xref: /illumos-gate/usr/src/uts/common/fs/zfs/zfs_ioctl.c (revision 63ea9ad24896f2939472f8f96f568086d190eb33)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 /*
22  * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 #include <sys/types.h>
27 #include <sys/param.h>
28 #include <sys/errno.h>
29 #include <sys/uio.h>
30 #include <sys/buf.h>
31 #include <sys/modctl.h>
32 #include <sys/open.h>
33 #include <sys/file.h>
34 #include <sys/kmem.h>
35 #include <sys/conf.h>
36 #include <sys/cmn_err.h>
37 #include <sys/stat.h>
38 #include <sys/zfs_ioctl.h>
39 #include <sys/zfs_znode.h>
40 #include <sys/zap.h>
41 #include <sys/spa.h>
42 #include <sys/spa_impl.h>
43 #include <sys/vdev.h>
44 #include <sys/vdev_impl.h>
45 #include <sys/dmu.h>
46 #include <sys/dsl_dir.h>
47 #include <sys/dsl_dataset.h>
48 #include <sys/dsl_prop.h>
49 #include <sys/dsl_deleg.h>
50 #include <sys/dmu_objset.h>
51 #include <sys/ddi.h>
52 #include <sys/sunddi.h>
53 #include <sys/sunldi.h>
54 #include <sys/policy.h>
55 #include <sys/zone.h>
56 #include <sys/nvpair.h>
57 #include <sys/pathname.h>
58 #include <sys/mount.h>
59 #include <sys/sdt.h>
60 #include <sys/fs/zfs.h>
61 #include <sys/zfs_ctldir.h>
62 #include <sys/zfs_dir.h>
63 #include <sys/zvol.h>
64 #include <sharefs/share.h>
65 #include <sys/dmu_objset.h>
66 
67 #include "zfs_namecheck.h"
68 #include "zfs_prop.h"
69 #include "zfs_deleg.h"
70 
71 extern struct modlfs zfs_modlfs;
72 
73 extern void zfs_init(void);
74 extern void zfs_fini(void);
75 
76 ldi_ident_t zfs_li = NULL;
77 dev_info_t *zfs_dip;
78 
79 typedef int zfs_ioc_func_t(zfs_cmd_t *);
80 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, cred_t *);
81 
82 typedef struct zfs_ioc_vec {
83 	zfs_ioc_func_t		*zvec_func;
84 	zfs_secpolicy_func_t	*zvec_secpolicy;
85 	enum {
86 		NO_NAME,
87 		POOL_NAME,
88 		DATASET_NAME
89 	} zvec_namecheck;
90 	boolean_t		zvec_his_log;
91 } zfs_ioc_vec_t;
92 
93 static void clear_props(char *dataset, nvlist_t *props, nvlist_t *newprops);
94 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
95     boolean_t *);
96 int zfs_set_prop_nvlist(const char *, nvlist_t *);
97 
98 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
99 void
100 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
101 {
102 	const char *newfile;
103 	char buf[256];
104 	va_list adx;
105 
106 	/*
107 	 * Get rid of annoying "../common/" prefix to filename.
108 	 */
109 	newfile = strrchr(file, '/');
110 	if (newfile != NULL) {
111 		newfile = newfile + 1; /* Get rid of leading / */
112 	} else {
113 		newfile = file;
114 	}
115 
116 	va_start(adx, fmt);
117 	(void) vsnprintf(buf, sizeof (buf), fmt, adx);
118 	va_end(adx);
119 
120 	/*
121 	 * To get this data, use the zfs-dprintf probe as so:
122 	 * dtrace -q -n 'zfs-dprintf \
123 	 *	/stringof(arg0) == "dbuf.c"/ \
124 	 *	{printf("%s: %s", stringof(arg1), stringof(arg3))}'
125 	 * arg0 = file name
126 	 * arg1 = function name
127 	 * arg2 = line number
128 	 * arg3 = message
129 	 */
130 	DTRACE_PROBE4(zfs__dprintf,
131 	    char *, newfile, char *, func, int, line, char *, buf);
132 }
133 
134 static void
135 history_str_free(char *buf)
136 {
137 	kmem_free(buf, HIS_MAX_RECORD_LEN);
138 }
139 
140 static char *
141 history_str_get(zfs_cmd_t *zc)
142 {
143 	char *buf;
144 
145 	if (zc->zc_history == NULL)
146 		return (NULL);
147 
148 	buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
149 	if (copyinstr((void *)(uintptr_t)zc->zc_history,
150 	    buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
151 		history_str_free(buf);
152 		return (NULL);
153 	}
154 
155 	buf[HIS_MAX_RECORD_LEN -1] = '\0';
156 
157 	return (buf);
158 }
159 
160 /*
161  * Check to see if the named dataset is currently defined as bootable
162  */
163 static boolean_t
164 zfs_is_bootfs(const char *name)
165 {
166 	spa_t *spa;
167 	boolean_t ret = B_FALSE;
168 
169 	if (spa_open(name, &spa, FTAG) == 0) {
170 		if (spa->spa_bootfs) {
171 			objset_t *os;
172 
173 			if (dmu_objset_open(name, DMU_OST_ZFS,
174 			    DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
175 				ret = (dmu_objset_id(os) == spa->spa_bootfs);
176 				dmu_objset_close(os);
177 			}
178 		}
179 		spa_close(spa, FTAG);
180 	}
181 	return (ret);
182 }
183 
184 /*
185  * zfs_earlier_version
186  *
187  *	Return non-zero if the spa version is less than requested version.
188  */
189 static int
190 zfs_earlier_version(const char *name, int version)
191 {
192 	spa_t *spa;
193 
194 	if (spa_open(name, &spa, FTAG) == 0) {
195 		if (spa_version(spa) < version) {
196 			spa_close(spa, FTAG);
197 			return (1);
198 		}
199 		spa_close(spa, FTAG);
200 	}
201 	return (0);
202 }
203 
204 /*
205  * zpl_earlier_version
206  *
207  * Return TRUE if the ZPL version is less than requested version.
208  */
209 static boolean_t
210 zpl_earlier_version(const char *name, int version)
211 {
212 	objset_t *os;
213 	boolean_t rc = B_TRUE;
214 
215 	if (dmu_objset_open(name, DMU_OST_ANY,
216 	    DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
217 		uint64_t zplversion;
218 
219 		if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
220 			rc = zplversion < version;
221 		dmu_objset_close(os);
222 	}
223 	return (rc);
224 }
225 
226 static void
227 zfs_log_history(zfs_cmd_t *zc)
228 {
229 	spa_t *spa;
230 	char *buf;
231 
232 	if ((buf = history_str_get(zc)) == NULL)
233 		return;
234 
235 	if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
236 		if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
237 			(void) spa_history_log(spa, buf, LOG_CMD_NORMAL);
238 		spa_close(spa, FTAG);
239 	}
240 	history_str_free(buf);
241 }
242 
243 /*
244  * Policy for top-level read operations (list pools).  Requires no privileges,
245  * and can be used in the local zone, as there is no associated dataset.
246  */
247 /* ARGSUSED */
248 static int
249 zfs_secpolicy_none(zfs_cmd_t *zc, cred_t *cr)
250 {
251 	return (0);
252 }
253 
254 /*
255  * Policy for dataset read operations (list children, get statistics).  Requires
256  * no privileges, but must be visible in the local zone.
257  */
258 /* ARGSUSED */
259 static int
260 zfs_secpolicy_read(zfs_cmd_t *zc, cred_t *cr)
261 {
262 	if (INGLOBALZONE(curproc) ||
263 	    zone_dataset_visible(zc->zc_name, NULL))
264 		return (0);
265 
266 	return (ENOENT);
267 }
268 
269 static int
270 zfs_dozonecheck(const char *dataset, cred_t *cr)
271 {
272 	uint64_t zoned;
273 	int writable = 1;
274 
275 	/*
276 	 * The dataset must be visible by this zone -- check this first
277 	 * so they don't see EPERM on something they shouldn't know about.
278 	 */
279 	if (!INGLOBALZONE(curproc) &&
280 	    !zone_dataset_visible(dataset, &writable))
281 		return (ENOENT);
282 
283 	if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
284 		return (ENOENT);
285 
286 	if (INGLOBALZONE(curproc)) {
287 		/*
288 		 * If the fs is zoned, only root can access it from the
289 		 * global zone.
290 		 */
291 		if (secpolicy_zfs(cr) && zoned)
292 			return (EPERM);
293 	} else {
294 		/*
295 		 * If we are in a local zone, the 'zoned' property must be set.
296 		 */
297 		if (!zoned)
298 			return (EPERM);
299 
300 		/* must be writable by this zone */
301 		if (!writable)
302 			return (EPERM);
303 	}
304 	return (0);
305 }
306 
307 int
308 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
309 {
310 	int error;
311 
312 	error = zfs_dozonecheck(name, cr);
313 	if (error == 0) {
314 		error = secpolicy_zfs(cr);
315 		if (error)
316 			error = dsl_deleg_access(name, perm, cr);
317 	}
318 	return (error);
319 }
320 
321 static int
322 zfs_secpolicy_setprop(const char *name, zfs_prop_t prop, cred_t *cr)
323 {
324 	/*
325 	 * Check permissions for special properties.
326 	 */
327 	switch (prop) {
328 	case ZFS_PROP_ZONED:
329 		/*
330 		 * Disallow setting of 'zoned' from within a local zone.
331 		 */
332 		if (!INGLOBALZONE(curproc))
333 			return (EPERM);
334 		break;
335 
336 	case ZFS_PROP_QUOTA:
337 		if (!INGLOBALZONE(curproc)) {
338 			uint64_t zoned;
339 			char setpoint[MAXNAMELEN];
340 			/*
341 			 * Unprivileged users are allowed to modify the
342 			 * quota on things *under* (ie. contained by)
343 			 * the thing they own.
344 			 */
345 			if (dsl_prop_get_integer(name, "zoned", &zoned,
346 			    setpoint))
347 				return (EPERM);
348 			if (!zoned || strlen(name) <= strlen(setpoint))
349 				return (EPERM);
350 		}
351 		break;
352 	}
353 
354 	return (zfs_secpolicy_write_perms(name, zfs_prop_to_name(prop), cr));
355 }
356 
357 int
358 zfs_secpolicy_fsacl(zfs_cmd_t *zc, cred_t *cr)
359 {
360 	int error;
361 
362 	error = zfs_dozonecheck(zc->zc_name, cr);
363 	if (error)
364 		return (error);
365 
366 	/*
367 	 * permission to set permissions will be evaluated later in
368 	 * dsl_deleg_can_allow()
369 	 */
370 	return (0);
371 }
372 
373 int
374 zfs_secpolicy_rollback(zfs_cmd_t *zc, cred_t *cr)
375 {
376 	int error;
377 	error = zfs_secpolicy_write_perms(zc->zc_name,
378 	    ZFS_DELEG_PERM_ROLLBACK, cr);
379 	if (error == 0)
380 		error = zfs_secpolicy_write_perms(zc->zc_name,
381 		    ZFS_DELEG_PERM_MOUNT, cr);
382 	return (error);
383 }
384 
385 int
386 zfs_secpolicy_send(zfs_cmd_t *zc, cred_t *cr)
387 {
388 	return (zfs_secpolicy_write_perms(zc->zc_name,
389 	    ZFS_DELEG_PERM_SEND, cr));
390 }
391 
392 static int
393 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, cred_t *cr)
394 {
395 	vnode_t *vp;
396 	int error;
397 
398 	if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
399 	    NO_FOLLOW, NULL, &vp)) != 0)
400 		return (error);
401 
402 	/* Now make sure mntpnt and dataset are ZFS */
403 
404 	if (vp->v_vfsp->vfs_fstype != zfsfstype ||
405 	    (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
406 	    zc->zc_name) != 0)) {
407 		VN_RELE(vp);
408 		return (EPERM);
409 	}
410 
411 	VN_RELE(vp);
412 	return (dsl_deleg_access(zc->zc_name,
413 	    ZFS_DELEG_PERM_SHARE, cr));
414 }
415 
416 int
417 zfs_secpolicy_share(zfs_cmd_t *zc, cred_t *cr)
418 {
419 	if (!INGLOBALZONE(curproc))
420 		return (EPERM);
421 
422 	if (secpolicy_nfs(cr) == 0) {
423 		return (0);
424 	} else {
425 		return (zfs_secpolicy_deleg_share(zc, cr));
426 	}
427 }
428 
429 int
430 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, cred_t *cr)
431 {
432 	if (!INGLOBALZONE(curproc))
433 		return (EPERM);
434 
435 	if (secpolicy_smb(cr) == 0) {
436 		return (0);
437 	} else {
438 		return (zfs_secpolicy_deleg_share(zc, cr));
439 	}
440 }
441 
442 static int
443 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
444 {
445 	char *cp;
446 
447 	/*
448 	 * Remove the @bla or /bla from the end of the name to get the parent.
449 	 */
450 	(void) strncpy(parent, datasetname, parentsize);
451 	cp = strrchr(parent, '@');
452 	if (cp != NULL) {
453 		cp[0] = '\0';
454 	} else {
455 		cp = strrchr(parent, '/');
456 		if (cp == NULL)
457 			return (ENOENT);
458 		cp[0] = '\0';
459 	}
460 
461 	return (0);
462 }
463 
464 int
465 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
466 {
467 	int error;
468 
469 	if ((error = zfs_secpolicy_write_perms(name,
470 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
471 		return (error);
472 
473 	return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
474 }
475 
476 static int
477 zfs_secpolicy_destroy(zfs_cmd_t *zc, cred_t *cr)
478 {
479 	return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
480 }
481 
482 /*
483  * Must have sys_config privilege to check the iscsi permission
484  */
485 /* ARGSUSED */
486 static int
487 zfs_secpolicy_iscsi(zfs_cmd_t *zc, cred_t *cr)
488 {
489 	return (secpolicy_zfs(cr));
490 }
491 
492 int
493 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
494 {
495 	char 	parentname[MAXNAMELEN];
496 	int	error;
497 
498 	if ((error = zfs_secpolicy_write_perms(from,
499 	    ZFS_DELEG_PERM_RENAME, cr)) != 0)
500 		return (error);
501 
502 	if ((error = zfs_secpolicy_write_perms(from,
503 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
504 		return (error);
505 
506 	if ((error = zfs_get_parent(to, parentname,
507 	    sizeof (parentname))) != 0)
508 		return (error);
509 
510 	if ((error = zfs_secpolicy_write_perms(parentname,
511 	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
512 		return (error);
513 
514 	if ((error = zfs_secpolicy_write_perms(parentname,
515 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
516 		return (error);
517 
518 	return (error);
519 }
520 
521 static int
522 zfs_secpolicy_rename(zfs_cmd_t *zc, cred_t *cr)
523 {
524 	return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
525 }
526 
527 static int
528 zfs_secpolicy_promote(zfs_cmd_t *zc, cred_t *cr)
529 {
530 	char 	parentname[MAXNAMELEN];
531 	objset_t *clone;
532 	int error;
533 
534 	error = zfs_secpolicy_write_perms(zc->zc_name,
535 	    ZFS_DELEG_PERM_PROMOTE, cr);
536 	if (error)
537 		return (error);
538 
539 	error = dmu_objset_open(zc->zc_name, DMU_OST_ANY,
540 	    DS_MODE_USER | DS_MODE_READONLY, &clone);
541 
542 	if (error == 0) {
543 		dsl_dataset_t *pclone = NULL;
544 		dsl_dir_t *dd;
545 		dd = clone->os->os_dsl_dataset->ds_dir;
546 
547 		rw_enter(&dd->dd_pool->dp_config_rwlock, RW_READER);
548 		error = dsl_dataset_hold_obj(dd->dd_pool,
549 		    dd->dd_phys->dd_origin_obj, FTAG, &pclone);
550 		rw_exit(&dd->dd_pool->dp_config_rwlock);
551 		if (error) {
552 			dmu_objset_close(clone);
553 			return (error);
554 		}
555 
556 		error = zfs_secpolicy_write_perms(zc->zc_name,
557 		    ZFS_DELEG_PERM_MOUNT, cr);
558 
559 		dsl_dataset_name(pclone, parentname);
560 		dmu_objset_close(clone);
561 		dsl_dataset_rele(pclone, FTAG);
562 		if (error == 0)
563 			error = zfs_secpolicy_write_perms(parentname,
564 			    ZFS_DELEG_PERM_PROMOTE, cr);
565 	}
566 	return (error);
567 }
568 
569 static int
570 zfs_secpolicy_receive(zfs_cmd_t *zc, cred_t *cr)
571 {
572 	int error;
573 
574 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
575 	    ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
576 		return (error);
577 
578 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
579 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
580 		return (error);
581 
582 	return (zfs_secpolicy_write_perms(zc->zc_name,
583 	    ZFS_DELEG_PERM_CREATE, cr));
584 }
585 
586 int
587 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
588 {
589 	int error;
590 
591 	if ((error = zfs_secpolicy_write_perms(name,
592 	    ZFS_DELEG_PERM_SNAPSHOT, cr)) != 0)
593 		return (error);
594 
595 	error = zfs_secpolicy_write_perms(name,
596 	    ZFS_DELEG_PERM_MOUNT, cr);
597 
598 	return (error);
599 }
600 
601 static int
602 zfs_secpolicy_snapshot(zfs_cmd_t *zc, cred_t *cr)
603 {
604 
605 	return (zfs_secpolicy_snapshot_perms(zc->zc_name, cr));
606 }
607 
608 static int
609 zfs_secpolicy_create(zfs_cmd_t *zc, cred_t *cr)
610 {
611 	char 	parentname[MAXNAMELEN];
612 	int 	error;
613 
614 	if ((error = zfs_get_parent(zc->zc_name, parentname,
615 	    sizeof (parentname))) != 0)
616 		return (error);
617 
618 	if (zc->zc_value[0] != '\0') {
619 		if ((error = zfs_secpolicy_write_perms(zc->zc_value,
620 		    ZFS_DELEG_PERM_CLONE, cr)) != 0)
621 			return (error);
622 	}
623 
624 	if ((error = zfs_secpolicy_write_perms(parentname,
625 	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
626 		return (error);
627 
628 	error = zfs_secpolicy_write_perms(parentname,
629 	    ZFS_DELEG_PERM_MOUNT, cr);
630 
631 	return (error);
632 }
633 
634 static int
635 zfs_secpolicy_umount(zfs_cmd_t *zc, cred_t *cr)
636 {
637 	int error;
638 
639 	error = secpolicy_fs_unmount(cr, NULL);
640 	if (error) {
641 		error = dsl_deleg_access(zc->zc_name, ZFS_DELEG_PERM_MOUNT, cr);
642 	}
643 	return (error);
644 }
645 
646 /*
647  * Policy for pool operations - create/destroy pools, add vdevs, etc.  Requires
648  * SYS_CONFIG privilege, which is not available in a local zone.
649  */
650 /* ARGSUSED */
651 static int
652 zfs_secpolicy_config(zfs_cmd_t *zc, cred_t *cr)
653 {
654 	if (secpolicy_sys_config(cr, B_FALSE) != 0)
655 		return (EPERM);
656 
657 	return (0);
658 }
659 
660 /*
661  * Just like zfs_secpolicy_config, except that we will check for
662  * mount permission on the dataset for permission to create/remove
663  * the minor nodes.
664  */
665 static int
666 zfs_secpolicy_minor(zfs_cmd_t *zc, cred_t *cr)
667 {
668 	if (secpolicy_sys_config(cr, B_FALSE) != 0) {
669 		return (dsl_deleg_access(zc->zc_name,
670 		    ZFS_DELEG_PERM_MOUNT, cr));
671 	}
672 
673 	return (0);
674 }
675 
676 /*
677  * Policy for fault injection.  Requires all privileges.
678  */
679 /* ARGSUSED */
680 static int
681 zfs_secpolicy_inject(zfs_cmd_t *zc, cred_t *cr)
682 {
683 	return (secpolicy_zinject(cr));
684 }
685 
686 static int
687 zfs_secpolicy_inherit(zfs_cmd_t *zc, cred_t *cr)
688 {
689 	zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
690 
691 	if (prop == ZPROP_INVAL) {
692 		if (!zfs_prop_user(zc->zc_value))
693 			return (EINVAL);
694 		return (zfs_secpolicy_write_perms(zc->zc_name,
695 		    ZFS_DELEG_PERM_USERPROP, cr));
696 	} else {
697 		if (!zfs_prop_inheritable(prop))
698 			return (EINVAL);
699 		return (zfs_secpolicy_setprop(zc->zc_name, prop, cr));
700 	}
701 }
702 
703 /*
704  * Returns the nvlist as specified by the user in the zfs_cmd_t.
705  */
706 static int
707 get_nvlist(uint64_t nvl, uint64_t size, nvlist_t **nvp)
708 {
709 	char *packed;
710 	int error;
711 	nvlist_t *list = NULL;
712 
713 	/*
714 	 * Read in and unpack the user-supplied nvlist.
715 	 */
716 	if (size == 0)
717 		return (EINVAL);
718 
719 	packed = kmem_alloc(size, KM_SLEEP);
720 
721 	if ((error = xcopyin((void *)(uintptr_t)nvl, packed, size)) != 0) {
722 		kmem_free(packed, size);
723 		return (error);
724 	}
725 
726 	if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
727 		kmem_free(packed, size);
728 		return (error);
729 	}
730 
731 	kmem_free(packed, size);
732 
733 	*nvp = list;
734 	return (0);
735 }
736 
737 static int
738 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
739 {
740 	char *packed = NULL;
741 	size_t size;
742 	int error;
743 
744 	VERIFY(nvlist_size(nvl, &size, NV_ENCODE_NATIVE) == 0);
745 
746 	if (size > zc->zc_nvlist_dst_size) {
747 		error = ENOMEM;
748 	} else {
749 		packed = kmem_alloc(size, KM_SLEEP);
750 		VERIFY(nvlist_pack(nvl, &packed, &size, NV_ENCODE_NATIVE,
751 		    KM_SLEEP) == 0);
752 		error = xcopyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
753 		    size);
754 		kmem_free(packed, size);
755 	}
756 
757 	zc->zc_nvlist_dst_size = size;
758 	return (error);
759 }
760 
761 static int
762 zfs_ioc_pool_create(zfs_cmd_t *zc)
763 {
764 	int error;
765 	nvlist_t *config, *props = NULL;
766 	nvlist_t *rootprops = NULL;
767 	nvlist_t *zplprops = NULL;
768 	char *buf;
769 
770 	if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
771 	    &config))
772 		return (error);
773 
774 	if (zc->zc_nvlist_src_size != 0 && (error =
775 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, &props))) {
776 		nvlist_free(config);
777 		return (error);
778 	}
779 
780 	if (props) {
781 		nvlist_t *nvl = NULL;
782 		uint64_t version = SPA_VERSION;
783 
784 		(void) nvlist_lookup_uint64(props,
785 		    zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
786 		if (version < SPA_VERSION_INITIAL || version > SPA_VERSION) {
787 			error = EINVAL;
788 			goto pool_props_bad;
789 		}
790 		(void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
791 		if (nvl) {
792 			error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
793 			if (error != 0) {
794 				nvlist_free(config);
795 				nvlist_free(props);
796 				return (error);
797 			}
798 			(void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
799 		}
800 		VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
801 		error = zfs_fill_zplprops_root(version, rootprops,
802 		    zplprops, NULL);
803 		if (error)
804 			goto pool_props_bad;
805 	}
806 
807 	buf = history_str_get(zc);
808 
809 	error = spa_create(zc->zc_name, config, props, buf, zplprops);
810 
811 	/*
812 	 * Set the remaining root properties
813 	 */
814 	if (!error &&
815 	    (error = zfs_set_prop_nvlist(zc->zc_name, rootprops)) != 0)
816 		(void) spa_destroy(zc->zc_name);
817 
818 	if (buf != NULL)
819 		history_str_free(buf);
820 
821 pool_props_bad:
822 	nvlist_free(rootprops);
823 	nvlist_free(zplprops);
824 	nvlist_free(config);
825 	nvlist_free(props);
826 
827 	return (error);
828 }
829 
830 static int
831 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
832 {
833 	int error;
834 	zfs_log_history(zc);
835 	error = spa_destroy(zc->zc_name);
836 	return (error);
837 }
838 
839 static int
840 zfs_ioc_pool_import(zfs_cmd_t *zc)
841 {
842 	int error;
843 	nvlist_t *config, *props = NULL;
844 	uint64_t guid;
845 
846 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
847 	    &config)) != 0)
848 		return (error);
849 
850 	if (zc->zc_nvlist_src_size != 0 && (error =
851 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size, &props))) {
852 		nvlist_free(config);
853 		return (error);
854 	}
855 
856 	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
857 	    guid != zc->zc_guid)
858 		error = EINVAL;
859 	else if (zc->zc_cookie)
860 		error = spa_import_faulted(zc->zc_name, config,
861 		    props);
862 	else
863 		error = spa_import(zc->zc_name, config, props);
864 
865 	nvlist_free(config);
866 
867 	if (props)
868 		nvlist_free(props);
869 
870 	return (error);
871 }
872 
873 static int
874 zfs_ioc_pool_export(zfs_cmd_t *zc)
875 {
876 	int error;
877 	boolean_t force = (boolean_t)zc->zc_cookie;
878 	boolean_t hardforce = (boolean_t)zc->zc_guid;
879 
880 	zfs_log_history(zc);
881 	error = spa_export(zc->zc_name, NULL, force, hardforce);
882 	return (error);
883 }
884 
885 static int
886 zfs_ioc_pool_configs(zfs_cmd_t *zc)
887 {
888 	nvlist_t *configs;
889 	int error;
890 
891 	if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
892 		return (EEXIST);
893 
894 	error = put_nvlist(zc, configs);
895 
896 	nvlist_free(configs);
897 
898 	return (error);
899 }
900 
901 static int
902 zfs_ioc_pool_stats(zfs_cmd_t *zc)
903 {
904 	nvlist_t *config;
905 	int error;
906 	int ret = 0;
907 
908 	error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
909 	    sizeof (zc->zc_value));
910 
911 	if (config != NULL) {
912 		ret = put_nvlist(zc, config);
913 		nvlist_free(config);
914 
915 		/*
916 		 * The config may be present even if 'error' is non-zero.
917 		 * In this case we return success, and preserve the real errno
918 		 * in 'zc_cookie'.
919 		 */
920 		zc->zc_cookie = error;
921 	} else {
922 		ret = error;
923 	}
924 
925 	return (ret);
926 }
927 
928 /*
929  * Try to import the given pool, returning pool stats as appropriate so that
930  * user land knows which devices are available and overall pool health.
931  */
932 static int
933 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
934 {
935 	nvlist_t *tryconfig, *config;
936 	int error;
937 
938 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
939 	    &tryconfig)) != 0)
940 		return (error);
941 
942 	config = spa_tryimport(tryconfig);
943 
944 	nvlist_free(tryconfig);
945 
946 	if (config == NULL)
947 		return (EINVAL);
948 
949 	error = put_nvlist(zc, config);
950 	nvlist_free(config);
951 
952 	return (error);
953 }
954 
955 static int
956 zfs_ioc_pool_scrub(zfs_cmd_t *zc)
957 {
958 	spa_t *spa;
959 	int error;
960 
961 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
962 		return (error);
963 
964 	error = spa_scrub(spa, zc->zc_cookie);
965 
966 	spa_close(spa, FTAG);
967 
968 	return (error);
969 }
970 
971 static int
972 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
973 {
974 	spa_t *spa;
975 	int error;
976 
977 	error = spa_open(zc->zc_name, &spa, FTAG);
978 	if (error == 0) {
979 		spa_freeze(spa);
980 		spa_close(spa, FTAG);
981 	}
982 	return (error);
983 }
984 
985 static int
986 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
987 {
988 	spa_t *spa;
989 	int error;
990 
991 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
992 		return (error);
993 
994 	if (zc->zc_cookie < spa_version(spa) || zc->zc_cookie > SPA_VERSION) {
995 		spa_close(spa, FTAG);
996 		return (EINVAL);
997 	}
998 
999 	spa_upgrade(spa, zc->zc_cookie);
1000 	spa_close(spa, FTAG);
1001 
1002 	return (error);
1003 }
1004 
1005 static int
1006 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1007 {
1008 	spa_t *spa;
1009 	char *hist_buf;
1010 	uint64_t size;
1011 	int error;
1012 
1013 	if ((size = zc->zc_history_len) == 0)
1014 		return (EINVAL);
1015 
1016 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1017 		return (error);
1018 
1019 	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1020 		spa_close(spa, FTAG);
1021 		return (ENOTSUP);
1022 	}
1023 
1024 	hist_buf = kmem_alloc(size, KM_SLEEP);
1025 	if ((error = spa_history_get(spa, &zc->zc_history_offset,
1026 	    &zc->zc_history_len, hist_buf)) == 0) {
1027 		error = xcopyout(hist_buf,
1028 		    (char *)(uintptr_t)zc->zc_history,
1029 		    zc->zc_history_len);
1030 	}
1031 
1032 	spa_close(spa, FTAG);
1033 	kmem_free(hist_buf, size);
1034 	return (error);
1035 }
1036 
1037 static int
1038 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1039 {
1040 	int error;
1041 
1042 	if (error = dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value))
1043 		return (error);
1044 
1045 	return (0);
1046 }
1047 
1048 static int
1049 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1050 {
1051 	objset_t *osp;
1052 	int error;
1053 
1054 	if ((error = dmu_objset_open(zc->zc_name, DMU_OST_ZFS,
1055 	    DS_MODE_USER | DS_MODE_READONLY, &osp)) != 0)
1056 		return (error);
1057 	error = zfs_obj_to_path(osp, zc->zc_obj, zc->zc_value,
1058 	    sizeof (zc->zc_value));
1059 	dmu_objset_close(osp);
1060 
1061 	return (error);
1062 }
1063 
1064 static int
1065 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1066 {
1067 	spa_t *spa;
1068 	int error;
1069 	nvlist_t *config, **l2cache, **spares;
1070 	uint_t nl2cache = 0, nspares = 0;
1071 
1072 	error = spa_open(zc->zc_name, &spa, FTAG);
1073 	if (error != 0)
1074 		return (error);
1075 
1076 	error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1077 	    &config);
1078 	(void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1079 	    &l2cache, &nl2cache);
1080 
1081 	(void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1082 	    &spares, &nspares);
1083 
1084 	/*
1085 	 * A root pool with concatenated devices is not supported.
1086 	 * Thus, can not add a device to a root pool.
1087 	 *
1088 	 * Intent log device can not be added to a rootpool because
1089 	 * during mountroot, zil is replayed, a seperated log device
1090 	 * can not be accessed during the mountroot time.
1091 	 *
1092 	 * l2cache and spare devices are ok to be added to a rootpool.
1093 	 */
1094 	if (spa->spa_bootfs != 0 && nl2cache == 0 && nspares == 0) {
1095 		spa_close(spa, FTAG);
1096 		return (EDOM);
1097 	}
1098 
1099 	if (error == 0) {
1100 		error = spa_vdev_add(spa, config);
1101 		nvlist_free(config);
1102 	}
1103 	spa_close(spa, FTAG);
1104 	return (error);
1105 }
1106 
1107 static int
1108 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1109 {
1110 	spa_t *spa;
1111 	int error;
1112 
1113 	error = spa_open(zc->zc_name, &spa, FTAG);
1114 	if (error != 0)
1115 		return (error);
1116 	error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1117 	spa_close(spa, FTAG);
1118 	return (error);
1119 }
1120 
1121 static int
1122 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1123 {
1124 	spa_t *spa;
1125 	int error;
1126 	vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1127 
1128 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1129 		return (error);
1130 	switch (zc->zc_cookie) {
1131 	case VDEV_STATE_ONLINE:
1132 		error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
1133 		break;
1134 
1135 	case VDEV_STATE_OFFLINE:
1136 		error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
1137 		break;
1138 
1139 	case VDEV_STATE_FAULTED:
1140 		error = vdev_fault(spa, zc->zc_guid);
1141 		break;
1142 
1143 	case VDEV_STATE_DEGRADED:
1144 		error = vdev_degrade(spa, zc->zc_guid);
1145 		break;
1146 
1147 	default:
1148 		error = EINVAL;
1149 	}
1150 	zc->zc_cookie = newstate;
1151 	spa_close(spa, FTAG);
1152 	return (error);
1153 }
1154 
1155 static int
1156 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
1157 {
1158 	spa_t *spa;
1159 	int replacing = zc->zc_cookie;
1160 	nvlist_t *config;
1161 	int error;
1162 
1163 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1164 		return (error);
1165 
1166 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1167 	    &config)) == 0) {
1168 		error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
1169 		nvlist_free(config);
1170 	}
1171 
1172 	spa_close(spa, FTAG);
1173 	return (error);
1174 }
1175 
1176 static int
1177 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
1178 {
1179 	spa_t *spa;
1180 	int error;
1181 
1182 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1183 		return (error);
1184 
1185 	error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
1186 
1187 	spa_close(spa, FTAG);
1188 	return (error);
1189 }
1190 
1191 static int
1192 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
1193 {
1194 	spa_t *spa;
1195 	char *path = zc->zc_value;
1196 	uint64_t guid = zc->zc_guid;
1197 	int error;
1198 
1199 	error = spa_open(zc->zc_name, &spa, FTAG);
1200 	if (error != 0)
1201 		return (error);
1202 
1203 	error = spa_vdev_setpath(spa, guid, path);
1204 	spa_close(spa, FTAG);
1205 	return (error);
1206 }
1207 
1208 /*
1209  * inputs:
1210  * zc_name		name of filesystem
1211  * zc_nvlist_dst_size	size of buffer for property nvlist
1212  *
1213  * outputs:
1214  * zc_objset_stats	stats
1215  * zc_nvlist_dst	property nvlist
1216  * zc_nvlist_dst_size	size of property nvlist
1217  */
1218 static int
1219 zfs_ioc_objset_stats(zfs_cmd_t *zc)
1220 {
1221 	objset_t *os = NULL;
1222 	int error;
1223 	nvlist_t *nv;
1224 
1225 	if (error = dmu_objset_open(zc->zc_name,
1226 	    DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os))
1227 		return (error);
1228 
1229 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1230 
1231 	if (zc->zc_nvlist_dst != 0 &&
1232 	    (error = dsl_prop_get_all(os, &nv, FALSE)) == 0) {
1233 		dmu_objset_stats(os, nv);
1234 		/*
1235 		 * NB: zvol_get_stats() will read the objset contents,
1236 		 * which we aren't supposed to do with a
1237 		 * DS_MODE_USER hold, because it could be
1238 		 * inconsistent.  So this is a bit of a workaround...
1239 		 */
1240 		if (!zc->zc_objset_stats.dds_inconsistent) {
1241 			if (dmu_objset_type(os) == DMU_OST_ZVOL)
1242 				VERIFY(zvol_get_stats(os, nv) == 0);
1243 		}
1244 		error = put_nvlist(zc, nv);
1245 		nvlist_free(nv);
1246 	}
1247 
1248 	dmu_objset_close(os);
1249 	return (error);
1250 }
1251 
1252 static int
1253 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
1254 {
1255 	uint64_t value;
1256 	int error;
1257 
1258 	/*
1259 	 * zfs_get_zplprop() will either find a value or give us
1260 	 * the default value (if there is one).
1261 	 */
1262 	if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
1263 		return (error);
1264 	VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
1265 	return (0);
1266 }
1267 
1268 /*
1269  * inputs:
1270  * zc_name		name of filesystem
1271  * zc_nvlist_dst_size	size of buffer for zpl property nvlist
1272  *
1273  * outputs:
1274  * zc_nvlist_dst	zpl property nvlist
1275  * zc_nvlist_dst_size	size of zpl property nvlist
1276  */
1277 static int
1278 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
1279 {
1280 	objset_t *os;
1281 	int err;
1282 
1283 	if (err = dmu_objset_open(zc->zc_name,
1284 	    DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os))
1285 		return (err);
1286 
1287 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
1288 
1289 	/*
1290 	 * NB: nvl_add_zplprop() will read the objset contents,
1291 	 * which we aren't supposed to do with a DS_MODE_USER
1292 	 * hold, because it could be inconsistent.
1293 	 */
1294 	if (zc->zc_nvlist_dst != NULL &&
1295 	    !zc->zc_objset_stats.dds_inconsistent &&
1296 	    dmu_objset_type(os) == DMU_OST_ZFS) {
1297 		nvlist_t *nv;
1298 
1299 		VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1300 		if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
1301 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
1302 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
1303 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
1304 			err = put_nvlist(zc, nv);
1305 		nvlist_free(nv);
1306 	} else {
1307 		err = ENOENT;
1308 	}
1309 	dmu_objset_close(os);
1310 	return (err);
1311 }
1312 
1313 /*
1314  * inputs:
1315  * zc_name		name of filesystem
1316  * zc_cookie		zap cursor
1317  * zc_nvlist_dst_size	size of buffer for property nvlist
1318  *
1319  * outputs:
1320  * zc_name		name of next filesystem
1321  * zc_objset_stats	stats
1322  * zc_nvlist_dst	property nvlist
1323  * zc_nvlist_dst_size	size of property nvlist
1324  */
1325 static int
1326 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
1327 {
1328 	objset_t *os;
1329 	int error;
1330 	char *p;
1331 
1332 	if (error = dmu_objset_open(zc->zc_name,
1333 	    DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os)) {
1334 		if (error == ENOENT)
1335 			error = ESRCH;
1336 		return (error);
1337 	}
1338 
1339 	p = strrchr(zc->zc_name, '/');
1340 	if (p == NULL || p[1] != '\0')
1341 		(void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
1342 	p = zc->zc_name + strlen(zc->zc_name);
1343 
1344 	/*
1345 	 * Pre-fetch the datasets.  dmu_objset_prefetch() always returns 0
1346 	 * but is not declared void because its called by dmu_objset_find().
1347 	 */
1348 	if (zc->zc_cookie == 0) {
1349 		uint64_t cookie = 0;
1350 		int len = sizeof (zc->zc_name) - (p - zc->zc_name);
1351 
1352 		while (dmu_dir_list_next(os, len, p, NULL, &cookie) == 0)
1353 			(void) dmu_objset_prefetch(p, NULL);
1354 	}
1355 
1356 	do {
1357 		error = dmu_dir_list_next(os,
1358 		    sizeof (zc->zc_name) - (p - zc->zc_name), p,
1359 		    NULL, &zc->zc_cookie);
1360 		if (error == ENOENT)
1361 			error = ESRCH;
1362 	} while (error == 0 && !INGLOBALZONE(curproc) &&
1363 	    !zone_dataset_visible(zc->zc_name, NULL));
1364 	dmu_objset_close(os);
1365 
1366 	/*
1367 	 * If it's a hidden dataset (ie. with a '$' in its name), don't
1368 	 * try to get stats for it.  Userland will skip over it.
1369 	 */
1370 	if (error == 0 && strchr(zc->zc_name, '$') == NULL)
1371 		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
1372 
1373 	return (error);
1374 }
1375 
1376 /*
1377  * inputs:
1378  * zc_name		name of filesystem
1379  * zc_cookie		zap cursor
1380  * zc_nvlist_dst_size	size of buffer for property nvlist
1381  *
1382  * outputs:
1383  * zc_name		name of next snapshot
1384  * zc_objset_stats	stats
1385  * zc_nvlist_dst	property nvlist
1386  * zc_nvlist_dst_size	size of property nvlist
1387  */
1388 static int
1389 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
1390 {
1391 	objset_t *os;
1392 	int error;
1393 
1394 	error = dmu_objset_open(zc->zc_name,
1395 	    DMU_OST_ANY, DS_MODE_USER | DS_MODE_READONLY, &os);
1396 	if (error)
1397 		return (error == ENOENT ? ESRCH : error);
1398 
1399 	if (zc->zc_cookie == 0)
1400 		(void) dmu_objset_find(zc->zc_name, dmu_objset_prefetch,
1401 		    NULL, DS_FIND_SNAPSHOTS);
1402 	/*
1403 	 * A dataset name of maximum length cannot have any snapshots,
1404 	 * so exit immediately.
1405 	 */
1406 	if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >= MAXNAMELEN) {
1407 		dmu_objset_close(os);
1408 		return (ESRCH);
1409 	}
1410 
1411 	error = dmu_snapshot_list_next(os,
1412 	    sizeof (zc->zc_name) - strlen(zc->zc_name),
1413 	    zc->zc_name + strlen(zc->zc_name), NULL, &zc->zc_cookie, NULL);
1414 	dmu_objset_close(os);
1415 	if (error == 0)
1416 		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
1417 	else if (error == ENOENT)
1418 		error = ESRCH;
1419 
1420 	/* if we failed, undo the @ that we tacked on to zc_name */
1421 	if (error)
1422 		*strchr(zc->zc_name, '@') = '\0';
1423 	return (error);
1424 }
1425 
1426 int
1427 zfs_set_prop_nvlist(const char *name, nvlist_t *nvl)
1428 {
1429 	nvpair_t *elem;
1430 	int error = 0;
1431 	uint64_t intval;
1432 	char *strval;
1433 	nvlist_t *genericnvl;
1434 
1435 	/*
1436 	 * First validate permission to set all of the properties
1437 	 */
1438 	elem = NULL;
1439 	while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) {
1440 		const char *propname = nvpair_name(elem);
1441 		zfs_prop_t prop = zfs_name_to_prop(propname);
1442 
1443 		if (prop == ZPROP_INVAL) {
1444 			/*
1445 			 * If this is a user-defined property, it must be a
1446 			 * string, and there is no further validation to do.
1447 			 */
1448 			if (!zfs_prop_user(propname) ||
1449 			    nvpair_type(elem) != DATA_TYPE_STRING)
1450 				return (EINVAL);
1451 
1452 			if (error = zfs_secpolicy_write_perms(name,
1453 			    ZFS_DELEG_PERM_USERPROP, CRED()))
1454 				return (error);
1455 			continue;
1456 		}
1457 
1458 		if ((error = zfs_secpolicy_setprop(name, prop, CRED())) != 0)
1459 			return (error);
1460 
1461 		/*
1462 		 * Check that this value is valid for this pool version
1463 		 */
1464 		switch (prop) {
1465 		case ZFS_PROP_COMPRESSION:
1466 			/*
1467 			 * If the user specified gzip compression, make sure
1468 			 * the SPA supports it. We ignore any errors here since
1469 			 * we'll catch them later.
1470 			 */
1471 			if (nvpair_type(elem) == DATA_TYPE_UINT64 &&
1472 			    nvpair_value_uint64(elem, &intval) == 0) {
1473 				if (intval >= ZIO_COMPRESS_GZIP_1 &&
1474 				    intval <= ZIO_COMPRESS_GZIP_9 &&
1475 				    zfs_earlier_version(name,
1476 				    SPA_VERSION_GZIP_COMPRESSION))
1477 					return (ENOTSUP);
1478 
1479 				/*
1480 				 * If this is a bootable dataset then
1481 				 * verify that the compression algorithm
1482 				 * is supported for booting. We must return
1483 				 * something other than ENOTSUP since it
1484 				 * implies a downrev pool version.
1485 				 */
1486 				if (zfs_is_bootfs(name) &&
1487 				    !BOOTFS_COMPRESS_VALID(intval))
1488 					return (ERANGE);
1489 			}
1490 			break;
1491 
1492 		case ZFS_PROP_COPIES:
1493 			if (zfs_earlier_version(name,
1494 			    SPA_VERSION_DITTO_BLOCKS))
1495 				return (ENOTSUP);
1496 			break;
1497 
1498 		case ZFS_PROP_SHARESMB:
1499 			if (zpl_earlier_version(name, ZPL_VERSION_FUID))
1500 				return (ENOTSUP);
1501 			break;
1502 
1503 		case ZFS_PROP_ACLINHERIT:
1504 			if (nvpair_type(elem) == DATA_TYPE_UINT64 &&
1505 			    nvpair_value_uint64(elem, &intval) == 0)
1506 				if (intval == ZFS_ACL_PASSTHROUGH_X &&
1507 				    zfs_earlier_version(name,
1508 				    SPA_VERSION_PASSTHROUGH_X))
1509 					return (ENOTSUP);
1510 		}
1511 	}
1512 
1513 	VERIFY(nvlist_alloc(&genericnvl, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1514 	elem = NULL;
1515 	while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) {
1516 		const char *propname = nvpair_name(elem);
1517 		zfs_prop_t prop = zfs_name_to_prop(propname);
1518 
1519 		if (prop == ZPROP_INVAL) {
1520 			VERIFY(nvpair_value_string(elem, &strval) == 0);
1521 			error = dsl_prop_set(name, propname, 1,
1522 			    strlen(strval) + 1, strval);
1523 			if (error == 0)
1524 				continue;
1525 			else
1526 				goto out;
1527 		}
1528 
1529 		switch (prop) {
1530 		case ZFS_PROP_QUOTA:
1531 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1532 			    (error = dsl_dir_set_quota(name, intval)) != 0)
1533 				goto out;
1534 			break;
1535 
1536 		case ZFS_PROP_REFQUOTA:
1537 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1538 			    (error = dsl_dataset_set_quota(name, intval)) != 0)
1539 				goto out;
1540 			break;
1541 
1542 		case ZFS_PROP_RESERVATION:
1543 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1544 			    (error = dsl_dir_set_reservation(name,
1545 			    intval)) != 0)
1546 				goto out;
1547 			break;
1548 
1549 		case ZFS_PROP_REFRESERVATION:
1550 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1551 			    (error = dsl_dataset_set_reservation(name,
1552 			    intval)) != 0)
1553 				goto out;
1554 			break;
1555 
1556 		case ZFS_PROP_VOLSIZE:
1557 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1558 			    (error = zvol_set_volsize(name,
1559 			    ddi_driver_major(zfs_dip), intval)) != 0)
1560 				goto out;
1561 			break;
1562 
1563 		case ZFS_PROP_VOLBLOCKSIZE:
1564 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1565 			    (error = zvol_set_volblocksize(name, intval)) != 0)
1566 				goto out;
1567 			break;
1568 
1569 		case ZFS_PROP_VERSION:
1570 			if ((error = nvpair_value_uint64(elem, &intval)) != 0 ||
1571 			    (error = zfs_set_version(name, intval)) != 0)
1572 				goto out;
1573 			break;
1574 
1575 		default:
1576 			if (nvpair_type(elem) == DATA_TYPE_STRING) {
1577 				if (zfs_prop_get_type(prop) !=
1578 				    PROP_TYPE_STRING) {
1579 					error = EINVAL;
1580 					goto out;
1581 				}
1582 			} else if (nvpair_type(elem) == DATA_TYPE_UINT64) {
1583 				const char *unused;
1584 
1585 				VERIFY(nvpair_value_uint64(elem, &intval) == 0);
1586 
1587 				switch (zfs_prop_get_type(prop)) {
1588 				case PROP_TYPE_NUMBER:
1589 					break;
1590 				case PROP_TYPE_STRING:
1591 					error = EINVAL;
1592 					goto out;
1593 				case PROP_TYPE_INDEX:
1594 					if (zfs_prop_index_to_string(prop,
1595 					    intval, &unused) != 0) {
1596 						error = EINVAL;
1597 						goto out;
1598 					}
1599 					break;
1600 				default:
1601 					cmn_err(CE_PANIC,
1602 					    "unknown property type");
1603 					break;
1604 				}
1605 			} else {
1606 				error = EINVAL;
1607 				goto out;
1608 			}
1609 			if ((error = nvlist_add_nvpair(genericnvl, elem)) != 0)
1610 				goto out;
1611 		}
1612 	}
1613 
1614 	if (nvlist_next_nvpair(genericnvl, NULL) != NULL) {
1615 		error = dsl_props_set(name, genericnvl);
1616 	}
1617 out:
1618 	nvlist_free(genericnvl);
1619 	return (error);
1620 }
1621 
1622 /*
1623  * inputs:
1624  * zc_name		name of filesystem
1625  * zc_value		name of property to set
1626  * zc_nvlist_src{_size}	nvlist of properties to apply
1627  * zc_cookie		clear existing local props?
1628  *
1629  * outputs:		none
1630  */
1631 static int
1632 zfs_ioc_set_prop(zfs_cmd_t *zc)
1633 {
1634 	nvlist_t *nvl;
1635 	int error;
1636 
1637 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1638 	    &nvl)) != 0)
1639 		return (error);
1640 
1641 	if (zc->zc_cookie) {
1642 		nvlist_t *origprops;
1643 		objset_t *os;
1644 
1645 		if (dmu_objset_open(zc->zc_name, DMU_OST_ANY,
1646 		    DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
1647 			if (dsl_prop_get_all(os, &origprops, TRUE) == 0) {
1648 				clear_props(zc->zc_name, origprops, nvl);
1649 				nvlist_free(origprops);
1650 			}
1651 			dmu_objset_close(os);
1652 		}
1653 
1654 	}
1655 
1656 	error = zfs_set_prop_nvlist(zc->zc_name, nvl);
1657 
1658 	nvlist_free(nvl);
1659 	return (error);
1660 }
1661 
1662 /*
1663  * inputs:
1664  * zc_name		name of filesystem
1665  * zc_value		name of property to inherit
1666  *
1667  * outputs:		none
1668  */
1669 static int
1670 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
1671 {
1672 	/* the property name has been validated by zfs_secpolicy_inherit() */
1673 	return (dsl_prop_set(zc->zc_name, zc->zc_value, 0, 0, NULL));
1674 }
1675 
1676 static int
1677 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
1678 {
1679 	nvlist_t *props;
1680 	spa_t *spa;
1681 	int error;
1682 	nvpair_t *elem;
1683 
1684 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1685 	    &props)))
1686 		return (error);
1687 
1688 	/*
1689 	 * If the only property is the configfile, then just do a spa_lookup()
1690 	 * to handle the faulted case.
1691 	 */
1692 	elem = nvlist_next_nvpair(props, NULL);
1693 	if (elem != NULL && strcmp(nvpair_name(elem),
1694 	    zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
1695 	    nvlist_next_nvpair(props, elem) == NULL) {
1696 		mutex_enter(&spa_namespace_lock);
1697 		if ((spa = spa_lookup(zc->zc_name)) != NULL) {
1698 			spa_configfile_set(spa, props, B_FALSE);
1699 			spa_config_sync(spa, B_FALSE, B_TRUE);
1700 		}
1701 		mutex_exit(&spa_namespace_lock);
1702 		if (spa != NULL)
1703 			return (0);
1704 	}
1705 
1706 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
1707 		nvlist_free(props);
1708 		return (error);
1709 	}
1710 
1711 	error = spa_prop_set(spa, props);
1712 
1713 	nvlist_free(props);
1714 	spa_close(spa, FTAG);
1715 
1716 	return (error);
1717 }
1718 
1719 static int
1720 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
1721 {
1722 	spa_t *spa;
1723 	int error;
1724 	nvlist_t *nvp = NULL;
1725 
1726 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
1727 		/*
1728 		 * If the pool is faulted, there may be properties we can still
1729 		 * get (such as altroot and cachefile), so attempt to get them
1730 		 * anyway.
1731 		 */
1732 		mutex_enter(&spa_namespace_lock);
1733 		if ((spa = spa_lookup(zc->zc_name)) != NULL)
1734 			error = spa_prop_get(spa, &nvp);
1735 		mutex_exit(&spa_namespace_lock);
1736 	} else {
1737 		error = spa_prop_get(spa, &nvp);
1738 		spa_close(spa, FTAG);
1739 	}
1740 
1741 	if (error == 0 && zc->zc_nvlist_dst != NULL)
1742 		error = put_nvlist(zc, nvp);
1743 	else
1744 		error = EFAULT;
1745 
1746 	nvlist_free(nvp);
1747 	return (error);
1748 }
1749 
1750 static int
1751 zfs_ioc_iscsi_perm_check(zfs_cmd_t *zc)
1752 {
1753 	nvlist_t *nvp;
1754 	int error;
1755 	uint32_t uid;
1756 	uint32_t gid;
1757 	uint32_t *groups;
1758 	uint_t group_cnt;
1759 	cred_t	*usercred;
1760 
1761 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1762 	    &nvp)) != 0) {
1763 		return (error);
1764 	}
1765 
1766 	if ((error = nvlist_lookup_uint32(nvp,
1767 	    ZFS_DELEG_PERM_UID, &uid)) != 0) {
1768 		nvlist_free(nvp);
1769 		return (EPERM);
1770 	}
1771 
1772 	if ((error = nvlist_lookup_uint32(nvp,
1773 	    ZFS_DELEG_PERM_GID, &gid)) != 0) {
1774 		nvlist_free(nvp);
1775 		return (EPERM);
1776 	}
1777 
1778 	if ((error = nvlist_lookup_uint32_array(nvp, ZFS_DELEG_PERM_GROUPS,
1779 	    &groups, &group_cnt)) != 0) {
1780 		nvlist_free(nvp);
1781 		return (EPERM);
1782 	}
1783 	usercred = cralloc();
1784 	if ((crsetugid(usercred, uid, gid) != 0) ||
1785 	    (crsetgroups(usercred, group_cnt, (gid_t *)groups) != 0)) {
1786 		nvlist_free(nvp);
1787 		crfree(usercred);
1788 		return (EPERM);
1789 	}
1790 	nvlist_free(nvp);
1791 	error = dsl_deleg_access(zc->zc_name,
1792 	    zfs_prop_to_name(ZFS_PROP_SHAREISCSI), usercred);
1793 	crfree(usercred);
1794 	return (error);
1795 }
1796 
1797 /*
1798  * inputs:
1799  * zc_name		name of filesystem
1800  * zc_nvlist_src{_size}	nvlist of delegated permissions
1801  * zc_perm_action	allow/unallow flag
1802  *
1803  * outputs:		none
1804  */
1805 static int
1806 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
1807 {
1808 	int error;
1809 	nvlist_t *fsaclnv = NULL;
1810 
1811 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1812 	    &fsaclnv)) != 0)
1813 		return (error);
1814 
1815 	/*
1816 	 * Verify nvlist is constructed correctly
1817 	 */
1818 	if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
1819 		nvlist_free(fsaclnv);
1820 		return (EINVAL);
1821 	}
1822 
1823 	/*
1824 	 * If we don't have PRIV_SYS_MOUNT, then validate
1825 	 * that user is allowed to hand out each permission in
1826 	 * the nvlist(s)
1827 	 */
1828 
1829 	error = secpolicy_zfs(CRED());
1830 	if (error) {
1831 		if (zc->zc_perm_action == B_FALSE) {
1832 			error = dsl_deleg_can_allow(zc->zc_name,
1833 			    fsaclnv, CRED());
1834 		} else {
1835 			error = dsl_deleg_can_unallow(zc->zc_name,
1836 			    fsaclnv, CRED());
1837 		}
1838 	}
1839 
1840 	if (error == 0)
1841 		error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
1842 
1843 	nvlist_free(fsaclnv);
1844 	return (error);
1845 }
1846 
1847 /*
1848  * inputs:
1849  * zc_name		name of filesystem
1850  *
1851  * outputs:
1852  * zc_nvlist_src{_size}	nvlist of delegated permissions
1853  */
1854 static int
1855 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
1856 {
1857 	nvlist_t *nvp;
1858 	int error;
1859 
1860 	if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
1861 		error = put_nvlist(zc, nvp);
1862 		nvlist_free(nvp);
1863 	}
1864 
1865 	return (error);
1866 }
1867 
1868 /*
1869  * inputs:
1870  * zc_name		name of volume
1871  *
1872  * outputs:		none
1873  */
1874 static int
1875 zfs_ioc_create_minor(zfs_cmd_t *zc)
1876 {
1877 	return (zvol_create_minor(zc->zc_name, ddi_driver_major(zfs_dip)));
1878 }
1879 
1880 /*
1881  * inputs:
1882  * zc_name		name of volume
1883  *
1884  * outputs:		none
1885  */
1886 static int
1887 zfs_ioc_remove_minor(zfs_cmd_t *zc)
1888 {
1889 	return (zvol_remove_minor(zc->zc_name));
1890 }
1891 
1892 /*
1893  * Search the vfs list for a specified resource.  Returns a pointer to it
1894  * or NULL if no suitable entry is found. The caller of this routine
1895  * is responsible for releasing the returned vfs pointer.
1896  */
1897 static vfs_t *
1898 zfs_get_vfs(const char *resource)
1899 {
1900 	struct vfs *vfsp;
1901 	struct vfs *vfs_found = NULL;
1902 
1903 	vfs_list_read_lock();
1904 	vfsp = rootvfs;
1905 	do {
1906 		if (strcmp(refstr_value(vfsp->vfs_resource), resource) == 0) {
1907 			VFS_HOLD(vfsp);
1908 			vfs_found = vfsp;
1909 			break;
1910 		}
1911 		vfsp = vfsp->vfs_next;
1912 	} while (vfsp != rootvfs);
1913 	vfs_list_unlock();
1914 	return (vfs_found);
1915 }
1916 
1917 /* ARGSUSED */
1918 static void
1919 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
1920 {
1921 	zfs_creat_t *zct = arg;
1922 
1923 	zfs_create_fs(os, cr, zct->zct_zplprops, tx);
1924 }
1925 
1926 #define	ZFS_PROP_UNDEFINED	((uint64_t)-1)
1927 
1928 /*
1929  * inputs:
1930  * createprops		list of properties requested by creator
1931  * default_zplver	zpl version to use if unspecified in createprops
1932  * fuids_ok		fuids allowed in this version of the spa?
1933  * os			parent objset pointer (NULL if root fs)
1934  *
1935  * outputs:
1936  * zplprops	values for the zplprops we attach to the master node object
1937  * is_ci	true if requested file system will be purely case-insensitive
1938  *
1939  * Determine the settings for utf8only, normalization and
1940  * casesensitivity.  Specific values may have been requested by the
1941  * creator and/or we can inherit values from the parent dataset.  If
1942  * the file system is of too early a vintage, a creator can not
1943  * request settings for these properties, even if the requested
1944  * setting is the default value.  We don't actually want to create dsl
1945  * properties for these, so remove them from the source nvlist after
1946  * processing.
1947  */
1948 static int
1949 zfs_fill_zplprops_impl(objset_t *os, uint64_t default_zplver,
1950     boolean_t fuids_ok, nvlist_t *createprops, nvlist_t *zplprops,
1951     boolean_t *is_ci)
1952 {
1953 	uint64_t zplver = default_zplver;
1954 	uint64_t sense = ZFS_PROP_UNDEFINED;
1955 	uint64_t norm = ZFS_PROP_UNDEFINED;
1956 	uint64_t u8 = ZFS_PROP_UNDEFINED;
1957 
1958 	ASSERT(zplprops != NULL);
1959 
1960 	/*
1961 	 * Pull out creator prop choices, if any.
1962 	 */
1963 	if (createprops) {
1964 		(void) nvlist_lookup_uint64(createprops,
1965 		    zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
1966 		(void) nvlist_lookup_uint64(createprops,
1967 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
1968 		(void) nvlist_remove_all(createprops,
1969 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE));
1970 		(void) nvlist_lookup_uint64(createprops,
1971 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
1972 		(void) nvlist_remove_all(createprops,
1973 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
1974 		(void) nvlist_lookup_uint64(createprops,
1975 		    zfs_prop_to_name(ZFS_PROP_CASE), &sense);
1976 		(void) nvlist_remove_all(createprops,
1977 		    zfs_prop_to_name(ZFS_PROP_CASE));
1978 	}
1979 
1980 	/*
1981 	 * If the zpl version requested is whacky or the file system
1982 	 * or pool is version is too "young" to support normalization
1983 	 * and the creator tried to set a value for one of the props,
1984 	 * error out.
1985 	 */
1986 	if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
1987 	    (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
1988 	    (zplver < ZPL_VERSION_NORMALIZATION &&
1989 	    (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
1990 	    sense != ZFS_PROP_UNDEFINED)))
1991 		return (ENOTSUP);
1992 
1993 	/*
1994 	 * Put the version in the zplprops
1995 	 */
1996 	VERIFY(nvlist_add_uint64(zplprops,
1997 	    zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
1998 
1999 	if (norm == ZFS_PROP_UNDEFINED)
2000 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
2001 	VERIFY(nvlist_add_uint64(zplprops,
2002 	    zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
2003 
2004 	/*
2005 	 * If we're normalizing, names must always be valid UTF-8 strings.
2006 	 */
2007 	if (norm)
2008 		u8 = 1;
2009 	if (u8 == ZFS_PROP_UNDEFINED)
2010 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
2011 	VERIFY(nvlist_add_uint64(zplprops,
2012 	    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
2013 
2014 	if (sense == ZFS_PROP_UNDEFINED)
2015 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
2016 	VERIFY(nvlist_add_uint64(zplprops,
2017 	    zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
2018 
2019 	if (is_ci)
2020 		*is_ci = (sense == ZFS_CASE_INSENSITIVE);
2021 
2022 	return (0);
2023 }
2024 
2025 static int
2026 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
2027     nvlist_t *zplprops, boolean_t *is_ci)
2028 {
2029 	boolean_t fuids_ok = B_TRUE;
2030 	uint64_t zplver = ZPL_VERSION;
2031 	objset_t *os = NULL;
2032 	char parentname[MAXNAMELEN];
2033 	char *cp;
2034 	int error;
2035 
2036 	(void) strlcpy(parentname, dataset, sizeof (parentname));
2037 	cp = strrchr(parentname, '/');
2038 	ASSERT(cp != NULL);
2039 	cp[0] = '\0';
2040 
2041 	if (zfs_earlier_version(dataset, SPA_VERSION_FUID)) {
2042 		zplver = ZPL_VERSION_FUID - 1;
2043 		fuids_ok = B_FALSE;
2044 	}
2045 
2046 	/*
2047 	 * Open parent object set so we can inherit zplprop values.
2048 	 */
2049 	if ((error = dmu_objset_open(parentname, DMU_OST_ANY,
2050 	    DS_MODE_USER | DS_MODE_READONLY, &os)) != 0)
2051 		return (error);
2052 
2053 	error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, createprops,
2054 	    zplprops, is_ci);
2055 	dmu_objset_close(os);
2056 	return (error);
2057 }
2058 
2059 static int
2060 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
2061     nvlist_t *zplprops, boolean_t *is_ci)
2062 {
2063 	boolean_t fuids_ok = B_TRUE;
2064 	uint64_t zplver = ZPL_VERSION;
2065 	int error;
2066 
2067 	if (spa_vers < SPA_VERSION_FUID) {
2068 		zplver = ZPL_VERSION_FUID - 1;
2069 		fuids_ok = B_FALSE;
2070 	}
2071 
2072 	error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, createprops,
2073 	    zplprops, is_ci);
2074 	return (error);
2075 }
2076 
2077 /*
2078  * inputs:
2079  * zc_objset_type	type of objset to create (fs vs zvol)
2080  * zc_name		name of new objset
2081  * zc_value		name of snapshot to clone from (may be empty)
2082  * zc_nvlist_src{_size}	nvlist of properties to apply
2083  *
2084  * outputs: none
2085  */
2086 static int
2087 zfs_ioc_create(zfs_cmd_t *zc)
2088 {
2089 	objset_t *clone;
2090 	int error = 0;
2091 	zfs_creat_t zct;
2092 	nvlist_t *nvprops = NULL;
2093 	void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
2094 	dmu_objset_type_t type = zc->zc_objset_type;
2095 
2096 	switch (type) {
2097 
2098 	case DMU_OST_ZFS:
2099 		cbfunc = zfs_create_cb;
2100 		break;
2101 
2102 	case DMU_OST_ZVOL:
2103 		cbfunc = zvol_create_cb;
2104 		break;
2105 
2106 	default:
2107 		cbfunc = NULL;
2108 		break;
2109 	}
2110 	if (strchr(zc->zc_name, '@') ||
2111 	    strchr(zc->zc_name, '%'))
2112 		return (EINVAL);
2113 
2114 	if (zc->zc_nvlist_src != NULL &&
2115 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2116 	    &nvprops)) != 0)
2117 		return (error);
2118 
2119 	zct.zct_zplprops = NULL;
2120 	zct.zct_props = nvprops;
2121 
2122 	if (zc->zc_value[0] != '\0') {
2123 		/*
2124 		 * We're creating a clone of an existing snapshot.
2125 		 */
2126 		zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
2127 		if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0) {
2128 			nvlist_free(nvprops);
2129 			return (EINVAL);
2130 		}
2131 
2132 		error = dmu_objset_open(zc->zc_value, type,
2133 		    DS_MODE_USER | DS_MODE_READONLY, &clone);
2134 		if (error) {
2135 			nvlist_free(nvprops);
2136 			return (error);
2137 		}
2138 
2139 		error = dmu_objset_create(zc->zc_name, type, clone, 0,
2140 		    NULL, NULL);
2141 		if (error) {
2142 			dmu_objset_close(clone);
2143 			nvlist_free(nvprops);
2144 			return (error);
2145 		}
2146 		dmu_objset_close(clone);
2147 	} else {
2148 		boolean_t is_insensitive = B_FALSE;
2149 
2150 		if (cbfunc == NULL) {
2151 			nvlist_free(nvprops);
2152 			return (EINVAL);
2153 		}
2154 
2155 		if (type == DMU_OST_ZVOL) {
2156 			uint64_t volsize, volblocksize;
2157 
2158 			if (nvprops == NULL ||
2159 			    nvlist_lookup_uint64(nvprops,
2160 			    zfs_prop_to_name(ZFS_PROP_VOLSIZE),
2161 			    &volsize) != 0) {
2162 				nvlist_free(nvprops);
2163 				return (EINVAL);
2164 			}
2165 
2166 			if ((error = nvlist_lookup_uint64(nvprops,
2167 			    zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
2168 			    &volblocksize)) != 0 && error != ENOENT) {
2169 				nvlist_free(nvprops);
2170 				return (EINVAL);
2171 			}
2172 
2173 			if (error != 0)
2174 				volblocksize = zfs_prop_default_numeric(
2175 				    ZFS_PROP_VOLBLOCKSIZE);
2176 
2177 			if ((error = zvol_check_volblocksize(
2178 			    volblocksize)) != 0 ||
2179 			    (error = zvol_check_volsize(volsize,
2180 			    volblocksize)) != 0) {
2181 				nvlist_free(nvprops);
2182 				return (error);
2183 			}
2184 		} else if (type == DMU_OST_ZFS) {
2185 			int error;
2186 
2187 			/*
2188 			 * We have to have normalization and
2189 			 * case-folding flags correct when we do the
2190 			 * file system creation, so go figure them out
2191 			 * now.
2192 			 */
2193 			VERIFY(nvlist_alloc(&zct.zct_zplprops,
2194 			    NV_UNIQUE_NAME, KM_SLEEP) == 0);
2195 			error = zfs_fill_zplprops(zc->zc_name, nvprops,
2196 			    zct.zct_zplprops, &is_insensitive);
2197 			if (error != 0) {
2198 				nvlist_free(nvprops);
2199 				nvlist_free(zct.zct_zplprops);
2200 				return (error);
2201 			}
2202 		}
2203 		error = dmu_objset_create(zc->zc_name, type, NULL,
2204 		    is_insensitive ? DS_FLAG_CI_DATASET : 0, cbfunc, &zct);
2205 		nvlist_free(zct.zct_zplprops);
2206 	}
2207 
2208 	/*
2209 	 * It would be nice to do this atomically.
2210 	 */
2211 	if (error == 0) {
2212 		if ((error = zfs_set_prop_nvlist(zc->zc_name, nvprops)) != 0)
2213 			(void) dmu_objset_destroy(zc->zc_name);
2214 	}
2215 	nvlist_free(nvprops);
2216 	return (error);
2217 }
2218 
2219 struct snap_prop_arg {
2220 	nvlist_t *nvprops;
2221 	const char *snapname;
2222 };
2223 
2224 static int
2225 set_snap_props(char *name, void *arg)
2226 {
2227 	struct snap_prop_arg *snpa = arg;
2228 	int len = strlen(name) + strlen(snpa->snapname) + 2;
2229 	char *buf = kmem_alloc(len, KM_SLEEP);
2230 	int err;
2231 
2232 	(void) snprintf(buf, len, "%s@%s", name, snpa->snapname);
2233 	err = zfs_set_prop_nvlist(buf, snpa->nvprops);
2234 	if (err)
2235 		(void) dmu_objset_destroy(buf);
2236 	kmem_free(buf, len);
2237 	return (err);
2238 }
2239 
2240 /*
2241  * inputs:
2242  * zc_name	name of filesystem
2243  * zc_value	short name of snapshot
2244  * zc_cookie	recursive flag
2245  *
2246  * outputs:	none
2247  */
2248 static int
2249 zfs_ioc_snapshot(zfs_cmd_t *zc)
2250 {
2251 	nvlist_t *nvprops = NULL;
2252 	int error;
2253 	boolean_t recursive = zc->zc_cookie;
2254 
2255 	if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
2256 		return (EINVAL);
2257 
2258 	if (zc->zc_nvlist_src != NULL &&
2259 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2260 	    &nvprops)) != 0)
2261 		return (error);
2262 
2263 	error = dmu_objset_snapshot(zc->zc_name, zc->zc_value, recursive);
2264 
2265 	/*
2266 	 * It would be nice to do this atomically.
2267 	 */
2268 	if (error == 0) {
2269 		struct snap_prop_arg snpa;
2270 		snpa.nvprops = nvprops;
2271 		snpa.snapname = zc->zc_value;
2272 		if (recursive) {
2273 			error = dmu_objset_find(zc->zc_name,
2274 			    set_snap_props, &snpa, DS_FIND_CHILDREN);
2275 			if (error) {
2276 				(void) dmu_snapshots_destroy(zc->zc_name,
2277 				    zc->zc_value);
2278 			}
2279 		} else {
2280 			error = set_snap_props(zc->zc_name, &snpa);
2281 		}
2282 	}
2283 	nvlist_free(nvprops);
2284 	return (error);
2285 }
2286 
2287 int
2288 zfs_unmount_snap(char *name, void *arg)
2289 {
2290 	vfs_t *vfsp = NULL;
2291 
2292 	if (arg) {
2293 		char *snapname = arg;
2294 		int len = strlen(name) + strlen(snapname) + 2;
2295 		char *buf = kmem_alloc(len, KM_SLEEP);
2296 
2297 		(void) strcpy(buf, name);
2298 		(void) strcat(buf, "@");
2299 		(void) strcat(buf, snapname);
2300 		vfsp = zfs_get_vfs(buf);
2301 		kmem_free(buf, len);
2302 	} else if (strchr(name, '@')) {
2303 		vfsp = zfs_get_vfs(name);
2304 	}
2305 
2306 	if (vfsp) {
2307 		/*
2308 		 * Always force the unmount for snapshots.
2309 		 */
2310 		int flag = MS_FORCE;
2311 		int err;
2312 
2313 		if ((err = vn_vfswlock(vfsp->vfs_vnodecovered)) != 0) {
2314 			VFS_RELE(vfsp);
2315 			return (err);
2316 		}
2317 		VFS_RELE(vfsp);
2318 		if ((err = dounmount(vfsp, flag, kcred)) != 0)
2319 			return (err);
2320 	}
2321 	return (0);
2322 }
2323 
2324 /*
2325  * inputs:
2326  * zc_name	name of filesystem
2327  * zc_value	short name of snapshot
2328  *
2329  * outputs:	none
2330  */
2331 static int
2332 zfs_ioc_destroy_snaps(zfs_cmd_t *zc)
2333 {
2334 	int err;
2335 
2336 	if (snapshot_namecheck(zc->zc_value, NULL, NULL) != 0)
2337 		return (EINVAL);
2338 	err = dmu_objset_find(zc->zc_name,
2339 	    zfs_unmount_snap, zc->zc_value, DS_FIND_CHILDREN);
2340 	if (err)
2341 		return (err);
2342 	return (dmu_snapshots_destroy(zc->zc_name, zc->zc_value));
2343 }
2344 
2345 /*
2346  * inputs:
2347  * zc_name		name of dataset to destroy
2348  * zc_objset_type	type of objset
2349  *
2350  * outputs:		none
2351  */
2352 static int
2353 zfs_ioc_destroy(zfs_cmd_t *zc)
2354 {
2355 	if (strchr(zc->zc_name, '@') && zc->zc_objset_type == DMU_OST_ZFS) {
2356 		int err = zfs_unmount_snap(zc->zc_name, NULL);
2357 		if (err)
2358 			return (err);
2359 	}
2360 
2361 	return (dmu_objset_destroy(zc->zc_name));
2362 }
2363 
2364 /*
2365  * inputs:
2366  * zc_name	name of dataset to rollback (to most recent snapshot)
2367  *
2368  * outputs:	none
2369  */
2370 static int
2371 zfs_ioc_rollback(zfs_cmd_t *zc)
2372 {
2373 	objset_t *os;
2374 	int error;
2375 	zfsvfs_t *zfsvfs = NULL;
2376 
2377 	/*
2378 	 * Get the zfsvfs for the receiving objset. There
2379 	 * won't be one if we're operating on a zvol, if the
2380 	 * objset doesn't exist yet, or is not mounted.
2381 	 */
2382 	error = dmu_objset_open(zc->zc_name, DMU_OST_ANY, DS_MODE_USER, &os);
2383 	if (error)
2384 		return (error);
2385 
2386 	if (dmu_objset_type(os) == DMU_OST_ZFS) {
2387 		mutex_enter(&os->os->os_user_ptr_lock);
2388 		zfsvfs = dmu_objset_get_user(os);
2389 		if (zfsvfs != NULL)
2390 			VFS_HOLD(zfsvfs->z_vfs);
2391 		mutex_exit(&os->os->os_user_ptr_lock);
2392 	}
2393 
2394 	if (zfsvfs != NULL) {
2395 		char *osname;
2396 		int mode;
2397 
2398 		osname = kmem_alloc(MAXNAMELEN, KM_SLEEP);
2399 		error = zfs_suspend_fs(zfsvfs, osname, &mode);
2400 		if (error == 0) {
2401 			int resume_err;
2402 
2403 			ASSERT(strcmp(osname, zc->zc_name) == 0);
2404 			error = dmu_objset_rollback(os);
2405 			resume_err = zfs_resume_fs(zfsvfs, osname, mode);
2406 			error = error ? error : resume_err;
2407 		} else {
2408 			dmu_objset_close(os);
2409 		}
2410 		kmem_free(osname, MAXNAMELEN);
2411 		VFS_RELE(zfsvfs->z_vfs);
2412 	} else {
2413 		error = dmu_objset_rollback(os);
2414 	}
2415 	/* Note, the dmu_objset_rollback() releases the objset for us. */
2416 
2417 	return (error);
2418 }
2419 
2420 /*
2421  * inputs:
2422  * zc_name	old name of dataset
2423  * zc_value	new name of dataset
2424  * zc_cookie	recursive flag (only valid for snapshots)
2425  *
2426  * outputs:	none
2427  */
2428 static int
2429 zfs_ioc_rename(zfs_cmd_t *zc)
2430 {
2431 	boolean_t recursive = zc->zc_cookie & 1;
2432 
2433 	zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
2434 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
2435 	    strchr(zc->zc_value, '%'))
2436 		return (EINVAL);
2437 
2438 	/*
2439 	 * Unmount snapshot unless we're doing a recursive rename,
2440 	 * in which case the dataset code figures out which snapshots
2441 	 * to unmount.
2442 	 */
2443 	if (!recursive && strchr(zc->zc_name, '@') != NULL &&
2444 	    zc->zc_objset_type == DMU_OST_ZFS) {
2445 		int err = zfs_unmount_snap(zc->zc_name, NULL);
2446 		if (err)
2447 			return (err);
2448 	}
2449 	return (dmu_objset_rename(zc->zc_name, zc->zc_value, recursive));
2450 }
2451 
2452 static void
2453 clear_props(char *dataset, nvlist_t *props, nvlist_t *newprops)
2454 {
2455 	zfs_cmd_t *zc;
2456 	nvpair_t *prop;
2457 
2458 	if (props == NULL)
2459 		return;
2460 	zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
2461 	(void) strcpy(zc->zc_name, dataset);
2462 	for (prop = nvlist_next_nvpair(props, NULL); prop;
2463 	    prop = nvlist_next_nvpair(props, prop)) {
2464 		if (newprops != NULL &&
2465 		    nvlist_exists(newprops, nvpair_name(prop)))
2466 			continue;
2467 		(void) strcpy(zc->zc_value, nvpair_name(prop));
2468 		if (zfs_secpolicy_inherit(zc, CRED()) == 0)
2469 			(void) zfs_ioc_inherit_prop(zc);
2470 	}
2471 	kmem_free(zc, sizeof (zfs_cmd_t));
2472 }
2473 
2474 /*
2475  * inputs:
2476  * zc_name		name of containing filesystem
2477  * zc_nvlist_src{_size}	nvlist of properties to apply
2478  * zc_value		name of snapshot to create
2479  * zc_string		name of clone origin (if DRR_FLAG_CLONE)
2480  * zc_cookie		file descriptor to recv from
2481  * zc_begin_record	the BEGIN record of the stream (not byteswapped)
2482  * zc_guid		force flag
2483  *
2484  * outputs:
2485  * zc_cookie		number of bytes read
2486  */
2487 static int
2488 zfs_ioc_recv(zfs_cmd_t *zc)
2489 {
2490 	file_t *fp;
2491 	objset_t *os;
2492 	dmu_recv_cookie_t drc;
2493 	zfsvfs_t *zfsvfs = NULL;
2494 	boolean_t force = (boolean_t)zc->zc_guid;
2495 	int error, fd;
2496 	offset_t off;
2497 	nvlist_t *props = NULL;
2498 	nvlist_t *origprops = NULL;
2499 	objset_t *origin = NULL;
2500 	char *tosnap;
2501 	char tofs[ZFS_MAXNAMELEN];
2502 
2503 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
2504 	    strchr(zc->zc_value, '@') == NULL ||
2505 	    strchr(zc->zc_value, '%'))
2506 		return (EINVAL);
2507 
2508 	(void) strcpy(tofs, zc->zc_value);
2509 	tosnap = strchr(tofs, '@');
2510 	*tosnap = '\0';
2511 	tosnap++;
2512 
2513 	if (zc->zc_nvlist_src != NULL &&
2514 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2515 	    &props)) != 0)
2516 		return (error);
2517 
2518 	fd = zc->zc_cookie;
2519 	fp = getf(fd);
2520 	if (fp == NULL) {
2521 		nvlist_free(props);
2522 		return (EBADF);
2523 	}
2524 
2525 	if (dmu_objset_open(tofs, DMU_OST_ANY,
2526 	    DS_MODE_USER | DS_MODE_READONLY, &os) == 0) {
2527 		/*
2528 		 * Try to get the zfsvfs for the receiving objset.
2529 		 * There won't be one if we're operating on a zvol,
2530 		 * if the objset doesn't exist yet, or is not mounted.
2531 		 */
2532 		mutex_enter(&os->os->os_user_ptr_lock);
2533 		if (zfsvfs = dmu_objset_get_user(os)) {
2534 			if (!mutex_tryenter(&zfsvfs->z_online_recv_lock)) {
2535 				mutex_exit(&os->os->os_user_ptr_lock);
2536 				dmu_objset_close(os);
2537 				zfsvfs = NULL;
2538 				error = EBUSY;
2539 				goto out;
2540 			}
2541 			VFS_HOLD(zfsvfs->z_vfs);
2542 		}
2543 		mutex_exit(&os->os->os_user_ptr_lock);
2544 
2545 		/*
2546 		 * If new properties are supplied, they are to completely
2547 		 * replace the existing ones, so stash away the existing ones.
2548 		 */
2549 		if (props)
2550 			(void) dsl_prop_get_all(os, &origprops, TRUE);
2551 
2552 		dmu_objset_close(os);
2553 	}
2554 
2555 	if (zc->zc_string[0]) {
2556 		error = dmu_objset_open(zc->zc_string, DMU_OST_ANY,
2557 		    DS_MODE_USER | DS_MODE_READONLY, &origin);
2558 		if (error)
2559 			goto out;
2560 	}
2561 
2562 	error = dmu_recv_begin(tofs, tosnap, &zc->zc_begin_record,
2563 	    force, origin, zfsvfs != NULL, &drc);
2564 	if (origin)
2565 		dmu_objset_close(origin);
2566 	if (error)
2567 		goto out;
2568 
2569 	/*
2570 	 * Reset properties.  We do this before we receive the stream
2571 	 * so that the properties are applied to the new data.
2572 	 */
2573 	if (props) {
2574 		clear_props(tofs, origprops, props);
2575 		/*
2576 		 * XXX - Note, this is all-or-nothing; should be best-effort.
2577 		 */
2578 		(void) zfs_set_prop_nvlist(tofs, props);
2579 	}
2580 
2581 	off = fp->f_offset;
2582 	error = dmu_recv_stream(&drc, fp->f_vnode, &off);
2583 
2584 	if (error == 0 && zfsvfs) {
2585 		char *osname;
2586 		int mode;
2587 
2588 		/* online recv */
2589 		osname = kmem_alloc(MAXNAMELEN, KM_SLEEP);
2590 		error = zfs_suspend_fs(zfsvfs, osname, &mode);
2591 		if (error == 0) {
2592 			int resume_err;
2593 
2594 			error = dmu_recv_end(&drc);
2595 			resume_err = zfs_resume_fs(zfsvfs, osname, mode);
2596 			error = error ? error : resume_err;
2597 		} else {
2598 			dmu_recv_abort_cleanup(&drc);
2599 		}
2600 		kmem_free(osname, MAXNAMELEN);
2601 	} else if (error == 0) {
2602 		error = dmu_recv_end(&drc);
2603 	}
2604 
2605 	zc->zc_cookie = off - fp->f_offset;
2606 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
2607 		fp->f_offset = off;
2608 
2609 	/*
2610 	 * On error, restore the original props.
2611 	 */
2612 	if (error && props) {
2613 		clear_props(tofs, props, NULL);
2614 		(void) zfs_set_prop_nvlist(tofs, origprops);
2615 	}
2616 out:
2617 	if (zfsvfs) {
2618 		mutex_exit(&zfsvfs->z_online_recv_lock);
2619 		VFS_RELE(zfsvfs->z_vfs);
2620 	}
2621 	nvlist_free(props);
2622 	nvlist_free(origprops);
2623 	releasef(fd);
2624 	return (error);
2625 }
2626 
2627 /*
2628  * inputs:
2629  * zc_name	name of snapshot to send
2630  * zc_value	short name of incremental fromsnap (may be empty)
2631  * zc_cookie	file descriptor to send stream to
2632  * zc_obj	fromorigin flag (mutually exclusive with zc_value)
2633  *
2634  * outputs: none
2635  */
2636 static int
2637 zfs_ioc_send(zfs_cmd_t *zc)
2638 {
2639 	objset_t *fromsnap = NULL;
2640 	objset_t *tosnap;
2641 	file_t *fp;
2642 	int error;
2643 	offset_t off;
2644 
2645 	error = dmu_objset_open(zc->zc_name, DMU_OST_ANY,
2646 	    DS_MODE_USER | DS_MODE_READONLY, &tosnap);
2647 	if (error)
2648 		return (error);
2649 
2650 	if (zc->zc_value[0] != '\0') {
2651 		char *buf;
2652 		char *cp;
2653 
2654 		buf = kmem_alloc(MAXPATHLEN, KM_SLEEP);
2655 		(void) strncpy(buf, zc->zc_name, MAXPATHLEN);
2656 		cp = strchr(buf, '@');
2657 		if (cp)
2658 			*(cp+1) = 0;
2659 		(void) strncat(buf, zc->zc_value, MAXPATHLEN);
2660 		error = dmu_objset_open(buf, DMU_OST_ANY,
2661 		    DS_MODE_USER | DS_MODE_READONLY, &fromsnap);
2662 		kmem_free(buf, MAXPATHLEN);
2663 		if (error) {
2664 			dmu_objset_close(tosnap);
2665 			return (error);
2666 		}
2667 	}
2668 
2669 	fp = getf(zc->zc_cookie);
2670 	if (fp == NULL) {
2671 		dmu_objset_close(tosnap);
2672 		if (fromsnap)
2673 			dmu_objset_close(fromsnap);
2674 		return (EBADF);
2675 	}
2676 
2677 	off = fp->f_offset;
2678 	error = dmu_sendbackup(tosnap, fromsnap, zc->zc_obj, fp->f_vnode, &off);
2679 
2680 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
2681 		fp->f_offset = off;
2682 	releasef(zc->zc_cookie);
2683 	if (fromsnap)
2684 		dmu_objset_close(fromsnap);
2685 	dmu_objset_close(tosnap);
2686 	return (error);
2687 }
2688 
2689 static int
2690 zfs_ioc_inject_fault(zfs_cmd_t *zc)
2691 {
2692 	int id, error;
2693 
2694 	error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
2695 	    &zc->zc_inject_record);
2696 
2697 	if (error == 0)
2698 		zc->zc_guid = (uint64_t)id;
2699 
2700 	return (error);
2701 }
2702 
2703 static int
2704 zfs_ioc_clear_fault(zfs_cmd_t *zc)
2705 {
2706 	return (zio_clear_fault((int)zc->zc_guid));
2707 }
2708 
2709 static int
2710 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
2711 {
2712 	int id = (int)zc->zc_guid;
2713 	int error;
2714 
2715 	error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
2716 	    &zc->zc_inject_record);
2717 
2718 	zc->zc_guid = id;
2719 
2720 	return (error);
2721 }
2722 
2723 static int
2724 zfs_ioc_error_log(zfs_cmd_t *zc)
2725 {
2726 	spa_t *spa;
2727 	int error;
2728 	size_t count = (size_t)zc->zc_nvlist_dst_size;
2729 
2730 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2731 		return (error);
2732 
2733 	error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
2734 	    &count);
2735 	if (error == 0)
2736 		zc->zc_nvlist_dst_size = count;
2737 	else
2738 		zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
2739 
2740 	spa_close(spa, FTAG);
2741 
2742 	return (error);
2743 }
2744 
2745 static int
2746 zfs_ioc_clear(zfs_cmd_t *zc)
2747 {
2748 	spa_t *spa;
2749 	vdev_t *vd;
2750 	int error;
2751 
2752 	/*
2753 	 * On zpool clear we also fix up missing slogs
2754 	 */
2755 	mutex_enter(&spa_namespace_lock);
2756 	spa = spa_lookup(zc->zc_name);
2757 	if (spa == NULL) {
2758 		mutex_exit(&spa_namespace_lock);
2759 		return (EIO);
2760 	}
2761 	if (spa->spa_log_state == SPA_LOG_MISSING) {
2762 		/* we need to let spa_open/spa_load clear the chains */
2763 		spa->spa_log_state = SPA_LOG_CLEAR;
2764 	}
2765 	mutex_exit(&spa_namespace_lock);
2766 
2767 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2768 		return (error);
2769 
2770 	spa_vdev_state_enter(spa);
2771 
2772 	if (zc->zc_guid == 0) {
2773 		vd = NULL;
2774 	} else {
2775 		vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
2776 		if (vd == NULL) {
2777 			(void) spa_vdev_state_exit(spa, NULL, ENODEV);
2778 			spa_close(spa, FTAG);
2779 			return (ENODEV);
2780 		}
2781 	}
2782 
2783 	vdev_clear(spa, vd);
2784 
2785 	(void) spa_vdev_state_exit(spa, NULL, 0);
2786 
2787 	/*
2788 	 * Resume any suspended I/Os.
2789 	 */
2790 	zio_resume(spa);
2791 
2792 	spa_close(spa, FTAG);
2793 
2794 	return (0);
2795 }
2796 
2797 /*
2798  * inputs:
2799  * zc_name	name of filesystem
2800  * zc_value	name of origin snapshot
2801  *
2802  * outputs:	none
2803  */
2804 static int
2805 zfs_ioc_promote(zfs_cmd_t *zc)
2806 {
2807 	char *cp;
2808 
2809 	/*
2810 	 * We don't need to unmount *all* the origin fs's snapshots, but
2811 	 * it's easier.
2812 	 */
2813 	cp = strchr(zc->zc_value, '@');
2814 	if (cp)
2815 		*cp = '\0';
2816 	(void) dmu_objset_find(zc->zc_value,
2817 	    zfs_unmount_snap, NULL, DS_FIND_SNAPSHOTS);
2818 	return (dsl_dataset_promote(zc->zc_name));
2819 }
2820 
2821 /*
2822  * We don't want to have a hard dependency
2823  * against some special symbols in sharefs
2824  * nfs, and smbsrv.  Determine them if needed when
2825  * the first file system is shared.
2826  * Neither sharefs, nfs or smbsrv are unloadable modules.
2827  */
2828 int (*znfsexport_fs)(void *arg);
2829 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
2830 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
2831 
2832 int zfs_nfsshare_inited;
2833 int zfs_smbshare_inited;
2834 
2835 ddi_modhandle_t nfs_mod;
2836 ddi_modhandle_t sharefs_mod;
2837 ddi_modhandle_t smbsrv_mod;
2838 kmutex_t zfs_share_lock;
2839 
2840 static int
2841 zfs_init_sharefs()
2842 {
2843 	int error;
2844 
2845 	ASSERT(MUTEX_HELD(&zfs_share_lock));
2846 	/* Both NFS and SMB shares also require sharetab support. */
2847 	if (sharefs_mod == NULL && ((sharefs_mod =
2848 	    ddi_modopen("fs/sharefs",
2849 	    KRTLD_MODE_FIRST, &error)) == NULL)) {
2850 		return (ENOSYS);
2851 	}
2852 	if (zshare_fs == NULL && ((zshare_fs =
2853 	    (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
2854 	    ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
2855 		return (ENOSYS);
2856 	}
2857 	return (0);
2858 }
2859 
2860 static int
2861 zfs_ioc_share(zfs_cmd_t *zc)
2862 {
2863 	int error;
2864 	int opcode;
2865 
2866 	switch (zc->zc_share.z_sharetype) {
2867 	case ZFS_SHARE_NFS:
2868 	case ZFS_UNSHARE_NFS:
2869 		if (zfs_nfsshare_inited == 0) {
2870 			mutex_enter(&zfs_share_lock);
2871 			if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
2872 			    KRTLD_MODE_FIRST, &error)) == NULL)) {
2873 				mutex_exit(&zfs_share_lock);
2874 				return (ENOSYS);
2875 			}
2876 			if (znfsexport_fs == NULL &&
2877 			    ((znfsexport_fs = (int (*)(void *))
2878 			    ddi_modsym(nfs_mod,
2879 			    "nfs_export", &error)) == NULL)) {
2880 				mutex_exit(&zfs_share_lock);
2881 				return (ENOSYS);
2882 			}
2883 			error = zfs_init_sharefs();
2884 			if (error) {
2885 				mutex_exit(&zfs_share_lock);
2886 				return (ENOSYS);
2887 			}
2888 			zfs_nfsshare_inited = 1;
2889 			mutex_exit(&zfs_share_lock);
2890 		}
2891 		break;
2892 	case ZFS_SHARE_SMB:
2893 	case ZFS_UNSHARE_SMB:
2894 		if (zfs_smbshare_inited == 0) {
2895 			mutex_enter(&zfs_share_lock);
2896 			if (smbsrv_mod == NULL && ((smbsrv_mod =
2897 			    ddi_modopen("drv/smbsrv",
2898 			    KRTLD_MODE_FIRST, &error)) == NULL)) {
2899 				mutex_exit(&zfs_share_lock);
2900 				return (ENOSYS);
2901 			}
2902 			if (zsmbexport_fs == NULL && ((zsmbexport_fs =
2903 			    (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
2904 			    "smb_server_share", &error)) == NULL)) {
2905 				mutex_exit(&zfs_share_lock);
2906 				return (ENOSYS);
2907 			}
2908 			error = zfs_init_sharefs();
2909 			if (error) {
2910 				mutex_exit(&zfs_share_lock);
2911 				return (ENOSYS);
2912 			}
2913 			zfs_smbshare_inited = 1;
2914 			mutex_exit(&zfs_share_lock);
2915 		}
2916 		break;
2917 	default:
2918 		return (EINVAL);
2919 	}
2920 
2921 	switch (zc->zc_share.z_sharetype) {
2922 	case ZFS_SHARE_NFS:
2923 	case ZFS_UNSHARE_NFS:
2924 		if (error =
2925 		    znfsexport_fs((void *)
2926 		    (uintptr_t)zc->zc_share.z_exportdata))
2927 			return (error);
2928 		break;
2929 	case ZFS_SHARE_SMB:
2930 	case ZFS_UNSHARE_SMB:
2931 		if (error = zsmbexport_fs((void *)
2932 		    (uintptr_t)zc->zc_share.z_exportdata,
2933 		    zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
2934 		    B_TRUE: B_FALSE)) {
2935 			return (error);
2936 		}
2937 		break;
2938 	}
2939 
2940 	opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
2941 	    zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
2942 	    SHAREFS_ADD : SHAREFS_REMOVE;
2943 
2944 	/*
2945 	 * Add or remove share from sharetab
2946 	 */
2947 	error = zshare_fs(opcode,
2948 	    (void *)(uintptr_t)zc->zc_share.z_sharedata,
2949 	    zc->zc_share.z_sharemax);
2950 
2951 	return (error);
2952 
2953 }
2954 
2955 ace_t full_access[] = {
2956 	{(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
2957 };
2958 
2959 /*
2960  * Remove all ACL files in shares dir
2961  */
2962 static int
2963 zfs_smb_acl_purge(znode_t *dzp)
2964 {
2965 	zap_cursor_t	zc;
2966 	zap_attribute_t	zap;
2967 	zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
2968 	int error;
2969 
2970 	for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
2971 	    (error = zap_cursor_retrieve(&zc, &zap)) == 0;
2972 	    zap_cursor_advance(&zc)) {
2973 		if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
2974 		    NULL, 0)) != 0)
2975 			break;
2976 	}
2977 	zap_cursor_fini(&zc);
2978 	return (error);
2979 }
2980 
2981 static int
2982 zfs_ioc_smb_acl(zfs_cmd_t *zc)
2983 {
2984 	vnode_t *vp;
2985 	znode_t *dzp;
2986 	vnode_t *resourcevp = NULL;
2987 	znode_t *sharedir;
2988 	zfsvfs_t *zfsvfs;
2989 	nvlist_t *nvlist;
2990 	char *src, *target;
2991 	vattr_t vattr;
2992 	vsecattr_t vsec;
2993 	int error = 0;
2994 
2995 	if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
2996 	    NO_FOLLOW, NULL, &vp)) != 0)
2997 		return (error);
2998 
2999 	/* Now make sure mntpnt and dataset are ZFS */
3000 
3001 	if (vp->v_vfsp->vfs_fstype != zfsfstype ||
3002 	    (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
3003 	    zc->zc_name) != 0)) {
3004 		VN_RELE(vp);
3005 		return (EINVAL);
3006 	}
3007 
3008 	dzp = VTOZ(vp);
3009 	zfsvfs = dzp->z_zfsvfs;
3010 
3011 	ZFS_ENTER(zfsvfs);
3012 
3013 	if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
3014 		ZFS_EXIT(zfsvfs);
3015 		return (error);
3016 	}
3017 
3018 	switch (zc->zc_cookie) {
3019 	case ZFS_SMB_ACL_ADD:
3020 		vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
3021 		vattr.va_type = VREG;
3022 		vattr.va_mode = S_IFREG|0777;
3023 		vattr.va_uid = 0;
3024 		vattr.va_gid = 0;
3025 
3026 		vsec.vsa_mask = VSA_ACE;
3027 		vsec.vsa_aclentp = &full_access;
3028 		vsec.vsa_aclentsz = sizeof (full_access);
3029 		vsec.vsa_aclcnt = 1;
3030 
3031 		error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
3032 		    &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
3033 		if (resourcevp)
3034 			VN_RELE(resourcevp);
3035 		break;
3036 
3037 	case ZFS_SMB_ACL_REMOVE:
3038 		error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
3039 		    NULL, 0);
3040 		break;
3041 
3042 	case ZFS_SMB_ACL_RENAME:
3043 		if ((error = get_nvlist(zc->zc_nvlist_src,
3044 		    zc->zc_nvlist_src_size, &nvlist)) != 0) {
3045 			VN_RELE(vp);
3046 			ZFS_EXIT(zfsvfs);
3047 			return (error);
3048 		}
3049 		if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
3050 		    nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
3051 		    &target)) {
3052 			VN_RELE(vp);
3053 			ZFS_EXIT(zfsvfs);
3054 			return (error);
3055 		}
3056 		error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
3057 		    kcred, NULL, 0);
3058 		nvlist_free(nvlist);
3059 		break;
3060 
3061 	case ZFS_SMB_ACL_PURGE:
3062 		error = zfs_smb_acl_purge(sharedir);
3063 		break;
3064 
3065 	default:
3066 		error = EINVAL;
3067 		break;
3068 	}
3069 
3070 	VN_RELE(vp);
3071 	VN_RELE(ZTOV(sharedir));
3072 
3073 	ZFS_EXIT(zfsvfs);
3074 
3075 	return (error);
3076 }
3077 
3078 /*
3079  * pool create, destroy, and export don't log the history as part of
3080  * zfsdev_ioctl, but rather zfs_ioc_pool_create, and zfs_ioc_pool_export
3081  * do the logging of those commands.
3082  */
3083 static zfs_ioc_vec_t zfs_ioc_vec[] = {
3084 	{ zfs_ioc_pool_create, zfs_secpolicy_config, POOL_NAME, B_FALSE },
3085 	{ zfs_ioc_pool_destroy,	zfs_secpolicy_config, POOL_NAME, B_FALSE },
3086 	{ zfs_ioc_pool_import, zfs_secpolicy_config, POOL_NAME, B_TRUE },
3087 	{ zfs_ioc_pool_export, zfs_secpolicy_config, POOL_NAME, B_FALSE },
3088 	{ zfs_ioc_pool_configs,	zfs_secpolicy_none, NO_NAME, B_FALSE },
3089 	{ zfs_ioc_pool_stats, zfs_secpolicy_read, POOL_NAME, B_FALSE },
3090 	{ zfs_ioc_pool_tryimport, zfs_secpolicy_config, NO_NAME, B_FALSE },
3091 	{ zfs_ioc_pool_scrub, zfs_secpolicy_config, POOL_NAME, B_TRUE },
3092 	{ zfs_ioc_pool_freeze, zfs_secpolicy_config, NO_NAME, B_FALSE },
3093 	{ zfs_ioc_pool_upgrade,	zfs_secpolicy_config, POOL_NAME, B_TRUE },
3094 	{ zfs_ioc_pool_get_history, zfs_secpolicy_config, POOL_NAME, B_FALSE },
3095 	{ zfs_ioc_vdev_add, zfs_secpolicy_config, POOL_NAME, B_TRUE },
3096 	{ zfs_ioc_vdev_remove, zfs_secpolicy_config, POOL_NAME, B_TRUE },
3097 	{ zfs_ioc_vdev_set_state, zfs_secpolicy_config,	POOL_NAME, B_TRUE },
3098 	{ zfs_ioc_vdev_attach, zfs_secpolicy_config, POOL_NAME, B_TRUE },
3099 	{ zfs_ioc_vdev_detach, zfs_secpolicy_config, POOL_NAME, B_TRUE },
3100 	{ zfs_ioc_vdev_setpath,	zfs_secpolicy_config, POOL_NAME, B_FALSE },
3101 	{ zfs_ioc_objset_stats,	zfs_secpolicy_read, DATASET_NAME, B_FALSE },
3102 	{ zfs_ioc_objset_zplprops, zfs_secpolicy_read, DATASET_NAME, B_FALSE },
3103 	{ zfs_ioc_dataset_list_next, zfs_secpolicy_read,
3104 	    DATASET_NAME, B_FALSE },
3105 	{ zfs_ioc_snapshot_list_next, zfs_secpolicy_read,
3106 	    DATASET_NAME, B_FALSE },
3107 	{ zfs_ioc_set_prop, zfs_secpolicy_none, DATASET_NAME, B_TRUE },
3108 	{ zfs_ioc_create_minor,	zfs_secpolicy_minor, DATASET_NAME, B_FALSE },
3109 	{ zfs_ioc_remove_minor,	zfs_secpolicy_minor, DATASET_NAME, B_FALSE },
3110 	{ zfs_ioc_create, zfs_secpolicy_create, DATASET_NAME, B_TRUE },
3111 	{ zfs_ioc_destroy, zfs_secpolicy_destroy, DATASET_NAME, B_TRUE },
3112 	{ zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME, B_TRUE },
3113 	{ zfs_ioc_rename, zfs_secpolicy_rename,	DATASET_NAME, B_TRUE },
3114 	{ zfs_ioc_recv, zfs_secpolicy_receive, DATASET_NAME, B_TRUE },
3115 	{ zfs_ioc_send, zfs_secpolicy_send, DATASET_NAME, B_TRUE },
3116 	{ zfs_ioc_inject_fault,	zfs_secpolicy_inject, NO_NAME, B_FALSE },
3117 	{ zfs_ioc_clear_fault, zfs_secpolicy_inject, NO_NAME, B_FALSE },
3118 	{ zfs_ioc_inject_list_next, zfs_secpolicy_inject, NO_NAME, B_FALSE },
3119 	{ zfs_ioc_error_log, zfs_secpolicy_inject, POOL_NAME, B_FALSE },
3120 	{ zfs_ioc_clear, zfs_secpolicy_config, POOL_NAME, B_TRUE },
3121 	{ zfs_ioc_promote, zfs_secpolicy_promote, DATASET_NAME, B_TRUE },
3122 	{ zfs_ioc_destroy_snaps, zfs_secpolicy_destroy,	DATASET_NAME, B_TRUE },
3123 	{ zfs_ioc_snapshot, zfs_secpolicy_snapshot, DATASET_NAME, B_TRUE },
3124 	{ zfs_ioc_dsobj_to_dsname, zfs_secpolicy_config, POOL_NAME, B_FALSE },
3125 	{ zfs_ioc_obj_to_path, zfs_secpolicy_config, NO_NAME, B_FALSE },
3126 	{ zfs_ioc_pool_set_props, zfs_secpolicy_config,	POOL_NAME, B_TRUE },
3127 	{ zfs_ioc_pool_get_props, zfs_secpolicy_read, POOL_NAME, B_FALSE },
3128 	{ zfs_ioc_set_fsacl, zfs_secpolicy_fsacl, DATASET_NAME, B_TRUE },
3129 	{ zfs_ioc_get_fsacl, zfs_secpolicy_read, DATASET_NAME, B_FALSE },
3130 	{ zfs_ioc_iscsi_perm_check, zfs_secpolicy_iscsi,
3131 	    DATASET_NAME, B_FALSE },
3132 	{ zfs_ioc_share, zfs_secpolicy_share, DATASET_NAME, B_FALSE },
3133 	{ zfs_ioc_inherit_prop, zfs_secpolicy_inherit, DATASET_NAME, B_TRUE },
3134 	{ zfs_ioc_smb_acl, zfs_secpolicy_smb_acl, DATASET_NAME, B_FALSE }
3135 };
3136 
3137 static int
3138 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
3139 {
3140 	zfs_cmd_t *zc;
3141 	uint_t vec;
3142 	int error, rc;
3143 
3144 	if (getminor(dev) != 0)
3145 		return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
3146 
3147 	vec = cmd - ZFS_IOC;
3148 	ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
3149 
3150 	if (vec >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
3151 		return (EINVAL);
3152 
3153 	zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
3154 
3155 	error = xcopyin((void *)arg, zc, sizeof (zfs_cmd_t));
3156 
3157 	if (error == 0)
3158 		error = zfs_ioc_vec[vec].zvec_secpolicy(zc, cr);
3159 
3160 	/*
3161 	 * Ensure that all pool/dataset names are valid before we pass down to
3162 	 * the lower layers.
3163 	 */
3164 	if (error == 0) {
3165 		zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
3166 		switch (zfs_ioc_vec[vec].zvec_namecheck) {
3167 		case POOL_NAME:
3168 			if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
3169 				error = EINVAL;
3170 			break;
3171 
3172 		case DATASET_NAME:
3173 			if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
3174 				error = EINVAL;
3175 			break;
3176 
3177 		case NO_NAME:
3178 			break;
3179 		}
3180 	}
3181 
3182 	if (error == 0)
3183 		error = zfs_ioc_vec[vec].zvec_func(zc);
3184 
3185 	rc = xcopyout(zc, (void *)arg, sizeof (zfs_cmd_t));
3186 	if (error == 0) {
3187 		error = rc;
3188 		if (zfs_ioc_vec[vec].zvec_his_log == B_TRUE)
3189 			zfs_log_history(zc);
3190 	}
3191 
3192 	kmem_free(zc, sizeof (zfs_cmd_t));
3193 	return (error);
3194 }
3195 
3196 static int
3197 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
3198 {
3199 	if (cmd != DDI_ATTACH)
3200 		return (DDI_FAILURE);
3201 
3202 	if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
3203 	    DDI_PSEUDO, 0) == DDI_FAILURE)
3204 		return (DDI_FAILURE);
3205 
3206 	zfs_dip = dip;
3207 
3208 	ddi_report_dev(dip);
3209 
3210 	return (DDI_SUCCESS);
3211 }
3212 
3213 static int
3214 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
3215 {
3216 	if (spa_busy() || zfs_busy() || zvol_busy())
3217 		return (DDI_FAILURE);
3218 
3219 	if (cmd != DDI_DETACH)
3220 		return (DDI_FAILURE);
3221 
3222 	zfs_dip = NULL;
3223 
3224 	ddi_prop_remove_all(dip);
3225 	ddi_remove_minor_node(dip, NULL);
3226 
3227 	return (DDI_SUCCESS);
3228 }
3229 
3230 /*ARGSUSED*/
3231 static int
3232 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
3233 {
3234 	switch (infocmd) {
3235 	case DDI_INFO_DEVT2DEVINFO:
3236 		*result = zfs_dip;
3237 		return (DDI_SUCCESS);
3238 
3239 	case DDI_INFO_DEVT2INSTANCE:
3240 		*result = (void *)0;
3241 		return (DDI_SUCCESS);
3242 	}
3243 
3244 	return (DDI_FAILURE);
3245 }
3246 
3247 /*
3248  * OK, so this is a little weird.
3249  *
3250  * /dev/zfs is the control node, i.e. minor 0.
3251  * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
3252  *
3253  * /dev/zfs has basically nothing to do except serve up ioctls,
3254  * so most of the standard driver entry points are in zvol.c.
3255  */
3256 static struct cb_ops zfs_cb_ops = {
3257 	zvol_open,	/* open */
3258 	zvol_close,	/* close */
3259 	zvol_strategy,	/* strategy */
3260 	nodev,		/* print */
3261 	zvol_dump,	/* dump */
3262 	zvol_read,	/* read */
3263 	zvol_write,	/* write */
3264 	zfsdev_ioctl,	/* ioctl */
3265 	nodev,		/* devmap */
3266 	nodev,		/* mmap */
3267 	nodev,		/* segmap */
3268 	nochpoll,	/* poll */
3269 	ddi_prop_op,	/* prop_op */
3270 	NULL,		/* streamtab */
3271 	D_NEW | D_MP | D_64BIT,		/* Driver compatibility flag */
3272 	CB_REV,		/* version */
3273 	nodev,		/* async read */
3274 	nodev,		/* async write */
3275 };
3276 
3277 static struct dev_ops zfs_dev_ops = {
3278 	DEVO_REV,	/* version */
3279 	0,		/* refcnt */
3280 	zfs_info,	/* info */
3281 	nulldev,	/* identify */
3282 	nulldev,	/* probe */
3283 	zfs_attach,	/* attach */
3284 	zfs_detach,	/* detach */
3285 	nodev,		/* reset */
3286 	&zfs_cb_ops,	/* driver operations */
3287 	NULL,		/* no bus operations */
3288 	NULL,		/* power */
3289 	ddi_quiesce_not_needed,	/* quiesce */
3290 };
3291 
3292 static struct modldrv zfs_modldrv = {
3293 	&mod_driverops,
3294 	"ZFS storage pool",
3295 	&zfs_dev_ops
3296 };
3297 
3298 static struct modlinkage modlinkage = {
3299 	MODREV_1,
3300 	(void *)&zfs_modlfs,
3301 	(void *)&zfs_modldrv,
3302 	NULL
3303 };
3304 
3305 
3306 uint_t zfs_fsyncer_key;
3307 extern uint_t rrw_tsd_key;
3308 
3309 int
3310 _init(void)
3311 {
3312 	int error;
3313 
3314 	spa_init(FREAD | FWRITE);
3315 	zfs_init();
3316 	zvol_init();
3317 
3318 	if ((error = mod_install(&modlinkage)) != 0) {
3319 		zvol_fini();
3320 		zfs_fini();
3321 		spa_fini();
3322 		return (error);
3323 	}
3324 
3325 	tsd_create(&zfs_fsyncer_key, NULL);
3326 	tsd_create(&rrw_tsd_key, NULL);
3327 
3328 	error = ldi_ident_from_mod(&modlinkage, &zfs_li);
3329 	ASSERT(error == 0);
3330 	mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
3331 
3332 	return (0);
3333 }
3334 
3335 int
3336 _fini(void)
3337 {
3338 	int error;
3339 
3340 	if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
3341 		return (EBUSY);
3342 
3343 	if ((error = mod_remove(&modlinkage)) != 0)
3344 		return (error);
3345 
3346 	zvol_fini();
3347 	zfs_fini();
3348 	spa_fini();
3349 	if (zfs_nfsshare_inited)
3350 		(void) ddi_modclose(nfs_mod);
3351 	if (zfs_smbshare_inited)
3352 		(void) ddi_modclose(smbsrv_mod);
3353 	if (zfs_nfsshare_inited || zfs_smbshare_inited)
3354 		(void) ddi_modclose(sharefs_mod);
3355 
3356 	tsd_destroy(&zfs_fsyncer_key);
3357 	ldi_ident_release(zfs_li);
3358 	zfs_li = NULL;
3359 	mutex_destroy(&zfs_share_lock);
3360 
3361 	return (error);
3362 }
3363 
3364 int
3365 _info(struct modinfo *modinfop)
3366 {
3367 	return (mod_info(&modlinkage, modinfop));
3368 }
3369