xref: /illumos-gate/usr/src/uts/common/fs/zfs/zfs_ioctl.c (revision 0dfe541e13279d277d838d6a27e55188b9486cb1)
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 /*
23  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24  */
25 
26 /*
27  * Copyright (c) 2011-2012 Pawel Jakub Dawidek. All rights reserved.
28  * Portions Copyright 2011 Martin Matuska
29  * Copyright 2015, OmniTI Computer Consulting, Inc. All rights reserved.
30  * Copyright 2018 Nexenta Systems, Inc.  All rights reserved.
31  * Copyright 2019 Joyent, Inc.
32  * Copyright (c) 2011, 2017 by Delphix. All rights reserved.
33  * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
34  * Copyright (c) 2013 Steven Hartland. All rights reserved.
35  * Copyright (c) 2014 Integros [integros.com]
36  * Copyright 2016 Toomas Soome <tsoome@me.com>
37  * Copyright (c) 2017, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
38  * Copyright 2017 RackTop Systems.
39  * Copyright (c) 2017, Datto, Inc. All rights reserved.
40  */
41 
42 /*
43  * ZFS ioctls.
44  *
45  * This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
46  * pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
47  *
48  * There are two ways that we handle ioctls: the legacy way where almost
49  * all of the logic is in the ioctl callback, and the new way where most
50  * of the marshalling is handled in the common entry point, zfsdev_ioctl().
51  *
52  * Non-legacy ioctls should be registered by calling
53  * zfs_ioctl_register() from zfs_ioctl_init().  The ioctl is invoked
54  * from userland by lzc_ioctl().
55  *
56  * The registration arguments are as follows:
57  *
58  * const char *name
59  *   The name of the ioctl.  This is used for history logging.  If the
60  *   ioctl returns successfully (the callback returns 0), and allow_log
61  *   is true, then a history log entry will be recorded with the input &
62  *   output nvlists.  The log entry can be printed with "zpool history -i".
63  *
64  * zfs_ioc_t ioc
65  *   The ioctl request number, which userland will pass to ioctl(2).
66  *   The ioctl numbers can change from release to release, because
67  *   the caller (libzfs) must be matched to the kernel.
68  *
69  * zfs_secpolicy_func_t *secpolicy
70  *   This function will be called before the zfs_ioc_func_t, to
71  *   determine if this operation is permitted.  It should return EPERM
72  *   on failure, and 0 on success.  Checks include determining if the
73  *   dataset is visible in this zone, and if the user has either all
74  *   zfs privileges in the zone (SYS_MOUNT), or has been granted permission
75  *   to do this operation on this dataset with "zfs allow".
76  *
77  * zfs_ioc_namecheck_t namecheck
78  *   This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
79  *   name, a dataset name, or nothing.  If the name is not well-formed,
80  *   the ioctl will fail and the callback will not be called.
81  *   Therefore, the callback can assume that the name is well-formed
82  *   (e.g. is null-terminated, doesn't have more than one '@' character,
83  *   doesn't have invalid characters).
84  *
85  * zfs_ioc_poolcheck_t pool_check
86  *   This specifies requirements on the pool state.  If the pool does
87  *   not meet them (is suspended or is readonly), the ioctl will fail
88  *   and the callback will not be called.  If any checks are specified
89  *   (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
90  *   Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED |
91  *   POOL_CHECK_READONLY).
92  *
93  * boolean_t smush_outnvlist
94  *   If smush_outnvlist is true, then the output is presumed to be a
95  *   list of errors, and it will be "smushed" down to fit into the
96  *   caller's buffer, by removing some entries and replacing them with a
97  *   single "N_MORE_ERRORS" entry indicating how many were removed.  See
98  *   nvlist_smush() for details.  If smush_outnvlist is false, and the
99  *   outnvlist does not fit into the userland-provided buffer, then the
100  *   ioctl will fail with ENOMEM.
101  *
102  * zfs_ioc_func_t *func
103  *   The callback function that will perform the operation.
104  *
105  *   The callback should return 0 on success, or an error number on
106  *   failure.  If the function fails, the userland ioctl will return -1,
107  *   and errno will be set to the callback's return value.  The callback
108  *   will be called with the following arguments:
109  *
110  *   const char *name
111  *     The name of the pool or dataset to operate on, from
112  *     zfs_cmd_t:zc_name.  The 'namecheck' argument specifies the
113  *     expected type (pool, dataset, or none).
114  *
115  *   nvlist_t *innvl
116  *     The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src.  Or
117  *     NULL if no input nvlist was provided.  Changes to this nvlist are
118  *     ignored.  If the input nvlist could not be deserialized, the
119  *     ioctl will fail and the callback will not be called.
120  *
121  *   nvlist_t *outnvl
122  *     The output nvlist, initially empty.  The callback can fill it in,
123  *     and it will be returned to userland by serializing it into
124  *     zfs_cmd_t:zc_nvlist_dst.  If it is non-empty, and serialization
125  *     fails (e.g. because the caller didn't supply a large enough
126  *     buffer), then the overall ioctl will fail.  See the
127  *     'smush_nvlist' argument above for additional behaviors.
128  *
129  *     There are two typical uses of the output nvlist:
130  *       - To return state, e.g. property values.  In this case,
131  *         smush_outnvlist should be false.  If the buffer was not large
132  *         enough, the caller will reallocate a larger buffer and try
133  *         the ioctl again.
134  *
135  *       - To return multiple errors from an ioctl which makes on-disk
136  *         changes.  In this case, smush_outnvlist should be true.
137  *         Ioctls which make on-disk modifications should generally not
138  *         use the outnvl if they succeed, because the caller can not
139  *         distinguish between the operation failing, and
140  *         deserialization failing.
141  */
142 
143 #include <sys/types.h>
144 #include <sys/param.h>
145 #include <sys/errno.h>
146 #include <sys/uio.h>
147 #include <sys/buf.h>
148 #include <sys/modctl.h>
149 #include <sys/open.h>
150 #include <sys/file.h>
151 #include <sys/kmem.h>
152 #include <sys/conf.h>
153 #include <sys/cmn_err.h>
154 #include <sys/stat.h>
155 #include <sys/zfs_ioctl.h>
156 #include <sys/zfs_vfsops.h>
157 #include <sys/zfs_znode.h>
158 #include <sys/zap.h>
159 #include <sys/spa.h>
160 #include <sys/spa_impl.h>
161 #include <sys/vdev.h>
162 #include <sys/priv_impl.h>
163 #include <sys/dmu.h>
164 #include <sys/dsl_dir.h>
165 #include <sys/dsl_dataset.h>
166 #include <sys/dsl_prop.h>
167 #include <sys/dsl_deleg.h>
168 #include <sys/dmu_objset.h>
169 #include <sys/dmu_impl.h>
170 #include <sys/dmu_tx.h>
171 #include <sys/ddi.h>
172 #include <sys/sunddi.h>
173 #include <sys/sunldi.h>
174 #include <sys/policy.h>
175 #include <sys/zone.h>
176 #include <sys/nvpair.h>
177 #include <sys/pathname.h>
178 #include <sys/mount.h>
179 #include <sys/sdt.h>
180 #include <sys/fs/zfs.h>
181 #include <sys/zfs_ctldir.h>
182 #include <sys/zfs_dir.h>
183 #include <sys/zfs_onexit.h>
184 #include <sys/zvol.h>
185 #include <sys/dsl_scan.h>
186 #include <sharefs/share.h>
187 #include <sys/dmu_objset.h>
188 #include <sys/dmu_recv.h>
189 #include <sys/dmu_send.h>
190 #include <sys/dsl_destroy.h>
191 #include <sys/dsl_bookmark.h>
192 #include <sys/dsl_userhold.h>
193 #include <sys/zfeature.h>
194 #include <sys/zcp.h>
195 #include <sys/zio_checksum.h>
196 #include <sys/vdev_removal.h>
197 #include <sys/vdev_impl.h>
198 #include <sys/vdev_initialize.h>
199 #include <sys/vdev_trim.h>
200 #include <sys/dsl_crypt.h>
201 
202 #include "zfs_namecheck.h"
203 #include "zfs_prop.h"
204 #include "zfs_deleg.h"
205 #include "zfs_comutil.h"
206 
207 #include "lua.h"
208 #include "lauxlib.h"
209 
210 extern struct modlfs zfs_modlfs;
211 
212 extern void zfs_init(void);
213 extern void zfs_fini(void);
214 
215 ldi_ident_t zfs_li = NULL;
216 dev_info_t *zfs_dip;
217 
218 uint_t zfs_fsyncer_key;
219 extern uint_t rrw_tsd_key;
220 static uint_t zfs_allow_log_key;
221 
222 typedef int zfs_ioc_legacy_func_t(zfs_cmd_t *);
223 typedef int zfs_ioc_func_t(const char *, nvlist_t *, nvlist_t *);
224 typedef int zfs_secpolicy_func_t(zfs_cmd_t *, nvlist_t *, cred_t *);
225 
226 typedef enum {
227 	NO_NAME,
228 	POOL_NAME,
229 	DATASET_NAME
230 } zfs_ioc_namecheck_t;
231 
232 typedef enum {
233 	POOL_CHECK_NONE		= 1 << 0,
234 	POOL_CHECK_SUSPENDED	= 1 << 1,
235 	POOL_CHECK_READONLY	= 1 << 2,
236 } zfs_ioc_poolcheck_t;
237 
238 typedef struct zfs_ioc_vec {
239 	zfs_ioc_legacy_func_t	*zvec_legacy_func;
240 	zfs_ioc_func_t		*zvec_func;
241 	zfs_secpolicy_func_t	*zvec_secpolicy;
242 	zfs_ioc_namecheck_t	zvec_namecheck;
243 	boolean_t		zvec_allow_log;
244 	zfs_ioc_poolcheck_t	zvec_pool_check;
245 	boolean_t		zvec_smush_outnvlist;
246 	const char		*zvec_name;
247 } zfs_ioc_vec_t;
248 
249 /* This array is indexed by zfs_userquota_prop_t */
250 static const char *userquota_perms[] = {
251 	ZFS_DELEG_PERM_USERUSED,
252 	ZFS_DELEG_PERM_USERQUOTA,
253 	ZFS_DELEG_PERM_GROUPUSED,
254 	ZFS_DELEG_PERM_GROUPQUOTA,
255 	ZFS_DELEG_PERM_USEROBJUSED,
256 	ZFS_DELEG_PERM_USEROBJQUOTA,
257 	ZFS_DELEG_PERM_GROUPOBJUSED,
258 	ZFS_DELEG_PERM_GROUPOBJQUOTA,
259 	ZFS_DELEG_PERM_PROJECTUSED,
260 	ZFS_DELEG_PERM_PROJECTQUOTA,
261 	ZFS_DELEG_PERM_PROJECTOBJUSED,
262 	ZFS_DELEG_PERM_PROJECTOBJQUOTA,
263 };
264 
265 static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
266 static int zfs_ioc_id_quota_upgrade(zfs_cmd_t *zc);
267 static int zfs_check_settable(const char *name, nvpair_t *property,
268     cred_t *cr);
269 static int zfs_check_clearable(char *dataset, nvlist_t *props,
270     nvlist_t **errors);
271 static int zfs_fill_zplprops_root(uint64_t, nvlist_t *, nvlist_t *,
272     boolean_t *);
273 int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *);
274 static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);
275 
276 static int zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature);
277 
278 /* _NOTE(PRINTFLIKE(4)) - this is printf-like, but lint is too whiney */
279 void
280 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
281 {
282 	const char *newfile;
283 	char buf[512];
284 	va_list adx;
285 
286 	/*
287 	 * Get rid of annoying "../common/" prefix to filename.
288 	 */
289 	newfile = strrchr(file, '/');
290 	if (newfile != NULL) {
291 		newfile = newfile + 1; /* Get rid of leading / */
292 	} else {
293 		newfile = file;
294 	}
295 
296 	va_start(adx, fmt);
297 	(void) vsnprintf(buf, sizeof (buf), fmt, adx);
298 	va_end(adx);
299 
300 	/*
301 	 * To get this data, use the zfs-dprintf probe as so:
302 	 * dtrace -q -n 'zfs-dprintf \
303 	 *	/stringof(arg0) == "dbuf.c"/ \
304 	 *	{printf("%s: %s", stringof(arg1), stringof(arg3))}'
305 	 * arg0 = file name
306 	 * arg1 = function name
307 	 * arg2 = line number
308 	 * arg3 = message
309 	 */
310 	DTRACE_PROBE4(zfs__dprintf,
311 	    char *, newfile, char *, func, int, line, char *, buf);
312 }
313 
314 static void
315 history_str_free(char *buf)
316 {
317 	kmem_free(buf, HIS_MAX_RECORD_LEN);
318 }
319 
320 static char *
321 history_str_get(zfs_cmd_t *zc)
322 {
323 	char *buf;
324 
325 	if (zc->zc_history == 0)
326 		return (NULL);
327 
328 	buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
329 	if (copyinstr((void *)(uintptr_t)zc->zc_history,
330 	    buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
331 		history_str_free(buf);
332 		return (NULL);
333 	}
334 
335 	buf[HIS_MAX_RECORD_LEN -1] = '\0';
336 
337 	return (buf);
338 }
339 
340 /*
341  * Check to see if the named dataset is currently defined as bootable
342  */
343 static boolean_t
344 zfs_is_bootfs(const char *name)
345 {
346 	objset_t *os;
347 
348 	if (dmu_objset_hold(name, FTAG, &os) == 0) {
349 		boolean_t ret;
350 		ret = (dmu_objset_id(os) == spa_bootfs(dmu_objset_spa(os)));
351 		dmu_objset_rele(os, FTAG);
352 		return (ret);
353 	}
354 	return (B_FALSE);
355 }
356 
357 /*
358  * Return non-zero if the spa version is less than requested version.
359  */
360 static int
361 zfs_earlier_version(const char *name, int version)
362 {
363 	spa_t *spa;
364 
365 	if (spa_open(name, &spa, FTAG) == 0) {
366 		if (spa_version(spa) < version) {
367 			spa_close(spa, FTAG);
368 			return (1);
369 		}
370 		spa_close(spa, FTAG);
371 	}
372 	return (0);
373 }
374 
375 /*
376  * Return TRUE if the ZPL version is less than requested version.
377  */
378 static boolean_t
379 zpl_earlier_version(const char *name, int version)
380 {
381 	objset_t *os;
382 	boolean_t rc = B_TRUE;
383 
384 	if (dmu_objset_hold(name, FTAG, &os) == 0) {
385 		uint64_t zplversion;
386 
387 		if (dmu_objset_type(os) != DMU_OST_ZFS) {
388 			dmu_objset_rele(os, FTAG);
389 			return (B_TRUE);
390 		}
391 		/* XXX reading from non-owned objset */
392 		if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
393 			rc = zplversion < version;
394 		dmu_objset_rele(os, FTAG);
395 	}
396 	return (rc);
397 }
398 
399 static void
400 zfs_log_history(zfs_cmd_t *zc)
401 {
402 	spa_t *spa;
403 	char *buf;
404 
405 	if ((buf = history_str_get(zc)) == NULL)
406 		return;
407 
408 	if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
409 		if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
410 			(void) spa_history_log(spa, buf);
411 		spa_close(spa, FTAG);
412 	}
413 	history_str_free(buf);
414 }
415 
416 /*
417  * Policy for top-level read operations (list pools).  Requires no privileges,
418  * and can be used in the local zone, as there is no associated dataset.
419  */
420 /* ARGSUSED */
421 static int
422 zfs_secpolicy_none(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
423 {
424 	return (0);
425 }
426 
427 /*
428  * Policy for dataset read operations (list children, get statistics).  Requires
429  * no privileges, but must be visible in the local zone.
430  */
431 /* ARGSUSED */
432 static int
433 zfs_secpolicy_read(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
434 {
435 	if (INGLOBALZONE(curproc) ||
436 	    zone_dataset_visible(zc->zc_name, NULL))
437 		return (0);
438 
439 	return (SET_ERROR(ENOENT));
440 }
441 
442 static int
443 zfs_dozonecheck_impl(const char *dataset, uint64_t zoned, cred_t *cr)
444 {
445 	int writable = 1;
446 
447 	/*
448 	 * The dataset must be visible by this zone -- check this first
449 	 * so they don't see EPERM on something they shouldn't know about.
450 	 */
451 	if (!INGLOBALZONE(curproc) &&
452 	    !zone_dataset_visible(dataset, &writable))
453 		return (SET_ERROR(ENOENT));
454 
455 	if (INGLOBALZONE(curproc)) {
456 		/*
457 		 * If the fs is zoned, only root can access it from the
458 		 * global zone.
459 		 */
460 		if (secpolicy_zfs(cr) && zoned)
461 			return (SET_ERROR(EPERM));
462 	} else {
463 		/*
464 		 * If we are in a local zone, the 'zoned' property must be set.
465 		 */
466 		if (!zoned)
467 			return (SET_ERROR(EPERM));
468 
469 		/* must be writable by this zone */
470 		if (!writable)
471 			return (SET_ERROR(EPERM));
472 	}
473 	return (0);
474 }
475 
476 static int
477 zfs_dozonecheck(const char *dataset, cred_t *cr)
478 {
479 	uint64_t zoned;
480 
481 	if (dsl_prop_get_integer(dataset, "zoned", &zoned, NULL))
482 		return (SET_ERROR(ENOENT));
483 
484 	return (zfs_dozonecheck_impl(dataset, zoned, cr));
485 }
486 
487 static int
488 zfs_dozonecheck_ds(const char *dataset, dsl_dataset_t *ds, cred_t *cr)
489 {
490 	uint64_t zoned;
491 
492 	if (dsl_prop_get_int_ds(ds, "zoned", &zoned))
493 		return (SET_ERROR(ENOENT));
494 
495 	return (zfs_dozonecheck_impl(dataset, zoned, cr));
496 }
497 
498 static int
499 zfs_secpolicy_write_perms_ds(const char *name, dsl_dataset_t *ds,
500     const char *perm, cred_t *cr)
501 {
502 	int error;
503 
504 	error = zfs_dozonecheck_ds(name, ds, cr);
505 	if (error == 0) {
506 		error = secpolicy_zfs(cr);
507 		if (error != 0)
508 			error = dsl_deleg_access_impl(ds, perm, cr);
509 	}
510 	return (error);
511 }
512 
513 static int
514 zfs_secpolicy_write_perms(const char *name, const char *perm, cred_t *cr)
515 {
516 	int error;
517 	dsl_dataset_t *ds;
518 	dsl_pool_t *dp;
519 
520 	/*
521 	 * First do a quick check for root in the global zone, which
522 	 * is allowed to do all write_perms.  This ensures that zfs_ioc_*
523 	 * will get to handle nonexistent datasets.
524 	 */
525 	if (INGLOBALZONE(curproc) && secpolicy_zfs(cr) == 0)
526 		return (0);
527 
528 	error = dsl_pool_hold(name, FTAG, &dp);
529 	if (error != 0)
530 		return (error);
531 
532 	error = dsl_dataset_hold(dp, name, FTAG, &ds);
533 	if (error != 0) {
534 		dsl_pool_rele(dp, FTAG);
535 		return (error);
536 	}
537 
538 	error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);
539 
540 	dsl_dataset_rele(ds, FTAG);
541 	dsl_pool_rele(dp, FTAG);
542 	return (error);
543 }
544 
545 /*
546  * Policy for setting the security label property.
547  *
548  * Returns 0 for success, non-zero for access and other errors.
549  */
550 static int
551 zfs_set_slabel_policy(const char *name, char *strval, cred_t *cr)
552 {
553 	char		ds_hexsl[MAXNAMELEN];
554 	bslabel_t	ds_sl, new_sl;
555 	boolean_t	new_default = FALSE;
556 	uint64_t	zoned;
557 	int		needed_priv = -1;
558 	int		error;
559 
560 	/* First get the existing dataset label. */
561 	error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
562 	    1, sizeof (ds_hexsl), &ds_hexsl, NULL);
563 	if (error != 0)
564 		return (SET_ERROR(EPERM));
565 
566 	if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
567 		new_default = TRUE;
568 
569 	/* The label must be translatable */
570 	if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
571 		return (SET_ERROR(EINVAL));
572 
573 	/*
574 	 * In a non-global zone, disallow attempts to set a label that
575 	 * doesn't match that of the zone; otherwise no other checks
576 	 * are needed.
577 	 */
578 	if (!INGLOBALZONE(curproc)) {
579 		if (new_default || !blequal(&new_sl, CR_SL(CRED())))
580 			return (SET_ERROR(EPERM));
581 		return (0);
582 	}
583 
584 	/*
585 	 * For global-zone datasets (i.e., those whose zoned property is
586 	 * "off", verify that the specified new label is valid for the
587 	 * global zone.
588 	 */
589 	if (dsl_prop_get_integer(name,
590 	    zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
591 		return (SET_ERROR(EPERM));
592 	if (!zoned) {
593 		if (zfs_check_global_label(name, strval) != 0)
594 			return (SET_ERROR(EPERM));
595 	}
596 
597 	/*
598 	 * If the existing dataset label is nondefault, check if the
599 	 * dataset is mounted (label cannot be changed while mounted).
600 	 * Get the zfsvfs; if there isn't one, then the dataset isn't
601 	 * mounted (or isn't a dataset, doesn't exist, ...).
602 	 */
603 	if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
604 		objset_t *os;
605 		static char *setsl_tag = "setsl_tag";
606 
607 		/*
608 		 * Try to own the dataset; abort if there is any error,
609 		 * (e.g., already mounted, in use, or other error).
610 		 */
611 		error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE, B_TRUE,
612 		    setsl_tag, &os);
613 		if (error != 0)
614 			return (SET_ERROR(EPERM));
615 
616 		dmu_objset_disown(os, B_TRUE, setsl_tag);
617 
618 		if (new_default) {
619 			needed_priv = PRIV_FILE_DOWNGRADE_SL;
620 			goto out_check;
621 		}
622 
623 		if (hexstr_to_label(strval, &new_sl) != 0)
624 			return (SET_ERROR(EPERM));
625 
626 		if (blstrictdom(&ds_sl, &new_sl))
627 			needed_priv = PRIV_FILE_DOWNGRADE_SL;
628 		else if (blstrictdom(&new_sl, &ds_sl))
629 			needed_priv = PRIV_FILE_UPGRADE_SL;
630 	} else {
631 		/* dataset currently has a default label */
632 		if (!new_default)
633 			needed_priv = PRIV_FILE_UPGRADE_SL;
634 	}
635 
636 out_check:
637 	if (needed_priv != -1)
638 		return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
639 	return (0);
640 }
641 
642 static int
643 zfs_secpolicy_setprop(const char *dsname, zfs_prop_t prop, nvpair_t *propval,
644     cred_t *cr)
645 {
646 	char *strval;
647 
648 	/*
649 	 * Check permissions for special properties.
650 	 */
651 	switch (prop) {
652 	case ZFS_PROP_ZONED:
653 		/*
654 		 * Disallow setting of 'zoned' from within a local zone.
655 		 */
656 		if (!INGLOBALZONE(curproc))
657 			return (SET_ERROR(EPERM));
658 		break;
659 
660 	case ZFS_PROP_QUOTA:
661 	case ZFS_PROP_FILESYSTEM_LIMIT:
662 	case ZFS_PROP_SNAPSHOT_LIMIT:
663 		if (!INGLOBALZONE(curproc)) {
664 			uint64_t zoned;
665 			char setpoint[ZFS_MAX_DATASET_NAME_LEN];
666 			/*
667 			 * Unprivileged users are allowed to modify the
668 			 * limit on things *under* (ie. contained by)
669 			 * the thing they own.
670 			 */
671 			if (dsl_prop_get_integer(dsname, "zoned", &zoned,
672 			    setpoint))
673 				return (SET_ERROR(EPERM));
674 			if (!zoned || strlen(dsname) <= strlen(setpoint))
675 				return (SET_ERROR(EPERM));
676 		}
677 		break;
678 
679 	case ZFS_PROP_MLSLABEL:
680 		if (!is_system_labeled())
681 			return (SET_ERROR(EPERM));
682 
683 		if (nvpair_value_string(propval, &strval) == 0) {
684 			int err;
685 
686 			err = zfs_set_slabel_policy(dsname, strval, CRED());
687 			if (err != 0)
688 				return (err);
689 		}
690 		break;
691 	}
692 
693 	return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
694 }
695 
696 /* ARGSUSED */
697 static int
698 zfs_secpolicy_set_fsacl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
699 {
700 	int error;
701 
702 	error = zfs_dozonecheck(zc->zc_name, cr);
703 	if (error != 0)
704 		return (error);
705 
706 	/*
707 	 * permission to set permissions will be evaluated later in
708 	 * dsl_deleg_can_allow()
709 	 */
710 	return (0);
711 }
712 
713 /* ARGSUSED */
714 static int
715 zfs_secpolicy_rollback(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
716 {
717 	return (zfs_secpolicy_write_perms(zc->zc_name,
718 	    ZFS_DELEG_PERM_ROLLBACK, cr));
719 }
720 
721 /* ARGSUSED */
722 static int
723 zfs_secpolicy_send(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
724 {
725 	dsl_pool_t *dp;
726 	dsl_dataset_t *ds;
727 	char *cp;
728 	int error;
729 
730 	/*
731 	 * Generate the current snapshot name from the given objsetid, then
732 	 * use that name for the secpolicy/zone checks.
733 	 */
734 	cp = strchr(zc->zc_name, '@');
735 	if (cp == NULL)
736 		return (SET_ERROR(EINVAL));
737 	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
738 	if (error != 0)
739 		return (error);
740 
741 	error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
742 	if (error != 0) {
743 		dsl_pool_rele(dp, FTAG);
744 		return (error);
745 	}
746 
747 	dsl_dataset_name(ds, zc->zc_name);
748 
749 	error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
750 	    ZFS_DELEG_PERM_SEND, cr);
751 	dsl_dataset_rele(ds, FTAG);
752 	dsl_pool_rele(dp, FTAG);
753 
754 	return (error);
755 }
756 
757 /* ARGSUSED */
758 static int
759 zfs_secpolicy_send_new(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
760 {
761 	return (zfs_secpolicy_write_perms(zc->zc_name,
762 	    ZFS_DELEG_PERM_SEND, cr));
763 }
764 
765 /* ARGSUSED */
766 static int
767 zfs_secpolicy_deleg_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
768 {
769 	vnode_t *vp;
770 	int error;
771 
772 	if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
773 	    NO_FOLLOW, NULL, &vp)) != 0)
774 		return (error);
775 
776 	/* Now make sure mntpnt and dataset are ZFS */
777 
778 	if (vp->v_vfsp->vfs_fstype != zfsfstype ||
779 	    (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
780 	    zc->zc_name) != 0)) {
781 		VN_RELE(vp);
782 		return (SET_ERROR(EPERM));
783 	}
784 
785 	VN_RELE(vp);
786 	return (dsl_deleg_access(zc->zc_name,
787 	    ZFS_DELEG_PERM_SHARE, cr));
788 }
789 
790 int
791 zfs_secpolicy_share(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
792 {
793 	if (secpolicy_nfs(cr) == 0) {
794 		return (0);
795 	} else {
796 		return (zfs_secpolicy_deleg_share(zc, innvl, cr));
797 	}
798 }
799 
800 int
801 zfs_secpolicy_smb_acl(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
802 {
803 	if (secpolicy_smb(cr) == 0) {
804 		return (0);
805 	} else {
806 		return (zfs_secpolicy_deleg_share(zc, innvl, cr));
807 	}
808 }
809 
810 static int
811 zfs_get_parent(const char *datasetname, char *parent, int parentsize)
812 {
813 	char *cp;
814 
815 	/*
816 	 * Remove the @bla or /bla from the end of the name to get the parent.
817 	 */
818 	(void) strncpy(parent, datasetname, parentsize);
819 	cp = strrchr(parent, '@');
820 	if (cp != NULL) {
821 		cp[0] = '\0';
822 	} else {
823 		cp = strrchr(parent, '/');
824 		if (cp == NULL)
825 			return (SET_ERROR(ENOENT));
826 		cp[0] = '\0';
827 	}
828 
829 	return (0);
830 }
831 
832 int
833 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
834 {
835 	int error;
836 
837 	if ((error = zfs_secpolicy_write_perms(name,
838 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
839 		return (error);
840 
841 	return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
842 }
843 
844 /* ARGSUSED */
845 static int
846 zfs_secpolicy_destroy(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
847 {
848 	return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
849 }
850 
851 /*
852  * Destroying snapshots with delegated permissions requires
853  * descendant mount and destroy permissions.
854  */
855 /* ARGSUSED */
856 static int
857 zfs_secpolicy_destroy_snaps(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
858 {
859 	nvlist_t *snaps;
860 	nvpair_t *pair, *nextpair;
861 	int error = 0;
862 
863 	if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
864 		return (SET_ERROR(EINVAL));
865 	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
866 	    pair = nextpair) {
867 		nextpair = nvlist_next_nvpair(snaps, pair);
868 		error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
869 		if (error == ENOENT) {
870 			/*
871 			 * Ignore any snapshots that don't exist (we consider
872 			 * them "already destroyed").  Remove the name from the
873 			 * nvl here in case the snapshot is created between
874 			 * now and when we try to destroy it (in which case
875 			 * we don't want to destroy it since we haven't
876 			 * checked for permission).
877 			 */
878 			fnvlist_remove_nvpair(snaps, pair);
879 			error = 0;
880 		}
881 		if (error != 0)
882 			break;
883 	}
884 
885 	return (error);
886 }
887 
888 int
889 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
890 {
891 	char	parentname[ZFS_MAX_DATASET_NAME_LEN];
892 	int	error;
893 
894 	if ((error = zfs_secpolicy_write_perms(from,
895 	    ZFS_DELEG_PERM_RENAME, cr)) != 0)
896 		return (error);
897 
898 	if ((error = zfs_secpolicy_write_perms(from,
899 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
900 		return (error);
901 
902 	if ((error = zfs_get_parent(to, parentname,
903 	    sizeof (parentname))) != 0)
904 		return (error);
905 
906 	if ((error = zfs_secpolicy_write_perms(parentname,
907 	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
908 		return (error);
909 
910 	if ((error = zfs_secpolicy_write_perms(parentname,
911 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
912 		return (error);
913 
914 	return (error);
915 }
916 
917 /* ARGSUSED */
918 static int
919 zfs_secpolicy_rename(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
920 {
921 	return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
922 }
923 
924 /* ARGSUSED */
925 static int
926 zfs_secpolicy_promote(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
927 {
928 	dsl_pool_t *dp;
929 	dsl_dataset_t *clone;
930 	int error;
931 
932 	error = zfs_secpolicy_write_perms(zc->zc_name,
933 	    ZFS_DELEG_PERM_PROMOTE, cr);
934 	if (error != 0)
935 		return (error);
936 
937 	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
938 	if (error != 0)
939 		return (error);
940 
941 	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);
942 
943 	if (error == 0) {
944 		char parentname[ZFS_MAX_DATASET_NAME_LEN];
945 		dsl_dataset_t *origin = NULL;
946 		dsl_dir_t *dd;
947 		dd = clone->ds_dir;
948 
949 		error = dsl_dataset_hold_obj(dd->dd_pool,
950 		    dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin);
951 		if (error != 0) {
952 			dsl_dataset_rele(clone, FTAG);
953 			dsl_pool_rele(dp, FTAG);
954 			return (error);
955 		}
956 
957 		error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
958 		    ZFS_DELEG_PERM_MOUNT, cr);
959 
960 		dsl_dataset_name(origin, parentname);
961 		if (error == 0) {
962 			error = zfs_secpolicy_write_perms_ds(parentname, origin,
963 			    ZFS_DELEG_PERM_PROMOTE, cr);
964 		}
965 		dsl_dataset_rele(clone, FTAG);
966 		dsl_dataset_rele(origin, FTAG);
967 	}
968 	dsl_pool_rele(dp, FTAG);
969 	return (error);
970 }
971 
972 /* ARGSUSED */
973 static int
974 zfs_secpolicy_recv(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
975 {
976 	int error;
977 
978 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
979 	    ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
980 		return (error);
981 
982 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
983 	    ZFS_DELEG_PERM_MOUNT, cr)) != 0)
984 		return (error);
985 
986 	return (zfs_secpolicy_write_perms(zc->zc_name,
987 	    ZFS_DELEG_PERM_CREATE, cr));
988 }
989 
990 int
991 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
992 {
993 	return (zfs_secpolicy_write_perms(name,
994 	    ZFS_DELEG_PERM_SNAPSHOT, cr));
995 }
996 
997 /*
998  * Check for permission to create each snapshot in the nvlist.
999  */
1000 /* ARGSUSED */
1001 static int
1002 zfs_secpolicy_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1003 {
1004 	nvlist_t *snaps;
1005 	int error = 0;
1006 	nvpair_t *pair;
1007 
1008 	if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
1009 		return (SET_ERROR(EINVAL));
1010 	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
1011 	    pair = nvlist_next_nvpair(snaps, pair)) {
1012 		char *name = nvpair_name(pair);
1013 		char *atp = strchr(name, '@');
1014 
1015 		if (atp == NULL) {
1016 			error = SET_ERROR(EINVAL);
1017 			break;
1018 		}
1019 		*atp = '\0';
1020 		error = zfs_secpolicy_snapshot_perms(name, cr);
1021 		*atp = '@';
1022 		if (error != 0)
1023 			break;
1024 	}
1025 	return (error);
1026 }
1027 
1028 /*
1029  * Check for permission to create each snapshot in the nvlist.
1030  */
1031 /* ARGSUSED */
1032 static int
1033 zfs_secpolicy_bookmark(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1034 {
1035 	int error = 0;
1036 
1037 	for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
1038 	    pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
1039 		char *name = nvpair_name(pair);
1040 		char *hashp = strchr(name, '#');
1041 
1042 		if (hashp == NULL) {
1043 			error = SET_ERROR(EINVAL);
1044 			break;
1045 		}
1046 		*hashp = '\0';
1047 		error = zfs_secpolicy_write_perms(name,
1048 		    ZFS_DELEG_PERM_BOOKMARK, cr);
1049 		*hashp = '#';
1050 		if (error != 0)
1051 			break;
1052 	}
1053 	return (error);
1054 }
1055 
1056 /* ARGSUSED */
1057 static int
1058 zfs_secpolicy_remap(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1059 {
1060 	return (zfs_secpolicy_write_perms(zc->zc_name,
1061 	    ZFS_DELEG_PERM_REMAP, cr));
1062 }
1063 
1064 /* ARGSUSED */
1065 static int
1066 zfs_secpolicy_destroy_bookmarks(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1067 {
1068 	nvpair_t *pair, *nextpair;
1069 	int error = 0;
1070 
1071 	for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1072 	    pair = nextpair) {
1073 		char *name = nvpair_name(pair);
1074 		char *hashp = strchr(name, '#');
1075 		nextpair = nvlist_next_nvpair(innvl, pair);
1076 
1077 		if (hashp == NULL) {
1078 			error = SET_ERROR(EINVAL);
1079 			break;
1080 		}
1081 
1082 		*hashp = '\0';
1083 		error = zfs_secpolicy_write_perms(name,
1084 		    ZFS_DELEG_PERM_DESTROY, cr);
1085 		*hashp = '#';
1086 		if (error == ENOENT) {
1087 			/*
1088 			 * Ignore any filesystems that don't exist (we consider
1089 			 * their bookmarks "already destroyed").  Remove
1090 			 * the name from the nvl here in case the filesystem
1091 			 * is created between now and when we try to destroy
1092 			 * the bookmark (in which case we don't want to
1093 			 * destroy it since we haven't checked for permission).
1094 			 */
1095 			fnvlist_remove_nvpair(innvl, pair);
1096 			error = 0;
1097 		}
1098 		if (error != 0)
1099 			break;
1100 	}
1101 
1102 	return (error);
1103 }
1104 
1105 /* ARGSUSED */
1106 static int
1107 zfs_secpolicy_log_history(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1108 {
1109 	/*
1110 	 * Even root must have a proper TSD so that we know what pool
1111 	 * to log to.
1112 	 */
1113 	if (tsd_get(zfs_allow_log_key) == NULL)
1114 		return (SET_ERROR(EPERM));
1115 	return (0);
1116 }
1117 
1118 static int
1119 zfs_secpolicy_create_clone(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1120 {
1121 	char	parentname[ZFS_MAX_DATASET_NAME_LEN];
1122 	int	error;
1123 	char	*origin;
1124 
1125 	if ((error = zfs_get_parent(zc->zc_name, parentname,
1126 	    sizeof (parentname))) != 0)
1127 		return (error);
1128 
1129 	if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
1130 	    (error = zfs_secpolicy_write_perms(origin,
1131 	    ZFS_DELEG_PERM_CLONE, cr)) != 0)
1132 		return (error);
1133 
1134 	if ((error = zfs_secpolicy_write_perms(parentname,
1135 	    ZFS_DELEG_PERM_CREATE, cr)) != 0)
1136 		return (error);
1137 
1138 	return (zfs_secpolicy_write_perms(parentname,
1139 	    ZFS_DELEG_PERM_MOUNT, cr));
1140 }
1141 
1142 /*
1143  * Policy for pool operations - create/destroy pools, add vdevs, etc.  Requires
1144  * SYS_CONFIG privilege, which is not available in a local zone.
1145  */
1146 /* ARGSUSED */
1147 static int
1148 zfs_secpolicy_config(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1149 {
1150 	if (secpolicy_sys_config(cr, B_FALSE) != 0)
1151 		return (SET_ERROR(EPERM));
1152 
1153 	return (0);
1154 }
1155 
1156 /*
1157  * Policy for object to name lookups.
1158  */
1159 /* ARGSUSED */
1160 static int
1161 zfs_secpolicy_diff(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1162 {
1163 	int error;
1164 
1165 	if ((error = secpolicy_sys_config(cr, B_FALSE)) == 0)
1166 		return (0);
1167 
1168 	error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
1169 	return (error);
1170 }
1171 
1172 /*
1173  * Policy for fault injection.  Requires all privileges.
1174  */
1175 /* ARGSUSED */
1176 static int
1177 zfs_secpolicy_inject(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1178 {
1179 	return (secpolicy_zinject(cr));
1180 }
1181 
1182 /* ARGSUSED */
1183 static int
1184 zfs_secpolicy_inherit_prop(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1185 {
1186 	zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);
1187 
1188 	if (prop == ZPROP_INVAL) {
1189 		if (!zfs_prop_user(zc->zc_value))
1190 			return (SET_ERROR(EINVAL));
1191 		return (zfs_secpolicy_write_perms(zc->zc_name,
1192 		    ZFS_DELEG_PERM_USERPROP, cr));
1193 	} else {
1194 		return (zfs_secpolicy_setprop(zc->zc_name, prop,
1195 		    NULL, cr));
1196 	}
1197 }
1198 
1199 static int
1200 zfs_secpolicy_userspace_one(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1201 {
1202 	int err = zfs_secpolicy_read(zc, innvl, cr);
1203 	if (err)
1204 		return (err);
1205 
1206 	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1207 		return (SET_ERROR(EINVAL));
1208 
1209 	if (zc->zc_value[0] == 0) {
1210 		/*
1211 		 * They are asking about a posix uid/gid.  If it's
1212 		 * themself, allow it.
1213 		 */
1214 		if (zc->zc_objset_type == ZFS_PROP_USERUSED ||
1215 		    zc->zc_objset_type == ZFS_PROP_USERQUOTA ||
1216 		    zc->zc_objset_type == ZFS_PROP_USEROBJUSED ||
1217 		    zc->zc_objset_type == ZFS_PROP_USEROBJQUOTA) {
1218 			if (zc->zc_guid == crgetuid(cr))
1219 				return (0);
1220 		} else if (zc->zc_objset_type == ZFS_PROP_GROUPUSED ||
1221 		    zc->zc_objset_type == ZFS_PROP_GROUPQUOTA ||
1222 		    zc->zc_objset_type == ZFS_PROP_GROUPOBJUSED ||
1223 		    zc->zc_objset_type == ZFS_PROP_GROUPOBJQUOTA) {
1224 			if (groupmember(zc->zc_guid, cr))
1225 				return (0);
1226 		}
1227 		/* else is for project quota/used */
1228 	}
1229 
1230 	return (zfs_secpolicy_write_perms(zc->zc_name,
1231 	    userquota_perms[zc->zc_objset_type], cr));
1232 }
1233 
1234 static int
1235 zfs_secpolicy_userspace_many(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1236 {
1237 	int err = zfs_secpolicy_read(zc, innvl, cr);
1238 	if (err)
1239 		return (err);
1240 
1241 	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
1242 		return (SET_ERROR(EINVAL));
1243 
1244 	return (zfs_secpolicy_write_perms(zc->zc_name,
1245 	    userquota_perms[zc->zc_objset_type], cr));
1246 }
1247 
1248 /* ARGSUSED */
1249 static int
1250 zfs_secpolicy_userspace_upgrade(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1251 {
1252 	return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
1253 	    NULL, cr));
1254 }
1255 
1256 /* ARGSUSED */
1257 static int
1258 zfs_secpolicy_hold(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1259 {
1260 	nvpair_t *pair;
1261 	nvlist_t *holds;
1262 	int error;
1263 
1264 	error = nvlist_lookup_nvlist(innvl, "holds", &holds);
1265 	if (error != 0)
1266 		return (SET_ERROR(EINVAL));
1267 
1268 	for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
1269 	    pair = nvlist_next_nvpair(holds, pair)) {
1270 		char fsname[ZFS_MAX_DATASET_NAME_LEN];
1271 		error = dmu_fsname(nvpair_name(pair), fsname);
1272 		if (error != 0)
1273 			return (error);
1274 		error = zfs_secpolicy_write_perms(fsname,
1275 		    ZFS_DELEG_PERM_HOLD, cr);
1276 		if (error != 0)
1277 			return (error);
1278 	}
1279 	return (0);
1280 }
1281 
1282 /* ARGSUSED */
1283 static int
1284 zfs_secpolicy_release(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1285 {
1286 	nvpair_t *pair;
1287 	int error;
1288 
1289 	for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
1290 	    pair = nvlist_next_nvpair(innvl, pair)) {
1291 		char fsname[ZFS_MAX_DATASET_NAME_LEN];
1292 		error = dmu_fsname(nvpair_name(pair), fsname);
1293 		if (error != 0)
1294 			return (error);
1295 		error = zfs_secpolicy_write_perms(fsname,
1296 		    ZFS_DELEG_PERM_RELEASE, cr);
1297 		if (error != 0)
1298 			return (error);
1299 	}
1300 	return (0);
1301 }
1302 
1303 /* ARGSUSED */
1304 static int
1305 zfs_secpolicy_load_key(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1306 {
1307 	return (zfs_secpolicy_write_perms(zc->zc_name,
1308 	    ZFS_DELEG_PERM_LOAD_KEY, cr));
1309 }
1310 
1311 /* ARGSUSED */
1312 static int
1313 zfs_secpolicy_change_key(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1314 {
1315 	return (zfs_secpolicy_write_perms(zc->zc_name,
1316 	    ZFS_DELEG_PERM_CHANGE_KEY, cr));
1317 }
1318 
1319 /*
1320  * Policy for allowing temporary snapshots to be taken or released
1321  */
1322 static int
1323 zfs_secpolicy_tmp_snapshot(zfs_cmd_t *zc, nvlist_t *innvl, cred_t *cr)
1324 {
1325 	/*
1326 	 * A temporary snapshot is the same as a snapshot,
1327 	 * hold, destroy and release all rolled into one.
1328 	 * Delegated diff alone is sufficient that we allow this.
1329 	 */
1330 	int error;
1331 
1332 	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
1333 	    ZFS_DELEG_PERM_DIFF, cr)) == 0)
1334 		return (0);
1335 
1336 	error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);
1337 	if (error == 0)
1338 		error = zfs_secpolicy_hold(zc, innvl, cr);
1339 	if (error == 0)
1340 		error = zfs_secpolicy_release(zc, innvl, cr);
1341 	if (error == 0)
1342 		error = zfs_secpolicy_destroy(zc, innvl, cr);
1343 	return (error);
1344 }
1345 
1346 /*
1347  * Returns the nvlist as specified by the user in the zfs_cmd_t.
1348  */
1349 static int
1350 get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
1351 {
1352 	char *packed;
1353 	int error;
1354 	nvlist_t *list = NULL;
1355 
1356 	/*
1357 	 * Read in and unpack the user-supplied nvlist.
1358 	 */
1359 	if (size == 0)
1360 		return (SET_ERROR(EINVAL));
1361 
1362 	packed = kmem_alloc(size, KM_SLEEP);
1363 
1364 	if ((error = ddi_copyin((void *)(uintptr_t)nvl, packed, size,
1365 	    iflag)) != 0) {
1366 		kmem_free(packed, size);
1367 		return (SET_ERROR(EFAULT));
1368 	}
1369 
1370 	if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
1371 		kmem_free(packed, size);
1372 		return (error);
1373 	}
1374 
1375 	kmem_free(packed, size);
1376 
1377 	*nvp = list;
1378 	return (0);
1379 }
1380 
1381 /*
1382  * Reduce the size of this nvlist until it can be serialized in 'max' bytes.
1383  * Entries will be removed from the end of the nvlist, and one int32 entry
1384  * named "N_MORE_ERRORS" will be added indicating how many entries were
1385  * removed.
1386  */
1387 static int
1388 nvlist_smush(nvlist_t *errors, size_t max)
1389 {
1390 	size_t size;
1391 
1392 	size = fnvlist_size(errors);
1393 
1394 	if (size > max) {
1395 		nvpair_t *more_errors;
1396 		int n = 0;
1397 
1398 		if (max < 1024)
1399 			return (SET_ERROR(ENOMEM));
1400 
1401 		fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
1402 		more_errors = nvlist_prev_nvpair(errors, NULL);
1403 
1404 		do {
1405 			nvpair_t *pair = nvlist_prev_nvpair(errors,
1406 			    more_errors);
1407 			fnvlist_remove_nvpair(errors, pair);
1408 			n++;
1409 			size = fnvlist_size(errors);
1410 		} while (size > max);
1411 
1412 		fnvlist_remove_nvpair(errors, more_errors);
1413 		fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
1414 		ASSERT3U(fnvlist_size(errors), <=, max);
1415 	}
1416 
1417 	return (0);
1418 }
1419 
1420 static int
1421 put_nvlist(zfs_cmd_t *zc, nvlist_t *nvl)
1422 {
1423 	char *packed = NULL;
1424 	int error = 0;
1425 	size_t size;
1426 
1427 	size = fnvlist_size(nvl);
1428 
1429 	if (size > zc->zc_nvlist_dst_size) {
1430 		error = SET_ERROR(ENOMEM);
1431 	} else {
1432 		packed = fnvlist_pack(nvl, &size);
1433 		if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
1434 		    size, zc->zc_iflags) != 0)
1435 			error = SET_ERROR(EFAULT);
1436 		fnvlist_pack_free(packed, size);
1437 	}
1438 
1439 	zc->zc_nvlist_dst_size = size;
1440 	zc->zc_nvlist_dst_filled = B_TRUE;
1441 	return (error);
1442 }
1443 
1444 int
1445 getzfsvfs_impl(objset_t *os, zfsvfs_t **zfvp)
1446 {
1447 	int error = 0;
1448 	if (dmu_objset_type(os) != DMU_OST_ZFS) {
1449 		return (SET_ERROR(EINVAL));
1450 	}
1451 
1452 	mutex_enter(&os->os_user_ptr_lock);
1453 	*zfvp = dmu_objset_get_user(os);
1454 	if (*zfvp) {
1455 		VFS_HOLD((*zfvp)->z_vfs);
1456 	} else {
1457 		error = SET_ERROR(ESRCH);
1458 	}
1459 	mutex_exit(&os->os_user_ptr_lock);
1460 	return (error);
1461 }
1462 
1463 int
1464 getzfsvfs(const char *dsname, zfsvfs_t **zfvp)
1465 {
1466 	objset_t *os;
1467 	int error;
1468 
1469 	error = dmu_objset_hold(dsname, FTAG, &os);
1470 	if (error != 0)
1471 		return (error);
1472 
1473 	error = getzfsvfs_impl(os, zfvp);
1474 	dmu_objset_rele(os, FTAG);
1475 	return (error);
1476 }
1477 
1478 /*
1479  * Find a zfsvfs_t for a mounted filesystem, or create our own, in which
1480  * case its z_vfs will be NULL, and it will be opened as the owner.
1481  * If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
1482  * which prevents all vnode ops from running.
1483  */
1484 static int
1485 zfsvfs_hold(const char *name, void *tag, zfsvfs_t **zfvp, boolean_t writer)
1486 {
1487 	int error = 0;
1488 
1489 	if (getzfsvfs(name, zfvp) != 0)
1490 		error = zfsvfs_create(name, B_FALSE, zfvp);
1491 	if (error == 0) {
1492 		rrm_enter(&(*zfvp)->z_teardown_lock, (writer) ? RW_WRITER :
1493 		    RW_READER, tag);
1494 		if ((*zfvp)->z_unmounted) {
1495 			/*
1496 			 * XXX we could probably try again, since the unmounting
1497 			 * thread should be just about to disassociate the
1498 			 * objset from the zfsvfs.
1499 			 */
1500 			rrm_exit(&(*zfvp)->z_teardown_lock, tag);
1501 			return (SET_ERROR(EBUSY));
1502 		}
1503 	}
1504 	return (error);
1505 }
1506 
1507 static void
1508 zfsvfs_rele(zfsvfs_t *zfsvfs, void *tag)
1509 {
1510 	rrm_exit(&zfsvfs->z_teardown_lock, tag);
1511 
1512 	if (zfsvfs->z_vfs) {
1513 		VFS_RELE(zfsvfs->z_vfs);
1514 	} else {
1515 		dmu_objset_disown(zfsvfs->z_os, B_TRUE, zfsvfs);
1516 		zfsvfs_free(zfsvfs);
1517 	}
1518 }
1519 
1520 static int
1521 zfs_ioc_pool_create(zfs_cmd_t *zc)
1522 {
1523 	int error;
1524 	nvlist_t *config, *props = NULL;
1525 	nvlist_t *rootprops = NULL;
1526 	nvlist_t *zplprops = NULL;
1527 	char *spa_name = zc->zc_name;
1528 	dsl_crypto_params_t *dcp = NULL;
1529 
1530 	if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1531 	    zc->zc_iflags, &config))
1532 		return (error);
1533 
1534 	if (zc->zc_nvlist_src_size != 0 && (error =
1535 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1536 	    zc->zc_iflags, &props))) {
1537 		nvlist_free(config);
1538 		return (error);
1539 	}
1540 
1541 	if (props) {
1542 		nvlist_t *nvl = NULL;
1543 		nvlist_t *hidden_args = NULL;
1544 		uint64_t version = SPA_VERSION;
1545 		char *tname;
1546 
1547 		(void) nvlist_lookup_uint64(props,
1548 		    zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
1549 		if (!SPA_VERSION_IS_SUPPORTED(version)) {
1550 			error = SET_ERROR(EINVAL);
1551 			goto pool_props_bad;
1552 		}
1553 		(void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
1554 		if (nvl) {
1555 			error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
1556 			if (error != 0) {
1557 				nvlist_free(config);
1558 				nvlist_free(props);
1559 				return (error);
1560 			}
1561 			(void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
1562 		}
1563 
1564 		(void) nvlist_lookup_nvlist(props, ZPOOL_HIDDEN_ARGS,
1565 		    &hidden_args);
1566 		error = dsl_crypto_params_create_nvlist(DCP_CMD_NONE,
1567 		    rootprops, hidden_args, &dcp);
1568 		if (error != 0) {
1569 			nvlist_free(config);
1570 			nvlist_free(props);
1571 			return (error);
1572 		}
1573 		(void) nvlist_remove_all(props, ZPOOL_HIDDEN_ARGS);
1574 
1575 		VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1576 		error = zfs_fill_zplprops_root(version, rootprops,
1577 		    zplprops, NULL);
1578 		if (error != 0)
1579 			goto pool_props_bad;
1580 
1581 		if (nvlist_lookup_string(props,
1582 		    zpool_prop_to_name(ZPOOL_PROP_TNAME), &tname) == 0)
1583 			spa_name = tname;
1584 	}
1585 
1586 	error = spa_create(zc->zc_name, config, props, zplprops, dcp);
1587 
1588 	/*
1589 	 * Set the remaining root properties
1590 	 */
1591 	if (!error && (error = zfs_set_prop_nvlist(spa_name,
1592 	    ZPROP_SRC_LOCAL, rootprops, NULL)) != 0)
1593 		(void) spa_destroy(spa_name);
1594 
1595 pool_props_bad:
1596 	nvlist_free(rootprops);
1597 	nvlist_free(zplprops);
1598 	nvlist_free(config);
1599 	nvlist_free(props);
1600 	dsl_crypto_params_free(dcp, !!error);
1601 
1602 	return (error);
1603 }
1604 
1605 static int
1606 zfs_ioc_pool_destroy(zfs_cmd_t *zc)
1607 {
1608 	int error;
1609 	zfs_log_history(zc);
1610 	error = spa_destroy(zc->zc_name);
1611 	if (error == 0)
1612 		zvol_remove_minors(zc->zc_name);
1613 	return (error);
1614 }
1615 
1616 static int
1617 zfs_ioc_pool_import(zfs_cmd_t *zc)
1618 {
1619 	nvlist_t *config, *props = NULL;
1620 	uint64_t guid;
1621 	int error;
1622 
1623 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1624 	    zc->zc_iflags, &config)) != 0)
1625 		return (error);
1626 
1627 	if (zc->zc_nvlist_src_size != 0 && (error =
1628 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
1629 	    zc->zc_iflags, &props))) {
1630 		nvlist_free(config);
1631 		return (error);
1632 	}
1633 
1634 	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 ||
1635 	    guid != zc->zc_guid)
1636 		error = SET_ERROR(EINVAL);
1637 	else
1638 		error = spa_import(zc->zc_name, config, props, zc->zc_cookie);
1639 
1640 	if (zc->zc_nvlist_dst != 0) {
1641 		int err;
1642 
1643 		if ((err = put_nvlist(zc, config)) != 0)
1644 			error = err;
1645 	}
1646 
1647 	nvlist_free(config);
1648 
1649 	nvlist_free(props);
1650 
1651 	return (error);
1652 }
1653 
1654 static int
1655 zfs_ioc_pool_export(zfs_cmd_t *zc)
1656 {
1657 	int error;
1658 	boolean_t force = (boolean_t)zc->zc_cookie;
1659 	boolean_t hardforce = (boolean_t)zc->zc_guid;
1660 
1661 	zfs_log_history(zc);
1662 	error = spa_export(zc->zc_name, NULL, force, hardforce);
1663 	if (error == 0)
1664 		zvol_remove_minors(zc->zc_name);
1665 	return (error);
1666 }
1667 
1668 static int
1669 zfs_ioc_pool_configs(zfs_cmd_t *zc)
1670 {
1671 	nvlist_t *configs;
1672 	int error;
1673 
1674 	if ((configs = spa_all_configs(&zc->zc_cookie)) == NULL)
1675 		return (SET_ERROR(EEXIST));
1676 
1677 	error = put_nvlist(zc, configs);
1678 
1679 	nvlist_free(configs);
1680 
1681 	return (error);
1682 }
1683 
1684 /*
1685  * inputs:
1686  * zc_name		name of the pool
1687  *
1688  * outputs:
1689  * zc_cookie		real errno
1690  * zc_nvlist_dst	config nvlist
1691  * zc_nvlist_dst_size	size of config nvlist
1692  */
1693 static int
1694 zfs_ioc_pool_stats(zfs_cmd_t *zc)
1695 {
1696 	nvlist_t *config;
1697 	int error;
1698 	int ret = 0;
1699 
1700 	error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
1701 	    sizeof (zc->zc_value));
1702 
1703 	if (config != NULL) {
1704 		ret = put_nvlist(zc, config);
1705 		nvlist_free(config);
1706 
1707 		/*
1708 		 * The config may be present even if 'error' is non-zero.
1709 		 * In this case we return success, and preserve the real errno
1710 		 * in 'zc_cookie'.
1711 		 */
1712 		zc->zc_cookie = error;
1713 	} else {
1714 		ret = error;
1715 	}
1716 
1717 	return (ret);
1718 }
1719 
1720 /*
1721  * Try to import the given pool, returning pool stats as appropriate so that
1722  * user land knows which devices are available and overall pool health.
1723  */
1724 static int
1725 zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
1726 {
1727 	nvlist_t *tryconfig, *config;
1728 	int error;
1729 
1730 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1731 	    zc->zc_iflags, &tryconfig)) != 0)
1732 		return (error);
1733 
1734 	config = spa_tryimport(tryconfig);
1735 
1736 	nvlist_free(tryconfig);
1737 
1738 	if (config == NULL)
1739 		return (SET_ERROR(EINVAL));
1740 
1741 	error = put_nvlist(zc, config);
1742 	nvlist_free(config);
1743 
1744 	return (error);
1745 }
1746 
1747 /*
1748  * inputs:
1749  * zc_name              name of the pool
1750  * zc_cookie            scan func (pool_scan_func_t)
1751  * zc_flags             scrub pause/resume flag (pool_scrub_cmd_t)
1752  */
1753 static int
1754 zfs_ioc_pool_scan(zfs_cmd_t *zc)
1755 {
1756 	spa_t *spa;
1757 	int error;
1758 
1759 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1760 		return (error);
1761 
1762 	if (zc->zc_flags >= POOL_SCRUB_FLAGS_END)
1763 		return (SET_ERROR(EINVAL));
1764 
1765 	if (zc->zc_flags == POOL_SCRUB_PAUSE)
1766 		error = spa_scrub_pause_resume(spa, POOL_SCRUB_PAUSE);
1767 	else if (zc->zc_cookie == POOL_SCAN_NONE)
1768 		error = spa_scan_stop(spa);
1769 	else
1770 		error = spa_scan(spa, zc->zc_cookie);
1771 
1772 	spa_close(spa, FTAG);
1773 
1774 	return (error);
1775 }
1776 
1777 static int
1778 zfs_ioc_pool_freeze(zfs_cmd_t *zc)
1779 {
1780 	spa_t *spa;
1781 	int error;
1782 
1783 	error = spa_open(zc->zc_name, &spa, FTAG);
1784 	if (error == 0) {
1785 		spa_freeze(spa);
1786 		spa_close(spa, FTAG);
1787 	}
1788 	return (error);
1789 }
1790 
1791 static int
1792 zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
1793 {
1794 	spa_t *spa;
1795 	int error;
1796 
1797 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1798 		return (error);
1799 
1800 	if (zc->zc_cookie < spa_version(spa) ||
1801 	    !SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
1802 		spa_close(spa, FTAG);
1803 		return (SET_ERROR(EINVAL));
1804 	}
1805 
1806 	spa_upgrade(spa, zc->zc_cookie);
1807 	spa_close(spa, FTAG);
1808 
1809 	return (error);
1810 }
1811 
1812 static int
1813 zfs_ioc_pool_get_history(zfs_cmd_t *zc)
1814 {
1815 	spa_t *spa;
1816 	char *hist_buf;
1817 	uint64_t size;
1818 	int error;
1819 
1820 	if ((size = zc->zc_history_len) == 0)
1821 		return (SET_ERROR(EINVAL));
1822 
1823 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1824 		return (error);
1825 
1826 	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
1827 		spa_close(spa, FTAG);
1828 		return (SET_ERROR(ENOTSUP));
1829 	}
1830 
1831 	hist_buf = kmem_alloc(size, KM_SLEEP);
1832 	if ((error = spa_history_get(spa, &zc->zc_history_offset,
1833 	    &zc->zc_history_len, hist_buf)) == 0) {
1834 		error = ddi_copyout(hist_buf,
1835 		    (void *)(uintptr_t)zc->zc_history,
1836 		    zc->zc_history_len, zc->zc_iflags);
1837 	}
1838 
1839 	spa_close(spa, FTAG);
1840 	kmem_free(hist_buf, size);
1841 	return (error);
1842 }
1843 
1844 static int
1845 zfs_ioc_pool_reguid(zfs_cmd_t *zc)
1846 {
1847 	spa_t *spa;
1848 	int error;
1849 
1850 	error = spa_open(zc->zc_name, &spa, FTAG);
1851 	if (error == 0) {
1852 		error = spa_change_guid(spa);
1853 		spa_close(spa, FTAG);
1854 	}
1855 	return (error);
1856 }
1857 
1858 static int
1859 zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
1860 {
1861 	return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
1862 }
1863 
1864 /*
1865  * inputs:
1866  * zc_name		name of filesystem
1867  * zc_obj		object to find
1868  *
1869  * outputs:
1870  * zc_value		name of object
1871  */
1872 static int
1873 zfs_ioc_obj_to_path(zfs_cmd_t *zc)
1874 {
1875 	objset_t *os;
1876 	int error;
1877 
1878 	/* XXX reading from objset not owned */
1879 	if ((error = dmu_objset_hold_flags(zc->zc_name, B_TRUE,
1880 	    FTAG, &os)) != 0)
1881 		return (error);
1882 	if (dmu_objset_type(os) != DMU_OST_ZFS) {
1883 		dmu_objset_rele_flags(os, B_TRUE, FTAG);
1884 		return (SET_ERROR(EINVAL));
1885 	}
1886 	error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
1887 	    sizeof (zc->zc_value));
1888 	dmu_objset_rele_flags(os, B_TRUE, FTAG);
1889 
1890 	return (error);
1891 }
1892 
1893 /*
1894  * inputs:
1895  * zc_name		name of filesystem
1896  * zc_obj		object to find
1897  *
1898  * outputs:
1899  * zc_stat		stats on object
1900  * zc_value		path to object
1901  */
1902 static int
1903 zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
1904 {
1905 	objset_t *os;
1906 	int error;
1907 
1908 	/* XXX reading from objset not owned */
1909 	if ((error = dmu_objset_hold_flags(zc->zc_name, B_TRUE,
1910 	    FTAG, &os)) != 0)
1911 		return (error);
1912 	if (dmu_objset_type(os) != DMU_OST_ZFS) {
1913 		dmu_objset_rele_flags(os, B_TRUE, FTAG);
1914 		return (SET_ERROR(EINVAL));
1915 	}
1916 	error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
1917 	    sizeof (zc->zc_value));
1918 	dmu_objset_rele_flags(os, B_TRUE, FTAG);
1919 
1920 	return (error);
1921 }
1922 
1923 static int
1924 zfs_ioc_vdev_add(zfs_cmd_t *zc)
1925 {
1926 	spa_t *spa;
1927 	int error;
1928 	nvlist_t *config, **l2cache, **spares;
1929 	uint_t nl2cache = 0, nspares = 0;
1930 
1931 	error = spa_open(zc->zc_name, &spa, FTAG);
1932 	if (error != 0)
1933 		return (error);
1934 
1935 	error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
1936 	    zc->zc_iflags, &config);
1937 	(void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_L2CACHE,
1938 	    &l2cache, &nl2cache);
1939 
1940 	(void) nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_SPARES,
1941 	    &spares, &nspares);
1942 
1943 	/*
1944 	 * A root pool with concatenated devices is not supported.
1945 	 * Thus, can not add a device to a root pool.
1946 	 *
1947 	 * Intent log device can not be added to a rootpool because
1948 	 * during mountroot, zil is replayed, a seperated log device
1949 	 * can not be accessed during the mountroot time.
1950 	 *
1951 	 * l2cache and spare devices are ok to be added to a rootpool.
1952 	 */
1953 	if (spa_bootfs(spa) != 0 && nl2cache == 0 && nspares == 0) {
1954 		nvlist_free(config);
1955 		spa_close(spa, FTAG);
1956 		return (SET_ERROR(EDOM));
1957 	}
1958 
1959 	if (error == 0) {
1960 		error = spa_vdev_add(spa, config);
1961 		nvlist_free(config);
1962 	}
1963 	spa_close(spa, FTAG);
1964 	return (error);
1965 }
1966 
1967 /*
1968  * inputs:
1969  * zc_name		name of the pool
1970  * zc_guid		guid of vdev to remove
1971  * zc_cookie		cancel removal
1972  */
1973 static int
1974 zfs_ioc_vdev_remove(zfs_cmd_t *zc)
1975 {
1976 	spa_t *spa;
1977 	int error;
1978 
1979 	error = spa_open(zc->zc_name, &spa, FTAG);
1980 	if (error != 0)
1981 		return (error);
1982 	if (zc->zc_cookie != 0) {
1983 		error = spa_vdev_remove_cancel(spa);
1984 	} else {
1985 		error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
1986 	}
1987 	spa_close(spa, FTAG);
1988 	return (error);
1989 }
1990 
1991 static int
1992 zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
1993 {
1994 	spa_t *spa;
1995 	int error;
1996 	vdev_state_t newstate = VDEV_STATE_UNKNOWN;
1997 
1998 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
1999 		return (error);
2000 	switch (zc->zc_cookie) {
2001 	case VDEV_STATE_ONLINE:
2002 		error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
2003 		break;
2004 
2005 	case VDEV_STATE_OFFLINE:
2006 		error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
2007 		break;
2008 
2009 	case VDEV_STATE_FAULTED:
2010 		if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
2011 		    zc->zc_obj != VDEV_AUX_EXTERNAL)
2012 			zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
2013 
2014 		error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
2015 		break;
2016 
2017 	case VDEV_STATE_DEGRADED:
2018 		if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
2019 		    zc->zc_obj != VDEV_AUX_EXTERNAL)
2020 			zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;
2021 
2022 		error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
2023 		break;
2024 
2025 	default:
2026 		error = SET_ERROR(EINVAL);
2027 	}
2028 	zc->zc_cookie = newstate;
2029 	spa_close(spa, FTAG);
2030 	return (error);
2031 }
2032 
2033 static int
2034 zfs_ioc_vdev_attach(zfs_cmd_t *zc)
2035 {
2036 	spa_t *spa;
2037 	int replacing = zc->zc_cookie;
2038 	nvlist_t *config;
2039 	int error;
2040 
2041 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2042 		return (error);
2043 
2044 	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
2045 	    zc->zc_iflags, &config)) == 0) {
2046 		error = spa_vdev_attach(spa, zc->zc_guid, config, replacing);
2047 		nvlist_free(config);
2048 	}
2049 
2050 	spa_close(spa, FTAG);
2051 	return (error);
2052 }
2053 
2054 static int
2055 zfs_ioc_vdev_detach(zfs_cmd_t *zc)
2056 {
2057 	spa_t *spa;
2058 	int error;
2059 
2060 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2061 		return (error);
2062 
2063 	error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);
2064 
2065 	spa_close(spa, FTAG);
2066 	return (error);
2067 }
2068 
2069 static int
2070 zfs_ioc_vdev_split(zfs_cmd_t *zc)
2071 {
2072 	spa_t *spa;
2073 	nvlist_t *config, *props = NULL;
2074 	int error;
2075 	boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);
2076 
2077 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
2078 		return (error);
2079 
2080 	if (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
2081 	    zc->zc_iflags, &config)) {
2082 		spa_close(spa, FTAG);
2083 		return (error);
2084 	}
2085 
2086 	if (zc->zc_nvlist_src_size != 0 && (error =
2087 	    get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2088 	    zc->zc_iflags, &props))) {
2089 		spa_close(spa, FTAG);
2090 		nvlist_free(config);
2091 		return (error);
2092 	}
2093 
2094 	error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);
2095 
2096 	spa_close(spa, FTAG);
2097 
2098 	nvlist_free(config);
2099 	nvlist_free(props);
2100 
2101 	return (error);
2102 }
2103 
2104 static int
2105 zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
2106 {
2107 	spa_t *spa;
2108 	char *path = zc->zc_value;
2109 	uint64_t guid = zc->zc_guid;
2110 	int error;
2111 
2112 	error = spa_open(zc->zc_name, &spa, FTAG);
2113 	if (error != 0)
2114 		return (error);
2115 
2116 	error = spa_vdev_setpath(spa, guid, path);
2117 	spa_close(spa, FTAG);
2118 	return (error);
2119 }
2120 
2121 static int
2122 zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
2123 {
2124 	spa_t *spa;
2125 	char *fru = zc->zc_value;
2126 	uint64_t guid = zc->zc_guid;
2127 	int error;
2128 
2129 	error = spa_open(zc->zc_name, &spa, FTAG);
2130 	if (error != 0)
2131 		return (error);
2132 
2133 	error = spa_vdev_setfru(spa, guid, fru);
2134 	spa_close(spa, FTAG);
2135 	return (error);
2136 }
2137 
2138 static int
2139 zfs_ioc_objset_stats_impl(zfs_cmd_t *zc, objset_t *os)
2140 {
2141 	int error = 0;
2142 	nvlist_t *nv;
2143 
2144 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2145 
2146 	if (zc->zc_nvlist_dst != 0 &&
2147 	    (error = dsl_prop_get_all(os, &nv)) == 0) {
2148 		dmu_objset_stats(os, nv);
2149 		/*
2150 		 * NB: zvol_get_stats() will read the objset contents,
2151 		 * which we aren't supposed to do with a
2152 		 * DS_MODE_USER hold, because it could be
2153 		 * inconsistent.  So this is a bit of a workaround...
2154 		 * XXX reading with out owning
2155 		 */
2156 		if (!zc->zc_objset_stats.dds_inconsistent &&
2157 		    dmu_objset_type(os) == DMU_OST_ZVOL) {
2158 			error = zvol_get_stats(os, nv);
2159 			if (error == EIO)
2160 				return (error);
2161 			VERIFY0(error);
2162 		}
2163 		error = put_nvlist(zc, nv);
2164 		nvlist_free(nv);
2165 	}
2166 
2167 	return (error);
2168 }
2169 
2170 /*
2171  * inputs:
2172  * zc_name		name of filesystem
2173  * zc_nvlist_dst_size	size of buffer for property nvlist
2174  *
2175  * outputs:
2176  * zc_objset_stats	stats
2177  * zc_nvlist_dst	property nvlist
2178  * zc_nvlist_dst_size	size of property nvlist
2179  */
2180 static int
2181 zfs_ioc_objset_stats(zfs_cmd_t *zc)
2182 {
2183 	objset_t *os;
2184 	int error;
2185 
2186 	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2187 	if (error == 0) {
2188 		error = zfs_ioc_objset_stats_impl(zc, os);
2189 		dmu_objset_rele(os, FTAG);
2190 	}
2191 
2192 	return (error);
2193 }
2194 
2195 /*
2196  * inputs:
2197  * zc_name		name of filesystem
2198  * zc_nvlist_dst_size	size of buffer for property nvlist
2199  *
2200  * outputs:
2201  * zc_nvlist_dst	received property nvlist
2202  * zc_nvlist_dst_size	size of received property nvlist
2203  *
2204  * Gets received properties (distinct from local properties on or after
2205  * SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
2206  * local property values.
2207  */
2208 static int
2209 zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
2210 {
2211 	int error = 0;
2212 	nvlist_t *nv;
2213 
2214 	/*
2215 	 * Without this check, we would return local property values if the
2216 	 * caller has not already received properties on or after
2217 	 * SPA_VERSION_RECVD_PROPS.
2218 	 */
2219 	if (!dsl_prop_get_hasrecvd(zc->zc_name))
2220 		return (SET_ERROR(ENOTSUP));
2221 
2222 	if (zc->zc_nvlist_dst != 0 &&
2223 	    (error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
2224 		error = put_nvlist(zc, nv);
2225 		nvlist_free(nv);
2226 	}
2227 
2228 	return (error);
2229 }
2230 
2231 static int
2232 nvl_add_zplprop(objset_t *os, nvlist_t *props, zfs_prop_t prop)
2233 {
2234 	uint64_t value;
2235 	int error;
2236 
2237 	/*
2238 	 * zfs_get_zplprop() will either find a value or give us
2239 	 * the default value (if there is one).
2240 	 */
2241 	if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
2242 		return (error);
2243 	VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
2244 	return (0);
2245 }
2246 
2247 /*
2248  * inputs:
2249  * zc_name		name of filesystem
2250  * zc_nvlist_dst_size	size of buffer for zpl property nvlist
2251  *
2252  * outputs:
2253  * zc_nvlist_dst	zpl property nvlist
2254  * zc_nvlist_dst_size	size of zpl property nvlist
2255  */
2256 static int
2257 zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
2258 {
2259 	objset_t *os;
2260 	int err;
2261 
2262 	/* XXX reading without owning */
2263 	if (err = dmu_objset_hold(zc->zc_name, FTAG, &os))
2264 		return (err);
2265 
2266 	dmu_objset_fast_stat(os, &zc->zc_objset_stats);
2267 
2268 	/*
2269 	 * NB: nvl_add_zplprop() will read the objset contents,
2270 	 * which we aren't supposed to do with a DS_MODE_USER
2271 	 * hold, because it could be inconsistent.
2272 	 */
2273 	if (zc->zc_nvlist_dst != 0 &&
2274 	    !zc->zc_objset_stats.dds_inconsistent &&
2275 	    dmu_objset_type(os) == DMU_OST_ZFS) {
2276 		nvlist_t *nv;
2277 
2278 		VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2279 		if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
2280 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
2281 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
2282 		    (err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
2283 			err = put_nvlist(zc, nv);
2284 		nvlist_free(nv);
2285 	} else {
2286 		err = SET_ERROR(ENOENT);
2287 	}
2288 	dmu_objset_rele(os, FTAG);
2289 	return (err);
2290 }
2291 
2292 static boolean_t
2293 dataset_name_hidden(const char *name)
2294 {
2295 	/*
2296 	 * Skip over datasets that are not visible in this zone,
2297 	 * internal datasets (which have a $ in their name), and
2298 	 * temporary datasets (which have a % in their name).
2299 	 */
2300 	if (strchr(name, '$') != NULL)
2301 		return (B_TRUE);
2302 	if (strchr(name, '%') != NULL)
2303 		return (B_TRUE);
2304 	if (!INGLOBALZONE(curproc) && !zone_dataset_visible(name, NULL))
2305 		return (B_TRUE);
2306 	return (B_FALSE);
2307 }
2308 
2309 /*
2310  * inputs:
2311  * zc_name		name of filesystem
2312  * zc_cookie		zap cursor
2313  * zc_nvlist_dst_size	size of buffer for property nvlist
2314  *
2315  * outputs:
2316  * zc_name		name of next filesystem
2317  * zc_cookie		zap cursor
2318  * zc_objset_stats	stats
2319  * zc_nvlist_dst	property nvlist
2320  * zc_nvlist_dst_size	size of property nvlist
2321  */
2322 static int
2323 zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
2324 {
2325 	objset_t *os;
2326 	int error;
2327 	char *p;
2328 	size_t orig_len = strlen(zc->zc_name);
2329 
2330 top:
2331 	if (error = dmu_objset_hold(zc->zc_name, FTAG, &os)) {
2332 		if (error == ENOENT)
2333 			error = SET_ERROR(ESRCH);
2334 		return (error);
2335 	}
2336 
2337 	p = strrchr(zc->zc_name, '/');
2338 	if (p == NULL || p[1] != '\0')
2339 		(void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
2340 	p = zc->zc_name + strlen(zc->zc_name);
2341 
2342 	do {
2343 		error = dmu_dir_list_next(os,
2344 		    sizeof (zc->zc_name) - (p - zc->zc_name), p,
2345 		    NULL, &zc->zc_cookie);
2346 		if (error == ENOENT)
2347 			error = SET_ERROR(ESRCH);
2348 	} while (error == 0 && dataset_name_hidden(zc->zc_name));
2349 	dmu_objset_rele(os, FTAG);
2350 
2351 	/*
2352 	 * If it's an internal dataset (ie. with a '$' in its name),
2353 	 * don't try to get stats for it, otherwise we'll return ENOENT.
2354 	 */
2355 	if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
2356 		error = zfs_ioc_objset_stats(zc); /* fill in the stats */
2357 		if (error == ENOENT) {
2358 			/* We lost a race with destroy, get the next one. */
2359 			zc->zc_name[orig_len] = '\0';
2360 			goto top;
2361 		}
2362 	}
2363 	return (error);
2364 }
2365 
2366 /*
2367  * inputs:
2368  * zc_name		name of filesystem
2369  * zc_cookie		zap cursor
2370  * zc_nvlist_dst_size	size of buffer for property nvlist
2371  * zc_simple		when set, only name is requested
2372  *
2373  * outputs:
2374  * zc_name		name of next snapshot
2375  * zc_objset_stats	stats
2376  * zc_nvlist_dst	property nvlist
2377  * zc_nvlist_dst_size	size of property nvlist
2378  */
2379 static int
2380 zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
2381 {
2382 	objset_t *os;
2383 	int error;
2384 
2385 	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
2386 	if (error != 0) {
2387 		return (error == ENOENT ? ESRCH : error);
2388 	}
2389 
2390 	/*
2391 	 * A dataset name of maximum length cannot have any snapshots,
2392 	 * so exit immediately.
2393 	 */
2394 	if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >=
2395 	    ZFS_MAX_DATASET_NAME_LEN) {
2396 		dmu_objset_rele(os, FTAG);
2397 		return (SET_ERROR(ESRCH));
2398 	}
2399 
2400 	error = dmu_snapshot_list_next(os,
2401 	    sizeof (zc->zc_name) - strlen(zc->zc_name),
2402 	    zc->zc_name + strlen(zc->zc_name), &zc->zc_obj, &zc->zc_cookie,
2403 	    NULL);
2404 
2405 	if (error == 0 && !zc->zc_simple) {
2406 		dsl_dataset_t *ds;
2407 		dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool;
2408 
2409 		error = dsl_dataset_hold_obj(dp, zc->zc_obj, FTAG, &ds);
2410 		if (error == 0) {
2411 			objset_t *ossnap;
2412 
2413 			error = dmu_objset_from_ds(ds, &ossnap);
2414 			if (error == 0)
2415 				error = zfs_ioc_objset_stats_impl(zc, ossnap);
2416 			dsl_dataset_rele(ds, FTAG);
2417 		}
2418 	} else if (error == ENOENT) {
2419 		error = SET_ERROR(ESRCH);
2420 	}
2421 
2422 	dmu_objset_rele(os, FTAG);
2423 	/* if we failed, undo the @ that we tacked on to zc_name */
2424 	if (error != 0)
2425 		*strchr(zc->zc_name, '@') = '\0';
2426 	return (error);
2427 }
2428 
2429 static int
2430 zfs_prop_set_userquota(const char *dsname, nvpair_t *pair)
2431 {
2432 	const char *propname = nvpair_name(pair);
2433 	uint64_t *valary;
2434 	unsigned int vallen;
2435 	const char *domain;
2436 	char *dash;
2437 	zfs_userquota_prop_t type;
2438 	uint64_t rid;
2439 	uint64_t quota;
2440 	zfsvfs_t *zfsvfs;
2441 	int err;
2442 
2443 	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2444 		nvlist_t *attrs;
2445 		VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2446 		if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2447 		    &pair) != 0)
2448 			return (SET_ERROR(EINVAL));
2449 	}
2450 
2451 	/*
2452 	 * A correctly constructed propname is encoded as
2453 	 * userquota@<rid>-<domain>.
2454 	 */
2455 	if ((dash = strchr(propname, '-')) == NULL ||
2456 	    nvpair_value_uint64_array(pair, &valary, &vallen) != 0 ||
2457 	    vallen != 3)
2458 		return (SET_ERROR(EINVAL));
2459 
2460 	domain = dash + 1;
2461 	type = valary[0];
2462 	rid = valary[1];
2463 	quota = valary[2];
2464 
2465 	err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
2466 	if (err == 0) {
2467 		err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
2468 		zfsvfs_rele(zfsvfs, FTAG);
2469 	}
2470 
2471 	return (err);
2472 }
2473 
2474 /*
2475  * If the named property is one that has a special function to set its value,
2476  * return 0 on success and a positive error code on failure; otherwise if it is
2477  * not one of the special properties handled by this function, return -1.
2478  *
2479  * XXX: It would be better for callers of the property interface if we handled
2480  * these special cases in dsl_prop.c (in the dsl layer).
2481  */
2482 static int
2483 zfs_prop_set_special(const char *dsname, zprop_source_t source,
2484     nvpair_t *pair)
2485 {
2486 	const char *propname = nvpair_name(pair);
2487 	zfs_prop_t prop = zfs_name_to_prop(propname);
2488 	uint64_t intval = 0;
2489 	char *strval = NULL;
2490 	int err = -1;
2491 
2492 	if (prop == ZPROP_INVAL) {
2493 		if (zfs_prop_userquota(propname))
2494 			return (zfs_prop_set_userquota(dsname, pair));
2495 		return (-1);
2496 	}
2497 
2498 	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2499 		nvlist_t *attrs;
2500 		VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
2501 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2502 		    &pair) == 0);
2503 	}
2504 
2505 	/* all special properties are numeric except for keylocation */
2506 	if (zfs_prop_get_type(prop) == PROP_TYPE_STRING) {
2507 		strval = fnvpair_value_string(pair);
2508 	} else {
2509 		intval = fnvpair_value_uint64(pair);
2510 	}
2511 
2512 	switch (prop) {
2513 	case ZFS_PROP_QUOTA:
2514 		err = dsl_dir_set_quota(dsname, source, intval);
2515 		break;
2516 	case ZFS_PROP_REFQUOTA:
2517 		err = dsl_dataset_set_refquota(dsname, source, intval);
2518 		break;
2519 	case ZFS_PROP_FILESYSTEM_LIMIT:
2520 	case ZFS_PROP_SNAPSHOT_LIMIT:
2521 		if (intval == UINT64_MAX) {
2522 			/* clearing the limit, just do it */
2523 			err = 0;
2524 		} else {
2525 			err = dsl_dir_activate_fs_ss_limit(dsname);
2526 		}
2527 		/*
2528 		 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2529 		 * default path to set the value in the nvlist.
2530 		 */
2531 		if (err == 0)
2532 			err = -1;
2533 		break;
2534 	case ZFS_PROP_KEYLOCATION:
2535 		err = dsl_crypto_can_set_keylocation(dsname, strval);
2536 
2537 		/*
2538 		 * Set err to -1 to force the zfs_set_prop_nvlist code down the
2539 		 * default path to set the value in the nvlist.
2540 		 */
2541 		if (err == 0)
2542 			err = -1;
2543 		break;
2544 	case ZFS_PROP_RESERVATION:
2545 		err = dsl_dir_set_reservation(dsname, source, intval);
2546 		break;
2547 	case ZFS_PROP_REFRESERVATION:
2548 		err = dsl_dataset_set_refreservation(dsname, source, intval);
2549 		break;
2550 	case ZFS_PROP_VOLSIZE:
2551 		err = zvol_set_volsize(dsname, intval);
2552 		break;
2553 	case ZFS_PROP_VERSION:
2554 	{
2555 		zfsvfs_t *zfsvfs;
2556 
2557 		if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
2558 			break;
2559 
2560 		err = zfs_set_version(zfsvfs, intval);
2561 		zfsvfs_rele(zfsvfs, FTAG);
2562 
2563 		if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
2564 			zfs_cmd_t *zc;
2565 
2566 			zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
2567 			(void) strcpy(zc->zc_name, dsname);
2568 			(void) zfs_ioc_userspace_upgrade(zc);
2569 			(void) zfs_ioc_id_quota_upgrade(zc);
2570 			kmem_free(zc, sizeof (zfs_cmd_t));
2571 		}
2572 		break;
2573 	}
2574 	default:
2575 		err = -1;
2576 	}
2577 
2578 	return (err);
2579 }
2580 
2581 /*
2582  * This function is best effort. If it fails to set any of the given properties,
2583  * it continues to set as many as it can and returns the last error
2584  * encountered. If the caller provides a non-NULL errlist, it will be filled in
2585  * with the list of names of all the properties that failed along with the
2586  * corresponding error numbers.
2587  *
2588  * If every property is set successfully, zero is returned and errlist is not
2589  * modified.
2590  */
2591 int
2592 zfs_set_prop_nvlist(const char *dsname, zprop_source_t source, nvlist_t *nvl,
2593     nvlist_t *errlist)
2594 {
2595 	nvpair_t *pair;
2596 	nvpair_t *propval;
2597 	int rv = 0;
2598 	uint64_t intval;
2599 	char *strval;
2600 	nvlist_t *genericnvl = fnvlist_alloc();
2601 	nvlist_t *retrynvl = fnvlist_alloc();
2602 
2603 retry:
2604 	pair = NULL;
2605 	while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2606 		const char *propname = nvpair_name(pair);
2607 		zfs_prop_t prop = zfs_name_to_prop(propname);
2608 		int err = 0;
2609 
2610 		/* decode the property value */
2611 		propval = pair;
2612 		if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2613 			nvlist_t *attrs;
2614 			attrs = fnvpair_value_nvlist(pair);
2615 			if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
2616 			    &propval) != 0)
2617 				err = SET_ERROR(EINVAL);
2618 		}
2619 
2620 		/* Validate value type */
2621 		if (err == 0 && source == ZPROP_SRC_INHERITED) {
2622 			/* inherited properties are expected to be booleans */
2623 			if (nvpair_type(propval) != DATA_TYPE_BOOLEAN)
2624 				err = SET_ERROR(EINVAL);
2625 		} else if (err == 0 && prop == ZPROP_INVAL) {
2626 			if (zfs_prop_user(propname)) {
2627 				if (nvpair_type(propval) != DATA_TYPE_STRING)
2628 					err = SET_ERROR(EINVAL);
2629 			} else if (zfs_prop_userquota(propname)) {
2630 				if (nvpair_type(propval) !=
2631 				    DATA_TYPE_UINT64_ARRAY)
2632 					err = SET_ERROR(EINVAL);
2633 			} else {
2634 				err = SET_ERROR(EINVAL);
2635 			}
2636 		} else if (err == 0) {
2637 			if (nvpair_type(propval) == DATA_TYPE_STRING) {
2638 				if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
2639 					err = SET_ERROR(EINVAL);
2640 			} else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
2641 				const char *unused;
2642 
2643 				intval = fnvpair_value_uint64(propval);
2644 
2645 				switch (zfs_prop_get_type(prop)) {
2646 				case PROP_TYPE_NUMBER:
2647 					break;
2648 				case PROP_TYPE_STRING:
2649 					err = SET_ERROR(EINVAL);
2650 					break;
2651 				case PROP_TYPE_INDEX:
2652 					if (zfs_prop_index_to_string(prop,
2653 					    intval, &unused) != 0)
2654 						err = SET_ERROR(EINVAL);
2655 					break;
2656 				default:
2657 					cmn_err(CE_PANIC,
2658 					    "unknown property type");
2659 				}
2660 			} else {
2661 				err = SET_ERROR(EINVAL);
2662 			}
2663 		}
2664 
2665 		/* Validate permissions */
2666 		if (err == 0)
2667 			err = zfs_check_settable(dsname, pair, CRED());
2668 
2669 		if (err == 0) {
2670 			if (source == ZPROP_SRC_INHERITED)
2671 				err = -1; /* does not need special handling */
2672 			else
2673 				err = zfs_prop_set_special(dsname, source,
2674 				    pair);
2675 			if (err == -1) {
2676 				/*
2677 				 * For better performance we build up a list of
2678 				 * properties to set in a single transaction.
2679 				 */
2680 				err = nvlist_add_nvpair(genericnvl, pair);
2681 			} else if (err != 0 && nvl != retrynvl) {
2682 				/*
2683 				 * This may be a spurious error caused by
2684 				 * receiving quota and reservation out of order.
2685 				 * Try again in a second pass.
2686 				 */
2687 				err = nvlist_add_nvpair(retrynvl, pair);
2688 			}
2689 		}
2690 
2691 		if (err != 0) {
2692 			if (errlist != NULL)
2693 				fnvlist_add_int32(errlist, propname, err);
2694 			rv = err;
2695 		}
2696 	}
2697 
2698 	if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
2699 		nvl = retrynvl;
2700 		goto retry;
2701 	}
2702 
2703 	if (!nvlist_empty(genericnvl) &&
2704 	    dsl_props_set(dsname, source, genericnvl) != 0) {
2705 		/*
2706 		 * If this fails, we still want to set as many properties as we
2707 		 * can, so try setting them individually.
2708 		 */
2709 		pair = NULL;
2710 		while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
2711 			const char *propname = nvpair_name(pair);
2712 			int err = 0;
2713 
2714 			propval = pair;
2715 			if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
2716 				nvlist_t *attrs;
2717 				attrs = fnvpair_value_nvlist(pair);
2718 				propval = fnvlist_lookup_nvpair(attrs,
2719 				    ZPROP_VALUE);
2720 			}
2721 
2722 			if (nvpair_type(propval) == DATA_TYPE_STRING) {
2723 				strval = fnvpair_value_string(propval);
2724 				err = dsl_prop_set_string(dsname, propname,
2725 				    source, strval);
2726 			} else if (nvpair_type(propval) == DATA_TYPE_BOOLEAN) {
2727 				err = dsl_prop_inherit(dsname, propname,
2728 				    source);
2729 			} else {
2730 				intval = fnvpair_value_uint64(propval);
2731 				err = dsl_prop_set_int(dsname, propname, source,
2732 				    intval);
2733 			}
2734 
2735 			if (err != 0) {
2736 				if (errlist != NULL) {
2737 					fnvlist_add_int32(errlist, propname,
2738 					    err);
2739 				}
2740 				rv = err;
2741 			}
2742 		}
2743 	}
2744 	nvlist_free(genericnvl);
2745 	nvlist_free(retrynvl);
2746 
2747 	return (rv);
2748 }
2749 
2750 /*
2751  * Check that all the properties are valid user properties.
2752  */
2753 static int
2754 zfs_check_userprops(const char *fsname, nvlist_t *nvl)
2755 {
2756 	nvpair_t *pair = NULL;
2757 	int error = 0;
2758 
2759 	while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
2760 		const char *propname = nvpair_name(pair);
2761 
2762 		if (!zfs_prop_user(propname) ||
2763 		    nvpair_type(pair) != DATA_TYPE_STRING)
2764 			return (SET_ERROR(EINVAL));
2765 
2766 		if (error = zfs_secpolicy_write_perms(fsname,
2767 		    ZFS_DELEG_PERM_USERPROP, CRED()))
2768 			return (error);
2769 
2770 		if (strlen(propname) >= ZAP_MAXNAMELEN)
2771 			return (SET_ERROR(ENAMETOOLONG));
2772 
2773 		if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
2774 			return (E2BIG);
2775 	}
2776 	return (0);
2777 }
2778 
2779 static void
2780 props_skip(nvlist_t *props, nvlist_t *skipped, nvlist_t **newprops)
2781 {
2782 	nvpair_t *pair;
2783 
2784 	VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2785 
2786 	pair = NULL;
2787 	while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
2788 		if (nvlist_exists(skipped, nvpair_name(pair)))
2789 			continue;
2790 
2791 		VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
2792 	}
2793 }
2794 
2795 static int
2796 clear_received_props(const char *dsname, nvlist_t *props,
2797     nvlist_t *skipped)
2798 {
2799 	int err = 0;
2800 	nvlist_t *cleared_props = NULL;
2801 	props_skip(props, skipped, &cleared_props);
2802 	if (!nvlist_empty(cleared_props)) {
2803 		/*
2804 		 * Acts on local properties until the dataset has received
2805 		 * properties at least once on or after SPA_VERSION_RECVD_PROPS.
2806 		 */
2807 		zprop_source_t flags = (ZPROP_SRC_NONE |
2808 		    (dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
2809 		err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
2810 	}
2811 	nvlist_free(cleared_props);
2812 	return (err);
2813 }
2814 
2815 /*
2816  * inputs:
2817  * zc_name		name of filesystem
2818  * zc_value		name of property to set
2819  * zc_nvlist_src{_size}	nvlist of properties to apply
2820  * zc_cookie		received properties flag
2821  *
2822  * outputs:
2823  * zc_nvlist_dst{_size} error for each unapplied received property
2824  */
2825 static int
2826 zfs_ioc_set_prop(zfs_cmd_t *zc)
2827 {
2828 	nvlist_t *nvl;
2829 	boolean_t received = zc->zc_cookie;
2830 	zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
2831 	    ZPROP_SRC_LOCAL);
2832 	nvlist_t *errors;
2833 	int error;
2834 
2835 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2836 	    zc->zc_iflags, &nvl)) != 0)
2837 		return (error);
2838 
2839 	if (received) {
2840 		nvlist_t *origprops;
2841 
2842 		if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
2843 			(void) clear_received_props(zc->zc_name,
2844 			    origprops, nvl);
2845 			nvlist_free(origprops);
2846 		}
2847 
2848 		error = dsl_prop_set_hasrecvd(zc->zc_name);
2849 	}
2850 
2851 	errors = fnvlist_alloc();
2852 	if (error == 0)
2853 		error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);
2854 
2855 	if (zc->zc_nvlist_dst != 0 && errors != NULL) {
2856 		(void) put_nvlist(zc, errors);
2857 	}
2858 
2859 	nvlist_free(errors);
2860 	nvlist_free(nvl);
2861 	return (error);
2862 }
2863 
2864 /*
2865  * inputs:
2866  * zc_name		name of filesystem
2867  * zc_value		name of property to inherit
2868  * zc_cookie		revert to received value if TRUE
2869  *
2870  * outputs:		none
2871  */
2872 static int
2873 zfs_ioc_inherit_prop(zfs_cmd_t *zc)
2874 {
2875 	const char *propname = zc->zc_value;
2876 	zfs_prop_t prop = zfs_name_to_prop(propname);
2877 	boolean_t received = zc->zc_cookie;
2878 	zprop_source_t source = (received
2879 	    ? ZPROP_SRC_NONE		/* revert to received value, if any */
2880 	    : ZPROP_SRC_INHERITED);	/* explicitly inherit */
2881 
2882 	if (received) {
2883 		nvlist_t *dummy;
2884 		nvpair_t *pair;
2885 		zprop_type_t type;
2886 		int err;
2887 
2888 		/*
2889 		 * zfs_prop_set_special() expects properties in the form of an
2890 		 * nvpair with type info.
2891 		 */
2892 		if (prop == ZPROP_INVAL) {
2893 			if (!zfs_prop_user(propname))
2894 				return (SET_ERROR(EINVAL));
2895 
2896 			type = PROP_TYPE_STRING;
2897 		} else if (prop == ZFS_PROP_VOLSIZE ||
2898 		    prop == ZFS_PROP_VERSION) {
2899 			return (SET_ERROR(EINVAL));
2900 		} else {
2901 			type = zfs_prop_get_type(prop);
2902 		}
2903 
2904 		VERIFY(nvlist_alloc(&dummy, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2905 
2906 		switch (type) {
2907 		case PROP_TYPE_STRING:
2908 			VERIFY(0 == nvlist_add_string(dummy, propname, ""));
2909 			break;
2910 		case PROP_TYPE_NUMBER:
2911 		case PROP_TYPE_INDEX:
2912 			VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
2913 			break;
2914 		default:
2915 			nvlist_free(dummy);
2916 			return (SET_ERROR(EINVAL));
2917 		}
2918 
2919 		pair = nvlist_next_nvpair(dummy, NULL);
2920 		err = zfs_prop_set_special(zc->zc_name, source, pair);
2921 		nvlist_free(dummy);
2922 		if (err != -1)
2923 			return (err); /* special property already handled */
2924 	} else {
2925 		/*
2926 		 * Only check this in the non-received case. We want to allow
2927 		 * 'inherit -S' to revert non-inheritable properties like quota
2928 		 * and reservation to the received or default values even though
2929 		 * they are not considered inheritable.
2930 		 */
2931 		if (prop != ZPROP_INVAL && !zfs_prop_inheritable(prop))
2932 			return (SET_ERROR(EINVAL));
2933 	}
2934 
2935 	/* property name has been validated by zfs_secpolicy_inherit_prop() */
2936 	return (dsl_prop_inherit(zc->zc_name, zc->zc_value, source));
2937 }
2938 
2939 static int
2940 zfs_ioc_pool_set_props(zfs_cmd_t *zc)
2941 {
2942 	nvlist_t *props;
2943 	spa_t *spa;
2944 	int error;
2945 	nvpair_t *pair;
2946 
2947 	if (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
2948 	    zc->zc_iflags, &props))
2949 		return (error);
2950 
2951 	/*
2952 	 * If the only property is the configfile, then just do a spa_lookup()
2953 	 * to handle the faulted case.
2954 	 */
2955 	pair = nvlist_next_nvpair(props, NULL);
2956 	if (pair != NULL && strcmp(nvpair_name(pair),
2957 	    zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
2958 	    nvlist_next_nvpair(props, pair) == NULL) {
2959 		mutex_enter(&spa_namespace_lock);
2960 		if ((spa = spa_lookup(zc->zc_name)) != NULL) {
2961 			spa_configfile_set(spa, props, B_FALSE);
2962 			spa_write_cachefile(spa, B_FALSE, B_TRUE);
2963 		}
2964 		mutex_exit(&spa_namespace_lock);
2965 		if (spa != NULL) {
2966 			nvlist_free(props);
2967 			return (0);
2968 		}
2969 	}
2970 
2971 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2972 		nvlist_free(props);
2973 		return (error);
2974 	}
2975 
2976 	error = spa_prop_set(spa, props);
2977 
2978 	nvlist_free(props);
2979 	spa_close(spa, FTAG);
2980 
2981 	return (error);
2982 }
2983 
2984 static int
2985 zfs_ioc_pool_get_props(zfs_cmd_t *zc)
2986 {
2987 	spa_t *spa;
2988 	int error;
2989 	nvlist_t *nvp = NULL;
2990 
2991 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
2992 		/*
2993 		 * If the pool is faulted, there may be properties we can still
2994 		 * get (such as altroot and cachefile), so attempt to get them
2995 		 * anyway.
2996 		 */
2997 		mutex_enter(&spa_namespace_lock);
2998 		if ((spa = spa_lookup(zc->zc_name)) != NULL)
2999 			error = spa_prop_get(spa, &nvp);
3000 		mutex_exit(&spa_namespace_lock);
3001 	} else {
3002 		error = spa_prop_get(spa, &nvp);
3003 		spa_close(spa, FTAG);
3004 	}
3005 
3006 	if (error == 0 && zc->zc_nvlist_dst != 0)
3007 		error = put_nvlist(zc, nvp);
3008 	else
3009 		error = SET_ERROR(EFAULT);
3010 
3011 	nvlist_free(nvp);
3012 	return (error);
3013 }
3014 
3015 /*
3016  * inputs:
3017  * zc_name		name of filesystem
3018  * zc_nvlist_src{_size}	nvlist of delegated permissions
3019  * zc_perm_action	allow/unallow flag
3020  *
3021  * outputs:		none
3022  */
3023 static int
3024 zfs_ioc_set_fsacl(zfs_cmd_t *zc)
3025 {
3026 	int error;
3027 	nvlist_t *fsaclnv = NULL;
3028 
3029 	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
3030 	    zc->zc_iflags, &fsaclnv)) != 0)
3031 		return (error);
3032 
3033 	/*
3034 	 * Verify nvlist is constructed correctly
3035 	 */
3036 	if ((error = zfs_deleg_verify_nvlist(fsaclnv)) != 0) {
3037 		nvlist_free(fsaclnv);
3038 		return (SET_ERROR(EINVAL));
3039 	}
3040 
3041 	/*
3042 	 * If we don't have PRIV_SYS_MOUNT, then validate
3043 	 * that user is allowed to hand out each permission in
3044 	 * the nvlist(s)
3045 	 */
3046 
3047 	error = secpolicy_zfs(CRED());
3048 	if (error != 0) {
3049 		if (zc->zc_perm_action == B_FALSE) {
3050 			error = dsl_deleg_can_allow(zc->zc_name,
3051 			    fsaclnv, CRED());
3052 		} else {
3053 			error = dsl_deleg_can_unallow(zc->zc_name,
3054 			    fsaclnv, CRED());
3055 		}
3056 	}
3057 
3058 	if (error == 0)
3059 		error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);
3060 
3061 	nvlist_free(fsaclnv);
3062 	return (error);
3063 }
3064 
3065 /*
3066  * inputs:
3067  * zc_name		name of filesystem
3068  *
3069  * outputs:
3070  * zc_nvlist_src{_size}	nvlist of delegated permissions
3071  */
3072 static int
3073 zfs_ioc_get_fsacl(zfs_cmd_t *zc)
3074 {
3075 	nvlist_t *nvp;
3076 	int error;
3077 
3078 	if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
3079 		error = put_nvlist(zc, nvp);
3080 		nvlist_free(nvp);
3081 	}
3082 
3083 	return (error);
3084 }
3085 
3086 /* ARGSUSED */
3087 static void
3088 zfs_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
3089 {
3090 	zfs_creat_t *zct = arg;
3091 
3092 	zfs_create_fs(os, cr, zct->zct_zplprops, tx);
3093 }
3094 
3095 #define	ZFS_PROP_UNDEFINED	((uint64_t)-1)
3096 
3097 /*
3098  * inputs:
3099  * os			parent objset pointer (NULL if root fs)
3100  * fuids_ok		fuids allowed in this version of the spa?
3101  * sa_ok		SAs allowed in this version of the spa?
3102  * createprops		list of properties requested by creator
3103  *
3104  * outputs:
3105  * zplprops	values for the zplprops we attach to the master node object
3106  * is_ci	true if requested file system will be purely case-insensitive
3107  *
3108  * Determine the settings for utf8only, normalization and
3109  * casesensitivity.  Specific values may have been requested by the
3110  * creator and/or we can inherit values from the parent dataset.  If
3111  * the file system is of too early a vintage, a creator can not
3112  * request settings for these properties, even if the requested
3113  * setting is the default value.  We don't actually want to create dsl
3114  * properties for these, so remove them from the source nvlist after
3115  * processing.
3116  */
3117 static int
3118 zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
3119     boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
3120     nvlist_t *zplprops, boolean_t *is_ci)
3121 {
3122 	uint64_t sense = ZFS_PROP_UNDEFINED;
3123 	uint64_t norm = ZFS_PROP_UNDEFINED;
3124 	uint64_t u8 = ZFS_PROP_UNDEFINED;
3125 
3126 	ASSERT(zplprops != NULL);
3127 
3128 	if (os != NULL && os->os_phys->os_type != DMU_OST_ZFS)
3129 		return (SET_ERROR(EINVAL));
3130 
3131 	/*
3132 	 * Pull out creator prop choices, if any.
3133 	 */
3134 	if (createprops) {
3135 		(void) nvlist_lookup_uint64(createprops,
3136 		    zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
3137 		(void) nvlist_lookup_uint64(createprops,
3138 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
3139 		(void) nvlist_remove_all(createprops,
3140 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE));
3141 		(void) nvlist_lookup_uint64(createprops,
3142 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
3143 		(void) nvlist_remove_all(createprops,
3144 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
3145 		(void) nvlist_lookup_uint64(createprops,
3146 		    zfs_prop_to_name(ZFS_PROP_CASE), &sense);
3147 		(void) nvlist_remove_all(createprops,
3148 		    zfs_prop_to_name(ZFS_PROP_CASE));
3149 	}
3150 
3151 	/*
3152 	 * If the zpl version requested is whacky or the file system
3153 	 * or pool is version is too "young" to support normalization
3154 	 * and the creator tried to set a value for one of the props,
3155 	 * error out.
3156 	 */
3157 	if ((zplver < ZPL_VERSION_INITIAL || zplver > ZPL_VERSION) ||
3158 	    (zplver >= ZPL_VERSION_FUID && !fuids_ok) ||
3159 	    (zplver >= ZPL_VERSION_SA && !sa_ok) ||
3160 	    (zplver < ZPL_VERSION_NORMALIZATION &&
3161 	    (norm != ZFS_PROP_UNDEFINED || u8 != ZFS_PROP_UNDEFINED ||
3162 	    sense != ZFS_PROP_UNDEFINED)))
3163 		return (SET_ERROR(ENOTSUP));
3164 
3165 	/*
3166 	 * Put the version in the zplprops
3167 	 */
3168 	VERIFY(nvlist_add_uint64(zplprops,
3169 	    zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);
3170 
3171 	if (norm == ZFS_PROP_UNDEFINED)
3172 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm) == 0);
3173 	VERIFY(nvlist_add_uint64(zplprops,
3174 	    zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);
3175 
3176 	/*
3177 	 * If we're normalizing, names must always be valid UTF-8 strings.
3178 	 */
3179 	if (norm)
3180 		u8 = 1;
3181 	if (u8 == ZFS_PROP_UNDEFINED)
3182 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8) == 0);
3183 	VERIFY(nvlist_add_uint64(zplprops,
3184 	    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);
3185 
3186 	if (sense == ZFS_PROP_UNDEFINED)
3187 		VERIFY(zfs_get_zplprop(os, ZFS_PROP_CASE, &sense) == 0);
3188 	VERIFY(nvlist_add_uint64(zplprops,
3189 	    zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);
3190 
3191 	if (is_ci)
3192 		*is_ci = (sense == ZFS_CASE_INSENSITIVE);
3193 
3194 	return (0);
3195 }
3196 
3197 static int
3198 zfs_fill_zplprops(const char *dataset, nvlist_t *createprops,
3199     nvlist_t *zplprops, boolean_t *is_ci)
3200 {
3201 	boolean_t fuids_ok, sa_ok;
3202 	uint64_t zplver = ZPL_VERSION;
3203 	objset_t *os = NULL;
3204 	char parentname[ZFS_MAX_DATASET_NAME_LEN];
3205 	char *cp;
3206 	spa_t *spa;
3207 	uint64_t spa_vers;
3208 	int error;
3209 
3210 	(void) strlcpy(parentname, dataset, sizeof (parentname));
3211 	cp = strrchr(parentname, '/');
3212 	ASSERT(cp != NULL);
3213 	cp[0] = '\0';
3214 
3215 	if ((error = spa_open(dataset, &spa, FTAG)) != 0)
3216 		return (error);
3217 
3218 	spa_vers = spa_version(spa);
3219 	spa_close(spa, FTAG);
3220 
3221 	zplver = zfs_zpl_version_map(spa_vers);
3222 	fuids_ok = (zplver >= ZPL_VERSION_FUID);
3223 	sa_ok = (zplver >= ZPL_VERSION_SA);
3224 
3225 	/*
3226 	 * Open parent object set so we can inherit zplprop values.
3227 	 */
3228 	if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
3229 		return (error);
3230 
3231 	error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
3232 	    zplprops, is_ci);
3233 	dmu_objset_rele(os, FTAG);
3234 	return (error);
3235 }
3236 
3237 static int
3238 zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
3239     nvlist_t *zplprops, boolean_t *is_ci)
3240 {
3241 	boolean_t fuids_ok;
3242 	boolean_t sa_ok;
3243 	uint64_t zplver = ZPL_VERSION;
3244 	int error;
3245 
3246 	zplver = zfs_zpl_version_map(spa_vers);
3247 	fuids_ok = (zplver >= ZPL_VERSION_FUID);
3248 	sa_ok = (zplver >= ZPL_VERSION_SA);
3249 
3250 	error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
3251 	    createprops, zplprops, is_ci);
3252 	return (error);
3253 }
3254 
3255 /*
3256  * innvl: {
3257  *     "type" -> dmu_objset_type_t (int32)
3258  *     (optional) "props" -> { prop -> value }
3259  *     (optional) "hidden_args" -> { "wkeydata" -> value }
3260  *         raw uint8_t array of encryption wrapping key data (32 bytes)
3261  * }
3262  *
3263  * outnvl: propname -> error code (int32)
3264  */
3265 static int
3266 zfs_ioc_create(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3267 {
3268 	int error = 0;
3269 	zfs_creat_t zct = { 0 };
3270 	nvlist_t *nvprops = NULL;
3271 	nvlist_t *hidden_args = NULL;
3272 	void (*cbfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx);
3273 	int32_t type32;
3274 	dmu_objset_type_t type;
3275 	boolean_t is_insensitive = B_FALSE;
3276 	dsl_crypto_params_t *dcp = NULL;
3277 
3278 	if (nvlist_lookup_int32(innvl, "type", &type32) != 0)
3279 		return (SET_ERROR(EINVAL));
3280 	type = type32;
3281 	(void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3282 	(void) nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);
3283 
3284 	switch (type) {
3285 	case DMU_OST_ZFS:
3286 		cbfunc = zfs_create_cb;
3287 		break;
3288 
3289 	case DMU_OST_ZVOL:
3290 		cbfunc = zvol_create_cb;
3291 		break;
3292 
3293 	default:
3294 		cbfunc = NULL;
3295 		break;
3296 	}
3297 	if (strchr(fsname, '@') ||
3298 	    strchr(fsname, '%'))
3299 		return (SET_ERROR(EINVAL));
3300 
3301 	zct.zct_props = nvprops;
3302 
3303 	if (cbfunc == NULL)
3304 		return (SET_ERROR(EINVAL));
3305 
3306 	if (type == DMU_OST_ZVOL) {
3307 		uint64_t volsize, volblocksize;
3308 
3309 		if (nvprops == NULL)
3310 			return (SET_ERROR(EINVAL));
3311 		if (nvlist_lookup_uint64(nvprops,
3312 		    zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
3313 			return (SET_ERROR(EINVAL));
3314 
3315 		if ((error = nvlist_lookup_uint64(nvprops,
3316 		    zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3317 		    &volblocksize)) != 0 && error != ENOENT)
3318 			return (SET_ERROR(EINVAL));
3319 
3320 		if (error != 0)
3321 			volblocksize = zfs_prop_default_numeric(
3322 			    ZFS_PROP_VOLBLOCKSIZE);
3323 
3324 		if ((error = zvol_check_volblocksize(
3325 		    volblocksize)) != 0 ||
3326 		    (error = zvol_check_volsize(volsize,
3327 		    volblocksize)) != 0)
3328 			return (error);
3329 	} else if (type == DMU_OST_ZFS) {
3330 		int error;
3331 
3332 		/*
3333 		 * We have to have normalization and
3334 		 * case-folding flags correct when we do the
3335 		 * file system creation, so go figure them out
3336 		 * now.
3337 		 */
3338 		VERIFY(nvlist_alloc(&zct.zct_zplprops,
3339 		    NV_UNIQUE_NAME, KM_SLEEP) == 0);
3340 		error = zfs_fill_zplprops(fsname, nvprops,
3341 		    zct.zct_zplprops, &is_insensitive);
3342 		if (error != 0) {
3343 			nvlist_free(zct.zct_zplprops);
3344 			return (error);
3345 		}
3346 	}
3347 
3348 	error = dsl_crypto_params_create_nvlist(DCP_CMD_NONE, nvprops,
3349 	    hidden_args, &dcp);
3350 	if (error != 0) {
3351 		nvlist_free(zct.zct_zplprops);
3352 		return (error);
3353 	}
3354 
3355 	error = dmu_objset_create(fsname, type,
3356 	    is_insensitive ? DS_FLAG_CI_DATASET : 0, dcp, cbfunc, &zct);
3357 
3358 	nvlist_free(zct.zct_zplprops);
3359 	dsl_crypto_params_free(dcp, !!error);
3360 
3361 	/*
3362 	 * It would be nice to do this atomically.
3363 	 */
3364 	if (error == 0) {
3365 		error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3366 		    nvprops, outnvl);
3367 		if (error != 0)
3368 			(void) dsl_destroy_head(fsname);
3369 	}
3370 	return (error);
3371 }
3372 
3373 /*
3374  * innvl: {
3375  *     "origin" -> name of origin snapshot
3376  *     (optional) "props" -> { prop -> value }
3377  *     (optional) "hidden_args" -> { "wkeydata" -> value }
3378  *         raw uint8_t array of encryption wrapping key data (32 bytes)
3379  * }
3380  *
3381  * outnvl: propname -> error code (int32)
3382  */
3383 static int
3384 zfs_ioc_clone(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3385 {
3386 	int error = 0;
3387 	nvlist_t *nvprops = NULL;
3388 	char *origin_name;
3389 
3390 	if (nvlist_lookup_string(innvl, "origin", &origin_name) != 0)
3391 		return (SET_ERROR(EINVAL));
3392 	(void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
3393 
3394 	if (strchr(fsname, '@') ||
3395 	    strchr(fsname, '%'))
3396 		return (SET_ERROR(EINVAL));
3397 
3398 	if (dataset_namecheck(origin_name, NULL, NULL) != 0)
3399 		return (SET_ERROR(EINVAL));
3400 
3401 	error = dmu_objset_clone(fsname, origin_name);
3402 
3403 	/*
3404 	 * It would be nice to do this atomically.
3405 	 */
3406 	if (error == 0) {
3407 		error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
3408 		    nvprops, outnvl);
3409 		if (error != 0)
3410 			(void) dsl_destroy_head(fsname);
3411 	}
3412 	return (error);
3413 }
3414 
3415 /* ARGSUSED */
3416 static int
3417 zfs_ioc_remap(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3418 {
3419 	if (strchr(fsname, '@') ||
3420 	    strchr(fsname, '%'))
3421 		return (SET_ERROR(EINVAL));
3422 
3423 	return (dmu_objset_remap_indirects(fsname));
3424 }
3425 
3426 /*
3427  * innvl: {
3428  *     "snaps" -> { snapshot1, snapshot2 }
3429  *     (optional) "props" -> { prop -> value (string) }
3430  * }
3431  *
3432  * outnvl: snapshot -> error code (int32)
3433  */
3434 static int
3435 zfs_ioc_snapshot(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3436 {
3437 	nvlist_t *snaps;
3438 	nvlist_t *props = NULL;
3439 	int error, poollen;
3440 	nvpair_t *pair;
3441 
3442 	(void) nvlist_lookup_nvlist(innvl, "props", &props);
3443 	if ((error = zfs_check_userprops(poolname, props)) != 0)
3444 		return (error);
3445 
3446 	if (!nvlist_empty(props) &&
3447 	    zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
3448 		return (SET_ERROR(ENOTSUP));
3449 
3450 	if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3451 		return (SET_ERROR(EINVAL));
3452 	poollen = strlen(poolname);
3453 	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3454 	    pair = nvlist_next_nvpair(snaps, pair)) {
3455 		const char *name = nvpair_name(pair);
3456 		const char *cp = strchr(name, '@');
3457 
3458 		/*
3459 		 * The snap name must contain an @, and the part after it must
3460 		 * contain only valid characters.
3461 		 */
3462 		if (cp == NULL ||
3463 		    zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3464 			return (SET_ERROR(EINVAL));
3465 
3466 		/*
3467 		 * The snap must be in the specified pool.
3468 		 */
3469 		if (strncmp(name, poolname, poollen) != 0 ||
3470 		    (name[poollen] != '/' && name[poollen] != '@'))
3471 			return (SET_ERROR(EXDEV));
3472 
3473 		/* This must be the only snap of this fs. */
3474 		for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
3475 		    pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
3476 			if (strncmp(name, nvpair_name(pair2), cp - name + 1)
3477 			    == 0) {
3478 				return (SET_ERROR(EXDEV));
3479 			}
3480 		}
3481 	}
3482 
3483 	error = dsl_dataset_snapshot(snaps, props, outnvl);
3484 	return (error);
3485 }
3486 
3487 /*
3488  * innvl: "message" -> string
3489  */
3490 /* ARGSUSED */
3491 static int
3492 zfs_ioc_log_history(const char *unused, nvlist_t *innvl, nvlist_t *outnvl)
3493 {
3494 	char *message;
3495 	spa_t *spa;
3496 	int error;
3497 	char *poolname;
3498 
3499 	/*
3500 	 * The poolname in the ioctl is not set, we get it from the TSD,
3501 	 * which was set at the end of the last successful ioctl that allows
3502 	 * logging.  The secpolicy func already checked that it is set.
3503 	 * Only one log ioctl is allowed after each successful ioctl, so
3504 	 * we clear the TSD here.
3505 	 */
3506 	poolname = tsd_get(zfs_allow_log_key);
3507 	(void) tsd_set(zfs_allow_log_key, NULL);
3508 	error = spa_open(poolname, &spa, FTAG);
3509 	strfree(poolname);
3510 	if (error != 0)
3511 		return (error);
3512 
3513 	if (nvlist_lookup_string(innvl, "message", &message) != 0)  {
3514 		spa_close(spa, FTAG);
3515 		return (SET_ERROR(EINVAL));
3516 	}
3517 
3518 	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
3519 		spa_close(spa, FTAG);
3520 		return (SET_ERROR(ENOTSUP));
3521 	}
3522 
3523 	error = spa_history_log(spa, message);
3524 	spa_close(spa, FTAG);
3525 	return (error);
3526 }
3527 
3528 /*
3529  * The dp_config_rwlock must not be held when calling this, because the
3530  * unmount may need to write out data.
3531  *
3532  * This function is best-effort.  Callers must deal gracefully if it
3533  * remains mounted (or is remounted after this call).
3534  *
3535  * Returns 0 if the argument is not a snapshot, or it is not currently a
3536  * filesystem, or we were able to unmount it.  Returns error code otherwise.
3537  */
3538 void
3539 zfs_unmount_snap(const char *snapname)
3540 {
3541 	vfs_t *vfsp = NULL;
3542 	zfsvfs_t *zfsvfs = NULL;
3543 
3544 	if (strchr(snapname, '@') == NULL)
3545 		return;
3546 
3547 	int err = getzfsvfs(snapname, &zfsvfs);
3548 	if (err != 0) {
3549 		ASSERT3P(zfsvfs, ==, NULL);
3550 		return;
3551 	}
3552 	vfsp = zfsvfs->z_vfs;
3553 
3554 	ASSERT(!dsl_pool_config_held(dmu_objset_pool(zfsvfs->z_os)));
3555 
3556 	err = vn_vfswlock(vfsp->vfs_vnodecovered);
3557 	VFS_RELE(vfsp);
3558 	if (err != 0)
3559 		return;
3560 
3561 	/*
3562 	 * Always force the unmount for snapshots.
3563 	 */
3564 	(void) dounmount(vfsp, MS_FORCE, kcred);
3565 }
3566 
3567 /* ARGSUSED */
3568 static int
3569 zfs_unmount_snap_cb(const char *snapname, void *arg)
3570 {
3571 	zfs_unmount_snap(snapname);
3572 	return (0);
3573 }
3574 
3575 /*
3576  * When a clone is destroyed, its origin may also need to be destroyed,
3577  * in which case it must be unmounted.  This routine will do that unmount
3578  * if necessary.
3579  */
3580 void
3581 zfs_destroy_unmount_origin(const char *fsname)
3582 {
3583 	int error;
3584 	objset_t *os;
3585 	dsl_dataset_t *ds;
3586 
3587 	error = dmu_objset_hold(fsname, FTAG, &os);
3588 	if (error != 0)
3589 		return;
3590 	ds = dmu_objset_ds(os);
3591 	if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
3592 		char originname[ZFS_MAX_DATASET_NAME_LEN];
3593 		dsl_dataset_name(ds->ds_prev, originname);
3594 		dmu_objset_rele(os, FTAG);
3595 		zfs_unmount_snap(originname);
3596 	} else {
3597 		dmu_objset_rele(os, FTAG);
3598 	}
3599 }
3600 
3601 /*
3602  * innvl: {
3603  *     "snaps" -> { snapshot1, snapshot2 }
3604  *     (optional boolean) "defer"
3605  * }
3606  *
3607  * outnvl: snapshot -> error code (int32)
3608  *
3609  */
3610 /* ARGSUSED */
3611 static int
3612 zfs_ioc_destroy_snaps(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3613 {
3614 	nvlist_t *snaps;
3615 	nvpair_t *pair;
3616 	boolean_t defer;
3617 
3618 	if (nvlist_lookup_nvlist(innvl, "snaps", &snaps) != 0)
3619 		return (SET_ERROR(EINVAL));
3620 	defer = nvlist_exists(innvl, "defer");
3621 
3622 	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
3623 	    pair = nvlist_next_nvpair(snaps, pair)) {
3624 		zfs_unmount_snap(nvpair_name(pair));
3625 	}
3626 
3627 	return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
3628 }
3629 
3630 /*
3631  * Create bookmarks.  Bookmark names are of the form <fs>#<bmark>.
3632  * All bookmarks must be in the same pool.
3633  *
3634  * innvl: {
3635  *     bookmark1 -> snapshot1, bookmark2 -> snapshot2
3636  * }
3637  *
3638  * outnvl: bookmark -> error code (int32)
3639  *
3640  */
3641 /* ARGSUSED */
3642 static int
3643 zfs_ioc_bookmark(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3644 {
3645 	for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3646 	    pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3647 		char *snap_name;
3648 
3649 		/*
3650 		 * Verify the snapshot argument.
3651 		 */
3652 		if (nvpair_value_string(pair, &snap_name) != 0)
3653 			return (SET_ERROR(EINVAL));
3654 
3655 
3656 		/* Verify that the keys (bookmarks) are unique */
3657 		for (nvpair_t *pair2 = nvlist_next_nvpair(innvl, pair);
3658 		    pair2 != NULL; pair2 = nvlist_next_nvpair(innvl, pair2)) {
3659 			if (strcmp(nvpair_name(pair), nvpair_name(pair2)) == 0)
3660 				return (SET_ERROR(EINVAL));
3661 		}
3662 	}
3663 
3664 	return (dsl_bookmark_create(innvl, outnvl));
3665 }
3666 
3667 /*
3668  * innvl: {
3669  *     property 1, property 2, ...
3670  * }
3671  *
3672  * outnvl: {
3673  *     bookmark name 1 -> { property 1, property 2, ... },
3674  *     bookmark name 2 -> { property 1, property 2, ... }
3675  * }
3676  *
3677  */
3678 static int
3679 zfs_ioc_get_bookmarks(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
3680 {
3681 	return (dsl_get_bookmarks(fsname, innvl, outnvl));
3682 }
3683 
3684 /*
3685  * innvl: {
3686  *     bookmark name 1, bookmark name 2
3687  * }
3688  *
3689  * outnvl: bookmark -> error code (int32)
3690  *
3691  */
3692 static int
3693 zfs_ioc_destroy_bookmarks(const char *poolname, nvlist_t *innvl,
3694     nvlist_t *outnvl)
3695 {
3696 	int error, poollen;
3697 
3698 	poollen = strlen(poolname);
3699 	for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
3700 	    pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
3701 		const char *name = nvpair_name(pair);
3702 		const char *cp = strchr(name, '#');
3703 
3704 		/*
3705 		 * The bookmark name must contain an #, and the part after it
3706 		 * must contain only valid characters.
3707 		 */
3708 		if (cp == NULL ||
3709 		    zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
3710 			return (SET_ERROR(EINVAL));
3711 
3712 		/*
3713 		 * The bookmark must be in the specified pool.
3714 		 */
3715 		if (strncmp(name, poolname, poollen) != 0 ||
3716 		    (name[poollen] != '/' && name[poollen] != '#'))
3717 			return (SET_ERROR(EXDEV));
3718 	}
3719 
3720 	error = dsl_bookmark_destroy(innvl, outnvl);
3721 	return (error);
3722 }
3723 
3724 static int
3725 zfs_ioc_channel_program(const char *poolname, nvlist_t *innvl,
3726     nvlist_t *outnvl)
3727 {
3728 	char *program;
3729 	uint64_t instrlimit, memlimit;
3730 	boolean_t sync_flag;
3731 	nvpair_t *nvarg = NULL;
3732 
3733 	if (0 != nvlist_lookup_string(innvl, ZCP_ARG_PROGRAM, &program)) {
3734 		return (EINVAL);
3735 	}
3736 	if (0 != nvlist_lookup_boolean_value(innvl, ZCP_ARG_SYNC, &sync_flag)) {
3737 		sync_flag = B_TRUE;
3738 	}
3739 	if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_INSTRLIMIT, &instrlimit)) {
3740 		instrlimit = ZCP_DEFAULT_INSTRLIMIT;
3741 	}
3742 	if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_MEMLIMIT, &memlimit)) {
3743 		memlimit = ZCP_DEFAULT_MEMLIMIT;
3744 	}
3745 	if (0 != nvlist_lookup_nvpair(innvl, ZCP_ARG_ARGLIST, &nvarg)) {
3746 		return (EINVAL);
3747 	}
3748 
3749 	if (instrlimit == 0 || instrlimit > zfs_lua_max_instrlimit)
3750 		return (EINVAL);
3751 	if (memlimit == 0 || memlimit > zfs_lua_max_memlimit)
3752 		return (EINVAL);
3753 
3754 	return (zcp_eval(poolname, program, sync_flag, instrlimit, memlimit,
3755 	    nvarg, outnvl));
3756 }
3757 
3758 /*
3759  * innvl: unused
3760  * outnvl: empty
3761  */
3762 /* ARGSUSED */
3763 static int
3764 zfs_ioc_pool_checkpoint(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3765 {
3766 	return (spa_checkpoint(poolname));
3767 }
3768 
3769 /*
3770  * innvl: unused
3771  * outnvl: empty
3772  */
3773 /* ARGSUSED */
3774 static int
3775 zfs_ioc_pool_discard_checkpoint(const char *poolname, nvlist_t *innvl,
3776     nvlist_t *outnvl)
3777 {
3778 	return (spa_checkpoint_discard(poolname));
3779 }
3780 
3781 /*
3782  * inputs:
3783  * zc_name		name of dataset to destroy
3784  * zc_defer_destroy	mark for deferred destroy
3785  *
3786  * outputs:		none
3787  */
3788 static int
3789 zfs_ioc_destroy(zfs_cmd_t *zc)
3790 {
3791 	objset_t *os;
3792 	dmu_objset_type_t ost;
3793 	int err;
3794 
3795 	err = dmu_objset_hold(zc->zc_name, FTAG, &os);
3796 	if (err != 0)
3797 		return (err);
3798 	ost = dmu_objset_type(os);
3799 	dmu_objset_rele(os, FTAG);
3800 
3801 	if (ost == DMU_OST_ZFS)
3802 		zfs_unmount_snap(zc->zc_name);
3803 
3804 	if (strchr(zc->zc_name, '@')) {
3805 		err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
3806 	} else {
3807 		err = dsl_destroy_head(zc->zc_name);
3808 		if (err == EEXIST) {
3809 			/*
3810 			 * It is possible that the given DS may have
3811 			 * hidden child (%recv) datasets - "leftovers"
3812 			 * resulting from the previously interrupted
3813 			 * 'zfs receive'.
3814 			 *
3815 			 * 6 extra bytes for /%recv
3816 			 */
3817 			char namebuf[ZFS_MAX_DATASET_NAME_LEN + 6];
3818 
3819 			if (snprintf(namebuf, sizeof (namebuf), "%s/%s",
3820 			    zc->zc_name, recv_clone_name) >=
3821 			    sizeof (namebuf))
3822 				return (SET_ERROR(EINVAL));
3823 
3824 			/*
3825 			 * Try to remove the hidden child (%recv) and after
3826 			 * that try to remove the target dataset.
3827 			 * If the hidden child (%recv) does not exist
3828 			 * the original error (EEXIST) will be returned
3829 			 */
3830 			err = dsl_destroy_head(namebuf);
3831 			if (err == 0)
3832 				err = dsl_destroy_head(zc->zc_name);
3833 			else if (err == ENOENT)
3834 				err = SET_ERROR(EEXIST);
3835 		}
3836 	}
3837 	if (ost == DMU_OST_ZVOL && err == 0)
3838 		(void) zvol_remove_minor(zc->zc_name);
3839 	return (err);
3840 }
3841 
3842 /*
3843  * innvl: {
3844  *     "initialize_command" -> POOL_INITIALIZE_{CANCEL|START|SUSPEND} (uint64)
3845  *     "initialize_vdevs": { -> guids to initialize (nvlist)
3846  *         "vdev_path_1": vdev_guid_1, (uint64),
3847  *         "vdev_path_2": vdev_guid_2, (uint64),
3848  *         ...
3849  *     },
3850  * }
3851  *
3852  * outnvl: {
3853  *     "initialize_vdevs": { -> initialization errors (nvlist)
3854  *         "vdev_path_1": errno, see function body for possible errnos (uint64)
3855  *         "vdev_path_2": errno, ... (uint64)
3856  *         ...
3857  *     }
3858  * }
3859  *
3860  * EINVAL is returned for an unknown command or if any of the provided vdev
3861  * guids have be specified with a type other than uint64.
3862  */
3863 static int
3864 zfs_ioc_pool_initialize(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3865 {
3866 	uint64_t cmd_type;
3867 	if (nvlist_lookup_uint64(innvl, ZPOOL_INITIALIZE_COMMAND,
3868 	    &cmd_type) != 0) {
3869 		return (SET_ERROR(EINVAL));
3870 	}
3871 
3872 	if (!(cmd_type == POOL_INITIALIZE_CANCEL ||
3873 	    cmd_type == POOL_INITIALIZE_START ||
3874 	    cmd_type == POOL_INITIALIZE_SUSPEND)) {
3875 		return (SET_ERROR(EINVAL));
3876 	}
3877 
3878 	nvlist_t *vdev_guids;
3879 	if (nvlist_lookup_nvlist(innvl, ZPOOL_INITIALIZE_VDEVS,
3880 	    &vdev_guids) != 0) {
3881 		return (SET_ERROR(EINVAL));
3882 	}
3883 
3884 	for (nvpair_t *pair = nvlist_next_nvpair(vdev_guids, NULL);
3885 	    pair != NULL; pair = nvlist_next_nvpair(vdev_guids, pair)) {
3886 		uint64_t vdev_guid;
3887 		if (nvpair_value_uint64(pair, &vdev_guid) != 0) {
3888 			return (SET_ERROR(EINVAL));
3889 		}
3890 	}
3891 
3892 	spa_t *spa;
3893 	int error = spa_open(poolname, &spa, FTAG);
3894 	if (error != 0)
3895 		return (error);
3896 
3897 	nvlist_t *vdev_errlist = fnvlist_alloc();
3898 	int total_errors = spa_vdev_initialize(spa, vdev_guids, cmd_type,
3899 	    vdev_errlist);
3900 
3901 	if (fnvlist_size(vdev_errlist) > 0) {
3902 		fnvlist_add_nvlist(outnvl, ZPOOL_INITIALIZE_VDEVS,
3903 		    vdev_errlist);
3904 	}
3905 	fnvlist_free(vdev_errlist);
3906 
3907 	spa_close(spa, FTAG);
3908 	return (total_errors > 0 ? EINVAL : 0);
3909 }
3910 
3911 /*
3912  * innvl: {
3913  *     "trim_command" -> POOL_TRIM_{CANCEL|START|SUSPEND} (uint64)
3914  *     "trim_vdevs": { -> guids to TRIM (nvlist)
3915  *         "vdev_path_1": vdev_guid_1, (uint64),
3916  *         "vdev_path_2": vdev_guid_2, (uint64),
3917  *         ...
3918  *     },
3919  *     "trim_rate" -> Target TRIM rate in bytes/sec.
3920  *     "trim_secure" -> Set to request a secure TRIM.
3921  * }
3922  *
3923  * outnvl: {
3924  *     "trim_vdevs": { -> TRIM errors (nvlist)
3925  *         "vdev_path_1": errno, see function body for possible errnos (uint64)
3926  *         "vdev_path_2": errno, ... (uint64)
3927  *         ...
3928  *     }
3929  * }
3930  *
3931  * EINVAL is returned for an unknown command or if any of the provided vdev
3932  * guids have be specified with a type other than uint64.
3933  */
3934 
3935 static int
3936 zfs_ioc_pool_trim(const char *poolname, nvlist_t *innvl, nvlist_t *outnvl)
3937 {
3938 	uint64_t cmd_type;
3939 	if (nvlist_lookup_uint64(innvl, ZPOOL_TRIM_COMMAND, &cmd_type) != 0)
3940 		return (SET_ERROR(EINVAL));
3941 
3942 	if (!(cmd_type == POOL_TRIM_CANCEL ||
3943 	    cmd_type == POOL_TRIM_START ||
3944 	    cmd_type == POOL_TRIM_SUSPEND)) {
3945 		return (SET_ERROR(EINVAL));
3946 	}
3947 
3948 	nvlist_t *vdev_guids;
3949 	if (nvlist_lookup_nvlist(innvl, ZPOOL_TRIM_VDEVS, &vdev_guids) != 0)
3950 		return (SET_ERROR(EINVAL));
3951 
3952 	for (nvpair_t *pair = nvlist_next_nvpair(vdev_guids, NULL);
3953 	    pair != NULL; pair = nvlist_next_nvpair(vdev_guids, pair)) {
3954 		uint64_t vdev_guid;
3955 		if (nvpair_value_uint64(pair, &vdev_guid) != 0) {
3956 			return (SET_ERROR(EINVAL));
3957 		}
3958 	}
3959 
3960 	/* Optional, defaults to maximum rate when not provided */
3961 	uint64_t rate;
3962 	if (nvlist_lookup_uint64(innvl, ZPOOL_TRIM_RATE, &rate) != 0)
3963 		rate = 0;
3964 
3965 	/* Optional, defaults to standard TRIM when not provided */
3966 	boolean_t secure;
3967 	if (nvlist_lookup_boolean_value(innvl, ZPOOL_TRIM_SECURE,
3968 	    &secure) != 0) {
3969 		secure = B_FALSE;
3970 	}
3971 
3972 	spa_t *spa;
3973 	int error = spa_open(poolname, &spa, FTAG);
3974 	if (error != 0)
3975 		return (error);
3976 
3977 	nvlist_t *vdev_errlist = fnvlist_alloc();
3978 	int total_errors = spa_vdev_trim(spa, vdev_guids, cmd_type,
3979 	    rate, !!zfs_trim_metaslab_skip, secure, vdev_errlist);
3980 
3981 	if (fnvlist_size(vdev_errlist) > 0)
3982 		fnvlist_add_nvlist(outnvl, ZPOOL_TRIM_VDEVS, vdev_errlist);
3983 
3984 	fnvlist_free(vdev_errlist);
3985 
3986 	spa_close(spa, FTAG);
3987 	return (total_errors > 0 ? EINVAL : 0);
3988 }
3989 
3990 /*
3991  * fsname is name of dataset to rollback (to most recent snapshot)
3992  *
3993  * innvl may contain name of expected target snapshot
3994  *
3995  * outnvl: "target" -> name of most recent snapshot
3996  * }
3997  */
3998 /* ARGSUSED */
3999 static int
4000 zfs_ioc_rollback(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl)
4001 {
4002 	zfsvfs_t *zfsvfs;
4003 	char *target = NULL;
4004 	int error;
4005 
4006 	(void) nvlist_lookup_string(innvl, "target", &target);
4007 	if (target != NULL) {
4008 		const char *cp = strchr(target, '@');
4009 
4010 		/*
4011 		 * The snap name must contain an @, and the part after it must
4012 		 * contain only valid characters.
4013 		 */
4014 		if (cp == NULL ||
4015 		    zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
4016 			return (SET_ERROR(EINVAL));
4017 	}
4018 
4019 	if (getzfsvfs(fsname, &zfsvfs) == 0) {
4020 		dsl_dataset_t *ds;
4021 
4022 		ds = dmu_objset_ds(zfsvfs->z_os);
4023 		error = zfs_suspend_fs(zfsvfs);
4024 		if (error == 0) {
4025 			int resume_err;
4026 
4027 			error = dsl_dataset_rollback(fsname, target, zfsvfs,
4028 			    outnvl);
4029 			resume_err = zfs_resume_fs(zfsvfs, ds);
4030 			error = error ? error : resume_err;
4031 		}
4032 		VFS_RELE(zfsvfs->z_vfs);
4033 	} else {
4034 		error = dsl_dataset_rollback(fsname, target, NULL, outnvl);
4035 	}
4036 	return (error);
4037 }
4038 
4039 static int
4040 recursive_unmount(const char *fsname, void *arg)
4041 {
4042 	const char *snapname = arg;
4043 	char fullname[ZFS_MAX_DATASET_NAME_LEN];
4044 
4045 	(void) snprintf(fullname, sizeof (fullname), "%s@%s", fsname, snapname);
4046 	zfs_unmount_snap(fullname);
4047 
4048 	return (0);
4049 }
4050 
4051 /*
4052  * inputs:
4053  * zc_name	old name of dataset
4054  * zc_value	new name of dataset
4055  * zc_cookie	recursive flag (only valid for snapshots)
4056  *
4057  * outputs:	none
4058  */
4059 static int
4060 zfs_ioc_rename(zfs_cmd_t *zc)
4061 {
4062 	objset_t *os;
4063 	dmu_objset_type_t ost;
4064 	boolean_t recursive = zc->zc_cookie & 1;
4065 	char *at;
4066 	int err;
4067 
4068 	/* "zfs rename" from and to ...%recv datasets should both fail */
4069 	zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
4070 	zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
4071 	if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0 ||
4072 	    dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4073 	    strchr(zc->zc_name, '%') || strchr(zc->zc_value, '%'))
4074 		return (SET_ERROR(EINVAL));
4075 
4076 	err = dmu_objset_hold(zc->zc_name, FTAG, &os);
4077 	if (err != 0)
4078 		return (err);
4079 	ost = dmu_objset_type(os);
4080 	dmu_objset_rele(os, FTAG);
4081 
4082 	at = strchr(zc->zc_name, '@');
4083 	if (at != NULL) {
4084 		/* snaps must be in same fs */
4085 		int error;
4086 
4087 		if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
4088 			return (SET_ERROR(EXDEV));
4089 		*at = '\0';
4090 		if (ost == DMU_OST_ZFS) {
4091 			error = dmu_objset_find(zc->zc_name,
4092 			    recursive_unmount, at + 1,
4093 			    recursive ? DS_FIND_CHILDREN : 0);
4094 			if (error != 0) {
4095 				*at = '@';
4096 				return (error);
4097 			}
4098 		}
4099 		error = dsl_dataset_rename_snapshot(zc->zc_name,
4100 		    at + 1, strchr(zc->zc_value, '@') + 1, recursive);
4101 		*at = '@';
4102 
4103 		return (error);
4104 	} else {
4105 		if (ost == DMU_OST_ZVOL)
4106 			(void) zvol_remove_minor(zc->zc_name);
4107 		return (dsl_dir_rename(zc->zc_name, zc->zc_value));
4108 	}
4109 }
4110 
4111 static int
4112 zfs_check_settable(const char *dsname, nvpair_t *pair, cred_t *cr)
4113 {
4114 	const char *propname = nvpair_name(pair);
4115 	boolean_t issnap = (strchr(dsname, '@') != NULL);
4116 	zfs_prop_t prop = zfs_name_to_prop(propname);
4117 	uint64_t intval;
4118 	int err;
4119 
4120 	if (prop == ZPROP_INVAL) {
4121 		if (zfs_prop_user(propname)) {
4122 			if (err = zfs_secpolicy_write_perms(dsname,
4123 			    ZFS_DELEG_PERM_USERPROP, cr))
4124 				return (err);
4125 			return (0);
4126 		}
4127 
4128 		if (!issnap && zfs_prop_userquota(propname)) {
4129 			const char *perm = NULL;
4130 			const char *uq_prefix =
4131 			    zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
4132 			const char *gq_prefix =
4133 			    zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
4134 			const char *uiq_prefix =
4135 			    zfs_userquota_prop_prefixes[ZFS_PROP_USEROBJQUOTA];
4136 			const char *giq_prefix =
4137 			    zfs_userquota_prop_prefixes[ZFS_PROP_GROUPOBJQUOTA];
4138 			const char *pq_prefix =
4139 			    zfs_userquota_prop_prefixes[ZFS_PROP_PROJECTQUOTA];
4140 			const char *piq_prefix = zfs_userquota_prop_prefixes[\
4141 			    ZFS_PROP_PROJECTOBJQUOTA];
4142 
4143 			if (strncmp(propname, uq_prefix,
4144 			    strlen(uq_prefix)) == 0) {
4145 				perm = ZFS_DELEG_PERM_USERQUOTA;
4146 			} else if (strncmp(propname, uiq_prefix,
4147 			    strlen(uiq_prefix)) == 0) {
4148 				perm = ZFS_DELEG_PERM_USEROBJQUOTA;
4149 			} else if (strncmp(propname, gq_prefix,
4150 			    strlen(gq_prefix)) == 0) {
4151 				perm = ZFS_DELEG_PERM_GROUPQUOTA;
4152 			} else if (strncmp(propname, giq_prefix,
4153 			    strlen(giq_prefix)) == 0) {
4154 				perm = ZFS_DELEG_PERM_GROUPOBJQUOTA;
4155 			} else if (strncmp(propname, pq_prefix,
4156 			    strlen(pq_prefix)) == 0) {
4157 				perm = ZFS_DELEG_PERM_PROJECTQUOTA;
4158 			} else if (strncmp(propname, piq_prefix,
4159 			    strlen(piq_prefix)) == 0) {
4160 				perm = ZFS_DELEG_PERM_PROJECTOBJQUOTA;
4161 			} else {
4162 				/* {USER|GROUP|PROJECT}USED are read-only */
4163 				return (SET_ERROR(EINVAL));
4164 			}
4165 
4166 			if (err = zfs_secpolicy_write_perms(dsname, perm, cr))
4167 				return (err);
4168 			return (0);
4169 		}
4170 
4171 		return (SET_ERROR(EINVAL));
4172 	}
4173 
4174 	if (issnap)
4175 		return (SET_ERROR(EINVAL));
4176 
4177 	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
4178 		/*
4179 		 * dsl_prop_get_all_impl() returns properties in this
4180 		 * format.
4181 		 */
4182 		nvlist_t *attrs;
4183 		VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
4184 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4185 		    &pair) == 0);
4186 	}
4187 
4188 	/*
4189 	 * Check that this value is valid for this pool version
4190 	 */
4191 	switch (prop) {
4192 	case ZFS_PROP_COMPRESSION:
4193 		/*
4194 		 * If the user specified gzip compression, make sure
4195 		 * the SPA supports it. We ignore any errors here since
4196 		 * we'll catch them later.
4197 		 */
4198 		if (nvpair_value_uint64(pair, &intval) == 0) {
4199 			if (intval >= ZIO_COMPRESS_GZIP_1 &&
4200 			    intval <= ZIO_COMPRESS_GZIP_9 &&
4201 			    zfs_earlier_version(dsname,
4202 			    SPA_VERSION_GZIP_COMPRESSION)) {
4203 				return (SET_ERROR(ENOTSUP));
4204 			}
4205 
4206 			if (intval == ZIO_COMPRESS_ZLE &&
4207 			    zfs_earlier_version(dsname,
4208 			    SPA_VERSION_ZLE_COMPRESSION))
4209 				return (SET_ERROR(ENOTSUP));
4210 
4211 			if (intval == ZIO_COMPRESS_LZ4) {
4212 				spa_t *spa;
4213 
4214 				if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4215 					return (err);
4216 
4217 				if (!spa_feature_is_enabled(spa,
4218 				    SPA_FEATURE_LZ4_COMPRESS)) {
4219 					spa_close(spa, FTAG);
4220 					return (SET_ERROR(ENOTSUP));
4221 				}
4222 				spa_close(spa, FTAG);
4223 			}
4224 
4225 			/*
4226 			 * If this is a bootable dataset then
4227 			 * verify that the compression algorithm
4228 			 * is supported for booting. We must return
4229 			 * something other than ENOTSUP since it
4230 			 * implies a downrev pool version.
4231 			 */
4232 			if (zfs_is_bootfs(dsname) &&
4233 			    !BOOTFS_COMPRESS_VALID(intval)) {
4234 				return (SET_ERROR(ERANGE));
4235 			}
4236 		}
4237 		break;
4238 
4239 	case ZFS_PROP_COPIES:
4240 		if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
4241 			return (SET_ERROR(ENOTSUP));
4242 		break;
4243 
4244 	case ZFS_PROP_RECORDSIZE:
4245 		/* Record sizes above 128k need the feature to be enabled */
4246 		if (nvpair_value_uint64(pair, &intval) == 0 &&
4247 		    intval > SPA_OLD_MAXBLOCKSIZE) {
4248 			spa_t *spa;
4249 
4250 			/*
4251 			 * We don't allow setting the property above 1MB,
4252 			 * unless the tunable has been changed.
4253 			 */
4254 			if (intval > zfs_max_recordsize ||
4255 			    intval > SPA_MAXBLOCKSIZE)
4256 				return (SET_ERROR(ERANGE));
4257 
4258 			if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4259 				return (err);
4260 
4261 			if (!spa_feature_is_enabled(spa,
4262 			    SPA_FEATURE_LARGE_BLOCKS)) {
4263 				spa_close(spa, FTAG);
4264 				return (SET_ERROR(ENOTSUP));
4265 			}
4266 			spa_close(spa, FTAG);
4267 		}
4268 		break;
4269 
4270 	case ZFS_PROP_DNODESIZE:
4271 		/* Dnode sizes above 512 need the feature to be enabled */
4272 		if (nvpair_value_uint64(pair, &intval) == 0 &&
4273 		    intval != ZFS_DNSIZE_LEGACY) {
4274 			spa_t *spa;
4275 
4276 			if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4277 				return (err);
4278 
4279 			if (!spa_feature_is_enabled(spa,
4280 			    SPA_FEATURE_LARGE_DNODE)) {
4281 				spa_close(spa, FTAG);
4282 				return (SET_ERROR(ENOTSUP));
4283 			}
4284 			spa_close(spa, FTAG);
4285 		}
4286 		break;
4287 
4288 	case ZFS_PROP_SPECIAL_SMALL_BLOCKS:
4289 		/*
4290 		 * This property could require the allocation classes
4291 		 * feature to be active for setting, however we allow
4292 		 * it so that tests of settable properties succeed.
4293 		 * The CLI will issue a warning in this case.
4294 		 */
4295 		break;
4296 
4297 	case ZFS_PROP_SHARESMB:
4298 		if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
4299 			return (SET_ERROR(ENOTSUP));
4300 		break;
4301 
4302 	case ZFS_PROP_ACLINHERIT:
4303 		if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
4304 		    nvpair_value_uint64(pair, &intval) == 0) {
4305 			if (intval == ZFS_ACL_PASSTHROUGH_X &&
4306 			    zfs_earlier_version(dsname,
4307 			    SPA_VERSION_PASSTHROUGH_X))
4308 				return (SET_ERROR(ENOTSUP));
4309 		}
4310 		break;
4311 
4312 	case ZFS_PROP_CHECKSUM:
4313 	case ZFS_PROP_DEDUP:
4314 	{
4315 		spa_feature_t feature;
4316 		spa_t *spa;
4317 
4318 		/* dedup feature version checks */
4319 		if (prop == ZFS_PROP_DEDUP &&
4320 		    zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
4321 			return (SET_ERROR(ENOTSUP));
4322 
4323 		if (nvpair_value_uint64(pair, &intval) != 0)
4324 			return (SET_ERROR(EINVAL));
4325 
4326 		/* check prop value is enabled in features */
4327 		feature = zio_checksum_to_feature(intval & ZIO_CHECKSUM_MASK);
4328 		if (feature == SPA_FEATURE_NONE)
4329 			break;
4330 
4331 		if ((err = spa_open(dsname, &spa, FTAG)) != 0)
4332 			return (err);
4333 
4334 		if (!spa_feature_is_enabled(spa, feature)) {
4335 			spa_close(spa, FTAG);
4336 			return (SET_ERROR(ENOTSUP));
4337 		}
4338 		spa_close(spa, FTAG);
4339 		break;
4340 	}
4341 	}
4342 
4343 	return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
4344 }
4345 
4346 /*
4347  * Checks for a race condition to make sure we don't increment a feature flag
4348  * multiple times.
4349  */
4350 static int
4351 zfs_prop_activate_feature_check(void *arg, dmu_tx_t *tx)
4352 {
4353 	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
4354 	spa_feature_t *featurep = arg;
4355 
4356 	if (!spa_feature_is_active(spa, *featurep))
4357 		return (0);
4358 	else
4359 		return (SET_ERROR(EBUSY));
4360 }
4361 
4362 /*
4363  * The callback invoked on feature activation in the sync task caused by
4364  * zfs_prop_activate_feature.
4365  */
4366 static void
4367 zfs_prop_activate_feature_sync(void *arg, dmu_tx_t *tx)
4368 {
4369 	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
4370 	spa_feature_t *featurep = arg;
4371 
4372 	spa_feature_incr(spa, *featurep, tx);
4373 }
4374 
4375 /*
4376  * Activates a feature on a pool in response to a property setting. This
4377  * creates a new sync task which modifies the pool to reflect the feature
4378  * as being active.
4379  */
4380 static int
4381 zfs_prop_activate_feature(spa_t *spa, spa_feature_t feature)
4382 {
4383 	int err;
4384 
4385 	/* EBUSY here indicates that the feature is already active */
4386 	err = dsl_sync_task(spa_name(spa),
4387 	    zfs_prop_activate_feature_check, zfs_prop_activate_feature_sync,
4388 	    &feature, 2, ZFS_SPACE_CHECK_RESERVED);
4389 
4390 	if (err != 0 && err != EBUSY)
4391 		return (err);
4392 	else
4393 		return (0);
4394 }
4395 
4396 /*
4397  * Removes properties from the given props list that fail permission checks
4398  * needed to clear them and to restore them in case of a receive error. For each
4399  * property, make sure we have both set and inherit permissions.
4400  *
4401  * Returns the first error encountered if any permission checks fail. If the
4402  * caller provides a non-NULL errlist, it also gives the complete list of names
4403  * of all the properties that failed a permission check along with the
4404  * corresponding error numbers. The caller is responsible for freeing the
4405  * returned errlist.
4406  *
4407  * If every property checks out successfully, zero is returned and the list
4408  * pointed at by errlist is NULL.
4409  */
4410 static int
4411 zfs_check_clearable(char *dataset, nvlist_t *props, nvlist_t **errlist)
4412 {
4413 	zfs_cmd_t *zc;
4414 	nvpair_t *pair, *next_pair;
4415 	nvlist_t *errors;
4416 	int err, rv = 0;
4417 
4418 	if (props == NULL)
4419 		return (0);
4420 
4421 	VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4422 
4423 	zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
4424 	(void) strcpy(zc->zc_name, dataset);
4425 	pair = nvlist_next_nvpair(props, NULL);
4426 	while (pair != NULL) {
4427 		next_pair = nvlist_next_nvpair(props, pair);
4428 
4429 		(void) strcpy(zc->zc_value, nvpair_name(pair));
4430 		if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 ||
4431 		    (err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
4432 			VERIFY(nvlist_remove_nvpair(props, pair) == 0);
4433 			VERIFY(nvlist_add_int32(errors,
4434 			    zc->zc_value, err) == 0);
4435 		}
4436 		pair = next_pair;
4437 	}
4438 	kmem_free(zc, sizeof (zfs_cmd_t));
4439 
4440 	if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
4441 		nvlist_free(errors);
4442 		errors = NULL;
4443 	} else {
4444 		VERIFY(nvpair_value_int32(pair, &rv) == 0);
4445 	}
4446 
4447 	if (errlist == NULL)
4448 		nvlist_free(errors);
4449 	else
4450 		*errlist = errors;
4451 
4452 	return (rv);
4453 }
4454 
4455 static boolean_t
4456 propval_equals(nvpair_t *p1, nvpair_t *p2)
4457 {
4458 	if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
4459 		/* dsl_prop_get_all_impl() format */
4460 		nvlist_t *attrs;
4461 		VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
4462 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4463 		    &p1) == 0);
4464 	}
4465 
4466 	if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
4467 		nvlist_t *attrs;
4468 		VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
4469 		VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
4470 		    &p2) == 0);
4471 	}
4472 
4473 	if (nvpair_type(p1) != nvpair_type(p2))
4474 		return (B_FALSE);
4475 
4476 	if (nvpair_type(p1) == DATA_TYPE_STRING) {
4477 		char *valstr1, *valstr2;
4478 
4479 		VERIFY(nvpair_value_string(p1, (char **)&valstr1) == 0);
4480 		VERIFY(nvpair_value_string(p2, (char **)&valstr2) == 0);
4481 		return (strcmp(valstr1, valstr2) == 0);
4482 	} else {
4483 		uint64_t intval1, intval2;
4484 
4485 		VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
4486 		VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
4487 		return (intval1 == intval2);
4488 	}
4489 }
4490 
4491 /*
4492  * Remove properties from props if they are not going to change (as determined
4493  * by comparison with origprops). Remove them from origprops as well, since we
4494  * do not need to clear or restore properties that won't change.
4495  */
4496 static void
4497 props_reduce(nvlist_t *props, nvlist_t *origprops)
4498 {
4499 	nvpair_t *pair, *next_pair;
4500 
4501 	if (origprops == NULL)
4502 		return; /* all props need to be received */
4503 
4504 	pair = nvlist_next_nvpair(props, NULL);
4505 	while (pair != NULL) {
4506 		const char *propname = nvpair_name(pair);
4507 		nvpair_t *match;
4508 
4509 		next_pair = nvlist_next_nvpair(props, pair);
4510 
4511 		if ((nvlist_lookup_nvpair(origprops, propname,
4512 		    &match) != 0) || !propval_equals(pair, match))
4513 			goto next; /* need to set received value */
4514 
4515 		/* don't clear the existing received value */
4516 		(void) nvlist_remove_nvpair(origprops, match);
4517 		/* don't bother receiving the property */
4518 		(void) nvlist_remove_nvpair(props, pair);
4519 next:
4520 		pair = next_pair;
4521 	}
4522 }
4523 
4524 /*
4525  * Extract properties that cannot be set PRIOR to the receipt of a dataset.
4526  * For example, refquota cannot be set until after the receipt of a dataset,
4527  * because in replication streams, an older/earlier snapshot may exceed the
4528  * refquota.  We want to receive the older/earlier snapshot, but setting
4529  * refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent
4530  * the older/earlier snapshot from being received (with EDQUOT).
4531  *
4532  * The ZFS test "zfs_receive_011_pos" demonstrates such a scenario.
4533  *
4534  * libzfs will need to be judicious handling errors encountered by props
4535  * extracted by this function.
4536  */
4537 static nvlist_t *
4538 extract_delay_props(nvlist_t *props)
4539 {
4540 	nvlist_t *delayprops;
4541 	nvpair_t *nvp, *tmp;
4542 	static const zfs_prop_t delayable[] = {
4543 		ZFS_PROP_REFQUOTA,
4544 		ZFS_PROP_KEYLOCATION,
4545 		0
4546 	};
4547 	int i;
4548 
4549 	VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
4550 
4551 	for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL;
4552 	    nvp = nvlist_next_nvpair(props, nvp)) {
4553 		/*
4554 		 * strcmp() is safe because zfs_prop_to_name() always returns
4555 		 * a bounded string.
4556 		 */
4557 		for (i = 0; delayable[i] != 0; i++) {
4558 			if (strcmp(zfs_prop_to_name(delayable[i]),
4559 			    nvpair_name(nvp)) == 0) {
4560 				break;
4561 			}
4562 		}
4563 		if (delayable[i] != 0) {
4564 			tmp = nvlist_prev_nvpair(props, nvp);
4565 			VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0);
4566 			VERIFY(nvlist_remove_nvpair(props, nvp) == 0);
4567 			nvp = tmp;
4568 		}
4569 	}
4570 
4571 	if (nvlist_empty(delayprops)) {
4572 		nvlist_free(delayprops);
4573 		delayprops = NULL;
4574 	}
4575 	return (delayprops);
4576 }
4577 
4578 #ifdef	DEBUG
4579 static boolean_t zfs_ioc_recv_inject_err;
4580 #endif
4581 
4582 /*
4583  * nvlist 'errors' is always allocated. It will contain descriptions of
4584  * encountered errors, if any. It's the callers responsibility to free.
4585  */
4586 static int
4587 zfs_ioc_recv_impl(char *tofs, char *tosnap, char *origin, nvlist_t *recvprops,
4588     nvlist_t *localprops, nvlist_t *hidden_args, boolean_t force,
4589     boolean_t resumable, int input_fd, dmu_replay_record_t *begin_record,
4590     int cleanup_fd, uint64_t *read_bytes, uint64_t *errflags,
4591     uint64_t *action_handle, nvlist_t **errors)
4592 {
4593 	dmu_recv_cookie_t drc;
4594 	int error = 0;
4595 	int props_error = 0;
4596 	offset_t off;
4597 	nvlist_t *local_delayprops = NULL;
4598 	nvlist_t *recv_delayprops = NULL;
4599 	nvlist_t *origprops = NULL; /* existing properties */
4600 	nvlist_t *origrecvd = NULL; /* existing received properties */
4601 	boolean_t first_recvd_props = B_FALSE;
4602 	file_t *input_fp;
4603 
4604 	*read_bytes = 0;
4605 	*errflags = 0;
4606 	*errors = fnvlist_alloc();
4607 
4608 	input_fp = getf(input_fd);
4609 	if (input_fp == NULL)
4610 		return (SET_ERROR(EBADF));
4611 
4612 	error = dmu_recv_begin(tofs, tosnap, begin_record, force,
4613 	    resumable, localprops, hidden_args, origin, &drc);
4614 	if (error != 0)
4615 		goto out;
4616 
4617 	/*
4618 	 * Set properties before we receive the stream so that they are applied
4619 	 * to the new data. Note that we must call dmu_recv_stream() if
4620 	 * dmu_recv_begin() succeeds.
4621 	 */
4622 	if (recvprops != NULL && !drc.drc_newfs) {
4623 		if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
4624 		    SPA_VERSION_RECVD_PROPS &&
4625 		    !dsl_prop_get_hasrecvd(tofs))
4626 			first_recvd_props = B_TRUE;
4627 
4628 		/*
4629 		 * If new received properties are supplied, they are to
4630 		 * completely replace the existing received properties,
4631 		 * so stash away the existing ones.
4632 		 */
4633 		if (dsl_prop_get_received(tofs, &origrecvd) == 0) {
4634 			nvlist_t *errlist = NULL;
4635 			/*
4636 			 * Don't bother writing a property if its value won't
4637 			 * change (and avoid the unnecessary security checks).
4638 			 *
4639 			 * The first receive after SPA_VERSION_RECVD_PROPS is a
4640 			 * special case where we blow away all local properties
4641 			 * regardless.
4642 			 */
4643 			if (!first_recvd_props)
4644 				props_reduce(recvprops, origrecvd);
4645 			if (zfs_check_clearable(tofs, origrecvd, &errlist) != 0)
4646 				(void) nvlist_merge(*errors, errlist, 0);
4647 			nvlist_free(errlist);
4648 
4649 			if (clear_received_props(tofs, origrecvd,
4650 			    first_recvd_props ? NULL : recvprops) != 0)
4651 				*errflags |= ZPROP_ERR_NOCLEAR;
4652 		} else {
4653 			*errflags |= ZPROP_ERR_NOCLEAR;
4654 		}
4655 	}
4656 
4657 	/*
4658 	 * Stash away existing properties so we can restore them on error unless
4659 	 * we're doing the first receive after SPA_VERSION_RECVD_PROPS, in which
4660 	 * case "origrecvd" will take care of that.
4661 	 */
4662 	if (localprops != NULL && !drc.drc_newfs && !first_recvd_props) {
4663 		objset_t *os;
4664 		if (dmu_objset_hold(tofs, FTAG, &os) == 0) {
4665 			if (dsl_prop_get_all(os, &origprops) != 0) {
4666 				*errflags |= ZPROP_ERR_NOCLEAR;
4667 			}
4668 			dmu_objset_rele(os, FTAG);
4669 		} else {
4670 			*errflags |= ZPROP_ERR_NOCLEAR;
4671 		}
4672 	}
4673 
4674 	if (recvprops != NULL) {
4675 		props_error = dsl_prop_set_hasrecvd(tofs);
4676 
4677 		if (props_error == 0) {
4678 			recv_delayprops = extract_delay_props(recvprops);
4679 			(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4680 			    recvprops, *errors);
4681 		}
4682 	}
4683 
4684 	if (localprops != NULL) {
4685 		nvlist_t *oprops = fnvlist_alloc();
4686 		nvlist_t *xprops = fnvlist_alloc();
4687 		nvpair_t *nvp = NULL;
4688 
4689 		while ((nvp = nvlist_next_nvpair(localprops, nvp)) != NULL) {
4690 			if (nvpair_type(nvp) == DATA_TYPE_BOOLEAN) {
4691 				/* -x property */
4692 				const char *name = nvpair_name(nvp);
4693 				zfs_prop_t prop = zfs_name_to_prop(name);
4694 				if (prop != ZPROP_INVAL) {
4695 					if (!zfs_prop_inheritable(prop))
4696 						continue;
4697 				} else if (!zfs_prop_user(name))
4698 					continue;
4699 				fnvlist_add_boolean(xprops, name);
4700 			} else {
4701 				/* -o property=value */
4702 				fnvlist_add_nvpair(oprops, nvp);
4703 			}
4704 		}
4705 
4706 		local_delayprops = extract_delay_props(oprops);
4707 		(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL,
4708 		    oprops, *errors);
4709 		(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_INHERITED,
4710 		    xprops, *errors);
4711 
4712 		nvlist_free(oprops);
4713 		nvlist_free(xprops);
4714 	}
4715 
4716 	off = input_fp->f_offset;
4717 	error = dmu_recv_stream(&drc, input_fp->f_vnode, &off, cleanup_fd,
4718 	    action_handle);
4719 
4720 	if (error == 0) {
4721 		zfsvfs_t *zfsvfs = NULL;
4722 
4723 		if (getzfsvfs(tofs, &zfsvfs) == 0) {
4724 			/* online recv */
4725 			dsl_dataset_t *ds;
4726 			int end_err;
4727 
4728 			ds = dmu_objset_ds(zfsvfs->z_os);
4729 			error = zfs_suspend_fs(zfsvfs);
4730 			/*
4731 			 * If the suspend fails, then the recv_end will
4732 			 * likely also fail, and clean up after itself.
4733 			 */
4734 			end_err = dmu_recv_end(&drc, zfsvfs);
4735 			if (error == 0)
4736 				error = zfs_resume_fs(zfsvfs, ds);
4737 			error = error ? error : end_err;
4738 			VFS_RELE(zfsvfs->z_vfs);
4739 		} else {
4740 			error = dmu_recv_end(&drc, NULL);
4741 		}
4742 
4743 		/* Set delayed properties now, after we're done receiving. */
4744 		if (recv_delayprops != NULL && error == 0) {
4745 			(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
4746 			    recv_delayprops, *errors);
4747 		}
4748 		if (local_delayprops != NULL && error == 0) {
4749 			(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL,
4750 			    local_delayprops, *errors);
4751 		}
4752 	}
4753 
4754 	/*
4755 	 * Merge delayed props back in with initial props, in case
4756 	 * we're DEBUG and zfs_ioc_recv_inject_err is set (which means
4757 	 * we have to make sure clear_received_props() includes
4758 	 * the delayed properties).
4759 	 *
4760 	 * Since zfs_ioc_recv_inject_err is only in DEBUG kernels,
4761 	 * using ASSERT() will be just like a VERIFY.
4762 	 */
4763 	if (recv_delayprops != NULL) {
4764 		ASSERT(nvlist_merge(recvprops, recv_delayprops, 0) == 0);
4765 		nvlist_free(recv_delayprops);
4766 	}
4767 	if (local_delayprops != NULL) {
4768 		ASSERT(nvlist_merge(localprops, local_delayprops, 0) == 0);
4769 		nvlist_free(local_delayprops);
4770 	}
4771 
4772 	*read_bytes = off - input_fp->f_offset;
4773 	if (VOP_SEEK(input_fp->f_vnode, input_fp->f_offset, &off, NULL) == 0)
4774 		input_fp->f_offset = off;
4775 
4776 #ifdef	DEBUG
4777 	if (zfs_ioc_recv_inject_err) {
4778 		zfs_ioc_recv_inject_err = B_FALSE;
4779 		error = 1;
4780 	}
4781 #endif
4782 
4783 	/*
4784 	 * On error, restore the original props.
4785 	 */
4786 	if (error != 0 && recvprops != NULL && !drc.drc_newfs) {
4787 		if (clear_received_props(tofs, recvprops, NULL) != 0) {
4788 			/*
4789 			 * We failed to clear the received properties.
4790 			 * Since we may have left a $recvd value on the
4791 			 * system, we can't clear the $hasrecvd flag.
4792 			 */
4793 			*errflags |= ZPROP_ERR_NORESTORE;
4794 		} else if (first_recvd_props) {
4795 			dsl_prop_unset_hasrecvd(tofs);
4796 		}
4797 
4798 		if (origrecvd == NULL && !drc.drc_newfs) {
4799 			/* We failed to stash the original properties. */
4800 			*errflags |= ZPROP_ERR_NORESTORE;
4801 		}
4802 
4803 		/*
4804 		 * dsl_props_set() will not convert RECEIVED to LOCAL on or
4805 		 * after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
4806 		 * explicitly if we're restoring local properties cleared in the
4807 		 * first new-style receive.
4808 		 */
4809 		if (origrecvd != NULL &&
4810 		    zfs_set_prop_nvlist(tofs, (first_recvd_props ?
4811 		    ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
4812 		    origrecvd, NULL) != 0) {
4813 			/*
4814 			 * We stashed the original properties but failed to
4815 			 * restore them.
4816 			 */
4817 			*errflags |= ZPROP_ERR_NORESTORE;
4818 		}
4819 	}
4820 	if (error != 0 && localprops != NULL && !drc.drc_newfs &&
4821 	    !first_recvd_props) {
4822 		nvlist_t *setprops;
4823 		nvlist_t *inheritprops;
4824 		nvpair_t *nvp;
4825 
4826 		if (origprops == NULL) {
4827 			/* We failed to stash the original properties. */
4828 			*errflags |= ZPROP_ERR_NORESTORE;
4829 			goto out;
4830 		}
4831 
4832 		/* Restore original props */
4833 		setprops = fnvlist_alloc();
4834 		inheritprops = fnvlist_alloc();
4835 		nvp = NULL;
4836 		while ((nvp = nvlist_next_nvpair(localprops, nvp)) != NULL) {
4837 			const char *name = nvpair_name(nvp);
4838 			const char *source;
4839 			nvlist_t *attrs;
4840 
4841 			if (!nvlist_exists(origprops, name)) {
4842 				/*
4843 				 * Property was not present or was explicitly
4844 				 * inherited before the receive, restore this.
4845 				 */
4846 				fnvlist_add_boolean(inheritprops, name);
4847 				continue;
4848 			}
4849 			attrs = fnvlist_lookup_nvlist(origprops, name);
4850 			source = fnvlist_lookup_string(attrs, ZPROP_SOURCE);
4851 
4852 			/* Skip received properties */
4853 			if (strcmp(source, ZPROP_SOURCE_VAL_RECVD) == 0)
4854 				continue;
4855 
4856 			if (strcmp(source, tofs) == 0) {
4857 				/* Property was locally set */
4858 				fnvlist_add_nvlist(setprops, name, attrs);
4859 			} else {
4860 				/* Property was implicitly inherited */
4861 				fnvlist_add_boolean(inheritprops, name);
4862 			}
4863 		}
4864 
4865 		if (zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL, setprops,
4866 		    NULL) != 0)
4867 			*errflags |= ZPROP_ERR_NORESTORE;
4868 		if (zfs_set_prop_nvlist(tofs, ZPROP_SRC_INHERITED, inheritprops,
4869 		    NULL) != 0)
4870 			*errflags |= ZPROP_ERR_NORESTORE;
4871 
4872 		nvlist_free(setprops);
4873 		nvlist_free(inheritprops);
4874 	}
4875 out:
4876 	releasef(input_fd);
4877 	nvlist_free(origrecvd);
4878 	nvlist_free(origprops);
4879 
4880 	if (error == 0)
4881 		error = props_error;
4882 
4883 	return (error);
4884 }
4885 
4886 /*
4887  * inputs:
4888  * zc_name		name of containing filesystem
4889  * zc_nvlist_src{_size}	nvlist of received properties to apply
4890  * zc_nvlist_conf{_size} nvlist of local properties to apply
4891  * zc_history_offset{_len} nvlist of hidden args { "wkeydata" -> value }
4892  * zc_value		name of snapshot to create
4893  * zc_string		name of clone origin (if DRR_FLAG_CLONE)
4894  * zc_cookie		file descriptor to recv from
4895  * zc_begin_record	the BEGIN record of the stream (not byteswapped)
4896  * zc_guid		force flag
4897  * zc_cleanup_fd	cleanup-on-exit file descriptor
4898  * zc_action_handle	handle for this guid/ds mapping (or zero on first call)
4899  * zc_resumable		if data is incomplete assume sender will resume
4900  *
4901  * outputs:
4902  * zc_cookie		number of bytes read
4903  * zc_nvlist_dst{_size} error for each unapplied received property
4904  * zc_obj		zprop_errflags_t
4905  * zc_action_handle	handle for this guid/ds mapping
4906  */
4907 static int
4908 zfs_ioc_recv(zfs_cmd_t *zc)
4909 {
4910 	dmu_replay_record_t begin_record;
4911 	nvlist_t *errors = NULL;
4912 	nvlist_t *recvdprops = NULL;
4913 	nvlist_t *localprops = NULL;
4914 	nvlist_t *hidden_args = NULL;
4915 	char *origin = NULL;
4916 	char *tosnap;
4917 	char tofs[ZFS_MAX_DATASET_NAME_LEN];
4918 	int error = 0;
4919 
4920 	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 ||
4921 	    strchr(zc->zc_value, '@') == NULL ||
4922 	    strchr(zc->zc_value, '%'))
4923 		return (SET_ERROR(EINVAL));
4924 
4925 	(void) strlcpy(tofs, zc->zc_value, sizeof (tofs));
4926 	tosnap = strchr(tofs, '@');
4927 	*tosnap++ = '\0';
4928 
4929 	if (zc->zc_nvlist_src != 0 &&
4930 	    (error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
4931 	    zc->zc_iflags, &recvdprops)) != 0)
4932 		return (error);
4933 
4934 	if (zc->zc_nvlist_conf != 0 &&
4935 	    (error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
4936 	    zc->zc_iflags, &localprops)) != 0)
4937 		return (error);
4938 
4939 	if (zc->zc_history_offset != 0 &&
4940 	    (error = get_nvlist(zc->zc_history_offset, zc->zc_history_len,
4941 	    zc->zc_iflags, &hidden_args)) != 0)
4942 		return (error);
4943 
4944 	if (zc->zc_string[0])
4945 		origin = zc->zc_string;
4946 
4947 	begin_record.drr_type = DRR_BEGIN;
4948 	begin_record.drr_payloadlen = zc->zc_begin_record.drr_payloadlen;
4949 	begin_record.drr_u.drr_begin = zc->zc_begin_record.drr_u.drr_begin;
4950 
4951 	error = zfs_ioc_recv_impl(tofs, tosnap, origin, recvdprops, localprops,
4952 	    hidden_args, zc->zc_guid, zc->zc_resumable, zc->zc_cookie,
4953 	    &begin_record, zc->zc_cleanup_fd, &zc->zc_cookie, &zc->zc_obj,
4954 	    &zc->zc_action_handle, &errors);
4955 	nvlist_free(recvdprops);
4956 	nvlist_free(localprops);
4957 
4958 	/*
4959 	 * Now that all props, initial and delayed, are set, report the prop
4960 	 * errors to the caller.
4961 	 */
4962 	if (zc->zc_nvlist_dst_size != 0 && errors != NULL &&
4963 	    (nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 ||
4964 	    put_nvlist(zc, errors) != 0)) {
4965 		/*
4966 		 * Caller made zc->zc_nvlist_dst less than the minimum expected
4967 		 * size or supplied an invalid address.
4968 		 */
4969 		error = SET_ERROR(EINVAL);
4970 	}
4971 
4972 	nvlist_free(errors);
4973 
4974 	return (error);
4975 }
4976 
4977 /*
4978  * inputs:
4979  * zc_name	name of snapshot to send
4980  * zc_cookie	file descriptor to send stream to
4981  * zc_obj	fromorigin flag (mutually exclusive with zc_fromobj)
4982  * zc_sendobj	objsetid of snapshot to send
4983  * zc_fromobj	objsetid of incremental fromsnap (may be zero)
4984  * zc_guid	if set, estimate size of stream only.  zc_cookie is ignored.
4985  *		output size in zc_objset_type.
4986  * zc_flags	lzc_send_flags
4987  *
4988  * outputs:
4989  * zc_objset_type	estimated size, if zc_guid is set
4990  */
4991 static int
4992 zfs_ioc_send(zfs_cmd_t *zc)
4993 {
4994 	int error;
4995 	offset_t off;
4996 	boolean_t estimate = (zc->zc_guid != 0);
4997 	boolean_t embedok = (zc->zc_flags & 0x1);
4998 	boolean_t large_block_ok = (zc->zc_flags & 0x2);
4999 	boolean_t compressok = (zc->zc_flags & 0x4);
5000 	boolean_t rawok = (zc->zc_flags & 0x8);
5001 
5002 	if (zc->zc_obj != 0) {
5003 		dsl_pool_t *dp;
5004 		dsl_dataset_t *tosnap;
5005 
5006 		error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5007 		if (error != 0)
5008 			return (error);
5009 
5010 		error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
5011 		if (error != 0) {
5012 			dsl_pool_rele(dp, FTAG);
5013 			return (error);
5014 		}
5015 
5016 		if (dsl_dir_is_clone(tosnap->ds_dir))
5017 			zc->zc_fromobj =
5018 			    dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj;
5019 		dsl_dataset_rele(tosnap, FTAG);
5020 		dsl_pool_rele(dp, FTAG);
5021 	}
5022 
5023 	if (estimate) {
5024 		dsl_pool_t *dp;
5025 		dsl_dataset_t *tosnap;
5026 		dsl_dataset_t *fromsnap = NULL;
5027 
5028 		error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5029 		if (error != 0)
5030 			return (error);
5031 
5032 		error = dsl_dataset_hold_obj(dp, zc->zc_sendobj,
5033 		    FTAG, &tosnap);
5034 		if (error != 0) {
5035 			dsl_pool_rele(dp, FTAG);
5036 			return (error);
5037 		}
5038 
5039 		if (zc->zc_fromobj != 0) {
5040 			error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
5041 			    FTAG, &fromsnap);
5042 			if (error != 0) {
5043 				dsl_dataset_rele(tosnap, FTAG);
5044 				dsl_pool_rele(dp, FTAG);
5045 				return (error);
5046 			}
5047 		}
5048 
5049 		error = dmu_send_estimate(tosnap, fromsnap, compressok || rawok,
5050 		    &zc->zc_objset_type);
5051 
5052 		if (fromsnap != NULL)
5053 			dsl_dataset_rele(fromsnap, FTAG);
5054 		dsl_dataset_rele(tosnap, FTAG);
5055 		dsl_pool_rele(dp, FTAG);
5056 	} else {
5057 		file_t *fp = getf(zc->zc_cookie);
5058 		if (fp == NULL)
5059 			return (SET_ERROR(EBADF));
5060 
5061 		off = fp->f_offset;
5062 		error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
5063 		    zc->zc_fromobj, embedok, large_block_ok, compressok, rawok,
5064 		    zc->zc_cookie, fp->f_vnode, &off);
5065 
5066 		if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5067 			fp->f_offset = off;
5068 		releasef(zc->zc_cookie);
5069 	}
5070 	return (error);
5071 }
5072 
5073 /*
5074  * inputs:
5075  * zc_name	name of snapshot on which to report progress
5076  * zc_cookie	file descriptor of send stream
5077  *
5078  * outputs:
5079  * zc_cookie	number of bytes written in send stream thus far
5080  */
5081 static int
5082 zfs_ioc_send_progress(zfs_cmd_t *zc)
5083 {
5084 	dsl_pool_t *dp;
5085 	dsl_dataset_t *ds;
5086 	dmu_sendarg_t *dsp = NULL;
5087 	int error;
5088 
5089 	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5090 	if (error != 0)
5091 		return (error);
5092 
5093 	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
5094 	if (error != 0) {
5095 		dsl_pool_rele(dp, FTAG);
5096 		return (error);
5097 	}
5098 
5099 	mutex_enter(&ds->ds_sendstream_lock);
5100 
5101 	/*
5102 	 * Iterate over all the send streams currently active on this dataset.
5103 	 * If there's one which matches the specified file descriptor _and_ the
5104 	 * stream was started by the current process, return the progress of
5105 	 * that stream.
5106 	 */
5107 	for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
5108 	    dsp = list_next(&ds->ds_sendstreams, dsp)) {
5109 		if (dsp->dsa_outfd == zc->zc_cookie &&
5110 		    dsp->dsa_proc == curproc)
5111 			break;
5112 	}
5113 
5114 	if (dsp != NULL)
5115 		zc->zc_cookie = *(dsp->dsa_off);
5116 	else
5117 		error = SET_ERROR(ENOENT);
5118 
5119 	mutex_exit(&ds->ds_sendstream_lock);
5120 	dsl_dataset_rele(ds, FTAG);
5121 	dsl_pool_rele(dp, FTAG);
5122 	return (error);
5123 }
5124 
5125 static int
5126 zfs_ioc_inject_fault(zfs_cmd_t *zc)
5127 {
5128 	int id, error;
5129 
5130 	error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
5131 	    &zc->zc_inject_record);
5132 
5133 	if (error == 0)
5134 		zc->zc_guid = (uint64_t)id;
5135 
5136 	return (error);
5137 }
5138 
5139 static int
5140 zfs_ioc_clear_fault(zfs_cmd_t *zc)
5141 {
5142 	return (zio_clear_fault((int)zc->zc_guid));
5143 }
5144 
5145 static int
5146 zfs_ioc_inject_list_next(zfs_cmd_t *zc)
5147 {
5148 	int id = (int)zc->zc_guid;
5149 	int error;
5150 
5151 	error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
5152 	    &zc->zc_inject_record);
5153 
5154 	zc->zc_guid = id;
5155 
5156 	return (error);
5157 }
5158 
5159 static int
5160 zfs_ioc_error_log(zfs_cmd_t *zc)
5161 {
5162 	spa_t *spa;
5163 	int error;
5164 	size_t count = (size_t)zc->zc_nvlist_dst_size;
5165 
5166 	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
5167 		return (error);
5168 
5169 	error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
5170 	    &count);
5171 	if (error == 0)
5172 		zc->zc_nvlist_dst_size = count;
5173 	else
5174 		zc->zc_nvlist_dst_size = spa_get_errlog_size(spa);
5175 
5176 	spa_close(spa, FTAG);
5177 
5178 	return (error);
5179 }
5180 
5181 static int
5182 zfs_ioc_clear(zfs_cmd_t *zc)
5183 {
5184 	spa_t *spa;
5185 	vdev_t *vd;
5186 	int error;
5187 
5188 	/*
5189 	 * On zpool clear we also fix up missing slogs
5190 	 */
5191 	mutex_enter(&spa_namespace_lock);
5192 	spa = spa_lookup(zc->zc_name);
5193 	if (spa == NULL) {
5194 		mutex_exit(&spa_namespace_lock);
5195 		return (SET_ERROR(EIO));
5196 	}
5197 	if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
5198 		/* we need to let spa_open/spa_load clear the chains */
5199 		spa_set_log_state(spa, SPA_LOG_CLEAR);
5200 	}
5201 	spa->spa_last_open_failed = 0;
5202 	mutex_exit(&spa_namespace_lock);
5203 
5204 	if (zc->zc_cookie & ZPOOL_NO_REWIND) {
5205 		error = spa_open(zc->zc_name, &spa, FTAG);
5206 	} else {
5207 		nvlist_t *policy;
5208 		nvlist_t *config = NULL;
5209 
5210 		if (zc->zc_nvlist_src == 0)
5211 			return (SET_ERROR(EINVAL));
5212 
5213 		if ((error = get_nvlist(zc->zc_nvlist_src,
5214 		    zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
5215 			error = spa_open_rewind(zc->zc_name, &spa, FTAG,
5216 			    policy, &config);
5217 			if (config != NULL) {
5218 				int err;
5219 
5220 				if ((err = put_nvlist(zc, config)) != 0)
5221 					error = err;
5222 				nvlist_free(config);
5223 			}
5224 			nvlist_free(policy);
5225 		}
5226 	}
5227 
5228 	if (error != 0)
5229 		return (error);
5230 
5231 	/*
5232 	 * If multihost is enabled, resuming I/O is unsafe as another
5233 	 * host may have imported the pool.
5234 	 */
5235 	if (spa_multihost(spa) && spa_suspended(spa))
5236 		return (SET_ERROR(EINVAL));
5237 
5238 	spa_vdev_state_enter(spa, SCL_NONE);
5239 
5240 	if (zc->zc_guid == 0) {
5241 		vd = NULL;
5242 	} else {
5243 		vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
5244 		if (vd == NULL) {
5245 			(void) spa_vdev_state_exit(spa, NULL, ENODEV);
5246 			spa_close(spa, FTAG);
5247 			return (SET_ERROR(ENODEV));
5248 		}
5249 	}
5250 
5251 	vdev_clear(spa, vd);
5252 
5253 	(void) spa_vdev_state_exit(spa, NULL, 0);
5254 
5255 	/*
5256 	 * Resume any suspended I/Os.
5257 	 */
5258 	if (zio_resume(spa) != 0)
5259 		error = SET_ERROR(EIO);
5260 
5261 	spa_close(spa, FTAG);
5262 
5263 	return (error);
5264 }
5265 
5266 static int
5267 zfs_ioc_pool_reopen(zfs_cmd_t *zc)
5268 {
5269 	spa_t *spa;
5270 	int error;
5271 
5272 	error = spa_open(zc->zc_name, &spa, FTAG);
5273 	if (error != 0)
5274 		return (error);
5275 
5276 	spa_vdev_state_enter(spa, SCL_NONE);
5277 
5278 	/*
5279 	 * If a resilver is already in progress then set the
5280 	 * spa_scrub_reopen flag to B_TRUE so that we don't restart
5281 	 * the scan as a side effect of the reopen. Otherwise, let
5282 	 * vdev_open() decided if a resilver is required.
5283 	 */
5284 	spa->spa_scrub_reopen = dsl_scan_resilvering(spa->spa_dsl_pool);
5285 	vdev_reopen(spa->spa_root_vdev);
5286 	spa->spa_scrub_reopen = B_FALSE;
5287 
5288 	(void) spa_vdev_state_exit(spa, NULL, 0);
5289 	spa_close(spa, FTAG);
5290 	return (0);
5291 }
5292 /*
5293  * inputs:
5294  * zc_name	name of filesystem
5295  *
5296  * outputs:
5297  * zc_string	name of conflicting snapshot, if there is one
5298  */
5299 static int
5300 zfs_ioc_promote(zfs_cmd_t *zc)
5301 {
5302 	dsl_pool_t *dp;
5303 	dsl_dataset_t *ds, *ods;
5304 	char origin[ZFS_MAX_DATASET_NAME_LEN];
5305 	char *cp;
5306 	int error;
5307 
5308 	zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
5309 	if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0 ||
5310 	    strchr(zc->zc_name, '%'))
5311 		return (SET_ERROR(EINVAL));
5312 
5313 	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5314 	if (error != 0)
5315 		return (error);
5316 
5317 	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
5318 	if (error != 0) {
5319 		dsl_pool_rele(dp, FTAG);
5320 		return (error);
5321 	}
5322 
5323 	if (!dsl_dir_is_clone(ds->ds_dir)) {
5324 		dsl_dataset_rele(ds, FTAG);
5325 		dsl_pool_rele(dp, FTAG);
5326 		return (SET_ERROR(EINVAL));
5327 	}
5328 
5329 	error = dsl_dataset_hold_obj(dp,
5330 	    dsl_dir_phys(ds->ds_dir)->dd_origin_obj, FTAG, &ods);
5331 	if (error != 0) {
5332 		dsl_dataset_rele(ds, FTAG);
5333 		dsl_pool_rele(dp, FTAG);
5334 		return (error);
5335 	}
5336 
5337 	dsl_dataset_name(ods, origin);
5338 	dsl_dataset_rele(ods, FTAG);
5339 	dsl_dataset_rele(ds, FTAG);
5340 	dsl_pool_rele(dp, FTAG);
5341 
5342 	/*
5343 	 * We don't need to unmount *all* the origin fs's snapshots, but
5344 	 * it's easier.
5345 	 */
5346 	cp = strchr(origin, '@');
5347 	if (cp)
5348 		*cp = '\0';
5349 	(void) dmu_objset_find(origin,
5350 	    zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
5351 	return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
5352 }
5353 
5354 /*
5355  * Retrieve a single {user|group|project}{used|quota}@... property.
5356  *
5357  * inputs:
5358  * zc_name	name of filesystem
5359  * zc_objset_type zfs_userquota_prop_t
5360  * zc_value	domain name (eg. "S-1-234-567-89")
5361  * zc_guid	RID/UID/GID
5362  *
5363  * outputs:
5364  * zc_cookie	property value
5365  */
5366 static int
5367 zfs_ioc_userspace_one(zfs_cmd_t *zc)
5368 {
5369 	zfsvfs_t *zfsvfs;
5370 	int error;
5371 
5372 	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
5373 		return (SET_ERROR(EINVAL));
5374 
5375 	error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
5376 	if (error != 0)
5377 		return (error);
5378 
5379 	error = zfs_userspace_one(zfsvfs,
5380 	    zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
5381 	zfsvfs_rele(zfsvfs, FTAG);
5382 
5383 	return (error);
5384 }
5385 
5386 /*
5387  * inputs:
5388  * zc_name		name of filesystem
5389  * zc_cookie		zap cursor
5390  * zc_objset_type	zfs_userquota_prop_t
5391  * zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
5392  *
5393  * outputs:
5394  * zc_nvlist_dst[_size]	data buffer (array of zfs_useracct_t)
5395  * zc_cookie	zap cursor
5396  */
5397 static int
5398 zfs_ioc_userspace_many(zfs_cmd_t *zc)
5399 {
5400 	zfsvfs_t *zfsvfs;
5401 	int bufsize = zc->zc_nvlist_dst_size;
5402 
5403 	if (bufsize <= 0)
5404 		return (SET_ERROR(ENOMEM));
5405 
5406 	int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
5407 	if (error != 0)
5408 		return (error);
5409 
5410 	void *buf = kmem_alloc(bufsize, KM_SLEEP);
5411 
5412 	error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
5413 	    buf, &zc->zc_nvlist_dst_size);
5414 
5415 	if (error == 0) {
5416 		error = xcopyout(buf,
5417 		    (void *)(uintptr_t)zc->zc_nvlist_dst,
5418 		    zc->zc_nvlist_dst_size);
5419 	}
5420 	kmem_free(buf, bufsize);
5421 	zfsvfs_rele(zfsvfs, FTAG);
5422 
5423 	return (error);
5424 }
5425 
5426 /*
5427  * inputs:
5428  * zc_name		name of filesystem
5429  *
5430  * outputs:
5431  * none
5432  */
5433 static int
5434 zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
5435 {
5436 	objset_t *os;
5437 	int error = 0;
5438 	zfsvfs_t *zfsvfs;
5439 
5440 	if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
5441 		if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
5442 			/*
5443 			 * If userused is not enabled, it may be because the
5444 			 * objset needs to be closed & reopened (to grow the
5445 			 * objset_phys_t).  Suspend/resume the fs will do that.
5446 			 */
5447 			dsl_dataset_t *ds, *newds;
5448 
5449 			ds = dmu_objset_ds(zfsvfs->z_os);
5450 			error = zfs_suspend_fs(zfsvfs);
5451 			if (error == 0) {
5452 				dmu_objset_refresh_ownership(ds, &newds,
5453 				    B_TRUE, zfsvfs);
5454 				error = zfs_resume_fs(zfsvfs, newds);
5455 			}
5456 		}
5457 		if (error == 0)
5458 			error = dmu_objset_userspace_upgrade(zfsvfs->z_os);
5459 		VFS_RELE(zfsvfs->z_vfs);
5460 	} else {
5461 		/* XXX kind of reading contents without owning */
5462 		error = dmu_objset_hold_flags(zc->zc_name, B_TRUE, FTAG, &os);
5463 		if (error != 0)
5464 			return (error);
5465 
5466 		error = dmu_objset_userspace_upgrade(os);
5467 		dmu_objset_rele_flags(os, B_TRUE, FTAG);
5468 	}
5469 
5470 	return (error);
5471 }
5472 
5473 /*
5474  * inputs:
5475  * zc_name		name of filesystem
5476  *
5477  * outputs:
5478  * none
5479  */
5480 static int
5481 zfs_ioc_id_quota_upgrade(zfs_cmd_t *zc)
5482 {
5483 	objset_t *os;
5484 	int error;
5485 
5486 	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
5487 	if (error != 0)
5488 		return (error);
5489 
5490 	dsl_dataset_long_hold(dmu_objset_ds(os), FTAG);
5491 	dsl_pool_rele(dmu_objset_pool(os), FTAG);
5492 
5493 	if (dmu_objset_userobjspace_upgradable(os) ||
5494 	    dmu_objset_projectquota_upgradable(os)) {
5495 		mutex_enter(&os->os_upgrade_lock);
5496 		if (os->os_upgrade_id == 0) {
5497 			/* clear potential error code and retry */
5498 			os->os_upgrade_status = 0;
5499 			mutex_exit(&os->os_upgrade_lock);
5500 
5501 			dmu_objset_id_quota_upgrade(os);
5502 		} else {
5503 			mutex_exit(&os->os_upgrade_lock);
5504 		}
5505 
5506 		taskq_wait_id(os->os_spa->spa_upgrade_taskq, os->os_upgrade_id);
5507 		error = os->os_upgrade_status;
5508 	}
5509 
5510 	dsl_dataset_long_rele(dmu_objset_ds(os), FTAG);
5511 	dsl_dataset_rele(dmu_objset_ds(os), FTAG);
5512 
5513 	return (error);
5514 }
5515 
5516 /*
5517  * We don't want to have a hard dependency
5518  * against some special symbols in sharefs
5519  * nfs, and smbsrv.  Determine them if needed when
5520  * the first file system is shared.
5521  * Neither sharefs, nfs or smbsrv are unloadable modules.
5522  */
5523 int (*znfsexport_fs)(void *arg);
5524 int (*zshare_fs)(enum sharefs_sys_op, share_t *, uint32_t);
5525 int (*zsmbexport_fs)(void *arg, boolean_t add_share);
5526 
5527 int zfs_nfsshare_inited;
5528 int zfs_smbshare_inited;
5529 
5530 ddi_modhandle_t nfs_mod;
5531 ddi_modhandle_t sharefs_mod;
5532 ddi_modhandle_t smbsrv_mod;
5533 kmutex_t zfs_share_lock;
5534 
5535 static int
5536 zfs_init_sharefs()
5537 {
5538 	int error;
5539 
5540 	ASSERT(MUTEX_HELD(&zfs_share_lock));
5541 	/* Both NFS and SMB shares also require sharetab support. */
5542 	if (sharefs_mod == NULL && ((sharefs_mod =
5543 	    ddi_modopen("fs/sharefs",
5544 	    KRTLD_MODE_FIRST, &error)) == NULL)) {
5545 		return (SET_ERROR(ENOSYS));
5546 	}
5547 	if (zshare_fs == NULL && ((zshare_fs =
5548 	    (int (*)(enum sharefs_sys_op, share_t *, uint32_t))
5549 	    ddi_modsym(sharefs_mod, "sharefs_impl", &error)) == NULL)) {
5550 		return (SET_ERROR(ENOSYS));
5551 	}
5552 	return (0);
5553 }
5554 
5555 static int
5556 zfs_ioc_share(zfs_cmd_t *zc)
5557 {
5558 	int error;
5559 	int opcode;
5560 
5561 	switch (zc->zc_share.z_sharetype) {
5562 	case ZFS_SHARE_NFS:
5563 	case ZFS_UNSHARE_NFS:
5564 		if (zfs_nfsshare_inited == 0) {
5565 			mutex_enter(&zfs_share_lock);
5566 			if (nfs_mod == NULL && ((nfs_mod = ddi_modopen("fs/nfs",
5567 			    KRTLD_MODE_FIRST, &error)) == NULL)) {
5568 				mutex_exit(&zfs_share_lock);
5569 				return (SET_ERROR(ENOSYS));
5570 			}
5571 			if (znfsexport_fs == NULL &&
5572 			    ((znfsexport_fs = (int (*)(void *))
5573 			    ddi_modsym(nfs_mod,
5574 			    "nfs_export", &error)) == NULL)) {
5575 				mutex_exit(&zfs_share_lock);
5576 				return (SET_ERROR(ENOSYS));
5577 			}
5578 			error = zfs_init_sharefs();
5579 			if (error != 0) {
5580 				mutex_exit(&zfs_share_lock);
5581 				return (SET_ERROR(ENOSYS));
5582 			}
5583 			zfs_nfsshare_inited = 1;
5584 			mutex_exit(&zfs_share_lock);
5585 		}
5586 		break;
5587 	case ZFS_SHARE_SMB:
5588 	case ZFS_UNSHARE_SMB:
5589 		if (zfs_smbshare_inited == 0) {
5590 			mutex_enter(&zfs_share_lock);
5591 			if (smbsrv_mod == NULL && ((smbsrv_mod =
5592 			    ddi_modopen("drv/smbsrv",
5593 			    KRTLD_MODE_FIRST, &error)) == NULL)) {
5594 				mutex_exit(&zfs_share_lock);
5595 				return (SET_ERROR(ENOSYS));
5596 			}
5597 			if (zsmbexport_fs == NULL && ((zsmbexport_fs =
5598 			    (int (*)(void *, boolean_t))ddi_modsym(smbsrv_mod,
5599 			    "smb_server_share", &error)) == NULL)) {
5600 				mutex_exit(&zfs_share_lock);
5601 				return (SET_ERROR(ENOSYS));
5602 			}
5603 			error = zfs_init_sharefs();
5604 			if (error != 0) {
5605 				mutex_exit(&zfs_share_lock);
5606 				return (SET_ERROR(ENOSYS));
5607 			}
5608 			zfs_smbshare_inited = 1;
5609 			mutex_exit(&zfs_share_lock);
5610 		}
5611 		break;
5612 	default:
5613 		return (SET_ERROR(EINVAL));
5614 	}
5615 
5616 	switch (zc->zc_share.z_sharetype) {
5617 	case ZFS_SHARE_NFS:
5618 	case ZFS_UNSHARE_NFS:
5619 		if (error =
5620 		    znfsexport_fs((void *)
5621 		    (uintptr_t)zc->zc_share.z_exportdata))
5622 			return (error);
5623 		break;
5624 	case ZFS_SHARE_SMB:
5625 	case ZFS_UNSHARE_SMB:
5626 		if (error = zsmbexport_fs((void *)
5627 		    (uintptr_t)zc->zc_share.z_exportdata,
5628 		    zc->zc_share.z_sharetype == ZFS_SHARE_SMB ?
5629 		    B_TRUE: B_FALSE)) {
5630 			return (error);
5631 		}
5632 		break;
5633 	}
5634 
5635 	opcode = (zc->zc_share.z_sharetype == ZFS_SHARE_NFS ||
5636 	    zc->zc_share.z_sharetype == ZFS_SHARE_SMB) ?
5637 	    SHAREFS_ADD : SHAREFS_REMOVE;
5638 
5639 	/*
5640 	 * Add or remove share from sharetab
5641 	 */
5642 	error = zshare_fs(opcode,
5643 	    (void *)(uintptr_t)zc->zc_share.z_sharedata,
5644 	    zc->zc_share.z_sharemax);
5645 
5646 	return (error);
5647 
5648 }
5649 
5650 ace_t full_access[] = {
5651 	{(uid_t)-1, ACE_ALL_PERMS, ACE_EVERYONE, 0}
5652 };
5653 
5654 /*
5655  * inputs:
5656  * zc_name		name of containing filesystem
5657  * zc_obj		object # beyond which we want next in-use object #
5658  *
5659  * outputs:
5660  * zc_obj		next in-use object #
5661  */
5662 static int
5663 zfs_ioc_next_obj(zfs_cmd_t *zc)
5664 {
5665 	objset_t *os = NULL;
5666 	int error;
5667 
5668 	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
5669 	if (error != 0)
5670 		return (error);
5671 
5672 	error = dmu_object_next(os, &zc->zc_obj, B_FALSE,
5673 	    dsl_dataset_phys(os->os_dsl_dataset)->ds_prev_snap_txg);
5674 
5675 	dmu_objset_rele(os, FTAG);
5676 	return (error);
5677 }
5678 
5679 /*
5680  * inputs:
5681  * zc_name		name of filesystem
5682  * zc_value		prefix name for snapshot
5683  * zc_cleanup_fd	cleanup-on-exit file descriptor for calling process
5684  *
5685  * outputs:
5686  * zc_value		short name of new snapshot
5687  */
5688 static int
5689 zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
5690 {
5691 	char *snap_name;
5692 	char *hold_name;
5693 	int error;
5694 	minor_t minor;
5695 
5696 	error = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
5697 	if (error != 0)
5698 		return (error);
5699 
5700 	snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
5701 	    (u_longlong_t)ddi_get_lbolt64());
5702 	hold_name = kmem_asprintf("%%%s", zc->zc_value);
5703 
5704 	error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
5705 	    hold_name);
5706 	if (error == 0)
5707 		(void) strcpy(zc->zc_value, snap_name);
5708 	strfree(snap_name);
5709 	strfree(hold_name);
5710 	zfs_onexit_fd_rele(zc->zc_cleanup_fd);
5711 	return (error);
5712 }
5713 
5714 /*
5715  * inputs:
5716  * zc_name		name of "to" snapshot
5717  * zc_value		name of "from" snapshot
5718  * zc_cookie		file descriptor to write diff data on
5719  *
5720  * outputs:
5721  * dmu_diff_record_t's to the file descriptor
5722  */
5723 static int
5724 zfs_ioc_diff(zfs_cmd_t *zc)
5725 {
5726 	file_t *fp;
5727 	offset_t off;
5728 	int error;
5729 
5730 	fp = getf(zc->zc_cookie);
5731 	if (fp == NULL)
5732 		return (SET_ERROR(EBADF));
5733 
5734 	off = fp->f_offset;
5735 
5736 	error = dmu_diff(zc->zc_name, zc->zc_value, fp->f_vnode, &off);
5737 
5738 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
5739 		fp->f_offset = off;
5740 	releasef(zc->zc_cookie);
5741 
5742 	return (error);
5743 }
5744 
5745 /*
5746  * Remove all ACL files in shares dir
5747  */
5748 static int
5749 zfs_smb_acl_purge(znode_t *dzp)
5750 {
5751 	zap_cursor_t	zc;
5752 	zap_attribute_t	zap;
5753 	zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
5754 	int error;
5755 
5756 	for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
5757 	    (error = zap_cursor_retrieve(&zc, &zap)) == 0;
5758 	    zap_cursor_advance(&zc)) {
5759 		if ((error = VOP_REMOVE(ZTOV(dzp), zap.za_name, kcred,
5760 		    NULL, 0)) != 0)
5761 			break;
5762 	}
5763 	zap_cursor_fini(&zc);
5764 	return (error);
5765 }
5766 
5767 static int
5768 zfs_ioc_smb_acl(zfs_cmd_t *zc)
5769 {
5770 	vnode_t *vp;
5771 	znode_t *dzp;
5772 	vnode_t *resourcevp = NULL;
5773 	znode_t *sharedir;
5774 	zfsvfs_t *zfsvfs;
5775 	nvlist_t *nvlist;
5776 	char *src, *target;
5777 	vattr_t vattr;
5778 	vsecattr_t vsec;
5779 	int error = 0;
5780 
5781 	if ((error = lookupname(zc->zc_value, UIO_SYSSPACE,
5782 	    NO_FOLLOW, NULL, &vp)) != 0)
5783 		return (error);
5784 
5785 	/* Now make sure mntpnt and dataset are ZFS */
5786 
5787 	if (vp->v_vfsp->vfs_fstype != zfsfstype ||
5788 	    (strcmp((char *)refstr_value(vp->v_vfsp->vfs_resource),
5789 	    zc->zc_name) != 0)) {
5790 		VN_RELE(vp);
5791 		return (SET_ERROR(EINVAL));
5792 	}
5793 
5794 	dzp = VTOZ(vp);
5795 	zfsvfs = dzp->z_zfsvfs;
5796 	ZFS_ENTER(zfsvfs);
5797 
5798 	/*
5799 	 * Create share dir if its missing.
5800 	 */
5801 	mutex_enter(&zfsvfs->z_lock);
5802 	if (zfsvfs->z_shares_dir == 0) {
5803 		dmu_tx_t *tx;
5804 
5805 		tx = dmu_tx_create(zfsvfs->z_os);
5806 		dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, TRUE,
5807 		    ZFS_SHARES_DIR);
5808 		dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
5809 		error = dmu_tx_assign(tx, TXG_WAIT);
5810 		if (error != 0) {
5811 			dmu_tx_abort(tx);
5812 		} else {
5813 			error = zfs_create_share_dir(zfsvfs, tx);
5814 			dmu_tx_commit(tx);
5815 		}
5816 		if (error != 0) {
5817 			mutex_exit(&zfsvfs->z_lock);
5818 			VN_RELE(vp);
5819 			ZFS_EXIT(zfsvfs);
5820 			return (error);
5821 		}
5822 	}
5823 	mutex_exit(&zfsvfs->z_lock);
5824 
5825 	ASSERT(zfsvfs->z_shares_dir);
5826 	if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &sharedir)) != 0) {
5827 		VN_RELE(vp);
5828 		ZFS_EXIT(zfsvfs);
5829 		return (error);
5830 	}
5831 
5832 	switch (zc->zc_cookie) {
5833 	case ZFS_SMB_ACL_ADD:
5834 		vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
5835 		vattr.va_type = VREG;
5836 		vattr.va_mode = S_IFREG|0777;
5837 		vattr.va_uid = 0;
5838 		vattr.va_gid = 0;
5839 
5840 		vsec.vsa_mask = VSA_ACE;
5841 		vsec.vsa_aclentp = &full_access;
5842 		vsec.vsa_aclentsz = sizeof (full_access);
5843 		vsec.vsa_aclcnt = 1;
5844 
5845 		error = VOP_CREATE(ZTOV(sharedir), zc->zc_string,
5846 		    &vattr, EXCL, 0, &resourcevp, kcred, 0, NULL, &vsec);
5847 		if (resourcevp)
5848 			VN_RELE(resourcevp);
5849 		break;
5850 
5851 	case ZFS_SMB_ACL_REMOVE:
5852 		error = VOP_REMOVE(ZTOV(sharedir), zc->zc_string, kcred,
5853 		    NULL, 0);
5854 		break;
5855 
5856 	case ZFS_SMB_ACL_RENAME:
5857 		if ((error = get_nvlist(zc->zc_nvlist_src,
5858 		    zc->zc_nvlist_src_size, zc->zc_iflags, &nvlist)) != 0) {
5859 			VN_RELE(vp);
5860 			VN_RELE(ZTOV(sharedir));
5861 			ZFS_EXIT(zfsvfs);
5862 			return (error);
5863 		}
5864 		if (nvlist_lookup_string(nvlist, ZFS_SMB_ACL_SRC, &src) ||
5865 		    nvlist_lookup_string(nvlist, ZFS_SMB_ACL_TARGET,
5866 		    &target)) {
5867 			VN_RELE(vp);
5868 			VN_RELE(ZTOV(sharedir));
5869 			ZFS_EXIT(zfsvfs);
5870 			nvlist_free(nvlist);
5871 			return (error);
5872 		}
5873 		error = VOP_RENAME(ZTOV(sharedir), src, ZTOV(sharedir), target,
5874 		    kcred, NULL, 0);
5875 		nvlist_free(nvlist);
5876 		break;
5877 
5878 	case ZFS_SMB_ACL_PURGE:
5879 		error = zfs_smb_acl_purge(sharedir);
5880 		break;
5881 
5882 	default:
5883 		error = SET_ERROR(EINVAL);
5884 		break;
5885 	}
5886 
5887 	VN_RELE(vp);
5888 	VN_RELE(ZTOV(sharedir));
5889 
5890 	ZFS_EXIT(zfsvfs);
5891 
5892 	return (error);
5893 }
5894 
5895 /*
5896  * innvl: {
5897  *     "holds" -> { snapname -> holdname (string), ... }
5898  *     (optional) "cleanup_fd" -> fd (int32)
5899  * }
5900  *
5901  * outnvl: {
5902  *     snapname -> error value (int32)
5903  *     ...
5904  * }
5905  */
5906 /* ARGSUSED */
5907 static int
5908 zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
5909 {
5910 	nvpair_t *pair;
5911 	nvlist_t *holds;
5912 	int cleanup_fd = -1;
5913 	int error;
5914 	minor_t minor = 0;
5915 
5916 	error = nvlist_lookup_nvlist(args, "holds", &holds);
5917 	if (error != 0)
5918 		return (SET_ERROR(EINVAL));
5919 
5920 	/* make sure the user didn't pass us any invalid (empty) tags */
5921 	for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
5922 	    pair = nvlist_next_nvpair(holds, pair)) {
5923 		char *htag;
5924 
5925 		error = nvpair_value_string(pair, &htag);
5926 		if (error != 0)
5927 			return (SET_ERROR(error));
5928 
5929 		if (strlen(htag) == 0)
5930 			return (SET_ERROR(EINVAL));
5931 	}
5932 
5933 	if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
5934 		error = zfs_onexit_fd_hold(cleanup_fd, &minor);
5935 		if (error != 0)
5936 			return (error);
5937 	}
5938 
5939 	error = dsl_dataset_user_hold(holds, minor, errlist);
5940 	if (minor != 0)
5941 		zfs_onexit_fd_rele(cleanup_fd);
5942 	return (error);
5943 }
5944 
5945 /*
5946  * innvl is not used.
5947  *
5948  * outnvl: {
5949  *    holdname -> time added (uint64 seconds since epoch)
5950  *    ...
5951  * }
5952  */
5953 /* ARGSUSED */
5954 static int
5955 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
5956 {
5957 	ASSERT3P(args, ==, NULL);
5958 	return (dsl_dataset_get_holds(snapname, outnvl));
5959 }
5960 
5961 /*
5962  * innvl: {
5963  *     snapname -> { holdname, ... }
5964  *     ...
5965  * }
5966  *
5967  * outnvl: {
5968  *     snapname -> error value (int32)
5969  *     ...
5970  * }
5971  */
5972 /* ARGSUSED */
5973 static int
5974 zfs_ioc_release(const char *pool, nvlist_t *holds, nvlist_t *errlist)
5975 {
5976 	return (dsl_dataset_user_release(holds, errlist));
5977 }
5978 
5979 /*
5980  * inputs:
5981  * zc_name		name of new filesystem or snapshot
5982  * zc_value		full name of old snapshot
5983  *
5984  * outputs:
5985  * zc_cookie		space in bytes
5986  * zc_objset_type	compressed space in bytes
5987  * zc_perm_action	uncompressed space in bytes
5988  */
5989 static int
5990 zfs_ioc_space_written(zfs_cmd_t *zc)
5991 {
5992 	int error;
5993 	dsl_pool_t *dp;
5994 	dsl_dataset_t *new, *old;
5995 
5996 	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
5997 	if (error != 0)
5998 		return (error);
5999 	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
6000 	if (error != 0) {
6001 		dsl_pool_rele(dp, FTAG);
6002 		return (error);
6003 	}
6004 	error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);
6005 	if (error != 0) {
6006 		dsl_dataset_rele(new, FTAG);
6007 		dsl_pool_rele(dp, FTAG);
6008 		return (error);
6009 	}
6010 
6011 	error = dsl_dataset_space_written(old, new, &zc->zc_cookie,
6012 	    &zc->zc_objset_type, &zc->zc_perm_action);
6013 	dsl_dataset_rele(old, FTAG);
6014 	dsl_dataset_rele(new, FTAG);
6015 	dsl_pool_rele(dp, FTAG);
6016 	return (error);
6017 }
6018 
6019 /*
6020  * innvl: {
6021  *     "firstsnap" -> snapshot name
6022  * }
6023  *
6024  * outnvl: {
6025  *     "used" -> space in bytes
6026  *     "compressed" -> compressed space in bytes
6027  *     "uncompressed" -> uncompressed space in bytes
6028  * }
6029  */
6030 static int
6031 zfs_ioc_space_snaps(const char *lastsnap, nvlist_t *innvl, nvlist_t *outnvl)
6032 {
6033 	int error;
6034 	dsl_pool_t *dp;
6035 	dsl_dataset_t *new, *old;
6036 	char *firstsnap;
6037 	uint64_t used, comp, uncomp;
6038 
6039 	if (nvlist_lookup_string(innvl, "firstsnap", &firstsnap) != 0)
6040 		return (SET_ERROR(EINVAL));
6041 
6042 	error = dsl_pool_hold(lastsnap, FTAG, &dp);
6043 	if (error != 0)
6044 		return (error);
6045 
6046 	error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
6047 	if (error == 0 && !new->ds_is_snapshot) {
6048 		dsl_dataset_rele(new, FTAG);
6049 		error = SET_ERROR(EINVAL);
6050 	}
6051 	if (error != 0) {
6052 		dsl_pool_rele(dp, FTAG);
6053 		return (error);
6054 	}
6055 	error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
6056 	if (error == 0 && !old->ds_is_snapshot) {
6057 		dsl_dataset_rele(old, FTAG);
6058 		error = SET_ERROR(EINVAL);
6059 	}
6060 	if (error != 0) {
6061 		dsl_dataset_rele(new, FTAG);
6062 		dsl_pool_rele(dp, FTAG);
6063 		return (error);
6064 	}
6065 
6066 	error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
6067 	dsl_dataset_rele(old, FTAG);
6068 	dsl_dataset_rele(new, FTAG);
6069 	dsl_pool_rele(dp, FTAG);
6070 	fnvlist_add_uint64(outnvl, "used", used);
6071 	fnvlist_add_uint64(outnvl, "compressed", comp);
6072 	fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
6073 	return (error);
6074 }
6075 
6076 /*
6077  * innvl: {
6078  *     "fd" -> file descriptor to write stream to (int32)
6079  *     (optional) "fromsnap" -> full snap name to send an incremental from
6080  *     (optional) "largeblockok" -> (value ignored)
6081  *         indicates that blocks > 128KB are permitted
6082  *     (optional) "embedok" -> (value ignored)
6083  *         presence indicates DRR_WRITE_EMBEDDED records are permitted
6084  *     (optional) "compressok" -> (value ignored)
6085  *         presence indicates compressed DRR_WRITE records are permitted
6086  *     (optional) "rawok" -> (value ignored)
6087  *         presence indicates raw encrypted records should be used.
6088  *     (optional) "resume_object" and "resume_offset" -> (uint64)
6089  *         if present, resume send stream from specified object and offset.
6090  * }
6091  *
6092  * outnvl is unused
6093  */
6094 /* ARGSUSED */
6095 static int
6096 zfs_ioc_send_new(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
6097 {
6098 	int error;
6099 	offset_t off;
6100 	char *fromname = NULL;
6101 	int fd;
6102 	boolean_t largeblockok;
6103 	boolean_t embedok;
6104 	boolean_t compressok;
6105 	boolean_t rawok;
6106 	uint64_t resumeobj = 0;
6107 	uint64_t resumeoff = 0;
6108 
6109 	error = nvlist_lookup_int32(innvl, "fd", &fd);
6110 	if (error != 0)
6111 		return (SET_ERROR(EINVAL));
6112 
6113 	(void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
6114 
6115 	largeblockok = nvlist_exists(innvl, "largeblockok");
6116 	embedok = nvlist_exists(innvl, "embedok");
6117 	compressok = nvlist_exists(innvl, "compressok");
6118 	rawok = nvlist_exists(innvl, "rawok");
6119 
6120 	(void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
6121 	(void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);
6122 
6123 	file_t *fp = getf(fd);
6124 	if (fp == NULL)
6125 		return (SET_ERROR(EBADF));
6126 
6127 	off = fp->f_offset;
6128 	error = dmu_send(snapname, fromname, embedok, largeblockok, compressok,
6129 	    rawok, fd, resumeobj, resumeoff, fp->f_vnode, &off);
6130 
6131 	if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
6132 		fp->f_offset = off;
6133 	releasef(fd);
6134 	return (error);
6135 }
6136 
6137 /*
6138  * Determine approximately how large a zfs send stream will be -- the number
6139  * of bytes that will be written to the fd supplied to zfs_ioc_send_new().
6140  *
6141  * innvl: {
6142  *     (optional) "from" -> full snap or bookmark name to send an incremental
6143  *                          from
6144  *     (optional) "largeblockok" -> (value ignored)
6145  *         indicates that blocks > 128KB are permitted
6146  *     (optional) "embedok" -> (value ignored)
6147  *         presence indicates DRR_WRITE_EMBEDDED records are permitted
6148  *     (optional) "compressok" -> (value ignored)
6149  *         presence indicates compressed DRR_WRITE records are permitted
6150  * }
6151  *
6152  * outnvl: {
6153  *     "space" -> bytes of space (uint64)
6154  * }
6155  */
6156 static int
6157 zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
6158 {
6159 	dsl_pool_t *dp;
6160 	dsl_dataset_t *tosnap;
6161 	int error;
6162 	char *fromname;
6163 	boolean_t compressok;
6164 	boolean_t rawok;
6165 	uint64_t space;
6166 
6167 	error = dsl_pool_hold(snapname, FTAG, &dp);
6168 	if (error != 0)
6169 		return (error);
6170 
6171 	error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
6172 	if (error != 0) {
6173 		dsl_pool_rele(dp, FTAG);
6174 		return (error);
6175 	}
6176 
6177 	compressok = nvlist_exists(innvl, "compressok");
6178 	rawok = nvlist_exists(innvl, "rawok");
6179 
6180 	error = nvlist_lookup_string(innvl, "from", &fromname);
6181 	if (error == 0) {
6182 		if (strchr(fromname, '@') != NULL) {
6183 			/*
6184 			 * If from is a snapshot, hold it and use the more
6185 			 * efficient dmu_send_estimate to estimate send space
6186 			 * size using deadlists.
6187 			 */
6188 			dsl_dataset_t *fromsnap;
6189 			error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
6190 			if (error != 0)
6191 				goto out;
6192 			error = dmu_send_estimate(tosnap, fromsnap,
6193 			    compressok || rawok, &space);
6194 			dsl_dataset_rele(fromsnap, FTAG);
6195 		} else if (strchr(fromname, '#') != NULL) {
6196 			/*
6197 			 * If from is a bookmark, fetch the creation TXG of the
6198 			 * snapshot it was created from and use that to find
6199 			 * blocks that were born after it.
6200 			 */
6201 			zfs_bookmark_phys_t frombm;
6202 
6203 			error = dsl_bookmark_lookup(dp, fromname, tosnap,
6204 			    &frombm);
6205 			if (error != 0)
6206 				goto out;
6207 			error = dmu_send_estimate_from_txg(tosnap,
6208 			    frombm.zbm_creation_txg, compressok || rawok,
6209 			    &space);
6210 		} else {
6211 			/*
6212 			 * from is not properly formatted as a snapshot or
6213 			 * bookmark
6214 			 */
6215 			error = SET_ERROR(EINVAL);
6216 			goto out;
6217 		}
6218 	} else {
6219 		/*
6220 		 * If estimating the size of a full send, use dmu_send_estimate.
6221 		 */
6222 		error = dmu_send_estimate(tosnap, NULL, compressok || rawok,
6223 		    &space);
6224 	}
6225 
6226 	fnvlist_add_uint64(outnvl, "space", space);
6227 
6228 out:
6229 	dsl_dataset_rele(tosnap, FTAG);
6230 	dsl_pool_rele(dp, FTAG);
6231 	return (error);
6232 }
6233 
6234 /*
6235  * Sync the currently open TXG to disk for the specified pool.
6236  * This is somewhat similar to 'zfs_sync()'.
6237  * For cases that do not result in error this ioctl will wait for
6238  * the currently open TXG to commit before returning back to the caller.
6239  *
6240  * innvl: {
6241  *  "force" -> when true, force uberblock update even if there is no dirty data.
6242  *             In addition this will cause the vdev configuration to be written
6243  *             out including updating the zpool cache file. (boolean_t)
6244  * }
6245  *
6246  * onvl is unused
6247  */
6248 /* ARGSUSED */
6249 static int
6250 zfs_ioc_pool_sync(const char *pool, nvlist_t *innvl, nvlist_t *onvl)
6251 {
6252 	int err;
6253 	boolean_t force;
6254 	spa_t *spa;
6255 
6256 	if ((err = spa_open(pool, &spa, FTAG)) != 0)
6257 		return (err);
6258 
6259 	force = fnvlist_lookup_boolean_value(innvl, "force");
6260 	if (force) {
6261 		spa_config_enter(spa, SCL_CONFIG, FTAG, RW_WRITER);
6262 		vdev_config_dirty(spa->spa_root_vdev);
6263 		spa_config_exit(spa, SCL_CONFIG, FTAG);
6264 	}
6265 	txg_wait_synced(spa_get_dsl(spa), 0);
6266 
6267 	spa_close(spa, FTAG);
6268 
6269 	return (err);
6270 }
6271 
6272 /*
6273  * Load a user's wrapping key into the kernel.
6274  * innvl: {
6275  *     "hidden_args" -> { "wkeydata" -> value }
6276  *         raw uint8_t array of encryption wrapping key data (32 bytes)
6277  *     (optional) "noop" -> (value ignored)
6278  *         presence indicated key should only be verified, not loaded
6279  * }
6280  */
6281 /* ARGSUSED */
6282 static int
6283 zfs_ioc_load_key(const char *dsname, nvlist_t *innvl, nvlist_t *outnvl)
6284 {
6285 	int ret = 0;
6286 	dsl_crypto_params_t *dcp = NULL;
6287 	nvlist_t *hidden_args;
6288 	boolean_t noop = nvlist_exists(innvl, "noop");
6289 
6290 	if (strchr(dsname, '@') != NULL || strchr(dsname, '%') != NULL) {
6291 		ret = SET_ERROR(EINVAL);
6292 		goto error;
6293 	}
6294 
6295 	ret = nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);
6296 	if (ret != 0) {
6297 		ret = SET_ERROR(EINVAL);
6298 		goto error;
6299 	}
6300 
6301 	ret = dsl_crypto_params_create_nvlist(DCP_CMD_NONE, NULL,
6302 	    hidden_args, &dcp);
6303 	if (ret != 0)
6304 		goto error;
6305 
6306 	ret = spa_keystore_load_wkey(dsname, dcp, noop);
6307 	if (ret != 0)
6308 		goto error;
6309 
6310 	dsl_crypto_params_free(dcp, noop);
6311 
6312 	return (0);
6313 
6314 error:
6315 	dsl_crypto_params_free(dcp, B_TRUE);
6316 	return (ret);
6317 }
6318 
6319 /*
6320  * Unload a user's wrapping key from the kernel.
6321  * Both innvl and outnvl are unused.
6322  */
6323 /* ARGSUSED */
6324 static int
6325 zfs_ioc_unload_key(const char *dsname, nvlist_t *innvl, nvlist_t *outnvl)
6326 {
6327 	int ret = 0;
6328 
6329 	if (strchr(dsname, '@') != NULL || strchr(dsname, '%') != NULL) {
6330 		ret = (SET_ERROR(EINVAL));
6331 		goto out;
6332 	}
6333 
6334 	ret = spa_keystore_unload_wkey(dsname);
6335 	if (ret != 0)
6336 		goto out;
6337 
6338 out:
6339 	return (ret);
6340 }
6341 
6342 /*
6343  * Changes a user's wrapping key used to decrypt a dataset. The keyformat,
6344  * keylocation, pbkdf2salt, and  pbkdf2iters properties can also be specified
6345  * here to change how the key is derived in userspace.
6346  *
6347  * innvl: {
6348  *    "hidden_args" (optional) -> { "wkeydata" -> value }
6349  *         raw uint8_t array of new encryption wrapping key data (32 bytes)
6350  *    "props" (optional) -> { prop -> value }
6351  * }
6352  *
6353  * outnvl is unused
6354  */
6355 /* ARGSUSED */
6356 static int
6357 zfs_ioc_change_key(const char *dsname, nvlist_t *innvl, nvlist_t *outnvl)
6358 {
6359 	int ret;
6360 	uint64_t cmd = DCP_CMD_NONE;
6361 	dsl_crypto_params_t *dcp = NULL;
6362 	nvlist_t *args = NULL, *hidden_args = NULL;
6363 
6364 	if (strchr(dsname, '@') != NULL || strchr(dsname, '%') != NULL) {
6365 		ret = (SET_ERROR(EINVAL));
6366 		goto error;
6367 	}
6368 
6369 	(void) nvlist_lookup_uint64(innvl, "crypt_cmd", &cmd);
6370 	(void) nvlist_lookup_nvlist(innvl, "props", &args);
6371 	(void) nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);
6372 
6373 	ret = dsl_crypto_params_create_nvlist(cmd, args, hidden_args, &dcp);
6374 	if (ret != 0)
6375 		goto error;
6376 
6377 	ret = spa_keystore_change_key(dsname, dcp);
6378 	if (ret != 0)
6379 		goto error;
6380 
6381 	dsl_crypto_params_free(dcp, B_FALSE);
6382 
6383 	return (0);
6384 
6385 error:
6386 	dsl_crypto_params_free(dcp, B_TRUE);
6387 	return (ret);
6388 }
6389 
6390 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
6391 
6392 static void
6393 zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6394     zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
6395     boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
6396 {
6397 	zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
6398 
6399 	ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
6400 	ASSERT3U(ioc, <, ZFS_IOC_LAST);
6401 	ASSERT3P(vec->zvec_legacy_func, ==, NULL);
6402 	ASSERT3P(vec->zvec_func, ==, NULL);
6403 
6404 	vec->zvec_legacy_func = func;
6405 	vec->zvec_secpolicy = secpolicy;
6406 	vec->zvec_namecheck = namecheck;
6407 	vec->zvec_allow_log = log_history;
6408 	vec->zvec_pool_check = pool_check;
6409 }
6410 
6411 /*
6412  * See the block comment at the beginning of this file for details on
6413  * each argument to this function.
6414  */
6415 static void
6416 zfs_ioctl_register(const char *name, zfs_ioc_t ioc, zfs_ioc_func_t *func,
6417     zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
6418     zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
6419     boolean_t allow_log)
6420 {
6421 	zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];
6422 
6423 	ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
6424 	ASSERT3U(ioc, <, ZFS_IOC_LAST);
6425 	ASSERT3P(vec->zvec_legacy_func, ==, NULL);
6426 	ASSERT3P(vec->zvec_func, ==, NULL);
6427 
6428 	/* if we are logging, the name must be valid */
6429 	ASSERT(!allow_log || namecheck != NO_NAME);
6430 
6431 	vec->zvec_name = name;
6432 	vec->zvec_func = func;
6433 	vec->zvec_secpolicy = secpolicy;
6434 	vec->zvec_namecheck = namecheck;
6435 	vec->zvec_pool_check = pool_check;
6436 	vec->zvec_smush_outnvlist = smush_outnvlist;
6437 	vec->zvec_allow_log = allow_log;
6438 }
6439 
6440 static void
6441 zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6442     zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
6443     zfs_ioc_poolcheck_t pool_check)
6444 {
6445 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
6446 	    POOL_NAME, log_history, pool_check);
6447 }
6448 
6449 static void
6450 zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6451     zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
6452 {
6453 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
6454 	    DATASET_NAME, B_FALSE, pool_check);
6455 }
6456 
6457 static void
6458 zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
6459 {
6460 	zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
6461 	    POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
6462 }
6463 
6464 static void
6465 zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6466     zfs_secpolicy_func_t *secpolicy)
6467 {
6468 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
6469 	    NO_NAME, B_FALSE, POOL_CHECK_NONE);
6470 }
6471 
6472 static void
6473 zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
6474     zfs_ioc_legacy_func_t *func, zfs_secpolicy_func_t *secpolicy)
6475 {
6476 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
6477 	    DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
6478 }
6479 
6480 static void
6481 zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
6482 {
6483 	zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
6484 	    zfs_secpolicy_read);
6485 }
6486 
6487 static void
6488 zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
6489     zfs_secpolicy_func_t *secpolicy)
6490 {
6491 	zfs_ioctl_register_legacy(ioc, func, secpolicy,
6492 	    DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
6493 }
6494 
6495 static void
6496 zfs_ioctl_init(void)
6497 {
6498 	zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
6499 	    zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
6500 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6501 
6502 	zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
6503 	    zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
6504 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
6505 
6506 	zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
6507 	    zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
6508 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
6509 
6510 	zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
6511 	    zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
6512 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
6513 
6514 	zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
6515 	    zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
6516 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
6517 
6518 	zfs_ioctl_register("create", ZFS_IOC_CREATE,
6519 	    zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
6520 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6521 
6522 	zfs_ioctl_register("clone", ZFS_IOC_CLONE,
6523 	    zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
6524 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6525 
6526 	zfs_ioctl_register("remap", ZFS_IOC_REMAP,
6527 	    zfs_ioc_remap, zfs_secpolicy_remap, DATASET_NAME,
6528 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
6529 
6530 	zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
6531 	    zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
6532 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6533 
6534 	zfs_ioctl_register("hold", ZFS_IOC_HOLD,
6535 	    zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
6536 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6537 	zfs_ioctl_register("release", ZFS_IOC_RELEASE,
6538 	    zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
6539 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6540 
6541 	zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
6542 	    zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
6543 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
6544 
6545 	zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
6546 	    zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
6547 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_TRUE);
6548 
6549 	zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
6550 	    zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
6551 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6552 
6553 	zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
6554 	    zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
6555 	    POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE);
6556 
6557 	zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
6558 	    zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
6559 	    POOL_NAME,
6560 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6561 
6562 	zfs_ioctl_register("channel_program", ZFS_IOC_CHANNEL_PROGRAM,
6563 	    zfs_ioc_channel_program, zfs_secpolicy_config,
6564 	    POOL_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE,
6565 	    B_TRUE);
6566 
6567 	zfs_ioctl_register("zpool_checkpoint", ZFS_IOC_POOL_CHECKPOINT,
6568 	    zfs_ioc_pool_checkpoint, zfs_secpolicy_config, POOL_NAME,
6569 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6570 
6571 	zfs_ioctl_register("zpool_discard_checkpoint",
6572 	    ZFS_IOC_POOL_DISCARD_CHECKPOINT, zfs_ioc_pool_discard_checkpoint,
6573 	    zfs_secpolicy_config, POOL_NAME,
6574 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6575 
6576 	zfs_ioctl_register("initialize", ZFS_IOC_POOL_INITIALIZE,
6577 	    zfs_ioc_pool_initialize, zfs_secpolicy_config, POOL_NAME,
6578 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6579 
6580 	zfs_ioctl_register("trim", ZFS_IOC_POOL_TRIM,
6581 	    zfs_ioc_pool_trim, zfs_secpolicy_config, POOL_NAME,
6582 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE, B_TRUE);
6583 
6584 	zfs_ioctl_register("sync", ZFS_IOC_POOL_SYNC,
6585 	    zfs_ioc_pool_sync, zfs_secpolicy_none, POOL_NAME,
6586 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
6587 
6588 	zfs_ioctl_register("load-key", ZFS_IOC_LOAD_KEY,
6589 	    zfs_ioc_load_key, zfs_secpolicy_load_key,
6590 	    DATASET_NAME, POOL_CHECK_SUSPENDED, B_TRUE, B_TRUE);
6591 	zfs_ioctl_register("unload-key", ZFS_IOC_UNLOAD_KEY,
6592 	    zfs_ioc_unload_key, zfs_secpolicy_load_key,
6593 	    DATASET_NAME, POOL_CHECK_SUSPENDED, B_TRUE, B_TRUE);
6594 	zfs_ioctl_register("change-key", ZFS_IOC_CHANGE_KEY,
6595 	    zfs_ioc_change_key, zfs_secpolicy_change_key,
6596 	    DATASET_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY,
6597 	    B_TRUE, B_TRUE);
6598 
6599 	/* IOCTLS that use the legacy function signature */
6600 
6601 	zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
6602 	    zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);
6603 
6604 	zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
6605 	    zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
6606 	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
6607 	    zfs_ioc_pool_scan);
6608 	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
6609 	    zfs_ioc_pool_upgrade);
6610 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
6611 	    zfs_ioc_vdev_add);
6612 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
6613 	    zfs_ioc_vdev_remove);
6614 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
6615 	    zfs_ioc_vdev_set_state);
6616 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
6617 	    zfs_ioc_vdev_attach);
6618 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
6619 	    zfs_ioc_vdev_detach);
6620 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
6621 	    zfs_ioc_vdev_setpath);
6622 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
6623 	    zfs_ioc_vdev_setfru);
6624 	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
6625 	    zfs_ioc_pool_set_props);
6626 	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
6627 	    zfs_ioc_vdev_split);
6628 	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
6629 	    zfs_ioc_pool_reguid);
6630 
6631 	zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
6632 	    zfs_ioc_pool_configs, zfs_secpolicy_none);
6633 	zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
6634 	    zfs_ioc_pool_tryimport, zfs_secpolicy_config);
6635 	zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
6636 	    zfs_ioc_inject_fault, zfs_secpolicy_inject);
6637 	zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
6638 	    zfs_ioc_clear_fault, zfs_secpolicy_inject);
6639 	zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
6640 	    zfs_ioc_inject_list_next, zfs_secpolicy_inject);
6641 
6642 	/*
6643 	 * pool destroy, and export don't log the history as part of
6644 	 * zfsdev_ioctl, but rather zfs_ioc_pool_export
6645 	 * does the logging of those commands.
6646 	 */
6647 	zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
6648 	    zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
6649 	zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
6650 	    zfs_secpolicy_config, B_FALSE, POOL_CHECK_NONE);
6651 
6652 	zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
6653 	    zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
6654 	zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
6655 	    zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
6656 
6657 	zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
6658 	    zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
6659 	zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
6660 	    zfs_ioc_dsobj_to_dsname,
6661 	    zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
6662 	zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
6663 	    zfs_ioc_pool_get_history,
6664 	    zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
6665 
6666 	zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
6667 	    zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
6668 
6669 	zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
6670 	    zfs_secpolicy_config, B_TRUE, POOL_CHECK_READONLY);
6671 	zfs_ioctl_register_pool(ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
6672 	    zfs_secpolicy_config, B_TRUE, POOL_CHECK_SUSPENDED);
6673 
6674 	zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
6675 	    zfs_ioc_space_written);
6676 	zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
6677 	    zfs_ioc_objset_recvd_props);
6678 	zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
6679 	    zfs_ioc_next_obj);
6680 	zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
6681 	    zfs_ioc_get_fsacl);
6682 	zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
6683 	    zfs_ioc_objset_stats);
6684 	zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
6685 	    zfs_ioc_objset_zplprops);
6686 	zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
6687 	    zfs_ioc_dataset_list_next);
6688 	zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
6689 	    zfs_ioc_snapshot_list_next);
6690 	zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
6691 	    zfs_ioc_send_progress);
6692 
6693 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
6694 	    zfs_ioc_diff, zfs_secpolicy_diff);
6695 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
6696 	    zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
6697 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
6698 	    zfs_ioc_obj_to_path, zfs_secpolicy_diff);
6699 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
6700 	    zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
6701 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
6702 	    zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
6703 	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
6704 	    zfs_ioc_send, zfs_secpolicy_send);
6705 
6706 	zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
6707 	    zfs_secpolicy_none);
6708 	zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
6709 	    zfs_secpolicy_destroy);
6710 	zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
6711 	    zfs_secpolicy_rename);
6712 	zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
6713 	    zfs_secpolicy_recv);
6714 	zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
6715 	    zfs_secpolicy_promote);
6716 	zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
6717 	    zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
6718 	zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
6719 	    zfs_secpolicy_set_fsacl);
6720 
6721 	zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
6722 	    zfs_secpolicy_share, POOL_CHECK_NONE);
6723 	zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
6724 	    zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
6725 	zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
6726 	    zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
6727 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
6728 	zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
6729 	    zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
6730 	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY);
6731 }
6732 
6733 int
6734 pool_status_check(const char *name, zfs_ioc_namecheck_t type,
6735     zfs_ioc_poolcheck_t check)
6736 {
6737 	spa_t *spa;
6738 	int error;
6739 
6740 	ASSERT(type == POOL_NAME || type == DATASET_NAME);
6741 
6742 	if (check & POOL_CHECK_NONE)
6743 		return (0);
6744 
6745 	error = spa_open(name, &spa, FTAG);
6746 	if (error == 0) {
6747 		if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
6748 			error = SET_ERROR(EAGAIN);
6749 		else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
6750 			error = SET_ERROR(EROFS);
6751 		spa_close(spa, FTAG);
6752 	}
6753 	return (error);
6754 }
6755 
6756 /*
6757  * Find a free minor number.
6758  */
6759 minor_t
6760 zfsdev_minor_alloc(void)
6761 {
6762 	static minor_t last_minor;
6763 	minor_t m;
6764 
6765 	ASSERT(MUTEX_HELD(&zfsdev_state_lock));
6766 
6767 	for (m = last_minor + 1; m != last_minor; m++) {
6768 		if (m > ZFSDEV_MAX_MINOR)
6769 			m = 1;
6770 		if (ddi_get_soft_state(zfsdev_state, m) == NULL) {
6771 			last_minor = m;
6772 			return (m);
6773 		}
6774 	}
6775 
6776 	return (0);
6777 }
6778 
6779 static int
6780 zfs_ctldev_init(dev_t *devp)
6781 {
6782 	minor_t minor;
6783 	zfs_soft_state_t *zs;
6784 
6785 	ASSERT(MUTEX_HELD(&zfsdev_state_lock));
6786 	ASSERT(getminor(*devp) == 0);
6787 
6788 	minor = zfsdev_minor_alloc();
6789 	if (minor == 0)
6790 		return (SET_ERROR(ENXIO));
6791 
6792 	if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS)
6793 		return (SET_ERROR(EAGAIN));
6794 
6795 	*devp = makedevice(getemajor(*devp), minor);
6796 
6797 	zs = ddi_get_soft_state(zfsdev_state, minor);
6798 	zs->zss_type = ZSST_CTLDEV;
6799 	zfs_onexit_init((zfs_onexit_t **)&zs->zss_data);
6800 
6801 	return (0);
6802 }
6803 
6804 static void
6805 zfs_ctldev_destroy(zfs_onexit_t *zo, minor_t minor)
6806 {
6807 	ASSERT(MUTEX_HELD(&zfsdev_state_lock));
6808 
6809 	zfs_onexit_destroy(zo);
6810 	ddi_soft_state_free(zfsdev_state, minor);
6811 }
6812 
6813 void *
6814 zfsdev_get_soft_state(minor_t minor, enum zfs_soft_state_type which)
6815 {
6816 	zfs_soft_state_t *zp;
6817 
6818 	zp = ddi_get_soft_state(zfsdev_state, minor);
6819 	if (zp == NULL || zp->zss_type != which)
6820 		return (NULL);
6821 
6822 	return (zp->zss_data);
6823 }
6824 
6825 static int
6826 zfsdev_open(dev_t *devp, int flag, int otyp, cred_t *cr)
6827 {
6828 	int error = 0;
6829 
6830 	if (getminor(*devp) != 0)
6831 		return (zvol_open(devp, flag, otyp, cr));
6832 
6833 	/* This is the control device. Allocate a new minor if requested. */
6834 	if (flag & FEXCL) {
6835 		mutex_enter(&zfsdev_state_lock);
6836 		error = zfs_ctldev_init(devp);
6837 		mutex_exit(&zfsdev_state_lock);
6838 	}
6839 
6840 	return (error);
6841 }
6842 
6843 static int
6844 zfsdev_close(dev_t dev, int flag, int otyp, cred_t *cr)
6845 {
6846 	zfs_onexit_t *zo;
6847 	minor_t minor = getminor(dev);
6848 
6849 	if (minor == 0)
6850 		return (0);
6851 
6852 	mutex_enter(&zfsdev_state_lock);
6853 	zo = zfsdev_get_soft_state(minor, ZSST_CTLDEV);
6854 	if (zo == NULL) {
6855 		mutex_exit(&zfsdev_state_lock);
6856 		return (zvol_close(dev, flag, otyp, cr));
6857 	}
6858 	zfs_ctldev_destroy(zo, minor);
6859 	mutex_exit(&zfsdev_state_lock);
6860 
6861 	return (0);
6862 }
6863 
6864 static int
6865 zfsdev_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp)
6866 {
6867 	zfs_cmd_t *zc;
6868 	uint_t vecnum;
6869 	int error, rc, len;
6870 	minor_t minor = getminor(dev);
6871 	const zfs_ioc_vec_t *vec;
6872 	char *saved_poolname = NULL;
6873 	nvlist_t *innvl = NULL;
6874 
6875 	if (minor != 0 &&
6876 	    zfsdev_get_soft_state(minor, ZSST_CTLDEV) == NULL)
6877 		return (zvol_ioctl(dev, cmd, arg, flag, cr, rvalp));
6878 
6879 	vecnum = cmd - ZFS_IOC_FIRST;
6880 	ASSERT3U(getmajor(dev), ==, ddi_driver_major(zfs_dip));
6881 
6882 	if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
6883 		return (SET_ERROR(EINVAL));
6884 	vec = &zfs_ioc_vec[vecnum];
6885 
6886 	zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
6887 
6888 	error = ddi_copyin((void *)arg, zc, sizeof (zfs_cmd_t), flag);
6889 	if (error != 0) {
6890 		error = SET_ERROR(EFAULT);
6891 		goto out;
6892 	}
6893 
6894 	zc->zc_iflags = flag & FKIOCTL;
6895 	if (zc->zc_nvlist_src_size != 0) {
6896 		error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
6897 		    zc->zc_iflags, &innvl);
6898 		if (error != 0)
6899 			goto out;
6900 	}
6901 
6902 	/*
6903 	 * Ensure that all pool/dataset names are valid before we pass down to
6904 	 * the lower layers.
6905 	 */
6906 	zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
6907 	switch (vec->zvec_namecheck) {
6908 	case POOL_NAME:
6909 		if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
6910 			error = SET_ERROR(EINVAL);
6911 		else
6912 			error = pool_status_check(zc->zc_name,
6913 			    vec->zvec_namecheck, vec->zvec_pool_check);
6914 		break;
6915 
6916 	case DATASET_NAME:
6917 		if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
6918 			error = SET_ERROR(EINVAL);
6919 		else
6920 			error = pool_status_check(zc->zc_name,
6921 			    vec->zvec_namecheck, vec->zvec_pool_check);
6922 		break;
6923 
6924 	case NO_NAME:
6925 		break;
6926 	}
6927 
6928 
6929 	if (error == 0)
6930 		error = vec->zvec_secpolicy(zc, innvl, cr);
6931 
6932 	if (error != 0)
6933 		goto out;
6934 
6935 	/* legacy ioctls can modify zc_name */
6936 	len = strcspn(zc->zc_name, "/@#") + 1;
6937 	saved_poolname = kmem_alloc(len, KM_SLEEP);
6938 	(void) strlcpy(saved_poolname, zc->zc_name, len);
6939 
6940 	if (vec->zvec_func != NULL) {
6941 		nvlist_t *outnvl;
6942 		int puterror = 0;
6943 		spa_t *spa;
6944 		nvlist_t *lognv = NULL;
6945 
6946 		ASSERT(vec->zvec_legacy_func == NULL);
6947 
6948 		/*
6949 		 * Add the innvl to the lognv before calling the func,
6950 		 * in case the func changes the innvl.
6951 		 */
6952 		if (vec->zvec_allow_log) {
6953 			lognv = fnvlist_alloc();
6954 			fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
6955 			    vec->zvec_name);
6956 			if (!nvlist_empty(innvl)) {
6957 				fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
6958 				    innvl);
6959 			}
6960 		}
6961 
6962 		outnvl = fnvlist_alloc();
6963 		error = vec->zvec_func(zc->zc_name, innvl, outnvl);
6964 
6965 		/*
6966 		 * Some commands can partially execute, modify state, and still
6967 		 * return an error.  In these cases, attempt to record what
6968 		 * was modified.
6969 		 */
6970 		if ((error == 0 ||
6971 		    (cmd == ZFS_IOC_CHANNEL_PROGRAM && error != EINVAL)) &&
6972 		    vec->zvec_allow_log &&
6973 		    spa_open(zc->zc_name, &spa, FTAG) == 0) {
6974 			if (!nvlist_empty(outnvl)) {
6975 				fnvlist_add_nvlist(lognv, ZPOOL_HIST_OUTPUT_NVL,
6976 				    outnvl);
6977 			}
6978 			if (error != 0) {
6979 				fnvlist_add_int64(lognv, ZPOOL_HIST_ERRNO,
6980 				    error);
6981 			}
6982 			(void) spa_history_log_nvl(spa, lognv);
6983 			spa_close(spa, FTAG);
6984 		}
6985 		fnvlist_free(lognv);
6986 
6987 		if (!nvlist_empty(outnvl) || zc->zc_nvlist_dst_size != 0) {
6988 			int smusherror = 0;
6989 			if (vec->zvec_smush_outnvlist) {
6990 				smusherror = nvlist_smush(outnvl,
6991 				    zc->zc_nvlist_dst_size);
6992 			}
6993 			if (smusherror == 0)
6994 				puterror = put_nvlist(zc, outnvl);
6995 		}
6996 
6997 		if (puterror != 0)
6998 			error = puterror;
6999 
7000 		nvlist_free(outnvl);
7001 	} else {
7002 		error = vec->zvec_legacy_func(zc);
7003 	}
7004 
7005 out:
7006 	nvlist_free(innvl);
7007 	rc = ddi_copyout(zc, (void *)arg, sizeof (zfs_cmd_t), flag);
7008 	if (error == 0 && rc != 0)
7009 		error = SET_ERROR(EFAULT);
7010 	if (error == 0 && vec->zvec_allow_log) {
7011 		char *s = tsd_get(zfs_allow_log_key);
7012 		if (s != NULL)
7013 			strfree(s);
7014 		(void) tsd_set(zfs_allow_log_key, saved_poolname);
7015 	} else {
7016 		if (saved_poolname != NULL)
7017 			strfree(saved_poolname);
7018 	}
7019 
7020 	kmem_free(zc, sizeof (zfs_cmd_t));
7021 	return (error);
7022 }
7023 
7024 static int
7025 zfs_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
7026 {
7027 	if (cmd != DDI_ATTACH)
7028 		return (DDI_FAILURE);
7029 
7030 	if (ddi_create_minor_node(dip, "zfs", S_IFCHR, 0,
7031 	    DDI_PSEUDO, 0) == DDI_FAILURE)
7032 		return (DDI_FAILURE);
7033 
7034 	zfs_dip = dip;
7035 
7036 	ddi_report_dev(dip);
7037 
7038 	return (DDI_SUCCESS);
7039 }
7040 
7041 static int
7042 zfs_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
7043 {
7044 	if (spa_busy() || zfs_busy() || zvol_busy())
7045 		return (DDI_FAILURE);
7046 
7047 	if (cmd != DDI_DETACH)
7048 		return (DDI_FAILURE);
7049 
7050 	zfs_dip = NULL;
7051 
7052 	ddi_prop_remove_all(dip);
7053 	ddi_remove_minor_node(dip, NULL);
7054 
7055 	return (DDI_SUCCESS);
7056 }
7057 
7058 /*ARGSUSED*/
7059 static int
7060 zfs_info(dev_info_t *dip, ddi_info_cmd_t infocmd, void *arg, void **result)
7061 {
7062 	switch (infocmd) {
7063 	case DDI_INFO_DEVT2DEVINFO:
7064 		*result = zfs_dip;
7065 		return (DDI_SUCCESS);
7066 
7067 	case DDI_INFO_DEVT2INSTANCE:
7068 		*result = (void *)0;
7069 		return (DDI_SUCCESS);
7070 	}
7071 
7072 	return (DDI_FAILURE);
7073 }
7074 
7075 /*
7076  * OK, so this is a little weird.
7077  *
7078  * /dev/zfs is the control node, i.e. minor 0.
7079  * /dev/zvol/[r]dsk/pool/dataset are the zvols, minor > 0.
7080  *
7081  * /dev/zfs has basically nothing to do except serve up ioctls,
7082  * so most of the standard driver entry points are in zvol.c.
7083  */
7084 static struct cb_ops zfs_cb_ops = {
7085 	zfsdev_open,	/* open */
7086 	zfsdev_close,	/* close */
7087 	zvol_strategy,	/* strategy */
7088 	nodev,		/* print */
7089 	zvol_dump,	/* dump */
7090 	zvol_read,	/* read */
7091 	zvol_write,	/* write */
7092 	zfsdev_ioctl,	/* ioctl */
7093 	nodev,		/* devmap */
7094 	nodev,		/* mmap */
7095 	nodev,		/* segmap */
7096 	nochpoll,	/* poll */
7097 	ddi_prop_op,	/* prop_op */
7098 	NULL,		/* streamtab */
7099 	D_NEW | D_MP | D_64BIT,		/* Driver compatibility flag */
7100 	CB_REV,		/* version */
7101 	nodev,		/* async read */
7102 	nodev,		/* async write */
7103 };
7104 
7105 static struct dev_ops zfs_dev_ops = {
7106 	DEVO_REV,	/* version */
7107 	0,		/* refcnt */
7108 	zfs_info,	/* info */
7109 	nulldev,	/* identify */
7110 	nulldev,	/* probe */
7111 	zfs_attach,	/* attach */
7112 	zfs_detach,	/* detach */
7113 	nodev,		/* reset */
7114 	&zfs_cb_ops,	/* driver operations */
7115 	NULL,		/* no bus operations */
7116 	NULL,		/* power */
7117 	ddi_quiesce_not_needed,	/* quiesce */
7118 };
7119 
7120 static struct modldrv zfs_modldrv = {
7121 	&mod_driverops,
7122 	"ZFS storage pool",
7123 	&zfs_dev_ops
7124 };
7125 
7126 static struct modlinkage modlinkage = {
7127 	MODREV_1,
7128 	(void *)&zfs_modlfs,
7129 	(void *)&zfs_modldrv,
7130 	NULL
7131 };
7132 
7133 static void
7134 zfs_allow_log_destroy(void *arg)
7135 {
7136 	char *poolname = arg;
7137 	strfree(poolname);
7138 }
7139 
7140 int
7141 _init(void)
7142 {
7143 	int error;
7144 
7145 	spa_init(FREAD | FWRITE);
7146 	zfs_init();
7147 	zvol_init();
7148 	zfs_ioctl_init();
7149 
7150 	if ((error = mod_install(&modlinkage)) != 0) {
7151 		zvol_fini();
7152 		zfs_fini();
7153 		spa_fini();
7154 		return (error);
7155 	}
7156 
7157 	tsd_create(&zfs_fsyncer_key, NULL);
7158 	tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
7159 	tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);
7160 
7161 	error = ldi_ident_from_mod(&modlinkage, &zfs_li);
7162 	ASSERT(error == 0);
7163 	mutex_init(&zfs_share_lock, NULL, MUTEX_DEFAULT, NULL);
7164 
7165 	return (0);
7166 }
7167 
7168 int
7169 _fini(void)
7170 {
7171 	int error;
7172 
7173 	if (spa_busy() || zfs_busy() || zvol_busy() || zio_injection_enabled)
7174 		return (SET_ERROR(EBUSY));
7175 
7176 	if ((error = mod_remove(&modlinkage)) != 0)
7177 		return (error);
7178 
7179 	zvol_fini();
7180 	zfs_fini();
7181 	spa_fini();
7182 	if (zfs_nfsshare_inited)
7183 		(void) ddi_modclose(nfs_mod);
7184 	if (zfs_smbshare_inited)
7185 		(void) ddi_modclose(smbsrv_mod);
7186 	if (zfs_nfsshare_inited || zfs_smbshare_inited)
7187 		(void) ddi_modclose(sharefs_mod);
7188 
7189 	tsd_destroy(&zfs_fsyncer_key);
7190 	ldi_ident_release(zfs_li);
7191 	zfs_li = NULL;
7192 	mutex_destroy(&zfs_share_lock);
7193 
7194 	return (error);
7195 }
7196 
7197 int
7198 _info(struct modinfo *modinfop)
7199 {
7200 	return (mod_info(&modlinkage, modinfop));
7201 }
7202