs_secpolicy_write_perms(fsname,
1861 		    ZFS_DELEG_PERM_USERPROP, CRED()))
1862 			return (error);
1863 
1864 		if (strlen(propname) >= ZAP_MAXNAMELEN)
1865 			return (ENAMETOOLONG);
1866 
1867 		VERIFY(nvpair_value_string(elem, &valstr) == 0);
1868 		if (strlen(valstr) >= ZAP_MAXVALUELEN)
1869 			return (E2BIG);
1870 	}
1871 	return (0);
1872 }
1873 
1874 /*
1875  * inputs:
1876  * zc_name		name of filesystem
1877  * zc_value		name of property to set
1878  * zc_nvlist_src{_size}	nvlist of properties to apply
1879  * zc_cookie		clear existing local props?
1880  *
1881  * outputs:		none
1882  */
1883 static int
1884 zfs_ioc_set_prop(zfs_cmd_t *zc)
1885 {
1886 	nvlist_t *nvl;
1887 	int error;
1888 
1889 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1890 	    &nvl)) != 0)
1891 		return (error);
1892 
1893 	if (zc->zc_cookie) {
1894 		nvlist_t *origprops;
1895 		objset_t *os;
1896 
1897 		if (dmu_objset_open(zc->zc_name, DMU_OST_ANY,
1898 		    DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
1899 			if (dsl_prop_get_all(os, &origprops, TRUE) == 0) {
1900 				clear_props(zc->zc_name, origprops, nvl);
1901 				nvlist_free(origprops);
1902 			}
1903 			dmu_objset_close(os);
1904 		}
1905 
1906 	}
1907 
1908 	error = zfs_set_prop_nvlist(zc->zc_name, nvl);
1909 
1910 	nvlist_free(nvl);
1911 	return (error);
1912 }
1913 
1914 /*
1915  * inputs:
1916  * zc_name		name of filesystem
1917  * zc_value		name of property to inherit
1918  *
1919  * outputs:		none
1920  */
1921 static int
1922 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
1923 {
1924 	/* the property name has been validated by zfs_secpolicy_inherit() */
1925 	return (dsl_prop_set(zc->zc_name, zc->zc_value, 0, 0, NULL));
1926 }
1927 
1928 static int
1929 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
1930 {
1931 	nvlist_t *props;
1932 	spa_t *spa;
1933 	int error;
1934 	nvpair_t *elem;
1935 
1936 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1937 	    &props)))
1938 		return (error);
1939 
1940 	/*
1941 	 * If the only property is the configfile, then just do a spa_lookup()
1942 	 * to handle the faulted case.
1943 	 */
1944 	elem = nvlist_next_nvpair(props, NULL);
1945 	if (elem != NULL && strcmp(nvpair_name(elem),
1946 	    zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
1947 	    nvlist_next_nvpair(props, elem) == NULL) {
1948 		mutex_enter(&spa_namespace_lock);
1949 		if ((spa = spa_lookup(zc->zc_name)) != NULL) {
1950 			spa_configfile_set(spa, props, B_FALSE);
1951 			spa_config_sync(spa, B_FALSE, B_TRUE);
1952 		}
1953 		mutex_exit(&spa_namespace_lock);
1954 		if (spa != NULL)
1955 			return (0);
1956 	}
1957 
1958 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
1959 		nvlist_free(props);
1960 		return (error);
1961 	}
1962 
1963 	error = spa_prop_set(spa, props);
1964 
1965 	nvlist_free(props);
1966 	spa_close(spa, FTAG);
1967 
1968 	return (error);
1969 }
1970 
1971 static int
1972 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
1973 {
1974 	spa_t *spa;
1975 	int error;
1976 	nvlist_t *nvp = NULL;
1977 
1978 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
1979 		/*
1980 		 * If the pool is faulted, there may be properties we can still
1981 		 * get (such as altroot and cachefile), so attempt to get them
1982 		 * anyway.
1983 		 */
1984 		mutex_enter(&spa_namespace_lock);
1985 		if ((spa = spa_lookup(zc->zc_name)) != NULL)
1986 			error = spa_prop_get(spa, &nvp);
1987 		mutex_exit(&spa_namespace_lock);
1988 	} else {
1989 		error = spa_prop_get(spa, &nvp);
1990 		spa_close(spa, FTAG);
1991 	}
1992 
1993 	if (error == 0 && zc->zc_nvlist_dst != NULL)
1994 		error = put_nvlist(zc, nvp);
1995 	else
1996 		error = EFAULT;
1997 
1998 	nvlist_free(nvp);
1999 	return (error);
2000 }
2001 
2002 static int
2003 zfs_ioc_iscsi_perm_check(zfs_cmd_t *zc)
2004 {
2005 	nvlist_t *nvp;
2006 	int error;
2007 	uint32_t uid;
2008 	uint32_t gid;
2009 	uint32_t *groups;
2010 	uint_t group_cnt;
2011 	cred_t	*usercred;
2012 
2013 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2014 	    &nvp)) != 0) {
2015 		return (error);
2016 	}
2017 
2018 	if ((error = nvlist_lookup_uint32(nvp,
2019 	    ZFS_DELEG_PERM_UID, &uid)) != 0) {
2020 		nvlist_free(nvp);
2021 		return (EPERM);
2022 	}
2023 
2024 	if ((error = nvlist_lookup_uint32(nvp,
2025 	    ZFS_DELEG_PERM_GID, &gid)) != 0) {
2026 		nvlist_free(nvp);
2027 		return (EPERM);
2028 	}
2029 
2030 	if ((error = nvlist_lookup_uint32_array(nvp, ZFS_DELEG_PERM_GROUPS,
2031 	    &groups, &group_cnt)) != 0) {
2032 		nvlist_free(nvp);
2033 		return (EPERM);
2034 	}
2035 	usercred = cralloc();
2036 	if ((crsetugid(usercred, uid, gid) != 0) ||
2037 	    (crsetgroups(usercred, group_cnt, (gid_t *)groups) != 0)) {
2038 		nvlist_free(nvp);
2039 		crfree(usercred);
2040 		return (EPERM);
2041 	}
2042 	nvlist_free(nvp);
2043 	error = dsl_deleg_access(zc->zc_name,
2044 	    zfs_prop_to_name(ZFS_PROP_SHAREISCSI), usercred);
2045 	crfree(usercred);
2046 	return (error);
2047 }
2048 
2049 /*
2050  * inputs:
2051  * zc_name		name of filesystem
2052  * zc_nvlist_src{_size}	nvlist of delegated permissions
2053  * zc_perm_action	allow/unallow flag
2054  *
2055  * outputs:		none
2056  */
2057 static int
2058 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
2059 {
2060 	int error;
2061 	nvlist_t *fsaclnv = NULL;
2062 
2063 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2064 	    &fsaclnv)) != 0)
2065 		return (error);
2066 
2067 	/*
2068 	 * Verify nvlist is constructed correctly
2069 	 */
2070 	if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
2071 		nvlist_free(fsaclnv);
2072 		return (EINVAL);
2073 	}
2074 
2075 	/*
2076 	 * If we don't have PRIV_SYS_MOUNT, then validate
2077 	 * that user is allowed to hand out each permission in
2078 	 * the nvlist(s)
2079 	 */
2080 
2081 	error = secpolicy_zfs(CRED());
2082 	if (error) {
2083 		if (zc->zc_perm_action == B_FALSE) {
2084 			error = dsl_deleg_can_allow(zc->zc_name,
2085 			    fsaclnv, CRED());
2086 		} else {
2087 			error = dsl_deleg_can_unallow(zc->zc_name,
2088 			    fsaclnv, CRED());
2089 		}
2090 	}
2091 
2092 	if (error == 0)
2093 		error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
2094 
2095 	nvlist_free(fsaclnv);
2096 	return (error);
2097 }
2098 
2099 /*
2100  * inputs:
2101  * zc_name		name of filesystem
2102  *
2103  * outputs:
2104  * zc_nvlist_src{_size}	nvlist of delegated permissions
2105  */
2106 static int
2107 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
2108 {
2109 	nvlist_t *nvp;
2110 	int error;
2111 
2112 	if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
2113 		error = put_nvlist(zc, nvp);
2114 		nvlist_free(nvp);
2115 	}
2116 
2117 	return (error);
2118 }
2119 
2120 /*
2121  * inputs:
2122  * zc_name		name of volume
2123  *
2124  * outputs:		none
2125  */
2126 static int
2127 zfs_ioc_create_minor(zfs_cmd_t *zc)
2128 {
2129 	return (zvol_create_minor(zc->zc_name, ddi_driver_major(zfs_dip)));
2130 }
2131 
2132 /*
2133  * inputs:
2134  * zc_name		name of volume
2135  *
2136  * outputs:		none
2137  */
2138 static int
2139 zfs_ioc_remove_minor(zfs_cmd_t *zc)
2140 {
2141 	return (zvol_remove_minor(zc->zc_name));
2142 }
2143 
2144 /*
2145  * Search the vfs list for a specified resource.  Returns a pointer to it
2146  * or NULL if no suitable entry is found. The caller of this routine
2147  * is responsible for releasing the returned vfs pointer.
2148  */
2149 static vfs_t *
2150 zfs_get_vfs(const char *resource)
2151 {
2152 	struct vfs *vfsp;
2153 	struct vfs *vfs_found = NULL;
2154 
2155 	vfs_list_read_lock();
2156 	vfsp = rootvfs;
2157 	do {
2158 		if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
2159 			VFS_HOLD(vfsp);
2160 			vfs_found = vfsp;
2161 			break;
2162 		}
2163 		vfsp = vfsp->vfs_next;
2164 	} while (vfsp != rootvfs);
2165 	vfs_list_unlock();
2166 	return (vfs_found);
2167 }
2168 
2169 /* ARGSUSED */
2170 static void
2171 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
2172 {
2173 	zfs_creat_t *zct = arg;
2174 
2175 	zfs_create_fs(os, cr, zct->zct_zplprops, tx);
2176 }
2177 
2178 #define	ZFS_PROP_UNDEFINED	((uint64_t)-1)
2179 
2180 /*
2181  * inputs:
2182  * createprops		list of properties requested by creator
2183  * default_zplver	zpl version to use if unspecified in createprops
2184  * fuids_ok		fuids allowed in this version of the spa?
2185  * os			parent objset pointer (NULL if root fs)
2186  *
2187  * outputs:
2188  * zplprops	values for the zplprops we attach to the master node object
2189  * is_ci	true if requested file system will be purely case-insensitive
2190  *
2191  * Determine the settings for utf8only, normalization and
2192  * casesensitivity.  Specific values may have been requested by the
2193  * creator and/or we can inherit values from the parent dataset.  If
2194  * the file system is of too early a vintage, a creator can not
2195  * request settings for these properties, even if the requested
2196  * setting is the default value.  We don't actually want to create dsl
2197  * properties for these, so remove them from the source nvlist after
2198  * processing.
2199  */
2200 static int
2201 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
2202     boolean_t fuids_ok, nvlist_t *createprops, nvlist_t *zplprops,
2203     boolean_t *is_ci)
2204 {
2205 	uint64_t sense = ZFS_PROP_UNDEFINED;
2206 	uint64_t norm = ZFS_PROP_UNDEFINED;
2207 	uint64_t u8 = ZFS_PROP_UNDEFINED;
2208 
2209 	ASSERT(zplprops != NULL);
2210 
2211 	/*
2212 	 * Pull out creator prop choices, if any.
2213 	 */
2214 	if (createprops) {
2215 		(void) nvlist_lookup_uint64(createprops,
2216 		    zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
2217 		(void) nvlist_lookup_uint64(createprops,
2218 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
2219 		(void) nvlist_remove_all(createprops,
2220 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE));
2221 		(void) nvlist_lookup_uint64(createprops,
2222 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
2223 		(void) nvlist_remove_all(createprops,
2224 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
2225 		(void) nvlist_lookup_uint64(createprops,
2226 		    zfs_prop_to_name(ZFS_PROP_CASE), &sense);
2227 		(void) nvlist_remove_all(createprops,
2228 		    zfs_prop_to_name(ZFS_PROP_CASE));
2229 	}
2230 
2231 	/*
2232 	 * If the zpl version requested is whacky or the file system
2233 	 * or pool is version is too "young" to support normalization
2234 	 * and the creator tried to set a value for one of the props,
2235 	 * error out.
2236 	 */
2237 	if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
2238 	    (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
2239 	    (zplver < ZPL_VERSION_NORMALIZATION &&
2240 	    (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
2241 	    sense != ZFS_PROP_UNDEFINED)))
2242 		return (ENOTSUP);
2243 
2244 	/*
2245 	 * Put the version in the zplprops
2246 	 */
2247 	VERIFY(nvlist_add_uint64(zplprops,
2248 	    zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
2249 
2250 	if (norm == ZFS_PROP_UNDEFINED)
2251 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
2252 	VERIFY(nvlist_add_uint64(zplprops,
2253 	    zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
2254 
2255 	/*
2256 	 * If we're normalizing, names must always be valid UTF-8 strings.
2257 	 */
2258 	if (norm)
2259 		u8 = 1;
2260 	if (u8 == ZFS_PROP_UNDEFINED)
2261 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
2262 	VERIFY(nvlist_add_uint64(zplprops,
2263 	    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
2264 
2265 	if (sense == ZFS_PROP_UNDEFINED)
2266 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
2267 	VERIFY(nvlist_add_uint64(zplprops,
2268 	    zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
2269 
2270 	if (is_ci)
2271 		*is_ci = (sense == ZFS_CASE_INSENSITIVE);
2272 
2273 	return (0);
2274 }
2275 
2276 static int
2277 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
2278     nvlist_t *zplprops, boolean_t *is_ci)
2279 {
2280 	boolean_t fuids_ok = B_TRUE;
2281 	uint64_t zplver = ZPL_VERSION;
2282 	objset_t *os = NULL;
2283 	char parentname[MAXNAMELEN];
2284 	char *cp;
2285 	int error;
2286 
2287 	(void) strlcpy(parentname, dataset, sizeof (parentname));
2288 	cp = strrchr(parentname, '/');
2289 	ASSERT(cp != NULL);
2290 	cp[0] = '\0';
2291 
2292 	if (zfs_earlier_version(dataset, SPA_VERSION_USERSPACE))
2293 		zplver = ZPL_VERSION_USERSPACE - 1;
2294 	if (zfs_earlier_version(dataset, SPA_VERSION_FUID)) {
2295 		zplver = ZPL_VERSION_FUID - 1;
2296 		fuids_ok = B_FALSE;
2297 	}
2298 
2299 	/*
2300 	 * Open parent object set so we can inherit zplprop values.
2301 	 */
2302 	if ((error = dmu_objset_open(parentname, DMU_OST_ANY,
2303 	    DS_MODE_USER | DS_MODE_READONLY, &os)) != 0)
2304 		return (error);
2305 
2306 	error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, createprops,
2307 	    zplprops, is_ci);
2308 	dmu_objset_close(os);
2309 	return (error);
2310 }
2311 
2312 static int
2313 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
2314     nvlist_t *zplprops, boolean_t *is_ci)
2315 {
2316 	boolean_t fuids_ok = B_TRUE;
2317 	uint64_t zplver = ZPL_VERSION;
2318 	int error;
2319 
2320 	if (spa_vers < SPA_VERSION_FUID) {
2321 		zplver = ZPL_VERSION_FUID - 1;
2322 		fuids_ok = B_FALSE;
2323 	}
2324 
2325 	error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, createprops,
2326 	    zplprops, is_ci);
2327 	return (error);
2328 }
2329 
2330 /*
2331  * inputs:
2332  * zc_objset_type	type of objset to create (fs vs zvol)
2333  * zc_name		name of new objset
2334  * zc_value		name of snapshot to clone from (may be empty)
2335  * zc_nvlist_src{_size}	nvlist of properties to apply
2336  *
2337  * outputs: none
2338  */
2339 static int
2340 zfs_ioc_create(zfs_cmd_t *zc)
2341 {
2342 	objset_t *clone;
2343 	int error = 0;
2344 	zfs_creat_t zct;
2345 	nvlist_t *nvprops = NULL;
2346 	void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
2347 	dmu_objset_type_t type = zc->zc_objset_type;
2348 
2349 	switch (type) {
2350 
2351 	case DMU_OST_ZFS:
2352 		cbfunc = zfs_create_cb;
2353 		break;
2354 
2355 	case DMU_OST_ZVOL:
2356 		cbfunc = zvol_create_cb;
2357 		break;
2358 
2359 	default:
2360 		cbfunc = NULL;
2361 		break;
2362 	}
2363 	if (strchr(zc->zc_name, '@') ||
2364 	    strchr(zc->zc_name, '%'))
2365 		return (EINVAL);
2366 
2367 	if (zc->zc_nvlist_src != NULL &&
2368 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2369 	    &nvprops)) != 0)
2370 		return (error);
2371 
2372 	zct.zct_zplprops = NULL;
2373 	zct.zct_props = nvprops;
2374 
2375 	if (zc->zc_value[0] != '\0') {
2376 		/*
2377 		 * We're creating a clone of an existing snapshot.
2378 		 */
2379 		zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
2380 		if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0) {
2381 			nvlist_free(nvprops);
2382 			return (EINVAL);
2383 		}
2384 
2385 		error = dmu_objset_open(zc->zc_value, type,
2386 		    DS_MODE_USER | DS_MODE_READONLY, &clone);
2387 		if (error) {
2388 			nvlist_free(nvprops);
2389 			return (error);
2390 		}
2391 
2392 		error = dmu_objset_create(zc->zc_name, type, clone, 0,
2393 		    NULL, NULL);
2394 		if (error) {
2395 			dmu_objset_close(clone);
2396 			nvlist_free(nvprops);
2397 			return (error);
2398 		}
2399 		dmu_objset_close(clone);
2400 	} else {
2401 		boolean_t is_insensitive = B_FALSE;
2402 
2403 		if (cbfunc == NULL) {
2404 			nvlist_free(nvprops);
2405 			return (EINVAL);
2406 		}
2407 
2408 		if (type == DMU_OST_ZVOL) {
2409 			uint64_t volsize, volblocksize;
2410 
2411 			if (nvprops == NULL ||
2412 			    nvlist_lookup_uint64(nvprops,
2413 			    zfs_prop_to_name(ZFS_PROP_VOLSIZE),
2414 			    &volsize) != 0) {
2415 				nvlist_free(nvprops);
2416 				return (EINVAL);
2417 			}
2418 
2419 			if ((error = nvlist_lookup_uint64(nvprops,
2420 			    zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
2421 			    &volblocksize)) != 0 && error != ENOENT) {
2422 				nvlist_free(nvprops);
2423 				return (EINVAL);
2424 			}
2425 
2426 			if (error != 0)
2427 				volblocksize = zfs_prop_default_numeric(
2428 				    ZFS_PROP_VOLBLOCKSIZE);
2429 
2430 			if ((error = zvol_check_volblocksize(
2431 			    volblocksize)) != 0 ||
2432 			    (error = zvol_check_volsize(volsize,
2433 			    volblocksize)) != 0) {
2434 				nvlist_free(nvprops);
2435 				return (error);
2436 			}
2437 		} else if (type == DMU_OST_ZFS) {
2438 			int error;
2439 
2440 			/*
2441 			 * We have to have normalization and
2442 			 * case-folding flags correct when we do the
2443 			 * file system creation, so go figure them out
2444 			 * now.
2445 			 */
2446 			VERIFY(nvlist_alloc(&zct.zct_zplprops,
2447 			    NV_UNIQUE_NAME, KM_SLEEP) == 0);
2448 			error = zfs_fill_zplprops(zc->zc_name, nvprops,
2449 			    zct.zct_zplprops, &is_insensitive);
2450 			if (error != 0) {
2451 				nvlist_free(nvprops);
2452 				nvlist_free(zct.zct_zplprops);
2453 				return (error);
2454 			}
2455 		}
2456 		error = dmu_objset_create(zc->zc_name, type, NULL,
2457 		    is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
2458 		nvlist_free(zct.zct_zplprops);
2459 	}
2460 
2461 	/*
2462 	 * It would be nice to do this atomically.
2463 	 */
2464 	if (error == 0) {
2465 		if ((error = zfs_set_prop_nvlist(zc->zc_name, nvprops)) != 0)
2466 			(void) dmu_objset_destroy(zc->zc_name);
2467 	}
2468 	nvlist_free(nvprops);
2469 	return (error);
2470 }
2471 
2472 /*
2473  * inputs:
2474  * zc_name	name of filesystem
2475  * zc_value	short name of snapshot
2476  * zc_cookie	recursive flag
2477  * zc_nvlist_src[_size] property list
2478  *
2479  * outputs:	none
2480  */
2481 static int
2482 zfs_ioc_snapshot(zfs_cmd_t *zc)
2483 {
2484 	nvlist_t *nvprops = NULL;
2485 	int error;
2486 	boolean_t recursive = zc->zc_cookie;
2487 
2488 	if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
2489 		return (EINVAL);
2490 
2491 	if (zc->zc_nvlist_src != NULL &&
2492 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2493 	    &nvprops)) != 0)
2494 		return (error);
2495 
2496 	error = zfs_check_userprops(zc->zc_name, nvprops);
2497 	if (error)
2498 		goto out;
2499 
2500 	if (nvprops != NULL && nvlist_next_nvpair(nvprops, NULL) != NULL &&
2501 	    zfs_earlier_version(zc->zc_name, SPA_VERSION_SNAP_PROPS)) {
2502 		error = ENOTSUP;
2503 		goto out;
2504 	}
2505 
2506 	error = dmu_objset_snapshot(zc->zc_name, zc->zc_value,
2507 	    nvprops, recursive);
2508 
2509 out:
2510 	nvlist_free(nvprops);
2511 	return (error);
2512 }
2513 
2514 int
2515 zfs_unmount_snap(char *name, void *arg)
2516 {
2517 	vfs_t *vfsp = NULL;
2518 
2519 	if (arg) {
2520 		char *snapname = arg;
2521 		int len = strlen(name) + strlen(snapname) + 2;
2522 		char *buf = kmem_alloc(len, KM_SLEEP);
2523 
2524 		(void) strcpy(buf, name);
2525 		(void) strcat(buf, "@");
2526 		(void) strcat(buf, snapname);
2527 		vfsp = zfs_get_vfs(buf);
2528 		kmem_free(buf, len);
2529 	} else if (strchr(name, '@')) {
2530 		vfsp = zfs_get_vfs(name);
2531 	}
2532 
2533 	if (vfsp) {
2534 		/*
2535 		 * Always force the unmount for snapshots.
2536 		 */
2537 		int flag = MS_FORCE;
2538 		int err;
2539 
2540 		if ((err = vn_vfswlock(vfsp->vfs_vnodecovered)) != 0) {
2541 			VFS_RELE(vfsp);
2542 			return (err);
2543 		}
2544 		VFS_RELE(vfsp);
2545 		if ((err = dounmount(vfsp, flag, kcred)) != 0)
2546 			return (err);
2547 	}
2548 	return (0);
2549 }
2550 
2551 /*
2552  * inputs:
2553  * zc_name	name of filesystem
2554  * zc_value	short name of snapshot
2555  *
2556  * outputs:	none
2557  */
2558 static int
2559 zfs_ioc_destroy_snaps(zfs_cmd_t *zc)
2560 {
2561 	int err;
2562 
2563 	if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
2564 		return (EINVAL);
2565 	err = dmu_objset_find(zc->zc_name,
2566 	    zfs_unmount_snap, zc->zc_value, DS_FIND_CHILDREN);
2567 	if (err)
2568 		return (err);
2569 	return (dmu_snapshots_destroy(zc->zc_name, zc->zc_value));
2570 }
2571 
2572 /*
2573  * inputs:
2574  * zc_name		name of dataset to destroy
2575  * zc_objset_type	type of objset
2576  *
2577  * outputs:		none
2578  */
2579 static int
2580 zfs_ioc_destroy(zfs_cmd_t *zc)
2581 {
2582 	if (strchr(zc->zc_name, '@') && zc->zc_objset_type == DMU_OST_ZFS) {
2583 		int err = zfs_unmount_snap(zc->zc_name, NULL);
2584 		if (err)
2585 			return (err);
2586 	}
2587 
2588 	return (dmu_objset_destroy(zc->zc_name));
2589 }
2590 
2591 /*
2592  * inputs:
2593  * zc_name	name of dataset to rollback (to most recent snapshot)
2594  *
2595  * outputs:	none
2596  */
2597 static int
2598 zfs_ioc_rollback(zfs_cmd_t *zc)
2599 {
2600 	objset_t *os;
2601 	int error;
2602 	zfsvfs_t *zfsvfs = NULL;
2603 
2604 	/*
2605 	 * Get the zfsvfs for the receiving objset. There
2606 	 * won't be one if we're operating on a zvol, if the
2607 	 * objset doesn't exist yet, or is not mounted.
2608 	 */
2609 	error = dmu_objset_open(zc->zc_name, DMU_OST_ANY, DS_MODE_USER, &os);
2610 	if (error)
2611 		return (error);
2612 
2613 	if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
2614 		int mode;
2615 
2616 		error = zfs_suspend_fs(zfsvfs, NULL, &mode);
2617 		if (error == 0) {
2618 			int resume_err;
2619 
2620 			error = dmu_objset_rollback(os);
2621 			resume_err = zfs_resume_fs(zfsvfs, zc->zc_name, mode);
2622 			error = error ? error : resume_err;
2623 		} else {
2624 			dmu_objset_close(os);
2625 		}
2626 		VFS_RELE(zfsvfs->z_vfs);
2627 	} else {
2628 		error = dmu_objset_rollback(os);
2629 	}
2630 	/* Note, the dmu_objset_rollback() releases the objset for us. */
2631 
2632 	return (error);
2633 }
2634 
2635 /*
2636  * inputs:
2637  * zc_name	old name of dataset
2638  * zc_value	new name of dataset
2639  * zc_cookie	recursive flag (only valid for snapshots)
2640  *
2641  * outputs:	none
2642  */
2643 static int
2644 zfs_ioc_rename(zfs_cmd_t *zc)
2645 {
2646 	boolean_t recursive = zc->zc_cookie & 1;
2647 
2648 	zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
2649 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
2650 	    strchr(zc->zc_value, '%'))
2651 		return (EINVAL);
2652 
2653 	/*
2654 	 * Unmount snapshot unless we're doing a recursive rename,
2655 	 * in which case the dataset code figures out which snapshots
2656 	 * to unmount.
2657 	 */
2658 	if (!recursive && strchr(zc->zc_name, '@') != NULL &&
2659 	    zc->zc_objset_type == DMU_OST_ZFS) {
2660 		int err = zfs_unmount_snap(zc->zc_name, NULL);
2661 		if (err)
2662 			return (err);
2663 	}
2664 	return (dmu_objset_rename(zc->zc_name, zc->zc_value, recursive));
2665 }
2666 
2667 static void
2668 clear_props(char *dataset, nvlist_t *props, nvlist_t *newprops)
2669 {
2670 	zfs_cmd_t *zc;
2671 	nvpair_t *prop;
2672 
2673 	if (props == NULL)
2674 		return;
2675 	zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
2676 	(void) strcpy(zc->zc_name, dataset);
2677 	for (prop = nvlist_next_nvpair(props, NULL); prop;
2678 	    prop = nvlist_next_nvpair(props, prop)) {
2679 		if (newprops != NULL &&
2680 		    nvlist_exists(newprops, nvpair_name(prop)))
2681 			continue;
2682 		(void) strcpy(zc->zc_value, nvpair_name(prop));
2683 		if (zfs_secpolicy_inherit(zc, CRED()) == 0)
2684 			(void) zfs_ioc_inherit_prop(zc);
2685 	}
2686 	kmem_free(zc, sizeof (zfs_cmd_t));
2687 }
2688 
2689 /*
2690  * inputs:
2691  * zc_name		name of containing filesystem
2692  * zc_nvlist_src{_size}	nvlist of properties to apply
2693  * zc_value		name of snapshot to create
2694  * zc_string		name of clone origin (if DRR_FLAG_CLONE)
2695  * zc_cookie		file descriptor to recv from
2696  * zc_begin_record	the BEGIN record of the stream (not byteswapped)
2697  * zc_guid		force flag
2698  *
2699  * outputs:
2700  * zc_cookie		number of bytes read
2701  */
2702 static int
2703 zfs_ioc_recv(zfs_cmd_t *zc)
2704 {
2705 	file_t *fp;
2706 	objset_t *os;
2707 	dmu_recv_cookie_t drc;
2708 	zfsvfs_t *zfsvfs = NULL;
2709 	boolean_t force = (boolean_t)zc->zc_guid;
2710 	int error, fd;
2711 	offset_t off;
2712 	nvlist_t *props = NULL;
2713 	nvlist_t *origprops = NULL;
2714 	objset_t *origin = NULL;
2715 	char *tosnap;
2716 	char tofs[ZFS_MAXNAMELEN];
2717 
2718 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
2719 	    strchr(zc->zc_value, '@') == NULL ||
2720 	    strchr(zc->zc_value, '%'))
2721 		return (EINVAL);
2722 
2723 	(void) strcpy(tofs, zc->zc_value);
2724 	tosnap = strchr(tofs, '@');
2725 	*tosnap = '\0';
2726 	tosnap++;
2727 
2728 	if (zc->zc_nvlist_src != NULL &&
2729 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2730 	    &props)) != 0)
2731 		return (error);
2732 
2733 	fd = zc->zc_cookie;
2734 	fp = getf(fd);
2735 	if (fp == NULL) {
2736 		nvlist_free(props);
2737 		return (EBADF);
2738 	}
2739 
2740 	if (getzfsvfs(tofs, &zfsvfs) == 0) {
2741 		if (!mutex_tryenter(&zfsvfs->z_online_recv_lock)) {
2742 			VFS_RELE(zfsvfs->z_vfs);
2743 			zfsvfs = NULL;
2744 			error = EBUSY;
2745 			goto out;
2746 		}
2747 		/*
2748 		 * If new properties are supplied, they are to completely
2749 		 * replace the existing ones, so stash away the existing ones.
2750 		 */
2751 		if (props)
2752 			(void) dsl_prop_get_all(zfsvfs->z_os, &origprops, TRUE);
2753 	} else if (props && dmu_objset_open(tofs, DMU_OST_ANY,
2754 	    DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
2755 		/*
2756 		 * Get the props even if there was no zfsvfs (zvol or
2757 		 * unmounted zpl).
2758 		 */
2759 		(void) dsl_prop_get_all(os, &origprops, TRUE);
2760 
2761 		dmu_objset_close(os);
2762 	}
2763 
2764 	if (zc->zc_string[0]) {
2765 		error = dmu_objset_open(zc->zc_string, DMU_OST_ANY,
2766 		    DS_MODE_USER | DS_MODE_READONLY, &origin);
2767 		if (error)
2768 			goto out;
2769 	}
2770 
2771 	error = dmu_recv_begin(tofs, tosnap, &zc->zc_begin_record,
2772 	    force, origin, zfsvfs != NULL, &drc);
2773 	if (origin)
2774 		dmu_objset_close(origin);
2775 	if (error)
2776 		goto out;
2777 
2778 	/*
2779 	 * Reset properties.  We do this before we receive the stream
2780 	 * so that the properties are applied to the new data.
2781 	 */
2782 	if (props) {
2783 		clear_props(tofs, origprops, props);
2784 		/*
2785 		 * XXX - Note, this is all-or-nothing; should be best-effort.
2786 		 */
2787 		(void) zfs_set_prop_nvlist(tofs, props);
2788 	}
2789 
2790 	off = fp->f_offset;
2791 	error = dmu_recv_stream(&drc, fp->f_vnode, &off);
2792 
2793 	if (error == 0 && zfsvfs) {
2794 		char *osname;
2795 		int mode;
2796 
2797 		/* online recv */
2798 		osname = kmem_alloc(MAXNAMELEN, KM_SLEEP);
2799 		error = zfs_suspend_fs(zfsvfs, osname, &mode);
2800 		if (error == 0) {
2801 			int resume_err;
2802 
2803 			error = dmu_recv_end(&drc);
2804 			resume_err = zfs_resume_fs(zfsvfs, osname, mode);
2805 			error = error ? error : resume_err;
2806 		} else {
2807 			dmu_recv_abort_cleanup(&drc);
2808 		}
2809 		kmem_free(osname, MAXNAMELEN);
2810 	} else if (error == 0) {
2811 		error = dmu_recv_end(&drc);
2812 	}
2813 
2814 	zc->zc_cookie = off - fp->f_offset;
2815 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
2816 		fp->f_offset = off;
2817 
2818 	/*
2819 	 * On error, restore the original props.
2820 	 */
2821 	if (error && props) {
2822 		clear_props(tofs, props, NULL);
2823 		(void) zfs_set_prop_nvlist(tofs, origprops);
2824 	}
2825 out:
2826 	if (zfsvfs) {
2827 		mutex_exit(&zfsvfs->z_online_recv_lock);
2828 		VFS_RELE(zfsvfs->z_vfs);
2829 	}
2830 	nvlist_free(props);
2831 	nvlist_free(origprops);
2832 	releasef(fd);
2833 	return (error);
2834 }
2835 
2836 /*
2837  * inputs:
2838  * zc_name	name of snapshot to send
2839  * zc_value	short name of incremental fromsnap (may be empty)
2840  * zc_cookie	file descriptor to send stream to
2841  * zc_obj	fromorigin flag (mutually exclusive with zc_value)
2842  *
2843  * outputs: none
2844  */
2845 static int
2846 zfs_ioc_send(zfs_cmd_t *zc)
2847 {
2848 	objset_t *fromsnap = NULL;
2849 	objset_t *tosnap;
2850 	file_t *fp;
2851 	int error;
2852 	offset_t off;
2853 
2854 	error = dmu_objset_open(zc->zc_name, DMU_OST_ANY,
2855 	    DS_MODE_USER | DS_MODE_READONLY, &tosnap);
2856 	if (error)
2857 		return (error);
2858 
2859 	if (zc->zc_value[0] != '\0') {
2860 		char *buf;
2861 		char *cp;
2862 
2863 		buf = kmem_alloc(MAXPATHLEN, KM_SLEEP);
2864 		(void) strncpy(buf, zc->zc_name, MAXPATHLEN);
2865 		cp = strchr(buf, '@');
2866 		if (cp)
2867 			*(cp+1) = 0;
2868 		(void) strncat(buf, zc->zc_value, MAXPATHLEN);
2869 		error = dmu_objset_open(buf, DMU_OST_ANY,
2870 		    DS_MODE_USER | DS_MODE_READONLY, &fromsnap);
2871 		kmem_free(buf, MAXPATHLEN);
2872 		if (error) {
2873 			dmu_objset_close(tosnap);
2874 			return (error);
2875 		}
2876 	}
2877 
2878 	fp = getf(zc->zc_cookie);
2879 	if (fp == NULL) {
2880 		dmu_objset_close(tosnap);
2881 		if (fromsnap)
2882 			dmu_objset_close(fromsnap);
2883 		return (EBADF);
2884 	}
2885 
2886 	off = fp->f_offset;
2887 	error = dmu_sendbackup(tosnap, fromsnap, zc->zc_obj, fp->f_vnode, &off);
2888 
2889 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
2890 		fp->f_offset = off;
2891 	releasef(zc->zc_cookie);
2892 	if (fromsnap)
2893 		dmu_objset_close(fromsnap);
2894 	dmu_objset_close(tosnap);
2895 	return (error);
2896 }
2897 
2898 static int
2899 zfs_ioc_inject_fault(zfs_cmd_t *zc)
2900 {
2901 	int id, error;
2902 
2903 	error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
2904 	    &zc->zc_inject_record);
2905 
2906 	if (error == 0)
2907 		zc->zc_guid = (uint64_t)id;
2908 
2909 	return (error);
2910 }
2911 
2912 static int
2913 zfs_ioc_clear_fault(zfs_cmd_t *zc)
2914 {
2915 	return (zio_clear_fault((int)zc->zc_guid));
2916 }
2917 
2918 static int
2919 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
2920 {
2921 	int id = (int)zc->zc_guid;
2922 	int error;
2923 
2924 	error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
2925 	    &zc->zc_inject_record);
2926 
2927 	zc->zc_guid = id;
2928 
2929 	return (error);
2930 }
2931 
2932 static int
2933 zfs_ioc_error_log(zfs_cmd_t *zc)
2934 {
2935 	spa_t *spa;
2936 	int error;
2937 	size_t count = (size_t)zc->zc_nvlist_dst_size;
2938 
2939 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2940 		return (error);
2941 
2942 	error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
2943 	    &count);
2944 	if (error == 0)
2945 		zc->zc_nvlist_dst_size = count;
2946 	else
2947 		zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
2948 
2949 	spa_close(spa, FTAG);
2950 
2951 	return (error);
2952 }
2953 
2954 static int
2955 zfs_ioc_clear(zfs_cmd_t *zc)
2956 {
2957 	spa_t *spa;
2958 	vdev_t *vd;
2959 	int error;
2960 
2961 	/*
2962 	 * On zpool clear we also fix up missing slogs
2963 	 */
2964 	mutex_enter(&spa_namespace_lock);
2965 	spa = spa_lookup(zc->zc_name);
2966 	if (spa == NULL) {
2967 		mutex_exit(&spa_namespace_lock);
2968 		return (EIO);
2969 	}
2970 	if (spa->spa_log_state == SPA_LOG_MISSING) {
2971 		/* we need to let spa_open/spa_load clear the chains */
2972 		spa->spa_log_state = SPA_LOG_CLEAR;
2973 	}
2974 	mutex_exit(&spa_namespace_lock);
2975 
2976 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2977 		return (error);
2978 
2979 	spa_vdev_state_enter(spa);
2980 
2981 	if (zc->zc_guid == 0) {
2982 		vd = NULL;
2983 	} else {
2984 		vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
2985 		if (vd == NULL) {
2986 			(void) spa_vdev_state_exit(spa, NULL, ENODEV);
2987 			spa_close(spa, FTAG);
2988 			return (ENODEV);
2989 		}
2990 	}
2991 
2992 	vdev_clear(spa, vd);
2993 
2994 	(void) spa_vdev_state_exit(spa, NULL, 0);
2995 
2996 	/*
2997 	 * Resume any suspended I/Os.
2998 	 */
2999 	if (zio_resume(spa) != 0)
3000 		error = EIO;
3001 
3002 	spa_close(spa, FTAG);
3003 
3004 	return (error);
3005 }
3006 
3007 /*
3008  * inputs:
3009  * zc_name	name of filesystem
3010  * zc_value	name of origin snapshot
3011  *
3012  * outputs:	none
3013  */
3014 static int
3015 zfs_ioc_promote(zfs_cmd_t *zc)
3016 {
3017 	char *cp;
3018 
3019 	/*
3020 	 * We don't need to unmount *all* the origin fs's snapshots, but
3021 	 * it's easier.
3022 	 */
3023 	cp = strchr(zc->zc_value, '@');
3024 	if (cp)
3025 		*cp = '\0';
3026 	(void) dmu_objset_find(zc->zc_value,
3027 	    zfs_unmount_snap, NULL, DS_FIND_SNAPSHOTS);
3028 	return (dsl_dataset_promote(zc->zc_name));
3029 }
3030 
3031 /*
3032  * Retrieve a single {user|group}{used|quota}@... property.
3033  *
3034  * inputs:
3035  * zc_name	name of filesystem
3036  * zc_objset_type zfs_userquota_prop_t
3037  * zc_value	domain name (eg. "S-1-234-567-89")
3038  * zc_guid	RID/UID/GID
3039  *
3040  * outputs:
3041  * zc_cookie	property value
3042  */
3043 static int
3044 zfs_ioc_userspace_one(zfs_cmd_t *zc)
3045 {
3046 	zfsvfs_t *zfsvfs;
3047 	int error;
3048 
3049 	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
3050 		return (EINVAL);
3051 
3052 	error = zfsvfs_hold(zc->zc_name, B_TRUE, FTAG, &zfsvfs);
3053 	if (error)
3054 		return (error);
3055 
3056 	error = zfs_userspace_one(zfsvfs,
3057 	    zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
3058 	zfsvfs_rele(zfsvfs, FTAG);
3059 
3060 	return (error);
3061 }
3062 
3063 /*
3064  * inputs:
3065  * zc_name		name of filesystem
3066  * zc_cookie		zap cursor
3067  * zc_objset_type	zfs_userquota_prop_t
3068  * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
3069  *
3070  * outputs:
3071  * zc_nvlist_dst[_size]	data buffer (array of zfs_useracct_t)
3072  * zc_cookie	zap cursor
3073  */
3074 static int
3075 zfs_ioc_userspace_many(zfs_cmd_t *zc)
3076 {
3077 	zfsvfs_t *zfsvfs;
3078 	int error;
3079 
3080 	error = zfsvfs_hold(zc->zc_name, B_TRUE, FTAG, &zfsvfs);
3081 	if (error)
3082 		return (error);
3083 
3084 	int bufsize = zc->zc_nvlist_dst_size;
3085 	void *buf = kmem_alloc(bufsize, KM_SLEEP);
3086 
3087 	error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
3088 	    buf, &zc->zc_nvlist_dst_size);
3089 
3090 	if (error == 0) {
3091 		error = xcopyout(buf,
3092 		    (void *)(uintptr_t)zc->zc_nvlist_dst,
3093 		    zc->zc_nvlist_dst_size);
3094 	}
3095 	kmem_free(buf, bufsize);
3096 	zfsvfs_rele(zfsvfs, FTAG);
3097 
3098 	return (error);
3099 }
3100 
3101 /*
3102  * inputs:
3103  * zc_name		name of filesystem
3104  *
3105  * outputs:
3106  * none
3107  */
3108 static int
3109 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
3110 {
3111 	objset_t *os;
3112 	int error;
3113 	zfsvfs_t *zfsvfs;
3114 
3115 	if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
3116 		if (!dmu_objset_userused_enabled(zfsvfs->z_os->os)) {
3117 			/*
3118 			 * If userused is not enabled, it may be because the
3119 			 * objset needs to be closed & reopened (to grow the
3120 			 * objset_phys_t).  Suspend/resume the fs will do that.
3121 			 */
3122 			int mode;
3123 			error = zfs_suspend_fs(zfsvfs, NULL, &mode);
3124 			if (error == 0) {
3125 				error = zfs_resume_fs(zfsvfs,
3126 				    zc->zc_name, mode);
3127 			}
3128 		}
3129 		if (error == 0)
3130 			error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
3131 		VFS_RELE(zfsvfs->z_vfs);
3132 	} else {
3133 		error = dmu_objset_open(zc->zc_name, DMU_OST_ANY,
3134 		    DS_MODE_USER, &os);
3135 		if (error)
3136 			return (error);
3137 
3138 		error = dmu_objset_userspace_upgrade(os);
3139 		dmu_objset_close(os);
3140 	}
3141 
3142 	return (error);
3143 }
3144 
3145 /*
3146  * We don't want to have a hard dependency
3147  * against some special symbols in sharefs
3148  * nfs, and smbsrv.  Determine them if needed when
3149  * the first file system is shared.
3150  * Neither sharefs, nfs or smbsrv are unloadable modules.
3151  */
3152 int (*znfsexport_fs)(void *arg);
3153 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
3154 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
3155 
3156 int zfs_nfsshare_inited;
3157 int zfs_smbshare_inited;
3158 
3159 ddi_modhandle_t nfs_mod;
3160 ddi_modhandle_t sharefs_mod;
3161 ddi_modhandle_t smbsrv_mod;
3162 kmutex_t zfs_share_lock;
3163 
3164 static int
3165 zfs_init_sharefs()
3166 {
3167 	int error;
3168 
3169 	ASSERT(MUTEX_HELD(&zfs_share_lock));
3170 	/* Both NFS and SMB shares also require sharetab support. */
3171 	if (sharefs_mod == NULL && ((sharefs_mod =
3172 	    ddi_modopen("fs/sharefs",
3173 	    KRTLD_MODE_FIRST, &error)) == NULL)) {
3174 		return (ENOSYS);
3175 	}
3176 	if (zshare_fs == NULL && ((zshare_fs =
3177 	    (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
3178 	    ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
3179 		return (ENOSYS);
3180 	}
3181 	return (0);
3182 }
3183 
3184 static int
3185 zfs_ioc_share(zfs_cmd_t *zc)
3186 {
3187 	int error;
3188 	int opcode;
3189 
3190 	switch (zc->zc_share.z_sharetype) {
3191 	case ZFS_SHARE_NFS:
3192 	case ZFS_UNSHARE_NFS:
3193 		if (zfs_nfsshare_inited == 0) {
3194 			mutex_enter(&zfs_share_lock);
3195 			if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
3196 			    KRTLD_MODE_FIRST, &error)) == NULL)) {
3197 				mutex_exit(&zfs_share_lock);
3198 				return (ENOSYS);
3199 			}
3200 			if (znfsexport_fs == NULL &&
3201 			    ((znfsexport_fs = (int (*)(void *))
3202 			    ddi_modsym(nfs_mod,
3203 			    "nfs_export", &error)) == NULL)) {
3204 				mutex_exit(&zfs_share_lock);
3205 				return (ENOSYS);
3206 			}
3207 			error = zfs_init_sharefs();
3208 			if (error) {
3209 				mutex_exit(&zfs_share_lock);
3210 				return (ENOSYS);
3211 			}
3212 			zfs_nfsshare_inited = 1;
3213 			mutex_exit(&zfs_share_lock);
3214 		}
3215 		break;
3216 	case ZFS_SHARE_SMB:
3217 	case ZFS_UNSHARE_SMB:
3218 		if (zfs_smbshare_inited == 0) {
3219 			mutex_enter(&zfs_share_lock);
3220 			if (smbsrv_mod == NULL && ((smbsrv_mod =
3221 			    ddi_modopen("drv/smbsrv",
3222 			    KRTLD_MODE_FIRST, &error)) == NULL)) {
3223 				mutex_exit(&zfs_share_lock);
3224 				return (ENOSYS);
3225 			}
3226 			if (zsmbexport_fs == NULL && ((zsmbexport_fs =
3227 			    (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
3228 			    "smb_server_share", &error)) == NULL)) {
3229 				mutex_exit(&zfs_share_lock);
3230 				return (ENOSYS);
3231 			}
3232 			error = zfs_init_sharefs();
3233 			if (error) {
3234 				mutex_exit(&zfs_share_lock);
3235 				return (ENOSYS);
3236 			}
3237 			zfs_smbshare_inited = 1;
3238 			mutex_exit(&zfs_share_lock);
3239 		}
3240 		break;
3241 	default:
3242 		return (EINVAL);
3243 	}
3244 
3245 	switch (zc->zc_share.z_sharetype) {
3246 	case ZFS_SHARE_NFS:
3247 	case ZFS_UNSHARE_NFS:
3248 		if (error =
3249 		    znfsexport_fs((void *)
3250 		    (uintptr_t)zc->zc_share.z_exportdata))
3251 			return (error);
3252 		break;
3253 	case ZFS_SHARE_SMB:
3254 	case ZFS_UNSHARE_SMB:
3255 		if (error = zsmbexport_fs((void *)
3256 		    (uintptr_t)zc->zc_share.z_exportdata,
3257 		    zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
3258 		    B_TRUE: B_FALSE)) {
3259 			return (error);
3260 		}
3261 		break;
3262 	}
3263 
3264 	opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
3265 	    zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
3266 	    SHAREFS_ADD : SHAREFS_REMOVE;
3267 
3268 	/*
3269 	 * Add or remove share from sharetab
3270 	 */
3271 	error = zshare_fs(opcode,
3272 	    (void *)(uintptr_t)zc->zc_share.z_sharedata,
3273 	    zc->zc_share.z_sharemax);
3274 
3275 	return (error);
3276 
3277 }
3278 
3279 ace_t full_access[] = {
3280 	{(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
3281 };
3282 
3283 /*
3284  * Remove all ACL files in shares dir
3285  */
3286 static int
3287 zfs_smb_acl_purge(znode_t *dzp)
3288 {
3289 	zap_cursor_t	zc;
3290 	zap_attribute_t	zap;
3291 	zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
3292 	int error;
3293 
3294 	for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
3295 	    (error = zap_cursor_retrieve(&zc, &zap)) == 0;
3296 	    zap_cursor_advance(&zc)) {
3297 		if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
3298 		    NULL, 0)) != 0)
3299 			break;
3300 	}
3301 	zap_cursor_fini(&zc);
3302 	return (error);
3303 }
3304 
3305 static int
3306 zfs_ioc_smb_acl(zfs_cmd_t *zc)
3307 {
3308 	vnode_t *vp;
3309 	znode_t *dzp;
3310 	vnode_t *resourcevp = NULL;
3311 	znode_t *sharedir;
3312 	zfsvfs_t *zfsvfs;
3313 	nvlist_t *nvlist;
3314 	char *src, *target;
3315 	vattr_t vattr;
3316 	vsecattr_t vsec;
3317 	int error = 0;
3318 
3319 	if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
3320 	    NO_FOLLOW, NULL, &vp)) != 0)
3321 		return (error);
3322 
3323 	/* Now make sure mntpnt and dataset are ZFS */
3324 
3325 	if (vp->v_vfsp->vfs_fstype != zfsfstype ||
3326 	    (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
3327 	    zc->zc_name) != 0)) {
3328 		VN_RELE(vp);
3329 		return (EINVAL);
3330 	}
3331 
3332 	dzp = VTOZ(vp);
3333 	zfsvfs = dzp->z_zfsvfs;
3334 	ZFS_ENTER(zfsvfs);
3335 
3336 	/*
3337 	 * Create share dir if its missing.
3338 	 */
3339 	mutex_enter(&zfsvfs->z_lock);
3340 	if (zfsvfs->z_shares_dir == 0) {
3341 		dmu_tx_t *tx;
3342 
3343 		tx = dmu_tx_create(zfsvfs->z_os);
3344 		dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
3345 		    ZFS_SHARES_DIR);
3346 		dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
3347 		error = dmu_tx_assign(tx, TXG_WAIT);
3348 		if (error) {
3349 			dmu_tx_abort(tx);
3350 		} else {
3351 			error = zfs_create_share_dir(zfsvfs, tx);
3352 			dmu_tx_commit(tx);
3353 		}
3354 		if (error) {
3355 			mutex_exit(&zfsvfs->z_lock);
3356 			VN_RELE(vp);
3357 			ZFS_EXIT(zfsvfs);
3358 			return (error);
3359 		}
3360 	}
3361 	mutex_exit(&zfsvfs->z_lock);
3362 
3363 	ASSERT(zfsvfs->z_shares_dir);
3364 	if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
3365 		VN_RELE(vp);
3366 		ZFS_EXIT(zfsvfs);
3367 		return (error);
3368 	}
3369 
3370 	switch (zc->zc_cookie) {
3371 	case ZFS_SMB_ACL_ADD:
3372 		vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
3373 		vattr.va_type = VREG;
3374 		vattr.va_mode = S_IFREG|0777;
3375 		vattr.va_uid = 0;
3376 		vattr.va_gid = 0;
3377 
3378 		vsec.vsa_mask = VSA_ACE;
3379 		vsec.vsa_aclentp = &full_access;
3380 		vsec.vsa_aclentsz = sizeof (full_access);
3381 		vsec.vsa_aclcnt = 1;
3382 
3383 		error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
3384 		    &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
3385 		if (resourcevp)
3386 			VN_RELE(resourcevp);
3387 		break;
3388 
3389 	case ZFS_SMB_ACL_REMOVE:
3390 		error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
3391 		    NULL, 0);
3392 		break;
3393 
3394 	case ZFS_SMB_ACL_RENAME:
3395 		if ((error = get_nvlist(zc->zc_nvlist_src,
3396 		    zc->zc_nvlist_src_size, &nvlist)) != 0) {
3397 			VN_RELE(vp);
3398 			ZFS_EXIT(zfsvfs);
3399 			return (error);
3400 		}
3401 		if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
3402 		    nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
3403 		    &target)) {
3404 			VN_RELE(vp);
3405 			VN_RELE(ZTOV(sharedir));
3406 			ZFS_EXIT(zfsvfs);
3407 			return (error);
3408 		}
3409 		error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
3410 		    kcred, NULL, 0);
3411 		nvlist_free(nvlist);
3412 		break;
3413 
3414 	case ZFS_SMB_ACL_PURGE:
3415 		error = zfs_smb_acl_purge(sharedir);
3416 		break;
3417 
3418 	default:
3419 		error = EINVAL;
3420 		break;
3421 	}
3422 
3423 	VN_RELE(vp);
3424 	VN_RELE(ZTOV(sharedir));
3425 
3426 	ZFS_EXIT(zfsvfs);
3427 
3428 	return (error);
3429 }
3430 
3431 /*
3432  * pool create, destroy, and export don't log the history as part of
3433  * zfsdev_ioctl, but rather zfs_ioc_pool_create, and zfs_ioc_pool_export
3434  * do the logging of those commands.
3435  */
3436 static zfs_ioc_vec_t zfs_ioc_vec[] = {
3437 	{ zfs_ioc_pool_create, zfs_secpolicy_config, POOL_NAME, B_FALSE,
3438 	    B_FALSE },
3439 	{ zfs_ioc_pool_destroy,	zfs_secpolicy_config, POOL_NAME, B_FALSE,
3440 	    B_FALSE },
3441 	{ zfs_ioc_pool_import, zfs_secpolicy_config, POOL_NAME, B_TRUE,
3442 	    B_FALSE },
3443 	{ zfs_ioc_pool_export, zfs_secpolicy_config, POOL_NAME, B_FALSE,
3444 	    B_FALSE },
3445 	{ zfs_ioc_pool_configs,	zfs_secpolicy_none, NO_NAME, B_FALSE,
3446 	    B_FALSE },
3447 	{ zfs_ioc_pool_stats, zfs_secpolicy_read, POOL_NAME, B_FALSE,
3448 	    B_FALSE },
3449 	{ zfs_ioc_pool_tryimport, zfs_secpolicy_config, NO_NAME, B_FALSE,
3450 	    B_FALSE },
3451 	{ zfs_ioc_pool_scrub, zfs_secpolicy_config, POOL_NAME, B_TRUE,
3452 	    B_TRUE },
3453 	{ zfs_ioc_pool_freeze, zfs_secpolicy_config, NO_NAME, B_FALSE,
3454 	    B_FALSE },
3455 	{ zfs_ioc_pool_upgrade,	zfs_secpolicy_config, POOL_NAME, B_TRUE,
3456 	    B_TRUE },
3457 	{ zfs_ioc_pool_get_history, zfs_secpolicy_config, POOL_NAME, B_FALSE,
3458 	    B_FALSE },
3459 	{ zfs_ioc_vdev_add, zfs_secpolicy_config, POOL_NAME, B_TRUE,
3460 	    B_TRUE },
3461 	{ zfs_ioc_vdev_remove, zfs_secpolicy_config, POOL_NAME, B_TRUE,
3462 	    B_TRUE },
3463 	{ zfs_ioc_vdev_set_state, zfs_secpolicy_config,	POOL_NAME, B_TRUE,
3464 	    B_FALSE },
3465 	{ zfs_ioc_vdev_attach, zfs_secpolicy_config, POOL_NAME, B_TRUE,
3466 	    B_TRUE },
3467 	{ zfs_ioc_vdev_detach, zfs_secpolicy_config, POOL_NAME, B_TRUE,
3468 	    B_TRUE },
3469 	{ zfs_ioc_vdev_setpath,	zfs_secpolicy_config, POOL_NAME, B_FALSE,
3470 	    B_TRUE },
3471 	{ zfs_ioc_vdev_setfru,	zfs_secpolicy_config, POOL_NAME, B_FALSE,
3472 	    B_TRUE },
3473 	{ zfs_ioc_objset_stats,	zfs_secpolicy_read, DATASET_NAME, B_FALSE,
3474 	    B_FALSE },
3475 	{ zfs_ioc_objset_zplprops, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
3476 	    B_FALSE },
3477 	{ zfs_ioc_dataset_list_next, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
3478 	    B_FALSE },
3479 	{ zfs_ioc_snapshot_list_next, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
3480 	    B_FALSE },
3481 	{ zfs_ioc_set_prop, zfs_secpolicy_none, DATASET_NAME, B_TRUE, B_TRUE },
3482 	{ zfs_ioc_create_minor,	zfs_secpolicy_minor, DATASET_NAME, B_FALSE,
3483 	    B_FALSE },
3484 	{ zfs_ioc_remove_minor,	zfs_secpolicy_minor, DATASET_NAME, B_FALSE,
3485 	    B_FALSE },
3486 	{ zfs_ioc_create, zfs_secpolicy_create, DATASET_NAME, B_TRUE, B_TRUE },
3487 	{ zfs_ioc_destroy, zfs_secpolicy_destroy, DATASET_NAME, B_TRUE,
3488 	    B_TRUE},
3489 	{ zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, B_TRUE,
3490 	    B_TRUE },
3491 	{ zfs_ioc_rename, zfs_secpolicy_rename,	DATASET_NAME, B_TRUE, B_TRUE },
3492 	{ zfs_ioc_recv, zfs_secpolicy_receive, DATASET_NAME, B_TRUE, B_TRUE },
3493 	{ zfs_ioc_send, zfs_secpolicy_send, DATASET_NAME, B_TRUE, B_FALSE },
3494 	{ zfs_ioc_inject_fault,	zfs_secpolicy_inject, NO_NAME, B_FALSE,
3495 	    B_FALSE },
3496 	{ zfs_ioc_clear_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE,
3497 	    B_FALSE },
3498 	{ zfs_ioc_inject_list_next, zfs_secpolicy_inject, NO_NAME, B_FALSE,
3499 	    B_FALSE },
3500 	{ zfs_ioc_error_log, zfs_secpolicy_inject, POOL_NAME, B_FALSE,
3501 	    B_FALSE },
3502 	{ zfs_ioc_clear, zfs_secpolicy_config, POOL_NAME, B_TRUE, B_FALSE },
3503 	{ zfs_ioc_promote, zfs_secpolicy_promote, DATASET_NAME, B_TRUE,
3504 	    B_TRUE },
3505 	{ zfs_ioc_destroy_snaps, zfs_secpolicy_destroy,	DATASET_NAME, B_TRUE,
3506 	    B_TRUE },
3507 	{ zfs_ioc_snapshot, zfs_secpolicy_snapshot, DATASET_NAME, B_TRUE,
3508 	    B_TRUE },
3509 	{ zfs_ioc_dsobj_to_dsname, zfs_secpolicy_config, POOL_NAME, B_FALSE,
3510 	    B_FALSE },
3511 	{ zfs_ioc_obj_to_path, zfs_secpolicy_config, NO_NAME, B_FALSE,
3512 	    B_FALSE },
3513 	{ zfs_ioc_pool_set_props, zfs_secpolicy_config,	POOL_NAME, B_TRUE,
3514 	    B_TRUE },
3515 	{ zfs_ioc_pool_get_props, zfs_secpolicy_read, POOL_NAME, B_FALSE,
3516 	    B_FALSE },
3517 	{ zfs_ioc_set_fsacl, zfs_secpolicy_fsacl, DATASET_NAME, B_TRUE,
3518 	    B_TRUE },
3519 	{ zfs_ioc_get_fsacl, zfs_secpolicy_read, DATASET_NAME, B_FALSE,
3520 	    B_FALSE },
3521 	{ zfs_ioc_iscsi_perm_check, zfs_secpolicy_iscsi, DATASET_NAME, B_FALSE,
3522 	    B_FALSE },
3523 	{ zfs_ioc_share, zfs_secpolicy_share, DATASET_NAME, B_FALSE, B_FALSE },
3524 	{ zfs_ioc_inherit_prop, zfs_secpolicy_inherit, DATASET_NAME, B_TRUE,
3525 	    B_TRUE },
3526 	{ zfs_ioc_smb_acl, zfs_secpolicy_smb_acl, DATASET_NAME, B_FALSE,
3527 	    B_FALSE },
3528 	{ zfs_ioc_userspace_one, zfs_secpolicy_userspace_one,
3529 	    DATASET_NAME, B_FALSE, B_FALSE },
3530 	{ zfs_ioc_userspace_many, zfs_secpolicy_userspace_many,
3531 	    DATASET_NAME, B_FALSE, B_FALSE },
3532 	{ zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
3533 	    DATASET_NAME, B_FALSE, B_TRUE },
3534 };
3535 
3536 int
3537 pool_status_check(const char *name, zfs_ioc_namecheck_t type)
3538 {
3539 	spa_t *spa;
3540 	int error;
3541 
3542 	ASSERT(type == POOL_NAME || type == DATASET_NAME);
3543 
3544 	error = spa_open(name, &spa, FTAG);
3545 	if (error == 0) {
3546 		if (spa_suspended(spa))
3547 			error = EAGAIN;
3548 		spa_close(spa, FTAG);
3549 	}
3550 	return (error);
3551 }
3552 
3553 static int
3554 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
3555 {
3556 	zfs_cmd_t *zc;
3557 	uint_t vec;
3558 	int error, rc;
3559 
3560 	if (getminor(dev) != 0)
3561 		return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
3562 
3563 	vec = cmd - ZFS_IOC;
3564 	ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
3565 
3566 	if (vec >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
3567 		return (EINVAL);
3568 
3569 	zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
3570 
3571 	error = xcopyin((void *)arg, zc, sizeof (zfs_cmd_t));
3572 
3573 	if (error == 0)
3574 		error = zfs_ioc_vec[vec].zvec_secpolicy(zc, cr);
3575 
3576 	/*
3577 	 * Ensure that all pool/dataset names are valid before we pass down to
3578 	 * the lower layers.
3579 	 */
3580 	if (error == 0) {
3581 		zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
3582 		switch (zfs_ioc_vec[vec].zvec_namecheck) {
3583 		case POOL_NAME:
3584 			if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
3585 				error = EINVAL;
3586 			if (zfs_ioc_vec[vec].zvec_pool_check)
3587 				error = pool_status_check(zc->zc_name,
3588 				    zfs_ioc_vec[vec].zvec_namecheck);
3589 			break;
3590 
3591 		case DATASET_NAME:
3592 			if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
3593 				error = EINVAL;
3594 			if (zfs_ioc_vec[vec].zvec_pool_check)
3595 				error = pool_status_check(zc->zc_name,
3596 				    zfs_ioc_vec[vec].zvec_namecheck);
3597 			break;
3598 
3599 		case NO_NAME:
3600 			break;
3601 		}
3602 	}
3603 
3604 	if (error == 0)
3605 		error = zfs_ioc_vec[vec].zvec_func(zc);
3606 
3607 	rc = xcopyout(zc, (void *)arg, sizeof (zfs_cmd_t));
3608 	if (error == 0) {
3609 		error = rc;
3610 		if (zfs_ioc_vec[vec].zvec_his_log)
3611 			zfs_log_history(zc);
3612 	}
3613 
3614 	kmem_free(zc, sizeof (zfs_cmd_t));
3615 	return (error);
3616 }
3617 
3618 static int
3619 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
3620 {
3621 	if (cmd != DDI_ATTACH)
3622 		return (DDI_FAILURE);
3623 
3624 	if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
3625 	    DDI_PSEUDO, 0) == DDI_FAILURE)
3626 		return (DDI_FAILURE);
3627 
3628 	zfs_dip = dip;
3629 
3630 	ddi_report_dev(dip);
3631 
3632 	return (DDI_SUCCESS);
3633 }
3634 
3635 static int
3636 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
3637 {
3638 	if (spa_busy() || zfs_busy() || zvol_busy())
3639 		return (DDI_FAILURE);
3640 
3641 	if (cmd != DDI_DETACH)
3642 		return (DDI_FAILURE);
3643 
3644 	zfs_dip = NULL;
3645 
3646 	ddi_prop_remove_all(dip);
3647 	ddi_remove_minor_node(dip, NULL);
3648 
3649 	return (DDI_SUCCESS);
3650 }
3651 
3652 /*ARGSUSED*/
3653 static int
3654 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
3655 {
3656 	switch (infocmd) {
3657 	case DDI_INFO_DEVT2DEVINFO:
3658 		*result = zfs_dip;
3659 		return (DDI_SUCCESS);
3660 
3661 	case DDI_INFO_DEVT2INSTANCE:
3662 		*result = (void *)0;
3663 		return (DDI_SUCCESS);
3664 	}
3665 
3666 	return (DDI_FAILURE);
3667 }
3668 
3669 /*
3670  * OK, so this is a little weird.
3671  *
3672  * /dev/zfs is the control node, i.e. minor 0.
3673  * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
3674  *
3675  * /dev/zfs has basically nothing to do except serve up ioctls,
3676  * so most of the standard driver entry points are in zvol.c.
3677  */
3678 static struct cb_ops zfs_cb_ops = {
3679 	zvol_open,	/* open */
3680 	zvol_close,	/* close */
3681 	zvol_strategy,	/* strategy */
3682 	nodev,		/* print */
3683 	zvol_dump,	/* dump */
3684 	zvol_read,	/* read */
3685 	zvol_write,	/* write */
3686 	zfsdev_ioctl,	/* ioctl */
3687 	nodev,		/* devmap */
3688 	nodev,		/* mmap */
3689 	nodev,		/* segmap */
3690 	nochpoll,	/* poll */
3691 	ddi_prop_op,	/* prop_op */
3692 	NULL,		/* streamtab */
3693 	D_NEW | D_MP | D_64BIT,		/* Driver compatibility flag */
3694 	CB_REV,		/* version */
3695 	nodev,		/* async read */
3696 	nodev,		/* async write */
3697 };
3698 
3699 static struct dev_ops zfs_dev_ops = {
3700 	DEVO_REV,	/* version */
3701 	0,		/* refcnt */
3702 	zfs_info,	/* info */
3703 	nulldev,	/* identify */
3704 	nulldev,	/* probe */
3705 	zfs_attach,	/* attach */
3706 	zfs_detach,	/* detach */
3707 	nodev,		/* reset */
3708 	&zfs_cb_ops,	/* driver operations */
3709 	NULL,		/* no bus operations */
3710 	NULL,		/* power */
3711 	ddi_quiesce_not_needed,	/* quiesce */
3712 };
3713 
3714 static struct modldrv zfs_modldrv = {
3715 	&mod_driverops,
3716 	"ZFS storage pool",
3717 	&zfs_dev_ops
3718 };
3719 
3720 static struct modlinkage modlinkage = {
3721 	MODREV_1,
3722 	(void *)&zfs_modlfs,
3723 	(void *)&zfs_modldrv,
3724 	NULL
3725 };
3726 
3727 
3728 uint_t zfs_fsyncer_key;
3729 extern uint_t rrw_tsd_key;
3730 
3731 int
3732 _init(void)
3733 {
3734 	int error;
3735 
3736 	spa_init(FREAD | FWRITE);
3737 	zfs_init();
3738 	zvol_init();
3739 
3740 	if ((error = mod_install(&modlinkage)) != 0) {
3741 		zvol_fini();
3742 		zfs_fini();
3743 		spa_fini();
3744 		return (error);
3745 	}
3746 
3747 	tsd_create(&zfs_fsyncer_key, NULL);
3748 	tsd_create(&rrw_tsd_key, NULL);
3749 
3750 	error = ldi_ident_from_mod(&modlinkage, &zfs_li);
3751 	ASSERT(error == 0);
3752 	mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
3753 
3754 	return (0);
3755 }
3756 
3757 int
3758 _fini(void)
3759 {
3760 	int error;
3761 
3762 	if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
3763 		return (EBUSY);
3764 
3765 	if ((error = mod_remove(&modlinkage)) != 0)
3766 		return (error);
3767 
3768 	zvol_fini();
3769 	zfs_fini();
3770 	spa_fini();
3771 	if (zfs_nfsshare_inited)
3772 		(void) ddi_modclose(nfs_mod);
3773 	if (zfs_smbshare_inited)
3774 		(void) ddi_modclose(smbsrv_mod);
3775 	if (zfs_nfsshare_inited || zfs_smbshare_inited)
3776 		(void) ddi_modclose(sharefs_mod);
3777 
3778 	tsd_destroy(&zfs_fsyncer_key);
3779 	ldi_ident_release(zfs_li);
3780 	zfs_li = NULL;
3781 	mutex_destroy(&zfs_share_lock);
3782 
3783 	return (error);
3784 }
3785 
3786 int
3787 _info(struct modinfo *modinfop)
3788 {
3789 	return (mod_info(&modlinkage, modinfop));
3790 }
3791